xref: /freebsd/sys/contrib/openzfs/lib/libzfs/libzfs_dataset.c (revision e6bfd18d21b225af6a0ed67ceeaf1293b7b9eba5)
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 https://opensource.org/licenses/CDDL-1.0.
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  * Copyright 2019 Joyent, Inc.
25  * Copyright (c) 2011, 2020 by Delphix. All rights reserved.
26  * Copyright (c) 2012 DEY Storage Systems, Inc.  All rights reserved.
27  * Copyright (c) 2012 Pawel Jakub Dawidek <pawel@dawidek.net>.
28  * Copyright (c) 2013 Martin Matuska. All rights reserved.
29  * Copyright (c) 2013 Steven Hartland. All rights reserved.
30  * Copyright 2017 Nexenta Systems, Inc.
31  * Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>
32  * Copyright 2017-2018 RackTop Systems.
33  * Copyright (c) 2019 Datto Inc.
34  * Copyright (c) 2019, loli10K <ezomori.nozomu@gmail.com>
35  * Copyright (c) 2021 Matt Fiddaman
36  */
37 
38 #include <ctype.h>
39 #include <errno.h>
40 #include <libintl.h>
41 #include <stdio.h>
42 #include <stdlib.h>
43 #include <strings.h>
44 #include <unistd.h>
45 #include <stddef.h>
46 #include <zone.h>
47 #include <fcntl.h>
48 #include <sys/mntent.h>
49 #include <sys/mount.h>
50 #include <pwd.h>
51 #include <grp.h>
52 #ifdef HAVE_IDMAP
53 #include <idmap.h>
54 #include <aclutils.h>
55 #include <directory.h>
56 #endif /* HAVE_IDMAP */
57 
58 #include <sys/dnode.h>
59 #include <sys/spa.h>
60 #include <sys/zap.h>
61 #include <sys/dsl_crypt.h>
62 #include <libzfs.h>
63 #include <libzutil.h>
64 
65 #include "zfs_namecheck.h"
66 #include "zfs_prop.h"
67 #include "libzfs_impl.h"
68 #include "zfs_deleg.h"
69 
70 static int userquota_propname_decode(const char *propname, boolean_t zoned,
71     zfs_userquota_prop_t *typep, char *domain, int domainlen, uint64_t *ridp);
72 
73 /*
74  * Given a single type (not a mask of types), return the type in a human
75  * readable form.
76  */
77 const char *
78 zfs_type_to_name(zfs_type_t type)
79 {
80 	switch (type) {
81 	case ZFS_TYPE_FILESYSTEM:
82 		return (dgettext(TEXT_DOMAIN, "filesystem"));
83 	case ZFS_TYPE_SNAPSHOT:
84 		return (dgettext(TEXT_DOMAIN, "snapshot"));
85 	case ZFS_TYPE_VOLUME:
86 		return (dgettext(TEXT_DOMAIN, "volume"));
87 	case ZFS_TYPE_POOL:
88 		return (dgettext(TEXT_DOMAIN, "pool"));
89 	case ZFS_TYPE_BOOKMARK:
90 		return (dgettext(TEXT_DOMAIN, "bookmark"));
91 	default:
92 		assert(!"unhandled zfs_type_t");
93 	}
94 
95 	return (NULL);
96 }
97 
98 /*
99  * Validate a ZFS path.  This is used even before trying to open the dataset, to
100  * provide a more meaningful error message.  We call zfs_error_aux() to
101  * explain exactly why the name was not valid.
102  */
103 int
104 zfs_validate_name(libzfs_handle_t *hdl, const char *path, int type,
105     boolean_t modifying)
106 {
107 	namecheck_err_t why;
108 	char what;
109 
110 	if (!(type & ZFS_TYPE_SNAPSHOT) && strchr(path, '@') != NULL) {
111 		if (hdl != NULL)
112 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
113 			    "snapshot delimiter '@' is not expected here"));
114 		return (0);
115 	}
116 
117 	if (type == ZFS_TYPE_SNAPSHOT && strchr(path, '@') == NULL) {
118 		if (hdl != NULL)
119 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
120 			    "missing '@' delimiter in snapshot name"));
121 		return (0);
122 	}
123 
124 	if (!(type & ZFS_TYPE_BOOKMARK) && strchr(path, '#') != NULL) {
125 		if (hdl != NULL)
126 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
127 			    "bookmark delimiter '#' is not expected here"));
128 		return (0);
129 	}
130 
131 	if (type == ZFS_TYPE_BOOKMARK && strchr(path, '#') == NULL) {
132 		if (hdl != NULL)
133 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
134 			    "missing '#' delimiter in bookmark name"));
135 		return (0);
136 	}
137 
138 	if (modifying && strchr(path, '%') != NULL) {
139 		if (hdl != NULL)
140 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
141 			    "invalid character %c in name"), '%');
142 		return (0);
143 	}
144 
145 	if (entity_namecheck(path, &why, &what) != 0) {
146 		if (hdl != NULL) {
147 			switch (why) {
148 			case NAME_ERR_TOOLONG:
149 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
150 				    "name is too long"));
151 				break;
152 
153 			case NAME_ERR_LEADING_SLASH:
154 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
155 				    "leading slash in name"));
156 				break;
157 
158 			case NAME_ERR_EMPTY_COMPONENT:
159 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
160 				    "empty component or misplaced '@'"
161 				    " or '#' delimiter in name"));
162 				break;
163 
164 			case NAME_ERR_TRAILING_SLASH:
165 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
166 				    "trailing slash in name"));
167 				break;
168 
169 			case NAME_ERR_INVALCHAR:
170 				zfs_error_aux(hdl,
171 				    dgettext(TEXT_DOMAIN, "invalid character "
172 				    "'%c' in name"), what);
173 				break;
174 
175 			case NAME_ERR_MULTIPLE_DELIMITERS:
176 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
177 				    "multiple '@' and/or '#' delimiters in "
178 				    "name"));
179 				break;
180 
181 			case NAME_ERR_NOLETTER:
182 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
183 				    "pool doesn't begin with a letter"));
184 				break;
185 
186 			case NAME_ERR_RESERVED:
187 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
188 				    "name is reserved"));
189 				break;
190 
191 			case NAME_ERR_DISKLIKE:
192 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
193 				    "reserved disk name"));
194 				break;
195 
196 			case NAME_ERR_SELF_REF:
197 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
198 				    "self reference, '.' is found in name"));
199 				break;
200 
201 			case NAME_ERR_PARENT_REF:
202 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
203 				    "parent reference, '..' is found in name"));
204 				break;
205 
206 			default:
207 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
208 				    "(%d) not defined"), why);
209 				break;
210 			}
211 		}
212 
213 		return (0);
214 	}
215 
216 	return (-1);
217 }
218 
219 int
220 zfs_name_valid(const char *name, zfs_type_t type)
221 {
222 	if (type == ZFS_TYPE_POOL)
223 		return (zpool_name_valid(NULL, B_FALSE, name));
224 	return (zfs_validate_name(NULL, name, type, B_FALSE));
225 }
226 
227 /*
228  * This function takes the raw DSL properties, and filters out the user-defined
229  * properties into a separate nvlist.
230  */
231 static nvlist_t *
232 process_user_props(zfs_handle_t *zhp, nvlist_t *props)
233 {
234 	libzfs_handle_t *hdl = zhp->zfs_hdl;
235 	nvpair_t *elem;
236 	nvlist_t *nvl;
237 
238 	if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0) {
239 		(void) no_memory(hdl);
240 		return (NULL);
241 	}
242 
243 	elem = NULL;
244 	while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
245 		if (!zfs_prop_user(nvpair_name(elem)))
246 			continue;
247 
248 		nvlist_t *propval = fnvpair_value_nvlist(elem);
249 		if (nvlist_add_nvlist(nvl, nvpair_name(elem), propval) != 0) {
250 			nvlist_free(nvl);
251 			(void) no_memory(hdl);
252 			return (NULL);
253 		}
254 	}
255 
256 	return (nvl);
257 }
258 
259 static zpool_handle_t *
260 zpool_add_handle(zfs_handle_t *zhp, const char *pool_name)
261 {
262 	libzfs_handle_t *hdl = zhp->zfs_hdl;
263 	zpool_handle_t *zph;
264 
265 	if ((zph = zpool_open_canfail(hdl, pool_name)) != NULL) {
266 		if (hdl->libzfs_pool_handles != NULL)
267 			zph->zpool_next = hdl->libzfs_pool_handles;
268 		hdl->libzfs_pool_handles = zph;
269 	}
270 	return (zph);
271 }
272 
273 static zpool_handle_t *
274 zpool_find_handle(zfs_handle_t *zhp, const char *pool_name, int len)
275 {
276 	libzfs_handle_t *hdl = zhp->zfs_hdl;
277 	zpool_handle_t *zph = hdl->libzfs_pool_handles;
278 
279 	while ((zph != NULL) &&
280 	    (strncmp(pool_name, zpool_get_name(zph), len) != 0))
281 		zph = zph->zpool_next;
282 	return (zph);
283 }
284 
285 /*
286  * Returns a handle to the pool that contains the provided dataset.
287  * If a handle to that pool already exists then that handle is returned.
288  * Otherwise, a new handle is created and added to the list of handles.
289  */
290 static zpool_handle_t *
291 zpool_handle(zfs_handle_t *zhp)
292 {
293 	char *pool_name;
294 	int len;
295 	zpool_handle_t *zph;
296 
297 	len = strcspn(zhp->zfs_name, "/@#") + 1;
298 	pool_name = zfs_alloc(zhp->zfs_hdl, len);
299 	(void) strlcpy(pool_name, zhp->zfs_name, len);
300 
301 	zph = zpool_find_handle(zhp, pool_name, len);
302 	if (zph == NULL)
303 		zph = zpool_add_handle(zhp, pool_name);
304 
305 	free(pool_name);
306 	return (zph);
307 }
308 
309 void
310 zpool_free_handles(libzfs_handle_t *hdl)
311 {
312 	zpool_handle_t *next, *zph = hdl->libzfs_pool_handles;
313 
314 	while (zph != NULL) {
315 		next = zph->zpool_next;
316 		zpool_close(zph);
317 		zph = next;
318 	}
319 	hdl->libzfs_pool_handles = NULL;
320 }
321 
322 /*
323  * Utility function to gather stats (objset and zpl) for the given object.
324  */
325 static int
326 get_stats_ioctl(zfs_handle_t *zhp, zfs_cmd_t *zc)
327 {
328 	libzfs_handle_t *hdl = zhp->zfs_hdl;
329 
330 	(void) strlcpy(zc->zc_name, zhp->zfs_name, sizeof (zc->zc_name));
331 
332 	while (zfs_ioctl(hdl, ZFS_IOC_OBJSET_STATS, zc) != 0) {
333 		if (errno == ENOMEM)
334 			zcmd_expand_dst_nvlist(hdl, zc);
335 		else
336 			return (-1);
337 	}
338 	return (0);
339 }
340 
341 /*
342  * Utility function to get the received properties of the given object.
343  */
344 static int
345 get_recvd_props_ioctl(zfs_handle_t *zhp)
346 {
347 	libzfs_handle_t *hdl = zhp->zfs_hdl;
348 	nvlist_t *recvdprops;
349 	zfs_cmd_t zc = {"\0"};
350 	int err;
351 
352 	zcmd_alloc_dst_nvlist(hdl, &zc, 0);
353 
354 	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
355 
356 	while (zfs_ioctl(hdl, ZFS_IOC_OBJSET_RECVD_PROPS, &zc) != 0) {
357 		if (errno == ENOMEM)
358 			zcmd_expand_dst_nvlist(hdl, &zc);
359 		else {
360 			zcmd_free_nvlists(&zc);
361 			return (-1);
362 		}
363 	}
364 
365 	err = zcmd_read_dst_nvlist(zhp->zfs_hdl, &zc, &recvdprops);
366 	zcmd_free_nvlists(&zc);
367 	if (err != 0)
368 		return (-1);
369 
370 	nvlist_free(zhp->zfs_recvd_props);
371 	zhp->zfs_recvd_props = recvdprops;
372 
373 	return (0);
374 }
375 
376 static int
377 put_stats_zhdl(zfs_handle_t *zhp, zfs_cmd_t *zc)
378 {
379 	nvlist_t *allprops, *userprops;
380 
381 	zhp->zfs_dmustats = zc->zc_objset_stats; /* structure assignment */
382 
383 	if (zcmd_read_dst_nvlist(zhp->zfs_hdl, zc, &allprops) != 0) {
384 		return (-1);
385 	}
386 
387 	/*
388 	 * XXX Why do we store the user props separately, in addition to
389 	 * storing them in zfs_props?
390 	 */
391 	if ((userprops = process_user_props(zhp, allprops)) == NULL) {
392 		nvlist_free(allprops);
393 		return (-1);
394 	}
395 
396 	nvlist_free(zhp->zfs_props);
397 	nvlist_free(zhp->zfs_user_props);
398 
399 	zhp->zfs_props = allprops;
400 	zhp->zfs_user_props = userprops;
401 
402 	return (0);
403 }
404 
405 static int
406 get_stats(zfs_handle_t *zhp)
407 {
408 	int rc = 0;
409 	zfs_cmd_t zc = {"\0"};
410 
411 	zcmd_alloc_dst_nvlist(zhp->zfs_hdl, &zc, 0);
412 
413 	if (get_stats_ioctl(zhp, &zc) != 0)
414 		rc = -1;
415 	else if (put_stats_zhdl(zhp, &zc) != 0)
416 		rc = -1;
417 	zcmd_free_nvlists(&zc);
418 	return (rc);
419 }
420 
421 /*
422  * Refresh the properties currently stored in the handle.
423  */
424 void
425 zfs_refresh_properties(zfs_handle_t *zhp)
426 {
427 	(void) get_stats(zhp);
428 }
429 
430 /*
431  * Makes a handle from the given dataset name.  Used by zfs_open() and
432  * zfs_iter_* to create child handles on the fly.
433  */
434 static int
435 make_dataset_handle_common(zfs_handle_t *zhp, zfs_cmd_t *zc)
436 {
437 	if (put_stats_zhdl(zhp, zc) != 0)
438 		return (-1);
439 
440 	/*
441 	 * We've managed to open the dataset and gather statistics.  Determine
442 	 * the high-level type.
443 	 */
444 	if (zhp->zfs_dmustats.dds_type == DMU_OST_ZVOL) {
445 		zhp->zfs_head_type = ZFS_TYPE_VOLUME;
446 	} else if (zhp->zfs_dmustats.dds_type == DMU_OST_ZFS) {
447 		zhp->zfs_head_type = ZFS_TYPE_FILESYSTEM;
448 	} else if (zhp->zfs_dmustats.dds_type == DMU_OST_OTHER) {
449 		errno = EINVAL;
450 		return (-1);
451 	} else if (zhp->zfs_dmustats.dds_inconsistent) {
452 		errno = EBUSY;
453 		return (-1);
454 	} else {
455 		abort();
456 	}
457 
458 	if (zhp->zfs_dmustats.dds_is_snapshot)
459 		zhp->zfs_type = ZFS_TYPE_SNAPSHOT;
460 	else if (zhp->zfs_dmustats.dds_type == DMU_OST_ZVOL)
461 		zhp->zfs_type = ZFS_TYPE_VOLUME;
462 	else if (zhp->zfs_dmustats.dds_type == DMU_OST_ZFS)
463 		zhp->zfs_type = ZFS_TYPE_FILESYSTEM;
464 	else
465 		abort();	/* we should never see any other types */
466 
467 	if ((zhp->zpool_hdl = zpool_handle(zhp)) == NULL)
468 		return (-1);
469 
470 	return (0);
471 }
472 
473 zfs_handle_t *
474 make_dataset_handle(libzfs_handle_t *hdl, const char *path)
475 {
476 	zfs_cmd_t zc = {"\0"};
477 
478 	zfs_handle_t *zhp = calloc(1, sizeof (zfs_handle_t));
479 
480 	if (zhp == NULL)
481 		return (NULL);
482 
483 	zhp->zfs_hdl = hdl;
484 	(void) strlcpy(zhp->zfs_name, path, sizeof (zhp->zfs_name));
485 	zcmd_alloc_dst_nvlist(hdl, &zc, 0);
486 
487 	if (get_stats_ioctl(zhp, &zc) == -1) {
488 		zcmd_free_nvlists(&zc);
489 		free(zhp);
490 		return (NULL);
491 	}
492 	if (make_dataset_handle_common(zhp, &zc) == -1) {
493 		free(zhp);
494 		zhp = NULL;
495 	}
496 	zcmd_free_nvlists(&zc);
497 	return (zhp);
498 }
499 
500 zfs_handle_t *
501 make_dataset_handle_zc(libzfs_handle_t *hdl, zfs_cmd_t *zc)
502 {
503 	zfs_handle_t *zhp = calloc(1, sizeof (zfs_handle_t));
504 
505 	if (zhp == NULL)
506 		return (NULL);
507 
508 	zhp->zfs_hdl = hdl;
509 	(void) strlcpy(zhp->zfs_name, zc->zc_name, sizeof (zhp->zfs_name));
510 	if (make_dataset_handle_common(zhp, zc) == -1) {
511 		free(zhp);
512 		return (NULL);
513 	}
514 	return (zhp);
515 }
516 
517 zfs_handle_t *
518 make_dataset_simple_handle_zc(zfs_handle_t *pzhp, zfs_cmd_t *zc)
519 {
520 	zfs_handle_t *zhp = calloc(1, sizeof (zfs_handle_t));
521 
522 	if (zhp == NULL)
523 		return (NULL);
524 
525 	zhp->zfs_hdl = pzhp->zfs_hdl;
526 	(void) strlcpy(zhp->zfs_name, zc->zc_name, sizeof (zhp->zfs_name));
527 	zhp->zfs_head_type = pzhp->zfs_type;
528 	zhp->zfs_type = ZFS_TYPE_SNAPSHOT;
529 	zhp->zpool_hdl = zpool_handle(zhp);
530 
531 	if (zc->zc_objset_stats.dds_creation_txg != 0) {
532 		/* structure assignment */
533 		zhp->zfs_dmustats = zc->zc_objset_stats;
534 	} else {
535 		if (get_stats_ioctl(zhp, zc) == -1) {
536 			zcmd_free_nvlists(zc);
537 			free(zhp);
538 			return (NULL);
539 		}
540 		if (make_dataset_handle_common(zhp, zc) == -1) {
541 			zcmd_free_nvlists(zc);
542 			free(zhp);
543 			return (NULL);
544 		}
545 	}
546 
547 	if (zhp->zfs_dmustats.dds_is_snapshot ||
548 	    strchr(zc->zc_name, '@') != NULL)
549 		zhp->zfs_type = ZFS_TYPE_SNAPSHOT;
550 	else if (zhp->zfs_dmustats.dds_type == DMU_OST_ZVOL)
551 		zhp->zfs_type = ZFS_TYPE_VOLUME;
552 	else if (zhp->zfs_dmustats.dds_type == DMU_OST_ZFS)
553 		zhp->zfs_type = ZFS_TYPE_FILESYSTEM;
554 
555 	return (zhp);
556 }
557 
558 zfs_handle_t *
559 zfs_handle_dup(zfs_handle_t *zhp_orig)
560 {
561 	zfs_handle_t *zhp = calloc(1, sizeof (zfs_handle_t));
562 
563 	if (zhp == NULL)
564 		return (NULL);
565 
566 	zhp->zfs_hdl = zhp_orig->zfs_hdl;
567 	zhp->zpool_hdl = zhp_orig->zpool_hdl;
568 	(void) strlcpy(zhp->zfs_name, zhp_orig->zfs_name,
569 	    sizeof (zhp->zfs_name));
570 	zhp->zfs_type = zhp_orig->zfs_type;
571 	zhp->zfs_head_type = zhp_orig->zfs_head_type;
572 	zhp->zfs_dmustats = zhp_orig->zfs_dmustats;
573 	if (zhp_orig->zfs_props != NULL) {
574 		if (nvlist_dup(zhp_orig->zfs_props, &zhp->zfs_props, 0) != 0) {
575 			(void) no_memory(zhp->zfs_hdl);
576 			zfs_close(zhp);
577 			return (NULL);
578 		}
579 	}
580 	if (zhp_orig->zfs_user_props != NULL) {
581 		if (nvlist_dup(zhp_orig->zfs_user_props,
582 		    &zhp->zfs_user_props, 0) != 0) {
583 			(void) no_memory(zhp->zfs_hdl);
584 			zfs_close(zhp);
585 			return (NULL);
586 		}
587 	}
588 	if (zhp_orig->zfs_recvd_props != NULL) {
589 		if (nvlist_dup(zhp_orig->zfs_recvd_props,
590 		    &zhp->zfs_recvd_props, 0)) {
591 			(void) no_memory(zhp->zfs_hdl);
592 			zfs_close(zhp);
593 			return (NULL);
594 		}
595 	}
596 	zhp->zfs_mntcheck = zhp_orig->zfs_mntcheck;
597 	if (zhp_orig->zfs_mntopts != NULL) {
598 		zhp->zfs_mntopts = zfs_strdup(zhp_orig->zfs_hdl,
599 		    zhp_orig->zfs_mntopts);
600 	}
601 	zhp->zfs_props_table = zhp_orig->zfs_props_table;
602 	return (zhp);
603 }
604 
605 boolean_t
606 zfs_bookmark_exists(const char *path)
607 {
608 	nvlist_t *bmarks;
609 	nvlist_t *props;
610 	char fsname[ZFS_MAX_DATASET_NAME_LEN];
611 	char *bmark_name;
612 	char *pound;
613 	int err;
614 	boolean_t rv;
615 
616 	(void) strlcpy(fsname, path, sizeof (fsname));
617 	pound = strchr(fsname, '#');
618 	if (pound == NULL)
619 		return (B_FALSE);
620 
621 	*pound = '\0';
622 	bmark_name = pound + 1;
623 	props = fnvlist_alloc();
624 	err = lzc_get_bookmarks(fsname, props, &bmarks);
625 	nvlist_free(props);
626 	if (err != 0) {
627 		nvlist_free(bmarks);
628 		return (B_FALSE);
629 	}
630 
631 	rv = nvlist_exists(bmarks, bmark_name);
632 	nvlist_free(bmarks);
633 	return (rv);
634 }
635 
636 zfs_handle_t *
637 make_bookmark_handle(zfs_handle_t *parent, const char *path,
638     nvlist_t *bmark_props)
639 {
640 	zfs_handle_t *zhp = calloc(1, sizeof (zfs_handle_t));
641 
642 	if (zhp == NULL)
643 		return (NULL);
644 
645 	/* Fill in the name. */
646 	zhp->zfs_hdl = parent->zfs_hdl;
647 	(void) strlcpy(zhp->zfs_name, path, sizeof (zhp->zfs_name));
648 
649 	/* Set the property lists. */
650 	if (nvlist_dup(bmark_props, &zhp->zfs_props, 0) != 0) {
651 		free(zhp);
652 		return (NULL);
653 	}
654 
655 	/* Set the types. */
656 	zhp->zfs_head_type = parent->zfs_head_type;
657 	zhp->zfs_type = ZFS_TYPE_BOOKMARK;
658 
659 	if ((zhp->zpool_hdl = zpool_handle(zhp)) == NULL) {
660 		nvlist_free(zhp->zfs_props);
661 		free(zhp);
662 		return (NULL);
663 	}
664 
665 	return (zhp);
666 }
667 
668 struct zfs_open_bookmarks_cb_data {
669 	const char *path;
670 	zfs_handle_t *zhp;
671 };
672 
673 static int
674 zfs_open_bookmarks_cb(zfs_handle_t *zhp, void *data)
675 {
676 	struct zfs_open_bookmarks_cb_data *dp = data;
677 
678 	/*
679 	 * Is it the one we are looking for?
680 	 */
681 	if (strcmp(dp->path, zfs_get_name(zhp)) == 0) {
682 		/*
683 		 * We found it.  Save it and let the caller know we are done.
684 		 */
685 		dp->zhp = zhp;
686 		return (EEXIST);
687 	}
688 
689 	/*
690 	 * Not found.  Close the handle and ask for another one.
691 	 */
692 	zfs_close(zhp);
693 	return (0);
694 }
695 
696 /*
697  * Opens the given snapshot, bookmark, filesystem, or volume.   The 'types'
698  * argument is a mask of acceptable types.  The function will print an
699  * appropriate error message and return NULL if it can't be opened.
700  */
701 zfs_handle_t *
702 zfs_open(libzfs_handle_t *hdl, const char *path, int types)
703 {
704 	zfs_handle_t *zhp;
705 	char errbuf[ERRBUFLEN];
706 	char *bookp;
707 
708 	(void) snprintf(errbuf, sizeof (errbuf),
709 	    dgettext(TEXT_DOMAIN, "cannot open '%s'"), path);
710 
711 	/*
712 	 * Validate the name before we even try to open it.
713 	 */
714 	if (!zfs_validate_name(hdl, path, types, B_FALSE)) {
715 		(void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
716 		errno = EINVAL;
717 		return (NULL);
718 	}
719 
720 	/*
721 	 * Bookmarks needs to be handled separately.
722 	 */
723 	bookp = strchr(path, '#');
724 	if (bookp == NULL) {
725 		/*
726 		 * Try to get stats for the dataset, which will tell us if it
727 		 * exists.
728 		 */
729 		errno = 0;
730 		if ((zhp = make_dataset_handle(hdl, path)) == NULL) {
731 			(void) zfs_standard_error(hdl, errno, errbuf);
732 			return (NULL);
733 		}
734 	} else {
735 		char dsname[ZFS_MAX_DATASET_NAME_LEN];
736 		zfs_handle_t *pzhp;
737 		struct zfs_open_bookmarks_cb_data cb_data = {path, NULL};
738 
739 		/*
740 		 * We need to cut out '#' and everything after '#'
741 		 * to get the parent dataset name only.
742 		 */
743 		assert(bookp - path < sizeof (dsname));
744 		(void) strlcpy(dsname, path,
745 		    MIN(sizeof (dsname), bookp - path + 1));
746 
747 		/*
748 		 * Create handle for the parent dataset.
749 		 */
750 		errno = 0;
751 		if ((pzhp = make_dataset_handle(hdl, dsname)) == NULL) {
752 			(void) zfs_standard_error(hdl, errno, errbuf);
753 			return (NULL);
754 		}
755 
756 		/*
757 		 * Iterate bookmarks to find the right one.
758 		 */
759 		errno = 0;
760 		if ((zfs_iter_bookmarks_v2(pzhp, 0, zfs_open_bookmarks_cb,
761 		    &cb_data) == 0) && (cb_data.zhp == NULL)) {
762 			(void) zfs_error(hdl, EZFS_NOENT, errbuf);
763 			zfs_close(pzhp);
764 			errno = ENOENT;
765 			return (NULL);
766 		}
767 		if (cb_data.zhp == NULL) {
768 			(void) zfs_standard_error(hdl, errno, errbuf);
769 			zfs_close(pzhp);
770 			return (NULL);
771 		}
772 		zhp = cb_data.zhp;
773 
774 		/*
775 		 * Cleanup.
776 		 */
777 		zfs_close(pzhp);
778 	}
779 
780 	if (!(types & zhp->zfs_type)) {
781 		(void) zfs_error(hdl, EZFS_BADTYPE, errbuf);
782 		zfs_close(zhp);
783 		errno = EINVAL;
784 		return (NULL);
785 	}
786 
787 	return (zhp);
788 }
789 
790 /*
791  * Release a ZFS handle.  Nothing to do but free the associated memory.
