[PATCH] trace/ring_buffer: fix section mismatch warning

[PATCH] trace/ring_buffer: fix section mismatch warning

[Heiko Carstens]

From: Heiko Carstens <heiko.carstens@de.ibm.com>

With linux-next I get this:

Section mismatch in reference from the function ring_buffer_alloc() to the function .cpuinit.text:rb_cpu_notify()
The function ring_buffer_alloc() references
the function __cpuinit rb_cpu_notify().
This is often because ring_buffer_alloc lacks a __cpuinit 
annotation or the annotation of rb_cpu_notify is wrong.

This is a false positive since rb_cpu_notify gets only referenced if
CONFIG_HOTPLUG_CPU=y.
To get rid of the warning annotate ring_buffer_alloc with __ref.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
---
 kernel/trace/ring_buffer.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Index: linux-next/kernel/trace/ring_buffer.c
===================================================================
--- linux-next.orig/kernel/trace/ring_buffer.c
+++ linux-next/kernel/trace/ring_buffer.c
@@ -549,7 +549,7 @@ static int __cpuinit rb_cpu_notify(struc
  * when the buffer wraps. If this flag is not set, the buffer will
  * drop data when the tail hits the head.
  */
-struct ring_buffer *ring_buffer_alloc(unsigned long size, unsigned flags)
+struct ring_buffer * __ref ring_buffer_alloc(unsigned long size, unsigned flags)
 {
 	struct ring_buffer *buffer;
 	int bsize;

Re: [PATCH] trace/ring_buffer: fix section mismatch warning

[Ingo Molnar]

* Heiko Carstens <heiko.carstens@de.ibm.com> wrote:

> From: Heiko Carstens <heiko.carstens@de.ibm.com>
> 
> With linux-next I get this:
> 
> Section mismatch in reference from the function 
> ring_buffer_alloc() to the function .cpuinit.text:rb_cpu_notify() 
> The function ring_buffer_alloc() references the function __cpuinit 
> rb_cpu_notify(). This is often because ring_buffer_alloc lacks a 
> __cpuinit annotation or the annotation of rb_cpu_notify is wrong.
> 
> This is a false positive since rb_cpu_notify gets only referenced 
> if CONFIG_HOTPLUG_CPU=y. To get rid of the warning annotate 
> ring_buffer_alloc with __ref.
> 
> Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
> ---
>  kernel/trace/ring_buffer.c |    2 +-

Frederic has beaten you to it narrowly:

09c9e84: tracing/ring-buffer: don't annotate rb_cpu_notify with __cpuinit

:)

	Ingo

Re: [PATCH] trace/ring_buffer: fix section mismatch warning

[Heiko Carstens]

On Sat, 21 Mar 2009 15:39:11 +0100
Ingo Molnar <mingo@elte.hu> wrote:
> > Section mismatch in reference from the function 
> > ring_buffer_alloc() to the function .cpuinit.text:rb_cpu_notify() 
> > The function ring_buffer_alloc() references the function __cpuinit 
> > rb_cpu_notify(). This is often because ring_buffer_alloc lacks a 
> > __cpuinit annotation or the annotation of rb_cpu_notify is wrong.
> > 
> > This is a false positive since rb_cpu_notify gets only referenced 
> > if CONFIG_HOTPLUG_CPU=y. To get rid of the warning annotate 
> > ring_buffer_alloc with __ref.
> > 
> > Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
> > ---
> >  kernel/trace/ring_buffer.c |    2 +-
> 
> Frederic has beaten you to it narrowly:
> 
> 09c9e84: tracing/ring-buffer: don't annotate rb_cpu_notify with __cpuinit

Btw. that was the only section mismatch warning I got on linux-next
with an allyesconfig on s390. We probably could enable the verbose
section mismatch warnings at compile time?

Sam?

Re: [PATCH] trace/ring_buffer: fix section mismatch warning

[Ingo Molnar]

* Heiko Carstens <heiko.carstens@de.ibm.com> wrote:

> On Sat, 21 Mar 2009 15:39:11 +0100
> Ingo Molnar <mingo@elte.hu> wrote:
> > > Section mismatch in reference from the function 
> > > ring_buffer_alloc() to the function .cpuinit.text:rb_cpu_notify() 
> > > The function ring_buffer_alloc() references the function __cpuinit 
> > > rb_cpu_notify(). This is often because ring_buffer_alloc lacks a 
> > > __cpuinit annotation or the annotation of rb_cpu_notify is wrong.
> > > 
> > > This is a false positive since rb_cpu_notify gets only referenced 
> > > if CONFIG_HOTPLUG_CPU=y. To get rid of the warning annotate 
> > > ring_buffer_alloc with __ref.
> > > 
> > > Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
> > > ---
> > >  kernel/trace/ring_buffer.c |    2 +-
> > 
> > Frederic has beaten you to it narrowly:
> > 
> > 09c9e84: tracing/ring-buffer: don't annotate rb_cpu_notify with __cpuinit
> 
> Btw. that was the only section mismatch warning I got on linux-next
> with an allyesconfig on s390. We probably could enable the verbose
> section mismatch warnings at compile time?
> 
> Sam?

There's still too many warnings in general on other platforms - but 
as long as it's per arch we could in fact propagate the warnings to 
build failures.

	Ingo

Re: [PATCH] trace/ring_buffer: fix section mismatch warning

[Sam Ravnborg]

On Sat, Mar 21, 2009 at 04:01:02PM +0100, Heiko Carstens wrote:
> On Sat, 21 Mar 2009 15:39:11 +0100
> Ingo Molnar <mingo@elte.hu> wrote:
> > > Section mismatch in reference from the function 
> > > ring_buffer_alloc() to the function .cpuinit.text:rb_cpu_notify() 
> > > The function ring_buffer_alloc() references the function __cpuinit 
> > > rb_cpu_notify(). This is often because ring_buffer_alloc lacks a 
> > > __cpuinit annotation or the annotation of rb_cpu_notify is wrong.
> > > 
> > > This is a false positive since rb_cpu_notify gets only referenced 
> > > if CONFIG_HOTPLUG_CPU=y. To get rid of the warning annotate 
> > > ring_buffer_alloc with __ref.
> > > 
> > > Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
> > > ---
> > >  kernel/trace/ring_buffer.c |    2 +-
> > 
> > Frederic has beaten you to it narrowly:
> > 
> > 09c9e84: tracing/ring-buffer: don't annotate rb_cpu_notify with __cpuinit
> 
> Btw. that was the only section mismatch warning I got on linux-next
> with an allyesconfig on s390. We probably could enable the verbose
> section mismatch warnings at compile time?
> 
> Sam?

When I get a bit more time (and a properly working setup again)
I will enable the verbose version.
But until I can dedicate enough time to support it I am not the
one to make it more verbose.

That may be in the timeframe of 2.6.30 but I'm not sure.

	Sam