xref: /freebsd/sys/contrib/openzfs/lib/libzfs/libzfs_pool.c (revision d8fbbd371ca11d9ad4b29b9d3a316885a5da0b15)
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  * Copyright (c) 2025 Hewlett Packard Enterprise Development LP.
35  */
36 
37 #include <errno.h>
38 #include <libintl.h>
39 #include <stdio.h>
40 #include <stdlib.h>
41 #include <strings.h>
42 #include <unistd.h>
43 #include <libgen.h>
44 #include <zone.h>
45 #include <sys/stat.h>
46 #include <sys/efi_partition.h>
47 #include <sys/systeminfo.h>
48 #include <sys/zfs_ioctl.h>
49 #include <sys/zfs_sysfs.h>
50 #include <sys/vdev_disk.h>
51 #include <sys/types.h>
52 #include <dlfcn.h>
53 #include <libzutil.h>
54 #include <fcntl.h>
55 
56 #include "zfs_namecheck.h"
57 #include "zfs_prop.h"
58 #include "libzfs_impl.h"
59 #include "zfs_comutil.h"
60 #include "zfeature_common.h"
61 
62 static boolean_t zpool_vdev_is_interior(const char *name);
63 
64 typedef struct prop_flags {
65 	unsigned int create:1;	/* Validate property on creation */
66 	unsigned int import:1;	/* Validate property on import */
67 	unsigned int vdevprop:1; /* Validate property as a VDEV property */
68 } prop_flags_t;
69 
70 /*
71  * ====================================================================
72  *   zpool property functions
73  * ====================================================================
74  */
75 
76 static int
zpool_get_all_props(zpool_handle_t * zhp)77 zpool_get_all_props(zpool_handle_t *zhp)
78 {
79 	zfs_cmd_t zc = {"\0"};
80 	libzfs_handle_t *hdl = zhp->zpool_hdl;
81 
82 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
83 
84 	if (zhp->zpool_n_propnames > 0) {
85 		nvlist_t *innvl = fnvlist_alloc();
86 		fnvlist_add_string_array(innvl, ZPOOL_GET_PROPS_NAMES,
87 		    zhp->zpool_propnames, zhp->zpool_n_propnames);
88 		zcmd_write_src_nvlist(hdl, &zc, innvl);
89 		fnvlist_free(innvl);
90 	}
91 
92 	zcmd_alloc_dst_nvlist(hdl, &zc, 0);
93 
94 	while (zfs_ioctl(hdl, ZFS_IOC_POOL_GET_PROPS, &zc) != 0) {
95 		if (errno == ENOMEM)
96 			zcmd_expand_dst_nvlist(hdl, &zc);
97 		else {
98 			zcmd_free_nvlists(&zc);
99 			return (-1);
100 		}
101 	}
102 
103 	if (zcmd_read_dst_nvlist(hdl, &zc, &zhp->zpool_props) != 0) {
104 		zcmd_free_nvlists(&zc);
105 		return (-1);
106 	}
107 
108 	zcmd_free_nvlists(&zc);
109 
110 	return (0);
111 }
112 
113 int
zpool_props_refresh(zpool_handle_t * zhp)114 zpool_props_refresh(zpool_handle_t *zhp)
115 {
116 	nvlist_t *old_props;
117 
118 	old_props = zhp->zpool_props;
119 
120 	if (zpool_get_all_props(zhp) != 0)
121 		return (-1);
122 
123 	nvlist_free(old_props);
124 	return (0);
125 }
126 
127 static const char *
zpool_get_prop_string(zpool_handle_t * zhp,zpool_prop_t prop,zprop_source_t * src)128 zpool_get_prop_string(zpool_handle_t *zhp, zpool_prop_t prop,
129     zprop_source_t *src)
130 {
131 	nvlist_t *nv, *nvl;
132 	const char *value;
133 	zprop_source_t source;
134 
135 	nvl = zhp->zpool_props;
136 	if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
137 		source = fnvlist_lookup_uint64(nv, ZPROP_SOURCE);
138 		value = fnvlist_lookup_string(nv, ZPROP_VALUE);
139 	} else {
140 		source = ZPROP_SRC_DEFAULT;
141 		if ((value = zpool_prop_default_string(prop)) == NULL)
142 			value = "-";
143 	}
144 
145 	if (src)
146 		*src = source;
147 
148 	return (value);
149 }
150 
151 uint64_t
zpool_get_prop_int(zpool_handle_t * zhp,zpool_prop_t prop,zprop_source_t * src)152 zpool_get_prop_int(zpool_handle_t *zhp, zpool_prop_t prop, zprop_source_t *src)
153 {
154 	nvlist_t *nv, *nvl;
155 	uint64_t value;
156 	zprop_source_t source;
157 
158 	if (zhp->zpool_props == NULL && zpool_get_all_props(zhp)) {
159 		/*
160 		 * zpool_get_all_props() has most likely failed because
161 		 * the pool is faulted, but if all we need is the top level
162 		 * vdev's guid then get it from the zhp config nvlist.
163 		 */
164 		if ((prop == ZPOOL_PROP_GUID) &&
165 		    (nvlist_lookup_nvlist(zhp->zpool_config,
166 		    ZPOOL_CONFIG_VDEV_TREE, &nv) == 0) &&
167 		    (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value)
168 		    == 0)) {
169 			return (value);
170 		}
171 		return (zpool_prop_default_numeric(prop));
172 	}
173 
174 	nvl = zhp->zpool_props;
175 	if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
176 		source = fnvlist_lookup_uint64(nv, ZPROP_SOURCE);
177 		value = fnvlist_lookup_uint64(nv, ZPROP_VALUE);
178 	} else {
179 		source = ZPROP_SRC_DEFAULT;
180 		value = zpool_prop_default_numeric(prop);
181 	}
182 
183 	if (src)
184 		*src = source;
185 
186 	return (value);
187 }
188 
189 /*
190  * Map VDEV STATE to printed strings.
191  */
192 const char *
zpool_state_to_name(vdev_state_t state,vdev_aux_t aux)193 zpool_state_to_name(vdev_state_t state, vdev_aux_t aux)
194 {
195 	switch (state) {
196 	case VDEV_STATE_CLOSED:
197 	case VDEV_STATE_OFFLINE:
198 		return (gettext("OFFLINE"));
199 	case VDEV_STATE_REMOVED:
200 		return (gettext("REMOVED"));
201 	case VDEV_STATE_CANT_OPEN:
202 		if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG)
203 			return (gettext("FAULTED"));
204 		else if (aux == VDEV_AUX_SPLIT_POOL)
205 			return (gettext("SPLIT"));
206 		else
207 			return (gettext("UNAVAIL"));
208 	case VDEV_STATE_FAULTED:
209 		return (gettext("FAULTED"));
210 	case VDEV_STATE_DEGRADED:
211 		return (gettext("DEGRADED"));
212 	case VDEV_STATE_HEALTHY:
213 		return (gettext("ONLINE"));
214 
215 	default:
216 		break;
217 	}
218 
219 	return (gettext("UNKNOWN"));
220 }
221 
222 /*
223  * Map POOL STATE to printed strings.
224  */
225 const char *
zpool_pool_state_to_name(pool_state_t state)226 zpool_pool_state_to_name(pool_state_t state)
227 {
228 	switch (state) {
229 	default:
230 		break;
231 	case POOL_STATE_ACTIVE:
232 		return (gettext("ACTIVE"));
233 	case POOL_STATE_EXPORTED:
234 		return (gettext("EXPORTED"));
235 	case POOL_STATE_DESTROYED:
236 		return (gettext("DESTROYED"));
237 	case POOL_STATE_SPARE:
238 		return (gettext("SPARE"));
239 	case POOL_STATE_L2CACHE:
240 		return (gettext("L2CACHE"));
241 	case POOL_STATE_UNINITIALIZED:
242 		return (gettext("UNINITIALIZED"));
243 	case POOL_STATE_UNAVAIL:
244 		return (gettext("UNAVAIL"));
245 	case POOL_STATE_POTENTIALLY_ACTIVE:
246 		return (gettext("POTENTIALLY_ACTIVE"));
247 	}
248 
249 	return (gettext("UNKNOWN"));
250 }
251 
252 /*
253  * Given a pool handle, return the pool health string ("ONLINE", "DEGRADED",
254  * "SUSPENDED", etc).
255  */
256 const char *
zpool_get_state_str(zpool_handle_t * zhp)257 zpool_get_state_str(zpool_handle_t *zhp)
258 {
259 	zpool_errata_t errata;
260 	zpool_status_t status;
261 	const char *str;
262 
263 	status = zpool_get_status(zhp, NULL, &errata);
264 
265 	if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
266 		str = gettext("FAULTED");
267 	} else if (status == ZPOOL_STATUS_IO_FAILURE_WAIT ||
268 	    status == ZPOOL_STATUS_IO_FAILURE_CONTINUE ||
269 	    status == ZPOOL_STATUS_IO_FAILURE_MMP) {
270 		str = gettext("SUSPENDED");
271 	} else {
272 		nvlist_t *nvroot = fnvlist_lookup_nvlist(
273 		    zpool_get_config(zhp, NULL), ZPOOL_CONFIG_VDEV_TREE);
274 		uint_t vsc;
275 		vdev_stat_t *vs = (vdev_stat_t *)fnvlist_lookup_uint64_array(
276 		    nvroot, ZPOOL_CONFIG_VDEV_STATS, &vsc);
277 		str = zpool_state_to_name(vs->vs_state, vs->vs_aux);
278 	}
279 	return (str);
280 }
281 
282 /*
283  * Get a zpool property value for 'prop' and return the value in
284  * a pre-allocated buffer.
285  */
286 int
zpool_get_prop(zpool_handle_t * zhp,zpool_prop_t prop,char * buf,size_t len,zprop_source_t * srctype,boolean_t literal)287 zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf,
288     size_t len, zprop_source_t *srctype, boolean_t literal)
289 {
290 	uint64_t intval;
291 	const char *strval;
292 	zprop_source_t src = ZPROP_SRC_NONE;
293 
294 	if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
295 		switch (prop) {
296 		case ZPOOL_PROP_NAME:
297 			(void) strlcpy(buf, zpool_get_name(zhp), len);
298 			break;
299 
300 		case ZPOOL_PROP_HEALTH:
301 			(void) strlcpy(buf, zpool_get_state_str(zhp), len);
302 			break;
303 
304 		case ZPOOL_PROP_GUID:
305 			intval = zpool_get_prop_int(zhp, prop, &src);
306 			(void) snprintf(buf, len, "%llu", (u_longlong_t)intval);
307 			break;
308 
309 		case ZPOOL_PROP_ALTROOT:
310 		case ZPOOL_PROP_CACHEFILE:
311 		case ZPOOL_PROP_COMMENT:
312 		case ZPOOL_PROP_COMPATIBILITY:
313 			if (zhp->zpool_props != NULL ||
314 			    zpool_get_all_props(zhp) == 0) {
315 				(void) strlcpy(buf,
316 				    zpool_get_prop_string(zhp, prop, &src),
317 				    len);
318 				break;
319 			}
320 			zfs_fallthrough;
321 		default:
322 			(void) strlcpy(buf, "-", len);
323 			break;
324 		}
325 
326 		if (srctype != NULL)
327 			*srctype = src;
328 		return (0);
329 	}
330 
331 	/*
332 	 * ZPOOL_PROP_DEDUPCACHED can be fetched by name only using
333 	 * the ZPOOL_GET_PROPS_NAMES mechanism
334 	 */
335 	if (prop == ZPOOL_PROP_DEDUPCACHED) {
336 		zpool_add_propname(zhp, ZPOOL_DEDUPCACHED_PROP_NAME);
337 		(void) zpool_props_refresh(zhp);
338 	}
339 
340 	if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) &&
341 	    prop != ZPOOL_PROP_NAME)
342 		return (-1);
343 
344 	switch (zpool_prop_get_type(prop)) {
345 	case PROP_TYPE_STRING:
346 		(void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src),
347 		    len);
348 		break;
349 
350 	case PROP_TYPE_NUMBER:
351 		intval = zpool_get_prop_int(zhp, prop, &src);
352 
353 		switch (prop) {
354 		case ZPOOL_PROP_DEDUP_TABLE_QUOTA:
355 			/*
356 			 * If dedup quota is 0, we translate this into 'none'
357 			 * (unless literal is set). And if it is UINT64_MAX
358 			 * we translate that as 'automatic' (limit to size of
359 			 * the dedicated dedup VDEV.  Otherwise, fall throught
360 			 * into the regular number formating.
361 			 */
362 			if (intval == 0) {
363 				(void) strlcpy(buf, literal ? "0" : "none",
364 				    len);
365 				break;
366 			} else if (intval == UINT64_MAX) {
367 				(void) strlcpy(buf, "auto", len);
368 				break;
369 			}
370 			zfs_fallthrough;
371 
372 		case ZPOOL_PROP_SIZE:
373 		case ZPOOL_PROP_NORMAL_SIZE:
374 		case ZPOOL_PROP_SPECIAL_SIZE:
375 		case ZPOOL_PROP_DEDUP_SIZE:
376 		case ZPOOL_PROP_LOG_SIZE:
377 		case ZPOOL_PROP_ELOG_SIZE:
378 		case ZPOOL_PROP_SELOG_SIZE:
379 		case ZPOOL_PROP_ALLOCATED:
380 		case ZPOOL_PROP_NORMAL_ALLOCATED:
381 		case ZPOOL_PROP_SPECIAL_ALLOCATED:
382 		case ZPOOL_PROP_DEDUP_ALLOCATED:
383 		case ZPOOL_PROP_LOG_ALLOCATED:
384 		case ZPOOL_PROP_ELOG_ALLOCATED:
385 		case ZPOOL_PROP_SELOG_ALLOCATED:
386 		case ZPOOL_PROP_AVAILABLE:
387 		case ZPOOL_PROP_NORMAL_AVAILABLE:
388 		case ZPOOL_PROP_SPECIAL_AVAILABLE:
389 		case ZPOOL_PROP_DEDUP_AVAILABLE:
390 		case ZPOOL_PROP_LOG_AVAILABLE:
391 		case ZPOOL_PROP_ELOG_AVAILABLE:
392 		case ZPOOL_PROP_SELOG_AVAILABLE:
393 		case ZPOOL_PROP_FREE:
394 		case ZPOOL_PROP_NORMAL_FREE:
395 		case ZPOOL_PROP_SPECIAL_FREE:
396 		case ZPOOL_PROP_DEDUP_FREE:
397 		case ZPOOL_PROP_LOG_FREE:
398 		case ZPOOL_PROP_ELOG_FREE:
399 		case ZPOOL_PROP_SELOG_FREE:
400 		case ZPOOL_PROP_USABLE:
401 		case ZPOOL_PROP_NORMAL_USABLE:
402 		case ZPOOL_PROP_SPECIAL_USABLE:
403 		case ZPOOL_PROP_DEDUP_USABLE:
404 		case ZPOOL_PROP_LOG_USABLE:
405 		case ZPOOL_PROP_ELOG_USABLE:
406 		case ZPOOL_PROP_SELOG_USABLE:
407 		case ZPOOL_PROP_USED:
408 		case ZPOOL_PROP_NORMAL_USED:
409 		case ZPOOL_PROP_SPECIAL_USED:
410 		case ZPOOL_PROP_DEDUP_USED:
411 		case ZPOOL_PROP_LOG_USED:
412 		case ZPOOL_PROP_ELOG_USED:
413 		case ZPOOL_PROP_SELOG_USED:
414 		case ZPOOL_PROP_FREEING:
415 		case ZPOOL_PROP_LEAKED:
416 		case ZPOOL_PROP_ASHIFT:
417 		case ZPOOL_PROP_MAXBLOCKSIZE:
418 		case ZPOOL_PROP_MAXDNODESIZE:
419 		case ZPOOL_PROP_BCLONESAVED:
420 		case ZPOOL_PROP_BCLONEUSED:
421 		case ZPOOL_PROP_DEDUP_TABLE_SIZE:
422 		case ZPOOL_PROP_DEDUPUSED:
423 		case ZPOOL_PROP_DEDUPSAVED:
424 		case ZPOOL_PROP_DEDUPCACHED:
425 			if (literal)
426 				(void) snprintf(buf, len, "%llu",
427 				    (u_longlong_t)intval);
428 			else
429 				(void) zfs_nicenum(intval, buf, len);
430 			break;
431 
432 		case ZPOOL_PROP_EXPANDSZ:
433 		case ZPOOL_PROP_NORMAL_EXPANDSZ:
434 		case ZPOOL_PROP_SPECIAL_EXPANDSZ:
435 		case ZPOOL_PROP_DEDUP_EXPANDSZ:
436 		case ZPOOL_PROP_LOG_EXPANDSZ:
437 		case ZPOOL_PROP_ELOG_EXPANDSZ:
438 		case ZPOOL_PROP_SELOG_EXPANDSZ:
439 		case ZPOOL_PROP_CHECKPOINT:
440 			if (intval == 0) {
441 				(void) strlcpy(buf, "-", len);
442 			} else if (literal) {
443 				(void) snprintf(buf, len, "%llu",
444 				    (u_longlong_t)intval);
445 			} else {
446 				(void) zfs_nicebytes(intval, buf, len);
447 			}
448 			break;
449 
450 		case ZPOOL_PROP_CAPACITY:
451 		case ZPOOL_PROP_NORMAL_CAPACITY:
452 		case ZPOOL_PROP_SPECIAL_CAPACITY:
453 		case ZPOOL_PROP_DEDUP_CAPACITY:
454 		case ZPOOL_PROP_LOG_CAPACITY:
455 		case ZPOOL_PROP_ELOG_CAPACITY:
456 		case ZPOOL_PROP_SELOG_CAPACITY:
457 			if (literal) {
458 				(void) snprintf(buf, len, "%llu",
459 				    (u_longlong_t)intval);
460 			} else {
461 				(void) snprintf(buf, len, "%llu%%",
462 				    (u_longlong_t)intval);
463 			}
464 			break;
465 
466 		case ZPOOL_PROP_FRAGMENTATION:
467 		case ZPOOL_PROP_NORMAL_FRAGMENTATION:
468 		case ZPOOL_PROP_SPECIAL_FRAGMENTATION:
469 		case ZPOOL_PROP_DEDUP_FRAGMENTATION:
470 		case ZPOOL_PROP_LOG_FRAGMENTATION:
471 		case ZPOOL_PROP_ELOG_FRAGMENTATION:
472 		case ZPOOL_PROP_SELOG_FRAGMENTATION:
473 			if (intval == ZFS_FRAG_INVALID) {
474 				(void) strlcpy(buf, "-", len);
475 			} else if (literal) {
476 				(void) snprintf(buf, len, "%llu",
477 				    (u_longlong_t)intval);
478 			} else {
479 				(void) snprintf(buf, len, "%llu%%",
480 				    (u_longlong_t)intval);
481 			}
482 			break;
483 
484 		case ZPOOL_PROP_BCLONERATIO:
485 		case ZPOOL_PROP_DEDUPRATIO:
486 			if (literal)
487 				(void) snprintf(buf, len, "%llu.%02llu",
488 				    (u_longlong_t)(intval / 100),
489 				    (u_longlong_t)(intval % 100));
490 			else
491 				(void) snprintf(buf, len, "%llu.%02llux",
492 				    (u_longlong_t)(intval / 100),
493 				    (u_longlong_t)(intval % 100));
494 			break;
495 
496 		case ZPOOL_PROP_HEALTH:
497 			(void) strlcpy(buf, zpool_get_state_str(zhp), len);
498 			break;
499 		case ZPOOL_PROP_VERSION:
500 			if (intval >= SPA_VERSION_FEATURES) {
501 				(void) snprintf(buf, len, "-");
502 				break;
503 			}
504 			zfs_fallthrough;
505 		default:
506 			(void) snprintf(buf, len, "%llu", (u_longlong_t)intval);
507 		}
508 		break;
509 
510 	case PROP_TYPE_INDEX:
511 		intval = zpool_get_prop_int(zhp, prop, &src);
512 		if (zpool_prop_index_to_string(prop, intval, &strval)
513 		    != 0)
514 			return (-1);
515 		(void) strlcpy(buf, strval, len);
516 		break;
517 
518 	default:
519 		abort();
520 	}
521 
522 	if (srctype)
523 		*srctype = src;
524 
525 	return (0);
526 }
527 
528 /*
529  * Get a zpool property value for 'propname' and return the value in
530  * a pre-allocated buffer.
531  */
532 int
zpool_get_userprop(zpool_handle_t * zhp,const char * propname,char * buf,size_t len,zprop_source_t * srctype)533 zpool_get_userprop(zpool_handle_t *zhp, const char *propname, char *buf,
534     size_t len, zprop_source_t *srctype)
535 {
536 	nvlist_t *nv;
537 	uint64_t ival;
538 	const char *value;
539 	zprop_source_t source = ZPROP_SRC_LOCAL;
540 
541 	if (zhp->zpool_props == NULL)
542 		zpool_get_all_props(zhp);
543 
544 	if (nvlist_lookup_nvlist(zhp->zpool_props, propname, &nv) == 0) {
545 		if (nvlist_lookup_uint64(nv, ZPROP_SOURCE, &ival) == 0)
546 			source = ival;
547 		verify(nvlist_lookup_string(nv, ZPROP_VALUE, &value) == 0);
548 	} else {
549 		source = ZPROP_SRC_DEFAULT;
550 		value = "-";
551 	}
552 
553 	if (srctype)
554 		*srctype = source;
555 
556 	(void) strlcpy(buf, value, len);
557 
558 	return (0);
559 }
560 
561 /*
562  * Check if the bootfs name has the same pool name as it is set to.
563  * Assuming bootfs is a valid dataset name.
564  */
565 static boolean_t
bootfs_name_valid(const char * pool,const char * bootfs)566 bootfs_name_valid(const char *pool, const char *bootfs)
567 {
568 	int len = strlen(pool);
569 	if (bootfs[0] == '\0')
570 		return (B_TRUE);
571 
572 	if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT))
573 		return (B_FALSE);
574 
575 	if (strncmp(pool, bootfs, len) == 0 &&
576 	    (bootfs[len] == '/' || bootfs[len] == '\0'))
577 		return (B_TRUE);
578 
579 	return (B_FALSE);
580 }
581 
582 /*
583  * Given an nvlist of zpool properties to be set, validate that they are
584  * correct, and parse any numeric properties (index, boolean, etc) if they are
585  * specified as strings.
586  */
587 static nvlist_t *
zpool_valid_proplist(libzfs_handle_t * hdl,const char * poolname,nvlist_t * props,uint64_t version,prop_flags_t flags,char * errbuf)588 zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname,
589     nvlist_t *props, uint64_t version, prop_flags_t flags, char *errbuf)
590 {
591 	nvpair_t *elem;
592 	nvlist_t *retprops;
593 	zpool_prop_t prop;
594 	const char *strval;
595 	uint64_t intval;
596 	const char *check;
597 	struct stat64 statbuf;
598 	zpool_handle_t *zhp;
599 	char *parent, *slash;
600 	char report[1024];
601 
602 	if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) {
603 		(void) no_memory(hdl);
604 		return (NULL);
605 	}
606 
607 	elem = NULL;
608 	while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
609 		const char *propname = nvpair_name(elem);
610 
611 		if (flags.vdevprop && zpool_prop_vdev(propname)) {
612 			vdev_prop_t vprop = vdev_name_to_prop(propname);
613 
614 			if (vdev_prop_readonly(vprop)) {
615 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
616 				    "is readonly"), propname);
617 				(void) zfs_error(hdl, EZFS_PROPREADONLY,
618 				    errbuf);
619 				goto error;
620 			}
621 
622 			if (zprop_parse_value(hdl, elem, vprop, ZFS_TYPE_VDEV,
623 			    retprops, &strval, &intval, errbuf) != 0)
624 				goto error;
625 
626 			continue;
627 		} else if (flags.vdevprop && vdev_prop_user(propname)) {
628 			if (nvlist_add_nvpair(retprops, elem) != 0) {
629 				(void) no_memory(hdl);
630 				goto error;
631 			}
632 			continue;
633 		} else if (flags.vdevprop) {
634 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
635 			    "invalid property: '%s'"), propname);
636 			(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
637 			goto error;
638 		}
639 
640 		prop = zpool_name_to_prop(propname);
641 		if (prop == ZPOOL_PROP_INVAL && zpool_prop_feature(propname)) {
642 			int err;
643 			char *fname = strchr(propname, '@') + 1;
644 
645 			err = zfeature_lookup_name(fname, NULL);
646 			if (err != 0) {
647 				ASSERT3U(err, ==, ENOENT);
648 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
649 				    "feature '%s' unsupported by kernel"),
650 				    fname);
651 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
652 				goto error;
653 			}
654 
655 			if (nvpair_type(elem) != DATA_TYPE_STRING) {
656 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
657 				    "'%s' must be a string"), propname);
658 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
659 				goto error;
660 			}
661 
662 			(void) nvpair_value_string(elem, &strval);
663 			if (strcmp(strval, ZFS_FEATURE_ENABLED) != 0 &&
664 			    strcmp(strval, ZFS_FEATURE_DISABLED) != 0) {
665 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
666 				    "property '%s' can only be set to "
667 				    "'enabled' or 'disabled'"), propname);
668 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
669 				goto error;
670 			}
671 
672 			if (!flags.create &&
673 			    strcmp(strval, ZFS_FEATURE_DISABLED) == 0) {
674 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
675 				    "property '%s' can only be set to "
676 				    "'disabled' at creation time"), propname);
677 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
678 				goto error;
679 			}
680 
681 			if (nvlist_add_uint64(retprops, propname, 0) != 0) {
682 				(void) no_memory(hdl);
683 				goto error;
684 			}
685 			continue;
686 		} else if (prop == ZPOOL_PROP_INVAL &&
687 		    zfs_prop_user(propname)) {
688 			/*
689 			 * This is a user property: make sure it's a
690 			 * string, and that it's less than ZAP_MAXNAMELEN.
691 			 */
692 			if (nvpair_type(elem) != DATA_TYPE_STRING) {
693 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
694 				    "'%s' must be a string"), propname);
695 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
696 				goto error;
697 			}
698 
699 			if (strlen(nvpair_name(elem)) >= ZAP_MAXNAMELEN) {
700 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
701 				    "property name '%s' is too long"),
702 				    propname);
703 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
704 				goto error;
705 			}
706 
707 			(void) nvpair_value_string(elem, &strval);
708 
709 			if (strlen(strval) >= ZFS_MAXPROPLEN) {
710 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
711 				    "property value '%s' is too long"),
712 				    strval);
713 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
714 				goto error;
715 			}
716 
717 			if (nvlist_add_string(retprops, propname,
718 			    strval) != 0) {
719 				(void) no_memory(hdl);
720 				goto error;
721 			}
722 
723 			continue;
724 		}
725 
726 		/*
727 		 * Make sure this property is valid and applies to this type.
728 		 */
729 		if (prop == ZPOOL_PROP_INVAL) {
730 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
731 			    "invalid property '%s'"), propname);
732 			(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
733 			goto error;
734 		}
735 
736 		if (zpool_prop_readonly(prop)) {
737 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
738 			    "is readonly"), propname);
739 			(void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
740 			goto error;
741 		}
742 
743 		if (!flags.create && zpool_prop_setonce(prop)) {
744 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
745 			    "property '%s' can only be set at "
746 			    "creation time"), propname);
747 			(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
748 			goto error;
749 		}
750 
751 		if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops,
752 		    &strval, &intval, errbuf) != 0)
753 			goto error;
754 
755 		/*
756 		 * Perform additional checking for specific properties.
757 		 */
758 		switch (prop) {
759 		case ZPOOL_PROP_VERSION:
760 			if (intval < version ||
761 			    !SPA_VERSION_IS_SUPPORTED(intval)) {
762 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
763 				    "property '%s' number %llu is invalid."),
764 				    propname, (unsigned long long)intval);
765 				(void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
766 				goto error;
767 			}
768 			break;
769 
770 		case ZPOOL_PROP_ASHIFT:
771 			if (intval != 0 &&
772 			    (intval < ASHIFT_MIN || intval > ASHIFT_MAX)) {
773 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
774 				    "property '%s' number %llu is invalid, "
775 				    "only values between %" PRId32 " and %"
776 				    PRId32 " are allowed."),
777 				    propname, (unsigned long long)intval,
778 				    ASHIFT_MIN, ASHIFT_MAX);
779 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
780 				goto error;
781 			}
782 			break;
783 
784 		case ZPOOL_PROP_BOOTFS:
785 			if (flags.create || flags.import) {
786 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
787 				    "property '%s' cannot be set at creation "
788 				    "or import time"), propname);
789 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
790 				goto error;
791 			}
792 
793 			if (version < SPA_VERSION_BOOTFS) {
794 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
795 				    "pool must be upgraded to support "
796 				    "'%s' property"), propname);
797 				(void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
798 				goto error;
799 			}
800 
801 			/*
802 			 * bootfs property value has to be a dataset name and
803 			 * the dataset has to be in the same pool as it sets to.
