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, 2015 by Delphix. All rights reserved.
25 * Copyright (c) 2013, Joyent, Inc. All rights reserved.
26 * Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>
27 * Copyright 2017 Nexenta Systems, Inc.
28 * Copyright (c) 2017 Datto Inc.
29 */
30
31 #include <ctype.h>
32 #include <errno.h>
33 #include <devid.h>
34 #include <fcntl.h>
35 #include <libintl.h>
36 #include <stdio.h>
37 #include <stdlib.h>
38 #include <strings.h>
39 #include <unistd.h>
40 #include <libgen.h>
41 #include <sys/efi_partition.h>
42 #include <sys/vtoc.h>
43 #include <sys/zfs_ioctl.h>
44 #include <dlfcn.h>
45
46 #include "zfs_namecheck.h"
47 #include "zfs_prop.h"
48 #include "libzfs_impl.h"
49 #include "zfs_comutil.h"
50 #include "zfeature_common.h"
51
52 static int read_efi_label(nvlist_t *, diskaddr_t *, boolean_t *);
53
54 #define BACKUP_SLICE "s2"
55
56 typedef struct prop_flags {
57 int create:1; /* Validate property on creation */
58 int import:1; /* Validate property on import */
59 } prop_flags_t;
60
61 /*
62 * ====================================================================
63 * zpool property functions
64 * ====================================================================
65 */
66
67 static int
68 zpool_get_all_props(zpool_handle_t *zhp)
69 {
70 zfs_cmd_t zc = { 0 };
71 libzfs_handle_t *hdl = zhp->zpool_hdl;
72
73 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
74
75 if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0)
76 return (-1);
77
78 while (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_PROPS, &zc) != 0) {
79 if (errno == ENOMEM) {
80 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
81 zcmd_free_nvlists(&zc);
82 return (-1);
83 }
84 } else {
85 zcmd_free_nvlists(&zc);
86 return (-1);
87 }
88 }
89
90 if (zcmd_read_dst_nvlist(hdl, &zc, &zhp->zpool_props) != 0) {
91 zcmd_free_nvlists(&zc);
92 return (-1);
93 }
94
95 zcmd_free_nvlists(&zc);
96
97 return (0);
98 }
99
100 static int
101 zpool_props_refresh(zpool_handle_t *zhp)
102 {
103 nvlist_t *old_props;
104
105 old_props = zhp->zpool_props;
106
107 if (zpool_get_all_props(zhp) != 0)
108 return (-1);
109
110 nvlist_free(old_props);
111 return (0);
112 }
113
114 static char *
115 zpool_get_prop_string(zpool_handle_t *zhp, zpool_prop_t prop,
116 zprop_source_t *src)
117 {
118 nvlist_t *nv, *nvl;
119 uint64_t ival;
120 char *value;
121 zprop_source_t source;
122
123 nvl = zhp->zpool_props;
124 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
125 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &ival) == 0);
126 source = ival;
127 verify(nvlist_lookup_string(nv, ZPROP_VALUE, &value) == 0);
128 } else {
129 source = ZPROP_SRC_DEFAULT;
130 if ((value = (char *)zpool_prop_default_string(prop)) == NULL)
131 value = "-";
132 }
133
134 if (src)
135 *src = source;
136
137 return (value);
138 }
139
140 uint64_t
141 zpool_get_prop_int(zpool_handle_t *zhp, zpool_prop_t prop, zprop_source_t *src)
142 {
143 nvlist_t *nv, *nvl;
144 uint64_t value;
145 zprop_source_t source;
146
147 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp)) {
148 /*
149 * zpool_get_all_props() has most likely failed because
150 * the pool is faulted, but if all we need is the top level
151 * vdev's guid then get it from the zhp config nvlist.
152 */
153 if ((prop == ZPOOL_PROP_GUID) &&
154 (nvlist_lookup_nvlist(zhp->zpool_config,
155 ZPOOL_CONFIG_VDEV_TREE, &nv) == 0) &&
156 (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value)
157 == 0)) {
158 return (value);
159 }
160 return (zpool_prop_default_numeric(prop));
161 }
162
163 nvl = zhp->zpool_props;
164 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
165 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &value) == 0);
166 source = value;
167 verify(nvlist_lookup_uint64(nv, ZPROP_VALUE, &value) == 0);
168 } else {
169 source = ZPROP_SRC_DEFAULT;
170 value = zpool_prop_default_numeric(prop);
171 }
172
173 if (src)
174 *src = source;
175
176 return (value);
177 }
178
179 /*
180 * Map VDEV STATE to printed strings.
181 */
182 const char *
183 zpool_state_to_name(vdev_state_t state, vdev_aux_t aux)
184 {
185 switch (state) {
186 case VDEV_STATE_CLOSED:
187 case VDEV_STATE_OFFLINE:
188 return (gettext("OFFLINE"));
189 case VDEV_STATE_REMOVED:
190 return (gettext("REMOVED"));
191 case VDEV_STATE_CANT_OPEN:
192 if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG)
193 return (gettext("FAULTED"));
194 else if (aux == VDEV_AUX_SPLIT_POOL)
195 return (gettext("SPLIT"));
196 else
197 return (gettext("UNAVAIL"));
198 case VDEV_STATE_FAULTED:
199 return (gettext("FAULTED"));
200 case VDEV_STATE_DEGRADED:
201 return (gettext("DEGRADED"));
202 case VDEV_STATE_HEALTHY:
203 return (gettext("ONLINE"));
204
205 default:
206 break;
207 }
208
209 return (gettext("UNKNOWN"));
210 }
211
212 /*
213 * Map POOL STATE to printed strings.
214 */
215 const char *
216 zpool_pool_state_to_name(pool_state_t state)
217 {
218 switch (state) {
219 case POOL_STATE_ACTIVE:
220 return (gettext("ACTIVE"));
221 case POOL_STATE_EXPORTED:
222 return (gettext("EXPORTED"));
223 case POOL_STATE_DESTROYED:
224 return (gettext("DESTROYED"));
225 case POOL_STATE_SPARE:
226 return (gettext("SPARE"));
227 case POOL_STATE_L2CACHE:
228 return (gettext("L2CACHE"));
229 case POOL_STATE_UNINITIALIZED:
230 return (gettext("UNINITIALIZED"));
231 case POOL_STATE_UNAVAIL:
232 return (gettext("UNAVAIL"));
233 case POOL_STATE_POTENTIALLY_ACTIVE:
234 return (gettext("POTENTIALLY_ACTIVE"));
235 }
236
237 return (gettext("UNKNOWN"));
238 }
239
240 /*
241 * Get a zpool property value for 'prop' and return the value in
242 * a pre-allocated buffer.
243 */
244 int
245 zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf, size_t len,
246 zprop_source_t *srctype, boolean_t literal)
247 {
248 uint64_t intval;
249 const char *strval;
250 zprop_source_t src = ZPROP_SRC_NONE;
251 nvlist_t *nvroot;
252 vdev_stat_t *vs;
253 uint_t vsc;
254
255 if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
256 switch (prop) {
257 case ZPOOL_PROP_NAME:
258 (void) strlcpy(buf, zpool_get_name(zhp), len);
259 break;
260
261 case ZPOOL_PROP_HEALTH:
262 (void) strlcpy(buf, "FAULTED", len);
263 break;
264
265 case ZPOOL_PROP_GUID:
266 intval = zpool_get_prop_int(zhp, prop, &src);
267 (void) snprintf(buf, len, "%llu", intval);
268 break;
269
270 case ZPOOL_PROP_ALTROOT:
271 case ZPOOL_PROP_CACHEFILE:
272 case ZPOOL_PROP_COMMENT:
273 if (zhp->zpool_props != NULL ||
274 zpool_get_all_props(zhp) == 0) {
275 (void) strlcpy(buf,
276 zpool_get_prop_string(zhp, prop, &src),
277 len);
278 break;
279 }
280 /* FALLTHROUGH */
281 default:
282 (void) strlcpy(buf, "-", len);
283 break;
284 }
285
286 if (srctype != NULL)
287 *srctype = src;
288 return (0);
289 }
290
291 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) &&
292 prop != ZPOOL_PROP_NAME)
293 return (-1);
294
295 switch (zpool_prop_get_type(prop)) {
296 case PROP_TYPE_STRING:
297 (void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src),
298 len);
299 break;
300
301 case PROP_TYPE_NUMBER:
302 intval = zpool_get_prop_int(zhp, prop, &src);
303
304 switch (prop) {
305 case ZPOOL_PROP_SIZE:
306 case ZPOOL_PROP_ALLOCATED:
307 case ZPOOL_PROP_FREE:
308 case ZPOOL_PROP_FREEING:
309 case ZPOOL_PROP_LEAKED:
310 if (literal) {
311 (void) snprintf(buf, len, "%llu",
312 (u_longlong_t)intval);
313 } else {
314 (void) zfs_nicenum(intval, buf, len);
315 }
316 break;
317 case ZPOOL_PROP_BOOTSIZE:
318 case ZPOOL_PROP_EXPANDSZ:
319 if (intval == 0) {
320 (void) strlcpy(buf, "-", len);
321 } else if (literal) {
322 (void) snprintf(buf, len, "%llu",
323 (u_longlong_t)intval);
324 } else {
325 (void) zfs_nicenum(intval, buf, len);
326 }
327 break;
328 case ZPOOL_PROP_CAPACITY:
329 if (literal) {
330 (void) snprintf(buf, len, "%llu",
331 (u_longlong_t)intval);
332 } else {
333 (void) snprintf(buf, len, "%llu%%",
334 (u_longlong_t)intval);
335 }
336 break;
337 case ZPOOL_PROP_FRAGMENTATION:
338 if (intval == UINT64_MAX) {
339 (void) strlcpy(buf, "-", len);
340 } else {
341 (void) snprintf(buf, len, "%llu%%",
342 (u_longlong_t)intval);
343 }
344 break;
345 case ZPOOL_PROP_DEDUPRATIO:
346 (void) snprintf(buf, len, "%llu.%02llux",
347 (u_longlong_t)(intval / 100),
348 (u_longlong_t)(intval % 100));
349 break;
350 case ZPOOL_PROP_HEALTH:
351 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
352 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
353 verify(nvlist_lookup_uint64_array(nvroot,
354 ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc)
355 == 0);
356
357 (void) strlcpy(buf, zpool_state_to_name(intval,
358 vs->vs_aux), len);
359 break;
360 case ZPOOL_PROP_VERSION:
361 if (intval >= SPA_VERSION_FEATURES) {
362 (void) snprintf(buf, len, "-");
363 break;
364 }
365 /* FALLTHROUGH */
366 default:
367 (void) snprintf(buf, len, "%llu", intval);
368 }
369 break;
370
371 case PROP_TYPE_INDEX:
372 intval = zpool_get_prop_int(zhp, prop, &src);
373 if (zpool_prop_index_to_string(prop, intval, &strval)
374 != 0)
375 return (-1);
376 (void) strlcpy(buf, strval, len);
377 break;
378
379 default:
380 abort();
381 }
382
383 if (srctype)
384 *srctype = src;
385
386 return (0);
387 }
388
389 /*
390 * Check if the bootfs name has the same pool name as it is set to.
391 * Assuming bootfs is a valid dataset name.
392 */
393 static boolean_t
394 bootfs_name_valid(const char *pool, char *bootfs)
395 {
396 int len = strlen(pool);
397
398 if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT))
399 return (B_FALSE);
400
401 if (strncmp(pool, bootfs, len) == 0 &&
402 (bootfs[len] == '/' || bootfs[len] == '\0'))
403 return (B_TRUE);
404
405 return (B_FALSE);
406 }
407
408 boolean_t
409 zpool_is_bootable(zpool_handle_t *zhp)
410 {
411 char bootfs[ZFS_MAX_DATASET_NAME_LEN];
412
413 return (zpool_get_prop(zhp, ZPOOL_PROP_BOOTFS, bootfs,
414 sizeof (bootfs), NULL, B_FALSE) == 0 && strncmp(bootfs, "-",
415 sizeof (bootfs)) != 0);
416 }
417
418
419 /*
420 * Given an nvlist of zpool properties to be set, validate that they are
421 * correct, and parse any numeric properties (index, boolean, etc) if they are
422 * specified as strings.
423 */
424 static nvlist_t *
425 zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname,
426 nvlist_t *props, uint64_t version, prop_flags_t flags, const char *errbuf)
427 {
428 nvpair_t *elem;
429 nvlist_t *retprops;
430 zpool_prop_t prop;
431 char *strval;
432 uint64_t intval;
433 char *slash, *check;
434 struct stat64 statbuf;
435 zpool_handle_t *zhp;
436
437 if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) {
438 (void) no_memory(hdl);
439 return (NULL);
440 }
441
442 elem = NULL;
443 while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
444 const char *propname = nvpair_name(elem);
445
446 prop = zpool_name_to_prop(propname);
447 if (prop == ZPROP_INVAL && zpool_prop_feature(propname)) {
448 int err;
449 char *fname = strchr(propname, '@') + 1;
450
451 err = zfeature_lookup_name(fname, NULL);
452 if (err != 0) {
453 ASSERT3U(err, ==, ENOENT);
454 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
455 "invalid feature '%s'"), fname);
456 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
457 goto error;
458 }
459
460 if (nvpair_type(elem) != DATA_TYPE_STRING) {
461 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
462 "'%s' must be a string"), propname);
463 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
464 goto error;
465 }
466
467 (void) nvpair_value_string(elem, &strval);
468 if (strcmp(strval, ZFS_FEATURE_ENABLED) != 0) {
469 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
470 "property '%s' can only be set to "
471 "'enabled'"), propname);
472 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
473 goto error;
474 }
475
476 if (nvlist_add_uint64(retprops, propname, 0) != 0) {
477 (void) no_memory(hdl);
478 goto error;
479 }
480 continue;
481 }
482
483 /*
484 * Make sure this property is valid and applies to this type.
485 */
486 if (prop == ZPROP_INVAL) {
487 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
488 "invalid property '%s'"), propname);
489 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
490 goto error;
491 }
492
493 if (zpool_prop_readonly(prop)) {
494 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
495 "is readonly"), propname);
496 (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
497 goto error;
498 }
499
500 if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops,
501 &strval, &intval, errbuf) != 0)
502 goto error;
503
504 /*
505 * Perform additional checking for specific properties.
506 */
507 switch (prop) {
508 case ZPOOL_PROP_VERSION:
509 if (intval < version ||
510 !SPA_VERSION_IS_SUPPORTED(intval)) {
511 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
512 "property '%s' number %d is invalid."),
513 propname, intval);
514 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
515 goto error;
516 }
517 break;
518
519 case ZPOOL_PROP_BOOTSIZE:
520 if (!flags.create) {
521 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
522 "property '%s' can only be set during pool "
523 "creation"), propname);
524 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
525 goto error;
526 }
527 break;
528
529 case ZPOOL_PROP_BOOTFS:
530 if (flags.create || flags.import) {
531 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
532 "property '%s' cannot be set at creation "
533 "or import time"), propname);
534 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
535 goto error;
536 }
537
538 if (version < SPA_VERSION_BOOTFS) {
539 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
540 "pool must be upgraded to support "
541 "'%s' property"), propname);
542 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
543 goto error;
544 }
545
546 /*
547 * bootfs property value has to be a dataset name and
548 * the dataset has to be in the same pool as it sets to.
549 */
550 if (strval[0] != '\0' && !bootfs_name_valid(poolname,
551 strval)) {
552 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
553 "is an invalid name"), strval);
554 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
555 goto error;
556 }
557
558 if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) {
559 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
560 "could not open pool '%s'"), poolname);
561 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
562 goto error;
563 }
564 zpool_close(zhp);
565 break;
566
567 case ZPOOL_PROP_ALTROOT:
568 if (!flags.create && !flags.import) {
569 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
570 "property '%s' can only be set during pool "
571 "creation or import"), propname);
572 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
573 goto error;
574 }
575
576 if (strval[0] != '/') {
577 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
578 "bad alternate root '%s'"), strval);
579 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
580 goto error;
581 }
582 break;
583
584 case ZPOOL_PROP_CACHEFILE:
585 if (strval[0] == '\0')
586 break;
587
588 if (strcmp(strval, "none") == 0)
589 break;
590
591 if (strval[0] != '/') {
592 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
593 "property '%s' must be empty, an "
594 "absolute path, or 'none'"), propname);
595 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
596 goto error;
597 }
598
599 slash = strrchr(strval, '/');
600
601 if (slash[1] == '\0' || strcmp(slash, "/.") == 0 ||
602 strcmp(slash, "/..") == 0) {
603 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
604 "'%s' is not a valid file"), strval);
605 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
606 goto error;
607 }
608
609 *slash = '\0';
610
611 if (strval[0] != '\0' &&
612 (stat64(strval, &statbuf) != 0 ||
613 !S_ISDIR(statbuf.st_mode))) {
614 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
615 "'%s' is not a valid directory"),
616 strval);
617 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
618 goto error;
619 }
620
621 *slash = '/';
622 break;
623
624 case ZPOOL_PROP_COMMENT:
625 for (check = strval; *check != '\0'; check++) {
626 if (!isprint(*check)) {
627 zfs_error_aux(hdl,
628 dgettext(TEXT_DOMAIN,
629 "comment may only have printable "
630 "characters"));
631 (void) zfs_error(hdl, EZFS_BADPROP,
632 errbuf);
633 goto error;
634 }
635 }
636 if (strlen(strval) > ZPROP_MAX_COMMENT) {
637 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
638 "comment must not exceed %d characters"),
639 ZPROP_MAX_COMMENT);
640 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
641 goto error;
642 }
643 break;
644 case ZPOOL_PROP_READONLY:
645 if (!flags.import) {
646 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
647 "property '%s' can only be set at "
648 "import time"), propname);
649 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
650 goto error;
651 }
652 break;
653
654 default:
655 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
656 "property '%s'(%d) not defined"), propname, prop);
657 break;
658 }
659 }
660
661 return (retprops);
662 error:
663 nvlist_free(retprops);
664 return (NULL);
665 }
666
667 /*
668 * Set zpool property : propname=propval.
669 */
670 int
671 zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval)
672 {
673 zfs_cmd_t zc = { 0 };
674 int ret = -1;
675 char errbuf[1024];
676 nvlist_t *nvl = NULL;
677 nvlist_t *realprops;
678 uint64_t version;
679 prop_flags_t flags = { 0 };
680
681 (void) snprintf(errbuf, sizeof (errbuf),
682 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
683 zhp->zpool_name);
684
685 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
686 return (no_memory(zhp->zpool_hdl));
687
688 if (nvlist_add_string(nvl, propname, propval) != 0) {
689 nvlist_free(nvl);
690 return (no_memory(zhp->zpool_hdl));
691 }
692
693 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
694 if ((realprops = zpool_valid_proplist(zhp->zpool_hdl,
695 zhp->zpool_name, nvl, version, flags, errbuf)) == NULL) {
696 nvlist_free(nvl);
697 return (-1);
698 }
699
700 nvlist_free(nvl);
701 nvl = realprops;
702
703 /*
704 * Execute the corresponding ioctl() to set this property.
