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, 2020 by Delphix. All rights reserved.
25 * Copyright 2020 Joyent, Inc.
26 * Copyright 2016 Nexenta Systems, Inc.
27 * Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>
28 * Copyright (c) 2017 Datto Inc.
29 * Copyright (c) 2017, Intel Corporation.
30 * Copyright 2022 OmniOS Community Edition (OmniOSce) Association.
31 * Copyright 2022 Oxide Computer Company
32 */
33
34 #include <ctype.h>
35 #include <errno.h>
36 #include <devid.h>
37 #include <fcntl.h>
38 #include <libintl.h>
39 #include <stdio.h>
40 #include <stdlib.h>
41 #include <strings.h>
42 #include <unistd.h>
43 #include <libgen.h>
44 #include <sys/dkio.h>
45 #include <sys/efi_partition.h>
46 #include <sys/vtoc.h>
47 #include <sys/zfs_ioctl.h>
48 #include <sys/modctl.h>
49 #include <sys/mkdev.h>
50 #include <dlfcn.h>
51 #include <libzutil.h>
52
53 #include "zfs_namecheck.h"
54 #include "zfs_prop.h"
55 #include "libzfs_impl.h"
56 #include "zfs_comutil.h"
57 #include "zfeature_common.h"
58
59 static int read_efi_label(nvlist_t *, diskaddr_t *, boolean_t *);
60 static boolean_t zpool_vdev_is_interior(const char *name);
61
62 #define BACKUP_SLICE "s2"
63
64 typedef struct prop_flags {
65 int create:1; /* Validate property on creation */
66 int import:1; /* Validate property on import */
67 } prop_flags_t;
68
69 /*
70 * ====================================================================
71 * zpool property functions
72 * ====================================================================
73 */
74
75 static int
zpool_get_all_props(zpool_handle_t * zhp)76 zpool_get_all_props(zpool_handle_t *zhp)
77 {
78 zfs_cmd_t zc = { 0 };
79 libzfs_handle_t *hdl = zhp->zpool_hdl;
80
81 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
82
83 if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0)
84 return (-1);
85
86 while (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_PROPS, &zc) != 0) {
87 if (errno == ENOMEM) {
88 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
89 zcmd_free_nvlists(&zc);
90 return (-1);
91 }
92 } else {
93 zcmd_free_nvlists(&zc);
94 return (-1);
95 }
96 }
97
98 if (zcmd_read_dst_nvlist(hdl, &zc, &zhp->zpool_props) != 0) {
99 zcmd_free_nvlists(&zc);
100 return (-1);
101 }
102
103 zcmd_free_nvlists(&zc);
104
105 return (0);
106 }
107
108 static int
zpool_props_refresh(zpool_handle_t * zhp)109 zpool_props_refresh(zpool_handle_t *zhp)
110 {
111 nvlist_t *old_props;
112
113 old_props = zhp->zpool_props;
114
115 if (zpool_get_all_props(zhp) != 0)
116 return (-1);
117
118 nvlist_free(old_props);
119 return (0);
120 }
121
122 static char *
zpool_get_prop_string(zpool_handle_t * zhp,zpool_prop_t prop,zprop_source_t * src)123 zpool_get_prop_string(zpool_handle_t *zhp, zpool_prop_t prop,
124 zprop_source_t *src)
125 {
126 nvlist_t *nv, *nvl;
127 uint64_t ival;
128 char *value;
129 zprop_source_t source;
130
131 nvl = zhp->zpool_props;
132 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
133 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &ival) == 0);
134 source = ival;
135 verify(nvlist_lookup_string(nv, ZPROP_VALUE, &value) == 0);
136 } else {
137 source = ZPROP_SRC_DEFAULT;
138 if ((value = (char *)zpool_prop_default_string(prop)) == NULL)
139 value = "-";
140 }
141
142 if (src)
143 *src = source;
144
145 return (value);
146 }
147
148 uint64_t
zpool_get_prop_int(zpool_handle_t * zhp,zpool_prop_t prop,zprop_source_t * src)149 zpool_get_prop_int(zpool_handle_t *zhp, zpool_prop_t prop, zprop_source_t *src)
150 {
151 nvlist_t *nv, *nvl;
152 uint64_t value;
153 zprop_source_t source;
154
155 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp)) {
156 /*
157 * zpool_get_all_props() has most likely failed because
158 * the pool is faulted, but if all we need is the top level
159 * vdev's guid then get it from the zhp config nvlist.
160 */
161 if ((prop == ZPOOL_PROP_GUID) &&
162 (nvlist_lookup_nvlist(zhp->zpool_config,
163 ZPOOL_CONFIG_VDEV_TREE, &nv) == 0) &&
164 (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value)
165 == 0)) {
166 return (value);
167 }
168 return (zpool_prop_default_numeric(prop));
169 }
170
171 nvl = zhp->zpool_props;
172 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
173 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &value) == 0);
174 source = value;
175 verify(nvlist_lookup_uint64(nv, ZPROP_VALUE, &value) == 0);
176 } else {
177 source = ZPROP_SRC_DEFAULT;
178 value = zpool_prop_default_numeric(prop);
179 }
180
181 if (src)
182 *src = source;
183
184 return (value);
185 }
186
187 /*
188 * Map VDEV STATE to printed strings.
189 */
190 const char *
zpool_state_to_name(vdev_state_t state,vdev_aux_t aux)191 zpool_state_to_name(vdev_state_t state, vdev_aux_t aux)
192 {
193 switch (state) {
194 case VDEV_STATE_CLOSED:
195 case VDEV_STATE_OFFLINE:
196 return (gettext("OFFLINE"));
197 case VDEV_STATE_REMOVED:
198 return (gettext("REMOVED"));
199 case VDEV_STATE_CANT_OPEN:
200 if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG)
201 return (gettext("FAULTED"));
202 else if (aux == VDEV_AUX_SPLIT_POOL)
203 return (gettext("SPLIT"));
204 else
205 return (gettext("UNAVAIL"));
206 case VDEV_STATE_FAULTED:
207 return (gettext("FAULTED"));
208 case VDEV_STATE_DEGRADED:
209 return (gettext("DEGRADED"));
210 case VDEV_STATE_HEALTHY:
211 return (gettext("ONLINE"));
212
213 default:
214 break;
215 }
216
217 return (gettext("UNKNOWN"));
218 }
219
220 /*
221 * Map POOL STATE to printed strings.
222 */
223 const char *
zpool_pool_state_to_name(pool_state_t state)224 zpool_pool_state_to_name(pool_state_t state)
225 {
226 switch (state) {
227 case POOL_STATE_ACTIVE:
228 return (gettext("ACTIVE"));
229 case POOL_STATE_EXPORTED:
230 return (gettext("EXPORTED"));
231 case POOL_STATE_DESTROYED:
232 return (gettext("DESTROYED"));
233 case POOL_STATE_SPARE:
234 return (gettext("SPARE"));
235 case POOL_STATE_L2CACHE:
236 return (gettext("L2CACHE"));
237 case POOL_STATE_UNINITIALIZED:
238 return (gettext("UNINITIALIZED"));
239 case POOL_STATE_UNAVAIL:
240 return (gettext("UNAVAIL"));
241 case POOL_STATE_POTENTIALLY_ACTIVE:
242 return (gettext("POTENTIALLY_ACTIVE"));
243 }
244
245 return (gettext("UNKNOWN"));
246 }
247
248 /*
249 * Get a zpool property value for 'prop' and return the value in
250 * a pre-allocated buffer.
251 */
252 int
zpool_get_prop(zpool_handle_t * zhp,zpool_prop_t prop,char * buf,size_t len,zprop_source_t * srctype,boolean_t literal)253 zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf, size_t len,
254 zprop_source_t *srctype, boolean_t literal)
255 {
256 uint64_t intval;
257 const char *strval;
258 zprop_source_t src = ZPROP_SRC_NONE;
259 nvlist_t *nvroot;
260 vdev_stat_t *vs;
261 uint_t vsc;
262
263 if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
264 switch (prop) {
265 case ZPOOL_PROP_NAME:
266 (void) strlcpy(buf, zpool_get_name(zhp), len);
267 break;
268
269 case ZPOOL_PROP_HEALTH:
270 (void) strlcpy(buf, "FAULTED", len);
271 break;
272
273 case ZPOOL_PROP_GUID:
274 intval = zpool_get_prop_int(zhp, prop, &src);
275 (void) snprintf(buf, len, "%llu", intval);
276 break;
277
278 case ZPOOL_PROP_ALTROOT:
279 case ZPOOL_PROP_CACHEFILE:
280 case ZPOOL_PROP_COMMENT:
281 if (zhp->zpool_props != NULL ||
282 zpool_get_all_props(zhp) == 0) {
283 (void) strlcpy(buf,
284 zpool_get_prop_string(zhp, prop, &src),
285 len);
286 break;
287 }
288 /* FALLTHROUGH */
289 default:
290 (void) strlcpy(buf, "-", len);
291 break;
292 }
293
294 if (srctype != NULL)
295 *srctype = src;
296 return (0);
297 }
298
299 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) &&
300 prop != ZPOOL_PROP_NAME)
301 return (-1);
302
303 switch (zpool_prop_get_type(prop)) {
304 case PROP_TYPE_STRING:
305 (void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src),
306 len);
307 break;
308
309 case PROP_TYPE_NUMBER:
310 intval = zpool_get_prop_int(zhp, prop, &src);
311
312 switch (prop) {
313 case ZPOOL_PROP_SIZE:
314 case ZPOOL_PROP_ALLOCATED:
315 case ZPOOL_PROP_FREE:
316 case ZPOOL_PROP_FREEING:
317 case ZPOOL_PROP_LEAKED:
318 case ZPOOL_PROP_ASHIFT:
319 if (literal) {
320 (void) snprintf(buf, len, "%llu",
321 (u_longlong_t)intval);
322 } else {
323 (void) zfs_nicenum(intval, buf, len);
324 }
325 break;
326 case ZPOOL_PROP_BOOTSIZE:
327 case ZPOOL_PROP_EXPANDSZ:
328 case ZPOOL_PROP_CHECKPOINT:
329 if (intval == 0) {
330 (void) strlcpy(buf, "-", len);
331 } else if (literal) {
332 (void) snprintf(buf, len, "%llu",
333 (u_longlong_t)intval);
334 } else {
335 (void) zfs_nicebytes(intval, buf, len);
336 }
337 break;
338 case ZPOOL_PROP_CAPACITY:
339 if (literal) {
340 (void) snprintf(buf, len, "%llu",
341 (u_longlong_t)intval);
342 } else {
343 (void) snprintf(buf, len, "%llu%%",
344 (u_longlong_t)intval);
345 }
346 break;
347 case ZPOOL_PROP_FRAGMENTATION:
348 if (intval == UINT64_MAX) {
349 (void) strlcpy(buf, "-", len);
350 } else if (literal) {
351 (void) snprintf(buf, len, "%llu",
352 (u_longlong_t)intval);
353 } else {
354 (void) snprintf(buf, len, "%llu%%",
355 (u_longlong_t)intval);
356 }
357 break;
358 case ZPOOL_PROP_DEDUPRATIO:
359 if (literal)
360 (void) snprintf(buf, len, "%llu.%02llu",
361 (u_longlong_t)(intval / 100),
362 (u_longlong_t)(intval % 100));
363 else
364 (void) snprintf(buf, len, "%llu.%02llux",
365 (u_longlong_t)(intval / 100),
366 (u_longlong_t)(intval % 100));
367 break;
368 case ZPOOL_PROP_HEALTH:
369 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
370 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
371 verify(nvlist_lookup_uint64_array(nvroot,
372 ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc)
373 == 0);
374
375 (void) strlcpy(buf, zpool_state_to_name(intval,
376 vs->vs_aux), len);
377 break;
378 case ZPOOL_PROP_VERSION:
379 if (intval >= SPA_VERSION_FEATURES) {
380 (void) snprintf(buf, len, "-");
381 break;
382 }
383 /* FALLTHROUGH */
384 default:
385 (void) snprintf(buf, len, "%llu", intval);
386 }
387 break;
388
389 case PROP_TYPE_INDEX:
390 intval = zpool_get_prop_int(zhp, prop, &src);
391 if (zpool_prop_index_to_string(prop, intval, &strval)
392 != 0)
393 return (-1);
394 (void) strlcpy(buf, strval, len);
395 break;
396
397 default:
398 abort();
399 }
400
401 if (srctype)
402 *srctype = src;
403
404 return (0);
405 }
406
407 /*
408 * Check if the bootfs name has the same pool name as it is set to.
409 * Assuming bootfs is a valid dataset name.
410 */
411 static boolean_t
bootfs_name_valid(const char * pool,const char * bootfs)412 bootfs_name_valid(const char *pool, const char *bootfs)
413 {
414 int len = strlen(pool);
415 if (bootfs[0] == '\0')
416 return (B_TRUE);
417
418 if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT))
419 return (B_FALSE);
420
421 if (strncmp(pool, bootfs, len) == 0 &&
422 (bootfs[len] == '/' || bootfs[len] == '\0'))
423 return (B_TRUE);
424
425 return (B_FALSE);
426 }
427
428 boolean_t
zpool_is_bootable(zpool_handle_t * zhp)429 zpool_is_bootable(zpool_handle_t *zhp)
430 {
431 char bootfs[ZFS_MAX_DATASET_NAME_LEN];
432
433 return (zpool_get_prop(zhp, ZPOOL_PROP_BOOTFS, bootfs,
434 sizeof (bootfs), NULL, B_FALSE) == 0 && strncmp(bootfs, "-",
435 sizeof (bootfs)) != 0);
436 }
437
438
439 /*
440 * Given an nvlist of zpool properties to be set, validate that they are
441 * correct, and parse any numeric properties (index, boolean, etc) if they are
442 * specified as strings.
443 */
444 static nvlist_t *
zpool_valid_proplist(libzfs_handle_t * hdl,const char * poolname,nvlist_t * props,uint64_t version,prop_flags_t flags,char * errbuf)445 zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname,
446 nvlist_t *props, uint64_t version, prop_flags_t flags, char *errbuf)
447 {
448 nvpair_t *elem;
449 nvlist_t *retprops;
450 zpool_prop_t prop;
451 char *strval;
452 uint64_t intval;
453 char *slash, *check;
454 struct stat64 statbuf;
455 zpool_handle_t *zhp;
456
457 if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) {
458 (void) no_memory(hdl);
459 return (NULL);
460 }
461
462 elem = NULL;
463 while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
464 const char *propname = nvpair_name(elem);
465
466 prop = zpool_name_to_prop(propname);
467 if (prop == ZPOOL_PROP_INVAL && zpool_prop_feature(propname)) {
468 int err;
469 char *fname = strchr(propname, '@') + 1;
470
471 err = zfeature_lookup_name(fname, NULL);
472 if (err != 0) {
473 ASSERT3U(err, ==, ENOENT);
474 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
475 "invalid feature '%s', '%s'"), fname,
476 propname);
477 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
478 goto error;
479 }
480
481 if (nvpair_type(elem) != DATA_TYPE_STRING) {
482 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
483 "'%s' must be a string"), propname);
484 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
485 goto error;
486 }
487
488 (void) nvpair_value_string(elem, &strval);
489 if (strcmp(strval, ZFS_FEATURE_ENABLED) != 0 &&
490 strcmp(strval, ZFS_FEATURE_DISABLED) != 0) {
491 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
492 "property '%s' can only be set to "
493 "'enabled' or 'disabled'"), propname);
494 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
495 goto error;
496 }
497
498 if (nvlist_add_uint64(retprops, propname, 0) != 0) {
499 (void) no_memory(hdl);
500 goto error;
501 }
502 continue;
503 }
504
505 /*
506 * Make sure this property is valid and applies to this type.
507 */
508 if (prop == ZPOOL_PROP_INVAL) {
509 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
510 "invalid property '%s'"), propname);
511 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
512 goto error;
513 }
514
515 if (zpool_prop_readonly(prop)) {
516 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
517 "is readonly"), propname);
518 (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
519 goto error;
520 }
521
522 if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops,
523 &strval, &intval, errbuf) != 0)
524 goto error;
525
526 /*
527 * Perform additional checking for specific properties.
528 */
529 switch (prop) {
530 case ZPOOL_PROP_VERSION:
531 if (intval < version ||
532 !SPA_VERSION_IS_SUPPORTED(intval)) {
533 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
534 "property '%s' number %d is invalid."),
535 propname, intval);
536 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
537 goto error;
538 }
539 break;
540
541 case ZPOOL_PROP_BOOTSIZE:
542 if (!flags.create) {
543 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
544 "property '%s' can only be set during pool "
545 "creation"), propname);
546 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
547 goto error;
548 }
549 break;
550
551 case ZPOOL_PROP_ASHIFT:
552 if (intval != 0 &&
553 (intval < ASHIFT_MIN || intval > ASHIFT_MAX)) {
554 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
555 "invalid '%s=%d' property: only values "
556 "between %" PRId32 " and %" PRId32 " "
557 "are allowed.\n"),
558 propname, intval, ASHIFT_MIN, ASHIFT_MAX);
559 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
560 goto error;
561 }
562 break;
563
564 case ZPOOL_PROP_BOOTFS:
565 if (flags.create || flags.import) {
566 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
567 "property '%s' cannot be set at creation "
568 "or import time"), propname);
569 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
570 goto error;
571 }
572
573 if (version < SPA_VERSION_BOOTFS) {
574 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
575 "pool must be upgraded to support "
576 "'%s' property"), propname);
577 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
578 goto error;
579 }
580
581 /*
582 * bootfs property value has to be a dataset name and
583 * the dataset has to be in the same pool as it sets to.
584 */
585 if (!bootfs_name_valid(poolname, strval)) {
586 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
587 "is an invalid name"), strval);
588 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
589 goto error;
590 }
591
592 if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) {
593 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
594 "could not open pool '%s'"), poolname);
595 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
596 goto error;
597 }
598 zpool_close(zhp);
599 break;
600
601 case ZPOOL_PROP_ALTROOT:
602 if (!flags.create && !flags.import) {
603 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
604 "property '%s' can only be set during pool "
605 "creation or import"), propname);
606 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
607 goto error;
608 }
609
610 if (strval[0] != '/') {
611 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
612 "bad alternate root '%s'"), strval);
613 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
614 goto error;
615 }
616 break;
617
618 case ZPOOL_PROP_CACHEFILE:
619 if (strval[0] == '\0')
620 break;
621
622 if (strcmp(strval, "none") == 0)
623 break;
624
625 if (strval[0] != '/') {
626 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
627 "property '%s' must be empty, an "
628 "absolute path, or 'none'"), propname);
629 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
630 goto error;
631 }
632
633 slash = strrchr(strval, '/');
634
635 if (slash[1] == '\0' || strcmp(slash, "/.") == 0 ||
636 strcmp(slash, "/..") == 0) {
637 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
638 "'%s' is not a valid file"), strval);
639 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
640 goto error;
641 }
642
643 *slash = '\0';
644
645 if (strval[0] != '\0' &&
646 (stat64(strval, &statbuf) != 0 ||
647 !S_ISDIR(statbuf.st_mode))) {
648 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
649 "'%s' is not a valid directory"),
650 strval);
651 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
652 goto error;
653 }
654
655 *slash = '/';
656 break;
657
658 case ZPOOL_PROP_COMMENT:
659 for (check = strval; *check != '\0'; check++) {
660 if (!isprint(*check)) {
661 zfs_error_aux(hdl,
662 dgettext(TEXT_DOMAIN,
663 "comment may only have printable "
664 "characters"));
665 (void) zfs_error(hdl, EZFS_BADPROP,
666 errbuf);
667 goto error;
668 }
669 }
670 if (strlen(strval) > ZPROP_MAX_COMMENT) {
671 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
672 "comment must not exceed %d characters"),
673 ZPROP_MAX_COMMENT);
674 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
675 goto error;
676 }
677 break;
678
679 case ZPOOL_PROP_READONLY:
680 if (!flags.import) {
681 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
682 "property '%s' can only be set at "
683 "import time"), propname);
684 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
685 goto error;
686 }
687 break;
688
689 case ZPOOL_PROP_TNAME:
690 if (!flags.create) {
691 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
692 "property '%s' can only be set at "
693 "creation time"), propname);
694 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
695 goto error;
696 }
697 break;
698
699 case ZPOOL_PROP_MULTIHOST:
700 if (get_system_hostid() == 0) {
701 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
702 "requires a non-zero system hostid"));
703 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
704 goto error;
705 }
706 break;
707
708 default:
709 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
710 "property '%s'(%d) not defined"), propname, prop);
711 break;
712 }
713 }
714
715 return (retprops);
716 error:
717 nvlist_free(retprops);
718 return (NULL);
719 }
720
721 /*
722 * Set zpool property : propname=propval.
