Re: [V2 PATCH 2/2] KVM test: Add vhost-net support

[Lucas Meneghel Rodrigues <lmr@redhat.com>]

On Fri, 2010-10-08 at 12:19 -0300, Lucas Meneghel Rodrigues wrote:
> On Sun, 2010-09-26 at 10:28 +0800, Jason Wang wrote:
> > Vhost is a kernel-level backend for virtio. This patch add a nic_params named
> > "vhost" to enable/disable vhost backend.
> 
> Jason, I had rebased your patches. The patches themselves look good, but
> apparently the syntax proposed here is invalid, and I've tried with
> different qemu variants:
> 
> - qemu from rhel6
> - qemu-kvm.git HEAD
> - qemu.git HEAD
> 
> All of them don't support vhost with netdev:

10:40:01 ERROR| VM could not be created; qemu command failed:
/root/autotest/client/tests/kvm/qemu -name 'vm1' -monitor unix:'/tmp/monitor-humanmonitor1-20101008-103957-4nMr',server,nowait -serial unix:'/tmp/serial-20101008-103957-4nMr',server,nowait -drive file='/tmp/kvm_autotest_root/images/f13-64.qcow2',index=0,if=virtio,boot=on -device virtio-net-pci,mac=9a:0b:5d:7a:c1:7b,netdev=idlTBuUt -netdev user,id=idlTBuUt,vhost=on,hostfwd=tcp::5000-:22 -m 512 -smp 2 -drive file='/tmp/kvm_autotest_root/isos/linux/Fedora-13-x86_64-DVD.iso',media=cdrom,index=1 -vnc :0
10:40:01 ERROR| Status: 1
10:40:01 ERROR| Output:
qemu: -drive file=/tmp/kvm_autotest_root/images/f13-64.qcow2,index=0,if=virtio,boot=on: Invalid parameter 'boot'
qemu: -netdev user,id=idlTBuUt,vhost=on,hostfwd=tcp::5000-:22: Invalid parameter 'vhost'

Ok, the 'drive' problem I am going to fix, let's consider only 'vhost'.
When I read all the different qemu helps, it seems that vhost is tied to
-net:

-net tap[,vlan=n][,name=str][,fd=h][,ifname=name][,script=file][,downscript=dfile][,sndbuf=nbytes][,vnet_hdr=on|off][,vhost=on|off][,vhostfd=h]
                connect the host TAP network interface to VLAN 'n' and use the
                network scripts 'file' (default=/etc/qemu-ifup)
                and 'dfile' (default=/etc/qemu-ifdown);
                use '[down]script=no' to disable script execution;
                use 'fd=h' to connect to an already opened TAP interface
                use 'sndbuf=nbytes' to limit the size of the send buffer; the
                default of 'sndbuf=1048576' can be disabled using 'sndbuf=0'
                use vnet_hdr=off to avoid enabling the IFF_VNET_HDR tap flag; use
                vnet_hdr=on to make the lack of IFF_VNET_HDR support an error condition
                use vhost=on to enable experimental in kernel accelerator
                use 'vhostfd=h' to connect to an already opened vhost net device

I don't think what I am missing here. I'd like to apply this, but since
I couldn't verify it working with none of the qemu versions I tried, I
am going to hold on for your comments.

Cheers,

Lucas