smtpd timeout

[Stocky60]

I always get timeouts receiving mail from my employer's Exchange system (as well as several other senders whose implementation I don't know).

tcpdump shows that I do receive the entire message content, and that the packet ends with "<CR><LF><CR><LF>.<CR><LF>QUIT<CR><LF>". It also shows that tcp immediately sends an ack for the packet but that smptd does not respond to either the data content or the QUIT, other than to send a "421 timeout exceeded" some five minutes later.

Any idea what could be going on here or what I could do about it?

Running Postfix 2.1.5 on Mac OS X Server 10.4.8, PPC G5 processor.

 

[Stocky60]

Rebuilt smtpd with a few debug statements thrown in. It seems that for those connections that "timeout after DATA", the process gets hung as soon as it begins trying to read the mail content, not succeeding at reading a single byte.

I'm sure this is related, but I just noticed something interesting in the tcpdump output: for those connections that succeed, the data sequence numbers in all the packets are consecutive from start to finish, but for those that timeout, the sequence numbers look as though I am missing a packet between the DATA command and the data content, and that missing packet is about 25% larger than the data content packet itself. I assume that my timeout is due to the read waiting for the missing packet.

Apparently something somewhere along the line is munging the data sequence numbers. Is there any way to find out who is causing this? It's not my firewall, because I turned it off for this test.

This is really frustrating, because I can't send myself mail from my employer, and I also am unable to receive mail from my bank.