[PATCH tip/core/rcu 0/15] Miscellaneous fixes for 3.18

[PATCH tip/core/rcu 0/15] Miscellaneous fixes for 3.18

[Paul E. McKenney]

Hello!

This series contains miscellaneous fixes for RCU:

1.	Remove more read-modify-write ACCESS_ONCE() calls, courtesy of
	Pranith Kumar.

2.	Fix sparse warning for rcu_batches_completed_preempt(), courtesy
	of Pranith Kumar.

3.	Use bool type for variable holding rcu_is_watching() return
	value, courtesy of Pranith Kumar.

4.	Use bool constant for bool return in rcu_try_advance_all_cbs(),
	courtesy of Pranith Kumar.

5.	Use bool constant for bool return in rcu_lockdep_current_cpu_online(),
	courtesy of Pranith Kumar.

6.	Use bool constant for assignments to bool variable isidle,
	courtesy of Pranith Kumar.

7.	Shrink the kernel a bit by uninlining rcu_read_lock_held(),
	courtesy of Oleg Nesterov.

8.	Shrink the kernel a bit more by defining tracepoint strings only
	if CONFIG_TRACING is set, courtesy of Ard Biesheuvel.

9.	Fix outdated pathnames in comments, courtesy of Pranith Kumar.

10.	Remove stale comment, courtesy of Pranith Kumar.

11.	Break additional call_rcu() deadlock involving scheduler and perf.

12.	Shrink the TINY_RCU side of the kernel by putting error checks
	under #ifdef.

13.	Eliminate unnecessary self-wakeups, courtesy of Pranith Kumar.

14.	Replace flush_signals() with WARN_ON(signal_pending()).

15.	Add ACCESS_ONCE() for RCU_INIT_POINTER().

							Thanx, Paul

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

 b/include/linux/rcupdate.h   |   40 ++-------------------------------------
 b/include/trace/events/rcu.h |    3 ++
 b/kernel/rcu/tiny.c          |    8 +++----
 b/kernel/rcu/tree.c          |   39 +++++++++++++++++++++++---------------
 b/kernel/rcu/tree.h          |    9 ++++++--
 b/kernel/rcu/tree_plugin.h   |   44 +++++++++++++++++++++++++++----------------
 b/kernel/rcu/update.c        |   32 +++++++++++++++++++++++++++++++
 7 files changed, 101 insertions(+), 74 deletions(-)

[PATCH tip/core/rcu 01/15] rcu: Remove remaining read-modify-write ACCESS_ONCE() calls

[Paul E. McKenney]

From: Pranith Kumar <bobby.prani@gmail.com>

Change the remaining uses of ACCESS_ONCE() so that each ACCESS_ONCE() either does a load or a store, but not both.

Signed-off-by: Pranith Kumar <bobby.prani@gmail.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
 kernel/rcu/tree.c        | 6 ++++--
 kernel/rcu/tree_plugin.h | 8 +++++---
 2 files changed, 9 insertions(+), 5 deletions(-)

diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index 1b70cb6fbe3c..4b526ca46801 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -1684,7 +1684,8 @@ static int rcu_gp_fqs(struct rcu_state *rsp, int fqs_state_in)
 	if (ACCESS_ONCE(rsp->gp_flags) & RCU_GP_FLAG_FQS) {
 		raw_spin_lock_irq(&rnp->lock);
 		smp_mb__after_unlock_lock();
-		ACCESS_ONCE(rsp->gp_flags) &= ~RCU_GP_FLAG_FQS;
+		ACCESS_ONCE(rsp->gp_flags) =
+			ACCESS_ONCE(rsp->gp_flags) & ~RCU_GP_FLAG_FQS;
 		raw_spin_unlock_irq(&rnp->lock);
 	}
 	return fqs_state;
@@ -2505,7 +2506,8 @@ static void force_quiescent_state(struct rcu_state *rsp)
 		raw_spin_unlock_irqrestore(&rnp_old->lock, flags);
 		return;  /* Someone beat us to it. */
 	}
-	ACCESS_ONCE(rsp->gp_flags) |= RCU_GP_FLAG_FQS;
+	ACCESS_ONCE(rsp->gp_flags) =
+		ACCESS_ONCE(rsp->gp_flags) | RCU_GP_FLAG_FQS;
 	raw_spin_unlock_irqrestore(&rnp_old->lock, flags);
 	wake_up(&rsp->gp_wq);  /* Memory barrier implied by wake_up() path. */
 }
diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h
index 00dc411e9676..2626a07ae8f2 100644
--- a/kernel/rcu/tree_plugin.h
+++ b/kernel/rcu/tree_plugin.h
@@ -897,7 +897,8 @@ void synchronize_rcu_expedited(void)
 
 	/* Clean up and exit. */
 	smp_mb(); /* ensure expedited GP seen before counter increment. */
-	ACCESS_ONCE(sync_rcu_preempt_exp_count)++;
+	ACCESS_ONCE(sync_rcu_preempt_exp_count) =
+					sync_rcu_preempt_exp_count + 1;
 unlock_mb_ret:
 	mutex_unlock(&sync_rcu_preempt_exp_mutex);
 mb_ret:
@@ -2428,8 +2429,9 @@ static int rcu_nocb_kthread(void *arg)
 			list = next;
 		}
 		trace_rcu_batch_end(rdp->rsp->name, c, !!list, 0, 0, 1);
-		ACCESS_ONCE(rdp->nocb_p_count) -= c;
-		ACCESS_ONCE(rdp->nocb_p_count_lazy) -= cl;
+		ACCESS_ONCE(rdp->nocb_p_count) = rdp->nocb_p_count - c;
+		ACCESS_ONCE(rdp->nocb_p_count_lazy) =
+						rdp->nocb_p_count_lazy - cl;
 		rdp->n_nocbs_invoked += c;
 	}
 	return 0;
-- 
1.8.1.5

