Do forked child processes use the same semaphore? - c

Question

Let's say I create a semaphore. If I fork a bunch of child processes, will they all still use that same semaphore?
Also, suppose I create a struct with semaphores inside and forked. Do all the child processes still use that same semaphore? If not, would storing that struct+semaphores in shared memory allow the child processes to use the same semaphores?
I'm really confused about how my forked child processes can use the same semaphores.

Answer 1

Let's say I create a semaphore. If I fork a bunch of child processes, will they all still use that same semaphore?
If you are using a SysV IPC semaphore (semctl), then yes. If you are using POSIX semaphores (sem_init), then yes, but only if you pass a true value for the pshared argument on creation and place it in shared memory.
Also, suppose I create a struct with semaphores inside and forked. Do all the child processes still use that same semaphore? If not, would storing that struct+semaphores in shared memory allow the child processes to use the same semaphores?
What do you mean be 'semaphores inside'? References to SysV IPC semaphores will be shared, because the semaphores don't belong to any process. If you're using POSIX semaphores, or constructing something out of pthreads mutexes and condvars, you will need to use shared memory, and the pshared attribute (pthreads has a pshared attribute for condvars and mutexes as well)
Note that anonymous mmaps created with the MAP_SHARED flag count as (anonymous) shared memory for these purposes, so it's not necessary to actually create a named shared memory segment. Ordinary heap memory will not be shared after a fork.

Answer 2

Let's say I create a semaphore. If I fork a bunch of child processes,
will they all still use that same semaphore?
It depends how you created the semaphore, to do that with an IPC semaphore see semaphore.c: Illustration of simple semaphore passing for an example.
Also, suppose I create a struct with semaphores inside and forked. Do
all the child processes still use that same semaphore? If not, would
storing that struct+semaphores in shared memory allow the child
processes to use the same semaphores?
For that to work your semaphore needs to be stored in an area shared between the parent and the child process like shared memory, and not just created on the stack or on the heap because it will be copied when the process forks.
I'm really confused about how my forked child processes can use the
same semaphores.
The semaphore can be shared across threads or processes. Cross-process sharing is implemented on the operating system level. Two or more different processes can share the same semaphore even if those processes were not created by forking a single parent process.
See this example for sharing an unnamed UNIX semaphore between a parent process and its child (to compile with gcc you'll need the -pthread flag):
#include <semaphore.h>
#include <stdio.h>
#include <errno.h>
#include <stdlib.h>
#include <unistd.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <sys/mman.h>
int main(void)
{
/* place semaphore in shared memory */
sem_t *sema = mmap(NULL, sizeof(*sema),
PROT_READ |PROT_WRITE,MAP_SHARED|MAP_ANONYMOUS,
-1, 0);
if (sema == MAP_FAILED) {
perror("mmap");
exit(EXIT_FAILURE);
}
/* create/initialize semaphore */
if ( sem_init(sema, 1, 0) < 0) {
perror("sem_init");
exit(EXIT_FAILURE);
}
int nloop=10;
int pid = fork();
if (pid < 0) {
perror("fork");
exit(EXIT_FAILURE);
}
if (pid == 0) {
/* child process*/
for (int i = 0; i < nloop; i++) {
printf("child unlocks semaphore: %d\n", i);
if (sem_post(sema) < 0) {
perror("sem_post");
}
sleep(1);
}
if (munmap(sema, sizeof(sema)) < 0) {
perror("munmap");
exit(EXIT_FAILURE);
}
exit(EXIT_SUCCESS);
}
if (pid > 0) {
/* back to parent process */
for (int i = 0; i < nloop; i++) {
printf("parent starts waiting: %d\n", i);
if (sem_wait(sema) < 0) {
perror("sem_wait");
}
printf("parent finished waiting: %d\n", i);
}
if (sem_destroy(sema) < 0) {
perror("sem_destroy failed");
exit(EXIT_FAILURE);
}
if (munmap(sema, sizeof(sema)) < 0) {
perror("munmap failed");
exit(EXIT_FAILURE);
}
exit(EXIT_SUCCESS);
}
}
The output will be:
parent starts waiting: 0
child unlocks semaphore: 0
parent finished waiting: 0
parent starts waiting: 1
child unlocks semaphore: 1
parent finished waiting: 1
...
You may want to read Semaphores in Linux as well, but be aware that the example of UNIX semaphores across fork given doesn't work because the author forgot to use the MAP_ANONYMOUS flag in mmap.

Answer 3

Try this
child and parent would increment the shared variable alternatively
#include <semaphore.h>
#include <stdio.h>
#include <errno.h>
#include <stdlib.h>
#include <unistd.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <sys/mman.h>
struct test {
sem_t mutex1;
sem_t mutex2;
int temp;
}test1;
int main(int argc, char **argv)
{
int fd, i,count=0,nloop=10,zero=0,*ptr;
struct test *testptr;
//open a file and map it into memory
sem_t mutex;
fd = open("log.txt",O_RDWR|O_CREAT,S_IRWXU);
write(fd,&zero,sizeof(int));
ptr = mmap(NULL, sizeof(struct test),PROT_READ |PROT_WRITE,MAP_SHARED,fd,0);
close(fd);
memcpy(ptr, &test1, sizeof(test1));
testptr = (struct test *)ptr;
// testptr = (struct test *)&test1;
/* create, initialize semaphore */
if( sem_init(&(testptr->mutex1),1,1) < 0)
{
perror("semaphore initilization");
exit(0);
}
/* create, initialize semaphore */
if( sem_init(&(testptr->mutex2),1,0) < 0)
{
perror("semaphore initilization");
exit(0);
}
if (fork() == 0) { /* child process*/
for (i = 0; i < nloop; i++) {
sem_wait(&(testptr->mutex2));
printf("child: %d\n", testptr->temp++);
sem_post(&(testptr->mutex1));
}
exit(0);
/* back to parent process */
for (i = 0; i < nloop; i++) {
sem_wait(&testptr->mutex1);
printf("parent: %d\n", testptr->temp++);
sem_post(&(testptr->mutex2));
}
exit(0);
}