Evolving genpd’s - sync_state() support
summary
Generic power domain sync_state support lands in v6.17-rc1 and provides a unified consumer-probe notification, but it leaves several concrete failures: domains remain powered when unused because devlink defaults to strict mode (waiting forever for missing consumers), and nodes that supply multiple domains cause multi-provider genpd all-or-nothing behavior where one missing consumer keeps every subdomain on. Proposed fixes include per-domain flags or timeouts, a provider-side list of expected consumer firmware nodes, and extending devlink to map sync_state to subresources rather than per-device.
AI-written summary · about
chapters (7)
- 0:00Introduction to sync_state()
- 1:50Background and sync_state() in genpd
- 4:51Unused power domains staying on
- 8:25Behavior change and regressions
- 11:58Timeout vs strict mode debate
- 13:23Multi-provider power domains issue
- 18:29Q&A and wrap-up
related talks
25mThe Power-Supply Subsystem
14mfw_devlink and parallelization updates
41mLinux Power Management Features & Their Interactions, Part 2
41mThe Case for an SoC Power Management Driver
43mLinux Power! (from the Perspective of a PMIC Vendor)
30mPCIe Endpoint Subsystem Open Items Discussion
27mRefactor Power Library for Vendor Agnostic Uncore APIs
41mThe Case for an SoC Power Management Driver
