<< | Thread Index | >> ]    [ << | Date Index | >> ]

Subject: Re: [PATCH][RFC] packets received from cipe device are exactly 5 bytes to big
From: Olaf Titz <olaf,AT,bigred,DOT,inka,DOT,de>
Date: Tue, 6 Nov 2001 21:03:03 +0100
In-reply-to: <20011101181743.A24339@gnu.org>

> See subject.  Packets received from a CIPE protocol 4 device (i.e. ethernet
> device) are exactly 5 bytes too big at the hardware level.  I suspect this
> has something to do with the 'pad up to 3 mod 8 and prepend 5 bytes' thing
> that the protocol documentation that I've briefly glanced over describes,
> but I'm not really sure.  I'm using the attached patch for now, but it
> isn't exactly clean.

Strange. The 5 bytes are already subtracted in cipe_decrypt, see the
last few lines of encaps.c (and since that code really hasn't changed
for years, I would be very surprised if such a basic bug could
survive... but then IP packets contain a length field of their own.)

Of course there must be another bug if that fixes it for you. Perhaps
something Ethernet-specific.

> This patch _finally_ makes it possible for me to ping with 5k packets over
> a bridged CIPE tunnel (with the MTU forced up to the ethernet MTU, see
> separate email) by the way, so I'm really wondering whether anyone
> has ever put this into production use succesfully.  But that's a wholly

At least _I_ have not put that into production use. ;-) I have tested
bridging but surely not all corner cases like maximum sized packets or
fragmentation. Since these are a recurring problem with all kinds of
tunnels I wouldn't be surprised of any bugs or incompatibilities in
that area.

This is an issue which needs further investigation.

Olaf





<< | Thread Index | >> ]    [ << | Date Index | >> ]