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