persistence luks mounting "Loading" error
authorstack <stack@inventati.org>
Tue, 19 Feb 2013 00:49:23 +0000 (01:49 +0100)
committerstack <stack@inventati.org>
Tue, 19 Feb 2013 00:49:23 +0000 (01:49 +0100)
loadkeys in load_keymap was outputting a message about Loading
/etc/boottime.kmap.gz and that was taken into account as return value
for open_luks_device(). loadkeys is now called with -q (quiet) switch.

This was a boot message example before the fix, notice the "mounting Loading":

    Begin: Running /scripts/live-realpremount ... done.
    Begin: Mounting "/live/medium/live/filesystem.squashfs" on "//filesystem.squashfs" via "/dev/loop0" ... done.
    Enter passphrase for /dev/sdb2:
    mount: mounting Loading on /live/persistence/Loading failed: No such device
    sort: /tmp/custom_mounts.list: No such file or directory
    /init: line 1: can't open /tmp/custom_mounts.list: no such file
    rm: can't remove '/tmp/custom_mounts.list': No such file or directory
    umount: can't umount /live/overlay: Device or resource busy
    done.
    Begin: Configuring fstab ... done.
    Begin: Preconfiguring networking ... done.

xfce-desktop/config/includes.chroot/lib/live/boot/9990-misc-helpers.sh [new file with mode: 0755]

diff --git a/xfce-desktop/config/includes.chroot/lib/live/boot/9990-misc-helpers.sh b/xfce-desktop/config/includes.chroot/lib/live/boot/9990-misc-helpers.sh
new file mode 100755 (executable)
index 0000000..0515f9b
--- /dev/null
@@ -0,0 +1,1600 @@
+#!/bin/sh
+
+#set -e
+
+is_live_path ()
+{
+       DIRECTORY="${1}"
+
+       if [ -d "${DIRECTORY}"/"${LIVE_MEDIA_PATH}" ]
+       then
+               for FILESYSTEM in squashfs ext2 ext3 ext4 xfs dir jffs2
+               do
+                       if [ "$(echo ${DIRECTORY}/${LIVE_MEDIA_PATH}/*.${FILESYSTEM})" != "${DIRECTORY}/${LIVE_MEDIA_PATH}/*.${FILESYSTEM}" ]
+                       then
+                               return 0
+                       fi
+               done
+       fi
+
+       return 1
+}
+
+matches_uuid ()
+{
+       if [ "${IGNORE_UUID}" ] || [ ! -e /conf/uuid.conf ]
+       then
+               return 0
+       fi
+
+       path="${1}"
+       uuid="$(cat /conf/uuid.conf)"
+
+       for try_uuid_file in "${path}/.disk/live-uuid"*
+       do
+               [ -e "${try_uuid_file}" ] || continue
+
+               try_uuid="$(cat "${try_uuid_file}")"
+
+               if [ "${uuid}" = "${try_uuid}" ]
+               then
+                       return 0
+               fi
+       done
+
+       return 1
+}
+
+get_backing_device ()
+{
+       case "${1}" in
+               *.squashfs|*.ext2|*.ext3|*.ext4|*.jffs2)
+                       echo $(setup_loop "${1}" "loop" "/sys/block/loop*" '0' "${LIVE_MEDIA_ENCRYPTION}" "${2}")
+                       ;;
+
+               *.dir)
+                       echo "directory"
+                       ;;
+
+               *)
+                       panic "Unrecognized live filesystem: ${1}"
+                       ;;
+       esac
+}
+
+match_files_in_dir ()
+{
+       # Does any files match pattern ${1} ?
+       local pattern
+       pattern="${1}"
+
+       if [ "$(echo ${pattern})" != "${pattern}" ]
+       then
+               return 0
+       fi
+
+       return 1
+}
+
+mount_images_in_directory ()
+{
+       directory="${1}"
+       rootmnt="${2}"
+       mac="${3}"
+
+       if match_files_in_dir "${directory}/${LIVE_MEDIA_PATH}/*.squashfs" ||
+               match_files_in_dir "${directory}/${LIVE_MEDIA_PATH}/*.ext2" ||
+               match_files_in_dir "${directory}/${LIVE_MEDIA_PATH}/*.ext3" ||
+               match_files_in_dir "${directory}/${LIVE_MEDIA_PATH}/*.ext4" ||
+               match_files_in_dir "${directory}/${LIVE_MEDIA_PATH}/*.jffs2" ||
+               match_files_in_dir "${directory}/${LIVE_MEDIA_PATH}/*.dir"
+       then
+               [ -n "${mac}" ] && adddirectory="${directory}/${LIVE_MEDIA_PATH}/${mac}"
+               setup_unionfs "${directory}/${LIVE_MEDIA_PATH}" "${rootmnt}" "${adddirectory}"
+       else
+               panic "No supported filesystem images found at /${LIVE_MEDIA_PATH}."
+       fi
+}
+
+is_nice_device ()
+{
+       sysfs_path="${1#/sys}"
+
+       if [ -e /lib/udev/path_id ]
+       then
+               # squeeze
+               PATH_ID="/lib/udev/path_id"
+       else
+               # wheezy/sid (udev >= 174)
+               PATH_ID="/sbin/udevadm test-builtin path_id"
+       fi
+
+       if ${PATH_ID} "${sysfs_path}" | egrep -q "ID_PATH=(usb|pci-[^-]*-(ide|sas|scsi|usb|virtio)|platform-sata_mv|platform-orion-ehci|platform-mmc|platform-mxsdhci)"
+       then
+               return 0
+       elif echo "${sysfs_path}" | grep -q '^/block/vd[a-z]$'
+       then
+               return 0
+       elif echo ${sysfs_path} | grep -q "^/block/dm-"
+       then
+               return 0
+       elif echo ${sysfs_path} | grep -q "^/block/mtdblock"
+       then
+               return 0
+       fi
+
+       return 1
+}
+
+check_dev ()
+{
+       sysdev="${1}"
+       devname="${2}"
+       skip_uuid_check="${3}"
+
+       # support for fromiso=.../isofrom=....
