Problem Description
https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html says that ROCm is supported on Ubuntu 26.04, kernel 7.0.
The Dockerfile used the operator uses for Ubuntu downloads the linux-modules-extra package:
|
linux-modules-extra-${KERNEL_FULL_VERSION} |
However, the linux-modules-extra package is not shipped for kernel 7.0 and newer: https://discourse.ubuntu.com/t/kernel-development-release-cadence-and-deprecation-of-linux-modules-extra/65176
Therefore, the Dockerfile build fails:
[+] Building 4.9s (6/14) docker:default
=> [internal] load build definition from Dockerfile 0.0s
=> => transferring dockerfile: 1.35kB 0.0s
=> WARN: FromAsCasing: 'as' and 'FROM' keywords' casing do not match (line 2) 0.0s
=> WARN: InvalidDefaultArgInFrom: Default value for ARG ubuntu:${OS_VERSION} results in empty or invalid base image name (line 2) 0.0s
=> WARN: InvalidDefaultArgInFrom: Default value for ARG ubuntu:${OS_VERSION} results in empty or invalid base image name (line 35) 0.0s
=> [internal] load metadata for docker.io/library/ubuntu:26.04 0.0s
=> [internal] load .dockerignore 0.0s
=> => transferring context: 2B 0.0s
=> [builder 1/7] FROM docker.io/library/ubuntu:26.04 0.0s
=> ERROR [builder 2/7] RUN apt-get update && apt-get install -y bc bison flex libelf-dev gnupg wget git make gcc 4.9s
=> CANCELED [stage-1 2/5] RUN apt-get update && apt-get install -y kmod 4.9s
------
> [builder 2/7] RUN apt-get update && apt-get install -y bc bison flex libelf-dev gnupg wget git make gcc ca-certificates linux-headers-7.0.0-30-generic linux-modules-extra-7.0.0-30-generic:
0.508 Get:1 http://security.ubuntu.com/ubuntu resolute-security InRelease [137 kB]
0.614 Get:2 http://security.ubuntu.com/ubuntu resolute-security/restricted amd64 Packages [431 kB]
0.709 Get:3 http://archive.ubuntu.com/ubuntu resolute InRelease [136 kB]
0.729 Get:4 http://security.ubuntu.com/ubuntu resolute-security/multiverse amd64 Packages [11.2 kB]
0.746 Get:5 http://security.ubuntu.com/ubuntu resolute-security/main amd64 Packages [511 kB]
0.817 Get:6 http://security.ubuntu.com/ubuntu resolute-security/universe amd64 Packages [196 kB]
1.026 Get:7 http://archive.ubuntu.com/ubuntu resolute-updates InRelease [137 kB]
1.229 Get:8 http://archive.ubuntu.com/ubuntu resolute-backports InRelease [136 kB]
1.309 Get:9 http://archive.ubuntu.com/ubuntu resolute/restricted amd64 Packages [189 kB]
1.336 Get:10 http://archive.ubuntu.com/ubuntu resolute/main amd64 Packages [1874 kB]
1.513 Get:11 http://archive.ubuntu.com/ubuntu resolute/universe amd64 Packages [20.1 MB]
3.027 Get:12 http://archive.ubuntu.com/ubuntu resolute/multiverse amd64 Packages [352 kB]
3.052 Get:13 http://archive.ubuntu.com/ubuntu resolute-updates/universe amd64 Packages [300 kB]
3.078 Get:14 http://archive.ubuntu.com/ubuntu resolute-updates/multiverse amd64 Packages [14.3 kB]
3.078 Get:15 http://archive.ubuntu.com/ubuntu resolute-updates/main amd64 Packages [599 kB]
3.123 Get:16 http://archive.ubuntu.com/ubuntu resolute-updates/restricted amd64 Packages [443 kB]
3.160 Get:17 http://archive.ubuntu.com/ubuntu resolute-backports/universe amd64 Packages [567 B]
3.539 Fetched 25.6 MB in 3s (7612 kB/s)
3.539 Reading package lists...
4.071 Reading package lists...
4.589 Building dependency tree...
4.738 Reading state information...
4.814 E: Unable to locate package linux-modules-extra-7.0.0-30-generic
4.814 E: Couldn't find any package by glob 'linux-modules-extra-7.0.0-30-generic'
4.814 E: Couldn't find any package by regex 'linux-modules-extra-7.0.0-30-generic'
Operating System
Ubuntu 26.04.0
CPU
N/A
GPU
N/A
ROCm Version
N/A
ROCm Component
No response
Steps to Reproduce
Download the Dockerfile used by the operator, then run:
#!/usr/bin/env bash
export OS_VERSION=26.04
export OS_CODENAME=resolute
export KERNEL_FULL_VERSION=7.0.0-30-generic
export DRIVERS_VERSION=31.40.1
export REPO_URL=https://repo.radeon.com
docker build \
--build-arg OS_VERSION=${OS_VERSION} \
--build-arg OS_CODENAME=${OS_CODENAME} \
--build-arg KERNEL_FULL_VERSION=${KERNEL_FULL_VERSION} \
--build-arg DRIVERS_VERSION=${DRIVERS_VERSION} \
--build-arg REPO_URL=${REPO_URL} \
-t registry.example.com/amdgpu-driver:ubuntu-${OS_VERSION}-${KERNEL_FULL_VERSION}-${DRIVERS_VERSION} \
- <precompiled.dockerfile
(Optional for Linux users) Output of /opt/rocm/bin/rocminfo --support
Additional Information
There is a note for the 1.4.0 operator release that appears to be related.
There are a few places in the operator repository that reference this package, including the example Dockerfile in the documentation.
Problem Description
https://rocm.docs.amd.com/en/latest/compatibility/compatibility-matrix.html says that ROCm is supported on Ubuntu 26.04, kernel 7.0.
The Dockerfile used the operator uses for Ubuntu downloads the
linux-modules-extrapackage:gpu-operator/internal/kmmmodule/dockerfiles/DockerfileTemplate.ubuntu
Line 23 in fa43b1c
However, the
linux-modules-extrapackage is not shipped for kernel 7.0 and newer: https://discourse.ubuntu.com/t/kernel-development-release-cadence-and-deprecation-of-linux-modules-extra/65176Therefore, the Dockerfile build fails:
Operating System
Ubuntu 26.04.0
CPU
N/A
GPU
N/A
ROCm Version
N/A
ROCm Component
No response
Steps to Reproduce
Download the Dockerfile used by the operator, then run:
(Optional for Linux users) Output of /opt/rocm/bin/rocminfo --support
Additional Information
There is a note for the 1.4.0 operator release that appears to be related.
There are a few places in the operator repository that reference this package, including the example Dockerfile in the documentation.