Topic: eth0 timeout problem

[s. herrera]

ETH0 TIMEOUT PROBLEM:
Hi Folks!  I'm currently using the e-smith server/gateway and have been enjoying the may features of this great product and the benefits of having a cable modem tied to my home network.  I have seen a recurring error which I hope someone can help me resolve.  

This error causes e-smith to stop communicating with the other computers tied to the hub although communications to the internet continue via the network interface card (eth1) - i.e. the internet test is always successful.  The error message I get displayed on the e-smith is always the same: "ETH0: transmit timeout, status 0d 0000".  The transmit timeouts seem to occur randomly.  I have recorded e-smith up-times of over 24 hours, others are 10, 8, and 4 hours.

Before I continue, let me provide additional details on e-smith hardware configuration.  I have two identical OvisLink Ethernet adapters( RealTek 8139) NICs installed on a Alton 575 MB with a CYRIX  M2-300 CPU and 32MB of memory.  Admittedly this is "cheap" hardware but I assure you that this configuration has been running linux off and on for over one year without problems.   In addition, I have looked at http://cesdis1.gsfc.nasa.gov/linux/drivers/rtl8139.html and the author of the RealTek 8139 driver, Donald Becker, does make the statement on the web site that "the RTL8129 series is a low-end design, and thus should be considered a "connectivity solution" rather a performance-oriented product. (That's a nice way of saying that the design sucks, but the price is right.)".  Mr. Becker also indicates that "..if you encounter Rx overflow errors and transmit timeouts you likely have the card in a non-bus-master slot".

I have taken the actions of switching slots and even switching the NIC used for ETH0 without solving this problem.  The alternate NICs installed as ETH0 included a RealTek 8029 and the Intel PCI EtherExpress PRO and both .  The confusion I have with this problem is that the timeout problem only manifests itself on ETH0 and never once on ETH1 (the NIC tied to the cable modem).  Has anyone else using e-smith seen a similar problem?  

Any suggestions would be welcomed.  Thanks in advanced.

Sal

[Joseph Morrison]

ETH0 TIMEOUT: Hello, Sal! My initial theory is that you have your ethernet card in a non- busmastering slot. Do you know for sure whether all of your PCI slots are busmastering? Your motherboard documentation would specify this. Some older motherboards have only one or two busmastering slots (typically the middle ones). In newer motherboards all PCI slots are busmastering. Both of your ethernet cards must be in busmastering slots. If one of them is in a non-busmastering slot, you might get timeout problems regardless of the model of ethernet card, and depending on your slot configuration, it wouldn't surprise me if the problem always occurred with eth0. (The card in the lowest slot number is always assigned eth0.) Hope this helps.

[sal]

Joseph, thank you for the help.  I did review the motherboard documentation and
the manufacturer claims that the four PCI slots have bus mastering capabilities.  I've also checked
the BIOS and the PCI is configured to auto/PNP.

I have continued to observe the ETH0 timeout problem, even after I switched the PCI slot
for the NIC tied to my hub.  Since my posting, I have obtained the following up-times for the   
E-Smith Server and Gateway (ESSAG): 2 days, 17 hrs; 6 hrs; 22 hrs.  It may be that my cheap
hardware configuration is causing some conflict, so I'll continue to monitor and provide you any new
information.  Meanwhile, if you folks have any other suggestions, I would welcome the input.  Thanks
again for your help.

Sal

[Kim Morrison]

ETHO Timeout: SAL, thanks for the update. Joseph and I talked about your problem some more.  We've been unable to figure out how the e-smith software    could generate an intermittent, ethernet level problem like this.  There are a few other possibilities for intermittent hardware failure. If it    isn't bus-mastering issues, other possible tests might be to try (1) plugging the cable into a different port on the hub, (2) testing the ethernet cable    between the hub and the server (3) and finally, try to connect only one client machine to the e-smith server and gateway using a cross over cable to see    whether the hub is the problem. Anyone else have ideas? Let us know how it comes out, Sal.

[Sal Herrera]

ETH0 Timeout Resolved:
Hi Folks.  I belive I've isolated the problem of the ETH0
timeout to my Alton motherboard.  It seems you were
right in your recommendations.  After several additona
timout problems with the ETH0 NIC, I decided to change
the motherboard in the E-Smith server.  Since replacing
the Alton motherboard with an Abit TX5 motherboard, I
can joyfuly report that the server has now been running
for over 4 days, 9 hours.

I want to thank you folks for your assitance in the resolution
of this problem.

Sal

[Kim Morrison]

ETHO Timeout Problem: Sal, thanks for the update. I'm so glad the problem was resolved!  K.