]>
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 | /* | |
23 | * Copyright 2007 Sun Microsystems, Inc. All rights reserved. | |
24 | * Use is subject to license terms. | |
25 | */ | |
26 | ||
27 | #pragma ident "@(#)zpool_vdev.c 1.15 07/11/09 SMI" | |
28 | ||
29 | /* | |
30 | * Functions to convert between a list of vdevs and an nvlist representing the | |
31 | * configuration. Each entry in the list can be one of: | |
32 | * | |
33 | * Device vdevs | |
34 | * disk=(path=..., devid=...) | |
35 | * file=(path=...) | |
36 | * | |
37 | * Group vdevs | |
38 | * raidz[1|2]=(...) | |
39 | * mirror=(...) | |
40 | * | |
41 | * Hot spares | |
42 | * | |
43 | * While the underlying implementation supports it, group vdevs cannot contain | |
44 | * other group vdevs. All userland verification of devices is contained within | |
45 | * this file. If successful, the nvlist returned can be passed directly to the | |
46 | * kernel; we've done as much verification as possible in userland. | |
47 | * | |
48 | * Hot spares are a special case, and passed down as an array of disk vdevs, at | |
49 | * the same level as the root of the vdev tree. | |
50 | * | |
51 | * The only function exported by this file is 'make_root_vdev'. The | |
52 | * function performs several passes: | |
53 | * | |
54 | * 1. Construct the vdev specification. Performs syntax validation and | |
55 | * makes sure each device is valid. | |
56 | * 2. Check for devices in use. Using libdiskmgt, makes sure that no | |
57 | * devices are also in use. Some can be overridden using the 'force' | |
58 | * flag, others cannot. | |
59 | * 3. Check for replication errors if the 'force' flag is not specified. | |
60 | * validates that the replication level is consistent across the | |
61 | * entire pool. | |
62 | * 4. Call libzfs to label any whole disks with an EFI label. | |
63 | */ | |
64 | ||
65 | #include <assert.h> | |
66 | #include <devid.h> | |
67 | #include <errno.h> | |
68 | #include <fcntl.h> | |
69 | #include <libdiskmgt.h> | |
70 | #include <libintl.h> | |
71 | #include <libnvpair.h> | |
72 | #include <stdio.h> | |
73 | #include <string.h> | |
74 | #include <unistd.h> | |
75 | #include <sys/efi_partition.h> | |
76 | #include <sys/stat.h> | |
77 | #include <sys/vtoc.h> | |
78 | #include <sys/mntent.h> | |
79 | ||
80 | #include "zpool_util.h" | |
81 | ||
82 | #define DISK_ROOT "/dev/dsk" | |
83 | #define RDISK_ROOT "/dev/rdsk" | |
84 | #define BACKUP_SLICE "s2" | |
85 | ||
86 | /* | |
87 | * For any given vdev specification, we can have multiple errors. The | |
88 | * vdev_error() function keeps track of whether we have seen an error yet, and | |
89 | * prints out a header if its the first error we've seen. | |
90 | */ | |
91 | boolean_t error_seen; | |
92 | boolean_t is_force; | |
93 | ||
94 | /*PRINTFLIKE1*/ | |
95 | static void | |
96 | vdev_error(const char *fmt, ...) | |
97 | { | |
98 | va_list ap; | |
99 | ||
100 | if (!error_seen) { | |
101 | (void) fprintf(stderr, gettext("invalid vdev specification\n")); | |
102 | if (!is_force) | |
103 | (void) fprintf(stderr, gettext("use '-f' to override " | |
104 | "the following errors:\n")); | |
105 | else | |
106 | (void) fprintf(stderr, gettext("the following errors " | |
107 | "must be manually repaired:\n")); | |
108 | error_seen = B_TRUE; | |
109 | } | |
110 | ||
111 | va_start(ap, fmt); | |
112 | (void) vfprintf(stderr, fmt, ap); | |
113 | va_end(ap); | |
114 | } | |
115 | ||
116 | static void | |
117 | libdiskmgt_error(int error) | |
118 | { | |
119 | /* | |
120 | * ENXIO/ENODEV is a valid error message if the device doesn't live in | |
121 | * /dev/dsk. Don't bother printing an error message in this case. | |
122 | */ | |
123 | if (error == ENXIO || error == ENODEV) | |
124 | return; | |
125 | ||
126 | (void) fprintf(stderr, gettext("warning: device in use checking " | |
127 | "failed: %s\n"), strerror(error)); | |
128 | } | |
129 | ||
130 | /* | |
131 | * Validate a device, passing the bulk of the work off to libdiskmgt. | |
132 | */ | |
133 | static int | |
134 | check_slice(const char *path, int force, boolean_t wholedisk, boolean_t isspare) | |
135 | { | |
136 | char *msg; | |
137 | int error = 0; | |
138 | dm_who_type_t who; | |
139 | ||
140 | if (force) | |
141 | who = DM_WHO_ZPOOL_FORCE; | |
142 | else if (isspare) | |
143 | who = DM_WHO_ZPOOL_SPARE; | |
144 | else | |
145 | who = DM_WHO_ZPOOL; | |
146 | ||
147 | if (dm_inuse((char *)path, &msg, who, &error) || error) { | |
148 | if (error != 0) { | |
149 | libdiskmgt_error(error); | |
150 | return (0); | |
151 | } else { | |
152 | vdev_error("%s", msg); | |
153 | free(msg); | |
154 | return (-1); | |
155 | } | |
156 | } | |
157 | ||
158 | /* | |
159 | * If we're given a whole disk, ignore overlapping slices since we're | |
160 | * about to label it anyway. | |
161 | */ | |
162 | error = 0; | |
163 | if (!wholedisk && !force && | |
164 | (dm_isoverlapping((char *)path, &msg, &error) || error)) { | |
165 | if (error == 0) { | |
166 | /* dm_isoverlapping returned -1 */ | |
167 | vdev_error(gettext("%s overlaps with %s\n"), path, msg); | |
168 | free(msg); | |
169 | return (-1); | |
170 | } else if (error != ENODEV) { | |
171 | /* libdiskmgt's devcache only handles physical drives */ | |
172 | libdiskmgt_error(error); | |
173 | return (0); | |
174 | } | |
175 | } | |
176 | ||
177 | return (0); | |
178 | } | |
179 | ||
180 | ||
181 | /* | |
182 | * Validate a whole disk. Iterate over all slices on the disk and make sure | |
183 | * that none is in use by calling check_slice(). | |
184 | */ | |
185 | static int | |
186 | check_disk(const char *name, dm_descriptor_t disk, int force, int isspare) | |
187 | { | |
188 | dm_descriptor_t *drive, *media, *slice; | |
