1

Is there a utility or linux built in to get the PMTU/MTU of a remote system quickly (sub 1 second preferably)? If so what is it and what syntax does it require?

Google searches have not been my friend today and the only solution I've found is a full blown traceroute which can take upwards of 30 seconds.

2
  • PMTUD depends ICMP messages being returned when a packet is too big for a link in the path, or it depends on you increasing the packet size from a small size until no response is received from the other end. You can manually test the path MTU across a network with tracepath, but I doubt you are going to get sub-1s discovery with anything.
    – Ron Maupin
    Feb 2, 2017 at 16:40
  • Related (not duplicate) question: networkengineering.stackexchange.com/questions/13417/…
    – Dan Pritts
    Feb 2, 2017 at 18:03

2 Answers 2

2

I found a solution after a lot of searching. Just going to link this here for anyone who cares:

https://github.com/ValdikSS/p0f-mtu

It has an API you can call with full socket support. So while technically it's a daemon it does the trick.

1

Things I'd try:

  • Try this nmap script, which uses nmap to perform PMTUD. I haven't tried it, don't know how fast it is.

  • write a small program that opens a TCP connection with PMTUD enabled, and read the value back from the kernel. This could potentially be pretty quick. I don't know enough about how to do this to tell you more, but the question i linked to in my comment has some details that should help get you going in the right direction.

  • if you're limited to existing shell commands, try sending pings of different sizes in parallel. ping -M do -s 1472 remote-host will send a 1500-byte packet (1472 data bytes + ICMP & IP headers). -M do tells it to explicitly do PMTUD. http://muzso.hu/2009/05/17/how-to-determine-the-proper-mtu-size-with-icmp-pings

  • you mention traceroute - if you're trying to figure out where the MTU is limited, but want something faster than traceroute, try mtr; it sends a bunch of packets with different timeouts in parallel, rather than sequentially like traceroute does.

Unfortunately I don't have immediate access to any systems with differing MTUs, so my ability to experiment is limited.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .