Re: [Qemu-devel] [PATCH 0/9] introduce virtio net dataplane

[Michael S. Tsirkin]

On Thu, Feb 21, 2013 at 10:15:58AM -0600, Anthony Liguori wrote:
> "Michael S. Tsirkin" <address@hidden> writes:
> 
> > On Thu, Feb 21, 2013 at 08:54:44PM +0800, Liu Ping Fan wrote:
> >> This is a emulation to virtio-blk dataplane, which push the data
> >> handling out of biglock. And it is a try to implement this process
> >> in userspace, while vhost-net in kernel.
> >
> > What's the motivation for doing this?
> 
> I haven't looked carefully at the patches, but I don't think we should
> "cheat" when it comes to virtio-net.  I think it's useful to have a
> baseline proof of concept but I think we should focus on making the
> network layer re-entrant.
> 
> In terms of why even bother, if we can make virtio-net a viable
> alternative to vhost-net, it's a huge win.  Being in userspace is a huge
> advantage.
> 
> I understand the challenges with zero-copy but we really need a proper
> userspace implementation to determine how much it really matters.
> Zero-copy tap is also not entirely outside the realm of possibility.

This is exactly what we have, vhost-net is the interface we use
for asynchronous communication. If you want to add yet another
asynchronous interface in kernel, it's certainly doable but then what's
the point?

> >> The iperf's result show it improving the perfermance of base line,
> >> but still slower than vhost-net (maybe the rx path need optimized).
> >> 
> >> Todo:
> >> implement fine lock for net-subsystem
> >
> > vhost-net is currently the only way to get zero copy transmit
> > on linux (and zero copy receive in the future)
> > so at least in theory it'll always be faster.
> 
> It might always be faster.  But that doesn't mean we should limit the
> performance of virtio-net in userspace.  Some people may be willing to
> take a small performance hit in order to obtain the security offered by
> being in userspace.
> 
> Regards,
> 
> Anthony Liguori

By 'being in userspace' I presume you mean the ring processing code.
Ring processing can be done in userspace if you are ready
to use the synchronous read/write operations, skipping
this bunch of code might improve security. But it requires a data
copy almost by definition.


> >
> >> Liu Ping Fan (9):
> >>   vring: split the modification and publish of used ring
> >>   vring: introduce vring_restore() to restore from img
> >>   event poll: make epoll work for normal fd
> >>   event poll: pass event type to event callback
> >>   event poll: enable event poll handle more than one event each time
> >>   virtio net: introduce dataplane for virtio net
> >>   tap: make tap peer work on dedicated data-plane thread
> >>   virtio net: enable dataplane for virtio net
> >>   configure: make virtio net dataplane configurable
> >> 
> >>  configure                  |   12 ++
> >>  hw/dataplane/Makefile.objs |    4 +-
> >>  hw/dataplane/event-poll.c  |   69 +++++--
> >>  hw/dataplane/event-poll.h  |   14 ++-
> >>  hw/dataplane/virtio-blk.c  |    8 +-
> >>  hw/dataplane/virtio-net.c  |  444 
> >> ++++++++++++++++++++++++++++++++++++++++++++
> >>  hw/dataplane/virtio-net.h  |   32 ++++
> >>  hw/dataplane/vring.c       |   37 ++++
> >>  hw/dataplane/vring.h       |    3 +
> >>  hw/virtio-net.c            |   94 +++++-----
> >>  hw/virtio-net.h            |   64 +++++++
> >>  hw/virtio-pci.c            |    2 +-
> >>  include/net/tap.h          |    6 +
> >>  net/tap.c                  |   92 +++++++++-
> >>  14 files changed, 806 insertions(+), 75 deletions(-)
> >>  create mode 100644 hw/dataplane/virtio-net.c
> >>  create mode 100644 hw/dataplane/virtio-net.h
> >> 
> >> -- 
> >> 1.7.4.4
> >>