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.
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.
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]
23 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Copyright (c) 2013 by Delphix. All rights reserved.
25 * Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>.
29 * Functions to convert between a list of vdevs and an nvlist representing the
30 * configuration. Each entry in the list can be one of:
33 * disk=(path=..., devid=...)
42 * While the underlying implementation supports it, group vdevs cannot contain
43 * other group vdevs. All userland verification of devices is contained within
44 * this file. If successful, the nvlist returned can be passed directly to the
45 * kernel; we've done as much verification as possible in userland.
47 * Hot spares are a special case, and passed down as an array of disk vdevs, at
48 * the same level as the root of the vdev tree.
50 * The only function exported by this file is 'make_root_vdev'. The
51 * function performs several passes:
53 * 1. Construct the vdev specification. Performs syntax validation and
54 * makes sure each device is valid.
55 * 2. Check for devices in use. Using libdiskmgt, makes sure that no
56 * devices are also in use. Some can be overridden using the 'force'
57 * flag, others cannot.
58 * 3. Check for replication errors if the 'force' flag is not specified.
59 * validates that the replication level is consistent across the
61 * 4. Call libzfs to label any whole disks with an EFI label.
69 #include <libnvpair.h>
77 #include <sys/mntent.h>
80 #include "zpool_util.h"
82 #define DISK_ROOT "/dev/dsk"
83 #define RDISK_ROOT "/dev/rdsk"
84 #define BACKUP_SLICE "s2"
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.
96 vdev_error(const char *fmt, ...)
101 (void) fprintf(stderr, gettext("invalid vdev specification\n"));
103 (void) fprintf(stderr, gettext("use '-f' to override "
104 "the following errors:\n"));
106 (void) fprintf(stderr, gettext("the following errors "
107 "must be manually repaired:\n"));
112 (void) vfprintf(stderr, fmt, ap);
118 libdiskmgt_error(int error)
121 * ENXIO/ENODEV is a valid error message if the device doesn't live in
122 * /dev/dsk. Don't bother printing an error message in this case.
124 if (error == ENXIO || error == ENODEV)
127 (void) fprintf(stderr, gettext("warning: device in use checking "
128 "failed: %s\n"), strerror(error));
132 * Validate a device, passing the bulk of the work off to libdiskmgt.
135 check_slice(const char *path, int force, boolean_t wholedisk, boolean_t isspare)
142 who = DM_WHO_ZPOOL_FORCE;
144 who = DM_WHO_ZPOOL_SPARE;
148 if (dm_inuse((char *)path, &msg, who, &error) || error) {
150 libdiskmgt_error(error);
153 vdev_error("%s", msg);
160 * If we're given a whole disk, ignore overlapping slices since we're
161 * about to label it anyway.
164 if (!wholedisk && !force &&
165 (dm_isoverlapping((char *)path, &msg, &error) || error)) {
167 /* dm_isoverlapping returned -1 */
168 vdev_error(gettext("%s overlaps with %s\n"), path, msg);
171 } else if (error != ENODEV) {
172 /* libdiskmgt's devcache only handles physical drives */
173 libdiskmgt_error(error);
183 * Validate a whole disk. Iterate over all slices on the disk and make sure
184 * that none is in use by calling check_slice().
187 check_disk(const char *name, dm_descriptor_t disk, int force, int isspare)
189 dm_descriptor_t *drive, *media, *slice;
195 * Get the drive associated with this disk. This should never fail,
196 * because we already have an alias handle open for the device.
198 if ((drive = dm_get_associated_descriptors(disk, DM_DRIVE,
199 &err)) == NULL || *drive == NULL) {
201 libdiskmgt_error(err);
205 if ((media = dm_get_associated_descriptors(*drive, DM_MEDIA,
207 dm_free_descriptors(drive);
209 libdiskmgt_error(err);
213 dm_free_descriptors(drive);
216 * It is possible that the user has specified a removable media drive,
217 * and the media is not present.
219 if (*media == NULL) {
220 dm_free_descriptors(media);
221 vdev_error(gettext("'%s' has no media in drive\n"), name);
225 if ((slice = dm_get_associated_descriptors(*media, DM_SLICE,
227 dm_free_descriptors(media);
229 libdiskmgt_error(err);
233 dm_free_descriptors(media);
238 * Iterate over all slices and report any errors. We don't care about
239 * overlapping slices because we are using the whole disk.
