ISR Problem

[zvoncika]

Hey everybody!
I've been trying to implement ISR for some time now.
I've been following Bran's Kernel Development tutorials.
After calling my isr_install() function, I divide by zero expecting the fault handler to be called.
But somehow, the system doesn't crash nor does it print anything.
Any idea what it might be?

EDIT: When I manually call the interrupt it works...

Code: Select all

asm volatile("int $0");

[Haoud]

Your compiler can optimize something and break your divide by zero code. Try to do a divide by zero in a volatile asm block

Code: Select all

asm volatile("Insert code here");

Can you show you code ?

[zvoncika]

Sure!
Here's asm

Code: Select all

global _isr0
global _isr1
global _isr2
global _isr3
global _isr4
global _isr5
global _isr6
global _isr7
global _isr8
global _isr9
global _isr10
global _isr11
global _isr12
global _isr13
global _isr14
global _isr15
global _isr16
global _isr17
global _isr18
global _isr19
global _isr20
global _isr21
global _isr22
global _isr23
global _isr24
global _isr25
global _isr26
global _isr27
global _isr28
global _isr29
global _isr30
global _isr31

;  0: Divide By Zero Exception
_isr0:
    cli
    push byte 0
    push byte 0
    jmp isr_common_stub

;  1: Debug Exception
_isr1:
    cli
    push byte 0
    push byte 1
    jmp isr_common_stub

;  2: Non Maskable Interrupt Exception
_isr2:
    cli
    push byte 0
    push byte 2
    jmp isr_common_stub

;  3: Int 3 Exception
_isr3:
    cli
    push byte 0
    push byte 3
    jmp isr_common_stub

;  4: INTO Exception
_isr4:
    cli
    push byte 0
    push byte 4
    jmp isr_common_stub

;  5: Out of Bounds Exception
_isr5:
    cli
    push byte 0
    push byte 5
    jmp isr_common_stub

;  6: Invalid Opcode Exception
_isr6:
    cli
    push byte 0
    push byte 6
    jmp isr_common_stub

;  7: Coprocessor Not Available Exception
_isr7:
    cli
    push byte 0
    push byte 7
    jmp isr_common_stub

;  8: Double Fault Exception (With Error Code!)
_isr8:
    cli
    push byte 8
    jmp isr_common_stub

;  9: Coprocessor Segment Overrun Exception
_isr9:
    cli
    push byte 0
    push byte 9
    jmp isr_common_stub

; 10: Bad TSS Exception (With Error Code!)
_isr10:
    cli
    push byte 10
    jmp isr_common_stub

; 11: Segment Not Present Exception (With Error Code!)
_isr11:
    cli
    push byte 11
    jmp isr_common_stub

; 12: Stack Fault Exception (With Error Code!)
_isr12:
    cli
    push byte 12
    jmp isr_common_stub

; 13: General Protection Fault Exception (With Error Code!)
_isr13:
    cli
    push byte 13
    jmp isr_common_stub

; 14: Page Fault Exception (With Error Code!)
_isr14:
    cli
    push byte 14
    jmp isr_common_stub

; 15: Reserved Exception
_isr15:
    cli
    push byte 0
    push byte 15
    jmp isr_common_stub

; 16: Floating Point Exception
_isr16:
    cli
    push byte 0
    push byte 16
    jmp isr_common_stub

; 17: Alignment Check Exception
_isr17:
    cli
    push byte 0
    push byte 17
    jmp isr_common_stub

; 18: Machine Check Exception
_isr18:
    cli
    push byte 0
    push byte 18
    jmp isr_common_stub

; 19: Reserved
_isr19:
    cli
    push byte 0
    push byte 19
    jmp isr_common_stub

; 20: Reserved
_isr20:
    cli
    push byte 0
    push byte 20
    jmp isr_common_stub

; 21: Reserved
_isr21:
    cli
    push byte 0
    push byte 21
    jmp isr_common_stub

; 22: Reserved
_isr22:
    cli
    push byte 0
    push byte 22
    jmp isr_common_stub

; 23: Reserved
_isr23:
    cli
    push byte 0
    push byte 23
    jmp isr_common_stub

; 24: Reserved
_isr24:
    cli
    push byte 0
    push byte 24
    jmp isr_common_stub

; 25: Reserved
_isr25:
    cli
    push byte 0
    push byte 25
    jmp isr_common_stub

; 26: Reserved
_isr26:
    cli
    push byte 0
    push byte 26
    jmp isr_common_stub

