Hi Srini,
Thank you for took a long time to help me, after your explain, now I know
what's the problem, thank you.
But I also do not know how to solve it...
After tests, the following is results for your question:
1.
a) ./dpctl add-flow unix:/var/run/test
dl_type=0x0806,dl_dst=0xFFFFFFFFFFFF,actions=output:FLOOD was incorrect, so
I do some changes like the following.
./dpctl add-flow unix:/var/run/test
dl_type=0x0806,dl_dst=ff:ff:ff:ff:ff:ff,actions=FLOOD
b) Flow table was no use, they were disappearance in seconds.
c) Ping was no use, two hosts could not ping mutually, and no ofpbuf error.
2.
a) Ping was no use, two hosts could not ping mutually, and there is no
ofpbuf errors.
b) Two flows were correct, because n_packets increases when two hosts ping
mutually, like the following:
duration_sec=11s, n_packets=2, n_bytes=148,
idle_timeout=60,hard_timeout=0,icmp,in_port=2,nw_dst=192.168.1.20,actions=mod_vlan_vid:2,output:3
(↑n_packet increase when 192.168.1.10 ping 192.168.1.20)
(↓n_packet increase when 192.168.1.20 ping 192.168.1.10)
duration_sec=19s, n_packets=0, n_bytes=0,
idle_timeout=60,hard_timeout=0,icmp,in_port=3,nw_dst=192.168.1.10,actions=output:2
3.
a) I was no connecting to controller in the 1 and 2 tests.
b) It's no occurrence of any errors or message in OpenFlow switch, I don't
know what log I need to dump.
2011/3/15 Srini Seetharaman <[email protected]>
> Hi Yi-Shou
> I don't see anything suspicious in your control traffic dump or in your NOX
> log. However, there is a mismatch. The OpenFlow switch log has
> packet_in/flow_mod for ARP and the control traffic you mailed had
> packet_in/flow_mod for ICMP. I'm assuming the problem exists in both. So,
> let's try three things:
>
> ==============================
> 1.
> a) Please clear the arp cache in each of the host by doing arp -d.
>
> b) Statically handle the ARP request and the reply. For this, I've crafted
> the same flow_mod through the dpctl command and pasted below. Could you
> please run this on your NetFPGA switch and let me know what happens. I don't
> think you need the controller running for this step.
>
> ./dpctl add-flow unix:/var/run/test
> dl_type=0x0806,dl_dst=0xFFFFFFFFFFFF,actions=output:FLOOD
>
> ./dpctl add-flow unix:/var/run/test
> dl_type=0x0806,in_port=2,dl_vlan=0xffff,dl_vlan_pcp=0x00,dl_src=00:0f:b0:df:38:18,dl_dst=00:17:31:b8:ee:8c,nw_src=0.0.0.0,nw_dst=0.0.0.0,nw_proto=0,tp_src=0,tp_dst=0,actions=mod_vlan_vid:2,output:3
>
> c) After this, please try the ping and see if the switch crashes with the
> ofpbuf error.
> =================================
> 2. Repeat the ping after adding the following static ARP entries in the
> hosts:
>
> arp -s 192.168.0.10 00:0f:b0:df:38:18
> arp -s 192.168.0.20 00:17:31:b8:ee:8c
>
> And, following flow-table entries added in the switch:
> ./dpctl add-flow unix:/var/run/test
> dl_type=0x0800,nw_proto=1,in_port=3,nw_dst=192.168.0.10,actions=output:2
> ./dpctl add-flow unix:/var/run/test
> dl_type=0x0800,nw_proto=1,in_port=2,nw_dst=192.168.0.20,actions=mod_vlan_vid:2,output:3
>
> =================================
> 3. In the OpenFlow switch screen dump you mailed, there is a problem. The
> pkt_in has all fields filled out and corresponds to a ARP reply. While the
> flow_mod for that ARP reply has most of the fields zeroed out!! This is a
> problem. I'm not sure why that happened. We need to check if this repeats
> with ICMP too. So, if you can provide the switch screen dump corresponding
> to the control traffic you mailed earlier, that will help a lot!
>
> Mar 14 02:57:19|00059|vconn|DBG|tcp:140.115.155.180:6633: sent (Success):
> packet_in (xid=0x0): total_len=106 in_port=2 data_len=106 buffer=0x00000104
>
> arp,in_port=2,dl_vlan=0xffff,dl_vlan_pcp=0x00,dl_src=00:0f:b0:df:38:18,dl_dst=00:17:31:b8:ee:8c,nw_src=192.168.0.10,nw_dst=192.168.0.20,nw_tos=0x00,nw_proto=2,tp_src=0,tp_dst=0,
>
> Mar 14 02:57:19|00060|poll_loop|DBG|[POLLIN] on fd 6: 0x8064b20 0x805470e
> 0xafae9c
> Mar 14 02:57:19|00061|vconn|DBG|tcp:140.115.155.180:6633: received:
> flow_mod
> (xid=0x0):arp,in_port=2,dl_vlan=0xffff,dl_vlan_pcp=0x00,dl_src=00:0f:b0:df:38:18,dl_dst=00:17:31:b8:ee:8c,nw_src=0.0.0.0,nw_dst=0.0.0.0,nw_proto=0,tp_src=0,tp_dst=0,
> ADD: cookie:0 idle:5 hard:0 pri:32768 buf:0x104 flg:0x1
> actions=output:3,mod_vlan_vid:2
>
> Thanks
> Srini.
>
--
Best Regards, Yi-Shou Chen (陳懿首),
Mobile and Broadband Networking Lab., CSIE, NCU,
03-4227151 #57968, 0922-717020
_______________________________________________
openflow-discuss mailing list
[email protected]
https://mailman.stanford.edu/mailman/listinfo/openflow-discuss
