Neutrino's Blog

There are many ways for processes to communicate on Unix. This post introduces a simple way to use signals. You can first read Beej’s introduction. As the name suggests, a signal is a notification sent and received by processes. For example, when you use a shell, Ctrl-C can interrupt a program because the shell catches the SIGINT signal triggered by Ctrl-C, recognizes it as an interrupt signal, and terminates the program.

To send signals, you can use sigaction() or signal(). I recommend sigaction because it’s newer. For a detailed comparison, see the Stack Overflow thread “What is the difference between sigaction and signal?”.

If you want to trigger a signal, you can use kill() or sigqueue(). The difference is that the latter is Linux-only, but it lets you attach additional information via siginfo. In addition, some system calls can trigger signals as well. For example, if you try to send() to a non-existent socket, you’ll get a SIGPIPE.

Continue reading

When you have many processes and want to implement shared memory to handle shared data, you can use shared memory to build the solution. To create shared memory, you can use mmap or System V shmget. However, according to the Stack Overflow answer “(How to use shared memory with Linux in C)”, shmget is somewhat outdated, while mmap is newer and more flexible.

Shared memory allows us to create a region of memory that can be shared. mmap returns a pointer to that region, with type void *. If we want to put data into it, we can use memcpy to copy objects, strings, or anything else into the shared region. We can also cast the void * directly to an object pointer—this way we create a shared object, and different processes can access the object directly.

Continue reading