Tuesday, August 27, 2019

RHEL 8.0 and support for removed adapters

In RHEL 8.0, support for a good number of hardware devices has been removed. A list of removed adapters with their device IDs can be found in this RHEL documentation. We provide support for some of those that are still fairly commonly used today. You can check your devices's IDs (as shown by lspci -nn) against our list of supported devices.

(1) Installation of the OS requires a driver for your hardware. We offer driver update disks (DUD). You can download them from here or the corresponding directory of our mirror sites. Each DUD image contains a driver in the form of a kmod package. The installer is supposed to find the driver. If this does not happen, you need to append the inst.dd option to the boot command line. For details please see Performing an assisted driver update. Also, a tutorial video that demonstrates how to use ELRepo's DUD is available.

(2) The installation process installs the kmod package for your adapter. Normally, because of the kABI-tracking nature of kmod, there is no need to reinstall the driver upon each new kernel update. However, it was found that the current version of dracut in RHEL 8.0 has a bug and the initramfs image of a new kernel does not contain the kernel module from the installed kmod package. As a result, the new kernel fails to boot. (This is no longer an issue. See the [Update] below.)

As an interim solution for the problem in (2), we provide a dracut package with a patch that fixes the bug here:

 http://elrepo.org/people/akemi/testing/el8/dracut/

Install this version of dracut and then update the kernel. The system now should boot normally.

[UPDATE]  dracut-049-10.git20190115.el8_0.1 released on Oct 29, 2019 has the patch that fixes the issue.

[UPDATE2]  When updating the OS from 8.0 to 8.1, ELRepo's kmod packages must be updated to their .el8_1 version for the system to boot the 8.1 kernel.

[Note for CentOS-8 users]  If you have switched to the centosplus kernel (kernel-plus) while on 8.0, you may want to uninstall the kmod package and no further action is required when updating to 8.1.

[Note for users who wish to install remotely] There is an excellent blog post written by Arrfab.

25 comments:

  1. Works great on my Dell T610 / PERC H310. Thanks!

    ReplyDelete
  2. Thanks for letting us know. Great to hear it worked.

    ReplyDelete
  3. Good evening,

    I'm trying to follow this to fix my fresh Linux install onto a Dell T320 / PERC H310. I downloaded the iso for megaraid, burned it, mounted the CD-ROM and ran rpm -ivh kmod-megaraid_sas***.rpm

    I got a warning for no key but it installed. Do I need to reboot at this point? How do I verify that it is working?

    Centos 8.0.1905

    ReplyDelete
    Replies
    1. I suspect that the "no key" message was about the secureboot key. You can ignore it.

      Delete
    2. Thanks for the update Tora. I'm still not sure what to do after I've installed it. Do I need to run something to see the hdd's attached to my raid controller? Sorry for the confusion on my part, first time having to deal with a raid controller on Linux.

      Delete
    3. It would be best if you would please make use of our mailing list [1] for support queries. More people will see your question and assistance may be obtained somewhat more promptly.

      [1] http://lists.elrepo.org/mailman/listinfo/elrepo

      Delete
  4. Thanks so much for the DUD. Worked great for me.

    ReplyDelete
  5. I have a Dell R710 running PERC H700 RAID. I am trying to install CentOS 8.0.1911. I downloaded the dd-mpt3sas-27.101.00.00-1.el8_0.elrepo.iso for the system, but it would not work. Two questions:

    1. Is that the correct ISO for the system?

    2. When trying to manually install this ISO, I get the following error: "modprobe: ERROR: could not insert 'mpt3sas': Invalid argument". Any ideas as to why?

    Thanks!

    ReplyDelete
    Replies
    1. It would be best if you would please make use of our mailing list [1] for support queries. More people will see your question and assistance may be obtained somewhat more promptly.

      [1] http://lists.elrepo.org/mailman/listinfo/elrepo

      Delete
    2. Thanks for the information!

      Delete
  6. hi boys, well done for everything you do. i'm on poweredge 2950 III perc 6 / ir and i still don't know which one to take as dd. I am trying 2

    ReplyDelete
  7. Hello all

    Just did it on PowerDell T310 (LS20089 and "almost" went well but the small advice here.
    If your BIOS setting is UEFI, you will not have access TAB at first menu to add the inst.dd parameter. Thus, be sure you're in BIOS mode before booting on CentOS. (that can be change back to UEFI after the installation).
    It is now all working fine seing all my RAID array in CentOS8

    ReplyDelete
    Replies
    1. Would you mind telling me what DUD iso you used? I have the same RAID controller and wasn't able to get it working with any of the megaraid_sas

      Delete
    2. I have T310 and I installed CentOS 8 on SATA SSD - used 4 SATA ports on the motherboard. Then I replaced HDD from the controller caddy to 2.5" SSDs (Just screwed them in the caddy and they inserted well - no adapter needed.) and created VD array from the controller menu but they were not visible to the CentOS. Thank you for the article - it is great help. This is what I did.

