virtualization-virtio-vhost-net
vhost-net
Virtio Backend(dataplane) in host kernel leverage vhost protocol.
The vhost-net is a kernel driver that implements the handler side of the vhost protocol to implement an efficient data plane, i.e., packet forwarding. In this implementation, qemu and the vhost-net kernel driver (handler) use ioctls to exchange vhost messages and a couple of eventfd-like file descriptors called irqfd and ioeventfd(created by Qemu process with ioctl with kvm, then pass it with ioctl with vhost-net) are used to exchange notifications with the guest.
When vhost-net kernel driver is loaded, it exposes a character device on /dev/vhost-net. When qemu is launched with vhost-net support it opens it and initializes the vhost-net instance with several ioctl(2) calls. These are necessary to associate the hypervisor(qemu) process with the vhost-net instance, prepare for virtio feature negotiation and pass the guest physical memory mapping to the vhost-net driver by this fd with ioctl.
During the initialization the vhost-net kernel driver creates a kernel thread called vhost-$pid, where $pid is the hypervisor(qemu) process pid. This thread is called the “vhost worker thread”.
A tap device is still used to communicate the VM with the host but now the worker thread handles the I/O events i.e. it polls for driver notifications or tap events, and forwards data.
Qemu allocates one eventfd and registers it to both vhost and KVM in order to achieve the notification bypass. The vhost-$pid kernel thread polls it, and KVM writes to it when the guest writes in a specific address. This mechanism is named ioeventfd. This way, a simple read/write operation to a specific guest memory address does not need to go through the expensive QEMU process wakeup and can be routed to the vhost worker thread directly. This also has the advantage of being asynchronous, no need for the vCPU to stop (so no need to do an immediate context switch).
On the other hand, qemu allocates another eventfd and registers it to both KVM and vhost again for direct vCPU interruption injection. This mechanism is called irqfd, and it allows any process in the host to inject vCPU interrupts to the guest by writing to it, with the same advantages (asynchronous, no need for immediate context switching, etc).
Sending pkt diagram
More detail with OVS(add tap interface to ovs or create ovs port by ovs-tool)
