Did pmg-smtp-filter catch a glitch?

AlekseyZi

New Member
Mar 7, 2025
3
0
1
PMG version 8.1.5

Hello everyone, we caught a bug\glitch with an empty whitelist.
At a certain point we noticed that there was more spam and the reason was the empty whitelist, after rebooting the problem went away.

Did pmg-smtp-filter catch a glitch?

CLI Whitelist:
root@mail:/etc/mail/spamassassin# pmgsh get /config/ruledb/who/38/objects
200 OK
[]
root@mail:/etc/mail/spamassassin# pmgsh get /config/ruledb/who/37/objects
200 OK
[]

Tracking Center, log of one of the letters:
2025-03-07T08:29:44 mail postfix/smtpd[1473912]: connect from unknown[192.253.236.92]
2025-03-07T08:29:45 mail postfix/smtpd[1473912]: NOQUEUE: client=unknown[192.253.236.92]
2025-03-07T08:29:46 mail pmg-smtp-filter[1475258]: 2001467CA844A09F2A: new mail message-id=<2025030726974210855525@oj749.repica.jp>#012
2025-03-07T08:29:46 mail pmg-smtp-filter[1475258]: 2001467CA844A09F2A: SA score=34/5 time=0.825 bayes=0.56 autolearn=spam autolearn_force=no =BAYES_50(0.3),DMARC_QUAR(0.1),ENA_SUBJ_LONG_WORD(2.2),GB_SUBJ25(0.5),HTML_MESSAGE(0.001),HTTP_ESCAPED_HOST(0.1),KAM_DMARC _QUARANTINE(2),KAM_DMARC_STATUS(0.01),KAM_VERY_BLACK_DBL(5),MIXED_HREF_CASE(1.997),RCVD_IN_HOSTKARMA_BL(2.5),RCVD_IN_MSPIKE_BL( 0.001),RCVD_IN_MSPIKE_ZBI(0.001),RCVD_IN_SBL(0.141),RCVD_IN_SBL_CSS(1.5),RDNS_NONE(1.5),SPF_HELO_NONE(0.001),SPF_SOF TFAIL(1.5),UNWANTED_LANGUAGE_BODY(2),URIBL_ABUSE_SURBL(1.25),URIBL_BLACK(1.7),URIBL_DBL_SPAM(10),URIBL_PH_SURBL(0.61)
2025-03-07T08:29:46 mail postfix/smtpd[1475336]: connect from localhost.localdomain[127.0.0.1]
2025-03-07T08:29:46 mail postfix/smtpd[1475336]: E0AE21449AF: client=localhost.localdomain[127.0.0.1], orig_client=unknown[192.253.236.92]
2025-03-07T08:29:46 mail postfix/cleanup[1475115]: E0AE21449AF: message-id=<2025030726974210855525@oj749.repica.jp>
2025-03-07T08:29:46 mail postfix/qmgr[3793576]: E0AE21449AF: from=<webmaster-psmbcarfd-Administrator-t5wM@accounts.nintendo.com>, size=8148, nrcpt=1 (queue active)
2025-03-07T08:29:46 mail pmg-smtp-filter[1475258]: 2001467CA844A09F2A: accept mail to <***@**** email> (E0AE21449AF) (rule: WL-recipients)
 
Last edited:
how does the rule look like? (pmgdb dump) - how did the rule look like before the reboot?

In general I don't think that a reboot should change much in processing of mails by the rule-system?
 
At the moment 03/07/2025 (Just in case, we disabled this rule):
RULE 24 (prio: 91, in, inactive): WL-recipients
TO group 38 (and=0, invert=0): WL-recipients-ZAO
TO group 37 (and=0, invert=0): WL-recipients-RT
ACTION group 17: Accept

In the nightly backup 03/06/2025:
pmg-night.png
But I forgot to mention that we also updated to 8.2.0
 
Last edited:
the difference I see is that WL-recipients-RT (group 37) had 'and=1' in your backup, while at the moment it's 'and=0'

These flags (match any, match all) were introduced in PMG 8.1 - if you have Match all then you get 'and=1, invert=0' - and in this case an empty
who-object matches all addresses (matching all of nothing is always true)

I hope this explains it.
 
Thank you very much, we will be more careful. Perhaps we accidentally switched to "all match" and did not notice, because there is no apply button.