How to Install Fedora 38 with Full Disk Encryption, Snapshot and Rollback Support

In this post, I will show you how to install Fedora 38 with full disk encryption, snapshot and rollback support. This allows you to take snapshots of the / and /home subvolumes and perform a full system rollback on the / volume if something goes wrong.

In my previous post on how to install Fedora 37 with LUKS full disk encryption, I used Fedora's default Anaconda installer. The Anaconda installer does not support encrypting the /boot directory. As a result, I had to first create a separate ext4 /boot partition, then after installation, move the contents of the /boot partition to the / volume, then delete the /boot partition, and finally merge the deleted space to the / volume. These are far too many steps.

So, instead of using the standard GUI Anaconda installer, I'll install Fedora via the command line interface with DNF installroot and chroot environment. This will be straightforward, with no unnecessary steps.

In terms of btrfs snapshot and rollback support, nothing has changed in Fedora 38. The DNF database is still in /var/lib/dnf and has not been moved to /usr.

I've also made some other changes in this updated guide. The changes are as follows:

Created new subvolumes on the system root to keep some directories out of root file system snapshots. Some of these directories contain temporary files and caches, others contain data that you do not want to lose if you roll back the system root, and some should be in Read-Write mode when booting into a Read-only snapshot.
Created one subvolume in the user home for the Firefox browser. This prevents you from losing bookmarks, passwords, and other user data when you undo a change to the user's home.
TPM2 support has been added for automatic disk decryption. You can now use either TPM2 or a key file to automatically decrypt your disk with the LUKS passphrase only being asked once. I provided both options.
Turned on automatic timeline snapshots for the system root in snapper. You can, however, turn off the timeline snapshots.
Moved grub-btrfs from the GitHub archived 'Add-systemd-volatile' branch to the active master branch.
Created a snapshot of the system root and set it as the default root file system. This makes managing snapshots and rollbacks a lot easier.

Here's a quick rundown of what needs to be done:

Boot from the Fedora 38 Workstation Live installation media. Create the necessary partitions on the disk where Fedora 38 will be installed.
Enable LUKS encryption on the partition where Fedora will be installed, and then format all the partitions with the appropriate Linux file systems.
Create and mount all necessary directories and subvolumes for Fedora installation in the /mnt directory.
Install a minimal Fedora distribution in the /mnt directory.
Finish the minimal installation by creating all of the required configuration files.
Boot from Fedora Minimal and configure TPM2 or a key file to automatically decrypt your disk.
Install the GNOME Desktop Environment.
Create any additional subvolumes that are required based on your desktop environment choice.
Install the snapper and grub-btrfs packages to enable snapshot and rollback support. Then complete the installation by creating a default subvolume for the system root to simplify snapshot and rollback management.

If you want to install Fedora 38 without LUKS Full Disk Encryption but still have Full System Rollback, please see my other article, 'How to Install Fedora 38 with Snapshot and Rollback Support'.

Although it appears to be overwhelming, it is not. I'll walk you through every step of the installation process. So let's get started.

Table of Contents

1. Disk and Subvolumes Layout for Fedora 38

I'll be using a hard disk with a capacity of 100 GiB. I'll create one 512 MiB EFI partition and use the rest of the disk space to create a btrfs volume. There is no need for a separate swap partition because Fedora creates a SwapOnZRAM during boot.

This is how the disk partition looks.

NAME           SIZE   FSTYPE  LABEL    MOUNTPOINT
/dev/vda       100G
├─/dev/vda1    512M   vfat    EFI      /boot/efi
└─/dev/vda2   99.5G   btrfs   FEDORA   /

The system root will be installed on the LUKS encrypted btrfs main volume (/dev/vda2). I'll then create the following subvolumes on the system root to keep them out of root file system snapshots.

NAME              MOUNTPOINT                 TYPE
[main]            /                          mainvolume
home              /home                      subvolume
opt               /opt                       subvolume
cache             /var/cache                 subvolume
crash             /var/crash                 subvolume
log               /var/log                   subvolume
spool             /var/spool                 subvolume
tmp               /var/tmp                   subvolume
www               /var/www                   subvolume
AccountsService   /var/lib/AccountsService   subvolume
gdm               /var/lib/gdm               subvolume
images            /var/lib/libvirt/images    subvolume
.mozilla          /home/<USER>/.mozilla      subvolume

The directories for which subvolumes are created and the reasons for doing so are listed below.

/home

Contains user data. It is created to keep user data separate and also to avoid loss of user data on system root rollbacks. To learn more about /home, check this link: /home.

/opt

Contains applications installed by a third party. It is excluded to prevent these applications from being uninstalled during rollbacks. To learn more about /opt, check this link: /opt.

/var/cache, /var/crash, /var/tmp

Directories contain temporary files and caches. So they are created to be excluded from snapshots. Fedora also mounts /tmp with tmpfs at boot, so a separate subvolume for /tmp is unnecessary. To learn more, check these links: /var/cache, /var/crash, /var/tmp, /tmp.

/var/log

Contains log files. It is created to avoid the loss of log data during rollbacks. To learn more about /var/log, check this link: /var/log.

/var/spool

Contains data that is awaiting some kind of later processing such as mail, mail queues, printing, printer spool, and so on. Created to avoid a loss of mail, printing, and spool data after a rollback. To learn more about /var/spool, check this link: /var/spool.

/var/www

Web server directory. It is created to keep hosted web server data separate and to avoid data loss on system root rollbacks.

/var/lib/AccountsService, /var/lib/gdm

Contains login user information as well as Gnome display information. These directories must always be writable. When you attempt to boot a snapshot from the GRUB menu, you are booting into a read-only snapshot, which causes the system to hang just before the Gnome login screen appears.

In my previous guides, I enabled the 'systemd.volatile=state' kernel parameter, which should have solved this problem, but some of my readers reported that they couldn't log in. As a result, I created subvolumes for these two directories so that they are always writable and thus permanently solve this problem.

If you are using a display manager other than Gnome, you must create a directory specific to that display manager. For XFCE, it might be /var/lib/lightdm. I haven't tested.

/var/lib/libvirt/images

The default location for libvirt-managed virtual machine images. Excluded to avoid replacing virtual machine images with older versions during a rollback.

/home/<USER>/.mozilla

By default, Fedora Workstation includes the Firefox browser. So, if you take a snapshot of your home directory and later undo any changes, you don't want to lose any new bookmarks, passwords, and other user data.

As a result, I created a separate subvolume for the '.mozilla' directory, keeping it separate from the /home subvolume and avoiding data loss during any undo changes.

Create a subvolume for the '/home/<USER>/.config/google-chrome' directory if you use the Chrome web browser.

2. Create Partitions and File Systems

Boot your system using the Fedora 38 Workstation Live installer in UEFI mode. In the 'Welcome to Fedora' window, click the 'Not Now' button to return to the Gnome desktop. Next, launch Gnome Terminal to begin preparing the system for Fedora 38 installation.

Locate the disk on which you want to install Fedora. For me, it's /dev/vda. Yours might be different.

$ sudo -i

# lsblk -pd
NAME       MAJ:MIN RM  SIZE RO TYPE MOUNTPOINTS
/dev/loop0   7:0    0  1.9G  1 loop 
/dev/loop1   7:1    0  7.6G  1 loop 
/dev/loop2   7:2    0   32G  0 loop 
/dev/sr0    11:0    1    2G  0 rom  /run/initramfs/live
/dev/zram0 251:0    0  3.8G  0 disk [SWAP]
/dev/vda   252:0    0  100G  0 disk

Create an EFI System Partition of 512 MiB. I'll name the partition 'EFI', but you can give it any name.

# sgdisk -n 1::+512M -t 1:ef00 -c 1:'EFI' /dev/vda

Create a Linux filesystem partition on which Fedora will be installed. I will use the entire remaining space. I'll name this partition 'Fedora 38', but you can give it any name.

# sgdisk -n 2:: -t 2:8300 -c 2:'Fedora 38' /dev/vda

List the partitions you just created to ensure that everything is in order.

# sgdisk -p /dev/vda
Disk /dev/vda: 209715200 sectors, 100.0 GiB
...
Total free space is 2014 sectors (1007.0 KiB)

Number  Start (sector)    End (sector)  Size       Code  Name
   1            2048         1050623   512.0 MiB   EF00  EFI
   2         1050624       209715166   99.5 GiB    8300  Fedora 38

Format the /dev/vda1 partition with the FAT32 file system. I'll name the volume 'EFI', but you may name it whatever you want.

# mkfs.fat -F 32 -n EFI /dev/vda1

Initialize the /dev/vda2 partition as a LUKS device and set the initial passphrase.

# cryptsetup luksFormat --type luks2 --label FEDORA_LUKS \
    --pbkdf pbkdf2 --pbkdf-force-iterations 500000 /dev/vda2

WARNING!
========
This will overwrite data on /dev/vda2 irrevocably.