; 27: Reserved
_isr27:
    cli
    push byte 0
    push byte 27
    jmp isr_common_stub

; 28: Reserved
_isr28:
    cli
    push byte 0
    push byte 28
    jmp isr_common_stub

; 29: Reserved
_isr29:
    cli
    push byte 0
    push byte 29
    jmp isr_common_stub

; 30: Reserved
_isr30:
    cli
    push byte 0
    push byte 30
    jmp isr_common_stub

; 31: Reserved
_isr31:
    cli
    push byte 0
    push byte 31
    jmp isr_common_stub


; We call a C function in here. We need to let the assembler know
; that '_fault_handler' exists in another file
extern _fault_handler

; This is our common ISR stub. It saves the processor state, sets
; up for kernel mode segments, calls the C-level fault handler,
; and finally restores the stack frame.
isr_common_stub:
    pusha
    push ds
    push es
    push fs
    push gs
    mov ax, 0x10
    mov ds, ax
    mov es, ax
    mov fs, ax
    mov gs, ax
    mov eax, esp
    push eax
    mov eax, _fault_handler
    call eax
    pop eax
    pop gs
    pop fs
    pop es
    pop ds
    popa
    add esp, 8
    iret

Here's C

Code: Select all

#include <isr/isr.h>
const char* fault_messages[] = 
{
	"Division By Zero",
    "Debug",
    "Non Maskable Interrupt",
    "Breakpoint",
    "Into Detected Overflow",
    "Out of Bounds",
    "Invalid Opcode",
    "No Coprocessor",

    "Double Fault",
    "Coprocessor Segment Overrun",
    "Bad TSS",
    "Segment Not Present",
    "Stack Fault",
    "General Protection Fault",
    "Page Fault",
    "Unknown Interrupt",

    "Coprocessor Fault",
    "Alignment Check",
    "Machine Check",
    "Reserved",
    "Reserved",
    "Reserved",
    "Reserved",
    "Reserved",

    "Reserved",
    "Reserved",
    "Reserved",
    "Reserved",
    "Reserved",
    "Reserved",
    "Reserved",
    "Reserved"
};
void _fault_handler(reg_dump_t* r)
{
	if(r->int_no < 32)
	{
		puts(" Exception: ");
		puts(fault_messages[r->int_no]);
		for(;;);
	}
}

void isrs_install()
{
	idt_set_gate(0,(unsigned)_isr0,0x08,0x8E);
	idt_set_gate(1,(unsigned)_isr1,0x08,0x8E);
	idt_set_gate(2,(unsigned)_isr2,0x08,0x8E);
	idt_set_gate(3,(unsigned)_isr3,0x08,0x8E);
	idt_set_gate(4,(unsigned)_isr4,0x08,0x8E);
	idt_set_gate(5,(unsigned)_isr5,0x08,0x8E);
	idt_set_gate(6,(unsigned)_isr6,0x08,0x8E);
	idt_set_gate(7,(unsigned)_isr7,0x08,0x8E);
	idt_set_gate(8,(unsigned)_isr8,0x08,0x8E);
	idt_set_gate(9,(unsigned)_isr9,0x08,0x8E);
	idt_set_gate(10,(unsigned)_isr10,0x08,0x8E);
	idt_set_gate(11,(unsigned)_isr11,0x08,0x8E);
	idt_set_gate(12,(unsigned)_isr12,0x08,0x8E);
	idt_set_gate(13,(unsigned)_isr13,0x08,0x8E);
	idt_set_gate(14,(unsigned)_isr14,0x08,0x8E);
	idt_set_gate(15,(unsigned)_isr15,0x08,0x8E);
	idt_set_gate(16,(unsigned)_isr16,0x08,0x8E);
	idt_set_gate(17,(unsigned)_isr17,0x08,0x8E);
	idt_set_gate(18,(unsigned)_isr18,0x08,0x8E);
	idt_set_gate(19,(unsigned)_isr19,0x08,0x8E);
	idt_set_gate(20,(unsigned)_isr20,0x08,0x8E);
	idt_set_gate(21,(unsigned)_isr21,0x08,0x8E);
	idt_set_gate(22,(unsigned)_isr22,0x08,0x8E);
	idt_set_gate(23,(unsigned)_isr23,0x08,0x8E);
	idt_set_gate(24,(unsigned)_isr24,0x08,0x8E);
	idt_set_gate(25,(unsigned)_isr25,0x08,0x8E);
	idt_set_gate(26,(unsigned)_isr26,0x08,0x8E);
	idt_set_gate(27,(unsigned)_isr27,0x08,0x8E);
	idt_set_gate(28,(unsigned)_isr28,0x08,0x8E);
	idt_set_gate(29,(unsigned)_isr29,0x08,0x8E);
	idt_set_gate(30,(unsigned)_isr30,0x08,0x8E);
	idt_set_gate(31,(unsigned)_isr31,0x08,0x8E);
}