      # lspci -nn | grep -i lsi
      03:00.0 SCSI storage controller [0100]: Broadcom / LSI SAS1068E PCI-Express Fusion-MPT SAS [1000:0058] (rev 08)

      Search 1000:0058 for the rpm ISO in the link

      wget https://elrepo.org/linux/dud/el8/x86_64/dd-mptsas-3.04.20-4.el8_3.elrepo.iso

      mkdir /mnt/iso
      mount -t iso dd-mptsas-3.04.20-4.el8_3.elrepo.iso /mnt/iso
      cd /mnt/iso
      cd rpms/
      cd x86_64/
      rpm -ivh kmod-mptsas-3.04.20-4.el8_3.elrepo.x86_64.rpm

      Then rebooted - all VDs are now seen.

      Delete
  8. Hi ELRepo team, any chance you could provide documentation for how to make a DUD? I am trying to netinstall CentOS 8 in an HP Z800 workstation and having no luck because the NIC is not recognized by the installer. I already have the DUD for my LSI SAS controller, but I need to make a DUD for the tg3 NIC driver and of course RedHat's documentation doesn't go into detail.

    ReplyDelete
    Replies
    1. The tg3 driver is included in RHEL/CentOS 8. Therefore you should not need a kmod package for that. If you require more assistance, please make use of our mailing list.

      Delete
  9. Wow, thank you, I could have gotten this machine back up and working without this resource.

    ReplyDelete
  10. Hi there toracat...

    Iam using intel sever S2600JF, node server using eRSTe (isci module), im not using raid or intel eRST2 because i want all the disk present like HBA.
    at intel website for centos8 they only support Intel RSTe2

    all the disk are attach using isci module, i cant see the whole drive while installation
    i was trying to inst.dd with your dd-isci-1.2.0-1.el8.elrepo.iso

    but after continue the error said :
    modprobe error : could not insert 'isci' unknown symbol in module, or unkown parameter (see dmesg)

    I wonder if anything I do wrong ?
    could you help me choose the right module to insert while installation sir ?

    Thank you.

    ReplyDelete
    Replies
    1. dd-isci-1.2.0-1.el8.elrepo.iso is for RHEL/CentOS 8.0. Is this the version you are trying to install? Not the latest 8.2?

      Delete
  11. I am trying to install centos 8 on Ibm SystemX 3650 m3 server, same SAS drive issue, disk not detect, tried to use dud method , inst.dd option, make usb bootable with iso mega raid file, usb is mount , but showing no driver detect, please support

    ReplyDelete
  12. Just migrated from centos 6 to rhel 8.3 as cent is going byebyes. Boot up rhel 8.3 but no mass storage...come here 4 minutes later its back. Saves more fuffing around.

    Thank you kindly.

    ReplyDelete
  13. trying to (safely) fix my problem - I installed Centos 8.2 (4.18.0.193) on a dell with H700 LSI controller with the extra megaraid-sas driver on usb stuck. install went fine. examining the lib/modules/weak-updates folder I see megaraid-sas as it should be. however when I did a kernel update to 4.18.0.240 and rebooted, the megaraid-sas driver was not installed so no hard drive on boot.
    Now examinging my folders on the 240 kernel there is no megaraid-sas under weak updates. I see this is due to the dracut bug, but I'm running dracut-049-95.git20200804.el8_3.4.x86_64 and though the bug had been fixed!
    This is a PRODUCTION server I'm trying to finesse any help would be appreciated.
    The kmod under 8.2 was kmod-megaraid_sas-07.710.50.00-1.el8_2.elrepo.x86_64
    How do I (with maximum safety and minimum risk of screw up get my 240 kernel updated to include the right kmod? (which I assume is no longer el8_2...)?
    Many thanks. and REDHAT WHAT WERE YOU THINKING? BAD MOVE SPACE CADET!

    ReplyDelete
    Replies
    1. For RHEL/CentOS 8.3, you need the package with a .el8_3 tag. It is:

      https://elrepo.org/linux/elrepo/el8/x86_64/RPMS/kmod-megaraid_sas-07.714.04.00-2.el8_3.elrepo.x86_64.rpm

      This is for the kernel -240 series.

      Delete
    2. I gathered as much, but I'm not overly confident about how to 'slip' this into the 240 kernel easily with a minimum of fuss?
      are there directions anywhere? and my dracut still seems to have the bug I forced it to rebuilt the 240 kernel, but still did not get the kmod driver....

      Delete
    3. If you are looking for support, the elrepo mailing list (or your distribution mailing list) is probably a better venue as very few people will see your comment here, let alone reply.

      Delete