After upgrading to NVIDIA 610.57.04, the external HDMI monitor on my hybrid AMD/NVIDIA laptop can enter a state where the desktop remains visible on the external monitor, but the display is effectively non-interactive.
X/Wayland still detects the monitor correctly. Under X11, the logical desktop geometry remains correct and the mouse pointer coordinates can move into the external monitor’s coordinate space, but the mouse is not usable/visible there and windows cannot be interacted with normally on that display.
The issue occurs across multiple desktop environments and display protocols:
- KDE Plasma – X11 and Wayland
- MATE – X11
- XFCE – X11
The laptop’s internal eDP display is connected to the AMD Vega 8 iGPU, while the HDMI output is connected to the NVIDIA RTX 2060 Mobile.
Downgrading from NVIDIA 610.57.04 to 580.178.04 resolves the problem. The same Linux 6.18 kernel works correctly with NVIDIA 580.178.04, including repeated HDMI hot-unplug/replug testing.
Hardware
Laptop: ASUS TUF Gaming FX505DV
GPUs:
NVIDIA TU106M [GeForce RTX 2060 Mobile]
PCI: 01:00.0
AMD Picasso/Raven 2 [Radeon Vega Series / Radeon Mobile Series]
PCI: 05:00.0
Internal display:
eDP
1920x1080
AMD Vega 8
External display:
ASUS MB16AMT
HDMI
1920x1080 @ 60 Hz
NVIDIA RTX 2060 Mobile
Broken configuration
Kernel: 6.18.45-1-MANJARO
NVIDIA: 610.57.04
Hybrid profile: video-hybrid-amd-nvidia-prime
The problem was reproduced using both the NVIDIA 610 open kernel module and proprietary kernel module.
With the open module:
linux618-nvidia-open 610.57.04
With the proprietary module:
linux618-nvidia 610.57.04
The display problem occurred with both.
Working configuration
Kernel: 6.18.45-1-MANJARO
NVIDIA: 580.178.04
Hybrid profile: video-hybrid-amd-nvidia-580xx-prime
Relevant kernel module:
linux618-nvidia-580xx-open 580.178.04
HDMI hotplug works normally with this configuration.
I tested this by placing a MATE Terminal window and the mouse on the external HDMI display, physically unplugging HDMI, and reconnecting it. The external monitor recovered normally and remained interactive.
X11 state while broken
xrandr --query still reported a valid 3840x1080 desktop:
Screen 0: minimum 320 x 200, current 3840 x 1080, maximum 16384 x 16384
eDP connected primary 1920x1080+0+0
HDMI-1-0 connected 1920x1080+1920+0
xrandr --listproviders:
Providers: number : 2
Provider 0:
Source Output, Sink Output, Source Offload, Sink Offload
AMD Radeon Vega 8 Graphics @ pci:0000:05:00.0
Provider 1:
Sink Output
NVIDIA-G0
xrandr --listmonitors:
Monitors: 2
0: +*eDP 1920/344x1080/194+0+0 eDP
1: +HDMI-1-0 1920/344x1080/194+1920+0 HDMI-1-0
Root desktop geometry was also correct:
_NET_DESKTOP_GEOMETRY(CARDINAL) = 3840, 1080
and:
dimensions: 3840x1080 pixels
Despite the external display being unusable, xdotool showed that X could move the pointer all the way to the far edge of the external monitor:
x:3839 y:731 screen:0
This suggests that the X screen/monitor geometry itself was not the cause.
PRIME
While the problem was present:
PRIME Synchronization: 0
supported: 0, 1
Attempting:
xrandr --output HDMI-1-0 --set "PRIME Synchronization" 1
returned exit status 0, but PRIME Synchronization remained 0.
Reconfiguring the HDMI output with xrandr did not resolve the underlying problem.
NVIDIA 610 open-module kernel errors
When using NVIDIA 610.57.04 with the open kernel module, the failure coincided with repeated NVRM errors:
NVRM: nvAssertFailedNoLog: Assertion failed: 0 @ alloc_free.c:263
NVRM: nvAssertOkFailedNoLog: Assertion failed:
Requested lock state not valid [NV_ERR_INVALID_LOCK_STATE] (0x0000002F)
NVRM: nvCheckOkFailedNoLog: Check failed:
Requested lock state not valid [NV_ERR_INVALID_LOCK_STATE] (0x0000002F)
returned from kgraphicsCreateBug4208224Channel_HAL(...)
NVRM: nvAssertFailedNoLog:
Assertion failed: 0 @ kernel_fifo.c:3108
NVRM: nvAssertOkFailedNoLog:
Requested lock state not valid [NV_ERR_INVALID_LOCK_STATE] (0x0000002F)
returned from gpuStateLoad(...)
@gpu_suspend.c:281
The errors occurred at approximately the same time the HDMI display entered the broken state.
The NVIDIA DRM parameters were:
nvidia_drm modeset: Y
nvidia_drm fbdev: Y
PreserveVideoMemoryAllocations: 1
DynamicPowerManagement: 2
DynamicPowerManagementVideoMemoryThreshold: 200
The NVIDIA GPU’s runtime PM state while testing was:
power/control: auto
power/runtime_status: active
Forcing:
power/control = on
did not resolve the problem.
Proprietary NVIDIA 610 test
I replaced linux618-nvidia-open with the proprietary linux618-nvidia module while retaining NVIDIA 610.57.04 userspace.
The external-monitor problem still occurred.
However, the NV_ERR_INVALID_LOCK_STATE messages seen with the open kernel module were no longer generated.
Therefore, those NVRM assertions appear related to the 610 open module, but are not necessarily the root cause of the external-display failure itself.
NVIDIA 580 test
I downgraded the hybrid driver to:
video-hybrid-amd-nvidia-580xx-prime
NVIDIA 580.178.04
with:
linux618-nvidia-580xx-open 580.178.04
using the same:
Linux 6.18.45-1-MANJARO
The external monitor then behaved normally.
Repeated HDMI unplug/replug testing, including unplugging the HDMI cable while an application and mouse pointer were actively on the external display, worked correctly.
Expected behaviour
The external HDMI display should remain usable after HDMI disconnect/reconnect and should behave as part of the extended desktop.
Actual behaviour with NVIDIA 610.57.04
The external display can remain visible and correctly detected by the display server, but becomes effectively non-interactive.
The display server continues to report the correct monitor layout and pointer coordinate space.
Regression comparison
Linux 6.18.45 + NVIDIA 610.57.04 open
BROKEN
NV_ERR_INVALID_LOCK_STATE errors observed
Linux 6.18.45 + NVIDIA 610.57.04 proprietary
BROKEN
NV_ERR_INVALID_LOCK_STATE errors not observed
Linux 6.18.45 + NVIDIA 580.178.04 open
WORKING
HDMI hot-unplug/replug works correctly
This appears to be a regression associated with the NVIDIA 610 driver branch on this hybrid AMD Vega 8 + NVIDIA RTX 2060 Mobile PRIME configuration rather than a desktop-environment-specific issue.
If there are no posts since your last post, edit that post instead include additional content. 
