LinuxQuestions.org
Go Job Hunting at the LQ Job Marketplace
Go Back   LinuxQuestions.org > Forums > Linux Forums > Linux - Hardware
User Name
Password
Linux - Hardware This forum is for Hardware issues.
Having trouble installing a piece of hardware? Want to know if that peripheral is compatible with Linux?

Notices

Reply
 
Search this Thread
Old 10-14-2004, 09:01 PM   #1
marcheikens
Member
 
Registered: Aug 2003
Location: Cleveland, OH
Distribution: Ubuntu 6.06
Posts: 66

Rep: Reputation: 15
Network switch causing problems?


This isn't a Linux problem, I don't think, but as this is a pretty vast knowledge base, I figured I'd ask it:

I recently bought a new 10/100 network switch, the Encore ENH908-NWY, to replace my old 10mbps hub. The price ($13) was right and the user reviews at newegg.com were entirely favorable. The connections in my network are as follows: Motorola SB5100 goes to Linksys BEFSR11 router goes to switch goes to a Windows XP machine, my Linux box, and my wireless access point.

Everything worked like a charm for about 18 hours, and then while I was connected to my computer from work, the connection went down. When I got home, the connection lights on the switch (only one light per port) were all slowly blinking in unison. I tried power cycling the switch, but that didn't work, and eventually I found that the problem lies with the router--power cycling the router temporarily fixes the problem for anywhere from 20 minutes to 8 hours. Is there some way that a router and a switch can be incompatible, or is the switch somehow messing things up? I've been searching and I can't find anything regarding this. I upgraded the firmware of the router to the latest version (Aug 31 2004), and it still happens.

Any ideas?
 
Old 10-14-2004, 10:29 PM   #2
mritch
Member
 
Registered: Nov 2003
Location: austria
Distribution: debian
Posts: 667

Rep: Reputation: 30
switches (and also the ethernet ports of your router) are some kind of intelligent - they autodetect the speed of a link, some routers check if a twisted or straight cable is connected and they have sort of route remember function so packets will be forwarded to the right port. maybe there is a problem related to one of these features. eg. one can't interact (understand) the other and is messing things up.
can you give any information (from the manual) what the unison blinking on your switch mean?
as restarting the router (which one - linksys/wireless?) fixes it i'd try to use a twisted (it's likely a straight now) cable as first step. if it works this way you're lucky.
the best thing to do would be to turn off autonegotiation/cable-recognizing to be sure here is the problem. however, it's not very common for switches to support that. but as routers sometimes run linux you maybe can do it there if you have access to a console (via telnet).
even better: bring it back and try another ( i assume it worked properly with the 10mbps).

sl mritch.
 
Old 10-14-2004, 11:37 PM   #3
marcheikens
Member
 
Registered: Aug 2003
Location: Cleveland, OH
Distribution: Ubuntu 6.06
Posts: 66

Original Poster
Rep: Reputation: 15
Well, I've tried using a straight-through cable and a crossover cable. I think the router itself (single port wired only from 2000, I believe) is simply overheating all of a sudden. I'm running a bit of an experiment--repositioned it to be vertical with plenty of airspace around it to see if I can't fix the problem that way. I'll see if that works overnight (downloading a large file from slow server). If it does, it's probably time to seek out a new router. Well, let's be honest, it was probably time to seek out a new router long ago.

With regards to negotiation, mii-tool didn't have any trouble re-negotiating a (100mbps full duplex) connection. Nor does my brother's WinXP machine.

Thanks--

marc
 
Old 10-15-2004, 01:13 AM   #4
mritch
Member
 
Registered: Nov 2003
Location: austria
Distribution: debian
Posts: 667

Rep: Reputation: 30
Quote:
Originally posted by marcheikens

With regards to negotiation, mii-tool didn't have any trouble re-negotiating a (100mbps full duplex) connection. Nor does my brother's WinXP machine.
yep, but if the link run dead for short and both ends concurrently are trying to autonegotiate/link-select the other and it's a stupid device...

iirc i've read about heatproblems with belkin routers.
as rma loss won't be your pain - take off the case.

sl mritch.
 
Old 10-22-2004, 12:52 PM   #5
marcheikens
Member
 
Registered: Aug 2003
Location: Cleveland, OH
Distribution: Ubuntu 6.06
Posts: 66

Original Poster
Rep: Reputation: 15
So this is still a problem. I separated all the components, and this seems to have helped a bit. The connection only seems to cut out a couple times a day. It seems to be related to bandwidth being used to the router, i.e. when I download at a sustained high speed (~300-400 KBps) for a minute or two, the connection craps out. [Edit: I'll have to try this on the WinXP machine to see if it's a computer-specific problem]

On my switch, lights on indicates a connection. Blinking light indicates access (Tx/Rx). Interestingly, when I unplug the ethernet cable from the switch that goes to my Linux box, all lights stop blinking. Plug my computer back into the switch, and lights resume blinking and I still can't connect. Power cycle the router, and it works again. So is my computer or network card somehow causing the problem? I wish I still had windows installed to test whether it's the network card or drivers or something, but that's long gone. I have a DLink DFE-530TX+ 10/100 card if it matters, using the RTL-8139 drivers.

