opportunistic service registration fails

[Miika Komu]

Opportunistic registration to e.g. RVS fails. From Joakim from the mailing list:

A short overview, just tell me if you need more info:

I have 1 RVS server and 2 clients. On the server I use the default hipl config, doing only a 'hipconf daemon add service rvs'.
Client 1 tries to register to the RVS, and client 2 ping client 1. No firewalls on the hosts (iptables accepts all). One of the client is behind a nat, the other (as well as the RVS) has a public ip. Doesn't matter which one registers, and which tries to ping, both cases produce the same errors.

When client 1 register, the RVS server produces a bunch of these:

error(libcore/builder.c:1975@hip_verify_network_header): Invalid version in received packet. Dropping
error(libcore/message.c:677@read_control_msg_all): verifying network header failed
error(libhipl/hip_socket.c:137@handle_nat_input): Reading network msg failed

But after a while, or more often after trying it again (doing another 'hipconf daemon add server rvs <ip> <time>'), it apparently gets through, logging on the RVS:
error(libcore/hostsfiles.c:432@hip_map_hit_to_lsi_from_hosts_fi: Failed to map id to hostname
info(libhipl/input.c:1775@hip_handle_i2): Reached ESTABLISHED state
error(modules/cert/hipd/cert.c:68@hip_add_certificate): Certificate is NULL

After that there's a lot of:
info(libhipl/input.c:534@hip_receive_control_packet): Src IP: <rvs client ip>
info(libhipl/input.c:535@hip_receive_control_packet): Dst IP: <rvs serve ip>

on the RVS server, as the client apparently has problems receiving the packets the server is sending, printing:
info(libhipl/input.c:534@hip_receive_control_packet): Src IP: <rvs server ip>
info(libhipl/input.c:535@hip_receive_control_packet): Dst IP: <rvs client ip>
error(libhipl/input.c:658@hip_receive_udp_control_packet): receiving of control packet failed

There seems to be some sort of retransmission-limit, as eventually it just dies out.. stops trying.

Then, on the second client, we try to ping the first one through the RVS:

'hipconf daemon add map <client 1 hit> <rvs server ip>':
error(libcore/gpl/nlink.c:879@parse_rtattr): Deficit len 28, rta_len=0
error(libcore/hostsfiles.c:123@for_each_hosts_file_line): Failed to open hosts file
error(libcore/hostsfiles.c:432@hip_map_hit_to_lsi_from_hosts_fi: Failed to map id to hostname

..which doesn't look good. But ignoring that, we try 'ping6 <client 1 hit>'.

The RVS server does, apparently, receive something, because the log starts getting full of:
error(libhipl/input.c:658@hip_receive_udp_control_packet): receiving of control packet failed

but nothing else, and this doesn't change. However, using another host as RVS may get you:
info(libhipl/input.c:534@hip_receive_control_packet): Src IP: <client 2 ip>
info(libhipl/input.c:535@hip_receive_control_packet): Dst IP: <rvs ip>
info(libhipl/input.c:1218@hip_check_i1): Matching relay record found.
error(libcore/gpl/nlink.c:879@parse_rtattr): Deficit len 28, rta_len=0
error(libhipl/pkt_handling.c:170@hip_run_handle_functions): Error after running registered handle function, dropping packet...

.. which seems ok. But on client 1 (the one registered to the rvs), there's just a lot of:
info(libhipl/input.c:534@hip_receive_control_packet): Src IP: <rvs ip>
info(libhipl/input.c:535@hip_receive_control_packet): Dst IP: <client 1 ip>

And nothing else happens. Does this sould familiar to anyone? Does anyone else have a problem with this or could it be something in the environment I'm using (a mix of debian stable/unstable/testing)?

[Jookos (jookos)]

Did a quick trace and the problem apparently is that during opportunistic bexes, when receiving the R1 a new hadb entry is created with the proper HITs, and the old one removed. While creating this new entry, the hip_version is never set. This results in a faulty I2.

Attached is a simple patch which wfm.

[Jookos (jookos)]

Here's the patch

[Miika Komu (miika-iki)]

Thanks for the patch, I tested it and seemed work just fine.