+       if [ -n "$FROMISO" ]
+       then
+               ISO_DEVICE=$(dirname $FROMISO)
+               if ! [ -b $ISO_DEVICE ]
+               then
+                       # to support unusual device names like /dev/cciss/c0d0p1
+                       # as well we have to identify the block device name, let's
+                       # do that for up to 15 levels
+                       i=15
+                       while [ -n "$ISO_DEVICE" ] && [ "$i" -gt 0 ]
+                       do
+                               ISO_DEVICE=$(dirname ${ISO_DEVICE})
+                               [ -b "$ISO_DEVICE" ] && break
+                               i=$(($i -1))
+                       done
+               fi
+
+               if [ "$ISO_DEVICE" = "/" ]
+               then
+                       echo "Warning: device for bootoption fromiso= ($FROMISO) not found.">>/boot.log
+               else
+                       fs_type=$(get_fstype "${ISO_DEVICE}")
+                       if is_supported_fs ${fs_type}
+                       then
+                               mkdir /live/fromiso
+                               mount -t $fs_type "$ISO_DEVICE" /live/fromiso
+                               ISO_NAME="$(echo $FROMISO | sed "s|$ISO_DEVICE||")"
+                               loopdevname=$(setup_loop "/live/fromiso/${ISO_NAME}" "loop" "/sys/block/loop*" "" '')
+                               devname="${loopdevname}"
+                       else
+                               echo "Warning: unable to mount $ISO_DEVICE." >>/boot.log
+                       fi
+               fi
+       fi
+
+       if [ -z "${devname}" ]
+       then
+               devname=$(sys2dev "${sysdev}")
+       fi
+
+       if [ -d "${devname}" ]
+       then
+               mount -o bind "${devname}" $mountpoint || continue
+
+               if is_live_path $mountpoint
+               then
+                       echo $mountpoint
+                       return 0
+               else
+                       umount $mountpoint
+               fi
+       fi
+
+       IFS=","
+       for device in ${devname}
+       do
+               case "$device" in
+                       *mapper*)
+                               # Adding lvm support
+                               if [ -x /scripts/local-top/lvm2 ]
+                               then
+                                       ROOT="$device" resume="" /scripts/local-top/lvm2
+                               fi
+                               ;;
+
+                       /dev/md*)
+                               # Adding raid support
+                               if [ -x /scripts/local-top/mdadm ]
+                               then
+                                       cp /conf/conf.d/md /conf/conf.d/md.orig
+                                       echo "MD_DEVS=$device " >> /conf/conf.d/md
+                                       /scripts/local-top/mdadm
+                                       mv /conf/conf.d/md.orig /conf/conf.d/md
+                               fi
+                               ;;
+               esac
+       done
+       unset IFS
+
+       [ -n "$device" ] && devname="$device"
+
+       [ -e "$devname" ] || continue
+
+       if [ -n "${LIVE_MEDIA_OFFSET}" ]
+       then
+               loopdevname=$(setup_loop "${devname}" "loop" "/sys/block/loop*" "${LIVE_MEDIA_OFFSET}" '')
+               devname="${loopdevname}"
+       fi
+
+       fstype=$(get_fstype "${devname}")
+
+       if is_supported_fs ${fstype}
+       then
+               devuid=$(blkid -o value -s UUID "$devname")
+               [ -n "$devuid" ] && grep -qs "\<$devuid\>" $tried && continue
+               mount -t ${fstype} -o ro,noatime "${devname}" ${mountpoint} || continue
+               [ -n "$devuid" ] && echo "$devuid" >> $tried
+
+               if [ -n "${FINDISO}" ]
+               then
+                       if [ -f ${mountpoint}/${FINDISO} ]
+                       then
+                               umount ${mountpoint}
+                               mkdir -p /live/findiso
+                               mount -t ${fstype} -o ro,noatime "${devname}" /live/findiso
+                               loopdevname=$(setup_loop "/live/findiso/${FINDISO}" "loop" "/sys/block/loop*" 0 "")
+                               devname="${loopdevname}"
+                               mount -t iso9660 -o ro,noatime "${devname}" ${mountpoint}
+                       else
+                               umount ${mountpoint}
+                       fi
+               fi
+
+               if is_live_path ${mountpoint} && \
+                       ([ "${skip_uuid_check}" ] || matches_uuid ${mountpoint})
+               then
+                       echo ${mountpoint}
+                       return 0
+               else
+                       umount ${mountpoint} 2>/dev/null
+               fi
+       fi
+
+       if [ -n "${LIVE_MEDIA_OFFSET}" ]
+       then
+               losetup -d "${loopdevname}"
+       fi
+
+       return 1
+}
+
+find_livefs ()
+{
+       timeout="${1}"
+
+       # don't start autodetection before timeout has expired
+       if [ -n "${LIVE_MEDIA_TIMEOUT}" ]
+       then
+               if [ "${timeout}" -lt "${LIVE_MEDIA_TIMEOUT}" ]
+               then
+                       return 1
+               fi
+       fi
+
+       # first look at the one specified in the command line
+       case "${LIVE_MEDIA}" in
+               removable-usb)
+                       for sysblock in $(removable_usb_dev "sys")
+                       do
+                               for dev in $(subdevices "${sysblock}")
+                               do
+                                       if check_dev "${dev}"
+                                       then
+                                               return 0
+                                       fi
+                               done
+                       done
+                       return 1
+                       ;;
+
+               removable)
+                       for sysblock in $(removable_dev "sys")
+                       do
+                               for dev in $(subdevices "${sysblock}")
+                               do
+                                       if check_dev "${dev}"
+                                       then
+                                               return 0
+                                       fi
+                               done
+                       done
+                       return 1
+                       ;;
+
+               *)
+                       if [ ! -z "${LIVE_MEDIA}" ]
+                       then
+                               if check_dev "null" "${LIVE_MEDIA}" "skip_uuid_check"
+                               then
+                                       return 0
+                               fi
+                       fi
+                       ;;
+       esac
+
+       # or do the scan of block devices
+       # prefer removable devices over non-removable devices, so scan them first
+       devices_to_scan="$(removable_dev 'sys') $(non_removable_dev 'sys')"
+
+       for sysblock in $devices_to_scan
+       do
+               devname=$(sys2dev "${sysblock}")
+               [ -e "$devname" ] || continue
+               fstype=$(get_fstype "${devname}")
+
+               if /lib/udev/cdrom_id ${devname} > /dev/null
+               then
+                       if check_dev "null" "${devname}"
+                       then
+                               return 0
+                       fi
+               elif is_nice_device "${sysblock}"
+               then
+                       for dev in $(subdevices "${sysblock}")
+                       do
+                               if check_dev "${dev}"
+                               then
+                                       return 0
+                               fi
+                       done
+               elif [ "${fstype}" = "squashfs" -o \
+                       "${fstype}" = "btrfs" -o \
+                       "${fstype}" = "ext2" -o \
+                       "${fstype}" = "ext3" -o \
+                       "${fstype}" = "ext4" -o \
+                       "${fstype}" = "jffs2" ]
+               then
+                       # This is an ugly hack situation, the block device has
+                       # an image directly on it.  It's hopefully
+                       # live-boot, so take it and run with it.
+                       ln -s "${devname}" "${devname}.${fstype}"
+                       echo "${devname}.${fstype}"
+                       return 0
+               fi
+       done
+
+       return 1
+}
+
+really_export ()
+{
+       STRING="${1}"
+       VALUE="$(eval echo -n \${$STRING})"
+
+       if [ -f /live.vars ] && grep -sq "export ${STRING}" /live.vars
+       then
+               sed -i -e 's/\('${STRING}'=\).*$/\1'${VALUE}'/' /live.vars
+       else
+               echo "export ${STRING}=\"${VALUE}\"" >> /live.vars
+       fi
+
+       eval export "${STRING}"="${VALUE}"
+}
+
+is_in_list_separator_helper ()
+{
+       local sep element list
+       sep=${1}
+       shift
+       element=${1}
+       shift
+       list=${*}
+       echo ${list} | grep -qe "^\(.*${sep}\)\?${element}\(${sep}.*\)\?$"
+}
+
+is_in_space_sep_list ()
+{
+       local element
+       element=${1}
+       shift
+       is_in_list_separator_helper "[[:space:]]" "${element}" "${*}"
+}
+
+is_in_comma_sep_list ()
+{
+       local element
+       element=${1}
+       shift
+       is_in_list_separator_helper "," "${element}" "${*}"
+}
+
+sys2dev ()
+{
+       sysdev=${1#/sys}
+       echo "/dev/$($udevinfo -q name -p ${sysdev} 2>/dev/null|| echo ${sysdev##*/})"
+}
+
+subdevices ()
+{
+       sysblock=${1}
+       r=""
+
+       for dev in "${sysblock}"/* "${sysblock}"
+       do
+               if [ -e "${dev}/dev" ]
+               then
+                       r="${r} ${dev}"
+               fi
+       done
+
+       echo ${r}
+}
+
+storage_devices()
+{
+       black_listed_devices="${1}"
+       white_listed_devices="${2}"
+
+       for sysblock in $(echo /sys/block/* | tr ' ' '\n' | grep -vE "loop|ram|fd")
+       do
+               fulldevname=$(sys2dev "${sysblock}")
+
+               if is_in_space_sep_list ${fulldevname} ${black_listed_devices} || \
+                       [ -n "${white_listed_devices}" ] && \
+                       ! is_in_space_sep_list ${fulldevname} ${white_listed_devices}
+               then
+                       # skip this device entirely
+                       continue
+               fi
+
+               for dev in $(subdevices "${sysblock}")
+               do
+                       devname=$(sys2dev "${dev}")
+
+                       if is_in_space_sep_list ${devname} ${black_listed_devices}
+                       then
+                               # skip this subdevice
+                               continue
+                       else
+                               echo "${devname}"
+                       fi
+               done
+       done
+}
+
+is_supported_fs ()
+{
+       fstype="${1}"
+
+       # Validate input first
+       if [ -z "${fstype}" ]
+       then
+               return 1
+       fi
+
+       # Try to look if it is already supported by the kernel
+       if grep -q ${fstype} /proc/filesystems
+       then
+               return 0
+       else
+               # Then try to add support for it the gentle way using the initramfs capabilities
+               modprobe ${fstype}
+               if grep -q ${fstype} /proc/filesystems
+               then
+                       return 0
+               # Then try the hard way if /root is already reachable
+               else
+                       kmodule="/root/lib/modules/`uname -r`/${fstype}/${fstype}.ko"
+                       if [ -e "${kmodule}" ]
+                       then
+                               insmod "${kmodule}"
+                               if grep -q ${fstype} /proc/filesystems
+                               then
+                                       return 0
+                               fi
+                       fi
+               fi
+       fi
+
+       return 1
+}
+
+get_fstype ()
+{
+       /sbin/blkid -s TYPE -o value $1 2>/dev/null
+}
+
+where_is_mounted ()
+{
+       device=${1}
+       # return first found
+       grep -m1 "^${device} " /proc/mounts | cut -f2 -d ' '
+}
+
+trim_path ()
+{
+       # remove all unnecessary /:s in the path, including last one (except
+       # if path is just "/")
+       echo ${1} | sed 's|//\+|/|g' | sed 's|^\(.*[^/]\)/$|\1|'
+}
+
+what_is_mounted_on ()
+{
+       local dir
+       dir="$(trim_path ${1})"
+       grep -m1 "^[^ ]\+ ${dir} " /proc/mounts | cut -d' ' -f1
+}
+
+chown_ref ()
+{
+       local reference targets owner
+       reference="${1}"
+       shift
+       targets=${@}
+       owner=$(stat -c %u:%g "${reference}")
+       chown -h ${owner} ${targets}
+}
+
+chmod_ref ()
+{
+       local reference targets rights
+       reference="${1}"
+       shift
+       targets=${@}
+       rights=$(stat -c %a "${reference}")
+       chmod ${rights} ${targets}
+}
+
+lastline ()
+{
+       while read lines
+       do
+               line=${lines}
+       done
+
+       echo "${line}"
+}
+
+base_path ()
+{
+       testpath="${1}"
+       mounts="$(awk '{print $2}' /proc/mounts)"
+       testpath="$(busybox realpath ${testpath})"
+
+       while true
+       do
+               if echo "${mounts}" | grep -qs "^${testpath}"
+               then
+                       set -- $(echo "${mounts}" | grep "^${testpath}" | lastline)
+                       echo ${1}
+                       break
+               else
+                       testpath=$(dirname $testpath)
+               fi
+       done
+}
+
+fs_size ()
+{
+       # Returns used/free fs kbytes + 5% more
+       # You could pass a block device as ${1} or the mount point as ${2}
+
+       dev="${1}"
+       mountp="${2}"
+       used="${3}"
+
+       if [ -z "${mountp}" ]
+       then
+               mountp="$(where_is_mounted ${dev})"
+
+               if [ -z "${mountp}" ]
+               then
+                       mountp="/mnt/tmp_fs_size"
+
+                       mkdir -p "${mountp}"
+                       mount -t $(get_fstype "${dev}") -o ro "${dev}" "${mountp}" || log_warning_msg "cannot mount -t $(get_fstype ${dev}) -o ro ${dev} ${mountp}"
+
+                       doumount=1
+               fi
+       fi
+
+       if [ "${used}" = "used" ]
+       then
+               size=$(du -ks ${mountp} | cut -f1)
+               size=$(expr ${size} + ${size} / 20 ) # FIXME: 5% more to be sure
+       else
+               # free space
+               size="$(df -kP | grep -s ${mountp} | awk '{print $4}')"
+       fi
+
+       if [ -n "${doumount}" ]
+       then
