grub2 needs to properly manage kernel device name changes (nvme-eui -> nvme-nvme)

[theofficialgman]

failed to install on upgrade on mantic

ProblemType: Package
DistroRelease: Ubuntu 23.10
Package: shim-signed 1.56+15.7-0ubuntu1
ProcVersionSignature: Ubuntu 6.5.0-7.7-generic 6.5.3
Uname: Linux 6.5.0-7-generic x86_64
.proc.sys.kernel.moksbstate_disabled: Error: [Errno 2] No such file or directory: '/proc/sys/kernel/moksbstate_disabled'
ApportVersion: 2.27.0-0ubuntu5
Architecture: amd64
BootEFIContents:
 BOOTX64.CSV
 grub.cfg
 grubx64.efi
 mmx64.efi
 shimx64.efi
CasperMD5CheckResult: unknown
Date: Wed Oct 11 17:41:29 2023
ErrorMessage: installed shim-signed package post-installation script subprocess returned error exit status 32
InstallationDate: Installed on 2021-06-21 (842 days ago)
InstallationMedia: Ubuntu 20.04.2.0 LTS "Focal Fossa" - Release amd64 (20210209.1)
Python3Details: /usr/bin/python3.11, Python 3.11.6, python3-minimal, 3.11.4-5
PythonDetails: /usr/bin/python3.11, Python 3.11.6, python-is-python3, 3.11.4-1
RelatedPackageVersions:
 dpkg 1.22.0ubuntu1
 apt 2.7.3
SecureBoot: 6 0 0 0 1
SourcePackage: shim-signed
Title: package shim-signed 1.56+15.7-0ubuntu1 failed to install/upgrade: installed shim-signed package post-installation script subprocess returned error exit status 32
UpgradeStatus: Upgraded to mantic on 2023-08-10 (62 days ago)

[Steve Langasek (vorlon)]

> Setting up shim-signed (1.56+15.7-0ubuntu1) ...
> mount: /var/lib/grub/esp: special device /dev/disk/by-id/nvme-eui.0025384361b46f51-part1 does not exist.
> dmesg(1) may have more information after failed mount system call.
> dpkg: error processing package shim-signed (--configure):
> installed shim-signed package post-installation script subprocess returned error exit status 32

Your df output shows that you do have an EFI system partition mounted at /boot/efi. Please show the output of ls /dev/disk/by-id on this system.

[theofficialgman (theofficialgman)]

This update was run from KDE software center (so no terminal) and that's where it failed.

When run from a terminal afterwards I was prompted with a terminal dialog prompt to regenerate grub due to (apparently) the id of the system drive changing.

[Steve Langasek (vorlon)]

Please show the output of ls /dev/disk/by-id on this system.

[theofficialgman (theofficialgman)]

Note that before your first response I already updated successfully via terminal with `sudo apt install --fix-broken` as I wrote in the previous post. So I have updated and rebooted since then and am not sure if this helps now at this point.

sudo ls /boot/efi
 EFI 'System Volume Information'

