PCI Endpoint Subsystem Open Items Discussion
summary
Three concrete open items: adding VIRTIO-based EPF drivers, device integration so EPF binds to EPC via devicetree, and replacing the custom outbound-window allocator. Three VIRTIO proposals exist; the earliest exposes two virtio-net devices and is messy, the vhost-based one never merged, and proposal three (vringh helpers) is the favored, simpler split library approach that manages vrings but currently records <1 Gbit/s measured throughput and relies on polling for host notifications. Device-tree work proposes an devicetree EPF node with 64-bit BAR array and interrupt properties for MHI-like functions, and the allocator change is to migrate allocator to genalloc/genpool.
AI-written summary · about
chapters (13)
- 0:00Introduction and agenda
- 0:43Overview of Vero standard
- 3:31PCI endpoint subsystem basics
- 6:40Vero in endpoint subsystem proposal
- 8:04First Vero proposal by Haian
- 11:18Second proposal by Kishon
- 14:50Third proposal by Shuk
- 17:42Comparison with vDPA and performance
- 20:31OpenAMP and broader use cases
- 22:45Symmetrical vs asymmetric protocols
- 23:49Vero EPF standardization
- 25:55Device integration with MHI
- 28:03Conclusion and Q&A
related talks
30mPCIe Endpoint Subsystem Open Items Discussion
41mVirtIo-Net PCIe Function Using Linux PCI Endpoint Framework- 41mVirtIo-Net PCIe Function Using Linux PCI Endpoint Framework
30mVirtio for PCI Endpoint Subsystem in Linux Kernel
21mVirtio hardware accelerators and other news
24mvDPA Support in Linux Kernel
31mHardware Friendly Vhost vDPA: Towards an Efficient and Migratable Device Model
28mVirtual Versus Physical: Virtio Changes for New Hardware