Dm-crypt/Encrypting an entire system: Difference between revisions

From ArchWiki
(→‎Configuring the boot loader: the grub package doesn't contain /boot/grub/ anymore, run grub-install before grub-mkconfig so that the directory gets created)
(→‎Configuring mkinitcpio: - add a note that systemd hook is a replacement to the default udev)
Line 289: Line 289:


  HOOKS=(base '''systemd''' autodetect '''keyboard''' '''sd-vconsole''' modconf block '''sd-encrypt''' '''sd-lvm2''' filesystems fsck)
  HOOKS=(base '''systemd''' autodetect '''keyboard''' '''sd-vconsole''' modconf block '''sd-encrypt''' '''sd-lvm2''' filesystems fsck)
Note that the {{ic|systemd}} hook replaces the default {{ic|udev}} hook.


See [[dm-crypt/System configuration#mkinitcpio]] for details and other hooks that you may need.
See [[dm-crypt/System configuration#mkinitcpio]] for details and other hooks that you may need.

Revision as of 21:55, 26 December 2018

The following are examples of common scenarios of full system encryption with dm-crypt. They explain all the adaptations that need to be done to the normal installation procedure. All the necessary tools are on the installation image.

Overview

Securing a root filesystem is where dm-crypt excels, feature and performance-wise. Unlike selectively encrypting non-root filesystems, an encrypted root filesystem can conceal information such as which programs are installed, the usernames of all user accounts, and common data-leakage vectors such as mlocate and /var/log/. Furthermore, an encrypted root filesystem makes tampering with the system far more difficult, as everything except the boot loader and (usually) the kernel is encrypted.

All scenarios illustrated in the following share these advantages, other pros and cons differentiating them are summarized below:

Scenarios Advantages Disadvantages
#Simple partition layout with LUKS

shows a basic and straight-forward set-up for a fully LUKS encrypted root.

  • Simple partitioning and setup
  • Inflexible; disk-space to be encrypted has to be pre-allocated
#LVM on LUKS

achieves partitioning flexibility by using LVM inside a single LUKS encrypted partition.

  • Simple partitioning with knowledge of LVM
  • Only one key required to unlock all volumes (e.g. easy resume-from-disk setup)
  • Volume layout not transparent when locked
  • Easiest method to allow suspension to disk
  • LVM adds an additional mapping layer and hook
  • Less useful, if a singular volume should receive a separate key
#LUKS on LVM

uses dm-crypt only after the LVM is setup.

  • LVM can be used to have encrypted volumes span multiple disks
  • Easy mix of un-/encrypted volume groups
  • Complex; changing volumes requires changing encryption mappers too
  • Volumes require individual keys
  • LVM layout is transparent when locked
#LUKS on software RAID

uses dm-crypt only after RAID is setup.

  • Analogous to LUKS on LVM
  • Analogous to LUKS on LVM
#Plain dm-crypt

uses dm-crypt plain mode, i.e. without a LUKS header and its options for multiple keys.
This scenario also employs USB devices for /boot and key storage, which may be applied to the other scenarios.

  • High care to all encryption parameters is required
  • Single encryption key and no option to change it
#Encrypted boot partition (GRUB)

shows how to encrypt the boot partition using the GRUB bootloader.
This scenario also employs an EFI system partition, which may be applied to the other scenarios.

  • Same advantages as the scenario the installation is based on (LVM on LUKS for this particular example)
  • Less data is left unencrypted, i.e. the boot loader and the EFI system partition, if present
  • Same disadvantages as the scenario the installation is based on (LVM on LUKS for this particular example)
  • More complicated configuration
  • Not supported by other boot loaders
#Btrfs subvolumes with swap

shows how to encrypt a Btrfs system, including the /boot directory, also adding a partition for swap, on UEFI hardware.

While all above scenarios provide much greater protection from outside threats than encrypted secondary filesystems, they also share a common disadvantage: any user in possession of the encryption key is able to decrypt the entire drive, and therefore can access other users' data. If that is of concern, it is possible to use a combination of blockdevice and stacked filesystem encryption and reap the advantages of both. See Disk encryption to plan ahead.

See dm-crypt/Drive preparation#Partitioning for a general overview of the partitioning strategies used in the scenarios.

Another area to consider is whether to set up an encrypted swap partition and what kind. See dm-crypt/Swap encryption for alternatives.

If you anticipate to protect the system's data not only against physical theft, but also have a requirement of precautions against logical tampering, see dm-crypt/Specialties#Securing the unencrypted boot partition for further possibilities after following one of the scenarios.

For solid state drives you might want to consider enabling TRIM support, but be warned, there are potential security implications. See dm-crypt/Specialties#Discard/TRIM support for solid state drives (SSD) for more information.

Warning:
  • In any scenario, never use file system repair software such as fsck directly on an encrypted volume, or it will destroy any means to recover the key used to decrypt your files. Such tools must be used on the decrypted (opened) device instead.
  • For the LUKS2 format:
    • GRUB does not support LUKS2. Do not use LUKS2 on partitions that GRUB needs to access.
    • The LUKS2 format has a high RAM usage per design, defaulting to 1GB per encrypted mapper. Machines with low RAM and/or multiple LUKS2 partitions unlocked in parallel may error on boot. See the --pbkdf-memory option to control memory usage.[1]

Simple partition layout with LUKS

This example covers a full system encryption with dmcrypt + LUKS in a simple partition layout:

+-----------------------+------------------------+-----------------------+
| Boot partition        | LUKS2 encrypted system | Optional free space   |
|                       | partition              | for additional        |
|                       |                        | partitions or swap    |
| /boot                 | /                      | to be setup later     |
|                       |                        |                       |
|                       | /dev/mapper/cryptroot  |                       |
|                       |------------------------|                       |
| /dev/sda1             | /dev/sda2              |                       |
+-----------------------+------------------------+-----------------------+

The first steps can be performed directly after booting the Arch Linux install image.

