NET_RX_DROP question

NET_RX_DROP question

[Joakim Tjernlund]

In our ethernet drivers RX path we have:
  if (netif_receive_skb(skb) == NET_RX_DROP)
     priv->stats.rx_dropped++;

Now we can see dropped counter counting dropped pkgs but we don't see
any corrupt pkgs etc. Is NET_RX_DROP really meant to be used like this?
It gives the impression that there is something wrong with the driver.

 Jocke

Re: NET_RX_DROP question

[Joakim Tjernlund]

On Thu, 2022-11-03 at 15:08 +0100, Joakim Tjernlund wrote:
> In our ethernet drivers RX path we have:
>   if (netif_receive_skb(skb) == NET_RX_DROP)
>      priv->stats.rx_dropped++;
> 
> Now we can see dropped counter counting dropped pkgs but we don't see
> any corrupt pkgs etc. Is NET_RX_DROP really meant to be used like this?
> It gives the impression that there is something wrong with the driver.
> 
>  Jocke

Looking into drivers/net/ethernet is seems just a few checks if (netif_receive_skb(skb) == NET_RX_DROP),
the majority does not which makes me this I should just drop this check.

Confirmation more than welcome.

 Jocke

Re: NET_RX_DROP question

[Francois Romieu]

Joakim Tjernlund <Joakim.Tjernlund@infinera.com> :
[...]
> Looking into drivers/net/ethernet is seems just a few checks if (netif_receive_skb(skb) == NET_RX_DROP),
> the majority does not which makes me this I should just drop this check.
> 
> Confirmation more than welcome.

The driver should not maintain stats for packets dropped by
the core network. They are already accounted for in core stats
(include/linux/netdevice.c::struct net_device.core_stats).

-- 
Ueimor