Hi Miika, I tried all of your instructions, and several recompiles... And still no luck. It seams to be the same issue as i sent before the IETF in Dallas, with kernel 2.6.15. Here is the relevant part of log (i think): --- LOG --- debug(hadb.c:615@hip_hadb_add_peer_addr): SPI is 0, set address as the bex address debug(nlink.c:567@hip_iproute_get): debug(debug.c:430@hip_print_hit): dst addr :: 2001:0690:2380:7770:0213:d4ff:fe8d:439e error(nlink.c:993@parse_rtattr): !!!Deficit len 28, rta_len=0 debug(debug.c:430@hip_print_hit): src: 2001:0690:2380:7770:0211:d8ff:feec:f02a debug(hadb.c:308@hip_hadb_add_peer_info): Source address found debug(hadb.c:325@hip_hadb_add_peer_info): HA: 0x8084b30, refcnt decremented to: 1 error(hipd.c:305@hip_exit): Signal: 2 debug(close.c:9@hip_send_close): msg=(nil) error(close.c:39@hip_xmit_close): State 0, not sending CLOSE --- LOG END --- Do you notice anything special ? Miika Komu wrote: > On Wed, 19 Apr 2006, Stephen Herborn wrote: > >> I did some experiments on HIPL in mid-2005 and have recently come back >> to extend on them. I'm now using the latest release of HIPL which is a >> bit different to the version I used last year. I am confused by the >> dummy0 interface and 'HITs' that are assigned to the dummy0 interface >> (e.g. "1197:fbdd:d462:...."), what is their purpose and why do they >> differ from the 'traditional' HITs (e.g. "40db:40c7:5b5e:....") that >> are returned after running 'hipconf add hi default'? Which HITs >> should I be using for the /etc/hip/hosts file or 'hipconf add map' >> command? > > The ones beginning with one. The format of HITs was changed in the > latest drafts. > >> Here is exactly what I am doing to test HIPL: >> >> 1. crash: modprobe hipmod; hipconf add hi default; hipd >> 2. oops: modprobe hipmod; hipconf add hi default; hipd > > The latest code does not require hipmod anymore (*). Also, as the hipd > is now in userspace, it reads the configuration files by itself, so > calling hipconf is not required anymore. > >> 3. crash: ping6 IPaddr_of_oops (just to check that that the routing >> tables etc. are OK) >> 4. oops: ping6 IPaddr_of_crash (just to check that that the routing >> tables etc. are OK) >> 5. crash: hipconf add map HIT_of_oops IPaddr_of_oops >> 6. oops: hipconf add map HIT_of_crash IPaddr_of_crash >> 7. crash: ping6 HIT_of_oops >> >> * When I try the above using any of the dummy0 HITs the ICMP echo >> requests are passed to the local dummy0 interface, but nothing happens >> after that. >> >> * When I try the above using any of the "traditional" HITs the ICMP >> echo requests are sent out from eth0 of crash, and received by eth0 of >> oops, but then dropped (dmesg says "ICMPv6 checksum failed"). > > I think you haven't touched your kernel for a long time. You have to > have to get vanilla linux kernel version 2.6.16.5 and patch it with the > following patches: > > patches/beet > * simple-beet-ph-patch-v1.0-2.6.16.5 > * interfamily-beet-ph-patch-v1.0-2.6.16.5 > > and these patches are also recommended: > > patches/kernel > * policy-sleep-2.6.16.5-v2.patch > * hipmod-2.6.16.5-v1.patch (currently optional, but may be useful later > (*)) > > See also the requirements from the manual: > > http://hipl.hiit.fi/hipl/manual > > that you have IPv6, IPsec ESP v4 and v6, SHA1, 3DES, XFRM_USER, AES and > supported. > > (*) We are working actively on another kernel module that will re-enable > support just for native HIP APIs, but nothing else. >