Text 2491, 54 rader
Skriven 2011-03-25 20:10:16 av glen herrmannsfeldt
Ärende: Re: Linux is stealing the internet from eCS...
======================================================
posting-host="VhYZ3ZDVkI1WvjO/5Jzlvw";
mail-complaints-to="abuse@eternal-september.org";	posting-account="U2FsdGVkX18C07iwJIuvHY7mtIukfiKt"

(x86_64))
comp.os.os2.networking.tcp-ip:185
From: glen herrmannsfeldt <gah@ugcs.caltech.edu>

In comp.os.os2.networking.tcp-ip Pierre Jelenc <rcpj@panix.com> wrote:

> I have just set up a Linux machine running Ubuntu 10.10/Gnome (that's the 
> way it came, I haven't even attempted to modify anything.) It plugs into 
> port #2 of a Netopia router, while the eCS machine uses port #1. It seems 
> that almost as soon as I do something via the internet on the Linux 
> machine, it disables completely internet access from the eCS one: http, 
> ssh, telnet, ftp all time out with no explanation, while things look OK on 
> the Linux side.
(snip)

> One peculiar thing is that both computers claim to be at 192.168.1.105 
> (from ifconfig lan0/eth0) which would point to a competition, but why 
> would they both get the same address? Why 105 rather than 100 or 101? 

That was about to me my guess.   The why is a good question, my 
guess is more than one DHCP server, but there are other possibilities.

As someone mentioned, another possibility is more than one with
the same MAC (ethernet) address.  With built-in ethernet interfaces,
that isn't so likely, but some years ago there were reports of
NICs with the same address coming from one company.  I have seen
the address 00:00:00:00:00:00 on nets before, (so far only one) that
presumably came when someone erased the NVRAM and didn't store
a new address.  

Many DHCP servers start with 100 for the last octet, leaving
the lower addresses for static assignment.  (Usually .1 for the router.)

Hosts likely to be used as servers should have static addresses.

> Why does Linux always seem to win regardless of whether it 
> is plugged in port #1 or #2, or whether it is not even 
> plugged in until well after eCS has already grabbed 192.168.1.105 
> and been shown to connect without problem?

The later host always wins (until the ARP cache times out).
Actually, it is more complicated, but it is ARP related.

I first saw this booting diskless Suns, accidentally assigning
the same address to two of them.  The later one takes over the
connection, leaving the first one stuck.

-- glen

--- Internet Rex 2.31
 * Origin: A noiseless patient Spider (1:261/20.999)