+               umount "${mountp}" || log_warning_msg "cannot umount ${mountp}"
+               rmdir "${mountp}"
+       fi
+
+       echo "${size}"
+}
+
+load_keymap ()
+{
+       # Load custom keymap
+       if [ -x /bin/loadkeys -a -r /etc/boottime.kmap.gz ]
+       then
+               loadkeys -q /etc/boottime.kmap.gz
+       fi
+}
+
+setup_loop ()
+{
+       local fspath module pattern offset encryption readonly
+       fspath=${1}
+       module=${2}
+       pattern=${3}
+       offset=${4}
+       encryption=${5}
+       readonly=${6}
+
+       # the output of setup_loop is evaluated in other functions,
+       # modprobe leaks kernel options like "libata.dma=0"
+       # as "options libata dma=0" on stdout, causing serious
+       # problems therefor, so instead always avoid output to stdout
+       modprobe -q -b "${module}" 1>/dev/null
+
+       udevadm settle
+
+       for loopdev in ${pattern}
+       do
+               if [ "$(cat ${loopdev}/size)" -eq 0 ]
+               then
+                       dev=$(sys2dev "${loopdev}")
+                       options=''
+
+                       if [ -n "${readonly}" ]
+                       then
+                               if losetup --help 2>&1 | grep -q -- "-r\b"
+                               then
+                                       options="${options} -r"
+                               fi
+                       fi
+
+                       if [ -n "${offset}" ] && [ 0 -lt "${offset}" ]
+                       then
+                               options="${options} -o ${offset}"
+                       fi
+
+                       if [ -z "${encryption}" ]
+                       then
+                               losetup ${options} "${dev}" "${fspath}"
+                       else
+                               # Loop AES encryption
+                               while true
+                               do
+                                       load_keymap
+
+                                       echo -n "Enter passphrase for root filesystem: " >&6
+                                       read -s passphrase
+                                       echo "${passphrase}" > /tmp/passphrase
+                                       unset passphrase
+                                       exec 9</tmp/passphrase
+                                       /sbin/losetup ${options} -e "${encryption}" -p 9 "${dev}" "${fspath}"
+                                       error=${?}
+                                       exec 9<&-
+                                       rm -f /tmp/passphrase
+
+                                       if [ 0 -eq ${error} ]
+                                       then
+                                               unset error
+                                               break
+                                       fi
+
+                                       echo
+                                       echo -n "There was an error decrypting the root filesystem ... Retry? [Y/n] " >&6
+                                       read answer
+
+                                       if [ "$(echo "${answer}" | cut -b1 | tr A-Z a-z)" = "n" ]
+                                       then
+                                               unset answer
+                                               break
+                                       fi
+                               done
+                       fi
+
+                       echo "${dev}"
+                       return 0
+               fi
+       done
+
+       panic "No loop devices available"
+}
+
+try_mount ()
+{
+       dev="${1}"
+       mountp="${2}"
+       opts="${3}"
+       fstype="${4}"
+
+       old_mountp="$(where_is_mounted ${dev})"
+
+       if [ -n "${old_mountp}" ]
+       then
+               if [ "${opts}" != "ro" ]
+               then
+                       mount -o remount,"${opts}" "${dev}" "${old_mountp}" || panic "Remounting ${dev} ${opts} on ${old_mountp} failed"
+               fi
+
+               mount -o bind "${old_mountp}" "${mountp}" || panic "Cannot bind-mount ${old_mountp} on ${mountp}"
+       else
+               if [ -z "${fstype}" ]
+               then
+                       fstype=$(get_fstype "${dev}")
+               fi
+               mount -t "${fstype}" -o "${opts}" "${dev}" "${mountp}" || \
+               ( echo "SKIPPING: Cannot mount ${dev} on ${mountp}, fstype=${fstype}, options=${opts}" > boot.log && return 0 )
+       fi
+}
+
+# Try to mount $device to the place expected by live-boot. If $device
+# is already mounted somewhere, move it to the expected place. If
+# we're only probing $device (to check if it has custom persistence)
+# $probe should be set, which suppresses warnings upon failure. On
+# success, print the mount point for $device.
+mount_persistence_media ()
+{
+       local device probe backing old_backing fstype mount_opts
+       device=${1}
+       probe=${2}
+
+       backing="/live/persistence/$(basename ${device})"
+
+       mkdir -p "${backing}"
+       old_backing="$(where_is_mounted ${device})"
+       if [ -z "${old_backing}" ]
+       then
+               fstype="$(get_fstype ${device})"
+               mount_opts="rw,noatime"
+               if [ -n "${PERSISTENCE_READONLY}" ]
+               then
+                       mount_opts="ro,noatime"
+               fi
+               if mount -t "${fstype}" -o "${mount_opts}" "${device}" "${backing}" >/dev/null
+               then
+                       echo ${backing}
+                       return 0
+               else
+                       [ -z "${probe}" ] && log_warning_msg "Failed to mount persistence media ${device}"
+                       rmdir "${backing}"
+                       return 1
+               fi
+       elif [ "${backing}" != "${old_backing}" ]
+       then
+               if mount --move ${old_backing} ${backing} >/dev/null
+               then
+                       echo ${backing}
+                       return 0
+               else
+                       [ -z "${probe}" ] && log_warning_msg "Failed to move persistence media ${device}"
+                       rmdir "${backing}"
+                       return 1
+               fi
+       else
+               # This means that $device has already been mounted on
+               # the place expected by live-boot, so we're done.
