cp — Copy Files and Directories

cp (copy) is one of the most fundamental Unix commands: it duplicates files and directories, creates backups, and deploys content across the filesystem. Day to day you reach for it to snapshot a config file before editing, copy a project tree recursively with -r, or make an exact clone with -a, preserving permissions, ownership and timestamps. One important pitfall: without -i, cp overwrites existing destination files silently – when in doubt, use -i (prompt), -n (never overwrite) or -b (keep a backup). On modern filesystems such as Btrfs or XFS, --reflink makes copies practically instant.

Basic Copying

cp <source> <destination> — Copy a file to a new location or name.

cp config.yml config-backup.yml

cp <file1> <file2> <directory> — Copy multiple files into a directory.

cp index.html style.css app.js /var/www/html/

cp <source> <directory>/ — Copy a file into a directory, keeping the original filename.

cp report.pdf /tmp/

cp *.txt <directory> — Copy all files matching a glob pattern into a directory.

cp *.jpg /backup/photos/

Directory Copying

cp -r <source_dir> <destination_dir> — Copy a directory recursively including all contents.

cp -r ./project/ /backup/project/

cp -R <source_dir> <destination_dir> — Synonym for -r. Copy directories recursively.

cp -R ./src/ /backup/src/

cp -r <source_dir>/ <destination_dir>/ — Copy the contents of a directory into another directory.

cp -r ./dist/ /var/www/html/

cp -r <dir1> <dir2> <dir3> <target_dir> — Copy multiple directories into a target directory.

cp -r ./css ./js ./images /var/www/html/

Preserving Attributes

cp -p <source> <destination> — Preserve file mode, ownership, and timestamps.

cp -p script.sh /usr/local/bin/

cp -a <source> <destination> — Archive mode. Same as -dR --preserve=all. Preserves everything including symlinks.

cp -a /var/www/html/ /backup/www/

cp --preserve=mode,timestamps <source> <destination> — Preserve specific attributes only.

cp --preserve=mode,timestamps deploy.sh /usr/local/bin/

cp --preserve=all <source> <destination> — Preserve all attributes: mode, ownership, timestamps, context, links, xattr.

cp --preserve=all important.conf /etc/app/

cp --no-preserve=ownership <source> <destination> — Copy but explicitly do not preserve ownership.

cp --no-preserve=ownership /root/config.yml /home/user/

Overwrite Control

cp -i <source> <destination> — Interactive mode. Prompt before overwriting an existing file.

cp -i new-config.yml /etc/app/config.yml

cp -n <source> <destination> — No clobber. Never overwrite an existing file.

cp -n defaults.conf /etc/app/config.conf

cp -f <source> <destination> — Force. Remove destination file if it cannot be opened, then copy.

cp -f updated.bin /usr/local/bin/app

cp -u <source> <destination> — Update. Only copy when source is newer than destination or destination is missing.

cp -u *.html /var/www/html/

cp --update=none <source> <destination> — Explicit update mode: none (same as -n), all (default), or older (same as -u).

cp --update=older ./assets/* /var/www/assets/

Symlinks & Special Files

cp -d <source> <destination> — Copy symbolic links as links, not the files they point to.

cp -d current-release /backup/current-release

cp -L <source> <destination> — Always follow symbolic links. Copy the file the symlink points to.

cp -L /etc/alternatives/editor ./editor-backup

cp -P <source> <destination> — Never follow symbolic links. Copy the symlink itself (default for -r).

cp -P link.txt /backup/

cp -s <source> <destination> — Create a symbolic link instead of copying the file.

cp -s /opt/app/config.yml /etc/app/config.yml

cp -l <source> <destination> — Create a hard link instead of copying. Saves disk space.

cp -l large-file.iso /backup/large-file.iso

Verbosity & Feedback

cp -v <source> <destination> — Verbose mode. Print each file as it is copied.

cp -v *.conf /etc/app/

cp -rv <source_dir> <destination_dir> — Recursively copy with verbose output. Shows every file being copied.

cp -rv ./project/ /backup/project/

cp -v --backup=numbered <source> <destination> — Create numbered backups of existing files and show output.

cp -v --backup=numbered config.yml /etc/app/config.yml

Backup Options

cp -b <source> <destination> — Create a backup of the destination file (appends ~ suffix) before overwriting.

cp -b new.conf /etc/app/app.conf

cp --backup=numbered <source> <destination> — Create numbered backups (.~1~, .~2~, etc.) instead of a single ~ backup.

cp --backup=numbered config.yml /etc/app/config.yml

cp --backup=existing <source> <destination> — Use numbered backups if any already exist, otherwise use simple ~ backup.

cp --backup=existing data.json /var/data/data.json

cp -S '<suffix>' <source> <destination> — Use a custom suffix for backup files instead of the default ~.

cp -b -S '.bak' config.yml /etc/app/config.yml

Sparse Files & Performance

cp --sparse=auto <source> <destination> — Auto-detect and preserve sparse files (default behavior).

cp --sparse=auto disk.img /backup/

cp --sparse=always <source> <destination> — Always try to create sparse files. Useful for disk images and virtual disks.

cp --sparse=always vm-disk.qcow2 /backup/

cp --reflink=auto <source> <destination> — Use copy-on-write if the filesystem supports it (Btrfs, XFS). Instant copies.

cp --reflink=auto large-file.tar /backup/

cp --reflink=always <source> <destination> — Force copy-on-write. Fails if filesystem does not support it.

cp --reflink=always database.db /snapshot/

Target Directory & Structure

cp -t <directory> <file1> <file2> — Specify the target directory first. Useful with xargs and find.

find . -name '*.log' -print0 | xargs -0 cp -t /backup/logs/

cp -T <source> <destination> — Treat destination as a normal file, not a directory. Fails if destination is a directory.

cp -T config-new.yml config.yml

cp --parents <source> <destination> — Preserve the full source path structure under the destination directory.

cp --parents src/app/main.js /backup/

Common Patterns

cp -a <source>/ <destination>/ — Mirror a directory tree preserving all attributes. Best for local backups.

cp -a /var/www/html/ /backup/www/

cp -rn <source>/ <destination>/ — Copy new files only, skip existing. Useful for incremental merges.

cp -rn ./new-assets/ /var/www/assets/

cp -ru <source>/ <destination>/ — Copy only newer or missing files recursively.

cp -ru ./updated-site/ /var/www/html/

cp -av <source>/ <destination>/ — Archive copy with verbose output. See exactly what gets copied.

cp -av /home/user/ /backup/user/

cp /dev/null <file> — Truncate a file to zero bytes (empty it) without deleting it.

cp /dev/null /var/log/app.log

find <dir> -name '<pattern>' -exec cp {} <dest>/ \; — Find files by pattern and copy them to a single destination directory.

find ./project/ -name '*.pdf' -exec cp {} /backup/pdfs/ \;

find <dir> -name '<pattern>' -print0 | xargs -0 cp -t <dest>/ — Find and copy files safely (handles filenames with spaces).

find ./docs/ -name '*.md' -print0 | xargs -0 cp -t /backup/docs/

Conclusion

cp is the workhorse for duplicating files: for everyday use, cp <source> <destination>, cp -r for directories and cp -a for exact mirrors with all attributes cover most needs. Keep in mind that without -i, cp overwrites existing destination files silently – when handling important data, -i, -n or -b are worth the keystrokes. Copy first, then modify the original is a good habit. On copy-on-write filesystems, --reflink=auto saves both time and space, and for large local backups cp -a is usually the right choice.