public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
* [PATCH] rcu: Remove superfluous versions of rcu_read_lock_sched_held()
@ 2016-03-23 15:11 Boqun Feng
  2016-03-23 16:38 ` Paul E. McKenney
  0 siblings, 1 reply; 3+ messages in thread
From: Boqun Feng @ 2016-03-23 15:11 UTC (permalink / raw)
  To: linux-kernel
  Cc: Paul E. McKenney, Josh Triplett, Steven Rostedt,
	Mathieu Desnoyers, Lai Jiangshan, Frederic Weisbecker, Boqun Feng

Currently, we have four versions of rcu_read_lock_sched_held(),
depending on the combined choices on PREEMPT_COUNT and DEBUG_LOCK_ALLOC.
But we actually don't need to specialize those for PREEMPT_COUNT=n
kernel. Because:

1.	For the implementations in DEBUG_LOCK_ALLOC=n kernel, we can use
	preemptible() to implement one rcu_read_lock_sched_held(), which
	gives us the same behavior as the current two.

2.	For the implementations in DEBUG_LOCK_ALLOC=y kernel, even when
	PREEMPT_COUNT=n, one CPU may block no grace period because of
	the same reason for the PREEMPT_COUNT=y and !PREEMPT kernel.
	(e.g. dynticks or cpu hotplug)

So unify the implementations of rcu_read_lock_sched_held() by using
macro preemptible() and tightening up the lock-held checking for
PREEMPT_COUNT=n kernel. And this will improve the readability, make the
debug checking as expected and save several lines of code.

Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
---
 include/linux/rcupdate.h | 17 +----------------
 kernel/rcu/update.c      |  4 ++--
 2 files changed, 3 insertions(+), 18 deletions(-)

diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h
index b5d48bd..e3f845b 100644
--- a/include/linux/rcupdate.h
+++ b/include/linux/rcupdate.h
@@ -510,14 +510,7 @@ int rcu_read_lock_bh_held(void);
  * CONFIG_DEBUG_LOCK_ALLOC, this assumes we are in an RCU-sched read-side
  * critical section unless it can prove otherwise.
  */
-#ifdef CONFIG_PREEMPT_COUNT
 int rcu_read_lock_sched_held(void);
-#else /* #ifdef CONFIG_PREEMPT_COUNT */
-static inline int rcu_read_lock_sched_held(void)
-{
-	return 1;
-}
-#endif /* #else #ifdef CONFIG_PREEMPT_COUNT */
 
 #else /* #ifdef CONFIG_DEBUG_LOCK_ALLOC */
 
@@ -534,18 +527,10 @@ static inline int rcu_read_lock_bh_held(void)
 	return 1;
 }
 
-#ifdef CONFIG_PREEMPT_COUNT
 static inline int rcu_read_lock_sched_held(void)
 {
-	return preempt_count() != 0 || irqs_disabled();
+	return !preemptible();
 }
-#else /* #ifdef CONFIG_PREEMPT_COUNT */
-static inline int rcu_read_lock_sched_held(void)
-{
-	return 1;
-}
-#endif /* #else #ifdef CONFIG_PREEMPT_COUNT */
-
 #endif /* #else #ifdef CONFIG_DEBUG_LOCK_ALLOC */
 
 #ifdef CONFIG_PROVE_RCU
diff --git a/kernel/rcu/update.c b/kernel/rcu/update.c
index ca828b4..3ccdc8e 100644
--- a/kernel/rcu/update.c
+++ b/kernel/rcu/update.c
@@ -67,7 +67,7 @@ static int rcu_normal_after_boot;
 module_param(rcu_normal_after_boot, int, 0);
 #endif /* #ifndef CONFIG_TINY_RCU */
 
-#if defined(CONFIG_DEBUG_LOCK_ALLOC) && defined(CONFIG_PREEMPT_COUNT)
+#ifdef CONFIG_DEBUG_LOCK_ALLOC
 /**
  * rcu_read_lock_sched_held() - might we be in RCU-sched read-side critical section?
  *
@@ -111,7 +111,7 @@ int rcu_read_lock_sched_held(void)
 		return 0;
 	if (debug_locks)
 		lockdep_opinion = lock_is_held(&rcu_sched_lock_map);
-	return lockdep_opinion || preempt_count() != 0 || irqs_disabled();
+	return lockdep_opinion || !preemptible();
 }
 EXPORT_SYMBOL(rcu_read_lock_sched_held);
 #endif
-- 
2.7.3