189 | int err = 0; | |
190 | int i; | |
191 | int ret; | |
192 | ||
193 | /* | |
194 | * Get the drive associated with this disk. This should never fail, | |
195 | * because we already have an alias handle open for the device. | |
196 | */ | |
197 | if ((drive = dm_get_associated_descriptors(disk, DM_DRIVE, | |
198 | &err)) == NULL || *drive == NULL) { | |
199 | if (err) | |
200 | libdiskmgt_error(err); | |
201 | return (0); | |
202 | } | |
203 | ||
204 | if ((media = dm_get_associated_descriptors(*drive, DM_MEDIA, | |
205 | &err)) == NULL) { | |
206 | dm_free_descriptors(drive); | |
207 | if (err) | |
208 | libdiskmgt_error(err); | |
209 | return (0); | |
210 | } | |
211 | ||
212 | dm_free_descriptors(drive); | |
213 | ||
214 | /* | |
215 | * It is possible that the user has specified a removable media drive, | |
216 | * and the media is not present. | |
217 | */ | |
218 | if (*media == NULL) { | |
219 | dm_free_descriptors(media); | |
220 | vdev_error(gettext("'%s' has no media in drive\n"), name); | |
221 | return (-1); | |
222 | } | |
223 | ||
224 | if ((slice = dm_get_associated_descriptors(*media, DM_SLICE, | |
225 | &err)) == NULL) { | |
226 | dm_free_descriptors(media); | |
227 | if (err) | |
228 | libdiskmgt_error(err); | |
229 | return (0); | |
230 | } | |
231 | ||
232 | dm_free_descriptors(media); | |
233 | ||
234 | ret = 0; | |
235 | ||
236 | /* | |
237 | * Iterate over all slices and report any errors. We don't care about | |
238 | * overlapping slices because we are using the whole disk. | |
239 | */ | |
240 | for (i = 0; slice[i] != NULL; i++) { | |
241 | char *name = dm_get_name(slice[i], &err); | |
242 | ||
243 | if (check_slice(name, force, B_TRUE, isspare) != 0) | |
244 | ret = -1; | |
245 | ||
246 | dm_free_name(name); | |
247 | } | |
248 | ||
249 | dm_free_descriptors(slice); | |
250 | return (ret); | |
251 | } | |
252 | ||
253 | /* | |
254 | * Validate a device. | |
255 | */ | |
256 | static int | |
257 | check_device(const char *path, boolean_t force, boolean_t isspare) | |
258 | { | |
259 | dm_descriptor_t desc; | |
260 | int err; | |
261 | char *dev; | |
262 | ||
263 | /* | |
264 | * For whole disks, libdiskmgt does not include the leading dev path. | |
265 | */ | |
266 | dev = strrchr(path, '/'); | |
267 | assert(dev != NULL); | |
268 | dev++; | |
269 | if ((desc = dm_get_descriptor_by_name(DM_ALIAS, dev, &err)) != NULL) { | |
270 | err = check_disk(path, desc, force, isspare); | |
271 | dm_free_descriptor(desc); | |
272 | return (err); | |
273 | } | |
274 | ||
275 | return (check_slice(path, force, B_FALSE, isspare)); | |
276 | } | |
277 | ||
278 | /* | |
279 | * Check that a file is valid. All we can do in this case is check that it's | |
280 | * not in use by another pool, and not in use by swap. | |
281 | */ | |
282 | static int | |
283 | check_file(const char *file, boolean_t force, boolean_t isspare) | |
284 | { | |
285 | char *name; | |
286 | int fd; | |
287 | int ret = 0; | |
288 | int err; | |
289 | pool_state_t state; | |
290 | boolean_t inuse; | |
291 | ||
292 | if (dm_inuse_swap(file, &err)) { | |
293 | if (err) | |
294 | libdiskmgt_error(err); | |
295 | else | |
296 | vdev_error(gettext("%s is currently used by swap. " | |
297 | "Please see swap(1M).\n"), file); | |
298 | return (-1); | |
299 | } | |
300 | ||
301 | if ((fd = open(file, O_RDONLY)) < 0) | |
302 | return (0); | |
303 | ||
304 | if (zpool_in_use(g_zfs, fd, &state, &name, &inuse) == 0 && inuse) { | |
305 | const char *desc; | |
306 | ||
307 | switch (state) { | |
308 | case POOL_STATE_ACTIVE: | |
309 | desc = gettext("active"); | |
310 | break; | |
311 | ||
312 | case POOL_STATE_EXPORTED: | |
313 | desc = gettext("exported"); | |
314 | break; | |
315 | ||
316 | case POOL_STATE_POTENTIALLY_ACTIVE: | |
317 | desc = gettext("potentially active"); | |
318 | break; | |
319 | ||
320 | default: | |
321 | desc = gettext("unknown"); | |
322 | break; | |
323 | } | |
324 | ||
325 | /* | |
326 | * Allow hot spares to be shared between pools. | |
327 | */ | |
328 | if (state == POOL_STATE_SPARE && isspare) | |
329 | return (0); | |
330 | ||
331 | if (state == POOL_STATE_ACTIVE || | |
332 | state == POOL_STATE_SPARE || !force) { | |
333 | switch (state) { | |
334 | case POOL_STATE_SPARE: | |
335 | vdev_error(gettext("%s is reserved as a hot " | |
336 | "spare for pool %s\n"), file, name); | |
337 | break; | |
338 | default: | |
339 | vdev_error(gettext("%s is part of %s pool " | |
340 | "'%s'\n"), file, desc, name); | |
341 | break; | |
342 | } | |
343 | ret = -1; | |
344 | } | |
345 | ||
346 | free(name); | |
347 | } | |
348 | ||
349 | (void) close(fd); | |
350 | return (ret); | |
351 | } | |
352 | ||
353 | ||
354 | /* | |
355 | * By "whole disk" we mean an entire physical disk (something we can | |
356 | * label, toggle the write cache on, etc.) as opposed to the full | |
357 | * capacity of a pseudo-device such as lofi or did. We act as if we | |
358 | * are labeling the disk, which should be a pretty good test of whether | |
359 | * it's a viable device or not. Returns B_TRUE if it is and B_FALSE if | |
360 | * it isn't. | |
361 | */ | |
362 | static boolean_t | |
363 | is_whole_disk(const char *arg) | |
364 | { | |
365 | struct dk_gpt *label; | |
366 | int fd; | |
367 | char path[MAXPATHLEN]; | |
368 | ||
369 | (void) snprintf(path, sizeof (path), "%s%s%s", | |
370 | RDISK_ROOT, strrchr(arg, '/'), BACKUP_SLICE); | |
371 | if ((fd = open(path, O_RDWR | O_NDELAY)) < 0) | |
372 | return (B_FALSE); | |
373 | if (efi_alloc_and_init(fd, EFI_NUMPAR, &label) != 0) { | |
374 | (void) close(fd); | |
375 | return (B_FALSE); | |
376 | } | |
377 | efi_free(label); | |
378 | (void) close(fd); | |
379 | return (B_TRUE); | |
380 | } | |
381 | ||
382 | /* | |
383 | * Create a leaf vdev. Determine if this is a file or a device. If it's a | |
384 | * device, fill in the device id to make a complete nvlist. Valid forms for a | |
385 | * leaf vdev are: | |
386 | * | |
387 | * /dev/dsk/xxx Complete disk path | |
388 | * /xxx Full path to file | |
389 | * xxx Shorthand for /dev/dsk/xxx | |
390 | */ | |
391 | static nvlist_t * | |
392 | make_leaf_vdev(const char *arg, uint64_t is_log) | |
393 | { | |
394 | char path[MAXPATHLEN]; | |
395 | struct stat64 statbuf; | |
396 | nvlist_t *vdev = NULL; | |
397 | char *type = NULL; | |
398 | boolean_t wholedisk = B_FALSE; | |
399 | ||
400 | /* | |
401 | * Determine what type of vdev this is, and put the full path into | |
402 | * 'path'. We detect whether this is a device of file afterwards by | |
403 | * checking the st_mode of the file. | |
404 | */ | |
405 | if (arg[0] == '/') { | |
406 | /* | |
407 | * Complete device or file path. Exact type is determined by | |
408 | * examining the file descriptor afterwards. | |
409 | */ | |
410 | wholedisk = is_whole_disk(arg); | |
411 | if (!wholedisk && (stat64(arg, &statbuf) != 0)) { | |
412 | (void) fprintf(stderr, | |
413 | gettext("cannot open '%s': %s\n"), | |
414 | arg, strerror(errno)); | |
415 | return (NULL); | |
416 | } | |
417 | ||
418 | (void) strlcpy(path, arg, sizeof (path)); | |
419 | } else { | |
420 | /* | |
421 | * This may be a short path for a device, or it could be total | |
422 | * gibberish. Check to see if it's a known device in | |
423 | * /dev/dsk/. As part of this check, see if we've been given a | |
424 | * an entire disk (minus the slice number). | |
425 | */ | |
426 | (void) snprintf(path, sizeof (path), "%s/%s", DISK_ROOT, | |
427 | arg); | |
428 | wholedisk = is_whole_disk(path); | |
429 | if (!wholedisk && (stat64(path, &statbuf) != 0)) { | |
430 | /* | |
431 | * If we got ENOENT, then the user gave us | |
432 | * gibberish, so try to direct them with a | |
433 | * reasonable error message. Otherwise, | |
434 | * regurgitate strerror() since it's the best we | |
435 | * can do. | |
436 | */ | |
437 | if (errno == ENOENT) { | |
438 | (void) fprintf(stderr, | |
439 | gettext("cannot open '%s': no such " | |
440 | "device in %s\n"), arg, DISK_ROOT); | |
441 | (void) fprintf(stderr, | |
442 | gettext("must be a full path or " | |
443 | "shorthand device name\n")); | |
444 | return (NULL); | |
445 | } else { | |
446 | (void) fprintf(stderr, | |
447 | gettext("cannot open '%s': %s\n"), | |
448 | path, strerror(errno)); | |
449 | return (NULL); | |
450 | } | |
451 | } | |
452 | } | |
453 | ||
454 | /* | |
455 | * Determine whether this is a device or a file. | |
456 | */ | |
457 | if (wholedisk || S_ISBLK(statbuf.st_mode)) { | |
458 | type = VDEV_TYPE_DISK; | |
459 | } else if (S_ISREG(statbuf.st_mode)) { | |
460 | type = VDEV_TYPE_FILE; | |
461 | } else { | |
462 | (void) fprintf(stderr, gettext("cannot use '%s': must be a " | |
463 | "block device or regular file\n"), path); | |
464 | return (NULL); | |
465 | } | |
466 | ||
467 | /* | |
468 | * Finally, we have the complete device or file, and we know that it is | |
469 | * acceptable to use. Construct the nvlist to describe this vdev. All | |
470 | * vdevs have a 'path' element, and devices also have a 'devid' element. | |
471 | */ | |
472 | verify(nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) == 0); | |
473 | verify(nvlist_add_string(vdev, ZPOOL_CONFIG_PATH, path) == 0); | |
474 | verify(nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE, type) == 0); | |
475 | verify(nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_LOG, is_log) == 0); | |
476 | if (strcmp(type, VDEV_TYPE_DISK) == 0) | |
477 | verify(nvlist_add_uint64(vdev, ZPOOL_CONFIG_WHOLE_DISK, | |
478 | (uint64_t)wholedisk) == 0); | |
479 | ||
480 | /* | |
481 | * For a whole disk, defer getting its devid until after labeling it. | |
482 | */ | |
483 | if (S_ISBLK(statbuf.st_mode) && !wholedisk) { | |
484 | /* | |
485 | * Get the devid for the device. | |
486 | */ | |
487 | int fd; | |
488 | ddi_devid_t devid; | |
489 | char *minor = NULL, *devid_str = NULL; | |
490 | ||
491 | if ((fd = open(path, O_RDONLY)) < 0) { | |
492 | (void) fprintf(stderr, gettext("cannot open '%s': " | |
493 | "%s\n"), path, strerror(errno)); | |
494 | nvlist_free(vdev); | |
495 | return (NULL); | |
496 | } | |
497 | ||
498 | if (devid_get(fd, &devid) == 0) { | |
499 | if (devid_get_minor_name(fd, &minor) == 0 && | |
500 | (devid_str = devid_str_encode(devid, minor)) != | |
501 | NULL) { | |
502 | verify(nvlist_add_string(vdev, | |
503 | ZPOOL_CONFIG_DEVID, devid_str) == 0); | |
504 | } | |
505 | if (devid_str != NULL) | |
506 | devid_str_free(devid_str); | |
507 | if (minor != NULL) | |
508 | devid_str_free(minor); | |
509 | devid_free(devid); | |
510 | } | |
511 | ||
512 | (void) close(fd); | |
513 | } | |
514 | ||
515 | return (vdev); | |
516 | } | |
517 | ||
518 | /* | |
519 | * Go through and verify the replication level of the pool is consistent. | |
520 | * Performs the following checks: | |
521 | * | |
522 | * For the new spec, verifies that devices in mirrors and raidz are the | |
523 | * same size. | |
524 | * | |
525 | * If the current configuration already has inconsistent replication | |
526 | * levels, ignore any other potential problems in the new spec. | |
527 | * | |
528 | * Otherwise, make sure that the current spec (if there is one) and the new | |
529 | * spec have consistent replication levels. | |
530 | */ | |
531 | typedef struct replication_level { | |
532 | char *zprl_type; | |
533 | uint64_t zprl_children; | |
534 | uint64_t zprl_parity; | |
535 | } replication_level_t; | |
536 | ||
537 | #define ZPOOL_FUZZ (16 * 1024 * 1024) | |
538 | ||
539 | /* | |
540 | * Given a list of toplevel vdevs, return the current replication level. If | |
541 | * the config is inconsistent, then NULL is returned. If 'fatal' is set, then | |
542 | * an error message will be displayed for each self-inconsistent vdev. | |
543 | */ | |
544 | static replication_level_t * | |
545 | get_replication(nvlist_t *nvroot, boolean_t fatal) | |
546 | { | |
547 | nvlist_t **top; | |
548 | uint_t t, toplevels; | |
549 | nvlist_t **child; | |
550 | uint_t c, children; | |
551 | nvlist_t *nv; | |
552 | char *type; | |
553 | replication_level_t lastrep, rep, *ret; | |
554 | boolean_t dontreport; | |
555 | ||
556 | ret = safe_malloc(sizeof (replication_level_t)); | |
557 | ||
558 | verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, | |
559 | &top, &toplevels) == 0); | |
560 | ||
561 | lastrep.zprl_type = NULL; | |
562 | for (t = 0; t < toplevels; t++) { | |
563 | uint64_t is_log = B_FALSE; | |
564 | ||
565 | nv = top[t]; | |
566 | ||
567 | /* | |
568 | * For separate logs we ignore the top level vdev replication | |
569 | * constraints. | |
570 | */ | |
571 | (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG, &is_log); | |
572 | if (is_log) | |
573 | continue; | |
574 | ||
575 | verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, | |
576 | &type) == 0); | |
577 | if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, | |
578 | &child, &children) != 0) { | |
579 | /* | |
580 | * This is a 'file' or 'disk' vdev. | |
581 | */ | |
582 | rep.zprl_type = type; | |
583 | rep.zprl_children = 1; | |
584 | rep.zprl_parity = 0; | |
585 | } else { | |
586 | uint64_t vdev_size; | |
587 | ||
588 | /* | |
589 | * This is a mirror or RAID-Z vdev. Go through and make | |
590 | * sure the contents are all the same (files vs. disks), | |
591 | * keeping track of the number of elements in the | |
592 | * process. | |
593 | * | |
594 | * We also check that the size of each vdev (if it can | |
595 | * be determined) is the same. | |
596 | */ | |
597 | rep.zprl_type = type; | |
598 | rep.zprl_children = 0; | |
599 | ||
600 | if (strcmp(type, VDEV_TYPE_RAIDZ) == 0) { | |
601 | verify(nvlist_lookup_uint64(nv, | |
602 | ZPOOL_CONFIG_NPARITY, | |
603 | &rep.zprl_parity) == 0); | |
604 | assert(rep.zprl_parity != 0); | |
605 | } else { | |
606 | rep.zprl_parity = 0; | |
607 | } | |
608 | ||
609 | /* | |
610 | * The 'dontreport' variable indicates that we've | |
611 | * already reported an error for this spec, so don't | |
612 | * bother doing it again. | |
613 | */ | |
614 | type = NULL; | |
615 | dontreport = 0; | |
616 | vdev_size = -1ULL; | |
617 | for (c = 0; c < children; c++) { | |
618 | nvlist_t *cnv = child[c]; | |
619 | char *path; | |
620 | struct stat64 statbuf; | |
621 | uint64_t size = -1ULL; | |
622 | char *childtype; | |
623 | int fd, err; | |
624 | ||
625 | rep.zprl_children++; | |
626 | ||
627 | verify(nvlist_lookup_string(cnv, | |
628 | ZPOOL_CONFIG_TYPE, &childtype) == 0); | |
629 | ||
630 | /* | |
631 | * If this is a replacing or spare vdev, then | |
632 | * get the real first child of the vdev. | |
633 | */ | |
634 | if (strcmp(childtype, | |
635 | VDEV_TYPE_REPLACING) == 0 || | |
636 | strcmp(childtype, VDEV_TYPE_SPARE) == 0) { | |
637 | nvlist_t **rchild; | |
638 | uint_t rchildren; | |
639 | ||
640 | verify(nvlist_lookup_nvlist_array(cnv, | |
641 | ZPOOL_CONFIG_CHILDREN, &rchild, | |
642 | &rchildren) == 0); | |
643 | assert(rchildren == 2); | |
644 | cnv = rchild[0]; | |
645 | ||
646 | verify(nvlist_lookup_string(cnv, | |
647 | ZPOOL_CONFIG_TYPE, | |
648 | &childtype) == 0); | |
649 | } | |
650 | ||
651 | verify(nvlist_lookup_string(cnv, | |
652 | ZPOOL_CONFIG_PATH, &path) == 0); | |
653 | ||
654 | /* | |
655 | * If we have a raidz/mirror that combines disks | |
656 | * with files, report it as an error. | |
657 | */ | |
658 | if (!dontreport && type != NULL && | |
659 | strcmp(type, childtype) != 0) { | |
660 | if (ret != NULL) | |
661 | free(ret); | |
662 | ret = NULL; | |
663 | if (fatal) | |
664 | vdev_error(gettext( | |
665 | "mismatched replication " | |
666 | "level: %s contains both " | |
667 | "files and devices\n"), | |
668 | rep.zprl_type); | |
669 | else | |
670 | return (NULL); | |
671 | dontreport = B_TRUE; | |
672 | } | |
673 | ||
674 | /* | |
675 | * According to stat(2), the value of 'st_size' | |
676 | * is undefined for block devices and character | |
677 | * devices. But there is no effective way to | |
678 | * determine the real size in userland. | |
679 | * | |
680 | * Instead, we'll take advantage of an | |
681 | * implementation detail of spec_size(). If the | |
682 | * device is currently open, then we (should) | |
683 | * return a valid size. | |
684 | * | |
685 | * If we still don't get a valid size (indicated | |
686 | * by a size of 0 or MAXOFFSET_T), then ignore | |
687 | * this device altogether. | |
688 | */ | |
689 | if ((fd = open(path, O_RDONLY)) >= 0) { | |
690 | err = fstat64(fd, &statbuf); | |
691 | (void) close(fd); | |
692 | } else { | |
693 | err = stat64(path, &statbuf); | |
694 | } | |
695 | ||
696 | if (err != 0 || | |
697 | statbuf.st_size == 0 || | |
698 | statbuf.st_size == MAXOFFSET_T) | |
699 | continue; | |
700 | ||
701 | size = statbuf.st_size; | |
702 | ||
703 | /* | |
704 | * Also make sure that devices and | |
705 | * slices have a consistent size. If | |
706 | * they differ by a significant amount | |
707 | * (~16MB) then report an error. | |
708 | */ | |
709 | if (!dontreport && | |
710 | (vdev_size != -1ULL && | |
711 | (labs(size - vdev_size) > | |
712 | ZPOOL_FUZZ))) { | |
713 | if (ret != NULL) | |
714 | free(ret); | |
715 | ret = NULL; | |
716 | if (fatal) | |