241 for (i = 0; slice[i] != NULL; i++) {
242 char *name = dm_get_name(slice[i], &err);
244 if (check_slice(name, force, B_TRUE, isspare) != 0)
250 dm_free_descriptors(slice);
258 check_device(const char *path, boolean_t force, boolean_t isspare)
260 dm_descriptor_t desc;
265 * For whole disks, libdiskmgt does not include the leading dev path.
267 dev = strrchr(path, '/');
270 if ((desc = dm_get_descriptor_by_name(DM_ALIAS, dev, &err)) != NULL) {
271 err = check_disk(path, desc, force, isspare);
272 dm_free_descriptor(desc);
276 return (check_slice(path, force, B_FALSE, isspare));
281 * Check that a file is valid. All we can do in this case is check that it's
282 * not in use by another pool, and not in use by swap.
285 check_file(const char *file, boolean_t force, boolean_t isspare)
295 if (dm_inuse_swap(file, &err)) {
297 libdiskmgt_error(err);
299 vdev_error(gettext("%s is currently used by swap. "
300 "Please see swap(1M).\n"), file);
305 if ((fd = open(file, O_RDONLY)) < 0)
308 if (zpool_in_use(g_zfs, fd, &state, &name, &inuse) == 0 && inuse) {
312 case POOL_STATE_ACTIVE:
313 desc = gettext("active");
316 case POOL_STATE_EXPORTED:
317 desc = gettext("exported");
320 case POOL_STATE_POTENTIALLY_ACTIVE:
321 desc = gettext("potentially active");
325 desc = gettext("unknown");
330 * Allow hot spares to be shared between pools.
332 if (state == POOL_STATE_SPARE && isspare)
335 if (state == POOL_STATE_ACTIVE ||
336 state == POOL_STATE_SPARE || !force) {
338 case POOL_STATE_SPARE:
339 vdev_error(gettext("%s is reserved as a hot "
340 "spare for pool %s\n"), file, name);
343 vdev_error(gettext("%s is part of %s pool "
344 "'%s'\n"), file, desc, name);
358 check_device(const char *name, boolean_t force, boolean_t isspare)
360 char path[MAXPATHLEN];
362 if (strncmp(name, _PATH_DEV, sizeof(_PATH_DEV) - 1) != 0)
363 snprintf(path, sizeof(path), "%s%s", _PATH_DEV, name);
365 strlcpy(path, name, sizeof(path));
367 return (check_file(path, force, isspare));
371 * By "whole disk" we mean an entire physical disk (something we can
372 * label, toggle the write cache on, etc.) as opposed to the full
373 * capacity of a pseudo-device such as lofi or did. We act as if we
374 * are labeling the disk, which should be a pretty good test of whether
375 * it's a viable device or not. Returns B_TRUE if it is and B_FALSE if
379 is_whole_disk(const char *arg)
382 struct dk_gpt *label;
384 char path[MAXPATHLEN];
386 (void) snprintf(path, sizeof (path), "%s%s%s",
387 RDISK_ROOT, strrchr(arg, '/'), BACKUP_SLICE);
388 if ((fd = open(path, O_RDWR | O_NDELAY)) < 0)
390 if (efi_alloc_and_init(fd, EFI_NUMPAR, &label) != 0) {
410 * Create a leaf vdev. Determine if this is a file or a device. If it's a
411 * device, fill in the device id to make a complete nvlist. Valid forms for a
414 * /dev/dsk/xxx Complete disk path
415 * /xxx Full path to file
416 * xxx Shorthand for /dev/dsk/xxx
419 make_leaf_vdev(const char *arg, uint64_t is_log)
421 char path[MAXPATHLEN];
422 struct stat64 statbuf;
423 nvlist_t *vdev = NULL;
425 boolean_t wholedisk = B_FALSE;
428 * Determine what type of vdev this is, and put the full path into
429 * 'path'. We detect whether this is a device of file afterwards by
430 * checking the st_mode of the file.
