Fix a docked external display staying dark after a long suspend
The problem is real (basecamp/omarchy#7328 exists with a matching title) and the sysfs force-reprobe is a legitimate workaround, but three things need fixing. (1) The connector example is wrong for the command given: ls -d /sys/class/drm/card*-* yields /sys/class/drm/card1-DP-5, not /sys/class/drm/card1/card1-DP-5. (2) The reasoning about echo detect is inverted, and the consequence matters: in drm_sysfs.c, status_store calls fill_modes for detect too (the || !connector->force branch), so detect *does* re-probe. What none of off/on/detect does is emit a hotplug uevent. That is why the compositor often does not notice, and why a hyprctl reload is needed afterwards (the sibling records for the same technique include one, but this one omits it). (3) The resume hook is genuinely harmful as written: it forces every DP connector off for a second on every single resume, including the ones that came back fine, which shuffles workspaces and windows on healthy multi-monitor setups. It also belongs in /etc/systemd/system-sleep/, because /usr/lib/systemd/system-sleep/ is the vendor directory for packages.
The sleep hook runs as root and writes to every DisplayPort connector on every resume. On a machine where the display is already working this forces a brief blank/re-modeset. Test the manual commands first and narrow the glob to your specific connector.
Symptom
Laptop in clamshell on a Thunderbolt/USB4 dock. Short sleeps come back fine, but after ~40 minutes or more of suspend the machine wakes (keyboard works, hyprctl answers, fans spin) and the external monitor never lights up. It sits in standby until you open the lid or reboot. Journal shows [drm] Skip DMUB HPD IRQ callback in suspend/resume and xhci_hcd 0000:03:00.0: Controller not ready at resume -19.
Cause
On resume, amdgpu discards hotplug interrupts raised inside the resume window, so re-detection depends on the dock re-asserting HPD *after* resume. After a long s2idle the Thunderbolt tunnel does not fully come back and that HPD never arrives. The kernel keeps the stale pre-suspend connected state, so nothing looks wrong from userspace and nothing re-modesets. This is a kernel/DRM state problem, not a compositor one, and hyprctl reload does not help.
Fix
Force the kernel to re-probe the connector (needs root). Find your connector path first:
ls -d /sys/class/drm/card*-* # e.g. /sys/class/drm/card1-DP-5
Then toggle it off and back on. Writing to status re-probes but emits no hotplug uevent, so nudge the compositor afterwards:
CONN=/sys/class/drm/card1-DP-5
echo off | sudo tee $CONN/status
sleep 1
echo on | sudo tee $CONN/status
echo detect | sudo tee $CONN/status
hyprctl reload
To automate it, drop a resume hook, but only re-probe connectors that came back *dis*connected, so a working monitor is never blanked on every resume:
sudo tee /etc/systemd/system-sleep/99-reprobe-drm <<'EOF'
#!/bin/bash
[ "$1" = post ] || exit 0
for c in /sys/class/drm/card*-DP-*; do
[ -e "$c/status" ] || continue
[ "$(cat "$c/status")" = disconnected ] || continue
echo off > "$c/status"; sleep 1
echo on > "$c/status"; echo detect > "$c/status"
done
EOF
sudo chmod +x /etc/systemd/system-sleep/99-reprobe-drm
(Use /etc/systemd/system-sleep/, not /usr/lib/systemd/system-sleep/: the latter is the package-owned vendor directory.)
Verify
After a >40 minute suspend, the external panel lights up on wake with the lid still closed, and hyprctl monitors reports the correct mode.