792  */
793 void
794 zfs_close(zfs_handle_t *zhp)
795 {
796 	if (zhp->zfs_mntopts)
797 		free(zhp->zfs_mntopts);
798 	nvlist_free(zhp->zfs_props);
799 	nvlist_free(zhp->zfs_user_props);
800 	nvlist_free(zhp->zfs_recvd_props);
801 	free(zhp);
802 }
803 
804 typedef struct mnttab_node {
805 	struct mnttab mtn_mt;
806 	avl_node_t mtn_node;
807 } mnttab_node_t;
808 
809 static int
810 libzfs_mnttab_cache_compare(const void *arg1, const void *arg2)
811 {
812 	const mnttab_node_t *mtn1 = (const mnttab_node_t *)arg1;
813 	const mnttab_node_t *mtn2 = (const mnttab_node_t *)arg2;
814 	int rv;
815 
816 	rv = strcmp(mtn1->mtn_mt.mnt_special, mtn2->mtn_mt.mnt_special);
817 
818 	return (TREE_ISIGN(rv));
819 }
820 
821 void
822 libzfs_mnttab_init(libzfs_handle_t *hdl)
823 {
824 	pthread_mutex_init(&hdl->libzfs_mnttab_cache_lock, NULL);
825 	assert(avl_numnodes(&hdl->libzfs_mnttab_cache) == 0);
826 	avl_create(&hdl->libzfs_mnttab_cache, libzfs_mnttab_cache_compare,
827 	    sizeof (mnttab_node_t), offsetof(mnttab_node_t, mtn_node));
828 }
829 
830 static int
831 libzfs_mnttab_update(libzfs_handle_t *hdl)
832 {
833 	FILE *mnttab;
834 	struct mnttab entry;
835 
836 	if ((mnttab = fopen(MNTTAB, "re")) == NULL)
837 		return (ENOENT);
838 
839 	while (getmntent(mnttab, &entry) == 0) {
840 		mnttab_node_t *mtn;
841 		avl_index_t where;
842 
843 		if (strcmp(entry.mnt_fstype, MNTTYPE_ZFS) != 0)
844 			continue;
845 
846 		mtn = zfs_alloc(hdl, sizeof (mnttab_node_t));
847 		mtn->mtn_mt.mnt_special = zfs_strdup(hdl, entry.mnt_special);
848 		mtn->mtn_mt.mnt_mountp = zfs_strdup(hdl, entry.mnt_mountp);
849 		mtn->mtn_mt.mnt_fstype = zfs_strdup(hdl, entry.mnt_fstype);
850 		mtn->mtn_mt.mnt_mntopts = zfs_strdup(hdl, entry.mnt_mntopts);
851 
852 		/* Exclude duplicate mounts */
853 		if (avl_find(&hdl->libzfs_mnttab_cache, mtn, &where) != NULL) {
854 			free(mtn->mtn_mt.mnt_special);
855 			free(mtn->mtn_mt.mnt_mountp);
856 			free(mtn->mtn_mt.mnt_fstype);
857 			free(mtn->mtn_mt.mnt_mntopts);
858 			free(mtn);
859 			continue;
860 		}
861 
862 		avl_add(&hdl->libzfs_mnttab_cache, mtn);
863 	}
864 
865 	(void) fclose(mnttab);
866 	return (0);
867 }
868 
869 void
870 libzfs_mnttab_fini(libzfs_handle_t *hdl)
871 {
872 	void *cookie = NULL;
873 	mnttab_node_t *mtn;
874 
875 	while ((mtn = avl_destroy_nodes(&hdl->libzfs_mnttab_cache, &cookie))
876 	    != NULL) {
877 		free(mtn->mtn_mt.mnt_special);
878 		free(mtn->mtn_mt.mnt_mountp);
879 		free(mtn->mtn_mt.mnt_fstype);
880 		free(mtn->mtn_mt.mnt_mntopts);
881 		free(mtn);
882 	}
883 	avl_destroy(&hdl->libzfs_mnttab_cache);
884 	(void) pthread_mutex_destroy(&hdl->libzfs_mnttab_cache_lock);
885 }
886 
887 void
888 libzfs_mnttab_cache(libzfs_handle_t *hdl, boolean_t enable)
889 {
890 	hdl->libzfs_mnttab_enable = enable;
891 }
892 
893 int
894 libzfs_mnttab_find(libzfs_handle_t *hdl, const char *fsname,
895     struct mnttab *entry)
896 {
897 	FILE *mnttab;
898 	mnttab_node_t find;
899 	mnttab_node_t *mtn;
900 	int ret = ENOENT;
901 
902 	if (!hdl->libzfs_mnttab_enable) {
903 		struct mnttab srch = { 0 };
904 
905 		if (avl_numnodes(&hdl->libzfs_mnttab_cache))
906 			libzfs_mnttab_fini(hdl);
907 
908 		if ((mnttab = fopen(MNTTAB, "re")) == NULL)
909 			return (ENOENT);
910 
911 		srch.mnt_special = (char *)fsname;
912 		srch.mnt_fstype = (char *)MNTTYPE_ZFS;
913 		ret = getmntany(mnttab, entry, &srch) ? ENOENT : 0;
914 		(void) fclose(mnttab);
915 		return (ret);
916 	}
917 
918 	pthread_mutex_lock(&hdl->libzfs_mnttab_cache_lock);
919 	if (avl_numnodes(&hdl->libzfs_mnttab_cache) == 0) {
920 		int error;
921 
922 		if ((error = libzfs_mnttab_update(hdl)) != 0) {
923 			pthread_mutex_unlock(&hdl->libzfs_mnttab_cache_lock);
924 			return (error);
925 		}
926 	}
927 
928 	find.mtn_mt.mnt_special = (char *)fsname;
929 	mtn = avl_find(&hdl->libzfs_mnttab_cache, &find, NULL);
930 	if (mtn) {
931 		*entry = mtn->mtn_mt;
932 		ret = 0;
933 	}
934 	pthread_mutex_unlock(&hdl->libzfs_mnttab_cache_lock);
935 	return (ret);
936 }
937 
938 void
939 libzfs_mnttab_add(libzfs_handle_t *hdl, const char *special,
940     const char *mountp, const char *mntopts)
941 {
942 	mnttab_node_t *mtn;
943 
944 	pthread_mutex_lock(&hdl->libzfs_mnttab_cache_lock);
945 	if (avl_numnodes(&hdl->libzfs_mnttab_cache) != 0) {
946 		mtn = zfs_alloc(hdl, sizeof (mnttab_node_t));
947 		mtn->mtn_mt.mnt_special = zfs_strdup(hdl, special);
948 		mtn->mtn_mt.mnt_mountp = zfs_strdup(hdl, mountp);
949 		mtn->mtn_mt.mnt_fstype = zfs_strdup(hdl, MNTTYPE_ZFS);
950 		mtn->mtn_mt.mnt_mntopts = zfs_strdup(hdl, mntopts);
951 		/*
952 		 * Another thread may have already added this entry
953 		 * via libzfs_mnttab_update. If so we should skip it.
954 		 */
955 		if (avl_find(&hdl->libzfs_mnttab_cache, mtn, NULL) != NULL) {
956 			free(mtn->mtn_mt.mnt_special);
957 			free(mtn->mtn_mt.mnt_mountp);
958 			free(mtn->mtn_mt.mnt_fstype);
959 			free(mtn->mtn_mt.mnt_mntopts);
960 			free(mtn);
961 		} else {
962 			avl_add(&hdl->libzfs_mnttab_cache, mtn);
963 		}
964 	}
965 	pthread_mutex_unlock(&hdl->libzfs_mnttab_cache_lock);
966 }
967 
968 void
969 libzfs_mnttab_remove(libzfs_handle_t *hdl, const char *fsname)
970 {
971 	mnttab_node_t find;
972 	mnttab_node_t *ret;
973 
974 	pthread_mutex_lock(&hdl->libzfs_mnttab_cache_lock);
975 	find.mtn_mt.mnt_special = (char *)fsname;
976 	if ((ret = avl_find(&hdl->libzfs_mnttab_cache, (void *)&find, NULL))
977 	    != NULL) {
978 		avl_remove(&hdl->libzfs_mnttab_cache, ret);
979 		free(ret->mtn_mt.mnt_special);
980 		free(ret->mtn_mt.mnt_mountp);
981 		free(ret->mtn_mt.mnt_fstype);
982 		free(ret->mtn_mt.mnt_mntopts);
983 		free(ret);
984 	}
985 	pthread_mutex_unlock(&hdl->libzfs_mnttab_cache_lock);
986 }
987 
988 int
989 zfs_spa_version(zfs_handle_t *zhp, int *spa_version)
990 {
991 	zpool_handle_t *zpool_handle = zhp->zpool_hdl;
992 
993 	if (zpool_handle == NULL)
994 		return (-1);
995 
996 	*spa_version = zpool_get_prop_int(zpool_handle,
997 	    ZPOOL_PROP_VERSION, NULL);
998 	return (0);
999 }
1000 
1001 /*
1002  * The choice of reservation property depends on the SPA version.
1003  */
1004 static int
1005 zfs_which_resv_prop(zfs_handle_t *zhp, zfs_prop_t *resv_prop)
1006 {
1007 	int spa_version;
1008 
1009 	if (zfs_spa_version(zhp, &spa_version) < 0)
1010 		return (-1);
1011 
1012 	if (spa_version >= SPA_VERSION_REFRESERVATION)
1013 		*resv_prop = ZFS_PROP_REFRESERVATION;
1014 	else
1015 		*resv_prop = ZFS_PROP_RESERVATION;
1016 
1017 	return (0);
1018 }
1019 
1020 /*
1021  * Given an nvlist of properties to set, validates that they are correct, and
1022  * parses any numeric properties (index, boolean, etc) if they are specified as
1023  * strings.
1024  */
1025 nvlist_t *
1026 zfs_valid_proplist(libzfs_handle_t *hdl, zfs_type_t type, nvlist_t *nvl,
1027     uint64_t zoned, zfs_handle_t *zhp, zpool_handle_t *zpool_hdl,
1028     boolean_t key_params_ok, const char *errbuf)
1029 {
1030 	nvpair_t *elem;
1031 	uint64_t intval;
1032 	const char *strval;
1033 	zfs_prop_t prop;
1034 	nvlist_t *ret;
1035 	int chosen_normal = -1;
1036 	int chosen_utf = -1;
1037 	int set_maxbs = 0;
1038 
1039 	if (nvlist_alloc(&ret, NV_UNIQUE_NAME, 0) != 0) {
1040 		(void) no_memory(hdl);
1041 		return (NULL);
1042 	}
1043 
1044 	/*
1045 	 * Make sure this property is valid and applies to this type.
1046 	 */
1047 
1048 	elem = NULL;
1049 	while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
1050 		const char *propname = nvpair_name(elem);
1051 
1052 		prop = zfs_name_to_prop(propname);
1053 		if (prop == ZPROP_USERPROP && zfs_prop_user(propname)) {
1054 			/*
1055 			 * This is a user property: make sure it's a
1056 			 * string, and that it's less than ZAP_MAXNAMELEN.
1057 			 */
1058 			if (nvpair_type(elem) != DATA_TYPE_STRING) {
1059 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1060 				    "'%s' must be a string"), propname);
1061 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1062 				goto error;
1063 			}
1064 
1065 			if (strlen(nvpair_name(elem)) >= ZAP_MAXNAMELEN) {
1066 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1067 				    "property name '%s' is too long"),
1068 				    propname);
1069 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1070 				goto error;
1071 			}
1072 
1073 			(void) nvpair_value_string(elem, &strval);
1074 			if (nvlist_add_string(ret, propname, strval) != 0) {
1075 				(void) no_memory(hdl);
1076 				goto error;
1077 			}
1078 			continue;
1079 		}
1080 
1081 		/*
1082 		 * Currently, only user properties can be modified on
1083 		 * snapshots.
1084 		 */
1085 		if (type == ZFS_TYPE_SNAPSHOT) {
1086 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1087 			    "this property can not be modified for snapshots"));
1088 			(void) zfs_error(hdl, EZFS_PROPTYPE, errbuf);
1089 			goto error;
1090 		}
1091 
1092 		if (prop == ZPROP_USERPROP && zfs_prop_userquota(propname)) {
1093 			zfs_userquota_prop_t uqtype;
1094 			char *newpropname = NULL;
1095 			char domain[128];
1096 			uint64_t rid;
1097 			uint64_t valary[3];
1098 			int rc;
1099 
1100 			if (userquota_propname_decode(propname, zoned,
1101 			    &uqtype, domain, sizeof (domain), &rid) != 0) {
1102 				zfs_error_aux(hdl,
1103 				    dgettext(TEXT_DOMAIN,
1104 				    "'%s' has an invalid user/group name"),
1105 				    propname);
1106 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1107 				goto error;
1108 			}
1109 
1110 			if (uqtype != ZFS_PROP_USERQUOTA &&
1111 			    uqtype != ZFS_PROP_GROUPQUOTA &&
1112 			    uqtype != ZFS_PROP_USEROBJQUOTA &&
1113 			    uqtype != ZFS_PROP_GROUPOBJQUOTA &&
1114 			    uqtype != ZFS_PROP_PROJECTQUOTA &&
1115 			    uqtype != ZFS_PROP_PROJECTOBJQUOTA) {
1116 				zfs_error_aux(hdl,
1117 				    dgettext(TEXT_DOMAIN, "'%s' is readonly"),
1118 				    propname);
1119 				(void) zfs_error(hdl, EZFS_PROPREADONLY,
1120 				    errbuf);
1121 				goto error;
1122 			}
1123 
1124 			if (nvpair_type(elem) == DATA_TYPE_STRING) {
1125 				(void) nvpair_value_string(elem, &strval);
1126 				if (strcmp(strval, "none") == 0) {
1127 					intval = 0;
1128 				} else if (zfs_nicestrtonum(hdl,
1129 				    strval, &intval) != 0) {
1130 					(void) zfs_error(hdl,
1131 					    EZFS_BADPROP, errbuf);
1132 					goto error;
1133 				}
1134 			} else if (nvpair_type(elem) ==
1135 			    DATA_TYPE_UINT64) {
1136 				(void) nvpair_value_uint64(elem, &intval);
1137 				if (intval == 0) {
1138 					zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1139 					    "use 'none' to disable "
1140 					    "{user|group|project}quota"));
1141 					goto error;
1142 				}
1143 			} else {
1144 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1145 				    "'%s' must be a number"), propname);
1146 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1147 				goto error;
1148 			}
1149 
1150 			/*
1151 			 * Encode the prop name as
1152 			 * userquota@<hex-rid>-domain, to make it easy
1153 			 * for the kernel to decode.
1154 			 */
1155 			rc = asprintf(&newpropname, "%s%llx-%s",
1156 			    zfs_userquota_prop_prefixes[uqtype],
1157 			    (longlong_t)rid, domain);
1158 			if (rc == -1 || newpropname == NULL) {
1159 				(void) no_memory(hdl);
1160 				goto error;
1161 			}
1162 
1163 			valary[0] = uqtype;
1164 			valary[1] = rid;
1165 			valary[2] = intval;
1166 			if (nvlist_add_uint64_array(ret, newpropname,
1167 			    valary, 3) != 0) {
1168 				free(newpropname);
1169 				(void) no_memory(hdl);
1170 				goto error;
1171 			}
1172 			free(newpropname);
1173 			continue;
1174 		} else if (prop == ZPROP_USERPROP &&
1175 		    zfs_prop_written(propname)) {
1176 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1177 			    "'%s' is readonly"),
1178 			    propname);
1179 			(void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
1180 			goto error;
1181 		}
1182 
1183 		if (prop == ZPROP_INVAL) {
1184 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1185 			    "invalid property '%s'"), propname);
1186 			(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1187 			goto error;
1188 		}
1189 
1190 		if (!zfs_prop_valid_for_type(prop, type, B_FALSE)) {
1191 			zfs_error_aux(hdl,
1192 			    dgettext(TEXT_DOMAIN, "'%s' does not "
1193 			    "apply to datasets of this type"), propname);
1194 			(void) zfs_error(hdl, EZFS_PROPTYPE, errbuf);
1195 			goto error;
1196 		}
1197 
1198 		if (zfs_prop_readonly(prop) &&
1199 		    !(zfs_prop_setonce(prop) && zhp == NULL) &&
1200 		    !(zfs_prop_encryption_key_param(prop) && key_params_ok)) {
1201 			zfs_error_aux(hdl,
1202 			    dgettext(TEXT_DOMAIN, "'%s' is readonly"),
1203 			    propname);
1204 			(void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
1205 			goto error;
1206 		}
1207 
1208 		if (zprop_parse_value(hdl, elem, prop, type, ret,
1209 		    &strval, &intval, errbuf) != 0)
1210 			goto error;
1211 
1212 		/*
1213 		 * Perform some additional checks for specific properties.
1214 		 */
1215 		switch (prop) {
1216 		case ZFS_PROP_VERSION:
1217 		{
1218 			int version;
1219 
1220 			if (zhp == NULL)
1221 				break;
1222 			version = zfs_prop_get_int(zhp, ZFS_PROP_VERSION);
1223 			if (intval < version) {
1224 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1225 				    "Can not downgrade; already at version %u"),
1226 				    version);
1227 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1228 				goto error;
1229 			}
1230 			break;
1231 		}
1232 
1233 		case ZFS_PROP_VOLBLOCKSIZE:
1234 		case ZFS_PROP_RECORDSIZE:
1235 		{
1236 			int maxbs = SPA_MAXBLOCKSIZE;
1237 			char buf[64];
1238 
1239 			if (zpool_hdl != NULL) {
1240 				maxbs = zpool_get_prop_int(zpool_hdl,
1241 				    ZPOOL_PROP_MAXBLOCKSIZE, NULL);
1242 			}
1243 			/*
1244 			 * The value must be a power of two between
1245 			 * SPA_MINBLOCKSIZE and maxbs.
1246 			 */
1247 			if (intval < SPA_MINBLOCKSIZE ||
1248 			    intval > maxbs || !ISP2(intval)) {
1249 				zfs_nicebytes(maxbs, buf, sizeof (buf));
1250 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1251 				    "'%s' must be power of 2 from 512B "
1252 				    "to %s"), propname, buf);
1253 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1254 				goto error;
1255 			}
1256 			/* save the ZFS_PROP_RECORDSIZE during create op */
1257 			if (zpool_hdl == NULL && prop == ZFS_PROP_RECORDSIZE) {
1258 				set_maxbs = intval;
1259 			}
1260 			break;
1261 		}
1262 
1263 		case ZFS_PROP_SPECIAL_SMALL_BLOCKS:
1264 		{
1265 			int maxbs =
1266 			    set_maxbs == 0 ? SPA_OLD_MAXBLOCKSIZE : set_maxbs;
1267 			char buf[64];
1268 
1269 			if (zpool_hdl != NULL) {
1270 				char state[64] = "";
1271 
1272 				maxbs = zpool_get_prop_int(zpool_hdl,
1273 				    ZPOOL_PROP_MAXBLOCKSIZE, NULL);
1274 
1275 				/*
1276 				 * Issue a warning but do not fail so that
1277 				 * tests for settable properties succeed.
1278 				 */
1279 				if (zpool_prop_get_feature(zpool_hdl,
1280 				    "feature@allocation_classes", state,
1281 				    sizeof (state)) != 0 ||
1282 				    strcmp(state, ZFS_FEATURE_ACTIVE) != 0) {
1283 					(void) fprintf(stderr, gettext(
1284 					    "%s: property requires a special "
1285 					    "device in the pool\n"), propname);
1286 				}
1287 			}
1288 			if (intval != 0 &&
1289 			    (intval < SPA_MINBLOCKSIZE ||
1290 			    intval > maxbs || !ISP2(intval))) {
1291 				zfs_nicebytes(maxbs, buf, sizeof (buf));
1292 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1293 				    "invalid '%s=%llu' property: must be zero "
1294 				    "or a power of 2 from 512B to %s"),
1295 				    propname, (unsigned long long)intval, buf);
1296 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1297 				goto error;
1298 			}
1299 			break;
1300 		}
1301 
1302 		case ZFS_PROP_MLSLABEL:
1303 		{
1304 #ifdef HAVE_MLSLABEL
1305 			/*
1306 			 * Verify the mlslabel string and convert to
1307 			 * internal hex label string.
1308 			 */
1309 
1310 			m_label_t *new_sl;
1311 			char *hex = NULL;	/* internal label string */
1312 
1313 			/* Default value is already OK. */
1314 			if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
1315 				break;
1316 
1317 			/* Verify the label can be converted to binary form */
1318 			if (((new_sl = m_label_alloc(MAC_LABEL)) == NULL) ||
1319 			    (str_to_label(strval, &new_sl, MAC_LABEL,
1320 			    L_NO_CORRECTION, NULL) == -1)) {
1321 				goto badlabel;
1322 			}
1323 
1324 			/* Now translate to hex internal label string */
1325 			if (label_to_str(new_sl, &hex, M_INTERNAL,
1326 			    DEF_NAMES) != 0) {
1327 				if (hex)
1328 					free(hex);
1329 				goto badlabel;
1330 			}
1331 			m_label_free(new_sl);
1332 
1333 			/* If string is already in internal form, we're done. */
1334 			if (strcmp(strval, hex) == 0) {
1335 				free(hex);
1336 				break;
1337 			}
1338 
1339 			/* Replace the label string with the internal form. */
1340 			(void) nvlist_remove(ret, zfs_prop_to_name(prop),
1341 			    DATA_TYPE_STRING);
1342 			fnvlist_add_string(ret, zfs_prop_to_name(prop), hex);
1343 			free(hex);
1344 
1345 			break;
1346 
1347 badlabel:
1348 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1349 			    "invalid mlslabel '%s'"), strval);
1350 			(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1351 			m_label_free(new_sl);	/* OK if null */
1352 			goto error;
1353 #else
1354 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1355 			    "mlslabels are unsupported"));
1356 			(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1357 			goto error;
1358 #endif /* HAVE_MLSLABEL */
1359 		}
1360 
1361 		case ZFS_PROP_MOUNTPOINT:
1362 		{
1363 			namecheck_err_t why;
1364 
1365 			if (strcmp(strval, ZFS_MOUNTPOINT_NONE) == 0 ||
1366 			    strcmp(strval, ZFS_MOUNTPOINT_LEGACY) == 0)
1367 				break;
1368 
1369 			if (mountpoint_namecheck(strval, &why)) {
1370 				switch (why) {
1371 				case NAME_ERR_LEADING_SLASH:
1372 					zfs_error_aux(hdl,
1373 					    dgettext(TEXT_DOMAIN,
1374 					    "'%s' must be an absolute path, "
1375 					    "'none', or 'legacy'"), propname);
1376 					break;
1377 				case NAME_ERR_TOOLONG:
1378 					zfs_error_aux(hdl,
1379 					    dgettext(TEXT_DOMAIN,
1380 					    "component of '%s' is too long"),
1381 					    propname);
1382 					break;
1383 
1384 				default:
1385 					zfs_error_aux(hdl,
1386 					    dgettext(TEXT_DOMAIN,
1387 					    "(%d) not defined"),
1388 					    why);
1389 					break;
1390 				}
1391 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1392 				goto error;
1393 			}
1394 			zfs_fallthrough;
1395 		}
1396 
1397 		case ZFS_PROP_SHARESMB:
1398 		case ZFS_PROP_SHARENFS:
1399 			/*
1400 			 * For the mountpoint and sharenfs or sharesmb
1401 			 * properties, check if it can be set in a
1402 			 * global/non-global zone based on
1403 			 * the zoned property value:
1404 			 *
1405 			 *		global zone	    non-global zone
1406 			 * --------------------------------------------------
1407 			 * zoned=on	mountpoint (no)	    mountpoint (yes)
1408 			 *		sharenfs (no)	    sharenfs (no)
1409 			 *		sharesmb (no)	    sharesmb (no)
1410 			 *
1411 			 * zoned=off	mountpoint (yes)	N/A
1412 			 *		sharenfs (yes)
1413 			 *		sharesmb (yes)
1414 			 */
1415 			if (zoned) {
1416 				if (getzoneid() == GLOBAL_ZONEID) {
1417 					zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1418 					    "'%s' cannot be set on "
1419 					    "dataset in a non-global zone"),
1420 					    propname);
1421 					(void) zfs_error(hdl, EZFS_ZONED,
1422 					    errbuf);
1423 					goto error;
1424 				} else if (prop == ZFS_PROP_SHARENFS ||
1425 				    prop == ZFS_PROP_SHARESMB) {
1426 					zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1427 					    "'%s' cannot be set in "
1428 					    "a non-global zone"), propname);
1429 					(void) zfs_error(hdl, EZFS_ZONED,
1430 					    errbuf);
1431 					goto error;
1432 				}
1433 			} else if (getzoneid() != GLOBAL_ZONEID) {
1434 				/*
1435 				 * If zoned property is 'off', this must be in
1436 				 * a global zone. If not, something is wrong.
1437 				 */
1438 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1439 				    "'%s' cannot be set while dataset "
1440 				    "'zoned' property is set"), propname);
1441 				(void) zfs_error(hdl, EZFS_ZONED, errbuf);
1442 				goto error;
1443 			}
1444 
1445 			/*
1446 			 * At this point, it is legitimate to set the
1447 			 * property. Now we want to make sure that the
1448 			 * property value is valid if it is sharenfs.
1449 			 */
1450 			if ((prop == ZFS_PROP_SHARENFS ||
1451 			    prop == ZFS_PROP_SHARESMB) &&
1452 			    strcmp(strval, "on") != 0 &&
1453 			    strcmp(strval, "off") != 0) {
1454 				enum sa_protocol proto;
1455 
1456 				if (prop == ZFS_PROP_SHARESMB)
1457 					proto = SA_PROTOCOL_SMB;
1458 				else
1459 					proto = SA_PROTOCOL_NFS;
1460 
1461 				if (sa_validate_shareopts(strval, proto) !=
1462 				    SA_OK) {
1463 					zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1464 					    "'%s' cannot be set to invalid "
1465 					    "options"), propname);
1466 					(void) zfs_error(hdl, EZFS_BADPROP,
1467 					    errbuf);
1468 					goto error;
1469 				}
1470 			}
1471 
1472 			break;
1473 
1474 		case ZFS_PROP_KEYLOCATION:
1475 			if (!zfs_prop_valid_keylocation(strval, B_FALSE)) {
1476 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1477 				    "invalid keylocation"));
1478 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1479 				goto error;
1480 			}
1481 
1482 			if (zhp != NULL) {
1483 				uint64_t crypt =
1484 				    zfs_prop_get_int(zhp, ZFS_PROP_ENCRYPTION);
1485 
1486 				if (crypt == ZIO_CRYPT_OFF &&
1487 				    strcmp(strval, "none") != 0) {
1488 					zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1489 					    "keylocation must be 'none' "
1490 					    "for unencrypted datasets"));
1491 					(void) zfs_error(hdl, EZFS_BADPROP,
1492 					    errbuf);
1493 					goto error;
1494 				} else if (crypt != ZIO_CRYPT_OFF &&
1495 				    strcmp(strval, "none") == 0) {
1496 					zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1497 					    "keylocation must not be 'none' "
1498 					    "for encrypted datasets"));
1499 					(void) zfs_error(hdl, EZFS_BADPROP,
1500 					    errbuf);
1501 					goto error;
1502 				}
1503 			}
1504 			break;
1505 
1506 		case ZFS_PROP_PBKDF2_ITERS:
1507 			if (intval < MIN_PBKDF2_ITERATIONS) {
1508 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1509 				    "minimum pbkdf2 iterations is %u"),
1510 				    MIN_PBKDF2_ITERATIONS);
1511 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1512 				goto error;
1513 			}
1514 			break;
1515 
1516 		case ZFS_PROP_UTF8ONLY:
1517 			chosen_utf = (int)intval;
1518 			break;
1519 
1520 		case ZFS_PROP_NORMALIZE:
1521 			chosen_normal = (int)intval;
1522 			break;
1523 
1524 		default:
1525 			break;
1526 		}
1527 
1528 		/*
1529 		 * For changes to existing volumes, we have some additional
1530 		 * checks to enforce.