434 * Complete device or file path. Exact type is determined by
435 * examining the file descriptor afterwards.
437 wholedisk = is_whole_disk(arg);
438 if (!wholedisk && (stat64(arg, &statbuf) != 0)) {
439 (void) fprintf(stderr,
440 gettext("cannot open '%s': %s\n"),
441 arg, strerror(errno));
445 (void) strlcpy(path, arg, sizeof (path));
448 * This may be a short path for a device, or it could be total
449 * gibberish. Check to see if it's a known device in
450 * /dev/dsk/. As part of this check, see if we've been given a
451 * an entire disk (minus the slice number).
453 if (strncmp(arg, _PATH_DEV, sizeof(_PATH_DEV) - 1) == 0)
454 strlcpy(path, arg, sizeof (path));
456 snprintf(path, sizeof (path), "%s%s", _PATH_DEV, arg);
457 wholedisk = is_whole_disk(path);
458 if (!wholedisk && (stat64(path, &statbuf) != 0)) {
460 * If we got ENOENT, then the user gave us
461 * gibberish, so try to direct them with a
462 * reasonable error message. Otherwise,
463 * regurgitate strerror() since it's the best we
466 if (errno == ENOENT) {
467 (void) fprintf(stderr,
468 gettext("cannot open '%s': no such "
469 "GEOM provider\n"), arg);
470 (void) fprintf(stderr,
471 gettext("must be a full path or "
472 "shorthand device name\n"));
475 (void) fprintf(stderr,
476 gettext("cannot open '%s': %s\n"),
477 path, strerror(errno));
484 if (S_ISCHR(statbuf.st_mode)) {
485 statbuf.st_mode &= ~S_IFCHR;
486 statbuf.st_mode |= S_IFBLK;
492 * Determine whether this is a device or a file.
494 if (wholedisk || S_ISBLK(statbuf.st_mode)) {
495 type = VDEV_TYPE_DISK;
496 } else if (S_ISREG(statbuf.st_mode)) {
497 type = VDEV_TYPE_FILE;
499 (void) fprintf(stderr, gettext("cannot use '%s': must be a "
500 "GEOM provider or regular file\n"), path);
505 * Finally, we have the complete device or file, and we know that it is
506 * acceptable to use. Construct the nvlist to describe this vdev. All
507 * vdevs have a 'path' element, and devices also have a 'devid' element.
509 verify(nvlist_alloc(&vdev, NV_UNIQUE_NAME, 0) == 0);
510 verify(nvlist_add_string(vdev, ZPOOL_CONFIG_PATH, path) == 0);
511 verify(nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE, type) == 0);
512 verify(nvlist_add_uint64(vdev, ZPOOL_CONFIG_IS_LOG, is_log) == 0);
513 if (strcmp(type, VDEV_TYPE_DISK) == 0)
514 verify(nvlist_add_uint64(vdev, ZPOOL_CONFIG_WHOLE_DISK,
515 (uint64_t)wholedisk) == 0);
519 * For a whole disk, defer getting its devid until after labeling it.
521 if (S_ISBLK(statbuf.st_mode) && !wholedisk) {
523 * Get the devid for the device.
527 char *minor = NULL, *devid_str = NULL;
529 if ((fd = open(path, O_RDONLY)) < 0) {
530 (void) fprintf(stderr, gettext("cannot open '%s': "
531 "%s\n"), path, strerror(errno));
536 if (devid_get(fd, &devid) == 0) {
537 if (devid_get_minor_name(fd, &minor) == 0 &&
538 (devid_str = devid_str_encode(devid, minor)) !=
540 verify(nvlist_add_string(vdev,
541 ZPOOL_CONFIG_DEVID, devid_str) == 0);
543 if (devid_str != NULL)
544 devid_str_free(devid_str);
546 devid_str_free(minor);
558 * Go through and verify the replication level of the pool is consistent.
559 * Performs the following checks:
561 * For the new spec, verifies that devices in mirrors and raidz are the
564 * If the current configuration already has inconsistent replication
565 * levels, ignore any other potential problems in the new spec.
567 * Otherwise, make sure that the current spec (if there is one) and the new
568 * spec have consistent replication levels.
