Overall design

The initial focus here is updating the ESP, but the overall design of bootupd contains a lot of abstraction to support different "components".

Ideal case

In the ideal case, an OS builder uses

to install all bootloader data, and thereafter it is fully (exclusively) managed by bootupd. It would e.g. be a bug/error for an administrator to manually invoke

grub2-install

e.g. again.

In other words, an end user system would simply invoke

as desired.

However, we're not in that ideal case. Thus bootupd has the concept of "adoption" where we start tracking the installed state as we find it.

Handling adoption

For Fedora CoreOS, currently the

file is created outside of the ostree inside

create_disk.sh

. So we aren't including any updates for it in the OSTree.

This type of problem is exactly what bootupd should be solving.

However, we need to be very cautious in handling this because we basically can't assume we own all of the state. We shouldn't touch any files that we don't know about.

Upgrade edges

We don't necessarily want to update the bootloader data, even if a new update happens to be provided. For example, Fedora does "mass rebuilds" usually once a release, but it's not strictly necessary to update users' bootloaders then.

A common policy in fact might be "only update bootloader for security issue or if it's strictly necessary".

A "strictly necessary" upgrade would be one like the GRUB BLS parsing support.

There is not yet any support for upgrade edges in the code apart from a stub structure.