1531 		 */
1532 		if (type == ZFS_TYPE_VOLUME && zhp != NULL) {
1533 			uint64_t blocksize = zfs_prop_get_int(zhp,
1534 			    ZFS_PROP_VOLBLOCKSIZE);
1535 			char buf[64];
1536 
1537 			switch (prop) {
1538 			case ZFS_PROP_VOLSIZE:
1539 				if (intval % blocksize != 0) {
1540 					zfs_nicebytes(blocksize, buf,
1541 					    sizeof (buf));
1542 					zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1543 					    "'%s' must be a multiple of "
1544 					    "volume block size (%s)"),
1545 					    propname, buf);
1546 					(void) zfs_error(hdl, EZFS_BADPROP,
1547 					    errbuf);
1548 					goto error;
1549 				}
1550 
1551 				if (intval == 0) {
1552 					zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1553 					    "'%s' cannot be zero"),
1554 					    propname);
1555 					(void) zfs_error(hdl, EZFS_BADPROP,
1556 					    errbuf);
1557 					goto error;
1558 				}
1559 				break;
1560 
1561 			default:
1562 				break;
1563 			}
1564 		}
1565 
1566 		/* check encryption properties */
1567 		if (zhp != NULL) {
1568 			int64_t crypt = zfs_prop_get_int(zhp,
1569 			    ZFS_PROP_ENCRYPTION);
1570 
1571 			switch (prop) {
1572 			case ZFS_PROP_COPIES:
1573 				if (crypt != ZIO_CRYPT_OFF && intval > 2) {
1574 					zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1575 					    "encrypted datasets cannot have "
1576 					    "3 copies"));
1577 					(void) zfs_error(hdl, EZFS_BADPROP,
1578 					    errbuf);
1579 					goto error;
1580 				}
1581 				break;
1582 			default:
1583 				break;
1584 			}
1585 		}
1586 	}
1587 
1588 	/*
1589 	 * If normalization was chosen, but no UTF8 choice was made,
1590 	 * enforce rejection of non-UTF8 names.
1591 	 *
1592 	 * If normalization was chosen, but rejecting non-UTF8 names
1593 	 * was explicitly not chosen, it is an error.
1594 	 *
1595 	 * If utf8only was turned off, but the parent has normalization,
1596 	 * turn off normalization.
1597 	 */
1598 	if (chosen_normal > 0 && chosen_utf < 0) {
1599 		if (nvlist_add_uint64(ret,
1600 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), 1) != 0) {
1601 			(void) no_memory(hdl);
1602 			goto error;
1603 		}
1604 	} else if (chosen_normal > 0 && chosen_utf == 0) {
1605 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1606 		    "'%s' must be set 'on' if normalization chosen"),
1607 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
1608 		(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1609 		goto error;
1610 	} else if (chosen_normal < 0 && chosen_utf == 0) {
1611 		if (nvlist_add_uint64(ret,
1612 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE), 0) != 0) {
1613 			(void) no_memory(hdl);
1614 			goto error;
1615 		}
1616 	}
1617 	return (ret);
1618 
1619 error:
1620 	nvlist_free(ret);
1621 	return (NULL);
1622 }
1623 
1624 static int
1625 zfs_add_synthetic_resv(zfs_handle_t *zhp, nvlist_t *nvl)
1626 {
1627 	uint64_t old_volsize;
1628 	uint64_t new_volsize;
1629 	uint64_t old_reservation;
1630 	uint64_t new_reservation;
1631 	zfs_prop_t resv_prop;
1632 	nvlist_t *props;
1633 	zpool_handle_t *zph = zpool_handle(zhp);
1634 
1635 	/*
1636 	 * If this is an existing volume, and someone is setting the volsize,
1637 	 * make sure that it matches the reservation, or add it if necessary.
1638 	 */
1639 	old_volsize = zfs_prop_get_int(zhp, ZFS_PROP_VOLSIZE);
1640 	if (zfs_which_resv_prop(zhp, &resv_prop) < 0)
1641 		return (-1);
1642 	old_reservation = zfs_prop_get_int(zhp, resv_prop);
1643 
1644 	props = fnvlist_alloc();
1645 	fnvlist_add_uint64(props, zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
1646 	    zfs_prop_get_int(zhp, ZFS_PROP_VOLBLOCKSIZE));
1647 
1648 	if ((zvol_volsize_to_reservation(zph, old_volsize, props) !=
1649 	    old_reservation) || nvlist_exists(nvl,
1650 	    zfs_prop_to_name(resv_prop))) {
1651 		fnvlist_free(props);
1652 		return (0);
1653 	}
1654 	if (nvlist_lookup_uint64(nvl, zfs_prop_to_name(ZFS_PROP_VOLSIZE),
1655 	    &new_volsize) != 0) {
1656 		fnvlist_free(props);
1657 		return (-1);
1658 	}
1659 	new_reservation = zvol_volsize_to_reservation(zph, new_volsize, props);
1660 	fnvlist_free(props);
1661 
1662 	if (nvlist_add_uint64(nvl, zfs_prop_to_name(resv_prop),
1663 	    new_reservation) != 0) {
1664 		(void) no_memory(zhp->zfs_hdl);
1665 		return (-1);
1666 	}
1667 	return (1);
1668 }
1669 
1670 /*
1671  * Helper for 'zfs {set|clone} refreservation=auto'.  Must be called after
1672  * zfs_valid_proplist(), as it is what sets the UINT64_MAX sentinel value.
1673  * Return codes must match zfs_add_synthetic_resv().
1674  */
1675 static int
1676 zfs_fix_auto_resv(zfs_handle_t *zhp, nvlist_t *nvl)
1677 {
1678 	uint64_t volsize;
1679 	uint64_t resvsize;
1680 	zfs_prop_t prop;
1681 	nvlist_t *props;
1682 
1683 	if (!ZFS_IS_VOLUME(zhp)) {
1684 		return (0);
1685 	}
1686 
1687 	if (zfs_which_resv_prop(zhp, &prop) != 0) {
1688 		return (-1);
1689 	}
1690 
1691 	if (prop != ZFS_PROP_REFRESERVATION) {
1692 		return (0);
1693 	}
1694 
1695 	if (nvlist_lookup_uint64(nvl, zfs_prop_to_name(prop), &resvsize) != 0) {
1696 		/* No value being set, so it can't be "auto" */
1697 		return (0);
1698 	}
1699 	if (resvsize != UINT64_MAX) {
1700 		/* Being set to a value other than "auto" */
1701 		return (0);
1702 	}
1703 
1704 	props = fnvlist_alloc();
1705 
1706 	fnvlist_add_uint64(props, zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
1707 	    zfs_prop_get_int(zhp, ZFS_PROP_VOLBLOCKSIZE));
1708 
1709 	if (nvlist_lookup_uint64(nvl, zfs_prop_to_name(ZFS_PROP_VOLSIZE),
1710 	    &volsize) != 0) {
1711 		volsize = zfs_prop_get_int(zhp, ZFS_PROP_VOLSIZE);
1712 	}
1713 
1714 	resvsize = zvol_volsize_to_reservation(zpool_handle(zhp), volsize,
1715 	    props);
1716 	fnvlist_free(props);
1717 
1718 	(void) nvlist_remove_all(nvl, zfs_prop_to_name(prop));
1719 	if (nvlist_add_uint64(nvl, zfs_prop_to_name(prop), resvsize) != 0) {
1720 		(void) no_memory(zhp->zfs_hdl);
1721 		return (-1);
1722 	}
1723 	return (1);
1724 }
1725 
1726 static boolean_t
1727 zfs_is_namespace_prop(zfs_prop_t prop)
1728 {
1729 	switch (prop) {
1730 
1731 	case ZFS_PROP_ATIME:
1732 	case ZFS_PROP_RELATIME:
1733 	case ZFS_PROP_DEVICES:
1734 	case ZFS_PROP_EXEC:
1735 	case ZFS_PROP_SETUID:
1736 	case ZFS_PROP_READONLY:
1737 	case ZFS_PROP_XATTR:
1738 	case ZFS_PROP_NBMAND:
1739 		return (B_TRUE);
1740 
1741 	default:
1742 		return (B_FALSE);
1743 	}
1744 }
1745 
1746 /*
1747  * Given a property name and value, set the property for the given dataset.
1748  */
1749 int
1750 zfs_prop_set(zfs_handle_t *zhp, const char *propname, const char *propval)
1751 {
1752 	int ret = -1;
1753 	char errbuf[ERRBUFLEN];
1754 	libzfs_handle_t *hdl = zhp->zfs_hdl;
1755 	nvlist_t *nvl = NULL;
1756 
1757 	(void) snprintf(errbuf, sizeof (errbuf),
1758 	    dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
1759 	    zhp->zfs_name);
1760 
1761 	if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0 ||
1762 	    nvlist_add_string(nvl, propname, propval) != 0) {
1763 		(void) no_memory(hdl);
1764 		goto error;
1765 	}
1766 
1767 	ret = zfs_prop_set_list(zhp, nvl);
1768 
1769 error:
1770 	nvlist_free(nvl);
1771 	return (ret);
1772 }
1773 
1774 
1775 
1776 /*
1777  * Given an nvlist of property names and values, set the properties for the
1778  * given dataset.
1779  */
1780 int
1781 zfs_prop_set_list(zfs_handle_t *zhp, nvlist_t *props)
1782 {
1783 	zfs_cmd_t zc = {"\0"};
1784 	int ret = -1;
1785 	prop_changelist_t **cls = NULL;
1786 	int cl_idx;
1787 	char errbuf[ERRBUFLEN];
1788 	libzfs_handle_t *hdl = zhp->zfs_hdl;
1789 	nvlist_t *nvl;
1790 	int nvl_len = 0;
1791 	int added_resv = 0;
1792 	zfs_prop_t prop = 0;
1793 	nvpair_t *elem;
1794 
1795 	(void) snprintf(errbuf, sizeof (errbuf),
1796 	    dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
1797 	    zhp->zfs_name);
1798 
1799 	if ((nvl = zfs_valid_proplist(hdl, zhp->zfs_type, props,
1800 	    zfs_prop_get_int(zhp, ZFS_PROP_ZONED), zhp, zhp->zpool_hdl,
1801 	    B_FALSE, errbuf)) == NULL)
1802 		goto error;
1803 
1804 	/*
1805 	 * We have to check for any extra properties which need to be added
1806 	 * before computing the length of the nvlist.
1807 	 */
1808 	for (elem = nvlist_next_nvpair(nvl, NULL);
1809 	    elem != NULL;
1810 	    elem = nvlist_next_nvpair(nvl, elem)) {
1811 		if (zfs_name_to_prop(nvpair_name(elem)) == ZFS_PROP_VOLSIZE &&
1812 		    (added_resv = zfs_add_synthetic_resv(zhp, nvl)) == -1) {
1813 			goto error;
1814 		}
1815 	}
1816 
1817 	if (added_resv != 1 &&
1818 	    (added_resv = zfs_fix_auto_resv(zhp, nvl)) == -1) {
1819 		goto error;
1820 	}
1821 
1822 	/*
1823 	 * Check how many properties we're setting and allocate an array to
1824 	 * store changelist pointers for postfix().
1825 	 */
1826 	for (elem = nvlist_next_nvpair(nvl, NULL);
1827 	    elem != NULL;
1828 	    elem = nvlist_next_nvpair(nvl, elem))
1829 		nvl_len++;
1830 	if ((cls = calloc(nvl_len, sizeof (prop_changelist_t *))) == NULL)
1831 		goto error;
1832 
1833 	cl_idx = 0;
1834 	for (elem = nvlist_next_nvpair(nvl, NULL);
1835 	    elem != NULL;
1836 	    elem = nvlist_next_nvpair(nvl, elem)) {
1837 
1838 		prop = zfs_name_to_prop(nvpair_name(elem));
1839 
1840 		assert(cl_idx < nvl_len);
1841 		/*
1842 		 * We don't want to unmount & remount the dataset when changing
1843 		 * its canmount property to 'on' or 'noauto'.  We only use
1844 		 * the changelist logic to unmount when setting canmount=off.
1845 		 */
1846 		if (prop != ZFS_PROP_CANMOUNT ||
1847 		    (fnvpair_value_uint64(elem) == ZFS_CANMOUNT_OFF &&
1848 		    zfs_is_mounted(zhp, NULL))) {
1849 			cls[cl_idx] = changelist_gather(zhp, prop, 0, 0);
1850 			if (cls[cl_idx] == NULL)
1851 				goto error;
1852 		}
1853 
1854 		if (prop == ZFS_PROP_MOUNTPOINT &&
1855 		    changelist_haszonedchild(cls[cl_idx])) {
1856 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1857 			    "child dataset with inherited mountpoint is used "
1858 			    "in a non-global zone"));
1859 			ret = zfs_error(hdl, EZFS_ZONED, errbuf);
1860 			goto error;
1861 		}
1862 
1863 		if (cls[cl_idx] != NULL &&
1864 		    (ret = changelist_prefix(cls[cl_idx])) != 0)
1865 			goto error;
1866 
1867 		cl_idx++;
1868 	}
1869 	assert(cl_idx == nvl_len);
1870 
1871 	/*
1872 	 * Execute the corresponding ioctl() to set this list of properties.
1873 	 */
1874 	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
1875 
1876 	zcmd_write_src_nvlist(hdl, &zc, nvl);
1877 	zcmd_alloc_dst_nvlist(hdl, &zc, 0);
1878 
1879 	ret = zfs_ioctl(hdl, ZFS_IOC_SET_PROP, &zc);
1880 
1881 	if (ret != 0) {
1882 		if (zc.zc_nvlist_dst_filled == B_FALSE) {
1883 			(void) zfs_standard_error(hdl, errno, errbuf);
1884 			goto error;
1885 		}
1886 
1887 		/* Get the list of unset properties back and report them. */
1888 		nvlist_t *errorprops = NULL;
1889 		if (zcmd_read_dst_nvlist(hdl, &zc, &errorprops) != 0)
1890 			goto error;
1891 		for (nvpair_t *elem = nvlist_next_nvpair(errorprops, NULL);
1892 		    elem != NULL;
1893 		    elem = nvlist_next_nvpair(errorprops, elem)) {
1894 			prop = zfs_name_to_prop(nvpair_name(elem));
1895 			zfs_setprop_error(hdl, prop, errno, errbuf);
1896 		}
1897 		nvlist_free(errorprops);
1898 
1899 		if (added_resv && errno == ENOSPC) {
1900 			/* clean up the volsize property we tried to set */
1901 			uint64_t old_volsize = zfs_prop_get_int(zhp,
1902 			    ZFS_PROP_VOLSIZE);
1903 			nvlist_free(nvl);
1904 			nvl = NULL;
1905 			zcmd_free_nvlists(&zc);
1906 
1907 			if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
1908 				goto error;
1909 			if (nvlist_add_uint64(nvl,
1910 			    zfs_prop_to_name(ZFS_PROP_VOLSIZE),
1911 			    old_volsize) != 0)
1912 				goto error;
1913 			zcmd_write_src_nvlist(hdl, &zc, nvl);
1914 			(void) zfs_ioctl(hdl, ZFS_IOC_SET_PROP, &zc);
1915 		}
1916 	} else {
1917 		for (cl_idx = 0; cl_idx < nvl_len; cl_idx++) {
1918 			if (cls[cl_idx] != NULL) {
1919 				int clp_err = changelist_postfix(cls[cl_idx]);
1920 				if (clp_err != 0)
1921 					ret = clp_err;
1922 			}
1923 		}
1924 
1925 		if (ret == 0) {
1926 			/*
1927 			 * Refresh the statistics so the new property
1928 			 * value is reflected.
1929 			 */
1930 			(void) get_stats(zhp);
1931 
1932 			/*
1933 			 * Remount the filesystem to propagate the change
1934 			 * if one of the options handled by the generic
1935 			 * Linux namespace layer has been modified.
1936 			 */
1937 			if (zfs_is_namespace_prop(prop) &&
1938 			    zfs_is_mounted(zhp, NULL))
1939 				ret = zfs_mount(zhp, MNTOPT_REMOUNT, 0);
1940 		}
1941 	}
1942 
1943 error:
1944 	nvlist_free(nvl);
1945 	zcmd_free_nvlists(&zc);
1946 	if (cls != NULL) {
1947 		for (cl_idx = 0; cl_idx < nvl_len; cl_idx++) {
1948 			if (cls[cl_idx] != NULL)
1949 				changelist_free(cls[cl_idx]);
1950 		}
1951 		free(cls);
1952 	}
1953 	return (ret);
1954 }
1955 
1956 /*
1957  * Given a property, inherit the value from the parent dataset, or if received
1958  * is TRUE, revert to the received value, if any.
1959  */
1960 int
1961 zfs_prop_inherit(zfs_handle_t *zhp, const char *propname, boolean_t received)
1962 {
1963 	zfs_cmd_t zc = {"\0"};
1964 	int ret;
1965 	prop_changelist_t *cl;
1966 	libzfs_handle_t *hdl = zhp->zfs_hdl;
1967 	char errbuf[ERRBUFLEN];
1968 	zfs_prop_t prop;
1969 
1970 	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1971 	    "cannot inherit %s for '%s'"), propname, zhp->zfs_name);
1972 
1973 	zc.zc_cookie = received;
1974 	if ((prop = zfs_name_to_prop(propname)) == ZPROP_USERPROP) {
1975 		/*
1976 		 * For user properties, the amount of work we have to do is very
1977 		 * small, so just do it here.
1978 		 */
1979 		if (!zfs_prop_user(propname)) {
1980 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1981 			    "invalid property"));
1982 			return (zfs_error(hdl, EZFS_BADPROP, errbuf));
1983 		}
1984 
1985 		(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
1986 		(void) strlcpy(zc.zc_value, propname, sizeof (zc.zc_value));
1987 
1988 		if (zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_INHERIT_PROP, &zc) != 0)
1989 			return (zfs_standard_error(hdl, errno, errbuf));
1990 
1991 		(void) get_stats(zhp);
1992 		return (0);
1993 	}
1994 
1995 	/*
1996 	 * Verify that this property is inheritable.
1997 	 */
1998 	if (zfs_prop_readonly(prop))
1999 		return (zfs_error(hdl, EZFS_PROPREADONLY, errbuf));
2000 
2001 	if (!zfs_prop_inheritable(prop) && !received)
2002 		return (zfs_error(hdl, EZFS_PROPNONINHERIT, errbuf));
2003 
2004 	/*
2005 	 * Check to see if the value applies to this type
2006 	 */
2007 	if (!zfs_prop_valid_for_type(prop, zhp->zfs_type, B_FALSE))
2008 		return (zfs_error(hdl, EZFS_PROPTYPE, errbuf));
2009 
2010 	/*
2011 	 * Normalize the name, to get rid of shorthand abbreviations.
2012 	 */
2013 	propname = zfs_prop_to_name(prop);
2014 	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
2015 	(void) strlcpy(zc.zc_value, propname, sizeof (zc.zc_value));
2016 
2017 	if (prop == ZFS_PROP_MOUNTPOINT && getzoneid() == GLOBAL_ZONEID &&
2018 	    zfs_prop_get_int(zhp, ZFS_PROP_ZONED)) {
2019 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2020 		    "dataset is used in a non-global zone"));
2021 		return (zfs_error(hdl, EZFS_ZONED, errbuf));
2022 	}
2023 
2024 	/*
2025 	 * Determine datasets which will be affected by this change, if any.
2026 	 */
2027 	if ((cl = changelist_gather(zhp, prop, 0, 0)) == NULL)
2028 		return (-1);
2029 
2030 	if (prop == ZFS_PROP_MOUNTPOINT && changelist_haszonedchild(cl)) {
2031 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2032 		    "child dataset with inherited mountpoint is used "
2033 		    "in a non-global zone"));
2034 		ret = zfs_error(hdl, EZFS_ZONED, errbuf);
2035 		goto error;
2036 	}
2037 
2038 	if ((ret = changelist_prefix(cl)) != 0)
2039 		goto error;
2040 
2041 	if (zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_INHERIT_PROP, &zc) != 0) {
2042 		changelist_free(cl);
2043 		return (zfs_standard_error(hdl, errno, errbuf));
2044 	} else {
2045 
2046 		if ((ret = changelist_postfix(cl)) != 0)
2047 			goto error;
2048 
2049 		/*
2050 		 * Refresh the statistics so the new property is reflected.
2051 		 */
2052 		(void) get_stats(zhp);
2053 
2054 		/*
2055 		 * Remount the filesystem to propagate the change
2056 		 * if one of the options handled by the generic
2057 		 * Linux namespace layer has been modified.
2058 		 */
2059 		if (zfs_is_namespace_prop(prop) &&
2060 		    zfs_is_mounted(zhp, NULL))
2061 			ret = zfs_mount(zhp, MNTOPT_REMOUNT, 0);
2062 	}
2063 
2064 error:
2065 	changelist_free(cl);
2066 	return (ret);
2067 }
2068 
2069 /*
2070  * True DSL properties are stored in an nvlist.  The following two functions
2071  * extract them appropriately.
2072  */
2073 uint64_t
2074 getprop_uint64(zfs_handle_t *zhp, zfs_prop_t prop, const char **source)
2075 {
2076 	nvlist_t *nv;
2077 	uint64_t value;
2078 
2079 	*source = NULL;
2080 	if (nvlist_lookup_nvlist(zhp->zfs_props,
2081 	    zfs_prop_to_name(prop), &nv) == 0) {
2082 		value = fnvlist_lookup_uint64(nv, ZPROP_VALUE);
2083 		(void) nvlist_lookup_string(nv, ZPROP_SOURCE, source);
2084 	} else {
2085 		verify(!zhp->zfs_props_table ||
2086 		    zhp->zfs_props_table[prop] == B_TRUE);
2087 		value = zfs_prop_default_numeric(prop);
2088 		*source = "";
2089 	}
2090 
2091 	return (value);
2092 }
2093 
2094 static const char *
2095 getprop_string(zfs_handle_t *zhp, zfs_prop_t prop, const char **source)
2096 {
2097 	nvlist_t *nv;
2098 	const char *value;
2099 
2100 	*source = NULL;
2101 	if (nvlist_lookup_nvlist(zhp->zfs_props,
2102 	    zfs_prop_to_name(prop), &nv) == 0) {
2103 		value = fnvlist_lookup_string(nv, ZPROP_VALUE);
2104 		(void) nvlist_lookup_string(nv, ZPROP_SOURCE, source);
2105 	} else {
2106 		verify(!zhp->zfs_props_table ||
2107 		    zhp->zfs_props_table[prop] == B_TRUE);
2108 		value = zfs_prop_default_string(prop);
2109 		*source = "";
2110 	}
2111 
2112 	return (value);
2113 }
2114 
2115 static boolean_t
2116 zfs_is_recvd_props_mode(zfs_handle_t *zhp)
2117 {
2118 	return (zhp->zfs_props != NULL &&
2119 	    zhp->zfs_props == zhp->zfs_recvd_props);
2120 }
2121 
2122 static void
2123 zfs_set_recvd_props_mode(zfs_handle_t *zhp, uintptr_t *cookie)
2124 {
2125 	*cookie = (uintptr_t)zhp->zfs_props;
2126 	zhp->zfs_props = zhp->zfs_recvd_props;
2127 }
2128 
2129 static void
2130 zfs_unset_recvd_props_mode(zfs_handle_t *zhp, uintptr_t *cookie)
2131 {
2132 	zhp->zfs_props = (nvlist_t *)*cookie;
2133 	*cookie = 0;
2134 }
2135 
2136 /*
2137  * Internal function for getting a numeric property.  Both zfs_prop_get() and
2138  * zfs_prop_get_int() are built using this interface.
2139  *
2140  * Certain properties can be overridden using 'mount -o'.  In this case, scan
2141  * the contents of the /proc/self/mounts entry, searching for the
2142  * appropriate options. If they differ from the on-disk values, report the
2143  * current values and mark the source "temporary".
2144  */
2145 static int
2146 get_numeric_property(zfs_handle_t *zhp, zfs_prop_t prop, zprop_source_t *src,
2147     const char **source, uint64_t *val)
2148 {
2149 	zfs_cmd_t zc = {"\0"};
2150 	nvlist_t *zplprops = NULL;
2151 	struct mnttab mnt;
2152 	const char *mntopt_on = NULL;
2153 	const char *mntopt_off = NULL;
2154 	boolean_t received = zfs_is_recvd_props_mode(zhp);
2155 
2156 	*source = NULL;
2157 
2158 	/*
2159 	 * If the property is being fetched for a snapshot, check whether
2160 	 * the property is valid for the snapshot's head dataset type.
2161 	 */
2162 	if (zhp->zfs_type == ZFS_TYPE_SNAPSHOT &&
2163 	    !zfs_prop_valid_for_type(prop, zhp->zfs_head_type, B_TRUE)) {
2164 		*val = zfs_prop_default_numeric(prop);
2165 		return (-1);
2166 	}
2167 
2168 	switch (prop) {
2169 	case ZFS_PROP_ATIME:
2170 		mntopt_on = MNTOPT_ATIME;
2171 		mntopt_off = MNTOPT_NOATIME;
2172 		break;
2173 
2174 	case ZFS_PROP_RELATIME:
2175 		mntopt_on = MNTOPT_RELATIME;
2176 		mntopt_off = MNTOPT_NORELATIME;
2177 		break;
2178 
2179 	case ZFS_PROP_DEVICES:
2180 		mntopt_on = MNTOPT_DEVICES;
2181 		mntopt_off = MNTOPT_NODEVICES;
2182 		break;
2183 
2184 	case ZFS_PROP_EXEC:
2185 		mntopt_on = MNTOPT_EXEC;
2186 		mntopt_off = MNTOPT_NOEXEC;
2187 		break;
2188 
2189 	case ZFS_PROP_READONLY:
2190 		mntopt_on = MNTOPT_RO;
2191 		mntopt_off = MNTOPT_RW;
2192 		break;
2193 
2194 	case ZFS_PROP_SETUID:
2195 		mntopt_on = MNTOPT_SETUID;
2196 		mntopt_off = MNTOPT_NOSETUID;
2197 		break;
2198 
2199 	case ZFS_PROP_XATTR:
2200 		mntopt_on = MNTOPT_XATTR;
2201 		mntopt_off = MNTOPT_NOXATTR;
2202 		break;
2203 
2204 	case ZFS_PROP_NBMAND:
2205 		mntopt_on = MNTOPT_NBMAND;
2206 		mntopt_off = MNTOPT_NONBMAND;
2207 		break;
2208 
2209 	default:
2210 		break;
2211 	}
2212 
2213 	/*
2214 	 * Because looking up the mount options is potentially expensive
2215 	 * (iterating over all of /proc/self/mounts), we defer its
2216 	 * calculation until we're looking up a property which requires
2217 	 * its presence.
2218 	 */
2219 	if (!zhp->zfs_mntcheck &&
2220 	    (mntopt_on != NULL || prop == ZFS_PROP_MOUNTED)) {
2221 		libzfs_handle_t *hdl = zhp->zfs_hdl;
2222 		struct mnttab entry;
2223 
2224 		if (libzfs_mnttab_find(hdl, zhp->zfs_name, &entry) == 0)
2225 			zhp->zfs_mntopts = zfs_strdup(hdl,
2226 			    entry.mnt_mntopts);
2227 
2228 		zhp->zfs_mntcheck = B_TRUE;
2229 	}
2230 
2231 	if (zhp->zfs_mntopts == NULL)
2232 		mnt.mnt_mntopts = (char *)"";
2233 	else
2234 		mnt.mnt_mntopts = zhp->zfs_mntopts;
2235 
2236 	switch (prop) {
2237 	case ZFS_PROP_ATIME:
2238 	case ZFS_PROP_RELATIME:
2239 	case ZFS_PROP_DEVICES:
2240 	case ZFS_PROP_EXEC:
2241 	case ZFS_PROP_READONLY:
2242 	case ZFS_PROP_SETUID:
2243 #ifndef __FreeBSD__
2244 	case ZFS_PROP_XATTR:
2245 #endif
2246 	case ZFS_PROP_NBMAND:
2247 		*val = getprop_uint64(zhp, prop, source);
2248 
2249 		if (received)
2250 			break;
2251 
2252 		if (hasmntopt(&mnt, mntopt_on) && !*val) {
2253 			*val = B_TRUE;
2254 			if (src)
2255 				*src = ZPROP_SRC_TEMPORARY;
2256 		} else if (hasmntopt(&mnt, mntopt_off) && *val) {
2257 			*val = B_FALSE;
2258 			if (src)
2259 				*src = ZPROP_SRC_TEMPORARY;
2260 		}
2261 		break;
2262 
2263 	case ZFS_PROP_CANMOUNT:
2264 	case ZFS_PROP_VOLSIZE:
2265 	case ZFS_PROP_QUOTA:
2266 	case ZFS_PROP_REFQUOTA:
2267 	case ZFS_PROP_RESERVATION:
2268 	case ZFS_PROP_REFRESERVATION:
2269 	case ZFS_PROP_FILESYSTEM_LIMIT:
2270 	case ZFS_PROP_SNAPSHOT_LIMIT:
2271 	case ZFS_PROP_FILESYSTEM_COUNT:
2272 	case ZFS_PROP_SNAPSHOT_COUNT:
2273 		*val = getprop_uint64(zhp, prop, source);
2274 
2275 		if (*source == NULL) {
2276 			/* not default, must be local */
2277 			*source = zhp->zfs_name;
2278 		}
2279 		break;
2280 
2281 	case ZFS_PROP_MOUNTED:
2282 		*val = (zhp->zfs_mntopts != NULL);
2283 		break;
2284 
2285 	case ZFS_PROP_NUMCLONES:
2286 		*val = zhp->zfs_dmustats.dds_num_clones;
2287 		break;
2288 
2289 	case ZFS_PROP_VERSION:
2290 	case ZFS_PROP_NORMALIZE:
2291 	case ZFS_PROP_UTF8ONLY:
2292 	case ZFS_PROP_CASE:
2293 		zcmd_alloc_dst_nvlist(zhp->zfs_hdl, &zc, 0);
2294 
2295 		(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
2296 		if (zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_OBJSET_ZPLPROPS, &zc)) {
2297 			zcmd_free_nvlists(&zc);
2298 			if (prop == ZFS_PROP_VERSION &&
2299 			    zhp->zfs_type == ZFS_TYPE_VOLUME)
2300 				*val = zfs_prop_default_numeric(prop);
2301 			return (-1);
2302 		}
2303 		if (zcmd_read_dst_nvlist(zhp->zfs_hdl, &zc, &zplprops) != 0 ||
2304 		    nvlist_lookup_uint64(zplprops, zfs_prop_to_name(prop),
2305 		    val) != 0) {
2306 			zcmd_free_nvlists(&zc);
2307 			return (-1);
2308 		}
2309 		nvlist_free(zplprops);
2310 		zcmd_free_nvlists(&zc);
2311 		break;
2312 
2313 	case ZFS_PROP_INCONSISTENT:
2314 		*val = zhp->zfs_dmustats.dds_inconsistent;
2315 		break;
2316 
2317 	case ZFS_PROP_REDACTED:
2318 		*val = zhp->zfs_dmustats.dds_redacted;
2319 		break;
2320 
2321 	case ZFS_PROP_GUID:
2322 		if (zhp->zfs_dmustats.dds_guid != 0)
2323 			*val = zhp->zfs_dmustats.dds_guid;
2324 		else
2325 			*val = getprop_uint64(zhp, prop, source);
2326 		break;
2327 
2328 	case ZFS_PROP_CREATETXG:
2329 		/*
2330 		 * We can directly read createtxg property from zfs
2331 		 * handle for Filesystem, Snapshot and ZVOL types.
2332 		 */
2333 		if (((zhp->zfs_type == ZFS_TYPE_FILESYSTEM) ||
2334 		    (zhp->zfs_type == ZFS_TYPE_SNAPSHOT) ||
2335 		    (zhp->zfs_type == ZFS_TYPE_VOLUME)) &&
2336 		    (zhp->zfs_dmustats.dds_creation_txg != 0)) {
2337 			*val = zhp->zfs_dmustats.dds_creation_txg;
2338 			break;
2339 		} else {
2340 			*val = getprop_uint64(zhp, prop, source);
2341 		}
2342 		zfs_fallthrough;
2343 	default:
2344 		switch (zfs_prop_get_type(prop)) {
2345 		case PROP_TYPE_NUMBER:
2346 		case PROP_TYPE_INDEX:
2347 			*val = getprop_uint64(zhp, prop, source);
2348 			/*
2349 			 * If we tried to use a default value for a
2350 			 * readonly property, it means that it was not
2351 			 * present.  Note this only applies to "truly"
2352 			 * readonly properties, not set-once properties
2353 			 * like volblocksize.
2354 			 */
2355 			if (zfs_prop_readonly(prop) &&
2356 			    !zfs_prop_setonce(prop) &&
2357 			    *source != NULL && (*source)[0] == '\0') {
2358 				*source = NULL;
2359 				return (-1);
2360 			}
2361 			break;
2362 
2363 		case PROP_TYPE_STRING:
2364 		default:
2365 			zfs_error_aux(zhp->zfs_hdl, dgettext(TEXT_DOMAIN,
2366 			    "cannot get non-numeric property"));
2367 			return (zfs_error(zhp->zfs_hdl, EZFS_BADPROP,
2368 			    dgettext(TEXT_DOMAIN, "internal error")));
2369 		}
2370 	}
2371 
2372 	return (0);
2373 }
2374 
2375 /*
2376  * Calculate the source type, given the raw source string.
2377  */
2378 static void
2379 get_source(zfs_handle_t *zhp, zprop_source_t *srctype, const char *source,
2380     char *statbuf, size_t statlen)
2381 {
2382 	if (statbuf == NULL ||
2383 	    srctype == NULL || *srctype == ZPROP_SRC_TEMPORARY) {
2384 		return;
2385 	}
2386 
2387 	if (source == NULL) {
2388 		*srctype = ZPROP_SRC_NONE;
2389 	} else if (source[0] == '\0') {
2390 		*srctype = ZPROP_SRC_DEFAULT;
2391 	} else if (strstr(source, ZPROP_SOURCE_VAL_RECVD) != NULL) {
2392 		*srctype = ZPROP_SRC_RECEIVED;
2393 	} else {
2394 		if (strcmp(source, zhp->zfs_name) == 0) {
2395 			*srctype = ZPROP_SRC_LOCAL;
2396 		} else {
2397 			(void) strlcpy(statbuf, source, statlen);
2398 			*srctype = ZPROP_SRC_INHERITED;
2399 		}
2400 	}
2401 
2402 }
2403 
2404 int
2405 zfs_prop_get_recvd(zfs_handle_t *zhp, const char *propname, char *propbuf,
2406     size_t proplen, boolean_t literal)
2407 {
2408 	zfs_prop_t prop;
2409 	int err = 0;
2410 
2411 	if (zhp->zfs_recvd_props == NULL)
2412 		if (get_recvd_props_ioctl(zhp) != 0)
2413 			return (-1);
2414 
2415 	prop = zfs_name_to_prop(propname);
2416 
2417 	if (prop != ZPROP_USERPROP) {
2418 		uintptr_t cookie;
2419 		if (!nvlist_exists(zhp->zfs_recvd_props, propname))
2420 			return (-1);
2421 		zfs_set_recvd_props_mode(zhp, &cookie);
2422 		err = zfs_prop_get(zhp, prop, propbuf, proplen,
2423 		    NULL, NULL, 0, literal);
2424 		zfs_unset_recvd_props_mode(zhp, &cookie);
2425 	} else {
2426 		nvlist_t *propval;
2427 		const char *recvdval;
2428 		if (nvlist_lookup_nvlist(zhp->zfs_recvd_props,
2429 		    propname, &propval) != 0)
2430 			return (-1);
2431 		recvdval = fnvlist_lookup_string(propval, ZPROP_VALUE);
2432 		(void) strlcpy(propbuf, recvdval, proplen);
2433 	}
2434 
2435 	return (err == 0 ? 0 : -1);
2436 }
2437 
2438 static int
2439 get_clones_string(zfs_handle_t *zhp, char *propbuf, size_t proplen)
2440 {
2441 	nvlist_t *value;
2442 	nvpair_t *pair;
2443 
2444 	value = zfs_get_clones_nvl(zhp);
2445 	if (value == NULL || nvlist_empty(value))
2446 		return (-1);
2447 
2448 	propbuf[0] = '\0';
2449 	for (pair = nvlist_next_nvpair(value, NULL); pair != NULL;
2450 	    pair = nvlist_next_nvpair(value, pair)) {
2451 		if (propbuf[0] != '\0')
2452 			(void) strlcat(propbuf, ",", proplen);
2453 		(void) strlcat(propbuf, nvpair_name(pair), proplen);
2454 	}
2455 
2456 	return (0);
2457 }
2458 
2459 struct get_clones_arg {
2460 	uint64_t numclones;
2461 	nvlist_t *value;
2462 	const char *origin;
2463 	char buf[ZFS_MAX_DATASET_NAME_LEN];
2464 };
2465 
2466 static int
2467 get_clones_cb(zfs_handle_t *zhp, void *arg)
2468 {
2469 	struct get_clones_arg *gca = arg;
2470 
2471 	if (gca->numclones == 0) {
2472 		zfs_close(zhp);
2473 		return (0);
2474 	}
2475 
2476 	if (zfs_prop_get(zhp, ZFS_PROP_ORIGIN, gca->buf, sizeof (gca->buf),
2477 	    NULL, NULL, 0, B_TRUE) != 0)
2478 		goto out;
2479 	if (strcmp(gca->buf, gca->origin) == 0) {
2480 		fnvlist_add_boolean(gca->value, zfs_get_name(zhp));
2481 		gca->numclones--;
2482 	}
2483 
2484 out:
2485 	(void) zfs_iter_children_v2(zhp, 0, get_clones_cb, gca);
2486 	zfs_close(zhp);
2487 	return (0);
2488 }
2489 
2490 nvlist_t *
2491 zfs_get_clones_nvl(zfs_handle_t *zhp)
2492 {
2493 	nvlist_t *nv, *value;
2494 
2495 	if (nvlist_lookup_nvlist(zhp->zfs_props,
2496 	    zfs_prop_to_name(ZFS_PROP_CLONES), &nv) != 0) {
2497 		struct get_clones_arg gca;
2498 
2499 		/*
2500 		 * if this is a snapshot, then the kernel wasn't able
2501 		 * to get the clones.  Do it by slowly iterating.
2502 		 */
2503 		if (zhp->zfs_type != ZFS_TYPE_SNAPSHOT)
2504 			return (NULL);
2505 		if (nvlist_alloc(&nv, NV_UNIQUE_NAME, 0) != 0)
2506 			return (NULL);
2507 		if (nvlist_alloc(&value, NV_UNIQUE_NAME, 0) != 0) {
2508 			nvlist_free(nv);
2509 			return (NULL);
2510 		}
2511 
2512 		gca.numclones = zfs_prop_get_int(zhp, ZFS_PROP_NUMCLONES);
2513 		gca.value = value;
2514 		gca.origin = zhp->zfs_name;
2515 
2516 		if (gca.numclones != 0) {
2517 			zfs_handle_t *root;
2518 			char pool[ZFS_MAX_DATASET_NAME_LEN];
2519 			char *cp = pool;
2520 
2521 			/* get the pool name */
2522 			(void) strlcpy(pool, zhp->zfs_name, sizeof (pool));
2523 			(void) strsep(&cp, "/@");
2524 			root = zfs_open(zhp->zfs_hdl, pool,
2525 			    ZFS_TYPE_FILESYSTEM);
2526 			if (root == NULL) {
2527 				nvlist_free(nv);
2528 				nvlist_free(value);
2529 				return (NULL);
2530 			}
2531 
2532 			(void) get_clones_cb(root, &gca);
2533 		}
2534 
2535 		if (gca.numclones != 0 ||
2536 		    nvlist_add_nvlist(nv, ZPROP_VALUE, value) != 0 ||
2537 		    nvlist_add_nvlist(zhp->zfs_props,
2538 		    zfs_prop_to_name(ZFS_PROP_CLONES), nv) != 0) {
2539 			nvlist_free(nv);
2540 			nvlist_free(value);
2541 			return (NULL);
2542 		}
2543 		nvlist_free(nv);
2544 		nvlist_free(value);
2545 		nv = fnvlist_lookup_nvlist(zhp->zfs_props,
2546 		    zfs_prop_to_name(ZFS_PROP_CLONES));
2547 	}
2548 
2549 	return (fnvlist_lookup_nvlist(nv, ZPROP_VALUE));
2550 }
2551 
2552 static int
2553 get_rsnaps_string(zfs_handle_t *zhp, char *propbuf, size_t proplen)
2554 {
2555 	nvlist_t *value;
2556 	uint64_t *snaps;
2557 	uint_t nsnaps;
2558 
2559 	if (nvlist_lookup_nvlist(zhp->zfs_props,
2560 	    zfs_prop_to_name(ZFS_PROP_REDACT_SNAPS), &value) != 0)
2561 		return (-1);
2562 	if (nvlist_lookup_uint64_array(value, ZPROP_VALUE, &snaps,
2563 	    &nsnaps) != 0)
2564 		return (-1);
2565 	if (nsnaps == 0) {
2566 		/* There's no redaction snapshots; pass a special value back */
2567 		(void) snprintf(propbuf, proplen, "none");
2568 		return (0);
2569 	}
2570 	propbuf[0] = '\0';
2571 	for (int i = 0; i < nsnaps; i++) {
2572 		char buf[128];
2573 		if (propbuf[0] != '\0')
2574 			(void) strlcat(propbuf, ",", proplen);
2575 		(void) snprintf(buf, sizeof (buf), "%llu",
2576 		    (u_longlong_t)snaps[i]);
2577 		(void) strlcat(propbuf, buf, proplen);
2578 	}
2579 
2580 	return (0);
2581 }
2582 
2583 /*
2584  * Accepts a property and value and checks that the value
2585  * matches the one found by the channel program. If they are
2586  * not equal, print both of them.
2587  */
2588 static void
2589 zcp_check(zfs_handle_t *zhp, zfs_prop_t prop, uint64_t intval,
2590     const char *strval)
2591 {
2592 	if (!zhp->zfs_hdl->libzfs_prop_debug)
2593 		return;
2594 	int error;
2595 	char *poolname = zhp->zpool_hdl->zpool_name;
2596 	const char *prop_name = zfs_prop_to_name(prop);
2597 	const char *program =
2598 	    "args = ...\n"
2599 	    "ds = args['dataset']\n"
2600 	    "prop = args['property']\n"
2601 	    "value, setpoint = zfs.get_prop(ds, prop)\n"
2602 	    "return {value=value, setpoint=setpoint}\n";
2603 	nvlist_t *outnvl;
2604 	nvlist_t *retnvl;
2605 	nvlist_t *argnvl = fnvlist_alloc();
2606 
2607 	fnvlist_add_string(argnvl, "dataset", zhp->zfs_name);
2608 	fnvlist_add_string(argnvl, "property", zfs_prop_to_name(prop));
2609 
2610 	error = lzc_channel_program_nosync(poolname, program,
2611 	    10 * 1000 * 1000, 10 * 1024 * 1024, argnvl, &outnvl);
2612 
2613 	if (error == 0) {
2614 		retnvl = fnvlist_lookup_nvlist(outnvl, "return");
2615 		if (zfs_prop_get_type(prop) == PROP_TYPE_NUMBER) {
2616 			int64_t ans;
2617 			error = nvlist_lookup_int64(retnvl, "value", &ans);
2618 			if (error != 0) {
2619 				(void) fprintf(stderr, "%s: zcp check error: "
2620 				    "%u\n", prop_name, error);
2621 				return;
2622 			}
2623 			if (ans != intval) {
2624 				(void) fprintf(stderr, "%s: zfs found %llu, "
2625 				    "but zcp found %llu\n", prop_name,
2626 				    (u_longlong_t)intval, (u_longlong_t)ans);
2627 			}
2628 		} else {
2629 			const char *str_ans;
2630 			error = nvlist_lookup_string(retnvl, "value", &str_ans);
2631 			if (error != 0) {
2632 				(void) fprintf(stderr, "%s: zcp check error: "
2633 				    "%u\n", prop_name, error);
2634 				return;
2635 			}
2636 			if (strcmp(strval, str_ans) != 0) {
2637 				(void) fprintf(stderr,
2638 				    "%s: zfs found '%s', but zcp found '%s'\n",
2639 				    prop_name, strval, str_ans);
2640 			}
2641 		}
2642 	} else {
2643 		(void) fprintf(stderr, "%s: zcp check failed, channel program "
2644 		    "error: %u\n", prop_name, error);
2645 	}
2646 	nvlist_free(argnvl);
2647 	nvlist_free(outnvl);
2648 }
2649 
2650 /*
2651  * Retrieve a property from the given object.  If 'literal' is specified, then
2652  * numbers are left as exact values.  Otherwise, numbers are converted to a
2653  * human-readable form.
2654  *
2655  * Returns 0 on success, or -1 on error.
2656  */
2657 int
2658 zfs_prop_get(zfs_handle_t *zhp, zfs_prop_t prop, char *propbuf, size_t proplen,
2659     zprop_source_t *src, char *statbuf, size_t statlen, boolean_t literal)
2660 {
2661 	const char *source = NULL;
2662 	uint64_t val;
2663 	const char *str;
2664 	const char *strval;
2665 	boolean_t received = zfs_is_recvd_props_mode(zhp);
2666 
2667 	/*
2668 	 * Check to see if this property applies to our object
2669 	 */
2670 	if (!zfs_prop_valid_for_type(prop, zhp->zfs_type, B_FALSE))
2671 		return (-1);
2672 
2673 	if (received && zfs_prop_readonly(prop))
2674 		return (-1);
2675 
2676 	if (src)
2677 		*src = ZPROP_SRC_NONE;
2678 
2679 	switch (prop) {
2680 	case ZFS_PROP_CREATION:
2681 		/*
2682 		 * 'creation' is a time_t stored in the statistics.  We convert
2683 		 * this into a string unless 'literal' is specified.
2684 		 */
2685 		{
2686 			val = getprop_uint64(zhp, prop, &source);
2687 			time_t time = (time_t)val;
2688 			struct tm t;
2689 
2690 			if (literal ||
2691 			    localtime_r(&time, &t) == NULL ||
2692 			    strftime(propbuf, proplen, "%a %b %e %k:%M %Y",
2693 			    &t) == 0)
2694 				(void) snprintf(propbuf, proplen, "%llu",
2695 				    (u_longlong_t)val);
2696 		}
2697 		zcp_check(zhp, prop, val, NULL);
2698 		break;
2699 
2700 	case ZFS_PROP_MOUNTPOINT:
2701 		/*
2702 		 * Getting the precise mountpoint can be tricky.
2703 		 *
2704 		 *  - for 'none' or 'legacy', return those values.
2705 		 *  - for inherited mountpoints, we want to take everything
2706 		 *    after our ancestor and append it to the inherited value.
2707 		 *
2708 		 * If the pool has an alternate root, we want to prepend that
2709 		 * root to any values we return.
2710 		 */
2711 
2712 		str = getprop_string(zhp, prop, &source);
2713 
2714 		if (str[0] == '/') {
2715 			char buf[MAXPATHLEN];
2716 			char *root = buf;
2717 			const char *relpath;
2718 
2719 			/*
2720 			 * If we inherit the mountpoint, even from a dataset
2721 			 * with a received value, the source will be the path of
2722 			 * the dataset we inherit from. If source is
2723 			 * ZPROP_SOURCE_VAL_RECVD, the received value is not
2724 			 * inherited.
2725 			 */
2726 			if (strcmp(source, ZPROP_SOURCE_VAL_RECVD) == 0) {
2727 				relpath = "";
2728 			} else {
2729 				relpath = zhp->zfs_name + strlen(source);
2730 				if (relpath[0] == '/')
2731 					relpath++;
2732 			}
2733 
2734 			if ((zpool_get_prop(zhp->zpool_hdl,
2735 			    ZPOOL_PROP_ALTROOT, buf, MAXPATHLEN, NULL,
2736 			    B_FALSE)) || (strcmp(root, "-") == 0))
2737 				root[0] = '\0';
2738 			/*
2739 			 * Special case an alternate root of '/'. This will
2740 			 * avoid having multiple leading slashes in the
2741 			 * mountpoint path.
2742 			 */
2743 			if (strcmp(root, "/") == 0)
2744 				root++;
2745 
2746 			/*
2747 			 * If the mountpoint is '/' then skip over this
2748 			 * if we are obtaining either an alternate root or
2749 			 * an inherited mountpoint.
2750 			 */
2751 			if (str[1] == '\0' && (root[0] != '\0' ||
2752 			    relpath[0] != '\0'))
2753 				str++;
2754 
2755 			if (relpath[0] == '\0')
2756 				(void) snprintf(propbuf, proplen, "%s%s",
2757 				    root, str);
2758 			else
2759 				(void) snprintf(propbuf, proplen, "%s%s%s%s",
2760 				    root, str, relpath[0] == '@' ? "" : "/",
2761 				    relpath);
2762 		} else {
2763 			/* 'legacy' or 'none' */
2764 			(void) strlcpy(propbuf, str, proplen);
2765 		}
2766 		zcp_check(zhp, prop, 0, propbuf);
2767 		break;
2768 
2769 	case ZFS_PROP_ORIGIN:
2770 		if (*zhp->zfs_dmustats.dds_origin != '\0') {
2771 			str = (char *)&zhp->zfs_dmustats.dds_origin;
2772 		} else {
2773 			str = getprop_string(zhp, prop, &source);
2774 		}
2775 		if (str == NULL || *str == '\0')
2776 			str = zfs_prop_default_string(prop);
2777 		if (str == NULL)
2778 			return (-1);
2779 		(void) strlcpy(propbuf, str, proplen);
2780 		zcp_check(zhp, prop, 0, str);
2781 		break;
2782 
2783 	case ZFS_PROP_REDACT_SNAPS:
2784 		if (get_rsnaps_string(zhp, propbuf, proplen) != 0)
2785 			return (-1);
2786 		break;
2787 
2788 	case ZFS_PROP_CLONES:
2789 		if (get_clones_string(zhp, propbuf, proplen) != 0)
2790 			return (-1);
2791 		break;
2792 
2793 	case ZFS_PROP_QUOTA:
2794 	case ZFS_PROP_REFQUOTA:
2795 	case ZFS_PROP_RESERVATION:
2796 	case ZFS_PROP_REFRESERVATION:
2797 
2798 		if (get_numeric_property(zhp, prop, src, &source, &val) != 0)
2799 			return (-1);
2800 		/*
2801 		 * If quota or reservation is 0, we translate this into 'none'
2802 		 * (unless literal is set), and indicate that it's the default
2803 		 * value.  Otherwise, we print the number nicely and indicate
2804 		 * that its set locally.
2805 		 */
2806 		if (val == 0) {
2807 			if (literal)
2808 				(void) strlcpy(propbuf, "0", proplen);
2809 			else
2810 				(void) strlcpy(propbuf, "none", proplen);
2811 		} else {
2812 			if (literal)
2813 				(void) snprintf(propbuf, proplen, "%llu",
2814 				    (u_longlong_t)val);
2815 			else
2816 				zfs_nicebytes(val, propbuf, proplen);
2817 		}
2818 		zcp_check(zhp, prop, val, NULL);
2819 		break;
2820 
2821 	case ZFS_PROP_FILESYSTEM_LIMIT:
2822 	case ZFS_PROP_SNAPSHOT_LIMIT:
2823 	case ZFS_PROP_FILESYSTEM_COUNT:
2824 	case ZFS_PROP_SNAPSHOT_COUNT:
2825 
2826 		if (get_numeric_property(zhp, prop, src, &source, &val) != 0)
2827 			return (-1);
2828 
2829 		/*
2830 		 * If limit is UINT64_MAX, we translate this into 'none', and
2831 		 * indicate that it's the default value. Otherwise, we print
2832 		 * the number nicely and indicate that it's set locally.
2833 		 */
2834 		if (val == UINT64_MAX) {
2835 			(void) strlcpy(propbuf, "none", proplen);
2836 		} else if (literal) {
2837 			(void) snprintf(propbuf, proplen, "%llu",
2838 			    (u_longlong_t)val);
2839 		} else {
2840 			zfs_nicenum(val, propbuf, proplen);
2841 		}
2842 
2843 		zcp_check(zhp, prop, val, NULL);
2844 		break;
2845 
2846 	case ZFS_PROP_REFRATIO:
2847 	case ZFS_PROP_COMPRESSRATIO:
2848 		if (get_numeric_property(zhp, prop, src, &source, &val) != 0)
2849 			return (-1);
2850 		if (literal)
2851 			(void) snprintf(propbuf, proplen, "%llu.%02llu",
2852 			    (u_longlong_t)(val / 100),
2853 			    (u_longlong_t)(val % 100));
2854 		else
2855 			(void) snprintf(propbuf, proplen, "%llu.%02llux",
2856 			    (u_longlong_t)(val / 100),
2857 			    (u_longlong_t)(val % 100));
2858 		zcp_check(zhp, prop, val, NULL);
2859 		break;
2860 
2861 	case ZFS_PROP_TYPE:
2862 		switch (zhp->zfs_type) {
2863 		case ZFS_TYPE_FILESYSTEM:
2864 			str = "filesystem";
2865 			break;
2866 		case ZFS_TYPE_VOLUME:
2867 			str = "volume";
2868 			break;
2869 		case ZFS_TYPE_SNAPSHOT:
2870 			str = "snapshot";
2871 			break;
2872 		case ZFS_TYPE_BOOKMARK:
2873 			str = "bookmark";
2874 			break;
2875 		default:
2876 			abort();
2877 		}
2878 		(void) snprintf(propbuf, proplen, "%s", str);
2879 		zcp_check(zhp, prop, 0, propbuf);
2880 		break;
2881 
2882 	case ZFS_PROP_MOUNTED:
2883 		/*
2884 		 * The 'mounted' property is a pseudo-property that described
2885 		 * whether the filesystem is currently mounted.  Even though
2886 		 * it's a boolean value, the typical values of "on" and "off"
2887 		 * don't make sense, so we translate to "yes" and "no".
