Server Images: Partitioning and file system are faulty #193

Open
opened 2026-04-29 09:31:12 +00:00 by pboy · 9 comments
Owner

a) KVM Guest image (qcow2)

  • In the partition table, the main data (payload) partition (partition 4) is listed as Linux (83), not as LVM (8e)

b) Aarch64 SBC raw image

  • In the partition table there is a gap of 16 mb before the first partition
  • In the partition table, the main data (payload) partition (partition 3) is listed as Linux (83), not as LVM (8e)
  • The LVM device file in /etc/lvm/devices is missing; only the ‘auto-import-root’ tag file exists. This could be due to the above error in the partition table. However, the file is present in the aarch64 raw image (which has also the wrong partition type).
a) KVM Guest image (qcow2) * In the partition table, the main data (payload) partition (partition 4) is listed as Linux (83), not as LVM (8e) b) Aarch64 SBC raw image * In the partition table there is a gap of 16 mb before the first partition * In the partition table, the main data (payload) partition (partition 3) is listed as Linux (83), not as LVM (8e) * The LVM device file in /etc/lvm/devices is missing; only the ‘auto-import-root’ tag file exists. This could be due to the above error in the partition table. However, the file is present in the aarch64 raw image (which has also the wrong partition type).
Author
Owner

@ngompa actioned to fix the LVM label part (the gap part needs further discussion)

@ngompa actioned to fix the LVM label part (the gap part needs further discussion)
pboy added this to the Maintenance project 2026-05-21 12:42:37 +00:00
Member

So it appears we set it correctly and then accidentally blow it away. I'm discussing how to fix it with upstream kiwi folks.

https://github.com/OSInside/kiwi/issues/3036

So it appears we set it correctly and then accidentally blow it away. I'm discussing how to fix it with upstream kiwi folks. https://github.com/OSInside/kiwi/issues/3036
Member

@pboy , can you specify precisely what image files you mean, and give the literal commands and output you see issues with?

I'm trying to determine whether the kiwi changed fixed this, but the description isn't really precise enough. It sort of seems like this isn't "fixed", when I test current Rawhide images (which are built with the updated kiwi), but can you confirm?

[adamw@toolbx fedora-toolbox-44 46]$ guestfish -a Fedora-Server-Guest-Generic-Rawhide-20260902.n.0.x86_64.qcow2 -i sfdisk-l /dev/sda
Disk /dev/sda: 10 GiB, 10737418240 bytes, 20971520 sectors
Disk model: QEMU HARDDISK   
Units: sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disklabel type: gpt
Disk identifier: 966E52FD-15D8-4DA3-912B-BC9C810774D9

Device       Start      End  Sectors  Size Type
/dev/sda1     2048     6143     4096    2M BIOS boot
/dev/sda2     6144  1030143  1024000  500M EFI System
/dev/sda3  1030144  5126143  4096000    2G Linux extended boot
/dev/sda4  5126144 20971486 15845343  7.6G Linux root (x86-64)
[adamw@toolbx fedora-toolbox-44 46]$ fdisk -l Fedora-Server-Host-Generic-Rawhide-20260902.n.0.aarch64.raw
Disk Fedora-Server-Host-Generic-Rawhide-20260902.n.0.aarch64.raw: 10 GiB, 10737418240 bytes, 20971520 sectors
Units: sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disklabel type: dos
Disk identifier: 0xac8d5b90

Device                                                       Boot   Start      End  Sectors  Size Id Type
Fedora-Server-Host-Generic-Rawhide-20260902.n.0.aarch64.raw1        34816  1058815  1024000  500M  6 FAT16
Fedora-Server-Host-Generic-Rawhide-20260902.n.0.aarch64.raw2      1058816  5154815  4096000    2G ea Linux extended boot
Fedora-Server-Host-Generic-Rawhide-20260902.n.0.aarch64.raw3      5154816 20971486 15816671  7.5G 83 Linux
@pboy , can you specify precisely what image files you mean, and give the literal commands and output you see issues with? I'm trying to determine whether the kiwi changed fixed this, but the description isn't really precise enough. It sort of seems like this isn't "fixed", when I test current Rawhide images (which are built with the updated kiwi), but can you confirm? ``` [adamw@toolbx fedora-toolbox-44 46]$ guestfish -a Fedora-Server-Guest-Generic-Rawhide-20260902.n.0.x86_64.qcow2 -i sfdisk-l /dev/sda Disk /dev/sda: 10 GiB, 10737418240 bytes, 20971520 sectors Disk model: QEMU HARDDISK Units: sectors of 1 * 512 = 512 bytes Sector size (logical/physical): 512 bytes / 512 bytes I/O size (minimum/optimal): 512 bytes / 512 bytes Disklabel type: gpt Disk identifier: 966E52FD-15D8-4DA3-912B-BC9C810774D9 Device Start End Sectors Size Type /dev/sda1 2048 6143 4096 2M BIOS boot /dev/sda2 6144 1030143 1024000 500M EFI System /dev/sda3 1030144 5126143 4096000 2G Linux extended boot /dev/sda4 5126144 20971486 15845343 7.6G Linux root (x86-64) ``` ``` [adamw@toolbx fedora-toolbox-44 46]$ fdisk -l Fedora-Server-Host-Generic-Rawhide-20260902.n.0.aarch64.raw Disk Fedora-Server-Host-Generic-Rawhide-20260902.n.0.aarch64.raw: 10 GiB, 10737418240 bytes, 20971520 sectors Units: sectors of 1 * 512 = 512 bytes Sector size (logical/physical): 512 bytes / 512 bytes I/O size (minimum/optimal): 512 bytes / 512 bytes Disklabel type: dos Disk identifier: 0xac8d5b90 Device Boot Start End Sectors Size Id Type Fedora-Server-Host-Generic-Rawhide-20260902.n.0.aarch64.raw1 34816 1058815 1024000 500M 6 FAT16 Fedora-Server-Host-Generic-Rawhide-20260902.n.0.aarch64.raw2 1058816 5154815 4096000 2G ea Linux extended boot Fedora-Server-Host-Generic-Rawhide-20260902.n.0.aarch64.raw3 5154816 20971486 15816671 7.5G 83 Linux ```
Author
Owner

