I’ve found a very strange behaviour with network adaptor detection on both openSUSE 11.4 and 12.1 while I updated my old 11.1 on a Windows 7 multiboot computer. I’m using Linux quite occasionally on that machine. It is a Intel Core i7 860 quad core 2.8 GHz on a Gigabyte P55A-UD4 F15 mainboard with Realtek RTL 8111D Gigabit Ethernet Controller integrated. I’ve spent the last three weeks on testing and installing the 11.4 and 12.1 distros to get networking running.
The following happend always during the installation: While I’ve chosen manual adjustments for installation YaST always failed the network connection test. The Realtek controller was detected correctly and it didn’t matter if I used ifup or NetworkManager for network administration. As a tradition for this computer I have a static IP configuration. I tried lots of ideas posted in this forum to get the network running. I was even driven into despair because normally I’m quite experienced to get a network running. I realized that during the initialization phase of the RTL 8111D the green LED just blinked once to acknowledge a 1 GBit/s link, then failed and fell back to 10 MBit/s. Linux was not able to handle that and all external IP’s where not pingable and the route command took the network timeout value to return to the prompt with correct results.
So it took me 12 days to get me to the right place where the real error showed up: When I had to use Windows again I realized that there also was only 10 MBit/s network speed available. My conclusion was that the onboard network chip was damaged. I deactivated the chip within BIOS and bought a dedicated network adapter with almost the same chip on it, a Realtek 8111E. Now everything worked. On the same evening I spoke with a friend who solved a similar strange behaviour on his Mac by unplugging the power supply for a reasonable amount of time. I thought it might be a good idea to test this with my computer too … I unplugged the PSU for 2 hours, activated the onboard LAN again … and I was left with two working Realtek controllers. So placed everything how it was with the onboard RTL 8111D only and now both Linux versions had a working network. No matter if I use ifup or NetworkManager - both are working.
But: it broke down again when I let YaST do the lots of updates which were quite many for 11.4 and reasonable some on 12.1 - after reboot the RTL 8111 chip was down again with 10 MBit/s. I repeated the procedure unplugging the PSU (for 20 minutes only) which recovered the functionality of the chip.
And quite interesting: on a spare disk I’ve tested another Linux distro Ubuntu desktop 11.10 which made the same thing. Installation ok, online updating ok while the machine did not reboot. After first reboot: no more network. Unplugging PSU for a while and restart did it.
So my questions here are:
What can cause a strange and weird disfunctionality on a mainboard which can’t be recovered by resetting the BIOS’es factory settings (which I also tried in the beginning)?
As it is difficult to reproduce because it only happened in the driver detection and installation part of the distro - where can I find logs or anything else to post them? I think it might be important for the developers to make a fix and I also think that a reasonable amount of network problems posted here could have similar basics.
Well I think I can reproduce it once as I want to upgrade my “productive” 11.4 system on a SSD to 12.1 in the next few days. So if I knew the right place(s) where to look I could retrieve some logs to post.
It seems to me that you have demonstrated that this is a hardware failure.
I have seen similar things, though not with a network card. Rebooting doesn’t fix it, but power down and leave the power off for a brief period does. My guess, at the time, was that this was some sort of runaway feedback effect in a transistor or similar circuit component.
I suggest that you junk the defective card and save yourself the frustration.
Ok, thanks - I’ll consider that. I’ve also heard from a friend who works for IT support that onboard LAN adapters quite often get damaged (for whatever reason).
So, when I deactivate the onboard card and use my PCI Express replacement the new card will be seen as a different PCI device. How can I prevent or change that the system/kernel will see the new one as eth1 instead of eth0? It’s just for cosmetical reasons that I like to have eth0 as long as I only have one active wired LAN adapter. I know how to do that in Windows 7 but likely not on Linux
On 12/27/2011 10:46 AM, konran wrote:
>
> Ok, thanks - I’ll consider that. I’ve also heard from a friend who works
> for IT support that onboard LAN adapters quite often get damaged (for
> whatever reason).
>
> So, when I deactivate the onboard card and use my PCI Express
> replacement the new card will be seen as a different PCI device. How can
> I prevent or change that the system/kernel will see the new one as eth1
> instead of eth0? It’s just for cosmetical reasons that I like to have
> eth0 as long as I only have one active wired LAN adapter. I know how to
> do that in Windows 7 but likely not on Linux
The file that controls that is /etc/udev/rules.d/70-persitent-net.rules. Delete
all lines that do a rename to “ethX” and restart.
Thanks Carlos and lwfinger - great answers. Very useful for me … it seems that I have to spend more time on it. Anyway I’ll see what result a Bugzilla will do…
Not for sure, but just a thought.
What if the chip was ‘smart’ enough to know that the network transmission speed (MTU ?) is never above 10mb/s and ‘down grades’ itself.
There could be a ‘power saving’ thing or something to that effect. If you had another system on the same concentrator (switch, router, whatever) that you could test transfer between them at 100+mb/s???
Once you leave your local net, NOTHING is over 10mb/s, not T-1’s, cable ISP etc… Nothing.
Landis, I’m not quite sure if I get that right how to test it…?
The PC’s LAN adapter is connected to a 8-port GBit/s switch (7Links TP-Link, old fashion with green/yellow dual-LED’s to show 1000 or 100/10 MBit/s link). One port of this switch is connected to a AVM Fritz!Box Fon WLAN 7170 which has max. 100MBit/s and configured as DSL router to pass through up to 16.000 kBit/s from WAN IP.
When the problem I described has happened it is permanently frozen in the RTL 8111D chip. I can reboot the PC and start Windows 7 and have also a 10MBit/s speed fixed. Windows then takes more time to get ready after start and activating the network … anyway it keeps the slow speed working. At the same time I can transfer at full 1GBit/s speed between a PowerMac G5 and a MacBook Pro connected to the same switch. I’ve changed the switch ports for the PC also and the cable and the slow link moves to the newly connected switch port.
On Windows I tried once to increase MTU to 7200 for jumbo packets but this didn’t make any difference.
So, is there anything left that I can test/check for your idea?
