From mboxrd@z Thu Jan  1 00:00:00 1970
From: "Michael S. Tsirkin" <mst@redhat.com>
Subject: Re: Is fallback vhost_net to =?utf-8?Q?qem?=
 =?utf-8?Q?u_for_live_migrate_available=EF=BC=9F?=
Date: Tue, 27 Aug 2013 07:19:35 +0300
Message-ID: <20130827041935.GA7263@redhat.com>
References: <521C1DCF.5090202@huawei.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: QUOTED-PRINTABLE
Cc: jasowang@redhat.com, kvm@vger.kernel.org, netdev@vger.kernel.org,
	qianhuibin@huawei.com
To: Qin Chuanyu <qinchuanyu@huawei.com>
Return-path: <kvm-owner@vger.kernel.org>
Content-Disposition: inline
In-Reply-To: <521C1DCF.5090202@huawei.com>
Sender: kvm-owner@vger.kernel.org
List-Id: netdev.vger.kernel.org

On Tue, Aug 27, 2013 at 11:32:31AM +0800, Qin Chuanyu wrote:
> Hi all
>=20
> I am participating in a project which try to port vhost_net on Xen=E3=
=80=82
>=20
> By change the memory copy and notify mechanism =EF=BC=8Ccurrently
> virtio-net with vhost_net could run on Xen with good
> performance=E3=80=82TCP receive throughput of single vnic from 2.77Gb=
ps up
> to 6Gps=E3=80=82In VM receive side=EF=BC=8CI instead grant_copy with =
grant_map +
> memcopy=EF=BC=8Cit efficiently reduce the cost of grant_table spin_lo=
ck of
> dom0=EF=BC=8CSo the hole server TCP performance from 5.33Gps up to 9.=
5Gps=E3=80=82
>=20
> Now I am consider the live migrate of vhost_net on Xen=EF=BC=8Cvhost_=
net
> use vhost_log for live migrate on Kvm=EF=BC=8Cbut qemu on Xen havn't =
manage
> the hole memory of VM=EF=BC=8CSo I am trying to fallback datapath fro=
m
> vhost_net to qemu when doing live migrate =EF=BC=8Cand fallback datap=
ath
> from qemu to
> vhost_net again after vm migrate to new server=E3=80=82
>=20
> My question is=EF=BC=9A
> 	why didn't vhost_net do the same fallback operation for live
> migrate on KVM=EF=BC=8Cbut use vhost_log to mark the dirty page=EF=BC=
=9F
> 	Is there any mechanism fault for the idea of fallback datapath from
> vhost_net to qemu for live migrate=EF=BC=9F
>=20
> any question about the detail of vhost_net on Xen is welcome=E3=80=82
>=20
> Thanks
>=20

It should work, in practice.

However, one issue with this approach that I see is that you
are running two instances of virtio-net on the host:
qemu and vhost-net, doubling your security surface
for guest to host attack.

I don't exactly see why does it matter that qemu doesn't manage
the whole memory of a VM - vhost only needs to log
memory writes that it performs.