I'm looking at routers right now, mritch, and once I'm sure I can get one quickly to replace the current one, I'll try and take the current one apart.

Last edited by marcheikens; 10-22-2004 at 12:53 PM.
 
Old 10-23-2004, 05:14 AM   #6
mritch
Member
 
Registered: Nov 2003
Location: austria
Distribution: debian
Posts: 667

Rep: Reputation: 30
the tcp stack works somewhat different on each os. you may look if some ipv4 settings in your /proc/sys directorie is causing this.

http://ipsysctl-tutorial.frozentux.n...-tutorial.html

mritch.
 
Old 10-24-2004, 10:28 AM   #7
marcheikens
Member
 
Registered: Aug 2003
Location: Cleveland, OH
Distribution: Ubuntu 6.06
Posts: 66

Original Poster
Rep: Reputation: 15
I'm browsing your link, but are there any specific things you'd suggest looking at? Today I'm downloading a large file with wget, limiting the bandwidth used to 100KB/s, and it seems to be working...any more ideas?

Thanks for taking the time to help--

marc
 
Old 10-24-2004, 11:05 AM   #8
randyding
Member
 
Registered: May 2004
Posts: 552

Rep: Reputation: 31
Hi, Regarding these switches I've noticed some similarities between your issues
and issues we found doing embedded microcontroller devices talking low-level
ethernet. We hired a netgear engineer come into our lab to help us troubleshoot
and this is what we found out.

Netgear's switches, and others too, can have lots of problems with these
so-called auto-detect auto-rate detection. First of all our embedded devices
were 10Mb/s, but the network was 100Mb/s, so the switch was buffering the
packets and retransmitting them at a different rate. It turns out that
when the embedded device or network pc responded too quickly to an
ethernet packet the packet was dropped inside the netgear. We had
to insert delays into the software for it to work. This didn't happen
when the network was forced to be 10Mb/s just like the embedded device.
I would assume it also wouldn't happen if the embedded device would
be 100Mb/s, but we didn't have that kind of hardware to test it.

Another huge problem with these switches was when the mac-address
of a embedded device changed, when one device was swapped for another.
The smart switch would stop routing packets to the embedded device
because it didn't realize that the mac address changed on that port.
It wasn't until the embedded device output a new unsolicited ethernet
packet that the connection went up again. This problem normally does
not happen with PCs connected to switches because when they boot
they automatically output some form of communication (like dhcp)
that tends to "Wake up" the switch equipment.

I think this is what your probably seeing with your router to switch
setup. If router reboots, or its been over some timeout period with
no packets, the switch forgets about your device. Then you can't
connect to it until some unsolicited traffic goes the other way.

I know this sounds strange but we spent 2 weeks with a lot of smart
people troubleshooting these problems in a lab with very high-
tech hardware ethernet scopes.

Sorry I don't have a solution for you, just wanted to let you know
what kind of issues there are with some of the lower-cost switches
on the market these days. BTW, we now tell our customers that use
our embedded devices to only use 10Mb/s dumb hubs to connect
their devices... and this works perfectly. No issues. Bummer.
 
Old 10-24-2004, 12:02 PM   #9
marcheikens
Member
 
Registered: Aug 2003
Location: Cleveland, OH
Distribution: Ubuntu 6.06
Posts: 66

Original Poster
Rep: Reputation: 15
What of the fact that this seems to be related to bandwidth usage? Fullspeed for longer than a few minutes causes this error, limiting the bandwidth to some undefined "lower" level seems to keep everything okay. I also experience this when using BitTorrent--any chance this could be caused by too many connections? Never had this trouble before...

btw, thanks for your explanation--I think I may have to stick to the dumb hub for now, but I was hoping to use the 10/100 so I could transfer files much faster between computers on my network.

Addendum: transferring files on the network at full speed via sftp--5MBps--causes no problems

Addendum 2: Strike that about lower levels. It crashed after about 3h45m of downloading at 100 KBps.

marc

Last edited by marcheikens; 10-24-2004 at 02:01 PM.
 
Old 10-24-2004, 02:02 PM   #10
randyding
Member
 
Registered: May 2004
Posts: 552

Rep: Reputation: 31
Gut feel makes me think the issue is related to lower layer 2 communication problem between switch and one of your routers, not TCP/IP, so probably not a "too many connections" issue. It wasn't clear to me but is one of your devices 10Mb/s, is that motorola unit a cable or dsl router? Sometimes those are fixed at 10Mb/s, I know my DSL router is for example. The bottom line of our testing showed that with some of the lower cost home switchs they tend to have more traffic related glitches when you mix 10/100 on the same switch, and the number of packets per second is higher.

