I've updated the AES encryption patch on my web page:

http://www.arnor.net/encryptingbridge/index.html

The code has been cleaned up a bit, with better error checking.  This
fixes crashes I found during testing.  Also, it has been rediffed
against 2.4.19-pre8.

The modified version of brctl has not changed.

Now, the RFC:

One of the things that's tricky about this patch is that it actually
makes the ethernet frames a little bigger as they go through the
bridge.  The data has to be padded to have a length which is a multiple
of 16 bytes before encrypting with AES.

This is almost always not a problem, as the sk_buff's are usually
allocated with enough spare space at the end.  But for some packets
(only small ones in my testing) there is not enough room at the end to
pad the packet.  I'm still learning how this works, but I'm afraid it
may depend on the allocation strategy used by the underlying network
card driver.

As a workaround for now, I just pass these "unpaddable" packets through
unencrypted. That works for my app for now, but is obviously not a real
solution.

I haven't decided the best approach to a proper fix, I suspect that
modifying the alloc_skb() and dev_alloc_skb() routines to always ensure
there's at least 16 bytes extra at the end of the buffer is the best way
to go.

Suggestions welcome...

Torrey

_______________________________________________
Bridge mailing list
[EMAIL PROTECTED]
http://www.math.leidenuniv.nl/mailman/listinfo/bridge

Reply via email to