Red Hat Bugzilla – Bug 463873
[LTC 6.0 FEAT] 201196:Idle CPU power management
Last modified: 2013-01-10 02:57:17 EST
Emily J. Ratliff <email@example.com> - 2008-09-24 13:54 EDT
1. Feature Overview:
Feature Id: 
a. Name of Feature: Idle CPU power management
b. Feature Description
Arbitrary kernel timer events, user space polling and interrupts wake otherwise idle CPUs,
shortening the CPU idle time. By instead moving to more asynchronous notifications for various
applications and device subsystems instead of using regular timer based wake-ups/polling, the time
spent by a CPU in lower C states during idle can be increased. Similarly scheduling of tasks with
active timers to active CPUs allows idle CPUs to remain idle and in low power state.
2. Feature Details:
Arch Specificity: Both
Affects Core Kernel: Yes
Delivery Mechanism: Direct from community
Request Type: Kernel - Enhancement from Upstream
d. Upstream Acceptance: In Progress
Sponsor Priority 1
f. Severity: High
IBM Confidential: no
Code Contribution: IBM code
g. Component Version Target: unknown
Performance Assistance: yes
3. Business Case
Increasingly the enterprise market is interested in mitigating and managing power consumption in
servers, where interest previously was mostly in a mobile market. Driving optimizations in the
power consumption area allows an enterprise Linux distribution advantages over other operating
systems in this emerging market.
4. Primary contact at Red Hat:
5. Primary contacts at Partner:
Project Management Contact:
Sarah Wright, firstname.lastname@example.org, 503-578-5145
Dipankar Sarma, email@example.com
Gautham R Shenoy, firstname.lastname@example.org
Jeffrey Heroux, email@example.com
We've done some work on this already in userspace in Fedora (using the powertop program to point out offenders.)
While we have no problem with the feature, there's nothing specific listed here that is actionable other than 'test and see what is causing wakeups and polling'.
Are there specific instances you know of that you'd like fixed?
This feature is to do with power-savings through task-consolidation, which is a scheduler feature offered through the interfaces sched_mc_power_savings and sched_smt_power_savings.
There has been enhancements to the existing interface, which was earlier boolean, but now has been made multi-valued to represent the aggressiveness of consolidation to obtain power savings. This feature is accepted in linux-2.6.29.
Further enhancements to the accepted code have been discussed in the community (Ref: http://lkml.org/lkml/2009/2/16/221)
In addition, there are other scheduler enhancements which are under discussion such as
- the timer-migration framework discussed (Ref: http://lkml.org/lkml/2009/2/20/121)
- Idle Load Balancer enhancements (Ref:http://lkml.org/lkml/2008/9/23/82 )
OK, marking as MODIFIED.
The feature requested has already been accepted into the upstream code base
planned for the next major release of Red Hat Enterprise Linux.
When the next milestone release of Red Hat Enterprise Linux 6 is available,
please verify that the feature requested is present and functioning as
Since this is in RH cvs tree, moving to Accepted state here
------- Comment From firstname.lastname@example.org 2009-11-18 08:53 EDT-------
Testing that this message will get over to the RH bug
------- Comment From email@example.com 2010-06-07 06:59 EDT-------
Marking this as verified as the dependant bugs https://bugzilla.linux.ibm.com/show_bug.cgi?id=52943 , https://bugzilla.linux.ibm.com/show_bug.cgi?id=52944 ,
https://bugzilla.linux.ibm.com/show_bug.cgi?id=52945 have been verified on RHEL 6 snapshot 1
code review... patches 1,4 and 5 from comment 2 are applied in 2.6.32-15.el6 source. No code from patches 2 and 3 found in the kernel source. Neither the surrounding code found, so I guess it was deleted, moved somewhere else.
Red Hat Enterprise Linux 6.0 is now available and should resolve
the problem described in this bug report. This report is therefore being closed
with a resolution of CURRENTRELEASE. You may reopen this bug report if the
solution does not work for you.