fib_trie: Use empty_children instead of counting empty nodes in stats collection
authorAlexander Duyck <alexander.h.duyck@redhat.com>
Thu, 22 Jan 2015 23:51:33 +0000 (15:51 -0800)
committerDavid S. Miller <davem@davemloft.net>
Sun, 25 Jan 2015 22:47:16 +0000 (14:47 -0800)
It doesn't make much sense to count the pointers ourselves when
empty_children already has a count for the number of NULL pointers stored
in the tnode.  As such save ourselves the cycles and just use
empty_children.

Signed-off-by: Alexander Duyck <alexander.h.duyck@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
net/ipv4/fib_trie.c

index f874e1811eabaca6ba6ad06bae0c7bbb19ec4f90..90654bb64e21f81c9cf56ca453043a6a2a49f59d 100644 (file)
@@ -1954,16 +1954,10 @@ static void trie_collect_stats(struct trie *t, struct trie_stat *s)
                        hlist_for_each_entry_rcu(li, &n->list, hlist)
                                ++s->prefixes;
                } else {
-                       unsigned long i;
-
                        s->tnodes++;
                        if (n->bits < MAX_STAT_DEPTH)
                                s->nodesizes[n->bits]++;
-
-                       for (i = tnode_child_length(n); i--;) {
-                               if (!rcu_access_pointer(n->child[i]))
-                                       s->nullpointers++;
-                       }
+                       s->nullpointers += n->empty_children;
                }
        }
        rcu_read_unlock();