audk/OvmfPkg/RiscVVirt
Laszlo Ersek 29cce3356a OvmfPkg/RiscVVirt/README: document direct kernel boot
Direct (fw_cfg) kernel boot works on the RiscVVirt firmware platform too;
I've tested it after extracting the kernel, initrd, and kernel command
line from "openSUSE-Tumbleweed-RISC-V-E20-efi.riscv64.raw". Document this
type of boot, because at least historically, fw_cfg kernel boot was
implemented differently between OVMF and ArmVirtQemu. Thanks: Drew, Sunil.

Cc: Andrei Warkentin <andrei.warkentin@intel.com>
Cc: Andrew Jones <ajones@ventanamicro.com>
Cc: Ard Biesheuvel <ardb+tianocore@kernel.org>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Jiewen Yao <jiewen.yao@intel.com>
Cc: Jordan Justen <jordan.l.justen@intel.com>
Cc: Sunil V L <sunilvl@ventanamicro.com>
Signed-off-by: Laszlo Ersek <lersek@redhat.com>
Reviewed-by: Andrew Jones <ajones@ventanamicro.com>
Reviewed-by: Sunil V L <sunilvl@ventanamicro.com>
2023-09-14 09:56:20 +00:00
..
Library OvmfPkg/RiscVVirt: Avoid printing hard coded timeout value 2023-07-21 05:56:55 +00:00
PciCpuIo2Dxe OvmfPkg/RiscVVirt: Add PciCpuIo2Dxe module 2023-02-16 05:53:28 +00:00
Sec OvmfPkg/RiscVVirt: SecEntry: Remove unnecessary assembly directives 2023-07-31 12:37:12 +00:00
README.md OvmfPkg/RiscVVirt/README: document direct kernel boot 2023-09-14 09:56:20 +00:00
RiscVVirt.dsc.inc MdeModulePkg: Duplicate BaseRngLibTimerLib to MdeModulePkg 2023-09-08 09:48:55 +00:00
RiscVVirt.fdf.inc OvmfPkg/RiscVVirt: Add support for separate code and variable store 2023-06-23 04:49:11 +00:00
RiscVVirtQemu.dsc OvmfPkg: move PciEncoding into AcpiPlatformLib 2023-06-23 17:26:37 +00:00
RiscVVirtQemu.fdf OvmfPkg/RiscVVirt: use 'auto' alignment and FIXED for XIP modules 2023-07-31 12:37:12 +00:00
VarStore.fdf.inc OvmfPkg/RiscVVirt: Fix issues in VarStore Blockmap config 2023-08-25 01:54:50 +00:00

README.md

Support for RISC-V QEMU virt platform

Overview

RISC-V QEMU 'virt' is a generic platform which does not correspond to any real hardware.

EDK2 for RISC-V virt platform is a payload (S-mode) for the previous stage M-mode firmware like OpenSBI. It follows PEI less design.

The minimum QEMU version required is 8.1 or with commit 7efd65423a which supports separate pflash devices for EDK2 code and variable storage.

Get edk2 sources

git clone --recurse-submodule git@github.com:tianocore/edk2.git

Build

Using GCC toolchain

Prerequisite: RISC-V GNU compiler toolchain should be installed.

export WORKSPACE=`pwd`
export GCC5_RISCV64_PREFIX=riscv64-linux-gnu-
export PACKAGES_PATH=$WORKSPACE/edk2
export EDK_TOOLS_PATH=$WORKSPACE/edk2/BaseTools
source edk2/edksetup.sh --reconfig
make -C edk2/BaseTools
source edk2/edksetup.sh BaseTools
build -a RISCV64 --buildtarget RELEASE -p OvmfPkg/RiscVVirt/RiscVVirtQemu.dsc -t GCC5

Using CLANGDWARF toolchain (clang + lld)

Prerequisite: LLVM toolchain with clang and lld should be installed.

export WORKSPACE=`pwd`
export CLANGDWARF_BIN=/usr/bin/
export PACKAGES_PATH=$WORKSPACE/edk2
export EDK_TOOLS_PATH=$WORKSPACE/edk2/BaseTools
source edk2/edksetup.sh --reconfig
make -C edk2/BaseTools
source edk2/edksetup.sh BaseTools
build -a RISCV64 --buildtarget RELEASE -p OvmfPkg/RiscVVirt/RiscVVirtQemu.dsc -t CLANGDWARF

After a successful build, two files namely RISCV_VIRT_CODE.fd and RISCV_VIRT_VARS.fd are created.