Are you sure? (Type 'yes' in capital letters): YES
Enter passphrase for /dev/vda2: 
Verify passphrase:

GRUB currently only supports pbkdf2 algorithm. For pbkdf2, the minimum iteration count is 1000 and the maximum is 4294967295. The more powerful your CPU, the higher the iteration count. And, the higher the iteration count, the longer it takes to unlock the LUKS in the GRUB. So I have set the iteration count to 500000. Lowering the iterations reduces security. To get around it, use a stronger passphrase with higher entropy (password entropy > 60 bits is suggested).

Now, open the LUKS device /dev/vda2 and map it to the temporary name tmpluks. I'll map the LUKS device with the proper UUID later. When prompted, enter the LUKS passphrase you provided.

# cryptsetup luksOpen /dev/vda2 tmpluks
Enter passphrase for /dev/vda2:

Format the mapped device with the btrfs file system. I'll name the volume FEDORA, but you may name it whatever you want.

# mkfs.btrfs -L FEDORA /dev/mapper/tmpluks

This is how your setup should look. The UUIDs will be unique to your system.

# udevadm trigger

# lsblk -p -o name,fstype,fsver,label,uuid /dev/vda
NAME                    FSTYPE      FSVER LABEL       UUID
/dev/vda                                              
├─/dev/vda1             vfat        FAT32 EFI         AFE6-47F9
└─/dev/vda2             crypto_LUKS 2     FEDORA_LUKS 6e6c2ee3-852a-4efb-9801-8277dccf0af1
  └─/dev/mapper/tmpluks btrfs             FEDORA      85826240-0537-4ad0-a2ee-c59166a18f7c

3. Create Directories and Subvolumes

First, mount /dev/mapper/tmpluks to the /mnt directory.

# mount -v -o compress=zstd:1 /dev/mapper/tmpluks /mnt
# restorecon -RFv /mnt

Then, create the necessary directories and subvolumes. As previously stated, these subvolumes are created on the system root to avoid being included in the rollback regime because some of these directories contain temporary files and caches, others contain data that you do not want to lose if you rollback the system root, and some should be in Read-Write mode when booting into a Read-only snapshot.

# mkdir -vp /mnt/boot/efi
# mkdir -vp /mnt/var/lib/libvirt

# mount -v /dev/vda1 /mnt/boot/efi

# btrfs subvolume create /mnt/home
# btrfs subvolume create /mnt/opt

# btrfs subvolume create /mnt/var/cache
# btrfs subvolume create /mnt/var/crash
# btrfs subvolume create /mnt/var/log
# btrfs subvolume create /mnt/var/spool
# btrfs subvolume create /mnt/var/tmp
# btrfs subvolume create /mnt/var/www

# btrfs subvolume create /mnt/var/lib/libvirt/images

Change the permissions of the directory /mnt/var/tmp.

# chmod 1777 /mnt/var/tmp

# ls -lhd /mnt/var/tmp
drwxrwxrwt. 1 root root 0 May  3 20:14 /mnt/var/tmp

When using QCOW2 images as virtual disks in KVM, the IO performance suffers. So disable the CoW feature in the /mnt/var/lib/libvirt/images directory.

# chattr -R +C /mnt/var/lib/libvirt/images

# lsattr -d /mnt/var/lib/libvirt/images
---------------C------ /mnt/var/lib/libvirt/images

List the subvolumes. This is how it should look.

# btrfs subvolume list /mnt
ID 256 gen 10 top level 5 path home
ID 257 gen 11 top level 5 path opt
ID 258 gen 12 top level 5 path var/cache
ID 259 gen 13 top level 5 path var/crash
ID 260 gen 14 top level 5 path var/log
ID 261 gen 15 top level 5 path var/spool
ID 262 gen 18 top level 5 path var/tmp
ID 263 gen 19 top level 5 path var/lib/libvirt/images

I'll create the remaining subvolumes when I install the desktop environment.

4. Bootstrap a Fedora System in a Chroot Environment

Create a Fedora system in the /mnt directory using DNF installroot. I'm going to install the basesystem package first, which defines the basic elements of a Fedora system.

# dnf --releasever=38 --installroot=/mnt install basesystem

Bind mount the directories /sys, /dev, and /proc into /mnt. Copy from 'for' to 'done', paste it into the terminal, and hit [Enter].

# for dir in sys dev proc ; do
    mount -v --rbind "/${dir}" "/mnt/${dir}" 
    mount -v --make-rslave "/mnt/${dir}"
done

Now that the directories have been bind mounted, install Fedora Minimal in the /mnt directory. I've included Fedora Minimal package groups, kernel, grub, and some other relevant packages.

# dnf --releasever=38 --installroot=/mnt install @core @standard \
    @networkmanager-submodules @guest-agents \
    kernel langpacks-en glibc-langpack-en glibc-all-langpacks \
    grub2-efi-x64 grub2-efi-x64-modules grub2-tools grub2-common shim \
    vim git inotify-tools make tpm2-tss finger

Copy the resolv.conf file from the host to the /mnt/etc/ directory so that you can connect to the internet. But first, you must delete the existing symbolic link to the resolv.conf file.

# rm -vf /mnt/etc/resolv.conf
# cp -v /etc/resolv.conf /mnt/etc/

Set the timezone, locale, keymap, hostname, and machine-id.

# systemd-firstboot --root=/mnt \
    --timezone=Asia/Kolkata \
    --locale=en_IN.UTF-8 \
    --keymap=in-eng \
    --hostname=fedora \
    --setup-machine-id

I set the time zone to India and the keyboard language to India English. Change the setting according to your location. I've also set the hostname to 'fedora', but you can name it whatever you want.

Check the following directory to determine your time zone.

# ls /usr/share/zoneinfo

To determine your locale, run the following command.

# localectl list-locales

To determine your keymap, run the following command.

# localectl list-keymaps

Finally, chroot into the /mnt directory.

# chroot /mnt /bin/bash

5. Create a User Account

Now that you're in a chroot environment, reinstall the polkit package before creating a user account so that the polkit's scripts are properly regenerated.

Enable the fastest mirrors to speed up package downloads. Copy the command from cat to EOF, paste it into your terminal, and hit [Enter].

# cat <<'EOF' >> /etc/dnf/dnf.conf
defaultyes=True
fastestmirror=True
max_parallel_downloads=10
EOF

Clear all previous caches and update the local DNF metadata cache.

# dnf clean all
# dnf makecache

Reinstall the polkit package.

# dnf -y reinstall polkit

After reinstalling the polkit package, proceed to create a user account.

A user account is required for logging into the Fedora Linux operating system. I'll create a user called 'madhu' with root privileges. Change the user name to something more specific to you.

# useradd madhu --comment 'Madhu Desai' --create-home --groups wheel \
    --shell /bin/bash

Set the user password.

# passwd madhu

Ensure that a user account has been created with the correct information

# finger madhu
Login: madhu          			Name: Madhu Desai
Directory: /home/madhu              	Shell: /bin/bash
Never logged in.
No mail.
No Plan.

6. Create the /etc/fstab File

You need to create /etc/fstab file so that the Fedora operating system will boot without any errors.

Set the btrfs subvolume mount options.

# OPTIONS="compress=zstd:1,x-systemd.device-timeout=0"

Get the UUID of your EFI partition.

# EFI_UUID="$(grub2-probe --target=fs_uuid /boot/efi)"

Get the UUID of your btrfs system root.

# ROOT_UUID="$(grub2-probe --target=fs_uuid /)"

Get your subvolmes.

# SUBVOLUMES=($(btrfs subvolume list / | awk '!/machines|portables/ {print $9}' | xargs))

Print system root mount in /etc/fstab file.

# printf "%-41s %-24s %-5s %-s %-s\n" \
    "UUID=${ROOT_UUID}" \
    "/" \
    "btrfs" \
    "defaults,${OPTIONS}" \
    "0 0" >> /etc/fstab

Print EFI partition mount in /etc/fstab file.

# printf "%-41s %-24s %-5s %-s %-s\n" \
    "UUID=${EFI_UUID}" \
    "/boot/efi" \
    "vfat" \
    "umask=0077,shortname=winnt" \
    "0 0" >> /etc/fstab

Print the rest of the subvolume mounts in /etc/fstab file. Copy from 'for' to 'done', paste it into the terminal, and hit [Enter].

# for dir in "${SUBVOLUMES[@]}" ; do
    printf "%-41s %-24s %-5s %-s %-s\n" \
    "UUID=${ROOT_UUID}" \
    "/${dir}" \
    "btrfs" \
    "subvol=${dir},${OPTIONS}" \
    "0 0" >> /etc/fstab
    restorecon -RF "/${dir}"
done

Examine your /etc/fstab file. This is what it should look like. The UUID and username will be unique to your system.

