Re: [PATCH v2] expand-on-first-boot: support resizing a btrfs root

["Moessbauer, Felix" <felix.moessbauer@siemens.com>]

On Mon, 2022-12-12 at 16:24 +0100, Henning Schild wrote:
> Am Fri, 22 Oct 2021 15:50:47 -0400
> schrieb Joe MacDonald <joe_macdonald@mentor.com>:
> 
> > resize2fs fails if the build sets the root filesystem for an image
> > to
> > btrfs though the partition is expanded properly. Since the rest of
> > the
> > script works as intended, simply checking the last partition type
> > with
> > lsblk and adding `btrfs resize` instead of `resize2fs` in the
> > appropriate circumstance allows everything to function as expected.
> > 
> > Signed-off-by: Joe MacDonald <joe_macdonald@mentor.com>
> > ---
> >  .../files/expand-last-partition.sh            | 22
> > ++++++++++++++----- 1 file changed, 17 insertions(+), 5 deletions(-
> > )
> > 
> > diff --git
> > a/meta/recipes-support/expand-on-first-boot/files/expand-last-
> > partition.sh
> > b/meta/recipes-support/expand-on-first-boot/files/expand-last-
> > partition.sh
> > index 7175dfd..6975d9c 100755 ---
> > a/meta/recipes-support/expand-on-first-boot/files/expand-last-
> > partition.sh
> > +++
> > b/meta/recipes-support/expand-on-first-boot/files/expand-last-
> > partition.sh
> > @@ -36,8 +36,20 @@ sfdisk -d "${BOOT_DEV}" 2>/dev/null | \ # Inform
> > the kernel about the partitioning change partx -u "${LAST_PART}" -#
> > Do not fail resize2fs if no mtab entry is found, e.g., -# when
> > using
> > systemd mount units. -export EXT2FS_NO_MTAB_OK=1
> > -
> > -resize2fs "${LAST_PART}"
> > +# Determine the filesystem type and perform the appropriate resize
> > function +ROOTFS_TYPE=$(lsblk -fno FSTYPE ${ROOT_DEV})
> 
> Calling "lsblk" suffers from the same udev race we have seen with
> systemd-growfs ... only very subtle where we get like "" ... no
> filesystem.
> 
> It seems to help to mount here and findmnt the mntpoint, we need the
> mounting for btrfs anyhow.

One problem remains: Where to mount.
If we mount on /mnt this will not work on an RO rootfs out of the box.
We currently have this issue in CIP core.

We could mount below /tmp which should be writable, but I don't know
which side effects this as the tmpfs itself will be mounted / created
by systemd.

Felix

> 
> Also lsblk can return multiple filesystems, like luks or lvm layers
> underneath. Where findmnt only returns the top.
> 
> Joe, i had to rewrite your patch considerably. So i think i will
> change
> the authorship to myself. I would really like to give you the credits
> but am afraid to give you blame ... this whole expand story seems
> never
> ending ...
> So the patch will mention you, but likely not as author.
> 
> Henning
> 
> > +case ${ROOTFS_TYPE} in
> > +   ext* )
> > +      # Do not fail resize2fs if no mtab entry is found, e.g.,
> > +      # when using systemd mount units.
> > +      export EXT2FS_NO_MTAB_OK=1
> > +      resize2fs "${LAST_PART}"
> > +      ;;
> > +   btrfs )
> > +      btrfs filesystem resize max /
> > +      ;;
> > +   * )
> > +      echo "Unrecognized filesystem type ${ROOTFS_TYPE} - no
> > resize
> > performed"
> > +      exit 1
> > +      ;;
> > +esac
>