Re: [Qemu-devel] [PATCH] Use existing tun/tap network interface

[Thomas Schwinge]

On Sun, Nov 28, 2004 at 05:36:48PM +0100, Lars Munch wrote:
> On Sun, Nov 28, 2004 at 05:07:20PM +0100, Thomas Schwinge wrote:
> > On Sat, Nov 27, 2004 at 11:54:30AM +0100, Lars Munch wrote:
> > > The attached patch adds a new option '-tun-if' which will enable you to
> > > use preconfigured tun/tap network interfaces as described here:
> > > http://user-mode-linux.sourceforge.net/UserModeLinux-HOWTO-6.html#ss6.7
> > 
> > I've been preparing a patch to achieve the same functionality some days
> > ago, but didn't have the time to finish it, yet.
> > It is, however, working: 'qemu-net-if.patch' is attached.
> > 
> > Things to be done:
> > * Make it possible to use '-net-if ...' and '-tun-fd ...' at the same
> >   time without interfering with each other - albeit I don't know if
> >   anybody will use both of them at the same time, ever.
> 
> My patch takes care of that.

Good.

> > * Disable the invocation of a tap/tun network init script when
> >   specifying '-net-if ...' or '-tun-fd ...'
> 
> With my patch you can still use a tap/tun network init script for the
> NICs not created by '-net-if' or '-tun-fd.

Yes, I forgot that the same script is used for all interfaces - you
could use that with my patch, too.
If every interface had it's own specific script, it would be reasonable
to disable it by default when using a preconfigured interface.
But since it is not, I'd also just leave that alone.

> > > Then start qemu with something like this:
> > > 
> > > # qemu -tun-if qemu0 .....
> > 
> > I'm adding '-n /bin/true' here.
> 
> I don't have to.

Of course I also don't have to, but I don't want to get
'/etc/qemu-ifup: could not launch network script' every time running
qemu.
;-)

> +static int net_if_init(NetDriverState *nd, char *ifname)
> +{
> +    pstrcpy(nd->ifname, sizeof(nd->ifname), ifname);
> +
> +    nd->fd = tun_open(nd->ifname, sizeof(nd->ifname));
> +    if (nd->fd < 0)
> +        return -1;
> +
> +    nd->send_packet = tun_send_packet;
> +    nd->add_read_packet = tun_add_read_packet;
> +    return 0;
> +}
> +

Why do you need this function?
(Sorry, I don't have the time to dig through qemu's networking code at
the moment.)
Is it because I am doing
#v+
            case QEMU_OPTION_net_if:
                net_if_type = NET_IF_TUN;
                if (nb_net_ifs < MAX_NICS) {
                    pstrcpy (nd_table[nb_net_ifs].ifname, 
sizeof(nd_table[nb_net_ifs].ifname), optarg);
                    nb_net_ifs++;
                }
                break;
#v-
..., whereas you are putting the devices' names into an array and copy it
to the interfaces' structure later?
What's the advantage of your approach compared to mine?

> +            case QEMU_OPTION_tun_if:
> +                net_if_type = NET_IF_TUN;
> +                if (nb_tun_fds+nb_tun_ifs < MAX_NICS) {
> +                    pstrcpy(tun_ifs[nb_tun_ifs], sizeof(tun_ifs[0]), optarg);
> +                    nb_tun_ifs++;
> +                }
> +                break;

> +            } else if (i < nb_tun_fds+nb_tun_ifs) {
> +                if (net_if_init(nd, tun_ifs[i-nb_tun_fds]) < 0)
> +                    net_dummy_init(nd);
>              } else {
>                  if (net_tun_init(nd) < 0)
>                      net_dummy_init(nd);


Regards,
 Thomas