570 typedef struct replication_level {
572 uint64_t zprl_children;
573 uint64_t zprl_parity;
574 } replication_level_t;
576 #define ZPOOL_FUZZ (16 * 1024 * 1024)
579 * Given a list of toplevel vdevs, return the current replication level. If
580 * the config is inconsistent, then NULL is returned. If 'fatal' is set, then
581 * an error message will be displayed for each self-inconsistent vdev.
583 static replication_level_t *
584 get_replication(nvlist_t *nvroot, boolean_t fatal)
592 replication_level_t lastrep = {0};
593 replication_level_t rep;
594 replication_level_t *ret;
595 boolean_t dontreport;
597 ret = safe_malloc(sizeof (replication_level_t));
599 verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
600 &top, &toplevels) == 0);
602 lastrep.zprl_type = NULL;
603 for (t = 0; t < toplevels; t++) {
604 uint64_t is_log = B_FALSE;
609 * For separate logs we ignore the top level vdev replication
612 (void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_IS_LOG, &is_log);
616 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE,
618 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
619 &child, &children) != 0) {
621 * This is a 'file' or 'disk' vdev.
623 rep.zprl_type = type;
624 rep.zprl_children = 1;
630 * This is a mirror or RAID-Z vdev. Go through and make
631 * sure the contents are all the same (files vs. disks),
632 * keeping track of the number of elements in the
635 * We also check that the size of each vdev (if it can
636 * be determined) is the same.
638 rep.zprl_type = type;
639 rep.zprl_children = 0;
641 if (strcmp(type, VDEV_TYPE_RAIDZ) == 0) {
642 verify(nvlist_lookup_uint64(nv,
643 ZPOOL_CONFIG_NPARITY,
644 &rep.zprl_parity) == 0);
645 assert(rep.zprl_parity != 0);
651 * The 'dontreport' variable indicates that we've
652 * already reported an error for this spec, so don't
653 * bother doing it again.
658 for (c = 0; c < children; c++) {
659 nvlist_t *cnv = child[c];
661 struct stat64 statbuf;
662 uint64_t size = -1ULL;
668 verify(nvlist_lookup_string(cnv,
669 ZPOOL_CONFIG_TYPE, &childtype) == 0);
672 * If this is a replacing or spare vdev, then
673 * get the real first child of the vdev.
675 if (strcmp(childtype,
676 VDEV_TYPE_REPLACING) == 0 ||
677 strcmp(childtype, VDEV_TYPE_SPARE) == 0) {
681 verify(nvlist_lookup_nvlist_array(cnv,
682 ZPOOL_CONFIG_CHILDREN, &rchild,
684 assert(rchildren == 2);
687 verify(nvlist_lookup_string(cnv,
692 verify(nvlist_lookup_string(cnv,
693 ZPOOL_CONFIG_PATH, &path) == 0);
696 * If we have a raidz/mirror that combines disks
697 * with files, report it as an error.
699 if (!dontreport && type != NULL &&
700 strcmp(type, childtype) != 0) {
706 "mismatched replication "
707 "level: %s contains both "
708 "files and devices\n"),
716 * According to stat(2), the value of 'st_size'
717 * is undefined for block devices and character
718 * devices. But there is no effective way to
719 * determine the real size in userland.
721 * Instead, we'll take advantage of an
722 * implementation detail of spec_size(). If the
723 * device is currently open, then we (should)
724 * return a valid size.
726 * If we still don't get a valid size (indicated
727 * by a size of 0 or MAXOFFSET_T), then ignore
728 * this device altogether.
730 if ((fd = open(path, O_RDONLY)) >= 0) {
731 err = fstat64(fd, &statbuf);
734 err = stat64(path, &statbuf);
738 statbuf.st_size == 0 ||
739 statbuf.st_size == MAXOFFSET_T)
742 size = statbuf.st_size;
745 * Also make sure that devices and
746 * slices have a consistent size. If
747 * they differ by a significant amount
748 * (~16MB) then report an error.
