xref: /freebsd/sys/contrib/openzfs/lib/libzfs/libzfs_pool.c (revision 9b37d84c87e69dabc69d818aa4d2fea718bd8b74)
1 // SPDX-License-Identifier: CDDL-1.0
2 /*
3  * CDDL HEADER START
4  *
5  * The contents of this file are subject to the terms of the
6  * Common Development and Distribution License (the "License").
7  * You may not use this file except in compliance with the License.
8  *
9  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
10  * or https://opensource.org/licenses/CDDL-1.0.
11  * See the License for the specific language governing permissions
12  * and limitations under the License.
13  *
14  * When distributing Covered Code, include this CDDL HEADER in each
15  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
16  * If applicable, add the following below this CDDL HEADER, with the
17  * fields enclosed by brackets "[]" replaced with your own identifying
18  * information: Portions Copyright [yyyy] [name of copyright owner]
19  *
20  * CDDL HEADER END
21  */
22 
23 /*
24  * Copyright 2015 Nexenta Systems, Inc.  All rights reserved.
25  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
26  * Copyright (c) 2011, 2024 by Delphix. All rights reserved.
27  * Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>
28  * Copyright (c) 2018 Datto Inc.
29  * Copyright (c) 2017 Open-E, Inc. All Rights Reserved.
30  * Copyright (c) 2017, Intel Corporation.
31  * Copyright (c) 2018, loli10K <ezomori.nozomu@gmail.com>
32  * Copyright (c) 2021, Colm Buckley <colm@tuatha.org>
33  * Copyright (c) 2021, 2023, Klara Inc.
34  */
35 
36 #include <errno.h>
37 #include <libintl.h>
38 #include <stdio.h>
39 #include <stdlib.h>
40 #include <strings.h>
41 #include <unistd.h>
42 #include <libgen.h>
43 #include <zone.h>
44 #include <sys/stat.h>
45 #include <sys/efi_partition.h>
46 #include <sys/systeminfo.h>
47 #include <sys/zfs_ioctl.h>
48 #include <sys/zfs_sysfs.h>
49 #include <sys/vdev_disk.h>
50 #include <sys/types.h>
51 #include <dlfcn.h>
52 #include <libzutil.h>
53 #include <fcntl.h>
54 
55 #include "zfs_namecheck.h"
56 #include "zfs_prop.h"
57 #include "libzfs_impl.h"
58 #include "zfs_comutil.h"
59 #include "zfeature_common.h"
60 
61 static boolean_t zpool_vdev_is_interior(const char *name);
62 
63 typedef struct prop_flags {
64 	unsigned int create:1;	/* Validate property on creation */
65 	unsigned int import:1;	/* Validate property on import */
66 	unsigned int vdevprop:1; /* Validate property as a VDEV property */
67 } prop_flags_t;
68 
69 /*
70  * ====================================================================
71  *   zpool property functions
72  * ====================================================================
73  */
74 
75 static int
76 zpool_get_all_props(zpool_handle_t *zhp)
77 {
78 	zfs_cmd_t zc = {"\0"};
79 	libzfs_handle_t *hdl = zhp->zpool_hdl;
80 
81 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
82 
83 	if (zhp->zpool_n_propnames > 0) {
84 		nvlist_t *innvl = fnvlist_alloc();
85 		fnvlist_add_string_array(innvl, ZPOOL_GET_PROPS_NAMES,
86 		    zhp->zpool_propnames, zhp->zpool_n_propnames);
87 		zcmd_write_src_nvlist(hdl, &zc, innvl);
88 		fnvlist_free(innvl);
89 	}
90 
91 	zcmd_alloc_dst_nvlist(hdl, &zc, 0);
92 
93 	while (zfs_ioctl(hdl, ZFS_IOC_POOL_GET_PROPS, &zc) != 0) {
94 		if (errno == ENOMEM)
95 			zcmd_expand_dst_nvlist(hdl, &zc);
96 		else {
97 			zcmd_free_nvlists(&zc);
98 			return (-1);
99 		}
100 	}
101 
102 	if (zcmd_read_dst_nvlist(hdl, &zc, &zhp->zpool_props) != 0) {
103 		zcmd_free_nvlists(&zc);
104 		return (-1);
105 	}
106 
107 	zcmd_free_nvlists(&zc);
108 
109 	return (0);
110 }
111 
112 int
113 zpool_props_refresh(zpool_handle_t *zhp)
114 {
115 	nvlist_t *old_props;
116 
117 	old_props = zhp->zpool_props;
118 
119 	if (zpool_get_all_props(zhp) != 0)
120 		return (-1);
121 
122 	nvlist_free(old_props);
123 	return (0);
124 }
125 
126 static const char *
127 zpool_get_prop_string(zpool_handle_t *zhp, zpool_prop_t prop,
128     zprop_source_t *src)
129 {
130 	nvlist_t *nv, *nvl;
131 	const char *value;
132 	zprop_source_t source;
133 
134 	nvl = zhp->zpool_props;
135 	if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
136 		source = fnvlist_lookup_uint64(nv, ZPROP_SOURCE);
137 		value = fnvlist_lookup_string(nv, ZPROP_VALUE);
138 	} else {
139 		source = ZPROP_SRC_DEFAULT;
140 		if ((value = zpool_prop_default_string(prop)) == NULL)
141 			value = "-";
142 	}
143 
144 	if (src)
145 		*src = source;
146 
147 	return (value);
148 }
149 
150 uint64_t
151 zpool_get_prop_int(zpool_handle_t *zhp, zpool_prop_t prop, zprop_source_t *src)
152 {
153 	nvlist_t *nv, *nvl;
154 	uint64_t value;
155 	zprop_source_t source;
156 
157 	if (zhp->zpool_props == NULL && zpool_get_all_props(zhp)) {
158 		/*
159 		 * zpool_get_all_props() has most likely failed because
160 		 * the pool is faulted, but if all we need is the top level
161 		 * vdev's guid then get it from the zhp config nvlist.
162 		 */
163 		if ((prop == ZPOOL_PROP_GUID) &&
164 		    (nvlist_lookup_nvlist(zhp->zpool_config,
165 		    ZPOOL_CONFIG_VDEV_TREE, &nv) == 0) &&
166 		    (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value)
167 		    == 0)) {
168 			return (value);
169 		}
170 		return (zpool_prop_default_numeric(prop));
171 	}
172 
173 	nvl = zhp->zpool_props;
174 	if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
175 		source = fnvlist_lookup_uint64(nv, ZPROP_SOURCE);
176 		value = fnvlist_lookup_uint64(nv, ZPROP_VALUE);
177 	} else {
178 		source = ZPROP_SRC_DEFAULT;
179 		value = zpool_prop_default_numeric(prop);
180 	}
181 
182 	if (src)
183 		*src = source;
184 
185 	return (value);
186 }
187 
188 /*
189  * Map VDEV STATE to printed strings.
190  */
191 const char *
192 zpool_state_to_name(vdev_state_t state, vdev_aux_t aux)
193 {
194 	switch (state) {
195 	case VDEV_STATE_CLOSED:
196 	case VDEV_STATE_OFFLINE:
197 		return (gettext("OFFLINE"));
198 	case VDEV_STATE_REMOVED:
199 		return (gettext("REMOVED"));
200 	case VDEV_STATE_CANT_OPEN:
201 		if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG)
202 			return (gettext("FAULTED"));
203 		else if (aux == VDEV_AUX_SPLIT_POOL)
204 			return (gettext("SPLIT"));
205 		else
206 			return (gettext("UNAVAIL"));
207 	case VDEV_STATE_FAULTED:
208 		return (gettext("FAULTED"));
209 	case VDEV_STATE_DEGRADED:
210 		return (gettext("DEGRADED"));
211 	case VDEV_STATE_HEALTHY:
212 		return (gettext("ONLINE"));
213 
214 	default:
215 		break;
216 	}
217 
218 	return (gettext("UNKNOWN"));
219 }
220 
221 /*
222  * Map POOL STATE to printed strings.
223  */
224 const char *
225 zpool_pool_state_to_name(pool_state_t state)
226 {
227 	switch (state) {
228 	default:
229 		break;
230 	case POOL_STATE_ACTIVE:
231 		return (gettext("ACTIVE"));
232 	case POOL_STATE_EXPORTED:
233 		return (gettext("EXPORTED"));
234 	case POOL_STATE_DESTROYED:
235 		return (gettext("DESTROYED"));
236 	case POOL_STATE_SPARE:
237 		return (gettext("SPARE"));
238 	case POOL_STATE_L2CACHE:
239 		return (gettext("L2CACHE"));
240 	case POOL_STATE_UNINITIALIZED:
241 		return (gettext("UNINITIALIZED"));
242 	case POOL_STATE_UNAVAIL:
243 		return (gettext("UNAVAIL"));
244 	case POOL_STATE_POTENTIALLY_ACTIVE:
245 		return (gettext("POTENTIALLY_ACTIVE"));
246 	}
247 
248 	return (gettext("UNKNOWN"));
249 }
250 
251 /*
252  * Given a pool handle, return the pool health string ("ONLINE", "DEGRADED",
253  * "SUSPENDED", etc).
254  */
255 const char *
256 zpool_get_state_str(zpool_handle_t *zhp)
257 {
258 	zpool_errata_t errata;
259 	zpool_status_t status;
260 	const char *str;
261 
262 	status = zpool_get_status(zhp, NULL, &errata);
263 
264 	if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
265 		str = gettext("FAULTED");
266 	} else if (status == ZPOOL_STATUS_IO_FAILURE_WAIT ||
267 	    status == ZPOOL_STATUS_IO_FAILURE_CONTINUE ||
268 	    status == ZPOOL_STATUS_IO_FAILURE_MMP) {
269 		str = gettext("SUSPENDED");
270 	} else {
271 		nvlist_t *nvroot = fnvlist_lookup_nvlist(
272 		    zpool_get_config(zhp, NULL), ZPOOL_CONFIG_VDEV_TREE);
273 		uint_t vsc;
274 		vdev_stat_t *vs = (vdev_stat_t *)fnvlist_lookup_uint64_array(
275 		    nvroot, ZPOOL_CONFIG_VDEV_STATS, &vsc);
276 		str = zpool_state_to_name(vs->vs_state, vs->vs_aux);
277 	}
278 	return (str);
279 }
280 
281 /*
282  * Get a zpool property value for 'prop' and return the value in
283  * a pre-allocated buffer.
284  */
285 int
286 zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf,
287     size_t len, zprop_source_t *srctype, boolean_t literal)
288 {
289 	uint64_t intval;
290 	const char *strval;
291 	zprop_source_t src = ZPROP_SRC_NONE;
292 
293 	if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
294 		switch (prop) {
295 		case ZPOOL_PROP_NAME:
296 			(void) strlcpy(buf, zpool_get_name(zhp), len);
297 			break;
298 
299 		case ZPOOL_PROP_HEALTH:
300 			(void) strlcpy(buf, zpool_get_state_str(zhp), len);
301 			break;
302 
303 		case ZPOOL_PROP_GUID:
304 			intval = zpool_get_prop_int(zhp, prop, &src);
305 			(void) snprintf(buf, len, "%llu", (u_longlong_t)intval);
306 			break;
307 
308 		case ZPOOL_PROP_ALTROOT:
309 		case ZPOOL_PROP_CACHEFILE:
310 		case ZPOOL_PROP_COMMENT:
311 		case ZPOOL_PROP_COMPATIBILITY:
312 			if (zhp->zpool_props != NULL ||
313 			    zpool_get_all_props(zhp) == 0) {
314 				(void) strlcpy(buf,
315 				    zpool_get_prop_string(zhp, prop, &src),
316 				    len);
317 				break;
318 			}
319 			zfs_fallthrough;
320 		default:
321 			(void) strlcpy(buf, "-", len);
322 			break;
323 		}
324 
325 		if (srctype != NULL)
326 			*srctype = src;
327 		return (0);
328 	}
329 
330 	/*
331 	 * ZPOOL_PROP_DEDUPCACHED can be fetched by name only using
332 	 * the ZPOOL_GET_PROPS_NAMES mechanism
333 	 */
334 	if (prop == ZPOOL_PROP_DEDUPCACHED) {
335 		zpool_add_propname(zhp, ZPOOL_DEDUPCACHED_PROP_NAME);
336 		(void) zpool_props_refresh(zhp);
337 	}
338 
339 	if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) &&
340 	    prop != ZPOOL_PROP_NAME)
341 		return (-1);
342 
343 	switch (zpool_prop_get_type(prop)) {
344 	case PROP_TYPE_STRING:
345 		(void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src),
346 		    len);
347 		break;
348 
349 	case PROP_TYPE_NUMBER:
350 		intval = zpool_get_prop_int(zhp, prop, &src);
351 
352 		switch (prop) {
353 		case ZPOOL_PROP_DEDUP_TABLE_QUOTA:
354 			/*
355 			 * If dedup quota is 0, we translate this into 'none'
356 			 * (unless literal is set). And if it is UINT64_MAX
357 			 * we translate that as 'automatic' (limit to size of
358 			 * the dedicated dedup VDEV.  Otherwise, fall throught
359 			 * into the regular number formating.
360 			 */
361 			if (intval == 0) {
362 				(void) strlcpy(buf, literal ? "0" : "none",
363 				    len);
364 				break;
365 			} else if (intval == UINT64_MAX) {
366 				(void) strlcpy(buf, "auto", len);
367 				break;
368 			}
369 			zfs_fallthrough;
370 
371 		case ZPOOL_PROP_SIZE:
372 		case ZPOOL_PROP_ALLOCATED:
373 		case ZPOOL_PROP_FREE:
374 		case ZPOOL_PROP_FREEING:
375 		case ZPOOL_PROP_LEAKED:
376 		case ZPOOL_PROP_ASHIFT:
377 		case ZPOOL_PROP_MAXBLOCKSIZE:
378 		case ZPOOL_PROP_MAXDNODESIZE:
379 		case ZPOOL_PROP_BCLONESAVED:
380 		case ZPOOL_PROP_BCLONEUSED:
381 		case ZPOOL_PROP_DEDUP_TABLE_SIZE:
382 		case ZPOOL_PROP_DEDUPCACHED:
383 			if (literal)
384 				(void) snprintf(buf, len, "%llu",
385 				    (u_longlong_t)intval);
386 			else
387 				(void) zfs_nicenum(intval, buf, len);
388 			break;
389 
390 		case ZPOOL_PROP_EXPANDSZ:
391 		case ZPOOL_PROP_CHECKPOINT:
392 			if (intval == 0) {
393 				(void) strlcpy(buf, "-", len);
394 			} else if (literal) {
395 				(void) snprintf(buf, len, "%llu",
396 				    (u_longlong_t)intval);
397 			} else {
398 				(void) zfs_nicebytes(intval, buf, len);
399 			}
400 			break;
401 
402 		case ZPOOL_PROP_CAPACITY:
403 			if (literal) {
404 				(void) snprintf(buf, len, "%llu",
405 				    (u_longlong_t)intval);
406 			} else {
407 				(void) snprintf(buf, len, "%llu%%",
408 				    (u_longlong_t)intval);
409 			}
410 			break;
411 
412 		case ZPOOL_PROP_FRAGMENTATION:
413 			if (intval == UINT64_MAX) {
414 				(void) strlcpy(buf, "-", len);
415 			} else if (literal) {
416 				(void) snprintf(buf, len, "%llu",
417 				    (u_longlong_t)intval);
418 			} else {
419 				(void) snprintf(buf, len, "%llu%%",
420 				    (u_longlong_t)intval);
421 			}
422 			break;
423 
424 		case ZPOOL_PROP_BCLONERATIO:
425 		case ZPOOL_PROP_DEDUPRATIO:
426 			if (literal)
427 				(void) snprintf(buf, len, "%llu.%02llu",
428 				    (u_longlong_t)(intval / 100),
429 				    (u_longlong_t)(intval % 100));
430 			else
431 				(void) snprintf(buf, len, "%llu.%02llux",
432 				    (u_longlong_t)(intval / 100),
433 				    (u_longlong_t)(intval % 100));
434 			break;
435 
436 		case ZPOOL_PROP_HEALTH:
437 			(void) strlcpy(buf, zpool_get_state_str(zhp), len);
438 			break;
439 		case ZPOOL_PROP_VERSION:
440 			if (intval >= SPA_VERSION_FEATURES) {
441 				(void) snprintf(buf, len, "-");
442 				break;
443 			}
444 			zfs_fallthrough;
445 		default:
446 			(void) snprintf(buf, len, "%llu", (u_longlong_t)intval);
447 		}
448 		break;
449 
450 	case PROP_TYPE_INDEX:
451 		intval = zpool_get_prop_int(zhp, prop, &src);
452 		if (zpool_prop_index_to_string(prop, intval, &strval)
453 		    != 0)
454 			return (-1);
455 		(void) strlcpy(buf, strval, len);
456 		break;
457 
458 	default:
459 		abort();
460 	}
461 
462 	if (srctype)
463 		*srctype = src;
464 
465 	return (0);
466 }
467 
468 /*
469  * Get a zpool property value for 'propname' and return the value in
470  * a pre-allocated buffer.
471  */
472 int
473 zpool_get_userprop(zpool_handle_t *zhp, const char *propname, char *buf,
474     size_t len, zprop_source_t *srctype)
475 {
476 	nvlist_t *nv;
477 	uint64_t ival;
478 	const char *value;
479 	zprop_source_t source = ZPROP_SRC_LOCAL;
480 
481 	if (zhp->zpool_props == NULL)
482 		zpool_get_all_props(zhp);
483 
484 	if (nvlist_lookup_nvlist(zhp->zpool_props, propname, &nv) == 0) {
485 		if (nvlist_lookup_uint64(nv, ZPROP_SOURCE, &ival) == 0)
486 			source = ival;
487 		verify(nvlist_lookup_string(nv, ZPROP_VALUE, &value) == 0);
488 	} else {
489 		source = ZPROP_SRC_DEFAULT;
490 		value = "-";
491 	}
492 
493 	if (srctype)
494 		*srctype = source;
495 
496 	(void) strlcpy(buf, value, len);
497 
498 	return (0);
499 }
500 
501 /*
502  * Check if the bootfs name has the same pool name as it is set to.
503  * Assuming bootfs is a valid dataset name.
504  */
505 static boolean_t
506 bootfs_name_valid(const char *pool, const char *bootfs)
507 {
508 	int len = strlen(pool);
509 	if (bootfs[0] == '\0')
510 		return (B_TRUE);
511 
512 	if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT))
513 		return (B_FALSE);
514 
515 	if (strncmp(pool, bootfs, len) == 0 &&
516 	    (bootfs[len] == '/' || bootfs[len] == '\0'))
517 		return (B_TRUE);
518 
519 	return (B_FALSE);
520 }
521 
522 /*
523  * Given an nvlist of zpool properties to be set, validate that they are
524  * correct, and parse any numeric properties (index, boolean, etc) if they are
525  * specified as strings.
526  */
527 static nvlist_t *
528 zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname,
529     nvlist_t *props, uint64_t version, prop_flags_t flags, char *errbuf)
530 {
531 	nvpair_t *elem;
532 	nvlist_t *retprops;
533 	zpool_prop_t prop;
534 	const char *strval;
535 	uint64_t intval;
536 	const char *check;
537 	struct stat64 statbuf;
538 	zpool_handle_t *zhp;
539 	char *parent, *slash;
540 	char report[1024];
541 
542 	if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) {
543 		(void) no_memory(hdl);
544 		return (NULL);
545 	}
546 
547 	elem = NULL;
548 	while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
549 		const char *propname = nvpair_name(elem);
550 
551 		if (flags.vdevprop && zpool_prop_vdev(propname)) {
552 			vdev_prop_t vprop = vdev_name_to_prop(propname);
553 
554 			if (vdev_prop_readonly(vprop)) {
555 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
556 				    "is readonly"), propname);
557 				(void) zfs_error(hdl, EZFS_PROPREADONLY,
558 				    errbuf);
559 				goto error;
560 			}
561 
562 			if (zprop_parse_value(hdl, elem, vprop, ZFS_TYPE_VDEV,
563 			    retprops, &strval, &intval, errbuf) != 0)
564 				goto error;
565 
566 			continue;
567 		} else if (flags.vdevprop && vdev_prop_user(propname)) {
568 			if (nvlist_add_nvpair(retprops, elem) != 0) {
569 				(void) no_memory(hdl);
570 				goto error;
571 			}
572 			continue;
573 		} else if (flags.vdevprop) {
574 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
575 			    "invalid property: '%s'"), propname);
576 			(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
577 			goto error;
578 		}
579 
580 		prop = zpool_name_to_prop(propname);
581 		if (prop == ZPOOL_PROP_INVAL && zpool_prop_feature(propname)) {
582 			int err;
583 			char *fname = strchr(propname, '@') + 1;
584 
585 			err = zfeature_lookup_name(fname, NULL);
586 			if (err != 0) {
587 				ASSERT3U(err, ==, ENOENT);
588 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
589 				    "feature '%s' unsupported by kernel"),
590 				    fname);
591 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
592 				goto error;
593 			}
594 
595 			if (nvpair_type(elem) != DATA_TYPE_STRING) {
596 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
597 				    "'%s' must be a string"), propname);
598 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
599 				goto error;
600 			}
601 
602 			(void) nvpair_value_string(elem, &strval);
603 			if (strcmp(strval, ZFS_FEATURE_ENABLED) != 0 &&
604 			    strcmp(strval, ZFS_FEATURE_DISABLED) != 0) {
605 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
606 				    "property '%s' can only be set to "
607 				    "'enabled' or 'disabled'"), propname);
608 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
609 				goto error;
610 			}
611 
612 			if (!flags.create &&
613 			    strcmp(strval, ZFS_FEATURE_DISABLED) == 0) {
614 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
615 				    "property '%s' can only be set to "
616 				    "'disabled' at creation time"), propname);
617 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
618 				goto error;
619 			}
620 
621 			if (nvlist_add_uint64(retprops, propname, 0) != 0) {
622 				(void) no_memory(hdl);
623 				goto error;
624 			}
625 			continue;
626 		} else if (prop == ZPOOL_PROP_INVAL &&
627 		    zfs_prop_user(propname)) {
628 			/*
629 			 * This is a user property: make sure it's a
630 			 * string, and that it's less than ZAP_MAXNAMELEN.
631 			 */
632 			if (nvpair_type(elem) != DATA_TYPE_STRING) {
633 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
634 				    "'%s' must be a string"), propname);
635 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
636 				goto error;
637 			}
638 
639 			if (strlen(nvpair_name(elem)) >= ZAP_MAXNAMELEN) {
640 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
641 				    "property name '%s' is too long"),
642 				    propname);
643 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
644 				goto error;
645 			}
646 
647 			(void) nvpair_value_string(elem, &strval);
648 
649 			if (strlen(strval) >= ZFS_MAXPROPLEN) {
650 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
651 				    "property value '%s' is too long"),
652 				    strval);
653 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
654 				goto error;
655 			}
656 
657 			if (nvlist_add_string(retprops, propname,
658 			    strval) != 0) {
659 				(void) no_memory(hdl);
660 				goto error;
661 			}
662 
663 			continue;
664 		}
665 
666 		/*
667 		 * Make sure this property is valid and applies to this type.
668 		 */
669 		if (prop == ZPOOL_PROP_INVAL) {
670 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
671 			    "invalid property '%s'"), propname);
672 			(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
673 			goto error;
674 		}
675 
676 		if (zpool_prop_readonly(prop)) {
677 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
678 			    "is readonly"), propname);
679 			(void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
680 			goto error;
681 		}
682 
683 		if (!flags.create && zpool_prop_setonce(prop)) {
684 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
685 			    "property '%s' can only be set at "
686 			    "creation time"), propname);
687 			(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
688 			goto error;
689 		}
690 
691 		if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops,
692 		    &strval, &intval, errbuf) != 0)
693 			goto error;
694 
695 		/*
696 		 * Perform additional checking for specific properties.
697 		 */
698 		switch (prop) {
699 		case ZPOOL_PROP_VERSION:
700 			if (intval < version ||
701 			    !SPA_VERSION_IS_SUPPORTED(intval)) {
702 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
703 				    "property '%s' number %llu is invalid."),
704 				    propname, (unsigned long long)intval);
705 				(void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
706 				goto error;
707 			}
708 			break;
709 
710 		case ZPOOL_PROP_ASHIFT:
711 			if (intval != 0 &&
712 			    (intval < ASHIFT_MIN || intval > ASHIFT_MAX)) {
713 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
714 				    "property '%s' number %llu is invalid, "
715 				    "only values between %" PRId32 " and %"
716 				    PRId32 " are allowed."),
717 				    propname, (unsigned long long)intval,
718 				    ASHIFT_MIN, ASHIFT_MAX);
719 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
720 				goto error;
721 			}
722 			break;
723 
724 		case ZPOOL_PROP_BOOTFS:
725 			if (flags.create || flags.import) {
726 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
727 				    "property '%s' cannot be set at creation "
728 				    "or import time"), propname);
729 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
730 				goto error;
731 			}
732 
733 			if (version < SPA_VERSION_BOOTFS) {
734 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
735 				    "pool must be upgraded to support "
736 				    "'%s' property"), propname);
737 				(void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
738 				goto error;
739 			}
740 
741 			/*
742 			 * bootfs property value has to be a dataset name and
743 			 * the dataset has to be in the same pool as it sets to.
744 			 */
745 			if (!bootfs_name_valid(poolname, strval)) {
746 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
747 				    "is an invalid name"), strval);
748 				(void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
749 				goto error;
750 			}
751 
752 			if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) {
753 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
754 				    "could not open pool '%s'"), poolname);
755 				(void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
756 				goto error;
757 			}
758 			zpool_close(zhp);
759 			break;
760 
761 		case ZPOOL_PROP_ALTROOT:
762 			if (!flags.create && !flags.import) {
763 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
764 				    "property '%s' can only be set during pool "
765 				    "creation or import"), propname);
766 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
767 				goto error;
768 			}
769 
770 			if (strval[0] != '/') {
771 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
772 				    "bad alternate root '%s'"), strval);
773 				(void) zfs_error(hdl, EZFS_BADPATH, errbuf);
774 				goto error;
775 			}
776 			break;
777 
778 		case ZPOOL_PROP_CACHEFILE:
779 			if (strval[0] == '\0')
780 				break;
781 
782 			if (strcmp(strval, "none") == 0)
783 				break;
784 
785 			if (strval[0] != '/') {
786 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
787 				    "property '%s' must be empty, an "
788 				    "absolute path, or 'none'"), propname);
789 				(void) zfs_error(hdl, EZFS_BADPATH, errbuf);
790 				goto error;
791 			}
792 
793 			parent = strdup(strval);
794 			if (parent == NULL) {
795 				(void) zfs_error(hdl, EZFS_NOMEM, errbuf);
796 				goto error;
797 			}
798 			slash = strrchr(parent, '/');
799 
800 			if (slash[1] == '\0' || strcmp(slash, "/.") == 0 ||
801 			    strcmp(slash, "/..") == 0) {
802 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
803 				    "'%s' is not a valid file"), parent);
804 				(void) zfs_error(hdl, EZFS_BADPATH, errbuf);
805 				free(parent);
806 				goto error;
807 			}
808 
809 			*slash = '\0';
810 
811 			if (parent[0] != '\0' &&
812 			    (stat64(parent, &statbuf) != 0 ||
813 			    !S_ISDIR(statbuf.st_mode))) {
814 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
815 				    "'%s' is not a valid directory"),
816 				    parent);
817 				(void) zfs_error(hdl, EZFS_BADPATH, errbuf);
818 				free(parent);
819 				goto error;
820 			}
821 			free(parent);
822 
823 			break;
824 
825 		case ZPOOL_PROP_COMPATIBILITY:
826 			switch (zpool_load_compat(strval, NULL, report, 1024)) {
827 			case ZPOOL_COMPATIBILITY_OK:
828 			case ZPOOL_COMPATIBILITY_WARNTOKEN:
829 				break;
830 			case ZPOOL_COMPATIBILITY_BADFILE:
831 			case ZPOOL_COMPATIBILITY_BADTOKEN:
832 			case ZPOOL_COMPATIBILITY_NOFILES:
833 				zfs_error_aux(hdl, "%s", report);
834 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
835 				goto error;
836 			}
837 			break;
838 
839 		case ZPOOL_PROP_COMMENT:
840 			for (check = strval; *check != '\0'; check++) {
841 				if (!isprint(*check)) {
842 					zfs_error_aux(hdl,
843 					    dgettext(TEXT_DOMAIN,
844 					    "comment may only have printable "
845 					    "characters"));
846 					(void) zfs_error(hdl, EZFS_BADPROP,
847 					    errbuf);
848 					goto error;
849 				}
850 			}
851 			if (strlen(strval) > ZPROP_MAX_COMMENT) {
852 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
853 				    "comment must not exceed %d characters"),
854 				    ZPROP_MAX_COMMENT);
855 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
856 				goto error;
857 			}
858 			break;
859 		case ZPOOL_PROP_READONLY:
860 			if (!flags.import) {
861 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
862 				    "property '%s' can only be set at "
863 				    "import time"), propname);
864 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
865 				goto error;
866 			}
867 			break;
868 		case ZPOOL_PROP_MULTIHOST:
869 			if (get_system_hostid() == 0) {
870 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
871 				    "requires a non-zero system hostid"));
872 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
873 				goto error;
874 			}
875 			break;
876 		case ZPOOL_PROP_DEDUPDITTO:
877 			printf("Note: property '%s' no longer has "
878 			    "any effect\n", propname);
879 			break;
880 
881 		default:
882 			break;
883 		}
884 	}
885 
886 	return (retprops);
887 error:
888 	nvlist_free(retprops);
889 	return (NULL);
890 }
891 
892 /*
893  * Set zpool property : propname=propval.
894  */
895 int
896 zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval)
897 {
898 	zfs_cmd_t zc = {"\0"};
899 	int ret = -1;
900 	char errbuf[ERRBUFLEN];
901 	nvlist_t *nvl = NULL;
902 	nvlist_t *realprops;
903 	uint64_t version;
904 	prop_flags_t flags = { 0 };
905 
906 	(void) snprintf(errbuf, sizeof (errbuf),
907 	    dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
908 	    zhp->zpool_name);
909 
910 	if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
911 		return (no_memory(zhp->zpool_hdl));
912 
913 	if (nvlist_add_string(nvl, propname, propval) != 0) {
914 		nvlist_free(nvl);
915 		return (no_memory(zhp->zpool_hdl));
916 	}
917 
918 	version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
919 	if ((realprops = zpool_valid_proplist(zhp->zpool_hdl,
920 	    zhp->zpool_name, nvl, version, flags, errbuf)) == NULL) {
921 		nvlist_free(nvl);
922 		return (-1);
923 	}
924 
925 	nvlist_free(nvl);
926 	nvl = realprops;
927 
928 	/*
929 	 * Execute the corresponding ioctl() to set this property.
930 	 */
931 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
932 
933 	zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl);
934 
935 	ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc);
936 
937 	zcmd_free_nvlists(&zc);
938 	nvlist_free(nvl);
939 
940 	if (ret)
941 		(void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
942 	else
943 		(void) zpool_props_refresh(zhp);
944 
945 	return (ret);
946 }
947 
948 int
949 zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp,
950     zfs_type_t type, boolean_t literal)
951 {
952 	libzfs_handle_t *hdl = zhp->zpool_hdl;
953 	zprop_list_t *entry;
954 	char buf[ZFS_MAXPROPLEN];
955 	nvlist_t *features = NULL;
956 	nvpair_t *nvp;
957 	zprop_list_t **last;
958 	boolean_t firstexpand = (NULL == *plp);
959 	int i;
960 
961 	if (zprop_expand_list(hdl, plp, type) != 0)
962 		return (-1);
963 
964 	if (type == ZFS_TYPE_VDEV)
965 		return (0);
966 
967 	last = plp;
968 	while (*last != NULL)
969 		last = &(*last)->pl_next;
970 
971 	if ((*plp)->pl_all)
972 		features = zpool_get_features(zhp);
973 
974 	if ((*plp)->pl_all && firstexpand) {
975 		/* Handle userprops in the all properties case */
976 		if (zhp->zpool_props == NULL && zpool_props_refresh(zhp))
977 			return (-1);
978 
979 		nvp = NULL;
980 		while ((nvp = nvlist_next_nvpair(zhp->zpool_props, nvp)) !=
981 		    NULL) {
982 			const char *propname = nvpair_name(nvp);
983 
984 			if (!zfs_prop_user(propname))
985 				continue;
986 
987 			entry = zfs_alloc(hdl, sizeof (zprop_list_t));
988 			entry->pl_prop = ZPROP_USERPROP;
989 			entry->pl_user_prop = zfs_strdup(hdl, propname);
990 			entry->pl_width = strlen(entry->pl_user_prop);
991 			entry->pl_all = B_TRUE;
992 
993 			*last = entry;
994 			last = &entry->pl_next;
995 		}
996 
997 		for (i = 0; i < SPA_FEATURES; i++) {
998 			entry = zfs_alloc(hdl, sizeof (zprop_list_t));
999 			entry->pl_prop = ZPROP_USERPROP;
1000 			entry->pl_user_prop = zfs_asprintf(hdl, "feature@%s",
1001 			    spa_feature_table[i].fi_uname);
1002 			entry->pl_width = strlen(entry->pl_user_prop);
1003 			entry->pl_all = B_TRUE;
1004 
1005 			*last = entry;
1006 			last = &entry->pl_next;
1007 		}
1008 	}
1009 
1010 	/* add any unsupported features */
1011 	for (nvp = nvlist_next_nvpair(features, NULL);
1012 	    nvp != NULL; nvp = nvlist_next_nvpair(features, nvp)) {
1013 		char *propname;
1014 		boolean_t found;
1015 
1016 		if (zfeature_is_supported(nvpair_name(nvp)))
1017 			continue;
1018 
1019 		propname = zfs_asprintf(hdl, "unsupported@%s",
1020 		    nvpair_name(nvp));
1021 
1022 		/*
1023 		 * Before adding the property to the list make sure that no
1024 		 * other pool already added the same property.
1025 		 */
1026 		found = B_FALSE;
1027 		entry = *plp;
1028 		while (entry != NULL) {
1029 			if (entry->pl_user_prop != NULL &&
1030 			    strcmp(propname, entry->pl_user_prop) == 0) {
1031 				found = B_TRUE;
1032 				break;
1033 			}
1034 			entry = entry->pl_next;
1035 		}
1036 		if (found) {
1037 			free(propname);
1038 			continue;
1039 		}
1040 
1041 		entry = zfs_alloc(hdl, sizeof (zprop_list_t));
1042 		entry->pl_prop = ZPROP_USERPROP;
1043 		entry->pl_user_prop = propname;
1044 		entry->pl_width = strlen(entry->pl_user_prop);
1045 		entry->pl_all = B_TRUE;
1046 
1047 		*last = entry;
1048 		last = &entry->pl_next;
1049 	}
1050 
1051 	for (entry = *plp; entry != NULL; entry = entry->pl_next) {
1052 		if (entry->pl_fixed && !literal)
1053 			continue;
1054 
1055 		if (entry->pl_prop != ZPROP_USERPROP &&
1056 		    zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf),
1057 		    NULL, literal) == 0) {
1058 			if (strlen(buf) > entry->pl_width)
1059 				entry->pl_width = strlen(buf);
1060 		} else if (entry->pl_prop == ZPROP_INVAL &&
1061 		    zfs_prop_user(entry->pl_user_prop) &&
1062 		    zpool_get_userprop(zhp, entry->pl_user_prop, buf,
1063 		    sizeof (buf), NULL) == 0) {
1064 			if (strlen(buf) > entry->pl_width)
1065 				entry->pl_width = strlen(buf);
1066 		}
1067 	}
1068 
1069 	return (0);
1070 }
1071 
1072 int
1073 vdev_expand_proplist(zpool_handle_t *zhp, const char *vdevname,
1074     zprop_list_t **plp)
1075 {
1076 	zprop_list_t *entry;
1077 	char buf[ZFS_MAXPROPLEN];
1078 	const char *strval = NULL;
1079 	int err = 0;
1080 	nvpair_t *elem = NULL;
1081 	nvlist_t *vprops = NULL;
1082 	nvlist_t *propval = NULL;
1083 	const char *propname;
1084 	vdev_prop_t prop;
1085 	zprop_list_t **last;
1086 
1087 	for (entry = *plp; entry != NULL; entry = entry->pl_next) {
1088 		if (entry->pl_fixed)
1089 			continue;
1090 
1091 		if (zpool_get_vdev_prop(zhp, vdevname, entry->pl_prop,
1092 		    entry->pl_user_prop, buf, sizeof (buf), NULL,
1093 		    B_FALSE) == 0) {
1094 			if (strlen(buf) > entry->pl_width)
1095 				entry->pl_width = strlen(buf);
1096 		}
1097 		if (entry->pl_prop == VDEV_PROP_NAME &&
1098 		    strlen(vdevname) > entry->pl_width)
1099 			entry->pl_width = strlen(vdevname);
1100 	}
1101 
1102 	/* Handle the all properties case */
1103 	last = plp;
1104 	if (*last != NULL && (*last)->pl_all == B_TRUE) {
1105 		while (*last != NULL)
1106 			last = &(*last)->pl_next;
1107 
1108 		err = zpool_get_all_vdev_props(zhp, vdevname, &vprops);
1109 		if (err != 0)
1110 			return (err);
1111 
1112 		while ((elem = nvlist_next_nvpair(vprops, elem)) != NULL) {
1113 			propname = nvpair_name(elem);
1114 
1115 			/* Skip properties that are not user defined */
1116 			if ((prop = vdev_name_to_prop(propname)) !=
1117 			    VDEV_PROP_USERPROP)
1118 				continue;
1119 
1120 			if (nvpair_value_nvlist(elem, &propval) != 0)
1121 				continue;
1122 
1123 			strval = fnvlist_lookup_string(propval, ZPROP_VALUE);
1124 
1125 			entry = zfs_alloc(zhp->zpool_hdl,
1126 			    sizeof (zprop_list_t));
1127 			entry->pl_prop = prop;
1128 			entry->pl_user_prop = zfs_strdup(zhp->zpool_hdl,
1129 			    propname);
1130 			entry->pl_width = strlen(strval);
1131 			entry->pl_all = B_TRUE;
1132 			*last = entry;
1133 			last = &entry->pl_next;
1134 		}
1135 	}
1136 
1137 	return (0);
1138 }
1139 
1140 /*
1141  * Get the state for the given feature on the given ZFS pool.
1142  */
1143 int
1144 zpool_prop_get_feature(zpool_handle_t *zhp, const char *propname, char *buf,
1145     size_t len)
1146 {
1147 	uint64_t refcount;
1148 	boolean_t found = B_FALSE;
1149 	nvlist_t *features = zpool_get_features(zhp);
1150 	boolean_t supported;
1151 	const char *feature = strchr(propname, '@') + 1;
1152 
1153 	supported = zpool_prop_feature(propname);
1154 	ASSERT(supported || zpool_prop_unsupported(propname));
1155 
1156 	/*
1157 	 * Convert from feature name to feature guid. This conversion is
1158 	 * unnecessary for unsupported@... properties because they already
1159 	 * use guids.
1160 	 */
1161 	if (supported) {
1162 		int ret;
1163 		spa_feature_t fid;
1164 
1165 		ret = zfeature_lookup_name(feature, &fid);
1166 		if (ret != 0) {
1167 			(void) strlcpy(buf, "-", len);
1168 			return (ENOTSUP);
1169 		}
1170 		feature = spa_feature_table[fid].fi_guid;
1171 	}
1172 
1173 	if (nvlist_lookup_uint64(features, feature, &refcount) == 0)
1174 		found = B_TRUE;
1175 
1176 	if (supported) {
1177 		if (!found) {
1178 			(void) strlcpy(buf, ZFS_FEATURE_DISABLED, len);
1179 		} else  {
1180 			if (refcount == 0)
1181 				(void) strlcpy(buf, ZFS_FEATURE_ENABLED, len);
1182 			else
1183 				(void) strlcpy(buf, ZFS_FEATURE_ACTIVE, len);
1184 		}
1185 	} else {
1186 		if (found) {
1187 			if (refcount == 0) {
1188 				(void) strcpy(buf, ZFS_UNSUPPORTED_INACTIVE);
1189 			} else {
1190 				(void) strcpy(buf, ZFS_UNSUPPORTED_READONLY);
1191 			}
1192 		} else {
1193 			(void) strlcpy(buf, "-", len);
1194 			return (ENOTSUP);
1195 		}
1196 	}
1197 
1198 	return (0);
1199 }
1200 
1201 /*
1202  * Validate the given pool name, optionally putting an extended error message in
1203  * 'buf'.
1204  */
1205 boolean_t
1206 zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool)
1207 {
1208 	namecheck_err_t why;
1209 	char what;
1210 	int ret;
1211 
1212 	ret = pool_namecheck(pool, &why, &what);
1213 
1214 	/*
1215 	 * The rules for reserved pool names were extended at a later point.
1216 	 * But we need to support users with existing pools that may now be
1217 	 * invalid.  So we only check for this expanded set of names during a
1218 	 * create (or import), and only in userland.
1219 	 */
1220 	if (ret == 0 && !isopen &&
1221 	    (strncmp(pool, "mirror", 6) == 0 ||
1222 	    strncmp(pool, "raidz", 5) == 0 ||
1223 	    strncmp(pool, "draid", 5) == 0 ||
1224 	    strncmp(pool, "spare", 5) == 0 ||
1225 	    strcmp(pool, "log") == 0)) {
1226 		if (hdl != NULL)
1227 			zfs_error_aux(hdl,
1228 			    dgettext(TEXT_DOMAIN, "name is reserved"));
1229 		return (B_FALSE);
1230 	}
1231 
1232 
1233 	if (ret != 0) {
1234 		if (hdl != NULL) {
1235 			switch (why) {
1236 			case NAME_ERR_TOOLONG:
1237 				zfs_error_aux(hdl,
1238 				    dgettext(TEXT_DOMAIN, "name is too long"));
1239 				break;
1240 
1241 			case NAME_ERR_INVALCHAR:
1242 				zfs_error_aux(hdl,
1243 				    dgettext(TEXT_DOMAIN, "invalid character "
1244 				    "'%c' in pool name"), what);
1245 				break;
1246 
1247 			case NAME_ERR_NOLETTER:
1248 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1249 				    "name must begin with a letter"));
1250 				break;
1251 
1252 			case NAME_ERR_RESERVED:
1253 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1254 				    "name is reserved"));
1255 				break;
1256 
1257 			case NAME_ERR_DISKLIKE:
1258 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1259 				    "pool name is reserved"));
1260 				break;
1261 
1262 			case NAME_ERR_LEADING_SLASH:
1263 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1264 				    "leading slash in name"));
1265 				break;
1266 
1267 			case NAME_ERR_EMPTY_COMPONENT:
1268 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1269 				    "empty component in name"));
1270 				break;
1271 
1272 			case NAME_ERR_TRAILING_SLASH:
1273 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1274 				    "trailing slash in name"));
1275 				break;
1276 
1277 			case NAME_ERR_MULTIPLE_DELIMITERS:
1278 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1279 				    "multiple '@' and/or '#' delimiters in "
1280 				    "name"));
1281 				break;
1282 
1283 			case NAME_ERR_NO_AT:
1284 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1285 				    "permission set is missing '@'"));
1286 				break;
1287 
1288 			default:
1289 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1290 				    "(%d) not defined"), why);
1291 				break;
1292 			}
1293 		}
1294 		return (B_FALSE);
1295 	}
1296 
1297 	return (B_TRUE);
1298 }
1299 
1300 /*
1301  * Open a handle to the given pool, even if the pool is currently in the FAULTED
1302  * state.
1303  */
1304 zpool_handle_t *
1305 zpool_open_canfail(libzfs_handle_t *hdl, const char *pool)
1306 {
1307 	zpool_handle_t *zhp;
1308 	boolean_t missing;
1309 
1310 	/*
1311 	 * Make sure the pool name is valid.
1312 	 */
1313 	if (!zpool_name_valid(hdl, B_TRUE, pool)) {
1314 		(void) zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1315 		    dgettext(TEXT_DOMAIN, "cannot open '%s'"),
1316 		    pool);
1317 		return (NULL);
1318 	}
1319 
1320 	zhp = zfs_alloc(hdl, sizeof (zpool_handle_t));
1321 
1322 	zhp->zpool_hdl = hdl;
1323 	(void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
1324 
1325 	if (zpool_refresh_stats(zhp, &missing) != 0) {
1326 		zpool_close(zhp);
1327 		return (NULL);
1328 	}
1329 
1330 	if (missing) {
1331 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool"));
1332 		(void) zfs_error_fmt(hdl, EZFS_NOENT,
1333 		    dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool);
1334 		zpool_close(zhp);
1335 		return (NULL);
1336 	}
1337 
1338 	return (zhp);
1339 }
1340 
1341 /*
1342  * Like the above, but silent on error.  Used when iterating over pools (because
1343  * the configuration cache may be out of date).
1344  */
1345 int
1346 zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret)
1347 {
1348 	zpool_handle_t *zhp;
1349 	boolean_t missing;
1350 
1351 	zhp = zfs_alloc(hdl, sizeof (zpool_handle_t));
1352 
1353 	zhp->zpool_hdl = hdl;
1354 	(void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
1355 
1356 	if (zpool_refresh_stats(zhp, &missing) != 0) {
1357 		zpool_close(zhp);
1358 		return (-1);
1359 	}
1360 
1361 	if (missing) {
1362 		zpool_close(zhp);
1363 		*ret = NULL;
1364 		return (0);
1365 	}
1366 
1367 	*ret = zhp;
1368 	return (0);
1369 }
1370 
1371 /*
1372  * Similar to zpool_open_canfail(), but refuses to open pools in the faulted
1373  * state.
1374  */
1375 zpool_handle_t *
1376 zpool_open(libzfs_handle_t *hdl, const char *pool)
1377 {
1378 	zpool_handle_t *zhp;
1379 
1380 	if ((zhp = zpool_open_canfail(hdl, pool)) == NULL)
1381 		return (NULL);
1382 
1383 	if (zhp->zpool_state == POOL_STATE_UNAVAIL) {
1384 		(void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
1385 		    dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name);
1386 		zpool_close(zhp);
1387 		return (NULL);
1388 	}
1389 
1390 	return (zhp);
1391 }
1392 
1393 /*
1394  * Close the handle.  Simply frees the memory associated with the handle.
1395  */
1396 void
1397 zpool_close(zpool_handle_t *zhp)
1398 {
1399 	nvlist_free(zhp->zpool_config);
1400 	nvlist_free(zhp->zpool_old_config);
1401 	nvlist_free(zhp->zpool_props);
1402 	free(zhp);
1403 }
1404 
1405 /*
1406  * Return the name of the pool.
1407  */
1408 const char *
1409 zpool_get_name(zpool_handle_t *zhp)
1410 {
1411 	return (zhp->zpool_name);
1412 }
1413 
1414 
1415 /*
1416  * Return the state of the pool (ACTIVE or UNAVAILABLE)
1417  */
1418 int
1419 zpool_get_state(zpool_handle_t *zhp)
1420 {
1421 	return (zhp->zpool_state);
1422 }
1423 
1424 /*
1425  * Check if vdev list contains a special vdev
1426  */
1427 static boolean_t
1428 zpool_has_special_vdev(nvlist_t *nvroot)
1429 {
1430 	nvlist_t **child;
1431 	uint_t children;
1432 
1433 	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, &child,
1434 	    &children) == 0) {
1435 		for (uint_t c = 0; c < children; c++) {
1436 			const char *bias;
1437 
1438 			if (nvlist_lookup_string(child[c],
1439 			    ZPOOL_CONFIG_ALLOCATION_BIAS, &bias) == 0 &&
1440 			    strcmp(bias, VDEV_ALLOC_BIAS_SPECIAL) == 0) {
1441 				return (B_TRUE);
1442 			}
1443 		}
1444 	}
1445 	return (B_FALSE);
1446 }
1447 
1448 /*
1449  * Check if vdev list contains a dRAID vdev
1450  */
1451 static boolean_t
1452 zpool_has_draid_vdev(nvlist_t *nvroot)
1453 {
1454 	nvlist_t **child;
1455 	uint_t children;
1456 
1457 	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
1458 	    &child, &children) == 0) {
1459 		for (uint_t c = 0; c < children; c++) {
1460 			const char *type;
1461 
1462 			if (nvlist_lookup_string(child[c],
1463 			    ZPOOL_CONFIG_TYPE, &type) == 0 &&
1464 			    strcmp(type, VDEV_TYPE_DRAID) == 0) {
1465 				return (B_TRUE);
1466 			}
1467 		}
1468 	}
1469 	return (B_FALSE);
1470 }
1471 
1472 /*
1473  * Output a dRAID top-level vdev name in to the provided buffer.
1474  */
1475 static char *
1476 zpool_draid_name(char *name, int len, uint64_t data, uint64_t parity,
1477     uint64_t spares, uint64_t children)
1478 {
1479 	snprintf(name, len, "%s%llu:%llud:%lluc:%llus",
1480 	    VDEV_TYPE_DRAID, (u_longlong_t)parity, (u_longlong_t)data,
1481 	    (u_longlong_t)children, (u_longlong_t)spares);
1482 
1483 	return (name);
1484 }
1485 
1486 /*
1487  * Return B_TRUE if the provided name is a dRAID spare name.
1488  */
1489 boolean_t
1490 zpool_is_draid_spare(const char *name)
1491 {
1492 	uint64_t spare_id, parity, vdev_id;
1493 
1494 	if (sscanf(name, VDEV_TYPE_DRAID "%llu-%llu-%llu",
1495 	    (u_longlong_t *)&parity, (u_longlong_t *)&vdev_id,
1496 	    (u_longlong_t *)&spare_id) == 3) {
1497 		return (B_TRUE);
1498 	}
1499 
1500 	return (B_FALSE);
1501 }
1502 
1503 /*
1504  * Create the named pool, using the provided vdev list.  It is assumed
1505  * that the consumer has already validated the contents of the nvlist, so we
1506  * don't have to worry about error semantics.
1507  */
1508 int
1509 zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot,
1510     nvlist_t *props, nvlist_t *fsprops)
1511 {
1512 	zfs_cmd_t zc = {"\0"};
1513 	nvlist_t *zc_fsprops = NULL;
1514 	nvlist_t *zc_props = NULL;
1515 	nvlist_t *hidden_args = NULL;
1516 	uint8_t *wkeydata = NULL;
1517 	uint_t wkeylen = 0;
1518 	char errbuf[ERRBUFLEN];
1519 	int ret = -1;
1520 
1521 	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1522 	    "cannot create '%s'"), pool);
1523 
1524 	if (!zpool_name_valid(hdl, B_FALSE, pool))
1525 		return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
1526 
1527 	zcmd_write_conf_nvlist(hdl, &zc, nvroot);
1528 
1529 	if (props) {
1530 		prop_flags_t flags = { .create = B_TRUE, .import = B_FALSE };
1531 
1532 		if ((zc_props = zpool_valid_proplist(hdl, pool, props,
1533 		    SPA_VERSION_1, flags, errbuf)) == NULL) {
1534 			goto create_failed;
1535 		}
1536 	}
1537 
1538 	if (fsprops) {
1539 		uint64_t zoned;
1540 		const char *zonestr;
1541 
1542 		zoned = ((nvlist_lookup_string(fsprops,
1543 		    zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) &&
1544 		    strcmp(zonestr, "on") == 0);
1545 
1546 		if ((zc_fsprops = zfs_valid_proplist(hdl, ZFS_TYPE_FILESYSTEM,
1547 		    fsprops, zoned, NULL, NULL, B_TRUE, errbuf)) == NULL) {
1548 			goto create_failed;
1549 		}
1550 
1551 		if (nvlist_exists(zc_fsprops,
1552 		    zfs_prop_to_name(ZFS_PROP_SPECIAL_SMALL_BLOCKS)) &&
1553 		    !zpool_has_special_vdev(nvroot)) {
1554 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1555 			    "%s property requires a special vdev"),
1556 			    zfs_prop_to_name(ZFS_PROP_SPECIAL_SMALL_BLOCKS));
1557 			(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
1558 			goto create_failed;
1559 		}
1560 
1561 		if (!zc_props &&
1562 		    (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) {
1563 			goto create_failed;
1564 		}
1565 		if (zfs_crypto_create(hdl, NULL, zc_fsprops, props, B_TRUE,
1566 		    &wkeydata, &wkeylen) != 0) {
1567 			zfs_error(hdl, EZFS_CRYPTOFAILED, errbuf);
1568 			goto create_failed;
1569 		}
1570 		if (nvlist_add_nvlist(zc_props,
1571 		    ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) {
1572 			goto create_failed;
1573 		}
1574 		if (wkeydata != NULL) {
1575 			if (nvlist_alloc(&hidden_args, NV_UNIQUE_NAME, 0) != 0)
1576 				goto create_failed;
1577 
1578 			if (nvlist_add_uint8_array(hidden_args, "wkeydata",
1579 			    wkeydata, wkeylen) != 0)
1580 				goto create_failed;
1581 
1582 			if (nvlist_add_nvlist(zc_props, ZPOOL_HIDDEN_ARGS,
1583 			    hidden_args) != 0)
1584 				goto create_failed;
1585 		}
1586 	}
1587 
1588 	if (zc_props)
1589 		zcmd_write_src_nvlist(hdl, &zc, zc_props);
1590 
1591 	(void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name));
1592 
1593 	if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) {
1594 
1595 		zcmd_free_nvlists(&zc);
1596 		nvlist_free(zc_props);
1597 		nvlist_free(zc_fsprops);
1598 		nvlist_free(hidden_args);
1599 		if (wkeydata != NULL)
1600 			free(wkeydata);
1601 
1602 		switch (errno) {
1603 		case EBUSY:
1604 			/*
1605 			 * This can happen if the user has specified the same
1606 			 * device multiple times.  We can't reliably detect this
1607 			 * until we try to add it and see we already have a
1608 			 * label.  This can also happen under if the device is
1609 			 * part of an active md or lvm device.
