On Tue, Mar 24, 2020 at 03:36:28PM +0000, Will Deacon wrote:
> Commit 860c8802ace1 ("rcu: Use WRITE_ONCE() for assignments to ->pprev
> for hlist_nulls") added WRITE_ONCE() invocations to hlist_nulls_add_head()
> and hlist_nulls_del().
>
> Since these functions should not ordinarily run concurrently with other
> list accessors, restore the plain C assignments so that KCSAN can yell
> if a data race occurs.
>
> Cc: Eric Dumazet <edumazet@xxxxxxxxxx>
> Cc: Paul E. McKenney <paulmck@xxxxxxxxxx>
> Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
> Signed-off-by: Will Deacon <will@xxxxxxxxxx>
And this means that the lockless uses of hlist_nulls_empty() need
attention, correct?
Thanx, Paul
> ---
> include/linux/list_nulls.h | 6 +++---
> 1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/include/linux/list_nulls.h b/include/linux/list_nulls.h
> index fa51a801bf32..fd154ceb5b0d 100644
> --- a/include/linux/list_nulls.h
> +++ b/include/linux/list_nulls.h
> @@ -80,10 +80,10 @@ static inline void hlist_nulls_add_head(struct hlist_nulls_node *n,
> struct hlist_nulls_node *first = h->first;
>
> n->next = first;
> - WRITE_ONCE(n->pprev, &h->first);
> + n->pprev = &h->first;
> h->first = n;
> if (!is_a_nulls(first))
> - WRITE_ONCE(first->pprev, &n->next);
> + first->pprev = &n->next;
> }
>
> static inline void __hlist_nulls_del(struct hlist_nulls_node *n)
> @@ -99,7 +99,7 @@ static inline void __hlist_nulls_del(struct hlist_nulls_node *n)
> static inline void hlist_nulls_del(struct hlist_nulls_node *n)
> {
> __hlist_nulls_del(n);
> - WRITE_ONCE(n->pprev, LIST_POISON2);
> + n->pprev = LIST_POISON2;
> }
>
> /**
> --
> 2.20.1
>