Preparing the disk

Prior to creating any partitions, you should inform yourself about the importance and methods to securely erase the disk, described in dm-crypt/Drive preparation.

Then create the needed partitions, at least one for / (e.g. /dev/sda2) and /boot (/dev/sda1). See Partitioning.

Preparing non-boot partitions

The following commands create and mount the encrypted root partition. They correspond to the procedure described in detail in dm-crypt/Encrypting a non-root file system#Partition (which, despite the title, can be applied to root partitions, as long as mkinitcpio and the boot loader are correctly configured). If you want to use particular non-default encryption options (e.g. cipher, key length), see the encryption options before executing the first command:

# cryptsetup -y -v luksFormat --type luks2 /dev/sda2
# cryptsetup open /dev/sda2 cryptroot
# mkfs.ext4 /dev/mapper/cryptroot
# mount /dev/mapper/cryptroot /mnt

Check the mapping works as intended:

# umount /mnt
# cryptsetup close cryptroot
# cryptsetup open /dev/sda2 cryptroot
# mount /dev/mapper/cryptroot /mnt

If you created separate partitions (e.g. /home), these steps have to be adapted and repeated for all of them, except for /boot. See dm-crypt/Encrypting a non-root file system#Automated unlocking and mounting on how to handle additional partitions at boot.

Note that each blockdevice requires its own passphrase. This may be inconvenient, because it results in a separate passphrase to be input during boot. An alternative is to use a keyfile stored in the system partition to unlock the separate partition via crypttab. See dm-crypt/Device encryption#Using LUKS to format partitions with a keyfile for instructions.

Preparing the boot partition

What you do have to setup is a non-encrypted /boot partition, which is needed for a encrypted root. For a standard MBR/non-EFI /boot partition, for example, execute:

# mkfs.ext4 /dev/sda1
# mkdir /mnt/boot
# mount /dev/sda1 /mnt/boot

Mounting the devices

At Installation guide#Mount the file systems you will have to mount the mapped devices, not the actual partitions. Of course /boot, which is not encrypted, will still have to be mounted directly.

Configuring mkinitcpio

Add the keyboard, keymap and encrypt hooks to mkinitcpio.conf. If the default US keymap is fine for you, you can omit the keymap hook.

HOOKS=(base udev autodetect keyboard keymap consolefont modconf block encrypt filesystems fsck)

If using the sd-encrypt hook with the systemd-based initramfs, the following needs to be set instead:

HOOKS=(base systemd autodetect keyboard sd-vconsole modconf block sd-encrypt filesystems fsck)

Depending on which other hooks are used, the order may be relevant. See dm-crypt/System configuration#mkinitcpio for details and other hooks that you may need.

Configuring the boot loader

In order to unlock the encrypted root partition at boot, the following kernel parameters need to be set by the boot loader:

cryptdevice=UUID=device-UUID:cryptroot root=/dev/mapper/cryptroot

If using the sd-encrypt hook, the following need to be set instead:

rd.luks.name=device-UUID=cryptroot root=/dev/mapper/cryptroot

See dm-crypt/System configuration#Boot loader for details.

The device-UUID refers to the UUID of /dev/sda2. See Persistent block device naming for details.

LVM on LUKS

The straight-forward method is to set up LVM on top of the encrypted partition instead of the other way round. Technically the LVM is setup inside one big encrypted blockdevice. Hence, the LVM is not transparent until the blockdevice is unlocked and the underlying volume structure is scanned and mounted during boot.

The disk layout in this example is:

+-----------------------------------------------------------------------+ +----------------+
| Logical volume 1      | Logical volume 2      | Logical volume 3      | | Boot partition |
|                       |                       |                       | |                |
| [SWAP]                | /                     | /home                 | | /boot          |
|                       |                       |                       | |                |
| /dev/MyVolGroup/swap  | /dev/MyVolGroup/root  | /dev/MyVolGroup/home  | |                |
|_ _ _ _ _ _ _ _ _ _ _ _|_ _ _ _ _ _ _ _ _ _ _ _|_ _ _ _ _ _ _ _ _ _ _ _| | (may be on     |
|                                                                       | | other device)  |
|                         LUKS2 encrypted partition                     | |                |
|                           /dev/sda1                                   | | /dev/sdb1      |
+-----------------------------------------------------------------------+ +----------------+
Note: While using the encrypt hook this method does not allow you to span the logical volumes over multiple disks; either use the sd-encrypt or see dm-crypt/Specialties#Modifying the encrypt hook for multiple partitions.
Tip: Three variants of this setup:

Preparing the disk

Prior to creating any partitions, you should inform yourself about the importance and methods to securely erase the disk, described in dm-crypt/Drive preparation.

Tip: When using the GRUB bootloader together with GPT, create a BIOS boot partition.

Create a partition to be mounted at /boot of type 8300 with a size of 200 MiB or more.

Create a partition of type 8E00, which will later contain the encrypted container.

Create the LUKS encrypted container at the "system" partition. Enter the chosen password twice.

# cryptsetup luksFormat --type luks2 /dev/sda1

For more information about the available cryptsetup options see the LUKS encryption options prior to above command.

Open the container:

# cryptsetup open /dev/sda1 cryptlvm

The decrypted container is now available at /dev/mapper/cryptlvm.

