View Issue Details

Jump to NotesJump to History
IDProjectCategoryView StatusDate SubmittedLast Update
0001585channel: elrepo/el8--elrepo--request-for-enhancement--public2026-02-13 08:262026-04-13 18:17
ReporterSunTekhnik Assigned Totqhoang  
PrioritynormalSeverityminorReproducibilityhave not tried
Status closedResolutionfixed 
OSOracle Linux ServerOS Version8.6 
Summary0001585: kmod-nvidia-367xx for RHEL8
Description[ 1380.843026] NVRM: No NVIDIA graphics adapter found!
[ 1380.844186] nvidia-nvlink: Unregistered the Nvlink Core, major device number 242
[ 1381.005341] nvidia-nvlink: Nvlink Core is being initialized, major device number 242
[ 1381.006020] NVRM: The NVIDIA GRID K1 GPU installed in this system is
               NVRM: supported through the NVIDIA 367.xx Legacy drivers. Please
               NVRM: visit http://www.nvidia.com/object/unix.html for more
               NVRM: information. The 390.157 NVIDIA driver will ignore
Steps To Reproducesudo dnf install -y nvidia-x11-drv-390xx kmod-nvidia-390xx
Additional InformationHello.
Can I expect kmod-nvidia-367xx to be available for RHEL8?
TagsNo tags attached.

Activities

tqhoang

2026-02-13 13:35

administrator   ~0010804

We do not have plans to officially support the legacy 367.xx branch.

In an un-official capacity, I can look into it's feasibility though. Basically if it requires AGP support like the 340.xx branch, then it's SOL.

tqhoang

2026-02-14 16:46

administrator   ~0010805

@SunTekhnik please let me know if you have the capability to test on a non-production system. I have something built, but no way to test since I don't have a GRID GPU. It's highly experimental since I had to essentially backport the 390xx nvidia-uvm changes from v390.138 to v390.157.

SunTekhnik

2026-02-17 02:02

reporter   ~0010806

@tqhoang, hello. I have a system, and it's not in production yet. But the graphics card won't work on the 390 version.

tqhoang

2026-02-17 22:48

administrator   ~0010807

I have some experimental packages available here: https://elrepo.org/people/tqhoang/bug-1585/

kmod-nvidia-367xx-367.134-0.el8_10.elrepo.x86_64.rpm (for RHEL 8.10 GA kernel-4.18.0-553.el8_10.x86_64)
kmod-nvidia-367xx-367.134-0.1.el8_10.elrepo.x86_64.rpm (for RHEL 8.10 errata kernel-4.18.0-553.75.1.el8_10.x86_64 and later)
nvidia-x11-drv-367xx-367.134-0.el8_10.elrepo.x86_64.rpm
nvidia-x11-drv-367xx-libs-367.134-0.el8_10.elrepo.x86_64.rpm
nvidia-x11-drv-367xx-libs-367.134-0.el8_10.elrepo.i686.rpm (for 32-bit libs)

Although I just realized you're running Oracle Linux Server 8.6. Our build systems are only setup for RHEL 8.10 compilation. Any chance you can upgrade 8.10?

SunTekhnik

2026-02-18 02:16

reporter   ~0010808

Yes, we'll try updating to version 8.10. I'll get back to you as soon as it's done.
You mentioned that version 390 needs to be tested further. We're happy to help. If I understand correctly.

SunTekhnik

2026-02-18 04:23

reporter   ~0010809

It's alive!)

nvidia-smi
Wed Feb 18 12:15:19 2026
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 367.134 Driver Version: 367.134 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 GRID K1 On | 0000:13:00.0 Off | N/A |
| N/A 22C P8 7W / 31W | 0MiB / 4036MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
| 1 GRID K1 On | 0000:1B:00.0 Off | N/A |
| N/A 22C P8 8W / 31W | 0MiB / 4036MiB | 0% Default |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| No running processes found |
+-----------------------------------------------------------------------------+

tqhoang

2026-02-18 08:53

administrator   ~0010810

Last edited: 2026-02-18 09:00

 Nice!

That's a promising start. The part that is very experimental is the nvidia-uvm kernel module.

