[PATCH RFC tip/core/rcu 0/5] Consolidate RCU dyntick-idle counter manipulation

[PATCH RFC tip/core/rcu 0/5] Consolidate RCU dyntick-idle counter manipulation

[Paul E. McKenney]

Hello!

The first four patches of this series consolidate RCU's dyntick-idle
counter manipulations into a set of access functions.  These are currently
queued for 4.12 (merge window after next).  The last patch in this series
uses the bottom bit of RCU's dyntick-idle counter to track a CPU's state,
and has been proposed to track the need to flush the local CPU's TLB.
This last patch will not be pushed until it has a use.

The patches are as follows:

1.	Abstract the dynticks momentary-idle operation.

2.	Abstract the dynticks snapshot operation.

3.	Abstract dynticks extended quiescent state enter/exit operations.

4.	Abstract extended quiescent state determination.

5.	Maintain special bits at bottom of ->dynticks counter, which
	might be used to track a given CPU's need to flush TLBs.

							Thanx, Paul

------------------------------------------------------------------------

 include/linux/rcutiny.h  |   11 +
 kernel/rcu/tree.c        |  263 ++++++++++++++++++++++++++++++++++-------------
 kernel/rcu/tree.h        |    3 
 kernel/rcu/tree_exp.h    |   12 --
 kernel/rcu/tree_plugin.h |    2 
 kernel/rcu/tree_trace.c  |    2 
 6 files changed, 216 insertions(+), 77 deletions(-)

[PATCH tip/core/rcu 1/5] rcu: Abstract the dynticks momentary-idle operation

[Paul E. McKenney]

This commit is the first step towards full abstraction of all accesses to
the ->dynticks counter, implementing the previously open-coded atomic add
of two in a new rcu_dynticks_momentary_idle() function.  This abstraction
will ease changes to the ->dynticks counter operation.

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
 kernel/rcu/tree.c | 19 ++++++++++++++-----
 1 file changed, 14 insertions(+), 5 deletions(-)

diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index 96c52e43f7ca..73b2b1da912d 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -278,6 +278,19 @@ static DEFINE_PER_CPU(struct rcu_dynticks, rcu_dynticks) = {
 #endif /* #ifdef CONFIG_NO_HZ_FULL_SYSIDLE */
 };
 
+/*
+ * Do a double-increment of the ->dynticks counter to emulate a
+ * momentary idle-CPU quiescent state.
+ */
+static void rcu_dynticks_momentary_idle(void)
+{
+	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
+	int special = atomic_add_return(2, &rdtp->dynticks);
+
+	/* It is illegal to call this from idle state. */
+	WARN_ON_ONCE(!(special & 0x1));
+}
+
 DEFINE_PER_CPU_SHARED_ALIGNED(unsigned long, rcu_qs_ctr);
 EXPORT_PER_CPU_SYMBOL_GPL(rcu_qs_ctr);
 
@@ -297,7 +310,6 @@ EXPORT_PER_CPU_SYMBOL_GPL(rcu_qs_ctr);
 static void rcu_momentary_dyntick_idle(void)
 {
 	struct rcu_data *rdp;
-	struct rcu_dynticks *rdtp;
 	int resched_mask;
 	struct rcu_state *rsp;
 
@@ -324,10 +336,7 @@ static void rcu_momentary_dyntick_idle(void)
 		 * quiescent state, with no need for this CPU to do anything
 		 * further.
 		 */
-		rdtp = this_cpu_ptr(&rcu_dynticks);
-		smp_mb__before_atomic(); /* Earlier stuff before QS. */
-		atomic_add(2, &rdtp->dynticks);  /* QS. */
-		smp_mb__after_atomic(); /* Later stuff after QS. */
+		rcu_dynticks_momentary_idle();
 		break;
 	}
 }
-- 
2.5.2

[PATCH tip/core/rcu 2/5] rcu: Abstract the dynticks snapshot operation

[Paul E. McKenney]

This commit is the second step towards full abstraction of all accesses to
the ->dynticks counter, implementing the previously open-coded atomic
add of zero in a new rcu_dynticks_snap() function.  This abstraction will
ease changes o the ->dynticks counter operation.

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
 kernel/rcu/tree.c     | 19 ++++++++++++++++---
 kernel/rcu/tree_exp.h |  6 ++----
 2 files changed, 18 insertions(+), 7 deletions(-)

diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index 73b2b1da912d..9bc60b29ea5c 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -279,6 +279,17 @@ static DEFINE_PER_CPU(struct rcu_dynticks, rcu_dynticks) = {
 };
 
 /*
+ * Snapshot the ->dynticks counter with full ordering so as to allow
+ * stable comparison of this counter with past and future snapshots.
+ */
+static int rcu_dynticks_snap(struct rcu_dynticks *rdtp)
+{
+	int snap = atomic_add_return(0, &rdtp->dynticks);
+
+	return snap;
+}
+
+/*
  * Do a double-increment of the ->dynticks counter to emulate a
  * momentary idle-CPU quiescent state.
  */
@@ -1046,7 +1057,9 @@ void rcu_nmi_exit(void)
  */
 bool notrace __rcu_is_watching(void)
 {
-	return atomic_read(this_cpu_ptr(&rcu_dynticks.dynticks)) & 0x1;
+	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
+
+	return atomic_read(&rdtp->dynticks) & 0x1;
 }
 
 /**
@@ -1129,7 +1142,7 @@ static int rcu_is_cpu_rrupt_from_idle(void)
 static int dyntick_save_progress_counter(struct rcu_data *rdp,
 					 bool *isidle, unsigned long *maxj)
 {
-	rdp->dynticks_snap = atomic_add_return(0, &rdp->dynticks->dynticks);
+	rdp->dynticks_snap = rcu_dynticks_snap(rdp->dynticks);
 	rcu_sysidle_check_cpu(rdp, isidle, maxj);
 	if ((rdp->dynticks_snap & 0x1) == 0) {
 		trace_rcu_fqs(rdp->rsp->name, rdp->gpnum, rdp->cpu, TPS("dti"));
@@ -1154,7 +1167,7 @@ static int rcu_implicit_dynticks_qs(struct rcu_data *rdp,
 	int *rcrmp;
 	unsigned int snap;
 
-	curr = (unsigned int)atomic_add_return(0, &rdp->dynticks->dynticks);
+	curr = (unsigned int)rcu_dynticks_snap(rdp->dynticks);
 	snap = (unsigned int)rdp->dynticks_snap;
 
