conntrack: Optimize recirculations.

Cache the 'conn' context and use it when it is valid.  The cached 'conn'
context will get reset if it is not expected to be valid; the cost to do
this is negligible.  Besides being most optimal, this also handles corner
cases, such as decapsulation leading to the same tuple, as in tunnel VPN
cases.  A negative test is added to check the resetting of the cached
'conn'.

Signed-off-by: Darrell Ball <dlu998@gmail.com>
Signed-off-by: Ben Pfaff <blp@ovn.org>

4 files changed, 76 insertions, 5 deletions

diff --git a/lib/conntrack.c b/lib/conntrack.c
index b56ef06ac..ad7583cbe 100644
--- a/lib/conntrack.c
+++ b/lib/conntrack.c
@@ -1089,6 +1089,46 @@ conn_update_state_alg(struct conntrack *ct, struct dp_packet *pkt,
 }
 
 static void
+set_cached_conn(const struct nat_action_info_t *nat_action_info,
+                const struct conn_lookup_ctx *ctx, struct conn *conn,
+                struct dp_packet *pkt)
+{
+    if (OVS_LIKELY(!nat_action_info)) {
+        pkt->md.conn = conn;
+        pkt->md.reply = ctx->reply;
+        pkt->md.icmp_related = ctx->icmp_related;
+    } else {
+        pkt->md.conn = NULL;
+    }
+}
+
+static void
+process_one_fast(uint16_t zone, const uint32_t *setmark,
+                 const struct ovs_key_ct_labels *setlabel,
+                 const struct nat_action_info_t *nat_action_info,
+                 struct conn *conn, struct dp_packet *pkt)
+{
+    if (nat_action_info) {
+        handle_nat(pkt, conn, zone, pkt->md.reply, pkt->md.icmp_related);
+        pkt->md.conn = NULL;
+    }
+
+    pkt->md.ct_zone = zone;
+    ovs_mutex_lock(&conn->lock);
+    pkt->md.ct_mark = conn->mark;
+    pkt->md.ct_label = conn->label;
+    ovs_mutex_unlock(&conn->lock);
+
+    if (setmark) {
+        set_mark(pkt, conn, setmark[0], setmark[1]);
+    }
+
+    if (setlabel) {
+        set_label(pkt, conn, &setlabel[0], &setlabel[1]);
+    }
+}
+
+static void
 process_one(struct conntrack *ct, struct dp_packet *pkt,
             struct conn_lookup_ctx *ctx, uint16_t zone,
             bool force, bool commit, long long now, const uint32_t *setmark,
@@ -1188,6 +1228,8 @@ process_one(struct conntrack *ct, struct dp_packet *pkt,
     }
 
     handle_alg_ctl(ct, ctx, pkt, ct_alg_ctl, conn, now, !!nat_action_info);
+
+    set_cached_conn(nat_action_info, ctx, conn, pkt);
 }
 
 /* Sends the packets in '*pkt_batch' through the connection tracker 'ct'.  All
@@ -1215,14 +1257,21 @@ conntrack_execute(struct conntrack *ct, struct dp_packet_batch *pkt_batch,
     struct conn_lookup_ctx ctx;
 
     DP_PACKET_BATCH_FOR_EACH (i, packet, pkt_batch) {
-        if (packet->md.ct_state == CS_INVALID
-            || !conn_key_extract(ct, packet, dl_type, &ctx, zone)) {
+        struct conn *conn = packet->md.conn;
+        if (OVS_UNLIKELY(packet->md.ct_state == CS_INVALID)) {
+            write_ct_md(packet, zone, NULL, NULL, NULL);
+        } else if (conn && conn->key.zone == zone && !force
+                   && !get_alg_ctl_type(packet, tp_src, tp_dst, helper)) {
+            process_one_fast(zone, setmark, setlabel, nat_action_info,
+                             conn, packet);
+        } else if (OVS_UNLIKELY(!conn_key_extract(ct, packet, dl_type, &ctx,
+                                zone))) {
             packet->md.ct_state = CS_INVALID;
             write_ct_md(packet, zone, NULL, NULL, NULL);
-            continue;
+        } else {
+            process_one(ct, packet, &ctx, zone, force, commit, now, setmark,
+                        setlabel, nat_action_info, tp_src, tp_dst, helper);
         }
-        process_one(ct, packet, &ctx, zone, force, commit, now, setmark,
-                    setlabel, nat_action_info, tp_src, tp_dst, helper);
     }
 
     ipf_postprocess_conntrack(ct->ipf, pkt_batch, now, dl_type);
@@ -1236,6 +1285,7 @@ conntrack_clear(struct dp_packet *packet)
     /* According to pkt_metadata_init(), ct_state == 0 is enough to make all of
      * the conntrack fields invalid. */
     packet->md.ct_state = 0;
+    pkt_metadata_init_conn(&packet->md);
 }
 