^ permalink raw reply related	[flat|nested] 3+ messages in thread

* Re: [PATCH] rcu: Remove superfluous versions of rcu_read_lock_sched_held()
  2016-03-23 15:11 [PATCH] rcu: Remove superfluous versions of rcu_read_lock_sched_held() Boqun Feng
@ 2016-03-23 16:38 ` Paul E. McKenney
  2016-03-24  1:02   ` Boqun Feng
  0 siblings, 1 reply; 3+ messages in thread
From: Paul E. McKenney @ 2016-03-23 16:38 UTC (permalink / raw)
  To: Boqun Feng
  Cc: linux-kernel, Josh Triplett, Steven Rostedt, Mathieu Desnoyers,
	Lai Jiangshan, Frederic Weisbecker

On Wed, Mar 23, 2016 at 11:11:48PM +0800, Boqun Feng wrote:
> Currently, we have four versions of rcu_read_lock_sched_held(),
> depending on the combined choices on PREEMPT_COUNT and DEBUG_LOCK_ALLOC.
> But we actually don't need to specialize those for PREEMPT_COUNT=n
> kernel. Because:
> 
> 1.	For the implementations in DEBUG_LOCK_ALLOC=n kernel, we can use
> 	preemptible() to implement one rcu_read_lock_sched_held(), which
> 	gives us the same behavior as the current two.
> 
> 2.	For the implementations in DEBUG_LOCK_ALLOC=y kernel, even when
> 	PREEMPT_COUNT=n, one CPU may block no grace period because of
> 	the same reason for the PREEMPT_COUNT=y and !PREEMPT kernel.
> 	(e.g. dynticks or cpu hotplug)
> 
> So unify the implementations of rcu_read_lock_sched_held() by using
> macro preemptible() and tightening up the lock-held checking for
> PREEMPT_COUNT=n kernel. And this will improve the readability, make the
> debug checking as expected and save several lines of code.
> 
> Signed-off-by: Boqun Feng <boqun.feng@gmail.com>

Looks like a nice consolidation!  I have queued this for review and
testing, updating the commit log as follows:

	Currently, we have four versions of rcu_read_lock_sched_held(),
	depending on the combined choices on PREEMPT_COUNT and
	DEBUG_LOCK_ALLOC.  However, there is an existing function
	preemptible() that already distinguishes between the
	PREEMPT_COUNT=y and PREEMPT_COUNT=n cases, and allows these four
	implementations to be consolidated down to two.
		    
	This commit therefore uses preemptible() to achieve this
	consolidation.	Note that there could be a small performance
	regression in the case of CONFIG_DEBUG_LOCK_ALLOC=y &&
	PREEMPT_COUNT=n.  However, given the overhead associated with
	CONFIG_DEBUG_LOCK_ALLOC=y, this should be down in the noise.

Does that capture it?

