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 http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21
22 /*
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright (c) 2011, 2017 by Delphix. All rights reserved.
25 * Copyright (c) 2013, Joyent, Inc. All rights reserved.
26 * Copyright 2016 Nexenta Systems, Inc.
27 * Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>
28 * Copyright (c) 2017 Datto Inc.
29 */
30
31 #include <ctype.h>
32 #include <errno.h>
33 #include <devid.h>
34 #include <fcntl.h>
35 #include <libintl.h>
36 #include <stdio.h>
37 #include <stdlib.h>
38 #include <strings.h>
39 #include <unistd.h>
40 #include <libgen.h>
41 #include <sys/efi_partition.h>
42 #include <sys/vtoc.h>
43 #include <sys/zfs_ioctl.h>
44 #include <dlfcn.h>
45
46 #include "zfs_namecheck.h"
47 #include "zfs_prop.h"
48 #include "libzfs_impl.h"
49 #include "zfs_comutil.h"
50 #include "zfeature_common.h"
51
52 static int read_efi_label(nvlist_t *, diskaddr_t *, boolean_t *);
53 static boolean_t zpool_vdev_is_interior(const char *name);
54
55 #define BACKUP_SLICE "s2"
56
57 typedef struct prop_flags {
58 int create:1; /* Validate property on creation */
59 int import:1; /* Validate property on import */
60 } prop_flags_t;
61
62 /*
63 * ====================================================================
64 * zpool property functions
65 * ====================================================================
66 */
67
68 static int
69 zpool_get_all_props(zpool_handle_t *zhp)
70 {
71 zfs_cmd_t zc = { 0 };
72 libzfs_handle_t *hdl = zhp->zpool_hdl;
73
74 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
75
76 if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0)
77 return (-1);
78
79 while (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_PROPS, &zc) != 0) {
80 if (errno == ENOMEM) {
81 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
82 zcmd_free_nvlists(&zc);
83 return (-1);
84 }
85 } else {
86 zcmd_free_nvlists(&zc);
87 return (-1);
88 }
89 }
90
91 if (zcmd_read_dst_nvlist(hdl, &zc, &zhp->zpool_props) != 0) {
92 zcmd_free_nvlists(&zc);
93 return (-1);
94 }
95
96 zcmd_free_nvlists(&zc);
97
98 return (0);
99 }
100
101 static int
102 zpool_props_refresh(zpool_handle_t *zhp)
103 {
104 nvlist_t *old_props;
105
106 old_props = zhp->zpool_props;
107
108 if (zpool_get_all_props(zhp) != 0)
109 return (-1);
110
111 nvlist_free(old_props);
112 return (0);
113 }
114
115 static char *
116 zpool_get_prop_string(zpool_handle_t *zhp, zpool_prop_t prop,
117 zprop_source_t *src)
118 {
119 nvlist_t *nv, *nvl;
120 uint64_t ival;
121 char *value;
122 zprop_source_t source;
123
124 nvl = zhp->zpool_props;
125 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
126 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &ival) == 0);
127 source = ival;
128 verify(nvlist_lookup_string(nv, ZPROP_VALUE, &value) == 0);
129 } else {
130 source = ZPROP_SRC_DEFAULT;
131 if ((value = (char *)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 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &value) == 0);
167 source = value;
168 verify(nvlist_lookup_uint64(nv, ZPROP_VALUE, &value) == 0);
169 } else {
170 source = ZPROP_SRC_DEFAULT;
171 value = zpool_prop_default_numeric(prop);
172 }
173
174 if (src)
175 *src = source;
176
177 return (value);
178 }
179
180 /*
181 * Map VDEV STATE to printed strings.
182 */
183 const char *
184 zpool_state_to_name(vdev_state_t state, vdev_aux_t aux)
185 {
186 switch (state) {
187 case VDEV_STATE_CLOSED:
188 case VDEV_STATE_OFFLINE:
189 return (gettext("OFFLINE"));
190 case VDEV_STATE_REMOVED:
191 return (gettext("REMOVED"));
192 case VDEV_STATE_CANT_OPEN:
193 if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG)
194 return (gettext("FAULTED"));
195 else if (aux == VDEV_AUX_SPLIT_POOL)
196 return (gettext("SPLIT"));
197 else
198 return (gettext("UNAVAIL"));
199 case VDEV_STATE_FAULTED:
200 return (gettext("FAULTED"));
201 case VDEV_STATE_DEGRADED:
202 return (gettext("DEGRADED"));
203 case VDEV_STATE_HEALTHY:
204 return (gettext("ONLINE"));
205
206 default:
207 break;
208 }
209
210 return (gettext("UNKNOWN"));
211 }
212
213 /*
214 * Map POOL STATE to printed strings.
215 */
216 const char *
217 zpool_pool_state_to_name(pool_state_t state)
218 {
219 switch (state) {
220 case POOL_STATE_ACTIVE:
221 return (gettext("ACTIVE"));
222 case POOL_STATE_EXPORTED:
223 return (gettext("EXPORTED"));
224 case POOL_STATE_DESTROYED:
225 return (gettext("DESTROYED"));
226 case POOL_STATE_SPARE:
227 return (gettext("SPARE"));
228 case POOL_STATE_L2CACHE:
229 return (gettext("L2CACHE"));
230 case POOL_STATE_UNINITIALIZED:
231 return (gettext("UNINITIALIZED"));
232 case POOL_STATE_UNAVAIL:
233 return (gettext("UNAVAIL"));
234 case POOL_STATE_POTENTIALLY_ACTIVE:
235 return (gettext("POTENTIALLY_ACTIVE"));
236 }
237
238 return (gettext("UNKNOWN"));
239 }
240
241 /*
242 * Get a zpool property value for 'prop' and return the value in
243 * a pre-allocated buffer.
244 */
245 int
246 zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf, size_t len,
247 zprop_source_t *srctype, boolean_t literal)
248 {
249 uint64_t intval;
250 const char *strval;
251 zprop_source_t src = ZPROP_SRC_NONE;
252 nvlist_t *nvroot;
253 vdev_stat_t *vs;
254 uint_t vsc;
255
256 if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
257 switch (prop) {
258 case ZPOOL_PROP_NAME:
259 (void) strlcpy(buf, zpool_get_name(zhp), len);
260 break;
261
262 case ZPOOL_PROP_HEALTH:
263 (void) strlcpy(buf, "FAULTED", len);
264 break;
265
266 case ZPOOL_PROP_GUID:
267 intval = zpool_get_prop_int(zhp, prop, &src);
268 (void) snprintf(buf, len, "%llu", intval);
269 break;
270
271 case ZPOOL_PROP_ALTROOT:
272 case ZPOOL_PROP_CACHEFILE:
273 case ZPOOL_PROP_COMMENT:
274 if (zhp->zpool_props != NULL ||
275 zpool_get_all_props(zhp) == 0) {
276 (void) strlcpy(buf,
277 zpool_get_prop_string(zhp, prop, &src),
278 len);
279 break;
280 }
281 /* FALLTHROUGH */
282 default:
283 (void) strlcpy(buf, "-", len);
284 break;
285 }
286
287 if (srctype != NULL)
288 *srctype = src;
289 return (0);
290 }
291
292 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) &&
293 prop != ZPOOL_PROP_NAME)
294 return (-1);
295
296 switch (zpool_prop_get_type(prop)) {
297 case PROP_TYPE_STRING:
298 (void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src),
299 len);
300 break;
301
302 case PROP_TYPE_NUMBER:
303 intval = zpool_get_prop_int(zhp, prop, &src);
304
305 switch (prop) {
306 case ZPOOL_PROP_SIZE:
307 case ZPOOL_PROP_ALLOCATED:
308 case ZPOOL_PROP_FREE:
309 case ZPOOL_PROP_FREEING:
310 case ZPOOL_PROP_LEAKED:
311 if (literal) {
312 (void) snprintf(buf, len, "%llu",
313 (u_longlong_t)intval);
314 } else {
315 (void) zfs_nicenum(intval, buf, len);
316 }
317 break;
318 case ZPOOL_PROP_BOOTSIZE:
319 case ZPOOL_PROP_EXPANDSZ:
320 case ZPOOL_PROP_CHECKPOINT:
321 if (intval == 0) {
322 (void) strlcpy(buf, "-", len);
323 } else if (literal) {
324 (void) snprintf(buf, len, "%llu",
325 (u_longlong_t)intval);
326 } else {
327 (void) zfs_nicenum(intval, buf, len);
328 }
329 break;
330 case ZPOOL_PROP_CAPACITY:
331 if (literal) {
332 (void) snprintf(buf, len, "%llu",
333 (u_longlong_t)intval);
334 } else {
335 (void) snprintf(buf, len, "%llu%%",
336 (u_longlong_t)intval);
337 }
338 break;
339 case ZPOOL_PROP_FRAGMENTATION:
340 if (intval == UINT64_MAX) {
341 (void) strlcpy(buf, "-", len);
342 } else {
343 (void) snprintf(buf, len, "%llu%%",
344 (u_longlong_t)intval);
345 }
346 break;
347 case ZPOOL_PROP_DEDUPRATIO:
348 (void) snprintf(buf, len, "%llu.%02llux",
349 (u_longlong_t)(intval / 100),
350 (u_longlong_t)(intval % 100));
351 break;
352 case ZPOOL_PROP_HEALTH:
353 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
354 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
355 verify(nvlist_lookup_uint64_array(nvroot,
356 ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc)
357 == 0);
358
359 (void) strlcpy(buf, zpool_state_to_name(intval,
360 vs->vs_aux), len);
361 break;
362 case ZPOOL_PROP_VERSION:
363 if (intval >= SPA_VERSION_FEATURES) {
364 (void) snprintf(buf, len, "-");
365 break;
366 }
367 /* FALLTHROUGH */
368 default:
369 (void) snprintf(buf, len, "%llu", intval);
370 }
371 break;
372
373 case PROP_TYPE_INDEX:
374 intval = zpool_get_prop_int(zhp, prop, &src);
375 if (zpool_prop_index_to_string(prop, intval, &strval)
376 != 0)
377 return (-1);
378 (void) strlcpy(buf, strval, len);
379 break;
380
381 default:
382 abort();
383 }
384
385 if (srctype)
386 *srctype = src;
387
388 return (0);
389 }
390
391 /*
392 * Check if the bootfs name has the same pool name as it is set to.
393 * Assuming bootfs is a valid dataset name.
394 */
395 static boolean_t
396 bootfs_name_valid(const char *pool, char *bootfs)
397 {
398 int len = strlen(pool);
399 if (bootfs[0] == '\0')
400 return (B_TRUE);
401
402 if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT))
403 return (B_FALSE);
404
405 if (strncmp(pool, bootfs, len) == 0 &&
406 (bootfs[len] == '/' || bootfs[len] == '\0'))
407 return (B_TRUE);
408
409 return (B_FALSE);
410 }
411
412 boolean_t
413 zpool_is_bootable(zpool_handle_t *zhp)
414 {
415 char bootfs[ZFS_MAX_DATASET_NAME_LEN];
416
417 return (zpool_get_prop(zhp, ZPOOL_PROP_BOOTFS, bootfs,
418 sizeof (bootfs), NULL, B_FALSE) == 0 && strncmp(bootfs, "-",
419 sizeof (bootfs)) != 0);
420 }
421
422
423 /*
424 * Given an nvlist of zpool properties to be set, validate that they are
425 * correct, and parse any numeric properties (index, boolean, etc) if they are
426 * specified as strings.
427 */
428 static nvlist_t *
429 zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname,
430 nvlist_t *props, uint64_t version, prop_flags_t flags, char *errbuf)
431 {
432 nvpair_t *elem;
433 nvlist_t *retprops;
434 zpool_prop_t prop;
435 char *strval;
436 uint64_t intval;
437 char *slash, *check;
438 struct stat64 statbuf;
439 zpool_handle_t *zhp;
440
441 if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) {
442 (void) no_memory(hdl);
443 return (NULL);
444 }
445
446 elem = NULL;
447 while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
448 const char *propname = nvpair_name(elem);
449
450 prop = zpool_name_to_prop(propname);
451 if (prop == ZPOOL_PROP_INVAL && zpool_prop_feature(propname)) {
452 int err;
453 char *fname = strchr(propname, '@') + 1;
454
455 err = zfeature_lookup_name(fname, NULL);
456 if (err != 0) {
457 ASSERT3U(err, ==, ENOENT);
458 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
459 "invalid feature '%s'"), fname);
460 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
461 goto error;
462 }
463
464 if (nvpair_type(elem) != DATA_TYPE_STRING) {
465 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
466 "'%s' must be a string"), propname);
467 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
468 goto error;
469 }
470
471 (void) nvpair_value_string(elem, &strval);
472 if (strcmp(strval, ZFS_FEATURE_ENABLED) != 0) {
473 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
474 "property '%s' can only be set to "
475 "'enabled'"), propname);
476 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
477 goto error;
478 }
479
480 if (nvlist_add_uint64(retprops, propname, 0) != 0) {
481 (void) no_memory(hdl);
482 goto error;
483 }
484 continue;
485 }
486
487 /*
488 * Make sure this property is valid and applies to this type.
489 */
490 if (prop == ZPOOL_PROP_INVAL) {
491 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
492 "invalid property '%s'"), propname);
493 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
494 goto error;
495 }
496
497 if (zpool_prop_readonly(prop)) {
498 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
499 "is readonly"), propname);
500 (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
501 goto error;
502 }
503
504 if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops,
505 &strval, &intval, errbuf) != 0)
506 goto error;
507
508 /*
509 * Perform additional checking for specific properties.
510 */
511 switch (prop) {
512 case ZPOOL_PROP_VERSION:
513 if (intval < version ||
514 !SPA_VERSION_IS_SUPPORTED(intval)) {
515 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
516 "property '%s' number %d is invalid."),
517 propname, intval);
518 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
519 goto error;
520 }
521 break;
522
523 case ZPOOL_PROP_BOOTSIZE:
524 if (!flags.create) {
525 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
526 "property '%s' can only be set during pool "
527 "creation"), propname);
528 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
529 goto error;
530 }
531 break;
532
533 case ZPOOL_PROP_BOOTFS:
534 if (flags.create || flags.import) {
535 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
536 "property '%s' cannot be set at creation "
537 "or import time"), propname);
538 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
539 goto error;
540 }
541
542 if (version < SPA_VERSION_BOOTFS) {
543 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
544 "pool must be upgraded to support "
545 "'%s' property"), propname);
546 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
547 goto error;
548 }
549
550 /*
551 * bootfs property value has to be a dataset name and
552 * the dataset has to be in the same pool as it sets to.
553 */
554 if (!bootfs_name_valid(poolname, strval)) {
555 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
556 "is an invalid name"), strval);
557 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
558 goto error;
559 }
560
561 if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) {
562 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
563 "could not open pool '%s'"), poolname);
564 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
565 goto error;
566 }
567 zpool_close(zhp);
568 break;
569
570 case ZPOOL_PROP_ALTROOT:
571 if (!flags.create && !flags.import) {
572 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
573 "property '%s' can only be set during pool "
574 "creation or import"), propname);
575 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
576 goto error;
577 }
578
579 if (strval[0] != '/') {
580 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
581 "bad alternate root '%s'"), strval);
582 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
583 goto error;
584 }
585 break;
586
587 case ZPOOL_PROP_CACHEFILE:
588 if (strval[0] == '\0')
589 break;
590
591 if (strcmp(strval, "none") == 0)
592 break;
593
594 if (strval[0] != '/') {
595 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
596 "property '%s' must be empty, an "
597 "absolute path, or 'none'"), propname);
598 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
599 goto error;
600 }
601
602 slash = strrchr(strval, '/');
603
604 if (slash[1] == '\0' || strcmp(slash, "/.") == 0 ||
605 strcmp(slash, "/..") == 0) {
606 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
607 "'%s' is not a valid file"), strval);
608 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
609 goto error;
610 }
611
612 *slash = '\0';
613
614 if (strval[0] != '\0' &&
615 (stat64(strval, &statbuf) != 0 ||
616 !S_ISDIR(statbuf.st_mode))) {
617 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
618 "'%s' is not a valid directory"),
619 strval);
620 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
621 goto error;
622 }
623
624 *slash = '/';
625 break;
626
627 case ZPOOL_PROP_COMMENT:
628 for (check = strval; *check != '\0'; check++) {
629 if (!isprint(*check)) {
630 zfs_error_aux(hdl,
631 dgettext(TEXT_DOMAIN,
632 "comment may only have printable "
633 "characters"));
634 (void) zfs_error(hdl, EZFS_BADPROP,
635 errbuf);
636 goto error;
637 }
638 }
639 if (strlen(strval) > ZPROP_MAX_COMMENT) {
640 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
641 "comment must not exceed %d characters"),
642 ZPROP_MAX_COMMENT);
643 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
644 goto error;
645 }
646 break;
647 case ZPOOL_PROP_READONLY:
648 if (!flags.import) {
649 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
650 "property '%s' can only be set at "
651 "import time"), propname);
652 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
653 goto error;
654 }
655 break;
656
657 default:
658 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
659 "property '%s'(%d) not defined"), propname, prop);
660 break;
661 }
662 }
663
664 return (retprops);
665 error:
666 nvlist_free(retprops);
667 return (NULL);
668 }
669
670 /*
671 * Set zpool property : propname=propval.
672 */
673 int
674 zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval)
675 {
676 zfs_cmd_t zc = { 0 };
677 int ret = -1;
678 char errbuf[1024];
679 nvlist_t *nvl = NULL;
680 nvlist_t *realprops;
681 uint64_t version;
682 prop_flags_t flags = { 0 };
683
684 (void) snprintf(errbuf, sizeof (errbuf),
685 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
686 zhp->zpool_name);
687
688 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
689 return (no_memory(zhp->zpool_hdl));
690
691 if (nvlist_add_string(nvl, propname, propval) != 0) {
692 nvlist_free(nvl);
693 return (no_memory(zhp->zpool_hdl));
694 }
695
696 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
697 if ((realprops = zpool_valid_proplist(zhp->zpool_hdl,
698 zhp->zpool_name, nvl, version, flags, errbuf)) == NULL) {
699 nvlist_free(nvl);
700 return (-1);
701 }
702
703 nvlist_free(nvl);
704 nvl = realprops;
705
706 /*
707 * Execute the corresponding ioctl() to set this property.
708 */
709 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
710
711 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) {
712 nvlist_free(nvl);
713 return (-1);
714 }
715
716 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc);
717
718 zcmd_free_nvlists(&zc);
719 nvlist_free(nvl);
720
721 if (ret)
722 (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
723 else
724 (void) zpool_props_refresh(zhp);
725
726 return (ret);
727 }
728
729 int
730 zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp)
731 {
732 libzfs_handle_t *hdl = zhp->zpool_hdl;
733 zprop_list_t *entry;
734 char buf[ZFS_MAXPROPLEN];
735 nvlist_t *features = NULL;
736 zprop_list_t **last;
737 boolean_t firstexpand = (NULL == *plp);
738
739 if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0)
740 return (-1);
741
742 last = plp;
743 while (*last != NULL)
744 last = &(*last)->pl_next;
745
746 if ((*plp)->pl_all)
747 features = zpool_get_features(zhp);
748
749 if ((*plp)->pl_all && firstexpand) {
750 for (int i = 0; i < SPA_FEATURES; i++) {
751 zprop_list_t *entry = zfs_alloc(hdl,
752 sizeof (zprop_list_t));
753 entry->pl_prop = ZPROP_INVAL;
754 entry->pl_user_prop = zfs_asprintf(hdl, "feature@%s",
755 spa_feature_table[i].fi_uname);
756 entry->pl_width = strlen(entry->pl_user_prop);
757 entry->pl_all = B_TRUE;
758
759 *last = entry;
760 last = &entry->pl_next;
761 }
762 }
763
764 /* add any unsupported features */
765 for (nvpair_t *nvp = nvlist_next_nvpair(features, NULL);
766 nvp != NULL; nvp = nvlist_next_nvpair(features, nvp)) {
767 char *propname;
768 boolean_t found;
769 zprop_list_t *entry;
770
771 if (zfeature_is_supported(nvpair_name(nvp)))
772 continue;
773
774 propname = zfs_asprintf(hdl, "unsupported@%s",
775 nvpair_name(nvp));
776
777 /*
778 * Before adding the property to the list make sure that no
779 * other pool already added the same property.
780 */
781 found = B_FALSE;
782 entry = *plp;
783 while (entry != NULL) {
784 if (entry->pl_user_prop != NULL &&
785 strcmp(propname, entry->pl_user_prop) == 0) {
786 found = B_TRUE;
787 break;
788 }
789 entry = entry->pl_next;
790 }
791 if (found) {
792 free(propname);
793 continue;
794 }
795
796 entry = zfs_alloc(hdl, sizeof (zprop_list_t));
797 entry->pl_prop = ZPROP_INVAL;
798 entry->pl_user_prop = propname;
799 entry->pl_width = strlen(entry->pl_user_prop);
800 entry->pl_all = B_TRUE;
801
802 *last = entry;
803 last = &entry->pl_next;
804 }
805
806 for (entry = *plp; entry != NULL; entry = entry->pl_next) {
807
808 if (entry->pl_fixed)
809 continue;
810
811 if (entry->pl_prop != ZPROP_INVAL &&
812 zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf),
813 NULL, B_FALSE) == 0) {
814 if (strlen(buf) > entry->pl_width)
815 entry->pl_width = strlen(buf);
816 }
817 }
818
819 return (0);
820 }
821
822 /*
823 * Get the state for the given feature on the given ZFS pool.
824 */
825 int
826 zpool_prop_get_feature(zpool_handle_t *zhp, const char *propname, char *buf,
827 size_t len)
828 {
829 uint64_t refcount;
830 boolean_t found = B_FALSE;
831 nvlist_t *features = zpool_get_features(zhp);
832 boolean_t supported;
833 const char *feature = strchr(propname, '@') + 1;
834
835 supported = zpool_prop_feature(propname);
836 ASSERT(supported || zpool_prop_unsupported(propname));
837
838 /*
839 * Convert from feature name to feature guid. This conversion is
840 * unecessary for unsupported@... properties because they already
841 * use guids.
842 */
843 if (supported) {
844 int ret;
845 spa_feature_t fid;
846
847 ret = zfeature_lookup_name(feature, &fid);
848 if (ret != 0) {
849 (void) strlcpy(buf, "-", len);
850 return (ENOTSUP);
851 }
852 feature = spa_feature_table[fid].fi_guid;
853 }
854
855 if (nvlist_lookup_uint64(features, feature, &refcount) == 0)
856 found = B_TRUE;
857
858 if (supported) {
859 if (!found) {
860 (void) strlcpy(buf, ZFS_FEATURE_DISABLED, len);
861 } else {
862 if (refcount == 0)
863 (void) strlcpy(buf, ZFS_FEATURE_ENABLED, len);
864 else
865 (void) strlcpy(buf, ZFS_FEATURE_ACTIVE, len);
866 }
867 } else {
868 if (found) {
869 if (refcount == 0) {
870 (void) strcpy(buf, ZFS_UNSUPPORTED_INACTIVE);
871 } else {
872 (void) strcpy(buf, ZFS_UNSUPPORTED_READONLY);
873 }
874 } else {
875 (void) strlcpy(buf, "-", len);
876 return (ENOTSUP);
877 }
878 }
879
880 return (0);
881 }
882
883 /*
884 * Don't start the slice at the default block of 34; many storage
885 * devices will use a stripe width of 128k, so start there instead.
886 */
887 #define NEW_START_BLOCK 256
888
889 /*
890 * Validate the given pool name, optionally putting an extended error message in
891 * 'buf'.
892 */
893 boolean_t
894 zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool)
895 {
896 namecheck_err_t why;
897 char what;
898 int ret;
899
900 ret = pool_namecheck(pool, &why, &what);
901
902 /*
903 * The rules for reserved pool names were extended at a later point.
904 * But we need to support users with existing pools that may now be
905 * invalid. So we only check for this expanded set of names during a
906 * create (or import), and only in userland.
907 */
908 if (ret == 0 && !isopen &&
909 (strncmp(pool, "mirror", 6) == 0 ||
910 strncmp(pool, "raidz", 5) == 0 ||
911 strncmp(pool, "spare", 5) == 0 ||
912 strcmp(pool, "log") == 0)) {
913 if (hdl != NULL)
914 zfs_error_aux(hdl,
915 dgettext(TEXT_DOMAIN, "name is reserved"));
916 return (B_FALSE);
917 }
918
919
920 if (ret != 0) {
921 if (hdl != NULL) {
922 switch (why) {
923 case NAME_ERR_TOOLONG:
924 zfs_error_aux(hdl,
925 dgettext(TEXT_DOMAIN, "name is too long"));
926 break;
927
928 case NAME_ERR_INVALCHAR:
929 zfs_error_aux(hdl,
930 dgettext(TEXT_DOMAIN, "invalid character "
931 "'%c' in pool name"), what);
932 break;
933
934 case NAME_ERR_NOLETTER:
935 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
936 "name must begin with a letter"));
937 break;
938
939 case NAME_ERR_RESERVED:
940 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
941 "name is reserved"));
942 break;
943
944 case NAME_ERR_DISKLIKE:
945 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
946 "pool name is reserved"));
947 break;
948
949 case NAME_ERR_LEADING_SLASH:
950 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
951 "leading slash in name"));
952 break;
953
954 case NAME_ERR_EMPTY_COMPONENT:
955 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
956 "empty component in name"));
957 break;
958
959 case NAME_ERR_TRAILING_SLASH:
960 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
961 "trailing slash in name"));
962 break;
963
964 case NAME_ERR_MULTIPLE_DELIMITERS:
965 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
966 "multiple '@' and/or '#' delimiters in "
967 "name"));
968 break;
969
970 default:
971 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
972 "(%d) not defined"), why);
973 break;
974 }
975 }
976 return (B_FALSE);
977 }
978
979 return (B_TRUE);
980 }
981
982 /*
983 * Open a handle to the given pool, even if the pool is currently in the FAULTED
984 * state.
985 */
986 zpool_handle_t *
987 zpool_open_canfail(libzfs_handle_t *hdl, const char *pool)
988 {
989 zpool_handle_t *zhp;
990 boolean_t missing;
991
992 /*
993 * Make sure the pool name is valid.
994 */
995 if (!zpool_name_valid(hdl, B_TRUE, pool)) {
996 (void) zfs_error_fmt(hdl, EZFS_INVALIDNAME,
997 dgettext(TEXT_DOMAIN, "cannot open '%s'"),
998 pool);
999 return (NULL);
1000 }
1001
1002 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
1003 return (NULL);
1004
1005 zhp->zpool_hdl = hdl;
1006 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
1007
1008 if (zpool_refresh_stats(zhp, &missing) != 0) {
1009 zpool_close(zhp);
1010 return (NULL);
1011 }
1012
1013 if (missing) {
1014 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool"));
1015 (void) zfs_error_fmt(hdl, EZFS_NOENT,
1016 dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool);
1017 zpool_close(zhp);
1018 return (NULL);
1019 }
1020
1021 return (zhp);
1022 }
1023
1024 /*
1025 * Like the above, but silent on error. Used when iterating over pools (because
1026 * the configuration cache may be out of date).
1027 */
1028 int
1029 zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret)
1030 {
1031 zpool_handle_t *zhp;
1032 boolean_t missing;
1033
1034 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
1035 return (-1);
1036
1037 zhp->zpool_hdl = hdl;
1038 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
1039
1040 if (zpool_refresh_stats(zhp, &missing) != 0) {
1041 zpool_close(zhp);
1042 return (-1);
1043 }
1044
1045 if (missing) {
1046 zpool_close(zhp);
1047 *ret = NULL;
1048 return (0);
1049 }
1050
1051 *ret = zhp;
1052 return (0);
1053 }
1054
1055 /*
1056 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted
1057 * state.
1058 */
1059 zpool_handle_t *
1060 zpool_open(libzfs_handle_t *hdl, const char *pool)
1061 {
1062 zpool_handle_t *zhp;
1063
1064 if ((zhp = zpool_open_canfail(hdl, pool)) == NULL)
1065 return (NULL);
1066
1067 if (zhp->zpool_state == POOL_STATE_UNAVAIL) {
1068 (void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
1069 dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name);
1070 zpool_close(zhp);
1071 return (NULL);
1072 }
1073
1074 return (zhp);
1075 }
1076
1077 /*
1078 * Close the handle. Simply frees the memory associated with the handle.
1079 */
1080 void
1081 zpool_close(zpool_handle_t *zhp)
1082 {
1083 nvlist_free(zhp->zpool_config);
1084 nvlist_free(zhp->zpool_old_config);
1085 nvlist_free(zhp->zpool_props);
1086 free(zhp);
1087 }
1088
1089 /*
1090 * Return the name of the pool.
1091 */
1092 const char *
1093 zpool_get_name(zpool_handle_t *zhp)
1094 {
1095 return (zhp->zpool_name);
1096 }
1097
1098
1099 /*
1100 * Return the state of the pool (ACTIVE or UNAVAILABLE)
1101 */
1102 int
1103 zpool_get_state(zpool_handle_t *zhp)
1104 {
1105 return (zhp->zpool_state);
1106 }
1107
1108 /*
1109 * Create the named pool, using the provided vdev list. It is assumed
1110 * that the consumer has already validated the contents of the nvlist, so we
1111 * don't have to worry about error semantics.
1112 */
1113 int
1114 zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot,
1115 nvlist_t *props, nvlist_t *fsprops)
1116 {
1117 zfs_cmd_t zc = { 0 };
1118 nvlist_t *zc_fsprops = NULL;
1119 nvlist_t *zc_props = NULL;
1120 char msg[1024];
1121 int ret = -1;
1122
1123 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1124 "cannot create '%s'"), pool);
1125
1126 if (!zpool_name_valid(hdl, B_FALSE, pool))
1127 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
1128
1129 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1130 return (-1);
1131
1132 if (props) {
1133 prop_flags_t flags = { .create = B_TRUE, .import = B_FALSE };
1134
1135 if ((zc_props = zpool_valid_proplist(hdl, pool, props,
1136 SPA_VERSION_1, flags, msg)) == NULL) {
1137 goto create_failed;
1138 }
1139 }
1140
1141 if (fsprops) {
1142 uint64_t zoned;
1143 char *zonestr;
1144
1145 zoned = ((nvlist_lookup_string(fsprops,
1146 zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) &&
1147 strcmp(zonestr, "on") == 0);
1148
1149 if ((zc_fsprops = zfs_valid_proplist(hdl, ZFS_TYPE_FILESYSTEM,
1150 fsprops, zoned, NULL, NULL, msg)) == NULL) {
1151 goto create_failed;
1152 }
1153 if (!zc_props &&
1154 (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) {
1155 goto create_failed;
1156 }
1157 if (nvlist_add_nvlist(zc_props,
1158 ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) {
1159 goto create_failed;
1160 }
1161 }
1162
1163 if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
1164 goto create_failed;
1165
1166 (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name));
1167
1168 if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) {
1169
1170 zcmd_free_nvlists(&zc);
1171 nvlist_free(zc_props);
1172 nvlist_free(zc_fsprops);
1173
1174 switch (errno) {
1175 case EBUSY:
1176 /*
1177 * This can happen if the user has specified the same
1178 * device multiple times. We can't reliably detect this
1179 * until we try to add it and see we already have a
1180 * label.
1181 */
1182 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1183 "one or more vdevs refer to the same device"));
1184 return (zfs_error(hdl, EZFS_BADDEV, msg));
1185
1186 case ERANGE:
1187 /*
1188 * This happens if the record size is smaller or larger
1189 * than the allowed size range, or not a power of 2.
1190 *
1191 * NOTE: although zfs_valid_proplist is called earlier,
1192 * this case may have slipped through since the
1193 * pool does not exist yet and it is therefore
1194 * impossible to read properties e.g. max blocksize
1195 * from the pool.
1196 */
1197 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1198 "record size invalid"));
1199 return (zfs_error(hdl, EZFS_BADPROP, msg));
1200
1201 case EOVERFLOW:
1202 /*
1203 * This occurs when one of the devices is below
1204 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1205 * device was the problem device since there's no
1206 * reliable way to determine device size from userland.
1207 */
1208 {
1209 char buf[64];
1210
1211 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
1212
1213 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1214 "one or more devices is less than the "
1215 "minimum size (%s)"), buf);
1216 }
1217 return (zfs_error(hdl, EZFS_BADDEV, msg));
1218
1219 case ENOSPC:
1220 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1221 "one or more devices is out of space"));
1222 return (zfs_error(hdl, EZFS_BADDEV, msg));
1223
1224 case ENOTBLK:
1225 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1226 "cache device must be a disk or disk slice"));
1227 return (zfs_error(hdl, EZFS_BADDEV, msg));
1228
1229 default:
1230 return (zpool_standard_error(hdl, errno, msg));
1231 }
1232 }
1233
1234 create_failed:
1235 zcmd_free_nvlists(&zc);
1236 nvlist_free(zc_props);
1237 nvlist_free(zc_fsprops);
1238 return (ret);
1239 }
1240
1241 /*
1242 * Destroy the given pool. It is up to the caller to ensure that there are no
1243 * datasets left in the pool.
