• Bug#1106854: e2fsprogs: mkfs.ext4 1.47.3 regression for existing use ca

    From Luca Boccassi@21:1/5 to All on Fri May 30 16:30:01 2025
    Package: e2fsprogs
    Version: 1.47.3~rc1-1
    Affects: src:systemd

    Dear Maintainer(s),

    autopkgtest on debci has highlighted that the 1.47.3~rc1 version of
    mkfs.ext4 fails when used by systemd-repart, while it works with the
    version in unstable and testing:

    2310s mke2fs 1.47.3-rc1 (28-May-2025)
    2310s Discarding device blocks: 0/2097152 done
    2310s Creating filesystem with 2097152 4k blocks and 524288 inodes
    2310s Filesystem UUID: 997c41e7-f079-4609-985b-7fde097cc539
    2310s Superblock backups stored on blocks:
    2310s 32768, 98304, 163840, 229376, 294912, 819200, 884736, 1605632
    2310s
    2310s Allocating group tables: 0/64 done
    2310s Writing inode tables: 0/64 done
    2310s Creating journal (16384 blocks): done
    2311s Copying files into the device: __populate_fs: Extent block checksum does not match extent block while writing file "tput"
    2311s mkfs.ext4: Extent block checksum does not match extent block while populating file system
    2311s (mkfs) failed with exit status 1.
    2312s ‣ "systemd-repart --empty=allow --size=auto --dry-run=no --json=pretty --no-pager --root=/buildroot --offline=yes --seed 67e37ec0-9485-4fe2-ab81-b86e7eaa581a /work/var/tmp/integration-tests.QQ0MYsxIx2/mkosi-workspace-vsbhym0i/staging/image.raw --
    empty=create --defer-partitions esp,xbootldr --generate-fstab=/etc/fstab --generate-crypttab=/etc/crypttab --definitions /work/tmp/autopkgtest.Pev0im/build.cSZ/src/mkosi/mkosi.repart --definitions /work/tmp/autopkgtest.Pev0im/build.cSZ/src/mkosi/mkosi.
    repart --certificate /work/tmp/autopkgtest.Pev0im/build.cSZ/src/mkosi.crt --private-key /work/tmp/autopkgtest.Pev0im/build.cSZ/src/mkosi.key" returned non-zero exit code 1.

    https://ci.debian.net/packages/s/systemd/unstable/amd64/61010146/

    The least-effort way to reproduce is probably to run autopkgtest. Get
    the src:systemd sources from unstable/testing and then from the source directory run:

    autopkgtest -B --test-name upstream . -- <preferred autopkgtest runner config>

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Theodore Ts'o@21:1/5 to Luca Boccassi on Fri May 30 17:20:01 2025
    On Fri, May 30, 2025 at 03:27:56PM +0100, Luca Boccassi wrote:
    autopkgtest on debci has highlighted that the 1.47.3~rc1 version of
    mkfs.ext4 fails when used by systemd-repart, while it works with the
    version in unstable and testing:

    2310s mke2fs 1.47.3-rc1 (28-May-2025)
    2310s Discarding device blocks: 0/2097152 done
    2310s Creating filesystem with 2097152 4k blocks and 524288 inodes
    2310s Filesystem UUID: 997c41e7-f079-4609-985b-7fde097cc539
    2310s Superblock backups stored on blocks:
    2310s 32768, 98304, 163840, 229376, 294912, 819200, 884736, 1605632 2310s
    2310s Allocating group tables: 0/64 done
    2310s Writing inode tables: 0/64 done
    2310s Creating journal (16384 blocks): done
    2311s Copying files into the device: __populate_fs: Extent block checksum does not match extent block while writing file "tput"
    2311s mkfs.ext4: Extent block checksum does not match extent block while populating file system
    2311s (mkfs) failed with exit status 1.

    Thanks for the bug report. For the record e2fsprogs 1.47.3-rc1 was
    only uploaded to experimental, and not unstable or testing, precisely
    so I could get early feedback.

    This does look like some kind of e2fsprogs bug, and I'll take a look
    at it before releasing 1.47.3 final.

    Cheers,

    - Ted

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Theodore Ts'o@21:1/5 to Luca Boccassi on Fri Jun 6 21:10:02 2025
    On Fri, May 30, 2025 at 03:27:56PM +0100, Luca Boccassi wrote:

    autopkgtest on debci has highlighted that the 1.47.3~rc1 version of
    mkfs.ext4 fails when used by systemd-repart, while it works with the
    version in unstable and testing:

    Hi, I'm having a lot of trouble trying to replicate the failure.

    autopkgtest is unfortunately, *not* obvious. The replication
    instructions mention "<preferred autopkgtest runner config>" but
    instructions for how to do this is hard to come by.

    I had a schroot which is suitable for use by sbuild, but when I try to
    use it, it blows up with an "su: Authentication failure".

