logind suspends laptop on DPMS-off when lid is closed

[Nikratio]

systemd version the issue has been seen with

258.7

Used distribution

Fedora 43

Linux kernel version used

6.19.13-200.fc43.x86_64

CPU architectures issue was seen on

x86_64

Component

systemd-logind

Expected behaviour you didn't see

I am using a laptop with a closed lid and an external display. When blanking the external display through DPMS, I expect this not to affect the state of the laptop.

Unexpected behaviour you saw

When powering off the screen, my system suspends.

This seems to be because logind enumerates DRM connectors and counts a connector only if its sysfs enabled attribute equals the literal string "enabled" - but when DPMS is on, it has a different value (at least for the amdgpu DRM driver, in report enabled based on whether the connector is currently driving an active CRTC, not whether the connector is physically present.)

Therefore, when a userspace compositor (swayidle in my case) sends a DPMS-off ioctl, the CRTC is detached and the kernel flips enabled from enabled to disabled — even though the cable is still plugged in and status remains connected.

Steps to reproduce the problem

1. Close laptop lid with a single external monitor connected via DisplayPort. Confirm HandleLidSwitch=suspend, HandleLidSwitchDocked=ignore (defaults), and IdleAction=ignore (default).
2. Confirm busctl get-property org.freedesktop.login1 /org/freedesktop/login1 org.freedesktop.login1.Manager Docked returns b true.
3. From a terminal on the laptop, run a DPMS-off command, e.g. on Sway: swaymsg "output * dpms off".
4. Observe: the laptop suspends within a fraction of a second.

Additional program output to the terminal or log subsystem illustrating the issue

before and after swaymsg "output * dpms off", on a Lenovo ThinkPad with a single DP monitor:                                                            
  
  # Before:                                                                                                                                                                            
  /sys/class/drm/card1-DP-3/status: connected (enabled=enabled)                                                                                                                      
  Docked = b true                                                                                                                                                                      
  
  # After:                                                                                                                                                                             
  /sys/class/drm/card1-DP-3/status: connected (enabled=disabled)                                                                                                                     
  Docked = b false     


Suggested patch (generated with Claude Code, not reviewed by me, I do not know enough about systemd):

In src/login/logind-core.c, manager_count_external_displays(), drop the enabled check entirely and rely on status != "disconnected" (which the code already does immediately afterwards).

Looking at the same function, there's a missing comma in the prefix list:                                                                   
  if (!STARTSWITH_SET(dash,
                      "VGA-", "DVI-I-", "DVI-D-", "DVI-A-"                                                                                                                             
                      "Composite-", "SVIDEO-", "Component-",                                                                                                                           
                      "DIN-", "DP-", "HDMI-A-", "HDMI-B-", "TV-"))                                                                                                                     

Adjacent string literal concatenation turns "DVI-A-" "Composite-" into the single token "DVI-A-Composite-", so DVI-A and Composite connectors are both never matched. (Probably    
  worth fixing in the same patch.)

[arvidjaar]

Looking at Intel GPU:

00:02.0 VGA compatible controller [0300]: Intel Corporation Raptor Lake-P [Intel Graphics] [8086:a7ab] (rev 04)

When the output is connected and DPMS is off in GNOME settings:

$ cat /sys/class/drm/card1-HDMI-A-1/{dpms,enabled,status}
On
enabled
connected
$

when the output is connected and DPMS is forced on in GNOME settings (busctl ... PowerSaveMode i 1):

$ cat /sys/class/drm/card1-HDMI-A-1/{dpms,enabled,status}
Off
disabled
connected
$

When the output is connected but disabled in the GNOME settings:

$ cat /sys/class/drm/card1-HDMI-A-1/{dpms,enabled,status}
Off
disabled
connected
$ 

when the output is logically disconnected (echo off > /sys/class/drm/.../status):

$ cat /sys/class/drm/card1-HDMI-A-1/{dpms,enabled,status}
Off
disabled
disconnected
$

Unfortunately it means there is no way to distinguish between "connector is disabled due to DPMS" and "connector is disabled in the configuration". How should we handle the situation when user explicitly disabled external monitor for whatever reason?

It sounds like something for kernel folks. They need to expose "this monitor is temporary out of service due to power saving" indication. Until then maybe systemd-logind could (optionally) ignore enabled attribute for the purpose of detecting external monitors.