Re: [RFC][PATCH 3/3] x86: Add workaround to NMI iret woes

[Steven Rostedt <rostedt@goodmis.org>]

On Thu, 2011-12-08 at 21:43 -0500, Steven Rostedt wrote:

> diff --git a/arch/x86/kernel/nmi.c b/arch/x86/kernel/nmi.c
> index b9c8628..d35f554 100644
> --- a/arch/x86/kernel/nmi.c
> +++ b/arch/x86/kernel/nmi.c
> @@ -407,13 +407,29 @@ static notrace __kprobes void default_do_nmi(struct pt_regs *regs)
>  dotraplinkage notrace __kprobes void
>  do_nmi(struct pt_regs *regs, long error_code)
>  {
> +	int update_debug_stack = 0;
> +
>  	nmi_enter();
>  
> +	/*
> +	 * If we interrupted a breakpoint, it is possible that
> +	 * the nmi handler will have breakpoints too. We need to
> +	 * change the IDT such that breakpoints that happen here
> +	 * continue to use the NMI stack.
> +	 */
> +	if (unlikely(is_debug_stack(regs->sp))) {

For kicks, I made the update happen every NMI, but also changed the IDT
entry of the NMI to also not change the stack. Then I was able to move a
lot of the hacks from assembly into C code, just before the nmi_enter()
above. It seems it can work too. I didn't finish it, I just wanted to
see if it was doable. But I'm not sure we want to change the IDT (twice)
for all NMIs.

-- Steve


> +		zero_debug_stack();
> +		update_debug_stack = 1;
> +	}
> +	
>  	inc_irq_stat(__nmi_count);
>  
>  	if (!ignore_nmis)
>  		default_do_nmi(regs);
>  
> +	if (unlikely(update_debug_stack))
> +		reset_debug_stack();
> +
>  	nmi_exit();
>  }