+               echo ${backing}
+               return 0
+       fi
+}
+
+close_persistence_media ()
+{
+       local device backing
+       device=${1}
+       backing="$(where_is_mounted ${device})"
+
+       if [ -d "${backing}" ]
+       then
+               umount "${backing}" >/dev/null 2>&1
+               rmdir "${backing}" >/dev/null 2>&1
+       fi
+
+       if is_active_luks_mapping ${device}
+       then
+               /sbin/cryptsetup luksClose ${device}
+       fi
+}
+
+open_luks_device ()
+{
+       dev="${1}"
+       name="$(basename ${dev})"
+       opts="--key-file=-"
+       if [ -n "${PERSISTENCE_READONLY}" ]
+       then
+               opts="${opts} --readonly"
+       fi
+
+       if /sbin/cryptsetup status "${name}" >/dev/null 2>&1
+       then
+               re="^[[:space:]]*device:[[:space:]]*\([^[:space:]]*\)$"
+               opened_dev=$(cryptsetup status ${name} 2>/dev/null | grep "${re}" | sed "s|${re}|\1|")
+               if [ "${opened_dev}" = "${dev}" ]
+               then
+                       luks_device="/dev/mapper/${name}"
+                       echo ${luks_device}
+                       return 0
+               else
+                       log_warning_msg "Cannot open luks device ${dev} since ${opened_dev} already is opened with its name"
+                       return 1
+               fi
+       fi
+
+       load_keymap
+
+       while true
+       do
+               /lib/cryptsetup/askpass "Enter passphrase for ${dev}: " | \
+                       /sbin/cryptsetup -T 1 luksOpen ${dev} ${name} ${opts}
+
+               if [ 0 -eq ${?} ]
+               then
+                       luks_device="/dev/mapper/${name}"
+                       echo ${luks_device}
+                       return 0
+               fi
+
+               echo >&6
+               echo -n "There was an error decrypting ${dev} ... Retry? [Y/n] " >&6
+               read answer
+
+               if [ "$(echo "${answer}" | cut -b1 | tr A-Z a-z)" = "n" ]
+               then
+                       return 2
+               fi
+       done
+}
+
+get_gpt_name ()
+{
+    local dev
+    dev="${1}"
+    /sbin/blkid -s PART_ENTRY_NAME -p -o value ${dev} 2>/dev/null
+}
+
+is_gpt_device ()
+{
+    local dev
+    dev="${1}"
+    [ "$(/sbin/blkid -s PART_ENTRY_SCHEME -p -o value ${dev} 2>/dev/null)" = "gpt" ]
+}
+
+probe_for_gpt_name ()
+{
+       local overlays dev gpt_dev gpt_name
+       overlays="${1}"
+       dev="${2}"
+
+       gpt_dev="${dev}"
+       if is_active_luks_mapping ${dev}
+       then
+               # if $dev is an opened luks device, we need to check
+               # GPT stuff on the backing device
+               gpt_dev=$(get_luks_backing_device "${dev}")
+       fi
+
+       if ! is_gpt_device ${gpt_dev}
+       then
+               return
+       fi
+
+       gpt_name=$(get_gpt_name ${gpt_dev})
+       for label in ${overlays}
+       do
+               if [ "${gpt_name}" = "${label}" ]
+               then
+                       echo "${label}=${dev}"
+               fi
+       done
+}
+
+probe_for_fs_label ()
+{
+       local overlays dev
+       overlays="${1}"
+       dev="${2}"
+
+       for label in ${overlays}
+       do
+               if [ "$(/sbin/blkid -s LABEL -o value $dev 2>/dev/null)" = "${label}" ]
+               then
+                       echo "${label}=${dev}"
+               fi
+       done
+}
+
+probe_for_file_name ()
+{
+       local overlays dev ret backing
+       overlays="${1}"
+       dev="${2}"
+
+       ret=""
+       backing="$(mount_persistence_media ${dev} probe)"
+       if [ -z "${backing}" ]
+       then
+           return
+       fi
+
+       for label in ${overlays}
+       do
+               path=${backing}/${PERSISTENCE_PATH}${label}
+               if [ -f "${path}" ]
+               then
+                       local loopdev
+                       loopdev=$(setup_loop "${path}" "loop" "/sys/block/loop*")
+                       ret="${ret} ${label}=${loopdev}"
+               fi
+       done
+
+       if [ -n "${ret}" ]
+       then
+               echo ${ret}
+       else
+               # unmount and remove mountpoint
+               umount ${backing} > /dev/null 2>&1 || true
+               rmdir ${backing} > /dev/null 2>&1 || true
+       fi
+}
+
+find_persistence_media ()
+{
+       # Scans devices for overlays, and returns a whitespace
+       # separated list of how to use them. Only overlays with a partition
+       # label or file name in ${overlays} are returned.
+       #
+       # When scanning a LUKS device, the user will be asked to enter the
+       # passphrase; on failure to enter it, or if no persistence partitions
+       # or files were found, the LUKS device is closed.
+       #
+       # For all other cases (overlay partition and overlay file) the
+       # return value is "${label}=${device}", where ${device} a device that
+       # can mount the content. In the case of an overlay file, the device
+       # containing the file will remain mounted as a side-effect.
+       #
+       # No devices in ${black_listed_devices} will be scanned, and if
+       # ${white_list_devices} is non-empty, only devices in it will be
+       # scanned.
+
+       local overlays white_listed_devices ret black_listed_devices
+       overlays="${1}"
+       white_listed_devices="${2}"
+       ret=""
+
+       black_listed_devices="$(what_is_mounted_on /live/medium)"
+
+       for dev in $(storage_devices "${black_listed_devices}" "${white_listed_devices}")
+       do
+               local result luks_device
+               result=""
+
+               luks_device=""
+               # Check if it's a luks device; we'll have to open the device
+               # in order to probe any filesystem it contains, like we do
+               # below. activate_custom_mounts() also depends on that any luks
+               # device already has been opened.
