Erros message nvidia [drm:nvidia_drm_gem_import_nvkms_memory [nvidia_drm]

When I enable modesetting with nvidia 375.20 I get this error message in dmesg
[drm:nvidia_drm_gem_import_nvkms_memory [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to import NVKMS memory to GEM object

I have Fedora 25 and I am using Xorg.

nvidia-bug-report.log.gz (95.2 KB)

dmesg shows
[ 59.221776] [drm:nvidia_drm_gem_prime_fence_init [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to lookup gem object for fencing init: 0x00000004
[ 59.222613] [drm:nvidia_drm_gem_prime_fence_init [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to lookup gem object for fencing init: 0x00000004
[ 59.222971] [drm:nvidia_drm_gem_prime_fence_init [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to lookup gem object for fencing init: 0x00000004

I believe this is related to GBM and EGL handling on Wayland. Recent gnome uses only Wayland by default. I am on Debian (Sid), so everything is up to date.

nvidia-bug-report.log.gz (106 KB)

Do you run latest kernel (staging) any chance?
There was recently big refactoring around fenceing infrastructure that might cause module ABIs break,
notably fence_ functions becoming dma_fence_ functions along with some other minor tweaks.
Not sure why it wouldn’t whine about missing symbols properly, but I guess it might be the case.

My kernel is 4.8.0, which is stable, the mainline kernel is 4.9.
I attached nvidia-bug-report.log.gz with all information on my environment.

Still present with nvidia 375.26 and kernel kernel-4.8.14-300.fc25.x86_64. Kernel mode setting is enabled.

[ 19.055857] [drm:nvidia_drm_gem_import_nvkms_memory [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to import NVKMS memory to GEM object
[ 19.127648] [drm:nvidia_drm_gem_import_nvkms_memory [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to import NVKMS memory to GEM object
[ 19.129347] [drm:nvidia_drm_gem_import_nvkms_memory [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to import NVKMS memory to GEM object

same here, but in this post

https://devtalk.nvidia.com/default/topic/952109/linux/kms-only-in-one-card-in-dual-card-config/post/5064084/#5064084

Same here:
[12872.942387] [drm:nvidia_drm_gem_prime_fence_force_signal [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000300] Failed to lookup gem object for fence attach: 0x00000004
[12873.073506] [drm:nvidia_drm_gem_prime_fence_force_signal [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000300] Failed to lookup gem object for fence attach: 0x00000004
[12873.074589] [drm:nvidia_drm_gem_prime_fence_force_signal [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000300] Failed to lookup gem object for fence attach: 0x00000004

Linux ERD228-01 4.13.0-36-generic #40~16.04.1-Ubuntu SMP Fri Feb 16 23:25:58 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux

Seen on fedora 31 w/ nvidia 440.44 kernel 5.3.16-300.fc31.x86_64

[ 2433.533786] [drm:nv_drm_gem_fence_attach_ioctl [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to lookup gem object for fence attach: 0x00000004
[ 2433.559944] [drm:nv_drm_gem_fence_attach_ioctl [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to lookup gem object for fence attach: 0x00000004
[ 2433.560694] [drm:nv_drm_gem_fence_attach_ioctl [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to lookup gem object for fence attach: 0x00000004
[ 2433.561094] [drm:nv_drm_gem_fence_attach_ioctl [nvidia_drm]] ERROR [nvidia-drm] [GPU ID 0x00000100] Failed to lookup gem object for fence attach: 0x00000004