I just tested with 0921 host and VM.
Same result. Using CLI with virt-install (our main user interface) infinite loop, same with Cockpit and serial console. After switching to graphical console (and restart required to activate graphic, according to message from Cockpit) the initial-setup screen showed up and I could complete i
Subsequent starts proceed as expected. The initial setup is disabled.

I still get error message: [FAILED] Failed to start systemd-repart.service - Repartition Root Disk.

And the LVM partition is still of the wrong type in fdisk / cfdisk.

I just tested with 0921 host and VM. Same result. Using CLI with virt-install (our main user interface) infinite loop, same with Cockpit and serial console. After switching to graphical console (and restart required to activate graphic, according to message from Cockpit) the initial-setup screen showed up and I could complete i Subsequent starts proceed as expected. The initial setup is disabled. I still get error message: [FAILED] Failed to start systemd-repart.service - Repartition Root Disk. And the LVM partition is still of the wrong type in fdisk / cfdisk.
Member

I still get error message: [FAILED] Failed to start systemd-repart.service - Repartition Root Disk.

That is an expected error as repart will fail for anything that isn't a GPT disk. Our bare metal ARM disks are MBR so that they are compatible with the Raspberry Pi 3.

> I still get error message: [FAILED] Failed to start systemd-repart.service - Repartition Root Disk. That is an expected error as repart will fail for anything that isn't a GPT disk. Our bare metal ARM disks are MBR so that they are compatible with the Raspberry Pi 3.
Member

We're testing x86_64 here, not ARM.

We're testing x86_64 here, not ARM.
Member

The Guest-Generic images are correct though, it's a GPT image, and it's set to the right values for GPT. Only the ARM Host-Generic images are MBR.

The Guest-Generic images are correct though, it's a GPT image, and it's set to the right values for GPT. Only the ARM Host-Generic images are MBR.

Both images' partition 3 use the partition type GUID/code defined in the UAPI Boot Loader Specification for the Linux extended boot partition. However, this partition is formatted XFS, therefore the type code is invalid. If the partition used for Linux boot will be XFS, then the type code needs to be linux file system data (83h on MBR, 0FC63DAF-8483-4772-8E79-3D69D8477DE4 on GPT).

On Fedora-Server-Guest-Generic-45-20260921.n.0.x86_64.qcow2, partition 4 is using the partition type code defined in the UAPI Discoverable Partitions Specification for a root partition on x86-64 architecture. This type code cannot be used for LVM members. The type code for this partition should be E6D6D379-F507-44C2-A23C-238F2A3DF928 on GPT and EAh on MBR. It's possible 0FC63DAF-8483-4772-8E79-3D69D8477DE4 and 83h are still OK. I'm not aware of any spec or tool that conflicts or enforces partition type codes on LVM members.

Both images' partition 3 use the partition type GUID/code defined in the UAPI Boot Loader Specification for the Linux extended boot partition. However, this partition is formatted XFS, therefore the type code is invalid. If the partition used for Linux boot will be XFS, then the type code needs to be linux file system data (83h on MBR, 0FC63DAF-8483-4772-8E79-3D69D8477DE4 on GPT). On Fedora-Server-Guest-Generic-45-20260921.n.0.x86_64.qcow2, partition 4 is using the partition type code defined in the UAPI Discoverable Partitions Specification for a root partition on x86-64 architecture. This type code cannot be used for LVM members. The type code for this partition should be E6D6D379-F507-44C2-A23C-238F2A3DF928 on GPT and EAh on MBR. It's possible 0FC63DAF-8483-4772-8E79-3D69D8477DE4 and 83h are still OK. I'm not aware of any spec or tool that conflicts or enforces partition type codes on LVM members.
Filed a bug here: https://github.com/OSInside/kiwi/issues/3090
Sign in to join this conversation.
No milestone
No project
No assignees
4 participants
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
server/tickets#193
No description provided.