xref: /freebsd/sys/contrib/openzfs/lib/libzfs/libzfs_dataset.c (revision f5f40dd63bc7acbb5312b26ac1ea1103c12352a6)
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 __thread struct passwd gpwd;
71 static __thread struct group ggrp;
72 static __thread char rpbuf[2048];
73 
74 static int userquota_propname_decode(const char *propname, boolean_t zoned,
75     zfs_userquota_prop_t *typep, char *domain, int domainlen, uint64_t *ridp);
76 
77 /*
78  * Given a single type (not a mask of types), return the type in a human
79  * readable form.
80  */
81 const char *
82 zfs_type_to_name(zfs_type_t type)
83 {
84 	switch (type) {
85 	case ZFS_TYPE_FILESYSTEM:
86 		return (dgettext(TEXT_DOMAIN, "filesystem"));
87 	case ZFS_TYPE_SNAPSHOT:
88 		return (dgettext(TEXT_DOMAIN, "snapshot"));
89 	case ZFS_TYPE_VOLUME:
90 		return (dgettext(TEXT_DOMAIN, "volume"));
91 	case ZFS_TYPE_POOL:
92 		return (dgettext(TEXT_DOMAIN, "pool"));
93 	case ZFS_TYPE_BOOKMARK:
94 		return (dgettext(TEXT_DOMAIN, "bookmark"));
95 	default:
96 		assert(!"unhandled zfs_type_t");
97 	}
98 
99 	return (NULL);
100 }
101 
102 /*
103  * Validate a ZFS path.  This is used even before trying to open the dataset, to
104  * provide a more meaningful error message.  We call zfs_error_aux() to
105  * explain exactly why the name was not valid.
106  */
107 int
108 zfs_validate_name(libzfs_handle_t *hdl, const char *path, int type,
109     boolean_t modifying)
110 {
111 	namecheck_err_t why;
112 	char what;
113 
114 	if (!(type & ZFS_TYPE_SNAPSHOT) && strchr(path, '@') != NULL) {
115 		if (hdl != NULL)
116 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
117 			    "snapshot delimiter '@' is not expected here"));
118 		return (0);
119 	}
120 
121 	if (type == ZFS_TYPE_SNAPSHOT && strchr(path, '@') == NULL) {
122 		if (hdl != NULL)
123 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
124 			    "missing '@' delimiter in snapshot name"));
125 		return (0);
126 	}
127 
128 	if (!(type & ZFS_TYPE_BOOKMARK) && strchr(path, '#') != NULL) {
129 		if (hdl != NULL)
130 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
131 			    "bookmark delimiter '#' is not expected here"));
132 		return (0);
133 	}
134 
135 	if (type == ZFS_TYPE_BOOKMARK && strchr(path, '#') == NULL) {
136 		if (hdl != NULL)
137 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
138 			    "missing '#' delimiter in bookmark name"));
139 		return (0);
140 	}
141 
142 	if (modifying && strchr(path, '%') != NULL) {
143 		if (hdl != NULL)
144 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
145 			    "invalid character %c in name"), '%');
146 		return (0);
147 	}
148 
149 	if (entity_namecheck(path, &why, &what) != 0) {
150 		if (hdl != NULL) {
151 			switch (why) {
152 			case NAME_ERR_TOOLONG:
153 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
154 				    "name is too long"));
155 				break;
156 
157 			case NAME_ERR_LEADING_SLASH:
158 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
159 				    "leading slash in name"));
160 				break;
161 
162 			case NAME_ERR_EMPTY_COMPONENT:
163 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
164 				    "empty component or misplaced '@'"
165 				    " or '#' delimiter in name"));
166 				break;
167 
168 			case NAME_ERR_TRAILING_SLASH:
169 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
170 				    "trailing slash in name"));
171 				break;
172 
173 			case NAME_ERR_INVALCHAR:
174 				zfs_error_aux(hdl,
175 				    dgettext(TEXT_DOMAIN, "invalid character "
176 				    "'%c' in name"), what);
177 				break;
178 
179 			case NAME_ERR_MULTIPLE_DELIMITERS:
180 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
181 				    "multiple '@' and/or '#' delimiters in "
182 				    "name"));
183 				break;
184 
185 			case NAME_ERR_NOLETTER:
186 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
187 				    "pool doesn't begin with a letter"));
188 				break;
189 
190 			case NAME_ERR_RESERVED:
191 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
192 				    "name is reserved"));
193 				break;
194 
195 			case NAME_ERR_DISKLIKE:
196 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
197 				    "reserved disk name"));
198 				break;
199 
200 			case NAME_ERR_SELF_REF:
201 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
202 				    "self reference, '.' is found in name"));
203 				break;
204 
205 			case NAME_ERR_PARENT_REF:
206 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
207 				    "parent reference, '..' is found in name"));
208 				break;
209 
210 			default:
211 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
212 				    "(%d) not defined"), why);
213 				break;
214 			}
215 		}
216 
217 		return (0);
218 	}
219 
220 	return (-1);
221 }
222 
223 int
224 zfs_name_valid(const char *name, zfs_type_t type)
225 {
226 	if (type == ZFS_TYPE_POOL)
227 		return (zpool_name_valid(NULL, B_FALSE, name));
228 	return (zfs_validate_name(NULL, name, type, B_FALSE));
229 }
230 
231 /*
232  * This function takes the raw DSL properties, and filters out the user-defined
233  * properties into a separate nvlist.
234  */
235 static nvlist_t *
236 process_user_props(zfs_handle_t *zhp, nvlist_t *props)
237 {
238 	libzfs_handle_t *hdl = zhp->zfs_hdl;
239 	nvpair_t *elem;
240 	nvlist_t *nvl;
241 
242 	if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0) {
243 		(void) no_memory(hdl);
244 		return (NULL);
245 	}
246 
247 	elem = NULL;
248 	while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
249 		if (!zfs_prop_user(nvpair_name(elem)))
250 			continue;
251 
252 		nvlist_t *propval = fnvpair_value_nvlist(elem);
253 		if (nvlist_add_nvlist(nvl, nvpair_name(elem), propval) != 0) {
254 			nvlist_free(nvl);
255 			(void) no_memory(hdl);
256 			return (NULL);
257 		}
258 	}
259 
260 	return (nvl);
261 }
262 
263 static zpool_handle_t *
264 zpool_add_handle(zfs_handle_t *zhp, const char *pool_name)
265 {
266 	libzfs_handle_t *hdl = zhp->zfs_hdl;
267 	zpool_handle_t *zph;
268 
269 	if ((zph = zpool_open_canfail(hdl, pool_name)) != NULL) {
270 		if (hdl->libzfs_pool_handles != NULL)
271 			zph->zpool_next = hdl->libzfs_pool_handles;
272 		hdl->libzfs_pool_handles = zph;
273 	}
274 	return (zph);
275 }
276 
277 static zpool_handle_t *
278 zpool_find_handle(zfs_handle_t *zhp, const char *pool_name, int len)
279 {
280 	libzfs_handle_t *hdl = zhp->zfs_hdl;
281 	zpool_handle_t *zph = hdl->libzfs_pool_handles;
282 
283 	while ((zph != NULL) &&
284 	    (strncmp(pool_name, zpool_get_name(zph), len) != 0))
285 		zph = zph->zpool_next;
286 	return (zph);
287 }
288 
289 /*
290  * Returns a handle to the pool that contains the provided dataset.
291  * If a handle to that pool already exists then that handle is returned.
292  * Otherwise, a new handle is created and added to the list of handles.
293  */
294 static zpool_handle_t *
295 zpool_handle(zfs_handle_t *zhp)
296 {
297 	char *pool_name;
298 	int len;
299 	zpool_handle_t *zph;
300 
301 	len = strcspn(zhp->zfs_name, "/@#") + 1;
302 	pool_name = zfs_alloc(zhp->zfs_hdl, len);
303 	(void) strlcpy(pool_name, zhp->zfs_name, len);
304 
305 	zph = zpool_find_handle(zhp, pool_name, len);
306 	if (zph == NULL)
307 		zph = zpool_add_handle(zhp, pool_name);
308 
309 	free(pool_name);
310 	return (zph);
311 }
312 
313 void
314 zpool_free_handles(libzfs_handle_t *hdl)
315 {
316 	zpool_handle_t *next, *zph = hdl->libzfs_pool_handles;
317 
318 	while (zph != NULL) {
319 		next = zph->zpool_next;
320 		zpool_close(zph);
321 		zph = next;
322 	}
323 	hdl->libzfs_pool_handles = NULL;
324 }
325 
326 /*
327  * Utility function to gather stats (objset and zpl) for the given object.
328  */
329 static int
330 get_stats_ioctl(zfs_handle_t *zhp, zfs_cmd_t *zc)
331 {
332 	libzfs_handle_t *hdl = zhp->zfs_hdl;
333 
334 	(void) strlcpy(zc->zc_name, zhp->zfs_name, sizeof (zc->zc_name));
335 
336 	while (zfs_ioctl(hdl, ZFS_IOC_OBJSET_STATS, zc) != 0) {
337 		if (errno == ENOMEM)
338 			zcmd_expand_dst_nvlist(hdl, zc);
339 		else
340 			return (-1);
341 	}
342 	return (0);
343 }
344 
345 /*
346  * Utility function to get the received properties of the given object.
347  */
348 static int
349 get_recvd_props_ioctl(zfs_handle_t *zhp)
350 {
351 	libzfs_handle_t *hdl = zhp->zfs_hdl;
352 	nvlist_t *recvdprops;
353 	zfs_cmd_t zc = {"\0"};
354 	int err;
355 
356 	zcmd_alloc_dst_nvlist(hdl, &zc, 0);
357 
358 	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
359 
360 	while (zfs_ioctl(hdl, ZFS_IOC_OBJSET_RECVD_PROPS, &zc) != 0) {
361 		if (errno == ENOMEM)
362 			zcmd_expand_dst_nvlist(hdl, &zc);
363 		else {
364 			zcmd_free_nvlists(&zc);
365 			return (-1);
366 		}
367 	}
368 
369 	err = zcmd_read_dst_nvlist(zhp->zfs_hdl, &zc, &recvdprops);
370 	zcmd_free_nvlists(&zc);
371 	if (err != 0)
372 		return (-1);
373 
374 	nvlist_free(zhp->zfs_recvd_props);
375 	zhp->zfs_recvd_props = recvdprops;
376 
377 	return (0);
378 }
379 
380 static int
381 put_stats_zhdl(zfs_handle_t *zhp, zfs_cmd_t *zc)
382 {
383 	nvlist_t *allprops, *userprops;
384 
385 	zhp->zfs_dmustats = zc->zc_objset_stats; /* structure assignment */
386 
387 	if (zcmd_read_dst_nvlist(zhp->zfs_hdl, zc, &allprops) != 0) {
388 		return (-1);
389 	}
390 
391 	/*
392 	 * XXX Why do we store the user props separately, in addition to
393 	 * storing them in zfs_props?
394 	 */
395 	if ((userprops = process_user_props(zhp, allprops)) == NULL) {
396 		nvlist_free(allprops);
397 		return (-1);
398 	}
399 
400 	nvlist_free(zhp->zfs_props);
401 	nvlist_free(zhp->zfs_user_props);
402 
403 	zhp->zfs_props = allprops;
404 	zhp->zfs_user_props = userprops;
405 
406 	return (0);
407 }
408 
409 static int
410 get_stats(zfs_handle_t *zhp)
411 {
412 	int rc = 0;
413 	zfs_cmd_t zc = {"\0"};
414 
415 	zcmd_alloc_dst_nvlist(zhp->zfs_hdl, &zc, 0);
416 
417 	if (get_stats_ioctl(zhp, &zc) != 0)
418 		rc = -1;
419 	else if (put_stats_zhdl(zhp, &zc) != 0)
420 		rc = -1;
421 	zcmd_free_nvlists(&zc);
422 	return (rc);
423 }
424 
425 /*
426  * Refresh the properties currently stored in the handle.
427  */
428 void
429 zfs_refresh_properties(zfs_handle_t *zhp)
430 {
431 	(void) get_stats(zhp);
432 }
433 
434 /*
435  * Makes a handle from the given dataset name.  Used by zfs_open() and
436  * zfs_iter_* to create child handles on the fly.
437  */
438 static int
439 make_dataset_handle_common(zfs_handle_t *zhp, zfs_cmd_t *zc)
440 {
441 	if (put_stats_zhdl(zhp, zc) != 0)
442 		return (-1);
443 
444 	/*
445 	 * We've managed to open the dataset and gather statistics.  Determine
446 	 * the high-level type.
447 	 */
448 	if (zhp->zfs_dmustats.dds_type == DMU_OST_ZVOL) {
449 		zhp->zfs_head_type = ZFS_TYPE_VOLUME;
450 	} else if (zhp->zfs_dmustats.dds_type == DMU_OST_ZFS) {
451 		zhp->zfs_head_type = ZFS_TYPE_FILESYSTEM;
452 	} else if (zhp->zfs_dmustats.dds_type == DMU_OST_OTHER) {
453 		errno = EINVAL;
454 		return (-1);
455 	} else if (zhp->zfs_dmustats.dds_inconsistent) {
456 		errno = EBUSY;
457 		return (-1);
458 	} else {
459 		abort();
460 	}
461 
462 	if (zhp->zfs_dmustats.dds_is_snapshot)
463 		zhp->zfs_type = ZFS_TYPE_SNAPSHOT;
464 	else if (zhp->zfs_dmustats.dds_type == DMU_OST_ZVOL)
465 		zhp->zfs_type = ZFS_TYPE_VOLUME;
466 	else if (zhp->zfs_dmustats.dds_type == DMU_OST_ZFS)
467 		zhp->zfs_type = ZFS_TYPE_FILESYSTEM;
468 	else
469 		abort();	/* we should never see any other types */
470 
471 	if ((zhp->zpool_hdl = zpool_handle(zhp)) == NULL)
472 		return (-1);
473 
474 	return (0);
475 }
476 
477 zfs_handle_t *
478 make_dataset_handle(libzfs_handle_t *hdl, const char *path)
479 {
480 	zfs_cmd_t zc = {"\0"};
481 
482 	zfs_handle_t *zhp = calloc(1, sizeof (zfs_handle_t));
483 
484 	if (zhp == NULL)
485 		return (NULL);
486 
487 	zhp->zfs_hdl = hdl;
488 	(void) strlcpy(zhp->zfs_name, path, sizeof (zhp->zfs_name));
489 	zcmd_alloc_dst_nvlist(hdl, &zc, 0);
490 
491 	if (get_stats_ioctl(zhp, &zc) == -1) {
492 		zcmd_free_nvlists(&zc);
493 		free(zhp);
494 		return (NULL);
495 	}
496 	if (make_dataset_handle_common(zhp, &zc) == -1) {
497 		free(zhp);
498 		zhp = NULL;
499 	}
500 	zcmd_free_nvlists(&zc);
501 	return (zhp);
502 }
503 
504 zfs_handle_t *
505 make_dataset_handle_zc(libzfs_handle_t *hdl, zfs_cmd_t *zc)
506 {
507 	zfs_handle_t *zhp = calloc(1, sizeof (zfs_handle_t));
508 
509 	if (zhp == NULL)
510 		return (NULL);
511 
512 	zhp->zfs_hdl = hdl;
513 	(void) strlcpy(zhp->zfs_name, zc->zc_name, sizeof (zhp->zfs_name));
514 	if (make_dataset_handle_common(zhp, zc) == -1) {
515 		free(zhp);
516 		return (NULL);
517 	}
518 	return (zhp);
519 }
520 
521 zfs_handle_t *
522 make_dataset_simple_handle_zc(zfs_handle_t *pzhp, zfs_cmd_t *zc)
523 {
524 	zfs_handle_t *zhp = calloc(1, sizeof (zfs_handle_t));
525 
526 	if (zhp == NULL)
527 		return (NULL);
528 
529 	zhp->zfs_hdl = pzhp->zfs_hdl;
530 	(void) strlcpy(zhp->zfs_name, zc->zc_name, sizeof (zhp->zfs_name));
531 	zhp->zfs_head_type = pzhp->zfs_type;
532 	zhp->zfs_type = ZFS_TYPE_SNAPSHOT;
533 	zhp->zpool_hdl = zpool_handle(zhp);
534 
535 	if (zc->zc_objset_stats.dds_creation_txg != 0) {
536 		/* structure assignment */
537 		zhp->zfs_dmustats = zc->zc_objset_stats;
538 	} else {
539 		if (get_stats_ioctl(zhp, zc) == -1) {
540 			zcmd_free_nvlists(zc);
541 			free(zhp);
542 			return (NULL);
543 		}
544 		if (make_dataset_handle_common(zhp, zc) == -1) {
545 			zcmd_free_nvlists(zc);
546 			free(zhp);
547 			return (NULL);
548 		}
549 	}
550 
551 	if (zhp->zfs_dmustats.dds_is_snapshot ||
552 	    strchr(zc->zc_name, '@') != NULL)
553 		zhp->zfs_type = ZFS_TYPE_SNAPSHOT;
554 	else if (zhp->zfs_dmustats.dds_type == DMU_OST_ZVOL)
555 		zhp->zfs_type = ZFS_TYPE_VOLUME;
556 	else if (zhp->zfs_dmustats.dds_type == DMU_OST_ZFS)
557 		zhp->zfs_type = ZFS_TYPE_FILESYSTEM;
558 
559 	return (zhp);
560 }
561 
562 zfs_handle_t *
563 zfs_handle_dup(zfs_handle_t *zhp_orig)
564 {
565 	zfs_handle_t *zhp = calloc(1, sizeof (zfs_handle_t));
566 
567 	if (zhp == NULL)
568 		return (NULL);
569 
570 	zhp->zfs_hdl = zhp_orig->zfs_hdl;
571 	zhp->zpool_hdl = zhp_orig->zpool_hdl;
572 	(void) strlcpy(zhp->zfs_name, zhp_orig->zfs_name,
573 	    sizeof (zhp->zfs_name));
574 	zhp->zfs_type = zhp_orig->zfs_type;
575 	zhp->zfs_head_type = zhp_orig->zfs_head_type;
576 	zhp->zfs_dmustats = zhp_orig->zfs_dmustats;
577 	if (zhp_orig->zfs_props != NULL) {
578 		if (nvlist_dup(zhp_orig->zfs_props, &zhp->zfs_props, 0) != 0) {
579 			(void) no_memory(zhp->zfs_hdl);
580 			zfs_close(zhp);
581 			return (NULL);
582 		}
583 	}
584 	if (zhp_orig->zfs_user_props != NULL) {
585 		if (nvlist_dup(zhp_orig->zfs_user_props,
586 		    &zhp->zfs_user_props, 0) != 0) {
587 			(void) no_memory(zhp->zfs_hdl);
588 			zfs_close(zhp);
589 			return (NULL);
590 		}
591 	}
592 	if (zhp_orig->zfs_recvd_props != NULL) {
593 		if (nvlist_dup(zhp_orig->zfs_recvd_props,
594 		    &zhp->zfs_recvd_props, 0)) {
595 			(void) no_memory(zhp->zfs_hdl);
596 			zfs_close(zhp);
597 			return (NULL);
598 		}
599 	}
600 	zhp->zfs_mntcheck = zhp_orig->zfs_mntcheck;
601 	if (zhp_orig->zfs_mntopts != NULL) {
602 		zhp->zfs_mntopts = zfs_strdup(zhp_orig->zfs_hdl,
603 		    zhp_orig->zfs_mntopts);
604 	}
605 	zhp->zfs_props_table = zhp_orig->zfs_props_table;
606 	return (zhp);
607 }
608 
609 boolean_t
610 zfs_bookmark_exists(const char *path)
611 {
612 	nvlist_t *bmarks;
613 	nvlist_t *props;
614 	char fsname[ZFS_MAX_DATASET_NAME_LEN];
615 	char *bmark_name;
616 	char *pound;
617 	int err;
618 	boolean_t rv;
619 
620 	(void) strlcpy(fsname, path, sizeof (fsname));
621 	pound = strchr(fsname, '#');
622 	if (pound == NULL)
623 		return (B_FALSE);
624 
625 	*pound = '\0';
626 	bmark_name = pound + 1;
627 	props = fnvlist_alloc();
628 	err = lzc_get_bookmarks(fsname, props, &bmarks);
629 	nvlist_free(props);
630 	if (err != 0) {
631 		nvlist_free(bmarks);
632 		return (B_FALSE);
633 	}
634 
635 	rv = nvlist_exists(bmarks, bmark_name);
636 	nvlist_free(bmarks);
637 	return (rv);
638 }
639 
640 zfs_handle_t *
641 make_bookmark_handle(zfs_handle_t *parent, const char *path,
642     nvlist_t *bmark_props)
643 {
644 	zfs_handle_t *zhp = calloc(1, sizeof (zfs_handle_t));
645 
646 	if (zhp == NULL)
647 		return (NULL);
648 
649 	/* Fill in the name. */
650 	zhp->zfs_hdl = parent->zfs_hdl;
651 	(void) strlcpy(zhp->zfs_name, path, sizeof (zhp->zfs_name));
652 
653 	/* Set the property lists. */
654 	if (nvlist_dup(bmark_props, &zhp->zfs_props, 0) != 0) {
655 		free(zhp);
656 		return (NULL);
657 	}
658 
659 	/* Set the types. */
660 	zhp->zfs_head_type = parent->zfs_head_type;
661 	zhp->zfs_type = ZFS_TYPE_BOOKMARK;
662 
663 	if ((zhp->zpool_hdl = zpool_handle(zhp)) == NULL) {
664 		nvlist_free(zhp->zfs_props);
665 		free(zhp);
666 		return (NULL);
667 	}
668 
669 	return (zhp);
670 }
671 
672 struct zfs_open_bookmarks_cb_data {
673 	const char *path;
674 	zfs_handle_t *zhp;
675 };
676 
677 static int
678 zfs_open_bookmarks_cb(zfs_handle_t *zhp, void *data)
679 {
680 	struct zfs_open_bookmarks_cb_data *dp = data;
681 
682 	/*
683 	 * Is it the one we are looking for?
684 	 */
685 	if (strcmp(dp->path, zfs_get_name(zhp)) == 0) {
686 		/*
687 		 * We found it.  Save it and let the caller know we are done.
688 		 */
689 		dp->zhp = zhp;
690 		return (EEXIST);
691 	}
692 
693 	/*
694 	 * Not found.  Close the handle and ask for another one.
695 	 */
696 	zfs_close(zhp);
697 	return (0);
698 }
699 
700 /*
701  * Opens the given snapshot, bookmark, filesystem, or volume.   The 'types'
702  * argument is a mask of acceptable types.  The function will print an
703  * appropriate error message and return NULL if it can't be opened.
704  */
705 zfs_handle_t *
706 zfs_open(libzfs_handle_t *hdl, const char *path, int types)
707 {
708 	zfs_handle_t *zhp;
709 	char errbuf[ERRBUFLEN];
710 	char *bookp;
711 
712 	(void) snprintf(errbuf, sizeof (errbuf),
713 	    dgettext(TEXT_DOMAIN, "cannot open '%s'"), path);
714 
715 	/*
716 	 * Validate the name before we even try to open it.
717 	 */
718 	if (!zfs_validate_name(hdl, path, types, B_FALSE)) {
719 		(void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
720 		errno = EINVAL;
721 		return (NULL);
722 	}
723 
724 	/*
725 	 * Bookmarks needs to be handled separately.
726 	 */
727 	bookp = strchr(path, '#');
728 	if (bookp == NULL) {
729 		/*
730 		 * Try to get stats for the dataset, which will tell us if it
731 		 * exists.
732 		 */
733 		errno = 0;
734 		if ((zhp = make_dataset_handle(hdl, path)) == NULL) {
735 			(void) zfs_standard_error(hdl, errno, errbuf);
736 			return (NULL);
737 		}
738 	} else {
739 		char dsname[ZFS_MAX_DATASET_NAME_LEN];
740 		zfs_handle_t *pzhp;
741 		struct zfs_open_bookmarks_cb_data cb_data = {path, NULL};
742 
743 		/*
744 		 * We need to cut out '#' and everything after '#'
745 		 * to get the parent dataset name only.
746 		 */
747 		assert(bookp - path < sizeof (dsname));
748 		(void) strlcpy(dsname, path,
749 		    MIN(sizeof (dsname), bookp - path + 1));
750 
751 		/*
752 		 * Create handle for the parent dataset.
753 		 */
754 		errno = 0;
755 		if ((pzhp = make_dataset_handle(hdl, dsname)) == NULL) {
756 			(void) zfs_standard_error(hdl, errno, errbuf);
757 			return (NULL);
758 		}
759 
760 		/*
761 		 * Iterate bookmarks to find the right one.
762 		 */
763 		errno = 0;
764 		if ((zfs_iter_bookmarks_v2(pzhp, 0, zfs_open_bookmarks_cb,
765 		    &cb_data) == 0) && (cb_data.zhp == NULL)) {
766 			(void) zfs_error(hdl, EZFS_NOENT, errbuf);
767 			zfs_close(pzhp);
768 			errno = ENOENT;
769 			return (NULL);
770 		}
771 		if (cb_data.zhp == NULL) {
772 			(void) zfs_standard_error(hdl, errno, errbuf);
773 			zfs_close(pzhp);
774 			return (NULL);
775 		}
776 		zhp = cb_data.zhp;
777 
778 		/*
779 		 * Cleanup.
780 		 */
781 		zfs_close(pzhp);
782 	}
783 
784 	if (!(types & zhp->zfs_type)) {
785 		(void) zfs_error(hdl, EZFS_BADTYPE, errbuf);
786 		zfs_close(zhp);
787 		errno = EINVAL;
788 		return (NULL);
789 	}
790 
791 	return (zhp);
792 }
793 
794 /*
795  * Release a ZFS handle.  Nothing to do but free the associated memory.
