CODEWITHSUNDEEP
cmultithreadingprocessoperating-system
JDS
JDS

Reputation: 16978

Operating systems: implementing process IDs

I'm working on a dummy Unix-like OS (OS161) and I will want to implement processes similar to how Unix does it, meaning fork(), execv(), getpid(), waitpid().

To start, I need to actually have the notion of a process, and give each one its own process ID, some unique integer value to differentiate them. From what I know, each process associates with itself an address space, and at least one thread; possibly more. Let's keep it to one for now.

So my first thought to create a the notion of a process is to give each thread and address space its associated with a unique process ID, and create a global list variable (a hash table? linked list?) containing all the PIDs. In particular, I currently have available:

struct thread; 
struct addr_space;

So I could go into these structures and add a "PID" field as an integer. Alternatively, I was thinking I could just create a wrapper "struct process" and include a thread (or list of threads) and an address space as members.

[Note: currently the thread structure I have has an address space struct as one of its members... Does this mean it's already essentially a process? I can provide more details, for reference I'm working with OS161]

Finally, there's the issue of how I should store the PIDs. I'm pretty sure it has to be a global list structure of some sort (similar to the various queues associated with threads) but not sot sure what would be advisable.

Cheers.

Upvotes: 1

Views: 947

Answers (2)

Jerry Coffin
Jerry Coffin

Reputation: 490098

Rather than a structure containing PIDs, I'd tend to think in terms of a table of process structures, and use indexes into that table as the process IDs.

Upvotes: 1

Jesus Ramos
Jesus Ramos

Reputation: 23268

In Unix systems theres an odd distinction between process and threads as they are both essentially managed by the same structure (task_struct in Linux).

Threads belonging to the same process share a PID and simply use parent/child pointers to track who spawned who for cleanup and creation. Hash table is probably a good option for storing them. As long as your address space keeps appropriate VM mappings then you should be okay with a similar scheme.

Upvotes: 1

Related Questions