I''ve played around with UnionFS a bit. I like the idea for a COW file system. Anyhow, I have an initrd file that I''d like to use to set up the unionfs root. Here is the linuxrc file: == SNIP:linuxrc =# unionfs module inserted previously # these directories already exist in the initrd image BASE=/mnt/basefs UNION=/mnt/union CHANGES=/mnt/userfs echo "Mount base filesystem.." mount -t ext3 /dev/sda1 $BASE echo "Mount user filesystem.." mount -t ext3 /dev/sda2 $CHANGES echo "Mount root UnionFS.." mount -t unionfs -o dirs=$CHANGES=rw:$BASE=ro none $UNION cd $UNION mkdir -p initrd pivot_root . initrd exec chroot . sbin/init 3 < /dev/console > /dev/console 2>&1 == END = So now my issue is: init: timeout opening/writing control channel /dev/initctl (After which kernel panic ) If I exec chroot . /bin/bash instead of sbin/init... I get a shell in the current unionfs root. (As expected) Any thoughts on the sbin/init problem? I''m using a gentoo base file system -Dennis _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Dennis wrote:> I''ve played around with UnionFS a bit. I like the idea for a COW file > system.It doesn''t work properly right now for a root FS running a completely unmodified distribution; there are some bad interactions with procfs (/proc/<pid>/exe ends up pointing at the particular UnionFS branch rather than at the UnionFS filesystem itself, which confuses a number of distributions'' tools for checking daemon status), and the unionfs patch I''ve been using to override this issue via changing /proc''s symlink resolution is a bit crashy (though there are newer proposed patches I haven''t been able to test with yet). See my posts on gmane.comp.file-systems.unionfs.general and the surrounding threads. _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users