]>
Commit | Line | Data |
---|---|---|
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 |
47 | static int read_efi_label(nvlist_t *config, diskaddr_t *sb); |
48 | ||
572e2857 BB |
49 | typedef 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 | ||
60 | static int | |
61 | zpool_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 | ||
93 | static int | |
94 | zpool_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 | ||
107 | static char * | |
108 | zpool_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 | ||
133 | uint64_t | |
134 | zpool_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 | */ | |
175 | char * | |
176 | zpool_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 | */ | |
208 | int | |
209 | zpool_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: | |
df30f566 | 273 | case ZPOOL_PROP_ASHIFT: |
34dc7c2f BB |
274 | (void) zfs_nicenum(intval, buf, len); |
275 | break; | |
276 | ||
277 | case ZPOOL_PROP_CAPACITY: | |
278 | (void) snprintf(buf, len, "%llu%%", | |
279 | (u_longlong_t)intval); | |
280 | break; | |
281 | ||
428870ff BB |
282 | case ZPOOL_PROP_DEDUPRATIO: |
283 | (void) snprintf(buf, len, "%llu.%02llux", | |
284 | (u_longlong_t)(intval / 100), | |
285 | (u_longlong_t)(intval % 100)); | |
286 | break; | |
287 | ||
34dc7c2f BB |
288 | case ZPOOL_PROP_HEALTH: |
289 | verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL), | |
290 | ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); | |
291 | verify(nvlist_lookup_uint64_array(nvroot, | |
428870ff BB |
292 | ZPOOL_CONFIG_VDEV_STATS, (uint64_t **)&vs, &vsc) |
293 | == 0); | |
34dc7c2f BB |
294 | |
295 | (void) strlcpy(buf, zpool_state_to_name(intval, | |
296 | vs->vs_aux), len); | |
297 | break; | |
298 | default: | |
b8864a23 | 299 | (void) snprintf(buf, len, "%llu", (u_longlong_t)intval); |
34dc7c2f BB |
300 | } |
301 | break; | |
302 | ||
303 | case PROP_TYPE_INDEX: | |
304 | intval = zpool_get_prop_int(zhp, prop, &src); | |
305 | if (zpool_prop_index_to_string(prop, intval, &strval) | |
306 | != 0) | |
307 | return (-1); | |
308 | (void) strlcpy(buf, strval, len); | |
309 | break; | |
310 | ||
311 | default: | |
312 | abort(); | |
313 | } | |
314 | ||
315 | if (srctype) | |
316 | *srctype = src; | |
317 | ||
318 | return (0); | |
319 | } | |
320 | ||
321 | /* | |
322 | * Check if the bootfs name has the same pool name as it is set to. | |
323 | * Assuming bootfs is a valid dataset name. | |
324 | */ | |
325 | static boolean_t | |
326 | bootfs_name_valid(const char *pool, char *bootfs) | |
327 | { | |
328 | int len = strlen(pool); | |
329 | ||
b128c09f | 330 | if (!zfs_name_valid(bootfs, ZFS_TYPE_FILESYSTEM|ZFS_TYPE_SNAPSHOT)) |
34dc7c2f BB |
331 | return (B_FALSE); |
332 | ||
333 | if (strncmp(pool, bootfs, len) == 0 && | |
334 | (bootfs[len] == '/' || bootfs[len] == '\0')) | |
335 | return (B_TRUE); | |
336 | ||
337 | return (B_FALSE); | |
338 | } | |
339 | ||
b128c09f BB |
340 | /* |
341 | * Inspect the configuration to determine if any of the devices contain | |
342 | * an EFI label. | |
343 | */ | |
344 | static boolean_t | |
345 | pool_uses_efi(nvlist_t *config) | |
346 | { | |
347 | nvlist_t **child; | |
348 | uint_t c, children; | |
349 | ||
350 | if (nvlist_lookup_nvlist_array(config, ZPOOL_CONFIG_CHILDREN, | |
351 | &child, &children) != 0) | |
352 | return (read_efi_label(config, NULL) >= 0); | |
353 | ||
354 | for (c = 0; c < children; c++) { | |
355 | if (pool_uses_efi(child[c])) | |
356 | return (B_TRUE); | |
357 | } | |
358 | return (B_FALSE); | |
359 | } | |
360 | ||
361 | static boolean_t | |
362 | pool_is_bootable(zpool_handle_t *zhp) | |
363 | { | |
364 | char bootfs[ZPOOL_MAXNAMELEN]; | |
365 | ||
366 | return (zpool_get_prop(zhp, ZPOOL_PROP_BOOTFS, bootfs, | |
367 | sizeof (bootfs), NULL) == 0 && strncmp(bootfs, "-", | |
368 | sizeof (bootfs)) != 0); | |
369 | } | |
370 | ||
371 | ||
34dc7c2f BB |
372 | /* |
373 | * Given an nvlist of zpool properties to be set, validate that they are | |
374 | * correct, and parse any numeric properties (index, boolean, etc) if they are | |
375 | * specified as strings. | |
376 | */ | |
377 | static nvlist_t * | |
b128c09f | 378 | zpool_valid_proplist(libzfs_handle_t *hdl, const char *poolname, |
572e2857 | 379 | nvlist_t *props, uint64_t version, prop_flags_t flags, char *errbuf) |
34dc7c2f BB |
380 | { |
381 | nvpair_t *elem; | |
382 | nvlist_t *retprops; | |
383 | zpool_prop_t prop; | |
384 | char *strval; | |
385 | uint64_t intval; | |
386 | char *slash; | |
387 | struct stat64 statbuf; | |
b128c09f BB |
388 | zpool_handle_t *zhp; |
389 | nvlist_t *nvroot; | |
34dc7c2f BB |
390 | |
391 | if (nvlist_alloc(&retprops, NV_UNIQUE_NAME, 0) != 0) { | |
392 | (void) no_memory(hdl); | |
393 | return (NULL); | |
394 | } | |
395 | ||
396 | elem = NULL; | |
397 | while ((elem = nvlist_next_nvpair(props, elem)) != NULL) { | |
398 | const char *propname = nvpair_name(elem); | |
399 | ||
400 | /* | |
401 | * Make sure this property is valid and applies to this type. | |
402 | */ | |
403 | if ((prop = zpool_name_to_prop(propname)) == ZPROP_INVAL) { | |
404 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
405 | "invalid property '%s'"), propname); | |
406 | (void) zfs_error(hdl, EZFS_BADPROP, errbuf); | |
407 | goto error; | |
408 | } | |
409 | ||
410 | if (zpool_prop_readonly(prop)) { | |
411 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' " | |
412 | "is readonly"), propname); | |
413 | (void) zfs_error(hdl, EZFS_PROPREADONLY, errbuf); | |
414 | goto error; | |
415 | } | |
416 | ||
417 | if (zprop_parse_value(hdl, elem, prop, ZFS_TYPE_POOL, retprops, | |
418 | &strval, &intval, errbuf) != 0) | |
419 | goto error; | |
420 | ||
421 | /* | |
422 | * Perform additional checking for specific properties. | |
423 | */ | |
424 | switch (prop) { | |
e75c13c3 BB |
425 | default: |
426 | break; | |
34dc7c2f BB |
427 | case ZPOOL_PROP_VERSION: |
428 | if (intval < version || intval > SPA_VERSION) { | |
429 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
430 | "property '%s' number %d is invalid."), | |
431 | propname, intval); | |
432 | (void) zfs_error(hdl, EZFS_BADVERSION, errbuf); | |
433 | goto error; | |
434 | } | |
435 | break; | |
436 | ||
df30f566 CK |
437 | case ZPOOL_PROP_ASHIFT: |
438 | if (!flags.create) { | |
439 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
440 | "property '%s' can only be set at " | |
441 | "creation time"), propname); | |
442 | (void) zfs_error(hdl, EZFS_BADPROP, errbuf); | |
443 | goto error; | |
444 | } | |
445 | ||
b41c9906 | 446 | if (intval != 0 && (intval < 9 || intval > 13)) { |
df30f566 CK |
447 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, |
448 | "property '%s' number %d is invalid."), | |
449 | propname, intval); | |
450 | (void) zfs_error(hdl, EZFS_BADPROP, errbuf); | |
451 | goto error; | |
452 | } | |
453 | break; | |
454 | ||
34dc7c2f | 455 | case ZPOOL_PROP_BOOTFS: |
572e2857 | 456 | if (flags.create || flags.import) { |
34dc7c2f BB |
457 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, |
458 | "property '%s' cannot be set at creation " | |
459 | "or import time"), propname); | |
460 | (void) zfs_error(hdl, EZFS_BADPROP, errbuf); | |
461 | goto error; | |
462 | } | |
463 | ||
464 | if (version < SPA_VERSION_BOOTFS) { | |
465 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
466 | "pool must be upgraded to support " | |
467 | "'%s' property"), propname); | |
468 | (void) zfs_error(hdl, EZFS_BADVERSION, errbuf); | |
469 | goto error; | |
470 | } | |
471 | ||
472 | /* | |
473 | * bootfs property value has to be a dataset name and | |
474 | * the dataset has to be in the same pool as it sets to. | |
475 | */ | |
476 | if (strval[0] != '\0' && !bootfs_name_valid(poolname, | |
477 | strval)) { | |
478 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "'%s' " | |
479 | "is an invalid name"), strval); | |
480 | (void) zfs_error(hdl, EZFS_INVALIDNAME, errbuf); | |
481 | goto error; | |
482 | } | |
b128c09f BB |
483 | |
484 | if ((zhp = zpool_open_canfail(hdl, poolname)) == NULL) { | |
485 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
486 | "could not open pool '%s'"), poolname); | |
487 | (void) zfs_error(hdl, EZFS_OPENFAILED, errbuf); | |
488 | goto error; | |
489 | } | |
490 | verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL), | |
491 | ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); | |
492 | ||
f783130a | 493 | #if defined(__sun__) || defined(__sun) |
b128c09f BB |
494 | /* |
495 | * bootfs property cannot be set on a disk which has | |
496 | * been EFI labeled. | |
497 | */ | |
498 | if (pool_uses_efi(nvroot)) { | |
499 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
500 | "property '%s' not supported on " | |
501 | "EFI labeled devices"), propname); | |
502 | (void) zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf); | |
503 | zpool_close(zhp); | |
504 | goto error; | |
505 | } | |
f783130a | 506 | #endif |
b128c09f | 507 | zpool_close(zhp); |
34dc7c2f BB |
508 | break; |
509 | ||
510 | case ZPOOL_PROP_ALTROOT: | |
572e2857 | 511 | if (!flags.create && !flags.import) { |
34dc7c2f BB |
512 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, |
513 | "property '%s' can only be set during pool " | |
514 | "creation or import"), propname); | |
515 | (void) zfs_error(hdl, EZFS_BADPROP, errbuf); | |
516 | goto error; | |
517 | } | |
518 | ||
519 | if (strval[0] != '/') { | |
520 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
521 | "bad alternate root '%s'"), strval); | |
522 | (void) zfs_error(hdl, EZFS_BADPATH, errbuf); | |
523 | goto error; | |
524 | } | |
525 | break; | |
526 | ||
527 | case ZPOOL_PROP_CACHEFILE: | |
528 | if (strval[0] == '\0') | |
529 | break; | |
530 | ||
531 | if (strcmp(strval, "none") == 0) | |
532 | break; | |
533 | ||
534 | if (strval[0] != '/') { | |
535 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
536 | "property '%s' must be empty, an " | |
537 | "absolute path, or 'none'"), propname); | |
538 | (void) zfs_error(hdl, EZFS_BADPATH, errbuf); | |
539 | goto error; | |
540 | } | |
541 | ||
542 | slash = strrchr(strval, '/'); | |
543 | ||
544 | if (slash[1] == '\0' || strcmp(slash, "/.") == 0 || | |
545 | strcmp(slash, "/..") == 0) { | |
546 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
547 | "'%s' is not a valid file"), strval); | |
548 | (void) zfs_error(hdl, EZFS_BADPATH, errbuf); | |
549 | goto error; | |
550 | } | |
551 | ||
552 | *slash = '\0'; | |
553 | ||
554 | if (strval[0] != '\0' && | |
555 | (stat64(strval, &statbuf) != 0 || | |
556 | !S_ISDIR(statbuf.st_mode))) { | |
557 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
558 | "'%s' is not a valid directory"), | |
559 | strval); | |
560 | (void) zfs_error(hdl, EZFS_BADPATH, errbuf); | |
561 | goto error; | |
562 | } | |
563 | ||
564 | *slash = '/'; | |
565 | break; | |
572e2857 BB |
566 | |
567 | case ZPOOL_PROP_READONLY: | |
568 | if (!flags.import) { | |
569 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
570 | "property '%s' can only be set at " | |
571 | "import time"), propname); | |
572 | (void) zfs_error(hdl, EZFS_BADPROP, errbuf); | |
573 | goto error; | |
574 | } | |
575 | break; | |
34dc7c2f BB |
576 | } |
577 | } | |
578 | ||
579 | return (retprops); | |
580 | error: | |
581 | nvlist_free(retprops); | |
582 | return (NULL); | |
583 | } | |
584 | ||
585 | /* | |
586 | * Set zpool property : propname=propval. | |
587 | */ | |
588 | int | |
589 | zpool_set_prop(zpool_handle_t *zhp, const char *propname, const char *propval) | |
590 | { | |
2598c001 | 591 | zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 }; |
34dc7c2f BB |
592 | int ret = -1; |
593 | char errbuf[1024]; | |
594 | nvlist_t *nvl = NULL; | |
595 | nvlist_t *realprops; | |
596 | uint64_t version; | |
572e2857 | 597 | prop_flags_t flags = { 0 }; |
34dc7c2f BB |
598 | |
599 | (void) snprintf(errbuf, sizeof (errbuf), | |
600 | dgettext(TEXT_DOMAIN, "cannot set property for '%s'"), | |
601 | zhp->zpool_name); | |
602 | ||
34dc7c2f BB |
603 | if (nvlist_alloc(&nvl, NV_UNIQUE_NAME, 0) != 0) |
604 | return (no_memory(zhp->zpool_hdl)); | |
605 | ||
606 | if (nvlist_add_string(nvl, propname, propval) != 0) { | |
607 | nvlist_free(nvl); | |
608 | return (no_memory(zhp->zpool_hdl)); | |
609 | } | |
610 | ||
611 | version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL); | |
b128c09f | 612 | if ((realprops = zpool_valid_proplist(zhp->zpool_hdl, |
572e2857 | 613 | zhp->zpool_name, nvl, version, flags, errbuf)) == NULL) { |
34dc7c2f BB |
614 | nvlist_free(nvl); |
615 | return (-1); | |
616 | } | |
617 | ||
618 | nvlist_free(nvl); | |
619 | nvl = realprops; | |
620 | ||
621 | /* | |
622 | * Execute the corresponding ioctl() to set this property. | |
623 | */ | |
624 | (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); | |
625 | ||
626 | if (zcmd_write_src_nvlist(zhp->zpool_hdl, &zc, nvl) != 0) { | |
627 | nvlist_free(nvl); | |
628 | return (-1); | |
629 | } | |
630 | ||
631 | ret = zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_SET_PROPS, &zc); | |
632 | ||
633 | zcmd_free_nvlists(&zc); | |
634 | nvlist_free(nvl); | |
635 | ||
636 | if (ret) | |
637 | (void) zpool_standard_error(zhp->zpool_hdl, errno, errbuf); | |
638 | else | |
639 | (void) zpool_props_refresh(zhp); | |
640 | ||
641 | return (ret); | |
642 | } | |
643 | ||
644 | int | |
645 | zpool_expand_proplist(zpool_handle_t *zhp, zprop_list_t **plp) | |
646 | { | |
647 | libzfs_handle_t *hdl = zhp->zpool_hdl; | |
648 | zprop_list_t *entry; | |
649 | char buf[ZFS_MAXPROPLEN]; | |
650 | ||
651 | if (zprop_expand_list(hdl, plp, ZFS_TYPE_POOL) != 0) | |
652 | return (-1); | |
653 | ||
654 | for (entry = *plp; entry != NULL; entry = entry->pl_next) { | |
655 | ||
656 | if (entry->pl_fixed) | |
657 | continue; | |
658 | ||
659 | if (entry->pl_prop != ZPROP_INVAL && | |
660 | zpool_get_prop(zhp, entry->pl_prop, buf, sizeof (buf), | |
661 | NULL) == 0) { | |
662 | if (strlen(buf) > entry->pl_width) | |
663 | entry->pl_width = strlen(buf); | |
664 | } | |
665 | } | |
666 | ||
667 | return (0); | |
668 | } | |
669 | ||
670 | ||
9babb374 BB |
671 | /* |
672 | * Don't start the slice at the default block of 34; many storage | |
d603ed6c BB |
673 | * devices will use a stripe width of 128k, other vendors prefer a 1m |
674 | * alignment. It is best to play it safe and ensure a 1m alignment | |
613d88ed NB |
675 | * given 512B blocks. When the block size is larger by a power of 2 |
676 | * we will still be 1m aligned. Some devices are sensitive to the | |
677 | * partition ending alignment as well. | |
9babb374 | 678 | */ |
613d88ed NB |
679 | #define NEW_START_BLOCK 2048 |
680 | #define PARTITION_END_ALIGNMENT 2048 | |
9babb374 | 681 | |
34dc7c2f BB |
682 | /* |
683 | * Validate the given pool name, optionally putting an extended error message in | |
684 | * 'buf'. | |
685 | */ | |
686 | boolean_t | |
687 | zpool_name_valid(libzfs_handle_t *hdl, boolean_t isopen, const char *pool) | |
688 | { | |
689 | namecheck_err_t why; | |
690 | char what; | |
691 | int ret; | |
692 | ||
693 | ret = pool_namecheck(pool, &why, &what); | |
694 | ||
695 | /* | |
696 | * The rules for reserved pool names were extended at a later point. | |
697 | * But we need to support users with existing pools that may now be | |
698 | * invalid. So we only check for this expanded set of names during a | |
699 | * create (or import), and only in userland. | |
700 | */ | |
701 | if (ret == 0 && !isopen && | |
702 | (strncmp(pool, "mirror", 6) == 0 || | |
703 | strncmp(pool, "raidz", 5) == 0 || | |
704 | strncmp(pool, "spare", 5) == 0 || | |
705 | strcmp(pool, "log") == 0)) { | |
706 | if (hdl != NULL) | |
707 | zfs_error_aux(hdl, | |
708 | dgettext(TEXT_DOMAIN, "name is reserved")); | |
709 | return (B_FALSE); | |
710 | } | |
711 | ||
712 | ||
713 | if (ret != 0) { | |
714 | if (hdl != NULL) { | |
715 | switch (why) { | |
716 | case NAME_ERR_TOOLONG: | |
717 | zfs_error_aux(hdl, | |
718 | dgettext(TEXT_DOMAIN, "name is too long")); | |
719 | break; | |
720 | ||
721 | case NAME_ERR_INVALCHAR: | |
722 | zfs_error_aux(hdl, | |
723 | dgettext(TEXT_DOMAIN, "invalid character " | |
724 | "'%c' in pool name"), what); | |
725 | break; | |
726 | ||
727 | case NAME_ERR_NOLETTER: | |
