Script to efficiently dump/restore whole disks using `sfdisk`&`fsarchiver` - useful for RasPi SD cards backup

disk-dump

#!/bin/bash

# Dumps and restores disk partition table and all disk partitions contents
# using fsarchiver. Useful for backing up and restoring complete SD cards
# in a size-efficient way.
#
# Supports testing that the dumped archives are restorable.

# Bash strict mode.
# see https://olivergondza.github.io/2019/10/01/bash-strict-mode.html
set -euo pipefail
trap 's=$?; $echo -e >&2 "\nERROR: line "$LINENO": $BASH_COMMAND\n"; exit $s' ERR

PTABLE_FILE=ptable.sfdisk
DISK_START_FILE=disk_start.img
DISK_START_LEN_MB=5
MD5SUM_FILE=md5sums.txt

USE_CORES=$(($(nproc) - 1))

#FSA_VERBOSE="-v"
FSA_VERBOSE=""
FSA_OPTS="-j$USE_CORES $FSA_VERBOSE"

# Define maximum disk size (in bytes) to avoid accidental operations on large disks.
MAX_DISK_SIZE=$((32 * 1024 * 1024 * 1024)) # 32GB

dd=/bin/dd
echo="/bin/echo -e"
lsblk=/bin/lsblk
sfdisk=/sbin/sfdisk
fsarchiver=/sbin/fsarchiver
wipefs=/sbin/wipefs
losetup=/sbin/losetup
truncate=/bin/truncate
md5sum=/bin/md5sum

title() {
  MSG="$1"

  $echo "**** $MSG"
}

debug() {
  MSG="$1"

  if [ ! -z "${DEBUG-}" ]; then
    $echo "D: $MSG"
  fi
}

check_device() {
  DEVICE="$1"

  if [ ! -b "$DEVICE" ]; then
    bad_arg "Device $DEVICE is not a block device."
  fi

  DEV_SIZE=$($lsblk -dnb -o size "$DEVICE")
  debug "Device size: ${DEV_SIZE}B"

  if [ "$DEV_SIZE" -gt "$MAX_DISK_SIZE" ]; then
    panic "Device $DEVICE size ${DEV_SIZE}B is greater than ${MAX_DISK_SIZE}B => probably not an SD card!"
  fi
}

check_command() {
  CMD="$1"

  if [ ! -x "$CMD" ]; then
    panic "Command $CMD not found or not executable."
  fi
}

usage() {
  $echo "Usage:"
  $echo
  $echo "$0 dump DEVICE"
  $echo "$0 restore DEVICE"
  $echo "$0 test"
  $echo
  $echo "dump    - dump partition table and start of disk"
  $echo "restore - restore partition table and start of disk"
  $echo "test    - test that archives are restorable"
  $echo "DEVICE  - block device to dump or restore to (e.g. /dev/sda)"
  $echo
  $echo "Examples:"
  $echo
  $echo "sudo $0 dump /dev/sdd 2>&1 | tee dump.log"
  $echo "sudo $0 restore /dev/sdd 2>&1 | tee restore.log"
  $echo "sudo $0 test 2>&1 | tee test.log"
}

bad_arg() {
  MSG="$1"

  $echo "\nERROR: $MSG\n"
  usage
  exit 1
}

panic() {
  MSG="$1"

  $echo "\nERROR: $MSG\n"
  exit 1
}

args_num_at_least() {
  AT_LEAST="$1"
  ACTUAL="$2"
  MSG="$3"

  if [ "$ACTUAL" -lt $AT_LEAST ]; then
    bad_arg "$MSG"
  fi
}

dump() {
  args_num_at_least 1 $# "DEVICE argument missing."
  DEVICE=$1

  check_command "$sfdisk"
  check_command "$dd"
  check_command "$fsarchiver"
  check_command "$md5sum"
  check_device "$DEVICE"

  title "Dumping partition table of $DEVICE to $PTABLE_FILE"
  $sfdisk -d "$DEVICE" > "$PTABLE_FILE"

  title "Dumping first ${DISK_START_LEN_MB}MB of $DEVICE to $DISK_START_FILE"
  $dd "if=$DEVICE" "of=$DISK_START_FILE" bs=1M "count=$DISK_START_LEN_MB"