And here's the idt_set_gate

Code: Select all

void idt_set_gate(uint8_t index,uint64_t base,uint16_t sel,uint8_t flags)
{
	// Base
	g(index).base_low = ((uintptr_t)base & 0xFFFF);
	g(index).base_hi = ((uintptr_t)base >> 16) & 0xFFFF;
	// Selector
	g(index).selector = sel;
	// Zero & Flags
	g(index).zero = 0;
	g(index).flags = flags | 0x60;
}

[zvoncika]

This also fires the exception:

Code: Select all

asm volatile("mov $0, %edx\n"
"mov $0, %eax\n"
"mov $0, %ecx\n"
"div %ecx");

[zvoncika]

Nevermind, fixed it!
I feel silly now...forgot to delete the "-O2" flag...

[Haoud]

No problem (this has already happened to me)
Good luck for the future

[Octocontrabass]

After calling my isr_install() function, I divide by zero expecting the fault handler to be called.

In C, division by zero is undefined behavior. That means when you try to divide by zero in a C program, the compiler may not actually generate code that tells the CPU to divide by zero.

If you want to ensure your program causes #DE, you'll have to use assembly. (Changing the optimization level is not guaranteed to work either!)

[zvoncika]

After calling my isr_install() function, I divide by zero expecting the fault handler to be called.

In C, division by zero is undefined behavior. That means when you try to divide by zero in a C program, the compiler may not actually generate code that tells the CPU to divide by zero.

If you want to ensure your program causes #DE, you'll have to use assembly. (Changing the optimization level is not guaranteed to work either!)

Thanks for letting be know!
I solved it by disabling optimization.

[Octocontrabass]

(Changing the optimization level is not guaranteed to work either!)

I solved it by disabling optimization.

Disabling optimization is also not guaranteed to work.

[MichaelPetch]

You can avoid the optimizer doing away with the div by zero if you put 0 in a volatile variable and use the volatile as the denominator of the division. Something like:

Code: Select all

volatile int zero = 0;
int divbyzero()
{
    return 1/zero;
}

Since the compiler can no longer assume the variable will be zero (might be modified asynchronously) it is forced to emit the division and it should produce an exception at run time. In your code you should be able to do this to generate the division:

Code: Select all

volatile int tmp = divbyzero();

Edit: This is still undefined behaviour and a poor way of attempting to do it since the compiler won't guarantee that even at runtime that it won't check for a denominator of zero (on x86 this usually isn't the case but the compiler is allowed to emit such instructions if it so chooses). The only way to guarantee this exception is raised is with an external assembly function or inline assembly that does div by zero. Something simple like this should work:

Code: Select all

__asm__  ("div %0" :: "r"(0));

. This should take whatever is in (e|r)ax (doesn't matter the value) and divide it by a register containing 0.

[zesterer]

Don't turn optimisation off. If ANY level of optimisation breaks your code, it's because the code is broken and you should really fix it. If you want to perform a division by zero, it's as simple as:

Code: Select all

volatile int i = 1 / 0;

'volatile' here ensures that the compiler won't optimise out the code.

[MichaelPetch]

Code: Select all

volatile int i = 1 / 0;

Unfortunately that is undefined behaviour. Some compilers will even replace this with a ud2 instruction.

[davidv1992]

Don't turn optimisation off. If ANY level of optimisation breaks your code, it's because the code is broken and you should really fix it. If you want to perform a division by zero, it's as simple as:

Code: Select all

volatile int i = 1 / 0;

'volatile' here ensures that the compiler won't optimise out the code.

I'm also afraid you misunderstand volatile here. The only thing it tells the compiler is that the variable i may be accessed/modified by other code it cant know about at (almost) any time. It doesn't actually do anything with the division.

[Octocontrabass]

You can avoid the optimizer doing away with the div by zero if you put 0 in a volatile variable and use the volatile as the denominator of the division. [...] Since the compiler can no longer assume the variable will be zero (might be modified asynchronously) it is forced to emit the division and it should produce an exception at run time.

Division by zero in C is still undefined behavior, even if it can't be detected at compile time. There's no guarantee the compiler won't insert code specifically to check if the denominator is zero and crash in some other way.