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

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

Le mardi 31 janvier 2012 à 16:50 -0800, Shawn Lu a écrit :
> TCP RST mechanism is broken in TCP md5(RFC2385). When
> connection is gone, md5 key is lost, sending RST
> without md5 hash is deem to ignored by peer. This can
> be a problem since RST help protocal like bgp to fast
> recove from peer crash.
> 
> In most case, users of tcp md5, such as bgp and ldp,
> have listeners on both sides. md5 keys for peers are
> saved in listening socket. When passive side connection
> is gone, we can still get md5 key from listening socket.
> When active side of connection is gone, we can try to
> find listening socket through source port, and then md5
> key.
> we are not loosing sercuriy here:
> packet is valified checked with md5 hash. No RST is generated
> if md5 hash doesn't match or no md5 key can be found.
> 
> Signed-off-by: Shawn Lu <shawn.lu@ericsson.com>
> ---
>  net/ipv4/tcp_ipv4.c |   45 ++++++++++++++++++++++++++++++++++++++++++---
>  net/ipv6/tcp_ipv6.c |   43 +++++++++++++++++++++++++++++++++++++++++--
>  2 files changed, 83 insertions(+), 5 deletions(-)
> 

Hmm, ok but using RCU for the md5 lookup (instead of mere sock lock as
done today) also means we need to protect struct tcp_md5sig_info itself

I send a patch for this in a separate patch first, then David can apply
your patch when I add my "Signed-off-by: ..."