Ubuntu
linux package

5.4.0-67.75 breaks compat with 3rd party Mellanox OFED drivers

Bug #1919392 reported by dann frazier
24
This bug affects 4 people
Affects Status Importance Assigned to Milestone
linux (Ubuntu)
Won't Fix
Undecided
Unassigned
Focal
Won't Fix
Undecided
Unassigned
Groovy
Won't Fix
Undecided
Unassigned
Hirsute
Won't Fix
Undecided
Unassigned

Bug Description

[Impact]
Releases of Mellanox OFED drivers no longer compile against latest focal kernel. While Ubuntu can not necessarily guarantee source compatibility with 3rd party trees, it seems like the MOFED stack has enough users that we'd want to avoid breaking it unnecessarily.

[Fix]
Temporarily reverting the following commits until MOFED can release a new version that supports them would do the trick:

f21f647e08bd tools headers UAPI: Sync linux/const.h with the kernel headers
37bbb8ed3e93 uapi: move constants from <linux/kernel.h> to <linux/const.h>

[Test Case]
Install MOFED driver stack:

export DEBIAN_FRONTEND=noninteractive
NVIDIA_DRIVER_VERSION=450-server
MELLANOX_OFED_VERSION=5.1-2.5.8.0

install_mellanox_ofed() {
    local MLNX_REPO="https://linux.mellanox.com/public/repo/mlnx_ofed"
    local mlnx_ofed_ver="$1"
    local ubuntu_ver
    ubuntu_ver="$(lsb_release -rs)"
    mlnx_url="${MLNX_REPO}/${mlnx_ofed_ver}/ubuntu${ubuntu_ver}/mellanox_mlnx_ofed.list"

    wget -qO - https://www.mellanox.com/downloads/ofed/RPM-GPG-KEY-Mellanox | \
 apt-key add -
    wget -q -O /etc/apt/sources.list.d/mellanox_mlnx_ofed.list \
  "$mlnx_url"
    apt update
    apt install mlnx-ofed-kernel-only -y
}

install_mellanox_ofed ${MELLANOX_OFED_VERSION}

Revision history for this message
dann frazier (dannf) wrote :
Revision history for this message
Ubuntu Kernel Bot (ubuntu-kernel-bot) wrote : Missing required logs.

This bug is missing log files that will aid in diagnosing the problem. While running an Ubuntu kernel (not a mainline or third-party kernel) please enter the following command in a terminal window:

apport-collect 1919392

and then change the status of the bug to 'Confirmed'.

If, due to the nature of the issue you have encountered, you are unable to run this command, please add a comment stating that fact and change the bug status to 'Confirmed'.

This change has been made by an automated script, maintained by the Ubuntu Kernel Team.

Changed in linux (Ubuntu):
status: New → Incomplete
Changed in linux (Ubuntu Focal):
status: New → Incomplete
Changed in linux (Ubuntu Groovy):
status: New → Incomplete
dann frazier (dannf)
Changed in linux (Ubuntu Hirsute):
status: Incomplete → Confirmed
Changed in linux (Ubuntu Groovy):
status: Incomplete → Confirmed
Changed in linux (Ubuntu Focal):
status: Incomplete → Confirmed
Revision history for this message
walkerstreet (dbonner) wrote :

This bug affects all Groovy kernels from 5.8.0-45 or higher.
This bug did not affect Groovy kernel 5.8.0-44 or earlier.

To fix it:

git clone --recursive git://git.launchpad.net/~ubuntu-kernel/ubuntu/+source/linux/+git/groovy

cd groovy

git log --grep="tools headers UAPI: Sync linux/const.h with the kernel headers"

git log --grep="uapi: move constants from <linux/kernel.h> to <linux/const.h>"

git revert ecd8d70a8da1ab4987b488510c476be9052def19

git revert 5edb59c030419e21af5961ab60610f39a3c218e1

sudo nano /etc/apt/sources.list
--------------------------------
Uncomment these lines:
.......................
deb-src http://archive.ubuntu.com/ubuntu groovy main restricted
...
deb-src http://archive.ubuntu.com/ubuntu groovy-updates main restricted
........................

sudo apt-get build-dep linux linux-image-$(uname -r)

sudo apt-get install libncurses-dev gawk flex bison openssl libssl-dev dkms libelf-dev libudev-dev libpci-dev libiberty-dev autoconf

LANG=C fakeroot debian/rules clean

LANG=C fakeroot debian/rules binary-headers binary-generic binary-perarch

cd ..

rm linux-cloud-tools*

sudo dpkg -i *

Revision history for this message
dann frazier (dannf) wrote :

The 5.3 release of MOFED, which is now available, should fix this.

https://www.mellanox.com/products/infiniband-drivers/linux/mlnx_ofed

Marking Won't Fix for the Ubuntu side.

Changed in linux (Ubuntu Hirsute):
status: Confirmed → Won't Fix
Changed in linux (Ubuntu Focal):
status: Confirmed → Won't Fix
Changed in linux (Ubuntu Groovy):
status: Confirmed → Won't Fix
dann frazier (dannf)
Changed in linux (Ubuntu):
status: Confirmed → Won't Fix
Revision history for this message
dann frazier (dannf) wrote :

fyi, there are now releases of the OFED 5.1 and 5.2 branches that support the latest Ubuntu kernels:

OFED-5.1-2.6.2.0
OFED-5.2-2.2.3.0

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.
Subscribing...

Other bug subscribers

Bug attachments

Remote bug watches

Bug watches keep track of this bug in other bug trackers.