eth0 not present anymore after card upgrade. Must use Windows first

narc · 04-22-2007, 09:38 AM

Hello. I'm using a "Linux from scratch" system which usually serves me well on a 3.2 gHZ system and with kernel 2.6.11.12. I had no problem whatsoever with my network until I upgraded my card from an ordinary SVGA to an nVidia AGP card. I used the latest drivers from nVidia and everything went fine, video-wise. But since then my PCI network card is not recognized at boot time anymore. My box dual-boots with Windows XP and if I run Windows first, I could reboot again and then the card would be recognized.

lsmod returns nothing since everything is compiled in the kernel.

dmesg returned some interesting messages which I present here, diffed between two boots, once with a failure, the other with a successful network boot.

Code:

45c45
< Detected 2400.756 MHz processor.
---
> Detected 2400.729 MHz processor.
93d92
< PCI: device 0000:00:08.0 has unknown header type 08, ignoring.
119c118
< audit(1177234387.489:0): initialized
---
> audit(1177235523.748:0): initialized
124c123
< ACPI: Thermal Zone [THRM] (45 C)
---
> ACPI: Thermal Zone [THRM] (64 C)
144a144,147
> ne2k-pci.c:v1.03 9/22/2003 D. Becker/P. Gortmaker
>   http://www.scyld.com/network/ne2k-pci.html
> ACPI: PCI interrupt 0000:00:08.0[A] -> GSI 19 (level, low) -> IRQ 19
> eth0: RealTek RTL-8029 found at 0xe800, IRQ 19, 00:00:B4:9C:F6:D5.
198c201
< intel8x0_measure_ac97_clock: measured 49970 usecs
---
> intel8x0_measure_ac97_clock: measured 49986 usecs
238,240c241,243
< /dev/vmmon[2973]: Module vmmon: registered with major=10 minor=165
< /dev/vmmon[2973]: Module vmmon: initialized
< /dev/vmnet: open called by PID 3035 (vmnet-bridge)
---
> /dev/vmmon[3026]: Module vmmon: registered with major=10 minor=165
> /dev/vmmon[3026]: Module vmmon: initialized
> /dev/vmnet: open called by PID 3088 (vmnet-bridge)
243c246,248
< bridge-eth0: peer interface eth0 not found, will wait for it to come up
---
> bridge-eth0: enabling the bridge
> bridge-eth0: up
> bridge-eth0: already up

As you can see, PCI device 0000:00:08.0 is rejected by the kernel for having a bad header. After running Windows and rebooting to Linux, device is being recognized by the kernel and sets it correctly to eth0.

I don't know much about interrupts. Could be possible that the network card is cconlflicting with the nVidia card - even though the nVidia card is in a AGP slot ? Any kernel settings that I might look into to force a correct reading the first time ?

Any hint would be appreciated.

-----

b0uncer · 04-22-2007, 10:29 AM

I can't be of much help I'm afraid, but here's one trick to try. A friend of mine had some time ago a Fedora box which had suffered a power loss during nighttime (presumably lightning strike, it had thundered) and in the morning the ethernet card didn't function. Basically the same story as yours, except that I couldn't test if it worked after booting Windows for there was no such OS on board (pure Linux). If you can't figure out anything else, remove the ethernet card from the machine and let it be off for some time (I had three quarters of an hour, while I was eating

heh); it might also do good to boot the machine without the card, shutdown and have power off (cords off too, and press power button to get rid of the remaining electricity..ground it too, by grabbing to the metal box while touching ground at the same time, to get rid of static electricity) also. I have no guarantee that after you plug the card in after this and reboot it works, but it did for me

I have not figured out any sane reason why, but I think the lightning had somehow caused problems in the machine that weren't fatal but caused malfunction with the ehternet card, and it helped to drain any electricity off the machine (as long as the cord is on the wall, even when the machine's power switches are set to "off", there can be electric potential in the components; pressing power button after unplugging power cord should get rid of that). Tried powering off, removing configs, plugging card out and in (while cords on the wall

) but apparently without any use; I know remaining electricity can cause nasty stuff with computers, but this was new to me.

MQMan · 04-23-2007, 12:54 PM

Maybe reset the PnP information, in the BIOS.

Cheers.