Virtualization - Aug 2023 - [PATCH vhost v11 05/10] virtio_ring: introduce virtqueue_dma_dev()

On Mon, 31 Jul 2023 19:36:06 -0700, Jakub Kicinski <kuba at kernel.org>
wrote:
> On Tue, 1 Aug 2023 10:03:44 +0800 Xuan Zhuo wrote:
> > > Virtio is either a SW
> > > construct or offloaded to very capable HW, so either way cost of
> > > creating an extra instance for DPDK or whatever else is very low.
> >
> > The extra instance is virtio-net?
> >
> > I think there is a gap. So let me give you a brief introduction of our
case.
> >
> > Firstly, we donot use dpdk. We use the AF_XDP, because of that the
AF_XDP is
> > more simpler and easy to deploy for the nginx.
> >
> > We use the AF_XDP to speedup the UDP of the quic. By the library, the
APP just
> > needs some simple change.
> >
> > On the AliYun, the net driver is virtio-net. So we want the virtio-net
support
> > the AF_XDP.
> >
> > I guess what you mean is that we can speed up through the cooperation
of devices
> > and drivers, but our machines are public clouds, and we cannot change
the
> > back-end devices of virtio under normal circumstances.
> >
> > Here I do not know the different of the real hw and the virtio-net.
>
> You have this working and benchmarked or this is just and idea?
This is not just an idea. I said that has been used on large scale.

This is the library for the APP to use the AF_XDP. We has open it.
https://gitee.com/anolis/libxudp

This is the Alibaba version of the nginx. That has been opened, that supported
to work with the libray to use AF_XDP.
http://tengine.taobao.org/

I supported this on our kernel release Anolis/Alinux.

The work was done about 2 years ago. You know, I pushed the first version to
enable AF_XDP on virtio-net about two years ago. I never thought the job would
be so difficult.

The nic (virtio-net) of AliYun can reach 24,000,000PPS.
So I think there is no different with the real HW on the performance.

With the AF_XDP, the UDP pps is seven times that of the kernel udp stack.
>
> What about io_uring zero copy w/ pre-registered buffers.
> You'll get csum offload, GSO, all the normal perf features.
We tried io-uring, but it was not suitable for our scenario.

Yes, now the AF_XDP does not support the csum offload and GSO.
This is indeed a small problem.

Thanks.