RCU safe comment extension handling

Signed-off-by: Jozsef Kadlecsik <kadlec@blackhole.kfki.hu>

4 files changed, 42 insertions, 20 deletions

diff --git a/kernel/include/linux/netfilter/ipset/ip_set.h b/kernel/include/linux/netfilter/ipset/ip_set.h
index 5c1c6a5..8894904 100644
--- a/kernel/include/linux/netfilter/ipset/ip_set.h
+++ b/kernel/include/linux/netfilter/ipset/ip_set.h
@@ -109,8 +109,13 @@ struct ip_set_counter {
 	atomic64_t packets;
 };
 
+struct ip_set_comment_rcu {
+	struct rcu_head rcu;
+	char str[0];
+};
+
 struct ip_set_comment {
-	char *str;
+	struct ip_set_comment_rcu __rcu *c;
 };
 
 struct ip_set_skbinfo {
diff --git a/kernel/include/linux/netfilter/ipset/ip_set_comment.h b/kernel/include/linux/netfilter/ipset/ip_set_comment.h
index 21217ea..42dce2b 100644
--- a/kernel/include/linux/netfilter/ipset/ip_set_comment.h
+++ b/kernel/include/linux/netfilter/ipset/ip_set_comment.h
@@ -20,37 +20,43 @@ static inline void
 ip_set_init_comment(struct ip_set_comment *comment,
 		    const struct ip_set_ext *ext)
 {
+	struct ip_set_comment_rcu *c = rcu_dereference_protected(comment->c, 1);
 	size_t len = ext->comment ? strlen(ext->comment) : 0;
 
-	if (unlikely(comment->str)) {
-		kfree(comment->str);
-		comment->str = NULL;
+	if (unlikely(c)) {
+		kfree_rcu(c, rcu);
+		rcu_assign_pointer(comment->c, NULL);
 	}
 	if (!len)
 		return;
 	if (unlikely(len > IPSET_MAX_COMMENT_SIZE))
 		len = IPSET_MAX_COMMENT_SIZE;
-	comment->str = kzalloc(len + 1, GFP_ATOMIC);
-	if (unlikely(!comment->str))
+	c = kzalloc(sizeof(*c) + len + 1, GFP_ATOMIC);
+	if (unlikely(!c))
 		return;
-	strlcpy(comment->str, ext->comment, len + 1);
+	strlcpy(c->str, ext->comment, len + 1);
+	rcu_assign_pointer(comment->c, c);
 }
 
 static inline int
 ip_set_put_comment(struct sk_buff *skb, struct ip_set_comment *comment)
 {
-	if (!comment->str)
+	struct ip_set_comment_rcu *c = rcu_dereference_bh(comment->c);
+
+	if (!c)
 		return 0;
-	return nla_put_string(skb, IPSET_ATTR_COMMENT, comment->str);
+	return nla_put_string(skb, IPSET_ATTR_COMMENT, c->str);
 }
 
 static inline void
 ip_set_comment_free(struct ip_set_comment *comment)
 {
-	if (unlikely(!comment->str))
+	struct ip_set_comment_rcu *c = rcu_dereference_bh(comment->c);
+
+	if (unlikely(!c))
 		return;
-	kfree(comment->str);
-	comment->str = NULL;
+	kfree_rcu(c, rcu);
+	rcu_assign_pointer(comment->c, NULL);
 }
 
 #endif
diff --git a/kernel/net/netfilter/ipset/ip_set_bitmap_gen.h b/kernel/net/netfilter/ipset/ip_set_bitmap_gen.h
index dba65b4..381a011 100644
--- a/kernel/net/netfilter/ipset/ip_set_bitmap_gen.h
+++ b/kernel/net/netfilter/ipset/ip_set_bitmap_gen.h
@@ -203,10 +203,13 @@ mtype_list(const struct ip_set *set,
 	struct nlattr *adt, *nested;
 	void *x;
 	u32 id, first = cb->args[IPSET_CB_ARG0];
+	int ret = 0;
 
 	adt = ipset_nest_start(skb, IPSET_ATTR_ADT);
 	if (!adt)
 		return -EMSGSIZE;
+	/* Extensions may be replaced */
+	rcu_read_lock();
 	for (; cb->args[IPSET_CB_ARG0] < map->elements;
 	     cb->args[IPSET_CB_ARG0]++) {
 		id = cb->args[IPSET_CB_ARG0];
@@ -222,7 +225,8 @@ mtype_list(const struct ip_set *set,
 		if (!nested) {
 			if (id == first) {
 				nla_nest_cancel(skb, adt);
-				return -EMSGSIZE;
+				ret = -EMSGSIZE;
+				goto out;
 			}
 
 			goto nla_put_failure;
@@ -239,16 +243,18 @@ mtype_list(const struct ip_set *set,
 	/* Set listing finished */
 	cb->args[IPSET_CB_ARG0] = 0;
 
-	return 0;
+	goto out;
 
 nla_put_failure:
 	nla_nest_cancel(skb, nested);
 	if (unlikely(id == first)) {
 		cb->args[IPSET_CB_ARG0] = 0;
-		return -EMSGSIZE;
+		ret = -EMSGSIZE;
 	}
 	ipset_nest_end(skb, adt);
-	return 0;
+out:
+	rcu_read_unlock();
+	return ret;
 }
 
 static void
diff --git a/kernel/net/netfilter/ipset/ip_set_list_set.c b/kernel/net/netfilter/ipset/ip_set_list_set.c
index 8edadb5..0b88000 100644
--- a/kernel/net/netfilter/ipset/ip_set_list_set.c
+++ b/kernel/net/netfilter/ipset/ip_set_list_set.c
@@ -478,6 +478,7 @@ list_set_list(const struct ip_set *set,
 	struct nlattr *atd, *nested;
 	u32 i = 0, first = cb->args[IPSET_CB_ARG0];
 	struct set_elem *e;
+	int ret = 0;
 
 	atd = ipset_nest_start(skb, IPSET_ATTR_ADT);
 	if (!atd)
@@ -488,6 +489,7 @@ list_set_list(const struct ip_set *set,
 		i++;
 	}
 
+	rcu_read_lock();
 	list_for_each_entry_from(e, &map->members, list) {
 		i++;
 		if (SET_WITH_TIMEOUT(set) &&
@@ -497,7 +499,8 @@ list_set_list(const struct ip_set *set,
 		if (!nested) {
 			if (i == first) {
 				nla_nest_cancel(skb, atd);
-				return -EMSGSIZE;
+				ret = -EMSGSIZE;
+				goto out;
 			}
 			goto nla_put_failure;
 		}
@@ -512,17 +515,19 @@ list_set_list(const struct ip_set *set,
 	ipset_nest_end(skb, atd);
 	/* Set listing finished */
 	cb->args[IPSET_CB_ARG0] = 0;
-	return 0;
+	goto out;
 
 nla_put_failure:
 	nla_nest_cancel(skb, nested);
 	if (unlikely(i == first)) {
 		cb->args[IPSET_CB_ARG0] = 0;
-		return -EMSGSIZE;
+		ret = -EMSGSIZE;
 	}
 	cb->args[IPSET_CB_ARG0] = i - 1;
 	ipset_nest_end(skb, atd);
-	return 0;
+out:
+	rcu_read_unlock();
+	return ret;
 }
 
 static bool