---------- Forwarded message ----------
From: BlueFlux
X-Sender: blueflux@swipnet.se
To: aaa-list
Cc: dc-stuff
Date: Wed, 6 May 1998 22:14:57 +0200 (MET DST)
Subject: Re: 2.1.99 sends ULTRIX death packets (fwd)
Ok, some people requested this information. I'm sorry i couldnt find the
original message since it's deleted. Anyways, this is all the info I can
get at the moment. Perhaps i'll get something more usefull to send but I
highly doubt it since this thread was almost totally dead until davem sent
this. For those who dont know, David S. Miller is the maintainer of linux
TCP/IP stack so he /should/ be considered a reliable source of
information=).
For those who dont know what this is about, SACK is a relatively new
package used by OS's and the TCP code of them. The ULTRIX OS' dont know
how to handle these packages so they will instantly die on receiving them.
The /big/ problem here is that ULTRIX is no longer mainained by any group
of developers so this bug will most probably stay around for quite some
time, On the other hand I also heard from the same list that there is an
NYC based company that is realy developing ULTRIX[1].
[1] Very uncertain about this, its greatly apreciated if someone else
could bring up some information about this since i dont have that message
lying around but that mail had no references to any company address or
anything but it at least stated that there was some company around.
P.S. This mail was requested by mea culpa and some more people on IRC
#dc-stuff, so i thought it would be better to send it to the list than
rather to each and one of the persons as i know there is more people out
there that is interested in this information.
--------------------------------------------------------------------------
http://blueflux.home.ml.org
--------------------------------------------------------------------------
"Give me a dozen atom smashers, an underground laboratory, and a girl in a
diaphanous veil waiting to be turned into a chimpanzee, and I care not who
writes the nations laws!"
--------------------------------------------------------------------------
-BlueFlux
---------- Forwarded message ----------
Date: Mon, 4 May 1998 13:51:01 -0700
From: "David S. Miller"
To: vonbrand@sleipnir.valparaiso.cl
Cc: joe@alla.as.utexas.edu, linux-kernel@vger.rutgers.edu, fletch1@mit.edu,
secretary@lists.gnacademy.org
Subject: Re: 2.1.99 sends ULTRIX death packets
Date: Mon, 04 May 1998 14:18:10 -0400
From: Horst von Brand
That depends... anyway, Linux should _not_ generate illegal
packages that might crash other systems. What happened to "Be
conservative with what you send, liberal with what you accept"?
This is not the realm I think the bug in question is within'.
I think the problem is question is more likely to be "Ultrix can crash
when fed TCP packets containing RFC1323/SACK TCP options as specified
in the RFC's" which did not exist at the time the Ultrix networking
code was written. However the RFC's which were present at the time of
the writing of the Ultrix networking code, say to ignore TCP options
which one does not understand. This is precisely why all TCP options
have a specific format (optiontype byte, optionlength byte, optiondata
bytes), and are negotiated at connection startup. These connection
startup negotiation options are probably what Ultrix is locking up
on.
But, I can't be sure of this until someone sends me a packet trace of
what gets sent over the wire which causes the Ultrix box to go down.
Later,
David S. Miller
davem@dm.cobaltmicro.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
|