diff mbox series

[net-next,1/1] l2tp: fix up inconsistent rx/tx statistics

Message ID 20200918102321.7036-1-tparkin@katalix.com
State Accepted
Delegated to: David Miller
Headers show
Series [net-next,1/1] l2tp: fix up inconsistent rx/tx statistics | expand

Commit Message

Tom Parkin Sept. 18, 2020, 10:23 a.m. UTC 
Historically L2TP core statistics count the L2TP header in the
per-session and per-tunnel byte counts tracked for transmission and
receipt.

Now that l2tp_xmit_skb updates tx stats, it is necessary for
l2tp_xmit_core to pass out the length of the transmitted packet so that
the statistics can be updated correctly.

Signed-off-by: Tom Parkin <tparkin@katalix.com>
---
 net/l2tp/l2tp_core.c | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)

Comments

David Miller Sept. 18, 2020, 9:37 p.m. UTC | #1 
From: Tom Parkin <tparkin@katalix.com>
Date: Fri, 18 Sep 2020 11:23:21 +0100

> Historically L2TP core statistics count the L2TP header in the
> per-session and per-tunnel byte counts tracked for transmission and
> receipt.
> 
> Now that l2tp_xmit_skb updates tx stats, it is necessary for
> l2tp_xmit_core to pass out the length of the transmitted packet so that
> the statistics can be updated correctly.
> 
> Signed-off-by: Tom Parkin <tparkin@katalix.com>

Applied.
diff mbox series

Patch

diff --git a/net/l2tp/l2tp_core.c b/net/l2tp/l2tp_core.c
index 7de05be4fc33..7be5103ff2a8 100644
--- a/net/l2tp/l2tp_core.c
+++ b/net/l2tp/l2tp_core.c
@@ -1006,7 +1006,7 @@  static int l2tp_xmit_queue(struct l2tp_tunnel *tunnel, struct sk_buff *skb, stru
 	return err >= 0 ? NET_XMIT_SUCCESS : NET_XMIT_DROP;
 }
 
-static int l2tp_xmit_core(struct l2tp_session *session, struct sk_buff *skb)
+static int l2tp_xmit_core(struct l2tp_session *session, struct sk_buff *skb, unsigned int *len)
 {
 	struct l2tp_tunnel *tunnel = session->tunnel;
 	unsigned int data_len = skb->len;
@@ -1054,6 +1054,11 @@  static int l2tp_xmit_core(struct l2tp_session *session, struct sk_buff *skb)
 		goto out_unlock;
 	}
 
+	/* Report transmitted length before we add encap header, which keeps
+	 * statistics consistent for both UDP and IP encap tx/rx paths.
+	 */
+	*len = skb->len;
+
 	inet = inet_sk(sk);
 	switch (tunnel->encap) {
 	case L2TP_ENCAPTYPE_UDP:
@@ -1095,10 +1100,10 @@  static int l2tp_xmit_core(struct l2tp_session *session, struct sk_buff *skb)
  */
 int l2tp_xmit_skb(struct l2tp_session *session, struct sk_buff *skb)
 {
-	unsigned int len = skb->len;
+	unsigned int len = 0;
 	int ret;
 
-	ret = l2tp_xmit_core(session, skb);
+	ret = l2tp_xmit_core(session, skb, &len);
 	if (ret == NET_XMIT_SUCCESS) {
 		atomic_long_inc(&session->tunnel->stats.tx_packets);
 		atomic_long_add(len, &session->tunnel->stats.tx_bytes);