723 */
724 int
zpool_set_prop(zpool_handle_t * zhp,const char * propname,const char * propval)725 zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval)
726 {
727 zfs_cmd_t zc = { 0 };
728 int ret = -1;
729 char errbuf[1024];
730 nvlist_t *nvl = NULL;
731 nvlist_t *realprops;
732 uint64_t version;
733 prop_flags_t flags = { 0 };
734
735 (void) snprintf(errbuf, sizeof (errbuf),
736 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
737 zhp->zpool_name);
738
739 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
740 return (no_memory(zhp->zpool_hdl));
741
742 if (nvlist_add_string(nvl, propname, propval) != 0) {
743 nvlist_free(nvl);
744 return (no_memory(zhp->zpool_hdl));
745 }
746
747 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
748 if ((realprops = zpool_valid_proplist(zhp->zpool_hdl,
749 zhp->zpool_name, nvl, version, flags, errbuf)) == NULL) {
750 nvlist_free(nvl);
751 return (-1);
752 }
753
754 nvlist_free(nvl);
755 nvl = realprops;
756
757 /*
758 * Execute the corresponding ioctl() to set this property.
759 */
760 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
761
762 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) {
763 nvlist_free(nvl);
764 return (-1);
765 }
766
767 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc);
768
769 zcmd_free_nvlists(&zc);
770 nvlist_free(nvl);
771
772 if (ret)
773 (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
774 else
775 (void) zpool_props_refresh(zhp);
776
777 return (ret);
778 }
779
780 int
zpool_expand_proplist(zpool_handle_t * zhp,zprop_list_t ** plp)781 zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp)
782 {
783 libzfs_handle_t *hdl = zhp->zpool_hdl;
784 zprop_list_t *entry;
785 char buf[ZFS_MAXPROPLEN];
786 nvlist_t *features = NULL;
787 zprop_list_t **last;
788 boolean_t firstexpand = (NULL == *plp);
789
790 if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0)
791 return (-1);
792
793 last = plp;
794 while (*last != NULL)
795 last = &(*last)->pl_next;
796
797 if ((*plp)->pl_all)
798 features = zpool_get_features(zhp);
799
800 if ((*plp)->pl_all && firstexpand) {
801 for (int i = 0; i < SPA_FEATURES; i++) {
802 zprop_list_t *entry = zfs_alloc(hdl,
803 sizeof (zprop_list_t));
804 entry->pl_prop = ZPROP_INVAL;
805 entry->pl_user_prop = zfs_asprintf(hdl, "feature@%s",
806 spa_feature_table[i].fi_uname);
807 entry->pl_width = strlen(entry->pl_user_prop);
808 entry->pl_all = B_TRUE;
809
810 *last = entry;
811 last = &entry->pl_next;
812 }
813 }
814
815 /* add any unsupported features */
816 for (nvpair_t *nvp = nvlist_next_nvpair(features, NULL);
817 nvp != NULL; nvp = nvlist_next_nvpair(features, nvp)) {
818 char *propname;
819 boolean_t found;
820 zprop_list_t *entry;
821
822 if (zfeature_is_supported(nvpair_name(nvp)))
823 continue;
824
825 propname = zfs_asprintf(hdl, "unsupported@%s",
826 nvpair_name(nvp));
827
828 /*
829 * Before adding the property to the list make sure that no
830 * other pool already added the same property.
831 */
832 found = B_FALSE;
833 entry = *plp;
834 while (entry != NULL) {
835 if (entry->pl_user_prop != NULL &&
836 strcmp(propname, entry->pl_user_prop) == 0) {
837 found = B_TRUE;
838 break;
839 }
840 entry = entry->pl_next;
841 }
842 if (found) {
843 free(propname);
844 continue;
845 }
846
847 entry = zfs_alloc(hdl, sizeof (zprop_list_t));
848 entry->pl_prop = ZPROP_INVAL;
849 entry->pl_user_prop = propname;
850 entry->pl_width = strlen(entry->pl_user_prop);
851 entry->pl_all = B_TRUE;
852
853 *last = entry;
854 last = &entry->pl_next;
855 }
856
857 for (entry = *plp; entry != NULL; entry = entry->pl_next) {
858
859 if (entry->pl_fixed)
860 continue;
861
862 if (entry->pl_prop != ZPROP_INVAL &&
863 zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf),
864 NULL, B_FALSE) == 0) {
865 if (strlen(buf) > entry->pl_width)
866 entry->pl_width = strlen(buf);
867 }
868 }
869
870 return (0);
871 }
872
873 /*
874 * Get the state for the given feature on the given ZFS pool.
875 */
876 int
zpool_prop_get_feature(zpool_handle_t * zhp,const char * propname,char * buf,size_t len)877 zpool_prop_get_feature(zpool_handle_t *zhp, const char *propname, char *buf,
878 size_t len)
879 {
880 uint64_t refcount;
881 boolean_t found = B_FALSE;
882 nvlist_t *features = zpool_get_features(zhp);
883 boolean_t supported;
884 const char *feature = strchr(propname, '@') + 1;
885
886 supported = zpool_prop_feature(propname);
887 ASSERT(supported || zpool_prop_unsupported(propname));
888
889 /*
890 * Convert from feature name to feature guid. This conversion is
891 * unecessary for unsupported@... properties because they already
892 * use guids.
893 */
894 if (supported) {
895 int ret;
896 spa_feature_t fid;
897
898 ret = zfeature_lookup_name(feature, &fid);
899 if (ret != 0) {
900 (void) strlcpy(buf, "-", len);
901 return (ENOTSUP);
902 }
903 feature = spa_feature_table[fid].fi_guid;
904 }
905
906 if (nvlist_lookup_uint64(features, feature, &refcount) == 0)
907 found = B_TRUE;
908
909 if (supported) {
910 if (!found) {
911 (void) strlcpy(buf, ZFS_FEATURE_DISABLED, len);
912 } else {
913 if (refcount == 0)
914 (void) strlcpy(buf, ZFS_FEATURE_ENABLED, len);
915 else
916 (void) strlcpy(buf, ZFS_FEATURE_ACTIVE, len);
917 }
918 } else {
919 if (found) {
920 if (refcount == 0) {
921 (void) strcpy(buf, ZFS_UNSUPPORTED_INACTIVE);
922 } else {
923 (void) strcpy(buf, ZFS_UNSUPPORTED_READONLY);
924 }
925 } else {
926 (void) strlcpy(buf, "-", len);
927 return (ENOTSUP);
928 }
929 }
930
931 return (0);
932 }
933
934 /*
935 * Don't start the slice at the default block of 34; many storage
936 * devices will use a stripe width of 128k, so start there instead.
937 */
938 #define NEW_START_BLOCK 256
939
940 /*
941 * Validate the given pool name, optionally putting an extended error message in
942 * 'buf'.
943 */
944 boolean_t
zpool_name_valid(libzfs_handle_t * hdl,boolean_t isopen,const char * pool)945 zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool)
946 {
947 namecheck_err_t why;
948 char what;
949 int ret;
950
951 ret = pool_namecheck(pool, &why, &what);
952
953 /*
954 * The rules for reserved pool names were extended at a later point.
955 * But we need to support users with existing pools that may now be
956 * invalid. So we only check for this expanded set of names during a
957 * create (or import), and only in userland.
958 */
959 if (ret == 0 && !isopen &&
960 (strncmp(pool, "mirror", 6) == 0 ||
961 strncmp(pool, "raidz", 5) == 0 ||
962 strncmp(pool, "spare", 5) == 0 ||
963 strcmp(pool, "log") == 0)) {
964 if (hdl != NULL)
965 zfs_error_aux(hdl,
966 dgettext(TEXT_DOMAIN, "name is reserved"));
967 return (B_FALSE);
968 }
969
970
971 if (ret != 0) {
972 if (hdl != NULL) {
973 switch (why) {
974 case NAME_ERR_TOOLONG:
975 zfs_error_aux(hdl,
976 dgettext(TEXT_DOMAIN, "name is too long"));
977 break;
978
979 case NAME_ERR_INVALCHAR:
980 zfs_error_aux(hdl,
981 dgettext(TEXT_DOMAIN, "invalid character "
982 "'%c' in pool name"), what);
983 break;
984
985 case NAME_ERR_NOLETTER:
986 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
987 "name must begin with a letter"));
988 break;
989
990 case NAME_ERR_RESERVED:
991 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
992 "name is reserved"));
993 break;
994
995 case NAME_ERR_DISKLIKE:
996 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
997 "pool name is reserved"));
998 break;
999
1000 case NAME_ERR_LEADING_SLASH:
1001 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1002 "leading slash in name"));
1003 break;
1004
1005 case NAME_ERR_EMPTY_COMPONENT:
1006 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1007 "empty component in name"));
1008 break;
1009
1010 case NAME_ERR_TRAILING_SLASH:
1011 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1012 "trailing slash in name"));
1013 break;
1014
1015 case NAME_ERR_MULTIPLE_DELIMITERS:
1016 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1017 "multiple '@' and/or '#' delimiters in "
1018 "name"));
1019 break;
1020
1021 default:
1022 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1023 "(%d) not defined"), why);
1024 break;
1025 }
1026 }
1027 return (B_FALSE);
1028 }
1029
1030 return (B_TRUE);
1031 }
1032
1033 /*
1034 * Open a handle to the given pool, even if the pool is currently in the FAULTED
1035 * state.
1036 */
1037 zpool_handle_t *
zpool_open_canfail(libzfs_handle_t * hdl,const char * pool)1038 zpool_open_canfail(libzfs_handle_t *hdl, const char *pool)
1039 {
1040 zpool_handle_t *zhp;
1041 boolean_t missing;
1042
1043 /*
1044 * Make sure the pool name is valid.
1045 */
1046 if (!zpool_name_valid(hdl, B_TRUE, pool)) {
1047 (void) zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1048 dgettext(TEXT_DOMAIN, "cannot open '%s'"),
1049 pool);
1050 return (NULL);
1051 }
1052
1053 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
1054 return (NULL);
1055
1056 zhp->zpool_hdl = hdl;
1057 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
1058
1059 if (zpool_refresh_stats(zhp, &missing) != 0) {
1060 zpool_close(zhp);
1061 return (NULL);
1062 }
1063
1064 if (missing) {
1065 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool"));
1066 (void) zfs_error_fmt(hdl, EZFS_NOENT,
1067 dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool);
1068 zpool_close(zhp);
1069 return (NULL);
1070 }
1071
1072 return (zhp);
1073 }
1074
1075 /*
1076 * Like the above, but silent on error. Used when iterating over pools (because
1077 * the configuration cache may be out of date).
1078 */
1079 int
zpool_open_silent(libzfs_handle_t * hdl,const char * pool,zpool_handle_t ** ret)1080 zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret)
1081 {
1082 zpool_handle_t *zhp;
1083 boolean_t missing;
1084
1085 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
1086 return (-1);
1087
1088 zhp->zpool_hdl = hdl;
1089 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
1090
1091 if (zpool_refresh_stats(zhp, &missing) != 0) {
1092 zpool_close(zhp);
1093 return (-1);
1094 }
1095
1096 if (missing) {
1097 zpool_close(zhp);
1098 *ret = NULL;
1099 return (0);
1100 }
1101
1102 *ret = zhp;
1103 return (0);
1104 }
1105
1106 /*
1107 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted
1108 * state.
1109 */
1110 zpool_handle_t *
zpool_open(libzfs_handle_t * hdl,const char * pool)1111 zpool_open(libzfs_handle_t *hdl, const char *pool)
1112 {
1113 zpool_handle_t *zhp;
1114
1115 if ((zhp = zpool_open_canfail(hdl, pool)) == NULL)
1116 return (NULL);
1117
1118 if (zhp->zpool_state == POOL_STATE_UNAVAIL) {
1119 (void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
1120 dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name);
1121 zpool_close(zhp);
1122 return (NULL);
1123 }
1124
1125 return (zhp);
1126 }
1127
1128 /*
1129 * Close the handle. Simply frees the memory associated with the handle.
1130 */
1131 void
zpool_close(zpool_handle_t * zhp)1132 zpool_close(zpool_handle_t *zhp)
1133 {
1134 nvlist_free(zhp->zpool_config);
1135 nvlist_free(zhp->zpool_old_config);
1136 nvlist_free(zhp->zpool_props);
1137 free(zhp);
1138 }
1139
1140 /*
1141 * Return the name of the pool.
1142 */
1143 const char *
zpool_get_name(zpool_handle_t * zhp)1144 zpool_get_name(zpool_handle_t *zhp)
1145 {
1146 return (zhp->zpool_name);
1147 }
1148
1149
1150 /*
1151 * Return the state of the pool (ACTIVE or UNAVAILABLE)
1152 */
1153 int
zpool_get_state(zpool_handle_t * zhp)1154 zpool_get_state(zpool_handle_t *zhp)
1155 {
1156 return (zhp->zpool_state);
1157 }
1158
1159 /*
1160 * Check if vdev list contains a special vdev
1161 */
1162 static boolean_t
zpool_has_special_vdev(nvlist_t * nvroot)1163 zpool_has_special_vdev(nvlist_t *nvroot)
1164 {
1165 nvlist_t **child;
1166 uint_t children;
1167
1168 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, &child,
1169 &children) == 0) {
1170 for (uint_t c = 0; c < children; c++) {
1171 char *bias;
1172
1173 if (nvlist_lookup_string(child[c],
1174 ZPOOL_CONFIG_ALLOCATION_BIAS, &bias) == 0 &&
1175 strcmp(bias, VDEV_ALLOC_BIAS_SPECIAL) == 0) {
1176 return (B_TRUE);
1177 }
1178 }
1179 }
1180 return (B_FALSE);
1181 }
1182
1183 /*
1184 * Create the named pool, using the provided vdev list. It is assumed
1185 * that the consumer has already validated the contents of the nvlist, so we
1186 * don't have to worry about error semantics.
1187 */
1188 int
zpool_create(libzfs_handle_t * hdl,const char * pool,nvlist_t * nvroot,nvlist_t * props,nvlist_t * fsprops)1189 zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot,
1190 nvlist_t *props, nvlist_t *fsprops)
1191 {
1192 zfs_cmd_t zc = { 0 };
1193 nvlist_t *zc_fsprops = NULL;
1194 nvlist_t *zc_props = NULL;
1195 nvlist_t *hidden_args = NULL;
1196 uint8_t *wkeydata = NULL;
1197 uint_t wkeylen = 0;
1198 char msg[1024];
1199 int ret = -1;
1200
1201 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1202 "cannot create '%s'"), pool);
1203
1204 if (!zpool_name_valid(hdl, B_FALSE, pool))
1205 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
1206
1207 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1208 return (-1);
1209
1210 if (props) {
1211 prop_flags_t flags = { .create = B_TRUE, .import = B_FALSE };
1212
1213 if ((zc_props = zpool_valid_proplist(hdl, pool, props,
1214 SPA_VERSION_1, flags, msg)) == NULL) {
1215 goto create_failed;
1216 }
1217 }
1218
1219 if (fsprops) {
1220 uint64_t zoned;
1221 char *zonestr;
1222
1223 zoned = ((nvlist_lookup_string(fsprops,
1224 zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) &&
1225 strcmp(zonestr, "on") == 0);
1226
1227 if ((zc_fsprops = zfs_valid_proplist(hdl, ZFS_TYPE_FILESYSTEM,
1228 fsprops, zoned, NULL, NULL, B_TRUE, msg)) == NULL) {
1229 goto create_failed;
1230 }
1231
1232 if (nvlist_exists(zc_fsprops,
1233 zfs_prop_to_name(ZFS_PROP_SPECIAL_SMALL_BLOCKS)) &&
1234 !zpool_has_special_vdev(nvroot)) {
1235 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1236 "%s property requires a special vdev"),
1237 zfs_prop_to_name(ZFS_PROP_SPECIAL_SMALL_BLOCKS));
1238 (void) zfs_error(hdl, EZFS_BADPROP, msg);
1239 goto create_failed;
1240 }
1241
1242 if (!zc_props &&
1243 (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) {
1244 goto create_failed;
1245 }
1246 if (zfs_crypto_create(hdl, NULL, zc_fsprops, props, B_TRUE,
1247 &wkeydata, &wkeylen) != 0) {
1248 (void) zfs_error(hdl, EZFS_CRYPTOFAILED, msg);
1249 goto create_failed;
1250 }
1251 if (nvlist_add_nvlist(zc_props,
1252 ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) {
1253 goto create_failed;
1254 }
1255 if (wkeydata != NULL) {
1256 if (nvlist_alloc(&hidden_args, NV_UNIQUE_NAME, 0) != 0)
1257 goto create_failed;
1258
1259 if (nvlist_add_uint8_array(hidden_args, "wkeydata",
1260 wkeydata, wkeylen) != 0)
1261 goto create_failed;
1262
1263 if (nvlist_add_nvlist(zc_props, ZPOOL_HIDDEN_ARGS,
1264 hidden_args) != 0)
1265 goto create_failed;
1266 }
1267 }
1268
1269 if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
1270 goto create_failed;
1271
1272 (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name));
1273
1274 if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) {
1275
1276 zcmd_free_nvlists(&zc);
1277 nvlist_free(zc_props);
1278 nvlist_free(zc_fsprops);
1279 nvlist_free(hidden_args);
1280 if (wkeydata != NULL)
1281 free(wkeydata);
1282
1283 switch (errno) {
1284 case EBUSY:
1285 /*
1286 * This can happen if the user has specified the same
1287 * device multiple times. We can't reliably detect this
1288 * until we try to add it and see we already have a
1289 * label.
1290 */
1291 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1292 "one or more vdevs refer to the same device"));
1293 return (zfs_error(hdl, EZFS_BADDEV, msg));
1294
1295 case EDOM:
1296 /*
1297 * This happens if the asize/ashift required by a disk
1298 * vdev is less than ASHIFT_MIN or greater than
1299 * ASHIFT_MAX.
1300 */
1301 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1302 "one or more vdevs require an invalid ashift"));
1303 return (zfs_error(hdl, EZFS_BADDEV, msg));
1304
1305 case ERANGE:
1306 /*
1307 * This happens if the record size is smaller or larger
1308 * than the allowed size range, or not a power of 2.
1309 *
1310 * NOTE: although zfs_valid_proplist is called earlier,
1311 * this case may have slipped through since the
1312 * pool does not exist yet and it is therefore
1313 * impossible to read properties e.g. max blocksize
1314 * from the pool.
1315 */
1316 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1317 "record size invalid"));
1318 return (zfs_error(hdl, EZFS_BADPROP, msg));
1319
1320 case EOVERFLOW:
1321 /*
1322 * This occurs when one of the devices is below
1323 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1324 * device was the problem device since there's no
1325 * reliable way to determine device size from userland.
1326 */
1327 {
1328 char buf[64];
1329
1330 zfs_nicebytes(SPA_MINDEVSIZE, buf,
1331 sizeof (buf));
1332
1333 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1334 "one or more devices is less than the "
1335 "minimum size (%s)"), buf);
1336 }
1337 return (zfs_error(hdl, EZFS_BADDEV, msg));
1338
1339 case ENOSPC:
1340 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1341 "one or more devices is out of space"));
1342 return (zfs_error(hdl, EZFS_BADDEV, msg));
1343
1344 default:
1345 return (zpool_standard_error(hdl, errno, msg));
1346 }
1347 }
1348
1349 create_failed:
1350 zcmd_free_nvlists(&zc);
1351 nvlist_free(zc_props);
1352 nvlist_free(zc_fsprops);
1353 nvlist_free(hidden_args);
1354 if (wkeydata != NULL)
1355 free(wkeydata);
1356 return (ret);
1357 }
1358
1359 /*
1360 * Destroy the given pool. It is up to the caller to ensure that there are no
1361 * datasets left in the pool.
1362 */
1363 int
zpool_destroy(zpool_handle_t * zhp,const char * log_str)1364 zpool_destroy(zpool_handle_t *zhp, const char *log_str)
1365 {
1366 zfs_cmd_t zc = { 0 };
1367 zfs_handle_t *zfp = NULL;
1368 libzfs_handle_t *hdl = zhp->zpool_hdl;
1369 char msg[1024];
1370
1371 if (zhp->zpool_state == POOL_STATE_ACTIVE &&
1372 (zfp = zfs_open(hdl, zhp->zpool_name, ZFS_TYPE_FILESYSTEM)) == NULL)
1373 return (-1);
1374
1375 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1376 zc.zc_history = (uint64_t)(uintptr_t)log_str;
1377
1378 if (zfs_ioctl(hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) {
1379 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1380 "cannot destroy '%s'"), zhp->zpool_name);
1381
1382 if (errno == EROFS) {
1383 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1384 "one or more devices is read only"));
1385 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1386 } else {
1387 (void) zpool_standard_error(hdl, errno, msg);
1388 }
1389
1390 if (zfp)
1391 zfs_close(zfp);
1392 return (-1);
1393 }
1394
1395 if (zfp) {
1396 remove_mountpoint(zfp);
1397 zfs_close(zfp);
1398 }
1399
1400 return (0);
1401 }
1402
1403 /*
1404 * Create a checkpoint in the given pool.
1405 */
1406 int
zpool_checkpoint(zpool_handle_t * zhp)1407 zpool_checkpoint(zpool_handle_t *zhp)
1408 {
1409 libzfs_handle_t *hdl = zhp->zpool_hdl;
1410 char msg[1024];
1411 int error;
1412
1413 error = lzc_pool_checkpoint(zhp->zpool_name);
1414 if (error != 0) {
1415 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1416 "cannot checkpoint '%s'"), zhp->zpool_name);
1417 (void) zpool_standard_error(hdl, error, msg);
1418 return (-1);
1419 }
1420
1421 return (0);
1422 }
1423
1424 /*
1425 * Discard the checkpoint from the given pool.
1426 */
1427 int
zpool_discard_checkpoint(zpool_handle_t * zhp)1428 zpool_discard_checkpoint(zpool_handle_t *zhp)
1429 {
1430 libzfs_handle_t *hdl = zhp->zpool_hdl;
1431 char msg[1024];
1432 int error;
1433
1434 error = lzc_pool_checkpoint_discard(zhp->zpool_name);
1435 if (error != 0) {
1436 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1437 "cannot discard checkpoint in '%s'"), zhp->zpool_name);
1438 (void) zpool_standard_error(hdl, error, msg);
1439 return (-1);
1440 }
1441
1442 return (0);
1443 }
1444
1445 /*
1446 * Add the given vdevs to the pool. The caller must have already performed the
1447 * necessary verification to ensure that the vdev specification is well-formed.
1448 */
1449 int
zpool_add(zpool_handle_t * zhp,nvlist_t * nvroot)1450 zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot)
1451 {
1452 zfs_cmd_t zc = { 0 };
1453 int ret;
1454 libzfs_handle_t *hdl = zhp->zpool_hdl;
1455 char msg[1024];
1456 nvlist_t **spares, **l2cache;
1457 uint_t nspares, nl2cache;
1458
1459 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1460 "cannot add to '%s'"), zhp->zpool_name);
1461
1462 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1463 SPA_VERSION_SPARES &&
1464 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
1465 &spares, &nspares) == 0) {
1466 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1467 "upgraded to add hot spares"));
1468 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1469 }
1470
1471 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1472 SPA_VERSION_L2CACHE &&
1473 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
1474 &l2cache, &nl2cache) == 0) {
1475 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1476 "upgraded to add cache devices"));
1477 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1478 }
1479
1480 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1481 return (-1);
1482 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1483
1484 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) {
1485 switch (errno) {
1486 case EBUSY:
1487 /*
1488 * This can happen if the user has specified the same
1489 * device multiple times. We can't reliably detect this
1490 * until we try to add it and see we already have a
1491 * label.
1492 */
1493 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1494 "one or more vdevs refer to the same device"));
1495 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1496 break;
1497
1498 case EINVAL:
1499 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1500 "invalid config; a pool with removing/removed "
1501 "vdevs does not support adding raidz vdevs"));
1502 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1503 break;
1504
1505 case EOVERFLOW:
1506 /*
1507 * This occurrs when one of the devices is below
1508 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1509 * device was the problem device since there's no
1510 * reliable way to determine device size from userland.
1511 */
1512 {
1513 char buf[64];
1514
1515 zfs_nicebytes(SPA_MINDEVSIZE, buf,
1516 sizeof (buf));
1517
1518 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1519 "device is less than the minimum "
1520 "size (%s)"), buf);
1521 }
1522 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1523 break;
1524
1525 case ENOTSUP:
1526 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1527 "pool must be upgraded to add these vdevs"));
1528 (void) zfs_error(hdl, EZFS_BADVERSION, msg);
1529 break;
1530
1531 default:
1532 (void) zpool_standard_error(hdl, errno, msg);
1533 }
1534
1535 ret = -1;
1536 } else {
1537 ret = 0;
1538 }
1539
1540 zcmd_free_nvlists(&zc);
1541
1542 return (ret);
1543 }
1544
1545 /*
1546 * Exports the pool from the system. The caller must ensure that there are no
1547 * mounted datasets in the pool.
1548 */
1549 static int
zpool_export_common(zpool_handle_t * zhp,boolean_t force,boolean_t hardforce,const char * log_str)1550 zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce,
1551 const char *log_str)
1552 {
1553 zfs_cmd_t zc = { 0 };
1554 char msg[1024];
1555
1556 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1557 "cannot export '%s'"), zhp->zpool_name);
1558
1559 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1560 zc.zc_cookie = force;
1561 zc.zc_guid = hardforce;
1562 zc.zc_history = (uint64_t)(uintptr_t)log_str;
1563
1564 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) {
1565 switch (errno) {
1566 case EXDEV:
1567 zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
1568 "use '-f' to override the following errors:\n"
1569 "'%s' has an active shared spare which could be"
1570 " used by other pools once '%s' is exported."),
1571 zhp->zpool_name, zhp->zpool_name);
1572 return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE,
1573 msg));
1574 default:
1575 return (zpool_standard_error_fmt(zhp->zpool_hdl, errno,
1576 msg));
1577 }
1578 }
1579
1580 return (0);
1581 }
1582
1583 int
zpool_export(zpool_handle_t * zhp,boolean_t force,const char * log_str)1584 zpool_export(zpool_handle_t *zhp, boolean_t force, const char *log_str)
1585 {
1586 return (zpool_export_common(zhp, force, B_FALSE, log_str));
1587 }
1588
1589 int
zpool_export_force(zpool_handle_t * zhp,const char * log_str)1590 zpool_export_force(zpool_handle_t *zhp, const char *log_str)
1591 {
1592 return (zpool_export_common(zhp, B_TRUE, B_TRUE, log_str));
1593 }
1594
1595 static void
zpool_rewind_exclaim(libzfs_handle_t * hdl,const char * name,boolean_t dryrun,nvlist_t * config)1596 zpool_rewind_exclaim(libzfs_handle_t *hdl, const char *name, boolean_t dryrun,
1597 nvlist_t *config)
1598 {
1599 nvlist_t *nv = NULL;
1600 uint64_t rewindto;
1601 int64_t loss = -1;
1602 struct tm t;
1603 char timestr[128];
1604
1605 if (!hdl->libzfs_printerr || config == NULL)
1606 return;
1607
1608 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
1609 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0) {
1610 return;
1611 }
1612
1613 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1614 return;
1615 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1616
1617 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1618 strftime(timestr, 128, 0, &t) != 0) {
1619 if (dryrun) {
1620 (void) printf(dgettext(TEXT_DOMAIN,
1621 "Would be able to return %s "
1622 "to its state as of %s.\n"),
1623 name, timestr);
1624 } else {
1625 (void) printf(dgettext(TEXT_DOMAIN,
1626 "Pool %s returned to its state as of %s.\n"),
1627 name, timestr);
1628 }
1629 if (loss > 120) {
1630 (void) printf(dgettext(TEXT_DOMAIN,
1631 "%s approximately %lld "),
1632 dryrun ? "Would discard" : "Discarded",
1633 (loss + 30) / 60);
1634 (void) printf(dgettext(TEXT_DOMAIN,
1635 "minutes of transactions.\n"));
1636 } else if (loss > 0) {
1637 (void) printf(dgettext(TEXT_DOMAIN,
1638 "%s approximately %lld "),
1639 dryrun ? "Would discard" : "Discarded", loss);
1640 (void) printf(dgettext(TEXT_DOMAIN,
1641 "seconds of transactions.\n"));
1642 }
1643 }
1644 }
1645
1646 void
zpool_explain_recover(libzfs_handle_t * hdl,const char * name,int reason,nvlist_t * config)1647 zpool_explain_recover(libzfs_handle_t *hdl, const char *name, int reason,
1648 nvlist_t *config)
1649 {
1650 nvlist_t *nv = NULL;
1651 int64_t loss = -1;
1652 uint64_t edata = UINT64_MAX;
1653 uint64_t rewindto;
1654 struct tm t;
1655 char timestr[128];
1656
1657 if (!hdl->libzfs_printerr)
1658 return;
1659
1660 if (reason >= 0)
1661 (void) printf(dgettext(TEXT_DOMAIN, "action: "));
1662 else
1663 (void) printf(dgettext(TEXT_DOMAIN, "\t"));
1664
1665 /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */
1666 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
1667 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0 ||
1668 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1669 goto no_info;
1670
1671 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1672 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_DATA_ERRORS,
1673 &edata);
1674
1675 (void) printf(dgettext(TEXT_DOMAIN,
1676 "Recovery is possible, but will result in some data loss.\n"));
1677
1678 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1679 strftime(timestr, 128, 0, &t) != 0) {
1680 (void) printf(dgettext(TEXT_DOMAIN,
1681 "\tReturning the pool to its state as of %s\n"
1682 "\tshould correct the problem. "),
1683 timestr);
1684 } else {
1685 (void) printf(dgettext(TEXT_DOMAIN,
1686 "\tReverting the pool to an earlier state "
1687 "should correct the problem.\n\t"));
1688 }
1689
1690 if (loss > 120) {
1691 (void) printf(dgettext(TEXT_DOMAIN,
1692 "Approximately %lld minutes of data\n"
1693 "\tmust be discarded, irreversibly. "), (loss + 30) / 60);
1694 } else if (loss > 0) {
1695 (void) printf(dgettext(TEXT_DOMAIN,
1696 "Approximately %lld seconds of data\n"
1697 "\tmust be discarded, irreversibly. "), loss);
1698 }
1699 if (edata != 0 && edata != UINT64_MAX) {
1700 if (edata == 1) {
1701 (void) printf(dgettext(TEXT_DOMAIN,
1702 "After rewind, at least\n"
1703 "\tone persistent user-data error will remain. "));
1704 } else {
1705 (void) printf(dgettext(TEXT_DOMAIN,
1706 "After rewind, several\n"
1707 "\tpersistent user-data errors will remain. "));
1708 }
1709 }
1710 (void) printf(dgettext(TEXT_DOMAIN,
1711 "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "),
1712 reason >= 0 ? "clear" : "import", name);
1713
1714 (void) printf(dgettext(TEXT_DOMAIN,
1715 "A scrub of the pool\n"
1716 "\tis strongly recommended after recovery.\n"));
1717 return;
1718
1719 no_info:
1720 (void) printf(dgettext(TEXT_DOMAIN,
1721 "Destroy and re-create the pool from\n\ta backup source.\n"));
1722 }
1723
1724 /*
1725 * zpool_import() is a contracted interface. Should be kept the same
1726 * if possible.
1727 *
1728 * Applications should use zpool_import_props() to import a pool with
1729 * new properties value to be set.
1730 */
1731 int
zpool_import(libzfs_handle_t * hdl,nvlist_t * config,const char * newname,char * altroot)1732 zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1733 char *altroot)
1734 {
1735 nvlist_t *props = NULL;
1736 int ret;
1737
1738 if (altroot != NULL) {
1739 if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) {
1740 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1741 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1742 newname));
1743 }
1744
1745 if (nvlist_add_string(props,
1746 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 ||
1747 nvlist_add_string(props,
1748 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) {
1749 nvlist_free(props);
1750 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1751 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1752 newname));
1753 }
1754 }
1755
1756 ret = zpool_import_props(hdl, config, newname, props,
1757 ZFS_IMPORT_NORMAL);
1758 nvlist_free(props);
1759 return (ret);
1760 }
1761
1762 static void
print_vdev_tree(libzfs_handle_t * hdl,const char * name,nvlist_t * nv,int indent)1763 print_vdev_tree(libzfs_handle_t *hdl, const char *name, nvlist_t *nv,
1764 int indent)
1765 {
1766 nvlist_t **child;
1767 uint_t c, children;
1768 char *vname;
1769 uint64_t is_log = 0;
1770
1771 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG,
1772 &is_log);
1773
1774 if (name != NULL)
1775 (void) printf("\t%*s%s%s\n", indent, "", name,
1776 is_log ? " [log]" : "");
1777
1778 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1779 &child, &children) != 0)
1780 return;
1781
1782 for (c = 0; c < children; c++) {
1783 vname = zpool_vdev_name(hdl, NULL, child[c], VDEV_NAME_TYPE_ID);
1784 print_vdev_tree(hdl, vname, child[c], indent + 2);
1785 free(vname);
1786 }
1787 }
1788
1789 void
zpool_print_unsup_feat(nvlist_t * config)1790 zpool_print_unsup_feat(nvlist_t *config)
1791 {
1792 nvlist_t *nvinfo, *unsup_feat;
1793
1794 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nvinfo) ==
1795 0);
1796 verify(nvlist_lookup_nvlist(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT,
1797 &unsup_feat) == 0);
1798
1799 for (nvpair_t *nvp = nvlist_next_nvpair(unsup_feat, NULL); nvp != NULL;
1800 nvp = nvlist_next_nvpair(unsup_feat, nvp)) {
1801 char *desc;
1802
1803 verify(nvpair_type(nvp) == DATA_TYPE_STRING);
1804 verify(nvpair_value_string(nvp, &desc) == 0);
1805
1806 if (strlen(desc) > 0)
1807 (void) printf("\t%s (%s)\n", nvpair_name(nvp), desc);
1808 else
1809 (void) printf("\t%s\n", nvpair_name(nvp));
1810 }
1811 }
1812
1813 /*
1814 * Import the given pool using the known configuration and a list of
1815 * properties to be set. The configuration should have come from
1816 * zpool_find_import(). The 'newname' parameters control whether the pool
1817 * is imported with a different name.
1818 */
1819 int
zpool_import_props(libzfs_handle_t * hdl,nvlist_t * config,const char * newname,nvlist_t * props,int flags)1820 zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1821 nvlist_t *props, int flags)
1822 {
1823 zfs_cmd_t zc = { 0 };
1824 zpool_load_policy_t policy;
1825 nvlist_t *nv = NULL;
1826 nvlist_t *nvinfo = NULL;
1827 nvlist_t *missing = NULL;
1828 char *thename;
1829 char *origname;
1830 int ret;
1831 int error = 0;
1832 char errbuf[1024];
1833
1834 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1835 &origname) == 0);
1836
1837 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1838 "cannot import pool '%s'"), origname);
1839
1840 if (newname != NULL) {
1841 if (!zpool_name_valid(hdl, B_FALSE, newname))
1842 return (zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1843 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1844 newname));
1845 thename = (char *)newname;
1846 } else {
1847 thename = origname;
1848 }
1849
1850 if (props != NULL) {
1851 uint64_t version;
1852 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
1853
1854 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
1855 &version) == 0);
1856
1857 if ((props = zpool_valid_proplist(hdl, origname,
1858 props, version, flags, errbuf)) == NULL)
1859 return (-1);
1860 if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) {
1861 nvlist_free(props);
1862 return (-1);
1863 }
1864 nvlist_free(props);
1865 }
1866
1867 (void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name));
1868
1869 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1870 &zc.zc_guid) == 0);
1871
1872 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) {
1873 zcmd_free_nvlists(&zc);
1874 return (-1);
1875 }
1876 if (zcmd_alloc_dst_nvlist(hdl, &zc, zc.zc_nvlist_conf_size * 2) != 0) {
1877 zcmd_free_nvlists(&zc);
1878 return (-1);
1879 }
1880
1881 zc.zc_cookie = flags;
1882 while ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc)) != 0 &&
1883 errno == ENOMEM) {
1884 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
1885 zcmd_free_nvlists(&zc);
1886 return (-1);
1887 }
1888 }
1889 if (ret != 0)
1890 error = errno;
1891
1892 (void) zcmd_read_dst_nvlist(hdl, &zc, &nv);
1893
1894 zcmd_free_nvlists(&zc);
1895
1896 zpool_get_load_policy(config, &policy);
1897
1898 if (error) {
1899 char desc[1024];
1900 char aux[256];
1901
1902 /*
1903 * Dry-run failed, but we print out what success
1904 * looks like if we found a best txg
1905 */
1906 if (policy.zlp_rewind & ZPOOL_TRY_REWIND) {
1907 zpool_rewind_exclaim(hdl, newname ? origname : thename,
1908 B_TRUE, nv);
1909 nvlist_free(nv);
1910 return (-1);
1911 }
1912
1913 if (newname == NULL)
1914 (void) snprintf(desc, sizeof (desc),
1915 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1916 thename);
1917 else
1918 (void) snprintf(desc, sizeof (desc),
1919 dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"),
1920 origname, thename);
1921
1922 switch (error) {
1923 case ENOTSUP:
1924 if (nv != NULL && nvlist_lookup_nvlist(nv,
1925 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
1926 nvlist_exists(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT)) {
1927 (void) printf(dgettext(TEXT_DOMAIN, "This "
1928 "pool uses the following feature(s) not "
1929 "supported by this system:\n"));
1930 zpool_print_unsup_feat(nv);
1931 if (nvlist_exists(nvinfo,
1932 ZPOOL_CONFIG_CAN_RDONLY)) {
1933 (void) printf(dgettext(TEXT_DOMAIN,
1934 "All unsupported features are only "
1935 "required for writing to the pool."
1936 "\nThe pool can be imported using "
1937 "'-o readonly=on'.\n"));
1938 }
1939 }
1940 /*
1941 * Unsupported version.
