I am doing the second BASH exercise from TLDP Bash-Scripting Guide, and I have most of it figured out up until the part when it comes time to copy the compressed files to an inserted USB.

Home Directory Listing

Perform a recursive directory listing on the user's home directory and save the information to a file. Compress the file, have the script prompt the user to insert a USB flash drive, then press ENTER. Finally, save the file to the flash drive after making certain the flash drive has properly mounted by parsing the output of df. Note that the flash drive must be unmounted before it is removed.

As I progress with the script it is becoming less ..elegant, and was wondering if there was a better way to do this. I know creating files is likely not the most efficient way to do the comparisons, but have not got the shell expansions figured yet, and intend to change those once I get it working.

The problem specifically is, to ensure that the usb is mounted and that I am writing to the USB and nowhere else. I am comparing the last line of df after the USB is plugged in with the last line of df from the diff between df before USB is plugged in and df after USB is plugged in, and looping until they match. Unfortunately, the diff result starts with a >, but I intend to use sed to get rid of that. The real problem is the path to where my usb is mounted is:

/media/flerb/"Title of USB with spaces"

To make this portable for USBs that may have different names is my best bet from here to do something with awk and field separators? And as a follow-up, I know this is pretty inelegant, and wonder if there is a cleaner way to go about this...especially because this is the second exercise and still in EASY.

The output from the df tails is:

/dev/sdb1                     15611904  8120352   7491552  53% /media/flerb/CENTOS 7 X8
> /dev/sdb1                     15611904  8120352   7491552  53% /media/flerb/CENTOS 7 X8

The script so far

 1 #!/bin/bash
  3 if [ "$UID" -eq 0 ] ; then
  4         echo "Don't run this as root"
  5         exit 1
  6 fi
  8 #Create a backup file with the date as title in a backup directory
  9 BACKUP_DIR="$HOME/backup"
 10 DATE_OF_COPY=$(date --rfc-3339=date)
 13 [ -d "$BACKUP_DIR" ] || mkdir -m 700 "$BACKUP_DIR"
 15 #find all files recursively in $HOME directory
 16 find -P $HOME >> "$BACKUP_FILE"
 18 #use lzma to compress
 19 xz -zk --format=auto --check=sha256 --threads=0 "$BACKUP_FILE"
 21 #making files to use in operations
 22 BEFORE="$BACKUP_DIR"/before_usb.txt
 23 AFTER="$BACKUP_DIR"/after_usb.txt
 24 DIFFERENCE="$BACKUP_DIR"/difference.txt
 26 df > "$BEFORE"
 27 read -p 'Enter USB and press any button' ok
 28 sleep 2
 29 df > "$AFTER"
 30 diff "$BEFORE" "$AFTER" > "$DIFFERENCE"
 31 sleep 2
 32 echo
 34 TAIL_AFTER=$(tail -n 1 "$AFTER")
 35 TAIL_DIFF=$(tail -n 1 "$DIFFERENCE")
 37 until [ "$TAIL_AFTER" == "$TAIL_DIFF" ] ;
 38 do
 39         echo "Not yet"
 40         df > "$AFTER"
 41         TAIL_AFTER=$(tail -n 1 "$AFTER")
 42         diff "$BEFORE" "$AFTER" > "$DIFFERENCE"
 43         TAIL_DIFF=$(tail -n 1 "$DIFFERENCE")
 44         echo "$TAIL_AFTER"
 45         echo "$TAIL_DIFF"
 46         sleep 1
 48 done
 49 exit $?
  • (1) is this on a Linux system, and (2) will the USB device always be /dev/sdb1?
    – Jeff Schaller
    Jul 12, 2017 at 17:58
  • Yes it is on Ubuntu 17.04, and I think it would be pretty slick if it didn't rely on the USB device being on /dev/sdb1, but not totally necessary.
    – flerb
    Jul 12, 2017 at 18:03
  • 1
    What happens if in another process, someone mounts/unmounts other media (e. g. a CD, or another tmpfs block)? It might be smarter to watch the output of awk 'NR!=1{print $NF}' <(df)|sort for the specific mount point you're looking for, remembering only that, and then watching for it to go away.
    – DopeGhoti
    Jul 12, 2017 at 18:05
  • 1
    along with DopeGhoti's suggestion, I'll suggest GNU df's --output=target option as one way to get the mountpoint in a simple fashion
    – Jeff Schaller
    Jul 12, 2017 at 18:08
  • I understand now why I need either the mount point to be constant or to know the name of the device. df --output=target is a nice way to get the mountpoint. The awk command gives only X8 (the last chunk of the USB's name) for me. Could you explain what you mean by watching for the specific mount point to go away?
    – flerb
    Jul 12, 2017 at 18:35

