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
648 case ZPOOL_PROP_READONLY:
649 if (!flags.import) {
650 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
651 "property '%s' can only be set at "
652 "import time"), propname);
653 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
654 goto error;
655 }
656 break;
657
658 case ZPOOL_PROP_TNAME:
659 if (!flags.create) {
660 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
661 "property '%s' can only be set at "
662 "creation time"), propname);
663 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
664 goto error;
665 }
666 break;
667
668 default:
669 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
670 "property '%s'(%d) not defined"), propname, prop);
671 break;
672 }
673 }
674
675 return (retprops);
676 error:
677 nvlist_free(retprops);
678 return (NULL);
679 }
680
681 /*
682 * Set zpool property : propname=propval.
683 */
684 int
685 zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval)
686 {
687 zfs_cmd_t zc = { 0 };
688 int ret = -1;
689 char errbuf[1024];
690 nvlist_t *nvl = NULL;
691 nvlist_t *realprops;
692 uint64_t version;
693 prop_flags_t flags = { 0 };
694
695 (void) snprintf(errbuf, sizeof (errbuf),
696 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
697 zhp->zpool_name);
698
699 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
700 return (no_memory(zhp->zpool_hdl));
701
702 if (nvlist_add_string(nvl, propname, propval) != 0) {
703 nvlist_free(nvl);
704 return (no_memory(zhp->zpool_hdl));
705 }
706
707 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
708 if ((realprops = zpool_valid_proplist(zhp->zpool_hdl,
709 zhp->zpool_name, nvl, version, flags, errbuf)) == NULL) {
710 nvlist_free(nvl);
711 return (-1);
712 }
713
714 nvlist_free(nvl);
715 nvl = realprops;
716
717 /*
718 * Execute the corresponding ioctl() to set this property.
719 */
720 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
721
722 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) {
723 nvlist_free(nvl);
724 return (-1);
725 }
726
727 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc);
728
729 zcmd_free_nvlists(&zc);
730 nvlist_free(nvl);
731
732 if (ret)
733 (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
734 else
735 (void) zpool_props_refresh(zhp);
736
737 return (ret);
738 }
739
740 int
741 zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp)
742 {
743 libzfs_handle_t *hdl = zhp->zpool_hdl;
744 zprop_list_t *entry;
745 char buf[ZFS_MAXPROPLEN];
746 nvlist_t *features = NULL;
747 zprop_list_t **last;
748 boolean_t firstexpand = (NULL == *plp);
749
750 if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0)
751 return (-1);
752
753 last = plp;
754 while (*last != NULL)
755 last = &(*last)->pl_next;
756
757 if ((*plp)->pl_all)
758 features = zpool_get_features(zhp);
759
760 if ((*plp)->pl_all && firstexpand) {
761 for (int i = 0; i < SPA_FEATURES; i++) {
762 zprop_list_t *entry = zfs_alloc(hdl,
763 sizeof (zprop_list_t));
764 entry->pl_prop = ZPROP_INVAL;
765 entry->pl_user_prop = zfs_asprintf(hdl, "feature@%s",
766 spa_feature_table[i].fi_uname);
767 entry->pl_width = strlen(entry->pl_user_prop);
768 entry->pl_all = B_TRUE;
769
770 *last = entry;
771 last = &entry->pl_next;
772 }
773 }
774
775 /* add any unsupported features */
776 for (nvpair_t *nvp = nvlist_next_nvpair(features, NULL);
777 nvp != NULL; nvp = nvlist_next_nvpair(features, nvp)) {
778 char *propname;
779 boolean_t found;
780 zprop_list_t *entry;
781
782 if (zfeature_is_supported(nvpair_name(nvp)))
783 continue;
784
785 propname = zfs_asprintf(hdl, "unsupported@%s",
786 nvpair_name(nvp));
787
788 /*
789 * Before adding the property to the list make sure that no
790 * other pool already added the same property.
791 */
792 found = B_FALSE;
793 entry = *plp;
794 while (entry != NULL) {
795 if (entry->pl_user_prop != NULL &&
796 strcmp(propname, entry->pl_user_prop) == 0) {
797 found = B_TRUE;
798 break;
799 }
800 entry = entry->pl_next;
801 }
802 if (found) {
803 free(propname);
804 continue;
805 }
806
807 entry = zfs_alloc(hdl, sizeof (zprop_list_t));
808 entry->pl_prop = ZPROP_INVAL;
809 entry->pl_user_prop = propname;
810 entry->pl_width = strlen(entry->pl_user_prop);
811 entry->pl_all = B_TRUE;
812
813 *last = entry;
814 last = &entry->pl_next;
815 }
816
817 for (entry = *plp; entry != NULL; entry = entry->pl_next) {
818
819 if (entry->pl_fixed)
820 continue;
821
822 if (entry->pl_prop != ZPROP_INVAL &&
823 zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf),
824 NULL, B_FALSE) == 0) {
825 if (strlen(buf) > entry->pl_width)
826 entry->pl_width = strlen(buf);
827 }
828 }
829
830 return (0);
831 }
832
833 /*
834 * Get the state for the given feature on the given ZFS pool.
835 */
836 int
837 zpool_prop_get_feature(zpool_handle_t *zhp, const char *propname, char *buf,
838 size_t len)
839 {
840 uint64_t refcount;
841 boolean_t found = B_FALSE;
842 nvlist_t *features = zpool_get_features(zhp);
843 boolean_t supported;
844 const char *feature = strchr(propname, '@') + 1;
845
846 supported = zpool_prop_feature(propname);
847 ASSERT(supported || zpool_prop_unsupported(propname));
848
849 /*
850 * Convert from feature name to feature guid. This conversion is
851 * unecessary for unsupported@... properties because they already
852 * use guids.
853 */
854 if (supported) {
855 int ret;
856 spa_feature_t fid;
857
858 ret = zfeature_lookup_name(feature, &fid);
859 if (ret != 0) {
860 (void) strlcpy(buf, "-", len);
861 return (ENOTSUP);
862 }
863 feature = spa_feature_table[fid].fi_guid;
864 }
865
866 if (nvlist_lookup_uint64(features, feature, &refcount) == 0)
867 found = B_TRUE;
868
869 if (supported) {
870 if (!found) {
871 (void) strlcpy(buf, ZFS_FEATURE_DISABLED, len);
872 } else {
873 if (refcount == 0)
874 (void) strlcpy(buf, ZFS_FEATURE_ENABLED, len);
875 else
876 (void) strlcpy(buf, ZFS_FEATURE_ACTIVE, len);
877 }
878 } else {
879 if (found) {
880 if (refcount == 0) {
881 (void) strcpy(buf, ZFS_UNSUPPORTED_INACTIVE);
882 } else {
883 (void) strcpy(buf, ZFS_UNSUPPORTED_READONLY);
884 }
885 } else {
886 (void) strlcpy(buf, "-", len);
887 return (ENOTSUP);
888 }
889 }
890
891 return (0);
892 }
893
894 /*
895 * Don't start the slice at the default block of 34; many storage
896 * devices will use a stripe width of 128k, so start there instead.
897 */
898 #define NEW_START_BLOCK 256
899
900 /*
901 * Validate the given pool name, optionally putting an extended error message in
902 * 'buf'.
903 */
904 boolean_t
905 zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool)
906 {
907 namecheck_err_t why;
908 char what;
909 int ret;
910
911 ret = pool_namecheck(pool, &why, &what);
912
913 /*
914 * The rules for reserved pool names were extended at a later point.
915 * But we need to support users with existing pools that may now be
916 * invalid. So we only check for this expanded set of names during a
917 * create (or import), and only in userland.
918 */
919 if (ret == 0 && !isopen &&
920 (strncmp(pool, "mirror", 6) == 0 ||
921 strncmp(pool, "raidz", 5) == 0 ||
922 strncmp(pool, "spare", 5) == 0 ||
923 strcmp(pool, "log") == 0)) {
924 if (hdl != NULL)
925 zfs_error_aux(hdl,
926 dgettext(TEXT_DOMAIN, "name is reserved"));
927 return (B_FALSE);
928 }
929
930
931 if (ret != 0) {
932 if (hdl != NULL) {
933 switch (why) {
934 case NAME_ERR_TOOLONG:
935 zfs_error_aux(hdl,
936 dgettext(TEXT_DOMAIN, "name is too long"));
937 break;
938
939 case NAME_ERR_INVALCHAR:
940 zfs_error_aux(hdl,
941 dgettext(TEXT_DOMAIN, "invalid character "
942 "'%c' in pool name"), what);
943 break;
944
945 case NAME_ERR_NOLETTER:
946 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
947 "name must begin with a letter"));
948 break;
949
950 case NAME_ERR_RESERVED:
951 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
952 "name is reserved"));
953 break;
954
955 case NAME_ERR_DISKLIKE:
956 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
957 "pool name is reserved"));
958 break;
959
960 case NAME_ERR_LEADING_SLASH:
961 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
962 "leading slash in name"));
963 break;
964
965 case NAME_ERR_EMPTY_COMPONENT:
966 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
967 "empty component in name"));
968 break;
969
970 case NAME_ERR_TRAILING_SLASH:
971 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
972 "trailing slash in name"));
973 break;
974
975 case NAME_ERR_MULTIPLE_DELIMITERS:
976 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
977 "multiple '@' and/or '#' delimiters in "
978 "name"));
979 break;
980
981 default:
982 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
983 "(%d) not defined"), why);
984 break;
985 }
986 }
987 return (B_FALSE);
988 }
989
990 return (B_TRUE);
991 }
992
993 /*
994 * Open a handle to the given pool, even if the pool is currently in the FAULTED
995 * state.
996 */
997 zpool_handle_t *
998 zpool_open_canfail(libzfs_handle_t *hdl, const char *pool)
999 {
1000 zpool_handle_t *zhp;
1001 boolean_t missing;
1002
1003 /*
1004 * Make sure the pool name is valid.
1005 */
1006 if (!zpool_name_valid(hdl, B_TRUE, pool)) {
1007 (void) zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1008 dgettext(TEXT_DOMAIN, "cannot open '%s'"),
1009 pool);
1010 return (NULL);
1011 }
1012
1013 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
1014 return (NULL);
1015
1016 zhp->zpool_hdl = hdl;
1017 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
1018
1019 if (zpool_refresh_stats(zhp, &missing) != 0) {
1020 zpool_close(zhp);
1021 return (NULL);
1022 }
1023
1024 if (missing) {
1025 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool"));
1026 (void) zfs_error_fmt(hdl, EZFS_NOENT,
1027 dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool);
1028 zpool_close(zhp);
1029 return (NULL);
1030 }
1031
1032 return (zhp);
1033 }
1034
1035 /*
1036 * Like the above, but silent on error. Used when iterating over pools (because
1037 * the configuration cache may be out of date).
1038 */
1039 int
1040 zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret)
1041 {
1042 zpool_handle_t *zhp;
1043 boolean_t missing;
1044
1045 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
1046 return (-1);
1047
1048 zhp->zpool_hdl = hdl;
1049 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
1050
1051 if (zpool_refresh_stats(zhp, &missing) != 0) {
1052 zpool_close(zhp);
1053 return (-1);
1054 }
1055
1056 if (missing) {
1057 zpool_close(zhp);
1058 *ret = NULL;
1059 return (0);
1060 }
1061
1062 *ret = zhp;
1063 return (0);
1064 }
1065
1066 /*
1067 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted
1068 * state.
1069 */
1070 zpool_handle_t *
1071 zpool_open(libzfs_handle_t *hdl, const char *pool)
1072 {
1073 zpool_handle_t *zhp;
1074
1075 if ((zhp = zpool_open_canfail(hdl, pool)) == NULL)
1076 return (NULL);
1077
1078 if (zhp->zpool_state == POOL_STATE_UNAVAIL) {
1079 (void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
1080 dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name);
1081 zpool_close(zhp);
1082 return (NULL);
1083 }
1084
1085 return (zhp);
1086 }
1087
1088 /*
1089 * Close the handle. Simply frees the memory associated with the handle.
1090 */
1091 void
1092 zpool_close(zpool_handle_t *zhp)
1093 {
1094 nvlist_free(zhp->zpool_config);
1095 nvlist_free(zhp->zpool_old_config);
1096 nvlist_free(zhp->zpool_props);
1097 free(zhp);
1098 }
1099
1100 /*
1101 * Return the name of the pool.
1102 */
1103 const char *
1104 zpool_get_name(zpool_handle_t *zhp)
1105 {
1106 return (zhp->zpool_name);
1107 }
1108
1109
1110 /*
1111 * Return the state of the pool (ACTIVE or UNAVAILABLE)
1112 */
1113 int
1114 zpool_get_state(zpool_handle_t *zhp)
1115 {
1116 return (zhp->zpool_state);
1117 }
1118
1119 /*
1120 * Create the named pool, using the provided vdev list. It is assumed
1121 * that the consumer has already validated the contents of the nvlist, so we
1122 * don't have to worry about error semantics.
1123 */
1124 int
1125 zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot,
1126 nvlist_t *props, nvlist_t *fsprops)
1127 {
1128 zfs_cmd_t zc = { 0 };
1129 nvlist_t *zc_fsprops = NULL;
1130 nvlist_t *zc_props = NULL;
1131 char msg[1024];
1132 int ret = -1;
1133
1134 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1135 "cannot create '%s'"), pool);
1136
1137 if (!zpool_name_valid(hdl, B_FALSE, pool))
1138 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
1139
1140 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1141 return (-1);
1142
1143 if (props) {
1144 prop_flags_t flags = { .create = B_TRUE, .import = B_FALSE };
1145
1146 if ((zc_props = zpool_valid_proplist(hdl, pool, props,
1147 SPA_VERSION_1, flags, msg)) == NULL) {
1148 goto create_failed;
1149 }
1150 }
1151
1152 if (fsprops) {
1153 uint64_t zoned;
1154 char *zonestr;
1155
1156 zoned = ((nvlist_lookup_string(fsprops,
1157 zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) &&
1158 strcmp(zonestr, "on") == 0);
1159
1160 if ((zc_fsprops = zfs_valid_proplist(hdl, ZFS_TYPE_FILESYSTEM,
1161 fsprops, zoned, NULL, NULL, msg)) == NULL) {
1162 goto create_failed;
1163 }
1164 if (!zc_props &&
1165 (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) {
1166 goto create_failed;
1167 }
1168 if (nvlist_add_nvlist(zc_props,
1169 ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) {
1170 goto create_failed;
1171 }
1172 }
1173
1174 if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
1175 goto create_failed;
1176
1177 (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name));
1178
1179 if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) {
1180
1181 zcmd_free_nvlists(&zc);
1182 nvlist_free(zc_props);
1183 nvlist_free(zc_fsprops);
1184
1185 switch (errno) {
1186 case EBUSY:
1187 /*
1188 * This can happen if the user has specified the same
1189 * device multiple times. We can't reliably detect this
1190 * until we try to add it and see we already have a
1191 * label.