1942 */
1943 (void) zfs_error(hdl, EZFS_BADVERSION, desc);
1944 break;
1945
1946 case EREMOTEIO:
1947 if (nv != NULL && nvlist_lookup_nvlist(nv,
1948 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0) {
1949 char *hostname = "<unknown>";
1950 uint64_t hostid = 0;
1951 mmp_state_t mmp_state;
1952
1953 mmp_state = fnvlist_lookup_uint64(nvinfo,
1954 ZPOOL_CONFIG_MMP_STATE);
1955
1956 if (nvlist_exists(nvinfo,
1957 ZPOOL_CONFIG_MMP_HOSTNAME))
1958 hostname = fnvlist_lookup_string(nvinfo,
1959 ZPOOL_CONFIG_MMP_HOSTNAME);
1960
1961 if (nvlist_exists(nvinfo,
1962 ZPOOL_CONFIG_MMP_HOSTID))
1963 hostid = fnvlist_lookup_uint64(nvinfo,
1964 ZPOOL_CONFIG_MMP_HOSTID);
1965
1966 if (mmp_state == MMP_STATE_ACTIVE) {
1967 (void) snprintf(aux, sizeof (aux),
1968 dgettext(TEXT_DOMAIN, "pool is imp"
1969 "orted on host '%s' (hostid=%lx).\n"
1970 "Export the pool on the other "
1971 "system, then run 'zpool import'."),
1972 hostname, (unsigned long) hostid);
1973 } else if (mmp_state == MMP_STATE_NO_HOSTID) {
1974 (void) snprintf(aux, sizeof (aux),
1975 dgettext(TEXT_DOMAIN, "pool has "
1976 "the multihost property on and "
1977 "the\nsystem's hostid is not "
1978 "set.\n"));
1979 }
1980
1981 (void) zfs_error_aux(hdl, aux);
1982 }
1983 (void) zfs_error(hdl, EZFS_ACTIVE_POOL, desc);
1984 break;
1985
1986 case EINVAL:
1987 (void) zfs_error(hdl, EZFS_INVALCONFIG, desc);
1988 break;
1989
1990 case EROFS:
1991 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1992 "one or more devices is read only"));
1993 (void) zfs_error(hdl, EZFS_BADDEV, desc);
1994 break;
1995
1996 case ENXIO:
1997 if (nv && nvlist_lookup_nvlist(nv,
1998 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
1999 nvlist_lookup_nvlist(nvinfo,
2000 ZPOOL_CONFIG_MISSING_DEVICES, &missing) == 0) {
2001 (void) printf(dgettext(TEXT_DOMAIN,
2002 "The devices below are missing or "
2003 "corrupted, use '-m' to import the pool "
2004 "anyway:\n"));
2005 print_vdev_tree(hdl, NULL, missing, 2);
2006 (void) printf("\n");
2007 }
2008 (void) zpool_standard_error(hdl, error, desc);
2009 break;
2010
2011 case EEXIST:
2012 (void) zpool_standard_error(hdl, error, desc);
2013 break;
2014 case ENAMETOOLONG:
2015 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2016 "new name of at least one dataset is longer than "
2017 "the maximum allowable length"));
2018 (void) zfs_error(hdl, EZFS_NAMETOOLONG, desc);
2019 break;
2020 default:
2021 (void) zpool_standard_error(hdl, error, desc);
2022 zpool_explain_recover(hdl,
2023 newname ? origname : thename, -error, nv);
2024 break;
2025 }
2026
2027 nvlist_free(nv);
2028 ret = -1;
2029 } else {
2030 zpool_handle_t *zhp;
2031
2032 /*
2033 * This should never fail, but play it safe anyway.
2034 */
2035 if (zpool_open_silent(hdl, thename, &zhp) != 0)
2036 ret = -1;
2037 else if (zhp != NULL)
2038 zpool_close(zhp);
2039 if (policy.zlp_rewind &
2040 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
2041 zpool_rewind_exclaim(hdl, newname ? origname : thename,
2042 ((policy.zlp_rewind & ZPOOL_TRY_REWIND) != 0), nv);
2043 }
2044 nvlist_free(nv);
2045 return (0);
2046 }
2047
2048 return (ret);
2049 }
2050
2051 /*
2052 * Translate vdev names to guids. If a vdev_path is determined to be
2053 * unsuitable then a vd_errlist is allocated and the vdev path and errno
2054 * are added to it.
2055 */
2056 static int
zpool_translate_vdev_guids(zpool_handle_t * zhp,nvlist_t * vds,nvlist_t * vdev_guids,nvlist_t * guids_to_paths,nvlist_t ** vd_errlist)2057 zpool_translate_vdev_guids(zpool_handle_t *zhp, nvlist_t *vds,
2058 nvlist_t *vdev_guids, nvlist_t *guids_to_paths, nvlist_t **vd_errlist)
2059 {
2060 nvlist_t *errlist = NULL;
2061 int error = 0;
2062
2063 for (nvpair_t *elem = nvlist_next_nvpair(vds, NULL); elem != NULL;
2064 elem = nvlist_next_nvpair(vds, elem)) {
2065 boolean_t spare, cache;
2066
2067 char *vd_path = nvpair_name(elem);
2068 nvlist_t *tgt = zpool_find_vdev(zhp, vd_path, &spare, &cache,
2069 NULL);
2070
2071 if ((tgt == NULL) || cache || spare) {
2072 if (errlist == NULL) {
2073 errlist = fnvlist_alloc();
2074 error = EINVAL;
2075 }
2076
2077 uint64_t err = (tgt == NULL) ? EZFS_NODEVICE :
2078 (spare ? EZFS_ISSPARE : EZFS_ISL2CACHE);
2079 fnvlist_add_int64(errlist, vd_path, err);
2080 continue;
2081 }
2082
2083 uint64_t guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
2084 fnvlist_add_uint64(vdev_guids, vd_path, guid);
2085
2086 char msg[MAXNAMELEN];
2087 (void) snprintf(msg, sizeof (msg), "%llu", (u_longlong_t)guid);
2088 fnvlist_add_string(guids_to_paths, msg, vd_path);
2089 }
2090
2091 if (error != 0) {
2092 verify(errlist != NULL);
2093 if (vd_errlist != NULL)
2094 *vd_errlist = errlist;
2095 else
2096 fnvlist_free(errlist);
2097 }
2098
2099 return (error);
2100 }
2101
2102 /*
2103 * Scan the pool.
2104 */
2105 int
zpool_scan(zpool_handle_t * zhp,pool_scan_func_t func,pool_scrub_cmd_t cmd)2106 zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func, pool_scrub_cmd_t cmd)
2107 {
2108 zfs_cmd_t zc = { 0 };
2109 char msg[1024];
2110 int err;
2111 libzfs_handle_t *hdl = zhp->zpool_hdl;
2112
2113 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2114 zc.zc_cookie = func;
2115 zc.zc_flags = cmd;
2116
2117 if (zfs_ioctl(hdl, ZFS_IOC_POOL_SCAN, &zc) == 0)
2118 return (0);
2119
2120 err = errno;
2121
2122 /* ECANCELED on a scrub means we resumed a paused scrub */
2123 if (err == ECANCELED && func == POOL_SCAN_SCRUB &&
2124 cmd == POOL_SCRUB_NORMAL)
2125 return (0);
2126
2127 if (err == ENOENT && func != POOL_SCAN_NONE && cmd == POOL_SCRUB_NORMAL)
2128 return (0);
2129
2130 if (func == POOL_SCAN_SCRUB) {
2131 if (cmd == POOL_SCRUB_PAUSE) {
2132 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2133 "cannot pause scrubbing %s"), zc.zc_name);
2134 } else {
2135 assert(cmd == POOL_SCRUB_NORMAL);
2136 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2137 "cannot scrub %s"), zc.zc_name);
2138 }
2139 } else if (func == POOL_SCAN_RESILVER) {
2140 assert(cmd == POOL_SCRUB_NORMAL);
2141 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2142 "cannot restart resilver on %s"), zc.zc_name);
2143 } else if (func == POOL_SCAN_NONE) {
2144 (void) snprintf(msg, sizeof (msg),
2145 dgettext(TEXT_DOMAIN, "cannot cancel scrubbing %s"),
2146 zc.zc_name);
2147 } else {
2148 assert(!"unexpected result");
2149 }
2150
2151 if (err == EBUSY) {
2152 nvlist_t *nvroot;
2153 pool_scan_stat_t *ps = NULL;
2154 uint_t psc;
2155
2156 verify(nvlist_lookup_nvlist(zhp->zpool_config,
2157 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
2158 (void) nvlist_lookup_uint64_array(nvroot,
2159 ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc);
2160 if (ps && ps->pss_func == POOL_SCAN_SCRUB) {
2161 if (cmd == POOL_SCRUB_PAUSE)
2162 return (zfs_error(hdl, EZFS_SCRUB_PAUSED, msg));
2163 else
2164 return (zfs_error(hdl, EZFS_SCRUBBING, msg));
2165 } else {
2166 return (zfs_error(hdl, EZFS_RESILVERING, msg));
2167 }
2168 } else if (err == ENOENT) {
2169 return (zfs_error(hdl, EZFS_NO_SCRUB, msg));
2170 } else if (err == ENOTSUP && func == POOL_SCAN_RESILVER) {
2171 return (zfs_error(hdl, EZFS_NO_RESILVER_DEFER, msg));
2172 } else {
2173 return (zpool_standard_error(hdl, err, msg));
2174 }
2175 }
2176
2177 static int
xlate_init_err(int err)2178 xlate_init_err(int err)
2179 {
2180 switch (err) {
2181 case ENODEV:
2182 return (EZFS_NODEVICE);
2183 case EINVAL:
2184 case EROFS:
2185 return (EZFS_BADDEV);
2186 case EBUSY:
2187 return (EZFS_INITIALIZING);
2188 case ESRCH:
2189 return (EZFS_NO_INITIALIZE);
2190 }
2191 return (err);
2192 }
2193
2194 /*
2195 * Begin, suspend, or cancel the initialization (initializing of all free
2196 * blocks) for the given vdevs in the given pool.
2197 */
2198 int
zpool_initialize(zpool_handle_t * zhp,pool_initialize_func_t cmd_type,nvlist_t * vds)2199 zpool_initialize(zpool_handle_t *zhp, pool_initialize_func_t cmd_type,
2200 nvlist_t *vds)
2201 {
2202 char msg[1024];
2203 int err;
2204
2205 nvlist_t *vdev_guids = fnvlist_alloc();
2206 nvlist_t *guids_to_paths = fnvlist_alloc();
2207 nvlist_t *vd_errlist = NULL;
2208 nvlist_t *errlist;
2209 nvpair_t *elem;
2210
2211 err = zpool_translate_vdev_guids(zhp, vds, vdev_guids,
2212 guids_to_paths, &vd_errlist);
2213
2214 if (err == 0) {
2215 err = lzc_initialize(zhp->zpool_name, cmd_type,
2216 vdev_guids, &errlist);
2217 if (err == 0) {
2218 fnvlist_free(vdev_guids);
2219 fnvlist_free(guids_to_paths);
2220 return (0);
2221 }
2222
2223 if (errlist != NULL) {
2224 vd_errlist = fnvlist_lookup_nvlist(errlist,
2225 ZPOOL_INITIALIZE_VDEVS);
2226 }
2227
2228 (void) snprintf(msg, sizeof (msg),
2229 dgettext(TEXT_DOMAIN, "operation failed"));
2230 } else {
2231 verify(vd_errlist != NULL);
2232 }
2233
2234 for (elem = nvlist_next_nvpair(vd_errlist, NULL); elem != NULL;
2235 elem = nvlist_next_nvpair(vd_errlist, elem)) {
2236 int64_t vd_error = xlate_init_err(fnvpair_value_int64(elem));
2237 char *path;
2238
2239 if (nvlist_lookup_string(guids_to_paths, nvpair_name(elem),
2240 &path) != 0)
2241 path = nvpair_name(elem);
2242
2243 (void) zfs_error_fmt(zhp->zpool_hdl, vd_error,
2244 "cannot initialize '%s'", path);
2245 }
2246
2247 fnvlist_free(vdev_guids);
2248 fnvlist_free(guids_to_paths);
2249
2250 if (vd_errlist != NULL) {
2251 fnvlist_free(vd_errlist);
2252 return (-1);
2253 }
2254
2255 return (zpool_standard_error(zhp->zpool_hdl, err, msg));
2256 }
2257
2258 static int
xlate_trim_err(int err)2259 xlate_trim_err(int err)
2260 {
2261 switch (err) {
2262 case ENODEV:
2263 return (EZFS_NODEVICE);
2264 case EINVAL:
2265 case EROFS:
2266 return (EZFS_BADDEV);
2267 case EBUSY:
2268 return (EZFS_TRIMMING);
2269 case ESRCH:
2270 return (EZFS_NO_TRIM);
2271 case EOPNOTSUPP:
2272 return (EZFS_TRIM_NOTSUP);
2273 }
2274 return (err);
2275 }
2276
2277 /*
2278 * Begin, suspend, or cancel the TRIM (discarding of all free blocks) for
2279 * the given vdevs in the given pool.
2280 */
2281 int
zpool_trim(zpool_handle_t * zhp,pool_trim_func_t cmd_type,nvlist_t * vds,trimflags_t * trim_flags)2282 zpool_trim(zpool_handle_t *zhp, pool_trim_func_t cmd_type, nvlist_t *vds,
2283 trimflags_t *trim_flags)
2284 {
2285 char msg[1024];
2286 int err;
2287
2288 nvlist_t *vdev_guids = fnvlist_alloc();
2289 nvlist_t *guids_to_paths = fnvlist_alloc();
2290 nvlist_t *vd_errlist = NULL;
2291 nvlist_t *errlist;
2292 nvpair_t *elem;
2293
2294 err = zpool_translate_vdev_guids(zhp, vds, vdev_guids,
2295 guids_to_paths, &vd_errlist);
2296 if (err == 0) {
2297 err = lzc_trim(zhp->zpool_name, cmd_type, trim_flags->rate,
2298 trim_flags->secure, vdev_guids, &errlist);
2299 if (err == 0) {
2300 fnvlist_free(vdev_guids);
2301 fnvlist_free(guids_to_paths);
2302 return (0);
2303 }
2304
2305 if (errlist != NULL) {
2306 vd_errlist = fnvlist_lookup_nvlist(errlist,
2307 ZPOOL_TRIM_VDEVS);
2308 }
2309
2310 (void) snprintf(msg, sizeof (msg),
2311 dgettext(TEXT_DOMAIN, "operation failed"));
2312 } else {
2313 verify(vd_errlist != NULL);
2314 }
2315
2316 for (elem = nvlist_next_nvpair(vd_errlist, NULL);
2317 elem != NULL; elem = nvlist_next_nvpair(vd_errlist, elem)) {
2318 int64_t vd_error = xlate_trim_err(fnvpair_value_int64(elem));
2319 char *path;
2320 /*
2321 * If only the pool was specified, and it was not a secure
2322 * trim then suppress warnings for individual vdevs which
2323 * do not support trimming.
2324 */
2325 if (vd_error == EZFS_TRIM_NOTSUP &&
2326 trim_flags->fullpool &&
2327 !trim_flags->secure) {
2328 continue;
2329 }
2330
2331 if (nvlist_lookup_string(guids_to_paths, nvpair_name(elem),
2332 &path) != 0)
2333 path = nvpair_name(elem);
2334
2335 (void) zfs_error_fmt(zhp->zpool_hdl, vd_error,
2336 "cannot trim '%s'", path);
2337 }
2338
2339 fnvlist_free(vdev_guids);
2340 fnvlist_free(guids_to_paths);
2341
2342 if (vd_errlist != NULL) {
2343 fnvlist_free(vd_errlist);
2344 return (-1);
2345 }
2346
2347 return (zpool_standard_error(zhp->zpool_hdl, err, msg));
2348 }
2349
2350 /*
2351 * This provides a very minimal check whether a given string is likely a
2352 * c#t#d# style string. Users of this are expected to do their own
2353 * verification of the s# part.
2354 */
2355 #define CTD_CHECK(str) (str && str[0] == 'c' && isdigit(str[1]))
2356
2357 /*
2358 * More elaborate version for ones which may start with "/dev/dsk/"
2359 * and the like.
2360 */
2361 static int
ctd_check_path(char * str)2362 ctd_check_path(char *str)
2363 {
2364 /*
2365 * If it starts with a slash, check the last component.
2366 */
2367 if (str && str[0] == '/') {
2368 char *tmp = strrchr(str, '/');
2369
2370 /*
2371 * If it ends in "/old", check the second-to-last
2372 * component of the string instead.
2373 */
2374 if (tmp != str && strcmp(tmp, "/old") == 0) {
2375 for (tmp--; *tmp != '/'; tmp--)
2376 ;
2377 }
2378 str = tmp + 1;
2379 }
2380 return (CTD_CHECK(str));
2381 }
2382
2383 /*
2384 * Find a vdev that matches the search criteria specified. We use the
2385 * the nvpair name to determine how we should look for the device.
2386 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
2387 * spare; but FALSE if its an INUSE spare.
2388 */
2389 static nvlist_t *
vdev_to_nvlist_iter(nvlist_t * nv,nvlist_t * search,boolean_t * avail_spare,boolean_t * l2cache,boolean_t * log)2390 vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare,
2391 boolean_t *l2cache, boolean_t *log)
2392 {
2393 uint_t c, children;
2394 nvlist_t **child;
2395 nvlist_t *ret;
2396 uint64_t is_log;
2397 char *srchkey;
2398 nvpair_t *pair = nvlist_next_nvpair(search, NULL);
2399
2400 /* Nothing to look for */
2401 if (search == NULL || pair == NULL)
2402 return (NULL);
2403
2404 /* Obtain the key we will use to search */
2405 srchkey = nvpair_name(pair);
2406
2407 switch (nvpair_type(pair)) {
2408 case DATA_TYPE_UINT64:
2409 if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) {
2410 uint64_t srchval, theguid;
2411
2412 verify(nvpair_value_uint64(pair, &srchval) == 0);
2413 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
2414 &theguid) == 0);
2415 if (theguid == srchval)
2416 return (nv);
2417 }
2418 break;
2419
2420 case DATA_TYPE_STRING: {
2421 char *srchval, *val;
2422
2423 verify(nvpair_value_string(pair, &srchval) == 0);
2424 if (nvlist_lookup_string(nv, srchkey, &val) != 0)
2425 break;
2426
2427 /*
2428 * Search for the requested value. Special cases:
2429 *
2430 * - ZPOOL_CONFIG_PATH for whole disk entries. To support
2431 * UEFI boot, these end in "s0" or "s0/old" or "s1" or
2432 * "s1/old". The "s0" or "s1" part is hidden from the user,
2433 * but included in the string, so this matches around it.
2434 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE).
2435 *
2436 * Otherwise, all other searches are simple string compares.
2437 */
2438 if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0 &&
2439 ctd_check_path(val)) {
2440 uint64_t wholedisk = 0;
2441
2442 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
2443 &wholedisk);
2444 if (wholedisk) {
2445 int slen = strlen(srchval);
2446 int vlen = strlen(val);
2447
2448 if (slen != vlen - 2)
2449 break;
2450
2451 /*
2452 * make_leaf_vdev() should only set
2453 * wholedisk for ZPOOL_CONFIG_PATHs which
2454 * will include "/dev/dsk/", giving plenty of
2455 * room for the indices used next.
2456 */
2457 ASSERT(vlen >= 6);
2458
2459 /*
2460 * strings identical except trailing "s0"
2461 */
2462 if ((strcmp(&val[vlen - 2], "s0") == 0 ||
2463 strcmp(&val[vlen - 2], "s1") == 0) &&
2464 strncmp(srchval, val, slen) == 0)
2465 return (nv);
2466
2467 /*
2468 * strings identical except trailing "s0/old"
2469 */
2470 if ((strcmp(&val[vlen - 6], "s0/old") == 0 ||
2471 strcmp(&val[vlen - 6], "s1/old") == 0) &&
2472 strcmp(&srchval[slen - 4], "/old") == 0 &&
2473 strncmp(srchval, val, slen - 4) == 0)
2474 return (nv);
2475
2476 break;
2477 }
2478 } else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0 && val) {
2479 char *type, *idx, *end, *p;
2480 uint64_t id, vdev_id;
2481
2482 /*
2483 * Determine our vdev type, keeping in mind
2484 * that the srchval is composed of a type and
2485 * vdev id pair (i.e. mirror-4).
2486 */
2487 if ((type = strdup(srchval)) == NULL)
2488 return (NULL);
2489
2490 if ((p = strrchr(type, '-')) == NULL) {
2491 free(type);
2492 break;
2493 }
2494 idx = p + 1;
2495 *p = '\0';
2496
2497 /*
2498 * If the types don't match then keep looking.
