Re: [PATCH V4 1/3] net/filter: Remove vnet_hdr from filter-mirror and filter-redirector

[Jason Wang]

On Wed, Oct 27, 2021 at 2:40 PM Zhang, Chen <chen.zhang@intel.com> wrote:
>
>
>
> > -----Original Message-----
> > From: Jason Wang <jasowang@redhat.com>
> > Sent: Wednesday, October 27, 2021 2:24 PM
> > To: Zhang, Chen <chen.zhang@intel.com>
> > Cc: qemu-dev <qemu-devel@nongnu.org>; Markus Armbruster
> > <armbru@redhat.com>; Li Zhijian <lizhijian@cn.fujitsu.com>
> > Subject: Re: [PATCH V4 1/3] net/filter: Remove vnet_hdr from filter-mirror
> > and filter-redirector
> >
> >
> > 在 2021/10/27 下午2:19, Zhang, Chen 写道:
> > >>
> > mirror,id=id,netdev=netdevid,outdev=chardevid,queue=all|rx|tx[,vnet_h
> > >> dr _support][,position=head|tail|id=<id>][,insert=behind|before]``
> > >>> +    ``-object
> > >>> + filter-mirror,id=id,netdev=netdevid,outdev=chardevid,queue=all|rx|
> > >>> + tx [,position=head|tail|id=<id>][,insert=behind|before]``
> > >> I wonder if we break management layer. If yes, maybe it's better to
> > >> keep the vnet_hdr_support here.
> > > Yes and no,   With this series of patches, filters have ability to 
> > > automatically
> > > Configure the appropriate vnet_hdr_support flag according to the current
> > environment.
> > > And can report error when can't fixing the vnet_hdr(The user cannot fix it
> > from the previous way ).
> > > So I think no need for the user to configure this option, some relevant
> > background knowledge required.
> > >
> > > For the management layer, keep the vnet_hdr_support may be
> > meaningless except for compatibility.
> > > In this situation, Do you think we still need to keep the vnet_hdr_support
> > for management layer?
> >
> >
> > So it depends on whether management layer like libvirt has already
> > supported this. If yes, we may get errors using new qemu with old libvirt?
>
> As far as I know, Current management layer like upstream libvirt is no COLO 
> official support yet.
> And some real CSPs use libvirt passthrough qmp command to Qemu for manage 
> COLO VM.

So the question still, it looks to me it requires the modification of
the layers on top of libvirt? If the answer is yes, we'd better keep
that compatibility.

> It is no harm to users to reduce some unnecessary parameters. But if you 
> think compatibility is
> more important, I will restore this parameter in next version.

Thanks

>
> Thanks
> Chen
>
>
>
>
> >
> > Thanks
> >
> > > Enable/disable it do the same things for filters.
> > >
> > > Thanks
> > > Chen
> > >
>