[Tutorial] Convert to systemd-boot

Zesko · August 26, 2022, 2:18pm

The keyfile is for automatically unlocking LUKS encryption at startup without manually entering password when you enable it in /etc/crypttab.

If you do not want automatic decryption, you can remove this keyfile or not.

dalto · August 26, 2022, 2:19pm

You should keep the keyfile for your second disk but ensure it isn’t being loaded into your initramfs.

manfredlotz · August 26, 2022, 2:27pm

Hm, how do I make sure that the keyfile won’t end up in the initramfs.

Currently, as far as I can see, the keyfile is mentioned in the /etc/crypttab only.

JKMooney · August 26, 2022, 11:22pm

Not as difficult as I expected. I would add, as a final step; Go into your motherboard boot settings and make sure “Linux Boot Manager” is selected as first option.

anon49550872 · August 27, 2022, 12:26am

I decided to give it a try, I’m using the kernel-install method.
Followed all the steps, created a script file as per instructions /usr/bin/convert.sh, chmod +x to it, ./convert.sh and rebooted, its working perfectly.

I’m sorry if these questions were asked before, but what should I do when a kernel update happens?
Should I run convert.sh or kernel-install? Or everything will be set automatically?

Also, in case I need to arch-chroot to recover my system, should I change something? That is not clear to me.

My kernel cmd line is:

$ cat /efi/loader/entries/b6cebfd9c59b41a79dd137badc328377-5.19.4-arch1-1.conf

title      EndeavourOS
version    5.19.4-arch1-1
machine-id b6cebfd9c59b41a79dd137badc328377
options    root=UUID=b1a08508-812b-448c-9e8b-ae7903e02f76 rw rootflags=subvol=@ loglevel=3 nowatchdog nvme_load=YES amd_pstate.shared_mem=1
linux      /b6cebfd9c59b41a79dd137badc328377/5.19.4-arch1-1/linux
initrd     /b6cebfd9c59b41a79dd137badc328377/5.19.4-arch1-1/amd-ucode.img
initrd     /b6cebfd9c59b41a79dd137badc328377/5.19.4-arch1-1/initrd

dalto · August 27, 2022, 12:33am

As long as you installed kernel-install-mkinitcpio, nothing special. It should run kernel-install automatically using hooks.

As long as you have /etc/kernel/cmdline populated, you shouldn’t have to do anything special when you chroot. If it isn’t populated now, it should be automatically the first time a kernel is installed updated with kernel-install-mkinitcpio. If you want to force it, you could always reinstall your current kernel.

anon49550872 · August 27, 2022, 12:57am

Really thanks Dalto, really easy tutorial, I really appreciate

If you don’t mind, one last question

I’m using BTRFS and since I don’t have grub and grub-btrfs anymore, to recover from a failed boot state, I would need to boot into the live USB and run the btrfs-assistant from there, right?
I tried that but its not working, is it safe to reinstall btrfs-assistant while in arch-chroot state?

[liveuser@eos-2022.08.05 ~]$ sudo mount -o subvol=@ /dev/sda2 /mnt
[liveuser@eos-2022.08.05 ~]$ sudo mount -o subvol=@log /dev/sda2 /mnt/var/log
[liveuser@eos-2022.08.05 ~]$ sudo mount -o subvol=@cache /dev/sda2 /mnt/var/cache
[liveuser@eos-2022.08.05 ~]$ sudo mount /dev/sda1 /mnt/efi
[liveuser@eos-2022.08.05 ~]$ 
[liveuser@eos-2022.08.05 ~]$ 
[liveuser@eos-2022.08.05 ~]$ sudo arch-chroot /mnt
[root@EndeavourOS /]# 
[root@EndeavourOS /]# sudo btrfs-assistant -l
Authorization required, but no authorization protocol specified

qt.qpa.xcb: could not connect to display :0.0
qt.qpa.plugin: Could not load the Qt platform plugin "xcb" in "" even though it was found.
This application failed to start because no Qt platform plugin could be initialized. Reinstalling the application may fix this problem.

Available platform plugins are: eglfs, linuxfb, minimal, minimalegl, offscreen, vnc, xcb.

