Re: [PATCH] ARCNET: fix hard_header_len limit

[Michael Grzeschik <mgr@pengutronix.de>]

On Thu, Jul 30, 2015 at 11:16:36AM -0700, David Miller wrote:
> From: Michael Grzeschik <m.grzeschik@pengutronix.de>
> Date: Thu, 30 Jul 2015 15:34:36 +0200
> 
> > The commit <9c7077622dd9> ("packet: make packet_snd fail on len smaller
> > than l2 header") adds the check for minimum packet length of the used l2.
> > For arcnet the hardware header length is not the complete archdr which
> > includes hard + soft header. This patch changes the length to
> > sizeof(arc_hardware).
> > 
> > Signed-off-by: Michael Grzeschik <m.grzeschik@pengutronix.de>
> 
> The hard header len is used for other purposes as well, are you sure
> those don't get broken by this change?

Its meaning is to represent the amount of the hardware (link layer)
data of one packet.

Which other purposes do you mean?
Can you point to some code?

> Code assumes that if the data at the SKB mac pointer is taken, for
> dev->hard_header_len bytes, that is exactly the link layer header.
> And that this can be used to compare two MAC headers, copy the
> MAC header from one packet to another, etc.

The link layer size of arcnet is 4 bytes long. 1 byte source, 1 byte
dest and two offset bytes. As described by struct arc_hardware in
if_arcnet.h . The above condition is fulfilled when the mac pointer
is 0.

The following pending bytes of struct archdr have a variable meaning
depending of the used protocol and are represented by an union.
(network layer)

In the case of raw packets, the payload comes immediately after the
hard_header.

Thanks,
Michael

-- 
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |