Re: [Qemu-devel] [PATCH 0/5] VMWare PVSCSI paravirtual device implementation

[Avi Kivity <avi@redhat.com>]

On 03/15/2012 01:47 PM, Stefan Hajnoczi wrote:
> On Thu, Mar 15, 2012 at 9:02 AM, Dmitry Fleytman
> <dmitry.fleytman@ravellosystems.com> wrote:
> > Below is the implementation of VMWare PVSCSI device and
> > command line parameters to configure vendor name and product name
> > for SCSI storage are implemented.
> > Latter is needed to make PVSCSI storage devices look exactly as
> > on VMWare hypervisors.
> >
> > With this and VMWARE3 patches V2V migration problem for VMWare
> > images should be solved relatively easy.
>
> What is the V2V strategy?
>
> Supporting these devices is fine if we have a way to convert guests to
> use virtio.  But if the plan is to keep the guests on VMware pv
> devices, then that will split the development effort on support and
> optimizing I/O devices.  All the performance work going into
> virtio-net (vhost-net), virtio-blk, and possibly virtio-scsi isn't
> easy to duplicate for VMware pv devices.  Also, we cannot extend
> VMware devices easily while retaining compatibility.
>

Note, it's possible to hook vmxnet to vhost-net:

 - process the vmxnet ring in qemu
 - translate it into a virtio ring in offguest memory (placing virtio
headers in offguest memory as well)
 - fire off vhost-net
 - etc.

So there are still extra context switches, but we can still get some of
the benefits of vhost-net.

-- 
error compiling committee.c: too many arguments to function