Commit Graph

29 Commits

Author SHA1 Message Date
Tobin Feldman-Fitzthum
0444d70704 rootfs: add runtime support for NVAT
Update NVIDIA rootfs builder to include runtime dependencies for NVAT
Rust bindings.

The nvattest package does not include the .so file, so we need to build
from source.

Signed-off-by: Tobin Feldman-Fitzthum <tfeldmanfitz@nvidia.com>
2026-04-06 17:51:12 +00:00
Manuel Huber
a9b222f91e gpu: Update chiseled rootfs with new CDH deps
With CDH requiring libcryptsetup, mkfs.ext4, dd, and their
dependencies, we will need to update the chiseled NVIDIA rootfs
accordingly.

Signed-off-by: Manuel Huber <manuelh@nvidia.com>
2026-03-16 09:43:17 -07:00
Joji Mekkattuparamban
1fd66db271 nvidia-gpu: add missing libraries to rootfs
Added the missing packages to the nvidia rootfs.

Fixes #12534

Signed-off-by: Joji Mekkattuparamban <jojim@nvidia.com>
2026-03-13 16:24:32 -07:00
Fabiano Fidêncio
e9894c0bd8 nvidia: Bump nvidia-container-toolkit to 1.18.1
Let's update the nvidia-container-toolkit to 1.18.1 (from 1.17.6).

We're, from now on, relying on the version set in the versions.yaml
file.

Signed-off-by: Fabiano Fidêncio <ffidencio@nvidia.com>
2026-03-05 11:53:09 +01:00
Zvonko Kaiser
1de7dd58f5 gpu: Add NVLSM daemon
We need to chissel the NVLSM daemon for NVL5 systems

Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
2026-02-20 11:39:59 +01:00
Manuel Huber
d9d1073cf1 gpu: Install packages for devkit
Introduce a new function to install additional packages into the
devkit flavor. With modprobe, we avoid errors on pod startup
related to loading nvidia kernel modules in the NVRC phase.
Note, the production flavor gets modprobe from busybox, see its
configuration file containing CONFIG_MODPROBE=y.

Signed-off-by: Manuel Huber <manuelh@nvidia.com>
2026-02-06 09:58:32 +01:00
Zvonko Kaiser
a59f791bf5 gpu: Move CUDA repo selection to versions.yaml
We want to enable local and remote CUDA repository builds.
Moving the cuda and tools repo to versions.yaml with a
unified build for both types.

Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
2026-01-26 22:19:40 +01:00
Zvonko Kaiser
428cc5d586 gpu: Chroot Cleanup
With the newest NVRC we do not need the supported GPUs
anymore.

Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
2026-01-17 19:27:24 +01:00
Zvonko Kaiser
ffc8725164 gpu: rootfs update decoupling
Remove all the driver build instructions,
sicne those are now done in the kernel target.

Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
2026-01-14 20:45:54 +01:00
Manuel Huber
3966864376 gpu: introduce devkit build flag
Introduce a new devkit parameter which will produce a rootfs
without chisselling. This results in a larger rootfs with various
packages and binaries being included, for instance, enabling the
use of the debug console.

Signed-off-by: Manuel Huber <manuelh@nvidia.com>
2025-11-19 15:50:03 +01:00
Manuel Huber
2c9e0f9f4f gpu: add signed-by to package sources
Pin to specific key. CUDA package sources in
/etc/apt/sources.list.d already use a specific key.

Signed-off-by: Manuel Huber <manuelh@nvidia.com>
2025-11-19 15:50:03 +01:00
Zvonko Kaiser
39848e0983 gpu: rootfs fixes
Build only from Ubuntu repositories do not mix with developer.nvidia.com

Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>

Update tools/osbuilder/rootfs-builder/nvidia/nvidia_chroot.sh

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
2025-10-26 19:36:55 +01:00
Manuel Huber
af34308c83 gpu: remove version suffixes for imex and nscq
This change ensures that the NVIDIA package repository for nvidia-imex
and libnvidia-nspc is being used as source.

The NVIDIA repository does not publish these packages with a -580
version suffix, which made us fall back to the packages from the
Ubuntu repository.

