dm-crypt/System configuration
Back to Dm-crypt.
1 mkinitcpio
When encrypting a system it is necessary to regenerate the initial ramdisk after properly configuring mkinitcpio. Depending on the particular scenarios, a subset of the following hooks will have to be enabled:
- encrypt: always needed when encrypting the root partition, or a partition that needs to be mounted before root. It is not needed in all the other cases, as system initialization scripts like /etc/crypttab take care of unlocking other encrypted partitions. This hook must be placed after the udev hook, if that is used.
- sd-encrypt: systemd version of encrypt hook and used instead of encrypt hook. Must be used with systemd hook.
- keymap: provides support for foreign keymaps for typing encryption passwords; it must come before the encrypt hook. Setting your keymap is done in /etc/vconsole.conf.
- keyboard: needed to make USB keyboards work in early userspace.
Other hooks needed should be clear from other manual steps followed during the installation of the system.
2 Boot loader
In order to enable booting an encrypted root partition, a subset of the following kernel parameters need to be set. See kernel parameters for instructions specific to your boot loader.
For example using GRUB the relevant parameters are best added to /etc/default/grub before generating the boot configuration. See also GRUB#Warning when installing in chroot as another point to be aware of when installing the GRUB loader.
The kernel parameters you need to specify depend on whether or not you are using the encrypt hook or the sd-encrypt hook.
2.1 Using encrypt hook
2.1.1 cryptdevice
This parameter will make the system prompt for the passphrase to unlock the device containing the encrypted root on a cold boot. It is parsed by the encrypt hook to identify which device contains the encrypted system:
cryptdevice=device:dmname
- device is the path to the device backing the encrypted device. Usage of Persistent block device naming is advisable.
- dmname is the device-mapper name given to the device after decryption, which will be available as /dev/mapper/dmname.
- If a LVM contains the encrypted root, the LVM gets activated first and the volume group containing the logical volume of the encrypted root serves as device. It is then followed by the respective volume group to be mapped to root. The parameter follows the form of cryptdevice=/dev/vgname/lvname:dmname.
2.1.2 root
The root= parameter specifies the device of the actual (decrypted) root file system:
root=device
- If the file system is formatted directly on the decrypted device file this will be /dev/mapper/dmname.
- If a LVM gets activated first and contains an encrypted logical rootvolume, the above form applies as well.
- If the root file system is contained in a logical volume of a fully encrypted LVM, the device mapper for it will be in the general form of root=/dev/mapper/volumegroup-logicalvolume.
2.1.3 resume
resume=device
- device is the device file of the decrypted (swap) filesystem used for suspend2disk. If swap is on a separate partition, it will be in the form of /dev/mapper/swap. See also Dm-crypt/Swap encryption.
2.1.4 cryptkey
This parameter specifies the location of a keyfile and is required by the encrypt hook for reading such a keyfile to unlock the cryptdevice (unless a key is in the default location, see below). It can have three parameter sets, depending on whether the keyfile exists as a file in a particular device, a bitstream starting on a specific location, or a file in the initramfs.
For a file in a device the format is:
cryptkey=device:fstype:path
- device is the raw block device where the key exists.
- fstype is the filesystem type of device (or auto).
- path is the absolute path of the keyfile within the device.
Example: cryptkey=/dev/usbstick:vfat:/secretkey
For a bitstream on a device the key's location is specified with the following:
cryptkey=device:offset:size
where the offset is in bytes and the size in bits. Example: cryptkey=/dev/sdZ:0:512 reads a 512 bit keyfile starting at the beginning of the device.
For a file included in the initramfs the format is[1]:
cryptkey=rootfs:path
Example: cryptkey=rootfs:/secretkey
Also note that if cryptkey is not specified, it defaults to /crypto_keyfile.bin (in the initramfs).[2]
See also Dm-crypt/Device encryption#Keyfiles.
2.1.5 crypto
This parameter is specific to pass dm-crypt plain mode options to the encrypt hook.
It takes the form
crypto=<hash>:<cipher>:<keysize>:<offset>:<skip>
The arguments relate directly to the cryptsetup options. See Dm-crypt/Device encryption#Encryption options for plain mode.
For a disk encrypted with just plain default options, the crypto arguments must be specified, but each entry can be left blank:
crypto=::::
A specific example of arguments is
crypto=sha512:twofish-xts-plain64:512:0:
2.2 Using sd-encrypt hook
In all of the following luks can be replaced with rd.luks. luks parameters are honored by both the main system and initrd. rd.luks parameters are only honored by the initrd. See man 8 systemd-cryptsetup-generator for more options and more details.
2.2.1 luks.uuid
luks.uuid=UUID
Specify the UUID of the device to be decrypted on boot with this flag. If the UUID is in /etc/crypttab, the options listed there will be used.
2.2.2 luks.name
luks.name=cryptroot
Specify the name of the mapped device after the LUKS partition is open. For example, specifying cryptroot causes the unlocked device to be located at /dev/mapper/cryptroot. If this is not specified the mapped device will be located at /dev/mapper/luks-UUID where UUID is the UUID of the LUKS partition.
This is equivalent to the second parameter of encrypt's cryptdevice.
2.2.3 luks.options
luks.options=UUID=options
or
luks.options=options
Specify options for the device listed after UUID or, if not specified, for all UUIDs not specified elsewhere (e.g., cryptab).
This is roughly equivalent to the third parameter of encrypt's cryptdevice.
2.2.4 luks.key
luks.key=mykeyfile
Specify the location of a password file used to decrypt the device specified in luks.UUID. There is no default location like there is with the encrypt hook parameter cryptkey.
3 crypttab
The /etc/crypttab (encrypted device table) file is similar to the fstab file and contains a list of encrypted devices to be unlocked during system boot up. This file can be used for automatically mounting encrypted swap devices or secondary file systems.
crypttab is read before fstab, so that dm-crypt containers can be unlocked before the file system inside is mounted. Note that crypttab is read after the system has booted up, therefore it is not a replacement for unlocking encrypted partitions by using mkinitcpio hooks and boot loader options as in the case of encrypting the root partition. crypttab processing at boot time is made by the systemd-cryptsetup-generator automatically.
See the crypttab man page for details, read below for some examples, and the #Mounting at boot time section for instructions on how to use UUIDs to mount an encrypted device.
- For --type plain devices with a keyfile, it is necessary to add the hash=plain option to crypttab due to a systemd incompatibility. Do not use systemd-cryptsetup manually for device creation to work around it.
- It may be further required to add the plain option explicitly to force systemd-cryptsetup to recognize a --type plain) device at boot. GitHub issue in question.
/etc/crypttab
# Example crypttab file. Fields are: name, underlying device, passphrase, cryptsetup options. # Mount /dev/lvm/swap re-encrypting it with a fresh key each reboot swap /dev/lvm/swap /dev/urandom swap,cipher=aes-xts-plain64,size=256 # Mount /dev/lvm/tmp as /dev/mapper/tmp using plain dm-crypt with a random passphrase, making its contents unrecoverable after it is dismounted. tmp /dev/lvm/tmp /dev/urandom tmp,cipher=aes-xts-plain64,size=256 # Mount /dev/lvm/home as /dev/mapper/home using LUKS, and prompt for the passphrase at boot time. home /dev/lvm/home # Mount /dev/sdb1 as /dev/mapper/backup using LUKS, with a passphrase stored in a file. backup /dev/sdb1 /home/alice/backup.key
3.1 Mounting at boot time
If you want to mount an encrypted drive at boot time, enter the device's UUID in /etc/crypttab. You get the UUID (partition) by using the command lsblk -f and adding it to crypttab in the form:
/etc/crypttab
externaldrive UUID=2f9a8428-ac69-478a-88a2-4aa458565431 none luks,timeout=180
The first parameter is your preferred device mapper's name for the encrypted drive. The option none will trigger a prompt during boot to type the passphrase for unlocking the partition. The timeout option defines a timeout in seconds for entering the decryption password during boot.
A keyfile can also be set up and referenced instead of none. This results in an automatic unlocking, if the keyfile is accessible during boot. Since LUKS offers the option to have multiple keys, the chosen option can also be changed later.
Use the device mapper's name you've defined in /etc/crypttab in /etc/fstab as follows:
/etc/fstab
/dev/mapper/externaldrive /mnt/backup ext4 defaults,errors=remount-ro 0 2
Since /dev/mapper/externaldrive already is the result of a unique partition mapping, there is no need to specify an UUID for it. In any case, the mapper with the filesystem will have a different UUID than the partition it is encrypted in.
3.1.1 Mounting a stacked blockdevice
The systemd generators also automatically process stacked block devices at boot.
For example, you can create a RAID setup, use cryptsetup on it and create an LVM logical volume with respective filesystem inside the encrypted block device. A resulting:
$ lsblk -f
─sdXX linux_raid_member │ └─md0 crypto_LUKS │ └─cryptedbackup LVM2_member │ └─vgraid-lvraid ext4 /mnt/backup └─sdYY linux_raid_member └─md0 crypto_LUKS └─cryptedbackup LVM2_member └─vgraid-lvraid ext4 /mnt/backup
will ask for the passphrase and mount automatically at boot.
Given you specify the correct corresponding crypttab (e.g. UUID for the crypto_LUKS device) and fstab (/dev/mapper/vgraid-lvraid) entries, there is no need to add additional mkinitcpio hooks/configuration, because /etc/crypttab processing applies to non-root mounts only. One exception is when the mdadm_udev hook is used already (e.g. for the root device). In this case /etc/madadm.conf and the initramfs need updating to achieve the correct root raid is picked first.