# cat /etc/fstab
UUID=85826240-0537-4ad0-a2ee-c59166a18f7c /                        btrfs defaults,compress=zstd:1,x-systemd.device-timeout=0 0 0
UUID=AFE6-47F9                            /boot/efi                vfat  umask=0077,shortname=winnt 0 0
UUID=85826240-0537-4ad0-a2ee-c59166a18f7c /home                    btrfs subvol=home,compress=zstd:1,x-systemd.device-timeout=0 0 0
UUID=85826240-0537-4ad0-a2ee-c59166a18f7c /opt                     btrfs subvol=opt,compress=zstd:1,x-systemd.device-timeout=0 0 0
UUID=85826240-0537-4ad0-a2ee-c59166a18f7c /var/cache               btrfs subvol=var/cache,compress=zstd:1,x-systemd.device-timeout=0 0 0
UUID=85826240-0537-4ad0-a2ee-c59166a18f7c /var/crash               btrfs subvol=var/crash,compress=zstd:1,x-systemd.device-timeout=0 0 0
UUID=85826240-0537-4ad0-a2ee-c59166a18f7c /var/log                 btrfs subvol=var/log,compress=zstd:1,x-systemd.device-timeout=0 0 0
UUID=85826240-0537-4ad0-a2ee-c59166a18f7c /var/spool               btrfs subvol=var/spool,compress=zstd:1,x-systemd.device-timeout=0 0 0
UUID=85826240-0537-4ad0-a2ee-c59166a18f7c /var/tmp                 btrfs subvol=var/tmp,compress=zstd:1,x-systemd.device-timeout=0 0 0
UUID=85826240-0537-4ad0-a2ee-c59166a18f7c /var/lib/libvirt/images  btrfs subvol=var/lib/libvirt/images,compress=zstd:1,x-systemd.device-timeout=0 0 0

Mount all of the subvolumes listed in the /etc/fstab file.

# mount -va
/                        : ignored
/boot/efi                : already mounted
/home                    : successfully mounted
/opt                     : successfully mounted
/var/cache               : successfully mounted
/var/crash               : successfully mounted
/var/log                 : successfully mounted
/var/spool               : successfully mounted
/var/tmp                 : successfully mounted
/var/lib/libvirt/images  : successfully mounted

Check your subvolumes.

# btrfs subvolume list /
ID 256 gen 72 top level 5 path home
ID 257 gen 72 top level 5 path opt
ID 258 gen 72 top level 5 path var/cache
ID 259 gen 72 top level 5 path var/crash
ID 260 gen 72 top level 5 path var/log
ID 261 gen 72 top level 5 path var/spool
ID 262 gen 72 top level 5 path var/tmp
ID 263 gen 72 top level 5 path var/lib/libvirt/images
ID 264 gen 43 top level 5 path var/lib/portables

Your setup should now look somewhat like this.

# lsblk -p /dev/vda
NAME                    MAJ:MIN RM  SIZE RO TYPE  MOUNTPOINTS
/dev/vda                252:0    0  100G  0 disk  
├─/dev/vda1             252:1    0  512M  0 part  /boot/efi
└─/dev/vda2             252:2    0 99.5G  0 part  
  └─/dev/mapper/tmpluks 253:2    0 99.5G  0 crypt /var/lib/libvirt/images
                                                  /var/tmp
                                                  /var/spool
                                                  /var/log
                                                  /var/crash
                                                  /var/cache
                                                  /opt
                                                  /home
                                                  /

7. Enable Crypto Disk and Snapshot Booting

Enable crypto disk and snapshot booting to allow encrypted /boot and snapper snapshots to boot directly from GRUB.

When you enable the GRUB_ENABLE_CRYPTODISK option in GRUB, it will check for encrypted disks and generate the extra commands needed to access them during boot.

With the SUSE_BTRFS_SNAPSHOT_BOOTING option in GRUB enabled, snapper snapshots can boot directly from GRUB. This is also useful when rolling back a snapshot.

Get the UUID of your LUKS partition.

# LUKS_DEV_UUID="$(cryptsetup luksUUID /dev/vda2)"

# echo $LUKS_DEV_UUID 
6e6c2ee3-852a-4efb-9801-8277dccf0af1

First, create a /etc/default/grub file with the UUID of your LUKS partition in it.

Copy the command from cat to EOF, paste it into your terminal, and hit [Enter].

# cat <<EOF > /etc/default/grub
GRUB_TIMEOUT=5
GRUB_DISTRIBUTOR="\$(sed 's, release .*$,,g' /etc/system-release)"
GRUB_DEFAULT=saved
GRUB_DISABLE_SUBMENU=true
GRUB_TERMINAL_OUTPUT="console"
GRUB_CMDLINE_LINUX="rd.luks.uuid=luks-$LUKS_DEV_UUID quiet"
GRUB_DISABLE_RECOVERY="true"
GRUB_ENABLE_BLSCFG=true
GRUB_ENABLE_CRYPTODISK=y
SUSE_BTRFS_SNAPSHOT_BOOTING="true"
EOF

This is how it should look. The UUID will be unique to your system.

# cat /etc/default/grub 
GRUB_TIMEOUT=5
GRUB_DISTRIBUTOR="$(sed 's, release .*$,,g' /etc/system-release)"
GRUB_DEFAULT=saved
GRUB_DISABLE_SUBMENU=true
GRUB_TERMINAL_OUTPUT="console"
GRUB_CMDLINE_LINUX="rd.luks.uuid=luks-6e6c2ee3-852a-4efb-9801-8277dccf0af1 quiet"
GRUB_DISABLE_RECOVERY="true"
GRUB_ENABLE_BLSCFG=true
GRUB_ENABLE_CRYPTODISK=y
SUSE_BTRFS_SNAPSHOT_BOOTING="true"

You must also modify the /boot/efi/EFI/fedora/grub.cfg file now that crypto disk and snapshot booting are enabled.

# sed -i "1i cryptomount -u ${LUKS_DEV_UUID//-/}" /boot/efi/EFI/fedora/grub.cfg
# sed -i '1i set btrfs_relative_path="yes"' /boot/efi/EFI/fedora/grub.cfg

This is how it should appear. The UUID in the cryptomount line is that of the /dev/vda2 partition, with no dashes (-).

# cat /boot/efi/EFI/fedora/grub.cfg
set btrfs_relative_path="yes"
cryptomount -u 6e6c2ee3852a4efb98018277dccf0af1
search --no-floppy --fs-uuid --set=dev 85826240-0537-4ad0-a2ee-c59166a18f7c
set prefix=($dev)/boot/grub2
export $prefix
configfile $prefix/grub.cfg

Then, update the grub.cfg file.

# grub2-mkconfig -o /boot/grub2/grub.cfg

This is how the GRUB BLS should appear.

# grubby --info=DEFAULT
index=0
kernel="/boot/vmlinuz-6.2.14-300.fc38.x86_64"
args="ro rd.luks.uuid=luks-6e6c2ee3-852a-4efb-9801-8277dccf0af1 quiet ${extra_cmdline}"
root="UUID=85826240-0537-4ad0-a2ee-c59166a18f7c"
initrd="/boot/initramfs-6.2.14-300.fc38.x86_64.img"
title="Fedora Linux (6.2.14-300.fc38.x86_64) 38 (Thirty Eight)"
id="7acd6f0f07714ec3b203b7a75f495fed-6.2.14-300.fc38.x86_64"

Add the encrypted block device to the /etc/crypttab file so that it is enabled when the system boots.

# echo "luks-${LUKS_DEV_UUID} UUID=${LUKS_DEV_UUID} none discard" >> /etc/crypttab

# cat /etc/crypttab
luks-6e6c2ee3-852a-4efb-9801-8277dccf0af1 UUID=6e6c2ee3-852a-4efb-9801-8277dccf0af1 none discard

Enable NetworkManager and auditd systemd services.

# systemctl enable NetworkManager
# systemctl enable auditd

Execute the following command so that the SELinux contexts are relabeled on the next boot.

# fixfiles onboot

Now exit the chroot environment, unmount all mounts, and restart your computer.

# exit
# umount -vR /mnt
# reboot

The system will reboot twice. The first time it boots, it relabels SELinux contexts, and the second time it logs in to the operating system.

You will also be prompted for the passphrase twice. The first time is before loading GRUB, and the second time is before mounting the root file system. In the following section, I will enable the single passphrase prompt.

Depending on your computer's CPU, it may take up to 15 seconds for the GRUB to display the menu. So please be patient.

When you input your passphrase in the first LUKS prompt, no "*" cursor indication appears. So you should assume you have entered your passphrase correctly. Simply enter your LUKS passphrase blindly and hit the [Enter] key.

The Fedora Minimal installation is now complete.

You may now remove the Fedora Live installation media. From now on, the configuration will be performed on Fedora 38 Minimal itself.

8. Bypass the Additional Passphrase Prompt

While GRUB requests a passphrase in order to decrypt the encrypted /boot files, this information is not passed on to initramfs. As a result, the root (/) must be unlocked once more during the initramfs stage. This means that the user must enter his passphrase twice.

I'll show you two methods for automatically decrypting your disk without requiring a passphrase: using a key file or TPM2. Choose your preferred method.

8.1 Using a Key File

If you want to decrypt your LUKS-encrypted volumes using a key file, use this method. SecureBoot does not need to be enabled for this.

