[Qemu-devel] Re: [PATCH-RFC 09/13] tap: add vhost/vhostfd options

["Michael S. Tsirkin" <mst@redhat.com>]

On Tue, Jan 12, 2010 at 04:39:52PM -0600, Anthony Liguori wrote:
> On 01/11/2010 11:22 AM, Michael S. Tsirkin wrote:
>> Looks like order got mixed up: vhost_net header
>> is added by a follow-up patch. Will be fixed
>> in the next revision.
>>
>> Signed-off-by: Michael S. Tsirkin<mst@redhat.com>
>> ---
>>   net.c           |    8 ++++++++
>>   net/tap.c       |   29 +++++++++++++++++++++++++++++
>>   qemu-options.hx |    4 +++-
>>   3 files changed, 40 insertions(+), 1 deletions(-)
>>
>> diff --git a/net.c b/net.c
>> index 6ef93e6..b942d03 100644
>> --- a/net.c
>> +++ b/net.c
>> @@ -976,6 +976,14 @@ static struct {
>>                   .name = "vnet_hdr",
>>                   .type = QEMU_OPT_BOOL,
>>                   .help = "enable the IFF_VNET_HDR flag on the tap interface"
>> +            }, {
>> +                .name = "vhost",
>> +                .type = QEMU_OPT_BOOL,
>> +                .help = "enable vhost-net network accelerator",
>> +            }, {
>> +                .name = "vhostfd",
>> +                .type = QEMU_OPT_STRING,
>> +                .help = "file descriptor of an already opened vhost net device",
>>               },
>>    
>
> Semantically, I think making vhost it's own backend makes more sense  
> from a user perspective.


It doesn't. Users mostly do not care that vhost is used: they just get
fast virtio and that's all.  Users do care about e.g. tap because they
need setup scripts, understand bridging etc.


Do you propose -net tap be replaced with -net vhost?  This means vhost will
need to get tap flags if it is attached to tap and raw flags if attached
to raw, etc.

A separate backend that only works with virtio frontend is also
ugly. OTOH an option that has effect only with virtio frontend
is pretty usual, vnet_hdr is one such example.


>
> I don't think it's a significant code change.
>
> Regards,
>
> Anthony Liguori

-- 
MST