[PATCH 1/2] Correct function definition for C++

[PATCH 1/2] Correct function definition for C++

[Joakim Tjernlund]

C++ does does not like the extra extern before asmlinkage, remove it.

Signed-off-by: Joakim Tjernlund <joakim.tjernlund@infinera.com>
---
 include/linux/printk.h | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/include/linux/printk.h b/include/linux/printk.h
index 3472cc6..be823f5 100644
--- a/include/linux/printk.h
+++ b/include/linux/printk.h
@@ -139,7 +139,7 @@ struct va_format {
 })
 
 #ifdef CONFIG_EARLY_PRINTK
-extern asmlinkage __printf(1, 2)
+asmlinkage  __printf(1, 2)
 void early_printk(const char *fmt, ...);
 #else
 static inline __printf(1, 2) __cold
@@ -270,7 +270,7 @@ static inline void show_regs_print_info(const char *log_lvl)
 }
 #endif
 
-extern asmlinkage void dump_stack(void) __cold;
+asmlinkage void dump_stack(void) __cold;
 
 #ifndef pr_fmt
 #define pr_fmt(fmt) fmt
-- 
2.10.2

[PATCH 2/2] compiler.h: fix C++ uninitialized const issue

[Joakim Tjernlund]

C++ does not like the union { typeof(x) __val; char __c[1]; } __u
construct for const types:
 error: uninitialized const member in
 'union atomic_read(const atomic_t*)::<anonymous>'
Address this by creating a C++ version of READ_ONCE where this union is
initialized: union { void _u(){}; typeof(x) __val; char __c[1]; } __u={0}

To please gcc 6.3.0 also add in a _u(){} as default ctor.

This makes C++ happy enough to build.

Signed-off-by: Joakim Tjernlund <joakim.tjernlund@infinera.com>
---
 include/linux/compiler.h | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/include/linux/compiler.h b/include/linux/compiler.h
index cf0fa5d..0a047fd 100644
--- a/include/linux/compiler.h
+++ b/include/linux/compiler.h
@@ -300,6 +300,7 @@ static __always_inline void __write_once_size(volatile void *p, void *res, int s
  * required ordering.
  */
 
+#ifndef __cplusplus
 #define __READ_ONCE(x, check)						\
 ({									\
 	union { typeof(x) __val; char __c[1]; } __u;			\
@@ -309,6 +310,17 @@ static __always_inline void __write_once_size(volatile void *p, void *res, int s
 		__read_once_size_nocheck(&(x), __u.__c, sizeof(x));	\
 	__u.__val;							\
 })
+#else
+#define __READ_ONCE(x, check)						\
+({									\
+	union { void _u(){}; typeof(x) __val; char __c[1]; } __u={0};	\
+	if (check)							\
+		__read_once_size(&(x), __u.__c, sizeof(x));		\
+	else								\
+		__read_once_size_nocheck(&(x), __u.__c, sizeof(x));	\
+	__u.__val;							\
+})
+#endif
 #define READ_ONCE(x) __READ_ONCE(x, 1)
 
 /*
-- 
2.10.2

Re: [PATCH 1/2] Correct function definition for C++

[Greg KH]

On Tue, Feb 21, 2017 at 04:24:04PM +0100, Joakim Tjernlund wrote:
> C++ does does not like the extra extern before asmlinkage, remove it.
> 
> Signed-off-by: Joakim Tjernlund <joakim.tjernlund@infinera.com>
> ---
>  include/linux/printk.h | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/include/linux/printk.h b/include/linux/printk.h
> index 3472cc6..be823f5 100644
> --- a/include/linux/printk.h
> +++ b/include/linux/printk.h


Why are you building this file with a C++ compiler?

