Mlnx-ofed-kernel-dkms fails to build on kernel 6.14.0-1015-nvidia (aarch64), succeeds on 6.14.0-1013-nvidia

Accelerated Computing DGX Spark / GB10 User Forum DGX Spark / GB10
1

Hi Nv member:
I’m seeing a DKMS build failure when installing mlnx-ofed-kernel-dkms (OFED 25.07) on an aarch64 system with NVIDIA kernels.
On the same machine, the DKMS build works fine on kernel 6.14.0-1013-nvidia, but fails on 6.14.0-1015-nvidia. Because both kernels are installed, the package installation fails overall.

The failure happens during the DKMS build step:

./ofed_scripts/pre_build.sh 6.14.0-1015-nvidia …

and DKMS reports a bad exit status for 6.14.0-1015-nvidia.

Kernel headers for 6.14.0-1015-nvidia are installed, and no changes were made between the two kernels other than the patch level. This makes it look like a kernel compatibility issue specific to 1015.

Or I need the install the driver by the following link?

Could you please advise whether kernel 6.14.0-1015-nvidia is expected to be supported by OFED 25.07, or if there is a recommended workaround or fix?

Found pre-existing /lib/modules/6.14.0-1013-nvidia/kernel/drivers/net/ethernet/mellanox/mlx5/core/mlx5_dpll.ko.zst, archiving for uninstallation
Installing /lib/modules/6.14.0-1013-nvidia/updates/dkms/mlx5_dpll.ko.zst
Installing /lib/modules/6.14.0-1013-nvidia/updates/dkms/fwctl.ko.zst
Installing /lib/modules/6.14.0-1013-nvidia/updates/dkms/mlx5_fwctl.ko.zst
Running the post_install script:

----- mlnx-ofed-kernel post-install ------

Package ‘srp-dkms’ is not installed, skipping module ‘srp’.

Package ‘iser-dkms’ is not installed, skipping module ‘iser’.

Package ‘isert-dkms’ is not installed, skipping module ‘isert’.

Package ‘mlnx-nfsrdma-dkms’ is not installed, skipping module ‘mlnx-nfsrdma’.

Package ‘mlnx-nvme-dkms’ is not installed, skipping module ‘mlnx-nvme’.

Running depmod… done.

Building initial module mlnx-ofed-kernel/25.07.OFED.25.07.0.9.7.1 for 6.14.0-1015-nvidia
Deprecated feature: CLEAN (/var/lib/dkms/mlnx-ofed-kernel/25.07.OFED.25.07.0.9.7.1/source/dkms.conf)
Sign command: /usr/bin/kmodsign
Signing key: /var/lib/shim-signed/mok/MOK.priv
Public certificate (MOK): /var/lib/shim-signed/mok/MOK.der

Building module(s)…(bad exit status: 1)
Failed command:
./ofed_scripts/pre_build.sh 6.14.0-1015-nvidia /lib/modules/6.14.0-1015-nvidia/build mlnx-ofed-kernel 25.07.OFED.25.07.0.9.7.1
ERROR: Cannot create report: [Errno 17] File exists: ‘/var/crash/mlnx-ofed-kernel-dkms.0.crash’

Error! Bad return status for module build on kernel: 6.14.0-1015-nvidia (aarch64)
Consult /var/lib/dkms/mlnx-ofed-kernel/25.07.OFED.25.07.0.9.7.1/build/make.log for more information.
dpkg: error processing package mlnx-ofed-kernel-dkms (–configure):
installed mlnx-ofed-kernel-dkms package post-installation script subprocess returned error exit status 10
Processing triggers for initramfs-tools (0.142ubuntu25.5) …
update-initramfs: Generating /boot/initrd.img-6.14.0-1015-nvidia
Errors were encountered while processing:
mlnx-ofed-kernel-dkms
needrestart is being skipped since dpkg has failed
E: Sub-process /usr/bin/dpkg returned an error code (1)

2

I am using the MLNX OFED driver from the NVIDIA link below and am experiencing the same DKMS issue. Could you please review the log below and advise accordingly?
Thank you.

image
image937×431 29.7 KB

~/Desktop/mlnx/MLNX_OFED_LINUX-24.10-3.2.5.0-ubuntu24.04-aarch64$ sudo ./mlnxofedinstall

Building initial module for 6.14.0-1015-nvidia
Deprecated feature: CLEAN (/var/lib/dkms/mlnx-ofed-kernel/24.10.OFED.24.10.3.2.5.1/source/dkms.conf)
ERROR: Cannot create report: [Errno 17] File exists: ‘/var/crash/mlnx-ofed-kernel-dkms.0.crash’

Error! Bad return status for module build on kernel: 6.14.0-1015-nvidia (aarch64)
Consult /var/lib/dkms/mlnx-ofed-kernel/24.10.OFED.24.10.3.2.5.1/build/make.log for more information.
dpkg: error processing package mlnx-ofed-kernel-dkms (–install):
installed mlnx-ofed-kernel-dkms package post-installation script subprocess returned error exit status 10
Processing triggers for initramfs-tools (0.142ubuntu25.5) …
update-initramfs: Generating /boot/initrd.img-6.14.0-1015-nvidia
Errors were encountered while processing:
mlnx-ofed-kernel-dkms

3

DOCA is not supported or recommended on DGX Spark. The main reason for this is because you would need to turn Secure Boot off to install and uninstall.

4

Hi @NVES :
The above explanation is based on Secure Boot already being disabled. This looks more like a driver version incompatibility issue.

Thank you!

5

Certain DOCA-OFED versions will not work with certain kernel versions. We do not test DOCA on the Spark so we cannot say which versions will be compatible currently or in the future

6

Hi @aniculescu :
I think the main issue is that the latest mlnx-ofed-kernel-dkms downloaded from NVIDIA’s website cannot be installed on 6.14.0-1015-nvidia. Do you have any suggestions on this?

Thank you.

7

Like I said, we do not test DOCA on the Spark GB10 platform so things may break.

8

Hi @aniculescu
Thank you for your reply. I understand that this is a compatibility issue. Currently, the MLNX-OFED driver fails on the 6.14.0-1015-nvidia kernel, which is expected behavior, and it only works properly on 6.14.0-1013-nvidia.