bpf: fix memory leak in lpm_trie map_free callback function

There is a memory leak happening in lpm_trie map_free callback
function trie_free. The trie structure itself does not get freed.

Also, trie_free function did not do synchronize_rcu before freeing
various data structures. This is incorrect as some rcu_read_lock
region(s) for lookup, update, delete or get_next_key may not complete yet.
The fix is to add synchronize_rcu in the beginning of trie_free.
The useless spin_lock is removed from this function as well.

Fixes: b95a5c4 ("bpf: add a longest prefix match trie map implementation")
Reported-by: Mathieu Malaterre <malat@debian.org>
Reported-by: Alexei Starovoitov <ast@kernel.org>
Tested-by: Mathieu Malaterre <malat@debian.org>
Signed-off-by: Yonghong Song <yhs@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

kernel/bpf/lpm_trie.c

@@ -555,7 +555,10 @@ static void trie_free(struct bpf_map *map)
 	struct lpm_trie_node __rcu **slot;
 	struct lpm_trie_node *node;
 
-	raw_spin_lock(&trie->lock);
+	/* Wait for outstanding programs to complete
+	 * update/lookup/delete/get_next_key and free the trie.
+	 */
+	synchronize_rcu();
 
 	/* Always start at the root and walk down to a node that has no
 	 * children. Then free that node, nullify its reference in the parent
@@ -569,7 +572,7 @@ static void trie_free(struct bpf_map *map)
 			node = rcu_dereference_protected(*slot,
 					lockdep_is_held(&trie->lock));
 			if (!node)
-				goto unlock;
+				goto out;
 
 			if (rcu_access_pointer(node->child[0])) {
 				slot = &node->child[0];
@@ -587,8 +590,8 @@ static void trie_free(struct bpf_map *map)
 		}
 	}
 
-unlock:
-	raw_spin_unlock(&trie->lock);
+out:
+	kfree(trie);
 }
 
 static int trie_get_next_key(struct bpf_map *map, void *_key, void *_next_key)