Show HN: A DOS-like hobby OS written in Rust and x86 assembly
github.comTo try it out, simply build the project yourself from source, or use attached bootable ISO image of the system (in Releases on Github) and run it in QEMU.
https://blog.vxn.dev/rou2exos-rusted-edition Memory-safe language. x86_64, with Arm on the roadmap. Networking stack. Boots from a CD and via multiboot. Your hobby project wipes the floor with DOS. Whoa there. Gotta run Doom and BASIC to compete with DOS. That is the officially recognized DOS-Kármán line. What a challenge! Need to implement some interrupts it seems then, to provide an API for filesystem and so... Thank you for such idea Doomgeneric has a very thin platform-specific layer: https://github.com/ozkl/doomgeneric?tab=readme-ov-file#porti... I want some TSRs and print spoolers... DOOM required DOS 5.0. rou2exOS is only the second take. Watch this area ;) Borland languages and xBase ecosystem as well Nobody cares about Lotus 1-2-3 support any longer :) What about Wordstar? The text editor JOE has a WordStar mode. WordPerfect was better anyway :) (SCNR) also, can't be a dos with the 'dir' command. Afaik there is a 'DIR' command in MS-DOS. Anyway, what would be a better command to list a directory? I could think of 'ls' maybe I would most likely end up with something like this: I suggest: To create a file, we should take inspiration from TOPS-10 and call the command `make` [0]. When issuing `make love`, it should ask for confirmation to the user: `not war?`. Very good design. Implements only the essential tasks an operator would need with concise and well-known mnemonics. I’m sure this will be immediately intuitive to anyone, with no potential for mistakes whatsoever. CAT was on the Amstrad CPC. Love it. As someone old enough to have started on MS-DOS 3.3, dir has been there all time. CATALOG, DSPFLR, Get-ChildItem It would seem to logically follow that a Disk Operating System would have a directory list utility for disks like ‘dir’. It was there in the first version called 86 DOS. `dir` was also in CP/M. Though it didn't have a concept of a file hierarchy, so it listed all files on the disk (but did support supplying a pattern to filter against). > Rust and x86 assembly then > Memory-safe language. What's the point? Looks like today Rust is like 3D printing was. As if it makes something better. Printing was hyped and advertised by printers sellers and manufacturers. Finally they run out of money. As for project, it's cool if compatible with old soft. Otherwise suitable mostly for education and masochism. Long way to become practical anyway even if it gets traction. The idea of Rust is not that it is 100% safe, but rather that it is able to encapsulate unsafety and divide the program in two parts: - unsafe code plus modules that support it (the "trusted base") - all the rest Rust's promise is that there is no way to trigger any undefined behavior from bugs happening in "all the rest" of the code. If that code makes for more than 95% of the total, then that's a huge win compared to a completely unsafe language. Also, Rust's support for inline assembly is in my opinion better than C's, it's much easier to specify and figure out the constraints on the boundary between Rust/assembly. You can write memory safe code in any language, but having a machine i.e. the Rust compiler check it for you is less error-prone than if a human does it. Also if you look at the repo, only 3% of the codebase is in Assembly.
IMO if >95% project is in Rust, you can definitely claim it's a Rust project. Finally, a positive comment about Rust that I can 100% agree with :) Well, the point is to experiment with Rust no_std+no_main environment while trying to educate myself on how the things work under the hood. The project itself is part "just" a rewritten system (from C to Rust), and part an enhancement of such system furthermore. It lacks the external program execution though yet. I love that the networking stack uses SLIP and slattach(1)! I was playing with a toy TCP/IP stack, and decided using SLIP over a pty on Linux was a great way to interface with the kernel. Unfortunately it looks like macOS previously shipped with slattach(1) a very long time ago, but no longer does. I'm curios if other people have used SLIP on macOS to get a dead-simple, cross-platform API to the networking stack? The alternative would be to use tun/tap on Linux and utun on macOS, but SLIP would be so much nicer. Out of curiosity, why x86? Is it the preponderance of resources? The weird instruction format? The complexity of the boot sequence? Are you specifically trying to mimic DOS? > A support for the ARM architecture (aarch) is coming soon too. Wow! How do you support a DOS-like OS across multiple architectures when DOS itself is tightly tied to interactions among the program, the system code, and the architecture? I have not looked at this project, but my guess would be: x86 is a widely available platform that, because of its history and relentless compatibility, contains a lot of legacy interfaces that make implement a very simple, thin-layer and “DOS-like OS” without the need to parse device trees, set up MMUs, deal with complex busses like PCI(e) and so on. It is much harder to bootstrap a simple OS in ARM, and it won’t stay very simple unless you accept significantly more limitations than you would under x86. (For example, you can’t do very much with the MMU off on ARM, and you also don’t have convenient BIOS interfaces that allow you to, say, read a sector, or wait for a keypress, with just a few lines of assembly). The x86 arch is used because this system iteration derives from the first one, which relies on BIOS interrupts and inline assembly in Turbo C. I am not trying to mimic (MS-)DOS exclusively, but both systems are highly inspired by it. IMO multiple archs could be supported as Rust compiler allows the target arch specification, so one would build a specific target before the build itself. I would have preferred something like this to the current UEFI environment and shell, a FLOSS 64-bit DOS-like. A cool retro boot manager and diagnostic env perhaps. Could this run from an efi system partition? Seems to support fat12, what about gpt? Does it poke video hardware like DOS, or have a terminal like output? Booting from