1192 */
1193 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1194 "one or more vdevs refer to the same device"));
1195 return (zfs_error(hdl, EZFS_BADDEV, msg));
1196
1197 case ERANGE:
1198 /*
1199 * This happens if the record size is smaller or larger
1200 * than the allowed size range, or not a power of 2.
1201 *
1202 * NOTE: although zfs_valid_proplist is called earlier,
1203 * this case may have slipped through since the
1204 * pool does not exist yet and it is therefore
1205 * impossible to read properties e.g. max blocksize
1206 * from the pool.
1207 */
1208 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1209 "record size invalid"));
1210 return (zfs_error(hdl, EZFS_BADPROP, msg));
1211
1212 case EOVERFLOW:
1213 /*
1214 * This occurs when one of the devices is below
1215 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1216 * device was the problem device since there's no
1217 * reliable way to determine device size from userland.
1218 */
1219 {
1220 char buf[64];
1221
1222 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
1223
1224 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1225 "one or more devices is less than the "
1226 "minimum size (%s)"), buf);
1227 }
1228 return (zfs_error(hdl, EZFS_BADDEV, msg));
1229
1230 case ENOSPC:
1231 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1232 "one or more devices is out of space"));
1233 return (zfs_error(hdl, EZFS_BADDEV, msg));
1234
1235 case ENOTBLK:
1236 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1237 "cache device must be a disk or disk slice"));
1238 return (zfs_error(hdl, EZFS_BADDEV, msg));
1239
1240 default:
1241 return (zpool_standard_error(hdl, errno, msg));
1242 }
1243 }
1244
1245 create_failed:
1246 zcmd_free_nvlists(&zc);
1247 nvlist_free(zc_props);
1248 nvlist_free(zc_fsprops);
1249 return (ret);
1250 }
1251
1252 /*
1253 * Destroy the given pool. It is up to the caller to ensure that there are no
1254 * datasets left in the pool.
1255 */
1256 int
1257 zpool_destroy(zpool_handle_t *zhp, const char *log_str)
1258 {
1259 zfs_cmd_t zc = { 0 };
1260 zfs_handle_t *zfp = NULL;
1261 libzfs_handle_t *hdl = zhp->zpool_hdl;
1262 char msg[1024];
1263
1264 if (zhp->zpool_state == POOL_STATE_ACTIVE &&
1265 (zfp = zfs_open(hdl, zhp->zpool_name, ZFS_TYPE_FILESYSTEM)) == NULL)
1266 return (-1);
1267
1268 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1269 zc.zc_history = (uint64_t)(uintptr_t)log_str;
1270
1271 if (zfs_ioctl(hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) {
1272 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1273 "cannot destroy '%s'"), zhp->zpool_name);
1274
1275 if (errno == EROFS) {
1276 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1277 "one or more devices is read only"));
1278 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1279 } else {
1280 (void) zpool_standard_error(hdl, errno, msg);
1281 }
1282
1283 if (zfp)
1284 zfs_close(zfp);
1285 return (-1);
1286 }
1287
1288 if (zfp) {
1289 remove_mountpoint(zfp);
1290 zfs_close(zfp);
1291 }
1292
1293 return (0);
1294 }
1295
1296 /*
1297 * Create a checkpoint in the given pool.
1298 */
1299 int
1300 zpool_checkpoint(zpool_handle_t *zhp)
1301 {
1302 libzfs_handle_t *hdl = zhp->zpool_hdl;
1303 char msg[1024];
1304 int error;
1305
1306 error = lzc_pool_checkpoint(zhp->zpool_name);
1307 if (error != 0) {
1308 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1309 "cannot checkpoint '%s'"), zhp->zpool_name);
1310 (void) zpool_standard_error(hdl, error, msg);
1311 return (-1);
1312 }
1313
1314 return (0);
1315 }
1316
1317 /*
1318 * Discard the checkpoint from the given pool.
1319 */
1320 int
1321 zpool_discard_checkpoint(zpool_handle_t *zhp)
1322 {
1323 libzfs_handle_t *hdl = zhp->zpool_hdl;
1324 char msg[1024];
1325 int error;
1326
1327 error = lzc_pool_checkpoint_discard(zhp->zpool_name);
1328 if (error != 0) {
1329 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1330 "cannot discard checkpoint in '%s'"), zhp->zpool_name);
1331 (void) zpool_standard_error(hdl, error, msg);
1332 return (-1);
1333 }
1334
1335 return (0);
1336 }
1337
1338 /*
1339 * Add the given vdevs to the pool. The caller must have already performed the
1340 * necessary verification to ensure that the vdev specification is well-formed.
1341 */
1342 int
1343 zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot)
1344 {
1345 zfs_cmd_t zc = { 0 };
1346 int ret;
1347 libzfs_handle_t *hdl = zhp->zpool_hdl;
1348 char msg[1024];
1349 nvlist_t **spares, **l2cache;
1350 uint_t nspares, nl2cache;
1351
1352 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1353 "cannot add to '%s'"), zhp->zpool_name);
1354
1355 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1356 SPA_VERSION_SPARES &&
1357 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
1358 &spares, &nspares) == 0) {
1359 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1360 "upgraded to add hot spares"));
1361 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1362 }
1363
1364 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1365 SPA_VERSION_L2CACHE &&
1366 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
1367 &l2cache, &nl2cache) == 0) {
1368 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1369 "upgraded to add cache devices"));
1370 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1371 }
1372
1373 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1374 return (-1);
1375 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1376
1377 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) {
1378 switch (errno) {
1379 case EBUSY:
1380 /*
1381 * This can happen if the user has specified the same
1382 * device multiple times. We can't reliably detect this
1383 * until we try to add it and see we already have a
1384 * label.
1385 */
1386 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1387 "one or more vdevs refer to the same device"));
1388 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1389 break;
1390
1391 case EINVAL:
1392 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1393 "invalid config; a pool with removing/removed "
1394 "vdevs does not support adding raidz vdevs"));
1395 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1396 break;
1397
1398 case EOVERFLOW:
1399 /*
1400 * This occurrs when one of the devices is below
1401 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1402 * device was the problem device since there's no
1403 * reliable way to determine device size from userland.
1404 */
1405 {
1406 char buf[64];
1407
1408 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
1409
1410 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1411 "device is less than the minimum "
1412 "size (%s)"), buf);
1413 }
1414 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1415 break;
1416
1417 case ENOTSUP:
1418 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1419 "pool must be upgraded to add these vdevs"));
1420 (void) zfs_error(hdl, EZFS_BADVERSION, msg);
1421 break;
1422
1423 case EDOM:
1424 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1425 "root pool can not have multiple vdevs"
1426 " or separate logs"));
1427 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, msg);
1428 break;
1429
1430 case ENOTBLK:
1431 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1432 "cache device must be a disk or disk slice"));
1433 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1434 break;
1435
1436 default:
1437 (void) zpool_standard_error(hdl, errno, msg);
1438 }
1439
1440 ret = -1;
1441 } else {
1442 ret = 0;
1443 }
1444
1445 zcmd_free_nvlists(&zc);
1446
1447 return (ret);
1448 }
1449
1450 /*
1451 * Exports the pool from the system. The caller must ensure that there are no
1452 * mounted datasets in the pool.
1453 */
1454 static int
1455 zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce,
1456 const char *log_str)
1457 {
1458 zfs_cmd_t zc = { 0 };
1459 char msg[1024];
1460
1461 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1462 "cannot export '%s'"), zhp->zpool_name);
1463
1464 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1465 zc.zc_cookie = force;
1466 zc.zc_guid = hardforce;
1467 zc.zc_history = (uint64_t)(uintptr_t)log_str;
1468
1469 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) {
1470 switch (errno) {
1471 case EXDEV:
1472 zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
1473 "use '-f' to override the following errors:\n"
1474 "'%s' has an active shared spare which could be"
1475 " used by other pools once '%s' is exported."),
1476 zhp->zpool_name, zhp->zpool_name);
1477 return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE,
1478 msg));
1479 default:
1480 return (zpool_standard_error_fmt(zhp->zpool_hdl, errno,
1481 msg));
1482 }
1483 }
1484
1485 return (0);
1486 }
1487
1488 int
1489 zpool_export(zpool_handle_t *zhp, boolean_t force, const char *log_str)
1490 {
1491 return (zpool_export_common(zhp, force, B_FALSE, log_str));
1492 }
1493
1494 int
1495 zpool_export_force(zpool_handle_t *zhp, const char *log_str)
1496 {
1497 return (zpool_export_common(zhp, B_TRUE, B_TRUE, log_str));
1498 }
1499
1500 static void
1501 zpool_rewind_exclaim(libzfs_handle_t *hdl, const char *name, boolean_t dryrun,
1502 nvlist_t *config)
1503 {
1504 nvlist_t *nv = NULL;
1505 uint64_t rewindto;
1506 int64_t loss = -1;
1507 struct tm t;
1508 char timestr[128];
1509
1510 if (!hdl->libzfs_printerr || config == NULL)
1511 return;
1512
1513 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
1514 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0) {
1515 return;
1516 }
1517
1518 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1519 return;
1520 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1521
1522 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1523 strftime(timestr, 128, 0, &t) != 0) {
1524 if (dryrun) {
1525 (void) printf(dgettext(TEXT_DOMAIN,
1526 "Would be able to return %s "
1527 "to its state as of %s.\n"),
1528 name, timestr);
1529 } else {
1530 (void) printf(dgettext(TEXT_DOMAIN,
1531 "Pool %s returned to its state as of %s.\n"),
1532 name, timestr);
1533 }
1534 if (loss > 120) {
1535 (void) printf(dgettext(TEXT_DOMAIN,
1536 "%s approximately %lld "),
1537 dryrun ? "Would discard" : "Discarded",
1538 (loss + 30) / 60);
1539 (void) printf(dgettext(TEXT_DOMAIN,
1540 "minutes of transactions.\n"));
1541 } else if (loss > 0) {
1542 (void) printf(dgettext(TEXT_DOMAIN,
1543 "%s approximately %lld "),
1544 dryrun ? "Would discard" : "Discarded", loss);
1545 (void) printf(dgettext(TEXT_DOMAIN,
1546 "seconds of transactions.\n"));
1547 }
1548 }
1549 }
1550
1551 void
1552 zpool_explain_recover(libzfs_handle_t *hdl, const char *name, int reason,
1553 nvlist_t *config)
1554 {
1555 nvlist_t *nv = NULL;
1556 int64_t loss = -1;
1557 uint64_t edata = UINT64_MAX;
1558 uint64_t rewindto;
1559 struct tm t;
1560 char timestr[128];
1561
1562 if (!hdl->libzfs_printerr)
1563 return;
1564
1565 if (reason >= 0)
1566 (void) printf(dgettext(TEXT_DOMAIN, "action: "));
1567 else
1568 (void) printf(dgettext(TEXT_DOMAIN, "\t"));
1569
1570 /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */
1571 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
1572 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0 ||
1573 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1574 goto no_info;
1575
1576 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1577 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_DATA_ERRORS,
1578 &edata);
1579
1580 (void) printf(dgettext(TEXT_DOMAIN,
1581 "Recovery is possible, but will result in some data loss.\n"));
1582
1583 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1584 strftime(timestr, 128, 0, &t) != 0) {
1585 (void) printf(dgettext(TEXT_DOMAIN,
1586 "\tReturning the pool to its state as of %s\n"
1587 "\tshould correct the problem. "),
1588 timestr);
1589 } else {
1590 (void) printf(dgettext(TEXT_DOMAIN,
1591 "\tReverting the pool to an earlier state "
1592 "should correct the problem.\n\t"));
1593 }
1594
1595 if (loss > 120) {
1596 (void) printf(dgettext(TEXT_DOMAIN,
1597 "Approximately %lld minutes of data\n"
1598 "\tmust be discarded, irreversibly. "), (loss + 30) / 60);
1599 } else if (loss > 0) {
1600 (void) printf(dgettext(TEXT_DOMAIN,
1601 "Approximately %lld seconds of data\n"
1602 "\tmust be discarded, irreversibly. "), loss);
1603 }
1604 if (edata != 0 && edata != UINT64_MAX) {
1605 if (edata == 1) {
1606 (void) printf(dgettext(TEXT_DOMAIN,
1607 "After rewind, at least\n"
1608 "\tone persistent user-data error will remain. "));
1609 } else {
1610 (void) printf(dgettext(TEXT_DOMAIN,
1611 "After rewind, several\n"
1612 "\tpersistent user-data errors will remain. "));
1613 }
1614 }
1615 (void) printf(dgettext(TEXT_DOMAIN,
1616 "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "),
1617 reason >= 0 ? "clear" : "import", name);
1618
1619 (void) printf(dgettext(TEXT_DOMAIN,
1620 "A scrub of the pool\n"
1621 "\tis strongly recommended after recovery.\n"));
1622 return;
1623
1624 no_info:
1625 (void) printf(dgettext(TEXT_DOMAIN,
1626 "Destroy and re-create the pool from\n\ta backup source.\n"));
1627 }
1628
1629 /*
1630 * zpool_import() is a contracted interface. Should be kept the same
1631 * if possible.
1632 *
1633 * Applications should use zpool_import_props() to import a pool with
1634 * new properties value to be set.
