TCPIP not functioning properly
 

I bought a Microtel whitebox from Walmart.com (I guess my first mistake) and installed Mandrake Linux 9.1 over the Lindows that was on it. After using the box successfully for a couple of weeks I had to work on a laptop so I took out the network cable without shutting down first and plugged into my laptop. I later found out that my tcpip was dysfunctional. I still had link lights but couldn't ping loopback or given IP address. I reinstalled and that fixed the issue. However, yesterday I had an ungraceful shutdown and now I can't communicate again! I used linuxconf to verify that the settings were still in place and a network stop (start) to try to jumpstart the connectivity but without success.

I'm looking for a solution where I don't have to reinstall the OS each time this happens. Thanx.

Scott.

Whoah dude.

Ok, really what you need is something like "network troubleshooting howto." Networking issues should never require a reinstall.

Try a couple things next time there is a problem and let us know some details.

1. What's the output of /sbin/ifconfig eth0
2. What's the output of /sbin/route
3. What can you ping and what can't you ping? Start with 127.0.0.1. This will tell you if the card is even functioning. Then try your internal IP, try your router's ip, if it has one, try other devices ip's on the network, try pinging external addresses both by IP and by name.

Most connectivity and DNS issues can be diagnosed with these methods, solving is another matter.

also, unplugging the network cable shouldn't cause any problems. It's not like unplugging a ps/2 port. It won't harm anything. Heck, even plugging/unplugging ps/2, serial etc ports can be ok for the most part but not recommended. Do what Gantrep said, post the results here.

Well actually, there could be an unlikely scenerio where the address expires about when he unplugs it and then is reassigned via dhcp to the laptop, he loses settings, and then maybe for some reason has difficult getting them back. But yes typically, loss of a link temporarily would not cause a prob. The protocol is robust and deals well with that kind of stuff. Hell, they call it ethernet for a reason. It's precurser was Alohanet, and relied on radio transmitters for data transfer.

You said you could'nt ping loopback. That could be a funky /etc/hosts prob. I was helping this guy with some network problems once and I started looking around and he had changed /etc/hosts so 127.0.0.1 did not resolve to loopback for some unknown reason. If you can't ping 127.0.0.1, I would say that is almost definately a hardware and driver problem, not a networking problem. Make sure the proper module is loaded.

Being a newbe to linux, it does make logical sense that somehow the driver was corrupted or a config file changed. The first time this happened I disabled the onboard nic, and installed an Intel Pro 100b - eepro driver I believe. Now its happened again with the Intel. Is it easy to uninstall and reinstall the driver without an OS reinstall?

If that is the correct module for your card(google around), you can try to load it by hand with

modprobe eepro

the command lsmod lists currently loaded modules. Uninstalling modules, I am not so familiar with. The command is rmmod, but if the module is loaded and working, i think it might tell you that the module is in use or busy and not do it. Someone else probably knows more about unloading modules than I.

In other words, if when you can't ping localhost, check if the appropriate module is loaded with lsmod, and if it isn't, try loading it by hand and let us know if that fixes it.

OK. I realized that I could not ping 127.0.0.1 nor 192.168.4.5 (itself) but could ping outside the subnet 192.168.4.0 such as 192.168.5.1 (another site). This seems to jive with what I received from the /sbin/route command result:

Destination Gateway GenMask flags IFace
192.168.4.0 * 255.255.255.0 U eth0
127.0.0.0 * 255.0.0.0 U lo
default 192.168.4.1 0.0.0.0 UG eth0

Output from /sbin/ifconfig eth0 is:
eth0 link encap: ethernet HW addr: 00:A0:C9:08:1E:7F
inet addr: 192.168.4.5 bcast: 192.168.4.255 mask:255.255.255.0 UP broadcast running multicast MTU:1500
metric:1
RX packets: 49 errors: 0...
TX packets: 1412 errors: 0 overruns: 1412
Interrupt 11 baseaddr 0xD000

Anyhow, my colleague on the 192.168.3.0 subnet was able to webmin into the box but we're still experiencing issues with sendmail and ssh which maybe because of this comm issue. Is the lack of an explicit gateway for my subnet and loopback the issue?

Wow this is odd. I don't know why you can ping others but can't ping localhost.

Yes, 127.0.0.1 not being accessible can cause some major issues. In fact, the guy I helped who did not have localhost mapped to 127.0.0.1 was also having problems with ssh. Some programs kind of expect to be able to connect to themselves in that manner.

since your ifconfig dowesn't show the interface up have you tried "ifconfig eth0 up"?