Also, this is not the correct way to submit patches for inclusion in the
stable kernel tree.  Please read Documentation/stable_kernel_rules.txt
for how to do this properly.

thanks,

greg k-h

Re: [PATCH 1/2] Correct function definition for C++

[Joakim Tjernlund]

On Wed, 2017-02-22 at 08:10 +0100, Greg KH wrote:
> On Tue, Feb 21, 2017 at 04:24:04PM +0100, Joakim Tjernlund wrote:
> > C++ does does not like the extra extern before asmlinkage, remove it.
> > 
> > Signed-off-by: Joakim Tjernlund <joakim.tjernlund@infinera.com>
> > ---
> >  include/linux/printk.h | 4 ++--
> >  1 file changed, 2 insertions(+), 2 deletions(-)
> > 
> > diff --git a/include/linux/printk.h b/include/linux/printk.h
> > index 3472cc6..be823f5 100644
> > --- a/include/linux/printk.h
> > +++ b/include/linux/printk.h
> 
> 
> Why are you building this file with a C++ compiler?

virtualbox uses C++ and includes various kernel headers and the build
fails, virtualbox guest additions has not build for quite some time now and
this is one of the problems.

> 
> Also, this is not the correct way to submit patches for inclusion in the
> stable kernel tree.  Please read Documentation/stable_kernel_rules.txt
> for how to do this properly.

Sorry, I was in a hurry(lazy:) so I just included stable directly as
I know it needs to be fixed in 4.9 as well.

 Jocke

Re: [PATCH 1/2] Correct function definition for C++

[greg]

On Wed, Feb 22, 2017 at 07:50:46AM +0000, Joakim Tjernlund wrote:
> On Wed, 2017-02-22 at 08:10 +0100, Greg KH wrote:
> > On Tue, Feb 21, 2017 at 04:24:04PM +0100, Joakim Tjernlund wrote:
> > > C++ does does not like the extra extern before asmlinkage, remove it.
> > > 
> > > Signed-off-by: Joakim Tjernlund <joakim.tjernlund@infinera.com>
> > > ---
> > >  include/linux/printk.h | 4 ++--
> > >  1 file changed, 2 insertions(+), 2 deletions(-)
> > > 
> > > diff --git a/include/linux/printk.h b/include/linux/printk.h
> > > index 3472cc6..be823f5 100644
> > > --- a/include/linux/printk.h
> > > +++ b/include/linux/printk.h
> > 
> > 
> > Why are you building this file with a C++ compiler?
> 
> virtualbox uses C++ and includes various kernel headers and the build
> fails, virtualbox guest additions has not build for quite some time now and
> this is one of the problems.

Virtualbox is a horrid pile of crap.  You can quote me on that.  We
don't care about out-of-tree drivers, the authors should work to get
them merged properly if they do care.

Kernel code should be C, not C++, and if you do want to use C++, then
you are on your own, sorry.

> > Also, this is not the correct way to submit patches for inclusion in the
> > stable kernel tree.  Please read Documentation/stable_kernel_rules.txt
> > for how to do this properly.
> 
> Sorry, I was in a hurry(lazy:) so I just included stable directly as
> I know it needs to be fixed in 4.9 as well.

But that's not how to get a patch into the stable tree :(

sorry,

greg k-h

Re: [PATCH 1/2] Correct function definition for C++

[Joakim Tjernlund]

On Wed, 2017-02-22 at 14:03 +0100, greg@kroah.com wrote:
> On Wed, Feb 22, 2017 at 07:50:46AM +0000, Joakim Tjernlund wrote:
> > On Wed, 2017-02-22 at 08:10 +0100, Greg KH wrote:
> > > On Tue, Feb 21, 2017 at 04:24:04PM +0100, Joakim Tjernlund wrote:
> > > > C++ does does not like the extra extern before asmlinkage, remove it.
> > > > 
> > > > Signed-off-by: Joakim Tjernlund <joakim.tjernlund@infinera.com>
> > > > ---
> > > >  include/linux/printk.h | 4 ++--
> > > >  1 file changed, 2 insertions(+), 2 deletions(-)
> > > > 
> > > > diff --git a/include/linux/printk.h b/include/linux/printk.h
> > > > index 3472cc6..be823f5 100644
> > > > --- a/include/linux/printk.h
> > > > +++ b/include/linux/printk.h
> > > 
> > > 
> > > Why are you building this file with a C++ compiler?
> > 
> > virtualbox uses C++ and includes various kernel headers and the build
> > fails, virtualbox guest additions has not build for quite some time now and
> > this is one of the problems.
> 
> Virtualbox is a horrid pile of crap.  You can quote me on that.  We
> don't care about out-of-tree drivers, the authors should work to get
> them merged properly if they do care.

