> > I just emailed you, it's not a fragment level problem but a packet level
> > 'backup' or 'bottleneck' and a protocol/program not designed to account
> > for such bottlenecks.
>
> No, this is not what is happening to me. It happens with all the pipes
> empty and the systems idle, there are no timeouts or bottlenecks involved.
>
> Setting:
>
> *idle* systems
> *idle* PPP (MTU/MRU 296)
> *idle* eth0
> always_defrag is turned on
>
> Some systems completely fail to forward, and changing the PPP MTU/MRU to
> 1500 fixes it. To me, this means they are not defragmenting, and of
> course the fragments can't masquerade. There is a bug in the
> always_defrag.
>
> -Tom
I confirm this! The important thing is the MTU. Even with MTU set to 527
(or whatever) the systems fails to forward... but not completely... I mean
that with MTU 527 the masquaraded hosts manage to see outside to the www
but only some hosts, some still not work, for example www.uu.net works but
www.hotmail.com doesn't. etc...
And, YES, changing the PPP MTU to 1500 fixes this misbehaviour... I noticed
this problem since 2.0.32 (before that I don't know).
Roby
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/