nvkvm v0.2.3
Both artifacts below were built from this tag by
.github/workflows/release.yml
on a GitHub-hosted runner and carry a build-provenance attestation.
Verify before you run either — an attestation nobody checks is decoration.
Container (the quick way in)
gh attestation verify oci://ghcr.io/reindertpelsma/nvkvm-pv:v0.2.3 --repo reindertpelsma/nvkvm-pv
docker run --rm -it --device /dev/kvm --gpus all \
-e NVIDIA_DRIVER_CAPABILITIES=compute,utility,graphics,display,video \
-p 127.0.0.1:2222:2222 -v nvkvm-guest:/opt/nvkvm-guest \
ghcr.io/reindertpelsma/nvkvm-pv:v0.2.3Needs /dev/kvm and, for the GPU, the NVIDIA container runtime. Note
that in a container the isolate falls back to UID separation, which is
weaker than the namespaced sandbox — see SECURITY.md.
Prebuilt tarball (bare host)
gh attestation verify nvkvm-v0.2.3-linux-x86_64.tar.gz --repo reindertpelsma/nvkvm-pv
tar xzf nvkvm-v0.2.3-linux-x86_64.tar.gz && cd nvkvm-v0.2.3
sudo cp -a qemu-nvkvm /opt/qemu-nvkvm
sudo install -Dm755 src/stub/nvkvm_stub /usr/lib/nvkvm/nvkvm_stubx86-64 only, headless QEMU, glibc floor and the full contents are in
RELEASE.md inside the tarball. The guest kernel module is not
prebuilt — it is compiled against your guest's kernel from
src/guest/, which is why the sources ship with it.
Full Changelog: v0.2.2...v0.2.3