  PARTITIONS=$($lsblk -n -o path "$DEVICE" | tail -n +2 | xargs)
  debug "Partitions found: $PARTITIONS"
  debug "Will use $USE_CORES cores for fsarchiver."
  for PARTITION in $PARTITIONS; do
    PARTNUM=$(echo "$PARTITION" | tr -d -c 0-9 )
    DUMPFILE="partition-${PARTNUM}.fsa"
    title "Dumping partition $PARTITION to $DUMPFILE"
    $fsarchiver savefs $FSA_OPTS -o "$DUMPFILE" "$PARTITION"
  done

  title "Storing MD5 checksums of dump files to $MD5SUM_FILE"
  $md5sum "$PTABLE_FILE" "$DISK_START_FILE" partition-*.fsa > "$MD5SUM_FILE"

  # Print date in case script output is logged to file.
  title "Dump OK. Date: $(date)"
}

restore() {
  args_num_at_least 1 $# "DEVICE argument missing."
  DEVICE=$1

  # Determine partition number prefix.
  # Needed for loop devices used in test_archives() where device name is
  # something like /dev/loop0 and partitions are /dev/loop0p1, /dev/loop0p2, ...
  # => prefix is "p".
  # For regular devices like /dev/sda, partitions are /dev/sda1, /dev/sda2, ...
  # => prefix is "".
  PARTNUM_PREFIX="${2-}"

  check_command "$wipefs"
  check_command "$dd"
  check_command "$sfdisk"
  check_command "$fsarchiver"
  check_device "$DEVICE"

  title "Wiping all partitions on $DEVICE"
  $wipefs -a "$DEVICE"

  title "Restoring first ${DISK_START_LEN_MB}MB of $DISK_START_FILE to $DEVICE"
  $dd "if=$DISK_START_FILE" "of=$DEVICE"

  title "Restoring partition table from $PTABLE_FILE to $DEVICE"
  $sfdisk "$DEVICE" < "$PTABLE_FILE"

  DUMPFILES=$(ls partition-*.fsa | sort | xargs)
  debug "Dump files found: $DUMPFILES"
  for DUMPFILE in $DUMPFILES; do
    PARTNUM=$(echo "$DUMPFILE" | tr -d -c 0-9 )
    PARTITION_DEVICE="${DEVICE}${PARTNUM_PREFIX}${PARTNUM}"
    title "Restoring $DUMPFILE to $PARTITION_DEVICE"
    $fsarchiver restfs "$DUMPFILE" "id=0,dest=$PARTITION_DEVICE"
  done

  # Print date in case script output is logged to file.
  title "Restore OK. Date: $(date)"
}

test_archives_cleanup() {
  IMAGE_FILE="$1"
  LOOP_DEVICE="$2"

  title "Destroying loop device $LOOP_DEVICE"
  $losetup -d "$LOOP_DEVICE"

  title "Deleting virtual disk image $IMAGE_FILE"
  rm "$IMAGE_FILE"
}

test_archives() {
  check_command "$md5sum"
  check_command "$truncate"
  check_command "$losetup"
  check_command "$sfdisk"

  title "Verifying MD5 checksums of dump files from $MD5SUM_FILE"
  $md5sum -c "$MD5SUM_FILE"

  IMAGE_FILE="tmp-virtual-disk.img"
  LOOP_DEVICE=$($losetup -f)

  title "Creating temporary virtual disk image $IMAGE_FILE"
  $truncate -s $MAX_DISK_SIZE "$IMAGE_FILE"

  title "Setting up loop device $LOOP_DEVICE for $IMAGE_FILE"
  $losetup "$LOOP_DEVICE" "$IMAGE_FILE"
  trap "test_archives_cleanup '$IMAGE_FILE' '$LOOP_DEVICE'" EXIT

  # We must dump partition table first to create partitions on the loop device
  # because partprobe /dev/loopX does not work reliably.
  title "Restoring partition table from $PTABLE_FILE and reattaching loop device"
  $sfdisk "$LOOP_DEVICE" < "$PTABLE_FILE"
  $losetup -d "$LOOP_DEVICE"
  # Wait a bit to ensure the device is fully detached.
  sleep 1
  $losetup -P "$LOOP_DEVICE" "$IMAGE_FILE"

  restore "$LOOP_DEVICE" p

  # Print date in case script output is logged to file.
  title "Test OK. Date: $(date)"
}

args_num_at_least 1 $# "Command not specified."
COMMAND=$1
shift
case $COMMAND in
dump)
  dump $@
  ;;
restore)
  restore $@
  ;;
test)
  test_archives $@
  ;;
*)
  bad_arg "Invalid command: $COMMAND"
esac