肖建彬的博客

作者: 肖建彬 | 可以转载, 转载时务必以超链接形式标明文章原始出处和作者信息及版权声明
网址：http://www.xiaojb.com/archives/mailtech/data_timeout_fr.shtml

question:
Morten Albrigtsen wrote:

I have a MS Exchange server on an internal network, which is supposed
to relay all mail to my postfix server so that mail doesnt get stuck
inside the building, but is accesible from the internet.

The exchange server sends the emails to the postfix server, but I
suspect it sends some sort of crazy format that doesnt properly end
the DATA transfer, log:

 > unknown[192.168.20.2]: 220 xxx.xxxx.xxx ESMTP Postfix (Trustix)
 unknown[192.168.20.2]: 250-xxxxx.xxxxx.xxxxx
 > unknown[192.168.20.2]: 250-PIPELINING
 > unknown[192.168.20.2]: 250-SIZE 10240000
 > unknown[192.168.20.2]: 250-VRFY
 > unknown[192.168.20.2]: 250-ETRN
 > unknown[192.168.20.2]: 250-STARTTLS
 > unknown[192.168.20.2]: 250 8BITMIME
 unknown[192.168.20.2]: 220 Ready to start TLS
 SIZE=3116
 > unknown[192.168.20.2]: 250 Ok

 > unknown[192.168.20.2]: 250 Ok
 unknown[192.168.20.2]: 354 End data with .
 > unknown[192.168.20.2]: 421 xxx.xxxxxxxx.xx Error: timeout exceeded

And:
: timeout after DATA from unknown[192.168.20.2]

I have googled now for about 2 days, trying a lot of crazy things but
things just dont seem to work.

This has been driving me crazy, anyone who have experienced anything
like it? Or have a clever tip?

The answers involve:

This is a standard problem with servers behind PPP over Ethernet.
Try a smaller MTU on your server ethernet interface, or configure
your router to spoof a smaller MTU in the replies from your server.

The client has a broken PIPELINING implementation. Turn off PIPELINING
on the Postfix side (requires Postfix 2.3 smtpd_discard_ehlo_keywords
or smtpd_discard_ehlo_keyword_address_maps feature).

These answers are archived world-wide.

Tags: Postfix