(usagi-users 03826) Re: mip6d-daemon multiple interfaces

Masahide NAKAMURA <[email protected]> Wed, 21 Mar 2007 13:49:42 +0900
Newsgroups gmane.linux.ipv6.usagi.users
Message-ID <[email protected]>
Hi,

Please find my comment inline:

Monday 12 March 2007 22:14, Boutin Maël wrote:
> >Can you show us more information just in case?
> > o the kernel version; Please include git commit-id if you are using git tree
> > o the oops log message if you can copy it
> 
> Kernel version : 2.6.20.1
> 

Please read below "Known issues".

http://www.linux-ipv6.org/memo/mipv6/

As described the above URL document, MN doesn't work on 2.6.20.1.


Friday 09 March 2007 22:34, Boutin Maël wrote:
> I loaded mipv6-demon and there is a problem with the content of file
> configuration (mip6d.conf) on mobile node side (MN) :
> 
> With only one interface written into the file there is no problem. But when
> I add an other interface (eth1) with a preference on eth0 in the mip6d.conf,
> there is a kernel oops in the following situations:
> 
> - If the home address are different on each interface, when i stop mipl
> there is a kernel oops.
> 
> - if the home address for the both interface is the same, a kernel oops
> appears after sending some BUs.

Finally I don't have kernel Oops at least when I've tested your two senario on
following versions even MN does not work well.
(Just I confirmed is that at one senario multiple HoA makes two BCEs on HA
correctly and at the another the HoA makes one BCE on HA using the priority device.
Packet sending is not tested.)

kernel:
	either kernel.org's 2.6.20.1, or
	our git's linux-2.6-mip6, branch=linux-2.6.21-rc2-mip6
daemon:
	our git's mipv6-daemon, branch umip-20070307

This is not a fix but can you try kernel without af_key.o since it seems that
reading null pointer is caused at pfkey path which is not required
if you want to test it even before MN basic behavior becomes stable?


Regards,


-- 
Masahide NAKAMURA