1 Answer 1


To get the right USB and it's path:

  • noticed that the first USB plugged into my system is mounted on the /dev/sdb1 filesystem and the second is mounted on the /dev/sdc1 filesystem. I added a section to switch to /dev/sdc1 if udevadm info -q all -n "/dev/sdb" | grep "ID_BUS=usb" returned true.

  • if diff between df --output=target | grep "/media/darren" before and after has "/media/darren" in it, that's presumably the new USB. USB is mounted.

  • if udevadm info -q all -n "/dev/sdb" | grep "ID_BUS=usb is true then that device mounted on "/dev/sdb" is a usb device

  • to get the path, and make sure I was grabbing the right one: df --output=source,avail,target | grep "$FILESYSTEM" | grep "$MOUNTPOINT" and some really nice awk

    udevadm info -q all -n "$FILESYSTEM" | grep "ID_BUS=usb" > $RESULT
    if grep "ID_BUS=usb" "$RESULT" ; then
        df --output=source,avail,target | grep "$FILESYSTEM" | grep "$MOUNTPOINT" > "$USB_ADDR"
        awk -i inplace '{$0=gensub(/\s*\S+/,"",1)}1' "$USB_ADDR"
        awk -i inplace '{$0=gensub(/\s*\S+/,"",1)}1' "$USB_ADDR"
        sed -i 's/^ *//' "$USB_ADDR"

Unfortunately diffing fails if other stuff is added or removed. But, at least it fails nicely if it grabs the wrong device when using the df filtered output and grepping for the filesystem and mountpoint.

Tried awk 'NR!=1{print $NF}' <(df)|sort. Still not great on bash expansions.


if [ "$UID" -eq 0 ] ; then
        echo "Don't run this as root"
        exit 50

#Create a backup file with the date as title in a backup directory
DATE_OF_COPY=$(date --rfc-3339=date)

[ -d "$BACKUP_DIR" ] || mkdir -m 700 "$BACKUP_DIR"

#find all files recursively in $HOME directory
find -P $HOME >> "$BACKUP_FILE"

#use lzma to compress
xz -zk --format=auto --check=sha256 --threads=0 "$BACKUP_FILE"

#making files to use in operations of before and after usb is inserted info and diff between them
#this is part of how we will insure that we are writing to the correct device
#MOUNTPOINT=where computer mounts USBs by default

#if there is something on the mountpoint before we plug in the usb, we want to know
#so the new device is written to
udevadm info -q all -n "$FILESYSTEM" | grep "ID_BUS=usb" > $RESULT
if grep "ID_BUS=usb" "$RESULT" ; then
        echo $FILESYSTEM

df --output=target | grep $MOUNTPOINT > "$BEFORE"

read -p 'Enter USB or hard drive and press any button' ok

df --output=target | grep $MOUNTPOINT > "$AFTER"
#dif them to find the new one

# Look in mountpoint in the difference between the before and after df outputs

until [ "$SEARCH_FOR_MOUNT" ] ;
        df --output=target | grep $MOUNTPOINT > "$AFTER"
        diff "$BEFORE" "$AFTER" > "$DIFFERENCE"

#check if new device plugged into filesystem is a USB
udevadm info -q all -n "$FILESYSTEM" | grep "ID_BUS=usb" > $RESULT
if grep "ID_BUS=usb" "$RESULT" ; then
        df --output=source,avail,target | grep "$FILESYSTEM" | grep "$MOUNTPOINT" > "$USB_ADDR"
        awk -i inplace '{$0=gensub(/\s*\S+/,"",1)}1' "$USB_ADDR"
        awk -i inplace '{$0=gensub(/\s*\S+/,"",1)}1' "$USB_ADDR"
        sed -i 's/^ *//' "$USB_ADDR"

        if [ ! -s $USB_ADDR ] ; then
              echo "Error finding USB Address"
              exit 100

        ADDR=$(cat "$USB_ADDR")
        cp "$BACKUP_FILE.xz" "$ADDR"

        echo "Device is not a USB"
        exit 52

exit $?

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .