hardlockup: detect hard lockups without NMIs using secondary cpus

Emulate NMIs on systems where they are not available by using timer
interrupts on other cpus.  Each cpu will use its softlockup hrtimer
to check that the next cpu is processing hrtimer interrupts by
verifying that a counter is increasing.

This patch is useful on systems where the hardlockup detector is not
available due to a lack of NMIs, for example most ARM SoCs.
Without this patch any cpu stuck with interrupts disabled can
cause a hardware watchdog reset with no debugging information,
but with this patch the kernel can detect the lockup and panic,
which can result in useful debugging info.

Change-Id: Ia5faf50243e19c1755201212e04c8892d929785a
Signed-off-by: Colin Cross <ccross@android.com>

diff --git a/include/linux/nmi.h b/include/linux/nmi.h
index 7ec5b86735f33d512a27086635ac15a221039101..74385351935d4cb20bfceeebd7eff30ed1f6bff7 100644 (file)
--- a/include/linux/nmi.h
+++ b/include/linux/nmi.h
@@ -14,8 +14,11 @@
  * may be used to reset the timeout - for code which intentionally
  * disables interrupts for a long time. This call is stateless.
  */
-#if defined(CONFIG_HAVE_NMI_WATCHDOG) || defined(CONFIG_HARDLOCKUP_DETECTOR)
+#if defined(CONFIG_HAVE_NMI_WATCHDOG) || defined(CONFIG_HARDLOCKUP_DETECTOR_NMI)
 #include <asm/nmi.h>
+#endif
+
+#if defined(CONFIG_HAVE_NMI_WATCHDOG) || defined(CONFIG_HARDLOCKUP_DETECTOR)
 extern void touch_nmi_watchdog(void);
 #else
 static inline void touch_nmi_watchdog(void)

diff --git a/kernel/watchdog.c b/kernel/watchdog.c
index 18f34cf75f741e2a63db6dc6c522dfa2e03d1a54..25955235ecdde0d973141c55d39f20251c52ba17 100644 (file)
--- a/kernel/watchdog.c
+++ b/kernel/watchdog.c
@@ -103,6 +103,11 @@ static DEFINE_PER_CPU(struct task_struct *, softlockup_task_ptr_saved);
 static DEFINE_PER_CPU(bool, hard_watchdog_warn);
 static DEFINE_PER_CPU(bool, watchdog_nmi_touch);
 static DEFINE_PER_CPU(unsigned long, hrtimer_interrupts_saved);
+#endif
+#ifdef CONFIG_HARDLOCKUP_DETECTOR_OTHER_CPU
+static cpumask_t __read_mostly watchdog_cpus;
+#endif
+#ifdef CONFIG_HARDLOCKUP_DETECTOR_NMI
 static DEFINE_PER_CPU(struct perf_event *, watchdog_ev);
 #endif
 static unsigned long soft_lockup_nmi_warn;
@@ -271,7 +276,7 @@ void touch_softlockup_watchdog_sync(void)
        __this_cpu_write(watchdog_touch_ts, 0);
 }
 
-#ifdef CONFIG_HARDLOCKUP_DETECTOR
+#ifdef CONFIG_HARDLOCKUP_DETECTOR_NMI
 /* watchdog detector functions */
 static bool is_hardlockup(void)
 {
@@ -285,6 +290,76 @@ static bool is_hardlockup(void)
 }
 #endif
 
+#ifdef CONFIG_HARDLOCKUP_DETECTOR_OTHER_CPU
+static unsigned int watchdog_next_cpu(unsigned int cpu)
+{
+       cpumask_t cpus = watchdog_cpus;
+       unsigned int next_cpu;
+
+       next_cpu = cpumask_next(cpu, &cpus);
+       if (next_cpu >= nr_cpu_ids)
+               next_cpu = cpumask_first(&cpus);
+
+       if (next_cpu == cpu)
+               return nr_cpu_ids;
+
+       return next_cpu;
+}
+
+static int is_hardlockup_other_cpu(unsigned int cpu)
+{
+       unsigned long hrint = per_cpu(hrtimer_interrupts, cpu);
+
+       if (per_cpu(hrtimer_interrupts_saved, cpu) == hrint)
+               return 1;
+
+       per_cpu(hrtimer_interrupts_saved, cpu) = hrint;
+       return 0;
+}
+
+static void watchdog_check_hardlockup_other_cpu(void)
+{
+       unsigned int next_cpu;
+
+       /*
+        * Test for hardlockups every 3 samples.  The sample period is
+        *  watchdog_thresh * 2 / 5, so 3 samples gets us back to slightly over
+        *  watchdog_thresh (over by 20%).
+        */
+       if (__this_cpu_read(hrtimer_interrupts) % 3 != 0)
+               return;
+
+       /* check for a hardlockup on the next cpu */
+       next_cpu = watchdog_next_cpu(smp_processor_id());
+       if (next_cpu >= nr_cpu_ids)
+               return;
+
+       smp_rmb();
+
+       if (per_cpu(watchdog_nmi_touch, next_cpu) == true) {
+               per_cpu(watchdog_nmi_touch, next_cpu) = false;
+               return;
+       }
+
+       if (is_hardlockup_other_cpu(next_cpu)) {
+               /* only warn once */
+               if (per_cpu(hard_watchdog_warn, next_cpu) == true)
+                       return;
+
+               if (hardlockup_panic)
+                       panic("Watchdog detected hard LOCKUP on cpu %u", next_cpu);
+               else
+                       WARN(1, "Watchdog detected hard LOCKUP on cpu %u", next_cpu);
+
+               per_cpu(hard_watchdog_warn, next_cpu) = true;
+       } else {
+               per_cpu(hard_watchdog_warn, next_cpu) = false;
+       }
+}
+#else
+static inline void watchdog_check_hardlockup_other_cpu(void) { return; }
+#endif
+
 static int is_softlockup(unsigned long touch_ts)
 {
        unsigned long now = get_timestamp();
@@ -297,7 +372,7 @@ static int is_softlockup(unsigned long touch_ts)
        return 0;
 }
 
-#ifdef CONFIG_HARDLOCKUP_DETECTOR
+#ifdef CONFIG_HARDLOCKUP_DETECTOR_NMI
 
 static struct perf_event_attr wd_hw_attr = {
        .type           = PERF_TYPE_HARDWARE,
@@ -360,7 +435,7 @@ static void watchdog_overflow_callback(struct perf_event *event,
        __this_cpu_write(hard_watchdog_warn, false);
        return;
 }
-#endif /* CONFIG_HARDLOCKUP_DETECTOR */
+#endif /* CONFIG_HARDLOCKUP_DETECTOR_NMI */
 
 static void watchdog_interrupt_count(void)
 {
@@ -384,6 +459,9 @@ static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer)
        /* kick the hardlockup detector */
        watchdog_interrupt_count();
 
+       /* test for hardlockups on the next cpu */
+       watchdog_check_hardlockup_other_cpu();
+
        /* kick the softlockup detector */
        wake_up_process(__this_cpu_read(softlockup_watchdog));
 
@@ -561,7 +639,7 @@ static void watchdog(unsigned int cpu)
                watchdog_nmi_disable(cpu);
 }
 
-#ifdef CONFIG_HARDLOCKUP_DETECTOR
+#ifdef CONFIG_HARDLOCKUP_DETECTOR_NMI
 /*
  * People like the simple clean cpu node info on boot.
  * Reduce the watchdog noise by only printing messages
@@ -659,10 +737,45 @@ static void watchdog_nmi_disable(unsigned int cpu)
        }
 }
 
+#else
+#ifdef CONFIG_HARDLOCKUP_DETECTOR_OTHER_CPU
+static int watchdog_nmi_enable(unsigned int cpu)
+{
+       /*
+        * The new cpu will be marked online before the first hrtimer interrupt
+        * runs on it.  If another cpu tests for a hardlockup on the new cpu
+        * before it has run its first hrtimer, it will get a false positive.
+        * Touch the watchdog on the new cpu to delay the first check for at
+        * least 3 sampling periods to guarantee one hrtimer has run on the new
+        * cpu.
+        */
+       per_cpu(watchdog_nmi_touch, cpu) = true;
+       smp_wmb();
+       cpumask_set_cpu(cpu, &watchdog_cpus);
+       return 0;
+}
+
+static void watchdog_nmi_disable(unsigned int cpu)
+{
+       unsigned int next_cpu = watchdog_next_cpu(cpu);
+
+       /*
+        * Offlining this cpu will cause the cpu before this one to start
+        * checking the one after this one.  If this cpu just finished checking
+        * the next cpu and updating hrtimer_interrupts_saved, and then the
+        * previous cpu checks it within one sample period, it will trigger a
+        * false positive.  Touch the watchdog on the next cpu to prevent it.
+        */
+       if (next_cpu < nr_cpu_ids)
+               per_cpu(watchdog_nmi_touch, next_cpu) = true;
+       smp_wmb();
+       cpumask_clear_cpu(cpu, &watchdog_cpus);
+}
 #else
 static int watchdog_nmi_enable(unsigned int cpu) { return 0; }
 static void watchdog_nmi_disable(unsigned int cpu) { return; }
-#endif /* CONFIG_HARDLOCKUP_DETECTOR */
+#endif /* CONFIG_HARDLOCKUP_DETECTOR_OTHER_CPU */
+#endif /* CONFIG_HARDLOCKUP_DETECTOR_NMI */
 
 static struct smp_hotplug_thread watchdog_threads = {
        .store                  = &softlockup_watchdog,

diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 8c15b29d5adc64214880cf7c066fb83349f612aa..fc9f2adc0c9b26b830e0e572c01602c70d871724 100644 (file)
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -707,15 +707,27 @@ config LOCKUP_DETECTOR
          The overhead should be minimal.  A periodic hrtimer runs to
          generate interrupts and kick the watchdog task every 4 seconds.
          An NMI is generated every 10 seconds or so to check for hardlockups.
+         If NMIs are not available on the platform, every 12 seconds the
+         hrtimer interrupt on one cpu will be used to check for hardlockups
+         on the next cpu.
 
          The frequency of hrtimer and NMI events and the soft and hard lockup
          thresholds can be controlled through the sysctl watchdog_thresh.
 
-config HARDLOCKUP_DETECTOR
+config HARDLOCKUP_DETECTOR_NMI
        def_bool y
        depends on LOCKUP_DETECTOR && !HAVE_NMI_WATCHDOG
        depends on PERF_EVENTS && HAVE_PERF_EVENTS_NMI
 
+config HARDLOCKUP_DETECTOR_OTHER_CPU
+       def_bool y
+       depends on LOCKUP_DETECTOR && SMP
+       depends on !HARDLOCKUP_DETECTOR_NMI && !HAVE_NMI_WATCHDOG
+
+config HARDLOCKUP_DETECTOR
+       def_bool y
+       depends on HARDLOCKUP_DETECTOR_NMI || HARDLOCKUP_DETECTOR_OTHER_CPU
+
 config BOOTPARAM_HARDLOCKUP_PANIC
        bool "Panic (Reboot) On Hard Lockups"
        depends on HARDLOCKUP_DETECTOR