ipv4: save cpu cycles from check_leaf()
Compiler is not smart enough to avoid double BSWAP instructions in
ntohl(inet_make_mask(plen)).
Lets cache this value in struct leaf_info, (fill a hole on 64bit arches)
With route cache disabled, this saves ~2% of cpu in udpflood bench on
x86_64 machine.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
diff --git a/net/ipv4/fib_trie.c b/net/ipv4/fib_trie.c
index 58c25ea..de9e297 100644
--- a/net/ipv4/fib_trie.c
+++ b/net/ipv4/fib_trie.c
@@ -110,9 +110,10 @@
struct leaf_info {
struct hlist_node hlist;
- struct rcu_head rcu;
int plen;
+ u32 mask_plen; /* ntohl(inet_make_mask(plen)) */
struct list_head falh;
+ struct rcu_head rcu;
};
struct tnode {
@@ -451,6 +452,7 @@
struct leaf_info *li = kmalloc(sizeof(struct leaf_info), GFP_KERNEL);
if (li) {
li->plen = plen;
+ li->mask_plen = ntohl(inet_make_mask(plen));
INIT_LIST_HEAD(&li->falh);
}
return li;
@@ -1359,10 +1361,8 @@
hlist_for_each_entry_rcu(li, node, hhead, hlist) {
struct fib_alias *fa;
- int plen = li->plen;
- __be32 mask = inet_make_mask(plen);
- if (l->key != (key & ntohl(mask)))
+ if (l->key != (key & li->mask_plen))
continue;
list_for_each_entry_rcu(fa, &li->falh, fa_list) {
@@ -1394,7 +1394,7 @@
#ifdef CONFIG_IP_FIB_TRIE_STATS
t->stats.semantic_match_passed++;
#endif
- res->prefixlen = plen;
+ res->prefixlen = li->plen;
res->nh_sel = nhsel;
res->type = fa->fa_type;
res->scope = fa->fa_info->fib_scope;
@@ -1402,7 +1402,7 @@
res->table = tb;
res->fa_head = &li->falh;
if (!(fib_flags & FIB_LOOKUP_NOREF))
- atomic_inc(&res->fi->fib_clntref);
+ atomic_inc(&fi->fib_clntref);
return 0;
}
}