link dies intermittently-seemingly at random- between win<->linux not linux<->linux??

takahaya · 09-20-2005, 07:57 AM

I have a mandrake 10.1 box/server which I use to firewall a laptop at home and mandrake and winXP installed on the laptop. When I'm connected to the server with linux on the laptop I *never* have any problems. Yet when I boot the laptop into XP the link to the server (and the Internet) intermittently fails. Like once every fifteen minutes or so the link just dies and I get no response when I ping the server. Then, within three or four minutes, the link is back up again.

Has anyone any ideas what might be causing this? I would suspect a faulty ethernet card if it wasn't for the fact that when I connect in linux it never fails.

Has anyone any ideas how I might go about tracing the problem.

It is really frustrating when you're working - as I am at the moment - at getting that piece of sh*t IExploder to render CSS properly in php generated pages and every few minutes puTTy dies on you and you've got to wait around twiddling your thumbs for it to come back up and then get back into the directory on the server etc etc etc.

Any help-ideas-suggestions greatly appreciated!

---------------------------------------------------------------
http://hasperacu.zapto.org

Snowbat · 09-20-2005, 08:03 AM

On the server:
cat /var/log/messages | grep eth

Any interesting errors/warnings?

takahaya · 09-20-2005, 08:20 AM

A *lot* of iptables dropped packets get logged much of which I have trouble understanding since I cobbled together the firewall script from various sources. A lot of stuff gets logged. As far as cat /var/log/messages | grep eth1 (the internal ethernet card) I notice something that seems to me a little strange:

Sep 20 08:58:30 pc65120 kernel: IPTABLES UDP-IN: IN=eth1 OUT= MAC=ff:ff:ff:ff:ff:ff:00:0a:42:6e:d8:54:08:00 SRC=172.24.0.1 DST=255.255.255.255 LEN=419 TOS=0x00 PREC=0x00 TTL=255 ID=12680 PROTO=UDP SPT=67 DPT=68 LEN=399

Sep 20 09:03:59 pc65120 kernel: IPTABLES UDP-OUT: IN= OUT=eth1 SRC=192.168.3.1 DST=192.168.3.255 LEN=247 TOS=0x00 PREC=0x00 TTL=64 ID=4375 DF PROTO=UDP SPT=138 DPT=138 LEN=227

Sep 20 09:09:16 pc65120 kernel: IPTABLES UDP-IN: IN=eth1 OUT= MAC=ff:ff:ff:ff:ff:ff:00:0a:42:6e:d8:54:08:00 SRC=172.24.0.1 DST=255.255.255.255 LEN=413 TOS=0x00 PREC=0x00 TTL=255 ID=12872 PROTO=UDP SPT=67 DPT=68 LEN=393

My laptop ip is set to 192.168.3.2 (notice the second logged udp packet). But what of the other two? Why is the source 172.24.0.1??

(Perhaps I should mention that I don't see any warnings/errors in the messages log.

Snowbat · 09-20-2005, 09:03 AM

172.24.0.1 is part of 172.16.0.0/12, reserved for private networks like 192.168.x.x.
Not sure if I'm reading this right but 00:0a:42 is a Cisco MAC address - do you have any Cisco gear on your network?

Back to the task at hand. I was expecting to see messages about link beat lost when your putty connection dies. Are the machines connected directly with a crossover cable or are they hooked up to a hub or switch? Could XP be pushing the laptop ethernet into a powersaving mode?

takahaya · 09-20-2005, 09:48 AM

The computers are connected through a hub. The 'server' is connected to a cable modem through the same hub. Stupid of me not to guess it's probably the modem that has the address 172.etc

Looking at the settings for 'power options' in the control panel (is that where I'd look?) I can't see anything that you'd think would affect the network.

Thanks for your interest so far.

Snowbat · 09-20-2005, 10:05 AM

A hub, not a switch...so the network runs at 10Mbit? Perhaps XP is trying 100Mbit every so often - you should be able to force 10Mbit in device manager/properties of NIC if you're using a driver from the vendor.

Other than that, as your hardware works fine in Linux, the readers of a XP forum would probably be better equipped to advise on the mysteries of Redmondware.

takahaya · 09-20-2005, 05:54 PM

Why would the network run at 10M because it's a hub not a switch? The hub is the dumb cheapo variety but it does have 10/100M etched on the casing. Anyway, I've had a squizz at the NIC properties (thanks for the suggestion) and it has an 'advanced' setting for link speed/duplex mode, currently set to auto-negotiation. I may just try out the other settings and see whether the link still fails. I'll let you know what happens though like you say, it may be better trying an XP forum (do they exist?)

takahaya · 09-21-2005, 03:51 AM

I was wondering if anyone might know of any decent Windows forums that might be in any way savvy to the problems I'm having?

---------------------------------
http://hasperacu.zapto.org

itz2000 · 07-07-2006, 04:24 AM

do this at the windows when this happens

Code:

ipconfig /release
ipconfig /renew

wait 140+ seconds, and enjoy.
I've heard that someone had this 140 second bug, I still think it's a process who tries to connect and give windows expected failure after 140 seconds.
I know it sound stupid, but try it and lemme know if it's the same bug.

takahaya · 07-13-2006, 05:43 PM

Thanks for the suggestions. I fixed the problem without ever figuring out why it only occurred while Windows was running. The firewall linux box was connected through a hub. Since I only have one computer here I need to connect to it I removed the hub in the middle and ever since it's worked fine.

So it was the hub? Well, I still have the linux box connected to the cable modem through the hub and there's never been any problems there so....

Perhaps one day I'll try your suggestion.

inspiron_Droid · 03-09-2007, 10:37 PM

You should be able to ditc the hub and use the linux router, server & firewall machine directly connected to one of your nickcards on the aftermentioned box.You might have to reconfigure your firewall again but if you have done this once already it shoould be amatter of just taking the hub out of the loop between the machine acting as the router and plugging the routers Cat5 cable into the nick card that linux has designated as eth0 and installing the correct packages to run a dhcp server on eth1 for your internal network.