tcp: increment sk_drops for dropped rx packets
authorEric Dumazet <edumazet@google.com>
Fri, 1 Apr 2016 15:52:19 +0000 (08:52 -0700)
committerDavid S. Miller <davem@davemloft.net>
Tue, 5 Apr 2016 02:11:20 +0000 (22:11 -0400)
Now ss can report sk_drops, we can instruct TCP to increment
this per socket counter when it drops an incoming frame, to refine
monitoring and debugging.

Following patch takes care of listeners drops.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
include/net/sock.h
net/ipv4/tcp_input.c
net/ipv4/tcp_ipv4.c
net/ipv6/tcp_ipv6.c

index 7ad73db..310c436 100644 (file)
@@ -2012,6 +2012,13 @@ sock_skb_set_dropcount(const struct sock *sk, struct sk_buff *skb)
        SOCK_SKB_CB(skb)->dropcount = atomic_read(&sk->sk_drops);
 }
 
+static inline void sk_drops_add(struct sock *sk, const struct sk_buff *skb)
+{
+       int segs = max_t(u16, 1, skb_shinfo(skb)->gso_segs);
+
+       atomic_add(segs, &sk->sk_drops);
+}
+
 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
                           struct sk_buff *skb);
 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
index a26e2d2..0ffcd07 100644 (file)
@@ -4307,6 +4307,12 @@ static bool tcp_try_coalesce(struct sock *sk,
        return true;
 }
 
+static void tcp_drop(struct sock *sk, struct sk_buff *skb)
+{
+       sk_drops_add(sk, skb);
+       __kfree_skb(skb);
+}
+
 /* This one checks to see if we can put data from the
  * out_of_order queue into the receive_queue.
  */
@@ -4331,7 +4337,7 @@ static void tcp_ofo_queue(struct sock *sk)
                __skb_unlink(skb, &tp->out_of_order_queue);
                if (!after(TCP_SKB_CB(skb)->end_seq, tp->rcv_nxt)) {
                        SOCK_DEBUG(sk, "ofo packet was already received\n");
-                       __kfree_skb(skb);
+                       tcp_drop(sk, skb);
                        continue;
                }
                SOCK_DEBUG(sk, "ofo requeuing : rcv_next %X seq %X - %X\n",
@@ -4383,7 +4389,7 @@ static void tcp_data_queue_ofo(struct sock *sk, struct sk_buff *skb)
 
        if (unlikely(tcp_try_rmem_schedule(sk, skb, skb->truesize))) {
                NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPOFODROP);
-               __kfree_skb(skb);
+               tcp_drop(sk, skb);
                return;
        }
 
@@ -4447,7 +4453,7 @@ static void tcp_data_queue_ofo(struct sock *sk, struct sk_buff *skb)
                if (!after(end_seq, TCP_SKB_CB(skb1)->end_seq)) {
                        /* All the bits are present. Drop. */
                        NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPOFOMERGE);
-                       __kfree_skb(skb);
+                       tcp_drop(sk, skb);
                        skb = NULL;
                        tcp_dsack_set(sk, seq, end_seq);
                        goto add_sack;
@@ -4486,7 +4492,7 @@ static void tcp_data_queue_ofo(struct sock *sk, struct sk_buff *skb)
                tcp_dsack_extend(sk, TCP_SKB_CB(skb1)->seq,
                                 TCP_SKB_CB(skb1)->end_seq);
                NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_TCPOFOMERGE);
-               __kfree_skb(skb1);
+               tcp_drop(sk, skb1);
        }
 
 add_sack:
@@ -4569,12 +4575,13 @@ err:
 static void tcp_data_queue(struct sock *sk, struct sk_buff *skb)
 {
        struct tcp_sock *tp = tcp_sk(sk);
-       int eaten = -1;
        bool fragstolen = false;
+       int eaten = -1;
 
-       if (TCP_SKB_CB(skb)->seq == TCP_SKB_CB(skb)->end_seq)
-               goto drop;
-
+       if (TCP_SKB_CB(skb)->seq == TCP_SKB_CB(skb)->end_seq) {
+               __kfree_skb(skb);
+               return;
+       }
        skb_dst_drop(skb);
        __skb_pull(skb, tcp_hdr(skb)->doff * 4);
 
@@ -4656,7 +4663,7 @@ out_of_window:
                tcp_enter_quickack_mode(sk);
                inet_csk_schedule_ack(sk);
 drop:
-               __kfree_skb(skb);
+               tcp_drop(sk, skb);
                return;
        }
 
@@ -5233,7 +5240,7 @@ syn_challenge:
        return true;
 
 discard:
-       __kfree_skb(skb);
+       tcp_drop(sk, skb);
        return false;
 }
 
@@ -5451,7 +5458,7 @@ csum_error:
        TCP_INC_STATS_BH(sock_net(sk), TCP_MIB_INERRS);
 
 discard:
-       __kfree_skb(skb);
+       tcp_drop(sk, skb);
 }
 EXPORT_SYMBOL(tcp_rcv_established);
 
@@ -5682,7 +5689,7 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,
                                                  TCP_DELACK_MAX, TCP_RTO_MAX);
 
 discard:
-                       __kfree_skb(skb);
+                       tcp_drop(sk, skb);
                        return 0;
                } else {
                        tcp_send_ack(sk);
@@ -6043,7 +6050,7 @@ int tcp_rcv_state_process(struct sock *sk, struct sk_buff *skb)
 
        if (!queued) {
 discard:
-               __kfree_skb(skb);
+               tcp_drop(sk, skb);
        }
        return 0;
 }
index e5f924b..059a98f 100644 (file)
@@ -1694,6 +1694,7 @@ discard_it:
        return 0;
 
 discard_and_relse:
+       sk_drops_add(sk, skb);
        if (refcounted)
                sock_put(sk);
        goto discard_it;
index f0422e7..5fa8fea 100644 (file)
@@ -1484,6 +1484,7 @@ discard_it:
        return 0;
 
 discard_and_relse:
+       sk_drops_add(sk, skb);
        if (refcounted)
                sock_put(sk);
        goto discard_it;