Red Hat Bugzilla – Bug 497185
[LTC 6.0 FEAT] Timer migration from idle CPUs [201926]
Last modified: 2010-05-10 13:25:19 EDT
=Comment: #0================================================= Emily J. Ratliff <ratliff@austin.ibm.com> - 1. Feature Overview: Feature Id: [201926] a. Name of Feature: Timer migration from idle CPUs b. Feature Description In an SMP system, tasks are scheduled on different CPUs by the scheduler, interrupts are managed by irqbalancer daemon, but timers are still stuck to the CPUs that they have been initialised. Timers queued by tasks gets re-queued on the CPU where the task gets to run next, but timers from IRQ context like the ones in device drivers are still stuck on the CPU they were initialised. 2. Feature Details: Sponsor: Linux Systems Tech Architectures: x86_64 ppc64 s390x Arch Specificity: Purely Common Code Affects Core Kernel: Yes Delivery Mechanism: Direct from community Category: Kernel Request Type: Kernel - Enhancement from Upstream d. Upstream Acceptance: Submitted Sponsor Priority 1 f. Severity: High IBM Confidential: no Code Contribution: IBM code g. Component Version Target: Patch and discussions: (Version 6) http://lkml.org/lkml/2009/4/16/45 Targetting 2.6.31 3. Business Case This framework will help move all 'movable timers' using a sysctl interface. 4. Primary contact at Red Hat: John Jarvis jjarvis@redhat.com 5. Primary contacts at Partner: Project Management Contact: Stephanie Glass, sglass@us.ibm.com, 512-838-9284 Technical contact(s): Vaidyanathan Srinivasan, svaidyan@in.ibm.com IBM Manager: Jeffrey Heroux, heroux@us.ibm.com
------- Comment From svaidyan@in.ibm.com 2009-05-20 02:08 EDT------- Upstream code in sched-tip development tree.
------- Comment From sglass@us.ibm.com 2009-06-17 16:44 EDT------- This is now in 2.6.31-rc kernel. ------- Comment From sglass@us.ibm.com 2009-06-17 16:45 EDT------- This is now in 2.6.31-rc
------- Comment From arunbharadwaj@in.ibm.com 2010-05-10 13:14 EDT------- Hi, I have verified that this feature is included in RHEL6 snap1 kernel and that this feature is functionally working correct. Here is how I have tested this feature: On an SMP (n >= 4), all CPUs are kept idle. Run a small workload pinned to one CPU. This increases the timer interrupt rate on this CPU, and also another CPU in the same package has high interrupt rate ( equal to HZ rate). This is the ILB CPU. Now start queuing timers on another CPU using a tool called cyclictest. Due to timer migration feature, all timers from this CPU are migrated onto the ILB CPU. This verifies that the feature is functionally working correct. thanks arun