I have read some articles related to set up a QEMU environment for enabling RISC-V Linux kernel development. Unfortunately, none of them describes all the required steps to get a fully prepared RISC-V 64bit system. In this article, I am describing how to build the latest (as of the moment of writing) Linux Kernel for RISC-V 64 architecture and deploy it along with minimal environment: busybox command-line. PREREQUISITES I use Ubuntu as my main desktop OS, so all the described steps related to it. Probably it could be reproduced on other Debian-based distributions. My precise Ubuntu version is: hedin@home:~ linux/riscv64-linux$ lsb_release -a
LSB Version:    core ubuntu2-noarch:security ubuntu2-noarch
Distributor ID: Ubuntu
Description:    Ubuntu LTS
Release: Codename:       focal /projects/ -11.1 .0 -11.1 .0 20.04 .2 20.04 The following is the project's catalog structure: .....
hedin@home:~ linux/riscv64-linux$ tree -L .
├── busybox
├── initramfs
└── qemu /projects/ 1 Their intention is following: - contains minimal command-line environment to work with the Kernel. busybox - initial RAM disk for the Kernel. initramfs - emulator I run the RISC-V enabled Kernel on. qemu From now on I am using "$" instead of the full path to save space. Note 1: I intentionally didn't include Linux Kernel in this project as it is a huge amount of source code and it makes no sense to copy it to every project. It is better to have only one shared Kernel repository for all the related projects and use Git branches for each of them. Note 2: SETUP Firstly, we need to install common development tools: $ sudo apt install autoconf automake autotools-dev curl libmpc-dev libmpfr-dev libgmp-dev gawk build-essential bison flex texinfo gperf libtool patchutils bc zlib1g-dev libexpat-dev git And RISC-V compiler: $ sudo apt-get install -y gcc-riscv64-linux-gnu Secondly, we need to install Qemu from the source. The reason is that most Linux distributions don't contain the latest versions and also don't contain (or uses limited) RISC-V support. $ git clone https: $ cd qemu
$ git checkout v5 $ ./configure --target-list=riscv64-softmmu
ERROR: Cannot find Ninja //github.com/qemu/qemu .2 .0 Something went wrong. Qemu requires a Ninja-build package. Let's install it: $ sudo apt install ninja-build And try again: $ ./configure --target-list=riscv64-softmmu
$ make -j $(nproc)
$ sudo make install OK, now it is built. Let's return to the project catalog and obtain the Linux Kernel: $ cd .. Get the Linux tree: $ git clone https: //github.com/torvalds/linux Check out the latest (as of this moment of writing) kernel version: $ git checkout -b risc-v v5 .11 Make default config: $ make ARCH=riscv CROSS_COMPILE=riscv64-unknown-linux-gnu- defconfig And get a new error message: compiler not found 'riscv64-unknown-linux-gnu-gcc' The issue is that there is no standard toolchain that required my meets: $ sudo apt search riscv64-unknown-
Sorting... Done
Full Text Search... Done
binutils-riscv64-unknown-elf/focal,now ubuntu1 amd64 [installed,automatic]
  GNU assembler, linker and binary utilities RISC-V processors
gcc-riscv64-unknown-elf/focal,now ubuntu1 amd64 [installed]
  GCC compiler embedded RISC-V chips 2.34 -0 for 9.3 .0 -0 for There are 2 approaches for solving this issue: Compile it manually as described in the following article: https://programmersought.com/article/4626958334/ Get the latest build from bootlin. I will use the latter approach: Goto https://toolchains.bootlin.com/ Select arch - riscv64 Select libc - glibc Copy the link and download it somewhere (not in my project catalog): $ wget https: //toolchains.bootlin.com/downloads/releases/toolchains/riscv64/tarballs/riscv64--glibc--bleeding-edge-2020.08-1.tar.bz2 5) sudo mkdir -p /opt/bootlin 6) sudo tar jxf riscv64--glibc--bleeding-edge tar.bz2 -C /opt/bootlin/ -2020.08 -1. Now we are ready to compile the kernel again: $ make ARCH=riscv CROSS_COMPILE= bootlin/riscv64--glibc--bleeding-edge /bin/riscv64-buildroot-linux-gnu- defconfig
$ make ARCH=riscv CROSS_COMPILE= bootlin/riscv64--glibc--bleeding-edge /bin/riscv64-buildroot-linux-gnu- -j $(nproc)
...............................................................
  LD      vmlinux.o
  MODPOST vmlinux.symvers
  MODINFO modules.builtin.modinfo
