On 22 Feb 99, at 16:30, Fuzzy Fox wrote about
    "[Masq]  Re: Masquerade access to se":

| 
| Markus Hansmair <[EMAIL PROTECTED]> wrote:
| >
| > [PORT-ftp will work, PASV-ftp won't]
| > 
| >     Am I right?
| 
| You are right.  I didn't notice an ftp server in my initial response.
| 
| Note that there is code in the ip_masq_ftp module which appears to be
| trying to modify any PASV-reply from a masqueraded ftp server, but other
| traffic on this list has pointed out that it does not seem to work at
| all.

There's no code in ip_masq_ftp.c that *modifies* PASV replies. All it 
does with them is set up a masquerade entry for the data connection 
with a "keepalive" link to the control connection.

The IPPORTFW patch does not bind masq modules for the ports it is 
forwarding.  Also the kernel does not check for a bound module based 
on the source port (unless IPAUTOFW is enabled). So the ip_masq_ftp 
module doesn't get called at all in the masqueraded server case.  
Fixing this takes a couple of very small kernel changes.

The ip_masq_ftp module is direction sensetive.  It only looks for 
PASV replies in *incoming* packets, and PORT commands in *outgoing* 
packets.  So when it's the server that is masqueraded and the client 
that is external, neither packet type is ever seen because they are 
going in the other directions.  Fixing this takes a pretty 
significant re-write of ip_masq_ftp.c.  It needs to be orthogonal: 
checking for both PORT commands and PASV replies in all packets, re-
writing both types in outgoing traffic and setting up the keep-alive 
entry for both types in incoming traffic.

- Fred Viles <mailto:[EMAIL PROTECTED]>




_______________________________________________
Masq maillist  -  [EMAIL PROTECTED]
http://tiffany.indyramp.com/mailman/listinfo/masq
Admin requests can be handled by web (above) or [EMAIL PROTECTED]

Reply via email to