[1/1] arch/powerpc: Rework local_paca to avoid LTO warnings

From: Alastair D'Silva <alastair@d-silva.org>

When building an LTO kernel, the existing code generates warnings:
    ./arch/powerpc/include/asm/paca.h:37:30: warning: register of
        ‘local_paca’ used for multiple global register variables
     register struct paca_struct *local_paca asm("r13");
                              ^
    ./arch/powerpc/include/asm/paca.h:37:30: note: conflicts with
        ‘local_paca’

This patch reworks local_paca into an inline getter & setter function,
which addresses the warning.

Generated ASM from this patch is broadly similar (addresses have
changed and the compiler uses different GPRs in some places).

Signed-off-by: Alastair D'Silva <alastair@d-silva.org>
---
 arch/powerpc/include/asm/paca.h | 44 +++++++++++++++++++++++----------
 arch/powerpc/kernel/paca.c      |  2 +-
 2 files changed, 32 insertions(+), 14 deletions(-)

Hello,

Le 13/03/2019 à 04:42, Alastair D'Silva a écrit :
> From: Alastair D'Silva <alastair@d-silva.org>
> 
> When building an LTO kernel, the existing code generates warnings:
>      ./arch/powerpc/include/asm/paca.h:37:30: warning: register of
>          ‘local_paca’ used for multiple global register variables
>       register struct paca_struct *local_paca asm("r13");
>                                ^
>      ./arch/powerpc/include/asm/paca.h:37:30: note: conflicts with
>          ‘local_paca’

How do you build a LTO kernel ?

> 
> This patch reworks local_paca into an inline getter & setter function,
> which addresses the warning.

This patch adds sparse warnings, see 
https://patchwork.ozlabs.org/patch/1055875/

> 
> Generated ASM from this patch is broadly similar (addresses have
> changed and the compiler uses different GPRs in some places).

Your text might be confusion. When I read it the first time I thought 
you were saying that the compiler was now using another GPR than r13.

> 
> Signed-off-by: Alastair D'Silva <alastair@d-silva.org>

I guess the same has to be done for current, see 
arch/powerpc/include/asm/current.h :

/*
  * We keep `current' in r2 for speed.
  */
register struct task_struct *current asm ("r2");

> ---
>   arch/powerpc/include/asm/paca.h | 44 +++++++++++++++++++++++----------
>   arch/powerpc/kernel/paca.c      |  2 +-
>   2 files changed, 32 insertions(+), 14 deletions(-)
> 
> diff --git a/arch/powerpc/include/asm/paca.h b/arch/powerpc/include/asm/paca.h
> index e843bc5d1a0f..9c9e2dea0f9b 100644
> --- a/arch/powerpc/include/asm/paca.h
> +++ b/arch/powerpc/include/asm/paca.h
> @@ -34,19 +34,6 @@
>   #include <asm/cpuidle.h>
>   #include <asm/atomic.h>
>   
> -register struct paca_struct *local_paca asm("r13");
> -
> -#if defined(CONFIG_DEBUG_PREEMPT) && defined(CONFIG_SMP)
> -extern unsigned int debug_smp_processor_id(void); /* from linux/smp.h */
> -/*
> - * Add standard checks that preemption cannot occur when using get_paca():
> - * otherwise the paca_struct it points to may be the wrong one just after.
> - */
> -#define get_paca()	((void) debug_smp_processor_id(), local_paca)
> -#else
> -#define get_paca()	local_paca
> -#endif
> -
>   #ifdef CONFIG_PPC_PSERIES
>   #define get_lppaca()	(get_paca()->lppaca_ptr)
>   #endif
> @@ -266,6 +253,37 @@ struct paca_struct {
>   #endif
>   } ____cacheline_aligned;
>   
> +#if defined(CONFIG_DEBUG_PREEMPT) && defined(CONFIG_SMP)
> +extern unsigned int debug_smp_processor_id(void); /* from linux/smp.h */
> +#endif

Why moving this down, why not leaving at the same place as before ?

If you really need to move it, you should remove the 'extern' at the 
same time to make checkpatch happy.

> +
> +static inline struct paca_struct *get_paca_no_preempt_check(void)
> +{
> +	register struct paca_struct *paca asm("r13");

Should be a blank line there.