705 */
706 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
707
708 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) {
709 nvlist_free(nvl);
710 return (-1);
711 }
712
713 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc);
714
715 zcmd_free_nvlists(&zc);
716 nvlist_free(nvl);
717
718 if (ret)
719 (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
720 else
721 (void) zpool_props_refresh(zhp);
722
723 return (ret);
724 }
725
726 /*
727 * Set zpool properties nvlist
728 */
729 int
730 zpool_set_proplist(zpool_handle_t *zhp, nvlist_t *nvl)
731 {
732 zfs_cmd_t zc = { 0 };
733 int ret = -1;
734 char errbuf[1024];
735 nvlist_t *realprops;
736 uint64_t version;
737 prop_flags_t flags = { 0 };
738
739 assert(nvl != NULL);
740
741 (void) snprintf(errbuf, sizeof (errbuf),
742 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
743 zhp->zpool_name);
744
745 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
746 if ((realprops = zpool_valid_proplist(zhp->zpool_hdl,
747 zhp->zpool_name, nvl, version, flags, errbuf)) == NULL) {
748 nvlist_free(nvl);
749 return (-1);
750 }
751
752 nvlist_free(nvl);
753 nvl = realprops;
754
755 /*
756 * Execute the corresponding ioctl() to set this property.
757 */
758 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
759
760 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) {
761 nvlist_free(nvl);
762 return (-1);
763 }
764
765 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc);
766
767 zcmd_free_nvlists(&zc);
768 nvlist_free(nvl);
769
770 if (ret)
771 (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
772 else
773 (void) zpool_props_refresh(zhp);
774
775 return (ret);
776 }
777
778 int
779 zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp)
780 {
781 libzfs_handle_t *hdl = zhp->zpool_hdl;
782 zprop_list_t *entry;
783 char buf[ZFS_MAXPROPLEN];
784 nvlist_t *features = NULL;
785 zprop_list_t **last;
786 boolean_t firstexpand = (NULL == *plp);
787
788 if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0)
789 return (-1);
790
791 last = plp;
792 while (*last != NULL)
793 last = &(*last)->pl_next;
794
795 if ((*plp)->pl_all)
796 features = zpool_get_features(zhp);
797
798 if ((*plp)->pl_all && firstexpand) {
799 for (int i = 0; i < SPA_FEATURES; i++) {
800 zprop_list_t *entry = zfs_alloc(hdl,
801 sizeof (zprop_list_t));
802 entry->pl_prop = ZPROP_INVAL;
803 entry->pl_user_prop = zfs_asprintf(hdl, "feature@%s",
804 spa_feature_table[i].fi_uname);
805 entry->pl_width = strlen(entry->pl_user_prop);
806 entry->pl_all = B_TRUE;
807
808 *last = entry;
809 last = &entry->pl_next;
810 }
811 }
812
813 /* add any unsupported features */
814 for (nvpair_t *nvp = nvlist_next_nvpair(features, NULL);
815 nvp != NULL; nvp = nvlist_next_nvpair(features, nvp)) {
816 char *propname;
817 boolean_t found;
818 zprop_list_t *entry;
819
820 if (zfeature_is_supported(nvpair_name(nvp)))
821 continue;
822
823 propname = zfs_asprintf(hdl, "unsupported@%s",
824 nvpair_name(nvp));
825
826 /*
827 * Before adding the property to the list make sure that no
828 * other pool already added the same property.
829 */
830 found = B_FALSE;
831 entry = *plp;
832 while (entry != NULL) {
833 if (entry->pl_user_prop != NULL &&
834 strcmp(propname, entry->pl_user_prop) == 0) {
835 found = B_TRUE;
836 break;
837 }
838 entry = entry->pl_next;
839 }
840 if (found) {
841 free(propname);
842 continue;
843 }
844
845 entry = zfs_alloc(hdl, sizeof (zprop_list_t));
846 entry->pl_prop = ZPROP_INVAL;
847 entry->pl_user_prop = propname;
848 entry->pl_width = strlen(entry->pl_user_prop);
849 entry->pl_all = B_TRUE;
850
851 *last = entry;
852 last = &entry->pl_next;
853 }
854
855 for (entry = *plp; entry != NULL; entry = entry->pl_next) {
856
857 if (entry->pl_fixed)
858 continue;
859
860 if (entry->pl_prop != ZPROP_INVAL &&
861 zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf),
862 NULL, B_FALSE) == 0) {
863 if (strlen(buf) > entry->pl_width)
864 entry->pl_width = strlen(buf);
865 }
866 }
867
868 return (0);
869 }
870
871 /*
872 * Get the state for the given feature on the given ZFS pool.
873 */
874 int
875 zpool_prop_get_feature(zpool_handle_t *zhp, const char *propname, char *buf,
876 size_t len)
877 {
878 uint64_t refcount;
879 boolean_t found = B_FALSE;
880 nvlist_t *features = zpool_get_features(zhp);
881 boolean_t supported;
882 const char *feature = strchr(propname, '@') + 1;
883
884 supported = zpool_prop_feature(propname);
885 ASSERT(supported || zpool_prop_unsupported(propname));
886
887 /*
888 * Convert from feature name to feature guid. This conversion is
889 * unecessary for unsupported@... properties because they already
890 * use guids.
891 */
892 if (supported) {
893 int ret;
894 spa_feature_t fid;
895
896 ret = zfeature_lookup_name(feature, &fid);
897 if (ret != 0) {
898 (void) strlcpy(buf, "-", len);
899 return (ENOTSUP);
900 }
901 feature = spa_feature_table[fid].fi_guid;
902 }
903
904 if (nvlist_lookup_uint64(features, feature, &refcount) == 0)
905 found = B_TRUE;
906
907 if (supported) {
908 if (!found) {
909 (void) strlcpy(buf, ZFS_FEATURE_DISABLED, len);
910 } else {
911 if (refcount == 0)
912 (void) strlcpy(buf, ZFS_FEATURE_ENABLED, len);
913 else
914 (void) strlcpy(buf, ZFS_FEATURE_ACTIVE, len);
915 }
916 } else {
917 if (found) {
918 if (refcount == 0) {
919 (void) strcpy(buf, ZFS_UNSUPPORTED_INACTIVE);
920 } else {
921 (void) strcpy(buf, ZFS_UNSUPPORTED_READONLY);
922 }
923 } else {
924 (void) strlcpy(buf, "-", len);
925 return (ENOTSUP);
926 }
927 }
928
929 return (0);
930 }
931
932 /*
933 * Don't start the slice at the default block of 34; many storage
934 * devices will use a stripe width of 128k, so start there instead.
935 */
936 #define NEW_START_BLOCK 256
937
938 /*
939 * Validate the given pool name, optionally putting an extended error message in
940 * 'buf'.
941 */
942 boolean_t
943 zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool)
944 {
945 namecheck_err_t why;
946 char what;
947 int ret;
948
949 ret = pool_namecheck(pool, &why, &what);
950
951 /*
952 * The rules for reserved pool names were extended at a later point.
953 * But we need to support users with existing pools that may now be
954 * invalid. So we only check for this expanded set of names during a
955 * create (or import), and only in userland.
956 */
957 if (ret == 0 && !isopen &&
958 (strncmp(pool, "mirror", 6) == 0 ||
959 strncmp(pool, "raidz", 5) == 0 ||
960 strncmp(pool, "spare", 5) == 0 ||
961 strcmp(pool, "log") == 0)) {
962 if (hdl != NULL)
963 zfs_error_aux(hdl,
964 dgettext(TEXT_DOMAIN, "name is reserved"));
965 return (B_FALSE);
966 }
967
968
969 if (ret != 0) {
970 if (hdl != NULL) {
971 switch (why) {
972 case NAME_ERR_TOOLONG:
973 zfs_error_aux(hdl,
974 dgettext(TEXT_DOMAIN, "name is too long"));
975 break;
976
977 case NAME_ERR_INVALCHAR:
978 zfs_error_aux(hdl,
979 dgettext(TEXT_DOMAIN, "invalid character "
980 "'%c' in pool name"), what);
981 break;
982
983 case NAME_ERR_NOLETTER:
984 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
985 "name must begin with a letter"));
986 break;
987
988 case NAME_ERR_RESERVED:
989 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
990 "name is reserved"));
991 break;
992
993 case NAME_ERR_DISKLIKE:
994 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
995 "pool name is reserved"));
996 break;
997
998 case NAME_ERR_LEADING_SLASH:
999 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1000 "leading slash in name"));
1001 break;
1002
1003 case NAME_ERR_EMPTY_COMPONENT:
1004 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1005 "empty component in name"));
1006 break;
1007
1008 case NAME_ERR_TRAILING_SLASH:
1009 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1010 "trailing slash in name"));
1011 break;
1012
1013 case NAME_ERR_MULTIPLE_DELIMITERS:
1014 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1015 "multiple '@' and/or '#' delimiters in "
1016 "name"));
1017 break;
1018
1019 default:
1020 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1021 "(%d) not defined"), why);
1022 break;
1023 }
1024 }
1025 return (B_FALSE);
1026 }
1027
1028 return (B_TRUE);
1029 }
1030
1031 /*
1032 * Open a handle to the given pool, even if the pool is currently in the FAULTED
1033 * state.
1034 */
1035 zpool_handle_t *
1036 zpool_open_canfail(libzfs_handle_t *hdl, const char *pool)
1037 {
1038 zpool_handle_t *zhp;
1039 boolean_t missing;
1040
1041 /*
1042 * Make sure the pool name is valid.
1043 */
1044 if (!zpool_name_valid(hdl, B_TRUE, pool)) {
1045 (void) zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1046 dgettext(TEXT_DOMAIN, "cannot open '%s'"),
1047 pool);
1048 return (NULL);
1049 }
1050
1051 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
1052 return (NULL);
1053
1054 zhp->zpool_hdl = hdl;
1055 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
1056
1057 if (zpool_refresh_stats(zhp, &missing) != 0) {
1058 zpool_close(zhp);
1059 return (NULL);
1060 }
1061
1062 if (missing) {
1063 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool"));
1064 (void) zfs_error_fmt(hdl, EZFS_NOENT,
1065 dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool);
1066 zpool_close(zhp);
1067 return (NULL);
1068 }
1069
1070 return (zhp);
1071 }
1072
1073 /*
1074 * Like the above, but silent on error. Used when iterating over pools (because
1075 * the configuration cache may be out of date).
1076 */
1077 int
1078 zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret)
1079 {
1080 zpool_handle_t *zhp;
1081 boolean_t missing;
1082
1083 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
1084 return (-1);
1085
1086 zhp->zpool_hdl = hdl;
1087 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
1088
1089 if (zpool_refresh_stats(zhp, &missing) != 0) {
1090 zpool_close(zhp);
1091 return (-1);
1092 }
1093
1094 if (missing) {
1095 zpool_close(zhp);
1096 *ret = NULL;
1097 return (0);
1098 }
1099
1100 *ret = zhp;
1101 return (0);
1102 }
1103
1104 /*
1105 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted
1106 * state.
1107 */
1108 zpool_handle_t *
1109 zpool_open(libzfs_handle_t *hdl, const char *pool)
1110 {
1111 zpool_handle_t *zhp;
1112
1113 if ((zhp = zpool_open_canfail(hdl, pool)) == NULL)
1114 return (NULL);
1115
1116 if (zhp->zpool_state == POOL_STATE_UNAVAIL) {
1117 (void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
1118 dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name);
1119 zpool_close(zhp);
1120 return (NULL);
1121 }
1122
1123 return (zhp);
1124 }
1125
1126 /*
1127 * Close the handle. Simply frees the memory associated with the handle.
1128 */
1129 void
1130 zpool_close(zpool_handle_t *zhp)
1131 {
1132 nvlist_free(zhp->zpool_config);
1133 nvlist_free(zhp->zpool_old_config);
1134 nvlist_free(zhp->zpool_props);
1135 free(zhp);
1136 }
1137
1138 /*
1139 * Return the name of the pool.
1140 */
1141 const char *
1142 zpool_get_name(zpool_handle_t *zhp)
1143 {
1144 return (zhp->zpool_name);
1145 }
1146
1147
1148 /*
1149 * Return the state of the pool (ACTIVE or UNAVAILABLE)
1150 */
1151 int
1152 zpool_get_state(zpool_handle_t *zhp)
1153 {
1154 return (zhp->zpool_state);
1155 }
1156
1157 /*
1158 * Create the named pool, using the provided vdev list. It is assumed
1159 * that the consumer has already validated the contents of the nvlist, so we
1160 * don't have to worry about error semantics.
1161 */
1162 int
1163 zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot,
1164 nvlist_t *props, nvlist_t *fsprops)
1165 {
1166 zfs_cmd_t zc = { 0 };
1167 nvlist_t *zc_fsprops = NULL;
1168 nvlist_t *zc_props = NULL;
1169 char msg[1024];
1170 int ret = -1;
1171 boolean_t wbc_mode_prop_exists = B_FALSE;
1172
1173 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1174 "cannot create '%s'"), pool);
1175
1176 if (!zpool_name_valid(hdl, B_FALSE, pool))
1177 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
1178
1179 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1180 return (-1);
1181
1182 if (props) {
1183 prop_flags_t flags = { .create = B_TRUE, .import = B_FALSE };
1184
1185 if ((zc_props = zpool_valid_proplist(hdl, pool, props,
1186 SPA_VERSION_1, flags, msg)) == NULL) {
1187 goto create_failed;
1188 }
1189 }
1190
1191 if (fsprops) {
1192 uint64_t zoned;
1193 char *zonestr;
1194
1195 zoned = ((nvlist_lookup_string(fsprops,
1196 zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) &&
1197 strcmp(zonestr, "on") == 0);
1198
1199 if ((zc_fsprops = zfs_valid_proplist(hdl, ZFS_TYPE_FILESYSTEM,
1200 fsprops, zoned, NULL, NULL, msg)) == NULL) {
1201 goto create_failed;
1202 }
1203
1204 if (nvlist_exists(zc_fsprops,
1205 zfs_prop_to_name(ZFS_PROP_WBC_MODE))) {
1206 wbc_mode_prop_exists = B_TRUE;
1207 }
1208
1209 if (!zc_props &&
1210 (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) {
1211 goto create_failed;
1212 }
1213 if (nvlist_add_nvlist(zc_props,
1214 ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) {
1215 goto create_failed;
1216 }
1217 }
1218
1219 if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
1220 goto create_failed;
1221
1222 (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name));
1223
1224 if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) {
1225
1226 zcmd_free_nvlists(&zc);
1227 nvlist_free(zc_props);
1228 nvlist_free(zc_fsprops);
1229
1230 switch (errno) {
1231 case EBUSY:
1232 /*
1233 * This can happen if the user has specified the same
1234 * device multiple times. We can't reliably detect this
1235 * until we try to add it and see we already have a
1236 * label.
1237 */
1238 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1239 "one or more vdevs refer to the same device"));
1240 return (zfs_error(hdl, EZFS_BADDEV, msg));
1241
1242 case ERANGE:
1243 /*
1244 * This happens if the record size is smaller or larger
1245 * than the allowed size range, or not a power of 2.
1246 *
1247 * NOTE: although zfs_valid_proplist is called earlier,
1248 * this case may have slipped through since the
1249 * pool does not exist yet and it is therefore
1250 * impossible to read properties e.g. max blocksize
1251 * from the pool.
1252 */
1253 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1254 "record size invalid"));
1255 return (zfs_error(hdl, EZFS_BADPROP, msg));
1256
1257 case EOVERFLOW:
1258 /*
1259 * This occurs when one of the devices is below
1260 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1261 * device was the problem device since there's no
1262 * reliable way to determine device size from userland.
1263 */
1264 {
1265 char buf[64];
1266
1267 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
1268
1269 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1270 "one or more devices is less than the "
1271 "minimum size (%s)"), buf);
1272 }
1273 return (zfs_error(hdl, EZFS_BADDEV, msg));
1274
1275 case ENOSPC:
1276 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1277 "one or more devices is out of space"));
1278 return (zfs_error(hdl, EZFS_BADDEV, msg));
1279
1280 case ENOTBLK:
1281 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1282 "cache device must be a disk or disk slice"));
1283 return (zfs_error(hdl, EZFS_BADDEV, msg));
1284
1285 case EALREADY:
1286 if (wbc_mode_prop_exists) {
1287 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1288 "WBC is already in the OFF state"));
1289 return (zfs_error(hdl, EZFS_WBCALREADY, msg));
1290 }
1291
1292 default:
1293 return (zpool_standard_error(hdl, errno, msg));
1294 }
1295 }
1296
1297 create_failed:
1298 zcmd_free_nvlists(&zc);
1299 nvlist_free(zc_props);
1300 nvlist_free(zc_fsprops);
1301 return (ret);
1302 }
1303
1304 /*
1305 * Destroy the given pool. It is up to the caller to ensure that there are no
1306 * datasets left in the pool.
1307 */
1308 int
1309 zpool_destroy(zpool_handle_t *zhp, const char *log_str)
1310 {
1311 zfs_cmd_t zc = { 0 };
1312 zfs_handle_t *zfp = NULL;
1313 libzfs_handle_t *hdl = zhp->zpool_hdl;
1314 char msg[1024];
1315
1316 if (zhp->zpool_state == POOL_STATE_ACTIVE &&
1317 (zfp = zfs_open(hdl, zhp->zpool_name, ZFS_TYPE_FILESYSTEM)) == NULL)
1318 return (-1);
1319
1320 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1321 zc.zc_history = (uint64_t)(uintptr_t)log_str;
1322
1323 if (zfs_ioctl(hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) {
1324 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1325 "cannot destroy '%s'"), zhp->zpool_name);
1326
1327 if (errno == EROFS) {
1328 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1329 "one or more devices is read only"));
1330 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1331 } else {
1332 (void) zpool_standard_error(hdl, errno, msg);
1333 }
1334
1335 if (zfp)
1336 zfs_close(zfp);
1337 return (-1);
1338 }
1339
1340 if (zfp) {
1341 remove_mountpoint(zfp);
1342 zfs_close(zfp);
1343 }
1344
1345 return (0);
1346 }
1347
1348 /*
1349 * Add the given vdevs to the pool. The caller must have already performed the
1350 * necessary verification to ensure that the vdev specification is well-formed.
1351 */
1352 int
1353 zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot)
1354 {
1355 zfs_cmd_t zc = { 0 };
1356 int ret;
1357 libzfs_handle_t *hdl = zhp->zpool_hdl;
1358 char msg[1024];
1359 nvlist_t **spares, **l2cache;
1360 uint_t nspares, nl2cache;
1361
1362 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1363 "cannot add to '%s'"), zhp->zpool_name);
1364
1365 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1366 SPA_VERSION_SPARES &&
1367 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
1368 &spares, &nspares) == 0) {
1369 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1370 "upgraded to add hot spares"));
1371 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1372 }
1373
1374 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1375 SPA_VERSION_L2CACHE &&
1376 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
1377 &l2cache, &nl2cache) == 0) {
1378 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1379 "upgraded to add cache devices"));
1380 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1381 }
1382
1383 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1384 return (-1);
1385 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1386
1387 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) {
1388 switch (errno) {
1389 case EBUSY:
1390 /*
1391 * This can happen if the user has specified the same
1392 * device multiple times. We can't reliably detect this
1393 * until we try to add it and see we already have a
1394 * label.
1395 */
1396 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1397 "one or more vdevs refer to the same device"));
1398 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1399 break;
1400
1401 case EOVERFLOW:
1402 /*
1403 * This occurrs when one of the devices is below
1404 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1405 * device was the problem device since there's no
1406 * reliable way to determine device size from userland.
