diff mbox

[ovs-dev,patch_v8,2/9] dpdk: Remove batch sorting in userspace conntrack.

Message ID 1496166573-90065-3-git-send-email-dlu998@gmail.com
State Accepted
Headers show

Commit Message

Darrell Ball May 30, 2017, 5:49 p.m. UTC 
Packet batch sorting is removed for three reasons:

1) The following patches for NAT change the locking
    marshalling so batching loses benefit.

2) For real mixtures of flows either in hypervisors
   or gateways, the batch sorting won't provide benefit
   and will just be a tax.

3) Code clarity.

Signed-off-by: Darrell Ball <dlu998@gmail.com>
Acked-by: Flavio Leitner <fbl@sysclose.org>
Acked-by: Daniele Di Proietto <diproiettod@ovn.org>
---
 lib/conntrack.c | 58 +++++++++++----------------------------------------------
 1 file changed, 11 insertions(+), 47 deletions(-)
diff mbox

Patch

diff --git a/lib/conntrack.c b/lib/conntrack.c
index ecc6177..d339d72 100644
--- a/lib/conntrack.c
+++ b/lib/conntrack.c
@@ -318,22 +318,9 @@  conntrack_execute(struct conntrack *ct, struct dp_packet_batch *pkt_batch,
 {
     struct dp_packet **pkts = pkt_batch->packets;
     size_t cnt = pkt_batch->count;
-#if !defined(__CHECKER__) && !defined(_WIN32)
-    const size_t KEY_ARRAY_SIZE = cnt;
-#else
-    enum { KEY_ARRAY_SIZE = NETDEV_MAX_BURST };
-#endif
-    struct conn_lookup_ctx ctxs[KEY_ARRAY_SIZE];
-    int8_t bucket_list[CONNTRACK_BUCKETS];
-    struct {
-        unsigned bucket;
-        unsigned long maps;
-    } arr[KEY_ARRAY_SIZE];
     long long now = time_msec();
+    struct conn_lookup_ctx ctx;
     size_t i = 0;
-    uint8_t arrcnt = 0;
-
-    BUILD_ASSERT_DECL(sizeof arr[0].maps * CHAR_BIT >= NETDEV_MAX_BURST);
 
     if (helper) {
         static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(5, 5);
@@ -342,48 +329,25 @@  conntrack_execute(struct conntrack *ct, struct dp_packet_batch *pkt_batch,
         /* Continue without the helper */
     }
 
-    memset(bucket_list, INT8_C(-1), sizeof bucket_list);
     for (i = 0; i < cnt; i++) {
-        unsigned bucket;
 
-        if (!conn_key_extract(ct, pkts[i], dl_type, &ctxs[i], zone)) {
+        if (!conn_key_extract(ct, pkts[i], dl_type, &ctx, zone)) {
             write_ct_md(pkts[i], CS_INVALID, zone, NULL, NULL);
             continue;
         }
 
-        bucket = hash_to_bucket(ctxs[i].hash);
-        if (bucket_list[bucket] == INT8_C(-1)) {
-            bucket_list[bucket] = arrcnt;
-
-            arr[arrcnt].maps = 0;
-            ULLONG_SET1(arr[arrcnt].maps, i);
-            arr[arrcnt++].bucket = bucket;
-        } else {
-            ULLONG_SET1(arr[bucket_list[bucket]].maps, i);
-        }
-    }
-
-    for (i = 0; i < arrcnt; i++) {
-        struct conntrack_bucket *ctb = &ct->buckets[arr[i].bucket];
-        size_t j;
-
+        struct conntrack_bucket *ctb = &ct->buckets[i];
         ct_lock_lock(&ctb->lock);
+        conn_key_lookup(ctb, &ctx, now);
+        struct conn *conn = process_one(ct, pkts[i], &ctx, zone,
+                                        force, commit, now);
 
-        ULLONG_FOR_EACH_1(j, arr[i].maps) {
-            struct conn *conn;
-
-            conn_key_lookup(ctb, &ctxs[j], now);
-
-            conn = process_one(ct, pkts[j], &ctxs[j], zone, force, commit,
-                               now);
-
-            if (conn && setmark) {
-                set_mark(pkts[j], conn, setmark[0], setmark[1]);
-            }
+        if (conn && setmark) {
+            set_mark(pkts[i], conn, setmark[0], setmark[1]);
+        }
 
-            if (conn && setlabel) {
-                set_label(pkts[j], conn, &setlabel[0], &setlabel[1]);
-            }
+        if (conn && setlabel) {
+                set_label(pkts[i], conn, &setlabel[0], &setlabel[1]);
         }
         ct_lock_unlock(&ctb->lock);
     }