[PATCH tip/core/rcu 02/15] rcu: Fix sparse warning about rcu_batches_completed_preempt() being non-static

[Paul E. McKenney]

From: Pranith Kumar <bobby.prani@gmail.com>

fix sparse warning about rcu_batches_completed_preempt() being non-static by
marking it as static

Signed-off-by: Pranith Kumar <bobby.prani@gmail.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
 kernel/rcu/tree_plugin.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h
index 2626a07ae8f2..8b7bded214b7 100644
--- a/kernel/rcu/tree_plugin.h
+++ b/kernel/rcu/tree_plugin.h
@@ -134,7 +134,7 @@ static void __init rcu_bootup_announce(void)
  * Return the number of RCU-preempt batches processed thus far
  * for debug and statistics.
  */
-long rcu_batches_completed_preempt(void)
+static long rcu_batches_completed_preempt(void)
 {
 	return rcu_preempt_state.completed;
 }
-- 
1.8.1.5

[PATCH tip/core/rcu 03/15] rcu: Use bool type for return value in rcu_is_watching()

[Paul E. McKenney]

From: Pranith Kumar <bobby.prani@gmail.com>

Use a bool type for return in rcu_is_watching().

Signed-off-by: Pranith Kumar <bobby.prani@gmail.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
 kernel/rcu/tree.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index 4b526ca46801..253ea55dc508 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -819,7 +819,7 @@ bool notrace __rcu_is_watching(void)
  */
 bool notrace rcu_is_watching(void)
 {
-	int ret;
+	bool ret;
 
 	preempt_disable();
 	ret = __rcu_is_watching();
-- 
1.8.1.5

[PATCH tip/core/rcu 04/15] rcu: Return bool type for rcu_try_advance_all_cbs()

[Paul E. McKenney]

From: Pranith Kumar <bobby.prani@gmail.com>

Return a bool type instead of 0 in rcu_try_advance_all_cbs().

Signed-off-by: Pranith Kumar <bobby.prani@gmail.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
 kernel/rcu/tree_plugin.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h
index 8b7bded214b7..b6acb9340192 100644
--- a/kernel/rcu/tree_plugin.h
+++ b/kernel/rcu/tree_plugin.h
@@ -1626,7 +1626,7 @@ static bool __maybe_unused rcu_try_advance_all_cbs(void)
 
 	/* Exit early if we advanced recently. */
 	if (jiffies == rdtp->last_advance_all)
-		return 0;
+		return false;
 	rdtp->last_advance_all = jiffies;
 
 	for_each_rcu_flavor(rsp) {
-- 
1.8.1.5

[PATCH tip/core/rcu 05/15] rcu: Return bool type in rcu_lockdep_current_cpu_online()

[Paul E. McKenney]

From: Pranith Kumar <bobby.prani@gmail.com>

Return true instead of 1 in rcu_lockdep_current_cpu_online() as this
has bool as return type.

Signed-off-by: Pranith Kumar <bobby.prani@gmail.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
 include/linux/rcupdate.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h
index d231aa17b1d7..7e47e44bce03 100644
--- a/include/linux/rcupdate.h
+++ b/include/linux/rcupdate.h
@@ -349,7 +349,7 @@ bool rcu_lockdep_current_cpu_online(void);
 #else /* #if defined(CONFIG_HOTPLUG_CPU) && defined(CONFIG_PROVE_RCU) */
 static inline bool rcu_lockdep_current_cpu_online(void)
 {
-	return 1;
+	return true;
 }
 #endif /* #else #if defined(CONFIG_HOTPLUG_CPU) && defined(CONFIG_PROVE_RCU) */
 
-- 
1.8.1.5

[PATCH tip/core/rcu 06/15] rcu: Use true/false instead of 1/0 for a bool type

[Paul E. McKenney]

From: Pranith Kumar <bobby.prani@gmail.com>

This commit uses true/false instead of 1/0 for bool types in rcu_gp_fqs()
and force_qs_rnp().

Signed-off-by: Pranith Kumar <bobby.prani@gmail.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
 kernel/rcu/tree.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index 253ea55dc508..2719978ea018 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -1668,7 +1668,7 @@ static int rcu_gp_fqs(struct rcu_state *rsp, int fqs_state_in)
 	if (fqs_state == RCU_SAVE_DYNTICK) {
 		/* Collect dyntick-idle snapshots. */
 		if (is_sysidle_rcu_state(rsp)) {
-			isidle = 1;
+			isidle = true;
 			maxj = jiffies - ULONG_MAX / 4;
 		}
 		force_qs_rnp(rsp, dyntick_save_progress_counter,
@@ -1677,7 +1677,7 @@ static int rcu_gp_fqs(struct rcu_state *rsp, int fqs_state_in)
 		fqs_state = RCU_FORCE_QS;
 	} else {
 		/* Handle dyntick-idle and offline CPUs. */
-		isidle = 0;
+		isidle = false;
 		force_qs_rnp(rsp, rcu_implicit_dynticks_qs, &isidle, &maxj);
 	}
 	/* Clear flag to prevent immediate re-entry. */
@@ -2450,7 +2450,7 @@ static void force_qs_rnp(struct rcu_state *rsp,
 		for (; cpu <= rnp->grphi; cpu++, bit <<= 1) {
 			if ((rnp->qsmask & bit) != 0) {
 				if ((rnp->qsmaskinit & bit) != 0)
-					*isidle = 0;
+					*isidle = false;
 				if (f(per_cpu_ptr(rsp->rda, cpu), isidle, maxj))
 					mask |= bit;
 			}
-- 
1.8.1.5