 	/*
diff --git a/kernel/rcu/tree_exp.h b/kernel/rcu/tree_exp.h
index d3053e99fdb6..75f079ac79f9 100644
--- a/kernel/rcu/tree_exp.h
+++ b/kernel/rcu/tree_exp.h
@@ -356,10 +356,9 @@ static void sync_rcu_exp_select_cpus(struct rcu_state *rsp,
 		mask_ofl_test = 0;
 		for_each_leaf_node_possible_cpu(rnp, cpu) {
 			struct rcu_data *rdp = per_cpu_ptr(rsp->rda, cpu);
-			struct rcu_dynticks *rdtp = &per_cpu(rcu_dynticks, cpu);
 
 			rdp->exp_dynticks_snap =
-				atomic_add_return(0, &rdtp->dynticks);
+				rcu_dynticks_snap(rdp->dynticks);
 			if (raw_smp_processor_id() == cpu ||
 			    !(rdp->exp_dynticks_snap & 0x1) ||
 			    !(rnp->qsmaskinitnext & rdp->grpmask))
@@ -380,12 +379,11 @@ static void sync_rcu_exp_select_cpus(struct rcu_state *rsp,
 		for_each_leaf_node_possible_cpu(rnp, cpu) {
 			unsigned long mask = leaf_node_cpu_bit(rnp, cpu);
 			struct rcu_data *rdp = per_cpu_ptr(rsp->rda, cpu);
-			struct rcu_dynticks *rdtp = &per_cpu(rcu_dynticks, cpu);
 
 			if (!(mask_ofl_ipi & mask))
 				continue;
 retry_ipi:
-			if (atomic_add_return(0, &rdtp->dynticks) !=
+			if (rcu_dynticks_snap(rdp->dynticks) !=
 			    rdp->exp_dynticks_snap) {
 				mask_ofl_test |= mask;
 				continue;
-- 
2.5.2

[PATCH tip/core/rcu 3/5] rcu: Abstract dynticks extended quiescent state enter/exit operations

[Paul E. McKenney]

This commit is the third step towards full abstraction of all accesses
to the ->dynticks counter, implementing the previously open-coded atomic
add of 1 and entry checks in a new rcu_dynticks_eqs_enter() function, and
the same but with exit checks in a new rcu_dynticks_eqs_exit() function.
This abstraction will ease changes to the ->dynticks counter operation.

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
 kernel/rcu/tree.c | 92 +++++++++++++++++++++++++++++++++++++++----------------
 1 file changed, 66 insertions(+), 26 deletions(-)

diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index 9bc60b29ea5c..52e844902a43 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -279,6 +279,65 @@ static DEFINE_PER_CPU(struct rcu_dynticks, rcu_dynticks) = {
 };
 
 /*
+ * Record entry into an extended quiescent state.  This is only to be
+ * called when not already in an extended quiescent state.
+ */
+static void rcu_dynticks_eqs_enter(void)
+{
+	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
+
+	/*
+	 * CPUs seeing atomic_inc() must see prior RCU read-side critical
+	 * sections, and we also must force ordering with the next idle
+	 * sojourn.
+	 */
+	smp_mb__before_atomic(); /* See above. */
+	atomic_inc(&rdtp->dynticks);
+	smp_mb__after_atomic(); /* See above. */
+	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
+		     atomic_read(&rdtp->dynticks) & 0x1);
+}
+
+/*
+ * Record exit from an extended quiescent state.  This is only to be
+ * called from an extended quiescent state.
+ */
+static void rcu_dynticks_eqs_exit(void)
+{
+	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
+
+	/*
+	 * CPUs seeing atomic_inc() must see prior idle sojourns,
+	 * and we also must force ordering with the next RCU read-side
+	 * critical section.
+	 */
+	smp_mb__before_atomic(); /* See above. */
+	atomic_inc(&rdtp->dynticks);
+	smp_mb__after_atomic(); /* See above. */
+	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
+		     !(atomic_read(&rdtp->dynticks) & 0x1));
+}
+
+/*
+ * Reset the current CPU's ->dynticks counter to indicate that the
+ * newly onlined CPU is no longer in an extended quiescent state.
+ * This will either leave the counter unchanged, or increment it
+ * to the next non-quiescent value.
+ *
+ * The non-atomic test/increment sequence works because the upper bits
+ * of the ->dynticks counter are manipulated only by the corresponding CPU,
+ * or when the corresponding CPU is offline.
+ */
+static void rcu_dynticks_eqs_online(void)
+{
+	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
+
+	if (atomic_read(&rdtp->dynticks) & 0x1)
+		return;
+	atomic_add(0x1, &rdtp->dynticks);
+}
+
+/*
  * Snapshot the ->dynticks counter with full ordering so as to allow
  * stable comparison of this counter with past and future snapshots.
  */
@@ -690,7 +749,7 @@ static void rcu_eqs_enter_common(long long oldval, bool user)
 {
 	struct rcu_state *rsp;
 	struct rcu_data *rdp;
-	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
+	struct rcu_dynticks __maybe_unused *rdtp = this_cpu_ptr(&rcu_dynticks);
 
 	trace_rcu_dyntick(TPS("Start"), oldval, rdtp->dynticks_nesting);
 	if (IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
@@ -709,12 +768,7 @@ static void rcu_eqs_enter_common(long long oldval, bool user)
 		do_nocb_deferred_wakeup(rdp);
 	}
 	rcu_prepare_for_idle();
-	/* CPUs seeing atomic_inc() must see prior RCU read-side crit sects */
-	smp_mb__before_atomic();  /* See above. */
-	atomic_inc(&rdtp->dynticks);
-	smp_mb__after_atomic();  /* Force ordering with next sojourn. */
-	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
-		     atomic_read(&rdtp->dynticks) & 0x1);
+	rcu_dynticks_eqs_enter();
 	rcu_dynticks_task_enter();
 
 	/*
@@ -843,15 +897,10 @@ void rcu_irq_exit_irqson(void)
  */
 static void rcu_eqs_exit_common(long long oldval, int user)
 {
-	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
+	struct rcu_dynticks __maybe_unused *rdtp = this_cpu_ptr(&rcu_dynticks);
 
 	rcu_dynticks_task_exit();
-	smp_mb__before_atomic();  /* Force ordering w/previous sojourn. */
-	atomic_inc(&rdtp->dynticks);
-	/* CPUs seeing atomic_inc() must see later RCU read-side crit sects */
-	smp_mb__after_atomic();  /* See above. */
-	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
-		     !(atomic_read(&rdtp->dynticks) & 0x1));
+	rcu_dynticks_eqs_exit();
 	rcu_cleanup_after_idle();
 	trace_rcu_dyntick(TPS("End"), oldval, rdtp->dynticks_nesting);
 	if (IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
@@ -998,11 +1047,7 @@ void rcu_nmi_enter(void)
 	 * period (observation due to Andy Lutomirski).
 	 */
 	if (!(atomic_read(&rdtp->dynticks) & 0x1)) {
-		smp_mb__before_atomic();  /* Force delay from prior write. */
-		atomic_inc(&rdtp->dynticks);
-		/* atomic_inc() before later RCU read-side crit sects */
-		smp_mb__after_atomic();  /* See above. */
-		WARN_ON_ONCE(!(atomic_read(&rdtp->dynticks) & 0x1));
+		rcu_dynticks_eqs_exit();
 		incby = 1;
 	}
 	rdtp->dynticks_nmi_nesting += incby;
@@ -1040,11 +1085,7 @@ void rcu_nmi_exit(void)
 
