xattr records not actually working? (Was: creating Hurd chroots on Linux using DPKG_ROOT chrootless mode)

[Samuel Thibault]

Hello,

Putting a more eye-catching title :)

Samuel

Johannes Schauer Marin Rodrigues, le lun. 15 juil. 2024 11:38:37 +0200, a ecrit:
> Hi,
> 
> Quoting Johannes Schauer Marin Rodrigues (2024-05-21 11:50:25)
> > In any case, things go much further now. The next problem is some missing
> > DPKG_ROOT support in the hurd maintainer script. I opened a merge request
> > here:
> > 
> > https://salsa.debian.org/hurd-team/hurd/-/merge_requests/1
> 
> thank you for uploading a new version of the hurd package including these
> DPKG_ROOT changes! I just confirmed that these indeed do work as intended and
> it is now possible to create minimal hurd tarballs containing kernel and
> sysvinit using chrootless mode on linux like this:
> 
> mmdebstrap --variant=apt \
>       
> --include=passwd,debian-ports-archive-keyring,mmdebstrap,sysvinit-core,sysv-rc
>  \
>       --customize-hook='chroot "$1" mmdebstrap \
>               --mode=chrootless --arch=hurd-i386 \
>               
> --include=sysvinit-core,sysv-rc,debian-ports-archive-keyring,gnumach-image-1-486
>  \
>               --variant=apt \
>               unstable /tmp/chroot.tar \
>               "deb http://ftp.ports.debian.org/debian-ports/ unstable main" \
>               "deb http://ftp.ports.debian.org/debian-ports/ unreleased 
> main"' \
>       --customize-hook='copy-out /tmp/chroot.tar .'\
>       unstable /dev/null
> 
> We could turn these tarballs into an ext4 file system but that would not be
> very useful yet, because all the translators would be missing from it. I was
> told by Samuel Thibault that since recently, translators are stored in 
> extended
> attributes, namely in the gnu.translator xattr. So it should be possible to
> mimic what settrans does on Hurd on Linux using setxattr(2) or, in shell,
> setfattr(1). Namely, it should be possible to special-case the st() and md()
> functions in /usr/lib/hurd/setup-translators, such that the script can be run
> on Linux.
> 
> To make sure that my settrans replacement on Linux does the right thing, I
> wanted to investigate how the extended attribute values look like on Hurd. For
> that purpose, I downloaded
> https://cdimage.debian.org/cdimage/ports/stable/hurd-i386/debian-hurd-20230608.img
> and also gnumach.gz, exec.static and ext2fs.static from the same location and
> then ran the qemu vm like this:
> 
> qemu-system-i386 -nographic -net user,hostfwd=tcp:127.0.0.1:2222-:22 -net 
> nic,model=e1000 \
>       -m 1G -kernel gnumach -append 'root=device:hd0s2 console=com0' \
>       --initrd './ext2fs.static 
> --multiboot-command-line=${kernel-command-line} --host-priv-port=${host-port} 
> --device-master-port=${device-port} --exec-server-task=${exec-task} -T typed 
> ${root} $(task-create) $(task-resume),./exec.static $(exec-task=task-create)' 
> \
>       -drive file=debian-hurd-20230608.img,format=raw
> 
> I put http://ftp.ports.debian.org/debian-ports/ unstable and unreleased main
> into the apt sources.list and upgraded everything to the latest versions. Then
> I turned the machine off, mounted the filesystem of the qemu vm and extracted
> /boot/gnumach-1.8-486-up.gz, /hurd/exec.static and /hurd/ext2fs.static. I
> gunzipped the gnumach kernel and then ran qemu again with the updated files.
> Just to be sure, I ran /usr/lib/hurd/setup-translators -K again and powered 
> off
> the machine. I then mounted the root filesystem and ran this:
> 
> sudo getfattr --dump --match=- /mnt/hurd/*
> 
> Unfortunately this came back empty. Should there not be extended attributes
> attached?
> 
> What could I be missing?
> 
> Thanks!
> 
> cheers, josch