Ghost Kernel

Since the beginning, Ghost was implemented solely for x86. Within the last weeks I made efforts to port it to x86_64.

With this change the system is also migrating from GRUB2 to Limine since I found its time to switch to something more modern and Limine looks like a great choice. This now also brings UEFI support to the project.

The changes for the port are now on master. The last major bugs with SMP support have been sorted out and now I felt it's time to merge.

The loader that was previously responsible for loading the kernel binary, after it was loaded by GRUB itself, was removed since it is no longer required in the new setup. This will allow me to clean up the codebase much more.

The port also increases compatibility with real hardware, especially since the EFI framebuffer is much more reliable than the old virtual8086 VESA setup:

For a long time the system only supported ramdisk access in the virtual filesystem. The reason was mostly that for building the remaining system architecture, that was sufficient. In order to be able to install the OS from the Live ISO onto a hard disk, this is obviously a crucial feature though.

The plan is to create separate PCI, AHCI and filesystem drivers as well as a general disk management service to take care of which device is managed by whom. All of them will be privileged userspace processes.

In previous versions, there was already a userspace protocol that would allow a process to register as a filesystem driver and provide a mountpoint. This communication is mostly via the means of messaging, while for actual file reading and writing shared memory will be the preferred way. In the course of this implementation this functionality will be restored.

For a very long time now the memory allocator that was in place on kernel level was pretty crappy. When implementing this first it was one of the "I'll do that properly later" pieces of code that then are never touched again since they work well enough. Well, it actually didn't and caused a lot of issues both performance and stability wise.

The allocator is now replaced by a new one that features a mix of bucket and chunk allocation which vastly improves performance for the huge amount of tiny allocations that the kernel likes to make.

Aside from that I've updated the build script to respect library dependency (currently by just defining it there) which should avoid issues when setting the project up initially.

Try the release now.

There's now a Docker image available which simplifies the process of setting up your local development system if you want to modify Ghost. Just clone the repositories, create a container and get started.

Check the documentation for more details!

After being inactive for a pretty long time (sorry :) I've recently refactored quite some parts of the system.

The wait handling has been extensively reworked so that sleeping an waiting for file I/O is way more efficient now. The scheduler was slightly improved with this change as well. I also added support for the legacy PIC again since all the code was there, so it works without the IO/APIC activated.

Many dumb performance bugs in the window server have been fixed. Here's something for the eye: