-
Bug
-
Resolution: Unresolved
-
Critical
-
None
-
4.19
-
Quality / Stability / Reliability
-
False
-
-
1
-
Moderate
-
None
-
x86_64
-
None
-
None
-
Ready to Pick
-
1
-
In Progress
-
Release Note Not Required
-
None
-
None
-
None
-
None
-
None
Description of problem:
We're enabling some early Intel 6th Generation Xeon hardware from partners called GNR-D that requires an update to the RHCOS boot images. The presently running versions for 4.19: https://mirror.openshift.com/pub/openshift-v4/dependencies/rhcos/4.19/latest/ do not contain fixes pushed as a part of https://issues.redhat.com/browse/RHEL-92555, and we're having provisioning trouble as a result. We would like to request that RHCOS re-spin the 4.19 images so that this hardware can be discovered properly during provisioning, picking up fixes from RHEL-92555.
Problem in depth:
Presently with the shipping images (4.19.0) we see interfaces for an onboard NIC (E825C) come up in safe mode (as a consequence of not having the proper DDP firmware contained in the linux-firmware package) and instantiate themselves as: eno1np0 through eno7np0 Log from the host: Aug 15 22:27:18 api.sno132.outbound.borball.bos2.lab node-image-pull.sh[5681]: Checking out node image content Aug 15 22:27:19 api.sno132.outbound.borball.bos2.lab systemd[1]: Finished Node Image Pull. Aug 15 22:27:19 api.sno132.outbound.borball.bos2.lab systemd[1]: Starting Node Image Overlay... Aug 15 22:27:19 api.sno132.outbound.borball.bos2.lab node-image-overlay.sh[5953]: Overlaying node image content The linux-firmware package is seemingly upgraded (per logs): Fri Aug 15 22:27:19 UTC 2025 linux-firmware-whence-20250415-146.5.el9_5.noarch linux-firmware-20250415-146.5.el9_5.noarch Fri Aug 15 22:27:21 UTC 2025 linux-firmware-whence-20250604-151.2.el9_6.noarch linux-firmware-20250604-151.2.el9_6.noarch However, this is insufficient. We need to have the proper firmware built as a part of the initrd / initramfs, so that early initialization and loading of the DDP firmware for these nics after the bootloader runs ensures that they are discovered properly in the bootimage and in subsequent boots. The proper interfaces should be: eno1 through eno7 Otherwise, incorrect network configuration results and the node is offline on the network (both ipv4 and ipv6) after the first reboot.
Version-Release number of selected component (if applicable):
4.19.0
How reproducible:
Every time on this hardware
Steps to Reproduce:
1.Deploy OpenShift 4.19.5+ on this hardware using the Agent Based Installer or Red Hat ZTP through the assisted service / MCE / ACM using nmstate for the network configuration and 4.19.0 boot images from mirror.openshift.com 2. Observe the server's first boot that the boot image / discovery ISO provision, however it doesn't load proper DDP firmware and the E825 NIC comes up in "safe mode" for its interfaces (eno1np0 through eno7np0), which is undesirable 3. The node reboots after successful provisioning and comes up with the proper DDP firmware (linux-firmware as of 4.19.5 is fixed with the proper firmware), however networkmanager and sysfs recognizes different interfaces (eno1 through eno7) and the node cannot be accessed over the network
Actual results:
Node comes up and is inaccessible over the network
Expected results:
Node is provisioned with the same interface in the bootimage / discovery ISO and the final deployment (eno1 through eno7).
Additional info:
- account is impacted by
-
OCPBUGS-60925 [4.19] Bootimage bump tracker
-
- Closed
-
-
OCPBUGS-60949 [4.20] Bootimage bump tracker
-
- Closed
-
- relates to
-
OCPBUGS-55068 [4.19] Bootimage bump tracker
-
- Closed
-
- links to
-
RHBA-2025:15397 OpenShift Container Platform 4.20.0 bug fix update