LUKS passphrase rejected at the boot prompt (keymap, or intermittent)
Issue 8618 is real and the keymap diagnosis for flavour (a) is correct. cryptsetup open --test-passphrase <dev> and luksAddKey/luksHeaderBackup are all valid invocations. But the systemd-initramfs advice is wrong in a way that can brick a boot: it says a UKI 'as Omarchy builds' needs base systemd ... sd-vconsole ... sd-encrypt. A UKI is a packaging format, not an initramfs flavour. Omarchy's actual shipped array (omarchy-settings' /etc/mkinitcpio.conf.d/omarchy_hooks.conf, quoted verbatim in issue 8471) is busybox-based: base udev plymouth keyboard autodetect microcode modconf kms keymap consolefont block encrypt filesystems fsck btrfs-overlayfs. Pasting the systemd array swaps udev/encrypt for systemd/sd-encrypt and drops plymouth, and the cmdline still says cryptdevice= rather than rd.luks.*. That is an unbootable machine. Also missing: a LUKS header backup is a decryption-capable secret and must not sit in ~ on the encrypted disk it unlocks.
DATA LOSS. Removing or overwriting the wrong LUKS keyslot, or writing a stale header backup over a live header, destroys access to the encrypted volume permanently. There is no recovery. Always add a new key before removing an old one, and store the header backup off the encrypted disk.
Symptom
At the encryption prompt the correct passphrase is refused:
A password is required to access the cryptroot volume:
Enter passphrase for /dev/nvme0n1p2:
No key available with this passphrase.
Two distinct flavours are reported. (a) It *never* works from the boot prompt but the same passphrase works from a live USB, almost always a keyboard-layout problem. (b) It works after several tries with nothing changed, reported on a ThinkPad T14s Gen 1 running Omarchy 4.0.0 (basecamp/omarchy#8618, still open upstream).
Cause
(a) The initramfs uses the US layout unless a keymap is baked in, so any non-alphanumeric character on a non-US keyboard produces a different byte than the one you enrolled. Num Lock state changes digits typed on the numpad the same way. (b) The intermittent case on Omarchy has not been root-caused. The reporter ruled out header corruption, Plymouth and USB/keyboard timing (the machine uses an i8042 PS/2 keyboard with clean logs).
Fix
**First, prove the passphrase is fine** from a live USB:
sudo cryptsetup luksDump /dev/nvme0n1p2
sudo cryptsetup open --test-passphrase /dev/nvme0n1p2
If that succeeds, the passphrase is right and the problem is early boot.
**Check which initramfs flavour you actually have before editing anything**. Do not assume, and note that building a UKI does not make it systemd-based:
grep -h '^HOOKS' /etc/mkinitcpio.conf /etc/mkinitcpio.conf.d/*.conf
Set the console layout:
# /etc/vconsole.conf
KEYMAP=uk
Then add the keymap hooks **to the array you already have**, keeping its flavour:
- If your array contains
udevandencrypt(this is Omarchy's default, and it shipsbase udev plymouth keyboard autodetect microcode modconf kms keymap consolefont block encrypt filesystems fsck btrfs-overlayfs), ensurekeyboardandkeymapare present. Do not switch to systemd hooks. - Only if your array already contains
systemdandsd-encrypt, usesd-vconsolein place ofkeymap.
Mixing the two families also requires changing the kernel cmdline (cryptdevice= vs rd.luks.name=), so never swap flavours to fix a keymap.
Rebuild:
sudo mkinitcpio -P # Arch/EndeavourOS/CachyOS
sudo limine-mkinitcpio # Omarchy
**Safety net: add a purely-ASCII second passphrase** so a layout problem can never lock you out:
sudo cryptsetup luksAddKey /dev/nvme0n1p2
**Back up the LUKS header first.** Treat the backup as equivalent to the disk's contents: anyone holding it plus any passphrase that was valid *at backup time* can decrypt the disk, even after you later change that passphrase. Write it to removable media you keep offline, never to the encrypted disk itself:
sudo cryptsetup luksHeaderBackup /dev/nvme0n1p2 \
--header-backup-file /run/media/$USER/USBSTICK/luks-header-nvme0n1p2.img
sudo chmod 600 /run/media/$USER/USBSTICK/luks-header-nvme0n1p2.img
For the intermittent Omarchy case (#8618), keep retrying and attach journalctl -b -1 to the issue.
Verify
Reboot and type the passphrase using the same physical keys as in the OS. It is accepted first time. sudo cryptsetup luksDump /dev/nvme0n1p2 shows the expected number of enabled keyslots.