> +	return paca;
> +}
> +
> +static inline struct paca_struct *get_paca(void)
> +{
> +#if defined(CONFIG_DEBUG_PREEMPT) && defined(CONFIG_SMP)
> +	/*
> +	 * Add standard checks that preemption cannot occur when using get_paca():
> +	 * otherwise the paca_struct it points to may be the wrong one just after.
> +	 */
> +	debug_smp_processor_id();
> +#endif
> +	return get_paca_no_preempt_check();
> +}
> +
> +#define local_paca	get_paca_no_preempt_check()
> +
> +static inline void set_paca(struct paca_struct *new)
> +{
> +	register struct paca_struct *paca asm("r13");

Blank line should be added here.

> +	paca = new;
> +}
> +
> +
>   extern void copy_mm_to_paca(struct mm_struct *mm);
>   extern struct paca_struct **paca_ptrs;
>   extern void initialise_paca(struct paca_struct *new_paca, int cpu);
> diff --git a/arch/powerpc/kernel/paca.c b/arch/powerpc/kernel/paca.c
> index 913bfca09c4f..ae5c243f9d5a 100644
> --- a/arch/powerpc/kernel/paca.c
> +++ b/arch/powerpc/kernel/paca.c
> @@ -172,7 +172,7 @@ void __init initialise_paca(struct paca_struct *new_paca, int cpu)
>   void setup_paca(struct paca_struct *new_paca)
>   {
>   	/* Setup r13 */
> -	local_paca = new_paca;
> +	set_paca(new_paca);
>   
>   #ifdef CONFIG_PPC_BOOK3E
>   	/* On Book3E, initialize the TLB miss exception frames */
> 

Christophe

"Alastair D'Silva" <alastair@au1.ibm.com> writes:

> From: Alastair D'Silva <alastair@d-silva.org>
>
> When building an LTO kernel, the existing code generates warnings:
>     ./arch/powerpc/include/asm/paca.h:37:30: warning: register of
>         ‘local_paca’ used for multiple global register variables
>      register struct paca_struct *local_paca asm("r13");
>                               ^
>     ./arch/powerpc/include/asm/paca.h:37:30: note: conflicts with
>         ‘local_paca’
>
> This patch reworks local_paca into an inline getter & setter function,
> which addresses the warning.
>
> Generated ASM from this patch is broadly similar (addresses have
> changed and the compiler uses different GPRs in some places).

Ditto to Christophe's comment; I'd love to know how to build this so I
can actually see the differences. Perhaps you could bundle up all the
required changes and send it as a patch series with a cover letter
explaining this?

> +static inline struct paca_struct *get_paca_no_preempt_check(void)
> +{
> +	register struct paca_struct *paca asm("r13");
> +	return paca;
> +}

Isn't the convention to have the { on the same line as the function, or
am I horrible mis-remembering things?

Should these functions be __always_inline?

Regards,
Daniel

> +
> +static inline struct paca_struct *get_paca(void)
> +{
> +#if defined(CONFIG_DEBUG_PREEMPT) && defined(CONFIG_SMP)
> +	/*
> +	 * Add standard checks that preemption cannot occur when using get_paca():
> +	 * otherwise the paca_struct it points to may be the wrong one just after.
> +	 */
> +	debug_smp_processor_id();
> +#endif
> +	return get_paca_no_preempt_check();
> +}
> +
> +#define local_paca	get_paca_no_preempt_check()
> +
> +static inline void set_paca(struct paca_struct *new)
> +{
> +	register struct paca_struct *paca asm("r13");
> +	paca = new;
> +}
> +
> +
>  extern void copy_mm_to_paca(struct mm_struct *mm);
>  extern struct paca_struct **paca_ptrs;
>  extern void initialise_paca(struct paca_struct *new_paca, int cpu);
> diff --git a/arch/powerpc/kernel/paca.c b/arch/powerpc/kernel/paca.c
> index 913bfca09c4f..ae5c243f9d5a 100644
> --- a/arch/powerpc/kernel/paca.c
> +++ b/arch/powerpc/kernel/paca.c
> @@ -172,7 +172,7 @@ void __init initialise_paca(struct paca_struct *new_paca, int cpu)
>  void setup_paca(struct paca_struct *new_paca)
>  {
>  	/* Setup r13 */
> -	local_paca = new_paca;
> +	set_paca(new_paca);
>  
>  #ifdef CONFIG_PPC_BOOK3E
>  	/* On Book3E, initialize the TLB miss exception frames */
> -- 
> 2.20.1

On 14/3/19 10:54 am, Daniel Axtens wrote:
>> +static inline struct paca_struct *get_paca_no_preempt_check(void)
>> +{
>> +	register struct paca_struct *paca asm("r13");
>> +	return paca;
>> +}
> 
> Isn't the convention to have the { on the same line as the function, or
> am I horrible mis-remembering things?

However, there is one special case, namely functions: they have the
opening brace at the beginning of the next line, thus:

         int function(int x)
         {
                 body of function
         }

Heretic people all over the world have claimed that this inconsistency
is ...  well ...  inconsistent, but all right-thinking people know that
(a) K&R are **right** and (b) K&R are right.  Besides, functions are
special anyway (you can't nest them in C).

On Wed, 2019-03-13 at 10:06 +0100, Christophe Leroy wrote:
> Hello,

Thanks for reviewing :)
> 
> Le 13/03/2019 à 04:42, Alastair D'Silva a écrit :
> > From: Alastair D'Silva <alastair@d-silva.org>
> > 
> > When building an LTO kernel, the existing code generates warnings:
> >      ./arch/powerpc/include/asm/paca.h:37:30: warning: register of
> >          ‘local_paca’ used for multiple global register variables
> >       register struct paca_struct *local_paca asm("r13");
> >                                ^
> >      ./arch/powerpc/include/asm/paca.h:37:30: note: conflicts with
> >          ‘local_paca’
> 
> How do you build a LTO kernel ?

I'm using Andi Kleen's LTO tree:
https://github.com/andikleen/linux-misc/tree/lto-420-1

with a few other patches:
https://github.com/andikleen/linux-misc/pull/27

You'll need to add the following to your .config:
CONFIG_LTO_MENU=y
CONFIG_LTO=y

> 
> > This patch reworks local_paca into an inline getter & setter
> > function,
> > which addresses the warning.
> 
> This patch adds sparse warnings, see 
> https://patchwork.ozlabs.org/patch/1055875/

These warnings are bogus, they replace warnings that flagged against
spinlock.h.

> > Generated ASM from this patch is broadly similar (addresses have
> > changed and the compiler uses different GPRs in some places).
> 
> Your text might be confusion. When I read it the first time I
> thought 
> you were saying that the compiler was now using another GPR than r13.
> 

I'll see if I can improve it.

> > Signed-off-by: Alastair D'Silva <alastair@d-silva.org>
> 
> I guess the same has to be done for current, see 
> arch/powerpc/include/asm/current.h :
> 
> /*
>   * We keep `current' in r2 for speed.
>   */
> register struct task_struct *current asm ("r2");

Hmm, I didn't see problems on PPC64 as that already uses an inline
function. I'll address this in another patch for the PPC32 case.

> > ---
> >   arch/powerpc/include/asm/paca.h | 44 +++++++++++++++++++++++-----
> > -----
> >   arch/powerpc/kernel/paca.c      |  2 +-
> >   2 files changed, 32 insertions(+), 14 deletions(-)
> > 
> > diff --git a/arch/powerpc/include/asm/paca.h
> > b/arch/powerpc/include/asm/paca.h
> > index e843bc5d1a0f..9c9e2dea0f9b 100644
> > --- a/arch/powerpc/include/asm/paca.h
> > +++ b/arch/powerpc/include/asm/paca.h
> > @@ -34,19 +34,6 @@
> >   #include <asm/cpuidle.h>
> >   #include <asm/atomic.h>
> >   
> > -register struct paca_struct *local_paca asm("r13");
> > -
> > -#if defined(CONFIG_DEBUG_PREEMPT) && defined(CONFIG_SMP)
> > -extern unsigned int debug_smp_processor_id(void); /* from
> > linux/smp.h */
> > -/*
> > - * Add standard checks that preemption cannot occur when using
> > get_paca():
> > - * otherwise the paca_struct it points to may be the wrong one
> > just after.
> > - */
> > -#define get_paca()	((void) debug_smp_processor_id(), local_paca)
> > -#else
> > -#define get_paca()	local_paca
> > -#endif
> > -
> >   #ifdef CONFIG_PPC_PSERIES
> >   #define get_lppaca()	(get_paca()->lppaca_ptr)
> >   #endif
> > @@ -266,6 +253,37 @@ struct paca_struct {
> >   #endif
> >   } ____cacheline_aligned;
> >   
> > +#if defined(CONFIG_DEBUG_PREEMPT) && defined(CONFIG_SMP)
> > +extern unsigned int debug_smp_processor_id(void); /* from
> > linux/smp.h */
> > +#endif
> 
> Why moving this down, why not leaving at the same place as before ?
> 
> If you really need to move it, you should remove the 'extern' at the 
> same time to make checkpatch happy.

I moved it to keep it close to the usage of it.

I suppose the new implementation should be in the same place though.

> > +
> > +static inline struct paca_struct *get_paca_no_preempt_check(void)
> > +{
> > +	register struct paca_struct *paca asm("r13");
> 
> Should be a blank line there.
Whoops, I thought I ran checkpatch, but clearly, I forgot. I'll
resubmit.

> > +	return paca;
> > +}
> > +
> > +static inline struct paca_struct *get_paca(void)
> > +{
> > +#if defined(CONFIG_DEBUG_PREEMPT) && defined(CONFIG_SMP)
> > +	/*
> > +	 * Add standard checks that preemption cannot occur when using
> > get_paca():
> > +	 * otherwise the paca_struct it points to may be the wrong one
> > just after.
> > +	 */
> > +	debug_smp_processor_id();
> > +#endif
> > +	return get_paca_no_preempt_check();
> > +}
> > +
> > +#define local_paca	get_paca_no_preempt_check()
> > +
> > +static inline void set_paca(struct paca_struct *new)
> > +{
> > +	register struct paca_struct *paca asm("r13");
> 
> Blank line should be added here.
> 
> > +	paca = new;
> > +}
> > +
> > +
> >   extern void copy_mm_to_paca(struct mm_struct *mm);
> >   extern struct paca_struct **paca_ptrs;
> >   extern void initialise_paca(struct paca_struct *new_paca, int
> > cpu);
> > diff --git a/arch/powerpc/kernel/paca.c
> > b/arch/powerpc/kernel/paca.c
> > index 913bfca09c4f..ae5c243f9d5a 100644
> > --- a/arch/powerpc/kernel/paca.c
> > +++ b/arch/powerpc/kernel/paca.c
> > @@ -172,7 +172,7 @@ void __init initialise_paca(struct paca_struct
> > *new_paca, int cpu)
> >   void setup_paca(struct paca_struct *new_paca)
> >   {
> >   	/* Setup r13 */
> > -	local_paca = new_paca;
> > +	set_paca(new_paca);
> >   
> >   #ifdef CONFIG_PPC_BOOK3E
> >   	/* On Book3E, initialize the TLB miss exception frames */
> > 
> 
> Christophe
>

On Thu, 2019-03-14 at 10:54 +1100, Daniel Axtens wrote:
> "Alastair D'Silva" <alastair@au1.ibm.com> writes:
> 
> > From: Alastair D'Silva <alastair@d-silva.org>
> > 
> > When building an LTO kernel, the existing code generates warnings:
> >     ./arch/powerpc/include/asm/paca.h:37:30: warning: register of
> >         ‘local_paca’ used for multiple global register variables
> >      register struct paca_struct *local_paca asm("r13");
> >                               ^
> >     ./arch/powerpc/include/asm/paca.h:37:30: note: conflicts with
> >         ‘local_paca’
> > 
> > This patch reworks local_paca into an inline getter & setter
> > function,
> > which addresses the warning.
> > 
> > Generated ASM from this patch is broadly similar (addresses have
> > changed and the compiler uses different GPRs in some places).
> 
> Ditto to Christophe's comment; I'd love to know how to build this so
> I
> can actually see the differences. Perhaps you could bundle up all the
> required changes and send it as a patch series with a cover letter
> explaining this?

The differences are visible in a normal build, but if you want to play
with LTO, see my comments to Christophe.

> 
> > +static inline struct paca_struct *get_paca_no_preempt_check(void)
> > +{
> > +	register struct paca_struct *paca asm("r13");
> > +	return paca;
> > +}
> 
> Isn't the convention to have the { on the same line as the function,
> or
> am I horrible mis-remembering things?
> 
You are :)

> Should these functions be __always_inline?
> 

Yes, they should, I'll add that to V3.