This function is used for every packet. siphash_4u64 is
about 25% faster compared to buffer+memset+siphash_unaligned as
reported by perf-diff.
Signed-off-by: Florian Westphal <fw@xxxxxxxxx>
---
net/netfilter/nf_conntrack_core.c | 31 +++++++++++++------------------
1 file changed, 13 insertions(+), 18 deletions(-)
diff --git a/net/netfilter/nf_conntrack_core.c b/net/netfilter/nf_conntrack_core.c
index f97bda06d2a9..48c0c146cef1 100644
--- a/net/netfilter/nf_conntrack_core.c
+++ b/net/netfilter/nf_conntrack_core.c
@@ -211,28 +211,23 @@ static u32 hash_conntrack_raw(const struct nf_conntrack_tuple *tuple,
unsigned int zoneid,
const struct net *net)
{
- struct {
- struct nf_conntrack_man src;
- union nf_inet_addr dst_addr;
- unsigned int zone;
- u32 net_mix;
- u16 dport;
- u16 proto;
- } __aligned(SIPHASH_ALIGNMENT) combined;
+ u64 a, b, c, d;
get_random_once(&nf_conntrack_hash_rnd, sizeof(nf_conntrack_hash_rnd));
- memset(&combined, 0, sizeof(combined));
-
/* The direction must be ignored, so handle usable members manually. */
- combined.src = tuple->src;
- combined.dst_addr = tuple->dst.u3;
- combined.zone = zoneid;
- combined.net_mix = net_hash_mix(net);
- combined.dport = (__force __u16)tuple->dst.u.all;
- combined.proto = tuple->dst.protonum;
-
- return (u32)siphash(&combined, sizeof(combined), &nf_conntrack_hash_rnd);
+ a = (u64)tuple->src.u3.all[0] << 32 | tuple->src.u3.all[1];
+ b = (u64)tuple->src.u3.all[2] << 32 | tuple->src.u3.all[3];
+
+ c = (u64)tuple->dst.u3.all[0] << 32 | tuple->dst.u3.all[1];
+ d = (u64)tuple->dst.u3.all[1] << 32 | tuple->dst.u3.all[2];
+
+ a ^= (__force __u64)tuple->dst.u.all;
+ b += tuple->dst.protonum;
+ c ^= net_hash_mix(net);
+ d += zoneid;
+
+ return (u32)siphash_4u64(a, b, c, d, &nf_conntrack_hash_rnd);
}
static u32 scale_hash(u32 hash)
--
2.37.3
