10 files changed, 171 insertions, 20 deletions

diff --git a/kernel/hrtimer.c b/kernel/hrtimer.c
index cb8a15c19583..b675a67c9ac3 100644
--- a/kernel/hrtimer.c
+++ b/kernel/hrtimer.c
@@ -43,6 +43,8 @@
 #include <linux/seq_file.h>
 #include <linux/err.h>
 #include <linux/debugobjects.h>
+#include <linux/sched.h>
+#include <linux/timer.h>
 
 #include <asm/uaccess.h>
 
@@ -193,12 +195,24 @@ struct hrtimer_clock_base *lock_hrtimer_base(const struct hrtimer *timer,
  * Switch the timer base to the current CPU when possible.
  */
 static inline struct hrtimer_clock_base *
-switch_hrtimer_base(struct hrtimer *timer, struct hrtimer_clock_base *base)
+switch_hrtimer_base(struct hrtimer *timer, struct hrtimer_clock_base *base,
+                    int pinned)
 {
         struct hrtimer_clock_base *new_base;
         struct hrtimer_cpu_base *new_cpu_base;
+        int cpu, preferred_cpu = -1;
+
+        cpu = smp_processor_id();
+#if defined(CONFIG_NO_HZ) && defined(CONFIG_SMP)
+        if (!pinned && get_sysctl_timer_migration() && idle_cpu(cpu)) {
+                preferred_cpu = get_nohz_load_balancer();
+                if (preferred_cpu >= 0)
+                        cpu = preferred_cpu;
+        }
+#endif
 
-        new_cpu_base = &__get_cpu_var(hrtimer_bases);
+again:
+        new_cpu_base = &per_cpu(hrtimer_bases, cpu);
         new_base = &new_cpu_base->clock_base[base->index];
 
         if (base != new_base) {
@@ -218,6 +232,40 @@ switch_hrtimer_base(struct hrtimer *timer, struct hrtimer_clock_base *base)
                 timer->base = NULL;
                 spin_unlock(&base->cpu_base->lock);
                 spin_lock(&new_base->cpu_base->lock);
+
+                /* Optimized away for NOHZ=n SMP=n */
+                if (cpu == preferred_cpu) {
+                        /* Calculate clock monotonic expiry time */
+#ifdef CONFIG_HIGH_RES_TIMERS
+                        ktime_t expires = ktime_sub(hrtimer_get_expires(timer),
+                                                        new_base->offset);
+#else
+                        ktime_t expires = hrtimer_get_expires(timer);
+#endif
+
+                        /*
+                         * Get the next event on target cpu from the
+                         * clock events layer.
+                         * This covers the highres=off nohz=on case as well.
+                         */
+                        ktime_t next = clockevents_get_next_event(cpu);
+
+                        ktime_t delta = ktime_sub(expires, next);
+
+                        /*
+                         * We do not migrate the timer when it is expiring
+                         * before the next event on the target cpu because
+                         * we cannot reprogram the target cpu hardware and
+                         * we would cause it to fire late.
+                         */
+                        if (delta.tv64 < 0) {
+                                cpu = smp_processor_id();
+                                spin_unlock(&new_base->cpu_base->lock);
+                                spin_lock(&base->cpu_base->lock);
+                                timer->base = base;
+                                goto again;
+                        }
+                }
                 timer->base = new_base;
         }
         return new_base;
@@ -235,7 +283,7 @@ lock_hrtimer_base(const struct hrtimer *timer, unsigned long *flags)
         return base;
 }
 
-# define switch_hrtimer_base(t, b)      (b)
+# define switch_hrtimer_base(t, b, p)   (b)
 
 #endif  /* !CONFIG_SMP */
 
@@ -907,9 +955,9 @@ int __hrtimer_start_range_ns(struct hrtimer *timer, ktime_t tim,
         ret = remove_hrtimer(timer, base);
 
         /* Switch the timer base, if necessary: */
-        new_base = switch_hrtimer_base(timer, base);
+        new_base = switch_hrtimer_base(timer, base, mode & HRTIMER_MODE_PINNED);
 
-        if (mode == HRTIMER_MODE_REL) {
+        if (mode & HRTIMER_MODE_REL) {
                 tim = ktime_add_safe(tim, new_base->get_time());
                 /*
                  * CONFIG_TIME_LOW_RES is a temporary way for architectures
diff --git a/kernel/sched.c b/kernel/sched.c
index 8ec9d13140be..8fb88a906aaa 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -240,7 +240,7 @@ static void start_rt_bandwidth(struct rt_bandwidth *rt_b)
                 hard = hrtimer_get_expires(&rt_b->rt_period_timer);
                 delta = ktime_to_ns(ktime_sub(hard, soft));
                 __hrtimer_start_range_ns(&rt_b->rt_period_timer, soft, delta,
-                                HRTIMER_MODE_ABS, 0);
+                                HRTIMER_MODE_ABS_PINNED, 0);
         }
         spin_unlock(&rt_b->rt_runtime_lock);
 }
@@ -1155,7 +1155,7 @@ static __init void init_hrtick(void)
 static void hrtick_start(struct rq *rq, u64 delay)
 {
         __hrtimer_start_range_ns(&rq->hrtick_timer, ns_to_ktime(delay), 0,
-                        HRTIMER_MODE_REL, 0);
+                        HRTIMER_MODE_REL_PINNED, 0);
 }
 
 static inline void init_hrtick(void)
@@ -4397,6 +4397,11 @@ static struct {
         .load_balancer = ATOMIC_INIT(-1),
 };
 
+int get_nohz_load_balancer(void)
+{
+        return atomic_read(&nohz.load_balancer);
+}
+
 #if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT)
 /**
  * lowest_flag_domain - Return lowest sched_domain containing flag.
@@ -9029,6 +9034,8 @@ void __init sched_init_smp(void)
 }
 #endif /* CONFIG_SMP */
 
+const_debug unsigned int sysctl_timer_migration = 1;
+
 int in_sched_functions(unsigned long addr)
 {
         return in_lock_functions(addr) ||
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index ce664f98e3fb..0e51a35a4486 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -328,6 +328,14 @@ static struct ctl_table kern_table[] = {
                 .mode           = 0644,
                 .proc_handler   = &proc_dointvec,
         },
+        {
+                .ctl_name       = CTL_UNNUMBERED,
+                .procname       = "timer_migration",
+                .data           = &sysctl_timer_migration,
+                .maxlen         = sizeof(unsigned int),
+                .mode           = 0644,
+                .proc_handler   = &proc_dointvec,
+        },
 #endif
         {
                 .ctl_name       = CTL_UNNUMBERED,
diff --git a/kernel/time/clockevents.c b/kernel/time/clockevents.c
index d13be216a790..1ad6dd461119 100644
--- a/kernel/time/clockevents.c
+++ b/kernel/time/clockevents.c
@@ -18,6 +18,7 @@
 #include <linux/notifier.h>
 #include <linux/smp.h>
 #include <linux/sysdev.h>
+#include <linux/tick.h>
 
 /* The registered clock event devices */
 static LIST_HEAD(clockevent_devices);
@@ -54,6 +55,7 @@ unsigned long clockevent_delta2ns(unsigned long latch,
 
         return (unsigned long) clc;
 }
+EXPORT_SYMBOL_GPL(clockevent_delta2ns);
 
 /**
  * clockevents_set_mode - set the operating mode of a clock event device
@@ -187,6 +189,7 @@ void clockevents_register_device(struct clock_event_device *dev)
 
         spin_unlock(&clockevents_lock);
 }
+EXPORT_SYMBOL_GPL(clockevents_register_device);
 
 /*
  * Noop handler when we shut down an event device
@@ -251,4 +254,15 @@ void clockevents_notify(unsigned long reason, void *arg)
         spin_unlock(&clockevents_lock);
 }
 EXPORT_SYMBOL_GPL(clockevents_notify);
+
+ktime_t clockevents_get_next_event(int cpu)
+{
+        struct tick_device *td;
+        struct clock_event_device *dev;
+
+        td = &per_cpu(tick_cpu_device, cpu);
+        dev = td->evtdev;
+
+        return dev->next_event;
+}
 #endif
diff --git a/kernel/time/clocksource.c b/kernel/time/clocksource.c
index 80189f6f1c5a..592bf584d1d2 100644
--- a/kernel/time/clocksource.c
+++ b/kernel/time/clocksource.c
@@ -509,6 +509,18 @@ static ssize_t sysfs_override_clocksource(struct sys_device *dev,
                 }
         }
 
+        /*
+         * Check to make sure we don't switch to a non-highres capable
+         * clocksource if the tick code is in oneshot mode (highres or nohz)
+         */
+        if (tick_oneshot_mode_active() &&
+            !(ovr->flags & CLOCK_SOURCE_VALID_FOR_HRES)) {
+                printk(KERN_WARNING "%s clocksource is not HRT compatible. "
+                        "Cannot switch while in HRT/NOHZ mode\n", ovr->name);
+                ovr = NULL;
+                override_name[0] = 0;
+        }
+
         /* Reselect, when the override name has changed */
         if (ovr != clocksource_override) {
                 clocksource_override = ovr;
@@ -537,7 +549,13 @@ sysfs_show_available_clocksources(struct sys_device *dev,
 
         spin_lock_irq(&clocksource_lock);
         list_for_each_entry(src, &clocksource_list, list) {
-                count += snprintf(buf + count,
+                /*
+                 * Don't show non-HRES clocksource if the tick code is
+                 * in one shot mode (highres=on or nohz=on)
+                 */
+                if (!tick_oneshot_mode_active() ||
+                    (src->flags & CLOCK_SOURCE_VALID_FOR_HRES))
+                        count += snprintf(buf + count,
                                   max((ssize_t)PAGE_SIZE - count, (ssize_t)0),
                                   "%s ", src->name);
         }
diff --git a/kernel/time/tick-broadcast.c b/kernel/time/tick-broadcast.c
index 118a3b3b3f9a..877dbedc3118 100644
--- a/kernel/time/tick-broadcast.c
+++ b/kernel/time/tick-broadcast.c
@@ -27,7 +27,7 @@
  * timer stops in C3 state.
  */
 
-struct tick_device tick_broadcast_device;
+static struct tick_device tick_broadcast_device;
 /* FIXME: Use cpumask_var_t. */
 static DECLARE_BITMAP(tick_broadcast_mask, NR_CPUS);
 static DECLARE_BITMAP(tmpmask, NR_CPUS);
diff --git a/kernel/time/tick-oneshot.c b/kernel/time/tick-oneshot.c
index 2e8de678e767..a96c0e2b89cf 100644
--- a/kernel/time/tick-oneshot.c
+++ b/kernel/time/tick-oneshot.c
@@ -128,6 +128,23 @@ int tick_switch_to_oneshot(void (*handler)(struct clock_event_device *))
         return 0;
 }
 
+/**
+ * tick_check_oneshot_mode - check whether the system is in oneshot mode
+ *
+ * returns 1 when either nohz or highres are enabled. otherwise 0.
+ */
+int tick_oneshot_mode_active(void)
+{
+        unsigned long flags;
+        int ret;
+
+        local_irq_save(flags);
+        ret = __get_cpu_var(tick_cpu_device).mode == TICKDEV_MODE_ONESHOT;
+        local_irq_restore(flags);
+
+        return ret;
+}
+
 #ifdef CONFIG_HIGH_RES_TIMERS
 /**
  * tick_init_highres - switch to high resolution mode
diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index d3f1ef4d5cbe..2aff39c6f10c 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -349,7 +349,7 @@ void tick_nohz_stop_sched_tick(int inidle)
 
                 if (ts->nohz_mode == NOHZ_MODE_HIGHRES) {
                         hrtimer_start(&ts->sched_timer, expires,
-                                      HRTIMER_MODE_ABS);
+                                      HRTIMER_MODE_ABS_PINNED);
                         /* Check, if the timer was already in the past */
                         if (hrtimer_active(&ts->sched_timer))
                                 goto out;
@@ -395,7 +395,7 @@ static void tick_nohz_restart(struct tick_sched *ts, ktime_t now)
 
                 if (ts->nohz_mode == NOHZ_MODE_HIGHRES) {
                         hrtimer_start_expires(&ts->sched_timer,
-                                      HRTIMER_MODE_ABS);
+                                              HRTIMER_MODE_ABS_PINNED);
                         /* Check, if the timer was already in the past */
                         if (hrtimer_active(&ts->sched_timer))
                                 break;
@@ -698,7 +698,8 @@ void tick_setup_sched_timer(void)
 
         for (;;) {
                 hrtimer_forward(&ts->sched_timer, now, tick_period);
-                hrtimer_start_expires(&ts->sched_timer, HRTIMER_MODE_ABS);
+                hrtimer_start_expires(&ts->sched_timer,
+                                      HRTIMER_MODE_ABS_PINNED);
                 /* Check, if the timer was already in the past */
                 if (hrtimer_active(&ts->sched_timer))
                         break;
diff --git a/kernel/timer.c b/kernel/timer.c
index faf2db897de4..54d3912f8cad 100644
--- a/kernel/timer.c
+++ b/kernel/timer.c
@@ -38,6 +38,7 @@
 #include <linux/tick.h>
 #include <linux/kallsyms.h>
 #include <linux/perf_counter.h>
+#include <linux/sched.h>
 
 #include <asm/uaccess.h>
 #include <asm/unistd.h>
@@ -605,13 +606,12 @@ static struct tvec_base *lock_timer_base(struct timer_list *timer,
 }
 
 static inline int
-__mod_timer(struct timer_list *timer, unsigned long expires, bool pending_only)
+__mod_timer(struct timer_list *timer, unsigned long expires,
+                                                bool pending_only, int pinned)
 {
         struct tvec_base *base, *new_base;
         unsigned long flags;
-        int ret;
-
-        ret = 0;
+        int ret = 0 , cpu;
 
         timer_stats_timer_set_start_info(timer);
         BUG_ON(!timer->function);
@@ -630,6 +630,18 @@ __mod_timer(struct timer_list *timer, unsigned long expires, bool pending_only)
 
         new_base = __get_cpu_var(tvec_bases);
 
+        cpu = smp_processor_id();
+
+#if defined(CONFIG_NO_HZ) && defined(CONFIG_SMP)
+        if (!pinned && get_sysctl_timer_migration() && idle_cpu(cpu)) {
+                int preferred_cpu = get_nohz_load_balancer();
+
+                if (preferred_cpu >= 0)
+                        cpu = preferred_cpu;
+        }
+#endif
+        new_base = per_cpu(tvec_bases, cpu);
+
         if (base != new_base) {
                 /*
                  * We are trying to schedule the timer on the local CPU.
@@ -669,7 +681,7 @@ out_unlock:
  */
 int mod_timer_pending(struct timer_list *timer, unsigned long expires)
 {
-        return __mod_timer(timer, expires, true);
+        return __mod_timer(timer, expires, true, TIMER_NOT_PINNED);
 }
 EXPORT_SYMBOL(mod_timer_pending);
 
@@ -703,11 +715,33 @@ int mod_timer(struct timer_list *timer, unsigned long expires)
         if (timer->expires == expires && timer_pending(timer))
                 return 1;
 
-        return __mod_timer(timer, expires, false);
+        return __mod_timer(timer, expires, false, TIMER_NOT_PINNED);
 }
 EXPORT_SYMBOL(mod_timer);
 
 /**
+ * mod_timer_pinned - modify a timer's timeout
+ * @timer: the timer to be modified
+ * @expires: new timeout in jiffies
+ *
+ * mod_timer_pinned() is a way to update the expire field of an
+ * active timer (if the timer is inactive it will be activated)
+ * and not allow the timer to be migrated to a different CPU.
+ *
+ * mod_timer_pinned(timer, expires) is equivalent to:
+ *
+ *     del_timer(timer); timer->expires = expires; add_timer(timer);
+ */
+int mod_timer_pinned(struct timer_list *timer, unsigned long expires)
+{
+        if (timer->expires == expires && timer_pending(timer))
+                return 1;
+
+        return __mod_timer(timer, expires, false, TIMER_PINNED);
+}
+EXPORT_SYMBOL(mod_timer_pinned);
+
+/**
  * add_timer - start a timer
  * @timer: the timer to be added
  *
@@ -1017,6 +1051,9 @@ cascade:
                 index = slot = timer_jiffies & TVN_MASK;
                 do {
                         list_for_each_entry(nte, varp->vec + slot, entry) {
+                                if (tbase_get_deferrable(nte->base))
+                                        continue;
+
                                 found = 1;
                                 if (time_before(nte->expires, expires))
                                         expires = nte->expires;
@@ -1307,7 +1344,7 @@ signed long __sched schedule_timeout(signed long timeout)
         expire = timeout + jiffies;
 
         setup_timer_on_stack(&timer, process_timeout, (unsigned long)current);
-        __mod_timer(&timer, expire, false);
+        __mod_timer(&timer, expire, false, TIMER_NOT_PINNED);
         schedule();
         del_singleshot_timer_sync(&timer);
 
diff --git a/kernel/trace/trace_sysprof.c b/kernel/trace/trace_sysprof.c
index e04b76cc238a..f6693969287d 100644
--- a/kernel/trace/trace_sysprof.c
+++ b/kernel/trace/trace_sysprof.c
@@ -203,7 +203,8 @@ static void start_stack_timer(void *unused)
         hrtimer_init(hrtimer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
         hrtimer->function = stack_trace_timer_fn;
 
-        hrtimer_start(hrtimer, ns_to_ktime(sample_period), HRTIMER_MODE_REL);
+        hrtimer_start(hrtimer, ns_to_ktime(sample_period),
+                      HRTIMER_MODE_REL_PINNED);
 }
 
 static void start_stack_timers(void)