On Tue, Mar 12, 2002 at 12:07:28PM -0800, Jonathan Stone wrote:
> 
> Current data shows there really *are* NIC errors, which do cause
> packets with valid MAC CRCs but incorrect IP headers, even on a single
> LAN hop.  Whether it's worth detecting these errors before they get to
> the (possily incorrect) endhost is an individual choice.

Many many moons ago we had problems with the zilog SCC transmitting
HDLC frames using DMA.  The SCC has to be setup to transmit a valid
CRC when the DMA underruns.  Usually this is ok, but is the system bus
locks up you send a short frame with a valid CRC!  This was (meant
to be) a resiliant system, when the secondary controller received
the short frame it tended to die :-(  Changing the MAC protocol
to include a length field helped somewhat!

I've also seen a VMEbus ethernet card that failed to driver the
data bus 'hard' enough to transfer repeats of the 3 byte pattern
0xff, 0xff, 0x00.

Both these problems could happen with modern hardware.

	David

-- 
David Laight: david@l8s.co.uk