Merge branch 'timers/nohz-v3' of git://git.kernel.org/pub/scm/linux/kernel/git/frederic/linux-dynticks into timers/nohz

Pull nohz improvements from Frederic Weisbecker:

 " It mostly contains fixes and full dynticks off-case optimizations. I believe that
   distros want to enable this feature so it seems important to optimize the case
   where the "nohz_full=" parameter is empty. ie: I'm trying to remove any performance
   regression that comes with NO_HZ_FULL=y when the feature is not used.

   This patchset improves the current situation a lot (off-case appears to be around 11% faster
   with hackbench, although I guess it may vary depending on the configuration but it should be
   significantly faster in any case) now there is still some work to do: I can still observe a
   remaining loss of 1.6% throughput seen with hackbench compared to CONFIG_NO_HZ_FULL=n. "

Signed-off-by: Ingo Molnar <mingo@kernel.org>

6 files changed, 116 insertions, 128 deletions

diff --git a/kernel/context_tracking.c b/kernel/context_tracking.c
index 383f8231e436..247091bf0587 100644
--- a/kernel/context_tracking.c
+++ b/kernel/context_tracking.c
@@ -20,22 +20,33 @@
 #include <linux/hardirq.h>
 #include <linux/export.h>
 
-DEFINE_PER_CPU(struct context_tracking, context_tracking) = {
-#ifdef CONFIG_CONTEXT_TRACKING_FORCE
-	.active = true,
-#endif
-};
+#define CREATE_TRACE_POINTS
+#include <trace/events/context_tracking.h>
+
+struct static_key context_tracking_enabled = STATIC_KEY_INIT_FALSE;
+EXPORT_SYMBOL_GPL(context_tracking_enabled);
+
+DEFINE_PER_CPU(struct context_tracking, context_tracking);
+EXPORT_SYMBOL_GPL(context_tracking);
+
+void context_tracking_cpu_set(int cpu)
+{
+	if (!per_cpu(context_tracking.active, cpu)) {
+		per_cpu(context_tracking.active, cpu) = true;
+		static_key_slow_inc(&context_tracking_enabled);
+	}
+}
 
 /**
- * user_enter - Inform the context tracking that the CPU is going to
- *              enter userspace mode.
+ * context_tracking_user_enter - Inform the context tracking that the CPU is going to
+ *                               enter userspace mode.
  *
  * This function must be called right before we switch from the kernel
  * to userspace, when it's guaranteed the remaining kernel instructions
  * to execute won't use any RCU read side critical section because this
  * function sets RCU in extended quiescent state.
  */
-void user_enter(void)
+void context_tracking_user_enter(void)
 {
 	unsigned long flags;
 
@@ -54,17 +65,32 @@ void user_enter(void)
 	WARN_ON_ONCE(!current->mm);
 
 	local_irq_save(flags);
-	if (__this_cpu_read(context_tracking.active) &&
-	    __this_cpu_read(context_tracking.state) != IN_USER) {
+	if ( __this_cpu_read(context_tracking.state) != IN_USER) {
+		if (__this_cpu_read(context_tracking.active)) {
+			trace_user_enter(0);
+			/*
+			 * At this stage, only low level arch entry code remains and
+			 * then we'll run in userspace. We can assume there won't be
+			 * any RCU read-side critical section until the next call to
+			 * user_exit() or rcu_irq_enter(). Let's remove RCU's dependency
+			 * on the tick.
+			 */
+			vtime_user_enter(current);
+			rcu_user_enter();
+		}
 		/*
-		 * At this stage, only low level arch entry code remains and
-		 * then we'll run in userspace. We can assume there won't be
-		 * any RCU read-side critical section until the next call to
-		 * user_exit() or rcu_irq_enter(). Let's remove RCU's dependency
-		 * on the tick.
+		 * Even if context tracking is disabled on this CPU, because it's outside
+		 * the full dynticks mask for example, we still have to keep track of the
+		 * context transitions and states to prevent inconsistency on those of
+		 * other CPUs.
+		 * If a task triggers an exception in userspace, sleep on the exception
+		 * handler and then migrate to another CPU, that new CPU must know where
+		 * the exception returns by the time we call exception_exit().
+		 * This information can only be provided by the previous CPU when it called
+		 * exception_enter().
+		 * OTOH we can spare the calls to vtime and RCU when context_tracking.active
+		 * is false because we know that CPU is not tickless.
 		 */
-		vtime_user_enter(current);
-		rcu_user_enter();
 		__this_cpu_write(context_tracking.state, IN_USER);
 	}
 	local_irq_restore(flags);
@@ -87,10 +113,9 @@ void user_enter(void)
  */
 void __sched notrace preempt_schedule_context(void)
 {
-	struct thread_info *ti = current_thread_info();
 	enum ctx_state prev_ctx;
 
