Networking problem

[Shane Dawalt]

On 05/14/2011 10:59 PM, James McKenzie wrote:
> On 5/14/11 7:41 PM, Kevin J. Cummings wrote:
>> On 05/14/2011 10:09 PM, JD wrote:
>>> On 05/14/11 18:45, James McKenzie wrote:
>>>> On 5/14/11 6:40 PM, JD wrote:
>>>>> On 05/14/11 18:24, Joe Zeff wrote:
>>>>>> On 05/14/2011 01:27 PM, JD wrote:
>>>>>>> I also brought the fedora firewall down, and retried to ping Fedora
>>>>>>> from Powerbook. No go!!
>>>>>> That means that it's not a firewall issue.  Check your router config to
>>>>>> see if it's set to allow pings inside the LAN.
>>>>> Thanx!
>>>>> I checked. The gateway has a built-in feature (program)
>>>>> to let you ping any client on the lan (or any ip on the public net).
>>>>> The gateway can ping both the powerbook and the fedora pc.
>>>>> no problems there.
>>>>> The fedora pc and the powerbook can ping the gw, and a third machine
>>>>> connected to the GW by ethernet, and can of course ping addresses
>>>>> on the public net.
>>>>> They (fedora pc and powerbook) cannot ping each other!
>>>>> Powerbook firewall is set to promiscuous mode.
>>>>> And as I had stated earlier, I even stopped iptables on the
>>>>> fedora pc, which puts it also in promiscuous mode (I assume).
>>>>> Still these two machines refuse to talk.
>>>>>
>>>> Can you use traceroute to communicate between the two of them?
>>>>
>>>> James McKenzie
>>>>
>>> Tried it.
>>> Tracerout is unable to get to target after 30 tries.
>>> All it shows is asterisks.
>> Sounds to me like traceroute is trying to go "direct" between machines....
>>
> That it is if the two devices exist on the same subnet, which is a bad
> thing for wireless.  Your suggestion on how to solve this is 'spot on'.
> Unless traffic is local, it should go to the gateway on wireless.  Wired
> is much different.

   Wired and wireless are essentially the same.  They all adhere to the 
IEEE 802.3 spec and the 802.1 bridging paradigms.  Copper, light, air 
... it's all in the physical layer as stipulated in IEEE 802.3.  Once 
the electrical signals are delivered, then it becomes IEEE 802.1 for 
bridging and all the RFCs for IP and ICMP, TCP and UDP and etc.  You can 
build a network with 254 hosts (192.168.1.1 - 192.168.1.254) using both 
wired and wireless and it works quite well.  I've done that in my house 
for the last 10 years.  I work at a place where we have 40 buildings 
with 20 wireless networks that span virtual-LANs across our campus 
connecting to routing interfaces throughout the campus.  But, as one 
poster has already pointed out, there are different modes of the radios 
that can cause problems.

   However, the OP appears to be pinging wired machine (having an en1 
interface) from a wireless machine (having a wlan interface).  If this 
is true then the radio mode shouldn't make any different.

   I'm almost ready to break down asking for ARP information.  But I 
think we have one other test before diving deep into the dungeons.

   One thing confuses me here.   The OP is trying to ping 192.168.1.60.  
Yet, the physical interface of the ifconfig listing for machine at 
192.168.1.60 shows en1 as having 192.168.1.70.  In fact, there is no 
physical interface on that computer having 192.168.1.60 configured on 
it.  However, there is a route table listing that says to get to 
192.168.1.60, packets should be routed to localhost (127.0.0.1).  Here's 
an excerpt of what I am describing:

192.168.1.60       127.0.0.1          UHS         0        0    lo0


   So, the next question is; is there something in the host that is 
actually listening for packets destined for 192.168.1.60?  Because at 
this point, it looks like any packets destined for 192.168.1.60 hit 
127.0.0.1 and then die for the lack of a responder.  The machine won't 
respond, because there is no physical interface having an address that 
matches the destination of the packet.

   My next test would be to try pinging 192.168.1.70 from the 
192.168.1.108 machine and see what happens.  (Of course, modify the 
iptables of the 192.168.1.108 machine appropriately.)

   Shane