On 02/22/2019 01:44 AM, Li RongQing wrote:
> using rcu_assign_pointer when setting, which has a memory
> barrier to ensure the initialization is seen first.
>
> using rcu_dereference when dereference this pointer
>
> Signed-off-by: Zhang Yu <zhangy...@baidu.com>
> Signed-off-by: Li RongQing <lirongq...@baidu.com>
> ---
> net/ipv6/ip6_fib.c | 6 ++++--
> 1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/net/ipv6/ip6_fib.c b/net/ipv6/ip6_fib.c
> index 6613d8dbb0e5..b73d40d68178 100644
> --- a/net/ipv6/ip6_fib.c
> +++ b/net/ipv6/ip6_fib.c
> @@ -1143,7 +1143,7 @@ static int fib6_add_rt2node(struct fib6_node *fn,
> struct fib6_info *rt,
>
> atomic_inc(&rt->fib6_ref);
> rcu_assign_pointer(rt->fib6_node, fn);
> - rt->fib6_next = iter->fib6_next;
> + rcu_assign_pointer(rt->fib6_next, iter->fib6_next);
We do not need a barrier here, the object is still private.
> rcu_assign_pointer(*ins, rt);
> if (!info->skip_notify)
> inet6_rt_notify(RTM_NEWROUTE, rt, info, NLM_F_REPLACE);
> @@ -1761,7 +1761,9 @@ static void fib6_del_route(struct fib6_table *table,
> struct fib6_node *fn,
> RT6_TRACE("fib6_del_route\n");
>
> /* Unlink it */
> - *rtp = rt->fib6_next;
> + *rtp = rcu_dereference_protected(rt->fib6_next,
> + lockdep_is_held(&rt->fib6_table->tb6_lock));
This change will likely add a sparse bug.
> +
> rt->fib6_node = NULL;
> net->ipv6.rt6_stats->fib_rt_entries--;
> net->ipv6.rt6_stats->fib_discarded_routes++;
>
Really I do not believe these changes are needed.
barriers should be added blindly.