1. Can you see if it is loaded? Ex: lsmod | grep nvidia

2. If not, can you load it ok? Ex: modprobe nvidia-uvm

3. If nvidia-uvm is loaded ok, do you have any CUDA applications to test it?

---

BTW I forgot to reply about the 390xx drivers. I *do not* need them tested any further. I was only saying the 390xx code changes (from v390.138 to v390.157) were my reference for fixing the nvidia-uvm module in 367xx.

SunTekhnik

2026-02-20 02:08

reporter   ~0010811

1.
lsmod | grep nvidia
nvidia 11784192 9
nvidia_drm 16384 0

2.
modprobe nvidia-uvm
lsmod | grep nvidia
nvidia_uvm 790528 0
nvidia 11784192 10 nvidia_uvm
nvidia_drm 16384 0

3.
So far, I've managed to install CUDA 7.5 and get deviceQuery working.
I get an error when trying to compile simple tests. The old CUDA 7.5 doesn't play nice with the new GCC.

./deviceQuery
./deviceQuery Starting...
 CUDA Device Query (Runtime API) version (CUDART static linking)
Detected 2 CUDA Capable device(s)
Device 0: "GRID K1"
  CUDA Driver Version / Runtime Version 8.0 / 7.5
  CUDA Capability Major/Minor version number: 3.0
  Total amount of global memory: 4036 MBytes (4232577024 bytes)
  ( 1) Multiprocessors, (192) CUDA Cores/MP: 192 CUDA Cores
  GPU Max Clock rate: 850 MHz (0.85 GHz)
  Memory Clock rate: 891 Mhz
  Memory Bus Width: 128-bit
  L2 Cache Size: 262144 bytes
  Maximum Texture Dimension Size (x,y,z) 1D=(65536), 2D=(65536, 65536), 3D=(4096, 4096, 4096)
  Maximum Layered 1D Texture Size, (num) layers 1D=(16384), 2048 layers
  Maximum Layered 2D Texture Size, (num) layers 2D=(16384, 16384), 2048 layers
  Total amount of constant memory: 65536 bytes
  Total amount of shared memory per block: 49152 bytes
  Total number of registers available per block: 65536
  Warp size: 32
  Maximum number of threads per multiprocessor: 2048
  Maximum number of threads per block: 1024
  Max dimension size of a thread block (x,y,z): (1024, 1024, 64)
  Max dimension size of a grid size (x,y,z): (2147483647, 65535, 65535)
  Maximum memory pitch: 2147483647 bytes
  Texture alignment: 512 bytes
  Concurrent copy and kernel execution: Yes with 1 copy engine(s)
  Run time limit on kernels: No
  Integrated GPU sharing Host Memory: No
  Support host page-locked memory mapping: Yes
  Alignment requirement for Surfaces: Yes
  Device has ECC support: Disabled
  Device supports Unified Addressing (UVA): Yes
  Device PCI Domain ID / Bus ID / location ID: 0 / 19 / 0
  Compute Mode:
     < Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >
Device 1: "GRID K1"
  CUDA Driver Version / Runtime Version 8.0 / 7.5
  CUDA Capability Major/Minor version number: 3.0
  Total amount of global memory: 4036 MBytes (4232577024 bytes)
  ( 1) Multiprocessors, (192) CUDA Cores/MP: 192 CUDA Cores
  GPU Max Clock rate: 850 MHz (0.85 GHz)
  Memory Clock rate: 891 Mhz
  Memory Bus Width: 128-bit
  L2 Cache Size: 262144 bytes
  Maximum Texture Dimension Size (x,y,z) 1D=(65536), 2D=(65536, 65536), 3D=(4096, 4096, 4096)
  Maximum Layered 1D Texture Size, (num) layers 1D=(16384), 2048 layers
  Maximum Layered 2D Texture Size, (num) layers 2D=(16384, 16384), 2048 layers
  Total amount of constant memory: 65536 bytes
  Total amount of shared memory per block: 49152 bytes
  Total number of registers available per block: 65536
  Warp size: 32
  Maximum number of threads per multiprocessor: 2048
  Maximum number of threads per block: 1024
  Max dimension size of a thread block (x,y,z): (1024, 1024, 64)
  Max dimension size of a grid size (x,y,z): (2147483647, 65535, 65535)
  Maximum memory pitch: 2147483647 bytes
  Texture alignment: 512 bytes
  Concurrent copy and kernel execution: Yes with 1 copy engine(s)
  Run time limit on kernels: No
  Integrated GPU sharing Host Memory: No
  Support host page-locked memory mapping: Yes
  Alignment requirement for Surfaces: Yes
  Device has ECC support: Disabled
  Device supports Unified Addressing (UVA): Yes
  Device PCI Domain ID / Bus ID / location ID: 0 / 27 / 0
  Compute Mode:
     < Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >
> Peer access from GRID K1 (GPU0) -> GRID K1 (GPU1) : No
> Peer access from GRID K1 (GPU1) -> GRID K1 (GPU0) : No
deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 8.0, CUDA Runtime Version = 7.5, NumDevs = 2, Device0 = GRID K1, Device1 = GRID K1