Sure, I believe you :)
 
But in this case it is not the kernel modules that fails, it is their
guest additions/xf86 video driver. Seem like that these need to include some
kernel herders there too.

> 
> Kernel code should be C, not C++, and if you do want to use C++, then
> you are on your own, sorry.

But there are already a lot of C++ adjustments!? Just grepping for __cplusplus shows
plenty of hits and what is the point of asmlinkage then:
#ifdef __cplusplus
#define CPP_ASMLINKAGE extern "C"
#else
#define CPP_ASMLINKAGE
#endif

#ifndef asmlinkage
#define asmlinkage CPP_ASMLINKAGE
#endif

I read all of this that C++ code should be able to at least include kernel headers
without fatal errors. 
You don't agree? In that case all __cplusplus/asmlinkage should be removed?

> 
> > > Also, this is not the correct way to submit patches for inclusion in the
> > > stable kernel tree.  Please read Documentation/stable_kernel_rules.txt
> > > for how to do this properly.
> > 
> > Sorry, I was in a hurry(lazy:) so I just included stable directly as
> > I know it needs to be fixed in 4.9 as well.
> 
> But that's not how to get a patch into the stable tree :(
> 
> sorry,

Right, my mistake. No need to be sorry, I got what I deserved :)

 Jocke

> 
> greg k-h

Re: [PATCH 1/2] Correct function definition for C++

[greg]

On Wed, Feb 22, 2017 at 02:00:53PM +0000, Joakim Tjernlund wrote:
> On Wed, 2017-02-22 at 14:03 +0100, greg@kroah.com wrote:
> > On Wed, Feb 22, 2017 at 07:50:46AM +0000, Joakim Tjernlund wrote:
> > > On Wed, 2017-02-22 at 08:10 +0100, Greg KH wrote:
> > > > On Tue, Feb 21, 2017 at 04:24:04PM +0100, Joakim Tjernlund wrote:
> > > > > C++ does does not like the extra extern before asmlinkage, remove it.
> > > > > 
> > > > > Signed-off-by: Joakim Tjernlund <joakim.tjernlund@infinera.com>
> > > > > ---
> > > > >  include/linux/printk.h | 4 ++--
> > > > >  1 file changed, 2 insertions(+), 2 deletions(-)
> > > > > 
> > > > > diff --git a/include/linux/printk.h b/include/linux/printk.h
> > > > > index 3472cc6..be823f5 100644
> > > > > --- a/include/linux/printk.h
> > > > > +++ b/include/linux/printk.h
> > > > 
> > > > 
> > > > Why are you building this file with a C++ compiler?
> > > 
> > > virtualbox uses C++ and includes various kernel headers and the build
> > > fails, virtualbox guest additions has not build for quite some time now and
> > > this is one of the problems.
> > 
> > Virtualbox is a horrid pile of crap.  You can quote me on that.  We
> > don't care about out-of-tree drivers, the authors should work to get
> > them merged properly if they do care.
> 
> Sure, I believe you :)
>  
> But in this case it is not the kernel modules that fails, it is their
> guest additions/xf86 video driver. Seem like that these need to include some
> kernel herders there too.

Seems like their build system is broken, it should not be including non
uapi header files at all.

> > Kernel code should be C, not C++, and if you do want to use C++, then
> > you are on your own, sorry.
> 
> But there are already a lot of C++ adjustments!? Just grepping for __cplusplus shows
> plenty of hits and what is the point of asmlinkage then:
> #ifdef __cplusplus
> #define CPP_ASMLINKAGE extern "C"
> #else
> #define CPP_ASMLINKAGE
> #endif
> 
> #ifndef asmlinkage
> #define asmlinkage CPP_ASMLINKAGE
> #endif

The usages in include/uapi/ make a bit of sense, but not much.
Everything outside of there makes no sense at all (really just some drm
and one scsi file.)  They should be removed.

> I read all of this that C++ code should be able to at least include kernel headers
> without fatal errors. 

For uapi stuff, yes.  But that's not what you are doing here.  You are
trying to build kernel code with a c++ compiler.  That's not going to
work.

How is a C++ compiler supposed to parse the following line in device.h
	struct class;
(one of my most proud moments...)

thanks,

greg k-h

Re: [PATCH 1/2] Correct function definition for C++

[greg]

On Wed, Feb 22, 2017 at 02:00:53PM +0000, Joakim Tjernlund wrote:
> On Wed, 2017-02-22 at 14:03 +0100, greg@kroah.com wrote:
> > On Wed, Feb 22, 2017 at 07:50:46AM +0000, Joakim Tjernlund wrote:
> > > On Wed, 2017-02-22 at 08:10 +0100, Greg KH wrote:
> > > > On Tue, Feb 21, 2017 at 04:24:04PM +0100, Joakim Tjernlund wrote:
> > > > > C++ does does not like the extra extern before asmlinkage, remove it.
> > > > > 
> > > > > Signed-off-by: Joakim Tjernlund <joakim.tjernlund@infinera.com>
> > > > > ---
> > > > >  include/linux/printk.h | 4 ++--
> > > > >  1 file changed, 2 insertions(+), 2 deletions(-)
> > > > > 
> > > > > diff --git a/include/linux/printk.h b/include/linux/printk.h
> > > > > index 3472cc6..be823f5 100644
> > > > > --- a/include/linux/printk.h
> > > > > +++ b/include/linux/printk.h
> > > > 
> > > > 
> > > > Why are you building this file with a C++ compiler?
> > > 
> > > virtualbox uses C++ and includes various kernel headers and the build
> > > fails, virtualbox guest additions has not build for quite some time now and
> > > this is one of the problems.
> > 
> > Virtualbox is a horrid pile of crap.  You can quote me on that.  We
> > don't care about out-of-tree drivers, the authors should work to get
> > them merged properly if they do care.
> 
> Sure, I believe you :)
>  
> But in this case it is not the kernel modules that fails, it is their
> guest additions/xf86 video driver. Seem like that these need to include some
> kernel herders there too.

Seems like their build system is broken, it should not be including non
uapi header files at all.

> > Kernel code should be C, not C++, and if you do want to use C++, then
> > you are on your own, sorry.
> 
> But there are already a lot of C++ adjustments!? Just grepping for __cplusplus shows
> plenty of hits and what is the point of asmlinkage then:
> #ifdef __cplusplus
> #define CPP_ASMLINKAGE extern "C"
> #else
> #define CPP_ASMLINKAGE
> #endif
> 
> #ifndef asmlinkage
> #define asmlinkage CPP_ASMLINKAGE
> #endif

The usages in include/uapi/ make a bit of sense, but not much.
Everything outside of there makes no sense at all (really just some drm
and one scsi file.)  They should be removed.

> I read all of this that C++ code should be able to at least include kernel headers
> without fatal errors.�

For uapi stuff, yes.  But that's not what you are doing here.  You are
trying to build kernel code with a c++ compiler.  That's not going to
work.

How is a C++ compiler supposed to parse the following line in device.h
	struct class;
(one of my most proud moments...)

thanks,

greg k-h

Re: [PATCH 1/2] Correct function definition for C++

[Joakim Tjernlund]

On Wed, 2017-02-22 at 15:22 +0100, greg@kroah.com wrote:
> On Wed, Feb 22, 2017 at 02:00:53PM +0000, Joakim Tjernlund wrote:
> > On Wed, 2017-02-22 at 14:03 +0100, greg@kroah.com wrote:
> > > On Wed, Feb 22, 2017 at 07:50:46AM +0000, Joakim Tjernlund wrote:
> > > > On Wed, 2017-02-22 at 08:10 +0100, Greg KH wrote:
> > > > > On Tue, Feb 21, 2017 at 04:24:04PM +0100, Joakim Tjernlund wrote:
> > > > > > C++ does does not like the extra extern before asmlinkage, remove it.
> > > > > > 
> > > > > > Signed-off-by: Joakim Tjernlund <joakim.tjernlund@infinera.com>
> > > > > > ---
> > > > > >  include/linux/printk.h | 4 ++--
> > > > > >  1 file changed, 2 insertions(+), 2 deletions(-)
> > > > > > 
> > > > > > diff --git a/include/linux/printk.h b/include/linux/printk.h
> > > > > > index 3472cc6..be823f5 100644
> > > > > > --- a/include/linux/printk.h
> > > > > > +++ b/include/linux/printk.h
> > > > > 
> > > > > 
> > > > > Why are you building this file with a C++ compiler?
> > > > 
> > > > virtualbox uses C++ and includes various kernel headers and the build
> > > > fails, virtualbox guest additions has not build for quite some time now and
> > > > this is one of the problems.
> > > 
> > > Virtualbox is a horrid pile of crap.  You can quote me on that.  We
> > > don't care about out-of-tree drivers, the authors should work to get
> > > them merged properly if they do care.
> > 
> > Sure, I believe you :)
> >  
> > But in this case it is not the kernel modules that fails, it is their
> > guest additions/xf86 video driver. Seem like that these need to include some
> > kernel herders there too.
> 
> Seems like their build system is broken, it should not be including non
> uapi header files at all.
> 
> > > Kernel code should be C, not C++, and if you do want to use C++, then
> > > you are on your own, sorry.
> > 
> > But there are already a lot of C++ adjustments!? Just grepping for __cplusplus shows
> > plenty of hits and what is the point of asmlinkage then:
> > #ifdef __cplusplus
> > #define CPP_ASMLINKAGE extern "C"
> > #else
> > #define CPP_ASMLINKAGE
> > #endif
> > 
> > #ifndef asmlinkage
> > #define asmlinkage CPP_ASMLINKAGE
> > #endif
> 
> The usages in include/uapi/ make a bit of sense, but not much.
> Everything outside of there makes no sense at all (really just some drm
> and one scsi file.)  They should be removed.
> 
> > I read all of this that C++ code should be able to at least include kernel headers
> > without fatal errors. 
> 
> For uapi stuff, yes.  But that's not what you are doing here.  You are
> trying to build kernel code with a c++ compiler.  That's not going to
> work.

Right, I looked into guest-addition and it builds a few kernel modules too
and include linux/string.h and that pulls in the offending code.

Anyhow, the code in printk.h is "wrong" to add an extra extern so that
change should not be a big deal as is, C++ or not.

> 
> How is a C++ compiler supposed to parse the following line in device.h
> 	struct class;
> (one of my most proud moments...)

 :)

Re: [PATCH 1/2] Correct function definition for C++

[Thomas Backlund]

Den 22.02.2017 kl. 09:50, skrev Joakim Tjernlund:
> On Wed, 2017-02-22 at 08:10 +0100, Greg KH wrote:
>> On Tue, Feb 21, 2017 at 04:24:04PM +0100, Joakim Tjernlund wrote:
>>> C++ does does not like the extra extern before asmlinkage, remove it.
>>>
>>> Signed-off-by: Joakim Tjernlund <joakim.tjernlund@infinera.com>
>>> ---
>>>  include/linux/printk.h | 4 ++--
>>>  1 file changed, 2 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/include/linux/printk.h b/include/linux/printk.h
>>> index 3472cc6..be823f5 100644
>>> --- a/include/linux/printk.h
>>> +++ b/include/linux/printk.h
>>
>>
>> Why are you building this file with a C++ compiler?
>
> virtualbox uses C++ and includes various kernel headers and the build
> fails, virtualbox guest additions has not build for quite some time now and
> this is one of the problems.


You need this fix for virtualbox:

https://www.virtualbox.org/changeset/65874/vbox/trunk/configure

--

Thomas