[realtek] Realtek 8139 problem with bridging

[Donald Becker]

On Tue, 12 Nov 2002, Orlando Andico wrote:
> http://www.math.leidenuniv.nl/pipermail/bridge/2001-February/000125.html
> 
> In this link, lennert gives a pretty good description as to why Realtek 
> 8139 cards work fine in IP mode but break in bridging mode. Is this 
> problem user-fixable? what changes do I have to do in the rtl8139.c 
> driver? is this going to be fixed in the next revision of the driver?

The problem above is that the driver fails to remove the 4 byte CRC from
the frame length reported to the queue layer.

This problem was fixed in mid-2000 with the rtl8139 driver, and was
verified to not exist with the other Scyld drivers soon after:
----------------------------
revision 1.24
date: 2000/07/15 01:24:47;  author: becker;  state: Exp;  lines: +11 -10
rtl8139.c: v1.11 7/14/2000

Corrected the recovery from a TxAborted 16-collision error to discard the
  packet rather than attempt retransmission, reported Paul Campbell.

Corrected the Rx packet length buglet -- it included the CRC.
----------------------------

The problem _does_ exist with various current non-Scyld drivers, such as
the ns83820.c driver (but not the similar Scyld "ns820.c" driver).
This doesn't affect just bridging -- some of our cluster low-level
software relies on the frame length to guess the protocol version, and
has to explicitly handle this driver bug.


-- 
Donald Becker				becker@scyld.com
Scyld Computing Corporation		http://www.scyld.com
410 Severn Ave. Suite 210		Scyld Beowulf cluster system
Annapolis MD 21403			410-990-9993