1635 */
1636 int
1637 zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1638 char *altroot)
1639 {
1640 nvlist_t *props = NULL;
1641 int ret;
1642
1643 if (altroot != NULL) {
1644 if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) {
1645 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1646 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1647 newname));
1648 }
1649
1650 if (nvlist_add_string(props,
1651 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 ||
1652 nvlist_add_string(props,
1653 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) {
1654 nvlist_free(props);
1655 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1656 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1657 newname));
1658 }
1659 }
1660
1661 ret = zpool_import_props(hdl, config, newname, props,
1662 ZFS_IMPORT_NORMAL);
1663 nvlist_free(props);
1664 return (ret);
1665 }
1666
1667 static void
1668 print_vdev_tree(libzfs_handle_t *hdl, const char *name, nvlist_t *nv,
1669 int indent)
1670 {
1671 nvlist_t **child;
1672 uint_t c, children;
1673 char *vname;
1674 uint64_t is_log = 0;
1675
1676 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG,
1677 &is_log);
1678
1679 if (name != NULL)
1680 (void) printf("\t%*s%s%s\n", indent, "", name,
1681 is_log ? " [log]" : "");
1682
1683 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1684 &child, &children) != 0)
1685 return;
1686
1687 for (c = 0; c < children; c++) {
1688 vname = zpool_vdev_name(hdl, NULL, child[c], B_TRUE);
1689 print_vdev_tree(hdl, vname, child[c], indent + 2);
1690 free(vname);
1691 }
1692 }
1693
1694 void
1695 zpool_print_unsup_feat(nvlist_t *config)
1696 {
1697 nvlist_t *nvinfo, *unsup_feat;
1698
1699 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nvinfo) ==
1700 0);
1701 verify(nvlist_lookup_nvlist(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT,
1702 &unsup_feat) == 0);
1703
1704 for (nvpair_t *nvp = nvlist_next_nvpair(unsup_feat, NULL); nvp != NULL;
1705 nvp = nvlist_next_nvpair(unsup_feat, nvp)) {
1706 char *desc;
1707
1708 verify(nvpair_type(nvp) == DATA_TYPE_STRING);
1709 verify(nvpair_value_string(nvp, &desc) == 0);
1710
1711 if (strlen(desc) > 0)
1712 (void) printf("\t%s (%s)\n", nvpair_name(nvp), desc);
1713 else
1714 (void) printf("\t%s\n", nvpair_name(nvp));
1715 }
1716 }
1717
1718 /*
1719 * Import the given pool using the known configuration and a list of
1720 * properties to be set. The configuration should have come from
1721 * zpool_find_import(). The 'newname' parameters control whether the pool
1722 * is imported with a different name.
1723 */
1724 int
1725 zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1726 nvlist_t *props, int flags)
1727 {
1728 zfs_cmd_t zc = { 0 };
1729 zpool_load_policy_t policy;
1730 nvlist_t *nv = NULL;
1731 nvlist_t *nvinfo = NULL;
1732 nvlist_t *missing = NULL;
1733 char *thename;
1734 char *origname;
1735 int ret;
1736 int error = 0;
1737 char errbuf[1024];
1738
1739 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1740 &origname) == 0);
1741
1742 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1743 "cannot import pool '%s'"), origname);
1744
1745 if (newname != NULL) {
1746 if (!zpool_name_valid(hdl, B_FALSE, newname))
1747 return (zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1748 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1749 newname));
1750 thename = (char *)newname;
1751 } else {
1752 thename = origname;
1753 }
1754
1755 if (props != NULL) {
1756 uint64_t version;
1757 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
1758
1759 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
1760 &version) == 0);
1761
1762 if ((props = zpool_valid_proplist(hdl, origname,
1763 props, version, flags, errbuf)) == NULL)
1764 return (-1);
1765 if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) {
1766 nvlist_free(props);
1767 return (-1);
1768 }
1769 nvlist_free(props);
1770 }
1771
1772 (void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name));
1773
1774 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1775 &zc.zc_guid) == 0);
1776
1777 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) {
1778 zcmd_free_nvlists(&zc);
1779 return (-1);
1780 }
1781 if (zcmd_alloc_dst_nvlist(hdl, &zc, zc.zc_nvlist_conf_size * 2) != 0) {
1782 zcmd_free_nvlists(&zc);
1783 return (-1);
1784 }
1785
1786 zc.zc_cookie = flags;
1787 while ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc)) != 0 &&
1788 errno == ENOMEM) {
1789 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
1790 zcmd_free_nvlists(&zc);
1791 return (-1);
1792 }
1793 }
1794 if (ret != 0)
1795 error = errno;
1796
1797 (void) zcmd_read_dst_nvlist(hdl, &zc, &nv);
1798
1799 zcmd_free_nvlists(&zc);
1800
1801 zpool_get_load_policy(config, &policy);
1802
1803 if (error) {
1804 char desc[1024];
1805
1806 /*
1807 * Dry-run failed, but we print out what success
1808 * looks like if we found a best txg
1809 */
1810 if (policy.zlp_rewind & ZPOOL_TRY_REWIND) {
1811 zpool_rewind_exclaim(hdl, newname ? origname : thename,
1812 B_TRUE, nv);
1813 nvlist_free(nv);
1814 return (-1);
1815 }
1816
1817 if (newname == NULL)
1818 (void) snprintf(desc, sizeof (desc),
1819 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1820 thename);
1821 else
1822 (void) snprintf(desc, sizeof (desc),
1823 dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"),
1824 origname, thename);
1825
1826 switch (error) {
1827 case ENOTSUP:
1828 if (nv != NULL && nvlist_lookup_nvlist(nv,
1829 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
1830 nvlist_exists(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT)) {
1831 (void) printf(dgettext(TEXT_DOMAIN, "This "
1832 "pool uses the following feature(s) not "
1833 "supported by this system:\n"));
1834 zpool_print_unsup_feat(nv);
1835 if (nvlist_exists(nvinfo,
1836 ZPOOL_CONFIG_CAN_RDONLY)) {
1837 (void) printf(dgettext(TEXT_DOMAIN,
1838 "All unsupported features are only "
1839 "required for writing to the pool."
1840 "\nThe pool can be imported using "
1841 "'-o readonly=on'.\n"));
1842 }
1843 }
1844 /*
1845 * Unsupported version.
1846 */
1847 (void) zfs_error(hdl, EZFS_BADVERSION, desc);
1848 break;
1849
1850 case EINVAL:
1851 (void) zfs_error(hdl, EZFS_INVALCONFIG, desc);
1852 break;
1853
1854 case EROFS:
1855 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1856 "one or more devices is read only"));
1857 (void) zfs_error(hdl, EZFS_BADDEV, desc);
1858 break;
1859
1860 case ENXIO:
1861 if (nv && nvlist_lookup_nvlist(nv,
1862 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
1863 nvlist_lookup_nvlist(nvinfo,
1864 ZPOOL_CONFIG_MISSING_DEVICES, &missing) == 0) {
1865 (void) printf(dgettext(TEXT_DOMAIN,
1866 "The devices below are missing or "
1867 "corrupted, use '-m' to import the pool "
1868 "anyway:\n"));
1869 print_vdev_tree(hdl, NULL, missing, 2);
1870 (void) printf("\n");
1871 }
1872 (void) zpool_standard_error(hdl, error, desc);
1873 break;
1874
1875 case EEXIST:
1876 (void) zpool_standard_error(hdl, error, desc);
1877 break;
1878 case ENAMETOOLONG:
1879 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1880 "new name of at least one dataset is longer than "
1881 "the maximum allowable length"));
1882 (void) zfs_error(hdl, EZFS_NAMETOOLONG, desc);
1883 break;
1884 default:
1885 (void) zpool_standard_error(hdl, error, desc);
1886 zpool_explain_recover(hdl,
1887 newname ? origname : thename, -error, nv);
1888 break;
1889 }
1890
1891 nvlist_free(nv);
1892 ret = -1;
1893 } else {
1894 zpool_handle_t *zhp;
1895
1896 /*
1897 * This should never fail, but play it safe anyway.
1898 */
1899 if (zpool_open_silent(hdl, thename, &zhp) != 0)
1900 ret = -1;
1901 else if (zhp != NULL)
1902 zpool_close(zhp);
1903 if (policy.zlp_rewind &
1904 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
1905 zpool_rewind_exclaim(hdl, newname ? origname : thename,
1906 ((policy.zlp_rewind & ZPOOL_TRY_REWIND) != 0), nv);
1907 }
1908 nvlist_free(nv);
1909 return (0);
1910 }
1911
1912 return (ret);
1913 }
1914
1915 /*
1916 * Scan the pool.
1917 */
1918 int
1919 zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func, pool_scrub_cmd_t cmd)
1920 {
1921 zfs_cmd_t zc = { 0 };
1922 char msg[1024];
1923 int err;
1924 libzfs_handle_t *hdl = zhp->zpool_hdl;
1925
1926 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1927 zc.zc_cookie = func;
1928 zc.zc_flags = cmd;
1929
1930 if (zfs_ioctl(hdl, ZFS_IOC_POOL_SCAN, &zc) == 0)
1931 return (0);
1932
1933 err = errno;
1934
1935 /* ECANCELED on a scrub means we resumed a paused scrub */
1936 if (err == ECANCELED && func == POOL_SCAN_SCRUB &&
1937 cmd == POOL_SCRUB_NORMAL)
1938 return (0);
1939
1940 if (err == ENOENT && func != POOL_SCAN_NONE && cmd == POOL_SCRUB_NORMAL)
1941 return (0);
1942
1943 if (func == POOL_SCAN_SCRUB) {
1944 if (cmd == POOL_SCRUB_PAUSE) {
1945 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1946 "cannot pause scrubbing %s"), zc.zc_name);
1947 } else {
1948 assert(cmd == POOL_SCRUB_NORMAL);
1949 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1950 "cannot scrub %s"), zc.zc_name);
1951 }
1952 } else if (func == POOL_SCAN_NONE) {
1953 (void) snprintf(msg, sizeof (msg),
1954 dgettext(TEXT_DOMAIN, "cannot cancel scrubbing %s"),
1955 zc.zc_name);
1956 } else {
1957 assert(!"unexpected result");
1958 }
1959
1960 if (err == EBUSY) {
1961 nvlist_t *nvroot;
1962 pool_scan_stat_t *ps = NULL;
1963 uint_t psc;
1964
1965 verify(nvlist_lookup_nvlist(zhp->zpool_config,
1966 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
1967 (void) nvlist_lookup_uint64_array(nvroot,
1968 ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc);
1969 if (ps && ps->pss_func == POOL_SCAN_SCRUB) {
1970 if (cmd == POOL_SCRUB_PAUSE)
1971 return (zfs_error(hdl, EZFS_SCRUB_PAUSED, msg));
1972 else
1973 return (zfs_error(hdl, EZFS_SCRUBBING, msg));
1974 } else {
1975 return (zfs_error(hdl, EZFS_RESILVERING, msg));
1976 }
1977 } else if (err == ENOENT) {
1978 return (zfs_error(hdl, EZFS_NO_SCRUB, msg));
1979 } else {
1980 return (zpool_standard_error(hdl, err, msg));
1981 }
1982 }
1983
1984 static int
1985 xlate_init_err(int err)
1986 {
1987 switch (err) {
1988 case ENODEV:
1989 return (EZFS_NODEVICE);
1990 case EINVAL:
1991 case EROFS:
1992 return (EZFS_BADDEV);
1993 case EBUSY:
1994 return (EZFS_INITIALIZING);
1995 case ESRCH:
1996 return (EZFS_NO_INITIALIZE);
1997 }
1998 return (err);
1999 }
2000
2001 /*
2002 * Begin, suspend, or cancel the initialization (initializing of all free
2003 * blocks) for the given vdevs in the given pool.
2004 */
2005 int
2006 zpool_initialize(zpool_handle_t *zhp, pool_initialize_func_t cmd_type,
2007 nvlist_t *vds)
2008 {
2009 char msg[1024];
2010 libzfs_handle_t *hdl = zhp->zpool_hdl;
2011
2012 nvlist_t *errlist;
2013
2014 /* translate vdev names to guids */
2015 nvlist_t *vdev_guids = fnvlist_alloc();
2016 nvlist_t *guids_to_paths = fnvlist_alloc();
2017 boolean_t spare, cache;
2018 nvlist_t *tgt;
2019 nvpair_t *elem;
2020
2021 for (elem = nvlist_next_nvpair(vds, NULL); elem != NULL;
2022 elem = nvlist_next_nvpair(vds, elem)) {
2023 char *vd_path = nvpair_name(elem);
2024 tgt = zpool_find_vdev(zhp, vd_path, &spare, &cache, NULL);
2025
2026 if ((tgt == NULL) || cache || spare) {
2027 (void) snprintf(msg, sizeof (msg),
2028 dgettext(TEXT_DOMAIN, "cannot initialize '%s'"),
2029 vd_path);
2030 int err = (tgt == NULL) ? EZFS_NODEVICE :
2031 (spare ? EZFS_ISSPARE : EZFS_ISL2CACHE);
2032 fnvlist_free(vdev_guids);
2033 fnvlist_free(guids_to_paths);
2034 return (zfs_error(hdl, err, msg));
2035 }
2036
2037 uint64_t guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
2038 fnvlist_add_uint64(vdev_guids, vd_path, guid);
2039
2040 (void) snprintf(msg, sizeof (msg), "%llu", guid);
2041 fnvlist_add_string(guids_to_paths, msg, vd_path);
2042 }
2043
2044 int err = lzc_initialize(zhp->zpool_name, cmd_type, vdev_guids,
2045 &errlist);
2046 fnvlist_free(vdev_guids);
2047
2048 if (err == 0) {
2049 fnvlist_free(guids_to_paths);
2050 return (0);
2051 }
2052
2053 nvlist_t *vd_errlist = NULL;
2054 if (errlist != NULL) {
2055 vd_errlist = fnvlist_lookup_nvlist(errlist,
2056 ZPOOL_INITIALIZE_VDEVS);
2057 }
2058
2059 (void) snprintf(msg, sizeof (msg),
2060 dgettext(TEXT_DOMAIN, "operation failed"));
2061
2062 for (elem = nvlist_next_nvpair(vd_errlist, NULL); elem != NULL;
2063 elem = nvlist_next_nvpair(vd_errlist, elem)) {
2064 int64_t vd_error = xlate_init_err(fnvpair_value_int64(elem));
2065 char *path = fnvlist_lookup_string(guids_to_paths,
2066 nvpair_name(elem));
2067 (void) zfs_error_fmt(hdl, vd_error, "cannot initialize '%s'",
2068 path);
2069 }
2070
2071 fnvlist_free(guids_to_paths);
2072 if (vd_errlist != NULL)
2073 return (-1);
2074
2075 return (zpool_standard_error(hdl, err, msg));
2076 }
2077
2078 /*
2079 * This provides a very minimal check whether a given string is likely a
2080 * c#t#d# style string. Users of this are expected to do their own
2081 * verification of the s# part.
2082 */
2083 #define CTD_CHECK(str) (str && str[0] == 'c' && isdigit(str[1]))
2084
2085 /*
2086 * More elaborate version for ones which may start with "/dev/dsk/"
2087 * and the like.
2088 */
2089 static int
2090 ctd_check_path(char *str)
2091 {
2092 /*
2093 * If it starts with a slash, check the last component.
2094 */
2095 if (str && str[0] == '/') {
2096 char *tmp = strrchr(str, '/');
2097
2098 /*
2099 * If it ends in "/old", check the second-to-last
2100 * component of the string instead.
2101 */
2102 if (tmp != str && strcmp(tmp, "/old") == 0) {
2103 for (tmp--; *tmp != '/'; tmp--)
2104 ;
2105 }
2106 str = tmp + 1;
2107 }
2108 return (CTD_CHECK(str));
2109 }
2110
2111 /*
2112 * Find a vdev that matches the search criteria specified. We use the
2113 * the nvpair name to determine how we should look for the device.
2114 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
2115 * spare; but FALSE if its an INUSE spare.
2116 */
2117 static nvlist_t *
2118 vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare,
2119 boolean_t *l2cache, boolean_t *log)
2120 {
2121 uint_t c, children;
2122 nvlist_t **child;
2123 nvlist_t *ret;
2124 uint64_t is_log;
2125 char *srchkey;
2126 nvpair_t *pair = nvlist_next_nvpair(search, NULL);
2127
2128 /* Nothing to look for */
2129 if (search == NULL || pair == NULL)
2130 return (NULL);
2131
2132 /* Obtain the key we will use to search */
2133 srchkey = nvpair_name(pair);
2134
2135 switch (nvpair_type(pair)) {
2136 case DATA_TYPE_UINT64:
2137 if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) {
2138 uint64_t srchval, theguid;
2139
2140 verify(nvpair_value_uint64(pair, &srchval) == 0);
2141 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
2142 &theguid) == 0);
2143 if (theguid == srchval)
2144 return (nv);
2145 }
2146 break;
2147
2148 case DATA_TYPE_STRING: {
2149 char *srchval, *val;
2150
2151 verify(nvpair_value_string(pair, &srchval) == 0);
2152 if (nvlist_lookup_string(nv, srchkey, &val) != 0)
2153 break;
2154
2155 /*
2156 * Search for the requested value. Special cases:
2157 *
2158 * - ZPOOL_CONFIG_PATH for whole disk entries. To support
2159 * UEFI boot, these end in "s0" or "s0/old" or "s1" or
2160 * "s1/old". The "s0" or "s1" part is hidden from the user,
2161 * but included in the string, so this matches around it.
2162 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE).
2163 *
2164 * Otherwise, all other searches are simple string compares.
2165 */
2166 if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0 &&
2167 ctd_check_path(val)) {
2168 uint64_t wholedisk = 0;
2169
2170 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
2171 &wholedisk);
2172 if (wholedisk) {
2173 int slen = strlen(srchval);
2174 int vlen = strlen(val);
2175
2176 if (slen != vlen - 2)
2177 break;
2178
2179 /*
2180 * make_leaf_vdev() should only set
2181 * wholedisk for ZPOOL_CONFIG_PATHs which
2182 * will include "/dev/dsk/", giving plenty of
2183 * room for the indices used next.
2184 */
2185 ASSERT(vlen >= 6);
2186
2187 /*
2188 * strings identical except trailing "s0"
2189 */
2190 if ((strcmp(&val[vlen - 2], "s0") == 0 ||
2191 strcmp(&val[vlen - 2], "s1") == 0) &&
2192 strncmp(srchval, val, slen) == 0)
2193 return (nv);
2194
2195 /*
2196 * strings identical except trailing "s0/old"
2197 */
2198 if ((strcmp(&val[vlen - 6], "s0/old") == 0 ||
2199 strcmp(&val[vlen - 6], "s1/old") == 0) &&
2200 strcmp(&srchval[slen - 4], "/old") == 0 &&
2201 strncmp(srchval, val, slen - 4) == 0)
2202 return (nv);
2203
2204 break;
2205 }
2206 } else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0 && val) {
2207 char *type, *idx, *end, *p;
2208 uint64_t id, vdev_id;
2209
2210 /*
2211 * Determine our vdev type, keeping in mind
2212 * that the srchval is composed of a type and
2213 * vdev id pair (i.e. mirror-4).
2214 */
2215 if ((type = strdup(srchval)) == NULL)
2216 return (NULL);
2217
2218 if ((p = strrchr(type, '-')) == NULL) {
2219 free(type);
2220 break;
2221 }
2222 idx = p + 1;
2223 *p = '\0';
2224
2225 /*
2226 * If the types don't match then keep looking.
2227 */
2228 if (strncmp(val, type, strlen(val)) != 0) {
2229 free(type);
2230 break;
2231 }
2232
2233 verify(zpool_vdev_is_interior(type));
2234 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
2235 &id) == 0);
2236
2237 errno = 0;
2238 vdev_id = strtoull(idx, &end, 10);
2239
2240 free(type);
2241 if (errno != 0)
2242 return (NULL);
2243
2244 /*
2245 * Now verify that we have the correct vdev id.
2246 */
2247 if (vdev_id == id)
2248 return (nv);
2249 }
2250
2251 /*
2252 * Common case
2253 */
2254 if (strcmp(srchval, val) == 0)
2255 return (nv);
2256 break;
2257 }
2258
2259 default:
2260 break;
2261 }
2262
2263 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
2264 &child, &children) != 0)
2265 return (NULL);
2266
2267 for (c = 0; c < children; c++) {
2268 if ((ret = vdev_to_nvlist_iter(child[c], search,
2269 avail_spare, l2cache, NULL)) != NULL) {
2270 /*
2271 * The 'is_log' value is only set for the toplevel
2272 * vdev, not the leaf vdevs. So we always lookup the
2273 * log device from the root of the vdev tree (where
2274 * 'log' is non-NULL).
2275 */
2276 if (log != NULL &&
2277 nvlist_lookup_uint64(child[c],
2278 ZPOOL_CONFIG_IS_LOG, &is_log) == 0 &&
2279 is_log) {
2280 *log = B_TRUE;
2281 }
2282 return (ret);
2283 }
2284 }
2285
2286 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
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 *avail_spare = B_TRUE;
2292 return (ret);
2293 }
2294 }
2295 }
2296
2297 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
2298 &child, &children) == 0) {
2299 for (c = 0; c < children; c++) {
2300 if ((ret = vdev_to_nvlist_iter(child[c], search,
2301 avail_spare, l2cache, NULL)) != NULL) {
2302 *l2cache = B_TRUE;
2303 return (ret);
2304 }
2305 }
2306 }
2307
2308 return (NULL);
2309 }
2310
2311 /*
2312 * Given a physical path (minus the "/devices" prefix), find the
2313 * associated vdev.
2314 */
2315 nvlist_t *
2316 zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath,
2317 boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log)
2318 {
2319 nvlist_t *search, *nvroot, *ret;
2320
2321 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2322 verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath) == 0);
2323
2324 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
2325 &nvroot) == 0);
2326
2327 *avail_spare = B_FALSE;
2328 *l2cache = B_FALSE;
2329 if (log != NULL)
2330 *log = B_FALSE;
2331 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
2332 nvlist_free(search);
2333
2334 return (ret);
2335 }
2336
2337 /*
2338 * Determine if we have an "interior" top-level vdev (i.e mirror/raidz).
2339 */
2340 static boolean_t
2341 zpool_vdev_is_interior(const char *name)
2342 {
2343 if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 ||
2344 strncmp(name, VDEV_TYPE_SPARE, strlen(VDEV_TYPE_SPARE)) == 0 ||
2345 strncmp(name,
2346 VDEV_TYPE_REPLACING, strlen(VDEV_TYPE_REPLACING)) == 0 ||
2347 strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0)
2348 return (B_TRUE);
2349 return (B_FALSE);
2350 }
2351
2352 nvlist_t *
2353 zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare,
2354 boolean_t *l2cache, boolean_t *log)
2355 {
2356 char buf[MAXPATHLEN];
2357 char *end;
2358 nvlist_t *nvroot, *search, *ret;
2359 uint64_t guid;
2360
2361 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2362
2363 guid = strtoull(path, &end, 10);
2364 if (guid != 0 && *end == '\0') {
2365 verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0);
2366 } else if (zpool_vdev_is_interior(path)) {
2367 verify(nvlist_add_string(search, ZPOOL_CONFIG_TYPE, path) == 0);
2368 } else if (path[0] != '/') {
2369 (void) snprintf(buf, sizeof (buf), "%s/%s", ZFS_DISK_ROOT,
2370 path);
2371 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, buf) == 0);
2372 } else {
2373 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0);
2374 }
2375
2376 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
2377 &nvroot) == 0);
2378
2379 *avail_spare = B_FALSE;
2380 *l2cache = B_FALSE;
2381 if (log != NULL)
2382 *log = B_FALSE;
2383 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
2384 nvlist_free(search);
2385
2386 return (ret);
2387 }
2388
2389 static int
2390 vdev_online(nvlist_t *nv)
2391 {
2392 uint64_t ival;
2393
2394 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 ||
2395 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 ||
2396 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0)
2397 return (0);
2398
2399 return (1);
2400 }
2401
2402 /*
2403 * Helper function for zpool_get_physpaths().
2404 */
2405 static int
2406 vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size,
2407 size_t *bytes_written)
2408 {
2409 size_t bytes_left, pos, rsz;
2410 char *tmppath;
2411 const char *format;
2412
2413 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH,
2414 &tmppath) != 0)
2415 return (EZFS_NODEVICE);
2416
2417 pos = *bytes_written;
2418 bytes_left = physpath_size - pos;
2419 format = (pos == 0) ? "%s" : " %s";
2420
2421 rsz = snprintf(physpath + pos, bytes_left, format, tmppath);
2422 *bytes_written += rsz;
2423
2424 if (rsz >= bytes_left) {
2425 /* if physpath was not copied properly, clear it */
2426 if (bytes_left != 0) {
2427 physpath[pos] = 0;
2428 }
2429 return (EZFS_NOSPC);
2430 }
2431 return (0);
2432 }
2433
2434 static int
2435 vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size,
2436 size_t *rsz, boolean_t is_spare)
2437 {
2438 char *type;
2439 int ret;
2440
2441 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0)
2442 return (EZFS_INVALCONFIG);
2443
2444 if (strcmp(type, VDEV_TYPE_DISK) == 0) {
2445 /*
2446 * An active spare device has ZPOOL_CONFIG_IS_SPARE set.
2447 * For a spare vdev, we only want to boot from the active
2448 * spare device.
2449 */
2450 if (is_spare) {
2451 uint64_t spare = 0;
2452 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE,
2453 &spare);
2454 if (!spare)
2455 return (EZFS_INVALCONFIG);
2456 }
2457
2458 if (vdev_online(nv)) {
2459 if ((ret = vdev_get_one_physpath(nv, physpath,
2460 phypath_size, rsz)) != 0)
2461 return (ret);
2462 }
2463 } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 ||
2464 strcmp(type, VDEV_TYPE_RAIDZ) == 0 ||
2465 strcmp(type, VDEV_TYPE_REPLACING) == 0 ||
2466 (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) {
2467 nvlist_t **child;
2468 uint_t count;
2469 int i, ret;
2470
2471 if (nvlist_lookup_nvlist_array(nv,
2472 ZPOOL_CONFIG_CHILDREN, &child, &count) != 0)
2473 return (EZFS_INVALCONFIG);
2474
2475 for (i = 0; i < count; i++) {
2476 ret = vdev_get_physpaths(child[i], physpath,
2477 phypath_size, rsz, is_spare);
2478 if (ret == EZFS_NOSPC)
2479 return (ret);
2480 }
2481 }
2482
2483 return (EZFS_POOL_INVALARG);
2484 }
2485
2486 /*
2487 * Get phys_path for a root pool config.
2488 * Return 0 on success; non-zero on failure.
2489 */
2490 static int
2491 zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size)
2492 {
2493 size_t rsz;
2494 nvlist_t *vdev_root;
2495 nvlist_t **child;
2496 uint_t count;
2497 char *type;
2498
2499 rsz = 0;
2500
2501 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
2502 &vdev_root) != 0)
2503 return (EZFS_INVALCONFIG);
2504
2505 if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 ||
2506 nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN,
2507 &child, &count) != 0)
2508 return (EZFS_INVALCONFIG);
2509
2510 /*
2511 * root pool can only have a single top-level vdev.
2512 */
2513 if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1)
2514 return (EZFS_POOL_INVALARG);
2515
2516 (void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz,
2517 B_FALSE);
2518
2519 /* No online devices */
2520 if (rsz == 0)
2521 return (EZFS_NODEVICE);
2522
2523 return (0);
2524 }
2525
2526 /*
2527 * Get phys_path for a root pool
2528 * Return 0 on success; non-zero on failure.
2529 */
2530 int
2531 zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size)
2532 {
2533 return (zpool_get_config_physpath(zhp->zpool_config, physpath,
2534 phypath_size));
2535 }
2536
2537 /*
2538 * If the device has being dynamically expanded then we need to relabel
2539 * the disk to use the new unallocated space.
2540 */
2541 static int
2542 zpool_relabel_disk(libzfs_handle_t *hdl, const char *name)
2543 {
2544 char path[MAXPATHLEN];
2545 char errbuf[1024];
2546 int fd, error;
2547 int (*_efi_use_whole_disk)(int);
2548
2549 if ((_efi_use_whole_disk = (int (*)(int))dlsym(RTLD_DEFAULT,
2550 "efi_use_whole_disk")) == NULL)
2551 return (-1);
2552
2553 (void) snprintf(path, sizeof (path), "%s/%s", ZFS_RDISK_ROOT, name);
2554
2555 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) {
2556 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2557 "relabel '%s': unable to open device"), name);
2558 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
2559 }
2560
2561 /*
2562 * It's possible that we might encounter an error if the device
2563 * does not have any unallocated space left. If so, we simply
2564 * ignore that error and continue on.
2565 */
2566 error = _efi_use_whole_disk(fd);
2567 (void) close(fd);
2568 if (error && error != VT_ENOSPC) {
2569 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2570 "relabel '%s': unable to read disk capacity"), name);
2571 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
2572 }
2573 return (0);
2574 }
2575
2576 /*
2577 * Bring the specified vdev online. The 'flags' parameter is a set of the
2578 * ZFS_ONLINE_* flags.