1407 */
1408 {
1409 char buf[64];
1410
1411 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
1412
1413 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1414 "device is less than the minimum "
1415 "size (%s)"), buf);
1416 }
1417 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1418 break;
1419
1420 case ENOTSUP:
1421 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1422 "pool must be upgraded to add these vdevs"));
1423 (void) zfs_error(hdl, EZFS_BADVERSION, msg);
1424 break;
1425
1426 case EDOM:
1427 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1428 "root pool can not have multiple vdevs"
1429 " or separate logs"));
1430 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, msg);
1431 break;
1432
1433 case ENOTBLK:
1434 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1435 "cache device must be a disk or disk slice"));
1436 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1437 break;
1438
1439 default:
1440 (void) zpool_standard_error(hdl, errno, msg);
1441 }
1442
1443 ret = -1;
1444 } else {
1445 ret = 0;
1446 }
1447
1448 zcmd_free_nvlists(&zc);
1449
1450 return (ret);
1451 }
1452
1453 /*
1454 * Exports the pool from the system. The caller must ensure that there are no
1455 * mounted datasets in the pool.
1456 */
1457 static int
1458 zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce,
1459 boolean_t saveconfig, const char *log_str)
1460 {
1461 zfs_cmd_t zc = { 0 };
1462 char msg[1024];
1463
1464 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1465 "cannot export '%s'"), zhp->zpool_name);
1466
1467 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1468 zc.zc_cookie = force;
1469 zc.zc_guid = hardforce;
1470 zc.zc_obj = saveconfig;
1471 zc.zc_history = (uint64_t)(uintptr_t)log_str;
1472
1473 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) {
1474 switch (errno) {
1475 case EXDEV:
1476 zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
1477 "use '-f' to override the following errors:\n"
1478 "'%s' has an active shared spare which could be"
1479 " used by other pools once '%s' is exported."),
1480 zhp->zpool_name, zhp->zpool_name);
1481 return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE,
1482 msg));
1483 default:
1484 return (zpool_standard_error_fmt(zhp->zpool_hdl, errno,
1485 msg));
1486 }
1487 }
1488
1489 return (0);
1490 }
1491
1492 int
1493 zpool_export(zpool_handle_t *zhp, boolean_t force, boolean_t saveconfig,
1494 const char *log_str)
1495 {
1496 return (zpool_export_common(zhp, force, B_FALSE, saveconfig, log_str));
1497 }
1498
1499 int
1500 zpool_export_force(zpool_handle_t *zhp, boolean_t saveconfig,
1501 const char *log_str)
1502 {
1503 return (zpool_export_common(zhp, B_TRUE, B_TRUE, saveconfig, log_str));
1504 }
1505
1506 static void
1507 zpool_rewind_exclaim(libzfs_handle_t *hdl, const char *name, boolean_t dryrun,
1508 nvlist_t *config)
1509 {
1510 nvlist_t *nv = NULL;
1511 uint64_t rewindto;
1512 int64_t loss = -1;
1513 struct tm t;
1514 char timestr[128];
1515
1516 if (!hdl->libzfs_printerr || config == NULL)
1517 return;
1518
1519 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
1520 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0) {
1521 return;
1522 }
1523
1524 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1525 return;
1526 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1527
1528 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1529 strftime(timestr, 128, 0, &t) != 0) {
1530 if (dryrun) {
1531 (void) printf(dgettext(TEXT_DOMAIN,
1532 "Would be able to return %s "
1533 "to its state as of %s.\n"),
1534 name, timestr);
1535 } else {
1536 (void) printf(dgettext(TEXT_DOMAIN,
1537 "Pool %s returned to its state as of %s.\n"),
1538 name, timestr);
1539 }
1540 if (loss > 120) {
1541 (void) printf(dgettext(TEXT_DOMAIN,
1542 "%s approximately %lld "),
1543 dryrun ? "Would discard" : "Discarded",
1544 (loss + 30) / 60);
1545 (void) printf(dgettext(TEXT_DOMAIN,
1546 "minutes of transactions.\n"));
1547 } else if (loss > 0) {
1548 (void) printf(dgettext(TEXT_DOMAIN,
1549 "%s approximately %lld "),
1550 dryrun ? "Would discard" : "Discarded", loss);
1551 (void) printf(dgettext(TEXT_DOMAIN,
1552 "seconds of transactions.\n"));
1553 }
1554 }
1555 }
1556
1557 void
1558 zpool_explain_recover(libzfs_handle_t *hdl, const char *name, int reason,
1559 nvlist_t *config)
1560 {
1561 nvlist_t *nv = NULL;
1562 int64_t loss = -1;
1563 uint64_t edata = UINT64_MAX;
1564 uint64_t rewindto;
1565 struct tm t;
1566 char timestr[128];
1567
1568 if (!hdl->libzfs_printerr)
1569 return;
1570
1571 if (reason >= 0)
1572 (void) printf(dgettext(TEXT_DOMAIN, "action: "));
1573 else
1574 (void) printf(dgettext(TEXT_DOMAIN, "\t"));
1575
1576 /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */
1577 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
1578 nvlist_lookup_nvlist(nv, ZPOOL_CONFIG_REWIND_INFO, &nv) != 0 ||
1579 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
1580 goto no_info;
1581
1582 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1583 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_DATA_ERRORS,
1584 &edata);
1585
1586 (void) printf(dgettext(TEXT_DOMAIN,
1587 "Recovery is possible, but will result in some data loss.\n"));
1588
1589 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
1590 strftime(timestr, 128, 0, &t) != 0) {
1591 (void) printf(dgettext(TEXT_DOMAIN,
1592 "\tReturning the pool to its state as of %s\n"
1593 "\tshould correct the problem. "),
1594 timestr);
1595 } else {
1596 (void) printf(dgettext(TEXT_DOMAIN,
1597 "\tReverting the pool to an earlier state "
1598 "should correct the problem.\n\t"));
1599 }
1600
1601 if (loss > 120) {
1602 (void) printf(dgettext(TEXT_DOMAIN,
1603 "Approximately %lld minutes of data\n"
1604 "\tmust be discarded, irreversibly. "), (loss + 30) / 60);
1605 } else if (loss > 0) {
1606 (void) printf(dgettext(TEXT_DOMAIN,
1607 "Approximately %lld seconds of data\n"
1608 "\tmust be discarded, irreversibly. "), loss);
1609 }
1610 if (edata != 0 && edata != UINT64_MAX) {
1611 if (edata == 1) {
1612 (void) printf(dgettext(TEXT_DOMAIN,
1613 "After rewind, at least\n"
1614 "\tone persistent user-data error will remain. "));
1615 } else {
1616 (void) printf(dgettext(TEXT_DOMAIN,
1617 "After rewind, several\n"
1618 "\tpersistent user-data errors will remain. "));
1619 }
1620 }
1621 (void) printf(dgettext(TEXT_DOMAIN,
1622 "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "),
1623 reason >= 0 ? "clear" : "import", name);
1624
1625 (void) printf(dgettext(TEXT_DOMAIN,
1626 "A scrub of the pool\n"
1627 "\tis strongly recommended after recovery.\n"));
1628 return;
1629
1630 no_info:
1631 (void) printf(dgettext(TEXT_DOMAIN,
1632 "Destroy and re-create the pool from\n\ta backup source.\n"));
1633 }
1634
1635 /*
1636 * zpool_import() is a contracted interface. Should be kept the same
1637 * if possible.
1638 *
1639 * Applications should use zpool_import_props() to import a pool with
1640 * new properties value to be set.
1641 */
1642 int
1643 zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1644 char *altroot)
1645 {
1646 nvlist_t *props = NULL;
1647 int ret;
1648
1649 if (altroot != NULL) {
1650 if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) {
1651 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1652 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1653 newname));
1654 }
1655
1656 if (nvlist_add_string(props,
1657 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 ||
1658 nvlist_add_string(props,
1659 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) {
1660 nvlist_free(props);
1661 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1662 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1663 newname));
1664 }
1665 }
1666
1667 ret = zpool_import_props(hdl, config, newname, props,
1668 ZFS_IMPORT_NORMAL);
1669 nvlist_free(props);
1670 return (ret);
1671 }
1672
1673 static void
1674 print_vdev_tree(libzfs_handle_t *hdl, const char *name, nvlist_t *nv,
1675 int indent)
1676 {
1677 nvlist_t **child;
1678 uint_t c, children;
1679 char *vname;
1680 uint64_t is_log = 0;
1681
1682 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG,
1683 &is_log);
1684
1685 if (name != NULL)
1686 (void) printf("\t%*s%s%s\n", indent, "", name,
1687 is_log ? " [log]" : "");
1688
1689 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1690 &child, &children) != 0)
1691 return;
1692
1693 for (c = 0; c < children; c++) {
1694 vname = zpool_vdev_name(hdl, NULL, child[c], B_TRUE);
1695 print_vdev_tree(hdl, vname, child[c], indent + 2);
1696 free(vname);
1697 }
1698 }
1699
1700 void
1701 zpool_print_unsup_feat(nvlist_t *config)
1702 {
1703 nvlist_t *nvinfo, *unsup_feat;
1704
1705 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nvinfo) ==
1706 0);
1707 verify(nvlist_lookup_nvlist(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT,
1708 &unsup_feat) == 0);
1709
1710 for (nvpair_t *nvp = nvlist_next_nvpair(unsup_feat, NULL); nvp != NULL;
1711 nvp = nvlist_next_nvpair(unsup_feat, nvp)) {
1712 char *desc;
1713
1714 verify(nvpair_type(nvp) == DATA_TYPE_STRING);
1715 verify(nvpair_value_string(nvp, &desc) == 0);
1716
1717 if (strlen(desc) > 0)
1718 (void) printf("\t%s (%s)\n", nvpair_name(nvp), desc);
1719 else
1720 (void) printf("\t%s\n", nvpair_name(nvp));
1721 }
1722 }
1723
1724 /*
1725 * Import the given pool using the known configuration and a list of
1726 * properties to be set. The configuration should have come from
1727 * zpool_find_import(). The 'newname' parameters control whether the pool
1728 * is imported with a different name.
1729 */
1730 int
1731 zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1732 nvlist_t *props, int flags)
1733 {
1734 zfs_cmd_t zc = { 0 };
1735 zpool_rewind_policy_t policy;
1736 nvlist_t *nv = NULL;
1737 nvlist_t *nvinfo = NULL;
1738 nvlist_t *missing = NULL;
1739 char *thename;
1740 char *origname;
1741 int ret;
1742 int error = 0;
1743 char errbuf[1024];
1744
1745 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1746 &origname) == 0);
1747
1748 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1749 "cannot import pool '%s'"), origname);
1750
1751 if (newname != NULL) {
1752 if (!zpool_name_valid(hdl, B_FALSE, newname))
1753 return (zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1754 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1755 newname));
1756 thename = (char *)newname;
1757 } else {
1758 thename = origname;
1759 }
1760
1761 if (props != NULL) {
1762 uint64_t version;
1763 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
1764
1765 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
1766 &version) == 0);
1767
1768 if ((props = zpool_valid_proplist(hdl, origname,
1769 props, version, flags, errbuf)) == NULL)
1770 return (-1);
1771 if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) {
1772 nvlist_free(props);
1773 return (-1);
1774 }
1775 nvlist_free(props);
1776 }
1777
1778 (void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name));
1779
1780 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1781 &zc.zc_guid) == 0);
1782
1783 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) {
1784 zcmd_free_nvlists(&zc);
1785 return (-1);
1786 }
1787 if (zcmd_alloc_dst_nvlist(hdl, &zc, zc.zc_nvlist_conf_size * 2) != 0) {
1788 zcmd_free_nvlists(&zc);
1789 return (-1);
1790 }
1791
1792 zc.zc_cookie = flags;
1793 while ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc)) != 0 &&
1794 errno == ENOMEM) {
1795 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
1796 zcmd_free_nvlists(&zc);
1797 return (-1);
1798 }
1799 }
1800 if (ret != 0)
1801 error = errno;
1802
1803 (void) zcmd_read_dst_nvlist(hdl, &zc, &nv);
1804
1805 zcmd_free_nvlists(&zc);
1806
1807 zpool_get_rewind_policy(config, &policy);
1808
1809 if (error) {
1810 char desc[1024];
1811
1812 /*
1813 * Dry-run failed, but we print out what success
1814 * looks like if we found a best txg
1815 */
1816 if (policy.zrp_request & ZPOOL_TRY_REWIND) {
1817 zpool_rewind_exclaim(hdl, newname ? origname : thename,
1818 B_TRUE, nv);
1819 nvlist_free(nv);
1820 return (-1);
1821 }
1822
1823 if (newname == NULL)
1824 (void) snprintf(desc, sizeof (desc),
1825 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1826 thename);
1827 else
1828 (void) snprintf(desc, sizeof (desc),
1829 dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"),
1830 origname, thename);
1831
1832 switch (error) {
1833 case ENOTSUP:
1834 if (nv != NULL && nvlist_lookup_nvlist(nv,
1835 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
1836 nvlist_exists(nvinfo, ZPOOL_CONFIG_UNSUP_FEAT)) {
1837 (void) printf(dgettext(TEXT_DOMAIN, "This "
1838 "pool uses the following feature(s) not "
1839 "supported by this system:\n"));
1840 zpool_print_unsup_feat(nv);
1841 if (nvlist_exists(nvinfo,
1842 ZPOOL_CONFIG_CAN_RDONLY)) {
1843 (void) printf(dgettext(TEXT_DOMAIN,
1844 "All unsupported features are only "
1845 "required for writing to the pool."
1846 "\nThe pool can be imported using "
1847 "'-o readonly=on'.\n"));
1848 }
1849 }
1850 /*
1851 * Unsupported version.
1852 */
1853 (void) zfs_error(hdl, EZFS_BADVERSION, desc);
1854 break;
1855
1856 case EINVAL:
1857 (void) zfs_error(hdl, EZFS_INVALCONFIG, desc);
1858 break;
1859
1860 case EROFS:
1861 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1862 "one or more devices is read only"));
1863 (void) zfs_error(hdl, EZFS_BADDEV, desc);
1864 break;
1865
1866 case ENXIO:
1867 if (nv && nvlist_lookup_nvlist(nv,
1868 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
1869 nvlist_lookup_nvlist(nvinfo,
1870 ZPOOL_CONFIG_MISSING_DEVICES, &missing) == 0) {
1871 (void) printf(dgettext(TEXT_DOMAIN,
1872 "The devices below are missing, use "
1873 "'-m' to import the pool anyway:\n"));
1874 print_vdev_tree(hdl, NULL, missing, 2);
1875 (void) printf("\n");
1876 }
1877 (void) zpool_standard_error(hdl, error, desc);
1878 break;
1879
1880 case EEXIST:
1881 (void) zpool_standard_error(hdl, error, desc);
1882 break;
1883 case ENAMETOOLONG:
1884 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1885 "new name of at least one dataset is longer than "
1886 "the maximum allowable length"));
1887 (void) zfs_error(hdl, EZFS_NAMETOOLONG, desc);
1888 break;
1889 default:
1890 (void) zpool_standard_error(hdl, error, desc);
1891 zpool_explain_recover(hdl,
1892 newname ? origname : thename, -error, nv);
1893 break;
1894 }
1895
1896 nvlist_free(nv);
1897 ret = -1;
1898 } else {
1899 zpool_handle_t *zhp;
1900
1901 /*
1902 * This should never fail, but play it safe anyway.
1903 */
1904 if (zpool_open_silent(hdl, thename, &zhp) != 0)
1905 ret = -1;
1906 else if (zhp != NULL)
1907 zpool_close(zhp);
1908 if (policy.zrp_request &
1909 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
1910 zpool_rewind_exclaim(hdl, newname ? origname : thename,
1911 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0), nv);
1912 }
1913 nvlist_free(nv);
1914 return (0);
1915 }
1916
1917 return (ret);
1918 }
1919
1920 /*
1921 * Scan the pool.
1922 */
1923 int
1924 zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func, pool_scrub_cmd_t cmd)
1925 {
1926 zfs_cmd_t zc = { 0 };
1927 char msg[1024];
1928 int err;
1929 libzfs_handle_t *hdl = zhp->zpool_hdl;
1930
1931 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1932 zc.zc_cookie = func;
1933 zc.zc_flags = cmd;
1934
1935 if (zfs_ioctl(hdl, ZFS_IOC_POOL_SCAN, &zc) == 0)
1936 return (0);
1937
1938 err = errno;
1939
1940 /* ECANCELED on a scrub means we resumed a paused scrub */
1941 if (err == ECANCELED && func == POOL_SCAN_SCRUB &&
1942 cmd == POOL_SCRUB_NORMAL)
1943 return (0);
1944
1945 if (err == ENOENT && func != POOL_SCAN_NONE && cmd == POOL_SCRUB_NORMAL)
1946 return (0);
1947
1948 if (func == POOL_SCAN_SCRUB) {
1949 if (cmd == POOL_SCRUB_PAUSE) {
1950 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1951 "cannot pause scrubbing %s"), zc.zc_name);
1952 } else {
1953 assert(cmd == POOL_SCRUB_NORMAL);
1954 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1955 "cannot scrub %s"), zc.zc_name);
1956 }
1957 } else if (func == POOL_SCAN_NONE) {
1958 (void) snprintf(msg, sizeof (msg),
1959 dgettext(TEXT_DOMAIN, "cannot cancel scrubbing %s"),
1960 zc.zc_name);
1961 } else {
1962 assert(!"unexpected result");
1963 }
1964
1965 if (err == EBUSY) {
1966 nvlist_t *nvroot;
1967 pool_scan_stat_t *ps = NULL;
1968 uint_t psc;
1969
1970 verify(nvlist_lookup_nvlist(zhp->zpool_config,
1971 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
1972 (void) nvlist_lookup_uint64_array(nvroot,
1973 ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc);
1974 if (ps && ps->pss_func == POOL_SCAN_SCRUB) {
1975 if (cmd == POOL_SCRUB_PAUSE)
1976 return (zfs_error(hdl, EZFS_SCRUB_PAUSED, msg));
1977 else
1978 return (zfs_error(hdl, EZFS_SCRUBBING, msg));
1979 } else {
1980 return (zfs_error(hdl, EZFS_RESILVERING, msg));
1981 }
1982 } else if (err == ENOENT) {
1983 return (zfs_error(hdl, EZFS_NO_SCRUB, msg));
1984 } else {
1985 return (zpool_standard_error(hdl, err, msg));
1986 }
1987 }
1988
1989 /*
1990 * Trim the pool.
1991 */
1992 int
1993 zpool_trim(zpool_handle_t *zhp, boolean_t start, uint64_t rate)
1994 {
1995 zfs_cmd_t zc = { 0 };
1996 char msg[1024];
1997 libzfs_handle_t *hdl = zhp->zpool_hdl;
1998 trim_cmd_info_t tci = { .tci_start = start, .tci_rate = rate };
1999
2000 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2001 zc.zc_cookie = (uintptr_t)&tci;
2002
2003 if (zfs_ioctl(hdl, ZFS_IOC_POOL_TRIM, &zc) == 0)
2004 return (0);
2005
2006 return (zpool_standard_error(hdl, errno, msg));
2007 }
2008
2009 /*
2010 * This provides a very minimal check whether a given string is likely a
2011 * c#t#d# style string. Users of this are expected to do their own
2012 * verification of the s# part.
