when to start writting my own os

Hey Warmaster199, I just checked out your bootstrap ZIP file. I looked at the code and all the comments in it. Don't forget to credit me for the turorial bootstrap that you obviously used in it ok guy? I mean hey, it's not everyday that I see my work in someone else's code.. LOL

On another note, I am sitting here trying to peace together CTOS. And for the life of me, I don't have all the peaces it seems. I have a bunch of icons and BMPs that your text files say will utilize it. I have the boot.bin (ASM file) that you altered from my bootstrap tutorial 1, and I have a mouse.c file. I looked it iver, I cannot figure out how to get it to work with your boot.bin file. I wanted to see your OS working with full mouse support. Is it ready yet? And can I have the source? I wanted to add something to it that I think you might like. LCARS. I wanted to see your work combined with mine.

Let me know what you think. I'm anxous to find out what your response if here..

It is possible, but I haven't compiled it yet. It's just a whole mess of code. Individual parts work, but the thing has not been tested as a whole. I couldn't get GCC to work. As for that bootstrap file... It has changed a bit. It is supposed to load the start-up code which initializes 32-bit protected mode, loads the kernel, and then jumps to the kernel's address.

Right now, I don't think that I should open source it. I am planning to start a company in the future, and probably sell CTOS, as well as games and utilities for CTOS and Windows and such. And of course... No one except me has all of CTOS's code. There is about 3MBytes of PURE TYPED C CODE(with some ASM)!!!

The bootstrap file is actually quite simple now that I think about it. The only tutorial that I used was, I believe, your second one??? It just prints a message... Now the bootstrap prints the start message, sets up the BPB, DPB, and opens the next sector on the disk(startup code). I will put your name in the comment for the boot code... But right now, I really got to go(In a hurry)

Well then could you do me one BIG favor then? I'll credit you for your work in my OS as well. I need a SIMPLE bootstrap that turns the mouse on and I can move the mouse around. Can you do that at least for me?

Thanks guy.

Knight Vision

That is ok. I do not need credit for your OS.

That code is in C. Isn't your OS in PURE assembler? You can't put C in assembler, but you can put ASM in C code. You need GNU C Compiler(GCC) for that code. To make it work, you need a put pixel function, and a set pixel function. Then, you need to call the mouseinterrupt function in that file every time you want to update the cursor and draw the mouse. The mouseinterrupt function automatically calls functions to copy the background, and draw the picture. When you do the mouse interrupt again, it places the previous background back in it's place(Clears the mouse cursor away), gets the next background where the mouse is to be drawn, then draws the pictures. It does that everytime you call it. This is basically what a mouse driver does, so you can call the file I gave you a mouse driver. The thing is, your OS will have to get a little more complicated. You will have to access the disk to load another cluster or two for more code space. There's actually some good tutorials for this on that www.nondot.org site.

Try to convert that code to assembler(Might be hard, but it'll be easier for you to implement because your code is assembler, right?)

I went and got the GNU C compiler. If you send me the whole working code from bootstrap to mouse I'll compile it. Also, could you add the compiled BIN to the ZIP as well? That way I can see what it does BEFORE I compile the code on my end? In that way, i'll know if I compiled it wrong or something.

Thanks for your help. My code is in Pure assembler, but i'll port it over to C if I can just get the mouse going.

Hey i have been reading all these posts and I am really getting interested in this OS creating. I have been wanting to know how it is structured and how to go about it for a long time. Right now all I know as far as programming is VB and very little C++. How do you program assembly. Could you give me some tips. Please email me


Jayman

Thanks

Jayman

Start with reading the FAQ here.
I would be more interested where I can find information on how an OS must be structured etc. Simply how to create a very simple OS (say just a prompt where the only working thing is an echo instruction).

I could make a simple command line like that in a maximum of 50 lines(C/C++).

I have basically done that, but in graphics mode(You need your own font!). All it does is get your input, then draw the corresponding font character on the screen in the specified area(It confines it like it's drawing to a window of some sort)

Here is what I was talking about above, but it's got a window drawn behind it and stuff some it looks good. It's 671 lines of C code(NOT INCLUDED - Classified CTOS System Console code). This is a lame start to what the real CTOS console will look like... With about 30 or so more lines I could use pointers and be independent of that BGI file included, but I'm a little lazy right now that it requires making offsets and stuff because 640x480x16 color is a 4-bit mode. I could do it in 320x240x256 in about 15 more lines, but this is just a simple test.

