Announcement

Collapse
No announcement yet.

2.6.27 Kernel Killing Network Hardware

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • 2.6.27 Kernel Killing Network Hardware

    Phoronix: 2.6.27 Kernel Killing Network Hardware

    In case you missed it, there's a rather serious regression with the e1000e network driver in the Linux 2.6.27 release candidate kernels. This Ethernet driver has been killing some Intel integrated Gigabit network adapters by corrupting the chip's EEPROM...

    http://www.phoronix.com/vr.php?view=Njc0Nw

  • #2
    Just a little correction for Ubuntu, there is not the mainline driver used, but

    filename: /lib/modules/2.6.27-4-generic/kernel/ubuntu/e1000e/e1000e.ko
    author: Intel Corporation, <linux.nics@intel.com>
    description: Intel(R) PRO/1000 Network Driver
    license: GPL
    version: 0.4.1.7

    But it seems to suffer from the same problem. That problem is nothing new, basically every distro with 2.6.27 kernel is affected and the first corruption I knew of was about 1 month ago already. Hopefully there will be a tool that can flash the eeprom back with working values...

    Comment


    • #3
      It's the RC, not mainline

      I found it confusing that the article states that it's the mainline kernel which is affected - in fact, it's the RC, which is not a production version. Moreover, it does not impact all major Linux distributions, but rather their current beta versions.

      I think that should be made clear in order to minimize FUD emission which this bug caused too much of already anyway.

      Comment


      • #4
        Originally posted by shamooki View Post
        I think that should be made clear in order to minimize FUD emission which this bug caused too much of already anyway.
        Apologizes if it wasn't clear enough, it's just been revised.
        Michael Larabel
        http://www.michaellarabel.com/

        Comment


        • #5
          Originally posted by shamooki View Post
          I found it confusing that the article states that it's the mainline kernel which is affected - in fact, it's the RC, which is not a production version. Moreover, it does not impact all major Linux distributions, but rather their current beta versions.

          I think that should be made clear in order to minimize FUD emission which this bug caused too much of already anyway.
          Good point

          Comment


          • #6
            Wording

            "Intel netcard suiciding on concurrent read/write"

            As little research as I did.
            No?

            Comment


            • #7
              Ok wait, so is any physical damage caused to the hardware, or is it just made not-functional until certain bit states are fixed?

              Comment


              • #8
                As I understand there is no physical damage, to return to working state you only must reprogram eeprom back to its default state.

                Comment


                • #9
                  Originally posted by blueskynis View Post
                  As I understand there is no physical damage, to return to working state you only must reprogram eeprom back to its default state.
                  From what I've heard this is not 100% the case. Some who have tried say that results in a MAC address of 00-00-00-00-00.

                  Comment


                  • #10
                    "you only must reprogram eeprom ": for 99,9% of users that sentence means = dead card.

                    but it's also true that a user that's running 2.6.27rc must be at least a user that understands a minimum of "computer things" so he's able somehow to reflash it.

                    If something like this happens for example with an Ubuntu stable release, oh well, I couldn't blame if Microsoft pays some newspaper to FUD about it..lol

                    Comment


                    • #11
                      How do I test to see what card I'm using so I don't encounter the problem when I update.

                      Also, isn't Canonical's work around just getting rid of that driver?

                      Also, would it be easier to just import the driver from a previous Kernel?

                      Possible problems I can see:
                      1. It not being compatible with the 2.6.27 kernel
                      2. It's not the driver that's corrupting (which that is being talked about in the bug report)

                      Thanks a lot.


                      Bas.

                      Comment


                      • #12
                        but it's also true that a user that's running 2.6.27rc must be at least a user that understands a minimum of "computer things" so he's able somehow to reflash it.
                        anybody running -rc* kernels must be ready for anything, be it hardware failure, your pc killing your dog, or kernel devs kidnapping your girlfriend :]

                        Also, would it be easier to just import the driver from a previous Kernel?
                        i think this one will nail down the problem - http://lkml.org/lkml/2008/9/23/431 .

                        the issue might lay deep within the kernel, and not the driver itself.

                        remember how rtorrent exposed a deep bug in the kernel in 2.6.18 ?

                        Comment


                        • #13
                          Originally posted by bulletxt View Post
                          If something like this happens for example with an Ubuntu stable release, oh well, I couldn't blame if Microsoft pays some newspaper to FUD about it..lol
                          What most likely happened is Microsoft paid a Linux kernel saboteur to have the kernel trash your LAN card.

                          See: The Linux Kernel SABOTEURS at
                          http://www.phoronix.com/forums/showthread.php?t=9509
                          Last edited by Jade; 09-26-2008, 07:04 PM.

                          Comment


                          • #14
                            oh....I was just joking but it seems the thing is more serious than I thought... I'll read your link!

                            Comment


                            • #15
                              Originally posted by bulletxt View Post
                              oh....I was just joking but it seems the thing is more serious than I thought... I'll read your link!
                              Being friends with one of the Intel linux devs, he told me that this issue is most likely related to core kernel changes rather than a bug with the actual driver. The nvram on the e1000* based cards are shared between the rest of the hardware components on the machine, as its a part of the address space. Anything can have a flaw that writes to the wrong address, such blame can be bugs in the kernel, a buggy X11 graphics driver writing to a bad memory address, or a combination of both could be the culprit here. From what I understand this is mainly happening with laptops, and no reported happenings with PCIe based server adapters yet.

                              Comment

                              Working...
                              X