 	/* This NMI interrupted an RCU-idle CPU, restore RCU-idleness. */
 	rdtp->dynticks_nmi_nesting = 0;
-	/* CPUs seeing atomic_inc() must see prior RCU read-side crit sects */
-	smp_mb__before_atomic();  /* See above. */
-	atomic_inc(&rdtp->dynticks);
-	smp_mb__after_atomic();  /* Force delay to next write. */
-	WARN_ON_ONCE(atomic_read(&rdtp->dynticks) & 0x1);
+	rcu_dynticks_eqs_enter();
 }
 
 /**
@@ -3797,8 +3838,7 @@ rcu_init_percpu_data(int cpu, struct rcu_state *rsp)
 		init_callback_list(rdp);  /* Re-enable callbacks on this CPU. */
 	rdp->dynticks->dynticks_nesting = DYNTICK_TASK_EXIT_IDLE;
 	rcu_sysidle_init_percpu_data(rdp->dynticks);
-	atomic_set(&rdp->dynticks->dynticks,
-		   (atomic_read(&rdp->dynticks->dynticks) & ~0x1) + 1);
+	rcu_dynticks_eqs_online();
 	raw_spin_unlock_rcu_node(rnp);		/* irqs remain disabled. */
 
 	/*
-- 
2.5.2

Re: [PATCH tip/core/rcu 3/5] rcu: Abstract dynticks extended quiescent state enter/exit operations

[Boqun Feng]

[-- Attachment #1: Type: text/plain, Size: 6627 bytes --]

Hi Paul, 

On Mon, Nov 14, 2016 at 10:30:19AM -0800, Paul E. McKenney wrote:
> This commit is the third step towards full abstraction of all accesses
> to the ->dynticks counter, implementing the previously open-coded atomic
> add of 1 and entry checks in a new rcu_dynticks_eqs_enter() function, and
> the same but with exit checks in a new rcu_dynticks_eqs_exit() function.
> This abstraction will ease changes to the ->dynticks counter operation.
> 
> Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
> ---
>  kernel/rcu/tree.c | 92 +++++++++++++++++++++++++++++++++++++++----------------
>  1 file changed, 66 insertions(+), 26 deletions(-)
> 
> diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> index 9bc60b29ea5c..52e844902a43 100644
> --- a/kernel/rcu/tree.c
> +++ b/kernel/rcu/tree.c
> @@ -279,6 +279,65 @@ static DEFINE_PER_CPU(struct rcu_dynticks, rcu_dynticks) = {
>  };
>  
>  /*
> + * Record entry into an extended quiescent state.  This is only to be
> + * called when not already in an extended quiescent state.
> + */
> +static void rcu_dynticks_eqs_enter(void)
> +{
> +	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
> +
> +	/*
> +	 * CPUs seeing atomic_inc() must see prior RCU read-side critical
> +	 * sections, and we also must force ordering with the next idle
> +	 * sojourn.
> +	 */
> +	smp_mb__before_atomic(); /* See above. */
> +	atomic_inc(&rdtp->dynticks);
> +	smp_mb__after_atomic(); /* See above. */

How about replacing three lines above with:

	(void)atomic_inc_return(&rdtp->dynticks);

?

*_return should have the full barrier semantics.

> +	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
> +		     atomic_read(&rdtp->dynticks) & 0x1);
> +}
> +
> +/*
> + * Record exit from an extended quiescent state.  This is only to be
> + * called from an extended quiescent state.
> + */
> +static void rcu_dynticks_eqs_exit(void)
> +{
> +	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
> +
> +	/*
> +	 * CPUs seeing atomic_inc() must see prior idle sojourns,
> +	 * and we also must force ordering with the next RCU read-side
> +	 * critical section.
> +	 */
> +	smp_mb__before_atomic(); /* See above. */
> +	atomic_inc(&rdtp->dynticks);
> +	smp_mb__after_atomic(); /* See above. */

Ditto.

Regards,
Boqun

> +	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
> +		     !(atomic_read(&rdtp->dynticks) & 0x1));
> +}
> +
> +/*
> + * Reset the current CPU's ->dynticks counter to indicate that the
> + * newly onlined CPU is no longer in an extended quiescent state.
> + * This will either leave the counter unchanged, or increment it
> + * to the next non-quiescent value.
> + *
> + * The non-atomic test/increment sequence works because the upper bits
> + * of the ->dynticks counter are manipulated only by the corresponding CPU,
> + * or when the corresponding CPU is offline.
> + */
> +static void rcu_dynticks_eqs_online(void)
> +{
> +	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
> +
> +	if (atomic_read(&rdtp->dynticks) & 0x1)
> +		return;
> +	atomic_add(0x1, &rdtp->dynticks);
> +}
> +
> +/*
>   * Snapshot the ->dynticks counter with full ordering so as to allow
>   * stable comparison of this counter with past and future snapshots.
>   */
> @@ -690,7 +749,7 @@ static void rcu_eqs_enter_common(long long oldval, bool user)
>  {
>  	struct rcu_state *rsp;
>  	struct rcu_data *rdp;
> -	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
> +	struct rcu_dynticks __maybe_unused *rdtp = this_cpu_ptr(&rcu_dynticks);
>  
>  	trace_rcu_dyntick(TPS("Start"), oldval, rdtp->dynticks_nesting);
>  	if (IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
> @@ -709,12 +768,7 @@ static void rcu_eqs_enter_common(long long oldval, bool user)
>  		do_nocb_deferred_wakeup(rdp);
>  	}
>  	rcu_prepare_for_idle();
> -	/* CPUs seeing atomic_inc() must see prior RCU read-side crit sects */
> -	smp_mb__before_atomic();  /* See above. */
> -	atomic_inc(&rdtp->dynticks);
> -	smp_mb__after_atomic();  /* Force ordering with next sojourn. */
> -	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
> -		     atomic_read(&rdtp->dynticks) & 0x1);
> +	rcu_dynticks_eqs_enter();
>  	rcu_dynticks_task_enter();
>  
>  	/*
> @@ -843,15 +897,10 @@ void rcu_irq_exit_irqson(void)
>   */
>  static void rcu_eqs_exit_common(long long oldval, int user)
>  {
> -	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
> +	struct rcu_dynticks __maybe_unused *rdtp = this_cpu_ptr(&rcu_dynticks);
>  
>  	rcu_dynticks_task_exit();
> -	smp_mb__before_atomic();  /* Force ordering w/previous sojourn. */
> -	atomic_inc(&rdtp->dynticks);
> -	/* CPUs seeing atomic_inc() must see later RCU read-side crit sects */
> -	smp_mb__after_atomic();  /* See above. */
> -	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
> -		     !(atomic_read(&rdtp->dynticks) & 0x1));
> +	rcu_dynticks_eqs_exit();
>  	rcu_cleanup_after_idle();
>  	trace_rcu_dyntick(TPS("End"), oldval, rdtp->dynticks_nesting);
>  	if (IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
> @@ -998,11 +1047,7 @@ void rcu_nmi_enter(void)
>  	 * period (observation due to Andy Lutomirski).
>  	 */
>  	if (!(atomic_read(&rdtp->dynticks) & 0x1)) {
> -		smp_mb__before_atomic();  /* Force delay from prior write. */
> -		atomic_inc(&rdtp->dynticks);
> -		/* atomic_inc() before later RCU read-side crit sects */
> -		smp_mb__after_atomic();  /* See above. */
> -		WARN_ON_ONCE(!(atomic_read(&rdtp->dynticks) & 0x1));
> +		rcu_dynticks_eqs_exit();
>  		incby = 1;
>  	}
>  	rdtp->dynticks_nmi_nesting += incby;
> @@ -1040,11 +1085,7 @@ void rcu_nmi_exit(void)
>  
>  	/* This NMI interrupted an RCU-idle CPU, restore RCU-idleness. */
>  	rdtp->dynticks_nmi_nesting = 0;
> -	/* CPUs seeing atomic_inc() must see prior RCU read-side crit sects */
> -	smp_mb__before_atomic();  /* See above. */
> -	atomic_inc(&rdtp->dynticks);
> -	smp_mb__after_atomic();  /* Force delay to next write. */
> -	WARN_ON_ONCE(atomic_read(&rdtp->dynticks) & 0x1);
> +	rcu_dynticks_eqs_enter();
>  }
>  
>  /**
> @@ -3797,8 +3838,7 @@ rcu_init_percpu_data(int cpu, struct rcu_state *rsp)
>  		init_callback_list(rdp);  /* Re-enable callbacks on this CPU. */
>  	rdp->dynticks->dynticks_nesting = DYNTICK_TASK_EXIT_IDLE;
>  	rcu_sysidle_init_percpu_data(rdp->dynticks);
> -	atomic_set(&rdp->dynticks->dynticks,
> -		   (atomic_read(&rdp->dynticks->dynticks) & ~0x1) + 1);
> +	rcu_dynticks_eqs_online();
>  	raw_spin_unlock_rcu_node(rnp);		/* irqs remain disabled. */
>  
>  	/*
> -- 
> 2.5.2
> 

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 455 bytes --]

[PATCH tip/core/rcu 4/5] rcu: Abstract extended quiescent state determination

[Paul E. McKenney]

This commit is the fourth step towards full abstraction of all accesses
to the ->dynticks counter, implementing previously open-coded checks and
comparisons in new rcu_dynticks_in_eqs() and rcu_dynticks_in_eqs_since()
functions.  This abstraction will ease changes to the ->dynticks counter
operation.

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
 include/linux/rcutiny.h  |  6 ++++++
 kernel/rcu/tree.c        | 52 +++++++++++++++++++++++++++++++++++-------------
 kernel/rcu/tree.h        |  2 ++
 kernel/rcu/tree_exp.h    |  6 +++---
 kernel/rcu/tree_plugin.h |  2 +-
 kernel/rcu/tree_trace.c  |  2 +-
 6 files changed, 51 insertions(+), 19 deletions(-)

diff --git a/include/linux/rcutiny.h b/include/linux/rcutiny.h
index ac81e4063b40..4f9b2fa2173d 100644
--- a/include/linux/rcutiny.h
+++ b/include/linux/rcutiny.h
@@ -27,6 +27,12 @@
 
 #include <linux/cache.h>
 
+struct rcu_dynticks;
+static inline int rcu_dynticks_snap(struct rcu_dynticks *rdtp)
+{
+	return 0;
+}
+
 static inline unsigned long get_state_synchronize_rcu(void)
 {
 	return 0;
diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index 52e844902a43..c2b2f5b591b7 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -338,10 +338,22 @@ static void rcu_dynticks_eqs_online(void)
 }
 
 /*
+ * Is the current CPU in an extended quiescent state?
+ *
+ * No ordering, as we are sampling CPU-local information.
+ */
+bool rcu_dynticks_curr_cpu_in_eqs(void)
+{
+	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
+
+	return !(atomic_read(&rdtp->dynticks) & 0x1);
+}
+
+/*
  * Snapshot the ->dynticks counter with full ordering so as to allow
  * stable comparison of this counter with past and future snapshots.
  */
-static int rcu_dynticks_snap(struct rcu_dynticks *rdtp)
+int rcu_dynticks_snap(struct rcu_dynticks *rdtp)
 {
 	int snap = atomic_add_return(0, &rdtp->dynticks);
 
@@ -349,6 +361,25 @@ static int rcu_dynticks_snap(struct rcu_dynticks *rdtp)
 }
 
 /*
+ * Return true if the snapshot returned from rcu_dynticks_snap()
+ * indicates that RCU is in an extended quiescent state.
+ */
+static bool rcu_dynticks_in_eqs(int snap)
+{
+	return !(snap & 0x1);
+}
+
+/*
+ * Return true if the CPU corresponding to the specified rcu_dynticks
+ * structure has spent some time in an extended quiescent state since
+ * rcu_dynticks_snap() returned the specified snapshot.
+ */
+static bool rcu_dynticks_in_eqs_since(struct rcu_dynticks *rdtp, int snap)
+{
+	return snap != rcu_dynticks_snap(rdtp);
+}
+
+/*
  * Do a double-increment of the ->dynticks counter to emulate a
  * momentary idle-CPU quiescent state.
  */
@@ -1046,7 +1077,7 @@ void rcu_nmi_enter(void)
 	 * to be in the outermost NMI handler that interrupted an RCU-idle
 	 * period (observation due to Andy Lutomirski).
 	 */
-	if (!(atomic_read(&rdtp->dynticks) & 0x1)) {
+	if (rcu_dynticks_curr_cpu_in_eqs()) {
 		rcu_dynticks_eqs_exit();
 		incby = 1;
 	}
@@ -1072,7 +1103,7 @@ void rcu_nmi_exit(void)
 	 * to us!)
 	 */
 	WARN_ON_ONCE(rdtp->dynticks_nmi_nesting <= 0);
-	WARN_ON_ONCE(!(atomic_read(&rdtp->dynticks) & 0x1));
+	WARN_ON_ONCE(rcu_dynticks_curr_cpu_in_eqs());
 
 	/*
 	 * If the nesting level is not 1, the CPU wasn't RCU-idle, so
@@ -1098,9 +1129,7 @@ void rcu_nmi_exit(void)
  */
 bool notrace __rcu_is_watching(void)
 {
-	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
-
-	return atomic_read(&rdtp->dynticks) & 0x1;
+	return !rcu_dynticks_curr_cpu_in_eqs();
 }
 
 /**
@@ -1185,7 +1214,7 @@ static int dyntick_save_progress_counter(struct rcu_data *rdp,
 {
 	rdp->dynticks_snap = rcu_dynticks_snap(rdp->dynticks);
 	rcu_sysidle_check_cpu(rdp, isidle, maxj);
-	if ((rdp->dynticks_snap & 0x1) == 0) {
+	if (rcu_dynticks_in_eqs(rdp->dynticks_snap)) {
 		trace_rcu_fqs(rdp->rsp->name, rdp->gpnum, rdp->cpu, TPS("dti"));
 		if (ULONG_CMP_LT(READ_ONCE(rdp->gpnum) + ULONG_MAX / 4,
 				 rdp->mynode->gpnum))
@@ -1204,12 +1233,7 @@ static int dyntick_save_progress_counter(struct rcu_data *rdp,
 static int rcu_implicit_dynticks_qs(struct rcu_data *rdp,
 				    bool *isidle, unsigned long *maxj)
 {
-	unsigned int curr;
 	int *rcrmp;
-	unsigned int snap;
-
-	curr = (unsigned int)rcu_dynticks_snap(rdp->dynticks);
-	snap = (unsigned int)rdp->dynticks_snap;
 
 	/*
 	 * If the CPU passed through or entered a dynticks idle phase with
@@ -1219,7 +1243,7 @@ static int rcu_implicit_dynticks_qs(struct rcu_data *rdp,
 	 * read-side critical section that started before the beginning
 	 * of the current RCU grace period.
 	 */
-	if ((curr & 0x1) == 0 || UINT_CMP_GE(curr, snap + 2)) {
+	if (rcu_dynticks_in_eqs_since(rdp->dynticks, rdp->dynticks_snap)) {
 		trace_rcu_fqs(rdp->rsp->name, rdp->gpnum, rdp->cpu, TPS("dti"));
 		rdp->dynticks_fqs++;
 		return 1;
@@ -3808,7 +3832,7 @@ rcu_boot_init_percpu_data(int cpu, struct rcu_state *rsp)
 	rdp->grpmask = leaf_node_cpu_bit(rdp->mynode, cpu);
 	rdp->dynticks = &per_cpu(rcu_dynticks, cpu);
 	WARN_ON_ONCE(rdp->dynticks->dynticks_nesting != DYNTICK_TASK_EXIT_IDLE);
-	WARN_ON_ONCE(atomic_read(&rdp->dynticks->dynticks) != 1);
+	WARN_ON_ONCE(rcu_dynticks_in_eqs(rcu_dynticks_snap(rdp->dynticks)));
 	rdp->cpu = cpu;
 	rdp->rsp = rsp;
 	rcu_boot_init_nocb_percpu_data(rdp);
diff --git a/kernel/rcu/tree.h b/kernel/rcu/tree.h
index fe98dd24adf8..3b953dcf6afc 100644
--- a/kernel/rcu/tree.h
+++ b/kernel/rcu/tree.h
@@ -595,6 +595,8 @@ extern struct rcu_state rcu_bh_state;
 extern struct rcu_state rcu_preempt_state;
 #endif /* #ifdef CONFIG_PREEMPT_RCU */
 
+int rcu_dynticks_snap(struct rcu_dynticks *rdtp);
+
 #ifdef CONFIG_RCU_BOOST
 DECLARE_PER_CPU(unsigned int, rcu_cpu_kthread_status);
 DECLARE_PER_CPU(int, rcu_cpu_kthread_cpu);
diff --git a/kernel/rcu/tree_exp.h b/kernel/rcu/tree_exp.h
index 75f079ac79f9..a3a8756670d1 100644
--- a/kernel/rcu/tree_exp.h
+++ b/kernel/rcu/tree_exp.h
@@ -360,7 +360,7 @@ static void sync_rcu_exp_select_cpus(struct rcu_state *rsp,
 			rdp->exp_dynticks_snap =
 				rcu_dynticks_snap(rdp->dynticks);
 			if (raw_smp_processor_id() == cpu ||
-			    !(rdp->exp_dynticks_snap & 0x1) ||
+			    rcu_dynticks_in_eqs(rdp->exp_dynticks_snap) ||
 			    !(rnp->qsmaskinitnext & rdp->grpmask))
 				mask_ofl_test |= rdp->grpmask;
 		}
@@ -383,8 +383,8 @@ static void sync_rcu_exp_select_cpus(struct rcu_state *rsp,
 			if (!(mask_ofl_ipi & mask))
 				continue;
 retry_ipi:
-			if (rcu_dynticks_snap(rdp->dynticks) !=
-			    rdp->exp_dynticks_snap) {
+			if (rcu_dynticks_in_eqs_since(rdp->dynticks,
+						      rdp->exp_dynticks_snap)) {
 				mask_ofl_test |= mask;
 				continue;
 			}
diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h
index 85c5a883c6e3..98cd997b078e 100644
--- a/kernel/rcu/tree_plugin.h
+++ b/kernel/rcu/tree_plugin.h
@@ -1643,7 +1643,7 @@ static void print_cpu_stall_info(struct rcu_state *rsp, int cpu)
 	       "o."[!!(rdp->grpmask & rdp->mynode->qsmaskinit)],
 	       "N."[!!(rdp->grpmask & rdp->mynode->qsmaskinitnext)],
 	       ticks_value, ticks_title,
-	       atomic_read(&rdtp->dynticks) & 0xfff,
+	       rcu_dynticks_snap(rdtp) & 0xfff,
 	       rdtp->dynticks_nesting, rdtp->dynticks_nmi_nesting,
 	       rdp->softirq_snap, kstat_softirqs_cpu(RCU_SOFTIRQ, cpu),
 	       READ_ONCE(rsp->n_force_qs) - rsp->n_force_qs_gpstart,
diff --git a/kernel/rcu/tree_trace.c b/kernel/rcu/tree_trace.c
index b1f28972872c..b833cd0a29e8 100644
--- a/kernel/rcu/tree_trace.c
+++ b/kernel/rcu/tree_trace.c
@@ -124,7 +124,7 @@ static void print_one_rcu_data(struct seq_file *m, struct rcu_data *rdp)
 		   rdp->rcu_qs_ctr_snap == per_cpu(rcu_qs_ctr, rdp->cpu),
 		   rdp->core_needs_qs);
 	seq_printf(m, " dt=%d/%llx/%d df=%lu",
-		   atomic_read(&rdp->dynticks->dynticks),
+		   rcu_dynticks_snap(rdp->dynticks),
 		   rdp->dynticks->dynticks_nesting,
 		   rdp->dynticks->dynticks_nmi_nesting,
 		   rdp->dynticks_fqs);
-- 
2.5.2

[PATCH tip/core/rcu 5/5] rcu: Maintain special bits at bottom of ->dynticks counter

[Paul E. McKenney]

Currently, IPIs are used to force other CPUs to invalidate their TLBs
in response to a kernel virtual-memory mapping change.  This works, but
degrades both battery lifetime (for idle CPUs) and real-time response
(for nohz_full CPUs), and in addition results in unnecessary IPIs due to
the fact that CPUs executing in usermode are unaffected by stale kernel
mappings.  It would be better to cause a CPU executing in usermode to
wait until it is entering kernel mode to do the flush, first to avoid
interrupting usemode tasks and second to handle multiple flush requests
with a single flush in the case of a long-running user task.

