omarchy update stops at migration 1789325478: "The Omarchy kernel has no Limine boot entry"
The installed /usr/share/omarchy/migrations/1789325478.sh is byte-identical to quattro and does what the cause says (omarchy-pkg-add of linux-omarchy and headers, BOOT_ORDER rewrite in /etc/default/limine, limine-mkinitcpio linux-omarchy, limine-entry-tool --tree check, exit 1). It is absent at v4.0.3 and present at v4.0.4. /usr/share/libalpm/scripts/limine-mkinitcpio-install prints 'mkinitcpio failed for kernel ..., skipping.' and process_kernel returns 0, confirmed. omarchy-migrate (identical to quattro) runs under set -euo pipefail with per-user markers in $HOME/.local/state/omarchy/migrations, and omarchy-update's set -e skips everything after it, confirmed. Issue 12044 supports the OOM make.log and the workaround, issue 5026's comment supports the 'dkms status installed but module not found, depmod fixes it' case. Three defects. (1) Fix step 2a ran 'dkms install --no-depmod' and then went straight to the UKI rebuild without depmod, which recreates the exact 'module not found' failure, and the 12044 reporter's working sequence includes depmod. (2) The cause stated that DKMS's quoting of 'make' exists so dkms's -j cannot lower it. The dkms.conf comment says the quoting is there to stop DKMS adding KERNELRELEASE. /usr/bin/dkms line 1603 only rewrites a leading bare make, so the loss of -j is a side effect, not the intent. (3) The cause presented issue 5026's '60-depmod runs before 70-dkms-install' as the mechanism, but /usr/share/libalpm/scripts/dkms runs depmod itself after each successful build (DKMS_DEPMOD=1 by default), so that mechanism is unconfirmed and was reworded. Not exercised: no NVIDIA rebuild or migration was run.
Do not reboot expecting the new kernel while limine-mkinitcpio linux-omarchy still prints skipping.: there is no linux-omarchy entry yet and the stock linux entry is the only working one, so do not remove linux at this stage. Running sudo omarchy-migrate replays every migration as root against root's home and can write root-owned files and settings where your user's belong.
Symptom
Updating to Omarchy 4.0.4 ends in the red "Something went wrong during the update!" screen. Just above it:
Running migration (1789325478)
Install the Omarchy kernel and make it the first Limine boot entry
Building UKI for linux-omarchy (7.2.5-3-omarchy)
...
==> ERROR: module not found: 'nvidia'
==> ERROR: module not found: 'nvidia_modeset'
==> ERROR: module not found: 'nvidia_uvm'
==> ERROR: module not found: 'nvidia_drm'
==> WARNING: errors were encountered during the build. The image may not be complete.
==> Creating unified kernel image: '/tmp/limine-mkinitcpio.BrnQbL/linux-omarchy.efi'
==> Unified kernel image generation successful
ERROR: mkinitcpio failed for kernel 7.2.5-3-omarchy, skipping.
The Omarchy kernel has no Limine boot entry; rerun omarchy-migrate after fixing the boot image build.
Every later omarchy update stops at the same place. The machine still boots, on the stock linux kernel.
Cause
Migration 1789325478.sh (Omarchy 4.0.4) installs linux-omarchy and linux-omarchy-headers, rewrites BOOT_ORDER in /etc/default/limine, then runs sudo limine-mkinitcpio linux-omarchy. limine-mkinitcpio-install prints mkinitcpio failed for kernel ..., skipping. and still exits 0 when mkinitcpio returns non-zero, so the migration checks limine-entry-tool --tree itself and exits 1 with that message when no linux-omarchy entry exists. The guard line is a consequence. The real failure is the line above it.
On NVIDIA machines mkinitcpio fails because /etc/mkinitcpio.conf.d/nvidia.conf (written by install/hardware/nvidia.sh) asks for nvidia nvidia_modeset nvidia_uvm nvidia_drm early and mkinitcpio cannot find those modules for 7.2.5-3-omarchy. Two situations are reported in issues 12044 and 5026:
- The
nvidia-open-dkmsbuild for the new kernel died. One reporter'smake.logshowedgcc: fatal error: Killed signal terminated program cc1, the OOM killer. The package'sdkms.confsetsMAKE[0]="'make' -j\nproc\...". Themakeis quoted to stop DKMS addingKERNELRELEASE, and a side effect is that DKMS's own-jsetting is not applied either, so the build always runs one job per CPU. - DKMS reports the modules as
installedfor the new kernel, but mkinitcpio still saysmodule not found, and runningdepmodfor that kernel fixes it. Issue 5026 attributes this to60-depmodrunning before70-dkms-install. Arch's DKMS hook script also runsdepmoditself after a successful build, so why the index was stale on those machines is not established. A manualdkms install --no-depmod, as in the 12044 workaround, also leaves it stale.
omarchy-migrate runs migrations in order under set -e and only writes the per-user marker after a migration succeeds, so this one stays pending, and every migration after it (including 1789444024, which installs missing kernel headers) is blocked behind it. The rest of the update after the migrate step (post-update hooks, AUR, mise, orphan cleanup, the reboot prompt) is skipped too.
Fix
Run these from a terminal in the working session (stock linux kernel). The versions below are the ones from the 4.0.4 reports. Use the ones dkms status and ls /usr/lib/modules/ print on your machine.
**1. Find which case you are in.**
uname -r
ls /usr/lib/modules/
pacman -Q linux-omarchy linux-omarchy-headers
dkms status
You need a line like nvidia/610.57.04, 7.2.5-3-omarchy, x86_64: installed. If headers are missing, install them (this passes Omarchy's pacman guard, which only stops -S combined with -u), and check that the headers version matches linux-omarchy:
sudo pacman -S --needed linux-omarchy-headers
pacman -Q linux-omarchy linux-omarchy-headers
**2a. DKMS shows no installed line for the -omarchy kernel: rebuild it and read why it failed.**
sudo dkms install nvidia/610.57.04 -k 7.2.5-3-omarchy
# on failure:
sudo tail -n 40 /var/lib/dkms/nvidia/610.57.04/build/make.log
journalctl -k -b | grep -iE 'out of memory|killed process'
If make.log shows Killed signal terminated program cc1, the build ran out of memory. Close large programs and limit the build to two jobs for this one run. The edit is to a package-owned file that the next nvidia-open-dkms upgrade overwrites:
sudo sed -i 's/-j`nproc`/-j2/' /usr/src/nvidia-610.57.04/dkms.conf
sudo dkms install nvidia/610.57.04 -k 7.2.5-3-omarchy
One report used sudo MAKEFLAGS="-j2" dkms install ... instead. GNU make gives a -j on its command line priority over MAKEFLAGS, so do not rely on that form.
**2b. In every case, refresh the module index for the new kernel before rebuilding the image.** This is the whole fix when DKMS already said installed but mkinitcpio said module not found.
sudo depmod 7.2.5-3-omarchy
modinfo -k 7.2.5-3-omarchy -F filename nvidia # must print a path, not an error
**3. Rebuild the Omarchy kernel's UKI and confirm the entry exists.**
sudo limine-mkinitcpio linux-omarchy
sudo limine-entry-tool --tree | grep linux-omarchy
The build output must not contain skipping.
**4. Finish the update.** Run it as your user. Never sudo omarchy-migrate: under sudo it reads root's state directory (/root/.local/state/omarchy/migrations), sees no markers there, and replays every migration as root.
omarchy update
That reruns the pending migrations, then the steps the failure skipped. The migration sets reboot-required when it completes. Reboot and check uname -r.
Verify
omarchy-migrate --pending prints nothing (exit status 1). sudo limine-entry-tool --tree lists linux-omarchy. After a reboot uname -r ends in -omarchy (unless Direct Boot is enabled, see the direct boot record), and nvidia-smi reports the driver.
Sources
- https://github.com/omacom/omarchy/issues/12044
- https://github.com/omacom/omarchy/issues/5026
- https://github.com/omacom/omarchy/blob/v4.0.4/migrations/1789325478.sh
- https://github.com/omacom/omarchy/blob/quattro/bin/omarchy-migrate
- https://github.com/omacom/omarchy/blob/v4.0.4/bin/omarchy-update
- https://wiki.archlinux.org/title/Dynamic_Kernel_Module_Support