Create the /etc/cryptsetup-keys.d directory if it does not already exist.

# mkdir -v /etc/cryptsetup-keys.d

Create a key file with 4KiB of random data. The key file has to be in volume.key format.

# dd if=/dev/random \
   of=/etc/cryptsetup-keys.d/luks-$(cryptsetup luksUUID /dev/vda2).key \
   bs=512 count=8

Ensure that only the root user has read access to the key file.

# chmod 0400 /etc/cryptsetup-keys.d/*.key

Attach the created key to the encrypted device’s available key slot. When prompted for a passphrase, enter the same LUKS passphrase.

# cryptsetup luksAddKey \
    --pbkdf pbkdf2 \
    --pbkdf-force-iterations 500000 \
    /dev/vda2 \
    /etc/cryptsetup-keys.d/luks-$(cryptsetup luksUUID /dev/vda2).key

Add the key to /etc/dracut.conf.d/cryptodisk.conf file to include it in the initramfs image.

# echo 'install_items+=" /etc/cryptsetup-keys.d/* "' \
    >> /etc/dracut.conf.d/cryptodisk.conf

Rebuild the initramfs image.

# dracut --verbose --force

Reboot the computer to ensure that everything is working properly. The LUKS passphrase should only be asked once before the GRUB menu appears.

# reboot

8.2 Using TPM2 Method

This method allows you to decrypt your LUKS-encrypted volumes using the TPM on your computer. You must have SecureBoot enabled for this method to work.

First, make sure SecureBoot is enabled.

# mokutil --sb-state
SecureBoot enabled

If it is disabled, you must enable it in the BIOS.

Next, see if your computer supports TPM.

# systemd-cryptenroll --tpm2-device=list
PATH        DEVICE      DRIVER 
/dev/tpmrm0 MSFT0101:00 tpm_crb

As you can see from the output, my TPM2 device node path is /dev/tpmrm0.

Now enroll your encrypted volume. When prompted for a passphrase, enter the same LUKS passphrase.

# systemd-cryptenroll /dev/vda2 --tpm2-device=/dev/tpmrm0
🔐 Please enter current passphrase for disk /dev/vda2: ***********             
New TPM2 token enrolled as key slot 1.

According to the systemd-cryptenroll man page, it should be sufficient for most applications to bind against PCR 7, which is also the default. For more information, see man systemd-cryptenroll.

Add the TPM2 device node path to the kernel command line parameter 'rd.luks.options='.

# grubby --update-kernel=ALL \
    --args="rd.luks.options=tpm2-device=/dev/tpmrm0"

Check to see if the TPM2 device node path has been added to the grub kernel parameter.

# grubby --info=DEFAULT
index=0
kernel="/boot/vmlinuz-6.2.14-300.fc38.x86_64"
args="ro rd.luks.uuid=luks-6e6c2ee3-852a-4efb-9801-8277dccf0af1 quiet ${extra_cmdline} rd.luks.options=tpm2-device=/dev/tpmrm0"
root="UUID=85826240-0537-4ad0-a2ee-c59166a18f7c"
initrd="/boot/initramfs-6.2.14-300.fc38.x86_64.img"
title="Fedora Linux (6.2.14-300.fc38.x86_64) 38 (Thirty Eight)"
id="7acd6f0f07714ec3b203b7a75f495fed-6.2.14-300.fc38.x86_64"

Changes should also be reflected in the /etc/crypttab file.

# LUKS_DEV_UUID="$(cryptsetup luksUUID /dev/vda2)"

# echo "luks-${LUKS_DEV_UUID} UUID=${LUKS_DEV_UUID} - tpm2-device=auto,discard" > /etc/crypttab

# cat /etc/crypttab
luks-6e6c2ee3-852a-4efb-9801-8277dccf0af1 UUID=6e6c2ee3-852a-4efb-9801-8277dccf0af1 - tpm2-device=auto,discard

Update the grub.cfg file.

# grub2-mkconfig -o /boot/grub2/grub.cfg

Run dracut to rebuild the initramfs.

# dracut --verbose --force

Check that the TPM has been added to the encrypted volume.

# systemd-cryptenroll /dev/vda2
SLOT TYPE    
   0 password
   1 tpm2

Finally, reboot.

# reboot

The LUKS passphrase should only be asked once before the GRUB menu appears.

9. Install the Desktop Environment

You can now install the desktop environment of your choice. To find out which desktop environments are available to you, run the following command.

# dnf group list available hidden --ids

I will, however, install the GNOME 44 desktop environment that comes with Fedora's official desktop edition 'Fedora Workstation'.

# dnf group install 'Fedora Workstation'

Create subvolumes for the directories /var/lib/AccountsService and /var/lib/gdm. These subvolumes are only relevant if you install GNOME desktop. They contain login user information as well as Gnome display information.

When you boot to a snapshot from the GRUB menu, the system freezes just before the GNOME login screen because the snapshots are read-only. These directories must always be writable. By separating these directories into subvolumes, you can ensure that they are always read-write and that the system does not freeze just before login.

Create AccountsService subvolume.

# mv -v /var/lib/AccountsService{,-old}
# btrfs subvolume create /var/lib/AccountsService
# cp -arv /var/lib/AccountsService{-old/.,/}
# restorecon -RFv /var/lib/AccountsService
# rm -rfv /var/lib/AccountsService-old

Create gdm subvolume.

# mv -v /var/lib/gdm{,-old}
# btrfs subvolume create /var/lib/gdm
# cp -arv /var/lib/gdm{-old/.,/}
# restorecon -RFv /var/lib/gdm
# rm -rfv /var/lib/gdm-old
# chmod 1770 /var/lib/gdm
# chown -R gdm:gdm /var/lib/gdm

In the user's home directory, create a subvolume called .mozilla.

# USERNAME="$(id -nu 1000)"

# echo $USERNAME
madhu

# btrfs subvolume create /home/$USERNAME/.mozilla
# chown -R $USERNAME: /home/$USERNAME/.mozilla
# restorecon -RFv /home/$USERNAME/.mozilla

In the /etc/fstab file, add the newly created subvolumes. The updated lines are displayed in amber. Replace UUID and user name with your setup's UUID and user name.

# cat /etc/fstab
UUID=85826240-0537-4ad0-a2ee-c59166a18f7c /                        btrfs defaults,compress=zstd:1,x-systemd.device-timeout=0 0 0
UUID=AFE6-47F9                            /boot/efi                vfat  umask=0077,shortname=winnt 0 0
UUID=85826240-0537-4ad0-a2ee-c59166a18f7c /home                    btrfs subvol=home,compress=zstd:1,x-systemd.device-timeout=0 0 0
UUID=85826240-0537-4ad0-a2ee-c59166a18f7c /opt                     btrfs subvol=opt,compress=zstd:1,x-systemd.device-timeout=0 0 0
UUID=85826240-0537-4ad0-a2ee-c59166a18f7c /var/cache               btrfs subvol=var/cache,compress=zstd:1,x-systemd.device-timeout=0 0 0
UUID=85826240-0537-4ad0-a2ee-c59166a18f7c /var/crash               btrfs subvol=var/crash,compress=zstd:1,x-systemd.device-timeout=0 0 0
UUID=85826240-0537-4ad0-a2ee-c59166a18f7c /var/log                 btrfs subvol=var/log,compress=zstd:1,x-systemd.device-timeout=0 0 0
UUID=85826240-0537-4ad0-a2ee-c59166a18f7c /var/spool               btrfs subvol=var/spool,compress=zstd:1,x-systemd.device-timeout=0 0 0
UUID=85826240-0537-4ad0-a2ee-c59166a18f7c /var/tmp                 btrfs subvol=var/tmp,compress=zstd:1,x-systemd.device-timeout=0 0 0
UUID=85826240-0537-4ad0-a2ee-c59166a18f7c /var/lib/libvirt/images  btrfs subvol=var/lib/libvirt/images,compress=zstd:1,x-systemd.device-timeout=0 0 0
UUID=85826240-0537-4ad0-a2ee-c59166a18f7c /var/lib/AccountsService btrfs subvol=var/lib/AccountsService,compress=zstd:1,x-systemd.device-timeout=0 0 0
UUID=85826240-0537-4ad0-a2ee-c59166a18f7c /var/lib/gdm             btrfs subvol=var/lib/gdm,compress=zstd:1,x-systemd.device-timeout=0 0 0
UUID=85826240-0537-4ad0-a2ee-c59166a18f7c /home/madhu/.mozilla     btrfs subvol=home/madhu/.mozilla,compress=zstd:1,x-systemd.device-timeout=0 0 0

Reload the /etc/fstab file to mount all filesystems.

# systemctl daemon-reload

# mount -va
/                        : ignored
/boot/efi                : already mounted
/home                    : already mounted
/opt                     : already mounted
/var/cache               : already mounted
/var/crash               : already mounted
/var/log                 : already mounted
/var/spool               : already mounted
/var/tmp                 : already mounted
/var/lib/libvirt/images  : already mounted
/var/lib/AccountsService : successfully mounted
/var/lib/gdm             : successfully mounted
/home/madhu/.mozilla     : successfully mounted

Check your subvolumes.

# btrfs subvolume list /
ID 256 gen 282 top level 5 path home
ID 257 gen 171 top level 5 path opt
ID 258 gen 272 top level 5 path var/cache
ID 259 gen 171 top level 5 path var/crash
ID 260 gen 289 top level 5 path var/log
ID 261 gen 270 top level 5 path var/spool
ID 262 gen 281 top level 5 path var/tmp
ID 263 gen 265 top level 5 path var/lib/libvirt/images
ID 264 gen 270 top level 5 path var/lib/portables
ID 265 gen 271 top level 5 path var/lib/machines
ID 266 gen 277 top level 5 path var/lib/AccountsService
ID 267 gen 280 top level 5 path var/lib/gdm
ID 268 gen 283 top level 256 path home/madhu/.mozilla

Your setup should now look somewhat like this.

# lsblk -o name,mountpoints /dev/vda
NAME                                          MOUNTPOINTS
vda                                           
├─vda1                                        /boot/efi
└─vda2                                        
  └─luks-6e6c2ee3-852a-4efb-9801-8277dccf0af1 /home/madhu/.mozilla
                                              /var/lib/gdm
                                              /var/lib/AccountsService
                                              /var/tmp
                                              /var/spool
                                              /var/log
                                              /var/lib/libvirt/images
                                              /var/crash
                                              /var/cache
                                              /opt
                                              /home
                                              /

Now that you've installed the desktop environment, enable boot animation and make the spinner theme your default.

# plymouth-set-default-theme spinner -R

Enable Red Hat Graphical Boot in /etc/default/grub as well.

# sed -i 's/quiet/rhgb quiet/' /etc/default/grub

Finally, update the grub.cfg file.

# grub2-mkconfig -o /boot/grub2/grub.cfg

Fedora Workstation installation is complete. Reboot the system now.

# reboot

You will now be logged into the Fedora 38 Workstation with the all-new Gnome 44 desktop interface.

How to Install Fedora 38 with Full Disk Encryption, Snapshot and Rollback Support Fedora Workstation

10. Install and Configure Snapper

Install snapper and the optional package python3-dnf-plugin-snapper. The python3-dnf-plugin-snapper package allows you to create pre and post snapshots every time you install a package on the system with the dnf package manager.

$ sudo dnf install snapper python3-dnf-plugin-snapper

Create new snapper configurations named root and home for the Btrfs volume at / and /home, respectively.

$ sudo snapper -c root create-config /
$ sudo snapper -c home create-config /home

Verify the snapper configuration files have been created.

$ sudo snapper list-configs
Config | Subvolume
-------+----------
home   | /home    
root   | /

Allow regular user to use snapper without requiring root privileges. Add your user name to the snapper's root and home configurations to set the ACL on the /.snapshots and /home/.snapshots directories.

$ sudo snapper -c root set-config ALLOW_USERS=$USER SYNC_ACL=yes
$ sudo snapper -c home set-config ALLOW_USERS=$USER SYNC_ACL=yes

Add the newly created snapshot subvolumes to the /etc/fstab file. The updated lines are in amber.

$ sudo vim /etc/fstab
...
UUID=85826240-0537-4ad0-a2ee-c59166a18f7c /home/madhu/.mozilla     btrfs subvol=home/madhu/.mozilla,compress=zstd:1,x-systemd.device-timeout=0 0 0
UUID=85826240-0537-4ad0-a2ee-c59166a18f7c /.snapshots              btrfs subvol=.snapshots,compress=zstd:1,x-systemd.device-timeout=0 0 0
UUID=85826240-0537-4ad0-a2ee-c59166a18f7c /home/.snapshots         btrfs subvol=home/.snapshots,compress=zstd:1,x-systemd.device-timeout=0 0 0

Reload the /etc/fstab file.

$ sudo systemctl daemon-reload
$ sudo mount -va

Your subvolumes will look like this.

$ sudo btrfs subvolume list /
ID 256 gen 324 top level 5 path home
ID 257 gen 171 top level 5 path opt
ID 258 gen 317 top level 5 path var/cache
ID 259 gen 171 top level 5 path var/crash
ID 260 gen 324 top level 5 path var/log
ID 261 gen 315 top level 5 path var/spool
ID 262 gen 318 top level 5 path var/tmp
ID 263 gen 304 top level 5 path var/lib/libvirt/images
ID 264 gen 315 top level 5 path var/lib/portables
ID 265 gen 315 top level 5 path var/lib/machines
ID 266 gen 315 top level 5 path var/lib/AccountsService
ID 267 gen 315 top level 5 path var/lib/gdm
ID 268 gen 304 top level 256 path home/madhu/.mozilla
ID 269 gen 319 top level 5 path .snapshots
ID 270 gen 320 top level 256 path home/.snapshots

And your setup should look something like this.

$ lsblk -o name,mountpoints /dev/vda
NAME                                          MOUNTPOINTS
vda                                           
├─vda1                                        /boot/efi
└─vda2                                        
  └─luks-6e6c2ee3-852a-4efb-9801-8277dccf0af1 /home/.snapshots
                                              /.snapshots
                                              /home/madhu/.mozilla
                                              /var/tmp
                                              /var/spool
                                              /var/log
                                              /var/lib/libvirt/images
                                              /var/lib/gdm
                                              /var/lib/AccountsService
                                              /var/crash
                                              /var/cache
                                              /opt
                                              /home
                                              /

Disable the indexing of the .snapshots directories by updatedb. It is enabled by default, which can cause significant slowdown and excessive memory usage if you have a large number of snapshots.

$ echo 'PRUNENAMES = ".snapshots"' | sudo tee -a /etc/updatedb.conf

List the snapshots for / volume. For the root, you may use snapper -c root ls or simply snapper ls. Both provide the same output.

$ snapper ls
 # | Type   | Pre # | Date | User | Cleanup | Description | Userdata
---+--------+-------+------+------+---------+-------------+---------
0  | single |       |      | root |         | current     |

List the snapshots for /home subvolume.

$ snapper -c home ls
 # | Type   | Pre # | Date | User | Cleanup | Description | Userdata
---+--------+-------+------+------+---------+-------------+---------
0  | single |       |      | root |         | current     |

At this stage, you do not have any snapshots, we will create some later.

11. Install and Configure Grub-Btrfs

The grub-btrfs package adds 'Fedora Linux snapshots' to the GRUB menu and allows you to test a snapshot in read-only mode before rolling back to it in read-write mode.

Clone the grub-btrfs repository.

$ git clone https://github.com/Antynea/grub-btrfs

The setup is pre-configured to work with Arch and Gentoo out of the box. You must make a few changes to the grub-btrfs config file to make it work with Fedora.

$ cd grub-btrfs

$ sed -i '/#GRUB_BTRFS_SNAPSHOT_KERNEL/a GRUB_BTRFS_SNAPSHOT_KERNEL_PARAMETERS="systemd.volatile=state"' config
$ sed -i '/#GRUB_BTRFS_GRUB_DIRNAME/a GRUB_BTRFS_GRUB_DIRNAME="/boot/grub2"' config
$ sed -i '/#GRUB_BTRFS_MKCONFIG=/a GRUB_BTRFS_MKCONFIG=/sbin/grub2-mkconfig' config
$ sed -i '/#GRUB_BTRFS_SCRIPT_CHECK=/a GRUB_BTRFS_SCRIPT_CHECK=grub2-script-check' config

Then install it.

$ sudo make install

Update grub.cfg and enable the grub-btrfsd.service.

$ sudo grub2-mkconfig -o /boot/grub2/grub.cfg
$ sudo systemctl enable --now grub-btrfsd.service

You will receive the 'No snapshots found' error since you have not yet created any snapshots, but don't worry, it will function properly after you do.

Your grub-btrfs installation is now complete. You may now delete the cloned grub-btrfs repository.

$ cd ..
$ rm -rvf grub-btrfs

12. Create a System Root Snapshot and Set It as the Default

Now that everything in Fedora has been installed and configured, it is time to create a snapshot of the system root and make it the default. This makes it much easier to manage snapshots and rollbacks.

Create a directory named '1' in the /.snapshots directory.

$ sudo mkdir -v /.snapshots/1

Create an XML file called info.xml in the /.snapshots/1/ directory. Copy the following command from sudo to EOF, paste it into your terminal, and hit the <Enter> button.

$ sudo bash -c "cat > /.snapshots/1/info.xml" <<EOF
<?xml version="1.0"?>
<snapshot>
  <type>single</type>
  <num>1</num>
  <date>$(date -u +"%F %T")</date>
  <description>first root subvolume</description>
</snapshot>
EOF

Now, create a read-write subvolume snapshot of the system root in the /.snapshots/1/ directory.

$ sudo btrfs subvolume snapshot / /.snapshots/1/snapshot

Get the subvolid of the /.snapshots/1/snapshot subvolume.

$ sudo btrfs inspect-internal rootid /.snapshots/1/snapshot
271

Using subvolid 271, set the /.snapshots/1/snapshot subvolume as the default subvolume for the root (/) filesystem.

$ sudo btrfs subvolume set-default 271 /

And reboot your system.

$ sudo reboot

After rebooting, confirm that the /.snapshots/1/snapshot subvolume is indeed the default for the / filesystem.

$ sudo btrfs subvolume get-default /
ID 271 gen 367 top level 269 path .snapshots/1/snapshot

Now check the snapshots in snapper.

$ snapper ls
 # | Type   | Pre # | Date                             | User | Cleanup | Description          | Userdata
---+--------+-------+----------------------------------+------+---------+----------------------+---------
0  | single |       |                                  | root |         | current              |         
1* | single |       | Saturday 06 May 2023 06:36:26 PM | root |         | first root subvolume |

As you can see, the /.snapshots/1/snapshot subvolume is also visible as snapshot #1 in snapper. The asterisk (*) indicates that this snapshot is the default and is currently active.

13. Snapper Tests

I'll run some tests here to make sure everything is in working order and that there are no problems.

Test 1: Undo the Changes Made by the DNF Installer

In this test, I will use the dnf package manager to install a small package named ps_mem. I'll check the system after it's installed to see what changes were made. I'll then attempt to undo the changes and see if I can get the system back to its pre-installation state.

Install the ps_mem package. It displays the core memory used per program (not per process).

$ sudo dnf install ps_mem

Run the ps_mem program to see if it was installed successfully.

$ sudo ps_mem
 Private  +   Shared  =  RAM used	Program

148.0 KiB +  13.5 KiB = 161.5 KiB	fusermount3
196.0 KiB +  13.5 KiB = 209.5 KiB	inotifywait
296.0 KiB +  15.5 KiB = 311.5 KiB	mcelog
...
 51.4 MiB +   3.5 MiB =  54.9 MiB	python3.11
204.2 MiB +  23.5 MiB = 227.6 MiB	gnome-shell
268.7 MiB +   2.0 MiB = 270.7 MiB	packagekitd
---------------------------------
                          1.2 GiB
=================================

Check the snapper for snapshots.

$ snapper ls
 # | Type   | Pre # | Date                             | User | Cleanup | Description                 | Userdata
---+--------+-------+----------------------------------+------+---------+-----------------------------+---------
0  | single |       |                                  | root |         | current                     |         
1* | single |       | Saturday 06 May 2023 06:36:26 PM | root |         | first root subvolume        |         
2  | pre    |       | Saturday 06 May 2023 06:47:54 PM | root | number  | /usr/bin/dnf install ps_mem |         
3  | post   |     2 | Saturday 06 May 2023 06:47:55 PM | root | number  | /usr/bin/dnf install ps_mem |

As you can see, the ps_mem package has pre (#2) and post (#3) snapshots.

Let's take a look at the changes it made to the system between snapshots #2 and #3.

$ snapper status 2..3
+..... /usr/bin/ps_mem
c..... /usr/lib/sysimage/rpm/rpmdb.sqlite-shm
c..... /usr/lib/sysimage/rpm/rpmdb.sqlite-wal
+..... /usr/share/doc/ps_mem
+..... /usr/share/doc/ps_mem/LICENSE
+..... /usr/share/man/man1/ps_mem.1.gz
c..... /var/lib/dnf/history.sqlite-shm
c..... /var/lib/dnf/history.sqlite-wal
-..... /var/lib/dnf/rpmdb_lock.pid

Now I'll undo the changes.

$ sudo snapper undochange 2..3
create:1 modify:4 delete:4

Check again to ensure that the ps_mem package was successfully uninstalled.

$ sudo ps_mem
sudo: ps_mem: command not found

Cool! The undo was successful. But, because I want to keep the package ps_mem, I'll undo the changes again. This time between snapshots #3 and #2.

$ sudo snapper undochange 3..2
create:4 modify:4 delete:1

Now run ps_mem again to see if it exists.

$ sudo ps_mem
 Private  +   Shared  =  RAM used	Program

148.0 KiB +  13.5 KiB = 161.5 KiB	fusermount3
196.0 KiB +  13.5 KiB = 209.5 KiB	inotifywait
296.0 KiB +  15.5 KiB = 311.5 KiB	mcelog
...
 51.4 MiB +   3.5 MiB =  54.9 MiB	python3.11
204.2 MiB +  23.5 MiB = 227.6 MiB	gnome-shell
268.7 MiB +   2.0 MiB = 270.7 MiB	packagekitd
---------------------------------
                          1.2 GiB
=================================

Test 1 is successful.

Test 2: Undo Any Changes Made to Individual Files

Assume you're configuring something and editing a configuration file. Then you realize you messed up the configuration file. You want to see what changes you made to the original file and possibly wish you could undo the changes.

You can use snapper to undo not only all changes between two snapshots, but also a single file.

In this test, I will simply delete the last line of the /etc/hosts file. Then I'll compare it to the same file in snapshot #3 to see what changed and if I can undo all of the changes and restore it to its original state.

This is how the original /etc/hosts file looks.

$ cat /etc/hosts
# Loopback entries; do not change.
# For historical reasons, localhost precedes localhost.localdomain:
127.0.0.1   localhost localhost.localdomain localhost4 localhost4.localdomain4
::1         localhost localhost.localdomain localhost6 localhost6.localdomain6
# See hosts(5) for proper format and other examples:
# 192.168.1.10 foo.mydomain.org foo
# 192.168.1.13 bar.mydomain.org bar

Now I'll remove the last line.

$ sudo sed -i '$d' /etc/hosts

$ cat /etc/hosts
# Loopback entries; do not change.
# For historical reasons, localhost precedes localhost.localdomain:
127.0.0.1   localhost localhost.localdomain localhost4 localhost4.localdomain4
::1         localhost localhost.localdomain localhost6 localhost6.localdomain6
# See hosts(5) for proper format and other examples:
# 192.168.1.10 foo.mydomain.org foo

Using snapper, compare the /etc/hosts file with the one in snapshot #3.

$ snapper diff 3..0 /etc/hosts
--- /.snapshots/3/snapshot/etc/hosts	2022-11-27 20:56:24.000000000 +0530
+++ /etc/hosts	2023-05-06 18:54:56.251963413 +0530
@@ -4,4 +4,3 @@
 ::1         localhost localhost.localdomain localhost6 localhost6.localdomain6
 # See hosts(5) for proper format and other examples:
 # 192.168.1.10 foo.mydomain.org foo
-# 192.168.1.13 bar.mydomain.org bar

In the output, 0 represents the current snapshot. As you can see in the last line, the /etc/hosts file in the current snapshot is missing that line compared to the one in snapshot #3. If you want to learn how to read the output, take a look at the man pages man snapper and man diff.

To replace the /etc/hosts file with the one in snapshot #3, use the following command.

$ sudo snapper undochange 3..0 /etc/hosts
create:0 modify:1 delete:0

Check to see if the /etc/hosts file has indeed been rolled back.

$ cat /etc/hosts
# Loopback entries; do not change.
# For historical reasons, localhost precedes localhost.localdomain:
127.0.0.1   localhost localhost.localdomain localhost4 localhost4.localdomain4
::1         localhost localhost.localdomain localhost6 localhost6.localdomain6
# See hosts(5) for proper format and other examples:
# 192.168.1.10 foo.mydomain.org foo
# 192.168.1.13 bar.mydomain.org bar

Test 2 is successful.

Test 3: Create Manual Pre-Post Snapshots and Undo the Changes

In my previous article, I performed a manual pre-post snapshots test for installing a package with Gnome Software. The concept for both tests is the same. If you want, you can also run that test.

In this test, I will begin by creating manual 'pre' snapshots of the / and /home directories. Then I'll install a small package called Color Picker (gcolor3) from GitHub by compiling it from the source file. When I'm finished, regardless of whether the compiling was successful or not, I'll create 'post' snapshots for the previously created 'pre' snapshots. I'll then look at the changes it made to the system between the pre-post snapshots and, finally, undo everything.

Create a manual 'pre' snapshot for both the / and /home directories.

$ snapper -c root create -t pre -c number -d 'Pre Color Picker'
$ snapper -c home create -t pre -c number -d 'Pre Color Picker'

List the snapshots.

$ snapper ls
 # | Type   | Pre # | Date                             | User  | Cleanup | Description                 | Userdata
---+--------+-------+----------------------------------+-------+---------+-----------------------------+---------
0  | single |       |                                  | root  |         | current                     |         
1* | single |       | Saturday 06 May 2023 06:36:26 PM | root  |         | first root subvolume        |         
2  | pre    |       | Saturday 06 May 2023 06:47:54 PM | root  | number  | /usr/bin/dnf install ps_mem |         
3  | post   |     2 | Saturday 06 May 2023 06:47:55 PM | root  | number  | /usr/bin/dnf install ps_mem |         
4  | pre    |       | Saturday 06 May 2023 07:00:52 PM | madhu | number  | Pre Color Picker            |         

$ snapper -c home ls
 # | Type   | Pre # | Date                             | User  | Cleanup | Description      | Userdata
---+--------+-------+----------------------------------+-------+---------+------------------+---------
0  | single |       |                                  | root  |         | current          |         
1  | pre    |       | Saturday 06 May 2023 07:01:00 PM | madhu | number  | Pre Color Picker |

Install the pre-requisite packages before compiling the source file.

$ sudo dnf install meson gcc-c++ libhandy-devel libportal-devel \
    libportal-gtk3-devel

Compile the source.

$ git clone https://github.com/Hjdskes/gcolor3.git
$ cd gcolor3
$ meson setup build
$ ninja-build -C build
$ sudo ninja-build -C build install

Make sure the Color Picker package has been installed properly. Enter gcolor3 into the terminal and hit <Enter>. The Color Picker app must be open and appear as shown.

How to Install Fedora 38 with Snapshot and Rollback Support gcolor3

Now that you have successfully installed the Color Picker application, close the application and create 'post' snapshots for the / and /home directories.

$ snapper -c home create -t post --pre-number 1 -c number -d 'Post Color Picker'
$ snapper -c root create -t post --pre-number 4 -c number -d 'Post Color Picker'

List the snapshots.

$ snapper -c root ls
 # | Type   | Pre # | Date                             | User  | Cleanup | Description                                                                            | Userdata
---+--------+-------+----------------------------------+-------+---------+----------------------------------------------------------------------------------------+---------
0  | single |       |                                  | root  |         | current                                                                                |         
1* | single |       | Saturday 06 May 2023 06:36:26 PM | root  |         | first root subvolume                                                                   |         
2  | pre    |       | Saturday 06 May 2023 06:47:54 PM | root  | number  | /usr/bin/dnf install ps_mem                                                            |         
3  | post   |     2 | Saturday 06 May 2023 06:47:55 PM | root  | number  | /usr/bin/dnf install ps_mem                                                            |         
4  | pre    |       | Saturday 06 May 2023 07:00:52 PM | madhu | number  | Pre Color Picker                                                                       |         
5  | pre    |       | Saturday 06 May 2023 07:04:12 PM | root  | number  | /usr/bin/dnf install meson gcc-c++ libhandy-devel libportal-devel libportal-gtk3-devel |         
6  | post   |     5 | Saturday 06 May 2023 07:04:42 PM | root  | number  | /usr/bin/dnf install meson gcc-c++ libhandy-devel libportal-devel libportal-gtk3-devel |         
7  | post   |     4 | Saturday 06 May 2023 07:05:55 PM | madhu | number  | Post Color Picker                                                                      |         

$ snapper -c home ls
 # | Type   | Pre # | Date                             | User  | Cleanup | Description       | Userdata
---+--------+-------+----------------------------------+-------+---------+-------------------+---------
0  | single |       |                                  | root  |         | current           |         
1  | pre    |       | Saturday 06 May 2023 07:01:00 PM | madhu | number  | Pre Color Picker  |         
2  | post   |     1 | Saturday 06 May 2023 07:05:44 PM | madhu | number  | Post Color Picker |

Review the changes made to the system between the pre and post snapshots.

$ snapper -c home status 1..2
+..... /home/madhu/.config/gcolor3
+..... /home/madhu/.config/gcolor3/config.ini
+..... /home/madhu/gcolor3
+..... /home/madhu/gcolor3/build
+..... /home/madhu/gcolor3/build/build.ninja
...
+..... /home/madhu/.local/share/gnome-shell/application_state
c..... /home/madhu/.local/share/gvfs-metadata/root
-..... /home/madhu/.local/share/gvfs-metadata/root-3616f454.log
+..... /home/madhu/.local/share/gvfs-metadata/root-54385136.log

$ snapper -c root status 4..7
c..... /boot/grub2/grub-btrfs.cfg
c..... /boot/grub2/grub.cfg
+..... /etc/alternatives/ld
c..... /etc/ld.so.cache
+..... /usr/bin/addr2line
...
+..... /usr/share/xml/dbus-1/catalog.xml
+..... /usr/share/xml/dbus-1/introspect.dtd
+..... /usr/share/zsh/site-functions/_meson
+..... /usr/share/zsh/site-functions/_ninja
+..... /var/lib/alternatives/ld
c..... /var/lib/dnf/history.sqlite
c..... /var/lib/dnf/history.sqlite-shm
c..... /var/lib/dnf/history.sqlite-wal

This is the total number of files added, removed, or modified.

$ snapper -c home status 1..2 | wc -l
273

$ snapper -c root status 4..7 | wc -l
14144

I'll now undo the changes in the / and /home directories.

$ sudo snapper -c home undochange 1..2
create:1 modify:1 delete:271

$ sudo snapper -c root undochange 4..7
create:0 modify:11 delete:14133

The Color Picker package will be completely removed from the system.

Log out and then log back in.

$ gnome-session-quit

Check again to see if the Color Picker has been completely removed.

$ gcolor3
bash: gcolor3: command not found...

Since you removed the Color Picker package, there is no reason to keep its pre-post snapshots if you do not intend to use it again. So you can delete those snapshots.

$ snapper -c home delete 1-2
$ snapper -c root delete 4-7

Test 3 is successful.

Test 4: Rollback to a Previous Snapshot from the GRUB Menu

In this rollback test, I will delete the most critical files and directories on which the Linux operating system relies. These include Linux kernel and initramfs files in the /boot directory, the /etc directory, which contains all configuration files, and the /usr directory, which contains all drivers, kernel modules, and library files.

Once I delete these files and directories, I will be unable to issue any commands because the Linux OS is pretty much dead, and the system will undoubtedly fail the next time it boots. So I'll hard boot the system and see if I can save the day by rolling back to a working snapshot.

So, before I delete them, here's how they appear.

$ sudo du -sch /boot /etc /usr
177M	/boot
32M	/etc
5.6G	/usr
5.8G	total

Now delete the files and directories.

$ sudo rm -rvf /boot/{vmlinuz,initramfs}* /etc /usr

Boom!!! All critical files have been removed. Your Linux system is no longer operational. Reboot the system. Your system should now fail to boot and display something like this.

How to Install Fedora 38 with Full Disk Encryption, Snapshot and Rollback Support Boot Error

Return to the GRUB menu and this time boot from snapshot #3.

How to Install Fedora 38 with Full Disk Encryption, Snapshot and Rollback Support GRUB Menu

How to Install Fedora 38 with Full Disk Encryption, Snapshot and Rollback Support Snapshot No 3

How to Install Fedora 38 with Full Disk Encryption, Snapshot and Rollback Support Latest Kernel

Open the terminal and make sure everything is back to normal.

$ sudo ls /boot /etc /usr

Because you booted into the snapshot, the '/' file system is read-only.

$ sudo btrfs property get -ts /
ro=true

Now that everything appears to be operating properly, it is time to roll back this snapshot in read-write mode.

$ snapper ls
 # | Type   | Pre # | Date                             | User | Cleanup | Description                 | Userdata
---+--------+-------+----------------------------------+------+---------+-----------------------------+---------
0  | single |       |                                  | root |         | current                     |         
1+ | single |       | Saturday 06 May 2023 06:36:26 PM | root |         | first root subvolume        |         
2  | pre    |       | Saturday 06 May 2023 06:47:54 PM | root | number  | /usr/bin/dnf install ps_mem |         
3- | post   |     2 | Saturday 06 May 2023 06:47:55 PM | root | number  | /usr/bin/dnf install ps_mem |

As you can see from the output, snapshot #1 has a '+' symbol, indicating that it is the default subvolume, whereas snapshot #3 has a '-' symbol, indicating that it is the currently booted subvolume.

$ sudo snapper rollback
Ambit is classic.
Creating read-only snapshot of default subvolume. (Snapshot 4.)
Creating read-write snapshot of current subvolume. (Snapshot 5.)
Setting default subvolume to snapshot 5.

Now, reboot your system.

$ sudo reboot

The first thing you should do after a rollback is to update the grub.cfg file so that the GRUB menu snapshots are always in sync.

$ sudo grub2-mkconfig -o /boot/grub2/grub.cfg

Check the system once again.

