Subject: Re: Clock drift on Miatas
To: None <port-alpha@netbsd.org>
From: Hal Murray <murray@pa.dec.com>
List: port-alpha
Date: 05/29/2000 01:45:55
[From Jason R Thorpe]
> NetBSD doesn't use this, I'm pretty sure.  It just knows that the clock
> interrupts 1024 times a second and goes from there.

Thanks.  That means the system has to see all the timer interrupts.  
(or the clock slips, like I'm seeing)

The -current version of kern_clock.c has the timer interrupt processing 
split between hardclock and softclock. 

Can somebody give me a quick lesson in interrupts?  Or tell me where 
to look...

The only way that I can see for the hard/soft split to make sense 
requires that the "hard" part of the timer interrupt be able to interrupt 
network interrupt code and anything else that might run for a long 
time.  (The alternative is to make sure that no interrupt routine 
runs "too long", and I haven't seen anything like that.  But maybe 
I wasn't looking in the right place.)

Is there a list of what devices/code run at what level?