man.bsd.lv manual page server

Raw or in-core label

The disk label resides close to or at the beginning of each disk slice. For faster access, the kernel maintains a copy in core at all times. By default, most forms of the disklabel64 command access the in-core copy of the label. To access the raw (on-disk) copy, use the -r option. This option allows a label to be installed on a disk without kernel support for a label, such as when labels are first installed on a system; it must be used when first installing a label on a disk. The specific effect of -r is described under each command.

Disk device name

All disklabel64 forms require a disk device name, which should always be the raw device name representing the disk or slice. DragonFly uses the following scheme for slice numbering: If the disk doesn't use GPT (typically laid out by gpt(8)), but e.g. MBR (typically laid out by fdisk(8)), then slice 0, e.g. da0s0, represents the entire disk regardless of any DOS partitioning. Slice 0 is called the compatibility slice, and slice 1 and onward, e.g. da0s1, represents a BSD slice. If the disk does use GPT, then all slices are BSD slices, slice 0 isn't special, it is just the first slice on the disk. You do not have to include the /dev/ path prefix when specifying the device. The disklabel64 utility will automatically prepend it.

Reading the disk label

To examine the label on a disk drive, use disklabel64 without options:

disklabel64 [-r] disk

disk represents the raw disk in question, and may be in the form da0s1 or /dev/da0s1. It will display all of the parameters associated with the drive and its partition layout. Unless the -r flag is given, the kernel's in-core copy of the label is displayed; if the disk has no label, or the partition types on the disk are incorrect, the kernel may have constructed or modified the label. If the -r flag is given, disklabel64 reads the label from the raw disk and displays it. Both versions are usually identical except in the case where a label has not yet been initialized or is corrupt.

Writing a standard label

To write a standard label, use the form

disklabel64 -w [-r] [-n] disk [disktype/auto [packid]]

The required argument to disklabel64 is the drive to be labeled. The first optional argument is the drive type as described in the disktab(5) file, from which the drive parameters and partitions are taken. If not specified, then the auto type is assumed to make a virgin label for the disk as described below. If different disks of the same physical type are to have different partitions, it will be necessary to have separate disktab entries describing each, or to edit the label after installation as described below. The optional argument is a pack identification string, up to 63 characters long. The pack id must be quoted if it contains blanks.

If the -n flag is given, no data will be written to the device, and instead the disklabel that would have been written will be printed to stdout.

If the -r flag is given, the disk sectors containing the label and bootstrap will be written directly. A side-effect of this is that any existing bootstrap code will be overwritten and the disk rendered unbootable. See the boot options below for a method of writing the label and the bootstrap at the same time. If -r is not specified, the existing label will be updated via the in-core copy and any bootstrap code will be unaffected. If the disk does not already have a label, the -r flag must be used. In either case, the kernel's in-core label is replaced.

For a virgin disk that is not known to disktab(5), disktype can be specified as auto. In this case, the driver is requested to produce a virgin label for the disk. This might or might not be successful, depending on whether the driver for the disk is able to get the required data without reading anything from the disk at all. It will likely succeed for all SCSI disks, most IDE disks, and vnode devices. Writing a label to the disk is the only supported operation.

For most harddisks, a label based on percentages for most partitions (and one partition with a size of ‘*’) will produce a reasonable configuration.

PC-based systems have special requirements in order for the BIOS to properly recognize a DragonFly disklabel. Older systems may require what is known as a “dangerously dedicated” disklabel, which creates a fake DOS partition to work around problems older BIOSes have with modern disk geometries. On newer systems you generally want to create a normal DOS partition using fdisk and then create a DragonFly disklabel within that slice. This is described later on in this page.

Installing a new disklabel does not in of itself allow your system to boot a kernel using that label. You must also install boot blocks, which is described later on in this manual page.

Editing an existing disk label

To edit an existing disk label, use the form

disklabel64 -e [-r] [-n] disk

This command reads the label from the in-core kernel copy, or directly from the disk if the -r flag is also specified. The label is written to a file in ASCII and then supplied to an editor for changes. If no editor is specified in an EDITOR environment variable, vi(1) is used. When the editor terminates, the label file is used to rewrite the disk label. Existing bootstrap code is unchanged regardless of whether -r was specified. If -n is specified, no data will be written to the device, and instead the disklabel that would have been written will be printed to stdout. This is useful to see how a partitioning scheme will work out for a specific disk.

Restoring a disk label from a file

To restore a disk label from a file, use the form

disklabel64 -R [-r] [-n] disk protofile

disklabel64 is capable of restoring a disk label that was previously saved in a file in ASCII format. The prototype file used to create the label should be in the same format as that produced when reading or editing a label. Comments are delimited by ‘#’ and newline. As when writing a new label, any existing bootstrap code will be clobbered if -r is specified and will be unaffected otherwise. See the boot options below for a method of restoring the label and writing the bootstrap at the same time. If -n is used, no data will be written to the device, and instead the disklabel that would have been written will be printed to stdout. This is useful to see how a partitioning scheme will work out for a specific disk.

Enabling and disabling writing to the disk label area

By default, it is not possible to write to the disk label area at the beginning of a disk. The disk driver arranges for write(2) and similar system calls to return EROFS on any attempt to do so. If you need to write to this area (for example, to obliterate the label), use the form

disklabel64 -W disk

To disallow writing to the label area after previously allowing it, use the command

disklabel64 -N disk

Installing bootstraps

The final three forms of disklabel64 are used to install bootstrap code, which allows boot from a HAMMER(5), HAMMER2(5), or UFS(5) file system. If you are creating a “dangerously-dedicated” slice for compatibility with older PC systems, you generally want to specify the compatibility slice, such as da0s0. If you are creating a label within an existing DOS slice, you should specify the slice name such as da0s1. Making a slice bootable can be tricky. If you are using a normal DOS slice you typically install (or leave) a standard MBR on the base disk and then install the DragonFly bootblocks in the slice.

disklabel64 -B [-b boot1 -s boot2] disk [disktype/auto]

This form installs the bootstrap only. It does not change the disk label. You should never use this command on the compatibility slice unless you intend to create a “dangerously-dedicated” disk, such as da0s0. This command is typically run on a BSD slice such as da0s1.

disklabel64 -w -B [-n] [-b boot1 -s boot2] disk [disktype/auto [packid]]

This form corresponds to the “write label” command described above. In addition to writing a new volume label, it also installs the bootstrap. If run on the compatibility slice this command will create a “dangerously-dedicated” label. This command is normally run on a BSD slice rather than the compatibility slice. If -n is used, no data will be written to the device, and instead the disklabel that would have been written will be printed to stdout.

disklabel64 -R -B [-n] [-b boot1 -s boot2] disk protofile [disktype/auto]

This form corresponds to the “restore label” command described above. In addition to restoring the volume label, it also installs the bootstrap. If run on the compatibility slice this command will create a “dangerously-dedicated” label. This command is normally run on a BSD slice rather than the compatibility slice.

The bootstrap commands always access the disk directly, so it is not necessary to specify the -r flag. If -n is used, no data will be written to the device, and instead the disklabel that would have been written will be printed to stdout.

The bootstrap code is comprised of two boot programs. Specify the name of the boot programs to be installed in one of these ways:

1. Specify the names explicitly with the -b and -s flags. -b indicates the primary boot program and -s the secondary boot program. The boot programs are normally located in /boot.
2. If the -b and -s flags are not specified, but disktype was specified, the names of the programs are taken from the “b0” and “b1” parameters of the disktab(5) entry for the disk if the disktab entry exists and includes those parameters.
3. Otherwise, the default boot image names are used: /boot/boot1_64 and /boot/boot2_64 for the standard stage1 and stage2 boot images.

Initializing/Formatting a bootable disk from scratch

To initialize a disk from scratch the following sequence is recommended. Please note that this will wipe everything that was previously on the disk, including any non-DragonFly slices.

1. Use gpt(8) or fdisk(8) to initialize the hard disk, and create a GPT or MBR slice table, referred to as the “partition table” in DOS.
2. Use disklabel64 or disklabel32(8) to define partitions on DragonFly slices created in the previous step.
3. Finally use newfs_hammer(8), newfs_hammer2(8), or newfs(8) to create file systems on new partitions.

A typical partitioning scheme would be to have an ‘a’ partition of approximately 1 GB for /boot to hold the current, old (and backup) kernels and modules, a ‘b’ partition for swap (suggested to be at least the same size as the physical memory), and finally a ‘d’ partition for the root file system (usually all remaining space). Your mileage may vary.