Recently, I’ve started to explore RISC-V. I experienced the journey as pretty refreshing, particularly because I’ve been working on x86 low-level software almost exclusively for about 10 years.
In this post, I want to quickly go over some high-level stumbling blocks I noticed as I was starting out. I’m probably going to write about more technical differences in subsequent blog posts.
While reading the rest of this post, please keep in mind: RISC-V is simple! If you managed to do any low-level coding on x86, you will find your way around RISC-V with little effort. It’s easy to forget how crazy some parts of x86 are. Just see my past posts.
Choosing a Development System
If you want to do any OS work, you need a system that implements the Privileged ISA. Supporting the Privileged ISA is synonymous to being able to run UNIX-like systems, because it brings user-/supervisor-mode distinction and paging. Working on real hardware is generally preferred to working in emulators, because the code eventually has to run on metal anyway and emulators can be too forgiving for certain classes of problems.
There is a plethora of RISC-V microcontrollers for a couple of dollars that all don’t support the Privileged ISA. At the time of writing, the only board you can buy that does support it is the $1000 HiFive Unleased, which was beyond my “I just want to play around with this” budget.
The next most realistic option is an FPGA emulating one of the open-source RISC-V implementations. BOOM in particular looks interesting, because it supports everything you need to get Linux going on it. Unfortunately, setting up the whole tooling to build and deploy BOOM on a FGPA or simulator is also sufficiently demotivating to make this viable for a hobby OS project.
Spike works, but is weird. First, there is no documentation of the machine it emulates and you have to go read the source code. The serial console in particular is strange and I’m not the only one to point this out.
Qemu turned out to be the emulator of choice, because its machine
model is easily explorable (type
info mtree) and because recent
versions include OpenSBI as a
firmware by default (
-bios default). Exploring the machine model
quickly answers questions like where to find RAM and OpenSBI is
helpful in getting early console output.
Problematic is that there is lots of misleading documentation out
there and that makes it look complicated to boot a simple ELF image
that wants to speak to a working
implementation. In Qemu >=4.1, this is as easy as
-M virt -bios default -device loader,file=kernel.elf.
Coming from the x86 world, you expect a gigantic PDF that tells you everything you need to know: The Intel Software Developer’s Manual. Given the complexity of x86, this document actually does a good job of telling you what you need to know.
RISC-V documentation relevant for kernel development is split between
the Privileged and Unprivileged ISA
specifications. This leads to
weird situations, for example when trying to figure out how
(basically the system call instruction) works.
The instruction itself is documented in the Unprivileged ISA, but there you find no details about what it actually does. The Privileged ISA enumerates the instruction again and gives a high-level description. It’s then up to careful reading of the CSR description to see what actually happens, specifically with interrupt masking.
If you compare this to the Intel SDM’s description of the
instruction, it’s pretty straight-forward to figure out what’s going
on. You go to the one list of all instructions and find
syscall. You read the pseudo-code description and it mentions all
changes to system state and every
MSR that is
involved. Additionally, there is a list of what exceptions will be
caused in what situations.
This is my main problem with the state of RISC-V documentation so far. Everything feels scattered. But don’t get me wrong: Because everything is also pretty simple, you will eventually piece it together. The worst case is trying to get the bigger picture from Linux source code comments.
If this post sounded overly negative, this is not intended. I’m personally very excited about RISC-V and its possiblities. If the amount of assembly I have to write to get a toy kernel going is any indication, RISC-V wins hands down.
The hurdles with hardware availability and documentation will go away over time. Eventually there will be cheap development boards (a Raspberry Five would be nice) and the OSDev Wiki will catch up to RISC-V. This will make this already very approachable architecture even easier to work with for beginners.
Stay tuned for more RISC-V content from an x86 angle in future posts.