[PATCH tip/core/rcu 07/15] rcu: Uninline rcu_read_lock_held()

[Paul E. McKenney]

From: Oleg Nesterov <oleg@redhat.com>

This commit uninlines rcu_read_lock_held(). According to "size vmlinux"
this saves 28549 in .text:

	- 5541731 3014560 14757888 23314179
	+ 5513182 3026848 14757888 23297918

Note: it looks as if the data grows by 12288 bytes but this is not true,
it does not actually grow. But .data starts with ALIGN(THREAD_SIZE) and
since .text shrinks the padding grows, and thus .data grows too as it
seen by /bin/size. diff System.map:

	- ffffffff81510000 D _sdata
	- ffffffff81510000 D init_thread_union
	+ ffffffff81509000 D _sdata
	+ ffffffff8150c000 D init_thread_union

Perhaps we can change vmlinux.lds.S to .data itself, so that /bin/size
can't "wrongly" report that .data grows if .text shinks.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
 include/linux/rcupdate.h | 36 +-----------------------------------
 kernel/rcu/update.c      | 32 ++++++++++++++++++++++++++++++++
 2 files changed, 33 insertions(+), 35 deletions(-)

diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h
index 7e47e44bce03..321ed0d4e675 100644
--- a/include/linux/rcupdate.h
+++ b/include/linux/rcupdate.h
@@ -371,41 +371,7 @@ extern struct lockdep_map rcu_sched_lock_map;
 extern struct lockdep_map rcu_callback_map;
 int debug_lockdep_rcu_enabled(void);
 
-/**
- * rcu_read_lock_held() - might we be in RCU read-side critical section?
- *
- * If CONFIG_DEBUG_LOCK_ALLOC is selected, returns nonzero iff in an RCU
- * read-side critical section.  In absence of CONFIG_DEBUG_LOCK_ALLOC,
- * this assumes we are in an RCU read-side critical section unless it can
- * prove otherwise.  This is useful for debug checks in functions that
- * require that they be called within an RCU read-side critical section.
- *
- * Checks debug_lockdep_rcu_enabled() to prevent false positives during boot
- * and while lockdep is disabled.
- *
- * Note that rcu_read_lock() and the matching rcu_read_unlock() must
- * occur in the same context, for example, it is illegal to invoke
- * rcu_read_unlock() in process context if the matching rcu_read_lock()
- * was invoked from within an irq handler.
- *
- * Note that rcu_read_lock() is disallowed if the CPU is either idle or
- * offline from an RCU perspective, so check for those as well.
- */
-static inline int rcu_read_lock_held(void)
-{
-	if (!debug_lockdep_rcu_enabled())
-		return 1;
-	if (!rcu_is_watching())
-		return 0;
-	if (!rcu_lockdep_current_cpu_online())
-		return 0;
-	return lock_is_held(&rcu_lock_map);
-}
-
-/*
- * rcu_read_lock_bh_held() is defined out of line to avoid #include-file
- * hell.
- */
+int rcu_read_lock_held(void);
 int rcu_read_lock_bh_held(void);
 
 /**
diff --git a/kernel/rcu/update.c b/kernel/rcu/update.c
index 4056d7992a6c..ea8ea7b16e11 100644
--- a/kernel/rcu/update.c
+++ b/kernel/rcu/update.c
@@ -137,6 +137,38 @@ int notrace debug_lockdep_rcu_enabled(void)
 EXPORT_SYMBOL_GPL(debug_lockdep_rcu_enabled);
 
 /**
+ * rcu_read_lock_held() - might we be in RCU read-side critical section?
+ *
+ * If CONFIG_DEBUG_LOCK_ALLOC is selected, returns nonzero iff in an RCU
+ * read-side critical section.  In absence of CONFIG_DEBUG_LOCK_ALLOC,
+ * this assumes we are in an RCU read-side critical section unless it can
+ * prove otherwise.  This is useful for debug checks in functions that
+ * require that they be called within an RCU read-side critical section.
+ *
+ * Checks debug_lockdep_rcu_enabled() to prevent false positives during boot
+ * and while lockdep is disabled.
+ *
+ * Note that rcu_read_lock() and the matching rcu_read_unlock() must
+ * occur in the same context, for example, it is illegal to invoke
+ * rcu_read_unlock() in process context if the matching rcu_read_lock()
+ * was invoked from within an irq handler.
+ *
+ * Note that rcu_read_lock() is disallowed if the CPU is either idle or
+ * offline from an RCU perspective, so check for those as well.
+ */
+int rcu_read_lock_held(void)
+{
+	if (!debug_lockdep_rcu_enabled())
+		return 1;
+	if (!rcu_is_watching())
+		return 0;
+	if (!rcu_lockdep_current_cpu_online())
+		return 0;
+	return lock_is_held(&rcu_lock_map);
+}
+EXPORT_SYMBOL_GPL(rcu_read_lock_held);
+
+/**
  * rcu_read_lock_bh_held() - might we be in RCU-bh read-side critical section?
  *
  * Check for bottom half being disabled, which covers both the
-- 
1.8.1.5