$ sudo du -sch /boot /etc /usr
177M    /boot
32M     /etc
5.6G    /usr
5.8G    total

$ sudo btrfs property get -ts /
ro=false

$ snapper ls
 # | Type   | Pre # | Date                             | User | Cleanup | Description                 | Userdata     
---+--------+-------+----------------------------------+------+---------+-----------------------------+--------------
0  | single |       |                                  | root |         | current                     |              
1  | single |       | Saturday 06 May 2023 06:36:26 PM | root | number  | first root subvolume        |              
2  | pre    |       | Saturday 06 May 2023 06:47:54 PM | root | number  | /usr/bin/dnf install ps_mem |              
3  | post   |     2 | Saturday 06 May 2023 06:47:55 PM | root | number  | /usr/bin/dnf install ps_mem |              
4  | single |       | Saturday 06 May 2023 07:16:22 PM | root | number  | rollback backup of #1       | important=yes
5* | single |       | Saturday 06 May 2023 07:16:22 PM | root |         | writable copy of #3         |

$ sudo btrfs subvolume list /
ID 256 gen 445 top level 5 path home
ID 257 gen 171 top level 5 path opt
ID 258 gen 443 top level 5 path var/cache
ID 259 gen 171 top level 5 path var/crash
ID 260 gen 446 top level 5 path var/log
ID 261 gen 444 top level 5 path var/spool
ID 262 gen 445 top level 5 path var/tmp
ID 263 gen 304 top level 5 path var/lib/libvirt/images
ID 264 gen 354 top level 5 path var/lib/portables
ID 265 gen 354 top level 5 path var/lib/machines
ID 266 gen 443 top level 5 path var/lib/AccountsService
ID 267 gen 443 top level 5 path var/lib/gdm
ID 268 gen 304 top level 256 path home/madhu/.mozilla
ID 269 gen 443 top level 5 path .snapshots
ID 270 gen 426 top level 256 path home/.snapshots
ID 271 gen 438 top level 269 path .snapshots/1/snapshot
ID 272 gen 382 top level 269 path .snapshots/2/snapshot
ID 273 gen 439 top level 269 path .snapshots/3/snapshot
ID 274 gen 438 top level 269 path .snapshots/4/snapshot
ID 275 gen 446 top level 269 path .snapshots/5/snapshot

$ sudo btrfs subvolume get-default /
ID 275 gen 446 top level 269 path .snapshots/5/snapshot

You successfully restored the system with the latest working snapshot in read and write mode.

Test 4 is successful.

14. Make a Snapshot the New System Root

Now that you've tested the snapper undo and rollback features and are hopefully satisfied with the results, you might want to set one of the snapshots as the new default system root and resume normal operating system use.

But first, let me show you how to determine how much space each snapshot takes up on your disk.

$ sudo btrfs filesystem du -s --human-readable /.snapshots/*/snapshot
     Total   Exclusive  Set shared  Filename
  46.98MiB       0.00B    33.39MiB  /.snapshots/1/snapshot
   5.48GiB    32.00KiB     2.88GiB  /.snapshots/2/snapshot
   5.48GiB    52.00KiB     2.88GiB  /.snapshots/3/snapshot
  46.98MiB       0.00B    33.39MiB  /.snapshots/4/snapshot
   5.48GiB     1.09MiB     2.88GiB  /.snapshots/5/snapshot

As you can see from the output above, snapshot #1 has an exclusive size of 0 bytes. This is because, as you may recall, I deleted all of the files recursively. Snapshot #4 is also 0 bytes in size because it is a read-only duplicate of snapshot #1.

To find out how the 'Total' and 'Set shared' values are calculated, see the man page man btrfs-filesystem.

Now, let's return to making a snapshot the new system root.

$ snapper ls
 # | Type   | Pre # | Date                             | User | Cleanup | Description                 | Userdata     
---+--------+-------+----------------------------------+------+---------+-----------------------------+--------------
0  | single |       |                                  | root |         | current                     |              
1  | single |       | Saturday 06 May 2023 06:36:26 PM | root | number  | first root subvolume        |              
2  | pre    |       | Saturday 06 May 2023 06:47:54 PM | root | number  | /usr/bin/dnf install ps_mem |              
3  | post   |     2 | Saturday 06 May 2023 06:47:55 PM | root | number  | /usr/bin/dnf install ps_mem |              
4  | single |       | Saturday 06 May 2023 07:16:22 PM | root | number  | rollback backup of #1       | important=yes
5* | single |       | Saturday 06 May 2023 07:16:22 PM | root |         | writable copy of #3         |

The output indicates that snapshot #5 is the default subvolume, and since it is operational, you can simply delete the remaining snapshots and move on.

However, I will use snapshot #2 as the default because it is a 'pre' snapshot that was created just before ps_mem was installed and is clean.

So, first, delete all of the snapshots except #2. You cannot delete snapshot #5 because it is the active one.

$ snapper delete 1
$ snapper delete 3-4

$ snapper ls
 # | Type   | Pre # | Date                             | User | Cleanup | Description                 | Userdata
---+--------+-------+----------------------------------+------+---------+-----------------------------+---------
0  | single |       |                                  | root |         | current                     |         
2  | pre    |       | Saturday 06 May 2023 06:47:54 PM | root | number  | /usr/bin/dnf install ps_mem |         
5* | single |       | Saturday 06 May 2023 07:16:22 PM | root |         | writable copy of #3         |

Create a directory named '1' in the /.snapshots directory.

$ sudo mkdir -v /.snapshots/1

Copy the info.xml file from /.snapshots/2/ to /.snapshots/1/.

$ sudo cp -v /.snapshots/2/info.xml /.snapshots/1/

Edit the /.snapshots/1/info.xml file and change the content from this...

$ sudo cat /.snapshots/1/info.xml
<?xml version="1.0"?>
<snapshot>
  <type>pre</type>
  <num>2</num>
  <date>2023-05-06 13:17:54</date>
  <description>/usr/bin/dnf install ps_mem</description>
  <cleanup>number</cleanup>
</snapshot>

...to this. Changes are in amber.

$ sudo cat /.snapshots/1/info.xml
<?xml version="1.0"?>
<snapshot>
  <type>single</type>
  <num>1</num>
  <date>2023-05-06 13:17:54</date>
  <description>new root subvolume</description>
</snapshot>

Create a read-write subvolume snapshot of snapshot #2 in the /.snapshots/1/ directory.

$ sudo btrfs subvolume snapshot /.snapshots/2/snapshot /.snapshots/1/snapshot

Get the subvolid of the /.snapshots/1/snapshot subvolume.

$ sudo btrfs inspect-internal rootid /.snapshots/1/snapshot
276

Using subvolid 276, set the /.snapshots/1/snapshot subvolume as the default subvolume for the root (/) filesystem.

$ sudo btrfs subvolume set-default 276 /

Then reboot.

$ sudo reboot

After rebooting, confirm that the /.snapshots/1/snapshot subvolume is indeed the default for the / filesystem.

$ sudo btrfs subvolume get-default /
ID 276 gen 480 top level 269 path .snapshots/1/snapshot

And is writable.

$ sudo btrfs property get -ts /
ro=false

Take a look at the snapper now.

$ snapper ls
 # | Type   | Pre # | Date                             | User | Cleanup | Description                 | Userdata
---+--------+-------+----------------------------------+------+---------+-----------------------------+---------
0  | single |       |                                  | root |         | current                     |         
1* | single |       | Saturday 06 May 2023 06:47:54 PM | root |         | new root subvolume          |         
2  | pre    |       | Saturday 06 May 2023 06:47:54 PM | root | number  | /usr/bin/dnf install ps_mem |         
5  | single |       | Saturday 06 May 2023 07:16:22 PM | root |         | writable copy of #3         |

As you can see, snapshot #1 is the default. You can now delete the remaining snapshots.

$ snapper delete 2-5

$ snapper ls
 # | Type   | Pre # | Date                             | User | Cleanup | Description        | Userdata
---+--------+-------+----------------------------------+------+---------+--------------------+---------
0  | single |       |                                  | root |         | current            |         
1* | single |       | Saturday 06 May 2023 06:47:54 PM | root |         | new root subvolume |         

$ sudo btrfs filesystem du -s --human-readable /.snapshots/*/snapshot
     Total   Exclusive  Set shared  Filename
   5.48GiB   448.00KiB     2.88GiB  /.snapshots/1/snapshot

You're done.

15. Enable Automatic Timeline Snapshots

Now that you've finished everything, you can enable automatic timeline snapshots. When the timeline is enabled, a snapshot is created once every hour. Once per day, the timeline cleanup algorithm cleans up the snapshots.

It is enabled by default in both the root and home configurations. Enable it only for the system root and disable it for the home. See the Arch Wiki page on 'Automatic timeline snapshots' for more information.

$ sudo snapper -c home set-config TIMELINE_CREATE=no
$ sudo systemctl enable --now snapper-timeline.timer
$ sudo systemctl enable --now snapper-cleanup.timer

That's all. Every hour from now on, a 'single' snapshot will be created and cleaned up every other day.

$ snapper ls
 # | Type   | Pre # | Date                             | User | Cleanup  | Description        | Userdata
---+--------+-------+----------------------------------+------+----------+--------------------+---------
0  | single |       |                                  | root |          | current            |         
1* | single |       | Saturday 06 May 2023 06:47:54 PM | root |          | new root subvolume |         
2  | single |       | Saturday 06 May 2023 08:00:04 PM | root | timeline | timeline           |

To stop timeline snapshots, disable the snapper timers.

$ sudo systemctl disable --now snapper-timeline.timer
$ sudo systemctl disable --now snapper-cleanup.timer

The installation of Fedora 38 with Full Disk Encryption, Snapshot and Rollback Support is now complete.