751 (vdev_size != -1ULL &&
752 (labs(size - vdev_size) >
759 "%s contains devices of "
760 "different sizes\n"),
773 * At this point, we have the replication of the last toplevel
774 * vdev in 'rep'. Compare it to 'lastrep' to see if its
777 if (lastrep.zprl_type != NULL) {
778 if (strcmp(lastrep.zprl_type, rep.zprl_type) != 0) {
784 "mismatched replication level: "
785 "both %s and %s vdevs are "
787 lastrep.zprl_type, rep.zprl_type);
790 } else if (lastrep.zprl_parity != rep.zprl_parity) {
796 "mismatched replication level: "
797 "both %llu and %llu device parity "
798 "%s vdevs are present\n"),
804 } else if (lastrep.zprl_children != rep.zprl_children) {
810 "mismatched replication level: "
811 "both %llu-way and %llu-way %s "
812 "vdevs are present\n"),
813 lastrep.zprl_children,
830 * Check the replication level of the vdev spec against the current pool. Calls
831 * get_replication() to make sure the new spec is self-consistent. If the pool
832 * has a consistent replication level, then we ignore any errors. Otherwise,
833 * report any difference between the two.
836 check_replication(nvlist_t *config, nvlist_t *newroot)
840 replication_level_t *current = NULL, *new;
844 * If we have a current pool configuration, check to see if it's
845 * self-consistent. If not, simply return success.
847 if (config != NULL) {
850 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
852 if ((current = get_replication(nvroot, B_FALSE)) == NULL)
856 * for spares there may be no children, and therefore no
857 * replication level to check
859 if ((nvlist_lookup_nvlist_array(newroot, ZPOOL_CONFIG_CHILDREN,
860 &child, &children) != 0) || (children == 0)) {
866 * If all we have is logs then there's no replication level to check.
868 if (num_logs(newroot) == children) {
874 * Get the replication level of the new vdev spec, reporting any
875 * inconsistencies found.
877 if ((new = get_replication(newroot, B_TRUE)) == NULL) {
883 * Check to see if the new vdev spec matches the replication level of
887 if (current != NULL) {
888 if (strcmp(current->zprl_type, new->zprl_type) != 0) {
890 "mismatched replication level: pool uses %s "
891 "and new vdev is %s\n"),
892 current->zprl_type, new->zprl_type);
894 } else if (current->zprl_parity != new->zprl_parity) {
896 "mismatched replication level: pool uses %llu "
897 "device parity and new vdev uses %llu\n"),
898 current->zprl_parity, new->zprl_parity);
900 } else if (current->zprl_children != new->zprl_children) {
902 "mismatched replication level: pool uses %llu-way "
903 "%s and new vdev uses %llu-way %s\n"),
904 current->zprl_children, current->zprl_type,
905 new->zprl_children, new->zprl_type);
919 * Go through and find any whole disks in the vdev specification, labelling them
920 * as appropriate. When constructing the vdev spec, we were unable to open this
921 * device in order to provide a devid. Now that we have labelled the disk and
922 * know that slice 0 is valid, we can construct the devid now.
924 * If the disk was already labeled with an EFI label, we will have gotten the
925 * devid already (because we were able to open the whole disk). Otherwise, we
926 * need to get the devid after we label the disk.
929 make_disks(zpool_handle_t *zhp, nvlist_t *nv)
933 char *type, *path, *diskname;
934 char buf[MAXPATHLEN];
939 char *minor = NULL, *devid_str = NULL;
941 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0);
943 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
944 &child, &children) != 0) {
946 if (strcmp(type, VDEV_TYPE_DISK) != 0)
950 * We have a disk device. Get the path to the device
951 * and see if it's a whole disk by appending the backup
952 * slice and stat()ing the device.
954 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0);
955 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
956 &wholedisk) != 0 || !wholedisk)
959 diskname = strrchr(path, '/');
960 assert(diskname != NULL);
962 if (zpool_label_disk(g_zfs, zhp, diskname) == -1)
966 * Fill in the devid, now that we've labeled the disk.