804 			 */
805 			if (!bootfs_name_valid(poolname, strval)) {
806 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
807 				    "is an invalid name"), strval);
808 				(void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
809 				goto error;
810 			}
811 
812 			if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) {
813 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
814 				    "could not open pool '%s'"), poolname);
815 				(void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
816 				goto error;
817 			}
818 			zpool_close(zhp);
819 			break;
820 
821 		case ZPOOL_PROP_ALTROOT:
822 			if (!flags.create && !flags.import) {
823 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
824 				    "property '%s' can only be set during pool "
825 				    "creation or import"), propname);
826 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
827 				goto error;
828 			}
829 
830 			if (strval[0] != '/') {
831 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
832 				    "bad alternate root '%s'"), strval);
833 				(void) zfs_error(hdl, EZFS_BADPATH, errbuf);
834 				goto error;
835 			}
836 			break;
837 
838 		case ZPOOL_PROP_CACHEFILE:
839 			if (strval[0] == '\0')
840 				break;
841 
842 			if (strcmp(strval, "none") == 0)
843 				break;
844 
845 			if (strval[0] != '/') {
846 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
847 				    "property '%s' must be empty, an "
848 				    "absolute path, or 'none'"), propname);
849 				(void) zfs_error(hdl, EZFS_BADPATH, errbuf);
850 				goto error;
851 			}
852 
853 			parent = strdup(strval);
854 			if (parent == NULL) {
855 				(void) zfs_error(hdl, EZFS_NOMEM, errbuf);
856 				goto error;
857 			}
858 			slash = strrchr(parent, '/');
859 
860 			if (slash[1] == '\0' || strcmp(slash, "/.") == 0 ||
861 			    strcmp(slash, "/..") == 0) {
862 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
863 				    "'%s' is not a valid file"), parent);
864 				(void) zfs_error(hdl, EZFS_BADPATH, errbuf);
865 				free(parent);
866 				goto error;
867 			}
868 
869 			*slash = '\0';
870 
871 			if (parent[0] != '\0' &&
872 			    (stat64(parent, &statbuf) != 0 ||
873 			    !S_ISDIR(statbuf.st_mode))) {
874 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
875 				    "'%s' is not a valid directory"),
876 				    parent);
877 				(void) zfs_error(hdl, EZFS_BADPATH, errbuf);
878 				free(parent);
879 				goto error;
880 			}
881 			free(parent);
882 
883 			break;
884 
885 		case ZPOOL_PROP_COMPATIBILITY:
886 			switch (zpool_load_compat(strval, NULL, report, 1024)) {
887 			case ZPOOL_COMPATIBILITY_OK:
888 			case ZPOOL_COMPATIBILITY_WARNTOKEN:
889 				break;
890 			case ZPOOL_COMPATIBILITY_BADFILE:
891 			case ZPOOL_COMPATIBILITY_BADTOKEN:
892 			case ZPOOL_COMPATIBILITY_NOFILES:
893 				zfs_error_aux(hdl, "%s", report);
894 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
895 				goto error;
896 			}
897 			break;
898 
899 		case ZPOOL_PROP_COMMENT:
900 			for (check = strval; *check != '\0'; check++) {
901 				if (!isprint(*check)) {
902 					zfs_error_aux(hdl,
903 					    dgettext(TEXT_DOMAIN,
904 					    "comment may only have printable "
905 					    "characters"));
906 					(void) zfs_error(hdl, EZFS_BADPROP,
907 					    errbuf);
908 					goto error;
909 				}
910 			}
911 			if (strlen(strval) > ZPROP_MAX_COMMENT) {
912 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
913 				    "comment must not exceed %d characters"),
914 				    ZPROP_MAX_COMMENT);
915 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
916 				goto error;
917 			}
918 			break;
919 		case ZPOOL_PROP_READONLY:
920 			if (!flags.import) {
921 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
922 				    "property '%s' can only be set at "
923 				    "import time"), propname);
924 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
925 				goto error;
926 			}
927 			break;
928 		case ZPOOL_PROP_MULTIHOST:
929 			if (get_system_hostid() == 0) {
930 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
931 				    "requires a non-zero system hostid"));
932 				(void) zfs_error(hdl, EZFS_BADPROP, errbuf);
933 				goto error;
934 			}
935 			break;
936 		case ZPOOL_PROP_DEDUPDITTO:
937 			printf("Note: property '%s' no longer has "
938 			    "any effect\n", propname);
939 			break;
940 
941 		default:
942 			break;
943 		}
944 	}
945 
946 	return (retprops);
947 error:
948 	nvlist_free(retprops);
949 	return (NULL);
950 }
951 
952 /*
953  * Set zpool property : propname=propval.
954  */
955 int
zpool_set_prop(zpool_handle_t * zhp,const char * propname,const char * propval)956 zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval)
957 {
958 	zfs_cmd_t zc = {"\0"};
959 	int ret;
960 	char errbuf[ERRBUFLEN];
961 	nvlist_t *nvl = NULL;
962 	nvlist_t *realprops;
963 	uint64_t version;
964 	prop_flags_t flags = { 0 };
965 
966 	(void) snprintf(errbuf, sizeof (errbuf),
967 	    dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
968 	    zhp->zpool_name);
969 
970 	if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
971 		return (no_memory(zhp->zpool_hdl));
972 
973 	if (nvlist_add_string(nvl, propname, propval) != 0) {
974 		nvlist_free(nvl);
975 		return (no_memory(zhp->zpool_hdl));
976 	}
977 
978 	version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
979 	if ((realprops = zpool_valid_proplist(zhp->zpool_hdl,
980 	    zhp->zpool_name, nvl, version, flags, errbuf)) == NULL) {
981 		nvlist_free(nvl);
982 		return (-1);
983 	}
984 
985 	nvlist_free(nvl);
986 	nvl = realprops;
987 
988 	/*
989 	 * Execute the corresponding ioctl() to set this property.
990 	 */
991 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
992 
993 	zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl);
994 
995 	ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc);
996 
997 	zcmd_free_nvlists(&zc);
998 	nvlist_free(nvl);
999 
1000 	if (ret)
1001 		(void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
1002 	else
1003 		(void) zpool_props_refresh(zhp);
1004 
1005 	return (ret);
1006 }
1007 
1008 int
zpool_expand_proplist(zpool_handle_t * zhp,zprop_list_t ** plp,zfs_type_t type,boolean_t literal)1009 zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp,
1010     zfs_type_t type, boolean_t literal)
1011 {
1012 	libzfs_handle_t *hdl = zhp->zpool_hdl;
1013 	zprop_list_t *entry;
1014 	char buf[ZFS_MAXPROPLEN];
1015 	nvlist_t *features = NULL;
1016 	nvpair_t *nvp;
1017 	zprop_list_t **last;
1018 	boolean_t firstexpand = (NULL == *plp);
1019 	int i;
1020 
1021 	if (zprop_expand_list(hdl, plp, type) != 0)
1022 		return (-1);
1023 
1024 	if (type == ZFS_TYPE_VDEV)
1025 		return (0);
1026 
1027 	last = plp;
1028 	while (*last != NULL)
1029 		last = &(*last)->pl_next;
1030 
1031 	if ((*plp)->pl_all)
1032 		features = zpool_get_features(zhp);
1033 
1034 	if ((*plp)->pl_all && firstexpand) {
1035 		/* Handle userprops in the all properties case */
1036 		if (zhp->zpool_props == NULL && zpool_props_refresh(zhp))
1037 			return (-1);
1038 
1039 		nvp = NULL;
1040 		while ((nvp = nvlist_next_nvpair(zhp->zpool_props, nvp)) !=
1041 		    NULL) {
1042 			const char *propname = nvpair_name(nvp);
1043 
1044 			if (!zfs_prop_user(propname))
1045 				continue;
1046 
1047 			entry = zfs_alloc(hdl, sizeof (zprop_list_t));
1048 			entry->pl_prop = ZPROP_USERPROP;
1049 			entry->pl_user_prop = zfs_strdup(hdl, propname);
1050 			entry->pl_width = strlen(entry->pl_user_prop);
1051 			entry->pl_all = B_TRUE;
1052 
1053 			*last = entry;
1054 			last = &entry->pl_next;
1055 		}
1056 
1057 		for (i = 0; i < SPA_FEATURES; i++) {
1058 			entry = zfs_alloc(hdl, sizeof (zprop_list_t));
1059 			entry->pl_prop = ZPROP_USERPROP;
1060 			entry->pl_user_prop = zfs_asprintf(hdl, "feature@%s",
1061 			    spa_feature_table[i].fi_uname);
1062 			entry->pl_width = strlen(entry->pl_user_prop);
1063 			entry->pl_all = B_TRUE;
1064 
1065 			*last = entry;
1066 			last = &entry->pl_next;
1067 		}
1068 	}
1069 
1070 	/* add any unsupported features */
1071 	for (nvp = nvlist_next_nvpair(features, NULL);
1072 	    nvp != NULL; nvp = nvlist_next_nvpair(features, nvp)) {
1073 		char *propname;
1074 		boolean_t found;
1075 
1076 		if (zfeature_is_supported(nvpair_name(nvp)))
1077 			continue;
1078 
1079 		propname = zfs_asprintf(hdl, "unsupported@%s",
1080 		    nvpair_name(nvp));
1081 
1082 		/*
1083 		 * Before adding the property to the list make sure that no
1084 		 * other pool already added the same property.
1085 		 */
1086 		found = B_FALSE;
1087 		entry = *plp;
1088 		while (entry != NULL) {
1089 			if (entry->pl_user_prop != NULL &&
1090 			    strcmp(propname, entry->pl_user_prop) == 0) {
1091 				found = B_TRUE;
1092 				break;
1093 			}
1094 			entry = entry->pl_next;
1095 		}
1096 		if (found) {
1097 			free(propname);
1098 			continue;
1099 		}
1100 
1101 		entry = zfs_alloc(hdl, sizeof (zprop_list_t));
1102 		entry->pl_prop = ZPROP_USERPROP;
1103 		entry->pl_user_prop = propname;
1104 		entry->pl_width = strlen(entry->pl_user_prop);
1105 		entry->pl_all = B_TRUE;
1106 
1107 		*last = entry;
1108 		last = &entry->pl_next;
1109 	}
1110 
1111 	for (entry = *plp; entry != NULL; entry = entry->pl_next) {
1112 		if (entry->pl_fixed && !literal)
1113 			continue;
1114 
1115 		if (entry->pl_prop != ZPROP_USERPROP &&
1116 		    zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf),
1117 		    NULL, literal) == 0) {
1118 			if (strlen(buf) > entry->pl_width)
1119 				entry->pl_width = strlen(buf);
1120 		} else if (entry->pl_prop == ZPROP_INVAL &&
1121 		    zfs_prop_user(entry->pl_user_prop) &&
1122 		    zpool_get_userprop(zhp, entry->pl_user_prop, buf,
1123 		    sizeof (buf), NULL) == 0) {
1124 			if (strlen(buf) > entry->pl_width)
1125 				entry->pl_width = strlen(buf);
1126 		}
1127 	}
1128 
1129 	return (0);
1130 }
1131 
1132 int
vdev_expand_proplist(zpool_handle_t * zhp,const char * vdevname,zprop_list_t ** plp)1133 vdev_expand_proplist(zpool_handle_t *zhp, const char *vdevname,
1134     zprop_list_t **plp)
1135 {
1136 	zprop_list_t *entry;
1137 	char buf[ZFS_MAXPROPLEN];
1138 	const char *strval = NULL;
1139 	int err = 0;
1140 	nvpair_t *elem = NULL;
1141 	nvlist_t *vprops = NULL;
1142 	nvlist_t *propval = NULL;
1143 	const char *propname;
1144 	vdev_prop_t prop;
1145 	zprop_list_t **last;
1146 
1147 	for (entry = *plp; entry != NULL; entry = entry->pl_next) {
1148 		if (entry->pl_fixed)
1149 			continue;
1150 
1151 		if (zpool_get_vdev_prop(zhp, vdevname, entry->pl_prop,
1152 		    entry->pl_user_prop, buf, sizeof (buf), NULL,
1153 		    B_FALSE) == 0) {
1154 			if (strlen(buf) > entry->pl_width)
1155 				entry->pl_width = strlen(buf);
1156 		}
1157 		if (entry->pl_prop == VDEV_PROP_NAME &&
1158 		    strlen(vdevname) > entry->pl_width)
1159 			entry->pl_width = strlen(vdevname);
1160 	}
1161 
1162 	/* Handle the all properties case */
1163 	last = plp;
1164 	if (*last != NULL && (*last)->pl_all == B_TRUE) {
1165 		while (*last != NULL)
1166 			last = &(*last)->pl_next;
1167 
1168 		err = zpool_get_all_vdev_props(zhp, vdevname, &vprops);
1169 		if (err != 0)
1170 			return (err);
1171 
1172 		while ((elem = nvlist_next_nvpair(vprops, elem)) != NULL) {
1173 			propname = nvpair_name(elem);
1174 
1175 			/* Skip properties that are not user defined */
1176 			if ((prop = vdev_name_to_prop(propname)) !=
1177 			    VDEV_PROP_USERPROP)
1178 				continue;
1179 
1180 			if (nvpair_value_nvlist(elem, &propval) != 0)
1181 				continue;
1182 
1183 			strval = fnvlist_lookup_string(propval, ZPROP_VALUE);
1184 
1185 			entry = zfs_alloc(zhp->zpool_hdl,
1186 			    sizeof (zprop_list_t));
1187 			entry->pl_prop = prop;
1188 			entry->pl_user_prop = zfs_strdup(zhp->zpool_hdl,
1189 			    propname);
1190 			entry->pl_width = strlen(strval);
1191 			entry->pl_all = B_TRUE;
1192 			*last = entry;
1193 			last = &entry->pl_next;
1194 		}
1195 	}
1196 
1197 	return (0);
1198 }
1199 
1200 /*
1201  * Get the state for the given feature on the given ZFS pool.
1202  */
1203 int
zpool_prop_get_feature(zpool_handle_t * zhp,const char * propname,char * buf,size_t len)1204 zpool_prop_get_feature(zpool_handle_t *zhp, const char *propname, char *buf,
1205     size_t len)
1206 {
1207 	uint64_t refcount;
1208 	boolean_t found = B_FALSE;
1209 	nvlist_t *features = zpool_get_features(zhp);
1210 	boolean_t supported;
1211 	const char *feature = strchr(propname, '@') + 1;
1212 
1213 	supported = zpool_prop_feature(propname);
1214 	ASSERT(supported || zpool_prop_unsupported(propname));
1215 
1216 	/*
1217 	 * Convert from feature name to feature guid. This conversion is
1218 	 * unnecessary for unsupported@... properties because they already
1219 	 * use guids.
1220 	 */
1221 	if (supported) {
1222 		int ret;
1223 		spa_feature_t fid;
1224 
1225 		ret = zfeature_lookup_name(feature, &fid);
1226 		if (ret != 0) {
1227 			(void) strlcpy(buf, "-", len);
1228 			return (ENOTSUP);
1229 		}
1230 		feature = spa_feature_table[fid].fi_guid;
1231 	}
1232 
1233 	if (nvlist_lookup_uint64(features, feature, &refcount) == 0)
1234 		found = B_TRUE;
1235 
1236 	if (supported) {
1237 		if (!found) {
1238 			(void) strlcpy(buf, ZFS_FEATURE_DISABLED, len);
1239 		} else  {
1240 			if (refcount == 0)
1241 				(void) strlcpy(buf, ZFS_FEATURE_ENABLED, len);
1242 			else
1243 				(void) strlcpy(buf, ZFS_FEATURE_ACTIVE, len);
1244 		}
1245 	} else {
1246 		if (found) {
1247 			if (refcount == 0) {
1248 				(void) strcpy(buf, ZFS_UNSUPPORTED_INACTIVE);
1249 			} else {
1250 				(void) strcpy(buf, ZFS_UNSUPPORTED_READONLY);
1251 			}
1252 		} else {
1253 			(void) strlcpy(buf, "-", len);
1254 			return (ENOTSUP);
1255 		}
1256 	}
1257 
1258 	return (0);
1259 }
1260 
1261 /*
1262  * Validate the given pool name, optionally putting an extended error message in
1263  * 'buf'.
1264  */
1265 boolean_t
zpool_name_valid(libzfs_handle_t * hdl,boolean_t isopen,const char * pool)1266 zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool)
1267 {
1268 	namecheck_err_t why;
1269 	char what;
1270 	int ret;
1271 
1272 	ret = pool_namecheck(pool, &why, &what);
1273 
1274 	/*
1275 	 * The rules for reserved pool names were extended at a later point.
1276 	 * But we need to support users with existing pools that may now be
1277 	 * invalid.  So we only check for this expanded set of names during a
1278 	 * create (or import), and only in userland.
1279 	 */
1280 	if (ret == 0 && !isopen &&
1281 	    (strncmp(pool, "mirror", 6) == 0 ||
1282 	    strncmp(pool, "raidz", 5) == 0 ||
1283 	    strncmp(pool, "draid", 5) == 0 ||
1284 	    strncmp(pool, "spare", 5) == 0 ||
1285 	    strcmp(pool, "log") == 0)) {
1286 		if (hdl != NULL)
1287 			zfs_error_aux(hdl,
1288 			    dgettext(TEXT_DOMAIN, "name is reserved"));
1289 		return (B_FALSE);
1290 	}
1291 
1292 
1293 	if (ret != 0) {
1294 		if (hdl != NULL) {
1295 			switch (why) {
1296 			case NAME_ERR_TOOLONG:
1297 				zfs_error_aux(hdl,
1298 				    dgettext(TEXT_DOMAIN, "name is too long"));
1299 				break;
1300 
1301 			case NAME_ERR_INVALCHAR:
1302 				zfs_error_aux(hdl,
1303 				    dgettext(TEXT_DOMAIN, "invalid character "
1304 				    "'%c' in pool name"), what);
1305 				break;
1306 
1307 			case NAME_ERR_NOLETTER:
1308 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1309 				    "name must begin with a letter"));
1310 				break;
1311 
1312 			case NAME_ERR_RESERVED:
1313 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1314 				    "name is reserved"));
1315 				break;
1316 
1317 			case NAME_ERR_DISKLIKE:
1318 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1319 				    "pool name is reserved"));
1320 				break;
1321 
1322 			case NAME_ERR_LEADING_SLASH:
1323 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1324 				    "leading slash in name"));
1325 				break;
1326 
1327 			case NAME_ERR_EMPTY_COMPONENT:
1328 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1329 				    "empty component in name"));
1330 				break;
1331 
1332 			case NAME_ERR_TRAILING_SLASH:
1333 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1334 				    "trailing slash in name"));
1335 				break;
1336 
1337 			case NAME_ERR_MULTIPLE_DELIMITERS:
1338 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1339 				    "multiple '@' and/or '#' delimiters in "
1340 				    "name"));
1341 				break;
1342 
1343 			case NAME_ERR_NO_AT:
1344 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1345 				    "permission set is missing '@'"));
1346 				break;
1347 
1348 			default:
1349 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1350 				    "(%d) not defined"), why);
1351 				break;
1352 			}
1353 		}
1354 		return (B_FALSE);
1355 	}
1356 
1357 	return (B_TRUE);
1358 }
1359 
1360 /*
1361  * Open a handle to the given pool, even if the pool is currently in the FAULTED
1362  * state.
1363  */
1364 zpool_handle_t *
zpool_open_canfail(libzfs_handle_t * hdl,const char * pool)1365 zpool_open_canfail(libzfs_handle_t *hdl, const char *pool)
1366 {
1367 	zpool_handle_t *zhp;
1368 	boolean_t missing;
1369 
1370 	/*
1371 	 * Make sure the pool name is valid.
1372 	 */
1373 	if (!zpool_name_valid(hdl, B_TRUE, pool)) {
1374 		(void) zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1375 		    dgettext(TEXT_DOMAIN, "cannot open '%s'"),
1376 		    pool);
1377 		return (NULL);
1378 	}
1379 
1380 	zhp = zfs_alloc(hdl, sizeof (zpool_handle_t));
1381 
1382 	zhp->zpool_hdl = hdl;
1383 	(void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
1384 
1385 	if (zpool_refresh_stats(zhp, &missing) != 0) {
1386 		zpool_close(zhp);
1387 		return (NULL);
1388 	}
1389 
1390 	if (missing) {
1391 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool"));
1392 		(void) zfs_error_fmt(hdl, EZFS_NOENT,
1393 		    dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool);
1394 		zpool_close(zhp);
1395 		return (NULL);
1396 	}
1397 
1398 	return (zhp);
1399 }
1400 
1401 /*
1402  * Like the above, but silent on error.  Used when iterating over pools (because
1403  * the configuration cache may be out of date).
1404  */
1405 int
zpool_open_silent(libzfs_handle_t * hdl,const char * pool,zpool_handle_t ** ret)1406 zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret)
1407 {
1408 	zpool_handle_t *zhp;
1409 	boolean_t missing;
1410 
1411 	zhp = zfs_alloc(hdl, sizeof (zpool_handle_t));
1412 
1413 	zhp->zpool_hdl = hdl;
1414 	(void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
1415 
1416 	if (zpool_refresh_stats(zhp, &missing) != 0) {
1417 		zpool_close(zhp);
1418 		return (-1);
1419 	}
1420 
1421 	if (missing) {
1422 		zpool_close(zhp);
1423 		*ret = NULL;
1424 		return (0);
1425 	}
1426 
1427 	*ret = zhp;
1428 	return (0);
1429 }
1430 
1431 /*
1432  * Similar to zpool_open_canfail(), but refuses to open pools in the faulted
1433  * state.
1434  */
1435 zpool_handle_t *
zpool_open(libzfs_handle_t * hdl,const char * pool)1436 zpool_open(libzfs_handle_t *hdl, const char *pool)
1437 {
1438 	zpool_handle_t *zhp;
1439 
1440 	if ((zhp = zpool_open_canfail(hdl, pool)) == NULL)
1441 		return (NULL);
1442 
1443 	if (zhp->zpool_state == POOL_STATE_UNAVAIL) {
1444 		(void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
1445 		    dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name);
1446 		zpool_close(zhp);
1447 		return (NULL);
1448 	}
1449 
1450 	return (zhp);
1451 }
1452 
1453 /*
1454  * Close the handle.  Simply frees the memory associated with the handle.
1455  */
1456 void
zpool_close(zpool_handle_t * zhp)1457 zpool_close(zpool_handle_t *zhp)
1458 {
1459 	nvlist_free(zhp->zpool_config);
1460 	nvlist_free(zhp->zpool_old_config);
1461 	nvlist_free(zhp->zpool_props);
1462 	free(zhp);
1463 }
1464 
1465 /*
1466  * Return the name of the pool.
1467  */
1468 const char *
zpool_get_name(zpool_handle_t * zhp)1469 zpool_get_name(zpool_handle_t *zhp)
1470 {
1471 	return (zhp->zpool_name);
1472 }
1473 
1474 
1475 /*
1476  * Return the state of the pool (ACTIVE or UNAVAILABLE)
1477  */
1478 int
zpool_get_state(zpool_handle_t * zhp)1479 zpool_get_state(zpool_handle_t *zhp)
1480 {
1481 	return (zhp->zpool_state);
1482 }
1483 
1484 /*
1485  * Check if vdev list contains a dRAID vdev
1486  */
1487 static boolean_t
zpool_has_draid_vdev(nvlist_t * nvroot)1488 zpool_has_draid_vdev(nvlist_t *nvroot)
1489 {
1490 	nvlist_t **child;
1491 	uint_t children;
1492 
1493 	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
1494 	    &child, &children) == 0) {
1495 		for (uint_t c = 0; c < children; c++) {
1496 			const char *type;
1497 
1498 			if (nvlist_lookup_string(child[c],
1499 			    ZPOOL_CONFIG_TYPE, &type) == 0 &&
1500 			    strcmp(type, VDEV_TYPE_DRAID) == 0) {
1501 				return (B_TRUE);
1502 			}
1503 		}
1504 	}
1505 	return (B_FALSE);
1506 }
1507 
1508 /*
1509  * Output a dRAID top-level vdev name in to the provided buffer.
1510  */
1511 static char *
zpool_draid_name(char * name,int len,uint64_t data,uint64_t parity,uint64_t spares,uint64_t children,uint64_t width)1512 zpool_draid_name(char *name, int len, uint64_t data, uint64_t parity,
1513     uint64_t spares, uint64_t children, uint64_t width)
1514 {
1515 	if (children < width)
1516 		snprintf(name, len, "%s%llu:%llud:%lluc:%lluw:%llus",
1517 		    VDEV_TYPE_DRAID, (u_longlong_t)parity, (u_longlong_t)data,
1518 		    (u_longlong_t)children, (u_longlong_t)width,
1519 		    (u_longlong_t)spares);
1520 	else
1521 		snprintf(name, len, "%s%llu:%llud:%lluc:%llus",
1522 		    VDEV_TYPE_DRAID, (u_longlong_t)parity, (u_longlong_t)data,
1523 		    (u_longlong_t)children, (u_longlong_t)spares);
1524 
1525 	return (name);
1526 }
1527 
1528 /*
1529  * Return B_TRUE if the provided name is a dRAID spare name.
1530  */
1531 boolean_t
zpool_is_draid_spare(const char * name)1532 zpool_is_draid_spare(const char *name)
1533 {
1534 	uint64_t spare_id, parity, vdev_id;
1535 
1536 	if (sscanf(name, VDEV_TYPE_DRAID "%llu-%llu-%llu",
1537 	    (u_longlong_t *)&parity, (u_longlong_t *)&vdev_id,
1538 	    (u_longlong_t *)&spare_id) == 3) {
1539 		return (B_TRUE);
1540 	}
1541 
1542 	return (B_FALSE);
1543 }
1544 
1545 
1546 /*
1547  * Extract device-specific error information from a failed pool creation.
1548  * If the kernel returned ZPOOL_CONFIG_CREATE_INFO in the ioctl output,
1549  * set an appropriate error aux message identifying the problematic device.
1550  */
1551 static int
zpool_create_info(libzfs_handle_t * hdl,zfs_cmd_t * zc)1552 zpool_create_info(libzfs_handle_t *hdl, zfs_cmd_t *zc)
1553 {
1554 	nvlist_t *outnv = NULL;
1555 	nvlist_t *info = NULL;
1556 	const char *vdev = NULL;
1557 	const char *pname = NULL;
1558 
1559 	if (zc->zc_nvlist_dst_size == 0)
1560 		return (ENOENT);
1561 
1562 	if (nvlist_unpack((void *)(uintptr_t)zc->zc_nvlist_dst,
1563 	    zc->zc_nvlist_dst_size, &outnv, 0) != 0 || outnv == NULL)
1564 		return (EINVAL);
1565 
1566 	if (nvlist_lookup_nvlist(outnv, ZPOOL_CONFIG_CREATE_INFO, &info) != 0) {
1567 		nvlist_free(outnv);
1568 		return (EINVAL);
1569 	}
1570 
1571 	if (nvlist_lookup_string(info, ZPOOL_CREATE_INFO_VDEV, &vdev) != 0) {
1572 		nvlist_free(outnv);
1573 		return (EINVAL);
1574 	}
1575 
1576 	if (nvlist_lookup_string(info, ZPOOL_CREATE_INFO_POOL, &pname) == 0) {
1577 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1578 		    "device '%s' is part of active pool '%s'"),
1579 		    vdev, pname);
1580 	} else {
1581 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1582 		    "device '%s' is in use"), vdev);
1583 	}
1584 
1585 	nvlist_free(outnv);
1586 	return (0);
1587 }
1588 
1589 /*
1590  * Create the named pool, using the provided vdev list.  It is assumed
1591  * that the consumer has already validated the contents of the nvlist, so we
1592  * don't have to worry about error semantics.