ls /dev/disk/by-id
nvme-nvme.144d-533243544e584148333035393838-53414d53554e47204d5a564c5635313248434a482d3030304831-00000001
nvme-nvme.144d-533243544e584148333035393838-53414d53554e47204d5a564c5635313248434a482d3030304831-00000001-part1
nvme-nvme.144d-533243544e584148333035393838-53414d53554e47204d5a564c5635313248434a482d3030304831-00000001-part2
nvme-nvme.144d-533243544e584148333035393838-53414d53554e47204d5a564c5635313248434a482d3030304831-00000001-part3
nvme-nvme.144d-533243544e584148333035393838-53414d53554e47204d5a564c5635313248434a482d3030304831-00000001-part4
nvme-nvme.144d-533243544e584148333035393838-53414d53554e47204d5a564c5635313248434a482d3030304831-00000001-part5
nvme-nvme.144d-533243544e584148333035393838-53414d53554e47204d5a564c5635313248434a482d3030304831-00000001-part6
nvme-SAMSUNG_MZVLV512HCJH-000H1_S2CTNXAH305988
nvme-SAMSUNG_MZVLV512HCJH-000H1_S2CTNXAH305988_1
nvme-SAMSUNG_MZVLV512HCJH-000H1_S2CTNXAH305988_1-part1
nvme-SAMSUNG_MZVLV512HCJH-000H1_S2CTNXAH305988_1-part2
nvme-SAMSUNG_MZVLV512HCJH-000H1_S2CTNXAH305988_1-part3
nvme-SAMSUNG_MZVLV512HCJH-000H1_S2CTNXAH305988_1-part4
nvme-SAMSUNG_MZVLV512HCJH-000H1_S2CTNXAH305988_1-part5
nvme-SAMSUNG_MZVLV512HCJH-000H1_S2CTNXAH305988_1-part6
nvme-SAMSUNG_MZVLV512HCJH-000H1_S2CTNXAH305988-part1
nvme-SAMSUNG_MZVLV512HCJH-000H1_S2CTNXAH305988-part2
nvme-SAMSUNG_MZVLV512HCJH-000H1_S2CTNXAH305988-part3
nvme-SAMSUNG_MZVLV512HCJH-000H1_S2CTNXAH305988-part4
nvme-SAMSUNG_MZVLV512HCJH-000H1_S2CTNXAH305988-part5
nvme-SAMSUNG_MZVLV512HCJH-000H1_S2CTNXAH305988-part6

[Steve Langasek (vorlon)]

Thank you. Your logs show that shim-signed was being update from 1.54+15.7-0ubuntu1 to 1.56+15.7-0ubuntu1.
shim-signed 1.54+15.7-0ubuntu1 was published to Ubuntu on 2023-03-31.

Between March and now, have you made any changes to the hardware in this system? Have you had to replace your hard drive at any point?

I am looking to establish that the change to the device name from nvme-eui.0025384361b46f51 to nvme-nvme.144d-533243544e584148333035393838-53414d53554e47204d5a564c5635313248434a482d3030304831 is the result of a behavior change in the kernel, and not as a result of a change to the physical hardware since the last time this command was successfully executed.

[Steve Langasek (vorlon)]

(Because if it is a result of a behavior in the kernel, it is our responsibility to automatically detect this and handle it instead of erroring out.)

[theofficialgman (theofficialgman)]

This occurred on my laptop so I can pretty confidently state that there have been no hardware changes to my system over that period of time (or ever).

[theofficialgman (theofficialgman)]

just did some quick searching. it does appear to have been caused by a kernel update for a user of rocky linux https://forums.rockylinux.org/t/lvm-device-identifiers-changed-after-kernel-updates-resulting-in-missing-vgs-pvs/10059 . same change from nvme-eui to nvme-nvme

not sure what kernel I would have been using back in march (I use whatever is default for the release I am on). I generally hop to the latest in development ubuntu release about 1-2 months before its release. So at that point in march 2023 I am sure I was already using the 23.04 (mantic) developement release. So probably was running this kernel at the time:
linux-meta (6.2.0.19.19) lunar; urgency=medium

  * Bump ABI 6.2.0-19

 -- Andrea Righi <email address hidden> Sat, 25 Mar 2023 09:47:55 +0100

[Steve Langasek (vorlon)]

> This occurred on my laptop so I can pretty confidently state that there
> have been no hardware changes to my system over that period of time (or
> ever).

Well it is your hardware in any case, so you would be in a position to know! (Swapping out hard drives in a laptop is certainly a possibility.)

> just did some quick searching. it does appear to have been caused by a
> kernel update for a user of rocky linux
> https://forums.rockylinux.org/t/lvm-device-identifiers-changed-after-
> kernel-updates-resulting-in-missing-vgs-pvs/10059 . same change from
> nvme-eui to nvme-nvme

Yes, so this is a change in kernel device names which we need to handle gracefully on the grub side instead of throwing errors / debconf prompts.

[Julian Andres Klode (juliank)]

Oh god this is horrible. While we plan to address this for UEFI by switching to UUID we also need to handle this for grub-pc, and we really need stable disk ids, kernel can't go around changing disk ids, I mean their entire purpose is to be a stable name and this subverts it.

[Mate Kukri (mkukri)]

In this case it's the kernel's fault, but this is still sort of a dupe of 1940723, and the only real long term solution is switching to UUID based installs as per 1940723