Linux per cpu idle injection
summary
Reuses and extends the existing powercap idle inject framework so the power clamp driver and per-CPU idle injector can coexist, adding callbacks for compensation and user-space policies that target sub-second responsiveness. Experiments run with 10–24 ms injections and aim for scale in the hundreds of milliseconds, revealing kernel warnings when stopping the tick and noticeable timer jitter. Fixes include a softirq-pending check to exit do_idle_loop quickly and a bounded idle loop to avoid IRQ-flood storms; overall approach works but needs kernel-cleanup and ML review.
AI-written summary · about
chapters (7)
- 0:00Introduction and motivation
- 4:00Use cases for idle injection
- 7:24Existing idle injection in Linux
- 10:12Challenges and fixes
- 13:05Timer and jitter issues
- 18:43Remaining open cases
- 20:08Q&A and conclusion
related talks
32mCPU Power Saving Methods for Real-time Workloads
19mDo nothing fast: How to scale idle cpus?
56mPMWG: CPU Idle governor and irq prediction
49mDebugging Methodologies for Realtime Issues in Linux Systems
52mEffectively Measure and Reduce Kernel Latencies for Real-time Constraints
34mA Journey into Linux Interrupt Handling Bottom Halves
32mImprove CPU Utilization by Designing a New Schedule Algorithm
51mTuning RT Kernel to Improve Schedule Performance with Intel Platform