Update on Sony Vaio w/DS21143

[Donald Becker]

On Thu, 8 Apr 1999 mbreuer@siac.com wrote:

> Also... I've done a little research...(please excuse my ignorance...
> drivers are new to me). 

> 1) Under Win98 & NT the 21143 works correctly using the generic Microsoft
> drivers.  I've examined the documentation available from Intel & DEC and
> I'm wondering about whether using the autonegotiate settings for 10baseT
> defined on page D-2 of the Hardware reference manual are actually used.
> According to the book, the correct sequence to set autonegotiate is:

>      CSR6<18>=0
>      CSR13=0000H
>      CSR14=3FFFFH
>      CSR13=0001H

The databook is confusing/wrong on this point.
This is for turning on the internal autonegotiation using a SYM (symbol
mode, non-MII) transceiver with a special design feature that lets it work
with the 21143.  It doesn't take into account the special settings from the
EEPROM/SROM media table that may be needed for certain designs.

> To be honest, I haven't had the time to fully analyze the manual or the code, but I don't see where this sequence is ever executed
> for a 21143 w/ mii and 21142_timer.

It's never executed with an MII transceiver.
In 0.90z its in t21142_start_nway().
In previous versions it was spread over several code sections, including
select_media().

> 2) After analyzing the SROM (using the 21x4 serial rom format v4.09 as
> reference), and after considering warnings regarding loss of power to the
> docking station (from Sony) I believe that this device does not, in fact,
> have an MII port.

That's likely true.  All MII transceivers are 10/100, and they typically can
be reliably detected without special general-purpose-I/O pin settings.

> It *must* have a 10BaseT half-duplex (device 0) and it *should* have a
> 10BaseT full-duplex (type 2) available.  If so, can this be forced in any
> way, shape or form.  I've tried insmod tulip options=0 (and options=2) to
> no avail.  Theoretically, I *should* be able to override the SROM and tell
> the driver via options (or by (re)defining appropriate table entries) and
> make this card work.  So far, I really don't have a clue how to proceed.

Try options=12, which forces 10baseT only.
You could also try forcing 10baseT-FDX.  'ping' will still work with a
duplex mismatch, although you'll get Tx errors with more intensive use.

> > The driver will then use generic values instead of the media table.  Some
> > 21143 designs will work with the generic values, but most power-saving
> > designs require the correct media table entries to switch on the transceiver
> > power.


Donald Becker					  becker@cesdis.gsfc.nasa.gov
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