My question is: what's the best way to implement mutual exclusion for a shared memory segment. I want to ensure that while updating a shared memory segment that another process will not get read part of the segment before and the other part after the update. (I also want to ensure that the update itself is not clobbered by another update.)
My update looks something like this:
Code:
semkey = ftok(CONFIG_FILE,1);
shmid = shmget(semkey, SHARED_POOL_SIZE, 0444);
if (-1 == shmid) {
/* new shared memory segment -- initialize */
isnew = 1;
shmid = shmget(semkey, SHARED_POOL_SIZE,
IPC_CREAT | IPC_CREAT | SHM_NORESERVE | shm_unlocked );
shmctl(shmid, IPC_STAT, &semstat);
}
else {
/* already exists -- grab pointer to it */
shmctl(shmid, IPC_STAT, &semstat);
stat(CONFIG_FILE, &confstat);
}
/* shmid is ID of our shared memory segment */
if (shmid < 0 || isnew || confstat.st_mtime > semstat.shm_ctime) {
/* update it */
/* parse the file */
config_parse(&config);
pconfig = shmat(shmid, NULL, 0);
/* update the shared segment */
if ((void*)-1 != pconfig) {
/* MUTAL WRITE EXCLUSION NEEDED HERE */
memcpy(pconfig, &config, sizeof(struct config_data));
/* update the timestamp */
shmctl(shmid, IPC_SET, &semstat);
/* END MUTEX */
shmdt(pconfig);
}
}
My reader code looks like this:
Code:
semkey = ftok(CONFIG_FILE,1);
shmid = shmget(semkey,128, 0444);
pconfig = shmat(shmid, NULL, SHM_RDONLY);
if ((void*)-1 == pconfig) {
/* error recovery */
}
else {
/* NEED MUTEX HERE */
memcpy(&config, pconfig, sizeof(struct config_data));
/* END MUTEX */
}
A semaphore with a value of one will make a good mutex. The first process in gets to decrement the semaphore, every other process has to wait. And when its done it increments the semaphore, letting the next process decrement it and take over.
That's a lot of code. I'm thinking that areas provisioned by shmget should already have a built-in mechanism for access control. The man pages say something about permissions, but say nothing about they actually work and whether or not they provide exclusion or not.
That's a lot of code. I'm thinking that areas provisioned by shmget should already have a built-in mechanism for access control. The man pages say something about permissions, but say nothing about they actually work and whether or not they provide exclusion or not.
Permissions describe who's allowed to use them, not when. IPC semaphores are a bit wordy, yes. But then, so is IPC shared mem.
Hello,
I wonder if anybody could help. I'm running SCO Xenix under Qemu on Xubuntu 16.04. I created a SYSV floppy image, but the files on it are poorly displayed when I mounted it under Xenix.
I tried to create and format this image under Ubuntu as well as under Xenix.
How could I create a... (6 Replies)
Hello,
how would i be able to call ps in C programming?
thanks,
---------- Post updated at 01:39 AM ---------- Previous update was at 01:31 AM ----------
here's the complete system call, ps -o pid -p %d, getpit() (2 Replies)
hi, i'm a noob
i have a quuestion: is possible to call and run the bash script by c++ program?
if so, is it posible in grafic? specially Qt ?
thanks (8 Replies)
Hi,
I'm new to UNIX system calls. Can someone share your knowledge as to how exactly system calls should be executed?
Can they be typed like commands such as mkdir on the terminal itself? Also, are there any websites which will show me an example of the output to expect when a system call like... (1 Reply)
Is there any facility to filter/identify the data calls and voice calls coming throug modem?
OR
Can we get the data or voice calls information through a script(preferably C Kermit)? (0 Replies)