Re: [PATCH] x86: handle both NMI kinds if they occur simultaneously

[Andrew Cooper <andrew.cooper3@citrix.com>]

On 17/01/13 10:28, Jan Beulich wrote:
> We shouldn't assume PCI SERR excludes IOCHK.
>
> Once at it, also remove the doubly redundant range restriction on
> "reason" - the variable already is "unsigned char".
>
> Signed-off-by: Jan Beulich <jbeulich@suse.com>

I am soon to embark on some work to do with NMI buttons, and was already
planning to make alterations along these lines as part of that work, so
for what it is worth,

Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>

>
> --- a/xen/arch/x86/traps.c
> +++ b/xen/arch/x86/traps.c
> @@ -3201,10 +3201,10 @@ void do_nmi(struct cpu_user_regs *regs)
>          reason = inb(0x61);
>          if ( reason & 0x80 )
>              pci_serr_error(regs);
> -        else if ( reason & 0x40 )
> +        if ( reason & 0x40 )
>              io_check_error(regs);
> -        else if ( !nmi_watchdog )
> -            unknown_nmi_error(regs, (unsigned char)(reason&0xff));
> +        if ( !(reason & 0xc0) && !nmi_watchdog )
> +            unknown_nmi_error(regs, reason);
>      }
>  }
>  
>
>
>