2499 */
2500 if (strncmp(val, type, strlen(val)) != 0) {
2501 free(type);
2502 break;
2503 }
2504
2505 verify(zpool_vdev_is_interior(type));
2506 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
2507 &id) == 0);
2508
2509 errno = 0;
2510 vdev_id = strtoull(idx, &end, 10);
2511
2512 free(type);
2513 if (errno != 0)
2514 return (NULL);
2515
2516 /*
2517 * Now verify that we have the correct vdev id.
2518 */
2519 if (vdev_id == id)
2520 return (nv);
2521 }
2522
2523 /*
2524 * Common case
2525 */
2526 if (strcmp(srchval, val) == 0)
2527 return (nv);
2528 break;
2529 }
2530
2531 default:
2532 break;
2533 }
2534
2535 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
2536 &child, &children) != 0)
2537 return (NULL);
2538
2539 for (c = 0; c < children; c++) {
2540 if ((ret = vdev_to_nvlist_iter(child[c], search,
2541 avail_spare, l2cache, NULL)) != NULL) {
2542 /*
2543 * The 'is_log' value is only set for the toplevel
2544 * vdev, not the leaf vdevs. So we always lookup the
2545 * log device from the root of the vdev tree (where
2546 * 'log' is non-NULL).
2547 */
2548 if (log != NULL &&
2549 nvlist_lookup_uint64(child[c],
2550 ZPOOL_CONFIG_IS_LOG, &is_log) == 0 &&
2551 is_log) {
2552 *log = B_TRUE;
2553 }
2554 return (ret);
2555 }
2556 }
2557
2558 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
2559 &child, &children) == 0) {
2560 for (c = 0; c < children; c++) {
2561 if ((ret = vdev_to_nvlist_iter(child[c], search,
2562 avail_spare, l2cache, NULL)) != NULL) {
2563 *avail_spare = B_TRUE;
2564 return (ret);
2565 }
2566 }
2567 }
2568
2569 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
2570 &child, &children) == 0) {
2571 for (c = 0; c < children; c++) {
2572 if ((ret = vdev_to_nvlist_iter(child[c], search,
2573 avail_spare, l2cache, NULL)) != NULL) {
2574 *l2cache = B_TRUE;
2575 return (ret);
2576 }
2577 }
2578 }
2579
2580 return (NULL);
2581 }
2582
2583 /*
2584 * Given a physical path (minus the "/devices" prefix), find the
2585 * associated vdev.
2586 */
2587 nvlist_t *
zpool_find_vdev_by_physpath(zpool_handle_t * zhp,const char * ppath,boolean_t * avail_spare,boolean_t * l2cache,boolean_t * log)2588 zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath,
2589 boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log)
2590 {
2591 nvlist_t *search, *nvroot, *ret;
2592
2593 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2594 verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath) == 0);
2595
2596 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
2597 &nvroot) == 0);
2598
2599 *avail_spare = B_FALSE;
2600 *l2cache = B_FALSE;
2601 if (log != NULL)
2602 *log = B_FALSE;
2603 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
2604 nvlist_free(search);
2605
2606 return (ret);
2607 }
2608
2609 /*
2610 * Determine if we have an "interior" top-level vdev (i.e mirror/raidz).
2611 */
2612 static boolean_t
zpool_vdev_is_interior(const char * name)2613 zpool_vdev_is_interior(const char *name)
2614 {
2615 if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 ||
2616 strncmp(name, VDEV_TYPE_SPARE, strlen(VDEV_TYPE_SPARE)) == 0 ||
2617 strncmp(name,
2618 VDEV_TYPE_REPLACING, strlen(VDEV_TYPE_REPLACING)) == 0 ||
2619 strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0)
2620 return (B_TRUE);
2621 return (B_FALSE);
2622 }
2623
2624 nvlist_t *
zpool_find_vdev(zpool_handle_t * zhp,const char * path,boolean_t * avail_spare,boolean_t * l2cache,boolean_t * log)2625 zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare,
2626 boolean_t *l2cache, boolean_t *log)
2627 {
2628 char buf[MAXPATHLEN];
2629 char *end;
2630 nvlist_t *nvroot, *search, *ret;
2631 uint64_t guid;
2632
2633 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2634
2635 guid = strtoull(path, &end, 10);
2636 if (guid != 0 && *end == '\0') {
2637 verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0);
2638 } else if (zpool_vdev_is_interior(path)) {
2639 verify(nvlist_add_string(search, ZPOOL_CONFIG_TYPE, path) == 0);
2640 } else if (path[0] != '/') {
2641 (void) snprintf(buf, sizeof (buf), "%s/%s", ZFS_DISK_ROOT,
2642 path);
2643 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, buf) == 0);
2644 } else {
2645 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0);
2646 }
2647
2648 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
2649 &nvroot) == 0);
2650
2651 *avail_spare = B_FALSE;
2652 *l2cache = B_FALSE;
2653 if (log != NULL)
2654 *log = B_FALSE;
2655 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
2656 nvlist_free(search);
2657
2658 return (ret);
2659 }
2660
2661 static int
vdev_is_online(nvlist_t * nv)2662 vdev_is_online(nvlist_t *nv)
2663 {
2664 uint64_t ival;
2665
2666 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 ||
2667 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 ||
2668 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0)
2669 return (0);
2670
2671 return (1);
2672 }
2673
2674 /*
2675 * Helper function for zpool_get_physpaths().
2676 */
2677 static int
vdev_get_one_physpath(nvlist_t * config,char * physpath,size_t physpath_size,size_t * bytes_written)2678 vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size,
2679 size_t *bytes_written)
2680 {
2681 size_t bytes_left, pos, rsz;
2682 char *tmppath;
2683 const char *format;
2684
2685 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH,
2686 &tmppath) != 0)
2687 return (EZFS_NODEVICE);
2688
2689 pos = *bytes_written;
2690 bytes_left = physpath_size - pos;
2691 format = (pos == 0) ? "%s" : " %s";
2692
2693 rsz = snprintf(physpath + pos, bytes_left, format, tmppath);
2694 *bytes_written += rsz;
2695
2696 if (rsz >= bytes_left) {
2697 /* if physpath was not copied properly, clear it */
2698 if (bytes_left != 0) {
2699 physpath[pos] = 0;
2700 }
2701 return (EZFS_NOSPC);
2702 }
2703 return (0);
2704 }
2705
2706 static int
vdev_get_physpaths(nvlist_t * nv,char * physpath,size_t phypath_size,size_t * rsz,boolean_t is_spare)2707 vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size,
2708 size_t *rsz, boolean_t is_spare)
2709 {
2710 char *type;
2711 int ret;
2712
2713 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0)
2714 return (EZFS_INVALCONFIG);
2715
2716 if (strcmp(type, VDEV_TYPE_DISK) == 0) {
2717 /*
2718 * An active spare device has ZPOOL_CONFIG_IS_SPARE set.
2719 * For a spare vdev, we only want to boot from the active
2720 * spare device.
2721 */
2722 if (is_spare) {
2723 uint64_t spare = 0;
2724 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE,
2725 &spare);
2726 if (!spare)
2727 return (EZFS_INVALCONFIG);
2728 }
2729
2730 if (vdev_is_online(nv)) {
2731 if ((ret = vdev_get_one_physpath(nv, physpath,
2732 phypath_size, rsz)) != 0)
2733 return (ret);
2734 }
2735 } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 ||
2736 strcmp(type, VDEV_TYPE_RAIDZ) == 0 ||
2737 strcmp(type, VDEV_TYPE_REPLACING) == 0 ||
2738 (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) {
2739 nvlist_t **child;
2740 uint_t count;
2741 int i, ret;
2742
2743 if (nvlist_lookup_nvlist_array(nv,
2744 ZPOOL_CONFIG_CHILDREN, &child, &count) != 0)
2745 return (EZFS_INVALCONFIG);
2746
2747 for (i = 0; i < count; i++) {
2748 ret = vdev_get_physpaths(child[i], physpath,
2749 phypath_size, rsz, is_spare);
2750 if (ret == EZFS_NOSPC)
2751 return (ret);
2752 }
2753 }
2754
2755 return (EZFS_POOL_INVALARG);
2756 }
2757
2758 /*
2759 * Get phys_path for a root pool config.
2760 * Return 0 on success; non-zero on failure.
2761 */
2762 static int
zpool_get_config_physpath(nvlist_t * config,char * physpath,size_t phypath_size)2763 zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size)
2764 {
2765 size_t rsz;
2766 nvlist_t *vdev_root;
2767 nvlist_t **child;
2768 uint_t count;
2769 char *type;
2770
2771 rsz = 0;
2772
2773 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
2774 &vdev_root) != 0)
2775 return (EZFS_INVALCONFIG);
2776
2777 if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 ||
2778 nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN,
2779 &child, &count) != 0)
2780 return (EZFS_INVALCONFIG);
2781
2782 /*
2783 * root pool can only have a single top-level vdev.
2784 */
2785 if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1)
2786 return (EZFS_POOL_INVALARG);
2787
2788 (void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz,
2789 B_FALSE);
2790
2791 /* No online devices */
2792 if (rsz == 0)
2793 return (EZFS_NODEVICE);
2794
2795 return (0);
2796 }
2797
2798 /*
2799 * Get phys_path for a root pool
2800 * Return 0 on success; non-zero on failure.
2801 */
2802 int
zpool_get_physpath(zpool_handle_t * zhp,char * physpath,size_t phypath_size)2803 zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size)
2804 {
2805 return (zpool_get_config_physpath(zhp->zpool_config, physpath,
2806 phypath_size));
2807 }
2808
2809 /*
2810 * If the device has being dynamically expanded then we need to relabel
2811 * the disk to use the new unallocated space.
2812 */
2813 static int
zpool_relabel_disk(libzfs_handle_t * hdl,const char * name,const char * msg)2814 zpool_relabel_disk(libzfs_handle_t *hdl, const char *name, const char *msg)
2815 {
2816 char path[MAXPATHLEN];
2817 int fd, error;
2818 int (*_efi_use_whole_disk)(int);
2819 char drv[MODMAXNAMELEN];
2820 major_t maj;
2821 struct stat st;
2822
2823 if ((_efi_use_whole_disk = (int (*)(int))dlsym(RTLD_DEFAULT,
2824 "efi_use_whole_disk")) == NULL)
2825 return (-1);
2826
2827 (void) snprintf(path, sizeof (path), "%s/%s", ZFS_RDISK_ROOT, name);
2828
2829 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) {
2830 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2831 "relabel '%s': unable to open device"), name);
2832 return (zfs_error(hdl, EZFS_OPENFAILED, msg));
2833 }
2834
2835 /*
2836 * It's possible that we might encounter an error if the device
2837 * does not have any unallocated space left. If so, we simply
2838 * ignore that error and continue on.
2839 */
2840 error = _efi_use_whole_disk(fd);
2841 if (error && error != VT_ENOSPC) {
2842 (void) close(fd);
2843 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2844 "relabel '%s': unable to read disk capacity"), name);
2845 return (zfs_error(hdl, EZFS_NOCAP, msg));
2846 }
2847
2848 /*
2849 * Writing a new EFI partition table to the disk will have marked
2850 * the geometry as needing re-validation. Before returning, force
2851 * it to be checked by querying the device state, otherwise the
2852 * subsequent vdev_reopen() will very likely fail to read the device
2853 * size, faulting the pool.
2854 *
2855 * The dkio(4I) ioctls are implemented by the disk driver rather than
2856 * some generic framework, so we limit its use here to drivers with
2857 * which it has been tested.
2858 */
2859 if (fstat(fd, &st) == 0 &&
2860 (maj = major(st.st_rdev)) != (major_t)NODEV &&
2861 modctl(MODGETNAME, drv, sizeof (drv), &maj) == 0 &&
2862 (strcmp(drv, "blkdev") == 0 || strcmp(drv, "sd") == 0)) {
2863 enum dkio_state dkst = DKIO_NONE;
2864 (void) ioctl(fd, DKIOCSTATE, &dkst);
2865 }
2866
2867 (void) close(fd);
2868
2869 return (0);
2870 }
2871
2872 /*
2873 * Bring the specified vdev online. The 'flags' parameter is a set of the
2874 * ZFS_ONLINE_* flags.
2875 */
2876 int
zpool_vdev_online(zpool_handle_t * zhp,const char * path,int flags,vdev_state_t * newstate)2877 zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags,
2878 vdev_state_t *newstate)
2879 {
2880 zfs_cmd_t zc = { 0 };
2881 char msg[1024];
2882 char *pathname;
2883 nvlist_t *tgt;
2884 boolean_t avail_spare, l2cache, islog;
2885 libzfs_handle_t *hdl = zhp->zpool_hdl;
2886 int error;
2887
2888 if (flags & ZFS_ONLINE_EXPAND) {
2889 (void) snprintf(msg, sizeof (msg),
2890 dgettext(TEXT_DOMAIN, "cannot expand %s"), path);
2891 } else {
2892 (void) snprintf(msg, sizeof (msg),
2893 dgettext(TEXT_DOMAIN, "cannot online %s"), path);
2894 }
2895
2896 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2897 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2898 &islog)) == NULL)
2899 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2900
2901 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2902
2903 if (avail_spare)
2904 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2905
2906 if ((flags & ZFS_ONLINE_EXPAND ||
2907 zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) &&
2908 nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH, &pathname) == 0) {
2909 uint64_t wholedisk = 0;
2910
2911 (void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK,
2912 &wholedisk);
2913
2914 /*
2915 * XXX - L2ARC 1.0 devices can't support expansion.
2916 */
2917 if (l2cache) {
2918 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2919 "cannot expand cache devices"));
2920 return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg));
2921 }
2922
2923 if (wholedisk) {
2924 pathname += strlen(ZFS_DISK_ROOT) + 1;
2925 error = zpool_relabel_disk(hdl, pathname, msg);
2926 if (error != 0)
2927 return (error);
2928 }
2929 }
2930
2931 zc.zc_cookie = VDEV_STATE_ONLINE;
2932 zc.zc_obj = flags;
2933
2934 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) {
2935 if (errno == EINVAL) {
2936 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split "
2937 "from this pool into a new one. Use '%s' "
2938 "instead"), "zpool detach");
2939 return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, msg));
2940 }
2941 return (zpool_standard_error(hdl, errno, msg));
2942 }
2943
2944 *newstate = zc.zc_cookie;
2945 return (0);
2946 }
2947
2948 /*
2949 * Take the specified vdev offline
2950 */
2951 int
zpool_vdev_offline(zpool_handle_t * zhp,const char * path,boolean_t istmp)2952 zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp)
2953 {
2954 zfs_cmd_t zc = { 0 };
2955 char msg[1024];
2956 nvlist_t *tgt;
2957 boolean_t avail_spare, l2cache;
2958 libzfs_handle_t *hdl = zhp->zpool_hdl;
2959
2960 (void) snprintf(msg, sizeof (msg),
2961 dgettext(TEXT_DOMAIN, "cannot offline %s"), path);
2962
2963 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2964 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2965 NULL)) == NULL)
2966 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2967
2968 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2969
2970 if (avail_spare)
2971 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2972
2973 zc.zc_cookie = VDEV_STATE_OFFLINE;
2974 zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0;
2975
2976 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2977 return (0);
2978
2979 switch (errno) {
2980 case EBUSY:
2981
2982 /*
2983 * There are no other replicas of this device.
2984 */
2985 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2986
2987 case EEXIST:
2988 /*
2989 * The log device has unplayed logs
2990 */
2991 return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg));
2992
2993 default:
2994 return (zpool_standard_error(hdl, errno, msg));
2995 }
2996 }
2997
2998 /*
2999 * Mark the given vdev faulted.
3000 */
3001 int
zpool_vdev_fault(zpool_handle_t * zhp,uint64_t guid,vdev_aux_t aux)3002 zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
3003 {
3004 zfs_cmd_t zc = { 0 };
3005 char msg[1024];
3006 libzfs_handle_t *hdl = zhp->zpool_hdl;
3007
3008 (void) snprintf(msg, sizeof (msg),
3009 dgettext(TEXT_DOMAIN, "cannot fault %llu"), guid);
3010
3011 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3012 zc.zc_guid = guid;
3013 zc.zc_cookie = VDEV_STATE_FAULTED;
3014 zc.zc_obj = aux;
3015
3016 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
3017 return (0);
3018
3019 switch (errno) {
3020 case EBUSY:
3021
3022 /*
3023 * There are no other replicas of this device.
3024 */
3025 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
3026
3027 default:
3028 return (zpool_standard_error(hdl, errno, msg));
3029 }
3030
3031 }
3032
3033 /*
3034 * Mark the given vdev degraded.
3035 */
3036 int
zpool_vdev_degrade(zpool_handle_t * zhp,uint64_t guid,vdev_aux_t aux)3037 zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
3038 {
3039 zfs_cmd_t zc = { 0 };
3040 char msg[1024];
3041 libzfs_handle_t *hdl = zhp->zpool_hdl;
3042
3043 (void) snprintf(msg, sizeof (msg),
3044 dgettext(TEXT_DOMAIN, "cannot degrade %llu"), guid);
3045
3046 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3047 zc.zc_guid = guid;
3048 zc.zc_cookie = VDEV_STATE_DEGRADED;
3049 zc.zc_obj = aux;
3050
3051 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
3052 return (0);
3053
3054 return (zpool_standard_error(hdl, errno, msg));
3055 }
3056
3057 /*
3058 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as
3059 * a hot spare.
3060 */
3061 static boolean_t
is_replacing_spare(nvlist_t * search,nvlist_t * tgt,int which)3062 is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which)
3063 {
3064 nvlist_t **child;
3065 uint_t c, children;
3066 char *type;
3067
3068 if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child,
3069 &children) == 0) {
3070 verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE,
3071 &type) == 0);
3072
3073 if (strcmp(type, VDEV_TYPE_SPARE) == 0 &&
3074 children == 2 && child[which] == tgt)
3075 return (B_TRUE);
3076
3077 for (c = 0; c < children; c++)
3078 if (is_replacing_spare(child[c], tgt, which))
3079 return (B_TRUE);
3080 }
3081
3082 return (B_FALSE);
3083 }
3084
3085 /*
3086 * Attach new_disk (fully described by nvroot) to old_disk.
3087 * If 'replacing' is specified, the new disk will replace the old one.
3088 */
3089 int
zpool_vdev_attach(zpool_handle_t * zhp,const char * old_disk,const char * new_disk,nvlist_t * nvroot,int replacing)3090 zpool_vdev_attach(zpool_handle_t *zhp,
3091 const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing)
3092 {
3093 zfs_cmd_t zc = { 0 };
3094 char msg[1024];
3095 int ret;
3096 nvlist_t *tgt, *newvd;
3097 boolean_t avail_spare, l2cache, islog;
3098 uint64_t val;
3099 char *newname;
3100 nvlist_t **child;
3101 uint_t children;
3102 nvlist_t *config_root;
3103 libzfs_handle_t *hdl = zhp->zpool_hdl;
3104 boolean_t rootpool = zpool_is_bootable(zhp);
3105
3106 if (replacing)
3107 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
3108 "cannot replace %s with %s"), old_disk, new_disk);
3109 else
3110 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
3111 "cannot attach %s to %s"), new_disk, old_disk);
3112
3113 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3114 if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache,
3115 &islog)) == NULL)
3116 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3117
3118 if (avail_spare)
3119 return (zfs_error(hdl, EZFS_ISSPARE, msg));
3120
3121 if (l2cache)
3122 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
3123
3124 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
3125 zc.zc_cookie = replacing;
3126
3127 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
3128 &child, &children) != 0 || children != 1) {
3129 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3130 "new device must be a single disk"));
3131 return (zfs_error(hdl, EZFS_INVALCONFIG, msg));
3132 }
3133
3134 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
3135 ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0);
3136
3137 if ((newname = zpool_vdev_name(NULL, NULL, child[0], 0)) == NULL)
3138 return (-1);
3139
3140 newvd = zpool_find_vdev(zhp, newname, &avail_spare, &l2cache, NULL);
3141 /*
3142 * If the target is a hot spare that has been swapped in, we can only
3143 * replace it with another hot spare.
