Troubles with a 3C905B-TX at 10Mb/s too!

[Jason M. Felice]

On Wed, Oct 13, 1999 at 12:11:50PM -0400, Donald Becker wrote:
> On Wed, 13 Oct 1999, Jason M. Felice wrote:
> 
> > This is same problem I'm having with two 3c905C cards in two different
> > machines, one a Dell and the other an IBM.  Both are fresh out of the box.
> > 
> > Looking through the archives of this list, I see someone else had some
> > similar problems.
> 
> Looking at the list archives isn't a good indicator.  Most people will post
> problems to the list, but never acknowledge later that the problem was
> really with their configuration.  It's especially true when they loudly
> proclaim that the driver is horribly broken.
> 
> I believe that the current 3c59x driver, v0.99L, works correctly with the
> following errata:
>   Incorrect EEPROM checksum calculation with some 3c905C cards.
>      (I don't know if 3Com changed the format again, or if these are
>      misprogrammed OEM cards.)

I don't believe this is the case with this card.  I fiddled with vortex-diag
to get some diagnostics, and I don't recall it failing the 3c905C checksum.
Of course, I had to pull the to 3c905C-TX-M cards and replace them with
3c905B-TX cards because that machine had to ship about two hours before my
previous post.  I've stolen one of the cards from my employer, will put it
on my home network, and get diag info as well as seeing what's happening
from the other end with tcpdump.  As it seems to be a receive issue, I doubt
that will provide any extra info, but ...

>   A work-around for a transceiver bug in some 3c905B cards that causes
>   occasional Rx CRC errors has not been implemented.
>      (I didn't know about this bug until the 3Com driver source was released.)

At least my problem is certainly not 'occasional'.  Where's the driver source?
I'd like to figure enough about the bug out so that I can determine if this is
related.  It doesn't seem to be in the standard 3Com driver section .. I hope
it's open to the public (this is all news to me).  If the problem occurs on
my chop-shop box at home, maybe I'll disable the checksum checking and see
what happens.

>   A rare transmit timeout problem with the 3c595 board.

Not the case.

> 
> All other confirmed problem reports have been caused by
>    Using an old version of the driver (usually v0.99E with newer 3c905B cards)

Not the case.  v0.99L is the only version which will see the card.

>    Forcing the media type to an invalid transceiver type.
>       (This will frequently *almost* work.)

Not the case.  Do the tranceiver types marked as 3c905B in the docs also
apply to the 3c905C-TX-M cards?  This I did not check - I didn't specify
a tranceiver.

>    Forcing full duplex, thinking it's a magic go-faster setting.

Not the case.

> 
> Finally, I have no interest in providing free consulting for setting up
> firewalls, configuring routing, or describing how to port a drivers to
> proprietary OSes.  Do not claim that there is a driver bug to trick me into
> doing these things.

Not the case.

> 
> Donald Becker
> Scyld Computing Corporation, and
> USRA-CESDIS, becker@cesdis.gsfc.nasa.gov

-Jay 'Eraserhead' Felice