Secure Boot Violation after a Limine-only update, with Secure Boot enforced
Checked on omarchy 4.0.2-1 (limine 12.6.0-1, limine-mkinitcpio-hook 1.37.1-1, sbctl NOT installed here, so every sbctl claim is from source). Held: sb_sign() in /usr/lib/limine/limine-common-functions calls sbctl sign with no -s and is byte-identical to the upstream 1.37.1 tag. sbctl.go Sign() only writes the file database when the enroll flag is set, and SignAll() iterates only that database. pacman matches hook targets with fnmatch(pattern, path, 0), so the glob is case-sensitive as the reporter said. /etc/pacman.d/hooks/99-omarchy-limine.hook is present here, owned by no package, and _write_limine_pacman_hook still writes it at omarchy-iso 1daf120. Wrong or missing: (1) the record credits the missing --save to itself. The second reporter (dot3x3q, 2026-09-14 comment) found it, and the record must say so for gate 1. (2) The 80 hook is owned by limine-mkinitcpio-hook (pacman -Qo), so the harvester was right that the reporter misattributed it, but the omission was wrong: the harvester compared the wrong upstream copy, install/arch-linux/limine-entry-tool/. The copy the package ships, install/arch-linux/limine-mkinitcpio-hook/usr/share/libalpm/hooks/80-limine-efi-deploy.hook at the same tag, is byte-identical to the installed file and is now cited. (3) The symptom and fix say any update touching limine or limine-mkinitcpio-hook. The installer hook triggers on Upgrade of limine only, so a limine-mkinitcpio-hook-only transaction leaves the file signed. (4) The title says Limine-only, and the second report is a kernel-bump transaction. (5) The fix claims -s makes sign-all cover the file on future updates. zz-sbctl.hook still does not fire on a limine-only transaction, so the check stays necessary, and the durable mitigation both reporters reached, disabling the installer hook, was missing. (6) danger claims the reporter had to sbctl create-keys. The issue says disabling Secure Boot can also mean re-doing custom key enrollment, nothing about new keys. The Option ROM warning was uncited. It is in docs/sbctl.8.txt at the pinned sha and is now cited. (7) Missing from the record: ENABLE_ENROLL_LIMINE_CONFIG=yes makes sbctl sign -s alone insufficient (second reporter), and upstream PR #12242, which removes the hook by migration, was closed unmerged, so 4.0.2-1 has no migration. Severity lowered to medium against the brief's scale: nothing is exposed by the defect, the harm is a hardening control (Secure Boot) left off by the obvious wrong fix. Not exercised: no sbctl, no hook, no ESP or firmware state was touched, and the chroot recovery was not run.
Do not answer the Secure Boot Violation by turning Secure Boot off in firmware setup and leaving it off. The issue's reporter writes that on their ASUS board disabling it can also mean redoing custom key enrollment from scratch, and until it is back on the machine has no pre-boot integrity check at all. Re-sign the file from a live system or chroot instead. If you do have to enroll again, keep -m: sbctl's manual says some devices have firmware that is signed and validated when Secure Boot is enabled, that failing to validate it could brick devices, and that it is recommended to enroll your own keys with Microsoft certificates.
If ENABLE_ENROLL_LIMINE_CONFIG=yes is set, sbctl sign -s on its own is not enough: the overwrite also dropped the enrolled config checksum and Limine stops on a config hash mismatch. Run sudo limine-enroll-config first.
When disabling the installer hook, do not touch 80-limine-efi-deploy.hook by mistake. It is the one that deploys and signs the loader on every limine and limine-mkinitcpio-hook upgrade.
Symptom
On a machine with Secure Boot enforced and sbctl custom keys enrolled, an omarchy update that upgrades the limine package, with or without a kernel in the same transaction, leaves the machine unable to boot. The firmware reports a Secure Boot Violation before Limine, the LUKS prompt or Plymouth runs. sudo sbctl verify, run before the reboot or from a chroot afterwards, lists /boot/EFI/limine/limine_x64.efi as not signed.
Cause
Three pacman hooks run on a limine upgrade, and their order is the defect.
/usr/share/libalpm/hooks/80-limine-efi-deploy.hook, owned bylimine-mkinitcpio-hook(not bylimineas the issue's reporter wrote, confirmed withpacman -Qoat 1.37.1-1), runs/usr/bin/limine-install, which copies/usr/share/limine/BOOTX64.EFIto/boot/EFI/limine/limine_x64.efiand signs it throughsb_sign()in/usr/lib/limine/limine-common-functions. That function callssbctl sign "$file_path"with no-s/--saveflag, identical on this workstation and in the upstream 1.37.1 tag, so the file is signed but never registered in sbctl's file database.
