Re: [PATCH][6/8] Arch agnostic completely out of line locks / arm

[Russell King <rmk+lkml@arm.linux.org.uk>]

On Thu, Sep 02, 2004 at 08:02:55PM -0400, Zwane Mwaikambo wrote:
> --- linux-2.6.9-rc1-mm1-stage/arch/arm/kernel/time.c	26 Aug 2004 13:13:04 -0000	1.1.1.1
> +++ linux-2.6.9-rc1-mm1-stage/arch/arm/kernel/time.c	2 Sep 2004 15:51:37 -0000
> @@ -52,6 +52,18 @@ EXPORT_SYMBOL(rtc_lock);
>  /* change this if you have some constant time drift */
>  #define USECS_PER_JIFFY	(1000000/HZ)
> 
> +#ifdef CONFIG_SMP
> +unsigned long profile_pc(struct pt_regs *regs)
> +{
> +	unsigned long pc = instruction_pointer(regs);
> +
> +	if (pc >= (unsigned long)&__lock_text_start &&
> +	    pc <= (unsigned long)&__lock_text_end)
> +		return regs->ARM_lr;
> +	return pc;
> +}
> +EXPORT_SYMBOL(profile_pc);
> +#endif

Looks good apart from this.  There's no guarantee that LR will be the
return address inside one of these lock functions - indeed the compiler
may have saved it onto the stack and decided to use the register for
something else.

Your best bet is to look at the get_wchan() code which walks the stack
frames (if present.)

-- 
Russell King
 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/
 maintainer of:  2.6 PCMCIA      - http://pcmcia.arm.linux.org.uk/
                 2.6 Serial core