As background, the RELEASE_NOTES for 3.8 mention:...>> 
postfix/psint/smtpd[27820]:
    esa.hc2802-61.iphmx.com[139.138.32.157]: TLS cipher list
   
"aNULL:-aNULL:HIGH:MEDIUM:!SEED:!IDEA:!3DES:!RC2:!RC4:!RC5:!kDH:!kECDH:!aDSS:!MD5:+RC4:@STRENGTH:!aNULL"

yep

Which is most of the above, but you also disable aNULL, which though
unrelated to the problem, is counter-productive.  Let the checklists
give you a lower score for this, they're all wrong:

     https://datatracker.ietf.org/doc/html/rfc7672#section-8.2

I've no idea in this case why aNULL is explicitly ref'd; for my own configs I 
don't call it out, rather stick with the default

and, TBH, don't get what's happening here

        "aNULL:-aNULL:...!aNULL"

is remove-all of those aNULL instnces in the cipher list what's advised by 
section-8.2?

Less verbose logging

well, that's at least where I started

PCAP capture of an attempted TLS handshake would have been more useful.

I've just been reading up on tcpdump'ing TLS -- and trying to figure out 
extracting cipher types; seems a useful add'l tool to have at hand.

latest naive attempt,

  tcpdump -i eth0 "tcp port 25 and (tcp[((tcp[12] & 0xf0) >> 2)] = 0x16)"

is still unhelpful.

My money is on an ECDSA server certificate
on the receiving server's end, right?

which is still a bit too "exotic" for some older SMTP clients.

here, checking

         postconf -n smtpd_tls_key_file smtpd_tls_cert_file
                smtpd_tls_key_file = 
/etc/sec/ssl/LE/deploy/${var_DOM_CMX0024}/priv.rsa.key smtpd_tls_cert_file = 
/etc/sec/ssl/LE/deploy/${var_DOM_CMX0024}/fullchain.rsa.crt.pem

that's ... odd
(hm. i don't see any postfix errors in startup; `postfix check` returns OK, too)

checking further

        grep smtpd_tls main.cf | grep file
                smtpd_tls_dh1024_param_file=${config_directory}/dh4096.pem
                smtpd_tls_eckey_file  = 
/etc/sec/ssl/LE/deploy/${var_DOM_CMX0024}/priv.ec.key
                smtpd_tls_eccert_file = 
/etc/sec/ssl/LE/deploy/${var_DOM_CMX0024}/fullchain.ec.crt.pem
                smtpd_tls_key_file    = 
/etc/sec/ssl/LE/deploy/${var_DOM_CMX0024}/priv.rsa.key
                 smtpd_tls_cert_file   = 
/etc/sec/ssl/LE/deploy/${var_DOM_CMX0024}/fullchain.rsa.crt.pem

and

        cat main.cf
                ...
                smtpd_tls_key_file    = 
/etc/sec/ssl/LE/deploy/${var_DOM_CMX0024}/priv.rsa.key

                 smtpd_tls_cert_file   = 
/etc/sec/ssl/LE/deploy/${var_DOM_CMX0024}/fullchain.rsa.crt.pem
                ...

note the errant leading-space.

sigh.  a continuation line.

which, iiuc, would lead to the ec cert being offered, but NOT the 'continued' 
rsa cert spec

and, indeed, rm'ing the space cures the problem.

not where i was looking.  i wouldn't have caught anytime soon(ish) without the 
hint. thx!

digging through this server's logs, the *vast* majority of these errors 
originate from 3 senders

        *.iphmx.com (healthcare org)
        *.jpmchase.com (finance org)
        *.comcast.net (evil empire)

each numbering in the 1000s; many others by ones & twos as well ...
one'd hope that banks and hospitals might be a bit more up-to-date on their end.

currently / so far, this server's config is
...
Though not relevant for the SMTP server, just stick with the defaults,
the above hodge-podge just leads to more email going out in the clear.

time to clean up some legacy bad habits!  i certainly carried such baggage 
around for way too long myself ...

thx agn!


_______________________________________________
Postfix-users mailing list -- postfix-users@postfix.org
To unsubscribe send an email to postfix-users-le...@postfix.org

Reply via email to