resctrl filesystem
summary
Resctrl is being refactored so the filesystem and core APIs support non-x86 QoS and monitoring like MPAM on ARM and CBQRI on RISC‑V; patches split locking, move plumbing into the FS, and add a driver layer so platform drivers can register resources. Google runs resctrl at scale and reports oddities: differing topology semantics (L3 vs memory controller), scarce PMU counters (example machine has 112 ARM IDs but few monitors), and unmount/cleanup bugs. Series on the list is near consensus; next steps are driver posts and schema additions to enumerate resource types.
AI-written summary · about
chapters (9)
- 0:00Introduction and background
- 3:12Challenges with multi-architecture support
- 8:57Resource naming and schema design
- 15:19Patch series status and next steps
- 21:44Monitoring and control ID space issues
- 27:04Cleanup and unmount behavior
- 31:07ABI stability and future directions
- 38:57Hardware-specific behavior and documentation
- 42:37Closing remarks and Q&A
related talks
01h 05mMicroconference for resctrl
42mcgroup2 Resource Control strategies w/ resctl-demo
42m"cgroup2 Resource Control strategies w/ resctl-demo"
24mReliability, Availability, and Serviceability (RAS) on ARM64 status
42mBoF: Realtime Linux
41mAn Essential Relationship between Real-time and Resource Partitioning
14mPCI Pwrctrl woes
47mReal-Time Linux: What Is Next?