728 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
729 | "name must begin with a letter")); | |
730 | break; | |
731 | ||
732 | case NAME_ERR_RESERVED: | |
733 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
734 | "name is reserved")); | |
735 | break; | |
736 | ||
737 | case NAME_ERR_DISKLIKE: | |
738 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
739 | "pool name is reserved")); | |
740 | break; | |
741 | ||
742 | case NAME_ERR_LEADING_SLASH: | |
743 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
744 | "leading slash in name")); | |
745 | break; | |
746 | ||
747 | case NAME_ERR_EMPTY_COMPONENT: | |
748 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
749 | "empty component in name")); | |
750 | break; | |
751 | ||
752 | case NAME_ERR_TRAILING_SLASH: | |
753 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
754 | "trailing slash in name")); | |
755 | break; | |
756 | ||
757 | case NAME_ERR_MULTIPLE_AT: | |
758 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
759 | "multiple '@' delimiters in name")); | |
760 | break; | |
e75c13c3 BB |
761 | case NAME_ERR_NO_AT: |
762 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
763 | "permission set is missing '@'")); | |
764 | break; | |
34dc7c2f BB |
765 | } |
766 | } | |
767 | return (B_FALSE); | |
768 | } | |
769 | ||
770 | return (B_TRUE); | |
771 | } | |
772 | ||
773 | /* | |
774 | * Open a handle to the given pool, even if the pool is currently in the FAULTED | |
775 | * state. | |
776 | */ | |
777 | zpool_handle_t * | |
778 | zpool_open_canfail(libzfs_handle_t *hdl, const char *pool) | |
779 | { | |
780 | zpool_handle_t *zhp; | |
781 | boolean_t missing; | |
782 | ||
783 | /* | |
784 | * Make sure the pool name is valid. | |
785 | */ | |
786 | if (!zpool_name_valid(hdl, B_TRUE, pool)) { | |
787 | (void) zfs_error_fmt(hdl, EZFS_INVALIDNAME, | |
788 | dgettext(TEXT_DOMAIN, "cannot open '%s'"), | |
789 | pool); | |
790 | return (NULL); | |
791 | } | |
792 | ||
793 | if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL) | |
794 | return (NULL); | |
795 | ||
796 | zhp->zpool_hdl = hdl; | |
797 | (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name)); | |
798 | ||
799 | if (zpool_refresh_stats(zhp, &missing) != 0) { | |
800 | zpool_close(zhp); | |
801 | return (NULL); | |
802 | } | |
803 | ||
804 | if (missing) { | |
805 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "no such pool")); | |
806 | (void) zfs_error_fmt(hdl, EZFS_NOENT, | |
807 | dgettext(TEXT_DOMAIN, "cannot open '%s'"), pool); | |
808 | zpool_close(zhp); | |
809 | return (NULL); | |
810 | } | |
811 | ||
812 | return (zhp); | |
813 | } | |
814 | ||
815 | /* | |
816 | * Like the above, but silent on error. Used when iterating over pools (because | |
817 | * the configuration cache may be out of date). | |
818 | */ | |
819 | int | |
820 | zpool_open_silent(libzfs_handle_t *hdl, const char *pool, zpool_handle_t **ret) | |
821 | { | |
822 | zpool_handle_t *zhp; | |
823 | boolean_t missing; | |
824 | ||
825 | if ((zhp = zfs_alloc(hdl, sizeof (zpool_handle_t))) == NULL) | |
826 | return (-1); | |
827 | ||
828 | zhp->zpool_hdl = hdl; | |
829 | (void) strlcpy(zhp->zpool_name, pool, sizeof (zhp->zpool_name)); | |
830 | ||
831 | if (zpool_refresh_stats(zhp, &missing) != 0) { | |
832 | zpool_close(zhp); | |
833 | return (-1); | |
834 | } | |
835 | ||
836 | if (missing) { | |
837 | zpool_close(zhp); | |
838 | *ret = NULL; | |
839 | return (0); | |
840 | } | |
841 | ||
842 | *ret = zhp; | |
843 | return (0); | |
844 | } | |
845 | ||
846 | /* | |
847 | * Similar to zpool_open_canfail(), but refuses to open pools in the faulted | |
848 | * state. | |
849 | */ | |
850 | zpool_handle_t * | |
851 | zpool_open(libzfs_handle_t *hdl, const char *pool) | |
852 | { | |
853 | zpool_handle_t *zhp; | |
854 | ||
855 | if ((zhp = zpool_open_canfail(hdl, pool)) == NULL) | |
856 | return (NULL); | |
857 | ||
858 | if (zhp->zpool_state == POOL_STATE_UNAVAIL) { | |
859 | (void) zfs_error_fmt(hdl, EZFS_POOLUNAVAIL, | |
860 | dgettext(TEXT_DOMAIN, "cannot open '%s'"), zhp->zpool_name); | |
861 | zpool_close(zhp); | |
862 | return (NULL); | |
863 | } | |
864 | ||
865 | return (zhp); | |
866 | } | |
867 | ||
868 | /* | |
869 | * Close the handle. Simply frees the memory associated with the handle. | |
870 | */ | |
871 | void | |
872 | zpool_close(zpool_handle_t *zhp) | |
873 | { | |
874 | if (zhp->zpool_config) | |
875 | nvlist_free(zhp->zpool_config); | |
876 | if (zhp->zpool_old_config) | |
877 | nvlist_free(zhp->zpool_old_config); | |
878 | if (zhp->zpool_props) | |
879 | nvlist_free(zhp->zpool_props); | |
880 | free(zhp); | |
881 | } | |
882 | ||
883 | /* | |
884 | * Return the name of the pool. | |
885 | */ | |
886 | const char * | |
887 | zpool_get_name(zpool_handle_t *zhp) | |
888 | { | |
889 | return (zhp->zpool_name); | |
890 | } | |
891 | ||
892 | ||
893 | /* | |
894 | * Return the state of the pool (ACTIVE or UNAVAILABLE) | |
895 | */ | |
896 | int | |
897 | zpool_get_state(zpool_handle_t *zhp) | |
898 | { | |
899 | return (zhp->zpool_state); | |
900 | } | |
901 | ||
902 | /* | |
903 | * Create the named pool, using the provided vdev list. It is assumed | |
904 | * that the consumer has already validated the contents of the nvlist, so we | |
905 | * don't have to worry about error semantics. | |
906 | */ | |
907 | int | |
908 | zpool_create(libzfs_handle_t *hdl, const char *pool, nvlist_t *nvroot, | |
b128c09f | 909 | nvlist_t *props, nvlist_t *fsprops) |
34dc7c2f | 910 | { |
2598c001 | 911 | zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 }; |
b128c09f BB |
912 | nvlist_t *zc_fsprops = NULL; |
913 | nvlist_t *zc_props = NULL; | |
34dc7c2f BB |
914 | char msg[1024]; |
915 | char *altroot; | |
b128c09f | 916 | int ret = -1; |
34dc7c2f BB |
917 | |
918 | (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, | |
919 | "cannot create '%s'"), pool); | |
920 | ||
921 | if (!zpool_name_valid(hdl, B_FALSE, pool)) | |
922 | return (zfs_error(hdl, EZFS_INVALIDNAME, msg)); | |
923 | ||
924 | if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0) | |
925 | return (-1); | |
926 | ||
b128c09f | 927 | if (props) { |
572e2857 BB |
928 | prop_flags_t flags = { .create = B_TRUE, .import = B_FALSE }; |
929 | ||
b128c09f | 930 | if ((zc_props = zpool_valid_proplist(hdl, pool, props, |
572e2857 | 931 | SPA_VERSION_1, flags, msg)) == NULL) { |
b128c09f BB |
932 | goto create_failed; |
933 | } | |
934 | } | |
34dc7c2f | 935 | |
b128c09f BB |
936 | if (fsprops) { |
937 | uint64_t zoned; | |
938 | char *zonestr; | |
939 | ||
940 | zoned = ((nvlist_lookup_string(fsprops, | |
941 | zfs_prop_to_name(ZFS_PROP_ZONED), &zonestr) == 0) && | |
942 | strcmp(zonestr, "on") == 0); | |
943 | ||
944 | if ((zc_fsprops = zfs_valid_proplist(hdl, | |
945 | ZFS_TYPE_FILESYSTEM, fsprops, zoned, NULL, msg)) == NULL) { | |
946 | goto create_failed; | |
947 | } | |
948 | if (!zc_props && | |
949 | (nvlist_alloc(&zc_props, NV_UNIQUE_NAME, 0) != 0)) { | |
950 | goto create_failed; | |
951 | } | |
952 | if (nvlist_add_nvlist(zc_props, | |
953 | ZPOOL_ROOTFS_PROPS, zc_fsprops) != 0) { | |
954 | goto create_failed; | |
955 | } | |
34dc7c2f BB |
956 | } |
957 | ||
b128c09f BB |
958 | if (zc_props && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0) |
959 | goto create_failed; | |
960 | ||
34dc7c2f BB |
961 | (void) strlcpy(zc.zc_name, pool, sizeof (zc.zc_name)); |
962 | ||
b128c09f | 963 | if ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_CREATE, &zc)) != 0) { |
34dc7c2f BB |
964 | |
965 | zcmd_free_nvlists(&zc); | |
b128c09f BB |
966 | nvlist_free(zc_props); |
967 | nvlist_free(zc_fsprops); | |
34dc7c2f BB |
968 | |
969 | switch (errno) { | |
970 | case EBUSY: | |
971 | /* | |
972 | * This can happen if the user has specified the same | |
973 | * device multiple times. We can't reliably detect this | |
974 | * until we try to add it and see we already have a | |
d603ed6c BB |
975 | * label. This can also happen under if the device is |
976 | * part of an active md or lvm device. | |
34dc7c2f BB |
977 | */ |
978 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
d603ed6c BB |
979 | "one or more vdevs refer to the same device, or one of\n" |
980 | "the devices is part of an active md or lvm device")); | |
34dc7c2f BB |
981 | return (zfs_error(hdl, EZFS_BADDEV, msg)); |
982 | ||
983 | case EOVERFLOW: | |
984 | /* | |
985 | * This occurs when one of the devices is below | |
986 | * SPA_MINDEVSIZE. Unfortunately, we can't detect which | |
987 | * device was the problem device since there's no | |
988 | * reliable way to determine device size from userland. | |
989 | */ | |
990 | { | |
991 | char buf[64]; | |
992 | ||
993 | zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf)); | |
994 | ||
995 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
996 | "one or more devices is less than the " | |
997 | "minimum size (%s)"), buf); | |
998 | } | |
999 | return (zfs_error(hdl, EZFS_BADDEV, msg)); | |
1000 | ||
1001 | case ENOSPC: | |
1002 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
1003 | "one or more devices is out of space")); | |
1004 | return (zfs_error(hdl, EZFS_BADDEV, msg)); | |
1005 | ||
1006 | case ENOTBLK: | |
1007 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
1008 | "cache device must be a disk or disk slice")); | |
1009 | return (zfs_error(hdl, EZFS_BADDEV, msg)); | |
1010 | ||
1011 | default: | |
1012 | return (zpool_standard_error(hdl, errno, msg)); | |
1013 | } | |
1014 | } | |
1015 | ||
1016 | /* | |
1017 | * If this is an alternate root pool, then we automatically set the | |
1018 | * mountpoint of the root dataset to be '/'. | |
1019 | */ | |
1020 | if (nvlist_lookup_string(props, zpool_prop_to_name(ZPOOL_PROP_ALTROOT), | |
1021 | &altroot) == 0) { | |
1022 | zfs_handle_t *zhp; | |
1023 | ||
1024 | verify((zhp = zfs_open(hdl, pool, ZFS_TYPE_DATASET)) != NULL); | |
1025 | verify(zfs_prop_set(zhp, zfs_prop_to_name(ZFS_PROP_MOUNTPOINT), | |
1026 | "/") == 0); | |
1027 | ||
1028 | zfs_close(zhp); | |
1029 | } | |
1030 | ||
b128c09f | 1031 | create_failed: |
34dc7c2f | 1032 | zcmd_free_nvlists(&zc); |
b128c09f BB |
1033 | nvlist_free(zc_props); |
1034 | nvlist_free(zc_fsprops); | |
1035 | return (ret); | |
34dc7c2f BB |
1036 | } |
1037 | ||
1038 | /* | |
1039 | * Destroy the given pool. It is up to the caller to ensure that there are no | |
1040 | * datasets left in the pool. | |
1041 | */ | |
1042 | int | |
1043 | zpool_destroy(zpool_handle_t *zhp) | |
1044 | { | |
2598c001 | 1045 | zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 }; |
34dc7c2f BB |
1046 | zfs_handle_t *zfp = NULL; |
1047 | libzfs_handle_t *hdl = zhp->zpool_hdl; | |
1048 | char msg[1024]; | |
1049 | ||
1050 | if (zhp->zpool_state == POOL_STATE_ACTIVE && | |
572e2857 | 1051 | (zfp = zfs_open(hdl, zhp->zpool_name, ZFS_TYPE_FILESYSTEM)) == NULL) |
34dc7c2f BB |
1052 | return (-1); |
1053 | ||
34dc7c2f BB |
1054 | (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); |
1055 | ||
572e2857 | 1056 | if (zfs_ioctl(hdl, ZFS_IOC_POOL_DESTROY, &zc) != 0) { |
34dc7c2f BB |
1057 | (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, |
1058 | "cannot destroy '%s'"), zhp->zpool_name); | |
1059 | ||
1060 | if (errno == EROFS) { | |
1061 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
1062 | "one or more devices is read only")); | |
1063 | (void) zfs_error(hdl, EZFS_BADDEV, msg); | |
1064 | } else { | |
1065 | (void) zpool_standard_error(hdl, errno, msg); | |
1066 | } | |
1067 | ||
1068 | if (zfp) | |
1069 | zfs_close(zfp); | |
1070 | return (-1); | |
1071 | } | |
1072 | ||
1073 | if (zfp) { | |
1074 | remove_mountpoint(zfp); | |
1075 | zfs_close(zfp); | |
1076 | } | |
1077 | ||
1078 | return (0); | |
1079 | } | |
1080 | ||
1081 | /* | |
1082 | * Add the given vdevs to the pool. The caller must have already performed the | |
1083 | * necessary verification to ensure that the vdev specification is well-formed. | |
1084 | */ | |
1085 | int | |
1086 | zpool_add(zpool_handle_t *zhp, nvlist_t *nvroot) | |
1087 | { | |
2598c001 | 1088 | zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 }; |
34dc7c2f BB |
1089 | int ret; |
1090 | libzfs_handle_t *hdl = zhp->zpool_hdl; | |
1091 | char msg[1024]; | |
1092 | nvlist_t **spares, **l2cache; | |
1093 | uint_t nspares, nl2cache; | |
1094 | ||
1095 | (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, | |
1096 | "cannot add to '%s'"), zhp->zpool_name); | |
1097 | ||
1098 | if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) < | |
1099 | SPA_VERSION_SPARES && | |
1100 | nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, | |
1101 | &spares, &nspares) == 0) { | |
1102 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be " | |
1103 | "upgraded to add hot spares")); | |
1104 | return (zfs_error(hdl, EZFS_BADVERSION, msg)); | |
1105 | } | |
1106 | ||
b128c09f BB |
1107 | if (pool_is_bootable(zhp) && nvlist_lookup_nvlist_array(nvroot, |
1108 | ZPOOL_CONFIG_SPARES, &spares, &nspares) == 0) { | |
1109 | uint64_t s; | |
1110 | ||
1111 | for (s = 0; s < nspares; s++) { | |
1112 | char *path; | |
1113 | ||
1114 | if (nvlist_lookup_string(spares[s], ZPOOL_CONFIG_PATH, | |
1115 | &path) == 0 && pool_uses_efi(spares[s])) { | |
1116 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
1117 | "device '%s' contains an EFI label and " | |
1118 | "cannot be used on root pools."), | |
428870ff BB |
1119 | zpool_vdev_name(hdl, NULL, spares[s], |
1120 | B_FALSE)); | |
b128c09f BB |
1121 | return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg)); |
1122 | } | |
1123 | } | |
1124 | } | |
1125 | ||
34dc7c2f BB |
1126 | if (zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL) < |
1127 | SPA_VERSION_L2CACHE && | |
1128 | nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, | |
1129 | &l2cache, &nl2cache) == 0) { | |
1130 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "pool must be " | |
1131 | "upgraded to add cache devices")); | |
1132 | return (zfs_error(hdl, EZFS_BADVERSION, msg)); | |
1133 | } | |
1134 | ||
1135 | if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0) | |
1136 | return (-1); | |
1137 | (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); | |
1138 | ||
572e2857 | 1139 | if (zfs_ioctl(hdl, ZFS_IOC_VDEV_ADD, &zc) != 0) { |
34dc7c2f BB |
1140 | switch (errno) { |
1141 | case EBUSY: | |
1142 | /* | |
1143 | * This can happen if the user has specified the same | |
1144 | * device multiple times. We can't reliably detect this | |
1145 | * until we try to add it and see we already have a | |
1146 | * label. | |
1147 | */ | |
1148 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
1149 | "one or more vdevs refer to the same device")); | |
1150 | (void) zfs_error(hdl, EZFS_BADDEV, msg); | |
1151 | break; | |
1152 | ||
1153 | case EOVERFLOW: | |
1154 | /* | |
1155 | * This occurrs when one of the devices is below | |
1156 | * SPA_MINDEVSIZE. Unfortunately, we can't detect which | |
1157 | * device was the problem device since there's no | |
1158 | * reliable way to determine device size from userland. | |
1159 | */ | |
1160 | { | |
1161 | char buf[64]; | |
1162 | ||
1163 | zfs_nicenum(SPA_MINDEVSIZE, buf, sizeof (buf)); | |
1164 | ||
1165 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
1166 | "device is less than the minimum " | |
1167 | "size (%s)"), buf); | |
1168 | } | |
1169 | (void) zfs_error(hdl, EZFS_BADDEV, msg); | |
1170 | break; | |
1171 | ||
1172 | case ENOTSUP: | |
1173 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
1174 | "pool must be upgraded to add these vdevs")); | |
1175 | (void) zfs_error(hdl, EZFS_BADVERSION, msg); | |
1176 | break; | |
1177 | ||
1178 | case EDOM: | |
1179 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
1180 | "root pool can not have multiple vdevs" | |
1181 | " or separate logs")); | |
1182 | (void) zfs_error(hdl, EZFS_POOL_NOTSUP, msg); | |
1183 | break; | |
1184 | ||
1185 | case ENOTBLK: | |
1186 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
1187 | "cache device must be a disk or disk slice")); | |
1188 | (void) zfs_error(hdl, EZFS_BADDEV, msg); | |
1189 | break; | |
1190 | ||
1191 | default: | |
1192 | (void) zpool_standard_error(hdl, errno, msg); | |
1193 | } | |
1194 | ||
1195 | ret = -1; | |
1196 | } else { | |
