I've been meddling with this with Jason and Paul... and I have
a patched clock.c file, but I'm pretty sure that I'm still getting the
effect from somewhere, so I'm throwing this out to the whole list.

	To put it bluntly, due to some quirk of the OS, the clock
interupts stop.

	What we've done so far is to realize that there was a race
condition between the zs (serial) interupts and the clock.  This was
patched by putting splhigh() around some of the clock interupt code.

	Now I have updated to current, and applied that patch to my
clock.c file.  However, I still have the same problem... the clock
stops --- but it stops far less often.

	To be honest, I hit my machine pretty hard.  It has a
reasonable portion of a full news feed.  It also feeds significant
feeds to four other machines.  This morning, the loadav was 5 when I
came in to find (at 8:30) that the clock had stopped at 5:17.

	I have also observed that the clock can stop when I'm running
sup --- usually only once or twice for a significant sup.  I figure
that I'm hitting the serial and disk subsystems pretty hard in both
cases.

	Has anyone else got any ideas where to poke the kernel?  I
have a small LKM that will clear the problem.  In normal cases, you
might run such a hack every  ten minutes or so.... but that won't work
in this case.  I can make this LKM available to people if they're
experiencing the problem, too.

Dave.

-- 
----------------------------------------------------------------------------
|David Gilbert, PCI, Richmond Hill, Ontario.  | Two things can only be     |
|Mail:      dgilbert@jaywon.pci.on.ca         |  equal if and only if they |
|http://www.pci.on.ca/~dgilbert               |   are precisely opposite.  |
---------------------------------------------------------GLO----------------