forcedeth freezes my NIC. How do I compile newer version?
 

My mainboard: MSI K9N SLI-2F (with dual gigabit ethernet)
My OS: Ubuntu Desktop 7.04
My driver: forcedeth 0.59

This happens:
eth0 "dies" on heavy load
i switched to eth1
eth1 "dies" on heavy load too

I tried:
/etc/init.d/networking restart
rmmod and modprobe forcedeth
I can't make the newer forcedeth compile

This guy seems to have the same problem, but I don't find any solution:
https://bugs.launchpad.net/ubuntu/+s...15/+bug/131737

I tried to follow this guide:
https://bugs.launchpad.net/ubuntu/+s....15/+bug/76346

My /var/log/messages (look for the "NETDEV WATCHDOG: eth0: transmit timed out")
http://indigo.ed.ntnu.no/jalla/files...t/messages.txt

I am kind of n00b and I need an übergeek to take a look at this... Please help!

That is well documented bug with the forcedeth driver: the NIC dies under heavy load.

I have a RTL8211 and am using the 8139 or whatever driver, the one in the kernel, and I haven't had the problem at all. I guess it occurs on one machine but not necessarily another.

EDIT- what NIC device do you have?

I'm sorry I haven't got any specific links for you, however I recommend using Google to see if there's a newer version, or a patch, for the forcedeth driver, which addresses the issue. If you can't find anything, I'll have a look too for you; I'm currently making supper, but after that, I'll definitely have a look for you if you haven't got anywhere.

nvidia chipset
 

Hi again! I have tried google and I can not find any solution to my problem. I am running forcedeth 0.59 and I think the newest is 0.60, is this small version difference that important? Yes I am very desperate for your help. If you want to I can paste my error messages from gcc when trying to compile the 0.60 version of forcedeth.

I am not sure what NIC I have and how to find out. MSI's webpage says "Dual LAN 10/100/1000 Fast Ethernet by Vitesse VSC8601"

Output from lshw:
*-bridge:0
description: Ethernet interface
product: MCP55 Ethernet
vendor: nVidia Corporation
physical id: 8
bus info: pci@00:08.0
logical name: eth0
version: a2
serial: 00:16:17:bb:25:71
width: 32 bits
clock: 66MHz
capabilities: bridge bus_master cap_list ethernet physical
configuration: broadcast=yes driver=forcedeth driverversion=0.59 latency=0 maxlatency=20 mingnt=1 multicast=yes
resources: iomemory:fbef6000-fbef6fff ioport:b400-b407 iomemory:fbefa800-fbefa8ff iomemory:fbefa400-fbefa40f irq:18
*-bridge:1
description: Ethernet interface
product: MCP55 Ethernet
vendor: nVidia Corporation
physical id: 9
bus info: pci@00:09.0
logical name: eth1
version: a2
serial: 00:16:17:b6:fa:62
width: 32 bits
clock: 66MHz
capabilities: bridge bus_master cap_list ethernet physical
configuration: broadcast=yes driver=forcedeth driverversion=0.59 ip=192.168.10.35 latency=0 maxlatency=20 mingnt=1 multicast=yes
resources: iomemory:fbef5000-fbef5fff ioport:b080-b087 iomemory:fbefa000-fbefa0ff iomemory:fbef4c00-fbef4c0f irq:19

Well, a tiny version change *can* be important, depending on what changed. Generally, a changelog will be included or viewable online, which describes any changes made to code when a new version of something is released.

I have never heard of Vitesse VSC8601. My nVidia board (an MSI also) uses the Realtek GBLan device.
Let me go and have a browse around and see if I can relocate some info on that bug I mentioned, and see if there's a patch or repair.

NOTE: Since the driver is one of the linux kernel's drivers, you wouldn't usually compile it all by itself. It would need to be placed into the source directory, added to the kernel config, and then you would recompile the kernel.

Anyway, before I get ahead of myself, lemme go look around. I'll get back to you.

OK, and, what kernel are you using? Since as I mentioned, this IS a kernel driver, the simplest solution will probably be to download a newer kernel and recompile it. That way, you won't have to go through any weird hoops, and the bug fix will already be incorporated.

If you are using a 2.6.xx kernel, why not upgrade, especially if yours is older than 2.6.22 because as I recall, 2.6.22.2 is whereabouts this issue was addressed.

http://www.mail-archive.com/linux-ke...msg199918.html

if you look at this link, you'll see that this is a patch-file, which fixes the forcedeth driver code for the Vitesse PHY driver bug. You could save the patch file and patch it manually, OR simply upgrade to a shiny new kernel :)

Hi again! I was running 2.6.20-15 and since I was not too keen on compiling, I upgraded to gutsy. Now I am running 2.6.22-11 :)

I have not tested with some heavy network load but if it crashes again I will be back whining more :P But I really think it did the trick! I hereby declare my eternal gratefulness! Thank you for your help!

No worries Jalla, glad to have helped, I hope. Keep us posted if that does the trick ;)

PS - Dangit! Are they up to -11 already!? Or is that just a Gutsy kernel? I think I just upgraded to 2.6.22.5 last week :O