+               if is_in_comma_sep_list luks ${PERSISTENCE_ENCRYPTION} && is_luks_partition ${dev}
+               then
+                       if luks_device=$(open_luks_device "${dev}")
+                       then
+                               dev="${luks_device}"
+                       else
+                               # skip $dev since we failed/chose not to open it
+                               continue
+                       fi
+               elif ! is_in_comma_sep_list none ${PERSISTENCE_ENCRYPTION}
+               then
+                       # skip $dev since we don't allow unencrypted storage
+                       continue
+               fi
+
+               # Probe for matching GPT partition names or filesystem labels
+               if is_in_comma_sep_list filesystem ${PERSISTENCE_STORAGE}
+               then
+                       result=$(probe_for_gpt_name "${overlays}" ${dev})
+                       if [ -n "${result}" ]
+                       then
+                               ret="${ret} ${result}"
+                               continue
+                       fi
+
+                       result=$(probe_for_fs_label "${overlays}" ${dev})
+                       if [ -n "${result}" ]
+                       then
+                               ret="${ret} ${result}"
+                               continue
+                       fi
+               fi
+
+               # Probe for files with matching name on mounted partition
+               if is_in_comma_sep_list file ${PERSISTENCE_STORAGE}
+               then
+                       result=$(probe_for_file_name "${overlays}" ${dev})
+                       if [ -n "${result}" ]
+                       then
+                               ret="${ret} ${result}"
+                               continue
+                       fi
+               fi
+
+               # Close luks device if it isn't used
+               if [ -z "${result}" ] && [ -n "${luks_device}" ] && is_active_luks_mapping "${luks_device}"
+               then
+                       /sbin/cryptsetup luksClose "${luks_device}"
+               fi
+       done
+
+       if [ -n "${ret}" ]
+       then
+               echo ${ret}
+       fi
+}
+
+get_mac ()
+{
+       mac=""
+
+       for adaptor in /sys/class/net/*
+       do
+               status="$(cat ${adaptor}/iflink)"
+
+               if [ "${status}" -eq 2 ]
+               then
+                       mac="$(cat ${adaptor}/address)"
+                       mac="$(echo ${mac} | sed 's/:/-/g' | tr '[a-z]' '[A-Z]')"
+               fi
+       done
+
+       echo ${mac}
+}
+
+is_luks_partition ()
+{
+       device="${1}"
+       /sbin/cryptsetup isLuks "${device}" 1>/dev/null 2>&1
+}
+
+is_active_luks_mapping ()
+{
+       device="${1}"
+       /sbin/cryptsetup status "${device}" 1>/dev/null 2>&1
+}
+
+get_luks_backing_device ()
+{
+       device=${1}
+       cryptsetup status ${device} 2> /dev/null | \
+               awk '{if ($1 == "device:") print $2}'
+}
+
+removable_dev ()
+{
+       output_format="${1}"
+       want_usb="${2}"
+       ret=
+
+       for sysblock in $(echo /sys/block/* | tr ' ' '\n' | grep -vE "/(loop|ram|dm-|fd)")
+       do
+               dev_ok=
+               if [ "$(cat ${sysblock}/removable)" = "1" ]
+               then
+                       if [ -z "${want_usb}" ]
+                       then
+                               dev_ok="true"
+                       else
+                               if readlink ${sysblock} | grep -q usb
+                               then
+                                       dev_ok="true"
+                               fi
+                       fi
+               fi
+
+               if [ "${dev_ok}" = "true" ]
+               then
+                       case "${output_format}" in
+                               sys)
+                                       ret="${ret} ${sysblock}"
+                                       ;;
+                               *)
+                                       devname=$(sys2dev "${sysblock}")
+                                       ret="${ret} ${devname}"
+                                       ;;
+                       esac
+               fi
+       done
+
+       echo "${ret}"
+}
+
+removable_usb_dev ()
+{
+       output_format="${1}"
+
+       removable_dev "${output_format}" "want_usb"
+}
+
+non_removable_dev ()
+{
+       output_format="${1}"
+       ret=
+
+       for sysblock in $(echo /sys/block/* | tr ' ' '\n' | grep -vE "/(loop|ram|dm-|fd)")
+       do
+               if [ "$(cat ${sysblock}/removable)" = "0" ]
+               then
+                       case "${output_format}" in
+                               sys)
+                                       ret="${ret} ${sysblock}"
+                                       ;;
+                               *)
+                                       devname=$(sys2dev "${sysblock}")
+                                       ret="${ret} ${devname}"
+                                       ;;
+                       esac
+               fi
+       done
+
+       echo "${ret}"
+}
+
+link_files ()
+{
+       # create source's directory structure in dest, and recursively
+       # create symlinks in dest to to all files in source. if mask
+       # is non-empty, remove mask from all source paths when
+       # creating links (will be necessary if we change root, which
+       # live-boot normally does (into $rootmnt)).
+       local src_dir dest_dir src_mask
+
+       # remove multiple /:s and ensure ending on /
+       src_dir="$(trim_path ${1})/"
+       dest_dir="$(trim_path ${2})/"
+       src_mask="${3}"
+
+       # This check can only trigger on the inital, non-recursive call since
+       # we create the destination before recursive calls
+       if [ ! -d "${dest_dir}" ]
+       then
+               log_warning_msg "Must link_files into a directory"
+               return
+       fi
+
+       find "${src_dir}" -mindepth 1 -maxdepth 1 | \
+       while read src
+       do
+               local dest final_src
+               dest="${dest_dir}$(basename "${src}")"
+               if [ -d "${src}" ]
+               then
+                       if [ -z "$(ls -A "${src}")" ]
+                       then
+                               continue
+                       fi
+                       if [ ! -d "${dest}" ]
+                       then
+                               mkdir -p "${dest}"
+                               chown_ref "${src}" "${dest}"
+                               chmod_ref "${src}" "${dest}"
+                       fi
+                       link_files "${src}" "${dest}" "${src_mask}"
+               else
+                       final_src=${src}
+                       if [ -n "${src_mask}" ]
+                       then
+                               final_src="$(echo ${final_src} | sed "s|^${src_mask}||")"
+                       fi
+                       rm -rf "${dest}" 2> /dev/null
+                       ln -s "${final_src}" "${dest}"
+                       chown_ref "${src}" "${dest}"
+               fi
+       done
+}
+
+do_union ()
+{
+       local unionmountpoint unionrw unionro
+       unionmountpoint="${1}"  # directory where the union is mounted
+       shift
+       unionrw="${1}"          # branch where the union changes are stored
+       shift
+       unionro="${*}"          # space separated list of read-only branches (optional)
+
+       case "${UNIONTYPE}" in
+               aufs)
+                       rw_opt="rw"
+                       ro_opt="rr+wh"
+                       noxino_opt="noxino"
+                       ;;
+
+               unionfs-fuse)
+                       rw_opt="RW"
+                       ro_opt="RO"
+                       ;;
+
+               *)
+                       rw_opt="rw"
+                       ro_opt="ro"
+                       ;;
+       esac
+
+       case "${UNIONTYPE}" in
+               unionfs-fuse)
+                       unionmountopts="-o cow -o noinitgroups -o default_permissions -o allow_other -o use_ino -o suid"
+                       unionmountopts="${unionmountopts} ${unionrw}=${rw_opt}"
+                       if [ -n "${unionro}" ]
+                       then
+                               for rofs in ${unionro}
+                               do
+                                       unionmountopts="${unionmountopts}:${rofs}=${ro_opt}"
+                               done
+                       fi
+                       ( sysctl -w fs.file-max=391524 ; ulimit -HSn 16384
+                       unionfs-fuse ${unionmountopts} "${unionmountpoint}" ) && \
+                       ( mkdir -p /run/sendsigs.omit.d
+                       pidof unionfs-fuse >> /run/sendsigs.omit.d/unionfs-fuse || true )
+                       ;;
+
+               overlayfs)
+                       # XXX: can multiple unionro be used? (overlayfs only handles two dirs, but perhaps they can be chained?)
+                       # XXX: and can unionro be optional? i.e. can overlayfs skip lowerdir?
+                       if echo ${unionro} | grep -q " "
+                       then
+                               panic "Multiple lower filesystems are currently not supported with overlayfs (unionro = ${unionro})."
