[PATCH] ipv4 tcp autobind problem

[PATCH] ipv4 tcp autobind problem

[Kovacs Krisztian]

[-- Attachment #1: Type: text/plain, Size: 1258 bytes --]


   Hi,

   While testing the tproxy (transparent proxying) patch for linux-2.4
(http://www.balabit.com/downloads/tproxy/linux-2.4), Stas Grabois has
found a quite strange aspect of Linux 2.4 TCP. Imagine the following
scenario: you create a new socket (AF_INET, SOCK_STREAM), bind it to local
port 0, and try to connect() to a closed port. Of course, the peer sends
back an RST, indicating no one is listening on that port. However, if your
application does not care about the return value of connect(), and calls
send() on the not connected socket, inet_autobind() is called and a new
local port is allocated for the socket. So, besides returning an error,
there is also a side effect of the send(). The same thing happens with an
established TCP session if the peer sends an RST between two send() calls,
the second call will autobind the socket, and then return error.

   Is this behaviour intentional? Isn't rebinding a TCP socket to a new
local port a bug? I mean, possibly inet_sendmsg() should check if the
socket is SOCK_STREAM before calling inet_autobind() if sk->num is zero. 
The attached patch adds this check to inet_sendmsg(). We've been using it 
for a while, and it looks it did not break anything.

-- 
   Regards,
     Krisztian KOVACS


[-- Attachment #2: inet_tcp_autobind_fix.diff --]
[-- Type: text/plain, Size: 400 bytes --]

--- linux-2.4.22/net/ipv4/af_inet.c.orig	Thu Sep 18 10:02:49 2003
+++ linux-2.4.22/net/ipv4/af_inet.c	Thu Sep 18 10:03:56 2003
@@ -751,7 +751,7 @@
 	struct sock *sk = sock->sk;
 
 	/* We may need to bind the socket. */
-	if (sk->num==0 && inet_autobind(sk) != 0)
+	if (sk->num==0 && sock->type != SOCK_STREAM && inet_autobind(sk) != 0)
 		return -EAGAIN;
 
 	return sk->prot->sendmsg(sk, msg, size);

Re: [PATCH] ipv4 tcp autobind problem

[David S. Miller]

Once connect() has been called on a socket, you may not ever again
perform any action that would try to connect that socket.

Said another way, a socket that has failed to connect() is a socket
that you may not use in any usable way ever again except to close
that file descriptor.

Re: [PATCH] ipv4 tcp autobind problem

[Kovacs Krisztian]

   Hi,

David S. Miller wrote:
> Once connect() has been called on a socket, you may not ever again
> perform any action that would try to connect that socket.
> 
> Said another way, a socket that has failed to connect() is a socket
> that you may not use in any usable way ever again except to close
> that file descriptor.

   Yes, I know that. Exactly this is why I felt it a problem that sk->num 
gets a new value if you call send() after an unsuccessful connect(). (I 
know that in theory one should not call send() in this case, but one _can_ 
call it in reality.) And, the problem occurs not only when send() is 
called after an unsuccessful connect(), but also when an RST is received 
between two send() calls, which is perfectly legal, and cannot be viewed 
as buggy user-space software.

   I don't really understand why the inet_sendmsg() calls inet_autobind() 
for SOCK_STREAM sockets: for these kind of sockets, one must call 
connect() anyway, before doing any other kind of operations. And, the side 
effect of the code

         /* We may need to bind the socket. */
         if (!inet_sk(sk)->num && inet_autobind(sk))
                 return -EAGAIN;

in inet_sendmsg() is that when an RST is received, sk->num is set to zero, 
and when the next inet_sendmsg() call occurs, the socket is rebound to a 
new port _before_ returning an error.

   In our transparent proxying patch, this causes problems, because it 
assumes that after a socket is bound (by connect(), for example), the 
local port (sk->sport) remains unchanged until the socket is closed. 
However, this is not true because of the mentioned side effect of 
inet_sendmsg(). This is why I proposed that inet_sendmsg() should call 
inet_autobind() only if it's not a SOCK_STREAM socket.

-- 
   Regards,
     Krisztian KOVACS

Re: Fw: Re: [PATCH] ipv4 tcp autobind problem

[kuznet]

Hello!

> - When the socket is closed (inet_release() is called, tproxy has its 
> "callback" function here), the corresponding entries are deleted from 
> tproxy's hash tables, again, based on the local ip:port pair.

The mistake is here. inet_release() has nothing to do to connection
state. Socket is not closed at this point, it goes to FIN-WAIT*, LAST-ACK or
something like this and can exist for long time after this and proxying
is to be continued all this time.

Right place to finish tracking is when socket is removed from TCP
hash tables, and to start tracking is when the socket is inserted
to TCP hash tables. BTW you would not see the problem with binding
if it was made right.

Alexey

Re: [PATCH] ipv4 tcp autobind problem

[Kovacs Krisztian]

   Hi,

kuznet@ms2.inr.ac.ru wrote:
> Right place to finish tracking is when socket is removed from TCP
> hash tables, and to start tracking is when the socket is inserted
> to TCP hash tables. BTW you would not see the problem with binding
> if it was made right.

   Thanks for the reply. In the meantime, although because of slightly 
different reasons, we also came to the same conclusion. Right now I'm 
testing a version in which our "unassign" hook is removed from 
inet_release(), and added to udp_v4_unhash() and tcp_put_port().
However, I'm not quite sure there won't be locking problems when calling 
our hook from these places.

-- 
   Regards,
     Krisztian KOVACS