2888 		 */
2889 		if (get_numeric_property(zhp, ZFS_PROP_MOUNTED,
2890 		    src, &source, &val) != 0)
2891 			return (-1);
2892 		if (val)
2893 			(void) strlcpy(propbuf, "yes", proplen);
2894 		else
2895 			(void) strlcpy(propbuf, "no", proplen);
2896 		break;
2897 
2898 	case ZFS_PROP_NAME:
2899 		/*
2900 		 * The 'name' property is a pseudo-property derived from the
2901 		 * dataset name.  It is presented as a real property to simplify
2902 		 * consumers.
2903 		 */
2904 		(void) strlcpy(propbuf, zhp->zfs_name, proplen);
2905 		zcp_check(zhp, prop, 0, propbuf);
2906 		break;
2907 
2908 	case ZFS_PROP_MLSLABEL:
2909 		{
2910 #ifdef HAVE_MLSLABEL
2911 			m_label_t *new_sl = NULL;
2912 			char *ascii = NULL;	/* human readable label */
2913 
2914 			(void) strlcpy(propbuf,
2915 			    getprop_string(zhp, prop, &source), proplen);
2916 
2917 			if (literal || (strcasecmp(propbuf,
2918 			    ZFS_MLSLABEL_DEFAULT) == 0))
2919 				break;
2920 
2921 			/*
2922 			 * Try to translate the internal hex string to
2923 			 * human-readable output.  If there are any
2924 			 * problems just use the hex string.
2925 			 */
2926 
2927 			if (str_to_label(propbuf, &new_sl, MAC_LABEL,
2928 			    L_NO_CORRECTION, NULL) == -1) {
2929 				m_label_free(new_sl);
2930 				break;
2931 			}
2932 
2933 			if (label_to_str(new_sl, &ascii, M_LABEL,
2934 			    DEF_NAMES) != 0) {
2935 				if (ascii)
2936 					free(ascii);
2937 				m_label_free(new_sl);
2938 				break;
2939 			}
2940 			m_label_free(new_sl);
2941 
2942 			(void) strlcpy(propbuf, ascii, proplen);
2943 			free(ascii);
2944 #else
2945 			(void) strlcpy(propbuf,
2946 			    getprop_string(zhp, prop, &source), proplen);
2947 #endif /* HAVE_MLSLABEL */
2948 		}
2949 		break;
2950 
2951 	case ZFS_PROP_GUID:
2952 	case ZFS_PROP_KEY_GUID:
2953 	case ZFS_PROP_IVSET_GUID:
2954 	case ZFS_PROP_CREATETXG:
2955 	case ZFS_PROP_OBJSETID:
2956 	case ZFS_PROP_PBKDF2_ITERS:
2957 		/*
2958 		 * These properties are stored as numbers, but they are
2959 		 * identifiers or counters.
2960 		 * We don't want them to be pretty printed, because pretty
2961 		 * printing truncates their values making them useless.
2962 		 */
2963 		if (get_numeric_property(zhp, prop, src, &source, &val) != 0)
2964 			return (-1);
2965 		(void) snprintf(propbuf, proplen, "%llu", (u_longlong_t)val);
2966 		zcp_check(zhp, prop, val, NULL);
2967 		break;
2968 
2969 	case ZFS_PROP_REFERENCED:
2970 	case ZFS_PROP_AVAILABLE:
2971 	case ZFS_PROP_USED:
2972 	case ZFS_PROP_USEDSNAP:
2973 	case ZFS_PROP_USEDDS:
2974 	case ZFS_PROP_USEDREFRESERV:
2975 	case ZFS_PROP_USEDCHILD:
2976 		if (get_numeric_property(zhp, prop, src, &source, &val) != 0)
2977 			return (-1);
2978 		if (literal) {
2979 			(void) snprintf(propbuf, proplen, "%llu",
2980 			    (u_longlong_t)val);
2981 		} else {
2982 			zfs_nicebytes(val, propbuf, proplen);
2983 		}
2984 		zcp_check(zhp, prop, val, NULL);
2985 		break;
2986 
2987 	case ZFS_PROP_SNAPSHOTS_CHANGED:
2988 		{
2989 			if ((get_numeric_property(zhp, prop, src, &source,
2990 			    &val) != 0) || val == 0) {
2991 				return (-1);
2992 			}
2993 
2994 			time_t time = (time_t)val;
2995 			struct tm t;
2996 
2997 			if (literal ||
2998 			    localtime_r(&time, &t) == NULL ||
2999 			    strftime(propbuf, proplen, "%a %b %e %k:%M:%S %Y",
3000 			    &t) == 0)
3001 				(void) snprintf(propbuf, proplen, "%llu",
3002 				    (u_longlong_t)val);
3003 		}
3004 		zcp_check(zhp, prop, val, NULL);
3005 		break;
3006 
3007 	default:
3008 		switch (zfs_prop_get_type(prop)) {
3009 		case PROP_TYPE_NUMBER:
3010 			if (get_numeric_property(zhp, prop, src,
3011 			    &source, &val) != 0) {
3012 				return (-1);
3013 			}
3014 
3015 			if (literal) {
3016 				(void) snprintf(propbuf, proplen, "%llu",
3017 				    (u_longlong_t)val);
3018 			} else {
3019 				zfs_nicenum(val, propbuf, proplen);
3020 			}
3021 			zcp_check(zhp, prop, val, NULL);
3022 			break;
3023 
3024 		case PROP_TYPE_STRING:
3025 			str = getprop_string(zhp, prop, &source);
3026 			if (str == NULL)
3027 				return (-1);
3028 
3029 			(void) strlcpy(propbuf, str, proplen);
3030 			zcp_check(zhp, prop, 0, str);
3031 			break;
3032 
3033 		case PROP_TYPE_INDEX:
3034 			if (get_numeric_property(zhp, prop, src,
3035 			    &source, &val) != 0)
3036 				return (-1);
3037 			if (zfs_prop_index_to_string(prop, val, &strval) != 0)
3038 				return (-1);
3039 
3040 			(void) strlcpy(propbuf, strval, proplen);
3041 			zcp_check(zhp, prop, 0, strval);
3042 			break;
3043 
3044 		default:
3045 			abort();
3046 		}
3047 	}
3048 
3049 	get_source(zhp, src, source, statbuf, statlen);
3050 
3051 	return (0);
3052 }
3053 
3054 /*
3055  * Utility function to get the given numeric property.  Does no validation that
3056  * the given property is the appropriate type; should only be used with
3057  * hard-coded property types.
3058  */
3059 uint64_t
3060 zfs_prop_get_int(zfs_handle_t *zhp, zfs_prop_t prop)
3061 {
3062 	const char *source;
3063 	uint64_t val = 0;
3064 
3065 	(void) get_numeric_property(zhp, prop, NULL, &source, &val);
3066 
3067 	return (val);
3068 }
3069 
3070 static int
3071 zfs_prop_set_int(zfs_handle_t *zhp, zfs_prop_t prop, uint64_t val)
3072 {
3073 	char buf[64];
3074 
3075 	(void) snprintf(buf, sizeof (buf), "%llu", (longlong_t)val);
3076 	return (zfs_prop_set(zhp, zfs_prop_to_name(prop), buf));
3077 }
3078 
3079 /*
3080  * Similar to zfs_prop_get(), but returns the value as an integer.
3081  */
3082 int
3083 zfs_prop_get_numeric(zfs_handle_t *zhp, zfs_prop_t prop, uint64_t *value,
3084     zprop_source_t *src, char *statbuf, size_t statlen)
3085 {
3086 	const char *source;
3087 
3088 	/*
3089 	 * Check to see if this property applies to our object
3090 	 */
3091 	if (!zfs_prop_valid_for_type(prop, zhp->zfs_type, B_FALSE)) {
3092 		return (zfs_error_fmt(zhp->zfs_hdl, EZFS_PROPTYPE,
3093 		    dgettext(TEXT_DOMAIN, "cannot get property '%s'"),
3094 		    zfs_prop_to_name(prop)));
3095 	}
3096 
3097 	if (src)
3098 		*src = ZPROP_SRC_NONE;
3099 
3100 	if (get_numeric_property(zhp, prop, src, &source, value) != 0)
3101 		return (-1);
3102 
3103 	get_source(zhp, src, source, statbuf, statlen);
3104 
3105 	return (0);
3106 }
3107 
3108 #ifdef HAVE_IDMAP
3109 static int
3110 idmap_id_to_numeric_domain_rid(uid_t id, boolean_t isuser,
3111     char **domainp, idmap_rid_t *ridp)
3112 {
3113 	idmap_get_handle_t *get_hdl = NULL;
3114 	idmap_stat status;
3115 	int err = EINVAL;
3116 
3117 	if (idmap_get_create(&get_hdl) != IDMAP_SUCCESS)
3118 		goto out;
3119 
3120 	if (isuser) {
3121 		err = idmap_get_sidbyuid(get_hdl, id,
3122 		    IDMAP_REQ_FLG_USE_CACHE, domainp, ridp, &status);
3123 	} else {
3124 		err = idmap_get_sidbygid(get_hdl, id,
3125 		    IDMAP_REQ_FLG_USE_CACHE, domainp, ridp, &status);
3126 	}
3127 	if (err == IDMAP_SUCCESS &&
3128 	    idmap_get_mappings(get_hdl) == IDMAP_SUCCESS &&
3129 	    status == IDMAP_SUCCESS)
3130 		err = 0;
3131 	else
3132 		err = EINVAL;
3133 out:
3134 	if (get_hdl)
3135 		idmap_get_destroy(get_hdl);
3136 	return (err);
3137 }
3138 #endif /* HAVE_IDMAP */
3139 
3140 /*
3141  * convert the propname into parameters needed by kernel
3142  * Eg: userquota@ahrens -> ZFS_PROP_USERQUOTA, "", 126829
3143  * Eg: userused@matt@domain -> ZFS_PROP_USERUSED, "S-1-123-456", 789
3144  * Eg: groupquota@staff -> ZFS_PROP_GROUPQUOTA, "", 1234
3145  * Eg: groupused@staff -> ZFS_PROP_GROUPUSED, "", 1234
3146  * Eg: projectquota@123 -> ZFS_PROP_PROJECTQUOTA, "", 123
3147  * Eg: projectused@789 -> ZFS_PROP_PROJECTUSED, "", 789
3148  */
3149 static int
3150 userquota_propname_decode(const char *propname, boolean_t zoned,
3151     zfs_userquota_prop_t *typep, char *domain, int domainlen, uint64_t *ridp)
3152 {
3153 	zfs_userquota_prop_t type;
3154 	char *cp;
3155 	boolean_t isuser;
3156 	boolean_t isgroup;
3157 	boolean_t isproject;
3158 	struct passwd *pw;
3159 	struct group *gr;
3160 
3161 	domain[0] = '\0';
3162 
3163 	/* Figure out the property type ({user|group|project}{quota|space}) */
3164 	for (type = 0; type < ZFS_NUM_USERQUOTA_PROPS; type++) {
3165 		if (strncmp(propname, zfs_userquota_prop_prefixes[type],
3166 		    strlen(zfs_userquota_prop_prefixes[type])) == 0)
3167 			break;
3168 	}
3169 	if (type == ZFS_NUM_USERQUOTA_PROPS)
3170 		return (EINVAL);
3171 	*typep = type;
3172 
3173 	isuser = (type == ZFS_PROP_USERQUOTA || type == ZFS_PROP_USERUSED ||
3174 	    type == ZFS_PROP_USEROBJQUOTA ||
3175 	    type == ZFS_PROP_USEROBJUSED);
3176 	isgroup = (type == ZFS_PROP_GROUPQUOTA || type == ZFS_PROP_GROUPUSED ||
3177 	    type == ZFS_PROP_GROUPOBJQUOTA ||
3178 	    type == ZFS_PROP_GROUPOBJUSED);
3179 	isproject = (type == ZFS_PROP_PROJECTQUOTA ||
3180 	    type == ZFS_PROP_PROJECTUSED || type == ZFS_PROP_PROJECTOBJQUOTA ||
3181 	    type == ZFS_PROP_PROJECTOBJUSED);
3182 
3183 	cp = strchr(propname, '@') + 1;
3184 
3185 	if (isuser && (pw = getpwnam(cp)) != NULL) {
3186 		if (zoned && getzoneid() == GLOBAL_ZONEID)
3187 			return (ENOENT);
3188 		*ridp = pw->pw_uid;
3189 	} else if (isgroup && (gr = getgrnam(cp)) != NULL) {
3190 		if (zoned && getzoneid() == GLOBAL_ZONEID)
3191 			return (ENOENT);
3192 		*ridp = gr->gr_gid;
3193 	} else if (!isproject && strchr(cp, '@')) {
3194 #ifdef HAVE_IDMAP
3195 		/*
3196 		 * It's a SID name (eg "user@domain") that needs to be
3197 		 * turned into S-1-domainID-RID.
3198 		 */
3199 		directory_error_t e;
3200 		char *numericsid = NULL;
3201 		char *end;
3202 
3203 		if (zoned && getzoneid() == GLOBAL_ZONEID)
3204 			return (ENOENT);
3205 		if (isuser) {
3206 			e = directory_sid_from_user_name(NULL,
3207 			    cp, &numericsid);
3208 		} else {
3209 			e = directory_sid_from_group_name(NULL,
3210 			    cp, &numericsid);
3211 		}
3212 		if (e != NULL) {
3213 			directory_error_free(e);
3214 			return (ENOENT);
3215 		}
3216 		if (numericsid == NULL)
3217 			return (ENOENT);
3218 		cp = numericsid;
3219 		(void) strlcpy(domain, cp, domainlen);
3220 		cp = strrchr(domain, '-');
3221 		*cp = '\0';
3222 		cp++;
3223 
3224 		errno = 0;
3225 		*ridp = strtoull(cp, &end, 10);
3226 		free(numericsid);
3227 
3228 		if (errno != 0 || *end != '\0')
3229 			return (EINVAL);
3230 #else
3231 		(void) domainlen;
3232 		return (ENOSYS);
3233 #endif /* HAVE_IDMAP */
3234 	} else {
3235 		/* It's a user/group/project ID (eg "12345"). */
3236 		uid_t id;
3237 		char *end;
3238 		id = strtoul(cp, &end, 10);
3239 		if (*end != '\0')
3240 			return (EINVAL);
3241 		if (id > MAXUID && !isproject) {
3242 #ifdef HAVE_IDMAP
3243 			/* It's an ephemeral ID. */
3244 			idmap_rid_t rid;
3245 			char *mapdomain;
3246 
3247 			if (idmap_id_to_numeric_domain_rid(id, isuser,
3248 			    &mapdomain, &rid) != 0)
3249 				return (ENOENT);
3250 			(void) strlcpy(domain, mapdomain, domainlen);
3251 			*ridp = rid;
3252 #else
3253 			return (ENOSYS);
3254 #endif /* HAVE_IDMAP */
3255 		} else {
3256 			*ridp = id;
3257 		}
3258 	}
3259 
3260 	return (0);
3261 }
3262 
3263 static int
3264 zfs_prop_get_userquota_common(zfs_handle_t *zhp, const char *propname,
3265     uint64_t *propvalue, zfs_userquota_prop_t *typep)
3266 {
3267 	int err;
3268 	zfs_cmd_t zc = {"\0"};
3269 
3270 	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
3271 
3272 	err = userquota_propname_decode(propname,
3273 	    zfs_prop_get_int(zhp, ZFS_PROP_ZONED),
3274 	    typep, zc.zc_value, sizeof (zc.zc_value), &zc.zc_guid);
3275 	zc.zc_objset_type = *typep;
3276 	if (err)
3277 		return (err);
3278 
3279 	err = zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_USERSPACE_ONE, &zc);
3280 	if (err)
3281 		return (err);
3282 
3283 	*propvalue = zc.zc_cookie;
3284 	return (0);
3285 }
3286 
3287 int
3288 zfs_prop_get_userquota_int(zfs_handle_t *zhp, const char *propname,
3289     uint64_t *propvalue)
3290 {
3291 	zfs_userquota_prop_t type;
3292 
3293 	return (zfs_prop_get_userquota_common(zhp, propname, propvalue,
3294 	    &type));
3295 }
3296 
3297 int
3298 zfs_prop_get_userquota(zfs_handle_t *zhp, const char *propname,
3299     char *propbuf, int proplen, boolean_t literal)
3300 {
3301 	int err;
3302 	uint64_t propvalue;
3303 	zfs_userquota_prop_t type;
3304 
3305 	err = zfs_prop_get_userquota_common(zhp, propname, &propvalue,
3306 	    &type);
3307 
3308 	if (err)
3309 		return (err);
3310 
3311 	if (literal) {
3312 		(void) snprintf(propbuf, proplen, "%llu",
3313 		    (u_longlong_t)propvalue);
3314 	} else if (propvalue == 0 &&
3315 	    (type == ZFS_PROP_USERQUOTA || type == ZFS_PROP_GROUPQUOTA ||
3316 	    type == ZFS_PROP_USEROBJQUOTA || type == ZFS_PROP_GROUPOBJQUOTA ||
3317 	    type == ZFS_PROP_PROJECTQUOTA ||
3318 	    type == ZFS_PROP_PROJECTOBJQUOTA)) {
3319 		(void) strlcpy(propbuf, "none", proplen);
3320 	} else if (type == ZFS_PROP_USERQUOTA || type == ZFS_PROP_GROUPQUOTA ||
3321 	    type == ZFS_PROP_USERUSED || type == ZFS_PROP_GROUPUSED ||
3322 	    type == ZFS_PROP_PROJECTUSED || type == ZFS_PROP_PROJECTQUOTA) {
3323 		zfs_nicebytes(propvalue, propbuf, proplen);
3324 	} else {
3325 		zfs_nicenum(propvalue, propbuf, proplen);
3326 	}
3327 	return (0);
3328 }
3329 
3330 /*
3331  * propname must start with "written@" or "written#".
3332  */
3333 int
3334 zfs_prop_get_written_int(zfs_handle_t *zhp, const char *propname,
3335     uint64_t *propvalue)
3336 {
3337 	int err;
3338 	zfs_cmd_t zc = {"\0"};
3339 	const char *snapname;
3340 
3341 	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
3342 
3343 	assert(zfs_prop_written(propname));
3344 	snapname = propname + strlen("written@");
3345 	if (strchr(snapname, '@') != NULL || strchr(snapname, '#') != NULL) {
3346 		/* full snapshot or bookmark name specified */
3347 		(void) strlcpy(zc.zc_value, snapname, sizeof (zc.zc_value));
3348 	} else {
3349 		/* snapname is the short name, append it to zhp's fsname */
3350 		char *cp;
3351 
3352 		(void) strlcpy(zc.zc_value, zhp->zfs_name,
3353 		    sizeof (zc.zc_value));
3354 		cp = strchr(zc.zc_value, '@');
3355 		if (cp != NULL)
3356 			*cp = '\0';
3357 		(void) strlcat(zc.zc_value, snapname - 1, sizeof (zc.zc_value));
3358 	}
3359 
3360 	err = zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_SPACE_WRITTEN, &zc);
3361 	if (err)
3362 		return (err);
3363 
3364 	*propvalue = zc.zc_cookie;
3365 	return (0);
3366 }
3367 
3368 int
3369 zfs_prop_get_written(zfs_handle_t *zhp, const char *propname,
3370     char *propbuf, int proplen, boolean_t literal)
3371 {
3372 	int err;
3373 	uint64_t propvalue;
3374 
3375 	err = zfs_prop_get_written_int(zhp, propname, &propvalue);
3376 
3377 	if (err)
3378 		return (err);
3379 
3380 	if (literal) {
3381 		(void) snprintf(propbuf, proplen, "%llu",
3382 		    (u_longlong_t)propvalue);
3383 	} else {
3384 		zfs_nicebytes(propvalue, propbuf, proplen);
3385 	}
3386 
3387 	return (0);
3388 }
3389 
3390 /*
3391  * Returns the name of the given zfs handle.
3392  */
3393 const char *
3394 zfs_get_name(const zfs_handle_t *zhp)
3395 {
3396 	return (zhp->zfs_name);
3397 }
3398 
3399 /*
3400  * Returns the name of the parent pool for the given zfs handle.
3401  */
3402 const char *
3403 zfs_get_pool_name(const zfs_handle_t *zhp)
3404 {
3405 	return (zhp->zpool_hdl->zpool_name);
3406 }
3407 
3408 /*
3409  * Returns the type of the given zfs handle.
3410  */
3411 zfs_type_t
3412 zfs_get_type(const zfs_handle_t *zhp)
3413 {
3414 	return (zhp->zfs_type);
3415 }
3416 
3417 /*
3418  * Returns the type of the given zfs handle,
3419  * or, if a snapshot, the type of the snapshotted dataset.
3420  */
3421 zfs_type_t
3422 zfs_get_underlying_type(const zfs_handle_t *zhp)
3423 {
3424 	return (zhp->zfs_head_type);
3425 }
3426 
3427 /*
3428  * Is one dataset name a child dataset of another?
3429  *
3430  * Needs to handle these cases:
3431  * Dataset 1	"a/foo"		"a/foo"		"a/foo"		"a/foo"
3432  * Dataset 2	"a/fo"		"a/foobar"	"a/bar/baz"	"a/foo/bar"
3433  * Descendant?	No.		No.		No.		Yes.
3434  */
3435 static boolean_t
3436 is_descendant(const char *ds1, const char *ds2)
3437 {
3438 	size_t d1len = strlen(ds1);
3439 
3440 	/* ds2 can't be a descendant if it's smaller */
3441 	if (strlen(ds2) < d1len)
3442 		return (B_FALSE);
3443 
3444 	/* otherwise, compare strings and verify that there's a '/' char */
3445 	return (ds2[d1len] == '/' && (strncmp(ds1, ds2, d1len) == 0));
3446 }
3447 
3448 /*
3449  * Given a complete name, return just the portion that refers to the parent.
3450  * Will return -1 if there is no parent (path is just the name of the
3451  * pool).
3452  */
3453 static int
3454 parent_name(const char *path, char *buf, size_t buflen)
3455 {
3456 	char *slashp;
3457 
3458 	(void) strlcpy(buf, path, buflen);
3459 
3460 	if ((slashp = strrchr(buf, '/')) == NULL)
3461 		return (-1);
3462 	*slashp = '\0';
3463 
3464 	return (0);
3465 }
3466 
3467 int
3468 zfs_parent_name(zfs_handle_t *zhp, char *buf, size_t buflen)
3469 {
3470 	return (parent_name(zfs_get_name(zhp), buf, buflen));
3471 }
3472 
3473 /*
3474  * If accept_ancestor is false, then check to make sure that the given path has
3475  * a parent, and that it exists.  If accept_ancestor is true, then find the
3476  * closest existing ancestor for the given path.  In prefixlen return the
3477  * length of already existing prefix of the given path.  We also fetch the
3478  * 'zoned' property, which is used to validate property settings when creating
3479  * new datasets.
3480  */
3481 static int
3482 check_parents(libzfs_handle_t *hdl, const char *path, uint64_t *zoned,
3483     boolean_t accept_ancestor, int *prefixlen)
3484 {
3485 	zfs_cmd_t zc = {"\0"};
3486 	char parent[ZFS_MAX_DATASET_NAME_LEN];
3487 	char *slash;
3488 	zfs_handle_t *zhp;
3489 	char errbuf[ERRBUFLEN];
3490 	uint64_t is_zoned;
3491 
3492 	(void) snprintf(errbuf, sizeof (errbuf),
3493 	    dgettext(TEXT_DOMAIN, "cannot create '%s'"), path);
3494 
3495 	/* get parent, and check to see if this is just a pool */
3496 	if (parent_name(path, parent, sizeof (parent)) != 0) {
3497 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3498 		    "missing dataset name"));
3499 		return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
3500 	}
3501 
3502 	/* check to see if the pool exists */
3503 	if ((slash = strchr(parent, '/')) == NULL)
3504 		slash = parent + strlen(parent);
3505 	(void) strlcpy(zc.zc_name, parent,
3506 	    MIN(sizeof (zc.zc_name), slash - parent + 1));
3507 	if (zfs_ioctl(hdl, ZFS_IOC_OBJSET_STATS, &zc) != 0 &&
3508 	    errno == ENOENT) {
3509 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3510 		    "no such pool '%s'"), zc.zc_name);
3511 		return (zfs_error(hdl, EZFS_NOENT, errbuf));
3512 	}
3513 
3514 	/* check to see if the parent dataset exists */
3515 	while ((zhp = make_dataset_handle(hdl, parent)) == NULL) {
3516 		if (errno == ENOENT && accept_ancestor) {
3517 			/*
3518 			 * Go deeper to find an ancestor, give up on top level.
3519 			 */
3520 			if (parent_name(parent, parent, sizeof (parent)) != 0) {
3521 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3522 				    "no such pool '%s'"), zc.zc_name);
3523 				return (zfs_error(hdl, EZFS_NOENT, errbuf));
3524 			}
3525 		} else if (errno == ENOENT) {
3526 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3527 			    "parent does not exist"));
3528 			return (zfs_error(hdl, EZFS_NOENT, errbuf));
3529 		} else
3530 			return (zfs_standard_error(hdl, errno, errbuf));
3531 	}
3532 
3533 	is_zoned = zfs_prop_get_int(zhp, ZFS_PROP_ZONED);
3534 	if (zoned != NULL)
3535 		*zoned = is_zoned;
3536 
3537 	/* we are in a non-global zone, but parent is in the global zone */
3538 	if (getzoneid() != GLOBAL_ZONEID && !is_zoned) {
3539 		(void) zfs_standard_error(hdl, EPERM, errbuf);
3540 		zfs_close(zhp);
3541 		return (-1);
3542 	}
3543 
3544 	/* make sure parent is a filesystem */
3545 	if (zfs_get_type(zhp) != ZFS_TYPE_FILESYSTEM) {
3546 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3547 		    "parent is not a filesystem"));
3548 		(void) zfs_error(hdl, EZFS_BADTYPE, errbuf);
3549 		zfs_close(zhp);
3550 		return (-1);
3551 	}
3552 
3553 	zfs_close(zhp);
3554 	if (prefixlen != NULL)
3555 		*prefixlen = strlen(parent);
3556 	return (0);
3557 }
3558 
3559 /*
3560  * Finds whether the dataset of the given type(s) exists.
3561  */
3562 boolean_t
3563 zfs_dataset_exists(libzfs_handle_t *hdl, const char *path, zfs_type_t types)
3564 {
3565 	zfs_handle_t *zhp;
3566 
3567 	if (!zfs_validate_name(hdl, path, types, B_FALSE))
3568 		return (B_FALSE);
3569 
3570 	/*
3571 	 * Try to get stats for the dataset, which will tell us if it exists.
3572 	 */
3573 	if ((zhp = make_dataset_handle(hdl, path)) != NULL) {
3574 		int ds_type = zhp->zfs_type;
3575 
3576 		zfs_close(zhp);
3577 		if (types & ds_type)
3578 			return (B_TRUE);
3579 	}
3580 	return (B_FALSE);
3581 }
3582 
3583 /*
3584  * Given a path to 'target', create all the ancestors between
3585  * the prefixlen portion of the path, and the target itself.
3586  * Fail if the initial prefixlen-ancestor does not already exist.