2013 */
2014 #define CTD_CHECK(str) (str && str[0] == 'c' && isdigit(str[1]))
2015
2016 /*
2017 * More elaborate version for ones which may start with "/dev/dsk/"
2018 * and the like.
2019 */
2020 static int
2021 ctd_check_path(char *str)
2022 {
2023 /*
2024 * If it starts with a slash, check the last component.
2025 */
2026 if (str && str[0] == '/') {
2027 char *tmp = strrchr(str, '/');
2028
2029 /*
2030 * If it ends in "/old", check the second-to-last
2031 * component of the string instead.
2032 */
2033 if (tmp != str && strcmp(tmp, "/old") == 0) {
2034 for (tmp--; *tmp != '/'; tmp--)
2035 ;
2036 }
2037 str = tmp + 1;
2038 }
2039 return (CTD_CHECK(str));
2040 }
2041
2042 /*
2043 * Find a vdev that matches the search criteria specified. We use the
2044 * the nvpair name to determine how we should look for the device.
2045 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
2046 * spare; but FALSE if its an INUSE spare.
2047 */
2048 static nvlist_t *
2049 vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare,
2050 boolean_t *l2cache, boolean_t *log, boolean_t *special)
2051 {
2052 uint_t c, children;
2053 nvlist_t **child;
2054 nvlist_t *ret;
2055 uint64_t is_log, is_special;
2056 char *srchkey;
2057 nvpair_t *pair = nvlist_next_nvpair(search, NULL);
2058
2059 /* Nothing to look for */
2060 if (search == NULL || pair == NULL)
2061 return (NULL);
2062
2063 /* Obtain the key we will use to search */
2064 srchkey = nvpair_name(pair);
2065
2066 switch (nvpair_type(pair)) {
2067 case DATA_TYPE_UINT64:
2068 if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) {
2069 uint64_t srchval, theguid;
2070
2071 verify(nvpair_value_uint64(pair, &srchval) == 0);
2072 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
2073 &theguid) == 0);
2074 if (theguid == srchval)
2075 return (nv);
2076 }
2077 break;
2078
2079 case DATA_TYPE_STRING: {
2080 char *srchval, *val;
2081
2082 verify(nvpair_value_string(pair, &srchval) == 0);
2083 if (nvlist_lookup_string(nv, srchkey, &val) != 0)
2084 break;
2085
2086 /*
2087 * Search for the requested value. Special cases:
2088 *
2089 * - ZPOOL_CONFIG_PATH for whole disk entries. To support
2090 * UEFI boot, these end in "s0" or "s0/old" or "s1" or
2091 * "s1/old". The "s0" or "s1" part is hidden from the user,
2092 * but included in the string, so this matches around it.
2093 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE).
2094 *
2095 * Otherwise, all other searches are simple string compares.
2096 */
2097 if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0 &&
2098 ctd_check_path(val)) {
2099 uint64_t wholedisk = 0;
2100
2101 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
2102 &wholedisk);
2103 if (wholedisk) {
2104 int slen = strlen(srchval);
2105 int vlen = strlen(val);
2106
2107 if (slen != vlen - 2)
2108 break;
2109
2110 /*
2111 * make_leaf_vdev() should only set
2112 * wholedisk for ZPOOL_CONFIG_PATHs which
2113 * will include "/dev/dsk/", giving plenty of
2114 * room for the indices used next.
2115 */
2116 ASSERT(vlen >= 6);
2117
2118 /*
2119 * strings identical except trailing "s0"
2120 */
2121 if ((strcmp(&val[vlen - 2], "s0") == 0 ||
2122 strcmp(&val[vlen - 2], "s1") == 0) &&
2123 strncmp(srchval, val, slen) == 0)
2124 return (nv);
2125
2126 /*
2127 * strings identical except trailing "s0/old"
2128 */
2129 if ((strcmp(&val[vlen - 6], "s0/old") == 0 ||
2130 strcmp(&val[vlen - 6], "s1/old") == 0) &&
2131 strcmp(&srchval[slen - 4], "/old") == 0 &&
2132 strncmp(srchval, val, slen - 4) == 0)
2133 return (nv);
2134
2135 break;
2136 }
2137 } else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0 && val) {
2138 char *type, *idx, *end, *p;
2139 uint64_t id, vdev_id;
2140
2141 /*
2142 * Determine our vdev type, keeping in mind
2143 * that the srchval is composed of a type and
2144 * vdev id pair (i.e. mirror-4).
2145 */
2146 if ((type = strdup(srchval)) == NULL)
2147 return (NULL);
2148
2149 if ((p = strrchr(type, '-')) == NULL) {
2150 free(type);
2151 break;
2152 }
2153 idx = p + 1;
2154 *p = '\0';
2155
2156 /*
2157 * If the types don't match then keep looking.
2158 */
2159 if (strncmp(val, type, strlen(val)) != 0) {
2160 free(type);
2161 break;
2162 }
2163
2164 verify(strncmp(type, VDEV_TYPE_RAIDZ,
2165 strlen(VDEV_TYPE_RAIDZ)) == 0 ||
2166 strncmp(type, VDEV_TYPE_MIRROR,
2167 strlen(VDEV_TYPE_MIRROR)) == 0);
2168 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
2169 &id) == 0);
2170
2171 errno = 0;
2172 vdev_id = strtoull(idx, &end, 10);
2173
2174 free(type);
2175 if (errno != 0)
2176 return (NULL);
2177
2178 /*
2179 * Now verify that we have the correct vdev id.
2180 */
2181 if (vdev_id == id)
2182 return (nv);
2183 }
2184
2185 /*
2186 * Common case
2187 */
2188 if (strcmp(srchval, val) == 0)
2189 return (nv);
2190 break;
2191 }
2192
2193 default:
2194 break;
2195 }
2196
2197 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
2198 &child, &children) != 0)
2199 return (NULL);
2200
2201 for (c = 0; c < children; c++) {
2202 if ((ret = vdev_to_nvlist_iter(child[c], search,
2203 avail_spare, l2cache, NULL, NULL)) != NULL) {
2204 /*
2205 * The 'is_log' value is only set for the toplevel
2206 * vdev, not the leaf vdevs. So we always lookup the
2207 * log device from the root of the vdev tree (where
2208 * 'log' is non-NULL).
2209 */
2210 if (log != NULL &&
2211 nvlist_lookup_uint64(child[c],
2212 ZPOOL_CONFIG_IS_LOG, &is_log) == 0 &&
2213 is_log) {
2214 *log = B_TRUE;
2215 }
2216
2217 if (special != NULL &&
2218 nvlist_lookup_uint64(child[c],
2219 ZPOOL_CONFIG_IS_SPECIAL, &is_special) == 0 &&
2220 is_special) {
2221 *special = B_TRUE;
2222 }
2223
2224 return (ret);
2225 }
2226 }
2227
2228 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
2229 &child, &children) == 0) {
2230 for (c = 0; c < children; c++) {
2231 if ((ret = vdev_to_nvlist_iter(child[c], search,
2232 avail_spare, l2cache, NULL, NULL)) != NULL) {
2233 *avail_spare = B_TRUE;
2234 return (ret);
2235 }
2236 }
2237 }
2238
2239 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
2240 &child, &children) == 0) {
2241 for (c = 0; c < children; c++) {
2242 if ((ret = vdev_to_nvlist_iter(child[c], search,
2243 avail_spare, l2cache, NULL, NULL)) != NULL) {
2244 *l2cache = B_TRUE;
2245 return (ret);
2246 }
2247 }
2248 }
2249
2250 return (NULL);
2251 }
2252
2253 /*
2254 * Given a physical path (minus the "/devices" prefix), find the
2255 * associated vdev.
2256 */
2257 nvlist_t *
2258 zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath,
2259 boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log)
2260 {
2261 nvlist_t *search, *nvroot, *ret;
2262
2263 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2264 verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath) == 0);
2265
2266 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
2267 &nvroot) == 0);
2268
2269 *avail_spare = B_FALSE;
2270 *l2cache = B_FALSE;
2271 if (log != NULL)
2272 *log = B_FALSE;
2273 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log,
2274 NULL);
2275 nvlist_free(search);
2276
2277 return (ret);
2278 }
2279
2280 /*
2281 * Determine if we have an "interior" top-level vdev (i.e mirror/raidz).
2282 */
2283 boolean_t
2284 zpool_vdev_is_interior(const char *name)
2285 {
2286 if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 ||
2287 strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0)
2288 return (B_TRUE);
2289 return (B_FALSE);
2290 }
2291
2292 nvlist_t *
2293 zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare,
2294 boolean_t *l2cache, boolean_t *log, boolean_t *special)
2295 {
2296 char buf[MAXPATHLEN];
2297 char *end;
2298 nvlist_t *nvroot, *search, *ret;
2299 uint64_t guid;
2300
2301 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
2302
2303 guid = strtoull(path, &end, 10);
2304 if (guid != 0 && *end == '\0') {
2305 verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0);
2306 } else if (zpool_vdev_is_interior(path)) {
2307 verify(nvlist_add_string(search, ZPOOL_CONFIG_TYPE, path) == 0);
2308 } else if (path[0] != '/') {
2309 (void) snprintf(buf, sizeof (buf), "%s/%s", ZFS_DISK_ROOT,
2310 path);
2311 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, buf) == 0);
2312 } else {
2313 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0);
2314 }
2315
2316 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
2317 &nvroot) == 0);
2318
2319 *avail_spare = B_FALSE;
2320 *l2cache = B_FALSE;
2321 if (log != NULL)
2322 *log = B_FALSE;
2323
2324 if (special != NULL)
2325 *special = B_FALSE;
2326
2327 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log,
2328 special);
2329 nvlist_free(search);
2330
2331 return (ret);
2332 }
2333
2334 static int
2335 vdev_online(nvlist_t *nv)
2336 {
2337 uint64_t ival;
2338
2339 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 ||
2340 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 ||
2341 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0)
2342 return (0);
2343
2344 return (1);
2345 }
2346
2347 /*
2348 * Helper function for zpool_get_physpaths().
2349 */
2350 static int
2351 vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size,
2352 size_t *bytes_written)
2353 {
2354 size_t bytes_left, pos, rsz;
2355 char *tmppath;
2356 const char *format;
2357
2358 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH,
2359 &tmppath) != 0)
2360 return (EZFS_NODEVICE);
2361
2362 pos = *bytes_written;
2363 bytes_left = physpath_size - pos;
2364 format = (pos == 0) ? "%s" : " %s";
2365
2366 rsz = snprintf(physpath + pos, bytes_left, format, tmppath);
2367 *bytes_written += rsz;
2368
2369 if (rsz >= bytes_left) {
2370 /* if physpath was not copied properly, clear it */
2371 if (bytes_left != 0) {
2372 physpath[pos] = 0;
2373 }
2374 return (EZFS_NOSPC);
2375 }
2376 return (0);
2377 }
2378
2379 static int
2380 vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size,
2381 size_t *rsz, boolean_t is_spare)
2382 {
2383 char *type;
2384 int ret;
2385
2386 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0)
2387 return (EZFS_INVALCONFIG);
2388
2389 if (strcmp(type, VDEV_TYPE_DISK) == 0) {
2390 /*
2391 * An active spare device has ZPOOL_CONFIG_IS_SPARE set.
2392 * For a spare vdev, we only want to boot from the active
2393 * spare device.
2394 */
2395 if (is_spare) {
2396 uint64_t spare = 0;
2397 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE,
2398 &spare);
2399 if (!spare)
2400 return (EZFS_INVALCONFIG);
2401 }
2402
2403 if (vdev_online(nv)) {
2404 if ((ret = vdev_get_one_physpath(nv, physpath,
2405 phypath_size, rsz)) != 0)
2406 return (ret);
2407 }
2408 } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 ||
2409 strcmp(type, VDEV_TYPE_RAIDZ) == 0 ||
2410 strcmp(type, VDEV_TYPE_REPLACING) == 0 ||
2411 (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) {
2412 nvlist_t **child;
2413 uint_t count;
2414 int i, ret;
2415
2416 if (nvlist_lookup_nvlist_array(nv,
2417 ZPOOL_CONFIG_CHILDREN, &child, &count) != 0)
2418 return (EZFS_INVALCONFIG);
2419
2420 for (i = 0; i < count; i++) {
2421 ret = vdev_get_physpaths(child[i], physpath,
2422 phypath_size, rsz, is_spare);
2423 if (ret == EZFS_NOSPC)
2424 return (ret);
2425 }
2426 }
2427
2428 return (EZFS_POOL_INVALARG);
2429 }
2430
2431 /*
2432 * Get phys_path for a root pool config.
2433 * Return 0 on success; non-zero on failure.
2434 */
2435 static int
2436 zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size)
2437 {
2438 size_t rsz;
2439 nvlist_t *vdev_root;
2440 nvlist_t **child;
2441 uint_t count;
2442 char *type;
2443
2444 rsz = 0;
2445
2446 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
2447 &vdev_root) != 0)
2448 return (EZFS_INVALCONFIG);
2449
2450 if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 ||
2451 nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN,
2452 &child, &count) != 0)
2453 return (EZFS_INVALCONFIG);
2454
2455 /*
2456 * root pool can only have a single top-level vdev.
2457 */
2458 if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1)
2459 return (EZFS_POOL_INVALARG);
2460
2461 (void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz,
2462 B_FALSE);
2463
2464 /* No online devices */
2465 if (rsz == 0)
2466 return (EZFS_NODEVICE);
2467
2468 return (0);
2469 }
2470
2471 /*
2472 * Get phys_path for a root pool
2473 * Return 0 on success; non-zero on failure.
2474 */
2475 int
2476 zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size)
2477 {
2478 return (zpool_get_config_physpath(zhp->zpool_config, physpath,
2479 phypath_size));
2480 }
2481
2482 /*
2483 * If the device has being dynamically expanded then we need to relabel
2484 * the disk to use the new unallocated space.
2485 */
2486 static int
2487 zpool_relabel_disk(libzfs_handle_t *hdl, const char *name)
2488 {
2489 char path[MAXPATHLEN];
2490 char errbuf[1024];
2491 int fd, error;
2492 int (*_efi_use_whole_disk)(int);
2493
2494 if ((_efi_use_whole_disk = (int (*)(int))dlsym(RTLD_DEFAULT,
2495 "efi_use_whole_disk")) == NULL)
2496 return (-1);
2497
2498 (void) snprintf(path, sizeof (path), "%s/%s", ZFS_RDISK_ROOT, name);
2499
2500 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) {
2501 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2502 "relabel '%s': unable to open device"), name);
2503 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
2504 }
2505
2506 /*
2507 * It's possible that we might encounter an error if the device
2508 * does not have any unallocated space left. If so, we simply
2509 * ignore that error and continue on.
2510 */
2511 error = _efi_use_whole_disk(fd);
2512 (void) close(fd);
2513 if (error && error != VT_ENOSPC) {
2514 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
2515 "relabel '%s': unable to read disk capacity"), name);
2516 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
2517 }
2518 return (0);
2519 }
2520
2521 /*
2522 * Bring the specified vdev online. The 'flags' parameter is a set of the
2523 * ZFS_ONLINE_* flags.
2524 */
2525 int
2526 zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags,
2527 vdev_state_t *newstate)
2528 {
2529 zfs_cmd_t zc = { 0 };
2530 char msg[1024];
2531 char *pathname;
2532 nvlist_t *tgt;
2533 boolean_t avail_spare, l2cache, islog;
2534 libzfs_handle_t *hdl = zhp->zpool_hdl;
2535
2536 if (flags & ZFS_ONLINE_EXPAND) {
2537 (void) snprintf(msg, sizeof (msg),
2538 dgettext(TEXT_DOMAIN, "cannot expand %s"), path);
2539 } else {
2540 (void) snprintf(msg, sizeof (msg),
2541 dgettext(TEXT_DOMAIN, "cannot online %s"), path);
2542 }
2543
2544 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2545 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2546 &islog, NULL)) == NULL)
2547 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2548
2549 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2550
2551 if ((flags & ZFS_ONLINE_EXPAND ||
2552 zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) &&
2553 nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH, &pathname) == 0) {
2554 uint64_t wholedisk = 0;
2555
2556 (void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK,
2557 &wholedisk);
2558
2559 /*
2560 * XXX - L2ARC 1.0 devices can't support expansion.
2561 */
2562 if (l2cache) {
2563 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2564 "cannot expand cache devices"));
2565 return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg));
2566 }
2567
2568 if (wholedisk) {
2569 pathname += strlen(ZFS_DISK_ROOT) + 1;
2570 (void) zpool_relabel_disk(hdl, pathname);
2571 }
2572 }
2573
2574 zc.zc_cookie = VDEV_STATE_ONLINE;
2575 zc.zc_obj = flags;
2576
2577 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) {
2578 if (errno == EINVAL) {
2579 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split "
2580 "from this pool into a new one. Use '%s' "
2581 "instead"), "zpool detach");
2582 return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, msg));
2583 }
2584 return (zpool_standard_error(hdl, errno, msg));
2585 }
2586
2587 *newstate = zc.zc_cookie;
2588 return (0);
2589 }
2590
2591 /*
2592 * Take the specified vdev offline
2593 */
2594 int
2595 zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp)
2596 {
2597 zfs_cmd_t zc = { 0 };
2598 char msg[1024];
2599 nvlist_t *tgt;
2600 boolean_t avail_spare, l2cache;
2601 libzfs_handle_t *hdl = zhp->zpool_hdl;
2602
2603 (void) snprintf(msg, sizeof (msg),
2604 dgettext(TEXT_DOMAIN, "cannot offline %s"), path);
2605
2606 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2607 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2608 NULL, NULL)) == NULL)
2609 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2610
2611 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2612
2613 if (avail_spare)
2614 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2615
2616 zc.zc_cookie = VDEV_STATE_OFFLINE;
2617 zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0;
2618
2619 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2620 return (0);
2621
2622 switch (errno) {
2623 case EBUSY:
2624
2625 /*
2626 * There are no other replicas of this device.
2627 */
2628 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2629
2630 case EEXIST:
2631 /*
2632 * The log device has unplayed logs
2633 */
2634 return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg));
2635
2636 default:
2637 return (zpool_standard_error(hdl, errno, msg));
2638 }
2639 }
2640
2641 /*
2642 * Mark the given vdev faulted.
2643 */
2644 int
2645 zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2646 {
2647 zfs_cmd_t zc = { 0 };
2648 char msg[1024];
2649 libzfs_handle_t *hdl = zhp->zpool_hdl;
2650
2651 (void) snprintf(msg, sizeof (msg),
2652 dgettext(TEXT_DOMAIN, "cannot fault %llu"), guid);
2653
2654 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2655 zc.zc_guid = guid;
2656 zc.zc_cookie = VDEV_STATE_FAULTED;
2657 zc.zc_obj = aux;
2658
2659 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2660 return (0);
2661
2662 switch (errno) {
2663 case EBUSY:
2664
2665 /*
2666 * There are no other replicas of this device.
2667 */
2668 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2669
2670 default:
2671 return (zpool_standard_error(hdl, errno, msg));
2672 }
2673
2674 }
2675
2676 /*
2677 * Mark the given vdev degraded.