    <tytso@trampoline> {/tmp/systemd-257.6}
    1210% autopkgtest -B --test-name upstream . -- autopkgtest-virt-schroot sid-amd64-sbuild
    autopkgtest [14:52:00]: starting date and time: 2025-06-06 14:52:00-0400 autopkgtest [14:52:00]: version 5.49
    autopkgtest [14:52:00]: host trampoline; command line: /bin/autopkgtest -B --test-name upstream . -- autopkgtest-virt-schroot sid-amd64-sbuild
    autopkgtest [14:52:00]: testbed dpkg architecture: amd64
    autopkgtest [14:52:00]: testbed apt version: 2.9.30
    autopkgtest [14:52:00]: @@@@@@@@@@@@@@@@@@@@ test bed setup
    autopkgtest [14:52:00]: testbed release detected to be: None
    autopkgtest [14:52:00]: testbed running kernel: Linux 6.12.27-amd64 #1 SMP PREEMPT_DYNAMIC Debian 6.12.27-1 (2025-05-06)
    autopkgtest [14:52:00]: @@@@@@@@@@@@@@@@@@@@ unbuilt-tree .
    su: Authentication failure
    blame: .
    badpkg: rules extract failed with exit code 1
    autopkgtest [14:52:01]: ERROR: erroneous package: rules extract failed with exit code 1

    Looking at the systemd's debian/test/upstream it appears to install an
    external third party github repository, https://github.com/systemd/mkosi
    but trying to figure out what the *heck* it is doing is... hard.

    Can you perhaps give me some reproduction instructions that even a
    kernel developer can follow? Many thanks!!

    Ideally, just a simple mke2fs command so I can see what systemd-repart
    is doing. It's just not obvious....

    - Ted

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Luca Boccassi@21:1/5 to Theodore Ts'o on Fri Jun 6 23:10:01 2025
    On Fri, 6 Jun 2025 at 19:59, Theodore Ts'o <tytso@mit.edu> wrote:

    On Fri, May 30, 2025 at 03:27:56PM +0100, Luca Boccassi wrote:

    autopkgtest on debci has highlighted that the 1.47.3~rc1 version of mkfs.ext4 fails when used by systemd-repart, while it works with the version in unstable and testing:

    Hi, I'm having a lot of trouble trying to replicate the failure.

    autopkgtest is unfortunately, *not* obvious. The replication
    instructions mention "<preferred autopkgtest runner config>" but
    instructions for how to do this is hard to come by.

    I had a schroot which is suitable for use by sbuild, but when I try to
    use it, it blows up with an "su: Authentication failure".

    <tytso@trampoline> {/tmp/systemd-257.6}
    1210% autopkgtest -B --test-name upstream . -- autopkgtest-virt-schroot sid-amd64-sbuild
    autopkgtest [14:52:00]: starting date and time: 2025-06-06 14:52:00-0400 autopkgtest [14:52:00]: version 5.49
    autopkgtest [14:52:00]: host trampoline; command line: /bin/autopkgtest -B --test-name upstream . -- autopkgtest-virt-schroot sid-amd64-sbuild
    autopkgtest [14:52:00]: testbed dpkg architecture: amd64
    autopkgtest [14:52:00]: testbed apt version: 2.9.30
    autopkgtest [14:52:00]: @@@@@@@@@@@@@@@@@@@@ test bed setup
    autopkgtest [14:52:00]: testbed release detected to be: None
    autopkgtest [14:52:00]: testbed running kernel: Linux 6.12.27-amd64 #1 SMP PREEMPT_DYNAMIC Debian 6.12.27-1 (2025-05-06)
    autopkgtest [14:52:00]: @@@@@@@@@@@@@@@@@@@@ unbuilt-tree .
    su: Authentication failure
    blame: .
    badpkg: rules extract failed with exit code 1
    autopkgtest [14:52:01]: ERROR: erroneous package: rules extract failed with exit code 1

    Looking at the systemd's debian/test/upstream it appears to install an external third party github repository, https://github.com/systemd/mkosi
    but trying to figure out what the *heck* it is doing is... hard.

    Can you perhaps give me some reproduction instructions that even a
    kernel developer can follow? Many thanks!!

    Ideally, just a simple mke2fs command so I can see what systemd-repart
    is doing. It's just not obvious....

    Unfortunately I don't know how to get a manual command that shows the
    same error. I guess the content of the full debian sid root directory
    being copied in the ext4 is what triggers the issue, as a naive run
    with some local directory as sources doesn't work.

    If you are not familiar with autopkgtest, it's best to use qemu as a
    runner. You can use autopkgtest-build-qemu to build a sid image:

    autopkgtest-build-qemu sid /path/to/autopkgtest-sid.img

    And then you can run the test after placing the new locally built or
    downloaded from experimental e2fsprogs packages in /path/to/pkgs/dir/
    :

    autopkgtest --test-name upstream -B . /path/to/pkgs/dir/ -- autopkgtest-virt-qemu --cpus 4 --ram-size 8192
    /path/to/autopkgtest-sid.img

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)