RISC-V IOMMU support

[br]

Summary

Support for RISC-V IOMMU spec v1.0.1 (ratified)

https://github.com/riscv-non-isa/riscv-iommu

Supports translation for PCI devices only.

Supports 1 or 2-level device-directory-table (DDT).

Supports SV39 and SV48 virtual memory system (on per-device basis).

Supports both "standard" and "extended" device-context (DC) structure.

Supports "bypass" mode to disable translation for a particular device.

Supports WSI (Wire-Signalled Interrupts) only.

This includes both PCI-bus and FDT attachment drivers.

Note in case of PCI-bus attachment, interrupts are not available. In this case no error report is provided in case of translation fault. Otherwise interrupts are not needed.

Test Plan

Tested in QEMU only, with virtio-block and virtio-network devices.

Ensure that "hw.busdma.default" kernel boot variable is enabled.

For virtio-network device (to sit behind MMU) apply this https://reviews.freebsd.org/D55492

Virtio-block should work without additional patch.

FDT:

./qemu-system-riscv64 -nographic \
	-machine virt,iommu-sys=on \
	-cpu 'rv64,h=true,sscofpmf=true' \
	-smp 8 \
	-netdev tap,ifname=tap0,script=no,id=net0 \
	-device virtio-blk-pci,iommu_platform=on,disable-legacy=on,drive=disk \
	-drive file=${DRIVE_FILE},id=disk,format=raw,if=none,cache=writeback \
	-device virtio-blk-pci,iommu_platform=on,disable-legacy=on,drive=disk2 \
	-drive file=${DRIVE2},id=disk2,format=raw,if=none,cache=writeback \
	-device virtio-net-pci,iommu_platform=on,disable-legacy=on,netdev=net0 \
	-m 16G \
	-bios /usr/local/share/opensbi/lp64/generic/firmware/fw_jump.elf \
	-kernel ${KERNEL}

PCI:

./qemu-system-riscv64 -nographic \
	-machine virt,iommu-sys=off \
	-device riscv-iommu-pci,addr=1.0 \
	-cpu 'rv64,h=true,sscofpmf=true' \
	-smp 8 \
	-netdev tap,ifname=tap0,script=no,id=net0 \
	-device virtio-blk-pci,iommu_platform=on,disable-legacy=on,drive=disk \
	-drive file=${DRIVE_FILE},id=disk,format=raw,if=none,cache=writeback \
	-device virtio-blk-pci,iommu_platform=on,disable-legacy=on,drive=disk2 \
	-drive file=${DRIVE2},id=disk2,format=raw,if=none,cache=writeback \
	-device virtio-net-pci,iommu_platform=on,disable-legacy=on,netdev=net0 \
	-m 16G \
	-bios /usr/local/share/opensbi/lp64/generic/firmware/fw_jump.elf \
	-kernel ${KERNEL}

[jrtc27]

sys/riscv/include/iommu.h

9

Just taking an initial cursory glance, all these #include <riscv/...> should be #include <machine/...>

[minsoochoo0122_proton.me]

sys/riscv/iommu/iommu.h

2–27

The preferred policy recommends not writing out the full license text, but ofc you can if you want.

same for other newly added files.

https://docs.freebsd.org/en/articles/license-guide/

/*
 * Copyright (c) [year] [your name]
 *
 * SPDX-License-Identifier: BSD-2-Clause
 */

[adrian]

sys/riscv/iommu/iommu_if.m

2

Would it make sense to push this layering abstraction up to sys/dev/iommu/ first? So arm64 can also implement >1 iommu provider?

[br]

sys/riscv/iommu/iommu_if.m

2

arm64 has one already:

sys/arm64/iommu/iommu_if.m

but we can unite them later in a single one

[br]

Regenerate