Skip to content

Can the kata container support the 4090? #1932

@ys928

Description

@ys928

I have successfully deployed and started the Kata container to run a pod, but I found that my 4090 still cannot be used and the log error "Data Center GPUs" is reported:

root@cuda-vectoradd-kata:/# cat /proc/driver/nvidia/gpus/0000\:02\:00.0/information
Model:           NVIDIA GeForce RTX 4090
IRQ:             40
GPU UUID:        GPU-e0549925-97e9-1e24-da74-aad41f9aacda
Video BIOS:      ??.??.??.??.??
Bus Type:        PCIe
DMA Size:        47 bits
DMA Mask:        0x7fffffffffff
Bus Location:    0000:02:00.0
Device Minor:    0
GPU Firmware:    N/A
GPU Excluded:    No
root@cuda-vectoradd-kata:/#
[    0.000000] Linux version 5.19.2-nvidia-gpu (root@1ffd18d2e2dc) (gcc (Ubuntu 11.3.0-1ubuntu1~22.04.1) 11.3.0, GNU ld (GNU Binutils for Ubuntu) 2.38) #1 SMP Wed Aug 2 17:50:59 UTC 2023
[    7.819857] nvidia: loading out-of-tree module taints kernel.
[    7.827841] nvidia: module verification failed: signature and/or required key missing - tainting kernel
[    7.832323] nvidia-nvlink: Nvlink Core is being initialized, major device number 245
[    7.832912] nvidia 0000:02:00.0: enabling device (0140 -> 0143)
[    7.844880] nvidia 0000:02:00.0: vgaarb: changed VGA decodes: olddecodes=io+mem,decodes=none:owns=none
[    7.982876] NVRM: loading NVIDIA UNIX Open Kernel Module for x86_64  535.86.10  Release Build  (dvs-builder@U16-I2-C05-14-2)  Wed Jul 26 23:15:31 UTC 2023
[    7.997821] nvidia-uvm: Loaded the UVM driver, major device number 243.
[    8.005662] nvidia-modeset: Loading NVIDIA UNIX Open Kernel Mode Setting Driver for x86_64  535.86.10  Release Build  (dvs-builder@U16-I2-C05-14-2)  Wed Jul 26 23:05:16 UTC 2023
[    9.607284] NVRM: Open nvidia.ko is only ready for use on Data Center GPUs.
[    9.607289] NVRM: To force use of Open nvidia.ko on other GPUs, see the
root@cuda-vectoradd-kata:/# /cuda-samples/sample
[Vector addition of 50000 elements]
Failed to allocate device vector A (error code CUDA driver version is insufficient for CUDA runtime version)!
root@cuda-vectoradd-kata:/#

Is there any way to make it support 4090?

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions