OPINIONATED OMARCHY
TROUBLESHOOTING CORPUS
790
RECORDS
790
AUDITED
0
UNCHECKED
2388
SOURCES
LIVE
gpu-drivers medium occasional DISPUTE

Fix Electron/Chromium apps stalling for a minute after boot on hybrid Intel+NVIDIA

omarchyarchendeavouroscachyosmanjarohyprlandwaylandnvidiaintellaptop
CORRECTED An auditor found the fix wrong or incomplete and rewrote it.
AUDIT NOTE

Checked against both cited pages fetched in full and against this omarchy 4.0.2-1 workstation (mkinitcpio 41.1-1, limine-mkinitcpio-hook 1.37.1-1, kernel 7.1.9-arch1-2). The symptom and the ordering fix are real and supported: the Hyprland NVIDIA page carries a warning box under "Early KMS, modeset and fbdev" stating that Electron or Chromium apps can stall for up to a minute after boot on Intel iGPU plus NVIDIA dGPU systems and that loading i915 before the NVIDIA modules fixes it. The second cited source does not support the claim. I fetched https://wiki.archlinux.org/index.php?title=NVIDIA&action=raw, 626 lines, and i915 appears zero times in it. It is kept on the record because it does support the surrounding mechanism, early module loading into the initramfs, and is the source of the hibernation caveat, but the ordering claim rests on the Hyprland page alone. Rewrote the cause to say that plainly: neither page gives a mechanism, so the record's original invented explanation about Electron enumerating the NVIDIA device first and blocking on it is unsourced and has been removed.

Four Omarchy 4 defects in the fix and its surroundings, all confirmed on this machine.

First, the fix did sudo tee /etc/mkinitcpio.conf.d/nvidia.conf, which overwrites a file Omarchy owns. /usr/share/omarchy/install/hardware/nvidia.sh:30 writes that exact path with cat > and the file here holds exactly MODULES+=(nvidia nvidia_modeset nvidia_uvm nvidia_drm). /etc/mkinitcpio.conf.d/omarchy_hooks.conf, owned by omarchy-settings 4.0.2-1, carries a comment naming nvidia.conf as sourced before it and reads MODULES to decide whether to drop the kms hook. Replaced that with a new drop-in, /etc/mkinitcpio.conf.d/00-i915.conf holding MODULES+=(i915), which sorts ahead of nvidia.conf and leaves Omarchy's file alone. The sort order is confirmed from mkinitcpio source at /usr/bin/mkinitcpio:1120, which builds the drop-in list with find ... -name '*.conf' -print0 | sed -z 's/.*\///' | LC_ALL=C.UTF-8 sort -zVu and then cats each onto the main config.

Second, sudo mkinitcpio -P fails here. /etc/mkinitcpio.d/ is empty, and /usr/bin/mkinitcpio:986 reads [[ -e "${_optpreset[0]}" ]] || die 'No presets found in %s' "$_d_presets". There is also a shim: /usr/local/bin/mkinitcpio, owned by limine-mkinitcpio-hook, is first on PATH and, on -P or -p, prints a warning that Limine entries were not updated and offers to run limine-mkinitcpio. The supported rebuild is sudo limine-mkinitcpio, which is what Omarchy's own migrations call.

Third, verify named /boot/initramfs-linux.img, which does not exist on Omarchy 4. /etc/limine-entry-tool.d/omarchy-uki.conf sets ENABLE_UKI=yes and omarchy-defaults.conf sets CUSTOM_UKI_NAME="omarchy", and /usr/share/libalpm/scripts/limine-mkinitcpio-install:95 builds UKI_PATH="${BOOT_PATH}/EFI/Linux/${UKI_PREFIX}_${KERNEL_NAME}.efi", so the image is /boot/EFI/Linux/omarchy_linux.efi. Replaced the primary check with journalctl -b -k -o short-monotonic | grep -iE 'i915|nvidia', which I ran here as an unprivileged user and which does show the nvidia load sequence with monotonic timestamps. The objcopy extraction is offered as a secondary check and was NOT exercised, because /boot is dmask=0077 and I have no sudo.

Fourth, the danger promised a fallback boot entry that does not exist on Omarchy 4. grep -rn MKINITCPIO_FALLBACK across /etc/limine-entry-tool.d/, /etc/default/limine and /usr/share/omarchy returns nothing, and limine-list on this machine prints Omarchy with one child linux, a Snapshots submenu with two entries, and a trailing EFI fallback. That last line is the EFI/BOOT/BOOTX64.EFI bootloader fallback from ENABLE_LIMINE_FALLBACK=yes, not a fallback initramfs. Rewrote the danger around the Snapshots submenu, which is the real recovery path, and kept the hibernation caveat both wiki pages give for early-loading NVIDIA modules.

Left severity medium and frequency occasional alone. The consequence is a slow first launch, not a broken machine, and it only reaches Intel plus NVIDIA hybrid laptops.

Not exercised: this is a single-GPU NVIDIA RTX 3090 desktop with no Intel iGPU, so the stall itself, the i915 load and the corrected drop-in could not be reproduced or tested. Nothing on this machine was modified, no initramfs was rebuilt, and mkinitcpio was read rather than run. The drop-in ordering, the preset failure and the UKI path are read from installed files and package scripts here, not from a live rebuild.

The cause above was rewritten on 2026-09-11 to match this note; the fix was corrected by the audit itself.
RISK

This rebuilds the initramfs, so a typo in the MODULES array can leave the machine unbootable.

Omarchy 4 has no fallback kernel entry to fall back on. MKINITCPIO_FALLBACK is not set in /etc/limine-entry-tool.conf, in /etc/limine-entry-tool.d/ or in /etc/default/limine, so no linux-fallback image is built. The EFI fallback line in the Limine menu is the EFI/BOOT/BOOTX64.EFI bootloader fallback from ENABLE_LIMINE_FALLBACK=yes and is not a second kernel entry.

Your recovery route is the Snapshots submenu in the Limine menu, from limine-snapper-sync. Confirm there is a recent snapshot before you rebuild:


limine-list

If there is none, boot the Omarchy ISO and chroot to repair. Adding only i915 is low risk in itself, because the NVIDIA modules were already early-loaded by Omarchy's own drop-in, so this changes order rather than content.

Separately, early-loading the NVIDIA modules at all, which Omarchy does by default on every machine with an NVIDIA card, is documented by both the Arch NVIDIA page and the Hyprland NVIDIA page as breaking resume from hibernation. This record does not change that either way, but Omarchy also ships /etc/mkinitcpio.conf.d/omarchy_resume.conf with HOOKS+=(resume), so hibernation is configured and may already be affected.

Symptom

On an Intel iGPU + NVIDIA dGPU laptop, the first launch of VS Code, Discord, Chromium etc. after boot hangs for up to a minute with a blank window before it finally appears. Later launches are fine.

Cause

Electron and CEF apps stall on their first launch after boot on Intel plus NVIDIA hybrid machines when the NVIDIA modules are loaded from the initramfs ahead of i915. Loading i915 first clears it. That is exactly what the Hyprland NVIDIA wiki page documents, in the warning box under "Early KMS, modeset and fbdev". Neither that page nor the Arch NVIDIA page explains why the order matters, and the Arch page does not mention i915 at all, so treat the ordering as an observed fix rather than an explained one.

On Omarchy 4 every hybrid machine starts in the bad order by default. /usr/share/omarchy/install/hardware/nvidia.sh writes /etc/mkinitcpio.conf.d/nvidia.conf containing MODULES+=(nvidia nvidia_modeset nvidia_uvm nvidia_drm) on any machine with an NVIDIA card, and /etc/mkinitcpio.conf ships stock with MODULES=(), so the NVIDIA modules are the first entries in the assembled list and nothing puts i915 before them.

Fix

Get i915 into the initramfs MODULES list ahead of the NVIDIA modules.

**Omarchy 4.** Do not edit /etc/mkinitcpio.conf.d/nvidia.conf. Omarchy's installer owns that file and rewrites it wholesale with cat >, and /etc/mkinitcpio.conf.d/omarchy_hooks.conf reads MODULES after it to decide whether to drop the kms hook. Add a separate drop-in that sorts ahead of it. mkinitcpio 41 concatenates /etc/mkinitcpio.conf.d/*.conf onto /etc/mkinitcpio.conf in sort -V order, so a filename starting with a digit lands first:


printf 'MODULES+=(i915)\n' | sudo tee /etc/mkinitcpio.conf.d/00-i915.conf

Check the assembled order before rebuilding. Read the files top to bottom in the order below and confirm i915 appears ahead of nvidia:


for f in /etc/mkinitcpio.conf /etc/mkinitcpio.conf.d/*.conf; do
  printf '== %s\n' "$f"
  grep -H MODULES "$f"
done

Then rebuild and reboot:


sudo limine-mkinitcpio
reboot

sudo mkinitcpio -P is wrong on Omarchy 4 and will not work. /etc/mkinitcpio.d/ is empty, so /usr/bin/mkinitcpio exits with No presets found in /etc/mkinitcpio.d, and /usr/local/bin/mkinitcpio, shipped by limine-mkinitcpio-hook, shadows the real binary on PATH and prompts to run limine-mkinitcpio for you instead of failing cleanly. Omarchy boots a unified kernel image that only limine-mkinitcpio rebuilds and re-registers.

**Plain Arch**, where MODULES normally lives in /etc/mkinitcpio.conf itself. Edit that line so i915 comes first:


MODULES=(i915 nvidia nvidia_modeset nvidia_uvm nvidia_drm)

sudo mkinitcpio -P
reboot

That branch does not apply on Omarchy 4, where /etc/mkinitcpio.conf is package-stock and unmodified with MODULES=(). Confirmed with pacman -Qkk mkinitcpio, which reports 103 files and 0 altered.

Verify

After a reboot the first cold launch of an Electron app returns promptly:


time chromium --version

Check the order the kernel actually took. This works on Omarchy 4 and on plain Arch, and needs no root:


journalctl -b -k -o short-monotonic | grep -iE 'i915|nvidia' | head

i915 should initialise before the nvidia, nvidia-modeset and nvidia-drm lines.

On Omarchy 4 do not reach for lsinitcpio /boot/initramfs-linux.img. That file does not exist. /etc/limine-entry-tool.d/omarchy-uki.conf sets ENABLE_UKI=yes and /etc/limine-entry-tool.d/omarchy-defaults.conf sets CUSTOM_UKI_NAME="omarchy", so the boot image is the unified /boot/EFI/Linux/omarchy_linux.efi, and /boot is a vfat ESP mounted dmask=0077 so an ordinary user cannot even list it. To read the bundled initramfs, pull it out of the UKI first, which needs binutils:


sudo objcopy -O binary --only-section=.initrd /boot/EFI/Linux/omarchy_linux.efi /tmp/initrd
lsinitcpio /tmp/initrd | grep -E 'i915|nvidia'

On plain Arch with a separate initramfs image, lsinitcpio /boot/initramfs-linux.img | grep -E 'i915|nvidia' still works.

Sources

igpu-electron-stall-i915-module-order