On The Theory of Interrupts

[Nikita]

Is it possible for an interrupt to change what address CPU returns to after the "nmiret"?

I'm trying to figure out how to make a thread scheduler.

[Black Phoenix]

Is it possible for an interrupt to change what address CPU returns to after the "nmiret"?

I'm trying to figure out how to make a thread scheduler.

The correct way to do thread switching is to simply switch stacks (returned address is stored on current threads stack - if you change stack to other threads stack, it will return there instead)

[CCFreak2K]

Is it possible for an interrupt to change what address CPU returns to after the "nmiret"?

I'm trying to figure out how to make a thread scheduler.

What you do is store ESP and SS separately, such as in an array in kernel-space. When the scheduler interrupt is called, you store ESP and SS, then restore ESP and SS for the thread that should run next and return from the interrupt.

[Nikita]

I'm trying to figure out how to get NMI working. So far I've done everything by the book:

stef
jmp Init
IntTable:
alloc 1024
Init:
sti
lidtr IntTable
mov eax, IntTable
add eax, 128
mov #eax, Interrupt32
add eax, 3
mov #eax, 64
jmp Main

However, when I send it the value of 32, it gives me the error 32, while the main code keeps running fine. The interrupt function at Interrupt32 label was not executed. Could someone point out what could I possible have got wrong there?

[Black Phoenix]

Try 96 instead of 64

[Nikita]

Thanks, that works.

Hmm, maybe you could include binary notation to the next version of the assembly editor? 0b1100000 might be handier to write in cases like this one...

An update to the documentation (and this thread's OP) would be nice too!

[aqez]

So.. I've been messing with cpu for a little bit, and understand asm thoroughly, but I'm having trouble with nmi's. For my example, I've got a cpu wired to an address bus that has a keyboard and a console screen. I also have the NMI output from cpu going to keyboard, and when I press the 'a' key on the keyboard, it triggers interrupt 97. I've got code in place to handle the interrupt, interrupts are enabled, and the code returns with nmiret, but for some reason the code isn't being called, and the cpu is locking at the interrupt, with the nmi number as the error output(obviously). So, thought maybe you gurus could take a look at the pieces of code I'll post here, and see what I'm doing wrong.

Near the top of my data section we have..

_interrupts:
    alloc 1024

A little bit further down I have an example interrupt handler, for this example 'a' pressed on a wired keybooard:

_interrupt_a:
    push eax // I push the registers onto the stack before
    push esi // each call's main work, to preserve whatever
    //may have been in the registers
    cmp #ConsoleCurrentChar, #ConsoleLastChar
    cge _screenshiftup
    mov eax, #ConsoleCurrentChar //Current console character
    mov esi, _a // _a is a pointer to the character 'a' in memory
    mov #eax, esi
    inc eax
    mov #eax, #ConsoleParams // Obvious..
    inc eax
    mov #ConsoleCurrentChar, eax
    sub #ModChar, 2
    cmp #ModChar, 2
    cl _resetmod //Dont worry about this, my way of keeping track of line endings
    pop esi
    pop eax
nmiret //Return from nmi

And a bit further down in my interrupt table initialization function..:

//------------------------------
// Function: Interrupt Setup
// Purpose: Set up interrupts
// Notes: 
// Should be called very soon 
// after boot.
//------------------------------
_interruptsetup:
    
    
    lidtr _interrupts //Tell cpu where interrupt table begins
    mov edi, _interrupts
    add edi, 388 //Move to interrupt address
    mov #edi, _interrupt_a //Move address of handler into the table
    add edi, 3
    mov #edi, 96
    
    inc edi
    mov #edi, _interrupt_b
    add edi, 3
    mov #edi, 96
    
    inc edi
    mov #edi, _interrupt_c
    add edi, 3
    mov #edi, 96
    
    ret

So for the life of me I can't figure out why this isnt working. I'll say it again since I didn't include it in the code: protected mode and interrupts are turned on.


Edit: So, I've figured out it IS calling the interrupt, its just taking forever to do so.. and once done with this, it quickly alternates between error 2 and 4.???. ??? being digits I can't identify because it swaps out so fast. So that brings about two more questions: Why does it take so long for the interrupt to be handled, and why does it alternate between errors once completed?

[Xandaros]

The errors are 2 and 4. 2 means "End of Program execution" and 4 means "Unknown Opcode".

Have you checked for compilation errors? (Console after uploading the code to the CPU)
You probably mistyped somewhere in the code. (not necessarily in the part of the code you have posted, it could be anywhere)

Fix Error 4 first, it might fix 2, too.