2012-05-02 20:24:08

by Chris Friesen

[permalink] [raw]
Subject: why would irq0 have many many more interrupts than LOC?


I have an x86_64 2.6.27 kernel running on a Sandy Bridge machine.

/proc/interrupts currently shows 75573114 counts for irq0 on cpu 0 and
no counts on other cpus and between 2973587 and 4185923 counts for "LOC"
spread across all cpus including cpu0.

The following are enabled.

CONFIG_GENERIC_HARDIRQS=y
CONFIG_PREEMPT_SOFTIRQS=y
CONFIG_PREEMPT_HARDIRQS=y
CONFIG_HIGH_RES_TIMERS=y
CONFIG_HZ_1000=y
CONFIG_HZ=1000


It almost seems like it's taking a timer interrupt on cpu0 for every
local timer interrupt on every cpu. Surely that's not design intent?

Any ideas on how to fix this?

Chris



--
Chris Friesen
Software Developer
GENBAND
[email protected]
http://www.genband.com