Industry


Ads by TechWords

See your link here


When Linux goes bad: the e1000e Ethernet bug

I know some people think I only say wonderful things about Linux and bad things about Windows. Wrong. I call them like I see them. Most of the time, while wearing my operating system umpire's cap, I see Linux making the right hits... Most of the time. Sometimes, though, Linux strikes out and it surely has with the recent e1000e Ethernet bug.

This bug is found only in the pre-release version of the 2.6.27 Linux kernel. If you're using a conventional Linux, you don't have to worry about it.

But, if you're living on the bleeding edge of technology and want a run a beta Linux that's using the 2.6.27 kernel, do not, do not run until you're sure your machine doesn't use the Intel ICH8 and ICH9 chipsets, with the 82566 and 82567 Ethernet chipsets. If you do, you're very likely to fry your Ethernet's firmware and you can say bye-bye to your network connection.

Don't have the foggiest idea what your Ethernet chipset is? In that case, open up a shell window and run the following command:

lspci | grep 8256[67]

If you can get any result, your system is vulnerable. Do not, unless you don't mind toasting your Ethernet, run any version of Linux with the currently existing version of the pre-release 2.6.27 kernel. If all you get is the return of your shell prompt, then you're good to go.

Now, it's possible that if you're using another chipset with the e1000e driver that they might get blasted. I'm not familiar with any reports of this happening, but then no one is exactly sure how this problem is occurring in the first place. If you want to be completely safe, check to see if you're using the e1000e driver by once more opening a shell interface and run this command:

/sbin/lsmod | grep e1000e

If you get no result, you're fine. Your system is not using this driver. If you do get a result, well, I, for one, wouldn't chance running my PC until we're all clear on what's happening here. Or, better still, the problem is fixed and I don't need to worry about it anymore.

That said, it's really only people like me who are always working with the newest and not quite ready for the big leagues Linuxes that need to be concerned. There are no shipping Linuxes with this bug; there are only alpha and beta builds.

To be exact, the major test builds to be worried about are the Mandriva Linux 2009 pre-releases; Novell's SUSE Linux Enterprise 11 Beta 1 and openSUSE 11 Beta 1; Fedora 10 release candidates; Gentoo Linux; and Ubuntu Intrepid Ibex.

Some people are claiming that you can fix the problem by using Intel's low-level IABUTIL.EXE. Fail!

What will happen is that you'll make an already losing situation even worse. Chris Jones, a Canonical system administrator, was one of the first to discover, the hard way, that this logical move actually only leads to the Ethernet hardware vanishing entirely from the system as far as his PC's bus and BIOS was concerned. In other words, you really don't want to call there.

The moral of the story is that there are no perfect operating systems and just because Linux usually works great, doesn't mean that it always leaves the operating system ballpark with a win.

What People Are Saying

I know people are mocking

I know people are mocking you for pointing out that this is a bug in a beta version but its a fairly major problem - one that could toast your ethernet connection so thanks. I am going to pass this on to a few people who might be at risk.

When post are just to post

This makes for a good public notice to those that haven't yet heard about the problem. As for an indication of how good an OS is, lame. Would it be Window's fault if Intel posted a bad driver? Was it Vista's fault that Nvidia's driver sucked? How is it Linux's fault for Intel's driver?

Plus this is a beta. No reason not to have an rc kernel in their if its going to be released before the final release of the distribution.

OMG OMG OMG

A bug in a beta release??? NOOOO! Say it isn't so!!! My whole world is crashing around me.

Stephen, what on earth were you smoking when you wrote this article? Is this a joke?

But, it's a Nasty bug

I mean, I can only think of one other bug off the top of my head that could zap PC hardware and that one was ages ago. In addition, today's bug shows up in a lot of popular beta Linux distributions.

So, I thought it worth pointing out.

Steven

I see your point

If I'm beta testing, I figure I'm risking my OS install, NOT my hardware.

2.9.27-rc7

As an embedded developer I have observed that this issue was seemingly created near rc6 and appears to be fixed at rc8.

Steven these are not released kernels. It is irresponsible to even compare release candidate kernels to anything after a year to release OS candidate from any other place.

You should place emphasis on:

If you do not know what your are doing you should not be running a rc7 kernel.

The fault in the system is the distro purveyors putting a rc7 kernel in a Beta 1 distro in their race to the bottom.

I am running 2.6.27-rc8 because I am running /arch/arm/mx2 ie. Freescale i.mx27 and this is the first off support for this arch.

Regards John

Not so fast!

"If you do not know what your are doing you should not be running a rc7 kernel."

I wouldn't run that line by any of the developers that got hosed - this issue didn't hit many of these folks out of stupidity.

Also, there is a need for folks to test out the different release candidates. (http://www.linux.com/feature/118454) However, those testers should know they are at some kind of risk and usually are encouraged to test kernels on equipment and data they can spare.

Linux says "Mmm tasty

Linux says "Mmm tasty chipset om nom nom nom" My apologies to those actually effected by the bug.

No need to be so negative with the moderation

C'mon, it's supposed to be funny. Laugh. :)

drivers are intel's

As I understand, the intel ethernet drivers are submitted by intel themselves, so this bug is, al least in part, responsibility of intel.

I remember some time ago, Mandriva had a similar issue with LG optical drives; some specific models died upon loading the install image, or booting into the system for the first time, but that was due to bad firmware in the drives.

Because of people who do this alpha and beta testing, some bugs like these can be found and corrected before they affects a lot of people.