  GEN     modules.builtin
  LD      .tmp_vmlinux.kallsyms1
  KSYMS   .tmp_vmlinux.kallsyms1.S
  AS      .tmp_vmlinux.kallsyms1.S
  LD      .tmp_vmlinux.kallsyms2
  KSYMS   .tmp_vmlinux.kallsyms2.S
  AS      .tmp_vmlinux.kallsyms2.S
  LD      vmlinux
  SYSMAP  System.map
  MODPOST Module.symvers
  OBJCOPY arch/riscv/boot/Image
  CC [M]  fs/efivarfs/efivarfs.mod.o
  CC [M]  fs/nfs/flexfilelayout/nfs_layout_flexfiles.mod.o
  GZIP    arch/riscv/boot/Image.gz
  LD [M]  fs/efivarfs/efivarfs.ko
  LD [M]  fs/nfs/flexfilelayout/nfs_layout_flexfiles.ko
  Kernel: arch/riscv/boot/Image.gz is ready /opt/ -2020.08 -1 /opt/ -2020.08 -1 The kernel is ready. It's time to compile BusyBox: Apply default config: $ CROSS_COMPILE= bootlin/riscv64--glibc--bleeding-edge /bin/riscv64-buildroot-linux-gnu- make defconfig /opt/ -2020.08 -1 Adjust one setting. I want a single monolith executable file without dependencies to dynamic libraries: $ CROSS_COMPILE= bootlin/riscv64--glibc--bleeding-edge /bin/riscv64-buildroot-linux-gnu- make menuconfig /opt/ -2020.08 -1 In the configuration change the following setting: Build binary (CONFIG_STATIC=y) static Build: $ CROSS_COMPILE= bootlin/riscv64--glibc--bleeding-edge /bin/riscv64-buildroot-linux-gnu- make -j $(nproc) /opt/ -2020.08 -1 Install: $ CROSS_COMPILE= bootlin/riscv64--glibc--bleeding-edge /bin/riscv64-buildroot-linux-gnu- make install /opt/ -2020.08 -1 The last thing we need to do is to create an initial RAM disk. And again I create it out of the project's tree as it will need for my purposes only once. After completion, we can remove it. So let's create it somewhere: $ mkdir _install
$ cd _install Create device tree catalog: $ mkdir -p dev And two required devices - console and ram: $ sudo mknod dev/ c $ sudo mknod dev/ram b console 5 1 1 0 And Init script, the first userspace process that the Kernel will run. This script creates required catalogs and mounts special-purpose kernel filesystems to them. Also, it measures how long this particular boot process took. And in the end, it runs a command interpreter, that is, bash-like busybox. $ vim init File content: echo mkdir /proc /sys /tmp
mount -t proc none /proc
mount -t sysfs none /sys
mount -t tmpfs none /tmp
echo -e exec /bin/sh #!/bin/sh "### INIT SCRIPT ###" "\nThis boot took $(cut -d' ' -f1 /proc/uptime) seconds\n" Let's pack file structure to the format the Kernel recognizes: find -print0 | cpio oH newc | gzip > -0 -9 /initramfs/initramfs.cpio.gz < > PATH_TO_OUR_PROJECT RUN Run qemu passing it the kernel, init RAM disk, and appropriate parameters: $ sudo qemu-system-riscv64 -nographic -machine virt -kernel ../linux-torvalds-tree/arch/riscv/boot/Image -initrd initramfs/initramfs.cpio.gz -append "root=/dev/vda ro console=ttyS0" ................................................................................................................................... I am in: / # ls
bin      init     proc     sbin     tmp
dev      linuxrc  root     sys      usr
/ # uname -a
Linux (none) # SMP Fri Mar : : MSK riscv64 GNU/Linux 5.11 .0 2 12 09 36 08 2021 References: https://risc-v-getting-started-guide.readthedocs.io/en/latest/linux-qemu.html https://programmersought.com/article/4626958334/ https://stackoverflow.com/questions/63393780/riscv64-linux-kernel-compilation-issue https://ibug.io/blog/2019/04/os-lab-1/

Target

Read My Stories

Too Long; Didn't Read

Make resilience your competitive advantage

How To Set Up The Environment for RISCV-64 Linux Kernel Development In Ubuntu 20.04

About Author

Comments

TOPICS

THIS ARTICLE WAS FEATURED IN

Related Stories

Untitled Story

Containerization In 2023: Strive for Maximum Modularity via the Cattle Model

Docker is dead. Long live the Unikernel.

Faster, Stronger, and 100% More Custom: A Grand Tour of Linux Kernels

How to Make a Linux Kernel with Nasm, Go Binary, Mini Linux

Kernel mode hooking, good features getting rotten

Containerization In 2023: Strive for Maximum Modularity via the Cattle Model

Docker is dead. Long live the Unikernel.

Faster, Stronger, and 100% More Custom: A Grand Tour of Linux Kernels

How to Make a Linux Kernel with Nasm, Go Binary, Mini Linux

Kernel mode hooking, good features getting rotten

Light-Mode

Classic

Newspaper

Dark-Mode

Neon Noir

Minty

HN StartUps