Christoph Pleger:
> I attached a file to this email.

Thanx.
But I cannot see the NFS mount.

tmpfs /live/cow tmpfs rw,relatime,mode=755 0 0
/dev/sda /live/image iso9660 ro,noatime,nojoliet,check=s,map=n,blocksize=2048 0 0

Is this iso9660 your rr branch?
More importantly, I am afraid you didn't "mount --move" your branches
before chroot.
The aufs manual describes as
----------------------------------------
The branch filesystems must be accessible, which means `not hidden.'
It means you need to `mount \-\-move' when you use initramfs and
switch_root(8), or chroot(8).
----------------------------------------

And there was a CVS tree in aufs1 era on SorceForge, and it contans a
sample called "diskless" which is for users who wants to mount aufs in
their initramfs.
Here is just a related part in linuxrc in the sample.
----------------------------------------
        umount /branch || :
        mkdir -p aufs/branch
        #mount -t tmpfs none aufs/branch
        set -x
        for i in branch/* branch/loop/*
        do
                test -d $i || continue
                mkdir -p aufs/$i
                #/bin/mount -n --move $i aufs/$i || :
                /bin/mount -n --move $i aufs/$i 2> /dev/null || :
        done

        cd aufs
        if [ "$dl_label" = "faunos" ]
        then
                cd /tfs
        else
                /bin/mount -n --bind /tmp branch
                mount -nt proc proc /proc
                egrep -v ' (root|cram)fs ' /proc/mounts > etc/mtab
                umount /proc
                /bin/mount -n -o remount,ro aufsroot .
        fi
        /bin/mount -n --move . /
        test "$dl_label" = "knoppix" && mount -nt proc proc /proc
        #/bin/sh -i
        trap "" 0
        if [ "$dl_label" = "faunos" ]
        then
                exec chroot . sbin/init s < dev/console > dev/console 2>&1 
dev/console
        else
                exec chroot . sbin/init < dev/console > dev/console 2>&1 
dev/console
        fi
----------------------------------------


> I have attached the output of strace. Somehow I suspect the ioctl in 
> line 2510 to be the problem, but I wonder how aufs can have an effect on 
> that TTY operation.

I don't think it is related.


J. R. Okajima

Reply via email to