Re: Some alpha problems in current

[Jarle Greipsland <jarle%uninett.no@localhost>]

"Michael L. Hitch" <mhitch%NetBSD.org@localhost> writes:
> On Wed, 6 Feb 2008, Jarle Greipsland wrote:
> >>    The machine survived 16 hours (until 04:55), at which point it paniced
> >> with "fpsave ipi didn't".  CPU 1 was hung and I couldn't get a backtrace
> >> from it.  This would also explain the panic - CPU 0 is waiting for the
> >> other CPU to process the fpusave ipi, but it's probably spinning somewhere
> >> and unable to process the ipi.  This type of panic used to be fairly
> >> frequent before, but I was able to track down the deadlock at the time and
> >> that particular problem was fixed.  It looks like it may be back again.
> > This is all very similar to the problems I reported in
> > port-alpha/37712.
> 
>    I'd say it's the same problem.
> 
>    I know what's happening, but not why or how to fix it.
Probably unrelated, but the timekeeping on my CS20 is also quite
wacky.  Shortly after booting a 4.99.53 -current kernel, the NTP
time synchronization can no longer keep the time in sync with the
NTP servers used.  Also, I am quite puzzled by the following
output:
# while true; do date; sleep 1; done
Thu Feb  7 13:09:54 CET 2008
Thu Feb  7 13:09:55 CET 2008
Thu Feb  7 13:09:55 CET 2008
Thu Feb  7 13:09:56 CET 2008
Thu Feb  7 13:09:56 CET 2008
Thu Feb  7 13:09:57 CET 2008
Thu Feb  7 13:09:57 CET 2008
Thu Feb  7 13:09:58 CET 2008
Thu Feb  7 13:09:58 CET 2008
Thu Feb  7 13:09:59 CET 2008
Thu Feb  7 13:09:59 CET 2008
Thu Feb  7 13:10:00 CET 2008
Thu Feb  7 13:10:00 CET 2008
Thu Feb  7 13:10:01 CET 2008
Thu Feb  7 13:10:01 CET 2008

The wall time seems to advance slower by a factor of 2 compared
to whatever timer the sleep call uses.

This was all with the time settings the kernel arrived on by
itself:
kern.timecounter.choice = PCC(q=1000, f=1666149600 Hz) clockinterrupt(q=0, 
f=1024 Hz) dummy(q=-1000000, f=1000000 Hz)
kern.timecounter.hardware = PCC
kern.timecounter.timestepwarnings = 0

I then switched the timecounter source to clockinterrupt, and
immediately the time keeping started behaving sanely again:
# while true; do date; sleep 1; done
Thu Feb  7 13:10:42 CET 2008
Thu Feb  7 13:10:43 CET 2008
Thu Feb  7 13:10:44 CET 2008
Thu Feb  7 13:10:45 CET 2008

I notice that the kernel thinks that the frequency of the PCC is
1666149600, while I seem to recall that the marketing blurb for
the CS20 claimed that the 21264 CPUs were clocked at 833MHz.  Of
course there may be some clock doubling going on somewhere, but
the factor of 2 made me a bit curious.
                                        -jarle