Hi, Pascal Hambourg <pas...@plouf.fr.eu.org> wrote (Sun, 12 Jan 2025 16:13:21 +0100): > On 12/01/2025 at 14:51, Holger Wansing wrote: > > Pascal Hambourg <pas...@plouf.fr.eu.org> wrote (Sun, 12 Jan 2025 14:13:40 > > +0100): > >> On 12/01/2025 at 11:53, Holger Wansing wrote: > >>> Cyril Brulebois <k...@debian.org> wrote (Sun, 12 Jan 2025 11:21:08 +0100): > >>>> oliviosu_ppc6...@tutanota.com <oliviosu_ppc6...@tutanota.com> > >>>> (2025-01-12): > >>>>> Hello, > >>>>> installing Trixie Alpha 1 on Tyan TN71-BP012 (POWER8) > >>>>> (/images/trixie_di_alpha1/ppc64el 2024/12/31) > >>>>> > >>>>> Installation goes well but installed system isn't bootable with > >>>>> Petitboot. > >>>>> Installation was made choosing automated partitioning for an entrier > >>>>> disk with LVM and luks. > >>>>> Doing that unencrypted /boot partition is ext4 and Petitboot can't see > >>>>> it. > >>>>> When partitioning if I edit /boot partition and set it to ext2 then > >>>>> after finishing installation, > >>>>> PetitBoot see it and I'm able to boot in the installed system. > >>>> > >>>> Sorry to hear about that. Looping in debian-boot@ who's responsible for > >>>> the installer components that implement those choices. > >> > >> The move from ext2 to ext4 for /boot was in response to bug #985463 > >> <https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=985463> after it was > >> considered that bootloaders such as GRUB and U-Boot currently used with > >> Debian supported ext4. > >> > >> Inndeed it is unfortunate that a modern boot loader like Petitboot does > >> not support ext4. It is even unexpected from a boot loader based on a > >> Linux kernel, as mentioned in > >> <https://open-power.github.io/petitboot/overview.html>. > >> > >>> thanks for pointing me on this. > >>> I have just pushed a fix for this, reverting back to ext2: > >>> https://salsa.debian.org/installer-team/partman-auto/-/commit/a87cccbe0198b4dea2cc46a39d5f5b99501aa095 > >> > >> Partitioning recipes for ppc64el create a separate /boot partition only > >> when using LVM (and this was never changed). When using plain > >> partitions, /boot is on the ext4 root partition so the above fix will > >> have no effect. > > > > The report was for guided partitioning with encrypted LVM. > > So I assumed the above would fit. > > Am I missing something? > > No, I just wanted to make sure everyone is aware that the revert fixes > only this specific use case (guided partitioning using LVM with or > without encryption). The issue already existed and remains when using > guided partitioniong without LVM (ext4 root, no separate /boot).
Ok, so that's fine. With this fix in place, there is at least the LVM partitioning scheme (with or without encryption), which works out of the box with default choice. And assuming we got no complains before about this situation on this machine, I guess it's fine this way. > >> According to > >> <https://github.com/open-power/petitboot/blob/master/README.md> and > >> <https://open-power.github.io/petitboot/platforms.html>, Petitboot is > >> also available for arm64 platforms. Partitioning recipes for arm64 also > >> use ext4 for /boot, so it would not work either. > > > > Hmm, is it widely used on arm64, or only a cornercase? > > Is it worth to change the default on arm64 just for petitboot? > > One could ask the same about ppc64el. Is Petitboot provided by IBM as > part of this platform firmware ? The recipes for ppc64el create a PReP > partition which, as far as I know, is intended for OpenFirmware + GRUB. > > I have no clue about Petitboot on arm64, I didn't know anything about > Petitboot before today. A quick web search shows that Petitboot is used > on arm64-based Odroid boards. An article [1] (in french, sorry) suggests > that it supports ext4, so I wonder why it does not on the OP's machine. Maybe the OP has some information on this... Holger -- Holger Wansing <hwans...@mailbox.org> PGP-Fingerprint: 496A C6E8 1442 4B34 8508 3529 59F1 87CA 156E B076