/etc/pacman.d/hooks/99-omarchy-limine.hook, owned by no package, is written by the Omarchy ISO installer (_write_limine_pacman_hookinomarchy-iso'sphases_impl.py, still present in the current installer) and triggers onOperation = Upgradeoflimineonly. It runs after hook 1 and copies the raw, unsigned package binary over the file hook 1 just signed:
Exec = /bin/sh -c "/usr/bin/cp /usr/share/limine/BOOTX64.EFI /boot/EFI/limine/limine_x64.efi"
zz-sbctl.hook, from thesbctlpackage, runssbctl sign-all -gand is the safety net meant to catch this. It cannot, for two independent reasons.sign-allonly iterates files already in sbctl's database (SignAll()reads the database and signs each entry), and step 1 never put this file there. And pacman matches the hook's path triggers withfnmatch()and no case folding, so on a transaction that upgrades onlylimine, whose sole EFI file isusr/share/limine/BOOTX64.EFI, the lowercaseusr/share/**/*.efi*trigger never matches and the hook does not run at all. On a transaction that also bumps the kernel the hook does run, becauseusr/lib/modules/*/vmlinuzmatches, and still signs nothing for this file.
The case-sensitive glob was found by the issue's reporter. The missing --save, which is what makes the glob beside the point, was found by the second reporter in the same thread, who hit the failure on a combined Omarchy 4.0.2 to 4.0.3 update with a kernel bump. Both are in omacom/omarchy#10945, open as of 2026-09-18. An upstream pull request, #12242, that deletes the installer hook by migration was closed without merging, and no such migration exists in 4.0.2-1, so every limine upgrade repeats the overwrite.
If you have set ENABLE_ENROLL_LIMINE_CONFIG=yes in /etc/default/limine (off by default, and unset on this workstation), the overwrite also discards the enrolled config checksum, so re-signing alone leaves Limine stopping on a config hash mismatch.
Fix
After any update that upgraded limine, check before rebooting:
sudo sbctl verify
If /boot/EFI/limine/limine_x64.efi is listed as not signed, sign it and register it in sbctl's database in the same step:
sudo sbctl sign -s /boot/EFI/limine/limine_x64.efi
sudo sbctl verify
-s is the flag the packaged signing path omits. Once the file is in the database, the plain sbctl sign that limine-install runs keeps it there, and sbctl sign-all covers it whenever zz-sbctl.hook fires. That hook still does not fire on a transaction that upgrades only limine, so the check above stays necessary after every such update for as long as 99-omarchy-limine.hook exists.
To stop the overwrite instead of repairing after it, take the installer hook out of pacman's path. 80-limine-efi-deploy.hook already deploys and signs the same file, which is the fix the issue's reporter proposed and the second reporter applied:
sudo mv /etc/pacman.d/hooks/99-omarchy-limine.hook /etc/pacman.d/hooks/99-omarchy-limine.hook.disabled
pacman reads only files ending in .hook, and no package owns this one, so nothing reinstalls it. If a later Omarchy migration removes it first, the mv finds nothing and that is fine.
If /etc/default/limine has ENABLE_ENROLL_LIMINE_CONFIG=yes, run sudo limine-enroll-config before sbctl sign -s, so the config checksum is enrolled again before the binary is signed.
If the machine is already locked out, boot the Omarchy ISO or any live Linux, switch to a second console, and run the same command from a chroot, since sbctl's keys live in /var/lib/sbctl/keys on the encrypted root:
cryptsetup open /dev/<your LUKS partition, from lsblk> root
mount -o subvol=@ /dev/mapper/root /mnt
mount /dev/<your ESP, the vfat partition from lsblk> /mnt/boot
arch-chroot /mnt
sbctl sign -s /boot/EFI/limine/limine_x64.efi
sbctl verify
Verify
sudo sbctl verify lists /boot/EFI/limine/limine_x64.efi as signed, sudo sbctl list-files includes it, and the machine boots with Secure Boot enforced. If you disabled the installer hook, ls /etc/pacman.d/hooks/ shows 99-omarchy-limine.hook.disabled and no 99-omarchy-limine.hook.
Sources
- https://github.com/omacom/omarchy/issues/10945
- https://github.com/omacom/omarchy-iso/blob/1daf1202d49147223341c8e88ed2cb5a65b75bed/configs/airootfs/usr/share/omarchy-iso/orchestrator/phases_impl.py
- https://gitlab.com/Zesko/limine-entry-tool/-/raw/26b1879c862a55bae4e8777d48b2f917c68ef347/install/arch-linux/limine-entry-tool/usr/lib/limine/limine-common-functions
- https://github.com/Foxboron/sbctl/blob/6f234197abc07b56092e0043b448f8b630dea7e4/cmd/sbctl/sign.go
- https://github.com/Foxboron/sbctl/blob/6f234197abc07b56092e0043b448f8b630dea7e4/cmd/sbctl/sign-all.go
- https://github.com/Foxboron/sbctl/blob/71ce775145dc4d01ef419804f445873ee2c837a7/contrib/pacman/ZZ-sbctl.hook
- https://github.com/omacom/omarchy/pull/12242
- https://gitlab.com/Zesko/limine-entry-tool/-/raw/26b1879c862a55bae4e8777d48b2f917c68ef347/install/arch-linux/limine-mkinitcpio-hook/usr/share/libalpm/hooks/80-limine-efi-deploy.hook
- https://github.com/Foxboron/sbctl/blob/6f234197abc07b56092e0043b448f8b630dea7e4/sbctl.go
- https://github.com/Foxboron/sbctl/blob/6f234197abc07b56092e0043b448f8b630dea7e4/docs/sbctl.8.txt
- https://gitlab.archlinux.org/pacman/pacman/-/blob/master/lib/libalpm/util.c