RE: [PATCH] tcp: md5: fix md5 RST when both sides have listener

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

Le mardi 31 janvier 2012 à 10:05 +0100, Eric Dumazet a écrit :

> Hmm... The second problem is that its not safe to call
> tcp_v4_md5_do_lookup() on an unlocked socket.
> 
> And locking a listener is way too expensive, since a listener socket is
> already a contention point.
> 
> An attacker could send forged tcp md5 packets to slow down a server.
> 
> A proper patch needs RCU conversion first.

I am working on this RCU conversion and big md5 cleanup, using a single
list of keys.

You can then add your fix on top of this work.


union tcp_md5_addr {
        struct in_addr  a4;
#if IS_ENABLED(CONFIG_IPV6)
        struct in6_addr a6;
#endif
};

/* - key database */
struct tcp_md5sig_key {
        struct hlist_node       node;
        u8                      keylen;
        u8                      family; /* AF_INET or AF_INET6 */
        union tcp_md5_addr      addr;
        u8                      key[TCP_MD5SIG_MAXKEYLEN];
        struct rcu_head         rcu;
};

/* - sock block */
struct tcp_md5sig_info {
        struct hlist_head       head;
};