Handle Interrupt on x86-64 kernel with zig
On x86_64, if we want to handle interrupts, we need to define IDT(Interrupt Descriptor Table), it has 256 elements.
And we also need to set IDTR(Interrupt Descriptor Table Register). In this article, we won’t discuss setting up IDT, that would be too long.
We just discuss how to handle interruptions gracefully, pay attention, what we are talking about here is the interrupt handling function, in other words, what you want to fill in the ide (it only fills part of the idt).
Tradition
Let we look at osdev(which is the richest os development website), the url is this: https://wiki.osdev.org/Interrupts_Tutorial.
In osdev, it will tell you to do this, using intel assembly syntax with macro feature:
%macro isr_err_stub 1
isr_stub_%+%1:
call exception_handler
iret
%endmacro
; if writing for 64-bit, use iretq instead
%macro isr_no_err_stub 1
isr_stub_%+%1:
call exception_handler
iret
%endmacroand this code just is a simple handle creator macro, it not store general register status, very unsafe and elegant.
With zig gracefully
For handling interrupts more precisely, we need to understand what the CPU does when an interrupt occurs, you can see this:{% post_link interrupt-function [interrupt-function] %}
Then we use feature Naked function in zig, it will throw away the function calling convention, usually used on inline assembly.
So we can we can get the following function:
fn generate_handle(comptime num: u8) fn () callconv(.Naked) void {
const error_code_list = [_]u8{ 8, 10, 11, 12, 13, 14, 17, 21, 29, 30 };
const public = std.fmt.comptimePrint(
\\ push ${}
\\ push %rax
\\ push %rbx
\\ push %rcx
\\ push %rdx
\\ push %rsp
\\ push %rbp
\\ push %rsi
\\ push %rdi
\\ push %r8
\\ push %r9
\\ push %r10
\\ push %r11
\\ push %r12
\\ push %r13
\\ push %r14
\\ push %r15
\\ mov %rsp, context
, .{num});
const save_status = if (for (error_code_list) |value| {
if (value == num) {
break true;
}
} else false)
public
else
\\ push $0b10000000000000000
\\
// Note: the Line breaks are very important
++
public;
const restore_status =
\\ mov context, %rsp
\\ pop %r15
\\ pop %r14
\\ pop %r13
\\ pop %r12
\\ pop %r11
\\ pop %r10
\\ pop %r9
\\ pop %r8
\\ pop %rdi
\\ pop %rsi
\\ pop %rbp
\\ pop %rsp
\\ pop %rdx
\\ pop %rcx
\\ pop %rbx
\\ pop %rax
\\ add $16, %rsp
\\ iretq
;
return struct {
fn handle() callconv(.Naked) void {
asm volatile (save_status ::: "memory");
// interruptDispatch just is a higher level interrupt handling function
asm volatile ("call interruptDispatch");
asm volatile (restore_status ::: "memory");
}
}.handle;
}we use the comptime feature of zig to generate a inline assembly function, the error_code_list is an array of EXCEPTION which has error_code.
So we assume that the top layout of the stack is:
bottom of stack
--------
SS
------
ESP
------
EFLAGS
------
CS
------
EIP
------
Error Code
------
Interrupt Number
------
General Purpose Registers
--------
top of stackWhen Interrupt number not has an error code, we will just push a fake error code to stack, here is $0b10000000000000000.
Then we will assign RSP to context(a structure pointer), then call interruptDispatch to execute more specific function logic.
After the execution is completed, we try to restore the original register context, so we pop most register, and use iretq to return to the original place and continue execution.
Why add 16 to RSP, because of maintaining the calling convention of iretq.