Debian 13, GTX 1080, Failed to allocate NvKmsKapiDevice with 580 drivers

Graphics / Linux Linux
,
1

I’m trying to get my old GTX 1080 up and running on Debian 13, fresh install, with the proprietary 580 drivers (last version that supports it) for use with headless CUDA for local Ollama execution.

I did manage to get it working, but then rebooted, and now even after purging drivers, reinstalling drivers, etc, I keep getting this:

[    4.863289] nvidia: loading out-of-tree module taints kernel.
[    4.863301] nvidia: module license 'NVIDIA' taints kernel.
[    4.863302] Disabling lock debugging due to kernel taint
[    4.863306] nvidia: module verification failed: signature and/or required key missing - tainting kernel
[    4.863308] nvidia: module license taints kernel.
[    5.495070] nvidia-nvlink: Nvlink Core is being initialized, major device number 243

[    5.504807] nvidia 0000:00:0c.0: vgaarb: VGA decodes changed: olddecodes=io+mem,decodes=none:owns=io+mem
[    5.621660] NVRM: loading NVIDIA UNIX x86_64 Kernel Module  580.105.08  Wed Oct 29 23:15:11 UTC 2025
[    5.684392] nvidia_uvm: module uses symbols nvUvmInterfaceDisableAccessCntr from proprietary module nvidia, inheriting taint.
[    5.703219] nvidia-modeset: Loading NVIDIA Kernel Mode Setting Driver for UNIX platforms  580.105.08  Wed Oct 29 22:15:26 UTC 2025
[    5.886821] NVRM: GPU 0000:00:0c.0: RmInitAdapter failed! (0x23:0xffff:1574)
[    5.886837] NVRM: GPU 0000:00:0c.0: rm_init_adapter failed, device minor number 0
[    5.897322] NVRM: GPU 0000:00:0c.0: RmInitAdapter failed! (0x23:0xffff:1574)
[    5.897330] NVRM: GPU 0000:00:0c.0: rm_init_adapter failed, device minor number 0
[    6.371431] NVRM: GPU 0000:00:0c.0: RmInitAdapter failed! (0x23:0xffff:1574)
[    6.371437] NVRM: GPU 0000:00:0c.0: rm_init_adapter failed, device minor number 0
[    6.371509] [drm] [nvidia-drm] [GPU ID 0x0000000c] Loading driver
[    6.378278] NVRM: GPU 0000:00:0c.0: RmInitAdapter failed! (0x23:0xffff:1574)
[    6.378283] NVRM: GPU 0000:00:0c.0: rm_init_adapter failed, device minor number 0
[    6.378865] [drm:nv_drm_dev_load [nvidia_drm]] *ERROR* [nvidia-drm] [GPU ID 0x0000000c] Failed to allocate NvKmsKapiDevice
[    6.385350] NVRM: GPU 0000:00:0c.0: RmInitAdapter failed! (0x23:0xffff:1574)
[    6.385355] NVRM: GPU 0000:00:0c.0: rm_init_adapter failed, device minor number 0
[    6.551984] NVRM: GPU 0000:00:0c.0: RmInitAdapter failed! (0x23:0xffff:1574)
[    6.551991] NVRM: GPU 0000:00:0c.0: rm_init_adapter failed, device minor number 0
[    6.559040] NVRM: GPU 0000:00:0c.0: RmInitAdapter failed! (0x23:0xffff:1574)
[    6.559045] NVRM: GPU 0000:00:0c.0: rm_init_adapter failed, device minor number 0

I’m following Debian — NVIDIA Driver Installation Guide instructions, installed these 2 packages nvidia-kernel-dkms nvidia-driver-cuda which then, of course, installed all the dependencies.

Hardware/kernel info

# uname -a
Linux darkllama 6.12.63+deb13-amd64 #1 SMP PREEMPT_DYNAMIC Debian 6.12.63-1 (2025-12-30) x86_64 GNU/Linux
# lspci | grep -i nvi
00:0c.0 VGA compatible controller: NVIDIA Corporation GP104 [GeForce GTX 1080] (rev a1)
00:0d.0 Audio device: NVIDIA Corporation GP104 High Definition Audio Controller (rev a1)
2

Unfortunately these codes are proprietary and NV engs never share what they mean, so we are in a dark here…

The fact that you have managed to get it working once however, makes me think, that this may be related to kernel module params: for example it may be possible that the first time the modules were loaded, modprobe config was not yet copied to /etc/modprobe.d/ or something like that. You can for example try to set modeset to 0 in /etc/modprobe.d/nvidia-modeset.conf and see if it helps (my best bet). There are also some settings in /etc/modprobe.d/nvidia.conf: if changing modeset won’t help you can try manipulating those also.

On a side note: any reason not to use a newer kernel from trixie-backports? (currently 6.17.13)

3

Well… the modeset = 0 got rid of that error, which probably makes sense since I’m not doing framebuffer/console/graphics at all.

However, nvidia-smi doesn’t see my device still.

Still getting multiple of these, before and after the nvidia-drm says it’s loading.

[    5.926110] NVRM: GPU 0000:00:0c.0: RmInitAdapter failed! (0x23:0xffff:1574)
[    5.926126] NVRM: GPU 0000:00:0c.0: rm_init_adapter failed, device minor number 0

Tried adding options nvidia NVreg_OpenRmEnableUnsupportedGpus=1 suggested in Nvidia-smi "No devices were found" - VMWare ESXI Ubuntu Server 20.04.03 with RTX3070 - #40 by LionelDerBoven , but no change.

I now tried rebooting the whole thing, and it seems to now be working. I wonder if there was some kind of something stuck, and needed a full reset of the GPU shrug I really hope that doesn’t happen regularly.

4

Sorry for being a bearer of bad news, but a GPU sometimes working sometimes not under the same conditions is usually a sign of a failing hardware… :/

Anyway, could you please post the exact set of options under which it currently works for reference for future users who may encounter similar problems?

Just to make sure: you are still using proprietary driver flavor, correct? (asking because AFAIR in the linked thread ppl were using NVreg_OpenRmEnableUnsupportedGpus for open ones)

5

I don’t have NVreg_OpenRmEnableUnsupportedGpus set currently, that was just a test.

So this is actually a PCIe passthrough FreeBSD host, Debian 13 VM. Looks like there was some kind of issue where during some rebooting/shutdown/whatever of the VM, the card got into a state it needed a full PCIe reset, so after rebooting the host, it’s working as expected.

6

…and previously you didn’t deem it worth mentioning somehow? geezzzz…

7

yeah minor detail :))