PATCH 2.2.15-pre15: tulip fix for 21041 chips

Hal Murray
Sat Mar 25 04:23:22 2000

> It certainly doesn't take 60 minutes for the driver to change media types.
> I meant that it takes that long to do just a transceiver test by hand.

> I know that seems like a long time, and it doesn't *feel* that long when I'm
> doing it, but I've casually timed it. ...

You didn't mention Hofstadter's Law. :)  Everytime I try something 
like that I blunder into some strange glitch that takes another hour 
or more to track down. 

I was actually thinking of checking out autonegotiation.  I was expecting 
your test recipe to involve setting up a machine with the new driver 
and (manually) moving the cable from one hub/switch to the next, 
stopping long enough to verify that it worked correctly.  It takes 
more than just ping to catch the full/half mixup.

Is 4 tests enough?  I'm thinking of connecting to a box that gets 
to each of 100 full, 100 half, 10 full, and 10 half.  Or are there 
interesting cases that need extra checks, maybe old gear that doesnt 
support autonegotiation?  (or new/fancy gear with some features disabled?)

It shouldn't take an hour, but I wouldn't be surprised if it did.

I'll be glad to help test autonegotation.  I can test both new and 
old DE500s and the Tulips on the motherboard in XP1000s and 600aus. 
I don't think I have any gear that runs at 10-full but I can cover 
the other 3 cases.  I can setup both 2.2.x and 2.3.x kernels. 

With the driver that came with 2.3.51, a pair of DE500s connected 
back to back with a crossover cable end up in 100-half rather than 
100-full.  It often takes a while to get synced up but I don't have 
hard data on that.  (I poked around some and got in over my head.) 
I'll get the tulip-diag info or turn on debugging if you want to 
look into this.  I can easily switch to 2.2.14 with whatever drivers 
are most convenient for you. 

Another thing on my wish list is a log entry on link state change.  
I think the hard part would be some hack to limit the clutter if 
things started flapping.
To unsubscribe send a message body containing "unsubscribe"