[v2] mm/percpu: prevent concurrency problem for pcpu_nr_populated read with spin lock

[PATCH v2] mm/percpu: prevent concurrency problem for pcpu_nr_populated read with spin lock

Posted by Jeongjun Park 3 months ago

pcpu_nr_pages() reads pcpu_nr_populated without any protection, which
causes a data race between read/write.

However, since this is an intended race, we should add a data_race
annotation instead of add a spin lock.

Reported-by: syzbot+e5bd32b79413e86f389e@syzkaller.appspotmail.com
Fixes: 7e8a6304d541 ("/proc/meminfo: add percpu populated pages count")
Suggested-by: Shakeel Butt <shakeel.butt@linux.dev>
Signed-off-by: Jeongjun Park <aha310510@gmail.com>
---
v2: Change it as suggested by Shakeel Butt
- Link to v1: https://lore.kernel.org/all/20250702082749.141616-1-aha310510@gmail.com/
---
 mm/percpu.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mm/percpu.c b/mm/percpu.c
index b35494c8ede2..782cc148b39c 100644
--- a/mm/percpu.c
+++ b/mm/percpu.c
@@ -3355,7 +3355,7 @@ void __init setup_per_cpu_areas(void)
  */
 unsigned long pcpu_nr_pages(void)
 {
-	return pcpu_nr_populated * pcpu_nr_units;
+	return data_race(READ_ONCE(pcpu_nr_populated) * pcpu_nr_units);
 }
 
 /*
--

Re: [PATCH v2] mm/percpu: prevent concurrency problem for pcpu_nr_populated read with spin lock

Posted by Vlastimil Babka 2 months, 4 weeks ago

On 7/3/25 08:56, Jeongjun Park wrote:
> pcpu_nr_pages() reads pcpu_nr_populated without any protection, which
> causes a data race between read/write.
> 
> However, since this is an intended race, we should add a data_race
> annotation instead of add a spin lock.
> 
> Reported-by: syzbot+e5bd32b79413e86f389e@syzkaller.appspotmail.com
> Fixes: 7e8a6304d541 ("/proc/meminfo: add percpu populated pages count")
> Suggested-by: Shakeel Butt <shakeel.butt@linux.dev>
> Signed-off-by: Jeongjun Park <aha310510@gmail.com>
> ---
> v2: Change it as suggested by Shakeel Butt
> - Link to v1: https://lore.kernel.org/all/20250702082749.141616-1-aha310510@gmail.com/
> ---
>  mm/percpu.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/mm/percpu.c b/mm/percpu.c
> index b35494c8ede2..782cc148b39c 100644
> --- a/mm/percpu.c
> +++ b/mm/percpu.c
> @@ -3355,7 +3355,7 @@ void __init setup_per_cpu_areas(void)
>   */
>  unsigned long pcpu_nr_pages(void)
>  {
> -	return pcpu_nr_populated * pcpu_nr_units;
> +	return data_race(READ_ONCE(pcpu_nr_populated) * pcpu_nr_units);

Nit: pcpu_nr_units should be excluded from the scope of data_race() as no
race can happen there.

>  }
>  
>  /*
> --

Re: [PATCH v2] mm/percpu: prevent concurrency problem for pcpu_nr_populated read with spin lock

Posted by Andrew Morton 2 months, 3 weeks ago

On Fri, 11 Jul 2025 17:56:34 +0200 Vlastimil Babka <vbabka@suse.cz> wrote:

> On 7/3/25 08:56, Jeongjun Park wrote:
> > pcpu_nr_pages() reads pcpu_nr_populated without any protection, which
> > causes a data race between read/write.
> > 
> > However, since this is an intended race, we should add a data_race
> > annotation instead of add a spin lock.
> > 
> > Reported-by: syzbot+e5bd32b79413e86f389e@syzkaller.appspotmail.com
> > Fixes: 7e8a6304d541 ("/proc/meminfo: add percpu populated pages count")
> > Suggested-by: Shakeel Butt <shakeel.butt@linux.dev>
> > Signed-off-by: Jeongjun Park <aha310510@gmail.com>
> > ---
> > v2: Change it as suggested by Shakeel Butt
> > - Link to v1: https://lore.kernel.org/all/20250702082749.141616-1-aha310510@gmail.com/
> > ---
> >  mm/percpu.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/mm/percpu.c b/mm/percpu.c
> > index b35494c8ede2..782cc148b39c 100644
> > --- a/mm/percpu.c
> > +++ b/mm/percpu.c
> > @@ -3355,7 +3355,7 @@ void __init setup_per_cpu_areas(void)
> >   */
> >  unsigned long pcpu_nr_pages(void)
> >  {
> > -	return pcpu_nr_populated * pcpu_nr_units;
> > +	return data_race(READ_ONCE(pcpu_nr_populated) * pcpu_nr_units);
> 
> Nit: pcpu_nr_units should be excluded from the scope of data_race() as no
> race can happen there.

This?

--- a/mm/percpu.c~mm-percpu-prevent-concurrency-problem-for-pcpu_nr_populated-read-with-spin-lock-fix
+++ a/mm/percpu.c
@@ -3355,7 +3355,7 @@ void __init setup_per_cpu_areas(void)
  */
 unsigned long pcpu_nr_pages(void)
 {
-	return data_race(READ_ONCE(pcpu_nr_populated) * pcpu_nr_units);
+	return data_race(READ_ONCE(pcpu_nr_populated)) * pcpu_nr_units;
 }
 
 /*
_

Re: [PATCH v2] mm/percpu: prevent concurrency problem for pcpu_nr_populated read with spin lock

Posted by Vlastimil Babka 2 months, 3 weeks ago

On 7/13/25 00:28, Andrew Morton wrote:
> On Fri, 11 Jul 2025 17:56:34 +0200 Vlastimil Babka <vbabka@suse.cz> wrote:
> 
>> On 7/3/25 08:56, Jeongjun Park wrote:
>> > pcpu_nr_pages() reads pcpu_nr_populated without any protection, which
>> > causes a data race between read/write.
>> > 
>> > However, since this is an intended race, we should add a data_race
>> > annotation instead of add a spin lock.
>> > 
>> > Reported-by: syzbot+e5bd32b79413e86f389e@syzkaller.appspotmail.com
>> > Fixes: 7e8a6304d541 ("/proc/meminfo: add percpu populated pages count")
>> > Suggested-by: Shakeel Butt <shakeel.butt@linux.dev>
>> > Signed-off-by: Jeongjun Park <aha310510@gmail.com>
>> > ---
>> > v2: Change it as suggested by Shakeel Butt
>> > - Link to v1: https://lore.kernel.org/all/20250702082749.141616-1-aha310510@gmail.com/
>> > ---
>> >  mm/percpu.c | 2 +-
>> >  1 file changed, 1 insertion(+), 1 deletion(-)
>> > 
>> > diff --git a/mm/percpu.c b/mm/percpu.c
>> > index b35494c8ede2..782cc148b39c 100644
>> > --- a/mm/percpu.c
>> > +++ b/mm/percpu.c
>> > @@ -3355,7 +3355,7 @@ void __init setup_per_cpu_areas(void)
>> >   */
>> >  unsigned long pcpu_nr_pages(void)
>> >  {
>> > -	return pcpu_nr_populated * pcpu_nr_units;
>> > +	return data_race(READ_ONCE(pcpu_nr_populated) * pcpu_nr_units);
>> 
>> Nit: pcpu_nr_units should be excluded from the scope of data_race() as no
>> race can happen there.
> 
> This?
> 
> --- a/mm/percpu.c~mm-percpu-prevent-concurrency-problem-for-pcpu_nr_populated-read-with-spin-lock-fix
> +++ a/mm/percpu.c
> @@ -3355,7 +3355,7 @@ void __init setup_per_cpu_areas(void)
>   */
>  unsigned long pcpu_nr_pages(void)
>  {
> -	return data_race(READ_ONCE(pcpu_nr_populated) * pcpu_nr_units);
> +	return data_race(READ_ONCE(pcpu_nr_populated)) * pcpu_nr_units;

Yes, thanks!

>  }
>  
>  /*
> _
>