2021-02-21 19:44:02

by Matthew Wilcox

Subject: [PATCH] mm: Use rcu_dereference in in_vfork

Fix a sparse warning by using rcu_dereference(). Technically this is a
bug and a sufficiently aggressive compiler could reload the `real_parent'
pointer outside the protection of the rcu lock (and access freed memory),
but I think it's pretty unlikely to happen.

Fixes: b18dc5f291c0 ("mm, oom: skip vforked tasks from being selected")
Signed-off-by: Matthew Wilcox (Oracle) <[email protected]>
---
include/linux/sched/mm.h | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/include/linux/sched/mm.h b/include/linux/sched/mm.h
index 1ae08b8462a4..90b2a0bce11c 100644
--- a/include/linux/sched/mm.h
+++ b/include/linux/sched/mm.h
@@ -140,7 +140,8 @@ static inline bool in_vfork(struct task_struct *tsk)
* another oom-unkillable task does this it should blame itself.
*/
rcu_read_lock();
- ret = tsk->vfork_done && tsk->real_parent->mm == tsk->mm;
+ ret = tsk->vfork_done &&
+ rcu_dereference(tsk->real_parent)->mm == tsk->mm;
rcu_read_unlock();

return ret;
--
2.29.2

2021-02-22 01:47:55

by Miaohe Lin

Subject: Re: [PATCH] mm: Use rcu_dereference in in_vfork

On 2021/2/22 3:42, Matthew Wilcox (Oracle) wrote:
> Fix a sparse warning by using rcu_dereference(). Technically this is a
> bug and a sufficiently aggressive compiler could reload the `real_parent'
> pointer outside the protection of the rcu lock (and access freed memory),
> but I think it's pretty unlikely to happen.
>
> Fixes: b18dc5f291c0 ("mm, oom: skip vforked tasks from being selected")
> Signed-off-by: Matthew Wilcox (Oracle) <[email protected]>

Looks good to me. Thanks!
Reviewed-by: Miaohe Lin <[email protected]>

> ---
> include/linux/sched/mm.h | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/include/linux/sched/mm.h b/include/linux/sched/mm.h
> index 1ae08b8462a4..90b2a0bce11c 100644
> --- a/include/linux/sched/mm.h
> +++ b/include/linux/sched/mm.h
> @@ -140,7 +140,8 @@ static inline bool in_vfork(struct task_struct *tsk)
> * another oom-unkillable task does this it should blame itself.
> */
> rcu_read_lock();
> - ret = tsk->vfork_done && tsk->real_parent->mm == tsk->mm;
> + ret = tsk->vfork_done &&
> + rcu_dereference(tsk->real_parent)->mm == tsk->mm;
> rcu_read_unlock();
>
> return ret;
>

2021-02-22 08:34:41

by Michal Hocko

Subject: Re: [PATCH] mm: Use rcu_dereference in in_vfork

On Sun 21-02-21 19:42:06, Matthew Wilcox wrote:
> Fix a sparse warning by using rcu_dereference(). Technically this is a
> bug and a sufficiently aggressive compiler could reload the `real_parent'
> pointer outside the protection of the rcu lock (and access freed memory),
> but I think it's pretty unlikely to happen.
>
> Fixes: b18dc5f291c0 ("mm, oom: skip vforked tasks from being selected")
> Signed-off-by: Matthew Wilcox (Oracle) <[email protected]>

I must have missed the RCU part.
Thanks!

Acked-by: Michal Hocko <[email protected]>

> ---
> include/linux/sched/mm.h | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/include/linux/sched/mm.h b/include/linux/sched/mm.h
> index 1ae08b8462a4..90b2a0bce11c 100644
> --- a/include/linux/sched/mm.h
> +++ b/include/linux/sched/mm.h
> @@ -140,7 +140,8 @@ static inline bool in_vfork(struct task_struct *tsk)
> * another oom-unkillable task does this it should blame itself.
> */
> rcu_read_lock();
> - ret = tsk->vfork_done && tsk->real_parent->mm == tsk->mm;
> + ret = tsk->vfork_done &&
> + rcu_dereference(tsk->real_parent)->mm == tsk->mm;
> rcu_read_unlock();
>
> return ret;
> --
> 2.29.2
>

--
Michal Hocko
SUSE Labs