1593  */
1594 int
zpool_create(libzfs_handle_t * hdl,const char * pool,nvlist_t * nvroot,nvlist_t * props,nvlist_t * fsprops)1595 zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot,
1596     nvlist_t *props, nvlist_t *fsprops)
1597 {
1598 	zfs_cmd_t zc = {"\0"};
1599 	nvlist_t *zc_fsprops = NULL;
1600 	nvlist_t *zc_props = NULL;
1601 	nvlist_t *hidden_args = NULL;
1602 	uint8_t *wkeydata = NULL;
1603 	uint_t wkeylen = 0;
1604 	char errbuf[ERRBUFLEN];
1605 	int ret = -1;
1606 
1607 	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1608 	    "cannot create '%s'"), pool);
1609 
1610 	if (!zpool_name_valid(hdl, B_FALSE, pool))
1611 		return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
1612 
1613 	zcmd_write_conf_nvlist(hdl, &zc, nvroot);
1614 
1615 	if (props) {
1616 		prop_flags_t flags = { .create = B_TRUE, .import = B_FALSE };
1617 
1618 		if ((zc_props = zpool_valid_proplist(hdl, pool, props,
1619 		    SPA_VERSION_1, flags, errbuf)) == NULL) {
1620 			goto create_failed;
1621 		}
1622 	}
1623 
1624 	if (fsprops) {
1625 		uint64_t zoned;
1626 		const char *zonestr;
1627 
1628 		zoned = ((nvlist_lookup_string(fsprops,
1629 		    zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) &&
1630 		    strcmp(zonestr, "on") == 0);
1631 
1632 		if ((zc_fsprops = zfs_valid_proplist(hdl, ZFS_TYPE_FILESYSTEM,
1633 		    fsprops, zoned, NULL, NULL, B_TRUE, errbuf)) == NULL) {
1634 			goto create_failed;
1635 		}
1636 
1637 		if (!zc_props &&
1638 		    (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) {
1639 			goto create_failed;
1640 		}
1641 		if (zfs_crypto_create(hdl, NULL, zc_fsprops, props, B_TRUE,
1642 		    &wkeydata, &wkeylen) != 0) {
1643 			zfs_error(hdl, EZFS_CRYPTOFAILED, errbuf);
1644 			goto create_failed;
1645 		}
1646 		if (nvlist_add_nvlist(zc_props,
1647 		    ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) {
1648 			goto create_failed;
1649 		}
1650 		if (wkeydata != NULL) {
1651 			if (nvlist_alloc(&hidden_args, NV_UNIQUE_NAME, 0) != 0)
1652 				goto create_failed;
1653 
1654 			if (nvlist_add_uint8_array(hidden_args, "wkeydata",
1655 			    wkeydata, wkeylen) != 0)
1656 				goto create_failed;
1657 
1658 			if (nvlist_add_nvlist(zc_props, ZPOOL_HIDDEN_ARGS,
1659 			    hidden_args) != 0)
1660 				goto create_failed;
1661 		}
1662 	}
1663 
1664 	if (zc_props)
1665 		zcmd_write_src_nvlist(hdl, &zc, zc_props);
1666 
1667 	(void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name));
1668 	zcmd_alloc_dst_nvlist(hdl, &zc, 4096);
1669 
1670 	if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) {
1671 		switch (errno) {
1672 		case EBUSY:
1673 			/*
1674 			 * This can happen if the user has specified the same
1675 			 * device multiple times.  We can't reliably detect this
1676 			 * until we try to add it and see we already have a
1677 			 * label.  This can also happen under if the device is
1678 			 * part of an active md or lvm device.
1679 			 */
1680 			if (zpool_create_info(hdl, &zc) != 0) {
1681 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1682 				    "one or more vdevs refer to the same "
1683 				    "device, or one of\nthe devices is "
1684 				    "part of an active md or lvm device"));
1685 			}
1686 			ret = zfs_error(hdl, EZFS_BADDEV, errbuf);
1687 			break;
1688 
1689 		case ERANGE:
1690 			/*
1691 			 * This happens if the record size is smaller or larger
1692 			 * than the allowed size range, or not a power of 2.
1693 			 *
1694 			 * NOTE: although zfs_valid_proplist is called earlier,
1695 			 * this case may have slipped through since the
1696 			 * pool does not exist yet and it is therefore
1697 			 * impossible to read properties e.g. max blocksize
1698 			 * from the pool.
1699 			 */
1700 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1701 			    "record size invalid"));
1702 			ret = zfs_error(hdl, EZFS_BADPROP, errbuf);
1703 			break;
1704 
1705 		case EOVERFLOW:
1706 			/*
1707 			 * This occurs when one of the devices is below
1708 			 * SPA_MINDEVSIZE.  Unfortunately, we can't detect which
1709 			 * device was the problem device since there's no
1710 			 * reliable way to determine device size from userland.
1711 			 */
1712 			{
1713 				char buf[64];
1714 
1715 				zfs_nicebytes(SPA_MINDEVSIZE, buf,
1716 				    sizeof (buf));
1717 
1718 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1719 				    "one or more devices is less than the "
1720 				    "minimum size (%s)"), buf);
1721 			}
1722 			ret = zfs_error(hdl, EZFS_BADDEV, errbuf);
1723 			break;
1724 
1725 		case ENOSPC:
1726 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1727 			    "one or more devices is out of space"));
1728 			ret = zfs_error(hdl, EZFS_BADDEV, errbuf);
1729 			break;
1730 
1731 		case EINVAL:
1732 			if (zpool_has_draid_vdev(nvroot) &&
1733 			    zfeature_lookup_name("draid", NULL) != 0) {
1734 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1735 				    "dRAID vdevs are unsupported by the "
1736 				    "kernel"));
1737 				ret = zfs_error(hdl, EZFS_BADDEV, errbuf);
1738 			} else {
1739 				ret = zpool_standard_error(hdl, errno, errbuf);
1740 			}
1741 			break;
1742 
1743 		case ENXIO:
1744 			if (zpool_create_info(hdl, &zc) == 0) {
1745 				ret = zfs_error(hdl, EZFS_BADDEV, errbuf);
1746 			} else {
1747 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1748 				    "one or more devices could not be "
1749 				    "opened"));
1750 				ret = zfs_error(hdl, EZFS_BADDEV, errbuf);
1751 			}
1752 			break;
1753 
1754 		case EDOM:
1755 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1756 			    "block size out of range or does not match"));
1757 			ret = zfs_error(hdl, EZFS_BADDEV, errbuf);
1758 			break;
1759 
1760 		default:
1761 			ret = zpool_standard_error(hdl, errno, errbuf);
1762 			break;
1763 		}
1764 	}
1765 
1766 create_failed:
1767 	zcmd_free_nvlists(&zc);
1768 	nvlist_free(zc_props);
1769 	nvlist_free(zc_fsprops);
1770 	nvlist_free(hidden_args);
1771 	if (wkeydata != NULL)
1772 		free(wkeydata);
1773 	return (ret);
1774 }
1775 
1776 /*
1777  * Destroy the given pool.  It is up to the caller to ensure that there are no
1778  * datasets left in the pool.
1779  */
1780 int
zpool_destroy(zpool_handle_t * zhp,const char * log_str)1781 zpool_destroy(zpool_handle_t *zhp, const char *log_str)
1782 {
1783 	zfs_cmd_t zc = {"\0"};
1784 	zfs_handle_t *zfp = NULL;
1785 	libzfs_handle_t *hdl = zhp->zpool_hdl;
1786 	char errbuf[ERRBUFLEN];
1787 
1788 	if (zhp->zpool_state == POOL_STATE_ACTIVE &&
1789 	    (zfp = zfs_open(hdl, zhp->zpool_name, ZFS_TYPE_FILESYSTEM)) == NULL)
1790 		return (-1);
1791 
1792 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1793 	zc.zc_history = (uint64_t)(uintptr_t)log_str;
1794 
1795 	if (zfs_ioctl(hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) {
1796 		(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1797 		    "cannot destroy '%s'"), zhp->zpool_name);
1798 
1799 		if (errno == EROFS) {
1800 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1801 			    "one or more devices is read only"));
1802 			(void) zfs_error(hdl, EZFS_BADDEV, errbuf);
1803 		} else {
1804 			(void) zpool_standard_error(hdl, errno, errbuf);
1805 		}
1806 
1807 		if (zfp)
1808 			zfs_close(zfp);
1809 		return (-1);
1810 	}
1811 
1812 	if (zfp) {
1813 		remove_mountpoint(zfp);
1814 		zfs_close(zfp);
1815 	}
1816 
1817 	return (0);
1818 }
1819 
1820 /*
1821  * Create a checkpoint in the given pool.
1822  */
1823 int
zpool_checkpoint(zpool_handle_t * zhp)1824 zpool_checkpoint(zpool_handle_t *zhp)
1825 {
1826 	libzfs_handle_t *hdl = zhp->zpool_hdl;
1827 	char errbuf[ERRBUFLEN];
1828 	int error;
1829 
1830 	error = lzc_pool_checkpoint(zhp->zpool_name);
1831 	if (error != 0) {
1832 		(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1833 		    "cannot checkpoint '%s'"), zhp->zpool_name);
1834 		(void) zpool_standard_error(hdl, error, errbuf);
1835 		return (-1);
1836 	}
1837 
1838 	return (0);
1839 }
1840 
1841 /*
1842  * Discard the checkpoint from the given pool.
1843  */
1844 int
zpool_discard_checkpoint(zpool_handle_t * zhp)1845 zpool_discard_checkpoint(zpool_handle_t *zhp)
1846 {
1847 	libzfs_handle_t *hdl = zhp->zpool_hdl;
1848 	char errbuf[ERRBUFLEN];
1849 	int error;
1850 
1851 	error = lzc_pool_checkpoint_discard(zhp->zpool_name);
1852 	if (error != 0) {
1853 		(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1854 		    "cannot discard checkpoint in '%s'"), zhp->zpool_name);
1855 		(void) zpool_standard_error(hdl, error, errbuf);
1856 		return (-1);
1857 	}
1858 
1859 	return (0);
1860 }
1861 
1862 /*
1863  * Load data type for the given pool.
1864  */
1865 int
zpool_prefetch(zpool_handle_t * zhp,zpool_prefetch_type_t type)1866 zpool_prefetch(zpool_handle_t *zhp, zpool_prefetch_type_t type)
1867 {
1868 	libzfs_handle_t *hdl = zhp->zpool_hdl;
1869 	char msg[1024];
1870 	int error;
1871 
1872 	error = lzc_pool_prefetch(zhp->zpool_name, type);
1873 	if (error != 0) {
1874 		const char *typename = "unknown";
1875 		if (type == ZPOOL_PREFETCH_DDT)
1876 			typename = "ddt";
1877 		else if (type == ZPOOL_PREFETCH_BRT)
1878 			typename = "brt";
1879 		(void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1880 		    "cannot prefetch %s in '%s'"), typename, zhp->zpool_name);
1881 		(void) zpool_standard_error(hdl, error, msg);
1882 		return (-1);
1883 	}
1884 
1885 	return (0);
1886 }
1887 
1888 /*
1889  * Add the given vdevs to the pool.  The caller must have already performed the
1890  * necessary verification to ensure that the vdev specification is well-formed.
1891  */
1892 int
zpool_add(zpool_handle_t * zhp,nvlist_t * nvroot,boolean_t check_ashift)1893 zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot, boolean_t check_ashift)
1894 {
1895 	zfs_cmd_t zc = {"\0"};
1896 	int ret;
1897 	libzfs_handle_t *hdl = zhp->zpool_hdl;
1898 	char errbuf[ERRBUFLEN];
1899 	nvlist_t **spares, **l2cache;
1900 	uint_t nspares, nl2cache;
1901 
1902 	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1903 	    "cannot add to '%s'"), zhp->zpool_name);
1904 
1905 	if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1906 	    SPA_VERSION_SPARES &&
1907 	    nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
1908 	    &spares, &nspares) == 0) {
1909 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1910 		    "upgraded to add hot spares"));
1911 		return (zfs_error(hdl, EZFS_BADVERSION, errbuf));
1912 	}
1913 
1914 	if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1915 	    SPA_VERSION_L2CACHE &&
1916 	    nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
1917 	    &l2cache, &nl2cache) == 0) {
1918 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1919 		    "upgraded to add cache devices"));
1920 		return (zfs_error(hdl, EZFS_BADVERSION, errbuf));
1921 	}
1922 
1923 	zcmd_write_conf_nvlist(hdl, &zc, nvroot);
1924 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1925 	zc.zc_flags = check_ashift;
1926 
1927 	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) {
1928 		switch (errno) {
1929 		case EBUSY:
1930 			/*
1931 			 * This can happen if the user has specified the same
1932 			 * device multiple times.  We can't reliably detect this
1933 			 * until we try to add it and see we already have a
1934 			 * label.
1935 			 */
1936 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1937 			    "one or more vdevs refer to the same device"));
1938 			(void) zfs_error(hdl, EZFS_BADDEV, errbuf);
1939 			break;
1940 
1941 		case EINVAL:
1942 
1943 			if (zpool_has_draid_vdev(nvroot) &&
1944 			    zfeature_lookup_name("draid", NULL) != 0) {
1945 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1946 				    "dRAID vdevs are unsupported by the "
1947 				    "kernel"));
1948 			} else {
1949 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1950 				    "invalid config; a pool with removing/"
1951 				    "removed vdevs does not support adding "
1952 				    "raidz or dRAID vdevs"));
1953 			}
1954 
1955 			(void) zfs_error(hdl, EZFS_BADDEV, errbuf);
1956 			break;
1957 
1958 		case EOVERFLOW:
1959 			/*
1960 			 * This occurs when one of the devices is below
1961 			 * SPA_MINDEVSIZE.  Unfortunately, we can't detect which
1962 			 * device was the problem device since there's no
1963 			 * reliable way to determine device size from userland.
1964 			 */
1965 			{
1966 				char buf[64];
1967 
1968 				zfs_nicebytes(SPA_MINDEVSIZE, buf,
1969 				    sizeof (buf));
1970 
1971 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1972 				    "device is less than the minimum "
1973 				    "size (%s)"), buf);
1974 			}
1975 			(void) zfs_error(hdl, EZFS_BADDEV, errbuf);
1976 			break;
1977 
1978 		case ENOTSUP:
1979 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1980 			    "pool must be upgraded to add these vdevs"));
1981 			(void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
1982 			break;
1983 
1984 		default:
1985 			(void) zpool_standard_error(hdl, errno, errbuf);
1986 		}
1987 
1988 		ret = -1;
1989 	} else {
1990 		ret = 0;
1991 	}
1992 
1993 	zcmd_free_nvlists(&zc);
1994 
1995 	return (ret);
1996 }
1997 
1998 /*
1999  * Exports the pool from the system.  The caller must ensure that there are no
2000  * mounted datasets in the pool.
2001  */
2002 static int
zpool_export_common(zpool_handle_t * zhp,boolean_t force,boolean_t hardforce,const char * log_str)2003 zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce,
2004     const char *log_str)
2005 {
2006 	zfs_cmd_t zc = {"\0"};
2007 
2008 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2009 	zc.zc_cookie = force;
2010 	zc.zc_guid = hardforce;
2011 	zc.zc_history = (uint64_t)(uintptr_t)log_str;
2012 
2013 	if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) {
2014 		switch (errno) {
2015 		case EXDEV:
2016 			zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
2017 			    "use '-f' to override the following errors:\n"
2018 			    "'%s' has an active shared spare which could be"
2019 			    " used by other pools once '%s' is exported."),
2020 			    zhp->zpool_name, zhp->zpool_name);
2021 			return (zfs_error_fmt(zhp->zpool_hdl, EZFS_ACTIVE_SPARE,
2022 			    dgettext(TEXT_DOMAIN, "cannot export '%s'"),
2023 			    zhp->zpool_name));
2024 		default:
2025 			return (zpool_standard_error_fmt(zhp->zpool_hdl, errno,
2026 			    dgettext(TEXT_DOMAIN, "cannot export '%s'"),
2027 			    zhp->zpool_name));
2028 		}
2029 	}
2030 
2031 	return (0);
2032 }
2033 
2034 int
zpool_export(zpool_handle_t * zhp,boolean_t force,const char * log_str)2035 zpool_export(zpool_handle_t *zhp, boolean_t force, const char *log_str)
2036 {
2037 	return (zpool_export_common(zhp, force, B_FALSE, log_str));
2038 }
2039 
2040 int
zpool_export_force(zpool_handle_t * zhp,const char * log_str)2041 zpool_export_force(zpool_handle_t *zhp, const char *log_str)
2042 {
2043 	return (zpool_export_common(zhp, B_TRUE, B_TRUE, log_str));
2044 }
2045 
2046 static void
zpool_rewind_exclaim(libzfs_handle_t * hdl,const char * name,boolean_t dryrun,nvlist_t * config)2047 zpool_rewind_exclaim(libzfs_handle_t *hdl, const char *name, boolean_t dryrun,
2048     nvlist_t *config)
2049 {
2050 	nvlist_t *nv = NULL;
2051 	uint64_t rewindto;
2052 	int64_t loss = -1;
2053 	struct tm t;
2054 	char timestr[128];
2055 
2056 	if (!hdl->libzfs_printerr || config == NULL)
2057 		return;
2058 
2059 	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
2060 	    nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0) {
2061 		return;
2062 	}
2063 
2064 	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
2065 		return;
2066 	(void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
2067 
2068 	if (localtime_r((time_t *)&rewindto, &t) != NULL &&
2069 	    ctime_r((time_t *)&rewindto, timestr) != NULL) {
2070 		timestr[24] = 0;
2071 		if (dryrun) {
2072 			(void) printf(dgettext(TEXT_DOMAIN,
2073 			    "Would be able to return %s "
2074 			    "to its state as of %s.\n"),
2075 			    name, timestr);
2076 		} else {
2077 			(void) printf(dgettext(TEXT_DOMAIN,
2078 			    "Pool %s returned to its state as of %s.\n"),
2079 			    name, timestr);
2080 		}
2081 		if (loss > 120) {
2082 			(void) printf(dgettext(TEXT_DOMAIN,
2083 			    "%s approximately %lld "),
2084 			    dryrun ? "Would discard" : "Discarded",
2085 			    ((longlong_t)loss + 30) / 60);
2086 			(void) printf(dgettext(TEXT_DOMAIN,
2087 			    "minutes of transactions.\n"));
2088 		} else if (loss > 0) {
2089 			(void) printf(dgettext(TEXT_DOMAIN,
2090 			    "%s approximately %lld "),
2091 			    dryrun ? "Would discard" : "Discarded",
2092 			    (longlong_t)loss);
2093 			(void) printf(dgettext(TEXT_DOMAIN,
2094 			    "seconds of transactions.\n"));
2095 		}
2096 	}
2097 }
2098 
2099 void
zpool_explain_recover(libzfs_handle_t * hdl,const char * name,int reason,nvlist_t * config,char * buf,size_t size)2100 zpool_explain_recover(libzfs_handle_t *hdl, const char *name, int reason,
2101     nvlist_t *config, char *buf, size_t size)
2102 {
2103 	nvlist_t *nv = NULL;
2104 	int64_t loss = -1;
2105 	uint64_t edata = UINT64_MAX;
2106 	uint64_t rewindto;
2107 	struct tm t;
2108 	char timestr[128], temp[1024];
2109 
2110 	if (!hdl->libzfs_printerr)
2111 		return;
2112 
2113 	/* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */
2114 	if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
2115 	    nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0 ||
2116 	    nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
2117 		goto no_info;
2118 
2119 	(void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
2120 	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_DATA_ERRORS,
2121 	    &edata);
2122 
2123 	(void) snprintf(buf, size, dgettext(TEXT_DOMAIN,
2124 	    "Recovery is possible, but will result in some data loss.\n"));
2125 
2126 	if (localtime_r((time_t *)&rewindto, &t) != NULL &&
2127 	    ctime_r((time_t *)&rewindto, timestr) != NULL) {
2128 		timestr[24] = 0;
2129 		(void) snprintf(temp, 1024, dgettext(TEXT_DOMAIN,
2130 		    "\tReturning the pool to its state as of %s\n"
2131 		    "\tshould correct the problem.  "), timestr);
2132 		(void) strlcat(buf, temp, size);
2133 	} else {
2134 		(void) strlcat(buf, dgettext(TEXT_DOMAIN,
2135 		    "\tReverting the pool to an earlier state "
2136 		    "should correct the problem.\n\t"), size);
2137 	}
2138 
2139 	if (loss > 120) {
2140 		(void) snprintf(temp, 1024, dgettext(TEXT_DOMAIN,
2141 		    "Approximately %lld minutes of data\n"
2142 		    "\tmust be discarded, irreversibly.  "),
2143 		    ((longlong_t)loss + 30) / 60);
2144 		(void) strlcat(buf, temp, size);
2145 	} else if (loss > 0) {
2146 		(void) snprintf(temp, 1024, dgettext(TEXT_DOMAIN,
2147 		    "Approximately %lld seconds of data\n"
2148 		    "\tmust be discarded, irreversibly.  "),
2149 		    (longlong_t)loss);
2150 		(void) strlcat(buf, temp, size);
2151 	}
2152 	if (edata != 0 && edata != UINT64_MAX) {
2153 		if (edata == 1) {
2154 			(void) strlcat(buf, dgettext(TEXT_DOMAIN,
2155 			    "After rewind, at least\n"
2156 			    "\tone persistent user-data error will remain.  "),
2157 			    size);
2158 		} else {
2159 			(void) strlcat(buf, dgettext(TEXT_DOMAIN,
2160 			    "After rewind, several\n"
2161 			    "\tpersistent user-data errors will remain.  "),
2162 			    size);
2163 		}
2164 	}
2165 	(void) snprintf(temp, 1024, dgettext(TEXT_DOMAIN,
2166 	    "Recovery can be attempted\n\tby executing 'zpool %s -F %s'.  "),
2167 	    reason >= 0 ? "clear" : "import", name);
2168 	(void) strlcat(buf, temp, size);
2169 
2170 	(void) strlcat(buf, dgettext(TEXT_DOMAIN,
2171 	    "A scrub of the pool\n"
2172 	    "\tis strongly recommended after recovery.\n"), size);
2173 	return;
2174 
2175 no_info:
2176 	(void) strlcat(buf, dgettext(TEXT_DOMAIN,
2177 	    "Ensure all pool devices are present and accessible, then "
2178 	    "retry the import.\n\tIf the problem persists, destroy and "
2179 	    "re-create the pool from a backup source.\n"), size);
2180 }
2181 
2182 /*
2183  * zpool_import() is a contracted interface. Should be kept the same
2184  * if possible.
2185  *
2186  * Applications should use zpool_import_props() to import a pool with
2187  * new properties value to be set.
2188  */
2189 int
zpool_import(libzfs_handle_t * hdl,nvlist_t * config,const char * newname,char * altroot)2190 zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
2191     char *altroot)
2192 {
2193 	nvlist_t *props = NULL;
2194 	int ret;
2195 
2196 	if (altroot != NULL) {
2197 		if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) {
2198 			return (zfs_error_fmt(hdl, EZFS_NOMEM,
2199 			    dgettext(TEXT_DOMAIN, "cannot import '%s'"),
2200 			    newname));
2201 		}
2202 
2203 		if (nvlist_add_string(props,
2204 		    zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 ||
2205 		    nvlist_add_string(props,
2206 		    zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) {
2207 			nvlist_free(props);
2208 			return (zfs_error_fmt(hdl, EZFS_NOMEM,
2209 			    dgettext(TEXT_DOMAIN, "cannot import '%s'"),
2210 			    newname));
2211 		}
2212 	}
2213 
2214 	ret = zpool_import_props(hdl, config, newname, props,
2215 	    ZFS_IMPORT_NORMAL);
2216 	nvlist_free(props);
2217 	return (ret);
2218 }
2219 
2220 static void
print_vdev_tree(libzfs_handle_t * hdl,const char * name,nvlist_t * nv,int indent)2221 print_vdev_tree(libzfs_handle_t *hdl, const char *name, nvlist_t *nv,
2222     int indent)
2223 {
2224 	nvlist_t **child;
2225 	uint_t c, children;
2226 	char *vname;
2227 	uint64_t is_log = 0;
2228 
2229 	(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG,
2230 	    &is_log);
2231 
2232 	if (name != NULL)
2233 		(void) printf("\t%*s%s%s\n", indent, "", name,
2234 		    is_log ? " [log]" : "");
2235 
2236 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
2237 	    &child, &children) != 0)
2238 		return;
2239 
2240 	for (c = 0; c < children; c++) {
2241 		vname = zpool_vdev_name(hdl, NULL, child[c], VDEV_NAME_TYPE_ID);
2242 		print_vdev_tree(hdl, vname, child[c], indent + 2);
2243 		free(vname);
2244 	}
2245 }
2246 
2247 void
zpool_collect_unsup_feat(nvlist_t * config,char * buf,size_t size)2248 zpool_collect_unsup_feat(nvlist_t *config, char *buf, size_t size)
2249 {
2250 	nvlist_t *nvinfo, *unsup_feat;
2251 	char temp[512];
2252 
2253 	nvinfo = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO);
2254 	unsup_feat = fnvlist_lookup_nvlist(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT);
2255 
2256 	for (nvpair_t *nvp = nvlist_next_nvpair(unsup_feat, NULL);
2257 	    nvp != NULL; nvp = nvlist_next_nvpair(unsup_feat, nvp)) {
2258 		const char *desc = fnvpair_value_string(nvp);
2259 		if (strlen(desc) > 0) {
2260 			(void) snprintf(temp, 512, "\t%s (%s)\n",
2261 			    nvpair_name(nvp), desc);
2262 			(void) strlcat(buf, temp, size);
2263 		} else {
2264 			(void) snprintf(temp, 512, "\t%s\n", nvpair_name(nvp));
2265 			(void) strlcat(buf, temp, size);
2266 		}
2267 	}
2268 }
2269 
2270 /*
2271  * Import the given pool using the known configuration and a list of
2272  * properties to be set. The configuration should have come from
2273  * zpool_find_import(). The 'newname' parameters control whether the pool
2274  * is imported with a different name.
2275  */
2276 int
zpool_import_props(libzfs_handle_t * hdl,nvlist_t * config,const char * newname,nvlist_t * props,int flags)2277 zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
2278     nvlist_t *props, int flags)
2279 {
2280 	zfs_cmd_t zc = {"\0"};
2281 	zpool_load_policy_t policy;
2282 	nvlist_t *nv = NULL;
2283 	nvlist_t *nvinfo = NULL;
2284 	nvlist_t *missing = NULL;
2285 	const char *thename;
2286 	const char *origname;
2287 	int ret;
2288 	int error = 0;
2289 	char buf[2048];
2290 	char errbuf[ERRBUFLEN];
2291 
2292 	origname = fnvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME);
2293 
2294 	(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
2295 	    "cannot import pool '%s'"), origname);
2296 
2297 	if (newname != NULL) {
2298 		if (!zpool_name_valid(hdl, B_FALSE, newname))
2299 			return (zfs_error_fmt(hdl, EZFS_INVALIDNAME,
2300 			    dgettext(TEXT_DOMAIN, "cannot import '%s'"),
2301 			    newname));
2302 		thename = newname;
2303 	} else {
2304 		thename = origname;
2305 	}
2306 
2307 	if (props != NULL) {
2308 		uint64_t version;
2309 		prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
2310 
2311 		version = fnvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION);
2312 
2313 		if ((props = zpool_valid_proplist(hdl, origname,
2314 		    props, version, flags, errbuf)) == NULL)
2315 			return (-1);
2316 		zcmd_write_src_nvlist(hdl, &zc, props);
2317 		nvlist_free(props);
2318 	}
2319 
2320 	(void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name));
2321 
2322 	zc.zc_guid = fnvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID);
2323 
2324 	zcmd_write_conf_nvlist(hdl, &zc, config);
2325 	zcmd_alloc_dst_nvlist(hdl, &zc, zc.zc_nvlist_conf_size * 2);
2326 
2327 	zc.zc_cookie = flags;
2328 	while ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc)) != 0 &&
2329 	    errno == ENOMEM)
2330 		zcmd_expand_dst_nvlist(hdl, &zc);
2331 	if (ret != 0)
2332 		error = errno;
2333 
2334 	(void) zcmd_read_dst_nvlist(hdl, &zc, &nv);
2335 
2336 	zcmd_free_nvlists(&zc);
2337 
2338 	zpool_get_load_policy(config, &policy);
2339 
2340 	if (getenv("ZFS_LOAD_INFO_DEBUG") && nv != NULL &&
2341 	    nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0) {
2342 		dump_nvlist(nvinfo, 4);
2343 	}
2344 
2345 	if (error) {
2346 		char desc[1024];
2347 		char aux[256];
2348 
2349 		/*
2350 		 * Dry-run failed, but we print out what success
2351 		 * looks like if we found a best txg
2352 		 */
2353 		if (policy.zlp_rewind & ZPOOL_TRY_REWIND) {
2354 			zpool_rewind_exclaim(hdl, newname ? origname : thename,
2355 			    B_TRUE, nv);
2356 			nvlist_free(nv);
2357 			return (-1);
2358 		}
2359 
2360 		if (newname == NULL)
2361 			(void) snprintf(desc, sizeof (desc),
2362 			    dgettext(TEXT_DOMAIN, "cannot import '%s'"),
2363 			    thename);
2364 		else
2365 			(void) snprintf(desc, sizeof (desc),
2366 			    dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"),
2367 			    origname, thename);
2368 
2369 		switch (error) {
2370 		case ENOTSUP:
2371 			if (nv != NULL && nvlist_lookup_nvlist(nv,
2372 			    ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
2373 			    nvlist_exists(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT)) {
2374 				(void) printf(dgettext(TEXT_DOMAIN, "This "
2375 				    "pool uses the following feature(s) not "
2376 				    "supported by this system:\n"));
2377 				memset(buf, 0, 2048);
2378 				zpool_collect_unsup_feat(nv, buf, 2048);
2379 				(void) printf("%s", buf);
2380 				if (nvlist_exists(nvinfo,
2381 				    ZPOOL_CONFIG_CAN_RDONLY)) {
2382 					(void) printf(dgettext(TEXT_DOMAIN,
2383 					    "All unsupported features are only "
2384 					    "required for writing to the pool."
2385 					    "\nThe pool can be imported using "
2386 					    "'-o readonly=on'.\n"));
2387 				}
2388 			}
2389 			/*
2390 			 * Unsupported version.
