Failed to write boot loader configuration with Silverblue 41 on Framework 13

[alexjg]

I’ve just received a brand new DIY Framework 13 and I wanted to give Silverblue a try. I created an install USB using Fedora Mediawriter and started the installation. I used the automatic disk formatting option but ticked the box to encrypt my data and chose a passphrase and then let the installer get on with it.

At some point I received an error that read something like “failed to write boot loader configuration”. I click to let it continue installing but on reboot I got a GRUB error screen.

At first I thought this might be a failure in the hardware, so just to check I created a Fedora Workstation install USB on the same USB stick and installed that using the same options and that worked fine.

This reproduces reliably. I've tried using different USB sticks and I've also encountered the same issue when attempting to install Bluefin. I would love to provide you with some logs but it wasn't immediately obvious to me how to get them off of the device during installation. I'm quite happy to run through the installation again and take photographs of the screen if that would help, just let me know what I should be getting photographs of.

[travier]

This is likely #284 / https://bugzilla.redhat.com/show_bug.cgi?id=1575957.

Make sure to wipe your /boot & /efi partitions before the installation, either via a Live ISO, the console, or in the advanced disk setup option.

[alexjg]

Can you describe what I should do in the advanced disk setup option to achieve this? It sounds like you're suggesting that I manually create the partitions in order to ensure that the /efi and /boot partitions are removed but I'm not sure what partitions I should actually create.

[travier]

You need to make sure they are wiped / reformatted. If you are not dual booting, you can remove all partitions, then use the default layout.

[alexjg]

So I ran the installation again intending to do as you suggest but it looks like the successful installation of Workstation (which was the last thing I did before submitting this issue) has also wiped /efi and /boot partitions because the installation of Silverblue succeeded this time.

Thus, one easy way to fix this particular problem is to install vanilla Workstation, then immediately install Silverblue.

[OnyxAzryn]

I actually just got my new Framework 13 today and tried to install Silverblue as well using automatic partitioning with encryption and faced the same exact issue on a brand new SSD. However, I cannot seem to get passed this problem no matter what I do. I tried formatting the partitions, deleting and recreating them, and even tried to install Workstation then Silverblue, but none of these worked.

[travier]

Please provide logs or at least a screenshot of the error message.

[valberg]

I'm experiencing the same error. Here are some (sorry for bad quality) photos I took of the error:

[travier]

Thanks for the screenshots. It looks like there is potentially a bug in bootupd when there is already an EFI boot entry matching what we are installing.

[valberg]

@travier I tried using Fedora Workstation Live USB to use efibootmgr to remove all entries and then tried installing again, resulting in the same error.

[travier]

Did you make sure to wipe the EFI partition when installing?

[fotnite-vevo]

I was having the same issue on the same device. It occured on a completely new Framework 13 with a new SSD. Even after wiping the drive before trying again, it still occured every time. The Fedora 41 Workstation installer didn't have the same issue.

However, once I installed Fedora 41 Workstation, updated my firmware from there with fwupdmgr, then wiped the drive and tried to install Silverblue, the installation was successful.

[sudul]

Thanks @fotnite-vevo, I had the exact same issue (brand new AMD Framework 13) and installing Workstation first and updating seems to fix Silverblue installation afterwards.

The issue is probably on Framework's rather than Silverblue then.

[notsyncing]

Same here on a non-Framework 13 machine, definitely not a Framework's issue. Wiped the whole disk again and again, this problem persists.

Finally I let the installer continue when the error occurred, then copied /boot/grub2, /boot/efi/EFI/fedora/grub.cfg, /boot/loader/grub.cfg from another silverblue 41 machine, and manually fixed the disk uuids, ostree paths and kernel and initrd paths in these files, and it successfuly booted.

EDIT: You also have to edit /boot/loader/entries/ostree-*.conf to add something like the following to the options line:

rd.luks.uuid=luks-<your root device uuid> rhgb quiet root=UUID=<decrypted root device uuid> rootflags=subvol=root 

or you won't be able to boot again if you rebooted.

[travier]

OK, this is really weird. I would be great if someone that can reproduce the issue could post the full Anaconda logs somewhere and share them here.

[dchats1]

I've been running into the same issue. Here is my anaconda.log from my last attempt. Let me know if there is any other info I can provide that would help.
anaconda.log

[travier]

@dchats1 I don't see the backtrace from the comments above in your log. The Anaconda logs are likely split into multiple files so we would need them all.

[calebegg]

I'm also running into this with a brand new Intel framework 13. Installing Workstation first did not help in my case.

[travier]

Suggestion: Install Fedora Workstation first, do all the package updates, reboot, do all the firmware updates, reboot, then try Silverblue again.

[calebegg]

@travier That actually did work -- I think it was the firmware updates (I tried about six times total with different combinations of things before I finally got it to work). Thanks! I'm successfully running silverblue now.