Peer, > If reinjection of mail fails, Amavis will send a DSN back to the sender to > notify him about this failure. > > But if D_REJECT is active this leads to *2* DSN back to the user: One DSN > from the sending MTA (because Amavis blocks that mail to the feeding > MTA), second DSN from Amavis himself.
A bug. Either a reject or a bounce should happen, not both. As Thomas said, it was not envisaged to have a back-end MTA reject mail at this late stage, so the situation was not thoroughly tested. > In my understanding Amavis should not send his own DSN in case of D_REJECT > if the mail has been already rejected against the feeding local MTA (and > in case of Postfix' smtpd_proxy_filter also against the external sending > MTA). Note there is no dedicated configuration variable to configure what should happen when a back-end MTA does not want to accept a message. Luckily there is a %final_destiny_by_ccat, which allows to assign a 'destiny' to a CC_MTA case: %final_destiny_by_ccat = ( CC_VIRUS, sub { c('final_virus_destiny') }, CC_BANNED, sub { c('final_banned_destiny') }, CC_SPAM, sub { c('final_spam_destiny') }, CC_BADH, sub { c('final_bad_header_destiny') }, CC_MTA.',2', D_REJECT, # in response to 5xx from MTA CC_OVERSIZED, D_BOUNCE, CC_CATCHALL, D_PASS, ); This should have sufficed, had it not been for a bug which forgets to set $r->recip_destiny in case of a MTA reject. The following patch fixes it (applicable to 2.6.4). It is more extensive than it need be, because it also makes possible to obtain a reject text from a %smtp_reason_by_ccat, which was introduced in 2.6.4. Also, as the D_PASS does not make any sense in case of a MTA failure (we can't forward anyway, even though we want to), the patch silently turns a D_PASS into a D_REJECT for the CC_MTA case, so adjusting %final_destiny_by_ccat is not really necessary, it's just nice to be explicit. --- amavisd.orig 2009-06-25 14:39:01.000000000 +0200 +++ amavisd 2009-07-10 19:55:58.000000000 +0200 @@ -1493,4 +1493,5 @@ CC_SPAM, sub { c('final_spam_destiny') }, CC_BADH, sub { c('final_bad_header_destiny') }, + CC_MTA.',2', D_REJECT, CC_OVERSIZED, D_BOUNCE, CC_CATCHALL, D_PASS, @@ -11317,4 +11318,30 @@ $msginfo->blocking_ccat($blocking_ccat) if !defined($msginfo->blocking_ccat); + my($final_destiny) = + $r->setting_by_contents_category(cr('final_destiny_by_ccat')); + if ($final_destiny == D_PASS) { + $final_destiny = D_REJECT; # impossible to pass, change to reject + } + local($1,$2); + $r->recip_destiny($final_destiny); + if ($final_destiny == D_DISCARD && $smtp_resp =~ /^5/) { + $smtp_resp =~ s{^5(\d\d) 5(\.\d\.\d)}{250 2$2}; # 5xx -> 250 + } + my($smtp_reason) = # get the custom smtp response reason text + $r->setting_by_contents_category(cr('smtp_reason_by_ccat')); + $smtp_reason = '' if !defined $smtp_reason; + if ($smtp_reason ne '') { + my(%mybuiltins) = %builtins; # make a local copy + $smtp_reason = expand(\$smtp_reason, \%mybuiltins); + $smtp_reason = !ref($smtp_reason) ? '' : $$smtp_reason; + chomp($smtp_reason); $smtp_reason = sanitize_str($smtp_reason,1); + $smtp_reason = substr($smtp_reason,0,100) . "..." + if length($smtp_reason) > 100+3; + } + $smtp_resp =~ /^(\d\d\d(?: \d\.\d\.\d)?)\s*(.*)\z/; + my($dis) = $final_destiny == D_DISCARD ? ' Discarded' : ''; + $r->recip_smtp_response("$1$dis $smtp_reason, $2"); + $r->recip_done(1); # fake a delivery (confirm delivery to a bit bucket) + # note that 5xx status rejects may later be converted to bounces } $msginfo->header_edits($hdr_edits); # restore original edits just in case @@ -17775,10 +17802,10 @@ sub enhance_smtp_response($$$$$) { my($smtp_resp,$am_id,$mta_id,$dflt_enhcode,$cmd_name) = @_; - local($1,$2,$3); my($resp_shortmsg,$resp_msg); + local($1,$2,$3); my($resp_msg); my($resp_code,$resp_enhcode) = ('451', '4.5.0'); if (!defined($smtp_resp) || $smtp_resp eq '') { - $resp_shortmsg = 'No resp. to '.$cmd_name; + $smtp_resp = sprintf('No resp. to %s', $cmd_name); } elsif ($smtp_resp !~ /^[245]\d{2}/) { - $resp_shortmsg = 'Bad resp. to '.$cmd_name; + $smtp_resp = sprintf('Bad resp. to %s: %s', $cmd_name,$smtp_resp); } elsif ($smtp_resp =~ /^ (\d{3}) [ \t]+ ([245] \. \d{1,3} \. \d{1,3})? \s* (.*) \z/xs) { @@ -17787,9 +17814,7 @@ if ($resp_enhcode eq '' && $resp_code =~ /^[245]/) { $resp_enhcode = $dflt_enhcode; $resp_enhcode =~ s/^\d*/$c/ } - $resp_shortmsg = $c eq '2' ? 'Ok' : $c eq '4' ? 'TempFailed' : 'Failed'; } - sprintf("%s %s %s, id=%s, from MTA(%s): %s", - $resp_code, $resp_enhcode, $resp_shortmsg, - $am_id, $mta_id, $smtp_resp); + sprintf("%s %s from MTA(%s): %s", + $resp_code, $resp_enhcode, $mta_id, $smtp_resp); } The range of choices is now fully supported through the $final_destiny_by_ccat{(CC_MTA)} setting, i.e. D_REJECT, D_BOUNCE and D_DISCARD. Usually one would chose a D_REJECT or D_BOUNCE. Note that even when a D_REJECT is chosen as a final_destiny_by_ccat for the CC_MTA case, it can still happen that amavisd responds with a 2xx smtp status and sends its own bounce. This can happen with multi-recipient mail, where some recipients are accepted and some rejected by MTA. As SMTP protocol does not allow for per-recipient responses, we have no other choice but to accept a message and send a bounce to the rejected recipients, while delivering to the accepted recipients. So, late header checks are alright, but late per-recipient restrictions are to be avoided. Mark ------------------------------------------------------------------------------ Enter the BlackBerry Developer Challenge This is your chance to win up to $100,000 in prizes! For a limited time, vendors submitting new applications to BlackBerry App World(TM) will have the opportunity to enter the BlackBerry Developer Challenge. See full prize details at: http://p.sf.net/sfu/Challenge _______________________________________________ AMaViS-user mailing list AMaViS-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/amavis-user AMaViS-FAQ:http://www.amavis.org/amavis-faq.php3 AMaViS-HowTos:http://www.amavis.org/howto/