netdev
[Top] [All Lists]

Re: IPv6: 2.4.2/3 autoconfiguration sometimes didn't work

To: Bill Robertson <bill.robertson@xxxxxx>, netdev@xxxxxxxxxxx
Subject: Re: IPv6: 2.4.2/3 autoconfiguration sometimes didn't work
From: Peter Bieringer <pb@xxxxxxxxxxxx>
Date: Sun, 22 Apr 2001 11:43:35 +0200
In-reply-to: <4.2.2.20010419075418.0224f5f0@pop.crc.ca>
References: <5.0.2.1.0.20010419092037.00af7d28@mail.bieringer.de>
Sender: owner-netdev@xxxxxxxxxxx
At 15:14 19.04.2001, Bill Robertson wrote:
I have a 2 host network with a 2.2.19 based router running radvd 0.6.2 (not p11) and a 2.2.19 based client and am having the same problem as Peter. Most of the time the client doesn't autoconfigure itself to the site local address being provided in the router advertisements.

I tried both "/etc/rc.d/init.d/network restart" and Peter's suggestion of "ifconfig eth0 down" followed about 60 seconds later by "ifconfig eth0 up". Neither seems to work.

Looks like the old (USAGI already knows about) IPv6 multicast bind bug isn't yet fixed.


After booting, following route occurs:
ff02::1/128 ff02::1 UAC 0 1 1 eth0


If network is restarted *and* network driver module wasn't removed, this route is not setup again by the kernel.
If you stop networking, remove the driver module and start networking again, this route is setup again and autoconfiguration works.


My randomness experiences perhaps depends on the module-autoremovable mechanism (but don't be shure).

Any hints, why this already known bug is still fixed now?

Unlike I hoped, the ndisc.c fix doesn't fix this problem.

        Peter


<Prev in Thread] Current Thread [Next in Thread>