Unable to deploy HWE kernel with sub-arch set to xgene-uboot

[Po-Hsu Lin]

When trying to deploy Moonshot nodes, the sub-arch must be set to xgene-uboot for it to be deployed with GA kernel.

However, when trying to deploy it with the HWE kernel it will complain that:
    Subarchitecture(xgene-uboot) must be generic when setting hwe_kernel.

[ deleted the bit about how generic/hwe fails, because that is expected. this machine requires xgene-uboot/hwe ]

[Po-Hsu Lin (cypressyew)]

Attachment is the console output on mcdivitt35-kernel with X + hwe-16.04

[Bill Wear (billwear)]

Don't know how much this helps, but in the troubleshooting guide at https://discourse.maas.io/t/troubleshooting-snap-3-0-ui/4134#heading--django-subarch-error, it mentions a Django error indicating that HWE kernels must choose "generic." The suggested workaround for Xenial is, for example, to try Focal.

I don't think that completely answers the question, but it might prompt someone to find the answer more easily.

[Bill Wear (billwear)]

Also, looking a the Xenial hardware enablement page, I see this:

Ubuntu 16.04 LTS - Xenial Xerus

The 16.04.2 and newer point releases will ship with an updated kernel and X stack by default for the desktop. Server installations will default to the GA kernel and provide the enablement kernel as optional.

The 16.04 HWE Stacks will follow a new Rolling Update Model as documented at the following location:

https://wiki.ubuntu.com/Kernel/RollingLTSEnablementStack

It is highly recommended to read the above documentation before executing the following commands, as the HWE model has changed in 16.04.

Installing the HWE stack is simple:

Desktop

 sudo apt-get install --install-recommends linux-generic-hwe-16.04 xserver-xorg-hwe-16.04

Server

 sudo apt-get install --install-recommends linux-generic-hwe-16.04

Also found this old bug (https://bugs.launchpad.net/curtin/+bug/1640519) about similar issues; this page of potential bits (http://ports.ubuntu.com/ubuntu-ports/dists/xenial/main/installer-arm64/current/images/netboot/ubuntu-installer/arm64/xgene-uboot/); and this related bug by Po-Hsu-Lin (https://bugs.launchpad.net/ubuntu/+source/linux-signed-hwe-5.8/+bug/1926440).

This makes me believe that we do not support anything but generic architecture on HWE kernels, but not marking as "Invalid" because I can't prove it. Instead, marking as "Incomplete" and asking the submitters to request more research (maybe from MAAS core team?) on whether or not this configuration is supported.

[Bill Wear (billwear)]

Patricia Domingues found the area in the code where the xgene-uboot kernel should be loaded, but fails to work, as part of #1918978. Marking this one as triaged.

[Jerzy Husakowski (jhusakowski)]

Unclear if the combination of subarch xgene-uboot and HWE is a valid one, asking the kernel team to advise.

[Ubuntu Kernel Bot (ubuntu-kernel-bot)]

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 1922910

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.

[Launchpad Janitor (janitor)]

[Expired for linux (Ubuntu) because there has been no activity for 60 days.]

[Launchpad Janitor (janitor)]

[Expired for MAAS because there has been no activity for 60 days.]

[dann frazier (dannf)]

I'm not on the kernel team, but I'm not sure what would make this flavor invalid w/ HWE. The xgene-uboot flavor is merely a wrapped version of the -generic kernel. This kernel works fine on the target platform, once you can get it installed.

[dann frazier (dannf)]

I wonder if we just need to treat xgene-uboot as an alias for generic in the MAAS code?

[Alberto Donato (ack)]

From the comments above, it seems the generic kernel does not work on this machine, and the xgene-uboot is required?

[dann frazier (dannf)]

Well, yes and no. xgene-uboot *is* required. But xgene-uboot *is* the generic kernel - it just has a u-boot wrapper prepended. So what I meant by comment #10, is that I wonder should allow xgene-uboot in the same cases where we allow -generic for HWE kernels, such as this code from src/maasserver/utils/osystems.py:

    if subarch != "generic" and (
        (hwe_kernel and validate_kernel_str(hwe_kernel))
        or (min_hwe_kernel and validate_kernel_str(min_hwe_kernel))
    ):
        raise ValidationError(
            "Subarchitecture(%s) must be generic when setting hwe_kernel."
            % subarch
        )

[Björn Tillenius (bjornt)]

It's not a simple as allowing xgene-uboot for HWE kernels in the MAAS code. If we'd do that, you'd end up with basically a generic subarchitecture.

We'd need to change the streams. Currently we overload subarchitecture a bit, so most likely we'd need to add new entries for xgene-uboot+hwe. But that also probably wouldn't be enough, we'd also have to change the MAAS code to cope with the new arhitecture/hwe combination.

So to sum it up, this is not a bug, since it was designed to work this way. If we want to add this, we'd need to get it on the roadmap and prioritized, since it's a significant amount of work.

[Björn Tillenius (bjornt)]

I'm closing this, since it's a feature request. Please add a Jira card for it in the product feedback board.