Etienne PerotandgVisor bot
578d8bf360
Prior to this change, setting `runsc` with `--nvproxy` and `--nvproxy-docker` caused the NVIDIA control files to be available in all sandboxes' filesystems, rather than only those which request GPU access. With this change, these devices are only mounted when requested for that sandbox, and the `nvproxy` driver isn't registered. This also has the advantage of not requiring the existence of these device files on the host when starting a non-GPU sandbox, which should help for possibly enabling this feature on by default and have `runsc` still work on machines without GPUs. The downside is that the spec has to be read twice, due to the boot ordering of `runsc`. `specutils.ReadSpecFromFile` already seeks to offset 0 when reading a spec file, so this should be OK. ```shell # On a machine without nvidia drivers loaded and no device files: $ docker run --rm --runtime=runsc debian sh -c 'echo /dev/nvidia*' /dev/nvidia* # Same machine, now requesting GPUs: $ docker run --rm --runtime=runsc --gpus=all debian sh -c 'echo /dev/nvidia*' /dev/nvidia-uvm /dev/nvidia0 /dev/nvidiactl # (As a side-effect of the previous container starting, kernel modules are # now loaded and /dev/nvidia0 exists on the host) # Same machine, but container no longer requesting GPUs: $ docker run --rm --runtime=runsc debian sh -c 'echo /dev/nvidia*' /dev/nvidia* # More tests with `--gpus` flag: $ docker run --rm --runtime=runsc --gpus=device=all debian sh -c 'echo /dev/nvidia*' /dev/nvidia-uvm /dev/nvidia0 /dev/nvidiactl $ docker run --rm --runtime=runsc --gpus=device=none debian sh -c 'echo /dev/nvidia*' /dev/nvidia-uvm /dev/nvidiactl $ docker run --rm --runtime=runsc --gpus=device=void debian sh -c 'echo /dev/nvidia*' /dev/nvidia* $ docker run --rm --runtime=runsc --gpus=device=0 debian sh -c 'echo /dev/nvidia*' /dev/nvidia-uvm /dev/nvidia0 /dev/nvidiactl $ docker run --rm --runtime=runsc --gpus=device=1 debian sh -c 'echo /dev/nvidia*' docker: Error response from daemon: failed to create task for container: failed to create shim task: OCI runtime create failed: creating container: nvidia-container-cli configure failed, err: exit status 1 stdout: stderr: nvidia-container-cli: device error: 1: unknown device: unknown. $ docker run --rm --runtime=runsc debian sh -c 'echo /dev/nvidia*' /dev/nvidia* ``` PiperOrigin-RevId: 542098578
What is gVisor?
gVisor is an application kernel, written in Go, that implements a
substantial portion of the Linux system surface. It includes an
Open Container Initiative (OCI) runtime called runsc that provides an
isolation boundary between the application and the host kernel. The runsc
runtime integrates with Docker and Kubernetes, making it simple to run sandboxed
containers.
Why does gVisor exist?
Containers are not a sandbox. While containers have revolutionized how we develop, package, and deploy applications, using them to run untrusted or potentially malicious code without additional isolation is not a good idea. While using a single, shared kernel allows for efficiency and performance gains, it also means that container escape is possible with a single vulnerability.
gVisor is an application kernel for containers. It limits the host kernel surface accessible to the application while still giving the application access to all the features it expects. Unlike most kernels, gVisor does not assume or require a fixed set of physical resources; instead, it leverages existing host kernel functionality and runs as a normal process. In other words, gVisor implements Linux by way of Linux.
gVisor should not be confused with technologies and tools to harden containers against external threats, provide additional integrity checks, or limit the scope of access for a service. One should always be careful about what data is made available to a container.
Documentation
User documentation and technical architecture, including quick start guides, can be found at gvisor.dev.
Installing from source
gVisor builds on x86_64 and ARM64. Other architectures may become available in the future.
For the purposes of these instructions, bazel and other build
dependencies are wrapped in a build container. It is possible to use
bazel directly, or type make help for standard targets.
Requirements
Make sure the following dependencies are installed:
- Linux 4.14.77+ (older linux)
- Docker version 17.09.0 or greater
Building
Build and install the runsc binary:
mkdir -p bin
make copy TARGETS=runsc DESTINATION=bin/
sudo cp ./bin/runsc /usr/local/bin
Testing
To run standard test suites, you can use:
make unit-tests
make tests
To run specific tests, you can specify the target:
make test TARGETS="//runsc:version_test"
Using go get
This project uses bazel to build and manage dependencies. A synthetic
go branch is maintained that is compatible with standard go tooling for
convenience.
For example, to build and install runsc directly from this branch:
echo "module runsc" > go.mod
GO111MODULE=on go get gvisor.dev/gvisor/runsc@go
CGO_ENABLED=0 GO111MODULE=on sudo -E go build -o /usr/local/bin/runsc gvisor.dev/gvisor/runsc
Subsequently, you can build and install the shim binary for containerd:
GO111MODULE=on sudo -E go build -o /usr/local/bin/containerd-shim-runsc-v1 gvisor.dev/gvisor/shim
Note that this branch is supported in a best effort capacity, and direct
development on this branch is not supported. Development should occur on the
master branch, which is then reflected into the go branch.
Community & Governance
See GOVERNANCE.md for project governance information.
The gvisor-users mailing list and gvisor-dev mailing list are good starting points for questions and discussion.
Security Policy
See SECURITY.md.
Contributing
See Contributing.md.