3144 */
3145 if (replacing &&
3146 nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 &&
3147 (newvd == NULL || !avail_spare) &&
3148 is_replacing_spare(config_root, tgt, 1)) {
3149 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3150 "can only be replaced by another hot spare"));
3151 free(newname);
3152 return (zfs_error(hdl, EZFS_BADTARGET, msg));
3153 }
3154
3155 free(newname);
3156
3157 if (replacing && avail_spare && !vdev_is_online(newvd)) {
3158 (void) zpool_standard_error(hdl, ENXIO, msg);
3159 return (-1);
3160 }
3161
3162 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
3163 return (-1);
3164
3165 ret = zfs_ioctl(hdl, ZFS_IOC_VDEV_ATTACH, &zc);
3166
3167 zcmd_free_nvlists(&zc);
3168
3169 if (ret == 0) {
3170 if (rootpool) {
3171 /*
3172 * XXX need a better way to prevent user from
3173 * booting up a half-baked vdev.
3174 */
3175 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make "
3176 "sure to wait until resilver is done "
3177 "before rebooting.\n"));
3178 }
3179 return (0);
3180 }
3181
3182 switch (errno) {
3183 case ENOTSUP:
3184 /*
3185 * Can't attach to or replace this type of vdev.
3186 */
3187 if (replacing) {
3188 uint64_t version = zpool_get_prop_int(zhp,
3189 ZPOOL_PROP_VERSION, NULL);
3190
3191 if (islog)
3192 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3193 "cannot replace a log with a spare"));
3194 else if (version >= SPA_VERSION_MULTI_REPLACE)
3195 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3196 "already in replacing/spare config; wait "
3197 "for completion or use 'zpool detach'"));
3198 else
3199 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3200 "cannot replace a replacing device"));
3201 } else {
3202 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3203 "can only attach to mirrors and top-level "
3204 "disks"));
3205 }
3206 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
3207 break;
3208
3209 case EINVAL:
3210 /*
3211 * The new device must be a single disk.
3212 */
3213 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3214 "new device must be a single disk"));
3215 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg);
3216 break;
3217
3218 case EBUSY:
3219 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy, "
3220 "or device removal is in progress"),
3221 new_disk);
3222 (void) zfs_error(hdl, EZFS_BADDEV, msg);
3223 break;
3224
3225 case EOVERFLOW:
3226 /*
3227 * The new device is too small.
3228 */
3229 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3230 "device is too small"));
3231 (void) zfs_error(hdl, EZFS_BADDEV, msg);
3232 break;
3233
3234 case EDOM:
3235 /*
3236 * The new device has a different optimal sector size.
3237 */
3238 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3239 "new device has a different optimal sector size; use the "
3240 "option '-o ashift=N' to override the optimal size"));
3241 (void) zfs_error(hdl, EZFS_BADDEV, msg);
3242 break;
3243
3244 case ENAMETOOLONG:
3245 /*
3246 * The resulting top-level vdev spec won't fit in the label.
3247 */
3248 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg);
3249 break;
3250
3251 default:
3252 (void) zpool_standard_error(hdl, errno, msg);
3253 }
3254
3255 return (-1);
3256 }
3257
3258 /*
3259 * Detach the specified device.
3260 */
3261 int
zpool_vdev_detach(zpool_handle_t * zhp,const char * path)3262 zpool_vdev_detach(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;
3268 libzfs_handle_t *hdl = zhp->zpool_hdl;
3269
3270 (void) snprintf(msg, sizeof (msg),
3271 dgettext(TEXT_DOMAIN, "cannot detach %s"), path);
3272
3273 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3274 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
3275 NULL)) == NULL)
3276 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3277
3278 if (avail_spare)
3279 return (zfs_error(hdl, EZFS_ISSPARE, msg));
3280
3281 if (l2cache)
3282 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
3283
3284 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
3285
3286 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0)
3287 return (0);
3288
3289 switch (errno) {
3290
3291 case ENOTSUP:
3292 /*
3293 * Can't detach from this type of vdev.
3294 */
3295 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only "
3296 "applicable to mirror and replacing vdevs"));
3297 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
3298 break;
3299
3300 case EBUSY:
3301 /*
3302 * There are no other replicas of this device.
3303 */
3304 (void) zfs_error(hdl, EZFS_NOREPLICAS, msg);
3305 break;
3306
3307 default:
3308 (void) zpool_standard_error(hdl, errno, msg);
3309 }
3310
3311 return (-1);
3312 }
3313
3314 /*
3315 * Find a mirror vdev in the source nvlist.
3316 *
3317 * The mchild array contains a list of disks in one of the top-level mirrors
3318 * of the source pool. The schild array contains a list of disks that the
3319 * user specified on the command line. We loop over the mchild array to
3320 * see if any entry in the schild array matches.
3321 *
3322 * If a disk in the mchild array is found in the schild array, we return
3323 * the index of that entry. Otherwise we return -1.
3324 */
3325 static int
find_vdev_entry(zpool_handle_t * zhp,nvlist_t ** mchild,uint_t mchildren,nvlist_t ** schild,uint_t schildren)3326 find_vdev_entry(zpool_handle_t *zhp, nvlist_t **mchild, uint_t mchildren,
3327 nvlist_t **schild, uint_t schildren)
3328 {
3329 uint_t mc;
3330
3331 for (mc = 0; mc < mchildren; mc++) {
3332 uint_t sc;
3333 char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp,
3334 mchild[mc], 0);
3335
3336 for (sc = 0; sc < schildren; sc++) {
3337 char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp,
3338 schild[sc], 0);
3339 boolean_t result = (strcmp(mpath, spath) == 0);
3340
3341 free(spath);
3342 if (result) {
3343 free(mpath);
3344 return (mc);
3345 }
3346 }
3347
3348 free(mpath);
3349 }
3350
3351 return (-1);
3352 }
3353
3354 /*
3355 * Split a mirror pool. If newroot points to null, then a new nvlist
3356 * is generated and it is the responsibility of the caller to free it.
3357 */
3358 int
zpool_vdev_split(zpool_handle_t * zhp,char * newname,nvlist_t ** newroot,nvlist_t * props,splitflags_t flags)3359 zpool_vdev_split(zpool_handle_t *zhp, char *newname, nvlist_t **newroot,
3360 nvlist_t *props, splitflags_t flags)
3361 {
3362 zfs_cmd_t zc = { 0 };
3363 char msg[1024];
3364 nvlist_t *tree, *config, **child, **newchild, *newconfig = NULL;
3365 nvlist_t **varray = NULL, *zc_props = NULL;
3366 uint_t c, children, newchildren, lastlog = 0, vcount, found = 0;
3367 libzfs_handle_t *hdl = zhp->zpool_hdl;
3368 uint64_t vers;
3369 boolean_t freelist = B_FALSE, memory_err = B_TRUE;
3370 int retval = 0;
3371
3372 (void) snprintf(msg, sizeof (msg),
3373 dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name);
3374
3375 if (!zpool_name_valid(hdl, B_FALSE, newname))
3376 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
3377
3378 if ((config = zpool_get_config(zhp, NULL)) == NULL) {
3379 (void) fprintf(stderr, gettext("Internal error: unable to "
3380 "retrieve pool configuration\n"));
3381 return (-1);
3382 }
3383
3384 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree)
3385 == 0);
3386 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &vers) == 0);
3387
3388 if (props) {
3389 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
3390 if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name,
3391 props, vers, flags, msg)) == NULL)
3392 return (-1);
3393 }
3394
3395 if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child,
3396 &children) != 0) {
3397 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3398 "Source pool is missing vdev tree"));
3399 nvlist_free(zc_props);
3400 return (-1);
3401 }
3402
3403 varray = zfs_alloc(hdl, children * sizeof (nvlist_t *));
3404 vcount = 0;
3405
3406 if (*newroot == NULL ||
3407 nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN,
3408 &newchild, &newchildren) != 0)
3409 newchildren = 0;
3410
3411 for (c = 0; c < children; c++) {
3412 uint64_t is_log = B_FALSE, is_hole = B_FALSE;
3413 char *type;
3414 nvlist_t **mchild, *vdev;
3415 uint_t mchildren;
3416 int entry;
3417
3418 /*
3419 * Unlike cache & spares, slogs are stored in the
3420 * ZPOOL_CONFIG_CHILDREN array. We filter them out here.
3421 */
3422 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
3423 &is_log);
3424 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
3425 &is_hole);
3426 if (is_log || is_hole) {
3427 /*
3428 * Create a hole vdev and put it in the config.
3429 */
3430 if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0)
3431 goto out;
3432 if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE,
3433 VDEV_TYPE_HOLE) != 0)
3434 goto out;
3435 if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE,
3436 1) != 0)
3437 goto out;
3438 if (lastlog == 0)
3439 lastlog = vcount;
3440 varray[vcount++] = vdev;
3441 continue;
3442 }
3443 lastlog = 0;
3444 verify(nvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE, &type)
3445 == 0);
3446 if (strcmp(type, VDEV_TYPE_MIRROR) != 0) {
3447 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3448 "Source pool must be composed only of mirrors\n"));
3449 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
3450 goto out;
3451 }
3452
3453 verify(nvlist_lookup_nvlist_array(child[c],
3454 ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0);
3455
3456 /* find or add an entry for this top-level vdev */
3457 if (newchildren > 0 &&
3458 (entry = find_vdev_entry(zhp, mchild, mchildren,
3459 newchild, newchildren)) >= 0) {
3460 /* We found a disk that the user specified. */
3461 vdev = mchild[entry];
3462 ++found;
3463 } else {
3464 /* User didn't specify a disk for this vdev. */
3465 vdev = mchild[mchildren - 1];
3466 }
3467
3468 if (nvlist_dup(vdev, &varray[vcount++], 0) != 0)
3469 goto out;
3470 }
3471
3472 /* did we find every disk the user specified? */
3473 if (found != newchildren) {
3474 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must "
3475 "include at most one disk from each mirror"));
3476 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
3477 goto out;
3478 }
3479
3480 /* Prepare the nvlist for populating. */
3481 if (*newroot == NULL) {
3482 if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0)
3483 goto out;
3484 freelist = B_TRUE;
3485 if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE,
3486 VDEV_TYPE_ROOT) != 0)
3487 goto out;
3488 } else {
3489 verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0);
3490 }
3491
3492 /* Add all the children we found */
3493 if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, varray,
3494 lastlog == 0 ? vcount : lastlog) != 0)
3495 goto out;
3496
3497 /*
3498 * If we're just doing a dry run, exit now with success.
3499 */
3500 if (flags.dryrun) {
3501 memory_err = B_FALSE;
3502 freelist = B_FALSE;
3503 goto out;
3504 }
3505
3506 /* now build up the config list & call the ioctl */
3507 if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0)
3508 goto out;
3509
3510 if (nvlist_add_nvlist(newconfig,
3511 ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 ||
3512 nvlist_add_string(newconfig,
3513 ZPOOL_CONFIG_POOL_NAME, newname) != 0 ||
3514 nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0)
3515 goto out;
3516
3517 /*
3518 * The new pool is automatically part of the namespace unless we
3519 * explicitly export it.
3520 */
3521 if (!flags.import)
3522 zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT;
3523 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3524 (void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string));
3525 if (zcmd_write_conf_nvlist(hdl, &zc, newconfig) != 0)
3526 goto out;
3527 if (zc_props != NULL && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
3528 goto out;
3529
3530 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) {
3531 retval = zpool_standard_error(hdl, errno, msg);
3532 goto out;
3533 }
3534
3535 freelist = B_FALSE;
3536 memory_err = B_FALSE;
3537
3538 out:
3539 if (varray != NULL) {
3540 int v;
3541
3542 for (v = 0; v < vcount; v++)
3543 nvlist_free(varray[v]);
3544 free(varray);
3545 }
3546 zcmd_free_nvlists(&zc);
3547 nvlist_free(zc_props);
3548 nvlist_free(newconfig);
3549 if (freelist) {
3550 nvlist_free(*newroot);
3551 *newroot = NULL;
3552 }
3553
3554 if (retval != 0)
3555 return (retval);
3556
3557 if (memory_err)
3558 return (no_memory(hdl));
3559
3560 return (0);
3561 }
3562
3563 /*
3564 * Remove the given device.
3565 */
3566 int
zpool_vdev_remove(zpool_handle_t * zhp,const char * path)3567 zpool_vdev_remove(zpool_handle_t *zhp, const char *path)
3568 {
3569 zfs_cmd_t zc = { 0 };
3570 char msg[1024];
3571 nvlist_t *tgt;
3572 boolean_t avail_spare, l2cache, islog;
3573 libzfs_handle_t *hdl = zhp->zpool_hdl;
3574 uint64_t version;
3575
3576 (void) snprintf(msg, sizeof (msg),
3577 dgettext(TEXT_DOMAIN, "cannot remove %s"), path);
3578
3579 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3580 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
3581 &islog)) == NULL)
3582 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3583
3584 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
3585 if (islog && version < SPA_VERSION_HOLES) {
3586 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3587 "pool must be upgraded to support log removal"));
3588 return (zfs_error(hdl, EZFS_BADVERSION, msg));
3589 }
3590
3591 zc.zc_guid = fnvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID);
3592
3593 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
3594 return (0);
3595
3596 switch (errno) {
3597
3598 case EINVAL:
3599 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3600 "invalid config; all top-level vdevs must "
3601 "have the same sector size and not be raidz."));
3602 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg);
3603 break;
3604
3605 case EBUSY:
3606 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3607 "Pool busy; removal may already be in progress"));
3608 (void) zfs_error(hdl, EZFS_BUSY, msg);
3609 break;
3610
3611 default:
3612 (void) zpool_standard_error(hdl, errno, msg);
3613 }
3614 return (-1);
3615 }
3616
3617 int
zpool_vdev_remove_cancel(zpool_handle_t * zhp)3618 zpool_vdev_remove_cancel(zpool_handle_t *zhp)
3619 {
3620 zfs_cmd_t zc = { 0 };
3621 char msg[1024];
3622 libzfs_handle_t *hdl = zhp->zpool_hdl;
3623
3624 (void) snprintf(msg, sizeof (msg),
3625 dgettext(TEXT_DOMAIN, "cannot cancel removal"));
3626
3627 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3628 zc.zc_cookie = 1;
3629
3630 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
3631 return (0);
3632
3633 return (zpool_standard_error(hdl, errno, msg));
3634 }
3635
3636 int
zpool_vdev_indirect_size(zpool_handle_t * zhp,const char * path,uint64_t * sizep)3637 zpool_vdev_indirect_size(zpool_handle_t *zhp, const char *path,
3638 uint64_t *sizep)
3639 {
3640 char msg[1024];
3641 nvlist_t *tgt;
3642 boolean_t avail_spare, l2cache, islog;
3643 libzfs_handle_t *hdl = zhp->zpool_hdl;
3644
3645 (void) snprintf(msg, sizeof (msg),
3646 dgettext(TEXT_DOMAIN, "cannot determine indirect size of %s"),
3647 path);
3648
3649 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
3650 &islog)) == NULL)
3651 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3652
3653 if (avail_spare || l2cache || islog) {
3654 *sizep = 0;
3655 return (0);
3656 }
3657
3658 if (nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_INDIRECT_SIZE, sizep) != 0) {
3659 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3660 "indirect size not available"));
3661 return (zfs_error(hdl, EINVAL, msg));
3662 }
3663 return (0);
3664 }
3665
3666 /*
3667 * Clear the errors for the pool, or the particular device if specified.
3668 */
3669 int
zpool_clear(zpool_handle_t * zhp,const char * path,nvlist_t * rewindnvl)3670 zpool_clear(zpool_handle_t *zhp, const char *path, nvlist_t *rewindnvl)
3671 {
3672 zfs_cmd_t zc = { 0 };
3673 char msg[1024];
3674 nvlist_t *tgt;
3675 zpool_load_policy_t policy;
3676 boolean_t avail_spare, l2cache;
3677 libzfs_handle_t *hdl = zhp->zpool_hdl;
3678 nvlist_t *nvi = NULL;
3679 int error;
3680
3681 if (path)
3682 (void) snprintf(msg, sizeof (msg),
3683 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
3684 path);
3685 else
3686 (void) snprintf(msg, sizeof (msg),
3687 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
3688 zhp->zpool_name);
3689
3690 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3691 if (path) {
3692 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare,
3693 &l2cache, NULL)) == NULL)
3694 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3695
3696 /*
3697 * Don't allow error clearing for hot spares. Do allow
3698 * error clearing for l2cache devices.
3699 */
3700 if (avail_spare)
3701 return (zfs_error(hdl, EZFS_ISSPARE, msg));
3702
3703 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID,
3704 &zc.zc_guid) == 0);
3705 }
3706
3707 zpool_get_load_policy(rewindnvl, &policy);
3708 zc.zc_cookie = policy.zlp_rewind;
3709
3710 if (zcmd_alloc_dst_nvlist(hdl, &zc, zhp->zpool_config_size * 2) != 0)
3711 return (-1);
3712
3713 if (zcmd_write_src_nvlist(hdl, &zc, rewindnvl) != 0)
3714 return (-1);
3715
3716 while ((error = zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc)) != 0 &&
3717 errno == ENOMEM) {
3718 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
3719 zcmd_free_nvlists(&zc);
3720 return (-1);
3721 }
3722 }
3723
3724 if (!error || ((policy.zlp_rewind & ZPOOL_TRY_REWIND) &&
3725 errno != EPERM && errno != EACCES)) {
3726 if (policy.zlp_rewind &
3727 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
3728 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi);
3729 zpool_rewind_exclaim(hdl, zc.zc_name,
3730 ((policy.zlp_rewind & ZPOOL_TRY_REWIND) != 0),
3731 nvi);
3732 nvlist_free(nvi);
3733 }
3734 zcmd_free_nvlists(&zc);
3735 return (0);
3736 }
3737
3738 zcmd_free_nvlists(&zc);
3739 return (zpool_standard_error(hdl, errno, msg));
3740 }
3741
3742 /*
3743 * Similar to zpool_clear(), but takes a GUID (used by fmd).
3744 */
3745 int
zpool_vdev_clear(zpool_handle_t * zhp,uint64_t guid)3746 zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid)
3747 {
3748 zfs_cmd_t zc = { 0 };
3749 char msg[1024];
3750 libzfs_handle_t *hdl = zhp->zpool_hdl;
3751
3752 (void) snprintf(msg, sizeof (msg),
3753 dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"),
3754 guid);
3755
3756 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3757 zc.zc_guid = guid;
3758 zc.zc_cookie = ZPOOL_NO_REWIND;
3759
3760 if (ioctl(hdl->libzfs_fd, ZFS_IOC_CLEAR, &zc) == 0)
3761 return (0);
3762
3763 return (zpool_standard_error(hdl, errno, msg));
3764 }
3765
3766 /*
3767 * Change the GUID for a pool.
3768 */
3769 int
zpool_reguid(zpool_handle_t * zhp)3770 zpool_reguid(zpool_handle_t *zhp)
3771 {
3772 char msg[1024];
3773 libzfs_handle_t *hdl = zhp->zpool_hdl;
3774 zfs_cmd_t zc = { 0 };
3775
3776 (void) snprintf(msg, sizeof (msg),
3777 dgettext(TEXT_DOMAIN, "cannot reguid '%s'"), zhp->zpool_name);
3778
3779 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3780 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REGUID, &zc) == 0)
3781 return (0);
3782
3783 return (zpool_standard_error(hdl, errno, msg));
3784 }
3785
3786 /*
3787 * Reopen the pool.
