Re: [PATCH net-next] net: keep original skb which only needs header checking during software GSO

[Eric Dumazet]

On Fri, 2014-09-19 at 14:38 +0800, Jason Wang wrote:
> Commit ce93718fb7cdbc064c3000ff59e4d3200bdfa744 ("net: Don't keep
> around original SKB when we software segment GSO frames") frees the
> original skb after software GSO even for dodgy gso skbs. This breaks
> the stream throughput from untrusted sources, since only header
> checking was done during software GSO instead of a true
> segmentation. This patch fixes this by freeing the original gso skb
> only when it was really segmented by software.
> 
> Fixes ce93718fb7cdbc064c3000ff59e4d3200bdfa744 ("net: Don't keep
> around original SKB when we software segment GSO frames.")
> 
> CC: David S. Miller <davem@xxxxxxxxxxxxx>
> Signed-off-by: Jason Wang <jasowang@xxxxxxxxxx>
> ---
>  net/core/dev.c | 6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/net/core/dev.c b/net/core/dev.c
> index e916ba8..b7a0e1d 100644
> --- a/net/core/dev.c
> +++ b/net/core/dev.c
> @@ -2694,10 +2694,12 @@ struct sk_buff *validate_xmit_skb(struct sk_buff *skb, struct net_device *dev)
>  		struct sk_buff *segs;
>  
>  		segs = skb_gso_segment(skb, features);
> -		kfree_skb(skb);
>  		if (IS_ERR(segs))
>  			
> -		skb = segs;
> +		else if (segs) {
> +			kfree_skb(skb);
> +			skb = segs;
> +		}
>  	} else {
>  		if (skb_needs_linearize(skb, features) &&
>  		    __skb_linearize(skb))

Good catch !

While we are at it, could you use consume_skb() instead of kfree_skb(),
and add missing {} (CodingStyle) ?

if (IS_ERR(segs)) {
	segs = NULL;
} else if (segs) {
	consume_skb(skb);
	skb = segs;
}


Thanks !


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/