-	if (likely(ti->preempt_count || irqs_disabled()))
+	if (likely(!preemptible()))
 		return;
 
 	/*
@@ -112,8 +137,8 @@ EXPORT_SYMBOL_GPL(preempt_schedule_context);
 #endif /* CONFIG_PREEMPT */
 
 /**
- * user_exit - Inform the context tracking that the CPU is
- *             exiting userspace mode and entering the kernel.
+ * context_tracking_user_exit - Inform the context tracking that the CPU is
+ *                              exiting userspace mode and entering the kernel.
  *
  * This function must be called after we entered the kernel from userspace
  * before any use of RCU read side critical section. This potentially include
@@ -122,7 +147,7 @@ EXPORT_SYMBOL_GPL(preempt_schedule_context);
  * This call supports re-entrancy. This way it can be called from any exception
  * handler without needing to know if we came from userspace or not.
  */
-void user_exit(void)
+void context_tracking_user_exit(void)
 {
 	unsigned long flags;
 
@@ -131,38 +156,22 @@ void user_exit(void)
 
 	local_irq_save(flags);
 	if (__this_cpu_read(context_tracking.state) == IN_USER) {
-		/*
-		 * We are going to run code that may use RCU. Inform
-		 * RCU core about that (ie: we may need the tick again).
-		 */
-		rcu_user_exit();
-		vtime_user_exit(current);
+		if (__this_cpu_read(context_tracking.active)) {
+			/*
+			 * We are going to run code that may use RCU. Inform
+			 * RCU core about that (ie: we may need the tick again).
+			 */
+			rcu_user_exit();
+			vtime_user_exit(current);
+			trace_user_exit(0);
+		}
 		__this_cpu_write(context_tracking.state, IN_KERNEL);
 	}
 	local_irq_restore(flags);
 }
 
-void guest_enter(void)
-{
-	if (vtime_accounting_enabled())
-		vtime_guest_enter(current);
-	else
-		__guest_enter();
-}
-EXPORT_SYMBOL_GPL(guest_enter);
-
-void guest_exit(void)
-{
-	if (vtime_accounting_enabled())
-		vtime_guest_exit(current);
-	else
-		__guest_exit();
-}
-EXPORT_SYMBOL_GPL(guest_exit);
-
-
 /**
- * context_tracking_task_switch - context switch the syscall callbacks
+ * __context_tracking_task_switch - context switch the syscall callbacks
  * @prev: the task that is being switched out
  * @next: the task that is being switched in
  *
@@ -174,11 +183,19 @@ EXPORT_SYMBOL_GPL(guest_exit);
  * migrate to some CPU that doesn't do the context tracking. As such the TIF
  * flag may not be desired there.
  */
-void context_tracking_task_switch(struct task_struct *prev,
-			     struct task_struct *next)
+void __context_tracking_task_switch(struct task_struct *prev,
+				    struct task_struct *next)
 {
-	if (__this_cpu_read(context_tracking.active)) {
-		clear_tsk_thread_flag(prev, TIF_NOHZ);
-		set_tsk_thread_flag(next, TIF_NOHZ);
-	}
+	clear_tsk_thread_flag(prev, TIF_NOHZ);
+	set_tsk_thread_flag(next, TIF_NOHZ);
 }
+
+#ifdef CONFIG_CONTEXT_TRACKING_FORCE
+void __init context_tracking_init(void)
+{
+	int cpu;
+
+	for_each_possible_cpu(cpu)
+		context_tracking_cpu_set(cpu);
+}
+#endif
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index b7c32cb7bfeb..3fb7acee7326 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -2510,13 +2510,11 @@ void __sched schedule_preempt_disabled(void)
  */
 asmlinkage void __sched notrace preempt_schedule(void)
 {
-	struct thread_info *ti = current_thread_info();
-
 	/*
 	 * If there is a non-zero preempt_count or interrupts are disabled,
 	 * we do not want to preempt the current task. Just return..
 	 */
-	if (likely(ti->preempt_count || irqs_disabled()))
+	if (likely(!preemptible()))
 		return;
 
 	do {
diff --git a/kernel/sched/cputime.c b/kernel/sched/cputime.c
index a7959e05a9d5..c1d7493825ae 100644
--- a/kernel/sched/cputime.c
+++ b/kernel/sched/cputime.c
@@ -378,11 +378,8 @@ static inline void irqtime_account_process_tick(struct task_struct *p, int user_
 #ifdef CONFIG_VIRT_CPU_ACCOUNTING
 
 #ifndef __ARCH_HAS_VTIME_TASK_SWITCH
-void vtime_task_switch(struct task_struct *prev)
+void vtime_common_task_switch(struct task_struct *prev)
 {
-	if (!vtime_accounting_enabled())
-		return;
-
 	if (is_idle_task(prev))
 		vtime_account_idle(prev);
 	else
@@ -404,11 +401,8 @@ void vtime_task_switch(struct task_struct *prev)
  * vtime_account().
  */
 #ifndef __ARCH_HAS_VTIME_ACCOUNT
-void vtime_account_irq_enter(struct task_struct *tsk)
+void vtime_common_account_irq_enter(struct task_struct *tsk)
 {
-	if (!vtime_accounting_enabled())
-		return;
-
 	if (!in_interrupt()) {
 		/*
 		 * If we interrupted user, context_tracking_in_user()
@@ -428,7 +422,7 @@ void vtime_account_irq_enter(struct task_struct *tsk)
 	}
 	vtime_account_system(tsk);
 }
-EXPORT_SYMBOL_GPL(vtime_account_irq_enter);
+EXPORT_SYMBOL_GPL(vtime_common_account_irq_enter);
 #endif /* __ARCH_HAS_VTIME_ACCOUNT */
 #endif /* CONFIG_VIRT_CPU_ACCOUNTING */
 
@@ -559,12 +553,6 @@ static void cputime_adjust(struct task_cputime *curr,
 {
 	cputime_t rtime, stime, utime, total;
 
-	if (vtime_accounting_enabled()) {
-		*ut = curr->utime;
-		*st = curr->stime;
-		return;
-	}
-
 	stime = curr->stime;
 	total = stime + curr->utime;
 
@@ -664,23 +652,17 @@ static void __vtime_account_system(struct task_struct *tsk)
 
 void vtime_account_system(struct task_struct *tsk)
 {
-	if (!vtime_accounting_enabled())
-		return;
-
 	write_seqlock(&tsk->vtime_seqlock);
 	__vtime_account_system(tsk);
 	write_sequnlock(&tsk->vtime_seqlock);
 }
 
-void vtime_account_irq_exit(struct task_struct *tsk)
+void vtime_gen_account_irq_exit(struct task_struct *tsk)
 {
-	if (!vtime_accounting_enabled())
-		return;
-
 	write_seqlock(&tsk->vtime_seqlock);
+	__vtime_account_system(tsk);
 	if (context_tracking_in_user())
 		tsk->vtime_snap_whence = VTIME_USER;
-	__vtime_account_system(tsk);
 	write_sequnlock(&tsk->vtime_seqlock);
 }
 
@@ -688,12 +670,8 @@ void vtime_account_user(struct task_struct *tsk)
 {
 	cputime_t delta_cpu;
 
-	if (!vtime_accounting_enabled())
-		return;
-
-	delta_cpu = get_vtime_delta(tsk);
-
 	write_seqlock(&tsk->vtime_seqlock);
+	delta_cpu = get_vtime_delta(tsk);
 	tsk->vtime_snap_whence = VTIME_SYS;
 	account_user_time(tsk, delta_cpu, cputime_to_scaled(delta_cpu));
 	write_sequnlock(&tsk->vtime_seqlock);
@@ -701,22 +679,27 @@ void vtime_account_user(struct task_struct *tsk)
 
 void vtime_user_enter(struct task_struct *tsk)
 {
-	if (!vtime_accounting_enabled())
-		return;
-
 	write_seqlock(&tsk->vtime_seqlock);
-	tsk->vtime_snap_whence = VTIME_USER;
 	__vtime_account_system(tsk);
+	tsk->vtime_snap_whence = VTIME_USER;
 	write_sequnlock(&tsk->vtime_seqlock);
 }
 
 void vtime_guest_enter(struct task_struct *tsk)
 {
+	/*
+	 * The flags must be updated under the lock with
+	 * the vtime_snap flush and update.
+	 * That enforces a right ordering and update sequence
+	 * synchronization against the reader (task_gtime())
+	 * that can thus safely catch up with a tickless delta.
+	 */
 	write_seqlock(&tsk->vtime_seqlock);
 	__vtime_account_system(tsk);
 	current->flags |= PF_VCPU;
 	write_sequnlock(&tsk->vtime_seqlock);
 }
+EXPORT_SYMBOL_GPL(vtime_guest_enter);
 
 void vtime_guest_exit(struct task_struct *tsk)
 {
@@ -725,6 +708,7 @@ void vtime_guest_exit(struct task_struct *tsk)
 	current->flags &= ~PF_VCPU;
 	write_sequnlock(&tsk->vtime_seqlock);
 }
+EXPORT_SYMBOL_GPL(vtime_guest_exit);
 
 void vtime_account_idle(struct task_struct *tsk)
 {
@@ -733,11 +717,6 @@ void vtime_account_idle(struct task_struct *tsk)
 	account_idle_time(delta_cpu);
 }
 
-bool vtime_accounting_enabled(void)
-{
-	return context_tracking_active();
-}
-
 void arch_vtime_task_switch(struct task_struct *prev)
 {
 	write_seqlock(&prev->vtime_seqlock);
diff --git a/kernel/time/Kconfig b/kernel/time/Kconfig
index 70f27e89012b..747bbc70f53b 100644
--- a/kernel/time/Kconfig
+++ b/kernel/time/Kconfig
@@ -105,7 +105,6 @@ config NO_HZ_FULL
 	select RCU_USER_QS
 	select RCU_NOCB_CPU
 	select VIRT_CPU_ACCOUNTING_GEN
-	select CONTEXT_TRACKING_FORCE
 	select IRQ_WORK
 	help
 	 Adaptively try to shutdown the tick whenever possible, even when
diff --git a/kernel/time/sched_clock.c b/kernel/time/sched_clock.c
index a326f27d7f09..0b479a6a22bb 100644
--- a/kernel/time/sched_clock.c
+++ b/kernel/time/sched_clock.c
@@ -121,7 +121,7 @@ void __init setup_sched_clock(u32 (*read)(void), int bits, unsigned long rate)
 	BUG_ON(bits > 32);
 	WARN_ON(!irqs_disabled());
 	read_sched_clock = read;
-	sched_clock_mask = (1 << bits) - 1;
+	sched_clock_mask = (1ULL << bits) - 1;
 	cd.rate = rate;
 
 	/* calculate the mult/shift to convert counter ticks to ns. */
diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index e77edc97e036..adea6fc3ba2a 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -23,6 +23,7 @@
 #include <linux/irq_work.h>
 #include <linux/posix-timers.h>
 #include <linux/perf_event.h>
+#include <linux/context_tracking.h>
 
 #include <asm/irq_regs.h>
 
@@ -148,8 +149,8 @@ static void tick_sched_handle(struct tick_sched *ts, struct pt_regs *regs)
 }
 
 #ifdef CONFIG_NO_HZ_FULL
-static cpumask_var_t nohz_full_mask;
-bool have_nohz_full_mask;
+cpumask_var_t tick_nohz_full_mask;
+bool tick_nohz_full_running;
 
 static bool can_stop_full_tick(void)
 {
@@ -182,7 +183,8 @@ static bool can_stop_full_tick(void)
 		 * Don't allow the user to think they can get
 		 * full NO_HZ with this machine.
 		 */
-		WARN_ONCE(1, "NO_HZ FULL will not work with unstable sched clock");
+		WARN_ONCE(tick_nohz_full_running,
+			  "NO_HZ FULL will not work with unstable sched clock");
 		return false;
 	}
 #endif
@@ -196,7 +198,7 @@ static void tick_nohz_restart_sched_tick(struct tick_sched *ts, ktime_t now);
  * Re-evaluate the need for the tick on the current CPU
  * and restart it if necessary.
  */
-void tick_nohz_full_check(void)
+void __tick_nohz_full_check(void)
 {
 	struct tick_sched *ts = &__get_cpu_var(tick_cpu_sched);
 
@@ -210,7 +212,7 @@ void tick_nohz_full_check(void)
 
 static void nohz_full_kick_work_func(struct irq_work *work)
 {
-	tick_nohz_full_check();
+	__tick_nohz_full_check();
 }
 
 static DEFINE_PER_CPU(struct irq_work, nohz_full_kick_work) = {
@@ -229,7 +231,7 @@ void tick_nohz_full_kick(void)
 
 static void nohz_full_kick_ipi(void *info)
 {
-	tick_nohz_full_check();
+	__tick_nohz_full_check();
 }
 
 /*
@@ -238,11 +240,11 @@ static void nohz_full_kick_ipi(void *info)
  */
 void tick_nohz_full_kick_all(void)
 {
-	if (!have_nohz_full_mask)
+	if (!tick_nohz_full_running)
 		return;
 
 	preempt_disable();
-	smp_call_function_many(nohz_full_mask,
+	smp_call_function_many(tick_nohz_full_mask,
 			       nohz_full_kick_ipi, NULL, false);
 	preempt_enable();
 }
@@ -252,7 +254,7 @@ void tick_nohz_full_kick_all(void)
  * It might need the tick due to per task/process properties:
  * perf events, posix cpu timers, ...
  */
-void tick_nohz_task_switch(struct task_struct *tsk)
+void __tick_nohz_task_switch(struct task_struct *tsk)
 {
 	unsigned long flags;
 
@@ -268,31 +270,23 @@ out:
 	local_irq_restore(flags);
 }
 
-int tick_nohz_full_cpu(int cpu)
-{
-	if (!have_nohz_full_mask)
-		return 0;
-
-	return cpumask_test_cpu(cpu, nohz_full_mask);
-}
-
 /* Parse the boot-time nohz CPU list from the kernel parameters. */
 static int __init tick_nohz_full_setup(char *str)
 {
 	int cpu;
 
-	alloc_bootmem_cpumask_var(&nohz_full_mask);
-	if (cpulist_parse(str, nohz_full_mask) < 0) {
+	alloc_bootmem_cpumask_var(&tick_nohz_full_mask);
+	if (cpulist_parse(str, tick_nohz_full_mask) < 0) {
 		pr_warning("NOHZ: Incorrect nohz_full cpumask\n");
 		return 1;
 	}
 
 	cpu = smp_processor_id();
-	if (cpumask_test_cpu(cpu, nohz_full_mask)) {
+	if (cpumask_test_cpu(cpu, tick_nohz_full_mask)) {
 		pr_warning("NO_HZ: Clearing %d from nohz_full range for timekeeping\n", cpu);
-		cpumask_clear_cpu(cpu, nohz_full_mask);
+		cpumask_clear_cpu(cpu, tick_nohz_full_mask);
 	}
-	have_nohz_full_mask = true;
+	tick_nohz_full_running = true;
 
 	return 1;
 }
@@ -310,7 +304,7 @@ static int tick_nohz_cpu_down_callback(struct notifier_block *nfb,
 		 * If we handle the timekeeping duty for full dynticks CPUs,
 		 * we can't safely shutdown that CPU.
 		 */
-		if (have_nohz_full_mask && tick_do_timer_cpu == cpu)
+		if (tick_nohz_full_running && tick_do_timer_cpu == cpu)
 			return NOTIFY_BAD;
 		break;
 	}
@@ -329,14 +323,14 @@ static int tick_nohz_init_all(void)
 	int err = -1;
 
 #ifdef CONFIG_NO_HZ_FULL_ALL
-	if (!alloc_cpumask_var(&nohz_full_mask, GFP_KERNEL)) {
+	if (!alloc_cpumask_var(&tick_nohz_full_mask, GFP_KERNEL)) {
 		pr_err("NO_HZ: Can't allocate full dynticks cpumask\n");
 		return err;
 	}
 	err = 0;
-	cpumask_setall(nohz_full_mask);
-	cpumask_clear_cpu(smp_processor_id(), nohz_full_mask);
-	have_nohz_full_mask = true;
+	cpumask_setall(tick_nohz_full_mask);
+	cpumask_clear_cpu(smp_processor_id(), tick_nohz_full_mask);
+	tick_nohz_full_running = true;
 #endif
 	return err;
 }
@@ -345,17 +339,18 @@ void __init tick_nohz_init(void)
 {
 	int cpu;
 
-	if (!have_nohz_full_mask) {
+	if (!tick_nohz_full_running) {
 		if (tick_nohz_init_all() < 0)
 			return;
 	}
 
+	for_each_cpu(cpu, tick_nohz_full_mask)
+		context_tracking_cpu_set(cpu);
+
 	cpu_notifier(tick_nohz_cpu_down_callback, 0);
-	cpulist_scnprintf(nohz_full_buf, sizeof(nohz_full_buf), nohz_full_mask);
+	cpulist_scnprintf(nohz_full_buf, sizeof(nohz_full_buf), tick_nohz_full_mask);
 	pr_info("NO_HZ: Full dynticks CPUs: %s.\n", nohz_full_buf);
 }
-#else
-#define have_nohz_full_mask (0)
 #endif
 
 /*
@@ -733,7 +728,7 @@ static bool can_stop_idle_tick(int cpu, struct tick_sched *ts)
 		return false;
 	}
 
-	if (have_nohz_full_mask) {
+	if (tick_nohz_full_enabled()) {
 		/*
 		 * Keep the tick alive to guarantee timekeeping progression
 		 * if there are full dynticks CPUs around