2391 			 */
2392 			(void) zfs_error(hdl, EZFS_BADVERSION, desc);
2393 			break;
2394 
2395 		case EREMOTEIO:
2396 			if (nv != NULL && nvlist_lookup_nvlist(nv,
2397 			    ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0) {
2398 				const char *hostname = "<unknown>";
2399 				uint64_t hostid = 0;
2400 				mmp_state_t mmp_state;
2401 
2402 				mmp_state = fnvlist_lookup_uint64(nvinfo,
2403 				    ZPOOL_CONFIG_MMP_STATE);
2404 
2405 				if (nvlist_exists(nvinfo,
2406 				    ZPOOL_CONFIG_MMP_HOSTNAME))
2407 					hostname = fnvlist_lookup_string(nvinfo,
2408 					    ZPOOL_CONFIG_MMP_HOSTNAME);
2409 
2410 				if (nvlist_exists(nvinfo,
2411 				    ZPOOL_CONFIG_MMP_HOSTID))
2412 					hostid = fnvlist_lookup_uint64(nvinfo,
2413 					    ZPOOL_CONFIG_MMP_HOSTID);
2414 
2415 				if (mmp_state == MMP_STATE_ACTIVE) {
2416 					(void) snprintf(aux, sizeof (aux),
2417 					    dgettext(TEXT_DOMAIN, "pool is imp"
2418 					    "orted on host '%s' (hostid=%lx).\n"
2419 					    "Export the pool on the other "
2420 					    "system, then run 'zpool import'."),
2421 					    hostname, (unsigned long) hostid);
2422 				} else if (mmp_state == MMP_STATE_NO_HOSTID) {
2423 					(void) snprintf(aux, sizeof (aux),
2424 					    dgettext(TEXT_DOMAIN, "pool has "
2425 					    "the multihost property on and "
2426 					    "the\nsystem's hostid is not set. "
2427 					    "Set a unique system hostid with "
2428 					    "the zgenhostid(8) command.\n"));
2429 				}
2430 
2431 				(void) zfs_error_aux(hdl, "%s", aux);
2432 			}
2433 			(void) zfs_error(hdl, EZFS_ACTIVE_POOL, desc);
2434 			break;
2435 
2436 		case EINVAL:
2437 			(void) zfs_error(hdl, EZFS_INVALCONFIG, desc);
2438 			break;
2439 
2440 		case EROFS:
2441 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2442 			    "one or more devices is read only"));
2443 			(void) zfs_error(hdl, EZFS_BADDEV, desc);
2444 			break;
2445 
2446 		case ENXIO:
2447 			if (nv && nvlist_lookup_nvlist(nv,
2448 			    ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
2449 			    nvlist_lookup_nvlist(nvinfo,
2450 			    ZPOOL_CONFIG_MISSING_DEVICES, &missing) == 0) {
2451 				(void) printf(dgettext(TEXT_DOMAIN,
2452 				    "The devices below are missing or "
2453 				    "corrupted, use '-m' to import the pool "
2454 				    "anyway:\n"));
2455 				print_vdev_tree(hdl, NULL, missing, 2);
2456 				(void) printf("\n");
2457 			}
2458 			(void) zpool_standard_error(hdl, error, desc);
2459 			break;
2460 
2461 		case EEXIST:
2462 			(void) zpool_standard_error(hdl, error, desc);
2463 			break;
2464 
2465 		case EBUSY:
2466 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2467 			    "one or more devices are already in use\n"));
2468 			(void) zfs_error(hdl, EZFS_BADDEV, desc);
2469 			break;
2470 		case ENAMETOOLONG:
2471 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2472 			    "new name of at least one dataset is longer than "
2473 			    "the maximum allowable length"));
2474 			(void) zfs_error(hdl, EZFS_NAMETOOLONG, desc);
2475 			break;
2476 		default:
2477 			(void) zpool_standard_error(hdl, error, desc);
2478 			memset(buf, 0, 2048);
2479 			zpool_explain_recover(hdl,
2480 			    newname ? origname : thename, -error, nv,
2481 			    buf, 2048);
2482 			(void) printf("\t%s", buf);
2483 			break;
2484 		}
2485 
2486 		nvlist_free(nv);
2487 		ret = -1;
2488 	} else {
2489 		zpool_handle_t *zhp;
2490 
2491 		/*
2492 		 * This should never fail, but play it safe anyway.
2493 		 */
2494 		if (zpool_open_silent(hdl, thename, &zhp) != 0)
2495 			ret = -1;
2496 		else if (zhp != NULL)
2497 			zpool_close(zhp);
2498 		if (policy.zlp_rewind &
2499 		    (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
2500 			zpool_rewind_exclaim(hdl, newname ? origname : thename,
2501 			    ((policy.zlp_rewind & ZPOOL_TRY_REWIND) != 0), nv);
2502 		}
2503 		nvlist_free(nv);
2504 	}
2505 
2506 	return (ret);
2507 }
2508 
2509 /*
2510  * Translate vdev names to guids.  If a vdev_path is determined to be
2511  * unsuitable then a vd_errlist is allocated and the vdev path and errno
2512  * are added to it.
2513  */
2514 static int
zpool_translate_vdev_guids(zpool_handle_t * zhp,nvlist_t * vds,nvlist_t * vdev_guids,nvlist_t * guids_to_paths,nvlist_t ** vd_errlist)2515 zpool_translate_vdev_guids(zpool_handle_t *zhp, nvlist_t *vds,
2516     nvlist_t *vdev_guids, nvlist_t *guids_to_paths, nvlist_t **vd_errlist)
2517 {
2518 	nvlist_t *errlist = NULL;
2519 	int error = 0;
2520 
2521 	for (nvpair_t *elem = nvlist_next_nvpair(vds, NULL); elem != NULL;
2522 	    elem = nvlist_next_nvpair(vds, elem)) {
2523 		boolean_t spare, cache;
2524 
2525 		const char *vd_path = nvpair_name(elem);
2526 		nvlist_t *tgt = zpool_find_vdev(zhp, vd_path, &spare, &cache,
2527 		    NULL);
2528 
2529 		if ((tgt == NULL) || cache || spare) {
2530 			if (errlist == NULL) {
2531 				errlist = fnvlist_alloc();
2532 				error = EINVAL;
2533 			}
2534 
2535 			uint64_t err = (tgt == NULL) ? EZFS_NODEVICE :
2536 			    (spare ? EZFS_ISSPARE : EZFS_ISL2CACHE);
2537 			fnvlist_add_int64(errlist, vd_path, err);
2538 			continue;
2539 		}
2540 
2541 		uint64_t guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
2542 		fnvlist_add_uint64(vdev_guids, vd_path, guid);
2543 
2544 		char msg[MAXNAMELEN];
2545 		(void) snprintf(msg, sizeof (msg), "%llu", (u_longlong_t)guid);
2546 		fnvlist_add_string(guids_to_paths, msg, vd_path);
2547 	}
2548 
2549 	if (error != 0) {
2550 		verify(errlist != NULL);
2551 		if (vd_errlist != NULL)
2552 			*vd_errlist = errlist;
2553 		else
2554 			fnvlist_free(errlist);
2555 	}
2556 
2557 	return (error);
2558 }
2559 
2560 static int
xlate_init_err(int err)2561 xlate_init_err(int err)
2562 {
2563 	switch (err) {
2564 	case ENODEV:
2565 		return (EZFS_NODEVICE);
2566 	case EINVAL:
2567 	case EROFS:
2568 		return (EZFS_BADDEV);
2569 	case EBUSY:
2570 		return (EZFS_INITIALIZING);
2571 	case ESRCH:
2572 		return (EZFS_NO_INITIALIZE);
2573 	}
2574 	return (err);
2575 }
2576 
2577 int
zpool_initialize_one(zpool_handle_t * zhp,void * data)2578 zpool_initialize_one(zpool_handle_t *zhp, void *data)
2579 {
2580 	int error;
2581 	libzfs_handle_t *hdl = zpool_get_handle(zhp);
2582 	const char *pool_name = zpool_get_name(zhp);
2583 	if (zpool_open_silent(hdl, pool_name, &zhp) != 0)
2584 		return (-1);
2585 	initialize_cbdata_t *cb = data;
2586 	nvlist_t *vdevs = fnvlist_alloc();
2587 
2588 	nvlist_t *config = zpool_get_config(zhp, NULL);
2589 	nvlist_t *nvroot = fnvlist_lookup_nvlist(config,
2590 	    ZPOOL_CONFIG_VDEV_TREE);
2591 	zpool_collect_leaves(zhp, nvroot, vdevs);
2592 	if (cb->wait)
2593 		error = zpool_initialize_wait(zhp, cb->cmd_type, vdevs);
2594 	else
2595 		error = zpool_initialize(zhp, cb->cmd_type, vdevs);
2596 	fnvlist_free(vdevs);
2597 
2598 	return (error);
2599 }
2600 
2601 /*
2602  * Begin, suspend, cancel, or uninit (clear) the initialization (initializing
2603  * of all free blocks) for the given vdevs in the given pool.
2604  */
2605 static int
zpool_initialize_impl(zpool_handle_t * zhp,pool_initialize_func_t cmd_type,nvlist_t * vds,boolean_t wait)2606 zpool_initialize_impl(zpool_handle_t *zhp, pool_initialize_func_t cmd_type,
2607     nvlist_t *vds, boolean_t wait)
2608 {
2609 	int err;
2610 
2611 	nvlist_t *vdev_guids = fnvlist_alloc();
2612 	nvlist_t *guids_to_paths = fnvlist_alloc();
2613 	nvlist_t *vd_errlist = NULL;
2614 	nvlist_t *errlist;
2615 	nvpair_t *elem;
2616 
2617 	err = zpool_translate_vdev_guids(zhp, vds, vdev_guids,
2618 	    guids_to_paths, &vd_errlist);
2619 
2620 	if (err != 0) {
2621 		verify(vd_errlist != NULL);
2622 		goto list_errors;
2623 	}
2624 
2625 	err = lzc_initialize(zhp->zpool_name, cmd_type,
2626 	    vdev_guids, &errlist);
2627 
2628 	if (err != 0) {
2629 		if (errlist != NULL && nvlist_lookup_nvlist(errlist,
2630 		    ZPOOL_INITIALIZE_VDEVS, &vd_errlist) == 0) {
2631 			goto list_errors;
2632 		}
2633 
2634 		if (err == EINVAL && cmd_type == POOL_INITIALIZE_UNINIT) {
2635 			zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
2636 			    "uninitialize is not supported by kernel"));
2637 		}
2638 
2639 		(void) zpool_standard_error(zhp->zpool_hdl, err,
2640 		    dgettext(TEXT_DOMAIN, "operation failed"));
2641 		goto out;
2642 	}
2643 
2644 	if (wait) {
2645 		for (elem = nvlist_next_nvpair(vdev_guids, NULL); elem != NULL;
2646 		    elem = nvlist_next_nvpair(vdev_guids, elem)) {
2647 
2648 			uint64_t guid = fnvpair_value_uint64(elem);
2649 
2650 			err = lzc_wait_tag(zhp->zpool_name,
2651 			    ZPOOL_WAIT_INITIALIZE, guid, NULL);
2652 			if (err != 0) {
2653 				(void) zpool_standard_error_fmt(zhp->zpool_hdl,
2654 				    err, dgettext(TEXT_DOMAIN, "error "
2655 				    "waiting for '%s' to initialize"),
2656 				    nvpair_name(elem));
2657 
2658 				goto out;
2659 			}
2660 		}
2661 	}
2662 	goto out;
2663 
2664 list_errors:
2665 	for (elem = nvlist_next_nvpair(vd_errlist, NULL); elem != NULL;
2666 	    elem = nvlist_next_nvpair(vd_errlist, elem)) {
2667 		int64_t vd_error = xlate_init_err(fnvpair_value_int64(elem));
2668 		const char *path;
2669 
2670 		if (nvlist_lookup_string(guids_to_paths, nvpair_name(elem),
2671 		    &path) != 0)
2672 			path = nvpair_name(elem);
2673 
2674 		(void) zfs_error_fmt(zhp->zpool_hdl, vd_error,
2675 		    "cannot initialize '%s'", path);
2676 	}
2677 
2678 out:
2679 	fnvlist_free(vdev_guids);
2680 	fnvlist_free(guids_to_paths);
2681 
2682 	if (vd_errlist != NULL)
2683 		fnvlist_free(vd_errlist);
2684 
2685 	return (err == 0 ? 0 : -1);
2686 }
2687 
2688 int
zpool_initialize(zpool_handle_t * zhp,pool_initialize_func_t cmd_type,nvlist_t * vds)2689 zpool_initialize(zpool_handle_t *zhp, pool_initialize_func_t cmd_type,
2690     nvlist_t *vds)
2691 {
2692 	return (zpool_initialize_impl(zhp, cmd_type, vds, B_FALSE));
2693 }
2694 
2695 int
zpool_initialize_wait(zpool_handle_t * zhp,pool_initialize_func_t cmd_type,nvlist_t * vds)2696 zpool_initialize_wait(zpool_handle_t *zhp, pool_initialize_func_t cmd_type,
2697     nvlist_t *vds)
2698 {
2699 	return (zpool_initialize_impl(zhp, cmd_type, vds, B_TRUE));
2700 }
2701 
2702 static int
xlate_trim_err(int err)2703 xlate_trim_err(int err)
2704 {
2705 	switch (err) {
2706 	case ENODEV:
2707 		return (EZFS_NODEVICE);
2708 	case EINVAL:
2709 	case EROFS:
2710 		return (EZFS_BADDEV);
2711 	case EBUSY:
2712 		return (EZFS_TRIMMING);
2713 	case ESRCH:
2714 		return (EZFS_NO_TRIM);
2715 	case EOPNOTSUPP:
2716 		return (EZFS_TRIM_NOTSUP);
2717 	}
2718 	return (err);
2719 }
2720 
2721 void
zpool_collect_leaves(zpool_handle_t * zhp,nvlist_t * nvroot,nvlist_t * res)2722 zpool_collect_leaves(zpool_handle_t *zhp, nvlist_t *nvroot, nvlist_t *res)
2723 {
2724 	libzfs_handle_t *hdl = zhp->zpool_hdl;
2725 	uint_t children = 0;
2726 	nvlist_t **child;
2727 	uint_t i;
2728 
2729 	(void) nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
2730 	    &child, &children);
2731 
2732 	if (children == 0) {
2733 		char *path = zpool_vdev_name(hdl, zhp, nvroot,
2734 		    VDEV_NAME_PATH);
2735 
2736 		if (strcmp(path, VDEV_TYPE_INDIRECT) != 0 &&
2737 		    strcmp(path, VDEV_TYPE_HOLE) != 0)
2738 			fnvlist_add_boolean(res, path);
2739 
2740 		free(path);
2741 		return;
2742 	}
2743 
2744 	for (i = 0; i < children; i++) {
2745 		zpool_collect_leaves(zhp, child[i], res);
2746 	}
2747 }
2748 
2749 int
zpool_trim_one(zpool_handle_t * zhp,void * data)2750 zpool_trim_one(zpool_handle_t *zhp, void *data)
2751 {
2752 	int error;
2753 	libzfs_handle_t *hdl = zpool_get_handle(zhp);
2754 	const char *pool_name = zpool_get_name(zhp);
2755 	if (zpool_open_silent(hdl, pool_name, &zhp) != 0)
2756 		return (-1);
2757 
2758 	trim_cbdata_t *cb = data;
2759 	nvlist_t *vdevs = fnvlist_alloc();
2760 
2761 	/* no individual leaf vdevs specified, so add them all */
2762 	nvlist_t *config = zpool_get_config(zhp, NULL);
2763 	nvlist_t *nvroot = fnvlist_lookup_nvlist(config,
2764 	    ZPOOL_CONFIG_VDEV_TREE);
2765 
2766 	zpool_collect_leaves(zhp, nvroot, vdevs);
2767 	error = zpool_trim(zhp, cb->cmd_type, vdevs, &cb->trim_flags);
2768 	fnvlist_free(vdevs);
2769 
2770 	return (error);
2771 }
2772 
2773 static int
zpool_trim_wait(zpool_handle_t * zhp,nvlist_t * vdev_guids)2774 zpool_trim_wait(zpool_handle_t *zhp, nvlist_t *vdev_guids)
2775 {
2776 	int err;
2777 	nvpair_t *elem;
2778 
2779 	for (elem = nvlist_next_nvpair(vdev_guids, NULL); elem != NULL;
2780 	    elem = nvlist_next_nvpair(vdev_guids, elem)) {
2781 
2782 		uint64_t guid = fnvpair_value_uint64(elem);
2783 
2784 		err = lzc_wait_tag(zhp->zpool_name,
2785 		    ZPOOL_WAIT_TRIM, guid, NULL);
2786 		if (err != 0) {
2787 			(void) zpool_standard_error_fmt(zhp->zpool_hdl,
2788 			    err, dgettext(TEXT_DOMAIN, "error "
2789 			    "waiting to trim '%s'"), nvpair_name(elem));
2790 
2791 			return (err);
2792 		}
2793 	}
2794 	return (0);
2795 }
2796 
2797 /*
2798  * Check errlist and report any errors, omitting ones which should be
2799  * suppressed. Returns B_TRUE if any errors were reported.
2800  */
2801 static boolean_t
check_trim_errs(zpool_handle_t * zhp,trimflags_t * trim_flags,nvlist_t * guids_to_paths,nvlist_t * vds,nvlist_t * errlist)2802 check_trim_errs(zpool_handle_t *zhp, trimflags_t *trim_flags,
2803     nvlist_t *guids_to_paths, nvlist_t *vds, nvlist_t *errlist)
2804 {
2805 	nvpair_t *elem;
2806 	boolean_t reported_errs = B_FALSE;
2807 	int num_vds = 0;
2808 	int num_suppressed_errs = 0;
2809 
2810 	for (elem = nvlist_next_nvpair(vds, NULL);
2811 	    elem != NULL; elem = nvlist_next_nvpair(vds, elem)) {
2812 		num_vds++;
2813 	}
2814 
2815 	for (elem = nvlist_next_nvpair(errlist, NULL);
2816 	    elem != NULL; elem = nvlist_next_nvpair(errlist, elem)) {
2817 		int64_t vd_error = xlate_trim_err(fnvpair_value_int64(elem));
2818 		const char *path;
2819 
2820 		/*
2821 		 * If only the pool was specified, and it was not a secure
2822 		 * trim then suppress warnings for individual vdevs which
2823 		 * do not support trimming.
2824 		 */
2825 		if (vd_error == EZFS_TRIM_NOTSUP &&
2826 		    trim_flags->fullpool &&
2827 		    !trim_flags->secure) {
2828 			num_suppressed_errs++;
2829 			continue;
2830 		}
2831 
2832 		reported_errs = B_TRUE;
2833 		if (nvlist_lookup_string(guids_to_paths, nvpair_name(elem),
2834 		    &path) != 0)
2835 			path = nvpair_name(elem);
2836 
2837 		(void) zfs_error_fmt(zhp->zpool_hdl, vd_error,
2838 		    "cannot trim '%s'", path);
2839 	}
2840 
2841 	if (num_suppressed_errs == num_vds) {
2842 		(void) zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
2843 		    "no devices in pool support trim operations"));
2844 		(void) (zfs_error(zhp->zpool_hdl, EZFS_TRIM_NOTSUP,
2845 		    dgettext(TEXT_DOMAIN, "cannot trim")));
2846 		reported_errs = B_TRUE;
2847 	}
2848 
2849 	return (reported_errs);
2850 }
2851 
2852 /*
2853  * Begin, suspend, or cancel the TRIM (discarding of all free blocks) for
2854  * the given vdevs in the given pool.
2855  */
2856 int
zpool_trim(zpool_handle_t * zhp,pool_trim_func_t cmd_type,nvlist_t * vds,trimflags_t * trim_flags)2857 zpool_trim(zpool_handle_t *zhp, pool_trim_func_t cmd_type, nvlist_t *vds,
2858     trimflags_t *trim_flags)
2859 {
2860 	int err;
2861 	int retval = 0;
2862 
2863 	nvlist_t *vdev_guids = fnvlist_alloc();
2864 	nvlist_t *guids_to_paths = fnvlist_alloc();
2865 	nvlist_t *errlist = NULL;
2866 
2867 	err = zpool_translate_vdev_guids(zhp, vds, vdev_guids,
2868 	    guids_to_paths, &errlist);
2869 	if (err != 0) {
2870 		check_trim_errs(zhp, trim_flags, guids_to_paths, vds, errlist);
2871 		retval = -1;
2872 		goto out;
2873 	}
2874 
2875 	err = lzc_trim(zhp->zpool_name, cmd_type, trim_flags->rate,
2876 	    trim_flags->secure, vdev_guids, &errlist);
2877 	if (err != 0) {
2878 		nvlist_t *vd_errlist;
2879 		if (errlist != NULL && nvlist_lookup_nvlist(errlist,
2880 		    ZPOOL_TRIM_VDEVS, &vd_errlist) == 0) {
2881 			if (check_trim_errs(zhp, trim_flags, guids_to_paths,
2882 			    vds, vd_errlist)) {
2883 				retval = -1;
2884 				goto out;
2885 			}
2886 		} else {
2887 			char errbuf[ERRBUFLEN];
2888 
2889 			(void) snprintf(errbuf, sizeof (errbuf),
2890 			    dgettext(TEXT_DOMAIN, "operation failed"));
2891 			zpool_standard_error(zhp->zpool_hdl, err, errbuf);
2892 			retval = -1;
2893 			goto out;
2894 		}
2895 	}
2896 
2897 
2898 	if (trim_flags->wait)
2899 		retval = zpool_trim_wait(zhp, vdev_guids);
2900 
2901 out:
2902 	if (errlist != NULL)
2903 		fnvlist_free(errlist);
2904 	fnvlist_free(vdev_guids);
2905 	fnvlist_free(guids_to_paths);
2906 	return (retval);
2907 }
2908 
2909 /*
2910  * Scan the pool.
2911  */
2912 int
zpool_scan(zpool_handle_t * zhp,pool_scan_func_t func,pool_scrub_cmd_t cmd)2913 zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func, pool_scrub_cmd_t cmd) {
2914 	return (zpool_scan_range(zhp, func, cmd, 0, 0));
2915 }
2916 
2917 int
zpool_scan_range(zpool_handle_t * zhp,pool_scan_func_t func,pool_scrub_cmd_t cmd,time_t date_start,time_t date_end)2918 zpool_scan_range(zpool_handle_t *zhp, pool_scan_func_t func,
2919     pool_scrub_cmd_t cmd, time_t date_start, time_t date_end)
2920 {
2921 	char errbuf[ERRBUFLEN];
2922 	int err;
2923 	libzfs_handle_t *hdl = zhp->zpool_hdl;
2924 
2925 	nvlist_t *args = fnvlist_alloc();
2926 	fnvlist_add_uint64(args, "scan_type", (uint64_t)func);
2927 	fnvlist_add_uint64(args, "scan_command", (uint64_t)cmd);
2928 	if (date_start != 0 || date_end != 0) {
2929 		fnvlist_add_uint64(args, "scan_date_start",
2930 		    (uint64_t)date_start);
2931 		fnvlist_add_uint64(args, "scan_date_end", (uint64_t)date_end);
2932 	}
2933 
2934 	err = lzc_scrub(ZFS_IOC_POOL_SCRUB, zhp->zpool_name, args, NULL);
2935 	fnvlist_free(args);
2936 
2937 	if (err == 0) {
2938 		return (0);
2939 	} else if (err == ZFS_ERR_IOC_CMD_UNAVAIL) {
2940 		zfs_cmd_t zc = {"\0"};
2941 		(void) strlcpy(zc.zc_name, zhp->zpool_name,
2942 		    sizeof (zc.zc_name));
2943 		zc.zc_cookie = func;
2944 		zc.zc_flags = cmd;
2945 
2946 		if (zfs_ioctl(hdl, ZFS_IOC_POOL_SCAN, &zc) == 0)
2947 			return (0);
2948 	}
2949 
2950 	/*
2951 	 * An ECANCELED on a scrub means one of the following:
2952 	 * 1. we resumed a paused scrub.
2953 	 * 2. we resumed a paused error scrub.
2954 	 * 3. Error scrub is not run because of no error log.
2955 	 *
2956 	 * Note that we no longer return ECANCELED in case 1 or 2. However, in
2957 	 * order to prevent problems where we have a newer userland than
2958 	 * kernel, we keep this check in place. That prevents erroneous
2959 	 * failures when an older kernel returns ECANCELED in those cases.
2960 	 */
2961 	if (err == ECANCELED && (func == POOL_SCAN_SCRUB ||
2962 	    func == POOL_SCAN_ERRORSCRUB) && cmd == POOL_SCRUB_NORMAL)
2963 		return (0);
2964 	/*
2965 	 * The following cases have been handled here:
2966 	 * 1. Paused a scrub/error scrub if there is none in progress.
2967 	 */
2968 	if (err == ENOENT && func != POOL_SCAN_NONE && cmd ==
2969 	    POOL_SCRUB_PAUSE) {
2970 		return (0);
2971 	}
2972 
2973 	ASSERT3U(func, >=, POOL_SCAN_NONE);
2974 	ASSERT3U(func, <, POOL_SCAN_FUNCS);
2975 
2976 	if (func == POOL_SCAN_SCRUB || func == POOL_SCAN_ERRORSCRUB) {
2977 		if (cmd == POOL_SCRUB_PAUSE) {
2978 			(void) snprintf(errbuf, sizeof (errbuf),
2979 			    dgettext(TEXT_DOMAIN, "cannot pause scrubbing %s"),
2980 			    zhp->zpool_name);
2981 		} else {
2982 			assert(cmd == POOL_SCRUB_NORMAL);
2983 			(void) snprintf(errbuf, sizeof (errbuf),
2984 			    dgettext(TEXT_DOMAIN, "cannot scrub %s"),
2985 			    zhp->zpool_name);
2986 		}
2987 	} else if (func == POOL_SCAN_RESILVER) {
2988 		assert(cmd == POOL_SCRUB_NORMAL);
2989 		(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
2990 		    "cannot restart resilver on %s"), zhp->zpool_name);
2991 	} else if (func == POOL_SCAN_NONE) {
2992 		(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
2993 		    "cannot cancel scrubbing %s"), zhp->zpool_name);
2994 	} else {
2995 		assert(!"unexpected result");
2996 	}
2997 
2998 	/*
2999 	 * With EBUSY, six cases are possible:
3000 	 *
3001 	 * Current state		Requested
3002 	 * 1. Normal Scrub Running	Normal Scrub or Error Scrub
3003 	 * 2. Normal Scrub Paused	Error Scrub
3004 	 * 3. Normal Scrub Paused 	Pause Normal Scrub
3005 	 * 4. Error Scrub Running	Normal Scrub or Error Scrub
3006 	 * 5. Error Scrub Paused	Pause Error Scrub
3007 	 * 6. Resilvering		Anything else
3008 	 */
3009 	if (err == EBUSY) {
3010 		nvlist_t *nvroot;
3011 		pool_scan_stat_t *ps = NULL;
3012 		uint_t psc;
3013 
3014 		nvroot = fnvlist_lookup_nvlist(zhp->zpool_config,
3015 		    ZPOOL_CONFIG_VDEV_TREE);
3016 		(void) nvlist_lookup_uint64_array(nvroot,
3017 		    ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc);
3018 		if (ps && ps->pss_func == POOL_SCAN_SCRUB &&
3019 		    ps->pss_state == DSS_SCANNING) {
3020 			if (ps->pss_pass_scrub_pause == 0) {
3021 				/* handles case 1 */
3022 				assert(cmd == POOL_SCRUB_NORMAL);
3023 				return (zfs_error(hdl, EZFS_SCRUBBING,
3024 				    errbuf));
3025 			} else {
3026 				if (func == POOL_SCAN_ERRORSCRUB) {
3027 					/* handles case 2 */
3028 					ASSERT3U(cmd, ==, POOL_SCRUB_NORMAL);
3029 					return (zfs_error(hdl,
3030 					    EZFS_SCRUB_PAUSED_TO_CANCEL,
3031 					    errbuf));
3032 				} else {
3033 					/* handles case 3 */
3034 					ASSERT3U(func, ==, POOL_SCAN_SCRUB);
3035 					ASSERT3U(cmd, ==, POOL_SCRUB_PAUSE);
3036 					return (zfs_error(hdl,
3037 					    EZFS_SCRUB_PAUSED, errbuf));
3038 				}
3039 			}
3040 		} else if (ps &&
3041 		    ps->pss_error_scrub_func == POOL_SCAN_ERRORSCRUB &&
3042 		    ps->pss_error_scrub_state == DSS_ERRORSCRUBBING) {
3043 			if (ps->pss_pass_error_scrub_pause == 0) {
3044 				/* handles case 4 */
3045 				ASSERT3U(cmd, ==, POOL_SCRUB_NORMAL);
3046 				return (zfs_error(hdl, EZFS_ERRORSCRUBBING,
3047 				    errbuf));
3048 			} else {
3049 				/* handles case 5 */
3050 				ASSERT3U(func, ==, POOL_SCAN_ERRORSCRUB);
3051 				ASSERT3U(cmd, ==, POOL_SCRUB_PAUSE);
3052 				return (zfs_error(hdl, EZFS_ERRORSCRUB_PAUSED,
3053 				    errbuf));
3054 			}
3055 		} else {
3056 			/* handles case 6 */
3057 			return (zfs_error(hdl, EZFS_RESILVERING, errbuf));
3058 		}
3059 	} else if (err == ENOENT) {
3060 		return (zfs_error(hdl, EZFS_NO_SCRUB, errbuf));
3061 	} else if (err == ENOTSUP && func == POOL_SCAN_RESILVER) {
3062 		return (zfs_error(hdl, EZFS_NO_RESILVER_DEFER, errbuf));
3063 	} else {
3064 		return (zpool_standard_error(hdl, err, errbuf));
3065 	}
3066 }
3067 
3068 /*
3069  * Find a vdev that matches the search criteria specified. We use the
3070  * the nvpair name to determine how we should look for the device.
