wrong compiled output code??
Posted: Tue May 31, 2011 2:50 am
hey hey!!
cant seem to get my head around this problem, im sure its something that im missing...
it *seems* that for a particular "for" loop in my C code, the compiler (GNU 4.5.2) has produced an endless loop.
the "for" loop should only loop 200 or so times, but when i step through the code in the debugger while running it on Bochs, the op-code that is being run *correctly* sets ebx to the counting variable's ("i" in the C code) initial value and *correctly* increments ebx on each loop iteration as i step through the running code, *BUT* there is no comparison of ebx to my "upper limit" at which the loop is supposed to cease - in the decompiled assembly there is only an unconditional "jmp .-25" and hence i get an endless loop...
anyway, i need to give some solid information so here it is:
C compiler: GNU 4.5.2
Bochs: my *debugging* version is custom-compiled with just about all features (SMP, APIC, USB etc) (yes i know its slow, but i have a "lean" version for fast non-debugging tests
)
here is a snippet of the C-Source code around the offending "for" loop (p.s. IDT_ENTRIES == 256):
and the corresponding dissassembled code... (im pretty sure this is it, but maybe ive not followed my debug-stepping closely enough due to lack of sleep... eitherway im going to be rechecking over it so ill update if what i post here turns out to be wrong)
any insight into what im overlooking would be wonderful!! anymore info let me know 
thanks!!
cant seem to get my head around this problem, im sure its something that im missing...
it *seems* that for a particular "for" loop in my C code, the compiler (GNU 4.5.2) has produced an endless loop.
the "for" loop should only loop 200 or so times, but when i step through the code in the debugger while running it on Bochs, the op-code that is being run *correctly* sets ebx to the counting variable's ("i" in the C code) initial value and *correctly* increments ebx on each loop iteration as i step through the running code, *BUT* there is no comparison of ebx to my "upper limit" at which the loop is supposed to cease - in the decompiled assembly there is only an unconditional "jmp .-25" and hence i get an endless loop...
anyway, i need to give some solid information so here it is:
C compiler: GNU 4.5.2
Bochs: my *debugging* version is custom-compiled with just about all features (SMP, APIC, USB etc) (yes i know its slow, but i have a "lean" version for fast non-debugging tests
)here is a snippet of the C-Source code around the offending "for" loop (p.s. IDT_ENTRIES == 256):
Code: Select all
//...
idt_SetGate(IRQ_Map[14], IRQ14, ring0_Code, IDT_FLAGS_RING0);
idt_SetGate(IRQ_Map[15], IRQ15, ring0_Code, IDT_FLAGS_RING0); /** eip = 0x0000be47 **/
// finished remapping PIC IRQs
// (for now)... just register the rest of the interrupts up to 255 to run ISR048
for(i = 48; i <= (IDT_ENTRIES - 1); i++) {
idt_SetGate(i, ISR048, ring0_Code, IDT_FLAGS_RING0);
}
// register the kernel data descriptor to use for interrupt routines
// (write to _ISR_kernel_data_descriptor, located in ISR.asm)
ISR_kernel_data_descriptor = ring0_Data;
// initialise all external interrupt drivers to NULL
struct InterruptHandler_s null_Handler;
null_Handler.handlerFunction = 0;
for(i = 0; i <= (IDT_ENTRIES - 1); i++) {
idt_RegisterInterruptHandler(i, null_Handler);
}
// set the new IDT
idt_Flush();
}
and the corresponding dissassembled code... (im pretty sure this is it, but maybe ive not followed my debug-stepping closely enough due to lack of sleep... eitherway im going to be rechecking over it so ill update if what i post here turns out to be wrong)
Code: Select all
L.Address (bytes) Opcode Mnemonic
---------------------------------------------------------
0x0000be47 (5) e858faffff call .-1448 (0x0000b8a4) <see corresponding C-code comment>
0x0000be4c (2) b330 mov bl, 0x30
0x...be4e (3) 0fb6c3 movzx eax, bl
be51 (7) c704248e000000 mov dword ptr ss:[esp], 0x0000008e
be58 (2) 89f1 mov ecx, esi
be5a (5) ba7fb40000 mov edx, 0x0000b47f
be5f (5) e840faffff call .-1472 (0x0000b8a4)
be64 (1) 43 inc ebx
be65 (2) ebe7 jmp .-25 (0x0000be4e)
be67 (1) 90 nop
be68 (5) 0fb6442404 movzx eax, byte ptr ss:[esp+4]
be6d (4) 8b542408 mov edx, dword ptr ss:[esp+8]
be71 (7) 89148514f00000 mov dword ptr ds:[eax*4+61460], edx
be78 (1) c3 ret
be79 (1) c3 ret
be7a (1) 57 push edi
be7b (1) 56 push esi
be7c (3) ... ...
--------------------------------------------------------------------------------
at eip = 0x0000be4c (just after the last function call returns, right before entering the offending "for" loop) the registers are:
eax = 0x0000008e
ebx = 0x0000000a
ecx = 0x0000e2b8
edx = 0x00000008
esi = 0x00000008
edi = 0x00007ac0
ebp = 0x00000000
esp = 0x000079a0
(segment registers are all correct GDT entries, etc)
thanks!!
