On Tue, 6 Feb 2001, Mark Abene wrote:

> On Mon, Feb 05, 2001 at 09:08:47AM -0800, Jason R Thorpe wrote:
> > 
> > The tsleep has a timeout on it, so if your clock interrupt is working
> > properly, then the process will be awakened after 2 seconds.
> > 
> > I suspect it actually *isn't* working properly.
> > 
> 
> If I stick a printf("*") in my clkintr handler, I see a constant stream of
> "*"'s at a fixed interval, which tells me it seems to be working.
> I'd be more than happy to send you my patches if you'll try them out on
> your Magnum.

Hi Mark,

By the sounds of it the hardintr() function isn't being called, or the
microtime() function is doing something weird.

Taking a look at the code I see that the hardintr() function is called
only if the clock_started variable has been set.  This is only specific
for the MAGNUM and is set in cpu_initclocks().  I assume that you have
checked this variable has been set correctly.

You might be able to get a better handle on what is happening by adding
the following debug printf to the end of the microtime() function

   printf("tv_sec=%ld, tv_usec=%ld\n", tvp->tv_sec, tvp->tv_usec);

If you could get into DDB and display the callout table - there should
only be a few entries and one will be endtsleep should be the event you
are after.  DDB should work at that point on the system - I just tried it
on one of my NetBSD/mipsco machines - You say it doesn't work on yours
which might indicate a more serious problem.

It might be possible that sleep/wakeup and the clock interrupts are all
working and the system goes into an infinite loop when it is actually
setting up the SCSI probes.

One problem I experienced in the mipsco port was interrupt's not being
serviced correctly, and the interrupt handler would never return.  Placing
a debug printf at the start of the interrupt handler showed me what was
going on (and filled the screen up real fast!)

Also on big-endian firmware if you hit the reset button and do a
printenv from the console you can get an idea of where it was looping from
the resetepc and resetra variables - that has saved my life many
times.   Not sure if ARC firmware has a similar feature.

Good luck!

Wayne