 static void
diff --git a/lib/netdev.c b/lib/netdev.c
index b1976d365..af8f8560d 100644
--- a/lib/netdev.c
+++ b/lib/netdev.c
@@ -837,6 +837,7 @@ netdev_pop_header(struct netdev *netdev, struct dp_packet_batch *batch)
              * interpretation in the further packet processing when
              * recirculated.*/
             dp_packet_reset_offload(packet);
+            pkt_metadata_init_conn(&packet->md);
             dp_packet_batch_refill(batch, packet, i);
         }
     }
diff --git a/lib/packets.c b/lib/packets.c
index 12053df57..fba69eed2 100644
--- a/lib/packets.c
+++ b/lib/packets.c
@@ -391,6 +391,8 @@ push_mpls(struct dp_packet *packet, ovs_be16 ethtype, ovs_be32 lse)
     header = dp_packet_resize_l2_5(packet, MPLS_HLEN);
     memmove(header, header + MPLS_HLEN, len);
     memcpy(header + len, &lse, sizeof lse);
+
+    pkt_metadata_init_conn(&packet->md);
 }
 
 /* If 'packet' is an MPLS packet, removes its outermost MPLS label stack entry.
@@ -983,6 +985,8 @@ packet_set_ipv4_addr(struct dp_packet *packet,
     ovs_be32 old_addr = get_16aligned_be32(addr);
     size_t l4_size = dp_packet_l4_size(packet);
 
+    pkt_metadata_init_conn(&packet->md);
+
     if (nh->ip_proto == IPPROTO_TCP && l4_size >= TCP_HEADER_LEN) {
         struct tcp_header *th = dp_packet_l4(packet);
 
@@ -1122,6 +1126,7 @@ packet_set_ipv6_addr(struct dp_packet *packet, uint8_t proto,
         packet_update_csum128(packet, proto, addr, new_addr);
     }
     memcpy(addr, new_addr, sizeof(ovs_be32[4]));
+    pkt_metadata_init_conn(&packet->md);
 }
 
 static void
@@ -1223,6 +1228,7 @@ packet_set_tcp_port(struct dp_packet *packet, ovs_be16 src, ovs_be16 dst)
 
     packet_set_port(&th->tcp_src, src, &th->tcp_csum);
     packet_set_port(&th->tcp_dst, dst, &th->tcp_csum);
+    pkt_metadata_init_conn(&packet->md);
 }
 
 /* Sets the UDP source and destination port ('src' and 'dst' respectively) of
@@ -1244,6 +1250,7 @@ packet_set_udp_port(struct dp_packet *packet, ovs_be16 src, ovs_be16 dst)
         uh->udp_src = src;
         uh->udp_dst = dst;
     }
+    pkt_metadata_init_conn(&packet->md);
 }
 
 /* Sets the SCTP source and destination port ('src' and 'dst' respectively) of
@@ -1265,6 +1272,7 @@ packet_set_sctp_port(struct dp_packet *packet, ovs_be16 src, ovs_be16 dst)
 
     new_csum = crc32c((void *)sh, tp_len);
     put_16aligned_be32(&sh->sctp_csum, old_csum ^ old_correct_csum ^ new_csum);
+    pkt_metadata_init_conn(&packet->md);
 }
 
 /* Sets the ICMP type and code of the ICMP header contained in 'packet'.
@@ -1283,6 +1291,7 @@ packet_set_icmp(struct dp_packet *packet, uint8_t type, uint8_t code)
 
         ih->icmp_csum = recalc_csum16(ih->icmp_csum, orig_tc, new_tc);
     }
+    pkt_metadata_init_conn(&packet->md);
 }
 
 /* Sets the IGMP type to IGMP_HOST_MEMBERSHIP_QUERY and populates the
diff --git a/lib/packets.h b/lib/packets.h
index c44009892..c78defb89 100644
--- a/lib/packets.h
+++ b/lib/packets.h
@@ -35,6 +35,7 @@
 #include "timeval.h"
 
 struct dp_packet;
+struct conn;
 struct ds;
 
 /* Purely internal to OVS userspace. These flags should never be exposed to
@@ -108,6 +109,9 @@ PADDED_MEMBERS_CACHELINE_MARKER(CACHE_LINE_SIZE, cacheline0,
     uint32_t ct_mark;           /* Connection mark. */
     ovs_u128 ct_label;          /* Connection label. */
     union flow_in_port in_port; /* Input port. */
+    struct conn *conn;          /* Cached conntrack connection. */
+    bool reply;                 /* True if reply direction. */
+    bool icmp_related;          /* True if ICMP related. */
 );
 
 PADDED_MEMBERS_CACHELINE_MARKER(CACHE_LINE_SIZE, cacheline1,
@@ -140,6 +144,12 @@ pkt_metadata_init_tnl(struct pkt_metadata *md)
 }
 
 static inline void
+pkt_metadata_init_conn(struct pkt_metadata *md)
+{
+    md->conn = NULL;
+}
+
+static inline void
 pkt_metadata_init(struct pkt_metadata *md, odp_port_t port)
 {
     /* This is called for every packet in userspace datapath and affects
@@ -157,6 +167,7 @@ pkt_metadata_init(struct pkt_metadata *md, odp_port_t port)
     md->tunnel.ip_dst = 0;
     md->tunnel.ipv6_dst = in6addr_any;
     md->in_port.odp_port = port;
+    md->conn = NULL;
 }
 
 /* This function prefetches the cachelines touched by pkt_metadata_init()