[Battlemesh] [Olsr-dev] OLSR src IP bug

[Markus Kittenberger]

On Tue, Mar 22, 2011 at 4:53 PM, Henning Rogge <hrogge at googlemail.com>wrote:

> On Tuesday 22 March 2011 16:44:18 Benjamin Henrion wrote:
> > Hi,
> >
> > During the WBMv4, we encountered a bug in OLSR which was running on an
> > ath0:4 interface, and it was taking the src ip address of the ath0
> > interface, which it should not.
>
To clarify this:

the olsrd didn't use the ath0 adress,..

the olsrd messges itself where using correct src adress (ath:4)
but the olsrd routes did not specify the src ip (so it`s up to the kernel to
choose *G)

but the kernel did choose the wrong source ip for local originated traffic
(e.g. ping) to "olsrd" routed targets that where only one hop away,..
(while targets multiple hops away didn't have this problem, and i have no
clue why the kernel's src-ip decision differs between one or more hops,.. it
might be even a kernel bug imho,..)

but infact this had no consequence on olsrd operation/performance in the
battlemesh setup, that's why we didn't even change/fix it,..
but in even more "artificial" setups as the battlemesh, this behaviour could
lead to severe problems,..

therefore olsrd will in future activate its srcip routing mode by default
(while currently it needs to be enabled in the configfile)

Markus

> >
> > According to Aaron, it was a modification done by some developer to
> > make things easier for most users.
>
just forget about this,.. *G

(as no previous version of olsrd would behaved differently regarding this,
and likely no previous kernel would,..)

> Why does this make things easier ?

i don`t know aswell *GG

>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://ml.ninux.org/pipermail/battlemesh/attachments/20110322/9da81159/attachment.htm>