Overview

semaphore can be used in shell scripts for locking

About

There are situations in bash scripts where you have to make sure that that section is executed only once even though multiple processes may be running in parallel executing the same script.

In order to assure integrity I have being using the following common method:

if mkdir /tmp/whatever; then
	doStuff that must be run only once at a given time
	rmdir /tmp/wahtever
fi

This however does not always work. I noticed several times that multiple processes entered the then-fi block at the same time. That's why I wrote this little c-program, enjoy.

Build

Just run

make

and it should compile it (provided gcc and install tool are present). Output looks like:

gcc  -o semaphore -lpthread -lrt semaphore.c
install -m 755 semaphore /usr/bin

Usage

usage: semaphore -{s sec|c|h|r|t|T|p} [-k sig] name
  -s sec  : Set and test semaphore whithout waiting for it. Return code = 0 means not used, you own the lock now. If lock is older than sec seconds we clear it.
  -c      : Clear/release semaphore.
  -r      : Read semaphore.
  -k sig  : If sig >0: Kill parent of old lock holder if timed out. sig = signal (number) to use for kill(2). If sig = 0, no kill.
  -t      : Show timer of semaphore. t=time, c=counter, p=PID, pp=parent PID
  -T      : Show timer in seconds since epoch.
  -l 0-9  : Set debuglevel.
  -p      : Print parent pid (pp).
  -h      : This short help.
Source can be found at: https://github.com/StarPet/semaphore
Note: Prefix used.........: "/dev/shm/semaphore."
Note: Value to sem_open(2): "semaphore"
Note: Only PIDs >1 will be killed.
Note: Debug log at /var/logs/semaphore.log

Use Cases

You may want to check out the example.sh bash script for a some example uses.

In the simple examples below I use $LOCK for the name of the LOCK. You need to have a LOCK=somename prior to get to that reference. If you have multiple places where you use semaphore for a different purpose, make sure to use a different value for $LOCK for each purpose.

Here are some use cases:

# Test and set: Here `-s 1` means a lock older than 1 sec is treated as free.
semaphore -s 1 $LOCK
rc=$?
# Here you should check the return code (stored in rc) if the lock is set.
# rc == 0: the lock is ours and set.
# rc != 0: the lock is set by some other process
<your code>
# Once done, clear the lock. But only if you own it (rc was 0)
semaphore -c $LOCK

# Test but do not set the LOCK
semaphore -t $LOCK
rc=$?
# Use

Important note:

The semaphore program does NOT check if the user is permitted to perform the change in the lock or not. Just the access to the used lock file (see SEMPREFIX below) is needed.

Environment

DEBUGLEVEL Can be used as an alternative to -l n (see usage).

DEBUG_FILE File used for debug log information, if DEBUGLEVEL greater than 0.

SEMPREFIX Prefix for the lock file. Defaults to /dev/shm/semaphore. which, together with the name of the lock (see usage) attached to the end it, will create a lock file for each name.

LOCKNAME Global lock used for the sem_open (3) call