968 (void) snprintf(buf, sizeof (buf), "%ss0", path);
969 if ((fd = open(buf, O_RDONLY)) < 0) {
970 (void) fprintf(stderr,
971 gettext("cannot open '%s': %s\n"),
972 buf, strerror(errno));
976 if (devid_get(fd, &devid) == 0) {
977 if (devid_get_minor_name(fd, &minor) == 0 &&
978 (devid_str = devid_str_encode(devid, minor)) !=
980 verify(nvlist_add_string(nv,
981 ZPOOL_CONFIG_DEVID, devid_str) == 0);
983 if (devid_str != NULL)
984 devid_str_free(devid_str);
986 devid_str_free(minor);
991 * Update the path to refer to the 's0' slice. The presence of
992 * the 'whole_disk' field indicates to the CLI that we should
993 * chop off the slice number when displaying the device in
996 verify(nvlist_add_string(nv, ZPOOL_CONFIG_PATH, buf) == 0);
1003 for (c = 0; c < children; c++)
1004 if ((ret = make_disks(zhp, child[c])) != 0)
1007 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
1008 &child, &children) == 0)
1009 for (c = 0; c < children; c++)
1010 if ((ret = make_disks(zhp, child[c])) != 0)
1013 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
1014 &child, &children) == 0)
1015 for (c = 0; c < children; c++)
1016 if ((ret = make_disks(zhp, child[c])) != 0)
1021 #endif /* illumos */
1024 * Determine if the given path is a hot spare within the given configuration.
1027 is_spare(nvlist_t *config, const char *path)
1033 uint64_t guid, spareguid;
1039 if ((fd = open(path, O_RDONLY)) < 0)
1042 if (zpool_in_use(g_zfs, fd, &state, &name, &inuse) != 0 ||
1044 state != POOL_STATE_SPARE ||
1045 zpool_read_label(fd, &label) != 0) {
1053 verify(nvlist_lookup_uint64(label, ZPOOL_CONFIG_GUID, &guid) == 0);
1056 verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
1058 if (nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
1059 &spares, &nspares) == 0) {
1060 for (i = 0; i < nspares; i++) {
1061 verify(nvlist_lookup_uint64(spares[i],
1062 ZPOOL_CONFIG_GUID, &spareguid) == 0);
1063 if (spareguid == guid)
1072 * Go through and find any devices that are in use. We rely on libdiskmgt for
1073 * the majority of this task.
1076 is_device_in_use(nvlist_t *config, nvlist_t *nv, boolean_t force,
1077 boolean_t replacing, boolean_t isspare)
1083 char buf[MAXPATHLEN];
1085 boolean_t anyinuse = B_FALSE;
1087 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0);
1089 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
1090 &child, &children) != 0) {
1092 verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0);
1095 * As a generic check, we look to see if this is a replace of a
1096 * hot spare within the same pool. If so, we allow it
1097 * regardless of what libdiskmgt or zpool_in_use() says.
1101 if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
1102 &wholedisk) == 0 && wholedisk)
1103 (void) snprintf(buf, sizeof (buf), "%ss0",
1107 (void) strlcpy(buf, path, sizeof (buf));
1109 if (is_spare(config, buf))
1113 if (strcmp(type, VDEV_TYPE_DISK) == 0)
1114 ret = check_device(path, force, isspare);
1115 else if (strcmp(type, VDEV_TYPE_FILE) == 0)
1116 ret = check_file(path, force, isspare);
1121 for (c = 0; c < children; c++)
1122 if (is_device_in_use(config, child[c], force, replacing,
1126 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES,
1127 &child, &children) == 0)
1128 for (c = 0; c < children; c++)
1129 if (is_device_in_use(config, child[c], force, replacing,
1133 if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE,
1134 &child, &children) == 0)
1135 for (c = 0; c < children; c++)
1136 if (is_device_in_use(config, child[c], force, replacing,
1144 is_grouping(const char *type, int *mindev, int *maxdev)
1146 if (strncmp(type, "raidz", 5) == 0) {
1147 const char *p = type + 5;
1153 } else if (*p == '0') {
1154 return (NULL); /* no zero prefixes allowed */
1157 nparity = strtol(p, &end, 10);
1158 if (errno != 0 || nparity < 1 || nparity >= 255 ||
1164 *mindev = nparity + 1;
1167 return (VDEV_TYPE_RAIDZ);
1173 if (strcmp(type, "mirror") == 0) {
1176 return (VDEV_TYPE_MIRROR);
1179 if (strcmp(type, "spare") == 0) {
1182 return (VDEV_TYPE_SPARE);
1185 if (strcmp(type, "log") == 0) {
1188 return (VDEV_TYPE_LOG);
1191 if (strcmp(type, "cache") == 0) {
1194 return (VDEV_TYPE_L2CACHE);
1201 * Construct a syntactically valid vdev specification,
1202 * and ensure that all devices and files exist and can be opened.
1203 * Note: we don't bother freeing anything in the error paths
1204 * because the program is just going to exit anyway.
1207 construct_spec(int argc, char **argv)
1209 nvlist_t *nvroot, *nv, **top, **spares, **l2cache;
1210 int t, toplevels, mindev, maxdev, nspares, nlogs, nl2cache;
1213 boolean_t seen_logs;
1223 seen_logs = B_FALSE;
1229 * If it's a mirror or raidz, the subsequent arguments are
1230 * its leaves -- until we encounter the next mirror or raidz.
1232 if ((type = is_grouping(argv[0], &mindev, &maxdev)) != NULL) {
1233 nvlist_t **child = NULL;
1234 int c, children = 0;
1236 if (strcmp(type, VDEV_TYPE_SPARE) == 0) {
1237 if (spares != NULL) {
1238 (void) fprintf(stderr,
1239 gettext("invalid vdev "
1240 "specification: 'spare' can be "
1241 "specified only once\n"));
1247 if (strcmp(type, VDEV_TYPE_LOG) == 0) {
1249 (void) fprintf(stderr,
1250 gettext("invalid vdev "
1251 "specification: 'log' can be "
1252 "specified only once\n"));
1260 * A log is not a real grouping device.
1261 * We just set is_log and continue.
1266 if (strcmp(type, VDEV_TYPE_L2CACHE) == 0) {
1267 if (l2cache != NULL) {
1268 (void) fprintf(stderr,
1269 gettext("invalid vdev "
1270 "specification: 'cache' can be "
1271 "specified only once\n"));
1278 if (strcmp(type, VDEV_TYPE_MIRROR) != 0) {
1279 (void) fprintf(stderr,
1280 gettext("invalid vdev "
1281 "specification: unsupported 'log' "
1282 "device: %s\n"), type);
1288 for (c = 1; c < argc; c++) {
1289 if (is_grouping(argv[c], NULL, NULL) != NULL)
1292 child = realloc(child,
1293 children * sizeof (nvlist_t *));
1296 if ((nv = make_leaf_vdev(argv[c], B_FALSE))
1299 child[children - 1] = nv;
1302 if (children < mindev) {
1303 (void) fprintf(stderr, gettext("invalid vdev "
1304 "specification: %s requires at least %d "
1305 "devices\n"), argv[0], mindev);
1309 if (children > maxdev) {
1310 (void) fprintf(stderr, gettext("invalid vdev "
1311 "specification: %s supports no more than "
1312 "%d devices\n"), argv[0], maxdev);
1319 if (strcmp(type, VDEV_TYPE_SPARE) == 0) {
1323 } else if (strcmp(type, VDEV_TYPE_L2CACHE) == 0) {
1325 nl2cache = children;
1328 verify(nvlist_alloc(&nv, NV_UNIQUE_NAME,
1330 verify(nvlist_add_string(nv, ZPOOL_CONFIG_TYPE,
1332 verify(nvlist_add_uint64(nv,
1333 ZPOOL_CONFIG_IS_LOG, is_log) == 0);
1334 if (strcmp(type, VDEV_TYPE_RAIDZ) == 0) {
1335 verify(nvlist_add_uint64(nv,
1336 ZPOOL_CONFIG_NPARITY,
1339 verify(nvlist_add_nvlist_array(nv,
1340 ZPOOL_CONFIG_CHILDREN, child,
1343 for (c = 0; c < children; c++)
1344 nvlist_free(child[c]);
1349 * We have a device. Pass off to make_leaf_vdev() to
1350 * construct the appropriate nvlist describing the vdev.