Oh ya, I forgot about this other example, I had at one time a linksys BEFSR41 router in my home network and every time I brought home and connected someone elses windows computer and hooked it up to this router, windows would about 3 times a minute popup a tray icon message saying the network disconnected.. then connected... then disconnected. This happened no matter who's computer I hooked up. One day I right clicked on the icon and changed the windows driver to force 10Mb/s instead of autodetect 10/100 and the problem went completely away. This never happened with linux, even when I switched the hardare over to linux from windows the problem goes away. Very strange, but makes me think more the problem is layer 2 issues with the router I had.
 
Old 10-24-2004, 02:39 PM   #11
marcheikens
Member
 
Registered: Aug 2003
Location: Cleveland, OH
Distribution: Ubuntu 6.06
Posts: 66

Original Poster
Rep: Reputation: 15
Okay, I'm not very (okay not at all) familiar with TCP/IP terminology...layer 2?

Also, is there a way to force 100Mbps in Linux without negotiating?

Here's how components compare: Not sure if the Motorola Modem is a 10 or 100 Mbps connection, but the router's WAN interface is strictly 10. It's LAN interface is auto 10/100. Switch is 10/100. All computers are 10/100, as is the WAP (Dlink DWL-800AP+).

Thanks--btw, do you think it's possible a new router would help or fix the issue?

marc
 
Old 10-24-2004, 03:43 PM   #12
randyding
Member
 
Registered: May 2004
Posts: 552

Rep: Reputation: 31
Layer 2 is the MAC layer, it consists of a raw ethernet packet.
TCP/IP packets are broken up into one or more ethernet packets.
The ethernet controller hardware (like the NE2000 chip, whatever)
negotiates go put these packets on the wire. There is 14 bytes of address
and length information followed by up to 1500 bytes data.

I'm afraid to advise on your hardware, I don't want to cause
you to buy new hardware that doesn't solve your problem.
I'd try to borrow someones router that is all 10/100 and see
what it does before buying one. Maybe you could just eliminate the router
for testing purposes, substitute in a borrowed switch or 10/100 hub.

Maybe try what I did with the windows computers, go into their driver
configuration dialog and turn off the 10/100 autosense, make it 10Mb/s only,
just to try it then put it back if it still has the problem.

Good luck.
 
Old 10-24-2004, 03:46 PM   #13
marcheikens
Member
 
Registered: Aug 2003
Location: Cleveland, OH
Distribution: Ubuntu 6.06
Posts: 66

Original Poster
Rep: Reputation: 15
It sounds like both of you are suggesting a problem with rate negotiation. I ran:

Code:
mii-tool -F 100baseTx-FD eth0
I hope this will work. Now instead of saying something about a negotiated rate, I get:

Code:
root@pants:/home/marc# mii-tool eth0
eth0: 100 Mbit, full duplex, link ok
We'll see.

marc
 
Old 10-24-2004, 06:19 PM   #14
marcheikens
Member
 
Registered: Aug 2003
Location: Cleveland, OH
Distribution: Ubuntu 6.06
Posts: 66

Original Poster
Rep: Reputation: 15
Okay, so that did NOT work. Still the problem. Interestingly enough, I ran
Code:
mii-tool --reset
and rebooted. After rebooting dmesg shows that my connection is running at 10Mbps/half duplex!
Code:
mii-tool --restart
correctly fixed the speed and duplex. Don't know what's going on. Interestingly enough, when the connection "dies," mii-tool reports everything's okay.

marc
 
Old 10-24-2004, 07:10 PM   #15
randyding
Member
 
Registered: May 2004
Posts: 552

Rep: Reputation: 31
At this point I'd go modem to computer directly with a new ethernet cabel and test that. Then add in one additional piece of hardware one at a time until the problem starts up again. Try to ID by trial and error what piece of hardware is the bummer here.
 
  


Reply


Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is Off
HTML code is Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
IPtables firewall causing network slowdown lucktsm Linux - Security 3 05-11-2004 11:37 PM
network problems causing KDE to hold while loading Yukanojo Linux - Wireless Networking 0 03-17-2004 09:37 PM
9.2 installation causing network problems centerice Linux - Networking 1 02-05-2004 07:21 AM
Switch causing Quake3 drop outs?? breakerfall Linux - Networking 0 01-18-2004 01:00 PM
Network causing unhappiness! Noobee Linux - Networking 4 07-18-2002 02:01 PM


All times are GMT -5. The time now is 12:11 AM.

Main Menu
My LQ
Write for LQ
LinuxQuestions.org is looking for people interested in writing Editorials, Articles, Reviews, and more. If you'd like to contribute content, let us know.
Main Menu
Syndicate
RSS1  Latest Threads
RSS1  LQ News
Twitter: @linuxquestions
identi.ca: @linuxquestions
Facebook: linuxquestions Google+: linuxquestions
Open Source Consulting | Domain Registration