zpool_vdev.c revision afefbcddfd8caf5f3b2da510d9439471ab225040
fa9e4066f08beec538e775443c5be79dd423fcabahrens * CDDL HEADER START
fa9e4066f08beec538e775443c5be79dd423fcabahrens * The contents of this file are subject to the terms of the
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Common Development and Distribution License, Version 1.0 only
fa9e4066f08beec538e775443c5be79dd423fcabahrens * (the "License"). You may not use this file except in compliance
fa9e4066f08beec538e775443c5be79dd423fcabahrens * with the License.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
fa9e4066f08beec538e775443c5be79dd423fcabahrens * See the License for the specific language governing permissions
fa9e4066f08beec538e775443c5be79dd423fcabahrens * and limitations under the License.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * When distributing Covered Code, include this CDDL HEADER in each
fa9e4066f08beec538e775443c5be79dd423fcabahrens * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * If applicable, add the following below this CDDL HEADER, with the
fa9e4066f08beec538e775443c5be79dd423fcabahrens * fields enclosed by brackets "[]" replaced with your own identifying
fa9e4066f08beec538e775443c5be79dd423fcabahrens * information: Portions Copyright [yyyy] [name of copyright owner]
fa9e4066f08beec538e775443c5be79dd423fcabahrens * CDDL HEADER END
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Copyright 2005 Sun Microsystems, Inc. All rights reserved.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Use is subject to license terms.
fa9e4066f08beec538e775443c5be79dd423fcabahrens#pragma ident "%Z%%M% %I% %E% SMI"
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Functions to convert between a list of vdevs and an nvlist representing the
fa9e4066f08beec538e775443c5be79dd423fcabahrens * configuration. Each entry in the list can be one of:
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Device vdevs
fa9e4066f08beec538e775443c5be79dd423fcabahrens * disk=(path=..., devid=...)
fa9e4066f08beec538e775443c5be79dd423fcabahrens * file=(path=...)
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Group vdevs
fa9e4066f08beec538e775443c5be79dd423fcabahrens * raidz=(...)
fa9e4066f08beec538e775443c5be79dd423fcabahrens * mirror=(...)
fa9e4066f08beec538e775443c5be79dd423fcabahrens * While the underlying implementation supports it, group vdevs cannot contain
fa9e4066f08beec538e775443c5be79dd423fcabahrens * other group vdevs. All userland verification of devices is contained within
fa9e4066f08beec538e775443c5be79dd423fcabahrens * this file. If successful, the nvlist returned can be passed directly to the
fa9e4066f08beec538e775443c5be79dd423fcabahrens * kernel; we've done as much verification as possible in userland.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * The only function exported by this file is 'get_vdev_spec'. The function
fa9e4066f08beec538e775443c5be79dd423fcabahrens * performs several passes:
fa9e4066f08beec538e775443c5be79dd423fcabahrens * 1. Construct the vdev specification. Performs syntax validation and
fa9e4066f08beec538e775443c5be79dd423fcabahrens * makes sure each device is valid.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * 2. Check for devices in use. Using libdiskmgt, makes sure that no
fa9e4066f08beec538e775443c5be79dd423fcabahrens * devices are also in use. Some can be overridden using the 'force'
fa9e4066f08beec538e775443c5be79dd423fcabahrens * flag, others cannot.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * 3. Check for replication errors if the 'force' flag is not specified.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * validates that the replication level is consistent across the
fa9e4066f08beec538e775443c5be79dd423fcabahrens * entire pool.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * 4. Label any whole disks with an EFI label.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * For any given vdev specification, we can have multiple errors. The
fa9e4066f08beec538e775443c5be79dd423fcabahrens * vdev_error() function keeps track of whether we have seen an error yet, and
fa9e4066f08beec538e775443c5be79dd423fcabahrens * prints out a header if its the first error we've seen.
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) fprintf(stderr, gettext("invalid vdev specification\n"));
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) fprintf(stderr, gettext("use '-f' to override "
fa9e4066f08beec538e775443c5be79dd423fcabahrens "the following errors:\n"));
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) fprintf(stderr, gettext("the following errors "
fa9e4066f08beec538e775443c5be79dd423fcabahrens "must be manually repaired:\n"));
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Some of the libdiskmgt stuff requires root privileges in order to
fa9e4066f08beec538e775443c5be79dd423fcabahrens * examine devices. Bail out gracefully in this case.
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) fprintf(stderr, gettext("cannot determine disk "
fa9e4066f08beec538e775443c5be79dd423fcabahrens "configuration: permission denied\n"));
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) fprintf(stderr, gettext("internal error: disk configuration "
fa9e4066f08beec538e775443c5be79dd423fcabahrens "error %d at line %d of file %s\n"), err, line, file);
fa9e4066f08beec538e775443c5be79dd423fcabahrens#define libdskmgt_error(err) (_libdskmgt_error((err), __FILE__, __LINE__))
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Checks whether a single slice overlaps with any of the slices in the provided
fa9e4066f08beec538e775443c5be79dd423fcabahrens * list. Called by check_overlapping().
fa9e4066f08beec538e775443c5be79dd423fcabahrensis_overlapping(dm_descriptor_t slice, dm_descriptor_t media,
fa9e4066f08beec538e775443c5be79dd423fcabahrens dm_descriptor_t *slice_list, int *error, char **overlaps_with)
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (*error != 0) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (-1);
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens *error = nvlist_lookup_uint64(media_attrs, DM_NACCESSIBLE, &media_size);
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (*error != 0) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (-1);
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (*error != 0) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (-1);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Not really possible, but the error above would catch any system
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens *error = nvlist_lookup_uint64(slice_attrs, DM_START, &start_block);
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (*error != 0) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (-1);
fa9e4066f08beec538e775443c5be79dd423fcabahrens *error = nvlist_lookup_uint64(slice_attrs, DM_SIZE, &size);
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (*error != 0) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (-1);
fa9e4066f08beec538e775443c5be79dd423fcabahrens *error = nvlist_lookup_uint32(slice_attrs, DM_INDEX, &in_snum);
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (*error != 0) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (-1);
fa9e4066f08beec538e775443c5be79dd423fcabahrens for (i = 0; slice_list[i]; i ++) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens nvlist_t *other_attrs = dm_get_attributes(slice_list[i], error);
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (*error != 0) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (-1);
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (-1);
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (-1);
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (-1);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Check to see if there are > 2 overlapping regions
fa9e4066f08beec538e775443c5be79dd423fcabahrens * on this media in the same region as this slice.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * This is done by assuming the following:
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Slice 2 is the backup slice if it is the size
fa9e4066f08beec538e775443c5be79dd423fcabahrens * of the whole disk
fa9e4066f08beec538e775443c5be79dd423fcabahrens * If slice 2 is the overlap and slice 2 is the size of
fa9e4066f08beec538e775443c5be79dd423fcabahrens * the whole disk, continue. If another slice is found
fa9e4066f08beec538e775443c5be79dd423fcabahrens * that overlaps with our slice, return it.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * There is the potential that there is more than one slice
fa9e4066f08beec538e775443c5be79dd423fcabahrens * that our slice overlaps with, however, we only return
fa9e4066f08beec538e775443c5be79dd423fcabahrens * the first overlapping slice we find.
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (start_block >= other_start && start_block <= other_end) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (*error != 0) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (-1);
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (1);
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (*error != 0) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (-1);
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (1);
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Check to see whether the given slice overlaps with any other slices. Get the
fa9e4066f08beec538e775443c5be79dd423fcabahrens * associated slice information and pass on to is_overlapping().
fa9e4066f08beec538e775443c5be79dd423fcabahrenscheck_overlapping(const char *slicename, dm_descriptor_t slice)
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Get the list of slices be fetching the associated media, and then all
fa9e4066f08beec538e775443c5be79dd423fcabahrens * associated slices.
fa9e4066f08beec538e775443c5be79dd423fcabahrens media = dm_get_associated_descriptors(slice, DM_MEDIA, &error);
fa9e4066f08beec538e775443c5be79dd423fcabahrens slices = dm_get_associated_descriptors(*media, DM_SLICE, &error);
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (is_overlapping(slice, *media, slices, &error, &overlaps)) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens vdev_error(gettext("device '%s' overlaps with '%s'\n"),
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Validate the given slice. If 'diskname' is non-NULL, then this is a single
fa9e4066f08beec538e775443c5be79dd423fcabahrens * slice on a complete disk. If 'force' is set, then the user specified '-f'
fa9e4066f08beec538e775443c5be79dd423fcabahrens * and we only want to report error for completely forbidden uses.
fa9e4066f08beec538e775443c5be79dd423fcabahrenscheck_slice(const char *slicename, dm_descriptor_t slice, int force,
fa9e4066f08beec538e775443c5be79dd423fcabahrens if ((stats = dm_get_stats(slice, DM_SLICE_STAT_USE, &err)) == NULL)
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Always check to see if this is used by an active ZFS pool.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * This slice is in use. Print out a descriptive message describing who
fa9e4066f08beec538e775443c5be79dd423fcabahrens * is using it. The 'used_by' nvlist is formatted as:
fa9e4066f08beec538e775443c5be79dd423fcabahrens * (used_by=what, used_name=desc, ...)
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Each 'used_by' must be accompanied by a 'used_name'.
fa9e4066f08beec538e775443c5be79dd423fcabahrens assert(strcmp(nvpair_name(nvwhat), DM_USED_BY) == 0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens assert(strcmp(nvpair_name(nvdesc), DM_USED_NAME) == 0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * For currently mounted filesystems, filesystems in
fa9e4066f08beec538e775443c5be79dd423fcabahrens * /etc/vfstab, or dedicated dump devices, we can never use
fa9e4066f08beec538e775443c5be79dd423fcabahrens * them, even if '-f' is specified. The rest of the errors
fa9e4066f08beec538e775443c5be79dd423fcabahrens * indicate that a filesystem was detected on disk, which can be
fa9e4066f08beec538e775443c5be79dd423fcabahrens * overridden with '-f'.
fa9e4066f08beec538e775443c5be79dd423fcabahrens "currently mounted on %s\n"),
fa9e4066f08beec538e775443c5be79dd423fcabahrens "mounted at %s in /etc/vfstab\n"),
fa9e4066f08beec538e775443c5be79dd423fcabahrens } else if (!force) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens * We should have already caught ZFS in-use
fa9e4066f08beec538e775443c5be79dd423fcabahrens * filesystems above. If the ZFS version is
fa9e4066f08beec538e775443c5be79dd423fcabahrens * different, or there was some other critical
fa9e4066f08beec538e775443c5be79dd423fcabahrens * failure, it's possible for fstyp to report it
fa9e4066f08beec538e775443c5be79dd423fcabahrens * as in-use, but zpool_open_by_dev() to fail.
fa9e4066f08beec538e775443c5be79dd423fcabahrens "outdated or damaged ZFS "
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Perform any overlap checking if requested to do so.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Validate a whole disk. Iterate over all slices on the disk and make sure
fa9e4066f08beec538e775443c5be79dd423fcabahrens * that none is in use by calling check_slice().
fa9e4066f08beec538e775443c5be79dd423fcabahrens/* ARGSUSED */
fa9e4066f08beec538e775443c5be79dd423fcabahrenscheck_disk(const char *name, dm_descriptor_t disk, int force)
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Get the drive associated with this disk. This should never fail,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * because we already have an alias handle open for the device.
fa9e4066f08beec538e775443c5be79dd423fcabahrens if ((drive = dm_get_associated_descriptors(disk, DM_DRIVE,
fa9e4066f08beec538e775443c5be79dd423fcabahrens if ((media = dm_get_associated_descriptors(*drive, DM_MEDIA,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * It is possible that the user has specified a removable media drive,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * and the media is not present.
fa9e4066f08beec538e775443c5be79dd423fcabahrens vdev_error(gettext("'%s' has no media in drive\n"), name);
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (-1);
fa9e4066f08beec538e775443c5be79dd423fcabahrens if ((slice = dm_get_associated_descriptors(*media, DM_SLICE,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Iterate over all slices and report any errors. We don't care about
fa9e4066f08beec538e775443c5be79dd423fcabahrens * overlapping slices because we are using the whole disk.
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (check_slice(dm_get_name(slice[i], &err), slice[i],
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Validate a device. Determines whether the device is a disk, slice, or
fa9e4066f08beec538e775443c5be79dd423fcabahrens * partition, and passes it off to an appropriate function.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * For whole disks, libdiskmgt does not include the leading dev path.
fa9e4066f08beec538e775443c5be79dd423fcabahrens if ((desc = dm_get_descriptor_by_name(DM_ALIAS, dev, &err)) != NULL)
fa9e4066f08beec538e775443c5be79dd423fcabahrens * If 'err' is not ENODEV, then we've had an unexpected error from
fa9e4066f08beec538e775443c5be79dd423fcabahrens * libdiskmgt. The only explanation is that we ran out of memory.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Determine if this is a slice.
fa9e4066f08beec538e775443c5be79dd423fcabahrens if ((desc = dm_get_descriptor_by_name(DM_SLICE, (char *)path, &err))
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Check for a partition. libdiskmgt expects path of /dev/rdsk when
fa9e4066f08beec538e775443c5be79dd423fcabahrens * dealing with partitions, so convert it.
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) snprintf(rpath, sizeof (rpath), "/dev/rdsk/%s", dev);
fa9e4066f08beec538e775443c5be79dd423fcabahrens if ((desc = dm_get_descriptor_by_name(DM_PARTITION, rpath, &err))
fa9e4066f08beec538e775443c5be79dd423fcabahrens /* XXZFS perform checking on partitions */
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * At this point, libdiskmgt failed to find the device as either a whole
fa9e4066f08beec538e775443c5be79dd423fcabahrens * disk or a slice. Ignore these errors, as we know that it at least a
fa9e4066f08beec538e775443c5be79dd423fcabahrens * block device. The user may have provided us with some unknown device
fa9e4066f08beec538e775443c5be79dd423fcabahrens * that libdiskmgt doesn't know about.
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Check that a file is valid. All we can do in this case is check that it's
fa9e4066f08beec538e775443c5be79dd423fcabahrens * not in use by another pool.
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (0);
fa9e4066f08beec538e775443c5be79dd423fcabahrensis_whole_disk(const char *arg, struct stat64 *statbuf)
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) snprintf(path, sizeof (path), "%s%s", arg, BACKUP_SLICE);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Create a leaf vdev. Determine if this is a file or a device. If it's a
fa9e4066f08beec538e775443c5be79dd423fcabahrens * device, fill in the device id to make a complete nvlist. Valid forms for a
fa9e4066f08beec538e775443c5be79dd423fcabahrens * leaf vdev are:
fa9e4066f08beec538e775443c5be79dd423fcabahrens * /dev/dsk/xxx Complete disk path
fa9e4066f08beec538e775443c5be79dd423fcabahrens * /xxx Full path to file
fa9e4066f08beec538e775443c5be79dd423fcabahrens * xxx Shorthand for /dev/dsk/xxx
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Determine what type of vdev this is, and put the full path into
fa9e4066f08beec538e775443c5be79dd423fcabahrens * 'path'. We detect whether this is a device of file afterwards by
fa9e4066f08beec538e775443c5be79dd423fcabahrens * checking the st_mode of the file.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Complete device or file path. Exact type is determined by
fa9e4066f08beec538e775443c5be79dd423fcabahrens * examining the file descriptor afterwards.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * This may be a short path for a device, or it could be total
fa9e4066f08beec538e775443c5be79dd423fcabahrens * gibberish. Check to see if it's a known device in
fa9e4066f08beec538e775443c5be79dd423fcabahrens * /dev/dsk/. As part of this check, see if we've been given a
fa9e4066f08beec538e775443c5be79dd423fcabahrens * an entire disk (minus the slice number).
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) snprintf(path, sizeof (path), "%s/%s", DISK_ROOT,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * If we got ENOENT, then the user gave us
fa9e4066f08beec538e775443c5be79dd423fcabahrens * gibberish, so try to direct them with a
fa9e4066f08beec538e775443c5be79dd423fcabahrens * reasonable error message. Otherwise,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * regurgitate strerror() since it's the best we
fa9e4066f08beec538e775443c5be79dd423fcabahrens "shorthand device name\n"));
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Determine whether this is a device or a file.
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) fprintf(stderr, gettext("cannot use '%s': must be a "
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Finally, we have the complete device or file, and we know that it is
fa9e4066f08beec538e775443c5be79dd423fcabahrens * acceptable to use. Construct the nvlist to describe this vdev. All
fa9e4066f08beec538e775443c5be79dd423fcabahrens * vdevs have a 'path' element, and devices also have a 'devid' element.
fa9e4066f08beec538e775443c5be79dd423fcabahrens verify(nvlist_add_string(vdev, ZPOOL_CONFIG_PATH, path) == 0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens verify(nvlist_add_string(vdev, ZPOOL_CONFIG_TYPE, type) == 0);
afefbcddfd8caf5f3b2da510d9439471ab225040eschrock verify(nvlist_add_uint64(vdev, ZPOOL_CONFIG_WHOLE_DISK,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * For a whole disk, defer getting its devid until after labeling it.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Get the devid for the device.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Go through and verify the replication level of the pool is consistent.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Performs the following checks:
fa9e4066f08beec538e775443c5be79dd423fcabahrens * For the new spec, verifies that devices in mirrors and raidz are the
fa9e4066f08beec538e775443c5be79dd423fcabahrens * same size.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * If the current configuration already has inconsistent replication
fa9e4066f08beec538e775443c5be79dd423fcabahrens * levels, ignore any other potential problems in the new spec.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Otherwise, make sure that the current spec (if there is one) and the new
fa9e4066f08beec538e775443c5be79dd423fcabahrens * spec have consistent replication levels.
fa9e4066f08beec538e775443c5be79dd423fcabahrenstypedef struct replication_level {
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Given a list of toplevel vdevs, return the current replication level. If
fa9e4066f08beec538e775443c5be79dd423fcabahrens * the config is inconsistent, then NULL is returned. If 'fatal' is set, then
fa9e4066f08beec538e775443c5be79dd423fcabahrens * an error message will be displayed for each self-inconsistent vdev.
fa9e4066f08beec538e775443c5be79dd423fcabahrens verify(nvlist_lookup_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
fa9e4066f08beec538e775443c5be79dd423fcabahrens for (t = 0; t < toplevels; t++) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * This is a 'file' or 'disk' vdev.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * This is a mirror or RAID-Z vdev. Go through and make
fa9e4066f08beec538e775443c5be79dd423fcabahrens * sure the contents are all the same (files vs. disks),
fa9e4066f08beec538e775443c5be79dd423fcabahrens * keeping track of the number of elements in the
fa9e4066f08beec538e775443c5be79dd423fcabahrens * We also check that the size of each vdev (if it can
fa9e4066f08beec538e775443c5be79dd423fcabahrens * be determined) is the same.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * The 'dontreport' variable indicatest that we've
fa9e4066f08beec538e775443c5be79dd423fcabahrens * already reported an error for this spec, so don't
fa9e4066f08beec538e775443c5be79dd423fcabahrens * bother doing it again.
fa9e4066f08beec538e775443c5be79dd423fcabahrens for (c = 0; c < children; c++) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens * If we have a raidz/mirror that combines disks
fa9e4066f08beec538e775443c5be79dd423fcabahrens * with files, report it as an error.
fa9e4066f08beec538e775443c5be79dd423fcabahrens "mismatched replication "
fa9e4066f08beec538e775443c5be79dd423fcabahrens "level: %s contains both "
fa9e4066f08beec538e775443c5be79dd423fcabahrens "files and devices\n"),
fa9e4066f08beec538e775443c5be79dd423fcabahrens * According to stat(2), the value of 'st_size'
fa9e4066f08beec538e775443c5be79dd423fcabahrens * is undefined for block devices and character
fa9e4066f08beec538e775443c5be79dd423fcabahrens * devices. But there is no effective way to
fa9e4066f08beec538e775443c5be79dd423fcabahrens * determine the real size in userland.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Instead, we'll take advantage of an
fa9e4066f08beec538e775443c5be79dd423fcabahrens * implementation detail of spec_size(). If the
fa9e4066f08beec538e775443c5be79dd423fcabahrens * device is currently open, then we (should)
fa9e4066f08beec538e775443c5be79dd423fcabahrens * return a valid size.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * If we still don't get a valid size (indicated
fa9e4066f08beec538e775443c5be79dd423fcabahrens * by a size of 0 or MAXOFFSET_T), then ignore
fa9e4066f08beec538e775443c5be79dd423fcabahrens * this device altogether.
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (err != 0 ||
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Also check the size of each device. If they
fa9e4066f08beec538e775443c5be79dd423fcabahrens * differ, then report an error.
fa9e4066f08beec538e775443c5be79dd423fcabahrens "%s contains devices of "
fa9e4066f08beec538e775443c5be79dd423fcabahrens "different sizes\n"),
fa9e4066f08beec538e775443c5be79dd423fcabahrens * At this point, we have the replication of the last toplevel
fa9e4066f08beec538e775443c5be79dd423fcabahrens * vdev in 'rep'. Compare it to 'lastrep' to see if its
fa9e4066f08beec538e775443c5be79dd423fcabahrens * different.
fa9e4066f08beec538e775443c5be79dd423fcabahrens "mismatched replication "
fa9e4066f08beec538e775443c5be79dd423fcabahrens "level: both %s and %s vdevs are "
fa9e4066f08beec538e775443c5be79dd423fcabahrens "present\n"),
fa9e4066f08beec538e775443c5be79dd423fcabahrens "mismatched replication "
fa9e4066f08beec538e775443c5be79dd423fcabahrens "level: %d-way %s and %d-way %s "
fa9e4066f08beec538e775443c5be79dd423fcabahrens "vdevs are present\n"),
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Check the replication level of the vdev spec against the current pool. Calls
fa9e4066f08beec538e775443c5be79dd423fcabahrens * get_replication() to make sure the new spec is self-consistent. If the pool
fa9e4066f08beec538e775443c5be79dd423fcabahrens * has a consistent replication level, then we ignore any errors. Otherwise,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * report any difference between the two.
fa9e4066f08beec538e775443c5be79dd423fcabahrenscheck_replication(nvlist_t *config, nvlist_t *newroot)
fa9e4066f08beec538e775443c5be79dd423fcabahrens * If we have a current pool configuration, check to see if it's
fa9e4066f08beec538e775443c5be79dd423fcabahrens * self-consistent. If not, simply return success.
fa9e4066f08beec538e775443c5be79dd423fcabahrens verify(nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE,
fa9e4066f08beec538e775443c5be79dd423fcabahrens if ((current = get_replication(nvroot, FALSE)) == NULL)
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Get the replication level of the new vdev spec, reporting any
fa9e4066f08beec538e775443c5be79dd423fcabahrens * inconsistencies found.
fa9e4066f08beec538e775443c5be79dd423fcabahrens if ((new = get_replication(newroot, TRUE)) == NULL) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (-1);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Check to see if the new vdev spec matches the replication level of
fa9e4066f08beec538e775443c5be79dd423fcabahrens * the current pool.
fa9e4066f08beec538e775443c5be79dd423fcabahrens "mismatched replication level: pool uses %d-way %s "
fa9e4066f08beec538e775443c5be79dd423fcabahrens "and new vdev uses %d-way %s\n"),
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Label an individual disk. The name provided is the short name, stripped of
fa9e4066f08beec538e775443c5be79dd423fcabahrens * any leading /dev path.
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) snprintf(path, sizeof (path), "%s/%s%s", RDISK_ROOT, name,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * This shouldn't happen. We've long since verified that this
fa9e4066f08beec538e775443c5be79dd423fcabahrens * is a valid device.
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) fprintf(stderr, gettext("cannot open '%s': %s\n"),
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (-1);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * The only way this can fail is if we run out of memory, or we
fa9e4066f08beec538e775443c5be79dd423fcabahrens * were unable to read the disk geometry.
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) fprintf(stderr, gettext("cannot label '%s': unable to "
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (-1);
fa9e4066f08beec538e775443c5be79dd423fcabahrens vtoc->efi_parts[0].p_size = vtoc->efi_last_u_lba + 1 -
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Why we use V_USR: V_BACKUP confuses users, and is considered
fa9e4066f08beec538e775443c5be79dd423fcabahrens * disposable by some EFI utilities (since EFI doesn't have a backup
fa9e4066f08beec538e775443c5be79dd423fcabahrens * slice). V_UNASSIGNED is supposed to be used only for zero size
fa9e4066f08beec538e775443c5be79dd423fcabahrens * partitions, and efi_write() will fail if we use it. V_ROOT, V_BOOT,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * etc. were all pretty specific. V_USR is as close to reality as we
fa9e4066f08beec538e775443c5be79dd423fcabahrens * can get, in the absence of V_OTHER.
fa9e4066f08beec538e775443c5be79dd423fcabahrens vtoc->efi_parts[8].p_start = vtoc->efi_last_u_lba + 1 - resv;
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Currently, EFI labels are not supported for IDE disks, and it
fa9e4066f08beec538e775443c5be79dd423fcabahrens * is likely that they will not be supported on other drives for
fa9e4066f08beec538e775443c5be79dd423fcabahrens * some time. Print out a helpful error message directing the
fa9e4066f08beec538e775443c5be79dd423fcabahrens * user to manually label the disk and give a specific slice.
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) fprintf(stderr, gettext("cannot label '%s': failed to "
fa9e4066f08beec538e775443c5be79dd423fcabahrens (void) fprintf(stderr, gettext("use fdisk(1M) to partition "
fa9e4066f08beec538e775443c5be79dd423fcabahrens "the disk, and provide a specific slice\n"));
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (-1);
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Go through and find any whole disks in the vdev specification, labelling them
fa9e4066f08beec538e775443c5be79dd423fcabahrens * as appropriate. When constructing the vdev spec, we were unable to open this
fa9e4066f08beec538e775443c5be79dd423fcabahrens * device in order to provide a devid. Now that we have labelled the disk and
fa9e4066f08beec538e775443c5be79dd423fcabahrens * know that slice 0 is valid, we can construct the devid now.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * If the disk was already labelled with an EFI label, we will have gotten the
fa9e4066f08beec538e775443c5be79dd423fcabahrens * devid already (because we were able to open the whole disk). Otherwise, we
fa9e4066f08beec538e775443c5be79dd423fcabahrens * need to get the devid after we label the disk.
fa9e4066f08beec538e775443c5be79dd423fcabahrens verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * We have a disk device. Get the path to the device
fa9e4066f08beec538e775443c5be79dd423fcabahrens * and see if its a whole disk by appending the backup
fa9e4066f08beec538e775443c5be79dd423fcabahrens * slice and stat()ing the device.
fa9e4066f08beec538e775443c5be79dd423fcabahrens verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0);
afefbcddfd8caf5f3b2da510d9439471ab225040eschrock if (nvlist_lookup_uint64(nv, ZPOOL_CONFIG_WHOLE_DISK,
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (-1);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Fill in the devid, now that we've labeled the disk.
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (-1);
afefbcddfd8caf5f3b2da510d9439471ab225040eschrock * Update the path to refer to the 's0' slice. The presence of
afefbcddfd8caf5f3b2da510d9439471ab225040eschrock * the 'whole_disk' field indicates to the CLI that we should
afefbcddfd8caf5f3b2da510d9439471ab225040eschrock * chop off the slice number when displaying the device in
afefbcddfd8caf5f3b2da510d9439471ab225040eschrock * future output.
afefbcddfd8caf5f3b2da510d9439471ab225040eschrock verify(nvlist_add_string(nv, ZPOOL_CONFIG_PATH, buf) == 0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens for (c = 0; c < children; c++)
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Go through and find any devices that are in use. We rely on libdiskmgt for
fa9e4066f08beec538e775443c5be79dd423fcabahrens * the majority of this task.
fa9e4066f08beec538e775443c5be79dd423fcabahrens verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_TYPE, &type) == 0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_CHILDREN,
fa9e4066f08beec538e775443c5be79dd423fcabahrens verify(nvlist_lookup_string(nv, ZPOOL_CONFIG_PATH, &path) == 0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens for (c = 0; c < children; c++)
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Construct a syntactically valid vdev specification,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * and ensure that all devices and files exist and can be opened.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Note: we don't bother freeing anything in the error paths
fa9e4066f08beec538e775443c5be79dd423fcabahrens * because the program is just going to exit anyway.
fa9e4066f08beec538e775443c5be79dd423fcabahrens while (argc > 0) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens * If it's a mirror or raidz, the subsequent arguments are
fa9e4066f08beec538e775443c5be79dd423fcabahrens * its leaves -- until we encounter the next mirror or raidz.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Mirrors and RAID-Z devices require at least
fa9e4066f08beec538e775443c5be79dd423fcabahrens * two components.
fa9e4066f08beec538e775443c5be79dd423fcabahrens for (c = 0; c < children; c++)
fa9e4066f08beec538e775443c5be79dd423fcabahrens * We have a device. Pass off to make_leaf_vdev() to
fa9e4066f08beec538e775443c5be79dd423fcabahrens * construct the appropriate nvlist describing the vdev.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Finally, create nvroot and add all top-level vdevs to it.
fa9e4066f08beec538e775443c5be79dd423fcabahrens verify(nvlist_alloc(&nvroot, NV_UNIQUE_NAME, 0) == 0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens verify(nvlist_add_nvlist_array(nvroot, ZPOOL_CONFIG_CHILDREN,
fa9e4066f08beec538e775443c5be79dd423fcabahrens for (t = 0; t < toplevels; t++)
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Get and validate the contents of the given vdev specification. This ensures
fa9e4066f08beec538e775443c5be79dd423fcabahrens * that the nvlist returned is well-formed, that all the devices exist, and that
fa9e4066f08beec538e775443c5be79dd423fcabahrens * they are not currently in use by any other known consumer. The 'poolconfig'
fa9e4066f08beec538e775443c5be79dd423fcabahrens * parameter is the current configuration of the pool when adding devices
fa9e4066f08beec538e775443c5be79dd423fcabahrens * existing pool, and is used to perform additional checks, such as changing the
fa9e4066f08beec538e775443c5be79dd423fcabahrens * replication level of the pool. It can be 'NULL' to indicate that this is a
fa9e4066f08beec538e775443c5be79dd423fcabahrens * new pool. The 'force' flag controls whether devices should be forcefully
fa9e4066f08beec538e775443c5be79dd423fcabahrens * added, even if they appear in use.
fa9e4066f08beec538e775443c5be79dd423fcabahrensmake_root_vdev(nvlist_t *poolconfig, int force, int check_rep,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Construct the vdev specification. If this is successful, we know
fa9e4066f08beec538e775443c5be79dd423fcabahrens * that we have a valid specification, and that all devices can be
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Validate each device to make sure that its not shared with another
fa9e4066f08beec538e775443c5be79dd423fcabahrens * subsystem. We do this even if 'force' is set, because there are some
fa9e4066f08beec538e775443c5be79dd423fcabahrens * uses (such as a dedicated dump device) that even '-f' cannot
fa9e4066f08beec538e775443c5be79dd423fcabahrens * override.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Check the replication level of the given vdevs and report any errors
fa9e4066f08beec538e775443c5be79dd423fcabahrens * found. We include the existing pool spec, if any, as we need to
fa9e4066f08beec538e775443c5be79dd423fcabahrens * catch changes against the existing replication level.
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (check_rep && check_replication(poolconfig, newroot) != 0) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Run through the vdev specification and label any whole disks found.