fixes to include/linux/rcupreempt.h

PATCH Applied Paul E. McKenney 3 participants 5 messages Aug 1, 2008 – Aug 21, 2008 RCU

lore

Share & tools

lore

Paul E. McKenney

author

2008-08-01 21:14 UTC #0 lore raw

PATCH 1/1 fixes to include/linux/rcupreempt.h

Hello!

Compared tip/core/rcu to my latest patchset, and found the following issues:

o the memory barrier in rcu_exit_nohz() somehow got out of place
(it is correct in mainline as of 2.6.26-rc7).

o There is a duplicate declaration of rcu_dyntick_sched.

The attached patch fixes these.

Signed-off-by Paul E. McKenney <[email protected]>

---

include/linux/rcupreempt.h +1 -2

index f04b64e..84678bf 100644
@@ -111,7 +111,6 @@ extern struct rcupreempt_trace *rcupreempt_trace_cpu(int cpu);
 struct softirq_action;
 
 #ifdef CONFIG_NO_HZ
-DECLARE_PER_CPU(struct rcu_dyntick_sched, rcu_dyntick_sched);
 
 static inline void rcu_enter_nohz(void)
 {

@@ -122,8 +121,8 @@ static inline void rcu_enter_nohz(void)
 
 static inline void rcu_exit_nohz(void)
 {
-	smp_mb(); /* CPUs seeing ++ must see later RCU read-side crit sects */
 	__get_cpu_var(rcu_dyntick_sched).dynticks++;
+	smp_mb(); /* CPUs seeing ++ must see later RCU read-side crit sects */
 	WARN_ON(!(__get_cpu_var(rcu_dyntick_sched).dynticks & 0x1));
 }

Ingo Molnar

Applied

2008-08-15 15:02 UTC #1 lore raw

Quoted 5 lines · Paul E. McKenney

Hello!
Compared tip/core/rcu to my latest patchset, and found the following issues:

Quoted 8 lines

o the memory barrier in rcu_exit_nohz() somehow got out of place
(it is correct in mainline as of 2.6.26-rc7).
o There is a duplicate declaration of rcu_dyntick_sched.
The attached patch fixes these.
Signed-off-by Paul E. McKenney <[email protected]>

applied to tip/core/rcu - thanks Paul.

Ingo

Steven Rostedt

reviewer

2008-08-20 16:17 UTC #2 lore raw

In reply to Paul E. McKenney

Sorry for the late post, I'm finally looking at all my -rt mail.

Quoted 20 lines · Paul E. McKenney

Hello!
Compared tip/core/rcu to my latest patchset, and found the following issues:
o the memory barrier in rcu_exit_nohz() somehow got out of place
(it is correct in mainline as of 2.6.26-rc7).
o There is a duplicate declaration of rcu_dyntick_sched.
The attached patch fixes these.
Signed-off-by Paul E. McKenney <[email protected]>
---
include/linux/rcupreempt.h +0 -0
index f04b64e..84678bf 100644

include/linux/rcupreempt.h extern struct rcupreempt_trace *rcupreempt_trace_cpu(int cpu); · line 114

Quoted 4 lines

include/linux/rcupreempt.h +0 -1

@@ -111,7 +111,6 @@ extern struct rcupreempt_trace *rcupreempt_trace_cpu(int cpu);
 struct softirq_action;
 
 #ifdef CONFIG_NO_HZ
-DECLARE_PER_CPU(struct rcu_dyntick_sched, rcu_dyntick_sched);

In -rt we deleted the first declaration. Is rcu_dyntick_sched defined on !CONFIG_NO_HZ?

-- Steve

Quoted 14 lines

include/linux/rcupreempt.h +1 -1

 static inline void rcu_enter_nohz(void)
 {

@@ -122,8 +121,8 @@ static inline void rcu_enter_nohz(void)
 
 static inline void rcu_exit_nohz(void)
 {
-	smp_mb(); /* CPUs seeing ++ must see later RCU read-side crit sects */
 	__get_cpu_var(rcu_dyntick_sched).dynticks++;
+	smp_mb(); /* CPUs seeing ++ must see later RCU read-side crit sects */
 	WARN_ON(!(__get_cpu_var(rcu_dyntick_sched).dynticks & 0x1));
 }

Paul E. McKenney

author

2008-08-20 22:34 UTC #3 lore raw

Quoted 28 lines · Steven Rostedt

Sorry for the late post, I'm finally looking at all my -rt mail.
Quoted 22 lines · Paul E. McKenney
Hello!
Compared tip/core/rcu to my latest patchset, and found the following issues:
o the memory barrier in rcu_exit_nohz() somehow got out of place
(it is correct in mainline as of 2.6.26-rc7).
o There is a duplicate declaration of rcu_dyntick_sched.
The attached patch fixes these.
Signed-off-by Paul E. McKenney <[email protected]>
---
include/linux/rcupreempt.h +0 -0
index f04b64e..84678bf 100644
@@ -111,7 +111,6 @@ extern struct rcupreempt_trace *rcupreempt_trace_cpu(int cpu);
 struct softirq_action;
 

Quoted 5 lines

include/linux/rcupreempt.h extern struct rcupreempt_trace *rcupreempt_trace_cpu(int cpu); · line 114
Quoted 2 lines
include/linux/rcupreempt.h +0 -1
 #ifdef CONFIG_NO_HZ
-DECLARE_PER_CPU(struct rcu_dyntick_sched, rcu_dyntick_sched);
In -rt we deleted the first declaration. Is rcu_dyntick_sched defined on !CONFIG_NO_HZ?

It builds for me in current tip/core/rcu without the above DECLARE_PER_CPU(). And the reason is the call_rcu_sched() function, which requires rcu_dyntick_sched.sched_qs and rcu_dyntick_sched.sched_qs_snap be defined even if !CONFIG_NO_HZ.

I suppose I could remove the other elements from the rcu_dyntick_sched structure if !CONFIG_NO_HZ. Worthwhile?

Thanx, Paul

Quoted 16 lines

-- Steve

Quoted 14 lines

include/linux/rcupreempt.h +1 -1

 static inline void rcu_enter_nohz(void)
 {

@@ -122,8 +121,8 @@ static inline void rcu_enter_nohz(void)
 
 static inline void rcu_exit_nohz(void)
 {
-	smp_mb(); /* CPUs seeing ++ must see later RCU read-side crit sects */
 	__get_cpu_var(rcu_dyntick_sched).dynticks++;
+	smp_mb(); /* CPUs seeing ++ must see later RCU read-side crit sects */
 	WARN_ON(!(__get_cpu_var(rcu_dyntick_sched).dynticks & 0x1));
 }

Steven Rostedt

reviewer

2008-08-21 01:02 UTC #4 lore raw

Quoted 5 lines · Paul E. McKenney

Quoted 4 lines
Quoted 1 line
-DECLARE_PER_CPU(struct rcu_dyntick_sched, rcu_dyntick_sched);
In -rt we deleted the first declaration. Is rcu_dyntick_sched defined on !CONFIG_NO_HZ?

Quoted 8 lines

It builds for me in current tip/core/rcu without the above DECLARE_PER_CPU(). And the reason is the call_rcu_sched() function, which requires rcu_dyntick_sched.sched_qs and rcu_dyntick_sched.sched_qs_snap be defined even if !CONFIG_NO_HZ.
I suppose I could remove the other elements from the rcu_dyntick_sched structure if !CONFIG_NO_HZ. Worthwhile?

Nah, forget about it. I've got a few different versions I'm dealing with.

Thanks,

-- Steve