This commit therefore reserves a bit at the bottom of the ->dynticks
counter, which is checked upon exit from extended quiescent states.
If it is set, it is cleared and then a new rcu_eqs_special_exit() macro is
invoked, which, if not supplied, is an empty single-pass do-while loop.
If this bottom bit is set on -entry- to an extended quiescent state,
then a WARN_ON_ONCE() triggers.

This bottom bit may be set using a new rcu_eqs_special_set() function,
which returns true if the bit was set, or false if the CPU turned
out to not be in an extended quiescent state.  Please note that this
function refuses to set the bit for a non-nohz_full CPU when that CPU
is executing in usermode because usermode execution is tracked by RCU
as a dyntick-idle extended quiescent state only for nohz_full CPUs.

Reported-by: Andy Lutomirski <luto@amacapital.net>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
 include/linux/rcutiny.h |  5 +++
 kernel/rcu/tree.c       | 81 +++++++++++++++++++++++++++++++++++++------------
 kernel/rcu/tree.h       |  1 +
 3 files changed, 67 insertions(+), 20 deletions(-)

diff --git a/include/linux/rcutiny.h b/include/linux/rcutiny.h
index 4f9b2fa2173d..7232d199a81c 100644
--- a/include/linux/rcutiny.h
+++ b/include/linux/rcutiny.h
@@ -33,6 +33,11 @@ static inline int rcu_dynticks_snap(struct rcu_dynticks *rdtp)
 	return 0;
 }
 
+static inline bool rcu_eqs_special_set(int cpu)
+{
+	return false;  /* Never flag non-existent other CPUs! */
+}
+
 static inline unsigned long get_state_synchronize_rcu(void)
 {
 	return 0;
diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index c2b2f5b591b7..2c399db6df6e 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -269,9 +269,19 @@ void rcu_bh_qs(void)
 
 static DEFINE_PER_CPU(int, rcu_sched_qs_mask);
 
+/*
+ * Steal a bit from the bottom of ->dynticks for idle entry/exit
+ * control.  Initially this is for TLB flushing.
+ */
+#define RCU_DYNTICK_CTRL_MASK 0x1
+#define RCU_DYNTICK_CTRL_CTR  (RCU_DYNTICK_CTRL_MASK + 1)
+#ifndef rcu_eqs_special_exit
+#define rcu_eqs_special_exit() do { } while (0)
+#endif
+
 static DEFINE_PER_CPU(struct rcu_dynticks, rcu_dynticks) = {
 	.dynticks_nesting = DYNTICK_TASK_EXIT_IDLE,
-	.dynticks = ATOMIC_INIT(1),
+	.dynticks = ATOMIC_INIT(RCU_DYNTICK_CTRL_CTR),
 #ifdef CONFIG_NO_HZ_FULL_SYSIDLE
 	.dynticks_idle_nesting = DYNTICK_TASK_NEST_VALUE,
 	.dynticks_idle = ATOMIC_INIT(1),
@@ -285,17 +295,20 @@ static DEFINE_PER_CPU(struct rcu_dynticks, rcu_dynticks) = {
 static void rcu_dynticks_eqs_enter(void)
 {
 	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
+	int seq;
 
 	/*
-	 * CPUs seeing atomic_inc() must see prior RCU read-side critical
-	 * sections, and we also must force ordering with the next idle
-	 * sojourn.
+	 * CPUs seeing atomic_inc_return() must see prior RCU read-side
+	 * critical sections, and we also must force ordering with the
+	 * next idle sojourn.
 	 */
-	smp_mb__before_atomic(); /* See above. */
-	atomic_inc(&rdtp->dynticks);
-	smp_mb__after_atomic(); /* See above. */
+	seq = atomic_add_return(RCU_DYNTICK_CTRL_CTR, &rdtp->dynticks);
+	/* Better be in an extended quiescent state! */
+	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
+		     (seq & RCU_DYNTICK_CTRL_CTR));
+	/* Better not have special action (TLB flush) pending! */
 	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
-		     atomic_read(&rdtp->dynticks) & 0x1);
+		     (seq & RCU_DYNTICK_CTRL_MASK));
 }
 
 /*
@@ -305,17 +318,22 @@ static void rcu_dynticks_eqs_enter(void)
 static void rcu_dynticks_eqs_exit(void)
 {
 	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
+	int seq;
 
 	/*
-	 * CPUs seeing atomic_inc() must see prior idle sojourns,
+	 * CPUs seeing atomic_inc_return() must see prior idle sojourns,
 	 * and we also must force ordering with the next RCU read-side
 	 * critical section.
 	 */
-	smp_mb__before_atomic(); /* See above. */
-	atomic_inc(&rdtp->dynticks);
-	smp_mb__after_atomic(); /* See above. */
+	seq = atomic_add_return(RCU_DYNTICK_CTRL_CTR, &rdtp->dynticks);
 	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
-		     !(atomic_read(&rdtp->dynticks) & 0x1));
+		     !(seq & RCU_DYNTICK_CTRL_CTR));
+	if (seq & RCU_DYNTICK_CTRL_MASK) {
+		rcu_eqs_special_exit();
+		/* Prefer duplicate flushes to losing a flush. */
+		smp_mb__before_atomic(); /* NMI safety. */
+		atomic_and(~RCU_DYNTICK_CTRL_MASK, &rdtp->dynticks);
+	}
 }
 
 /*
@@ -332,9 +350,9 @@ static void rcu_dynticks_eqs_online(void)
 {
 	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
 
-	if (atomic_read(&rdtp->dynticks) & 0x1)
+	if (atomic_read(&rdtp->dynticks) & RCU_DYNTICK_CTRL_CTR)
 		return;
-	atomic_add(0x1, &rdtp->dynticks);
+	atomic_add(RCU_DYNTICK_CTRL_CTR, &rdtp->dynticks);
 }
 
 /*
@@ -346,7 +364,7 @@ bool rcu_dynticks_curr_cpu_in_eqs(void)
 {
 	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
 
-	return !(atomic_read(&rdtp->dynticks) & 0x1);
+	return !(atomic_read(&rdtp->dynticks) & RCU_DYNTICK_CTRL_CTR);
 }
 
