Re: __update_max_tr: rcu_read_lock() used illegally while idle!

[Paul E. McKenney]

On Tue, Jul 31, 2012 at 01:24:57PM -0400, Steven Rostedt wrote:
> On Tue, 2012-07-31 at 10:17 -0700, Paul E. McKenney wrote:
> 
> > rcu: Permit RCU_NONIDLE() to be used from interrupt context
> > 
> > There is a need to use RCU from interrupt context, but either before
> > rcu_irq_enter() is called or after rcu_irq_exit() is called.  If the
> > interrupt occurs from idle, then lockdep-RCU will complain about such
> > uses, as they appear to be illegal uses of RCU from the idle loop.
> > In other environments, RCU_NONIDLE() could be used to properly protect
> > the use of RCU, but RCU_NONIDLE() currently cannot be invoked except
> > from process context.
> > 
> > This commit therefore modifies RCU_NONIDLE() to permit its use more
> > globally.
> > 
> > Signed-off-by: Paul E. McKenney <paul.mckenney@xxxxxxxxxx>
> > Signed-off-by: Paul E. McKenney <paulmck@xxxxxxxxxxxxxxxxxx>
> > 
> > diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h
> > index 115ead2..4eab81a 100644
> > --- a/include/linux/rcupdate.h
> > +++ b/include/linux/rcupdate.h
> > @@ -210,14 +210,12 @@ extern void exit_rcu(void);
> >   * to nest RCU_NONIDLE() wrappers, but the nesting level is currently
> >   * quite limited.  If deeper nesting is required, it will be necessary
> >   * to adjust DYNTICK_TASK_NESTING_VALUE accordingly.
> > - *
> > - * This macro may be used from process-level code only.
> >   */
> >  #define RCU_NONIDLE(a) \
> >  	do { \
> > -		rcu_idle_exit(); \
> > +		rcu_irq_exit(); \
> 
> Don't you want that to be rcu_irq_enter? And is this safe to happen
> everywhere in the kernel?

<facepalm>

You are quite correct, please see below.

>From RCU's viewpoint, this just makes it look like another level of
interrupt happened.  So the main issue would be if you call this from
NMI handlers.  Especially if you call this from NMI handlers in places
where in_nmi() returns false...

So do you need to invoke this from NMI handlers, and if so, do you
need to do so in places where in_nmi() returns false?

> The code that is calling rcu is called by the irqsoff tracer. Which can
> happen at all points where interrupts are enabled.

OK, I interpret this as excluding NMI handlers, but please let me
know if I am still being naive.  ;-)

							Thanx, Paul

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

rcu: Permit RCU_NONIDLE() to be used from interrupt context

There is a need to use RCU from interrupt context, but either before
rcu_irq_enter() is called or after rcu_irq_exit() is called.  If the
interrupt occurs from idle, then lockdep-RCU will complain about such
uses, as they appear to be illegal uses of RCU from the idle loop.
In other environments, RCU_NONIDLE() could be used to properly protect
the use of RCU, but RCU_NONIDLE() currently cannot be invoked except
from process context.

This commit therefore modifies RCU_NONIDLE() to permit its use more
globally.

Signed-off-by: Paul E. McKenney <paul.mckenney@xxxxxxxxxx>
Signed-off-by: Paul E. McKenney <paulmck@xxxxxxxxxxxxxxxxxx>

diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h
index 115ead2..0fbbd52 100644
--- a/include/linux/rcupdate.h
+++ b/include/linux/rcupdate.h
@@ -210,14 +210,12 @@ extern void exit_rcu(void);
  * to nest RCU_NONIDLE() wrappers, but the nesting level is currently
  * quite limited.  If deeper nesting is required, it will be necessary
  * to adjust DYNTICK_TASK_NESTING_VALUE accordingly.
- *
- * This macro may be used from process-level code only.
  */
 #define RCU_NONIDLE(a) \
 	do { \
-		rcu_idle_exit(); \
+		rcu_irq_enter(); \
 		do { a; } while (0); \
-		rcu_idle_enter(); \
+		rcu_irq_exit(); \
 	} while (0)
 
 /*
diff --git a/kernel/rcutiny.c b/kernel/rcutiny.c
index 31a10f9..aee6df8 100644
--- a/kernel/rcutiny.c
+++ b/kernel/rcutiny.c
@@ -115,6 +115,7 @@ void rcu_irq_exit(void)
 	rcu_idle_enter_common(newval);
 	local_irq_restore(flags);
 }
+EXPORT_SYMBOL_GPL(rcu_irq_exit);
 
 /* Common code for rcu_idle_exit() and rcu_irq_enter(), see kernel/rcutree.c. */
 static void rcu_idle_exit_common(long long oldval)
@@ -172,6 +173,7 @@ void rcu_irq_enter(void)
 	rcu_idle_exit_common(oldval);
 	local_irq_restore(flags);
 }
+EXPORT_SYMBOL_GPL(rcu_irq_enter);
 
 #ifdef CONFIG_DEBUG_LOCK_ALLOC
 
diff --git a/kernel/rcutree.c b/kernel/rcutree.c
index 44609c3..2175b13 100644
--- a/kernel/rcutree.c
+++ b/kernel/rcutree.c
@@ -455,6 +455,7 @@ void rcu_irq_exit(void)
 		rcu_idle_enter_common(rdtp, oldval);
 	local_irq_restore(flags);
 }
+EXPORT_SYMBOL_GPL(rcu_irq_exit);
 
 /*
  * rcu_idle_exit_common - inform RCU that current CPU is moving away from idle
@@ -550,6 +551,7 @@ void rcu_irq_enter(void)
 		rcu_idle_exit_common(rdtp, oldval);
 	local_irq_restore(flags);
 }
+EXPORT_SYMBOL_GPL(rcu_irq_enter);
 
 /**
  * rcu_nmi_enter - inform RCU of entry to NMI context

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/