Re: [PATCH] alternate stack dump fix.

[Oleg Nesterov <oleg@tv-sign.ru>]

Andrew Morton wrote:
> 
> In that case I'm not understanding you.  Are you saying that your patch
> fixes the same problem as that which Kirill's patch fixed?

It seems to me, yes.

Here the pseudo code for CONFIG_FRAME_POINTER:

int valid_stack_ptr(struct thread_info *tinfo, void *p)
{
	return	p > (void *)tinfo &&
		p < (void *)tinfo + THREAD_SIZE - 3;
}

long print_context_stack(thread_info *tinfo, long *stack, long ebp)
{
	while (valid_stack_ptr(tinfo, ebp))
	{
		print_symbol("%s", *(ebp+4));
		ebp = *(unsigned long *)ebp;
	}

	return ebp;
}

void show_trace(struct task_struct *task, unsigned long * stack)
{
	while (1) {
		struct thread_info *context = (struct thread_info *)
			((unsigned long)stack & (~(THREAD_SIZE - 1)));

		ebp = print_context_stack(context, stack, ebp);

		stack = (unsigned long*)context->previous_esp;
		if (!stack)
			break;
		printk(" =======================\n");
	}
}

show_trace() now does not use task argument in the main
loop. Instead, it converts stack to thread_info* context,
and passes it to print_context_stack() and (implicitly)
to valid_stack_ptr().

valid_stack_ptr() does not care now whether or not it is
irq stack, it just does bounds checking.

Please note, i simply deleted this printk("Stack pointer is garbage"),
it can be restored, if neccessary.

Did i miss something?

> That wasn't at all clear from your earlier comments.

Yes, sorry.

Oleg.