On Fri, 2004-06-18 at 07:47, Pekka Savola wrote: > Jeroen: I'm using Red Hat's 2.4.20 kernels (RHL73 and RHL9). I'm > having difficulty testing this more widely, as almost everywhere where > I have a Linux box, I also have IPv6 routes ;-) I know about that 'no boxes without IPv6 routes' problem, same here ;) > I just tested this again: after 'modprobe ipv6' and no RA, after a > couple of seconds, the default routes were added on all the > interfaces. When I restarted RA's, the default routes were removed, > replaced by the default route in the RA. /me just remembered he had a redhat box stashed somewhere and amaze: $ ip -6 ro <SNIP others> default dev eth1 proto kernel metric 256 mtu 1500 advmss 1440 default via fe80::207:4fff:fef7:fd dev eth0 proto kernel metric 1024 expires 1731sec mtu 1500 advmss 1440 unreachable default dev lo proto none metric -1 error -101 A default route on eth1 while we have a real one on eth0. Thus some searching in the kernel: From: 2.4.20/net/ipv6/addrconf.c 1575 printk(KERN_DEBUG "%s: no IPv6 routers present\n", 1576 ifp->idev->dev->name); 1577 1578 memset(&rtmsg, 0, sizeof(struct in6_rtmsg)); 1579 rtmsg.rtmsg_type = RTMSG_NEWROUTE; 1580 rtmsg.rtmsg_metric = IP6_RT_PRIO_ADDRCONF; 1581 rtmsg.rtmsg_flags = (RTF_ALLONLINK | RTF_ADDRCONF | 1582 RTF_DEFAULT | RTF_UP); 1583 1584 rtmsg.rtmsg_ifindex = ifp->idev->dev->ifindex; 1585 1586 ip6_route_add(&rtmsg, NULL); Here we go, a default route because there is no IPv6 router. Explains exactly the behaviour above. Oddly enough this is also on 2.6.7 and in 2.4.17. I thus wonder why I never have seen this behavior, which is in one word: bad. And this little piece of code should really be removed. The only scenario where this could be handy is where one configures a lot of addresses on one box, any other box on the same link can then still reach it. This could be the case when RA failed maybe, still I don't like this mechanism at all. Greets, Jeroen
Attachment:
signature.asc
Description: This is a digitally signed message part