3071  * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
3072  * spare; but FALSE if its an INUSE spare.
3073  *
3074  * If 'return_parent' is set, then return the *parent* of the vdev you're
3075  * searching for rather than the vdev itself.
3076  */
3077 static nvlist_t *
vdev_to_nvlist_iter(nvlist_t * nv,nvlist_t * search,boolean_t * avail_spare,boolean_t * l2cache,boolean_t * log,boolean_t return_parent)3078 vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare,
3079     boolean_t *l2cache, boolean_t *log, boolean_t return_parent)
3080 {
3081 	uint_t c, children;
3082 	nvlist_t **child;
3083 	nvlist_t *ret;
3084 	uint64_t is_log;
3085 	const char *srchkey;
3086 	nvpair_t *pair = nvlist_next_nvpair(search, NULL);
3087 	const char *tmp = NULL;
3088 	boolean_t is_root;
3089 
3090 	/* Nothing to look for */
3091 	if (search == NULL || pair == NULL)
3092 		return (NULL);
3093 
3094 	/* Obtain the key we will use to search */
3095 	srchkey = nvpair_name(pair);
3096 
3097 	nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &tmp);
3098 	if (strcmp(tmp, "root") == 0)
3099 		is_root = B_TRUE;
3100 	else
3101 		is_root = B_FALSE;
3102 
3103 	switch (nvpair_type(pair)) {
3104 	case DATA_TYPE_UINT64:
3105 		if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) {
3106 			uint64_t srchval = fnvpair_value_uint64(pair);
3107 			uint64_t theguid = fnvlist_lookup_uint64(nv,
3108 			    ZPOOL_CONFIG_GUID);
3109 			if (theguid == srchval)
3110 				return (nv);
3111 		}
3112 		break;
3113 
3114 	case DATA_TYPE_STRING: {
3115 		const char *srchval, *val;
3116 
3117 		srchval = fnvpair_value_string(pair);
3118 		if (nvlist_lookup_string(nv, srchkey, &val) != 0)
3119 			break;
3120 
3121 		/*
3122 		 * Search for the requested value. Special cases:
3123 		 *
3124 		 * - ZPOOL_CONFIG_PATH for whole disk entries.  These end in
3125 		 *   "-part1", or "p1".  The suffix is hidden from the user,
3126 		 *   but included in the string, so this matches around it.
3127 		 * - ZPOOL_CONFIG_PATH for short names zfs_strcmp_shortname()
3128 		 *   is used to check all possible expanded paths.
3129 		 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE).
3130 		 *
3131 		 * Otherwise, all other searches are simple string compares.
3132 		 */
3133 		if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0) {
3134 			uint64_t wholedisk = 0;
3135 
3136 			(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
3137 			    &wholedisk);
3138 			if (zfs_strcmp_pathname(srchval, val, wholedisk) == 0)
3139 				return (nv);
3140 
3141 		} else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0) {
3142 			char *type, *idx, *end, *p;
3143 			uint64_t id, vdev_id;
3144 
3145 			/*
3146 			 * Determine our vdev type, keeping in mind
3147 			 * that the srchval is composed of a type and
3148 			 * vdev id pair (i.e. mirror-4).
3149 			 */
3150 			if ((type = strdup(srchval)) == NULL)
3151 				return (NULL);
3152 
3153 			if ((p = strrchr(type, '-')) == NULL) {
3154 				free(type);
3155 				break;
3156 			}
3157 			idx = p + 1;
3158 			*p = '\0';
3159 
3160 			/*
3161 			 * draid names are presented like: draid2:4d:6c:0s
3162 			 * We match them up to the first ':' so we can still
3163 			 * do the parity check below, but the other params
3164 			 * are ignored.
3165 			 */
3166 			if ((p = strchr(type, ':')) != NULL) {
3167 				if (strncmp(type, VDEV_TYPE_DRAID,
3168 				    strlen(VDEV_TYPE_DRAID)) == 0)
3169 					*p = '\0';
3170 			}
3171 
3172 			/*
3173 			 * If the types don't match then keep looking.
3174 			 */
3175 			if (strncmp(val, type, strlen(val)) != 0) {
3176 				free(type);
3177 				break;
3178 			}
3179 
3180 			verify(zpool_vdev_is_interior(type));
3181 
3182 			id = fnvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID);
3183 			errno = 0;
3184 			vdev_id = strtoull(idx, &end, 10);
3185 
3186 			/*
3187 			 * If we are looking for a raidz and a parity is
3188 			 * specified, make sure it matches.
3189 			 */
3190 			int rzlen = strlen(VDEV_TYPE_RAIDZ);
3191 			assert(rzlen == strlen(VDEV_TYPE_DRAID));
3192 			int typlen = strlen(type);
3193 			if ((strncmp(type, VDEV_TYPE_RAIDZ, rzlen) == 0 ||
3194 			    strncmp(type, VDEV_TYPE_DRAID, rzlen) == 0) &&
3195 			    typlen != rzlen) {
3196 				uint64_t vdev_parity;
3197 				int parity = *(type + rzlen) - '0';
3198 
3199 				if (parity <= 0 || parity > 3 ||
3200 				    (typlen - rzlen) != 1) {
3201 					/*
3202 					 * Nonsense parity specified, can
3203 					 * never match
3204 					 */
3205 					free(type);
3206 					return (NULL);
3207 				}
3208 				vdev_parity = fnvlist_lookup_uint64(nv,
3209 				    ZPOOL_CONFIG_NPARITY);
3210 				if ((int)vdev_parity != parity) {
3211 					free(type);
3212 					break;
3213 				}
3214 			}
3215 
3216 			free(type);
3217 			if (errno != 0)
3218 				return (NULL);
3219 
3220 			/*
3221 			 * Now verify that we have the correct vdev id.
3222 			 */
3223 			if (vdev_id == id)
3224 				return (nv);
3225 		}
3226 
3227 		/*
3228 		 * Common case
3229 		 */
3230 		if (strcmp(srchval, val) == 0)
3231 			return (nv);
3232 		break;
3233 	}
3234 
3235 	default:
3236 		break;
3237 	}
3238 
3239 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
3240 	    &child, &children) != 0)
3241 		return (NULL);
3242 
3243 	for (c = 0; c < children; c++) {
3244 		if ((ret = vdev_to_nvlist_iter(child[c], search,
3245 		    avail_spare, l2cache, NULL, return_parent)) != NULL) {
3246 			/*
3247 			 * The 'is_log' value is only set for the toplevel
3248 			 * vdev, not the leaf vdevs.  So we always lookup the
3249 			 * log device from the root of the vdev tree (where
3250 			 * 'log' is non-NULL).
3251 			 */
3252 			if (log != NULL &&
3253 			    nvlist_lookup_uint64(child[c],
3254 			    ZPOOL_CONFIG_IS_LOG, &is_log) == 0 &&
3255 			    is_log) {
3256 				*log = B_TRUE;
3257 			}
3258 			return (ret && return_parent && !is_root ? nv : ret);
3259 		}
3260 	}
3261 
3262 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
3263 	    &child, &children) == 0) {
3264 		for (c = 0; c < children; c++) {
3265 			if ((ret = vdev_to_nvlist_iter(child[c], search,
3266 			    avail_spare, l2cache, NULL, return_parent))
3267 			    != NULL) {
3268 				*avail_spare = B_TRUE;
3269 				return (ret && return_parent &&
3270 				    !is_root ? nv : ret);
3271 			}
3272 		}
3273 	}
3274 
3275 	if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
3276 	    &child, &children) == 0) {
3277 		for (c = 0; c < children; c++) {
3278 			if ((ret = vdev_to_nvlist_iter(child[c], search,
3279 			    avail_spare, l2cache, NULL, return_parent))
3280 			    != NULL) {
3281 				*l2cache = B_TRUE;
3282 				return (ret && return_parent &&
3283 				    !is_root ? nv : ret);
3284 			}
3285 		}
3286 	}
3287 
3288 	return (NULL);
3289 }
3290 
3291 /*
3292  * Given a physical path or guid, find the associated vdev.
3293  */
3294 nvlist_t *
zpool_find_vdev_by_physpath(zpool_handle_t * zhp,const char * ppath,boolean_t * avail_spare,boolean_t * l2cache,boolean_t * log)3295 zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath,
3296     boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log)
3297 {
3298 	nvlist_t *search, *nvroot, *ret;
3299 	uint64_t guid;
3300 	char *end;
3301 
3302 	search = fnvlist_alloc();
3303 
3304 	guid = strtoull(ppath, &end, 0);
3305 	if (guid != 0 && *end == '\0') {
3306 		fnvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid);
3307 	} else {
3308 		fnvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath);
3309 	}
3310 
3311 	nvroot = fnvlist_lookup_nvlist(zhp->zpool_config,
3312 	    ZPOOL_CONFIG_VDEV_TREE);
3313 
3314 	*avail_spare = B_FALSE;
3315 	*l2cache = B_FALSE;
3316 	if (log != NULL)
3317 		*log = B_FALSE;
3318 	ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log,
3319 	    B_FALSE);
3320 	fnvlist_free(search);
3321 
3322 	return (ret);
3323 }
3324 
3325 /*
3326  * Determine if we have an "interior" top-level vdev (i.e mirror/raidz).
3327  */
3328 static boolean_t
zpool_vdev_is_interior(const char * name)3329 zpool_vdev_is_interior(const char *name)
3330 {
3331 	if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 ||
3332 	    strncmp(name, VDEV_TYPE_SPARE, strlen(VDEV_TYPE_SPARE)) == 0 ||
3333 	    strncmp(name,
3334 	    VDEV_TYPE_REPLACING, strlen(VDEV_TYPE_REPLACING)) == 0 ||
3335 	    strncmp(name, VDEV_TYPE_ROOT, strlen(VDEV_TYPE_ROOT)) == 0 ||
3336 	    strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0)
3337 		return (B_TRUE);
3338 
3339 	if (strncmp(name, VDEV_TYPE_DRAID, strlen(VDEV_TYPE_DRAID)) == 0 &&
3340 	    !zpool_is_draid_spare(name))
3341 		return (B_TRUE);
3342 
3343 	return (B_FALSE);
3344 }
3345 
3346 /*
3347  * Lookup the nvlist for a given vdev or vdev's parent (depending on
3348  * if 'return_parent' is set).
3349  */
3350 static nvlist_t *
__zpool_find_vdev(zpool_handle_t * zhp,const char * path,boolean_t * avail_spare,boolean_t * l2cache,boolean_t * log,boolean_t return_parent)3351 __zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare,
3352     boolean_t *l2cache, boolean_t *log, boolean_t return_parent)
3353 {
3354 	char *end;
3355 	nvlist_t *nvroot, *search, *ret;
3356 	uint64_t guid;
3357 	boolean_t __avail_spare, __l2cache, __log;
3358 
3359 	search = fnvlist_alloc();
3360 
3361 	guid = strtoull(path, &end, 0);
3362 	if (guid != 0 && *end == '\0') {
3363 		fnvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid);
3364 	} else if (zpool_vdev_is_interior(path)) {
3365 		fnvlist_add_string(search, ZPOOL_CONFIG_TYPE, path);
3366 	} else {
3367 		fnvlist_add_string(search, ZPOOL_CONFIG_PATH, path);
3368 	}
3369 
3370 	nvroot = fnvlist_lookup_nvlist(zhp->zpool_config,
3371 	    ZPOOL_CONFIG_VDEV_TREE);
3372 
3373 	/*
3374 	 * User can pass NULL for avail_spare, l2cache, and log, but
3375 	 * we still need to provide variables to vdev_to_nvlist_iter(), so
3376 	 * just point them to junk variables here.
3377 	 */
3378 	if (!avail_spare)
3379 		avail_spare = &__avail_spare;
3380 	if (!l2cache)
3381 		l2cache = &__l2cache;
3382 	if (!log)
3383 		log = &__log;
3384 
3385 	*avail_spare = B_FALSE;
3386 	*l2cache = B_FALSE;
3387 	if (log != NULL)
3388 		*log = B_FALSE;
3389 	ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log,
3390 	    return_parent);
3391 	fnvlist_free(search);
3392 
3393 	return (ret);
3394 }
3395 
3396 nvlist_t *
zpool_find_vdev(zpool_handle_t * zhp,const char * path,boolean_t * avail_spare,boolean_t * l2cache,boolean_t * log)3397 zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare,
3398     boolean_t *l2cache, boolean_t *log)
3399 {
3400 	return (__zpool_find_vdev(zhp, path, avail_spare, l2cache, log,
3401 	    B_FALSE));
3402 }
3403 
3404 /* Given a vdev path, return its parent's nvlist */
3405 nvlist_t *
zpool_find_parent_vdev(zpool_handle_t * zhp,const char * path,boolean_t * avail_spare,boolean_t * l2cache,boolean_t * log)3406 zpool_find_parent_vdev(zpool_handle_t *zhp, const char *path,
3407     boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log)
3408 {
3409 	return (__zpool_find_vdev(zhp, path, avail_spare, l2cache, log,
3410 	    B_TRUE));
3411 }
3412 
3413 /*
3414  * Convert a vdev path to a GUID.  Returns GUID or 0 on error.
3415  *
3416  * If is_spare, is_l2cache, or is_log is non-NULL, then store within it
3417  * if the VDEV is a spare, l2cache, or log device.  If they're NULL then
3418  * ignore them.
3419  */
3420 static uint64_t
zpool_vdev_path_to_guid_impl(zpool_handle_t * zhp,const char * path,boolean_t * is_spare,boolean_t * is_l2cache,boolean_t * is_log)3421 zpool_vdev_path_to_guid_impl(zpool_handle_t *zhp, const char *path,
3422     boolean_t *is_spare, boolean_t *is_l2cache, boolean_t *is_log)
3423 {
3424 	boolean_t spare = B_FALSE, l2cache = B_FALSE, log = B_FALSE;
3425 	nvlist_t *tgt;
3426 
3427 	if ((tgt = zpool_find_vdev(zhp, path, &spare, &l2cache,
3428 	    &log)) == NULL)
3429 		return (0);
3430 
3431 	if (is_spare != NULL)
3432 		*is_spare = spare;
3433 	if (is_l2cache != NULL)
3434 		*is_l2cache = l2cache;
3435 	if (is_log != NULL)
3436 		*is_log = log;
3437 
3438 	return (fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID));
3439 }
3440 
3441 /* Convert a vdev path to a GUID.  Returns GUID or 0 on error. */
3442 uint64_t
zpool_vdev_path_to_guid(zpool_handle_t * zhp,const char * path)3443 zpool_vdev_path_to_guid(zpool_handle_t *zhp, const char *path)
3444 {
3445 	return (zpool_vdev_path_to_guid_impl(zhp, path, NULL, NULL, NULL));
3446 }
3447 
3448 /*
3449  * Bring the specified vdev online.   The 'flags' parameter is a set of the
3450  * ZFS_ONLINE_* flags.
3451  */
3452 int
zpool_vdev_online(zpool_handle_t * zhp,const char * path,int flags,vdev_state_t * newstate)3453 zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags,
3454     vdev_state_t *newstate)
3455 {
3456 	zfs_cmd_t zc = {"\0"};
3457 	char errbuf[ERRBUFLEN];
3458 	nvlist_t *tgt;
3459 	boolean_t avail_spare, l2cache, islog;
3460 	libzfs_handle_t *hdl = zhp->zpool_hdl;
3461 
3462 	if (flags & ZFS_ONLINE_EXPAND) {
3463 		(void) snprintf(errbuf, sizeof (errbuf),
3464 		    dgettext(TEXT_DOMAIN, "cannot expand %s"), path);
3465 	} else {
3466 		(void) snprintf(errbuf, sizeof (errbuf),
3467 		    dgettext(TEXT_DOMAIN, "cannot online %s"), path);
3468 	}
3469 
3470 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3471 	if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
3472 	    &islog)) == NULL)
3473 		return (zfs_error(hdl, EZFS_NODEVICE, errbuf));
3474 
3475 	zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
3476 
3477 	if (!(flags & ZFS_ONLINE_SPARE) && avail_spare)
3478 		return (zfs_error(hdl, EZFS_ISSPARE, errbuf));
3479 
3480 #ifndef __FreeBSD__
3481 	const char *pathname;
3482 	if ((flags & ZFS_ONLINE_EXPAND ||
3483 	    zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) &&
3484 	    nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH, &pathname) == 0) {
3485 		uint64_t wholedisk = 0;
3486 
3487 		(void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK,
3488 		    &wholedisk);
3489 
3490 		/*
3491 		 * XXX - L2ARC 1.0 devices can't support expansion.
3492 		 */
3493 		if (l2cache) {
3494 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3495 			    "cannot expand cache devices"));
3496 			return (zfs_error(hdl, EZFS_VDEVNOTSUP, errbuf));
3497 		}
3498 
3499 		if (wholedisk) {
3500 			const char *fullpath = path;
3501 			char buf[MAXPATHLEN];
3502 			int error;
3503 
3504 			if (path[0] != '/') {
3505 				error = zfs_resolve_shortname(path, buf,
3506 				    sizeof (buf));
3507 				if (error != 0)
3508 					return (zfs_error(hdl, EZFS_NODEVICE,
3509 					    errbuf));
3510 
3511 				fullpath = buf;
3512 			}
3513 
3514 			error = zpool_relabel_disk(hdl, fullpath, errbuf);
3515 			if (error != 0)
3516 				return (error);
3517 		}
3518 	}
3519 #endif
3520 
3521 	zc.zc_cookie = VDEV_STATE_ONLINE;
3522 	zc.zc_obj = flags;
3523 
3524 	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) {
3525 		if (errno == EINVAL) {
3526 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split "
3527 			    "from this pool into a new one.  Use '%s' "
3528 			    "instead"), "zpool detach");
3529 			return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, errbuf));
3530 		}
3531 		return (zpool_standard_error(hdl, errno, errbuf));
3532 	}
3533 
3534 	*newstate = zc.zc_cookie;
3535 	return (0);
3536 }
3537 
3538 /*
3539  * Take the specified vdev offline
3540  */
3541 int
zpool_vdev_offline(zpool_handle_t * zhp,const char * path,boolean_t istmp)3542 zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp)
3543 {
3544 	zfs_cmd_t zc = {"\0"};
3545 	char errbuf[ERRBUFLEN];
3546 	nvlist_t *tgt;
3547 	boolean_t avail_spare, l2cache;
3548 	libzfs_handle_t *hdl = zhp->zpool_hdl;
3549 
3550 	(void) snprintf(errbuf, sizeof (errbuf),
3551 	    dgettext(TEXT_DOMAIN, "cannot offline %s"), path);
3552 
3553 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3554 	if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
3555 	    NULL)) == NULL)
3556 		return (zfs_error(hdl, EZFS_NODEVICE, errbuf));
3557 
3558 	zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
3559 
3560 	if (avail_spare)
3561 		return (zfs_error(hdl, EZFS_ISSPARE, errbuf));
3562 
3563 	zc.zc_cookie = VDEV_STATE_OFFLINE;
3564 	zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0;
3565 
3566 	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
3567 		return (0);
3568 
3569 	switch (errno) {
3570 	case EBUSY:
3571 
3572 		/*
3573 		 * There are no other replicas of this device.
3574 		 */
3575 		return (zfs_error(hdl, EZFS_NOREPLICAS, errbuf));
3576 
3577 	case EEXIST:
3578 		/*
3579 		 * The log device has unplayed logs
3580 		 */
3581 		return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, errbuf));
3582 
3583 	default:
3584 		return (zpool_standard_error(hdl, errno, errbuf));
3585 	}
3586 }
3587 
3588 /*
3589  * Remove the specified vdev asynchronously from the configuration, so
3590  * that it may come ONLINE if reinserted. This is called from zed on
3591  * Udev remove event.
3592  * Note: We also have a similar function zpool_vdev_remove() that
3593  * removes the vdev from the pool.
3594  */
3595 int
zpool_vdev_remove_wanted(zpool_handle_t * zhp,const char * path)3596 zpool_vdev_remove_wanted(zpool_handle_t *zhp, const char *path)
3597 {
3598 	zfs_cmd_t zc = {"\0"};
3599 	char errbuf[ERRBUFLEN];
3600 	nvlist_t *tgt;
3601 	boolean_t avail_spare, l2cache;
3602 	libzfs_handle_t *hdl = zhp->zpool_hdl;
3603 
3604 	(void) snprintf(errbuf, sizeof (errbuf),
3605 	    dgettext(TEXT_DOMAIN, "cannot remove %s"), path);
3606 
3607 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3608 	if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
3609 	    NULL)) == NULL)
3610 		return (zfs_error(hdl, EZFS_NODEVICE, errbuf));
3611 
3612 	zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
3613 
3614 	zc.zc_cookie = VDEV_STATE_REMOVED;
3615 
3616 	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
3617 		return (0);
3618 
3619 	return (zpool_standard_error(hdl, errno, errbuf));
3620 }
3621 
3622 /*
3623  * Mark the given vdev faulted.
3624  */
3625 int
zpool_vdev_fault(zpool_handle_t * zhp,uint64_t guid,vdev_aux_t aux)3626 zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
3627 {
3628 	zfs_cmd_t zc = {"\0"};
3629 	char errbuf[ERRBUFLEN];
3630 	libzfs_handle_t *hdl = zhp->zpool_hdl;
3631 	nvlist_t *vdev_nv;
3632 	boolean_t avail_spare, l2cache;
3633 	char *vdev_name;
3634 	char guid_str[21]; /* 64-bit num + '\0' */
3635 	boolean_t is_draid_spare = B_FALSE;
3636 	const char *vdev_type;
3637 
3638 	(void) snprintf(errbuf, sizeof (errbuf),
3639 	    dgettext(TEXT_DOMAIN, "cannot fault %llu"), (u_longlong_t)guid);
3640 
3641 	snprintf(guid_str, sizeof (guid_str), "%llu", (u_longlong_t)guid);
3642 	if ((vdev_nv = zpool_find_vdev(zhp, guid_str, &avail_spare,
3643 	    &l2cache, NULL)) == NULL)
3644 		return (zfs_error(hdl, EZFS_NODEVICE, errbuf));
3645 
3646 	vdev_name = zpool_vdev_name(hdl, zhp, vdev_nv, 0);
3647 	if (vdev_name != NULL) {
3648 		/*
3649 		 * We have the actual vdev name, so use that instead of the GUID
3650 		 * in any error messages.
3651 		 */
3652 		(void) snprintf(errbuf, sizeof (errbuf),
3653 		    dgettext(TEXT_DOMAIN, "cannot fault %s"), vdev_name);
3654 		free(vdev_name);
3655 	}
3656 
3657 	/*
3658 	 * Spares (traditional or draid) cannot be faulted by libzfs, except:
3659 	 *
3660 	 * - Any spare type that exceeds it's errors can be faulted (aux =
3661 	 *   VDEV_AUX_ERR_EXCEEDED).  This is only used by zed.
3662 	 *
3663 	 * - Traditional spares that are active can be force faulted.
3664 	 */
3665 	if (nvlist_lookup_string(vdev_nv, ZPOOL_CONFIG_TYPE, &vdev_type) == 0)
3666 		if (strcmp(vdev_type, VDEV_TYPE_DRAID_SPARE) == 0)
3667 			is_draid_spare = B_TRUE;
3668 
3669 	/*
3670 	 * If vdev is a spare that is not being used, or is a dRAID spare (in
3671 	 * use or not), then don't allow it to be force-faulted.  However, an
3672 	 * in-use dRAID spare can be faulted by ZED if see too many errors
3673 	 * (aux = VDEV_AUX_ERR_EXCEEDED).
3674 	 */
3675 	if (avail_spare || (is_draid_spare && aux != VDEV_AUX_ERR_EXCEEDED))
3676 		return (zfs_error(hdl, EZFS_ISSPARE, errbuf));
3677 
3678 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3679 	zc.zc_guid = guid;
3680 	zc.zc_cookie = VDEV_STATE_FAULTED;
3681 	zc.zc_obj = aux;
3682 
3683 	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
3684 		return (0);
3685 
3686 	switch (errno) {
3687 	case EBUSY:
3688 
3689 		/*
3690 		 * There are no other replicas of this device.
3691 		 */
3692 		return (zfs_error(hdl, EZFS_NOREPLICAS, errbuf));
3693 
3694 	default:
3695 		return (zpool_standard_error(hdl, errno, errbuf));
3696 	}
3697 
3698 }
3699 
3700 /*
3701  * Generic set vdev state function
3702  */
3703 static int
zpool_vdev_set_state(zpool_handle_t * zhp,uint64_t guid,vdev_aux_t aux,vdev_state_t state)3704 zpool_vdev_set_state(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux,
3705     vdev_state_t state)
3706 {
3707 	zfs_cmd_t zc = {"\0"};
3708 	char errbuf[ERRBUFLEN];
3709 	libzfs_handle_t *hdl = zhp->zpool_hdl;
3710 
3711 	(void) snprintf(errbuf, sizeof (errbuf),
3712 	    dgettext(TEXT_DOMAIN, "cannot set %s %llu"),
3713 	    zpool_state_to_name(state, aux), (u_longlong_t)guid);
3714 
3715 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3716 	zc.zc_guid = guid;
3717 	zc.zc_cookie = state;
3718 	zc.zc_obj = aux;
3719 
3720 	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
3721 		return (0);
3722 
3723 	return (zpool_standard_error(hdl, errno, errbuf));
3724 }
3725 
3726 /*
3727  * Mark the given vdev degraded.
3728  */
3729 int
zpool_vdev_degrade(zpool_handle_t * zhp,uint64_t guid,vdev_aux_t aux)3730 zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
3731 {
3732 	return (zpool_vdev_set_state(zhp, guid, aux, VDEV_STATE_DEGRADED));
3733 }
3734 
3735 /*
3736  * Mark the given vdev as in a removed state (as if the device does not exist).
3737  *
3738  * This is different than zpool_vdev_remove() which does a removal of a device
3739  * from the pool (but the device does exist).
3740  */
3741 int
zpool_vdev_set_removed_state(zpool_handle_t * zhp,uint64_t guid,vdev_aux_t aux)3742 zpool_vdev_set_removed_state(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
3743 {
3744 	return (zpool_vdev_set_state(zhp, guid, aux, VDEV_STATE_REMOVED));
3745 }
3746 
3747 /*
3748  * Returns TRUE if the given nvlist is a vdev that was originally swapped in as
3749  * a hot spare.
3750  */
3751 static boolean_t
is_replacing_spare(nvlist_t * search,nvlist_t * tgt,int which)3752 is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which)
3753 {
3754 	nvlist_t **child;
3755 	uint_t c, children;
3756 
3757 	if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child,
3758 	    &children) == 0) {
3759 		const char *type = fnvlist_lookup_string(search,
3760 		    ZPOOL_CONFIG_TYPE);
3761 		if ((strcmp(type, VDEV_TYPE_SPARE) == 0 ||
3762 		    strcmp(type, VDEV_TYPE_DRAID_SPARE) == 0) &&
3763 		    children == 2 && child[which] == tgt)
3764 			return (B_TRUE);
3765 
3766 		for (c = 0; c < children; c++)
3767 			if (is_replacing_spare(child[c], tgt, which))
3768 				return (B_TRUE);
3769 	}
3770 
3771 	return (B_FALSE);
3772 }
3773 
3774 /*
3775  * Attach new_disk (fully described by nvroot) to old_disk.
3776  * If 'replacing' is specified, the new disk will replace the old one.
3777  */
3778 int
zpool_vdev_attach(zpool_handle_t * zhp,const char * old_disk,const char * new_disk,nvlist_t * nvroot,int replacing,boolean_t rebuild)3779 zpool_vdev_attach(zpool_handle_t *zhp, const char *old_disk,
3780     const char *new_disk, nvlist_t *nvroot, int replacing, boolean_t rebuild)
3781 {
3782 	zfs_cmd_t zc = {"\0"};
3783 	char errbuf[ERRBUFLEN];
3784 	int ret;
3785 	nvlist_t *tgt;
3786 	boolean_t avail_spare, l2cache, islog;
3787 	uint64_t val;
3788 	char *newname;
3789 	const char *type;
3790 	nvlist_t **child;
3791 	uint_t children;
3792 	nvlist_t *config_root;
3793 	libzfs_handle_t *hdl = zhp->zpool_hdl;
3794 
3795 	if (replacing)
3796 		(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
3797 		    "cannot replace %s with %s"), old_disk, new_disk);
3798 	else
3799 		(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
3800 		    "cannot attach %s to %s"), new_disk, old_disk);
3801 
3802 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3803 	if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache,
3804 	    &islog)) == NULL)
3805 		return (zfs_error(hdl, EZFS_NODEVICE, errbuf));
3806 
3807 	if (avail_spare)
3808 		return (zfs_error(hdl, EZFS_ISSPARE, errbuf));
3809 
3810 	if (l2cache)
3811 		return (zfs_error(hdl, EZFS_ISL2CACHE, errbuf));
3812 
3813 	zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
3814 	zc.zc_cookie = replacing;
3815 	zc.zc_simple = rebuild;
3816 
3817 	if (rebuild &&
3818 	    zfeature_lookup_guid("org.openzfs:device_rebuild", NULL) != 0) {
3819 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3820 		    "the loaded zfs module doesn't support device rebuilds"));
3821 		return (zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf));
3822 	}
3823 
3824 	type = fnvlist_lookup_string(tgt, ZPOOL_CONFIG_TYPE);
3825 	if (strcmp(type, VDEV_TYPE_RAIDZ) == 0 &&
3826 	    zfeature_lookup_guid("org.openzfs:raidz_expansion", NULL) != 0) {
3827 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3828 		    "the loaded zfs module doesn't support raidz expansion"));
3829 		return (zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf));
3830 	}
3831 
3832 	if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
3833 	    &child, &children) != 0 || children != 1) {
3834 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3835 		    "new device must be a single disk"));
3836 		return (zfs_error(hdl, EZFS_INVALCONFIG, errbuf));
3837 	}
3838 
3839 	config_root = fnvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
3840 	    ZPOOL_CONFIG_VDEV_TREE);
3841 
3842 	if ((newname = zpool_vdev_name(NULL, NULL, child[0], 0)) == NULL)
3843 		return (-1);
3844 
3845 	/*
3846 	 * If the target is a hot spare that has been swapped in, we can only
3847 	 * replace it with another hot spare.
3848 	 */
3849 	if (replacing &&
3850 	    nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 &&
3851 	    (zpool_find_vdev(zhp, newname, &avail_spare, &l2cache,
3852 	    NULL) == NULL || !avail_spare) &&
3853 	    is_replacing_spare(config_root, tgt, 1)) {
3854 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3855 		    "can only be replaced by another hot spare"));
3856 		free(newname);
3857 		return (zfs_error(hdl, EZFS_BADTARGET, errbuf));
3858 	}
3859 
3860 	free(newname);
3861 
3862 	zcmd_write_conf_nvlist(hdl, &zc, nvroot);
3863 
3864 	ret = zfs_ioctl(hdl, ZFS_IOC_VDEV_ATTACH, &zc);
3865 
3866 	zcmd_free_nvlists(&zc);
3867 
3868 	if (ret == 0)
3869 		return (0);
3870 
3871 	switch (errno) {
3872 	case ENOTSUP:
3873 		/*
3874 		 * Can't attach to or replace this type of vdev.
3875 		 */
3876 		if (replacing) {
3877 			uint64_t version = zpool_get_prop_int(zhp,
3878 			    ZPOOL_PROP_VERSION, NULL);
3879 
3880 			if (islog) {
3881 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3882 				    "cannot replace a log with a spare"));
3883 			} else if (rebuild) {
3884 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3885 				    "only mirror and dRAID vdevs support "
3886 				    "sequential reconstruction"));
3887 			} else if (zpool_is_draid_spare(new_disk)) {
3888 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3889 				    "dRAID spares can only replace child "
3890 				    "devices in their parent's dRAID vdev"));
3891 			} else if (version >= SPA_VERSION_MULTI_REPLACE) {
3892 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3893 				    "already in replacing/spare config; wait "
3894 				    "for completion or use 'zpool detach'"));
3895 			} else {
3896 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3897 				    "cannot replace a replacing device"));
3898 			}
3899 		} else if (strcmp(type, VDEV_TYPE_RAIDZ) == 0) {
3900 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3901 			    "raidz_expansion feature must be enabled "
3902 			    "in order to attach a device to raidz"));
3903 		} else {
3904 			char status[64] = {0};
3905 			zpool_prop_get_feature(zhp,
3906 			    "feature@device_rebuild", status, 63);
3907 			if (rebuild &&
3908 			    strncmp(status, ZFS_FEATURE_DISABLED, 64) == 0) {
3909 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3910 				    "device_rebuild feature must be enabled "
3911 				    "in order to use sequential "
3912 				    "reconstruction"));
3913 			} else {
3914 				zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3915 				    "can only attach to mirrors and top-level "
3916 				    "disks"));
3917 			}
3918 		}
3919 		(void) zfs_error(hdl, EZFS_BADTARGET, errbuf);
3920 		break;
3921 
3922 	case EINVAL:
3923 		/*
3924 		 * The new device must be a single disk.
3925 		 */
3926 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3927 		    "new device must be a single disk"));
3928 		(void) zfs_error(hdl, EZFS_INVALCONFIG, errbuf);
3929 		break;
3930 
3931 	case EBUSY:
3932 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy"),
3933 		    new_disk);
3934 		(void) zfs_error(hdl, EZFS_BADDEV, errbuf);
3935 		break;
3936 
3937 	case EOVERFLOW:
3938 		/*
3939 		 * The new device is too small.
3940 		 */
3941 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3942 		    "device is too small"));
3943 		(void) zfs_error(hdl, EZFS_BADDEV, errbuf);
3944 		break;
3945 
3946 	case EDOM:
3947 		/*
3948 		 * The new device has a different optimal sector size.
3949 		 */
3950 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3951 		    "new device has a different optimal sector size; use the "
3952 		    "option '-o ashift=N' to override the optimal size"));
3953 		(void) zfs_error(hdl, EZFS_BADDEV, errbuf);
3954 		break;
3955 
3956 	case ENAMETOOLONG:
3957 		/*
3958 		 * The resulting top-level vdev spec won't fit in the label.
3959 		 */
3960 		(void) zfs_error(hdl, EZFS_DEVOVERFLOW, errbuf);
3961 		break;
3962 
3963 	case ENXIO:
3964 		/*
3965 		 * The existing raidz vdev has offline children
3966 		 */
3967 		if (strcmp(type, VDEV_TYPE_RAIDZ) == 0) {
3968 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3969 			    "raidz vdev has devices that are are offline or "
3970 			    "being replaced"));
3971 			(void) zfs_error(hdl, EZFS_BADDEV, errbuf);
3972 			break;
3973 		} else {
3974 			(void) zpool_standard_error(hdl, errno, errbuf);
3975 		}
3976 		break;
3977 
3978 	case EADDRINUSE:
3979 		/*
3980 		 * The boot reserved area is already being used (FreeBSD)
3981 		 */
3982 		if (strcmp(type, VDEV_TYPE_RAIDZ) == 0) {
3983 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3984 			    "the reserved boot area needed for the expansion "
3985 			    "is already being used by a boot loader"));
3986 			(void) zfs_error(hdl, EZFS_BADDEV, errbuf);
3987 		} else {
3988 			(void) zpool_standard_error(hdl, errno, errbuf);
3989 		}
3990 		break;
3991 
3992 	case ZFS_ERR_ASHIFT_MISMATCH:
3993 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3994 		    "The new device cannot have a higher alignment requirement "
3995 		    "than the top-level vdev."));
3996 		(void) zfs_error(hdl, EZFS_BADTARGET, errbuf);
3997 		break;
3998 	default:
3999 		(void) zpool_standard_error(hdl, errno, errbuf);
4000 	}
4001 
4002 	return (-1);
4003 }
4004 
4005 /*
4006  * Detach the specified device.
4007  */
4008 int
zpool_vdev_detach(zpool_handle_t * zhp,const char * path)4009 zpool_vdev_detach(zpool_handle_t *zhp, const char *path)
4010 {
4011 	zfs_cmd_t zc = {"\0"};
4012 	char errbuf[ERRBUFLEN];
4013 	nvlist_t *tgt;
4014 	boolean_t avail_spare, l2cache;
4015 	libzfs_handle_t *hdl = zhp->zpool_hdl;
4016 
4017 	(void) snprintf(errbuf, sizeof (errbuf),
4018 	    dgettext(TEXT_DOMAIN, "cannot detach %s"), path);
4019 
4020 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4021 	if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
4022 	    NULL)) == NULL)
4023 		return (zfs_error(hdl, EZFS_NODEVICE, errbuf));
4024 
4025 	if (avail_spare)
4026 		return (zfs_error(hdl, EZFS_ISSPARE, errbuf));
4027 
4028 	if (l2cache)
4029 		return (zfs_error(hdl, EZFS_ISL2CACHE, errbuf));
4030 
4031 	zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
4032 
4033 	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0)
4034 		return (0);
4035 
4036 	switch (errno) {
4037 
4038 	case ENOTSUP:
4039 		/*
4040 		 * Can't detach from this type of vdev.
4041 		 */
4042 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only "
4043 		    "applicable to mirror and replacing vdevs"));
4044 		(void) zfs_error(hdl, EZFS_BADTARGET, errbuf);
4045 		break;
4046 
4047 	case EBUSY:
4048 		/*
4049 		 * There are no other replicas of this device.
4050 		 */
4051 		(void) zfs_error(hdl, EZFS_NOREPLICAS, errbuf);
4052 		break;
4053 
4054 	default:
4055 		(void) zpool_standard_error(hdl, errno, errbuf);
4056 	}
4057 
4058 	return (-1);
4059 }
4060 
4061 /*
4062  * Find a mirror vdev in the source nvlist.
4063  *
4064  * The mchild array contains a list of disks in one of the top-level mirrors
4065  * of the source pool.  The schild array contains a list of disks that the
4066  * user specified on the command line.  We loop over the mchild array to
4067  * see if any entry in the schild array matches.
4068  *
4069  * If a disk in the mchild array is found in the schild array, we return
4070  * the index of that entry.  Otherwise we return -1.
4071  */
4072 static int
find_vdev_entry(zpool_handle_t * zhp,nvlist_t ** mchild,uint_t mchildren,nvlist_t ** schild,uint_t schildren)4073 find_vdev_entry(zpool_handle_t *zhp, nvlist_t **mchild, uint_t mchildren,
4074     nvlist_t **schild, uint_t schildren)
4075 {
4076 	uint_t mc;
4077 
4078 	for (mc = 0; mc < mchildren; mc++) {
4079 		uint_t sc;
4080 		char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp,
4081 		    mchild[mc], 0);
4082 
4083 		for (sc = 0; sc < schildren; sc++) {
4084 			char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp,
4085 			    schild[sc], 0);
4086 			boolean_t result = (strcmp(mpath, spath) == 0);
4087 
4088 			free(spath);
4089 			if (result) {
4090 				free(mpath);
4091 				return (mc);
4092 			}
4093 		}
4094 
4095 		free(mpath);
4096 	}
4097 
4098 	return (-1);
4099 }
4100 
4101 /*
4102  * Split a mirror pool.  If newroot points to null, then a new nvlist
4103  * is generated and it is the responsibility of the caller to free it.
4104  */
4105 int
zpool_vdev_split(zpool_handle_t * zhp,char * newname,nvlist_t ** newroot,nvlist_t * props,splitflags_t flags)4106 zpool_vdev_split(zpool_handle_t *zhp, char *newname, nvlist_t **newroot,
4107     nvlist_t *props, splitflags_t flags)
4108 {
4109 	zfs_cmd_t zc = {"\0"};
4110 	char errbuf[ERRBUFLEN];
4111 	const char *bias;
4112 	nvlist_t *tree, *config, **child, **newchild, *newconfig = NULL;
4113 	nvlist_t **varray = NULL, *zc_props = NULL;
4114 	uint_t c, children, newchildren, lastlog = 0, vcount, found = 0;
4115 	libzfs_handle_t *hdl = zhp->zpool_hdl;
4116 	uint64_t vers, readonly = B_FALSE;
4117 	boolean_t freelist = B_FALSE, memory_err = B_TRUE;
4118 	int retval = 0;
4119 
4120 	(void) snprintf(errbuf, sizeof (errbuf),
4121 	    dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name);
4122 
4123 	if (!zpool_name_valid(hdl, B_FALSE, newname))
4124 		return (zfs_error(hdl, EZFS_INVALIDNAME, errbuf));
4125 
4126 	if ((config = zpool_get_config(zhp, NULL)) == NULL) {
4127 		(void) fprintf(stderr, gettext("Internal error: unable to "
4128 		    "retrieve pool configuration\n"));
4129 		return (-1);
4130 	}
4131 
4132 	tree = fnvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE);
4133 	vers = fnvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION);
4134 
4135 	if (props) {
4136 		prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
4137 		if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name,
4138 		    props, vers, flags, errbuf)) == NULL)
4139 			return (-1);
4140 		(void) nvlist_lookup_uint64(zc_props,
4141 		    zpool_prop_to_name(ZPOOL_PROP_READONLY), &readonly);
4142 		if (readonly) {
4143 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4144 			    "property %s can only be set at import time"),
4145 			    zpool_prop_to_name(ZPOOL_PROP_READONLY));
4146 			return (-1);
4147 		}
4148 	}
4149 
4150 	if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child,
4151 	    &children) != 0) {
4152 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4153 		    "Source pool is missing vdev tree"));
4154 		nvlist_free(zc_props);
4155 		return (-1);
4156 	}
4157 
4158 	varray = zfs_alloc(hdl, children * sizeof (nvlist_t *));
4159 	vcount = 0;
4160 
4161 	if (*newroot == NULL ||
4162 	    nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN,
4163 	    &newchild, &newchildren) != 0)
4164 		newchildren = 0;
4165 
4166 	for (c = 0; c < children; c++) {
4167 		uint64_t is_log = B_FALSE, is_hole = B_FALSE;
4168 		boolean_t is_special = B_FALSE, is_dedup = B_FALSE;
4169 		const char *type;
4170 		nvlist_t **mchild, *vdev;
4171 		uint_t mchildren;
4172 		int entry;
4173 
4174 		/*
4175 		 * Unlike cache & spares, slogs are stored in the
4176 		 * ZPOOL_CONFIG_CHILDREN array.  We filter them out here.
4177 		 */
4178 		(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
4179 		    &is_log);
4180 		(void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
4181 		    &is_hole);
4182 		if (is_log || is_hole) {
4183 			/*
4184 			 * Create a hole vdev and put it in the config.
4185 			 */
4186 			if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0)
4187 				goto out;
4188 			if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE,
4189 			    VDEV_TYPE_HOLE) != 0)
4190 				goto out;
4191 			if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE,
4192 			    1) != 0)
4193 				goto out;
4194 			if (lastlog == 0)
4195 				lastlog = vcount;
4196 			varray[vcount++] = vdev;
4197 			continue;
4198 		}
4199 		lastlog = 0;
4200 		type = fnvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE);
4201 
4202 		if (strcmp(type, VDEV_TYPE_INDIRECT) == 0) {
4203 			vdev = child[c];
4204 			if (nvlist_dup(vdev, &varray[vcount++], 0) != 0)
4205 				goto out;
4206 			continue;
4207 		} else if (strcmp(type, VDEV_TYPE_MIRROR) != 0) {
4208 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4209 			    "Source pool must be composed only of mirrors\n"));
4210 			retval = zfs_error(hdl, EZFS_INVALCONFIG, errbuf);
4211 			goto out;
4212 		}
4213 
4214 		if (nvlist_lookup_string(child[c],
4215 		    ZPOOL_CONFIG_ALLOCATION_BIAS, &bias) == 0) {
4216 			if (strcmp(bias, VDEV_ALLOC_BIAS_SPECIAL) == 0)
4217 				is_special = B_TRUE;
4218 			else if (strcmp(bias, VDEV_ALLOC_BIAS_DEDUP) == 0)
4219 				is_dedup = B_TRUE;
4220 		}
4221 		verify(nvlist_lookup_nvlist_array(child[c],
4222 		    ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0);
4223 
4224 		/* find or add an entry for this top-level vdev */
4225 		if (newchildren > 0 &&
4226 		    (entry = find_vdev_entry(zhp, mchild, mchildren,
4227 		    newchild, newchildren)) >= 0) {
4228 			/* We found a disk that the user specified. */
4229 			vdev = mchild[entry];
4230 			++found;
4231 		} else {
4232 			/* User didn't specify a disk for this vdev. */
4233 			vdev = mchild[mchildren - 1];
4234 		}
4235 
4236 		if (nvlist_dup(vdev, &varray[vcount++], 0) != 0)
4237 			goto out;
4238 
4239 		if (flags.dryrun != 0) {
4240 			if (is_dedup == B_TRUE) {
4241 				if (nvlist_add_string(varray[vcount - 1],
4242 				    ZPOOL_CONFIG_ALLOCATION_BIAS,
4243 				    VDEV_ALLOC_BIAS_DEDUP) != 0)
4244 					goto out;
4245 			} else if (is_special == B_TRUE) {
4246 				if (nvlist_add_string(varray[vcount - 1],
4247 				    ZPOOL_CONFIG_ALLOCATION_BIAS,
4248 				    VDEV_ALLOC_BIAS_SPECIAL) != 0)
4249 					goto out;
4250 			}
4251 		}
4252 	}
4253 
4254 	/* did we find every disk the user specified? */
4255 	if (found != newchildren) {
4256 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must "
4257 		    "include at most one disk from each mirror"));
4258 		retval = zfs_error(hdl, EZFS_INVALCONFIG, errbuf);
4259 		goto out;
4260 	}
4261 
4262 	/* Prepare the nvlist for populating. */
4263 	if (*newroot == NULL) {
4264 		if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0)
4265 			goto out;
4266 		freelist = B_TRUE;
4267 		if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE,
4268 		    VDEV_TYPE_ROOT) != 0)
4269 			goto out;
4270 	} else {
4271 		verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0);
4272 	}
4273 
4274 	/* Add all the children we found */
4275 	if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN,
4276 	    (const nvlist_t **)varray, lastlog == 0 ? vcount : lastlog) != 0)
4277 		goto out;
4278 
4279 	/*
4280 	 * If we're just doing a dry run, exit now with success.
4281 	 */
4282 	if (flags.dryrun) {
4283 		memory_err = B_FALSE;
4284 		freelist = B_FALSE;
4285 		goto out;
4286 	}
4287 
4288 	/* now build up the config list & call the ioctl */
4289 	if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0)
4290 		goto out;
4291 
4292 	if (nvlist_add_nvlist(newconfig,
4293 	    ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 ||
4294 	    nvlist_add_string(newconfig,
4295 	    ZPOOL_CONFIG_POOL_NAME, newname) != 0 ||
4296 	    nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0)
4297 		goto out;
4298 
4299 	/*
4300 	 * The new pool is automatically part of the namespace unless we
4301 	 * explicitly export it.
4302 	 */
4303 	if (!flags.import)
4304 		zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT;
4305 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4306 	(void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string));
4307 	zcmd_write_conf_nvlist(hdl, &zc, newconfig);
4308 	if (zc_props != NULL)
4309 		zcmd_write_src_nvlist(hdl, &zc, zc_props);
4310 
4311 	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) {
4312 		retval = zpool_standard_error(hdl, errno, errbuf);
4313 		goto out;
4314 	}
4315 
4316 	freelist = B_FALSE;
4317 	memory_err = B_FALSE;
4318 
4319 out:
4320 	if (varray != NULL) {
4321 		int v;
4322 
4323 		for (v = 0; v < vcount; v++)
4324 			nvlist_free(varray[v]);
4325 		free(varray);
4326 	}
4327 	zcmd_free_nvlists(&zc);
4328 	nvlist_free(zc_props);
4329 	nvlist_free(newconfig);
4330 	if (freelist) {
4331 		nvlist_free(*newroot);
4332 		*newroot = NULL;
4333 	}
4334 
4335 	if (retval != 0)
4336 		return (retval);
4337 
4338 	if (memory_err)
4339 		return (no_memory(hdl));
4340 
4341 	return (0);
4342 }
4343 
4344 /*
4345  * Remove the given device.
4346  */
4347 int
zpool_vdev_remove(zpool_handle_t * zhp,const char * path)4348 zpool_vdev_remove(zpool_handle_t *zhp, const char *path)
4349 {
4350 	zfs_cmd_t zc = {"\0"};
4351 	char errbuf[ERRBUFLEN];
4352 	nvlist_t *tgt;
4353 	boolean_t avail_spare, l2cache, islog;
4354 	libzfs_handle_t *hdl = zhp->zpool_hdl;
4355 	uint64_t version;
4356 
4357 	(void) snprintf(errbuf, sizeof (errbuf),
4358 	    dgettext(TEXT_DOMAIN, "cannot remove %s"), path);
4359 
4360 	if (zpool_is_draid_spare(path)) {
4361 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4362 		    "dRAID spares cannot be removed"));
4363 		return (zfs_error(hdl, EZFS_NODEVICE, errbuf));
4364 	}
4365 
4366 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4367 	if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
4368 	    &islog)) == NULL)
4369 		return (zfs_error(hdl, EZFS_NODEVICE, errbuf));
4370 
4371 	version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
4372 	if (islog && version < SPA_VERSION_HOLES) {
4373 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4374 		    "pool must be upgraded to support log removal"));
4375 		return (zfs_error(hdl, EZFS_BADVERSION, errbuf));
4376 	}
4377 
4378 	zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
4379 
4380 	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
4381 		return (0);
4382 
4383 	switch (errno) {
4384 
4385 	case EALREADY:
4386 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4387 		    "removal for this vdev is already in progress."));
4388 		(void) zfs_error(hdl, EZFS_BUSY, errbuf);
4389 		break;
4390 
4391 	case EINVAL:
4392 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4393 		    "invalid config; all top-level vdevs must "
4394 		    "have the same sector size and not be raidz."));
4395 		(void) zfs_error(hdl, EZFS_INVALCONFIG, errbuf);
4396 		break;
4397 
4398 	case EBUSY:
4399 		if (islog) {
4400 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4401 			    "Mount encrypted datasets to replay logs."));
4402 		} else {
4403 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4404 			    "Pool busy; removal may already be in progress"));
4405 		}
4406 		(void) zfs_error(hdl, EZFS_BUSY, errbuf);
4407 		break;
4408 
4409 	case EACCES:
4410 		if (islog) {
4411 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4412 			    "Mount encrypted datasets to replay logs."));
4413 			(void) zfs_error(hdl, EZFS_BUSY, errbuf);
4414 		} else {
4415 			(void) zpool_standard_error(hdl, errno, errbuf);
4416 		}
4417 		break;
4418 
4419 	default:
4420 		(void) zpool_standard_error(hdl, errno, errbuf);
4421 	}
4422 	return (-1);
4423 }
4424 
4425 int
zpool_vdev_remove_cancel(zpool_handle_t * zhp)4426 zpool_vdev_remove_cancel(zpool_handle_t *zhp)
4427 {
4428 	zfs_cmd_t zc = {{0}};
4429 	char errbuf[ERRBUFLEN];
4430 	libzfs_handle_t *hdl = zhp->zpool_hdl;
4431 
4432 	(void) snprintf(errbuf, sizeof (errbuf),
4433 	    dgettext(TEXT_DOMAIN, "cannot cancel removal"));
4434 
4435 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4436 	zc.zc_cookie = 1;
4437 
4438 	if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
4439 		return (0);
4440 
4441 	return (zpool_standard_error(hdl, errno, errbuf));
4442 }
4443 
4444 int
zpool_vdev_indirect_size(zpool_handle_t * zhp,const char * path,uint64_t * sizep)4445 zpool_vdev_indirect_size(zpool_handle_t *zhp, const char *path,
4446     uint64_t *sizep)
4447 {
4448 	char errbuf[ERRBUFLEN];
4449 	nvlist_t *tgt;
4450 	boolean_t avail_spare, l2cache, islog;
4451 	libzfs_handle_t *hdl = zhp->zpool_hdl;
4452 
4453 	(void) snprintf(errbuf, sizeof (errbuf),
4454 	    dgettext(TEXT_DOMAIN, "cannot determine indirect size of %s"),
4455 	    path);
4456 
4457 	if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
4458 	    &islog)) == NULL)
4459 		return (zfs_error(hdl, EZFS_NODEVICE, errbuf));
4460 
4461 	if (avail_spare || l2cache || islog) {
4462 		*sizep = 0;
4463 		return (0);
4464 	}
4465 
4466 	if (nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_INDIRECT_SIZE, sizep) != 0) {
4467 		zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4468 		    "indirect size not available"));
4469 		return (zfs_error(hdl, EINVAL, errbuf));
4470 	}
4471 	return (0);
4472 }
4473 
4474 /*
4475  * Clear the errors for the pool, or the particular device if specified.
4476  */
4477 int
zpool_clear(zpool_handle_t * zhp,const char * path,nvlist_t * rewindnvl)4478 zpool_clear(zpool_handle_t *zhp, const char *path, nvlist_t *rewindnvl)
4479 {
4480 	zfs_cmd_t zc = {"\0"};
4481 	char errbuf[ERRBUFLEN];
4482 	nvlist_t *tgt;
4483 	zpool_load_policy_t policy;
4484 	boolean_t avail_spare, l2cache;
4485 	libzfs_handle_t *hdl = zhp->zpool_hdl;
4486 	nvlist_t *nvi = NULL;
4487 	int error;
4488 
4489 	if (path)
4490 		(void) snprintf(errbuf, sizeof (errbuf),
4491 		    dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
4492 		    path);
4493 	else
4494 		(void) snprintf(errbuf, sizeof (errbuf),
4495 		    dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
4496 		    zhp->zpool_name);
4497 
4498 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4499 	if (path) {
4500 		if ((tgt = zpool_find_vdev(zhp, path, &avail_spare,
4501 		    &l2cache, NULL)) == NULL)
4502 			return (zfs_error(hdl, EZFS_NODEVICE, errbuf));
4503 
4504 		/*
4505 		 * Don't allow error clearing for hot spares.  Do allow
4506 		 * error clearing for l2cache devices.
4507 		 */
4508 		if (avail_spare)
4509 			return (zfs_error(hdl, EZFS_ISSPARE, errbuf));
4510 
4511 		zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
4512 	}
4513 
4514 	zpool_get_load_policy(rewindnvl, &policy);
4515 	zc.zc_cookie = policy.zlp_rewind;
4516 
4517 	zcmd_alloc_dst_nvlist(hdl, &zc, zhp->zpool_config_size * 2);
4518 	if (rewindnvl != NULL)
4519 		zcmd_write_src_nvlist(hdl, &zc, rewindnvl);
4520 
4521 	while ((error = zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc)) != 0 &&
4522 	    errno == ENOMEM)
4523 		zcmd_expand_dst_nvlist(hdl, &zc);
4524 
4525 	if (!error || ((policy.zlp_rewind & ZPOOL_TRY_REWIND) &&
4526 	    errno != EPERM && errno != EACCES)) {
4527 		if (policy.zlp_rewind &
4528 		    (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
4529 			(void) zcmd_read_dst_nvlist(hdl, &zc, &nvi);
4530 			zpool_rewind_exclaim(hdl, zc.zc_name,
4531 			    ((policy.zlp_rewind & ZPOOL_TRY_REWIND) != 0),
4532 			    nvi);
4533 			nvlist_free(nvi);
4534 		}
4535 		zcmd_free_nvlists(&zc);
4536 		return (0);
4537 	}
4538 
4539 	zcmd_free_nvlists(&zc);
4540 	return (zpool_standard_error(hdl, errno, errbuf));
4541 }
4542 
4543 /*
4544  * Similar to zpool_clear(), but takes a GUID (used by fmd).
4545  */
4546 int
zpool_vdev_clear(zpool_handle_t * zhp,uint64_t guid)4547 zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid)
4548 {
4549 	zfs_cmd_t zc = {"\0"};
4550 	char errbuf[ERRBUFLEN];
4551 	libzfs_handle_t *hdl = zhp->zpool_hdl;
4552 
4553 	(void) snprintf(errbuf, sizeof (errbuf),
4554 	    dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"),
4555 	    (u_longlong_t)guid);
4556 
4557 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4558 	zc.zc_guid = guid;
4559 	zc.zc_cookie = ZPOOL_NO_REWIND;
4560 
4561 	if (zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc) == 0)
4562 		return (0);
4563 
4564 	return (zpool_standard_error(hdl, errno, errbuf));
4565 }
4566 
4567 /*
4568  * Change the GUID for a pool.
4569  *
4570  * Similar to zpool_reguid(), but may take a GUID.
4571  *
4572  * If the guid argument is NULL, then no GUID is passed in the nvlist to the
4573  * ioctl().
