RISC-V IOMMU
summary
Specification ratified this year that defines a system IOMMU for RISC‑V implementing two-stage address translation with per‑transaction device and process directories (device ID is 16 bits), an address translation cache (ATC), asynchronous command and fault/event queues, PCI ATS/PRI support and a MSI page table for direct guest interrupt delivery. It enables secure device passthrough, shared virtual addressing for pointer sharing with accelerators, page request/response flow, up to 64 hardware interrupt files, 31 programmable event counters, and multiple integration models.
AI-written summary · about
chapters (13)
- 0:00Introduction to RISC-V IOMMU
- 1:24Key IOMMU capabilities
- 3:11IOMMU use cases explained
- 6:41Two-stage address translation
- 8:27Address translation caching
- 10:13Handling faults and page requests
- 11:37Recoverable faults with ATS
- 14:03Direct interrupt delivery
- 16:54Memory resident interrupt files
- 17:56Hardware performance monitoring
- 18:37Endianness and ISA support
- 20:02IOMMU system integration
- 22:07Summary and capabilities
related talks
36mShared Virtual Addressing for High Performance Arm Infrastructure Platforms
38mPrivileged ISA
12mA Peek Inside a New RISC-V CPU for Autonomous Vehicles
15mRISC V Memory Consistency Model Task Group Update
19mRISC V ISA & Foundation Overview
31mFast Interrupts for RISC-V
15mRISC-V processor IP product line
32mDeveloping the RISC-V Hypervisor Extensions in QEMU
