]> git.proxmox.com Git - mirror_zfs-debian.git/blame - lib/libzfs/libzfs_pool.c
projects / mirror_zfs-debian.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Support shorthand names with zpool remove
[mirror_zfs-debian.git] / lib / libzfs / libzfs_pool.c
CommitLineData
34dc7c2f
BB		 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/*
428870ff 23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
34dc7c2f
BB		 24 */
25
34dc7c2f
BB		 26#include <ctype.h>
27#include <errno.h>
28#include <devid.h>
34dc7c2f
BB		 29#include <fcntl.h>
30#include <libintl.h>
31#include <stdio.h>
32#include <stdlib.h>
33#include <strings.h>
34#include <unistd.h>
d603ed6c
BB		 35#include <zone.h>
36#include <sys/stat.h>
34dc7c2f
BB		 37#include <sys/efi_partition.h>
38#include <sys/vtoc.h>
39#include <sys/zfs_ioctl.h>
9babb374 40#include <dlfcn.h>
34dc7c2f
BB		 41
42#include "zfs_namecheck.h"
43#include "zfs_prop.h"
44#include "libzfs_impl.h"
428870ff 45#include "zfs_comutil.h"
34dc7c2f 46
b128c09f
BB		 47static int read_efi_label(nvlist_t *config, diskaddr_t *sb);
48
572e2857
BB		 49typedef struct prop_flags {
50 int create:1; /* Validate property on creation */
51 int import:1; /* Validate property on import */
52} prop_flags_t;
53
34dc7c2f
BB		 54/*
55 * ====================================================================
56 * zpool property functions
57 * ====================================================================
58 */
59
60static int
61zpool_get_all_props(zpool_handle_t *zhp)
62{
2598c001 63 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
34dc7c2f
BB		 64 libzfs_handle_t *hdl = zhp->zpool_hdl;
65
66 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
67
68 if (zcmd_alloc_dst_nvlist(hdl, &zc, 0) != 0)
69 return (-1);
70
71 while (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_PROPS, &zc) != 0) {
72 if (errno == ENOMEM) {
73 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
74 zcmd_free_nvlists(&zc);
75 return (-1);
76 }
77 } else {
78 zcmd_free_nvlists(&zc);
79 return (-1);
80 }
81 }
82
83 if (zcmd_read_dst_nvlist(hdl, &zc, &zhp->zpool_props) != 0) {
84 zcmd_free_nvlists(&zc);
85 return (-1);
86 }
87
88 zcmd_free_nvlists(&zc);
89
90 return (0);
91}
92
93static int
94zpool_props_refresh(zpool_handle_t *zhp)
95{
96 nvlist_t *old_props;
97
98 old_props = zhp->zpool_props;
99
100 if (zpool_get_all_props(zhp) != 0)
101 return (-1);
102
103 nvlist_free(old_props);
104 return (0);
105}
106
107static char *
108zpool_get_prop_string(zpool_handle_t *zhp, zpool_prop_t prop,
109 zprop_source_t *src)
110{
111 nvlist_t *nv, *nvl;
112 uint64_t ival;
113 char *value;
114 zprop_source_t source;
115
116 nvl = zhp->zpool_props;
117 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
118 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &ival) == 0);
119 source = ival;
120 verify(nvlist_lookup_string(nv, ZPROP_VALUE, &value) == 0);
121 } else {
122 source = ZPROP_SRC_DEFAULT;
123 if ((value = (char *)zpool_prop_default_string(prop)) == NULL)
124 value = "-";
125 }
126
127 if (src)
128 *src = source;
129
130 return (value);
131}
132
133uint64_t
134zpool_get_prop_int(zpool_handle_t *zhp, zpool_prop_t prop, zprop_source_t *src)
135{
136 nvlist_t *nv, *nvl;
137 uint64_t value;
138 zprop_source_t source;
139
b128c09f
BB		 140 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp)) {
141 /*
142 * zpool_get_all_props() has most likely failed because
143 * the pool is faulted, but if all we need is the top level
144 * vdev's guid then get it from the zhp config nvlist.
145 */
146 if ((prop == ZPOOL_PROP_GUID) &&
147 (nvlist_lookup_nvlist(zhp->zpool_config,
148 ZPOOL_CONFIG_VDEV_TREE, &nv) == 0) &&
149 (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, &value)
150 == 0)) {
151 return (value);
152 }
34dc7c2f 153 return (zpool_prop_default_numeric(prop));
b128c09f 154 }
34dc7c2f
BB		 155
156 nvl = zhp->zpool_props;
157 if (nvlist_lookup_nvlist(nvl, zpool_prop_to_name(prop), &nv) == 0) {
158 verify(nvlist_lookup_uint64(nv, ZPROP_SOURCE, &value) == 0);
159 source = value;
160 verify(nvlist_lookup_uint64(nv, ZPROP_VALUE, &value) == 0);
161 } else {
162 source = ZPROP_SRC_DEFAULT;
163 value = zpool_prop_default_numeric(prop);
164 }
165
166 if (src)
167 *src = source;
168
169 return (value);
170}
171
172/*
173 * Map VDEV STATE to printed strings.
174 */
175char *
176zpool_state_to_name(vdev_state_t state, vdev_aux_t aux)
177{
178 switch (state) {
e75c13c3
BB		 179 default:
180 break;
34dc7c2f
BB		 181 case VDEV_STATE_CLOSED:
182 case VDEV_STATE_OFFLINE:
183 return (gettext("OFFLINE"));
184 case VDEV_STATE_REMOVED:
185 return (gettext("REMOVED"));
186 case VDEV_STATE_CANT_OPEN:
b128c09f 187 if (aux == VDEV_AUX_CORRUPT_DATA || aux == VDEV_AUX_BAD_LOG)
34dc7c2f 188 return (gettext("FAULTED"));
428870ff
BB		 189 else if (aux == VDEV_AUX_SPLIT_POOL)
190 return (gettext("SPLIT"));
34dc7c2f
BB		 191 else
192 return (gettext("UNAVAIL"));
193 case VDEV_STATE_FAULTED:
194 return (gettext("FAULTED"));
195 case VDEV_STATE_DEGRADED:
196 return (gettext("DEGRADED"));
197 case VDEV_STATE_HEALTHY:
198 return (gettext("ONLINE"));
199 }
200
201 return (gettext("UNKNOWN"));
202}
203
204/*
205 * Get a zpool property value for 'prop' and return the value in
206 * a pre-allocated buffer.
207 */
208int
209zpool_get_prop(zpool_handle_t *zhp, zpool_prop_t prop, char *buf, size_t len,
210 zprop_source_t *srctype)
211{
212 uint64_t intval;
213 const char *strval;
214 zprop_source_t src = ZPROP_SRC_NONE;
215 nvlist_t *nvroot;
216 vdev_stat_t *vs;
217 uint_t vsc;
218
219 if (zpool_get_state(zhp) == POOL_STATE_UNAVAIL) {
d164b209
BB		 220 switch (prop) {
221 case ZPOOL_PROP_NAME:
34dc7c2f 222 (void) strlcpy(buf, zpool_get_name(zhp), len);
d164b209
BB		 223 break;
224
225 case ZPOOL_PROP_HEALTH:
34dc7c2f 226 (void) strlcpy(buf, "FAULTED", len);
d164b209
BB		 227 break;
228
229 case ZPOOL_PROP_GUID:
230 intval = zpool_get_prop_int(zhp, prop, &src);
b8864a23 231 (void) snprintf(buf, len, "%llu", (u_longlong_t)intval);
d164b209
BB		 232 break;
233
234 case ZPOOL_PROP_ALTROOT:
235 case ZPOOL_PROP_CACHEFILE:
236 if (zhp->zpool_props != NULL ||
237 zpool_get_all_props(zhp) == 0) {
238 (void) strlcpy(buf,
239 zpool_get_prop_string(zhp, prop, &src),
240 len);
241 if (srctype != NULL)
242 *srctype = src;
243 return (0);
244 }
245 /* FALLTHROUGH */
246 default:
34dc7c2f 247 (void) strlcpy(buf, "-", len);
d164b209
BB		 248 break;
249 }
250
251 if (srctype != NULL)
252 *srctype = src;
34dc7c2f
BB		 253 return (0);
254 }
255
256 if (zhp->zpool_props == NULL && zpool_get_all_props(zhp) &&
257 prop != ZPOOL_PROP_NAME)
258 return (-1);
259
260 switch (zpool_prop_get_type(prop)) {
261 case PROP_TYPE_STRING:
262 (void) strlcpy(buf, zpool_get_prop_string(zhp, prop, &src),
263 len);
264 break;
265
266 case PROP_TYPE_NUMBER:
267 intval = zpool_get_prop_int(zhp, prop, &src);
268
269 switch (prop) {
270 case ZPOOL_PROP_SIZE:
428870ff
BB		 271 case ZPOOL_PROP_ALLOCATED:
272 case ZPOOL_PROP_FREE:
34dc7c2f
BB		 273 (void) zfs_nicenum(intval, buf, len);
274 break;
275
276 case ZPOOL_PROP_CAPACITY:
277 (void) snprintf(buf, len, "%llu%%",
278 (u_longlong_t)intval);
279 break;
280
428870ff
BB		 281 case ZPOOL_PROP_DEDUPRATIO:
282 (void) snprintf(buf, len, "%llu.%02llux",
283 (u_longlong_t)(intval / 100),
284 (u_longlong_t)(intval % 100));
285 break;
286
34dc7c2f
BB		 287 case ZPOOL_PROP_HEALTH:
288 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
289 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
290 verify(nvlist_lookup_uint64_array(nvroot,
428870ff
BB		 291 ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc)
292 == 0);
34dc7c2f
BB		 293
294 (void) strlcpy(buf, zpool_state_to_name(intval,
295 vs->vs_aux), len);
296 break;
297 default:
b8864a23 298 (void) snprintf(buf, len, "%llu", (u_longlong_t)intval);
34dc7c2f
BB		 299 }
300 break;
301
302 case PROP_TYPE_INDEX:
303 intval = zpool_get_prop_int(zhp, prop, &src);
304 if (zpool_prop_index_to_string(prop, intval, &strval)
305 != 0)
306 return (-1);
307 (void) strlcpy(buf, strval, len);
308 break;
309
310 default:
311 abort();
312 }
313
314 if (srctype)
315 *srctype = src;
316
317 return (0);
318}
319
320/*
321 * Check if the bootfs name has the same pool name as it is set to.
322 * Assuming bootfs is a valid dataset name.
323 */
324static boolean_t
325bootfs_name_valid(const char *pool, char *bootfs)
326{
327 int len = strlen(pool);
328
b128c09f 329 if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT))
34dc7c2f
BB		 330 return (B_FALSE);
331
332 if (strncmp(pool, bootfs, len) == 0 &&
333 (bootfs[len] == '/' || bootfs[len] == '\0'))
334 return (B_TRUE);
335
336 return (B_FALSE);
337}
338
b128c09f
BB		 339/*
340 * Inspect the configuration to determine if any of the devices contain
341 * an EFI label.
342 */
343static boolean_t
344pool_uses_efi(nvlist_t *config)
345{
346 nvlist_t **child;
347 uint_t c, children;
348
349 if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN,
350 &child, &children) != 0)
351 return (read_efi_label(config, NULL) >= 0);
352
353 for (c = 0; c < children; c++) {
354 if (pool_uses_efi(child[c]))
355 return (B_TRUE);
356 }
357 return (B_FALSE);
358}
359
360static boolean_t
361pool_is_bootable(zpool_handle_t *zhp)
362{
363 char bootfs[ZPOOL_MAXNAMELEN];
364
365 return (zpool_get_prop(zhp, ZPOOL_PROP_BOOTFS, bootfs,
366 sizeof (bootfs), NULL) == 0 && strncmp(bootfs, "-",
367 sizeof (bootfs)) != 0);
368}
369
370
34dc7c2f
BB		 371/*
372 * Given an nvlist of zpool properties to be set, validate that they are
373 * correct, and parse any numeric properties (index, boolean, etc) if they are
374 * specified as strings.
375 */
376static nvlist_t *
b128c09f 377zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname,
572e2857 378 nvlist_t *props, uint64_t version, prop_flags_t flags, char *errbuf)
34dc7c2f
BB		 379{
380 nvpair_t *elem;
381 nvlist_t *retprops;
382 zpool_prop_t prop;
383 char *strval;
384 uint64_t intval;
385 char *slash;
386 struct stat64 statbuf;
b128c09f
BB		 387 zpool_handle_t *zhp;
388 nvlist_t *nvroot;
34dc7c2f
BB		 389
390 if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) {
391 (void) no_memory(hdl);
392 return (NULL);
393 }
394
395 elem = NULL;
396 while ((elem = nvlist_next_nvpair(props, elem)) != NULL) {
397 const char *propname = nvpair_name(elem);
398
399 /*
400 * Make sure this property is valid and applies to this type.
401 */
402 if ((prop = zpool_name_to_prop(propname)) == ZPROP_INVAL) {
403 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
404 "invalid property '%s'"), propname);
405 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
406 goto error;
407 }
408
409 if (zpool_prop_readonly(prop)) {
410 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
411 "is readonly"), propname);
412 (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf);
413 goto error;
414 }
415
416 if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops,
417 &strval, &intval, errbuf) != 0)
418 goto error;
419
420 /*
421 * Perform additional checking for specific properties.
422 */
423 switch (prop) {
e75c13c3
BB		 424 default:
425 break;
34dc7c2f
BB		 426 case ZPOOL_PROP_VERSION:
427 if (intval < version || intval > SPA_VERSION) {
428 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
429 "property '%s' number %d is invalid."),
430 propname, intval);
431 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
432 goto error;
433 }
434 break;
435
436 case ZPOOL_PROP_BOOTFS:
572e2857 437 if (flags.create || flags.import) {
34dc7c2f
BB		 438 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
439 "property '%s' cannot be set at creation "
440 "or import time"), propname);
441 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
442 goto error;
443 }
444
445 if (version < SPA_VERSION_BOOTFS) {
446 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
447 "pool must be upgraded to support "
448 "'%s' property"), propname);
449 (void) zfs_error(hdl, EZFS_BADVERSION, errbuf);
450 goto error;
451 }
452
453 /*
454 * bootfs property value has to be a dataset name and
455 * the dataset has to be in the same pool as it sets to.
456 */
457 if (strval[0] != '\0' && !bootfs_name_valid(poolname,
458 strval)) {
459 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' "
460 "is an invalid name"), strval);
461 (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf);
462 goto error;
463 }
b128c09f
BB		 464
465 if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) {
466 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
467 "could not open pool '%s'"), poolname);
468 (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf);
469 goto error;
470 }
471 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
472 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
473
474 /*
475 * bootfs property cannot be set on a disk which has
476 * been EFI labeled.
