[3432] in linux-net channel archive
Re: MTU problems
daemon@ATHENA.MIT.EDU (really kuznet@ms2.inr.ac.ru)
Sun Jun 23 17:05:12 1996
From: inr-linux-net@ms2.inr.ac.ru (really kuznet@ms2.inr.ac.ru)
To: linux-net@vger.rutgers.edu
Date: 23 Jun 1996 23:35:28 +0400
Michael Lausch (mla@gams.co.AT) wrote:
: kuznet> How loki guessed that mss=236 ?
: That's one thing that bothered me. I tried this setup with different
: MRU values for the PPP client.
: ....
: Well loki has a frech compiled linux 2.0.0 kernel. I don't think user
: programs play any role in this, because PMTU is done at kernel level.
Got it. Apparently, PMTU discovery was "completed" :-) before
you started tcpdump. Let us make the following experiment:
- run tcpdump -i eth0 host loki on loki
- change something in route table (it will flush all cached information on MTU)
(I usually make it by: "route add -host 1.2.3.4 lo; route del 1.2.3.4")
- ping -c 1 other end of PPP line
- look at /proc/rt_cache -> MTU for other end of PPP should be 1500
- retry your TCP session.
- look at /proc/rt_cache and result of tcpdump now.
Now we will able to understand from where loki got this impossible
value for mss. OK?
Alexey Kuznetsov.
