Strange, strange problem...one of those things that makes you scratch your head

H

hardwarephreak

[H]ard|Gawd
Joined
Jul 13, 2002
Messages
1,283
Alright, I have a bit of a mystery on my hands here...like anything in computers and/or networking there is more than one way to fix an issue, so I have the issue resolved, yet the underlying issue is bothering me...

So we have a customer, he is having issues with broadband internet connectivity, namely cable. He comes in, we plug his PC into both of our test connections (DSL and Cable) and the PC works fine. Grabs an IP, surfs the web, resolves names to IPs, no problem.

Customer goes home, can't connect. Cable ISP sends out a tech, his laptop works fine. However the customers system cannot pull an IP (and neither can his router - more on this later). So he brings his PC back in, we test it again, this time a bit more thoroughly, and yet everything still works fine. Customer comes in, again, he verifies that everything is working properly, we release and renew the IP several times while he watches, no problems.

Goes home...still nothing.

So he brings it back. One of the technicians asks to bring down the entire sales floor demo network by plugging the customers system directly into the cable modem. Sure, why not, I give him the OK. Hooks up the customers system, power cycles the modem (the local cable co assigns IPs through the modems by MAC addy, so you can't just unplug and plug into a different PC/Router/etc.) The modem recycles, and what do you know...the system (win98se) can't pull an IP. So all the usual crap ensues, removing and reinstalling drivers, windows networking, blah blah blah...stuff we have all been doing since 98 came out. Still nothing. So I bring up one of our test USB NICs. Install. Still can't pull an IP. Interesting. Try BartPE (based off of XP Pro), still can't pull an IP..on either NIC. So my tech is just looking at me like WTF, this is REALLy weird.

So I start eliminating things we know it isn't...

Well the NIC works fine...it pulls an IP when behind a router of some type. Also if our test NIC doesn't work either, then it definitely isnt the NIC. On top of that, both the USB NIC and the cusotmers NIC cannot pull an IP from the cable modem from within Bart, so it isnt the OS, or a corrupted driver, or registry, etc... So I start thinking, the guy said he has a router...and the router can't pull an IP either...we have the same cable modem (model) feeding the sales floor demos as this guy, how much do you want to bet that the customer's router is bad, and on top of that, how much do you want to bet he has never used his cable connection without a router, and that is why he has never noticed this before.

So the customer goes and fetches his router. Brings it up, and what do you know the thing is bad. Won't pull an IP in several different test settings. So he buys a new router, goes home, plugs it all up and it works. So the customer is happy and the problem is solved.......not quite, not for me.

WTF is going on here...a compatibility issue between the cable modem and the computer...not the NIC...the computer, remember two different NICs, two different OSes and no IP, however, all of the above work fine with a router in the middle. You could say it is something else in the PC, but why, it all works when not directly connected to the modem. And no matter what you did to this PC, swap NICs, swap OSes it didn't work when connected directly to the modem, what the hell would cause the incompatibility.

maybe it was the end of the day, and we were all just tired and not thinking straight...as an after thought I wanted to try Knoppix, but the customer won't bring it all back in just to satisfy my curiousity. What am I missing?

Strange.
 
The Knoppix would have been a good test the Barts boot disk is not fool proff when it comes to working %100. Neither is Knoppix, but you have more to work with if it doesent.

There are also a lot if viruses that mess with win98's stack. Did you do a complete scan with up to date virus definiions?
 
Chaulk it up to win98 oddness. One of my favorites, btw. I've run into several systems where random parts have corrupted libs or bins, so you get some very odd behavior out of the system.

At that point, I usually wipe and reload. Not worth my time to waste diagnosing it further.
 
XOR != OR said:
Not worth my time to waste diagnosing it further.
Click to expand...

qft


sadly...


altho he is saying it was doing the same things on BARTPE... did you try any nics besides the USB and the one that was already installed? what kind was the one already installed? did you try a standard type 3com or intel PCI nic in there?

i've found just about as bizarre problems with networks, and it usually ends up stemming from an odd registry setting, or something specific to your NIC drivers
 
I am really not looking for troubleshooting assistance, just someone to come along and say, well I have seen that or something like that before and it was "x". I think a few of you ( XOR != OR and goodcooper) are missing the point, or did not fully understand the orignal post.

