I’ve had to change my approach the past week or so with this independent study. When confronted with the source code, I think I began to feel overwhelmed by the amount of concepts I had to dive into. As a result, I attempted to take it from square one and relearn many systems concepts while also working on understanding the scheduler. As it turned out, this was a bit stressful for me. So I have decided to instead look at the relevant source code and, line by line, take notes learn things as they come. Perhaps this is the traditional way to handle diving into a large system of work, but since I don’t have a large amount of experience in large-scale work this is a learning process for me. It seems that this new direction is working a bit better for me.
Let’s start with /minix/servers/sched/main.c. The main() function is the primary function of the scheduler. When a message is passed into the scheduler, the main() function defines variables for the message, system call number, the caller’s number, and the result of the system call. Then, it enters an infinite loop. This loop saves the message’s info in the variables that were defined for it, and then checks for special situations such as system notifications, etc. (I’m not totally positive on the function of that, but it says that the balance_queues() function is called in this event.) Then, based on the call_nr (the system call number), a switch statement determines what call from /sched/schedule.c should be executed next, with functions like do_noquantum() (which executes when a process is out of quantum) and do_start_scheduling() (which seems to start the scheduling of the process). So long as the process is executed correctly, a reply is sent back that communicates success, and the loop continues on to the next kernel message.
So hopefully soon, I’m actually going to start tinkering with the scheduling policy and see what I can come up with. In the scheduling report I discussed in my previous post, it was suggested that with the current interface a Round Robin, Priority Scheduling, Staircase Scheduler, or Rotating Staircase Deadline algorithm can be easily implemented, so I’ll learn one of those and aim for that. I’m sure it’s going to take me longer than just a week to fully implement, but we’ll see what kind of magic I can work. I’m also sure I’m going to have to write a couple more of these research-related blog posts before I fully understand the workings and can proceed forward, but having changed my plan of attack, I’d like to finish those this week. Ready to move forward once again!
From the blog CS@Worcester – James Blash by jwblash and used with permission of the author. All other rights reserved by the author.