tqhoang

2026-02-23 00:41

administrator   ~0010812

Last edited: 2026-02-25 10:26

 Have you tried CUDA 8.0.44? That is the max version for the NVIDIA R367 driver.

Please keep me posted on how things are going. If it's 100% working, I'll add this driver to our EL8 repo.

SunTekhnik

2026-03-03 02:28

reporter   ~0010819

@tqhoang Thanks for your help. The person handling this issue promised to test it this week.

SunTekhnik

2026-03-11 10:41

reporter   ~0010825

I couldn't compile llama.cpp and alpaka.cpp for CUDA8. I tried to compile for OpenCL, but the server was taken down. The driver seems to be working)).

tqhoang

2026-04-13 18:17

administrator   ~0010853

FYI I officially added the driver to our EL8 repo. Going to close this ticket as fixed.

kmod-nvidia-367xx-367.134-1.el8_10.elrepo.x86_64.rpm (for RHEL 8.10 GA kernel-4.18.0-553.el8_10.x86_64)
kmod-nvidia-367xx-367.134-1.1.el8_10.elrepo.x86_64.rpm (for RHEL 8.10 errata kernel-4.18.0-553.75.1.el8_10.x86_64 and later)
nvidia-x11-drv-367xx-367.134-1.el8_10.elrepo.x86_64.rpm
nvidia-x11-drv-367xx-libs-367.134-1.el8_10.elrepo.x86_64.rpm
nvidia-x11-drv-367xx-libs-367.134-1.el8_10.elrepo.i686.rpm (for 32-bit libs)

Note: These drivers do not support the X11 version in EL 8.10. They should be adequate for compute capability provided it's using the supported CUDA version.

Issue History

Date Modified Username Field Change
2026-02-13 08:26 SunTekhnik New Issue
2026-02-13 08:26 SunTekhnik Status new => assigned
2026-02-13 08:26 SunTekhnik Assigned To => tqhoang
2026-02-13 13:35 tqhoang Note Added: 0010804
2026-02-14 16:46 tqhoang Note Added: 0010805
2026-02-15 14:32 tqhoang Category kmod-nvidia => --elrepo--request-for-enhancement--
2026-02-17 02:02 SunTekhnik Note Added: 0010806
2026-02-17 22:48 tqhoang Note Added: 0010807
2026-02-17 22:48 tqhoang Status assigned => feedback
2026-02-18 02:16 SunTekhnik Note Added: 0010808
2026-02-18 02:16 SunTekhnik Status feedback => assigned
2026-02-18 04:23 SunTekhnik Note Added: 0010809
2026-02-18 08:53 tqhoang Note Added: 0010810
2026-02-18 09:00 tqhoang Note Edited: 0010810
2026-02-18 09:01 tqhoang Status assigned => feedback
2026-02-20 02:08 SunTekhnik Note Added: 0010811
2026-02-20 02:08 SunTekhnik Status feedback => assigned
2026-02-23 00:41 tqhoang Note Added: 0010812
2026-02-23 09:39 tqhoang Note Edited: 0010812
2026-02-25 10:26 tqhoang Note Edited: 0010812
2026-03-03 02:28 SunTekhnik Note Added: 0010819
2026-03-11 10:41 SunTekhnik Note Added: 0010825
2026-04-13 18:17 tqhoang Status assigned => closed
2026-04-13 18:17 tqhoang Resolution open => fixed
2026-04-13 18:17 tqhoang Note Added: 0010853