It fails to connect to the modem...in both a 98SE and an XP (Bart) enviroment..

Using two different cards, which use two different sets of drivers for each OS.

However, despite the above...both NICs work without issue in either environment when a router is present between the PC and the Modem.

The issue was resolved...but why would the router make the difference is the question?

Oh and I am quite sure Bart works with the USB nics, as we use them on a daily basis, in and out of the Bart environment. I just wanted to throw knoppix at it just to make things more interesting
 
I did get your point. I was saying win98 was famous around these parts for randomly corrupting it's libs and bins. You'd get all manner of strange behavior after that, not the least of which was network weirdness.

I just mentioned that when this would happen, I'd just wipe and reload, but that was incidental.
 
I've had a nic that would not do web pages....though could do ftp and network shares okay. This was replicated in more than one other machine. We chucked it and put a new NIC in and all was well.
 
The fact that your USB NIC works in other PC's with Barts bootdisk doesent mean it will with this machine. There are several USB chipset vendors and this could easly cause the USB nic to not work.

Is your Barts bootdisk the DOS based one? If so I had thought that there were no DOS USB NIC Drivers.
 
first off i little when it comes to networking, so what i say may be totally not related

but it seem you are saying the nic's would not hook up to his modem, but would to his router (but router was bad and had to be replaced, and once replaced it worked fine)

sounds to me that modem is setup in bridged mode which would not allow a a computer to hook directly to the modem and then needs a router to obtain an ip address and once he had a working router the issue was resolved
 
moetop said:
The fact that your USB NIC works in other PC's with Barts bootdisk doesent mean it will with this machine. There are several USB chipset vendors and this could easly cause the USB nic to not work.

Is your Barts bootdisk the DOS based one? If so I had thought that there were no DOS USB NIC Drivers.
Click to expand...

Please read the both of my posts again. When a router is involved BOTH NICs work, in either his original OS or BART. However, When you remove the router, it does not work. It has nothing to do with USB chipsets in this situation.

pstang said:
sounds to me that modem is setup in bridged mode which would not allow a a computer to hook directly to the modem and then needs a router to obtain an ip address and once he had a working router the issue was resolved.
Click to expand...

Except you can hook a computer directly to the modem and it work...just not his computer...without a router

The question is, why is the router making it work.
 
I know it's a shot in the dark, but the latency involved in pulling a DHCP address over a cable modem is higher than that of the router. Just something to think about. Maybe some piece of hardware didn't handle the higher latency well and the symptom of that problem is what you experienced. Not that I'm right on this, it's just very strange...
 
MorfiusX said:
I know it's a shot in the dark, but the latency involved in pulling a DHCP address over a cable modem is higher than that of the router. Just something to think about. Maybe some piece of hardware didn't handle the higher latency well and the symptom of that problem is what you experienced. Not that I'm right on this, it's just very strange...
Click to expand...

My gut tells me no...I keep thinking the problem is somewhere deep in one of the network layers...I just can't think of what it would be, and why the router makes the difference.
 
a troubleshooting thought that I didn't see mentioned as a test would be:

plug the modem and computer into a switch or hub. Verify link lights and test connectivity.


This issue could be related to auto-negotiation of one of the NICs, but it is doubtful it'd be related to both the internal NIC and the USB NIC.

When you swapped NICs, did you swap them with like kind? Was the machine hurting for speed? Were there any other devices sharing an IRQ with this NIC? Those are most of the thoughts I have on the issue.
 
This is a Windows 98SE bug.

This is the exact fix; it does not always work. Take it up with Microsoft.

Uninstall ALL TCP/IP instances.
Reboot.
Uninstall Ethernet drivers.
Reboot.
Install Ethernet drivers.
Reboot.
Install TCP/IP.
Reboot.
Make sure you have the PPP updates installed. (Just trust me; they screw with stacks.)
Attempt to get an IP again.

Lather rinse repeat.

Here's exactly what happens; the machine sends out a DHCP_REQ. The DHCP server (I'd bet money it's isc-dhcpd or NetID in this case, as they're the worst for this) sends an OFFER. Windows, rather than refusing the offer because it doesn't like it, just sits there looking stupid. You can verify this easily; get the ISP to permit you to set a static IP for testing. Pretty much 99.9% chance of that immediately working. Hell, get an IP with another machine, set it static on the 98SE box, and it'll start working. I have never gotten a straight answer from Microsoft as to the cause. Sometimes the above works. Sometimes replacing the ethernet card works. Sometimes you have to replace the cablemodem. Sometimes it just starts working.
There is no garaunteed fix for this, though Symantec is a garaunteed break - especially Norton AntiVirus. The port-proxy for email scanning exacerbates the hell out of this problem, because it screws with the TCP/IP stack. You have to uninstall, use Symantec's special remover, boot into Safe Mode w/out Networking, and run SFC.

Good luck. You're in for a special brand of hell. :(
 
AreEss said:
This is a Windows 98SE bug.

This is the exact fix; it does not always work. Take it up with Microsoft.

Uninstall ALL TCP/IP instances.
Reboot.
Uninstall Ethernet drivers.
Reboot.
Install Ethernet drivers.
Reboot.
Install TCP/IP.
Reboot.
Make sure you have the PPP updates installed. (Just trust me; they screw with stacks.)
Attempt to get an IP again.

Lather rinse repeat.

Here's exactly what happens; the machine sends out a DHCP_REQ. The DHCP server (I'd bet money it's isc-dhcpd or NetID in this case, as they're the worst for this) sends an OFFER. Windows, rather than refusing the offer because it doesn't like it, just sits there looking stupid. You can verify this easily; get the ISP to permit you to set a static IP for testing. Pretty much 99.9% chance of that immediately working. Hell, get an IP with another machine, set it static on the 98SE box, and it'll start working. I have never gotten a straight answer from Microsoft as to the cause. Sometimes the above works. Sometimes replacing the ethernet card works. Sometimes you have to replace the cablemodem. Sometimes it just starts working.
There is no garaunteed fix for this, though Symantec is a garaunteed break - especially Norton AntiVirus. The port-proxy for email scanning exacerbates the hell out of this problem, because it screws with the TCP/IP stack. You have to uninstall, use Symantec's special remover, boot into Safe Mode w/out Networking, and run SFC.

Good luck. You're in for a special brand of hell. :(
Click to expand...

OK, so lets say just for the sake of discussion the above fixed the problem (just a side note, we did do all of the above). Why on earth would it pull a DHCP address from the router without any problem? Why would it not work with Bart with either NIC? I relize that if its a general problem with the networking subsystem of Windows 98 it would affect any NIC I plugged in, however, Bart is based of off Windows XP Pro, completely different than 98SE, and it exhibited the exact same issue.

Like I said though, the problem was fixed when the guy bought a new router. With the router in place it all works great. I just can't let something this weird pass me by without trying to figure it out.
 
Your issue is baffling but the one that is giving me fits is the lack of reading skills here in this forum. For the smartest group of guys in the [H] to. just. keep. misreading. is killing me. :p

Although I don't know enough about the nuts and bolts of the stacks to give a truly qualified opinion, I think that it does have something to do with a hardware spec. Or maybe something that is only part of the win32 os interaction. It has to be something common to both OS'es (win98se and the stripped down XP of BARTPE) and both NICs that has an issue with the cable modem, something the router doesn't have. What doesn't the router NIC have that the other two combinations do? Or what does it do that other two combo's don't? See, I told you I wouldn't be any help. :)

I don't normally troubleshoot these things this deep but have often wondered about problem with seemingly inexplicable solutions. I face so many in my day to day that I tend to thank the Maker and move on but surely the tools exist that would allow you to monitor and log the handshake process such that one could see where the issue resides. Now I am dying to know the answer as well.

Rhino
 
Exactly, I fix stuff like this all day long...this one just happened to make me wonder a bit more about WHY it was doing what it was doing.
 
oh i read it all correctly, and i noted that it was a perplexing problem in my first post...

my next guess


MTU settings, have you looked into this? granted it seems to be doing the same thing in xp and 98, but who knows, it still may be the problem...

and i wish you would have been able to try a couple more nics besides those two
 
Why would the router solve the MTU problem? Cable internet is like a big WAN (LAN if you want to disect it down to the nodes)...you would have the standard MTU set for compatibility across the board. Or maybe you are refering to something I haven't thought of...
 
You must log in or register to reply here.
Back
Top