You are not logged in.
Enabling DRM KMS as instructed in https://wiki.archlinux.org/index.php/NV … de_setting makes my system hang on boot when the 375.26 driver is installed. It yields the following:
[drm:nvidia_drm_gem_import_nvkms_memory [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Failed to import NVKMS memory to GEM object
[drm:nvidia_drm_gem_import_nvkms_memory [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Failed to import NVKMS memory to GEM object
[drm:nvidia_drm_gem_import_nvkms_memory [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x00000100] Failed to import NVKMS memory to GEM object
Switching to a different TTY would make the screen flicker for a minute and then the commandline would be accessible but the graphical.target wouldn't be started. Reverting back to 375.20 would yield the same errors, KMS wouldn't start, but the system would boot and the graphical target would be reached just fine without KMS. Removing the modeset=1 kernel parameter with 375.26 doesn't solve the issue either. Any help concerning the subject would be appreciated. So far, all I've found online was a comment by a Fedora 25 user with the same exact issue: https://devtalk.nvidia.com/default/topi … st/5038170
Thanks!
Offline
Thanks to a faulty pacman hook, I was able to get it to boot by running /usr/bin/mkinitcpio -p linux after installing the driver. This, however, still prevents KMS from loading and thus X loads without KMS as if it's disabled. The 'Failed to import NVKMS memory to GEM object' messages are still visible in the logs. I'm starting to believe that there's a missing config somewhere. Any tips would be appreciated.
Offline
Had the same issue on both, Fedora 25 and Arch
This solves it:
https://devtalk.nvidia.com/default/topi … 5/#5039885
Offline
Thanks for the link but I don't think my case is the same as the one discussed. I got the driver working with Xorg a few days ago. My issue is with KMS still being disabled at boot even though I reviewed my configs over and over again. I'm trying to get weston running natively which is why I need KMS to start.
Offline
Sorry to bump an old thread but I'm having this same issue, even on the latest 378.13 and 381.22 drivers. I've done some research on "Failed to import NVKMS memory to GEM object" and it appears to be related to CONFIG_HARDENED_USERCOPY added in kernel 4.9, but Nvidia fixed that in the 375 branch a while back...
Were you able to resolve this without building a custom kernel with the above flag toggled off? (or downgrading)
Offline