Strange thing about IDT

Jeko · Post by **Jeko** » Sat Mar 01, 2008 3:55 am

I was programming the interrupt handling functions, more precisely the IDT, when a strange thing happened.

My code wasn't working:

Code: Select all

struct idt_t {
	unsigned short offset0_15;
	unsigned short segment;
	unsigned short reserved : 5;
	unsigned short options : 11;
	unsigned short offset16_31;
};

struct idt_t idt[256];

asm("cli");
...   //Here each entry in the IDT
...   //is initialized

unsigned int idt_reg[2];

idt_reg[0] = (256*8) << 16;

idt_reg[1] = (unsigned int)idt;

__asm__ __volatile__ ("lidt (%0)": :"g" ((char*)idt_reg+2));
asm("sti");

Then I was very angry after a lot of tests and I wrote:

Code: Select all

printk("****!");

Immediately before 'asm("sti");'
and... It worked...
So now I wrote:

Code: Select all

printk("");

and it continues to work...
Why????????

JamesM · Post by **JamesM** » Sat Mar 01, 2008 8:00 am

Pull every line of code out of printk() until it breaks again. Then you've found your problem.

Jeko · Post by **Jeko** » Sat Mar 01, 2008 12:55 pm

JamesM wrote:Pull every line of code out of printk() until it breaks again. Then you've found your problem.

Maybe I hadn't explained very well...
If I insert a printk, also with an empty string, my code works and exceptions a re handled.
If I don't insert a printk, bochs (and also real computers) reboots...
This is the log of bochs when there isn't printk:

Code: Select all

00005787616i[CPU0 ] >> mov word ptr ds:0xe00b8000, 0x0009 : 66C70500800BE00900
00005787616e[CPU0 ] exception(): 3rd (13) exception with no resolution, shutdown status is 00h, resetting

When there is the printk:

Code: Select all

//On the screen is written: "Page fault exception"
00005768624i[CPU0 ] WARNING: HLT instruction with IF=0!

So it works with the printk and not without the printk...

pcmattman · Post by **pcmattman** » Sat Mar 01, 2008 4:52 pm

What's *AFTER* the printk?

Looks to me like there might be dud code executing after the asm( "sti" ); which your printk stops thanks to a page fault

.

JamesM · Post by **JamesM** » Sun Mar 02, 2008 3:03 am

No, you explained yourself well, and my answer still stands.

Some line(s) of code in printk are changing the result. So go through it and pull out everything until it breaks again. Then you know what it causing it.

Unfortunately too often in OSDev we use intrusive methods of debugging, and this is one of the consequences.

Jeko · Post by **Jeko** » Sun Mar 02, 2008 4:21 am

JamesM wrote:No, you explained yourself well, and my answer still stands.

Some line(s) of code in printk are changing the result. So go through it and pull out everything until it breaks again. Then you know what it causing it.

Unfortunately too often in OSDev we use intrusive methods of debugging, and this is one of the consequences.

But with printk it doesn't break. With printk my exception handler works fine. Without printk my exception handler doesn't work.

Edit: my printk is based on your monitor_write

JamesM · Post by **JamesM** » Sun Mar 02, 2008 7:48 am

MarkOS: Please re-read my answer. You have not understood it.

Jeko · Post by **Jeko** » Sun Mar 02, 2008 11:01 am

JamesM wrote:MarkOS: Please re-read my answer. You have not understood it.

What must I pull out?

JamesM · Post by **JamesM** » Sun Mar 02, 2008 11:26 am

I'll say it again. Remove every line from printk, until you find the line that causes your function to stop breaking. Then you will know how to fix it!

Jeko · Post by **Jeko** » Sun Mar 02, 2008 11:36 am

JamesM wrote:I'll say it again. Remove every line from printk, until you find the line that causes your function to stop breaking. Then you will know how to fix it!

ah now I understand!

Sorry but I don't know english very well...sorry!

Jeko · Post by **Jeko** » Sun Mar 02, 2008 11:41 am

ehm...

Now it's very very strange...
I commented all the function printk:

Code: Select all

void printk(char *format, ...)
{
	/*unsigned int curchar, varnumber=0, curpos=0;
	unsigned char *tmp;
	va_list pt = pt;	//Serve solo ad eliminare un warning (probabilmente)

	while((curchar = format[curpos++]))
		if(curchar == '%')
			varnumber += 1;

	if(varnumber)
		va_start(pt, format);

	while((curchar = *format++)) {	//Loop attraverso la stringa
		switch(curchar) {
			case '%':
				curchar = *format++;
				switch(curchar) {
					case 'd':
					case 'l':
						print_dec((int)va_arg(pt, int));
					break;

					case 'u':
						print_unsigned((unsigned int)va_arg(pt, unsigned int));
					break;

					case 'f':
						print_float((float)va_arg(pt, float));
					break;

					case 'b':
						print_binary(va_arg(pt, unsigned int));
					break;

					case 'x':
						print_hex(va_arg(pt, unsigned int));
					break;

					case 's':
						tmp = va_arg(pt, unsigned char *);
						print_string(tmp);
					break;

					case 'c':
						putc(va_arg(pt, unsigned char));
					break;
				}
			break;

			default:
				putc(curchar);
		}
	}
	va_end(pt);*/
}

and...it continues to work. Without the all commented printk, it doesn't work...