							Thanx, Paul

> ---
>  include/linux/rcupdate.h | 17 +----------------
>  kernel/rcu/update.c      |  4 ++--
>  2 files changed, 3 insertions(+), 18 deletions(-)
> 
> diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h
> index b5d48bd..e3f845b 100644
> --- a/include/linux/rcupdate.h
> +++ b/include/linux/rcupdate.h
> @@ -510,14 +510,7 @@ int rcu_read_lock_bh_held(void);
>   * CONFIG_DEBUG_LOCK_ALLOC, this assumes we are in an RCU-sched read-side
>   * critical section unless it can prove otherwise.
>   */
> -#ifdef CONFIG_PREEMPT_COUNT
>  int rcu_read_lock_sched_held(void);
> -#else /* #ifdef CONFIG_PREEMPT_COUNT */
> -static inline int rcu_read_lock_sched_held(void)
> -{
> -	return 1;
> -}
> -#endif /* #else #ifdef CONFIG_PREEMPT_COUNT */
> 
>  #else /* #ifdef CONFIG_DEBUG_LOCK_ALLOC */
> 
> @@ -534,18 +527,10 @@ static inline int rcu_read_lock_bh_held(void)
>  	return 1;
>  }
> 
> -#ifdef CONFIG_PREEMPT_COUNT
>  static inline int rcu_read_lock_sched_held(void)
>  {
> -	return preempt_count() != 0 || irqs_disabled();
> +	return !preemptible();
>  }
> -#else /* #ifdef CONFIG_PREEMPT_COUNT */
> -static inline int rcu_read_lock_sched_held(void)
> -{
> -	return 1;
> -}
> -#endif /* #else #ifdef CONFIG_PREEMPT_COUNT */
> -
>  #endif /* #else #ifdef CONFIG_DEBUG_LOCK_ALLOC */
> 
>  #ifdef CONFIG_PROVE_RCU
> diff --git a/kernel/rcu/update.c b/kernel/rcu/update.c
> index ca828b4..3ccdc8e 100644
> --- a/kernel/rcu/update.c
> +++ b/kernel/rcu/update.c
> @@ -67,7 +67,7 @@ static int rcu_normal_after_boot;
>  module_param(rcu_normal_after_boot, int, 0);
>  #endif /* #ifndef CONFIG_TINY_RCU */
> 
> -#if defined(CONFIG_DEBUG_LOCK_ALLOC) && defined(CONFIG_PREEMPT_COUNT)
> +#ifdef CONFIG_DEBUG_LOCK_ALLOC
>  /**
>   * rcu_read_lock_sched_held() - might we be in RCU-sched read-side critical section?
>   *
> @@ -111,7 +111,7 @@ int rcu_read_lock_sched_held(void)
>  		return 0;
>  	if (debug_locks)
>  		lockdep_opinion = lock_is_held(&rcu_sched_lock_map);
> -	return lockdep_opinion || preempt_count() != 0 || irqs_disabled();
> +	return lockdep_opinion || !preemptible();
>  }
>  EXPORT_SYMBOL(rcu_read_lock_sched_held);
>  #endif
> -- 
> 2.7.3
> 

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: [PATCH] rcu: Remove superfluous versions of rcu_read_lock_sched_held()
  2016-03-23 16:38 ` Paul E. McKenney
@ 2016-03-24  1:02   ` Boqun Feng
  0 siblings, 0 replies; 3+ messages in thread
From: Boqun Feng @ 2016-03-24  1:02 UTC (permalink / raw)
  To: Paul E. McKenney
  Cc: linux-kernel, Josh Triplett, Steven Rostedt, Mathieu Desnoyers,
	Lai Jiangshan, Frederic Weisbecker

[-- Attachment #1: Type: text/plain, Size: 4469 bytes --]

On Wed, Mar 23, 2016 at 09:38:46AM -0700, Paul E. McKenney wrote:
> On Wed, Mar 23, 2016 at 11:11:48PM +0800, Boqun Feng wrote:
> > Currently, we have four versions of rcu_read_lock_sched_held(),
> > depending on the combined choices on PREEMPT_COUNT and DEBUG_LOCK_ALLOC.
> > But we actually don't need to specialize those for PREEMPT_COUNT=n
> > kernel. Because:
> > 
> > 1.	For the implementations in DEBUG_LOCK_ALLOC=n kernel, we can use
> > 	preemptible() to implement one rcu_read_lock_sched_held(), which
> > 	gives us the same behavior as the current two.
> > 
> > 2.	For the implementations in DEBUG_LOCK_ALLOC=y kernel, even when
> > 	PREEMPT_COUNT=n, one CPU may block no grace period because of
> > 	the same reason for the PREEMPT_COUNT=y and !PREEMPT kernel.
> > 	(e.g. dynticks or cpu hotplug)
> > 
> > So unify the implementations of rcu_read_lock_sched_held() by using
> > macro preemptible() and tightening up the lock-held checking for
> > PREEMPT_COUNT=n kernel. And this will improve the readability, make the
> > debug checking as expected and save several lines of code.
> > 
> > Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
> 
> Looks like a nice consolidation!  I have queued this for review and
> testing, updating the commit log as follows:
> 
> 	Currently, we have four versions of rcu_read_lock_sched_held(),
> 	depending on the combined choices on PREEMPT_COUNT and
> 	DEBUG_LOCK_ALLOC.  However, there is an existing function
> 	preemptible() that already distinguishes between the
> 	PREEMPT_COUNT=y and PREEMPT_COUNT=n cases, and allows these four
> 	implementations to be consolidated down to two.
> 		    
> 	This commit therefore uses preemptible() to achieve this
> 	consolidation.	Note that there could be a small performance
> 	regression in the case of CONFIG_DEBUG_LOCK_ALLOC=y &&
> 	PREEMPT_COUNT=n.  However, given the overhead associated with
> 	CONFIG_DEBUG_LOCK_ALLOC=y, this should be down in the noise.
> 
> Does that capture it?
> 

Yes, thank you ;-)

Regards,
Boqun

> 							Thanx, Paul
> 
> > ---
> >  include/linux/rcupdate.h | 17 +----------------
> >  kernel/rcu/update.c      |  4 ++--
> >  2 files changed, 3 insertions(+), 18 deletions(-)
> > 
> > diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h
> > index b5d48bd..e3f845b 100644
> > --- a/include/linux/rcupdate.h
> > +++ b/include/linux/rcupdate.h
> > @@ -510,14 +510,7 @@ int rcu_read_lock_bh_held(void);
> >   * CONFIG_DEBUG_LOCK_ALLOC, this assumes we are in an RCU-sched read-side
> >   * critical section unless it can prove otherwise.
> >   */
> > -#ifdef CONFIG_PREEMPT_COUNT
> >  int rcu_read_lock_sched_held(void);
> > -#else /* #ifdef CONFIG_PREEMPT_COUNT */
> > -static inline int rcu_read_lock_sched_held(void)
> > -{
> > -	return 1;
> > -}
> > -#endif /* #else #ifdef CONFIG_PREEMPT_COUNT */
> > 
> >  #else /* #ifdef CONFIG_DEBUG_LOCK_ALLOC */
> > 
> > @@ -534,18 +527,10 @@ static inline int rcu_read_lock_bh_held(void)
> >  	return 1;
> >  }
> > 
> > -#ifdef CONFIG_PREEMPT_COUNT
> >  static inline int rcu_read_lock_sched_held(void)
> >  {
> > -	return preempt_count() != 0 || irqs_disabled();
> > +	return !preemptible();
> >  }
> > -#else /* #ifdef CONFIG_PREEMPT_COUNT */
> > -static inline int rcu_read_lock_sched_held(void)
> > -{
> > -	return 1;
> > -}
> > -#endif /* #else #ifdef CONFIG_PREEMPT_COUNT */
> > -
> >  #endif /* #else #ifdef CONFIG_DEBUG_LOCK_ALLOC */
> > 
> >  #ifdef CONFIG_PROVE_RCU
> > diff --git a/kernel/rcu/update.c b/kernel/rcu/update.c
> > index ca828b4..3ccdc8e 100644
> > --- a/kernel/rcu/update.c
> > +++ b/kernel/rcu/update.c
> > @@ -67,7 +67,7 @@ static int rcu_normal_after_boot;
> >  module_param(rcu_normal_after_boot, int, 0);
> >  #endif /* #ifndef CONFIG_TINY_RCU */
> > 
> > -#if defined(CONFIG_DEBUG_LOCK_ALLOC) && defined(CONFIG_PREEMPT_COUNT)
> > +#ifdef CONFIG_DEBUG_LOCK_ALLOC
> >  /**
> >   * rcu_read_lock_sched_held() - might we be in RCU-sched read-side critical section?
> >   *
> > @@ -111,7 +111,7 @@ int rcu_read_lock_sched_held(void)
> >  		return 0;
> >  	if (debug_locks)
> >  		lockdep_opinion = lock_is_held(&rcu_sched_lock_map);
> > -	return lockdep_opinion || preempt_count() != 0 || irqs_disabled();
> > +	return lockdep_opinion || !preemptible();
> >  }
> >  EXPORT_SYMBOL(rcu_read_lock_sched_held);
> >  #endif
> > -- 
> > 2.7.3
> > 
> 

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 473 bytes --]

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2016-03-24  1:03 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2016-03-23 15:11 [PATCH] rcu: Remove superfluous versions of rcu_read_lock_sched_held() Boqun Feng
2016-03-23 16:38 ` Paul E. McKenney
2016-03-24  1:02   ` Boqun Feng

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox