[for-next][PATCH 1/4] ring-buffer: Replace this_cpu_*() with __this_cpu_*()

[for-next][PATCH 1/4] ring-buffer: Replace this_cpu_*() with __this_cpu_*()

[Steven Rostedt]

[-- Attachment #1: 0001-ring-buffer-Replace-this_cpu_-with-__this_cpu_.patch --]
[-- Type: text/plain, Size: 3003 bytes --]

From: Steven Rostedt <rostedt@goodmis.org>

It has come to my attention that this_cpu_read/write are horrible on
architectures other than x86. Worse yet, they actually disable
preemption or interrupts! This caused some unexpected tracing results
on ARM.

   101.356868: preempt_count_add <-ring_buffer_lock_reserve
   101.356870: preempt_count_sub <-ring_buffer_lock_reserve

The ring_buffer_lock_reserve has recursion protection that requires
accessing a per cpu variable. But since preempt_disable() is traced, it
too got traced while accessing the variable that is suppose to prevent
recursion like this.

The generic version of this_cpu_read() and write() are:

 #define this_cpu_generic_read(pcp)					\
 ({	typeof(pcp) ret__;						\
	preempt_disable();						\
	ret__ = *this_cpu_ptr(&(pcp));					\
	preempt_enable();						\
	ret__;								\
 })

 #define this_cpu_generic_to_op(pcp, val, op)				\
 do {									\
	unsigned long flags;						\
	raw_local_irq_save(flags);					\
	*__this_cpu_ptr(&(pcp)) op val;					\
	raw_local_irq_restore(flags);					\
 } while (0)

Which is unacceptable for locations that know they are within preempt
disabled or interrupt disabled locations.

Paul McKenney stated that __this_cpu_() versions produce much better code on
other architectures than this_cpu_() does, if we know that the call is done in
a preempt disabled location.

I also changed the recursive_unlock() to use two local variables instead
of accessing the per_cpu variable twice.

Link: http://lkml.kernel.org/r/20150317114411.GE3589@linux.vnet.ibm.com
Link: http://lkml.kernel.org/r/20150317104038.312e73d1@gandalf.local.home

Cc: stable@vger.kernel.org
Acked-by: Christoph Lameter <cl@linux.com>
Reported-by: Uwe Kleine-Koenig <u.kleine-koenig@pengutronix.de>
Tested-by: Uwe Kleine-Koenig <u.kleine-koenig@pengutronix.de>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
---
 kernel/trace/ring_buffer.c | 11 +++++------
 1 file changed, 5 insertions(+), 6 deletions(-)

diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c
index 5040d44fe5a3..922048a0f7ea 100644
--- a/kernel/trace/ring_buffer.c
+++ b/kernel/trace/ring_buffer.c
@@ -2679,7 +2679,7 @@ static DEFINE_PER_CPU(unsigned int, current_context);
 
 static __always_inline int trace_recursive_lock(void)
 {
-	unsigned int val = this_cpu_read(current_context);
+	unsigned int val = __this_cpu_read(current_context);
 	int bit;
 
 	if (in_interrupt()) {
@@ -2696,18 +2696,17 @@ static __always_inline int trace_recursive_lock(void)
 		return 1;
 
 	val |= (1 << bit);
-	this_cpu_write(current_context, val);
+	__this_cpu_write(current_context, val);
 
 	return 0;
 }
 
 static __always_inline void trace_recursive_unlock(void)
 {
-	unsigned int val = this_cpu_read(current_context);
+	unsigned int val = __this_cpu_read(current_context);
 
-	val--;
-	val &= this_cpu_read(current_context);
-	this_cpu_write(current_context, val);
+	val &= val & (val - 1);
+	__this_cpu_write(current_context, val);
 }
 
 #else
-- 
2.1.4

Re: [for-next][PATCH 1/4] ring-buffer: Replace this_cpu_*() with __this_cpu_*()

[Christoph Lameter]

On Wed, 25 Mar 2015, Steven Rostedt wrote:

> It has come to my attention that this_cpu_read/write are horrible on
> architectures other than x86. Worse yet, they actually disable
> preemption or interrupts! This caused some unexpected tracing results
> on ARM.

