[lwip-users] Time to live of icmp echo-reply is correct ?

[Claudio Leonel]

Hi all,

I noticed that lwIP returns an echo-reply packet
with the same TTL as the received echo-request packet.
Is this correct ?

I executed some tests in a local network (0 hops)
pinging several types of machines with different
TTLs and got the following results:
- Linux box always returns echo-replies with TTL=64
- Windows 2000 box always returns echo-replies with TTL=128
- An SMSC gateway/router always returns echo-replies with TTL=255

I didn't find any comments in the RFCs about the supposed value
of TTL for echo-replies.

I noticed this behaviour when I did a traceroute
(through the Internet) to a server and to an embedded system
running lwIP in the same network.
The traceroute of the lwIP embedded system had TWO TIMES
the number of hops of the server.

This happens, I think, because the icmp echo-reply packet
sent by lwIP will reach the traceroute system only when
the TTL of the original packet have two times the number
of hops of the route. The first half is spent reaching
the lwIP system and the other half is spent on the
way back to traceroute system.

So, I really think that for traceroute to work OK with
lwIP, the TTL of the echo-reply packet should be 64 or
greater.
Any comments ?


Regards,
Claudio