1244 */
1245 int
1246 zpool_destroy(zpool_handle_t *zhp, const char *log_str)
1247 {
1248 zfs_cmd_t zc = { 0 };
1249 zfs_handle_t *zfp = NULL;
1250 libzfs_handle_t *hdl = zhp->zpool_hdl;
1251 char msg[1024];
1252
1253 if (zhp->zpool_state == POOL_STATE_ACTIVE &&
1254 (zfp = zfs_open(hdl, zhp->zpool_name, ZFS_TYPE_FILESYSTEM)) == NULL)
1255 return (-1);
1256
1257 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1258 zc.zc_history = (uint64_t)(uintptr_t)log_str;
1259
1260 if (zfs_ioctl(hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) {
1261 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1262 "cannot destroy '%s'"), zhp->zpool_name);
1263
1264 if (errno == EROFS) {
1265 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1266 "one or more devices is read only"));
1267 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1268 } else {
1269 (void) zpool_standard_error(hdl, errno, msg);
1270 }
1271
1272 if (zfp)
1273 zfs_close(zfp);
1274 return (-1);
1275 }
1276
1277 if (zfp) {
1278 remove_mountpoint(zfp);
1279 zfs_close(zfp);
1280 }
1281
1282 return (0);
1283 }
1284
1285 /*
1286 * Create a checkpoint in the given pool.
1287 */
1288 int
1289 zpool_checkpoint(zpool_handle_t *zhp)
1290 {
1291 libzfs_handle_t *hdl = zhp->zpool_hdl;
1292 char msg[1024];
1293 int error;
1294
1295 error = lzc_pool_checkpoint(zhp->zpool_name);
1296 if (error != 0) {
1297 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1298 "cannot checkpoint '%s'"), zhp->zpool_name);
1299 (void) zpool_standard_error(hdl, error, msg);
1300 return (-1);
1301 }
1302
1303 return (0);
1304 }
1305
1306 /*
1307 * Discard the checkpoint from the given pool.
1308 */
1309 int
1310 zpool_discard_checkpoint(zpool_handle_t *zhp)
1311 {
1312 libzfs_handle_t *hdl = zhp->zpool_hdl;
1313 char msg[1024];
1314 int error;
1315
1316 error = lzc_pool_checkpoint_discard(zhp->zpool_name);
1317 if (error != 0) {
1318 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1319 "cannot discard checkpoint in '%s'"), zhp->zpool_name);
1320 (void) zpool_standard_error(hdl, error, msg);
1321 return (-1);
1322 }
1323
1324 return (0);
1325 }
1326
1327 /*
1328 * Add the given vdevs to the pool. The caller must have already performed the
1329 * necessary verification to ensure that the vdev specification is well-formed.
1330 */
1331 int
1332 zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot)
1333 {
1334 zfs_cmd_t zc = { 0 };
1335 int ret;
1336 libzfs_handle_t *hdl = zhp->zpool_hdl;
1337 char msg[1024];
1338 nvlist_t **spares, **l2cache;
1339 uint_t nspares, nl2cache;
1340
1341 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1342 "cannot add to '%s'"), zhp->zpool_name);
1343
1344 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1345 SPA_VERSION_SPARES &&
1346 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
1347 &spares, &nspares) == 0) {
1348 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1349 "upgraded to add hot spares"));
1350 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1351 }
1352
1353 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1354 SPA_VERSION_L2CACHE &&
1355 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
1356 &l2cache, &nl2cache) == 0) {
1357 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1358 "upgraded to add cache devices"));
1359 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1360 }
1361
1362 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1363 return (-1);
1364 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1365
1366 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) {
1367 switch (errno) {
1368 case EBUSY:
1369 /*
1370 * This can happen if the user has specified the same
1371 * device multiple times. We can't reliably detect this
1372 * until we try to add it and see we already have a
1373 * label.
1374 */
1375 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1376 "one or more vdevs refer to the same device"));
1377 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1378 break;
1379
1380 case EINVAL:
1381 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1382 "invalid config; a pool with removing/removed "
1383 "vdevs does not support adding raidz vdevs"));
1384 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1385 break;
1386
1387 case EOVERFLOW:
1388 /*
1389 * This occurrs when one of the devices is below
1390 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1391 * device was the problem device since there's no
1392 * reliable way to determine device size from userland.
1393 */
1394 {
1395 char buf[64];
1396
1397 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
1398
1399 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1400 "device is less than the minimum "
1401 "size (%s)"), buf);
1402 }
1403 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1404 break;
1405
1406 case ENOTSUP:
1407 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1408 "pool must be upgraded to add these vdevs"));
1409 (void) zfs_error(hdl, EZFS_BADVERSION, msg);
1410 break;
1411
1412 case EDOM:
1413 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1414 "root pool can not have multiple vdevs"
1415 " or separate logs"));
1416 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, msg);
1417 break;
1418
1419 case ENOTBLK:
1420 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1421 "cache device must be a disk or disk slice"));
1422 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1423 break;
1424
1425 default:
1426 (void) zpool_standard_error(hdl, errno, msg);
1427 }
1428
1429 ret = -1;
1430 } else {
1431 ret = 0;
1432 }
1433
1434 zcmd_free_nvlists(&zc);
1435
1436 return (ret);
1437 }
1438
1439 /*
1440 * Exports the pool from the system. The caller must ensure that there are no
1441 * mounted datasets in the pool.
1442 */
1443 static int
1444 zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce,
1445 const char *log_str)
1446 {
1447 zfs_cmd_t zc = { 0 };
1448 char msg[1024];
1449
1450 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1451 "cannot export '%s'"), zhp->zpool_name);
1452
1453 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1454 zc.zc_cookie = force;
1455 zc.zc_guid = hardforce;
1456 zc.zc_history = (uint64_t)(uintptr_t)log_str;
1457
1458 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) {
1459 switch (errno) {
1460 case EXDEV:
1461 zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
1462 "use '-f' to override the following errors:\n"
1463 "'%s' has an active shared spare which could be"
1464 " used by other pools once '%s' is exported."),
1465 zhp->zpool_name, zhp->zpool_name);
1466 return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE,
1467 msg));
1468 default:
1469 return (zpool_standard_error_fmt(zhp->zpool_hdl, errno,
1470 msg));
1471 }
1472 }
1473
1474 return (0);
1475 }
1476
1477 int
1478 zpool_export(zpool_handle_t *zhp, boolean_t force, const char *log_str)
1479 {
1480 return (zpool_export_common(zhp, force, B_FALSE, log_str));
1481 }
1482
1483 int
1484 zpool_export_force(zpool_handle_t *zhp, const char *log_str)
1485 {
1486 return (zpool_export_common(zhp, B_TRUE, B_TRUE, log_str));
1487 }
1488
1489 static void
1490 zpool_rewind_exclaim(libzfs_handle_t *hdl, const char *name, boolean_t dryrun,
1491 nvlist_t *config)
1492 {
1493 nvlist_t *nv = NULL;
1494 uint64_t rewindto;
1495 int64_t loss = -1;
1496 struct tm t;
1497 char timestr[128];
1498
1499 if (!hdl->libzfs_printerr || config == NULL)
1500 return;
1501
1502 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
1503 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0) {
1504 return;
1505 }
1506
1507 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1508 return;
1509 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1510
1511 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1512 strftime(timestr, 128, 0, &t) != 0) {
1513 if (dryrun) {
1514 (void) printf(dgettext(TEXT_DOMAIN,
1515 "Would be able to return %s "
1516 "to its state as of %s.\n"),
1517 name, timestr);
1518 } else {
1519 (void) printf(dgettext(TEXT_DOMAIN,
1520 "Pool %s returned to its state as of %s.\n"),
1521 name, timestr);
1522 }
1523 if (loss > 120) {
1524 (void) printf(dgettext(TEXT_DOMAIN,
1525 "%s approximately %lld "),
1526 dryrun ? "Would discard" : "Discarded",
1527 (loss + 30) / 60);
1528 (void) printf(dgettext(TEXT_DOMAIN,
1529 "minutes of transactions.\n"));
1530 } else if (loss > 0) {
1531 (void) printf(dgettext(TEXT_DOMAIN,
1532 "%s approximately %lld "),
1533 dryrun ? "Would discard" : "Discarded", loss);
1534 (void) printf(dgettext(TEXT_DOMAIN,
1535 "seconds of transactions.\n"));
1536 }
1537 }
1538 }
1539
1540 void
1541 zpool_explain_recover(libzfs_handle_t *hdl, const char *name, int reason,
1542 nvlist_t *config)
1543 {
1544 nvlist_t *nv = NULL;
1545 int64_t loss = -1;
1546 uint64_t edata = UINT64_MAX;
1547 uint64_t rewindto;
1548 struct tm t;
1549 char timestr[128];
1550
1551 if (!hdl->libzfs_printerr)
1552 return;
1553
1554 if (reason >= 0)
1555 (void) printf(dgettext(TEXT_DOMAIN, "action: "));
1556 else
1557 (void) printf(dgettext(TEXT_DOMAIN, "\t"));
1558
1559 /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */
1560 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
1561 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0 ||
1562 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1563 goto no_info;
1564
1565 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1566 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_DATA_ERRORS,
1567 &edata);
1568
1569 (void) printf(dgettext(TEXT_DOMAIN,
1570 "Recovery is possible, but will result in some data loss.\n"));
1571
1572 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1573 strftime(timestr, 128, 0, &t) != 0) {
1574 (void) printf(dgettext(TEXT_DOMAIN,
1575 "\tReturning the pool to its state as of %s\n"
1576 "\tshould correct the problem. "),
1577 timestr);
1578 } else {
1579 (void) printf(dgettext(TEXT_DOMAIN,
1580 "\tReverting the pool to an earlier state "
1581 "should correct the problem.\n\t"));
1582 }
1583
1584 if (loss > 120) {
1585 (void) printf(dgettext(TEXT_DOMAIN,
1586 "Approximately %lld minutes of data\n"
1587 "\tmust be discarded, irreversibly. "), (loss + 30) / 60);
1588 } else if (loss > 0) {
1589 (void) printf(dgettext(TEXT_DOMAIN,
1590 "Approximately %lld seconds of data\n"
1591 "\tmust be discarded, irreversibly. "), loss);
1592 }
1593 if (edata != 0 && edata != UINT64_MAX) {
1594 if (edata == 1) {
1595 (void) printf(dgettext(TEXT_DOMAIN,
1596 "After rewind, at least\n"
1597 "\tone persistent user-data error will remain. "));
1598 } else {
1599 (void) printf(dgettext(TEXT_DOMAIN,
1600 "After rewind, several\n"
1601 "\tpersistent user-data errors will remain. "));
1602 }
1603 }
1604 (void) printf(dgettext(TEXT_DOMAIN,
1605 "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "),
1606 reason >= 0 ? "clear" : "import", name);
1607
1608 (void) printf(dgettext(TEXT_DOMAIN,
1609 "A scrub of the pool\n"
1610 "\tis strongly recommended after recovery.\n"));
1611 return;
1612
1613 no_info:
1614 (void) printf(dgettext(TEXT_DOMAIN,
1615 "Destroy and re-create the pool from\n\ta backup source.\n"));
1616 }
1617
1618 /*
1619 * zpool_import() is a contracted interface. Should be kept the same
1620 * if possible.
1621 *
1622 * Applications should use zpool_import_props() to import a pool with
1623 * new properties value to be set.
1624 */
1625 int
1626 zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1627 char *altroot)
1628 {
1629 nvlist_t *props = NULL;
1630 int ret;
1631
1632 if (altroot != NULL) {
1633 if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) {
1634 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1635 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1636 newname));
1637 }
1638
1639 if (nvlist_add_string(props,
1640 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 ||
1641 nvlist_add_string(props,
1642 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) {
1643 nvlist_free(props);
1644 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1645 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1646 newname));
1647 }
1648 }
1649
1650 ret = zpool_import_props(hdl, config, newname, props,
1651 ZFS_IMPORT_NORMAL);
1652 nvlist_free(props);
1653 return (ret);
1654 }
1655
1656 static void
1657 print_vdev_tree(libzfs_handle_t *hdl, const char *name, nvlist_t *nv,
1658 int indent)
1659 {
1660 nvlist_t **child;
1661 uint_t c, children;
1662 char *vname;
1663 uint64_t is_log = 0;
1664
1665 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG,
1666 &is_log);
1667
1668 if (name != NULL)
1669 (void) printf("\t%*s%s%s\n", indent, "", name,
1670 is_log ? " [log]" : "");
1671
1672 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1673 &child, &children) != 0)
1674 return;
1675
1676 for (c = 0; c < children; c++) {
1677 vname = zpool_vdev_name(hdl, NULL, child[c], B_TRUE);
1678 print_vdev_tree(hdl, vname, child[c], indent + 2);
1679 free(vname);
1680 }
1681 }
1682
1683 void
1684 zpool_print_unsup_feat(nvlist_t *config)
1685 {
1686 nvlist_t *nvinfo, *unsup_feat;
1687
1688 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nvinfo) ==
1689 0);
1690 verify(nvlist_lookup_nvlist(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT,
1691 &unsup_feat) == 0);
1692
1693 for (nvpair_t *nvp = nvlist_next_nvpair(unsup_feat, NULL); nvp != NULL;
1694 nvp = nvlist_next_nvpair(unsup_feat, nvp)) {
1695 char *desc;
1696
1697 verify(nvpair_type(nvp) == DATA_TYPE_STRING);
1698 verify(nvpair_value_string(nvp, &desc) == 0);
1699
1700 if (strlen(desc) > 0)
1701 (void) printf("\t%s (%s)\n", nvpair_name(nvp), desc);
1702 else
1703 (void) printf("\t%s\n", nvpair_name(nvp));
1704 }
1705 }
1706
1707 /*
1708 * Import the given pool using the known configuration and a list of
1709 * properties to be set. The configuration should have come from
1710 * zpool_find_import(). The 'newname' parameters control whether the pool
1711 * is imported with a different name.
1712 */
1713 int
1714 zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1715 nvlist_t *props, int flags)
1716 {
1717 zfs_cmd_t zc = { 0 };
1718 zpool_load_policy_t policy;
1719 nvlist_t *nv = NULL;
1720 nvlist_t *nvinfo = NULL;
1721 nvlist_t *missing = NULL;
1722 char *thename;
1723 char *origname;
1724 int ret;
1725 int error = 0;
1726 char errbuf[1024];
1727
1728 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1729 &origname) == 0);
1730
1731 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1732 "cannot import pool '%s'"), origname);
1733
1734 if (newname != NULL) {
1735 if (!zpool_name_valid(hdl, B_FALSE, newname))
1736 return (zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1737 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1738 newname));
1739 thename = (char *)newname;
1740 } else {
1741 thename = origname;
1742 }
1743
1744 if (props != NULL) {
1745 uint64_t version;
1746 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
1747
1748 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
1749 &version) == 0);
1750
1751 if ((props = zpool_valid_proplist(hdl, origname,
1752 props, version, flags, errbuf)) == NULL)
1753 return (-1);
1754 if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) {
1755 nvlist_free(props);
1756 return (-1);
1757 }
1758 nvlist_free(props);
1759 }
1760
1761 (void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name));
1762
1763 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1764 &zc.zc_guid) == 0);
1765
1766 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) {
1767 zcmd_free_nvlists(&zc);
1768 return (-1);
1769 }
1770 if (zcmd_alloc_dst_nvlist(hdl, &zc, zc.zc_nvlist_conf_size * 2) != 0) {
1771 zcmd_free_nvlists(&zc);
1772 return (-1);
1773 }
1774
1775 zc.zc_cookie = flags;
1776 while ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc)) != 0 &&
1777 errno == ENOMEM) {
1778 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
1779 zcmd_free_nvlists(&zc);
1780 return (-1);
1781 }
1782 }
1783 if (ret != 0)
1784 error = errno;
1785
1786 (void) zcmd_read_dst_nvlist(hdl, &zc, &nv);
1787
1788 zcmd_free_nvlists(&zc);
1789
1790 zpool_get_load_policy(config, &policy);
1791
1792 if (error) {
1793 char desc[1024];
1794
1795 /*
1796 * Dry-run failed, but we print out what success
1797 * looks like if we found a best txg
1798 */
1799 if (policy.zlp_rewind & ZPOOL_TRY_REWIND) {
1800 zpool_rewind_exclaim(hdl, newname ? origname : thename,
1801 B_TRUE, nv);
1802 nvlist_free(nv);
1803 return (-1);
1804 }
1805
1806 if (newname == NULL)
1807 (void) snprintf(desc, sizeof (desc),
1808 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1809 thename);
1810 else
1811 (void) snprintf(desc, sizeof (desc),
1812 dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"),
1813 origname, thename);
1814
1815 switch (error) {
1816 case ENOTSUP:
1817 if (nv != NULL && nvlist_lookup_nvlist(nv,
1818 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
1819 nvlist_exists(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT)) {
1820 (void) printf(dgettext(TEXT_DOMAIN, "This "
1821 "pool uses the following feature(s) not "
1822 "supported by this system:\n"));
1823 zpool_print_unsup_feat(nv);
1824 if (nvlist_exists(nvinfo,
1825 ZPOOL_CONFIG_CAN_RDONLY)) {
1826 (void) printf(dgettext(TEXT_DOMAIN,
1827 "All unsupported features are only "
1828 "required for writing to the pool."