[PATCH tip/core/rcu 08/15] rcu: Define tracepoint strings only if CONFIG_TRACING is set

[Paul E. McKenney]

From: Ard Biesheuvel <ard.biesheuvel@linaro.org>

Commit f7f7bac9cb1c ("rcu: Have the RCU tracepoints use the tracepoint_string
infrastructure") unconditionally populates the __tracepoint_str input section,
but this section is not assigned an output section if CONFIG_TRACING is not set.
This results in the __tracepoint_str turning up in unexpected places, i.e.,
after _edata.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
 kernel/rcu/tree.c | 15 ++++++++++++---
 1 file changed, 12 insertions(+), 3 deletions(-)

diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index 2719978ea018..dc52dc3b8c3e 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -79,9 +79,18 @@ static struct lock_class_key rcu_fqs_class[RCU_NUM_LVLS];
  * the tracing userspace tools to be able to decipher the string
  * address to the matching string.
  */
-#define RCU_STATE_INITIALIZER(sname, sabbr, cr) \
+#ifdef CONFIG_TRACING
+# define DEFINE_RCU_TPS(sname) \
 static char sname##_varname[] = #sname; \
-static const char *tp_##sname##_varname __used __tracepoint_string = sname##_varname; \
+static const char *tp_##sname##_varname __used __tracepoint_string = sname##_varname;
+# define RCU_STATE_NAME(sname) sname##_varname
+#else
+# define DEFINE_RCU_TPS(sname)
+# define RCU_STATE_NAME(sname) __stringify(sname)
+#endif
+
+#define RCU_STATE_INITIALIZER(sname, sabbr, cr) \
+DEFINE_RCU_TPS(sname) \
 struct rcu_state sname##_state = { \
 	.level = { &sname##_state.node[0] }, \
 	.call = cr, \
@@ -93,7 +102,7 @@ struct rcu_state sname##_state = { \
 	.orphan_donetail = &sname##_state.orphan_donelist, \
 	.barrier_mutex = __MUTEX_INITIALIZER(sname##_state.barrier_mutex), \
 	.onoff_mutex = __MUTEX_INITIALIZER(sname##_state.onoff_mutex), \
-	.name = sname##_varname, \
+	.name = RCU_STATE_NAME(sname), \
 	.abbr = sabbr, \
 }; \
 DEFINE_PER_CPU(struct rcu_data, sname##_data)
-- 
1.8.1.5

[PATCH tip/core/rcu 09/15] rcu: Update tiny.c references to tree.c

[Paul E. McKenney]

From: Pranith Kumar <bobby.prani@gmail.com>

This commit updates the references to rcutree.c which is now rcu/tree.c

Signed-off-by: Pranith Kumar <bobby.prani@gmail.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
 kernel/rcu/tiny.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/rcu/tiny.c b/kernel/rcu/tiny.c
index d9efcc13008c..6bd785c34add 100644
--- a/kernel/rcu/tiny.c
+++ b/kernel/rcu/tiny.c
@@ -51,7 +51,7 @@ static long long rcu_dynticks_nesting = DYNTICK_TASK_EXIT_IDLE;
 
 #include "tiny_plugin.h"
 
-/* Common code for rcu_idle_enter() and rcu_irq_exit(), see kernel/rcutree.c. */
+/* Common code for rcu_idle_enter() and rcu_irq_exit(), see kernel/rcu/tree.c. */
 static void rcu_idle_enter_common(long long newval)
 {
 	if (newval) {
@@ -114,7 +114,7 @@ void rcu_irq_exit(void)
 }
 EXPORT_SYMBOL_GPL(rcu_irq_exit);
 
-/* Common code for rcu_idle_exit() and rcu_irq_enter(), see kernel/rcutree.c. */
+/* Common code for rcu_idle_exit() and rcu_irq_enter(), see kernel/rcu/tree.c. */
 static void rcu_idle_exit_common(long long oldval)
 {
 	if (oldval) {
-- 
1.8.1.5

[PATCH tip/core/rcu 10/15] rcu: Remove stale comment in tree.c

[Paul E. McKenney]

From: Pranith Kumar <bobby.prani@gmail.com>

This commit removes a stale comment in rcu/tree.c which was left
out when some code was moved around previously in commit 2036d94a7b61
("rcu:  Rework detection of use of RCU by offline CPUs") For reference,
the following updated comment exists a few lines below this which means
the same:

/* Remove the outgoing CPU from the masks in the rcu_node hierarchy. */

Signed-off-by: Pranith Kumar <bobby.prani@gmail.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Reviewed-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
 kernel/rcu/tree.c | 2 --
 1 file changed, 2 deletions(-)

diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index dc52dc3b8c3e..dd6c8b519691 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -2220,8 +2220,6 @@ static void rcu_cleanup_dead_cpu(int cpu, struct rcu_state *rsp)
 	/* Adjust any no-longer-needed kthreads. */
 	rcu_boost_kthread_setaffinity(rnp, -1);
 
-	/* Remove the dead CPU from the bitmasks in the rcu_node hierarchy. */
-
 	/* Exclude any attempts to start a new grace period. */
 	mutex_lock(&rsp->onoff_mutex);
 	raw_spin_lock_irqsave(&rsp->orphan_lock, flags);
-- 
1.8.1.5

[PATCH tip/core/rcu 11/15] rcu: Break more call_rcu() deadlock involving scheduler and perf

[Paul E. McKenney]

From: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>

Commit 96d3fd0d315a9 (rcu: Break call_rcu() deadlock involving scheduler
and perf) covered the case where __call_rcu_nocb_enqueue() needs to wake
the rcuo kthread due to the queue being initially empty, but did not
do anything for the case where the queue was overflowing.  This commit
therefore also defers wakeup for the overflow case.

