4 messages in net.sourceforge.lists.courier-usersRe: [courier-users] scanning for spam...
FromSent OnAttachments
Hari KurupApr 27, 2005 7:58 am 
Ben KennedyApr 27, 2005 8:20 am.txt
Hari KurupApr 27, 2005 11:34 pm 
Ben KennedyApr 28, 2005 8:34 am.txt
Actions with this message:
Paste this link in email or IM:
Paste this link in email or IM:
Atom feed for this thread
Paste this URL into your reader:
Subject:Re: [courier-users] scanning for spam on relay serverActions...
From:Ben Kennedy (be@zygoat.ca)
Date:Apr 28, 2005 8:34:52 am
List:net.sourceforge.lists.courier-users
Attachments:

Hari Kurup wrote at 9:32 am (+0300) on Thu 28 Apr 2005:

Thanks Ben. It works but as you mentioned, my CPU utilisation is always not less than 80%. I do not require to check for authenticated sessions as I don't do any authenticated smtp so could you kindly send me the earlier version of your filter that does not have this?

Certainly -- attached.

I suspect that the high load situation is caused by the header-parsing bit looping indefinitely on an EOF condition in some circumstances; I suspect it can be solved by changing the while(1) to while(eof()) on line 63 but I have not verified this yet.

In the mean time here is the simpler version with the auth check rolled back which seems to work reliably.

-ben

-- Ben Kennedy, chief magician zygoat creative technical services 613-228-3392 | 1-866-466-4628 http://www.zygoat.ca

# ClamAV and etc filter for Courier perlfitler. See the perlfilter-example that # comes with Courier for the comments.

# Installed by BRK on 07 July 2004 from http://karmak.org/2004/courier-clamav/ # BRK also altered return status from 500 to 577 which seems more consistent
with rfc1893.

# 05 Nov 2004: BRK modified to first scan the incoming message and reject if # there is no From: header provided. Seems there are spammy things getting
through # with no header, so by virtue of maildrop etc. it ends up misleadingly with the # recipient user's own address.

# 13 Dec 2004: BRK modified again to also call spamc and check for an egregious
spam score # following a successful virus check. This is an effort to cull out the most
obvious # crap before it even gets to the maildrop process.

# 20 Apr 2005: BRK modified; we now check the first Received: header to see
whether the # message came from an authenticated session. If so, we bypass the virus and
spam checks. # 27 Apr - rolled back; somehow this is getting stuck and eating lots of cpu.

use IO::File;

my $filedesc=shift @ARGV; my $socket=new IO::File "+<&$filedesc"; die "$!" unless defined $socket;

my $line; my $first=1; my $errmsg="200 Ok";

while (defined ($line=<$socket>)) { my $msg; chomp $line; last unless $line; if ($first) { $msg=filterdata($line); } else { $msg=filtercontrol($line); } $first=0; $errmsg=$msg if $msg; }

$errmsg .= "\n" unless $errmsg =~ /\n$/; print $socket $errmsg;

$socket->close;

sub filterdata { my $filename=shift;

# BRK: First scan the input file to ensure there is a From: header.

my $goodfrom = 0; # whether we have found presence of a From: header

open(FILE, "<${filename}"); my $nextline = <FILE>; # fetch first line while(<FILE>) { my $p = $_;

# If this is a From header, make note of its presence. if ($p =~ /^From\: /) { $goodfrom = 1; }

# If this is the end of the headers, or if we've found the From:, # no need to continue parsing. if ($p eq '' || ($goodfrom)) { last; } }

close FILE;

if (!$goodfrom) # If no From: header existed, barf up with the error. { return '550 Please provide a From: header in the message.'; }

# We passed; now see if it smells like fish.

# diag: dump env. #open(E, '>/tmp/msg.env'); #my $key; #foreach $key (sort(keys %ENV)) # { print E, $key . ' = ' . $ENV{$key} . "\n"; } #close E;

open(SPAMOUT, "cat ${filename} | spamc -c -U /var/run/spamd.sock |"); my $result = <SPAMOUT>; close(SPAMOUT); # Fetch result which will be "score/threshold", e.g. "6.9/4.0". $result =~ m/^([0-9.]+)/; # Extract the score. $result = $1; # Fail if the score is reasonably large. if ($result > 4.9) { return "571 Identified spam ($result). Contact postmaster\@zygoat.ca if this
is in error."; }

# Now, carry on with the original CLAMAV scanning.

open(CLAMOUT, "cat ${filename} | clamdscan --stdout - |"); my $result = <CLAMOUT>; close(CLAMOUT); if ( $result =~ m/ FOUND$/ ) { $result =~ s/.*stream: (.*) FOUND/$1/; chomp $result; # strip trailing newline (?) return "577 Virus detected (${result})."; }

# Otherwise all looks good, so return usual.

return ""; }

sub filtercontrol { my $filename=shift; return ""; }