Hello.
I setup local ip interface on openflow vlan (openflow-only mode not hybrid) on 5500EI (5.20.99 R2221P02).
The interface is not working and generate many openflow errors.
Partial config:
vlan 10
#
interface Vlan-interface10
ip address 10.10.10.101 255.255.255.0
#
openflow instance 10
controller 1 address ip 192.168.56.7
description vlan10
classification vlan 10
#
Interface is "visible" in openflow instance as "VLAN10" (see vlanif_ports.png).
There is many errors on setup and discovery (see vlanif_setup_1.png and vlanif_setup_2.png) (It should be corrected too.).
The main problem is simple ping (see vlanif_ping.png):
1) ping begins with arp request/response (host is connected on port2 and ip is 10.10.10.15, vlan10 interface on switch have ip 10.10.10.101)
- 22:28:46.751 - arp request accepted from switch (PACKET_IN)
- 22:28:46.755 - flooded (PACKET_OUT)
- 22:28:46.756 - arp resonse accepted (PACKET_IN)
- 22:28:46.758 - flooded (PACKET_OUT)
2) problem A
- 22:28:46.844 - unexpected error for flooded packet (22:28:46.758), but the host received the packet without problem, on switch there was no debug error message (debugging openflow error + terminal debugging).
3) now the host has mapping ip/mac address and send icmp ping
- 22:28:52.012 - icmp echo request accepted from switch (PACKET_IN)
- 22:28:52.012 - flooded (PACKET_OUT)
4) problem B
- 22:28:52.018 - unexpected error for flooded packet (22:28:52.012), packed does not continue to destination and response (icmp echo pong) is not generated
Questions:
1) How to interpret errors "Type: OFPET_BAD_REQUEST (1)/Code: OFPBRC_EPERM (5)" ?
2) Is it needed to reconfigure something on switch or it is a firmware bug ?
3) Is the "VLAN10" interface useful (eg. Can it be used as routing destination output port instead of action "NORMAL" ?) ?
There is also other useful files other.zip:
- vlanif_ping.pcapng
- vlanif_setup.pcapng
- there are screenshots of correct communication on the same switch between two host (hwif_*)
Additional questions for normal correct flows (see hwif_ping.png in other.zip):
+1) Why pathdeamon connect "arp" path after arp request/response from both sides (eg. after 4 packets - 20:58:30.650). I think the there is possibility to connect "arp-path" and "ip-path" after first arp response ?
+2) There is the same situation for "ip-path" (after ip packet flooded from on side to another - 20:58:30.861 + 20:58:31.605). Arp is not sufficient or secure ?
Thanks for answers, Martin Cerveny