477 */
478 if (pool_uses_efi(nvroot)) {
479 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
480 "property '%s' not supported on "
481 "EFI labeled devices"), propname);
482 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf);
483 zpool_close(zhp);
484 goto error;
485 }
486 zpool_close(zhp);
34dc7c2f
BB		 487 break;
488
489 case ZPOOL_PROP_ALTROOT:
572e2857 490 if (!flags.create && !flags.import) {
34dc7c2f
BB		 491 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
492 "property '%s' can only be set during pool "
493 "creation or import"), propname);
494 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
495 goto error;
496 }
497
498 if (strval[0] != '/') {
499 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
500 "bad alternate root '%s'"), strval);
501 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
502 goto error;
503 }
504 break;
505
506 case ZPOOL_PROP_CACHEFILE:
507 if (strval[0] == '\0')
508 break;
509
510 if (strcmp(strval, "none") == 0)
511 break;
512
513 if (strval[0] != '/') {
514 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
515 "property '%s' must be empty, an "
516 "absolute path, or 'none'"), propname);
517 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
518 goto error;
519 }
520
521 slash = strrchr(strval, '/');
522
523 if (slash[1] == '\0' || strcmp(slash, "/.") == 0 ||
524 strcmp(slash, "/..") == 0) {
525 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
526 "'%s' is not a valid file"), strval);
527 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
528 goto error;
529 }
530
531 *slash = '\0';
532
533 if (strval[0] != '\0' &&
534 (stat64(strval, &statbuf) != 0 ||
535 !S_ISDIR(statbuf.st_mode))) {
536 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
537 "'%s' is not a valid directory"),
538 strval);
539 (void) zfs_error(hdl, EZFS_BADPATH, errbuf);
540 goto error;
541 }
542
543 *slash = '/';
544 break;
572e2857
BB		 545
546 case ZPOOL_PROP_READONLY:
547 if (!flags.import) {
548 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
549 "property '%s' can only be set at "
550 "import time"), propname);
551 (void) zfs_error(hdl, EZFS_BADPROP, errbuf);
552 goto error;
553 }
554 break;
34dc7c2f
BB		 555 }
556 }
557
558 return (retprops);
559error:
560 nvlist_free(retprops);
561 return (NULL);
562}
563
564/*
565 * Set zpool property : propname=propval.
566 */
567int
568zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval)
569{
2598c001 570 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
34dc7c2f
BB		 571 int ret = -1;
572 char errbuf[1024];
573 nvlist_t *nvl = NULL;
574 nvlist_t *realprops;
575 uint64_t version;
572e2857 576 prop_flags_t flags = { 0 };
34dc7c2f
BB		 577
578 (void) snprintf(errbuf, sizeof (errbuf),
579 dgettext(TEXT_DOMAIN, "cannot set property for '%s'"),
580 zhp->zpool_name);
581
34dc7c2f
BB		 582 if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0)
583 return (no_memory(zhp->zpool_hdl));
584
585 if (nvlist_add_string(nvl, propname, propval) != 0) {
586 nvlist_free(nvl);
587 return (no_memory(zhp->zpool_hdl));
588 }
589
590 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
b128c09f 591 if ((realprops = zpool_valid_proplist(zhp->zpool_hdl,
572e2857 592 zhp->zpool_name, nvl, version, flags, errbuf)) == NULL) {
34dc7c2f
BB		 593 nvlist_free(nvl);
594 return (-1);
595 }
596
597 nvlist_free(nvl);
598 nvl = realprops;
599
600 /*
601 * Execute the corresponding ioctl() to set this property.
602 */
603 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
604
605 if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) {
606 nvlist_free(nvl);
607 return (-1);
608 }
609
610 ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc);
611
612 zcmd_free_nvlists(&zc);
613 nvlist_free(nvl);
614
615 if (ret)
616 (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf);
617 else
618 (void) zpool_props_refresh(zhp);
619
620 return (ret);
621}
622
623int
624zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp)
625{
626 libzfs_handle_t *hdl = zhp->zpool_hdl;
627 zprop_list_t *entry;
628 char buf[ZFS_MAXPROPLEN];
629
630 if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0)
631 return (-1);
632
633 for (entry = *plp; entry != NULL; entry = entry->pl_next) {
634
635 if (entry->pl_fixed)
636 continue;
637
638 if (entry->pl_prop != ZPROP_INVAL &&
639 zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf),
640 NULL) == 0) {
641 if (strlen(buf) > entry->pl_width)
642 entry->pl_width = strlen(buf);
643 }
644 }
645
646 return (0);
647}
648
649
9babb374
BB		 650/*
651 * Don't start the slice at the default block of 34; many storage
d603ed6c
BB		 652 * devices will use a stripe width of 128k, other vendors prefer a 1m
653 * alignment. It is best to play it safe and ensure a 1m alignment
654 * give 512b blocks. When the block size is larger by a power of 2
655 * we will still be 1m aligned.
9babb374 656 */
d603ed6c 657#define NEW_START_BLOCK 2048
9babb374 658
34dc7c2f
BB		 659/*
660 * Validate the given pool name, optionally putting an extended error message in
661 * 'buf'.
662 */
663boolean_t
664zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool)
665{
666 namecheck_err_t why;
667 char what;
668 int ret;
669
670 ret = pool_namecheck(pool, &why, &what);
671
672 /*
673 * The rules for reserved pool names were extended at a later point.
674 * But we need to support users with existing pools that may now be
675 * invalid. So we only check for this expanded set of names during a
676 * create (or import), and only in userland.
677 */
678 if (ret == 0 && !isopen &&
679 (strncmp(pool, "mirror", 6) == 0 ||
680 strncmp(pool, "raidz", 5) == 0 ||
681 strncmp(pool, "spare", 5) == 0 ||
682 strcmp(pool, "log") == 0)) {
683 if (hdl != NULL)
684 zfs_error_aux(hdl,
685 dgettext(TEXT_DOMAIN, "name is reserved"));
686 return (B_FALSE);
687 }
688
689
690 if (ret != 0) {
691 if (hdl != NULL) {
692 switch (why) {
693 case NAME_ERR_TOOLONG:
694 zfs_error_aux(hdl,
695 dgettext(TEXT_DOMAIN, "name is too long"));
696 break;
697
698 case NAME_ERR_INVALCHAR:
699 zfs_error_aux(hdl,
700 dgettext(TEXT_DOMAIN, "invalid character "
701 "'%c' in pool name"), what);
702 break;
703
704 case NAME_ERR_NOLETTER:
705 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
706 "name must begin with a letter"));
707 break;
708
709 case NAME_ERR_RESERVED:
710 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
711 "name is reserved"));
712 break;
713
714 case NAME_ERR_DISKLIKE:
715 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
716 "pool name is reserved"));
717 break;
718
719 case NAME_ERR_LEADING_SLASH:
720 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
721 "leading slash in name"));
722 break;
723
724 case NAME_ERR_EMPTY_COMPONENT:
725 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
726 "empty component in name"));
727 break;
728
729 case NAME_ERR_TRAILING_SLASH:
730 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
731 "trailing slash in name"));
732 break;
733
734 case NAME_ERR_MULTIPLE_AT:
735 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
736 "multiple '@' delimiters in name"));
737 break;
e75c13c3
BB		 738 case NAME_ERR_NO_AT:
739 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
740 "permission set is missing '@'"));
741 break;
34dc7c2f
BB		 742 }
743 }
744 return (B_FALSE);
745 }
746
747 return (B_TRUE);
748}
749
750/*
751 * Open a handle to the given pool, even if the pool is currently in the FAULTED
752 * state.
753 */
754zpool_handle_t *
755zpool_open_canfail(libzfs_handle_t *hdl, const char *pool)
756{
757 zpool_handle_t *zhp;
758 boolean_t missing;
759
760 /*
761 * Make sure the pool name is valid.
762 */
763 if (!zpool_name_valid(hdl, B_TRUE, pool)) {
764 (void) zfs_error_fmt(hdl, EZFS_INVALIDNAME,
765 dgettext(TEXT_DOMAIN, "cannot open '%s'"),
766 pool);
767 return (NULL);
768 }
769
770 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
771 return (NULL);
772
773 zhp->zpool_hdl = hdl;
774 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
775
776 if (zpool_refresh_stats(zhp, &missing) != 0) {
777 zpool_close(zhp);
778 return (NULL);
779 }
780
781 if (missing) {
782 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool"));
783 (void) zfs_error_fmt(hdl, EZFS_NOENT,
784 dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool);
785 zpool_close(zhp);
786 return (NULL);
787 }
788
789 return (zhp);
790}
791
792/*
793 * Like the above, but silent on error. Used when iterating over pools (because
794 * the configuration cache may be out of date).
795 */
796int
797zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret)
798{
799 zpool_handle_t *zhp;
800 boolean_t missing;
801
802 if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL)
803 return (-1);
804
805 zhp->zpool_hdl = hdl;
806 (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name));
807
808 if (zpool_refresh_stats(zhp, &missing) != 0) {
809 zpool_close(zhp);
810 return (-1);
811 }
812
813 if (missing) {
814 zpool_close(zhp);
815 *ret = NULL;
816 return (0);
817 }
818
819 *ret = zhp;
820 return (0);
821}
822
823/*
824 * Similar to zpool_open_canfail(), but refuses to open pools in the faulted
825 * state.
826 */
827zpool_handle_t *
828zpool_open(libzfs_handle_t *hdl, const char *pool)
829{
830 zpool_handle_t *zhp;
831
832 if ((zhp = zpool_open_canfail(hdl, pool)) == NULL)
833 return (NULL);
834
835 if (zhp->zpool_state == POOL_STATE_UNAVAIL) {
836 (void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
837 dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name);
838 zpool_close(zhp);
839 return (NULL);
840 }
841
842 return (zhp);
843}
844
845/*
846 * Close the handle. Simply frees the memory associated with the handle.
847 */
848void
849zpool_close(zpool_handle_t *zhp)
850{
851 if (zhp->zpool_config)
852 nvlist_free(zhp->zpool_config);
853 if (zhp->zpool_old_config)
854 nvlist_free(zhp->zpool_old_config);
855 if (zhp->zpool_props)
856 nvlist_free(zhp->zpool_props);
857 free(zhp);
858}
859
860/*
861 * Return the name of the pool.
862 */
863const char *
864zpool_get_name(zpool_handle_t *zhp)
865{
866 return (zhp->zpool_name);
867}
868
869
870/*
871 * Return the state of the pool (ACTIVE or UNAVAILABLE)
872 */
873int
874zpool_get_state(zpool_handle_t *zhp)
875{
876 return (zhp->zpool_state);
877}
878
879/*
880 * Create the named pool, using the provided vdev list. It is assumed
881 * that the consumer has already validated the contents of the nvlist, so we
882 * don't have to worry about error semantics.
883 */
884int
885zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot,
b128c09f 886 nvlist_t *props, nvlist_t *fsprops)
34dc7c2f 887{
2598c001 888 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
b128c09f
BB		 889 nvlist_t *zc_fsprops = NULL;
890 nvlist_t *zc_props = NULL;
34dc7c2f
BB		 891 char msg[1024];
892 char *altroot;
b128c09f 893 int ret = -1;
34dc7c2f
BB		 894
895 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
896 "cannot create '%s'"), pool);
897
898 if (!zpool_name_valid(hdl, B_FALSE, pool))
899 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
900
901 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
902 return (-1);
903
b128c09f 904 if (props) {
572e2857
BB		 905 prop_flags_t flags = { .create = B_TRUE, .import = B_FALSE };
906
b128c09f 907 if ((zc_props = zpool_valid_proplist(hdl, pool, props,
572e2857 908 SPA_VERSION_1, flags, msg)) == NULL) {
b128c09f
BB		 909 goto create_failed;
910 }
911 }
34dc7c2f 912
b128c09f
BB		 913 if (fsprops) {
914 uint64_t zoned;
915 char *zonestr;
916
917 zoned = ((nvlist_lookup_string(fsprops,
918 zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) &&
919 strcmp(zonestr, "on") == 0);
920
921 if ((zc_fsprops = zfs_valid_proplist(hdl,
922 ZFS_TYPE_FILESYSTEM, fsprops, zoned, NULL, msg)) == NULL) {
923 goto create_failed;
924 }
925 if (!zc_props &&
926 (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) {
927 goto create_failed;
928 }
929 if (nvlist_add_nvlist(zc_props,
930 ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) {
931 goto create_failed;
932 }
34dc7c2f
BB		 933 }
934
b128c09f
BB		 935 if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
936 goto create_failed;
937
34dc7c2f
BB		 938 (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name));
939
b128c09f 940 if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) {
34dc7c2f
BB		 941
942 zcmd_free_nvlists(&zc);
b128c09f
BB		 943 nvlist_free(zc_props);
944 nvlist_free(zc_fsprops);
34dc7c2f
BB		 945
946 switch (errno) {
947 case EBUSY:
948 /*
949 * This can happen if the user has specified the same
950 * device multiple times. We can't reliably detect this
951 * until we try to add it and see we already have a
d603ed6c
BB		 952 * label. This can also happen under if the device is
953 * part of an active md or lvm device.
34dc7c2f
BB		 954 */
955 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
d603ed6c
BB		 956 "one or more vdevs refer to the same device, or one of\n"
957 "the devices is part of an active md or lvm device"));
34dc7c2f
BB		 958 return (zfs_error(hdl, EZFS_BADDEV, msg));
959
960 case EOVERFLOW:
961 /*
962 * This occurs when one of the devices is below
963 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
964 * device was the problem device since there's no
965 * reliable way to determine device size from userland.
966 */
967 {
968 char buf[64];
969
970 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
971
972 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
973 "one or more devices is less than the "
974 "minimum size (%s)"), buf);
975 }
976 return (zfs_error(hdl, EZFS_BADDEV, msg));
977
978 case ENOSPC:
979 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
980 "one or more devices is out of space"));
981 return (zfs_error(hdl, EZFS_BADDEV, msg));
982
983 case ENOTBLK:
984 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
985 "cache device must be a disk or disk slice"));
986 return (zfs_error(hdl, EZFS_BADDEV, msg));
987
988 default:
989 return (zpool_standard_error(hdl, errno, msg));
990 }
991 }
992
993 /*
994 * If this is an alternate root pool, then we automatically set the
995 * mountpoint of the root dataset to be '/'.
996 */
997 if (nvlist_lookup_string(props, zpool_prop_to_name(ZPOOL_PROP_ALTROOT),
998 &altroot) == 0) {
999 zfs_handle_t *zhp;
1000
1001 verify((zhp = zfs_open(hdl, pool, ZFS_TYPE_DATASET)) != NULL);
1002 verify(zfs_prop_set(zhp, zfs_prop_to_name(ZFS_PROP_MOUNTPOINT),
1003 "/") == 0);
1004
1005 zfs_close(zhp);
1006 }
1007
b128c09f 1008create_failed:
34dc7c2f 1009 zcmd_free_nvlists(&zc);
b128c09f
BB		 1010 nvlist_free(zc_props);
1011 nvlist_free(zc_fsprops);
1012 return (ret);
34dc7c2f
BB		 1013}
1014
1015/*
1016 * Destroy the given pool. It is up to the caller to ensure that there are no
1017 * datasets left in the pool.
1018 */
1019int
1020zpool_destroy(zpool_handle_t *zhp)
1021{
2598c001 1022 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
34dc7c2f
BB		 1023 zfs_handle_t *zfp = NULL;
1024 libzfs_handle_t *hdl = zhp->zpool_hdl;
1025 char msg[1024];
1026
1027 if (zhp->zpool_state == POOL_STATE_ACTIVE &&
572e2857 1028 (zfp = zfs_open(hdl, zhp->zpool_name, ZFS_TYPE_FILESYSTEM)) == NULL)
34dc7c2f
BB		 1029 return (-1);
1030
34dc7c2f
BB		 1031 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1032
572e2857 1033 if (zfs_ioctl(hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) {
34dc7c2f
BB		 1034 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1035 "cannot destroy '%s'"), zhp->zpool_name);
1036
1037 if (errno == EROFS) {
1038 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1039 "one or more devices is read only"));
1040 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1041 } else {
1042 (void) zpool_standard_error(hdl, errno, msg);
1043 }
1044
1045 if (zfp)
1046 zfs_close(zfp);
1047 return (-1);
1048 }
1049
1050 if (zfp) {
1051 remove_mountpoint(zfp);
1052 zfs_close(zfp);
1053 }
1054
1055 return (0);
1056}
1057
1058/*
1059 * Add the given vdevs to the pool. The caller must have already performed the
1060 * necessary verification to ensure that the vdev specification is well-formed.