+                       elif [ -z "${unionro}"  ]
+                       then
+                               panic "Overlayfs needs at least one lower filesystem (read-only branch)."
+                       fi
+                       unionmountopts="-o noatime,lowerdir=${unionro},upperdir=${unionrw}"
+                       mount -t ${UNIONTYPE} ${unionmountopts} ${UNIONTYPE} "${unionmountpoint}"
+                       ;;
+
+               *)
+                       unionmountopts="-o noatime,${noxino_opt},dirs=${unionrw}=${rw_opt}"
+                       if [ -n "${unionro}" ]
+                       then
+                               for rofs in ${unionro}
+                               do
+                                       unionmountopts="${unionmountopts}:${rofs}=${ro_opt}"
+                               done
+                       fi
+                       mount -t ${UNIONTYPE} ${unionmountopts} ${UNIONTYPE} "${unionmountpoint}"
+                       ;;
+       esac
+}
+
+get_custom_mounts ()
+{
+       # Side-effect: leaves $devices with persistence.conf mounted in /live/persistence
+       # Side-effect: prints info to file $custom_mounts
+
+       local custom_mounts devices bindings links
+       custom_mounts=${1}
+       shift
+       devices=${@}
+
+       bindings="/tmp/bindings.list"
+       links="/tmp/links.list"
+       rm -rf ${bindings} ${links} 2> /dev/null
+
+       for device in ${devices}
+       do
+               if [ ! -b "${device}" ]
+               then
+                       continue
+               fi
+
+               local device_name backing include_list
+               device_name="$(basename ${device})"
+               backing=$(mount_persistence_media ${device})
+               if [ -z "${backing}" ]
+               then
+                       continue
+               fi
+
+               if [ -r "${backing}/${persistence_list}" ]
+               then
+                       include_list="${backing}/${persistence_list}"
+               elif [ -r "${backing}/${old_persistence_list}" ]
+               then
+                       include_list="${backing}/${old_persistence_list}"
+               else
+                       continue
+               fi
+
+               if [ -n "${DEBUG}" ] && [ -e "${include_list}" ]
+               then
+                       cp ${include_list} /live/persistence/${persistence_list}.${device_name}
+               fi
+
+               while read dir options # < ${include_list}
+               do
+                       if echo ${dir} | grep -qe "^[[:space:]]*\(#.*\)\?$"
+                       then
+                               # skipping empty or commented lines
+                               continue
+                       fi
+
+                       if trim_path ${dir} | grep -q -e "^[^/]" -e "^/lib" -e "^/lib/live\(/.*\)\?$" -e "^/\(.*/\)\?\.\.\?\(/.*\)\?$"
+                       then
+                               log_warning_msg "Skipping unsafe custom mount ${dir}: must be an absolute path containing neither the \".\" nor \"..\" special dirs, and cannot be \"/lib\", or \"/lib/live\" or any of its sub-directories."
+                               continue
+                       fi
+
+                       local opt_source opt_link source full_source full_dest
+                       opt_source=""
+                       opt_link=""
+                       for opt in $(echo ${options} | tr ',' ' ');
+                       do
+                               case "${opt}" in
+                                       source=*)
+                                               opt_source=${opt#source=}
+                                               ;;
+                                       link)
+                                               opt_link="true"
+                                               ;;
+                                       union|bind)
+                                               ;;
+                                       *)
+                                               log_warning_msg "Skipping custom mount with unkown option: ${opt}"
+                                               continue 2
+                                               ;;
+                               esac
+                       done
+
+                       source="${dir}"
+                       if [ -n "${opt_source}" ]
+                       then
+                               if echo ${opt_source} | grep -q -e "^/" -e "^\(.*/\)\?\.\.\?\(/.*\)\?$" && [ "${opt_source}" != "." ]
+                               then
+                                       log_warning_msg "Skipping unsafe custom mount with option source=${opt_source}: must be either \".\" (the media root) or a relative path w.r.t. the media root that contains neither comas, nor the special \".\" and \"..\" path components"
+                                       continue
+                               else
+                                       source="${opt_source}"
+                               fi
+                       fi
+
+                       full_source="$(trim_path ${backing}/${source})"
+                       full_dest="$(trim_path ${rootmnt}/${dir})"
+                       if [ -n "${opt_link}" ]
+                       then
+                               echo "${device} ${full_source} ${full_dest} ${options}" >> ${links}
+                       else
+                               echo "${device} ${full_source} ${full_dest} ${options}" >> ${bindings}
+                       fi
+               done < ${include_list}
+       done
+
+       # We sort the list according to destination so we're sure that
+       # we won't hide a previous mount. We also ignore duplicate
+       # destinations in a more or less arbitrary way.
+       [ -e "${bindings}" ] && sort -k3 -sbu ${bindings} >> ${custom_mounts} && rm ${bindings}
+
+       # After all mounts are considered we add symlinks so they
+       # won't be hidden by some mount.
+       [ -e "${links}" ] && cat ${links} >> ${custom_mounts} && rm ${links}
+
+       # We need to make sure that no two custom mounts have the same sources
+       # or are nested; if that is the case, too much weird stuff can happen.
+       local prev_source prev_dest
+       prev_source="impossible source" # first iteration must not match
+       prev_dest=""
+       # This sort will ensure that a source /a comes right before a source
+       # /a/b so we only need to look at the previous source
+       sort -k2 -b ${custom_mounts} |
+       while read device source dest options
+       do
+               if echo ${source} | grep -qe "^${prev_source}\(/.*\)\?$"
+               then
+                       panic "Two persistence mounts have the same or nested sources: ${source} on ${dest}, and ${prev_source} on ${prev_dest}"
+               fi
+               prev_source=${source}
+               prev_dest=${dest}
+       done
+}
+
+activate_custom_mounts ()
+{
+       local custom_mounts used_devices
+       custom_mounts="${1}" # the ouput from get_custom_mounts()
+       used_devices=""
+
+       while read device source dest options # < ${custom_mounts}
+       do
+               local opt_bind opt_link opt_union
+               opt_bind="true"
+               opt_link=""
+               opt_union=""
+               for opt in $(echo ${options} | tr ',' ' ');
+               do
+                       case "${opt}" in
+                               bind)
+                                       opt_bind="true"
+                                       unset opt_link opt_union
+                                       ;;
+                               link)
+                                       opt_link="true"
+                                       unset opt_bind opt_union
+                                       ;;
+                               union)
+                                       opt_union="true"
+                                       unset opt_bind opt_link
+                                       ;;
+                       esac
+               done
+
+               if [ -n "$(what_is_mounted_on "${dest}")" ]
+               then
+                       if [ "${dest}" = "${rootmnt}" ]
+                       then
+                               umount "${dest}"
+                       else
+                               log_warning_msg "Skipping custom mount ${dest}: $(what_is_mounted_on "${dest}") is already mounted there"
+                               continue
+                       fi
+               fi
+
+               if [ ! -d "${dest}" ]
+               then
+                       # create the destination and delete existing files in
+                       # its path that are in the way
+                       path="/"
+                       for dir in $(echo ${dest} | sed -e 's|/\+| |g')
+                       do
+                               path=$(trim_path ${path}/${dir})
+                               if [ -f ${path} ]
+                               then
+                                       rm -f ${path}
+                               fi
+                               if [ ! -e ${path} ]
+                               then
+                                       mkdir -p ${path}
+                                       if echo ${path} | grep -qe "^${rootmnt}/*home/[^/]\+"
+                                       then
+                                               # if ${dest} is in /home try fixing proper ownership by assuming that the intended user is the first, which is usually the case
+                                               # FIXME: this should really be handled by live-config since we don't know for sure which uid a certain user has until then
+                                               chown 1000:1000 ${path}
+                                       fi
+                               fi
+                       done
+               fi
+
+               # if ${source} doesn't exist on our persistence media
+               # we bootstrap it with $dest from the live filesystem.
