Re: [PATCH] inet: add RCU protection to inet->opt

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

Le jeudi 28 avril 2011 à 13:52 -0700, David Miller a écrit :
> From: Eric Dumazet <eric.dumazet@gmail.com>
> Date: Thu, 28 Apr 2011 22:49:19 +0200
> 
> > Are you sure socket is locked at this point ? ( by a priori call to
> > lock_sock())
> 
> Ugh, it's not, I'll commit this fix which is safe because we're only
> peeking at inet_opt to fetch the faddr value then we never reference
> it again:
> 
> diff --git a/net/l2tp/l2tp_ip.c b/net/l2tp/l2tp_ip.c
> index 962a607..e13c166 100644
> --- a/net/l2tp/l2tp_ip.c
> +++ b/net/l2tp/l2tp_ip.c
> @@ -472,13 +472,15 @@ static int l2tp_ip_sendmsg(struct kiocb *iocb, struct sock *sk, struct msghdr *m
>  	if (rt == NULL) {
>  		struct ip_options_rcu *inet_opt;
>  
> -		inet_opt = rcu_dereference_protected(inet->inet_opt,
> -						     sock_owned_by_user(sk));
> +		rcu_read_lock();
> +		inet_opt = rcu_dereference(inet->inet_opt);
>  
>  		/* Use correct destination address if we have options. */
>  		if (inet_opt && inet_opt->opt.srr)
>  			daddr = inet_opt->opt.faddr;
>  
> +		rcu_read_unlock();
> +
>  		/* If this fails, retransmit mechanism of transport layer will
>  		 * keep trying until route appears or the connection times
>  		 * itself out.


Thats perfect, thanks !