4574  */
4575 int
zpool_set_guid(zpool_handle_t * zhp,const uint64_t * guid)4576 zpool_set_guid(zpool_handle_t *zhp, const uint64_t *guid)
4577 {
4578 	char errbuf[ERRBUFLEN];
4579 	libzfs_handle_t *hdl = zhp->zpool_hdl;
4580 	nvlist_t *nvl = NULL;
4581 	zfs_cmd_t zc = {"\0"};
4582 	int error;
4583 
4584 	if (guid != NULL) {
4585 		if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
4586 			return (no_memory(hdl));
4587 
4588 		if (nvlist_add_uint64(nvl, ZPOOL_REGUID_GUID, *guid) != 0) {
4589 			nvlist_free(nvl);
4590 			return (no_memory(hdl));
4591 		}
4592 
4593 		zcmd_write_src_nvlist(hdl, &zc, nvl);
4594 	}
4595 
4596 	(void) snprintf(errbuf, sizeof (errbuf),
4597 	    dgettext(TEXT_DOMAIN, "cannot reguid '%s'"), zhp->zpool_name);
4598 
4599 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4600 	error = zfs_ioctl(hdl, ZFS_IOC_POOL_REGUID, &zc);
4601 	if (error) {
4602 		return (zpool_standard_error(hdl, errno, errbuf));
4603 	}
4604 	if (guid != NULL) {
4605 		zcmd_free_nvlists(&zc);
4606 		nvlist_free(nvl);
4607 	}
4608 	return (0);
4609 }
4610 
4611 /*
4612  * Change the GUID for a pool.
4613  */
4614 int
zpool_reguid(zpool_handle_t * zhp)4615 zpool_reguid(zpool_handle_t *zhp)
4616 {
4617 	return (zpool_set_guid(zhp, NULL));
4618 }
4619 
4620 /*
4621  * Reopen the pool.
4622  */
4623 int
zpool_reopen_one(zpool_handle_t * zhp,void * data)4624 zpool_reopen_one(zpool_handle_t *zhp, void *data)
4625 {
4626 	libzfs_handle_t *hdl = zpool_get_handle(zhp);
4627 	const char *pool_name = zpool_get_name(zhp);
4628 	boolean_t *scrub_restart = data;
4629 	int error;
4630 
4631 	error = lzc_reopen(pool_name, *scrub_restart);
4632 	if (error) {
4633 		return (zpool_standard_error_fmt(hdl, error,
4634 		    dgettext(TEXT_DOMAIN, "cannot reopen '%s'"), pool_name));
4635 	}
4636 
4637 	return (0);
4638 }
4639 
4640 /* call into libzfs_core to execute the sync IOCTL per pool */
4641 int
zpool_sync_one(zpool_handle_t * zhp,void * data)4642 zpool_sync_one(zpool_handle_t *zhp, void *data)
4643 {
4644 	int ret;
4645 	libzfs_handle_t *hdl = zpool_get_handle(zhp);
4646 	const char *pool_name = zpool_get_name(zhp);
4647 	boolean_t *force = data;
4648 	nvlist_t *innvl = fnvlist_alloc();
4649 
4650 	fnvlist_add_boolean_value(innvl, "force", *force);
4651 	if ((ret = lzc_sync(pool_name, innvl, NULL)) != 0) {
4652 		nvlist_free(innvl);
4653 		return (zpool_standard_error_fmt(hdl, ret,
4654 		    dgettext(TEXT_DOMAIN, "sync '%s' failed"), pool_name));
4655 	}
4656 	nvlist_free(innvl);
4657 
4658 	return (0);
4659 }
4660 
4661 #define	PATH_BUF_LEN	64
4662 
4663 /*
4664  * Given a vdev, return the name to display in iostat.  If the vdev has a path,
4665  * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
4666  * We also check if this is a whole disk, in which case we strip off the
4667  * trailing 's0' slice name.
4668  *
4669  * This routine is also responsible for identifying when disks have been
4670  * reconfigured in a new location.  The kernel will have opened the device by
4671  * devid, but the path will still refer to the old location.  To catch this, we
4672  * first do a path -> devid translation (which is fast for the common case).  If
4673  * the devid matches, we're done.  If not, we do a reverse devid -> path
4674  * translation and issue the appropriate ioctl() to update the path of the vdev.
4675  * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
4676  * of these checks.
4677  */
4678 char *
zpool_vdev_name(libzfs_handle_t * hdl,zpool_handle_t * zhp,nvlist_t * nv,int name_flags)4679 zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv,
4680     int name_flags)
4681 {
4682 	const char *type, *tpath;
4683 	const char *path;
4684 	uint64_t value;
4685 	char buf[PATH_BUF_LEN];
4686 	char tmpbuf[PATH_BUF_LEN * 2];
4687 
4688 	/*
4689 	 * vdev_name will be "root"/"root-0" for the root vdev, but it is the
4690 	 * zpool name that will be displayed to the user.
4691 	 */
4692 	type = fnvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE);
4693 	if (zhp != NULL && strcmp(type, "root") == 0)
4694 		return (zfs_strdup(hdl, zpool_get_name(zhp)));
4695 
4696 	if (libzfs_envvar_is_set("ZPOOL_VDEV_NAME_PATH"))
4697 		name_flags |= VDEV_NAME_PATH;
4698 	if (libzfs_envvar_is_set("ZPOOL_VDEV_NAME_GUID"))
4699 		name_flags |= VDEV_NAME_GUID;
4700 	if (libzfs_envvar_is_set("ZPOOL_VDEV_NAME_FOLLOW_LINKS"))
4701 		name_flags |= VDEV_NAME_FOLLOW_LINKS;
4702 
4703 	if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, &value) == 0 ||
4704 	    name_flags & VDEV_NAME_GUID) {
4705 		(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value);
4706 		(void) snprintf(buf, sizeof (buf), "%llu", (u_longlong_t)value);
4707 		path = buf;
4708 	} else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &tpath) == 0) {
4709 		path = tpath;
4710 
4711 		if (name_flags & VDEV_NAME_FOLLOW_LINKS) {
4712 			char *rp = realpath(path, NULL);
4713 			if (rp) {
4714 				strlcpy(buf, rp, sizeof (buf));
4715 				path = buf;
4716 				free(rp);
4717 			}
4718 		}
4719 
4720 		/*
4721 		 * For a block device only use the name.
4722 		 */
4723 		if ((strcmp(type, VDEV_TYPE_DISK) == 0) &&
4724 		    !(name_flags & VDEV_NAME_PATH)) {
4725 			path = zfs_strip_path(path);
4726 		}
4727 
4728 		/*
4729 		 * Remove the partition from the path if this is a whole disk.
4730 		 */
4731 		if (strcmp(type, VDEV_TYPE_DRAID_SPARE) != 0 &&
4732 		    nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, &value)
4733 		    == 0 && value && !(name_flags & VDEV_NAME_PATH)) {
4734 			return (zfs_strip_partition(path));
4735 		}
4736 	} else {
4737 		path = type;
4738 
4739 		/*
4740 		 * If it's a raidz device, we need to stick in the parity level.
4741 		 */
4742 		if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) {
4743 			value = fnvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY);
4744 			(void) snprintf(buf, sizeof (buf), "%s%llu", path,
4745 			    (u_longlong_t)value);
4746 			path = buf;
4747 		}
4748 
4749 		/*
4750 		 * If it's a dRAID device, we add parity, groups, and spares.
4751 		 */
4752 		if (strcmp(path, VDEV_TYPE_DRAID) == 0) {
4753 			uint64_t ndata, nparity, nspares, children;
4754 			nvlist_t **child;
4755 			uint_t width;
4756 
4757 			verify(nvlist_lookup_nvlist_array(nv,
4758 			    ZPOOL_CONFIG_CHILDREN, &child, &width) == 0);
4759 			nparity = fnvlist_lookup_uint64(nv,
4760 			    ZPOOL_CONFIG_NPARITY);
4761 			ndata = fnvlist_lookup_uint64(nv,
4762 			    ZPOOL_CONFIG_DRAID_NDATA);
4763 			nspares = fnvlist_lookup_uint64(nv,
4764 			    ZPOOL_CONFIG_DRAID_NSPARES);
4765 
4766 			if (nvlist_lookup_uint64(nv,
4767 			    ZPOOL_CONFIG_DRAID_NCHILDREN, &children) != 0)
4768 				children = width;
4769 
4770 			path = zpool_draid_name(buf, sizeof (buf), ndata,
4771 			    nparity, nspares, children, width);
4772 		}
4773 
4774 		/*
4775 		 * We identify each top-level vdev by using a <type-id>
4776 		 * naming convention.
4777 		 */
4778 		if (name_flags & VDEV_NAME_TYPE_ID) {
4779 			uint64_t id = fnvlist_lookup_uint64(nv,
4780 			    ZPOOL_CONFIG_ID);
4781 			(void) snprintf(tmpbuf, sizeof (tmpbuf), "%s-%llu",
4782 			    path, (u_longlong_t)id);
4783 			path = tmpbuf;
4784 		}
4785 	}
4786 
4787 	return (zfs_strdup(hdl, path));
4788 }
4789 
4790 static int
zbookmark_mem_compare(const void * a,const void * b)4791 zbookmark_mem_compare(const void *a, const void *b)
4792 {
4793 	return (memcmp(a, b, sizeof (zbookmark_phys_t)));
4794 }
4795 
4796 void
zpool_add_propname(zpool_handle_t * zhp,const char * propname)4797 zpool_add_propname(zpool_handle_t *zhp, const char *propname)
4798 {
4799 	assert(zhp->zpool_n_propnames < ZHP_MAX_PROPNAMES);
4800 	zhp->zpool_propnames[zhp->zpool_n_propnames] = propname;
4801 	zhp->zpool_n_propnames++;
4802 }
4803 
4804 /*
4805  * Retrieve the persistent error log, uniquify the members, and return to the
4806  * caller.
4807  */
4808 int
zpool_get_errlog(zpool_handle_t * zhp,nvlist_t ** nverrlistp)4809 zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp)
4810 {
4811 	zfs_cmd_t zc = {"\0"};
4812 	libzfs_handle_t *hdl = zhp->zpool_hdl;
4813 	zbookmark_phys_t *buf;
4814 	uint64_t buflen = 10000; /* approx. 1MB of RAM */
4815 
4816 	if (fnvlist_lookup_uint64(zhp->zpool_config,
4817 	    ZPOOL_CONFIG_ERRCOUNT) == 0)
4818 		return (0);
4819 
4820 	/*
4821 	 * Retrieve the raw error list from the kernel.  If it doesn't fit,
4822 	 * allocate a larger buffer and retry.
4823 	 */
4824 	(void) strcpy(zc.zc_name, zhp->zpool_name);
4825 	for (;;) {
4826 		buf = zfs_alloc(zhp->zpool_hdl,
4827 		    buflen * sizeof (zbookmark_phys_t));
4828 		zc.zc_nvlist_dst = (uintptr_t)buf;
4829 		zc.zc_nvlist_dst_size = buflen;
4830 		if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_ERROR_LOG,
4831 		    &zc) != 0) {
4832 			free(buf);
4833 			if (errno == ENOMEM) {
4834 				buflen *= 2;
4835 			} else {
4836 				return (zpool_standard_error_fmt(hdl, errno,
4837 				    dgettext(TEXT_DOMAIN, "errors: List of "
4838 				    "errors unavailable")));
4839 			}
4840 		} else {
4841 			break;
4842 		}
4843 	}
4844 
4845 	/*
4846 	 * Sort the resulting bookmarks.  This is a little confusing due to the
4847 	 * implementation of ZFS_IOC_ERROR_LOG.  The bookmarks are copied last
4848 	 * to first, and 'zc_nvlist_dst_size' indicates the number of bookmarks
4849 	 * _not_ copied as part of the process.  So we point the start of our
4850 	 * array appropriate and decrement the total number of elements.
4851 	 */
4852 	zbookmark_phys_t *zb = buf + zc.zc_nvlist_dst_size;
4853 	uint64_t zblen = buflen - zc.zc_nvlist_dst_size;
4854 
4855 	qsort(zb, zblen, sizeof (zbookmark_phys_t), zbookmark_mem_compare);
4856 
4857 	verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0);
4858 
4859 	/*
4860 	 * Fill in the nverrlistp with nvlist's of dataset and object numbers.
4861 	 */
4862 	for (uint64_t i = 0; i < zblen; i++) {
4863 		nvlist_t *nv;
4864 
4865 		/* ignoring zb_blkid and zb_level for now */
4866 		if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset &&
4867 		    zb[i-1].zb_object == zb[i].zb_object)
4868 			continue;
4869 
4870 		if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0)
4871 			goto nomem;
4872 		if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET,
4873 		    zb[i].zb_objset) != 0) {
4874 			nvlist_free(nv);
4875 			goto nomem;
4876 		}
4877 		if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT,
4878 		    zb[i].zb_object) != 0) {
4879 			nvlist_free(nv);
4880 			goto nomem;
4881 		}
4882 		if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) {
4883 			nvlist_free(nv);
4884 			goto nomem;
4885 		}
4886 		nvlist_free(nv);
4887 	}
4888 
4889 	free(buf);
4890 	return (0);
4891 
4892 nomem:
4893 	free(buf);
4894 	return (no_memory(zhp->zpool_hdl));
4895 }
4896 
4897 /*
4898  * Upgrade a ZFS pool to the latest on-disk version.
4899  */
4900 int
zpool_upgrade(zpool_handle_t * zhp,uint64_t new_version)4901 zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version)
4902 {
4903 	zfs_cmd_t zc = {"\0"};
4904 	libzfs_handle_t *hdl = zhp->zpool_hdl;
4905 
4906 	(void) strcpy(zc.zc_name, zhp->zpool_name);
4907 	zc.zc_cookie = new_version;
4908 
4909 	if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0)
4910 		return (zpool_standard_error_fmt(hdl, errno,
4911 		    dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"),
4912 		    zhp->zpool_name));
4913 	return (0);
4914 }
4915 
4916 void
zfs_save_arguments(int argc,char ** argv,char * string,int len)4917 zfs_save_arguments(int argc, char **argv, char *string, int len)
4918 {
4919 	int i;
4920 
4921 	(void) strlcpy(string, zfs_basename(argv[0]), len);
4922 	for (i = 1; i < argc; i++) {
4923 		(void) strlcat(string, " ", len);
4924 		(void) strlcat(string, argv[i], len);
4925 	}
4926 }
4927 
4928 int
zpool_log_history(libzfs_handle_t * hdl,const char * message)4929 zpool_log_history(libzfs_handle_t *hdl, const char *message)
4930 {
4931 	zfs_cmd_t zc = {"\0"};
4932 	nvlist_t *args;
4933 
4934 	args = fnvlist_alloc();
4935 	fnvlist_add_string(args, "message", message);
4936 	zcmd_write_src_nvlist(hdl, &zc, args);
4937 	int err = zfs_ioctl(hdl, ZFS_IOC_LOG_HISTORY, &zc);
4938 	nvlist_free(args);
4939 	zcmd_free_nvlists(&zc);
4940 	return (err);
4941 }
4942 
4943 /*
4944  * Perform ioctl to get some command history of a pool.
4945  *
4946  * 'buf' is the buffer to fill up to 'len' bytes.  'off' is the
4947  * logical offset of the history buffer to start reading from.
4948  *
4949  * Upon return, 'off' is the next logical offset to read from and
4950  * 'len' is the actual amount of bytes read into 'buf'.
4951  */
4952 static int
get_history(zpool_handle_t * zhp,char * buf,uint64_t * off,uint64_t * len)4953 get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len)
4954 {
4955 	zfs_cmd_t zc = {"\0"};
4956 	libzfs_handle_t *hdl = zhp->zpool_hdl;
4957 
4958 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4959 
4960 	zc.zc_history = (uint64_t)(uintptr_t)buf;
4961 	zc.zc_history_len = *len;
4962 	zc.zc_history_offset = *off;
4963 
4964 	if (zfs_ioctl(hdl, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) {
4965 		switch (errno) {
4966 		case EPERM:
4967 			return (zfs_error_fmt(hdl, EZFS_PERM,
4968 			    dgettext(TEXT_DOMAIN,
4969 			    "cannot show history for pool '%s'"),
4970 			    zhp->zpool_name));
4971 		case ENOENT:
4972 			return (zfs_error_fmt(hdl, EZFS_NOHISTORY,
4973 			    dgettext(TEXT_DOMAIN, "cannot get history for pool "
4974 			    "'%s'"), zhp->zpool_name));
4975 		case ENOTSUP:
4976 			return (zfs_error_fmt(hdl, EZFS_BADVERSION,
4977 			    dgettext(TEXT_DOMAIN, "cannot get history for pool "
4978 			    "'%s', pool must be upgraded"), zhp->zpool_name));
4979 		default:
4980 			return (zpool_standard_error_fmt(hdl, errno,
4981 			    dgettext(TEXT_DOMAIN,
4982 			    "cannot get history for '%s'"), zhp->zpool_name));
4983 		}
4984 	}
4985 
4986 	*len = zc.zc_history_len;
4987 	*off = zc.zc_history_offset;
4988 
4989 	return (0);
4990 }
4991 
4992 /*
4993  * Retrieve the command history of a pool.
4994  */
4995 int
zpool_get_history(zpool_handle_t * zhp,nvlist_t ** nvhisp,uint64_t * off,boolean_t * eof)4996 zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp, uint64_t *off,
4997     boolean_t *eof)
4998 {
4999 	libzfs_handle_t *hdl = zhp->zpool_hdl;
5000 	char *buf;
5001 	int buflen = 128 * 1024;
5002 	nvlist_t **records = NULL;
5003 	uint_t numrecords = 0;
5004 	int err = 0, i;
5005 	uint64_t start = *off;
5006 
5007 	buf = zfs_alloc(hdl, buflen);
5008 
5009 	/* process about 1MiB a time */
5010 	while (*off - start < 1024 * 1024) {
5011 		uint64_t bytes_read = buflen;
5012 		uint64_t leftover;
5013 
5014 		if ((err = get_history(zhp, buf, off, &bytes_read)) != 0)
5015 			break;
5016 
5017 		/* if nothing else was read in, we're at EOF, just return */
5018 		if (!bytes_read) {
5019 			*eof = B_TRUE;
5020 			break;
5021 		}
5022 
5023 		if ((err = zpool_history_unpack(buf, bytes_read,
5024 		    &leftover, &records, &numrecords)) != 0) {
5025 			zpool_standard_error_fmt(hdl, err,
5026 			    dgettext(TEXT_DOMAIN,
5027 			    "cannot get history for '%s'"), zhp->zpool_name);
5028 			break;
5029 		}
5030 		*off -= leftover;
5031 		if (leftover == bytes_read) {
5032 			/*
5033 			 * no progress made, because buffer is not big enough
5034 			 * to hold this record; resize and retry.
5035 			 */
5036 			buflen *= 2;
5037 			free(buf);
5038 			buf = zfs_alloc(hdl, buflen);
5039 		}
5040 	}
5041 
5042 	free(buf);
5043 
5044 	if (!err) {
5045 		*nvhisp = fnvlist_alloc();
5046 		fnvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD,
5047 		    (const nvlist_t **)records, numrecords);
5048 	}
5049 	for (i = 0; i < numrecords; i++)
5050 		nvlist_free(records[i]);
5051 	free(records);
5052 
5053 	return (err);
5054 }
5055 
5056 /*
5057  * Retrieve the next event given the passed 'zevent_fd' file descriptor.
5058  * If there is a new event available 'nvp' will contain a newly allocated
5059  * nvlist and 'dropped' will be set to the number of missed events since
5060  * the last call to this function.  When 'nvp' is set to NULL it indicates
5061  * no new events are available.  In either case the function returns 0 and
5062  * it is up to the caller to free 'nvp'.  In the case of a fatal error the
5063  * function will return a non-zero value.  When the function is called in
5064  * blocking mode (the default, unless the ZEVENT_NONBLOCK flag is passed),
5065  * it will not return until a new event is available.
5066  */
5067 int
zpool_events_next(libzfs_handle_t * hdl,nvlist_t ** nvp,int * dropped,unsigned flags,int zevent_fd)5068 zpool_events_next(libzfs_handle_t *hdl, nvlist_t **nvp,
5069     int *dropped, unsigned flags, int zevent_fd)
5070 {
5071 	zfs_cmd_t zc = {"\0"};
5072 	int error = 0;
5073 
5074 	*nvp = NULL;
5075 	*dropped = 0;
5076 	zc.zc_cleanup_fd = zevent_fd;
5077 
5078 	if (flags & ZEVENT_NONBLOCK)
5079 		zc.zc_guid = ZEVENT_NONBLOCK;
5080 
5081 	zcmd_alloc_dst_nvlist(hdl, &zc, ZEVENT_SIZE);
5082 
5083 retry:
5084 	if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_NEXT, &zc) != 0) {
5085 		switch (errno) {
5086 		case ESHUTDOWN:
5087 			error = zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
5088 			    dgettext(TEXT_DOMAIN, "zfs shutdown"));
5089 			goto out;
5090 		case ENOENT:
5091 			/* Blocking error case should not occur */
5092 			if (!(flags & ZEVENT_NONBLOCK))
5093 				error = zpool_standard_error_fmt(hdl, errno,
5094 				    dgettext(TEXT_DOMAIN, "cannot get event"));
5095 
5096 			goto out;
5097 		case ENOMEM:
5098 			zcmd_expand_dst_nvlist(hdl, &zc);
5099 			goto retry;
5100 		default:
5101 			error = zpool_standard_error_fmt(hdl, errno,
5102 			    dgettext(TEXT_DOMAIN, "cannot get event"));
5103 			goto out;
5104 		}
5105 	}
5106 
5107 	error = zcmd_read_dst_nvlist(hdl, &zc, nvp);
5108 	if (error != 0)
5109 		goto out;
5110 
5111 	*dropped = (int)zc.zc_cookie;
5112 out:
5113 	zcmd_free_nvlists(&zc);
5114 
5115 	return (error);
5116 }
5117 
5118 /*
5119  * Clear all events.
5120  */
5121 int
zpool_events_clear(libzfs_handle_t * hdl,int * count)5122 zpool_events_clear(libzfs_handle_t *hdl, int *count)
5123 {
5124 	zfs_cmd_t zc = {"\0"};
5125 
5126 	if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_CLEAR, &zc) != 0)
5127 		return (zpool_standard_error(hdl, errno,
5128 		    dgettext(TEXT_DOMAIN, "cannot clear events")));
5129 
5130 	if (count != NULL)
5131 		*count = (int)zc.zc_cookie; /* # of events cleared */
5132 
5133 	return (0);
5134 }
5135 
5136 /*
5137  * Seek to a specific EID, ZEVENT_SEEK_START, or ZEVENT_SEEK_END for
5138  * the passed zevent_fd file handle.  On success zero is returned,
5139  * otherwise -1 is returned and hdl->libzfs_error is set to the errno.
5140  */
5141 int
zpool_events_seek(libzfs_handle_t * hdl,uint64_t eid,int zevent_fd)5142 zpool_events_seek(libzfs_handle_t *hdl, uint64_t eid, int zevent_fd)
5143 {
5144 	zfs_cmd_t zc = {"\0"};
5145 	int error = 0;
5146 
5147 	zc.zc_guid = eid;
5148 	zc.zc_cleanup_fd = zevent_fd;
5149 
5150 	if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_SEEK, &zc) != 0) {
5151 		switch (errno) {
5152 		case ENOENT:
5153 			error = zfs_error_fmt(hdl, EZFS_NOENT,
5154 			    dgettext(TEXT_DOMAIN, "cannot get event"));
5155 			break;
5156 
5157 		case ENOMEM:
5158 			error = zfs_error_fmt(hdl, EZFS_NOMEM,
5159 			    dgettext(TEXT_DOMAIN, "cannot get event"));
5160 			break;
5161 
5162 		default:
5163 			error = zpool_standard_error_fmt(hdl, errno,
5164 			    dgettext(TEXT_DOMAIN, "cannot get event"));
5165 			break;
5166 		}
5167 	}
5168 
5169 	return (error);
5170 }
5171 
5172 static void
zpool_obj_to_path_impl(zpool_handle_t * zhp,uint64_t dsobj,uint64_t obj,char * pathname,size_t len,boolean_t always_unmounted)5173 zpool_obj_to_path_impl(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
5174     char *pathname, size_t len, boolean_t always_unmounted)
5175 {
5176 	zfs_cmd_t zc = {"\0"};
5177 	boolean_t mounted = B_FALSE;
5178 	char *mntpnt = NULL;
5179 	char dsname[ZFS_MAX_DATASET_NAME_LEN];
5180 
5181 	if (dsobj == 0) {
5182 		/* special case for the MOS */
5183 		(void) snprintf(pathname, len, "<metadata>:<0x%llx>",
5184 		    (longlong_t)obj);
5185 		return;
5186 	}
5187 
5188 	/* get the dataset's name */
5189 	(void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
5190 	zc.zc_obj = dsobj;
5191 	if (zfs_ioctl(zhp->zpool_hdl,
5192 	    ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) {
5193 		/* just write out a path of two object numbers */
5194 		(void) snprintf(pathname, len, "<0x%llx>:<0x%llx>",
5195 		    (longlong_t)dsobj, (longlong_t)obj);
5196 		return;
5197 	}
5198 	(void) strlcpy(dsname, zc.zc_value, sizeof (dsname));
5199 
5200 	/* find out if the dataset is mounted */
5201 	mounted = !always_unmounted && is_mounted(zhp->zpool_hdl, dsname,
5202 	    &mntpnt);
5203 
5204 	/* get the corrupted object's path */
5205 	(void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name));
5206 	zc.zc_obj = obj;
5207 	if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_OBJ_TO_PATH,
5208 	    &zc) == 0) {
5209 		if (mounted) {
5210 			(void) snprintf(pathname, len, "%s%s", mntpnt,
5211 			    zc.zc_value);
5212 		} else {
5213 			(void) snprintf(pathname, len, "%s:%s",
5214 			    dsname, zc.zc_value);
5215 		}
5216 	} else {
5217 		(void) snprintf(pathname, len, "%s:<0x%llx>", dsname,
5218 		    (longlong_t)obj);
5219 	}
5220 	free(mntpnt);
5221 }
5222 
5223 void
zpool_obj_to_path(zpool_handle_t * zhp,uint64_t dsobj,uint64_t obj,char * pathname,size_t len)5224 zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
5225     char *pathname, size_t len)
5226 {
5227 	zpool_obj_to_path_impl(zhp, dsobj, obj, pathname, len, B_FALSE);
5228 }
5229 
5230 void
zpool_obj_to_path_ds(zpool_handle_t * zhp,uint64_t dsobj,uint64_t obj,char * pathname,size_t len)5231 zpool_obj_to_path_ds(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
5232     char *pathname, size_t len)
5233 {
5234 	zpool_obj_to_path_impl(zhp, dsobj, obj, pathname, len, B_TRUE);
5235 }
5236 /*
5237  * Wait while the specified activity is in progress in the pool.
5238  */
5239 int
zpool_wait(zpool_handle_t * zhp,zpool_wait_activity_t activity)5240 zpool_wait(zpool_handle_t *zhp, zpool_wait_activity_t activity)
5241 {
5242 	boolean_t missing;
5243 
5244 	int error = zpool_wait_status(zhp, activity, &missing, NULL);
5245 
5246 	if (missing) {
5247 		(void) zpool_standard_error_fmt(zhp->zpool_hdl, ENOENT,
5248 		    dgettext(TEXT_DOMAIN, "error waiting in pool '%s'"),
5249 		    zhp->zpool_name);
5250 		return (ENOENT);
5251 	} else {
5252 		return (error);
5253 	}
5254 }
5255 
5256 /*
5257  * Wait for the given activity and return the status of the wait (whether or not
5258  * any waiting was done) in the 'waited' parameter. Non-existent pools are
5259  * reported via the 'missing' parameter, rather than by printing an error
5260  * message. This is convenient when this function is called in a loop over a
5261  * long period of time (as it is, for example, by zpool's wait cmd). In that
5262  * scenario, a pool being exported or destroyed should be considered a normal
5263  * event, so we don't want to print an error when we find that the pool doesn't
5264  * exist.
5265  */
5266 int
zpool_wait_status(zpool_handle_t * zhp,zpool_wait_activity_t activity,boolean_t * missing,boolean_t * waited)5267 zpool_wait_status(zpool_handle_t *zhp, zpool_wait_activity_t activity,
5268     boolean_t *missing, boolean_t *waited)
5269 {
5270 	int error = lzc_wait(zhp->zpool_name, activity, waited);
5271 	*missing = (error == ENOENT);
5272 	if (*missing)
5273 		return (0);
5274 
5275 	if (error != 0) {
5276 		(void) zpool_standard_error_fmt(zhp->zpool_hdl, error,
5277 		    dgettext(TEXT_DOMAIN, "error waiting in pool '%s'"),
5278 		    zhp->zpool_name);
5279 	}
5280 
5281 	return (error);
5282 }
5283 
5284 int
zpool_set_bootenv(zpool_handle_t * zhp,const nvlist_t * envmap)5285 zpool_set_bootenv(zpool_handle_t *zhp, const nvlist_t *envmap)
5286 {
5287 	int error = lzc_set_bootenv(zhp->zpool_name, envmap);
5288 	if (error != 0) {
5289 		(void) zpool_standard_error_fmt(zhp->zpool_hdl, error,
5290 		    dgettext(TEXT_DOMAIN,
5291 		    "error setting bootenv in pool '%s'"), zhp->zpool_name);
5292 	}
5293 
5294 	return (error);
5295 }
5296 
5297 int
zpool_get_bootenv(zpool_handle_t * zhp,nvlist_t ** nvlp)5298 zpool_get_bootenv(zpool_handle_t *zhp, nvlist_t **nvlp)
5299 {
5300 	nvlist_t *nvl;
5301 	int error;
5302 
5303 	nvl = NULL;
5304 	error = lzc_get_bootenv(zhp->zpool_name, &nvl);
5305 	if (error != 0) {
5306 		(void) zpool_standard_error_fmt(zhp->zpool_hdl, error,
5307 		    dgettext(TEXT_DOMAIN,
5308 		    "error getting bootenv in pool '%s'"), zhp->zpool_name);
5309 	} else {
5310 		*nvlp = nvl;
5311 	}
5312 
5313 	return (error);
5314 }
5315 
5316 /*
5317  * Attempt to read and parse feature file(s) (from "compatibility" property).
