I am making my custom 8x86 64 bit OS, and I have some problem for an interrupt handler.
When I added interrupt handlers for common exceptions and interrupts, I enabled interrupt by "sti" instruction.
Then, "General Protection Fault" occurs right after enabling interrupt.
So, I checked GDB.
The stack is
(gdb) bt
#0 kISR::kISRGeneralProtection () at /home/xaliver/WorkSpace/kOdin/kernel64/./ISR.cpp:70
#1 0x000000000000fee8 in ?? ()
#2 0x0000000000202e30 in _kISRTimer ()
Backtrace stopped: frame did not save the PC
The weird part is 0xfee8. I uses memory below 0x10000 as a stack for the 32 bit protected mode. Now, the system is in 64 bit mode, so the value is empty as "00 00".
So, I checked the frame 2, _kISRTimer.
The saved rip of 0xfee8 is 0x202e30. It is "iretq" instruction of _kISRTimer.
here is assembly code.
; #32, Timer ISR
_kISRTimer:
KSAVECONTEXT ; Store the context and change selector to
202dd3: 55 push %rbp
202dd4: 48 89 e5 mov %rsp,%rbp
202dd7: 50 push %rax
202dd8: 53 push %rbx
202dd9: 51 push %rcx
202dda: 52 push %rdx
202ddb: 57 push %rdi
202ddc: 56 push %rsi
202ddd: 41 50 push %r8
202ddf: 41 51 push %r9
202de1: 41 52 push %r10
202de3: 41 53 push %r11
202de5: 41 54 push %r12
202de7: 41 55 push %r13
202de9: 41 56 push %r14
202deb: 41 57 push %r15
202ded: 66 8c d8 mov %ds,%ax
202df0: 50 push %rax
202df1: 66 8c c0 mov %es,%ax
202df4: 50 push %rax
202df5: 0f a0 pushq %fs
202df7: 0f a8 pushq %gs
202df9: 66 b8 10 00 mov $0x10,%ax
202dfd: 8e d8 mov %eax,%ds
202dff: 8e c0 mov %eax,%es
202e01: 8e e8 mov %eax,%gs
202e03: 8e e0 mov %eax,%fs
; kernel data descriptor
; Insert interrupt number to the hander and call the handler
mov rdi, 32
202e05: bf 20 00 00 00 mov $0x20,%edi
call _kCommonInterruptHandler
202e0a: e8 75 e3 ff ff callq 201184 <_kCommonInterruptHandler>
KLOADCONTEXT ; Restore the context
202e0f: 0f a9 popq %gs
202e11: 0f a1 popq %fs
202e13: 58 pop %rax
202e14: 8e c0 mov %eax,%es
202e16: 58 pop %rax
202e17: 8e d8 mov %eax,%ds
202e19: 41 5f pop %r15
202e1b: 41 5e pop %r14
202e1d: 41 5d pop %r13
202e1f: 41 5c pop %r12
202e21: 41 5b pop %r11
202e23: 41 5a pop %r10
202e25: 41 59 pop %r9
202e27: 41 58 pop %r8
202e29: 5e pop %rsi
202e2a: 5f pop %rdi
202e2b: 5a pop %rdx
202e2c: 59 pop %rcx
202e2d: 5b pop %rbx
202e2e: 58 pop %rax
202e2f: 5d pop %rbp
iretq ; Return the event point after interrupt
202e30: 48 cf iretq
I think there is no point to jump 0xfee8.
I also tried to remove the interrupt handler for timer. However, "GPF" occurs at other interrupt handler. The other interrupt handler is the same as _kISRTimer, but the call function is different.
Do you know why GPF occurs from the code? Or, the jump is not from code? Please let me know why this problem happens. Thank you.
The iretq instruction, or return from interrupt handler to a 64-bit address, expects the return address and the flags to be on the top of the stack, and produces a general protection fault if the return address is not valid (for example, because it points to a no-execute memory page). Therefore, you should inspect the values on the stack when your handler gets called, the calling code, and the CS selector. If any of these have changed, your stack might contain garbage. Your comment about “memory below 0x10000 as a stack for the 32 bit protected mode” suggests that one or both of these might be the case.