Test

Below example shows how to boot openSUSE Tumbleweed E20.

  1. RISC-V QEMU pflash devices should be of of size 32MiB.

    truncate -s 32M RISCV_VIRT_CODE.fd

    truncate -s 32M RISCV_VIRT_VARS.fd

  2. Running QEMU

     qemu-system-riscv64 \
     -M virt,pflash0=pflash0,pflash1=pflash1,acpi=off \
     -m 4096 -smp 2 \
     -serial mon:stdio \
     -device virtio-gpu-pci -full-screen \
     -device qemu-xhci \
     -device usb-kbd \
     -device virtio-rng-pci \
     -blockdev node-name=pflash0,driver=file,read-only=on,filename=RISCV_VIRT_CODE.fd \
     -blockdev node-name=pflash1,driver=file,filename=RISCV_VIRT_VARS.fd \
     -netdev user,id=net0 \
     -device virtio-net-pci,netdev=net0 \
     -device virtio-blk-device,drive=hd0 \
     -drive file=openSUSE-Tumbleweed-RISC-V-E20-efi.riscv64.raw,format=raw,id=hd0
    

    Note: the acpi=off machine property is specified because Linux guest support for ACPI (that is, the ACPI consumer side) is a work in progress. Currently, acpi=off is recommended unless you are developing ACPI support yourself.

  3. Running QEMU with direct kernel boot

    The following example boots the same guest, but loads the kernel image and the initial RAM disk (which were extracted from openSUSE-Tumbleweed-RISC-V-E20-efi.riscv64.raw) from the host filesystem. It also sets the guest kernel command line on the QEMU command line.

     CMDLINE=(root=UUID=76d9b92d-09e9-4df0-8262-c1a7a466f2bc
              systemd.show_status=1
              ignore_loglevel
              console=ttyS0
              earlycon=uart8250,mmio,0x10000000)
    
     qemu-system-riscv64 \
     -M virt,pflash0=pflash0,pflash1=pflash1,acpi=off \
     -m 4096 -smp 2 \
     -serial mon:stdio \
     -device virtio-gpu-pci -full-screen \
     -device qemu-xhci \
     -device usb-kbd \
     -device virtio-rng-pci \
     -blockdev node-name=pflash0,driver=file,read-only=on,filename=RISCV_VIRT_CODE.fd \
     -blockdev node-name=pflash1,driver=file,filename=RISCV_VIRT_VARS.fd \
     -netdev user,id=net0 \
     -device virtio-net-pci,netdev=net0 \
     -device virtio-blk-device,drive=hd0 \
     -drive file=openSUSE-Tumbleweed-RISC-V-E20-efi.riscv64.raw,format=raw,id=hd0 \
     -kernel Image-6.5.2-1-default \
     -initrd initrd-6.5.2-1-default \
     -append "${CMDLINE[*]}"
    

Test with your own OpenSBI binary

Using the above QEMU command lines, RISCV_VIRT_CODE.fd is launched by the OpenSBI binary that is bundled with QEMU. You can build your own OpenSBI binary as well:

OPENSBI_DIR=...
git clone https://github.com/riscv/opensbi.git $OPENSBI_DIR
make -C $OPENSBI_DIR \
    -j $(getconf _NPROCESSORS_ONLN) \
    CROSS_COMPILE=riscv64-linux-gnu- \
    PLATFORM=generic

then specify that binary for QEMU, with the following additional command line option:

-bios $OPENSBI_DIR/build/platform/generic/firmware/fw_dynamic.bin

Note that the above only makes a difference with software emulation (which you can force with -M accel=tcg). With hardware virtualization (-M accel=kvm), KVM services the SBI (Supervisor Binary Interface) calls internally, therefore any OpenSBI binary specified with -bios is rejected.