dovecot - Jan 2015 - quota_full_tempfail = yes broken with lmtp_rcpt_check_quota (was: Re: lmtp_rcpt_check_quota not working)

Axel Luttgens <AxelLuttgens at swing.be> wrote:

Hi Axel,
> Le 11 f?vr. 2014 ? 08:34, Bernhard Schmidt a ?crit :
>
>> [...]
>> so Dovecot should reject in RCPT TO phase, but according to the logs of
the
>> upstream Postfix it does so after END-OF-DATA
>> 
>> (host xxx[2001:xxx] said: 452 4.2.2
>> <user at domain.de> Quota exceeded (mailbox for user is
>> full) (in reply to end of DATA command))
>> [...]
>
> Hello Bernhard,
>
> Could you exercise your quota-status service with variations of:
>
> 	# telnet xxx 12340
> 	Trying xxx...
> 	Connected to xxx.
> 	Escape character is '^]'.
> 	recipient=john.doe at example.com
> 	size=1000000
>
> 	action=OK
>
> 	^]
>
> and see whether the replies are the expected one?
>
> If yes, this would point to some error in Postfix' configuration.
>
> If not, well... could be you are facing the same kind of problem I'm
> having here (never managed to have the service reply anything else
> than "OK" or "REJECT Unknown user"...)
quota-status works absolutely fine, our frontend MXes defer mails for
mailboxes over quota. But when we try to send a mail to the mailbox
nevertheless through LMTP (i.e. because it was generated on a server
that doesn't check quota-status), it is deferred in LMTP in the
END-OF-DATA phase, not in RCPT TO as expected.

Bernhard

Bernhard Schmidt <berni at birkenwald.de> wrote:

Hi,
> Axel Luttgens <AxelLuttgens at swing.be> wrote:
>
> Hi Axel,
>
>> Le 11 f?vr. 2014 ? 08:34, Bernhard Schmidt a ?crit :
>>
>>> [...]
>>> so Dovecot should reject in RCPT TO phase, but according to the
logs of the
>>> upstream Postfix it does so after END-OF-DATA
>>> 
>>> (host xxx[2001:xxx] said: 452 4.2.2
>>> <user at domain.de> Quota exceeded (mailbox for user is
>>> full) (in reply to end of DATA command))
>>> [...]
>>
>> Hello Bernhard,
>>
>> Could you exercise your quota-status service with variations of:
>>
>> 	# telnet xxx 12340
>> 	Trying xxx...
>> 	Connected to xxx.
>> 	Escape character is '^]'.
>> 	recipient=john.doe at example.com
>> 	size=1000000
>>
>> 	action=OK
>>
>> 	^]
>>
>> and see whether the replies are the expected one?
>>
>> If yes, this would point to some error in Postfix' configuration.
>>
>> If not, well... could be you are facing the same kind of problem
I'm
>> having here (never managed to have the service reply anything else
>> than "OK" or "REJECT Unknown user"...)
>
> quota-status works absolutely fine, our frontend MXes defer mails for
> mailboxes over quota. But when we try to send a mail to the mailbox
> nevertheless through LMTP (i.e. because it was generated on a server
> that doesn't check quota-status), it is deferred in LMTP in the
> END-OF-DATA phase, not in RCPT TO as expected.
To follow up here:

Dovecot 2.2.13 (from Debian testing compiled on wheezy) is still showing
the same behaviour as 2.2.9. lmtp_rcpt_check_quota is apparently
ignored, the over-quota defer happens after the DATA phase.

# 2.2.13: /etc/dovecot/dovecot.conf
# OS: Linux 3.2.0-4-amd64 x86_64 Debian 7.8 
[...]
lmtp_rcpt_check_quota = yes
quota_full_tempfail = yes
protocol lmtp {
  mail_plugins = quota sieve
}

 -> MAIL FROM:<sender at domain.de>
<-  250 2.1.0 OK
 -> RCPT TO:<recipient at domain.de>
<-  250 2.1.5 OK
 -> DATA
<-  354 OK
 -> Date: Thu, 29 Jan 2015 13:35:51 +0100
 -> To: recipient at domain.de
 -> From: sender at domain.de
 -> Subject: test Thu, 29 Jan 2015 13:35:51 +0100
 -> X-Mailer: swaks v20120320.0 jetmore.org/john/code/swaks/
 -> 
 -> This is a test mailing
 -> 
 -> .
<** 452 4.2.2 <recipient at domain.de> Quota exceeded (mailbox for user
is full)

With 2.2.15 from dovecot-ee it is different. The quota is now checked at
RCPT stage

# 2.2.15.8 (912dd04d7a14+): /etc/dovecot/dovecot.conf
# Pigeonhole version 0.4.6 (4b9b9a88ac9b)
# OS: Linux 3.2.0-4-amd64 x86_64 Debian 7.8 

 -> MAIL FROM:<sender at domain.de>
<-  250 2.1.0 OK
 -> RCPT TO:<recipient at domain.de>
<** 552 5.2.2 <recipient at domain.de> Quota exceeded (mailbox for user
is full)

But as you can see, the setting for quota_full_tempfail is now ignored.
It should be deferring the message, not rejecting it. When I disable
lmtp_rcpt_check_quota it behaves like 2.2.13.

Summary:
- In 2.2.9 and 2.2.13 lmtp_rcpt_check_quota is ignored
- In 2.2.15 lmtp_rcpt_check_quota works, but quota_full_tempfail is
  ignored at the RCPT TO stage. It still works at the DATA stage, so
  disabling lmtp_rcpt_check_quota is a workaround falling back to 2.2.13
  behaviour

Bernhard