Subject: Re: i386 interrupt counters
To: Jonathan Stone <email@example.com>
From: Chris G Demetriou <Chris_G_Demetriou@ux2.sp.cs.cmu.edu>
Date: 08/12/1996 14:55:14
> >And, having SOMETHING is far better than nothing.
> Indeed. It suggests this machine may be losing clock ticks. Or
> alternatively, truncation of a true clock rate rate somewhere between
> 99 and 100 is showing up, somewhat misleadingly, as 99.
More likely: adjtime was used to set the time forward.
the boottime variable is adjusted to compensate for time _sets_, but
not for time adjustments.
I've found the same effect on the Alpha, when timed causes the time to
be set forward by a small enough amount that adjtime is used.
There really should be 'monoboottime', or something, and vmstat -i
should use monotime and monoboottime. The only problem with that
is that it requires several syscalls per time read...