1197 | ret = 0; | |
1198 | } | |
1199 | ||
1200 | zcmd_free_nvlists(&zc); | |
1201 | ||
1202 | return (ret); | |
1203 | } | |
1204 | ||
1205 | /* | |
1206 | * Exports the pool from the system. The caller must ensure that there are no | |
1207 | * mounted datasets in the pool. | |
1208 | */ | |
1209 | int | |
fb5f0bc8 | 1210 | zpool_export_common(zpool_handle_t *zhp, boolean_t force, boolean_t hardforce) |
34dc7c2f | 1211 | { |
2598c001 | 1212 | zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 }; |
b128c09f | 1213 | char msg[1024]; |
34dc7c2f | 1214 | |
b128c09f BB |
1215 | (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, |
1216 | "cannot export '%s'"), zhp->zpool_name); | |
1217 | ||
34dc7c2f | 1218 | (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); |
b128c09f | 1219 | zc.zc_cookie = force; |
fb5f0bc8 | 1220 | zc.zc_guid = hardforce; |
b128c09f BB |
1221 | |
1222 | if (zfs_ioctl(zhp->zpool_hdl, ZFS_IOC_POOL_EXPORT, &zc) != 0) { | |
1223 | switch (errno) { | |
1224 | case EXDEV: | |
1225 | zfs_error_aux(zhp->zpool_hdl, dgettext(TEXT_DOMAIN, | |
1226 | "use '-f' to override the following errors:\n" | |
1227 | "'%s' has an active shared spare which could be" | |
1228 | " used by other pools once '%s' is exported."), | |
1229 | zhp->zpool_name, zhp->zpool_name); | |
1230 | return (zfs_error(zhp->zpool_hdl, EZFS_ACTIVE_SPARE, | |
1231 | msg)); | |
1232 | default: | |
1233 | return (zpool_standard_error_fmt(zhp->zpool_hdl, errno, | |
1234 | msg)); | |
1235 | } | |
1236 | } | |
34dc7c2f | 1237 | |
34dc7c2f BB |
1238 | return (0); |
1239 | } | |
1240 | ||
fb5f0bc8 BB |
1241 | int |
1242 | zpool_export(zpool_handle_t *zhp, boolean_t force) | |
1243 | { | |
1244 | return (zpool_export_common(zhp, force, B_FALSE)); | |
1245 | } | |
1246 | ||
1247 | int | |
1248 | zpool_export_force(zpool_handle_t *zhp) | |
1249 | { | |
1250 | return (zpool_export_common(zhp, B_TRUE, B_TRUE)); | |
1251 | } | |
1252 | ||
428870ff BB |
1253 | static void |
1254 | zpool_rewind_exclaim(libzfs_handle_t *hdl, const char *name, boolean_t dryrun, | |
572e2857 | 1255 | nvlist_t *config) |
428870ff | 1256 | { |
572e2857 | 1257 | nvlist_t *nv = NULL; |
428870ff BB |
1258 | uint64_t rewindto; |
1259 | int64_t loss = -1; | |
1260 | struct tm t; | |
1261 | char timestr[128]; | |
1262 | ||
572e2857 BB |
1263 | if (!hdl->libzfs_printerr || config == NULL) |
1264 | return; | |
1265 | ||
1266 | if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0) | |
428870ff BB |
1267 | return; |
1268 | ||
572e2857 | 1269 | if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0) |
428870ff | 1270 | return; |
572e2857 | 1271 | (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss); |
428870ff BB |
1272 | |
1273 | if (localtime_r((time_t *)&rewindto, &t) != NULL && | |
b8864a23 | 1274 | strftime(timestr, 128, "%c", &t) != 0) { |
428870ff BB |
1275 | if (dryrun) { |
1276 | (void) printf(dgettext(TEXT_DOMAIN, | |
1277 | "Would be able to return %s " | |
1278 | "to its state as of %s.\n"), | |
1279 | name, timestr); | |
1280 | } else { | |
1281 | (void) printf(dgettext(TEXT_DOMAIN, | |
1282 | "Pool %s returned to its state as of %s.\n"), | |
1283 | name, timestr); | |
1284 | } | |
1285 | if (loss > 120) { | |
1286 | (void) printf(dgettext(TEXT_DOMAIN, | |
1287 | "%s approximately %lld "), | |
1288 | dryrun ? "Would discard" : "Discarded", | |
b8864a23 | 1289 | ((longlong_t)loss + 30) / 60); |
428870ff BB |
1290 | (void) printf(dgettext(TEXT_DOMAIN, |
1291 | "minutes of transactions.\n")); | |
1292 | } else if (loss > 0) { | |
1293 | (void) printf(dgettext(TEXT_DOMAIN, | |
1294 | "%s approximately %lld "), | |
b8864a23 BB |
1295 | dryrun ? "Would discard" : "Discarded", |
1296 | (longlong_t)loss); | |
428870ff BB |
1297 | (void) printf(dgettext(TEXT_DOMAIN, |
1298 | "seconds of transactions.\n")); | |
1299 | } | |
1300 | } | |
1301 | } | |
1302 | ||
1303 | void | |
1304 | zpool_explain_recover(libzfs_handle_t *hdl, const char *name, int reason, | |
1305 | nvlist_t *config) | |
1306 | { | |
572e2857 | 1307 | nvlist_t *nv = NULL; |
428870ff BB |
1308 | int64_t loss = -1; |
1309 | uint64_t edata = UINT64_MAX; | |
1310 | uint64_t rewindto; | |
1311 | struct tm t; | |
1312 | char timestr[128]; | |
1313 | ||
1314 | if (!hdl->libzfs_printerr) | |
1315 | return; | |
1316 | ||
1317 | if (reason >= 0) | |
1318 | (void) printf(dgettext(TEXT_DOMAIN, "action: ")); | |
1319 | else | |
1320 | (void) printf(dgettext(TEXT_DOMAIN, "\t")); | |
1321 | ||
1322 | /* All attempted rewinds failed if ZPOOL_CONFIG_LOAD_TIME missing */ | |
572e2857 BB |
1323 | if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_LOAD_INFO, &nv) != 0 || |
1324 | nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_TIME, &rewindto) != 0) | |
428870ff BB |
1325 | goto no_info; |
1326 | ||
572e2857 BB |
1327 | (void) nvlist_lookup_int64(nv, ZPOOL_CONFIG_REWIND_TIME, &loss); |
1328 | (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_LOAD_DATA_ERRORS, | |
428870ff BB |
1329 | &edata); |
1330 | ||
1331 | (void) printf(dgettext(TEXT_DOMAIN, | |
1332 | "Recovery is possible, but will result in some data loss.\n")); | |
1333 | ||
1334 | if (localtime_r((time_t *)&rewindto, &t) != NULL && | |
b8864a23 | 1335 | strftime(timestr, 128, "%c", &t) != 0) { |
428870ff BB |
1336 | (void) printf(dgettext(TEXT_DOMAIN, |
1337 | "\tReturning the pool to its state as of %s\n" | |
1338 | "\tshould correct the problem. "), | |
1339 | timestr); | |
1340 | } else { | |
1341 | (void) printf(dgettext(TEXT_DOMAIN, | |
1342 | "\tReverting the pool to an earlier state " | |
1343 | "should correct the problem.\n\t")); | |
1344 | } | |
1345 | ||
1346 | if (loss > 120) { | |
1347 | (void) printf(dgettext(TEXT_DOMAIN, | |
1348 | "Approximately %lld minutes of data\n" | |
b8864a23 BB |
1349 | "\tmust be discarded, irreversibly. "), |
1350 | ((longlong_t)loss + 30) / 60); | |
428870ff BB |
1351 | } else if (loss > 0) { |
1352 | (void) printf(dgettext(TEXT_DOMAIN, | |
1353 | "Approximately %lld seconds of data\n" | |
b8864a23 BB |
1354 | "\tmust be discarded, irreversibly. "), |
1355 | (longlong_t)loss); | |
428870ff BB |
1356 | } |
1357 | if (edata != 0 && edata != UINT64_MAX) { | |
1358 | if (edata == 1) { | |
1359 | (void) printf(dgettext(TEXT_DOMAIN, | |
1360 | "After rewind, at least\n" | |
1361 | "\tone persistent user-data error will remain. ")); | |
1362 | } else { | |
1363 | (void) printf(dgettext(TEXT_DOMAIN, | |
1364 | "After rewind, several\n" | |
1365 | "\tpersistent user-data errors will remain. ")); | |
1366 | } | |
1367 | } | |
1368 | (void) printf(dgettext(TEXT_DOMAIN, | |
1369 | "Recovery can be attempted\n\tby executing 'zpool %s -F %s'. "), | |
1370 | reason >= 0 ? "clear" : "import", name); | |
1371 | ||
1372 | (void) printf(dgettext(TEXT_DOMAIN, | |
1373 | "A scrub of the pool\n" | |
1374 | "\tis strongly recommended after recovery.\n")); | |
1375 | return; | |
1376 | ||
1377 | no_info: | |
1378 | (void) printf(dgettext(TEXT_DOMAIN, | |
1379 | "Destroy and re-create the pool from\n\ta backup source.\n")); | |
1380 | } | |
1381 | ||
34dc7c2f BB |
1382 | /* |
1383 | * zpool_import() is a contracted interface. Should be kept the same | |
1384 | * if possible. | |
1385 | * | |
1386 | * Applications should use zpool_import_props() to import a pool with | |
1387 | * new properties value to be set. | |
1388 | */ | |
1389 | int | |
1390 | zpool_import(libzfs_handle_t *hdl, nvlist_t *config, const char *newname, | |
1391 | char *altroot) | |
1392 | { | |
1393 | nvlist_t *props = NULL; | |
1394 | int ret; | |
1395 | ||
1396 | if (altroot != NULL) { | |
1397 | if (nvlist_alloc(&props, NV_UNIQUE_NAME, 0) != 0) { | |
1398 | return (zfs_error_fmt(hdl, EZFS_NOMEM, | |
1399 | dgettext(TEXT_DOMAIN, "cannot import '%s'"), | |
1400 | newname)); | |
1401 | } | |
1402 | ||
1403 | if (nvlist_add_string(props, | |
fb5f0bc8 BB |
1404 | zpool_prop_to_name(ZPOOL_PROP_ALTROOT), altroot) != 0 || |
1405 | nvlist_add_string(props, | |
1406 | zpool_prop_to_name(ZPOOL_PROP_CACHEFILE), "none") != 0) { | |
34dc7c2f BB |
1407 | nvlist_free(props); |
1408 | return (zfs_error_fmt(hdl, EZFS_NOMEM, | |
1409 | dgettext(TEXT_DOMAIN, "cannot import '%s'"), | |
1410 | newname)); | |
1411 | } | |
1412 | } | |
1413 | ||
572e2857 BB |
1414 | ret = zpool_import_props(hdl, config, newname, props, |
1415 | ZFS_IMPORT_NORMAL); | |
34dc7c2f BB |
1416 | if (props) |
1417 | nvlist_free(props); | |
1418 | return (ret); | |
1419 | } | |
1420 | ||
572e2857 BB |
1421 | static void |
1422 | print_vdev_tree(libzfs_handle_t *hdl, const char *name, nvlist_t *nv, | |
1423 | int indent) | |
1424 | { | |
1425 | nvlist_t **child; | |
1426 | uint_t c, children; | |
1427 | char *vname; | |
1428 | uint64_t is_log = 0; | |
1429 | ||
1430 | (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG, | |
1431 | &is_log); | |
1432 | ||
1433 | if (name != NULL) | |
1434 | (void) printf("\t%*s%s%s\n", indent, "", name, | |
1435 | is_log ? " [log]" : ""); | |
1436 | ||
1437 | if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, | |
1438 | &child, &children) != 0) | |
1439 | return; | |
1440 | ||
1441 | for (c = 0; c < children; c++) { | |
1442 | vname = zpool_vdev_name(hdl, NULL, child[c], B_TRUE); | |
1443 | print_vdev_tree(hdl, vname, child[c], indent + 2); | |
1444 | free(vname); | |
1445 | } | |
1446 | } | |
1447 | ||
34dc7c2f BB |
1448 | /* |
1449 | * Import the given pool using the known configuration and a list of | |
1450 | * properties to be set. The configuration should have come from | |
1451 | * zpool_find_import(). The 'newname' parameters control whether the pool | |
1452 | * is imported with a different name. | |
1453 | */ | |
1454 | int | |
1455 | zpool_import_props(libzfs_handle_t *hdl, nvlist_t *config, const char *newname, | |
572e2857 | 1456 | nvlist_t *props, int flags) |
34dc7c2f | 1457 | { |
2598c001 | 1458 | zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 }; |
428870ff | 1459 | zpool_rewind_policy_t policy; |
572e2857 BB |
1460 | nvlist_t *nv = NULL; |
1461 | nvlist_t *nvinfo = NULL; | |
1462 | nvlist_t *missing = NULL; | |
34dc7c2f BB |
1463 | char *thename; |
1464 | char *origname; | |
1465 | int ret; | |
572e2857 | 1466 | int error = 0; |
34dc7c2f BB |
1467 | char errbuf[1024]; |
1468 | ||
1469 | verify(nvlist_lookup_string(config, ZPOOL_CONFIG_POOL_NAME, | |
1470 | &origname) == 0); | |
1471 | ||
1472 | (void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN, | |
1473 | "cannot import pool '%s'"), origname); | |
1474 | ||
1475 | if (newname != NULL) { | |
1476 | if (!zpool_name_valid(hdl, B_FALSE, newname)) | |
1477 | return (zfs_error_fmt(hdl, EZFS_INVALIDNAME, | |
1478 | dgettext(TEXT_DOMAIN, "cannot import '%s'"), | |
1479 | newname)); | |
1480 | thename = (char *)newname; | |
1481 | } else { | |
1482 | thename = origname; | |
1483 | } | |
1484 | ||
1485 | if (props) { | |
1486 | uint64_t version; | |
572e2857 | 1487 | prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE }; |
34dc7c2f BB |
1488 | |
1489 | verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, | |
1490 | &version) == 0); | |
1491 | ||
b128c09f | 1492 | if ((props = zpool_valid_proplist(hdl, origname, |
572e2857 | 1493 | props, version, flags, errbuf)) == NULL) { |
34dc7c2f BB |
1494 | return (-1); |
1495 | } else if (zcmd_write_src_nvlist(hdl, &zc, props) != 0) { | |
1496 | nvlist_free(props); | |
1497 | return (-1); | |
1498 | } | |
1499 | } | |
1500 | ||
1501 | (void) strlcpy(zc.zc_name, thename, sizeof (zc.zc_name)); | |
1502 | ||
1503 | verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_POOL_GUID, | |
1504 | &zc.zc_guid) == 0); | |
1505 | ||
1506 | if (zcmd_write_conf_nvlist(hdl, &zc, config) != 0) { | |
1507 | nvlist_free(props); | |
1508 | return (-1); | |
1509 | } | |
572e2857 | 1510 | if (zcmd_alloc_dst_nvlist(hdl, &zc, zc.zc_nvlist_conf_size * 2) != 0) { |
428870ff BB |
1511 | nvlist_free(props); |
1512 | return (-1); | |
1513 | } | |
34dc7c2f | 1514 | |
572e2857 BB |
1515 | zc.zc_cookie = flags; |
1516 | while ((ret = zfs_ioctl(hdl, ZFS_IOC_POOL_IMPORT, &zc)) != 0 && | |
1517 | errno == ENOMEM) { | |
1518 | if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { | |
1519 | zcmd_free_nvlists(&zc); | |
1520 | return (-1); | |
1521 | } | |
1522 | } | |
1523 | if (ret != 0) | |
1524 | error = errno; | |
1525 | ||
1526 | (void) zcmd_read_dst_nvlist(hdl, &zc, &nv); | |
1527 | zpool_get_rewind_policy(config, &policy); | |
1528 | ||
1529 | if (error) { | |
34dc7c2f | 1530 | char desc[1024]; |
428870ff | 1531 | |
428870ff BB |
1532 | /* |
1533 | * Dry-run failed, but we print out what success | |
1534 | * looks like if we found a best txg | |
1535 | */ | |
572e2857 | 1536 | if (policy.zrp_request & ZPOOL_TRY_REWIND) { |
428870ff | 1537 | zpool_rewind_exclaim(hdl, newname ? origname : thename, |
572e2857 BB |
1538 | B_TRUE, nv); |
1539 | nvlist_free(nv); | |
428870ff BB |
1540 | return (-1); |
1541 | } | |
1542 | ||
34dc7c2f BB |
1543 | if (newname == NULL) |
1544 | (void) snprintf(desc, sizeof (desc), | |
1545 | dgettext(TEXT_DOMAIN, "cannot import '%s'"), | |
1546 | thename); | |
1547 | else | |
1548 | (void) snprintf(desc, sizeof (desc), | |
1549 | dgettext(TEXT_DOMAIN, "cannot import '%s' as '%s'"), | |
1550 | origname, thename); | |
1551 | ||
572e2857 | 1552 | switch (error) { |
34dc7c2f BB |
1553 | case ENOTSUP: |
1554 | /* | |
1555 | * Unsupported version. | |
1556 | */ | |
1557 | (void) zfs_error(hdl, EZFS_BADVERSION, desc); | |
1558 | break; | |
1559 | ||
1560 | case EINVAL: | |
1561 | (void) zfs_error(hdl, EZFS_INVALCONFIG, desc); | |
1562 | break; | |
1563 | ||
428870ff BB |
1564 | case EROFS: |
1565 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
1566 | "one or more devices is read only")); | |
1567 | (void) zfs_error(hdl, EZFS_BADDEV, desc); | |
1568 | break; | |
1569 | ||
572e2857 BB |
1570 | case ENXIO: |
1571 | if (nv && nvlist_lookup_nvlist(nv, | |
1572 | ZPOOL_CONFIG_LOAD_INFO, &nvinfo) == 0 && | |
1573 | nvlist_lookup_nvlist(nvinfo, | |
1574 | ZPOOL_CONFIG_MISSING_DEVICES, &missing) == 0) { | |
1575 | (void) printf(dgettext(TEXT_DOMAIN, | |
1576 | "The devices below are missing, use " | |
1577 | "'-m' to import the pool anyway:\n")); | |
1578 | print_vdev_tree(hdl, NULL, missing, 2); | |
1579 | (void) printf("\n"); | |
1580 | } | |
1581 | (void) zpool_standard_error(hdl, error, desc); | |
1582 | break; | |
1583 | ||
1584 | case EEXIST: | |
1585 | (void) zpool_standard_error(hdl, error, desc); | |
1586 | break; | |
1587 | ||
34dc7c2f | 1588 | default: |
572e2857 | 1589 | (void) zpool_standard_error(hdl, error, desc); |
428870ff | 1590 | zpool_explain_recover(hdl, |
572e2857 | 1591 | newname ? origname : thename, -error, nv); |
428870ff | 1592 | break; |
34dc7c2f BB |
1593 | } |
1594 | ||
572e2857 | 1595 | nvlist_free(nv); |
34dc7c2f BB |
1596 | ret = -1; |
1597 | } else { | |
1598 | zpool_handle_t *zhp; | |
1599 | ||
1600 | /* | |
1601 | * This should never fail, but play it safe anyway. | |
1602 | */ | |
428870ff | 1603 | if (zpool_open_silent(hdl, thename, &zhp) != 0) |
34dc7c2f | 1604 | ret = -1; |
428870ff | 1605 | else if (zhp != NULL) |
34dc7c2f | 1606 | zpool_close(zhp); |
428870ff BB |
1607 | if (policy.zrp_request & |
1608 | (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) { | |
1609 | zpool_rewind_exclaim(hdl, newname ? origname : thename, | |
572e2857 | 1610 | ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0), nv); |
34dc7c2f | 1611 | } |
572e2857 | 1612 | nvlist_free(nv); |
428870ff | 1613 | return (0); |
34dc7c2f BB |
1614 | } |
1615 | ||
1616 | zcmd_free_nvlists(&zc); | |
1617 | nvlist_free(props); | |
1618 | ||
1619 | return (ret); | |
1620 | } | |
1621 | ||
1622 | /* | |
428870ff | 1623 | * Scan the pool. |
34dc7c2f BB |
1624 | */ |
1625 | int | |
428870ff | 1626 | zpool_scan(zpool_handle_t *zhp, pool_scan_func_t func) |
34dc7c2f | 1627 | { |
2598c001 | 1628 | zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 }; |
34dc7c2f BB |
1629 | char msg[1024]; |
1630 | libzfs_handle_t *hdl = zhp->zpool_hdl; | |
1631 | ||
1632 | (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); | |
428870ff | 1633 | zc.zc_cookie = func; |
34dc7c2f | 1634 | |
572e2857 | 1635 | if (zfs_ioctl(hdl, ZFS_IOC_POOL_SCAN, &zc) == 0 || |
