Nvidia-smi failed: Unable to communicate with the NVIDIA driver on AlmaLinux 10.1 (RTX 4000 Blackwell)

General Discussion & Chat
1

Hello,

I am experiencing an issue with my NVIDIA GPU driver and would like your assistance.

1. Environment

  • OS: AlmaLinux 10.1

  • Kernel: 6.12.0-124.49.1.el10_1.x86_64

  • CPU: Ryzen 9 9950X

  • GPU: NVIDIA RTX PRO 4000 Blackwell

2. Problem Description and Error Message
After installing the NVIDIA driver, I attempted to verify the GPU status, but I encountered the following error. The driver does not seem to be communicating with the hardware correctly.

Error Message
nvidia-smi
NVIDIA-SMI has failed because it couldn’t communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

3. Installed Packages

Bash

# rpm -qa | grep nvidia
almalinux-release-nvidia-driver-10-5.el10_1.x86_64
libnvidia-gpucomp-595.58.03-1.el10.x86_64
nvidia-modprobe-595.58.03-1.el10.x86_64
libnvidia-ml-595.58.03-1.el10.x86_64
nvidia-driver-libs-595.58.03-1.el10.x86_64
nvidia-kmod-common-595.58.03-1.el10.noarch
kmod-nvidia-open-dkms-595.58.03-1.el10.noarch
nvidia-driver-595.58.03-1.el10.x86_64
nvidia-open-kmod-595.58.03-1.el10_1.x86_64
libnvidia-cfg-595.58.03-1.el10.x86_64
nvidia-driver-cuda-libs-595.58.03-1.el10.x86_64
nvidia-persistenced-595.58.03-1.el10.x86_64
nvidia-driver-cuda-595.58.03-1.el10.x86_64

4. Question

Is there any way to resolve this issue?
Secure Boot is enabled. (Note: I am looking for a solution that allows me to keep Secure Boot enabled.)

2

Are the kernel modules loaded?

lsmod | grep ^nvidia

If not, then sudo modprobe nvidia

3

Thank you for your reply.

I have completed the installation steps and attempted modprobe. The results are shown below. Please note that I would like to keep Secure Boot enabled.

Bash

# dnf install almalinux-release-nvidia-driver
# dnf install nvidia-open-kmod nvidia-driver-cuda
# modprobe nvidia
modprobe: ERROR: could not insert 'nvidia': Key was rejected by service
4

A kernel module from AlmaLinux ought to be signed by same key as the rest of AlmaLinux kernel. One could check what the module says:

modinfo nvidia | grep sig
5

Thank you for your reply.

As requested, here is the modinfo output. It seems the module was signed by a local DKMS key instead of the AlmaLinux official key, which is likely causing the ‘Key rejected’ error under Secure Boot.

modinfo nvidia | grep sig

sig_id: PKCS#7
signer: DKMS module signing key
sig_key: 08:AF:EC:CC:1E:6C:57:AF:C7:60:30:93:9F:F4:DE:5D:81:0F:12:7C
sig_hashalgo: sha512
signature: 16:34:5C:DD:53:08:48:B5:55:9D:7A:6E:C2:20:39:48:1D:27:B0:90:

Now we are trying to fix it. Thx.

6

One method would be to import the cert that DKMS does use:

mokutil --import /var/lib/dkms/mok.pub
reboot

and complete the import in “BIOS setup”.

The other method would be to figure out why there kernel modules are signed by DKMS – as happens when one installs module package from NVidia’s repo – and not by AlmaLinux’s repo (which should also filter out the NVidia’s package).

7

Thank you for your suggestion.

We tried the first method (importing the MOK certificate), but the situation has not changed.

# mokutil --import /var/lib/dkms/mok.pub
input password:
input password again:

# reboot

# nvidia-smi

NVIDIA-SMI has failed because it couldn’t communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running.

Could you please advise on the second method — figuring out why the kernel modules are signed by DKMS instead of AlmaLinux’s repo?

8

There were steps after reboot and before the machine starts.
NVidia describes them in UEFI Secure Boot - NVIDIA Docs
Look for text:

A reboot must be performed.

Just before loading GRUB, shim displays a blue screen

Did you get that blue dialog and did you complete the import there?

I have heard that some UEFI versions have something else than such blue, but I have got it on Dell machines.

Note: the ‘rpm’ can parse globber, so ‘grep’ is not necessary:
# rpm -qa \*nvidia\*

The ‘dnf’ tells more. Example:

# dnf list \*nvidia\*
Last metadata expiration check: 0:00:24 ago on Tue 21 Apr 2026 10:28:29 AM EEST.
Installed Packages
dnf-plugin-nvidia.noarch          2.3-1.el9         @cuda-rhel9-x86_64
kmod-nvidia-open-dkms.noarch      3:590.48.01-1.el9 @cuda-rhel9-x86_64
libnvidia-cfg.x86_64              3:590.48.01-1.el9 @cuda-rhel9-x86_64
libnvidia-gpucomp.x86_64          3:590.48.01-1.el9 @cuda-rhel9-x86_64
libnvidia-ml.x86_64               3:590.48.01-1.el9 @cuda-rhel9-x86_64
nvidia-driver.x86_64              3:590.48.01-1.el9 @cuda-rhel9-x86_64
nvidia-driver-cuda.x86_64         3:590.48.01-1.el9 @cuda-rhel9-x86_64
nvidia-driver-cuda-libs.x86_64    3:590.48.01-1.el9 @cuda-rhel9-x86_64
nvidia-driver-libs.x86_64         3:590.48.01-1.el9 @cuda-rhel9-x86_64
nvidia-kmod-common.noarch         3:590.48.01-1.el9 @cuda-rhel9-x86_64
nvidia-libXNVCtrl.x86_64          3:590.48.01-1.el9 @cuda-rhel9-x86_64
nvidia-modprobe.x86_64            3:590.48.01-1.el9 @cuda-rhel9-x86_64
nvidia-persistenced.x86_64        3:590.48.01-1.el9 @cuda-rhel9-x86_64
nvidia-settings.x86_64            3:590.48.01-1.el9 @cuda-rhel9-x86_64
nvidia-xconfig.x86_64             3:590.48.01-1.el9 @cuda-rhel9-x86_64
xorg-x11-nvidia.x86_64            3:590.48.01-1.el9 @cuda-rhel9-x86_64
Available Packages
almalinux-release-nvidia-driver.x86_64
                                  9-5.el9           extras            
pcp-pmda-nvidia-gpu.x86_64        6.3.7-5.el9       appstream

I don’t have package from AlmaLinux in this machine – all of the driver from NVidia’s repo (cuda-rhel9-x86_64).

9

Thanks for the detailed explanation and the useful dnf tips. I decided to rebuild the entire environment (OS and drivers). This time, the driver was recognized correctly without disabling Secure Boot. It seems like the fresh setup handled the signing/MOK process better. I appreciate your support!

1 Like