1829 "\nThe pool can be imported using "
1830 "'-o readonly=on'.\n"));
1831 }
1832 }
1833 /*
1834 * Unsupported version.
1835 */
1836 (void) zfs_error(hdl, EZFS_BADVERSION, desc);
1837 break;
1838
1839 case EINVAL:
1840 (void) zfs_error(hdl, EZFS_INVALCONFIG, desc);
1841 break;
1842
1843 case EROFS:
1844 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1845 "one or more devices is read only"));
1846 (void) zfs_error(hdl, EZFS_BADDEV, desc);
1847 break;
1848
1849 case ENXIO:
1850 if (nv && nvlist_lookup_nvlist(nv,
1851 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
1852 nvlist_lookup_nvlist(nvinfo,
1853 ZPOOL_CONFIG_MISSING_DEVICES, &missing) == 0) {
1854 (void) printf(dgettext(TEXT_DOMAIN,
1855 "The devices below are missing or "
1856 "corrupted, use '-m' to import the pool "
1857 "anyway:\n"));
1858 print_vdev_tree(hdl, NULL, missing, 2);
1859 (void) printf("\n");
1860 }
1861 (void) zpool_standard_error(hdl, error, desc);
1862 break;
1863
1864 case EEXIST:
1865 (void) zpool_standard_error(hdl, error, desc);
1866 break;
1867 case ENAMETOOLONG:
1868 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1869 "new name of at least one dataset is longer than "
1870 "the maximum allowable length"));
1871 (void) zfs_error(hdl, EZFS_NAMETOOLONG, desc);
1872 break;
1873 default:
1874 (void) zpool_standard_error(hdl, error, desc);
1875 zpool_explain_recover(hdl,
1876 newname ? origname : thename, -error, nv);
1877 break;
1878 }
1879
1880 nvlist_free(nv);
1881 ret = -1;
1882 } else {
1883 zpool_handle_t *zhp;
1884
1885 /*
1886 * This should never fail, but play it safe anyway.
1887 */
1888 if (zpool_open_silent(hdl, thename, &zhp) != 0)
1889 ret = -1;
1890 else if (zhp != NULL)
1891 zpool_close(zhp);
1892 if (policy.zlp_rewind &
1893 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
1894 zpool_rewind_exclaim(hdl, newname ? origname : thename,
1895 ((policy.zlp_rewind & ZPOOL_TRY_REWIND) != 0), nv);
1896 }
1897 nvlist_free(nv);
1898 return (0);
1899 }
1900
1901 return (ret);
1902 }
1903
1904 /*
1905 * Scan the pool.
1906 */
1907 int
1908 zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func, pool_scrub_cmd_t cmd)
1909 {
1910 zfs_cmd_t zc = { 0 };
1911 char msg[1024];
1912 int err;
1913 libzfs_handle_t *hdl = zhp->zpool_hdl;
1914
1915 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1916 zc.zc_cookie = func;
1917 zc.zc_flags = cmd;
1918
1919 if (zfs_ioctl(hdl, ZFS_IOC_POOL_SCAN, &zc) == 0)
1920 return (0);
1921
1922 err = errno;
1923
1924 /* ECANCELED on a scrub means we resumed a paused scrub */
1925 if (err == ECANCELED && func == POOL_SCAN_SCRUB &&
1926 cmd == POOL_SCRUB_NORMAL)
1927 return (0);
1928
1929 if (err == ENOENT && func != POOL_SCAN_NONE && cmd == POOL_SCRUB_NORMAL)
1930 return (0);
1931
1932 if (func == POOL_SCAN_SCRUB) {
1933 if (cmd == POOL_SCRUB_PAUSE) {
1934 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1935 "cannot pause scrubbing %s"), zc.zc_name);
1936 } else {
1937 assert(cmd == POOL_SCRUB_NORMAL);
1938 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1939 "cannot scrub %s"), zc.zc_name);
1940 }
1941 } else if (func == POOL_SCAN_NONE) {
1942 (void) snprintf(msg, sizeof (msg),
1943 dgettext(TEXT_DOMAIN, "cannot cancel scrubbing %s"),
1944 zc.zc_name);
1945 } else {
1946 assert(!"unexpected result");
1947 }
1948
1949 if (err == EBUSY) {
1950 nvlist_t *nvroot;
1951 pool_scan_stat_t *ps = NULL;
1952 uint_t psc;
1953
1954 verify(nvlist_lookup_nvlist(zhp->zpool_config,
1955 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
1956 (void) nvlist_lookup_uint64_array(nvroot,
1957 ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc);
1958 if (ps && ps->pss_func == POOL_SCAN_SCRUB) {
1959 if (cmd == POOL_SCRUB_PAUSE)
1960 return (zfs_error(hdl, EZFS_SCRUB_PAUSED, msg));
1961 else
1962 return (zfs_error(hdl, EZFS_SCRUBBING, msg));
1963 } else {
1964 return (zfs_error(hdl, EZFS_RESILVERING, msg));
1965 }
1966 } else if (err == ENOENT) {
1967 return (zfs_error(hdl, EZFS_NO_SCRUB, msg));
1968 } else {
1969 return (zpool_standard_error(hdl, err, msg));
1970 }
1971 }
1972
1973 static int
1974 xlate_init_err(int err)
1975 {
1976 switch (err) {
1977 case ENODEV:
1978 return (EZFS_NODEVICE);
1979 case EINVAL:
1980 case EROFS:
1981 return (EZFS_BADDEV);
1982 case EBUSY:
1983 return (EZFS_INITIALIZING);
1984 case ESRCH:
1985 return (EZFS_NO_INITIALIZE);
1986 }
1987 return (err);
1988 }
1989
1990 /*
1991 * Begin, suspend, or cancel the initialization (initializing of all free
1992 * blocks) for the given vdevs in the given pool.
1993 */
1994 int
1995 zpool_initialize(zpool_handle_t *zhp, pool_initialize_func_t cmd_type,
1996 nvlist_t *vds)
1997 {
1998 char msg[1024];
1999 libzfs_handle_t *hdl = zhp->zpool_hdl;
2000
2001 nvlist_t *errlist;
2002
2003 /* translate vdev names to guids */
2004 nvlist_t *vdev_guids = fnvlist_alloc();
2005 nvlist_t *guids_to_paths = fnvlist_alloc();
2006 boolean_t spare, cache;
2007 nvlist_t *tgt;
2008 nvpair_t *elem;
2009
2010 for (elem = nvlist_next_nvpair(vds, NULL); elem != NULL;
2011 elem = nvlist_next_nvpair(vds, elem)) {
2012 char *vd_path = nvpair_name(elem);
2013 tgt = zpool_find_vdev(zhp, vd_path, &spare, &cache, NULL);
2014
2015 if ((tgt == NULL) || cache || spare) {
2016 (void) snprintf(msg, sizeof (msg),
2017 dgettext(TEXT_DOMAIN, "cannot initialize '%s'"),
2018 vd_path);
2019 int err = (tgt == NULL) ? EZFS_NODEVICE :
2020 (spare ? EZFS_ISSPARE : EZFS_ISL2CACHE);
2021 fnvlist_free(vdev_guids);
2022 fnvlist_free(guids_to_paths);
2023 return (zfs_error(hdl, err, msg));
2024 }
2025
2026 uint64_t guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
2027 fnvlist_add_uint64(vdev_guids, vd_path, guid);
2028
2029 (void) snprintf(msg, sizeof (msg), "%llu", guid);
2030 fnvlist_add_string(guids_to_paths, msg, vd_path);
2031 }
2032
2033 int err = lzc_initialize(zhp->zpool_name, cmd_type, vdev_guids,
2034 &errlist);
2035 fnvlist_free(vdev_guids);
2036
2037 if (err == 0) {
2038 fnvlist_free(guids_to_paths);
2039 return (0);
2040 }
2041
2042 nvlist_t *vd_errlist = NULL;
2043 if (errlist != NULL) {
2044 vd_errlist = fnvlist_lookup_nvlist(errlist,
2045 ZPOOL_INITIALIZE_VDEVS);
2046 }
2047
2048 (void) snprintf(msg, sizeof (msg),
2049 dgettext(TEXT_DOMAIN, "operation failed"));
2050
2051 for (elem = nvlist_next_nvpair(vd_errlist, NULL); elem != NULL;
2052 elem = nvlist_next_nvpair(vd_errlist, elem)) {
2053 int64_t vd_error = xlate_init_err(fnvpair_value_int64(elem));
2054 char *path = fnvlist_lookup_string(guids_to_paths,
2055 nvpair_name(elem));
2056 (void) zfs_error_fmt(hdl, vd_error, "cannot initialize '%s'",
2057 path);
2058 }
2059
2060 fnvlist_free(guids_to_paths);
2061 if (vd_errlist != NULL)
2062 return (-1);
2063
2064 return (zpool_standard_error(hdl, err, msg));
2065 }
2066
2067 /*
2068 * This provides a very minimal check whether a given string is likely a
2069 * c#t#d# style string. Users of this are expected to do their own
2070 * verification of the s# part.
2071 */
2072 #define CTD_CHECK(str) (str && str[0] == 'c' && isdigit(str[1]))
2073
2074 /*
2075 * More elaborate version for ones which may start with "/dev/dsk/"
2076 * and the like.
2077 */
2078 static int
2079 ctd_check_path(char *str)
2080 {
2081 /*
2082 * If it starts with a slash, check the last component.
2083 */
2084 if (str && str[0] == '/') {
2085 char *tmp = strrchr(str, '/');
2086
2087 /*
2088 * If it ends in "/old", check the second-to-last
2089 * component of the string instead.
2090 */
2091 if (tmp != str && strcmp(tmp, "/old") == 0) {
2092 for (tmp--; *tmp != '/'; tmp--)
2093 ;
2094 }
2095 str = tmp + 1;
2096 }
2097 return (CTD_CHECK(str));
2098 }
2099
2100 /*
2101 * Find a vdev that matches the search criteria specified. We use the
2102 * the nvpair name to determine how we should look for the device.
2103 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
2104 * spare; but FALSE if its an INUSE spare.
2105 */
2106 static nvlist_t *
2107 vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare,
2108 boolean_t *l2cache, boolean_t *log)
2109 {
2110 uint_t c, children;
2111 nvlist_t **child;
2112 nvlist_t *ret;
2113 uint64_t is_log;
2114 char *srchkey;
2115 nvpair_t *pair = nvlist_next_nvpair(search, NULL);
2116
2117 /* Nothing to look for */
2118 if (search == NULL || pair == NULL)
2119 return (NULL);
2120
2121 /* Obtain the key we will use to search */
2122 srchkey = nvpair_name(pair);
2123
2124 switch (nvpair_type(pair)) {
2125 case DATA_TYPE_UINT64:
2126 if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) {
2127 uint64_t srchval, theguid;
2128
2129 verify(nvpair_value_uint64(pair, &srchval) == 0);
2130 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
2131 &theguid) == 0);
2132 if (theguid == srchval)
2133 return (nv);
2134 }
2135 break;
2136
2137 case DATA_TYPE_STRING: {
2138 char *srchval, *val;
2139
2140 verify(nvpair_value_string(pair, &srchval) == 0);
2141 if (nvlist_lookup_string(nv, srchkey, &val) != 0)
2142 break;
2143
2144 /*
2145 * Search for the requested value. Special cases:
2146 *
2147 * - ZPOOL_CONFIG_PATH for whole disk entries. To support
2148 * UEFI boot, these end in "s0" or "s0/old" or "s1" or
2149 * "s1/old". The "s0" or "s1" part is hidden from the user,
2150 * but included in the string, so this matches around it.
2151 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE).
2152 *
2153 * Otherwise, all other searches are simple string compares.
2154 */
2155 if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0 &&
2156 ctd_check_path(val)) {
2157 uint64_t wholedisk = 0;
2158
2159 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
2160 &wholedisk);
2161 if (wholedisk) {
2162 int slen = strlen(srchval);
2163 int vlen = strlen(val);
2164
2165 if (slen != vlen - 2)
2166 break;
2167
2168 /*
2169 * make_leaf_vdev() should only set
2170 * wholedisk for ZPOOL_CONFIG_PATHs which
2171 * will include "/dev/dsk/", giving plenty of
2172 * room for the indices used next.
2173 */
2174 ASSERT(vlen >= 6);
2175
2176 /*
2177 * strings identical except trailing "s0"
2178 */
2179 if ((strcmp(&val[vlen - 2], "s0") == 0 ||
2180 strcmp(&val[vlen - 2], "s1") == 0) &&
2181 strncmp(srchval, val, slen) == 0)
2182 return (nv);
2183
2184 /*
2185 * strings identical except trailing "s0/old"
2186 */
2187 if ((strcmp(&val[vlen - 6], "s0/old") == 0 ||
2188 strcmp(&val[vlen - 6], "s1/old") == 0) &&
2189 strcmp(&srchval[slen - 4], "/old") == 0 &&
2190 strncmp(srchval, val, slen - 4) == 0)
2191 return (nv);
2192
2193 break;
2194 }
2195 } else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0 && val) {
2196 char *type, *idx, *end, *p;
2197 uint64_t id, vdev_id;
2198
2199 /*
2200 * Determine our vdev type, keeping in mind
2201 * that the srchval is composed of a type and
2202 * vdev id pair (i.e. mirror-4).
2203 */
2204 if ((type = strdup(srchval)) == NULL)
2205 return (NULL);
2206
2207 if ((p = strrchr(type, '-')) == NULL) {
2208 free(type);
2209 break;
2210 }
2211 idx = p + 1;
2212 *p = '\0';
2213
2214 /*
2215 * If the types don't match then keep looking.
2216 */
2217 if (strncmp(val, type, strlen(val)) != 0) {
2218 free(type);
2219 break;
2220 }
2221
2222 verify(zpool_vdev_is_interior(type));
2223 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
2224 &id) == 0);
2225
2226 errno = 0;
2227 vdev_id = strtoull(idx, &end, 10);
2228
2229 free(type);
2230 if (errno != 0)
2231 return (NULL);
2232
2233 /*
2234 * Now verify that we have the correct vdev id.
2235 */
2236 if (vdev_id == id)
2237 return (nv);
2238 }
2239
2240 /*
2241 * Common case
2242 */
2243 if (strcmp(srchval, val) == 0)
2244 return (nv);
2245 break;
2246 }
2247
2248 default:
2249 break;
2250 }
2251
2252 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
2253 &child, &children) != 0)
2254 return (NULL);
2255
2256 for (c = 0; c < children; c++) {
2257 if ((ret = vdev_to_nvlist_iter(child[c], search,
2258 avail_spare, l2cache, NULL)) != NULL) {
2259 /*
2260 * The 'is_log' value is only set for the toplevel
2261 * vdev, not the leaf vdevs. So we always lookup the
2262 * log device from the root of the vdev tree (where
2263 * 'log' is non-NULL).
2264 */
2265 if (log != NULL &&
2266 nvlist_lookup_uint64(child[c],
2267 ZPOOL_CONFIG_IS_LOG, &is_log) == 0 &&
2268 is_log) {
2269 *log = B_TRUE;
2270 }
2271 return (ret);
2272 }
2273 }
2274
2275 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
2276 &child, &children) == 0) {
2277 for (c = 0; c < children; c++) {
2278 if ((ret = vdev_to_nvlist_iter(child[c], search,
2279 avail_spare, l2cache, NULL)) != NULL) {
2280 *avail_spare = B_TRUE;
2281 return (ret);
2282 }
2283 }
2284 }
2285
2286 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
2287 &child, &children) == 0) {
2288 for (c = 0; c < children; c++) {
2289 if ((ret = vdev_to_nvlist_iter(child[c], search,
2290 avail_spare, l2cache, NULL)) != NULL) {
2291 *l2cache = B_TRUE;
2292 return (ret);
2293 }
2294 }
2295 }
2296
2297 return (NULL);
2298 }
2299
2300 /*
2301 * Given a physical path (minus the "/devices" prefix), find the
2302 * associated vdev.
2303 */
2304 nvlist_t *
2305 zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath,
2306 boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log)
2307 {
2308 nvlist_t *search, *nvroot, *ret;
2309
2310 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2311 verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath) == 0);
2312
2313 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
2314 &nvroot) == 0);
2315
2316 *avail_spare = B_FALSE;
2317 *l2cache = B_FALSE;
2318 if (log != NULL)
2319 *log = B_FALSE;
2320 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
2321 nvlist_free(search);
2322
2323 return (ret);
2324 }
2325
2326 /*
2327 * Determine if we have an "interior" top-level vdev (i.e mirror/raidz).
2328 */
2329 static boolean_t
2330 zpool_vdev_is_interior(const char *name)
2331 {
2332 if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 ||
2333 strncmp(name, VDEV_TYPE_SPARE, strlen(VDEV_TYPE_SPARE)) == 0 ||
2334 strncmp(name,
2335 VDEV_TYPE_REPLACING, strlen(VDEV_TYPE_REPLACING)) == 0 ||
2336 strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0)
2337 return (B_TRUE);
2338 return (B_FALSE);
2339 }
2340
2341 nvlist_t *
2342 zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare,
2343 boolean_t *l2cache, boolean_t *log)
2344 {
2345 char buf[MAXPATHLEN];
2346 char *end;
2347 nvlist_t *nvroot, *search, *ret;
2348 uint64_t guid;
2349
2350 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2351
2352 guid = strtoull(path, &end, 10);
2353 if (guid != 0 && *end == '\0') {
2354 verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0);
2355 } else if (zpool_vdev_is_interior(path)) {
2356 verify(nvlist_add_string(search, ZPOOL_CONFIG_TYPE, path) == 0);
2357 } else if (path[0] != '/') {
2358 (void) snprintf(buf, sizeof (buf), "%s/%s", ZFS_DISK_ROOT,
2359 path);
2360 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, buf) == 0);
2361 } else {
2362 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0);
2363 }
2364
2365 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
2366 &nvroot) == 0);
2367
2368 *avail_spare = B_FALSE;
2369 *l2cache = B_FALSE;
2370 if (log != NULL)
2371 *log = B_FALSE;
2372 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
2373 nvlist_free(search);
2374
2375 return (ret);
2376 }
2377
2378 static int
2379 vdev_online(nvlist_t *nv)
2380 {
2381 uint64_t ival;
2382
2383 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 ||
2384 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 ||
2385 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0)
2386 return (0);
2387
2388 return (1);
2389 }
2390
2391 /*
2392 * Helper function for zpool_get_physpaths().
2393 */
2394 static int
2395 vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size,
2396 size_t *bytes_written)
2397 {
2398 size_t bytes_left, pos, rsz;
2399 char *tmppath;
2400 const char *format;
2401
2402 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH,
2403 &tmppath) != 0)
2404 return (EZFS_NODEVICE);
2405
2406 pos = *bytes_written;
2407 bytes_left = physpath_size - pos;
2408 format = (pos == 0) ? "%s" : " %s";
2409
2410 rsz = snprintf(physpath + pos, bytes_left, format, tmppath);
2411 *bytes_written += rsz;
2412
2413 if (rsz >= bytes_left) {
2414 /* if physpath was not copied properly, clear it */
2415 if (bytes_left != 0) {
2416 physpath[pos] = 0;
2417 }
2418 return (EZFS_NOSPC);
2419 }
2420 return (0);
2421 }
2422
2423 static int
2424 vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size,
2425 size_t *rsz, boolean_t is_spare)
2426 {
2427 char *type;
2428 int ret;
2429
2430 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0)
2431 return (EZFS_INVALCONFIG);
2432
2433 if (strcmp(type, VDEV_TYPE_DISK) == 0) {
2434 /*
2435 * An active spare device has ZPOOL_CONFIG_IS_SPARE set.
2436 * For a spare vdev, we only want to boot from the active
2437 * spare device.
2438 */
2439 if (is_spare) {
2440 uint64_t spare = 0;
2441 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE,
2442 &spare);
2443 if (!spare)
2444 return (EZFS_INVALCONFIG);
2445 }
2446
2447 if (vdev_online(nv)) {
2448 if ((ret = vdev_get_one_physpath(nv, physpath,
2449 phypath_size, rsz)) != 0)
2450 return (ret);
2451 }
2452 } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 ||
2453 strcmp(type, VDEV_TYPE_RAIDZ) == 0 ||
2454 strcmp(type, VDEV_TYPE_REPLACING) == 0 ||
2455 (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) {
2456 nvlist_t **child;
2457 uint_t count;
2458 int i, ret;
2459
2460 if (nvlist_lookup_nvlist_array(nv,
2461 ZPOOL_CONFIG_CHILDREN, &child, &count) != 0)
2462 return (EZFS_INVALCONFIG);
2463
2464 for (i = 0; i < count; i++) {
2465 ret = vdev_get_physpaths(child[i], physpath,
2466 phypath_size, rsz, is_spare);
2467 if (ret == EZFS_NOSPC)
2468 return (ret);
2469 }
2470 }
2471
2472 return (EZFS_POOL_INVALARG);
2473 }
2474
2475 /*
2476 * Get phys_path for a root pool config.
2477 * Return 0 on success; non-zero on failure.
2478 */
2479 static int
2480 zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size)
2481 {
2482 size_t rsz;
2483 nvlist_t *vdev_root;
2484 nvlist_t **child;
2485 uint_t count;
2486 char *type;
2487
2488 rsz = 0;
2489
2490 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
2491 &vdev_root) != 0)
2492 return (EZFS_INVALCONFIG);
2493
2494 if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 ||
2495 nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN,
2496 &child, &count) != 0)
2497 return (EZFS_INVALCONFIG);
2498
2499 /*
2500 * root pool can only have a single top-level vdev.
2501 */
2502 if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1)
2503 return (EZFS_POOL_INVALARG);
2504
2505 (void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz,
2506 B_FALSE);
2507
2508 /* No online devices */
2509 if (rsz == 0)
2510 return (EZFS_NODEVICE);
2511
2512 return (0);
2513 }
2514
2515 /*
2516 * Get phys_path for a root pool
2517 * Return 0 on success; non-zero on failure.
2518 */
2519 int
2520 zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size)
2521 {
2522 return (zpool_get_config_physpath(zhp->zpool_config, physpath,
2523 phypath_size));
2524 }
2525
2526 /*
2527 * If the device has being dynamically expanded then we need to relabel
2528 * the disk to use the new unallocated space.
2529 */
2530 static int
2531 zpool_relabel_disk(libzfs_handle_t *hdl, const char *name)
2532 {
2533 char path[MAXPATHLEN];
2534 char errbuf[1024];
2535 int fd, error;
2536 int (*_efi_use_whole_disk)(int);
2537
2538 if ((_efi_use_whole_disk = (int (*)(int))dlsym(RTLD_DEFAULT,
2539 "efi_use_whole_disk")) == NULL)
2540 return (-1);
2541
2542 (void) snprintf(path, sizeof (path), "%s/%s", ZFS_RDISK_ROOT, name);
2543
2544 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) {
2545 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2546 "relabel '%s': unable to open device"), name);
2547 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
2548 }
2549
2550 /*
2551 * It's possible that we might encounter an error if the device
2552 * does not have any unallocated space left. If so, we simply
2553 * ignore that error and continue on.
2554 */
2555 error = _efi_use_whole_disk(fd);
2556 (void) close(fd);
2557 if (error && error != VT_ENOSPC) {
2558 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2559 "relabel '%s': unable to read disk capacity"), name);
2560 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
2561 }
2562 return (0);
2563 }
2564
2565 /*
2566 * Bring the specified vdev online. The 'flags' parameter is a set of the
2567 * ZFS_ONLINE_* flags.
2568 */
2569 int
2570 zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags,
2571 vdev_state_t *newstate)
2572 {
2573 zfs_cmd_t zc = { 0 };
2574 char msg[1024];
2575 char *pathname;
2576 nvlist_t *tgt;
2577 boolean_t avail_spare, l2cache, islog;
2578 libzfs_handle_t *hdl = zhp->zpool_hdl;
2579
2580 if (flags & ZFS_ONLINE_EXPAND) {
2581 (void) snprintf(msg, sizeof (msg),
2582 dgettext(TEXT_DOMAIN, "cannot expand %s"), path);
2583 } else {
2584 (void) snprintf(msg, sizeof (msg),
2585 dgettext(TEXT_DOMAIN, "cannot online %s"), path);
2586 }
2587
2588 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2589 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2590 &islog)) == NULL)
2591 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2592
2593 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2594
2595 if (avail_spare)
2596 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2597
2598 if ((flags & ZFS_ONLINE_EXPAND ||
2599 zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) &&
2600 nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH, &pathname) == 0) {
2601 uint64_t wholedisk = 0;
2602
2603 (void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK,
2604 &wholedisk);
2605
2606 /*
2607 * XXX - L2ARC 1.0 devices can't support expansion.
2608 */
2609 if (l2cache) {
2610 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2611 "cannot expand cache devices"));
2612 return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg));
2613 }
2614
2615 if (wholedisk) {
2616 pathname += strlen(ZFS_DISK_ROOT) + 1;
2617 (void) zpool_relabel_disk(hdl, pathname);
2618 }
2619 }
2620
2621 zc.zc_cookie = VDEV_STATE_ONLINE;
2622 zc.zc_obj = flags;
2623
2624 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) {
2625 if (errno == EINVAL) {
2626 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split "
2627 "from this pool into a new one. Use '%s' "
2628 "instead"), "zpool detach");
2629 return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, msg));
2630 }
2631 return (zpool_standard_error(hdl, errno, msg));
2632 }
2633
2634 *newstate = zc.zc_cookie;
2635 return (0);
2636 }
2637
2638 /*
2639 * Take the specified vdev offline
2640 */
2641 int
2642 zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp)
2643 {
2644 zfs_cmd_t zc = { 0 };
2645 char msg[1024];
2646 nvlist_t *tgt;
2647 boolean_t avail_spare, l2cache;
2648 libzfs_handle_t *hdl = zhp->zpool_hdl;
2649
2650 (void) snprintf(msg, sizeof (msg),
2651 dgettext(TEXT_DOMAIN, "cannot offline %s"), path);
2652
2653 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2654 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2655 NULL)) == NULL)
2656 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2657
2658 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2659
2660 if (avail_spare)
2661 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2662
2663 zc.zc_cookie = VDEV_STATE_OFFLINE;
2664 zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0;
2665
2666 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2667 return (0);
2668
2669 switch (errno) {
2670 case EBUSY:
2671
2672 /*
2673 * There are no other replicas of this device.
2674 */
2675 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2676
2677 case EEXIST:
2678 /*
2679 * The log device has unplayed logs
2680 */
2681 return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg));
2682
2683 default:
2684 return (zpool_standard_error(hdl, errno, msg));
2685 }
2686 }
2687
2688 /*
2689 * Mark the given vdev faulted.
2690 */
2691 int
2692 zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2693 {
2694 zfs_cmd_t zc = { 0 };
2695 char msg[1024];
2696 libzfs_handle_t *hdl = zhp->zpool_hdl;
2697
2698 (void) snprintf(msg, sizeof (msg),
2699 dgettext(TEXT_DOMAIN, "cannot fault %llu"), guid);
2700
2701 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2702 zc.zc_guid = guid;
2703 zc.zc_cookie = VDEV_STATE_FAULTED;
2704 zc.zc_obj = aux;
2705
2706 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2707 return (0);
2708
2709 switch (errno) {
2710 case EBUSY:
2711
2712 /*
2713 * There are no other replicas of this device.
2714 */
2715 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2716
2717 default:
2718 return (zpool_standard_error(hdl, errno, msg));
2719 }
2720
2721 }
2722
2723 /*
2724 * Mark the given vdev degraded.
2725 */
2726 int
2727 zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2728 {
2729 zfs_cmd_t zc = { 0 };
2730 char msg[1024];
2731 libzfs_handle_t *hdl = zhp->zpool_hdl;
2732
2733 (void) snprintf(msg, sizeof (msg),
2734 dgettext(TEXT_DOMAIN, "cannot degrade %llu"), guid);
2735
2736 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2737 zc.zc_guid = guid;
2738 zc.zc_cookie = VDEV_STATE_DEGRADED;
2739 zc.zc_obj = aux;
2740
2741 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2742 return (0);
2743
2744 return (zpool_standard_error(hdl, errno, msg));
2745 }
2746
2747 /*
2748 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as
2749 * a hot spare.
2750 */
2751 static boolean_t
2752 is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which)
2753 {
2754 nvlist_t **child;
2755 uint_t c, children;
2756 char *type;
2757
2758 if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child,
2759 &children) == 0) {
2760 verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE,
2761 &type) == 0);
2762
2763 if (strcmp(type, VDEV_TYPE_SPARE) == 0 &&
2764 children == 2 && child[which] == tgt)
2765 return (B_TRUE);
2766
2767 for (c = 0; c < children; c++)
2768 if (is_replacing_spare(child[c], tgt, which))
2769 return (B_TRUE);
2770 }
2771
2772 return (B_FALSE);
2773 }
2774
2775 /*
2776 * Attach new_disk (fully described by nvroot) to old_disk.
2777 * If 'replacing' is specified, the new disk will replace the old one.
2778 */
2779 int
2780 zpool_vdev_attach(zpool_handle_t *zhp,
2781 const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing)
2782 {
2783 zfs_cmd_t zc = { 0 };
2784 char msg[1024];
2785 int ret;
2786 nvlist_t *tgt;
2787 boolean_t avail_spare, l2cache, islog;
2788 uint64_t val;
2789 char *newname;
2790 nvlist_t **child;
2791 uint_t children;
2792 nvlist_t *config_root;
2793 libzfs_handle_t *hdl = zhp->zpool_hdl;
2794 boolean_t rootpool = zpool_is_bootable(zhp);
2795
2796 if (replacing)
2797 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2798 "cannot replace %s with %s"), old_disk, new_disk);
2799 else
2800 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2801 "cannot attach %s to %s"), new_disk, old_disk);
2802
2803 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2804 if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache,
2805 &islog)) == NULL)
2806 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2807
2808 if (avail_spare)
2809 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2810
2811 if (l2cache)
2812 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2813
2814 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2815 zc.zc_cookie = replacing;
2816
2817 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
2818 &child, &children) != 0 || children != 1) {
2819 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2820 "new device must be a single disk"));
2821 return (zfs_error(hdl, EZFS_INVALCONFIG, msg));
2822 }
2823
2824 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
2825 ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0);
2826
2827 if ((newname = zpool_vdev_name(NULL, NULL, child[0], B_FALSE)) == NULL)
2828 return (-1);
2829
2830 /*
2831 * If the target is a hot spare that has been swapped in, we can only
2832 * replace it with another hot spare.
2833 */
2834 if (replacing &&
2835 nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 &&
2836 (zpool_find_vdev(zhp, newname, &avail_spare, &l2cache,
2837 NULL) == NULL || !avail_spare) &&
2838 is_replacing_spare(config_root, tgt, 1)) {
2839 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2840 "can only be replaced by another hot spare"));
2841 free(newname);
2842 return (zfs_error(hdl, EZFS_BADTARGET, msg));
2843 }
2844
2845 free(newname);
2846
2847 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
2848 return (-1);
2849
2850 ret = zfs_ioctl(hdl, ZFS_IOC_VDEV_ATTACH, &zc);
2851
2852 zcmd_free_nvlists(&zc);
2853
2854 if (ret == 0) {
2855 if (rootpool) {
2856 /*
2857 * XXX need a better way to prevent user from
2858 * booting up a half-baked vdev.
2859 */
2860 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make "
2861 "sure to wait until resilver is done "
2862 "before rebooting.\n"));
2863 }
2864 return (0);
2865 }
2866
2867 switch (errno) {
2868 case ENOTSUP:
2869 /*
2870 * Can't attach to or replace this type of vdev.
