2024-04-27 08:52:22

by linke li

Subject: [PATCH] mm/slub: mark racy access on slab->freelist

In deactivate_slab(), slab->freelist can be changed concurrently. Mark
data race on slab->freelist as benign using READ_ONCE.

This patch is aimed at reducing the number of benign races reported by
KCSAN in order to focus future debugging effort on harmful races.

Signed-off-by: linke li <[email protected]>
---
mm/slub.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mm/slub.c b/mm/slub.c
index 0d700f6ca547..bb0e05a321e9 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -2813,7 +2813,7 @@ static void deactivate_slab(struct kmem_cache *s, struct slab *slab,
struct slab new;
struct slab old;

- if (slab->freelist) {
+ if (READ_ONCE(slab->freelist)) {
stat(s, DEACTIVATE_REMOTE_FREES);
tail = DEACTIVATE_TO_TAIL;
}
--
2.39.3 (Apple Git-146)

2024-04-28 08:10:49

by David Rientjes

Subject: Re: [PATCH] mm/slub: mark racy access on slab->freelist

On Sat, 27 Apr 2024, linke li wrote:

> In deactivate_slab(), slab->freelist can be changed concurrently. Mark
> data race on slab->freelist as benign using READ_ONCE.
>
> This patch is aimed at reducing the number of benign races reported by
> KCSAN in order to focus future debugging effort on harmful races.
>

Thanks! Do you have a data race report to copy+paste into the commit
description so people can search for it if they stumble across the same
thing?

> Signed-off-by: linke li <[email protected]>
> ---
> mm/slub.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/mm/slub.c b/mm/slub.c
> index 0d700f6ca547..bb0e05a321e9 100644
> --- a/mm/slub.c
> +++ b/mm/slub.c
> @@ -2813,7 +2813,7 @@ static void deactivate_slab(struct kmem_cache *s, struct slab *slab,
> struct slab new;
> struct slab old;
>
> - if (slab->freelist) {
> + if (READ_ONCE(slab->freelist)) {
> stat(s, DEACTIVATE_REMOTE_FREES);
> tail = DEACTIVATE_TO_TAIL;
> }
> --
> 2.39.3 (Apple Git-146)
>
>

2024-04-29 02:30:39

by linke li

Subject: Re: [PATCH] mm/slub: mark racy access on slab->freelist

> Thanks! Do you have a data race report to copy+paste into the commit
> description so people can search for it if they stumble across the same
> thing?

I don't have a data race report, just try to analyze the code statically
and then check it manually. I hope I could help in some way.

2024-05-02 12:23:13

by Vlastimil Babka

Subject: Re: [PATCH] mm/slub: mark racy access on slab->freelist

On 4/27/24 10:51, linke li wrote:
> In deactivate_slab(), slab->freelist can be changed concurrently. Mark
> data race on slab->freelist as benign using READ_ONCE.
>
> This patch is aimed at reducing the number of benign races reported by
> KCSAN in order to focus future debugging effort on harmful races.
>
> Signed-off-by: linke li <[email protected]>

Added to slab/for-6.10, thanks

> ---
> mm/slub.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/mm/slub.c b/mm/slub.c
> index 0d700f6ca547..bb0e05a321e9 100644
> --- a/mm/slub.c
> +++ b/mm/slub.c
> @@ -2813,7 +2813,7 @@ static void deactivate_slab(struct kmem_cache *s, struct slab *slab,
> struct slab new;
> struct slab old;
>
> - if (slab->freelist) {
> + if (READ_ONCE(slab->freelist)) {
> stat(s, DEACTIVATE_REMOTE_FREES);
> tail = DEACTIVATE_TO_TAIL;
> }