Re: [PATCH net-next 05/10] udp: refine __udp_enqueue_schedule_skb() test

[Willem de Bruijn <willemdebruijn.kernel@gmail.com>]

Eric Dumazet wrote:
> Commit 5a465a0da13e ("udp: Fix multiple wraparounds
> of sk->sk_rmem_alloc.") allowed to slightly overshoot
> sk->sk_rmem_alloc, when many cpus are trying
> to feed packets to a common UDP socket.
> 
> This patch, combined with the following one reduces
> false sharing on the victim socket under DDOS.

It also changes the behavior. There was likely a reason to allow
at least one packet if the buffer is small. Kuniyuki?

> Signed-off-by: Eric Dumazet <edumazet@google.com>
> ---
>  net/ipv4/udp.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c
> index cca41c569f37621404829e096306ba7d78ce4d43..edd846fee90ff7850356a5cb3400ce96856e5429 100644
> --- a/net/ipv4/udp.c
> +++ b/net/ipv4/udp.c
> @@ -1739,8 +1739,8 @@ int __udp_enqueue_schedule_skb(struct sock *sk, struct sk_buff *skb)
>  		if (rcvbuf > INT_MAX >> 1)
>  			goto drop;
>  
> -		/* Always allow at least one packet for small buffer. */
> -		if (rmem > rcvbuf)
> +		/* Accept the packet if queue is empty. */
> +		if (rmem)
>  			goto drop;
>  	}
>  
> -- 
> 2.51.0.384.g4c02a37b29-goog
>