2678 */
2679 int
2680 zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
2681 {
2682 zfs_cmd_t zc = { 0 };
2683 char msg[1024];
2684 libzfs_handle_t *hdl = zhp->zpool_hdl;
2685
2686 (void) snprintf(msg, sizeof (msg),
2687 dgettext(TEXT_DOMAIN, "cannot degrade %llu"), guid);
2688
2689 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2690 zc.zc_guid = guid;
2691 zc.zc_cookie = VDEV_STATE_DEGRADED;
2692 zc.zc_obj = aux;
2693
2694 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
2695 return (0);
2696
2697 return (zpool_standard_error(hdl, errno, msg));
2698 }
2699
2700 /*
2701 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as
2702 * a hot spare.
2703 */
2704 static boolean_t
2705 is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which)
2706 {
2707 nvlist_t **child;
2708 uint_t c, children;
2709 char *type;
2710
2711 if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child,
2712 &children) == 0) {
2713 verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE,
2714 &type) == 0);
2715
2716 if (strcmp(type, VDEV_TYPE_SPARE) == 0 &&
2717 children == 2 && child[which] == tgt)
2718 return (B_TRUE);
2719
2720 for (c = 0; c < children; c++)
2721 if (is_replacing_spare(child[c], tgt, which))
2722 return (B_TRUE);
2723 }
2724
2725 return (B_FALSE);
2726 }
2727
2728 /*
2729 * Attach new_disk (fully described by nvroot) to old_disk.
2730 * If 'replacing' is specified, the new disk will replace the old one.
2731 */
2732 int
2733 zpool_vdev_attach(zpool_handle_t *zhp,
2734 const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing)
2735 {
2736 zfs_cmd_t zc = { 0 };
2737 char msg[1024];
2738 int ret;
2739 nvlist_t *tgt;
2740 boolean_t avail_spare, l2cache, islog;
2741 uint64_t val;
2742 char *newname;
2743 nvlist_t **child;
2744 uint_t children;
2745 nvlist_t *config_root;
2746 libzfs_handle_t *hdl = zhp->zpool_hdl;
2747 boolean_t rootpool = zpool_is_bootable(zhp);
2748
2749 if (replacing)
2750 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2751 "cannot replace %s with %s"), old_disk, new_disk);
2752 else
2753 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2754 "cannot attach %s to %s"), new_disk, old_disk);
2755
2756 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2757 if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache,
2758 &islog, NULL)) == 0)
2759 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2760
2761 if (avail_spare)
2762 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2763
2764 if (l2cache)
2765 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2766
2767 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2768 zc.zc_cookie = replacing;
2769
2770 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
2771 &child, &children) != 0 || children != 1) {
2772 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2773 "new device must be a single disk"));
2774 return (zfs_error(hdl, EZFS_INVALCONFIG, msg));
2775 }
2776
2777 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
2778 ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0);
2779
2780 if ((newname = zpool_vdev_name(NULL, NULL, child[0], B_FALSE)) == NULL)
2781 return (-1);
2782
2783 /*
2784 * If the target is a hot spare that has been swapped in, we can only
2785 * replace it with another hot spare.
2786 */
2787 if (replacing &&
2788 nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 &&
2789 (zpool_find_vdev(zhp, newname, &avail_spare, &l2cache,
2790 NULL, NULL) == NULL || !avail_spare) &&
2791 is_replacing_spare(config_root, tgt, 1)) {
2792 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2793 "can only be replaced by another hot spare"));
2794 free(newname);
2795 return (zfs_error(hdl, EZFS_BADTARGET, msg));
2796 }
2797
2798 free(newname);
2799
2800 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
2801 return (-1);
2802
2803 ret = zfs_ioctl(hdl, ZFS_IOC_VDEV_ATTACH, &zc);
2804
2805 zcmd_free_nvlists(&zc);
2806
2807 if (ret == 0) {
2808 if (rootpool) {
2809 /*
2810 * XXX need a better way to prevent user from
2811 * booting up a half-baked vdev.
2812 */
2813 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make "
2814 "sure to wait until resilver is done "
2815 "before rebooting.\n"));
2816 }
2817 return (0);
2818 }
2819
2820 switch (errno) {
2821 case ENOTSUP:
2822 /*
2823 * Can't attach to or replace this type of vdev.
2824 */
2825 if (replacing) {
2826 uint64_t version = zpool_get_prop_int(zhp,
2827 ZPOOL_PROP_VERSION, NULL);
2828
2829 if (islog)
2830 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2831 "cannot replace a log with a spare"));
2832 else if (version >= SPA_VERSION_MULTI_REPLACE)
2833 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2834 "already in replacing/spare config; wait "
2835 "for completion or use 'zpool detach'"));
2836 else
2837 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2838 "cannot replace a replacing device"));
2839 } else {
2840 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2841 "can only attach to mirrors and top-level "
2842 "disks"));
2843 }
2844 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
2845 break;
2846
2847 case EINVAL:
2848 /*
2849 * The new device must be a single disk.
2850 */
2851 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2852 "new device must be a single disk"));
2853 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg);
2854 break;
2855
2856 case EBUSY:
2857 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy"),
2858 new_disk);
2859 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2860 break;
2861
2862 case EOVERFLOW:
2863 /*
2864 * The new device is too small.
2865 */
2866 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2867 "device is too small"));
2868 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2869 break;
2870
2871 case EDOM:
2872 /*
2873 * The new device has a different alignment requirement.
2874 */
2875 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2876 "devices have different sector alignment"));
2877 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2878 break;
2879
2880 case ENAMETOOLONG:
2881 /*
2882 * The resulting top-level vdev spec won't fit in the label.
2883 */
2884 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg);
2885 break;
2886
2887 default:
2888 (void) zpool_standard_error(hdl, errno, msg);
2889 }
2890
2891 return (-1);
2892 }
2893
2894 /*
2895 * Detach the specified device.
2896 */
2897 int
2898 zpool_vdev_detach(zpool_handle_t *zhp, const char *path)
2899 {
2900 zfs_cmd_t zc = { 0 };
2901 char msg[1024];
2902 nvlist_t *tgt;
2903 boolean_t avail_spare, l2cache;
2904 libzfs_handle_t *hdl = zhp->zpool_hdl;
2905
2906 (void) snprintf(msg, sizeof (msg),
2907 dgettext(TEXT_DOMAIN, "cannot detach %s"), path);
2908
2909 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2910 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2911 NULL, NULL)) == 0)
2912 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2913
2914 if (avail_spare)
2915 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2916
2917 if (l2cache)
2918 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2919
2920 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2921
2922 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0)
2923 return (0);
2924
2925 switch (errno) {
2926
2927 case ENOTSUP:
2928 /*
2929 * Can't detach from this type of vdev.
2930 */
2931 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only "
2932 "applicable to mirror and replacing vdevs"));
2933 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
2934 break;
2935
2936 case EBUSY:
2937 /*
2938 * There are no other replicas of this device.
2939 */
2940 (void) zfs_error(hdl, EZFS_NOREPLICAS, msg);
2941 break;
2942
2943 default:
2944 (void) zpool_standard_error(hdl, errno, msg);
2945 }
2946
2947 return (-1);
2948 }
2949
2950 /*
2951 * Find a mirror vdev in the source nvlist.
2952 *
2953 * The mchild array contains a list of disks in one of the top-level mirrors
2954 * of the source pool. The schild array contains a list of disks that the
2955 * user specified on the command line. We loop over the mchild array to
2956 * see if any entry in the schild array matches.
2957 *
2958 * If a disk in the mchild array is found in the schild array, we return
2959 * the index of that entry. Otherwise we return -1.
2960 */
2961 static int
2962 find_vdev_entry(zpool_handle_t *zhp, nvlist_t **mchild, uint_t mchildren,
2963 nvlist_t **schild, uint_t schildren)
2964 {
2965 uint_t mc;
2966
2967 for (mc = 0; mc < mchildren; mc++) {
2968 uint_t sc;
2969 char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp,
2970 mchild[mc], B_FALSE);
2971
2972 for (sc = 0; sc < schildren; sc++) {
2973 char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp,
2974 schild[sc], B_FALSE);
2975 boolean_t result = (strcmp(mpath, spath) == 0);
2976
2977 free(spath);
2978 if (result) {
2979 free(mpath);
2980 return (mc);
2981 }
2982 }
2983
2984 free(mpath);
2985 }
2986
2987 return (-1);
2988 }
2989
2990 /*
2991 * Split a mirror pool. If newroot points to null, then a new nvlist
2992 * is generated and it is the responsibility of the caller to free it.
2993 */
2994 int
2995 zpool_vdev_split(zpool_handle_t *zhp, char *newname, nvlist_t **newroot,
2996 nvlist_t *props, splitflags_t flags)
2997 {
2998 zfs_cmd_t zc = { 0 };
2999 char msg[1024];
3000 nvlist_t *tree, *config, **child, **newchild, *newconfig = NULL;
3001 nvlist_t **varray = NULL, *zc_props = NULL;
3002 uint_t c, children, newchildren, lastlog = 0, vcount, found = 0;
3003 libzfs_handle_t *hdl = zhp->zpool_hdl;
3004 uint64_t vers;
3005 boolean_t freelist = B_FALSE, memory_err = B_TRUE;
3006 int retval = 0;
3007
3008 (void) snprintf(msg, sizeof (msg),
3009 dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name);
3010
3011 if (!zpool_name_valid(hdl, B_FALSE, newname))
3012 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
3013
3014 if ((config = zpool_get_config(zhp, NULL)) == NULL) {
3015 (void) fprintf(stderr, gettext("Internal error: unable to "
3016 "retrieve pool configuration\n"));
3017 return (-1);
3018 }
3019
3020 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree)
3021 == 0);
3022 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &vers) == 0);
3023
3024 if (props) {
3025 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
3026 if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name,
3027 props, vers, flags, msg)) == NULL)
3028 return (-1);
3029 }
3030
3031 if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child,
3032 &children) != 0) {
3033 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3034 "Source pool is missing vdev tree"));
3035 nvlist_free(zc_props);
3036 return (-1);
3037 }
3038
3039 varray = zfs_alloc(hdl, children * sizeof (nvlist_t *));
3040 vcount = 0;
3041
3042 if (*newroot == NULL ||
3043 nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN,
3044 &newchild, &newchildren) != 0)
3045 newchildren = 0;
3046
3047 for (c = 0; c < children; c++) {
3048 uint64_t is_log = B_FALSE, is_hole = B_FALSE;
3049 char *type;
3050 nvlist_t **mchild, *vdev;
3051 uint_t mchildren;
3052 int entry;
3053
3054 /*
3055 * Unlike cache & spares, slogs are stored in the
3056 * ZPOOL_CONFIG_CHILDREN array. We filter them out here.
3057 */
3058 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
3059 &is_log);
3060 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
3061 &is_hole);
3062 if (is_log || is_hole) {
3063 /*
3064 * Create a hole vdev and put it in the config.
3065 */
3066 if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0)
3067 goto out;
3068 if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE,
3069 VDEV_TYPE_HOLE) != 0)
3070 goto out;
3071 if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE,
3072 1) != 0)
3073 goto out;
3074 if (lastlog == 0)
3075 lastlog = vcount;
3076 varray[vcount++] = vdev;
3077 continue;
3078 }
3079 lastlog = 0;
3080 verify(nvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE, &type)
3081 == 0);
3082 if (strcmp(type, VDEV_TYPE_MIRROR) != 0) {
3083 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3084 "Source pool must be composed only of mirrors\n"));
3085 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
3086 goto out;
3087 }
3088
3089 verify(nvlist_lookup_nvlist_array(child[c],
3090 ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0);
3091
3092 /* find or add an entry for this top-level vdev */
3093 if (newchildren > 0 &&
3094 (entry = find_vdev_entry(zhp, mchild, mchildren,
3095 newchild, newchildren)) >= 0) {
3096 /* We found a disk that the user specified. */
3097 vdev = mchild[entry];
3098 ++found;
3099 } else {
3100 /* User didn't specify a disk for this vdev. */
3101 vdev = mchild[mchildren - 1];
3102 }
3103
3104 if (nvlist_dup(vdev, &varray[vcount++], 0) != 0)
3105 goto out;
3106 }
3107
3108 /* did we find every disk the user specified? */
3109 if (found != newchildren) {
3110 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must "
3111 "include at most one disk from each mirror"));
3112 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
3113 goto out;
3114 }
3115
3116 /* Prepare the nvlist for populating. */
3117 if (*newroot == NULL) {
3118 if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0)
3119 goto out;
3120 freelist = B_TRUE;
3121 if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE,
3122 VDEV_TYPE_ROOT) != 0)
3123 goto out;
3124 } else {
3125 verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0);
3126 }
3127
3128 /* Add all the children we found */
3129 if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, varray,
3130 lastlog == 0 ? vcount : lastlog) != 0)
3131 goto out;
3132
3133 /*
3134 * If we're just doing a dry run, exit now with success.
3135 */
3136 if (flags.dryrun) {
3137 memory_err = B_FALSE;
3138 freelist = B_FALSE;
3139 goto out;
3140 }
3141
3142 /* now build up the config list & call the ioctl */
3143 if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0)
3144 goto out;
3145
3146 if (nvlist_add_nvlist(newconfig,
3147 ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 ||
3148 nvlist_add_string(newconfig,
3149 ZPOOL_CONFIG_POOL_NAME, newname) != 0 ||
3150 nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0)
3151 goto out;
3152
3153 /*
3154 * The new pool is automatically part of the namespace unless we
3155 * explicitly export it.
3156 */
3157 if (!flags.import)
3158 zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT;
3159 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3160 (void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string));
3161 if (zcmd_write_conf_nvlist(hdl, &zc, newconfig) != 0)
3162 goto out;
3163 if (zc_props != NULL && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
3164 goto out;
3165
3166 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) {
3167 retval = zpool_standard_error(hdl, errno, msg);
3168 goto out;
3169 }
3170
3171 freelist = B_FALSE;
3172 memory_err = B_FALSE;
3173
3174 out:
3175 if (varray != NULL) {
3176 int v;
3177
3178 for (v = 0; v < vcount; v++)
3179 nvlist_free(varray[v]);
3180 free(varray);
3181 }
3182 zcmd_free_nvlists(&zc);
3183 nvlist_free(zc_props);
3184 nvlist_free(newconfig);
3185 if (freelist) {
3186 nvlist_free(*newroot);
3187 *newroot = NULL;
3188 }
3189
3190 if (retval != 0)
3191 return (retval);
3192
3193 if (memory_err)
3194 return (no_memory(hdl));
3195
3196 return (0);
3197 }
3198
3199 /*
3200 * Remove the given device. Currently, this is supported only for hot spares
3201 * and level 2 cache devices.
3202 */
3203 int
3204 zpool_vdev_remove(zpool_handle_t *zhp, const char *path)
3205 {
3206 zfs_cmd_t zc = { 0 };
3207 char msg[1024];
3208 nvlist_t *tgt;
3209 boolean_t avail_spare, l2cache, islog, isspecial;
3210 libzfs_handle_t *hdl = zhp->zpool_hdl;
3211 uint64_t version;
3212
3213 (void) snprintf(msg, sizeof (msg),
3214 dgettext(TEXT_DOMAIN, "cannot remove %s"), path);
3215
3216 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3217 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
3218 &islog, &isspecial)) == 0)
3219 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3220 /*
3221 * XXX - this should just go away.
3222 */
3223 if (!avail_spare && !l2cache && !islog && !isspecial) {
3224 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3225 "only inactive hot spares, cache, top-level, "
3226 "log, or special devices can be removed"));
3227 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3228 }
3229
3230 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
3231 if (islog && version < SPA_VERSION_HOLES) {
3232 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3233 "pool must be upgrade to support log removal"));
3234 return (zfs_error(hdl, EZFS_BADVERSION, msg));
3235 }
3236
3237 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
3238
3239 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
3240 return (0);
3241
3242 if (isspecial && errno == EEXIST) {
3243 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3244 "special device contains metadata"));
3245 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg));
3246 } else if (isspecial && errno == EBUSY) {
3247 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3248 "wbc feature flag is active"));
3249 return (zfs_error(hdl, EZFS_WBCCHILD, msg));
3250 }
3251
3252 return (zpool_standard_error(hdl, errno, msg));
3253 }
3254
3255 /*
3256 * Clear the errors for the pool, or the particular device if specified.
3257 */
3258 int
3259 zpool_clear(zpool_handle_t *zhp, const char *path, nvlist_t *rewindnvl)
3260 {
3261 zfs_cmd_t zc = { 0 };
3262 char msg[1024];
3263 nvlist_t *tgt;
3264 zpool_rewind_policy_t policy;
3265 boolean_t avail_spare, l2cache;
3266 libzfs_handle_t *hdl = zhp->zpool_hdl;
3267 nvlist_t *nvi = NULL;
3268 int error;
3269
3270 if (path)
3271 (void) snprintf(msg, sizeof (msg),
3272 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
3273 path);
3274 else
3275 (void) snprintf(msg, sizeof (msg),
3276 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
3277 zhp->zpool_name);
3278
3279 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3280 if (path) {
3281 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare,
3282 &l2cache, NULL, NULL)) == 0)
3283 return (zfs_error(hdl, EZFS_NODEVICE, msg));
3284
3285 /*
3286 * Don't allow error clearing for hot spares. Do allow
3287 * error clearing for l2cache devices.
3288 */
3289 if (avail_spare)
3290 return (zfs_error(hdl, EZFS_ISSPARE, msg));
3291
3292 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID,
3293 &zc.zc_guid) == 0);
3294 }
3295
3296 zpool_get_rewind_policy(rewindnvl, &policy);
3297 zc.zc_cookie = policy.zrp_request;
3298
3299 if (zcmd_alloc_dst_nvlist(hdl, &zc, zhp->zpool_config_size * 2) != 0)
3300 return (-1);
3301
3302 if (zcmd_write_src_nvlist(hdl, &zc, rewindnvl) != 0)
3303 return (-1);
3304
3305 while ((error = zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc)) != 0 &&
3306 errno == ENOMEM) {
3307 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
3308 zcmd_free_nvlists(&zc);
3309 return (-1);
3310 }
3311 }
3312
3313 if (!error || ((policy.zrp_request & ZPOOL_TRY_REWIND) &&
3314 errno != EPERM && errno != EACCES)) {
3315 if (policy.zrp_request &
3316 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
3317 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi);
3318 zpool_rewind_exclaim(hdl, zc.zc_name,
3319 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0),
3320 nvi);
3321 nvlist_free(nvi);
3322 }
3323 zcmd_free_nvlists(&zc);
3324 return (0);
3325 }
3326
3327 zcmd_free_nvlists(&zc);
3328 return (zpool_standard_error(hdl, errno, msg));
3329 }
3330
3331 /*
3332 * Similar to zpool_clear(), but takes a GUID (used by fmd).
3333 */
3334 int
3335 zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid)
3336 {
3337 zfs_cmd_t zc = { 0 };
3338 char msg[1024];
3339 libzfs_handle_t *hdl = zhp->zpool_hdl;
3340
3341 (void) snprintf(msg, sizeof (msg),
3342 dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"),
3343 guid);
3344
3345 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3346 zc.zc_guid = guid;
3347 zc.zc_cookie = ZPOOL_NO_REWIND;
3348
3349 if (ioctl(hdl->libzfs_fd, ZFS_IOC_CLEAR, &zc) == 0)
3350 return (0);
3351
3352 return (zpool_standard_error(hdl, errno, msg));
3353 }
3354
3355 /*
3356 * Change the GUID for a pool.
