Hi folks...

The `tlp' driver now supports the Macronix clone chips, i.e. the MX98713,
MX98713A, MX98715, and MX98725.  I have only been able to test the MX98715:

tlp2 at pci0 dev 11 function 0: Macronix MX98715 Ethernet, pass 2.0
tlp2: interrupting at kn20aa irq 0
tlp2: Ethernet address 00:80:c6:f7:9d:06
tlp2: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, auto

This is a SOHOWare 10/100 board.

I haven't tested the autosense support yet, but will try to do so tomorrow.

I have also added support for the ADMtek AL981, but have not yet been able
to test it (lack of sample boards), and haven't written a specific driver
for its built-in PHY (ukphy should work for now).

I have also fixed a few bugs:

	- Set `heartbeat disable' as appropriate (i.e. in anything except
	  10Mb/s half-duplex mode).  This could cause spurious CRC and
	  dribbling bit errors due to SQE pulses being transmitted when
	  they shouldn't be.

	- Fixed transmit stats gathering (helps to look at the correct
	  descriptor).

	- Hopefully fix a problem with non-ISV 21140 boards which use
	  MII-over-SIO (i.e. Asante) ... though I can't test this
	  directly.  Please report back on this!

There is one nasty bug remaining:  On the clone chips, transmissions
in 100Mb/s mode are pathetically slow.  1/10th the speed for 10Mb/s mode.
The genuine DECchip 21140 does NOT have this problem.  I have a theory,
which I hope to examine more closely tomorrow.

I'm going to work on the Lite-On PNIC II tonight while I'm on the train
(yes, the majority of this driver has been written during my daily CalTrain
commute between San Francisco and Mountain View).  This should be pretty
straight forward, as it's basically a MX98715 clone (get that!  A clone of
a clone!) with a 128-bit multicast hash table (rather than the standard
512-bit).  This chip is on the new LinkSys LNE100TX Version 2.0 boards
(blue box).

Have fun!

        -- Jason R. Thorpe <thorpej@nas.nasa.gov>