md5 all files in a directory tree
I would like to create a bash script that makes a md5 checksum of every file in this directory. I want to be able to type the script name in the CLI and then the path to the directory I want to hash and have it work. I’m sure there are many ways to accomplish this. Currently I have:
#!/bin/bash for file in "$1" ; do md5 >> "$__checksums.md5" done This just hangs and it not working. Perhaps I should use find? One caveat — the directories I want to hash will have files with different extensions and may not always have this exact same tree structure. I want something that will work in these different situations, as well.
8 Answers 8
Using md5deep
md5deep -r path/to/dir > sums.md5 Using find and md5sum
find relative/path/to/dir -type f -exec md5sum <> + > sums.md5 Be aware, that when you run check on your MD5 sums with md5sum -c sums.md5 , you need to run it from the same directory from which you generated sums.md5 file. This is because find outputs paths that are relative to your current location, which are then put into sums.md5 file.
If this is a problem you can make relative/path/to/dir absolute (e.g. by puting $PWD/ in front of your path). This way you can run check on sums.md5 from any location. Disadvantage is, that now sums.md5 contains absolute paths, which makes it bigger.
Fully featured function using find and md5sum
You can put this function to your .bashrc file (located in your $HOME directory):
function md5sums < if [ "$#" -lt 1 ]; then echo -e "At least one parameter is expected\n" \ "Usage: md5sums [OPTIONS] dir" else local OUTPUT="checksums.md5" local CHECK=false local MD5SUM_OPTIONS="" while [[ $# >1 ]]; do local key="$1" case $key in -c|--check) CHECK=true ;; -o|--output) OUTPUT=$2 shift ;; *) MD5SUM_OPTIONS="$MD5SUM_OPTIONS $1" ;; esac shift done local DIR=$1 if [ -d "$DIR" ]; then # if $DIR directory exists cd $DIR # change to $DIR directory if [ "$CHECK" = true ]; then # if -c or --check option specified md5sum --check $MD5SUM_OPTIONS $OUTPUT # check MD5 sums in $OUTPUT file else # else find . -type f ! -name "$OUTPUT" -exec md5sum $MD5SUM_OPTIONS <> + > $OUTPUT # Calculate MD5 sums for files in current directory and subdirectories excluding $OUTPUT file and save result in $OUTPUT file fi cd - > /dev/null # change to previous directory else cd $DIR # if $DIR doesn't exists, change to it to generate localized error message fi fi > After you run source ~/.bashrc , you can use md5sums like normal command:
will generate checksums.md5 file in path/to/dir directory, containing MD5 sums of all files in this directory and subdirectories. Use:
to check sums from path/to/dir/checksums.md5 file.
Note that path/to/dir can be relative or absolute, md5sums will work fine either way. Resulting checksums.md5 file always contains paths relative to path/to/dir . You can use different file name then default checksums.md5 by supplying -o or —output option. All options, other then -c , —check , -o and —output are passed to md5sum .
First half of md5sums function definition is responsible for parsing options. See this answer for more information about it. Second half contains explanatory comments.
Note: In my experience using -exec command <> + variant of exec makes it run faster than when variant -exec command ; is used.
find /path/you/need -type f -exec md5sum <> \; > checksums.md5
Update#1: Improved the command based on @twalberg’s recommendation to handle white spaces in file names.
Update#2: Improved based on @jil’s suggestion, to remove unnecessary xargs call and use -exec option of find instead.
Update#3: @Blake a naive implementation of your script would look something like this:
#!/bin/bash # Usage: checksumchecker.sh find "$1" -type f -exec md5sum <> \; > "$1"__checksums.md5 I would recommend find /path -type f -print0 | xargs -0 md5sum , to deal with file names that otherwise might get unintentionally split due to whitespace.
Thank you @taskalman. «You can build the path and output file name from $1 if we put it in your script. Note that you will have to handle the slashes in your path parameter to make it part of the filename in your script.» Can you explain this a bit further? I don’t quite understand what you mean.
@Blake sorry, misunderstood the logic, you van ignore that part, I remove it from the answer. I didn’t realize you want tó save the file __checksums.md5 in the root of your search path.
@taskalman I still want to use the $1 so that I don’t haveto have to type out a find command each time, rather just a script name, say «makechecksum», and drop the directory to be hashed into the CLI.
Updated Answer
If you like the answer below, or any of the others, you can make a function that does the command for you. So, to test it, type the following into Terminal to declare a function:
function sumthem() < find "$1" -type f -print0 | parallel -0 -X md5 >checksums.md5; > sumthem /Users/somebody/somewhere If that works how you like, you can add that line to the end of your «bash profile» and the function will be declared and available whenever you are logged in. Your «bash profile» is probably in $HOME/.profile
Original Answer
Why not get all your CPU cores working in parallel for you?
find . -type f -print0 | parallel -0 -X md5sum This finds all the files ( -type f ) in the current directory ( . ) and prints them with a null byte at the end. These are then passed passed into GNU Parallel, which is told that the filenames end with a null byte ( -0 ) and that it should do as many files as possible at a time ( -X ) to save creating a new process for each file and it should md5sum the files.
This approach will pay the largest bonus, in terms off speed, with big images like Photoshop files.
Yes, this would work, but I would have to continually re-type the whole command. I want a script where the directory to be hashed is a variable and the name of the file output is based on that variable. I don’t want to have to type out a find command each time — just a script name, say «makechecksum», and drop the directory to be hashed into the CLI.
#!/bin/bash shopt -s globstar md5sum "$1"/** > "$__checksums.md5" Explanation: shopt -s globstar (manual) enables ** recursive glob wildcard. It will mean that «$1″/** will expand to list of all the files recursively under the directory given as parameter $1 . Then the script simply calls md5sum with this file list as parameter and > «$__checksums.md5» redirects the output to the file.
your answer may be correct , but it would be a lot more helpful if you could explain what it does or how it works
@Blake Be aware that this won’t hash hidden files. If you want hidden files to not be ignored, then activate dotglob option: shopt -s dotglob [source].
@jil this is just hanging for me #!/usr/bin/env bash shopt -s globstar for file in «$1″/** ; do md5deep -br >> «$<1>__checksums.md5″ done
md5deep -r $your_directory | awk | sort | md5sum | awk
Please edit your answer to include some explanation. Code-only answers do very little to educate future SO readers. Your answer is in the moderation queue for being low-quality.
Use find command to list all files in directory tree, then use xargs to provide input to md5sum command
find dirname -type f | xargs md5sum > checksums.md5 In case you prefer to have separate checksum files in every directory, rather than a single file, you can
- find all subdirectories
- keep only those which actually contain files (not only other subdirs)
- cd to each of them and create a checksums.md5 file inside that directory
Here is a an example script which does that:
#!/bin/bash # Do separate md5 files in each subdirectory md5_filename=checksums.md5 dir="$1" [ -z "$dir" ] && dir="." # Check OS to select md5 command if [[ "$OSTYPE" == "linux-gnu"* ]]; then is_linux=1 md5cmd="md5sum" elif [[ "$OSTYPE" == "darwin"* ]]; then md5cmd="md5 -r" else echo "Error: unknown OS '$OSTYPE'. Don't know correct md5 command." exit 1 fi # go to base directory after saving where we started start_dir="$PWD" cd "$dir" # if we're in a symlink cd to the real path if [ ! "$dir" = "$(pwd -P)" ]; then dir="$(pwd -P)" cd "$dir" fi if [ "$PWD" = "/" ]; then die "Refusing to do it on system root '$PWD'" fi # Find all folders to process declare -a subdirs=() declare -a wanted=() # find all non-hidden subdirectories (not if the name begins with "." like ".Trashes", ".Spotlight-V100", etc.) while IFS= read -r; do subdirs+=("$PWD/$REPLY"); done < <(find . -type d -not -name ".*" | LC_ALL=C sort) # count files and if there are any, add dir to "wanted" array echo "Counting files and sizes to process . " for d in "$dir" "$"; do # include "$dir" itself, not only it's subdirs files_here=0 while IFS= read -r ; do (( files_here += 1 )) done < <(find "$d" -maxdepth 1 -type f -not -name "*.md5") (( files_here )) && wanted+=("$d") done echo "Found $<#wanted[@]>folders to process:" printf " * %s\n" "$" if [ "$" = 0 ]; then echo "Nothing to do. Exiting." exit 0 fi for d in "$"; do cd "$d" find . -maxdepth 1 -type f -not -name "$md5_filename" -print0 \ | LC_ALL=C sort -z \ | while IFS= read -rd '' f; do $md5cmd "$f" | tee -a "$md5_filename" done cd "$dir" done cd "$start_dir" (This is actually a very simplified version of this «md5dirs» script on Github. The original is quite specific and more complex, making it less illustrative as an example, and more difficult to adapt to other different needs.)
How to checksum and verify all files on a filesystem
I would like to run sha1sum or md5sum on all my files recursively, store the results to a file (or a couple of files) and then be able to scan through the files again to verify the integrity of the files. tools like sha1sum / md5sum does not allow for recursive, and I would also like to spread the load so it uses more than one CPU core. I do not yet feel safe putting my files on BTRFS and besides BTRFS does not yet work properly on raid5/6 so I am looking for a «quick» way to checksum all my files and verify them whenever I want. That way I would at least know if silent data corruption have hampered any of my files. Does any tools exist that is designed for this purpose (yes, I am aware of SnapRAID)
How about ZFS? It is available on Linux, too, from what I’ve heard. I mean, I would avoid manual periodic scripts. This is not very reliable.
@Hastur : In Debian your command does not work. If anything it have to look like this: ‘find -type f /path/ -exec md5sum <> \; >checksums’ and to check it you could use ‘md5sum -c checksums | grep «:FAILED» >checksumfailures’
2 Answers 2
These tools are intended for more or less this use case:
Or you could write something on your own;) I wrote this script in Python merging other people snippets from the web. It can be a starting point.
#!/usr/bin/python import os import hashlib from joblib import Parallel, delayed import json BASE_DIR = "/home/p/music" NUM_PROC = 2 OUT_FILE = "hashes.json" # list of files to be processed file_list = [] for root,dir,files in os.walk(BASE_DIR): for name in files : file_list.append(os.path.join(root, name)) # -- md5sum computation for a single file def hashOne(f): BLOCKSIZE = 65536 hasher = hashlib.md5() with open(f, 'rb') as afile: print f buf = afile.read(BLOCKSIZE) while len(buf) > 0: hasher.update(buf) buf = afile.read(BLOCKSIZE) return hasher.hexdigest() # parallel calculation of md5sum md5 = Parallel(n_jobs=NUM_PROC)(delayed(hashOne)(f) for f in file_list) # make an hash of file_names -> md5sum out_hash = <> for i in range(len(file_list)): out_hash[file_list[i]] = md5[i] # write out the hash "filename -> md5sum" to disk for future dump = json.dumps(out_hash) with open(OUT_FILE, "wb") as file: file.write(dump)