Description of CTOS windows: There are 2 different titlebars, each of which have an active and non-active (zorders taken into account). The small buttons in the top left of the windows are as follows: A blue triangle facing left for minimizing to the taskbar which is on the LEFT side on CTOS, A pin button(black) to pin/unpin the window to/from the VDesktop(Pin makes window follow desktop, un-pin makes window follow screen), a Maximize button(Full screen - black symbol), and a window close button(a red 'X')

Please check the following file: Extract into any directory, run constest.exe. Type and see what happens. Press [ESC] to quit.

That's cool. Where did you learn to make such things?

If you want to learn about OS structure a good book is

Operating Systems, Design and Implementation

ISBN:- 0-13-638677-6

Quote:

---

Originally posted by CornedBee
That's cool. Where did you learn to make such things?

---

Where did I learn it? A few documents for the fonts, on the net but that's it. The docu's are just something I came across. Everything else is just my programming skill :). If only I knew enough assembler to do that in it...

I actually mean where did you learn to write e.g. the OS loader? The "classified" part :)
Anything on the net? I'm interested, but not that interested that I would buy a book.

But you're writing an OS and therefore you must somewhere have learned how to do it.

All it is is different numbers and address locations and signals. For the low-down stuff like outportb and inportb(Output and input to ports respectively), I found those in a file that comes with DJGPP the GNU C Compiler. Those are useful for writing / getting values from different ports on your computer whether it's a hard disk and a sound card. Alot of that information can be found on the links to sites that are given in this thread. A really good site for learning drawing algorythms is http://www.brackeen.com/home/vga

There you will find setting modes, drawing pixels, drawing lines, circles, etc... Another good site for finding information on the VESA video modes is http://www.VESA.org

Basically, the information your need is just port values and the meaning of information that you send to the ports.

I can crash the contest project when I press backspace about 100 times...

I see... I've got to fix that. But right now, I am almost finished adding IDE Hard Disk support. The code allows for up to 4 hard disk controllers, which means that you can have 8 IDE drives. Imagine 8x 100GByte drives... The problem is that my code MAY only support up to 8GBytes. I'll look into it later on today. As for that error you found, I know what the bug is, I'll fix it later when I start working on the GUI again (At least I'll make an effort to fix my bugs, unlike Microsoft :) )

are there any commands for the console or just typing?

I coded that in about 30-45 minutes. I didn't include any commands yet, but I will AFTER I get the GUI running(Ability to move windows with the mouse, click buttons...). Most of that coding time is thinking on how to draw a good window and figuring out a way to use my font array.

To answer your question: No commands are supported yet. Why is everyone making such a big deal of this little thing? :D :rolleyes: I was just a simple report on what the O.S. will look like... Oh well :p

Ok, I gtg 2 school now (BTW, my programming teacher says that that console is amazing. Some of the code in there she didn't understand...)

it does look cool, a little too much like windows but cool.

Hey Guys

Though i would add something to the post. I amd trying to write an OS and so far have written a very basic bootloader and laid down the concepts behind my file system, memory managment and I/O allocation. Anyway what i was wondering is does anyone know where i can get a 64bit assembler and 64bit emulator that will run under windows as i want to develop a 64bit OS.

Peter

A 64-bit O.S.? Hmmm... You won't be able to run it on any Pentium chips... You'll need a chip made by SGI such as the R4000(The chip inside the N64). I believe that a SUN HyperSPARC is also a 64-bit CPU... You may possibly make your O.S. for the Macintosh G4 cube... But I think that's 128-bit... Oh well... good luck...

One last thing: a 64-bit emulator would be VERY SLOW. One for the N64(SGI R4000 - a SIMPLE 64-bit chip) needs a P3 @ 450MHz.

phew! I've been keeping up with this thread and I understand
nothing! I made a boostrap b4 but looks like its ant meets
elephant! You guyz are really good in ASM!
I bet u guyz could push down M$!

