more 3C900-combo data points

[Hans-Peter Jansen]

Donald Becker wrote:
> 
> Normally the 3c900 advice is to force the media type selection, since
> detecting the media types without link beat (10base2 or AUI) isn't reliable.
> But I suspect that this is a simple routing problem.
> 
> You shouldn't need to update your driver unless you have a 3c900B
> (Cyclone-based).

Well Donald, just want to let you know, that I just received the 3c docs.
When you told me, the 3c90x card family are bit****, I thought, you were 
sarcastic, but now I know, its the absolute truth (when the is one...)

BTW: current 3c59x.c v0.99E works fine with the 3c905B TX type. Thanks!

To all netbooters:

Just tried to get them net booting. actual etherboot fails for me (may be 
don't like my dhcp-server (V1-0-0) handling bootp). 

Actual netboot with new 3c90x pktdrv don't works either. The netboot 
kernel seems to be unhappy with the new packet driver (just freeze).
Only a patched 0.73 works for me currently (from floppy). Haven't the 
time to figure out, what's broken. Will flush stuff on bootroms tomorrow.   

Does anybody know, why netboot run the 3c90x pktdrv with -r (to switch
off the early receive feature). Does this affect 3c59x driver initialization?

I had some strange effects when pushing reset during boot, and starting
linux again, but they could be 3c59x.c v 0.49? related (can't remember
them after driver update). 

The B type seems to be a bit faster. Does anybody tried to use the 
hardware check summing with recent cards on linux, or is a decent CPU
IP cksumming faster? (hamachi, 3c90*B, ...)

Hans-Peter (who always liked the idea of intelligent subsystems)

> 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