cloud-hypervisor/docs/fs.md

How to use virtio-fs

In the context of virtualization, it is always convenient to be able to share a directory from the host with the guest.

virtio-fs, also known as vhost-user-fs is a virtual device defined by the VIRTIO specification which allows any VMM to perform filesystem sharing.

Pre-requisites

The daemon

This virtual device relies on the vhost-user protocol, which assumes the backend (device emulation) is handled by a dedicated process running on the host. This daemon is called virtiofsd and needs to be present on the host.

Build virtiofsd

git clone --depth 1 "https://github.com/sboeuf/qemu.git" -b "virtio-fs" $VIRTIOFSD_DIR
cd $VIRTIOFSD_DIR
./configure --prefix=$PWD --target-list=x86_64-softmmu
make virtiofsd -j `nproc`
sudo setcap cap_sys_admin+epi "virtiofsd"

Create shared directory

mkdir /tmp/shared_dir

Run virtiofsd

./virtiofsd \
    -d \
    --socket-path=/tmp/virtiofs \
    -o source=/tmp/shared_dir \
    -o cache=none

The cache=none option should be the default when using virtiofsd with the cloud-hypervisor VMM. This prevents from using the guest page cache, which reduces the memory footprint of the guest. When running multiple virtual machines on the same host, this will let the host deal with page cache, which will increase the density of virtual machines which can be launched.

The cache=always option will allow for the guest page cache to be used, which will increase the memory footprint of the guest. This option should be used only for specific use cases where a single VM is going to be running on a host.

The kernel

In order to leverage virtio-fs support from within the guest, and because the code has not been merged in upstream Linux kernel yet, it is required to build a custom kernel embedding the patches.

The following branch virtio-fs-virtio-iommu on the repository https://github.com/cloud-hypervisor/linux.git includes all the needed patches to support virtio-fs.

Make sure to build a kernel out of this branch that can be then used to boot the VM.

How to share directories with cloud-hypervisor

Start the VM

Once the daemon is running, the option --fs from cloud-hypervisor needs to be used.

Direct kernel boot option is preferred since we need to provide the custom kernel including the virtio-fs patches. We could boot from hypervisor-fw if we had previously edited the image to replace the kernel binary.

Because vhost-user expects a dedicated process (virtiofsd in this case) to be able to access the guest RAM to communicate through the virtqueues with the driver running in the guest, --memory option needs to be slightly modified. It needs to specify a backing file for the memory so that an external process can access it.

Assuming you have clear-kvm.img and custom-vmlinux.bin on your system, here is the cloud-hypervisor command you need to run:

./cloud-hypervisor \
    --cpus 4 \
    --memory "size=512,file=/dev/shm" \
    --disk path=clear-kvm.img \
    --kernel custom-vmlinux.bin \
    --cmdline "console=ttyS0 reboot=k panic=1 nomodules root=/dev/vda3" \ 
    --fs tag=myfs,sock=/tmp/virtiofs,num_queues=1,queue_size=512

By default, DAX is enabled with a cache window of 8GiB. You can specify a custom size (let's say 4GiB for this example) for the cache by explicitly setting DAX and the cache size:

--fs tag=virtiofs,sock=/tmp/virtiofs,num_queues=1,queue_size=512,dax=on,cache_size=4G

In case you don't want to use a shared window of cache to pass the shared files content, this means you will have to explicitly disable DAX with dax=off. Note that in this case, the cache_size parameter will be ignored.

--fs tag=virtiofs,sock=/tmp/virtiofs,num_queues=1,queue_size=512,dax=off

Mount the shared directory

The last step is to mount the shared directory inside the guest, using the virtiofs filesystem type.

mkdir mount_dir
mount -t virtiofs -o dax myfs mount_dir/

The tag needs to be consistent with what has been provided through the cloud-hypervisor command line, which happens to be myfs in this example.

The -o dax option must be removed in case the shared cache region is not enabled from the VMM.