diff --git a/include/net/ip_fib.h b/include/net/ip_fib.h
index cba4b7c32935..825cb2800908 100644
--- a/include/net/ip_fib.h
+++ b/include/net/ip_fib.h
@@ -185,6 +185,7 @@ struct fib_table {
u32 tb_id;
int tb_default;
int tb_num_default;
+ struct rcu_head rcu;
unsigned long tb_data[0];
};
@@ -206,12 +207,16 @@ void fib_free_table(struct fib_table *tb);
static inline struct fib_table *fib_get_table(struct net *net, u32 id)
{
+ struct hlist_node *tb_hlist;
struct hlist_head *ptr;
ptr = id == RT_TABLE_LOCAL ?
&net->ipv4.fib_table_hash[TABLE_LOCAL_INDEX] :
&net->ipv4.fib_table_hash[TABLE_MAIN_INDEX];
- return hlist_entry(ptr->first, struct fib_table, tb_hlist);
+
+ tb_hlist = rcu_dereference_rtnl(hlist_first_rcu(ptr));
+
+ return hlist_entry(tb_hlist, struct fib_table, tb_hlist);
}
static inline struct fib_table *fib_new_table(struct net *net, u32 id)
@@ -222,15 +227,19 @@ static inline struct fib_table *fib_new_table(struct net *net, u32 id)
static inline int fib_lookup(struct net *net, const struct flowi4 *flp,
struct fib_result *res)
{
- int err = -ENETUNREACH;
+ struct fib_table *tb;
+ int err;
rcu_read_lock();
- if (!fib_table_lookup(fib_get_table(net, RT_TABLE_LOCAL), flp, res,
- FIB_LOOKUP_NOREF) ||
- !fib_table_lookup(fib_get_table(net, RT_TABLE_MAIN), flp, res,
- FIB_LOOKUP_NOREF))
- err = 0;
+ for (err = 0; !err; err = -ENETUNREACH) {
+ tb = fib_get_table(net, RT_TABLE_LOCAL);
+ if (tb && !fib_table_lookup(tb, flp, res, FIB_LOOKUP_NOREF))
+ break;
+ tb = fib_get_table(net, RT_TABLE_MAIN);
+ if (tb && !fib_table_lookup(tb, flp, res, FIB_LOOKUP_NOREF))
+ break;
+ }
rcu_read_unlock();
@@ -249,28 +258,33 @@ int __fib_lookup(struct net *net, struct flowi4 *flp, struct fib_result *res);
static inline int fib_lookup(struct net *net, struct flowi4 *flp,
struct fib_result *res)
{
- if (!net->ipv4.fib_has_custom_rules) {
- int err = -ENETUNREACH;
+ struct fib_table *tb;
+ int err;
- rcu_read_lock();
+ if (net->ipv4.fib_has_custom_rules)
+ return __fib_lookup(net, flp, res);
- res->tclassid = 0;
- if ((net->ipv4.fib_local &&
- !fib_table_lookup(net->ipv4.fib_local, flp, res,
- FIB_LOOKUP_NOREF)) ||
- (net->ipv4.fib_main &&
- !fib_table_lookup(net->ipv4.fib_main, flp, res,
- FIB_LOOKUP_NOREF)) ||
- (net->ipv4.fib_default &&
- !fib_table_lookup(net->ipv4.fib_default, flp, res,
- FIB_LOOKUP_NOREF)))
- err = 0;
+ rcu_read_lock();
- rcu_read_unlock();
+ res->tclassid = 0;
- return err;
+ for (err = 0; !err; err = -ENETUNREACH) {
+ tb = rcu_dereference_rtnl(net->ipv4.fib_local);
+ if (tb && !fib_table_lookup(tb, flp, res, FIB_LOOKUP_NOREF))
+ break;
+
+ tb = rcu_dereference_rtnl(net->ipv4.fib_main);
+ if (tb && !fib_table_lookup(tb, flp, res, FIB_LOOKUP_NOREF))
+ break;
+
+ tb = rcu_dereference_rtnl(net->ipv4.fib_default);
+ if (tb && !fib_table_lookup(tb, flp, res, FIB_LOOKUP_NOREF))
+ break;
}
- return __fib_lookup(net, flp, res);
+
+ rcu_read_unlock();
+
+ return err;
}
#endif /* CONFIG_IP_MULTIPLE_TABLES */
