Old Hardware - PXE-E05

  1. 2 years ago
    Edited 2 years ago by Pwu

    Hello,

    I'm currently working on old hardware. I have approximately 2000 devices that I have to reinstall.

    My POC with CloneDeploy is really successful (the installation take less that 5 minutes instead 45 with the previous system), but I still have to solve a blocker problem : After booting on the kernels proposed by clonedeploy, the network card seems to be stuck, I got that error from the BIOS : PXE-E05.

    The only way I have is unplug/replug the power cable. Has you understand, I don't want to pass over 2000 devices to unplug/replug :D

    My questions are :

    Thanks for reply and thanks for clonedeploy :D
    Kind regards,

    Answered by Pwu

    See post in context

    Hello,

    I'm glad to announce that is working now :D

    I based my solution on the 3.18.35 linux kernel finally to be fully compatible with clonedeploy.

    About the driver error, here are the lines that I had to comment to make the network working at the reboot :

    linux-3.18.35/drivers/net/ethernet/intel/e1000e/netdev.c

    if ((adapter->flags & FLAG_IS_ICH) &&
    (adapter->flags & FLAG_READ_ONLY_NVM))
    e1000e_write_protect_nvm_ich8lan(&adapter->hw);

    /* if ((adapter->flags & FLAG_IS_ICH) && */
    /* (adapter->flags & FLAG_READ_ONLY_NVM)) */
    /* e1000e_write_protect_nvm_ich8lan(&adapter->hw); */

    Thanks for support :D

  2. clonedeploy

    14 Jun 2016 Administrator

    If I understand you correctly, you can pxe boot the first time and image the machine, but when it reboots and tries to boot from the nic again it gets the pxe-e05 error?

  3. Hello,

    Yes. it's the case.

    In my investigation, I tried multiple kernels in clonedeploy (got the kernel from the .deb file from debian and ubuntu communities)

    In the case of the debian kernels, i got a "kernel panic not syncing vfs unable to mount root fs on unknown-block(0 0)"

    In the case of the ubuntu kernels, with the 3.2 kernels binary, I can boot the clonedeploy solution but the diagnostic system says "Network card maybe not compatible" and the keyboard is not working anymore.
    Above 3.4 ubuntu kernels, I got back the pxe-e05 error.

    To reproduce :
    get the deb here : http://packages.ubuntu.com/fr/precise/linux-image-3.2.0-23-generic
    dpkg-deb -x the_deb_file.deb /tmp/extract
    In the extract directory, get the boot/vmlinuz file for clonedeploy

    So I will try 2 ways to solve :

    • Try to understand the VFS trouble and create custom kernel.
    • Try to make the ubuntu kernel working with my network interface (and the keyboard :D)

    Regards,

  4. clonedeploy

    14 Jun 2016 Administrator

    I would try some of these older kernels from my last project
    https://sourceforge.net/projects/cruciblewds/files/Kernels/

    Perhaps 3.3.4, there are two kernels in that folder one called speed and the other compatibility

  5. Edited 2 years ago by Pwu

    I tried the compatibility one, diagnostic passed with some errors, but everything seems to be fine, but I got back the pxe-e05 error :'-s

  6. clonedeploy

    14 Jun 2016 Administrator

    Are you certain it is the kernel doing this and not the pxe boot loader? Have you tried changing the pxe mode to ipxe?

  7. Hello,

    I was sure, but I tried, the result is the same. I can concluded that is really come from the kernel.

    Also, that is very instructif :
    https://thesorcerer.wordpress.com/2011/07/01/guide-intel-82573l-gigabit-ethernet-with-ubuntu-11-04-and-fix-pxe-e05/
    and then
    http://ubuntuforums.org/archive/index.php/t-1276211.html

    I can resume, the problem is related to the driver e1000 for the network cart, which is making a NVM check. That check make trouble on the eeprom. The second link explain how to rebuild kernel with the NVM check disable. So let's try that :D

    Regards,

  8. clonedeploy

    14 Jun 2016 Administrator

    I may be a little before I can make this change. If you want to try yourself, here is documentation for compiling a kernel from my last project, should be mostly relevant for CloneDeploy.

    http://docs.cruciblewds.org/2.3.3/r1/kernel.html

  9. clonedeploy

    14 Jun 2016 Administrator

    Another thing you could try is the vmware kernel I made a while back.
    https://sourceforge.net/projects/cruciblewds/files/Kernels/VMware/

    I made that specifically for issues I was having with vmware with the e1000e nic. Not sure if it is related.

  10. Hello,

    I tested 3.12.39-WDS but not working (no network in clondeploy "maybe not compatible + pxe-0e5)
    I'm currently building a kernel with a modification in the netdev.c from e1000e driver. (link explain on a link in my previous post)

    If it doesn't work, I'll try a custom kernel like you explain.

    Thanks for support!

  11. Hello,

    I'm back on that. I tested other tools, but clonedeploy is the most close to my project.

    So I tried to build my own kernel, based on 4.6.2. I have 2 questions

    With the new kernel, the name of the default network interface is enp0s3 instead eth0. The result is that I have no network connection. Is it possible to change that somewhere ?

    The second one is that you mention you had troubles with vmware and the e1000e, can you explain a bit more, because it's exactly the same driver.

    Regards,

  12. Pwu

    21 Jun 2016 Answer

    Hello,

    I'm glad to announce that is working now :D

    I based my solution on the 3.18.35 linux kernel finally to be fully compatible with clonedeploy.

    About the driver error, here are the lines that I had to comment to make the network working at the reboot :

    linux-3.18.35/drivers/net/ethernet/intel/e1000e/netdev.c

    if ((adapter->flags & FLAG_IS_ICH) &&
    (adapter->flags & FLAG_READ_ONLY_NVM))
    e1000e_write_protect_nvm_ich8lan(&adapter->hw);

    /* if ((adapter->flags & FLAG_IS_ICH) && */
    /* (adapter->flags & FLAG_READ_ONLY_NVM)) */
    /* e1000e_write_protect_nvm_ich8lan(&adapter->hw); */

    Thanks for support :D

  13. clonedeploy

    21 Jun 2016 Administrator

    Thanks for the info, I'll be sure to add that to my process when releasing future kernels.

 

or Sign Up to reply!