Re: [ANNOUNCE] v4.14.29-rt25

[Sebastian Andrzej Siewior <bigeasy@linutronix.de>]

On 2018-03-27 13:01:07 [-0500], Grygorii Strashko wrote:
> Hi Sebastian,
Hi Grygorii,

> I've took this RT version and applied "[RT] kernel/time/posix-timer: avoid schedule()
> while holding the RCU lock" [1] on top. Then I run below tests:
…
> no stall or crashes were observed, but I've caught two "rcu_note_context_switch()" warnings

the warning is unrelated to the patch I posted. This should mute it:

Subject: [PATCH RT] rtmutex: annotate sleeping lock context

The RCU code complains on schedule() within a rcu_readlock() section.
The valid scenario on -RT is if a sleeping is held. In order to suppress
the warning the mirgrate_disable counter was used to identify the
invocation of schedule() due to lock contention.

Grygorii Strashko report that during CPU hotplug we might see the
warning via
 rt_spin_lock() -> migrate_disable() -> pin_current_cpu() -> __read_rt_lock()

because the counter is not yet set.
It is also possible to trigger the warning from cpu_chill()
(seen on a kblockd_mod_delayed_work_on() caller).

To address this RCU warning I annotate the sleeping lock context. The
counter is incremented before migrate_disable() so the warning Grygorii
should not trigger anymore. Additionally I use that counter in
cpu_chill() to avoid the RCU warning from there.

Reported-by: Grygorii Strashko <grygorii.strashko@ti.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
---
 include/linux/sched.h      | 20 ++++++++++++++++++++
 kernel/locking/rtmutex.c   | 12 ++++++++++--
 kernel/locking/rwlock-rt.c | 18 ++++++++++++++----
 kernel/rcu/tree_plugin.h   |  8 ++++----
 kernel/time/hrtimer.c      |  2 ++
 5 files changed, 50 insertions(+), 10 deletions(-)

diff --git a/include/linux/sched.h b/include/linux/sched.h
index cb78e2785f60..e90a15f38bf5 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -613,6 +613,9 @@ struct task_struct {
 	int				migrate_disable_atomic;
 # endif
 #endif
+#ifdef CONFIG_PREEMPT_RT_FULL
+	int				sleeping_lock;
+#endif
 
 #ifdef CONFIG_PREEMPT_RCU
 	int				rcu_read_lock_nesting;
@@ -1774,6 +1777,23 @@ static __always_inline bool need_resched(void)
 	return unlikely(tif_need_resched());
 }
 
+#ifdef CONFIG_PREEMPT_RT_FULL
+static inline void sleeping_lock_inc(void)
+{
+	current->sleeping_lock++;
+}
+
+static inline void sleeping_lock_dec(void)
+{
+	current->sleeping_lock--;
+}
+
+#else
+
+static inline void sleeping_lock_inc(void) { }
+static inline void sleeping_lock_dec(void) { }
+#endif
+
 /*
  * Wrappers for p->thread_info->cpu access. No-op on UP.
  */
diff --git a/kernel/locking/rtmutex.c b/kernel/locking/rtmutex.c
index 65cf3f6851eb..761a910b8489 100644
--- a/kernel/locking/rtmutex.c
+++ b/kernel/locking/rtmutex.c
@@ -1141,6 +1141,7 @@ void __sched rt_spin_lock_slowunlock(struct rt_mutex *lock)
 
 void __lockfunc rt_spin_lock(spinlock_t *lock)
 {
+	sleeping_lock_inc();
 	migrate_disable();
 	spin_acquire(&lock->dep_map, 0, 0, _RET_IP_);
 	rt_spin_lock_fastlock(&lock->lock, rt_spin_lock_slowlock);
@@ -1155,6 +1156,7 @@ void __lockfunc __rt_spin_lock(struct rt_mutex *lock)
 #ifdef CONFIG_DEBUG_LOCK_ALLOC
 void __lockfunc rt_spin_lock_nested(spinlock_t *lock, int subclass)
 {
+	sleeping_lock_inc();
 	migrate_disable();
 	spin_acquire(&lock->dep_map, subclass, 0, _RET_IP_);
 	rt_spin_lock_fastlock(&lock->lock, rt_spin_lock_slowlock);
@@ -1168,6 +1170,7 @@ void __lockfunc rt_spin_unlock(spinlock_t *lock)
 	spin_release(&lock->dep_map, 1, _RET_IP_);
 	rt_spin_lock_fastunlock(&lock->lock, rt_spin_lock_slowunlock);
 	migrate_enable();
+	sleeping_lock_dec();
 }
 EXPORT_SYMBOL(rt_spin_unlock);
 
@@ -1193,12 +1196,15 @@ int __lockfunc rt_spin_trylock(spinlock_t *lock)
 {
 	int ret;
 
