\begin{Darrell Burkey}
> I have two Linux boxes connected to separate Bigpond Direct accounts via
> modems that dial in to the same Telstra router. Anytime the two boxes talk
> to each other I get the following messages on the consoles of both machines:
>
> Dec 29 08:45:26 adminserver kernel: Advised path = 203.48.63.65 ->
> 139.130.12.155, tos 00
> Dec 29 08:45:27 adminserver kernel: Redirect from 139.130.12.129 on ppp0
> about 139.130.12.155 ignored.
>
> This is also filling up the system logs and basically it's very annoying. My
> understanding is that the Cisco router at Telstra is simply advising the
> Linux boxes that there is a direct path between them and that the routing
> tables should be updated to reflect this. ???
yep. ICMP redirect messages. normal and proper.
> 1. Eliminate the cause of the messages.
> I don't have control over the router generating the message so that option
> is out. I also thought that perhaps I could add a static route on each box
> to the other but that doesn't seem to work. I find it a bit confusing as
> each box has a network address plus a local and remote ip address being fed
> to pppd via options. These addresses were assigned by Telstra. Anyway, I've
> tried every combination of routing table entries and I still get the
> messages.
>
> 2. Replace klogd with something like megalog that allows regex matching so I
> can filter out the messages. I've read up on this and it looks like a strong
> possibility but I'd rather eliminate the cause of the messages if possible.
>
> So, anyone have any tricks up their sleeves?
a quick grep of my nearest kernel source (2.2.19), showed:
net/ipv4/route.c:
reject_redirect:
#ifdef CONFIG_IP_ROUTE_VERBOSE
if (IN_DEV_LOG_MARTIANS(in_dev) && net_ratelimit())
printk(KERN_INFO "Redirect from %X/%s to %X ignored."
"Path = %X -> %X, tos %02x\n",
ntohl(old_gw), dev->name, ntohl(new_gw),
ntohl(saddr), ntohl(daddr), tos);
#endif
there's numerous reasons the code jumps to reject_redirect:
(from my quick understanding of the if blocks)
new gateway == old gateway, some multicast stuff, new address is 0.0.0.0,
if the new address isn't in the range of the network cards addresses,
its a unicast interface
i'll let you work out why its rejecting the redirect.
so, to not see these messages, do one of the following:
work out why its being rejected, and avoid that. (will probably involve
reading the relevant section of route.c)
undefine CONFIG_IP_ROUTE_VERBOSE (its a kernel config option, iirc)
make sure the kernel isn't logging KERN_INFO messages to the console
directly:
check /proc/sys/kernel/printk and the relevant section in
Documentation/sysctl/kernel.txt
make sure syslog isn't logging kern.info messages to the console
(/etc/syslog.conf and syslog.conf(5))
--
- Gus
--
SLUG - Sydney Linux User Group Mailing List - http://slug.org.au/
More Info: http://lists.slug.org.au/listinfo/slug