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
 include/trace/events/rcu.h |  3 +++
 kernel/rcu/tree.h          |  9 +++++++--
 kernel/rcu/tree_plugin.h   | 26 ++++++++++++++++++--------
 3 files changed, 28 insertions(+), 10 deletions(-)

diff --git a/include/trace/events/rcu.h b/include/trace/events/rcu.h
index aca382266411..9b56f37148cf 100644
--- a/include/trace/events/rcu.h
+++ b/include/trace/events/rcu.h
@@ -180,9 +180,12 @@ TRACE_EVENT(rcu_grace_period_init,
  * argument is a string as follows:
  *
  *	"WakeEmpty": Wake rcuo kthread, first CB to empty list.
+ *	"WakeEmptyIsDeferred": Wake rcuo kthread later, first CB to empty list.
  *	"WakeOvf": Wake rcuo kthread, CB list is huge.
+ *	"WakeOvfIsDeferred": Wake rcuo kthread later, CB list is huge.
  *	"WakeNot": Don't wake rcuo kthread.
  *	"WakeNotPoll": Don't wake rcuo kthread because it is polling.
+ *	"DeferredWake": Carried out the "IsDeferred" wakeup.
  *	"Poll": Start of new polling cycle for rcu_nocb_poll.
  *	"Sleep": Sleep waiting for CBs for !rcu_nocb_poll.
  *	"WokeEmpty": rcuo kthread woke to find empty list.
diff --git a/kernel/rcu/tree.h b/kernel/rcu/tree.h
index 71e64c718f75..0abc1a5c63f9 100644
--- a/kernel/rcu/tree.h
+++ b/kernel/rcu/tree.h
@@ -350,7 +350,7 @@ struct rcu_data {
 	int nocb_p_count_lazy;		/*  (approximate). */
 	wait_queue_head_t nocb_wq;	/* For nocb kthreads to sleep on. */
 	struct task_struct *nocb_kthread;
-	bool nocb_defer_wakeup;		/* Defer wakeup of nocb_kthread. */
+	int nocb_defer_wakeup;		/* Defer wakeup of nocb_kthread. */
 
 	/* The following fields are used by the leader, hence own cacheline. */
 	struct rcu_head *nocb_gp_head ____cacheline_internodealigned_in_smp;
@@ -383,6 +383,11 @@ struct rcu_data {
 #define RCU_FORCE_QS		3	/* Need to force quiescent state. */
 #define RCU_SIGNAL_INIT		RCU_SAVE_DYNTICK
 
+/* Values for nocb_defer_wakeup field in struct rcu_data. */
+#define RCU_NOGP_WAKE_NOT	0
+#define RCU_NOGP_WAKE		1
+#define RCU_NOGP_WAKE_FORCE	2
+
 #define RCU_JIFFIES_TILL_FORCE_QS (1 + (HZ > 250) + (HZ > 500))
 					/* For jiffies_till_first_fqs and */
 					/*  and jiffies_till_next_fqs. */
@@ -589,7 +594,7 @@ static bool __call_rcu_nocb(struct rcu_data *rdp, struct rcu_head *rhp,
 static bool rcu_nocb_adopt_orphan_cbs(struct rcu_state *rsp,
 				      struct rcu_data *rdp,
 				      unsigned long flags);
-static bool rcu_nocb_need_deferred_wakeup(struct rcu_data *rdp);
+static int rcu_nocb_need_deferred_wakeup(struct rcu_data *rdp);
 static void do_nocb_deferred_wakeup(struct rcu_data *rdp);
 static void rcu_boot_init_nocb_percpu_data(struct rcu_data *rdp);
 static void rcu_spawn_nocb_kthreads(struct rcu_state *rsp);
diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h
index b6acb9340192..a1af86099a67 100644
--- a/kernel/rcu/tree_plugin.h
+++ b/kernel/rcu/tree_plugin.h
@@ -2121,16 +2121,23 @@ static void __call_rcu_nocb_enqueue(struct rcu_data *rdp,
 			trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
 					    TPS("WakeEmpty"));
 		} else {
-			rdp->nocb_defer_wakeup = true;
+			rdp->nocb_defer_wakeup = RCU_NOGP_WAKE;
 			trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
 					    TPS("WakeEmptyIsDeferred"));
 		}
 		rdp->qlen_last_fqs_check = 0;
 	} else if (len > rdp->qlen_last_fqs_check + qhimark) {
 		/* ... or if many callbacks queued. */
-		wake_nocb_leader(rdp, true);
+		if (!irqs_disabled_flags(flags)) {
+			wake_nocb_leader(rdp, true);
+			trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
+					    TPS("WakeOvf"));
+		} else {
+			rdp->nocb_defer_wakeup = RCU_NOGP_WAKE_FORCE;
+			trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
+					    TPS("WakeOvfIsDeferred"));
+		}
 		rdp->qlen_last_fqs_check = LONG_MAX / 2;
-		trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu, TPS("WakeOvf"));
 	} else {
 		trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu, TPS("WakeNot"));
 	}
@@ -2438,7 +2445,7 @@ static int rcu_nocb_kthread(void *arg)
 }
 