Anyway, I want to know how you guyz learn all this?
I keep on reading the tutorials but still not understand...

Au contraire(sp?)... I don't really know that much about assembler, I just know some of the syntax and some of the opcodes... But, I do know lots of things in C. My C++ teacher calls me the top in the class and some of the things I tell her in C++ she didn't even know was possible. Example:

struct xy{int x, int y}

struct xy myfunc(int x, int y)
{
struct xy temp;
xy.x = x, xy.y = y;
return xy;
}

She did not know that it is possible to return a structure from a function. I believe you can do this in ASM as well, but you have to say word x, word y(2 byte = one word). Long is double word... I think that some machines such as some MACs support quad words. Anyways, for CTOS, I got the some of the IDE interface ready. It should be able to identify the geometry of different drives and read and write to the drives. I can write to 4 different controller port for a total of 4 controllers or 8 IDE Hard disks. Today, I started looking up information on the FAT12/16/32 filesystems(FloppyDOS, HDDDOS, and HDDWINDOWs, respectively). I have already coded structures for holding the Boot Parameter Block, the FAT12/16 BPB at offset 36, the FAT32 BPB at offset 36, the FATInfo sector, and the 32byte Directory/file entries. These are just simple structures taken from a document taken from the OS heaven, http://www.nondot.org/sabre/os

All I have to do is load sectors using my hard disk routines and use the info from the BPB to access files by reading sectors at calculated offsets given by the FAT Info. It's not the most difficult thing to do, but it IS possible.

I looked into that. Up to 12 bytes, the return value will be in ecx:edx:eax
After that the compiler automatically creates a storage area in the calling function's stack and passes a pointer to this area.

I can only help people who have VB :(

VBTZH:

Learn C. It's a bit more difficult than VB (Actually it's more than a bit...), but you get pointers(direct memory access... Raw power and speed... <drool> ). Your programs will run faster(lots faster!), but they take longer to code... unless your a code guru like some people I know(I could POSSIBLY be a C Code guru). You can also hit the hardware directly and manipulate data quicker and easier.

Take it a bit further with assembler and then your programs run from 2 to 20 times faster than in C, and the executables are super small (A 'Hello world prog in C: 4KBytes. Hello world prog in ASM is about 100-150 BYTES). Some instructions are even faster! The big draw back of ASM is that it'll take a while for you to code, and you need lines upon lines just to display a bit of text... Basically, if you need tons of speed and need to do ultimate direct hardware access and are short on space(Bootsectors - big time! Short on space...), do it in ASM!

Hey Guys

To all the people who have started writing OS's im curious what ideas have you come up with for memory managment. Have you just stuck to the standard shove a program in where theres space or have you com up with better ways of dealing with it.

I had an idea for memory managment put i think it would require big processor overhead. I was thinking though processors are getting ridiculously powerful now so this may not be a problem anymore. What do you guys think?

Basically my idea is to have a program stored all over memory anywhere there is free space to put a bit of it. This would allow for 100% memory usage the problem is whenever the program jumps or gets to the end of one of teh blocks the processor would have to check to make sure and then redirect the jmp or tell teh program where its next line of code is. Do you reckon this is worth implementing or would it just be too slow.


Peter

I don't really think that that method would be too efficient. Even with fast processors that we have today, they will be greatly slowed down because most of a processor's actions deal with memory accessing, whether it be adding or whatever, you still need to access memory. Having bits and pieces of a program lying all around your whole memory space is almost as bad as the FAT filesystem. Sorry to destroy your idea like that, it seems good at first(Good waste management), but analyse it and you will see the great dissadvantage(Possibiliy that a program could get lost, and great speed loss). One more thing, if you actually decide to implement this idea, look into "Garbage Collecting" algorythms for memory management.

One last thing of advice and then onto a big lecture: Use alot of malloc(), realloc(), and free(). If you didn't make these, then do it! These are important. Use malloc in your MM to make space for the programs, and use realloc for re-organizing to make more space. These are dynamic memory management techniques and are very important.

