OPINIONATED OMARCHY
TROUBLESHOOTING CORPUS
790
RECORDS
790
AUDITED
0
UNCHECKED
2388
SOURCES
LIVE
power-suspend high very-common DISPUTE

Fix a black or frozen screen after ordinary suspend on AMD graphics

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

Checked on this workstation (omarchy 4.0.2-1, omarchy-settings 4.0.2-1, Hyprland 0.56.2-1, kernel 7.1.9-arch1-2). It is a DESKTOP with an NVIDIA card, so there is no AMD GPU here and the symptom, the parameters and the suspend cycles could NOT be exercised. Everything below is a config and source check. What held: modinfo -p amdgpu on kernel 7.1.9 still lists all three parameters (dcdebugmask, sg_display, runpm), so none has been removed. wiki.archlinux.org/title/AMDGPU section 'Frozen or unresponsive display (flip_done timed out)' still gives 0x10 and 0x12 with the record's near-verbatim wording, 'Screen flickering white/gray' still gives amdgpu.sg_display=0, and the dGPU power-management section still gives amdgpu.runpm=0. I read bbs.archlinux.org/viewtopic.php?id=309052 in full: it is dated 2025-10-03, its dmesg matches the record's symptom block verbatim, and the reporter's closing post shows 6.16.10-arch1-1, so the record's 'resolved purely by moving from 6.16.7 to 6.16.10' claim is exactly right. /usr/bin/omarchy-update-pacman-guard aborts only when both sync and sysupgrade are present, so pacman -S linux-lts is allowed as the record says. Nothing under /usr/share/omarchy mentions amdgpu at all, so the advice collides with nothing Omarchy ships. Three things were wrong. First and worst, step 2. /etc/limine-entry-tool.d/omarchy-uki.conf sets ENABLE_UKI=yes, so the command line is embedded in /boot/EFI/Linux/omarchy_linux.efi, and systemd-stub's own documentation says an invocation command line is used to 'override' the embedded one and is ignored outright under Secure Boot. Limine's CONFIG.md confirms cmdline is a general entry option that applies to the efi protocol. So a menu edit holding only amdgpu.dcdebugmask=0x10 replaces the whole command line and drops cryptdevice=, root= and rootflags=subvol=@, which is a failed boot, not the 'no config edits, nothing to undo' step the record promises. The rewrite prints the real command line with limine-entry-tool --get-cmdline linux, which I ran unprivileged here and which returned the full string. Second, step 3 named /etc/default/limine, which does work (it is loaded last and highest priority per /usr/lib/limine/limine-common-functions lines 99 to 129, and Omarchy's own copy uses +=), but Omarchy 4 keeps its command line in /etc/limine-entry-tool.d/*.conf and ships resume.conf and omarchy-defaults.conf there, so the drop-in is now the primary branch with /etc/default/limine kept as the labelled alternative. Third, the symptom was not distinguishable from a lock screen failure, which is the specific Omarchy 4 trap here: omarchy-system-sleep-monitor holds a logind delay inhibitor and runs omarchy-system-sleep-lock before every suspend, so you always resume into an ext-session-lock surface drawn by omarchy-shell, and a lock that fails to draw gives the identical 'black screen, ssh works' picture. Step 1 now separates the two with omarchy-hyprland-session-locked before anything is changed. Minor fixes folded in: sudo limine-update after installing linux-lts is redundant because 90-mkinitcpio-install.hook already builds and registers the entry, and the danger now names Snapshots as the recovery path because Omarchy 4 builds no fallback kernel entry (MKINITCPIO_FALLBACK is unset everywhere and limine-list shows only Omarchy > linux, Snapshots and the EFI fallback bootloader). I could not read /boot/limine.conf to see whether a UKI entry carries a cmdline: line at all, because /boot is mounted dmask=0077 and I did not use sudo. Sources: the cited raw.githubusercontent.com/basecamp/omarchy/master/default/limine/default.conf still returns 200, but master is the branch this project treats as stale and its content no longer matches what 4.0.2 ships, which splits that file between /etc/default/limine (ESP_PATH plus the cmdline line) and /etc/limine-entry-tool.d/omarchy-defaults.conf (everything else). I replaced it with the two quattro paths I fetched through the gh API.

The cause above was not rewritten and may still contain the error described. The fix below is the corrected version.
RISK

amdgpu.runpm=0 stops the discrete GPU powering down at runtime. On a hybrid-graphics laptop that costs several watts of idle battery and makes the machine run hotter, so only keep it if it is genuinely the fix.

On Omarchy 4 the Limine menu editor replaces the UKI's embedded command line rather than adding to it, so a cmdline: line holding only the new parameter boots to an initramfs emergency shell with no root filesystem. Copy the whole command line from limine-entry-tool --get-cmdline linux before you edit. The edit is not persisted, so a plain reboot recovers.

Adding a drop-in and running limine-mkinitcpio or limine-update rewrites your boot entries. Keep the Limine menu visible (do NOT enable Setup > Direct Boot while you are experimenting). Omarchy 4 builds no fallback kernel entry, so if the machine stops booting the recovery path is a Snapshots entry, not a fallback entry.

Symptom

I close the lid (or suspend the desktop), open it again, and the screen stays black. The machine is clearly alive (the Caps Lock LED toggles, I can SSH in, music keeps playing) but nothing ever draws. Sometimes the panel comes back but frozen or full of corruption. journalctl -b -1 -k has:


amdgpu 0000:03:00.0: amdgpu: SMU: I'm not done with your previous command: SMN_C2PMSG_66:0x0000003A
amdgpu 0000:03:00.0: amdgpu: RunBtc failed!
amdgpu 0000:03:00.0: amdgpu: Failed to setup smc hw!
amdgpu 0000:03:00.0: amdgpu: resume of IP block <smu> failed -62
[drm:dc_dmub_srv_wait_idle] *ERROR* [CRTC:82:crtc-0] flip_done timed out

Cause

Several distinct amdgpu resume bugs land on the same symptom. On laptop eDP panels the usual culprit is PSR (Panel Self Refresh): the panel keeps showing its own stale frame and the display controller never re-arms after resume, producing flip_done timed out. On dGPUs the SMU (power-management microcontroller) can fail to re-initialise, giving resume of IP block <smu> failed. A third variant is scatter-gather display on APUs, which flickers white or stays blank when the framebuffer is re-created. All three are workaround-able from the kernel command line, and several have been fixed and re-broken across kernel releases. One Arch BBS report was resolved purely by moving from 6.16.7 to 6.16.10.

Fix

**1. Rule out the lock screen first, then confirm which amdgpu failure you have**

On Omarchy 4 every suspend is preceded by a lock. omarchy-system-sleep-monitor holds a logind delay inhibitor and runs omarchy-system-sleep-lock, which asks omarchy-shell for an ext-session-lock surface before the machine sleeps. So you always resume into a lock screen, and a lock surface that fails to draw looks exactly like this record's symptom: black panel, machine alive, ssh works. Separate the two before touching the kernel command line:


omarchy-hyprland-session-locked          # LOCK in solitaryBlockedBy means the compositor is locked
sudo journalctl -b -1 -k --no-pager | grep -iE 'amdgpu|drm|flip_done|PM: suspend'

If the compositor reports itself locked and the journal shows no amdgpu, flip_done or resume of IP block lines, this is a lock screen or compositor problem, not an amdgpu resume bug, and nothing below will help. Omarchy 4's lock cannot be released headlessly, so type the password on the console or use virsh send-key on a VM.

**2. Try the parameter live at the boot menu, but type the whole command line**

Omarchy 4 boots a UKI with the command line embedded in its .cmdline PE section. systemd-stub uses an invocation command line **instead of** the embedded one, not in addition to it, and ignores it altogether when Secure Boot is on. So a cmdline: line typed at the Limine menu that holds only the new parameter drops cryptdevice=, root= and rootflags=subvol=@, and the machine lands in an initramfs emergency shell. Print the real command line first:


limine-entry-tool --get-cmdline linux --no-mutex --no-hooks

At the Limine menu press e on the Omarchy entry, set cmdline: to that entire string plus the parameter, and boot:


cmdline: <everything the command above printed> amdgpu.dcdebugmask=0x10

Nothing is written to disk, so a reboot undoes it. 0x10 disables PSR v1 and Panel Self Refresh-Selectively Updated. If that is not enough, try amdgpu.dcdebugmask=0x12, which additionally disables memory stutter mode.

Other candidates, one at a time:

  • amdgpu.sg_display=0 for laptops and APUs that flicker white or stay blank when a display is attached or re-attached.
  • amdgpu.runpm=0 for discrete AMD cards that vanish or fail to come back after being powered down.

All three parameters exist on kernel 7.1.9: modinfo -p amdgpu | grep -E 'dcdebugmask|sg_display|runpm'.

**3. Make the winning parameter permanent on Omarchy 4 (Limine + UKI)**

Omarchy sets its own command line from drop-ins, so add one of your own rather than editing a shipped file. Any name works, += is what makes it append:


printf 'KERNEL_CMDLINE[default]+=" amdgpu.dcdebugmask=0x10"\n' \
  | sudo tee /etc/limine-entry-tool.d/zz-amdgpu.conf
sudo limine-mkinitcpio
sudo reboot

/etc/default/limine also works and is read last, after every drop-in, so += there appends too. It is owned by no package, so it takes no .pacnew. Either file is fine, the drop-in just matches how Omarchy ships resume.conf. sudo limine-update is the same thing plus a bootloader reinstall.

On GRUB systems instead: add it to GRUB_CMDLINE_LINUX_DEFAULT in /etc/default/grub, then sudo grub-mkconfig -o /boot/grub/grub.cfg. On systemd-boot: append it to /etc/kernel/cmdline or the entry's options line. Neither applies to Omarchy 4.

**4. If no parameter helps, it is a kernel regression, so change kernels rather than keep guessing**


sudo pacman -S linux-lts linux-lts-headers

The 90-mkinitcpio-install.hook pacman hook builds the UKI and registers the Limine entry for you, so no further command is needed. This is not caught by Omarchy's update guard, which aborts only when both -S and -u are present. Reboot into linux-lts from the Limine menu. If resume works there, stay on it until the mainline fix lands and re-test after each omarchy update.

**5. Firmware**

AMD laptop resume bugs are frequently EC or BIOS side. Check for an update:


sudo fwupdmgr refresh --force && sudo fwupdmgr get-updates

On Omarchy this is also exposed as Update > Firmware.

Verify

cat /proc/cmdline shows the parameter after reboot. Run three suspend/resume cycles (sudo systemctl suspend, wake, repeat) and then sudo journalctl -b -k | grep -iE 'flip_done|resume of IP block|SMU'. A clean run has no matches, and grep -c 'PM: suspend exit' <<< "$(journalctl -b -k)" counts the resumes that completed.

Sources

amdgpu-black-screen-on-resume-from-suspend