1352 if ((nv = make_leaf_vdev(argv[0], is_log)) == NULL)
1361 top = realloc(top, toplevels * sizeof (nvlist_t *));
1364 top[toplevels - 1] = nv;
1367 if (toplevels == 0 && nspares == 0 && nl2cache == 0) {
1368 (void) fprintf(stderr, gettext("invalid vdev "
1369 "specification: at least one toplevel vdev must be "
1374 if (seen_logs && nlogs == 0) {
1375 (void) fprintf(stderr, gettext("invalid vdev specification: "
1376 "log requires at least 1 device\n"));
1381 * Finally, create nvroot and add all top-level vdevs to it.
1383 verify(nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) == 0);
1384 verify(nvlist_add_string(nvroot, ZPOOL_CONFIG_TYPE,
1385 VDEV_TYPE_ROOT) == 0);
1386 verify(nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
1387 top, toplevels) == 0);
1389 verify(nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_SPARES,
1390 spares, nspares) == 0);
1392 verify(nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_L2CACHE,
1393 l2cache, nl2cache) == 0);
1395 for (t = 0; t < toplevels; t++)
1396 nvlist_free(top[t]);
1397 for (t = 0; t < nspares; t++)
1398 nvlist_free(spares[t]);
1399 for (t = 0; t < nl2cache; t++)
1400 nvlist_free(l2cache[t]);
1411 split_mirror_vdev(zpool_handle_t *zhp, char *newname, nvlist_t *props,
1412 splitflags_t flags, int argc, char **argv)
1414 nvlist_t *newroot = NULL, **child;
1418 if ((newroot = construct_spec(argc, argv)) == NULL) {
1419 (void) fprintf(stderr, gettext("Unable to build a "
1420 "pool from the specified devices\n"));
1425 if (!flags.dryrun && make_disks(zhp, newroot) != 0) {
1426 nvlist_free(newroot);
1431 /* avoid any tricks in the spec */
1432 verify(nvlist_lookup_nvlist_array(newroot,
1433 ZPOOL_CONFIG_CHILDREN, &child, &children) == 0);
1434 for (c = 0; c < children; c++) {
1439 verify(nvlist_lookup_string(child[c],
1440 ZPOOL_CONFIG_PATH, &path) == 0);
1441 if ((type = is_grouping(path, &min, &max)) != NULL) {
1442 (void) fprintf(stderr, gettext("Cannot use "
1443 "'%s' as a device for splitting\n"), type);
1444 nvlist_free(newroot);
1450 if (zpool_vdev_split(zhp, newname, &newroot, props, flags) != 0) {
1451 nvlist_free(newroot);
1459 * Get and validate the contents of the given vdev specification. This ensures
1460 * that the nvlist returned is well-formed, that all the devices exist, and that
1461 * they are not currently in use by any other known consumer. The 'poolconfig'
1462 * parameter is the current configuration of the pool when adding devices
1463 * existing pool, and is used to perform additional checks, such as changing the
1464 * replication level of the pool. It can be 'NULL' to indicate that this is a
1465 * new pool. The 'force' flag controls whether devices should be forcefully
1466 * added, even if they appear in use.
1469 make_root_vdev(zpool_handle_t *zhp, int force, int check_rep,
1470 boolean_t replacing, boolean_t dryrun, int argc, char **argv)
1473 nvlist_t *poolconfig = NULL;
1477 * Construct the vdev specification. If this is successful, we know
1478 * that we have a valid specification, and that all devices can be
1481 if ((newroot = construct_spec(argc, argv)) == NULL)
1484 if (zhp && ((poolconfig = zpool_get_config(zhp, NULL)) == NULL))
1488 * Validate each device to make sure that its not shared with another
1489 * subsystem. We do this even if 'force' is set, because there are some
1490 * uses (such as a dedicated dump device) that even '-f' cannot
1493 if (is_device_in_use(poolconfig, newroot, force, replacing, B_FALSE)) {
1494 nvlist_free(newroot);
1499 * Check the replication level of the given vdevs and report any errors
1500 * found. We include the existing pool spec, if any, as we need to
1501 * catch changes against the existing replication level.
1503 if (check_rep && check_replication(poolconfig, newroot) != 0) {
1504 nvlist_free(newroot);
1510 * Run through the vdev specification and label any whole disks found.
1512 if (!dryrun && make_disks(zhp, newroot) != 0) {
1513 nvlist_free(newroot);