Re: [PATCH] Port SquashFS to 2.6

From: Jörn Engel (
Date: Tue Jul 22 2003 - 05:47:23 EST

On Tue, 22 July 2003 11:36:07 +0100, John Bradford wrote:
> > If you look closely at the kernel, there is currently no way of
> > telling whether it contains stack overflows waiting to happen, or not.
> It would be an interesting experiment to deliberately make the kernel
> stack smaller, and see what happens. If no problems seem apparent
> with a reduced kernel stack, it gives more weight to the argument that
> the default one is OK.

Been there, done that. :)

Use the patch below, enable the stack checking in your .config and
watch the system log get filled. 5k is quite agressive, I agree, but
with 4k stacks in mind, it is just 1k of slack left, plus the
measurement is a bit fuzzy, so you really want some slack.

Maybe there is also a patch flying around that initializes the stack
with some magic number on a fork and logs the number of used bytes on
exit. That way you get rid of the fuzzyness, but you loose the exact
call trace that led to this number.


Measure. Don't tune for speed until you've measured, and even then
don't unless one part of the code overwhelms the rest.
-- Rob Pike

--- linux-2.5.67/arch/i386/kernel/irq.c~stack_overflow 2003-04-07 19:30:39.000000000 +0200 +++ linux-2.5.67/arch/i386/kernel/irq.c 2003-04-14 20:22:01.000000000 +0200 @@ -342,7 +342,15 @@ __asm__ __volatile__("andl %%esp,%0" : "=r" (esp) : "0" (8191)); +#if 0 if (unlikely(esp < (sizeof(struct thread_info) + 1024))) { +#else + /* We check for 5k for now. The kernel stack still is 8k, + * but should shrink to 4k, so this test makes sense. + * Once the stack is 4k, we go back to the old test. + */ + if (unlikely(esp < (sizeof(struct thread_info) + 5120))) { +#endif printk("do_IRQ: stack overflow: %ld\n", esp - sizeof(struct thread_info)); dump_stack(); - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to More majordomo info at Please read the FAQ at

This archive was generated by hypermail 2b29 : Wed Jul 23 2003 - 22:00:46 EST