3357 */
3358 int
3359 zpool_reguid(zpool_handle_t *zhp)
3360 {
3361 char msg[1024];
3362 libzfs_handle_t *hdl = zhp->zpool_hdl;
3363 zfs_cmd_t zc = { 0 };
3364
3365 (void) snprintf(msg, sizeof (msg),
3366 dgettext(TEXT_DOMAIN, "cannot reguid '%s'"), zhp->zpool_name);
3367
3368 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3369 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REGUID, &zc) == 0)
3370 return (0);
3371
3372 return (zpool_standard_error(hdl, errno, msg));
3373 }
3374
3375 /*
3376 * Reopen the pool.
3377 */
3378 int
3379 zpool_reopen(zpool_handle_t *zhp)
3380 {
3381 zfs_cmd_t zc = { 0 };
3382 char msg[1024];
3383 libzfs_handle_t *hdl = zhp->zpool_hdl;
3384
3385 (void) snprintf(msg, sizeof (msg),
3386 dgettext(TEXT_DOMAIN, "cannot reopen '%s'"),
3387 zhp->zpool_name);
3388
3389 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3390 if (zfs_ioctl(hdl, ZFS_IOC_POOL_REOPEN, &zc) == 0)
3391 return (0);
3392 return (zpool_standard_error(hdl, errno, msg));
3393 }
3394
3395 /*
3396 * Convert from a devid string to a path.
3397 */
3398 static char *
3399 devid_to_path(char *devid_str)
3400 {
3401 ddi_devid_t devid;
3402 char *minor;
3403 char *path;
3404 devid_nmlist_t *list = NULL;
3405 int ret;
3406
3407 if (devid_str_decode(devid_str, &devid, &minor) != 0)
3408 return (NULL);
3409
3410 ret = devid_deviceid_to_nmlist("/dev", devid, minor, &list);
3411
3412 devid_str_free(minor);
3413 devid_free(devid);
3414
3415 if (ret != 0)
3416 return (NULL);
3417
3418 /*
3419 * In a case the strdup() fails, we will just return NULL below.
3420 */
3421 path = strdup(list[0].devname);
3422
3423 devid_free_nmlist(list);
3424
3425 return (path);
3426 }
3427
3428 /*
3429 * Convert from a path to a devid string.
3430 */
3431 static char *
3432 path_to_devid(const char *path)
3433 {
3434 int fd;
3435 ddi_devid_t devid;
3436 char *minor, *ret;
3437
3438 if ((fd = open(path, O_RDONLY | O_NDELAY)) < 0)
3439 return (NULL);
3440
3441 minor = NULL;
3442 ret = NULL;
3443 if (devid_get(fd, &devid) == 0) {
3444 if (devid_get_minor_name(fd, &minor) == 0)
3445 ret = devid_str_encode(devid, minor);
3446 if (minor != NULL)
3447 devid_str_free(minor);
3448 devid_free(devid);
3449 }
3450 (void) close(fd);
3451
3452 return (ret);
3453 }
3454
3455 /*
3456 * Issue the necessary ioctl() to update the stored path value for the vdev. We
3457 * ignore any failure here, since a common case is for an unprivileged user to
3458 * type 'zpool status', and we'll display the correct information anyway.
3459 */
3460 static void
3461 set_path(zpool_handle_t *zhp, nvlist_t *nv, const char *path)
3462 {
3463 zfs_cmd_t zc = { 0 };
3464
3465 (void) strncpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3466 (void) strncpy(zc.zc_value, path, sizeof (zc.zc_value));
3467 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
3468 &zc.zc_guid) == 0);
3469
3470 (void) ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SETPATH, &zc);
3471 }
3472
3473 /*
3474 * Given a vdev, return the name to display in iostat. If the vdev has a path,
3475 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
3476 * We also check if this is a whole disk, in which case we strip off the
3477 * trailing 's0' slice name.
3478 *
3479 * This routine is also responsible for identifying when disks have been
3480 * reconfigured in a new location. The kernel will have opened the device by
3481 * devid, but the path will still refer to the old location. To catch this, we
3482 * first do a path -> devid translation (which is fast for the common case). If
3483 * the devid matches, we're done. If not, we do a reverse devid -> path
3484 * translation and issue the appropriate ioctl() to update the path of the vdev.
3485 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
3486 * of these checks.
3487 */
3488 char *
3489 zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv,
3490 boolean_t verbose)
3491 {
3492 char *path, *devid;
3493 uint64_t value;
3494 char buf[64];
3495 vdev_stat_t *vs;
3496 uint_t vsc;
3497
3498 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
3499 &value) == 0) {
3500 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
3501 &value) == 0);
3502 (void) snprintf(buf, sizeof (buf), "%llu",
3503 (u_longlong_t)value);
3504 path = buf;
3505 } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) {
3506
3507 /*
3508 * If the device is dead (faulted, offline, etc) then don't
3509 * bother opening it. Otherwise we may be forcing the user to
3510 * open a misbehaving device, which can have undesirable
3511 * effects.
3512 */
3513 if ((nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
3514 (uint64_t **)&vs, &vsc) != 0 ||
3515 vs->vs_state >= VDEV_STATE_DEGRADED) &&
3516 zhp != NULL &&
3517 nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) {
3518 /*
3519 * Determine if the current path is correct.
3520 */
3521 char *newdevid = path_to_devid(path);
3522
3523 if (newdevid == NULL ||
3524 strcmp(devid, newdevid) != 0) {
3525 char *newpath;
3526
3527 if ((newpath = devid_to_path(devid)) != NULL) {
3528 /*
3529 * Update the path appropriately.
3530 */
3531 set_path(zhp, nv, newpath);
3532 if (nvlist_add_string(nv,
3533 ZPOOL_CONFIG_PATH, newpath) == 0)
3534 verify(nvlist_lookup_string(nv,
3535 ZPOOL_CONFIG_PATH,
3536 &path) == 0);
3537 free(newpath);
3538 }
3539 }
3540
3541 if (newdevid)
3542 devid_str_free(newdevid);
3543 }
3544
3545 if (strncmp(path, ZFS_DISK_ROOTD, strlen(ZFS_DISK_ROOTD)) == 0)
3546 path += strlen(ZFS_DISK_ROOTD);
3547
3548 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
3549 &value) == 0 && value) {
3550 int pathlen = strlen(path);
3551 char *tmp = zfs_strdup(hdl, path);
3552
3553 /*
3554 * If it starts with c#, and ends with "s0" or "s1",
3555 * chop the slice off, or if it ends with "s0/old" or
3556 * "s1/old", remove the slice from the middle.
3557 */
3558 if (CTD_CHECK(tmp)) {
3559 if (strcmp(&tmp[pathlen - 2], "s0") == 0 ||
3560 strcmp(&tmp[pathlen - 2], "s1") == 0) {
3561 tmp[pathlen - 2] = '\0';
3562 } else if (pathlen > 6 &&
3563 (strcmp(&tmp[pathlen - 6], "s0/old") == 0 ||
3564 strcmp(&tmp[pathlen - 6], "s1/old") == 0)) {
3565 (void) strcpy(&tmp[pathlen - 6],
3566 "/old");
3567 }
3568 }
3569 return (tmp);
3570 }
3571 } else {
3572 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &path) == 0);
3573
3574 /*
3575 * If it's a raidz device, we need to stick in the parity level.
3576 */
3577 if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) {
3578 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY,
3579 &value) == 0);
3580 (void) snprintf(buf, sizeof (buf), "%s%llu", path,
3581 (u_longlong_t)value);
3582 path = buf;
3583 }
3584
3585 /*
3586 * We identify each top-level vdev by using a <type-id>
3587 * naming convention.
3588 */
3589 if (verbose) {
3590 uint64_t id;
3591
3592 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
3593 &id) == 0);
3594 (void) snprintf(buf, sizeof (buf), "%s-%llu", path,
3595 (u_longlong_t)id);
3596 path = buf;
3597 }
3598 }
3599
3600 return (zfs_strdup(hdl, path));
3601 }
3602
3603 static int
3604 zbookmark_mem_compare(const void *a, const void *b)
3605 {
3606 return (memcmp(a, b, sizeof (zbookmark_phys_t)));
3607 }
3608
3609 /*
3610 * Retrieve the persistent error log, uniquify the members, and return to the
3611 * caller.
3612 */
3613 int
3614 zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp)
3615 {
3616 zfs_cmd_t zc = { 0 };
3617 uint64_t count;
3618 zbookmark_phys_t *zb = NULL;
3619 int i;
3620
3621 /*
3622 * Retrieve the raw error list from the kernel. If the number of errors
3623 * has increased, allocate more space and continue until we get the
3624 * entire list.
3625 */
3626 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT,
3627 &count) == 0);
3628 if (count == 0)
3629 return (0);
3630 if ((zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl,
3631 count * sizeof (zbookmark_phys_t))) == (uintptr_t)NULL)
3632 return (-1);
3633 zc.zc_nvlist_dst_size = count;
3634 (void) strcpy(zc.zc_name, zhp->zpool_name);
3635 for (;;) {
3636 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_ERROR_LOG,
3637 &zc) != 0) {
3638 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3639 if (errno == ENOMEM) {
3640 void *dst;
3641
3642 count = zc.zc_nvlist_dst_size;
3643 dst = zfs_alloc(zhp->zpool_hdl, count *
3644 sizeof (zbookmark_phys_t));
3645 if (dst == NULL)
3646 return (-1);
3647 zc.zc_nvlist_dst = (uintptr_t)dst;
3648 } else {
3649 return (-1);
3650 }
3651 } else {
3652 break;
3653 }
3654 }
3655
3656 /*
3657 * Sort the resulting bookmarks. This is a little confusing due to the
3658 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last
3659 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks
3660 * _not_ copied as part of the process. So we point the start of our
3661 * array appropriate and decrement the total number of elements.
3662 */
3663 zb = ((zbookmark_phys_t *)(uintptr_t)zc.zc_nvlist_dst) +
3664 zc.zc_nvlist_dst_size;
3665 count -= zc.zc_nvlist_dst_size;
3666
3667 qsort(zb, count, sizeof (zbookmark_phys_t), zbookmark_mem_compare);
3668
3669 verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0);
3670
3671 /*
3672 * Fill in the nverrlistp with nvlist's of dataset and object numbers.
3673 */
3674 for (i = 0; i < count; i++) {
3675 nvlist_t *nv;
3676
3677 /* ignoring zb_blkid and zb_level for now */
3678 if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset &&
3679 zb[i-1].zb_object == zb[i].zb_object)
3680 continue;
3681
3682 if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0)
3683 goto nomem;
3684 if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET,
3685 zb[i].zb_objset) != 0) {
3686 nvlist_free(nv);
3687 goto nomem;
3688 }
3689 if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT,
3690 zb[i].zb_object) != 0) {
3691 nvlist_free(nv);
3692 goto nomem;
3693 }
3694 if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) {
3695 nvlist_free(nv);
3696 goto nomem;
3697 }
3698 nvlist_free(nv);
3699 }
3700
3701 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3702 return (0);
3703
3704 nomem:
3705 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3706 return (no_memory(zhp->zpool_hdl));
3707 }
3708
3709 /*
3710 * Upgrade a ZFS pool to the latest on-disk version.
3711 */
3712 int
3713 zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version)
3714 {
3715 zfs_cmd_t zc = { 0 };
3716 libzfs_handle_t *hdl = zhp->zpool_hdl;
3717
3718 (void) strcpy(zc.zc_name, zhp->zpool_name);
3719 zc.zc_cookie = new_version;
3720
3721 if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0)
3722 return (zpool_standard_error_fmt(hdl, errno,
3723 dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"),
3724 zhp->zpool_name));
3725 return (0);
3726 }
3727
3728 void
3729 zfs_save_arguments(int argc, char **argv, char *string, int len)
3730 {
3731 (void) strlcpy(string, basename(argv[0]), len);
3732 for (int i = 1; i < argc; i++) {
3733 (void) strlcat(string, " ", len);
3734 (void) strlcat(string, argv[i], len);
3735 }
3736 }
3737
3738 int
3739 zpool_log_history(libzfs_handle_t *hdl, const char *message)
3740 {
3741 zfs_cmd_t zc = { 0 };
3742 nvlist_t *args;
3743 int err;
3744
3745 args = fnvlist_alloc();
3746 fnvlist_add_string(args, "message", message);
3747 err = zcmd_write_src_nvlist(hdl, &zc, args);
3748 if (err == 0)
3749 err = ioctl(hdl->libzfs_fd, ZFS_IOC_LOG_HISTORY, &zc);
3750 nvlist_free(args);
3751 zcmd_free_nvlists(&zc);
3752 return (err);
3753 }
3754
3755 int
3756 zpool_stage_history(libzfs_handle_t *hdl, const char *history_str)
3757 {
3758 if (history_str == NULL)
3759 return (EINVAL);
3760
3761 if (hdl->libzfs_log_str != NULL)
3762 free(hdl->libzfs_log_str);
3763
3764 if ((hdl->libzfs_log_str = strdup(history_str)) == NULL)
3765 return (no_memory(hdl));
3766
3767 return (0);
3768 }
3769
3770 /*
3771 * Perform ioctl to get some command history of a pool.
3772 *
3773 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the
3774 * logical offset of the history buffer to start reading from.
3775 *
3776 * Upon return, 'off' is the next logical offset to read from and
3777 * 'len' is the actual amount of bytes read into 'buf'.
3778 */
3779 static int
3780 get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len)
3781 {
3782 zfs_cmd_t zc = { 0 };
3783 libzfs_handle_t *hdl = zhp->zpool_hdl;
3784
3785 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3786
3787 zc.zc_history = (uint64_t)(uintptr_t)buf;
3788 zc.zc_history_len = *len;
3789 zc.zc_history_offset = *off;
3790
3791 if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) {
3792 switch (errno) {
3793 case EPERM:
3794 return (zfs_error_fmt(hdl, EZFS_PERM,
3795 dgettext(TEXT_DOMAIN,
3796 "cannot show history for pool '%s'"),
3797 zhp->zpool_name));
3798 case ENOENT:
3799 return (zfs_error_fmt(hdl, EZFS_NOHISTORY,
3800 dgettext(TEXT_DOMAIN, "cannot get history for pool "
3801 "'%s'"), zhp->zpool_name));
3802 case ENOTSUP:
3803 return (zfs_error_fmt(hdl, EZFS_BADVERSION,
3804 dgettext(TEXT_DOMAIN, "cannot get history for pool "
3805 "'%s', pool must be upgraded"), zhp->zpool_name));
3806 default:
3807 return (zpool_standard_error_fmt(hdl, errno,
3808 dgettext(TEXT_DOMAIN,
3809 "cannot get history for '%s'"), zhp->zpool_name));
3810 }
3811 }
3812
3813 *len = zc.zc_history_len;
3814 *off = zc.zc_history_offset;
3815
3816 return (0);
3817 }
3818
3819 /*
3820 * Process the buffer of nvlists, unpacking and storing each nvlist record
3821 * into 'records'. 'leftover' is set to the number of bytes that weren't
3822 * processed as there wasn't a complete record.
3823 */
3824 int
3825 zpool_history_unpack(char *buf, uint64_t bytes_read, uint64_t *leftover,
3826 nvlist_t ***records, uint_t *numrecords)
3827 {
3828 uint64_t reclen;
3829 nvlist_t *nv;
3830 int i;
3831
3832 while (bytes_read > sizeof (reclen)) {
3833
3834 /* get length of packed record (stored as little endian) */
3835 for (i = 0, reclen = 0; i < sizeof (reclen); i++)
3836 reclen += (uint64_t)(((uchar_t *)buf)[i]) << (8*i);
3837
3838 if (bytes_read < sizeof (reclen) + reclen)
3839 break;
3840
3841 /* unpack record */
3842 if (nvlist_unpack(buf + sizeof (reclen), reclen, &nv, 0) != 0)
3843 return (ENOMEM);
3844 bytes_read -= sizeof (reclen) + reclen;
3845 buf += sizeof (reclen) + reclen;
3846
3847 /* add record to nvlist array */
3848 (*numrecords)++;
3849 if (ISP2(*numrecords + 1)) {
3850 *records = realloc(*records,
3851 *numrecords * 2 * sizeof (nvlist_t *));
3852 }
3853 (*records)[*numrecords - 1] = nv;
3854 }
3855
3856 *leftover = bytes_read;
3857 return (0);
3858 }
3859
3860 /*
3861 * Retrieve the command history of a pool.
3862 */
3863 int
3864 zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp)
3865 {
3866 char *buf;
3867 int buflen = 128 * 1024;
3868 uint64_t off = 0;
3869 nvlist_t **records = NULL;
3870 uint_t numrecords = 0;
3871 int err, i;
3872
3873 buf = malloc(buflen);
3874 if (buf == NULL)
3875 return (ENOMEM);
3876 do {
3877 uint64_t bytes_read = buflen;
3878 uint64_t leftover;
3879
3880 if ((err = get_history(zhp, buf, &off, &bytes_read)) != 0)
3881 break;
3882
3883 /* if nothing else was read in, we're at EOF, just return */
3884 if (!bytes_read)
3885 break;
3886
3887 if ((err = zpool_history_unpack(buf, bytes_read,
3888 &leftover, &records, &numrecords)) != 0)
3889 break;
3890 off -= leftover;
3891 if (leftover == bytes_read) {
3892 /*
3893 * no progress made, because buffer is not big enough
3894 * to hold this record; resize and retry.
3895 */
3896 buflen *= 2;
3897 free(buf);
3898 buf = malloc(buflen);
3899 if (buf == NULL)
3900 return (ENOMEM);
3901 }
3902
3903 /* CONSTCOND */
3904 } while (1);
3905
3906 free(buf);
3907
3908 if (!err) {
3909 verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0);
3910 verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD,
3911 records, numrecords) == 0);
3912 }
3913 for (i = 0; i < numrecords; i++)
3914 nvlist_free(records[i]);
3915 free(records);
3916
3917 return (err);
3918 }
3919
3920 void
3921 zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
3922 char *pathname, size_t len)
3923 {
3924 zfs_cmd_t zc = { 0 };
3925 boolean_t mounted = B_FALSE;
3926 char *mntpnt = NULL;
3927 char dsname[ZFS_MAX_DATASET_NAME_LEN];
3928
3929 if (dsobj == 0) {
3930 /* special case for the MOS */
3931 (void) snprintf(pathname, len, "<metadata>:<0x%llx>", obj);
3932 return;
3933 }
3934
3935 /* get the dataset's name */
3936 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3937 zc.zc_obj = dsobj;
3938 if (ioctl(zhp->zpool_hdl->libzfs_fd,
3939 ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) {
3940 /* just write out a path of two object numbers */
3941 (void) snprintf(pathname, len, "<0x%llx>:<0x%llx>",
3942 dsobj, obj);
3943 return;
3944 }
3945 (void) strlcpy(dsname, zc.zc_value, sizeof (dsname));
3946
3947 /* find out if the dataset is mounted */
3948 mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt);
3949
3950 /* get the corrupted object's path */
3951 (void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name));
3952 zc.zc_obj = obj;
3953 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_OBJ_TO_PATH,
3954 &zc) == 0) {
3955 if (mounted) {
3956 (void) snprintf(pathname, len, "%s%s", mntpnt,
3957 zc.zc_value);
3958 } else {
3959 (void) snprintf(pathname, len, "%s:%s",
3960 dsname, zc.zc_value);
3961 }
3962 } else {
3963 (void) snprintf(pathname, len, "%s:<0x%llx>", dsname, obj);
3964 }
3965 free(mntpnt);
3966 }
3967
3968 /*
3969 * Read the EFI label from the config, if a label does not exist then
3970 * pass back the error to the caller. If the caller has passed a non-NULL
3971 * diskaddr argument then we set it to the starting address of the EFI
3972 * partition. If the caller has passed a non-NULL boolean argument, then
3973 * we set it to indicate if the disk does have efi system partition.
