0

I'm trying to make some (temporary) use of an old machine. It has no disks of its own, but has a boot-capable network interface (Intel's LANDesk 0.99b).

I got it to obtain the IP-address and other options (including the TFTP-server) via DHCP.

However, when it tries actually download the specified file (pxeboot12), both sides complain of timeout. The server (FreeBSD-11.x) logs:

tftpd[16109]: Filename: 'pxeboot12'
tftpd[16109]: Mode: 'octet'
tftpd[16109]: receive_packet: timeout

and the client:

PXE-E32: TFTP open timeout.

There is no firewall in between the machines -- they plug into the same switch.

With tcpdump I can see the initial TFTP-request -- and the immediate response to it -- but nothing happens afterwards.

Using the tftp-client from another machine on the LAN, I was able to download the same boot-file in a fraction of a second.

What could be happening?

Update: based on this article, I disabled the MTU Discovery on the server (sysctl net.inet.tcp.path_mtu_discovery=0) and also lowered the server's MTU to 700. Still noworky...

Mikhail T.
  • 2,338
  • 1
  • 24
  • 55

1 Answers1

0

Ok, adding the -o option to tftpd helped. The option is described in the man-page thus:

-o      Disable support for RFC2347 style TFTP Options.

Ha-ha...

Mikhail T.
  • 2,338
  • 1
  • 24
  • 55