AMDGPU DisplayPort Hotplug Detection Fix

March 30, 2026 · View on GitHub

Part of the Patches Index.

Problem

On systems with AMD Radeon RX 7700 XT / 7800 XT (Navi 32, gfx1101) GPUs running Linux kernel 6.14+, DisplayPort monitors can fail to be detected at boot. The GPU misses the HPD (Hot-Plug Detect) signal during early initialization, leaving the monitor in a disconnected state despite being physically connected and powered on.

HDMI connections are unaffected because they use a simpler detection mechanism.

Symptoms

  • One or more DisplayPort monitors show as disconnected in kscreen-doctor -o or /sys/class/drm/card*-DP-*/status after boot
  • The monitor works fine if you power-cycle it (turn off and on) after boot
  • HDMI-connected monitors on the same GPU detect correctly
  • dmesg shows REG_WAIT timeout during display initialization: 
    amdgpu 0000:03:00.0: [drm] REG_WAIT timeout 1us * 100 tries - dcn32_program_compbuf_size line:138
  • No EDID data on the affected DP connector (/sys/class/drm/card1-DP-*/edid is empty)

Affected Hardware

  • GPU: AMD Radeon RX 7700 XT / 7800 XT (Navi 32, DCN 3.2)
  • Driver: amdgpu (kernel module)
  • Kernel: 6.14.0-15-generic (Ubuntu 25.10), likely affects other 6.x kernels
  • Display server: Wayland (KDE Plasma), but the issue is at the kernel/DRM level

Root Cause

DisplayPort requires an active handshake via the AUX channel + link training, unlike HDMI which is more passive. During early boot, the amdgpu display manager (amdgpu_dm) probes all connectors. If a DP monitor isn't ready fast enough (common with certain monitors that have slow power-on sequences), the AUX channel probe times out and the connector is marked as disconnected. The driver does not retry.

The REG_WAIT timeout in dcn32_program_compbuf_size confirms the display controller timed out during initialization of the display pipeline for that connector.

Fix

A systemd service that runs after the graphical target is reached, checks for disconnected DP ports, and forces the driver to re-scan.

Install

# Copy the rescan script
sudo cp dp-rescan.sh /usr/local/bin/dp-rescan.sh
sudo chmod +x /usr/local/bin/dp-rescan.sh

# Install the systemd service
sudo cp dp-rescan.service /etc/systemd/system/dp-rescan.service
sudo systemctl daemon-reload
sudo systemctl enable dp-rescan.service

How It Works

  1. Waits 5 seconds after graphical target for monitors to fully power on
  2. Checks if DP-2 shows as disconnected
  3. First attempts a soft re-detect via the connector status sysfs node
  4. If that fails, toggles the HPD (Hot-Plug Detect) mechanism on the display controller to force a full re-scan
  5. Logs all actions to the journal (journalctl -t dp-rescan)

Customization

If your affected port is not DP-2, edit dp-rescan.sh and change the connector path. Check your connectors with:

for conn in /sys/class/drm/card*-DP-*/status; do
    echo "$(basename $(dirname $conn)): $(cat $conn)"
done

After the Fix

After the DP port is re-detected, apply your screen layout. For KDE Plasma on Wayland:

kscreen-doctor \
  output.DP-2.enable output.DP-2.mode.1920x1080@60 output.DP-2.position.0,0 \
  output.HDMI-A-1.enable output.HDMI-A-1.mode.1920x1080@60 output.HDMI-A-1.position.1920,0 \
  output.DP-1.enable output.DP-1.mode.1920x1080@60 output.DP-1.position.3840,0

Diagnostics

Useful commands for investigating DP detection issues:

# Check connector status
cat /sys/class/drm/card1-DP-*/status

# Check if EDID is present (empty = no handshake)
test -s /sys/class/drm/card1-DP-2/edid && echo "EDID present" || echo "EDID missing"

# Check kernel logs for display issues
sudo dmesg | grep -i -E "REG_WAIT|hotplug|hpd|dp.*link|amdgpu_dm"

# Current display state
kscreen-doctor -o

# Check service logs
journalctl -t dp-rescan

License

MIT