3788 */
3789 int
zpool_reopen(zpool_handle_t * zhp)3790 zpool_reopen(zpool_handle_t *zhp)
3791 {
3792 zfs_cmd_t zc = { 0 };
3793 char msg[1024];
3794 libzfs_handle_t *hdl = zhp->zpool_hdl;
3795
3796 (void) snprintf(msg, sizeof (msg),
3797 dgettext(TEXT_DOMAIN, "cannot reopen '%s'"),
3798 zhp->zpool_name);
3799
3800 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3801 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REOPEN, &zc) == 0)
3802 return (0);
3803 return (zpool_standard_error(hdl, errno, msg));
3804 }
3805
3806 /* call into libzfs_core to execute the sync IOCTL per pool */
3807 int
zpool_sync_one(zpool_handle_t * zhp,void * data)3808 zpool_sync_one(zpool_handle_t *zhp, void *data)
3809 {
3810 int ret;
3811 libzfs_handle_t *hdl = zpool_get_handle(zhp);
3812 const char *pool_name = zpool_get_name(zhp);
3813 boolean_t *force = data;
3814 nvlist_t *innvl = fnvlist_alloc();
3815
3816 fnvlist_add_boolean_value(innvl, "force", *force);
3817 if ((ret = lzc_sync(pool_name, innvl, NULL)) != 0) {
3818 nvlist_free(innvl);
3819 return (zpool_standard_error_fmt(hdl, ret,
3820 dgettext(TEXT_DOMAIN, "sync '%s' failed"), pool_name));
3821 }
3822 nvlist_free(innvl);
3823
3824 return (0);
3825 }
3826
3827 /*
3828 * Convert from a devid string to a path.
3829 */
3830 static char *
devid_to_path(char * devid_str)3831 devid_to_path(char *devid_str)
3832 {
3833 ddi_devid_t devid;
3834 char *minor;
3835 char *path;
3836 devid_nmlist_t *list = NULL;
3837 int ret;
3838
3839 if (devid_str_decode(devid_str, &devid, &minor) != 0)
3840 return (NULL);
3841
3842 ret = devid_deviceid_to_nmlist("/dev", devid, minor, &list);
3843
3844 devid_str_free(minor);
3845 devid_free(devid);
3846
3847 if (ret != 0)
3848 return (NULL);
3849
3850 /*
3851 * In a case the strdup() fails, we will just return NULL below.
3852 */
3853 path = strdup(list[0].devname);
3854
3855 devid_free_nmlist(list);
3856
3857 return (path);
3858 }
3859
3860 /*
3861 * Convert from a path to a devid string.
3862 */
3863 static char *
path_to_devid(const char * path)3864 path_to_devid(const char *path)
3865 {
3866 int fd;
3867 ddi_devid_t devid;
3868 char *minor, *ret;
3869
3870 if ((fd = open(path, O_RDONLY)) < 0)
3871 return (NULL);
3872
3873 minor = NULL;
3874 ret = NULL;
3875 if (devid_get(fd, &devid) == 0) {
3876 if (devid_get_minor_name(fd, &minor) == 0)
3877 ret = devid_str_encode(devid, minor);
3878 if (minor != NULL)
3879 devid_str_free(minor);
3880 devid_free(devid);
3881 }
3882 (void) close(fd);
3883
3884 return (ret);
3885 }
3886
3887 struct path_from_physpath_walker_args {
3888 char *pfpwa_path;
3889 };
3890
3891 /*
3892 * Walker for use with di_devlink_walk(). Stores the "/dev" path of the first
3893 * primary devlink (i.e., the first devlink which refers to our "/devices"
3894 * node) and stops walking.
3895 */
3896 static int
path_from_physpath_walker(di_devlink_t devlink,void * arg)3897 path_from_physpath_walker(di_devlink_t devlink, void *arg)
3898 {
3899 struct path_from_physpath_walker_args *pfpwa = arg;
3900
3901 if (di_devlink_type(devlink) != DI_PRIMARY_LINK) {
3902 return (DI_WALK_CONTINUE);
3903 }
3904
3905 verify(pfpwa->pfpwa_path == NULL);
3906 if ((pfpwa->pfpwa_path = strdup(di_devlink_path(devlink))) != NULL) {
3907 return (DI_WALK_TERMINATE);
3908 }
3909
3910 return (DI_WALK_CONTINUE);
3911 }
3912
3913 /*
3914 * Search for a "/dev" path that refers to our physical path. Returns the new
3915 * path if one is found and it does not match the existing "path" value. If
3916 * the value is unchanged, or one could not be found, returns NULL.
3917 */
3918 static char *
path_from_physpath(libzfs_handle_t * hdl,const char * path,const char * physpath)3919 path_from_physpath(libzfs_handle_t *hdl, const char *path,
3920 const char *physpath)
3921 {
3922 struct path_from_physpath_walker_args pfpwa;
3923
3924 if (physpath == NULL) {
3925 return (NULL);
3926 }
3927
3928 if (hdl->libzfs_devlink == NULL) {
3929 if ((hdl->libzfs_devlink = di_devlink_init(NULL, 0)) ==
3930 DI_LINK_NIL) {
3931 /*
3932 * We may not be able to open a handle if this process
3933 * is insufficiently privileged, or we are too early in
3934 * boot for devfsadm to be ready. Ignore this error
3935 * and defer the path check to a subsequent run.
3936 */
3937 return (NULL);
3938 }
3939 }
3940
3941 pfpwa.pfpwa_path = NULL;
3942 (void) di_devlink_walk(hdl->libzfs_devlink, NULL, physpath,
3943 DI_PRIMARY_LINK, &pfpwa, path_from_physpath_walker);
3944
3945 if (path != NULL && pfpwa.pfpwa_path != NULL &&
3946 strcmp(path, pfpwa.pfpwa_path) == 0) {
3947 /*
3948 * If the path is already correct, no change is required.
3949 */
3950 free(pfpwa.pfpwa_path);
3951 return (NULL);
3952 }
3953
3954 return (pfpwa.pfpwa_path);
3955 }
3956
3957 /*
3958 * Issue the necessary ioctl() to update the stored path value for the vdev. We
3959 * ignore any failure here, since a common case is for an unprivileged user to
3960 * type 'zpool status', and we'll display the correct information anyway.
3961 */
3962 static void
set_path(zpool_handle_t * zhp,nvlist_t * nv,const char * path)3963 set_path(zpool_handle_t *zhp, nvlist_t *nv, const char *path)
3964 {
3965 zfs_cmd_t zc = { 0 };
3966
3967 (void) strncpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3968 (void) strncpy(zc.zc_value, path, sizeof (zc.zc_value));
3969 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
3970 &zc.zc_guid) == 0);
3971
3972 (void) ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SETPATH, &zc);
3973 }
3974
3975 /*
3976 * This routine is responsible for identifying when disks have been
3977 * reconfigured in a new location. The kernel will have opened the device by
3978 * devid, but the path will still refer to the old location. To catch this, we
3979 * first do a path -> devid translation (which is fast for the common case).
3980 * If the devid matches, we're done. If not, we do a reverse devid -> path
3981 * translation and issue the appropriate ioctl() to update the path of the
3982 * vdev.
3983 */
3984 void
zpool_vdev_refresh_path(libzfs_handle_t * hdl,zpool_handle_t * zhp,nvlist_t * nv)3985 zpool_vdev_refresh_path(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv)
3986 {
3987 char *path = NULL;
3988 char *newpath = NULL;
3989 char *physpath = NULL;
3990 char *devid = NULL;
3991
3992 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) != 0) {
3993 return;
3994 }
3995
3996 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) {
3997 /*
3998 * This vdev has a devid. We can use it to check the current
3999 * path.
4000 */
4001 char *newdevid = path_to_devid(path);
4002
4003 if (newdevid == NULL || strcmp(devid, newdevid) != 0) {
4004 newpath = devid_to_path(devid);
4005 }
4006
4007 if (newdevid != NULL) {
4008 devid_str_free(newdevid);
4009 }
4010
4011 } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PHYS_PATH,
4012 &physpath) == 0) {
4013 /*
4014 * This vdev does not have a devid, but it does have a physical
4015 * path. Attempt to translate this to a /dev path.
4016 */
4017 newpath = path_from_physpath(hdl, path, physpath);
4018 }
4019
4020 if (newpath == NULL) {
4021 /*
4022 * No path update is required.
4023 */
4024 return;
4025 }
4026
4027 set_path(zhp, nv, newpath);
4028 fnvlist_add_string(nv, ZPOOL_CONFIG_PATH, newpath);
4029
4030 free(newpath);
4031 }
4032
4033 /*
4034 * Given a vdev, return the name to display in iostat. If the vdev has a path,
4035 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
4036 * We will confirm that the path and name of the vdev are current, and update
4037 * them if not. We also check if this is a whole disk, in which case we strip
4038 * off the trailing 's0' slice name.
4039 *
4040 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
4041 * of these checks.
4042 */
4043 char *
zpool_vdev_name(libzfs_handle_t * hdl,zpool_handle_t * zhp,nvlist_t * nv,int name_flags)4044 zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv,
4045 int name_flags)
4046 {
4047 char *path, *type, *env;
4048 uint64_t value;
4049
4050 /*
4051 * vdev_name will be "root"/"root-0" for the root vdev, but it is the
4052 * zpool name that will be displayed to the user.
4053 */
4054 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0);
4055 if (zhp != NULL && strcmp(type, "root") == 0)
4056 return (zfs_strdup(hdl, zpool_get_name(zhp)));
4057
4058 env = getenv("ZPOOL_VDEV_NAME_PATH");
4059 if (env && (strtoul(env, NULL, 0) > 0 ||
4060 !strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2)))
4061 name_flags |= VDEV_NAME_PATH;
4062
4063 env = getenv("ZPOOL_VDEV_NAME_GUID");
4064 if (env && (strtoul(env, NULL, 0) > 0 ||
4065 !strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2)))
4066 name_flags |= VDEV_NAME_GUID;
4067
4068 env = getenv("ZPOOL_VDEV_NAME_FOLLOW_LINKS");
4069 if (env && (strtoul(env, NULL, 0) > 0 ||
4070 !strncasecmp(env, "YES", 3) || !strncasecmp(env, "ON", 2)))
4071 name_flags |= VDEV_NAME_FOLLOW_LINKS;
4072
4073 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, &value) == 0 ||
4074 name_flags & VDEV_NAME_GUID) {
4075 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value);
4076 path = zfs_asprintf(hdl, "%llu", (u_longlong_t)value);
4077 } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) {
4078 vdev_stat_t *vs;
4079 uint_t vsc;
4080
4081 /*
4082 * If the device is dead (faulted, offline, etc) then don't
4083 * bother opening it. Otherwise we may be forcing the user to
4084 * open a misbehaving device, which can have undesirable
4085 * effects.
4086 */
4087 if (nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
4088 (uint64_t **)&vs, &vsc) != 0 ||
4089 vs->vs_state < VDEV_STATE_DEGRADED ||
4090 zhp == NULL) {
4091 path = zfs_strdup(hdl, path);
4092 goto after_open;
4093 }
4094
4095 /*
4096 * Refresh the /dev path for this vdev if required, then ensure
4097 * we're using the latest path value:
4098 */
4099 zpool_vdev_refresh_path(hdl, zhp, nv);
4100 path = fnvlist_lookup_string(nv, ZPOOL_CONFIG_PATH);
4101
4102 if (name_flags & VDEV_NAME_FOLLOW_LINKS) {
4103 char *rp = realpath(path, NULL);
4104 if (rp == NULL)
4105 no_memory(hdl);
4106 path = rp;
4107 } else {
4108 path = zfs_strdup(hdl, path);
4109 }
4110
4111 after_open:
4112 if (strncmp(path, ZFS_DISK_ROOTD,
4113 sizeof (ZFS_DISK_ROOTD) - 1) == 0) {
4114 const char *p2 = path + sizeof (ZFS_DISK_ROOTD) - 1;
4115
4116 memmove(path, p2, strlen(p2) + 1);
4117 }
4118
4119 /*
4120 * Remove the partition from the path it this is a whole disk.
4121 */
4122 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, &value)
4123 == 0 && value && !(name_flags & VDEV_NAME_PATH)) {
4124 int pathlen = strlen(path);
4125
4126 /*
4127 * If it starts with c#, and ends with "s0" or "s1",
4128 * chop the slice off, or if it ends with "s0/old" or
4129 * "s1/old", remove the slice from the middle.
4130 */
4131 if (CTD_CHECK(path)) {
4132 if (strcmp(&path[pathlen - 2], "s0") == 0 ||
4133 strcmp(&path[pathlen - 2], "s1") == 0) {
4134 path[pathlen - 2] = '\0';
4135 } else if (pathlen > 6 &&
4136 (strcmp(&path[pathlen - 6],
4137 "s0/old") == 0 ||
4138 strcmp(&path[pathlen - 6],
4139 "s1/old") == 0)) {
4140 (void) strcpy(&path[pathlen - 6],
4141 "/old");
4142 }
4143 }
4144 return (path);
4145 }
4146 } else {
4147 /*
4148 * If it's a raidz device, we need to stick in the parity level.
4149 */
4150 if (strcmp(type, VDEV_TYPE_RAIDZ) == 0) {
4151 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY,
4152 &value) == 0);
4153 path = zfs_asprintf(hdl, "%s%llu", type,
4154 (u_longlong_t)value);
4155 } else {
4156 path = zfs_strdup(hdl, type);
4157 }
4158
4159 /*
4160 * We identify each top-level vdev by using a <type-id>
4161 * naming convention.
4162 */
4163 if (name_flags & VDEV_NAME_TYPE_ID) {
4164 uint64_t id;
4165 char *tmp;
4166
4167 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
4168 &id) == 0);
4169 tmp = zfs_asprintf(hdl, "%s-%llu", path,
4170 (u_longlong_t)id);
4171 free(path);
4172 path = tmp;
4173 }
4174 }
4175
4176 return (path);
4177 }
4178
4179 static int
zbookmark_mem_compare(const void * a,const void * b)4180 zbookmark_mem_compare(const void *a, const void *b)
4181 {
4182 return (memcmp(a, b, sizeof (zbookmark_phys_t)));
4183 }
4184
4185 /*
4186 * Retrieve the persistent error log, uniquify the members, and return to the
4187 * caller.
4188 */
4189 int
zpool_get_errlog(zpool_handle_t * zhp,nvlist_t ** nverrlistp)4190 zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp)
4191 {
4192 zfs_cmd_t zc = { 0 };
4193 uint64_t count;
4194 zbookmark_phys_t *zb = NULL;
4195 int i;
4196
4197 /*
4198 * Retrieve the raw error list from the kernel. If the number of errors
4199 * has increased, allocate more space and continue until we get the
4200 * entire list.
4201 */
4202 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT,
4203 &count) == 0);
4204 if (count == 0)
4205 return (0);
4206 if ((zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl,
4207 count * sizeof (zbookmark_phys_t))) == (uintptr_t)NULL)
4208 return (-1);
4209 zc.zc_nvlist_dst_size = count;
4210 (void) strcpy(zc.zc_name, zhp->zpool_name);
4211 for (;;) {
4212 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_ERROR_LOG,
4213 &zc) != 0) {
4214 free((void *)(uintptr_t)zc.zc_nvlist_dst);
4215 if (errno == ENOMEM) {
4216 void *dst;
4217
4218 count = zc.zc_nvlist_dst_size;
4219 dst = zfs_alloc(zhp->zpool_hdl, count *
4220 sizeof (zbookmark_phys_t));
4221 if (dst == NULL)
4222 return (-1);
4223 zc.zc_nvlist_dst = (uintptr_t)dst;
4224 } else {
4225 return (-1);
4226 }
4227 } else {
4228 break;
4229 }
4230 }
4231
4232 /*
4233 * Sort the resulting bookmarks. This is a little confusing due to the
4234 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last
4235 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks
4236 * _not_ copied as part of the process. So we point the start of our
4237 * array appropriate and decrement the total number of elements.
4238 */
4239 zb = ((zbookmark_phys_t *)(uintptr_t)zc.zc_nvlist_dst) +
4240 zc.zc_nvlist_dst_size;
4241 count -= zc.zc_nvlist_dst_size;
4242
4243 qsort(zb, count, sizeof (zbookmark_phys_t), zbookmark_mem_compare);
4244
4245 verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0);
4246
4247 /*
4248 * Fill in the nverrlistp with nvlist's of dataset and object numbers.
4249 */
4250 for (i = 0; i < count; i++) {
4251 nvlist_t *nv;
4252
4253 /* ignoring zb_blkid and zb_level for now */
4254 if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset &&
4255 zb[i-1].zb_object == zb[i].zb_object)
4256 continue;
4257
4258 if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0)
4259 goto nomem;
4260 if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET,
4261 zb[i].zb_objset) != 0) {
4262 nvlist_free(nv);
4263 goto nomem;
4264 }
4265 if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT,
4266 zb[i].zb_object) != 0) {
4267 nvlist_free(nv);
4268 goto nomem;
4269 }
4270 if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) {
4271 nvlist_free(nv);
4272 goto nomem;
4273 }
4274 nvlist_free(nv);
4275 }
4276
4277 free((void *)(uintptr_t)zc.zc_nvlist_dst);
4278 return (0);
4279
4280 nomem:
4281 free((void *)(uintptr_t)zc.zc_nvlist_dst);
4282 return (no_memory(zhp->zpool_hdl));
4283 }
4284
4285 /*
4286 * Upgrade a ZFS pool to the latest on-disk version.
4287 */
4288 int
zpool_upgrade(zpool_handle_t * zhp,uint64_t new_version)4289 zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version)
4290 {
4291 zfs_cmd_t zc = { 0 };
4292 libzfs_handle_t *hdl = zhp->zpool_hdl;
4293
4294 (void) strcpy(zc.zc_name, zhp->zpool_name);
4295 zc.zc_cookie = new_version;
4296
4297 if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0)
4298 return (zpool_standard_error_fmt(hdl, errno,
4299 dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"),
4300 zhp->zpool_name));
4301 return (0);
4302 }
4303
4304 void
zfs_save_arguments(int argc,char ** argv,char * string,int len)4305 zfs_save_arguments(int argc, char **argv, char *string, int len)
4306 {
4307 (void) strlcpy(string, basename(argv[0]), len);
4308 for (int i = 1; i < argc; i++) {
4309 (void) strlcat(string, " ", len);
4310 (void) strlcat(string, argv[i], len);
4311 }
4312 }
4313
4314 int
zpool_log_history(libzfs_handle_t * hdl,const char * message)4315 zpool_log_history(libzfs_handle_t *hdl, const char *message)
4316 {
4317 zfs_cmd_t zc = { 0 };
4318 nvlist_t *args;
4319 int err;
4320
4321 args = fnvlist_alloc();
4322 fnvlist_add_string(args, "message", message);
4323 err = zcmd_write_src_nvlist(hdl, &zc, args);
4324 if (err == 0)
4325 err = ioctl(hdl->libzfs_fd, ZFS_IOC_LOG_HISTORY, &zc);
4326 nvlist_free(args);
4327 zcmd_free_nvlists(&zc);
4328 return (err);
4329 }
4330
4331 /*
4332 * Perform ioctl to get some command history of a pool.
4333 *
4334 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the
4335 * logical offset of the history buffer to start reading from.
4336 *
4337 * Upon return, 'off' is the next logical offset to read from and
4338 * 'len' is the actual amount of bytes read into 'buf'.
4339 */
4340 static int
get_history(zpool_handle_t * zhp,char * buf,uint64_t * off,uint64_t * len)4341 get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len)
4342 {
4343 zfs_cmd_t zc = { 0 };
4344 libzfs_handle_t *hdl = zhp->zpool_hdl;
4345
4346 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4347
4348 zc.zc_history = (uint64_t)(uintptr_t)buf;
4349 zc.zc_history_len = *len;
4350 zc.zc_history_offset = *off;
4351
4352 if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) {
4353 switch (errno) {
4354 case EPERM:
4355 return (zfs_error_fmt(hdl, EZFS_PERM,
4356 dgettext(TEXT_DOMAIN,
4357 "cannot show history for pool '%s'"),
4358 zhp->zpool_name));
4359 case ENOENT:
4360 return (zfs_error_fmt(hdl, EZFS_NOHISTORY,
4361 dgettext(TEXT_DOMAIN, "cannot get history for pool "
4362 "'%s'"), zhp->zpool_name));
4363 case ENOTSUP:
4364 return (zfs_error_fmt(hdl, EZFS_BADVERSION,
4365 dgettext(TEXT_DOMAIN, "cannot get history for pool "
4366 "'%s', pool must be upgraded"), zhp->zpool_name));
4367 default:
4368 return (zpool_standard_error_fmt(hdl, errno,
4369 dgettext(TEXT_DOMAIN,
4370 "cannot get history for '%s'"), zhp->zpool_name));
4371 }
4372 }
4373
4374 *len = zc.zc_history_len;
4375 *off = zc.zc_history_offset;
4376
4377 return (0);
4378 }
4379
4380 /*
4381 * Retrieve the command history of a pool.
