Bug#1102604: rescue-mode: Provide rootfs subvolume selection menu
ti 13.5.2025 klo 23.26 Nicholas D Steeves (sten@debian.org) kirjoitti:
> Unless it's security-related, stable updates are minimal, and always
> occur after testing updates. Consequently, there needs to be a trixie
> D-I and rescue-mode first. Maybe that will be alpha2, maybe that will
> be the cycle after. Bookworm D-I and rescue-mode changes can maybe
> happen after that, and that's when you were asked to ping us. In other
> words: not now.
I just tested DI Trixie RC1. Starting from this release, it apparently
uses something else than GRUB-EFI to boot into EFI mode, and this
successfully got me into the rescue mode. The patch to support default
btrfs subvolumes works well: it immediately found the right subvolume
to mount, then asked me if I wanted to mount the EFI partition too.
Kudos to everyone who implemented this.
> > Btw, there's another compelling reason for backporting this: The
> > GRUB-EFI version that ships with Trixie currently barfs on common ASUS
> > motherboards. This makes Bookworm d-i the only usable way to rescue
> > those EFI hosts but, without that resuce mode backport, it requires a
> > lot of manual typing of mount commands just to get around the
> > subvolume issue.
>
> I hope you will consider reanalysing your conclusion vis à vis the facts
> you presented:
>
> Given
>
> 1. A trixie (testing) issue at what sounds like RC severity that makes
> the default installation unusable for a number of popular systems (ASUS
> motherboards) that has no workaround in trixie.
Kinda. No workaround if we use GRUB-EFI. However, systemd-boot works
fine on the same host. If someone doesn't need GRUB's boot menu, it's
a good enough workaround.
Also, as noted above, now that DI uses something else than GRUB-EFI to
boot in EFI mode, we succesfully get into the rescue mode, even on
btrfs hosts.
> 2. A bookworm bug of normal severity for a non-default installation
> that has a workaround in bookworm, but that is inconvenient.
If you meant that something else than ext4 is non-default, then yes.
However, btrfs very much is among the options one can select for
installation.
> Do you really think that an inconvenience is more significant than what
> you've presented as a release critical bug that affects the default
> installation?
It's more than an inconvenience. In a situation when someone needs to
use the rescue mode, their mind has to focus on rescueing the host,
not on memorizing a series of commands to type to compensate for the
rescue mode's lack of support for the default subvolume that was used
when installing the system.
> P.S. For the record, I'm Debian's btrfs-progs maintainer, I believe in
> btrfs and want to see it succeed, and I'm sorry for the inconvenience in
> bookworm's rescue-mode.
Which is precisely why I fully trust that this will eventually work.
In fact, now that DI no longer uses GRUB-EFI, it is already fixed for
Trixie on amd64. However, given how 2 architectures got dropped from
Trixie (i386, mips64el), we still need a backport, if we wanna avoid
doing the CLI dance before we can rescue a host.
Martin-Éric
Reply to: