Bill Cole via Postfix-users <[email protected]> wrote:
> On 2026-05-01 at 09:11:26 UTC-0400 (Fri, 1 May 2026 15:11:26 +0200)
> Michael Grimm via Postfix-users <[email protected]>
> is rumored to have said:

>>> Presumably, pf cannot prevent a non-Postfix pdocess from sendfing
>>> email directly to remote port 25, 465, and 587.
>> 
>> Yes, that's impossible (to my knowledge). Any process in that jail trying to 
>> send spam will pass that firewall rules, sadly.
> 
> I don't know if pf can use the 'user' parameter on the host on packets coming 
> from the jail, but if the user is preserved, you can make it a little less 
> obvious how to send spam by requiring that the packets are owned by the 
> postfix user.

I have to admit that I never thought about such an approach and that I wasn't 
aware of such a functionality.

Yes, pf can filter by uid and gid. But according to the manual page this only 
works on the host running the firewall. A quick check confirms it; postfix' uid 
running in a jail is unknown to the firewall [1].

But because I am running VNET jails I can simply create an additional firewall 
inside postfix' jail. And there I should be able to block outgoing traffic not 
initiated by a postfix process.

I will have to evaluate and test that approach, though, but I am quite 
confident that it will work as expected.

Thanks for pointing me into this direction and regards,
Michael

[1] For forwarded connections, where the firewall is not a connection endpoint, 
the user and group are unknown.


_______________________________________________
Postfix-users mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to