1610 			 */
1611 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1612 			    "one or more vdevs refer to the same device, or "
1613 			    "one of\nthe devices is part of an active md or "
1614 			    "lvm device"));
1615 			return (zfs_error(hdl, EZFS_BADDEV, errbuf));
1616 
1617 		case ERANGE:
1618 			/*
1619 			 * This happens if the record size is smaller or larger
1620 			 * than the allowed size range, or not a power of 2.
1621 			 *
1622 			 * NOTE: although zfs_valid_proplist is called earlier,
1623 			 * this case may have slipped through since the
1624 			 * pool does not exist yet and it is therefore
1625 			 * impossible to read properties e.g. max blocksize
1626 			 * from the pool.
1627 			 */
1628 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1629 			    "record size invalid"));
1630 			return (zfs_error(hdl, EZFS_BADPROP, errbuf));
1631 
1632 		case EOVERFLOW:
1633 			/*
1634 			 * This occurs when one of the devices is below
1635 			 * SPA_MINDEVSIZE.  Unfortunately, we can't detect which
1636 			 * device was the problem device since there's no
1637 			 * reliable way to determine device size from userland.
1638 			 */
1639 			{
1640 				char buf[64];
1641 
1642 				zfs_nicebytes(SPA_MINDEVSIZE, buf,
1643 				    sizeof (buf));
1644 
1645 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1646 				    "one or more devices is less than the "
1647 				    "minimum size (%s)"), buf);
1648 			}
1649 			return (zfs_error(hdl, EZFS_BADDEV, errbuf));
1650 
1651 		case ENOSPC:
1652 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1653 			    "one or more devices is out of space"));
1654 			return (zfs_error(hdl, EZFS_BADDEV, errbuf));
1655 
1656 		case EINVAL:
1657 			if (zpool_has_draid_vdev(nvroot) &&
1658 			    zfeature_lookup_name("draid", NULL) != 0) {
1659 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1660 				    "dRAID vdevs are unsupported by the "
1661 				    "kernel"));
1662 				return (zfs_error(hdl, EZFS_BADDEV, errbuf));
1663 			} else {
1664 				return (zpool_standard_error(hdl, errno,
1665 				    errbuf));
1666 			}
1667 
1668 		default:
1669 			return (zpool_standard_error(hdl, errno, errbuf));
1670 		}
1671 	}
1672 
1673 create_failed:
1674 	zcmd_free_nvlists(&zc);
1675 	nvlist_free(zc_props);
1676 	nvlist_free(zc_fsprops);
1677 	nvlist_free(hidden_args);
1678 	if (wkeydata != NULL)
1679 		free(wkeydata);
1680 	return (ret);
1681 }
1682 
1683 /*
1684  * Destroy the given pool.  It is up to the caller to ensure that there are no
1685  * datasets left in the pool.
1686  */
1687 int
1688 zpool_destroy(zpool_handle_t *zhp, const char *log_str)
1689 {
1690 	zfs_cmd_t zc = {"\0"};
1691 	zfs_handle_t *zfp = NULL;
1692 	libzfs_handle_t *hdl = zhp->zpool_hdl;
1693 	char errbuf[ERRBUFLEN];
1694 
1695 	if (zhp->zpool_state == POOL_STATE_ACTIVE &&
1696 	    (zfp = zfs_open(hdl, zhp->zpool_name, ZFS_TYPE_FILESYSTEM)) == NULL)
1697 		return (-1);
1698 
1699 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1700 	zc.zc_history = (uint64_t)(uintptr_t)log_str;
1701 
1702 	if (zfs_ioctl(hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) {
1703 		(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1704 		    "cannot destroy '%s'"), zhp->zpool_name);
1705 
1706 		if (errno == EROFS) {
1707 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1708 			    "one or more devices is read only"));
1709 			(void) zfs_error(hdl, EZFS_BADDEV, errbuf);
1710 		} else {
1711 			(void) zpool_standard_error(hdl, errno, errbuf);
1712 		}
1713 
1714 		if (zfp)
1715 			zfs_close(zfp);
1716 		return (-1);
1717 	}
1718 
1719 	if (zfp) {
1720 		remove_mountpoint(zfp);
1721 		zfs_close(zfp);
1722 	}
1723 
1724 	return (0);
1725 }
1726 
1727 /*
1728  * Create a checkpoint in the given pool.
1729  */
1730 int
1731 zpool_checkpoint(zpool_handle_t *zhp)
1732 {
1733 	libzfs_handle_t *hdl = zhp->zpool_hdl;
1734 	char errbuf[ERRBUFLEN];
1735 	int error;
1736 
1737 	error = lzc_pool_checkpoint(zhp->zpool_name);
1738 	if (error != 0) {
1739 		(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1740 		    "cannot checkpoint '%s'"), zhp->zpool_name);
1741 		(void) zpool_standard_error(hdl, error, errbuf);
1742 		return (-1);
1743 	}
1744 
1745 	return (0);
1746 }
1747 
1748 /*
1749  * Discard the checkpoint from the given pool.
1750  */
1751 int
1752 zpool_discard_checkpoint(zpool_handle_t *zhp)
1753 {
1754 	libzfs_handle_t *hdl = zhp->zpool_hdl;
1755 	char errbuf[ERRBUFLEN];
1756 	int error;
1757 
1758 	error = lzc_pool_checkpoint_discard(zhp->zpool_name);
1759 	if (error != 0) {
1760 		(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1761 		    "cannot discard checkpoint in '%s'"), zhp->zpool_name);
1762 		(void) zpool_standard_error(hdl, error, errbuf);
1763 		return (-1);
1764 	}
1765 
1766 	return (0);
1767 }
1768 
1769 /*
1770  * Load data type for the given pool.
1771  */
1772 int
1773 zpool_prefetch(zpool_handle_t *zhp, zpool_prefetch_type_t type)
1774 {
1775 	libzfs_handle_t *hdl = zhp->zpool_hdl;
1776 	char msg[1024];
1777 	int error;
1778 
1779 	error = lzc_pool_prefetch(zhp->zpool_name, type);
1780 	if (error != 0) {
1781 		(void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1782 		    "cannot prefetch %s in '%s'"),
1783 		    type == ZPOOL_PREFETCH_DDT ? "ddt" : "", zhp->zpool_name);
1784 		(void) zpool_standard_error(hdl, error, msg);
1785 		return (-1);
1786 	}
1787 
1788 	return (0);
1789 }
1790 
1791 /*
1792  * Add the given vdevs to the pool.  The caller must have already performed the
1793  * necessary verification to ensure that the vdev specification is well-formed.
1794  */
1795 int
1796 zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot, boolean_t check_ashift)
1797 {
1798 	zfs_cmd_t zc = {"\0"};
1799 	int ret;
1800 	libzfs_handle_t *hdl = zhp->zpool_hdl;
1801 	char errbuf[ERRBUFLEN];
1802 	nvlist_t **spares, **l2cache;
1803 	uint_t nspares, nl2cache;
1804 
1805 	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1806 	    "cannot add to '%s'"), zhp->zpool_name);
1807 
1808 	if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1809 	    SPA_VERSION_SPARES &&
1810 	    nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
1811 	    &spares, &nspares) == 0) {
1812 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1813 		    "upgraded to add hot spares"));
1814 		return (zfs_error(hdl, EZFS_BADVERSION, errbuf));
1815 	}
1816 
1817 	if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1818 	    SPA_VERSION_L2CACHE &&
1819 	    nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
1820 	    &l2cache, &nl2cache) == 0) {
1821 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1822 		    "upgraded to add cache devices"));
1823 		return (zfs_error(hdl, EZFS_BADVERSION, errbuf));
1824 	}
1825 
1826 	zcmd_write_conf_nvlist(hdl, &zc, nvroot);
1827 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1828 	zc.zc_flags = check_ashift;
1829 
1830 	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) {
1831 		switch (errno) {
1832 		case EBUSY:
1833 			/*
1834 			 * This can happen if the user has specified the same
1835 			 * device multiple times.  We can't reliably detect this
1836 			 * until we try to add it and see we already have a
1837 			 * label.
1838 			 */
1839 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1840 			    "one or more vdevs refer to the same device"));
1841 			(void) zfs_error(hdl, EZFS_BADDEV, errbuf);
1842 			break;
1843 
1844 		case EINVAL:
1845 
1846 			if (zpool_has_draid_vdev(nvroot) &&
1847 			    zfeature_lookup_name("draid", NULL) != 0) {
1848 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1849 				    "dRAID vdevs are unsupported by the "
1850 				    "kernel"));
1851 			} else {
1852 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1853 				    "invalid config; a pool with removing/"
1854 				    "removed vdevs does not support adding "
1855 				    "raidz or dRAID vdevs"));
1856 			}
1857 
1858 			(void) zfs_error(hdl, EZFS_BADDEV, errbuf);
1859 			break;
1860 
1861 		case EOVERFLOW:
1862 			/*
1863 			 * This occurs when one of the devices is below
1864 			 * SPA_MINDEVSIZE.  Unfortunately, we can't detect which
1865 			 * device was the problem device since there's no
1866 			 * reliable way to determine device size from userland.
1867 			 */
1868 			{
1869 				char buf[64];
1870 
1871 				zfs_nicebytes(SPA_MINDEVSIZE, buf,
1872 				    sizeof (buf));
1873 
1874 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1875 				    "device is less than the minimum "
1876 				    "size (%s)"), buf);
1877 			}
1878 			(void) zfs_error(hdl, EZFS_BADDEV, errbuf);
1879 			break;
1880 
1881 		case ENOTSUP:
1882 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1883 			    "pool must be upgraded to add these vdevs"));
1884 			(void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
1885 			break;
1886 
1887 		default:
1888 			(void) zpool_standard_error(hdl, errno, errbuf);
1889 		}
1890 
1891 		ret = -1;
1892 	} else {
1893 		ret = 0;
1894 	}
1895 
1896 	zcmd_free_nvlists(&zc);
1897 
1898 	return (ret);
1899 }
1900 
1901 /*
1902  * Exports the pool from the system.  The caller must ensure that there are no
1903  * mounted datasets in the pool.
1904  */
1905 static int
1906 zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce,
1907     const char *log_str)
1908 {
1909 	zfs_cmd_t zc = {"\0"};
1910 
1911 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1912 	zc.zc_cookie = force;
1913 	zc.zc_guid = hardforce;
1914 	zc.zc_history = (uint64_t)(uintptr_t)log_str;
1915 
1916 	if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) {
1917 		switch (errno) {
1918 		case EXDEV:
1919 			zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
1920 			    "use '-f' to override the following errors:\n"
1921 			    "'%s' has an active shared spare which could be"
1922 			    " used by other pools once '%s' is exported."),
1923 			    zhp->zpool_name, zhp->zpool_name);
1924 			return (zfs_error_fmt(zhp->zpool_hdl, EZFS_ACTIVE_SPARE,
1925 			    dgettext(TEXT_DOMAIN, "cannot export '%s'"),
1926 			    zhp->zpool_name));
1927 		default:
1928 			return (zpool_standard_error_fmt(zhp->zpool_hdl, errno,
1929 			    dgettext(TEXT_DOMAIN, "cannot export '%s'"),
1930 			    zhp->zpool_name));
1931 		}
1932 	}
1933 
1934 	return (0);
1935 }
1936 
1937 int
1938 zpool_export(zpool_handle_t *zhp, boolean_t force, const char *log_str)
1939 {
1940 	return (zpool_export_common(zhp, force, B_FALSE, log_str));
1941 }
1942 
1943 int
1944 zpool_export_force(zpool_handle_t *zhp, const char *log_str)
1945 {
1946 	return (zpool_export_common(zhp, B_TRUE, B_TRUE, log_str));
1947 }
1948 
1949 static void
1950 zpool_rewind_exclaim(libzfs_handle_t *hdl, const char *name, boolean_t dryrun,
1951     nvlist_t *config)
1952 {
1953 	nvlist_t *nv = NULL;
1954 	uint64_t rewindto;
1955 	int64_t loss = -1;
1956 	struct tm t;
1957 	char timestr[128];
1958 
1959 	if (!hdl->libzfs_printerr || config == NULL)
1960 		return;
1961 
1962 	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
1963 	    nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0) {
1964 		return;
1965 	}
1966 
1967 	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1968 		return;
1969 	(void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1970 
1971 	if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1972 	    ctime_r((time_t *)&rewindto, timestr) != NULL) {
1973 		timestr[24] = 0;
1974 		if (dryrun) {
1975 			(void) printf(dgettext(TEXT_DOMAIN,
1976 			    "Would be able to return %s "
1977 			    "to its state as of %s.\n"),
1978 			    name, timestr);
1979 		} else {
1980 			(void) printf(dgettext(TEXT_DOMAIN,
1981 			    "Pool %s returned to its state as of %s.\n"),
1982 			    name, timestr);
1983 		}
1984 		if (loss > 120) {
1985 			(void) printf(dgettext(TEXT_DOMAIN,
1986 			    "%s approximately %lld "),
1987 			    dryrun ? "Would discard" : "Discarded",
1988 			    ((longlong_t)loss + 30) / 60);
1989 			(void) printf(dgettext(TEXT_DOMAIN,
1990 			    "minutes of transactions.\n"));
1991 		} else if (loss > 0) {
1992 			(void) printf(dgettext(TEXT_DOMAIN,
1993 			    "%s approximately %lld "),
1994 			    dryrun ? "Would discard" : "Discarded",
1995 			    (longlong_t)loss);
1996 			(void) printf(dgettext(TEXT_DOMAIN,
1997 			    "seconds of transactions.\n"));
1998 		}
1999 	}
2000 }
2001 
2002 void
2003 zpool_explain_recover(libzfs_handle_t *hdl, const char *name, int reason,
2004     nvlist_t *config, char *buf, size_t size)
2005 {
2006 	nvlist_t *nv = NULL;
2007 	int64_t loss = -1;
2008 	uint64_t edata = UINT64_MAX;
2009 	uint64_t rewindto;
2010 	struct tm t;
2011 	char timestr[128], temp[1024];
2012 
2013 	if (!hdl->libzfs_printerr)
2014 		return;
2015 
2016 	/* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */
2017 	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
2018 	    nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0 ||
2019 	    nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
2020 		goto no_info;
2021 
2022 	(void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
2023 	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_DATA_ERRORS,
2024 	    &edata);
2025 
2026 	(void) snprintf(buf, size, dgettext(TEXT_DOMAIN,
2027 	    "Recovery is possible, but will result in some data loss.\n"));
2028 
2029 	if (localtime_r((time_t *)&rewindto, &t) != NULL &&
2030 	    ctime_r((time_t *)&rewindto, timestr) != NULL) {
2031 		timestr[24] = 0;
2032 		(void) snprintf(temp, 1024, dgettext(TEXT_DOMAIN,
2033 		    "\tReturning the pool to its state as of %s\n"
2034 		    "\tshould correct the problem.  "), timestr);
2035 		(void) strlcat(buf, temp, size);
2036 	} else {
2037 		(void) strlcat(buf, dgettext(TEXT_DOMAIN,
2038 		    "\tReverting the pool to an earlier state "
2039 		    "should correct the problem.\n\t"), size);
2040 	}
2041 
2042 	if (loss > 120) {
2043 		(void) snprintf(temp, 1024, dgettext(TEXT_DOMAIN,
2044 		    "Approximately %lld minutes of data\n"
2045 		    "\tmust be discarded, irreversibly.  "),
2046 		    ((longlong_t)loss + 30) / 60);
2047 		(void) strlcat(buf, temp, size);
2048 	} else if (loss > 0) {
2049 		(void) snprintf(temp, 1024, dgettext(TEXT_DOMAIN,
2050 		    "Approximately %lld seconds of data\n"
2051 		    "\tmust be discarded, irreversibly.  "),
2052 		    (longlong_t)loss);
2053 		(void) strlcat(buf, temp, size);
2054 	}
2055 	if (edata != 0 && edata != UINT64_MAX) {
2056 		if (edata == 1) {
2057 			(void) strlcat(buf, dgettext(TEXT_DOMAIN,
2058 			    "After rewind, at least\n"
2059 			    "\tone persistent user-data error will remain.  "),
2060 			    size);
2061 		} else {
2062 			(void) strlcat(buf, dgettext(TEXT_DOMAIN,
2063 			    "After rewind, several\n"
2064 			    "\tpersistent user-data errors will remain.  "),
2065 			    size);
2066 		}
2067 	}
2068 	(void) snprintf(temp, 1024, dgettext(TEXT_DOMAIN,
2069 	    "Recovery can be attempted\n\tby executing 'zpool %s -F %s'.  "),
2070 	    reason >= 0 ? "clear" : "import", name);
2071 	(void) strlcat(buf, temp, size);
2072 
2073 	(void) strlcat(buf, dgettext(TEXT_DOMAIN,
2074 	    "A scrub of the pool\n"
2075 	    "\tis strongly recommended after recovery.\n"), size);
2076 	return;
2077 
2078 no_info:
2079 	(void) strlcat(buf, dgettext(TEXT_DOMAIN,
2080 	    "Destroy and re-create the pool from\n\ta backup source.\n"), size);
2081 }
2082 
2083 /*
2084  * zpool_import() is a contracted interface. Should be kept the same
2085  * if possible.
2086  *
2087  * Applications should use zpool_import_props() to import a pool with
2088  * new properties value to be set.
2089  */
2090 int
2091 zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
2092     char *altroot)
2093 {
2094 	nvlist_t *props = NULL;
2095 	int ret;
2096 
2097 	if (altroot != NULL) {
2098 		if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) {
2099 			return (zfs_error_fmt(hdl, EZFS_NOMEM,
2100 			    dgettext(TEXT_DOMAIN, "cannot import '%s'"),
2101 			    newname));
2102 		}
2103 
2104 		if (nvlist_add_string(props,
2105 		    zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 ||
2106 		    nvlist_add_string(props,
2107 		    zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) {
2108 			nvlist_free(props);
2109 			return (zfs_error_fmt(hdl, EZFS_NOMEM,
2110 			    dgettext(TEXT_DOMAIN, "cannot import '%s'"),
2111 			    newname));
2112 		}
2113 	}
2114 
2115 	ret = zpool_import_props(hdl, config, newname, props,
2116 	    ZFS_IMPORT_NORMAL);
2117 	nvlist_free(props);
2118 	return (ret);
2119 }
2120 
2121 static void
2122 print_vdev_tree(libzfs_handle_t *hdl, const char *name, nvlist_t *nv,
2123     int indent)
2124 {
2125 	nvlist_t **child;
2126 	uint_t c, children;
2127 	char *vname;
2128 	uint64_t is_log = 0;
2129 
2130 	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG,
2131 	    &is_log);
2132 
2133 	if (name != NULL)
2134 		(void) printf("\t%*s%s%s\n", indent, "", name,
2135 		    is_log ? " [log]" : "");
2136 
2137 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
2138 	    &child, &children) != 0)
2139 		return;
2140 
2141 	for (c = 0; c < children; c++) {
2142 		vname = zpool_vdev_name(hdl, NULL, child[c], VDEV_NAME_TYPE_ID);
2143 		print_vdev_tree(hdl, vname, child[c], indent + 2);
2144 		free(vname);
2145 	}
2146 }
2147 
2148 void
2149 zpool_collect_unsup_feat(nvlist_t *config, char *buf, size_t size)
2150 {
2151 	nvlist_t *nvinfo, *unsup_feat;
2152 	char temp[512];
2153 
2154 	nvinfo = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO);
2155 	unsup_feat = fnvlist_lookup_nvlist(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT);
2156 
2157 	for (nvpair_t *nvp = nvlist_next_nvpair(unsup_feat, NULL);
2158 	    nvp != NULL; nvp = nvlist_next_nvpair(unsup_feat, nvp)) {
2159 		const char *desc = fnvpair_value_string(nvp);
2160 		if (strlen(desc) > 0) {
2161 			(void) snprintf(temp, 512, "\t%s (%s)\n",
2162 			    nvpair_name(nvp), desc);
2163 			(void) strlcat(buf, temp, size);
2164 		} else {
2165 			(void) snprintf(temp, 512, "\t%s\n", nvpair_name(nvp));
2166 			(void) strlcat(buf, temp, size);
2167 		}
2168 	}
2169 }
2170 
2171 /*
2172  * Import the given pool using the known configuration and a list of
2173  * properties to be set. The configuration should have come from
2174  * zpool_find_import(). The 'newname' parameters control whether the pool
2175  * is imported with a different name.
2176  */
2177 int
2178 zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
2179     nvlist_t *props, int flags)
2180 {
2181 	zfs_cmd_t zc = {"\0"};
2182 	zpool_load_policy_t policy;
2183 	nvlist_t *nv = NULL;
2184 	nvlist_t *nvinfo = NULL;
2185 	nvlist_t *missing = NULL;
2186 	const char *thename;
2187 	const char *origname;
2188 	int ret;
2189 	int error = 0;
2190 	char buf[2048];
2191 	char errbuf[ERRBUFLEN];
2192 
2193 	origname = fnvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME);
2194 
2195 	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
2196 	    "cannot import pool '%s'"), origname);
2197 
2198 	if (newname != NULL) {
2199 		if (!zpool_name_valid(hdl, B_FALSE, newname))
2200 			return (zfs_error_fmt(hdl, EZFS_INVALIDNAME,
2201 			    dgettext(TEXT_DOMAIN, "cannot import '%s'"),
2202 			    newname));
2203 		thename = newname;
2204 	} else {
2205 		thename = origname;
2206 	}
2207 
2208 	if (props != NULL) {
2209 		uint64_t version;
2210 		prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
2211 
2212 		version = fnvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION);
2213 
2214 		if ((props = zpool_valid_proplist(hdl, origname,
2215 		    props, version, flags, errbuf)) == NULL)
2216 			return (-1);
2217 		zcmd_write_src_nvlist(hdl, &zc, props);
2218 		nvlist_free(props);
2219 	}
2220 
2221 	(void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name));
2222 
2223 	zc.zc_guid = fnvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID);
2224 
2225 	zcmd_write_conf_nvlist(hdl, &zc, config);
2226 	zcmd_alloc_dst_nvlist(hdl, &zc, zc.zc_nvlist_conf_size * 2);
2227 
2228 	zc.zc_cookie = flags;
2229 	while ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc)) != 0 &&
2230 	    errno == ENOMEM)
2231 		zcmd_expand_dst_nvlist(hdl, &zc);
2232 	if (ret != 0)
2233 		error = errno;
2234 
2235 	(void) zcmd_read_dst_nvlist(hdl, &zc, &nv);
2236 
2237 	zcmd_free_nvlists(&zc);
2238 
2239 	zpool_get_load_policy(config, &policy);
2240 
2241 	if (error) {
2242 		char desc[1024];
2243 		char aux[256];
2244 
2245 		/*
2246 		 * Dry-run failed, but we print out what success
2247 		 * looks like if we found a best txg
2248 		 */
2249 		if (policy.zlp_rewind & ZPOOL_TRY_REWIND) {
2250 			zpool_rewind_exclaim(hdl, newname ? origname : thename,
2251 			    B_TRUE, nv);
2252 			nvlist_free(nv);
2253 			return (-1);
2254 		}
2255 
2256 		if (newname == NULL)
2257 			(void) snprintf(desc, sizeof (desc),
2258 			    dgettext(TEXT_DOMAIN, "cannot import '%s'"),
2259 			    thename);
2260 		else
2261 			(void) snprintf(desc, sizeof (desc),
2262 			    dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"),
2263 			    origname, thename);
2264 
2265 		switch (error) {
2266 		case ENOTSUP:
2267 			if (nv != NULL && nvlist_lookup_nvlist(nv,
2268 			    ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
2269 			    nvlist_exists(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT)) {
2270 				(void) printf(dgettext(TEXT_DOMAIN, "This "
2271 				    "pool uses the following feature(s) not "
2272 				    "supported by this system:\n"));
2273 				memset(buf, 0, 2048);
2274 				zpool_collect_unsup_feat(nv, buf, 2048);
2275 				(void) printf("%s", buf);
2276 				if (nvlist_exists(nvinfo,
2277 				    ZPOOL_CONFIG_CAN_RDONLY)) {
2278 					(void) printf(dgettext(TEXT_DOMAIN,
2279 					    "All unsupported features are only "
2280 					    "required for writing to the pool."
2281 					    "\nThe pool can be imported using "
2282 					    "'-o readonly=on'.\n"));
2283 				}
2284 			}
2285 			/*
2286 			 * Unsupported version.