+	sleeping_lock_inc();
 	migrate_disable();
 	ret = __rt_mutex_trylock(&lock->lock);
-	if (ret)
+	if (ret) {
 		spin_acquire(&lock->dep_map, 0, 1, _RET_IP_);
-	else
+	} else {
 		migrate_enable();
+		sleeping_lock_dec();
+	}
 	return ret;
 }
 EXPORT_SYMBOL(rt_spin_trylock);
@@ -1210,6 +1216,7 @@ int __lockfunc rt_spin_trylock_bh(spinlock_t *lock)
 	local_bh_disable();
 	ret = __rt_mutex_trylock(&lock->lock);
 	if (ret) {
+		sleeping_lock_inc();
 		migrate_disable();
 		spin_acquire(&lock->dep_map, 0, 1, _RET_IP_);
 	} else
@@ -1225,6 +1232,7 @@ int __lockfunc rt_spin_trylock_irqsave(spinlock_t *lock, unsigned long *flags)
 	*flags = 0;
 	ret = __rt_mutex_trylock(&lock->lock);
 	if (ret) {
+		sleeping_lock_inc();
 		migrate_disable();
 		spin_acquire(&lock->dep_map, 0, 1, _RET_IP_);
 	}
diff --git a/kernel/locking/rwlock-rt.c b/kernel/locking/rwlock-rt.c
index aebb7ce25bc6..f2e155b2c4a8 100644
--- a/kernel/locking/rwlock-rt.c
+++ b/kernel/locking/rwlock-rt.c
@@ -305,12 +305,15 @@ int __lockfunc rt_read_trylock(rwlock_t *rwlock)
 {
 	int ret;
 
+	sleeping_lock_inc();
 	migrate_disable();
 	ret = do_read_rt_trylock(rwlock);
-	if (ret)
+	if (ret) {
 		rwlock_acquire_read(&rwlock->dep_map, 0, 1, _RET_IP_);
-	else
+	} else {
 		migrate_enable();
+		sleeping_lock_dec();
+	}
 	return ret;
 }
 EXPORT_SYMBOL(rt_read_trylock);
@@ -319,18 +322,22 @@ int __lockfunc rt_write_trylock(rwlock_t *rwlock)
 {
 	int ret;
 
+	sleeping_lock_inc();
 	migrate_disable();
 	ret = do_write_rt_trylock(rwlock);
-	if (ret)
+	if (ret) {
 		rwlock_acquire(&rwlock->dep_map, 0, 1, _RET_IP_);
-	else
+	} else {
 		migrate_enable();
+		sleeping_lock_dec();
+	}
 	return ret;
 }
 EXPORT_SYMBOL(rt_write_trylock);
 
 void __lockfunc rt_read_lock(rwlock_t *rwlock)
 {
+	sleeping_lock_inc();
 	migrate_disable();
 	rwlock_acquire_read(&rwlock->dep_map, 0, 0, _RET_IP_);
 	do_read_rt_lock(rwlock);
@@ -339,6 +346,7 @@ EXPORT_SYMBOL(rt_read_lock);
 
 void __lockfunc rt_write_lock(rwlock_t *rwlock)
 {
+	sleeping_lock_inc();
 	migrate_disable();
 	rwlock_acquire(&rwlock->dep_map, 0, 0, _RET_IP_);
 	do_write_rt_lock(rwlock);
@@ -350,6 +358,7 @@ void __lockfunc rt_read_unlock(rwlock_t *rwlock)
 	rwlock_release(&rwlock->dep_map, 1, _RET_IP_);
 	do_read_rt_unlock(rwlock);
 	migrate_enable();
+	sleeping_lock_dec();
 }
 EXPORT_SYMBOL(rt_read_unlock);
 
@@ -358,6 +367,7 @@ void __lockfunc rt_write_unlock(rwlock_t *rwlock)
 	rwlock_release(&rwlock->dep_map, 1, _RET_IP_);
 	do_write_rt_unlock(rwlock);
 	migrate_enable();
+	sleeping_lock_dec();
 }
 EXPORT_SYMBOL(rt_write_unlock);
 
diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h
index 3933026838eb..181afc1ab6e2 100644
--- a/kernel/rcu/tree_plugin.h
+++ b/kernel/rcu/tree_plugin.h
@@ -300,13 +300,13 @@ static void rcu_preempt_note_context_switch(bool preempt)
 	struct task_struct *t = current;
 	struct rcu_data *rdp;
 	struct rcu_node *rnp;
-	int mg_counter = 0;
+	int sleeping_l = 0;
 
 	lockdep_assert_irqs_disabled();
-#if defined(CONFIG_PREEMPT_RT_BASE)
-	mg_counter = t->migrate_disable;
+#if defined(CONFIG_PREEMPT_RT_FULL)
+	sleeping_l = t->sleeping_lock;
 #endif
-	WARN_ON_ONCE(!preempt && t->rcu_read_lock_nesting > 0 && !mg_counter);
+	WARN_ON_ONCE(!preempt && t->rcu_read_lock_nesting > 0 && !sleeping_l);
 	if (t->rcu_read_lock_nesting > 0 &&
 	    !t->rcu_read_unlock_special.b.blocked) {
 
diff --git a/kernel/time/hrtimer.c b/kernel/time/hrtimer.c
index 43c2f571a842..9e1b766dc35c 100644
--- a/kernel/time/hrtimer.c
+++ b/kernel/time/hrtimer.c
@@ -1854,7 +1854,9 @@ void cpu_chill(void)
 	chill_time = ktime_set(0, NSEC_PER_MSEC);
 	set_current_state(TASK_UNINTERRUPTIBLE);
 	current->flags |= PF_NOFREEZE;
+	sleeping_lock_inc();
 	schedule_hrtimeout(&chill_time, HRTIMER_MODE_REL_HARD);
+	sleeping_lock_dec();
 	if (!freeze_flag)
 		current->flags &= ~PF_NOFREEZE;
 }
-- 
2.17.0