Re: [PATCH -v2 7/7] x86, NMI, Remove do_nmi_callback logic

[Don Zickus <dzickus@redhat.com>]

On Tue, Sep 28, 2010 at 08:28:25AM +0800, Huang Ying wrote:
> Hi, Don,
> 
> On Mon, 2010-09-27 at 23:16 +0800, Don Zickus wrote:
> > On Mon, Sep 27, 2010 at 03:43:41PM +0200, Robert Richter wrote:
> > > On 27.09.10 08:56:44, huang ying wrote:
> > > 
> > > > >> -static int unknown_nmi_panic_callback(struct pt_regs *regs, int cpu)
> > > > >> -{
> > > > >> -     unsigned char reason = get_nmi_reason();
> > > > >> -     char buf[64];
> > > > >> -
> > > > >> -     sprintf(buf, "NMI received for unknown reason %02x\n", reason);
> > > > >> -     die_nmi(buf, regs, 1); /* Always panic here */
> > > > >> -     return 0;
> > > > >
> > > > > You are dropping this code that is different to panic().
> > > > 
> > > > What is the difference? Is it relevant?
> > > 
> > > I think yes, since the code behaves different. Otherwise we could
> > > remove die_nmi() completly and replace it by panic(). But both are
> > > different implementions. Maybe we can merge the code, but I didn't
> > > look at it closly.
> > 
> > Actually die_nmi is a wrapper around panic with two important pieces.
> > One, it dumps some registers and two it does another notifier call to
> > DIE_NMIWATCHDOG (which correlates to another discussion in this patch
> > series).
> > 
> > So if we do any consolidation between panic and die_nmi, it should be
> > convert to die_nmi.  But then I wonder if that breaks the original
> > semantics of 'panic_on_unrecovered_nmi'.  I don't think so though.
> 
> Please take a look at the original code:
> 
> 
>                 if (nmi_watchdog_tick(regs, reason))
>                         return;
>                 if (!do_nmi_callback(regs, cpu))
> #endif /* !CONFIG_LOCKUP_DETECTOR */
>                         unknown_nmi_error(reason, regs);
> #else
>                 unknown_nmi_error(reason, regs);
> #endif
> 
> If NMI comes from watchdog, nmi_watchdog_tick() will return 1. So
> do_nmi_callback() is NOT for watchdog NMI, but for unknown NMI. Why do
> we call DIE_NMIWATCHDOG for unknown NMI (NOT watchdog NMI)? die_nmi is
> for watchdog, not unknown NMI.

I think watchdog is an overloaded term.  I was under the impression that
once the nmi watchdog determined a problem, it called the DIE_NMIWATCHDOG
die chain to see if any other drivers wanted to clean up or do their thing
first before panic'ing (namely drivers in drivers/char/watchdog).

Cheers,
Don