 /* Is a deferred wakeup of rcu_nocb_kthread() required? */
-static bool rcu_nocb_need_deferred_wakeup(struct rcu_data *rdp)
+static int rcu_nocb_need_deferred_wakeup(struct rcu_data *rdp)
 {
 	return ACCESS_ONCE(rdp->nocb_defer_wakeup);
 }
@@ -2446,11 +2453,14 @@ static bool rcu_nocb_need_deferred_wakeup(struct rcu_data *rdp)
 /* Do a deferred wakeup of rcu_nocb_kthread(). */
 static void do_nocb_deferred_wakeup(struct rcu_data *rdp)
 {
+	int ndw;
+
 	if (!rcu_nocb_need_deferred_wakeup(rdp))
 		return;
-	ACCESS_ONCE(rdp->nocb_defer_wakeup) = false;
-	wake_nocb_leader(rdp, false);
-	trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu, TPS("DeferredWakeEmpty"));
+	ndw = ACCESS_ONCE(rdp->nocb_defer_wakeup);
+	ACCESS_ONCE(rdp->nocb_defer_wakeup) = RCU_NOGP_WAKE_NOT;
+	wake_nocb_leader(rdp, ndw == RCU_NOGP_WAKE_FORCE);
+	trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu, TPS("DeferredWake"));
 }
 
 /* Initialize per-rcu_data variables for no-CBs CPUs. */
@@ -2557,7 +2567,7 @@ static void __init rcu_boot_init_nocb_percpu_data(struct rcu_data *rdp)
 {
 }
 
-static bool rcu_nocb_need_deferred_wakeup(struct rcu_data *rdp)
+static int rcu_nocb_need_deferred_wakeup(struct rcu_data *rdp)
 {
 	return false;
 }
-- 
1.8.1.5

Re: [PATCH tip/core/rcu 11/15] rcu: Break more call_rcu() deadlock involving scheduler and perf

[Peter Zijlstra]

On Thu, Aug 28, 2014 at 11:26:30AM -0700, Paul E. McKenney wrote:

> diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h
> index b6acb9340192..a1af86099a67 100644
> --- a/kernel/rcu/tree_plugin.h
> +++ b/kernel/rcu/tree_plugin.h
> @@ -2121,16 +2121,23 @@ static void __call_rcu_nocb_enqueue(struct rcu_data *rdp,
>  			trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
>  					    TPS("WakeEmpty"));
>  		} else {
> -			rdp->nocb_defer_wakeup = true;
> +			rdp->nocb_defer_wakeup = RCU_NOGP_WAKE;
>  			trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
>  					    TPS("WakeEmptyIsDeferred"));
>  		}
>  		rdp->qlen_last_fqs_check = 0;
>  	} else if (len > rdp->qlen_last_fqs_check + qhimark) {
>  		/* ... or if many callbacks queued. */
> -		wake_nocb_leader(rdp, true);
> +		if (!irqs_disabled_flags(flags)) {
> +			wake_nocb_leader(rdp, true);
> +			trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
> +					    TPS("WakeOvf"));
> +		} else {
> +			rdp->nocb_defer_wakeup = RCU_NOGP_WAKE_FORCE;
> +			trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
> +					    TPS("WakeOvfIsDeferred"));
> +		}
>  		rdp->qlen_last_fqs_check = LONG_MAX / 2;
> -		trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu, TPS("WakeOvf"));
>  	} else {
>  		trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu, TPS("WakeNot"));
>  	}

Is it possible for the RCU_NOCP_WAKE write to overwrite a WAKE_FORCE ?
If not, why not? (Would make a good comment thereabouts).

Re: [PATCH tip/core/rcu 11/15] rcu: Break more call_rcu() deadlock involving scheduler and perf

[Paul E. McKenney]

On Mon, Sep 01, 2014 at 01:29:34PM +0200, Peter Zijlstra wrote:
> On Thu, Aug 28, 2014 at 11:26:30AM -0700, Paul E. McKenney wrote:
> 
> > diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h
> > index b6acb9340192..a1af86099a67 100644
> > --- a/kernel/rcu/tree_plugin.h
> > +++ b/kernel/rcu/tree_plugin.h
> > @@ -2121,16 +2121,23 @@ static void __call_rcu_nocb_enqueue(struct rcu_data *rdp,
> >  			trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
> >  					    TPS("WakeEmpty"));
> >  		} else {
> > -			rdp->nocb_defer_wakeup = true;
> > +			rdp->nocb_defer_wakeup = RCU_NOGP_WAKE;
> >  			trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
> >  					    TPS("WakeEmptyIsDeferred"));
> >  		}
> >  		rdp->qlen_last_fqs_check = 0;
> >  	} else if (len > rdp->qlen_last_fqs_check + qhimark) {
> >  		/* ... or if many callbacks queued. */
> > -		wake_nocb_leader(rdp, true);
> > +		if (!irqs_disabled_flags(flags)) {
> > +			wake_nocb_leader(rdp, true);
> > +			trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
> > +					    TPS("WakeOvf"));
> > +		} else {
> > +			rdp->nocb_defer_wakeup = RCU_NOGP_WAKE_FORCE;
> > +			trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
> > +					    TPS("WakeOvfIsDeferred"));
> > +		}
> >  		rdp->qlen_last_fqs_check = LONG_MAX / 2;
> > -		trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu, TPS("WakeOvf"));
> >  	} else {
> >  		trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu, TPS("WakeNot"));
> >  	}
> 
> Is it possible for the RCU_NOCP_WAKE write to overwrite a WAKE_FORCE ?
> If not, why not? (Would make a good comment thereabouts).

