Hobby kernel + userspace, built mostly from scratch. Composited GUI, dynamically linked ELF binaries, networking, Python applications, and more. https://toaruos.org/
|K. Lange 31af34c73d boot: switch to tar ramdisks||2 years ago|
|apps||2 years ago|
|base||2 years ago|
|boot||2 years ago|
|ext||2 years ago|
|kernel||2 years ago|
|lib||2 years ago|
|libc||2 years ago|
|linker||2 years ago|
|modules||2 years ago|
|util||2 years ago|
|.gitignore||2 years ago|
|.travis.yml||2 years ago|
|AUTHORS||2 years ago|
|LICENSE||3 years ago|
|Makefile||2 years ago|
|README.md||2 years ago|
ToaruOS is a hobbyist, educational, Unix-like operating system built entirely from scratch and capable of hosting Python 3 and GCC. It includes a kernel, bootloader, dynamic linker, C standard library, composited windowing system, package manager, and several utilities and applications. All components of the core operating system are original, providing a complete environment in approximately 70,000 lines of C and assembly, all of which is included in this repository.
The ToaruOS project began in December 2010 and has its roots in an independent student project. The goals of the project have changed throughout its history, initially as a learning experience for the authors, and more recently as a complete, from-scratch ecosystem.
The 1.6.x release series represented the merging of a project (the "NIH" branch) to replace Newlib, the third-party C standard library employed by earlier versions of the operating system, with a new, in-house libc. With the ports of Python and GCC to this new C library completed, this project has been concluded and considered as success.
The current release series is 1.9.x, which aims to finish porting the remaining parts of older Python desktop environment to C.
dlopening of additional libraries.
The following projects are currently in progress:
To build ToaruOS from source, it is currently recommended you use a recent Debian- or Ubuntu-derived Linux host environment.
Several packages are necessary:
build-essential (to build the cross-compiler),
xorriso (to create CD images),
genext2fs (to build ramdisks),
python3 (various build scripts),
mtools (for building FAT EFI system partitions),
gnu-efi (for building the EFI loaders).
Beyond package installation, no part of the build needs root privileges.
The build process has two parts: building a cross-compiler, and building the operating system. The cross-compiler uses GCC 6.4.0 and will be built automatically by
make if other dependencies have been met. This only needs to be done once, and the cross-compiler does not depend on any of the components built for the operating system itself, though it is attached to the base directory of the repository so you may need to rebuild the toolchain if you move your checkout. Once the cross-compiler has been built,
make will continue to build the operating system itself.
You can skip the process of building the cross-compiler toolchain (which doesn't get updated very often anyway) by using our pre-built toolchain through Docker:
git clone https://github.com/klange/toaruos cd toaruos docker pull toaruos/build-tools:1.8.x docker run -v `pwd`:/root/toaruos -w /root/toaruos -e LANG=C.UTF-8 -t toaruos/build-tools:1.8.x util/build-travis.sh
After building like this, you can run the various utility targets (
make run, etc.) by setting
TOOLCHAIN=none make shell to run a ToaruOS shell (using QEMU and a network socket - you'll need netcat for this to work).
Makefile first checks to see if a toolchain is available and activates it (appends its
bin directory to
$PATH). If a toolchain is not available, users are prompted if they would like to build one. This process downloads and patches both Binutils and GCC and then builds and installs them locally.
Makefile then uses a Python tool,
auto-dep.py, to generate additional Makefiles for the userspace applications libraries, automatically resolving dependencies based on
In an indeterminate order, C library, kernel, modules, userspace librares and applications are built. Three boot loaders (one BIOS ATAPI CD loader for emulators, and both a 32-bit and 64-bit EFI loader for general use) are then built. Deployed binaries are stored in
base which is converted into an EXT2 filesystem image with
genext2fs. This image, along with the bootloader files and kernel are then placed in
fatbase which is converted into a FAT image for use as the EFI boot payload. That image is then placed in
cdbase along with shadow files representing each of the files in the FAT image, and
cdbase is compiled into an ISO 9660 CD El Torito image. The CD image is then passed through a tool to map the shadow files to their actual data from the FAT image, creating a hybrid ISO 9660 / FAT.
The kernel and driver modules have been successfully built with Clang using the
i686-elf target. If you would like to experiment with using Clang to build ToaruOS, pass
make from a clean build. You may confirm that Clang was used to build the kernel by examining
/proc/compiler within the OS.
Prior to ToaruOS 1.6.x, many third-party components were included by default (Python, libpng, zlib, Cairo, freetype, and so on). These are no longer part of the default distribution or build process and must be built manually. Complete guides for building these components are currently being drafted. The instructions for building Python are complete and available from the wiki (note that a host installation of Python 3.6 is required to build Python 3.6 and satisfying this is left as an exercise to the reader).
Freetype and Cairo have also been successfully built under the new in-house C library. When either of these are available, optional extension bindings may be built with
make ext-freetype and
make ext-cairo respectively. When the Freetype extension binding is available in the OS, alongside required Truetype font files, Freetype will be used to render text in several applications (including the terminal emulator, menus, and window decorations). When the Cairo extension binding is available, it is used by the compositor to provide improved performance.
base/usr/include, as well as graphical resources for the compositor and window decorator.
It is highly recommended that interested users run ToaruOS from virtual machines. While we have done some testing on real hardware, driver support is still limited and virtual machines provide easily tested environments where we can guarantee some level of useful functionality.
QEMU and VirtualBox are recommended and provide the most functonality. Audio support is not yet available in VMware. In VirtualBox and VMware, automatic guest display resizing is available (and a tool is available to provide similar functionality in QEMU). All three of the major VMs also support absolute mouse input.
1GB of RAM and an Intel AC'97 sound chip are recommended:
qemu-system-i386 -cdrom image.iso -serial mon:stdio -m 1G -soundhw ac97,pcspk -enable-kvm -rtc base=localtime
You may also use OVMF with the appropriate QEMU system target. Our EFI loader supports both IA32 and X64 EFIs:
qemu-system-x86_64 -cdrom image.iso -serial mon:stdio -m 1G -soundhw ac97,pcspk -enable-kvm -rtc base=localtime \ -bios /usr/share/qemu/OVMF.fd
qemu-system-i386 -cdrom image.iso -serial mon:stdio -m 1G -soundhw ac97,pcspk -enable-kvm -rtc base=localtime \ -bios /path/to/OVMFia32.fd
Additionally, a tool is available for running QEMU, under specific environments, with automatic support for resizing the guest display resolution when the QEMU window changes size:
ToaruOS should function either as an "Other/Unknown" guest or an "Other/Unknown 64-bit" guest with EFI.
All network chipset options should work except for
virtio-net (work on virtio drivers has not yet begun).
It is highly recommended, due to the existence of Guest Additions drivers, that you provide your VM with at least 32MB of video memory to support larger display resolutions - especially if you are using a 4K display.
Ensure that the audio controller is set to ICH AC97 and that audio output is enabled (as it is disabled by default in some versions of VirtualBox).
Keep the system chipset set to PIIX3 for best compatibility. 1GB of RAM is recommended.
Support for VMWare is experimental, though it has improved significantly in recent months. Optional support is provided for VMware's automatic guest display sizing, which is enabled by default and can be disabled from the bootloader menu.
Using Bochs to run ToaruOS is not advised; however the following configuration options are recommended if you wish to try it:
pcivgadevice is enabled or ToaruOS will not be able to find the video card through PCI.
e1000network device using the
sync=realtime, time0=local, rtc_sync=1are recommended.
ToaruOS is regularly mirrored to multiple Git hosting sites.
#toaruos on Freenode (
While the complete ISO build process requires some tools that have not been ported, ToaruOS is capable of building its own core userspace and kernel given an installation of GCC and Binutils, which we believe is sufficient to be considered "self-hosting". The other tools required for the ISO build are in the process of being replaced with native tools. There are also plans to build our own C compiler, as GCC itself requires a plethora of tools to compile.
As stated several times in this document, ToaruOS's kernel is written entirely from scratch. In addition to legal restrictions related to Linux's and ToaruOS's licensing terms, the very goal of the project prevents us from taking code from Linux in any way. Beyond that, the entire core userspace, all drivers, and the bootloader are also built from scratch. ToaruOS is its own project in all ways beyond the look and feel of its APIs and tools, which are based on Unix.
With the development of ToaruOS's "NIH" branch, a secondary goal in removing third-party dependencies was to make the operating system more viable for a 64-bit port. That said, the actual development of a 64-bit kernel is currently on pause while other goals are pursued. Due to the limited size of the development team, it is not feasible to continue work on the 64-bit kernel at this time.
SMP support likely poses a larger challenge as the early toy design for ToaruOS did not take into account multiprocessor systems and thus many challenges exist in getting the kernel to a functioning state with SMP.