5318  * Files contain zpool feature names, comma or whitespace-separated.
5319  * Comments (# character to next newline) are discarded.
5320  *
5321  * Arguments:
5322  *  compatibility : string containing feature filenames
5323  *  features : either NULL or pointer to array of boolean
5324  *  report : either NULL or pointer to string buffer
5325  *  rlen : length of "report" buffer
5326  *
5327  * compatibility is NULL (unset), "", "off", "legacy", or list of
5328  * comma-separated filenames. filenames should either be absolute,
5329  * or relative to:
5330  *   1) ZPOOL_SYSCONF_COMPAT_D (eg: /etc/zfs/compatibility.d) or
5331  *   2) ZPOOL_DATA_COMPAT_D (eg: /usr/share/zfs/compatibility.d).
5332  * (Unset), "" or "off" => enable all features
5333  * "legacy" => disable all features
5334  *
5335  * Any feature names read from files which match unames in spa_feature_table
5336  * will have the corresponding boolean set in the features array (if non-NULL).
5337  * If more than one feature set specified, only features present in *all* of
5338  * them will be set.
5339  *
5340  * "report" if not NULL will be populated with a suitable status message.
5341  *
5342  * Return values:
5343  *   ZPOOL_COMPATIBILITY_OK : files read and parsed ok
5344  *   ZPOOL_COMPATIBILITY_BADFILE : file too big or not a text file
5345  *   ZPOOL_COMPATIBILITY_BADTOKEN : SYSCONF file contains invalid feature name
5346  *   ZPOOL_COMPATIBILITY_WARNTOKEN : DATA file contains invalid feature name
5347  *   ZPOOL_COMPATIBILITY_NOFILES : no feature files found
5348  */
5349 zpool_compat_status_t
zpool_load_compat(const char * compat,boolean_t * features,char * report,size_t rlen)5350 zpool_load_compat(const char *compat, boolean_t *features, char *report,
5351     size_t rlen)
5352 {
5353 	int sdirfd, ddirfd, featfd;
5354 	struct stat fs;
5355 	char *fc;
5356 	char *ps, *ls, *ws;
5357 	char *file, *line, *word;
5358 
5359 	char l_compat[ZFS_MAXPROPLEN];
5360 
5361 	boolean_t ret_nofiles = B_TRUE;
5362 	boolean_t ret_badfile = B_FALSE;
5363 	boolean_t ret_badtoken = B_FALSE;
5364 	boolean_t ret_warntoken = B_FALSE;
5365 
5366 	/* special cases (unset), "" and "off" => enable all features */
5367 	if (compat == NULL || compat[0] == '\0' ||
5368 	    strcmp(compat, ZPOOL_COMPAT_OFF) == 0) {
5369 		if (features != NULL) {
5370 			for (uint_t i = 0; i < SPA_FEATURES; i++)
5371 				features[i] = B_TRUE;
5372 		}
5373 		if (report != NULL)
5374 			strlcpy(report, gettext("all features enabled"), rlen);
5375 		return (ZPOOL_COMPATIBILITY_OK);
5376 	}
5377 
5378 	/* Final special case "legacy" => disable all features */
5379 	if (strcmp(compat, ZPOOL_COMPAT_LEGACY) == 0) {
5380 		if (features != NULL)
5381 			for (uint_t i = 0; i < SPA_FEATURES; i++)
5382 				features[i] = B_FALSE;
5383 		if (report != NULL)
5384 			strlcpy(report, gettext("all features disabled"), rlen);
5385 		return (ZPOOL_COMPATIBILITY_OK);
5386 	}
5387 
5388 	/*
5389 	 * Start with all true; will be ANDed with results from each file
5390 	 */
5391 	if (features != NULL)
5392 		for (uint_t i = 0; i < SPA_FEATURES; i++)
5393 			features[i] = B_TRUE;
5394 
5395 	char err_badfile[ZFS_MAXPROPLEN] = "";
5396 	char err_badtoken[ZFS_MAXPROPLEN] = "";
5397 
5398 	/*
5399 	 * We ignore errors from the directory open()
5400 	 * as they're only needed if the filename is relative
5401 	 * which will be checked during the openat().
5402 	 */
5403 
5404 /* O_PATH safer than O_RDONLY if system allows it */
5405 #if defined(O_PATH)
5406 #define	ZC_DIR_FLAGS (O_DIRECTORY | O_CLOEXEC | O_PATH)
5407 #else
5408 #define	ZC_DIR_FLAGS (O_DIRECTORY | O_CLOEXEC | O_RDONLY)
5409 #endif
5410 
5411 	sdirfd = open(ZPOOL_SYSCONF_COMPAT_D, ZC_DIR_FLAGS);
5412 	ddirfd = open(ZPOOL_DATA_COMPAT_D, ZC_DIR_FLAGS);
5413 
5414 	(void) strlcpy(l_compat, compat, ZFS_MAXPROPLEN);
5415 
5416 	for (file = strtok_r(l_compat, ",", &ps);
5417 	    file != NULL;
5418 	    file = strtok_r(NULL, ",", &ps)) {
5419 
5420 		boolean_t l_features[SPA_FEATURES];
5421 
5422 		enum { Z_SYSCONF, Z_DATA } source;
5423 
5424 		/* try sysconfdir first, then datadir */
5425 		source = Z_SYSCONF;
5426 		if ((featfd = openat(sdirfd, file, O_RDONLY | O_CLOEXEC)) < 0) {
5427 			featfd = openat(ddirfd, file, O_RDONLY | O_CLOEXEC);
5428 			source = Z_DATA;
5429 		}
5430 
5431 		/* File readable and correct size? */
5432 		if (featfd < 0 ||
5433 		    fstat(featfd, &fs) < 0 ||
5434 		    fs.st_size < 1 ||
5435 		    fs.st_size > ZPOOL_COMPAT_MAXSIZE) {
5436 			(void) close(featfd);
5437 			strlcat(err_badfile, file, ZFS_MAXPROPLEN);
5438 			strlcat(err_badfile, " ", ZFS_MAXPROPLEN);
5439 			ret_badfile = B_TRUE;
5440 			continue;
5441 		}
5442 
5443 /* Prefault the file if system allows */
5444 #if defined(MAP_POPULATE)
5445 #define	ZC_MMAP_FLAGS (MAP_PRIVATE | MAP_POPULATE)
5446 #elif defined(MAP_PREFAULT_READ)
5447 #define	ZC_MMAP_FLAGS (MAP_PRIVATE | MAP_PREFAULT_READ)
5448 #else
5449 #define	ZC_MMAP_FLAGS (MAP_PRIVATE)
5450 #endif
5451 
5452 		/* private mmap() so we can strtok safely */
5453 		fc = (char *)mmap(NULL, fs.st_size, PROT_READ | PROT_WRITE,
5454 		    ZC_MMAP_FLAGS, featfd, 0);
5455 		(void) close(featfd);
5456 
5457 		/* map ok, and last character == newline? */
5458 		if (fc == MAP_FAILED || fc[fs.st_size - 1] != '\n') {
5459 			(void) munmap((void *) fc, fs.st_size);
5460 			strlcat(err_badfile, file, ZFS_MAXPROPLEN);
5461 			strlcat(err_badfile, " ", ZFS_MAXPROPLEN);
5462 			ret_badfile = B_TRUE;
5463 			continue;
5464 		}
5465 
5466 		ret_nofiles = B_FALSE;
5467 
5468 		for (uint_t i = 0; i < SPA_FEATURES; i++)
5469 			l_features[i] = B_FALSE;
5470 
5471 		/* replace final newline with NULL to ensure string ends */
5472 		fc[fs.st_size - 1] = '\0';
5473 
5474 		for (line = strtok_r(fc, "\n", &ls);
5475 		    line != NULL;
5476 		    line = strtok_r(NULL, "\n", &ls)) {
5477 			/* discard comments */
5478 			char *r = strchr(line, '#');
5479 			if (r != NULL)
5480 				*r = '\0';
5481 
5482 			for (word = strtok_r(line, ", \t", &ws);
5483 			    word != NULL;
5484 			    word = strtok_r(NULL, ", \t", &ws)) {
5485 				/* Find matching feature name */
5486 				uint_t f;
5487 				for (f = 0; f < SPA_FEATURES; f++) {
5488 					zfeature_info_t *fi =
5489 					    &spa_feature_table[f];
5490 					if (strcmp(word, fi->fi_uname) == 0) {
5491 						l_features[f] = B_TRUE;
5492 						break;
5493 					}
5494 				}
5495 				if (f < SPA_FEATURES)
5496 					continue;
5497 
5498 				/* found an unrecognized word */
5499 				/* lightly sanitize it */
5500 				if (strlen(word) > 32)
5501 					word[32] = '\0';
5502 				for (char *c = word; *c != '\0'; c++)
5503 					if (!isprint(*c))
5504 						*c = '?';
5505 
5506 				strlcat(err_badtoken, word, ZFS_MAXPROPLEN);
5507 				strlcat(err_badtoken, " ", ZFS_MAXPROPLEN);
5508 				if (source == Z_SYSCONF)
5509 					ret_badtoken = B_TRUE;
5510 				else
5511 					ret_warntoken = B_TRUE;
5512 			}
5513 		}
5514 		(void) munmap((void *) fc, fs.st_size);
5515 
5516 		if (features != NULL)
5517 			for (uint_t i = 0; i < SPA_FEATURES; i++)
5518 				features[i] &= l_features[i];
5519 	}
5520 	(void) close(sdirfd);
5521 	(void) close(ddirfd);
5522 
5523 	/* Return the most serious error */
5524 	if (ret_badfile) {
5525 		if (report != NULL)
5526 			snprintf(report, rlen, gettext("could not read/"
5527 			    "parse feature file(s): %s"), err_badfile);
5528 		return (ZPOOL_COMPATIBILITY_BADFILE);
5529 	}
5530 	if (ret_nofiles) {
5531 		if (report != NULL)
5532 			strlcpy(report,
5533 			    gettext("no valid compatibility files specified"),
5534 			    rlen);
5535 		return (ZPOOL_COMPATIBILITY_NOFILES);
5536 	}
5537 	if (ret_badtoken) {
5538 		if (report != NULL)
5539 			snprintf(report, rlen, gettext("invalid feature "
5540 			    "name(s) in local compatibility files: %s"),
5541 			    err_badtoken);
5542 		return (ZPOOL_COMPATIBILITY_BADTOKEN);
5543 	}
5544 	if (ret_warntoken) {
5545 		if (report != NULL)
5546 			snprintf(report, rlen, gettext("unrecognized feature "
5547 			    "name(s) in distribution compatibility files: %s"),
5548 			    err_badtoken);
5549 		return (ZPOOL_COMPATIBILITY_WARNTOKEN);
5550 	}
5551 	if (report != NULL)
5552 		strlcpy(report, gettext("compatibility set ok"), rlen);
5553 	return (ZPOOL_COMPATIBILITY_OK);
5554 }
5555 
5556 static int
zpool_vdev_guid(zpool_handle_t * zhp,const char * vdevname,uint64_t * vdev_guid)5557 zpool_vdev_guid(zpool_handle_t *zhp, const char *vdevname, uint64_t *vdev_guid)
5558 {
5559 	nvlist_t *tgt;
5560 	boolean_t avail_spare, l2cache;
5561 
5562 	verify(zhp != NULL);
5563 	if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
5564 		char errbuf[ERRBUFLEN];
5565 		(void) snprintf(errbuf, sizeof (errbuf),
5566 		    dgettext(TEXT_DOMAIN, "pool is in an unavailable state"));
5567 		return (zfs_error(zhp->zpool_hdl, EZFS_POOLUNAVAIL, errbuf));
5568 	}
5569 
5570 	if ((tgt = zpool_find_vdev(zhp, vdevname, &avail_spare, &l2cache,
5571 	    NULL)) == NULL) {
5572 		char errbuf[ERRBUFLEN];
5573 		(void) snprintf(errbuf, sizeof (errbuf),
5574 		    dgettext(TEXT_DOMAIN, "can not find %s in %s"),
5575 		    vdevname, zhp->zpool_name);
5576 		return (zfs_error(zhp->zpool_hdl, EZFS_NODEVICE, errbuf));
5577 	}
5578 
5579 	*vdev_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
5580 	return (0);
5581 }
5582 
5583 /*
5584  * Get a vdev property value for 'prop' and return the value in
5585  * a pre-allocated buffer.
5586  */
5587 int
zpool_get_vdev_prop_value(nvlist_t * nvprop,vdev_prop_t prop,char * prop_name,char * buf,size_t len,zprop_source_t * srctype,boolean_t literal)5588 zpool_get_vdev_prop_value(nvlist_t *nvprop, vdev_prop_t prop, char *prop_name,
5589     char *buf, size_t len, zprop_source_t *srctype, boolean_t literal)
5590 {
5591 	nvlist_t *nv;
5592 	const char *strval;
5593 	uint64_t intval;
5594 	zprop_source_t src = ZPROP_SRC_NONE;
5595 
5596 	if (prop == VDEV_PROP_USERPROP) {
5597 		/* user property, prop_name must contain the property name */
5598 		assert(prop_name != NULL);
5599 		if (nvlist_lookup_nvlist(nvprop, prop_name, &nv) == 0) {
5600 			src = fnvlist_lookup_uint64(nv, ZPROP_SOURCE);
5601 			strval = fnvlist_lookup_string(nv, ZPROP_VALUE);
5602 		} else {
5603 			/* user prop not found */
5604 			src = ZPROP_SRC_DEFAULT;
5605 			strval = "-";
5606 		}
5607 		(void) strlcpy(buf, strval, len);
5608 		if (srctype)
5609 			*srctype = src;
5610 		return (0);
5611 	}
5612 
5613 	if (prop_name == NULL)
5614 		prop_name = (char *)vdev_prop_to_name(prop);
5615 
5616 	switch (vdev_prop_get_type(prop)) {
5617 	case PROP_TYPE_STRING:
5618 		if (nvlist_lookup_nvlist(nvprop, prop_name, &nv) == 0) {
5619 			src = fnvlist_lookup_uint64(nv, ZPROP_SOURCE);
5620 			strval = fnvlist_lookup_string(nv, ZPROP_VALUE);
5621 		} else {
5622 			src = ZPROP_SRC_DEFAULT;
5623 			if ((strval = vdev_prop_default_string(prop)) == NULL)
5624 				strval = "-";
5625 		}
5626 		(void) strlcpy(buf, strval, len);
5627 		break;
5628 
5629 	case PROP_TYPE_NUMBER:
5630 		if (nvlist_lookup_nvlist(nvprop, prop_name, &nv) == 0) {
5631 			src = fnvlist_lookup_uint64(nv, ZPROP_SOURCE);
5632 			intval = fnvlist_lookup_uint64(nv, ZPROP_VALUE);
5633 		} else {
5634 			src = ZPROP_SRC_DEFAULT;
5635 			intval = vdev_prop_default_numeric(prop);
5636 		}
5637 
5638 		switch (prop) {
5639 		case VDEV_PROP_ASIZE:
5640 		case VDEV_PROP_PSIZE:
5641 		case VDEV_PROP_SIZE:
5642 		case VDEV_PROP_BOOTSIZE:
5643 		case VDEV_PROP_ALLOCATED:
5644 		case VDEV_PROP_FREE:
5645 		case VDEV_PROP_READ_ERRORS:
5646 		case VDEV_PROP_WRITE_ERRORS:
5647 		case VDEV_PROP_CHECKSUM_ERRORS:
5648 		case VDEV_PROP_INITIALIZE_ERRORS:
5649 		case VDEV_PROP_TRIM_ERRORS:
5650 		case VDEV_PROP_SLOW_IOS:
5651 		case VDEV_PROP_OPS_NULL:
5652 		case VDEV_PROP_OPS_READ:
5653 		case VDEV_PROP_OPS_WRITE:
5654 		case VDEV_PROP_OPS_FREE:
5655 		case VDEV_PROP_OPS_CLAIM:
5656 		case VDEV_PROP_OPS_TRIM:
5657 		case VDEV_PROP_BYTES_NULL:
5658 		case VDEV_PROP_BYTES_READ:
5659 		case VDEV_PROP_BYTES_WRITE:
5660 		case VDEV_PROP_BYTES_FREE:
5661 		case VDEV_PROP_BYTES_CLAIM:
5662 		case VDEV_PROP_BYTES_TRIM:
5663 			if (literal) {
5664 				(void) snprintf(buf, len, "%llu",
5665 				    (u_longlong_t)intval);
5666 			} else {
5667 				(void) zfs_nicenum(intval, buf, len);
5668 			}
5669 			break;
5670 		case VDEV_PROP_EXPANDSZ:
5671 			if (intval == 0) {
5672 				(void) strlcpy(buf, "-", len);
5673 			} else if (literal) {
5674 				(void) snprintf(buf, len, "%llu",
5675 				    (u_longlong_t)intval);
5676 			} else {
5677 				(void) zfs_nicenum(intval, buf, len);
5678 			}
5679 			break;
5680 		case VDEV_PROP_CAPACITY:
5681 			if (literal) {
5682 				(void) snprintf(buf, len, "%llu",
5683 				    (u_longlong_t)intval);
5684 			} else {
5685 				(void) snprintf(buf, len, "%llu%%",
5686 				    (u_longlong_t)intval);
5687 			}
5688 			break;
5689 		case VDEV_PROP_CHECKSUM_N:
5690 		case VDEV_PROP_CHECKSUM_T:
5691 		case VDEV_PROP_IO_N:
5692 		case VDEV_PROP_IO_T:
5693 		case VDEV_PROP_SLOW_IO_N:
5694 		case VDEV_PROP_SLOW_IO_T:
5695 		case VDEV_PROP_FDOMAIN:
5696 		case VDEV_PROP_FGROUP:
5697 			if (intval == UINT64_MAX) {
5698 				(void) strlcpy(buf, "-", len);
5699 			} else {
5700 				(void) snprintf(buf, len, "%llu",
5701 				    (u_longlong_t)intval);
5702 			}
5703 			break;
5704 		case VDEV_PROP_FRAGMENTATION:
5705 			if (intval == UINT64_MAX) {
5706 				(void) strlcpy(buf, "-", len);
5707 			} else {
5708 				(void) snprintf(buf, len, "%llu%%",
5709 				    (u_longlong_t)intval);
5710 			}
5711 			break;
5712 		case VDEV_PROP_STATE:
5713 			if (literal) {
5714 				(void) snprintf(buf, len, "%llu",
5715 				    (u_longlong_t)intval);
5716 			} else {
5717 				(void) strlcpy(buf, zpool_state_to_name(intval,
5718 				    VDEV_AUX_NONE), len);
5719 			}
5720 			break;
5721 		default:
5722 			(void) snprintf(buf, len, "%llu",
5723 			    (u_longlong_t)intval);
5724 		}
5725 		break;
5726 
5727 	case PROP_TYPE_INDEX:
5728 		if (nvlist_lookup_nvlist(nvprop, prop_name, &nv) == 0) {
5729 			src = fnvlist_lookup_uint64(nv, ZPROP_SOURCE);
5730 			intval = fnvlist_lookup_uint64(nv, ZPROP_VALUE);
5731 		} else {
5732 			/* 'trim_support' only valid for leaf vdevs */
5733 			if (prop == VDEV_PROP_TRIM_SUPPORT) {
5734 				(void) strlcpy(buf, "-", len);
5735 				break;
5736 			}
5737 			src = ZPROP_SRC_DEFAULT;
5738 			intval = vdev_prop_default_numeric(prop);
5739 			/* Only use if provided by the RAIDZ VDEV above */
5740 			if (prop == VDEV_PROP_RAIDZ_EXPANDING)
5741 				return (ENOENT);
5742 			if (prop == VDEV_PROP_SIT_OUT)
5743 				return (ENOENT);
5744 		}
5745 		if (vdev_prop_index_to_string(prop, intval,
5746 		    (const char **)&strval) != 0)
5747 			return (-1);
5748 		(void) strlcpy(buf, strval, len);
5749 		break;
5750 
5751 	default:
5752 		abort();
5753 	}
5754 
5755 	if (srctype)
5756 		*srctype = src;
5757 
5758 	return (0);
5759 }
5760 
5761 /*
5762  * Get a vdev property value for 'prop_name' and return the value in
5763  * a pre-allocated buffer.
5764  */
5765 int
zpool_get_vdev_prop(zpool_handle_t * zhp,const char * vdevname,vdev_prop_t prop,char * prop_name,char * buf,size_t len,zprop_source_t * srctype,boolean_t literal)5766 zpool_get_vdev_prop(zpool_handle_t *zhp, const char *vdevname, vdev_prop_t prop,
5767     char *prop_name, char *buf, size_t len, zprop_source_t *srctype,
5768     boolean_t literal)
5769 {
5770 	nvlist_t *reqnvl, *reqprops;
5771 	nvlist_t *retprops = NULL;
5772 	uint64_t vdev_guid = 0;
5773 	int ret;
5774 
5775 	if ((ret = zpool_vdev_guid(zhp, vdevname, &vdev_guid)) != 0)
5776 		return (ret);
5777 
5778 	if (nvlist_alloc(&reqnvl, NV_UNIQUE_NAME, 0) != 0)
5779 		return (no_memory(zhp->zpool_hdl));
5780 	if (nvlist_alloc(&reqprops, NV_UNIQUE_NAME, 0) != 0)
5781 		return (no_memory(zhp->zpool_hdl));
5782 
5783 	fnvlist_add_uint64(reqnvl, ZPOOL_VDEV_PROPS_GET_VDEV, vdev_guid);
5784 
5785 	if (prop != VDEV_PROP_USERPROP) {
5786 		/* prop_name overrides prop value */
5787 		if (prop_name != NULL)
5788 			prop = vdev_name_to_prop(prop_name);
5789 		else
5790 			prop_name = (char *)vdev_prop_to_name(prop);
5791 		assert(prop < VDEV_NUM_PROPS);
5792 	}
5793 
5794 	assert(prop_name != NULL);
5795 	if (nvlist_add_uint64(reqprops, prop_name, prop) != 0) {
5796 		nvlist_free(reqnvl);
5797 		nvlist_free(reqprops);
5798 		return (no_memory(zhp->zpool_hdl));
5799 	}
5800 
5801 	fnvlist_add_nvlist(reqnvl, ZPOOL_VDEV_PROPS_GET_PROPS, reqprops);
5802 
5803 	ret = lzc_get_vdev_prop(zhp->zpool_name, reqnvl, &retprops);
5804 
5805 	if (ret == 0) {
5806 		ret = zpool_get_vdev_prop_value(retprops, prop, prop_name, buf,
5807 		    len, srctype, literal);
5808 	} else {
5809 		char errbuf[ERRBUFLEN];
5810 		(void) snprintf(errbuf, sizeof (errbuf),
5811 		    dgettext(TEXT_DOMAIN, "cannot get vdev property %s from"
5812 		    " %s in %s"), prop_name, vdevname, zhp->zpool_name);
5813 		(void) zpool_standard_error(zhp->zpool_hdl, ret, errbuf);
5814 	}
5815 
5816 	nvlist_free(reqnvl);
5817 	nvlist_free(reqprops);
5818 	nvlist_free(retprops);
5819 
5820 	return (ret);
5821 }
5822 
5823 /*
5824  * Get all vdev properties
5825  */
5826 int
zpool_get_all_vdev_props(zpool_handle_t * zhp,const char * vdevname,nvlist_t ** outnvl)5827 zpool_get_all_vdev_props(zpool_handle_t *zhp, const char *vdevname,
5828     nvlist_t **outnvl)
5829 {
5830 	nvlist_t *nvl = NULL;
5831 	uint64_t vdev_guid = 0;
5832 	int ret;
5833 
5834 	if ((ret = zpool_vdev_guid(zhp, vdevname, &vdev_guid)) != 0)
5835 		return (ret);
5836 
5837 	if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
5838 		return (no_memory(zhp->zpool_hdl));
5839 
5840 	fnvlist_add_uint64(nvl, ZPOOL_VDEV_PROPS_GET_VDEV, vdev_guid);
5841 
5842 	ret = lzc_get_vdev_prop(zhp->zpool_name, nvl, outnvl);
5843 
5844 	nvlist_free(nvl);
5845 
5846 	if (ret) {
5847 		char errbuf[ERRBUFLEN];
5848 		(void) snprintf(errbuf, sizeof (errbuf),
5849 		    dgettext(TEXT_DOMAIN, "cannot get vdev properties for"
5850 		    " %s in %s"), vdevname, zhp->zpool_name);
5851 		(void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
5852 	}
5853 
5854 	return (ret);
5855 }
5856 
5857 /*
5858  * Set vdev property
5859  */
5860 int
zpool_set_vdev_prop(zpool_handle_t * zhp,const char * vdevname,const char * propname,const char * propval)5861 zpool_set_vdev_prop(zpool_handle_t *zhp, const char *vdevname,
5862     const char *propname, const char *propval)
5863 {
5864 	int ret;
5865 	nvlist_t *nvl = NULL;
5866 	nvlist_t *outnvl = NULL;
5867 	nvlist_t *props;
5868 	nvlist_t *realprops;
5869 	prop_flags_t flags = { 0 };
5870 	uint64_t version;
5871 	uint64_t vdev_guid;
5872 
5873 	if ((ret = zpool_vdev_guid(zhp, vdevname, &vdev_guid)) != 0)
5874 		return (ret);
5875 
5876 	if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
5877 		return (no_memory(zhp->zpool_hdl));
5878 	if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0)
5879 		return (no_memory(zhp->zpool_hdl));
5880 
5881 	fnvlist_add_uint64(nvl, ZPOOL_VDEV_PROPS_SET_VDEV, vdev_guid);
5882 
5883 	if (nvlist_add_string(props, propname, propval) != 0) {
5884 		nvlist_free(props);
5885 		return (no_memory(zhp->zpool_hdl));
5886 	}
5887 
5888 	char errbuf[ERRBUFLEN];
5889 	(void) snprintf(errbuf, sizeof (errbuf),
5890 	    dgettext(TEXT_DOMAIN, "cannot set property %s for %s on %s"),
5891 	    propname, vdevname, zhp->zpool_name);
5892 
5893 	flags.vdevprop = 1;
5894 	version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
5895 	if ((realprops = zpool_valid_proplist(zhp->zpool_hdl,
5896 	    zhp->zpool_name, props, version, flags, errbuf)) == NULL) {
5897 		nvlist_free(props);
5898 		nvlist_free(nvl);
5899 		return (-1);
5900 	}
5901 
5902 	nvlist_free(props);
5903 	props = realprops;
5904 
5905 	fnvlist_add_nvlist(nvl, ZPOOL_VDEV_PROPS_SET_PROPS, props);
5906 
5907 	ret = lzc_set_vdev_prop(zhp->zpool_name, nvl, &outnvl);
5908 
5909 	nvlist_free(props);
5910 	nvlist_free(nvl);
5911 	nvlist_free(outnvl);
5912 
5913 	if (ret) {
5914 		if (errno == ENOTSUP) {
5915 			zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
5916 			    "property not supported for this vdev"));
5917 			(void) zfs_error(zhp->zpool_hdl, EZFS_PROPTYPE, errbuf);
5918 		} else {
5919 			(void) zpool_standard_error(zhp->zpool_hdl, errno,
5920 			    errbuf);
5921 		}
5922 	}
5923 
5924 	return (ret);
5925 }
5926 
5927 /*
5928  * Prune older entries from the DDT to reclaim space under the quota
5929  */
5930 int
zpool_ddt_prune(zpool_handle_t * zhp,zpool_ddt_prune_unit_t unit,uint64_t amount)5931 zpool_ddt_prune(zpool_handle_t *zhp, zpool_ddt_prune_unit_t unit,
5932     uint64_t amount)
5933 {
5934 	int error = lzc_ddt_prune(zhp->zpool_name, unit, amount);
5935 	if (error != 0) {
5936 		libzfs_handle_t *hdl = zhp->zpool_hdl;
5937 		char errbuf[ERRBUFLEN];
5938 
5939 		(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
5940 		    "cannot prune dedup table on '%s'"), zhp->zpool_name);
5941 
5942 		if (error == EALREADY) {
5943 			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
5944 			    "a prune operation is already in progress"));
5945 			(void) zfs_error(hdl, EZFS_BUSY, errbuf);
5946 		} else {
5947 			(void) zpool_standard_error(hdl, errno, errbuf);
5948 		}
5949 		return (-1);
5950 	}
5951 
5952 	return (0);
5953 }
5954