diff --git a/include/net/netns/ipv4.h b/include/net/netns/ipv4.h
index 1b26c6c3fd7c..db1db158a00e 100644
--- a/include/net/netns/ipv4.h
+++ b/include/net/netns/ipv4.h
@@ -7,6 +7,7 @@
#include <linux/uidgid.h>
#include <net/inet_frag.h>
+#include <linux/rcupdate.h>
struct tcpm_hash_bucket;
struct ctl_table_header;
@@ -38,9 +39,9 @@ struct netns_ipv4 {
#ifdef CONFIG_IP_MULTIPLE_TABLES
struct fib_rules_ops *rules_ops;
bool fib_has_custom_rules;
- struct fib_table *fib_local;
- struct fib_table *fib_main;
- struct fib_table *fib_default;
+ struct fib_table __rcu *fib_local;
+ struct fib_table __rcu *fib_main;
+ struct fib_table __rcu *fib_default;
#endif
#ifdef CONFIG_IP_ROUTE_CLASSID
int fib_num_tclassid_users;
diff --git a/net/ipv4/fib_frontend.c b/net/ipv4/fib_frontend.c
index 57be71dd6a9e..220c4b4af4cf 100644
--- a/net/ipv4/fib_frontend.c
+++ b/net/ipv4/fib_frontend.c
@@ -89,17 +89,14 @@ struct fib_table *fib_new_table(struct net *net, u32 id)
switch (id) {
case RT_TABLE_LOCAL:
- net->ipv4.fib_local = tb;
+ rcu_assign_pointer(net->ipv4.fib_local, tb);
break;
-
case RT_TABLE_MAIN:
- net->ipv4.fib_main = tb;
+ rcu_assign_pointer(net->ipv4.fib_main, tb);
break;
-
case RT_TABLE_DEFAULT:
- net->ipv4.fib_default = tb;
+ rcu_assign_pointer(net->ipv4.fib_default, tb);
break;
-
default:
break;
}
@@ -132,13 +129,14 @@ struct fib_table *fib_get_table(struct net *net, u32 id)
static void fib_flush(struct net *net)
{
int flushed = 0;
- struct fib_table *tb;
- struct hlist_head *head;
unsigned int h;
for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
- head = &net->ipv4.fib_table_hash[h];
- hlist_for_each_entry(tb, head, tb_hlist)
+ struct hlist_head *head = &net->ipv4.fib_table_hash[h];
+ struct hlist_node *tmp;
+ struct fib_table *tb;
+
+ hlist_for_each_entry_safe(tb, tmp, head, tb_hlist)
flushed += fib_table_flush(tb);
}
@@ -665,10 +663,12 @@ static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
s_h = cb->args[0];
s_e = cb->args[1];
+ rcu_read_lock();
+
for (h = s_h; h < FIB_TABLE_HASHSZ; h++, s_e = 0) {
e = 0;
head = &net->ipv4.fib_table_hash[h];
- hlist_for_each_entry(tb, head, tb_hlist) {
+ hlist_for_each_entry_rcu(tb, head, tb_hlist) {
if (e < s_e)
goto next;
if (dumped)
@@ -682,6 +682,8 @@ static int inet_dump_fib(struct sk_buff *skb, struct netlink_callback *cb)
}
}
out:
+ rcu_read_unlock();
+
cb->args[1] = e;
cb->args[0] = h;
@@ -1117,14 +1119,34 @@ static void ip_fib_net_exit(struct net *net)
rtnl_lock();
for (i = 0; i < FIB_TABLE_HASHSZ; i++) {
- struct fib_table *tb;
- struct hlist_head *head;
+ struct hlist_head *head = &net->ipv4.fib_table_hash[i];
struct hlist_node *tmp;
+ struct fib_table *tb;
- head = &net->ipv4.fib_table_hash[i];
- hlist_for_each_entry_safe(tb, tmp, head, tb_hlist) {
- hlist_del(&tb->tb_hlist);
+ /* this is done in two passes as flushing the table could
+ * cause it to be reallocated in order to accommodate new
+ * tnodes at the root as the table shrinks.