/usr/bin/btrfs-assistant: line 42:     8 Aborted                 (core dumped) btrfs-assistant-bin ${params}

dalto · August 27, 2022, 1:00am

Don’t run btrfs-assistant from within a chroot.

You can just install it while booted on the iso and run it.

The only thing you need to do is mount your btrfs partition. One of the subvols or the root of the partition itself.

anon49550872 · August 27, 2022, 2:25am

It worked, steps I took to restore a snapshot through live usb using btrfs-assistant and systemd-boot.

- Boot to live USB
- Update mirrors through Welcome app
# pacman -Syy
# pacman -S snapper
$ yay -S btrfs-assistant
# mount /dev/sda2 /mnt -o subvolid=5
# cp /mnt/@/etc/snapper/configs /etc/snapper/configs
# cp /mnt/@/etc/conf.d/snapper /etc/conf.d/snapper

Then, open btrfs-assistant and restore the snapshot…

limotux · August 27, 2022, 9:02am

First of all, many thanks to @dalto for this tutorial.

I believe in KISS, the simpler the better.

As far as I understand, systemd-boot will not allow me to boot to earlier snapshots (I am on BTRFS, Snapper… etc), but can I say:
1- No matter what I will be able to boot.
2- Even if systemd-boot breaks it is much easier to fix.
3- Even if things break I still can boot and restore an earlier snapshot:
a- through command line
b- through a live session, install BTRFS Assitant, restore the earlier snapshot.

I hope I got it right this time!
Any feedback welcome.
Thank you.

limotux · August 27, 2022, 9:22am

Does this mean we do not go with the first post and just do:

yay -Syu eos-systemd-boot

I just want to be sure and not to mess things up!

Zesko · August 27, 2022, 9:56am

If Linux Kernel crashes or does not support your hardware, but you can not restore Linux Kernel when using systemd-boot, because Linux Kernel is in /efi using FAT32, not BTRFS. You need to install or fix Linux Kernel via arch-chroot in Live USB manually.

GRUB can help you to restore Linux Kernel because Kernel is in /boot using BTRFS, not FAT32, but you can not restore GRUB in /boot/efi using default FAT32.

That should be yay -S kernel-install-mkinitcpio instead eos-systemd-boot

limotux · August 27, 2022, 10:01am

@Zesko for the rescue again as usual.

So, I just

yay -S kernel-install-mkinitcpio

and reboot?
That’s all?

limotux · August 27, 2022, 10:05am

Sorry, this is a bit techie for me.
I installed LTS, and still have the default kernel as well.
I believe LTS kernel is safe enough, even though, there is another kernel to boot to! Right?!

Hey, you mean I can not live boot → BTRFS Assitant → Restore Snapshot?!

The problem now with Grub… I lost trust! Just my simple point of view!

Zesko · August 27, 2022, 10:11am

How many times has Grub crashed 2 years ago until now, I heard maybe only once.
Did you know many people have problems with Linux Kernels when using rolling releases?

No, you should read Dalto’s instructions

Just copy exact the commands and follow “kernel-install” instead “manual”

limotux · August 27, 2022, 10:27am

To me it is just a matter of principle!
systemd-boot is less prone to breaking and even if it breaks it is easier to fix (my humble knowledge till now), while Grub broke once in 30 years and was difficult to fix! So, better KISS, as a matter of principle! (sorry I might sound strange but… this is how I am… and maybe many of my age +60)

I know a rolling release is “theoretically” problematic and less stable. But actually it is working fine!

I go for a rolling release because it is an install it and forget it… (sort of), just update. I know there might be problems… but it is OK with me…

I believe an LTS and the standard/default kernel… I am OK, or at least in a little bit better situation!

Still I hope you advise me… is it just

yay -S kernel-install-mkinitcpio

and reboot?

P.S. Just thinking out loud with you… Is it possible to make systemd-boot consider snapshots as “other” operating systems and automatically point to them so you can boot to a snapshot?!

sradjoker · August 27, 2022, 10:32am

So one bad update/commit is bad principle?

limotux · August 27, 2022, 10:32am

OK… I will go for it!

limotux · August 27, 2022, 10:38am

Oh no… not exactly… the most important principle for me is the simplicity. Which is simpler and easier to fix! As long as there is software there are bugs!
So, breaking even a few times does not make any software “bad”.

