Merge branch 'master'

[linux-2.6-omap-h63xx.git] / arch / i386 / kernel / nmi.c

diff --git a/arch/i386/kernel/nmi.c b/arch/i386/kernel/nmi.c
index 0178457db72130b54d3170dfd8150742b8addfe1..d661703ac1cb713db7327de628e4f7f5b09efcb1 100644 (file)
--- a/arch/i386/kernel/nmi.c
+++ b/arch/i386/kernel/nmi.c
@@ -15,7 +15,6 @@
 
 #include <linux/config.h>
 #include <linux/mm.h>
-#include <linux/irq.h>
 #include <linux/delay.h>
 #include <linux/bootmem.h>
 #include <linux/smp_lock.h>
@@ -101,16 +100,44 @@ int nmi_active;
        (P4_CCCR_OVF_PMI0|P4_CCCR_THRESHOLD(15)|P4_CCCR_COMPLEMENT|     \
         P4_CCCR_COMPARE|P4_CCCR_REQUIRED|P4_CCCR_ESCR_SELECT(4)|P4_CCCR_ENABLE)
 
+#ifdef CONFIG_SMP
+/* The performance counters used by NMI_LOCAL_APIC don't trigger when
+ * the CPU is idle. To make sure the NMI watchdog really ticks on all
+ * CPUs during the test make them busy.
+ */
+static __init void nmi_cpu_busy(void *data)
+{
+       volatile int *endflag = data;
+       local_irq_enable();
+       /* Intentionally don't use cpu_relax here. This is
+          to make sure that the performance counter really ticks,
+          even if there is a simulator or similar that catches the
+          pause instruction. On a real HT machine this is fine because
+          all other CPUs are busy with "useless" delay loops and don't
+          care if they get somewhat less cycles. */
+       while (*endflag == 0)
+               barrier();
+}
+#endif
+
 static int __init check_nmi_watchdog(void)
 {
-       unsigned int prev_nmi_count[NR_CPUS];
+       volatile int endflag = 0;
+       unsigned int *prev_nmi_count;
        int cpu;
 
        if (nmi_watchdog == NMI_NONE)
                return 0;
 
+       prev_nmi_count = kmalloc(NR_CPUS * sizeof(int), GFP_KERNEL);
+       if (!prev_nmi_count)
+               return -1;
+
        printk(KERN_INFO "Testing NMI watchdog ... ");
 
+       if (nmi_watchdog == NMI_LOCAL_APIC)
+               smp_call_function(nmi_cpu_busy, (void *)&endflag, 0, 0);
+
        for (cpu = 0; cpu < NR_CPUS; cpu++)
                prev_nmi_count[cpu] = per_cpu(irq_stat, cpu).__nmi_count;
        local_irq_enable();
@@ -124,12 +151,18 @@ static int __init check_nmi_watchdog(void)
                        continue;
 #endif
                if (nmi_count(cpu) - prev_nmi_count[cpu] <= 5) {
-                       printk("CPU#%d: NMI appears to be stuck!\n", cpu);
+                       endflag = 1;
+                       printk("CPU#%d: NMI appears to be stuck (%d->%d)!\n",
+                               cpu,
+                               prev_nmi_count[cpu],
+                               nmi_count(cpu));
                        nmi_active = 0;
                        lapic_nmi_owner &= ~LAPIC_NMI_WATCHDOG;
+                       kfree(prev_nmi_count);
                        return -1;
                }
        }
+       endflag = 1;
        printk("OK.\n");
 
        /* now that we know it works we can reduce NMI frequency to
@@ -137,6 +170,7 @@ static int __init check_nmi_watchdog(void)
        if (nmi_watchdog == NMI_LOCAL_APIC)
                nmi_hz = 1;
 
+       kfree(prev_nmi_count);
        return 0;
 }
 /* This needs to happen later in boot so counters are working */