Good point.

So RCU_NOGP_WAKE only happens on the empty-to-non-empty transition,
and that RCU_NOGP_WAKE_FORCE only happens once the queue has a large
number of entries (10,000 by default).  Because this code runs with irqs
disabled (and must because you can do call_rcu() from interrupt handlers),
RCU_NOGP_WAKE_FORCE could overwrite RCU_NOGP_WAKE, but not vice versa.

However, the same effect can be obtained by having the setting of
RCU_NOGP_WAKE_FORCE race with the call to do_nocb_deferred_wakeup().
However, this would require that the call to do_nocb_deferred_wakeup()
was delayed long enough for 9,999 additional callbacks be registered,
and the next 10,000 callbacks will do another RCU_NOGP_WAKE_FORCE.
In addition, the RCU_NOGP_WAKE_FORCE shortens delays in the rcuo kthreads
rather than waking a rcuo kthread that would otherwise sleep indefinitely.
So I didn't see the point of closing this window.

							Thanx, Paul

[PATCH tip/core/rcu 12/15] rcu: Make TINY_RCU tinier by putting error checks under #ifdef

[Paul E. McKenney]

From: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>

The rcu_idle_enter_common() and rcu_idle_exit_common() functions contain
error checks that have to the best of my knowledge have never triggered
over the past several years.  These are nevertheless valuable when
creating new architectures or doing other low-level changes, so the
checks should not be deleted.  This commit instead places these checks
under #ifdef CONFIG_RCU_TRACE so that they are executed only when
specifically requested.

The savings are significant:

	Before:

	   text    data     bss     dec     hex filename
	   1749      39       0    1788     6fc /tmp/b/kernel/rcu/tiny.o
	    632     152       0     784     310 /tmp/b/kernel/rcu/update.o
				   ----
				   2572

	After:

	   text    data     bss     dec     hex filename
	   1281      37       0    1318     526 /tmp/b/kernel/rcu/tiny.o
	    632     152       0     784     310 /tmp/b/kernel/rcu/update.o
				   ----
				   2102

This amounts to 470 bytes, or 18% of the original.

Switched from #ifdef to IS_ENABLED() on Josh Triplett's advice.

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
---
 kernel/rcu/tiny.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/rcu/tiny.c b/kernel/rcu/tiny.c
index 6bd785c34add..4a55a2416e3c 100644
--- a/kernel/rcu/tiny.c
+++ b/kernel/rcu/tiny.c
@@ -62,7 +62,7 @@ static void rcu_idle_enter_common(long long newval)
 	}
 	RCU_TRACE(trace_rcu_dyntick(TPS("Start"),
 				    rcu_dynticks_nesting, newval));
-	if (!is_idle_task(current)) {
+	if (IS_ENABLED(CONFIG_RCU_TRACE) && !is_idle_task(current)) {
 		struct task_struct *idle __maybe_unused = idle_task(smp_processor_id());
 
 		RCU_TRACE(trace_rcu_dyntick(TPS("Entry error: not idle task"),
@@ -123,7 +123,7 @@ static void rcu_idle_exit_common(long long oldval)
 		return;
 	}
 	RCU_TRACE(trace_rcu_dyntick(TPS("End"), oldval, rcu_dynticks_nesting));
-	if (!is_idle_task(current)) {
+	if (IS_ENABLED(CONFIG_RCU_TRACE) && !is_idle_task(current)) {
 		struct task_struct *idle __maybe_unused = idle_task(smp_processor_id());
 
 		RCU_TRACE(trace_rcu_dyntick(TPS("Exit error: not idle task"),
-- 
1.8.1.5

[PATCH tip/core/rcu 13/15] rcu: Use rcu_gp_kthread_wake() to wake up grace period kthreads

[Paul E. McKenney]

From: Pranith Kumar <bobby.prani@gmail.com>

The rcu_gp_kthread_wake() function checks for three conditions before
waking up grace period kthreads:

*  Is the thread we are trying to wake up the current thread?
*  Are the gp_flags zero? (all threads wait on non-zero gp_flags condition)
*  Is there no thread created for this flavour, hence nothing to wake up?

If any one of these condition is true, we do not call wake_up().
It was found that there are quite a few avoidable wake ups both during
idle time and under stress induced by rcutorture.

Idle:

Total:66000, unnecessary:66000, case1:61827, case2:66000, case3:0
Total:68000, unnecessary:68000, case1:63696, case2:68000, case3:0

rcutorture:

Total:254000, unnecessary:254000, case1:199913, case2:254000, case3:0
Total:256000, unnecessary:256000, case1:201784, case2:256000, case3:0

Here case{1-3} are the cases listed above. We can avoid these wake
ups by using rcu_gp_kthread_wake() to conditionally wake up the grace
period kthreads.

There is a comment about an implied barrier supplied by the wake_up()
logic.  This barrier is necessary for the awakened thread to see the
updated ->gp_flags.  This flag is always being updated with the root node
lock held. Also, the awakened thread tries to acquire the root node lock
before reading ->gp_flags because of which there is proper ordering.

