The functions calls get_cpu()/put_cpu() meaningless because the actual
CPU that would execute the caller's function is not necessarily the
current one.

The smp_call_function_single() which is called by
smp_call_function_any() does the right get/put protection.

Signed-off-by: Yury Norov (NVIDIA) <yury.norov@gmail.com>
---
 kernel/smp.c | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)

diff --git a/kernel/smp.c b/kernel/smp.c
index 02f52291fae4..fa50ed459703 100644
--- a/kernel/smp.c
+++ b/kernel/smp.c
@@ -754,17 +754,13 @@ EXPORT_SYMBOL_GPL(smp_call_function_single_async);
 int smp_call_function_any(const struct cpumask *mask,
 			  smp_call_func_t func, void *info, int wait)
 {
-	unsigned int cpu;
-	int ret;
+	unsigned int cpu = smp_processor_id();
 
 	/* Try for same CPU (cheapest) */
-	cpu = get_cpu();
 	if (!cpumask_test_cpu(cpu, mask))
 		cpu = sched_numa_find_nth_cpu(mask, 0, cpu_to_node(cpu));
 
-	ret = smp_call_function_single(cpu, func, info, wait);
-	put_cpu();
-	return ret;
+	return smp_call_function_single(cpu, func, info, wait);
 }
 EXPORT_SYMBOL_GPL(smp_call_function_any);
 
-- 
2.43.0

On 2025-10-08 12:57, Yury Norov (NVIDIA) wrote:
> The functions calls get_cpu()/put_cpu() meaningless because the actual
> CPU that would execute the caller's function is not necessarily the
> current one.
> 
> The smp_call_function_single() which is called by
> smp_call_function_any() does the right get/put protection.
> 
> Signed-off-by: Yury Norov (NVIDIA) <yury.norov@gmail.com>
> ---
>   kernel/smp.c | 8 ++------
>   1 file changed, 2 insertions(+), 6 deletions(-)
> 
> diff --git a/kernel/smp.c b/kernel/smp.c
> index 02f52291fae4..fa50ed459703 100644
> --- a/kernel/smp.c
> +++ b/kernel/smp.c
> @@ -754,17 +754,13 @@ EXPORT_SYMBOL_GPL(smp_call_function_single_async);
>   int smp_call_function_any(const struct cpumask *mask,
>   			  smp_call_func_t func, void *info, int wait)
>   {
> -	unsigned int cpu;
> -	int ret;
> +	unsigned int cpu = smp_processor_id();

I wonder whether this passes any moderate testing with kernel debug
options enabled. I would at the very least expect a
raw_smp_processor_id() call here not to trip debug warnings.

AFAIU smp_call_function_any call be called from preemptible context,
right ?

Thanks,

Mathieu

>   
>   	/* Try for same CPU (cheapest) */
> -	cpu = get_cpu();
>   	if (!cpumask_test_cpu(cpu, mask))
>   		cpu = sched_numa_find_nth_cpu(mask, 0, cpu_to_node(cpu));
>   
> -	ret = smp_call_function_single(cpu, func, info, wait);
> -	put_cpu();
> -	return ret;
> +	return smp_call_function_single(cpu, func, info, wait);
>   }
>   EXPORT_SYMBOL_GPL(smp_call_function_any);
>   


-- 
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com

On Wed, Oct 08, 2025 at 01:06:18PM -0400, Mathieu Desnoyers wrote:
> On 2025-10-08 12:57, Yury Norov (NVIDIA) wrote:
> > The functions calls get_cpu()/put_cpu() meaningless because the actual
> > CPU that would execute the caller's function is not necessarily the
> > current one.
> > 
> > The smp_call_function_single() which is called by
> > smp_call_function_any() does the right get/put protection.
> > 
> > Signed-off-by: Yury Norov (NVIDIA) <yury.norov@gmail.com>
> > ---
> >   kernel/smp.c | 8 ++------
> >   1 file changed, 2 insertions(+), 6 deletions(-)
> > 
> > diff --git a/kernel/smp.c b/kernel/smp.c
> > index 02f52291fae4..fa50ed459703 100644
> > --- a/kernel/smp.c
> > +++ b/kernel/smp.c
> > @@ -754,17 +754,13 @@ EXPORT_SYMBOL_GPL(smp_call_function_single_async);
> >   int smp_call_function_any(const struct cpumask *mask,
> >   			  smp_call_func_t func, void *info, int wait)
> >   {
> > -	unsigned int cpu;
> > -	int ret;
> > +	unsigned int cpu = smp_processor_id();
> 
> I wonder whether this passes any moderate testing with kernel debug
> options enabled. I would at the very least expect a
> raw_smp_processor_id() call here not to trip debug warnings.
> 
> AFAIU smp_call_function_any call be called from preemptible context,
> right ?

You're right, we need to retain current CPU unless the work is
scheduled. I need to test better. Sorry for the noise.