2287 			 */
2288 			(void) zfs_error(hdl, EZFS_BADVERSION, desc);
2289 			break;
2290 
2291 		case EREMOTEIO:
2292 			if (nv != NULL && nvlist_lookup_nvlist(nv,
2293 			    ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0) {
2294 				const char *hostname = "<unknown>";
2295 				uint64_t hostid = 0;
2296 				mmp_state_t mmp_state;
2297 
2298 				mmp_state = fnvlist_lookup_uint64(nvinfo,
2299 				    ZPOOL_CONFIG_MMP_STATE);
2300 
2301 				if (nvlist_exists(nvinfo,
2302 				    ZPOOL_CONFIG_MMP_HOSTNAME))
2303 					hostname = fnvlist_lookup_string(nvinfo,
2304 					    ZPOOL_CONFIG_MMP_HOSTNAME);
2305 
2306 				if (nvlist_exists(nvinfo,
2307 				    ZPOOL_CONFIG_MMP_HOSTID))
2308 					hostid = fnvlist_lookup_uint64(nvinfo,
2309 					    ZPOOL_CONFIG_MMP_HOSTID);
2310 
2311 				if (mmp_state == MMP_STATE_ACTIVE) {
2312 					(void) snprintf(aux, sizeof (aux),
2313 					    dgettext(TEXT_DOMAIN, "pool is imp"
2314 					    "orted on host '%s' (hostid=%lx).\n"
2315 					    "Export the pool on the other "
2316 					    "system, then run 'zpool import'."),
2317 					    hostname, (unsigned long) hostid);
2318 				} else if (mmp_state == MMP_STATE_NO_HOSTID) {
2319 					(void) snprintf(aux, sizeof (aux),
2320 					    dgettext(TEXT_DOMAIN, "pool has "
2321 					    "the multihost property on and "
2322 					    "the\nsystem's hostid is not set. "
2323 					    "Set a unique system hostid with "
2324 					    "the zgenhostid(8) command.\n"));
2325 				}
2326 
2327 				(void) zfs_error_aux(hdl, "%s", aux);
2328 			}
2329 			(void) zfs_error(hdl, EZFS_ACTIVE_POOL, desc);
2330 			break;
2331 
2332 		case EINVAL:
2333 			(void) zfs_error(hdl, EZFS_INVALCONFIG, desc);
2334 			break;
2335 
2336 		case EROFS:
2337 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2338 			    "one or more devices is read only"));
2339 			(void) zfs_error(hdl, EZFS_BADDEV, desc);
2340 			break;
2341 
2342 		case ENXIO:
2343 			if (nv && nvlist_lookup_nvlist(nv,
2344 			    ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
2345 			    nvlist_lookup_nvlist(nvinfo,
2346 			    ZPOOL_CONFIG_MISSING_DEVICES, &missing) == 0) {
2347 				(void) printf(dgettext(TEXT_DOMAIN,
2348 				    "The devices below are missing or "
2349 				    "corrupted, use '-m' to import the pool "
2350 				    "anyway:\n"));
2351 				print_vdev_tree(hdl, NULL, missing, 2);
2352 				(void) printf("\n");
2353 			}
2354 			(void) zpool_standard_error(hdl, error, desc);
2355 			break;
2356 
2357 		case EEXIST:
2358 			(void) zpool_standard_error(hdl, error, desc);
2359 			break;
2360 
2361 		case EBUSY:
2362 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2363 			    "one or more devices are already in use\n"));
2364 			(void) zfs_error(hdl, EZFS_BADDEV, desc);
2365 			break;
2366 		case ENAMETOOLONG:
2367 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2368 			    "new name of at least one dataset is longer than "
2369 			    "the maximum allowable length"));
2370 			(void) zfs_error(hdl, EZFS_NAMETOOLONG, desc);
2371 			break;
2372 		default:
2373 			(void) zpool_standard_error(hdl, error, desc);
2374 			memset(buf, 0, 2048);
2375 			zpool_explain_recover(hdl,
2376 			    newname ? origname : thename, -error, nv,
2377 			    buf, 2048);
2378 			(void) printf("\t%s", buf);
2379 			break;
2380 		}
2381 
2382 		nvlist_free(nv);
2383 		ret = -1;
2384 	} else {
2385 		zpool_handle_t *zhp;
2386 
2387 		/*
2388 		 * This should never fail, but play it safe anyway.
2389 		 */
2390 		if (zpool_open_silent(hdl, thename, &zhp) != 0)
2391 			ret = -1;
2392 		else if (zhp != NULL)
2393 			zpool_close(zhp);
2394 		if (policy.zlp_rewind &
2395 		    (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
2396 			zpool_rewind_exclaim(hdl, newname ? origname : thename,
2397 			    ((policy.zlp_rewind & ZPOOL_TRY_REWIND) != 0), nv);
2398 		}
2399 		nvlist_free(nv);
2400 	}
2401 
2402 	return (ret);
2403 }
2404 
2405 /*
2406  * Translate vdev names to guids.  If a vdev_path is determined to be
2407  * unsuitable then a vd_errlist is allocated and the vdev path and errno
2408  * are added to it.
2409  */
2410 static int
2411 zpool_translate_vdev_guids(zpool_handle_t *zhp, nvlist_t *vds,
2412     nvlist_t *vdev_guids, nvlist_t *guids_to_paths, nvlist_t **vd_errlist)
2413 {
2414 	nvlist_t *errlist = NULL;
2415 	int error = 0;
2416 
2417 	for (nvpair_t *elem = nvlist_next_nvpair(vds, NULL); elem != NULL;
2418 	    elem = nvlist_next_nvpair(vds, elem)) {
2419 		boolean_t spare, cache;
2420 
2421 		const char *vd_path = nvpair_name(elem);
2422 		nvlist_t *tgt = zpool_find_vdev(zhp, vd_path, &spare, &cache,
2423 		    NULL);
2424 
2425 		if ((tgt == NULL) || cache || spare) {
2426 			if (errlist == NULL) {
2427 				errlist = fnvlist_alloc();
2428 				error = EINVAL;
2429 			}
2430 
2431 			uint64_t err = (tgt == NULL) ? EZFS_NODEVICE :
2432 			    (spare ? EZFS_ISSPARE : EZFS_ISL2CACHE);
2433 			fnvlist_add_int64(errlist, vd_path, err);
2434 			continue;
2435 		}
2436 
2437 		uint64_t guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
2438 		fnvlist_add_uint64(vdev_guids, vd_path, guid);
2439 
2440 		char msg[MAXNAMELEN];
2441 		(void) snprintf(msg, sizeof (msg), "%llu", (u_longlong_t)guid);
2442 		fnvlist_add_string(guids_to_paths, msg, vd_path);
2443 	}
2444 
2445 	if (error != 0) {
2446 		verify(errlist != NULL);
2447 		if (vd_errlist != NULL)
2448 			*vd_errlist = errlist;
2449 		else
2450 			fnvlist_free(errlist);
2451 	}
2452 
2453 	return (error);
2454 }
2455 
2456 static int
2457 xlate_init_err(int err)
2458 {
2459 	switch (err) {
2460 	case ENODEV:
2461 		return (EZFS_NODEVICE);
2462 	case EINVAL:
2463 	case EROFS:
2464 		return (EZFS_BADDEV);
2465 	case EBUSY:
2466 		return (EZFS_INITIALIZING);
2467 	case ESRCH:
2468 		return (EZFS_NO_INITIALIZE);
2469 	}
2470 	return (err);
2471 }
2472 
2473 /*
2474  * Begin, suspend, cancel, or uninit (clear) the initialization (initializing
2475  * of all free blocks) for the given vdevs in the given pool.
2476  */
2477 static int
2478 zpool_initialize_impl(zpool_handle_t *zhp, pool_initialize_func_t cmd_type,
2479     nvlist_t *vds, boolean_t wait)
2480 {
2481 	int err;
2482 
2483 	nvlist_t *vdev_guids = fnvlist_alloc();
2484 	nvlist_t *guids_to_paths = fnvlist_alloc();
2485 	nvlist_t *vd_errlist = NULL;
2486 	nvlist_t *errlist;
2487 	nvpair_t *elem;
2488 
2489 	err = zpool_translate_vdev_guids(zhp, vds, vdev_guids,
2490 	    guids_to_paths, &vd_errlist);
2491 
2492 	if (err != 0) {
2493 		verify(vd_errlist != NULL);
2494 		goto list_errors;
2495 	}
2496 
2497 	err = lzc_initialize(zhp->zpool_name, cmd_type,
2498 	    vdev_guids, &errlist);
2499 
2500 	if (err != 0) {
2501 		if (errlist != NULL && nvlist_lookup_nvlist(errlist,
2502 		    ZPOOL_INITIALIZE_VDEVS, &vd_errlist) == 0) {
2503 			goto list_errors;
2504 		}
2505 
2506 		if (err == EINVAL && cmd_type == POOL_INITIALIZE_UNINIT) {
2507 			zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
2508 			    "uninitialize is not supported by kernel"));
2509 		}
2510 
2511 		(void) zpool_standard_error(zhp->zpool_hdl, err,
2512 		    dgettext(TEXT_DOMAIN, "operation failed"));
2513 		goto out;
2514 	}
2515 
2516 	if (wait) {
2517 		for (elem = nvlist_next_nvpair(vdev_guids, NULL); elem != NULL;
2518 		    elem = nvlist_next_nvpair(vdev_guids, elem)) {
2519 
2520 			uint64_t guid = fnvpair_value_uint64(elem);
2521 
2522 			err = lzc_wait_tag(zhp->zpool_name,
2523 			    ZPOOL_WAIT_INITIALIZE, guid, NULL);
2524 			if (err != 0) {
2525 				(void) zpool_standard_error_fmt(zhp->zpool_hdl,
2526 				    err, dgettext(TEXT_DOMAIN, "error "
2527 				    "waiting for '%s' to initialize"),
2528 				    nvpair_name(elem));
2529 
2530 				goto out;
2531 			}
2532 		}
2533 	}
2534 	goto out;
2535 
2536 list_errors:
2537 	for (elem = nvlist_next_nvpair(vd_errlist, NULL); elem != NULL;
2538 	    elem = nvlist_next_nvpair(vd_errlist, elem)) {
2539 		int64_t vd_error = xlate_init_err(fnvpair_value_int64(elem));
2540 		const char *path;
2541 
2542 		if (nvlist_lookup_string(guids_to_paths, nvpair_name(elem),
2543 		    &path) != 0)
2544 			path = nvpair_name(elem);
2545 
2546 		(void) zfs_error_fmt(zhp->zpool_hdl, vd_error,
2547 		    "cannot initialize '%s'", path);
2548 	}
2549 
2550 out:
2551 	fnvlist_free(vdev_guids);
2552 	fnvlist_free(guids_to_paths);
2553 
2554 	if (vd_errlist != NULL)
2555 		fnvlist_free(vd_errlist);
2556 
2557 	return (err == 0 ? 0 : -1);
2558 }
2559 
2560 int
2561 zpool_initialize(zpool_handle_t *zhp, pool_initialize_func_t cmd_type,
2562     nvlist_t *vds)
2563 {
2564 	return (zpool_initialize_impl(zhp, cmd_type, vds, B_FALSE));
2565 }
2566 
2567 int
2568 zpool_initialize_wait(zpool_handle_t *zhp, pool_initialize_func_t cmd_type,
2569     nvlist_t *vds)
2570 {
2571 	return (zpool_initialize_impl(zhp, cmd_type, vds, B_TRUE));
2572 }
2573 
2574 static int
2575 xlate_trim_err(int err)
2576 {
2577 	switch (err) {
2578 	case ENODEV:
2579 		return (EZFS_NODEVICE);
2580 	case EINVAL:
2581 	case EROFS:
2582 		return (EZFS_BADDEV);
2583 	case EBUSY:
2584 		return (EZFS_TRIMMING);
2585 	case ESRCH:
2586 		return (EZFS_NO_TRIM);
2587 	case EOPNOTSUPP:
2588 		return (EZFS_TRIM_NOTSUP);
2589 	}
2590 	return (err);
2591 }
2592 
2593 static int
2594 zpool_trim_wait(zpool_handle_t *zhp, nvlist_t *vdev_guids)
2595 {
2596 	int err;
2597 	nvpair_t *elem;
2598 
2599 	for (elem = nvlist_next_nvpair(vdev_guids, NULL); elem != NULL;
2600 	    elem = nvlist_next_nvpair(vdev_guids, elem)) {
2601 
2602 		uint64_t guid = fnvpair_value_uint64(elem);
2603 
2604 		err = lzc_wait_tag(zhp->zpool_name,
2605 		    ZPOOL_WAIT_TRIM, guid, NULL);
2606 		if (err != 0) {
2607 			(void) zpool_standard_error_fmt(zhp->zpool_hdl,
2608 			    err, dgettext(TEXT_DOMAIN, "error "
2609 			    "waiting to trim '%s'"), nvpair_name(elem));
2610 
2611 			return (err);
2612 		}
2613 	}
2614 	return (0);
2615 }
2616 
2617 /*
2618  * Check errlist and report any errors, omitting ones which should be
2619  * suppressed. Returns B_TRUE if any errors were reported.
2620  */
2621 static boolean_t
2622 check_trim_errs(zpool_handle_t *zhp, trimflags_t *trim_flags,
2623     nvlist_t *guids_to_paths, nvlist_t *vds, nvlist_t *errlist)
2624 {
2625 	nvpair_t *elem;
2626 	boolean_t reported_errs = B_FALSE;
2627 	int num_vds = 0;
2628 	int num_suppressed_errs = 0;
2629 
2630 	for (elem = nvlist_next_nvpair(vds, NULL);
2631 	    elem != NULL; elem = nvlist_next_nvpair(vds, elem)) {
2632 		num_vds++;
2633 	}
2634 
2635 	for (elem = nvlist_next_nvpair(errlist, NULL);
2636 	    elem != NULL; elem = nvlist_next_nvpair(errlist, elem)) {
2637 		int64_t vd_error = xlate_trim_err(fnvpair_value_int64(elem));
2638 		const char *path;
2639 
2640 		/*
2641 		 * If only the pool was specified, and it was not a secure
2642 		 * trim then suppress warnings for individual vdevs which
2643 		 * do not support trimming.
2644 		 */
2645 		if (vd_error == EZFS_TRIM_NOTSUP &&
2646 		    trim_flags->fullpool &&
2647 		    !trim_flags->secure) {
2648 			num_suppressed_errs++;
2649 			continue;
2650 		}
2651 
2652 		reported_errs = B_TRUE;
2653 		if (nvlist_lookup_string(guids_to_paths, nvpair_name(elem),
2654 		    &path) != 0)
2655 			path = nvpair_name(elem);
2656 
2657 		(void) zfs_error_fmt(zhp->zpool_hdl, vd_error,
2658 		    "cannot trim '%s'", path);
2659 	}
2660 
2661 	if (num_suppressed_errs == num_vds) {
2662 		(void) zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
2663 		    "no devices in pool support trim operations"));
2664 		(void) (zfs_error(zhp->zpool_hdl, EZFS_TRIM_NOTSUP,
2665 		    dgettext(TEXT_DOMAIN, "cannot trim")));
2666 		reported_errs = B_TRUE;
2667 	}
2668 
2669 	return (reported_errs);
2670 }
2671 
2672 /*
2673  * Begin, suspend, or cancel the TRIM (discarding of all free blocks) for
2674  * the given vdevs in the given pool.
2675  */
2676 int
2677 zpool_trim(zpool_handle_t *zhp, pool_trim_func_t cmd_type, nvlist_t *vds,
2678     trimflags_t *trim_flags)
2679 {
2680 	int err;
2681 	int retval = 0;
2682 
2683 	nvlist_t *vdev_guids = fnvlist_alloc();
2684 	nvlist_t *guids_to_paths = fnvlist_alloc();
2685 	nvlist_t *errlist = NULL;
2686 
2687 	err = zpool_translate_vdev_guids(zhp, vds, vdev_guids,
2688 	    guids_to_paths, &errlist);
2689 	if (err != 0) {
2690 		check_trim_errs(zhp, trim_flags, guids_to_paths, vds, errlist);
2691 		retval = -1;
2692 		goto out;
2693 	}
2694 
2695 	err = lzc_trim(zhp->zpool_name, cmd_type, trim_flags->rate,
2696 	    trim_flags->secure, vdev_guids, &errlist);
2697 	if (err != 0) {
2698 		nvlist_t *vd_errlist;
2699 		if (errlist != NULL && nvlist_lookup_nvlist(errlist,
2700 		    ZPOOL_TRIM_VDEVS, &vd_errlist) == 0) {
2701 			if (check_trim_errs(zhp, trim_flags, guids_to_paths,
2702 			    vds, vd_errlist)) {
2703 				retval = -1;
2704 				goto out;
2705 			}
2706 		} else {
2707 			char errbuf[ERRBUFLEN];
2708 
2709 			(void) snprintf(errbuf, sizeof (errbuf),
2710 			    dgettext(TEXT_DOMAIN, "operation failed"));
2711 			zpool_standard_error(zhp->zpool_hdl, err, errbuf);
2712 			retval = -1;
2713 			goto out;
2714 		}
2715 	}
2716 
2717 
2718 	if (trim_flags->wait)
2719 		retval = zpool_trim_wait(zhp, vdev_guids);
2720 
2721 out:
2722 	if (errlist != NULL)
2723 		fnvlist_free(errlist);
2724 	fnvlist_free(vdev_guids);
2725 	fnvlist_free(guids_to_paths);
2726 	return (retval);
2727 }
2728 
2729 /*
2730  * Scan the pool.
2731  */
2732 int
2733 zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func, pool_scrub_cmd_t cmd)
2734 {
2735 	char errbuf[ERRBUFLEN];
2736 	int err;
2737 	libzfs_handle_t *hdl = zhp->zpool_hdl;
2738 
2739 	nvlist_t *args = fnvlist_alloc();
2740 	fnvlist_add_uint64(args, "scan_type", (uint64_t)func);
2741 	fnvlist_add_uint64(args, "scan_command", (uint64_t)cmd);
2742 
2743 	err = lzc_scrub(ZFS_IOC_POOL_SCRUB, zhp->zpool_name, args, NULL);
2744 	fnvlist_free(args);
2745 
2746 	if (err == 0) {
2747 		return (0);
2748 	} else if (err == ZFS_ERR_IOC_CMD_UNAVAIL) {
2749 		zfs_cmd_t zc = {"\0"};
2750 		(void) strlcpy(zc.zc_name, zhp->zpool_name,
2751 		    sizeof (zc.zc_name));
2752 		zc.zc_cookie = func;
2753 		zc.zc_flags = cmd;
2754 
2755 		if (zfs_ioctl(hdl, ZFS_IOC_POOL_SCAN, &zc) == 0)
2756 			return (0);
2757 	}
2758 
2759 	/*
2760 	 * An ECANCELED on a scrub means one of the following:
2761 	 * 1. we resumed a paused scrub.
2762 	 * 2. we resumed a paused error scrub.
2763 	 * 3. Error scrub is not run because of no error log.
2764 	 */
2765 	if (err == ECANCELED && (func == POOL_SCAN_SCRUB ||
2766 	    func == POOL_SCAN_ERRORSCRUB) && cmd == POOL_SCRUB_NORMAL)
2767 		return (0);
2768 	/*
2769 	 * The following cases have been handled here:
2770 	 * 1. Paused a scrub/error scrub if there is none in progress.
2771 	 */
2772 	if (err == ENOENT && func != POOL_SCAN_NONE && cmd ==
2773 	    POOL_SCRUB_PAUSE) {
2774 		return (0);
2775 	}
2776 
2777 	ASSERT3U(func, >=, POOL_SCAN_NONE);
2778 	ASSERT3U(func, <, POOL_SCAN_FUNCS);
2779 
2780 	if (func == POOL_SCAN_SCRUB || func == POOL_SCAN_ERRORSCRUB) {
2781 		if (cmd == POOL_SCRUB_PAUSE) {
2782 			(void) snprintf(errbuf, sizeof (errbuf),
2783 			    dgettext(TEXT_DOMAIN, "cannot pause scrubbing %s"),
2784 			    zhp->zpool_name);
2785 		} else {
2786 			assert(cmd == POOL_SCRUB_NORMAL);
2787 			(void) snprintf(errbuf, sizeof (errbuf),
2788 			    dgettext(TEXT_DOMAIN, "cannot scrub %s"),
2789 			    zhp->zpool_name);
2790 		}
2791 	} else if (func == POOL_SCAN_RESILVER) {
2792 		assert(cmd == POOL_SCRUB_NORMAL);
2793 		(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
2794 		    "cannot restart resilver on %s"), zhp->zpool_name);
2795 	} else if (func == POOL_SCAN_NONE) {
2796 		(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
2797 		    "cannot cancel scrubbing %s"), zhp->zpool_name);
2798 	} else {
2799 		assert(!"unexpected result");
2800 	}
2801 
2802 	/*
2803 	 * With EBUSY, six cases are possible:
2804 	 *
2805 	 * Current state		Requested
2806 	 * 1. Normal Scrub Running	Normal Scrub or Error Scrub
2807 	 * 2. Normal Scrub Paused	Error Scrub
2808 	 * 3. Normal Scrub Paused 	Pause Normal Scrub
2809 	 * 4. Error Scrub Running	Normal Scrub or Error Scrub
2810 	 * 5. Error Scrub Paused	Pause Error Scrub
2811 	 * 6. Resilvering		Anything else
2812 	 */
2813 	if (err == EBUSY) {
2814 		nvlist_t *nvroot;
2815 		pool_scan_stat_t *ps = NULL;
2816 		uint_t psc;
2817 
2818 		nvroot = fnvlist_lookup_nvlist(zhp->zpool_config,
2819 		    ZPOOL_CONFIG_VDEV_TREE);
2820 		(void) nvlist_lookup_uint64_array(nvroot,
2821 		    ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc);
2822 		if (ps && ps->pss_func == POOL_SCAN_SCRUB &&
2823 		    ps->pss_state == DSS_SCANNING) {
2824 			if (ps->pss_pass_scrub_pause == 0) {
2825 				/* handles case 1 */
2826 				assert(cmd == POOL_SCRUB_NORMAL);
2827 				return (zfs_error(hdl, EZFS_SCRUBBING,
2828 				    errbuf));
2829 			} else {
2830 				if (func == POOL_SCAN_ERRORSCRUB) {
2831 					/* handles case 2 */
2832 					ASSERT3U(cmd, ==, POOL_SCRUB_NORMAL);
2833 					return (zfs_error(hdl,
2834 					    EZFS_SCRUB_PAUSED_TO_CANCEL,
2835 					    errbuf));
2836 				} else {
2837 					/* handles case 3 */
2838 					ASSERT3U(func, ==, POOL_SCAN_SCRUB);
2839 					ASSERT3U(cmd, ==, POOL_SCRUB_PAUSE);
2840 					return (zfs_error(hdl,
2841 					    EZFS_SCRUB_PAUSED, errbuf));
2842 				}
2843 			}
2844 		} else if (ps &&
2845 		    ps->pss_error_scrub_func == POOL_SCAN_ERRORSCRUB &&
2846 		    ps->pss_error_scrub_state == DSS_ERRORSCRUBBING) {
2847 			if (ps->pss_pass_error_scrub_pause == 0) {
2848 				/* handles case 4 */
2849 				ASSERT3U(cmd, ==, POOL_SCRUB_NORMAL);
2850 				return (zfs_error(hdl, EZFS_ERRORSCRUBBING,
2851 				    errbuf));
2852 			} else {
2853 				/* handles case 5 */
2854 				ASSERT3U(func, ==, POOL_SCAN_ERRORSCRUB);
2855 				ASSERT3U(cmd, ==, POOL_SCRUB_PAUSE);
2856 				return (zfs_error(hdl, EZFS_ERRORSCRUB_PAUSED,
2857 				    errbuf));
2858 			}
2859 		} else {
2860 			/* handles case 6 */
2861 			return (zfs_error(hdl, EZFS_RESILVERING, errbuf));
2862 		}
2863 	} else if (err == ENOENT) {
2864 		return (zfs_error(hdl, EZFS_NO_SCRUB, errbuf));
2865 	} else if (err == ENOTSUP && func == POOL_SCAN_RESILVER) {
2866 		return (zfs_error(hdl, EZFS_NO_RESILVER_DEFER, errbuf));
2867 	} else {
2868 		return (zpool_standard_error(hdl, err, errbuf));
2869 	}
2870 }
2871 
2872 /*
2873  * Find a vdev that matches the search criteria specified. We use the
2874  * the nvpair name to determine how we should look for the device.
2875  * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
2876  * spare; but FALSE if its an INUSE spare.
2877  *
2878  * If 'return_parent' is set, then return the *parent* of the vdev you're
2879  * searching for rather than the vdev itself.
2880  */
2881 static nvlist_t *
2882 vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare,
2883     boolean_t *l2cache, boolean_t *log, boolean_t return_parent)
2884 {
2885 	uint_t c, children;
2886 	nvlist_t **child;
2887 	nvlist_t *ret;
2888 	uint64_t is_log;
2889 	const char *srchkey;
2890 	nvpair_t *pair = nvlist_next_nvpair(search, NULL);
2891 	const char *tmp = NULL;
2892 	boolean_t is_root;
2893 
2894 	/* Nothing to look for */
2895 	if (search == NULL || pair == NULL)
2896 		return (NULL);
2897 
2898 	/* Obtain the key we will use to search */
2899 	srchkey = nvpair_name(pair);
2900 
2901 	nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &tmp);
2902 	if (strcmp(tmp, "root") == 0)
2903 		is_root = B_TRUE;
2904 	else
2905 		is_root = B_FALSE;
2906 
2907 	switch (nvpair_type(pair)) {
2908 	case DATA_TYPE_UINT64:
2909 		if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) {
2910 			uint64_t srchval = fnvpair_value_uint64(pair);
2911 			uint64_t theguid = fnvlist_lookup_uint64(nv,
2912 			    ZPOOL_CONFIG_GUID);
2913 			if (theguid == srchval)
2914 				return (nv);
2915 		}
2916 		break;
2917 
2918 	case DATA_TYPE_STRING: {
2919 		const char *srchval, *val;
2920 
2921 		srchval = fnvpair_value_string(pair);
2922 		if (nvlist_lookup_string(nv, srchkey, &val) != 0)
2923 			break;
2924 
2925 		/*
2926 		 * Search for the requested value. Special cases:
2927 		 *
2928 		 * - ZPOOL_CONFIG_PATH for whole disk entries.  These end in
2929 		 *   "-part1", or "p1".  The suffix is hidden from the user,
2930 		 *   but included in the string, so this matches around it.
2931 		 * - ZPOOL_CONFIG_PATH for short names zfs_strcmp_shortname()
2932 		 *   is used to check all possible expanded paths.
2933 		 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE).
2934 		 *
2935 		 * Otherwise, all other searches are simple string compares.
2936 		 */
2937 		if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0) {
2938 			uint64_t wholedisk = 0;
2939 
2940 			(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
2941 			    &wholedisk);
2942 			if (zfs_strcmp_pathname(srchval, val, wholedisk) == 0)
2943 				return (nv);
2944 
2945 		} else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0) {
2946 			char *type, *idx, *end, *p;
2947 			uint64_t id, vdev_id;
2948 
2949 			/*
2950 			 * Determine our vdev type, keeping in mind
2951 			 * that the srchval is composed of a type and
2952 			 * vdev id pair (i.e. mirror-4).
2953 			 */
2954 			if ((type = strdup(srchval)) == NULL)
2955 				return (NULL);
2956 
2957 			if ((p = strrchr(type, '-')) == NULL) {
2958 				free(type);
2959 				break;
2960 			}
2961 			idx = p + 1;
2962 			*p = '\0';
2963 
2964 			/*
2965 			 * draid names are presented like: draid2:4d:6c:0s
2966 			 * We match them up to the first ':' so we can still
2967 			 * do the parity check below, but the other params
2968 			 * are ignored.
2969 			 */
2970 			if ((p = strchr(type, ':')) != NULL) {
2971 				if (strncmp(type, VDEV_TYPE_DRAID,
2972 				    strlen(VDEV_TYPE_DRAID)) == 0)
2973 					*p = '\0';
2974 			}
2975 
2976 			/*
2977 			 * If the types don't match then keep looking.
2978 			 */
2979 			if (strncmp(val, type, strlen(val)) != 0) {
2980 				free(type);
2981 				break;
2982 			}
2983 
2984 			verify(zpool_vdev_is_interior(type));
2985 
2986 			id = fnvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID);
2987 			errno = 0;
2988 			vdev_id = strtoull(idx, &end, 10);
2989 
2990 			/*
2991 			 * If we are looking for a raidz and a parity is
2992 			 * specified, make sure it matches.
2993 			 */
2994 			int rzlen = strlen(VDEV_TYPE_RAIDZ);
2995 			assert(rzlen == strlen(VDEV_TYPE_DRAID));
2996 			int typlen = strlen(type);
2997 			if ((strncmp(type, VDEV_TYPE_RAIDZ, rzlen) == 0 ||
2998 			    strncmp(type, VDEV_TYPE_DRAID, rzlen) == 0) &&
2999 			    typlen != rzlen) {
3000 				uint64_t vdev_parity;
3001 				int parity = *(type + rzlen) - '0';
3002 
3003 				if (parity <= 0 || parity > 3 ||
3004 				    (typlen - rzlen) != 1) {
3005 					/*
3006 					 * Nonsense parity specified, can
3007 					 * never match
3008 					 */
3009 					free(type);
3010 					return (NULL);
3011 				}
3012 				vdev_parity = fnvlist_lookup_uint64(nv,
3013 				    ZPOOL_CONFIG_NPARITY);
3014 				if ((int)vdev_parity != parity) {
3015 					free(type);
3016 					break;
3017 				}
3018 			}
3019 
3020 			free(type);
3021 			if (errno != 0)
3022 				return (NULL);
3023 
3024 			/*
3025 			 * Now verify that we have the correct vdev id.
3026 			 */
3027 			if (vdev_id == id)
3028 				return (nv);
3029 		}
3030 
3031 		/*
3032 		 * Common case
3033 		 */
3034 		if (strcmp(srchval, val) == 0)
3035 			return (nv);
3036 		break;
3037 	}
3038 
3039 	default:
3040 		break;
3041 	}
3042 
3043 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
3044 	    &child, &children) != 0)
3045 		return (NULL);
3046 
3047 	for (c = 0; c < children; c++) {
3048 		if ((ret = vdev_to_nvlist_iter(child[c], search,
3049 		    avail_spare, l2cache, NULL, return_parent)) != NULL) {
3050 			/*
3051 			 * The 'is_log' value is only set for the toplevel
3052 			 * vdev, not the leaf vdevs.  So we always lookup the
3053 			 * log device from the root of the vdev tree (where
3054 			 * 'log' is non-NULL).
3055 			 */
3056 			if (log != NULL &&
3057 			    nvlist_lookup_uint64(child[c],
3058 			    ZPOOL_CONFIG_IS_LOG, &is_log) == 0 &&
3059 			    is_log) {
3060 				*log = B_TRUE;
3061 			}
3062 			return (ret && return_parent && !is_root ? nv : ret);
3063 		}
3064 	}
3065 
3066 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
3067 	    &child, &children) == 0) {
3068 		for (c = 0; c < children; c++) {
3069 			if ((ret = vdev_to_nvlist_iter(child[c], search,
3070 			    avail_spare, l2cache, NULL, return_parent))
3071 			    != NULL) {
3072 				*avail_spare = B_TRUE;
3073 				return (ret && return_parent &&
3074 				    !is_root ? nv : ret);
3075 			}
3076 		}
3077 	}
3078 
3079 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
3080 	    &child, &children) == 0) {
3081 		for (c = 0; c < children; c++) {
3082 			if ((ret = vdev_to_nvlist_iter(child[c], search,
3083 			    avail_spare, l2cache, NULL, return_parent))
3084 			    != NULL) {
3085 				*l2cache = B_TRUE;
3086 				return (ret && return_parent &&
3087 				    !is_root ? nv : ret);
3088 			}
3089 		}
3090 	}
3091 
3092 	return (NULL);
3093 }
3094 
3095 /*
3096  * Given a physical path or guid, find the associated vdev.
3097  */
3098 nvlist_t *
3099 zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath,
3100     boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log)
3101 {
3102 	nvlist_t *search, *nvroot, *ret;
3103 	uint64_t guid;
3104 	char *end;
3105 
3106 	search = fnvlist_alloc();
3107 
3108 	guid = strtoull(ppath, &end, 0);
3109 	if (guid != 0 && *end == '\0') {
3110 		fnvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid);
3111 	} else {
3112 		fnvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath);
3113 	}
3114 
3115 	nvroot = fnvlist_lookup_nvlist(zhp->zpool_config,
3116 	    ZPOOL_CONFIG_VDEV_TREE);
3117 
3118 	*avail_spare = B_FALSE;
3119 	*l2cache = B_FALSE;
3120 	if (log != NULL)
3121 		*log = B_FALSE;
3122 	ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log,
3123 	    B_FALSE);
3124 	fnvlist_free(search);
3125 
3126 	return (ret);
3127 }
3128 
3129 /*
3130  * Determine if we have an "interior" top-level vdev (i.e mirror/raidz).
3131  */
3132 static boolean_t
3133 zpool_vdev_is_interior(const char *name)
3134 {
3135 	if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 ||
3136 	    strncmp(name, VDEV_TYPE_SPARE, strlen(VDEV_TYPE_SPARE)) == 0 ||
3137 	    strncmp(name,
3138 	    VDEV_TYPE_REPLACING, strlen(VDEV_TYPE_REPLACING)) == 0 ||
3139 	    strncmp(name, VDEV_TYPE_ROOT, strlen(VDEV_TYPE_ROOT)) == 0 ||
3140 	    strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0)
3141 		return (B_TRUE);
3142 
3143 	if (strncmp(name, VDEV_TYPE_DRAID, strlen(VDEV_TYPE_DRAID)) == 0 &&
3144 	    !zpool_is_draid_spare(name))
3145 		return (B_TRUE);
3146 
3147 	return (B_FALSE);
3148 }
3149 
3150 /*
3151  * Lookup the nvlist for a given vdev or vdev's parent (depending on
3152  * if 'return_parent' is set).
3153  */
3154 static nvlist_t *
3155 __zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare,
3156     boolean_t *l2cache, boolean_t *log, boolean_t return_parent)
3157 {
3158 	char *end;
3159 	nvlist_t *nvroot, *search, *ret;
3160 	uint64_t guid;
3161 	boolean_t __avail_spare, __l2cache, __log;
3162 
3163 	search = fnvlist_alloc();
3164 
3165 	guid = strtoull(path, &end, 0);
3166 	if (guid != 0 && *end == '\0') {
3167 		fnvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid);
3168 	} else if (zpool_vdev_is_interior(path)) {
3169 		fnvlist_add_string(search, ZPOOL_CONFIG_TYPE, path);
3170 	} else {
3171 		fnvlist_add_string(search, ZPOOL_CONFIG_PATH, path);
3172 	}
3173 
3174 	nvroot = fnvlist_lookup_nvlist(zhp->zpool_config,
3175 	    ZPOOL_CONFIG_VDEV_TREE);
3176 
3177 	/*
3178 	 * User can pass NULL for avail_spare, l2cache, and log, but
3179 	 * we still need to provide variables to vdev_to_nvlist_iter(), so
3180 	 * just point them to junk variables here.
3181 	 */
3182 	if (!avail_spare)
3183 		avail_spare = &__avail_spare;
3184 	if (!l2cache)
3185 		l2cache = &__l2cache;
3186 	if (!log)
3187 		log = &__log;
3188 
3189 	*avail_spare = B_FALSE;
3190 	*l2cache = B_FALSE;
3191 	if (log != NULL)
3192 		*log = B_FALSE;
3193 	ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log,
3194 	    return_parent);
3195 	fnvlist_free(search);
3196 
3197 	return (ret);
3198 }
3199 
3200 nvlist_t *
3201 zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare,
3202     boolean_t *l2cache, boolean_t *log)
3203 {
3204 	return (__zpool_find_vdev(zhp, path, avail_spare, l2cache, log,
3205 	    B_FALSE));
3206 }
3207 
3208 /* Given a vdev path, return its parent's nvlist */
3209 nvlist_t *
3210 zpool_find_parent_vdev(zpool_handle_t *zhp, const char *path,
3211     boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log)
3212 {
3213 	return (__zpool_find_vdev(zhp, path, avail_spare, l2cache, log,
3214 	    B_TRUE));
3215 }
3216 
3217 /*
3218  * Convert a vdev path to a GUID.  Returns GUID or 0 on error.
3219  *
3220  * If is_spare, is_l2cache, or is_log is non-NULL, then store within it
3221  * if the VDEV is a spare, l2cache, or log device.  If they're NULL then
3222  * ignore them.
3223  */
3224 static uint64_t
3225 zpool_vdev_path_to_guid_impl(zpool_handle_t *zhp, const char *path,
3226     boolean_t *is_spare, boolean_t *is_l2cache, boolean_t *is_log)
3227 {
3228 	boolean_t spare = B_FALSE, l2cache = B_FALSE, log = B_FALSE;
3229 	nvlist_t *tgt;
3230 
3231 	if ((tgt = zpool_find_vdev(zhp, path, &spare, &l2cache,
3232 	    &log)) == NULL)
3233 		return (0);
3234 
3235 	if (is_spare != NULL)
3236 		*is_spare = spare;
3237 	if (is_l2cache != NULL)
3238 		*is_l2cache = l2cache;
3239 	if (is_log != NULL)
3240 		*is_log = log;
3241 
3242 	return (fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID));
3243 }
3244 
3245 /* Convert a vdev path to a GUID.  Returns GUID or 0 on error. */
3246 uint64_t
3247 zpool_vdev_path_to_guid(zpool_handle_t *zhp, const char *path)
3248 {
3249 	return (zpool_vdev_path_to_guid_impl(zhp, path, NULL, NULL, NULL));
3250 }
3251 
3252 /*
3253  * Bring the specified vdev online.   The 'flags' parameter is a set of the
3254  * ZFS_ONLINE_* flags.
3255  */
3256 int
3257 zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags,
3258     vdev_state_t *newstate)
3259 {
3260 	zfs_cmd_t zc = {"\0"};
3261 	char errbuf[ERRBUFLEN];
3262 	nvlist_t *tgt;
3263 	boolean_t avail_spare, l2cache, islog;
3264 	libzfs_handle_t *hdl = zhp->zpool_hdl;
3265 
3266 	if (flags & ZFS_ONLINE_EXPAND) {
3267 		(void) snprintf(errbuf, sizeof (errbuf),
3268 		    dgettext(TEXT_DOMAIN, "cannot expand %s"), path);
3269 	} else {
3270 		(void) snprintf(errbuf, sizeof (errbuf),
3271 		    dgettext(TEXT_DOMAIN, "cannot online %s"), path);
3272 	}
3273 
3274 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3275 	if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
3276 	    &islog)) == NULL)
3277 		return (zfs_error(hdl, EZFS_NODEVICE, errbuf));
3278 
3279 	zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
3280 
3281 	if (!(flags & ZFS_ONLINE_SPARE) && avail_spare)
3282 		return (zfs_error(hdl, EZFS_ISSPARE, errbuf));
3283 
3284 #ifndef __FreeBSD__
3285 	const char *pathname;
3286 	if ((flags & ZFS_ONLINE_EXPAND ||
3287 	    zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) &&
3288 	    nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH, &pathname) == 0) {
3289 		uint64_t wholedisk = 0;
3290 
3291 		(void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK,
3292 		    &wholedisk);
3293 
3294 		/*
3295 		 * XXX - L2ARC 1.0 devices can't support expansion.
3296 		 */
3297 		if (l2cache) {
3298 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3299 			    "cannot expand cache devices"));
3300 			return (zfs_error(hdl, EZFS_VDEVNOTSUP, errbuf));
3301 		}
3302 
3303 		if (wholedisk) {
3304 			const char *fullpath = path;
3305 			char buf[MAXPATHLEN];
3306 			int error;
3307 
3308 			if (path[0] != '/') {
3309 				error = zfs_resolve_shortname(path, buf,
3310 				    sizeof (buf));
3311 				if (error != 0)
3312 					return (zfs_error(hdl, EZFS_NODEVICE,
3313 					    errbuf));
3314 
3315 				fullpath = buf;
3316 			}
3317 
3318 			error = zpool_relabel_disk(hdl, fullpath, errbuf);
3319 			if (error != 0)
3320 				return (error);
3321 		}
3322 	}
3323 #endif
3324 
3325 	zc.zc_cookie = VDEV_STATE_ONLINE;
3326 	zc.zc_obj = flags;
3327 
3328 	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) {
3329 		if (errno == EINVAL) {
3330 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split "
3331 			    "from this pool into a new one.  Use '%s' "
3332 			    "instead"), "zpool detach");
3333 			return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, errbuf));
3334 		}
3335 		return (zpool_standard_error(hdl, errno, errbuf));
3336 	}
3337 
3338 	*newstate = zc.zc_cookie;
3339 	return (0);
3340 }
3341 
3342 /*
3343  * Take the specified vdev offline
3344  */
3345 int
3346 zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp)
3347 {
3348 	zfs_cmd_t zc = {"\0"};
3349 	char errbuf[ERRBUFLEN];
3350 	nvlist_t *tgt;
3351 	boolean_t avail_spare, l2cache;
3352 	libzfs_handle_t *hdl = zhp->zpool_hdl;
3353 
3354 	(void) snprintf(errbuf, sizeof (errbuf),
3355 	    dgettext(TEXT_DOMAIN, "cannot offline %s"), path);
3356 
3357 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3358 	if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
3359 	    NULL)) == NULL)
3360 		return (zfs_error(hdl, EZFS_NODEVICE, errbuf));
3361 
3362 	zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
3363 
3364 	if (avail_spare)
3365 		return (zfs_error(hdl, EZFS_ISSPARE, errbuf));
3366 
3367 	zc.zc_cookie = VDEV_STATE_OFFLINE;
3368 	zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0;
3369 
3370 	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
3371 		return (0);
3372 
3373 	switch (errno) {
3374 	case EBUSY:
3375 
3376 		/*
3377 		 * There are no other replicas of this device.
3378 		 */
3379 		return (zfs_error(hdl, EZFS_NOREPLICAS, errbuf));
3380 
3381 	case EEXIST:
3382 		/*
3383 		 * The log device has unplayed logs
3384 		 */
3385 		return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, errbuf));
3386 
3387 	default:
3388 		return (zpool_standard_error(hdl, errno, errbuf));
3389 	}
3390 }
3391 
3392 /*
3393  * Remove the specified vdev asynchronously from the configuration, so
3394  * that it may come ONLINE if reinserted. This is called from zed on
3395  * Udev remove event.
3396  * Note: We also have a similar function zpool_vdev_remove() that
3397  * removes the vdev from the pool.
3398  */
3399 int
3400 zpool_vdev_remove_wanted(zpool_handle_t *zhp, const char *path)
3401 {
3402 	zfs_cmd_t zc = {"\0"};
3403 	char errbuf[ERRBUFLEN];
3404 	nvlist_t *tgt;
3405 	boolean_t avail_spare, l2cache;
3406 	libzfs_handle_t *hdl = zhp->zpool_hdl;
3407 
3408 	(void) snprintf(errbuf, sizeof (errbuf),
3409 	    dgettext(TEXT_DOMAIN, "cannot remove %s"), path);
3410 
3411 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3412 	if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
3413 	    NULL)) == NULL)
3414 		return (zfs_error(hdl, EZFS_NODEVICE, errbuf));
3415 
3416 	zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
3417 
3418 	zc.zc_cookie = VDEV_STATE_REMOVED;
3419 
3420 	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
3421 		return (0);
3422 
3423 	return (zpool_standard_error(hdl, errno, errbuf));
3424 }
3425 
3426 /*
3427  * Mark the given vdev faulted.
3428  */
3429 int
3430 zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
3431 {
3432 	zfs_cmd_t zc = {"\0"};
3433 	char errbuf[ERRBUFLEN];
3434 	libzfs_handle_t *hdl = zhp->zpool_hdl;
3435 
3436 	(void) snprintf(errbuf, sizeof (errbuf),
3437 	    dgettext(TEXT_DOMAIN, "cannot fault %llu"), (u_longlong_t)guid);
3438 
3439 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3440 	zc.zc_guid = guid;
3441 	zc.zc_cookie = VDEV_STATE_FAULTED;
3442 	zc.zc_obj = aux;
3443 
3444 	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
3445 		return (0);
3446 
3447 	switch (errno) {
3448 	case EBUSY:
3449 
3450 		/*
3451 		 * There are no other replicas of this device.
3452 		 */
3453 		return (zfs_error(hdl, EZFS_NOREPLICAS, errbuf));
3454 
3455 	default:
3456 		return (zpool_standard_error(hdl, errno, errbuf));
3457 	}
3458 
3459 }
3460 
3461 /*
3462  * Generic set vdev state function
3463  */
3464 static int
3465 zpool_vdev_set_state(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux,
3466     vdev_state_t state)
3467 {
3468 	zfs_cmd_t zc = {"\0"};
3469 	char errbuf[ERRBUFLEN];
3470 	libzfs_handle_t *hdl = zhp->zpool_hdl;
3471 
3472 	(void) snprintf(errbuf, sizeof (errbuf),
3473 	    dgettext(TEXT_DOMAIN, "cannot set %s %llu"),
3474 	    zpool_state_to_name(state, aux), (u_longlong_t)guid);
3475 
3476 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3477 	zc.zc_guid = guid;
3478 	zc.zc_cookie = state;
3479 	zc.zc_obj = aux;
3480 
3481 	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
3482 		return (0);
3483 
3484 	return (zpool_standard_error(hdl, errno, errbuf));
3485 }
3486 
3487 /*
3488  * Mark the given vdev degraded.
3489  */
3490 int
3491 zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
3492 {
3493 	return (zpool_vdev_set_state(zhp, guid, aux, VDEV_STATE_DEGRADED));
3494 }
3495 
3496 /*
3497  * Mark the given vdev as in a removed state (as if the device does not exist).
3498  *
3499  * This is different than zpool_vdev_remove() which does a removal of a device
3500  * from the pool (but the device does exist).
3501  */
3502 int
3503 zpool_vdev_set_removed_state(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
3504 {
3505 	return (zpool_vdev_set_state(zhp, guid, aux, VDEV_STATE_REMOVED));
3506 }
3507 
3508 /*
3509  * Returns TRUE if the given nvlist is a vdev that was originally swapped in as
3510  * a hot spare.
3511  */
3512 static boolean_t
3513 is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which)
3514 {
3515 	nvlist_t **child;
3516 	uint_t c, children;
3517 
3518 	if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child,
3519 	    &children) == 0) {
3520 		const char *type = fnvlist_lookup_string(search,
3521 		    ZPOOL_CONFIG_TYPE);
3522 		if ((strcmp(type, VDEV_TYPE_SPARE) == 0 ||
3523 		    strcmp(type, VDEV_TYPE_DRAID_SPARE) == 0) &&
3524 		    children == 2 && child[which] == tgt)
3525 			return (B_TRUE);
3526 
3527 		for (c = 0; c < children; c++)
3528 			if (is_replacing_spare(child[c], tgt, which))
3529 				return (B_TRUE);
3530 	}
3531 
3532 	return (B_FALSE);
3533 }
3534 
3535 /*
3536  * Attach new_disk (fully described by nvroot) to old_disk.
3537  * If 'replacing' is specified, the new disk will replace the old one.
3538  */
3539 int
3540 zpool_vdev_attach(zpool_handle_t *zhp, const char *old_disk,
3541     const char *new_disk, nvlist_t *nvroot, int replacing, boolean_t rebuild)
3542 {
3543 	zfs_cmd_t zc = {"\0"};
3544 	char errbuf[ERRBUFLEN];
3545 	int ret;
3546 	nvlist_t *tgt;
3547 	boolean_t avail_spare, l2cache, islog;
3548 	uint64_t val;
3549 	char *newname;
3550 	const char *type;
3551 	nvlist_t **child;
3552 	uint_t children;
3553 	nvlist_t *config_root;
3554 	libzfs_handle_t *hdl = zhp->zpool_hdl;
3555 
3556 	if (replacing)
3557 		(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
3558 		    "cannot replace %s with %s"), old_disk, new_disk);
3559 	else
3560 		(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
3561 		    "cannot attach %s to %s"), new_disk, old_disk);
3562 
3563 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3564 	if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache,
3565 	    &islog)) == NULL)
3566 		return (zfs_error(hdl, EZFS_NODEVICE, errbuf));
3567 
3568 	if (avail_spare)
3569 		return (zfs_error(hdl, EZFS_ISSPARE, errbuf));
3570 
3571 	if (l2cache)
3572 		return (zfs_error(hdl, EZFS_ISL2CACHE, errbuf));
3573 
3574 	zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
3575 	zc.zc_cookie = replacing;
3576 	zc.zc_simple = rebuild;
3577 
3578 	if (rebuild &&
3579 	    zfeature_lookup_guid("org.openzfs:device_rebuild", NULL) != 0) {
3580 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3581 		    "the loaded zfs module doesn't support device rebuilds"));
3582 		return (zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf));
3583 	}
3584 
3585 	type = fnvlist_lookup_string(tgt, ZPOOL_CONFIG_TYPE);
3586 	if (strcmp(type, VDEV_TYPE_RAIDZ) == 0 &&
3587 	    zfeature_lookup_guid("org.openzfs:raidz_expansion", NULL) != 0) {
3588 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3589 		    "the loaded zfs module doesn't support raidz expansion"));
3590 		return (zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf));
3591 	}
3592 
3593 	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
3594 	    &child, &children) != 0 || children != 1) {
3595 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3596 		    "new device must be a single disk"));
3597 		return (zfs_error(hdl, EZFS_INVALCONFIG, errbuf));
3598 	}
3599 
3600 	config_root = fnvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
3601 	    ZPOOL_CONFIG_VDEV_TREE);
3602 
3603 	if ((newname = zpool_vdev_name(NULL, NULL, child[0], 0)) == NULL)
3604 		return (-1);
3605 
3606 	/*
3607 	 * If the target is a hot spare that has been swapped in, we can only
3608 	 * replace it with another hot spare.
3609 	 */
3610 	if (replacing &&
3611 	    nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 &&
3612 	    (zpool_find_vdev(zhp, newname, &avail_spare, &l2cache,
3613 	    NULL) == NULL || !avail_spare) &&
3614 	    is_replacing_spare(config_root, tgt, 1)) {
3615 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3616 		    "can only be replaced by another hot spare"));
3617 		free(newname);
3618 		return (zfs_error(hdl, EZFS_BADTARGET, errbuf));
3619 	}
3620 
3621 	free(newname);
3622 
3623 	zcmd_write_conf_nvlist(hdl, &zc, nvroot);
3624 
3625 	ret = zfs_ioctl(hdl, ZFS_IOC_VDEV_ATTACH, &zc);
3626 
3627 	zcmd_free_nvlists(&zc);
3628 
3629 	if (ret == 0)
3630 		return (0);
3631 
3632 	switch (errno) {
3633 	case ENOTSUP:
3634 		/*
3635 		 * Can't attach to or replace this type of vdev.
3636 		 */
3637 		if (replacing) {
3638 			uint64_t version = zpool_get_prop_int(zhp,
3639 			    ZPOOL_PROP_VERSION, NULL);
3640 
3641 			if (islog) {
3642 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3643 				    "cannot replace a log with a spare"));
3644 			} else if (rebuild) {
3645 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3646 				    "only mirror and dRAID vdevs support "
3647 				    "sequential reconstruction"));
3648 			} else if (zpool_is_draid_spare(new_disk)) {
3649 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3650 				    "dRAID spares can only replace child "
3651 				    "devices in their parent's dRAID vdev"));
3652 			} else if (version >= SPA_VERSION_MULTI_REPLACE) {
3653 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3654 				    "already in replacing/spare config; wait "
3655 				    "for completion or use 'zpool detach'"));
3656 			} else {
3657 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3658 				    "cannot replace a replacing device"));
3659 			}
3660 		} else if (strcmp(type, VDEV_TYPE_RAIDZ) == 0) {
3661 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3662 			    "raidz_expansion feature must be enabled "
3663 			    "in order to attach a device to raidz"));
3664 		} else {
3665 			char status[64] = {0};
3666 			zpool_prop_get_feature(zhp,
3667 			    "feature@device_rebuild", status, 63);
3668 			if (rebuild &&
3669 			    strncmp(status, ZFS_FEATURE_DISABLED, 64) == 0) {
3670 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3671 				    "device_rebuild feature must be enabled "
3672 				    "in order to use sequential "
3673 				    "reconstruction"));
3674 			} else {
3675 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3676 				    "can only attach to mirrors and top-level "
3677 				    "disks"));
3678 			}
3679 		}
3680 		(void) zfs_error(hdl, EZFS_BADTARGET, errbuf);
3681 		break;
3682 
3683 	case EINVAL:
3684 		/*
3685 		 * The new device must be a single disk.
