On 01/07/13 21:40, David Miller wrote:

From: Duan Fugang-B38611 <redacted>
Date: Fri, 28 Jun 2013 02:11:30 +0000

quoted

On 06/28/13 02:25, Jim Baxter wrote:

quoted

diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
index ed6180e..05a5f76 100644
--- a/drivers/net/ethernet/freescale/fec_main.c
+++ b/drivers/net/ethernet/freescale/fec_main.c

@@ -738,6 +738,7 @@ fec_enet_tx(struct net_device *ndev)
 				ndev->stats.tx_carrier_errors++;
 		} else {
 			ndev->stats.tx_packets++;
+			ndev->stats.tx_bytes += bdp->cbd_datlen;
 		}
 
 		if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS) &&
--

You cannot stat. tx_bytes in here, since stat.tx_bytes indicate that all bytes sent by MAC regardless whether there have error packets or not.
You must add the stat. at xmit function as below:

I completely disagree.

tx_bytes indicates what actually made it to the wires, so Jim's original
patch is the correct one.

Every single driver I have ever written, always increments tx_bytes in
the transmit completion handler, not when the device layer gives the
packet to the driver.

If this should be the first version of the patch, can I cancel the
second version and set the first version to new in patchwork, or is it
best to submit a third version?

Thank you,
Jim