These two packages were recently updated by Ubuntu to depend on
nvidia-kernel-common-580-server (this happened from version
580.82.07-0ubuntu1 to version 580.95.05-0ubuntu1). This conflicts
with nvidia-kernel-common-580 which gets installed by
nvidia-headless-no-dkms-580-open, thus causing a build failure.

Signed-off-by: Manuel Huber <manuelh@nvidia.com>
2025-10-21 15:42:51 +02:00
Zvonko Kaiser
b00013c717 kernel: Add KBUILD_SIGN_PIN pass through
This is needed to the kernel setup picks up the correct
config values from our fragments directories.

Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
2025-10-10 15:45:34 -04:00
Zvonko Kaiser
37bd5e3c9d gpu: Add kernel CONFIG check
We need to make sure that the kernel we're using has the
correct configs set, otherwise the module signing will not work.

Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
2025-10-10 15:45:34 -04:00
Zvonko Kaiser
91739d4425 gpu: PPCIE support DGX like systems
For DGX like systems we need additional binaries and libraries,
enable the Kata AND CoCo use-case.

Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>

Update tools/osbuilder/rootfs-builder/nvidia/nvidia_rootfs.sh

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
2025-10-09 00:00:12 +00:00
Zvonko Kaiser
7061f64db5 gpu: Fix confidential build
NVRC introduced the confidential feature flag and we
haven't updated the rootfs build to accomodate.
If rootfs_type==confidential user --feature=confidential

Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
2025-10-08 10:01:27 +02:00
Zvonko Kaiser
2260f66339 gpu: Some fixes regarding the rootfs v580
With the 580 driver version we need new dependencies
in the rootfs.

Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
2025-10-08 10:01:27 +02:00
Zvonko Kaiser
2693daf503 gpu: Install dcgm export from the CUDA repo
Do not use the repo to install the exporter,
we rely on the version tested with Ubuntu <version>

Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
2025-10-02 18:05:13 +02:00
Zvonko Kaiser
56c6512781 gpu: Bump to noble and rearrange repos
Moving the CUDA repo to the top for all essential packages
and adding a repo priority favouring NVIDIA based repos.

Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
2025-10-02 18:05:13 +02:00
Zvonko Kaiser
e6f12d8f86 gpu: Add latest driver per default
Lets make sure that we use latest driver for CI and release.
There was a sort step missing.

Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
2025-09-20 23:50:35 +00:00
Zvonko Kaiser
da17b06d28 gpu: Pin toolkit version
New versions have incompatibilites, pin toolkit to a working
version

Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
2025-07-14 22:07:21 +00:00
Zvonko Kaiser
445cad7754 gpu: Set the ARCH explicilty for driver builds
Kernel Makefiles changed how to deduce the right arch
lets set it explicilty to enable arm and amd builds.

Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
2025-05-01 17:13:20 +00:00
Zvonko Kaiser
af1d6c2407 shecllcheck: Update nvidia_chroot.sh
Make shellcheck happy with the new rules new updates
needed

Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
2025-02-28 16:27:51 +00:00
Zvonko Kaiser
39d3b7fb90 gpu: Update NVIDIA chroot script
We need to place the signing key and cert at the right place
and hide the KBUILD_SIGN_PIN from echo'ing or xtrace

Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
2025-02-28 01:31:35 +00:00
Zvonko Kaiser
eeacd8fd74 gpu: Adapt rootfs build for multi-arch
Add aarch64 and x86_64 handling. Especially build the Rust
dependency with the correct rust musl target.

Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
2025-02-04 16:44:21 +00:00
Zvonko Kaiser
98e0dc1676 gpu: Add set -u to scripts
Make the scripts more robust by failing on unset varaibles

Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
2025-01-27 17:56:21 +00:00
Zvonko Kaiser
f153229865 gpu: Add driver version selection
Besides latest and lts options add an option to specify
the exact driver version.

Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
2025-01-27 17:56:21 +00:00
Zvonko Kaiser
0debf77770 gpu: NVIDIA gpu initrd/image build
With each release make sure we ship a GPU enabled rootfs/initrd

Fixes: #6554

Signed-off-by: Zvonko Kaiser <zkaiser@nvidia.com>
2024-11-21 18:57:23 +00:00