796  */
797 void
798 zfs_close(zfs_handle_t *zhp)
799 {
800 	if (zhp->zfs_mntopts)
801 		free(zhp->zfs_mntopts);
802 	nvlist_free(zhp->zfs_props);
803 	nvlist_free(zhp->zfs_user_props);
804 	nvlist_free(zhp->zfs_recvd_props);
805 	free(zhp);
806 }
807 
808 typedef struct mnttab_node {
809 	struct mnttab mtn_mt;
810 	avl_node_t mtn_node;
811 } mnttab_node_t;
812 
813 static int
814 libzfs_mnttab_cache_compare(const void *arg1, const void *arg2)
815 {
816 	const mnttab_node_t *mtn1 = (const mnttab_node_t *)arg1;
817 	const mnttab_node_t *mtn2 = (const mnttab_node_t *)arg2;
818 	int rv;
819 
820 	rv = strcmp(mtn1->mtn_mt.mnt_special, mtn2->mtn_mt.mnt_special);
821 
822 	return (TREE_ISIGN(rv));
823 }
824 
825 void
826 libzfs_mnttab_init(libzfs_handle_t *hdl)
827 {
828 	pthread_mutex_init(&hdl->libzfs_mnttab_cache_lock, NULL);
829 	assert(avl_numnodes(&hdl->libzfs_mnttab_cache) == 0);
830 	avl_create(&hdl->libzfs_mnttab_cache, libzfs_mnttab_cache_compare,
831 	    sizeof (mnttab_node_t), offsetof(mnttab_node_t, mtn_node));
832 }
833 
834 static int
835 libzfs_mnttab_update(libzfs_handle_t *hdl)
836 {
837 	FILE *mnttab;
838 	struct mnttab entry;
839 
840 	if ((mnttab = fopen(MNTTAB, "re")) == NULL)
841 		return (ENOENT);
842 
843 	while (getmntent(mnttab, &entry) == 0) {
844 		mnttab_node_t *mtn;
845 		avl_index_t where;
846 
847 		if (strcmp(entry.mnt_fstype, MNTTYPE_ZFS) != 0)
848 			continue;
849 
850 		mtn = zfs_alloc(hdl, sizeof (mnttab_node_t));
851 		mtn->mtn_mt.mnt_special = zfs_strdup(hdl, entry.mnt_special);
852 		mtn->mtn_mt.mnt_mountp = zfs_strdup(hdl, entry.mnt_mountp);
853 		mtn->mtn_mt.mnt_fstype = zfs_strdup(hdl, entry.mnt_fstype);
854 		mtn->mtn_mt.mnt_mntopts = zfs_strdup(hdl, entry.mnt_mntopts);
855 
856 		/* Exclude duplicate mounts */
857 		if (avl_find(&hdl->libzfs_mnttab_cache, mtn, &where) != NULL) {
858 			free(mtn->mtn_mt.mnt_special);
859 			free(mtn->mtn_mt.mnt_mountp);
860 			free(mtn->mtn_mt.mnt_fstype);
861 			free(mtn->mtn_mt.mnt_mntopts);
862 			free(mtn);
863 			continue;
864 		}
865 
866 		avl_add(&hdl->libzfs_mnttab_cache, mtn);
867 	}
868 
869 	(void) fclose(mnttab);
870 	return (0);
871 }
872 
873 void
874 libzfs_mnttab_fini(libzfs_handle_t *hdl)
875 {
876 	void *cookie = NULL;
877 	mnttab_node_t *mtn;
878 
879 	while ((mtn = avl_destroy_nodes(&hdl->libzfs_mnttab_cache, &cookie))
880 	    != NULL) {
881 		free(mtn->mtn_mt.mnt_special);
882 		free(mtn->mtn_mt.mnt_mountp);
883 		free(mtn->mtn_mt.mnt_fstype);
884 		free(mtn->mtn_mt.mnt_mntopts);
885 		free(mtn);
886 	}
887 	avl_destroy(&hdl->libzfs_mnttab_cache);
888 	(void) pthread_mutex_destroy(&hdl->libzfs_mnttab_cache_lock);
889 }
890 
891 void
892 libzfs_mnttab_cache(libzfs_handle_t *hdl, boolean_t enable)
893 {
894 	hdl->libzfs_mnttab_enable = enable;
895 }
896 
897 int
898 libzfs_mnttab_find(libzfs_handle_t *hdl, const char *fsname,
899     struct mnttab *entry)
900 {
901 	FILE *mnttab;
902 	mnttab_node_t find;
903 	mnttab_node_t *mtn;
904 	int ret = ENOENT;
905 
906 	if (!hdl->libzfs_mnttab_enable) {
907 		struct mnttab srch = { 0 };
908 
909 		if (avl_numnodes(&hdl->libzfs_mnttab_cache))
910 			libzfs_mnttab_fini(hdl);
911 
912 		if ((mnttab = fopen(MNTTAB, "re")) == NULL)
913 			return (ENOENT);
914 
915 		srch.mnt_special = (char *)fsname;
916 		srch.mnt_fstype = (char *)MNTTYPE_ZFS;
917 		ret = getmntany(mnttab, entry, &srch) ? ENOENT : 0;
918 		(void) fclose(mnttab);
919 		return (ret);
920 	}
921 
922 	pthread_mutex_lock(&hdl->libzfs_mnttab_cache_lock);
923 	if (avl_numnodes(&hdl->libzfs_mnttab_cache) == 0) {
924 		int error;
925 
926 		if ((error = libzfs_mnttab_update(hdl)) != 0) {
927 			pthread_mutex_unlock(&hdl->libzfs_mnttab_cache_lock);
928 			return (error);
929 		}
930 	}
931 
932 	find.mtn_mt.mnt_special = (char *)fsname;
933 	mtn = avl_find(&hdl->libzfs_mnttab_cache, &find, NULL);
934 	if (mtn) {
935 		*entry = mtn->mtn_mt;
936 		ret = 0;
937 	}
938 	pthread_mutex_unlock(&hdl->libzfs_mnttab_cache_lock);
939 	return (ret);
940 }
941 
942 void
943 libzfs_mnttab_add(libzfs_handle_t *hdl, const char *special,
944     const char *mountp, const char *mntopts)
945 {
946 	mnttab_node_t *mtn;
947 
948 	pthread_mutex_lock(&hdl->libzfs_mnttab_cache_lock);
949 	if (avl_numnodes(&hdl->libzfs_mnttab_cache) != 0) {
950 		mtn = zfs_alloc(hdl, sizeof (mnttab_node_t));
951 		mtn->mtn_mt.mnt_special = zfs_strdup(hdl, special);
952 		mtn->mtn_mt.mnt_mountp = zfs_strdup(hdl, mountp);
953 		mtn->mtn_mt.mnt_fstype = zfs_strdup(hdl, MNTTYPE_ZFS);
954 		mtn->mtn_mt.mnt_mntopts = zfs_strdup(hdl, mntopts);
955 		/*
956 		 * Another thread may have already added this entry
957 		 * via libzfs_mnttab_update. If so we should skip it.
958 		 */
959 		if (avl_find(&hdl->libzfs_mnttab_cache, mtn, NULL) != NULL) {
960 			free(mtn->mtn_mt.mnt_special);
961 			free(mtn->mtn_mt.mnt_mountp);
962 			free(mtn->mtn_mt.mnt_fstype);
963 			free(mtn->mtn_mt.mnt_mntopts);
964 			free(mtn);
965 		} else {
966 			avl_add(&hdl->libzfs_mnttab_cache, mtn);
967 		}
968 	}
969 	pthread_mutex_unlock(&hdl->libzfs_mnttab_cache_lock);
970 }
971 
972 void
973 libzfs_mnttab_remove(libzfs_handle_t *hdl, const char *fsname)
974 {
975 	mnttab_node_t find;
976 	mnttab_node_t *ret;
977 
978 	pthread_mutex_lock(&hdl->libzfs_mnttab_cache_lock);
979 	find.mtn_mt.mnt_special = (char *)fsname;
980 	if ((ret = avl_find(&hdl->libzfs_mnttab_cache, (void *)&find, NULL))
981 	    != NULL) {
982 		avl_remove(&hdl->libzfs_mnttab_cache, ret);
983 		free(ret->mtn_mt.mnt_special);
984 		free(ret->mtn_mt.mnt_mountp);
985 		free(ret->mtn_mt.mnt_fstype);
986 		free(ret->mtn_mt.mnt_mntopts);
987 		free(ret);
988 	}
989 	pthread_mutex_unlock(&hdl->libzfs_mnttab_cache_lock);
990 }
991 
992 int
993 zfs_spa_version(zfs_handle_t *zhp, int *spa_version)
994 {
995 	zpool_handle_t *zpool_handle = zhp->zpool_hdl;
996 
997 	if (zpool_handle == NULL)
998 		return (-1);
999 
1000 	*spa_version = zpool_get_prop_int(zpool_handle,
1001 	    ZPOOL_PROP_VERSION, NULL);
1002 	return (0);
1003 }
1004 
1005 /*
1006  * The choice of reservation property depends on the SPA version.
1007  */
1008 static int
1009 zfs_which_resv_prop(zfs_handle_t *zhp, zfs_prop_t *resv_prop)
1010 {
1011 	int spa_version;
1012 
1013 	if (zfs_spa_version(zhp, &spa_version) < 0)
1014 		return (-1);
1015 
1016 	if (spa_version >= SPA_VERSION_REFRESERVATION)
1017 		*resv_prop = ZFS_PROP_REFRESERVATION;
1018 	else
1019 		*resv_prop = ZFS_PROP_RESERVATION;
1020 
1021 	return (0);
1022 }
1023 
1024 /*
1025  * Given an nvlist of properties to set, validates that they are correct, and
1026  * parses any numeric properties (index, boolean, etc) if they are specified as
1027  * strings.
1028  */
1029 nvlist_t *
1030 zfs_valid_proplist(libzfs_handle_t *hdl, zfs_type_t type, nvlist_t *nvl,
1031     uint64_t zoned, zfs_handle_t *zhp, zpool_handle_t *zpool_hdl,
1032     boolean_t key_params_ok, const char *errbuf)
1033 {
1034 	nvpair_t *elem;
1035 	uint64_t intval;
1036 	const char *strval;
1037 	zfs_prop_t prop;
1038 	nvlist_t *ret;
1039 	int chosen_normal = -1;
1040 	int chosen_utf = -1;
1041 	int set_maxbs = 0;
1042 
1043 	if (nvlist_alloc(&ret, NV_UNIQUE_NAME, 0) != 0) {
1044 		(void) no_memory(hdl);
1045 		return (NULL);
1046 	}
1047 
1048 	/*
1049 	 * Make sure this property is valid and applies to this type.
1050 	 */
1051 
1052 	elem = NULL;
1053 	while ((elem = nvlist_next_nvpair(nvl, elem)) != NULL) {
1054 		const char *propname = nvpair_name(elem);
1055 
1056 		prop = zfs_name_to_prop(propname);
1057 		if (prop == ZPROP_USERPROP && zfs_prop_user(propname)) {
1058 			/*
1059 			 * This is a user property: make sure it's a
1060 			 * string, and that it's less than ZAP_MAXNAMELEN.
1061 			 */
1062 			if (nvpair_type(elem) != DATA_TYPE_STRING) {
1063 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1064 				    "'%s' must be a string"), propname);
1065 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1066 				goto error;
1067 			}
1068 
1069 			if (strlen(nvpair_name(elem)) >= ZAP_MAXNAMELEN) {
1070 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1071 				    "property name '%s' is too long"),
1072 				    propname);
1073 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1074 				goto error;
1075 			}
1076 
1077 			(void) nvpair_value_string(elem, &strval);
1078 			if (nvlist_add_string(ret, propname, strval) != 0) {
1079 				(void) no_memory(hdl);
1080 				goto error;
1081 			}
1082 			continue;
1083 		}
1084 
1085 		/*
1086 		 * Currently, only user properties can be modified on
1087 		 * snapshots.
1088 		 */
1089 		if (type == ZFS_TYPE_SNAPSHOT) {
1090 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1091 			    "this property can not be modified for snapshots"));
1092 			(void) zfs_error(hdl, EZFS_PROPTYPE, errbuf);
1093 			goto error;
1094 		}
1095 
1096 		if (prop == ZPROP_USERPROP && zfs_prop_userquota(propname)) {
1097 			zfs_userquota_prop_t uqtype;
1098 			char *newpropname = NULL;
1099 			char domain[128];
1100 			uint64_t rid;
1101 			uint64_t valary[3];
1102 			int rc;
1103 
1104 			if (userquota_propname_decode(propname, zoned,
1105 			    &uqtype, domain, sizeof (domain), &rid) != 0) {
1106 				zfs_error_aux(hdl,
1107 				    dgettext(TEXT_DOMAIN,
1108 				    "'%s' has an invalid user/group name"),
1109 				    propname);
1110 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1111 				goto error;
1112 			}
1113 
1114 			if (uqtype != ZFS_PROP_USERQUOTA &&
1115 			    uqtype != ZFS_PROP_GROUPQUOTA &&
1116 			    uqtype != ZFS_PROP_USEROBJQUOTA &&
1117 			    uqtype != ZFS_PROP_GROUPOBJQUOTA &&
1118 			    uqtype != ZFS_PROP_PROJECTQUOTA &&
1119 			    uqtype != ZFS_PROP_PROJECTOBJQUOTA) {
1120 				zfs_error_aux(hdl,
1121 				    dgettext(TEXT_DOMAIN, "'%s' is readonly"),
1122 				    propname);
1123 				(void) zfs_error(hdl, EZFS_PROPREADONLY,
1124 				    errbuf);
1125 				goto error;
1126 			}
1127 
1128 			if (nvpair_type(elem) == DATA_TYPE_STRING) {
1129 				(void) nvpair_value_string(elem, &strval);
1130 				if (strcmp(strval, "none") == 0) {
1131 					intval = 0;
1132 				} else if (zfs_nicestrtonum(hdl,
1133 				    strval, &intval) != 0) {
1134 					(void) zfs_error(hdl,
1135 					    EZFS_BADPROP, errbuf);
1136 					goto error;
1137 				}
1138 			} else if (nvpair_type(elem) ==
1139 			    DATA_TYPE_UINT64) {
1140 				(void) nvpair_value_uint64(elem, &intval);
1141 				if (intval == 0) {
1142 					zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1143 					    "use 'none' to disable "
1144 					    "{user|group|project}quota"));
1145 					goto error;
1146 				}
1147 			} else {
1148 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1149 				    "'%s' must be a number"), propname);
1150 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1151 				goto error;
1152 			}
1153 
1154 			/*
1155 			 * Encode the prop name as
1156 			 * userquota@<hex-rid>-domain, to make it easy
1157 			 * for the kernel to decode.
1158 			 */
1159 			rc = asprintf(&newpropname, "%s%llx-%s",
1160 			    zfs_userquota_prop_prefixes[uqtype],
1161 			    (longlong_t)rid, domain);
1162 			if (rc == -1 || newpropname == NULL) {
1163 				(void) no_memory(hdl);
1164 				goto error;
1165 			}
1166 
1167 			valary[0] = uqtype;
1168 			valary[1] = rid;
1169 			valary[2] = intval;
1170 			if (nvlist_add_uint64_array(ret, newpropname,
1171 			    valary, 3) != 0) {
1172 				free(newpropname);
1173 				(void) no_memory(hdl);
1174 				goto error;
1175 			}
1176 			free(newpropname);
1177 			continue;
1178 		} else if (prop == ZPROP_USERPROP &&
1179 		    zfs_prop_written(propname)) {
1180 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1181 			    "'%s' is readonly"),
1182 			    propname);
1183 			(void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
1184 			goto error;
1185 		}
1186 
1187 		if (prop == ZPROP_INVAL) {
1188 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1189 			    "invalid property '%s'"), propname);
1190 			(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1191 			goto error;
1192 		}
1193 
1194 		if (!zfs_prop_valid_for_type(prop, type, B_FALSE)) {
1195 			zfs_error_aux(hdl,
1196 			    dgettext(TEXT_DOMAIN, "'%s' does not "
1197 			    "apply to datasets of this type"), propname);
1198 			(void) zfs_error(hdl, EZFS_PROPTYPE, errbuf);
1199 			goto error;
1200 		}
1201 
1202 		if (zfs_prop_readonly(prop) &&
1203 		    !(zfs_prop_setonce(prop) && zhp == NULL) &&
1204 		    !(zfs_prop_encryption_key_param(prop) && key_params_ok)) {
1205 			zfs_error_aux(hdl,
1206 			    dgettext(TEXT_DOMAIN, "'%s' is readonly"),
1207 			    propname);
1208 			(void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
1209 			goto error;
1210 		}
1211 
1212 		if (zprop_parse_value(hdl, elem, prop, type, ret,
1213 		    &strval, &intval, errbuf) != 0)
1214 			goto error;
1215 
1216 		/*
1217 		 * Perform some additional checks for specific properties.
1218 		 */
1219 		switch (prop) {
1220 		case ZFS_PROP_VERSION:
1221 		{
1222 			int version;
1223 
1224 			if (zhp == NULL)
1225 				break;
1226 			version = zfs_prop_get_int(zhp, ZFS_PROP_VERSION);
1227 			if (intval < version) {
1228 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1229 				    "Can not downgrade; already at version %u"),
1230 				    version);
1231 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1232 				goto error;
1233 			}
1234 			break;
1235 		}
1236 
1237 		case ZFS_PROP_VOLBLOCKSIZE:
1238 		case ZFS_PROP_RECORDSIZE:
1239 		{
1240 			int maxbs = SPA_MAXBLOCKSIZE;
1241 			char buf[64];
1242 
1243 			if (zpool_hdl != NULL) {
1244 				maxbs = zpool_get_prop_int(zpool_hdl,
1245 				    ZPOOL_PROP_MAXBLOCKSIZE, NULL);
1246 			}
1247 			/*
1248 			 * The value must be a power of two between
1249 			 * SPA_MINBLOCKSIZE and maxbs.
1250 			 */
1251 			if (intval < SPA_MINBLOCKSIZE ||
1252 			    intval > maxbs || !ISP2(intval)) {
1253 				zfs_nicebytes(maxbs, buf, sizeof (buf));
1254 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1255 				    "'%s' must be power of 2 from 512B "
1256 				    "to %s"), propname, buf);
1257 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1258 				goto error;
1259 			}
1260 			/* save the ZFS_PROP_RECORDSIZE during create op */
1261 			if (zpool_hdl == NULL && prop == ZFS_PROP_RECORDSIZE) {
1262 				set_maxbs = intval;
1263 			}
1264 			break;
1265 		}
1266 
1267 		case ZFS_PROP_SPECIAL_SMALL_BLOCKS:
1268 		{
1269 			int maxbs =
1270 			    set_maxbs == 0 ? SPA_OLD_MAXBLOCKSIZE : set_maxbs;
1271 			char buf[64];
1272 
1273 			if (zpool_hdl != NULL) {
1274 				char state[64] = "";
1275 
1276 				maxbs = zpool_get_prop_int(zpool_hdl,
1277 				    ZPOOL_PROP_MAXBLOCKSIZE, NULL);
1278 
1279 				/*
1280 				 * Issue a warning but do not fail so that
1281 				 * tests for settable properties succeed.
1282 				 */
1283 				if (zpool_prop_get_feature(zpool_hdl,
1284 				    "feature@allocation_classes", state,
1285 				    sizeof (state)) != 0 ||
1286 				    strcmp(state, ZFS_FEATURE_ACTIVE) != 0) {
1287 					(void) fprintf(stderr, gettext(
1288 					    "%s: property requires a special "
1289 					    "device in the pool\n"), propname);
1290 				}
1291 			}
1292 			if (intval != 0 &&
1293 			    (intval < SPA_MINBLOCKSIZE ||
1294 			    intval > maxbs || !ISP2(intval))) {
1295 				zfs_nicebytes(maxbs, buf, sizeof (buf));
1296 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1297 				    "invalid '%s=%llu' property: must be zero "
1298 				    "or a power of 2 from 512B to %s"),
1299 				    propname, (unsigned long long)intval, buf);
1300 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1301 				goto error;
1302 			}
1303 			break;
1304 		}
1305 
1306 		case ZFS_PROP_MLSLABEL:
1307 		{
1308 #ifdef HAVE_MLSLABEL
1309 			/*
1310 			 * Verify the mlslabel string and convert to
1311 			 * internal hex label string.
1312 			 */
1313 
1314 			m_label_t *new_sl;
1315 			char *hex = NULL;	/* internal label string */
1316 
1317 			/* Default value is already OK. */
1318 			if (strcasecmp(strval, ZFS_MLSLABEL_DEFAULT) == 0)
1319 				break;
1320 
1321 			/* Verify the label can be converted to binary form */
1322 			if (((new_sl = m_label_alloc(MAC_LABEL)) == NULL) ||
1323 			    (str_to_label(strval, &new_sl, MAC_LABEL,
1324 			    L_NO_CORRECTION, NULL) == -1)) {
1325 				goto badlabel;
1326 			}
1327 
1328 			/* Now translate to hex internal label string */
1329 			if (label_to_str(new_sl, &hex, M_INTERNAL,
1330 			    DEF_NAMES) != 0) {
1331 				if (hex)
1332 					free(hex);
1333 				goto badlabel;
1334 			}
1335 			m_label_free(new_sl);
1336 
1337 			/* If string is already in internal form, we're done. */
1338 			if (strcmp(strval, hex) == 0) {
1339 				free(hex);
1340 				break;
1341 			}
1342 
1343 			/* Replace the label string with the internal form. */
1344 			(void) nvlist_remove(ret, zfs_prop_to_name(prop),
1345 			    DATA_TYPE_STRING);
1346 			fnvlist_add_string(ret, zfs_prop_to_name(prop), hex);
1347 			free(hex);
1348 
1349 			break;
1350 
1351 badlabel:
1352 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1353 			    "invalid mlslabel '%s'"), strval);
1354 			(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1355 			m_label_free(new_sl);	/* OK if null */
1356 			goto error;
1357 #else
1358 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1359 			    "mlslabels are unsupported"));
1360 			(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1361 			goto error;
1362 #endif /* HAVE_MLSLABEL */
1363 		}
1364 
1365 		case ZFS_PROP_MOUNTPOINT:
1366 		{
1367 			namecheck_err_t why;
1368 
1369 			if (strcmp(strval, ZFS_MOUNTPOINT_NONE) == 0 ||
1370 			    strcmp(strval, ZFS_MOUNTPOINT_LEGACY) == 0)
1371 				break;
1372 
1373 			if (mountpoint_namecheck(strval, &why)) {
1374 				switch (why) {
1375 				case NAME_ERR_LEADING_SLASH:
1376 					zfs_error_aux(hdl,
1377 					    dgettext(TEXT_DOMAIN,
1378 					    "'%s' must be an absolute path, "
1379 					    "'none', or 'legacy'"), propname);
1380 					break;
1381 				case NAME_ERR_TOOLONG:
1382 					zfs_error_aux(hdl,
1383 					    dgettext(TEXT_DOMAIN,
1384 					    "component of '%s' is too long"),
1385 					    propname);
1386 					break;
1387 
1388 				default:
1389 					zfs_error_aux(hdl,
1390 					    dgettext(TEXT_DOMAIN,
1391 					    "(%d) not defined"),
1392 					    why);
1393 					break;
1394 				}
1395 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1396 				goto error;
1397 			}
1398 			zfs_fallthrough;
1399 		}
1400 
1401 		case ZFS_PROP_SHARESMB:
1402 		case ZFS_PROP_SHARENFS:
1403 			/*
1404 			 * For the mountpoint and sharenfs or sharesmb
1405 			 * properties, check if it can be set in a
1406 			 * global/non-global zone based on
1407 			 * the zoned property value:
1408 			 *
1409 			 *		global zone	    non-global zone
1410 			 * --------------------------------------------------
1411 			 * zoned=on	mountpoint (no)	    mountpoint (yes)
1412 			 *		sharenfs (no)	    sharenfs (no)
1413 			 *		sharesmb (no)	    sharesmb (no)
1414 			 *
1415 			 * zoned=off	mountpoint (yes)	N/A
1416 			 *		sharenfs (yes)
1417 			 *		sharesmb (yes)
1418 			 */
1419 			if (zoned) {
1420 				if (getzoneid() == GLOBAL_ZONEID) {
1421 					zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1422 					    "'%s' cannot be set on "
1423 					    "dataset in a non-global zone"),
1424 					    propname);
1425 					(void) zfs_error(hdl, EZFS_ZONED,
1426 					    errbuf);
1427 					goto error;
1428 				} else if (prop == ZFS_PROP_SHARENFS ||
1429 				    prop == ZFS_PROP_SHARESMB) {
1430 					zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1431 					    "'%s' cannot be set in "
1432 					    "a non-global zone"), propname);
1433 					(void) zfs_error(hdl, EZFS_ZONED,
1434 					    errbuf);
1435 					goto error;
1436 				}
1437 			} else if (getzoneid() != GLOBAL_ZONEID) {
1438 				/*
1439 				 * If zoned property is 'off', this must be in
1440 				 * a global zone. If not, something is wrong.
1441 				 */
1442 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1443 				    "'%s' cannot be set while dataset "
1444 				    "'zoned' property is set"), propname);
1445 				(void) zfs_error(hdl, EZFS_ZONED, errbuf);
1446 				goto error;
1447 			}
1448 
1449 			/*
1450 			 * At this point, it is legitimate to set the
1451 			 * property. Now we want to make sure that the
1452 			 * property value is valid if it is sharenfs.
1453 			 */
1454 			if ((prop == ZFS_PROP_SHARENFS ||
1455 			    prop == ZFS_PROP_SHARESMB) &&
1456 			    strcmp(strval, "on") != 0 &&
1457 			    strcmp(strval, "off") != 0) {
1458 				enum sa_protocol proto;
1459 
1460 				if (prop == ZFS_PROP_SHARESMB)
1461 					proto = SA_PROTOCOL_SMB;
1462 				else
1463 					proto = SA_PROTOCOL_NFS;
1464 
1465 				if (sa_validate_shareopts(strval, proto) !=
1466 				    SA_OK) {
1467 					zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1468 					    "'%s' cannot be set to invalid "
1469 					    "options"), propname);
1470 					(void) zfs_error(hdl, EZFS_BADPROP,
1471 					    errbuf);
1472 					goto error;
1473 				}
1474 			}
1475 
1476 			break;
1477 
1478 		case ZFS_PROP_KEYLOCATION:
1479 			if (!zfs_prop_valid_keylocation(strval, B_FALSE)) {
1480 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1481 				    "invalid keylocation"));
1482 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1483 				goto error;
1484 			}
1485 
1486 			if (zhp != NULL) {
1487 				uint64_t crypt =
1488 				    zfs_prop_get_int(zhp, ZFS_PROP_ENCRYPTION);
1489 
1490 				if (crypt == ZIO_CRYPT_OFF &&
1491 				    strcmp(strval, "none") != 0) {
1492 					zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1493 					    "keylocation must be 'none' "
1494 					    "for unencrypted datasets"));
1495 					(void) zfs_error(hdl, EZFS_BADPROP,
1496 					    errbuf);
1497 					goto error;
1498 				} else if (crypt != ZIO_CRYPT_OFF &&
1499 				    strcmp(strval, "none") == 0) {
1500 					zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1501 					    "keylocation must not be 'none' "
1502 					    "for encrypted datasets"));
1503 					(void) zfs_error(hdl, EZFS_BADPROP,
1504 					    errbuf);
1505 					goto error;
1506 				}
1507 			}
1508 			break;
1509 
1510 		case ZFS_PROP_PBKDF2_ITERS:
1511 			if (intval < MIN_PBKDF2_ITERATIONS) {
1512 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1513 				    "minimum pbkdf2 iterations is %u"),
1514 				    MIN_PBKDF2_ITERATIONS);
1515 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1516 				goto error;
1517 			}
1518 			break;
1519 
1520 		case ZFS_PROP_UTF8ONLY:
1521 			chosen_utf = (int)intval;
1522 			break;
1523 
1524 		case ZFS_PROP_NORMALIZE:
1525 			chosen_normal = (int)intval;
1526 			break;
1527 
1528 		default:
1529 			break;
1530 		}
1531 
1532 		/*
1533 		 * For changes to existing volumes, we have some additional
1534 		 * checks to enforce.
