Override a firmware EDID when the monitor reports garbage or none (KVM, long HDMI run, cheap adapter)
Almost everything is verified verbatim against the Arch KMS page: the quoted sentence about a skewed or missing EDID, /usr/lib/firmware/edid/, drm.edid_firmware=edid/file.bin, the CONNECTOR: scoping, the comma-separated multi-connector form, cat file > /sys/kernel/debug/dri/0/HDMI-A-2/edid_override, echo -n reset > ...edid_override, echo 1 > ...trigger_hotplug, the lockdown-mode caveat, echo edid/file.bin > /sys/module/drm/parameters/edid_firmware taking effect only for newly plugged displays, the video=<conn>:<xres>x<yres>[M][R][-<bpp>][@<refresh>][i][m][eDd] syntax, the video=DVI-I-1:1024x768@85 video=TV-1:d example, and the note that video= "is useful for all Wayland compositors". read-edid (providing get-edid) is in extra. The Limine claims hold: /etc/limine-entry-tool.d/omarchy-defaults.conf exists with exactly the KERNEL_CMDLINE[default]+=" ... " form and ENABLE_UKI=yes, limine-update is a real command Omarchy calls, and /etc/mkinitcpio.conf.d/ is a directory Omarchy already populates with kms in HOOKS (so the early-KMS warning is right). The gap is step 4: on Omarchy 4 the initramfs *and the UKI Limine actually boots* are rebuilt by limine-mkinitcpio, not by mkinitcpio -P. Omarchy's own scripts make this explicit: omarchy-hibernation-setup writes a KERNEL_CMDLINE drop-in and then runs sudo limine-mkinitcpio with the comment "limine-mkinitcpio rebuilds initramfs/UKI for all kernels and updates the /boot/limine.conf entries via limine-entry-tool", and omarchy-plymouth-set uses sudo mkinitcpio -P only as the fallback when limine-mkinitcpio is absent. A user who runs mkinitcpio -P and reboots may find the EDID silently absent from the UKI. Also minor: the connector-discovery loop's card?- pattern breaks on card10 and above.
A wrong or corrupt EDID binary can leave you with no display at all from early boot onward, before any TTY is usable. Always test with the edid_override debugfs path first. It is reversible with echo -n reset and survives nothing more than a reboot. Editing the kernel command line and running mkinitcpio -P/limine-update rewrites your boot entries: keep the fallback entry that Omarchy generates (ENABLE_LIMINE_FALLBACK=yes) and know how to reach the Limine menu, because if Direct Boot is on you may need to pick Limine from the firmware boot menu to get a rescue entry. Do not run mkinitcpio -P in the middle of an interrupted pacman/omarchy update transaction.
Symptom
Through a KVM switch, a 15 m HDMI run, or a USB-C→HDMI dongle, the monitor comes up at 1024x768 or 640x480 and hyprctl monitors all lists only two or three junk modes. The native 2560x1440 is simply not there. Sometimes the output shows up enabled but 0x0, or the display is not detected at all. Plugged straight into the GPU with a short cable it is perfect, so the panel is fine. The EDID is being mangled in transit.
Cause
The monitor's EDID is read over the DDC/I²C sideband channel, which is the first thing to fail on a marginal link. A KVM, a long passive cable, or a cheap active adapter can return a truncated, corrupted, or empty EDID. The kernel then falls back to a generic mode list and the compositor can only offer what the kernel gave it. Arch's KMS page states this directly: *"If your native resolution is not automatically configured or no display at all is detected, then your monitor might send none or just a skewed EDID file. The kernel will try to catch this case and will set one of the most typical resolutions."* The remedy is to hand the kernel a known-good EDID binary for that connector.
Fix
Steps 1, 2, 3 and 5 and the video= alternative are correct as written. Two changes.
**Step 1: make the connector listing robust.** ${con#*/card?-} only matches single-digit card numbers, so it silently mangles card10-DP-1 on machines with several DRM devices. Use:
for p in /sys/class/drm/card*-*/status; do
d=${p%/status}
printf '%s\t%s\n' "${d##*/}" "$(cat "$p")"
done
That prints the full directory name (card1-DP-1), which is what you feed to cp. The connector name Hyprland uses is the part after the first - (DP-1).
**Step 4: rebuild the right image.** On Omarchy 4 (and any Limine + limine-entry-tool setup with ENABLE_UKI=yes), the cmdline and FILES are baked into a UKI that plain mkinitcpio -P does not regenerate the boot entry for. Write the drop-in as the record says:
sudo tee /etc/mkinitcpio.conf.d/edid.conf >/dev/null <<'EOF'
FILES+=(/usr/lib/firmware/edid/u2723qe.bin)
EOF
Then rebuild with the Limine-aware command, exactly as Omarchy's own omarchy-hibernation-setup and omarchy-plymouth-set do:
if command -v limine-mkinitcpio >/dev/null; then
sudo limine-mkinitcpio # rebuilds initramfs + UKI for all kernels, updates /boot/limine.conf
else
sudo mkinitcpio -P # plain mkinitcpio setups only
fi
If you also added the KERNEL_CMDLINE drop-in in step 3, sudo limine-mkinitcpio covers both. sudo limine-update also works but additionally re-deploys the bootloader binary and rebuilds a second time, which is why Omarchy prefers limine-mkinitcpio for a cmdline-only change.
**Verify before rebooting away from a working display:**
lsinitcpio /boot/initramfs-linux.img | grep -i edid # or check the UKI was re-dated
cat /proc/cmdline # after reboot, confirm drm.edid_firmware is present
Verify
After reboot, hyprctl monitors all lists the monitor's full availableModes including the native one, and dmesg | grep -i edid shows the firmware EDID being loaded rather than a fallback. cat /sys/class/drm/card*-HDMI-A-2/edid | wc -c should be 128 or 256 bytes, not 0.
Sources
- https://wiki.archlinux.org/rest.php/v1/page/Kernel_mode_setting
- https://wiki.archlinux.org/title/Kernel_mode_setting
- https://raw.githubusercontent.com/basecamp/omarchy/quattro/etc/limine-entry-tool.d/omarchy-defaults.conf
- https://raw.githubusercontent.com/basecamp/omarchy/quattro/default/limine/default.conf
- https://raw.githubusercontent.com/hyprwm/hyprland-wiki/main/content/Configuring/Basics/Monitors.md