 /*
@@ -357,7 +375,7 @@ int rcu_dynticks_snap(struct rcu_dynticks *rdtp)
 {
 	int snap = atomic_add_return(0, &rdtp->dynticks);
 
-	return snap;
+	return snap & ~RCU_DYNTICK_CTRL_MASK;
 }
 
 /*
@@ -366,7 +384,7 @@ int rcu_dynticks_snap(struct rcu_dynticks *rdtp)
  */
 static bool rcu_dynticks_in_eqs(int snap)
 {
-	return !(snap & 0x1);
+	return !(snap & RCU_DYNTICK_CTRL_CTR);
 }
 
 /*
@@ -386,10 +404,33 @@ static bool rcu_dynticks_in_eqs_since(struct rcu_dynticks *rdtp, int snap)
 static void rcu_dynticks_momentary_idle(void)
 {
 	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
-	int special = atomic_add_return(2, &rdtp->dynticks);
+	int special = atomic_add_return(2 * RCU_DYNTICK_CTRL_CTR,
+					&rdtp->dynticks);
 
 	/* It is illegal to call this from idle state. */
-	WARN_ON_ONCE(!(special & 0x1));
+	WARN_ON_ONCE(!(special & RCU_DYNTICK_CTRL_CTR));
+}
+
+/*
+ * Set the special (bottom) bit of the specified CPU so that it
+ * will take special action (such as flushing its TLB) on the
+ * next exit from an extended quiescent state.  Returns true if
+ * the bit was successfully set, or false if the CPU was not in
+ * an extended quiescent state.
+ */
+bool rcu_eqs_special_set(int cpu)
+{
+	int old;
+	int new;
+	struct rcu_dynticks *rdtp = &per_cpu(rcu_dynticks, cpu);
+
+	do {
+		old = atomic_read(&rdtp->dynticks);
+		if (old & RCU_DYNTICK_CTRL_CTR)
+			return false;
+		new = old | RCU_DYNTICK_CTRL_MASK;
+	} while (atomic_cmpxchg(&rdtp->dynticks, old, new) != old);
+	return true;
 }
 
 DEFINE_PER_CPU_SHARED_ALIGNED(unsigned long, rcu_qs_ctr);
diff --git a/kernel/rcu/tree.h b/kernel/rcu/tree.h
index 3b953dcf6afc..7dcdd59d894c 100644
--- a/kernel/rcu/tree.h
+++ b/kernel/rcu/tree.h
@@ -596,6 +596,7 @@ extern struct rcu_state rcu_preempt_state;
 #endif /* #ifdef CONFIG_PREEMPT_RCU */
 
 int rcu_dynticks_snap(struct rcu_dynticks *rdtp);
+bool rcu_eqs_special_set(int cpu);
 
 #ifdef CONFIG_RCU_BOOST
 DECLARE_PER_CPU(unsigned int, rcu_cpu_kthread_status);
-- 
2.5.2

Re: [PATCH tip/core/rcu 5/5] rcu: Maintain special bits at bottom of ->dynticks counter

[Boqun Feng]

[-- Attachment #1: Type: text/plain, Size: 8465 bytes --]

On Mon, Nov 14, 2016 at 10:30:21AM -0800, Paul E. McKenney wrote:
> Currently, IPIs are used to force other CPUs to invalidate their TLBs
> in response to a kernel virtual-memory mapping change.  This works, but
> degrades both battery lifetime (for idle CPUs) and real-time response
> (for nohz_full CPUs), and in addition results in unnecessary IPIs due to
> the fact that CPUs executing in usermode are unaffected by stale kernel
> mappings.  It would be better to cause a CPU executing in usermode to
> wait until it is entering kernel mode to do the flush, first to avoid
> interrupting usemode tasks and second to handle multiple flush requests
> with a single flush in the case of a long-running user task.
> 
> This commit therefore reserves a bit at the bottom of the ->dynticks
> counter, which is checked upon exit from extended quiescent states.
> If it is set, it is cleared and then a new rcu_eqs_special_exit() macro is
> invoked, which, if not supplied, is an empty single-pass do-while loop.
> If this bottom bit is set on -entry- to an extended quiescent state,
> then a WARN_ON_ONCE() triggers.
> 
> This bottom bit may be set using a new rcu_eqs_special_set() function,
> which returns true if the bit was set, or false if the CPU turned
> out to not be in an extended quiescent state.  Please note that this
> function refuses to set the bit for a non-nohz_full CPU when that CPU
> is executing in usermode because usermode execution is tracked by RCU
> as a dyntick-idle extended quiescent state only for nohz_full CPUs.
> 
> Reported-by: Andy Lutomirski <luto@amacapital.net>
> Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
> ---
>  include/linux/rcutiny.h |  5 +++
>  kernel/rcu/tree.c       | 81 +++++++++++++++++++++++++++++++++++++------------
>  kernel/rcu/tree.h       |  1 +
>  3 files changed, 67 insertions(+), 20 deletions(-)
> 
> diff --git a/include/linux/rcutiny.h b/include/linux/rcutiny.h
> index 4f9b2fa2173d..7232d199a81c 100644
> --- a/include/linux/rcutiny.h
> +++ b/include/linux/rcutiny.h
> @@ -33,6 +33,11 @@ static inline int rcu_dynticks_snap(struct rcu_dynticks *rdtp)
>  	return 0;
>  }
>  
> +static inline bool rcu_eqs_special_set(int cpu)
> +{
> +	return false;  /* Never flag non-existent other CPUs! */
> +}
> +
>  static inline unsigned long get_state_synchronize_rcu(void)
>  {
>  	return 0;
> diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> index c2b2f5b591b7..2c399db6df6e 100644
> --- a/kernel/rcu/tree.c
> +++ b/kernel/rcu/tree.c
> @@ -269,9 +269,19 @@ void rcu_bh_qs(void)
>  
>  static DEFINE_PER_CPU(int, rcu_sched_qs_mask);
>  
> +/*
> + * Steal a bit from the bottom of ->dynticks for idle entry/exit
> + * control.  Initially this is for TLB flushing.
> + */
> +#define RCU_DYNTICK_CTRL_MASK 0x1
> +#define RCU_DYNTICK_CTRL_CTR  (RCU_DYNTICK_CTRL_MASK + 1)
> +#ifndef rcu_eqs_special_exit
> +#define rcu_eqs_special_exit() do { } while (0)
> +#endif
> +
>  static DEFINE_PER_CPU(struct rcu_dynticks, rcu_dynticks) = {
>  	.dynticks_nesting = DYNTICK_TASK_EXIT_IDLE,
> -	.dynticks = ATOMIC_INIT(1),
> +	.dynticks = ATOMIC_INIT(RCU_DYNTICK_CTRL_CTR),
>  #ifdef CONFIG_NO_HZ_FULL_SYSIDLE
>  	.dynticks_idle_nesting = DYNTICK_TASK_NEST_VALUE,
>  	.dynticks_idle = ATOMIC_INIT(1),
> @@ -285,17 +295,20 @@ static DEFINE_PER_CPU(struct rcu_dynticks, rcu_dynticks) = {
>  static void rcu_dynticks_eqs_enter(void)
>  {
>  	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
> +	int seq;
>  
>  	/*
> -	 * CPUs seeing atomic_inc() must see prior RCU read-side critical
> -	 * sections, and we also must force ordering with the next idle
> -	 * sojourn.
> +	 * CPUs seeing atomic_inc_return() must see prior RCU read-side
> +	 * critical sections, and we also must force ordering with the
> +	 * next idle sojourn.
>  	 */
> -	smp_mb__before_atomic(); /* See above. */
> -	atomic_inc(&rdtp->dynticks);
> -	smp_mb__after_atomic(); /* See above. */
> +	seq = atomic_add_return(RCU_DYNTICK_CTRL_CTR, &rdtp->dynticks);
> +	/* Better be in an extended quiescent state! */
> +	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
> +		     (seq & RCU_DYNTICK_CTRL_CTR));
> +	/* Better not have special action (TLB flush) pending! */

Ah.. you did the clean-up here ;-)

Never mind my previous comment on patch #3 ;-)

Regards,
Boqun

>  	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
> -		     atomic_read(&rdtp->dynticks) & 0x1);
> +		     (seq & RCU_DYNTICK_CTRL_MASK));
>  }
>  
>  /*
> @@ -305,17 +318,22 @@ static void rcu_dynticks_eqs_enter(void)
>  static void rcu_dynticks_eqs_exit(void)
>  {
>  	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
> +	int seq;
>  
>  	/*
> -	 * CPUs seeing atomic_inc() must see prior idle sojourns,
> +	 * CPUs seeing atomic_inc_return() must see prior idle sojourns,
>  	 * and we also must force ordering with the next RCU read-side
>  	 * critical section.
>  	 */
> -	smp_mb__before_atomic(); /* See above. */
> -	atomic_inc(&rdtp->dynticks);
> -	smp_mb__after_atomic(); /* See above. */
> +	seq = atomic_add_return(RCU_DYNTICK_CTRL_CTR, &rdtp->dynticks);
>  	WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
> -		     !(atomic_read(&rdtp->dynticks) & 0x1));
> +		     !(seq & RCU_DYNTICK_CTRL_CTR));
> +	if (seq & RCU_DYNTICK_CTRL_MASK) {
> +		rcu_eqs_special_exit();
> +		/* Prefer duplicate flushes to losing a flush. */
> +		smp_mb__before_atomic(); /* NMI safety. */
> +		atomic_and(~RCU_DYNTICK_CTRL_MASK, &rdtp->dynticks);
> +	}
>  }
>  
>  /*
> @@ -332,9 +350,9 @@ static void rcu_dynticks_eqs_online(void)
>  {
>  	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
>  
> -	if (atomic_read(&rdtp->dynticks) & 0x1)
> +	if (atomic_read(&rdtp->dynticks) & RCU_DYNTICK_CTRL_CTR)
>  		return;
> -	atomic_add(0x1, &rdtp->dynticks);
> +	atomic_add(RCU_DYNTICK_CTRL_CTR, &rdtp->dynticks);
>  }
>  
>  /*
> @@ -346,7 +364,7 @@ bool rcu_dynticks_curr_cpu_in_eqs(void)
>  {
>  	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
>  
> -	return !(atomic_read(&rdtp->dynticks) & 0x1);
> +	return !(atomic_read(&rdtp->dynticks) & RCU_DYNTICK_CTRL_CTR);
>  }
>  
>  /*
> @@ -357,7 +375,7 @@ int rcu_dynticks_snap(struct rcu_dynticks *rdtp)
>  {
>  	int snap = atomic_add_return(0, &rdtp->dynticks);
>  
> -	return snap;
> +	return snap & ~RCU_DYNTICK_CTRL_MASK;
>  }
>  
>  /*
> @@ -366,7 +384,7 @@ int rcu_dynticks_snap(struct rcu_dynticks *rdtp)
>   */
>  static bool rcu_dynticks_in_eqs(int snap)
>  {
> -	return !(snap & 0x1);
> +	return !(snap & RCU_DYNTICK_CTRL_CTR);
>  }
>  
>  /*
> @@ -386,10 +404,33 @@ static bool rcu_dynticks_in_eqs_since(struct rcu_dynticks *rdtp, int snap)
>  static void rcu_dynticks_momentary_idle(void)
>  {
>  	struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
> -	int special = atomic_add_return(2, &rdtp->dynticks);
> +	int special = atomic_add_return(2 * RCU_DYNTICK_CTRL_CTR,
> +					&rdtp->dynticks);
>  
>  	/* It is illegal to call this from idle state. */
> -	WARN_ON_ONCE(!(special & 0x1));
> +	WARN_ON_ONCE(!(special & RCU_DYNTICK_CTRL_CTR));
> +}
> +
> +/*
> + * Set the special (bottom) bit of the specified CPU so that it
> + * will take special action (such as flushing its TLB) on the
> + * next exit from an extended quiescent state.  Returns true if
> + * the bit was successfully set, or false if the CPU was not in
> + * an extended quiescent state.
> + */
> +bool rcu_eqs_special_set(int cpu)
> +{
> +	int old;
> +	int new;
> +	struct rcu_dynticks *rdtp = &per_cpu(rcu_dynticks, cpu);
> +
> +	do {
> +		old = atomic_read(&rdtp->dynticks);
> +		if (old & RCU_DYNTICK_CTRL_CTR)
> +			return false;
> +		new = old | RCU_DYNTICK_CTRL_MASK;
> +	} while (atomic_cmpxchg(&rdtp->dynticks, old, new) != old);
> +	return true;
>  }
>  
>  DEFINE_PER_CPU_SHARED_ALIGNED(unsigned long, rcu_qs_ctr);
> diff --git a/kernel/rcu/tree.h b/kernel/rcu/tree.h
> index 3b953dcf6afc..7dcdd59d894c 100644
> --- a/kernel/rcu/tree.h
> +++ b/kernel/rcu/tree.h
> @@ -596,6 +596,7 @@ extern struct rcu_state rcu_preempt_state;
>  #endif /* #ifdef CONFIG_PREEMPT_RCU */
>  
>  int rcu_dynticks_snap(struct rcu_dynticks *rdtp);
> +bool rcu_eqs_special_set(int cpu);
>  
>  #ifdef CONFIG_RCU_BOOST
>  DECLARE_PER_CPU(unsigned int, rcu_cpu_kthread_status);
> -- 
> 2.5.2
> 

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 455 bytes --]