+ */
+ hlist_for_each_entry_safe(tb, tmp, head, tb_hlist)
fib_table_flush(tb);
+
+ hlist_for_each_entry_safe(tb, tmp, head, tb_hlist) {
+#ifdef CONFIG_IP_MULTIPLE_TABLES
+ switch (tb->tb_id) {
+ case RT_TABLE_LOCAL:
+ RCU_INIT_POINTER(net->ipv4.fib_local, NULL);
+ break;
+ case RT_TABLE_MAIN:
+ RCU_INIT_POINTER(net->ipv4.fib_main, NULL);
+ break;
+ case RT_TABLE_DEFAULT:
+ RCU_INIT_POINTER(net->ipv4.fib_default, NULL);
+ break;
+ default:
+ break;
+ }
+#endif
+ hlist_del(&tb->tb_hlist);
fib_free_table(tb);
}
}
diff --git a/net/ipv4/fib_trie.c b/net/ipv4/fib_trie.c
index 2233ebf2aae8..3642b17c8726 100644
--- a/net/ipv4/fib_trie.c
+++ b/net/ipv4/fib_trie.c
@@ -193,6 +193,13 @@ static inline struct tnode *tnode_get_child_rcu(const struct tnode *tn,
return rcu_dereference_rtnl(tn->tnode[i]);
}
+static inline struct fib_table *trie_get_table(struct trie *t)
+{
+ unsigned long *tb_data = (unsigned long *)t;
+
+ return container_of(tb_data, struct fib_table, tb_data[0]);
+}
+
/* To understand this stuff, an understanding of keys and all their bits is
* necessary. Every node in the trie has a key associated with it, but not
* all of the bits in that key are significant.
@@ -1593,8 +1600,9 @@ int fib_table_flush(struct fib_table *tb)
return found;
}
-void fib_free_table(struct fib_table *tb)
+static void __trie_free_rcu(struct rcu_head *head)
{
+ struct fib_table *tb = container_of(head, struct fib_table, rcu);
#ifdef CONFIG_IP_FIB_TRIE_STATS
struct trie *t = (struct trie *)tb->tb_data;
@@ -1603,6 +1611,11 @@ void fib_free_table(struct fib_table *tb)
kfree(tb);
}
+void fib_free_table(struct fib_table *tb)
+{
+ call_rcu(&tb->rcu, __trie_free_rcu);
+}
+
static int fn_trie_dump_leaf(struct tnode *l, struct fib_table *tb,
struct sk_buff *skb, struct netlink_callback *cb)
{
@@ -1639,6 +1652,7 @@ static int fn_trie_dump_leaf(struct tnode *l, struct fib_table *tb,
return skb->len;
}
+/* rcu_read_lock needs to be hold by caller from readside */
int fib_table_dump(struct fib_table *tb, struct sk_buff *skb,
struct netlink_callback *cb)
{
@@ -1650,15 +1664,12 @@ int fib_table_dump(struct fib_table *tb, struct sk_buff *skb,
int count = cb->args[2];
t_key key = cb->args[3];
- rcu_read_lock();
-
tp = rcu_dereference_rtnl(t->trie);
while ((l = leaf_walk_rcu(&tp, key)) != NULL) {
if (fn_trie_dump_leaf(l, tb, skb, cb) < 0) {
cb->args[3] = key;
cb->args[2] = count;
- rcu_read_unlock();
return -1;
}
@@ -1673,8 +1684,6 @@ int fib_table_dump(struct fib_table *tb, struct sk_buff *skb,
break;
}
- rcu_read_unlock();
-
cb->args[3] = key;
cb->args[2] = count;