428870ff | 1636 | (errno == ENOENT && func != POOL_SCAN_NONE)) |
34dc7c2f BB |
1637 | return (0); |
1638 | ||
428870ff BB |
1639 | if (func == POOL_SCAN_SCRUB) { |
1640 | (void) snprintf(msg, sizeof (msg), | |
1641 | dgettext(TEXT_DOMAIN, "cannot scrub %s"), zc.zc_name); | |
1642 | } else if (func == POOL_SCAN_NONE) { | |
1643 | (void) snprintf(msg, sizeof (msg), | |
1644 | dgettext(TEXT_DOMAIN, "cannot cancel scrubbing %s"), | |
1645 | zc.zc_name); | |
1646 | } else { | |
1647 | assert(!"unexpected result"); | |
1648 | } | |
34dc7c2f | 1649 | |
428870ff BB |
1650 | if (errno == EBUSY) { |
1651 | nvlist_t *nvroot; | |
1652 | pool_scan_stat_t *ps = NULL; | |
1653 | uint_t psc; | |
1654 | ||
1655 | verify(nvlist_lookup_nvlist(zhp->zpool_config, | |
1656 | ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); | |
1657 | (void) nvlist_lookup_uint64_array(nvroot, | |
1658 | ZPOOL_CONFIG_SCAN_STATS, (uint64_t **)&ps, &psc); | |
1659 | if (ps && ps->pss_func == POOL_SCAN_SCRUB) | |
1660 | return (zfs_error(hdl, EZFS_SCRUBBING, msg)); | |
1661 | else | |
1662 | return (zfs_error(hdl, EZFS_RESILVERING, msg)); | |
1663 | } else if (errno == ENOENT) { | |
1664 | return (zfs_error(hdl, EZFS_NO_SCRUB, msg)); | |
1665 | } else { | |
34dc7c2f | 1666 | return (zpool_standard_error(hdl, errno, msg)); |
428870ff BB |
1667 | } |
1668 | } | |
1669 | ||
34dc7c2f | 1670 | /* |
9babb374 BB |
1671 | * Find a vdev that matches the search criteria specified. We use the |
1672 | * the nvpair name to determine how we should look for the device. | |
34dc7c2f BB |
1673 | * 'avail_spare' is set to TRUE if the provided guid refers to an AVAIL |
1674 | * spare; but FALSE if its an INUSE spare. | |
1675 | */ | |
1676 | static nvlist_t * | |
9babb374 BB |
1677 | vdev_to_nvlist_iter(nvlist_t *nv, nvlist_t *search, boolean_t *avail_spare, |
1678 | boolean_t *l2cache, boolean_t *log) | |
34dc7c2f BB |
1679 | { |
1680 | uint_t c, children; | |
1681 | nvlist_t **child; | |
34dc7c2f | 1682 | nvlist_t *ret; |
b128c09f | 1683 | uint64_t is_log; |
9babb374 BB |
1684 | char *srchkey; |
1685 | nvpair_t *pair = nvlist_next_nvpair(search, NULL); | |
1686 | ||
1687 | /* Nothing to look for */ | |
1688 | if (search == NULL || pair == NULL) | |
1689 | return (NULL); | |
1690 | ||
1691 | /* Obtain the key we will use to search */ | |
1692 | srchkey = nvpair_name(pair); | |
1693 | ||
1694 | switch (nvpair_type(pair)) { | |
572e2857 | 1695 | case DATA_TYPE_UINT64: |
9babb374 | 1696 | if (strcmp(srchkey, ZPOOL_CONFIG_GUID) == 0) { |
572e2857 BB |
1697 | uint64_t srchval, theguid; |
1698 | ||
1699 | verify(nvpair_value_uint64(pair, &srchval) == 0); | |
1700 | verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, | |
1701 | &theguid) == 0); | |
1702 | if (theguid == srchval) | |
1703 | return (nv); | |
9babb374 BB |
1704 | } |
1705 | break; | |
9babb374 BB |
1706 | |
1707 | case DATA_TYPE_STRING: { | |
1708 | char *srchval, *val; | |
1709 | ||
1710 | verify(nvpair_value_string(pair, &srchval) == 0); | |
1711 | if (nvlist_lookup_string(nv, srchkey, &val) != 0) | |
1712 | break; | |
34dc7c2f | 1713 | |
9babb374 | 1714 | /* |
428870ff BB |
1715 | * Search for the requested value. Special cases: |
1716 | * | |
a2c6816c NB |
1717 | * - ZPOOL_CONFIG_PATH for whole disk entries. These end in with a |
1718 | * partition suffix "1", "-part1", or "p1". The suffix is hidden | |
1719 | * from the user, but included in the string, so this matches around | |
1720 | * it. | |
428870ff BB |
1721 | * - looking for a top-level vdev name (i.e. ZPOOL_CONFIG_TYPE). |
1722 | * | |
1723 | * Otherwise, all other searches are simple string compares. | |
9babb374 | 1724 | */ |
a2c6816c | 1725 | if (strcmp(srchkey, ZPOOL_CONFIG_PATH) == 0) { |
9babb374 BB |
1726 | uint64_t wholedisk = 0; |
1727 | ||
1728 | (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, | |
1729 | &wholedisk); | |
1730 | if (wholedisk) { | |
a2c6816c | 1731 | char buf[MAXPATHLEN]; |
428870ff | 1732 | |
a2c6816c NB |
1733 | zfs_append_partition(srchval, buf, sizeof (buf)); |
1734 | if (strcmp(val, buf) == 0) | |
428870ff BB |
1735 | return (nv); |
1736 | ||
9babb374 BB |
1737 | break; |
1738 | } | |
428870ff BB |
1739 | } else if (strcmp(srchkey, ZPOOL_CONFIG_TYPE) == 0 && val) { |
1740 | char *type, *idx, *end, *p; | |
1741 | uint64_t id, vdev_id; | |
1742 | ||
1743 | /* | |
1744 | * Determine our vdev type, keeping in mind | |
1745 | * that the srchval is composed of a type and | |
1746 | * vdev id pair (i.e. mirror-4). | |
1747 | */ | |
1748 | if ((type = strdup(srchval)) == NULL) | |
1749 | return (NULL); | |
1750 | ||
1751 | if ((p = strrchr(type, '-')) == NULL) { | |
1752 | free(type); | |
1753 | break; | |
1754 | } | |
1755 | idx = p + 1; | |
1756 | *p = '\0'; | |
1757 | ||
1758 | /* | |
1759 | * If the types don't match then keep looking. | |
1760 | */ | |
1761 | if (strncmp(val, type, strlen(val)) != 0) { | |
1762 | free(type); | |
1763 | break; | |
1764 | } | |
1765 | ||
1766 | verify(strncmp(type, VDEV_TYPE_RAIDZ, | |
1767 | strlen(VDEV_TYPE_RAIDZ)) == 0 || | |
1768 | strncmp(type, VDEV_TYPE_MIRROR, | |
1769 | strlen(VDEV_TYPE_MIRROR)) == 0); | |
1770 | verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID, | |
1771 | &id) == 0); | |
1772 | ||
1773 | errno = 0; | |
1774 | vdev_id = strtoull(idx, &end, 10); | |
1775 | ||
1776 | free(type); | |
1777 | if (errno != 0) | |
1778 | return (NULL); | |
1779 | ||
1780 | /* | |
1781 | * Now verify that we have the correct vdev id. | |
1782 | */ | |
1783 | if (vdev_id == id) | |
1784 | return (nv); | |
9babb374 | 1785 | } |
34dc7c2f | 1786 | |
34dc7c2f | 1787 | /* |
9babb374 | 1788 | * Common case |
34dc7c2f | 1789 | */ |
9babb374 | 1790 | if (strcmp(srchval, val) == 0) |
34dc7c2f | 1791 | return (nv); |
9babb374 BB |
1792 | break; |
1793 | } | |
1794 | ||
1795 | default: | |
1796 | break; | |
34dc7c2f BB |
1797 | } |
1798 | ||
1799 | if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, | |
1800 | &child, &children) != 0) | |
1801 | return (NULL); | |
1802 | ||
b128c09f | 1803 | for (c = 0; c < children; c++) { |
9babb374 | 1804 | if ((ret = vdev_to_nvlist_iter(child[c], search, |
b128c09f BB |
1805 | avail_spare, l2cache, NULL)) != NULL) { |
1806 | /* | |
1807 | * The 'is_log' value is only set for the toplevel | |
1808 | * vdev, not the leaf vdevs. So we always lookup the | |
1809 | * log device from the root of the vdev tree (where | |
1810 | * 'log' is non-NULL). | |
1811 | */ | |
1812 | if (log != NULL && | |
1813 | nvlist_lookup_uint64(child[c], | |
1814 | ZPOOL_CONFIG_IS_LOG, &is_log) == 0 && | |
1815 | is_log) { | |
1816 | *log = B_TRUE; | |
1817 | } | |
34dc7c2f | 1818 | return (ret); |
b128c09f BB |
1819 | } |
1820 | } | |
34dc7c2f BB |
1821 | |
1822 | if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES, | |
1823 | &child, &children) == 0) { | |
1824 | for (c = 0; c < children; c++) { | |
9babb374 | 1825 | if ((ret = vdev_to_nvlist_iter(child[c], search, |
b128c09f | 1826 | avail_spare, l2cache, NULL)) != NULL) { |
34dc7c2f BB |
1827 | *avail_spare = B_TRUE; |
1828 | return (ret); | |
1829 | } | |
1830 | } | |
1831 | } | |
1832 | ||
1833 | if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE, | |
1834 | &child, &children) == 0) { | |
1835 | for (c = 0; c < children; c++) { | |
9babb374 | 1836 | if ((ret = vdev_to_nvlist_iter(child[c], search, |
b128c09f | 1837 | avail_spare, l2cache, NULL)) != NULL) { |
34dc7c2f BB |
1838 | *l2cache = B_TRUE; |
1839 | return (ret); | |
1840 | } | |
1841 | } | |
1842 | } | |
1843 | ||
1844 | return (NULL); | |
1845 | } | |
1846 | ||
9babb374 BB |
1847 | /* |
1848 | * Given a physical path (minus the "/devices" prefix), find the | |
1849 | * associated vdev. | |
1850 | */ | |
1851 | nvlist_t * | |
1852 | zpool_find_vdev_by_physpath(zpool_handle_t *zhp, const char *ppath, | |
1853 | boolean_t *avail_spare, boolean_t *l2cache, boolean_t *log) | |
1854 | { | |
1855 | nvlist_t *search, *nvroot, *ret; | |
1856 | ||
1857 | verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0); | |
1858 | verify(nvlist_add_string(search, ZPOOL_CONFIG_PHYS_PATH, ppath) == 0); | |
1859 | ||
1860 | verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE, | |
1861 | &nvroot) == 0); | |
1862 | ||
1863 | *avail_spare = B_FALSE; | |
572e2857 BB |
1864 | *l2cache = B_FALSE; |
1865 | if (log != NULL) | |
1866 | *log = B_FALSE; | |
9babb374 BB |
1867 | ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log); |
1868 | nvlist_free(search); | |
1869 | ||
1870 | return (ret); | |
1871 | } | |
1872 | ||
428870ff BB |
1873 | /* |
1874 | * Determine if we have an "interior" top-level vdev (i.e mirror/raidz). | |
1875 | */ | |
1876 | boolean_t | |
1877 | zpool_vdev_is_interior(const char *name) | |
1878 | { | |
1879 | if (strncmp(name, VDEV_TYPE_RAIDZ, strlen(VDEV_TYPE_RAIDZ)) == 0 || | |
1880 | strncmp(name, VDEV_TYPE_MIRROR, strlen(VDEV_TYPE_MIRROR)) == 0) | |
1881 | return (B_TRUE); | |
1882 | return (B_FALSE); | |
1883 | } | |
1884 | ||
34dc7c2f BB |
1885 | nvlist_t * |
1886 | zpool_find_vdev(zpool_handle_t *zhp, const char *path, boolean_t *avail_spare, | |
b128c09f | 1887 | boolean_t *l2cache, boolean_t *log) |
34dc7c2f BB |
1888 | { |
1889 | char buf[MAXPATHLEN]; | |
34dc7c2f | 1890 | char *end; |
9babb374 | 1891 | nvlist_t *nvroot, *search, *ret; |
34dc7c2f BB |
1892 | uint64_t guid; |
1893 | ||
9babb374 BB |
1894 | verify(nvlist_alloc(&search, NV_UNIQUE_NAME, KM_SLEEP) == 0); |
1895 | ||
34dc7c2f BB |
1896 | guid = strtoull(path, &end, 10); |
1897 | if (guid != 0 && *end == '\0') { | |
9babb374 | 1898 | verify(nvlist_add_uint64(search, ZPOOL_CONFIG_GUID, guid) == 0); |
428870ff BB |
1899 | } else if (zpool_vdev_is_interior(path)) { |
1900 | verify(nvlist_add_string(search, ZPOOL_CONFIG_TYPE, path) == 0); | |
34dc7c2f | 1901 | } else if (path[0] != '/') { |
a2c6816c NB |
1902 | if (zfs_resolve_shortname(path, buf, sizeof (buf)) < 0) { |
1903 | nvlist_free(search); | |
1904 | return (NULL); | |
1905 | } | |
9babb374 | 1906 | verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, buf) == 0); |
34dc7c2f | 1907 | } else { |
9babb374 | 1908 | verify(nvlist_add_string(search, ZPOOL_CONFIG_PATH, path) == 0); |
34dc7c2f BB |
1909 | } |
1910 | ||
1911 | verify(nvlist_lookup_nvlist(zhp->zpool_config, ZPOOL_CONFIG_VDEV_TREE, | |
1912 | &nvroot) == 0); | |
1913 | ||
1914 | *avail_spare = B_FALSE; | |
1915 | *l2cache = B_FALSE; | |
b128c09f BB |
1916 | if (log != NULL) |
1917 | *log = B_FALSE; | |
9babb374 BB |
1918 | ret = vdev_to_nvlist_iter(nvroot, search, avail_spare, l2cache, log); |
1919 | nvlist_free(search); | |
1920 | ||
1921 | return (ret); | |
b128c09f BB |
1922 | } |
1923 | ||
1924 | static int | |
1925 | vdev_online(nvlist_t *nv) | |
1926 | { | |
1927 | uint64_t ival; | |
1928 | ||
1929 | if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_OFFLINE, &ival) == 0 || | |
1930 | nvlist_lookup_uint64(nv, ZPOOL_CONFIG_FAULTED, &ival) == 0 || | |
1931 | nvlist_lookup_uint64(nv, ZPOOL_CONFIG_REMOVED, &ival) == 0) | |
1932 | return (0); | |
1933 | ||
1934 | return (1); | |
1935 | } | |
1936 | ||
1937 | /* | |
9babb374 | 1938 | * Helper function for zpool_get_physpaths(). |
b128c09f | 1939 | */ |
9babb374 BB |
1940 | static int |
1941 | vdev_get_one_physpath(nvlist_t *config, char *physpath, size_t physpath_size, | |
1942 | size_t *bytes_written) | |
1943 | { | |
1944 | size_t bytes_left, pos, rsz; | |
1945 | char *tmppath; | |
1946 | const char *format; | |
1947 | ||
1948 | if (nvlist_lookup_string(config, ZPOOL_CONFIG_PHYS_PATH, | |
1949 | &tmppath) != 0) | |
1950 | return (EZFS_NODEVICE); | |
1951 | ||
1952 | pos = *bytes_written; | |
1953 | bytes_left = physpath_size - pos; | |
1954 | format = (pos == 0) ? "%s" : " %s"; | |
1955 | ||
1956 | rsz = snprintf(physpath + pos, bytes_left, format, tmppath); | |
1957 | *bytes_written += rsz; | |
1958 | ||
1959 | if (rsz >= bytes_left) { | |
1960 | /* if physpath was not copied properly, clear it */ | |
1961 | if (bytes_left != 0) { | |
1962 | physpath[pos] = 0; | |
1963 | } | |
1964 | return (EZFS_NOSPC); | |
1965 | } | |
1966 | return (0); | |
1967 | } | |
1968 | ||
1969 | static int | |
1970 | vdev_get_physpaths(nvlist_t *nv, char *physpath, size_t phypath_size, | |
1971 | size_t *rsz, boolean_t is_spare) | |
1972 | { | |
1973 | char *type; | |
1974 | int ret; | |
1975 | ||
1976 | if (nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) != 0) | |
1977 | return (EZFS_INVALCONFIG); | |
1978 | ||
1979 | if (strcmp(type, VDEV_TYPE_DISK) == 0) { | |
1980 | /* | |
1981 | * An active spare device has ZPOOL_CONFIG_IS_SPARE set. | |
1982 | * For a spare vdev, we only want to boot from the active | |
1983 | * spare device. | |
1984 | */ | |
1985 | if (is_spare) { | |
1986 | uint64_t spare = 0; | |
1987 | (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_SPARE, | |
1988 | &spare); | |
1989 | if (!spare) | |
1990 | return (EZFS_INVALCONFIG); | |
1991 | } | |
1992 | ||
1993 | if (vdev_online(nv)) { | |
1994 | if ((ret = vdev_get_one_physpath(nv, physpath, | |
1995 | phypath_size, rsz)) != 0) | |
1996 | return (ret); | |
1997 | } | |
1998 | } else if (strcmp(type, VDEV_TYPE_MIRROR) == 0 || | |
1999 | strcmp(type, VDEV_TYPE_REPLACING) == 0 || | |
2000 | (is_spare = (strcmp(type, VDEV_TYPE_SPARE) == 0))) { | |
2001 | nvlist_t **child; | |
2002 | uint_t count; | |
2003 | int i, ret; | |
2004 | ||
2005 | if (nvlist_lookup_nvlist_array(nv, | |
2006 | ZPOOL_CONFIG_CHILDREN, &child, &count) != 0) | |
2007 | return (EZFS_INVALCONFIG); | |
2008 | ||
2009 | for (i = 0; i < count; i++) { | |
2010 | ret = vdev_get_physpaths(child[i], physpath, | |
2011 | phypath_size, rsz, is_spare); | |
2012 | if (ret == EZFS_NOSPC) | |
2013 | return (ret); | |
2014 | } | |
2015 | } | |
2016 | ||
2017 | return (EZFS_POOL_INVALARG); | |
2018 | } | |
2019 | ||
2020 | /* | |
2021 | * Get phys_path for a root pool config. | |
2022 | * Return 0 on success; non-zero on failure. | |
2023 | */ | |
2024 | static int | |
2025 | zpool_get_config_physpath(nvlist_t *config, char *physpath, size_t phypath_size) | |
b128c09f | 2026 | { |
9babb374 | 2027 | size_t rsz; |
b128c09f BB |
2028 | nvlist_t *vdev_root; |
2029 | nvlist_t **child; | |
2030 | uint_t count; | |
9babb374 | 2031 | char *type; |
b128c09f | 2032 | |
9babb374 | 2033 | rsz = 0; |
b128c09f | 2034 | |
9babb374 BB |
2035 | if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, |
2036 | &vdev_root) != 0) | |
2037 | return (EZFS_INVALCONFIG); | |
b128c09f | 2038 | |
9babb374 BB |
2039 | if (nvlist_lookup_string(vdev_root, ZPOOL_CONFIG_TYPE, &type) != 0 || |
2040 | nvlist_lookup_nvlist_array(vdev_root, ZPOOL_CONFIG_CHILDREN, | |
b128c09f | 2041 | &child, &count) != 0) |
9babb374 | 2042 | return (EZFS_INVALCONFIG); |
b128c09f | 2043 | |
9babb374 BB |
2044 | /* |
2045 | * root pool can not have EFI labeled disks and can only have | |
2046 | * a single top-level vdev. | |
2047 | */ | |
2048 | if (strcmp(type, VDEV_TYPE_ROOT) != 0 || count != 1 || | |
2049 | pool_uses_efi(vdev_root)) | |
2050 | return (EZFS_POOL_INVALARG); | |
b128c09f | 2051 | |
9babb374 BB |
2052 | (void) vdev_get_physpaths(child[0], physpath, phypath_size, &rsz, |
2053 | B_FALSE); | |
2054 | ||
2055 | /* No online devices */ | |
2056 | if (rsz == 0) | |
2057 | return (EZFS_NODEVICE); | |
b128c09f BB |
2058 | |
2059 | return (0); | |
34dc7c2f BB |
2060 | } |
2061 | ||
9babb374 BB |
2062 | /* |
2063 | * Get phys_path for a root pool | |
2064 | * Return 0 on success; non-zero on failure. | |
2065 | */ | |
2066 | int | |
2067 | zpool_get_physpath(zpool_handle_t *zhp, char *physpath, size_t phypath_size) | |
2068 | { | |
2069 | return (zpool_get_config_physpath(zhp->zpool_config, physpath, | |
2070 | phypath_size)); | |
2071 | } | |
2072 | ||
9babb374 BB |
2073 | /* |
2074 | * If the device has being dynamically expanded then we need to relabel | |
2075 | * the disk to use the new unallocated space. | |
2076 | */ | |
2077 | static int | |
d603ed6c | 2078 | zpool_relabel_disk(libzfs_handle_t *hdl, const char *path) |
9babb374 | 2079 | { |
9babb374 BB |
2080 | char errbuf[1024]; |
2081 | int fd, error; | |
9babb374 | 2082 | |
d603ed6c | 2083 | if ((fd = open(path, O_RDWR|O_DIRECT)) < 0) { |
9babb374 | 2084 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot " |
d603ed6c | 2085 | "relabel '%s': unable to open device"), path); |
9babb374 BB |
2086 | return (zfs_error(hdl, EZFS_OPENFAILED, errbuf)); |
2087 | } | |
2088 | ||
2089 | /* | |
2090 | * It's possible that we might encounter an error if the device | |
2091 | * does not have any unallocated space left. If so, we simply | |
2092 | * ignore that error and continue on. | |
2093 | */ | |
d603ed6c | 2094 | error = efi_use_whole_disk(fd); |
9babb374 BB |
2095 | (void) close(fd); |
2096 | if (error && error != VT_ENOSPC) { | |
2097 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "cannot " | |
d603ed6c | 2098 | "relabel '%s': unable to read disk capacity"), path); |
9babb374 BB |
2099 | return (zfs_error(hdl, EZFS_NOCAP, errbuf)); |
2100 | } | |
2101 | return (0); | |
2102 | } | |
2103 | ||
34dc7c2f BB |
2104 | /* |
2105 | * Bring the specified vdev online. The 'flags' parameter is a set of the | |
2106 | * ZFS_ONLINE_* flags. | |
2107 | */ | |
2108 | int | |
2109 | zpool_vdev_online(zpool_handle_t *zhp, const char *path, int flags, | |
2110 | vdev_state_t *newstate) | |
2111 | { | |
2598c001 | 2112 | zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 }; |
34dc7c2f BB |
2113 | char msg[1024]; |
2114 | nvlist_t *tgt; | |
9babb374 | 2115 | boolean_t avail_spare, l2cache, islog; |
34dc7c2f BB |
2116 | libzfs_handle_t *hdl = zhp->zpool_hdl; |
2117 | ||
9babb374 BB |
2118 | if (flags & ZFS_ONLINE_EXPAND) { |
2119 | (void) snprintf(msg, sizeof (msg), | |
2120 | dgettext(TEXT_DOMAIN, "cannot expand %s"), path); | |
2121 | } else { | |
2122 | (void) snprintf(msg, sizeof (msg), | |
2123 | dgettext(TEXT_DOMAIN, "cannot online %s"), path); | |
2124 | } | |
34dc7c2f BB |
2125 | |
2126 | (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); | |
b128c09f | 2127 | if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, |
9babb374 | 2128 | &islog)) == NULL) |
34dc7c2f BB |
2129 | return (zfs_error(hdl, EZFS_NODEVICE, msg)); |
2130 | ||
2131 | verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); | |
2132 | ||
428870ff | 2133 | if (avail_spare) |
34dc7c2f BB |
2134 | return (zfs_error(hdl, EZFS_ISSPARE, msg)); |
2135 | ||
9babb374 BB |
2136 | if (flags & ZFS_ONLINE_EXPAND || |
2137 | zpool_get_prop_int(zhp, ZPOOL_PROP_AUTOEXPAND, NULL)) { | |
2138 | char *pathname = NULL; | |
2139 | uint64_t wholedisk = 0; | |
2140 | ||
2141 | (void) nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_WHOLE_DISK, | |
2142 | &wholedisk); | |
2143 | verify(nvlist_lookup_string(tgt, ZPOOL_CONFIG_PATH, | |
2144 | &pathname) == 0); | |
2145 | ||
2146 | /* | |
2147 | * XXX - L2ARC 1.0 devices can't support expansion. | |
2148 | */ | |
2149 | if (l2cache) { | |
2150 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
2151 | "cannot expand cache devices")); | |
2152 | return (zfs_error(hdl, EZFS_VDEVNOTSUP, msg)); | |
2153 | } | |
2154 | ||
2155 | if (wholedisk) { | |
2156 | pathname += strlen(DISK_ROOT) + 1; | |
572e2857 | 2157 | (void) zpool_relabel_disk(hdl, pathname); |
9babb374 BB |
2158 | } |
2159 | } | |
2160 | ||
34dc7c2f BB |
2161 | zc.zc_cookie = VDEV_STATE_ONLINE; |
2162 | zc.zc_obj = flags; | |
2163 | ||
572e2857 | 2164 | if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) != 0) { |
428870ff BB |
2165 | if (errno == EINVAL) { |
2166 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "was split " | |
2167 | "from this pool into a new one. Use '%s' " | |
2168 | "instead"), "zpool detach"); | |
2169 | return (zfs_error(hdl, EZFS_POSTSPLIT_ONLINE, msg)); | |
2170 | } | |
34dc7c2f | 2171 | return (zpool_standard_error(hdl, errno, msg)); |
428870ff | 2172 | } |
34dc7c2f BB |
2173 | |
2174 | *newstate = zc.zc_cookie; | |
2175 | return (0); | |
2176 | } | |
2177 | ||
2178 | /* | |
2179 | * Take the specified vdev offline | |
2180 | */ | |
2181 | int | |
2182 | zpool_vdev_offline(zpool_handle_t *zhp, const char *path, boolean_t istmp) | |
2183 | { | |
2598c001 | 2184 | zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 }; |
34dc7c2f BB |
2185 | char msg[1024]; |
2186 | nvlist_t *tgt; | |
2187 | boolean_t avail_spare, l2cache; | |
2188 | libzfs_handle_t *hdl = zhp->zpool_hdl; | |
2189 | ||
2190 | (void) snprintf(msg, sizeof (msg), | |
2191 | dgettext(TEXT_DOMAIN, "cannot offline %s"), path); | |
2192 | ||
2193 | (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); | |
b128c09f BB |
2194 | if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, |
2195 | NULL)) == NULL) | |
34dc7c2f BB |
2196 | return (zfs_error(hdl, EZFS_NODEVICE, msg)); |
2197 | ||
2198 | verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); | |
2199 | ||
428870ff | 2200 | if (avail_spare) |
34dc7c2f BB |
2201 | return (zfs_error(hdl, EZFS_ISSPARE, msg)); |
2202 | ||
34dc7c2f BB |
2203 | zc.zc_cookie = VDEV_STATE_OFFLINE; |
2204 | zc.zc_obj = istmp ? ZFS_OFFLINE_TEMPORARY : 0; | |
2205 | ||
572e2857 | 2206 | if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) |
34dc7c2f BB |
2207 | return (0); |
2208 | ||
2209 | switch (errno) { | |
2210 | case EBUSY: | |
2211 | ||
2212 | /* | |
2213 | * There are no other replicas of this device. | |
2214 | */ | |
2215 | return (zfs_error(hdl, EZFS_NOREPLICAS, msg)); | |
2216 | ||
9babb374 BB |
2217 | case EEXIST: |
2218 | /* | |
2219 | * The log device has unplayed logs | |
2220 | */ | |
2221 | return (zfs_error(hdl, EZFS_UNPLAYED_LOGS, msg)); | |
2222 | ||
34dc7c2f BB |
2223 | default: |
2224 | return (zpool_standard_error(hdl, errno, msg)); | |
2225 | } | |
2226 | } | |
2227 | ||
2228 | /* | |
2229 | * Mark the given vdev faulted. | |
2230 | */ | |
2231 | int | |
428870ff | 2232 | zpool_vdev_fault(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux) |
34dc7c2f | 2233 | { |
2598c001 | 2234 | zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 }; |
34dc7c2f BB |
2235 | char msg[1024]; |
2236 | libzfs_handle_t *hdl = zhp->zpool_hdl; | |
2237 | ||
2238 | (void) snprintf(msg, sizeof (msg), | |
b8864a23 | 2239 | dgettext(TEXT_DOMAIN, "cannot fault %llu"), (u_longlong_t)guid); |
34dc7c2f BB |
2240 | |
2241 | (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); | |
2242 | zc.zc_guid = guid; | |
2243 | zc.zc_cookie = VDEV_STATE_FAULTED; | |
428870ff | 2244 | zc.zc_obj = aux; |
34dc7c2f | 2245 | |
572e2857 | 2246 | if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) |
34dc7c2f BB |
2247 | return (0); |
2248 | ||
2249 | switch (errno) { | |
2250 | case EBUSY: | |
2251 | ||
2252 | /* | |
2253 | * There are no other replicas of this device. | |
2254 | */ | |
2255 | return (zfs_error(hdl, EZFS_NOREPLICAS, msg)); | |
2256 | ||
2257 | default: | |
2258 | return (zpool_standard_error(hdl, errno, msg)); | |
2259 | } | |
2260 | ||
2261 | } | |
2262 | ||
2263 | /* | |
2264 | * Mark the given vdev degraded. | |
2265 | */ | |
2266 | int | |
428870ff | 2267 | zpool_vdev_degrade(zpool_handle_t *zhp, uint64_t guid, vdev_aux_t aux) |
34dc7c2f | 2268 | { |
2598c001 | 2269 | zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 }; |
34dc7c2f BB |
2270 | char msg[1024]; |
2271 | libzfs_handle_t *hdl = zhp->zpool_hdl; | |
2272 | ||
2273 | (void) snprintf(msg, sizeof (msg), | |
b8864a23 | 2274 | dgettext(TEXT_DOMAIN, "cannot degrade %llu"), (u_longlong_t)guid); |
34dc7c2f BB |
2275 | |
2276 | (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); | |
2277 | zc.zc_guid = guid; | |
2278 | zc.zc_cookie = VDEV_STATE_DEGRADED; | |
428870ff | 2279 | zc.zc_obj = aux; |
34dc7c2f | 2280 | |
572e2857 | 2281 | if (ioctl(hdl->libzfs_fd, ZFS_IOC_VDEV_SET_STATE, &zc) == 0) |
34dc7c2f BB |
2282 | return (0); |
2283 | ||
2284 | return (zpool_standard_error(hdl, errno, msg)); | |
2285 | } | |
2286 | ||
2287 | /* | |
2288 | * Returns TRUE if the given nvlist is a vdev that was originally swapped in as | |
2289 | * a hot spare. | |
2290 | */ | |
2291 | static boolean_t | |
2292 | is_replacing_spare(nvlist_t *search, nvlist_t *tgt, int which) | |
2293 | { | |
2294 | nvlist_t **child; | |
2295 | uint_t c, children; | |
2296 | char *type; | |
2297 | ||
2298 | if (nvlist_lookup_nvlist_array(search, ZPOOL_CONFIG_CHILDREN, &child, | |
2299 | &children) == 0) { | |
2300 | verify(nvlist_lookup_string(search, ZPOOL_CONFIG_TYPE, | |
2301 | &type) == 0); | |
2302 | ||
2303 | if (strcmp(type, VDEV_TYPE_SPARE) == 0 && | |
2304 | children == 2 && child[which] == tgt) | |
2305 | return (B_TRUE); | |
2306 | ||
2307 | for (c = 0; c < children; c++) | |
2308 | if (is_replacing_spare(child[c], tgt, which)) | |
2309 | return (B_TRUE); | |
2310 | } | |
2311 | ||
2312 | return (B_FALSE); | |
2313 | } | |
2314 | ||
2315 | /* | |
2316 | * Attach new_disk (fully described by nvroot) to old_disk. | |
2317 | * If 'replacing' is specified, the new disk will replace the old one. | |
2318 | */ | |
2319 | int | |
2320 | zpool_vdev_attach(zpool_handle_t *zhp, | |
2321 | const char *old_disk, const char *new_disk, nvlist_t *nvroot, int replacing) | |
2322 | { | |
2598c001 | 2323 | zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 }; |
34dc7c2f BB |
2324 | char msg[1024]; |
2325 | int ret; | |
2326 | nvlist_t *tgt; | |
b128c09f BB |
2327 | boolean_t avail_spare, l2cache, islog; |
2328 | uint64_t val; | |
572e2857 | 2329 | char *newname; |
34dc7c2f BB |
2330 | nvlist_t **child; |
2331 | uint_t children; | |
2332 | nvlist_t *config_root; | |
2333 | libzfs_handle_t *hdl = zhp->zpool_hdl; | |
b128c09f | 2334 | boolean_t rootpool = pool_is_bootable(zhp); |
34dc7c2f BB |
2335 | |
2336 | if (replacing) | |
2337 | (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, | |
2338 | "cannot replace %s with %s"), old_disk, new_disk); | |
2339 | else | |
2340 | (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, | |
2341 | "cannot attach %s to %s"), new_disk, old_disk); | |
2342 | ||
b128c09f BB |
2343 | /* |
2344 | * If this is a root pool, make sure that we're not attaching an | |
2345 | * EFI labeled device. | |
2346 | */ | |
2347 | if (rootpool && pool_uses_efi(nvroot)) { | |
2348 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
2349 | "EFI labeled devices are not supported on root pools.")); | |
2350 | return (zfs_error(hdl, EZFS_POOL_NOTSUP, msg)); | |
2351 | } | |
2352 | ||
34dc7c2f | 2353 | (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); |
b128c09f BB |
2354 | if ((tgt = zpool_find_vdev(zhp, old_disk, &avail_spare, &l2cache, |
2355 | &islog)) == 0) | |
34dc7c2f BB |
2356 | return (zfs_error(hdl, EZFS_NODEVICE, msg)); |
2357 | ||
2358 | if (avail_spare) | |
2359 | return (zfs_error(hdl, EZFS_ISSPARE, msg)); | |
2360 | ||
2361 | if (l2cache) | |
2362 | return (zfs_error(hdl, EZFS_ISL2CACHE, msg)); | |
2363 | ||
2364 | verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); | |
2365 | zc.zc_cookie = replacing; | |
2366 | ||
2367 | if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, | |
2368 | &child, &children) != 0 || children != 1) { | |
2369 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
2370 | "new device must be a single disk")); | |
2371 | return (zfs_error(hdl, EZFS_INVALCONFIG, msg)); | |
2372 | } | |
2373 | ||
2374 | verify(nvlist_lookup_nvlist(zpool_get_config(zhp, NULL), | |
2375 | ZPOOL_CONFIG_VDEV_TREE, &config_root) == 0); | |
2376 | ||
428870ff | 2377 | if ((newname = zpool_vdev_name(NULL, NULL, child[0], B_FALSE)) == NULL) |
b128c09f BB |
2378 | return (-1); |
2379 | ||
34dc7c2f BB |
2380 | /* |
2381 | * If the target is a hot spare that has been swapped in, we can only | |
2382 | * replace it with another hot spare. | |
2383 | */ | |
2384 | if (replacing && | |
2385 | nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 && | |
b128c09f BB |
2386 | (zpool_find_vdev(zhp, newname, &avail_spare, &l2cache, |
2387 | NULL) == NULL || !avail_spare) && | |
2388 | is_replacing_spare(config_root, tgt, 1)) { | |
34dc7c2f BB |
2389 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, |
2390 | "can only be replaced by another hot spare")); | |
b128c09f | 2391 | free(newname); |
34dc7c2f BB |
2392 | return (zfs_error(hdl, EZFS_BADTARGET, msg)); |
2393 | } | |
2394 | ||
b128c09f BB |
2395 | free(newname); |
2396 | ||
34dc7c2f BB |
2397 | if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0) |
2398 | return (-1); | |
2399 | ||
572e2857 | 2400 | ret = zfs_ioctl(hdl, ZFS_IOC_VDEV_ATTACH, &zc); |
34dc7c2f BB |
2401 | |
2402 | zcmd_free_nvlists(&zc); | |
2403 | ||
b128c09f BB |
2404 | if (ret == 0) { |
2405 | if (rootpool) { | |
9babb374 BB |
2406 | /* |
2407 | * XXX need a better way to prevent user from | |
2408 | * booting up a half-baked vdev. | |
2409 | */ | |
2410 | (void) fprintf(stderr, dgettext(TEXT_DOMAIN, "Make " | |
2411 | "sure to wait until resilver is done " | |
2412 | "before rebooting.\n")); | |
b128c09f | 2413 | } |
34dc7c2f | 2414 | return (0); |
b128c09f | 2415 | } |
34dc7c2f BB |
2416 | |
2417 | switch (errno) { | |
2418 | case ENOTSUP: | |
2419 | /* | |
2420 | * Can't attach to or replace this type of vdev. | |
2421 | */ | |
2422 | if (replacing) { | |
572e2857 BB |
2423 | uint64_t version = zpool_get_prop_int(zhp, |
2424 | ZPOOL_PROP_VERSION, NULL); | |
2425 | ||
b128c09f | 2426 | if (islog) |
34dc7c2f BB |
2427 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, |
2428 | "cannot replace a log with a spare")); | |
572e2857 BB |
2429 | else if (version >= SPA_VERSION_MULTI_REPLACE) |
2430 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
2431 | "already in replacing/spare config; wait " | |
2432 | "for completion or use 'zpool detach'")); | |
34dc7c2f BB |
2433 | else |
2434 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
2435 | "cannot replace a replacing device")); | |
2436 | } else { | |
2437 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
2438 | "can only attach to mirrors and top-level " | |
2439 | "disks")); | |
2440 | } | |
2441 | (void) zfs_error(hdl, EZFS_BADTARGET, msg); | |
2442 | break; | |
2443 | ||
2444 | case EINVAL: | |
2445 | /* | |
2446 | * The new device must be a single disk. | |
2447 | */ | |
2448 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
2449 | "new device must be a single disk")); | |
2450 | (void) zfs_error(hdl, EZFS_INVALCONFIG, msg); | |
2451 | break; | |
2452 | ||
2453 | case EBUSY: | |
2454 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "%s is busy"), | |
2455 | new_disk); | |
2456 | (void) zfs_error(hdl, EZFS_BADDEV, msg); | |
2457 | break; | |
2458 | ||
2459 | case EOVERFLOW: | |
2460 | /* | |
2461 | * The new device is too small. | |
2462 | */ | |
2463 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
2464 | "device is too small")); | |
2465 | (void) zfs_error(hdl, EZFS_BADDEV, msg); | |
2466 | break; | |
2467 | ||
2468 | case EDOM: | |
2469 | /* | |
2470 | * The new device has a different alignment requirement. | |
2471 | */ | |
2472 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
2473 | "devices have different sector alignment")); | |
2474 | (void) zfs_error(hdl, EZFS_BADDEV, msg); | |
2475 | break; | |
2476 | ||
2477 | case ENAMETOOLONG: | |
2478 | /* | |
2479 | * The resulting top-level vdev spec won't fit in the label. | |
2480 | */ | |
2481 | (void) zfs_error(hdl, EZFS_DEVOVERFLOW, msg); | |
2482 | break; | |
2483 | ||
2484 | default: | |
2485 | (void) zpool_standard_error(hdl, errno, msg); | |
2486 | } | |
2487 | ||
2488 | return (-1); | |
2489 | } | |
2490 | ||
2491 | /* | |
2492 | * Detach the specified device. | |
2493 | */ | |
2494 | int | |
2495 | zpool_vdev_detach(zpool_handle_t *zhp, const char *path) | |
2496 | { | |
2598c001 | 2497 | zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 }; |
34dc7c2f BB |
2498 | char msg[1024]; |
2499 | nvlist_t *tgt; | |
2500 | boolean_t avail_spare, l2cache; | |
2501 | libzfs_handle_t *hdl = zhp->zpool_hdl; | |
2502 | ||
2503 | (void) snprintf(msg, sizeof (msg), | |
2504 | dgettext(TEXT_DOMAIN, "cannot detach %s"), path); | |
2505 | ||
2506 | (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); | |
b128c09f BB |
2507 | if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, |
2508 | NULL)) == 0) | |
34dc7c2f BB |
2509 | return (zfs_error(hdl, EZFS_NODEVICE, msg)); |
2510 | ||
2511 | if (avail_spare) | |
2512 | return (zfs_error(hdl, EZFS_ISSPARE, msg)); | |
2513 | ||
2514 | if (l2cache) | |
2515 | return (zfs_error(hdl, EZFS_ISL2CACHE, msg)); | |
2516 | ||
2517 | verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); | |