This isnt something new and I thought the comment was dropped from the
patch? This is a plain error in using this_cpu_* where __this_cpu_* would
have been sufficient. Code was uselessly disabling preemption twice.

> Which is unacceptable for locations that know they are within preempt
> disabled or interrupt disabled locations.

Well yes. Thats why the __this_cpu ops are there to avoid this
overhead.

> I also changed the recursive_unlock() to use two local variables instead
> of accessing the per_cpu variable twice.

Ok gotta look at that.

>  static __always_inline void trace_recursive_unlock(void)
>  {
> -	unsigned int val = this_cpu_read(current_context);
> +	unsigned int val = __this_cpu_read(current_context);
>
> -	val--;
> -	val &= this_cpu_read(current_context);
> -	this_cpu_write(current_context, val);
> +	val &= val & (val - 1);
> +	__this_cpu_write(current_context, val);
>  }

Ummm... This is does not look like an equivalent thing. Should this not
be:

	unsigned int val = __this_cpu_read(current_context);
	unsigned int newval = val - 1;

	newval &= val;
	__this_cpu_write(current_context, newval);

or more compact

	unsigned int val = __this_cpu_read(current_context);

	__this_cpu_write(current_context, val & (val - 1));

Re: [for-next][PATCH 1/4] ring-buffer: Replace this_cpu_*() with __this_cpu_*()

[Steven Rostedt]

On Fri, 27 Mar 2015 14:41:44 -0500 (CDT)
Christoph Lameter <cl@linux.com> wrote:

> On Wed, 25 Mar 2015, Steven Rostedt wrote:
> 
> > It has come to my attention that this_cpu_read/write are horrible on
> > architectures other than x86. Worse yet, they actually disable
> > preemption or interrupts! This caused some unexpected tracing results
> > on ARM.
> 
> This isnt something new and I thought the comment was dropped from the
> patch? This is a plain error in using this_cpu_* where __this_cpu_* would
> have been sufficient. Code was uselessly disabling preemption twice.
> 

Where in the patch do you see the comment? Or were you talking about
the change log? The original patch did have a comment, an it was
dropped, that's what I thought you were talking about.

> > Which is unacceptable for locations that know they are within preempt
> > disabled or interrupt disabled locations.
> 
> Well yes. Thats why the __this_cpu ops are there to avoid this
> overhead.
> 
> > I also changed the recursive_unlock() to use two local variables instead
> > of accessing the per_cpu variable twice.
> 
> Ok gotta look at that.
> 
> >  static __always_inline void trace_recursive_unlock(void)
> >  {
> > -	unsigned int val = this_cpu_read(current_context);
> > +	unsigned int val = __this_cpu_read(current_context);
> >
> > -	val--;
> > -	val &= this_cpu_read(current_context);
> > -	this_cpu_write(current_context, val);
> > +	val &= val & (val - 1);
> > +	__this_cpu_write(current_context, val);
> >  }
> 
> Ummm... This is does not look like an equivalent thing. Should this not
> be:
> 
> 	unsigned int val = __this_cpu_read(current_context);
> 	unsigned int newval = val - 1;
> 
> 	newval &= val;
> 	__this_cpu_write(current_context, newval);

Actually, it is equivalent, but I do see a issue with my patch.

	val &= val & (val - 1);

is the same as the more reasonable:

	val &= val - 1;

I think I meant to replace &= with = :-/

> 
> or more compact
> 
> 	unsigned int val = __this_cpu_read(current_context);
> 
> 	__this_cpu_write(current_context, val & (val - 1));

Maybe I'll just use your compact version.

Thanks,

-- Steve

[PATCH] ring-buffer: Remove duplicate use of '&' in recursive code

[Steven Rostedt]

Steven Rostedt (Red Hat) (1):
      ring-buffer: Remove duplicate use of '&' in recursive code

----
 kernel/trace/ring_buffer.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
---------------------------
commit 7eb867195b9f3990da60738b1f26d0a71f37f77f
Author: Steven Rostedt (Red Hat) <rostedt@goodmis.org>
Date:   Fri Mar 27 17:39:49 2015 -0400

    ring-buffer: Remove duplicate use of '&' in recursive code
    
    A clean up of the recursive protection code changed
    
      val = this_cpu_read(current_context);
      val--;
      val &= this_cpu_read(current_context);
    
    to
    
      val = this_cpu_read(current_context);
      val &= val & (val - 1);
    
    Which has a duplicate use of '&' as the above is the same as
    
      val = val & (val - 1);
    
    Actually, it would be best to remove that line altogether and
    just add it to where it is used.
    
    Link: http://lkml.kernel.org/alpine.DEB.2.11.1503271423580.23114@gentwo.org
    
    Suggested-by: Christoph Lameter <cl@linux.com>
    Signed-off-by: Steven Rostedt <rostedt@goodmis.org>

diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c
index 922048a0f7ea..93caf56567cb 100644
--- a/kernel/trace/ring_buffer.c
+++ b/kernel/trace/ring_buffer.c
@@ -2705,8 +2705,7 @@ static __always_inline void trace_recursive_unlock(void)
 {
 	unsigned int val = __this_cpu_read(current_context);
 
-	val &= val & (val - 1);
-	__this_cpu_write(current_context, val);
+	__this_cpu_write(current_context, val & (val - 1));
 }
 
 #else

Re: [for-next][PATCH 1/4] ring-buffer: Replace this_cpu_*() with __this_cpu_*()

[Christoph Lameter]

On Fri, 27 Mar 2015, Steven Rostedt wrote:

> Where in the patch do you see the comment? Or were you talking about
> the change log? The original patch did have a comment, an it was
> dropped, that's what I thought you were talking about.

Sorry yes the changelog.

> Actually, it is equivalent, but I do see a issue with my patch.
>
> 	val &= val & (val - 1);
>
> is the same as the more reasonable:
>
> 	val &= val - 1;
>
> I think I meant to replace &= with = :-/
>
> >
> > or more compact
> >
> > 	unsigned int val = __this_cpu_read(current_context);
> >
> > 	__this_cpu_write(current_context, val & (val - 1));
>
> Maybe I'll just use your compact version.

Hmmm... It could even be more compact

__this_cpu_and(current_context, __this_cpu_read(current_context) - 1);

Re: [for-next][PATCH 1/4] ring-buffer: Replace this_cpu_*() with __this_cpu_*()

[Steven Rostedt]

On Mon, 30 Mar 2015 07:44:30 -0500 (CDT)
Christoph Lameter <cl@linux.com> wrote:


> > >
> > > or more compact
> > >
> > > 	unsigned int val = __this_cpu_read(current_context);
> > >
> > > 	__this_cpu_write(current_context, val & (val - 1));
> >
> > Maybe I'll just use your compact version.
> 
> Hmmm... It could even be more compact
> 
> __this_cpu_and(current_context, __this_cpu_read(current_context) - 1);

Hmm,  I didn't realize there was an "and" version. I'm guessing this
would bring down the instruction count even more?

/me tries it.

I just finished testing my previous version. If this does prove to be
more compact, I'll have to replace that one with this one.

-- Steve

Re: [for-next][PATCH 1/4] ring-buffer: Replace this_cpu_*() with __this_cpu_*()

[Christoph Lameter]

On Mon, 30 Mar 2015, Steven Rostedt wrote:

> Hmm,  I didn't realize there was an "and" version. I'm guessing this
> would bring down the instruction count even more?

Yes two segment prefixed instructions and a decrement.