2871 */
2872 if (replacing) {
2873 uint64_t version = zpool_get_prop_int(zhp,
2874 ZPOOL_PROP_VERSION, NULL);
2875
2876 if (islog)
2877 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2878 "cannot replace a log with a spare"));
2879 else if (version >= SPA_VERSION_MULTI_REPLACE)
2880 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2881 "already in replacing/spare config; wait "
2882 "for completion or use 'zpool detach'"));
2883 else
2884 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2885 "cannot replace a replacing device"));
2886 } else {
2887 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2888 "can only attach to mirrors and top-level "
2889 "disks"));
2890 }
2891 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
2892 break;
2893
2894 case EINVAL:
2895 /*
2896 * The new device must be a single disk.
2897 */
2898 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2899 "new device must be a single disk"));
2900 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg);
2901 break;
2902
2903 case EBUSY:
2904 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy, "
2905 "or device removal is in progress"),
2906 new_disk);
2907 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2908 break;
2909
2910 case EOVERFLOW:
2911 /*
2912 * The new device is too small.
2913 */
2914 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2915 "device is too small"));
2916 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2917 break;
2918
2919 case EDOM:
2920 /*
2921 * The new device has a different alignment requirement.
2922 */
2923 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2924 "devices have different sector alignment"));
2925 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2926 break;
2927
2928 case ENAMETOOLONG:
2929 /*
2930 * The resulting top-level vdev spec won't fit in the label.
2931 */
2932 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg);
2933 break;
2934
2935 default:
2936 (void) zpool_standard_error(hdl, errno, msg);
2937 }
2938
2939 return (-1);
2940 }
2941
2942 /*
2943 * Detach the specified device.
2944 */
2945 int
2946 zpool_vdev_detach(zpool_handle_t *zhp, const char *path)
2947 {
2948 zfs_cmd_t zc = { 0 };
2949 char msg[1024];
2950 nvlist_t *tgt;
2951 boolean_t avail_spare, l2cache;
2952 libzfs_handle_t *hdl = zhp->zpool_hdl;
2953
2954 (void) snprintf(msg, sizeof (msg),
2955 dgettext(TEXT_DOMAIN, "cannot detach %s"), path);
2956
2957 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2958 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2959 NULL)) == NULL)
2960 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2961
2962 if (avail_spare)
2963 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2964
2965 if (l2cache)
2966 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2967
2968 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2969
2970 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0)
2971 return (0);
2972
2973 switch (errno) {
2974
2975 case ENOTSUP:
2976 /*
2977 * Can't detach from this type of vdev.
2978 */
2979 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only "
2980 "applicable to mirror and replacing vdevs"));
2981 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
2982 break;
2983
2984 case EBUSY:
2985 /*
2986 * There are no other replicas of this device.
2987 */
2988 (void) zfs_error(hdl, EZFS_NOREPLICAS, msg);
2989 break;
2990
2991 default:
2992 (void) zpool_standard_error(hdl, errno, msg);
2993 }
2994
2995 return (-1);
2996 }
2997
2998 /*
2999 * Find a mirror vdev in the source nvlist.
3000 *
3001 * The mchild array contains a list of disks in one of the top-level mirrors
3002 * of the source pool. The schild array contains a list of disks that the
3003 * user specified on the command line. We loop over the mchild array to
3004 * see if any entry in the schild array matches.
3005 *
3006 * If a disk in the mchild array is found in the schild array, we return
3007 * the index of that entry. Otherwise we return -1.
3008 */
3009 static int
3010 find_vdev_entry(zpool_handle_t *zhp, nvlist_t **mchild, uint_t mchildren,
3011 nvlist_t **schild, uint_t schildren)
3012 {
3013 uint_t mc;
3014
3015 for (mc = 0; mc < mchildren; mc++) {
3016 uint_t sc;
3017 char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp,
3018 mchild[mc], B_FALSE);
3019
3020 for (sc = 0; sc < schildren; sc++) {
3021 char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp,
3022 schild[sc], B_FALSE);
3023 boolean_t result = (strcmp(mpath, spath) == 0);
3024
3025 free(spath);
3026 if (result) {
3027 free(mpath);
3028 return (mc);
3029 }
3030 }
3031
3032 free(mpath);
3033 }
3034
3035 return (-1);
3036 }
3037
3038 /*
3039 * Split a mirror pool. If newroot points to null, then a new nvlist
3040 * is generated and it is the responsibility of the caller to free it.
3041 */
3042 int
3043 zpool_vdev_split(zpool_handle_t *zhp, char *newname, nvlist_t **newroot,
3044 nvlist_t *props, splitflags_t flags)
3045 {
3046 zfs_cmd_t zc = { 0 };
3047 char msg[1024];
3048 nvlist_t *tree, *config, **child, **newchild, *newconfig = NULL;
3049 nvlist_t **varray = NULL, *zc_props = NULL;
3050 uint_t c, children, newchildren, lastlog = 0, vcount, found = 0;
3051 libzfs_handle_t *hdl = zhp->zpool_hdl;
3052 uint64_t vers;
3053 boolean_t freelist = B_FALSE, memory_err = B_TRUE;
3054 int retval = 0;
3055
3056 (void) snprintf(msg, sizeof (msg),
3057 dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name);
3058
3059 if (!zpool_name_valid(hdl, B_FALSE, newname))
3060 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
3061
3062 if ((config = zpool_get_config(zhp, NULL)) == NULL) {
3063 (void) fprintf(stderr, gettext("Internal error: unable to "
3064 "retrieve pool configuration\n"));
3065 return (-1);
3066 }
3067
3068 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree)
3069 == 0);
3070 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &vers) == 0);
3071
3072 if (props) {
3073 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
3074 if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name,
3075 props, vers, flags, msg)) == NULL)
3076 return (-1);
3077 }
3078
3079 if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child,
3080 &children) != 0) {
3081 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3082 "Source pool is missing vdev tree"));
3083 nvlist_free(zc_props);
3084 return (-1);
3085 }
3086
3087 varray = zfs_alloc(hdl, children * sizeof (nvlist_t *));
3088 vcount = 0;
3089
3090 if (*newroot == NULL ||
3091 nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN,
3092 &newchild, &newchildren) != 0)
3093 newchildren = 0;
3094
3095 for (c = 0; c < children; c++) {
3096 uint64_t is_log = B_FALSE, is_hole = B_FALSE;
3097 char *type;
3098 nvlist_t **mchild, *vdev;
3099 uint_t mchildren;
3100 int entry;
3101
3102 /*
3103 * Unlike cache & spares, slogs are stored in the
3104 * ZPOOL_CONFIG_CHILDREN array. We filter them out here.
3105 */
3106 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
3107 &is_log);
3108 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
3109 &is_hole);
3110 if (is_log || is_hole) {
3111 /*
3112 * Create a hole vdev and put it in the config.
3113 */
3114 if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0)
3115 goto out;
3116 if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE,
3117 VDEV_TYPE_HOLE) != 0)
3118 goto out;
3119 if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE,
3120 1) != 0)
3121 goto out;
3122 if (lastlog == 0)
3123 lastlog = vcount;
3124 varray[vcount++] = vdev;
3125 continue;
3126 }
3127 lastlog = 0;
3128 verify(nvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE, &type)
3129 == 0);
3130 if (strcmp(type, VDEV_TYPE_MIRROR) != 0) {
3131 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3132 "Source pool must be composed only of mirrors\n"));
3133 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
3134 goto out;
3135 }
3136
3137 verify(nvlist_lookup_nvlist_array(child[c],
3138 ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0);
3139
3140 /* find or add an entry for this top-level vdev */
3141 if (newchildren > 0 &&
3142 (entry = find_vdev_entry(zhp, mchild, mchildren,
3143 newchild, newchildren)) >= 0) {
3144 /* We found a disk that the user specified. */
3145 vdev = mchild[entry];
3146 ++found;
3147 } else {
3148 /* User didn't specify a disk for this vdev. */
3149 vdev = mchild[mchildren - 1];
3150 }
3151
3152 if (nvlist_dup(vdev, &varray[vcount++], 0) != 0)
3153 goto out;
3154 }
3155
3156 /* did we find every disk the user specified? */
3157 if (found != newchildren) {
3158 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must "
3159 "include at most one disk from each mirror"));
3160 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
3161 goto out;
3162 }
3163
3164 /* Prepare the nvlist for populating. */
3165 if (*newroot == NULL) {
3166 if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0)
3167 goto out;
3168 freelist = B_TRUE;
3169 if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE,
3170 VDEV_TYPE_ROOT) != 0)
3171 goto out;
3172 } else {
3173 verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0);
3174 }
3175
3176 /* Add all the children we found */
3177 if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, varray,
3178 lastlog == 0 ? vcount : lastlog) != 0)
3179 goto out;
3180
3181 /*
3182 * If we're just doing a dry run, exit now with success.
3183 */
3184 if (flags.dryrun) {
3185 memory_err = B_FALSE;
3186 freelist = B_FALSE;
3187 goto out;
3188 }
3189
3190 /* now build up the config list & call the ioctl */
3191 if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0)
3192 goto out;
3193
3194 if (nvlist_add_nvlist(newconfig,
3195 ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 ||
3196 nvlist_add_string(newconfig,
3197 ZPOOL_CONFIG_POOL_NAME, newname) != 0 ||
3198 nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0)
3199 goto out;
3200
3201 /*
3202 * The new pool is automatically part of the namespace unless we
3203 * explicitly export it.
3204 */
3205 if (!flags.import)
3206 zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT;
3207 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3208 (void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string));
3209 if (zcmd_write_conf_nvlist(hdl, &zc, newconfig) != 0)
3210 goto out;
3211 if (zc_props != NULL && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
3212 goto out;
3213
3214 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) {
3215 retval = zpool_standard_error(hdl, errno, msg);
3216 goto out;
3217 }
3218
3219 freelist = B_FALSE;
3220 memory_err = B_FALSE;
3221
3222 out:
3223 if (varray != NULL) {
3224 int v;
3225
3226 for (v = 0; v < vcount; v++)
3227 nvlist_free(varray[v]);
3228 free(varray);
3229 }
3230 zcmd_free_nvlists(&zc);
3231 nvlist_free(zc_props);
3232 nvlist_free(newconfig);
3233 if (freelist) {
3234 nvlist_free(*newroot);
3235 *newroot = NULL;
3236 }
3237
3238 if (retval != 0)
3239 return (retval);
3240
3241 if (memory_err)
3242 return (no_memory(hdl));
3243
3244 return (0);
3245 }
3246
3247 /*
3248 * Remove the given device.
3249 */
3250 int
3251 zpool_vdev_remove(zpool_handle_t *zhp, const char *path)
3252 {
3253 zfs_cmd_t zc = { 0 };
3254 char msg[1024];
3255 nvlist_t *tgt;
3256 boolean_t avail_spare, l2cache, islog;
3257 libzfs_handle_t *hdl = zhp->zpool_hdl;
3258 uint64_t version;
3259
3260 (void) snprintf(msg, sizeof (msg),
3261 dgettext(TEXT_DOMAIN, "cannot remove %s"), path);
3262
3263 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3264 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
3265 &islog)) == NULL)
3266 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3267
3268 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
3269 if (islog && version < SPA_VERSION_HOLES) {
3270 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3271 "pool must be upgraded to support log removal"));
3272 return (zfs_error(hdl, EZFS_BADVERSION, msg));
3273 }
3274
3275 if (!islog && !avail_spare && !l2cache && zpool_is_bootable(zhp)) {
3276 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3277 "root pool can not have removed devices, "
3278 "because GRUB does not understand them"));
3279 return (zfs_error(hdl, EINVAL, msg));
3280 }
3281
3282 zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
3283
3284 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
3285 return (0);
3286
3287 switch (errno) {
3288
3289 case EINVAL:
3290 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3291 "invalid config; all top-level vdevs must "
3292 "have the same sector size and not be raidz."));
3293 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg);
3294 break;
3295
3296 case EBUSY:
3297 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3298 "Pool busy; removal may already be in progress"));
3299 (void) zfs_error(hdl, EZFS_BUSY, msg);
3300 break;
3301
3302 default:
3303 (void) zpool_standard_error(hdl, errno, msg);
3304 }
3305 return (-1);
3306 }
3307
3308 int
3309 zpool_vdev_remove_cancel(zpool_handle_t *zhp)
3310 {
3311 zfs_cmd_t zc = { 0 };
3312 char msg[1024];
3313 libzfs_handle_t *hdl = zhp->zpool_hdl;
3314
3315 (void) snprintf(msg, sizeof (msg),
3316 dgettext(TEXT_DOMAIN, "cannot cancel removal"));
3317
3318 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3319 zc.zc_cookie = 1;
3320
3321 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
3322 return (0);
3323
3324 return (zpool_standard_error(hdl, errno, msg));
3325 }
3326
3327 int
3328 zpool_vdev_indirect_size(zpool_handle_t *zhp, const char *path,
3329 uint64_t *sizep)
3330 {
3331 char msg[1024];
3332 nvlist_t *tgt;
3333 boolean_t avail_spare, l2cache, islog;
3334 libzfs_handle_t *hdl = zhp->zpool_hdl;
3335
3336 (void) snprintf(msg, sizeof (msg),
3337 dgettext(TEXT_DOMAIN, "cannot determine indirect size of %s"),
3338 path);
3339
3340 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
3341 &islog)) == NULL)
3342 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3343
3344 if (avail_spare || l2cache || islog) {
3345 *sizep = 0;
3346 return (0);
3347 }
3348
3349 if (nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_INDIRECT_SIZE, sizep) != 0) {
3350 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3351 "indirect size not available"));
3352 return (zfs_error(hdl, EINVAL, msg));
3353 }
3354 return (0);
3355 }
3356
3357 /*
3358 * Clear the errors for the pool, or the particular device if specified.
3359 */
3360 int
3361 zpool_clear(zpool_handle_t *zhp, const char *path, nvlist_t *rewindnvl)
3362 {
3363 zfs_cmd_t zc = { 0 };
3364 char msg[1024];
3365 nvlist_t *tgt;
3366 zpool_load_policy_t policy;
3367 boolean_t avail_spare, l2cache;
3368 libzfs_handle_t *hdl = zhp->zpool_hdl;
3369 nvlist_t *nvi = NULL;
3370 int error;
3371
3372 if (path)
3373 (void) snprintf(msg, sizeof (msg),
3374 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
3375 path);
3376 else
3377 (void) snprintf(msg, sizeof (msg),
3378 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
3379 zhp->zpool_name);
3380
3381 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3382 if (path) {
3383 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare,
3384 &l2cache, NULL)) == NULL)
3385 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3386
3387 /*
3388 * Don't allow error clearing for hot spares. Do allow
3389 * error clearing for l2cache devices.
3390 */
3391 if (avail_spare)
3392 return (zfs_error(hdl, EZFS_ISSPARE, msg));
3393
3394 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID,
3395 &zc.zc_guid) == 0);
3396 }
3397
3398 zpool_get_load_policy(rewindnvl, &policy);
3399 zc.zc_cookie = policy.zlp_rewind;
3400
3401 if (zcmd_alloc_dst_nvlist(hdl, &zc, zhp->zpool_config_size * 2) != 0)
3402 return (-1);
3403
3404 if (zcmd_write_src_nvlist(hdl, &zc, rewindnvl) != 0)
3405 return (-1);
3406
3407 while ((error = zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc)) != 0 &&
3408 errno == ENOMEM) {
3409 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
3410 zcmd_free_nvlists(&zc);
3411 return (-1);
3412 }
3413 }
3414
3415 if (!error || ((policy.zlp_rewind & ZPOOL_TRY_REWIND) &&
3416 errno != EPERM && errno != EACCES)) {
3417 if (policy.zlp_rewind &
3418 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
3419 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi);
3420 zpool_rewind_exclaim(hdl, zc.zc_name,
3421 ((policy.zlp_rewind & ZPOOL_TRY_REWIND) != 0),
3422 nvi);
3423 nvlist_free(nvi);
3424 }
3425 zcmd_free_nvlists(&zc);
3426 return (0);
3427 }
3428
3429 zcmd_free_nvlists(&zc);
3430 return (zpool_standard_error(hdl, errno, msg));
3431 }
3432
3433 /*
3434 * Similar to zpool_clear(), but takes a GUID (used by fmd).
3435 */
3436 int
3437 zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid)
3438 {
3439 zfs_cmd_t zc = { 0 };
3440 char msg[1024];
3441 libzfs_handle_t *hdl = zhp->zpool_hdl;
3442
3443 (void) snprintf(msg, sizeof (msg),
3444 dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"),
3445 guid);
3446
3447 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3448 zc.zc_guid = guid;
3449 zc.zc_cookie = ZPOOL_NO_REWIND;
3450
3451 if (ioctl(hdl->libzfs_fd, ZFS_IOC_CLEAR, &zc) == 0)
3452 return (0);
3453
3454 return (zpool_standard_error(hdl, errno, msg));
3455 }
3456
3457 /*
3458 * Change the GUID for a pool.
3459 */
3460 int
3461 zpool_reguid(zpool_handle_t *zhp)
3462 {
3463 char msg[1024];
3464 libzfs_handle_t *hdl = zhp->zpool_hdl;
3465 zfs_cmd_t zc = { 0 };
3466
3467 (void) snprintf(msg, sizeof (msg),
3468 dgettext(TEXT_DOMAIN, "cannot reguid '%s'"), zhp->zpool_name);
3469
3470 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3471 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REGUID, &zc) == 0)
3472 return (0);
3473
3474 return (zpool_standard_error(hdl, errno, msg));
3475 }
3476
3477 /*
3478 * Reopen the pool.
3479 */
3480 int
3481 zpool_reopen(zpool_handle_t *zhp)
3482 {
3483 zfs_cmd_t zc = { 0 };
3484 char msg[1024];
3485 libzfs_handle_t *hdl = zhp->zpool_hdl;
3486
3487 (void) snprintf(msg, sizeof (msg),
3488 dgettext(TEXT_DOMAIN, "cannot reopen '%s'"),
3489 zhp->zpool_name);
3490
3491 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3492 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REOPEN, &zc) == 0)
3493 return (0);
3494 return (zpool_standard_error(hdl, errno, msg));
3495 }
3496
3497 /*
3498 * Convert from a devid string to a path.
3499 */
3500 static char *
3501 devid_to_path(char *devid_str)
3502 {
3503 ddi_devid_t devid;
3504 char *minor;
3505 char *path;
3506 devid_nmlist_t *list = NULL;
3507 int ret;
3508
3509 if (devid_str_decode(devid_str, &devid, &minor) != 0)
3510 return (NULL);
3511
3512 ret = devid_deviceid_to_nmlist("/dev", devid, minor, &list);
3513
3514 devid_str_free(minor);
3515 devid_free(devid);
3516
3517 if (ret != 0)
3518 return (NULL);
3519
3520 /*
3521 * In a case the strdup() fails, we will just return NULL below.
3522 */
3523 path = strdup(list[0].devname);
3524
3525 devid_free_nmlist(list);
3526
3527 return (path);
3528 }
3529
3530 /*
3531 * Convert from a path to a devid string.
3532 */
3533 static char *
3534 path_to_devid(const char *path)
3535 {
3536 int fd;
3537 ddi_devid_t devid;
3538 char *minor, *ret;
3539
3540 if ((fd = open(path, O_RDONLY)) < 0)
3541 return (NULL);
3542
3543 minor = NULL;
3544 ret = NULL;
3545 if (devid_get(fd, &devid) == 0) {
3546 if (devid_get_minor_name(fd, &minor) == 0)
3547 ret = devid_str_encode(devid, minor);
3548 if (minor != NULL)
3549 devid_str_free(minor);
3550 devid_free(devid);
3551 }
3552 (void) close(fd);
3553
3554 return (ret);
3555 }
3556
3557 /*
3558 * Issue the necessary ioctl() to update the stored path value for the vdev. We
3559 * ignore any failure here, since a common case is for an unprivileged user to
3560 * type 'zpool status', and we'll display the correct information anyway.
3561 */
3562 static void
3563 set_path(zpool_handle_t *zhp, nvlist_t *nv, const char *path)
3564 {
3565 zfs_cmd_t zc = { 0 };
3566
3567 (void) strncpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3568 (void) strncpy(zc.zc_value, path, sizeof (zc.zc_value));
3569 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
3570 &zc.zc_guid) == 0);
3571
3572 (void) ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SETPATH, &zc);
3573 }
3574
3575 /*
3576 * Given a vdev, return the name to display in iostat. If the vdev has a path,
3577 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
3578 * We also check if this is a whole disk, in which case we strip off the
3579 * trailing 's0' slice name.
3580 *
3581 * This routine is also responsible for identifying when disks have been
3582 * reconfigured in a new location. The kernel will have opened the device by
3583 * devid, but the path will still refer to the old location. To catch this, we
3584 * first do a path -> devid translation (which is fast for the common case). If
3585 * the devid matches, we're done. If not, we do a reverse devid -> path
3586 * translation and issue the appropriate ioctl() to update the path of the vdev.
3587 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
3588 * of these checks.
3589 */
3590 char *
3591 zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv,
3592 boolean_t verbose)
3593 {
3594 char *path, *devid;
3595 uint64_t value;
3596 char buf[64];
3597 vdev_stat_t *vs;
3598 uint_t vsc;
3599
3600 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
3601 &value) == 0) {
3602 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
3603 &value) == 0);
3604 (void) snprintf(buf, sizeof (buf), "%llu",
3605 (u_longlong_t)value);
3606 path = buf;
3607 } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) {
3608
3609 /*
3610 * If the device is dead (faulted, offline, etc) then don't
3611 * bother opening it. Otherwise we may be forcing the user to
3612 * open a misbehaving device, which can have undesirable
3613 * effects.
3614 */
3615 if ((nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
3616 (uint64_t **)&vs, &vsc) != 0 ||
3617 vs->vs_state >= VDEV_STATE_DEGRADED) &&
3618 zhp != NULL &&
3619 nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) {
3620 /*
3621 * Determine if the current path is correct.
3622 */
3623 char *newdevid = path_to_devid(path);
3624
3625 if (newdevid == NULL ||
3626 strcmp(devid, newdevid) != 0) {
3627 char *newpath;
3628
3629 if ((newpath = devid_to_path(devid)) != NULL) {
3630 /*
3631 * Update the path appropriately.
3632 */
3633 set_path(zhp, nv, newpath);
3634 if (nvlist_add_string(nv,
3635 ZPOOL_CONFIG_PATH, newpath) == 0)
3636 verify(nvlist_lookup_string(nv,
3637 ZPOOL_CONFIG_PATH,
3638 &path) == 0);
3639 free(newpath);
3640 }
3641 }
3642
3643 if (newdevid)
3644 devid_str_free(newdevid);
3645 }
3646
3647 if (strncmp(path, ZFS_DISK_ROOTD, strlen(ZFS_DISK_ROOTD)) == 0)
3648 path += strlen(ZFS_DISK_ROOTD);
3649
3650 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
3651 &value) == 0 && value) {
3652 int pathlen = strlen(path);
3653 char *tmp = zfs_strdup(hdl, path);
3654
3655 /*
3656 * If it starts with c#, and ends with "s0" or "s1",
3657 * chop the slice off, or if it ends with "s0/old" or
3658 * "s1/old", remove the slice from the middle.
3659 */
3660 if (CTD_CHECK(tmp)) {
3661 if (strcmp(&tmp[pathlen - 2], "s0") == 0 ||
3662 strcmp(&tmp[pathlen - 2], "s1") == 0) {
3663 tmp[pathlen - 2] = '\0';
3664 } else if (pathlen > 6 &&
3665 (strcmp(&tmp[pathlen - 6], "s0/old") == 0 ||
3666 strcmp(&tmp[pathlen - 6], "s1/old") == 0)) {
3667 (void) strcpy(&tmp[pathlen - 6],
3668 "/old");
3669 }
3670 }
3671 return (tmp);
3672 }
3673 } else {
3674 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &path) == 0);
3675
3676 /*
3677 * If it's a raidz device, we need to stick in the parity level.
3678 */
3679 if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) {
3680 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY,
3681 &value) == 0);
3682 (void) snprintf(buf, sizeof (buf), "%s%llu", path,
3683 (u_longlong_t)value);
3684 path = buf;
3685 }
3686
3687 /*
3688 * We identify each top-level vdev by using a <type-id>
3689 * naming convention.
3690 */
3691 if (verbose) {
3692 uint64_t id;
3693
3694 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
3695 &id) == 0);
3696 (void) snprintf(buf, sizeof (buf), "%s-%llu", path,
3697 (u_longlong_t)id);
3698 path = buf;
3699 }
3700 }
3701
3702 return (zfs_strdup(hdl, path));
3703 }
3704
3705 static int
3706 zbookmark_mem_compare(const void *a, const void *b)
3707 {
3708 return (memcmp(a, b, sizeof (zbookmark_phys_t)));
3709 }
3710
3711 /*
3712 * Retrieve the persistent error log, uniquify the members, and return to the
3713 * caller.
3714 */
3715 int
3716 zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp)
3717 {
3718 zfs_cmd_t zc = { 0 };
3719 uint64_t count;
3720 zbookmark_phys_t *zb = NULL;
3721 int i;
3722
3723 /*
3724 * Retrieve the raw error list from the kernel. If the number of errors
3725 * has increased, allocate more space and continue until we get the
3726 * entire list.
3727 */
3728 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT,
3729 &count) == 0);
3730 if (count == 0)
3731 return (0);
3732 if ((zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl,
3733 count * sizeof (zbookmark_phys_t))) == (uintptr_t)NULL)
3734 return (-1);
3735 zc.zc_nvlist_dst_size = count;
3736 (void) strcpy(zc.zc_name, zhp->zpool_name);
3737 for (;;) {
3738 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_ERROR_LOG,
3739 &zc) != 0) {
3740 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3741 if (errno == ENOMEM) {
3742 void *dst;
3743
3744 count = zc.zc_nvlist_dst_size;
3745 dst = zfs_alloc(zhp->zpool_hdl, count *
3746 sizeof (zbookmark_phys_t));
3747 if (dst == NULL)
3748 return (-1);
3749 zc.zc_nvlist_dst = (uintptr_t)dst;
3750 } else {
3751 return (-1);
3752 }
3753 } else {
3754 break;
3755 }
3756 }
3757
3758 /*
3759 * Sort the resulting bookmarks. This is a little confusing due to the
3760 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last
3761 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks
3762 * _not_ copied as part of the process. So we point the start of our
3763 * array appropriate and decrement the total number of elements.
3764 */
3765 zb = ((zbookmark_phys_t *)(uintptr_t)zc.zc_nvlist_dst) +
3766 zc.zc_nvlist_dst_size;
3767 count -= zc.zc_nvlist_dst_size;
3768
3769 qsort(zb, count, sizeof (zbookmark_phys_t), zbookmark_mem_compare);
3770
3771 verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0);
3772
3773 /*
3774 * Fill in the nverrlistp with nvlist's of dataset and object numbers.
3775 */
3776 for (i = 0; i < count; i++) {
3777 nvlist_t *nv;
3778
3779 /* ignoring zb_blkid and zb_level for now */
3780 if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset &&
3781 zb[i-1].zb_object == zb[i].zb_object)
3782 continue;
3783
3784 if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0)
3785 goto nomem;
3786 if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET,
3787 zb[i].zb_objset) != 0) {
3788 nvlist_free(nv);
3789 goto nomem;
3790 }
3791 if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT,
3792 zb[i].zb_object) != 0) {
3793 nvlist_free(nv);
3794 goto nomem;
3795 }
3796 if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) {
3797 nvlist_free(nv);
3798 goto nomem;
3799 }
3800 nvlist_free(nv);
3801 }
3802
3803 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3804 return (0);
3805
3806 nomem:
3807 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3808 return (no_memory(zhp->zpool_hdl));
3809 }
3810
3811 /*
3812 * Upgrade a ZFS pool to the latest on-disk version.
3813 */
3814 int
3815 zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version)
3816 {
3817 zfs_cmd_t zc = { 0 };
3818 libzfs_handle_t *hdl = zhp->zpool_hdl;
3819
3820 (void) strcpy(zc.zc_name, zhp->zpool_name);
3821 zc.zc_cookie = new_version;
3822
3823 if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0)
3824 return (zpool_standard_error_fmt(hdl, errno,
3825 dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"),
3826 zhp->zpool_name));
3827 return (0);
3828 }
3829
3830 void
3831 zfs_save_arguments(int argc, char **argv, char *string, int len)
3832 {
3833 (void) strlcpy(string, basename(argv[0]), len);
3834 for (int i = 1; i < argc; i++) {
3835 (void) strlcat(string, " ", len);
3836 (void) strlcat(string, argv[i], len);
3837 }
3838 }
3839
3840 int
3841 zpool_log_history(libzfs_handle_t *hdl, const char *message)
3842 {
3843 zfs_cmd_t zc = { 0 };
3844 nvlist_t *args;
3845 int err;
3846
3847 args = fnvlist_alloc();
3848 fnvlist_add_string(args, "message", message);
3849 err = zcmd_write_src_nvlist(hdl, &zc, args);
3850 if (err == 0)
3851 err = ioctl(hdl->libzfs_fd, ZFS_IOC_LOG_HISTORY, &zc);
3852 nvlist_free(args);
3853 zcmd_free_nvlists(&zc);
3854 return (err);
3855 }
3856
3857 /*
3858 * Perform ioctl to get some command history of a pool.
3859 *
3860 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the
3861 * logical offset of the history buffer to start reading from.
3862 *
3863 * Upon return, 'off' is the next logical offset to read from and
3864 * 'len' is the actual amount of bytes read into 'buf'.
3865 */
3866 static int
3867 get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len)
3868 {
3869 zfs_cmd_t zc = { 0 };
3870 libzfs_handle_t *hdl = zhp->zpool_hdl;
3871
3872 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3873
3874 zc.zc_history = (uint64_t)(uintptr_t)buf;
3875 zc.zc_history_len = *len;
3876 zc.zc_history_offset = *off;
3877
3878 if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) {
3879 switch (errno) {
3880 case EPERM:
3881 return (zfs_error_fmt(hdl, EZFS_PERM,
3882 dgettext(TEXT_DOMAIN,
3883 "cannot show history for pool '%s'"),
3884 zhp->zpool_name));
3885 case ENOENT:
3886 return (zfs_error_fmt(hdl, EZFS_NOHISTORY,
3887 dgettext(TEXT_DOMAIN, "cannot get history for pool "
3888 "'%s'"), zhp->zpool_name));
3889 case ENOTSUP:
3890 return (zfs_error_fmt(hdl, EZFS_BADVERSION,
3891 dgettext(TEXT_DOMAIN, "cannot get history for pool "
3892 "'%s', pool must be upgraded"), zhp->zpool_name));
3893 default:
3894 return (zpool_standard_error_fmt(hdl, errno,
3895 dgettext(TEXT_DOMAIN,
3896 "cannot get history for '%s'"), zhp->zpool_name));
3897 }
3898 }
3899
3900 *len = zc.zc_history_len;
3901 *off = zc.zc_history_offset;
3902
3903 return (0);
3904 }
3905
3906 /*
3907 * Process the buffer of nvlists, unpacking and storing each nvlist record
3908 * into 'records'. 'leftover' is set to the number of bytes that weren't
3909 * processed as there wasn't a complete record.
3910 */
3911 int
3912 zpool_history_unpack(char *buf, uint64_t bytes_read, uint64_t *leftover,
3913 nvlist_t ***records, uint_t *numrecords)
3914 {
3915 uint64_t reclen;
3916 nvlist_t *nv;
3917 int i;
3918
3919 while (bytes_read > sizeof (reclen)) {
3920
3921 /* get length of packed record (stored as little endian) */
3922 for (i = 0, reclen = 0; i < sizeof (reclen); i++)
3923 reclen += (uint64_t)(((uchar_t *)buf)[i]) << (8*i);
3924
3925 if (bytes_read < sizeof (reclen) + reclen)
3926 break;
3927
3928 /* unpack record */
3929 if (nvlist_unpack(buf + sizeof (reclen), reclen, &nv, 0) != 0)
3930 return (ENOMEM);
3931 bytes_read -= sizeof (reclen) + reclen;
3932 buf += sizeof (reclen) + reclen;
3933
3934 /* add record to nvlist array */
3935 (*numrecords)++;
3936 if (ISP2(*numrecords + 1)) {
3937 *records = realloc(*records,
3938 *numrecords * 2 * sizeof (nvlist_t *));
3939 }
3940 (*records)[*numrecords - 1] = nv;
3941 }
3942
3943 *leftover = bytes_read;
3944 return (0);
3945 }
3946
3947 /*
3948 * Retrieve the command history of a pool.
