IRQ problems

[WinExperements]

Hello! I am trying to port my kernel to x86_64 and i have some IRQ handling problems.
Each time i got an timer IRQ, the stack begin corrupted after the iret begin executed, the stack pointer is restored as it be before the IRQ. I also set up the TSS and set rsp0 pointer to special stack address that is specially allocated before.

So, here is my code:
Timer IRQ handler code:

Code: Select all

scheduler_irq:
	;push qword 0
;	push qword 32
;	jmp irq_common_stub
	call interrupt_sendEOI
	iretq

The TSS setup code:

Code: Select all

struct x86_64_tss {
	uint32_t reserved0;
	uint64_t rsp[3];	// When privilege change occur.
	uint64_t reserved1;
	uint64_t ist[7];
	uint64_t reserved2;
	uint16_t reserved3;
	uint16_t iomap_base;
} __attribute__((packed));
struct GdtTssExtend {
	uint32_t base_uper;
	uint32_t reserved;
	uint32_t zero;
	uint32_t reserved1;
} __attribute__((packed));

void putTSS64() {
/*	uint32_t addr = (uint32_t)(uint64_t)liminePhys(&tssEntry);
	gdt_set_gate(5,addr,sizeof(tssEntry)-1,0x89,0);
*/
	uint32_t addr1 = (uint32_t)((uint64_t)&tssEntry & 0xFFFFFFFF);
	uint32_t addr2 = (uint32_t)(((uint64_t)&tssEntry >> 32) & 0xFFFFFFFF);
	uint64_t limit = (uint64_t)&tssEntry + sizeof(struct x86_64_tss);
	gdt_set_gate(5,addr1,limit & 0xFFFFFFFF,0x89,0);
	struct GdtTssExtend ext_tss;
	ext_tss.base_uper = addr2;
	ext_tss.reserved = 0;
	ext_tss.zero = 0;
	ext_tss.reserved1 = 0;
	gdt_entries[6] = *((gdt_entry_t*)&ext_tss);
}

Maybe i doing something wrong in the IRQ handling process

I also tried it with the actual registers save/restore code:

Code: Select all

extern stack_top
irq_common_stub:
	; Here we are...
	push rax
	push rbx
	push rcx
	push rdx
	push rsi
	push rdi
	push rbp
	push r8
	push r9
	push r10
	push r11
	push r12
	push r13
	push r14
	push r15
	mov rdi,rsp
	call x86_irq_handler
;	mov rsp,rax
	pop r15
	pop r14
	pop r13
	pop r12
	pop r11
	pop r10
	pop r9
	pop r8
	pop rbp
	pop rdi
	pop rsi
	pop rdx
	pop rcx
	pop rbx
	pop rax	
	add rsp,16	; two 64-bit values.
	iretq

And that doesn’t work too

[Octocontrabass]

the stack pointer is restored as it be before the IRQ.

That's normally how it's supposed to work. Why are you expecting the stack pointer to change?

[WinExperements]

the stack pointer is restored as it be before the IRQ.

That's normally how it's supposed to work. Why are you expecting the stack pointer to change?

I know that, I just can’t understand why the stack become have random data after the IRQ handler calls the iret. I don’t know much currently about the differences between IRQ handling in X86_64 and x86

I tried to debug it before and found nothing that can be strange, the context saved and restored correctly, but still, the function arguments where the IRQ begin fired is corrupting by some reason

[Octocontrabass]

I know that, I just can’t understand why the stack become have random data after the IRQ handler calls the iret.

Which part of the stack? If it's at addresses less than RSP, that part of the stack is the red zone, and it's getting corrupted because interrupts don't switch stacks in ring 0 (unless you use the IST). You can use -mno-red-zone to tell your compiler to stop using the red zone.

[WinExperements]

Thanks. It’s looks like I misunderstood the 64 bit TSS a little bit