2579 */
2580 int
2581 zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags,
2582 vdev_state_t *newstate)
2583 {
2584 zfs_cmd_t zc = { 0 };
2585 char msg[1024];
2586 char *pathname;
2587 nvlist_t *tgt;
2588 boolean_t avail_spare, l2cache, islog;
2589 libzfs_handle_t *hdl = zhp->zpool_hdl;
2590
2591 if (flags & ZFS_ONLINE_EXPAND) {
2592 (void) snprintf(msg, sizeof (msg),
2593 dgettext(TEXT_DOMAIN, "cannot expand %s"), path);
2594 } else {
2595 (void) snprintf(msg, sizeof (msg),
2596 dgettext(TEXT_DOMAIN, "cannot online %s"), path);
2597 }
2598
2599 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2600 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2601 &islog)) == NULL)
2602 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2603
2604 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2605
2606 if (avail_spare)
2607 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2608
2609 if ((flags & ZFS_ONLINE_EXPAND ||
2610 zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) &&
2611 nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH, &pathname) == 0) {
2612 uint64_t wholedisk = 0;
2613
2614 (void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK,
2615 &wholedisk);
2616
2617 /*
2618 * XXX - L2ARC 1.0 devices can't support expansion.
2619 */
2620 if (l2cache) {
2621 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2622 "cannot expand cache devices"));
2623 return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg));
2624 }
2625
2626 if (wholedisk) {
2627 pathname += strlen(ZFS_DISK_ROOT) + 1;
2628 (void) zpool_relabel_disk(hdl, pathname);
2629 }
2630 }
2631
2632 zc.zc_cookie = VDEV_STATE_ONLINE;
2633 zc.zc_obj = flags;
2634
2635 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) {
2636 if (errno == EINVAL) {
2637 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split "
2638 "from this pool into a new one. Use '%s' "
2639 "instead"), "zpool detach");
2640 return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, msg));
2641 }
2642 return (zpool_standard_error(hdl, errno, msg));
2643 }
2644
2645 *newstate = zc.zc_cookie;
2646 return (0);
2647 }
2648
2649 /*
2650 * Take the specified vdev offline
2651 */
2652 int
2653 zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp)
2654 {
2655 zfs_cmd_t zc = { 0 };
2656 char msg[1024];
2657 nvlist_t *tgt;
2658 boolean_t avail_spare, l2cache;
2659 libzfs_handle_t *hdl = zhp->zpool_hdl;
2660
2661 (void) snprintf(msg, sizeof (msg),
2662 dgettext(TEXT_DOMAIN, "cannot offline %s"), path);
2663
2664 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2665 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2666 NULL)) == NULL)
2667 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2668
2669 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2670
2671 if (avail_spare)
2672 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2673
2674 zc.zc_cookie = VDEV_STATE_OFFLINE;
2675 zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0;
2676
2677 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2678 return (0);
2679
2680 switch (errno) {
2681 case EBUSY:
2682
2683 /*
2684 * There are no other replicas of this device.
2685 */
2686 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2687
2688 case EEXIST:
2689 /*
2690 * The log device has unplayed logs
2691 */
2692 return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg));
2693
2694 default:
2695 return (zpool_standard_error(hdl, errno, msg));
2696 }
2697 }
2698
2699 /*
2700 * Mark the given vdev faulted.
2701 */
2702 int
2703 zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2704 {
2705 zfs_cmd_t zc = { 0 };
2706 char msg[1024];
2707 libzfs_handle_t *hdl = zhp->zpool_hdl;
2708
2709 (void) snprintf(msg, sizeof (msg),
2710 dgettext(TEXT_DOMAIN, "cannot fault %llu"), guid);
2711
2712 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2713 zc.zc_guid = guid;
2714 zc.zc_cookie = VDEV_STATE_FAULTED;
2715 zc.zc_obj = aux;
2716
2717 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2718 return (0);
2719
2720 switch (errno) {
2721 case EBUSY:
2722
2723 /*
2724 * There are no other replicas of this device.
2725 */
2726 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2727
2728 default:
2729 return (zpool_standard_error(hdl, errno, msg));
2730 }
2731
2732 }
2733
2734 /*
2735 * Mark the given vdev degraded.
2736 */
2737 int
2738 zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2739 {
2740 zfs_cmd_t zc = { 0 };
2741 char msg[1024];
2742 libzfs_handle_t *hdl = zhp->zpool_hdl;
2743
2744 (void) snprintf(msg, sizeof (msg),
2745 dgettext(TEXT_DOMAIN, "cannot degrade %llu"), guid);
2746
2747 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2748 zc.zc_guid = guid;
2749 zc.zc_cookie = VDEV_STATE_DEGRADED;
2750 zc.zc_obj = aux;
2751
2752 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2753 return (0);
2754
2755 return (zpool_standard_error(hdl, errno, msg));
2756 }
2757
2758 /*
2759 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as
2760 * a hot spare.
2761 */
2762 static boolean_t
2763 is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which)
2764 {
2765 nvlist_t **child;
2766 uint_t c, children;
2767 char *type;
2768
2769 if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child,
2770 &children) == 0) {
2771 verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE,
2772 &type) == 0);
2773
2774 if (strcmp(type, VDEV_TYPE_SPARE) == 0 &&
2775 children == 2 && child[which] == tgt)
2776 return (B_TRUE);
2777
2778 for (c = 0; c < children; c++)
2779 if (is_replacing_spare(child[c], tgt, which))
2780 return (B_TRUE);
2781 }
2782
2783 return (B_FALSE);
2784 }
2785
2786 /*
2787 * Attach new_disk (fully described by nvroot) to old_disk.
2788 * If 'replacing' is specified, the new disk will replace the old one.
2789 */
2790 int
2791 zpool_vdev_attach(zpool_handle_t *zhp,
2792 const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing)
2793 {
2794 zfs_cmd_t zc = { 0 };
2795 char msg[1024];
2796 int ret;
2797 nvlist_t *tgt;
2798 boolean_t avail_spare, l2cache, islog;
2799 uint64_t val;
2800 char *newname;
2801 nvlist_t **child;
2802 uint_t children;
2803 nvlist_t *config_root;
2804 libzfs_handle_t *hdl = zhp->zpool_hdl;
2805 boolean_t rootpool = zpool_is_bootable(zhp);
2806
2807 if (replacing)
2808 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2809 "cannot replace %s with %s"), old_disk, new_disk);
2810 else
2811 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2812 "cannot attach %s to %s"), new_disk, old_disk);
2813
2814 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2815 if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache,
2816 &islog)) == NULL)
2817 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2818
2819 if (avail_spare)
2820 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2821
2822 if (l2cache)
2823 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2824
2825 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2826 zc.zc_cookie = replacing;
2827
2828 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
2829 &child, &children) != 0 || children != 1) {
2830 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2831 "new device must be a single disk"));
2832 return (zfs_error(hdl, EZFS_INVALCONFIG, msg));
2833 }
2834
2835 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
2836 ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0);
2837
2838 if ((newname = zpool_vdev_name(NULL, NULL, child[0], B_FALSE)) == NULL)
2839 return (-1);
2840
2841 /*
2842 * If the target is a hot spare that has been swapped in, we can only
2843 * replace it with another hot spare.
2844 */
2845 if (replacing &&
2846 nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 &&
2847 (zpool_find_vdev(zhp, newname, &avail_spare, &l2cache,
2848 NULL) == NULL || !avail_spare) &&
2849 is_replacing_spare(config_root, tgt, 1)) {
2850 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2851 "can only be replaced by another hot spare"));
2852 free(newname);
2853 return (zfs_error(hdl, EZFS_BADTARGET, msg));
2854 }
2855
2856 free(newname);
2857
2858 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
2859 return (-1);
2860
2861 ret = zfs_ioctl(hdl, ZFS_IOC_VDEV_ATTACH, &zc);
2862
2863 zcmd_free_nvlists(&zc);
2864
2865 if (ret == 0) {
2866 if (rootpool) {
2867 /*
2868 * XXX need a better way to prevent user from
2869 * booting up a half-baked vdev.
2870 */
2871 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make "
2872 "sure to wait until resilver is done "
2873 "before rebooting.\n"));
2874 }
2875 return (0);
2876 }
2877
2878 switch (errno) {
2879 case ENOTSUP:
2880 /*
2881 * Can't attach to or replace this type of vdev.
2882 */
2883 if (replacing) {
2884 uint64_t version = zpool_get_prop_int(zhp,
2885 ZPOOL_PROP_VERSION, NULL);
2886
2887 if (islog)
2888 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2889 "cannot replace a log with a spare"));
2890 else if (version >= SPA_VERSION_MULTI_REPLACE)
2891 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2892 "already in replacing/spare config; wait "
2893 "for completion or use 'zpool detach'"));
2894 else
2895 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2896 "cannot replace a replacing device"));
2897 } else {
2898 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2899 "can only attach to mirrors and top-level "
2900 "disks"));
2901 }
2902 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
2903 break;
2904
2905 case EINVAL:
2906 /*
2907 * The new device must be a single disk.
2908 */
2909 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2910 "new device must be a single disk"));
2911 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg);
2912 break;
2913
2914 case EBUSY:
2915 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy, "
2916 "or device removal is in progress"),
2917 new_disk);
2918 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2919 break;
2920
2921 case EOVERFLOW:
2922 /*
2923 * The new device is too small.
2924 */
2925 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2926 "device is too small"));
2927 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2928 break;
2929
2930 case EDOM:
2931 /*
2932 * The new device has a different alignment requirement.
2933 */
2934 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2935 "devices have different sector alignment"));
2936 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2937 break;
2938
2939 case ENAMETOOLONG:
2940 /*
2941 * The resulting top-level vdev spec won't fit in the label.
2942 */
2943 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg);
2944 break;
2945
2946 default:
2947 (void) zpool_standard_error(hdl, errno, msg);
2948 }
2949
2950 return (-1);
2951 }
2952
2953 /*
2954 * Detach the specified device.
2955 */
2956 int
2957 zpool_vdev_detach(zpool_handle_t *zhp, const char *path)
2958 {
2959 zfs_cmd_t zc = { 0 };
2960 char msg[1024];
2961 nvlist_t *tgt;
2962 boolean_t avail_spare, l2cache;
2963 libzfs_handle_t *hdl = zhp->zpool_hdl;
2964
2965 (void) snprintf(msg, sizeof (msg),
2966 dgettext(TEXT_DOMAIN, "cannot detach %s"), path);
2967
2968 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2969 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2970 NULL)) == NULL)
2971 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2972
2973 if (avail_spare)
2974 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2975
2976 if (l2cache)
2977 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2978
2979 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2980
2981 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0)
2982 return (0);
2983
2984 switch (errno) {
2985
2986 case ENOTSUP:
2987 /*
2988 * Can't detach from this type of vdev.
2989 */
2990 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only "
2991 "applicable to mirror and replacing vdevs"));
2992 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
2993 break;
2994
2995 case EBUSY:
2996 /*
2997 * There are no other replicas of this device.
2998 */
2999 (void) zfs_error(hdl, EZFS_NOREPLICAS, msg);
3000 break;
3001
3002 default:
3003 (void) zpool_standard_error(hdl, errno, msg);
3004 }
3005
3006 return (-1);
3007 }
3008
3009 /*
3010 * Find a mirror vdev in the source nvlist.
3011 *
3012 * The mchild array contains a list of disks in one of the top-level mirrors
3013 * of the source pool. The schild array contains a list of disks that the
3014 * user specified on the command line. We loop over the mchild array to
3015 * see if any entry in the schild array matches.
3016 *
3017 * If a disk in the mchild array is found in the schild array, we return
3018 * the index of that entry. Otherwise we return -1.
3019 */
3020 static int
3021 find_vdev_entry(zpool_handle_t *zhp, nvlist_t **mchild, uint_t mchildren,
3022 nvlist_t **schild, uint_t schildren)
3023 {
3024 uint_t mc;
3025
3026 for (mc = 0; mc < mchildren; mc++) {
3027 uint_t sc;
3028 char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp,
3029 mchild[mc], B_FALSE);
3030
3031 for (sc = 0; sc < schildren; sc++) {
3032 char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp,
3033 schild[sc], B_FALSE);
3034 boolean_t result = (strcmp(mpath, spath) == 0);
3035
3036 free(spath);
3037 if (result) {
3038 free(mpath);
3039 return (mc);
3040 }
3041 }
3042
3043 free(mpath);
3044 }
3045
3046 return (-1);
3047 }
3048
3049 /*
3050 * Split a mirror pool. If newroot points to null, then a new nvlist
3051 * is generated and it is the responsibility of the caller to free it.
3052 */
3053 int
3054 zpool_vdev_split(zpool_handle_t *zhp, char *newname, nvlist_t **newroot,
3055 nvlist_t *props, splitflags_t flags)
3056 {
3057 zfs_cmd_t zc = { 0 };
3058 char msg[1024];
3059 nvlist_t *tree, *config, **child, **newchild, *newconfig = NULL;
3060 nvlist_t **varray = NULL, *zc_props = NULL;
3061 uint_t c, children, newchildren, lastlog = 0, vcount, found = 0;
3062 libzfs_handle_t *hdl = zhp->zpool_hdl;
3063 uint64_t vers;
3064 boolean_t freelist = B_FALSE, memory_err = B_TRUE;
3065 int retval = 0;
3066
3067 (void) snprintf(msg, sizeof (msg),
3068 dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name);
3069
3070 if (!zpool_name_valid(hdl, B_FALSE, newname))
3071 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
3072
3073 if ((config = zpool_get_config(zhp, NULL)) == NULL) {
3074 (void) fprintf(stderr, gettext("Internal error: unable to "
3075 "retrieve pool configuration\n"));
3076 return (-1);
3077 }
3078
3079 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree)
3080 == 0);
3081 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &vers) == 0);
3082
3083 if (props) {
3084 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
3085 if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name,
3086 props, vers, flags, msg)) == NULL)
3087 return (-1);
3088 }
3089
3090 if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child,
3091 &children) != 0) {
3092 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3093 "Source pool is missing vdev tree"));
3094 nvlist_free(zc_props);
3095 return (-1);
3096 }
3097
3098 varray = zfs_alloc(hdl, children * sizeof (nvlist_t *));
3099 vcount = 0;
3100
3101 if (*newroot == NULL ||
3102 nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN,
3103 &newchild, &newchildren) != 0)
3104 newchildren = 0;
3105
3106 for (c = 0; c < children; c++) {
3107 uint64_t is_log = B_FALSE, is_hole = B_FALSE;
3108 char *type;
3109 nvlist_t **mchild, *vdev;
3110 uint_t mchildren;
3111 int entry;
3112
3113 /*
3114 * Unlike cache & spares, slogs are stored in the
3115 * ZPOOL_CONFIG_CHILDREN array. We filter them out here.
