From: stack Date: Tue, 19 Feb 2013 00:49:23 +0000 (+0100) Subject: persistence luks mounting "Loading" error X-Git-Url: https://v.licheni.net/stack/code/debianlive.git/commitdiff_plain/d6cf774555166ee949904291232a250992956219 persistence luks mounting "Loading" error 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. --- 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 index 0000000..0515f9b --- /dev/null +++ b/xfce-desktop/config/includes.chroot/lib/live/boot/9990-misc-helpers.sh @@ -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&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}/..") ] +}