How do Unix systems keep track of shared memory when processes fork()?

Question

Process A creates a shared memory region and shares it with Process B. Then Process A forks and thereby creates Process C. By definition, C now has access to the same shared memory regions. If A dies, C and B still have access to the memory region. Only when C and B also die, the physical memory is free'd again.

How do Unix systems achieve this? Do they store information about which memory is shared with whom per-process or is this information global?

Answer 1

The shared memory region has a reference counter. It gets incremented whenever a process opens the shared memory or gets it by being forked from a process that had it open. It gets decremented when a process closes the shared memory or exits. When the counter goes to 0, the shared memory region is discarded.