[PATCH RFC tip/core/rcu 0/2] rcu: make RCU lockdep handle early boot in organized manner

[PATCH RFC tip/core/rcu 0/2] rcu: make RCU lockdep handle early boot in organized manner

[Paul E. McKenney]

Hello!

This patch set makes RCU lockdep handle early boot in an organized
manner:

1.	Make rcu_read_lock_held(), rcu_read_lock_bh_held(),
	rcu_read_lock_sched_held(), and rcu_dereference_check()
	suppress complaints if !rcu_scheduler_active.  The rationale
	for this approach is that RCU grace periods are handled very
	differently during early boot, and rcu_scheduler_active marks
	the point that normal runtime RCU behavior starts.

2.	Remove the !rcu_scheduler_active check from calls to
	rcu_dereference_check(), because rcu_dereference_check() already
	does the check in question.

Testing in progress, but want to get these out for review.

							Thanx, Paul

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

 cgroup.h   |    2 --
 rcupdate.h |   27 +++++++++++++++++----------
 2 files changed, 17 insertions(+), 12 deletions(-)

[PATCH RFC tip/core/rcu 1/2] rcu: suppress RCU lockdep warnings during early boot

[Paul E. McKenney]

RCU is used during very early boot, before RCU and lockdep have been
initialized.  So make the underlying primitives (rcu_read_lock_held(),
rcu_read_lock_bh_held(), rcu_read_lock_sched_held(), and
rcu_dereference_check()) check for early boot via the rcu_scheduler_active
flag.  This will suppress false positives.

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
 include/linux/rcupdate.h |   26 +++++++++++++++++---------
 1 files changed, 17 insertions(+), 9 deletions(-)

diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h
index c843736..af51d5f 100644
--- a/include/linux/rcupdate.h
+++ b/include/linux/rcupdate.h
@@ -104,12 +104,14 @@ extern struct lockdep_map rcu_sched_lock_map;
  * an RCU read-side critical section.  In absence of CONFIG_PROVE_LOCKING,
  * this assumes we are in an RCU read-side critical section unless it can
  * prove otherwise.
+ *
+ * Check rcu_scheduler_active to prevent false positives during boot.
  */
 static inline int rcu_read_lock_held(void)
 {
-	if (debug_locks)
-		return lock_is_held(&rcu_lock_map);
-	return 1;
+	if (!rcu_scheduler_active || !debug_locks)
+		return 1;
+	return lock_is_held(&rcu_lock_map);
 }
 
 /**
@@ -119,12 +121,14 @@ static inline int rcu_read_lock_held(void)
  * an RCU-bh read-side critical section.  In absence of CONFIG_PROVE_LOCKING,
  * this assumes we are in an RCU-bh read-side critical section unless it can
  * prove otherwise.
+ *
+ * Check rcu_scheduler_active to prevent false positives during boot.
  */
 static inline int rcu_read_lock_bh_held(void)
 {
-	if (debug_locks)
-		return lock_is_held(&rcu_bh_lock_map);
-	return 1;
+	if (!rcu_scheduler_active || !debug_locks)
+		return 1;
+	return lock_is_held(&rcu_bh_lock_map);
 }
 
 /**
@@ -135,14 +139,18 @@ static inline int rcu_read_lock_bh_held(void)
  * this assumes we are in an RCU-sched read-side critical section unless it
  * can prove otherwise.  Note that disabling of preemption (including
  * disabling irqs) counts as an RCU-sched read-side critical section.
+ *
+ * Check rcu_scheduler_active to prevent false positives during boot.
  */
 static inline int rcu_read_lock_sched_held(void)
 {
 	int lockdep_opinion = 0;
 
+	if (!rcu_scheduler_active || !debug_locks)
+		return 1;
 	if (debug_locks)
 		lockdep_opinion = lock_is_held(&rcu_sched_lock_map);
-	return lockdep_opinion || preempt_count() != 0 || !rcu_scheduler_active;
+	return lockdep_opinion || preempt_count() != 0;
 }
 
 #else /* #ifdef CONFIG_DEBUG_LOCK_ALLOC */
@@ -166,7 +174,7 @@ static inline int rcu_read_lock_bh_held(void)
 
 static inline int rcu_read_lock_sched_held(void)
 {
-	return preempt_count() != 0 || !rcu_scheduler_active;
+	return !rcu_scheduler_active || preempt_count() != 0;
 }
 
 #endif /* #else #ifdef CONFIG_DEBUG_LOCK_ALLOC */
@@ -184,7 +192,7 @@ static inline int rcu_read_lock_sched_held(void)
  */
 #define rcu_dereference_check(p, c) \
 	({ \
-		if (debug_locks && !(c)) \
+		if (rcu_scheduler_active && debug_locks && !(c)) \
 			lockdep_rcu_dereference(__FILE__, __LINE__); \
 		rcu_dereference_raw(p); \
 	})
-- 
1.6.6

Re: [PATCH RFC tip/core/rcu 1/2] rcu: suppress RCU lockdep warnings during early boot

[Ingo Molnar]

* Paul E. McKenney <paulmck@linux.vnet.ibm.com> wrote:

> +	if (!rcu_scheduler_active || !debug_locks)
> +		return 1;
> +	return lock_is_held(&rcu_lock_map);

> +	if (!rcu_scheduler_active || !debug_locks)
> +		return 1;
> +	return lock_is_held(&rcu_bh_lock_map);

i guess there could be a common helper here?

Also, could we clear rcu_scheduler_active when we clear debug_locks? That way 
only a single test is needed, a generic 'is lock debugging active'.

(Which test should probably be unlikely() as well?)

	Ingo

Re: [PATCH RFC tip/core/rcu 1/2] rcu: suppress RCU lockdep warnings during early boot

[Paul E. McKenney]

On Tue, Mar 02, 2010 at 01:20:59PM +0100, Ingo Molnar wrote:
> 
> * Paul E. McKenney <paulmck@linux.vnet.ibm.com> wrote:
> 
> > +	if (!rcu_scheduler_active || !debug_locks)
> > +		return 1;
> > +	return lock_is_held(&rcu_lock_map);
> 
> > +	if (!rcu_scheduler_active || !debug_locks)
> > +		return 1;
> > +	return lock_is_held(&rcu_bh_lock_map);
> 
> i guess there could be a common helper here?

Will do!

> Also, could we clear rcu_scheduler_active when we clear debug_locks? That way 
> only a single test is needed, a generic 'is lock debugging active'.

Doing that will break synchronize_rcu(), which returns immediately if
!rcu_scheduler_active.

> (Which test should probably be unlikely() as well?)

Good point, will fix.

							Thanx, Paul

Re: [PATCH RFC tip/core/rcu 1/2] rcu: suppress RCU lockdep warnings during early boot

[Ingo Molnar]

* Paul E. McKenney <paulmck@linux.vnet.ibm.com> wrote:

> On Tue, Mar 02, 2010 at 01:20:59PM +0100, Ingo Molnar wrote:
> > 
> > * Paul E. McKenney <paulmck@linux.vnet.ibm.com> wrote:
> > 
> > > +	if (!rcu_scheduler_active || !debug_locks)
> > > +		return 1;
> > > +	return lock_is_held(&rcu_lock_map);
> > 
> > > +	if (!rcu_scheduler_active || !debug_locks)
> > > +		return 1;
> > > +	return lock_is_held(&rcu_bh_lock_map);
> > 
> > i guess there could be a common helper here?
> 
> Will do!
> 
> > Also, could we clear rcu_scheduler_active when we clear debug_locks? That way 
> > only a single test is needed, a generic 'is lock debugging active'.
> 
> Doing that will break synchronize_rcu(), which returns immediately if
> !rcu_scheduler_active.

Ok - then have an debug_rcu flag which is cleared appropriately - so that the 
fastpath impact is reduced?

	Ingo

Re: [PATCH RFC tip/core/rcu 1/2] rcu: suppress RCU lockdep warnings during early boot

[Paul E. McKenney]

On Tue, Mar 02, 2010 at 03:12:57PM +0100, Ingo Molnar wrote:
> 
> * Paul E. McKenney <paulmck@linux.vnet.ibm.com> wrote:
> 
> > On Tue, Mar 02, 2010 at 01:20:59PM +0100, Ingo Molnar wrote:
> > > 
> > > * Paul E. McKenney <paulmck@linux.vnet.ibm.com> wrote:
> > > 
> > > > +	if (!rcu_scheduler_active || !debug_locks)
> > > > +		return 1;
> > > > +	return lock_is_held(&rcu_lock_map);
> > > 
> > > > +	if (!rcu_scheduler_active || !debug_locks)
> > > > +		return 1;
> > > > +	return lock_is_held(&rcu_bh_lock_map);
> > > 
> > > i guess there could be a common helper here?
> > 
> > Will do!
> > 
> > > Also, could we clear rcu_scheduler_active when we clear debug_locks? That way 
> > > only a single test is needed, a generic 'is lock debugging active'.
> > 
> > Doing that will break synchronize_rcu(), which returns immediately if
> > !rcu_scheduler_active.
> 
> Ok - then have an debug_rcu flag which is cleared appropriately - so that the 
> fastpath impact is reduced?

Good point, will do!

							Thanx, Paul

[PATCH RFC tip/core/rcu 2/2] rcu: revert 1883c79a: early boot now handled by lockdep-RCU

[Paul E. McKenney]

This patch removes the check for !rcu_scheduler_active because this
check has been incorporated into rcu_dereference_check().

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
---
 include/linux/cgroup.h |    1 -
 1 files changed, 0 insertions(+), 1 deletions(-)

diff --git a/include/linux/cgroup.h b/include/linux/cgroup.h
index a73e1ce..c9bbcb2 100644
--- a/include/linux/cgroup.h
+++ b/include/linux/cgroup.h
@@ -489,7 +489,6 @@ static inline struct cgroup_subsys_state *task_subsys_state(
 {
 	return rcu_dereference_check(task->cgroups->subsys[subsys_id],
 				     rcu_read_lock_held() ||
-				     !rcu_scheduler_active ||
 				     cgroup_lock_is_held());
 }
 
-- 
1.6.6