sched: cfs core, kernel/sched_rt.c
authorIngo Molnar <mingo@elte.hu>
Mon, 9 Jul 2007 16:51:58 +0000 (18:51 +0200)
committerIngo Molnar <mingo@elte.hu>
Mon, 9 Jul 2007 16:51:58 +0000 (18:51 +0200)
add kernel/sched_rt.c: SCHED_FIFO/SCHED_RR support. The behavior
and semantics of SCHED_FIFO/SCHED_RR tasks is unchanged.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
kernel/sched_rt.c [new file with mode: 0644]

diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c
new file mode 100644 (file)
index 0000000..1192a27
--- /dev/null
@@ -0,0 +1,255 @@
+/*
+ * Real-Time Scheduling Class (mapped to the SCHED_FIFO and SCHED_RR
+ * policies)
+ */
+
+/*
+ * Update the current task's runtime statistics. Skip current tasks that
+ * are not in our scheduling class.
+ */
+static inline void update_curr_rt(struct rq *rq, u64 now)
+{
+       struct task_struct *curr = rq->curr;
+       u64 delta_exec;
+
+       if (!task_has_rt_policy(curr))
+               return;
+
+       delta_exec = now - curr->se.exec_start;
+       if (unlikely((s64)delta_exec < 0))
+               delta_exec = 0;
+       if (unlikely(delta_exec > curr->se.exec_max))
+               curr->se.exec_max = delta_exec;
+
+       curr->se.sum_exec_runtime += delta_exec;
+       curr->se.exec_start = now;
+}
+
+static void
+enqueue_task_rt(struct rq *rq, struct task_struct *p, int wakeup, u64 now)
+{
+       struct rt_prio_array *array = &rq->rt.active;
+
+       list_add_tail(&p->run_list, array->queue + p->prio);
+       __set_bit(p->prio, array->bitmap);
+}
+
+/*
+ * Adding/removing a task to/from a priority array:
+ */
+static void
+dequeue_task_rt(struct rq *rq, struct task_struct *p, int sleep, u64 now)
+{
+       struct rt_prio_array *array = &rq->rt.active;
+
+       update_curr_rt(rq, now);
+
+       list_del(&p->run_list);
+       if (list_empty(array->queue + p->prio))
+               __clear_bit(p->prio, array->bitmap);
+}
+
+/*
+ * Put task to the end of the run list without the overhead of dequeue
+ * followed by enqueue.
+ */
+static void requeue_task_rt(struct rq *rq, struct task_struct *p)
+{
+       struct rt_prio_array *array = &rq->rt.active;
+
+       list_move_tail(&p->run_list, array->queue + p->prio);
+}
+
+static void
+yield_task_rt(struct rq *rq, struct task_struct *p)
+{
+       requeue_task_rt(rq, p);
+}
+
+/*
+ * Preempt the current task with a newly woken task if needed:
+ */
+static void check_preempt_curr_rt(struct rq *rq, struct task_struct *p)
+{
+       if (p->prio < rq->curr->prio)
+               resched_task(rq->curr);
+}
+
+static struct task_struct *pick_next_task_rt(struct rq *rq, u64 now)
+{
+       struct rt_prio_array *array = &rq->rt.active;
+       struct task_struct *next;
+       struct list_head *queue;
+       int idx;
+
+       idx = sched_find_first_bit(array->bitmap);
+       if (idx >= MAX_RT_PRIO)
+               return NULL;
+
+       queue = array->queue + idx;
+       next = list_entry(queue->next, struct task_struct, run_list);
+
+       next->se.exec_start = now;
+
+       return next;
+}
+
+static void put_prev_task_rt(struct rq *rq, struct task_struct *p, u64 now)
+{
+       update_curr_rt(rq, now);
+       p->se.exec_start = 0;
+}
+
+/*
+ * Load-balancing iterator. Note: while the runqueue stays locked
+ * during the whole iteration, the current task might be
+ * dequeued so the iterator has to be dequeue-safe. Here we
+ * achieve that by always pre-iterating before returning
+ * the current task:
+ */
+static struct task_struct *load_balance_start_rt(void *arg)
+{
+       struct rq *rq = arg;
+       struct rt_prio_array *array = &rq->rt.active;
+       struct list_head *head, *curr;
+       struct task_struct *p;
+       int idx;
+
+       idx = sched_find_first_bit(array->bitmap);
+       if (idx >= MAX_RT_PRIO)
+               return NULL;
+
+       head = array->queue + idx;
+       curr = head->prev;
+
+       p = list_entry(curr, struct task_struct, run_list);
+
+       curr = curr->prev;
+
+       rq->rt.rt_load_balance_idx = idx;
+       rq->rt.rt_load_balance_head = head;
+       rq->rt.rt_load_balance_curr = curr;
+
+       return p;
+}
+
+static struct task_struct *load_balance_next_rt(void *arg)
+{
+       struct rq *rq = arg;
+       struct rt_prio_array *array = &rq->rt.active;
+       struct list_head *head, *curr;
+       struct task_struct *p;
+       int idx;
+
+       idx = rq->rt.rt_load_balance_idx;
+       head = rq->rt.rt_load_balance_head;
+       curr = rq->rt.rt_load_balance_curr;
+
+       /*
+        * If we arrived back to the head again then
+        * iterate to the next queue (if any):
+        */
+       if (unlikely(head == curr)) {
+               int next_idx = find_next_bit(array->bitmap, MAX_RT_PRIO, idx+1);
+
+               if (next_idx >= MAX_RT_PRIO)
+                       return NULL;
+
+               idx = next_idx;
+               head = array->queue + idx;
+               curr = head->prev;
+
+               rq->rt.rt_load_balance_idx = idx;
+               rq->rt.rt_load_balance_head = head;
+       }
+
+       p = list_entry(curr, struct task_struct, run_list);
+
+       curr = curr->prev;
+
+       rq->rt.rt_load_balance_curr = curr;
+
+       return p;
+}
+
+static int
+load_balance_rt(struct rq *this_rq, int this_cpu, struct rq *busiest,
+                       unsigned long max_nr_move, unsigned long max_load_move,
+                       struct sched_domain *sd, enum cpu_idle_type idle,
+                       int *all_pinned, unsigned long *load_moved)
+{
+       int this_best_prio, best_prio, best_prio_seen = 0;
+       int nr_moved;
+       struct rq_iterator rt_rq_iterator;
+
+       best_prio = sched_find_first_bit(busiest->rt.active.bitmap);
+       this_best_prio = sched_find_first_bit(this_rq->rt.active.bitmap);
+
+       /*
+        * Enable handling of the case where there is more than one task
+        * with the best priority.   If the current running task is one
+        * of those with prio==best_prio we know it won't be moved
+        * and therefore it's safe to override the skip (based on load)
+        * of any task we find with that prio.
+        */
+       if (busiest->curr->prio == best_prio)
+               best_prio_seen = 1;
+
+       rt_rq_iterator.start = load_balance_start_rt;
+       rt_rq_iterator.next = load_balance_next_rt;
+       /* pass 'busiest' rq argument into
+        * load_balance_[start|next]_rt iterators
+        */
+       rt_rq_iterator.arg = busiest;
+
+       nr_moved = balance_tasks(this_rq, this_cpu, busiest, max_nr_move,
+                       max_load_move, sd, idle, all_pinned, load_moved,
+                       this_best_prio, best_prio, best_prio_seen,
+                       &rt_rq_iterator);
+
+       return nr_moved;
+}
+
+static void task_tick_rt(struct rq *rq, struct task_struct *p)
+{
+       /*
+        * RR tasks need a special form of timeslice management.
+        * FIFO tasks have no timeslices.
+        */
+       if (p->policy != SCHED_RR)
+               return;
+
+       if (--p->time_slice)
+               return;
+
+       p->time_slice = static_prio_timeslice(p->static_prio);
+       set_tsk_need_resched(p);
+
+       /* put it at the end of the queue: */
+       requeue_task_rt(rq, p);
+}
+
+/*
+ * No parent/child timeslice management necessary for RT tasks,
+ * just activate them:
+ */
+static void task_new_rt(struct rq *rq, struct task_struct *p)
+{
+       activate_task(rq, p, 1);
+}
+
+static struct sched_class rt_sched_class __read_mostly = {
+       .enqueue_task           = enqueue_task_rt,
+       .dequeue_task           = dequeue_task_rt,
+       .yield_task             = yield_task_rt,
+
+       .check_preempt_curr     = check_preempt_curr_rt,
+
+       .pick_next_task         = pick_next_task_rt,
+       .put_prev_task          = put_prev_task_rt,
+
+       .load_balance           = load_balance_rt,
+
+       .task_tick              = task_tick_rt,
+       .task_new               = task_new_rt,
+};