Just simplicity, and well… it is serious and critical to have the boot loader break… better avoid it… to something less prone to break and easier to fix though systemd-boot can break simply because it is a software!. This is the principle.

limotux · August 27, 2022, 10:41am

OK, just for the record I am following @Zesko advice to follow @dalto 's instructions (first post in this thread - important to read) using

Now, updating and rebooting! (Done)

 du -sh /boot
127M    /boot

I have a lot of space!
Next we need to remove grub:

sudo pacman -Rc grub

sudo mkdir /efi

efidevice=$(findmnt /boot/efi -no SOURCE) # save the efi partition location

sudo umount /boot/efi

sudo mount ${efidevice} /efi

To make the mount change permanent, edit `/etc/fstab` and change where it reads `/boot/efi` to `/efi`

<dump>  <pass>
UUID=FDBB-97B1                            /efi      vfat    defaults,noatime 0 2

[limo@lenovo ~]$ sudo bootctl install
Created "/efi/EFI/systemd".
Created "/efi/loader".
Created "/efi/loader/entries".
Created "/efi/EFI/Linux".
Copied "/usr/lib/systemd/boot/efi/systemd-bootx64.efi" to "/efi/EFI/systemd/systemd-bootx64.efi".
Copied "/usr/lib/systemd/boot/efi/systemd-bootx64.efi" to "/efi/EFI/BOOT/BOOTX64.EFI".
Random seed file /efi/loader/random-seed successfully written (32 bytes).
Successfully initialized system token in EFI variable with 32 bytes.
Created EFI boot entry "Linux Boot Manager".

# Edit the file `/efi/loader/loader.conf` and uncomment the "timeout" line.

timeout 3
#console-mode keep

Technically speaking, you have now successfully installed systemd-boot. Congrats!

> kernel-install:

yay -S kernel-install-mkinitcpio

> You can either save the below as a script (I wonder what should I name this script? Where to save it?)
Here is the text of the script which I called dalto.sh

#!/usr/bin/env bash

# Find the configured esp
esp=$(bootctl -p)

# Prepare the efi partition for kernel-install
machineid=$(cat /etc/machine-id)
if [[ ${machineid} ]]; then
    mkdir ${esp}/${machineid}
else
    echo "Failed to get the machine ID"
fi

# Run kernel install for all the installed kernels
while read -r kernel; do
    kernelversion=$(basename "${kernel%/vmlinuz}")
    echo "Installing kernel ${kernelversion}"
    kernel-install add ${kernelversion} ${kernel}
done < <(find /usr/lib/modules -maxdepth 2 -type f -name vmlinuz)

Well I thought it can be just executed so I did:

[limo@lenovo MyScripts]$ sudo bash dalto.sh

and got lots of stuff With a few warnings about drivers and fonts… I get occasionally, and it ended up with:

=> WARNING: Possibly missing firmware for module: wd719x
==> WARNING: Possibly missing firmware for module: xhci_pci
  -> Running build hook: [keyboard]
  -> Running build hook: [keymap]
  -> Running build hook: [consolefont]
==> WARNING: consolefont: no font found in configuration
  -> Running build hook: [filesystems]
/dev/stdin: line 22: grub-btrfs-overlayfs: command not found
==> Generating module dependencies
==> Creating zstd-compressed initcpio image: /efi/2ba377af56ef4fb0898b46fcc244bcb9/5.19.4-arch1-1/initrd-fallback
==> Image generation successful

Which I believe is OK as “==> Image generation successful”
Then

sudo rm -r /boot/efi /boot/grub /boot/initramfs* /boot/vmlinuz*

As @dalto said there

Now you can reboot into your new system. As you install/remove kernels, the whole process should be automated.

Rebooting… hopefully I will be back soon… alive!

i am back… still alive and kicking!
the greatest of the great distros, and the greatest fo the great community.
Billions of thanks and appreciation to @dalto and @Zesko

[Tutorial] Convert to systemd-boot

To make the mount change permanent, edit /etc/fstab and change where it reads /boot/efi to /efi

To make the mount change permanent, edit `/etc/fstab` and change where it reads `/boot/efi` to `/efi`