1061 */
1062int
1063zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot)
1064{
2598c001 1065 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
34dc7c2f
BB		 1066 int ret;
1067 libzfs_handle_t *hdl = zhp->zpool_hdl;
1068 char msg[1024];
1069 nvlist_t **spares, **l2cache;
1070 uint_t nspares, nl2cache;
1071
1072 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1073 "cannot add to '%s'"), zhp->zpool_name);
1074
1075 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1076 SPA_VERSION_SPARES &&
1077 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
1078 &spares, &nspares) == 0) {
1079 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1080 "upgraded to add hot spares"));
1081 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1082 }
1083
b128c09f
BB		 1084 if (pool_is_bootable(zhp) && nvlist_lookup_nvlist_array(nvroot,
1085 ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0) {
1086 uint64_t s;
1087
1088 for (s = 0; s < nspares; s++) {
1089 char *path;
1090
1091 if (nvlist_lookup_string(spares[s], ZPOOL_CONFIG_PATH,
1092 &path) == 0 && pool_uses_efi(spares[s])) {
1093 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1094 "device '%s' contains an EFI label and "
1095 "cannot be used on root pools."),
428870ff
BB		 1096 zpool_vdev_name(hdl, NULL, spares[s],
1097 B_FALSE));
b128c09f
BB		 1098 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg));
1099 }
1100 }
1101 }
1102
34dc7c2f
BB		 1103 if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) <
1104 SPA_VERSION_L2CACHE &&
1105 nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
1106 &l2cache, &nl2cache) == 0) {
1107 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be "
1108 "upgraded to add cache devices"));
1109 return (zfs_error(hdl, EZFS_BADVERSION, msg));
1110 }
1111
1112 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
1113 return (-1);
1114 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
1115
572e2857 1116 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) {
34dc7c2f
BB		 1117 switch (errno) {
1118 case EBUSY:
1119 /*
1120 * This can happen if the user has specified the same
1121 * device multiple times. We can't reliably detect this
1122 * until we try to add it and see we already have a
1123 * label.
1124 */
1125 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1126 "one or more vdevs refer to the same device"));
1127 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1128 break;
1129
1130 case EOVERFLOW:
1131 /*
1132 * This occurrs when one of the devices is below
1133 * SPA_MINDEVSIZE. Unfortunately, we can't detect which
1134 * device was the problem device since there's no
1135 * reliable way to determine device size from userland.
1136 */
1137 {
1138 char buf[64];
1139
1140 zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf));
1141
1142 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1143 "device is less than the minimum "
1144 "size (%s)"), buf);
1145 }
1146 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1147 break;
1148
1149 case ENOTSUP:
1150 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1151 "pool must be upgraded to add these vdevs"));
1152 (void) zfs_error(hdl, EZFS_BADVERSION, msg);
1153 break;
1154
1155 case EDOM:
1156 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1157 "root pool can not have multiple vdevs"
1158 " or separate logs"));
1159 (void) zfs_error(hdl, EZFS_POOL_NOTSUP, msg);
1160 break;
1161
1162 case ENOTBLK:
1163 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1164 "cache device must be a disk or disk slice"));
1165 (void) zfs_error(hdl, EZFS_BADDEV, msg);
1166 break;
1167
1168 default:
1169 (void) zpool_standard_error(hdl, errno, msg);
1170 }
1171
1172 ret = -1;
1173 } else {
1174 ret = 0;
1175 }
1176
1177 zcmd_free_nvlists(&zc);
1178
1179 return (ret);
1180}
1181
1182/*
1183 * Exports the pool from the system. The caller must ensure that there are no
1184 * mounted datasets in the pool.
1185 */
1186int
fb5f0bc8 1187zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce)
34dc7c2f 1188{
2598c001 1189 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
b128c09f 1190 char msg[1024];
34dc7c2f 1191
b128c09f
BB		 1192 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
1193 "cannot export '%s'"), zhp->zpool_name);
1194
34dc7c2f 1195 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
b128c09f 1196 zc.zc_cookie = force;
fb5f0bc8 1197 zc.zc_guid = hardforce;
b128c09f
BB		 1198
1199 if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) {
1200 switch (errno) {
1201 case EXDEV:
1202 zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN,
1203 "use '-f' to override the following errors:\n"
1204 "'%s' has an active shared spare which could be"
1205 " used by other pools once '%s' is exported."),
1206 zhp->zpool_name, zhp->zpool_name);
1207 return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE,
1208 msg));
1209 default:
1210 return (zpool_standard_error_fmt(zhp->zpool_hdl, errno,
1211 msg));
1212 }
1213 }
34dc7c2f 1214
34dc7c2f
BB		 1215 return (0);
1216}
1217
fb5f0bc8
BB		 1218int
1219zpool_export(zpool_handle_t *zhp, boolean_t force)
1220{
1221 return (zpool_export_common(zhp, force, B_FALSE));
1222}
1223
1224int
1225zpool_export_force(zpool_handle_t *zhp)
1226{
1227 return (zpool_export_common(zhp, B_TRUE, B_TRUE));
1228}
1229
428870ff
BB		 1230static void
1231zpool_rewind_exclaim(libzfs_handle_t *hdl, const char *name, boolean_t dryrun,
572e2857 1232 nvlist_t *config)
428870ff 1233{
572e2857 1234 nvlist_t *nv = NULL;
428870ff
BB		 1235 uint64_t rewindto;
1236 int64_t loss = -1;
1237 struct tm t;
1238 char timestr[128];
1239
572e2857
BB		 1240 if (!hdl->libzfs_printerr || config == NULL)
1241 return;
1242
1243 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0)
428870ff
BB		 1244 return;
1245
572e2857 1246 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
428870ff 1247 return;
572e2857 1248 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
428870ff
BB		 1249
1250 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
b8864a23 1251 strftime(timestr, 128, "%c", &t) != 0) {
428870ff
BB		 1252 if (dryrun) {
1253 (void) printf(dgettext(TEXT_DOMAIN,
1254 "Would be able to return %s "
1255 "to its state as of %s.\n"),
1256 name, timestr);
1257 } else {
1258 (void) printf(dgettext(TEXT_DOMAIN,
1259 "Pool %s returned to its state as of %s.\n"),
1260 name, timestr);
1261 }
1262 if (loss > 120) {
1263 (void) printf(dgettext(TEXT_DOMAIN,
1264 "%s approximately %lld "),
1265 dryrun ? "Would discard" : "Discarded",
b8864a23 1266 ((longlong_t)loss + 30) / 60);
428870ff
BB		 1267 (void) printf(dgettext(TEXT_DOMAIN,
1268 "minutes of transactions.\n"));
1269 } else if (loss > 0) {
1270 (void) printf(dgettext(TEXT_DOMAIN,
1271 "%s approximately %lld "),
b8864a23
BB		 1272 dryrun ? "Would discard" : "Discarded",
1273 (longlong_t)loss);
428870ff
BB		 1274 (void) printf(dgettext(TEXT_DOMAIN,
1275 "seconds of transactions.\n"));
1276 }
1277 }
1278}
1279
1280void
1281zpool_explain_recover(libzfs_handle_t *hdl, const char *name, int reason,
1282 nvlist_t *config)
1283{
572e2857 1284 nvlist_t *nv = NULL;
428870ff
BB		 1285 int64_t loss = -1;
1286 uint64_t edata = UINT64_MAX;
1287 uint64_t rewindto;
1288 struct tm t;
1289 char timestr[128];
1290
1291 if (!hdl->libzfs_printerr)
1292 return;
1293
1294 if (reason >= 0)
1295 (void) printf(dgettext(TEXT_DOMAIN, "action: "));
1296 else
1297 (void) printf(dgettext(TEXT_DOMAIN, "\t"));
1298
1299 /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */
572e2857
BB		 1300 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 ||
1301 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0)
428870ff
BB		 1302 goto no_info;
1303
572e2857
BB		 1304 (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss);
1305 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_DATA_ERRORS,
428870ff
BB		 1306 &edata);
1307
1308 (void) printf(dgettext(TEXT_DOMAIN,
1309 "Recovery is possible, but will result in some data loss.\n"));
1310
1311 if (localtime_r((time_t *)&rewindto, &t) != NULL &&
b8864a23 1312 strftime(timestr, 128, "%c", &t) != 0) {
428870ff
BB		 1313 (void) printf(dgettext(TEXT_DOMAIN,
1314 "\tReturning the pool to its state as of %s\n"
1315 "\tshould correct the problem. "),
1316 timestr);
1317 } else {
1318 (void) printf(dgettext(TEXT_DOMAIN,
1319 "\tReverting the pool to an earlier state "
1320 "should correct the problem.\n\t"));
1321 }
1322
1323 if (loss > 120) {
1324 (void) printf(dgettext(TEXT_DOMAIN,
1325 "Approximately %lld minutes of data\n"
b8864a23
BB		 1326 "\tmust be discarded, irreversibly. "),
1327 ((longlong_t)loss + 30) / 60);
428870ff
BB		 1328 } else if (loss > 0) {
1329 (void) printf(dgettext(TEXT_DOMAIN,
1330 "Approximately %lld seconds of data\n"
b8864a23
BB		 1331 "\tmust be discarded, irreversibly. "),
1332 (longlong_t)loss);
428870ff
BB		 1333 }
1334 if (edata != 0 && edata != UINT64_MAX) {
1335 if (edata == 1) {
1336 (void) printf(dgettext(TEXT_DOMAIN,
1337 "After rewind, at least\n"
1338 "\tone persistent user-data error will remain. "));
1339 } else {
1340 (void) printf(dgettext(TEXT_DOMAIN,
1341 "After rewind, several\n"
1342 "\tpersistent user-data errors will remain. "));
1343 }
1344 }
1345 (void) printf(dgettext(TEXT_DOMAIN,
1346 "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "),
1347 reason >= 0 ? "clear" : "import", name);
1348
1349 (void) printf(dgettext(TEXT_DOMAIN,
1350 "A scrub of the pool\n"
1351 "\tis strongly recommended after recovery.\n"));
1352 return;
1353
1354no_info:
1355 (void) printf(dgettext(TEXT_DOMAIN,
1356 "Destroy and re-create the pool from\n\ta backup source.\n"));
1357}
1358
34dc7c2f
BB		 1359/*
1360 * zpool_import() is a contracted interface. Should be kept the same
1361 * if possible.
1362 *
1363 * Applications should use zpool_import_props() to import a pool with
1364 * new properties value to be set.
1365 */
1366int
1367zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
1368 char *altroot)
1369{
1370 nvlist_t *props = NULL;
1371 int ret;
1372
1373 if (altroot != NULL) {
1374 if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) {
1375 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1376 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1377 newname));
1378 }
1379
1380 if (nvlist_add_string(props,
fb5f0bc8
BB		 1381 zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 ||
1382 nvlist_add_string(props,
1383 zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) {
34dc7c2f
BB		 1384 nvlist_free(props);
1385 return (zfs_error_fmt(hdl, EZFS_NOMEM,
1386 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1387 newname));
1388 }
1389 }
1390
572e2857
BB		 1391 ret = zpool_import_props(hdl, config, newname, props,
1392 ZFS_IMPORT_NORMAL);
34dc7c2f
BB		 1393 if (props)
1394 nvlist_free(props);
1395 return (ret);
1396}
1397
572e2857
BB		 1398static void
1399print_vdev_tree(libzfs_handle_t *hdl, const char *name, nvlist_t *nv,
1400 int indent)
1401{
1402 nvlist_t **child;
1403 uint_t c, children;
1404 char *vname;
1405 uint64_t is_log = 0;
1406
1407 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG,
1408 &is_log);
1409
1410 if (name != NULL)
1411 (void) printf("\t%*s%s%s\n", indent, "", name,
1412 is_log ? " [log]" : "");
1413
1414 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1415 &child, &children) != 0)
1416 return;
1417
1418 for (c = 0; c < children; c++) {
1419 vname = zpool_vdev_name(hdl, NULL, child[c], B_TRUE);
1420 print_vdev_tree(hdl, vname, child[c], indent + 2);
1421 free(vname);
1422 }
1423}
1424
34dc7c2f
BB		 1425/*
1426 * Import the given pool using the known configuration and a list of
1427 * properties to be set. The configuration should have come from
1428 * zpool_find_import(). The 'newname' parameters control whether the pool
1429 * is imported with a different name.
1430 */
1431int
1432zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname,
572e2857 1433 nvlist_t *props, int flags)
34dc7c2f 1434{
2598c001 1435 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
428870ff 1436 zpool_rewind_policy_t policy;
572e2857
BB		 1437 nvlist_t *nv = NULL;
1438 nvlist_t *nvinfo = NULL;
1439 nvlist_t *missing = NULL;
34dc7c2f
BB		 1440 char *thename;
1441 char *origname;
1442 int ret;
572e2857 1443 int error = 0;
34dc7c2f
BB		 1444 char errbuf[1024];
1445
1446 verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME,
1447 &origname) == 0);
1448
1449 (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
1450 "cannot import pool '%s'"), origname);
1451
1452 if (newname != NULL) {
1453 if (!zpool_name_valid(hdl, B_FALSE, newname))
1454 return (zfs_error_fmt(hdl, EZFS_INVALIDNAME,
1455 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1456 newname));
1457 thename = (char *)newname;
1458 } else {
1459 thename = origname;
1460 }
1461
1462 if (props) {
1463 uint64_t version;
572e2857 1464 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
34dc7c2f
BB		 1465
1466 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION,
1467 &version) == 0);
1468
b128c09f 1469 if ((props = zpool_valid_proplist(hdl, origname,
572e2857 1470 props, version, flags, errbuf)) == NULL) {
34dc7c2f
BB		 1471 return (-1);
1472 } else if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) {
1473 nvlist_free(props);
1474 return (-1);
1475 }
1476 }
1477
1478 (void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name));
1479
1480 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID,
1481 &zc.zc_guid) == 0);
1482
1483 if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) {
1484 nvlist_free(props);
1485 return (-1);
1486 }
572e2857 1487 if (zcmd_alloc_dst_nvlist(hdl, &zc, zc.zc_nvlist_conf_size * 2) != 0) {
428870ff
BB		 1488 nvlist_free(props);
1489 return (-1);
1490 }
34dc7c2f 1491
572e2857
BB		 1492 zc.zc_cookie = flags;
1493 while ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc)) != 0 &&
1494 errno == ENOMEM) {
1495 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
1496 zcmd_free_nvlists(&zc);
1497 return (-1);
1498 }
1499 }
1500 if (ret != 0)
1501 error = errno;
1502
1503 (void) zcmd_read_dst_nvlist(hdl, &zc, &nv);
1504 zpool_get_rewind_policy(config, &policy);
1505
1506 if (error) {
34dc7c2f 1507 char desc[1024];
428870ff 1508
428870ff
BB		 1509 /*
1510 * Dry-run failed, but we print out what success
1511 * looks like if we found a best txg
1512 */
572e2857 1513 if (policy.zrp_request & ZPOOL_TRY_REWIND) {
428870ff 1514 zpool_rewind_exclaim(hdl, newname ? origname : thename,
572e2857
BB		 1515 B_TRUE, nv);
1516 nvlist_free(nv);
428870ff
BB		 1517 return (-1);
1518 }
1519
34dc7c2f
BB		 1520 if (newname == NULL)
1521 (void) snprintf(desc, sizeof (desc),
1522 dgettext(TEXT_DOMAIN, "cannot import '%s'"),
1523 thename);
1524 else
1525 (void) snprintf(desc, sizeof (desc),
1526 dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"),
1527 origname, thename);
1528
572e2857 1529 switch (error) {
34dc7c2f
BB		 1530 case ENOTSUP:
1531 /*
1532 * Unsupported version.
1533 */
1534 (void) zfs_error(hdl, EZFS_BADVERSION, desc);
1535 break;
1536
1537 case EINVAL:
1538 (void) zfs_error(hdl, EZFS_INVALCONFIG, desc);
1539 break;
1540
428870ff
BB		 1541 case EROFS:
1542 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
1543 "one or more devices is read only"));
1544 (void) zfs_error(hdl, EZFS_BADDEV, desc);
1545 break;
1546
572e2857
BB		 1547 case ENXIO:
1548 if (nv && nvlist_lookup_nvlist(nv,
1549 ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 &&
1550 nvlist_lookup_nvlist(nvinfo,
1551 ZPOOL_CONFIG_MISSING_DEVICES, &missing) == 0) {
1552 (void) printf(dgettext(TEXT_DOMAIN,
1553 "The devices below are missing, use "
1554 "'-m' to import the pool anyway:\n"));
1555 print_vdev_tree(hdl, NULL, missing, 2);
1556 (void) printf("\n");
1557 }
1558 (void) zpool_standard_error(hdl, error, desc);
1559 break;
1560
1561 case EEXIST:
1562 (void) zpool_standard_error(hdl, error, desc);
1563 break;
1564
34dc7c2f 1565 default:
572e2857 1566 (void) zpool_standard_error(hdl, error, desc);
428870ff 1567 zpool_explain_recover(hdl,
572e2857 1568 newname ? origname : thename, -error, nv);
428870ff 1569 break;
34dc7c2f
BB		 1570 }
1571
572e2857 1572 nvlist_free(nv);
34dc7c2f
BB		 1573 ret = -1;
1574 } else {
1575 zpool_handle_t *zhp;
1576
1577 /*
1578 * This should never fail, but play it safe anyway.
1579 */
428870ff 1580 if (zpool_open_silent(hdl, thename, &zhp) != 0)
34dc7c2f 1581 ret = -1;
428870ff 1582 else if (zhp != NULL)
34dc7c2f 1583 zpool_close(zhp);
428870ff
BB		 1584 if (policy.zrp_request &
1585 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
1586 zpool_rewind_exclaim(hdl, newname ? origname : thename,
572e2857 1587 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0), nv);
34dc7c2f 1588 }
572e2857 1589 nvlist_free(nv);
428870ff 1590 return (0);
34dc7c2f
BB		 1591 }
1592
1593 zcmd_free_nvlists(&zc);
1594 nvlist_free(props);
1595
1596 return (ret);
1597}
1598
1599/*
428870ff 1600 * Scan the pool.
34dc7c2f
BB		 1601 */
1602int
428870ff 1603zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func)
34dc7c2f 1604{
2598c001 1605 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
34dc7c2f
BB		 1606 char msg[1024];
1607 libzfs_handle_t *hdl = zhp->zpool_hdl;
1608
1609 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
428870ff 1610 zc.zc_cookie = func;
34dc7c2f 1611
572e2857 1612 if (zfs_ioctl(hdl, ZFS_IOC_POOL_SCAN, &zc) == 0 ||
428870ff 1613 (errno == ENOENT && func != POOL_SCAN_NONE))
34dc7c2f
BB		 1614 return (0);
1615
428870ff
BB		 1616 if (func == POOL_SCAN_SCRUB) {
1617 (void) snprintf(msg, sizeof (msg),
1618 dgettext(TEXT_DOMAIN, "cannot scrub %s"), zc.zc_name);
1619 } else if (func == POOL_SCAN_NONE) {
1620 (void) snprintf(msg, sizeof (msg),
1621 dgettext(TEXT_DOMAIN, "cannot cancel scrubbing %s"),
1622 zc.zc_name);
1623 } else {
1624 assert(!"unexpected result");
1625 }
34dc7c2f 1626
428870ff
BB		 1627 if (errno == EBUSY) {
1628 nvlist_t *nvroot;
1629 pool_scan_stat_t *ps = NULL;
1630 uint_t psc;
1631
1632 verify(nvlist_lookup_nvlist(zhp->zpool_config,
1633 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
1634 (void) nvlist_lookup_uint64_array(nvroot,
1635 ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc);
1636 if (ps && ps->pss_func == POOL_SCAN_SCRUB)
1637 return (zfs_error(hdl, EZFS_SCRUBBING, msg));
1638 else
1639 return (zfs_error(hdl, EZFS_RESILVERING, msg));
1640 } else if (errno == ENOENT) {
1641 return (zfs_error(hdl, EZFS_NO_SCRUB, msg));
1642 } else {
34dc7c2f 1643 return (zpool_standard_error(hdl, errno, msg));
428870ff
BB		 1644 }
1645}
1646
34dc7c2f 1647/*
9babb374
BB		 1648 * Find a vdev that matches the search criteria specified. We use the
1649 * the nvpair name to determine how we should look for the device.
34dc7c2f
BB		 1650 * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL
1651 * spare; but FALSE if its an INUSE spare.
1652 */
1653static nvlist_t *
9babb374
BB		 1654vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare,
1655 boolean_t *l2cache, boolean_t *log)
34dc7c2f
BB		 1656{
1657 uint_t c, children;
1658 nvlist_t **child;
34dc7c2f 1659 nvlist_t *ret;
b128c09f 1660 uint64_t is_log;
9babb374
BB		 1661 char *srchkey;
1662 nvpair_t *pair = nvlist_next_nvpair(search, NULL);
1663
1664 /* Nothing to look for */
1665 if (search == NULL || pair == NULL)
1666 return (NULL);
1667
1668 /* Obtain the key we will use to search */
1669 srchkey = nvpair_name(pair);
1670
1671 switch (nvpair_type(pair)) {
572e2857 1672 case DATA_TYPE_UINT64:
9babb374 1673 if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) {
572e2857
BB		 1674 uint64_t srchval, theguid;
1675
1676 verify(nvpair_value_uint64(pair, &srchval) == 0);
1677 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
1678 &theguid) == 0);
1679 if (theguid == srchval)
1680 return (nv);
9babb374
BB		 1681 }
1682 break;
9babb374
BB		 1683
1684 case DATA_TYPE_STRING: {
1685 char *srchval, *val;
1686
1687 verify(nvpair_value_string(pair, &srchval) == 0);
1688 if (nvlist_lookup_string(nv, srchkey, &val) != 0)
1689 break;
34dc7c2f 1690
9babb374 1691 /*
428870ff
BB		 1692 * Search for the requested value. Special cases:
1693 *
a2c6816c
NB		 1694 * - ZPOOL_CONFIG_PATH for whole disk entries. These end in with a
1695 * partition suffix "1", "-part1", or "p1". The suffix is hidden
1696 * from the user, but included in the string, so this matches around
1697 * it.
428870ff
BB		 1698 * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE).
1699 *
1700 * Otherwise, all other searches are simple string compares.
9babb374 1701 */
a2c6816c 1702 if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0) {
9babb374
BB		 1703 uint64_t wholedisk = 0;
1704
1705 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
1706 &wholedisk);
1707 if (wholedisk) {
a2c6816c 1708 char buf[MAXPATHLEN];
428870ff 1709
a2c6816c
NB		 1710 zfs_append_partition(srchval, buf, sizeof (buf));
1711 if (strcmp(val, buf) == 0)
428870ff
BB		 1712 return (nv);
1713
9babb374
BB		 1714 break;
1715 }
428870ff
BB		 1716 } else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0 && val) {
1717 char *type, *idx, *end, *p;
1718 uint64_t id, vdev_id;
1719
1720 /*
1721 * Determine our vdev type, keeping in mind
1722 * that the srchval is composed of a type and
1723 * vdev id pair (i.e. mirror-4).
1724 */
1725 if ((type = strdup(srchval)) == NULL)
1726 return (NULL);
1727
1728 if ((p = strrchr(type, '-')) == NULL) {
1729 free(type);
1730 break;
1731 }
1732 idx = p + 1;
1733 *p = '\0';
1734
1735 /*
1736 * If the types don't match then keep looking.
1737 */
1738 if (strncmp(val, type, strlen(val)) != 0) {
1739 free(type);
1740 break;
1741 }
1742
1743 verify(strncmp(type, VDEV_TYPE_RAIDZ,
1744 strlen(VDEV_TYPE_RAIDZ)) == 0 ||
1745 strncmp(type, VDEV_TYPE_MIRROR,
1746 strlen(VDEV_TYPE_MIRROR)) == 0);
1747 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
1748 &id) == 0);
1749
1750 errno = 0;
1751 vdev_id = strtoull(idx, &end, 10);
1752
1753 free(type);
1754 if (errno != 0)
1755 return (NULL);
1756
1757 /*
1758 * Now verify that we have the correct vdev id.
1759 */
1760 if (vdev_id == id)
1761 return (nv);
9babb374 1762 }
34dc7c2f 1763
34dc7c2f 1764 /*
9babb374 1765 * Common case
34dc7c2f 1766 */
9babb374 1767 if (strcmp(srchval, val) == 0)
34dc7c2f 1768 return (nv);
9babb374
BB		 1769 break;
1770 }
1771
1772 default:
1773 break;
34dc7c2f
BB		 1774 }
1775
1776 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1777 &child, &children) != 0)
1778 return (NULL);
1779
b128c09f 1780 for (c = 0; c < children; c++) {
9babb374 1781 if ((ret = vdev_to_nvlist_iter(child[c], search,
b128c09f
BB		 1782 avail_spare, l2cache, NULL)) != NULL) {
1783 /*
1784 * The 'is_log' value is only set for the toplevel
1785 * vdev, not the leaf vdevs. So we always lookup the
1786 * log device from the root of the vdev tree (where
1787 * 'log' is non-NULL).
1788 */
1789 if (log != NULL &&
1790 nvlist_lookup_uint64(child[c],
1791 ZPOOL_CONFIG_IS_LOG, &is_log) == 0 &&
1792 is_log) {
1793 *log = B_TRUE;
1794 }
34dc7c2f 1795 return (ret);
b128c09f
BB		 1796 }
1797 }
34dc7c2f
BB		 1798
1799 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
1800 &child, &children) == 0) {
1801 for (c = 0; c < children; c++) {
9babb374 1802 if ((ret = vdev_to_nvlist_iter(child[c], search,
b128c09f 1803 avail_spare, l2cache, NULL)) != NULL) {
34dc7c2f
BB		 1804 *avail_spare = B_TRUE;
1805 return (ret);
1806 }
1807 }
1808 }
1809
1810 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
1811 &child, &children) == 0) {
1812 for (c = 0; c < children; c++) {
9babb374 1813 if ((ret = vdev_to_nvlist_iter(child[c], search,
b128c09f 1814 avail_spare, l2cache, NULL)) != NULL) {
34dc7c2f
BB		 1815 *l2cache = B_TRUE;
1816 return (ret);
1817 }
1818 }
1819 }
1820
1821 return (NULL);
1822}
1823
9babb374
BB		 1824/*
1825 * Given a physical path (minus the "/devices" prefix), find the
1826 * associated vdev.
1827 */
1828nvlist_t *
1829zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath,
1830 boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log)
1831{
1832 nvlist_t *search, *nvroot, *ret;
1833
1834 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1835 verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath) == 0);
1836
1837 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
1838 &nvroot) == 0);
1839
1840 *avail_spare = B_FALSE;
572e2857
BB		 1841 *l2cache = B_FALSE;
1842 if (log != NULL)
1843 *log = B_FALSE;
9babb374
BB		 1844 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
1845 nvlist_free(search);
1846
1847 return (ret);
1848}
1849
428870ff
BB		 1850/*
1851 * Determine if we have an "interior" top-level vdev (i.e mirror/raidz).
1852 */
1853boolean_t
1854zpool_vdev_is_interior(const char *name)
1855{
1856 if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 ||
1857 strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0)
1858 return (B_TRUE);
1859 return (B_FALSE);
1860}
1861
34dc7c2f
BB		 1862nvlist_t *
1863zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare,
b128c09f 1864 boolean_t *l2cache, boolean_t *log)
34dc7c2f
BB		 1865{
1866 char buf[MAXPATHLEN];
34dc7c2f 1867 char *end;
9babb374 1868 nvlist_t *nvroot, *search, *ret;
34dc7c2f
BB		 1869 uint64_t guid;
1870
9babb374
BB		 1871 verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0);
1872
34dc7c2f
BB		 1873 guid = strtoull(path, &end, 10);
1874 if (guid != 0 && *end == '\0') {
9babb374 1875 verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0);
428870ff
BB		 1876 } else if (zpool_vdev_is_interior(path)) {
1877 verify(nvlist_add_string(search, ZPOOL_CONFIG_TYPE, path) == 0);
34dc7c2f 1878 } else if (path[0] != '/') {
a2c6816c
NB		 1879 if (zfs_resolve_shortname(path, buf, sizeof (buf)) < 0) {
1880 nvlist_free(search);
1881 return (NULL);
1882 }
9babb374 1883 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, buf) == 0);
34dc7c2f 1884 } else {
9babb374 1885 verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0);
34dc7c2f
BB		 1886 }
1887
1888 verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE,
1889 &nvroot) == 0);
1890
1891 *avail_spare = B_FALSE;
1892 *l2cache = B_FALSE;
b128c09f
BB		 1893 if (log != NULL)
1894 *log = B_FALSE;
9babb374
BB		 1895 ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log);
1896 nvlist_free(search);
1897
1898 return (ret);
b128c09f
BB		 1899}
1900
1901static int
1902vdev_online(nvlist_t *nv)
1903{
1904 uint64_t ival;
1905
1906 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 ||
1907 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 ||
1908 nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0)
1909 return (0);
1910
1911 return (1);
1912}
1913
1914/*
9babb374 1915 * Helper function for zpool_get_physpaths().
b128c09f 1916 */
9babb374
BB		 1917static int
1918vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size,
1919 size_t *bytes_written)
1920{
1921 size_t bytes_left, pos, rsz;
1922 char *tmppath;
1923 const char *format;
1924
1925 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH,
1926 &tmppath) != 0)
1927 return (EZFS_NODEVICE);
1928
1929 pos = *bytes_written;
1930 bytes_left = physpath_size - pos;
1931 format = (pos == 0) ? "%s" : " %s";
1932
1933 rsz = snprintf(physpath + pos, bytes_left, format, tmppath);
1934 *bytes_written += rsz;
1935
1936 if (rsz >= bytes_left) {
1937 /* if physpath was not copied properly, clear it */
1938 if (bytes_left != 0) {
1939 physpath[pos] = 0;
1940 }
1941 return (EZFS_NOSPC);
1942 }
1943 return (0);
1944}
1945
1946static int
1947vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size,
1948 size_t *rsz, boolean_t is_spare)
1949{
1950 char *type;
1951 int ret;
1952
1953 if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0)
1954 return (EZFS_INVALCONFIG);
1955
1956 if (strcmp(type, VDEV_TYPE_DISK) == 0) {
1957 /*
1958 * An active spare device has ZPOOL_CONFIG_IS_SPARE set.
1959 * For a spare vdev, we only want to boot from the active
1960 * spare device.
1961 */
1962 if (is_spare) {
1963 uint64_t spare = 0;
1964 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE,
1965 &spare);
1966 if (!spare)
1967 return (EZFS_INVALCONFIG);
1968 }
1969
1970 if (vdev_online(nv)) {
1971 if ((ret = vdev_get_one_physpath(nv, physpath,
1972 phypath_size, rsz)) != 0)
1973 return (ret);
1974 }
1975 } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 ||
1976 strcmp(type, VDEV_TYPE_REPLACING) == 0 ||
1977 (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) {
1978 nvlist_t **child;
1979 uint_t count;
1980 int i, ret;
1981
1982 if (nvlist_lookup_nvlist_array(nv,
1983 ZPOOL_CONFIG_CHILDREN, &child, &count) != 0)
1984 return (EZFS_INVALCONFIG);
1985
1986 for (i = 0; i < count; i++) {
1987 ret = vdev_get_physpaths(child[i], physpath,
1988 phypath_size, rsz, is_spare);
1989 if (ret == EZFS_NOSPC)
1990 return (ret);
1991 }
1992 }
1993
1994 return (EZFS_POOL_INVALARG);
1995}
1996
1997/*
1998 * Get phys_path for a root pool config.
1999 * Return 0 on success; non-zero on failure.
2000 */
2001static int
2002zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size)
b128c09f 2003{
9babb374 2004 size_t rsz;
b128c09f
BB		 2005 nvlist_t *vdev_root;
2006 nvlist_t **child;
2007 uint_t count;
9babb374 2008 char *type;
b128c09f 2009
9babb374 2010 rsz = 0;
b128c09f 2011
9babb374
BB		 2012 if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
2013 &vdev_root) != 0)
2014 return (EZFS_INVALCONFIG);
b128c09f 2015
9babb374
BB		 2016 if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 ||
2017 nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN,
b128c09f 2018 &child, &count) != 0)
9babb374 2019 return (EZFS_INVALCONFIG);
b128c09f 2020
9babb374
BB		 2021 /*
2022 * root pool can not have EFI labeled disks and can only have
2023 * a single top-level vdev.
2024 */
2025 if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1 ||
2026 pool_uses_efi(vdev_root))
2027 return (EZFS_POOL_INVALARG);
b128c09f 2028
9babb374
BB		 2029 (void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz,
2030 B_FALSE);
2031
2032 /* No online devices */
2033 if (rsz == 0)
2034 return (EZFS_NODEVICE);
b128c09f
BB		 2035
2036 return (0);
34dc7c2f
BB		 2037}
2038
9babb374
BB		 2039/*
2040 * Get phys_path for a root pool
2041 * Return 0 on success; non-zero on failure.
2042 */
2043int
2044zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size)
2045{
2046 return (zpool_get_config_physpath(zhp->zpool_config, physpath,
2047 phypath_size));
2048}
2049
9babb374
BB		 2050/*
2051 * If the device has being dynamically expanded then we need to relabel
2052 * the disk to use the new unallocated space.
2053 */
2054static int
d603ed6c 2055zpool_relabel_disk(libzfs_handle_t *hdl, const char *path)
9babb374 2056{
9babb374
BB		 2057 char errbuf[1024];
2058 int fd, error;
9babb374 2059
d603ed6c 2060 if ((fd = open(path, O_RDWR|O_DIRECT)) < 0) {
9babb374 2061 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
d603ed6c 2062 "relabel '%s': unable to open device"), path);
9babb374
BB		 2063 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
2064 }
2065
2066 /*
2067 * It's possible that we might encounter an error if the device
2068 * does not have any unallocated space left. If so, we simply
2069 * ignore that error and continue on.
2070 */
d603ed6c 2071 error = efi_use_whole_disk(fd);
9babb374
BB		 2072 (void) close(fd);
2073 if (error && error != VT_ENOSPC) {
2074 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot "
d603ed6c 2075 "relabel '%s': unable to read disk capacity"), path);
9babb374
BB		 2076 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
2077 }
2078 return (0);
2079}
2080
34dc7c2f
BB		 2081/*
2082 * Bring the specified vdev online. The 'flags' parameter is a set of the
2083 * ZFS_ONLINE_* flags.
2084 */
2085int
2086zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags,
2087 vdev_state_t *newstate)
2088{
2598c001 2089 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
34dc7c2f
BB		 2090 char msg[1024];
2091 nvlist_t *tgt;
9babb374 2092 boolean_t avail_spare, l2cache, islog;
34dc7c2f
BB		 2093 libzfs_handle_t *hdl = zhp->zpool_hdl;
2094
9babb374
BB		 2095 if (flags & ZFS_ONLINE_EXPAND) {
2096 (void) snprintf(msg, sizeof (msg),
2097 dgettext(TEXT_DOMAIN, "cannot expand %s"), path);
2098 } else {
2099 (void) snprintf(msg, sizeof (msg),
2100 dgettext(TEXT_DOMAIN, "cannot online %s"), path);
2101 }
34dc7c2f
BB		 2102
2103 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
b128c09f 2104 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
9babb374 2105 &islog)) == NULL)
34dc7c2f
BB		 2106 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2107
2108 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2109
428870ff 2110 if (avail_spare)
34dc7c2f
BB		 2111 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2112
9babb374
BB		 2113 if (flags & ZFS_ONLINE_EXPAND ||
2114 zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) {
2115 char *pathname = NULL;
2116 uint64_t wholedisk = 0;
2117
2118 (void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK,
2119 &wholedisk);
2120 verify(nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH,
2121 &pathname) == 0);
2122
2123 /*
2124 * XXX - L2ARC 1.0 devices can't support expansion.
2125 */
2126 if (l2cache) {
2127 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2128 "cannot expand cache devices"));
2129 return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg));
2130 }
2131
2132 if (wholedisk) {
2133 pathname += strlen(DISK_ROOT) + 1;
572e2857 2134 (void) zpool_relabel_disk(hdl, pathname);
9babb374
BB		 2135 }
2136 }
2137
34dc7c2f
BB		 2138 zc.zc_cookie = VDEV_STATE_ONLINE;
2139 zc.zc_obj = flags;
2140
572e2857 2141 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) {
428870ff
BB		 2142 if (errno == EINVAL) {
2143 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split "
2144 "from this pool into a new one. Use '%s' "
2145 "instead"), "zpool detach");
2146 return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, msg));
2147 }
34dc7c2f 2148 return (zpool_standard_error(hdl, errno, msg));
428870ff 2149 }
34dc7c2f
BB		 2150
2151 *newstate = zc.zc_cookie;
2152 return (0);
2153}
2154
2155/*
2156 * Take the specified vdev offline
2157 */
2158int
2159zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp)
2160{
2598c001 2161 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
34dc7c2f
BB		 2162 char msg[1024];
2163 nvlist_t *tgt;
2164 boolean_t avail_spare, l2cache;
2165 libzfs_handle_t *hdl = zhp->zpool_hdl;
2166
2167 (void) snprintf(msg, sizeof (msg),
2168 dgettext(TEXT_DOMAIN, "cannot offline %s"), path);
2169
2170 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
b128c09f
BB		 2171 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2172 NULL)) == NULL)
34dc7c2f
BB		 2173 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2174
2175 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2176
428870ff 2177 if (avail_spare)
34dc7c2f
BB		 2178 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2179
34dc7c2f
BB		 2180 zc.zc_cookie = VDEV_STATE_OFFLINE;
2181 zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0;
2182
572e2857 2183 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
34dc7c2f
BB		 2184 return (0);
2185
2186 switch (errno) {
2187 case EBUSY:
2188
2189 /*
2190 * There are no other replicas of this device.
2191 */
2192 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2193
9babb374
BB		 2194 case EEXIST:
2195 /*
2196 * The log device has unplayed logs
2197 */
2198 return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg));
2199
34dc7c2f
BB		 2200 default:
2201 return (zpool_standard_error(hdl, errno, msg));
2202 }
2203}
2204
2205/*
2206 * Mark the given vdev faulted.
2207 */
2208int
428870ff 2209zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
34dc7c2f 2210{
2598c001 2211 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
34dc7c2f
BB		 2212 char msg[1024];
2213 libzfs_handle_t *hdl = zhp->zpool_hdl;
2214
2215 (void) snprintf(msg, sizeof (msg),
b8864a23 2216 dgettext(TEXT_DOMAIN, "cannot fault %llu"), (u_longlong_t)guid);
34dc7c2f
BB		 2217
2218 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2219 zc.zc_guid = guid;
2220 zc.zc_cookie = VDEV_STATE_FAULTED;
428870ff 2221 zc.zc_obj = aux;
34dc7c2f 2222
572e2857 2223 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
34dc7c2f
BB		 2224 return (0);
2225
2226 switch (errno) {
2227 case EBUSY:
2228
2229 /*
2230 * There are no other replicas of this device.
2231 */
2232 return (zfs_error(hdl, EZFS_NOREPLICAS, msg));
2233
2234 default:
2235 return (zpool_standard_error(hdl, errno, msg));
2236 }
2237
2238}
2239
2240/*
2241 * Mark the given vdev degraded.
2242 */
2243int
428870ff 2244zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux)
34dc7c2f 2245{
2598c001 2246 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
34dc7c2f
BB		 2247 char msg[1024];
2248 libzfs_handle_t *hdl = zhp->zpool_hdl;
2249
2250 (void) snprintf(msg, sizeof (msg),
b8864a23 2251 dgettext(TEXT_DOMAIN, "cannot degrade %llu"), (u_longlong_t)guid);
34dc7c2f
BB		 2252
2253 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2254 zc.zc_guid = guid;
2255 zc.zc_cookie = VDEV_STATE_DEGRADED;
428870ff 2256 zc.zc_obj = aux;
34dc7c2f 2257
572e2857 2258 if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0)
34dc7c2f
BB		 2259 return (0);
2260
2261 return (zpool_standard_error(hdl, errno, msg));
2262}
2263
2264/*
2265 * Returns TRUE if the given nvlist is a vdev that was originally swapped in as
2266 * a hot spare.
2267 */
2268static boolean_t
2269is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which)
2270{
2271 nvlist_t **child;
2272 uint_t c, children;
2273 char *type;
2274
2275 if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child,
2276 &children) == 0) {
2277 verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE,
2278 &type) == 0);
2279
2280 if (strcmp(type, VDEV_TYPE_SPARE) == 0 &&
2281 children == 2 && child[which] == tgt)
2282 return (B_TRUE);
2283
2284 for (c = 0; c < children; c++)
2285 if (is_replacing_spare(child[c], tgt, which))
2286 return (B_TRUE);
2287 }
2288
2289 return (B_FALSE);
2290}
2291
2292/*
2293 * Attach new_disk (fully described by nvroot) to old_disk.
2294 * If 'replacing' is specified, the new disk will replace the old one.
2295 */
2296int
2297zpool_vdev_attach(zpool_handle_t *zhp,
2298 const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing)
2299{
2598c001 2300 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
34dc7c2f
BB		 2301 char msg[1024];
2302 int ret;
2303 nvlist_t *tgt;
b128c09f
BB		 2304 boolean_t avail_spare, l2cache, islog;
2305 uint64_t val;
572e2857 2306 char *newname;
34dc7c2f
BB		 2307 nvlist_t **child;
2308 uint_t children;
2309 nvlist_t *config_root;
2310 libzfs_handle_t *hdl = zhp->zpool_hdl;
b128c09f 2311 boolean_t rootpool = pool_is_bootable(zhp);
34dc7c2f
BB		 2312
2313 if (replacing)
2314 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2315 "cannot replace %s with %s"), old_disk, new_disk);
2316 else
2317 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
2318 "cannot attach %s to %s"), new_disk, old_disk);
2319
b128c09f
BB		 2320 /*
2321 * If this is a root pool, make sure that we're not attaching an
2322 * EFI labeled device.
2323 */
2324 if (rootpool && pool_uses_efi(nvroot)) {
2325 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2326 "EFI labeled devices are not supported on root pools."));
2327 return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg));
2328 }
2329
34dc7c2f 2330 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
b128c09f
BB		 2331 if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache,
2332 &islog)) == 0)
34dc7c2f
BB		 2333 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2334
2335 if (avail_spare)
2336 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2337
2338 if (l2cache)
2339 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2340
2341 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2342 zc.zc_cookie = replacing;
2343
2344 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
2345 &child, &children) != 0 || children != 1) {
2346 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2347 "new device must be a single disk"));
2348 return (zfs_error(hdl, EZFS_INVALCONFIG, msg));
2349 }
2350
2351 verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL),
2352 ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0);
2353
428870ff 2354 if ((newname = zpool_vdev_name(NULL, NULL, child[0], B_FALSE)) == NULL)
b128c09f
BB		 2355 return (-1);
2356
34dc7c2f
BB		 2357 /*
2358 * If the target is a hot spare that has been swapped in, we can only
2359 * replace it with another hot spare.
2360 */
2361 if (replacing &&
2362 nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 &&
b128c09f
BB		 2363 (zpool_find_vdev(zhp, newname, &avail_spare, &l2cache,
2364 NULL) == NULL || !avail_spare) &&
2365 is_replacing_spare(config_root, tgt, 1)) {
34dc7c2f
BB		 2366 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2367 "can only be replaced by another hot spare"));
b128c09f 2368 free(newname);
34dc7c2f
BB		 2369 return (zfs_error(hdl, EZFS_BADTARGET, msg));
2370 }
2371
b128c09f
BB		 2372 free(newname);
2373
34dc7c2f
BB		 2374 if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0)
2375 return (-1);
2376
572e2857 2377 ret = zfs_ioctl(hdl, ZFS_IOC_VDEV_ATTACH, &zc);
34dc7c2f
BB		 2378
2379 zcmd_free_nvlists(&zc);
2380
b128c09f
BB		 2381 if (ret == 0) {
2382 if (rootpool) {
9babb374
BB		 2383 /*
2384 * XXX need a better way to prevent user from
2385 * booting up a half-baked vdev.
2386 */
2387 (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make "
2388 "sure to wait until resilver is done "
2389 "before rebooting.\n"));
b128c09f 2390 }
34dc7c2f 2391 return (0);
b128c09f 2392 }
34dc7c2f
BB		 2393
2394 switch (errno) {
2395 case ENOTSUP:
2396 /*
2397 * Can't attach to or replace this type of vdev.
2398 */
2399 if (replacing) {
572e2857
BB		 2400 uint64_t version = zpool_get_prop_int(zhp,
2401 ZPOOL_PROP_VERSION, NULL);
2402
b128c09f 2403 if (islog)
34dc7c2f
BB		 2404 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2405 "cannot replace a log with a spare"));
572e2857
BB		 2406 else if (version >= SPA_VERSION_MULTI_REPLACE)
2407 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2408 "already in replacing/spare config; wait "
2409 "for completion or use 'zpool detach'"));
34dc7c2f
BB		 2410 else
2411 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2412 "cannot replace a replacing device"));
2413 } else {
2414 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2415 "can only attach to mirrors and top-level "
2416 "disks"));
2417 }
2418 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
2419 break;
2420
2421 case EINVAL:
2422 /*
2423 * The new device must be a single disk.
2424 */
2425 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2426 "new device must be a single disk"));
2427 (void) zfs_error(hdl, EZFS_INVALCONFIG, msg);
2428 break;
2429
2430 case EBUSY:
2431 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy"),
2432 new_disk);
2433 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2434 break;
2435
2436 case EOVERFLOW:
2437 /*
2438 * The new device is too small.
2439 */
2440 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2441 "device is too small"));
2442 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2443 break;
2444
2445 case EDOM:
2446 /*
2447 * The new device has a different alignment requirement.
2448 */
2449 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2450 "devices have different sector alignment"));
2451 (void) zfs_error(hdl, EZFS_BADDEV, msg);
2452 break;
2453
2454 case ENAMETOOLONG:
2455 /*
2456 * The resulting top-level vdev spec won't fit in the label.
2457 */
2458 (void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg);
2459 break;
2460
2461 default:
2462 (void) zpool_standard_error(hdl, errno, msg);
2463 }
2464
2465 return (-1);
2466}
2467
2468/*
2469 * Detach the specified device.
2470 */
2471int
2472zpool_vdev_detach(zpool_handle_t *zhp, const char *path)
2473{
2598c001 2474 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
34dc7c2f
BB		 2475 char msg[1024];
2476 nvlist_t *tgt;
2477 boolean_t avail_spare, l2cache;
2478 libzfs_handle_t *hdl = zhp->zpool_hdl;
2479
2480 (void) snprintf(msg, sizeof (msg),
2481 dgettext(TEXT_DOMAIN, "cannot detach %s"), path);
2482
2483 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
b128c09f
BB		 2484 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
2485 NULL)) == 0)
34dc7c2f
BB		 2486 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2487
2488 if (avail_spare)
2489 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2490
2491 if (l2cache)
2492 return (zfs_error(hdl, EZFS_ISL2CACHE, msg));
2493
2494 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2495
2496 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0)
2497 return (0);
2498
2499 switch (errno) {
2500
2501 case ENOTSUP:
2502 /*
2503 * Can't detach from this type of vdev.
2504 */
2505 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only "
2506 "applicable to mirror and replacing vdevs"));
572e2857 2507 (void) zfs_error(hdl, EZFS_BADTARGET, msg);
34dc7c2f
BB		 2508 break;
2509
2510 case EBUSY:
2511 /*
2512 * There are no other replicas of this device.
2513 */
2514 (void) zfs_error(hdl, EZFS_NOREPLICAS, msg);
2515 break;
2516
2517 default:
2518 (void) zpool_standard_error(hdl, errno, msg);
2519 }
2520
2521 return (-1);
2522}
2523
428870ff
BB		 2524/*
2525 * Find a mirror vdev in the source nvlist.
2526 *
2527 * The mchild array contains a list of disks in one of the top-level mirrors
2528 * of the source pool. The schild array contains a list of disks that the
2529 * user specified on the command line. We loop over the mchild array to
2530 * see if any entry in the schild array matches.
2531 *
2532 * If a disk in the mchild array is found in the schild array, we return
2533 * the index of that entry. Otherwise we return -1.
2534 */
2535static int
2536find_vdev_entry(zpool_handle_t *zhp, nvlist_t **mchild, uint_t mchildren,
2537 nvlist_t **schild, uint_t schildren)
2538{
2539 uint_t mc;
2540
2541 for (mc = 0; mc < mchildren; mc++) {
2542 uint_t sc;
2543 char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp,
2544 mchild[mc], B_FALSE);
2545
2546 for (sc = 0; sc < schildren; sc++) {
2547 char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp,
2548 schild[sc], B_FALSE);
2549 boolean_t result = (strcmp(mpath, spath) == 0);
2550
2551 free(spath);
2552 if (result) {
2553 free(mpath);
2554 return (mc);
2555 }
2556 }
2557
2558 free(mpath);
2559 }
2560
2561 return (-1);
2562}
2563
2564/*
2565 * Split a mirror pool. If newroot points to null, then a new nvlist
2566 * is generated and it is the responsibility of the caller to free it.
2567 */
2568int
2569zpool_vdev_split(zpool_handle_t *zhp, char *newname, nvlist_t **newroot,
2570 nvlist_t *props, splitflags_t flags)
2571{
2598c001 2572 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
428870ff
BB		 2573 char msg[1024];
2574 nvlist_t *tree, *config, **child, **newchild, *newconfig = NULL;
2575 nvlist_t **varray = NULL, *zc_props = NULL;
2576 uint_t c, children, newchildren, lastlog = 0, vcount, found = 0;
2577 libzfs_handle_t *hdl = zhp->zpool_hdl;
2578 uint64_t vers;
2579 boolean_t freelist = B_FALSE, memory_err = B_TRUE;
2580 int retval = 0;
2581
2582 (void) snprintf(msg, sizeof (msg),
2583 dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name);
2584
2585 if (!zpool_name_valid(hdl, B_FALSE, newname))
2586 return (zfs_error(hdl, EZFS_INVALIDNAME, msg));
2587
2588 if ((config = zpool_get_config(zhp, NULL)) == NULL) {
2589 (void) fprintf(stderr, gettext("Internal error: unable to "
2590 "retrieve pool configuration\n"));
2591 return (-1);
2592 }
2593
2594 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree)
2595 == 0);
2596 verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &vers) == 0);
2597
2598 if (props) {
572e2857 2599 prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE };
428870ff 2600 if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name,
572e2857 2601 props, vers, flags, msg)) == NULL)
428870ff
BB		 2602 return (-1);
2603 }
2604
2605 if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child,
2606 &children) != 0) {
2607 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2608 "Source pool is missing vdev tree"));
2609 if (zc_props)
2610 nvlist_free(zc_props);
2611 return (-1);
2612 }
2613
2614 varray = zfs_alloc(hdl, children * sizeof (nvlist_t *));
2615 vcount = 0;
2616
2617 if (*newroot == NULL ||
2618 nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN,
2619 &newchild, &newchildren) != 0)
2620 newchildren = 0;
2621
2622 for (c = 0; c < children; c++) {
2623 uint64_t is_log = B_FALSE, is_hole = B_FALSE;
2624 char *type;
2625 nvlist_t **mchild, *vdev;
2626 uint_t mchildren;
2627 int entry;
2628
2629 /*
2630 * Unlike cache & spares, slogs are stored in the
2631 * ZPOOL_CONFIG_CHILDREN array. We filter them out here.
2632 */
2633 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG,
2634 &is_log);
2635 (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE,
2636 &is_hole);
2637 if (is_log || is_hole) {
2638 /*
2639 * Create a hole vdev and put it in the config.
2640 */
2641 if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0)
2642 goto out;
2643 if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE,
2644 VDEV_TYPE_HOLE) != 0)
2645 goto out;
2646 if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE,
2647 1) != 0)
2648 goto out;
2649 if (lastlog == 0)
2650 lastlog = vcount;
2651 varray[vcount++] = vdev;
2652 continue;
2653 }
2654 lastlog = 0;
2655 verify(nvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE, &type)
2656 == 0);
2657 if (strcmp(type, VDEV_TYPE_MIRROR) != 0) {
2658 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2659 "Source pool must be composed only of mirrors\n"));
2660 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
2661 goto out;
2662 }
2663
2664 verify(nvlist_lookup_nvlist_array(child[c],
2665 ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0);
2666
2667 /* find or add an entry for this top-level vdev */
2668 if (newchildren > 0 &&
2669 (entry = find_vdev_entry(zhp, mchild, mchildren,
2670 newchild, newchildren)) >= 0) {
2671 /* We found a disk that the user specified. */
2672 vdev = mchild[entry];
2673 ++found;
2674 } else {
2675 /* User didn't specify a disk for this vdev. */
2676 vdev = mchild[mchildren - 1];
2677 }
2678
2679 if (nvlist_dup(vdev, &varray[vcount++], 0) != 0)
2680 goto out;
2681 }
2682
2683 /* did we find every disk the user specified? */
2684 if (found != newchildren) {
2685 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must "
2686 "include at most one disk from each mirror"));
2687 retval = zfs_error(hdl, EZFS_INVALCONFIG, msg);
2688 goto out;
2689 }
2690
2691 /* Prepare the nvlist for populating. */
2692 if (*newroot == NULL) {
2693 if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0)
2694 goto out;
2695 freelist = B_TRUE;
2696 if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE,
2697 VDEV_TYPE_ROOT) != 0)
2698 goto out;
2699 } else {
2700 verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0);
2701 }
2702
2703 /* Add all the children we found */
2704 if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, varray,
2705 lastlog == 0 ? vcount : lastlog) != 0)
2706 goto out;
2707
2708 /*
2709 * If we're just doing a dry run, exit now with success.
2710 */
2711 if (flags.dryrun) {
2712 memory_err = B_FALSE;
2713 freelist = B_FALSE;
2714 goto out;
2715 }
2716
2717 /* now build up the config list & call the ioctl */
2718 if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0)
2719 goto out;
2720
2721 if (nvlist_add_nvlist(newconfig,
2722 ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 ||
2723 nvlist_add_string(newconfig,
2724 ZPOOL_CONFIG_POOL_NAME, newname) != 0 ||
2725 nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0)
2726 goto out;
2727
2728 /*
2729 * The new pool is automatically part of the namespace unless we
2730 * explicitly export it.
2731 */
2732 if (!flags.import)
2733 zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT;
2734 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2735 (void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string));
2736 if (zcmd_write_conf_nvlist(hdl, &zc, newconfig) != 0)
2737 goto out;
2738 if (zc_props != NULL && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0)
2739 goto out;
2740
2741 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) {
2742 retval = zpool_standard_error(hdl, errno, msg);
2743 goto out;
2744 }
2745
2746 freelist = B_FALSE;
2747 memory_err = B_FALSE;
2748
2749out:
2750 if (varray != NULL) {
2751 int v;
2752
2753 for (v = 0; v < vcount; v++)
2754 nvlist_free(varray[v]);
2755 free(varray);
2756 }
2757 zcmd_free_nvlists(&zc);
2758 if (zc_props)
2759 nvlist_free(zc_props);
2760 if (newconfig)
2761 nvlist_free(newconfig);
2762 if (freelist) {
2763 nvlist_free(*newroot);
2764 *newroot = NULL;
2765 }
2766
2767 if (retval != 0)
2768 return (retval);
2769
2770 if (memory_err)
2771 return (no_memory(hdl));
2772
2773 return (0);
2774}
2775
34dc7c2f
BB		 2776/*
2777 * Remove the given device. Currently, this is supported only for hot spares
2778 * and level 2 cache devices.
2779 */
2780int
2781zpool_vdev_remove(zpool_handle_t *zhp, const char *path)
2782{
2598c001 2783 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
34dc7c2f
BB		 2784 char msg[1024];
2785 nvlist_t *tgt;
428870ff 2786 boolean_t avail_spare, l2cache, islog;
34dc7c2f 2787 libzfs_handle_t *hdl = zhp->zpool_hdl;
428870ff 2788 uint64_t version;
34dc7c2f
BB		 2789
2790 (void) snprintf(msg, sizeof (msg),
2791 dgettext(TEXT_DOMAIN, "cannot remove %s"), path);
2792
2793 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
b128c09f 2794 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache,
428870ff 2795 &islog)) == 0)
34dc7c2f 2796 return (zfs_error(hdl, EZFS_NODEVICE, msg));
428870ff
BB		 2797 /*
2798 * XXX - this should just go away.
2799 */
2800 if (!avail_spare && !l2cache && !islog) {
34dc7c2f 2801 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
428870ff
BB		 2802 "only inactive hot spares, cache, top-level, "
2803 "or log devices can be removed"));
34dc7c2f
BB		 2804 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2805 }
2806
428870ff
BB		 2807 version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL);
2808 if (islog && version < SPA_VERSION_HOLES) {
2809 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
2810 "pool must be upgrade to support log removal"));
2811 return (zfs_error(hdl, EZFS_BADVERSION, msg));
2812 }
2813
34dc7c2f
BB		 2814 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0);
2815
2816 if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0)
2817 return (0);
2818
2819 return (zpool_standard_error(hdl, errno, msg));
2820}
2821
2822/*
2823 * Clear the errors for the pool, or the particular device if specified.
2824 */
2825int
428870ff 2826zpool_clear(zpool_handle_t *zhp, const char *path, nvlist_t *rewindnvl)
34dc7c2f 2827{
2598c001 2828 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
34dc7c2f
BB		 2829 char msg[1024];
2830 nvlist_t *tgt;
428870ff 2831 zpool_rewind_policy_t policy;
34dc7c2f
BB		 2832 boolean_t avail_spare, l2cache;
2833 libzfs_handle_t *hdl = zhp->zpool_hdl;
428870ff 2834 nvlist_t *nvi = NULL;
572e2857 2835 int error;
34dc7c2f
BB		 2836
2837 if (path)
2838 (void) snprintf(msg, sizeof (msg),
2839 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
2840 path);
2841 else
2842 (void) snprintf(msg, sizeof (msg),
2843 dgettext(TEXT_DOMAIN, "cannot clear errors for %s"),
2844 zhp->zpool_name);
2845
2846 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2847 if (path) {
2848 if ((tgt = zpool_find_vdev(zhp, path, &avail_spare,
b128c09f 2849 &l2cache, NULL)) == 0)
34dc7c2f
BB		 2850 return (zfs_error(hdl, EZFS_NODEVICE, msg));
2851
2852 /*
2853 * Don't allow error clearing for hot spares. Do allow
2854 * error clearing for l2cache devices.
2855 */
2856 if (avail_spare)
2857 return (zfs_error(hdl, EZFS_ISSPARE, msg));
2858
2859 verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID,
2860 &zc.zc_guid) == 0);
2861 }
2862
428870ff
BB		 2863 zpool_get_rewind_policy(rewindnvl, &policy);
2864 zc.zc_cookie = policy.zrp_request;
2865
572e2857 2866 if (zcmd_alloc_dst_nvlist(hdl, &zc, zhp->zpool_config_size * 2) != 0)
428870ff
BB		 2867 return (-1);
2868
572e2857 2869 if (zcmd_write_src_nvlist(hdl, &zc, rewindnvl) != 0)
428870ff
BB		 2870 return (-1);
2871
572e2857
BB		 2872 while ((error = zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc)) != 0 &&
2873 errno == ENOMEM) {
2874 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
2875 zcmd_free_nvlists(&zc);
2876 return (-1);
2877 }
2878 }
2879
2880 if (!error || ((policy.zrp_request & ZPOOL_TRY_REWIND) &&
428870ff
BB		 2881 errno != EPERM && errno != EACCES)) {
2882 if (policy.zrp_request &
2883 (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) {
2884 (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi);
2885 zpool_rewind_exclaim(hdl, zc.zc_name,
2886 ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0),
2887 nvi);
2888 nvlist_free(nvi);
2889 }
2890 zcmd_free_nvlists(&zc);
34dc7c2f 2891 return (0);
428870ff 2892 }
34dc7c2f 2893
428870ff 2894 zcmd_free_nvlists(&zc);
34dc7c2f
BB		 2895 return (zpool_standard_error(hdl, errno, msg));
2896}
2897
2898/*
2899 * Similar to zpool_clear(), but takes a GUID (used by fmd).
2900 */
2901int
2902zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid)
2903{
2598c001 2904 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
34dc7c2f
BB		 2905 char msg[1024];
2906 libzfs_handle_t *hdl = zhp->zpool_hdl;
2907
2908 (void) snprintf(msg, sizeof (msg),
2909 dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"),
b8864a23 2910 (u_longlong_t)guid);
34dc7c2f
BB		 2911
2912 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2913 zc.zc_guid = guid;
428870ff 2914 zc.zc_cookie = ZPOOL_NO_REWIND;
34dc7c2f
BB		 2915
2916 if (ioctl(hdl->libzfs_fd, ZFS_IOC_CLEAR, &zc) == 0)
2917 return (0);
2918
2919 return (zpool_standard_error(hdl, errno, msg));
2920}
2921
34dc7c2f
BB		 2922/*
2923 * Convert from a devid string to a path.