3116 */
3117 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
3118 &is_log);
3119 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
3120 &is_hole);
3121 if (is_log || is_hole) {
3122 /*
3123 * Create a hole vdev and put it in the config.
3124 */
3125 if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0)
3126 goto out;
3127 if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE,
3128 VDEV_TYPE_HOLE) != 0)
3129 goto out;
3130 if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE,
3131 1) != 0)
3132 goto out;
3133 if (lastlog == 0)
3134 lastlog = vcount;
3135 varray[vcount++] = vdev;
3136 continue;
3137 }
3138 lastlog = 0;
3139 verify(nvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE, &type)
3140 == 0);
3141 if (strcmp(type, VDEV_TYPE_MIRROR) != 0) {
3142 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3143 "Source pool must be composed only of mirrors\n"));
3144 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
3145 goto out;
3146 }
3147
3148 verify(nvlist_lookup_nvlist_array(child[c],
3149 ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0);
3150
3151 /* find or add an entry for this top-level vdev */
3152 if (newchildren > 0 &&
3153 (entry = find_vdev_entry(zhp, mchild, mchildren,
3154 newchild, newchildren)) >= 0) {
3155 /* We found a disk that the user specified. */
3156 vdev = mchild[entry];
3157 ++found;
3158 } else {
3159 /* User didn't specify a disk for this vdev. */
3160 vdev = mchild[mchildren - 1];
3161 }
3162
3163 if (nvlist_dup(vdev, &varray[vcount++], 0) != 0)
3164 goto out;
3165 }
3166
3167 /* did we find every disk the user specified? */
3168 if (found != newchildren) {
3169 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must "
3170 "include at most one disk from each mirror"));
3171 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
3172 goto out;
3173 }
3174
3175 /* Prepare the nvlist for populating. */
3176 if (*newroot == NULL) {
3177 if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0)
3178 goto out;
3179 freelist = B_TRUE;
3180 if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE,
3181 VDEV_TYPE_ROOT) != 0)
3182 goto out;
3183 } else {
3184 verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0);
3185 }
3186
3187 /* Add all the children we found */
3188 if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, varray,
3189 lastlog == 0 ? vcount : lastlog) != 0)
3190 goto out;
3191
3192 /*
3193 * If we're just doing a dry run, exit now with success.
3194 */
3195 if (flags.dryrun) {
3196 memory_err = B_FALSE;
3197 freelist = B_FALSE;
3198 goto out;
3199 }
3200
3201 /* now build up the config list & call the ioctl */
3202 if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0)
3203 goto out;
3204
3205 if (nvlist_add_nvlist(newconfig,
3206 ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 ||
3207 nvlist_add_string(newconfig,
3208 ZPOOL_CONFIG_POOL_NAME, newname) != 0 ||
3209 nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0)
3210 goto out;
3211
3212 /*
3213 * The new pool is automatically part of the namespace unless we
3214 * explicitly export it.
3215 */
3216 if (!flags.import)
3217 zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT;
3218 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3219 (void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string));
3220 if (zcmd_write_conf_nvlist(hdl, &zc, newconfig) != 0)
3221 goto out;
3222 if (zc_props != NULL && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
3223 goto out;
3224
3225 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) {
3226 retval = zpool_standard_error(hdl, errno, msg);
3227 goto out;
3228 }
3229
3230 freelist = B_FALSE;
3231 memory_err = B_FALSE;
3232
3233 out:
3234 if (varray != NULL) {
3235 int v;
3236
3237 for (v = 0; v < vcount; v++)
3238 nvlist_free(varray[v]);
3239 free(varray);
3240 }
3241 zcmd_free_nvlists(&zc);
3242 nvlist_free(zc_props);
3243 nvlist_free(newconfig);
3244 if (freelist) {
3245 nvlist_free(*newroot);
3246 *newroot = NULL;
3247 }
3248
3249 if (retval != 0)
3250 return (retval);
3251
3252 if (memory_err)
3253 return (no_memory(hdl));
3254
3255 return (0);
3256 }
3257
3258 /*
3259 * Remove the given device.
3260 */
3261 int
3262 zpool_vdev_remove(zpool_handle_t *zhp, const char *path)
3263 {
3264 zfs_cmd_t zc = { 0 };
3265 char msg[1024];
3266 nvlist_t *tgt;
3267 boolean_t avail_spare, l2cache, islog;
3268 libzfs_handle_t *hdl = zhp->zpool_hdl;
3269 uint64_t version;
3270
3271 (void) snprintf(msg, sizeof (msg),
3272 dgettext(TEXT_DOMAIN, "cannot remove %s"), path);
3273
3274 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3275 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
3276 &islog)) == NULL)
3277 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3278
3279 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
3280 if (islog && version < SPA_VERSION_HOLES) {
3281 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3282 "pool must be upgraded to support log removal"));
3283 return (zfs_error(hdl, EZFS_BADVERSION, msg));
3284 }
3285
3286 if (!islog && !avail_spare && !l2cache && zpool_is_bootable(zhp)) {
3287 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3288 "root pool can not have removed devices, "
3289 "because GRUB does not understand them"));
3290 return (zfs_error(hdl, EINVAL, msg));
3291 }
3292
3293 zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
3294
3295 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
3296 return (0);
3297
3298 switch (errno) {
3299
3300 case EINVAL:
3301 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3302 "invalid config; all top-level vdevs must "
3303 "have the same sector size and not be raidz."));
3304 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg);
3305 break;
3306
3307 case EBUSY:
3308 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3309 "Pool busy; removal may already be in progress"));
3310 (void) zfs_error(hdl, EZFS_BUSY, msg);
3311 break;
3312
3313 default:
3314 (void) zpool_standard_error(hdl, errno, msg);
3315 }
3316 return (-1);
3317 }
3318
3319 int
3320 zpool_vdev_remove_cancel(zpool_handle_t *zhp)
3321 {
3322 zfs_cmd_t zc = { 0 };
3323 char msg[1024];
3324 libzfs_handle_t *hdl = zhp->zpool_hdl;
3325
3326 (void) snprintf(msg, sizeof (msg),
3327 dgettext(TEXT_DOMAIN, "cannot cancel removal"));
3328
3329 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3330 zc.zc_cookie = 1;
3331
3332 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
3333 return (0);
3334
3335 return (zpool_standard_error(hdl, errno, msg));
3336 }
3337
3338 int
3339 zpool_vdev_indirect_size(zpool_handle_t *zhp, const char *path,
3340 uint64_t *sizep)
3341 {
3342 char msg[1024];
3343 nvlist_t *tgt;
3344 boolean_t avail_spare, l2cache, islog;
3345 libzfs_handle_t *hdl = zhp->zpool_hdl;
3346
3347 (void) snprintf(msg, sizeof (msg),
3348 dgettext(TEXT_DOMAIN, "cannot determine indirect size of %s"),
3349 path);
3350
3351 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
3352 &islog)) == NULL)
3353 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3354
3355 if (avail_spare || l2cache || islog) {
3356 *sizep = 0;
3357 return (0);
3358 }
3359
3360 if (nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_INDIRECT_SIZE, sizep) != 0) {
3361 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3362 "indirect size not available"));
3363 return (zfs_error(hdl, EINVAL, msg));
3364 }
3365 return (0);
3366 }
3367
3368 /*
3369 * Clear the errors for the pool, or the particular device if specified.
3370 */
3371 int
3372 zpool_clear(zpool_handle_t *zhp, const char *path, nvlist_t *rewindnvl)
3373 {
3374 zfs_cmd_t zc = { 0 };
3375 char msg[1024];
3376 nvlist_t *tgt;
3377 zpool_load_policy_t policy;
3378 boolean_t avail_spare, l2cache;
3379 libzfs_handle_t *hdl = zhp->zpool_hdl;
3380 nvlist_t *nvi = NULL;
3381 int error;
3382
3383 if (path)
3384 (void) snprintf(msg, sizeof (msg),
3385 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
3386 path);
3387 else
3388 (void) snprintf(msg, sizeof (msg),
3389 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
3390 zhp->zpool_name);
3391
3392 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3393 if (path) {
3394 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare,
3395 &l2cache, NULL)) == NULL)
3396 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3397
3398 /*
3399 * Don't allow error clearing for hot spares. Do allow
3400 * error clearing for l2cache devices.
3401 */
3402 if (avail_spare)
3403 return (zfs_error(hdl, EZFS_ISSPARE, msg));
3404
3405 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID,
3406 &zc.zc_guid) == 0);
3407 }
3408
3409 zpool_get_load_policy(rewindnvl, &policy);
3410 zc.zc_cookie = policy.zlp_rewind;
3411
3412 if (zcmd_alloc_dst_nvlist(hdl, &zc, zhp->zpool_config_size * 2) != 0)
3413 return (-1);
3414
3415 if (zcmd_write_src_nvlist(hdl, &zc, rewindnvl) != 0)
3416 return (-1);
3417
3418 while ((error = zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc)) != 0 &&
3419 errno == ENOMEM) {
3420 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
3421 zcmd_free_nvlists(&zc);
3422 return (-1);
3423 }
3424 }
3425
3426 if (!error || ((policy.zlp_rewind & ZPOOL_TRY_REWIND) &&
3427 errno != EPERM && errno != EACCES)) {
3428 if (policy.zlp_rewind &
3429 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
3430 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi);
3431 zpool_rewind_exclaim(hdl, zc.zc_name,
3432 ((policy.zlp_rewind & ZPOOL_TRY_REWIND) != 0),
3433 nvi);
3434 nvlist_free(nvi);
3435 }
3436 zcmd_free_nvlists(&zc);
3437 return (0);
3438 }
3439
3440 zcmd_free_nvlists(&zc);
3441 return (zpool_standard_error(hdl, errno, msg));
3442 }
3443
3444 /*
3445 * Similar to zpool_clear(), but takes a GUID (used by fmd).
3446 */
3447 int
3448 zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid)
3449 {
3450 zfs_cmd_t zc = { 0 };
3451 char msg[1024];
3452 libzfs_handle_t *hdl = zhp->zpool_hdl;
3453
3454 (void) snprintf(msg, sizeof (msg),
3455 dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"),
3456 guid);
3457
3458 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3459 zc.zc_guid = guid;
3460 zc.zc_cookie = ZPOOL_NO_REWIND;
3461
3462 if (ioctl(hdl->libzfs_fd, ZFS_IOC_CLEAR, &zc) == 0)
3463 return (0);
3464
3465 return (zpool_standard_error(hdl, errno, msg));
3466 }
3467
3468 /*
3469 * Change the GUID for a pool.
3470 */
3471 int
3472 zpool_reguid(zpool_handle_t *zhp)
3473 {
3474 char msg[1024];
3475 libzfs_handle_t *hdl = zhp->zpool_hdl;
3476 zfs_cmd_t zc = { 0 };
3477
3478 (void) snprintf(msg, sizeof (msg),
3479 dgettext(TEXT_DOMAIN, "cannot reguid '%s'"), zhp->zpool_name);
3480
3481 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3482 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REGUID, &zc) == 0)
3483 return (0);
3484
3485 return (zpool_standard_error(hdl, errno, msg));
3486 }
3487
3488 /*
3489 * Reopen the pool.
3490 */
3491 int
3492 zpool_reopen(zpool_handle_t *zhp)
3493 {
3494 zfs_cmd_t zc = { 0 };
3495 char msg[1024];
3496 libzfs_handle_t *hdl = zhp->zpool_hdl;
3497
3498 (void) snprintf(msg, sizeof (msg),
3499 dgettext(TEXT_DOMAIN, "cannot reopen '%s'"),
3500 zhp->zpool_name);
3501
3502 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3503 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REOPEN, &zc) == 0)
3504 return (0);
3505 return (zpool_standard_error(hdl, errno, msg));
3506 }
3507
3508 /*
3509 * Convert from a devid string to a path.
3510 */
3511 static char *
3512 devid_to_path(char *devid_str)
3513 {
3514 ddi_devid_t devid;
3515 char *minor;
3516 char *path;
3517 devid_nmlist_t *list = NULL;
3518 int ret;
3519
3520 if (devid_str_decode(devid_str, &devid, &minor) != 0)
3521 return (NULL);
3522
3523 ret = devid_deviceid_to_nmlist("/dev", devid, minor, &list);
3524
3525 devid_str_free(minor);
3526 devid_free(devid);
3527
3528 if (ret != 0)
3529 return (NULL);
3530
3531 /*
3532 * In a case the strdup() fails, we will just return NULL below.
3533 */
3534 path = strdup(list[0].devname);
3535
3536 devid_free_nmlist(list);
3537
3538 return (path);
3539 }
3540
3541 /*
3542 * Convert from a path to a devid string.
3543 */
3544 static char *
3545 path_to_devid(const char *path)
3546 {
3547 int fd;
3548 ddi_devid_t devid;
3549 char *minor, *ret;
3550
3551 if ((fd = open(path, O_RDONLY)) < 0)
3552 return (NULL);
3553
3554 minor = NULL;
3555 ret = NULL;
3556 if (devid_get(fd, &devid) == 0) {
3557 if (devid_get_minor_name(fd, &minor) == 0)
3558 ret = devid_str_encode(devid, minor);
3559 if (minor != NULL)
3560 devid_str_free(minor);
3561 devid_free(devid);
3562 }
3563 (void) close(fd);
3564
3565 return (ret);
3566 }
3567
3568 /*
3569 * Issue the necessary ioctl() to update the stored path value for the vdev. We
3570 * ignore any failure here, since a common case is for an unprivileged user to
3571 * type 'zpool status', and we'll display the correct information anyway.
3572 */
3573 static void
3574 set_path(zpool_handle_t *zhp, nvlist_t *nv, const char *path)
3575 {
3576 zfs_cmd_t zc = { 0 };
3577
3578 (void) strncpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3579 (void) strncpy(zc.zc_value, path, sizeof (zc.zc_value));
3580 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
3581 &zc.zc_guid) == 0);
3582
3583 (void) ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SETPATH, &zc);
3584 }
3585
3586 /*
3587 * Given a vdev, return the name to display in iostat. If the vdev has a path,
3588 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
3589 * We also check if this is a whole disk, in which case we strip off the
3590 * trailing 's0' slice name.
3591 *
3592 * This routine is also responsible for identifying when disks have been
3593 * reconfigured in a new location. The kernel will have opened the device by
3594 * devid, but the path will still refer to the old location. To catch this, we
3595 * first do a path -> devid translation (which is fast for the common case). If
3596 * the devid matches, we're done. If not, we do a reverse devid -> path
3597 * translation and issue the appropriate ioctl() to update the path of the vdev.
3598 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
3599 * of these checks.
3600 */
3601 char *
3602 zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv,
3603 boolean_t verbose)
3604 {
3605 char *path, *devid;
3606 uint64_t value;
3607 char buf[64];
3608 vdev_stat_t *vs;
3609 uint_t vsc;
3610
3611 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
3612 &value) == 0) {
3613 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
3614 &value) == 0);
3615 (void) snprintf(buf, sizeof (buf), "%llu",
3616 (u_longlong_t)value);
3617 path = buf;
3618 } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) {
3619
3620 /*
3621 * If the device is dead (faulted, offline, etc) then don't
3622 * bother opening it. Otherwise we may be forcing the user to
3623 * open a misbehaving device, which can have undesirable
3624 * effects.
3625 */
3626 if ((nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
3627 (uint64_t **)&vs, &vsc) != 0 ||
3628 vs->vs_state >= VDEV_STATE_DEGRADED) &&
3629 zhp != NULL &&
3630 nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) {
3631 /*
3632 * Determine if the current path is correct.
3633 */
3634 char *newdevid = path_to_devid(path);
3635
3636 if (newdevid == NULL ||
3637 strcmp(devid, newdevid) != 0) {
3638 char *newpath;
3639
3640 if ((newpath = devid_to_path(devid)) != NULL) {
3641 /*
3642 * Update the path appropriately.
3643 */
3644 set_path(zhp, nv, newpath);
3645 if (nvlist_add_string(nv,
3646 ZPOOL_CONFIG_PATH, newpath) == 0)
3647 verify(nvlist_lookup_string(nv,
3648 ZPOOL_CONFIG_PATH,
3649 &path) == 0);
3650 free(newpath);
3651 }
3652 }
3653
3654 if (newdevid)
3655 devid_str_free(newdevid);
3656 }
3657
3658 if (strncmp(path, ZFS_DISK_ROOTD, strlen(ZFS_DISK_ROOTD)) == 0)
3659 path += strlen(ZFS_DISK_ROOTD);
3660
3661 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
3662 &value) == 0 && value) {
3663 int pathlen = strlen(path);
3664 char *tmp = zfs_strdup(hdl, path);
3665
3666 /*
3667 * If it starts with c#, and ends with "s0" or "s1",
3668 * chop the slice off, or if it ends with "s0/old" or
3669 * "s1/old", remove the slice from the middle.
3670 */
3671 if (CTD_CHECK(tmp)) {
3672 if (strcmp(&tmp[pathlen - 2], "s0") == 0 ||
3673 strcmp(&tmp[pathlen - 2], "s1") == 0) {
3674 tmp[pathlen - 2] = '\0';
3675 } else if (pathlen > 6 &&
3676 (strcmp(&tmp[pathlen - 6], "s0/old") == 0 ||
3677 strcmp(&tmp[pathlen - 6], "s1/old") == 0)) {
3678 (void) strcpy(&tmp[pathlen - 6],
3679 "/old");
3680 }
3681 }
3682 return (tmp);
3683 }
3684 } else {
3685 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &path) == 0);
3686
3687 /*
3688 * If it's a raidz device, we need to stick in the parity level.
3689 */
3690 if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) {
3691 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY,
3692 &value) == 0);
3693 (void) snprintf(buf, sizeof (buf), "%s%llu", path,
3694 (u_longlong_t)value);
3695 path = buf;
3696 }
3697
3698 /*
3699 * We identify each top-level vdev by using a <type-id>
3700 * naming convention.
3701 */
3702 if (verbose) {
3703 uint64_t id;
3704
3705 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
3706 &id) == 0);
3707 (void) snprintf(buf, sizeof (buf), "%s-%llu", path,
3708 (u_longlong_t)id);
3709 path = buf;
3710 }
3711 }
3712
3713 return (zfs_strdup(hdl, path));
3714 }
3715
3716 static int
3717 zbookmark_mem_compare(const void *a, const void *b)
3718 {
3719 return (memcmp(a, b, sizeof (zbookmark_phys_t)));
3720 }
3721
3722 /*
3723 * Retrieve the persistent error log, uniquify the members, and return to the
3724 * caller.
3725 */
3726 int
3727 zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp)
3728 {
3729 zfs_cmd_t zc = { 0 };
3730 uint64_t count;
3731 zbookmark_phys_t *zb = NULL;
3732 int i;
3733
3734 /*
3735 * Retrieve the raw error list from the kernel. If the number of errors
3736 * has increased, allocate more space and continue until we get the
3737 * entire list.
3738 */
3739 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT,
3740 &count) == 0);
3741 if (count == 0)
3742 return (0);
3743 if ((zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl,
3744 count * sizeof (zbookmark_phys_t))) == (uintptr_t)NULL)
3745 return (-1);
3746 zc.zc_nvlist_dst_size = count;
3747 (void) strcpy(zc.zc_name, zhp->zpool_name);
3748 for (;;) {
3749 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_ERROR_LOG,
3750 &zc) != 0) {
3751 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3752 if (errno == ENOMEM) {
3753 void *dst;
3754
3755 count = zc.zc_nvlist_dst_size;
3756 dst = zfs_alloc(zhp->zpool_hdl, count *
3757 sizeof (zbookmark_phys_t));
3758 if (dst == NULL)
3759 return (-1);
3760 zc.zc_nvlist_dst = (uintptr_t)dst;
3761 } else {
3762 return (-1);
3763 }
3764 } else {
3765 break;
3766 }
3767 }
3768
3769 /*
3770 * Sort the resulting bookmarks. This is a little confusing due to the
3771 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last
3772 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks
3773 * _not_ copied as part of the process. So we point the start of our
3774 * array appropriate and decrement the total number of elements.
3775 */
3776 zb = ((zbookmark_phys_t *)(uintptr_t)zc.zc_nvlist_dst) +
3777 zc.zc_nvlist_dst_size;
3778 count -= zc.zc_nvlist_dst_size;
3779
3780 qsort(zb, count, sizeof (zbookmark_phys_t), zbookmark_mem_compare);
3781
3782 verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0);
3783
3784 /*
3785 * Fill in the nverrlistp with nvlist's of dataset and object numbers.
3786 */
3787 for (i = 0; i < count; i++) {
3788 nvlist_t *nv;
3789
3790 /* ignoring zb_blkid and zb_level for now */
3791 if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset &&
3792 zb[i-1].zb_object == zb[i].zb_object)
3793 continue;
3794
3795 if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0)
3796 goto nomem;
3797 if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET,
3798 zb[i].zb_objset) != 0) {
3799 nvlist_free(nv);
3800 goto nomem;
3801 }
3802 if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT,
3803 zb[i].zb_object) != 0) {
3804 nvlist_free(nv);
3805 goto nomem;
3806 }
3807 if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) {
3808 nvlist_free(nv);
3809 goto nomem;
3810 }
3811 nvlist_free(nv);
3812 }
3813
3814 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3815 return (0);
3816
3817 nomem:
3818 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3819 return (no_memory(zhp->zpool_hdl));
3820 }
3821
3822 /*
3823 * Upgrade a ZFS pool to the latest on-disk version.
3824 */
3825 int
3826 zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version)
3827 {
3828 zfs_cmd_t zc = { 0 };
3829 libzfs_handle_t *hdl = zhp->zpool_hdl;
3830
3831 (void) strcpy(zc.zc_name, zhp->zpool_name);
3832 zc.zc_cookie = new_version;
3833
3834 if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0)
3835 return (zpool_standard_error_fmt(hdl, errno,
3836 dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"),
3837 zhp->zpool_name));
3838 return (0);
3839 }
3840
3841 void
3842 zfs_save_arguments(int argc, char **argv, char *string, int len)
3843 {
3844 (void) strlcpy(string, basename(argv[0]), len);
3845 for (int i = 1; i < argc; i++) {
3846 (void) strlcat(string, " ", len);
3847 (void) strlcat(string, argv[i], len);
3848 }
3849 }
3850
3851 int
3852 zpool_log_history(libzfs_handle_t *hdl, const char *message)
3853 {
3854 zfs_cmd_t zc = { 0 };
3855 nvlist_t *args;
3856 int err;
3857
3858 args = fnvlist_alloc();
3859 fnvlist_add_string(args, "message", message);
3860 err = zcmd_write_src_nvlist(hdl, &zc, args);
3861 if (err == 0)
3862 err = ioctl(hdl->libzfs_fd, ZFS_IOC_LOG_HISTORY, &zc);
3863 nvlist_free(args);
3864 zcmd_free_nvlists(&zc);
3865 return (err);
3866 }
3867
3868 /*
3869 * Perform ioctl to get some command history of a pool.
3870 *
3871 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the
3872 * logical offset of the history buffer to start reading from.
3873 *
3874 * Upon return, 'off' is the next logical offset to read from and
3875 * 'len' is the actual amount of bytes read into 'buf'.
3876 */
3877 static int
3878 get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len)
3879 {
3880 zfs_cmd_t zc = { 0 };
3881 libzfs_handle_t *hdl = zhp->zpool_hdl;
3882
3883 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3884
3885 zc.zc_history = (uint64_t)(uintptr_t)buf;
3886 zc.zc_history_len = *len;
3887 zc.zc_history_offset = *off;
3888
3889 if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) {
3890 switch (errno) {
3891 case EPERM:
3892 return (zfs_error_fmt(hdl, EZFS_PERM,
3893 dgettext(TEXT_DOMAIN,
3894 "cannot show history for pool '%s'"),
3895 zhp->zpool_name));
3896 case ENOENT:
3897 return (zfs_error_fmt(hdl, EZFS_NOHISTORY,
3898 dgettext(TEXT_DOMAIN, "cannot get history for pool "
3899 "'%s'"), zhp->zpool_name));
3900 case ENOTSUP:
3901 return (zfs_error_fmt(hdl, EZFS_BADVERSION,
3902 dgettext(TEXT_DOMAIN, "cannot get history for pool "
3903 "'%s', pool must be upgraded"), zhp->zpool_name));
3904 default:
3905 return (zpool_standard_error_fmt(hdl, errno,
3906 dgettext(TEXT_DOMAIN,
3907 "cannot get history for '%s'"), zhp->zpool_name));
3908 }
3909 }
3910
3911 *len = zc.zc_history_len;
3912 *off = zc.zc_history_offset;
3913
3914 return (0);
3915 }
3916
3917 /*
3918 * Process the buffer of nvlists, unpacking and storing each nvlist record
3919 * into 'records'. 'leftover' is set to the number of bytes that weren't
3920 * processed as there wasn't a complete record.
3921 */
3922 int
3923 zpool_history_unpack(char *buf, uint64_t bytes_read, uint64_t *leftover,
3924 nvlist_t ***records, uint_t *numrecords)
3925 {
3926 uint64_t reclen;
3927 nvlist_t *nv;
3928 int i;
3929
3930 while (bytes_read > sizeof (reclen)) {
3931
3932 /* get length of packed record (stored as little endian) */
3933 for (i = 0, reclen = 0; i < sizeof (reclen); i++)
3934 reclen += (uint64_t)(((uchar_t *)buf)[i]) << (8*i);
3935
3936 if (bytes_read < sizeof (reclen) + reclen)
3937 break;
3938
3939 /* unpack record */
3940 if (nvlist_unpack(buf + sizeof (reclen), reclen, &nv, 0) != 0)
3941 return (ENOMEM);
3942 bytes_read -= sizeof (reclen) + reclen;
3943 buf += sizeof (reclen) + reclen;
3944
3945 /* add record to nvlist array */
3946 (*numrecords)++;
3947 if (ISP2(*numrecords + 1)) {
3948 *records = realloc(*records,
3949 *numrecords * 2 * sizeof (nvlist_t *));
3950 }
3951 (*records)[*numrecords - 1] = nv;
3952 }
3953
3954 *leftover = bytes_read;
3955 return (0);
3956 }
3957
3958 /*
3959 * Retrieve the command history of a pool.
3960 */
3961 int
3962 zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp)
3963 {
3964 char *buf;
3965 int buflen = 128 * 1024;
3966 uint64_t off = 0;
3967 nvlist_t **records = NULL;
3968 uint_t numrecords = 0;
3969 int err, i;
3970
3971 buf = malloc(buflen);
3972 if (buf == NULL)
3973 return (ENOMEM);
3974 do {
3975 uint64_t bytes_read = buflen;
3976 uint64_t leftover;
3977
3978 if ((err = get_history(zhp, buf, &off, &bytes_read)) != 0)
3979 break;
3980
3981 /* if nothing else was read in, we're at EOF, just return */
3982 if (!bytes_read)
3983 break;
3984
3985 if ((err = zpool_history_unpack(buf, bytes_read,
3986 &leftover, &records, &numrecords)) != 0)
3987 break;
3988 off -= leftover;
3989 if (leftover == bytes_read) {
3990 /*
3991 * no progress made, because buffer is not big enough
3992 * to hold this record; resize and retry.
3993 */
3994 buflen *= 2;
3995 free(buf);
3996 buf = malloc(buflen);
3997 if (buf == NULL)
3998 return (ENOMEM);
3999 }
4000
4001 /* CONSTCOND */
4002 } while (1);
4003
4004 free(buf);
4005
4006 if (!err) {
4007 verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0);
4008 verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD,
4009 records, numrecords) == 0);
4010 }
4011 for (i = 0; i < numrecords; i++)
4012 nvlist_free(records[i]);
4013 free(records);
4014
4015 return (err);
4016 }
4017
4018 void
4019 zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
4020 char *pathname, size_t len)
4021 {
4022 zfs_cmd_t zc = { 0 };
4023 boolean_t mounted = B_FALSE;
4024 char *mntpnt = NULL;
4025 char dsname[ZFS_MAX_DATASET_NAME_LEN];
4026
4027 if (dsobj == 0) {
4028 /* special case for the MOS */
4029 (void) snprintf(pathname, len, "<metadata>:<0x%llx>", obj);
4030 return;
4031 }
4032
4033 /* get the dataset's name */
4034 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4035 zc.zc_obj = dsobj;
4036 if (ioctl(zhp->zpool_hdl->libzfs_fd,
4037 ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) {
4038 /* just write out a path of two object numbers */
4039 (void) snprintf(pathname, len, "<0x%llx>:<0x%llx>",
4040 dsobj, obj);
4041 return;
4042 }
4043 (void) strlcpy(dsname, zc.zc_value, sizeof (dsname));
4044
4045 /* find out if the dataset is mounted */
4046 mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt);
4047
4048 /* get the corrupted object's path */
4049 (void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name));
4050 zc.zc_obj = obj;
4051 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_OBJ_TO_PATH,
4052 &zc) == 0) {
4053 if (mounted) {
4054 (void) snprintf(pathname, len, "%s%s", mntpnt,
4055 zc.zc_value);
4056 } else {
4057 (void) snprintf(pathname, len, "%s:%s",
4058 dsname, zc.zc_value);
4059 }
4060 } else {
4061 (void) snprintf(pathname, len, "%s:<0x%llx>", dsname, obj);
4062 }
4063 free(mntpnt);
4064 }
4065
4066 /*
4067 * Read the EFI label from the config, if a label does not exist then
4068 * pass back the error to the caller. If the caller has passed a non-NULL
4069 * diskaddr argument then we set it to the starting address of the EFI
4070 * partition. If the caller has passed a non-NULL boolean argument, then
4071 * we set it to indicate if the disk does have efi system partition.