1535 		 */
1536 		if (type == ZFS_TYPE_VOLUME && zhp != NULL) {
1537 			uint64_t blocksize = zfs_prop_get_int(zhp,
1538 			    ZFS_PROP_VOLBLOCKSIZE);
1539 			char buf[64];
1540 
1541 			switch (prop) {
1542 			case ZFS_PROP_VOLSIZE:
1543 				if (intval % blocksize != 0) {
1544 					zfs_nicebytes(blocksize, buf,
1545 					    sizeof (buf));
1546 					zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1547 					    "'%s' must be a multiple of "
1548 					    "volume block size (%s)"),
1549 					    propname, buf);
1550 					(void) zfs_error(hdl, EZFS_BADPROP,
1551 					    errbuf);
1552 					goto error;
1553 				}
1554 
1555 				if (intval == 0) {
1556 					zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1557 					    "'%s' cannot be zero"),
1558 					    propname);
1559 					(void) zfs_error(hdl, EZFS_BADPROP,
1560 					    errbuf);
1561 					goto error;
1562 				}
1563 				break;
1564 
1565 			default:
1566 				break;
1567 			}
1568 		}
1569 
1570 		/* check encryption properties */
1571 		if (zhp != NULL) {
1572 			int64_t crypt = zfs_prop_get_int(zhp,
1573 			    ZFS_PROP_ENCRYPTION);
1574 
1575 			switch (prop) {
1576 			case ZFS_PROP_COPIES:
1577 				if (crypt != ZIO_CRYPT_OFF && intval > 2) {
1578 					zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1579 					    "encrypted datasets cannot have "
1580 					    "3 copies"));
1581 					(void) zfs_error(hdl, EZFS_BADPROP,
1582 					    errbuf);
1583 					goto error;
1584 				}
1585 				break;
1586 			default:
1587 				break;
1588 			}
1589 		}
1590 	}
1591 
1592 	/*
1593 	 * If normalization was chosen, but no UTF8 choice was made,
1594 	 * enforce rejection of non-UTF8 names.
1595 	 *
1596 	 * If normalization was chosen, but rejecting non-UTF8 names
1597 	 * was explicitly not chosen, it is an error.
1598 	 *
1599 	 * If utf8only was turned off, but the parent has normalization,
1600 	 * turn off normalization.
1601 	 */
1602 	if (chosen_normal > 0 && chosen_utf < 0) {
1603 		if (nvlist_add_uint64(ret,
1604 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY), 1) != 0) {
1605 			(void) no_memory(hdl);
1606 			goto error;
1607 		}
1608 	} else if (chosen_normal > 0 && chosen_utf == 0) {
1609 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1610 		    "'%s' must be set 'on' if normalization chosen"),
1611 		    zfs_prop_to_name(ZFS_PROP_UTF8ONLY));
1612 		(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1613 		goto error;
1614 	} else if (chosen_normal < 0 && chosen_utf == 0) {
1615 		if (nvlist_add_uint64(ret,
1616 		    zfs_prop_to_name(ZFS_PROP_NORMALIZE), 0) != 0) {
1617 			(void) no_memory(hdl);
1618 			goto error;
1619 		}
1620 	}
1621 	return (ret);
1622 
1623 error:
1624 	nvlist_free(ret);
1625 	return (NULL);
1626 }
1627 
1628 static int
1629 zfs_add_synthetic_resv(zfs_handle_t *zhp, nvlist_t *nvl)
1630 {
1631 	uint64_t old_volsize;
1632 	uint64_t new_volsize;
1633 	uint64_t old_reservation;
1634 	uint64_t new_reservation;
1635 	zfs_prop_t resv_prop;
1636 	nvlist_t *props;
1637 	zpool_handle_t *zph = zpool_handle(zhp);
1638 
1639 	/*
1640 	 * If this is an existing volume, and someone is setting the volsize,
1641 	 * make sure that it matches the reservation, or add it if necessary.
1642 	 */
1643 	old_volsize = zfs_prop_get_int(zhp, ZFS_PROP_VOLSIZE);
1644 	if (zfs_which_resv_prop(zhp, &resv_prop) < 0)
1645 		return (-1);
1646 	old_reservation = zfs_prop_get_int(zhp, resv_prop);
1647 
1648 	props = fnvlist_alloc();
1649 	fnvlist_add_uint64(props, zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
1650 	    zfs_prop_get_int(zhp, ZFS_PROP_VOLBLOCKSIZE));
1651 
1652 	if ((zvol_volsize_to_reservation(zph, old_volsize, props) !=
1653 	    old_reservation) || nvlist_exists(nvl,
1654 	    zfs_prop_to_name(resv_prop))) {
1655 		fnvlist_free(props);
1656 		return (0);
1657 	}
1658 	if (nvlist_lookup_uint64(nvl, zfs_prop_to_name(ZFS_PROP_VOLSIZE),
1659 	    &new_volsize) != 0) {
1660 		fnvlist_free(props);
1661 		return (-1);
1662 	}
1663 	new_reservation = zvol_volsize_to_reservation(zph, new_volsize, props);
1664 	fnvlist_free(props);
1665 
1666 	if (nvlist_add_uint64(nvl, zfs_prop_to_name(resv_prop),
1667 	    new_reservation) != 0) {
1668 		(void) no_memory(zhp->zfs_hdl);
1669 		return (-1);
1670 	}
1671 	return (1);
1672 }
1673 
1674 /*
1675  * Helper for 'zfs {set|clone} refreservation=auto'.  Must be called after
1676  * zfs_valid_proplist(), as it is what sets the UINT64_MAX sentinel value.
1677  * Return codes must match zfs_add_synthetic_resv().
1678  */
1679 static int
1680 zfs_fix_auto_resv(zfs_handle_t *zhp, nvlist_t *nvl)
1681 {
1682 	uint64_t volsize;
1683 	uint64_t resvsize;
1684 	zfs_prop_t prop;
1685 	nvlist_t *props;
1686 
1687 	if (!ZFS_IS_VOLUME(zhp)) {
1688 		return (0);
1689 	}
1690 
1691 	if (zfs_which_resv_prop(zhp, &prop) != 0) {
1692 		return (-1);
1693 	}
1694 
1695 	if (prop != ZFS_PROP_REFRESERVATION) {
1696 		return (0);
1697 	}
1698 
1699 	if (nvlist_lookup_uint64(nvl, zfs_prop_to_name(prop), &resvsize) != 0) {
1700 		/* No value being set, so it can't be "auto" */
1701 		return (0);
1702 	}
1703 	if (resvsize != UINT64_MAX) {
1704 		/* Being set to a value other than "auto" */
1705 		return (0);
1706 	}
1707 
1708 	props = fnvlist_alloc();
1709 
1710 	fnvlist_add_uint64(props, zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
1711 	    zfs_prop_get_int(zhp, ZFS_PROP_VOLBLOCKSIZE));
1712 
1713 	if (nvlist_lookup_uint64(nvl, zfs_prop_to_name(ZFS_PROP_VOLSIZE),
1714 	    &volsize) != 0) {
1715 		volsize = zfs_prop_get_int(zhp, ZFS_PROP_VOLSIZE);
1716 	}
1717 
1718 	resvsize = zvol_volsize_to_reservation(zpool_handle(zhp), volsize,
1719 	    props);
1720 	fnvlist_free(props);
1721 
1722 	(void) nvlist_remove_all(nvl, zfs_prop_to_name(prop));
1723 	if (nvlist_add_uint64(nvl, zfs_prop_to_name(prop), resvsize) != 0) {
1724 		(void) no_memory(zhp->zfs_hdl);
1725 		return (-1);
1726 	}
1727 	return (1);
1728 }
1729 
1730 static boolean_t
1731 zfs_is_namespace_prop(zfs_prop_t prop)
1732 {
1733 	switch (prop) {
1734 
1735 	case ZFS_PROP_ATIME:
1736 	case ZFS_PROP_RELATIME:
1737 	case ZFS_PROP_DEVICES:
1738 	case ZFS_PROP_EXEC:
1739 	case ZFS_PROP_SETUID:
1740 	case ZFS_PROP_READONLY:
1741 	case ZFS_PROP_XATTR:
1742 	case ZFS_PROP_NBMAND:
1743 		return (B_TRUE);
1744 
1745 	default:
1746 		return (B_FALSE);
1747 	}
1748 }
1749 
1750 /*
1751  * Given a property name and value, set the property for the given dataset.
1752  */
1753 int
1754 zfs_prop_set(zfs_handle_t *zhp, const char *propname, const char *propval)
1755 {
1756 	int ret = -1;
1757 	char errbuf[ERRBUFLEN];
1758 	libzfs_handle_t *hdl = zhp->zfs_hdl;
1759 	nvlist_t *nvl = NULL;
1760 
1761 	(void) snprintf(errbuf, sizeof (errbuf),
1762 	    dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
1763 	    zhp->zfs_name);
1764 
1765 	if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0 ||
1766 	    nvlist_add_string(nvl, propname, propval) != 0) {
1767 		(void) no_memory(hdl);
1768 		goto error;
1769 	}
1770 
1771 	ret = zfs_prop_set_list(zhp, nvl);
1772 
1773 error:
1774 	nvlist_free(nvl);
1775 	return (ret);
1776 }
1777 
1778 /*
1779  * Given an nvlist of property names and values, set the properties for the
1780  * given dataset.
1781  */
1782 int
1783 zfs_prop_set_list(zfs_handle_t *zhp, nvlist_t *props)
1784 {
1785 	return (zfs_prop_set_list_flags(zhp, props, 0));
1786 }
1787 
1788 /*
1789  * Given an nvlist of property names, values and flags, set the properties
1790  * for the given dataset. If ZFS_SET_NOMOUNT is set, it allows to update
1791  * mountpoint, sharenfs and sharesmb properties without (un/re)mounting
1792  * and (un/re)sharing the dataset.
1793  */
1794 int
1795 zfs_prop_set_list_flags(zfs_handle_t *zhp, nvlist_t *props, int flags)
1796 {
1797 	zfs_cmd_t zc = {"\0"};
1798 	int ret = -1;
1799 	prop_changelist_t **cls = NULL;
1800 	int cl_idx;
1801 	char errbuf[ERRBUFLEN];
1802 	libzfs_handle_t *hdl = zhp->zfs_hdl;
1803 	nvlist_t *nvl;
1804 	int nvl_len = 0;
1805 	int added_resv = 0;
1806 	zfs_prop_t prop;
1807 	boolean_t nsprop = B_FALSE;
1808 	nvpair_t *elem;
1809 
1810 	(void) snprintf(errbuf, sizeof (errbuf),
1811 	    dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
1812 	    zhp->zfs_name);
1813 
1814 	if ((nvl = zfs_valid_proplist(hdl, zhp->zfs_type, props,
1815 	    zfs_prop_get_int(zhp, ZFS_PROP_ZONED), zhp, zhp->zpool_hdl,
1816 	    B_FALSE, errbuf)) == NULL)
1817 		goto error;
1818 
1819 	/*
1820 	 * We have to check for any extra properties which need to be added
1821 	 * before computing the length of the nvlist.
1822 	 */
1823 	for (elem = nvlist_next_nvpair(nvl, NULL);
1824 	    elem != NULL;
1825 	    elem = nvlist_next_nvpair(nvl, elem)) {
1826 		if (zfs_name_to_prop(nvpair_name(elem)) == ZFS_PROP_VOLSIZE &&
1827 		    (added_resv = zfs_add_synthetic_resv(zhp, nvl)) == -1) {
1828 			goto error;
1829 		}
1830 	}
1831 
1832 	if (added_resv != 1 &&
1833 	    (added_resv = zfs_fix_auto_resv(zhp, nvl)) == -1) {
1834 		goto error;
1835 	}
1836 
1837 	/*
1838 	 * Check how many properties we're setting and allocate an array to
1839 	 * store changelist pointers for postfix().
1840 	 */
1841 	for (elem = nvlist_next_nvpair(nvl, NULL);
1842 	    elem != NULL;
1843 	    elem = nvlist_next_nvpair(nvl, elem))
1844 		nvl_len++;
1845 	if ((cls = calloc(nvl_len, sizeof (prop_changelist_t *))) == NULL)
1846 		goto error;
1847 
1848 	cl_idx = 0;
1849 	for (elem = nvlist_next_nvpair(nvl, NULL);
1850 	    elem != NULL;
1851 	    elem = nvlist_next_nvpair(nvl, elem)) {
1852 
1853 		prop = zfs_name_to_prop(nvpair_name(elem));
1854 		nsprop |= zfs_is_namespace_prop(prop);
1855 
1856 		assert(cl_idx < nvl_len);
1857 		/*
1858 		 * We don't want to unmount & remount the dataset when changing
1859 		 * its canmount property to 'on' or 'noauto'.  We only use
1860 		 * the changelist logic to unmount when setting canmount=off.
1861 		 */
1862 		if (prop != ZFS_PROP_CANMOUNT ||
1863 		    (fnvpair_value_uint64(elem) == ZFS_CANMOUNT_OFF &&
1864 		    zfs_is_mounted(zhp, NULL))) {
1865 			cls[cl_idx] = changelist_gather(zhp, prop,
1866 			    ((flags & ZFS_SET_NOMOUNT) ?
1867 			    CL_GATHER_DONT_UNMOUNT : 0), 0);
1868 			if (cls[cl_idx] == NULL)
1869 				goto error;
1870 		}
1871 
1872 		if (prop == ZFS_PROP_MOUNTPOINT &&
1873 		    changelist_haszonedchild(cls[cl_idx])) {
1874 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1875 			    "child dataset with inherited mountpoint is used "
1876 			    "in a non-global zone"));
1877 			ret = zfs_error(hdl, EZFS_ZONED, errbuf);
1878 			goto error;
1879 		}
1880 
1881 		if (cls[cl_idx] != NULL &&
1882 		    (ret = changelist_prefix(cls[cl_idx])) != 0)
1883 			goto error;
1884 
1885 		cl_idx++;
1886 	}
1887 	assert(cl_idx == nvl_len);
1888 
1889 	/*
1890 	 * Execute the corresponding ioctl() to set this list of properties.
1891 	 */
1892 	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
1893 
1894 	zcmd_write_src_nvlist(hdl, &zc, nvl);
1895 	zcmd_alloc_dst_nvlist(hdl, &zc, 0);
1896 
1897 	ret = zfs_ioctl(hdl, ZFS_IOC_SET_PROP, &zc);
1898 
1899 	if (ret != 0) {
1900 		if (zc.zc_nvlist_dst_filled == B_FALSE) {
1901 			(void) zfs_standard_error(hdl, errno, errbuf);
1902 			goto error;
1903 		}
1904 
1905 		/* Get the list of unset properties back and report them. */
1906 		nvlist_t *errorprops = NULL;
1907 		if (zcmd_read_dst_nvlist(hdl, &zc, &errorprops) != 0)
1908 			goto error;
1909 		for (nvpair_t *elem = nvlist_next_nvpair(errorprops, NULL);
1910 		    elem != NULL;
1911 		    elem = nvlist_next_nvpair(errorprops, elem)) {
1912 			prop = zfs_name_to_prop(nvpair_name(elem));
1913 			zfs_setprop_error(hdl, prop, errno, errbuf);
1914 		}
1915 		nvlist_free(errorprops);
1916 
1917 		if (added_resv && errno == ENOSPC) {
1918 			/* clean up the volsize property we tried to set */
1919 			uint64_t old_volsize = zfs_prop_get_int(zhp,
1920 			    ZFS_PROP_VOLSIZE);
1921 			nvlist_free(nvl);
1922 			nvl = NULL;
1923 			zcmd_free_nvlists(&zc);
1924 
1925 			if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
1926 				goto error;
1927 			if (nvlist_add_uint64(nvl,
1928 			    zfs_prop_to_name(ZFS_PROP_VOLSIZE),
1929 			    old_volsize) != 0)
1930 				goto error;
1931 			zcmd_write_src_nvlist(hdl, &zc, nvl);
1932 			(void) zfs_ioctl(hdl, ZFS_IOC_SET_PROP, &zc);
1933 		}
1934 	} else {
1935 		for (cl_idx = 0; cl_idx < nvl_len; cl_idx++) {
1936 			if (cls[cl_idx] != NULL) {
1937 				int clp_err = changelist_postfix(cls[cl_idx]);
1938 				if (clp_err != 0)
1939 					ret = clp_err;
1940 			}
1941 		}
1942 
1943 		if (ret == 0) {
1944 			/*
1945 			 * Refresh the statistics so the new property
1946 			 * value is reflected.
1947 			 */
1948 			(void) get_stats(zhp);
1949 
1950 			/*
1951 			 * Remount the filesystem to propagate the change
1952 			 * if one of the options handled by the generic
1953 			 * Linux namespace layer has been modified.
1954 			 */
1955 			if (nsprop && zfs_is_mounted(zhp, NULL))
1956 				ret = zfs_mount(zhp, MNTOPT_REMOUNT, 0);
1957 		}
1958 	}
1959 
1960 error:
1961 	nvlist_free(nvl);
1962 	zcmd_free_nvlists(&zc);
1963 	if (cls != NULL) {
1964 		for (cl_idx = 0; cl_idx < nvl_len; cl_idx++) {
1965 			if (cls[cl_idx] != NULL)
1966 				changelist_free(cls[cl_idx]);
1967 		}
1968 		free(cls);
1969 	}
1970 	return (ret);
1971 }
1972 
1973 /*
1974  * Given a property, inherit the value from the parent dataset, or if received
1975  * is TRUE, revert to the received value, if any.
1976  */
1977 int
1978 zfs_prop_inherit(zfs_handle_t *zhp, const char *propname, boolean_t received)
1979 {
1980 	zfs_cmd_t zc = {"\0"};
1981 	int ret;
1982 	prop_changelist_t *cl;
1983 	libzfs_handle_t *hdl = zhp->zfs_hdl;
1984 	char errbuf[ERRBUFLEN];
1985 	zfs_prop_t prop;
1986 
1987 	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1988 	    "cannot inherit %s for '%s'"), propname, zhp->zfs_name);
1989 
1990 	zc.zc_cookie = received;
1991 	if ((prop = zfs_name_to_prop(propname)) == ZPROP_USERPROP) {
1992 		/*
1993 		 * For user properties, the amount of work we have to do is very
1994 		 * small, so just do it here.
1995 		 */
1996 		if (!zfs_prop_user(propname)) {
1997 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1998 			    "invalid property"));
1999 			return (zfs_error(hdl, EZFS_BADPROP, errbuf));
2000 		}
2001 
2002 		(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
2003 		(void) strlcpy(zc.zc_value, propname, sizeof (zc.zc_value));
2004 
2005 		if (zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_INHERIT_PROP, &zc) != 0)
2006 			return (zfs_standard_error(hdl, errno, errbuf));
2007 
2008 		(void) get_stats(zhp);
2009 		return (0);
2010 	}
2011 
2012 	/*
2013 	 * Verify that this property is inheritable.
2014 	 */
2015 	if (zfs_prop_readonly(prop))
2016 		return (zfs_error(hdl, EZFS_PROPREADONLY, errbuf));
2017 
2018 	if (!zfs_prop_inheritable(prop) && !received)
2019 		return (zfs_error(hdl, EZFS_PROPNONINHERIT, errbuf));
2020 
2021 	/*
2022 	 * Check to see if the value applies to this type
2023 	 */
2024 	if (!zfs_prop_valid_for_type(prop, zhp->zfs_type, B_FALSE))
2025 		return (zfs_error(hdl, EZFS_PROPTYPE, errbuf));
2026 
2027 	/*
2028 	 * Normalize the name, to get rid of shorthand abbreviations.
2029 	 */
2030 	propname = zfs_prop_to_name(prop);
2031 	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
2032 	(void) strlcpy(zc.zc_value, propname, sizeof (zc.zc_value));
2033 
2034 	if (prop == ZFS_PROP_MOUNTPOINT && getzoneid() == GLOBAL_ZONEID &&
2035 	    zfs_prop_get_int(zhp, ZFS_PROP_ZONED)) {
2036 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2037 		    "dataset is used in a non-global zone"));
2038 		return (zfs_error(hdl, EZFS_ZONED, errbuf));
2039 	}
2040 
2041 	/*
2042 	 * Determine datasets which will be affected by this change, if any.
2043 	 */
2044 	if ((cl = changelist_gather(zhp, prop, 0, 0)) == NULL)
2045 		return (-1);
2046 
2047 	if (prop == ZFS_PROP_MOUNTPOINT && changelist_haszonedchild(cl)) {
2048 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2049 		    "child dataset with inherited mountpoint is used "
2050 		    "in a non-global zone"));
2051 		ret = zfs_error(hdl, EZFS_ZONED, errbuf);
2052 		goto error;
2053 	}
2054 
2055 	if ((ret = changelist_prefix(cl)) != 0)
2056 		goto error;
2057 
2058 	if (zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_INHERIT_PROP, &zc) != 0) {
2059 		changelist_free(cl);
2060 		return (zfs_standard_error(hdl, errno, errbuf));
2061 	} else {
2062 
2063 		if ((ret = changelist_postfix(cl)) != 0)
2064 			goto error;
2065 
2066 		/*
2067 		 * Refresh the statistics so the new property is reflected.
2068 		 */
2069 		(void) get_stats(zhp);
2070 
2071 		/*
2072 		 * Remount the filesystem to propagate the change
2073 		 * if one of the options handled by the generic
2074 		 * Linux namespace layer has been modified.
2075 		 */
2076 		if (zfs_is_namespace_prop(prop) &&
2077 		    zfs_is_mounted(zhp, NULL))
2078 			ret = zfs_mount(zhp, MNTOPT_REMOUNT, 0);
2079 	}
2080 
2081 error:
2082 	changelist_free(cl);
2083 	return (ret);
2084 }
2085 
2086 /*
2087  * True DSL properties are stored in an nvlist.  The following two functions
2088  * extract them appropriately.
2089  */
2090 uint64_t
2091 getprop_uint64(zfs_handle_t *zhp, zfs_prop_t prop, const char **source)
2092 {
2093 	nvlist_t *nv;
2094 	uint64_t value;
2095 
2096 	*source = NULL;
2097 	if (nvlist_lookup_nvlist(zhp->zfs_props,
2098 	    zfs_prop_to_name(prop), &nv) == 0) {
2099 		value = fnvlist_lookup_uint64(nv, ZPROP_VALUE);
2100 		(void) nvlist_lookup_string(nv, ZPROP_SOURCE, source);
2101 	} else {
2102 		verify(!zhp->zfs_props_table ||
2103 		    zhp->zfs_props_table[prop] == B_TRUE);
2104 		value = zfs_prop_default_numeric(prop);
2105 		*source = "";
2106 	}
2107 
2108 	return (value);
2109 }
2110 
2111 static const char *
2112 getprop_string(zfs_handle_t *zhp, zfs_prop_t prop, const char **source)
2113 {
2114 	nvlist_t *nv;
2115 	const char *value;
2116 
2117 	*source = NULL;
2118 	if (nvlist_lookup_nvlist(zhp->zfs_props,
2119 	    zfs_prop_to_name(prop), &nv) == 0) {
2120 		value = fnvlist_lookup_string(nv, ZPROP_VALUE);
2121 		(void) nvlist_lookup_string(nv, ZPROP_SOURCE, source);
2122 	} else {
2123 		verify(!zhp->zfs_props_table ||
2124 		    zhp->zfs_props_table[prop] == B_TRUE);
2125 		value = zfs_prop_default_string(prop);
2126 		*source = "";
2127 	}
2128 
2129 	return (value);
2130 }
2131 
2132 static boolean_t
2133 zfs_is_recvd_props_mode(zfs_handle_t *zhp)
2134 {
2135 	return (zhp->zfs_props != NULL &&
2136 	    zhp->zfs_props == zhp->zfs_recvd_props);
2137 }
2138 
2139 static void
2140 zfs_set_recvd_props_mode(zfs_handle_t *zhp, uintptr_t *cookie)
2141 {
2142 	*cookie = (uintptr_t)zhp->zfs_props;
2143 	zhp->zfs_props = zhp->zfs_recvd_props;
2144 }
2145 
2146 static void
2147 zfs_unset_recvd_props_mode(zfs_handle_t *zhp, uintptr_t *cookie)
2148 {
2149 	zhp->zfs_props = (nvlist_t *)*cookie;
2150 	*cookie = 0;
2151 }
2152 
2153 /*
2154  * Internal function for getting a numeric property.  Both zfs_prop_get() and
2155  * zfs_prop_get_int() are built using this interface.
2156  *
2157  * Certain properties can be overridden using 'mount -o'.  In this case, scan
2158  * the contents of the /proc/self/mounts entry, searching for the
2159  * appropriate options. If they differ from the on-disk values, report the
2160  * current values and mark the source "temporary".
2161  */
2162 static int
2163 get_numeric_property(zfs_handle_t *zhp, zfs_prop_t prop, zprop_source_t *src,
2164     const char **source, uint64_t *val)
2165 {
2166 	zfs_cmd_t zc = {"\0"};
2167 	nvlist_t *zplprops = NULL;
2168 	struct mnttab mnt;
2169 	const char *mntopt_on = NULL;
2170 	const char *mntopt_off = NULL;
2171 	boolean_t received = zfs_is_recvd_props_mode(zhp);
2172 
2173 	*source = NULL;
2174 
2175 	/*
2176 	 * If the property is being fetched for a snapshot, check whether
2177 	 * the property is valid for the snapshot's head dataset type.
2178 	 */
2179 	if (zhp->zfs_type == ZFS_TYPE_SNAPSHOT &&
2180 	    !zfs_prop_valid_for_type(prop, zhp->zfs_head_type, B_TRUE)) {
2181 		*val = zfs_prop_default_numeric(prop);
2182 		return (-1);
2183 	}
2184 
2185 	switch (prop) {
2186 	case ZFS_PROP_ATIME:
2187 		mntopt_on = MNTOPT_ATIME;
2188 		mntopt_off = MNTOPT_NOATIME;
2189 		break;
2190 
2191 	case ZFS_PROP_RELATIME:
2192 		mntopt_on = MNTOPT_RELATIME;
2193 		mntopt_off = MNTOPT_NORELATIME;
2194 		break;
2195 
2196 	case ZFS_PROP_DEVICES:
2197 		mntopt_on = MNTOPT_DEVICES;
2198 		mntopt_off = MNTOPT_NODEVICES;
2199 		break;
2200 
2201 	case ZFS_PROP_EXEC:
2202 		mntopt_on = MNTOPT_EXEC;
2203 		mntopt_off = MNTOPT_NOEXEC;
2204 		break;
2205 
2206 	case ZFS_PROP_READONLY:
2207 		mntopt_on = MNTOPT_RO;
2208 		mntopt_off = MNTOPT_RW;
2209 		break;
2210 
2211 	case ZFS_PROP_SETUID:
2212 		mntopt_on = MNTOPT_SETUID;
2213 		mntopt_off = MNTOPT_NOSETUID;
2214 		break;
2215 
2216 	case ZFS_PROP_XATTR:
2217 		mntopt_on = MNTOPT_XATTR;
2218 		mntopt_off = MNTOPT_NOXATTR;
2219 		break;
2220 
2221 	case ZFS_PROP_NBMAND:
2222 		mntopt_on = MNTOPT_NBMAND;
2223 		mntopt_off = MNTOPT_NONBMAND;
2224 		break;
2225 
2226 	default:
2227 		break;
2228 	}
2229 
2230 	/*
2231 	 * Because looking up the mount options is potentially expensive
2232 	 * (iterating over all of /proc/self/mounts), we defer its
2233 	 * calculation until we're looking up a property which requires
2234 	 * its presence.