3587  */
3588 int
3589 create_parents(libzfs_handle_t *hdl, char *target, int prefixlen)
3590 {
3591 	zfs_handle_t *h;
3592 	char *cp;
3593 	const char *opname;
3594 
3595 	/* make sure prefix exists */
3596 	cp = target + prefixlen;
3597 	if (*cp != '/') {
3598 		assert(strchr(cp, '/') == NULL);
3599 		h = zfs_open(hdl, target, ZFS_TYPE_FILESYSTEM);
3600 	} else {
3601 		*cp = '\0';
3602 		h = zfs_open(hdl, target, ZFS_TYPE_FILESYSTEM);
3603 		*cp = '/';
3604 	}
3605 	if (h == NULL)
3606 		return (-1);
3607 	zfs_close(h);
3608 
3609 	/*
3610 	 * Attempt to create, mount, and share any ancestor filesystems,
3611 	 * up to the prefixlen-long one.
3612 	 */
3613 	for (cp = target + prefixlen + 1;
3614 	    (cp = strchr(cp, '/')) != NULL; *cp = '/', cp++) {
3615 
3616 		*cp = '\0';
3617 
3618 		h = make_dataset_handle(hdl, target);
3619 		if (h) {
3620 			/* it already exists, nothing to do here */
3621 			zfs_close(h);
3622 			continue;
3623 		}
3624 
3625 		if (zfs_create(hdl, target, ZFS_TYPE_FILESYSTEM,
3626 		    NULL) != 0) {
3627 			opname = dgettext(TEXT_DOMAIN, "create");
3628 			goto ancestorerr;
3629 		}
3630 
3631 		h = zfs_open(hdl, target, ZFS_TYPE_FILESYSTEM);
3632 		if (h == NULL) {
3633 			opname = dgettext(TEXT_DOMAIN, "open");
3634 			goto ancestorerr;
3635 		}
3636 
3637 		if (zfs_mount(h, NULL, 0) != 0) {
3638 			opname = dgettext(TEXT_DOMAIN, "mount");
3639 			goto ancestorerr;
3640 		}
3641 
3642 		if (zfs_share(h, NULL) != 0) {
3643 			opname = dgettext(TEXT_DOMAIN, "share");
3644 			goto ancestorerr;
3645 		}
3646 
3647 		zfs_close(h);
3648 	}
3649 	zfs_commit_shares(NULL);
3650 
3651 	return (0);
3652 
3653 ancestorerr:
3654 	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3655 	    "failed to %s ancestor '%s'"), opname, target);
3656 	return (-1);
3657 }
3658 
3659 /*
3660  * Creates non-existing ancestors of the given path.
3661  */
3662 int
3663 zfs_create_ancestors(libzfs_handle_t *hdl, const char *path)
3664 {
3665 	int prefix;
3666 	char *path_copy;
3667 	char errbuf[ERRBUFLEN];
3668 	int rc = 0;
3669 
3670 	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
3671 	    "cannot create '%s'"), path);
3672 
3673 	/*
3674 	 * Check that we are not passing the nesting limit
3675 	 * before we start creating any ancestors.
3676 	 */
3677 	if (dataset_nestcheck(path) != 0) {
3678 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3679 		    "maximum name nesting depth exceeded"));
3680 		return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
3681 	}
3682 
3683 	if (check_parents(hdl, path, NULL, B_TRUE, &prefix) != 0)
3684 		return (-1);
3685 
3686 	if ((path_copy = strdup(path)) != NULL) {
3687 		rc = create_parents(hdl, path_copy, prefix);
3688 		free(path_copy);
3689 	}
3690 	if (path_copy == NULL || rc != 0)
3691 		return (-1);
3692 
3693 	return (0);
3694 }
3695 
3696 /*
3697  * Create a new filesystem or volume.
3698  */
3699 int
3700 zfs_create(libzfs_handle_t *hdl, const char *path, zfs_type_t type,
3701     nvlist_t *props)
3702 {
3703 	int ret;
3704 	uint64_t size = 0;
3705 	uint64_t blocksize = zfs_prop_default_numeric(ZFS_PROP_VOLBLOCKSIZE);
3706 	uint64_t zoned;
3707 	enum lzc_dataset_type ost;
3708 	zpool_handle_t *zpool_handle;
3709 	uint8_t *wkeydata = NULL;
3710 	uint_t wkeylen = 0;
3711 	char errbuf[ERRBUFLEN];
3712 	char parent[ZFS_MAX_DATASET_NAME_LEN];
3713 
3714 	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
3715 	    "cannot create '%s'"), path);
3716 
3717 	/* validate the path, taking care to note the extended error message */
3718 	if (!zfs_validate_name(hdl, path, type, B_TRUE))
3719 		return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
3720 
3721 	if (dataset_nestcheck(path) != 0) {
3722 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3723 		    "maximum name nesting depth exceeded"));
3724 		return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
3725 	}
3726 
3727 	/* validate parents exist */
3728 	if (check_parents(hdl, path, &zoned, B_FALSE, NULL) != 0)
3729 		return (-1);
3730 
3731 	/*
3732 	 * The failure modes when creating a dataset of a different type over
3733 	 * one that already exists is a little strange.  In particular, if you
3734 	 * try to create a dataset on top of an existing dataset, the ioctl()
3735 	 * will return ENOENT, not EEXIST.  To prevent this from happening, we
3736 	 * first try to see if the dataset exists.
3737 	 */
3738 	if (zfs_dataset_exists(hdl, path, ZFS_TYPE_DATASET)) {
3739 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3740 		    "dataset already exists"));
3741 		return (zfs_error(hdl, EZFS_EXISTS, errbuf));
3742 	}
3743 
3744 	if (type == ZFS_TYPE_VOLUME)
3745 		ost = LZC_DATSET_TYPE_ZVOL;
3746 	else
3747 		ost = LZC_DATSET_TYPE_ZFS;
3748 
3749 	/* open zpool handle for prop validation */
3750 	char pool_path[ZFS_MAX_DATASET_NAME_LEN];
3751 	(void) strlcpy(pool_path, path, sizeof (pool_path));
3752 
3753 	/* truncate pool_path at first slash */
3754 	char *p = strchr(pool_path, '/');
3755 	if (p != NULL)
3756 		*p = '\0';
3757 
3758 	if ((zpool_handle = zpool_open(hdl, pool_path)) == NULL)
3759 		return (-1);
3760 
3761 	if (props && (props = zfs_valid_proplist(hdl, type, props,
3762 	    zoned, NULL, zpool_handle, B_TRUE, errbuf)) == 0) {
3763 		zpool_close(zpool_handle);
3764 		return (-1);
3765 	}
3766 	zpool_close(zpool_handle);
3767 
3768 	if (type == ZFS_TYPE_VOLUME) {
3769 		/*
3770 		 * If we are creating a volume, the size and block size must
3771 		 * satisfy a few restraints.  First, the blocksize must be a
3772 		 * valid block size between SPA_{MIN,MAX}BLOCKSIZE.  Second, the
3773 		 * volsize must be a multiple of the block size, and cannot be
3774 		 * zero.
3775 		 */
3776 		if (props == NULL || nvlist_lookup_uint64(props,
3777 		    zfs_prop_to_name(ZFS_PROP_VOLSIZE), &size) != 0) {
3778 			nvlist_free(props);
3779 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3780 			    "missing volume size"));
3781 			return (zfs_error(hdl, EZFS_BADPROP, errbuf));
3782 		}
3783 
3784 		if ((ret = nvlist_lookup_uint64(props,
3785 		    zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3786 		    &blocksize)) != 0) {
3787 			if (ret == ENOENT) {
3788 				blocksize = zfs_prop_default_numeric(
3789 				    ZFS_PROP_VOLBLOCKSIZE);
3790 			} else {
3791 				nvlist_free(props);
3792 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3793 				    "missing volume block size"));
3794 				return (zfs_error(hdl, EZFS_BADPROP, errbuf));
3795 			}
3796 		}
3797 
3798 		if (size == 0) {
3799 			nvlist_free(props);
3800 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3801 			    "volume size cannot be zero"));
3802 			return (zfs_error(hdl, EZFS_BADPROP, errbuf));
3803 		}
3804 
3805 		if (size % blocksize != 0) {
3806 			nvlist_free(props);
3807 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3808 			    "volume size must be a multiple of volume block "
3809 			    "size"));
3810 			return (zfs_error(hdl, EZFS_BADPROP, errbuf));
3811 		}
3812 	}
3813 
3814 	(void) parent_name(path, parent, sizeof (parent));
3815 	if (zfs_crypto_create(hdl, parent, props, NULL, B_TRUE,
3816 	    &wkeydata, &wkeylen) != 0) {
3817 		nvlist_free(props);
3818 		return (zfs_error(hdl, EZFS_CRYPTOFAILED, errbuf));
3819 	}
3820 
3821 	/* create the dataset */
3822 	ret = lzc_create(path, ost, props, wkeydata, wkeylen);
3823 	nvlist_free(props);
3824 	if (wkeydata != NULL)
3825 		free(wkeydata);
3826 
3827 	/* check for failure */
3828 	if (ret != 0) {
3829 		switch (errno) {
3830 		case ENOENT:
3831 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3832 			    "no such parent '%s'"), parent);
3833 			return (zfs_error(hdl, EZFS_NOENT, errbuf));
3834 
3835 		case ENOTSUP:
3836 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3837 			    "pool must be upgraded to set this "
3838 			    "property or value"));
3839 			return (zfs_error(hdl, EZFS_BADVERSION, errbuf));
3840 
3841 		case EACCES:
3842 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3843 			    "encryption root's key is not loaded "
3844 			    "or provided"));
3845 			return (zfs_error(hdl, EZFS_CRYPTOFAILED, errbuf));
3846 
3847 		case ERANGE:
3848 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3849 			    "invalid property value(s) specified"));
3850 			return (zfs_error(hdl, EZFS_BADPROP, errbuf));
3851 #ifdef _ILP32
3852 		case EOVERFLOW:
3853 			/*
3854 			 * This platform can't address a volume this big.
3855 			 */
3856 			if (type == ZFS_TYPE_VOLUME)
3857 				return (zfs_error(hdl, EZFS_VOLTOOBIG,
3858 				    errbuf));
3859 			zfs_fallthrough;
3860 #endif
3861 		default:
3862 			return (zfs_standard_error(hdl, errno, errbuf));
3863 		}
3864 	}
3865 
3866 	return (0);
3867 }
3868 
3869 /*
3870  * Destroys the given dataset.  The caller must make sure that the filesystem
3871  * isn't mounted, and that there are no active dependents. If the file system
3872  * does not exist this function does nothing.
3873  */
3874 int
3875 zfs_destroy(zfs_handle_t *zhp, boolean_t defer)
3876 {
3877 	int error;
3878 
3879 	if (zhp->zfs_type != ZFS_TYPE_SNAPSHOT && defer)
3880 		return (EINVAL);
3881 
3882 	if (zhp->zfs_type == ZFS_TYPE_BOOKMARK) {
3883 		nvlist_t *nv = fnvlist_alloc();
3884 		fnvlist_add_boolean(nv, zhp->zfs_name);
3885 		error = lzc_destroy_bookmarks(nv, NULL);
3886 		fnvlist_free(nv);
3887 		if (error != 0) {
3888 			return (zfs_standard_error_fmt(zhp->zfs_hdl, error,
3889 			    dgettext(TEXT_DOMAIN, "cannot destroy '%s'"),
3890 			    zhp->zfs_name));
3891 		}
3892 		return (0);
3893 	}
3894 
3895 	if (zhp->zfs_type == ZFS_TYPE_SNAPSHOT) {
3896 		nvlist_t *nv = fnvlist_alloc();
3897 		fnvlist_add_boolean(nv, zhp->zfs_name);
3898 		error = lzc_destroy_snaps(nv, defer, NULL);
3899 		fnvlist_free(nv);
3900 	} else {
3901 		error = lzc_destroy(zhp->zfs_name);
3902 	}
3903 
3904 	if (error != 0 && error != ENOENT) {
3905 		return (zfs_standard_error_fmt(zhp->zfs_hdl, errno,
3906 		    dgettext(TEXT_DOMAIN, "cannot destroy '%s'"),
3907 		    zhp->zfs_name));
3908 	}
3909 
3910 	remove_mountpoint(zhp);
3911 
3912 	return (0);
3913 }
3914 
3915 struct destroydata {
3916 	nvlist_t *nvl;
3917 	const char *snapname;
3918 };
3919 
3920 static int
3921 zfs_check_snap_cb(zfs_handle_t *zhp, void *arg)
3922 {
3923 	struct destroydata *dd = arg;
3924 	char name[ZFS_MAX_DATASET_NAME_LEN];
3925 	int rv = 0;
3926 
3927 	if (snprintf(name, sizeof (name), "%s@%s", zhp->zfs_name,
3928 	    dd->snapname) >= sizeof (name))
3929 		return (EINVAL);
3930 
3931 	if (lzc_exists(name))
3932 		fnvlist_add_boolean(dd->nvl, name);
3933 
3934 	rv = zfs_iter_filesystems_v2(zhp, 0, zfs_check_snap_cb, dd);
3935 	zfs_close(zhp);
3936 	return (rv);
3937 }
3938 
3939 /*
3940  * Destroys all snapshots with the given name in zhp & descendants.
3941  */
3942 int
3943 zfs_destroy_snaps(zfs_handle_t *zhp, char *snapname, boolean_t defer)
3944 {
3945 	int ret;
3946 	struct destroydata dd = { 0 };
3947 
3948 	dd.snapname = snapname;
3949 	dd.nvl = fnvlist_alloc();
3950 	(void) zfs_check_snap_cb(zfs_handle_dup(zhp), &dd);
3951 
3952 	if (nvlist_empty(dd.nvl)) {
3953 		ret = zfs_standard_error_fmt(zhp->zfs_hdl, ENOENT,
3954 		    dgettext(TEXT_DOMAIN, "cannot destroy '%s@%s'"),
3955 		    zhp->zfs_name, snapname);
3956 	} else {
3957 		ret = zfs_destroy_snaps_nvl(zhp->zfs_hdl, dd.nvl, defer);
3958 	}
3959 	fnvlist_free(dd.nvl);
3960 	return (ret);
3961 }
3962 
3963 /*
3964  * Destroys all the snapshots named in the nvlist.
3965  */
3966 int
3967 zfs_destroy_snaps_nvl(libzfs_handle_t *hdl, nvlist_t *snaps, boolean_t defer)
3968 {
3969 	nvlist_t *errlist = NULL;
3970 	nvpair_t *pair;
3971 
3972 	int ret = zfs_destroy_snaps_nvl_os(hdl, snaps);
3973 	if (ret != 0)
3974 		return (ret);
3975 
3976 	ret = lzc_destroy_snaps(snaps, defer, &errlist);
3977 
3978 	if (ret == 0) {
3979 		nvlist_free(errlist);
3980 		return (0);
3981 	}
3982 
3983 	if (nvlist_empty(errlist)) {
3984 		char errbuf[ERRBUFLEN];
3985 		(void) snprintf(errbuf, sizeof (errbuf),
3986 		    dgettext(TEXT_DOMAIN, "cannot destroy snapshots"));
3987 
3988 		ret = zfs_standard_error(hdl, ret, errbuf);
3989 	}
3990 	for (pair = nvlist_next_nvpair(errlist, NULL);
3991 	    pair != NULL; pair = nvlist_next_nvpair(errlist, pair)) {
3992 		char errbuf[ERRBUFLEN];
3993 		(void) snprintf(errbuf, sizeof (errbuf),
3994 		    dgettext(TEXT_DOMAIN, "cannot destroy snapshot %s"),
3995 		    nvpair_name(pair));
3996 
3997 		switch (fnvpair_value_int32(pair)) {
3998 		case EEXIST:
3999 			zfs_error_aux(hdl,
4000 			    dgettext(TEXT_DOMAIN, "snapshot is cloned"));
4001 			ret = zfs_error(hdl, EZFS_EXISTS, errbuf);
4002 			break;
4003 		default:
4004 			ret = zfs_standard_error(hdl, errno, errbuf);
4005 			break;
4006 		}
4007 	}
4008 
4009 	nvlist_free(errlist);
4010 	return (ret);
4011 }
4012 
4013 /*
4014  * Clones the given dataset.  The target must be of the same type as the source.
4015  */
4016 int
4017 zfs_clone(zfs_handle_t *zhp, const char *target, nvlist_t *props)
4018 {
4019 	char parent[ZFS_MAX_DATASET_NAME_LEN];
4020 	int ret;
4021 	char errbuf[ERRBUFLEN];
4022 	libzfs_handle_t *hdl = zhp->zfs_hdl;
4023 	uint64_t zoned;
4024 
4025 	assert(zhp->zfs_type == ZFS_TYPE_SNAPSHOT);
4026 
4027 	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
4028 	    "cannot create '%s'"), target);
4029 
4030 	/* validate the target/clone name */
4031 	if (!zfs_validate_name(hdl, target, ZFS_TYPE_FILESYSTEM, B_TRUE))
4032 		return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
4033 
4034 	/* validate parents exist */
4035 	if (check_parents(hdl, target, &zoned, B_FALSE, NULL) != 0)
4036 		return (-1);
4037 
4038 	(void) parent_name(target, parent, sizeof (parent));
4039 
4040 	/* do the clone */
4041 
4042 	if (props) {
4043 		zfs_type_t type = ZFS_TYPE_FILESYSTEM;
4044 
4045 		if (ZFS_IS_VOLUME(zhp))
4046 			type = ZFS_TYPE_VOLUME;
4047 		if ((props = zfs_valid_proplist(hdl, type, props, zoned,
4048 		    zhp, zhp->zpool_hdl, B_TRUE, errbuf)) == NULL)
4049 			return (-1);
4050 		if (zfs_fix_auto_resv(zhp, props) == -1) {
4051 			nvlist_free(props);
4052 			return (-1);
4053 		}
4054 	}
4055 
4056 	if (zfs_crypto_clone_check(hdl, zhp, parent, props) != 0) {
4057 		nvlist_free(props);
4058 		return (zfs_error(hdl, EZFS_CRYPTOFAILED, errbuf));
4059 	}
4060 
4061 	ret = lzc_clone(target, zhp->zfs_name, props);
4062 	nvlist_free(props);
4063 
4064 	if (ret != 0) {
4065 		switch (errno) {
4066 
4067 		case ENOENT:
4068 			/*
4069 			 * The parent doesn't exist.  We should have caught this
4070 			 * above, but there may a race condition that has since
4071 			 * destroyed the parent.
4072 			 *
4073 			 * At this point, we don't know whether it's the source
4074 			 * that doesn't exist anymore, or whether the target
4075 			 * dataset doesn't exist.
4076 			 */
4077 			zfs_error_aux(zhp->zfs_hdl, dgettext(TEXT_DOMAIN,
4078 			    "no such parent '%s'"), parent);
4079 			return (zfs_error(zhp->zfs_hdl, EZFS_NOENT, errbuf));
4080 
4081 		case EXDEV:
4082 			zfs_error_aux(zhp->zfs_hdl, dgettext(TEXT_DOMAIN,
4083 			    "source and target pools differ"));
4084 			return (zfs_error(zhp->zfs_hdl, EZFS_CROSSTARGET,
4085 			    errbuf));
4086 
4087 		default:
4088 			return (zfs_standard_error(zhp->zfs_hdl, errno,
4089 			    errbuf));
4090 		}
4091 	}
4092 
4093 	return (ret);
4094 }
4095 
4096 /*
4097  * Promotes the given clone fs to be the clone parent.
4098  */
4099 int
4100 zfs_promote(zfs_handle_t *zhp)
4101 {
4102 	libzfs_handle_t *hdl = zhp->zfs_hdl;
4103 	char snapname[ZFS_MAX_DATASET_NAME_LEN];
4104 	int ret;
4105 	char errbuf[ERRBUFLEN];
4106 
4107 	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
4108 	    "cannot promote '%s'"), zhp->zfs_name);
4109 
4110 	if (zhp->zfs_type == ZFS_TYPE_SNAPSHOT) {
4111 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4112 		    "snapshots can not be promoted"));
4113 		return (zfs_error(hdl, EZFS_BADTYPE, errbuf));
4114 	}
4115 
4116 	if (zhp->zfs_dmustats.dds_origin[0] == '\0') {
4117 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4118 		    "not a cloned filesystem"));
4119 		return (zfs_error(hdl, EZFS_BADTYPE, errbuf));
4120 	}
4121 
4122 	if (!zfs_validate_name(hdl, zhp->zfs_name, zhp->zfs_type, B_TRUE))
4123 		return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
4124 
4125 	ret = lzc_promote(zhp->zfs_name, snapname, sizeof (snapname));
4126 
4127 	if (ret != 0) {
4128 		switch (ret) {
4129 		case EACCES:
4130 			/*
4131 			 * Promoting encrypted dataset outside its
4132 			 * encryption root.
4133 			 */
4134 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4135 			    "cannot promote dataset outside its "
4136 			    "encryption root"));
4137 			return (zfs_error(hdl, EZFS_EXISTS, errbuf));
4138 
4139 		case EEXIST:
4140 			/* There is a conflicting snapshot name. */
4141 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4142 			    "conflicting snapshot '%s' from parent '%s'"),
4143 			    snapname, zhp->zfs_dmustats.dds_origin);
4144 			return (zfs_error(hdl, EZFS_EXISTS, errbuf));
4145 
4146 		default:
4147 			return (zfs_standard_error(hdl, ret, errbuf));
4148 		}
4149 	}
4150 	return (ret);
4151 }
4152 
4153 typedef struct snapdata {
4154 	nvlist_t *sd_nvl;
4155 	const char *sd_snapname;
4156 } snapdata_t;
4157 
4158 static int
4159 zfs_snapshot_cb(zfs_handle_t *zhp, void *arg)
4160 {
4161 	snapdata_t *sd = arg;
4162 	char name[ZFS_MAX_DATASET_NAME_LEN];
4163 	int rv = 0;
4164 
4165 	if (zfs_prop_get_int(zhp, ZFS_PROP_INCONSISTENT) == 0) {
4166 		if (snprintf(name, sizeof (name), "%s@%s", zfs_get_name(zhp),
4167 		    sd->sd_snapname) >= sizeof (name))
4168 			return (EINVAL);
4169 
4170 		fnvlist_add_boolean(sd->sd_nvl, name);
4171 
4172 		rv = zfs_iter_filesystems_v2(zhp, 0, zfs_snapshot_cb, sd);
4173 	}
4174 	zfs_close(zhp);
4175 
4176 	return (rv);
4177 }
4178 
4179 /*
4180  * Creates snapshots.  The keys in the snaps nvlist are the snapshots to be
4181  * created.
4182  */
4183 int
4184 zfs_snapshot_nvl(libzfs_handle_t *hdl, nvlist_t *snaps, nvlist_t *props)
4185 {
4186 	int ret;
4187 	char errbuf[ERRBUFLEN];
4188 	nvpair_t *elem;
4189 	nvlist_t *errors;
4190 	zpool_handle_t *zpool_hdl;
4191 	char pool[ZFS_MAX_DATASET_NAME_LEN];
4192 
4193 	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
4194 	    "cannot create snapshots "));
4195 
4196 	elem = NULL;
4197 	while ((elem = nvlist_next_nvpair(snaps, elem)) != NULL) {
4198 		const char *snapname = nvpair_name(elem);
4199 
4200 		/* validate the target name */
4201 		if (!zfs_validate_name(hdl, snapname, ZFS_TYPE_SNAPSHOT,
4202 		    B_TRUE)) {
4203 			(void) snprintf(errbuf, sizeof (errbuf),
4204 			    dgettext(TEXT_DOMAIN,
4205 			    "cannot create snapshot '%s'"), snapname);
4206 			return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
4207 		}
4208 	}
4209 
4210 	/*
4211 	 * get pool handle for prop validation. assumes all snaps are in the
4212 	 * same pool, as does lzc_snapshot (below).
4213 	 */
4214 	elem = nvlist_next_nvpair(snaps, NULL);
4215 	if (elem == NULL)
4216 		return (-1);
4217 	(void) strlcpy(pool, nvpair_name(elem), sizeof (pool));
4218 	pool[strcspn(pool, "/@")] = '\0';
4219 	zpool_hdl = zpool_open(hdl, pool);
4220 	if (zpool_hdl == NULL)
4221 		return (-1);
4222 
4223 	if (props != NULL &&
4224 	    (props = zfs_valid_proplist(hdl, ZFS_TYPE_SNAPSHOT,
4225 	    props, B_FALSE, NULL, zpool_hdl, B_FALSE, errbuf)) == NULL) {
4226 		zpool_close(zpool_hdl);
4227 		return (-1);
4228 	}
4229 	zpool_close(zpool_hdl);
4230 
4231 	ret = lzc_snapshot(snaps, props, &errors);
4232 
4233 	if (ret != 0) {
4234 		boolean_t printed = B_FALSE;
4235 		for (elem = nvlist_next_nvpair(errors, NULL);
4236 		    elem != NULL;
4237 		    elem = nvlist_next_nvpair(errors, elem)) {
4238 			(void) snprintf(errbuf, sizeof (errbuf),
4239 			    dgettext(TEXT_DOMAIN,
4240 			    "cannot create snapshot '%s'"), nvpair_name(elem));
4241 			(void) zfs_standard_error(hdl,
4242 			    fnvpair_value_int32(elem), errbuf);
4243 			printed = B_TRUE;
4244 		}
4245 		if (!printed) {
4246 			switch (ret) {
4247 			case EXDEV:
4248 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4249 				    "multiple snapshots of same "
4250 				    "fs not allowed"));
4251 				(void) zfs_error(hdl, EZFS_EXISTS, errbuf);
4252 
4253 				break;
4254 			default:
4255 				(void) zfs_standard_error(hdl, ret, errbuf);
4256 			}
4257 		}
4258 	}
4259 
4260 	nvlist_free(props);
4261 	nvlist_free(errors);
4262 	return (ret);
4263 }
4264 
4265 int
4266 zfs_snapshot(libzfs_handle_t *hdl, const char *path, boolean_t recursive,
4267     nvlist_t *props)
4268 {
4269 	int ret;
4270 	snapdata_t sd = { 0 };
4271 	char fsname[ZFS_MAX_DATASET_NAME_LEN];
4272 	char *cp;
4273 	zfs_handle_t *zhp;
4274 	char errbuf[ERRBUFLEN];
4275 
4276 	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
4277 	    "cannot snapshot %s"), path);
4278 
4279 	if (!zfs_validate_name(hdl, path, ZFS_TYPE_SNAPSHOT, B_TRUE))
4280 		return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
4281 
4282 	(void) strlcpy(fsname, path, sizeof (fsname));
4283 	cp = strchr(fsname, '@');
4284 	*cp = '\0';
4285 	sd.sd_snapname = cp + 1;
4286 
4287 	if ((zhp = zfs_open(hdl, fsname, ZFS_TYPE_FILESYSTEM |
4288 	    ZFS_TYPE_VOLUME)) == NULL) {
4289 		return (-1);
4290 	}
4291 
4292 	sd.sd_nvl = fnvlist_alloc();
4293 	if (recursive) {
4294 		(void) zfs_snapshot_cb(zfs_handle_dup(zhp), &sd);
4295 	} else {
4296 		fnvlist_add_boolean(sd.sd_nvl, path);
4297 	}
4298 
4299 	ret = zfs_snapshot_nvl(hdl, sd.sd_nvl, props);
4300 	fnvlist_free(sd.sd_nvl);
4301 	zfs_close(zhp);
4302 	return (ret);
4303 }
4304 
4305 /*
4306  * Destroy any more recent snapshots.  We invoke this callback on any dependents
4307  * of the snapshot first.  If the 'cb_dependent' member is non-zero, then this
4308  * is a dependent and we should just destroy it without checking the transaction
4309  * group.
4310  */
4311 typedef struct rollback_data {
4312 	const char	*cb_target;		/* the snapshot */
4313 	uint64_t	cb_create;		/* creation time reference */
4314 	boolean_t	cb_error;
4315 	boolean_t	cb_force;
4316 } rollback_data_t;
4317 
4318 static int
4319 rollback_destroy_dependent(zfs_handle_t *zhp, void *data)
4320 {
4321 	rollback_data_t *cbp = data;
4322 	prop_changelist_t *clp;
4323 
4324 	/* We must destroy this clone; first unmount it */
4325 	clp = changelist_gather(zhp, ZFS_PROP_NAME, 0,
4326 	    cbp->cb_force ? MS_FORCE: 0);
4327 	if (clp == NULL || changelist_prefix(clp) != 0) {
4328 		cbp->cb_error = B_TRUE;
4329 		zfs_close(zhp);
4330 		return (0);
4331 	}
4332 	if (zfs_destroy(zhp, B_FALSE) != 0)
4333 		cbp->cb_error = B_TRUE;
4334 	else
4335 		changelist_remove(clp, zhp->zfs_name);
4336 	(void) changelist_postfix(clp);
4337 	changelist_free(clp);
4338 
4339 	zfs_close(zhp);
4340 	return (0);
4341 }
4342 
4343 static int
4344 rollback_destroy(zfs_handle_t *zhp, void *data)
4345 {
4346 	rollback_data_t *cbp = data;
4347 
4348 	if (zfs_prop_get_int(zhp, ZFS_PROP_CREATETXG) > cbp->cb_create) {
4349 		cbp->cb_error |= zfs_iter_dependents_v2(zhp, 0, B_FALSE,
4350 		    rollback_destroy_dependent, cbp);
4351 
4352 		cbp->cb_error |= zfs_destroy(zhp, B_FALSE);
4353 	}
4354 
4355 	zfs_close(zhp);
4356 	return (0);
4357 }
4358 
4359 /*
4360  * Given a dataset, rollback to a specific snapshot, discarding any
4361  * data changes since then and making it the active dataset.
4362  *
4363  * Any snapshots and bookmarks more recent than the target are
4364  * destroyed, along with their dependents (i.e. clones).
4365  */
4366 int
4367 zfs_rollback(zfs_handle_t *zhp, zfs_handle_t *snap, boolean_t force)
4368 {
4369 	rollback_data_t cb = { 0 };
4370 	int err;
4371 	boolean_t restore_resv = 0;
4372 	uint64_t old_volsize = 0, new_volsize;
4373 	zfs_prop_t resv_prop = { 0 };
4374 	uint64_t min_txg = 0;
4375 
4376 	assert(zhp->zfs_type == ZFS_TYPE_FILESYSTEM ||
4377 	    zhp->zfs_type == ZFS_TYPE_VOLUME);
4378 
4379 	/*
4380 	 * Destroy all recent snapshots and their dependents.
4381 	 */
4382 	cb.cb_force = force;
4383 	cb.cb_target = snap->zfs_name;
4384 	cb.cb_create = zfs_prop_get_int(snap, ZFS_PROP_CREATETXG);
4385 
4386 	if (cb.cb_create > 0)
4387 		min_txg = cb.cb_create;
4388 
4389 	(void) zfs_iter_snapshots_v2(zhp, 0, rollback_destroy, &cb,
4390 	    min_txg, 0);
4391 
4392 	(void) zfs_iter_bookmarks_v2(zhp, 0, rollback_destroy, &cb);
4393 
4394 	if (cb.cb_error)
4395 		return (-1);
4396 
4397 	/*
4398 	 * Now that we have verified that the snapshot is the latest,
4399 	 * rollback to the given snapshot.
4400 	 */
4401 
4402 	if (zhp->zfs_type == ZFS_TYPE_VOLUME) {
4403 		if (zfs_which_resv_prop(zhp, &resv_prop) < 0)
4404 			return (-1);
4405 		old_volsize = zfs_prop_get_int(zhp, ZFS_PROP_VOLSIZE);
4406 		restore_resv =
4407 		    (old_volsize == zfs_prop_get_int(zhp, resv_prop));
4408 	}
4409 
4410 	/*
4411 	 * Pass both the filesystem and the wanted snapshot names,
4412 	 * we would get an error back if the snapshot is destroyed or
4413 	 * a new snapshot is created before this request is processed.
4414 	 */
4415 	err = lzc_rollback_to(zhp->zfs_name, snap->zfs_name);
4416 	if (err != 0) {
4417 		char errbuf[ERRBUFLEN];
4418 
4419 		(void) snprintf(errbuf, sizeof (errbuf),
4420 		    dgettext(TEXT_DOMAIN, "cannot rollback '%s'"),
4421 		    zhp->zfs_name);
4422 		switch (err) {
4423 		case EEXIST:
4424 			zfs_error_aux(zhp->zfs_hdl, dgettext(TEXT_DOMAIN,
4425 			    "there is a snapshot or bookmark more recent "
4426 			    "than '%s'"), snap->zfs_name);
4427 			(void) zfs_error(zhp->zfs_hdl, EZFS_EXISTS, errbuf);
4428 			break;
4429 		case ESRCH:
4430 			zfs_error_aux(zhp->zfs_hdl, dgettext(TEXT_DOMAIN,
4431 			    "'%s' is not found among snapshots of '%s'"),
4432 			    snap->zfs_name, zhp->zfs_name);
4433 			(void) zfs_error(zhp->zfs_hdl, EZFS_NOENT, errbuf);
4434 			break;
4435 		case EINVAL:
4436 			(void) zfs_error(zhp->zfs_hdl, EZFS_BADTYPE, errbuf);
4437 			break;
4438 		default:
4439 			(void) zfs_standard_error(zhp->zfs_hdl, err, errbuf);
4440 		}
4441 		return (err);
4442 	}
4443 
4444 	/*
4445 	 * For volumes, if the pre-rollback volsize matched the pre-
4446 	 * rollback reservation and the volsize has changed then set
4447 	 * the reservation property to the post-rollback volsize.
4448 	 * Make a new handle since the rollback closed the dataset.
4449 	 */
4450 	if ((zhp->zfs_type == ZFS_TYPE_VOLUME) &&
4451 	    (zhp = make_dataset_handle(zhp->zfs_hdl, zhp->zfs_name))) {
4452 		if (restore_resv) {
4453 			new_volsize = zfs_prop_get_int(zhp, ZFS_PROP_VOLSIZE);
4454 			if (old_volsize != new_volsize)
4455 				err = zfs_prop_set_int(zhp, resv_prop,
4456 				    new_volsize);
4457 		}
4458 		zfs_close(zhp);
4459 	}
4460 	return (err);
4461 }
4462 
4463 /*
4464  * Renames the given dataset.
4465  */
4466 int
4467 zfs_rename(zfs_handle_t *zhp, const char *target, renameflags_t flags)
4468 {
4469 	int ret = 0;
4470 	zfs_cmd_t zc = {"\0"};
4471 	char *delim;
4472 	prop_changelist_t *cl = NULL;
4473 	char parent[ZFS_MAX_DATASET_NAME_LEN];
4474 	char property[ZFS_MAXPROPLEN];
4475 	libzfs_handle_t *hdl = zhp->zfs_hdl;
4476 	char errbuf[ERRBUFLEN];
4477 
4478 	/* if we have the same exact name, just return success */
4479 	if (strcmp(zhp->zfs_name, target) == 0)
4480 		return (0);
4481 
4482 	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
4483 	    "cannot rename to '%s'"), target);
4484 
4485 	/* make sure source name is valid */
4486 	if (!zfs_validate_name(hdl, zhp->zfs_name, zhp->zfs_type, B_TRUE))
4487 		return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
4488 
4489 	/*
4490 	 * Make sure the target name is valid
4491 	 */
4492 	if (zhp->zfs_type == ZFS_TYPE_SNAPSHOT) {
4493 		if ((strchr(target, '@') == NULL) ||
4494 		    *target == '@') {
4495 			/*
4496 			 * Snapshot target name is abbreviated,
4497 			 * reconstruct full dataset name
4498 			 */
4499 			(void) strlcpy(parent, zhp->zfs_name,
4500 			    sizeof (parent));
4501 			delim = strchr(parent, '@');
4502 			if (strchr(target, '@') == NULL)
4503 				*(++delim) = '\0';
4504 			else
4505 				*delim = '\0';
4506 			(void) strlcat(parent, target, sizeof (parent));
4507 			target = parent;
4508 		} else {
4509 			/*
4510 			 * Make sure we're renaming within the same dataset.
4511 			 */
4512 			delim = strchr(target, '@');
4513 			if (strncmp(zhp->zfs_name, target, delim - target)
4514 			    != 0 || zhp->zfs_name[delim - target] != '@') {
4515 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4516 				    "snapshots must be part of same "
4517 				    "dataset"));
4518 				return (zfs_error(hdl, EZFS_CROSSTARGET,
4519 				    errbuf));
4520 			}
4521 		}
4522 
4523 		if (!zfs_validate_name(hdl, target, zhp->zfs_type, B_TRUE))
4524 			return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
4525 	} else {
4526 		if (flags.recursive) {
4527 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4528 			    "recursive rename must be a snapshot"));
4529 			return (zfs_error(hdl, EZFS_BADTYPE, errbuf));
4530 		}
4531 
4532 		if (!zfs_validate_name(hdl, target, zhp->zfs_type, B_TRUE))
4533 			return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
4534 
4535 		/* validate parents */
4536 		if (check_parents(hdl, target, NULL, B_FALSE, NULL) != 0)
4537 			return (-1);
4538 
4539 		/* make sure we're in the same pool */
4540 		verify((delim = strchr(target, '/')) != NULL);
4541 		if (strncmp(zhp->zfs_name, target, delim - target) != 0 ||
4542 		    zhp->zfs_name[delim - target] != '/') {
4543 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4544 			    "datasets must be within same pool"));
4545 			return (zfs_error(hdl, EZFS_CROSSTARGET, errbuf));
4546 		}
4547 
4548 		/* new name cannot be a child of the current dataset name */
4549 		if (is_descendant(zhp->zfs_name, target)) {
4550 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4551 			    "New dataset name cannot be a descendant of "
4552 			    "current dataset name"));
4553 			return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
4554 		}
4555 	}
4556 
4557 	(void) snprintf(errbuf, sizeof (errbuf),
4558 	    dgettext(TEXT_DOMAIN, "cannot rename '%s'"), zhp->zfs_name);
4559 
4560 	if (getzoneid() == GLOBAL_ZONEID &&
4561 	    zfs_prop_get_int(zhp, ZFS_PROP_ZONED)) {
4562 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4563 		    "dataset is used in a non-global zone"));
4564 		return (zfs_error(hdl, EZFS_ZONED, errbuf));
4565 	}
4566 
4567 	/*
4568 	 * Avoid unmounting file systems with mountpoint property set to
4569 	 * 'legacy' or 'none' even if -u option is not given.
4570 	 */
4571 	if (zhp->zfs_type == ZFS_TYPE_FILESYSTEM &&
4572 	    !flags.recursive && !flags.nounmount &&
4573 	    zfs_prop_get(zhp, ZFS_PROP_MOUNTPOINT, property,
4574 	    sizeof (property), NULL, NULL, 0, B_FALSE) == 0 &&
4575 	    (strcmp(property, "legacy") == 0 ||
4576 	    strcmp(property, "none") == 0)) {
4577 		flags.nounmount = B_TRUE;
4578 	}
4579 	if (flags.recursive) {
4580 		char *parentname = zfs_strdup(zhp->zfs_hdl, zhp->zfs_name);
4581 		delim = strchr(parentname, '@');
4582 		*delim = '\0';
4583 		zfs_handle_t *zhrp = zfs_open(zhp->zfs_hdl, parentname,
4584 		    ZFS_TYPE_DATASET);
4585 		free(parentname);
4586 		if (zhrp == NULL) {
4587 			ret = -1;
4588 			goto error;
4589 		}
4590 		zfs_close(zhrp);
4591 	} else if (zhp->zfs_type != ZFS_TYPE_SNAPSHOT) {
4592 		if ((cl = changelist_gather(zhp, ZFS_PROP_NAME,
4593 		    flags.nounmount ? CL_GATHER_DONT_UNMOUNT :
4594 		    CL_GATHER_ITER_MOUNTED,
4595 		    flags.forceunmount ? MS_FORCE : 0)) == NULL)
4596 			return (-1);
4597 
4598 		if (changelist_haszonedchild(cl)) {
4599 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4600 			    "child dataset with inherited mountpoint is used "
4601 			    "in a non-global zone"));
4602 			(void) zfs_error(hdl, EZFS_ZONED, errbuf);
4603 			ret = -1;
4604 			goto error;
4605 		}
4606 
4607 		if ((ret = changelist_prefix(cl)) != 0)
4608 			goto error;
4609 	}
4610 
4611 	if (ZFS_IS_VOLUME(zhp))
4612 		zc.zc_objset_type = DMU_OST_ZVOL;
4613 	else
4614 		zc.zc_objset_type = DMU_OST_ZFS;
4615 
4616 	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
4617 	(void) strlcpy(zc.zc_value, target, sizeof (zc.zc_value));
4618 
4619 	zc.zc_cookie = !!flags.recursive;
4620 	zc.zc_cookie |= (!!flags.nounmount) << 1;
4621 
4622 	if ((ret = zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_RENAME, &zc)) != 0) {
4623 		/*
4624 		 * if it was recursive, the one that actually failed will
4625 		 * be in zc.zc_name
4626 		 */
4627 		(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
4628 		    "cannot rename '%s'"), zc.zc_name);
4629 
4630 		if (flags.recursive && errno == EEXIST) {
4631 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4632 			    "a child dataset already has a snapshot "
4633 			    "with the new name"));
4634 			(void) zfs_error(hdl, EZFS_EXISTS, errbuf);
4635 		} else if (errno == EACCES) {
4636 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4637 			    "cannot move encrypted child outside of "
4638 			    "its encryption root"));
4639 			(void) zfs_error(hdl, EZFS_CRYPTOFAILED, errbuf);
4640 		} else {
4641 			(void) zfs_standard_error(zhp->zfs_hdl, errno, errbuf);
4642 		}
4643 
4644 		/*
4645 		 * On failure, we still want to remount any filesystems that
4646 		 * were previously mounted, so we don't alter the system state.
4647 		 */
4648 		if (cl != NULL)
4649 			(void) changelist_postfix(cl);
4650 	} else {
4651 		if (cl != NULL) {
4652 			changelist_rename(cl, zfs_get_name(zhp), target);
4653 			ret = changelist_postfix(cl);
4654 		}
4655 	}
4656 
4657 error:
4658 	if (cl != NULL) {
4659 		changelist_free(cl);
4660 	}
4661 	return (ret);
4662 }
4663 
4664 nvlist_t *
4665 zfs_get_all_props(zfs_handle_t *zhp)
4666 {
4667 	return (zhp->zfs_props);
4668 }
4669 
4670 nvlist_t *
4671 zfs_get_recvd_props(zfs_handle_t *zhp)
4672 {
4673 	if (zhp->zfs_recvd_props == NULL)
4674 		if (get_recvd_props_ioctl(zhp) != 0)
4675 			return (NULL);
4676 	return (zhp->zfs_recvd_props);
4677 }
4678 
4679 nvlist_t *
4680 zfs_get_user_props(zfs_handle_t *zhp)
4681 {
4682 	return (zhp->zfs_user_props);
4683 }
4684 
4685 /*
4686  * This function is used by 'zfs list' to determine the exact set of columns to
4687  * display, and their maximum widths.  This does two main things:
4688  *
4689  *      - If this is a list of all properties, then expand the list to include
4690  *        all native properties, and set a flag so that for each dataset we look
4691  *        for new unique user properties and add them to the list.
4692  *
4693  *      - For non fixed-width properties, keep track of the maximum width seen
4694  *        so that we can size the column appropriately. If the user has
4695  *        requested received property values, we also need to compute the width
4696  *        of the RECEIVED column.
4697  */
4698 int
4699 zfs_expand_proplist(zfs_handle_t *zhp, zprop_list_t **plp, boolean_t received,
4700     boolean_t literal)
4701 {
4702 	libzfs_handle_t *hdl = zhp->zfs_hdl;
4703 	zprop_list_t *entry;
4704 	zprop_list_t **last, **start;
4705 	nvlist_t *userprops, *propval;
4706 	nvpair_t *elem;
4707 	const char *strval;
4708 	char buf[ZFS_MAXPROPLEN];
4709 
4710 	if (zprop_expand_list(hdl, plp, ZFS_TYPE_DATASET) != 0)
4711 		return (-1);
4712 
4713 	userprops = zfs_get_user_props(zhp);
4714 
4715 	entry = *plp;
4716 	if (entry->pl_all && nvlist_next_nvpair(userprops, NULL) != NULL) {
4717 		/*
4718 		 * Go through and add any user properties as necessary.  We
4719 		 * start by incrementing our list pointer to the first
4720 		 * non-native property.
4721 		 */
4722 		start = plp;
4723 		while (*start != NULL) {
4724 			if ((*start)->pl_prop == ZPROP_USERPROP)
4725 				break;
4726 			start = &(*start)->pl_next;
4727 		}
4728 
4729 		elem = NULL;
4730 		while ((elem = nvlist_next_nvpair(userprops, elem)) != NULL) {
4731 			/*
4732 			 * See if we've already found this property in our list.
4733 			 */
4734 			for (last = start; *last != NULL;
4735 			    last = &(*last)->pl_next) {
4736 				if (strcmp((*last)->pl_user_prop,
4737 				    nvpair_name(elem)) == 0)
4738 					break;
4739 			}
4740 
4741 			if (*last == NULL) {
4742 				entry = zfs_alloc(hdl, sizeof (zprop_list_t));
4743 				entry->pl_user_prop =
4744 				    zfs_strdup(hdl, nvpair_name(elem));
4745 				entry->pl_prop = ZPROP_USERPROP;
4746 				entry->pl_width = strlen(nvpair_name(elem));
4747 				entry->pl_all = B_TRUE;
4748 				*last = entry;
4749 			}
4750 		}
4751 	}
4752 
4753 	/*
4754 	 * Now go through and check the width of any non-fixed columns
4755 	 */
4756 	for (entry = *plp; entry != NULL; entry = entry->pl_next) {
4757 		if (entry->pl_fixed && !literal)
4758 			continue;
4759 
4760 		if (entry->pl_prop != ZPROP_USERPROP) {
4761 			if (zfs_prop_get(zhp, entry->pl_prop,
4762 			    buf, sizeof (buf), NULL, NULL, 0, literal) == 0) {
4763 				if (strlen(buf) > entry->pl_width)
4764 					entry->pl_width = strlen(buf);
4765 			}
4766 			if (received && zfs_prop_get_recvd(zhp,
4767 			    zfs_prop_to_name(entry->pl_prop),
4768 			    buf, sizeof (buf), literal) == 0)
4769 				if (strlen(buf) > entry->pl_recvd_width)
4770 					entry->pl_recvd_width = strlen(buf);
4771 		} else {
4772 			if (nvlist_lookup_nvlist(userprops, entry->pl_user_prop,
4773 			    &propval) == 0) {
4774 				strval = fnvlist_lookup_string(propval,
4775 				    ZPROP_VALUE);
4776 				if (strlen(strval) > entry->pl_width)
4777 					entry->pl_width = strlen(strval);
4778 			}
4779 			if (received && zfs_prop_get_recvd(zhp,
4780 			    entry->pl_user_prop,
4781 			    buf, sizeof (buf), literal) == 0)
4782 				if (strlen(buf) > entry->pl_recvd_width)
4783 					entry->pl_recvd_width = strlen(buf);
4784 		}
4785 	}
4786 
4787 	return (0);
4788 }
4789 
4790 void
4791 zfs_prune_proplist(zfs_handle_t *zhp, uint8_t *props)
4792 {
4793 	nvpair_t *curr;
4794 	nvpair_t *next;
4795 
4796 	/*
4797 	 * Keep a reference to the props-table against which we prune the
4798 	 * properties.
4799 	 */
4800 	zhp->zfs_props_table = props;
4801 
4802 	curr = nvlist_next_nvpair(zhp->zfs_props, NULL);
4803 
4804 	while (curr) {
4805 		zfs_prop_t zfs_prop = zfs_name_to_prop(nvpair_name(curr));
4806 		next = nvlist_next_nvpair(zhp->zfs_props, curr);
4807 
4808 		/*
4809 		 * User properties will result in ZPROP_USERPROP (an alias
4810 		 * for ZPROP_INVAL), and since we
4811 		 * only know how to prune standard ZFS properties, we always
4812 		 * leave these in the list.  This can also happen if we
4813 		 * encounter an unknown DSL property (when running older
4814 		 * software, for example).
4815 		 */
4816 		if (zfs_prop != ZPROP_USERPROP && props[zfs_prop] == B_FALSE)
4817 			(void) nvlist_remove(zhp->zfs_props,
4818 			    nvpair_name(curr), nvpair_type(curr));
4819 		curr = next;
4820 	}
4821 }
4822 
4823 static int
4824 zfs_smb_acl_mgmt(libzfs_handle_t *hdl, char *dataset, char *path,
4825     zfs_smb_acl_op_t cmd, char *resource1, char *resource2)
4826 {
4827 	zfs_cmd_t zc = {"\0"};
4828 	nvlist_t *nvlist = NULL;
4829 	int error;
4830 
4831 	(void) strlcpy(zc.zc_name, dataset, sizeof (zc.zc_name));
4832 	(void) strlcpy(zc.zc_value, path, sizeof (zc.zc_value));
4833 	zc.zc_cookie = (uint64_t)cmd;
4834 
4835 	if (cmd == ZFS_SMB_ACL_RENAME) {
4836 		if (nvlist_alloc(&nvlist, NV_UNIQUE_NAME, 0) != 0) {
4837 			(void) no_memory(hdl);
4838 			return (0);
4839 		}
4840 	}
4841 
4842 	switch (cmd) {
4843 	case ZFS_SMB_ACL_ADD:
4844 	case ZFS_SMB_ACL_REMOVE:
4845 		(void) strlcpy(zc.zc_string, resource1, sizeof (zc.zc_string));
4846 		break;
4847 	case ZFS_SMB_ACL_RENAME:
4848 		if (nvlist_add_string(nvlist, ZFS_SMB_ACL_SRC,
4849 		    resource1) != 0) {
4850 				(void) no_memory(hdl);
4851 				return (-1);
4852 		}
4853 		if (nvlist_add_string(nvlist, ZFS_SMB_ACL_TARGET,
4854 		    resource2) != 0) {
4855 				(void) no_memory(hdl);
4856 				return (-1);
4857 		}
4858 		zcmd_write_src_nvlist(hdl, &zc, nvlist);
4859 		break;
4860 	case ZFS_SMB_ACL_PURGE:
4861 		break;
4862 	default:
4863 		return (-1);
4864 	}
4865 	error = ioctl(hdl->libzfs_fd, ZFS_IOC_SMB_ACL, &zc);
4866 	nvlist_free(nvlist);
4867 	return (error);
4868 }
4869 
4870 int
4871 zfs_smb_acl_add(libzfs_handle_t *hdl, char *dataset,
4872     char *path, char *resource)
4873 {
4874 	return (zfs_smb_acl_mgmt(hdl, dataset, path, ZFS_SMB_ACL_ADD,
4875 	    resource, NULL));
4876 }
4877 
4878 int
4879 zfs_smb_acl_remove(libzfs_handle_t *hdl, char *dataset,
4880     char *path, char *resource)
4881 {
4882 	return (zfs_smb_acl_mgmt(hdl, dataset, path, ZFS_SMB_ACL_REMOVE,
4883 	    resource, NULL));
4884 }
4885 
4886 int
4887 zfs_smb_acl_purge(libzfs_handle_t *hdl, char *dataset, char *path)
4888 {
4889 	return (zfs_smb_acl_mgmt(hdl, dataset, path, ZFS_SMB_ACL_PURGE,
4890 	    NULL, NULL));
4891 }
4892 
4893 int
4894 zfs_smb_acl_rename(libzfs_handle_t *hdl, char *dataset, char *path,
4895     char *oldname, char *newname)
4896 {
4897 	return (zfs_smb_acl_mgmt(hdl, dataset, path, ZFS_SMB_ACL_RENAME,
4898 	    oldname, newname));
4899 }
4900 
4901 int
4902 zfs_userspace(zfs_handle_t *zhp, zfs_userquota_prop_t type,
4903     zfs_userspace_cb_t func, void *arg)
4904 {
4905 	zfs_cmd_t zc = {"\0"};
4906 	zfs_useracct_t buf[100];
4907 	libzfs_handle_t *hdl = zhp->zfs_hdl;
4908 	int ret;
4909 
4910 	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
4911 
4912 	zc.zc_objset_type = type;
4913 	zc.zc_nvlist_dst = (uintptr_t)buf;
4914 
4915 	for (;;) {
4916 		zfs_useracct_t *zua = buf;
4917 
4918 		zc.zc_nvlist_dst_size = sizeof (buf);
4919 		if (zfs_ioctl(hdl, ZFS_IOC_USERSPACE_MANY, &zc) != 0) {
4920 			if ((errno == ENOTSUP &&
4921 			    (type == ZFS_PROP_USEROBJUSED ||
4922 			    type == ZFS_PROP_GROUPOBJUSED ||
4923 			    type == ZFS_PROP_USEROBJQUOTA ||
4924 			    type == ZFS_PROP_GROUPOBJQUOTA ||
4925 			    type == ZFS_PROP_PROJECTOBJUSED ||
4926 			    type == ZFS_PROP_PROJECTOBJQUOTA ||
4927 			    type == ZFS_PROP_PROJECTUSED ||
4928 			    type == ZFS_PROP_PROJECTQUOTA)))
4929 				break;
4930 
4931 			return (zfs_standard_error_fmt(hdl, errno,
4932 			    dgettext(TEXT_DOMAIN,
4933 			    "cannot get used/quota for %s"), zc.zc_name));
4934 		}
4935 		if (zc.zc_nvlist_dst_size == 0)
4936 			break;
4937 
4938 		while (zc.zc_nvlist_dst_size > 0) {
4939 			if ((ret = func(arg, zua->zu_domain, zua->zu_rid,
4940 			    zua->zu_space)) != 0)
4941 				return (ret);
4942 			zua++;
4943 			zc.zc_nvlist_dst_size -= sizeof (zfs_useracct_t);
4944 		}
4945 	}
4946 
4947 	return (0);
4948 }
4949 
4950 struct holdarg {
4951 	nvlist_t *nvl;
4952 	const char *snapname;
4953 	const char *tag;
4954 	boolean_t recursive;
4955 	int error;
4956 };
4957 
4958 static int
4959 zfs_hold_one(zfs_handle_t *zhp, void *arg)
4960 {
4961 	struct holdarg *ha = arg;
4962 	char name[ZFS_MAX_DATASET_NAME_LEN];
4963 	int rv = 0;
4964 
4965 	if (snprintf(name, sizeof (name), "%s@%s", zhp->zfs_name,
4966 	    ha->snapname) >= sizeof (name))
4967 		return (EINVAL);
4968 
4969 	if (lzc_exists(name))
4970 		fnvlist_add_string(ha->nvl, name, ha->tag);
4971 
4972 	if (ha->recursive)
4973 		rv = zfs_iter_filesystems_v2(zhp, 0, zfs_hold_one, ha);
4974 	zfs_close(zhp);
4975 	return (rv);
4976 }
4977 
4978 int
4979 zfs_hold(zfs_handle_t *zhp, const char *snapname, const char *tag,
4980     boolean_t recursive, int cleanup_fd)
4981 {
4982 	int ret;
4983 	struct holdarg ha;
4984 
4985 	ha.nvl = fnvlist_alloc();
4986 	ha.snapname = snapname;
4987 	ha.tag = tag;
4988 	ha.recursive = recursive;
4989 	(void) zfs_hold_one(zfs_handle_dup(zhp), &ha);
4990 
4991 	if (nvlist_empty(ha.nvl)) {
4992 		char errbuf[ERRBUFLEN];
4993 
4994 		fnvlist_free(ha.nvl);
4995 		ret = ENOENT;
4996 		(void) snprintf(errbuf, sizeof (errbuf),
4997 		    dgettext(TEXT_DOMAIN,
4998 		    "cannot hold snapshot '%s@%s'"),
4999 		    zhp->zfs_name, snapname);
5000 		(void) zfs_standard_error(zhp->zfs_hdl, ret, errbuf);
5001 		return (ret);
5002 	}
5003 
5004 	ret = zfs_hold_nvl(zhp, cleanup_fd, ha.nvl);
5005 	fnvlist_free(ha.nvl);
5006 
5007 	return (ret);
5008 }
5009 
5010 int
5011 zfs_hold_nvl(zfs_handle_t *zhp, int cleanup_fd, nvlist_t *holds)
5012 {
5013 	int ret;
5014 	nvlist_t *errors;
5015 	libzfs_handle_t *hdl = zhp->zfs_hdl;
5016 	char errbuf[ERRBUFLEN];
5017 	nvpair_t *elem;
5018 
5019 	errors = NULL;
5020 	ret = lzc_hold(holds, cleanup_fd, &errors);
5021 
5022 	if (ret == 0) {
5023 		/* There may be errors even in the success case. */
5024 		fnvlist_free(errors);
5025 		return (0);
5026 	}
5027 
5028 	if (nvlist_empty(errors)) {
5029 		/* no hold-specific errors */
5030 		(void) snprintf(errbuf, sizeof (errbuf),
5031 		    dgettext(TEXT_DOMAIN, "cannot hold"));
5032 		switch (ret) {
5033 		case ENOTSUP:
5034 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
5035 			    "pool must be upgraded"));
5036 			(void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
5037 			break;
5038 		case EINVAL:
5039 			(void) zfs_error(hdl, EZFS_BADTYPE, errbuf);
5040 			break;
5041 		default:
5042 			(void) zfs_standard_error(hdl, ret, errbuf);
5043 		}
5044 	}
5045 
5046 	for (elem = nvlist_next_nvpair(errors, NULL);
5047 	    elem != NULL;
5048 	    elem = nvlist_next_nvpair(errors, elem)) {
5049 		(void) snprintf(errbuf, sizeof (errbuf),
5050 		    dgettext(TEXT_DOMAIN,
5051 		    "cannot hold snapshot '%s'"), nvpair_name(elem));
5052 		switch (fnvpair_value_int32(elem)) {
5053 		case E2BIG:
5054 			/*
5055 			 * Temporary tags wind up having the ds object id
5056 			 * prepended. So even if we passed the length check
5057 			 * above, it's still possible for the tag to wind
5058 			 * up being slightly too long.
