It looks like dmfe is not part of OpenSolaris, but is only available for
SPARC systems.  Is there a reason for this?  The chip is a commodity
chip that is readily available on PCs, and indeed, it would be nice to
have a common driver for it.

Davicom parts are basically another tulip clone variant, and so I can't
imagine there is any compelling reason to keep the sources for dmfe
closed.  It would be nice to have this available in OpenSolaris.

If push came to shove, I could probably write a dmfe driver trivially by
modifying my own afe/mxfe drivers (Davicom chips are pretty "clean"
tulip clones), and make that public.  It is also supported, IIUC, by
Masayuki's "tu" driver.

Frankly, I am not really fond of the fact that there are multiple
drivers supporting the same chip, which is the _main_ reason I've never
bothered to write my own dmfe driver.  FWIW, I have a driver for PNIC
chips (another tulip clone) as well, that is about 85% working right
now, but I sort of lost interest because you can't really find those
chips on the market anymore.

Here's more or less a brief overview of what drivers for tulip chips exist:

dnet: in Solaris, but only built for x86 by default (why?).  Only
supports actual DEC tulips.  This code is really stale the last time I
looked.

dmfe: in Solaris (closed source), SPARC only.  Only supports Davicom
parts.  Code has gotten some maintenance.

afe/mxfe: My unbundled open source (SPARC/x86).  Fully DDI and
DR/suspend compliant.  Supports ADMtek, Macronix, and PNIC2 chips only. 
(These are the most common tulip variants found today, especially in
cardbus 10/100 cards where the admtek centaur is the main contender.)

tu: Masayuki's rewrite/port from Linux, supports pretty much all tulip
clones, but probably lacks some features (DDI suspend safety, ndd
tunables, etc.)  Because of all the various bugs/quirks in odd tulip
clones, the driver is quite a bit more complicated than it otherwise
might be.

Personally, I prefer my approach of providing the best possible Solaris
support for a few of the most common chips to Masayuki's approach of
providing more basic support for all the variants under the sun, but of
course my opinion is probably not 100% objective. :-)

If Sun wants to incorporate mxfe/afe into OpenSolaris, I'd be happy to
update them to nemo and make them support VLAN and Jumbo frames.  The
hardware can certainly do it.  I don't want to update them _before_ that
point, though, because then I'd have to maintain two different versions,
one for nemo and one for GLDv2.

-- 
Garrett D'Amore, Principal Software Engineer
Tadpole Computer / Computing Technologies Division,
General Dynamics C4 Systems
http://www.tadpolecomputer.com/
Phone: 951 325-2134  Fax: 951 325-2191


Reply via email to