2235 	 */
2236 	if (!zhp->zfs_mntcheck &&
2237 	    (mntopt_on != NULL || prop == ZFS_PROP_MOUNTED)) {
2238 		libzfs_handle_t *hdl = zhp->zfs_hdl;
2239 		struct mnttab entry;
2240 
2241 		if (libzfs_mnttab_find(hdl, zhp->zfs_name, &entry) == 0)
2242 			zhp->zfs_mntopts = zfs_strdup(hdl,
2243 			    entry.mnt_mntopts);
2244 
2245 		zhp->zfs_mntcheck = B_TRUE;
2246 	}
2247 
2248 	if (zhp->zfs_mntopts == NULL)
2249 		mnt.mnt_mntopts = (char *)"";
2250 	else
2251 		mnt.mnt_mntopts = zhp->zfs_mntopts;
2252 
2253 	switch (prop) {
2254 	case ZFS_PROP_ATIME:
2255 	case ZFS_PROP_RELATIME:
2256 	case ZFS_PROP_DEVICES:
2257 	case ZFS_PROP_EXEC:
2258 	case ZFS_PROP_READONLY:
2259 	case ZFS_PROP_SETUID:
2260 #ifndef __FreeBSD__
2261 	case ZFS_PROP_XATTR:
2262 #endif
2263 	case ZFS_PROP_NBMAND:
2264 		*val = getprop_uint64(zhp, prop, source);
2265 
2266 		if (received)
2267 			break;
2268 
2269 		if (hasmntopt(&mnt, mntopt_on) && !*val) {
2270 			*val = B_TRUE;
2271 			if (src)
2272 				*src = ZPROP_SRC_TEMPORARY;
2273 		} else if (hasmntopt(&mnt, mntopt_off) && *val) {
2274 			*val = B_FALSE;
2275 			if (src)
2276 				*src = ZPROP_SRC_TEMPORARY;
2277 		}
2278 		break;
2279 
2280 	case ZFS_PROP_CANMOUNT:
2281 	case ZFS_PROP_VOLSIZE:
2282 	case ZFS_PROP_QUOTA:
2283 	case ZFS_PROP_REFQUOTA:
2284 	case ZFS_PROP_RESERVATION:
2285 	case ZFS_PROP_REFRESERVATION:
2286 	case ZFS_PROP_FILESYSTEM_LIMIT:
2287 	case ZFS_PROP_SNAPSHOT_LIMIT:
2288 	case ZFS_PROP_FILESYSTEM_COUNT:
2289 	case ZFS_PROP_SNAPSHOT_COUNT:
2290 		*val = getprop_uint64(zhp, prop, source);
2291 
2292 		if (*source == NULL) {
2293 			/* not default, must be local */
2294 			*source = zhp->zfs_name;
2295 		}
2296 		break;
2297 
2298 	case ZFS_PROP_MOUNTED:
2299 		*val = (zhp->zfs_mntopts != NULL);
2300 		break;
2301 
2302 	case ZFS_PROP_NUMCLONES:
2303 		*val = zhp->zfs_dmustats.dds_num_clones;
2304 		break;
2305 
2306 	case ZFS_PROP_VERSION:
2307 	case ZFS_PROP_NORMALIZE:
2308 	case ZFS_PROP_UTF8ONLY:
2309 	case ZFS_PROP_CASE:
2310 		zcmd_alloc_dst_nvlist(zhp->zfs_hdl, &zc, 0);
2311 
2312 		(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
2313 		if (zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_OBJSET_ZPLPROPS, &zc)) {
2314 			zcmd_free_nvlists(&zc);
2315 			if (prop == ZFS_PROP_VERSION &&
2316 			    zhp->zfs_type == ZFS_TYPE_VOLUME)
2317 				*val = zfs_prop_default_numeric(prop);
2318 			return (-1);
2319 		}
2320 		if (zcmd_read_dst_nvlist(zhp->zfs_hdl, &zc, &zplprops) != 0 ||
2321 		    nvlist_lookup_uint64(zplprops, zfs_prop_to_name(prop),
2322 		    val) != 0) {
2323 			zcmd_free_nvlists(&zc);
2324 			return (-1);
2325 		}
2326 		nvlist_free(zplprops);
2327 		zcmd_free_nvlists(&zc);
2328 		break;
2329 
2330 	case ZFS_PROP_INCONSISTENT:
2331 		*val = zhp->zfs_dmustats.dds_inconsistent;
2332 		break;
2333 
2334 	case ZFS_PROP_REDACTED:
2335 		*val = zhp->zfs_dmustats.dds_redacted;
2336 		break;
2337 
2338 	case ZFS_PROP_GUID:
2339 		if (zhp->zfs_dmustats.dds_guid != 0)
2340 			*val = zhp->zfs_dmustats.dds_guid;
2341 		else
2342 			*val = getprop_uint64(zhp, prop, source);
2343 		break;
2344 
2345 	case ZFS_PROP_CREATETXG:
2346 		/*
2347 		 * We can directly read createtxg property from zfs
2348 		 * handle for Filesystem, Snapshot and ZVOL types.
2349 		 */
2350 		if (((zhp->zfs_type == ZFS_TYPE_FILESYSTEM) ||
2351 		    (zhp->zfs_type == ZFS_TYPE_SNAPSHOT) ||
2352 		    (zhp->zfs_type == ZFS_TYPE_VOLUME)) &&
2353 		    (zhp->zfs_dmustats.dds_creation_txg != 0)) {
2354 			*val = zhp->zfs_dmustats.dds_creation_txg;
2355 			break;
2356 		} else {
2357 			*val = getprop_uint64(zhp, prop, source);
2358 		}
2359 		zfs_fallthrough;
2360 	default:
2361 		switch (zfs_prop_get_type(prop)) {
2362 		case PROP_TYPE_NUMBER:
2363 		case PROP_TYPE_INDEX:
2364 			*val = getprop_uint64(zhp, prop, source);
2365 			/*
2366 			 * If we tried to use a default value for a
2367 			 * readonly property, it means that it was not
2368 			 * present.  Note this only applies to "truly"
2369 			 * readonly properties, not set-once properties
2370 			 * like volblocksize.
2371 			 */
2372 			if (zfs_prop_readonly(prop) &&
2373 			    !zfs_prop_setonce(prop) &&
2374 			    *source != NULL && (*source)[0] == '\0') {
2375 				*source = NULL;
2376 				return (-1);
2377 			}
2378 			break;
2379 
2380 		case PROP_TYPE_STRING:
2381 		default:
2382 			zfs_error_aux(zhp->zfs_hdl, dgettext(TEXT_DOMAIN,
2383 			    "cannot get non-numeric property"));
2384 			return (zfs_error(zhp->zfs_hdl, EZFS_BADPROP,
2385 			    dgettext(TEXT_DOMAIN, "internal error")));
2386 		}
2387 	}
2388 
2389 	return (0);
2390 }
2391 
2392 /*
2393  * Calculate the source type, given the raw source string.
2394  */
2395 static void
2396 get_source(zfs_handle_t *zhp, zprop_source_t *srctype, const char *source,
2397     char *statbuf, size_t statlen)
2398 {
2399 	if (statbuf == NULL ||
2400 	    srctype == NULL || *srctype == ZPROP_SRC_TEMPORARY) {
2401 		return;
2402 	}
2403 
2404 	if (source == NULL) {
2405 		*srctype = ZPROP_SRC_NONE;
2406 	} else if (source[0] == '\0') {
2407 		*srctype = ZPROP_SRC_DEFAULT;
2408 	} else if (strstr(source, ZPROP_SOURCE_VAL_RECVD) != NULL) {
2409 		*srctype = ZPROP_SRC_RECEIVED;
2410 	} else {
2411 		if (strcmp(source, zhp->zfs_name) == 0) {
2412 			*srctype = ZPROP_SRC_LOCAL;
2413 		} else {
2414 			(void) strlcpy(statbuf, source, statlen);
2415 			*srctype = ZPROP_SRC_INHERITED;
2416 		}
2417 	}
2418 
2419 }
2420 
2421 int
2422 zfs_prop_get_recvd(zfs_handle_t *zhp, const char *propname, char *propbuf,
2423     size_t proplen, boolean_t literal)
2424 {
2425 	zfs_prop_t prop;
2426 	int err = 0;
2427 
2428 	if (zhp->zfs_recvd_props == NULL)
2429 		if (get_recvd_props_ioctl(zhp) != 0)
2430 			return (-1);
2431 
2432 	prop = zfs_name_to_prop(propname);
2433 
2434 	if (prop != ZPROP_USERPROP) {
2435 		uintptr_t cookie;
2436 		if (!nvlist_exists(zhp->zfs_recvd_props, propname))
2437 			return (-1);
2438 		zfs_set_recvd_props_mode(zhp, &cookie);
2439 		err = zfs_prop_get(zhp, prop, propbuf, proplen,
2440 		    NULL, NULL, 0, literal);
2441 		zfs_unset_recvd_props_mode(zhp, &cookie);
2442 	} else {
2443 		nvlist_t *propval;
2444 		const char *recvdval;
2445 		if (nvlist_lookup_nvlist(zhp->zfs_recvd_props,
2446 		    propname, &propval) != 0)
2447 			return (-1);
2448 		recvdval = fnvlist_lookup_string(propval, ZPROP_VALUE);
2449 		(void) strlcpy(propbuf, recvdval, proplen);
2450 	}
2451 
2452 	return (err == 0 ? 0 : -1);
2453 }
2454 
2455 static int
2456 get_clones_string(zfs_handle_t *zhp, char *propbuf, size_t proplen)
2457 {
2458 	nvlist_t *value;
2459 	nvpair_t *pair;
2460 
2461 	value = zfs_get_clones_nvl(zhp);
2462 	if (value == NULL || nvlist_empty(value))
2463 		return (-1);
2464 
2465 	propbuf[0] = '\0';
2466 	for (pair = nvlist_next_nvpair(value, NULL); pair != NULL;
2467 	    pair = nvlist_next_nvpair(value, pair)) {
2468 		if (propbuf[0] != '\0')
2469 			(void) strlcat(propbuf, ",", proplen);
2470 		(void) strlcat(propbuf, nvpair_name(pair), proplen);
2471 	}
2472 
2473 	return (0);
2474 }
2475 
2476 struct get_clones_arg {
2477 	uint64_t numclones;
2478 	nvlist_t *value;
2479 	const char *origin;
2480 	char buf[ZFS_MAX_DATASET_NAME_LEN];
2481 };
2482 
2483 static int
2484 get_clones_cb(zfs_handle_t *zhp, void *arg)
2485 {
2486 	struct get_clones_arg *gca = arg;
2487 
2488 	if (gca->numclones == 0) {
2489 		zfs_close(zhp);
2490 		return (0);
2491 	}
2492 
2493 	if (zfs_prop_get(zhp, ZFS_PROP_ORIGIN, gca->buf, sizeof (gca->buf),
2494 	    NULL, NULL, 0, B_TRUE) != 0)
2495 		goto out;
2496 	if (strcmp(gca->buf, gca->origin) == 0) {
2497 		fnvlist_add_boolean(gca->value, zfs_get_name(zhp));
2498 		gca->numclones--;
2499 	}
2500 
2501 out:
2502 	(void) zfs_iter_children_v2(zhp, 0, get_clones_cb, gca);
2503 	zfs_close(zhp);
2504 	return (0);
2505 }
2506 
2507 nvlist_t *
2508 zfs_get_clones_nvl(zfs_handle_t *zhp)
2509 {
2510 	nvlist_t *nv, *value;
2511 
2512 	if (nvlist_lookup_nvlist(zhp->zfs_props,
2513 	    zfs_prop_to_name(ZFS_PROP_CLONES), &nv) != 0) {
2514 		struct get_clones_arg gca;
2515 
2516 		/*
2517 		 * if this is a snapshot, then the kernel wasn't able
2518 		 * to get the clones.  Do it by slowly iterating.
2519 		 */
2520 		if (zhp->zfs_type != ZFS_TYPE_SNAPSHOT)
2521 			return (NULL);
2522 		if (nvlist_alloc(&nv, NV_UNIQUE_NAME, 0) != 0)
2523 			return (NULL);
2524 		if (nvlist_alloc(&value, NV_UNIQUE_NAME, 0) != 0) {
2525 			nvlist_free(nv);
2526 			return (NULL);
2527 		}
2528 
2529 		gca.numclones = zfs_prop_get_int(zhp, ZFS_PROP_NUMCLONES);
2530 		gca.value = value;
2531 		gca.origin = zhp->zfs_name;
2532 
2533 		if (gca.numclones != 0) {
2534 			zfs_handle_t *root;
2535 			char pool[ZFS_MAX_DATASET_NAME_LEN];
2536 			char *cp = pool;
2537 
2538 			/* get the pool name */
2539 			(void) strlcpy(pool, zhp->zfs_name, sizeof (pool));
2540 			(void) strsep(&cp, "/@");
2541 			root = zfs_open(zhp->zfs_hdl, pool,
2542 			    ZFS_TYPE_FILESYSTEM);
2543 			if (root == NULL) {
2544 				nvlist_free(nv);
2545 				nvlist_free(value);
2546 				return (NULL);
2547 			}
2548 
2549 			(void) get_clones_cb(root, &gca);
2550 		}
2551 
2552 		if (gca.numclones != 0 ||
2553 		    nvlist_add_nvlist(nv, ZPROP_VALUE, value) != 0 ||
2554 		    nvlist_add_nvlist(zhp->zfs_props,
2555 		    zfs_prop_to_name(ZFS_PROP_CLONES), nv) != 0) {
2556 			nvlist_free(nv);
2557 			nvlist_free(value);
2558 			return (NULL);
2559 		}
2560 		nvlist_free(nv);
2561 		nvlist_free(value);
2562 		nv = fnvlist_lookup_nvlist(zhp->zfs_props,
2563 		    zfs_prop_to_name(ZFS_PROP_CLONES));
2564 	}
2565 
2566 	return (fnvlist_lookup_nvlist(nv, ZPROP_VALUE));
2567 }
2568 
2569 static int
2570 get_rsnaps_string(zfs_handle_t *zhp, char *propbuf, size_t proplen)
2571 {
2572 	nvlist_t *value;
2573 	uint64_t *snaps;
2574 	uint_t nsnaps;
2575 
2576 	if (nvlist_lookup_nvlist(zhp->zfs_props,
2577 	    zfs_prop_to_name(ZFS_PROP_REDACT_SNAPS), &value) != 0)
2578 		return (-1);
2579 	if (nvlist_lookup_uint64_array(value, ZPROP_VALUE, &snaps,
2580 	    &nsnaps) != 0)
2581 		return (-1);
2582 	if (nsnaps == 0) {
2583 		/* There's no redaction snapshots; pass a special value back */
2584 		(void) snprintf(propbuf, proplen, "none");
2585 		return (0);
2586 	}
2587 	propbuf[0] = '\0';
2588 	for (int i = 0; i < nsnaps; i++) {
2589 		char buf[128];
2590 		if (propbuf[0] != '\0')
2591 			(void) strlcat(propbuf, ",", proplen);
2592 		(void) snprintf(buf, sizeof (buf), "%llu",
2593 		    (u_longlong_t)snaps[i]);
2594 		(void) strlcat(propbuf, buf, proplen);
2595 	}
2596 
2597 	return (0);
2598 }
2599 
2600 /*
2601  * Accepts a property and value and checks that the value
2602  * matches the one found by the channel program. If they are
2603  * not equal, print both of them.
2604  */
2605 static void
2606 zcp_check(zfs_handle_t *zhp, zfs_prop_t prop, uint64_t intval,
2607     const char *strval)
2608 {
2609 	if (!zhp->zfs_hdl->libzfs_prop_debug)
2610 		return;
2611 	int error;
2612 	char *poolname = zhp->zpool_hdl->zpool_name;
2613 	const char *prop_name = zfs_prop_to_name(prop);
2614 	const char *program =
2615 	    "args = ...\n"
2616 	    "ds = args['dataset']\n"
2617 	    "prop = args['property']\n"
2618 	    "value, setpoint = zfs.get_prop(ds, prop)\n"
2619 	    "return {value=value, setpoint=setpoint}\n";
2620 	nvlist_t *outnvl;
2621 	nvlist_t *retnvl;
2622 	nvlist_t *argnvl = fnvlist_alloc();
2623 
2624 	fnvlist_add_string(argnvl, "dataset", zhp->zfs_name);
2625 	fnvlist_add_string(argnvl, "property", zfs_prop_to_name(prop));
2626 
2627 	error = lzc_channel_program_nosync(poolname, program,
2628 	    10 * 1000 * 1000, 10 * 1024 * 1024, argnvl, &outnvl);
2629 
2630 	if (error == 0) {
2631 		retnvl = fnvlist_lookup_nvlist(outnvl, "return");
2632 		if (zfs_prop_get_type(prop) == PROP_TYPE_NUMBER) {
2633 			int64_t ans;
2634 			error = nvlist_lookup_int64(retnvl, "value", &ans);
2635 			if (error != 0) {
2636 				(void) fprintf(stderr, "%s: zcp check error: "
2637 				    "%u\n", prop_name, error);
2638 				return;
2639 			}
2640 			if (ans != intval) {
2641 				(void) fprintf(stderr, "%s: zfs found %llu, "
2642 				    "but zcp found %llu\n", prop_name,
2643 				    (u_longlong_t)intval, (u_longlong_t)ans);
2644 			}
2645 		} else {
2646 			const char *str_ans;
2647 			error = nvlist_lookup_string(retnvl, "value", &str_ans);
2648 			if (error != 0) {
2649 				(void) fprintf(stderr, "%s: zcp check error: "
2650 				    "%u\n", prop_name, error);
2651 				return;
2652 			}
2653 			if (strcmp(strval, str_ans) != 0) {
2654 				(void) fprintf(stderr,
2655 				    "%s: zfs found '%s', but zcp found '%s'\n",
2656 				    prop_name, strval, str_ans);
2657 			}
2658 		}
2659 	} else {
2660 		(void) fprintf(stderr, "%s: zcp check failed, channel program "
2661 		    "error: %u\n", prop_name, error);
2662 	}
2663 	nvlist_free(argnvl);
2664 	nvlist_free(outnvl);
2665 }
2666 
2667 /*
2668  * Retrieve a property from the given object.  If 'literal' is specified, then
2669  * numbers are left as exact values.  Otherwise, numbers are converted to a
2670  * human-readable form.
2671  *
2672  * Returns 0 on success, or -1 on error.
2673  */
2674 int
2675 zfs_prop_get(zfs_handle_t *zhp, zfs_prop_t prop, char *propbuf, size_t proplen,
2676     zprop_source_t *src, char *statbuf, size_t statlen, boolean_t literal)
2677 {
2678 	const char *source = NULL;
2679 	uint64_t val;
2680 	const char *str;
2681 	const char *strval;
2682 	boolean_t received = zfs_is_recvd_props_mode(zhp);
2683 
2684 	/*
2685 	 * Check to see if this property applies to our object
2686 	 */
2687 	if (!zfs_prop_valid_for_type(prop, zhp->zfs_type, B_FALSE))
2688 		return (-1);
2689 
2690 	if (received && zfs_prop_readonly(prop))
2691 		return (-1);
2692 
2693 	if (src)
2694 		*src = ZPROP_SRC_NONE;
2695 
2696 	switch (prop) {
2697 	case ZFS_PROP_CREATION:
2698 		/*
2699 		 * 'creation' is a time_t stored in the statistics.  We convert
2700 		 * this into a string unless 'literal' is specified.
2701 		 */
2702 		{
2703 			val = getprop_uint64(zhp, prop, &source);
2704 			time_t time = (time_t)val;
2705 			struct tm t;
2706 
2707 			if (literal ||
2708 			    localtime_r(&time, &t) == NULL ||
2709 			    strftime(propbuf, proplen, "%a %b %e %k:%M %Y",
2710 			    &t) == 0)
2711 				(void) snprintf(propbuf, proplen, "%llu",
2712 				    (u_longlong_t)val);
2713 		}
2714 		zcp_check(zhp, prop, val, NULL);
2715 		break;
2716 
2717 	case ZFS_PROP_MOUNTPOINT:
2718 		/*
2719 		 * Getting the precise mountpoint can be tricky.
2720 		 *
2721 		 *  - for 'none' or 'legacy', return those values.
2722 		 *  - for inherited mountpoints, we want to take everything
2723 		 *    after our ancestor and append it to the inherited value.
2724 		 *
2725 		 * If the pool has an alternate root, we want to prepend that
2726 		 * root to any values we return.
2727 		 */
2728 
2729 		str = getprop_string(zhp, prop, &source);
2730 
2731 		if (str[0] == '/') {
2732 			char buf[MAXPATHLEN];
2733 			char *root = buf;
2734 			const char *relpath;
2735 
2736 			/*
2737 			 * If we inherit the mountpoint, even from a dataset
2738 			 * with a received value, the source will be the path of
2739 			 * the dataset we inherit from. If source is
2740 			 * ZPROP_SOURCE_VAL_RECVD, the received value is not
2741 			 * inherited.
2742 			 */
2743 			if (strcmp(source, ZPROP_SOURCE_VAL_RECVD) == 0) {
2744 				relpath = "";
2745 			} else {
2746 				relpath = zhp->zfs_name + strlen(source);
2747 				if (relpath[0] == '/')
2748 					relpath++;
2749 			}
2750 
2751 			if ((zpool_get_prop(zhp->zpool_hdl,
2752 			    ZPOOL_PROP_ALTROOT, buf, MAXPATHLEN, NULL,
2753 			    B_FALSE)) || (strcmp(root, "-") == 0))
2754 				root[0] = '\0';
2755 			/*
2756 			 * Special case an alternate root of '/'. This will
2757 			 * avoid having multiple leading slashes in the
2758 			 * mountpoint path.
2759 			 */
2760 			if (strcmp(root, "/") == 0)
2761 				root++;
2762 
2763 			/*
2764 			 * If the mountpoint is '/' then skip over this
2765 			 * if we are obtaining either an alternate root or
2766 			 * an inherited mountpoint.
2767 			 */
2768 			if (str[1] == '\0' && (root[0] != '\0' ||
2769 			    relpath[0] != '\0'))
2770 				str++;
2771 
2772 			if (relpath[0] == '\0')
2773 				(void) snprintf(propbuf, proplen, "%s%s",
2774 				    root, str);
2775 			else
2776 				(void) snprintf(propbuf, proplen, "%s%s%s%s",
2777 				    root, str, relpath[0] == '@' ? "" : "/",
2778 				    relpath);
2779 		} else {
2780 			/* 'legacy' or 'none' */
2781 			(void) strlcpy(propbuf, str, proplen);
2782 		}
2783 		zcp_check(zhp, prop, 0, propbuf);
2784 		break;
2785 
2786 	case ZFS_PROP_ORIGIN:
2787 		if (*zhp->zfs_dmustats.dds_origin != '\0') {
2788 			str = (char *)&zhp->zfs_dmustats.dds_origin;
2789 		} else {
2790 			str = getprop_string(zhp, prop, &source);
2791 		}
2792 		if (str == NULL || *str == '\0')
2793 			str = zfs_prop_default_string(prop);
2794 		if (str == NULL)
2795 			return (-1);
2796 		(void) strlcpy(propbuf, str, proplen);
2797 		zcp_check(zhp, prop, 0, str);
2798 		break;
2799 
2800 	case ZFS_PROP_REDACT_SNAPS:
2801 		if (get_rsnaps_string(zhp, propbuf, proplen) != 0)
2802 			return (-1);
2803 		break;
2804 
2805 	case ZFS_PROP_CLONES:
2806 		if (get_clones_string(zhp, propbuf, proplen) != 0)
2807 			return (-1);
2808 		break;
2809 
2810 	case ZFS_PROP_QUOTA:
2811 	case ZFS_PROP_REFQUOTA:
2812 	case ZFS_PROP_RESERVATION:
2813 	case ZFS_PROP_REFRESERVATION:
2814 
2815 		if (get_numeric_property(zhp, prop, src, &source, &val) != 0)
2816 			return (-1);
2817 		/*
2818 		 * If quota or reservation is 0, we translate this into 'none'
2819 		 * (unless literal is set), and indicate that it's the default
2820 		 * value.  Otherwise, we print the number nicely and indicate
2821 		 * that its set locally.
2822 		 */
2823 		if (val == 0) {
2824 			if (literal)
2825 				(void) strlcpy(propbuf, "0", proplen);
2826 			else
2827 				(void) strlcpy(propbuf, "none", proplen);
2828 		} else {
2829 			if (literal)
2830 				(void) snprintf(propbuf, proplen, "%llu",
2831 				    (u_longlong_t)val);
2832 			else
2833 				zfs_nicebytes(val, propbuf, proplen);
2834 		}
2835 		zcp_check(zhp, prop, val, NULL);
2836 		break;
2837 
2838 	case ZFS_PROP_FILESYSTEM_LIMIT:
2839 	case ZFS_PROP_SNAPSHOT_LIMIT:
2840 	case ZFS_PROP_FILESYSTEM_COUNT:
2841 	case ZFS_PROP_SNAPSHOT_COUNT:
2842 
2843 		if (get_numeric_property(zhp, prop, src, &source, &val) != 0)
2844 			return (-1);
2845 
2846 		/*
2847 		 * If limit is UINT64_MAX, we translate this into 'none', and
2848 		 * indicate that it's the default value. Otherwise, we print
2849 		 * the number nicely and indicate that it's set locally.
2850 		 */
2851 		if (val == UINT64_MAX) {
2852 			(void) strlcpy(propbuf, "none", proplen);
2853 		} else if (literal) {
2854 			(void) snprintf(propbuf, proplen, "%llu",
2855 			    (u_longlong_t)val);
2856 		} else {
2857 			zfs_nicenum(val, propbuf, proplen);
2858 		}
2859 
2860 		zcp_check(zhp, prop, val, NULL);
2861 		break;
2862 
2863 	case ZFS_PROP_REFRATIO:
2864 	case ZFS_PROP_COMPRESSRATIO:
2865 		if (get_numeric_property(zhp, prop, src, &source, &val) != 0)
2866 			return (-1);
2867 		if (literal)
2868 			(void) snprintf(propbuf, proplen, "%llu.%02llu",
2869 			    (u_longlong_t)(val / 100),
2870 			    (u_longlong_t)(val % 100));
2871 		else
2872 			(void) snprintf(propbuf, proplen, "%llu.%02llux",
2873 			    (u_longlong_t)(val / 100),
2874 			    (u_longlong_t)(val % 100));
2875 		zcp_check(zhp, prop, val, NULL);
2876 		break;
2877 
2878 	case ZFS_PROP_TYPE:
2879 		switch (zhp->zfs_type) {
2880 		case ZFS_TYPE_FILESYSTEM:
2881 			str = "filesystem";
2882 			break;
2883 		case ZFS_TYPE_VOLUME:
2884 			str = "volume";
2885 			break;
2886 		case ZFS_TYPE_SNAPSHOT:
2887 			str = "snapshot";
2888 			break;
2889 		case ZFS_TYPE_BOOKMARK:
2890 			str = "bookmark";
2891 			break;
2892 		default:
2893 			abort();
2894 		}
2895 		(void) snprintf(propbuf, proplen, "%s", str);
2896 		zcp_check(zhp, prop, 0, propbuf);
2897 		break;
2898 
2899 	case ZFS_PROP_MOUNTED:
2900 		/*
2901 		 * The 'mounted' property is a pseudo-property that described
2902 		 * whether the filesystem is currently mounted.  Even though
2903 		 * it's a boolean value, the typical values of "on" and "off"
2904 		 * don't make sense, so we translate to "yes" and "no".
2905 		 */
2906 		if (get_numeric_property(zhp, ZFS_PROP_MOUNTED,
2907 		    src, &source, &val) != 0)
2908 			return (-1);
2909 		if (val)
2910 			(void) strlcpy(propbuf, "yes", proplen);
2911 		else
2912 			(void) strlcpy(propbuf, "no", proplen);
2913 		break;
2914 
2915 	case ZFS_PROP_NAME:
2916 		/*
2917 		 * The 'name' property is a pseudo-property derived from the
2918 		 * dataset name.  It is presented as a real property to simplify
2919 		 * consumers.
2920 		 */
2921 		(void) strlcpy(propbuf, zhp->zfs_name, proplen);
2922 		zcp_check(zhp, prop, 0, propbuf);
2923 		break;
2924 
2925 	case ZFS_PROP_MLSLABEL:
2926 		{
2927 #ifdef HAVE_MLSLABEL
2928 			m_label_t *new_sl = NULL;
2929 			char *ascii = NULL;	/* human readable label */
2930 
2931 			(void) strlcpy(propbuf,
2932 			    getprop_string(zhp, prop, &source), proplen);
2933 
2934 			if (literal || (strcasecmp(propbuf,
2935 			    ZFS_MLSLABEL_DEFAULT) == 0))
2936 				break;
2937 
2938 			/*
2939 			 * Try to translate the internal hex string to
2940 			 * human-readable output.  If there are any
2941 			 * problems just use the hex string.
2942 			 */
2943 
2944 			if (str_to_label(propbuf, &new_sl, MAC_LABEL,
2945 			    L_NO_CORRECTION, NULL) == -1) {
2946 				m_label_free(new_sl);
2947 				break;
2948 			}
2949 
2950 			if (label_to_str(new_sl, &ascii, M_LABEL,
2951 			    DEF_NAMES) != 0) {
2952 				if (ascii)
2953 					free(ascii);
2954 				m_label_free(new_sl);
2955 				break;
2956 			}
2957 			m_label_free(new_sl);
2958 
2959 			(void) strlcpy(propbuf, ascii, proplen);
2960 			free(ascii);
2961 #else
2962 			(void) strlcpy(propbuf,
2963 			    getprop_string(zhp, prop, &source), proplen);
2964 #endif /* HAVE_MLSLABEL */
2965 		}
2966 		break;
2967 
2968 	case ZFS_PROP_GUID:
2969 	case ZFS_PROP_KEY_GUID:
2970 	case ZFS_PROP_IVSET_GUID:
2971 	case ZFS_PROP_CREATETXG:
2972 	case ZFS_PROP_OBJSETID:
2973 	case ZFS_PROP_PBKDF2_ITERS:
2974 		/*
2975 		 * These properties are stored as numbers, but they are
2976 		 * identifiers or counters.
2977 		 * We don't want them to be pretty printed, because pretty
2978 		 * printing truncates their values making them useless.
2979 		 */
2980 		if (get_numeric_property(zhp, prop, src, &source, &val) != 0)
2981 			return (-1);
2982 		(void) snprintf(propbuf, proplen, "%llu", (u_longlong_t)val);
2983 		zcp_check(zhp, prop, val, NULL);
2984 		break;
2985 
2986 	case ZFS_PROP_REFERENCED:
2987 	case ZFS_PROP_AVAILABLE:
2988 	case ZFS_PROP_USED:
2989 	case ZFS_PROP_USEDSNAP:
2990 	case ZFS_PROP_USEDDS:
2991 	case ZFS_PROP_USEDREFRESERV:
2992 	case ZFS_PROP_USEDCHILD:
2993 		if (get_numeric_property(zhp, prop, src, &source, &val) != 0)
2994 			return (-1);
2995 		if (literal) {
2996 			(void) snprintf(propbuf, proplen, "%llu",
2997 			    (u_longlong_t)val);
2998 		} else {
2999 			zfs_nicebytes(val, propbuf, proplen);
3000 		}
3001 		zcp_check(zhp, prop, val, NULL);
3002 		break;
3003 
3004 	case ZFS_PROP_SNAPSHOTS_CHANGED:
3005 		{
3006 			if ((get_numeric_property(zhp, prop, src, &source,
3007 			    &val) != 0) || val == 0) {
3008 				return (-1);
3009 			}
3010 
3011 			time_t time = (time_t)val;
3012 			struct tm t;
3013 
3014 			if (literal ||
3015 			    localtime_r(&time, &t) == NULL ||
3016 			    strftime(propbuf, proplen, "%a %b %e %k:%M:%S %Y",
3017 			    &t) == 0)
3018 				(void) snprintf(propbuf, proplen, "%llu",
3019 				    (u_longlong_t)val);
3020 		}
3021 		zcp_check(zhp, prop, val, NULL);
3022 		break;
3023 
3024 	default:
3025 		switch (zfs_prop_get_type(prop)) {
3026 		case PROP_TYPE_NUMBER:
3027 			if (get_numeric_property(zhp, prop, src,
3028 			    &source, &val) != 0) {
3029 				return (-1);
3030 			}
3031 
3032 			if (literal) {
3033 				(void) snprintf(propbuf, proplen, "%llu",
3034 				    (u_longlong_t)val);
3035 			} else {
3036 				zfs_nicenum(val, propbuf, proplen);
3037 			}
3038 			zcp_check(zhp, prop, val, NULL);
3039 			break;
3040 
3041 		case PROP_TYPE_STRING:
3042 			str = getprop_string(zhp, prop, &source);
3043 			if (str == NULL)
3044 				return (-1);
3045 
3046 			(void) strlcpy(propbuf, str, proplen);
3047 			zcp_check(zhp, prop, 0, str);
3048 			break;
3049 
3050 		case PROP_TYPE_INDEX:
3051 			if (get_numeric_property(zhp, prop, src,
3052 			    &source, &val) != 0)
3053 				return (-1);
3054 			if (zfs_prop_index_to_string(prop, val, &strval) != 0)
3055 				return (-1);
3056 
3057 			(void) strlcpy(propbuf, strval, proplen);
3058 			zcp_check(zhp, prop, 0, strval);
3059 			break;
3060 
3061 		default:
3062 			abort();
3063 		}
3064 	}
3065 
3066 	get_source(zhp, src, source, statbuf, statlen);
3067 
3068 	return (0);
3069 }
3070 
3071 /*
3072  * Utility function to get the given numeric property.  Does no validation that
3073  * the given property is the appropriate type; should only be used with
3074  * hard-coded property types.
3075  */
3076 uint64_t
3077 zfs_prop_get_int(zfs_handle_t *zhp, zfs_prop_t prop)
3078 {
3079 	const char *source;
3080 	uint64_t val = 0;
3081 
3082 	(void) get_numeric_property(zhp, prop, NULL, &source, &val);
3083 
3084 	return (val);
3085 }
3086 
3087 static int
3088 zfs_prop_set_int(zfs_handle_t *zhp, zfs_prop_t prop, uint64_t val)
3089 {
3090 	char buf[64];
3091 
3092 	(void) snprintf(buf, sizeof (buf), "%llu", (longlong_t)val);
3093 	return (zfs_prop_set(zhp, zfs_prop_to_name(prop), buf));
3094 }
3095 
3096 /*
3097  * Similar to zfs_prop_get(), but returns the value as an integer.
3098  */
3099 int
3100 zfs_prop_get_numeric(zfs_handle_t *zhp, zfs_prop_t prop, uint64_t *value,
3101     zprop_source_t *src, char *statbuf, size_t statlen)
3102 {
3103 	const char *source;
3104 
3105 	/*
3106 	 * Check to see if this property applies to our object
3107 	 */
3108 	if (!zfs_prop_valid_for_type(prop, zhp->zfs_type, B_FALSE)) {
3109 		return (zfs_error_fmt(zhp->zfs_hdl, EZFS_PROPTYPE,
3110 		    dgettext(TEXT_DOMAIN, "cannot get property '%s'"),
3111 		    zfs_prop_to_name(prop)));
3112 	}
3113 
3114 	if (src)
3115 		*src = ZPROP_SRC_NONE;
3116 
3117 	if (get_numeric_property(zhp, prop, src, &source, value) != 0)
3118 		return (-1);
3119 
3120 	get_source(zhp, src, source, statbuf, statlen);
3121 
3122 	return (0);
3123 }
3124 
3125 #ifdef HAVE_IDMAP
3126 static int
3127 idmap_id_to_numeric_domain_rid(uid_t id, boolean_t isuser,
3128     char **domainp, idmap_rid_t *ridp)
3129 {
3130 	idmap_get_handle_t *get_hdl = NULL;
3131 	idmap_stat status;
3132 	int err = EINVAL;
3133 
3134 	if (idmap_get_create(&get_hdl) != IDMAP_SUCCESS)
3135 		goto out;
3136 
3137 	if (isuser) {
3138 		err = idmap_get_sidbyuid(get_hdl, id,
3139 		    IDMAP_REQ_FLG_USE_CACHE, domainp, ridp, &status);
3140 	} else {
3141 		err = idmap_get_sidbygid(get_hdl, id,
3142 		    IDMAP_REQ_FLG_USE_CACHE, domainp, ridp, &status);
3143 	}
3144 	if (err == IDMAP_SUCCESS &&
3145 	    idmap_get_mappings(get_hdl) == IDMAP_SUCCESS &&
3146 	    status == IDMAP_SUCCESS)
3147 		err = 0;
3148 	else
3149 		err = EINVAL;
3150 out:
3151 	if (get_hdl)
3152 		idmap_get_destroy(get_hdl);
3153 	return (err);
3154 }
3155 #endif /* HAVE_IDMAP */
3156 
3157 /*
3158  * convert the propname into parameters needed by kernel
3159  * Eg: userquota@ahrens -> ZFS_PROP_USERQUOTA, "", 126829
3160  * Eg: userused@matt@domain -> ZFS_PROP_USERUSED, "S-1-123-456", 789
3161  * Eg: groupquota@staff -> ZFS_PROP_GROUPQUOTA, "", 1234
3162  * Eg: groupused@staff -> ZFS_PROP_GROUPUSED, "", 1234
3163  * Eg: projectquota@123 -> ZFS_PROP_PROJECTQUOTA, "", 123
3164  * Eg: projectused@789 -> ZFS_PROP_PROJECTUSED, "", 789
3165  */
3166 static int
3167 userquota_propname_decode(const char *propname, boolean_t zoned,
3168     zfs_userquota_prop_t *typep, char *domain, int domainlen, uint64_t *ridp)
3169 {
3170 	zfs_userquota_prop_t type;
3171 	char *cp;
3172 	boolean_t isuser;
3173 	boolean_t isgroup;
3174 	boolean_t isproject;
3175 	struct passwd *pw;
3176 	struct group *gr;
3177 
3178 	domain[0] = '\0';
3179 
3180 	/* Figure out the property type ({user|group|project}{quota|space}) */
3181 	for (type = 0; type < ZFS_NUM_USERQUOTA_PROPS; type++) {
3182 		if (strncmp(propname, zfs_userquota_prop_prefixes[type],
3183 		    strlen(zfs_userquota_prop_prefixes[type])) == 0)
3184 			break;
3185 	}
3186 	if (type == ZFS_NUM_USERQUOTA_PROPS)
3187 		return (EINVAL);
3188 	*typep = type;
3189 
3190 	isuser = (type == ZFS_PROP_USERQUOTA || type == ZFS_PROP_USERUSED ||
3191 	    type == ZFS_PROP_USEROBJQUOTA ||
3192 	    type == ZFS_PROP_USEROBJUSED);
3193 	isgroup = (type == ZFS_PROP_GROUPQUOTA || type == ZFS_PROP_GROUPUSED ||
3194 	    type == ZFS_PROP_GROUPOBJQUOTA ||
3195 	    type == ZFS_PROP_GROUPOBJUSED);
3196 	isproject = (type == ZFS_PROP_PROJECTQUOTA ||
3197 	    type == ZFS_PROP_PROJECTUSED || type == ZFS_PROP_PROJECTOBJQUOTA ||
3198 	    type == ZFS_PROP_PROJECTOBJUSED);
3199 
3200 	cp = strchr(propname, '@') + 1;
3201 
3202 	if (isuser &&
3203 	    getpwnam_r(cp, &gpwd, rpbuf, sizeof (rpbuf), &pw) == 0 &&
3204 	    pw != NULL) {
3205 		if (zoned && getzoneid() == GLOBAL_ZONEID)
3206 			return (ENOENT);
3207 		*ridp = pw->pw_uid;
3208 	} else if (isgroup &&
3209 	    getgrnam_r(cp, &ggrp, rpbuf, sizeof (rpbuf), &gr) == 0 &&
3210 	    gr != NULL) {
3211 		if (zoned && getzoneid() == GLOBAL_ZONEID)
3212 			return (ENOENT);
3213 		*ridp = gr->gr_gid;
3214 	} else if (!isproject && strchr(cp, '@')) {
3215 #ifdef HAVE_IDMAP
3216 		/*
3217 		 * It's a SID name (eg "user@domain") that needs to be
3218 		 * turned into S-1-domainID-RID.
3219 		 */
3220 		directory_error_t e;
3221 		char *numericsid = NULL;
3222 		char *end;
3223 
3224 		if (zoned && getzoneid() == GLOBAL_ZONEID)
3225 			return (ENOENT);
3226 		if (isuser) {
3227 			e = directory_sid_from_user_name(NULL,
3228 			    cp, &numericsid);
3229 		} else {
3230 			e = directory_sid_from_group_name(NULL,
3231 			    cp, &numericsid);
3232 		}
3233 		if (e != NULL) {
3234 			directory_error_free(e);
3235 			return (ENOENT);
3236 		}
3237 		if (numericsid == NULL)
3238 			return (ENOENT);
3239 		cp = numericsid;
3240 		(void) strlcpy(domain, cp, domainlen);
3241 		cp = strrchr(domain, '-');
3242 		*cp = '\0';
3243 		cp++;
3244 
3245 		errno = 0;
3246 		*ridp = strtoull(cp, &end, 10);
3247 		free(numericsid);
3248 
3249 		if (errno != 0 || *end != '\0')
3250 			return (EINVAL);
3251 #else
3252 		(void) domainlen;
3253 		return (ENOSYS);
3254 #endif /* HAVE_IDMAP */
3255 	} else {
3256 		/* It's a user/group/project ID (eg "12345"). */
3257 		uid_t id;
3258 		char *end;
3259 		id = strtoul(cp, &end, 10);
3260 		if (*end != '\0')
3261 			return (EINVAL);
3262 		if (id > MAXUID && !isproject) {
3263 #ifdef HAVE_IDMAP
3264 			/* It's an ephemeral ID. */
3265 			idmap_rid_t rid;
3266 			char *mapdomain;
3267 
3268 			if (idmap_id_to_numeric_domain_rid(id, isuser,
3269 			    &mapdomain, &rid) != 0)
3270 				return (ENOENT);
3271 			(void) strlcpy(domain, mapdomain, domainlen);
3272 			*ridp = rid;
3273 #else
3274 			return (ENOSYS);
3275 #endif /* HAVE_IDMAP */
3276 		} else {
3277 			*ridp = id;
3278 		}
3279 	}
3280 
3281 	return (0);
3282 }
3283 
3284 static int
3285 zfs_prop_get_userquota_common(zfs_handle_t *zhp, const char *propname,
3286     uint64_t *propvalue, zfs_userquota_prop_t *typep)
3287 {
3288 	int err;
3289 	zfs_cmd_t zc = {"\0"};
3290 
3291 	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
3292 
3293 	err = userquota_propname_decode(propname,
3294 	    zfs_prop_get_int(zhp, ZFS_PROP_ZONED),
3295 	    typep, zc.zc_value, sizeof (zc.zc_value), &zc.zc_guid);
3296 	zc.zc_objset_type = *typep;
3297 	if (err)
3298 		return (err);
3299 
3300 	err = zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_USERSPACE_ONE, &zc);
3301 	if (err)
3302 		return (err);
3303 
3304 	*propvalue = zc.zc_cookie;
3305 	return (0);
3306 }
3307 
3308 int
3309 zfs_prop_get_userquota_int(zfs_handle_t *zhp, const char *propname,
3310     uint64_t *propvalue)
3311 {
3312 	zfs_userquota_prop_t type;
3313 
3314 	return (zfs_prop_get_userquota_common(zhp, propname, propvalue,
3315 	    &type));
3316 }
3317 
3318 int
3319 zfs_prop_get_userquota(zfs_handle_t *zhp, const char *propname,
3320     char *propbuf, int proplen, boolean_t literal)
3321 {
3322 	int err;
3323 	uint64_t propvalue;
3324 	zfs_userquota_prop_t type;
3325 
3326 	err = zfs_prop_get_userquota_common(zhp, propname, &propvalue,
3327 	    &type);
3328 
3329 	if (err)
3330 		return (err);
3331 
3332 	if (literal) {
3333 		(void) snprintf(propbuf, proplen, "%llu",
3334 		    (u_longlong_t)propvalue);
3335 	} else if (propvalue == 0 &&
3336 	    (type == ZFS_PROP_USERQUOTA || type == ZFS_PROP_GROUPQUOTA ||
3337 	    type == ZFS_PROP_USEROBJQUOTA || type == ZFS_PROP_GROUPOBJQUOTA ||
3338 	    type == ZFS_PROP_PROJECTQUOTA ||
3339 	    type == ZFS_PROP_PROJECTOBJQUOTA)) {
3340 		(void) strlcpy(propbuf, "none", proplen);
3341 	} else if (type == ZFS_PROP_USERQUOTA || type == ZFS_PROP_GROUPQUOTA ||
3342 	    type == ZFS_PROP_USERUSED || type == ZFS_PROP_GROUPUSED ||
3343 	    type == ZFS_PROP_PROJECTUSED || type == ZFS_PROP_PROJECTQUOTA) {
3344 		zfs_nicebytes(propvalue, propbuf, proplen);
3345 	} else {
3346 		zfs_nicenum(propvalue, propbuf, proplen);
3347 	}
3348 	return (0);
3349 }
3350 
3351 /*
3352  * propname must start with "written@" or "written#".
3353  */
3354 int
3355 zfs_prop_get_written_int(zfs_handle_t *zhp, const char *propname,
3356     uint64_t *propvalue)
3357 {
3358 	int err;
3359 	zfs_cmd_t zc = {"\0"};
3360 	const char *snapname;
3361 
3362 	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
3363 
3364 	assert(zfs_prop_written(propname));
3365 	snapname = propname + strlen("written@");
3366 	if (strchr(snapname, '@') != NULL || strchr(snapname, '#') != NULL) {
3367 		/* full snapshot or bookmark name specified */
3368 		(void) strlcpy(zc.zc_value, snapname, sizeof (zc.zc_value));
3369 	} else {
3370 		/* snapname is the short name, append it to zhp's fsname */
3371 		char *cp;
3372 
3373 		(void) strlcpy(zc.zc_value, zhp->zfs_name,
3374 		    sizeof (zc.zc_value));
3375 		cp = strchr(zc.zc_value, '@');
3376 		if (cp != NULL)
3377 			*cp = '\0';
3378 		(void) strlcat(zc.zc_value, snapname - 1, sizeof (zc.zc_value));
3379 	}
3380 
3381 	err = zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_SPACE_WRITTEN, &zc);
3382 	if (err)
3383 		return (err);
3384 
3385 	*propvalue = zc.zc_cookie;
3386 	return (0);
3387 }
3388 
3389 int
3390 zfs_prop_get_written(zfs_handle_t *zhp, const char *propname,
3391     char *propbuf, int proplen, boolean_t literal)
3392 {
3393 	int err;
3394 	uint64_t propvalue;
3395 
3396 	err = zfs_prop_get_written_int(zhp, propname, &propvalue);
3397 
3398 	if (err)
3399 		return (err);
3400 
3401 	if (literal) {
3402 		(void) snprintf(propbuf, proplen, "%llu",
3403 		    (u_longlong_t)propvalue);
3404 	} else {
3405 		zfs_nicebytes(propvalue, propbuf, proplen);
3406 	}
3407 
3408 	return (0);
3409 }
3410 
3411 /*
3412  * Returns the name of the given zfs handle.
3413  */
3414 const char *
3415 zfs_get_name(const zfs_handle_t *zhp)
3416 {
3417 	return (zhp->zfs_name);
3418 }
3419 
3420 /*
3421  * Returns the name of the parent pool for the given zfs handle.
3422  */
3423 const char *
3424 zfs_get_pool_name(const zfs_handle_t *zhp)
3425 {
3426 	return (zhp->zpool_hdl->zpool_name);
3427 }
3428 
3429 /*
3430  * Returns the type of the given zfs handle.
3431  */
3432 zfs_type_t
3433 zfs_get_type(const zfs_handle_t *zhp)
3434 {
3435 	return (zhp->zfs_type);
3436 }
3437 
3438 /*
3439  * Returns the type of the given zfs handle,
3440  * or, if a snapshot, the type of the snapshotted dataset.
3441  */
3442 zfs_type_t
3443 zfs_get_underlying_type(const zfs_handle_t *zhp)
3444 {
3445 	return (zhp->zfs_head_type);
3446 }
3447 
3448 /*
3449  * Is one dataset name a child dataset of another?
3450  *
3451  * Needs to handle these cases:
3452  * Dataset 1	"a/foo"		"a/foo"		"a/foo"		"a/foo"
3453  * Dataset 2	"a/fo"		"a/foobar"	"a/bar/baz"	"a/foo/bar"
3454  * Descendant?	No.		No.		No.		Yes.
3455  */
3456 static boolean_t
3457 is_descendant(const char *ds1, const char *ds2)
3458 {
3459 	size_t d1len = strlen(ds1);
3460 
3461 	/* ds2 can't be a descendant if it's smaller */
3462 	if (strlen(ds2) < d1len)
3463 		return (B_FALSE);
3464 
3465 	/* otherwise, compare strings and verify that there's a '/' char */
3466 	return (ds2[d1len] == '/' && (strncmp(ds1, ds2, d1len) == 0));
3467 }
3468 
3469 /*
3470  * Given a complete name, return just the portion that refers to the parent.
3471  * Will return -1 if there is no parent (path is just the name of the
3472  * pool).
3473  */
3474 static int
3475 parent_name(const char *path, char *buf, size_t buflen)
3476 {
3477 	char *slashp;
3478 
3479 	(void) strlcpy(buf, path, buflen);
3480 
3481 	if ((slashp = strrchr(buf, '/')) == NULL)
3482 		return (-1);
3483 	*slashp = '\0';
3484 
3485 	return (0);
3486 }
3487 
3488 int
3489 zfs_parent_name(zfs_handle_t *zhp, char *buf, size_t buflen)
3490 {
3491 	return (parent_name(zfs_get_name(zhp), buf, buflen));
3492 }
3493 
3494 /*
3495  * If accept_ancestor is false, then check to make sure that the given path has
3496  * a parent, and that it exists.  If accept_ancestor is true, then find the
3497  * closest existing ancestor for the given path.  In prefixlen return the
3498  * length of already existing prefix of the given path.  We also fetch the
3499  * 'zoned' property, which is used to validate property settings when creating
3500  * new datasets.
3501  */
3502 static int
3503 check_parents(libzfs_handle_t *hdl, const char *path, uint64_t *zoned,
3504     boolean_t accept_ancestor, int *prefixlen)
3505 {
3506 	zfs_cmd_t zc = {"\0"};
3507 	char parent[ZFS_MAX_DATASET_NAME_LEN];
3508 	char *slash;
3509 	zfs_handle_t *zhp;
3510 	char errbuf[ERRBUFLEN];
3511 	uint64_t is_zoned;
3512 
3513 	(void) snprintf(errbuf, sizeof (errbuf),
3514 	    dgettext(TEXT_DOMAIN, "cannot create '%s'"), path);
3515 
3516 	/* get parent, and check to see if this is just a pool */
3517 	if (parent_name(path, parent, sizeof (parent)) != 0) {
3518 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3519 		    "missing dataset name"));
3520 		return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
3521 	}
3522 
3523 	/* check to see if the pool exists */
3524 	if ((slash = strchr(parent, '/')) == NULL)
3525 		slash = parent + strlen(parent);
3526 	(void) strlcpy(zc.zc_name, parent,
3527 	    MIN(sizeof (zc.zc_name), slash - parent + 1));
3528 	if (zfs_ioctl(hdl, ZFS_IOC_OBJSET_STATS, &zc) != 0 &&
3529 	    errno == ENOENT) {
3530 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3531 		    "no such pool '%s'"), zc.zc_name);
3532 		return (zfs_error(hdl, EZFS_NOENT, errbuf));
3533 	}
3534 
3535 	/* check to see if the parent dataset exists */
3536 	while ((zhp = make_dataset_handle(hdl, parent)) == NULL) {
3537 		if (errno == ENOENT && accept_ancestor) {
3538 			/*
3539 			 * Go deeper to find an ancestor, give up on top level.
3540 			 */
3541 			if (parent_name(parent, parent, sizeof (parent)) != 0) {
3542 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3543 				    "no such pool '%s'"), zc.zc_name);
3544 				return (zfs_error(hdl, EZFS_NOENT, errbuf));
3545 			}
3546 		} else if (errno == ENOENT) {
3547 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3548 			    "parent does not exist"));
3549 			return (zfs_error(hdl, EZFS_NOENT, errbuf));
3550 		} else
3551 			return (zfs_standard_error(hdl, errno, errbuf));
3552 	}
3553 
3554 	is_zoned = zfs_prop_get_int(zhp, ZFS_PROP_ZONED);
3555 	if (zoned != NULL)
3556 		*zoned = is_zoned;
3557 
3558 	/* we are in a non-global zone, but parent is in the global zone */
3559 	if (getzoneid() != GLOBAL_ZONEID && !is_zoned) {
3560 		(void) zfs_standard_error(hdl, EPERM, errbuf);
3561 		zfs_close(zhp);
3562 		return (-1);
3563 	}
3564 
3565 	/* make sure parent is a filesystem */
3566 	if (zfs_get_type(zhp) != ZFS_TYPE_FILESYSTEM) {
3567 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3568 		    "parent is not a filesystem"));
3569 		(void) zfs_error(hdl, EZFS_BADTYPE, errbuf);
3570 		zfs_close(zhp);
3571 		return (-1);
3572 	}
3573 
3574 	zfs_close(zhp);
3575 	if (prefixlen != NULL)
3576 		*prefixlen = strlen(parent);
3577 	return (0);
3578 }
3579 
3580 /*
3581  * Finds whether the dataset of the given type(s) exists.
3582  */
3583 boolean_t
3584 zfs_dataset_exists(libzfs_handle_t *hdl, const char *path, zfs_type_t types)
3585 {
3586 	zfs_handle_t *zhp;
3587 
3588 	if (!zfs_validate_name(hdl, path, types, B_FALSE))
3589 		return (B_FALSE);
3590 
3591 	/*
3592 	 * Try to get stats for the dataset, which will tell us if it exists.
3593 	 */
3594 	if ((zhp = make_dataset_handle(hdl, path)) != NULL) {
3595 		int ds_type = zhp->zfs_type;
3596 
3597 		zfs_close(zhp);
3598 		if (types & ds_type)
3599 			return (B_TRUE);
3600 	}
3601 	return (B_FALSE);
3602 }
3603 
3604 /*
3605  * Given a path to 'target', create all the ancestors between
3606  * the prefixlen portion of the path, and the target itself.
3607  * Fail if the initial prefixlen-ancestor does not already exist.
3608  */
3609 int
3610 create_parents(libzfs_handle_t *hdl, char *target, int prefixlen)
3611 {
3612 	zfs_handle_t *h;
3613 	char *cp;
3614 	const char *opname;
3615 
3616 	/* make sure prefix exists */
3617 	cp = target + prefixlen;
3618 	if (*cp != '/') {
3619 		assert(strchr(cp, '/') == NULL);
3620 		h = zfs_open(hdl, target, ZFS_TYPE_FILESYSTEM);
3621 	} else {
3622 		*cp = '\0';
3623 		h = zfs_open(hdl, target, ZFS_TYPE_FILESYSTEM);
3624 		*cp = '/';
3625 	}
3626 	if (h == NULL)
3627 		return (-1);
3628 	zfs_close(h);
3629 
3630 	/*
3631 	 * Attempt to create, mount, and share any ancestor filesystems,
3632 	 * up to the prefixlen-long one.
3633 	 */
3634 	for (cp = target + prefixlen + 1;
3635 	    (cp = strchr(cp, '/')) != NULL; *cp = '/', cp++) {
3636 
3637 		*cp = '\0';
3638 
3639 		h = make_dataset_handle(hdl, target);
3640 		if (h) {
3641 			/* it already exists, nothing to do here */
3642 			zfs_close(h);
3643 			continue;
3644 		}
3645 
3646 		if (zfs_create(hdl, target, ZFS_TYPE_FILESYSTEM,
3647 		    NULL) != 0) {
3648 			opname = dgettext(TEXT_DOMAIN, "create");
3649 			goto ancestorerr;
3650 		}
3651 
3652 		h = zfs_open(hdl, target, ZFS_TYPE_FILESYSTEM);
3653 		if (h == NULL) {
3654 			opname = dgettext(TEXT_DOMAIN, "open");
3655 			goto ancestorerr;
3656 		}
3657 
3658 		if (zfs_mount(h, NULL, 0) != 0) {
3659 			opname = dgettext(TEXT_DOMAIN, "mount");
3660 			goto ancestorerr;
3661 		}
3662 
3663 		if (zfs_share(h, NULL) != 0) {
3664 			opname = dgettext(TEXT_DOMAIN, "share");
3665 			goto ancestorerr;
3666 		}
3667 
3668 		zfs_close(h);
3669 	}
3670 	zfs_commit_shares(NULL);
3671 
3672 	return (0);
3673 
3674 ancestorerr:
3675 	zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3676 	    "failed to %s ancestor '%s'"), opname, target);
3677 	return (-1);
3678 }
3679 
3680 /*
3681  * Creates non-existing ancestors of the given path.
3682  */
3683 int
3684 zfs_create_ancestors(libzfs_handle_t *hdl, const char *path)
3685 {
3686 	int prefix;
3687 	char *path_copy;
3688 	char errbuf[ERRBUFLEN];
3689 	int rc = 0;
3690 
3691 	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
3692 	    "cannot create '%s'"), path);
3693 
3694 	/*
3695 	 * Check that we are not passing the nesting limit
3696 	 * before we start creating any ancestors.
3697 	 */
3698 	if (dataset_nestcheck(path) != 0) {
3699 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3700 		    "maximum name nesting depth exceeded"));
3701 		return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
3702 	}
3703 
3704 	if (check_parents(hdl, path, NULL, B_TRUE, &prefix) != 0)
3705 		return (-1);
3706 
3707 	if ((path_copy = strdup(path)) != NULL) {
3708 		rc = create_parents(hdl, path_copy, prefix);
3709 		free(path_copy);
3710 	}
3711 	if (path_copy == NULL || rc != 0)
3712 		return (-1);
3713 
3714 	return (0);
3715 }
3716 
3717 /*
3718  * Create a new filesystem or volume.
3719  */
3720 int
3721 zfs_create(libzfs_handle_t *hdl, const char *path, zfs_type_t type,
3722     nvlist_t *props)
3723 {
3724 	int ret;
3725 	uint64_t size = 0;
3726 	uint64_t blocksize = zfs_prop_default_numeric(ZFS_PROP_VOLBLOCKSIZE);
3727 	uint64_t zoned;
3728 	enum lzc_dataset_type ost;
3729 	zpool_handle_t *zpool_handle;
3730 	uint8_t *wkeydata = NULL;
3731 	uint_t wkeylen = 0;
3732 	char errbuf[ERRBUFLEN];
3733 	char parent[ZFS_MAX_DATASET_NAME_LEN];
3734 
3735 	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
3736 	    "cannot create '%s'"), path);
3737 
3738 	/* validate the path, taking care to note the extended error message */
3739 	if (!zfs_validate_name(hdl, path, type, B_TRUE))
3740 		return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
3741 
3742 	if (dataset_nestcheck(path) != 0) {
3743 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3744 		    "maximum name nesting depth exceeded"));
3745 		return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
3746 	}
3747 
3748 	/* validate parents exist */
3749 	if (check_parents(hdl, path, &zoned, B_FALSE, NULL) != 0)
3750 		return (-1);
3751 
3752 	/*
3753 	 * The failure modes when creating a dataset of a different type over
3754 	 * one that already exists is a little strange.  In particular, if you
3755 	 * try to create a dataset on top of an existing dataset, the ioctl()
3756 	 * will return ENOENT, not EEXIST.  To prevent this from happening, we
3757 	 * first try to see if the dataset exists.
3758 	 */
3759 	if (zfs_dataset_exists(hdl, path, ZFS_TYPE_DATASET)) {
3760 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3761 		    "dataset already exists"));
3762 		return (zfs_error(hdl, EZFS_EXISTS, errbuf));
3763 	}
3764 
3765 	if (type == ZFS_TYPE_VOLUME)
3766 		ost = LZC_DATSET_TYPE_ZVOL;
3767 	else
3768 		ost = LZC_DATSET_TYPE_ZFS;
3769 
3770 	/* open zpool handle for prop validation */
3771 	char pool_path[ZFS_MAX_DATASET_NAME_LEN];
3772 	(void) strlcpy(pool_path, path, sizeof (pool_path));
3773 
3774 	/* truncate pool_path at first slash */
3775 	char *p = strchr(pool_path, '/');
3776 	if (p != NULL)
3777 		*p = '\0';
3778 
3779 	if ((zpool_handle = zpool_open(hdl, pool_path)) == NULL)
3780 		return (-1);
3781 
3782 	if (props && (props = zfs_valid_proplist(hdl, type, props,
3783 	    zoned, NULL, zpool_handle, B_TRUE, errbuf)) == 0) {
3784 		zpool_close(zpool_handle);
3785 		return (-1);
3786 	}
3787 	zpool_close(zpool_handle);
3788 
3789 	if (type == ZFS_TYPE_VOLUME) {
3790 		/*
3791 		 * If we are creating a volume, the size and block size must
3792 		 * satisfy a few restraints.  First, the blocksize must be a
3793 		 * valid block size between SPA_{MIN,MAX}BLOCKSIZE.  Second, the
3794 		 * volsize must be a multiple of the block size, and cannot be
3795 		 * zero.
3796 		 */
3797 		if (props == NULL || nvlist_lookup_uint64(props,
3798 		    zfs_prop_to_name(ZFS_PROP_VOLSIZE), &size) != 0) {
3799 			nvlist_free(props);
3800 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3801 			    "missing volume size"));
3802 			return (zfs_error(hdl, EZFS_BADPROP, errbuf));
3803 		}
3804 
3805 		if ((ret = nvlist_lookup_uint64(props,
3806 		    zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
3807 		    &blocksize)) != 0) {
3808 			if (ret == ENOENT) {
3809 				blocksize = zfs_prop_default_numeric(
3810 				    ZFS_PROP_VOLBLOCKSIZE);
3811 			} else {
3812 				nvlist_free(props);
3813 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3814 				    "missing volume block size"));
3815 				return (zfs_error(hdl, EZFS_BADPROP, errbuf));
3816 			}
3817 		}
3818 
3819 		if (size == 0) {
3820 			nvlist_free(props);
3821 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3822 			    "volume size cannot be zero"));
3823 			return (zfs_error(hdl, EZFS_BADPROP, errbuf));
3824 		}
3825 
3826 		if (size % blocksize != 0) {
3827 			nvlist_free(props);
3828 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3829 			    "volume size must be a multiple of volume block "
3830 			    "size"));
3831 			return (zfs_error(hdl, EZFS_BADPROP, errbuf));
3832 		}
3833 	}
3834 
3835 	(void) parent_name(path, parent, sizeof (parent));
3836 	if (zfs_crypto_create(hdl, parent, props, NULL, B_TRUE,
3837 	    &wkeydata, &wkeylen) != 0) {
3838 		nvlist_free(props);
3839 		return (zfs_error(hdl, EZFS_CRYPTOFAILED, errbuf));
3840 	}
3841 
3842 	/* create the dataset */
3843 	ret = lzc_create(path, ost, props, wkeydata, wkeylen);
3844 	nvlist_free(props);
3845 	if (wkeydata != NULL)
3846 		free(wkeydata);
3847 
3848 	/* check for failure */
3849 	if (ret != 0) {
3850 		switch (errno) {
3851 		case ENOENT:
3852 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3853 			    "no such parent '%s'"), parent);
3854 			return (zfs_error(hdl, EZFS_NOENT, errbuf));
3855 
3856 		case ENOTSUP:
3857 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3858 			    "pool must be upgraded to set this "
3859 			    "property or value"));
3860 			return (zfs_error(hdl, EZFS_BADVERSION, errbuf));
3861 
3862 		case EACCES:
3863 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3864 			    "encryption root's key is not loaded "
3865 			    "or provided"));
3866 			return (zfs_error(hdl, EZFS_CRYPTOFAILED, errbuf));
3867 
3868 		case ERANGE:
3869 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3870 			    "invalid property value(s) specified"));
3871 			return (zfs_error(hdl, EZFS_BADPROP, errbuf));
3872 #ifdef _ILP32
3873 		case EOVERFLOW:
3874 			/*
3875 			 * This platform can't address a volume this big.
3876 			 */
3877 			if (type == ZFS_TYPE_VOLUME)
3878 				return (zfs_error(hdl, EZFS_VOLTOOBIG,
3879 				    errbuf));
3880 			zfs_fallthrough;
3881 #endif
3882 		default:
3883 			return (zfs_standard_error(hdl, errno, errbuf));
3884 		}
3885 	}
3886 
3887 	return (0);
3888 }
3889 
3890 /*
3891  * Destroys the given dataset.  The caller must make sure that the filesystem
3892  * isn't mounted, and that there are no active dependents. If the file system
3893  * does not exist this function does nothing.
3894  */
3895 int
3896 zfs_destroy(zfs_handle_t *zhp, boolean_t defer)
3897 {
3898 	int error;
3899 
3900 	if (zhp->zfs_type != ZFS_TYPE_SNAPSHOT && defer)
3901 		return (EINVAL);
3902 
3903 	if (zhp->zfs_type == ZFS_TYPE_BOOKMARK) {
3904 		nvlist_t *nv = fnvlist_alloc();
3905 		fnvlist_add_boolean(nv, zhp->zfs_name);
3906 		error = lzc_destroy_bookmarks(nv, NULL);
3907 		fnvlist_free(nv);
3908 		if (error != 0) {
3909 			return (zfs_standard_error_fmt(zhp->zfs_hdl, error,
3910 			    dgettext(TEXT_DOMAIN, "cannot destroy '%s'"),
3911 			    zhp->zfs_name));
3912 		}
3913 		return (0);
3914 	}
3915 
3916 	if (zhp->zfs_type == ZFS_TYPE_SNAPSHOT) {
3917 		nvlist_t *nv = fnvlist_alloc();
3918 		fnvlist_add_boolean(nv, zhp->zfs_name);
3919 		error = lzc_destroy_snaps(nv, defer, NULL);
3920 		fnvlist_free(nv);
3921 	} else {
3922 		error = lzc_destroy(zhp->zfs_name);
3923 	}
3924 
3925 	if (error != 0 && error != ENOENT) {
3926 		return (zfs_standard_error_fmt(zhp->zfs_hdl, errno,
3927 		    dgettext(TEXT_DOMAIN, "cannot destroy '%s'"),
3928 		    zhp->zfs_name));
3929 	}
3930 
3931 	remove_mountpoint(zhp);
3932 
3933 	return (0);
3934 }
3935 
3936 struct destroydata {
3937 	nvlist_t *nvl;
3938 	const char *snapname;
3939 };
3940 
3941 static int
3942 zfs_check_snap_cb(zfs_handle_t *zhp, void *arg)
3943 {
3944 	struct destroydata *dd = arg;
3945 	char name[ZFS_MAX_DATASET_NAME_LEN];
3946 	int rv = 0;
3947 
3948 	if (snprintf(name, sizeof (name), "%s@%s", zhp->zfs_name,
3949 	    dd->snapname) >= sizeof (name))
3950 		return (EINVAL);
3951 
3952 	if (lzc_exists(name))
3953 		fnvlist_add_boolean(dd->nvl, name);
3954 
3955 	rv = zfs_iter_filesystems_v2(zhp, 0, zfs_check_snap_cb, dd);
3956 	zfs_close(zhp);
3957 	return (rv);
3958 }
3959 
3960 /*
3961  * Destroys all snapshots with the given name in zhp & descendants.
3962  */
3963 int
3964 zfs_destroy_snaps(zfs_handle_t *zhp, char *snapname, boolean_t defer)
3965 {
3966 	int ret;
3967 	struct destroydata dd = { 0 };
3968 
3969 	dd.snapname = snapname;
3970 	dd.nvl = fnvlist_alloc();
3971 	(void) zfs_check_snap_cb(zfs_handle_dup(zhp), &dd);
3972 
3973 	if (nvlist_empty(dd.nvl)) {
3974 		ret = zfs_standard_error_fmt(zhp->zfs_hdl, ENOENT,
3975 		    dgettext(TEXT_DOMAIN, "cannot destroy '%s@%s'"),
3976 		    zhp->zfs_name, snapname);
3977 	} else {
3978 		ret = zfs_destroy_snaps_nvl(zhp->zfs_hdl, dd.nvl, defer);
3979 	}
3980 	fnvlist_free(dd.nvl);
3981 	return (ret);
3982 }
3983 
3984 /*
3985  * Destroys all the snapshots named in the nvlist.
3986  */
3987 int
3988 zfs_destroy_snaps_nvl(libzfs_handle_t *hdl, nvlist_t *snaps, boolean_t defer)
3989 {
3990 	nvlist_t *errlist = NULL;
3991 	nvpair_t *pair;
3992 
3993 	int ret = zfs_destroy_snaps_nvl_os(hdl, snaps);
3994 	if (ret != 0)
3995 		return (ret);
3996 
3997 	ret = lzc_destroy_snaps(snaps, defer, &errlist);
3998 
3999 	if (ret == 0) {
4000 		nvlist_free(errlist);
4001 		return (0);
4002 	}
4003 
4004 	if (nvlist_empty(errlist)) {
4005 		char errbuf[ERRBUFLEN];
4006 		(void) snprintf(errbuf, sizeof (errbuf),
4007 		    dgettext(TEXT_DOMAIN, "cannot destroy snapshots"));
4008 
4009 		ret = zfs_standard_error(hdl, ret, errbuf);
4010 	}
4011 	for (pair = nvlist_next_nvpair(errlist, NULL);
4012 	    pair != NULL; pair = nvlist_next_nvpair(errlist, pair)) {
4013 		char errbuf[ERRBUFLEN];
4014 		(void) snprintf(errbuf, sizeof (errbuf),
4015 		    dgettext(TEXT_DOMAIN, "cannot destroy snapshot %s"),
4016 		    nvpair_name(pair));
4017 
4018 		switch (fnvpair_value_int32(pair)) {
4019 		case EEXIST:
4020 			zfs_error_aux(hdl,
4021 			    dgettext(TEXT_DOMAIN, "snapshot is cloned"));
4022 			ret = zfs_error(hdl, EZFS_EXISTS, errbuf);
4023 			break;
4024 		default:
4025 			ret = zfs_standard_error(hdl, errno, errbuf);
4026 			break;
4027 		}
4028 	}
4029 
4030 	nvlist_free(errlist);
4031 	return (ret);
4032 }
4033 
4034 /*
4035  * Clones the given dataset.  The target must be of the same type as the source.
4036  */
4037 int
4038 zfs_clone(zfs_handle_t *zhp, const char *target, nvlist_t *props)
4039 {
4040 	char parent[ZFS_MAX_DATASET_NAME_LEN];
4041 	int ret;
4042 	char errbuf[ERRBUFLEN];
4043 	libzfs_handle_t *hdl = zhp->zfs_hdl;
4044 	uint64_t zoned;
4045 
4046 	assert(zhp->zfs_type == ZFS_TYPE_SNAPSHOT);
4047 
4048 	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
4049 	    "cannot create '%s'"), target);
4050 
4051 	/* validate the target/clone name */
4052 	if (!zfs_validate_name(hdl, target, ZFS_TYPE_FILESYSTEM, B_TRUE))
4053 		return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
4054 
4055 	/* validate parents exist */
4056 	if (check_parents(hdl, target, &zoned, B_FALSE, NULL) != 0)
4057 		return (-1);
4058 
4059 	(void) parent_name(target, parent, sizeof (parent));
4060 
4061 	/* do the clone */
4062 
4063 	if (props) {
4064 		zfs_type_t type = ZFS_TYPE_FILESYSTEM;
4065 
4066 		if (ZFS_IS_VOLUME(zhp))
4067 			type = ZFS_TYPE_VOLUME;
4068 		if ((props = zfs_valid_proplist(hdl, type, props, zoned,
4069 		    zhp, zhp->zpool_hdl, B_TRUE, errbuf)) == NULL)
4070 			return (-1);
4071 		if (zfs_fix_auto_resv(zhp, props) == -1) {
4072 			nvlist_free(props);
4073 			return (-1);
4074 		}
4075 	}
4076 
4077 	if (zfs_crypto_clone_check(hdl, zhp, parent, props) != 0) {
4078 		nvlist_free(props);
4079 		return (zfs_error(hdl, EZFS_CRYPTOFAILED, errbuf));
4080 	}
4081 
4082 	ret = lzc_clone(target, zhp->zfs_name, props);
4083 	nvlist_free(props);
4084 
4085 	if (ret != 0) {
4086 		switch (errno) {
4087 
4088 		case ENOENT:
4089 			/*
4090 			 * The parent doesn't exist.  We should have caught this
4091 			 * above, but there may a race condition that has since
4092 			 * destroyed the parent.
4093 			 *
4094 			 * At this point, we don't know whether it's the source
4095 			 * that doesn't exist anymore, or whether the target
4096 			 * dataset doesn't exist.
4097 			 */
4098 			zfs_error_aux(zhp->zfs_hdl, dgettext(TEXT_DOMAIN,
4099 			    "no such parent '%s'"), parent);
4100 			return (zfs_error(zhp->zfs_hdl, EZFS_NOENT, errbuf));
4101 
4102 		case EXDEV:
4103 			zfs_error_aux(zhp->zfs_hdl, dgettext(TEXT_DOMAIN,
4104 			    "source and target pools differ"));
4105 			return (zfs_error(zhp->zfs_hdl, EZFS_CROSSTARGET,
4106 			    errbuf));
4107 
4108 		default:
4109 			return (zfs_standard_error(zhp->zfs_hdl, errno,
4110 			    errbuf));
4111 		}
4112 	}
4113 
4114 	return (ret);
4115 }
4116 
4117 /*
4118  * Promotes the given clone fs to be the clone parent.
4119  */
4120 int
4121 zfs_promote(zfs_handle_t *zhp)
4122 {
4123 	libzfs_handle_t *hdl = zhp->zfs_hdl;
4124 	char snapname[ZFS_MAX_DATASET_NAME_LEN];
4125 	int ret;
4126 	char errbuf[ERRBUFLEN];
4127 
4128 	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
4129 	    "cannot promote '%s'"), zhp->zfs_name);
4130 
4131 	if (zhp->zfs_type == ZFS_TYPE_SNAPSHOT) {
4132 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4133 		    "snapshots can not be promoted"));
4134 		return (zfs_error(hdl, EZFS_BADTYPE, errbuf));
4135 	}
4136 
4137 	if (zhp->zfs_dmustats.dds_origin[0] == '\0') {
4138 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4139 		    "not a cloned filesystem"));
4140 		return (zfs_error(hdl, EZFS_BADTYPE, errbuf));
4141 	}
4142 
4143 	if (!zfs_validate_name(hdl, zhp->zfs_name, zhp->zfs_type, B_TRUE))
4144 		return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
4145 
4146 	ret = lzc_promote(zhp->zfs_name, snapname, sizeof (snapname));
4147 
4148 	if (ret != 0) {
4149 		switch (ret) {
4150 		case EACCES:
4151 			/*
4152 			 * Promoting encrypted dataset outside its
4153 			 * encryption root.
4154 			 */
4155 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4156 			    "cannot promote dataset outside its "
4157 			    "encryption root"));
4158 			return (zfs_error(hdl, EZFS_EXISTS, errbuf));
4159 
4160 		case EEXIST:
4161 			/* There is a conflicting snapshot name. */
4162 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4163 			    "conflicting snapshot '%s' from parent '%s'"),
4164 			    snapname, zhp->zfs_dmustats.dds_origin);
4165 			return (zfs_error(hdl, EZFS_EXISTS, errbuf));
4166 
4167 		default:
4168 			return (zfs_standard_error(hdl, ret, errbuf));
4169 		}
4170 	}
4171 	return (ret);
4172 }
4173 
4174 typedef struct snapdata {
4175 	nvlist_t *sd_nvl;
4176 	const char *sd_snapname;
4177 } snapdata_t;
4178 
4179 static int
4180 zfs_snapshot_cb(zfs_handle_t *zhp, void *arg)
4181 {
4182 	snapdata_t *sd = arg;
4183 	char name[ZFS_MAX_DATASET_NAME_LEN];
4184 	int rv = 0;
4185 
4186 	if (zfs_prop_get_int(zhp, ZFS_PROP_INCONSISTENT) == 0) {
4187 		if (snprintf(name, sizeof (name), "%s@%s", zfs_get_name(zhp),
4188 		    sd->sd_snapname) >= sizeof (name))
4189 			return (EINVAL);
4190 
4191 		fnvlist_add_boolean(sd->sd_nvl, name);
4192 
4193 		rv = zfs_iter_filesystems_v2(zhp, 0, zfs_snapshot_cb, sd);
4194 	}
4195 	zfs_close(zhp);
4196 
4197 	return (rv);
4198 }
4199 
4200 /*
4201  * Creates snapshots.  The keys in the snaps nvlist are the snapshots to be
4202  * created.
4203  */
4204 int
4205 zfs_snapshot_nvl(libzfs_handle_t *hdl, nvlist_t *snaps, nvlist_t *props)
4206 {
4207 	int ret;
4208 	char errbuf[ERRBUFLEN];
4209 	nvpair_t *elem;
4210 	nvlist_t *errors;
4211 	zpool_handle_t *zpool_hdl;
4212 	char pool[ZFS_MAX_DATASET_NAME_LEN];
4213 
4214 	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
4215 	    "cannot create snapshots "));
4216 
4217 	elem = NULL;
4218 	while ((elem = nvlist_next_nvpair(snaps, elem)) != NULL) {
4219 		const char *snapname = nvpair_name(elem);
4220 
4221 		/* validate the target name */
4222 		if (!zfs_validate_name(hdl, snapname, ZFS_TYPE_SNAPSHOT,
4223 		    B_TRUE)) {
4224 			(void) snprintf(errbuf, sizeof (errbuf),
4225 			    dgettext(TEXT_DOMAIN,
4226 			    "cannot create snapshot '%s'"), snapname);
4227 			return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
4228 		}
4229 	}
4230 
4231 	/*
4232 	 * get pool handle for prop validation. assumes all snaps are in the
4233 	 * same pool, as does lzc_snapshot (below).
4234 	 */
4235 	elem = nvlist_next_nvpair(snaps, NULL);
4236 	if (elem == NULL)
4237 		return (-1);
4238 	(void) strlcpy(pool, nvpair_name(elem), sizeof (pool));
4239 	pool[strcspn(pool, "/@")] = '\0';
4240 	zpool_hdl = zpool_open(hdl, pool);
4241 	if (zpool_hdl == NULL)
4242 		return (-1);
4243 
4244 	if (props != NULL &&
4245 	    (props = zfs_valid_proplist(hdl, ZFS_TYPE_SNAPSHOT,
4246 	    props, B_FALSE, NULL, zpool_hdl, B_FALSE, errbuf)) == NULL) {
4247 		zpool_close(zpool_hdl);
4248 		return (-1);
4249 	}
4250 	zpool_close(zpool_hdl);
4251 
4252 	ret = lzc_snapshot(snaps, props, &errors);
4253 
4254 	if (ret != 0) {
4255 		boolean_t printed = B_FALSE;
4256 		for (elem = nvlist_next_nvpair(errors, NULL);
4257 		    elem != NULL;
4258 		    elem = nvlist_next_nvpair(errors, elem)) {
4259 			(void) snprintf(errbuf, sizeof (errbuf),
4260 			    dgettext(TEXT_DOMAIN,
4261 			    "cannot create snapshot '%s'"), nvpair_name(elem));
4262 			(void) zfs_standard_error(hdl,
4263 			    fnvpair_value_int32(elem), errbuf);
4264 			printed = B_TRUE;
4265 		}
4266 		if (!printed) {
4267 			switch (ret) {
4268 			case EXDEV:
4269 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4270 				    "multiple snapshots of same "
4271 				    "fs not allowed"));
4272 				(void) zfs_error(hdl, EZFS_EXISTS, errbuf);
4273 
4274 				break;
4275 			default:
4276 				(void) zfs_standard_error(hdl, ret, errbuf);
4277 			}
4278 		}
4279 	}
4280 
4281 	nvlist_free(props);
4282 	nvlist_free(errors);
4283 	return (ret);
4284 }
4285 
4286 int
4287 zfs_snapshot(libzfs_handle_t *hdl, const char *path, boolean_t recursive,
4288     nvlist_t *props)
4289 {
4290 	int ret;
4291 	snapdata_t sd = { 0 };
4292 	char fsname[ZFS_MAX_DATASET_NAME_LEN];
4293 	char *cp;
4294 	zfs_handle_t *zhp;
4295 	char errbuf[ERRBUFLEN];
4296 
4297 	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
4298 	    "cannot snapshot %s"), path);
4299 
4300 	if (!zfs_validate_name(hdl, path, ZFS_TYPE_SNAPSHOT, B_TRUE))
4301 		return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
4302 
4303 	(void) strlcpy(fsname, path, sizeof (fsname));
4304 	cp = strchr(fsname, '@');
4305 	*cp = '\0';
4306 	sd.sd_snapname = cp + 1;
4307 
4308 	if ((zhp = zfs_open(hdl, fsname, ZFS_TYPE_FILESYSTEM |
4309 	    ZFS_TYPE_VOLUME)) == NULL) {
4310 		return (-1);
4311 	}
4312 
4313 	sd.sd_nvl = fnvlist_alloc();
4314 	if (recursive) {
4315 		(void) zfs_snapshot_cb(zfs_handle_dup(zhp), &sd);
4316 	} else {
4317 		fnvlist_add_boolean(sd.sd_nvl, path);
4318 	}
4319 
4320 	ret = zfs_snapshot_nvl(hdl, sd.sd_nvl, props);
4321 	fnvlist_free(sd.sd_nvl);
4322 	zfs_close(zhp);
4323 	return (ret);
4324 }
4325 
4326 /*
4327  * Destroy any more recent snapshots.  We invoke this callback on any dependents
4328  * of the snapshot first.  If the 'cb_dependent' member is non-zero, then this
4329  * is a dependent and we should just destroy it without checking the transaction
4330  * group.
