On Sun, 10 Feb 2002, Havard Eidnes wrote:

> Hi,
>
> I think there must be something wrong with how NetBSD/sgimips counts
> its disk statistics.  If I start a "cvs update" of the source tree, in
> the early phases of the update, the bottom of the "systat vm" screen
> may look like this (sd1 is my NetBSD disk on this system):
>
> Discs  sd0  sd1                                        64 fmin
> seeks                                                  85 ftarg
> xfers                                                     itarg
> Kbyte                                                1053 wired
> %busy      55.9                                           pdfre
>
> Notice that it says 55.9% busy, but no transfers and no data
> transferred (!).
>
> Anyone have an idea where the error might be?

Haven't looked at disk stats, but have noticed some cronic clock problems
in the last few months but haven't looked at them yet.
There are times when the machine pratically "dies" for around 10 seconds
at a time.

Maybe the count/compare interrupt is being missed (or compare
loaded with a value below the count value) and the machine has to wait
2^32 to come back.   microtime() looks to go backwards from time to
time too....   This could skew lots of stats!!

-- 
Wayne Knowles			NetBSD/mipsco port maintainer
wdk@netbsd.org			http://www.netbsd.org