[630] in linux-net channel archive
Re: Bad c'sums and skput:over in 1.3.7
daemon@ATHENA.MIT.EDU (Alan Cox)
Sat Jul 8 01:38:29 1995
From: iialan@iifeak.swan.ac.uk (Alan Cox)
To: gpg109@rsphy1.anu.edu.au (Paul Gortmaker)
Date: Fri, 7 Jul 1995 09:34:36 +0100 (BST)
Cc: iialan@www.linux.org.uk, linux-net@vger.rutgers.edu
In-Reply-To: <9507070417.AA00421@rsphy9.anu.edu.au> from "Paul Gortmaker" at Jul 7, 95 02:17:47 pm
> Kernel Panic skput:over: 001be180:60
> Kernel Panic skput:over: 001be180:247
> Kernel Panic skput:over: 001be180:92
That looks a pretty definite bug. Will check now (I dont have a lance to
test on so I changed the driver blind). Ok its a stupidity
There is an extra skb_put() just after the dev_alloc_skb(). This
causes the panic. Remove the line that says /* Make Room */
> On another box (ne2000) it boots okay, but there is a strange problem that
> appears to be bogons. I have a hunch that it is sending out packets with
> bad checksums as they seem to be silently ignored.
There is a bug in Tom May's last minute checksum code fix. See Linus fix
to the net channel.
> But if I repeat the above with a different host (also perfectly valid
> name) everything works fine.
It depends on the last 3 bytes of the packet and its size.
>
> Oh, and another minor nit -- the prototype for eth_copy_and_sum() should
> be stuck in some header file somewhere so gcc doesn't complain when
> compiling the lance driver (and others that use it).
Yep.. minor oversight.
Alan