4382 */
4383 int
zpool_get_history(zpool_handle_t * zhp,nvlist_t ** nvhisp,uint64_t * off,boolean_t * eof)4384 zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp, uint64_t *off,
4385 boolean_t *eof)
4386 {
4387 char *buf;
4388 int buflen = 128 * 1024;
4389 nvlist_t **records = NULL;
4390 uint_t numrecords = 0;
4391 int err = 0, i;
4392 uint64_t start = *off;
4393
4394 buf = malloc(buflen);
4395 if (buf == NULL)
4396 return (ENOMEM);
4397 /* process about 1MB a time */
4398 while (*off - start < 1024 * 1024) {
4399 uint64_t bytes_read = buflen;
4400 uint64_t leftover;
4401
4402 if ((err = get_history(zhp, buf, off, &bytes_read)) != 0)
4403 break;
4404
4405 /* if nothing else was read in, we're at EOF, just return */
4406 if (!bytes_read) {
4407 *eof = B_TRUE;
4408 break;
4409 }
4410
4411 if ((err = zpool_history_unpack(buf, bytes_read,
4412 &leftover, &records, &numrecords)) != 0)
4413 break;
4414 *off -= leftover;
4415 if (leftover == bytes_read) {
4416 /*
4417 * no progress made, because buffer is not big enough
4418 * to hold this record; resize and retry.
4419 */
4420 buflen *= 2;
4421 free(buf);
4422 buf = malloc(buflen);
4423 if (buf == NULL)
4424 return (ENOMEM);
4425 }
4426 }
4427
4428 free(buf);
4429
4430 if (!err) {
4431 verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0);
4432 verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD,
4433 records, numrecords) == 0);
4434 }
4435 for (i = 0; i < numrecords; i++)
4436 nvlist_free(records[i]);
4437 free(records);
4438
4439 return (err);
4440 }
4441
4442 void
zpool_obj_to_path(zpool_handle_t * zhp,uint64_t dsobj,uint64_t obj,char * pathname,size_t len)4443 zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
4444 char *pathname, size_t len)
4445 {
4446 zfs_cmd_t zc = { 0 };
4447 boolean_t mounted = B_FALSE;
4448 char *mntpnt = NULL;
4449 char dsname[ZFS_MAX_DATASET_NAME_LEN];
4450
4451 if (dsobj == 0) {
4452 /* special case for the MOS */
4453 (void) snprintf(pathname, len, "<metadata>:<0x%llx>", obj);
4454 return;
4455 }
4456
4457 /* get the dataset's name */
4458 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4459 zc.zc_obj = dsobj;
4460 if (ioctl(zhp->zpool_hdl->libzfs_fd,
4461 ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) {
4462 /* just write out a path of two object numbers */
4463 (void) snprintf(pathname, len, "<0x%llx>:<0x%llx>",
4464 dsobj, obj);
4465 return;
4466 }
4467 (void) strlcpy(dsname, zc.zc_value, sizeof (dsname));
4468
4469 /* find out if the dataset is mounted */
4470 mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt);
4471
4472 /* get the corrupted object's path */
4473 (void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name));
4474 zc.zc_obj = obj;
4475 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_OBJ_TO_PATH,
4476 &zc) == 0) {
4477 if (mounted) {
4478 (void) snprintf(pathname, len, "%s%s", mntpnt,
4479 zc.zc_value);
4480 } else {
4481 (void) snprintf(pathname, len, "%s:%s",
4482 dsname, zc.zc_value);
4483 }
4484 } else {
4485 (void) snprintf(pathname, len, "%s:<0x%llx>", dsname, obj);
4486 }
4487 free(mntpnt);
4488 }
4489
4490 int
zpool_set_bootenv(zpool_handle_t * zhp,const nvlist_t * envmap)4491 zpool_set_bootenv(zpool_handle_t *zhp, const nvlist_t *envmap)
4492 {
4493 int error = lzc_set_bootenv(zhp->zpool_name, envmap);
4494 if (error != 0) {
4495 (void) zpool_standard_error_fmt(zhp->zpool_hdl, error,
4496 dgettext(TEXT_DOMAIN,
4497 "error setting bootenv in pool '%s'"), zhp->zpool_name);
4498 }
4499
4500 return (error);
4501 }
4502
4503 int
zpool_get_bootenv(zpool_handle_t * zhp,nvlist_t ** nvlp)4504 zpool_get_bootenv(zpool_handle_t *zhp, nvlist_t **nvlp)
4505 {
4506 nvlist_t *nvl;
4507 int error;
4508
4509 nvl = NULL;
4510 error = lzc_get_bootenv(zhp->zpool_name, &nvl);
4511 if (error != 0) {
4512 (void) zpool_standard_error_fmt(zhp->zpool_hdl, error,
4513 dgettext(TEXT_DOMAIN,
4514 "error getting bootenv in pool '%s'"), zhp->zpool_name);
4515 } else {
4516 *nvlp = nvl;
4517 }
4518
4519 return (error);
4520 }
4521
4522 /*
4523 * Read the EFI label from the config, if a label does not exist then
4524 * pass back the error to the caller. If the caller has passed a non-NULL
4525 * diskaddr argument then we set it to the starting address of the EFI
4526 * partition. If the caller has passed a non-NULL boolean argument, then
4527 * we set it to indicate if the disk does have efi system partition.
4528 */
4529 static int
read_efi_label(nvlist_t * config,diskaddr_t * sb,boolean_t * system)4530 read_efi_label(nvlist_t *config, diskaddr_t *sb, boolean_t *system)
4531 {
4532 char *path;
4533 int fd;
4534 char diskname[MAXPATHLEN];
4535 boolean_t boot = B_FALSE;
4536 int err = -1;
4537 int slice;
4538
4539 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PATH, &path) != 0)
4540 return (err);
4541
4542 (void) snprintf(diskname, sizeof (diskname), "%s%s", ZFS_RDISK_ROOT,
4543 strrchr(path, '/'));
4544 if ((fd = open(diskname, O_RDONLY|O_NDELAY)) >= 0) {
4545 struct dk_gpt *vtoc;
4546
4547 if ((err = efi_alloc_and_read(fd, &vtoc)) >= 0) {
4548 for (slice = 0; slice < vtoc->efi_nparts; slice++) {
4549 if (vtoc->efi_parts[slice].p_tag == V_SYSTEM)
4550 boot = B_TRUE;
4551 if (vtoc->efi_parts[slice].p_tag == V_USR)
4552 break;
4553 }
4554 if (sb != NULL && vtoc->efi_parts[slice].p_tag == V_USR)
4555 *sb = vtoc->efi_parts[slice].p_start;
4556 if (system != NULL)
4557 *system = boot;
4558 efi_free(vtoc);
4559 }
4560 (void) close(fd);
4561 }
4562 return (err);
4563 }
4564
4565 /*
4566 * determine where a partition starts on a disk in the current
4567 * configuration
4568 */
4569 static diskaddr_t
find_start_block(nvlist_t * config)4570 find_start_block(nvlist_t *config)
4571 {
4572 nvlist_t **child;
4573 uint_t c, children;
4574 diskaddr_t sb = MAXOFFSET_T;
4575 uint64_t wholedisk;
4576
4577 if (nvlist_lookup_nvlist_array(config,
4578 ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) {
4579 if (nvlist_lookup_uint64(config,
4580 ZPOOL_CONFIG_WHOLE_DISK,
4581 &wholedisk) != 0 || !wholedisk) {
4582 return (MAXOFFSET_T);
4583 }
4584 if (read_efi_label(config, &sb, NULL) < 0)
4585 sb = MAXOFFSET_T;
4586 return (sb);
4587 }
4588
4589 for (c = 0; c < children; c++) {
4590 sb = find_start_block(child[c]);
4591 if (sb != MAXOFFSET_T) {
4592 return (sb);
4593 }
4594 }
4595 return (MAXOFFSET_T);
4596 }
4597
4598 /*
4599 * Label an individual disk. The name provided is the short name,
4600 * stripped of any leading /dev path.
4601 */
4602 int
zpool_label_disk(libzfs_handle_t * hdl,zpool_handle_t * zhp,const char * name,zpool_boot_label_t boot_type,uint64_t boot_size,int * slice)4603 zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, const char *name,
4604 zpool_boot_label_t boot_type, uint64_t boot_size, int *slice)
4605 {
4606 char path[MAXPATHLEN];
4607 struct dk_gpt *vtoc;
4608 int fd;
4609 size_t resv;
4610 uint64_t slice_size;
4611 diskaddr_t start_block;
4612 char errbuf[1024];
4613
4614 /* prepare an error message just in case */
4615 (void) snprintf(errbuf, sizeof (errbuf),
4616 dgettext(TEXT_DOMAIN, "cannot label '%s'"), name);
4617
4618 if (zhp) {
4619 nvlist_t *nvroot;
4620
4621 verify(nvlist_lookup_nvlist(zhp->zpool_config,
4622 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
4623
4624 if (zhp->zpool_start_block == 0)
4625 start_block = find_start_block(nvroot);
4626 else
4627 start_block = zhp->zpool_start_block;
4628 zhp->zpool_start_block = start_block;
4629 } else {
4630 /* new pool */
4631 start_block = NEW_START_BLOCK;
4632 }
4633
4634 (void) snprintf(path, sizeof (path), "%s/%s%s", ZFS_RDISK_ROOT, name,
4635 BACKUP_SLICE);
4636
4637 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) {
4638 /*
4639 * This shouldn't happen. We've long since verified that this
4640 * is a valid device.
4641 */
4642 zfs_error_aux(hdl,
4643 dgettext(TEXT_DOMAIN, "unable to open device"));
4644 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
4645 }
4646
4647 if (efi_alloc_and_init(fd, EFI_NUMPAR, &vtoc) != 0) {
4648 /*
4649 * The only way this can fail is if we run out of memory, or we
4650 * were unable to read the disk's capacity
4651 */
4652 if (errno == ENOMEM)
4653 (void) no_memory(hdl);
4654
4655 (void) close(fd);
4656 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4657 "unable to read disk capacity"), name);
4658
4659 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
4660 }
4661 resv = efi_reserved_sectors(vtoc);
4662
4663 /*
4664 * Why we use V_USR: V_BACKUP confuses users, and is considered
4665 * disposable by some EFI utilities (since EFI doesn't have a backup
4666 * slice). V_UNASSIGNED is supposed to be used only for zero size
4667 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT,
4668 * etc. were all pretty specific. V_USR is as close to reality as we
4669 * can get, in the absence of V_OTHER.
4670 */
4671 /* first fix the partition start block */
4672 if (start_block == MAXOFFSET_T)
4673 start_block = NEW_START_BLOCK;
4674
4675 /*
4676 * EFI System partition is using slice 0.
4677 * ZFS is on slice 1 and slice 8 is reserved.
4678 * We assume the GPT partition table without system
4679 * partition has zfs p_start == NEW_START_BLOCK.
4680 * If start_block != NEW_START_BLOCK, it means we have
4681 * system partition. Correct solution would be to query/cache vtoc
4682 * from existing vdev member.
4683 */
4684 if (boot_type == ZPOOL_CREATE_BOOT_LABEL) {
4685 if (boot_size % vtoc->efi_lbasize != 0) {
4686 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4687 "boot partition size must be a multiple of %d"),
4688 vtoc->efi_lbasize);
4689 (void) close(fd);
4690 efi_free(vtoc);
4691 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
4692 }
4693 /*
4694 * System partition size checks.
4695 * Note the 1MB is quite arbitrary value, since we
4696 * are creating dedicated pool, it should be enough
4697 * to hold fat + efi bootloader. May need to be
4698 * adjusted if the bootloader size will grow.
4699 */
4700 if (boot_size < 1024 * 1024) {
4701 char buf[64];
4702 zfs_nicenum(boot_size, buf, sizeof (buf));
4703 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4704 "Specified size %s for EFI System partition is too "
4705 "small, the minimum size is 1MB."), buf);
4706 (void) close(fd);
4707 efi_free(vtoc);
4708 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
4709 }
4710 /* 33MB is tested with mkfs -F pcfs */
4711 if (hdl->libzfs_printerr &&
4712 ((vtoc->efi_lbasize == 512 &&
4713 boot_size < 33 * 1024 * 1024) ||
4714 (vtoc->efi_lbasize == 4096 &&
4715 boot_size < 256 * 1024 * 1024))) {
4716 char buf[64];
4717 zfs_nicenum(boot_size, buf, sizeof (buf));
4718 (void) fprintf(stderr, dgettext(TEXT_DOMAIN,
4719 "Warning: EFI System partition size %s is "
4720 "not allowing to create FAT32 file\nsystem, which "
4721 "may result in unbootable system.\n"), buf);
4722 }
4723 /* Adjust zfs partition start by size of system partition. */
4724 start_block += boot_size / vtoc->efi_lbasize;
4725 }
4726
4727 if (start_block == NEW_START_BLOCK) {
4728 /*
4729 * Use default layout.
4730 * ZFS is on slice 0 and slice 8 is reserved.
4731 */
4732 slice_size = vtoc->efi_last_u_lba + 1;
4733 slice_size -= resv;
4734 slice_size -= start_block;
4735 if (slice != NULL)
4736 *slice = 0;
4737
4738 vtoc->efi_parts[0].p_start = start_block;
4739 vtoc->efi_parts[0].p_size = slice_size;
4740
4741 vtoc->efi_parts[0].p_tag = V_USR;
4742 (void) strcpy(vtoc->efi_parts[0].p_name, "zfs");
4743
4744 vtoc->efi_parts[8].p_start = slice_size + start_block;
4745 vtoc->efi_parts[8].p_size = resv;
4746 vtoc->efi_parts[8].p_tag = V_RESERVED;
4747 } else {
4748 slice_size = start_block - NEW_START_BLOCK;
4749 vtoc->efi_parts[0].p_start = NEW_START_BLOCK;
4750 vtoc->efi_parts[0].p_size = slice_size;
4751 vtoc->efi_parts[0].p_tag = V_SYSTEM;
4752 (void) strcpy(vtoc->efi_parts[0].p_name, "loader");
4753 if (slice != NULL)
4754 *slice = 1;
4755 /* prepare slice 1 */
4756 slice_size = vtoc->efi_last_u_lba + 1 - slice_size;
4757 slice_size -= resv;
4758 slice_size -= NEW_START_BLOCK;
4759 vtoc->efi_parts[1].p_start = start_block;
4760 vtoc->efi_parts[1].p_size = slice_size;
4761 vtoc->efi_parts[1].p_tag = V_USR;
4762 (void) strcpy(vtoc->efi_parts[1].p_name, "zfs");
4763
4764 vtoc->efi_parts[8].p_start = slice_size + start_block;
4765 vtoc->efi_parts[8].p_size = resv;
4766 vtoc->efi_parts[8].p_tag = V_RESERVED;
4767 }
4768
4769 if (efi_write(fd, vtoc) != 0) {
4770 /*
4771 * Some block drivers (like pcata) may not support EFI
4772 * GPT labels. Print out a helpful error message dir-
4773 * ecting the user to manually label the disk and give
4774 * a specific slice.
4775 */
4776 (void) close(fd);
4777 efi_free(vtoc);
4778
4779 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4780 "try using fdisk(8) and then provide a specific slice"));
4781 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
4782 }
4783
4784 (void) close(fd);
4785 efi_free(vtoc);
4786 return (0);
4787 }
4788
4789 static boolean_t
supported_dump_vdev_type(libzfs_handle_t * hdl,nvlist_t * config,char * errbuf)4790 supported_dump_vdev_type(libzfs_handle_t *hdl, nvlist_t *config, char *errbuf)
4791 {
4792 char *type;
4793 nvlist_t **child;
4794 uint_t children, c;
4795
4796 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_TYPE, &type) == 0);
4797 if (strcmp(type, VDEV_TYPE_FILE) == 0 ||
4798 strcmp(type, VDEV_TYPE_HOLE) == 0 ||
4799 strcmp(type, VDEV_TYPE_MISSING) == 0) {
4800 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4801 "vdev type '%s' is not supported"), type);
4802 (void) zfs_error(hdl, EZFS_VDEVNOTSUP, errbuf);
4803 return (B_FALSE);
4804 }
4805 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN,
4806 &child, &children) == 0) {
4807 for (c = 0; c < children; c++) {
4808 if (!supported_dump_vdev_type(hdl, child[c], errbuf))
4809 return (B_FALSE);
4810 }
4811 }
4812 return (B_TRUE);
4813 }
4814
4815 /*
4816 * Check if this zvol is allowable for use as a dump device; zero if
4817 * it is, > 0 if it isn't, < 0 if it isn't a zvol.
4818 *
4819 * Allowable storage configurations include mirrors, all raidz variants, and
4820 * pools with log, cache, and spare devices. Pools which are backed by files or
4821 * have missing/hole vdevs are not suitable.
4822 */
4823 int
zvol_check_dump_config(char * arg)4824 zvol_check_dump_config(char *arg)
4825 {
4826 zpool_handle_t *zhp = NULL;
4827 nvlist_t *config, *nvroot;
4828 char *p, *volname;
4829 nvlist_t **top;
4830 uint_t toplevels;
4831 libzfs_handle_t *hdl;
4832 char errbuf[1024];
4833 char poolname[ZFS_MAX_DATASET_NAME_LEN];
4834 int pathlen = strlen(ZVOL_FULL_DEV_DIR);
4835 int ret = 1;
4836
4837 if (strncmp(arg, ZVOL_FULL_DEV_DIR, pathlen)) {
4838 return (-1);
4839 }
4840
4841 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
4842 "dump is not supported on device '%s'"), arg);
4843
4844 if ((hdl = libzfs_init()) == NULL)
4845 return (1);
4846 libzfs_print_on_error(hdl, B_TRUE);
4847
4848 volname = arg + pathlen;
4849
4850 /* check the configuration of the pool */
4851 if ((p = strchr(volname, '/')) == NULL) {
4852 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4853 "malformed dataset name"));
4854 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
4855 return (1);
4856 } else if (p - volname >= ZFS_MAX_DATASET_NAME_LEN) {
4857 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4858 "dataset name is too long"));
4859 (void) zfs_error(hdl, EZFS_NAMETOOLONG, errbuf);
4860 return (1);
4861 } else {
4862 (void) strncpy(poolname, volname, p - volname);
4863 poolname[p - volname] = '\0';
4864 }
4865
4866 if ((zhp = zpool_open(hdl, poolname)) == NULL) {
4867 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4868 "could not open pool '%s'"), poolname);
4869 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
4870 goto out;
4871 }
4872 config = zpool_get_config(zhp, NULL);
4873 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
4874 &nvroot) != 0) {
4875 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4876 "could not obtain vdev configuration for '%s'"), poolname);
4877 (void) zfs_error(hdl, EZFS_INVALCONFIG, errbuf);
4878 goto out;
4879 }
4880
4881 verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
4882 &top, &toplevels) == 0);
4883
4884 if (!supported_dump_vdev_type(hdl, top[0], errbuf)) {
4885 goto out;
4886 }
4887 ret = 0;
4888
4889 out:
4890 if (zhp)
4891 zpool_close(zhp);
4892 libzfs_fini(hdl);
4893 return (ret);
4894 }
4895