The vm-virtio crate provides a virtio device implementation for the following:
- A virtio virtqueue and Descriptor chain API,
- A virtio device trait (
VirtioDevice), - Virtio block device abstractions.
We offer support only for virtio v1.0+ (the legacy interface is not supported).
Virtio device implementations will implement the VirtioDevice trait.
Typically the VMM will implement virtio devices and provide the paravirtualized
IO emulation for all supported virtio devices. For example a VMM willing to emulate
a virtio networking device will implement the VirtioDevice trait.
The VirtioDevice implementation will be discovered by the guest through a virtio
transport layer (MMIO, PCI or channel IO). The transport layer implementation will
then call into the VirtioDevice activate method to notify the device that the
guest driver is done configuring it and that it should be ready to handle IO from
the guest.
On the other hand, the VirtioDevice reset method will be called by the transport
layer when the guest driver needs to reset and release all the emulated device resources.
The vm-virtio queue API is to be consumed by virtio device implementations.
The virtio device (net, block, rng, etc) implementation will call into the
queue API to allocate and manage its specific virtio queues:
-
new()to allocate its queues. -
iter()to iterate over the queue descriptors and consume virtio queue buffers made available by the guest driver. -
add_used()to place a virtio queue buffer into the queue and have the guest driver consume it.
Our Continuous Integration (CI) pipeline is implemented on top of Buildkite. For the complete list of tests, check our CI pipeline.
Each individual test runs in a container. To reproduce a test locally, you can use the dev-container on both x86 and arm64.
container_version=5
docker run -it \
--security-opt seccomp=unconfined \
--volume $(pwd):/vm-virtio \
rustvmm/dev:v${container_version}
cd vm-virtio/
cargo test