Another Suse 9.1 + router/NIC problem
 

Hi,

I have just installed Suse 9.1 Pro, dual-booting on a WinsXP machine connected to an external Speedstream 5100 DSL modem.

My original config was 2 NICs, one for the LAN and the other for the DSL connection, with XP running ICS to distribute the net connection round the network. Linux detected the ethernet connection to the net correctly and all was fine.

I then acquired a router/switch, so I installed this and removed one of the NICs. Now Suse will not connect, though XP running on the same PC works fine. If I disconnect the router and connect directly to the modem, Suse connects up fine, but as soon as I replace the router the connection dies.

The router itself is working fine and none of the Windows PCs on the LAN are having any problems with their DSL connections.

My ISP is sbcglobal who provide a dynamic IP. I have configured the NIC for DCHP and entered the router address as the default gateway, I can't think of any other info that Suse might need.

All I get is a message during boot up to the effect that the eth0 interface cannot be set up, but no other explanation. It must be running OK as Suse connects up fine with it when there is no router there.

I'm stuck for ideas as to what to do next and would appreciate any you might have.

Thanks!

After suse having booted to graphical env., even with the eth0 failed, open a console and, as root, do:
ifdown eth0

then,

ifup eth0

Post again and tell me what happened.

sounds like you have a static IP on the Ethernet card you are trying to use to connect to the router.

type the command #ifconfig and post the screen output here.

We can further analyze it afterwards.

My interface is configured for DHCP - no static IP.

The output from the ifup command is:

Starting DHCP Client Daemon on eth0 ... . . . . no IP address yet ... backgrounding


It also issues the same warning message as I get on booting, to the effect that another interface (which does not exist) has not yet been configured. This is the old NIC that I removed when I got the router, as I mentioned above. It has been removed from the config with YaST but keeps reappearing on boot and in the shell console.

Here is some output from a couple of days ago, from the ifconfig and route commands. I didn't do one today as there seems to be no way of copying Suse text files to my Windows disk and I didn't feel like wasting another CD-R on a text file!

eth0 Link encap:Ethernet HWaddr XX:XX:XX:XX:XX:XX
inet6 addr: fe80::240:5ff:fe7c:7176/64 Scope:Link
UP BROADCAST NOTRAILERS RUNNING MULTICAST MTU:1500 Metric:1
RX packets:18 errors:0 dropped:0 overruns:0 frame:0
TX packets:17 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:1000
RX bytes:2225 (2.1 Kb) TX bytes:10030 (9.7 Kb)
Interrupt:5 Base address:0xae00

lo Link encap:Local Loopback
inet addr:127.0.0.1 Mask:255.0.0.0
inet6 addr: ::1/128 Scope:Host
UP LOOPBACK RUNNING MTU:16436 Metric:1
RX packets:556 errors:0 dropped:0 overruns:0 frame:0
TX packets:556 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:0
RX bytes:36456 (35.6 Kb) TX bytes:36456 (35.6 Kb)

linux:/home/chris # route
Kernel IP routing table
Destination Gateway Genmask Flags Metric Ref Use Iface

(I was advised elsewhere to 'X out' the hardware address)

The route command returns no information. If I try to add a default gateway I just get a message that the network is unavailable. I have included the router's IP address in the default gateway box in the NIC configuration, anyway.

This NIC runs perfectly in WindowsXP on the same PC and has required no special configuration.

I tried booting from my Suse Personal evaluation CD to see what that would make of things ... exactly the same result, although of course no errors about the missing NIC as it wasn't preconfigured anyway. No network and unable to access the web interface of the router (192.168.10.10 in my case). It's as if there's nothing there at all. But the live CD did correctly recognise my NIC and configure it for DHCP during startup. It just then does nothing with it.

I know for sure that both Suse versions will connect fine with a direct connection to the modem, so this has to be an issue with recognising the router, but I really haven't a clue why. It's all set up OK, works fine in Windows, and I've checked with the router manufacturers that I have the correct details set up. It's becoming an embarrassment that, having installed the router so I can play around in Linux without disturbing the XP users on my home network, the only box that won't work is the Linux one!

All I can see from the above outputs is that the dhcp client can´t grab an IP. The ifconfig won´t display one, and the dhcp broadcast was supposed to be successful.
When suse was connected to the modem, was the modem the dhcp server, or your isp?
The only thing I can think of is playing around with the dhcp trial time, in /etc/sysconfig/network/dhcp. Maybe its giving up too early.
Also, you might want to remove the missing nic from your config, with yast2.

I'll certainly try to find the file you mention and see if I can edit it. As for the other NIC, I keep telling everyone that I have already removed it from the configuration as well as physically. I have no idea where the boot sequence is finding it from, as it does not appear in the YaST config forms at all (and nor should it, of course).

Without the router, the IP comes from my ISP. I think this is what you are asking - my modem is not the router version but a plain vanilla one without any bells and whistles to play with.

I've got this problem posted on two Linux forums now in the hope that one or the other will come through with an answer; so far nobody's got a clue so don't feel bad if you can't figure it out - nobody else can, either!

Well it doesn't seem to be that either. The default time for DHCP to connect is still set at 999999 seconds.

I did also try setting DHCLIENT_SLEEP from 0 to 20 and DHCLIENT_WAIT_AT_BOOT from 5 to 20.

No change :(

Oh I also checked in the configuration files, and despite the old NIC reappearing in the boot sequence all the time, there is NO configuration file for it in the sysconfig directory :confused:

I see...
Its odd, because no matter where the router (or dhcp server) is, suse is talking to a dhcp server, in the end. If its the new one you just installed or your isps, is irrelevant. At least, was supposed to be.

Some dsl modems do dhcp, thats what I meant. If suse could run the dhcp client and get an IP well, it supposedly, could do the same with the router.

Can you have support from suse? maybe they know whats going on.

My last try: you can workaround the problem by setting static ip's. Is a small network, won't be a problem. Try to disable the dhcp serving feature on the router and set it to static ip. Then, set WinXP to 192.168.0.1/255.255.255.0 If everything goes fine, try to set suse to the same internal ip, 192.168.0.1, etc...

I wouldn't get support from Suse, I didn't buy the boxed version as I couldn't afford it. I've posted at length on the Suse forums though, and drawn a complete blank there too.

With respect, I don't want a workaround, I want a solution after all this time trying to figure out what's going on - I'm sure you'll understand that.

Since my last posting I have run Knoppix and that doesn't connect either, hence this is a "generic" linux issue. I also disconnected the router and connected direct to the modem, adding DSL information to Suse in order to do so, and it connected first time with no problems. I am sure the live CD and Knoppix will do likewise, as they have done before.

I really have little more to add and don't know what else I can post, so here's a summary of where I'm at in case any potential Sherlock Holmes's come along with a solution. Meanwhile I have given up on Linux entirely.

Firstly, the setup:

(1) I have a single NIC which is correctly configured for DHCP and works perfectly on the same PC in Windows XP. I am on a DSL connection via a Speedtream 5100 modem and my ISP, SBCGlobal.net, issues a dynamic IP when I log on.

(2) I have a router and switch which are correctly installed and work perfectly in Windows XP on the same PC.

(3) When I got the router I removed the (now) unused second NIC I had previously installed for my LAN.

(4) All the Windows PCs on my home LAN connect to the net without problems using the same router and switch.

Now the problems:

(1) On boot, Suse (other than the "live CD" version) finds the NIC which is no longer there and no longer configured and issues a warning message that it has not (yet) been set up. It does not appear in YaST nor does it have a configuration file of its own, so I have no idea whatsoever where the boot sequence is finding it from.

(2) The DHCP daemon starts and backgrounds the connection, but I receive an error message to the effect that eth0 cannot be set up. I have tried extending the time for the interface to start up and for the DHCP connection to be established, but in vain.

(3) I cannot access the net, nor can I access the router, nor can I ping anything at all in Suse.

(4) Every time I do an "up" and "down" on the network interface I get a message that /etc/resolv.conf does not exist. It is there but blank. I tried adding the nameservers of my ISP to it but this did nothing. I also get repeated warnings about the non-existent NIC.

(5) Although the NIC is configured for DHCP and is present and correct and correctly identified by Suse, I just get repeated messages that the "network is unreachable"

Logic tells me that since everything works fine if I switch to Windows, there is nothing wrong with the hardware. Since all versions of Linux connect perfectly if the router is not there, logically there has to be an issue with the router and Linux which is not present in Windows 98 or Windows XP. The router is correctly configured for PPPoE, otherwise none of the other Windows PCs on the LAN would be able to connect. I have confirmed with the manufacturers that I have configured the router correctly.

So from logic alone the issue must lie with the router, yet it is correctly configured for PPPoE and DHCP otherwise the other PCs on the LAN would be experiencing the same problems. Logic also suggests that somewhere within the Suse configuration there is a file containing information about the now-absent and unnecessary NIC which I removed last week (it seems an age ago!) yet there is nothing in the files that I can find at all.

The end result is that, since I use the PC primarily for the Net, I will have to dump Linux and go back to Windows because it's proven to be the only OS that works. I am not going to abandon the router and go back to Microsoft's Internet Connection Sharing scheme, as not only is a router a much more elegant and practical solution, it also allows me to turn my PC off now and then and get a bit of peace and quiet.

I have to say I'm deeply disappointed and frustrated at the moment because I'm stuck back with Microsoft again but I've asked everyone I can think of to ask and drawn a blank. A workaround might do, but I've spent years finding workarounds for Windows problems and wanted to move to Linux to get awaty from all that.

Hopefully someone will keep this thread alive and find a solution ... meanwhile, thanks for all the advice and I'm truly sorry it didn't work.

With respect, I don't want a workaround, I want a solution after all this time trying to figure out what's going on - I'm sure you'll understand that.


Of course. Never mind.


. meanwhile, thanks for all the advice and I'm truly sorry it didn't work.


So am I. Its not good to know that such a standard connection type will be so troubled to set.
I'm sure that somewhere on the internet, someone should have found out what is this, and worked on a fix. I'll try to search.

Until a few months ago I was dualbooting here, and there was a strange issue with dhcp, also: when in windows I could grab an IP at boot time and, before switch to linux I couldn't release the IP, so that linux could get one. No idea why. The win commands did their job, but linux couldn't get IP. I had to phone the guys in the isp and ask them to release my ip at the server, everytime. It was a pain.

After a few trials, I discovered that suse won't have such issue. It just grabs an IP, regardless of whether I released that under win or not. I had tried redhat and mdk before, and the same thing kept going, but suse apparently didn't had that. So, I switched to suse. It was the main point to my choice.

Anyway, I was trying to find another thread that may be related to yours. This other guy had problems connecting with suse, but I can find it. If I can, I'll post here.
Cheers,


Obs - And if everything goes well here, I'll be removing my 2nd nic from my pc, and experiencing this problem, too.

A small update...

I had a hard drive failure so I now have a shiny new drive and a shiny new installation of Suse, and it still does not work.

Since last posting I have used Knoppix, Damn Small Linux, and Slax - and all of them connect faultlessly without any manual configuration at all. Suse just does not want to know. I also tried the Suse evaluation CD and that wouldn't connect either, so it is increasingly looking like this is solely a Suse issue after all. I apologise for some misleading info about Knoppix earlier - it did generate an error message when I ran the connection app but I believe this was because it had already connected at boot anyway, something I wasn't expecting. It definitely does connect.

So I'm still stuck. I have mailed the router manufacturers to see if they have any ideas, but since it runs fine with every OS I've tried except Suse, I doubt they'll come up with anything.

One other random thought, I'm a newbie at networking so this may be irrelevant, but the output of the ifconfig in Suse shows no inet address - even though the router is obviously there to issue one. It does show a "inet6" address and I vaguely understand that there are issues with this format. Is it possible that for some reason Suse is only reading the v6 address and not the v4 one and that is causing some problem? All other versions of Linux that I've run do show both the inet and the v6 addresses.

I don't think the IPv6 has anything to do with it at this stage because it looks as though your NIC is recognised by the system - it's just not getting the IPv4 address so therefore the DHCP connection is just not getting set up correctly. If I unplug my ethernet cable do ifdown eth0 and ifup eth0 I get an ifconfig with just an inet6 reference in it too!

Have a look in /etc/sysconfig/network/ where there are a few default files config, dhcp, route and wireless together with the specific files for your NIC which will either be ifcfg-eth0 or maybe ifcfg-eth-id-0:00:11:22:33:44:55 (where 00:11:22:33:44:55 is the HW address of your card.) There's also an extremely useful ifcfg.template which has all the things in it you can set in your specific ifcfg-xxxx.

The bits of interest that should make it work properly are:

STARTMODE='onboot'
BOOTPROTO='dhcp'
DHCLIENT_MODIFY_RESOLV_CONF='yes'
DHCLIENT_SET_DEFAULT_ROUTE='yes'

which you can put into your specific ifcfg-xxxx to make sure they take precedence over any defaults.

If you do that then ifdown eth0 followed by ifup eth0 then it should grab an IP address. You can look in /etc/ for resolv.conf. If it's worked you'll see a temporary resolv.conf created by dhcp and the saved version of the original called resolv.conf.saved.by.dhcpcd.

Hope this helps.

Thanks, that was a really constructive bit of help.

Firstly my NIC file was there but did not have the DHCLIENT ... lines, so I added those, and ran the interface down and up as you suggested. Still no joy. I then looked at resolv.conf and it is empty, just the words "domain site" and nothing else. No other resolv... file present either.

So, I booted up from Knoppix and took a look at the resolv.conf file:

search ISS
nameserver 206.13.28.12
nameserver 206.13.29.12

This immediately suggests to me that Suse simply cannot "see" the network at all. Trying to ping the router returns a "network unreachable" message too. Windows, Knoppix and the other flavors of Linux obviously do see the network and are getting the info back from the router, but Suse is not even seeing the router. So I reckon that whether I configure Suse for DHCP or static IP is irrelevant, as it isn't even recognising that there is a network there.

While this has not solved the problem, it has been revealing about what the problem really is. Thanks for that!

Very strange. Do you have NAT enabled on your router/switch?

In my set up the router is a DHCP server, has an IP address of 192.168.1.1 , allocates IP adresses to PCs in that range and the resolv.conf contains nameserver 192.168.1.1 which is what I had assumed was normal! To share the internet connection my router has to have NAT enabled so that the many local IPs can talk to my one external ISP. Therefore the router is the default gateway and also the DNS forwarder as only it knows about the external IP addresses.

Are you using the same machine for the comparisons?

I know you've been through this before but can you post the results of ifconfig and route for the Knoppix and the SuSE setups so we can see the differences - both via the router and directly attached to the modem?

I'd very much like to help you sort this one out as I don't like unsolved mysteries!