2924 */
2925static char *
2926devid_to_path(char *devid_str)
2927{
2928 ddi_devid_t devid;
2929 char *minor;
2930 char *path;
2931 devid_nmlist_t *list = NULL;
2932 int ret;
2933
2934 if (devid_str_decode(devid_str, &devid, &minor) != 0)
2935 return (NULL);
2936
2937 ret = devid_deviceid_to_nmlist("/dev", devid, minor, &list);
2938
2939 devid_str_free(minor);
2940 devid_free(devid);
2941
2942 if (ret != 0)
2943 return (NULL);
2944
2945 if ((path = strdup(list[0].devname)) == NULL)
2946 return (NULL);
2947
2948 devid_free_nmlist(list);
2949
2950 return (path);
2951}
2952
2953/*
2954 * Convert from a path to a devid string.
2955 */
2956static char *
2957path_to_devid(const char *path)
2958{
2959 int fd;
2960 ddi_devid_t devid;
2961 char *minor, *ret;
2962
2963 if ((fd = open(path, O_RDONLY)) < 0)
2964 return (NULL);
2965
2966 minor = NULL;
2967 ret = NULL;
2968 if (devid_get(fd, &devid) == 0) {
2969 if (devid_get_minor_name(fd, &minor) == 0)
2970 ret = devid_str_encode(devid, minor);
2971 if (minor != NULL)
2972 devid_str_free(minor);
2973 devid_free(devid);
2974 }
2975 (void) close(fd);
2976
2977 return (ret);
2978}
2979
2980/*
2981 * Issue the necessary ioctl() to update the stored path value for the vdev. We
2982 * ignore any failure here, since a common case is for an unprivileged user to
2983 * type 'zpool status', and we'll display the correct information anyway.
2984 */
2985static void
2986set_path(zpool_handle_t *zhp, nvlist_t *nv, const char *path)
2987{
2598c001 2988 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
34dc7c2f
BB		 2989
2990 (void) strncpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
2991 (void) strncpy(zc.zc_value, path, sizeof (zc.zc_value));
2992 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
2993 &zc.zc_guid) == 0);
2994
2995 (void) ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SETPATH, &zc);
2996}
2997
83c62c93
NB		 2998/*
2999 * Remove partition suffix from a vdev path. Partition suffixes may take three
3000 * forms: "-partX", "pX", or "X", where X is a string of digits. The second
3001 * case only occurs when the suffix is preceded by a digit, i.e. "md0p0" The
3002 * third case only occurs when preceded by a string matching the regular
3003 * expression "^[hs]d[a-z]+", i.e. a scsi or ide disk.
3004 */
3005static char *
3006strip_partition(libzfs_handle_t *hdl, char *path)
3007{
3008 char *tmp = zfs_strdup(hdl, path);
3009 char *part = NULL, *d = NULL;
3010
3011 if ((part = strstr(tmp, "-part")) && part != tmp) {
3012 d = part + 5;
3013 } else if ((part = strrchr(tmp, 'p')) &&
3014 part > tmp + 1 && isdigit(*(part-1))) {
3015 d = part + 1;
3016 } else if ((tmp[0] == 'h' || tmp[0] == 's') && tmp[1] == 'd') {
3017 for (d = &tmp[2]; isalpha(*d); part = ++d);
3018 }
3019 if (part && d && *d != '\0') {
3020 for (; isdigit(*d); d++);
3021 if (*d == '\0')
3022 *part = '\0';
3023 }
3024 return (tmp);
3025}
3026
858219cc
NB		 3027#define PATH_BUF_LEN 64
3028
34dc7c2f
BB		 3029/*
3030 * Given a vdev, return the name to display in iostat. If the vdev has a path,
3031 * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type.
3032 * We also check if this is a whole disk, in which case we strip off the
3033 * trailing 's0' slice name.
3034 *
3035 * This routine is also responsible for identifying when disks have been
3036 * reconfigured in a new location. The kernel will have opened the device by
3037 * devid, but the path will still refer to the old location. To catch this, we
3038 * first do a path -> devid translation (which is fast for the common case). If
3039 * the devid matches, we're done. If not, we do a reverse devid -> path
3040 * translation and issue the appropriate ioctl() to update the path of the vdev.
3041 * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any
3042 * of these checks.
3043 */
3044char *
428870ff
BB		 3045zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv,
3046 boolean_t verbose)
34dc7c2f 3047{
d603ed6c 3048 char *path, *devid, *type;
34dc7c2f 3049 uint64_t value;
858219cc 3050 char buf[PATH_BUF_LEN];
34dc7c2f
BB		 3051 vdev_stat_t *vs;
3052 uint_t vsc;
3053
3054 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT,
3055 &value) == 0) {
3056 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID,
3057 &value) == 0);
3058 (void) snprintf(buf, sizeof (buf), "%llu",
3059 (u_longlong_t)value);
3060 path = buf;
3061 } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) {
34dc7c2f
BB		 3062 /*
3063 * If the device is dead (faulted, offline, etc) then don't
3064 * bother opening it. Otherwise we may be forcing the user to
3065 * open a misbehaving device, which can have undesirable
3066 * effects.
3067 */
428870ff 3068 if ((nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS,
34dc7c2f
BB		 3069 (uint64_t **)&vs, &vsc) != 0 ||
3070 vs->vs_state >= VDEV_STATE_DEGRADED) &&
3071 zhp != NULL &&
3072 nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) {
3073 /*
3074 * Determine if the current path is correct.
3075 */
3076 char *newdevid = path_to_devid(path);
3077
3078 if (newdevid == NULL ||
3079 strcmp(devid, newdevid) != 0) {
3080 char *newpath;
3081
3082 if ((newpath = devid_to_path(devid)) != NULL) {
3083 /*
3084 * Update the path appropriately.
3085 */
3086 set_path(zhp, nv, newpath);
3087 if (nvlist_add_string(nv,
3088 ZPOOL_CONFIG_PATH, newpath) == 0)
3089 verify(nvlist_lookup_string(nv,
3090 ZPOOL_CONFIG_PATH,
3091 &path) == 0);
3092 free(newpath);
3093 }
3094 }
3095
3096 if (newdevid)
3097 devid_str_free(newdevid);
3098 }
3099
d603ed6c
BB		 3100 /*
3101 * For a block device only use the name.
3102 */
3103 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0);
3104 if (strcmp(type, VDEV_TYPE_DISK) == 0) {
3105 path = strrchr(path, '/');
3106 path++;
3107 }
34dc7c2f 3108
d603ed6c 3109 /*
83c62c93 3110 * Remove the partition from the path it this is a whole disk.
d603ed6c 3111 */
34dc7c2f
BB		 3112 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
3113 &value) == 0 && value) {
83c62c93 3114 return strip_partition(hdl, path);
34dc7c2f
BB		 3115 }
3116 } else {
3117 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &path) == 0);
3118
3119 /*
3120 * If it's a raidz device, we need to stick in the parity level.
3121 */
3122 if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) {
858219cc
NB		 3123 char tmpbuf[PATH_BUF_LEN];
3124
34dc7c2f
BB		 3125 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY,
3126 &value) == 0);
858219cc 3127 (void) snprintf(tmpbuf, sizeof (tmpbuf), "%s%llu", path,
34dc7c2f 3128 (u_longlong_t)value);
858219cc 3129 path = tmpbuf;
34dc7c2f 3130 }
428870ff
BB		 3131
3132 /*
3133 * We identify each top-level vdev by using a <type-id>
3134 * naming convention.
3135 */
3136 if (verbose) {
3137 uint64_t id;
3138
3139 verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID,
3140 &id) == 0);
3141 (void) snprintf(buf, sizeof (buf), "%s-%llu", path,
3142 (u_longlong_t)id);
3143 path = buf;
3144 }
34dc7c2f
BB		 3145 }
3146
3147 return (zfs_strdup(hdl, path));
3148}
3149
3150static int
3151zbookmark_compare(const void *a, const void *b)
3152{
3153 return (memcmp(a, b, sizeof (zbookmark_t)));
3154}
3155
3156/*
3157 * Retrieve the persistent error log, uniquify the members, and return to the
3158 * caller.
3159 */
3160int
3161zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp)
3162{
2598c001 3163 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
34dc7c2f
BB		 3164 uint64_t count;
3165 zbookmark_t *zb = NULL;
3166 int i;
3167
3168 /*
3169 * Retrieve the raw error list from the kernel. If the number of errors
3170 * has increased, allocate more space and continue until we get the
3171 * entire list.
3172 */
3173 verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT,
3174 &count) == 0);
3175 if (count == 0)
3176 return (0);
3177 if ((zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl,
3178 count * sizeof (zbookmark_t))) == (uintptr_t)NULL)
3179 return (-1);
3180 zc.zc_nvlist_dst_size = count;
3181 (void) strcpy(zc.zc_name, zhp->zpool_name);
3182 for (;;) {
3183 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_ERROR_LOG,
3184 &zc) != 0) {
3185 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3186 if (errno == ENOMEM) {
3187 count = zc.zc_nvlist_dst_size;
3188 if ((zc.zc_nvlist_dst = (uintptr_t)
3189 zfs_alloc(zhp->zpool_hdl, count *
3190 sizeof (zbookmark_t))) == (uintptr_t)NULL)
3191 return (-1);
3192 } else {
3193 return (-1);
3194 }
3195 } else {
3196 break;
3197 }
3198 }
3199
3200 /*
3201 * Sort the resulting bookmarks. This is a little confusing due to the
3202 * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last
3203 * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks
3204 * _not_ copied as part of the process. So we point the start of our
3205 * array appropriate and decrement the total number of elements.
3206 */
3207 zb = ((zbookmark_t *)(uintptr_t)zc.zc_nvlist_dst) +
3208 zc.zc_nvlist_dst_size;
3209 count -= zc.zc_nvlist_dst_size;
3210
3211 qsort(zb, count, sizeof (zbookmark_t), zbookmark_compare);
3212
3213 verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0);
3214
3215 /*
3216 * Fill in the nverrlistp with nvlist's of dataset and object numbers.
3217 */
3218 for (i = 0; i < count; i++) {
3219 nvlist_t *nv;
3220
3221 /* ignoring zb_blkid and zb_level for now */
3222 if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset &&
3223 zb[i-1].zb_object == zb[i].zb_object)
3224 continue;
3225
3226 if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0)
3227 goto nomem;
3228 if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET,
3229 zb[i].zb_objset) != 0) {
3230 nvlist_free(nv);
3231 goto nomem;
3232 }
3233 if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT,
3234 zb[i].zb_object) != 0) {
3235 nvlist_free(nv);
3236 goto nomem;
3237 }
3238 if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) {
3239 nvlist_free(nv);
3240 goto nomem;
3241 }
3242 nvlist_free(nv);
3243 }
3244
3245 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3246 return (0);
3247
3248nomem:
3249 free((void *)(uintptr_t)zc.zc_nvlist_dst);
3250 return (no_memory(zhp->zpool_hdl));
3251}
3252
3253/*
3254 * Upgrade a ZFS pool to the latest on-disk version.
3255 */
3256int
3257zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version)
3258{
2598c001 3259 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
34dc7c2f
BB		 3260 libzfs_handle_t *hdl = zhp->zpool_hdl;
3261
3262 (void) strcpy(zc.zc_name, zhp->zpool_name);
3263 zc.zc_cookie = new_version;
3264
3265 if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0)
3266 return (zpool_standard_error_fmt(hdl, errno,
3267 dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"),
3268 zhp->zpool_name));
3269 return (0);
3270}
3271
3272void
3273zpool_set_history_str(const char *subcommand, int argc, char **argv,
3274 char *history_str)
3275{
3276 int i;
3277
3278 (void) strlcpy(history_str, subcommand, HIS_MAX_RECORD_LEN);
3279 for (i = 1; i < argc; i++) {
3280 if (strlen(history_str) + 1 + strlen(argv[i]) >
3281 HIS_MAX_RECORD_LEN)
3282 break;
3283 (void) strlcat(history_str, " ", HIS_MAX_RECORD_LEN);
3284 (void) strlcat(history_str, argv[i], HIS_MAX_RECORD_LEN);
3285 }
3286}
3287
3288/*
3289 * Stage command history for logging.
3290 */
3291int
3292zpool_stage_history(libzfs_handle_t *hdl, const char *history_str)
3293{
3294 if (history_str == NULL)
3295 return (EINVAL);
3296
3297 if (strlen(history_str) > HIS_MAX_RECORD_LEN)
3298 return (EINVAL);
3299
3300 if (hdl->libzfs_log_str != NULL)
3301 free(hdl->libzfs_log_str);
3302
3303 if ((hdl->libzfs_log_str = strdup(history_str)) == NULL)
3304 return (no_memory(hdl));
3305
3306 return (0);
3307}
3308
3309/*
3310 * Perform ioctl to get some command history of a pool.
3311 *
3312 * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the
3313 * logical offset of the history buffer to start reading from.
3314 *
3315 * Upon return, 'off' is the next logical offset to read from and
3316 * 'len' is the actual amount of bytes read into 'buf'.
3317 */
3318static int
3319get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len)
3320{
2598c001 3321 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
34dc7c2f
BB		 3322 libzfs_handle_t *hdl = zhp->zpool_hdl;
3323
3324 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3325
3326 zc.zc_history = (uint64_t)(uintptr_t)buf;
3327 zc.zc_history_len = *len;
3328 zc.zc_history_offset = *off;
3329
3330 if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) {
3331 switch (errno) {
3332 case EPERM:
3333 return (zfs_error_fmt(hdl, EZFS_PERM,
3334 dgettext(TEXT_DOMAIN,
3335 "cannot show history for pool '%s'"),
3336 zhp->zpool_name));
3337 case ENOENT:
3338 return (zfs_error_fmt(hdl, EZFS_NOHISTORY,
3339 dgettext(TEXT_DOMAIN, "cannot get history for pool "
3340 "'%s'"), zhp->zpool_name));
3341 case ENOTSUP:
3342 return (zfs_error_fmt(hdl, EZFS_BADVERSION,
3343 dgettext(TEXT_DOMAIN, "cannot get history for pool "
3344 "'%s', pool must be upgraded"), zhp->zpool_name));
3345 default:
3346 return (zpool_standard_error_fmt(hdl, errno,
3347 dgettext(TEXT_DOMAIN,
3348 "cannot get history for '%s'"), zhp->zpool_name));
3349 }
3350 }
3351
3352 *len = zc.zc_history_len;
3353 *off = zc.zc_history_offset;
3354
3355 return (0);
3356}
3357
3358/*
3359 * Process the buffer of nvlists, unpacking and storing each nvlist record
3360 * into 'records'. 'leftover' is set to the number of bytes that weren't
3361 * processed as there wasn't a complete record.
3362 */
428870ff 3363int
34dc7c2f
BB		 3364zpool_history_unpack(char *buf, uint64_t bytes_read, uint64_t *leftover,
3365 nvlist_t ***records, uint_t *numrecords)
3366{
3367 uint64_t reclen;
3368 nvlist_t *nv;
3369 int i;
3370
3371 while (bytes_read > sizeof (reclen)) {
3372
3373 /* get length of packed record (stored as little endian) */
3374 for (i = 0, reclen = 0; i < sizeof (reclen); i++)
3375 reclen += (uint64_t)(((uchar_t *)buf)[i]) << (8*i);
3376
3377 if (bytes_read < sizeof (reclen) + reclen)
3378 break;
3379
3380 /* unpack record */
3381 if (nvlist_unpack(buf + sizeof (reclen), reclen, &nv, 0) != 0)
3382 return (ENOMEM);
3383 bytes_read -= sizeof (reclen) + reclen;
3384 buf += sizeof (reclen) + reclen;
3385
3386 /* add record to nvlist array */
3387 (*numrecords)++;
3388 if (ISP2(*numrecords + 1)) {
3389 *records = realloc(*records,
3390 *numrecords * 2 * sizeof (nvlist_t *));
3391 }
3392 (*records)[*numrecords - 1] = nv;
3393 }
3394
3395 *leftover = bytes_read;
3396 return (0);
3397}
3398
3399#define HIS_BUF_LEN (128*1024)
3400
3401/*
3402 * Retrieve the command history of a pool.
