Re: [Qemu-devel] [Qemu-discuss] changing from net to netdev with vde switches and double nics

[Thomas Huth]

 Hi,

On 28.08.2017 15:58, Paolo wrote:
> Hello everyone,
> 
> I'm at loss as to how to translate from net to netdev syntax the following 
> lines:
> 
> NET0="-net vde,vlan=0,sock=/var/run/vde.ctl00 -net 
> nic,vlan=0,model=rtl8139,macaddr=$NIC0MAC -net 
> tap,vlan=0,ifname=pub2,script=no,downscript=no"
> 
> NET1="-net vde,vlan=1,sock=/var/run/vde.ctl01 -net 
> nic,vlan=1,model=rtl8139,macaddr=$NIC1MAC -net 
> tap,vlan=1,ifname=prv2,script=no,downscript=no"
> 
> Usage for lines above is: qemu blah blah blah $NET0 $NET1 blah blah
> (obviously NIC0MAC and NIC1MAC are declared before in the script)
> 
> these syntax, which has been working flawlessly for ages on my pc, creates 
> two network cards (pubX, prvX) for each one of my virtual machines. Each of 
> the two network cards is attached to a vde switch, so all of pubX cards (X is 
> the machine) are attached to vsw0 (virtaul switch 0) and all of prvX cards 
> are attached to vsw1. The rationale behind this is to attach each qemu 
> machine to a private vlan (172.16.y.z) and to a public vlan (192.168.w.t) at 
> the same time.
> This way the public lan is used when connected to the internet, the private 
> lan works always (i.e. with the router inaccessible) for accessing samba 
> directories on the host (of course, firewalled, ip restricted and so on).
> 
> Translation from net to netdev is simple for the nic (-device) part and for 
> the tap part alike, but how can I tell qemu that the vdeswitch with socket 
> /var/run/vde.ctl00 belongs to the same virtual lan of tap pub2 with device 
> rtl8139 and mac address NIC0MAC?
> 
> Translating from net to netdev we lose the vlan=X info.

That's interesting, I think you're the first person I know in months or
even years who really seems to be using the vlan=x parameter (with x >=
1)...

You're right, you can not translate your setup 1:1 to -netdev anymore,
since there are no 'vlan's (or rather hubs, since 'vlan' is rather a
misnomer here) available with the -netdev parameter.

Question is: Why do you need it at all? I see your point that you want
to have two network cards, but why does each NIC has to be wired to two
host networks (vde *and* tap)? Isn't it sufficient if you connect each
NIC to one VDE network?

 Thomas