3686 		 */
3687 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3688 		    "new device must be a single disk"));
3689 		(void) zfs_error(hdl, EZFS_INVALCONFIG, errbuf);
3690 		break;
3691 
3692 	case EBUSY:
3693 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy"),
3694 		    new_disk);
3695 		(void) zfs_error(hdl, EZFS_BADDEV, errbuf);
3696 		break;
3697 
3698 	case EOVERFLOW:
3699 		/*
3700 		 * The new device is too small.
3701 		 */
3702 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3703 		    "device is too small"));
3704 		(void) zfs_error(hdl, EZFS_BADDEV, errbuf);
3705 		break;
3706 
3707 	case EDOM:
3708 		/*
3709 		 * The new device has a different optimal sector size.
3710 		 */
3711 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3712 		    "new device has a different optimal sector size; use the "
3713 		    "option '-o ashift=N' to override the optimal size"));
3714 		(void) zfs_error(hdl, EZFS_BADDEV, errbuf);
3715 		break;
3716 
3717 	case ENAMETOOLONG:
3718 		/*
3719 		 * The resulting top-level vdev spec won't fit in the label.
3720 		 */
3721 		(void) zfs_error(hdl, EZFS_DEVOVERFLOW, errbuf);
3722 		break;
3723 
3724 	case ENXIO:
3725 		/*
3726 		 * The existing raidz vdev has offline children
3727 		 */
3728 		if (strcmp(type, VDEV_TYPE_RAIDZ) == 0) {
3729 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3730 			    "raidz vdev has devices that are are offline or "
3731 			    "being replaced"));
3732 			(void) zfs_error(hdl, EZFS_BADDEV, errbuf);
3733 			break;
3734 		} else {
3735 			(void) zpool_standard_error(hdl, errno, errbuf);
3736 		}
3737 		break;
3738 
3739 	case EADDRINUSE:
3740 		/*
3741 		 * The boot reserved area is already being used (FreeBSD)
3742 		 */
3743 		if (strcmp(type, VDEV_TYPE_RAIDZ) == 0) {
3744 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3745 			    "the reserved boot area needed for the expansion "
3746 			    "is already being used by a boot loader"));
3747 			(void) zfs_error(hdl, EZFS_BADDEV, errbuf);
3748 		} else {
3749 			(void) zpool_standard_error(hdl, errno, errbuf);
3750 		}
3751 		break;
3752 
3753 	case ZFS_ERR_ASHIFT_MISMATCH:
3754 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3755 		    "The new device cannot have a higher alignment requirement "
3756 		    "than the top-level vdev."));
3757 		(void) zfs_error(hdl, EZFS_BADTARGET, errbuf);
3758 		break;
3759 	default:
3760 		(void) zpool_standard_error(hdl, errno, errbuf);
3761 	}
3762 
3763 	return (-1);
3764 }
3765 
3766 /*
3767  * Detach the specified device.
3768  */
3769 int
3770 zpool_vdev_detach(zpool_handle_t *zhp, const char *path)
3771 {
3772 	zfs_cmd_t zc = {"\0"};
3773 	char errbuf[ERRBUFLEN];
3774 	nvlist_t *tgt;
3775 	boolean_t avail_spare, l2cache;
3776 	libzfs_handle_t *hdl = zhp->zpool_hdl;
3777 
3778 	(void) snprintf(errbuf, sizeof (errbuf),
3779 	    dgettext(TEXT_DOMAIN, "cannot detach %s"), path);
3780 
3781 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3782 	if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
3783 	    NULL)) == NULL)
3784 		return (zfs_error(hdl, EZFS_NODEVICE, errbuf));
3785 
3786 	if (avail_spare)
3787 		return (zfs_error(hdl, EZFS_ISSPARE, errbuf));
3788 
3789 	if (l2cache)
3790 		return (zfs_error(hdl, EZFS_ISL2CACHE, errbuf));
3791 
3792 	zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
3793 
3794 	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0)
3795 		return (0);
3796 
3797 	switch (errno) {
3798 
3799 	case ENOTSUP:
3800 		/*
3801 		 * Can't detach from this type of vdev.
3802 		 */
3803 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only "
3804 		    "applicable to mirror and replacing vdevs"));
3805 		(void) zfs_error(hdl, EZFS_BADTARGET, errbuf);
3806 		break;
3807 
3808 	case EBUSY:
3809 		/*
3810 		 * There are no other replicas of this device.
3811 		 */
3812 		(void) zfs_error(hdl, EZFS_NOREPLICAS, errbuf);
3813 		break;
3814 
3815 	default:
3816 		(void) zpool_standard_error(hdl, errno, errbuf);
3817 	}
3818 
3819 	return (-1);
3820 }
3821 
3822 /*
3823  * Find a mirror vdev in the source nvlist.
3824  *
3825  * The mchild array contains a list of disks in one of the top-level mirrors
3826  * of the source pool.  The schild array contains a list of disks that the
3827  * user specified on the command line.  We loop over the mchild array to
3828  * see if any entry in the schild array matches.
3829  *
3830  * If a disk in the mchild array is found in the schild array, we return
3831  * the index of that entry.  Otherwise we return -1.
3832  */
3833 static int
3834 find_vdev_entry(zpool_handle_t *zhp, nvlist_t **mchild, uint_t mchildren,
3835     nvlist_t **schild, uint_t schildren)
3836 {
3837 	uint_t mc;
3838 
3839 	for (mc = 0; mc < mchildren; mc++) {
3840 		uint_t sc;
3841 		char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp,
3842 		    mchild[mc], 0);
3843 
3844 		for (sc = 0; sc < schildren; sc++) {
3845 			char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp,
3846 			    schild[sc], 0);
3847 			boolean_t result = (strcmp(mpath, spath) == 0);
3848 
3849 			free(spath);
3850 			if (result) {
3851 				free(mpath);
3852 				return (mc);
3853 			}
3854 		}
3855 
3856 		free(mpath);
3857 	}
3858 
3859 	return (-1);
3860 }
3861 
3862 /*
3863  * Split a mirror pool.  If newroot points to null, then a new nvlist
3864  * is generated and it is the responsibility of the caller to free it.
3865  */
3866 int
3867 zpool_vdev_split(zpool_handle_t *zhp, char *newname, nvlist_t **newroot,
3868     nvlist_t *props, splitflags_t flags)
3869 {
3870 	zfs_cmd_t zc = {"\0"};
3871 	char errbuf[ERRBUFLEN];
3872 	const char *bias;
3873 	nvlist_t *tree, *config, **child, **newchild, *newconfig = NULL;
3874 	nvlist_t **varray = NULL, *zc_props = NULL;
3875 	uint_t c, children, newchildren, lastlog = 0, vcount, found = 0;
3876 	libzfs_handle_t *hdl = zhp->zpool_hdl;
3877 	uint64_t vers, readonly = B_FALSE;
3878 	boolean_t freelist = B_FALSE, memory_err = B_TRUE;
3879 	int retval = 0;
3880 
3881 	(void) snprintf(errbuf, sizeof (errbuf),
3882 	    dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name);
3883 
3884 	if (!zpool_name_valid(hdl, B_FALSE, newname))
3885 		return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
3886 
3887 	if ((config = zpool_get_config(zhp, NULL)) == NULL) {
3888 		(void) fprintf(stderr, gettext("Internal error: unable to "
3889 		    "retrieve pool configuration\n"));
3890 		return (-1);
3891 	}
3892 
3893 	tree = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE);
3894 	vers = fnvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION);
3895 
3896 	if (props) {
3897 		prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
3898 		if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name,
3899 		    props, vers, flags, errbuf)) == NULL)
3900 			return (-1);
3901 		(void) nvlist_lookup_uint64(zc_props,
3902 		    zpool_prop_to_name(ZPOOL_PROP_READONLY), &readonly);
3903 		if (readonly) {
3904 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3905 			    "property %s can only be set at import time"),
3906 			    zpool_prop_to_name(ZPOOL_PROP_READONLY));
3907 			return (-1);
3908 		}
3909 	}
3910 
3911 	if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child,
3912 	    &children) != 0) {
3913 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3914 		    "Source pool is missing vdev tree"));
3915 		nvlist_free(zc_props);
3916 		return (-1);
3917 	}
3918 
3919 	varray = zfs_alloc(hdl, children * sizeof (nvlist_t *));
3920 	vcount = 0;
3921 
3922 	if (*newroot == NULL ||
3923 	    nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN,
3924 	    &newchild, &newchildren) != 0)
3925 		newchildren = 0;
3926 
3927 	for (c = 0; c < children; c++) {
3928 		uint64_t is_log = B_FALSE, is_hole = B_FALSE;
3929 		boolean_t is_special = B_FALSE, is_dedup = B_FALSE;
3930 		const char *type;
3931 		nvlist_t **mchild, *vdev;
3932 		uint_t mchildren;
3933 		int entry;
3934 
3935 		/*
3936 		 * Unlike cache & spares, slogs are stored in the
3937 		 * ZPOOL_CONFIG_CHILDREN array.  We filter them out here.
3938 		 */
3939 		(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
3940 		    &is_log);
3941 		(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
3942 		    &is_hole);
3943 		if (is_log || is_hole) {
3944 			/*
3945 			 * Create a hole vdev and put it in the config.
3946 			 */
3947 			if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0)
3948 				goto out;
3949 			if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE,
3950 			    VDEV_TYPE_HOLE) != 0)
3951 				goto out;
3952 			if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE,
3953 			    1) != 0)
3954 				goto out;
3955 			if (lastlog == 0)
3956 				lastlog = vcount;
3957 			varray[vcount++] = vdev;
3958 			continue;
3959 		}
3960 		lastlog = 0;
3961 		type = fnvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE);
3962 
3963 		if (strcmp(type, VDEV_TYPE_INDIRECT) == 0) {
3964 			vdev = child[c];
3965 			if (nvlist_dup(vdev, &varray[vcount++], 0) != 0)
3966 				goto out;
3967 			continue;
3968 		} else if (strcmp(type, VDEV_TYPE_MIRROR) != 0) {
3969 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3970 			    "Source pool must be composed only of mirrors\n"));
3971 			retval = zfs_error(hdl, EZFS_INVALCONFIG, errbuf);
3972 			goto out;
3973 		}
3974 
3975 		if (nvlist_lookup_string(child[c],
3976 		    ZPOOL_CONFIG_ALLOCATION_BIAS, &bias) == 0) {
3977 			if (strcmp(bias, VDEV_ALLOC_BIAS_SPECIAL) == 0)
3978 				is_special = B_TRUE;
3979 			else if (strcmp(bias, VDEV_ALLOC_BIAS_DEDUP) == 0)
3980 				is_dedup = B_TRUE;
3981 		}
3982 		verify(nvlist_lookup_nvlist_array(child[c],
3983 		    ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0);
3984 
3985 		/* find or add an entry for this top-level vdev */
3986 		if (newchildren > 0 &&
3987 		    (entry = find_vdev_entry(zhp, mchild, mchildren,
3988 		    newchild, newchildren)) >= 0) {
3989 			/* We found a disk that the user specified. */
3990 			vdev = mchild[entry];
3991 			++found;
3992 		} else {
3993 			/* User didn't specify a disk for this vdev. */
3994 			vdev = mchild[mchildren - 1];
3995 		}
3996 
3997 		if (nvlist_dup(vdev, &varray[vcount++], 0) != 0)
3998 			goto out;
3999 
4000 		if (flags.dryrun != 0) {
4001 			if (is_dedup == B_TRUE) {
4002 				if (nvlist_add_string(varray[vcount - 1],
4003 				    ZPOOL_CONFIG_ALLOCATION_BIAS,
4004 				    VDEV_ALLOC_BIAS_DEDUP) != 0)
4005 					goto out;
4006 			} else if (is_special == B_TRUE) {
4007 				if (nvlist_add_string(varray[vcount - 1],
4008 				    ZPOOL_CONFIG_ALLOCATION_BIAS,
4009 				    VDEV_ALLOC_BIAS_SPECIAL) != 0)
4010 					goto out;
4011 			}
4012 		}
4013 	}
4014 
4015 	/* did we find every disk the user specified? */
4016 	if (found != newchildren) {
4017 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must "
4018 		    "include at most one disk from each mirror"));
4019 		retval = zfs_error(hdl, EZFS_INVALCONFIG, errbuf);
4020 		goto out;
4021 	}
4022 
4023 	/* Prepare the nvlist for populating. */
4024 	if (*newroot == NULL) {
4025 		if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0)
4026 			goto out;
4027 		freelist = B_TRUE;
4028 		if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE,
4029 		    VDEV_TYPE_ROOT) != 0)
4030 			goto out;
4031 	} else {
4032 		verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0);
4033 	}
4034 
4035 	/* Add all the children we found */
4036 	if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN,
4037 	    (const nvlist_t **)varray, lastlog == 0 ? vcount : lastlog) != 0)
4038 		goto out;
4039 
4040 	/*
4041 	 * If we're just doing a dry run, exit now with success.
4042 	 */
4043 	if (flags.dryrun) {
4044 		memory_err = B_FALSE;
4045 		freelist = B_FALSE;
4046 		goto out;
4047 	}
4048 
4049 	/* now build up the config list & call the ioctl */
4050 	if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0)
4051 		goto out;
4052 
4053 	if (nvlist_add_nvlist(newconfig,
4054 	    ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 ||
4055 	    nvlist_add_string(newconfig,
4056 	    ZPOOL_CONFIG_POOL_NAME, newname) != 0 ||
4057 	    nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0)
4058 		goto out;
4059 
4060 	/*
4061 	 * The new pool is automatically part of the namespace unless we
4062 	 * explicitly export it.
4063 	 */
4064 	if (!flags.import)
4065 		zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT;
4066 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4067 	(void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string));
4068 	zcmd_write_conf_nvlist(hdl, &zc, newconfig);
4069 	if (zc_props != NULL)
4070 		zcmd_write_src_nvlist(hdl, &zc, zc_props);
4071 
4072 	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) {
4073 		retval = zpool_standard_error(hdl, errno, errbuf);
4074 		goto out;
4075 	}
4076 
4077 	freelist = B_FALSE;
4078 	memory_err = B_FALSE;
4079 
4080 out:
4081 	if (varray != NULL) {
4082 		int v;
4083 
4084 		for (v = 0; v < vcount; v++)
4085 			nvlist_free(varray[v]);
4086 		free(varray);
4087 	}
4088 	zcmd_free_nvlists(&zc);
4089 	nvlist_free(zc_props);
4090 	nvlist_free(newconfig);
4091 	if (freelist) {
4092 		nvlist_free(*newroot);
4093 		*newroot = NULL;
4094 	}
4095 
4096 	if (retval != 0)
4097 		return (retval);
4098 
4099 	if (memory_err)
4100 		return (no_memory(hdl));
4101 
4102 	return (0);
4103 }
4104 
4105 /*
4106  * Remove the given device.
4107  */
4108 int
4109 zpool_vdev_remove(zpool_handle_t *zhp, const char *path)
4110 {
4111 	zfs_cmd_t zc = {"\0"};
4112 	char errbuf[ERRBUFLEN];
4113 	nvlist_t *tgt;
4114 	boolean_t avail_spare, l2cache, islog;
4115 	libzfs_handle_t *hdl = zhp->zpool_hdl;
4116 	uint64_t version;
4117 
4118 	(void) snprintf(errbuf, sizeof (errbuf),
4119 	    dgettext(TEXT_DOMAIN, "cannot remove %s"), path);
4120 
4121 	if (zpool_is_draid_spare(path)) {
4122 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4123 		    "dRAID spares cannot be removed"));
4124 		return (zfs_error(hdl, EZFS_NODEVICE, errbuf));
4125 	}
4126 
4127 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4128 	if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
4129 	    &islog)) == NULL)
4130 		return (zfs_error(hdl, EZFS_NODEVICE, errbuf));
4131 
4132 	version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
4133 	if (islog && version < SPA_VERSION_HOLES) {
4134 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4135 		    "pool must be upgraded to support log removal"));
4136 		return (zfs_error(hdl, EZFS_BADVERSION, errbuf));
4137 	}
4138 
4139 	zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
4140 
4141 	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
4142 		return (0);
4143 
4144 	switch (errno) {
4145 
4146 	case EALREADY:
4147 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4148 		    "removal for this vdev is already in progress."));
4149 		(void) zfs_error(hdl, EZFS_BUSY, errbuf);
4150 		break;
4151 
4152 	case EINVAL:
4153 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4154 		    "invalid config; all top-level vdevs must "
4155 		    "have the same sector size and not be raidz."));
4156 		(void) zfs_error(hdl, EZFS_INVALCONFIG, errbuf);
4157 		break;
4158 
4159 	case EBUSY:
4160 		if (islog) {
4161 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4162 			    "Mount encrypted datasets to replay logs."));
4163 		} else {
4164 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4165 			    "Pool busy; removal may already be in progress"));
4166 		}
4167 		(void) zfs_error(hdl, EZFS_BUSY, errbuf);
4168 		break;
4169 
4170 	case EACCES:
4171 		if (islog) {
4172 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4173 			    "Mount encrypted datasets to replay logs."));
4174 			(void) zfs_error(hdl, EZFS_BUSY, errbuf);
4175 		} else {
4176 			(void) zpool_standard_error(hdl, errno, errbuf);
4177 		}
4178 		break;
4179 
4180 	default:
4181 		(void) zpool_standard_error(hdl, errno, errbuf);
4182 	}
4183 	return (-1);
4184 }
4185 
4186 int
4187 zpool_vdev_remove_cancel(zpool_handle_t *zhp)
4188 {
4189 	zfs_cmd_t zc = {{0}};
4190 	char errbuf[ERRBUFLEN];
4191 	libzfs_handle_t *hdl = zhp->zpool_hdl;
4192 
4193 	(void) snprintf(errbuf, sizeof (errbuf),
4194 	    dgettext(TEXT_DOMAIN, "cannot cancel removal"));
4195 
4196 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4197 	zc.zc_cookie = 1;
4198 
4199 	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
4200 		return (0);
4201 
4202 	return (zpool_standard_error(hdl, errno, errbuf));
4203 }
4204 
4205 int
4206 zpool_vdev_indirect_size(zpool_handle_t *zhp, const char *path,
4207     uint64_t *sizep)
4208 {
4209 	char errbuf[ERRBUFLEN];
4210 	nvlist_t *tgt;
4211 	boolean_t avail_spare, l2cache, islog;
4212 	libzfs_handle_t *hdl = zhp->zpool_hdl;
4213 
4214 	(void) snprintf(errbuf, sizeof (errbuf),
4215 	    dgettext(TEXT_DOMAIN, "cannot determine indirect size of %s"),
4216 	    path);
4217 
4218 	if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
4219 	    &islog)) == NULL)
4220 		return (zfs_error(hdl, EZFS_NODEVICE, errbuf));
4221 
4222 	if (avail_spare || l2cache || islog) {
4223 		*sizep = 0;
4224 		return (0);
4225 	}
4226 
4227 	if (nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_INDIRECT_SIZE, sizep) != 0) {
4228 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4229 		    "indirect size not available"));
4230 		return (zfs_error(hdl, EINVAL, errbuf));
4231 	}
4232 	return (0);
4233 }
4234 
4235 /*
4236  * Clear the errors for the pool, or the particular device if specified.
4237  */
4238 int
4239 zpool_clear(zpool_handle_t *zhp, const char *path, nvlist_t *rewindnvl)
4240 {
4241 	zfs_cmd_t zc = {"\0"};
4242 	char errbuf[ERRBUFLEN];
4243 	nvlist_t *tgt;
4244 	zpool_load_policy_t policy;
4245 	boolean_t avail_spare, l2cache;
4246 	libzfs_handle_t *hdl = zhp->zpool_hdl;
4247 	nvlist_t *nvi = NULL;
4248 	int error;
4249 
4250 	if (path)
4251 		(void) snprintf(errbuf, sizeof (errbuf),
4252 		    dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
4253 		    path);
4254 	else
4255 		(void) snprintf(errbuf, sizeof (errbuf),
4256 		    dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
4257 		    zhp->zpool_name);
4258 
4259 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4260 	if (path) {
4261 		if ((tgt = zpool_find_vdev(zhp, path, &avail_spare,
4262 		    &l2cache, NULL)) == NULL)
4263 			return (zfs_error(hdl, EZFS_NODEVICE, errbuf));
4264 
4265 		/*
4266 		 * Don't allow error clearing for hot spares.  Do allow
4267 		 * error clearing for l2cache devices.
4268 		 */
4269 		if (avail_spare)
4270 			return (zfs_error(hdl, EZFS_ISSPARE, errbuf));
4271 
4272 		zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
4273 	}
4274 
4275 	zpool_get_load_policy(rewindnvl, &policy);
4276 	zc.zc_cookie = policy.zlp_rewind;
4277 
4278 	zcmd_alloc_dst_nvlist(hdl, &zc, zhp->zpool_config_size * 2);
4279 	zcmd_write_src_nvlist(hdl, &zc, rewindnvl);
4280 
4281 	while ((error = zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc)) != 0 &&
4282 	    errno == ENOMEM)
4283 		zcmd_expand_dst_nvlist(hdl, &zc);
4284 
4285 	if (!error || ((policy.zlp_rewind & ZPOOL_TRY_REWIND) &&
4286 	    errno != EPERM && errno != EACCES)) {
4287 		if (policy.zlp_rewind &
4288 		    (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
4289 			(void) zcmd_read_dst_nvlist(hdl, &zc, &nvi);
4290 			zpool_rewind_exclaim(hdl, zc.zc_name,
4291 			    ((policy.zlp_rewind & ZPOOL_TRY_REWIND) != 0),
4292 			    nvi);
4293 			nvlist_free(nvi);
4294 		}
4295 		zcmd_free_nvlists(&zc);
4296 		return (0);
4297 	}
4298 
4299 	zcmd_free_nvlists(&zc);
4300 	return (zpool_standard_error(hdl, errno, errbuf));
4301 }
4302 
4303 /*
4304  * Similar to zpool_clear(), but takes a GUID (used by fmd).
4305  */
4306 int
4307 zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid)
4308 {
4309 	zfs_cmd_t zc = {"\0"};
4310 	char errbuf[ERRBUFLEN];
4311 	libzfs_handle_t *hdl = zhp->zpool_hdl;
4312 
4313 	(void) snprintf(errbuf, sizeof (errbuf),
4314 	    dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"),
4315 	    (u_longlong_t)guid);
4316 
4317 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4318 	zc.zc_guid = guid;
4319 	zc.zc_cookie = ZPOOL_NO_REWIND;
4320 
4321 	if (zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc) == 0)
4322 		return (0);
4323 
4324 	return (zpool_standard_error(hdl, errno, errbuf));
4325 }
4326 
4327 /*
4328  * Change the GUID for a pool.
4329  *
4330  * Similar to zpool_reguid(), but may take a GUID.
4331  *
4332  * If the guid argument is NULL, then no GUID is passed in the nvlist to the
4333  * ioctl().
4334  */
4335 int
4336 zpool_set_guid(zpool_handle_t *zhp, const uint64_t *guid)
4337 {
4338 	char errbuf[ERRBUFLEN];
4339 	libzfs_handle_t *hdl = zhp->zpool_hdl;
4340 	nvlist_t *nvl = NULL;
4341 	zfs_cmd_t zc = {"\0"};
4342 	int error = -1;
4343 
4344 	if (guid != NULL) {
4345 		if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
4346 			return (no_memory(hdl));
4347 
4348 		if (nvlist_add_uint64(nvl, ZPOOL_REGUID_GUID, *guid) != 0) {
4349 			nvlist_free(nvl);
4350 			return (no_memory(hdl));
4351 		}
4352 
4353 		zcmd_write_src_nvlist(hdl, &zc, nvl);
4354 	}
4355 
4356 	(void) snprintf(errbuf, sizeof (errbuf),
4357 	    dgettext(TEXT_DOMAIN, "cannot reguid '%s'"), zhp->zpool_name);
4358 
4359 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4360 	error = zfs_ioctl(hdl, ZFS_IOC_POOL_REGUID, &zc);
4361 	if (error) {
4362 		return (zpool_standard_error(hdl, errno, errbuf));
4363 	}
4364 	if (guid != NULL) {
4365 		zcmd_free_nvlists(&zc);
4366 		nvlist_free(nvl);
4367 	}
4368 	return (0);
4369 }
4370 
4371 /*
4372  * Change the GUID for a pool.
4373  */
4374 int
4375 zpool_reguid(zpool_handle_t *zhp)
4376 {
4377 	return (zpool_set_guid(zhp, NULL));
4378 }
4379 
4380 /*
4381  * Reopen the pool.
4382  */
4383 int
4384 zpool_reopen_one(zpool_handle_t *zhp, void *data)
4385 {
4386 	libzfs_handle_t *hdl = zpool_get_handle(zhp);
4387 	const char *pool_name = zpool_get_name(zhp);
4388 	boolean_t *scrub_restart = data;
4389 	int error;
4390 
4391 	error = lzc_reopen(pool_name, *scrub_restart);
4392 	if (error) {
4393 		return (zpool_standard_error_fmt(hdl, error,
4394 		    dgettext(TEXT_DOMAIN, "cannot reopen '%s'"), pool_name));
4395 	}
4396 
4397 	return (0);
4398 }
4399 
4400 /* call into libzfs_core to execute the sync IOCTL per pool */
4401 int
4402 zpool_sync_one(zpool_handle_t *zhp, void *data)
4403 {
4404 	int ret;
4405 	libzfs_handle_t *hdl = zpool_get_handle(zhp);
4406 	const char *pool_name = zpool_get_name(zhp);
4407 	boolean_t *force = data;
4408 	nvlist_t *innvl = fnvlist_alloc();
4409 
4410 	fnvlist_add_boolean_value(innvl, "force", *force);
4411 	if ((ret = lzc_sync(pool_name, innvl, NULL)) != 0) {
4412 		nvlist_free(innvl);
4413 		return (zpool_standard_error_fmt(hdl, ret,
4414 		    dgettext(TEXT_DOMAIN, "sync '%s' failed"), pool_name));
4415 	}
4416 	nvlist_free(innvl);
4417 
4418 	return (0);
4419 }
4420 
4421 #define	PATH_BUF_LEN	64
4422 
4423 /*
4424  * Given a vdev, return the name to display in iostat.  If the vdev has a path,
4425  * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
4426  * We also check if this is a whole disk, in which case we strip off the
4427  * trailing 's0' slice name.
4428  *
4429  * This routine is also responsible for identifying when disks have been
4430  * reconfigured in a new location.  The kernel will have opened the device by
4431  * devid, but the path will still refer to the old location.  To catch this, we
4432  * first do a path -> devid translation (which is fast for the common case).  If
4433  * the devid matches, we're done.  If not, we do a reverse devid -> path
4434  * translation and issue the appropriate ioctl() to update the path of the vdev.
4435  * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
4436  * of these checks.
4437  */
4438 char *
4439 zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv,
4440     int name_flags)
4441 {
4442 	const char *type, *tpath;
4443 	const char *path;
4444 	uint64_t value;
4445 	char buf[PATH_BUF_LEN];
4446 	char tmpbuf[PATH_BUF_LEN * 2];
4447 
4448 	/*
4449 	 * vdev_name will be "root"/"root-0" for the root vdev, but it is the
4450 	 * zpool name that will be displayed to the user.
4451 	 */
4452 	type = fnvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE);
4453 	if (zhp != NULL && strcmp(type, "root") == 0)
4454 		return (zfs_strdup(hdl, zpool_get_name(zhp)));
4455 
4456 	if (libzfs_envvar_is_set("ZPOOL_VDEV_NAME_PATH"))
4457 		name_flags |= VDEV_NAME_PATH;
4458 	if (libzfs_envvar_is_set("ZPOOL_VDEV_NAME_GUID"))
4459 		name_flags |= VDEV_NAME_GUID;
4460 	if (libzfs_envvar_is_set("ZPOOL_VDEV_NAME_FOLLOW_LINKS"))
4461 		name_flags |= VDEV_NAME_FOLLOW_LINKS;
4462 
4463 	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, &value) == 0 ||
4464 	    name_flags & VDEV_NAME_GUID) {
4465 		(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value);
4466 		(void) snprintf(buf, sizeof (buf), "%llu", (u_longlong_t)value);
4467 		path = buf;
4468 	} else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &tpath) == 0) {
4469 		path = tpath;
4470 
4471 		if (name_flags & VDEV_NAME_FOLLOW_LINKS) {
4472 			char *rp = realpath(path, NULL);
4473 			if (rp) {
4474 				strlcpy(buf, rp, sizeof (buf));
4475 				path = buf;
4476 				free(rp);
4477 			}
4478 		}
4479 
4480 		/*
4481 		 * For a block device only use the name.
4482 		 */
4483 		if ((strcmp(type, VDEV_TYPE_DISK) == 0) &&
4484 		    !(name_flags & VDEV_NAME_PATH)) {
4485 			path = zfs_strip_path(path);
4486 		}
4487 
4488 		/*
4489 		 * Remove the partition from the path if this is a whole disk.
4490 		 */
4491 		if (strcmp(type, VDEV_TYPE_DRAID_SPARE) != 0 &&
4492 		    nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, &value)
4493 		    == 0 && value && !(name_flags & VDEV_NAME_PATH)) {
4494 			return (zfs_strip_partition(path));
4495 		}
4496 	} else {
4497 		path = type;
4498 
4499 		/*
4500 		 * If it's a raidz device, we need to stick in the parity level.
4501 		 */
4502 		if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) {
4503 			value = fnvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY);
4504 			(void) snprintf(buf, sizeof (buf), "%s%llu", path,
4505 			    (u_longlong_t)value);
4506 			path = buf;
4507 		}
4508 
4509 		/*
4510 		 * If it's a dRAID device, we add parity, groups, and spares.
4511 		 */
4512 		if (strcmp(path, VDEV_TYPE_DRAID) == 0) {
4513 			uint64_t ndata, nparity, nspares;
4514 			nvlist_t **child;
4515 			uint_t children;
4516 
4517 			verify(nvlist_lookup_nvlist_array(nv,
4518 			    ZPOOL_CONFIG_CHILDREN, &child, &children) == 0);
4519 			nparity = fnvlist_lookup_uint64(nv,
4520 			    ZPOOL_CONFIG_NPARITY);
4521 			ndata = fnvlist_lookup_uint64(nv,
4522 			    ZPOOL_CONFIG_DRAID_NDATA);
4523 			nspares = fnvlist_lookup_uint64(nv,
4524 			    ZPOOL_CONFIG_DRAID_NSPARES);
4525 
4526 			path = zpool_draid_name(buf, sizeof (buf), ndata,
4527 			    nparity, nspares, children);
4528 		}
4529 
4530 		/*
4531 		 * We identify each top-level vdev by using a <type-id>
4532 		 * naming convention.
4533 		 */
4534 		if (name_flags & VDEV_NAME_TYPE_ID) {
4535 			uint64_t id = fnvlist_lookup_uint64(nv,
4536 			    ZPOOL_CONFIG_ID);
4537 			(void) snprintf(tmpbuf, sizeof (tmpbuf), "%s-%llu",
4538 			    path, (u_longlong_t)id);
4539 			path = tmpbuf;
4540 		}
4541 	}
4542 
4543 	return (zfs_strdup(hdl, path));
4544 }
4545 
4546 static int
4547 zbookmark_mem_compare(const void *a, const void *b)
4548 {
4549 	return (memcmp(a, b, sizeof (zbookmark_phys_t)));
4550 }
4551 
4552 void
4553 zpool_add_propname(zpool_handle_t *zhp, const char *propname)
4554 {
4555 	assert(zhp->zpool_n_propnames < ZHP_MAX_PROPNAMES);
4556 	zhp->zpool_propnames[zhp->zpool_n_propnames] = propname;
4557 	zhp->zpool_n_propnames++;
4558 }
4559 
4560 /*
4561  * Retrieve the persistent error log, uniquify the members, and return to the
4562  * caller.
4563  */
4564 int
4565 zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp)
4566 {
4567 	zfs_cmd_t zc = {"\0"};
4568 	libzfs_handle_t *hdl = zhp->zpool_hdl;
4569 	zbookmark_phys_t *buf;
4570 	uint64_t buflen = 10000; /* approx. 1MB of RAM */
4571 
4572 	if (fnvlist_lookup_uint64(zhp->zpool_config,
4573 	    ZPOOL_CONFIG_ERRCOUNT) == 0)
4574 		return (0);
4575 
4576 	/*
4577 	 * Retrieve the raw error list from the kernel.  If it doesn't fit,
4578 	 * allocate a larger buffer and retry.
4579 	 */
4580 	(void) strcpy(zc.zc_name, zhp->zpool_name);
4581 	for (;;) {
4582 		buf = zfs_alloc(zhp->zpool_hdl,
4583 		    buflen * sizeof (zbookmark_phys_t));
4584 		zc.zc_nvlist_dst = (uintptr_t)buf;
4585 		zc.zc_nvlist_dst_size = buflen;
4586 		if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_ERROR_LOG,
4587 		    &zc) != 0) {
4588 			free(buf);
4589 			if (errno == ENOMEM) {
4590 				buflen *= 2;
4591 			} else {
4592 				return (zpool_standard_error_fmt(hdl, errno,
4593 				    dgettext(TEXT_DOMAIN, "errors: List of "
4594 				    "errors unavailable")));
4595 			}
4596 		} else {
4597 			break;
4598 		}
4599 	}
4600 
4601 	/*
4602 	 * Sort the resulting bookmarks.  This is a little confusing due to the
4603 	 * implementation of ZFS_IOC_ERROR_LOG.  The bookmarks are copied last
4604 	 * to first, and 'zc_nvlist_dst_size' indicates the number of bookmarks
4605 	 * _not_ copied as part of the process.  So we point the start of our
4606 	 * array appropriate and decrement the total number of elements.
4607 	 */
4608 	zbookmark_phys_t *zb = buf + zc.zc_nvlist_dst_size;
4609 	uint64_t zblen = buflen - zc.zc_nvlist_dst_size;
4610 
4611 	qsort(zb, zblen, sizeof (zbookmark_phys_t), zbookmark_mem_compare);
4612 
4613 	verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0);
4614 
4615 	/*
4616 	 * Fill in the nverrlistp with nvlist's of dataset and object numbers.
4617 	 */
4618 	for (uint64_t i = 0; i < zblen; i++) {
4619 		nvlist_t *nv;
4620 
4621 		/* ignoring zb_blkid and zb_level for now */
4622 		if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset &&
4623 		    zb[i-1].zb_object == zb[i].zb_object)
4624 			continue;
4625 
4626 		if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0)
4627 			goto nomem;
4628 		if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET,
4629 		    zb[i].zb_objset) != 0) {
4630 			nvlist_free(nv);
4631 			goto nomem;
4632 		}
4633 		if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT,
4634 		    zb[i].zb_object) != 0) {
4635 			nvlist_free(nv);
4636 			goto nomem;
4637 		}
4638 		if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) {
4639 			nvlist_free(nv);
4640 			goto nomem;
4641 		}
4642 		nvlist_free(nv);
4643 	}
4644 
4645 	free(buf);
4646 	return (0);
4647 
4648 nomem:
4649 	free(buf);
4650 	return (no_memory(zhp->zpool_hdl));
4651 }
4652 
4653 /*
4654  * Upgrade a ZFS pool to the latest on-disk version.
4655  */
4656 int
4657 zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version)
4658 {
4659 	zfs_cmd_t zc = {"\0"};
4660 	libzfs_handle_t *hdl = zhp->zpool_hdl;
4661 
4662 	(void) strcpy(zc.zc_name, zhp->zpool_name);
4663 	zc.zc_cookie = new_version;
4664 
4665 	if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0)
4666 		return (zpool_standard_error_fmt(hdl, errno,
4667 		    dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"),
4668 		    zhp->zpool_name));
4669 	return (0);
4670 }
4671 
4672 void
4673 zfs_save_arguments(int argc, char **argv, char *string, int len)
4674 {
4675 	int i;
4676 
4677 	(void) strlcpy(string, zfs_basename(argv[0]), len);
4678 	for (i = 1; i < argc; i++) {
4679 		(void) strlcat(string, " ", len);
4680 		(void) strlcat(string, argv[i], len);
4681 	}
4682 }
4683 
4684 int
4685 zpool_log_history(libzfs_handle_t *hdl, const char *message)
4686 {
4687 	zfs_cmd_t zc = {"\0"};
4688 	nvlist_t *args;
4689 
4690 	args = fnvlist_alloc();
4691 	fnvlist_add_string(args, "message", message);
4692 	zcmd_write_src_nvlist(hdl, &zc, args);
4693 	int err = zfs_ioctl(hdl, ZFS_IOC_LOG_HISTORY, &zc);
4694 	nvlist_free(args);
4695 	zcmd_free_nvlists(&zc);
4696 	return (err);
4697 }
4698 
4699 /*
4700  * Perform ioctl to get some command history of a pool.
4701  *
4702  * 'buf' is the buffer to fill up to 'len' bytes.  'off' is the
4703  * logical offset of the history buffer to start reading from.
4704  *
4705  * Upon return, 'off' is the next logical offset to read from and
4706  * 'len' is the actual amount of bytes read into 'buf'.
4707  */
4708 static int
4709 get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len)
4710 {
4711 	zfs_cmd_t zc = {"\0"};
4712 	libzfs_handle_t *hdl = zhp->zpool_hdl;
4713 
4714 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4715 
4716 	zc.zc_history = (uint64_t)(uintptr_t)buf;
4717 	zc.zc_history_len = *len;
4718 	zc.zc_history_offset = *off;
4719 
4720 	if (zfs_ioctl(hdl, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) {
4721 		switch (errno) {
4722 		case EPERM:
4723 			return (zfs_error_fmt(hdl, EZFS_PERM,
4724 			    dgettext(TEXT_DOMAIN,
4725 			    "cannot show history for pool '%s'"),
4726 			    zhp->zpool_name));
4727 		case ENOENT:
4728 			return (zfs_error_fmt(hdl, EZFS_NOHISTORY,
4729 			    dgettext(TEXT_DOMAIN, "cannot get history for pool "
4730 			    "'%s'"), zhp->zpool_name));
4731 		case ENOTSUP:
4732 			return (zfs_error_fmt(hdl, EZFS_BADVERSION,
4733 			    dgettext(TEXT_DOMAIN, "cannot get history for pool "
4734 			    "'%s', pool must be upgraded"), zhp->zpool_name));
4735 		default:
4736 			return (zpool_standard_error_fmt(hdl, errno,
4737 			    dgettext(TEXT_DOMAIN,
4738 			    "cannot get history for '%s'"), zhp->zpool_name));
4739 		}
4740 	}
4741 
4742 	*len = zc.zc_history_len;
4743 	*off = zc.zc_history_offset;
4744 
4745 	return (0);
4746 }
4747 
4748 /*
4749  * Retrieve the command history of a pool.
4750  */
4751 int
4752 zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp, uint64_t *off,
4753     boolean_t *eof)
4754 {
4755 	libzfs_handle_t *hdl = zhp->zpool_hdl;
4756 	char *buf;
4757 	int buflen = 128 * 1024;
4758 	nvlist_t **records = NULL;
4759 	uint_t numrecords = 0;
4760 	int err = 0, i;
4761 	uint64_t start = *off;
4762 
4763 	buf = zfs_alloc(hdl, buflen);
4764 
4765 	/* process about 1MiB a time */
4766 	while (*off - start < 1024 * 1024) {
4767 		uint64_t bytes_read = buflen;
4768 		uint64_t leftover;
4769 
4770 		if ((err = get_history(zhp, buf, off, &bytes_read)) != 0)
4771 			break;
4772 
4773 		/* if nothing else was read in, we're at EOF, just return */
4774 		if (!bytes_read) {
4775 			*eof = B_TRUE;
4776 			break;
4777 		}
4778 
4779 		if ((err = zpool_history_unpack(buf, bytes_read,
4780 		    &leftover, &records, &numrecords)) != 0) {
4781 			zpool_standard_error_fmt(hdl, err,
4782 			    dgettext(TEXT_DOMAIN,
4783 			    "cannot get history for '%s'"), zhp->zpool_name);
4784 			break;
4785 		}
4786 		*off -= leftover;
4787 		if (leftover == bytes_read) {
4788 			/*
4789 			 * no progress made, because buffer is not big enough
4790 			 * to hold this record; resize and retry.
4791 			 */
4792 			buflen *= 2;
4793 			free(buf);
4794 			buf = zfs_alloc(hdl, buflen);
4795 		}
4796 	}
4797 
4798 	free(buf);
4799 
4800 	if (!err) {
4801 		*nvhisp = fnvlist_alloc();
4802 		fnvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD,
4803 		    (const nvlist_t **)records, numrecords);
4804 	}
4805 	for (i = 0; i < numrecords; i++)
4806 		nvlist_free(records[i]);
4807 	free(records);
4808 
4809 	return (err);
4810 }
4811 
4812 /*
4813  * Retrieve the next event given the passed 'zevent_fd' file descriptor.
4814  * If there is a new event available 'nvp' will contain a newly allocated
4815  * nvlist and 'dropped' will be set to the number of missed events since
4816  * the last call to this function.  When 'nvp' is set to NULL it indicates
4817  * no new events are available.  In either case the function returns 0 and
4818  * it is up to the caller to free 'nvp'.  In the case of a fatal error the
4819  * function will return a non-zero value.  When the function is called in
4820  * blocking mode (the default, unless the ZEVENT_NONBLOCK flag is passed),
4821  * it will not return until a new event is available.
4822  */
4823 int
4824 zpool_events_next(libzfs_handle_t *hdl, nvlist_t **nvp,
4825     int *dropped, unsigned flags, int zevent_fd)
4826 {
4827 	zfs_cmd_t zc = {"\0"};
4828 	int error = 0;
4829 
4830 	*nvp = NULL;
4831 	*dropped = 0;
4832 	zc.zc_cleanup_fd = zevent_fd;
4833 
4834 	if (flags & ZEVENT_NONBLOCK)
4835 		zc.zc_guid = ZEVENT_NONBLOCK;
4836 
4837 	zcmd_alloc_dst_nvlist(hdl, &zc, ZEVENT_SIZE);
4838 
4839 retry:
4840 	if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_NEXT, &zc) != 0) {
4841 		switch (errno) {
4842 		case ESHUTDOWN:
4843 			error = zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
4844 			    dgettext(TEXT_DOMAIN, "zfs shutdown"));
4845 			goto out;
4846 		case ENOENT:
4847 			/* Blocking error case should not occur */
4848 			if (!(flags & ZEVENT_NONBLOCK))
4849 				error = zpool_standard_error_fmt(hdl, errno,
4850 				    dgettext(TEXT_DOMAIN, "cannot get event"));
4851 
4852 			goto out;
4853 		case ENOMEM:
4854 			zcmd_expand_dst_nvlist(hdl, &zc);
4855 			goto retry;
4856 		default:
4857 			error = zpool_standard_error_fmt(hdl, errno,
4858 			    dgettext(TEXT_DOMAIN, "cannot get event"));
4859 			goto out;
4860 		}
4861 	}
4862 
4863 	error = zcmd_read_dst_nvlist(hdl, &zc, nvp);
4864 	if (error != 0)
4865 		goto out;
4866 
4867 	*dropped = (int)zc.zc_cookie;
4868 out:
4869 	zcmd_free_nvlists(&zc);
4870 
4871 	return (error);
4872 }
4873 
4874 /*
4875  * Clear all events.
4876  */
4877 int
4878 zpool_events_clear(libzfs_handle_t *hdl, int *count)
4879 {
4880 	zfs_cmd_t zc = {"\0"};
4881 
4882 	if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_CLEAR, &zc) != 0)
4883 		return (zpool_standard_error(hdl, errno,
4884 		    dgettext(TEXT_DOMAIN, "cannot clear events")));
4885 
4886 	if (count != NULL)
4887 		*count = (int)zc.zc_cookie; /* # of events cleared */
4888 
4889 	return (0);
4890 }
4891 
4892 /*
4893  * Seek to a specific EID, ZEVENT_SEEK_START, or ZEVENT_SEEK_END for
4894  * the passed zevent_fd file handle.  On success zero is returned,
4895  * otherwise -1 is returned and hdl->libzfs_error is set to the errno.
4896  */
4897 int
4898 zpool_events_seek(libzfs_handle_t *hdl, uint64_t eid, int zevent_fd)
4899 {
4900 	zfs_cmd_t zc = {"\0"};
4901 	int error = 0;
4902 
4903 	zc.zc_guid = eid;
4904 	zc.zc_cleanup_fd = zevent_fd;
4905 
4906 	if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_SEEK, &zc) != 0) {
4907 		switch (errno) {
4908 		case ENOENT:
4909 			error = zfs_error_fmt(hdl, EZFS_NOENT,
4910 			    dgettext(TEXT_DOMAIN, "cannot get event"));
4911 			break;
4912 
4913 		case ENOMEM:
4914 			error = zfs_error_fmt(hdl, EZFS_NOMEM,
4915 			    dgettext(TEXT_DOMAIN, "cannot get event"));
4916 			break;
4917 
4918 		default:
4919 			error = zpool_standard_error_fmt(hdl, errno,
4920 			    dgettext(TEXT_DOMAIN, "cannot get event"));
4921 			break;
4922 		}
4923 	}
4924 
4925 	return (error);
4926 }
4927 
4928 static void
4929 zpool_obj_to_path_impl(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
4930     char *pathname, size_t len, boolean_t always_unmounted)
4931 {
4932 	zfs_cmd_t zc = {"\0"};
4933 	boolean_t mounted = B_FALSE;
4934 	char *mntpnt = NULL;
4935 	char dsname[ZFS_MAX_DATASET_NAME_LEN];
4936 
4937 	if (dsobj == 0) {
4938 		/* special case for the MOS */
4939 		(void) snprintf(pathname, len, "<metadata>:<0x%llx>",
4940 		    (longlong_t)obj);
4941 		return;
4942 	}
4943 
4944 	/* get the dataset's name */
4945 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4946 	zc.zc_obj = dsobj;
4947 	if (zfs_ioctl(zhp->zpool_hdl,
4948 	    ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) {
4949 		/* just write out a path of two object numbers */
4950 		(void) snprintf(pathname, len, "<0x%llx>:<0x%llx>",
4951 		    (longlong_t)dsobj, (longlong_t)obj);
4952 		return;
4953 	}
4954 	(void) strlcpy(dsname, zc.zc_value, sizeof (dsname));
4955 
4956 	/* find out if the dataset is mounted */
4957 	mounted = !always_unmounted && is_mounted(zhp->zpool_hdl, dsname,
4958 	    &mntpnt);
4959 
4960 	/* get the corrupted object's path */
4961 	(void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name));
4962 	zc.zc_obj = obj;
4963 	if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_OBJ_TO_PATH,
4964 	    &zc) == 0) {
4965 		if (mounted) {
4966 			(void) snprintf(pathname, len, "%s%s", mntpnt,
4967 			    zc.zc_value);
4968 		} else {
4969 			(void) snprintf(pathname, len, "%s:%s",
4970 			    dsname, zc.zc_value);
4971 		}
4972 	} else {
4973 		(void) snprintf(pathname, len, "%s:<0x%llx>", dsname,
4974 		    (longlong_t)obj);
4975 	}
4976 	free(mntpnt);
4977 }
4978 
4979 void
4980 zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
4981     char *pathname, size_t len)
4982 {
4983 	zpool_obj_to_path_impl(zhp, dsobj, obj, pathname, len, B_FALSE);
4984 }
4985 
4986 void
4987 zpool_obj_to_path_ds(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
4988     char *pathname, size_t len)
4989 {
4990 	zpool_obj_to_path_impl(zhp, dsobj, obj, pathname, len, B_TRUE);
4991 }
4992 /*
4993  * Wait while the specified activity is in progress in the pool.
4994  */
4995 int
4996 zpool_wait(zpool_handle_t *zhp, zpool_wait_activity_t activity)
4997 {
4998 	boolean_t missing;
4999 
5000 	int error = zpool_wait_status(zhp, activity, &missing, NULL);
5001 
5002 	if (missing) {
5003 		(void) zpool_standard_error_fmt(zhp->zpool_hdl, ENOENT,
5004 		    dgettext(TEXT_DOMAIN, "error waiting in pool '%s'"),
5005 		    zhp->zpool_name);
5006 		return (ENOENT);
5007 	} else {
5008 		return (error);
5009 	}
5010 }
5011 
5012 /*
5013  * Wait for the given activity and return the status of the wait (whether or not
5014  * any waiting was done) in the 'waited' parameter. Non-existent pools are
5015  * reported via the 'missing' parameter, rather than by printing an error
5016  * message. This is convenient when this function is called in a loop over a
5017  * long period of time (as it is, for example, by zpool's wait cmd). In that
5018  * scenario, a pool being exported or destroyed should be considered a normal
5019  * event, so we don't want to print an error when we find that the pool doesn't
5020  * exist.
5021  */
5022 int
5023 zpool_wait_status(zpool_handle_t *zhp, zpool_wait_activity_t activity,
5024     boolean_t *missing, boolean_t *waited)
5025 {
5026 	int error = lzc_wait(zhp->zpool_name, activity, waited);
5027 	*missing = (error == ENOENT);
5028 	if (*missing)
5029 		return (0);
5030 
5031 	if (error != 0) {
5032 		(void) zpool_standard_error_fmt(zhp->zpool_hdl, error,
5033 		    dgettext(TEXT_DOMAIN, "error waiting in pool '%s'"),
5034 		    zhp->zpool_name);
5035 	}
5036 
5037 	return (error);
5038 }
5039 
5040 int
5041 zpool_set_bootenv(zpool_handle_t *zhp, const nvlist_t *envmap)
5042 {
5043 	int error = lzc_set_bootenv(zhp->zpool_name, envmap);
5044 	if (error != 0) {
5045 		(void) zpool_standard_error_fmt(zhp->zpool_hdl, error,
5046 		    dgettext(TEXT_DOMAIN,
5047 		    "error setting bootenv in pool '%s'"), zhp->zpool_name);
5048 	}
5049 
5050 	return (error);
5051 }
5052 
5053 int
5054 zpool_get_bootenv(zpool_handle_t *zhp, nvlist_t **nvlp)
5055 {
5056 	nvlist_t *nvl;
5057 	int error;
5058 
5059 	nvl = NULL;
5060 	error = lzc_get_bootenv(zhp->zpool_name, &nvl);
5061 	if (error != 0) {
5062 		(void) zpool_standard_error_fmt(zhp->zpool_hdl, error,
5063 		    dgettext(TEXT_DOMAIN,
5064 		    "error getting bootenv in pool '%s'"), zhp->zpool_name);
5065 	} else {
5066 		*nvlp = nvl;
5067 	}
5068 
5069 	return (error);
5070 }
5071 
5072 /*
5073  * Attempt to read and parse feature file(s) (from "compatibility" property).
5074  * Files contain zpool feature names, comma or whitespace-separated.
5075  * Comments (# character to next newline) are discarded.
5076  *
5077  * Arguments:
5078  *  compatibility : string containing feature filenames
5079  *  features : either NULL or pointer to array of boolean
5080  *  report : either NULL or pointer to string buffer
5081  *  rlen : length of "report" buffer
5082  *
5083  * compatibility is NULL (unset), "", "off", "legacy", or list of
5084  * comma-separated filenames. filenames should either be absolute,
5085  * or relative to:
5086  *   1) ZPOOL_SYSCONF_COMPAT_D (eg: /etc/zfs/compatibility.d) or
5087  *   2) ZPOOL_DATA_COMPAT_D (eg: /usr/share/zfs/compatibility.d).
5088  * (Unset), "" or "off" => enable all features
5089  * "legacy" => disable all features
5090  *
5091  * Any feature names read from files which match unames in spa_feature_table
5092  * will have the corresponding boolean set in the features array (if non-NULL).
5093  * If more than one feature set specified, only features present in *all* of
5094  * them will be set.
5095  *
5096  * "report" if not NULL will be populated with a suitable status message.
5097  *
5098  * Return values:
5099  *   ZPOOL_COMPATIBILITY_OK : files read and parsed ok
5100  *   ZPOOL_COMPATIBILITY_BADFILE : file too big or not a text file
5101  *   ZPOOL_COMPATIBILITY_BADTOKEN : SYSCONF file contains invalid feature name
5102  *   ZPOOL_COMPATIBILITY_WARNTOKEN : DATA file contains invalid feature name
5103  *   ZPOOL_COMPATIBILITY_NOFILES : no feature files found
5104  */
5105 zpool_compat_status_t
5106 zpool_load_compat(const char *compat, boolean_t *features, char *report,
5107     size_t rlen)
5108 {
5109 	int sdirfd, ddirfd, featfd;
5110 	struct stat fs;
5111 	char *fc;
5112 	char *ps, *ls, *ws;
5113 	char *file, *line, *word;
5114 
5115 	char l_compat[ZFS_MAXPROPLEN];
5116 
5117 	boolean_t ret_nofiles = B_TRUE;
5118 	boolean_t ret_badfile = B_FALSE;
5119 	boolean_t ret_badtoken = B_FALSE;
5120 	boolean_t ret_warntoken = B_FALSE;
5121 
5122 	/* special cases (unset), "" and "off" => enable all features */
5123 	if (compat == NULL || compat[0] == '\0' ||
5124 	    strcmp(compat, ZPOOL_COMPAT_OFF) == 0) {
5125 		if (features != NULL)
5126 			for (uint_t i = 0; i < SPA_FEATURES; i++)
5127 				features[i] = B_TRUE;
5128 		if (report != NULL)
5129 			strlcpy(report, gettext("all features enabled"), rlen);
5130 		return (ZPOOL_COMPATIBILITY_OK);
5131 	}
5132 
5133 	/* Final special case "legacy" => disable all features */
5134 	if (strcmp(compat, ZPOOL_COMPAT_LEGACY) == 0) {
5135 		if (features != NULL)
5136 			for (uint_t i = 0; i < SPA_FEATURES; i++)
5137 				features[i] = B_FALSE;
5138 		if (report != NULL)
5139 			strlcpy(report, gettext("all features disabled"), rlen);
5140 		return (ZPOOL_COMPATIBILITY_OK);
5141 	}
5142 
5143 	/*
5144 	 * Start with all true; will be ANDed with results from each file
5145 	 */
5146 	if (features != NULL)
5147 		for (uint_t i = 0; i < SPA_FEATURES; i++)
5148 			features[i] = B_TRUE;
5149 
5150 	char err_badfile[ZFS_MAXPROPLEN] = "";
5151 	char err_badtoken[ZFS_MAXPROPLEN] = "";
5152 
5153 	/*
5154 	 * We ignore errors from the directory open()
5155 	 * as they're only needed if the filename is relative
5156 	 * which will be checked during the openat().
5157 	 */
5158 
5159 /* O_PATH safer than O_RDONLY if system allows it */
5160 #if defined(O_PATH)
5161 #define	ZC_DIR_FLAGS (O_DIRECTORY | O_CLOEXEC | O_PATH)
5162 #else
5163 #define	ZC_DIR_FLAGS (O_DIRECTORY | O_CLOEXEC | O_RDONLY)
5164 #endif
5165 
5166 	sdirfd = open(ZPOOL_SYSCONF_COMPAT_D, ZC_DIR_FLAGS);
5167 	ddirfd = open(ZPOOL_DATA_COMPAT_D, ZC_DIR_FLAGS);
5168 
5169 	(void) strlcpy(l_compat, compat, ZFS_MAXPROPLEN);
5170 
5171 	for (file = strtok_r(l_compat, ",", &ps);
5172 	    file != NULL;
5173 	    file = strtok_r(NULL, ",", &ps)) {
5174 
5175 		boolean_t l_features[SPA_FEATURES];
5176 
5177 		enum { Z_SYSCONF, Z_DATA } source;
5178 
5179 		/* try sysconfdir first, then datadir */
5180 		source = Z_SYSCONF;
5181 		if ((featfd = openat(sdirfd, file, O_RDONLY | O_CLOEXEC)) < 0) {
5182 			featfd = openat(ddirfd, file, O_RDONLY | O_CLOEXEC);
5183 			source = Z_DATA;
5184 		}
5185 
5186 		/* File readable and correct size? */
5187 		if (featfd < 0 ||
5188 		    fstat(featfd, &fs) < 0 ||
5189 		    fs.st_size < 1 ||
5190 		    fs.st_size > ZPOOL_COMPAT_MAXSIZE) {
5191 			(void) close(featfd);
5192 			strlcat(err_badfile, file, ZFS_MAXPROPLEN);
5193 			strlcat(err_badfile, " ", ZFS_MAXPROPLEN);
5194 			ret_badfile = B_TRUE;
5195 			continue;
5196 		}
5197 
5198 /* Prefault the file if system allows */
5199 #if defined(MAP_POPULATE)
5200 #define	ZC_MMAP_FLAGS (MAP_PRIVATE | MAP_POPULATE)
5201 #elif defined(MAP_PREFAULT_READ)
5202 #define	ZC_MMAP_FLAGS (MAP_PRIVATE | MAP_PREFAULT_READ)
5203 #else
5204 #define	ZC_MMAP_FLAGS (MAP_PRIVATE)
5205 #endif
5206 
5207 		/* private mmap() so we can strtok safely */
5208 		fc = (char *)mmap(NULL, fs.st_size, PROT_READ | PROT_WRITE,
5209 		    ZC_MMAP_FLAGS, featfd, 0);
5210 		(void) close(featfd);
5211 
5212 		/* map ok, and last character == newline? */
5213 		if (fc == MAP_FAILED || fc[fs.st_size - 1] != '\n') {
5214 			(void) munmap((void *) fc, fs.st_size);
5215 			strlcat(err_badfile, file, ZFS_MAXPROPLEN);
5216 			strlcat(err_badfile, " ", ZFS_MAXPROPLEN);
5217 			ret_badfile = B_TRUE;
5218 			continue;
5219 		}
5220 
5221 		ret_nofiles = B_FALSE;
5222 
5223 		for (uint_t i = 0; i < SPA_FEATURES; i++)
5224 			l_features[i] = B_FALSE;
5225 
5226 		/* replace final newline with NULL to ensure string ends */
5227 		fc[fs.st_size - 1] = '\0';
5228 
5229 		for (line = strtok_r(fc, "\n", &ls);
5230 		    line != NULL;
5231 		    line = strtok_r(NULL, "\n", &ls)) {
5232 			/* discard comments */
5233 			char *r = strchr(line, '#');
5234 			if (r != NULL)
5235 				*r = '\0';
5236 
5237 			for (word = strtok_r(line, ", \t", &ws);
5238 			    word != NULL;
5239 			    word = strtok_r(NULL, ", \t", &ws)) {
5240 				/* Find matching feature name */
5241 				uint_t f;
5242 				for (f = 0; f < SPA_FEATURES; f++) {
5243 					zfeature_info_t *fi =
5244 					    &spa_feature_table[f];
5245 					if (strcmp(word, fi->fi_uname) == 0) {
5246 						l_features[f] = B_TRUE;
5247 						break;
5248 					}
5249 				}
5250 				if (f < SPA_FEATURES)
5251 					continue;
5252 
5253 				/* found an unrecognized word */
5254 				/* lightly sanitize it */
5255 				if (strlen(word) > 32)
5256 					word[32] = '\0';
5257 				for (char *c = word; *c != '\0'; c++)
5258 					if (!isprint(*c))
5259 						*c = '?';
5260 
5261 				strlcat(err_badtoken, word, ZFS_MAXPROPLEN);
5262 				strlcat(err_badtoken, " ", ZFS_MAXPROPLEN);
5263 				if (source == Z_SYSCONF)
5264 					ret_badtoken = B_TRUE;
5265 				else
5266 					ret_warntoken = B_TRUE;
5267 			}
5268 		}
5269 		(void) munmap((void *) fc, fs.st_size);
5270 
5271 		if (features != NULL)
5272 			for (uint_t i = 0; i < SPA_FEATURES; i++)
5273 				features[i] &= l_features[i];
5274 	}
5275 	(void) close(sdirfd);
5276 	(void) close(ddirfd);
5277 
5278 	/* Return the most serious error */
5279 	if (ret_badfile) {
5280 		if (report != NULL)
5281 			snprintf(report, rlen, gettext("could not read/"
5282 			    "parse feature file(s): %s"), err_badfile);
5283 		return (ZPOOL_COMPATIBILITY_BADFILE);
5284 	}
5285 	if (ret_nofiles) {
5286 		if (report != NULL)
5287 			strlcpy(report,
5288 			    gettext("no valid compatibility files specified"),
5289 			    rlen);
5290 		return (ZPOOL_COMPATIBILITY_NOFILES);
5291 	}
5292 	if (ret_badtoken) {
5293 		if (report != NULL)
5294 			snprintf(report, rlen, gettext("invalid feature "
5295 			    "name(s) in local compatibility files: %s"),
5296 			    err_badtoken);
5297 		return (ZPOOL_COMPATIBILITY_BADTOKEN);
5298 	}
5299 	if (ret_warntoken) {
5300 		if (report != NULL)
5301 			snprintf(report, rlen, gettext("unrecognized feature "
5302 			    "name(s) in distribution compatibility files: %s"),
5303 			    err_badtoken);
5304 		return (ZPOOL_COMPATIBILITY_WARNTOKEN);
5305 	}
5306 	if (report != NULL)
5307 		strlcpy(report, gettext("compatibility set ok"), rlen);
5308 	return (ZPOOL_COMPATIBILITY_OK);
5309 }
5310 
5311 static int
5312 zpool_vdev_guid(zpool_handle_t *zhp, const char *vdevname, uint64_t *vdev_guid)
5313 {
5314 	nvlist_t *tgt;
5315 	boolean_t avail_spare, l2cache;
5316 
5317 	verify(zhp != NULL);
5318 	if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
5319 		char errbuf[ERRBUFLEN];
5320 		(void) snprintf(errbuf, sizeof (errbuf),
5321 		    dgettext(TEXT_DOMAIN, "pool is in an unavailable state"));
5322 		return (zfs_error(zhp->zpool_hdl, EZFS_POOLUNAVAIL, errbuf));
5323 	}
5324 
5325 	if ((tgt = zpool_find_vdev(zhp, vdevname, &avail_spare, &l2cache,
5326 	    NULL)) == NULL) {
5327 		char errbuf[ERRBUFLEN];
5328 		(void) snprintf(errbuf, sizeof (errbuf),
5329 		    dgettext(TEXT_DOMAIN, "can not find %s in %s"),
5330 		    vdevname, zhp->zpool_name);
5331 		return (zfs_error(zhp->zpool_hdl, EZFS_NODEVICE, errbuf));
5332 	}
5333 
5334 	*vdev_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
5335 	return (0);
5336 }
5337 
5338 /*
5339  * Get a vdev property value for 'prop' and return the value in
5340  * a pre-allocated buffer.
5341  */
5342 int
5343 zpool_get_vdev_prop_value(nvlist_t *nvprop, vdev_prop_t prop, char *prop_name,
5344     char *buf, size_t len, zprop_source_t *srctype, boolean_t literal)
5345 {
5346 	nvlist_t *nv;
5347 	const char *strval;
5348 	uint64_t intval;
5349 	zprop_source_t src = ZPROP_SRC_NONE;
5350 
5351 	if (prop == VDEV_PROP_USERPROP) {
5352 		/* user property, prop_name must contain the property name */
5353 		assert(prop_name != NULL);
5354 		if (nvlist_lookup_nvlist(nvprop, prop_name, &nv) == 0) {
5355 			src = fnvlist_lookup_uint64(nv, ZPROP_SOURCE);
5356 			strval = fnvlist_lookup_string(nv, ZPROP_VALUE);
5357 		} else {
5358 			/* user prop not found */
5359 			src = ZPROP_SRC_DEFAULT;
5360 			strval = "-";
5361 		}
5362 		(void) strlcpy(buf, strval, len);
5363 		if (srctype)
5364 			*srctype = src;
5365 		return (0);
5366 	}
5367 
5368 	if (prop_name == NULL)
5369 		prop_name = (char *)vdev_prop_to_name(prop);
5370 
5371 	switch (vdev_prop_get_type(prop)) {
5372 	case PROP_TYPE_STRING:
5373 		if (nvlist_lookup_nvlist(nvprop, prop_name, &nv) == 0) {
5374 			src = fnvlist_lookup_uint64(nv, ZPROP_SOURCE);
5375 			strval = fnvlist_lookup_string(nv, ZPROP_VALUE);
5376 		} else {
5377 			src = ZPROP_SRC_DEFAULT;
5378 			if ((strval = vdev_prop_default_string(prop)) == NULL)
5379 				strval = "-";
5380 		}
5381 		(void) strlcpy(buf, strval, len);
5382 		break;
5383 
5384 	case PROP_TYPE_NUMBER:
5385 		if (nvlist_lookup_nvlist(nvprop, prop_name, &nv) == 0) {
5386 			src = fnvlist_lookup_uint64(nv, ZPROP_SOURCE);
5387 			intval = fnvlist_lookup_uint64(nv, ZPROP_VALUE);
5388 		} else {
5389 			src = ZPROP_SRC_DEFAULT;
5390 			intval = vdev_prop_default_numeric(prop);
5391 		}
5392 
5393 		switch (prop) {
5394 		case VDEV_PROP_ASIZE:
5395 		case VDEV_PROP_PSIZE:
5396 		case VDEV_PROP_SIZE:
5397 		case VDEV_PROP_BOOTSIZE:
5398 		case VDEV_PROP_ALLOCATED:
5399 		case VDEV_PROP_FREE:
5400 		case VDEV_PROP_READ_ERRORS:
5401 		case VDEV_PROP_WRITE_ERRORS:
5402 		case VDEV_PROP_CHECKSUM_ERRORS:
5403 		case VDEV_PROP_INITIALIZE_ERRORS:
5404 		case VDEV_PROP_TRIM_ERRORS:
5405 		case VDEV_PROP_SLOW_IOS:
5406 		case VDEV_PROP_OPS_NULL:
5407 		case VDEV_PROP_OPS_READ:
5408 		case VDEV_PROP_OPS_WRITE:
5409 		case VDEV_PROP_OPS_FREE:
5410 		case VDEV_PROP_OPS_CLAIM:
5411 		case VDEV_PROP_OPS_TRIM:
5412 		case VDEV_PROP_BYTES_NULL:
5413 		case VDEV_PROP_BYTES_READ:
5414 		case VDEV_PROP_BYTES_WRITE:
5415 		case VDEV_PROP_BYTES_FREE:
5416 		case VDEV_PROP_BYTES_CLAIM:
5417 		case VDEV_PROP_BYTES_TRIM:
5418 			if (literal) {
5419 				(void) snprintf(buf, len, "%llu",
5420 				    (u_longlong_t)intval);
5421 			} else {
5422 				(void) zfs_nicenum(intval, buf, len);
5423 			}
5424 			break;
5425 		case VDEV_PROP_EXPANDSZ:
5426 			if (intval == 0) {
5427 				(void) strlcpy(buf, "-", len);
5428 			} else if (literal) {
5429 				(void) snprintf(buf, len, "%llu",
5430 				    (u_longlong_t)intval);
5431 			} else {
5432 				(void) zfs_nicenum(intval, buf, len);
5433 			}
5434 			break;
5435 		case VDEV_PROP_CAPACITY:
5436 			if (literal) {
5437 				(void) snprintf(buf, len, "%llu",
5438 				    (u_longlong_t)intval);
5439 			} else {
5440 				(void) snprintf(buf, len, "%llu%%",
5441 				    (u_longlong_t)intval);
5442 			}
5443 			break;
5444 		case VDEV_PROP_CHECKSUM_N:
5445 		case VDEV_PROP_CHECKSUM_T:
5446 		case VDEV_PROP_IO_N:
5447 		case VDEV_PROP_IO_T:
5448 		case VDEV_PROP_SLOW_IO_N:
5449 		case VDEV_PROP_SLOW_IO_T:
5450 			if (intval == UINT64_MAX) {
5451 				(void) strlcpy(buf, "-", len);
5452 			} else {
5453 				(void) snprintf(buf, len, "%llu",
5454 				    (u_longlong_t)intval);
5455 			}
5456 			break;
5457 		case VDEV_PROP_FRAGMENTATION:
5458 			if (intval == UINT64_MAX) {
5459 				(void) strlcpy(buf, "-", len);
5460 			} else {
5461 				(void) snprintf(buf, len, "%llu%%",
5462 				    (u_longlong_t)intval);
5463 			}
5464 			break;
5465 		case VDEV_PROP_STATE:
5466 			if (literal) {
5467 				(void) snprintf(buf, len, "%llu",
5468 				    (u_longlong_t)intval);
5469 			} else {
5470 				(void) strlcpy(buf, zpool_state_to_name(intval,
5471 				    VDEV_AUX_NONE), len);
5472 			}
5473 			break;
5474 		default:
5475 			(void) snprintf(buf, len, "%llu",
5476 			    (u_longlong_t)intval);
5477 		}
5478 		break;
5479 
5480 	case PROP_TYPE_INDEX:
5481 		if (nvlist_lookup_nvlist(nvprop, prop_name, &nv) == 0) {
5482 			src = fnvlist_lookup_uint64(nv, ZPROP_SOURCE);
5483 			intval = fnvlist_lookup_uint64(nv, ZPROP_VALUE);
5484 		} else {
5485 			/* 'trim_support' only valid for leaf vdevs */
5486 			if (prop == VDEV_PROP_TRIM_SUPPORT) {
5487 				(void) strlcpy(buf, "-", len);
5488 				break;
5489 			}
5490 			src = ZPROP_SRC_DEFAULT;
5491 			intval = vdev_prop_default_numeric(prop);
5492 			/* Only use if provided by the RAIDZ VDEV above */
5493 			if (prop == VDEV_PROP_RAIDZ_EXPANDING)
5494 				return (ENOENT);
5495 		}
5496 		if (vdev_prop_index_to_string(prop, intval,
5497 		    (const char **)&strval) != 0)
5498 			return (-1);
5499 		(void) strlcpy(buf, strval, len);
5500 		break;
5501 
5502 	default:
5503 		abort();
5504 	}
5505 
5506 	if (srctype)
5507 		*srctype = src;
5508 
5509 	return (0);
5510 }
5511 
5512 /*
5513  * Get a vdev property value for 'prop_name' and return the value in
5514  * a pre-allocated buffer.
5515  */
5516 int
5517 zpool_get_vdev_prop(zpool_handle_t *zhp, const char *vdevname, vdev_prop_t prop,
5518     char *prop_name, char *buf, size_t len, zprop_source_t *srctype,
5519     boolean_t literal)
5520 {
5521 	nvlist_t *reqnvl, *reqprops;
5522 	nvlist_t *retprops = NULL;
5523 	uint64_t vdev_guid = 0;
5524 	int ret;
5525 
5526 	if ((ret = zpool_vdev_guid(zhp, vdevname, &vdev_guid)) != 0)
5527 		return (ret);
5528 
5529 	if (nvlist_alloc(&reqnvl, NV_UNIQUE_NAME, 0) != 0)
5530 		return (no_memory(zhp->zpool_hdl));
5531 	if (nvlist_alloc(&reqprops, NV_UNIQUE_NAME, 0) != 0)
5532 		return (no_memory(zhp->zpool_hdl));
5533 
5534 	fnvlist_add_uint64(reqnvl, ZPOOL_VDEV_PROPS_GET_VDEV, vdev_guid);
5535 
5536 	if (prop != VDEV_PROP_USERPROP) {
5537 		/* prop_name overrides prop value */
5538 		if (prop_name != NULL)
5539 			prop = vdev_name_to_prop(prop_name);
5540 		else
5541 			prop_name = (char *)vdev_prop_to_name(prop);
5542 		assert(prop < VDEV_NUM_PROPS);
5543 	}
5544 
5545 	assert(prop_name != NULL);
5546 	if (nvlist_add_uint64(reqprops, prop_name, prop) != 0) {
5547 		nvlist_free(reqnvl);
5548 		nvlist_free(reqprops);
5549 		return (no_memory(zhp->zpool_hdl));
5550 	}
5551 
5552 	fnvlist_add_nvlist(reqnvl, ZPOOL_VDEV_PROPS_GET_PROPS, reqprops);
5553 
5554 	ret = lzc_get_vdev_prop(zhp->zpool_name, reqnvl, &retprops);
5555 
5556 	if (ret == 0) {
5557 		ret = zpool_get_vdev_prop_value(retprops, prop, prop_name, buf,
5558 		    len, srctype, literal);
5559 	} else {
5560 		char errbuf[ERRBUFLEN];
5561 		(void) snprintf(errbuf, sizeof (errbuf),
5562 		    dgettext(TEXT_DOMAIN, "cannot get vdev property %s from"
5563 		    " %s in %s"), prop_name, vdevname, zhp->zpool_name);
5564 		(void) zpool_standard_error(zhp->zpool_hdl, ret, errbuf);
5565 	}
5566 
5567 	nvlist_free(reqnvl);
5568 	nvlist_free(reqprops);
5569 	nvlist_free(retprops);
5570 
5571 	return (ret);
5572 }
5573 
5574 /*
5575  * Get all vdev properties
5576  */
5577 int
5578 zpool_get_all_vdev_props(zpool_handle_t *zhp, const char *vdevname,
5579     nvlist_t **outnvl)
5580 {
5581 	nvlist_t *nvl = NULL;
5582 	uint64_t vdev_guid = 0;
5583 	int ret;
5584 
5585 	if ((ret = zpool_vdev_guid(zhp, vdevname, &vdev_guid)) != 0)
5586 		return (ret);
5587 
5588 	if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
5589 		return (no_memory(zhp->zpool_hdl));
5590 
5591 	fnvlist_add_uint64(nvl, ZPOOL_VDEV_PROPS_GET_VDEV, vdev_guid);
5592 
5593 	ret = lzc_get_vdev_prop(zhp->zpool_name, nvl, outnvl);
5594 
5595 	nvlist_free(nvl);
5596 
5597 	if (ret) {
5598 		char errbuf[ERRBUFLEN];
5599 		(void) snprintf(errbuf, sizeof (errbuf),
5600 		    dgettext(TEXT_DOMAIN, "cannot get vdev properties for"
5601 		    " %s in %s"), vdevname, zhp->zpool_name);
5602 		(void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
5603 	}
5604 
5605 	return (ret);
5606 }
5607 
5608 /*
5609  * Set vdev property
5610  */
5611 int
5612 zpool_set_vdev_prop(zpool_handle_t *zhp, const char *vdevname,
5613     const char *propname, const char *propval)
5614 {
5615 	int ret;
5616 	nvlist_t *nvl = NULL;
5617 	nvlist_t *outnvl = NULL;
5618 	nvlist_t *props;
5619 	nvlist_t *realprops;
5620 	prop_flags_t flags = { 0 };
5621 	uint64_t version;
5622 	uint64_t vdev_guid;
5623 
5624 	if ((ret = zpool_vdev_guid(zhp, vdevname, &vdev_guid)) != 0)
5625 		return (ret);
5626 
5627 	if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
5628 		return (no_memory(zhp->zpool_hdl));
5629 	if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0)
5630 		return (no_memory(zhp->zpool_hdl));
5631 
5632 	fnvlist_add_uint64(nvl, ZPOOL_VDEV_PROPS_SET_VDEV, vdev_guid);
5633 
5634 	if (nvlist_add_string(props, propname, propval) != 0) {
5635 		nvlist_free(props);
5636 		return (no_memory(zhp->zpool_hdl));
5637 	}
5638 
5639 	char errbuf[ERRBUFLEN];
5640 	(void) snprintf(errbuf, sizeof (errbuf),
5641 	    dgettext(TEXT_DOMAIN, "cannot set property %s for %s on %s"),
5642 	    propname, vdevname, zhp->zpool_name);
5643 
5644 	flags.vdevprop = 1;
5645 	version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
5646 	if ((realprops = zpool_valid_proplist(zhp->zpool_hdl,
5647 	    zhp->zpool_name, props, version, flags, errbuf)) == NULL) {
5648 		nvlist_free(props);
5649 		nvlist_free(nvl);
5650 		return (-1);
5651 	}
5652 
5653 	nvlist_free(props);
5654 	props = realprops;
5655 
5656 	fnvlist_add_nvlist(nvl, ZPOOL_VDEV_PROPS_SET_PROPS, props);
5657 
5658 	ret = lzc_set_vdev_prop(zhp->zpool_name, nvl, &outnvl);
5659 
5660 	nvlist_free(props);
5661 	nvlist_free(nvl);
5662 	nvlist_free(outnvl);
5663 
5664 	if (ret)
5665 		(void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
5666 
5667 	return (ret);
5668 }
5669 
5670 /*
5671  * Prune older entries from the DDT to reclaim space under the quota
5672  */
5673 int
5674 zpool_ddt_prune(zpool_handle_t *zhp, zpool_ddt_prune_unit_t unit,
5675     uint64_t amount)
5676 {
5677 	int error = lzc_ddt_prune(zhp->zpool_name, unit, amount);
5678 	if (error != 0) {
5679 		libzfs_handle_t *hdl = zhp->zpool_hdl;
5680 		char errbuf[ERRBUFLEN];
5681 
5682 		(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
5683 		    "cannot prune dedup table on '%s'"), zhp->zpool_name);
5684 
5685 		if (error == EALREADY) {
5686 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
5687 			    "a prune operation is already in progress"));
5688 			(void) zfs_error(hdl, EZFS_BUSY, errbuf);
5689 		} else {
5690 			(void) zpool_standard_error(hdl, errno, errbuf);
5691 		}
5692 		return (-1);
5693 	}
5694 
5695 	return (0);
5696 }
5697