2518 | ||
2519 | if (zfs_ioctl(hdl, ZFS_IOC_VDEV_DETACH, &zc) == 0) | |
2520 | return (0); | |
2521 | ||
2522 | switch (errno) { | |
2523 | ||
2524 | case ENOTSUP: | |
2525 | /* | |
2526 | * Can't detach from this type of vdev. | |
2527 | */ | |
2528 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "only " | |
2529 | "applicable to mirror and replacing vdevs")); | |
572e2857 | 2530 | (void) zfs_error(hdl, EZFS_BADTARGET, msg); |
34dc7c2f BB |
2531 | break; |
2532 | ||
2533 | case EBUSY: | |
2534 | /* | |
2535 | * There are no other replicas of this device. | |
2536 | */ | |
2537 | (void) zfs_error(hdl, EZFS_NOREPLICAS, msg); | |
2538 | break; | |
2539 | ||
2540 | default: | |
2541 | (void) zpool_standard_error(hdl, errno, msg); | |
2542 | } | |
2543 | ||
2544 | return (-1); | |
2545 | } | |
2546 | ||
428870ff BB |
2547 | /* |
2548 | * Find a mirror vdev in the source nvlist. | |
2549 | * | |
2550 | * The mchild array contains a list of disks in one of the top-level mirrors | |
2551 | * of the source pool. The schild array contains a list of disks that the | |
2552 | * user specified on the command line. We loop over the mchild array to | |
2553 | * see if any entry in the schild array matches. | |
2554 | * | |
2555 | * If a disk in the mchild array is found in the schild array, we return | |
2556 | * the index of that entry. Otherwise we return -1. | |
2557 | */ | |
2558 | static int | |
2559 | find_vdev_entry(zpool_handle_t *zhp, nvlist_t **mchild, uint_t mchildren, | |
2560 | nvlist_t **schild, uint_t schildren) | |
2561 | { | |
2562 | uint_t mc; | |
2563 | ||
2564 | for (mc = 0; mc < mchildren; mc++) { | |
2565 | uint_t sc; | |
2566 | char *mpath = zpool_vdev_name(zhp->zpool_hdl, zhp, | |
2567 | mchild[mc], B_FALSE); | |
2568 | ||
2569 | for (sc = 0; sc < schildren; sc++) { | |
2570 | char *spath = zpool_vdev_name(zhp->zpool_hdl, zhp, | |
2571 | schild[sc], B_FALSE); | |
2572 | boolean_t result = (strcmp(mpath, spath) == 0); | |
2573 | ||
2574 | free(spath); | |
2575 | if (result) { | |
2576 | free(mpath); | |
2577 | return (mc); | |
2578 | } | |
2579 | } | |
2580 | ||
2581 | free(mpath); | |
2582 | } | |
2583 | ||
2584 | return (-1); | |
2585 | } | |
2586 | ||
2587 | /* | |
2588 | * Split a mirror pool. If newroot points to null, then a new nvlist | |
2589 | * is generated and it is the responsibility of the caller to free it. | |
2590 | */ | |
2591 | int | |
2592 | zpool_vdev_split(zpool_handle_t *zhp, char *newname, nvlist_t **newroot, | |
2593 | nvlist_t *props, splitflags_t flags) | |
2594 | { | |
2598c001 | 2595 | zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 }; |
428870ff BB |
2596 | char msg[1024]; |
2597 | nvlist_t *tree, *config, **child, **newchild, *newconfig = NULL; | |
2598 | nvlist_t **varray = NULL, *zc_props = NULL; | |
2599 | uint_t c, children, newchildren, lastlog = 0, vcount, found = 0; | |
2600 | libzfs_handle_t *hdl = zhp->zpool_hdl; | |
2601 | uint64_t vers; | |
2602 | boolean_t freelist = B_FALSE, memory_err = B_TRUE; | |
2603 | int retval = 0; | |
2604 | ||
2605 | (void) snprintf(msg, sizeof (msg), | |
2606 | dgettext(TEXT_DOMAIN, "Unable to split %s"), zhp->zpool_name); | |
2607 | ||
2608 | if (!zpool_name_valid(hdl, B_FALSE, newname)) | |
2609 | return (zfs_error(hdl, EZFS_INVALIDNAME, msg)); | |
2610 | ||
2611 | if ((config = zpool_get_config(zhp, NULL)) == NULL) { | |
2612 | (void) fprintf(stderr, gettext("Internal error: unable to " | |
2613 | "retrieve pool configuration\n")); | |
2614 | return (-1); | |
2615 | } | |
2616 | ||
2617 | verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree) | |
2618 | == 0); | |
2619 | verify(nvlist_lookup_uint64(config, ZPOOL_CONFIG_VERSION, &vers) == 0); | |
2620 | ||
2621 | if (props) { | |
572e2857 | 2622 | prop_flags_t flags = { .create = B_FALSE, .import = B_TRUE }; |
428870ff | 2623 | if ((zc_props = zpool_valid_proplist(hdl, zhp->zpool_name, |
572e2857 | 2624 | props, vers, flags, msg)) == NULL) |
428870ff BB |
2625 | return (-1); |
2626 | } | |
2627 | ||
2628 | if (nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &child, | |
2629 | &children) != 0) { | |
2630 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
2631 | "Source pool is missing vdev tree")); | |
2632 | if (zc_props) | |
2633 | nvlist_free(zc_props); | |
2634 | return (-1); | |
2635 | } | |
2636 | ||
2637 | varray = zfs_alloc(hdl, children * sizeof (nvlist_t *)); | |
2638 | vcount = 0; | |
2639 | ||
2640 | if (*newroot == NULL || | |
2641 | nvlist_lookup_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, | |
2642 | &newchild, &newchildren) != 0) | |
2643 | newchildren = 0; | |
2644 | ||
2645 | for (c = 0; c < children; c++) { | |
2646 | uint64_t is_log = B_FALSE, is_hole = B_FALSE; | |
2647 | char *type; | |
2648 | nvlist_t **mchild, *vdev; | |
2649 | uint_t mchildren; | |
2650 | int entry; | |
2651 | ||
2652 | /* | |
2653 | * Unlike cache & spares, slogs are stored in the | |
2654 | * ZPOOL_CONFIG_CHILDREN array. We filter them out here. | |
2655 | */ | |
2656 | (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_LOG, | |
2657 | &is_log); | |
2658 | (void) nvlist_lookup_uint64(child[c], ZPOOL_CONFIG_IS_HOLE, | |
2659 | &is_hole); | |
2660 | if (is_log || is_hole) { | |
2661 | /* | |
2662 | * Create a hole vdev and put it in the config. | |
2663 | */ | |
2664 | if (nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) != 0) | |
2665 | goto out; | |
2666 | if (nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE, | |
2667 | VDEV_TYPE_HOLE) != 0) | |
2668 | goto out; | |
2669 | if (nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_HOLE, | |
2670 | 1) != 0) | |
2671 | goto out; | |
2672 | if (lastlog == 0) | |
2673 | lastlog = vcount; | |
2674 | varray[vcount++] = vdev; | |
2675 | continue; | |
2676 | } | |
2677 | lastlog = 0; | |
2678 | verify(nvlist_lookup_string(child[c], ZPOOL_CONFIG_TYPE, &type) | |
2679 | == 0); | |
2680 | if (strcmp(type, VDEV_TYPE_MIRROR) != 0) { | |
2681 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
2682 | "Source pool must be composed only of mirrors\n")); | |
2683 | retval = zfs_error(hdl, EZFS_INVALCONFIG, msg); | |
2684 | goto out; | |
2685 | } | |
2686 | ||
2687 | verify(nvlist_lookup_nvlist_array(child[c], | |
2688 | ZPOOL_CONFIG_CHILDREN, &mchild, &mchildren) == 0); | |
2689 | ||
2690 | /* find or add an entry for this top-level vdev */ | |
2691 | if (newchildren > 0 && | |
2692 | (entry = find_vdev_entry(zhp, mchild, mchildren, | |
2693 | newchild, newchildren)) >= 0) { | |
2694 | /* We found a disk that the user specified. */ | |
2695 | vdev = mchild[entry]; | |
2696 | ++found; | |
2697 | } else { | |
2698 | /* User didn't specify a disk for this vdev. */ | |
2699 | vdev = mchild[mchildren - 1]; | |
2700 | } | |
2701 | ||
2702 | if (nvlist_dup(vdev, &varray[vcount++], 0) != 0) | |
2703 | goto out; | |
2704 | } | |
2705 | ||
2706 | /* did we find every disk the user specified? */ | |
2707 | if (found != newchildren) { | |
2708 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "Device list must " | |
2709 | "include at most one disk from each mirror")); | |
2710 | retval = zfs_error(hdl, EZFS_INVALCONFIG, msg); | |
2711 | goto out; | |
2712 | } | |
2713 | ||
2714 | /* Prepare the nvlist for populating. */ | |
2715 | if (*newroot == NULL) { | |
2716 | if (nvlist_alloc(newroot, NV_UNIQUE_NAME, 0) != 0) | |
2717 | goto out; | |
2718 | freelist = B_TRUE; | |
2719 | if (nvlist_add_string(*newroot, ZPOOL_CONFIG_TYPE, | |
2720 | VDEV_TYPE_ROOT) != 0) | |
2721 | goto out; | |
2722 | } else { | |
2723 | verify(nvlist_remove_all(*newroot, ZPOOL_CONFIG_CHILDREN) == 0); | |
2724 | } | |
2725 | ||
2726 | /* Add all the children we found */ | |
2727 | if (nvlist_add_nvlist_array(*newroot, ZPOOL_CONFIG_CHILDREN, varray, | |
2728 | lastlog == 0 ? vcount : lastlog) != 0) | |
2729 | goto out; | |
2730 | ||
2731 | /* | |
2732 | * If we're just doing a dry run, exit now with success. | |
2733 | */ | |
2734 | if (flags.dryrun) { | |
2735 | memory_err = B_FALSE; | |
2736 | freelist = B_FALSE; | |
2737 | goto out; | |
2738 | } | |
2739 | ||
2740 | /* now build up the config list & call the ioctl */ | |
2741 | if (nvlist_alloc(&newconfig, NV_UNIQUE_NAME, 0) != 0) | |
2742 | goto out; | |
2743 | ||
2744 | if (nvlist_add_nvlist(newconfig, | |
2745 | ZPOOL_CONFIG_VDEV_TREE, *newroot) != 0 || | |
2746 | nvlist_add_string(newconfig, | |
2747 | ZPOOL_CONFIG_POOL_NAME, newname) != 0 || | |
2748 | nvlist_add_uint64(newconfig, ZPOOL_CONFIG_VERSION, vers) != 0) | |
2749 | goto out; | |
2750 | ||
2751 | /* | |
2752 | * The new pool is automatically part of the namespace unless we | |
2753 | * explicitly export it. | |
2754 | */ | |
2755 | if (!flags.import) | |
2756 | zc.zc_cookie = ZPOOL_EXPORT_AFTER_SPLIT; | |
2757 | (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); | |
2758 | (void) strlcpy(zc.zc_string, newname, sizeof (zc.zc_string)); | |
2759 | if (zcmd_write_conf_nvlist(hdl, &zc, newconfig) != 0) | |
2760 | goto out; | |
2761 | if (zc_props != NULL && zcmd_write_src_nvlist(hdl, &zc, zc_props) != 0) | |
2762 | goto out; | |
2763 | ||
2764 | if (zfs_ioctl(hdl, ZFS_IOC_VDEV_SPLIT, &zc) != 0) { | |
2765 | retval = zpool_standard_error(hdl, errno, msg); | |
2766 | goto out; | |
2767 | } | |
2768 | ||
2769 | freelist = B_FALSE; | |
2770 | memory_err = B_FALSE; | |
2771 | ||
2772 | out: | |
2773 | if (varray != NULL) { | |
2774 | int v; | |
2775 | ||
2776 | for (v = 0; v < vcount; v++) | |
2777 | nvlist_free(varray[v]); | |
2778 | free(varray); | |
2779 | } | |
2780 | zcmd_free_nvlists(&zc); | |
2781 | if (zc_props) | |
2782 | nvlist_free(zc_props); | |
2783 | if (newconfig) | |
2784 | nvlist_free(newconfig); | |
2785 | if (freelist) { | |
2786 | nvlist_free(*newroot); | |
2787 | *newroot = NULL; | |
2788 | } | |
2789 | ||
2790 | if (retval != 0) | |
2791 | return (retval); | |
2792 | ||
2793 | if (memory_err) | |
2794 | return (no_memory(hdl)); | |
2795 | ||
2796 | return (0); | |
2797 | } | |
2798 | ||
34dc7c2f BB |
2799 | /* |
2800 | * Remove the given device. Currently, this is supported only for hot spares | |
2801 | * and level 2 cache devices. | |
2802 | */ | |
2803 | int | |
2804 | zpool_vdev_remove(zpool_handle_t *zhp, const char *path) | |
2805 | { | |
2598c001 | 2806 | zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 }; |
34dc7c2f BB |
2807 | char msg[1024]; |
2808 | nvlist_t *tgt; | |
428870ff | 2809 | boolean_t avail_spare, l2cache, islog; |
34dc7c2f | 2810 | libzfs_handle_t *hdl = zhp->zpool_hdl; |
428870ff | 2811 | uint64_t version; |
34dc7c2f BB |
2812 | |
2813 | (void) snprintf(msg, sizeof (msg), | |
2814 | dgettext(TEXT_DOMAIN, "cannot remove %s"), path); | |
2815 | ||
2816 | (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); | |
b128c09f | 2817 | if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, &l2cache, |
428870ff | 2818 | &islog)) == 0) |
34dc7c2f | 2819 | return (zfs_error(hdl, EZFS_NODEVICE, msg)); |
428870ff BB |
2820 | /* |
2821 | * XXX - this should just go away. | |
2822 | */ | |
2823 | if (!avail_spare && !l2cache && !islog) { | |
34dc7c2f | 2824 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, |
428870ff BB |
2825 | "only inactive hot spares, cache, top-level, " |
2826 | "or log devices can be removed")); | |
34dc7c2f BB |
2827 | return (zfs_error(hdl, EZFS_NODEVICE, msg)); |
2828 | } | |
2829 | ||
428870ff BB |
2830 | version = zpool_get_prop_int(zhp, ZPOOL_PROP_VERSION, NULL); |
2831 | if (islog && version < SPA_VERSION_HOLES) { | |
2832 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
2833 | "pool must be upgrade to support log removal")); | |
2834 | return (zfs_error(hdl, EZFS_BADVERSION, msg)); | |
2835 | } | |
2836 | ||
34dc7c2f BB |
2837 | verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, &zc.zc_guid) == 0); |
2838 | ||
2839 | if (zfs_ioctl(hdl, ZFS_IOC_VDEV_REMOVE, &zc) == 0) | |
2840 | return (0); | |
2841 | ||
2842 | return (zpool_standard_error(hdl, errno, msg)); | |
2843 | } | |
2844 | ||
2845 | /* | |
2846 | * Clear the errors for the pool, or the particular device if specified. | |
2847 | */ | |
2848 | int | |
428870ff | 2849 | zpool_clear(zpool_handle_t *zhp, const char *path, nvlist_t *rewindnvl) |
34dc7c2f | 2850 | { |
2598c001 | 2851 | zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 }; |
34dc7c2f BB |
2852 | char msg[1024]; |
2853 | nvlist_t *tgt; | |
428870ff | 2854 | zpool_rewind_policy_t policy; |
34dc7c2f BB |
2855 | boolean_t avail_spare, l2cache; |
2856 | libzfs_handle_t *hdl = zhp->zpool_hdl; | |
428870ff | 2857 | nvlist_t *nvi = NULL; |
572e2857 | 2858 | int error; |
34dc7c2f BB |
2859 | |
2860 | if (path) | |
2861 | (void) snprintf(msg, sizeof (msg), | |
2862 | dgettext(TEXT_DOMAIN, "cannot clear errors for %s"), | |
2863 | path); | |
2864 | else | |
2865 | (void) snprintf(msg, sizeof (msg), | |
2866 | dgettext(TEXT_DOMAIN, "cannot clear errors for %s"), | |
2867 | zhp->zpool_name); | |
2868 | ||
2869 | (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); | |
2870 | if (path) { | |
2871 | if ((tgt = zpool_find_vdev(zhp, path, &avail_spare, | |
b128c09f | 2872 | &l2cache, NULL)) == 0) |
34dc7c2f BB |
2873 | return (zfs_error(hdl, EZFS_NODEVICE, msg)); |
2874 | ||
2875 | /* | |
2876 | * Don't allow error clearing for hot spares. Do allow | |
2877 | * error clearing for l2cache devices. | |
2878 | */ | |
2879 | if (avail_spare) | |
2880 | return (zfs_error(hdl, EZFS_ISSPARE, msg)); | |
2881 | ||
2882 | verify(nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_GUID, | |
2883 | &zc.zc_guid) == 0); | |
2884 | } | |
2885 | ||
428870ff BB |
2886 | zpool_get_rewind_policy(rewindnvl, &policy); |
2887 | zc.zc_cookie = policy.zrp_request; | |
2888 | ||
572e2857 | 2889 | if (zcmd_alloc_dst_nvlist(hdl, &zc, zhp->zpool_config_size * 2) != 0) |
428870ff BB |
2890 | return (-1); |
2891 | ||
572e2857 | 2892 | if (zcmd_write_src_nvlist(hdl, &zc, rewindnvl) != 0) |
428870ff BB |
2893 | return (-1); |
2894 | ||
572e2857 BB |
2895 | while ((error = zfs_ioctl(hdl, ZFS_IOC_CLEAR, &zc)) != 0 && |
2896 | errno == ENOMEM) { | |
2897 | if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { | |
2898 | zcmd_free_nvlists(&zc); | |
2899 | return (-1); | |
2900 | } | |
2901 | } | |
2902 | ||
2903 | if (!error || ((policy.zrp_request & ZPOOL_TRY_REWIND) && | |
428870ff BB |
2904 | errno != EPERM && errno != EACCES)) { |
2905 | if (policy.zrp_request & | |
2906 | (ZPOOL_DO_REWIND | ZPOOL_TRY_REWIND)) { | |
2907 | (void) zcmd_read_dst_nvlist(hdl, &zc, &nvi); | |
2908 | zpool_rewind_exclaim(hdl, zc.zc_name, | |
2909 | ((policy.zrp_request & ZPOOL_TRY_REWIND) != 0), | |
2910 | nvi); | |
2911 | nvlist_free(nvi); | |
2912 | } | |
2913 | zcmd_free_nvlists(&zc); | |
34dc7c2f | 2914 | return (0); |
428870ff | 2915 | } |
34dc7c2f | 2916 | |
428870ff | 2917 | zcmd_free_nvlists(&zc); |
34dc7c2f BB |
2918 | return (zpool_standard_error(hdl, errno, msg)); |
2919 | } | |
2920 | ||
2921 | /* | |
2922 | * Similar to zpool_clear(), but takes a GUID (used by fmd). | |
2923 | */ | |
2924 | int | |
2925 | zpool_vdev_clear(zpool_handle_t *zhp, uint64_t guid) | |
2926 | { | |
2598c001 | 2927 | zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 }; |
34dc7c2f BB |
2928 | char msg[1024]; |
2929 | libzfs_handle_t *hdl = zhp->zpool_hdl; | |
2930 | ||
2931 | (void) snprintf(msg, sizeof (msg), | |
2932 | dgettext(TEXT_DOMAIN, "cannot clear errors for %llx"), | |
b8864a23 | 2933 | (u_longlong_t)guid); |
34dc7c2f BB |
2934 | |
2935 | (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); | |
2936 | zc.zc_guid = guid; | |
428870ff | 2937 | zc.zc_cookie = ZPOOL_NO_REWIND; |
34dc7c2f BB |
2938 | |
2939 | if (ioctl(hdl->libzfs_fd, ZFS_IOC_CLEAR, &zc) == 0) | |
2940 | return (0); | |
2941 | ||
2942 | return (zpool_standard_error(hdl, errno, msg)); | |
2943 | } | |
2944 | ||
34dc7c2f BB |
2945 | /* |
2946 | * Convert from a devid string to a path. | |
2947 | */ | |
2948 | static char * | |
2949 | devid_to_path(char *devid_str) | |
2950 | { | |
2951 | ddi_devid_t devid; | |
2952 | char *minor; | |
2953 | char *path; | |
2954 | devid_nmlist_t *list = NULL; | |
2955 | int ret; | |
2956 | ||
2957 | if (devid_str_decode(devid_str, &devid, &minor) != 0) | |
2958 | return (NULL); | |
2959 | ||
