Synthetic Events and Tracing Latency Within the Kernel
Talkat Open Source Summit Japan + AI_dev: Open Source GenAI & ML Summit + Automotive Linux Summit 2025 · 2025-12-08..12-10·
summary
Describes how synthetic events join two kernel tracepoints into one record that carries both fields and a time delta, created by histogram triggers under tracefs; shows real examples: filtering kmem cache allocs to >500 bytes and finding a 6592‑byte request seen nine times, isolating CPU frequency onchange events, and building a wakeup latency synthetic event that reports 11, 14 and 7 microsecond wake‑to‑run times (and a 2.7 ms blocked case), all reproducible and driven via trace-cmd sqlhist.
AI-written summary · about
chapters (9)
- 0:00Synthetic events in kernel
- 5:25Event triggers and filters
- 11:50Creating synthetic events
- 18:14Wake-up latency measurement
- 23:59Synthetic events API challenges
- 25:45SQL histogram commands
- 33:10Histogram features
- 37:48Snapshot latency analysis
- 39:56Q&A
related talks
49mAdding Inter-event Capabilities to Linux Kernel Trace Events
46mMeasuring and Summarizing Latencies using the Trace Event Subsystem
55mAdvanced Tracing Features to See What Your Application is Waiting On
48mCyclictest latency, find out why with tracing
42mHow Much Is Tracing? Measuring the Overhead Caused by the Tracing Infrastructure
39mTracing VMs Seeing How Host and Guest Interact
35mTracing is More Than Traces: The Insights in Trace Aggregates
38mLTTng: Beyond Ring-Buffer Based Tracing