Think of memory as a type of filesystem(Hard disks are non-volatile, slow[not any more], magnetic storage). The programs are your 'files' in this analogy. You have tables and descriptors to tell the processor where everything is. But what will be faster? A filesystem that spreads files all over and fragments the data(FAT12/16/32 will do this alot - run defragmentor), or a filesystem that stores everything linearly(in a straight line and together) and rearrages data to resize the data(Such as EXT2[:)][extended FS v2] and my new filesystem ALFS[Advanced Lookup FS])?

My OS, CTOS, stores everything together in clumps(Blocks?). The memory manager has sort of a table to save information about the allocated data(Where and how big it is). It is possible to move the whole linked list or move just a section of it where there is more space for realloc()ing.

Actually, my memory management code DOES need work, I REALLY need to work on this(The mem.c file is about 13KB of code. The rest of the CTOS kernel takes up like 3MB. The MM code definitely needs a facelift :D ). I have the great idea, but it's just NOT FUN to implement. Later, I plan to make virtual memory and stuff using file. But, I need to get the rest of the kernel AND the filesystems up and running first off ;)

BTW: I hate DJGPP(GCC)! I can't get rhide to show any errors that I have. It just says "There were some errors" and that's it! Can ANYONE help me out? I'm going to have to do a heck of alot of debugging :(... :D :rolleyes:

Yeah i here what your syaing but surely because memory isn't a rotational medium having the program all over memory shouldn't make a difference to access speed as it just has to address the next space and not rotate round to then read it off.


As for DJGPP doesn't it output the errors to a file i migth be wrong but i think it does.


Peter

Yeah... But it is slower. The processor needs to compute the next part of the program given the packet/block information that you have in your descriptor for the program. This could mean an extra memory read cycle and add / subtract cycle for EACH access. usually memory can be accessed once a clock cycle. The jumping CAN double the time needed to access SOME piece of program in memory(Moving to different packets that are computed).

One more thing: I didn't know that you even could rip up programs and put the parts into memory in totally different places. Oh well... maybe windows does it (that COULD explain some of the crashes[losing address locations and such])

Your method does sound like it could be a neat experimental thing. Code it up and see it. Make a timer for normal mem accesses, and a timer for your method. Move around huge quantities of memory then stop the timer(do for each method). If the time isn't vastly different, then try it for your OS :)

Thats a very good point i completly forgot that the adress of the next bit would be stored in memory so would have to be accessed.

I might still code it up though it would be interesting to experiment and see what its like.


Peter

Well, I've made a bootstrap.
Now I am starting the kernel from scratch.
I wanna start with file management.
How am I to create and access FAT32?
Then later on I'll just have to pass the ASCIIZ and the mode I am
opening it,ie text,binary,random...

Or any advice is also welcomed.

FAT32... I need help with that as well... I sort of get some of it, but I don't exactly get how it all fits together. I think I'll review my documentation again. Go to...

http://nondot.org/sabre/os/filesystems

...and click on anything that deals with FAT12/16/32. When you make your code, don't just right away open the drive as a FAT, as you might get errors. First, open the partition table on the drive to see if it actually is a FAT drive. Here: code from...
CTOS/drivers/block/ide.c

Quote:

---

void ata_partProbe(void)
{
struct partition part_entry;
unsigned char buffer[512], drive, part;
unsigned char lheads, heads, sects, Temp3;
unsigned int Temp, track, Temp2;
unsigned short offset, cyls;
unsigned char* cp;
unsigned int i;
unsigned char str[255];
unsigned int d_cyl, d_head, d_sec;

printk("IDE Disk Partition probe: ( )) <--- \n");

for(drive=0; drive < MAX_INTERFACES; drive++)
{
if (ide_interfaces[drive].detect == 0) continue;
if (ide_interfaces[drive].flags & ATA_FLG_ATAPI) continue;