2960 | ret = devid_deviceid_to_nmlist("/dev", devid, minor, &list); | |
2961 | ||
2962 | devid_str_free(minor); | |
2963 | devid_free(devid); | |
2964 | ||
2965 | if (ret != 0) | |
2966 | return (NULL); | |
2967 | ||
2968 | if ((path = strdup(list[0].devname)) == NULL) | |
2969 | return (NULL); | |
2970 | ||
2971 | devid_free_nmlist(list); | |
2972 | ||
2973 | return (path); | |
2974 | } | |
2975 | ||
2976 | /* | |
2977 | * Convert from a path to a devid string. | |
2978 | */ | |
2979 | static char * | |
2980 | path_to_devid(const char *path) | |
2981 | { | |
2982 | int fd; | |
2983 | ddi_devid_t devid; | |
2984 | char *minor, *ret; | |
2985 | ||
2986 | if ((fd = open(path, O_RDONLY)) < 0) | |
2987 | return (NULL); | |
2988 | ||
2989 | minor = NULL; | |
2990 | ret = NULL; | |
2991 | if (devid_get(fd, &devid) == 0) { | |
2992 | if (devid_get_minor_name(fd, &minor) == 0) | |
2993 | ret = devid_str_encode(devid, minor); | |
2994 | if (minor != NULL) | |
2995 | devid_str_free(minor); | |
2996 | devid_free(devid); | |
2997 | } | |
2998 | (void) close(fd); | |
2999 | ||
3000 | return (ret); | |
3001 | } | |
3002 | ||
3003 | /* | |
3004 | * Issue the necessary ioctl() to update the stored path value for the vdev. We | |
3005 | * ignore any failure here, since a common case is for an unprivileged user to | |
3006 | * type 'zpool status', and we'll display the correct information anyway. | |
3007 | */ | |
3008 | static void | |
3009 | set_path(zpool_handle_t *zhp, nvlist_t *nv, const char *path) | |
3010 | { | |
2598c001 | 3011 | zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 }; |
34dc7c2f BB |
3012 | |
3013 | (void) strncpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); | |
3014 | (void) strncpy(zc.zc_value, path, sizeof (zc.zc_value)); | |
3015 | verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, | |
3016 | &zc.zc_guid) == 0); | |
3017 | ||
3018 | (void) ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_VDEV_SETPATH, &zc); | |
3019 | } | |
3020 | ||
83c62c93 NB |
3021 | /* |
3022 | * Remove partition suffix from a vdev path. Partition suffixes may take three | |
3023 | * forms: "-partX", "pX", or "X", where X is a string of digits. The second | |
3024 | * case only occurs when the suffix is preceded by a digit, i.e. "md0p0" The | |
3025 | * third case only occurs when preceded by a string matching the regular | |
3026 | * expression "^[hs]d[a-z]+", i.e. a scsi or ide disk. | |
3027 | */ | |
3028 | static char * | |
3029 | strip_partition(libzfs_handle_t *hdl, char *path) | |
3030 | { | |
3031 | char *tmp = zfs_strdup(hdl, path); | |
3032 | char *part = NULL, *d = NULL; | |
3033 | ||
3034 | if ((part = strstr(tmp, "-part")) && part != tmp) { | |
3035 | d = part + 5; | |
3036 | } else if ((part = strrchr(tmp, 'p')) && | |
3037 | part > tmp + 1 && isdigit(*(part-1))) { | |
3038 | d = part + 1; | |
3039 | } else if ((tmp[0] == 'h' || tmp[0] == 's') && tmp[1] == 'd') { | |
3040 | for (d = &tmp[2]; isalpha(*d); part = ++d); | |
3041 | } | |
3042 | if (part && d && *d != '\0') { | |
3043 | for (; isdigit(*d); d++); | |
3044 | if (*d == '\0') | |
3045 | *part = '\0'; | |
3046 | } | |
3047 | return (tmp); | |
3048 | } | |
3049 | ||
858219cc NB |
3050 | #define PATH_BUF_LEN 64 |
3051 | ||
34dc7c2f BB |
3052 | /* |
3053 | * Given a vdev, return the name to display in iostat. If the vdev has a path, | |
3054 | * we use that, stripping off any leading "/dev/dsk/"; if not, we use the type. | |
3055 | * We also check if this is a whole disk, in which case we strip off the | |
3056 | * trailing 's0' slice name. | |
3057 | * | |
3058 | * This routine is also responsible for identifying when disks have been | |
3059 | * reconfigured in a new location. The kernel will have opened the device by | |
3060 | * devid, but the path will still refer to the old location. To catch this, we | |
3061 | * first do a path -> devid translation (which is fast for the common case). If | |
3062 | * the devid matches, we're done. If not, we do a reverse devid -> path | |
3063 | * translation and issue the appropriate ioctl() to update the path of the vdev. | |
3064 | * If 'zhp' is NULL, then this is an exported pool, and we don't need to do any | |
3065 | * of these checks. | |
3066 | */ | |
3067 | char * | |
428870ff BB |
3068 | zpool_vdev_name(libzfs_handle_t *hdl, zpool_handle_t *zhp, nvlist_t *nv, |
3069 | boolean_t verbose) | |
34dc7c2f | 3070 | { |
d603ed6c | 3071 | char *path, *devid, *type; |
34dc7c2f | 3072 | uint64_t value; |
858219cc | 3073 | char buf[PATH_BUF_LEN]; |
34dc7c2f BB |
3074 | vdev_stat_t *vs; |
3075 | uint_t vsc; | |
3076 | ||
3077 | if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NOT_PRESENT, | |
3078 | &value) == 0) { | |
3079 | verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_GUID, | |
3080 | &value) == 0); | |
3081 | (void) snprintf(buf, sizeof (buf), "%llu", | |
3082 | (u_longlong_t)value); | |
3083 | path = buf; | |
3084 | } else if (nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0) { | |
34dc7c2f BB |
3085 | /* |
3086 | * If the device is dead (faulted, offline, etc) then don't | |
3087 | * bother opening it. Otherwise we may be forcing the user to | |
3088 | * open a misbehaving device, which can have undesirable | |
3089 | * effects. | |
3090 | */ | |
428870ff | 3091 | if ((nvlist_lookup_uint64_array(nv, ZPOOL_CONFIG_VDEV_STATS, |
34dc7c2f BB |
3092 | (uint64_t **)&vs, &vsc) != 0 || |
3093 | vs->vs_state >= VDEV_STATE_DEGRADED) && | |
3094 | zhp != NULL && | |
3095 | nvlist_lookup_string(nv, ZPOOL_CONFIG_DEVID, &devid) == 0) { | |
3096 | /* | |
3097 | * Determine if the current path is correct. | |
3098 | */ | |
3099 | char *newdevid = path_to_devid(path); | |
3100 | ||
3101 | if (newdevid == NULL || | |
3102 | strcmp(devid, newdevid) != 0) { | |
3103 | char *newpath; | |
3104 | ||
3105 | if ((newpath = devid_to_path(devid)) != NULL) { | |
3106 | /* | |
3107 | * Update the path appropriately. | |
3108 | */ | |
3109 | set_path(zhp, nv, newpath); | |
3110 | if (nvlist_add_string(nv, | |
3111 | ZPOOL_CONFIG_PATH, newpath) == 0) | |
3112 | verify(nvlist_lookup_string(nv, | |
3113 | ZPOOL_CONFIG_PATH, | |
3114 | &path) == 0); | |
3115 | free(newpath); | |
3116 | } | |
3117 | } | |
3118 | ||
3119 | if (newdevid) | |
3120 | devid_str_free(newdevid); | |
3121 | } | |
3122 | ||
d603ed6c BB |
3123 | /* |
3124 | * For a block device only use the name. | |
3125 | */ | |
3126 | verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0); | |
3127 | if (strcmp(type, VDEV_TYPE_DISK) == 0) { | |
3128 | path = strrchr(path, '/'); | |
3129 | path++; | |
3130 | } | |
34dc7c2f | 3131 | |
d603ed6c | 3132 | /* |
83c62c93 | 3133 | * Remove the partition from the path it this is a whole disk. |
d603ed6c | 3134 | */ |
34dc7c2f BB |
3135 | if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, |
3136 | &value) == 0 && value) { | |
83c62c93 | 3137 | return strip_partition(hdl, path); |
34dc7c2f BB |
3138 | } |
3139 | } else { | |
3140 | verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &path) == 0); | |
3141 | ||
3142 | /* | |
3143 | * If it's a raidz device, we need to stick in the parity level. | |
3144 | */ | |
3145 | if (strcmp(path, VDEV_TYPE_RAIDZ) == 0) { | |
858219cc NB |
3146 | char tmpbuf[PATH_BUF_LEN]; |
3147 | ||
34dc7c2f BB |
3148 | verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_NPARITY, |
3149 | &value) == 0); | |
858219cc | 3150 | (void) snprintf(tmpbuf, sizeof (tmpbuf), "%s%llu", path, |
34dc7c2f | 3151 | (u_longlong_t)value); |
858219cc | 3152 | path = tmpbuf; |
34dc7c2f | 3153 | } |
428870ff BB |
3154 | |
3155 | /* | |
3156 | * We identify each top-level vdev by using a <type-id> | |
3157 | * naming convention. | |
3158 | */ | |
3159 | if (verbose) { | |
3160 | uint64_t id; | |
3161 | ||
3162 | verify(nvlist_lookup_uint64(nv, ZPOOL_CONFIG_ID, | |
3163 | &id) == 0); | |
3164 | (void) snprintf(buf, sizeof (buf), "%s-%llu", path, | |
3165 | (u_longlong_t)id); | |
3166 | path = buf; | |
3167 | } | |
34dc7c2f BB |
3168 | } |
3169 | ||
3170 | return (zfs_strdup(hdl, path)); | |
3171 | } | |
3172 | ||
3173 | static int | |
3174 | zbookmark_compare(const void *a, const void *b) | |
3175 | { | |
3176 | return (memcmp(a, b, sizeof (zbookmark_t))); | |
3177 | } | |
3178 | ||
3179 | /* | |
3180 | * Retrieve the persistent error log, uniquify the members, and return to the | |
3181 | * caller. | |
3182 | */ | |
3183 | int | |
3184 | zpool_get_errlog(zpool_handle_t *zhp, nvlist_t **nverrlistp) | |
3185 | { | |
2598c001 | 3186 | zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 }; |
34dc7c2f BB |
3187 | uint64_t count; |
3188 | zbookmark_t *zb = NULL; | |
3189 | int i; | |
3190 | ||
3191 | /* | |
3192 | * Retrieve the raw error list from the kernel. If the number of errors | |
3193 | * has increased, allocate more space and continue until we get the | |
3194 | * entire list. | |
3195 | */ | |
3196 | verify(nvlist_lookup_uint64(zhp->zpool_config, ZPOOL_CONFIG_ERRCOUNT, | |
3197 | &count) == 0); | |
3198 | if (count == 0) | |
3199 | return (0); | |
3200 | if ((zc.zc_nvlist_dst = (uintptr_t)zfs_alloc(zhp->zpool_hdl, | |
3201 | count * sizeof (zbookmark_t))) == (uintptr_t)NULL) | |
3202 | return (-1); | |
3203 | zc.zc_nvlist_dst_size = count; | |
3204 | (void) strcpy(zc.zc_name, zhp->zpool_name); | |
3205 | for (;;) { | |
3206 | if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_ERROR_LOG, | |
3207 | &zc) != 0) { | |
3208 | free((void *)(uintptr_t)zc.zc_nvlist_dst); | |
3209 | if (errno == ENOMEM) { | |
3210 | count = zc.zc_nvlist_dst_size; | |
3211 | if ((zc.zc_nvlist_dst = (uintptr_t) | |
3212 | zfs_alloc(zhp->zpool_hdl, count * | |
3213 | sizeof (zbookmark_t))) == (uintptr_t)NULL) | |
3214 | return (-1); | |
3215 | } else { | |
3216 | return (-1); | |
3217 | } | |
3218 | } else { | |
3219 | break; | |
3220 | } | |
3221 | } | |
3222 | ||
3223 | /* | |
3224 | * Sort the resulting bookmarks. This is a little confusing due to the | |
3225 | * implementation of ZFS_IOC_ERROR_LOG. The bookmarks are copied last | |
3226 | * to first, and 'zc_nvlist_dst_size' indicates the number of boomarks | |
3227 | * _not_ copied as part of the process. So we point the start of our | |
3228 | * array appropriate and decrement the total number of elements. | |
3229 | */ | |
3230 | zb = ((zbookmark_t *)(uintptr_t)zc.zc_nvlist_dst) + | |
3231 | zc.zc_nvlist_dst_size; | |
3232 | count -= zc.zc_nvlist_dst_size; | |
3233 | ||
3234 | qsort(zb, count, sizeof (zbookmark_t), zbookmark_compare); | |
3235 | ||
3236 | verify(nvlist_alloc(nverrlistp, 0, KM_SLEEP) == 0); | |
3237 | ||
3238 | /* | |
3239 | * Fill in the nverrlistp with nvlist's of dataset and object numbers. | |
3240 | */ | |
3241 | for (i = 0; i < count; i++) { | |
3242 | nvlist_t *nv; | |
3243 | ||
3244 | /* ignoring zb_blkid and zb_level for now */ | |
3245 | if (i > 0 && zb[i-1].zb_objset == zb[i].zb_objset && | |
3246 | zb[i-1].zb_object == zb[i].zb_object) | |
3247 | continue; | |
3248 | ||
3249 | if (nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) != 0) | |
3250 | goto nomem; | |
3251 | if (nvlist_add_uint64(nv, ZPOOL_ERR_DATASET, | |
3252 | zb[i].zb_objset) != 0) { | |
3253 | nvlist_free(nv); | |
3254 | goto nomem; | |
3255 | } | |
3256 | if (nvlist_add_uint64(nv, ZPOOL_ERR_OBJECT, | |
3257 | zb[i].zb_object) != 0) { | |
3258 | nvlist_free(nv); | |
3259 | goto nomem; | |
3260 | } | |
3261 | if (nvlist_add_nvlist(*nverrlistp, "ejk", nv) != 0) { | |
3262 | nvlist_free(nv); | |
3263 | goto nomem; | |
3264 | } | |
3265 | nvlist_free(nv); | |
3266 | } | |
3267 | ||
3268 | free((void *)(uintptr_t)zc.zc_nvlist_dst); | |
3269 | return (0); | |
3270 | ||
3271 | nomem: | |
3272 | free((void *)(uintptr_t)zc.zc_nvlist_dst); | |
3273 | return (no_memory(zhp->zpool_hdl)); | |
3274 | } | |
3275 | ||
3276 | /* | |
3277 | * Upgrade a ZFS pool to the latest on-disk version. | |
3278 | */ | |
3279 | int | |
3280 | zpool_upgrade(zpool_handle_t *zhp, uint64_t new_version) | |
3281 | { | |
2598c001 | 3282 | zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 }; |
34dc7c2f BB |
3283 | libzfs_handle_t *hdl = zhp->zpool_hdl; |
3284 | ||
3285 | (void) strcpy(zc.zc_name, zhp->zpool_name); | |
3286 | zc.zc_cookie = new_version; | |
3287 | ||
3288 | if (zfs_ioctl(hdl, ZFS_IOC_POOL_UPGRADE, &zc) != 0) | |
3289 | return (zpool_standard_error_fmt(hdl, errno, | |
3290 | dgettext(TEXT_DOMAIN, "cannot upgrade '%s'"), | |
3291 | zhp->zpool_name)); | |
3292 | return (0); | |
3293 | } | |
3294 | ||
3295 | void | |
3296 | zpool_set_history_str(const char *subcommand, int argc, char **argv, | |
3297 | char *history_str) | |
3298 | { | |
3299 | int i; | |
3300 | ||
3301 | (void) strlcpy(history_str, subcommand, HIS_MAX_RECORD_LEN); | |
3302 | for (i = 1; i < argc; i++) { | |
3303 | if (strlen(history_str) + 1 + strlen(argv[i]) > | |
3304 | HIS_MAX_RECORD_LEN) | |
3305 | break; | |
3306 | (void) strlcat(history_str, " ", HIS_MAX_RECORD_LEN); | |
3307 | (void) strlcat(history_str, argv[i], HIS_MAX_RECORD_LEN); | |
3308 | } | |
3309 | } | |
3310 | ||
3311 | /* | |
3312 | * Stage command history for logging. | |
3313 | */ | |
3314 | int | |
3315 | zpool_stage_history(libzfs_handle_t *hdl, const char *history_str) | |
3316 | { | |
3317 | if (history_str == NULL) | |
3318 | return (EINVAL); | |
3319 | ||
3320 | if (strlen(history_str) > HIS_MAX_RECORD_LEN) | |
3321 | return (EINVAL); | |
3322 | ||
3323 | if (hdl->libzfs_log_str != NULL) | |
3324 | free(hdl->libzfs_log_str); | |
3325 | ||
3326 | if ((hdl->libzfs_log_str = strdup(history_str)) == NULL) | |
3327 | return (no_memory(hdl)); | |
3328 | ||
3329 | return (0); | |
3330 | } | |
3331 | ||
3332 | /* | |
3333 | * Perform ioctl to get some command history of a pool. | |
3334 | * | |
3335 | * 'buf' is the buffer to fill up to 'len' bytes. 'off' is the | |
3336 | * logical offset of the history buffer to start reading from. | |
3337 | * | |
3338 | * Upon return, 'off' is the next logical offset to read from and | |
3339 | * 'len' is the actual amount of bytes read into 'buf'. | |
3340 | */ | |
3341 | static int | |
3342 | get_history(zpool_handle_t *zhp, char *buf, uint64_t *off, uint64_t *len) | |
3343 | { | |
2598c001 | 3344 | zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 }; |
34dc7c2f BB |
3345 | libzfs_handle_t *hdl = zhp->zpool_hdl; |
3346 | ||
3347 | (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); | |
3348 | ||
3349 | zc.zc_history = (uint64_t)(uintptr_t)buf; | |
3350 | zc.zc_history_len = *len; | |
3351 | zc.zc_history_offset = *off; | |
3352 | ||
3353 | if (ioctl(hdl->libzfs_fd, ZFS_IOC_POOL_GET_HISTORY, &zc) != 0) { | |
3354 | switch (errno) { | |
3355 | case EPERM: | |
3356 | return (zfs_error_fmt(hdl, EZFS_PERM, | |
3357 | dgettext(TEXT_DOMAIN, | |
3358 | "cannot show history for pool '%s'"), | |
3359 | zhp->zpool_name)); | |
3360 | case ENOENT: | |
3361 | return (zfs_error_fmt(hdl, EZFS_NOHISTORY, | |
3362 | dgettext(TEXT_DOMAIN, "cannot get history for pool " | |
3363 | "'%s'"), zhp->zpool_name)); | |
3364 | case ENOTSUP: | |
3365 | return (zfs_error_fmt(hdl, EZFS_BADVERSION, | |
3366 | dgettext(TEXT_DOMAIN, "cannot get history for pool " | |
3367 | "'%s', pool must be upgraded"), zhp->zpool_name)); | |
3368 | default: | |
3369 | return (zpool_standard_error_fmt(hdl, errno, | |
3370 | dgettext(TEXT_DOMAIN, | |
3371 | "cannot get history for '%s'"), zhp->zpool_name)); | |
3372 | } | |
3373 | } | |
3374 | ||
3375 | *len = zc.zc_history_len; | |
3376 | *off = zc.zc_history_offset; | |
3377 | ||
3378 | return (0); | |
3379 | } | |
3380 | ||
3381 | /* | |
3382 | * Process the buffer of nvlists, unpacking and storing each nvlist record | |
3383 | * into 'records'. 'leftover' is set to the number of bytes that weren't | |
3384 | * processed as there wasn't a complete record. | |
3385 | */ | |
428870ff | 3386 | int |
34dc7c2f BB |
3387 | zpool_history_unpack(char *buf, uint64_t bytes_read, uint64_t *leftover, |
3388 | nvlist_t ***records, uint_t *numrecords) | |
3389 | { | |
3390 | uint64_t reclen; | |
3391 | nvlist_t *nv; | |
3392 | int i; | |
3393 | ||
3394 | while (bytes_read > sizeof (reclen)) { | |
3395 | ||
3396 | /* get length of packed record (stored as little endian) */ | |
3397 | for (i = 0, reclen = 0; i < sizeof (reclen); i++) | |