4331  */
4332 typedef struct rollback_data {
4333 	const char	*cb_target;		/* the snapshot */
4334 	uint64_t	cb_create;		/* creation time reference */
4335 	boolean_t	cb_error;
4336 	boolean_t	cb_force;
4337 } rollback_data_t;
4338 
4339 static int
4340 rollback_destroy_dependent(zfs_handle_t *zhp, void *data)
4341 {
4342 	rollback_data_t *cbp = data;
4343 	prop_changelist_t *clp;
4344 
4345 	/* We must destroy this clone; first unmount it */
4346 	clp = changelist_gather(zhp, ZFS_PROP_NAME, 0,
4347 	    cbp->cb_force ? MS_FORCE: 0);
4348 	if (clp == NULL || changelist_prefix(clp) != 0) {
4349 		cbp->cb_error = B_TRUE;
4350 		zfs_close(zhp);
4351 		return (0);
4352 	}
4353 	if (zfs_destroy(zhp, B_FALSE) != 0)
4354 		cbp->cb_error = B_TRUE;
4355 	else
4356 		changelist_remove(clp, zhp->zfs_name);
4357 	(void) changelist_postfix(clp);
4358 	changelist_free(clp);
4359 
4360 	zfs_close(zhp);
4361 	return (0);
4362 }
4363 
4364 static int
4365 rollback_destroy(zfs_handle_t *zhp, void *data)
4366 {
4367 	rollback_data_t *cbp = data;
4368 
4369 	if (zfs_prop_get_int(zhp, ZFS_PROP_CREATETXG) > cbp->cb_create) {
4370 		cbp->cb_error |= zfs_iter_dependents_v2(zhp, 0, B_FALSE,
4371 		    rollback_destroy_dependent, cbp);
4372 
4373 		cbp->cb_error |= zfs_destroy(zhp, B_FALSE);
4374 	}
4375 
4376 	zfs_close(zhp);
4377 	return (0);
4378 }
4379 
4380 /*
4381  * Given a dataset, rollback to a specific snapshot, discarding any
4382  * data changes since then and making it the active dataset.
4383  *
4384  * Any snapshots and bookmarks more recent than the target are
4385  * destroyed, along with their dependents (i.e. clones).
4386  */
4387 int
4388 zfs_rollback(zfs_handle_t *zhp, zfs_handle_t *snap, boolean_t force)
4389 {
4390 	rollback_data_t cb = { 0 };
4391 	int err;
4392 	boolean_t restore_resv = 0;
4393 	uint64_t old_volsize = 0, new_volsize;
4394 	zfs_prop_t resv_prop = { 0 };
4395 	uint64_t min_txg = 0;
4396 
4397 	assert(zhp->zfs_type == ZFS_TYPE_FILESYSTEM ||
4398 	    zhp->zfs_type == ZFS_TYPE_VOLUME);
4399 
4400 	/*
4401 	 * Destroy all recent snapshots and their dependents.
4402 	 */
4403 	cb.cb_force = force;
4404 	cb.cb_target = snap->zfs_name;
4405 	cb.cb_create = zfs_prop_get_int(snap, ZFS_PROP_CREATETXG);
4406 
4407 	if (cb.cb_create > 0)
4408 		min_txg = cb.cb_create;
4409 
4410 	(void) zfs_iter_snapshots_v2(zhp, 0, rollback_destroy, &cb,
4411 	    min_txg, 0);
4412 
4413 	(void) zfs_iter_bookmarks_v2(zhp, 0, rollback_destroy, &cb);
4414 
4415 	if (cb.cb_error)
4416 		return (-1);
4417 
4418 	/*
4419 	 * Now that we have verified that the snapshot is the latest,
4420 	 * rollback to the given snapshot.
4421 	 */
4422 
4423 	if (zhp->zfs_type == ZFS_TYPE_VOLUME) {
4424 		if (zfs_which_resv_prop(zhp, &resv_prop) < 0)
4425 			return (-1);
4426 		old_volsize = zfs_prop_get_int(zhp, ZFS_PROP_VOLSIZE);
4427 		restore_resv =
4428 		    (old_volsize == zfs_prop_get_int(zhp, resv_prop));
4429 	}
4430 
4431 	/*
4432 	 * Pass both the filesystem and the wanted snapshot names,
4433 	 * we would get an error back if the snapshot is destroyed or
4434 	 * a new snapshot is created before this request is processed.
4435 	 */
4436 	err = lzc_rollback_to(zhp->zfs_name, snap->zfs_name);
4437 	if (err != 0) {
4438 		char errbuf[ERRBUFLEN];
4439 
4440 		(void) snprintf(errbuf, sizeof (errbuf),
4441 		    dgettext(TEXT_DOMAIN, "cannot rollback '%s'"),
4442 		    zhp->zfs_name);
4443 		switch (err) {
4444 		case EEXIST:
4445 			zfs_error_aux(zhp->zfs_hdl, dgettext(TEXT_DOMAIN,
4446 			    "there is a snapshot or bookmark more recent "
4447 			    "than '%s'"), snap->zfs_name);
4448 			(void) zfs_error(zhp->zfs_hdl, EZFS_EXISTS, errbuf);
4449 			break;
4450 		case ESRCH:
4451 			zfs_error_aux(zhp->zfs_hdl, dgettext(TEXT_DOMAIN,
4452 			    "'%s' is not found among snapshots of '%s'"),
4453 			    snap->zfs_name, zhp->zfs_name);
4454 			(void) zfs_error(zhp->zfs_hdl, EZFS_NOENT, errbuf);
4455 			break;
4456 		case EINVAL:
4457 			(void) zfs_error(zhp->zfs_hdl, EZFS_BADTYPE, errbuf);
4458 			break;
4459 		default:
4460 			(void) zfs_standard_error(zhp->zfs_hdl, err, errbuf);
4461 		}
4462 		return (err);
4463 	}
4464 
4465 	/*
4466 	 * For volumes, if the pre-rollback volsize matched the pre-
4467 	 * rollback reservation and the volsize has changed then set
4468 	 * the reservation property to the post-rollback volsize.
4469 	 * Make a new handle since the rollback closed the dataset.
4470 	 */
4471 	if ((zhp->zfs_type == ZFS_TYPE_VOLUME) &&
4472 	    (zhp = make_dataset_handle(zhp->zfs_hdl, zhp->zfs_name))) {
4473 		if (restore_resv) {
4474 			new_volsize = zfs_prop_get_int(zhp, ZFS_PROP_VOLSIZE);
4475 			if (old_volsize != new_volsize)
4476 				err = zfs_prop_set_int(zhp, resv_prop,
4477 				    new_volsize);
4478 		}
4479 		zfs_close(zhp);
4480 	}
4481 	return (err);
4482 }
4483 
4484 /*
4485  * Renames the given dataset.
4486  */
4487 int
4488 zfs_rename(zfs_handle_t *zhp, const char *target, renameflags_t flags)
4489 {
4490 	int ret = 0;
4491 	zfs_cmd_t zc = {"\0"};
4492 	char *delim;
4493 	prop_changelist_t *cl = NULL;
4494 	char parent[ZFS_MAX_DATASET_NAME_LEN];
4495 	char property[ZFS_MAXPROPLEN];
4496 	libzfs_handle_t *hdl = zhp->zfs_hdl;
4497 	char errbuf[ERRBUFLEN];
4498 
4499 	/* if we have the same exact name, just return success */
4500 	if (strcmp(zhp->zfs_name, target) == 0)
4501 		return (0);
4502 
4503 	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
4504 	    "cannot rename to '%s'"), target);
4505 
4506 	/* make sure source name is valid */
4507 	if (!zfs_validate_name(hdl, zhp->zfs_name, zhp->zfs_type, B_TRUE))
4508 		return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
4509 
4510 	/*
4511 	 * Make sure the target name is valid
4512 	 */
4513 	if (zhp->zfs_type == ZFS_TYPE_SNAPSHOT) {
4514 		if ((strchr(target, '@') == NULL) ||
4515 		    *target == '@') {
4516 			/*
4517 			 * Snapshot target name is abbreviated,
4518 			 * reconstruct full dataset name
4519 			 */
4520 			(void) strlcpy(parent, zhp->zfs_name,
4521 			    sizeof (parent));
4522 			delim = strchr(parent, '@');
4523 			if (strchr(target, '@') == NULL)
4524 				*(++delim) = '\0';
4525 			else
4526 				*delim = '\0';
4527 			(void) strlcat(parent, target, sizeof (parent));
4528 			target = parent;
4529 		} else {
4530 			/*
4531 			 * Make sure we're renaming within the same dataset.
4532 			 */
4533 			delim = strchr(target, '@');
4534 			if (strncmp(zhp->zfs_name, target, delim - target)
4535 			    != 0 || zhp->zfs_name[delim - target] != '@') {
4536 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4537 				    "snapshots must be part of same "
4538 				    "dataset"));
4539 				return (zfs_error(hdl, EZFS_CROSSTARGET,
4540 				    errbuf));
4541 			}
4542 		}
4543 
4544 		if (!zfs_validate_name(hdl, target, zhp->zfs_type, B_TRUE))
4545 			return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
4546 	} else {
4547 		if (flags.recursive) {
4548 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4549 			    "recursive rename must be a snapshot"));
4550 			return (zfs_error(hdl, EZFS_BADTYPE, errbuf));
4551 		}
4552 
4553 		if (!zfs_validate_name(hdl, target, zhp->zfs_type, B_TRUE))
4554 			return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
4555 
4556 		/* validate parents */
4557 		if (check_parents(hdl, target, NULL, B_FALSE, NULL) != 0)
4558 			return (-1);
4559 
4560 		/* make sure we're in the same pool */
4561 		verify((delim = strchr(target, '/')) != NULL);
4562 		if (strncmp(zhp->zfs_name, target, delim - target) != 0 ||
4563 		    zhp->zfs_name[delim - target] != '/') {
4564 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4565 			    "datasets must be within same pool"));
4566 			return (zfs_error(hdl, EZFS_CROSSTARGET, errbuf));
4567 		}
4568 
4569 		/* new name cannot be a child of the current dataset name */
4570 		if (is_descendant(zhp->zfs_name, target)) {
4571 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4572 			    "New dataset name cannot be a descendant of "
4573 			    "current dataset name"));
4574 			return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
4575 		}
4576 	}
4577 
4578 	(void) snprintf(errbuf, sizeof (errbuf),
4579 	    dgettext(TEXT_DOMAIN, "cannot rename '%s'"), zhp->zfs_name);
4580 
4581 	if (getzoneid() == GLOBAL_ZONEID &&
4582 	    zfs_prop_get_int(zhp, ZFS_PROP_ZONED)) {
4583 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4584 		    "dataset is used in a non-global zone"));
4585 		return (zfs_error(hdl, EZFS_ZONED, errbuf));
4586 	}
4587 
4588 	/*
4589 	 * Avoid unmounting file systems with mountpoint property set to
4590 	 * 'legacy' or 'none' even if -u option is not given.
4591 	 */
4592 	if (zhp->zfs_type == ZFS_TYPE_FILESYSTEM &&
4593 	    !flags.recursive && !flags.nounmount &&
4594 	    zfs_prop_get(zhp, ZFS_PROP_MOUNTPOINT, property,
4595 	    sizeof (property), NULL, NULL, 0, B_FALSE) == 0 &&
4596 	    (strcmp(property, "legacy") == 0 ||
4597 	    strcmp(property, "none") == 0)) {
4598 		flags.nounmount = B_TRUE;
4599 	}
4600 	if (flags.recursive) {
4601 		char *parentname = zfs_strdup(zhp->zfs_hdl, zhp->zfs_name);
4602 		delim = strchr(parentname, '@');
4603 		*delim = '\0';
4604 		zfs_handle_t *zhrp = zfs_open(zhp->zfs_hdl, parentname,
4605 		    ZFS_TYPE_DATASET);
4606 		free(parentname);
4607 		if (zhrp == NULL) {
4608 			ret = -1;
4609 			goto error;
4610 		}
4611 		zfs_close(zhrp);
4612 	} else if (zhp->zfs_type != ZFS_TYPE_SNAPSHOT) {
4613 		if ((cl = changelist_gather(zhp, ZFS_PROP_NAME,
4614 		    flags.nounmount ? CL_GATHER_DONT_UNMOUNT :
4615 		    CL_GATHER_ITER_MOUNTED,
4616 		    flags.forceunmount ? MS_FORCE : 0)) == NULL)
4617 			return (-1);
4618 
4619 		if (changelist_haszonedchild(cl)) {
4620 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4621 			    "child dataset with inherited mountpoint is used "
4622 			    "in a non-global zone"));
4623 			(void) zfs_error(hdl, EZFS_ZONED, errbuf);
4624 			ret = -1;
4625 			goto error;
4626 		}
4627 
4628 		if ((ret = changelist_prefix(cl)) != 0)
4629 			goto error;
4630 	}
4631 
4632 	if (ZFS_IS_VOLUME(zhp))
4633 		zc.zc_objset_type = DMU_OST_ZVOL;
4634 	else
4635 		zc.zc_objset_type = DMU_OST_ZFS;
4636 
4637 	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
4638 	(void) strlcpy(zc.zc_value, target, sizeof (zc.zc_value));
4639 
4640 	zc.zc_cookie = !!flags.recursive;
4641 	zc.zc_cookie |= (!!flags.nounmount) << 1;
4642 
4643 	if ((ret = zfs_ioctl(zhp->zfs_hdl, ZFS_IOC_RENAME, &zc)) != 0) {
4644 		/*
4645 		 * if it was recursive, the one that actually failed will
4646 		 * be in zc.zc_name
4647 		 */
4648 		(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
4649 		    "cannot rename '%s'"), zc.zc_name);
4650 
4651 		if (flags.recursive && errno == EEXIST) {
4652 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4653 			    "a child dataset already has a snapshot "
4654 			    "with the new name"));
4655 			(void) zfs_error(hdl, EZFS_EXISTS, errbuf);
4656 		} else if (errno == EACCES) {
4657 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4658 			    "cannot move encrypted child outside of "
4659 			    "its encryption root"));
4660 			(void) zfs_error(hdl, EZFS_CRYPTOFAILED, errbuf);
4661 		} else {
4662 			(void) zfs_standard_error(zhp->zfs_hdl, errno, errbuf);
4663 		}
4664 
4665 		/*
4666 		 * On failure, we still want to remount any filesystems that
4667 		 * were previously mounted, so we don't alter the system state.
4668 		 */
4669 		if (cl != NULL)
4670 			(void) changelist_postfix(cl);
4671 	} else {
4672 		if (cl != NULL) {
4673 			changelist_rename(cl, zfs_get_name(zhp), target);
4674 			ret = changelist_postfix(cl);
4675 		}
4676 	}
4677 
4678 error:
4679 	if (cl != NULL) {
4680 		changelist_free(cl);
4681 	}
4682 	return (ret);
4683 }
4684 
4685 nvlist_t *
4686 zfs_get_all_props(zfs_handle_t *zhp)
4687 {
4688 	return (zhp->zfs_props);
4689 }
4690 
4691 nvlist_t *
4692 zfs_get_recvd_props(zfs_handle_t *zhp)
4693 {
4694 	if (zhp->zfs_recvd_props == NULL)
4695 		if (get_recvd_props_ioctl(zhp) != 0)
4696 			return (NULL);
4697 	return (zhp->zfs_recvd_props);
4698 }
4699 
4700 nvlist_t *
4701 zfs_get_user_props(zfs_handle_t *zhp)
4702 {
4703 	return (zhp->zfs_user_props);
4704 }
4705 
4706 /*
4707  * This function is used by 'zfs list' to determine the exact set of columns to
4708  * display, and their maximum widths.  This does two main things:
4709  *
4710  *      - If this is a list of all properties, then expand the list to include
4711  *        all native properties, and set a flag so that for each dataset we look
4712  *        for new unique user properties and add them to the list.
4713  *
4714  *      - For non fixed-width properties, keep track of the maximum width seen
4715  *        so that we can size the column appropriately. If the user has
4716  *        requested received property values, we also need to compute the width
4717  *        of the RECEIVED column.
4718  */
4719 int
4720 zfs_expand_proplist(zfs_handle_t *zhp, zprop_list_t **plp, boolean_t received,
4721     boolean_t literal)
4722 {
4723 	libzfs_handle_t *hdl = zhp->zfs_hdl;
4724 	zprop_list_t *entry;
4725 	zprop_list_t **last, **start;
4726 	nvlist_t *userprops, *propval;
4727 	nvpair_t *elem;
4728 	const char *strval;
4729 	char buf[ZFS_MAXPROPLEN];
4730 
4731 	if (zprop_expand_list(hdl, plp, ZFS_TYPE_DATASET) != 0)
4732 		return (-1);
4733 
4734 	userprops = zfs_get_user_props(zhp);
4735 
4736 	entry = *plp;
4737 	if (entry->pl_all && nvlist_next_nvpair(userprops, NULL) != NULL) {
4738 		/*
4739 		 * Go through and add any user properties as necessary.  We
4740 		 * start by incrementing our list pointer to the first
4741 		 * non-native property.
4742 		 */
4743 		start = plp;
4744 		while (*start != NULL) {
4745 			if ((*start)->pl_prop == ZPROP_USERPROP)
4746 				break;
4747 			start = &(*start)->pl_next;
4748 		}
4749 
4750 		elem = NULL;
4751 		while ((elem = nvlist_next_nvpair(userprops, elem)) != NULL) {
4752 			/*
4753 			 * See if we've already found this property in our list.
4754 			 */
4755 			for (last = start; *last != NULL;
4756 			    last = &(*last)->pl_next) {
4757 				if (strcmp((*last)->pl_user_prop,
4758 				    nvpair_name(elem)) == 0)
4759 					break;
4760 			}
4761 
4762 			if (*last == NULL) {
4763 				entry = zfs_alloc(hdl, sizeof (zprop_list_t));
4764 				entry->pl_user_prop =
4765 				    zfs_strdup(hdl, nvpair_name(elem));
4766 				entry->pl_prop = ZPROP_USERPROP;
4767 				entry->pl_width = strlen(nvpair_name(elem));
4768 				entry->pl_all = B_TRUE;
4769 				*last = entry;
4770 			}
4771 		}
4772 	}
4773 
4774 	/*
4775 	 * Now go through and check the width of any non-fixed columns
4776 	 */
4777 	for (entry = *plp; entry != NULL; entry = entry->pl_next) {
4778 		if (entry->pl_fixed && !literal)
4779 			continue;
4780 
4781 		if (entry->pl_prop != ZPROP_USERPROP) {
4782 			if (zfs_prop_get(zhp, entry->pl_prop,
4783 			    buf, sizeof (buf), NULL, NULL, 0, literal) == 0) {
4784 				if (strlen(buf) > entry->pl_width)
4785 					entry->pl_width = strlen(buf);
4786 			}
4787 			if (received && zfs_prop_get_recvd(zhp,
4788 			    zfs_prop_to_name(entry->pl_prop),
4789 			    buf, sizeof (buf), literal) == 0)
4790 				if (strlen(buf) > entry->pl_recvd_width)
4791 					entry->pl_recvd_width = strlen(buf);
4792 		} else {
4793 			if (nvlist_lookup_nvlist(userprops, entry->pl_user_prop,
4794 			    &propval) == 0) {
4795 				strval = fnvlist_lookup_string(propval,
4796 				    ZPROP_VALUE);
4797 				if (strlen(strval) > entry->pl_width)
4798 					entry->pl_width = strlen(strval);
4799 			}
4800 			if (received && zfs_prop_get_recvd(zhp,
4801 			    entry->pl_user_prop,
4802 			    buf, sizeof (buf), literal) == 0)
4803 				if (strlen(buf) > entry->pl_recvd_width)
4804 					entry->pl_recvd_width = strlen(buf);
4805 		}
4806 	}
4807 
4808 	return (0);
4809 }
4810 
4811 void
4812 zfs_prune_proplist(zfs_handle_t *zhp, uint8_t *props)
4813 {
4814 	nvpair_t *curr;
4815 	nvpair_t *next;
4816 
4817 	/*
4818 	 * Keep a reference to the props-table against which we prune the
4819 	 * properties.
4820 	 */
4821 	zhp->zfs_props_table = props;
4822 
4823 	curr = nvlist_next_nvpair(zhp->zfs_props, NULL);
4824 
4825 	while (curr) {
4826 		zfs_prop_t zfs_prop = zfs_name_to_prop(nvpair_name(curr));
4827 		next = nvlist_next_nvpair(zhp->zfs_props, curr);
4828 
4829 		/*
4830 		 * User properties will result in ZPROP_USERPROP (an alias
4831 		 * for ZPROP_INVAL), and since we
4832 		 * only know how to prune standard ZFS properties, we always
4833 		 * leave these in the list.  This can also happen if we
4834 		 * encounter an unknown DSL property (when running older
4835 		 * software, for example).
4836 		 */
4837 		if (zfs_prop != ZPROP_USERPROP && props[zfs_prop] == B_FALSE)
4838 			(void) nvlist_remove(zhp->zfs_props,
4839 			    nvpair_name(curr), nvpair_type(curr));
4840 		curr = next;
4841 	}
4842 }
4843 
4844 static int
4845 zfs_smb_acl_mgmt(libzfs_handle_t *hdl, char *dataset, char *path,
4846     zfs_smb_acl_op_t cmd, char *resource1, char *resource2)
4847 {
4848 	zfs_cmd_t zc = {"\0"};
4849 	nvlist_t *nvlist = NULL;
4850 	int error;
4851 
4852 	(void) strlcpy(zc.zc_name, dataset, sizeof (zc.zc_name));
4853 	(void) strlcpy(zc.zc_value, path, sizeof (zc.zc_value));
4854 	zc.zc_cookie = (uint64_t)cmd;
4855 
4856 	if (cmd == ZFS_SMB_ACL_RENAME) {
4857 		if (nvlist_alloc(&nvlist, NV_UNIQUE_NAME, 0) != 0) {
4858 			(void) no_memory(hdl);
4859 			return (0);
4860 		}
4861 	}
4862 
4863 	switch (cmd) {
4864 	case ZFS_SMB_ACL_ADD:
4865 	case ZFS_SMB_ACL_REMOVE:
4866 		(void) strlcpy(zc.zc_string, resource1, sizeof (zc.zc_string));
4867 		break;
4868 	case ZFS_SMB_ACL_RENAME:
4869 		if (nvlist_add_string(nvlist, ZFS_SMB_ACL_SRC,
4870 		    resource1) != 0) {
4871 				(void) no_memory(hdl);
4872 				return (-1);
4873 		}
4874 		if (nvlist_add_string(nvlist, ZFS_SMB_ACL_TARGET,
4875 		    resource2) != 0) {
4876 				(void) no_memory(hdl);
4877 				return (-1);
4878 		}
4879 		zcmd_write_src_nvlist(hdl, &zc, nvlist);
4880 		break;
4881 	case ZFS_SMB_ACL_PURGE:
4882 		break;
4883 	default:
4884 		return (-1);
4885 	}
4886 	error = ioctl(hdl->libzfs_fd, ZFS_IOC_SMB_ACL, &zc);
4887 	nvlist_free(nvlist);
4888 	return (error);
4889 }
4890 
4891 int
4892 zfs_smb_acl_add(libzfs_handle_t *hdl, char *dataset,
4893     char *path, char *resource)
4894 {
4895 	return (zfs_smb_acl_mgmt(hdl, dataset, path, ZFS_SMB_ACL_ADD,
4896 	    resource, NULL));
4897 }
4898 
4899 int
4900 zfs_smb_acl_remove(libzfs_handle_t *hdl, char *dataset,
4901     char *path, char *resource)
4902 {
4903 	return (zfs_smb_acl_mgmt(hdl, dataset, path, ZFS_SMB_ACL_REMOVE,
4904 	    resource, NULL));
4905 }
4906 
4907 int
4908 zfs_smb_acl_purge(libzfs_handle_t *hdl, char *dataset, char *path)
4909 {
4910 	return (zfs_smb_acl_mgmt(hdl, dataset, path, ZFS_SMB_ACL_PURGE,
4911 	    NULL, NULL));
4912 }
4913 
4914 int
4915 zfs_smb_acl_rename(libzfs_handle_t *hdl, char *dataset, char *path,
4916     char *oldname, char *newname)
4917 {
4918 	return (zfs_smb_acl_mgmt(hdl, dataset, path, ZFS_SMB_ACL_RENAME,
4919 	    oldname, newname));
4920 }
4921 
4922 int
4923 zfs_userspace(zfs_handle_t *zhp, zfs_userquota_prop_t type,
4924     zfs_userspace_cb_t func, void *arg)
4925 {
4926 	zfs_cmd_t zc = {"\0"};
4927 	zfs_useracct_t buf[100];
4928 	libzfs_handle_t *hdl = zhp->zfs_hdl;
4929 	int ret;
4930 
4931 	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
4932 
4933 	zc.zc_objset_type = type;
4934 	zc.zc_nvlist_dst = (uintptr_t)buf;
4935 
4936 	for (;;) {
4937 		zfs_useracct_t *zua = buf;
4938 
4939 		zc.zc_nvlist_dst_size = sizeof (buf);
4940 		if (zfs_ioctl(hdl, ZFS_IOC_USERSPACE_MANY, &zc) != 0) {
4941 			if ((errno == ENOTSUP &&
4942 			    (type == ZFS_PROP_USEROBJUSED ||
4943 			    type == ZFS_PROP_GROUPOBJUSED ||
4944 			    type == ZFS_PROP_USEROBJQUOTA ||
4945 			    type == ZFS_PROP_GROUPOBJQUOTA ||
4946 			    type == ZFS_PROP_PROJECTOBJUSED ||
4947 			    type == ZFS_PROP_PROJECTOBJQUOTA ||
4948 			    type == ZFS_PROP_PROJECTUSED ||
4949 			    type == ZFS_PROP_PROJECTQUOTA)))
4950 				break;
4951 
4952 			return (zfs_standard_error_fmt(hdl, errno,
4953 			    dgettext(TEXT_DOMAIN,
4954 			    "cannot get used/quota for %s"), zc.zc_name));
4955 		}
4956 		if (zc.zc_nvlist_dst_size == 0)
4957 			break;
4958 
4959 		while (zc.zc_nvlist_dst_size > 0) {
4960 			if ((ret = func(arg, zua->zu_domain, zua->zu_rid,
4961 			    zua->zu_space)) != 0)
4962 				return (ret);
4963 			zua++;
4964 			zc.zc_nvlist_dst_size -= sizeof (zfs_useracct_t);
4965 		}
4966 	}
4967 
4968 	return (0);
4969 }
4970 
4971 struct holdarg {
4972 	nvlist_t *nvl;
4973 	const char *snapname;
4974 	const char *tag;
4975 	boolean_t recursive;
4976 	int error;
4977 };
4978 
4979 static int
4980 zfs_hold_one(zfs_handle_t *zhp, void *arg)
4981 {
4982 	struct holdarg *ha = arg;
4983 	char name[ZFS_MAX_DATASET_NAME_LEN];
4984 	int rv = 0;
4985 
4986 	if (snprintf(name, sizeof (name), "%s@%s", zhp->zfs_name,
4987 	    ha->snapname) >= sizeof (name))
4988 		return (EINVAL);
4989 
4990 	if (lzc_exists(name))
4991 		fnvlist_add_string(ha->nvl, name, ha->tag);
4992 
4993 	if (ha->recursive)
4994 		rv = zfs_iter_filesystems_v2(zhp, 0, zfs_hold_one, ha);
4995 	zfs_close(zhp);
4996 	return (rv);
4997 }
4998 
4999 int
5000 zfs_hold(zfs_handle_t *zhp, const char *snapname, const char *tag,
5001     boolean_t recursive, int cleanup_fd)
5002 {
5003 	int ret;
5004 	struct holdarg ha;
5005 
5006 	ha.nvl = fnvlist_alloc();
5007 	ha.snapname = snapname;
5008 	ha.tag = tag;
5009 	ha.recursive = recursive;
5010 	(void) zfs_hold_one(zfs_handle_dup(zhp), &ha);
5011 
5012 	if (nvlist_empty(ha.nvl)) {
5013 		char errbuf[ERRBUFLEN];
5014 
5015 		fnvlist_free(ha.nvl);
5016 		ret = ENOENT;
5017 		(void) snprintf(errbuf, sizeof (errbuf),
5018 		    dgettext(TEXT_DOMAIN,
5019 		    "cannot hold snapshot '%s@%s'"),
5020 		    zhp->zfs_name, snapname);
5021 		(void) zfs_standard_error(zhp->zfs_hdl, ret, errbuf);
5022 		return (ret);
5023 	}
5024 
5025 	ret = zfs_hold_nvl(zhp, cleanup_fd, ha.nvl);
5026 	fnvlist_free(ha.nvl);
5027 
5028 	return (ret);
5029 }
5030 
5031 int
5032 zfs_hold_nvl(zfs_handle_t *zhp, int cleanup_fd, nvlist_t *holds)
5033 {
5034 	int ret;
5035 	nvlist_t *errors;
5036 	libzfs_handle_t *hdl = zhp->zfs_hdl;
5037 	char errbuf[ERRBUFLEN];
5038 	nvpair_t *elem;
5039 
5040 	errors = NULL;
5041 	ret = lzc_hold(holds, cleanup_fd, &errors);
5042 
5043 	if (ret == 0) {
5044 		/* There may be errors even in the success case. */
5045 		fnvlist_free(errors);
5046 		return (0);
5047 	}
5048 
5049 	if (nvlist_empty(errors)) {
5050 		/* no hold-specific errors */
5051 		(void) snprintf(errbuf, sizeof (errbuf),
5052 		    dgettext(TEXT_DOMAIN, "cannot hold"));
5053 		switch (ret) {
5054 		case ENOTSUP:
5055 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
5056 			    "pool must be upgraded"));
5057 			(void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
5058 			break;
5059 		case EINVAL:
5060 			(void) zfs_error(hdl, EZFS_BADTYPE, errbuf);
5061 			break;
5062 		default:
5063 			(void) zfs_standard_error(hdl, ret, errbuf);
5064 		}
5065 	}
5066 
5067 	for (elem = nvlist_next_nvpair(errors, NULL);
5068 	    elem != NULL;
5069 	    elem = nvlist_next_nvpair(errors, elem)) {
5070 		(void) snprintf(errbuf, sizeof (errbuf),
5071 		    dgettext(TEXT_DOMAIN,
5072 		    "cannot hold snapshot '%s'"), nvpair_name(elem));
5073 		switch (fnvpair_value_int32(elem)) {
5074 		case E2BIG:
5075 			/*
5076 			 * Temporary tags wind up having the ds object id
5077 			 * prepended. So even if we passed the length check
5078 			 * above, it's still possible for the tag to wind
5079 			 * up being slightly too long.
