Messages in this thread | | | Subject | Re: [PATCH 2/7] sched: Add NEED_RESCHED to the preempt_count | From | Mike Galbraith <> | Date | Thu, 12 Sep 2013 08:01:41 +0200 |
| |
On Wed, 2013-09-11 at 15:34 +0200, Mike Galbraith wrote: > On Wed, 2013-09-11 at 13:06 +0200, Peter Zijlstra wrote:
> > Mike does it fix the funny you saw? > > Yup, modulo test_need_resched() not existing in master.
...
> while I haven't yet booted Q6600 box, core2 > Toshiba Satellite lappy is using mwait_idle_with_hints() in master.
Heh. Core2 lappy uses mwait_idle_with_hints().. unless you boot processor.max_cstate=1 so it doesn't use piece of shite hpet. Q6600 always has idle woes, seemingly because box doesn't do deep enough cstate to make something (acpi?) happy, despite mwait_idle() having worked fine for years.
Seems I need to find the cstate dependency spot, and rap it upside the head if I want my core2 boxen to regain happy campers status.
-Mike
| |