Hence this commit tries to avoid calling wake_up() whenever we can by
using rcu_gp_kthread_wake() function.

Signed-off-by: Pranith Kumar <bobby.prani@gmail.com>
CC: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
 kernel/rcu/tree.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index dd6c8b519691..9e83cd9a32f1 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -1938,7 +1938,7 @@ static void rcu_report_qs_rsp(struct rcu_state *rsp, unsigned long flags)
 {
 	WARN_ON_ONCE(!rcu_gp_in_progress(rsp));
 	raw_spin_unlock_irqrestore(&rcu_get_root(rsp)->lock, flags);
-	wake_up(&rsp->gp_wq);  /* Memory barrier implied by wake_up() path. */
+	rcu_gp_kthread_wake(rsp);
 }
 
 /*
@@ -2516,7 +2516,7 @@ static void force_quiescent_state(struct rcu_state *rsp)
 	ACCESS_ONCE(rsp->gp_flags) =
 		ACCESS_ONCE(rsp->gp_flags) | RCU_GP_FLAG_FQS;
 	raw_spin_unlock_irqrestore(&rnp_old->lock, flags);
-	wake_up(&rsp->gp_wq);  /* Memory barrier implied by wake_up() path. */
+	rcu_gp_kthread_wake(rsp);
 }
 
 /*
-- 
1.8.1.5

[PATCH tip/core/rcu 14/15] rcu: Replace flush_signals() with WARN_ON(signal_pending())

[Paul E. McKenney]

From: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>

Currently, when RCU awakens from a wait_event_interruptible() that
might have awakened prematurely, it does a flush_signals(). This is
done on the off-chance that someone figured out how to deliver a signal
to a kthread, which is supposed to be impossible.  Given that this
is supposed to be impossible, this commit changes the flush_signals()
calls into WARN_ON(signal_pending()).

Reported-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
 kernel/rcu/tree.c        | 4 ++--
 kernel/rcu/tree_plugin.h | 6 +++---
 2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index 9e83cd9a32f1..3e002c1cb441 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -1796,7 +1796,7 @@ static int __noreturn rcu_gp_kthread(void *arg)
 			if (rcu_gp_init(rsp))
 				break;
 			cond_resched();
-			flush_signals(current);
+			WARN_ON(signal_pending(current));
 			trace_rcu_grace_period(rsp->name,
 					       ACCESS_ONCE(rsp->gpnum),
 					       TPS("reqwaitsig"));
@@ -1842,7 +1842,7 @@ static int __noreturn rcu_gp_kthread(void *arg)
 			} else {
 				/* Deal with stray signal. */
 				cond_resched();
-				flush_signals(current);
+				WARN_ON(signal_pending(current));
 				trace_rcu_grace_period(rsp->name,
 						       ACCESS_ONCE(rsp->gpnum),
 						       TPS("fqswaitsig"));
diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h
index a1af86099a67..ce62cf3de2a2 100644
--- a/kernel/rcu/tree_plugin.h
+++ b/kernel/rcu/tree_plugin.h
@@ -2237,7 +2237,7 @@ static void rcu_nocb_wait_gp(struct rcu_data *rdp)
 			(d = ULONG_CMP_GE(ACCESS_ONCE(rnp->completed), c)));
 		if (likely(d))
 			break;
-		flush_signals(current);
+		WARN_ON(signal_pending(current));
 		trace_rcu_future_gp(rnp, rdp, c, TPS("ResumeWait"));
 	}
 	trace_rcu_future_gp(rnp, rdp, c, TPS("EndWait"));
@@ -2296,7 +2296,7 @@ wait_again:
 		if (!rcu_nocb_poll)
 			trace_rcu_nocb_wake(my_rdp->rsp->name, my_rdp->cpu,
 					    "WokeEmpty");
-		flush_signals(current);
+		WARN_ON(signal_pending(current));
 		schedule_timeout_interruptible(1);
 
 		/* Rescan in case we were a victim of memory ordering. */
@@ -2375,7 +2375,7 @@ static void nocb_follower_wait(struct rcu_data *rdp)
 		if (!rcu_nocb_poll)
 			trace_rcu_nocb_wake(rdp->rsp->name, rdp->cpu,
 					    "WokeEmpty");
-		flush_signals(current);
+		WARN_ON(signal_pending(current));
 		schedule_timeout_interruptible(1);
 	}
 }
-- 
1.8.1.5

[PATCH tip/core/rcu 15/15] rcu: Add ACCESS_ONCE() for RCU_INIT_POINTER()

[Paul E. McKenney]

From: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>

It is permissible to use RCU_INIT_POINTER() instead of rcu_assign_pointer()
in a couple of cases where readers might be concurrently accessing the
pointer being assigned to: (1) When assigning NULL and (2) When assigning
a pointer that is already visible to readers.  In these two cases, we
should really be using ACCESS_ONCE() to prevent compiler mischief.  This
commit therefore adds the ACCESS_ONCE().

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
 include/linux/rcupdate.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h
index 321ed0d4e675..7ba3f2470396 100644
--- a/include/linux/rcupdate.h
+++ b/include/linux/rcupdate.h
@@ -989,7 +989,7 @@ static inline notrace void rcu_read_unlock_sched_notrace(void)
  */
 #define RCU_INIT_POINTER(p, v) \
 	do { \
-		p = RCU_INITIALIZER(v); \
+		ACCESS_ONCE(p) = (v); \
 	} while (0)
 
 /**
-- 
1.8.1.5