Ship
Pending review
Devblog 18
So close to going over 10h however I managed to fix the bug.
So a missing feature has been writing to the disk and not only read it. But the current file system was fat16 and the attempts I have made to implement writing has failed.
What is the logical thing to do in this scenario, well I wrote my own filesystem: NeoFS. It is not feature rich compared to fat16 but that was why I could make it write.
And the performance may be horibble when having larger files. But if it works I will keep it!
So the test just shows writing/reading of a file.
– WilGulf
Devblog 17
I have developed my own ramfs.
The old filesystem only supported fat16 with only reading capabilities which maded it feel very limited with what you could do with it.
So the first step I dicided to do with adding reading was to implement filesystem that lives in ram which is really easy to write to.
Now the D: drive is used for the ramfs (D as in dev used by linux). In the screenshot test.elf creates a file: D:/test.txt and writes HELLO FROM RAMFS to it. Then read.elf (linux cat) can read that file.
– WilGulf
Devblog 16
I wanted more programs for NeoOS but I felt really limited to just the print and getkey functions. So I rewrite ncurses (not with all ncurses functions) into my own curses for NeoOS.
This allowed me to write programs that update the terminal like a screen. Here I shouw the NeoOS logo being displayed with a CRT-screen effect.
Now I am In the decision of what game to make with this new library.
– WilGulf
Devblog 15
I have now added colored printing that I showcase here in the fetch program.
The printing made me redo how the writer works to support the decoding of the default ascii sequence used to change color: \033[xxm , where xx is the color.
It follows the standard C way which makes it pretty easy to use colored printing.
– WilGulf
Devblog 14
I have worked on multiple commands for the shell but the one that i want to showcase now is the process management.
Running.elf lists all running processes (including itself) with their name, pid and privilege level.
Then I execute kill.elf with “1” as a argument which is the id of the shell. That kills the shell which leaves the os with just the blinking underscore cause there is no longer a shell running.
Other commands are the sysinfo (uname) and updating fetch.elf with some new info.
Devblog 13
Not much to show on the screen, however in the background i have worked on a security feature.
So it is kind of like the pledge feature in open-bsd: so the kernel first need to know from the program what it wants to do, like print to the screen. Not until it has told what it will do it won’t be able to do anything. And if the program ever does anything it hasn’t told it would do the kernel will just kill the process.
The program promise to not use any other functions except the one it has specified. And the kernel won’t let it break it’s promise. That is why the c function is called promise.
–WilGulf
Devblog 12
A shell
This is the first shell for NeoOS, it only has the clear command and a fetch program to display the logo. It took some time implementing all syscalls it needed, i needed to make many kernel functions available in the userprogram and that took most my time.
I have yet to implement a way to terminate the program so now when fetch is called it never ends and you never get back to the shell again.
– WilGulf
Devblog 11
NeoOS has now got ps2 keyboard drivers!
So now I have implemented a driver for the US layout on ps2 keyboards. I am also looking into changing it to support my keyboards (Swedish layout), but that would take some time to map the right keys to the right number when the US layout can be found right on the os dev wiki.
So this keyboard writing is also done the real way, the kernel boots, the kernel starts a program, that program asks for keyboard input, when it gets a input it asks the kernel to print that input.
Now you can see how this could later change into a shell: you write the name of a command and it runs it.
–WilGulf
Ship
Devblog 11
First i completed all information on linux, but it isn’t much to show. Livefetch on Linux now works like on macOS.
But as macOS got a homebrew tap I though Linux needed it’s own. I found a tool called “nfpm” that with one configuration can build packages for multiple distros, really recommend it!
But maintaining multiple package repositories looked like a lot of work so I then programmed my own shell script that takes the latest version of livefetch and publish it to all repositories.
Took some work but it is now easier for anyone to install livefetch. In the image you can see livefetch being installed with apt on debian.
– WilGulf
Devblog 10
Now a binary file is able to be loaded AND executed.
So from that qemu picture you can see the “Hello from a binary program” message, that message is printed by that print.s code.
This is a huge milestone. There are still some functions to implement. But after that it is just writing all the userland programs.
I must say none of this was easy, and choosing to develop a OS for stardance has so far been a really big challange. And now I need to push through coding the last bit so i can get this shipped before the end of stardance.
–WilGulf
Devblog 9
I found a course for os development to follow and it really helped me with this.
With the course explaining, I got my kernel to read a fat16 disk and print the content of hello.txt.
This was a lot of debugging, first the file did open but couldn’t be
read, then later the file was read but the data was just wierd
characters.
Now this is a huge step, the reading is done and now I only need it to be able to understand executables and read them.
– WilGulf
Devblog 10
I have been working on some small improvments that togther can add up and make livefetch a bit easier to understand and use.
The help message has been updated and improved. Then i have this new option that prints the available modules for your system. I have also been working on the README to be up to date and more useful.
I have also been working on getting livefetch to the aur, but the website hasn’t been working for me. Trying to register a account on the aur just gives me a 503 error. I hope that gets fixed so next ship can be installed using yay.
–WilGulf
– WilGulf
Devblog 8
The real malloc and free is implemented. And if you know C you may understand what is going on in the image. All ptr are allocated using malloc and the first one is free’d before allocating ptr4, that is why their values are the same: ptr4 reuses ptr1 because ptr1 is free’d.
Now that I don’t only have the allocation but also freeing which means I can safely implement a filesystem driver without worrying about the memory overflowing.
– WilGulf
Devblog 7
I got to usermode, there I could run all the programs like a shell. But when jumping to usermode you need somewhere to land, some address where the cpu can start executing code. I don’t have that.
So the next thing would be to implement a filesystem driver to read a disk and execute an executable. But my memory management is to simple to do it in a good way, I can only allocate and never delocate.
So now I am going to revamp my memory management.
–WilGulf
Devblog 9
I didn’t have access to my linux machine, so I needed to work on something else. And this is a large update for livefetch.
Livefetch now has a –spin-logo true option that spins the logo around it’s y-axis. It took a while to get the math right and place all the ascii characters at the right place and right Z-level.
Sometimes the logo disappeard, sometimes the color was wrong, sometimes the logo didn’t even show.
But now the spinning works, at least for the logos I have tested it with.
–WilGulf
Loading more…