Re: [RFC patch net-next-2.6] net: convert bonding to use rx_handler

[Eric Dumazet <eric.dumazet@gmail.com>]

Le jeudi 17 février 2011 à 13:52 +0100, Jiri Pirko a écrit :
> Hello.
> 
> This is an attempt to convert bonding to use rx_handler. Result should be
> cleaner __netif_receive_skb() with much less exceptions needed. I think I
> covered all aspects, not sure though. I gave this quick smoke test on my
> testing env. Please comment, test.
> 
> Thanks!
> 
> Signed-off-by: Jiri Pirko <jpirko@redhat.com>
> ---
>  drivers/net/bonding/bond_main.c |   75 ++++++++++++++++++++-
>  include/linux/skbuff.h          |    1 +
>  net/core/dev.c                  |  144 +++++++++++---------------------------
>  3 files changed, 117 insertions(+), 103 deletions(-)
> 
> diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c


> +
> +static struct sk_buff *bond_handle_frame(struct sk_buff *skb)
> +{
> +	struct net_device *slave_dev;
> +	struct net_device *bond_dev;
> +
> +	skb = skb_share_check(skb, GFP_ATOMIC);
> +	if (unlikely(!skb))
> +		return NULL;
> +	slave_dev = skb->dev;
> +	bond_dev = slave_dev->master;

I suggest being 10%% safe here :

	bond_dev = ACCESS_ONCE(slave_dev->master);

> +	if (unlikely(!bond_dev))
> +		return skb;
> +
> +	if (bond_dev->priv_flags & IFF_MASTER_ARPMON)
> +		slave_dev->last_rx = jiffies;
> +
> +	if (bond_should_deliver_exact_match(skb, slave_dev, bond_dev)) {
> +		skb->deliver_no_wcard = 1;
> +		return skb;
> +	}
> +
> +	skb->dev = bond_dev;
> +
> +	if (bond_dev->priv_flags & IFF_MASTER_ALB &&
> +	    bond_dev->priv_flags & IFF_BRIDGE_PORT &&
> +	    skb->pkt_type == PACKET_HOST) {
> +		u16 *dest = (u16 *) eth_hdr(skb)->h_dest;
> +
> +		memcpy(dest, bond_dev->dev_addr, ETH_ALEN);
> +	}
> +
> +	netif_rx(skb);
> +	return NULL;
> +}
> +


> diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
> index 31f02d0..15b54ea 100644
> --- a/include/linux/skbuff.h
> +++ b/include/linux/skbuff.h
> @@ -325,6 +325,7 @@ struct sk_buff {
>  
>  	struct sock		*sk;
>  	struct net_device	*dev;
> +	struct net_device	*orig_dev;
>  
>  	/*
>  	 * This is the control buffer. It is free to use for every

Thats a problem. lifetime of this field is so small, I wonder if you
cant find a solution to handle this differently. Maybe a percpu
variable, or in cb[] ?