On Mon, 12 Oct 1998, Philip Blundell wrote:
> >alignment traps. When copying packets (or with a better chip) you can do
> >skb_reserve(skb,2) to align the IP header.
>
> I did wonder whether it was possible to use the Tulip's chained-buffers scheme
> to arrange for the 14-byte header to be dumped in one buffer and the data
> itself to land in another. Then the driver would only have to copy the header
> around rather than the entire packet to put them back together.
Alas, no. The buffers must be longword aligned and buffer sizes must be a
multiple of 4. So the byte alignment will never change. I suspect that the
Rx FIFO data doesn't go through the "byte-steering" logic that the Tx FIFO
has.
Note: Some older chips require a specific mimimum amount of data buffer,
usually the minimum frame size, in the first descriptor. That way a
collision never requires the chip to "back up". Modern chips, including the
Tulip, just have an Rx FIFO threshold setting high enough that data isn't
transferred until a normal collision can no longer occur. Out-of-window
collisions are marked as CRC errors.
Donald Becker [EMAIL PROTECTED]
USRA-CESDIS, Center of Excellence in Space Data and Information Sciences.
Code 930.5, Goddard Space Flight Center, Greenbelt, MD. 20771
301-286-0882 http://cesdis.gsfc.nasa.gov/people/becker/whoiam.html
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to [EMAIL PROTECTED]
