Subject: tftp -e hang
To: None <netbsd-help@netbsd.org>
From: Stephen Borrill <netbsd@precedence.co.uk>
List: netbsd-help
Date: 07/04/2006 15:51:35
I've an embedded system running with a cutdown kernel and with root on md. 
I've just noticed that if I pass the -e flag to tftp, the connection 
hangs. tcpdump seems to suggest it's something to do with fragmentation.

17:44:58.815025 IP (tos 0x0, ttl  64, id 26633, offset 0, flags [+], length: 1500)
 	 192.168.2.254.64929 > 192.168.2.16.65512: UDP, length: 65468
17:44:58.815028 IP (tos 0x0, ttl  64, id 26633, offset 1480, flags [+], length: 1500)
 	 192.168.2.254 > 192.168.2.16: udp
[snip offset increasing]
17:44:58.815078 IP (tos 0x0, ttl  64, id 26633, offset 63640, flags [+], length: 1500)
 	 192.168.2.254 > 192.168.2.16: udp
17:44:58.815079 IP (tos 0x0, ttl  64, id 26633, offset 65120, flags [none], length: 376)
 	 192.168.2.254 > 192.168.2.16: udp

No data is written by the client (it creates a zero length file).

It works without the -e and it works with any full NetBSD installation. 
I'm not sure where to look to see what's missing from my cutdown system 
that's causing it.

-- 
Stephen