klange's OS Development Repo

I'm writing an OS because I'm bored and want a massive project to suck up the little pieces of my time for the next few years.

Testing it Out

Grab bootdisk.img from the git repo and load it into a virtual machine. The bootdisk will start GRUB with a single menu entry to boot the kernel off of the same diskette.

For example, you can boot the disk in qemu:

qemu -fda bootdisk.img

Or you could set up a VirtualBox machine and load the floppy image.

Goals and Roadmap

Overall, the goal of this project is to write a relatively POSIX-compatible OS from the ground up. With a focus on generic hardware functionality and universal specifications like VESA, I hope to eventually get something fairly complete in terms of what an OS should be. At some times, I may focus on an actual piece of complex hardware (I am looking to write a basic driver for Intel graphics cards based on the X driver and the Mesa components), but in general, I will stick to generic interfaces.

Basic Goals

Create a working modular monolithic kernel capable of executing arbitrary ELF binaries
Write, from scratch, a C standard library using past experience in writing standard library functions
Support POSIX threads
Implement an existing file system, specifically EXT2
Be able to manipulate VESA modes to run at an optimal resolution for graphics
Handle basic networking on a virtual Ethernet device (DHCP, TCP, etc.)

Loftier Goals

Some things are far easier said than done, but I like to say them anyway. The time span for these depends greatly on how quickly the basic goals are completed and can range anywhere from a few months to years from now.

Create a working implementation of Wayland under VESA (will be slow)
Port Qt (under Wayland) and some Qt apps
Support audio in a way that doesn't suck like Linux's mess of libraries and mixers
Various hardware-specific drivers (primarily for my T410):
- Intel graphics driver, with acceleration so Wayland isn't slow
- Realtek wireless driver, with WPA2
- Specific drivers for the Thinkpad itself (or just acpi?)

Roadmap

Currently, I have a kernel capable of reading its multiboot parameters, which is terribly un-useful. The current, ordered, plan of attack is as follows:

Finish James M's tutorial (second half), which covers:
- Paging
- Heap
- VFS
- Initial RAM Disk (except I'll probably use my own format for the directory structure)
- Multitasking
- User mode
Finish basic kernel functionality
- Loading ELF binaries and executing them in user mode
- Complete system call table
- Get a better semblance of users and groups
Write a file system driver for a real file system
- Target is EXT2, but might do FAT
- Move OS development images to some form of virtual drive (VDI or something QEMU compatible)
Implement a VESA mode handler
- Requires a Virtual 8086 monitor
- Need to be able to use graphics modes and still have output, so write a framebuffer terminal
Complete libc
- Enough to run basic unix tools...
Port some basic UNIX tools
- a shell (bash and zsh, because I like bash, but the office uses zsh)
- ls, mv, rm, etc.
- here's a real test: perl
Implement networking
- IPv4
- Ethernet driver for QEMU or VirtualBox

Anything beyond this point is part of the 'Loftier Goals' section

Wayland compositor
- based on specifications for a Wayland environment
- Port some of the Wayland sample applications
- Write my own!
Port Qt
- Specifically, for Wayland
- Qt is huge and has its own standard library, might need more extensive libc
- Need to support C++-built stuff, so will probably need a C++ stdlib.
Audio drivers
- Maybe before Qt?
- Should support software mixing at least, hardware under a virtual machine, maybe my Intel hw

References

I'll be more detailed here eventually, but for the most part, I have been using:

4.2 KiB Raw Blame History