On Fri, 25 May 2012, Thomas Gleixner wrote:
> On Fri, 25 May 2012, Chen Gong wrote:
> 
> > 于 2012/5/25 14:24, Borislav Petkov 写道:
> > > On Thu, May 24, 2012 at 05:54:52PM +0000, Thomas Gleixner wrote:
> > >> Intentionally left blank to be filled out by someone who wants
> > >> that and can explain the reason for this better than me.
> > > 
> > > That'll be Intel folk :)
> > > 
> > >> Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
> > > 
> > > Looks good to me too, thanks Thomas for doing this!
> > > 
> > > I'll run it next week just in case.
> > > 
> > > Acked-by: Borislav Petkov <borislav.petkov@amd.com>
> > > 
> > 
> > Oh, Oh, wait. First I need to thank Thomas to improve it. I don't
> > reply you at the first time because I have some thoughts and now
> > I'm testing it. The basic test shows it hangs the system after
> > sb_edac is removed and when error count increases to the threshold
> > it hangs again, and when trying to reboot the system no hang happnes
> > (not reach the threshold) the system oops. I need to time to debug
> > and give the valid feedback. Please be patient :-). Of course,
> 
> As I said it's completely untested. I just made sure it compiles :)

I just had a quick look again and noticed that nothing is polling the
cmci part in case of the storm mode.

Delta fix below.

Thanks,

	tglx

Index: linux-2.6/arch/x86/kernel/cpu/mcheck/mce-internal.h
===================================================================
--- linux-2.6.orig/arch/x86/kernel/cpu/mcheck/mce-internal.h
+++ linux-2.6/arch/x86/kernel/cpu/mcheck/mce-internal.h
@@ -30,8 +30,10 @@ extern struct mce_bank *mce_banks;
 
 #ifdef CONFIG_X86_MCE_INTEL
 unsigned long mce_intel_adjust_timer(unsigned long interval);
+void mce_intel_cmci_poll(void);
 #else
 # define mce_intel_adjust_timer mce_adjust_timer_default
+static inline void mce_intel_cmci_poll(void) { }
 #endif
 
 void mce_timer_kick(unsigned long interval);
Index: linux-2.6/arch/x86/kernel/cpu/mcheck/mce.c
===================================================================
--- linux-2.6.orig/arch/x86/kernel/cpu/mcheck/mce.c
+++ linux-2.6/arch/x86/kernel/cpu/mcheck/mce.c
@@ -1260,6 +1260,7 @@ static void mce_timer_fn(unsigned long d
 	if (mce_available(__this_cpu_ptr(&cpu_info))) {
 		machine_check_poll(MCP_TIMESTAMP,
 				&__get_cpu_var(mce_poll_banks));
+		mce_intel_cmci_poll();
 	}
 
 	/*
Index: linux-2.6/arch/x86/kernel/cpu/mcheck/mce_intel.c
===================================================================
--- linux-2.6.orig/arch/x86/kernel/cpu/mcheck/mce_intel.c
+++ linux-2.6/arch/x86/kernel/cpu/mcheck/mce_intel.c
@@ -70,6 +70,13 @@ static int cmci_supported(int *banks)
 	return !!(cap & MCG_CMCI_P);
 }
 
+void mce_intel_cmci_poll(void)
+{
+	if (__this_cpu_read(cmci_storm_state) == CMCI_STORM_NONE)
+		return;
+	machine_check_poll(MCP_TIMESTAMP, &__get_cpu_var(mce_banks_owned));
+}
+
 unsigned long mce_intel_adjust_timer(unsigned long interval)
 {
 	if (interval < CMCI_POLL_INTERVAL)