717 | vdev_error(gettext( | |
718 | "%s contains devices of " | |
719 | "different sizes\n"), | |
720 | rep.zprl_type); | |
721 | else | |
722 | return (NULL); | |
723 | dontreport = B_TRUE; | |
724 | } | |
725 | ||
726 | type = childtype; | |
727 | vdev_size = size; | |
728 | } | |
729 | } | |
730 | ||
731 | /* | |
732 | * At this point, we have the replication of the last toplevel | |
733 | * vdev in 'rep'. Compare it to 'lastrep' to see if its | |
734 | * different. | |
735 | */ | |
736 | if (lastrep.zprl_type != NULL) { | |
737 | if (strcmp(lastrep.zprl_type, rep.zprl_type) != 0) { | |
738 | if (ret != NULL) | |
739 | free(ret); | |
740 | ret = NULL; | |
741 | if (fatal) | |
742 | vdev_error(gettext( | |
743 | "mismatched replication level: " | |
744 | "both %s and %s vdevs are " | |
745 | "present\n"), | |
746 | lastrep.zprl_type, rep.zprl_type); | |
747 | else | |
748 | return (NULL); | |
749 | } else if (lastrep.zprl_parity != rep.zprl_parity) { | |
750 | if (ret) | |
751 | free(ret); | |
752 | ret = NULL; | |
753 | if (fatal) | |
754 | vdev_error(gettext( | |
755 | "mismatched replication level: " | |
756 | "both %llu and %llu device parity " | |
757 | "%s vdevs are present\n"), | |
758 | lastrep.zprl_parity, | |
759 | rep.zprl_parity, | |
760 | rep.zprl_type); | |
761 | else | |
762 | return (NULL); | |
763 | } else if (lastrep.zprl_children != rep.zprl_children) { | |
764 | if (ret) | |
765 | free(ret); | |
766 | ret = NULL; | |
767 | if (fatal) | |
768 | vdev_error(gettext( | |
769 | "mismatched replication level: " | |
770 | "both %llu-way and %llu-way %s " | |
771 | "vdevs are present\n"), | |
772 | lastrep.zprl_children, | |
773 | rep.zprl_children, | |
774 | rep.zprl_type); | |
775 | else | |
776 | return (NULL); | |
777 | } | |
778 | } | |
779 | lastrep = rep; | |
780 | } | |
781 | ||
782 | if (ret != NULL) | |
783 | *ret = rep; | |
784 | ||
785 | return (ret); | |
786 | } | |
787 | ||
788 | /* | |
789 | * Check the replication level of the vdev spec against the current pool. Calls | |
790 | * get_replication() to make sure the new spec is self-consistent. If the pool | |
791 | * has a consistent replication level, then we ignore any errors. Otherwise, | |
792 | * report any difference between the two. | |
793 | */ | |
794 | static int | |
795 | check_replication(nvlist_t *config, nvlist_t *newroot) | |
796 | { | |
797 | nvlist_t **child; | |
798 | uint_t children; | |
799 | replication_level_t *current = NULL, *new; | |
800 | int ret; | |
801 | ||
802 | /* | |
803 | * If we have a current pool configuration, check to see if it's | |
804 | * self-consistent. If not, simply return success. | |
805 | */ | |
806 | if (config != NULL) { | |
807 | nvlist_t *nvroot; | |
808 | ||
809 | verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, | |
810 | &nvroot) == 0); | |
811 | if ((current = get_replication(nvroot, B_FALSE)) == NULL) | |
812 | return (0); | |
813 | } | |
814 | /* | |
815 | * for spares there may be no children, and therefore no | |
816 | * replication level to check | |
817 | */ | |
818 | if ((nvlist_lookup_nvlist_array(newroot, ZPOOL_CONFIG_CHILDREN, | |
819 | &child, &children) != 0) || (children == 0)) { | |
820 | free(current); | |
821 | return (0); | |
822 | } | |
823 | ||
824 | /* | |
825 | * If all we have is logs then there's no replication level to check. | |
826 | */ | |
827 | if (num_logs(newroot) == children) { | |
828 | free(current); | |
829 | return (0); | |
830 | } | |
831 | ||
832 | /* | |
833 | * Get the replication level of the new vdev spec, reporting any | |
834 | * inconsistencies found. | |
835 | */ | |
836 | if ((new = get_replication(newroot, B_TRUE)) == NULL) { | |
837 | free(current); | |
838 | return (-1); | |
839 | } | |
840 | ||
841 | /* | |
842 | * Check to see if the new vdev spec matches the replication level of | |
843 | * the current pool. | |
844 | */ | |
845 | ret = 0; | |
846 | if (current != NULL) { | |
847 | if (strcmp(current->zprl_type, new->zprl_type) != 0) { | |
848 | vdev_error(gettext( | |
849 | "mismatched replication level: pool uses %s " | |
850 | "and new vdev is %s\n"), | |
851 | current->zprl_type, new->zprl_type); | |
852 | ret = -1; | |
853 | } else if (current->zprl_parity != new->zprl_parity) { | |
854 | vdev_error(gettext( | |
855 | "mismatched replication level: pool uses %llu " | |
856 | "device parity and new vdev uses %llu\n"), | |
857 | current->zprl_parity, new->zprl_parity); | |
858 | ret = -1; | |
859 | } else if (current->zprl_children != new->zprl_children) { | |
860 | vdev_error(gettext( | |
861 | "mismatched replication level: pool uses %llu-way " | |
862 | "%s and new vdev uses %llu-way %s\n"), | |
863 | current->zprl_children, current->zprl_type, | |
864 | new->zprl_children, new->zprl_type); | |
865 | ret = -1; | |
866 | } | |
867 | } | |
868 | ||
869 | free(new); | |
870 | if (current != NULL) | |
871 | free(current); | |
872 | ||
873 | return (ret); | |
874 | } | |
875 | ||
876 | /* | |
877 | * Go through and find any whole disks in the vdev specification, labelling them | |
878 | * as appropriate. When constructing the vdev spec, we were unable to open this | |
879 | * device in order to provide a devid. Now that we have labelled the disk and | |
880 | * know that slice 0 is valid, we can construct the devid now. | |
881 | * | |
882 | * If the disk was already labeled with an EFI label, we will have gotten the | |
883 | * devid already (because we were able to open the whole disk). Otherwise, we | |
884 | * need to get the devid after we label the disk. | |
885 | */ | |
886 | static int | |
887 | make_disks(zpool_handle_t *zhp, nvlist_t *nv) | |
888 | { | |
889 | nvlist_t **child; | |
890 | uint_t c, children; | |
891 | char *type, *path, *diskname; | |
892 | char buf[MAXPATHLEN]; | |
893 | uint64_t wholedisk; | |
894 | int fd; | |
895 | int ret; | |
896 | ddi_devid_t devid; | |
897 | char *minor = NULL, *devid_str = NULL; | |
898 | ||
899 | verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0); | |
900 | ||
901 | if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, | |
902 | &child, &children) != 0) { | |
903 | ||
904 | if (strcmp(type, VDEV_TYPE_DISK) != 0) | |
905 | return (0); | |
906 | ||
907 | /* | |
908 | * We have a disk device. Get the path to the device | |
909 | * and see if it's a whole disk by appending the backup | |
910 | * slice and stat()ing the device. | |
911 | */ | |
912 | verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0); | |
913 | if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, | |
914 | &wholedisk) != 0 || !wholedisk) | |
915 | return (0); | |
916 | ||
917 | diskname = strrchr(path, '/'); | |
918 | assert(diskname != NULL); | |
919 | diskname++; | |
920 | if (zpool_label_disk(g_zfs, zhp, diskname) == -1) | |
921 | return (-1); | |
922 | ||
923 | /* | |
924 | * Fill in the devid, now that we've labeled the disk. | |
925 | */ | |
926 | (void) snprintf(buf, sizeof (buf), "%ss0", path); | |
927 | if ((fd = open(buf, O_RDONLY)) < 0) { | |
928 | (void) fprintf(stderr, | |
929 | gettext("cannot open '%s': %s\n"), | |
930 | buf, strerror(errno)); | |
931 | return (-1); | |
932 | } | |
933 | ||
934 | if (devid_get(fd, &devid) == 0) { | |
935 | if (devid_get_minor_name(fd, &minor) == 0 && | |
936 | (devid_str = devid_str_encode(devid, minor)) != | |
937 | NULL) { | |
938 | verify(nvlist_add_string(nv, | |
939 | ZPOOL_CONFIG_DEVID, devid_str) == 0); | |
940 | } | |
941 | if (devid_str != NULL) | |
942 | devid_str_free(devid_str); | |
943 | if (minor != NULL) | |
944 | devid_str_free(minor); | |
945 | devid_free(devid); | |
946 | } | |
947 | ||
948 | /* | |
949 | * Update the path to refer to the 's0' slice. The presence of | |
950 | * the 'whole_disk' field indicates to the CLI that we should | |
951 | * chop off the slice number when displaying the device in | |
952 | * future output. | |
953 | */ | |
954 | verify(nvlist_add_string(nv, ZPOOL_CONFIG_PATH, buf) == 0); | |
955 | ||
956 | (void) close(fd); | |
957 | ||
958 | return (0); | |
959 | } | |
960 | ||
961 | for (c = 0; c < children; c++) | |
962 | if ((ret = make_disks(zhp, child[c])) != 0) | |
963 | return (ret); | |
964 | ||
965 | if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES, | |
966 | &child, &children) == 0) | |
967 | for (c = 0; c < children; c++) | |
968 | if ((ret = make_disks(zhp, child[c])) != 0) | |
969 | return (ret); | |
970 | ||
971 | if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE, | |
972 | &child, &children) == 0) | |
973 | for (c = 0; c < children; c++) | |
974 | if ((ret = make_disks(zhp, child[c])) != 0) | |
975 | return (ret); | |
976 | ||
977 | return (0); | |
978 | } | |
979 | ||
980 | /* | |
981 | * Determine if the given path is a hot spare within the given configuration. | |
982 | */ | |
983 | static boolean_t | |
984 | is_spare(nvlist_t *config, const char *path) | |
985 | { | |
986 | int fd; | |
987 | pool_state_t state; | |
988 | char *name = NULL; | |
989 | nvlist_t *label; | |
990 | uint64_t guid, spareguid; | |
991 | nvlist_t *nvroot; | |
992 | nvlist_t **spares; | |
993 | uint_t i, nspares; | |
994 | boolean_t inuse; | |
995 | ||
996 | if ((fd = open(path, O_RDONLY)) < 0) | |
997 | return (B_FALSE); | |
998 | ||
999 | if (zpool_in_use(g_zfs, fd, &state, &name, &inuse) != 0 || | |
1000 | !inuse || | |
1001 | state != POOL_STATE_SPARE || | |
1002 | zpool_read_label(fd, &label) != 0) { | |
1003 | free(name); | |
1004 | (void) close(fd); | |
1005 | return (B_FALSE); | |
1006 | } | |
1007 | free(name); | |
1008 | ||
1009 | (void) close(fd); | |
1010 | verify(nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID, &guid) == 0); | |
1011 | nvlist_free(label); | |
1012 | ||
1013 | verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, | |
1014 | &nvroot) == 0); | |
1015 | if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, | |
1016 | &spares, &nspares) == 0) { | |
1017 | for (i = 0; i < nspares; i++) { | |
1018 | verify(nvlist_lookup_uint64(spares[i], | |
1019 | ZPOOL_CONFIG_GUID, &spareguid) == 0); | |
1020 | if (spareguid == guid) | |
1021 | return (B_TRUE); | |
1022 | } | |
1023 | } | |
1024 | ||
1025 | return (B_FALSE); | |
1026 | } | |
1027 | ||
1028 | /* | |
1029 | * Go through and find any devices that are in use. We rely on libdiskmgt for | |
1030 | * the majority of this task. | |
1031 | */ | |
1032 | static int | |
1033 | check_in_use(nvlist_t *config, nvlist_t *nv, int force, int isreplacing, | |
1034 | int isspare) | |
1035 | { | |
1036 | nvlist_t **child; | |
1037 | uint_t c, children; | |
1038 | char *type, *path; | |
1039 | int ret; | |
1040 | char buf[MAXPATHLEN]; | |
1041 | uint64_t wholedisk; | |
1042 | ||
1043 | verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0); | |
1044 | ||
1045 | if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN, | |
1046 | &child, &children) != 0) { | |
1047 | ||
1048 | verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0); | |
1049 | ||
1050 | /* | |
1051 | * As a generic check, we look to see if this is a replace of a | |
1052 | * hot spare within the same pool. If so, we allow it | |
1053 | * regardless of what libdiskmgt or zpool_in_use() says. | |
1054 | */ | |
1055 | if (isreplacing) { | |
1056 | if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK, | |
1057 | &wholedisk) == 0 && wholedisk) | |
1058 | (void) snprintf(buf, sizeof (buf), "%ss0", | |
1059 | path); | |
1060 | else | |
1061 | (void) strlcpy(buf, path, sizeof (buf)); | |
1062 | if (is_spare(config, buf)) | |
1063 | return (0); | |
1064 | } | |
1065 | ||
1066 | if (strcmp(type, VDEV_TYPE_DISK) == 0) | |
1067 | ret = check_device(path, force, isspare); | |
1068 | ||
1069 | if (strcmp(type, VDEV_TYPE_FILE) == 0) | |
1070 | ret = check_file(path, force, isspare); | |
1071 | ||
1072 | return (ret); | |
1073 | } | |
1074 | ||
1075 | for (c = 0; c < children; c++) | |
1076 | if ((ret = check_in_use(config, child[c], force, | |
1077 | isreplacing, B_FALSE)) != 0) | |
1078 | return (ret); | |
1079 | ||
1080 | if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES, | |
1081 | &child, &children) == 0) | |
1082 | for (c = 0; c < children; c++) | |
1083 | if ((ret = check_in_use(config, child[c], force, | |
1084 | isreplacing, B_TRUE)) != 0) | |
1085 | return (ret); | |
1086 | ||
1087 | if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE, | |
1088 | &child, &children) == 0) | |
1089 | for (c = 0; c < children; c++) | |
1090 | if ((ret = check_in_use(config, child[c], force, | |
1091 | isreplacing, B_FALSE)) != 0) | |
1092 | return (ret); | |
1093 | ||
1094 | return (0); | |
1095 | } | |
1096 | ||
1097 | static const char * | |
1098 | is_grouping(const char *type, int *mindev) | |
1099 | { | |
1100 | if (strcmp(type, "raidz") == 0 || strcmp(type, "raidz1") == 0) { | |
1101 | if (mindev != NULL) | |
1102 | *mindev = 2; | |
1103 | return (VDEV_TYPE_RAIDZ); | |
1104 | } | |
1105 | ||
1106 | if (strcmp(type, "raidz2") == 0) { | |
1107 | if (mindev != NULL) | |
1108 | *mindev = 3; | |
1109 | return (VDEV_TYPE_RAIDZ); | |
1110 | } | |
1111 | ||
1112 | if (strcmp(type, "mirror") == 0) { | |
1113 | if (mindev != NULL) | |
1114 | *mindev = 2; | |
1115 | return (VDEV_TYPE_MIRROR); | |
1116 | } | |
1117 | ||
1118 | if (strcmp(type, "spare") == 0) { | |
1119 | if (mindev != NULL) | |
1120 | *mindev = 1; | |
1121 | return (VDEV_TYPE_SPARE); | |
1122 | } | |
1123 | ||
1124 | if (strcmp(type, "log") == 0) { | |
1125 | if (mindev != NULL) | |
1126 | *mindev = 1; | |
1127 | return (VDEV_TYPE_LOG); | |
1128 | } | |
1129 | ||
1130 | if (strcmp(type, "cache") == 0) { | |
1131 | if (mindev != NULL) | |
1132 | *mindev = 1; | |
1133 | return (VDEV_TYPE_L2CACHE); | |
1134 | } | |
1135 | ||
1136 | return (NULL); | |
1137 | } | |
1138 | ||
1139 | /* | |
1140 | * Construct a syntactically valid vdev specification, | |
1141 | * and ensure that all devices and files exist and can be opened. | |
1142 | * Note: we don't bother freeing anything in the error paths | |
1143 | * because the program is just going to exit anyway. | |
1144 | */ | |
1145 | nvlist_t * | |
1146 | construct_spec(int argc, char **argv) | |
1147 | { | |
1148 | nvlist_t *nvroot, *nv, **top, **spares, **l2cache; | |
1149 | int t, toplevels, mindev, nspares, nlogs, nl2cache; | |
1150 | const char *type; | |
1151 | uint64_t is_log; | |
1152 | boolean_t seen_logs; | |
1153 | ||
1154 | top = NULL; | |
1155 | toplevels = 0; | |
1156 | spares = NULL; | |
1157 | l2cache = NULL; | |
1158 | nspares = 0; | |
1159 | nlogs = 0; | |
1160 | nl2cache = 0; | |
1161 | is_log = B_FALSE; | |
1162 | seen_logs = B_FALSE; | |
1163 | ||
1164 | while (argc > 0) { | |
1165 | nv = NULL; | |
1166 | ||
1167 | /* | |
1168 | * If it's a mirror or raidz, the subsequent arguments are | |
1169 | * its leaves -- until we encounter the next mirror or raidz. | |
1170 | */ | |
1171 | if ((type = is_grouping(argv[0], &mindev)) != NULL) { | |
1172 | nvlist_t **child = NULL; | |
1173 | int c, children = 0; | |
1174 | ||
1175 | if (strcmp(type, VDEV_TYPE_SPARE) == 0) { | |
1176 | if (spares != NULL) { | |
1177 | (void) fprintf(stderr, | |
1178 | gettext("invalid vdev " | |
1179 | "specification: 'spare' can be " | |
1180 | "specified only once\n")); | |
1181 | return (NULL); | |
1182 | } | |
1183 | is_log = B_FALSE; | |
1184 | } | |
1185 | ||
1186 | if (strcmp(type, VDEV_TYPE_LOG) == 0) { | |
1187 | if (seen_logs) { | |
1188 | (void) fprintf(stderr, | |
1189 | gettext("invalid vdev " | |
1190 | "specification: 'log' can be " | |
1191 | "specified only once\n")); | |
1192 | return (NULL); | |
1193 | } | |
1194 | seen_logs = B_TRUE; | |
1195 | is_log = B_TRUE; | |
1196 | argc--; | |
1197 | argv++; | |
1198 | /* | |
1199 | * A log is not a real grouping device. | |
1200 | * We just set is_log and continue. | |
1201 | */ | |
1202 | continue; | |
1203 | } | |
1204 | ||
1205 | if (strcmp(type, VDEV_TYPE_L2CACHE) == 0) { | |
1206 | if (l2cache != NULL) { | |
1207 | (void) fprintf(stderr, | |
1208 | gettext("invalid vdev " | |
1209 | "specification: 'cache' can be " | |
1210 | "specified only once\n")); | |
1211 | return (NULL); | |
1212 | } | |
1213 | is_log = B_FALSE; | |
1214 | } | |
1215 | ||
1216 | if (is_log) { | |
1217 | if (strcmp(type, VDEV_TYPE_MIRROR) != 0) { | |
1218 | (void) fprintf(stderr, | |
1219 | gettext("invalid vdev " | |
1220 | "specification: unsupported 'log' " | |
1221 | "device: %s\n"), type); | |
1222 | return (NULL); | |
1223 | } | |
1224 | nlogs++; | |
1225 | } | |
1226 | ||
1227 | for (c = 1; c < argc; c++) { | |
1228 | if (is_grouping(argv[c], NULL) != NULL) | |
1229 | break; | |
1230 | children++; | |
1231 | child = realloc(child, | |
1232 | children * sizeof (nvlist_t *)); | |
1233 | if (child == NULL) | |
1234 | zpool_no_memory(); | |
1235 | if ((nv = make_leaf_vdev(argv[c], B_FALSE)) | |
1236 | == NULL) | |
1237 | return (NULL); | |
1238 | child[children - 1] = nv; | |
1239 | } | |
1240 | ||
1241 | if (children < mindev) { | |
1242 | (void) fprintf(stderr, gettext("invalid vdev " | |
1243 | "specification: %s requires at least %d " | |
1244 | "devices\n"), argv[0], mindev); | |
1245 | return (NULL); | |
1246 | } | |
1247 | ||
1248 | argc -= c; | |
1249 | argv += c; | |
1250 | ||
1251 | if (strcmp(type, VDEV_TYPE_SPARE) == 0) { | |
1252 | spares = child; | |
1253 | nspares = children; | |
1254 | continue; | |
1255 | } else if (strcmp(type, VDEV_TYPE_L2CACHE) == 0) { | |
1256 | l2cache = child; | |
1257 | nl2cache = children; | |
1258 | continue; | |
1259 | } else { | |
1260 | verify(nvlist_alloc(&nv, NV_UNIQUE_NAME, | |
1261 | 0) == 0); | |
1262 | verify(nvlist_add_string(nv, ZPOOL_CONFIG_TYPE, | |
1263 | type) == 0); | |
1264 | verify(nvlist_add_uint64(nv, | |
1265 | ZPOOL_CONFIG_IS_LOG, is_log) == 0); | |
1266 | if (strcmp(type, VDEV_TYPE_RAIDZ) == 0) { | |
1267 | verify(nvlist_add_uint64(nv, | |
1268 | ZPOOL_CONFIG_NPARITY, | |
1269 | mindev - 1) == 0); | |
1270 | } | |
1271 | verify(nvlist_add_nvlist_array(nv, | |
1272 | ZPOOL_CONFIG_CHILDREN, child, | |
1273 | children) == 0); | |
1274 | ||
1275 | for (c = 0; c < children; c++) | |
1276 | nvlist_free(child[c]); | |
1277 | free(child); | |
1278 | } | |
1279 | } else { | |
1280 | /* | |
1281 | * We have a device. Pass off to make_leaf_vdev() to | |
1282 | * construct the appropriate nvlist describing the vdev. | |
1283 | */ | |
1284 | if ((nv = make_leaf_vdev(argv[0], is_log)) == NULL) | |
1285 | return (NULL); | |
1286 | if (is_log) | |
1287 | nlogs++; | |
1288 | argc--; | |
1289 | argv++; | |
1290 | } | |
1291 | ||
1292 | toplevels++; | |
1293 | top = realloc(top, toplevels * sizeof (nvlist_t *)); | |
1294 | if (top == NULL) | |
1295 | zpool_no_memory(); | |
1296 | top[toplevels - 1] = nv; | |
1297 | } | |
1298 | ||
1299 | if (toplevels == 0 && nspares == 0 && nl2cache == 0) { | |
1300 | (void) fprintf(stderr, gettext("invalid vdev " | |
1301 | "specification: at least one toplevel vdev must be " | |
1302 | "specified\n")); | |
1303 | return (NULL); | |
1304 | } | |
1305 | ||
1306 | if (seen_logs && nlogs == 0) { | |
1307 | (void) fprintf(stderr, gettext("invalid vdev specification: " | |
1308 | "log requires at least 1 device\n")); | |
1309 | return (NULL); | |
1310 | } | |
1311 | ||
1312 | /* | |
1313 | * Finally, create nvroot and add all top-level vdevs to it. | |
1314 | */ | |
1315 | verify(nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) == 0); | |
1316 | verify(nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE, | |
1317 | VDEV_TYPE_ROOT) == 0); | |
1318 | verify(nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN, | |
1319 | top, toplevels) == 0); | |
1320 | if (nspares != 0) | |
1321 | verify(nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES, | |
1322 | spares, nspares) == 0); | |
1323 | if (nl2cache != 0) | |
1324 | verify(nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE, | |
1325 | l2cache, nl2cache) == 0); | |
1326 | ||
1327 | for (t = 0; t < toplevels; t++) | |
1328 | nvlist_free(top[t]); | |
1329 | for (t = 0; t < nspares; t++) | |
1330 | nvlist_free(spares[t]); | |
1331 | for (t = 0; t < nl2cache; t++) | |
1332 | nvlist_free(l2cache[t]); | |
1333 | if (spares) | |
1334 | free(spares); | |
1335 | if (l2cache) | |
1336 | free(l2cache); | |
1337 | free(top); | |
1338 | ||
1339 | return (nvroot); | |
1340 | } | |
1341 | ||
1342 | ||
1343 | /* | |
1344 | * Get and validate the contents of the given vdev specification. This ensures | |
1345 | * that the nvlist returned is well-formed, that all the devices exist, and that | |
1346 | * they are not currently in use by any other known consumer. The 'poolconfig' | |
1347 | * parameter is the current configuration of the pool when adding devices | |
1348 | * existing pool, and is used to perform additional checks, such as changing the | |
1349 | * replication level of the pool. It can be 'NULL' to indicate that this is a | |
1350 | * new pool. The 'force' flag controls whether devices should be forcefully | |
1351 | * added, even if they appear in use. | |
1352 | */ | |
1353 | nvlist_t * | |
1354 | make_root_vdev(zpool_handle_t *zhp, int force, int check_rep, | |
1355 | boolean_t isreplacing, int argc, char **argv) | |
1356 | { | |
1357 | nvlist_t *newroot; | |
1358 | nvlist_t *poolconfig = NULL; | |
1359 | is_force = force; | |
1360 | ||
1361 | /* | |
1362 | * Construct the vdev specification. If this is successful, we know | |
1363 | * that we have a valid specification, and that all devices can be | |
1364 | * opened. | |
1365 | */ | |
1366 | if ((newroot = construct_spec(argc, argv)) == NULL) | |
1367 | return (NULL); | |
1368 | ||
1369 | if (zhp && ((poolconfig = zpool_get_config(zhp, NULL)) == NULL)) | |
1370 | return (NULL); | |
1371 | ||
1372 | /* | |
1373 | * Validate each device to make sure that its not shared with another | |
1374 | * subsystem. We do this even if 'force' is set, because there are some | |
1375 | * uses (such as a dedicated dump device) that even '-f' cannot | |
1376 | * override. | |
1377 | */ | |
1378 | if (check_in_use(poolconfig, newroot, force, isreplacing, | |
1379 | B_FALSE) != 0) { | |
1380 | nvlist_free(newroot); | |
1381 | return (NULL); | |
1382 | } | |
1383 | ||
1384 | /* | |
1385 | * Check the replication level of the given vdevs and report any errors | |
1386 | * found. We include the existing pool spec, if any, as we need to | |
1387 | * catch changes against the existing replication level. | |
1388 | */ | |
1389 | if (check_rep && check_replication(poolconfig, newroot) != 0) { | |
1390 | nvlist_free(newroot); | |
1391 | return (NULL); | |
1392 | } | |
1393 | ||
1394 | /* | |
1395 | * Run through the vdev specification and label any whole disks found. | |
1396 | */ | |
1397 | if (make_disks(zhp, newroot) != 0) { | |
1398 | nvlist_free(newroot); | |
1399 | return (NULL); | |
1400 | } | |
1401 | ||
1402 | return (newroot); | |
1403 | } |