[tulip] Tulip media types 5 and 6 ? (Allied Telesyn AT-2800TX Cardbus)

[Donald Becker]

On Wed, 3 Jan 2001, Matti Aarnio wrote:
> On Wed, Jan 03, 2001 at 11:39:04AM -0500, Donald Becker wrote:
> > >   I have strange media types 5 and 6 at an Allied Telesyn AT-2800TX
> > 
> > These are new SROM entry types.
> > They are pretty obscure, and I didn't expect to see them.
> 
>    Famous last words, or that effect...

I was being nice.  Mostly I was thinking "what kind of low-grade crack were
they smoking to come up with _that_ pile..."

> > My driver uses block type 5 only for SYM transceivers.
> 
>    I haven't understood the difference of MII and SYM tranceivers,
>    specifically, can both types of tranceivers exist at some card ?

In theory, yes, SROM table could list both types.  But electrically it's
pretty much impossible.   The MII interface uses 4 data lines in each
direction, while SYM interfaces use 5 "4B/5B" scrambled data lines.

Also, due to a design quirk, the whole chip has to be reset to switch
between SYM and MII transceivers.  Every configuration bit is cleared except
for the MII/SYM selection bit!

> > Type 6 appeared to be a hack to work around a bug in a specific piece of
...
>    So that interface seems to work with MII in both modes.
>    What could be the point with those other media specifications
>    at this card ?

I don't know for certain.
This card couldn't use the standard example media table from Intel.  It
might be that an inexperienced designer had to write a new table, and was
just covering all of the bases.  Or perhaps the Digital/Intel reference
driver didn't handle MII resets correctly, and thus they added this reset as
a work-around.

An all-too-common bug is using the example SYM media table for a unique
design, and then hacking the reference driver to support only that card.
At least these guys tried to work with a generic driver.

Given that this has a type 6 table entry, and this is a CardBus card, I'm
wondering if the transceiver is in the dongle, and a dongle disconnect
requires the reset sequence.  The type 6 entry might be a sleazy way of
indicating this, even though it violates the current SROM specification.

> > I'm guessing that you added this case to the tulip-diag program.
> 
>    Now that you mention it, yes, very possibly.
>    Downloading your baseline version....

I'll send you a copy of the beta tulip-diag, which adds 21145, Comet and
Conexant support.  Please send a report.


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