Hi,

I have the following hardware:

Desktop with 3 GB Hdd with Mandrake 9.1 installed
Laptop with win98 SE
1 ethernet hub

I hv been trying to setup a small network between the 2 machines for sometime now.. I currently dont hv any internet connection at home so all I am trying to establish is a small LAN ..

Configured the linux network config, setting the ip to 192.168.0.1, the workgroup name is set as "home".

Win98 has an ip 192.168.0.2 and is also located in the same workgroup "home".

How can i setup my network to start running? I am as yet not interested in setting up samba.. just pinging the 2 machines from each other wud be great for now!

pinging the linbox from the laptop gives a request timed out and the viceversa gives a "icmp_sq =1 Destination host unreachable"
and I dont hv any kind of firewall installed.

tia,
sg

Stupid questions first.

Are you pinging using the IP addresses or the computer's name? It does look like you're using the IP addresses.

What kind of cables are you using? I can link up my Linux box to my WinME laptop with no problem. In that case, I use a cross-over cable. If you are using the hub in this connection, then you would use two straight cables.

Do you have the cables plugged into the hub right? One of the hub connections should be for the up-link and should not be used. Also, my hub has the up-link connector tied to the connector beside it, so only one or the other may be used, but not both (yes, I know that trick because I did fall for it!).

Then, do you have TCP/IP installed and working? Can both machines ping themselves? Not just ping localhost, but also ping their own explicit IP address. The former just tests the software of the stack while the latter should also test the NIC hardware.

Yes to all ..:)

I mean..

1, Yes am pinging each other using their respective IPs.

2. Since they r connected to the HUB, both use straight cables.

3. The cables r plugged into slot 4 & 5 of the Hub and not to the uplink.

4. TCP/IP is installed on both.. and the machines can ping themselves via their respective IPs.
I can ping the windows machine ip from the windows machine and the linux eth ip from the linux box
the LEDs light up on both the network cards and at the hub and when i issue a ping command the leds on the hub and nic flicker..

thnx,
sg

Nothing hits me right off. All I can do is toss out a few guesses.

Try different hub slots, like 2 & 3 or 2&4. I'm not sure how your hub is set up and I'd like to shy away from slot 5. BTW, which end is the uplink connector on -- next to 1 or next to 5? Also, if we know what hub it is then we could check the online documentation.

Have you tried eliminating the hub as a possible problem by connecting the two boxes directly with a cross-over cable?

Also, do you have another network device (eg, another computer or a router) that you could try to connect to each one and see if you can ping it?

Oh... this bit me once. Is either box multi-homed, ie has more than one NIC? If so, then they should be on different networks.

Yes, Tried all the slots, the Uplink is next to the Hub slot No. 1. Can get some info on the Hub only on Monday as it is at home.. But its new and I myself had used it at another place successfully...

Dont hv a crossover cable (not sure i want to buy one when i hv a hub and a bunch of straight cables!) and both machines hv only single NIC cards... and no, dont hv any extra network device to test. atleast not for now...

Further, I noticed something strange.. as I said before, the win 98 laptop has an ip 192.168.0.2 and the linbox 192.168.0.1 and as soon as i connect the cables to the HUB, the LEDs light on both sides and I could ping each machine by itself. BUT when I disconnected the cables from the HUB, I still could ping each machine by itself!!! IS this possible??

Can I still ping each machine by its own IP address even if they r not connected to the HUb ???

thnx,
sg

Yes, you can ping your own IP address without having a cable connected (I just did it in order to prove it empirically). I'm not completely sure of the exact electronics, but there must be a kind of loop-back connection in a NIC so that it also reads its own transmit line -- a Cat-5 cable has separate transmit and receive lines, which is why there's a difference between cross-over and straight cables.

The addresses certainly seem right. I'm sure that the subnet mask is either 255.255.0.0 or 255.255.255.0, so they are on the same network. Just as a stupid question, could you please carefully read and verify that both the IP address and subnet mask on both machines are correct?

Another stupid question: Are you sure you have two straight cables? I have yet to see an indication on a cable to identify how it's wired; cross-overs are often red or reddish, but not all red cables are cross-overs.

For GPs, try connecting them directly first with the cable and then with the second. Just to eliminate that possibility.

Links between network devices (ie, between routers and hubs, but not between routers/hubs and hosts) normally require cross-over cables. Some hubs have a switch to change the uplink connecter from cross-over to straight -- I think that the pairing-up of two connectors is the other way of providing this capability. As an alternative test, you should be able to use the uplink connector to "create" a crossover.

Then, are the NICs' indicator lights visible? When you try to ping one box from the other, do the destination's indicator lights indicate activity? That should at least show whether the signal is getting to the destination.

I hope somebody else jumps in here soon, because I'm running out of ideas.

Just a quick note before i leave for the weekend!

Everything that you have suggested has been checked out, not once , not twice but more than a dozen times.. (its all hv been doing the whole of this week!!)