3974 */
3975 static int
3976 read_efi_label(nvlist_t *config, diskaddr_t *sb, boolean_t *system)
3977 {
3978 char *path;
3979 int fd;
3980 char diskname[MAXPATHLEN];
3981 boolean_t boot = B_FALSE;
3982 int err = -1;
3983 int slice;
3984
3985 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PATH, &path) != 0)
3986 return (err);
3987
3988 (void) snprintf(diskname, sizeof (diskname), "%s%s", ZFS_RDISK_ROOT,
3989 strrchr(path, '/'));
3990 if ((fd = open(diskname, O_RDONLY|O_NDELAY)) >= 0) {
3991 struct dk_gpt *vtoc;
3992
3993 if ((err = efi_alloc_and_read(fd, &vtoc)) >= 0) {
3994 for (slice = 0; slice < vtoc->efi_nparts; slice++) {
3995 if (vtoc->efi_parts[slice].p_tag == V_SYSTEM)
3996 boot = B_TRUE;
3997 if (vtoc->efi_parts[slice].p_tag == V_USR)
3998 break;
3999 }
4000 if (sb != NULL && vtoc->efi_parts[slice].p_tag == V_USR)
4001 *sb = vtoc->efi_parts[slice].p_start;
4002 if (system != NULL)
4003 *system = boot;
4004 efi_free(vtoc);
4005 }
4006 (void) close(fd);
4007 }
4008 return (err);
4009 }
4010
4011 /*
4012 * determine where a partition starts on a disk in the current
4013 * configuration
4014 */
4015 static diskaddr_t
4016 find_start_block(nvlist_t *config)
4017 {
4018 nvlist_t **child;
4019 uint_t c, children;
4020 diskaddr_t sb = MAXOFFSET_T;
4021 uint64_t wholedisk;
4022
4023 if (nvlist_lookup_nvlist_array(config,
4024 ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) {
4025 if (nvlist_lookup_uint64(config,
4026 ZPOOL_CONFIG_WHOLE_DISK,
4027 &wholedisk) != 0 || !wholedisk) {
4028 return (MAXOFFSET_T);
4029 }
4030 if (read_efi_label(config, &sb, NULL) < 0)
4031 sb = MAXOFFSET_T;
4032 return (sb);
4033 }
4034
4035 for (c = 0; c < children; c++) {
4036 sb = find_start_block(child[c]);
4037 if (sb != MAXOFFSET_T) {
4038 return (sb);
4039 }
4040 }
4041 return (MAXOFFSET_T);
4042 }
4043
4044 /*
4045 * Label an individual disk. The name provided is the short name,
4046 * stripped of any leading /dev path.
4047 */
4048 int
4049 zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, const char *name,
4050 zpool_boot_label_t boot_type, uint64_t boot_size, int *slice)
4051 {
4052 char path[MAXPATHLEN];
4053 struct dk_gpt *vtoc;
4054 int fd;
4055 size_t resv = EFI_MIN_RESV_SIZE;
4056 uint64_t slice_size;
4057 diskaddr_t start_block;
4058 char errbuf[1024];
4059
4060 /* prepare an error message just in case */
4061 (void) snprintf(errbuf, sizeof (errbuf),
4062 dgettext(TEXT_DOMAIN, "cannot label '%s'"), name);
4063
4064 if (zhp) {
4065 nvlist_t *nvroot;
4066
4067 verify(nvlist_lookup_nvlist(zhp->zpool_config,
4068 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
4069
4070 if (zhp->zpool_start_block == 0)
4071 start_block = find_start_block(nvroot);
4072 else
4073 start_block = zhp->zpool_start_block;
4074 zhp->zpool_start_block = start_block;
4075 } else {
4076 /* new pool */
4077 start_block = NEW_START_BLOCK;
4078 }
4079
4080 (void) snprintf(path, sizeof (path), "%s/%s%s", ZFS_RDISK_ROOT, name,
4081 BACKUP_SLICE);
4082
4083 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) {
4084 /*
4085 * This shouldn't happen. We've long since verified that this
4086 * is a valid device.
4087 */
4088 zfs_error_aux(hdl,
4089 dgettext(TEXT_DOMAIN, "unable to open device"));
4090 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
4091 }
4092
4093 if (efi_alloc_and_init(fd, EFI_NUMPAR, &vtoc) != 0) {
4094 /*
4095 * The only way this can fail is if we run out of memory, or we
4096 * were unable to read the disk's capacity
4097 */
4098 if (errno == ENOMEM)
4099 (void) no_memory(hdl);
4100
4101 (void) close(fd);
4102 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4103 "unable to read disk capacity"), name);
4104
4105 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
4106 }
4107
4108 /*
4109 * Why we use V_USR: V_BACKUP confuses users, and is considered
4110 * disposable by some EFI utilities (since EFI doesn't have a backup
4111 * slice). V_UNASSIGNED is supposed to be used only for zero size
4112 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT,
4113 * etc. were all pretty specific. V_USR is as close to reality as we
4114 * can get, in the absence of V_OTHER.
4115 */
4116 /* first fix the partition start block */
4117 if (start_block == MAXOFFSET_T)
4118 start_block = NEW_START_BLOCK;
4119
4120 /*
4121 * EFI System partition is using slice 0.
4122 * ZFS is on slice 1 and slice 8 is reserved.
4123 * We assume the GPT partition table without system
4124 * partition has zfs p_start == NEW_START_BLOCK.
4125 * If start_block != NEW_START_BLOCK, it means we have
4126 * system partition. Correct solution would be to query/cache vtoc
4127 * from existing vdev member.
4128 */
4129 if (boot_type == ZPOOL_CREATE_BOOT_LABEL) {
4130 if (boot_size % vtoc->efi_lbasize != 0) {
4131 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4132 "boot partition size must be a multiple of %d"),
4133 vtoc->efi_lbasize);
4134 (void) close(fd);
4135 efi_free(vtoc);
4136 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
4137 }
4138 /*
4139 * System partition size checks.
4140 * Note the 1MB is quite arbitrary value, since we
4141 * are creating dedicated pool, it should be enough
4142 * to hold fat + efi bootloader. May need to be
4143 * adjusted if the bootloader size will grow.
4144 */
4145 if (boot_size < 1024 * 1024) {
4146 char buf[64];
4147 zfs_nicenum(boot_size, buf, sizeof (buf));
4148 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4149 "Specified size %s for EFI System partition is too "
4150 "small, the minimum size is 1MB."), buf);
4151 (void) close(fd);
4152 efi_free(vtoc);
4153 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
4154 }
4155 /* 33MB is tested with mkfs -F pcfs */
4156 if (hdl->libzfs_printerr &&
4157 ((vtoc->efi_lbasize == 512 &&
4158 boot_size < 33 * 1024 * 1024) ||
4159 (vtoc->efi_lbasize == 4096 &&
4160 boot_size < 256 * 1024 * 1024))) {
4161 char buf[64];
4162 zfs_nicenum(boot_size, buf, sizeof (buf));
4163 (void) fprintf(stderr, dgettext(TEXT_DOMAIN,
4164 "Warning: EFI System partition size %s is "
4165 "not allowing to create FAT32 file\nsystem, which "
4166 "may result in unbootable system.\n"), buf);
4167 }
4168 /* Adjust zfs partition start by size of system partition. */
4169 start_block += boot_size / vtoc->efi_lbasize;
4170 }
4171
4172 if (start_block == NEW_START_BLOCK) {
4173 /*
4174 * Use default layout.
4175 * ZFS is on slice 0 and slice 8 is reserved.
4176 */
4177 slice_size = vtoc->efi_last_u_lba + 1;
4178 slice_size -= EFI_MIN_RESV_SIZE;
4179 slice_size -= start_block;
4180 if (slice != NULL)
4181 *slice = 0;
4182
4183 vtoc->efi_parts[0].p_start = start_block;
4184 vtoc->efi_parts[0].p_size = slice_size;
4185
4186 vtoc->efi_parts[0].p_tag = V_USR;
4187 (void) strcpy(vtoc->efi_parts[0].p_name, "zfs");
4188
4189 vtoc->efi_parts[8].p_start = slice_size + start_block;
4190 vtoc->efi_parts[8].p_size = resv;
4191 vtoc->efi_parts[8].p_tag = V_RESERVED;
4192 } else {
4193 slice_size = start_block - NEW_START_BLOCK;
4194 vtoc->efi_parts[0].p_start = NEW_START_BLOCK;
4195 vtoc->efi_parts[0].p_size = slice_size;
4196 vtoc->efi_parts[0].p_tag = V_SYSTEM;
4197 (void) strcpy(vtoc->efi_parts[0].p_name, "loader");
4198 if (slice != NULL)
4199 *slice = 1;
4200 /* prepare slice 1 */
4201 slice_size = vtoc->efi_last_u_lba + 1 - slice_size;
4202 slice_size -= resv;
4203 slice_size -= NEW_START_BLOCK;
4204 vtoc->efi_parts[1].p_start = start_block;
4205 vtoc->efi_parts[1].p_size = slice_size;
4206 vtoc->efi_parts[1].p_tag = V_USR;
4207 (void) strcpy(vtoc->efi_parts[1].p_name, "zfs");
4208
4209 vtoc->efi_parts[8].p_start = slice_size + start_block;
4210 vtoc->efi_parts[8].p_size = resv;
4211 vtoc->efi_parts[8].p_tag = V_RESERVED;
4212 }
4213
4214 if (efi_write(fd, vtoc) != 0) {
4215 /*
4216 * Some block drivers (like pcata) may not support EFI
4217 * GPT labels. Print out a helpful error message dir-
4218 * ecting the user to manually label the disk and give
4219 * a specific slice.
4220 */
4221 (void) close(fd);
4222 efi_free(vtoc);
4223
4224 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4225 "try using fdisk(1M) and then provide a specific slice"));
4226 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
4227 }
4228
4229 (void) close(fd);
4230 efi_free(vtoc);
4231 return (0);
4232 }
4233
4234 static boolean_t
4235 supported_dump_vdev_type(libzfs_handle_t *hdl, nvlist_t *config, char *errbuf)
4236 {
4237 char *type;
4238 nvlist_t **child;
4239 uint_t children, c;
4240
4241 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_TYPE, &type) == 0);
4242 if (strcmp(type, VDEV_TYPE_FILE) == 0 ||
4243 strcmp(type, VDEV_TYPE_HOLE) == 0 ||
4244 strcmp(type, VDEV_TYPE_MISSING) == 0) {
4245 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4246 "vdev type '%s' is not supported"), type);
4247 (void) zfs_error(hdl, EZFS_VDEVNOTSUP, errbuf);
4248 return (B_FALSE);
4249 }
4250 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN,
4251 &child, &children) == 0) {
4252 for (c = 0; c < children; c++) {
4253 if (!supported_dump_vdev_type(hdl, child[c], errbuf))
4254 return (B_FALSE);
4255 }
4256 }
4257 return (B_TRUE);
4258 }
4259
4260 /*
4261 * Check if this zvol is allowable for use as a dump device; zero if
4262 * it is, > 0 if it isn't, < 0 if it isn't a zvol.
4263 *
4264 * Allowable storage configurations include mirrors, all raidz variants, and
4265 * pools with log, cache, and spare devices. Pools which are backed by files or
4266 * have missing/hole vdevs are not suitable.
4267 */
4268 int
4269 zvol_check_dump_config(char *arg)
4270 {
4271 zpool_handle_t *zhp = NULL;
4272 nvlist_t *config, *nvroot;
4273 char *p, *volname;
4274 nvlist_t **top;
4275 uint_t toplevels;
4276 libzfs_handle_t *hdl;
4277 char errbuf[1024];
4278 char poolname[ZFS_MAX_DATASET_NAME_LEN];
4279 int pathlen = strlen(ZVOL_FULL_DEV_DIR);
4280 int ret = 1;
4281
4282 if (strncmp(arg, ZVOL_FULL_DEV_DIR, pathlen)) {
4283 return (-1);
4284 }
4285
4286 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
4287 "dump is not supported on device '%s'"), arg);
4288
4289 if ((hdl = libzfs_init()) == NULL)
4290 return (1);
4291 libzfs_print_on_error(hdl, B_TRUE);
4292
4293 volname = arg + pathlen;
4294
4295 /* check the configuration of the pool */
4296 if ((p = strchr(volname, '/')) == NULL) {
4297 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4298 "malformed dataset name"));
4299 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
4300 return (1);
4301 } else if (p - volname >= ZFS_MAX_DATASET_NAME_LEN) {
4302 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4303 "dataset name is too long"));
4304 (void) zfs_error(hdl, EZFS_NAMETOOLONG, errbuf);
4305 return (1);
4306 } else {
4307 (void) strncpy(poolname, volname, p - volname);
4308 poolname[p - volname] = '\0';
4309 }
4310
4311 if ((zhp = zpool_open(hdl, poolname)) == NULL) {
4312 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4313 "could not open pool '%s'"), poolname);
4314 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
4315 goto out;
4316 }
4317 config = zpool_get_config(zhp, NULL);
4318 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
4319 &nvroot) != 0) {
4320 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4321 "could not obtain vdev configuration for '%s'"), poolname);
4322 (void) zfs_error(hdl, EZFS_INVALCONFIG, errbuf);
4323 goto out;
4324 }
4325
4326 verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
4327 &top, &toplevels) == 0);
4328
4329 if (!supported_dump_vdev_type(hdl, top[0], errbuf)) {
4330 goto out;
4331 }
4332 ret = 0;
4333
4334 out:
4335 if (zhp)
4336 zpool_close(zhp);
4337 libzfs_fini(hdl);
4338 return (ret);
4339 }
4340
4341 /*
4342 * Vdev props
4343 */
4344 static int
4345 vdev_get_guid(zpool_handle_t *zhp, const char *path, uint64_t *guid)
4346 {
4347 nvlist_t *nvl;
4348 boolean_t avail_spare, l2cache, islog;
4349
4350 if ((nvl = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
4351 &islog, NULL)) == NULL)
4352 return (1);
4353 verify(nvlist_lookup_uint64(nvl, ZPOOL_CONFIG_GUID, guid) == 0);
4354 return (0);
4355 }
4356
4357 /*
4358 * Given an nvlist of vdev properties to be set, validate that they are
4359 * correct, and parse any numeric properties (index, boolean, etc) if they are
4360 * specified as strings.
4361 */
4362 /*ARGSUSED*/
4363 static nvlist_t *
4364 vdev_valid_proplist(libzfs_handle_t *hdl, nvlist_t *props,
4365 uint64_t version, prop_flags_t flags, const char *errbuf)
4366 {
4367 nvpair_t *elem;
4368 nvlist_t *retprops;
4369 vdev_prop_t prop;
4370 char *strval;
4371 uint64_t intval;
4372
4373 if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) {
4374 (void) no_memory(hdl);
4375 return (NULL);
4376 }
4377
4378 elem = NULL;
4379 while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
4380 const char *propname = nvpair_name(elem);
4381
4382 /*
4383 * Make sure this property is valid and applies to this type.
4384 */
4385 if ((prop = vdev_name_to_prop(propname)) == ZPROP_INVAL) {
4386 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4387 "invalid vdev property '%s'"), propname);
4388 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
4389 goto error;
4390 }
4391
4392 if (vdev_prop_readonly(prop)) {
4393 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
4394 "is readonly"), propname);
4395 (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
4396 goto error;
4397 }
4398
4399 if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_VDEV, retprops,
4400 &strval, &intval, errbuf) != 0)
4401 goto error;
4402
4403 /*
4404 * Perform additional checking for specific properties.
4405 */
4406 switch (prop) {
4407 case VDEV_PROP_PREFERRED_READ:
4408 if (intval > 100) {
4409 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4410 "'%s' must be from 0 to 100"), propname);
4411 (void) zfs_error(hdl, EZFS_BADPROP,
4412 errbuf);
4413 goto error;
4414 }
4415 break;
4416 case VDEV_PROP_READ_MINACTIVE:
4417 case VDEV_PROP_READ_MAXACTIVE:
4418 case VDEV_PROP_AREAD_MINACTIVE:
4419 case VDEV_PROP_AREAD_MAXACTIVE:
4420 case VDEV_PROP_WRITE_MINACTIVE:
4421 case VDEV_PROP_WRITE_MAXACTIVE:
4422 case VDEV_PROP_AWRITE_MINACTIVE:
4423 case VDEV_PROP_AWRITE_MAXACTIVE:
4424 case VDEV_PROP_SCRUB_MINACTIVE:
4425 case VDEV_PROP_SCRUB_MAXACTIVE:
4426 if (intval > 1000) {
4427 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4428 "'%s' must be from 0 to 1000"), propname);
4429 (void) zfs_error(hdl, EZFS_BADPROP,
4430 errbuf);
4431 goto error;
4432 }
4433 break;
4434 default:
4435 break;
4436 }
4437 }
4438
4439 return (retprops);
4440 error:
4441 nvlist_free(retprops);
4442 return (NULL);
4443 }
4444
4445 static int
4446 vdev_get_all_props(zpool_handle_t *zhp, uint64_t vdev_guid, nvlist_t **nvp)
4447 {
4448 zfs_cmd_t zc = { 0 };
4449 libzfs_handle_t *hdl = zhp->zpool_hdl;
4450
4451 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4452
4453 if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0)
4454 return (-1);
4455
4456 zc.zc_guid = vdev_guid;
4457 while (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_GET_PROPS, &zc) != 0) {
4458 if (errno == ENOMEM) {
4459 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
4460 zcmd_free_nvlists(&zc);
4461 return (-1);
4462 }
4463 } else {
4464 zcmd_free_nvlists(&zc);
4465 (void) zfs_error(hdl, EZFS_BADDEV,
4466 "failed to get vdev properties");
4467 return (-1);
4468 }
4469 }
4470
4471 if (zcmd_read_dst_nvlist(hdl, &zc, nvp) != 0) {
4472 zcmd_free_nvlists(&zc);
4473 return (-1);
4474 }
4475
4476 zcmd_free_nvlists(&zc);
4477
4478 return (0);
4479 }
4480
4481 int
4482 vdev_get_prop(zpool_handle_t *zhp, const char *vdev, vdev_prop_t prop,
4483 char *buf, size_t len)
4484 {
4485 uint64_t vdev_guid;
4486 uint64_t intval;
4487 const char *strval;
4488 nvlist_t *nvl;
4489 char errbuf[1024];
4490
4491 (void) snprintf(errbuf, sizeof (errbuf),
4492 dgettext(TEXT_DOMAIN, "cannot get property for '%s'"),
4493 zhp->zpool_name);
4494
4495 if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL)
4496 return (-1);
4497
4498 if (vdev_get_guid(zhp, vdev, &vdev_guid) != 0) {
4499 (void) zfs_error(zhp->zpool_hdl, EZFS_BADDEV, errbuf);
4500 return (-1);
4501 }
4502
4503 if (vdev_get_all_props(zhp, vdev_guid, &nvl) != 0)
4504 return (-1);
4505
4506 switch (vdev_prop_get_type(prop)) {
4507 case PROP_TYPE_STRING:
4508 if (nvlist_lookup_string(nvl, vdev_prop_to_name(prop),
4509 (char **)&strval) != 0)
4510 if ((strval = (char *)vdev_prop_default_string(prop))
4511 == NULL)
4512 strval = "-";
4513
4514 (void) strlcpy(buf, strval, len);
4515 break;
4516
4517 case PROP_TYPE_NUMBER:
4518 if (nvlist_lookup_uint64(nvl, vdev_prop_to_name(prop),
4519 &intval) != 0)
4520 intval = vdev_prop_default_numeric(prop);
4521 (void) snprintf(buf, len, "%llu", (u_longlong_t)intval);
4522 break;
4523
4524 case PROP_TYPE_INDEX:
4525 if (nvlist_lookup_uint64(nvl, vdev_prop_to_name(prop),
4526 &intval) != 0)
4527 intval = vdev_prop_default_numeric(prop);
4528 if (vdev_prop_index_to_string(prop, intval, &strval) != 0) {
4529 (void) zfs_error(zhp->zpool_hdl, EZFS_BADPROP, errbuf);
4530 nvlist_free(nvl);
4531 return (-1);
4532 }
4533 (void) strlcpy(buf, strval, len);
4534 break;
4535
4536 default:
4537 abort();
4538 }
4539
4540 nvlist_free(nvl);
4541
4542 return (0);
4543 }
4544
4545 /*
4546 * Set vdev property : propname=propval.