5059 			 */
5060 			(void) zfs_error(hdl, EZFS_TAGTOOLONG, errbuf);
5061 			break;
5062 		case EINVAL:
5063 			(void) zfs_error(hdl, EZFS_BADTYPE, errbuf);
5064 			break;
5065 		case EEXIST:
5066 			(void) zfs_error(hdl, EZFS_REFTAG_HOLD, errbuf);
5067 			break;
5068 		default:
5069 			(void) zfs_standard_error(hdl,
5070 			    fnvpair_value_int32(elem), errbuf);
5071 		}
5072 	}
5073 
5074 	fnvlist_free(errors);
5075 	return (ret);
5076 }
5077 
5078 static int
5079 zfs_release_one(zfs_handle_t *zhp, void *arg)
5080 {
5081 	struct holdarg *ha = arg;
5082 	char name[ZFS_MAX_DATASET_NAME_LEN];
5083 	int rv = 0;
5084 	nvlist_t *existing_holds;
5085 
5086 	if (snprintf(name, sizeof (name), "%s@%s", zhp->zfs_name,
5087 	    ha->snapname) >= sizeof (name)) {
5088 		ha->error = EINVAL;
5089 		rv = EINVAL;
5090 	}
5091 
5092 	if (lzc_get_holds(name, &existing_holds) != 0) {
5093 		ha->error = ENOENT;
5094 	} else if (!nvlist_exists(existing_holds, ha->tag)) {
5095 		ha->error = ESRCH;
5096 	} else {
5097 		nvlist_t *torelease = fnvlist_alloc();
5098 		fnvlist_add_boolean(torelease, ha->tag);
5099 		fnvlist_add_nvlist(ha->nvl, name, torelease);
5100 		fnvlist_free(torelease);
5101 	}
5102 
5103 	if (ha->recursive)
5104 		rv = zfs_iter_filesystems_v2(zhp, 0, zfs_release_one, ha);
5105 	zfs_close(zhp);
5106 	return (rv);
5107 }
5108 
5109 int
5110 zfs_release(zfs_handle_t *zhp, const char *snapname, const char *tag,
5111     boolean_t recursive)
5112 {
5113 	int ret;
5114 	struct holdarg ha;
5115 	nvlist_t *errors = NULL;
5116 	nvpair_t *elem;
5117 	libzfs_handle_t *hdl = zhp->zfs_hdl;
5118 	char errbuf[ERRBUFLEN];
5119 
5120 	ha.nvl = fnvlist_alloc();
5121 	ha.snapname = snapname;
5122 	ha.tag = tag;
5123 	ha.recursive = recursive;
5124 	ha.error = 0;
5125 	(void) zfs_release_one(zfs_handle_dup(zhp), &ha);
5126 
5127 	if (nvlist_empty(ha.nvl)) {
5128 		fnvlist_free(ha.nvl);
5129 		ret = ha.error;
5130 		(void) snprintf(errbuf, sizeof (errbuf),
5131 		    dgettext(TEXT_DOMAIN,
5132 		    "cannot release hold from snapshot '%s@%s'"),
5133 		    zhp->zfs_name, snapname);
5134 		if (ret == ESRCH) {
5135 			(void) zfs_error(hdl, EZFS_REFTAG_RELE, errbuf);
5136 		} else {
5137 			(void) zfs_standard_error(hdl, ret, errbuf);
5138 		}
5139 		return (ret);
5140 	}
5141 
5142 	ret = lzc_release(ha.nvl, &errors);
5143 	fnvlist_free(ha.nvl);
5144 
5145 	if (ret == 0) {
5146 		/* There may be errors even in the success case. */
5147 		fnvlist_free(errors);
5148 		return (0);
5149 	}
5150 
5151 	if (nvlist_empty(errors)) {
5152 		/* no hold-specific errors */
5153 		(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
5154 		    "cannot release"));
5155 		switch (errno) {
5156 		case ENOTSUP:
5157 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
5158 			    "pool must be upgraded"));
5159 			(void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
5160 			break;
5161 		default:
5162 			(void) zfs_standard_error(hdl, errno, errbuf);
5163 		}
5164 	}
5165 
5166 	for (elem = nvlist_next_nvpair(errors, NULL);
5167 	    elem != NULL;
5168 	    elem = nvlist_next_nvpair(errors, elem)) {
5169 		(void) snprintf(errbuf, sizeof (errbuf),
5170 		    dgettext(TEXT_DOMAIN,
5171 		    "cannot release hold from snapshot '%s'"),
5172 		    nvpair_name(elem));
5173 		switch (fnvpair_value_int32(elem)) {
5174 		case ESRCH:
5175 			(void) zfs_error(hdl, EZFS_REFTAG_RELE, errbuf);
5176 			break;
5177 		case EINVAL:
5178 			(void) zfs_error(hdl, EZFS_BADTYPE, errbuf);
5179 			break;
5180 		default:
5181 			(void) zfs_standard_error(hdl,
5182 			    fnvpair_value_int32(elem), errbuf);
5183 		}
5184 	}
5185 
5186 	fnvlist_free(errors);
5187 	return (ret);
5188 }
5189 
5190 int
5191 zfs_get_fsacl(zfs_handle_t *zhp, nvlist_t **nvl)
5192 {
5193 	zfs_cmd_t zc = {"\0"};
5194 	libzfs_handle_t *hdl = zhp->zfs_hdl;
5195 	int nvsz = 2048;
5196 	void *nvbuf;
5197 	int err = 0;
5198 	char errbuf[ERRBUFLEN];
5199 
5200 	assert(zhp->zfs_type == ZFS_TYPE_VOLUME ||
5201 	    zhp->zfs_type == ZFS_TYPE_FILESYSTEM);
5202 
5203 tryagain:
5204 
5205 	nvbuf = malloc(nvsz);
5206 	if (nvbuf == NULL) {
5207 		err = (zfs_error(hdl, EZFS_NOMEM, strerror(errno)));
5208 		goto out;
5209 	}
5210 
5211 	zc.zc_nvlist_dst_size = nvsz;
5212 	zc.zc_nvlist_dst = (uintptr_t)nvbuf;
5213 
5214 	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
5215 
5216 	if (zfs_ioctl(hdl, ZFS_IOC_GET_FSACL, &zc) != 0) {
5217 		(void) snprintf(errbuf, sizeof (errbuf),
5218 		    dgettext(TEXT_DOMAIN, "cannot get permissions on '%s'"),
5219 		    zc.zc_name);
5220 		switch (errno) {
5221 		case ENOMEM:
5222 			free(nvbuf);
5223 			nvsz = zc.zc_nvlist_dst_size;
5224 			goto tryagain;
5225 
5226 		case ENOTSUP:
5227 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
5228 			    "pool must be upgraded"));
5229 			err = zfs_error(hdl, EZFS_BADVERSION, errbuf);
5230 			break;
5231 		case EINVAL:
5232 			err = zfs_error(hdl, EZFS_BADTYPE, errbuf);
5233 			break;
5234 		case ENOENT:
5235 			err = zfs_error(hdl, EZFS_NOENT, errbuf);
5236 			break;
5237 		default:
5238 			err = zfs_standard_error(hdl, errno, errbuf);
5239 			break;
5240 		}
5241 	} else {
5242 		/* success */
5243 		int rc = nvlist_unpack(nvbuf, zc.zc_nvlist_dst_size, nvl, 0);
5244 		if (rc) {
5245 			err = zfs_standard_error_fmt(hdl, rc, dgettext(
5246 			    TEXT_DOMAIN, "cannot get permissions on '%s'"),
5247 			    zc.zc_name);
5248 		}
5249 	}
5250 
5251 	free(nvbuf);
5252 out:
5253 	return (err);
5254 }
5255 
5256 int
5257 zfs_set_fsacl(zfs_handle_t *zhp, boolean_t un, nvlist_t *nvl)
5258 {
5259 	zfs_cmd_t zc = {"\0"};
5260 	libzfs_handle_t *hdl = zhp->zfs_hdl;
5261 	char *nvbuf;
5262 	char errbuf[ERRBUFLEN];
5263 	size_t nvsz;
5264 	int err;
5265 
5266 	assert(zhp->zfs_type == ZFS_TYPE_VOLUME ||
5267 	    zhp->zfs_type == ZFS_TYPE_FILESYSTEM);
5268 
5269 	err = nvlist_size(nvl, &nvsz, NV_ENCODE_NATIVE);
5270 	assert(err == 0);
5271 
5272 	nvbuf = malloc(nvsz);
5273 
5274 	err = nvlist_pack(nvl, &nvbuf, &nvsz, NV_ENCODE_NATIVE, 0);
5275 	assert(err == 0);
5276 
5277 	zc.zc_nvlist_src_size = nvsz;
5278 	zc.zc_nvlist_src = (uintptr_t)nvbuf;
5279 	zc.zc_perm_action = un;
5280 
5281 	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
5282 
5283 	if (zfs_ioctl(hdl, ZFS_IOC_SET_FSACL, &zc) != 0) {
5284 		(void) snprintf(errbuf, sizeof (errbuf),
5285 		    dgettext(TEXT_DOMAIN, "cannot set permissions on '%s'"),
5286 		    zc.zc_name);
5287 		switch (errno) {
5288 		case ENOTSUP:
5289 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
5290 			    "pool must be upgraded"));
5291 			err = zfs_error(hdl, EZFS_BADVERSION, errbuf);
5292 			break;
5293 		case EINVAL:
5294 			err = zfs_error(hdl, EZFS_BADTYPE, errbuf);
5295 			break;
5296 		case ENOENT:
5297 			err = zfs_error(hdl, EZFS_NOENT, errbuf);
5298 			break;
5299 		default:
5300 			err = zfs_standard_error(hdl, errno, errbuf);
5301 			break;
5302 		}
5303 	}
5304 
5305 	free(nvbuf);
5306 
5307 	return (err);
5308 }
5309 
5310 int
5311 zfs_get_holds(zfs_handle_t *zhp, nvlist_t **nvl)
5312 {
5313 	int err;
5314 	char errbuf[ERRBUFLEN];
5315 
5316 	err = lzc_get_holds(zhp->zfs_name, nvl);
5317 
5318 	if (err != 0) {
5319 		libzfs_handle_t *hdl = zhp->zfs_hdl;
5320 
5321 		(void) snprintf(errbuf, sizeof (errbuf),
5322 		    dgettext(TEXT_DOMAIN, "cannot get holds for '%s'"),
5323 		    zhp->zfs_name);
5324 		switch (err) {
5325 		case ENOTSUP:
5326 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
5327 			    "pool must be upgraded"));
5328 			err = zfs_error(hdl, EZFS_BADVERSION, errbuf);
5329 			break;
5330 		case EINVAL:
5331 			err = zfs_error(hdl, EZFS_BADTYPE, errbuf);
5332 			break;
5333 		case ENOENT:
5334 			err = zfs_error(hdl, EZFS_NOENT, errbuf);
5335 			break;
5336 		default:
5337 			err = zfs_standard_error(hdl, errno, errbuf);
5338 			break;
5339 		}
5340 	}
5341 
5342 	return (err);
5343 }
5344 
5345 /*
5346  * The theory of raidz space accounting
5347  *
5348  * The "referenced" property of RAIDZ vdevs is scaled such that a 128KB block
5349  * will "reference" 128KB, even though it allocates more than that, to store the
5350  * parity information (and perhaps skip sectors). This concept of the
5351  * "referenced" (and other DMU space accounting) being lower than the allocated
5352  * space by a constant factor is called "raidz deflation."
5353  *
5354  * As mentioned above, the constant factor for raidz deflation assumes a 128KB
5355  * block size. However, zvols typically have a much smaller block size (default
5356  * 8KB). These smaller blocks may require proportionally much more parity
5357  * information (and perhaps skip sectors). In this case, the change to the
5358  * "referenced" property may be much more than the logical block size.
5359  *
5360  * Suppose a raidz vdev has 5 disks with ashift=12.  A 128k block may be written
5361  * as follows.
5362  *
5363  * +-------+-------+-------+-------+-------+
5364  * | disk1 | disk2 | disk3 | disk4 | disk5 |
5365  * +-------+-------+-------+-------+-------+
5366  * |  P0   |  D0   |  D8   |  D16  |  D24  |
5367  * |  P1   |  D1   |  D9   |  D17  |  D25  |
5368  * |  P2   |  D2   |  D10  |  D18  |  D26  |
5369  * |  P3   |  D3   |  D11  |  D19  |  D27  |
5370  * |  P4   |  D4   |  D12  |  D20  |  D28  |
5371  * |  P5   |  D5   |  D13  |  D21  |  D29  |
5372  * |  P6   |  D6   |  D14  |  D22  |  D30  |
5373  * |  P7   |  D7   |  D15  |  D23  |  D31  |
5374  * +-------+-------+-------+-------+-------+
5375  *
5376  * Above, notice that 160k was allocated: 8 x 4k parity sectors + 32 x 4k data
5377  * sectors.  The dataset's referenced will increase by 128k and the pool's
5378  * allocated and free properties will be adjusted by 160k.
5379  *
5380  * A 4k block written to the same raidz vdev will require two 4k sectors.  The
5381  * blank cells represent unallocated space.
5382  *
5383  * +-------+-------+-------+-------+-------+
5384  * | disk1 | disk2 | disk3 | disk4 | disk5 |
5385  * +-------+-------+-------+-------+-------+
5386  * |  P0   |  D0   |       |       |       |
5387  * +-------+-------+-------+-------+-------+
5388  *
5389  * Above, notice that the 4k block required one sector for parity and another
5390  * for data.  vdev_raidz_asize() will return 8k and as such the pool's allocated
5391  * and free properties will be adjusted by 8k.  The dataset will not be charged
5392  * 8k.  Rather, it will be charged a value that is scaled according to the
5393  * overhead of the 128k block on the same vdev.  This 8k allocation will be
5394  * charged 8k * 128k / 160k.  128k is from SPA_OLD_MAXBLOCKSIZE and 160k is as
5395  * calculated in the 128k block example above.
5396  *
5397  * Every raidz allocation is sized to be a multiple of nparity+1 sectors.  That
5398  * is, every raidz1 allocation will be a multiple of 2 sectors, raidz2
5399  * allocations are a multiple of 3 sectors, and raidz3 allocations are a
5400  * multiple of of 4 sectors.  When a block does not fill the required number of
5401  * sectors, skip blocks (sectors) are used.
5402  *
5403  * An 8k block being written to a raidz vdev may be written as follows:
5404  *
5405  * +-------+-------+-------+-------+-------+
5406  * | disk1 | disk2 | disk3 | disk4 | disk5 |
5407  * +-------+-------+-------+-------+-------+
5408  * |  P0   |  D0   |  D1   |  S0   |       |
5409  * +-------+-------+-------+-------+-------+
5410  *
5411  * In order to maintain the nparity+1 allocation size, a skip block (S0) was
5412  * added.  For this 8k block, the pool's allocated and free properties are
5413  * adjusted by 16k and the dataset's referenced is increased by 16k * 128k /
5414  * 160k.  Again, 128k is from SPA_OLD_MAXBLOCKSIZE and 160k is as calculated in
5415  * the 128k block example above.
5416  *
5417  * The situation is slightly different for dRAID since the minimum allocation
5418  * size is the full group width.  The same 8K block above would be written as
5419  * follows in a dRAID group:
5420  *
5421  * +-------+-------+-------+-------+-------+
5422  * | disk1 | disk2 | disk3 | disk4 | disk5 |
5423  * +-------+-------+-------+-------+-------+
5424  * |  P0   |  D0   |  D1   |  S0   |  S1   |
5425  * +-------+-------+-------+-------+-------+
5426  *
5427  * Compression may lead to a variety of block sizes being written for the same
5428  * volume or file.  There is no clear way to reserve just the amount of space
5429  * that will be required, so the worst case (no compression) is assumed.
5430  * Note that metadata blocks will typically be compressed, so the reservation
5431  * size returned by zvol_volsize_to_reservation() will generally be slightly
5432  * larger than the maximum that the volume can reference.
5433  */
5434 
5435 /*
5436  * Derived from function of same name in module/zfs/vdev_raidz.c.  Returns the
5437  * amount of space (in bytes) that will be allocated for the specified block
5438  * size. Note that the "referenced" space accounted will be less than this, but
5439  * not necessarily equal to "blksize", due to RAIDZ deflation.
5440  */
5441 static uint64_t
5442 vdev_raidz_asize(uint64_t ndisks, uint64_t nparity, uint64_t ashift,
5443     uint64_t blksize)
5444 {
5445 	uint64_t asize, ndata;
5446 
5447 	ASSERT3U(ndisks, >, nparity);
5448 	ndata = ndisks - nparity;
5449 	asize = ((blksize - 1) >> ashift) + 1;
5450 	asize += nparity * ((asize + ndata - 1) / ndata);
5451 	asize = roundup(asize, nparity + 1) << ashift;
5452 
5453 	return (asize);
5454 }
5455 
5456 /*
5457  * Derived from function of same name in module/zfs/vdev_draid.c.  Returns the
5458  * amount of space (in bytes) that will be allocated for the specified block
5459  * size.
5460  */
5461 static uint64_t
5462 vdev_draid_asize(uint64_t ndisks, uint64_t nparity, uint64_t ashift,
5463     uint64_t blksize)
5464 {
5465 	ASSERT3U(ndisks, >, nparity);
5466 	uint64_t ndata = ndisks - nparity;
5467 	uint64_t rows = ((blksize - 1) / (ndata << ashift)) + 1;
5468 	uint64_t asize = (rows * ndisks) << ashift;
5469 
5470 	return (asize);
5471 }
5472 
5473 /*
5474  * Determine how much space will be allocated if it lands on the most space-
5475  * inefficient top-level vdev.  Returns the size in bytes required to store one
5476  * copy of the volume data.  See theory comment above.
5477  */
5478 static uint64_t
5479 volsize_from_vdevs(zpool_handle_t *zhp, uint64_t nblocks, uint64_t blksize)
5480 {
5481 	nvlist_t *config, *tree, **vdevs;
5482 	uint_t nvdevs;
5483 	uint64_t ret = 0;
5484 
5485 	config = zpool_get_config(zhp, NULL);
5486 	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree) != 0 ||
5487 	    nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN,
5488 	    &vdevs, &nvdevs) != 0) {
5489 		return (nblocks * blksize);
5490 	}
5491 
5492 	for (int v = 0; v < nvdevs; v++) {
5493 		const char *type;
5494 		uint64_t nparity, ashift, asize, tsize;
5495 		uint64_t volsize;
5496 
5497 		if (nvlist_lookup_string(vdevs[v], ZPOOL_CONFIG_TYPE,
5498 		    &type) != 0)
5499 			continue;
5500 
5501 		if (strcmp(type, VDEV_TYPE_RAIDZ) != 0 &&
5502 		    strcmp(type, VDEV_TYPE_DRAID) != 0)
5503 			continue;
5504 
5505 		if (nvlist_lookup_uint64(vdevs[v],
5506 		    ZPOOL_CONFIG_NPARITY, &nparity) != 0)
5507 			continue;
5508 
5509 		if (nvlist_lookup_uint64(vdevs[v],
5510 		    ZPOOL_CONFIG_ASHIFT, &ashift) != 0)
5511 			continue;
5512 
5513 		if (strcmp(type, VDEV_TYPE_RAIDZ) == 0) {
5514 			nvlist_t **disks;
5515 			uint_t ndisks;
5516 
5517 			if (nvlist_lookup_nvlist_array(vdevs[v],
5518 			    ZPOOL_CONFIG_CHILDREN, &disks, &ndisks) != 0)
5519 				continue;
5520 
5521 			/* allocation size for the "typical" 128k block */
5522 			tsize = vdev_raidz_asize(ndisks, nparity, ashift,
5523 			    SPA_OLD_MAXBLOCKSIZE);
5524 
5525 			/* allocation size for the blksize block */
5526 			asize = vdev_raidz_asize(ndisks, nparity, ashift,
5527 			    blksize);
5528 		} else {
5529 			uint64_t ndata;
5530 
5531 			if (nvlist_lookup_uint64(vdevs[v],
5532 			    ZPOOL_CONFIG_DRAID_NDATA, &ndata) != 0)
5533 				continue;
5534 
5535 			/* allocation size for the "typical" 128k block */
5536 			tsize = vdev_draid_asize(ndata + nparity, nparity,
5537 			    ashift, SPA_OLD_MAXBLOCKSIZE);
5538 
5539 			/* allocation size for the blksize block */
5540 			asize = vdev_draid_asize(ndata + nparity, nparity,
5541 			    ashift, blksize);
5542 		}
5543 
5544 		/*
5545 		 * Scale this size down as a ratio of 128k / tsize.
5546 		 * See theory statement above.
5547 		 */
5548 		volsize = nblocks * asize * SPA_OLD_MAXBLOCKSIZE / tsize;
5549 		if (volsize > ret) {
5550 			ret = volsize;
5551 		}
5552 	}
5553 
5554 	if (ret == 0) {
5555 		ret = nblocks * blksize;
5556 	}
5557 
5558 	return (ret);
5559 }
5560 
5561 /*
5562  * Convert the zvol's volume size to an appropriate reservation.  See theory
5563  * comment above.
5564  *
5565  * Note: If this routine is updated, it is necessary to update the ZFS test
5566  * suite's shell version in reservation.shlib.
5567  */
5568 uint64_t
5569 zvol_volsize_to_reservation(zpool_handle_t *zph, uint64_t volsize,
5570     nvlist_t *props)
5571 {
5572 	uint64_t numdb;
5573 	uint64_t nblocks, volblocksize;
5574 	int ncopies;
5575 	const char *strval;
5576 
5577 	if (nvlist_lookup_string(props,
5578 	    zfs_prop_to_name(ZFS_PROP_COPIES), &strval) == 0)
5579 		ncopies = atoi(strval);
5580 	else
5581 		ncopies = 1;
5582 	if (nvlist_lookup_uint64(props,
5583 	    zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
5584 	    &volblocksize) != 0)
5585 		volblocksize = ZVOL_DEFAULT_BLOCKSIZE;
5586 
5587 	nblocks = volsize / volblocksize;
5588 	/*
5589 	 * Metadata defaults to using 128k blocks, not volblocksize blocks.  For
5590 	 * this reason, only the data blocks are scaled based on vdev config.
5591 	 */
5592 	volsize = volsize_from_vdevs(zph, nblocks, volblocksize);
5593 
5594 	/* start with metadnode L0-L6 */
5595 	numdb = 7;
5596 	/* calculate number of indirects */
5597 	while (nblocks > 1) {
5598 		nblocks += DNODES_PER_LEVEL - 1;
5599 		nblocks /= DNODES_PER_LEVEL;
5600 		numdb += nblocks;
5601 	}
5602 	numdb *= MIN(SPA_DVAS_PER_BP, ncopies + 1);
5603 	volsize *= ncopies;
5604 	/*
5605 	 * this is exactly DN_MAX_INDBLKSHIFT when metadata isn't
5606 	 * compressed, but in practice they compress down to about
5607 	 * 1100 bytes
5608 	 */
5609 	numdb *= 1ULL << DN_MAX_INDBLKSHIFT;
5610 	volsize += numdb;
5611 	return (volsize);
5612 }
5613 
5614 /*
5615  * Wait for the given activity and return the status of the wait (whether or not
5616  * any waiting was done) in the 'waited' parameter. Non-existent fses are
5617  * reported via the 'missing' parameter, rather than by printing an error
5618  * message. This is convenient when this function is called in a loop over a
5619  * long period of time (as it is, for example, by zfs's wait cmd). In that
5620  * scenario, a fs being exported or destroyed should be considered a normal
5621  * event, so we don't want to print an error when we find that the fs doesn't
5622  * exist.
5623  */
5624 int
5625 zfs_wait_status(zfs_handle_t *zhp, zfs_wait_activity_t activity,
5626     boolean_t *missing, boolean_t *waited)
5627 {
5628 	int error = lzc_wait_fs(zhp->zfs_name, activity, waited);
5629 	*missing = (error == ENOENT);
5630 	if (*missing)
5631 		return (0);
5632 
5633 	if (error != 0) {
5634 		(void) zfs_standard_error_fmt(zhp->zfs_hdl, error,
5635 		    dgettext(TEXT_DOMAIN, "error waiting in fs '%s'"),
5636 		    zhp->zfs_name);
5637 	}
5638 
5639 	return (error);
5640 }
5641