Maximum PC

How is it done. And is it even possible?

Its possible, but you would have to be a geek to the nines.

There are books on the subject. The problem is that you have to start very low level code wise - you have to talk to the chip, and thats not easy. You gotta know things only the king of geeks wants to know.

Manta

First, create a bootloader. This must be done in pure assembly -- no C yet. It has to go on the master boot record. You can call code from other sectors, but you must have the beginning code there.

Then, you can create a compiler (if you create your own executable format), or use GCC if you want to use ELF. Once you have a compiler and linker working, create a kernel. This will handle all the OS functions: memory management, drivers, multitasking (optional, but not so optional these days), etc. This can (and should!) be made in C (so you should have a compiler ). Now you need to build the other parts of the OS: graphics libraries, drivers for hardware, etc.

Thats interesting I thought that the majority of the OS was programmed in assembly not C.

Several people have done it.

Read a book. I'm sorry if this sound patronising, but creating an OS is one of the most complicated tasks a programmer can undertake. There are so many issues, concepts, and problems to address before you can even begin, that a solid grounding in CS and OS fundamentals is an absolute must.

For a single person, to create a 'modern' multimedia capable OS from scratch, is several lifetimes of work.

It's possible, considering that we're typing this on an OS of some sort.

Here's a link for you:

http://mega-tokyo.com/osfaq2/

In essence, you need to have your code be able to talk to the hardware directly. First you have to sit down and write the bootloader in Assembly (x86 is very tricky, BTW. x86 launches stuff in 16-bit protected mode then you somehow have to convert over to 32-bit if you want to write for a 32-bit OS). Then you have to design a kernel, now, whether or not you want to use a monolithic design, a microkernel design, or an exokernel. Each have their own advantages, monolithic being the easiest to design but old and moldy, microkernels being middle of the road, and exokernels which give you pure performance at the expense of not using a HAL (Hardware Abstraction Layer).

There are thousands of things to do, heck, even Linus didn't write all of Linux. He just wrote what he needed to write, and let the rest of the people handle the work for him. Don't foget, you also need to have a shell to interface with the kernel for you.

Kewl Site DJ,

Thanx - that will keep gamer busy for a while (and me too if I am not careful).

On another note:
I just found a book on my shelf - Assembler for the IBM PC and PC-XT, by Peter Abel (1984). Its amazing what you find when you are moving. I am sure I never read it - can't fathom why I have it.


Manta

Merry Holidays All

w00t! Thanx!

I still have my IA-32 Assembly book, but no time to read it though.

So where those long hours of reading c++ books a waste to amke a os? Anywaze to start off i wanna make a really really simple os. ONe that is all text based and just has a text editor and maybe a telnet client. I will progress from there. Thanks for the link, and i will buy those books from you depneding on how OOD they are. Thanks,

gamer

Then you should do everything except make a graphics library and a few drivers.

You can use standard int 10h for graphics.

For a telnet client you will need to create a driver for the NIC card you use and you must create a sockets implementation .

You must manage memory and support multitasking.

Unless you plan on using a different file format, plan on making your own loader and compiler, too.

Making an operating system, even that simple, is no easy task.


I suggest modifying Linux code instead of creating a new OS from scratch.

No, it's not wasted. C++ has its place in OS Design, but so does Assembly and whatnow. You can't expect to build a house with just a hammer, you use a multitude of tools to get the job done, the same applies with making your own OS. That's just the way things are, so expect it.

BTW, a friend of mine is writing his own OS for fun. So far, Assembly's pissing him off, and even something as simple as opening a socket and sending messages to a static address is a difficult and long process. So be prepared to be stuck at home for a long time.

http://cboard.cprogramming.com/showthread.php?t=59720

Thanks...but he was going to use C for sockets, he's using Assembly to write the bootloader. I should have been more specific.

No, he was going to use assembly, but then people told him he could do it in C

yeah, it does hold some truth...more like half-truths.

Gamerfreak,

I wouldn't want to discourage anyone from attempting a programming project, but you're definitely jumping way out on a limb with this project. Why not write a couple of simple games first and then move on to maybe a compiler and finally an OS instead?

A lot of people think that about OS, games, etc. It isn't true - and hasn't been true for a long time now.

I think everyone forgot to mention that you need to develop your own filesystem as well. Perhaps an easy way to make your own operating system is to just take the core of Linux and modify it to your desire. The source code is all out there, and there's probably nothing about that core that you wouldn't want, anyway.

Don't poo poo the man... he just asked a question...

Tell you what man! Everything you wanted to know on OS's and even some stuff you didn't!

http://dmoz.org/Computers/Software/Operating_Systems/

Read, read and oh yea! Read!

check out the guides. There are several good Alternate OS lists with links to os projects. Many are GPLed so you can grab them and play with them. Look for small simple OSes like V2_OS (all done is ASM), SPIN (modular os designed at U of Utah) or NewOS (being designed by a former BeOS developer)

V2 is actually pretty cool. They have most of the hard work done so you're free to jump in and play around more... I've fiddled with early version a while back and what they had was pretty impressive.

NewOS is probabbly the most mature and "professional" it's being developed by a former BeOS programmer and he's creating a system very similar in design to what BeOS was. That's why the OpenBeOS project (now called Haiku) is helping with NewOS development in order to get a solid Kernel that will allow them to create a open source compatible version of BeOS... Fun stuff!

I never said i would actually do it..........