A Linux daemon that monitors the filesystem

• README.md: design document
• src/backupd.c: C source file containing code for the backupd daemon
• src/makefile: Makefile to build project

backupd is a Linux daemon written in C. Its primary purpose is to monitor select portions of the user's filesystem for changes. If it notices a volume of changes exceeding a user-specified threshold, it will trigger a backup operation of the user's choice.

Notable characteristics:

• The daemon runs continuously without manual intervention
• The daemon's behavior is specified by use of a config file
• All daemon activities are logged to both a custom log file and to the system logs.

The backupd daemon can be started manually via the command line with the following syntax:

./backupd config_file

To test, you will need to edit the .config file. The provided sample.config is an example config file. Just replace the filepaths with paths that you want on your system. (Note: the target directories must already exist).

Then, create files in the directories you set a target on. Each created file should show up as an event on the log. When the incident limit is reached, the backup command you specified in the config file will execute.

backupd may be installed as a systemd unit. This will allow the daemon to start on system boot, as well as give you control over the daemon's behavior using systemctl. Follow these steps to install the backupd systemd unit. Note: This will require admin privileges.

1. cd setup
2. ./setup.sh config_file

config_file is a filepath pointing to the location of the backupd config file. Note that if you change the location of this config file, you will need to reinstall backupd using this same setup procedure. If you just want to make changes to the config file, the following command will load those changes:

systemctl reload backupd

To uninstall the backupd systemd unit, run the remove.sh script in the setup/ directory.

The backupd daemon is controlled by the settings written to its configuration file. The config file contains whitespace separated data in a strict layout. The following information is contained:

• Log file location: filesystem path for custom log file
• Incident limit: the number of filesystem incidents backupd needs to see before invoking the backup action
• Backup action: a shell command to be invoked when the number of filesystem incidents reaches the predefined limit.
• Targets: list of directories to observe
• Target location
• Criteria for filesystem events to count as incidents in this target (file creation, file modification, etc...)

All of the above information is read into the struct config data type, using the read_config method and various *_config helper functions.

Each target is stored in the struct target data type. This struct contains all the information required to properly process an event detected in that specific target dir.

For an example config file, see sample.config

The daemonize method transforms the backupd process into a full-fledged daemon process.

backupd uses the start_log function to initialize its logs, and the logger method to record messages to them. end_log will release any logging resources used.

The daemon uses a global file descriptor, p_log, to refer to its custom log file (the location of which is set by the config file.)

If syslogging is enabled, all messages are sent with the identifier "backupd", along with the process ID. All messages have priority LOG_INFO.

Extreme conditions may require communication with the backupd daemon. Since it is not connected to a TTY, we must communicate with it via signals. The signal_handler function is essentially a big switch statement that allows backupd to respond to a variety of signals.

backupd starts a new thread for each of the target directories in its config file. That thread uses the select library function to block until it receives an inotify event indicating that a filesystem event has occured in the target directory. The thread can then hand off the event to the process_event function (along with the target struct).

backupd maintains an incident_counter global counter. This counter is incremented each time one of the threads detects an incident. It is kept threadsafe using a semaphore.

The process_event function is responsible for updating the incident_counter state, and potentially invoking the backup action.

The backup_action function is responsible for invoking whatever shell command is set in the config file. It is advised that this backup command write some sort of results elsewhere in the filesystem, as its progress is not recorded by the backupd daemon itself.

The log file below is the results of running the daemon with the below config file. Note that this config file does not contain all the information the final version will contain.

Initializing backupd
Starting target thread: /home/bpc30/test
Received event in /home/bpc30/test
File: /home/bpc30/test/file1
Incident Count: 1
File: /home/bpc30/test/file2
Incident Count: 2
File: /home/bpc30/test/file3
Incident Count: 3
File: /home/bpc30/test/file4
Incident Count: 4
File: /home/bpc30/test/file5
Incident Count: 5
File: /home/bpc30/test/file6
Incident Count: 6
File: /home/bpc30/test/file7
Incident Count: 7
File: /home/bpc30/test/file8
Incident Count: 8
File: /home/bpc30/test/file9
Incident Count: 9
File: /home/bpc30/test/file10
Incident Count: 10
Reached incident limit! Invoking backup_action.
Calling backup command: echo "<external update cmd output>" >> /home/bpc30/backupd.log 2>&1
<external update cmd output>
SIGTERM received
Closing log.

backupd has at least the following error codes: