While doing __rcu annotations work on net/netfilter I found following
bug. On some arches, it is possible we publish a table while its content
is not yet committed to memory, and lockless reader can dereference wild
pointer.
Signed-off-by: Eric Dumazet <eric.dumazet@xxxxxxxxx>
---
net/netfilter/nf_conntrack_proto.c | 6 ++++++
1 files changed, 6 insertions(+)
diff --git a/net/netfilter/nf_conntrack_proto.c b/net/netfilter/nf_conntrack_proto.c
index ed6d929..dc7bb74 100644
--- a/net/netfilter/nf_conntrack_proto.c
+++ b/net/netfilter/nf_conntrack_proto.c
@@ -292,6 +292,12 @@ int nf_conntrack_l4proto_register(struct nf_conntrack_l4proto *l4proto)
for (i = 0; i < MAX_NF_CT_PROTO; i++)
proto_array[i] = &nf_conntrack_l4proto_generic;
+
+ /* Before making proto_array visible to lockless readers,
+ * we must make sure its content is committed to memory.
+ */
+ smp_wmb();
+
nf_ct_protos[l4proto->l3proto] = proto_array;
} else if (nf_ct_protos[l4proto->l3proto][l4proto->l4proto] !=
&nf_conntrack_l4proto_generic) {
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
