Issues with porting sdbootutil to Arch Linux

[silverhadch]

PKGBUILD for Arch:

https://github.com/silverhadch/sdbootutil-arch/blob/main/PKGBUILD

Issues Encountered:

1. Snapshot Naming Conflict:

   sdbootutil assumes Btrfs snapshots follow the @root, @home format. However, Arch and CachyOS use numeric snapshot IDs (e.g., 1, 2, etc.), causing sdbootutil to fail when trying to locate the correct snapshot for bootloader configuration.

2. Bootloader Installation Error:

   systemd-boot installation fails because sdbootutil expects the kernel/initramfs inside a non-existent @-formatted snapshot path.

   The error message:

install: der Aufruf von stat für '/.snapshots/@/snapshot/usr/lib/systemd/boot/efi/systemd-bootx64.efi' ist nicht möglich: Datei oder Verzeichnis nicht gefunden

This happens because the tool isn't recognizing numeric snapshot IDs correctly.

3. Dropped Features for Compatibility:

• Joey: Removed due to being specific to openSUSE.
• TPC: Dropped because some of its functions depend on RPM-based package management, which is not present in Arch-based systems.
• Dracut: Replaced with Mkinitcpio, which is standard in Arch-based systems.

---

Current Roadblock:

The snapshot naming issue prevents sdbootutil from properly integrating with Snapper in Arch-based setups. The tool needs to be modified to fallback to checking if the snapshots are numbered and are at the path:

/.snapshots/<number>/snapshot/usr/lib/systemd/boot/efi/systemd-bootx64.efi

instead of expecting the @ format. Without this fix, sdbootutil cannot correctly generate boot entries or configure systemd-boot on Arch-based systems.

[LevitatingBusinessMan]

I am running into this issue myself, I am using OpenSUSE tumbleweed but my subvolumes are named differently.

my error is
install: cannot stat '/.snapshots/@opensuse/snapshot/usr/lib/systemd/boot/efi/systemd-bootx64.efi': No such file or directory

my root snapshots are saved at /@opensuse/.snapshots/

[aplanas]

The code is very aligned on how openSUSE was designed ... but happy to accept PRs that generalize the code (like the one that work without snapshots)

[LevitatingBusinessMan]

So with my configuration the root isn't a snapshot, which creates some problems.

I was able to run sdbootutil install by forcing have_snapshots=0: https://github.com/openSUSE/sdbootutil/blob/main/sdbootutil#L3097.

And I was then able to add my current kernels (with the exception of one that errored).

3097,3099c3097,3100
< 	have_snapshots=1
< 	root_subvol=$(btrfs subvol show / 2>/dev/null|head -1)
< 	subvol_prefix="${root_subvol%/.snapshots/*}"
---
> 	log_info Lying about having snapshots
> 	#have_snapshots=1
> 	#root_subvol=$(btrfs subvol show / 2>/dev/null|head -1)
> 	#subvol_prefix="${root_subvol%/.snapshots/*}"

[silverhadch]

The code is very aligned on how openSUSE was designed ... but happy to accept PRs that generalize the code (like the one that work without snapshots)

Well the main point of sdbotutil is having snapshots and i cant think of a way to make the snapshot detection flexible enough to dynamically detect the path.

[silverhadch]

@aplanas @LevitatingBusinessMan
I came up with this. I haven't tested this yet but i think it should work.

find_sdboot() {
    local prefix=""
    [ -z "$have_snapshots" ] || prefix="/.snapshots/${1-$root_snapshot}/snapshot"
    
    # Check standard @root format
    local sdboot="$prefix/usr/lib/systemd-boot/systemd-boot$firmware_arch.efi"
    [ -e "$sdboot" ] || sdboot="$prefix/usr/lib/systemd/boot/efi/systemd-boot$firmware_arch.efi"
    if [ -e "$sdboot" ]; then
        echo "$sdboot"
        return
    fi
    
    # If not found, search for numbered snapshots (allowing multiple digits)
    for snapshot in /.snapshots/*; do
        if [[ "$snapshot" =~ /.snapshots/[0-9]+$ ]]; then
            local alt_path="$snapshot/snapshot/usr/lib/systemd/boot/efi/systemd-boot$firmware_arch.efi"
            if [ -e "$alt_path" ]; then
                echo "$alt_path"
                return
            fi
        fi
    done
    
    # If nothing is found, print an error
    echo "Error: No valid systemd-boot path found!" >&2
    exit 1
}

[LevitatingBusinessMan]

superficially that appears to work

[silverhadch]

I think i must do that for every hardcoded path...