/* load sector 0 (the partition table) - 1 block, 1 count - 512 Bytes */
Temp = ata_read(drive, 1, 1, Buffer);
if (Temp < 0) continue;

if (buffer[0x1FE] != 0x55 || buffer[0x1FF] != 0xAA)
{
error("Invalid parition table on IDE Hard Disk device %d! Recommend you format it",drive);
continue;
}
lheads = ide_interfaces[drive].max_head;
for (part=0; part < MAX_INTERFACES; part++)
{
offset = 0x1BE + 16 * part;
Temp=*(unsigned int *)(Buffer + Offset + 8);
if (Temp == 0) continue;
sects = Temp % ide_interfaces[drive].max_sec + 1;
track = Temp / ide_interfaces[drive].max_sec;
heads = track % lheads;
cyls = track / lheads;
Temp2=*(unsigned int *)(buffer + offset + 12);
Temp3=*(unsigned char *)(buffer + offset + 4);
switch (Temp3)
{
case 0x01: sprintf(str,"FAT12"); break;
case 0x04: sprintf(str,"FAT16"); break;
case 0x05: sprintf(str,"Extended"); break;
case 0x06: sprintf(str,"FAT16 >32MB"); break;
case 0x07: sprintf(str,"HPFS"); break;
case 0x0B: sprintf(str,"FAT32"); break;
case 0x0C: sprintf(str,"FAT32"); break;
case 0x0E: sprintf(str,"FAT16"); break;
}
printk("HDD%d%c: start LBA=%d, start CHS=%d:%d:%d, Size: %dMB %s", drive, 'a' + part, Temp, Cyls, Heads, Sects, Temp2 * 512/1024/1024, str);
drive_para [drive * 4 + 4 + part].start_block = Temp;
sprintf(str,"HDD%d%c",drive,'a' + part);
}
}
}

---

I condensed some of that code so that it's not too ugly(removed some comments). The code above basically checks all the IDE hard disks and asks them for their first sector(boot block) for the FS IDs. When making your boot block, be sure to include the FAT identifier table. I'll show you people the CTOS boot block as an attachment :)

Have fun! (It's actually a "Stapled together" mess of boot blocks, but just take the table at the top of the file and use that.

I'd like to use OUT...:)
I see your C code...
Shouldn't the first sector be the boot sector?
Why it is the partition table?
I thought it's after boot sector...

The code reads block 1 (CHS = 0/0/0). This block the first sector and it contains the boot sector(AKA boot block). The drive's partition table is INSIDE the boot block's code.

Yes, I see what you mean.
I am not quite familiar with C codes, because I seldom use them,
though I know C fairly well.
Sometimes I feel C is more difficult than ASM. :)
I am no way good in FAT.
So, my naive thought is:I want to write couples functions that
will receive an ASCIIZ string and do all the hard stuffs.
Compatible with MS-DOS.

Uh,never mind. I'll check out www.freedos.org

New update for my OS. I have 8MBytes of stuff in total, 2.4MBytes are pure code for the OS - 2 years of data by just me(I haven't really worked too much on it the first 6 months and these last 6 months).

I have designed the GUI, but have not coded it(It's all on paper, in my head, and in bitmaps on my computer). This is going to be one big chore. Here comes another 150,000 lines of C code(The GUI) to double the size of my OS.

Check out these pictures I created (I already made algorythms to draw the alternating patterns on the windows - get a closer look). I like my ideas, WinXP looks like M$ collided with Fisher Price or something.

This is what CTOS will look like in possibly 2 or 3 years:

Hi,
Interesting reading on this thread. I have a couple of big picture questions that are not so clear to me.

I understand the CPU calls an address that points to BIOS. BIOS does it stuff and then searches for Boot Record....

Now, when you folks develop a bootstrap, are you talking about programming the BIOS...or programming a bootable sector, MBR, etc. so that the CPU has someplace to go after BIOS program?

Do you test your OS on the PC you use to develop it?

I imagine you do something like this...using Windows/DOS, editor, compiler, you build operating system software and compile it....then you place it somewhere...boot disk? Then you do a warm boot and then test it. Is that right?

Please, I need to be clear on the basic process you folks follow. I understand software and hardware pretty well but I have not done this OS thing before.

Regards,
ChuckB

The first thing you generally do is program you Bootsector, this is the first sector on a disk. This is loaded at 0x07C0 in memory by the bios after it preforms its POST checks. It is the resposibility of the Bootsector to call the second stage of the bootloader or go ahead and load the full OS. It is quite possible to test your OS on your development PC but you really should get an old 486 or something to test it on. All you do (atleast untill you got a bootloader on a harddisk) is write the bootsector to a floppy disk using debug (the dos program) then copy the other files to disk (atleast if you are using a FAT filesystem).