Hello,
Marten Lehmann <lehmann@???> (Tue Feb 22 19:15:21 2011):
> Hello,
>
> I wonder why Exim is using arbitrary interfaces so send mail from.
>
> Our remote_smtp transport on our webservers is pretty simple:
>
> remote_smtp:
> drive = smtp
>
> I know that it is possible to specify an interface address, but we
> don't want to do this, because the configuration shall be identical
> on all webservers.
>
> So far either Exim sent from the first interface (eth0), or in case
> of SSL-hosts from the IP address where the incoming https-connection
> went to.
I do not see any relation between HTTPS *to* your server and Exims
*choice* for the outgoing interface.
Actually if the above transport is used, Exim does not bind to any
interface for outgoing connections, thus it is totally up to the
underlying OS to choose some interface.
> But recently we noticed, that a lot of email is sent by arbitrary
> interfaces. Does exim rely on the operating system or does it decide
> on its own which interface to choose?
From smtp_out.c (around line 200) it is visible, that bind() is called
only if the transport specifies an interface.
> The problem is, that not all
> additional interfaces have valid reverse-dns records and even if
> they had, sometimes mails what have been sent with a wrong sender
> specifying a reverse-dns record, that has nothing to do with the
> actual user.
The sender (I suppose you're talking about the sender address) is not
related to reverse-dns of some interface. The sender should be checked
by other hosts using the MX records for the sender's address.
(Except you're using SPF records for the sending domains, then it might
be important to choose the right sending interface.)
--
Heiko :: dresden : linux : SCHLITTERMANN.de
GPG Key 48D0359B : 3061 CFBF 2D88 F034 E8D2 7E92 EE4E AC98 48D0 359B