an EFI system partition has not been tested yet. FAT12 is the only filesystem (ok, there is a memdisk implementation, but it won't work now) supported, so GPT is not supported at the moment too (yet). Kinda aiming for FAT32 implementation to be the very next implemented (flash disks are usually FAT32 iirc). Not sure about the last question: the OS utilizes/directly writes to the VGA buffer in memory, the provided resolution is 80x25 by GRUB. So MBR partitions? Or no partitions, like from a floppy? Or perhaps it doesn’t know because grub handles that part. For FAT12, it reads the first sector (0 or bootsector) of the floppy provided to gather information like bytes per sector, reserved sectors count, LBA of the root directory, etc. Good job OP, what a missed opportunity to say: "just a hobby, won't be big and professional like Linux" (: Are you planning adding support for Czech diacritics? I am not at the moment. The goal is to keep this iteration in English for now. The first iteration was in Czech in the beginning though. Is this using a custom VGA driver from scratch in Rust?? Sort of, but IMO that module(s) is just a wrapper for video buffer. DOS-like but not DOS-compatible, correct? You are right. The first iteration however is 16bit and is very close to MS-DOS in terms of compatibility. Moreover, any OS that can handle simple disk I/O ops could be considered a DOS system too, innit? Correct, there is a difference between MS-DOS and IBM-PC compatible and a DOS (ex: all the DOSes that existed for Amiga/Apple II/Commodore etc). There are many DOSes (and even MSDOSes, because yay early PC era incompatibilities!), but there is a very dubious list of things needed to be MS-DOS and IBM-PC compatible. You can probably do it if you're willing to setup a hypervisor and emulate some hardware. NGL one of my long term projects was/is something exactly like this but UEFI and secure boot. The idea being to use the VM extensions to create IBM-PC and DOS compatible environments. For anything using DPMI[1] I'd probably do the same trick as Win95 did and just replace it with my own implementation so it's not too overburdened with layers. [1] https://en.wikipedia.org/wiki/DOS_Protected_Mode_Interface A version of (say) FreeDOS that was layered on top of the EFI API instead of PC98 firmware interrupts would be quite interesting. That would be a major architectural change to most of the programs, of course. But one would have provided the EFI Shell with essentially a complete suite of MS-DOS (albeit not PC-DOS or DR-DOS) commands. That could probably be quite easily ported to (say) ARM whereas the original still has x86isms. On the other hand, did you see https://github.com/FlyGoat/csmwrap when it came up a few weeks ago? * https://news.ycombinator.com/item?id=44101828 There are already projects to provide replacements for the vanished Compatibility Support Module. I did see that and for people that need that specific functionality it seemed like a good solution. My goal was more "What if DOS hadn't ended and but kept up support for modern hardware" along with emulation of common things in DOS gaming. So for example you would be able to set up a PIV that mapped certain resources directly or emulated them depending on the need. Could I use DOSBox for this... yes, but this is a "why not" sort of thing. I figured it would be a good excuse to learn OS dev. But life has kept me busy for now. The 16bit one also looks fun[0] and would run on old PCs the new one does not. Have you considered open sourcing that one? NGL, the codebase of the first iteration is pretty much a mess. It had been written in my mid-teenage years, so the C code is not very consistent across the project. There is a filesystem prototype written in Go as well... Also, it is quite complex to even compile RoureXOS: it needs Borland TCC and TASM (ca 1989 btw) to properly build a 16bit executable(s). It is feasible to compile it using DOSBox though. Maybe some day I could open source that one repository too. ...meaning MS-DOS compatible :) I.e.: runs Alley Cat and Dune 2 - and Doom. And prevents Lotus 1-2-3 from running? (-: That's the litmus test. No version of DOS is complete until Lotus 1-2-3 no longer runs. it needs event queue to support async runtimes. How about the event loop in its completeness? Can it run Crysis?? :-)) I don’t understand: what makes this “DOS-like”? Not even the command names match DOS, except for dir and cls, and the architecture and feature set are completely different. The original blog post has more details: https://krusty.space/projects/rourexos/ Ultimately, it's an eye of the beholder type thing, but it seems very fair to call it DOS-like to me. The DOS inspiration is pretty blatant all throughout it from its aesthetic, commands (more than just dir and cls is shared), booting from a 1.44 floppy, etc. And if you are coming from a primarily windows computing background, then "DOS-like" is entirely appropriate to me. It looks like a fun project. Ah yes, the first version does indeed sound DOS-like. The second one diverges quite a bit though. Fun project indeed. You are right, it goes way ahead the first iteration. I call it DOS-like because of its resemblance to the Text mode-only operating systems (and to MS-DOS especially). On the other hand, it shares very little as the architecture, command set, or hardware utilization approach are concerned when compared to MS-DOS. Single-user, single-task, VGA text-mode operating system with a basic filesystem (fat12), I guess. Note the “guest” username in the prompt is a const, not multi-user:
https://github.com/krustowski/rou2exOS/blob/6f85955dd339f09d...
...and so on. CAT CATalogue - output the contents of current directory
RM to Raster Memory - load contents of named file in framebuffer
MV Make Virtual - map the file into memory and output the address
LS Load System - attempt to reboot using the named file as the kernel
CD Create Directory - self-explanatory
SH System Halt - immediately stop all processing
This would be a much simpler command interface. RM Relocate me - you could also use the "-rf" flag to relative fuzzy matching, for instance `rm -rf tmp` to find any temp-like folder near your current location. Pretty standard stuff.
CD Clear data - `cd -L /temp` would "liquidate" the directory.