net: trans_start cleanups

Now that core network takes care of trans_start updates, dont do it
in drivers themselves, if possible. Drivers can avoid one cache miss
(on dev->trans_start) in their start_xmit() handler.

Exceptions are NETIF_F_LLTX drivers

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

1 files changed, 0 insertions, 3 deletions

diff --git a/drivers/net/3c523.c b/drivers/net/3c523.c
index 55d219e585f..a7b0e5e43a5 100644
--- a/drivers/net/3c523.c
+++ b/drivers/net/3c523.c
@@ -1152,7 +1152,6 @@ static netdev_tx_t elmc_send_packet(struct sk_buff *skb, struct net_device *dev)
                 p->scb->cmd = CUC_START;
                 p->xmit_cmds[0]->cmd_status = 0;
                         elmc_attn586();
-                dev->trans_start = jiffies;
                 if (!i) {
                         dev_kfree_skb(skb);
                 }
@@ -1176,7 +1175,6 @@ static netdev_tx_t elmc_send_packet(struct sk_buff *skb, struct net_device *dev)
         p->xmit_cmds[0]->cmd_status = p->nop_cmds[next_nop]->cmd_status = 0;
 
         p->nop_cmds[p->nop_point]->cmd_link = make16((p->xmit_cmds[0]));
-        dev->trans_start = jiffies;
         p->nop_point = next_nop;
         dev_kfree_skb(skb);
 #endif
@@ -1190,7 +1188,6 @@ static netdev_tx_t elmc_send_packet(struct sk_buff *skb, struct net_device *dev)
             = make16((p->nop_cmds[next_nop]));
         p->nop_cmds[next_nop]->cmd_status = 0;
                 p->nop_cmds[p->xmit_count]->cmd_link = make16((p->xmit_cmds[p->xmit_count]));
-        dev->trans_start = jiffies;
         p->xmit_count = next_nop;
         if (p->xmit_count != p->xmit_last)
                 netif_wake_queue(dev);