[RFC] Collecting options to debug "kernel pause" issues

Stefan Bader stefan.bader at canonical.com
Mon Mar 2 19:09:00 UTC 2009


Starting with Intrepid we saw several reports of machines that had boot/install
issues that manifested themselves by complete stops of all activity until any
key (even shift) was pressed. [1][2][3]

There have been lots of submissions to those bugs but somehow it still looks 
like pieces are missing. Probably the best approach would to add a call for 
information and let everyone gather a certain portion of information.
Loosely I was thinking of the matrix below, but maybe there are more ideas to 
narrow down a certain source of the problem. Are there any other 
options/combinations I should ask for? Some mre data that should be gathered?

Required one time:
cat /proc/version_signature (if exists)
dmidecode
lspci -vvnn

Required per run:
dmseg
cat /proc/interrupts
cat /proc/timer_list

Boot option to try:

Timers/clockevents
Use jiffies as clocksource	clocksource=jiffies
Disable hpet:			hpet=disable
Do not be tickless		nohz=off (with highres=disable or separate?)

Interrupts:
Prevent use of iopaic:		noapic lapic
Do not use acpi irq routing:	acpi=noirq
Not use acpi for pci setup:	pci=noacpi(,routeirq?)

Power states:
Only got to C1:			idle=halt
Do not use mwait:		idle=nomwait

[1] https://bugs.launchpad.net/ubuntu/+source/linux/+bug/217849
[2] https://bugs.launchpad.net/bugs/254668
[3] https://bugs.launchpad.net/bugs/272247
-- 

When all other means of communication fail, try words!







More information about the kernel-team mailing list