P-FP: simplify boost_priority()

boost_priority() is only applied to already-scheduled tasks. Remove
the (untested and unneeded) case handling unscheduled tasks, which was
likely not correct anyway.

1 files changed, 4 insertions, 11 deletions

diff --git a/litmus/sched_pfp.c b/litmus/sched_pfp.c
index c425e15c8777..97581f689508 100644
--- a/litmus/sched_pfp.c
+++ b/litmus/sched_pfp.c
@@ -465,17 +465,10 @@ static void boost_priority(struct task_struct* t, lt_t priority_point)
         /* tie-break by protocol-specific priority point */
         tsk_rt(t)->boost_start_time = priority_point;
 
-        if (pfp->scheduled != t) {
-                /* holder may be queued: first stop queue changes */
-                raw_spin_lock(&pfp->domain.release_lock);
-                if (is_queued(t) &&
-                    /* If it is queued, then we need to re-order. */
-                    bheap_decrease(fp_ready_order, tsk_rt(t)->heap_node) &&
-                    /* If we bubbled to the top, then we need to check for preemptions. */
-                    fp_preemption_needed(&pfp->ready_queue, pfp->scheduled))
-                                preempt(pfp);
-                raw_spin_unlock(&pfp->domain.release_lock);
-        } /* else: nothing to do since the job is not queued while scheduled */
+        /* Priority boosting currently only takes effect for already-scheduled
+         * tasks. This is sufficient since priority boosting only kicks in as
+         * part of lock acquisitions. */
+        BUG_ON(pfp->scheduled != t);
 
         raw_spin_unlock_irqrestore(&pfp->slock, flags);
 }