On Sat, May 28, 2005 at 09:22:51PM -0400, Miles Nordin wrote:
> I tried with 'options ALTQ_NOPCC', and it didn't help.  but...looking
> at sys/altq/altq_subr.c I see that's silly because the option does
> nothing on sparc64---it is for i386 and alpha only.  On sparc64 it
> uses microtime().  not sure this has anything to do with the problem
> whatsoever though.

Hm, on NetBSD it should probably make use of the cpu_counter framework if
__HAVE_CPU_COUNTER is defined - machine/cpu_counter.h will define
cpu_hascounter() to test for presence of such counters and cpu_counter() or
cpu_counter32() will return current values.

If time passes somehow odd (and this is causing your problem), wouldn't the
ALTQ counters reflect the drops?

Can you test a 3.0 beta or -current?

Martin