RFC: VETH patch to zero timestamp.

RFC: VETH patch to zero timestamp.

[Ben Greear]

The patch below helps when using VETH and bridge(s)
This makes sure that the pkt timestamp is properly (re)calculated on
receiving the packet on the peer veth device.

I'm sure this patch is white-space damaged.  If this looks
useful, I'll generate a clean patch and send as attachment.

Signed-off-by:  Ben Greear<greearb@candelatech.com>


diff --git a/drivers/net/veth.c b/drivers/net/veth.c
index 1097c72..e9136af 100644
--- a/drivers/net/veth.c
+++ b/drivers/net/veth.c
@@ -170,7 +173,12 @@ static int veth_xmit(struct sk_buff *skb, struct net_device *dev)

         if (skb->len > (rcv->mtu + MTU_PAD))
                 goto rx_drop;
-
+
+       /* Zero out the time-stamp so that receiving code is forced
+        * to recalculate it.
+        */
+       skb->tstamp.tv64 = 0;
+
         skb->pkt_type = PACKET_HOST;
         skb->protocol = eth_type_trans(skb, rcv);
         if (dev->features & NETIF_F_NO_CSUM)


Thanks,
Ben

-- 
Ben Greear <greearb@candelatech.com>
Candela Technologies Inc  http://www.candelatech.com

Re: RFC: VETH patch to zero timestamp.

[Eric W. Biederman]

Ben Greear <greearb@candelatech.com> writes:

> The patch below helps when using VETH and bridge(s)
> This makes sure that the pkt timestamp is properly (re)calculated on
> receiving the packet on the peer veth device.
>
> I'm sure this patch is white-space damaged.  If this looks
> useful, I'll generate a clean patch and send as attachment.

We reset everything else that needs to be reset.  So we might as well
reset the timestamps as well.  Unless there is a big performance
impact by doing so.

> Signed-off-by:  Ben Greear<greearb@candelatech.com>
>
>
> diff --git a/drivers/net/veth.c b/drivers/net/veth.c
> index 1097c72..e9136af 100644
> --- a/drivers/net/veth.c
> +++ b/drivers/net/veth.c
> @@ -170,7 +173,12 @@ static int veth_xmit(struct sk_buff *skb, struct net_device *dev)
>
>         if (skb->len > (rcv->mtu + MTU_PAD))
>                 goto rx_drop;
> -
> +
> +       /* Zero out the time-stamp so that receiving code is forced
> +        * to recalculate it.
> +        */
> +       skb->tstamp.tv64 = 0;
> +
>         skb->pkt_type = PACKET_HOST;
>         skb->protocol = eth_type_trans(skb, rcv);
>         if (dev->features & NETIF_F_NO_CSUM)
>
>
> Thanks,
> Ben
>
> -- 
> Ben Greear <greearb@candelatech.com>
> Candela Technologies Inc  http://www.candelatech.com