[ovs-discuss] How to properly forward certain protocols to a port?

[Can Zhang]

I got:

# ovs-dpctl dump-flows ovs
skb_priority(0x6),in_port(2),eth(src=a2:05:92:71:79:69,dst=0a:00:27:00:00:00),eth_type(0x0800),ipv4(src=192.168.56.100,dst=192.168.56.1,proto=6,tos=0x10,ttl=64,frag=no),tcp(src=22,dst=56241),
packets:117, bytes:18730, used:0.264s, flags:P., actions:1
in_port(1),eth(src=0a:00:27:00:00:00,dst=a2:05:92:71:79:69),eth_type(0x0800),ipv4(src=192.168.56.1,dst=192.168.56.100,proto=6,tos=0,ttl=64,frag=no),tcp(src=56241,dst=22),
packets:168, bytes:13488, used:0.264s, flags:P., actions:2

And the settings are:

virtualbox host(my laptop,ip 192.168.56.1) --- eth0(no ip) ---- ovs ----
veth0(no ip) --- veth1(ip 192.168.56.100)

And I could ping from host to guest and also ssh to this interface. I think
what I should do next is to install some flow entries to forward those
ARP/ICMPv6/TCP packets to veth1




2013/7/30 Jesse Gross <jesse at nicira.com>

> What do you see if you run ovs-dpctl dump-flows while this is running?
>
> On Mon, Jul 29, 2013 at 7:40 AM, Can Zhang <can at canx.me> wrote:
> > I think I've replied but actually it's in the draft box. Sorry for the
> > delay.
> >
> > I captured packets on both interface using wireshark and as I described
> in
> > the first mail, I can see ARP request at eth0 but not tap0.
> >
> > And now I'm considering using a veth pair, say veth0 and veth1. Attach
> veth0
> > to an ovs and configure an IP address on veth1. As:
> >
> > peer packets --- eth0 ---- ovs ---- veth0 --- veth1(with IP address)
> >
> > The flows installed are similar as before. So the peer could establish
> TCP
> > connections with my machine via veth1's IP address. What do you think of
> it?
> >
> >
> > 2013/7/25 Jesse Gross <jesse at nicira.com>
> >>
> >> And you ran tcpdump on both ports and you see packets on one but not the
> >> other?
> >>
> >> On Wed, Jul 24, 2013 at 5:54 AM, Can Zhang <can at canx.me> wrote:
> >> > NXST_FLOW reply (xid=0x4):
> >> >  cookie=0x0, duration=2938.314s, table=0, n_packets=0, n_bytes=0,
> >> > idle_age=2938, in_port=1 actions=output:2
> >> >  cookie=0x0, duration=2943.819s, table=0, n_packets=130,
> n_bytes=20082,
> >> > idle_age=11, in_port=2 actions=output:1
> >> >
> >> >
> >> > Just forward all flow from 1 to 2 and 2 to 1.
> >> >
> >> >
> >> > 2013/7/24 Jesse Gross <jesse at nicira.com>
> >> >>
> >> >> On Tue, Jul 23, 2013 at 7:11 AM, Can Zhang <can at canx.me> wrote:
> >> >> > Hello,
> >> >> >
> >> >> > Firstly I'll describe a simple scenario I tried:
> >> >> >
> >> >> > peer packets --- eth0 ---- ovs ---- tap0
> >> >> >
> >> >> > I configured tap0 with IP 192.168.57.100, two flows to forward
> >> >> > packets
> >> >> > from eth0 and tap0 to the other side, and pinged from outside to
> the
> >> >> > IP. I
> >> >> > did capture ARP request on eth0 using wireshark, but could not see
> >> >> > any
> >> >> > packets on tap0. It seems that after attach a port to ovs, the IP
> >> >> > address is
> >> >> > no longer effective.
> >> >> >
> >> >> > My goal is to use a machine installed with ovs as a openflow switch
> >> >> > but
> >> >> > it can also handle some packets(say, a normal socket server can
> still
> >> >> > work).
> >> >> > So I'm thinking of extract some packets to tap0 and make the server
> >> >> > listen
> >> >> > on it. But the experiments I did failed as above. I have no idea
> what
> >> >> > to try
> >> >> > next so I come to see if somebody could give me some suggestions.
> >> >>
> >> >> What do the flows that you installed look like? At the Open vSwitch
> >> >> level, IP addresses on the interfaces do not matter.
> >> >
> >> >
> >> >
> >> >
> >> > --
> >> > Best regards,
> >> > Can Zhang
> >
> >
> >
> >
> > --
> > Best regards,
> > Can Zhang
>



-- 
Best regards,
Can Zhang
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://openvswitch.org/pipermail/ovs-discuss/attachments/20130730/85f7d17a/attachment.html>