4547 */
4548 int
4549 vdev_set_prop(zpool_handle_t *zhp, const char *vdev,
4550 const char *propname, const char *propval)
4551 {
4552 zfs_cmd_t zc = { 0 };
4553 int ret = -1;
4554 char errbuf[1024];
4555 nvlist_t *nvl = NULL;
4556 nvlist_t *realprops;
4557 uint64_t version;
4558 uint64_t guid;
4559 prop_flags_t flags = { 0 };
4560
4561 (void) snprintf(errbuf, sizeof (errbuf),
4562 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
4563 zhp->zpool_name);
4564
4565 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
4566 return (no_memory(zhp->zpool_hdl));
4567
4568 if (nvlist_add_string(nvl, propname, propval) != 0) {
4569 nvlist_free(nvl);
4570 return (no_memory(zhp->zpool_hdl));
4571 }
4572
4573 if (vdev_get_guid(zhp, vdev, &guid) != 0) {
4574 (void) zfs_error(zhp->zpool_hdl, EZFS_BADDEV, errbuf);
4575 return (-1);
4576 }
4577
4578 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
4579 if ((realprops = vdev_valid_proplist(zhp->zpool_hdl, nvl,
4580 version, flags, errbuf)) == NULL) {
4581 nvlist_free(nvl);
4582 return (-1);
4583 }
4584
4585 nvlist_free(nvl);
4586 nvl = realprops;
4587
4588 /*
4589 * Execute the corresponding ioctl() to set this property.
4590 */
4591 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4592 zc.zc_guid = guid;
4593
4594 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) {
4595 nvlist_free(nvl);
4596 return (-1);
4597 }
4598
4599 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_VDEV_SET_PROPS, &zc);
4600
4601 zcmd_free_nvlists(&zc);
4602 nvlist_free(nvl);
4603
4604 if (ret)
4605 (void) zpool_vprop_standard_error(zhp->zpool_hdl,
4606 errno, errbuf);
4607
4608 return (ret);
4609 }
4610
4611 /*
4612 * Set vdev properties nvlist
4613 */
4614 int
4615 vdev_set_proplist(zpool_handle_t *zhp, const char *vdev, nvlist_t *nvl)
4616 {
4617 zfs_cmd_t zc = { 0 };
4618 int ret = -1;
4619 char errbuf[1024];
4620 nvlist_t *realprops;
4621 uint64_t version;
4622 uint64_t guid;
4623 prop_flags_t flags = { 0 };
4624 libzfs_handle_t *hdl = zhp->zpool_hdl;
4625
4626 assert(nvl != NULL);
4627
4628 (void) snprintf(errbuf, sizeof (errbuf),
4629 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
4630 zhp->zpool_name);
4631
4632 if (vdev_get_guid(zhp, vdev, &guid) != 0) {
4633 (void) zfs_error(hdl, EZFS_BADDEV, errbuf);
4634 return (-1);
4635 }
4636
4637 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
4638 if ((realprops = vdev_valid_proplist(zhp->zpool_hdl, nvl,
4639 version, flags, errbuf)) == NULL) {
4640 nvlist_free(nvl);
4641 return (-1);
4642 }
4643
4644 nvlist_free(nvl);
4645 nvl = realprops;
4646
4647 /*
4648 * Execute the corresponding ioctl() to set this property.
4649 */
4650 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4651 zc.zc_guid = guid;
4652
4653 if (zcmd_write_src_nvlist(hdl, &zc, nvl) != 0) {
4654 nvlist_free(nvl);
4655 return (-1);
4656 }
4657
4658 ret = zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_PROPS, &zc);
4659
4660 zcmd_free_nvlists(&zc);
4661 nvlist_free(nvl);
4662
4663 if (ret)
4664 (void) zpool_vprop_standard_error(hdl, errno, errbuf);
4665
4666 return (ret);
4667 }
4668
4669 typedef struct vdev_cb {
4670 zpool_handle_t *vcb_zhp;
4671 char *vcb_name;
4672 uint64_t vcb_guid;
4673 boolean_t vcb_is_leaf;
4674 boolean_t vcb_success;
4675 void *vcb_data;
4676 } vdev_cb_t;
4677
4678 typedef int (*vdev_callback_t)(vdev_cb_t *);
4679
4680 /*
4681 * Invoke dev_callback for all vdevs in the pool
4682 */
4683 int
4684 for_each_vdev(zpool_handle_t *zhp, nvlist_t *root,
4685 vdev_callback_t vdev_callback, vdev_cb_t *cb)
4686 {
4687 int ret = 0;
4688 nvlist_t *config, *nvroot, **child, **child_list;
4689 uint32_t children, child_children, c;
4690 libzfs_handle_t *hdl = zhp->zpool_hdl;
4691
4692 if (!root) {
4693 config = zpool_get_config(zhp, NULL);
4694 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
4695 &nvroot) == 0);
4696 } else {
4697 nvroot = root;
4698 }
4699
4700 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
4701 &child, &children) != 0) {
4702 (void) fprintf(stderr, gettext("Failed to get the vdev "
4703 "children details from the root nvlist\n"));
4704 return (-1);
4705 }
4706
4707 cb->vcb_zhp = zhp;
4708 for (c = 0; c < children; c++) {
4709 cb->vcb_is_leaf = B_TRUE;
4710 if (nvlist_lookup_nvlist_array(child[c], ZPOOL_CONFIG_CHILDREN,
4711 &child_list, &child_children) == 0 &&
4712 child_children > 0) {
4713 ret = for_each_vdev(zhp, child[c], vdev_callback, cb);
4714 if (ret)
4715 return (ret);
4716 cb->vcb_is_leaf = B_FALSE;
4717 }
4718
4719 cb->vcb_name = zpool_vdev_name(hdl, NULL, child[c], B_TRUE);
4720 verify(nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_GUID,
4721 &cb->vcb_guid) == 0);
4722
4723 ret |= vdev_callback(cb);
4724 free(cb->vcb_name);
4725 if (ret)
4726 return (ret);
4727 }
4728
4729 return (0);
4730 }
4731
4732 int
4733 get_vdev_guid_callback(vdev_cb_t *cb)
4734 {
4735 if (!cb->vcb_is_leaf)
4736 return (0);
4737
4738 if (strncmp(cb->vcb_name, (char *)cb->vcb_data,
4739 strlen(cb->vcb_name)) == 0) {
4740 cb->vcb_success = B_TRUE;
4741 }
4742
4743 return (0);
4744 }
4745
4746 /*
4747 * Class of Storage (COS)
4748 */
4749 /*ARGSUSED*/
4750 static nvlist_t *
4751 cos_valid_proplist(libzfs_handle_t *hdl, nvlist_t *props,
4752 uint64_t version, prop_flags_t flags, const char *errbuf)
4753 {
4754 nvpair_t *elem;
4755 nvlist_t *retprops;
4756 cos_prop_t prop;
4757 char *strval;
4758 uint64_t intval;
4759
4760 if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) {
4761 (void) no_memory(hdl);
4762 return (NULL);
4763 }
4764
4765 elem = NULL;
4766 while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
4767 const char *propname = nvpair_name(elem);
4768
4769 /*
4770 * Make sure this property is valid and applies to this type.
4771 */
4772 if ((prop = cos_name_to_prop(propname)) == ZPROP_INVAL) {
4773 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4774 "invalid cos property '%s'"), propname);
4775 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
4776 goto error;
4777 }
4778
4779 if (cos_prop_readonly(prop)) {
4780 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
4781 "is readonly"), propname);
4782 (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
4783 goto error;
4784 }
4785
4786 if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_COS, retprops,
4787 &strval, &intval, errbuf) != 0)
4788 goto error;
4789
4790 /*
4791 * Perform additional checking for specific properties.
4792 */
4793 switch (prop) {
4794 case COS_PROP_PREFERRED_READ:
4795 if (intval > 100) {
4796 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4797 "'%s' must be from 0 to 100"), propname);
4798 (void) zfs_error(hdl, EZFS_BADPROP,
4799 errbuf);
4800 goto error;
4801 }
4802 break;
4803 case COS_PROP_READ_MINACTIVE:
4804 case COS_PROP_AREAD_MINACTIVE:
4805 case COS_PROP_WRITE_MINACTIVE:
4806 case COS_PROP_AWRITE_MINACTIVE:
4807 case COS_PROP_SCRUB_MINACTIVE:
4808 case COS_PROP_READ_MAXACTIVE:
4809 case COS_PROP_AREAD_MAXACTIVE:
4810 case COS_PROP_WRITE_MAXACTIVE:
4811 case COS_PROP_AWRITE_MAXACTIVE:
4812 case COS_PROP_SCRUB_MAXACTIVE:
4813 if (intval > 1000) {
4814 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
4815 "'%s' must be from 0 to 1000"), propname);
4816 (void) zfs_error(hdl, EZFS_BADPROP,
4817 errbuf);
4818 goto error;
4819 }
4820 break;
4821 default:
4822 break;
4823 }
4824 }
4825
4826 return (retprops);
4827 error:
4828 nvlist_free(retprops);
4829 return (NULL);
4830 }
4831
4832 int
4833 cos_alloc(zpool_handle_t *zhp, char *cosname, nvlist_t *nvl)
4834 {
4835 zfs_cmd_t zc = { 0 };
4836 libzfs_handle_t *hdl = zhp->zpool_hdl;
4837 char errbuf[1024];
4838 int error = 0;
4839
4840 (void) snprintf(errbuf, sizeof (errbuf),
4841 dgettext(TEXT_DOMAIN, "cannot allocate CoS descriptor for '%s'"),
4842 zhp->zpool_name);
4843
4844 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4845 (void) strlcpy(zc.zc_string, cosname, sizeof (zc.zc_string));
4846
4847 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) {
4848 nvlist_free(nvl);
4849 return (-1);
4850 }
4851
4852 error = ioctl(hdl->libzfs_fd, ZFS_IOC_COS_ALLOC, &zc);
4853
4854 if (error)
4855 (void) zpool_vprop_standard_error(hdl, errno, errbuf);
4856
4857 return (error);
4858 }
4859
4860 int
4861 cos_free(zpool_handle_t *zhp, char *cosname, uint64_t guid, boolean_t force)
4862 {
4863 zfs_cmd_t zc = { 0 };
4864 libzfs_handle_t *hdl = zhp->zpool_hdl;
4865 char errbuf[1024];
4866 int error = 0;
4867
4868 (void) snprintf(errbuf, sizeof (errbuf),
4869 dgettext(TEXT_DOMAIN, "cannot free CoS descriptor for '%s'"),
4870 zhp->zpool_name);
4871
4872 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4873 (void) strlcpy(zc.zc_string, cosname, sizeof (zc.zc_string));
4874 zc.zc_guid = guid;
4875
4876 zc.zc_cookie = (uint64_t)force;
4877
4878 error = ioctl(hdl->libzfs_fd, ZFS_IOC_COS_FREE, &zc);
4879
4880 if (error)
4881 (void) zpool_vprop_standard_error(hdl, errno, errbuf);
4882
4883 return (error);
4884 }
4885
4886 int
4887 cos_list(zpool_handle_t *zhp, nvlist_t **nvp)
4888 {
4889 zfs_cmd_t zc = { 0 };
4890 libzfs_handle_t *hdl = zhp->zpool_hdl;
4891
4892 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4893
4894 if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0)
4895 return (-1);
4896
4897 while (ioctl(hdl->libzfs_fd, ZFS_IOC_COS_LIST, &zc) != 0) {
4898 if (errno == ENOMEM) {
4899 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
4900 zcmd_free_nvlists(&zc);
4901 return (-1);
4902 }
4903 } else {
4904 zcmd_free_nvlists(&zc);
4905 return (-1);
4906 }
4907 }
4908
4909 if (zcmd_read_dst_nvlist(hdl, &zc, nvp) != 0) {
4910 zcmd_free_nvlists(&zc);
4911 return (-1);
4912 }
4913
4914 zcmd_free_nvlists(&zc);
4915
4916 return (0);
4917 }
4918
4919 int
4920 cos_get_all_props(zpool_handle_t *zhp, const char *cos, nvlist_t **nvp)
4921 {
4922 uint64_t cos_id;
4923 zfs_cmd_t zc = { 0 };
4924 libzfs_handle_t *hdl = zhp->zpool_hdl;
4925 char *endp;
4926 char errbuf[1024];
4927
4928 (void) snprintf(errbuf, sizeof (errbuf),
4929 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
4930 zhp->zpool_name);
4931
4932 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
4933
4934 if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0)
4935 return (-1);
4936
4937 cos_id = strtoll(cos, &endp, 10);
4938 zc.zc_guid = cos_id;
4939 if (cos_id == 0 && cos == endp)
4940 (void) strlcpy(zc.zc_string, cos, sizeof (zc.zc_string));
4941 else
4942 zc.zc_string[0] = '\0';
4943
4944 while (ioctl(hdl->libzfs_fd, ZFS_IOC_COS_GET_PROPS, &zc) != 0) {
4945 if (errno == ENOMEM) {
4946 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
4947 zcmd_free_nvlists(&zc);
4948 return (-1);
4949 }
4950 } else {
4951 zcmd_free_nvlists(&zc);
4952 (void) zpool_standard_error(hdl, errno, errbuf);
4953 return (-1);
4954 }
4955 }
4956
4957 if (zcmd_read_dst_nvlist(hdl, &zc, nvp) != 0) {
4958 zcmd_free_nvlists(&zc);
4959 return (-1);
4960 }
4961
4962 zcmd_free_nvlists(&zc);
4963
4964 return (0);
4965 }
4966
4967 int
4968 cos_get_prop(zpool_handle_t *zhp, const char *cos, cos_prop_t prop,
4969 char *buf, size_t len, nvlist_t **nvp)
4970 {
4971 uint64_t intval;
4972 const char *strval;
4973 nvlist_t *nvl;
4974 char errbuf[1024];
4975
4976 (void) snprintf(errbuf, sizeof (errbuf),
4977 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
4978 zhp->zpool_name);
4979
4980 assert(nvp != NULL);
4981
4982 if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL)
4983 return (-1);
4984
4985 if (*nvp == NULL && cos_get_all_props(zhp, cos, nvp) != 0)
4986 return (-1);
4987 nvl = *nvp;
4988
4989 switch (cos_prop_get_type(prop)) {
4990 case PROP_TYPE_STRING:
4991 if (nvlist_lookup_string(nvl, cos_prop_to_name(prop),
4992 (char **)&strval) != 0)
4993 if ((strval = (char *)cos_prop_default_string(prop))
4994 == NULL)
4995 strval = "-";
4996
4997 (void) strlcpy(buf, strval, len);
4998 break;
4999
5000 case PROP_TYPE_NUMBER:
5001 if (nvlist_lookup_uint64(nvl, cos_prop_to_name(prop),
5002 &intval) != 0)
5003 intval = cos_prop_default_numeric(prop);
5004 (void) snprintf(buf, len, "%llu", (u_longlong_t)intval);
5005 break;
5006
5007 case PROP_TYPE_INDEX:
5008 if (nvlist_lookup_uint64(nvl, cos_prop_to_name(prop),
5009 &intval) != 0)
5010 intval = cos_prop_default_numeric(prop);
5011 if (cos_prop_index_to_string(prop, intval, &strval) != 0) {
5012 (void) zfs_error(zhp->zpool_hdl, EZFS_BADPROP, errbuf);
5013 return (-1);
5014 }
5015 (void) strlcpy(buf, strval, len);
5016 break;
5017
5018 default:
5019 abort();
5020 }
5021
5022 return (0);
5023 }
5024
5025 /*
5026 * Set cos properties nvlist
5027 */
5028 int
5029 cos_set_proplist(zpool_handle_t *zhp, const char *cos, nvlist_t *nvl)
5030 {
5031 zfs_cmd_t zc = { 0 };
5032 int ret = -1;
5033 char errbuf[1024];
5034 char *endp;
5035 nvlist_t *realprops;
5036 uint64_t version;
5037 uint64_t cos_id;
5038 prop_flags_t flags = { 0 };
5039
5040 assert(nvl != NULL);
5041
5042 (void) snprintf(errbuf, sizeof (errbuf),
5043 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
5044 zhp->zpool_name);
5045
5046 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
5047 if ((realprops = cos_valid_proplist(zhp->zpool_hdl, nvl,
5048 version, flags, errbuf)) == NULL) {
5049 nvlist_free(nvl);
5050 return (-1);
5051 }
5052
5053 nvlist_free(nvl);
5054 nvl = realprops;
5055
5056 /*
5057 * Execute the corresponding ioctl() to set this property.
5058 */
5059 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
5060
5061 cos_id = strtoll(cos, &endp, 10);
5062 zc.zc_guid = cos_id;
5063 if (cos_id == 0 && cos == endp)
5064 (void) strlcpy(zc.zc_string, cos, sizeof (zc.zc_string));
5065 else
5066 zc.zc_string[0] = '\0';
5067
5068 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) {
5069 nvlist_free(nvl);
5070 return (-1);
5071 }
5072
5073 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_COS_SET_PROPS, &zc);
5074
5075 zcmd_free_nvlists(&zc);
5076 nvlist_free(nvl);
5077
5078 if (ret)
5079 (void) zpool_vprop_standard_error_fmt(zhp->zpool_hdl,
5080 errno, errbuf);
5081
5082 return (ret);
5083 }