Never been more frustrated (atleast so far!)

thnx for the suggestions.. if anything crosses ur mind, pl lemme know...

thnx,
sg

The frustrating thing for me, besides not being able to figure it out, is the sneaky suspicion that it's due to something fairly simple that would jump out at me if I were on the scene looking at it.

OK, how's about using tcpdump to see what's going on? If you don't have tcpdump already installed, it should be in a package on your distro CDs. Otherwise, I'm sure it can be found on the Web.

Run tcpdump on your Linux box. I'm pretty sure you'll need to be root to do this, or even to find it with which or whereis. Because you only have the two computers on the network, there shouldn't be any need for you to apply filters; just run tcpdump with no arguments.

Then try to ping the Linux box from the Win98 box. Here is a capture of when I pinged my Linux box from my Win98SE box:

[root@pc10593 /root]# tcpdump icmp
Kernel filter, protocol ALL, datagram packet socket
tcpdump: listening on all devices
11:33:52.797643 eth0 < pc14402 > pc10593: icmp: echo request
11:33:52.797778 eth0 > pc10593 > pc14402: icmp: echo reply
11:33:53.800172 eth0 < pc14402 > pc10593: icmp: echo request
11:33:53.800272 eth0 > pc10593 > pc14402: icmp: echo reply
11:33:54.805115 eth0 < pc14402 > pc10593: icmp: echo request
11:33:54.805198 eth0 > pc10593 > pc14402: icmp: echo reply
11:33:55.810058 eth0 < pc14402 > pc10593: icmp: echo request
11:33:55.810150 eth0 > pc10593 > pc14402: icmp: echo reply

8 packets received by filter
[root@pc10593 /root]#

My Win98SE box is pc14402 and my Linux box (RH7.0) is pc10593. Because I'm on a big network, I applied a filter to only look at the icmp protocol, which ping uses, and then did a ^C when my Win98SE completed its ping. You will want to apply tcpdump without the "icmp" after it, because you want to know if ANYthing makes it across.

Similarly, when I had my Linux box ping its own IP address, I got:

[root@pc10593 /root]# tcpdump icmp
Kernel filter, protocol ALL, datagram packet socket
tcpdump: listening on all devices
11:49:28.145811 lo > pc10593 > pc10593: icmp: echo request
11:49:28.145811 lo < pc10593 > pc10593: icmp: echo request
11:49:28.145955 lo > pc10593 > pc10593: icmp: echo reply
11:49:28.145955 lo < pc10593 > pc10593: icmp: echo reply
11:49:29.141927 lo > pc10593 > pc10593: icmp: echo request
11:49:29.141927 lo < pc10593 > pc10593: icmp: echo request
11:49:29.142027 lo > pc10593 > pc10593: icmp: echo reply
11:49:29.142027 lo < pc10593 > pc10593: icmp: echo reply
11:49:30.141902 lo > pc10593 > pc10593: icmp: echo request
11:49:30.141902 lo < pc10593 > pc10593: icmp: echo request
11:49:30.141997 lo > pc10593 > pc10593: icmp: echo reply
11:49:30.141997 lo < pc10593 > pc10593: icmp: echo reply

12 packets received by filter

So you can use that to verify tcpdump's operation and/or other tests you can think of.

tcpdump has also been ported to Windows as WinDump, as has a packet-sniffer front-end, Ethereal. For the time being, tcpdump would probably be easier to get set up.

BTW, you could also use tcpdump to capture the packets as you ping from your Linux box to the Win98SE and/or try other networking operations.

One that comes immediately to mind that you should look at are the ARP messages. The sender will broadcast an ARP request in order to get the MAC address of the destination NIC, but only if it's in the same network and it's not already in its ARP table (run arp from the command line to view manipulate the ARP table). If the IP address is on a different network, then the sender will not attempt an ARP, but rather will send the outgoing packet to the gateway router.

Now THAT could be a worthwhile test. See if the sender of the ping does an ARP or sends it to the gateway instead. Be sure to check the ARP table before you try. You may have to clear the ARP table in order to do the test. Come to think of it, if the other box is already in the ARP table, then you know it had been able to do it before. The arp command exists on both Linux and Windows.

BTW, what is the address of the gateway router? Common practice is to make it host 1 -- eg, 192.168.0.1 -- , but that's the IP address you gave to the Linux box. If the two got set to the same IP address, I don't know what would happen. Ditto the DNS server addresses.

Hi dwise1_aol,

Thnx a lot for your help.. Solved the problem but am still unable to believe in what I found.. Over the weekend, I brought over 2 more Hubs and All of a sudden I could ping the machines with one of the HUB.

What makes it difficult to believe is that the 2 other HUBs were purchased more recently and are working fine in another network !!!! and the LEDs also light up fine when the cables are connected...

Baffled with this problem though happy to hv solved it.. If u hv any ideas as to why this is happening pl share it with me..

thnx a bunch,

Sg

The only thing I can think of is that it might have something to do with the hub not supporting (or properly supporting) 10/100 auto-detect. Another engineer encountered it here at work when he discovered that our ethernet board (which has auto-detect) refused to work on a couple hubs. Our supplier's engineer told him that might be the problem.

Glad you finally got that up and running.