The Case for an SoC Power Management Driver
summary
Argues for an SoC PM driver that registers generic PM domains and attaches them to platform devices so device drivers no longer sequence clocks, resets or regulators. Implements a prototype that splits of_platform_default_populate to insert PM domains, adds one extra compatible string on the SoC node, and centralizes power-on/power-off logic in the SoC driver. Prototype powers devices on before probe, powers unused resources off afterward, preserves existing device tree bindings, and lets drivers keep consumer APIs only when they need fine-grained tuning.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 0:42SoC power management basics
- 3:46Current power management approaches
- 5:35Power management in ACPI
- 8:49Problems with current designs
- 12:19Previous and ongoing attempts
- 16:58Proposed solution with PM domains
- 19:51Barriers to implementation
- 22:25Proposed SoC PM driver architecture
- 23:56Implementation strategy
- 26:20Requirements and conclusion
- 31:36Q&A session
related talks
41mThe Case for an SoC Power Management Driver
29mPMIC: First One to Turn On and Last One to Turn Off
41mLinux Power Management Features & Their Interactions, Part 2
40mPower Management Integrated Circuits: Keep the Power in Your Hands
43mLinux Power! (from the Perspective of a PMIC Vendor)
39mDesigning a Multi-PMIC Linux Driver: Key Decisions for Effective Mainline Submission
57mImplementing ACPI 5 Features Device Enumeration and PM
14mPCI Pwrctrl woes