Scheduling policy

[Daniel Baluta]

On Mon, Apr 4, 2011 at 4:13 PM, Laurențiu Dascălu
<dascalu.laurentziu at gmail.com> wrote:
> Hello,
>
> I'm trying to implement a simple scheduling policy in the Linux kernel,
> but I'm not sure if I correctly understand the sched_class interface.
> Specifically, I would like to know more about the following functions:

There's some useful information about sched_class interface
inside [1], check 'Scheduler Classes', page 89.

>
> * put_prev_task(struct rq *rq, struct task_struct *prev)
> Q: It means the prev was preempted and I have to reschedule another
> task? Or I have just to add prev to the running queue?

put_prev_task first announces to the scheduler class that the currently running
task is going to be replaced by another one.

So I guess, that prev was preempted but it is still in the running queue. Now
the scheduler has to dequeue it, and to schedule another task.

>
> The code should look something like:
>
> if (prev->se.on_rq)
> {
>   enqueue_task(rq, prev);
> }
>
> or I'm wrong?

Why do you want to enqueue prev? Isn't rq supposed to hold
the current running tasks?

> * task_tick(struct rq *rq, struct task_struct *p, int queued)
> Q: Should I preempt the task p, by calling resched_task? What does
> "queued" means?


thanks,
Daniel.

[1] Professional Linux Kernel Architecture, W. Mauerer