4072 */
4073 static int
4074 read_efi_label(nvlist_t *config, diskaddr_t *sb, boolean_t *system)
4075 {
4076 char *path;
4077 int fd;
4078 char diskname[MAXPATHLEN];
4079 boolean_t boot = B_FALSE;
4080 int err = -1;
4081 int slice;
4082
4083 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PATH, &path) != 0)
4084 return (err);
4085
4086 (void) snprintf(diskname, sizeof (diskname), "%s%s", ZFS_RDISK_ROOT,
4087 strrchr(path, '/'));
4088 if ((fd = open(diskname, O_RDONLY|O_NDELAY)) >= 0) {
4089 struct dk_gpt *vtoc;
4090
4091 if ((err = efi_alloc_and_read(fd, &vtoc)) >= 0) {
4092 for (slice = 0; slice < vtoc->efi_nparts; slice++) {
4093 if (vtoc->efi_parts[slice].p_tag == V_SYSTEM)
4094 boot = B_TRUE;
4095 if (vtoc->efi_parts[slice].p_tag == V_USR)
4096 break;
4097 }
4098 if (sb != NULL && vtoc->efi_parts[slice].p_tag == V_USR)
4099 *sb = vtoc->efi_parts[slice].p_start;
4100 if (system != NULL)
4101 *system = boot;
4102 efi_free(vtoc);
4103 }
4104 (void) close(fd);
4105 }
4106 return (err);
4107 }
4108
4109 /*
4110 * determine where a partition starts on a disk in the current
4111 * configuration
4112 */
4113 static diskaddr_t
4114 find_start_block(nvlist_t *config)
4115 {
4116 nvlist_t **child;
4117 uint_t c, children;
4118 diskaddr_t sb = MAXOFFSET_T;
4119 uint64_t wholedisk;
4120
4121 if (nvlist_lookup_nvlist_array(config,
4122 ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) {
4123 if (nvlist_lookup_uint64(config,
4124 ZPOOL_CONFIG_WHOLE_DISK,
4125 &wholedisk) != 0 || !wholedisk) {
4126 return (MAXOFFSET_T);
4127 }
4128 if (read_efi_label(config, &sb, NULL) < 0)
4129 sb = MAXOFFSET_T;
4130 return (sb);
4131 }
4132
4133 for (c = 0; c < children; c++) {
4134 sb = find_start_block(child[c]);
4135 if (sb != MAXOFFSET_T) {
4136 return (sb);
4137 }
4138 }
4139 return (MAXOFFSET_T);
4140 }
4141
4142 /*
4143 * Label an individual disk. The name provided is the short name,
4144 * stripped of any leading /dev path.
4145 */
4146 int
4147 zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, const char *name,
4148 zpool_boot_label_t boot_type, uint64_t boot_size, int *slice)
4149 {
4150 char path[MAXPATHLEN];
4151 struct dk_gpt *vtoc;
4152 int fd;
4153 size_t resv = EFI_MIN_RESV_SIZE;
4154 uint64_t slice_size;
4155 diskaddr_t start_block;
4156 char errbuf[1024];
4157
4158 /* prepare an error message just in case */
4159 (void) snprintf(errbuf, sizeof (errbuf),
4160 dgettext(TEXT_DOMAIN, "cannot label '%s'"), name);
4161
4162 if (zhp) {
4163 nvlist_t *nvroot;
4164
4165 verify(nvlist_lookup_nvlist(zhp->zpool_config,
4166 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
4167
4168 if (zhp->zpool_start_block == 0)
4169 start_block = find_start_block(nvroot);
4170 else
4171 start_block = zhp->zpool_start_block;
4172 zhp->zpool_start_block = start_block;
4173 } else {
4174 /* new pool */
4175 start_block = NEW_START_BLOCK;
4176 }
4177
4178 (void) snprintf(path, sizeof (path), "%s/%s%s", ZFS_RDISK_ROOT, name,
4179 BACKUP_SLICE);
4180
4181 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) {
4182 /*
4183 * This shouldn't happen. We've long since verified that this
4184 * is a valid device.
4185 */
4186 zfs_error_aux(hdl,
4187 dgettext(TEXT_DOMAIN, "unable to open device"));
4188 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
4189 }
4190
4191 if (efi_alloc_and_init(fd, EFI_NUMPAR, &vtoc) != 0) {
4192 /*
4193 * The only way this can fail is if we run out of memory, or we
4194 * were unable to read the disk's capacity
4195 */
4196 if (errno == ENOMEM)
4197 (void) no_memory(hdl);
4198
4199 (void) close(fd);
4200 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4201 "unable to read disk capacity"), name);
4202
4203 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
4204 }
4205
4206 /*
4207 * Why we use V_USR: V_BACKUP confuses users, and is considered
4208 * disposable by some EFI utilities (since EFI doesn't have a backup
4209 * slice). V_UNASSIGNED is supposed to be used only for zero size
4210 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT,
4211 * etc. were all pretty specific. V_USR is as close to reality as we
4212 * can get, in the absence of V_OTHER.
4213 */
4214 /* first fix the partition start block */
4215 if (start_block == MAXOFFSET_T)
4216 start_block = NEW_START_BLOCK;
4217
4218 /*
4219 * EFI System partition is using slice 0.
4220 * ZFS is on slice 1 and slice 8 is reserved.
4221 * We assume the GPT partition table without system
4222 * partition has zfs p_start == NEW_START_BLOCK.
4223 * If start_block != NEW_START_BLOCK, it means we have
4224 * system partition. Correct solution would be to query/cache vtoc
4225 * from existing vdev member.
4226 */
4227 if (boot_type == ZPOOL_CREATE_BOOT_LABEL) {
4228 if (boot_size % vtoc->efi_lbasize != 0) {
4229 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4230 "boot partition size must be a multiple of %d"),
4231 vtoc->efi_lbasize);
4232 (void) close(fd);
4233 efi_free(vtoc);
4234 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
4235 }
4236 /*
4237 * System partition size checks.
4238 * Note the 1MB is quite arbitrary value, since we
4239 * are creating dedicated pool, it should be enough
4240 * to hold fat + efi bootloader. May need to be
4241 * adjusted if the bootloader size will grow.
4242 */
4243 if (boot_size < 1024 * 1024) {
4244 char buf[64];
4245 zfs_nicenum(boot_size, buf, sizeof (buf));
4246 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4247 "Specified size %s for EFI System partition is too "
4248 "small, the minimum size is 1MB."), buf);
4249 (void) close(fd);
4250 efi_free(vtoc);
4251 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
4252 }
4253 /* 33MB is tested with mkfs -F pcfs */
4254 if (hdl->libzfs_printerr &&
4255 ((vtoc->efi_lbasize == 512 &&
4256 boot_size < 33 * 1024 * 1024) ||
4257 (vtoc->efi_lbasize == 4096 &&
4258 boot_size < 256 * 1024 * 1024))) {
4259 char buf[64];
4260 zfs_nicenum(boot_size, buf, sizeof (buf));
4261 (void) fprintf(stderr, dgettext(TEXT_DOMAIN,
4262 "Warning: EFI System partition size %s is "
4263 "not allowing to create FAT32 file\nsystem, which "
4264 "may result in unbootable system.\n"), buf);
4265 }
4266 /* Adjust zfs partition start by size of system partition. */
4267 start_block += boot_size / vtoc->efi_lbasize;
4268 }
4269
4270 if (start_block == NEW_START_BLOCK) {
4271 /*
4272 * Use default layout.
4273 * ZFS is on slice 0 and slice 8 is reserved.
4274 */
4275 slice_size = vtoc->efi_last_u_lba + 1;
4276 slice_size -= EFI_MIN_RESV_SIZE;
4277 slice_size -= start_block;
4278 if (slice != NULL)
4279 *slice = 0;
4280
4281 vtoc->efi_parts[0].p_start = start_block;
4282 vtoc->efi_parts[0].p_size = slice_size;
4283
4284 vtoc->efi_parts[0].p_tag = V_USR;
4285 (void) strcpy(vtoc->efi_parts[0].p_name, "zfs");
4286
4287 vtoc->efi_parts[8].p_start = slice_size + start_block;
4288 vtoc->efi_parts[8].p_size = resv;
4289 vtoc->efi_parts[8].p_tag = V_RESERVED;
4290 } else {
4291 slice_size = start_block - NEW_START_BLOCK;
4292 vtoc->efi_parts[0].p_start = NEW_START_BLOCK;
4293 vtoc->efi_parts[0].p_size = slice_size;
4294 vtoc->efi_parts[0].p_tag = V_SYSTEM;
4295 (void) strcpy(vtoc->efi_parts[0].p_name, "loader");
4296 if (slice != NULL)
4297 *slice = 1;
4298 /* prepare slice 1 */
4299 slice_size = vtoc->efi_last_u_lba + 1 - slice_size;
4300 slice_size -= resv;
4301 slice_size -= NEW_START_BLOCK;
4302 vtoc->efi_parts[1].p_start = start_block;
4303 vtoc->efi_parts[1].p_size = slice_size;
4304 vtoc->efi_parts[1].p_tag = V_USR;
4305 (void) strcpy(vtoc->efi_parts[1].p_name, "zfs");
4306
4307 vtoc->efi_parts[8].p_start = slice_size + start_block;
4308 vtoc->efi_parts[8].p_size = resv;
4309 vtoc->efi_parts[8].p_tag = V_RESERVED;
4310 }
4311
4312 if (efi_write(fd, vtoc) != 0) {
4313 /*
4314 * Some block drivers (like pcata) may not support EFI
4315 * GPT labels. Print out a helpful error message dir-
4316 * ecting the user to manually label the disk and give
4317 * a specific slice.
4318 */
4319 (void) close(fd);
4320 efi_free(vtoc);
4321
4322 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4323 "try using fdisk(1M) and then provide a specific slice"));
4324 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
4325 }
4326
4327 (void) close(fd);
4328 efi_free(vtoc);
4329 return (0);
4330 }
4331
4332 static boolean_t
4333 supported_dump_vdev_type(libzfs_handle_t *hdl, nvlist_t *config, char *errbuf)
4334 {
4335 char *type;
4336 nvlist_t **child;
4337 uint_t children, c;
4338
4339 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_TYPE, &type) == 0);
4340 if (strcmp(type, VDEV_TYPE_FILE) == 0 ||
4341 strcmp(type, VDEV_TYPE_HOLE) == 0 ||
4342 strcmp(type, VDEV_TYPE_MISSING) == 0) {
4343 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4344 "vdev type '%s' is not supported"), type);
4345 (void) zfs_error(hdl, EZFS_VDEVNOTSUP, errbuf);
4346 return (B_FALSE);
4347 }
4348 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN,
4349 &child, &children) == 0) {
4350 for (c = 0; c < children; c++) {
4351 if (!supported_dump_vdev_type(hdl, child[c], errbuf))
4352 return (B_FALSE);
4353 }
4354 }
4355 return (B_TRUE);
4356 }
4357
4358 /*
4359 * Check if this zvol is allowable for use as a dump device; zero if
4360 * it is, > 0 if it isn't, < 0 if it isn't a zvol.
4361 *
4362 * Allowable storage configurations include mirrors, all raidz variants, and
4363 * pools with log, cache, and spare devices. Pools which are backed by files or
4364 * have missing/hole vdevs are not suitable.
4365 */
4366 int
4367 zvol_check_dump_config(char *arg)
4368 {
4369 zpool_handle_t *zhp = NULL;
4370 nvlist_t *config, *nvroot;
4371 char *p, *volname;
4372 nvlist_t **top;
4373 uint_t toplevels;
4374 libzfs_handle_t *hdl;
4375 char errbuf[1024];
4376 char poolname[ZFS_MAX_DATASET_NAME_LEN];
4377 int pathlen = strlen(ZVOL_FULL_DEV_DIR);
4378 int ret = 1;
4379
4380 if (strncmp(arg, ZVOL_FULL_DEV_DIR, pathlen)) {
4381 return (-1);
4382 }
4383
4384 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
4385 "dump is not supported on device '%s'"), arg);
4386
4387 if ((hdl = libzfs_init()) == NULL)
4388 return (1);
4389 libzfs_print_on_error(hdl, B_TRUE);
4390
4391 volname = arg + pathlen;
4392
4393 /* check the configuration of the pool */
4394 if ((p = strchr(volname, '/')) == NULL) {
4395 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4396 "malformed dataset name"));
4397 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
4398 return (1);
4399 } else if (p - volname >= ZFS_MAX_DATASET_NAME_LEN) {
4400 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4401 "dataset name is too long"));
4402 (void) zfs_error(hdl, EZFS_NAMETOOLONG, errbuf);
4403 return (1);
4404 } else {
4405 (void) strncpy(poolname, volname, p - volname);
4406 poolname[p - volname] = '\0';
4407 }
4408
4409 if ((zhp = zpool_open(hdl, poolname)) == NULL) {
4410 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4411 "could not open pool '%s'"), poolname);
4412 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
4413 goto out;
4414 }
4415 config = zpool_get_config(zhp, NULL);
4416 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
4417 &nvroot) != 0) {
4418 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4419 "could not obtain vdev configuration for '%s'"), poolname);
4420 (void) zfs_error(hdl, EZFS_INVALCONFIG, errbuf);
4421 goto out;
4422 }
4423
4424 verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
4425 &top, &toplevels) == 0);
4426
4427 if (!supported_dump_vdev_type(hdl, top[0], errbuf)) {
4428 goto out;
4429 }
4430 ret = 0;
4431
4432 out:
4433 if (zhp)
4434 zpool_close(zhp);
4435 libzfs_fini(hdl);
4436 return (ret);
4437 }