[595.71.05] RTX 3060 / Xorg: DisplayPort monitor lost, no display until Xorg restart
Summary
The NVIDIA Linux display driver intermittently loses the connected
DisplayPort monitor and the Xorg session does not recover properly. The
monitor goes into sleep/no-signal mode, making local login impossible. The
only reliable recovery I have found is to kill Xorg and start a new login
session.
This did not happen before. It appears to be a regression introduced
sometime during the last 6-12 months (maybe even later).
It does not happen very often, but it has happened repeatedly.
Impact
When the problem happens:
- the monitor goes to sleep / no signal
- the existing Xorg session becomes unusable locally
- the display does not recover even though the monitor is later detected
again in the log - I have to kill Xorg to be able to log in again
Hardware
- GPU: NVIDIA GeForce RTX 3060 desktop, 12 GB
- PCI device:
01:00.0 VGA compatible controller: NVIDIA Corporation GA106 [GeForce RTX 3060 Lite Hash Rate] (rev a1)
- Monitor seen in the log: DELL U2715H
- Connection in the log: DisplayPort / DFP-1 / DP-0
Driver
NVIDIA-Linux-x86_64-595.71.05
Desktop/session
- Xorg session started by GDM
- XFCE desktop
- xfce4-screensaver is running
Expected result
If the monitor temporarily disconnects, goes to sleep, or is re-detected,
the Xorg session should recover and keep at least one usable display output.
Actual result
The driver/Xorg stack temporarily reports the display as disconnected,
removes all MetaModes for DP-0, then reports:
No enabled display devices found; starting anyway because
AllowEmptyInitialConfiguration is enabled
Shortly afterwards, the same monitor is detected again as connected, but the
session does not recover to a usable local display.
Relevant log excerpt
Jun 27 15:30:18 flashdance /usr/libexec/gdm-x-session[5550]: (--) NVIDIA(GPU-0): DFP-1: disconnected
Jun 27 15:30:18 flashdance /usr/libexec/gdm-x-session[5550]: (--) NVIDIA(GPU-0): DFP-1: Internal DisplayPort
Jun 27 15:30:18 flashdance /usr/libexec/gdm-x-session[5550]: (--) NVIDIA(GPU-0): DFP-1: 2670.0 MHz maximum pixel clock
Jun 27 15:30:18 flashdance xfce4-screensav[6087]: gdk_monitor_get_scale_factor: assertion 'GDK_IS_MONITOR (monitor)' failed
Jun 27 15:30:18 flashdance xfce4-screensav[6087]: gdk_monitor_get_workarea: assertion 'GDK_IS_MONITOR (monitor)' failed
Jun 27 15:30:18 flashdance /usr/libexec/gdm-x-session[5550]: (WW) NVIDIA(0): MetaMode "DP-0:2048x1152" could not be re-validated against
Jun 27 15:30:18 flashdance /usr/libexec/gdm-x-session[5550]: (WW) NVIDIA(0): the current hardware configuration; removing.
Jun 27 15:30:18 flashdance /usr/libexec/gdm-x-session[5550]: (WW) NVIDIA(0): MetaMode "DP-0:1920x1200" could not be re-validated against
Jun 27 15:30:18 flashdance /usr/libexec/gdm-x-session[5550]: (WW) NVIDIA(0): the current hardware configuration; removing.
Jun 27 15:30:18 flashdance /usr/libexec/gdm-x-session[5550]: (WW) NVIDIA(0): MetaMode "DP-0:1920x1080" could not be re-validated against
Jun 27 15:30:18 flashdance /usr/libexec/gdm-x-session[5550]: (WW) NVIDIA(0): the current hardware configuration; removing.
Jun 27 15:30:18 flashdance /usr/libexec/gdm-x-session[5550]: (--) NVIDIA(0): No enabled display devices found; starting anyway because
Jun 27 15:30:18 flashdance /usr/libexec/gdm-x-session[5550]: (--) NVIDIA(0): AllowEmptyInitialConfiguration is enabled
Jun 27 15:30:19 flashdance /usr/libexec/gdm-x-session[5550]: (--) NVIDIA(GPU-0): DELL U2715H (DFP-1): connected
Jun 27 15:30:19 flashdance /usr/libexec/gdm-x-session[5550]: (--) NVIDIA(GPU-0): DELL U2715H (DFP-1): Internal DisplayPort
Jun 27 15:30:19 flashdance /usr/libexec/gdm-x-session[5550]: (--) NVIDIA(GPU-0): DELL U2715H (DFP-1): 2670.0 MHz maximum pixel clock
The same disconnect/reconnect pattern is also visible earlier at
15:18:41, followed by repeated GDK_IS_MONITOR assertions from
xfce4-screensaver.
Reproduction
I do not have a deterministic reproducer. It appears to happen while the
desktop is idle/locked or while the display/screensaver power management is
active. The failure is intermittent.
Regression
This was not observed previously on the same system. The issue appeared
sometime during the last 6-12 months (maybe even later).
Recovery
Killing Xorg allows me to log in again. Without killing Xorg, the local
display remains unusable because the monitor is in sleep/no-signal mode.
Attachments
nvidia-bug-report-flashdance-20260627-155420-sanitized.log.gz (925.7 KB)
I ran nvidia-bug-report.sh via SSH immediately after the problem occurred, before killing Xorg, so the attached bug report should contain the broken display state.