3398 | reclen += (uint64_t)(((uchar_t *)buf)[i]) << (8*i); | |
3399 | ||
3400 | if (bytes_read < sizeof (reclen) + reclen) | |
3401 | break; | |
3402 | ||
3403 | /* unpack record */ | |
3404 | if (nvlist_unpack(buf + sizeof (reclen), reclen, &nv, 0) != 0) | |
3405 | return (ENOMEM); | |
3406 | bytes_read -= sizeof (reclen) + reclen; | |
3407 | buf += sizeof (reclen) + reclen; | |
3408 | ||
3409 | /* add record to nvlist array */ | |
3410 | (*numrecords)++; | |
3411 | if (ISP2(*numrecords + 1)) { | |
3412 | *records = realloc(*records, | |
3413 | *numrecords * 2 * sizeof (nvlist_t *)); | |
3414 | } | |
3415 | (*records)[*numrecords - 1] = nv; | |
3416 | } | |
3417 | ||
3418 | *leftover = bytes_read; | |
3419 | return (0); | |
3420 | } | |
3421 | ||
3422 | #define HIS_BUF_LEN (128*1024) | |
3423 | ||
3424 | /* | |
3425 | * Retrieve the command history of a pool. | |
3426 | */ | |
3427 | int | |
3428 | zpool_get_history(zpool_handle_t *zhp, nvlist_t **nvhisp) | |
3429 | { | |
3430 | char buf[HIS_BUF_LEN]; | |
3431 | uint64_t off = 0; | |
3432 | nvlist_t **records = NULL; | |
3433 | uint_t numrecords = 0; | |
3434 | int err, i; | |
3435 | ||
3436 | do { | |
3437 | uint64_t bytes_read = sizeof (buf); | |
3438 | uint64_t leftover; | |
3439 | ||
3440 | if ((err = get_history(zhp, buf, &off, &bytes_read)) != 0) | |
3441 | break; | |
3442 | ||
3443 | /* if nothing else was read in, we're at EOF, just return */ | |
3444 | if (!bytes_read) | |
3445 | break; | |
3446 | ||
3447 | if ((err = zpool_history_unpack(buf, bytes_read, | |
3448 | &leftover, &records, &numrecords)) != 0) | |
3449 | break; | |
3450 | off -= leftover; | |
3451 | ||
3452 | /* CONSTCOND */ | |
3453 | } while (1); | |
3454 | ||
3455 | if (!err) { | |
3456 | verify(nvlist_alloc(nvhisp, NV_UNIQUE_NAME, 0) == 0); | |
3457 | verify(nvlist_add_nvlist_array(*nvhisp, ZPOOL_HIST_RECORD, | |
3458 | records, numrecords) == 0); | |
3459 | } | |
3460 | for (i = 0; i < numrecords; i++) | |
3461 | nvlist_free(records[i]); | |
3462 | free(records); | |
3463 | ||
3464 | return (err); | |
3465 | } | |
3466 | ||
26685276 BB |
3467 | /* |
3468 | * Retrieve the next event. If there is a new event available 'nvp' will | |
3469 | * contain a newly allocated nvlist and 'dropped' will be set to the number | |
3470 | * of missed events since the last call to this function. When 'nvp' is | |
3471 | * set to NULL it indicates no new events are available. In either case | |
3472 | * the function returns 0 and it is up to the caller to free 'nvp'. In | |
3473 | * the case of a fatal error the function will return a non-zero value. | |
3474 | * When the function is called in blocking mode it will not return until | |
3475 | * a new event is available. | |
3476 | */ | |
3477 | int | |
3478 | zpool_events_next(libzfs_handle_t *hdl, nvlist_t **nvp, | |
3479 | int *dropped, int block, int cleanup_fd) | |
3480 | { | |
3481 | zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 }; | |
3482 | int error = 0; | |
3483 | ||
3484 | *nvp = NULL; | |
3485 | *dropped = 0; | |
3486 | zc.zc_cleanup_fd = cleanup_fd; | |
3487 | ||
3488 | if (!block) | |
3489 | zc.zc_guid = ZEVENT_NONBLOCK; | |
3490 | ||
3491 | if (zcmd_alloc_dst_nvlist(hdl, &zc, ZEVENT_SIZE) != 0) | |
3492 | return (-1); | |
3493 | ||
3494 | retry: | |
3495 | if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_NEXT, &zc) != 0) { | |
3496 | switch (errno) { | |
3497 | case ESHUTDOWN: | |
3498 | error = zfs_error_fmt(hdl, EZFS_POOLUNAVAIL, | |
3499 | dgettext(TEXT_DOMAIN, "zfs shutdown")); | |
3500 | goto out; | |
3501 | case ENOENT: | |
3502 | /* Blocking error case should not occur */ | |
3503 | if (block) | |
3504 | error = zpool_standard_error_fmt(hdl, errno, | |
3505 | dgettext(TEXT_DOMAIN, "cannot get event")); | |
3506 | ||
3507 | goto out; | |
3508 | case ENOMEM: | |
3509 | if (zcmd_expand_dst_nvlist(hdl, &zc) != 0) { | |
3510 | error = zfs_error_fmt(hdl, EZFS_NOMEM, | |
3511 | dgettext(TEXT_DOMAIN, "cannot get event")); | |
3512 | goto out; | |
3513 | } else { | |
3514 | goto retry; | |
3515 | } | |
3516 | default: | |
3517 | error = zpool_standard_error_fmt(hdl, errno, | |
3518 | dgettext(TEXT_DOMAIN, "cannot get event")); | |
3519 | goto out; | |
3520 | } | |
3521 | } | |
3522 | ||
3523 | error = zcmd_read_dst_nvlist(hdl, &zc, nvp); | |
3524 | if (error != 0) | |
3525 | goto out; | |
3526 | ||
3527 | *dropped = (int)zc.zc_cookie; | |
3528 | out: | |
3529 | zcmd_free_nvlists(&zc); | |
3530 | ||
3531 | return (error); | |
3532 | } | |
3533 | ||
3534 | /* | |
3535 | * Clear all events. | |
3536 | */ | |
3537 | int | |
3538 | zpool_events_clear(libzfs_handle_t *hdl, int *count) | |
3539 | { | |
3540 | zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 }; | |
3541 | char msg[1024]; | |
3542 | ||
3543 | (void) snprintf(msg, sizeof (msg), dgettext(TEXT_DOMAIN, | |
3544 | "cannot clear events")); | |
3545 | ||
3546 | if (zfs_ioctl(hdl, ZFS_IOC_EVENTS_CLEAR, &zc) != 0) | |
3547 | return (zpool_standard_error_fmt(hdl, errno, msg)); | |
3548 | ||
3549 | if (count != NULL) | |
3550 | *count = (int)zc.zc_cookie; /* # of events cleared */ | |
3551 | ||
3552 | return (0); | |
3553 | } | |
3554 | ||
34dc7c2f BB |
3555 | void |
3556 | zpool_obj_to_path(zpool_handle_t *zhp, uint64_t dsobj, uint64_t obj, | |
3557 | char *pathname, size_t len) | |
3558 | { | |
2598c001 | 3559 | zfs_cmd_t zc = { "\0", "\0", "\0", "\0", 0 }; |
34dc7c2f BB |
3560 | boolean_t mounted = B_FALSE; |
3561 | char *mntpnt = NULL; | |
3562 | char dsname[MAXNAMELEN]; | |
3563 | ||
3564 | if (dsobj == 0) { | |
3565 | /* special case for the MOS */ | |
b8864a23 | 3566 | (void) snprintf(pathname, len, "<metadata>:<0x%llx>", (longlong_t)obj); |
34dc7c2f BB |
3567 | return; |
3568 | } | |
3569 | ||
3570 | /* get the dataset's name */ | |
3571 | (void) strlcpy(zc.zc_name, zhp->zpool_name, sizeof (zc.zc_name)); | |
3572 | zc.zc_obj = dsobj; | |
3573 | if (ioctl(zhp->zpool_hdl->libzfs_fd, | |
3574 | ZFS_IOC_DSOBJ_TO_DSNAME, &zc) != 0) { | |
3575 | /* just write out a path of two object numbers */ | |
3576 | (void) snprintf(pathname, len, "<0x%llx>:<0x%llx>", | |
b8864a23 | 3577 | (longlong_t)dsobj, (longlong_t)obj); |
34dc7c2f BB |
3578 | return; |
3579 | } | |
3580 | (void) strlcpy(dsname, zc.zc_value, sizeof (dsname)); | |
3581 | ||
3582 | /* find out if the dataset is mounted */ | |
3583 | mounted = is_mounted(zhp->zpool_hdl, dsname, &mntpnt); | |
3584 | ||
3585 | /* get the corrupted object's path */ | |
3586 | (void) strlcpy(zc.zc_name, dsname, sizeof (zc.zc_name)); | |
3587 | zc.zc_obj = obj; | |
3588 | if (ioctl(zhp->zpool_hdl->libzfs_fd, ZFS_IOC_OBJ_TO_PATH, | |
3589 | &zc) == 0) { | |
3590 | if (mounted) { | |
3591 | (void) snprintf(pathname, len, "%s%s", mntpnt, | |
3592 | zc.zc_value); | |
3593 | } else { | |
3594 | (void) snprintf(pathname, len, "%s:%s", | |
3595 | dsname, zc.zc_value); | |
3596 | } | |
3597 | } else { | |
b8864a23 | 3598 | (void) snprintf(pathname, len, "%s:<0x%llx>", dsname, (longlong_t)obj); |
34dc7c2f BB |
3599 | } |
3600 | free(mntpnt); | |
3601 | } | |
3602 | ||
b128c09f BB |
3603 | /* |
3604 | * Read the EFI label from the config, if a label does not exist then | |
3605 | * pass back the error to the caller. If the caller has passed a non-NULL | |
3606 | * diskaddr argument then we set it to the starting address of the EFI | |
3607 | * partition. | |
3608 | */ | |
3609 | static int | |
3610 | read_efi_label(nvlist_t *config, diskaddr_t *sb) | |
3611 | { | |
3612 | char *path; | |
3613 | int fd; | |
3614 | char diskname[MAXPATHLEN]; | |
3615 | int err = -1; | |
3616 | ||
3617 | if (nvlist_lookup_string(config, ZPOOL_CONFIG_PATH, &path) != 0) | |
3618 | return (err); | |
3619 | ||
3620 | (void) snprintf(diskname, sizeof (diskname), "%s%s", RDISK_ROOT, | |
3621 | strrchr(path, '/')); | |
d603ed6c | 3622 | if ((fd = open(diskname, O_RDWR|O_DIRECT)) >= 0) { |
b128c09f BB |
3623 | struct dk_gpt *vtoc; |
3624 | ||
3625 | if ((err = efi_alloc_and_read(fd, &vtoc)) >= 0) { | |
3626 | if (sb != NULL) | |
3627 | *sb = vtoc->efi_parts[0].p_start; | |
3628 | efi_free(vtoc); | |
3629 | } | |
3630 | (void) close(fd); | |
3631 | } | |
3632 | return (err); | |
3633 | } | |
3634 | ||
34dc7c2f BB |
3635 | /* |
3636 | * determine where a partition starts on a disk in the current | |
3637 | * configuration | |
3638 | */ | |
3639 | static diskaddr_t | |
3640 | find_start_block(nvlist_t *config) | |
3641 | { | |
3642 | nvlist_t **child; | |
3643 | uint_t c, children; | |
34dc7c2f | 3644 | diskaddr_t sb = MAXOFFSET_T; |
34dc7c2f BB |
3645 | uint64_t wholedisk; |
3646 | ||
3647 | if (nvlist_lookup_nvlist_array(config, | |
3648 | ZPOOL_CONFIG_CHILDREN, &child, &children) != 0) { | |
3649 | if (nvlist_lookup_uint64(config, | |
3650 | ZPOOL_CONFIG_WHOLE_DISK, | |
3651 | &wholedisk) != 0 || !wholedisk) { | |
3652 | return (MAXOFFSET_T); | |
3653 | } | |
b128c09f BB |
3654 | if (read_efi_label(config, &sb) < 0) |
3655 | sb = MAXOFFSET_T; | |
34dc7c2f BB |
3656 | return (sb); |
3657 | } | |
3658 | ||
3659 | for (c = 0; c < children; c++) { | |
3660 | sb = find_start_block(child[c]); | |
3661 | if (sb != MAXOFFSET_T) { | |
3662 | return (sb); | |
3663 | } | |
3664 | } | |
3665 | return (MAXOFFSET_T); | |
3666 | } | |
3667 | ||
d603ed6c BB |
3668 | int |
3669 | zpool_label_disk_wait(char *path, int timeout) | |
3670 | { | |
3671 | struct stat64 statbuf; | |
3672 | int i; | |
3673 | ||
3674 | /* | |
3675 | * Wait timeout miliseconds for a newly created device to be available | |
3676 | * from the given path. There is a small window when a /dev/ device | |
3677 | * will exist and the udev link will not, so we must wait for the | |
3678 | * symlink. Depending on the udev rules this may take a few seconds. | |
3679 | */ | |
3680 | for (i = 0; i < timeout; i++) { | |
3681 | usleep(1000); | |
3682 | ||
3683 | errno = 0; | |
3684 | if ((stat64(path, &statbuf) == 0) && (errno == 0)) | |
3685 | return (0); | |
3686 | } | |
3687 | ||
3688 | return (ENOENT); | |
3689 | } | |
3690 | ||
3691 | int | |
3692 | zpool_label_disk_check(char *path) | |
3693 | { | |
3694 | struct dk_gpt *vtoc; | |
3695 | int fd, err; | |
3696 | ||
3697 | if ((fd = open(path, O_RDWR|O_DIRECT)) < 0) | |
3698 | return errno; | |
3699 | ||
3700 | if ((err = efi_alloc_and_read(fd, &vtoc)) != 0) { | |
3701 | (void) close(fd); | |
3702 | return err; | |
3703 | } | |
3704 | ||
3705 | if (vtoc->efi_flags & EFI_GPT_PRIMARY_CORRUPT) { | |
3706 | efi_free(vtoc); | |
3707 | (void) close(fd); | |
3708 | return EIDRM; | |
3709 | } | |
3710 | ||
3711 | efi_free(vtoc); | |
3712 | (void) close(fd); | |
3713 | return 0; | |
3714 | } | |
3715 | ||
34dc7c2f BB |
3716 | /* |
3717 | * Label an individual disk. The name provided is the short name, | |
3718 | * stripped of any leading /dev path. | |
3719 | */ | |
3720 | int | |
3721 | zpool_label_disk(libzfs_handle_t *hdl, zpool_handle_t *zhp, char *name) | |
3722 | { | |
3723 | char path[MAXPATHLEN]; | |
3724 | struct dk_gpt *vtoc; | |
d603ed6c | 3725 | int rval, fd; |
34dc7c2f BB |
3726 | size_t resv = EFI_MIN_RESV_SIZE; |
3727 | uint64_t slice_size; | |
3728 | diskaddr_t start_block; | |
3729 | char errbuf[1024]; | |
3730 | ||
3731 | /* prepare an error message just in case */ | |
3732 | (void) snprintf(errbuf, sizeof (errbuf), | |
3733 | dgettext(TEXT_DOMAIN, "cannot label '%s'"), name); | |
3734 | ||
3735 | if (zhp) { | |
3736 | nvlist_t *nvroot; | |
3737 | ||
b128c09f BB |
3738 | if (pool_is_bootable(zhp)) { |
3739 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
3740 | "EFI labeled devices are not supported on root " | |
3741 | "pools.")); | |
3742 | return (zfs_error(hdl, EZFS_POOL_NOTSUP, errbuf)); | |
3743 | } | |
3744 | ||
34dc7c2f BB |
3745 | verify(nvlist_lookup_nvlist(zhp->zpool_config, |
3746 | ZPOOL_CONFIG_VDEV_TREE, &nvroot) == 0); | |
3747 | ||
3748 | if (zhp->zpool_start_block == 0) | |
3749 | start_block = find_start_block(nvroot); | |
3750 | else | |
3751 | start_block = zhp->zpool_start_block; | |
3752 | zhp->zpool_start_block = start_block; | |
3753 | } else { | |
3754 | /* new pool */ | |
3755 | start_block = NEW_START_BLOCK; | |
3756 | } | |
3757 | ||
3758 | (void) snprintf(path, sizeof (path), "%s/%s%s", RDISK_ROOT, name, | |
3759 | BACKUP_SLICE); | |
3760 | ||
d603ed6c | 3761 | if ((fd = open(path, O_RDWR|O_DIRECT)) < 0) { |
34dc7c2f BB |
3762 | /* |
3763 | * This shouldn't happen. We've long since verified that this | |
3764 | * is a valid device. | |
3765 | */ | |
d603ed6c BB |
3766 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, |
3767 | "unable to open device '%s': %d"), path, errno); | |
34dc7c2f BB |
3768 | return (zfs_error(hdl, EZFS_OPENFAILED, errbuf)); |
3769 | } | |
3770 | ||
3771 | if (efi_alloc_and_init(fd, EFI_NUMPAR, &vtoc) != 0) { | |
3772 | /* | |
3773 | * The only way this can fail is if we run out of memory, or we | |
3774 | * were unable to read the disk's capacity | |
3775 | */ | |
3776 | if (errno == ENOMEM) | |
3777 | (void) no_memory(hdl); | |
3778 | ||
3779 | (void) close(fd); | |
3780 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, | |
3781 | "unable to read disk capacity"), name); | |
3782 | ||
3783 | return (zfs_error(hdl, EZFS_NOCAP, errbuf)); | |
3784 | } | |
3785 | ||
3786 | slice_size = vtoc->efi_last_u_lba + 1; | |
3787 | slice_size -= EFI_MIN_RESV_SIZE; | |
3788 | if (start_block == MAXOFFSET_T) | |
3789 | start_block = NEW_START_BLOCK; | |
3790 | slice_size -= start_block; | |
613d88ed | 3791 | slice_size = P2ALIGN(slice_size, PARTITION_END_ALIGNMENT); |
34dc7c2f BB |
3792 | |
3793 | vtoc->efi_parts[0].p_start = start_block; | |
3794 | vtoc->efi_parts[0].p_size = slice_size; | |
3795 | ||
3796 | /* | |
3797 | * Why we use V_USR: V_BACKUP confuses users, and is considered | |
3798 | * disposable by some EFI utilities (since EFI doesn't have a backup | |
3799 | * slice). V_UNASSIGNED is supposed to be used only for zero size | |
3800 | * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT, | |
3801 | * etc. were all pretty specific. V_USR is as close to reality as we | |
3802 | * can get, in the absence of V_OTHER. | |
3803 | */ | |
3804 | vtoc->efi_parts[0].p_tag = V_USR; | |
3805 | (void) strcpy(vtoc->efi_parts[0].p_name, "zfs"); | |
3806 | ||
3807 | vtoc->efi_parts[8].p_start = slice_size + start_block; | |
3808 | vtoc->efi_parts[8].p_size = resv; | |
3809 | vtoc->efi_parts[8].p_tag = V_RESERVED; | |
3810 | ||
d603ed6c | 3811 | if ((rval = efi_write(fd, vtoc)) != 0) { |
34dc7c2f BB |
3812 | /* |
3813 | * Some block drivers (like pcata) may not support EFI | |
3814 | * GPT labels. Print out a helpful error message dir- | |
3815 | * ecting the user to manually label the disk and give | |
3816 | * a specific slice. | |
3817 | */ | |
3818 | (void) close(fd); | |
3819 | efi_free(vtoc); | |
3820 | ||
d603ed6c BB |
3821 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "try using " |
3822 | "parted(8) and then provide a specific slice: %d"), rval); | |
34dc7c2f BB |
3823 | return (zfs_error(hdl, EZFS_LABELFAILED, errbuf)); |
3824 | } | |
3825 | ||
3826 | (void) close(fd); | |
3827 | efi_free(vtoc); | |
34dc7c2f | 3828 | |
d603ed6c BB |
3829 | /* Wait for the first expected slice to appear. */ |
3830 | (void) snprintf(path, sizeof (path), "%s/%s%s%s", DISK_ROOT, name, | |
3831 | isdigit(name[strlen(name)-1]) ? "p" : "", FIRST_SLICE); | |
3832 | rval = zpool_label_disk_wait(path, 3000); | |
3833 | if (rval) { | |
3834 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "failed to " | |
3835 | "detect device partitions on '%s': %d"), path, rval); | |
3836 | return (zfs_error(hdl, EZFS_LABELFAILED, errbuf)); | |
34dc7c2f BB |
3837 | } |
3838 | ||
d603ed6c BB |
3839 | /* We can't be to paranoid. Read the label back and verify it. */ |
3840 | (void) snprintf(path, sizeof (path), "%s/%s", DISK_ROOT, name); | |
3841 | rval = zpool_label_disk_check(path); | |
3842 | if (rval) { | |
3843 | zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "freshly written " | |
3844 | "EFI label on '%s' is damaged. Ensure\nthis device " | |
3845 | "is not in in use, and is functioning properly: %d"), | |
3846 | path, rval); | |
3847 | return (zfs_error(hdl, EZFS_LABELFAILED, errbuf)); | |
34dc7c2f | 3848 | } |
34dc7c2f | 3849 | |
d603ed6c | 3850 | return 0; |
34dc7c2f | 3851 | } |