PCIe Endpoint Subsystem Open Items Discussion
summary
Proposal replaces atomic_notifier with a simple callback event_ops (coordinate, linkup) so EPF callbacks run in the same context as EPC, removing sleeping-in-atomic bugs and notifier overhead; downstream drivers already use four events (coord, linkup, linkdown, BME) successfully. Adds a minimal device-tree child node for MHI so EPC creates EPF devices at boot and avoids configfs, with the caveat that controllers should invoke callbacks from threaded/non-IRQ context to avoid blocking IRQ handlers.
AI-written summary · about
chapters (4)
- 0:00PCI endpoint notifications
- 5:13Addressing latency concerns
- 9:49Device tree integration
- 14:34Framework design Q&A
related talks
29mPCI Endpoint Subsystem Open Items Discussion
14mPCI Pwrctrl woes
29mNon-discoverable devices in PCI devices
50mHotplug of Non-discoverable Hardware: Status and Future Directions
16mElectrode: Accelerating Distributed Protocols with eBPF
40mVisions for the Linux Kernel PWM Subsystem
41mVirtIo-Net PCIe Function Using Linux PCI Endpoint Framework
19mPower Sequencing for Enumerable Busses - What's Still Missing From the Device Tree...