Subject: More on Alteon Gigabit and clocks on Miatas
To: None <>
From: Hal Murray <>
List: port-alpha
Date: 05/30/2000 02:25:26
I left some tests running overnight, bashing on the Alteon cards.  
No clock troubles.

I think the thing that I did to fix it was commenting out a can't 
get cluster printf.  (I assume it was getting called from interrupt 

My big test script then shifted over to running on FDDI cards.  The 
clocks are getting glitched again.  The log file shows things like: 

  May 29 22:53:27 mckinley /netbsd:  can't alloc cluster
  May 29 22:53:27 mckinley /netbsd: fpa0: can't alloc cluster
  May 29 22:53:27 mckinley last message repeated 312 times

I tried commenting out that printf but that didn't fix things.

Note that there is positive feedback if this happens at interrupt 
time.  The buffers are all sitting on the input queue waiting be 
be processed which can't happen if the interrupt routine is working 
or thrashing in a printf. 

Mumble. I'll put this on the back burner and work on updating to 

There is another can of worms in this area.  When the Alteon driver 
can't get a buffer, it bumps the input errors counter.  It would 
be nice if there were separate counters for software/configuration 
errors (like no buffer) and link errors (like CRC).