Timekeeping hosed by factor 3, high lapic[01] interrupt rates
Doug White
dwhite at gumbysoft.com
Wed May 18 15:04:40 PDT 2005
On Mon, 16 May 2005, Jens Schweikhardt wrote:
> hello, world\n
>
> the timekeeping on my CURRENT system as of May 15 is very strange. The
> time as reported by date(1) increases too slow by a factor of 3. Things
> with second intervals like sleep 1, tail -f, iostat 1, changing folders
> in mutt all go slower by a factor of 3.
>
> The system bios date however is correct, and a kernel as of March does
> not have this problem, so it's clearly the kernel or some other software
> problem. I have investigated a bit and found the interrupt rates for the
> lapic[01] being three times the value on the broken kernel (about 2kHz
> vs 6kHz),
Are you running with kern.hz or HZ set to something other than the
default?
> schweikh at hal9000:~ $ vmstat -i # Good kernel
> interrupt total rate
> irq1: atkbd0 250 0
> irq3: sio1 2 0
> irq4: sio0 2 0
> irq12: psm0 9079 25
> irq13: npx0 1 0
> irq14: ata0 84 0
> irq15: ata1 77 0
> irq18: em0 1091 3
> irq24: ahd0 9112 25
> irq25: ahc0 16 0
> lapic0: timer 722947 1997 <-- ~6057 on a bad system
> irq0: clk 361525 998
> lapic1: timer 708251 1956 <-- dito
> Total 1812437 5006 <-- greater than 12000
the lapic timer values should run about 2*hz.
--
Doug White | FreeBSD: The Power to Serve
dwhite at gumbysoft.com | www.FreeBSD.org
More information about the freebsd-current
mailing list