+               # this both makes sense and is critical if we're
+               # dealing with /etc or other system dir.
+               if [ ! -d "${source}" ]
+               then
+                       if [ -n "${PERSISTENCE_READONLY}" ]
+                       then
+                               continue
+                       elif [ -n "${opt_union}" ] || [ -n "${opt_link}" ]
+                       then
+                               # unions and don't need to be bootstrapped
+                               # link dirs can't be bootstrapped in a sensible way
+                               mkdir -p "${source}"
+                               chown_ref "${dest}" "${source}"
+                               chmod_ref "${dest}" "${source}"
+                       elif [ -n "${opt_bind}" ]
+                       then
+                               # ensure that $dest is not copied *into* $source
+                               mkdir -p "$(dirname ${source})"
+                               cp -a "${dest}" "${source}"
+                       fi
+               fi
+
+               # XXX: If CONFIG_AUFS_ROBR is added to the Debian kernel we can
+               # ignore the loop below and set rootfs_dest_backing=$dest
+               local rootfs_dest_backing
+               rootfs_dest_backing=""
+               if [ -n "${opt_link}"]
+               then
+                       for d in /live/rootfs/*
+                       do
+                               if [ -n "${rootmnt}" ]
+                               then
+                                       fs="${d}/$(echo ${dest} | sed -e "s|${rootmnt}||")"
+                               else
+                                       fs="${d}/${dest}"
+                               fi
+                               if [ -d "${fs}" ]
+                               then
+                                       rootfs_dest_backing="${rootfs_dest_backing} ${fs}"
+                               fi
+                       done
+               fi
+
+               local cow_dir links_source
+               if [ -n "${opt_link}" ] && [ -z "${PERSISTENCE_READONLY}" ]
+               then
+                       link_files ${source} ${dest} ${rootmnt}
+               elif [ -n "${opt_link}" ] && [ -n "${PERSISTENCE_READONLY}" ]
+               then
+                       mkdir -p ${rootmnt}/lib/live/mount/persistence
+                       links_source=$(mktemp -d ${rootmnt}/lib/live/mount/persistence/links-source-XXXXXX)
+                       chown_ref ${source} ${links_source}
+                       chmod_ref ${source} ${links_source}
+                       # We put the cow dir in the below strange place to
+                       # make it absolutely certain that the link source
+                       # has its own directory and isn't nested with some
+                       # other custom mount (if so that mount's files would
+                       # be linked, causing breakage.
+                       cow_dir="/live/overlay/lib/live/mount/persistence/$(basename ${links_source})"
+                       mkdir -p ${cow_dir}
+                       chown_ref "${source}" "${cow_dir}"
+                       chmod_ref "${source}" "${cow_dir}"
+                       do_union ${links_source} ${cow_dir} ${source} ${rootfs_dest_backing}
+                       link_files ${links_source} ${dest} ${rootmnt}
+               elif [ -n "${opt_union}" ] && [ -z "${PERSISTENCE_READONLY}" ]
+               then
+                       do_union ${dest} ${source} ${rootfs_dest_backing}
+               elif [ -n "${opt_bind}" ] && [ -z "${PERSISTENCE_READONLY}" ]
+               then
+                       mount --bind "${source}" "${dest}"
+               elif [ -n "${opt_bind}" -o -n "${opt_union}" ] && [ -n "${PERSISTENCE_READONLY}" ]
+               then
+                       # bind-mount and union mount are handled the same
+                       # in read-only mode, but note that rootfs_dest_backing
+                       # is non-empty (and necessary) only for unions
+                       cow_dir="/live/overlay/${dest}"
+                       if [ -e "${cow_dir}" ] && [ -z "${opt_link}" ]
+                       then
+                               # If an earlier custom mount has files here
+                               # it will "block" the current mount's files
+                               # which is undesirable
+                               rm -rf "${cow_dir}"
+                       fi
+                       mkdir -p ${cow_dir}
+                       chown_ref "${source}" "${cow_dir}"
+                       chmod_ref "${source}" "${cow_dir}"
+                       do_union ${dest} ${cow_dir} ${source} ${rootfs_dest_backing}
+               fi
+
+               PERSISTENCE_IS_ON="1"
+               export PERSISTENCE_IS_ON
+
+               if echo ${used_devices} | grep -qve "^\(.* \)\?${device}\( .*\)\?$"
+               then
+                       used_devices="${used_devices} ${device}"
+               fi
+       done < ${custom_mounts}
+
+       echo ${used_devices}
+}
+
+fix_backwards_compatibility ()
+{
+       local device dir opt backing include_list
+       device=${1}
+       dir=${2}
+       opt=${3}
+
+       if [ -n "${PERSISTENCE_READONLY}" ]
+       then
+               return
+       fi
+
+       backing="$(mount_persistence_media ${device})"
+       if [ -z "${backing}" ]
+       then
+               return
+       fi
+
+       include_list="${backing}/${persistence_list}"
+       if [ ! -r "${include_list}" ] && [ ! -r "${backing}/${old_persistence_list}" ]
+       then
+               echo "# persistence backwards compatibility:
+${dir} ${opt},source=." > "${include_list}"
+       fi
+}
+
+is_mountpoint ()
+{
+       directory="$1"
+
+       [ $(stat -fc%d:%D "${directory}") != $(stat -fc%d:%D "${directory}/..") ]
+}