5080 			 */
5081 			(void) zfs_error(hdl, EZFS_TAGTOOLONG, errbuf);
5082 			break;
5083 		case EINVAL:
5084 			(void) zfs_error(hdl, EZFS_BADTYPE, errbuf);
5085 			break;
5086 		case EEXIST:
5087 			(void) zfs_error(hdl, EZFS_REFTAG_HOLD, errbuf);
5088 			break;
5089 		default:
5090 			(void) zfs_standard_error(hdl,
5091 			    fnvpair_value_int32(elem), errbuf);
5092 		}
5093 	}
5094 
5095 	fnvlist_free(errors);
5096 	return (ret);
5097 }
5098 
5099 static int
5100 zfs_release_one(zfs_handle_t *zhp, void *arg)
5101 {
5102 	struct holdarg *ha = arg;
5103 	char name[ZFS_MAX_DATASET_NAME_LEN];
5104 	int rv = 0;
5105 	nvlist_t *existing_holds;
5106 
5107 	if (snprintf(name, sizeof (name), "%s@%s", zhp->zfs_name,
5108 	    ha->snapname) >= sizeof (name)) {
5109 		ha->error = EINVAL;
5110 		rv = EINVAL;
5111 	}
5112 
5113 	if (lzc_get_holds(name, &existing_holds) != 0) {
5114 		ha->error = ENOENT;
5115 	} else if (!nvlist_exists(existing_holds, ha->tag)) {
5116 		ha->error = ESRCH;
5117 	} else {
5118 		nvlist_t *torelease = fnvlist_alloc();
5119 		fnvlist_add_boolean(torelease, ha->tag);
5120 		fnvlist_add_nvlist(ha->nvl, name, torelease);
5121 		fnvlist_free(torelease);
5122 	}
5123 
5124 	if (ha->recursive)
5125 		rv = zfs_iter_filesystems_v2(zhp, 0, zfs_release_one, ha);
5126 	zfs_close(zhp);
5127 	return (rv);
5128 }
5129 
5130 int
5131 zfs_release(zfs_handle_t *zhp, const char *snapname, const char *tag,
5132     boolean_t recursive)
5133 {
5134 	int ret;
5135 	struct holdarg ha;
5136 	nvlist_t *errors = NULL;
5137 	nvpair_t *elem;
5138 	libzfs_handle_t *hdl = zhp->zfs_hdl;
5139 	char errbuf[ERRBUFLEN];
5140 
5141 	ha.nvl = fnvlist_alloc();
5142 	ha.snapname = snapname;
5143 	ha.tag = tag;
5144 	ha.recursive = recursive;
5145 	ha.error = 0;
5146 	(void) zfs_release_one(zfs_handle_dup(zhp), &ha);
5147 
5148 	if (nvlist_empty(ha.nvl)) {
5149 		fnvlist_free(ha.nvl);
5150 		ret = ha.error;
5151 		(void) snprintf(errbuf, sizeof (errbuf),
5152 		    dgettext(TEXT_DOMAIN,
5153 		    "cannot release hold from snapshot '%s@%s'"),
5154 		    zhp->zfs_name, snapname);
5155 		if (ret == ESRCH) {
5156 			(void) zfs_error(hdl, EZFS_REFTAG_RELE, errbuf);
5157 		} else {
5158 			(void) zfs_standard_error(hdl, ret, errbuf);
5159 		}
5160 		return (ret);
5161 	}
5162 
5163 	ret = lzc_release(ha.nvl, &errors);
5164 	fnvlist_free(ha.nvl);
5165 
5166 	if (ret == 0) {
5167 		/* There may be errors even in the success case. */
5168 		fnvlist_free(errors);
5169 		return (0);
5170 	}
5171 
5172 	if (nvlist_empty(errors)) {
5173 		/* no hold-specific errors */
5174 		(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
5175 		    "cannot release"));
5176 		switch (errno) {
5177 		case ENOTSUP:
5178 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
5179 			    "pool must be upgraded"));
5180 			(void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
5181 			break;
5182 		default:
5183 			(void) zfs_standard_error(hdl, errno, errbuf);
5184 		}
5185 	}
5186 
5187 	for (elem = nvlist_next_nvpair(errors, NULL);
5188 	    elem != NULL;
5189 	    elem = nvlist_next_nvpair(errors, elem)) {
5190 		(void) snprintf(errbuf, sizeof (errbuf),
5191 		    dgettext(TEXT_DOMAIN,
5192 		    "cannot release hold from snapshot '%s'"),
5193 		    nvpair_name(elem));
5194 		switch (fnvpair_value_int32(elem)) {
5195 		case ESRCH:
5196 			(void) zfs_error(hdl, EZFS_REFTAG_RELE, errbuf);
5197 			break;
5198 		case EINVAL:
5199 			(void) zfs_error(hdl, EZFS_BADTYPE, errbuf);
5200 			break;
5201 		default:
5202 			(void) zfs_standard_error(hdl,
5203 			    fnvpair_value_int32(elem), errbuf);
5204 		}
5205 	}
5206 
5207 	fnvlist_free(errors);
5208 	return (ret);
5209 }
5210 
5211 int
5212 zfs_get_fsacl(zfs_handle_t *zhp, nvlist_t **nvl)
5213 {
5214 	zfs_cmd_t zc = {"\0"};
5215 	libzfs_handle_t *hdl = zhp->zfs_hdl;
5216 	int nvsz = 2048;
5217 	void *nvbuf;
5218 	int err = 0;
5219 	char errbuf[ERRBUFLEN];
5220 
5221 	assert(zhp->zfs_type == ZFS_TYPE_VOLUME ||
5222 	    zhp->zfs_type == ZFS_TYPE_FILESYSTEM);
5223 
5224 tryagain:
5225 
5226 	nvbuf = malloc(nvsz);
5227 	if (nvbuf == NULL) {
5228 		err = (zfs_error(hdl, EZFS_NOMEM, zfs_strerror(errno)));
5229 		goto out;
5230 	}
5231 
5232 	zc.zc_nvlist_dst_size = nvsz;
5233 	zc.zc_nvlist_dst = (uintptr_t)nvbuf;
5234 
5235 	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
5236 
5237 	if (zfs_ioctl(hdl, ZFS_IOC_GET_FSACL, &zc) != 0) {
5238 		(void) snprintf(errbuf, sizeof (errbuf),
5239 		    dgettext(TEXT_DOMAIN, "cannot get permissions on '%s'"),
5240 		    zc.zc_name);
5241 		switch (errno) {
5242 		case ENOMEM:
5243 			free(nvbuf);
5244 			nvsz = zc.zc_nvlist_dst_size;
5245 			goto tryagain;
5246 
5247 		case ENOTSUP:
5248 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
5249 			    "pool must be upgraded"));
5250 			err = zfs_error(hdl, EZFS_BADVERSION, errbuf);
5251 			break;
5252 		case EINVAL:
5253 			err = zfs_error(hdl, EZFS_BADTYPE, errbuf);
5254 			break;
5255 		case ENOENT:
5256 			err = zfs_error(hdl, EZFS_NOENT, errbuf);
5257 			break;
5258 		default:
5259 			err = zfs_standard_error(hdl, errno, errbuf);
5260 			break;
5261 		}
5262 	} else {
5263 		/* success */
5264 		int rc = nvlist_unpack(nvbuf, zc.zc_nvlist_dst_size, nvl, 0);
5265 		if (rc) {
5266 			err = zfs_standard_error_fmt(hdl, rc, dgettext(
5267 			    TEXT_DOMAIN, "cannot get permissions on '%s'"),
5268 			    zc.zc_name);
5269 		}
5270 	}
5271 
5272 	free(nvbuf);
5273 out:
5274 	return (err);
5275 }
5276 
5277 int
5278 zfs_set_fsacl(zfs_handle_t *zhp, boolean_t un, nvlist_t *nvl)
5279 {
5280 	zfs_cmd_t zc = {"\0"};
5281 	libzfs_handle_t *hdl = zhp->zfs_hdl;
5282 	char *nvbuf;
5283 	char errbuf[ERRBUFLEN];
5284 	size_t nvsz;
5285 	int err;
5286 
5287 	assert(zhp->zfs_type == ZFS_TYPE_VOLUME ||
5288 	    zhp->zfs_type == ZFS_TYPE_FILESYSTEM);
5289 
5290 	err = nvlist_size(nvl, &nvsz, NV_ENCODE_NATIVE);
5291 	assert(err == 0);
5292 
5293 	nvbuf = malloc(nvsz);
5294 
5295 	err = nvlist_pack(nvl, &nvbuf, &nvsz, NV_ENCODE_NATIVE, 0);
5296 	assert(err == 0);
5297 
5298 	zc.zc_nvlist_src_size = nvsz;
5299 	zc.zc_nvlist_src = (uintptr_t)nvbuf;
5300 	zc.zc_perm_action = un;
5301 
5302 	(void) strlcpy(zc.zc_name, zhp->zfs_name, sizeof (zc.zc_name));
5303 
5304 	if (zfs_ioctl(hdl, ZFS_IOC_SET_FSACL, &zc) != 0) {
5305 		(void) snprintf(errbuf, sizeof (errbuf),
5306 		    dgettext(TEXT_DOMAIN, "cannot set permissions on '%s'"),
5307 		    zc.zc_name);
5308 		switch (errno) {
5309 		case ENOTSUP:
5310 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
5311 			    "pool must be upgraded"));
5312 			err = zfs_error(hdl, EZFS_BADVERSION, errbuf);
5313 			break;
5314 		case EINVAL:
5315 			err = zfs_error(hdl, EZFS_BADTYPE, errbuf);
5316 			break;
5317 		case ENOENT:
5318 			err = zfs_error(hdl, EZFS_NOENT, errbuf);
5319 			break;
5320 		default:
5321 			err = zfs_standard_error(hdl, errno, errbuf);
5322 			break;
5323 		}
5324 	}
5325 
5326 	free(nvbuf);
5327 
5328 	return (err);
5329 }
5330 
5331 int
5332 zfs_get_holds(zfs_handle_t *zhp, nvlist_t **nvl)
5333 {
5334 	int err;
5335 	char errbuf[ERRBUFLEN];
5336 
5337 	err = lzc_get_holds(zhp->zfs_name, nvl);
5338 
5339 	if (err != 0) {
5340 		libzfs_handle_t *hdl = zhp->zfs_hdl;
5341 
5342 		(void) snprintf(errbuf, sizeof (errbuf),
5343 		    dgettext(TEXT_DOMAIN, "cannot get holds for '%s'"),
5344 		    zhp->zfs_name);
5345 		switch (err) {
5346 		case ENOTSUP:
5347 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
5348 			    "pool must be upgraded"));
5349 			err = zfs_error(hdl, EZFS_BADVERSION, errbuf);
5350 			break;
5351 		case EINVAL:
5352 			err = zfs_error(hdl, EZFS_BADTYPE, errbuf);
5353 			break;
5354 		case ENOENT:
5355 			err = zfs_error(hdl, EZFS_NOENT, errbuf);
5356 			break;
5357 		default:
5358 			err = zfs_standard_error(hdl, errno, errbuf);
5359 			break;
5360 		}
5361 	}
5362 
5363 	return (err);
5364 }
5365 
5366 /*
5367  * The theory of raidz space accounting
5368  *
5369  * The "referenced" property of RAIDZ vdevs is scaled such that a 128KB block
5370  * will "reference" 128KB, even though it allocates more than that, to store the
5371  * parity information (and perhaps skip sectors). This concept of the
5372  * "referenced" (and other DMU space accounting) being lower than the allocated
5373  * space by a constant factor is called "raidz deflation."
5374  *
5375  * As mentioned above, the constant factor for raidz deflation assumes a 128KB
5376  * block size. However, zvols typically have a much smaller block size (default
5377  * 8KB). These smaller blocks may require proportionally much more parity
5378  * information (and perhaps skip sectors). In this case, the change to the
5379  * "referenced" property may be much more than the logical block size.
5380  *
5381  * Suppose a raidz vdev has 5 disks with ashift=12.  A 128k block may be written
5382  * as follows.
5383  *
5384  * +-------+-------+-------+-------+-------+
5385  * | disk1 | disk2 | disk3 | disk4 | disk5 |
5386  * +-------+-------+-------+-------+-------+
5387  * |  P0   |  D0   |  D8   |  D16  |  D24  |
5388  * |  P1   |  D1   |  D9   |  D17  |  D25  |
5389  * |  P2   |  D2   |  D10  |  D18  |  D26  |
5390  * |  P3   |  D3   |  D11  |  D19  |  D27  |
5391  * |  P4   |  D4   |  D12  |  D20  |  D28  |
5392  * |  P5   |  D5   |  D13  |  D21  |  D29  |
5393  * |  P6   |  D6   |  D14  |  D22  |  D30  |
5394  * |  P7   |  D7   |  D15  |  D23  |  D31  |
5395  * +-------+-------+-------+-------+-------+
5396  *
5397  * Above, notice that 160k was allocated: 8 x 4k parity sectors + 32 x 4k data
5398  * sectors.  The dataset's referenced will increase by 128k and the pool's
5399  * allocated and free properties will be adjusted by 160k.
5400  *
5401  * A 4k block written to the same raidz vdev will require two 4k sectors.  The
5402  * blank cells represent unallocated space.
5403  *
5404  * +-------+-------+-------+-------+-------+
5405  * | disk1 | disk2 | disk3 | disk4 | disk5 |
5406  * +-------+-------+-------+-------+-------+
5407  * |  P0   |  D0   |       |       |       |
5408  * +-------+-------+-------+-------+-------+
5409  *
5410  * Above, notice that the 4k block required one sector for parity and another
5411  * for data.  vdev_raidz_asize() will return 8k and as such the pool's allocated
5412  * and free properties will be adjusted by 8k.  The dataset will not be charged
5413  * 8k.  Rather, it will be charged a value that is scaled according to the
5414  * overhead of the 128k block on the same vdev.  This 8k allocation will be
5415  * charged 8k * 128k / 160k.  128k is from SPA_OLD_MAXBLOCKSIZE and 160k is as
5416  * calculated in the 128k block example above.
5417  *
5418  * Every raidz allocation is sized to be a multiple of nparity+1 sectors.  That
5419  * is, every raidz1 allocation will be a multiple of 2 sectors, raidz2
5420  * allocations are a multiple of 3 sectors, and raidz3 allocations are a
5421  * multiple of of 4 sectors.  When a block does not fill the required number of
5422  * sectors, skip blocks (sectors) are used.
5423  *
5424  * An 8k block being written to a raidz vdev may be written as follows:
5425  *
5426  * +-------+-------+-------+-------+-------+
5427  * | disk1 | disk2 | disk3 | disk4 | disk5 |
5428  * +-------+-------+-------+-------+-------+
5429  * |  P0   |  D0   |  D1   |  S0   |       |
5430  * +-------+-------+-------+-------+-------+
5431  *
5432  * In order to maintain the nparity+1 allocation size, a skip block (S0) was
5433  * added.  For this 8k block, the pool's allocated and free properties are
5434  * adjusted by 16k and the dataset's referenced is increased by 16k * 128k /
5435  * 160k.  Again, 128k is from SPA_OLD_MAXBLOCKSIZE and 160k is as calculated in
5436  * the 128k block example above.
5437  *
5438  * The situation is slightly different for dRAID since the minimum allocation
5439  * size is the full group width.  The same 8K block above would be written as
5440  * follows in a dRAID group:
5441  *
5442  * +-------+-------+-------+-------+-------+
5443  * | disk1 | disk2 | disk3 | disk4 | disk5 |
5444  * +-------+-------+-------+-------+-------+
5445  * |  P0   |  D0   |  D1   |  S0   |  S1   |
5446  * +-------+-------+-------+-------+-------+
5447  *
5448  * Compression may lead to a variety of block sizes being written for the same
5449  * volume or file.  There is no clear way to reserve just the amount of space
5450  * that will be required, so the worst case (no compression) is assumed.
5451  * Note that metadata blocks will typically be compressed, so the reservation
5452  * size returned by zvol_volsize_to_reservation() will generally be slightly
5453  * larger than the maximum that the volume can reference.
5454  */
5455 
5456 /*
5457  * Derived from function of same name in module/zfs/vdev_raidz.c.  Returns the
5458  * amount of space (in bytes) that will be allocated for the specified block
5459  * size. Note that the "referenced" space accounted will be less than this, but
5460  * not necessarily equal to "blksize", due to RAIDZ deflation.
5461  */
5462 static uint64_t
5463 vdev_raidz_asize(uint64_t ndisks, uint64_t nparity, uint64_t ashift,
5464     uint64_t blksize)
5465 {
5466 	uint64_t asize, ndata;
5467 
5468 	ASSERT3U(ndisks, >, nparity);
5469 	ndata = ndisks - nparity;
5470 	asize = ((blksize - 1) >> ashift) + 1;
5471 	asize += nparity * ((asize + ndata - 1) / ndata);
5472 	asize = roundup(asize, nparity + 1) << ashift;
5473 
5474 	return (asize);
5475 }
5476 
5477 /*
5478  * Derived from function of same name in module/zfs/vdev_draid.c.  Returns the
5479  * amount of space (in bytes) that will be allocated for the specified block
5480  * size.
5481  */
5482 static uint64_t
5483 vdev_draid_asize(uint64_t ndisks, uint64_t nparity, uint64_t ashift,
5484     uint64_t blksize)
5485 {
5486 	ASSERT3U(ndisks, >, nparity);
5487 	uint64_t ndata = ndisks - nparity;
5488 	uint64_t rows = ((blksize - 1) / (ndata << ashift)) + 1;
5489 	uint64_t asize = (rows * ndisks) << ashift;
5490 
5491 	return (asize);
5492 }
5493 
5494 /*
5495  * Determine how much space will be allocated if it lands on the most space-
5496  * inefficient top-level vdev.  Returns the size in bytes required to store one
5497  * copy of the volume data.  See theory comment above.
5498  */
5499 static uint64_t
5500 volsize_from_vdevs(zpool_handle_t *zhp, uint64_t nblocks, uint64_t blksize)
5501 {
5502 	nvlist_t *config, *tree, **vdevs;
5503 	uint_t nvdevs;
5504 	uint64_t ret = 0;
5505 
5506 	config = zpool_get_config(zhp, NULL);
5507 	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree) != 0 ||
5508 	    nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN,
5509 	    &vdevs, &nvdevs) != 0) {
5510 		return (nblocks * blksize);
5511 	}
5512 
5513 	for (int v = 0; v < nvdevs; v++) {
5514 		const char *type;
5515 		uint64_t nparity, ashift, asize, tsize;
5516 		uint64_t volsize;
5517 
5518 		if (nvlist_lookup_string(vdevs[v], ZPOOL_CONFIG_TYPE,
5519 		    &type) != 0)
5520 			continue;
5521 
5522 		if (strcmp(type, VDEV_TYPE_RAIDZ) != 0 &&
5523 		    strcmp(type, VDEV_TYPE_DRAID) != 0)
5524 			continue;
5525 
5526 		if (nvlist_lookup_uint64(vdevs[v],
5527 		    ZPOOL_CONFIG_NPARITY, &nparity) != 0)
5528 			continue;
5529 
5530 		if (nvlist_lookup_uint64(vdevs[v],
5531 		    ZPOOL_CONFIG_ASHIFT, &ashift) != 0)
5532 			continue;
5533 
5534 		if (strcmp(type, VDEV_TYPE_RAIDZ) == 0) {
5535 			nvlist_t **disks;
5536 			uint_t ndisks;
5537 
5538 			if (nvlist_lookup_nvlist_array(vdevs[v],
5539 			    ZPOOL_CONFIG_CHILDREN, &disks, &ndisks) != 0)
5540 				continue;
5541 
5542 			/* allocation size for the "typical" 128k block */
5543 			tsize = vdev_raidz_asize(ndisks, nparity, ashift,
5544 			    SPA_OLD_MAXBLOCKSIZE);
5545 
5546 			/* allocation size for the blksize block */
5547 			asize = vdev_raidz_asize(ndisks, nparity, ashift,
5548 			    blksize);
5549 		} else {
5550 			uint64_t ndata;
5551 
5552 			if (nvlist_lookup_uint64(vdevs[v],
5553 			    ZPOOL_CONFIG_DRAID_NDATA, &ndata) != 0)
5554 				continue;
5555 
5556 			/* allocation size for the "typical" 128k block */
5557 			tsize = vdev_draid_asize(ndata + nparity, nparity,
5558 			    ashift, SPA_OLD_MAXBLOCKSIZE);
5559 
5560 			/* allocation size for the blksize block */
5561 			asize = vdev_draid_asize(ndata + nparity, nparity,
5562 			    ashift, blksize);
5563 		}
5564 
5565 		/*
5566 		 * Scale this size down as a ratio of 128k / tsize.
5567 		 * See theory statement above.
5568 		 *
5569 		 * Bitshift is to avoid the case of nblocks * asize < tsize
5570 		 * producing a size of 0.
5571 		 */
5572 		volsize = (nblocks * asize) / (tsize >> SPA_MINBLOCKSHIFT);
5573 		/*
5574 		 * If we would blow UINT64_MAX with this next multiplication,
5575 		 * don't.
5576 		 */
5577 		if (volsize >
5578 		    (UINT64_MAX / (SPA_OLD_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT)))
5579 			volsize = UINT64_MAX;
5580 		else
5581 			volsize *= (SPA_OLD_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT);
5582 
5583 		if (volsize > ret) {
5584 			ret = volsize;
5585 		}
5586 	}
5587 
5588 	if (ret == 0) {
5589 		ret = nblocks * blksize;
5590 	}
5591 
5592 	return (ret);
5593 }
5594 
5595 /*
5596  * Convert the zvol's volume size to an appropriate reservation.  See theory
5597  * comment above.
5598  *
5599  * Note: If this routine is updated, it is necessary to update the ZFS test
5600  * suite's shell version in reservation.shlib.
5601  */
5602 uint64_t
5603 zvol_volsize_to_reservation(zpool_handle_t *zph, uint64_t volsize,
5604     nvlist_t *props)
5605 {
5606 	uint64_t numdb;
5607 	uint64_t nblocks, volblocksize;
5608 	int ncopies;
5609 	const char *strval;
5610 
5611 	if (nvlist_lookup_string(props,
5612 	    zfs_prop_to_name(ZFS_PROP_COPIES), &strval) == 0)
5613 		ncopies = atoi(strval);
5614 	else
5615 		ncopies = 1;
5616 	if (nvlist_lookup_uint64(props,
5617 	    zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE),
5618 	    &volblocksize) != 0)
5619 		volblocksize = ZVOL_DEFAULT_BLOCKSIZE;
5620 
5621 	nblocks = volsize / volblocksize;
5622 	/*
5623 	 * Metadata defaults to using 128k blocks, not volblocksize blocks.  For
5624 	 * this reason, only the data blocks are scaled based on vdev config.
5625 	 */
5626 	volsize = volsize_from_vdevs(zph, nblocks, volblocksize);
5627 
5628 	/* start with metadnode L0-L6 */
5629 	numdb = 7;
5630 	/* calculate number of indirects */
5631 	while (nblocks > 1) {
5632 		nblocks += DNODES_PER_LEVEL - 1;
5633 		nblocks /= DNODES_PER_LEVEL;
5634 		numdb += nblocks;
5635 	}
5636 	numdb *= MIN(SPA_DVAS_PER_BP, ncopies + 1);
5637 	volsize *= ncopies;
5638 	/*
5639 	 * this is exactly DN_MAX_INDBLKSHIFT when metadata isn't
5640 	 * compressed, but in practice they compress down to about
5641 	 * 1100 bytes
5642 	 */
5643 	numdb *= 1ULL << DN_MAX_INDBLKSHIFT;
5644 	volsize += numdb;
5645 	return (volsize);
5646 }
5647 
5648 /*
5649  * Wait for the given activity and return the status of the wait (whether or not
5650  * any waiting was done) in the 'waited' parameter. Non-existent fses are
5651  * reported via the 'missing' parameter, rather than by printing an error
5652  * message. This is convenient when this function is called in a loop over a
5653  * long period of time (as it is, for example, by zfs's wait cmd). In that
5654  * scenario, a fs being exported or destroyed should be considered a normal
5655  * event, so we don't want to print an error when we find that the fs doesn't
5656  * exist.
5657  */
5658 int
5659 zfs_wait_status(zfs_handle_t *zhp, zfs_wait_activity_t activity,
5660     boolean_t *missing, boolean_t *waited)
5661 {
5662 	int error = lzc_wait_fs(zhp->zfs_name, activity, waited);
5663 	*missing = (error == ENOENT);
5664 	if (*missing)
5665 		return (0);
5666 
5667 	if (error != 0) {
5668 		(void) zfs_standard_error_fmt(zhp->zfs_hdl, error,
5669 		    dgettext(TEXT_DOMAIN, "error waiting in fs '%s'"),
5670 		    zhp->zfs_name);
5671 	}
5672 
5673 	return (error);
5674 }
5675