EPIC Fail - eth0 dissappeared: via-rhine: probe of blablabla failed with error -5

Mr_Nuke · 12-07-2010, 03:58 PM

I apologize in advance for the brief explanation. I am too surprised to be able to react.
Running Fedora 14 x86_64.
I did a BIOS upgrade, rebooted, the next thing I know, eth0 was gone. For good:

Code:

[root@nukeserv2 mrnuke]# ifconfig eth0 up
eth0: unknown interface: No such device

However, when I run lspci, it's there:

Code:

[root@nukeserv2 mrnuke]# lspci |grep Ether
00:0b.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL-8169 Gigabit Ethernet (rev 10)
00:0c.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL-8169 Gigabit Ethernet (rev 10)
00:12.0 Ethernet controller: VIA Technologies, Inc. VT6102 [Rhine-II] (rev 78)

looking in the kernel log yields

Code:

[root@nukeserv2 mrnuke]# dmesg |grep rhine
[   11.891169] via-rhine.c:v1.10-LK1.4.3 2007-03-06 Written by Donald Becker
[   12.111662] via-rhine 0000:00:12.0: PCI INT A -> GSI 23 (level, low) -> IRQ 23
[   12.111840] via-rhine: probe of 0000:00:12.0 failed with error -5

checking /var/log/messages, I see the "error -5" first appeared around the time I started having the problem. I tried going back to the initial BIOS with no avail (the issue persists)

Removing /etc/udev/rules.d/70***net***, udev recreated it with rules for eth1 and eth2 (two Realtek NICs), but not eth0.

Does anyone have any idea what on earth may be going on?

foodown · 12-08-2010, 02:21 AM

Negative error numbers, as far as I know, generally indicate a hardware failure relating to memory. It sounds like your card may be fried.

Try re-seating the card in its slot.

If that doesn't change anything, try booting a live CD or installation CD and see what the dmesg looks like there. At least if you get the same probe error, you've ruled out a software problem.

As a last result, you could always see if there's an EEPROM flash of some kind available for your NIC and try that.

I have to say that, if the card is registering on the PCI bus, but failing to initialize with an error like that, the outlook is not good. I wish I had better news.

Mr_Nuke · 12-08-2010, 12:14 PM

The card is on-board, so I can't re-seat it. About the liveusb, I tried both Fedora14 and CentOS 5.5 liveusb's and they all return the same error.

I don't see how the card may be fried, as the light blinks when I plug in a cable, just as it should.

There is something I overlooked in /var/log/messages:

Code:

Dec  8 18:21:35 nukeserv2 kernel: [   12.302642] via-rhine 0000:00:12.0: PCI INT A -> GSI 23 (level, low) -> IRQ 23
Dec  8 18:21:35 nukeserv2 kernel: [   12.302714] Invalid MAC address
Dec  8 18:21:35 nukeserv2 kernel: [   12.302756] via-rhine: probe of 0000:00:12.0 failed with error -5

Could that "Invalid MAC" be fixable in any way?

Skaperen · 12-08-2010, 12:44 PM

Quote:

Originally Posted by Mr_Nuke

I don't see how the card may be fried, as the light blinks when I plug in a cable, just as it should.

Its interface to the bus and to the south bridge is what may be fried. The rest of the chip could be doing everything in can to tell the CPU about traffic it gets, and the CPU can't hear it.

Quote:

Originally Posted by Mr_Nuke

There is something I overlooked in /var/log/messages:

Code:

Dec  8 18:21:35 nukeserv2 kernel: [   12.302642] via-rhine 0000:00:12.0: PCI INT A -> GSI 23 (level, low) -> IRQ 23
Dec  8 18:21:35 nukeserv2 kernel: [   12.302714] Invalid MAC address
Dec  8 18:21:35 nukeserv2 kernel: [   12.302756] via-rhine: probe of 0000:00:12.0 failed with error -5

Could that "Invalid MAC" be fixable in any way?

It could be the cause of the problem, or related to it.

Just a thought, have you tried to do a configuration reset, or restore to factory default, in BIOS?

TobiSGD · 12-08-2010, 12:57 PM

Did you load the BIOS defaults after the upgrade? I have seen many weird errors if that is not done after an BIOS-upgrade.

Mr_Nuke · 12-08-2010, 01:23 PM

Yes, I did load the defaults afterward. I'll try to use the reset jumper and see it that works.

EDIT: I just reset the BIOS using the jumper method, booted with the defaults, and the error is still there.

@Skaperen: You said that the interface between the NIC and the SB may be fried. Wouldn't that mean that the card should be undetectable?

EDIT2: I've looked at via-rhine.c. It does stop initialisation once an invalid MAC is found. I'll try to comment that out and rebuild the driver to see if it the NIC works this way.

Mr_Nuke · 12-09-2010, 01:17 PM

If I remove the MAC chech, rebuild and reload the module, the adapter initializes, but other parts of the kernel prevent it from being used because of the invalid MAC; however, with

Code:

# ifconfig eth0 hw ether <valid MAC>

I can then use the NIC normally. I have contacted the driver maintainer to see if it might be worth "spoofing" the MAC if an invalid one is found. For now, I've put the modified module and set up a script that automatically gives the NIC a valid MAC, and it works.

gkasica · 01-24-2017, 04:11 PM

I'm not sure what file you have there in your distro but Ubuntu has this:
vi /etc/udev/rules.d/70-persistent-net.rules

What you end up doing is removing the added lines and rebooting and the correct items get automatically re-added.
Here's a URL on it:
https://www.serenux.com/2009/11/howt...ox-to-another/
Similar it may help you.

George

sundialsvcs · 01-24-2017, 06:55 PM

Please be careful not to resurrect seven-year-old threads ...