3403 */
3404int
3405zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp)
3406{
3407 char buf[HIS_BUF_LEN];
3408 uint64_t off = 0;
3409 nvlist_t **records = NULL;
3410 uint_t numrecords = 0;
3411 int err, i;
3412
3413 do {
3414 uint64_t bytes_read = sizeof (buf);
3415 uint64_t leftover;
3416
3417 if ((err = get_history(zhp, buf, &off, &bytes_read)) != 0)
3418 break;
3419
3420 /* if nothing else was read in, we're at EOF, just return */
3421 if (!bytes_read)
3422 break;
3423
3424 if ((err = zpool_history_unpack(buf, bytes_read,
3425 &leftover, &records, &numrecords)) != 0)
3426 break;
3427 off -= leftover;
3428
3429 /* CONSTCOND */
3430 } while (1);
3431
3432 if (!err) {
3433 verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0);
3434 verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD,
3435 records, numrecords) == 0);
3436 }
3437 for (i = 0; i < numrecords; i++)
3438 nvlist_free(records[i]);
3439 free(records);
3440
3441 return (err);
3442}
3443
26685276
BB		 3444/*
3445 * Retrieve the next event. If there is a new event available 'nvp' will
3446 * contain a newly allocated nvlist and 'dropped' will be set to the number
3447 * of missed events since the last call to this function. When 'nvp' is
3448 * set to NULL it indicates no new events are available. In either case
3449 * the function returns 0 and it is up to the caller to free 'nvp'. In
3450 * the case of a fatal error the function will return a non-zero value.
3451 * When the function is called in blocking mode it will not return until
3452 * a new event is available.
3453 */
3454int
3455zpool_events_next(libzfs_handle_t *hdl, nvlist_t **nvp,
3456 int *dropped, int block, int cleanup_fd)
3457{
3458 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
3459 int error = 0;
3460
3461 *nvp = NULL;
3462 *dropped = 0;
3463 zc.zc_cleanup_fd = cleanup_fd;
3464
3465 if (!block)
3466 zc.zc_guid = ZEVENT_NONBLOCK;
3467
3468 if (zcmd_alloc_dst_nvlist(hdl, &zc, ZEVENT_SIZE) != 0)
3469 return (-1);
3470
3471retry:
3472 if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_NEXT, &zc) != 0) {
3473 switch (errno) {
3474 case ESHUTDOWN:
3475 error = zfs_error_fmt(hdl, EZFS_POOLUNAVAIL,
3476 dgettext(TEXT_DOMAIN, "zfs shutdown"));
3477 goto out;
3478 case ENOENT:
3479 /* Blocking error case should not occur */
3480 if (block)
3481 error = zpool_standard_error_fmt(hdl, errno,
3482 dgettext(TEXT_DOMAIN, "cannot get event"));
3483
3484 goto out;
3485 case ENOMEM:
3486 if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) {
3487 error = zfs_error_fmt(hdl, EZFS_NOMEM,
3488 dgettext(TEXT_DOMAIN, "cannot get event"));
3489 goto out;
3490 } else {
3491 goto retry;
3492 }
3493 default:
3494 error = zpool_standard_error_fmt(hdl, errno,
3495 dgettext(TEXT_DOMAIN, "cannot get event"));
3496 goto out;
3497 }
3498 }
3499
3500 error = zcmd_read_dst_nvlist(hdl, &zc, nvp);
3501 if (error != 0)
3502 goto out;
3503
3504 *dropped = (int)zc.zc_cookie;
3505out:
3506 zcmd_free_nvlists(&zc);
3507
3508 return (error);
3509}
3510
3511/*
3512 * Clear all events.
3513 */
3514int
3515zpool_events_clear(libzfs_handle_t *hdl, int *count)
3516{
3517 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
3518 char msg[1024];
3519
3520 (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN,
3521 "cannot clear events"));
3522
3523 if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_CLEAR, &zc) != 0)
3524 return (zpool_standard_error_fmt(hdl, errno, msg));
3525
3526 if (count != NULL)
3527 *count = (int)zc.zc_cookie; /* # of events cleared */
3528
3529 return (0);
3530}
3531
34dc7c2f
BB		 3532void
3533zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj,
3534 char *pathname, size_t len)
3535{
2598c001 3536 zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 };
34dc7c2f
BB		 3537 boolean_t mounted = B_FALSE;
3538 char *mntpnt = NULL;
3539 char dsname[MAXNAMELEN];
3540
3541 if (dsobj == 0) {
3542 /* special case for the MOS */
b8864a23 3543 (void) snprintf(pathname, len, "<metadata>:<0x%llx>", (longlong_t)obj);
34dc7c2f
BB		 3544 return;
3545 }
3546
3547 /* get the dataset's name */
3548 (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name));
3549 zc.zc_obj = dsobj;
3550 if (ioctl(zhp->zpool_hdl->libzfs_fd,
3551 ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) {
3552 /* just write out a path of two object numbers */
3553 (void) snprintf(pathname, len, "<0x%llx>:<0x%llx>",
b8864a23 3554 (longlong_t)dsobj, (longlong_t)obj);
34dc7c2f
BB		 3555 return;
3556 }
3557 (void) strlcpy(dsname, zc.zc_value, sizeof (dsname));
3558
3559 /* find out if the dataset is mounted */
3560 mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt);
3561
3562 /* get the corrupted object's path */
3563 (void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name));
3564 zc.zc_obj = obj;
3565 if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_OBJ_TO_PATH,
3566 &zc) == 0) {
3567 if (mounted) {
3568 (void) snprintf(pathname, len, "%s%s", mntpnt,
3569 zc.zc_value);
3570 } else {
3571 (void) snprintf(pathname, len, "%s:%s",
3572 dsname, zc.zc_value);
3573 }
3574 } else {
b8864a23 3575 (void) snprintf(pathname, len, "%s:<0x%llx>", dsname, (longlong_t)obj);
34dc7c2f
BB		 3576 }
3577 free(mntpnt);
3578}
3579
b128c09f
BB		 3580/*
3581 * Read the EFI label from the config, if a label does not exist then
3582 * pass back the error to the caller. If the caller has passed a non-NULL
3583 * diskaddr argument then we set it to the starting address of the EFI
3584 * partition.
3585 */
3586static int
3587read_efi_label(nvlist_t *config, diskaddr_t *sb)
3588{
3589 char *path;
3590 int fd;
3591 char diskname[MAXPATHLEN];
3592 int err = -1;
3593
3594 if (nvlist_lookup_string(config, ZPOOL_CONFIG_PATH, &path) != 0)
3595 return (err);
3596
3597 (void) snprintf(diskname, sizeof (diskname), "%s%s", RDISK_ROOT,
3598 strrchr(path, '/'));
d603ed6c 3599 if ((fd = open(diskname, O_RDWR|O_DIRECT)) >= 0) {
b128c09f
BB		 3600 struct dk_gpt *vtoc;
3601
3602 if ((err = efi_alloc_and_read(fd, &vtoc)) >= 0) {
3603 if (sb != NULL)
3604 *sb = vtoc->efi_parts[0].p_start;
3605 efi_free(vtoc);
3606 }
3607 (void) close(fd);
3608 }
3609 return (err);
3610}
3611
34dc7c2f
BB		 3612/*
3613 * determine where a partition starts on a disk in the current
3614 * configuration
3615 */
3616static diskaddr_t
3617find_start_block(nvlist_t *config)
3618{
3619 nvlist_t **child;
3620 uint_t c, children;
34dc7c2f 3621 diskaddr_t sb = MAXOFFSET_T;
34dc7c2f
BB		 3622 uint64_t wholedisk;
3623
3624 if (nvlist_lookup_nvlist_array(config,
3625 ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) {
3626 if (nvlist_lookup_uint64(config,
3627 ZPOOL_CONFIG_WHOLE_DISK,
3628 &wholedisk) != 0 || !wholedisk) {
3629 return (MAXOFFSET_T);
3630 }
b128c09f
BB		 3631 if (read_efi_label(config, &sb) < 0)
3632 sb = MAXOFFSET_T;
34dc7c2f
BB		 3633 return (sb);
3634 }
3635
3636 for (c = 0; c < children; c++) {
3637 sb = find_start_block(child[c]);
3638 if (sb != MAXOFFSET_T) {
3639 return (sb);
3640 }
3641 }
3642 return (MAXOFFSET_T);
3643}
3644
d603ed6c
BB		 3645int
3646zpool_label_disk_wait(char *path, int timeout)
3647{
3648 struct stat64 statbuf;
3649 int i;
3650
3651 /*
3652 * Wait timeout miliseconds for a newly created device to be available
3653 * from the given path. There is a small window when a /dev/ device
3654 * will exist and the udev link will not, so we must wait for the
3655 * symlink. Depending on the udev rules this may take a few seconds.
3656 */
3657 for (i = 0; i < timeout; i++) {
3658 usleep(1000);
3659
3660 errno = 0;
3661 if ((stat64(path, &statbuf) == 0) && (errno == 0))
3662 return (0);
3663 }
3664
3665 return (ENOENT);
3666}
3667
3668int
3669zpool_label_disk_check(char *path)
3670{
3671 struct dk_gpt *vtoc;
3672 int fd, err;
3673
3674 if ((fd = open(path, O_RDWR|O_DIRECT)) < 0)
3675 return errno;
3676
3677 if ((err = efi_alloc_and_read(fd, &vtoc)) != 0) {
3678 (void) close(fd);
3679 return err;
3680 }
3681
3682 if (vtoc->efi_flags & EFI_GPT_PRIMARY_CORRUPT) {
3683 efi_free(vtoc);
3684 (void) close(fd);
3685 return EIDRM;
3686 }
3687
3688 efi_free(vtoc);
3689 (void) close(fd);
3690 return 0;
3691}
3692
34dc7c2f
BB		 3693/*
3694 * Label an individual disk. The name provided is the short name,
3695 * stripped of any leading /dev path.
3696 */
3697int
3698zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, char *name)
3699{
3700 char path[MAXPATHLEN];
3701 struct dk_gpt *vtoc;
d603ed6c 3702 int rval, fd;
34dc7c2f
BB		 3703 size_t resv = EFI_MIN_RESV_SIZE;
3704 uint64_t slice_size;
3705 diskaddr_t start_block;
3706 char errbuf[1024];
3707
3708 /* prepare an error message just in case */
3709 (void) snprintf(errbuf, sizeof (errbuf),
3710 dgettext(TEXT_DOMAIN, "cannot label '%s'"), name);
3711
3712 if (zhp) {
3713 nvlist_t *nvroot;
3714
b128c09f
BB		 3715 if (pool_is_bootable(zhp)) {
3716 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3717 "EFI labeled devices are not supported on root "
3718 "pools."));
3719 return (zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf));
3720 }
3721
34dc7c2f
BB		 3722 verify(nvlist_lookup_nvlist(zhp->zpool_config,
3723 ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0);
3724
3725 if (zhp->zpool_start_block == 0)
3726 start_block = find_start_block(nvroot);
3727 else
3728 start_block = zhp->zpool_start_block;
3729 zhp->zpool_start_block = start_block;
3730 } else {
3731 /* new pool */
3732 start_block = NEW_START_BLOCK;
3733 }
3734
3735 (void) snprintf(path, sizeof (path), "%s/%s%s", RDISK_ROOT, name,
3736 BACKUP_SLICE);
3737
d603ed6c 3738 if ((fd = open(path, O_RDWR|O_DIRECT)) < 0) {
34dc7c2f
BB		 3739 /*
3740 * This shouldn't happen. We've long since verified that this
3741 * is a valid device.
3742 */
d603ed6c
BB		 3743 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3744 "unable to open device '%s': %d"), path, errno);
34dc7c2f
BB		 3745 return (zfs_error(hdl, EZFS_OPENFAILED, errbuf));
3746 }
3747
3748 if (efi_alloc_and_init(fd, EFI_NUMPAR, &vtoc) != 0) {
3749 /*
3750 * The only way this can fail is if we run out of memory, or we
3751 * were unable to read the disk's capacity
3752 */
3753 if (errno == ENOMEM)
3754 (void) no_memory(hdl);
3755
3756 (void) close(fd);
3757 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
3758 "unable to read disk capacity"), name);
3759
3760 return (zfs_error(hdl, EZFS_NOCAP, errbuf));
3761 }
3762
3763 slice_size = vtoc->efi_last_u_lba + 1;
3764 slice_size -= EFI_MIN_RESV_SIZE;
3765 if (start_block == MAXOFFSET_T)
3766 start_block = NEW_START_BLOCK;
3767 slice_size -= start_block;
3768
3769 vtoc->efi_parts[0].p_start = start_block;
3770 vtoc->efi_parts[0].p_size = slice_size;
3771
3772 /*
3773 * Why we use V_USR: V_BACKUP confuses users, and is considered
3774 * disposable by some EFI utilities (since EFI doesn't have a backup
3775 * slice). V_UNASSIGNED is supposed to be used only for zero size
3776 * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT,
3777 * etc. were all pretty specific. V_USR is as close to reality as we
3778 * can get, in the absence of V_OTHER.
3779 */
3780 vtoc->efi_parts[0].p_tag = V_USR;
3781 (void) strcpy(vtoc->efi_parts[0].p_name, "zfs");
3782
3783 vtoc->efi_parts[8].p_start = slice_size + start_block;
3784 vtoc->efi_parts[8].p_size = resv;
3785 vtoc->efi_parts[8].p_tag = V_RESERVED;
3786
d603ed6c 3787 if ((rval = efi_write(fd, vtoc)) != 0) {
34dc7c2f
BB		 3788 /*
3789 * Some block drivers (like pcata) may not support EFI
3790 * GPT labels. Print out a helpful error message dir-
3791 * ecting the user to manually label the disk and give
3792 * a specific slice.
3793 */
3794 (void) close(fd);
3795 efi_free(vtoc);
3796
d603ed6c
BB		 3797 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "try using "
3798 "parted(8) and then provide a specific slice: %d"), rval);
34dc7c2f
BB		 3799 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
3800 }
3801
3802 (void) close(fd);
3803 efi_free(vtoc);
34dc7c2f 3804
d603ed6c
BB		 3805 /* Wait for the first expected slice to appear. */
3806 (void) snprintf(path, sizeof (path), "%s/%s%s%s", DISK_ROOT, name,
3807 isdigit(name[strlen(name)-1]) ? "p" : "", FIRST_SLICE);
3808 rval = zpool_label_disk_wait(path, 3000);
3809 if (rval) {
3810 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "failed to "
3811 "detect device partitions on '%s': %d"), path, rval);
3812 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
34dc7c2f
BB		 3813 }
3814
d603ed6c
BB		 3815 /* We can't be to paranoid. Read the label back and verify it. */
3816 (void) snprintf(path, sizeof (path), "%s/%s", DISK_ROOT, name);
3817 rval = zpool_label_disk_check(path);
3818 if (rval) {
3819 zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "freshly written "
3820 "EFI label on '%s' is damaged. Ensure\nthis device "
3821 "is not in in use, and is functioning properly: %d"),
3822 path, rval);
3823 return (zfs_error(hdl, EZFS_LABELFAILED, errbuf));
34dc7c2f 3824 }
34dc7c2f 3825
d603ed6c 3826 return 0;
34dc7c2f 3827}
Mirror of ZFSOnLinux for Debian