diff --git a/arch/i386/kernel/nmi.c b/arch/i386/kernel/nmi.c
index eaafe233a5da..171194ccb7bc 100644
--- a/arch/i386/kernel/nmi.c
+++ b/arch/i386/kernel/nmi.c
@@ -22,6 +22,7 @@
 #include <linux/percpu.h>
 #include <linux/dmi.h>
 #include <linux/kprobes.h>
+#include <linux/cpumask.h>
 
 #include <asm/smp.h>
 #include <asm/nmi.h>
@@ -42,6 +43,8 @@ int nmi_watchdog_enabled;
 static DEFINE_PER_CPU(unsigned long, perfctr_nmi_owner);
 static DEFINE_PER_CPU(unsigned long, evntsel_nmi_owner[3]);
 
+static cpumask_t backtrace_mask = CPU_MASK_NONE;
+
 /* this number is calculated from Intel's MSR_P4_CRU_ESCR5 register and it's
  * offset from MSR_P4_BSU_ESCR0.  It will be the max for all platforms (for now)
  */
@@ -907,6 +910,16 @@ __kprobes int nmi_watchdog_tick(struct pt_regs * regs, unsigned reason)
 		touched = 1;
 	}
 
+	if (cpu_isset(cpu, backtrace_mask)) {
+		static DEFINE_SPINLOCK(lock);	/* Serialise the printks */
+
+		spin_lock(&lock);
+		printk("NMI backtrace for cpu %d\n", cpu);
+		dump_stack();
+		spin_unlock(&lock);
+		cpu_clear(cpu, backtrace_mask);
+	}
+
 	sum = per_cpu(irq_stat, cpu).apic_timer_irqs;
 
 	/* if the apic timer isn't firing, this cpu isn't doing much */
@@ -1033,6 +1046,19 @@ int proc_nmi_enabled(struct ctl_table *table, int write, struct file *file,
 
 #endif
 
+void __trigger_all_cpu_backtrace(void)
+{
+	int i;
+
+	backtrace_mask = cpu_online_map;
+	/* Wait for up to 10 seconds for all CPUs to do the backtrace */
+	for (i = 0; i < 10 * 1000; i++) {
+		if (cpus_empty(backtrace_mask))
+			break;
+		mdelay(1);
+	}
+}
+
 EXPORT_SYMBOL(nmi_active);
 EXPORT_SYMBOL(nmi_watchdog);
 EXPORT_SYMBOL(avail_to_resrv_perfctr_nmi);
diff --git a/arch/x86_64/kernel/nmi.c b/arch/x86_64/kernel/nmi.c
index 7af9cb3e2d99..27e95e7922c1 100644
--- a/arch/x86_64/kernel/nmi.c
+++ b/arch/x86_64/kernel/nmi.c
@@ -12,14 +12,15 @@
  *  Mikael Pettersson	: PM converted to driver model. Disable/enable API.
  */
 
+#include <linux/nmi.h>
 #include <linux/mm.h>
 #include <linux/delay.h>
 #include <linux/interrupt.h>
 #include <linux/module.h>
 #include <linux/sysdev.h>
-#include <linux/nmi.h>
 #include <linux/sysctl.h>
 #include <linux/kprobes.h>
+#include <linux/cpumask.h>
 
 #include <asm/smp.h>
 #include <asm/nmi.h>
@@ -41,6 +42,8 @@ int panic_on_unrecovered_nmi;
 static DEFINE_PER_CPU(unsigned, perfctr_nmi_owner);
 static DEFINE_PER_CPU(unsigned, evntsel_nmi_owner[2]);
 
+static cpumask_t backtrace_mask = CPU_MASK_NONE;
+
 /* this number is calculated from Intel's MSR_P4_CRU_ESCR5 register and it's
  * offset from MSR_P4_BSU_ESCR0.  It will be the max for all platforms (for now)
  */
@@ -782,6 +785,7 @@ int __kprobes nmi_watchdog_tick(struct pt_regs * regs, unsigned reason)
 {
 	int sum;
 	int touched = 0;
+	int cpu = smp_processor_id();
 	struct nmi_watchdog_ctlblk *wd = &__get_cpu_var(nmi_watchdog_ctlblk);
 	u64 dummy;
 	int rc=0;
@@ -799,6 +803,16 @@ int __kprobes nmi_watchdog_tick(struct pt_regs * regs, unsigned reason)
 		touched = 1;
 	}
 
+	if (cpu_isset(cpu, backtrace_mask)) {
+		static DEFINE_SPINLOCK(lock);	/* Serialise the printks */
+
+		spin_lock(&lock);
+		printk("NMI backtrace for cpu %d\n", cpu);
+		dump_stack();
+		spin_unlock(&lock);
+		cpu_clear(cpu, backtrace_mask);
+	}
+
 #ifdef CONFIG_X86_MCE
 	/* Could check oops_in_progress here too, but it's safer
 	   not too */
@@ -931,6 +945,19 @@ int proc_nmi_enabled(struct ctl_table *table, int write, struct file *file,
 
 #endif
 
+void __trigger_all_cpu_backtrace(void)
+{
+	int i;
+
+	backtrace_mask = cpu_online_map;
+	/* Wait for up to 10 seconds for all CPUs to do the backtrace */
+	for (i = 0; i < 10 * 1000; i++) {
+		if (cpus_empty(backtrace_mask))
+			break;
+		mdelay(1);
+	}
+}
+
 EXPORT_SYMBOL(nmi_active);
 EXPORT_SYMBOL(nmi_watchdog);
 EXPORT_SYMBOL(avail_to_resrv_perfctr_nmi);
diff --git a/include/asm-i386/nmi.h b/include/asm-i386/nmi.h
index 269d315719ca..b04333ea6f31 100644
--- a/include/asm-i386/nmi.h
+++ b/include/asm-i386/nmi.h
@@ -5,6 +5,9 @@
 #define ASM_NMI_H
 
 #include <linux/pm.h>
+#include <asm/irq.h>
+
+#ifdef ARCH_HAS_NMI_WATCHDOG
 
 /**
  * do_nmi_callback
@@ -42,4 +45,9 @@ extern int proc_nmi_enabled(struct ctl_table *, int , struct file *,
 			void __user *, size_t *, loff_t *);
 extern int unknown_nmi_panic;
 
+void __trigger_all_cpu_backtrace(void);
+#define trigger_all_cpu_backtrace() __trigger_all_cpu_backtrace()
+
+#endif
+
 #endif /* ASM_NMI_H */
diff --git a/include/asm-x86_64/nmi.h b/include/asm-x86_64/nmi.h
index f367d4014b42..72375e7d32a8 100644
--- a/include/asm-x86_64/nmi.h
+++ b/include/asm-x86_64/nmi.h
@@ -77,4 +77,7 @@ extern int proc_nmi_enabled(struct ctl_table *, int , struct file *,
 
 extern int unknown_nmi_panic;
 
+void __trigger_all_cpu_backtrace(void);
+#define trigger_all_cpu_backtrace() __trigger_all_cpu_backtrace()
+
 #endif /* ASM_NMI_H */
diff --git a/include/linux/nmi.h b/include/linux/nmi.h
index e16904e28c3a..acb4ed130247 100644
--- a/include/linux/nmi.h
+++ b/include/linux/nmi.h
@@ -15,9 +15,14 @@
  * disables interrupts for a long time. This call is stateless.
  */
 #ifdef ARCH_HAS_NMI_WATCHDOG
+#include <asm/nmi.h>
 extern void touch_nmi_watchdog(void);
 #else
 # define touch_nmi_watchdog() touch_softlockup_watchdog()
 #endif
 
+#ifndef trigger_all_cpu_backtrace
+#define trigger_all_cpu_backtrace() do { } while (0)
+#endif
+
 #endif
diff --git a/lib/spinlock_debug.c b/lib/spinlock_debug.c
index b6c4f898197c..479fd462eaa9 100644
--- a/lib/spinlock_debug.c
+++ b/lib/spinlock_debug.c
@@ -7,6 +7,7 @@
  */
 
 #include <linux/spinlock.h>
+#include <linux/nmi.h>
 #include <linux/interrupt.h>
 #include <linux/debug_locks.h>
 #include <linux/delay.h>
@@ -117,6 +118,9 @@ static void __spin_lock_debug(spinlock_t *lock)
 				raw_smp_processor_id(), current->comm,
 				current->pid, lock);
 			dump_stack();
+#ifdef CONFIG_SMP
+			trigger_all_cpu_backtrace();
+#endif
 		}
 	}
 }