Preparing the logical volumes

Create a physical volume on top of the opened LUKS container:

# pvcreate /dev/mapper/cryptlvm

Create the volume group named MyVolGroup (or whatever you want), adding the previously created physical volume to it:

# vgcreate MyVolGroup /dev/mapper/cryptlvm

Create all your logical volumes on the volume group:

# lvcreate -L 8G MyVolGroup -n swap
# lvcreate -L 32G MyVolGroup -n root
# lvcreate -l 100%FREE MyVolGroup -n home

Format your filesystems on each logical volume:

# mkfs.ext4 /dev/MyVolGroup/root
# mkfs.ext4 /dev/MyVolGroup/home
# mkswap /dev/MyVolGroup/swap

Mount your filesystems:

# mount /dev/MyVolGroup/root /mnt
# mkdir /mnt/home
# mount /dev/MyVolGroup/home /mnt/home
# swapon /dev/MyVolGroup/swap

Preparing the boot partition

The bootloader loads the kernel, initramfs, and its own configuration files from the /boot directory. This directory must be located on a separate unencrypted filesystem.

Create a filesystem on the partition intended for /boot. Any filesystem that can be read by the bootloader is eligible.

# mkfs.ext4 /dev/sdb1

Create the directory /mnt/boot:

# mkdir /mnt/boot

Mount the partition to /mnt/boot:

# mount /dev/sdb1 /mnt/boot

Configuring mkinitcpio

Add the keyboard, encrypt and lvm2 hooks to mkinitcpio.conf:

HOOKS=(base udev autodetect keyboard keymap consolefont modconf block encrypt lvm2 filesystems fsck)

If using the sd-encrypt hook with the systemd-based initramfs, the following needs to be set instead:

HOOKS=(base systemd autodetect keyboard sd-vconsole modconf block sd-encrypt sd-lvm2 filesystems fsck)

Note that the systemd hook replaces the default udev hook.

See dm-crypt/System configuration#mkinitcpio for details and other hooks that you may need.

Configuring the boot loader

In order to unlock the encrypted root partition at boot, the following kernel parameter needs to be set by the boot loader:

cryptdevice=UUID=device-UUID:cryptlvm root=/dev/MyVolGroup/root

If using the sd-encrypt hook, the following need to be set instead:

rd.luks.name=device-UUID=cryptlvm root=/dev/MyVolGroup/root

The device-UUID refers to the UUID of /dev/sda1. See Persistent block device naming for details.

See dm-crypt/System configuration#Boot loader for details.

LUKS on LVM

To use encryption on top of LVM, the LVM volumes are set up first and then used as the base for the encrypted partitions. This way, a mixture of encrypted and non-encrypted volumes/partitions is possible as well.

Tip: Unlike #LVM on LUKS, this method allows normally spanning the logical volumes over multiple disks.

The following short example creates a LUKS on LVM setup and mixes in the use of a key-file for the /home partition and temporary crypt volumes for /tmp and /swap. The latter is considered desirable from a security perspective, because no potentially sensitive temporary data survives the reboot, when the encryption is re-initialised. If you are experienced with LVM, you will be able to ignore/replace LVM and other specifics according to your plan.

If you want to span a logical volume over multiple disks that have already been set up, or expand the logical volume for /home (or any other volume), a procedure to do so is described in dm-crypt/Specialties#Expanding LVM on multiple disks. It is important to note that the LUKS encrypted container has to be resized as well.

This article or section needs expansion.

Reason: The intro of this scenario needs some adjustment now that a comparison has been added to #Overview. A suggested structure is to make it similar to the #Simple partition layout with LUKS intro. (Discuss in Talk:Dm-crypt/Encrypting an entire system)

Preparing the disk

Partitioning scheme:

+----------------+-------------------------------------------------------------------------------------------------+
| Boot partition | dm-crypt plain encrypted volume | LUKS2 encrypted volume        | LUKS2 encrypted volume        |
|                |                                 |                               |                               |
| /boot          | [SWAP]                          | /                             | /home                         |
|                |                                 |                               |                               |
|                | /dev/mapper/swap                | /dev/mapper/root              | /dev/mapper/home              |
|                |_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _|_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _|_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _|
|                | Logical volume 1                | Logical volume 2              | Logical volume 3              |
|                | /dev/MyVolGroup/cryptswap       | /dev/MyVolGroup/cryptroot     | /dev/MyVolGroup/crypthome     |
|                |_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _|_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _|_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _|
|                |                                                                                                 |
|   /dev/sda1    |                                   /dev/sda2                                                     |
+----------------+-------------------------------------------------------------------------------------------------+

Randomise /dev/sda2 according to dm-crypt/Drive preparation#dm-crypt wipe on an empty disk or partition.

Preparing the logical volumes

# pvcreate /dev/sda2
# vgcreate MyVolGroup /dev/sda2
# lvcreate -L 32G -n cryptroot MyVolGroup
# lvcreate -L 500M -n cryptswap MyVolGroup
# lvcreate -L 500M -n crypttmp MyVolGroup
# lvcreate -l 100%FREE -n crypthome MyVolGroup
# cryptsetup luksFormat --type luks2 /dev/MyVolGroup/cryptroot
# cryptsetup open /dev/MyVolGroup/cryptroot root
# mkfs.ext4 /dev/mapper/root
# mount /dev/mapper/root /mnt

More information about the encryption options can be found in dm-crypt/Device encryption#Encryption options for LUKS mode. Note that /home will be encrypted in #Encrypting logical volume /home.

Tip: If you ever have to access the encrypted root from the Arch-ISO, the above open action will allow you to after the LVM shows up.

Preparing the boot partition

# dd if=/dev/zero of=/dev/sda1 bs=1M status=progress
# mkfs.ext4 /dev/sda1
# mkdir /mnt/boot
# mount /dev/sda1 /mnt/boot

Configuring mkinitcpio

Add the keyboard, lvm2 and encrypt hooks to mkinitcpio.conf:

HOOKS=(base udev autodetect keyboard keymap consolefont modconf block lvm2 encrypt filesystems fsck)

If using the sd-encrypt hook with the systemd-based initramfs, the following needs to be set instead:

HOOKS=(base systemd autodetect keyboard sd-vconsole modconf block sd-encrypt sd-lvm2 filesystems fsck)

See dm-crypt/System configuration#mkinitcpio for details and other hooks that you may need.

Configuring the boot loader

In order to unlock the encrypted root partition at boot, the following kernel parameters need to be set by the boot loader:

cryptdevice=UUID=device-UUID:root root=/dev/mapper/root

If using the sd-encrypt hook, the following need to be set instead:

rd.luks.name=device-UUID=root root=/dev/mapper/root

The device-UUID refers to the UUID of /dev/MyVolGroup/cryptroot. See Persistent block device naming for details.

See dm-crypt/System configuration#Boot loader for details.

Configuring fstab and crypttab

Both crypttab and fstab entries are required to both unlock the device and mount the filesystems, respectively. The following lines will re-encrypt the temporary filesystems on each reboot:

/etc/crypttab
swap	/dev/MyVolGroup/cryptswap	/dev/urandom	swap,cipher=aes-xts-plain64,size=256
tmp	/dev/MyVolGroup/crypttmp	/dev/urandom	tmp,cipher=aes-xts-plain64,size=256
/etc/fstab
/dev/mapper/root        /       ext4            defaults        0       1
/dev/sda1               /boot   ext4            defaults        0       2
/dev/mapper/tmp         /tmp    tmpfs           defaults        0       0
/dev/mapper/swap        none    swap            sw              0       0

Encrypting logical volume /home

Since this scenario uses LVM as the primary and dm-crypt as secondary mapper, each encrypted logical volume requires its own encryption. Yet, unlike the temporary filesystems configured with volatile encryption above, the logical volume for /home should of course be persistent. The following assumes you have rebooted into the installed system, otherwise you have to adjust paths. To save on entering a second passphrase at boot, a keyfile is created:

# mkdir -m 700 /etc/luks-keys
# dd if=/dev/random of=/etc/luks-keys/home bs=1 count=256 status=progress

The logical volume is encrypted with it:

# cryptsetup luksFormat --type luks2 -v /dev/MyVolGroup/crypthome /etc/luks-keys/home
# cryptsetup -d /etc/luks-keys/home open /dev/MyVolGroup/crypthome home
# mkfs.ext4 /dev/mapper/home
# mount /dev/mapper/home /home

The encrypted mount is configured in both crypttab and fstab:

/etc/crypttab
home	/dev/MyVolGroup/crypthome   /etc/luks-keys/home
/etc/fstab
/dev/mapper/home        /home   ext4        defaults        0       2

LUKS on software RAID

This example is based on a real-world setup for a workstation class laptop equipped with two SSDs of equal size, and an additional HDD for bulk storage. The end result is LUKS based full disk encryption (including /boot) for all drives, with the SSDs in a RAID0 array, and keyfiles used to unlock all encryption after GRUB is given a correct passphrase at boot.

This setup utilizes a very simplistic partitioning scheme, with all the available RAID storage being mounted at / (no separate /boot partition), and the decrypted HDD being mounted at /mnt/data. It is also worth mentioning that the system in this example boots in BIOS mode and the drives are partitioned with GPT partitions.

Please note that regular backups are very important in this setup. If either of the SSDs fail, the data contained in the RAID array will be practically impossible to recover. You may wish to select a different RAID level if fault tolerance is important to you.

The encryption is not deniable in this setup.

For the sake of the instructions below, the following block devices are used:

/dev/sda = first SSD
/dev/sdb = second SSD
/dev/sdc = HDD

Be sure to substitute them with the appropriate device designations for your setup, as they may be different.

Preparing the disks

Prior to creating any partitions, you should inform yourself about the importance and methods to securely erase the disk, described in dm-crypt/Drive preparation.

When using the GRUB bootloader together with GPT, create a BIOS boot partition. For this setup, this includes a 1 MiB partition for BIOS/GPT boot at /dev/sda1 and the remaining space on the drive being partitioned for "Linux RAID" at /dev/sda2.

Once partitions have been created on /dev/sda, the following commands can be used to clone them to /dev/sdb.

# sfdisk -d /dev/sda > sda.dump
# sfdisk /dev/sdb < sda.dump

The HDD is prepared with a single Linux partition covering the whole drive at /dev/sdc1.

Building the RAID array

Create the RAID array for the SSDs. This example utilizes RAID0, you may wish to substitute a different level based on your preferences or requirements.

# mdadm --create --verbose --level=0 --metadata=1.2 --raid-devices=2 /dev/md0 /dev/sda2 /dev/sdb2

Preparing the block devices

As explained in dm-crypt/Drive preparation, the devices are wiped with random data utilizing /dev/zero and a crypt device with a random key. Alternatively, you could use dd with /dev/random or /dev/urandom, though it will be much slower.

# cryptsetup open --type plain /dev/md0 container --key-file /dev/random
# dd if=/dev/zero of=/dev/mapper/container bs=1M status=progress
# cryptsetup close container

And repeat above for the HDD (/dev/sdc1 in this example).

Set up encryption for /dev/md0:

# cryptsetup -y -v luksFormat --type luks2 /dev/md0
# cryptsetup open /dev/md0 cryptroot
# mkfs.ext4 /dev/mapper/cryptroot
# mount /dev/mapper/cryptroot /mnt

And repeat for the HDD:

# cryptsetup -y -v luksFormat --type luks2 /dev/sdc1
# cryptsetup open /dev/sdc1 cryptdata
# mkfs.ext4 /dev/mapper/cryptdata
# mkdir -p /mnt/mnt/data
# mount /dev/mapper/cryptdata /mnt/mnt/data

Configuring the boot loader

Configure GRUB for the encrypted system by editing /etc/default/grub with the following:

GRUB_CMDLINE_LINUX="cryptdevice=/dev/md0:cryptroot root=/dev/mapper/cryptroot"
GRUB_ENABLE_CRYPTODISK=y

See dm-crypt/System configuration#Boot loader and GRUB#Encrypted /boot for details.

Complete the GRUB install to both SSDs (in reality, installing only to /dev/sda will work).

# grub-install --target=i386-pc /dev/sda
# grub-install --target=i386-pc /dev/sdb
# grub-mkconfig -o /boot/grub/grub.cfg

Creating the keyfiles

The next steps save you from entering your passphrase twice when you boot the system (once so GRUB can unlock the encryption, and second time once the initramfs assumes control of the system). This is done by creating a keyfile for the encryption and adding it to the initramfs image to allow the encrypt hook to unlock the root device. See dm-crypt/Device encryption#With a keyfile embedded in the initramfs for details.

Configuring the system

Edit fstab to mount the cryptroot and cryptdata block devices:

/dev/mapper/cryptroot  /           ext4    rw,noatime  0   1 
/dev/mapper/cryptdata  /mnt/data   ext4    defaults            0   2  

Save the RAID configuration:

# mdadm --detail --scan > /etc/mdadm.conf 

Edit mkinitcpio.conf to include your keyfile and add the proper hooks:

FILES=(/crypto_keyfile.bin)
HOOKS=(base udev autodetect keyboard keymap consolefont modconf block mdadm_udev encrypt filesystems fsck)

See dm-crypt/System configuration#mkinitcpio for details.

Plain dm-crypt

Contrary to LUKS, dm-crypt plain mode does not require a header on the encrypted device: this scenario exploits this feature to set up a system on an unpartitioned, encrypted disk that will be indistinguishable from a disk filled with random data, which could allow deniable encryption. See also wikipedia:Disk encryption#Full disk encryption.

Note that if full-disk encryption is not required, the methods using LUKS described in the sections above are better options for both system encryption and encrypted partitions. LUKS features like key management with multiple passphrases/key-files or re-encrypting a device in-place are unavailable with plain mode.

Plain dm-crypt encryption can be more resilient to damage than LUKS, because it does not rely on an encryption master-key which can be a single-point of failure if damaged. However, using plain mode also requires more manual configuration of encryption options to achieve the same cryptographic strength. See also Disk encryption#Cryptographic metadata. Using plain mode could also be considered if concerned with the problems explained in dm-crypt/Specialties#Discard/TRIM support for solid state drives (SSD).

Tip: If headerless encryption is your goal but you are unsure about the lack of key-derivation with plain mode, then two alternatives are:

The scenario uses two USB sticks:

  • one for the boot device, which also allows storing the options required to open/unlock the plain encrypted device in the boot loader configuration, since typing them on each boot would be error prone;
  • another for the encryption key file, assuming it stored as raw bits so that to the eyes of an unaware attacker who might get the usbkey the encryption key will appear as random data instead of being visible as a normal file. See also Wikipedia:Security through obscurity, follow dm-crypt/Device encryption#Keyfiles to prepare the keyfile.

The disk layout is:

+----------------------+----------------------+----------------------+ +----------------+ +----------------+
| Logical volume 1     | Logical volume 2     | Logical volume 3     | | Boot device    | | Encryption key |
|                      |                      |                      | |                | | file storage   |
| /                    | [SWAP]               | /home                | | /boot          | | (unpartitioned |
|                      |                      |                      | |                | | in example)    |
| /dev/MyVolGroup/root | /dev/MyVolGroup/swap | /dev/MyVolGroup/home | | /dev/sdb1      | | /dev/sdc       |
|----------------------+----------------------+----------------------| |----------------| |----------------|
| disk drive /dev/sda encrypted using plain mode and LVM             | | USB stick 1    | | USB stick 2    |
+--------------------------------------------------------------------+ +----------------+ +----------------+
Tip:
  • It is also possible to use a single USB key by copying the keyfile to the initramfs directly. An example keyfile /etc/keyfile gets copied to the initramfs image by setting FILES=(/etc/keyfile) in /etc/mkinitcpio.conf. The way to instruct the encrypt hook to read the keyfile in the initramfs image is using rootfs: prefix before the filename, e.g. cryptkey=rootfs:/etc/keyfile.
  • Another option is using a passphrase with good entropy.

Preparing the disk

It is vital that the mapped device is filled with random data. In particular this applies to the scenario use case we apply here.

See dm-crypt/Drive preparation and dm-crypt/Drive preparation#dm-crypt specific methods

Preparing the non-boot partitions

See dm-crypt/Device encryption#Encryption options for plain mode for details.

Using the device /dev/sda, with the twofish-xts cipher with a 512 bit key size and using a keyfile we have the following options for this scenario:

# cryptsetup --cipher=twofish-xts-plain64 --offset=0 --key-file=/dev/sdc --key-size=512 open --type=plain /dev/sda cryptlvm

Unlike encrypting with LUKS, the above command must be executed in full whenever the mapping needs to be re-established, so it is important to remember the cipher, and key file details.

We can now check a mapping entry has been made for /dev/mapper/cryptlvm:

# fdisk -l
Tip: A simpler alternative to using LVM, advocated in the cryptsetup FAQ for cases where LVM is not necessary, is to just create a filesystem on the entirety of the mapped dm-crypt device.

Next, we setup LVM logical volumes on the mapped device. See LVM#Installing Arch Linux on LVM for further details:

# pvcreate /dev/mapper/cryptlvm
# vgcreate MyVolGroup /dev/mapper/cryptlvm
# lvcreate -L 32G MyVolGroup -n root
# lvcreate -L 10G MyVolGroup -n swap
# lvcreate -l 100%FREE MyVolGroup -n home

We format and mount them and activate swap. See File systems#Create a file system for further details:

# mkfs.ext4 /dev/MyVolGroup/root
# mkfs.ext4 /dev/MyVolGroup/home
# mount /dev/MyVolGroup/root /mnt
# mkdir /mnt/home
# mount /dev/MyVolGroup/home /mnt/home
# mkswap /dev/MyVolGroup/swap
# swapon /dev/MyVolGroup/swap

Preparing the boot partition

The /boot partition can be installed on the standard vfat partition of a USB stick, if required. But if manual partitioning is needed, then a small 200 MiB partition is all that is required. Create the partition using a partitioning tool of your choice.

Create a filesystem on the partition intended for /boot, if it is not already formatted as vfat:

# mkfs.ext4 /dev/sdb1
# mkdir /mnt/boot
# mount /dev/sdb1 /mnt/boot

Configuring mkinitcpio

Add the keyboard, encrypt and lvm2 hooks to mkinitcpio.conf:

HOOKS=(base udev autodetect keyboard keymap consolefont modconf block encrypt lvm2 filesystems fsck)

See dm-crypt/System configuration#mkinitcpio for details and other hooks that you may need.

Configuring the boot loader

In order to boot the encrypted root partition, the following kernel parameters need to be set by the boot loader (note that 64 is the number of bytes in 512 bits):

cryptdevice=/dev/disk/by-id/disk-ID-of-sda:cryptlvm cryptkey=/dev/disk/by-id/disk-ID-of-sdc:0:64 crypto=:twofish-xts-plain64:512:0:

The disk-ID-of-disk refers to the id of the referenced disk. See Persistent block device naming for details.

See dm-crypt/System configuration#Boot loader for details and other parameters that you may need.

Tip: If using GRUB, you can install it on the same USB as the /boot partition with:
# grub-install --recheck /dev/sdb

Post-installation

You may wish to remove the USB sticks after booting. Since the /boot partition is not usually needed, the noauto option can be added to the relevant line in /etc/fstab:

/etc/fstab
# /dev/sdb1
/dev/sdb1 /boot ext4 noauto,rw,noatime 0 2

However, when an update to the kernel or bootloader is required, the /boot partition must be present and mounted. As the entry in fstab already exists, it can be mounted simply with:

# mount /boot

Encrypted boot partition (GRUB)

This setup utilizes the same partition layout and configuration for the system's root partition as the previous #LVM on LUKS section, with the difference that a special feature of the GRUB bootloader is used to additionally encrypt the boot partition /boot. See also GRUB#Encrypted /boot.

The disk layout in this example is:

+---------------------+----------------------+----------------+----------------------+----------------------+----------------------+
| BIOS boot partition | EFI system partition | Boot partition | Logical volume 1     | Logical volume 2     | Logical volume 3     |
|                     |                      |                |                      |                      |                      |
|                     | /efi                 | /boot          | /                    | [SWAP]               | /home                |
|                     |                      |                |                      |                      |                      |
|                     |                      |                | /dev/MyVolGroup/root | /dev/MyVolGroup/swap | /dev/MyVolGroup/home |
| /dev/sda1           | /dev/sda2            | /dev/sda3      +----------------------+----------------------+----------------------+
| unencrypted         | unencrypted          | LUKS encrypted | /dev/sda4 encrypted using LVM on LUKS2                             |
+---------------------+----------------------+----------------+--------------------------------------------------------------------+
Tip:
  • All scenarios are intended as examples. It is, of course, possible to apply both of the two above distinct installation steps with the other scenarios as well. See also the variants linked in #LVM on LUKS.
  • You can use cryptboot script from cryptbootAUR package for simplified encrypted boot management (mounting, unmounting, upgrading packages) and as a defense against Evil Maid attacks with UEFI Secure Boot. For more information and limitations see cryptboot project page.

Preparing the disk

Prior to creating any partitions, you should inform yourself about the importance and methods to securely erase the disk, described in dm-crypt/Drive preparation.

For BIOS systems create a BIOS boot partition with size of 1 MiB for GRUB to store the second stage of BIOS bootloader. Do not mount the partition.

For UEFI systems create an EFI system partition with an appropriate size, it will later be mounted at /efi.

Create a partition to be mounted at /boot of type 8300 with a size of 200 MiB or more.

Create a partition of type 8E00, which will later contain the encrypted container.

Create the LUKS encrypted container at the "system" partition.

# cryptsetup luksFormat --type luks2 /dev/sda4

For more information about the available cryptsetup options see the LUKS encryption options prior to above command.

Your partition layout should look similar to this:

# gdisk /dev/sda
Number  Start (sector)    End (sector)  Size       Code  Name
   1            2048            4095   1024.0 KiB  EF02  BIOS boot partition
   2            4096         1130495   550.0 MiB   EF00  EFI System
   3         1130496         1540095   200.0 MiB   8300  Linux filesystem
   4         1540096        69205982   32.3 GiB    8E00  Linux LVM

Open the container:

# cryptsetup open /dev/sda4 cryptlvm

The decrypted container is now available at /dev/mapper/cryptlvm.

Preparing the logical volumes

The LVM logical volumes of this example follow the exact layout as the #LVM on LUKS scenario. Therefore, please follow #Preparing the logical volumes above and adjust as required.

Preparing the boot partition

Warning: GRUB does not support LUKS2. Do not use LUKS2 on partitions that GRUB needs to access.

The bootloader loads the kernel, initramfs, and its own configuration files from the /boot directory.

First, create the LUKS container where the files will be located and installed into:

# cryptsetup luksFormat /dev/sda3

Next, open it:

# cryptsetup open /dev/sda3 cryptboot

Create a filesystem on the partition intended for /boot. Any filesystem that can be read by the bootloader is eligible:

# mkfs.ext4 /dev/mapper/cryptboot

Create the directory /mnt/boot:

# mkdir /mnt/boot

Mount the partition to /mnt/boot:

# mount /dev/mapper/cryptboot /mnt/boot

Create a mountpoint for the EFI system partition at /efi for compatibility with grub-install and mount it:

# mkdir /mnt/efi
# mount /dev/sda2 /mnt/efi

At this point, you should have the following partitions and logical volumes inside of /mnt:

$ lsblk
NAME                  MAJ:MIN RM   SIZE RO TYPE  MOUNTPOINT
sda                   8:0      0   200G  0 disk
├─sda1                8:1      0     1M  0 part
├─sda2                8:2      0   550M  0 part  /efi
├─sda3                8:3      0   200M  0 part
│ └─cryptboot         254:0    0   198M  0 crypt /boot
└─sda4                8:4      0   100G  0 part
  └─cryptlvm          254:1    0   100G  0 crypt
    ├─MyVolGroup-swap 254:2    0     8G  0 lvm   [SWAP]
    ├─MyVolGroup-root 254:3    0    32G  0 lvm   /
    └─MyVolGroup-home 254:4    0    60G  0 lvm   /home

Configuring mkinitcpio

Add the keyboard, encrypt and lvm2 hooks to mkinitcpio.conf:

HOOKS=(base udev autodetect keyboard keymap consolefont modconf block encrypt lvm2 filesystems fsck)

If using the sd-encrypt hook with the systemd-based initramfs, the following needs to be set instead:

HOOKS=(base systemd autodetect keyboard sd-vconsole modconf block sd-encrypt sd-lvm2 filesystems fsck)

See dm-crypt/System configuration#mkinitcpio for details and other hooks that you may need.

Configuring the boot loader

install GRUB to the mounted ESP for UEFI booting:

# grub-install --target=x86_64-efi --efi-directory=/efi --bootloader-id=GRUB --recheck

install GRUB to the disk for BIOS booting:

# grub-install --target=i386-pc --recheck /dev/sda

Configure GRUB to recognize the LUKS encrypted /boot partition and unlock the encrypted root partition at boot:

/etc/default/grub
GRUB_CMDLINE_LINUX="... cryptdevice=UUID=device-UUID:cryptlvm ..."
GRUB_ENABLE_CRYPTODISK=y

If using the sd-encrypt hook, the following need to be set instead:

/etc/default/grub
GRUB_CMDLINE_LINUX="... rd.luks.name=device-UUID=cryptlvm" ...
GRUB_ENABLE_CRYPTODISK=y

See dm-crypt/System configuration#Boot loader and GRUB#Encrypted /boot for details. The device-UUID refers to the UUID of /dev/sda4 (the partition which holds the lvm containing the root filesystem). See Persistent block device naming.

Generate GRUB's configuration file:

# grub-mkconfig -o /boot/grub/grub.cfg

If all commands finished without errors, GRUB should prompt for the passphrase to unlock the /boot partition after the next reboot.

Configuring fstab and crypttab

This section deals with extra configuration to let the system mount the encrypted /boot.

While GRUB asks for a passphrase to unlock the encrypted /boot after above instructions, the partition unlock is not passed on to the initramfs. Hence, /boot will not be available after the system has re-/booted, because the encrypt hook only unlocks the system's root.

If you used the genfstab script during installation, it will have generated /etc/fstab entries for the /boot and /efi mount points already, but the system will fail to find the generated device mapper for the boot partition. To make it available, add it to crypttab. For example:

/etc/crypttab
cryptboot  /dev/sda3      none        luks

will make the system ask for the passphrase again (i.e. you have to enter it twice at boot: once for GRUB and once for systemd init). To avoid the double entry for unlocking /boot, follow the instructions at dm-crypt/Device encryption#Keyfiles to:

  1. Create a randomtext keyfile,
  2. Add the keyfile to the (/dev/sda3) boot partition's LUKS header and
  3. Check the /etc/fstab entry and add the /etc/crypttab line to unlock it automatically at boot.

If for some reason the keyfile fails to unlock the boot partition, systemd will fallback to ask for a passphrase to unlock and, in case that is correct, continue booting.

Tip: Optional post-installation steps:
  • It may be worth considering to add the GRUB bootloader to the ignore list of /etc/pacman.conf in order to take particular control of when the bootloader (which includes its own encryption modules) is updated.
  • If you want to encrypt the /boot partition to protect against offline tampering threats, the mkinitcpio-chkcryptoboot hook has been contributed to help.

Btrfs subvolumes with swap

The following example creates a full system encryption with LUKS using Btrfs subvolumes to simulate partitions.

If using UEFI, an EFI system partition (ESP) is required. /boot itself may reside on / and be encrypted; however, the ESP itself cannot be encrypted. In this example layout, the ESP is /dev/sda1 and is mounted at /efi. /boot itself is located on the system partition, /dev/sda2.

Since /boot resides on the encrypted /, GRUB must be used as the bootloader because only GRUB can load modules necessary to decrypt /boot (e.g., crypto.mod, cryptodisk.mod and luks.mod) [2].

Additionally an optional plain-encrypted swap partition is shown.

Warning: Do not use a swap file instead of a separate partition, because this may result in data loss. See Btrfs#Swap file.
+----------------------+----------------------+----------------------+
| EFI system partition | System partition     | Swap partition       |
| unencrypted          | LUKS2-encrypted      | plain-encrypted      |
|                      |                      |                      |
| /efi                 | /                    | [SWAP]               |
| /dev/sda1            | /dev/sda2            | /dev/sda3            |
|----------------------+----------------------+----------------------+

Preparing the disk

Note: It is not possible to use btrfs partitioning as described in Btrfs#Partitionless Btrfs disk when using LUKS. Traditional partitioning must be used, even if it is just to create one partition.

Prior to creating any partitions, you should inform yourself about the importance and methods to securely erase the disk, described in dm-crypt/Drive preparation. If you are using UEFI create an EFI system partition with an appropriate size. It will later be mounted at /efi. If you are going to create an encrypted swap partition, create the partition for it, but do not mark it as swap, since plain dm-crypt will be used with the partition.

Create the needed partitions, at least one for / (e.g. /dev/sda2). See the Partitioning article.

Preparing the system partition

Create LUKS container

Follow dm-crypt/Device encryption#Encrypting devices with LUKS mode to setup /dev/sda2 for LUKS. See the dm-crypt/Device encryption#Encryption options for LUKS mode before doing so for a list of encryption options.

Unlock LUKS container

Now follow dm-crypt/Device encryption#Unlocking/Mapping LUKS partitions with the device mapper to unlock the LUKS container and map it.

Format mapped device

Proceed to format the mapped device as described in Btrfs#File system on a single device, where /dev/partition is the name of the mapped device (i.e., cryptroot) and not /dev/sda2.

Mount mapped device

Finally, mount the now-formatted mapped device (i.e., /dev/mapper/cryptroot) to /mnt.

Tip: You may want to use the compress=lzo mount option. See Btrfs#Compression for more information.

Creating btrfs subvolumes

This article or section is a candidate for merging with Btrfs.

Notes: The subvolume layout is not specific to an encrypted system. (Discuss in Talk:Dm-crypt/Encrypting an entire system)

Layout

Subvolumes will be used to simulate partitions, but other (nested) subvolumes will also be created. Here is a partial representation of what the following example will generate:

subvolid=5 (/dev/sda2)
   |
   ├── @ (mounted as /)
   |       |
   |       ├── /bin (directory)
   |       |
   |       ├── /home (mounted @home subvolume)
   |       |
   |       ├── /usr (directory)
   |       |
   |       ├── /.snapshots (mounted @snapshots subvolume)
   |       |
   |       ├── /var/cache/pacman/pkg (nested subvolume)
   |       |
   |       ├── ... (other directories and nested subvolumes)
   |
   ├── @snapshots (mounted as /.snapshots)
   |
   ├── @home (mounted as /home)
   |
   └── @... (additional subvolumes you wish to use as mount points)

This section follows the Snapper#Suggested filesystem layout, which is most useful when used with Snapper. You should also consult Btrfs Wiki SysadminGuide#Layout.

Create top-level subvolumes

Here we are using the convention of prefixing @ to subvolume names that will be used as mount points, and @ will be the subvolume that is mounted as /.

Following the Btrfs#Creating a subvolume article, create subvolumes at /mnt/@, /mnt/@snapshots, and /mnt/@home.

Create any additional subvolumes you wish to use as mount points now.

Mount top-level subvolumes

Unmount the system partition at /mnt.

Now mount the newly created @ subvolume which will serve as / to /mnt using the subvol= mount option. Assuming the mapped device is named cryptroot, the command would look like:

# mount -o compress=lzo,subvol=@ /dev/mapper/cryptroot /mnt

See Btrfs#Mounting subvolumes for more details.

Also mount the other subvolumes to their respective mount points: @home to /mnt/home and @snapshots to /mnt/.snapshots.

Create nested subvolumes

Create any subvolumes you do not want to have snapshots of when taking a snapshot of /. For example, you probably do not want to take snapshots of /var/cache/pacman/pkg. These subvolumes will be nested under the @ subvolume, but just as easily could have been created earlier at the same level as @ according to your preference.

Since the @ subvolume is mounted at /mnt you will need to create a subvolume at /mnt/var/cache/pacman/pkg for this example. You may have to create any parent directories first.

Other directories you may wish to do this with are /var/abs, /var/tmp, and /srv.

Mount ESP

If you prepared an EFI system partition earlier, create its mount point and mount it now.

Note: Btrfs snapshots will exclude /efi, since it is not a btrfs file system.

At the pacstrap installation step, the btrfs-progs must be installed in addition to the base group.

Configuring mkinitcpio

Create keyfile

In order for GRUB to open the LUKS partition without having the user enter his passphrase twice, we will use a keyfile embedded in the initramfs. Follow dm-crypt/Device encryption#With a keyfile embedded in the initramfs making sure to add the key to /dev/sda2 at the luksAddKey step.

Edit mkinitcpio.conf

After creating, adding, and embedding the key as described above, add the encrypt hook to mkinitcpio.conf as well as any other hooks you require. See dm-crypt/System configuration#mkinitcpio for detailed information.

Tip: You may want to add BINARIES=(/usr/bin/btrfs) to your mkinitcpio.conf. See the Btrfs#Corruption recovery article.

Configuring the boot loader

Install GRUB to /dev/sda. Then, edit /etc/default/grub as instructed in the GRUB#Additional arguments and GRUB#Encrypted /boot, following both the instructions for an encrypted root and boot partition. Finally, generate the GRUB configuration file.

Configuring swap

If you created a partition to be used for encrypted swap, now is the time to configure it. Follow the instructions at dm-crypt/Swap encryption.