3949 */
3950 int
3951 zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp)
3952 {
3953 char *buf;
3954 int buflen = 128 * 1024;
3955 uint64_t off = 0;
3956 nvlist_t **records = NULL;
3957 uint_t numrecords = 0;
3958 int err, i;
3959
3960 buf = malloc(buflen);
3961 if (buf == NULL)
3962 return (ENOMEM);
3963 do {
3964 uint64_t bytes_read = buflen;
3965 uint64_t leftover;
3966
3967 if ((err = get_history(zhp, buf, &off, &bytes_read)) != 0)
3968 break;
3969
3970 /* if nothing else was read in, we're at EOF, just return */
3971 if (!bytes_read)
3972 break;
3973
3974 if ((err = zpool_history_unpack(buf, bytes_read,
3975 &leftover, &records, &numrecords)) != 0)
3976 break;
3977 off -= leftover;
3978 if (leftover == bytes_read) {
3979 /*
3980 * no progress made, because buffer is not big enough
3981 * to hold this record; resize and retry.
3982 */
3983 buflen *= 2;
3984 free(buf);
3985 buf = malloc(buflen);
3986 if (buf == NULL)
3987 return (ENOMEM);
3988 }
3989
3990 /* CONSTCOND */
3991 } while (1);
3992
3993 free(buf);
3994
3995 if (!err) {
3996 verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0);
3997 verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD,
3998 records, numrecords) == 0);
3999 }
4000 for (i = 0; i < numrecords; i++)
4001 nvlist_free(records[i]);
4002 free(records);
4003
4004 return (err);
4005 }
4006
4007 void
4008 zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
4009 char *pathname, size_t len)
4010 {
4011 zfs_cmd_t zc = { 0 };
4012 boolean_t mounted = B_FALSE;
4013 char *mntpnt = NULL;
4014 char dsname[ZFS_MAX_DATASET_NAME_LEN];
4015
4016 if (dsobj == 0) {
4017 /* special case for the MOS */
4018 (void) snprintf(pathname, len, "<metadata>:<0x%llx>", obj);
4019 return;
4020 }
4021
4022 /* get the dataset's name */
4023 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4024 zc.zc_obj = dsobj;
4025 if (ioctl(zhp->zpool_hdl->libzfs_fd,
4026 ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) {
4027 /* just write out a path of two object numbers */
4028 (void) snprintf(pathname, len, "<0x%llx>:<0x%llx>",
4029 dsobj, obj);
4030 return;
4031 }
4032 (void) strlcpy(dsname, zc.zc_value, sizeof (dsname));
4033
4034 /* find out if the dataset is mounted */
4035 mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt);
4036
4037 /* get the corrupted object's path */
4038 (void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name));
4039 zc.zc_obj = obj;
4040 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_OBJ_TO_PATH,
4041 &zc) == 0) {
4042 if (mounted) {
4043 (void) snprintf(pathname, len, "%s%s", mntpnt,
4044 zc.zc_value);
4045 } else {
4046 (void) snprintf(pathname, len, "%s:%s",
4047 dsname, zc.zc_value);
4048 }
4049 } else {
4050 (void) snprintf(pathname, len, "%s:<0x%llx>", dsname, obj);
4051 }
4052 free(mntpnt);
4053 }
4054
4055 /*
4056 * Read the EFI label from the config, if a label does not exist then
4057 * pass back the error to the caller. If the caller has passed a non-NULL
4058 * diskaddr argument then we set it to the starting address of the EFI
4059 * partition. If the caller has passed a non-NULL boolean argument, then
4060 * we set it to indicate if the disk does have efi system partition.
4061 */
4062 static int
4063 read_efi_label(nvlist_t *config, diskaddr_t *sb, boolean_t *system)
4064 {
4065 char *path;
4066 int fd;
4067 char diskname[MAXPATHLEN];
4068 boolean_t boot = B_FALSE;
4069 int err = -1;
4070 int slice;
4071
4072 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PATH, &path) != 0)
4073 return (err);
4074
4075 (void) snprintf(diskname, sizeof (diskname), "%s%s", ZFS_RDISK_ROOT,
4076 strrchr(path, '/'));
4077 if ((fd = open(diskname, O_RDONLY|O_NDELAY)) >= 0) {
4078 struct dk_gpt *vtoc;
4079
4080 if ((err = efi_alloc_and_read(fd, &vtoc)) >= 0) {
4081 for (slice = 0; slice < vtoc->efi_nparts; slice++) {
4082 if (vtoc->efi_parts[slice].p_tag == V_SYSTEM)
4083 boot = B_TRUE;
4084 if (vtoc->efi_parts[slice].p_tag == V_USR)
4085 break;
4086 }
4087 if (sb != NULL && vtoc->efi_parts[slice].p_tag == V_USR)
4088 *sb = vtoc->efi_parts[slice].p_start;
4089 if (system != NULL)
4090 *system = boot;
4091 efi_free(vtoc);
4092 }
4093 (void) close(fd);
4094 }
4095 return (err);
4096 }
4097
4098 /*
4099 * determine where a partition starts on a disk in the current
4100 * configuration
4101 */
4102 static diskaddr_t
4103 find_start_block(nvlist_t *config)
4104 {
4105 nvlist_t **child;
4106 uint_t c, children;
4107 diskaddr_t sb = MAXOFFSET_T;
4108 uint64_t wholedisk;
4109
4110 if (nvlist_lookup_nvlist_array(config,
4111 ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) {
4112 if (nvlist_lookup_uint64(config,
4113 ZPOOL_CONFIG_WHOLE_DISK,
4114 &wholedisk) != 0 || !wholedisk) {
4115 return (MAXOFFSET_T);
4116 }
4117 if (read_efi_label(config, &sb, NULL) < 0)
4118 sb = MAXOFFSET_T;
4119 return (sb);
4120 }
4121
4122 for (c = 0; c < children; c++) {
4123 sb = find_start_block(child[c]);
4124 if (sb != MAXOFFSET_T) {
4125 return (sb);
4126 }
4127 }
4128 return (MAXOFFSET_T);
4129 }
4130
4131 /*
4132 * Label an individual disk. The name provided is the short name,
4133 * stripped of any leading /dev path.
4134 */
4135 int
4136 zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, const char *name,
4137 zpool_boot_label_t boot_type, uint64_t boot_size, int *slice)
4138 {
4139 char path[MAXPATHLEN];
4140 struct dk_gpt *vtoc;
4141 int fd;
4142 size_t resv = EFI_MIN_RESV_SIZE;
4143 uint64_t slice_size;
4144 diskaddr_t start_block;
4145 char errbuf[1024];
4146
4147 /* prepare an error message just in case */
4148 (void) snprintf(errbuf, sizeof (errbuf),
4149 dgettext(TEXT_DOMAIN, "cannot label '%s'"), name);
4150
4151 if (zhp) {
4152 nvlist_t *nvroot;
4153
4154 verify(nvlist_lookup_nvlist(zhp->zpool_config,
4155 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
4156
4157 if (zhp->zpool_start_block == 0)
4158 start_block = find_start_block(nvroot);
4159 else
4160 start_block = zhp->zpool_start_block;
4161 zhp->zpool_start_block = start_block;
4162 } else {
4163 /* new pool */
4164 start_block = NEW_START_BLOCK;
4165 }
4166
4167 (void) snprintf(path, sizeof (path), "%s/%s%s", ZFS_RDISK_ROOT, name,
4168 BACKUP_SLICE);
4169
4170 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) {
4171 /*
4172 * This shouldn't happen. We've long since verified that this
4173 * is a valid device.
4174 */
4175 zfs_error_aux(hdl,
4176 dgettext(TEXT_DOMAIN, "unable to open device"));
4177 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
4178 }
4179
4180 if (efi_alloc_and_init(fd, EFI_NUMPAR, &vtoc) != 0) {
4181 /*
4182 * The only way this can fail is if we run out of memory, or we
4183 * were unable to read the disk's capacity
4184 */
4185 if (errno == ENOMEM)
4186 (void) no_memory(hdl);
4187
4188 (void) close(fd);
4189 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4190 "unable to read disk capacity"), name);
4191
4192 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
4193 }
4194
4195 /*
4196 * Why we use V_USR: V_BACKUP confuses users, and is considered
4197 * disposable by some EFI utilities (since EFI doesn't have a backup
4198 * slice). V_UNASSIGNED is supposed to be used only for zero size
4199 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT,
4200 * etc. were all pretty specific. V_USR is as close to reality as we
4201 * can get, in the absence of V_OTHER.
4202 */
4203 /* first fix the partition start block */
4204 if (start_block == MAXOFFSET_T)
4205 start_block = NEW_START_BLOCK;
4206
4207 /*
4208 * EFI System partition is using slice 0.
4209 * ZFS is on slice 1 and slice 8 is reserved.
4210 * We assume the GPT partition table without system
4211 * partition has zfs p_start == NEW_START_BLOCK.
4212 * If start_block != NEW_START_BLOCK, it means we have
4213 * system partition. Correct solution would be to query/cache vtoc
4214 * from existing vdev member.
4215 */
4216 if (boot_type == ZPOOL_CREATE_BOOT_LABEL) {
4217 if (boot_size % vtoc->efi_lbasize != 0) {
4218 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4219 "boot partition size must be a multiple of %d"),
4220 vtoc->efi_lbasize);
4221 (void) close(fd);
4222 efi_free(vtoc);
4223 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
4224 }
4225 /*
4226 * System partition size checks.
4227 * Note the 1MB is quite arbitrary value, since we
4228 * are creating dedicated pool, it should be enough
4229 * to hold fat + efi bootloader. May need to be
4230 * adjusted if the bootloader size will grow.
4231 */
4232 if (boot_size < 1024 * 1024) {
4233 char buf[64];
4234 zfs_nicenum(boot_size, buf, sizeof (buf));
4235 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4236 "Specified size %s for EFI System partition is too "
4237 "small, the minimum size is 1MB."), buf);
4238 (void) close(fd);
4239 efi_free(vtoc);
4240 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
4241 }
4242 /* 33MB is tested with mkfs -F pcfs */
4243 if (hdl->libzfs_printerr &&
4244 ((vtoc->efi_lbasize == 512 &&
4245 boot_size < 33 * 1024 * 1024) ||
4246 (vtoc->efi_lbasize == 4096 &&
4247 boot_size < 256 * 1024 * 1024))) {
4248 char buf[64];
4249 zfs_nicenum(boot_size, buf, sizeof (buf));
4250 (void) fprintf(stderr, dgettext(TEXT_DOMAIN,
4251 "Warning: EFI System partition size %s is "
4252 "not allowing to create FAT32 file\nsystem, which "
4253 "may result in unbootable system.\n"), buf);
4254 }
4255 /* Adjust zfs partition start by size of system partition. */
4256 start_block += boot_size / vtoc->efi_lbasize;
4257 }
4258
4259 if (start_block == NEW_START_BLOCK) {
4260 /*
4261 * Use default layout.
4262 * ZFS is on slice 0 and slice 8 is reserved.
4263 */
4264 slice_size = vtoc->efi_last_u_lba + 1;
4265 slice_size -= EFI_MIN_RESV_SIZE;
4266 slice_size -= start_block;
4267 if (slice != NULL)
4268 *slice = 0;
4269
4270 vtoc->efi_parts[0].p_start = start_block;
4271 vtoc->efi_parts[0].p_size = slice_size;
4272
4273 vtoc->efi_parts[0].p_tag = V_USR;
4274 (void) strcpy(vtoc->efi_parts[0].p_name, "zfs");
4275
4276 vtoc->efi_parts[8].p_start = slice_size + start_block;
4277 vtoc->efi_parts[8].p_size = resv;
4278 vtoc->efi_parts[8].p_tag = V_RESERVED;
4279 } else {
4280 slice_size = start_block - NEW_START_BLOCK;
4281 vtoc->efi_parts[0].p_start = NEW_START_BLOCK;
4282 vtoc->efi_parts[0].p_size = slice_size;
4283 vtoc->efi_parts[0].p_tag = V_SYSTEM;
4284 (void) strcpy(vtoc->efi_parts[0].p_name, "loader");
4285 if (slice != NULL)
4286 *slice = 1;
4287 /* prepare slice 1 */
4288 slice_size = vtoc->efi_last_u_lba + 1 - slice_size;
4289 slice_size -= resv;
4290 slice_size -= NEW_START_BLOCK;
4291 vtoc->efi_parts[1].p_start = start_block;
4292 vtoc->efi_parts[1].p_size = slice_size;
4293 vtoc->efi_parts[1].p_tag = V_USR;
4294 (void) strcpy(vtoc->efi_parts[1].p_name, "zfs");
4295
4296 vtoc->efi_parts[8].p_start = slice_size + start_block;
4297 vtoc->efi_parts[8].p_size = resv;
4298 vtoc->efi_parts[8].p_tag = V_RESERVED;
4299 }
4300
4301 if (efi_write(fd, vtoc) != 0) {
4302 /*
4303 * Some block drivers (like pcata) may not support EFI
4304 * GPT labels. Print out a helpful error message dir-
4305 * ecting the user to manually label the disk and give
4306 * a specific slice.
4307 */
4308 (void) close(fd);
4309 efi_free(vtoc);
4310
4311 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4312 "try using fdisk(1M) and then provide a specific slice"));
4313 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
4314 }
4315
4316 (void) close(fd);
4317 efi_free(vtoc);
4318 return (0);
4319 }
4320
4321 static boolean_t
4322 supported_dump_vdev_type(libzfs_handle_t *hdl, nvlist_t *config, char *errbuf)
4323 {
4324 char *type;
4325 nvlist_t **child;
4326 uint_t children, c;
4327
4328 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_TYPE, &type) == 0);
4329 if (strcmp(type, VDEV_TYPE_FILE) == 0 ||
4330 strcmp(type, VDEV_TYPE_HOLE) == 0 ||
4331 strcmp(type, VDEV_TYPE_MISSING) == 0) {
4332 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4333 "vdev type '%s' is not supported"), type);
4334 (void) zfs_error(hdl, EZFS_VDEVNOTSUP, errbuf);
4335 return (B_FALSE);
4336 }
4337 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN,
4338 &child, &children) == 0) {
4339 for (c = 0; c < children; c++) {
4340 if (!supported_dump_vdev_type(hdl, child[c], errbuf))
4341 return (B_FALSE);
4342 }
4343 }
4344 return (B_TRUE);
4345 }
4346
4347 /*
4348 * Check if this zvol is allowable for use as a dump device; zero if
4349 * it is, > 0 if it isn't, < 0 if it isn't a zvol.
4350 *
4351 * Allowable storage configurations include mirrors, all raidz variants, and
4352 * pools with log, cache, and spare devices. Pools which are backed by files or
4353 * have missing/hole vdevs are not suitable.
4354 */
4355 int
4356 zvol_check_dump_config(char *arg)
4357 {
4358 zpool_handle_t *zhp = NULL;
4359 nvlist_t *config, *nvroot;
4360 char *p, *volname;
4361 nvlist_t **top;
4362 uint_t toplevels;
4363 libzfs_handle_t *hdl;
4364 char errbuf[1024];
4365 char poolname[ZFS_MAX_DATASET_NAME_LEN];
4366 int pathlen = strlen(ZVOL_FULL_DEV_DIR);
4367 int ret = 1;
4368
4369 if (strncmp(arg, ZVOL_FULL_DEV_DIR, pathlen)) {
4370 return (-1);
4371 }
4372
4373 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
4374 "dump is not supported on device '%s'"), arg);
4375
4376 if ((hdl = libzfs_init()) == NULL)
4377 return (1);
4378 libzfs_print_on_error(hdl, B_TRUE);
4379
4380 volname = arg + pathlen;
4381
4382 /* check the configuration of the pool */
4383 if ((p = strchr(volname, '/')) == NULL) {
4384 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4385 "malformed dataset name"));
4386 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
4387 return (1);
4388 } else if (p - volname >= ZFS_MAX_DATASET_NAME_LEN) {
4389 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4390 "dataset name is too long"));
4391 (void) zfs_error(hdl, EZFS_NAMETOOLONG, errbuf);
4392 return (1);
4393 } else {
4394 (void) strncpy(poolname, volname, p - volname);
4395 poolname[p - volname] = '\0';
4396 }
4397
4398 if ((zhp = zpool_open(hdl, poolname)) == NULL) {
4399 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4400 "could not open pool '%s'"), poolname);
4401 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
4402 goto out;
4403 }
4404 config = zpool_get_config(zhp, NULL);
4405 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
4406 &nvroot) != 0) {
4407 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4408 "could not obtain vdev configuration for '%s'"), poolname);
4409 (void) zfs_error(hdl, EZFS_INVALCONFIG, errbuf);
4410 goto out;
4411 }
4412
4413 verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
4414 &top, &toplevels) == 0);
4415
4416 if (!supported_dump_vdev_type(hdl, top[0], errbuf)) {
4417 goto out;
4418 }
4419 ret = 0;
4420
4421 out:
4422 if (zhp)
4423 zpool_close(zhp);
4424 libzfs_fini(hdl);
4425 return (ret);
4426 }