Re: [PATCH net-next 1/1] af-packet: fix - avoid reading stale data

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

Le mardi 12 juillet 2011 à 21:42 -0400, Chetan Loke a écrit :
> Currently we flush tp_status and then flush the remainder of the header+payload.
> tp_status should be flushed in the end to avoid stale data being read by user-space.
> 
> 
> Signed-off-by: Chetan Loke <loke.chetan@gmail.com>
> ---
>  net/packet/af_packet.c |    4 ++--
>  1 files changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/net/packet/af_packet.c b/net/packet/af_packet.c
> index d2294ad..f7be71f 100644
> --- a/net/packet/af_packet.c
> +++ b/net/packet/af_packet.c
> @@ -1129,8 +1129,6 @@ static int tpacket_rcv(struct sk_buff *skb, struct net_device *dev,
>  	else
>  		sll->sll_ifindex = dev->ifindex;
>  
> -	__packet_set_status(po, h.raw, status);
> -	smp_mb();
>  #if ARCH_IMPLEMENTS_FLUSH_DCACHE_PAGE == 1
>  	{
>  		u8 *start, *end;
> @@ -1140,6 +1138,8 @@ static int tpacket_rcv(struct sk_buff *skb, struct net_device *dev,
>  			flush_dcache_page(pgv_to_page(start));
>  	}
>  #endif
> +	smp_mb();
> +	__packet_set_status(po, h.raw, status);
>  
>  	sk->sk_data_ready(sk, 0);
>  

Are you sure we can perform the flush_dcache_page() before smp_mb() ?

I believe we should do one smp_wmb() before __packet_set_status() as
well.