Fiwix: Small Unix-Like Kernel
41 comments
·February 5, 2025PeterWhittaker
bitwize
You can write a fully 32-bit kernel, delegate entering protected mode and using the BIOS to load from disk to the bootloader, and tell your students "this bit won't be on the exam".
MisterTea
> i386 implies BIOS, AFAIK, and requires pre-x86_64 assembler, which is only partially transferable to modern CPUs.
Nothing wrong with this IMO. You dont want a 64-bit Arm/x86 hegemony. If the OS is designed properly you won't notice the platform underneath which means you have the right abstractions. That is an important challenge for students - decoupling platform dependency from the OS architecture. It also teaches them older tech which gives them a history lesson so they understand how and why we got here.
6SixTy
If a 64 bit ARM/x86 hegemony is a big concern, then why not drop i386 and go with RISC-V? There's both a 32 and 64 bit variants, and a free emulator. Also, it's more useful academically and practically.
MisterTea
Well Risc-v isn't as established as 32bit x86 which has a wide breadth of information available. There are also plenty of 32 bit x86 machines still around so testing your OS on real hardware and watching your programs work is quite a treat.
pabs3
In use by the Bootstrappable Builds folks to be able to build the Linux kernel before having a Linux or other kernel binary, since Linux requires compilers that are only bootstrapped later in the process.
https://bootstrappable.org/ https://lwn.net/Articles/983340/
bitwize
Just a hobby, not big and professional like GNU. I love it.
Rochus
What's the difference to e.g. xv6?
There are already so many Unix like operating systems of all sizes written in C (e.g. Minix, xv6, BSD, Darwin, Hurd, Lemon, MentOS, SanOS). If I were to go to such an effort, I would try another language, or even my own (like e.g. Bunnix written in Hare). It is meanwhile more than clear that it can be done in C. But C is not really a good programming language (weird syntax, incomplete type checking, many specification gaps, etc.), so it would actually be more interesting to explore the suitability of other languages for such a task.
BirAdam
C has been the standard system language almost since the time of its creation. Teaching that standard is valuable even if we’re in the midst of a change.
As for why, a system designed to be pedantic will necessarily be different from one intending to be production ready.
Why i386 and BIOS? It is simple enough for the purpose of teaching. Inundating a student with modern x86-64, UEFI (where most implementations do not meet standards), and such is not a great idea. People just get overwhelmed and shut down. It may be worthwhile to start with something simpler honestly, like DOS and 8088, or even CP/M and 8080. Working one up through time exposes the individual to simpler things and then progressively more complex things. It is useful.
As a side benefit, teaching people about these older things instructs them on how and why things came to be the way they are which may help create better entrepreneurs.
Rochus
> even if we’re in the midst of a change.
What change do you refer to?
Concerning C: C has obviously proven itself in countless system and embedded projects of all sizes. Nevertheless, C essentially represents the state of knowledge of the early seventies. Projects like the one discussed would be a good opportunity to test newer languages for their suitability for operating system and embedded development, without the pressure of a commercial project.
> Why i386 and BIOS? It is simple enough for the purpose of teaching.
Agree; that's also one of the reasons operating systems like xv6 dedicated for teaching do so; I'm therefore interested in how Fiwix positions itself in relation to xv6.
sjsdaiuasgdia
When you write one, you can make those choices however you like.
This target was sufficiently interesting to the people who pursued it. It doesn't have to meet your arbitrary standards.
Rochus
Unfortunately, your comment neither answers my question nor is it helpful in relation to the other arguments.
sjsdaiuasgdia
Other people's hobbies don't have to align to your priorities.
In other words: https://i.kym-cdn.com/entries/icons/facebook/000/043/517/i_d...
Paul-Craft
This one is just a hobby (won't be big and professional like gnu), after all. ;-)
Rochus
Most other mentioned are also just hobby kernels; actually also Linux and BSD started that way.
homarp
parent is quoting this https://fossbytes.com/linus-torvaldss-famous-email-first-lin...
glonq
I hope this newcomer Fiwix doesn't steal too much mindshare away from HURD...
pabs3
IIRC Fiwix has syscalls compatible with Linux, which could be useful in some cases.
Rochus
Thanks for the first useful answer so far. Does this mean that Fiwix would allow ELF binaries compiled for Linux to be executed natively in a Fiwix system without emulation? Have you tried this?
pabs3
I think that is correct, depending on how modern the syscalls used are I guess. I expect it doesn't implement a lot of them. I haven't tried it though.
null
Interesting. While I do appreciate that it runs on i386, I have to question its being aimed at students - or at least students of anything other than the history of computing. i386 implies BIOS, AFAIK, and requires pre-x86_64 assembler, which is only partially transferable to modern CPUs.
Oh, for sure, one would learn an awful lot working with those technologies, but as soon as one wished to work on more modern systems, one would have to forgot all of the former and all the ugly of the latter.
This isn't really a criticism, more an expression of puzzlement. Normally, I'm a pretty good contrarian, but I don't get it.
True, there was this Finn who did this some time ago and it worked out pretty well, but I still don't get why anyone would start with i386 and BIOS.