sched: wakeup preemption fix

[linux-2.6] / kernel / sched_fair.c

diff --git a/kernel/sched_fair.c b/kernel/sched_fair.c
index c495dcf7031b2ec62200f68bd3fff8f6a87df832..a3badf52bba2352e489c94bc768c9617bc4cb707 100644 (file)
--- a/kernel/sched_fair.c
+++ b/kernel/sched_fair.c
@@ -22,7 +22,7 @@
 
 /*
  * Targeted preemption latency for CPU-bound tasks:
- * (default: 20ms, units: nanoseconds)
+ * (default: 20ms * ilog(ncpus), units: nanoseconds)
  *
  * NOTE: this latency value is not the same as the concept of
  * 'timeslice length' - timeslices in CFS are of variable length
@@ -32,18 +32,18 @@
  * (to see the precise effective timeslice length of your workload,
  *  run vmstat and monitor the context-switches (cs) field)
  */
-const_debug unsigned int sysctl_sched_latency = 20000000ULL;
+unsigned int sysctl_sched_latency = 20000000ULL;
 
 /*
  * Minimal preemption granularity for CPU-bound tasks:
- * (default: 1 msec, units: nanoseconds)
+ * (default: 1 msec * ilog(ncpus), units: nanoseconds)
  */
-const_debug unsigned int sysctl_sched_min_granularity = 1000000ULL;
+unsigned int sysctl_sched_min_granularity = 1000000ULL;
 
 /*
  * is kept at sysctl_sched_latency / sysctl_sched_min_granularity
  */
-const_debug unsigned int sched_nr_latency = 20;
+unsigned int sched_nr_latency = 20;
 
 /*
  * After fork, child runs first. (default) If set to 0 then
@@ -61,23 +61,23 @@ unsigned int __read_mostly sysctl_sched_compat_yield;
 
 /*
  * SCHED_BATCH wake-up granularity.
- * (default: 10 msec, units: nanoseconds)
+ * (default: 10 msec * ilog(ncpus), units: nanoseconds)
  *
  * This option delays the preemption effects of decoupled workloads
  * and reduces their over-scheduling. Synchronous workloads will still
  * have immediate wakeup/sleep latencies.
  */
-const_debug unsigned int sysctl_sched_batch_wakeup_granularity = 10000000UL;
+unsigned int sysctl_sched_batch_wakeup_granularity = 10000000UL;
 
 /*
  * SCHED_OTHER wake-up granularity.
- * (default: 10 msec, units: nanoseconds)
+ * (default: 10 msec * ilog(ncpus), units: nanoseconds)
  *
  * This option delays the preemption effects of decoupled workloads
  * and reduces their over-scheduling. Synchronous workloads will still
  * have immediate wakeup/sleep latencies.
  */
-const_debug unsigned int sysctl_sched_wakeup_granularity = 10000000UL;
+unsigned int sysctl_sched_wakeup_granularity = 10000000UL;
 
 const_debug unsigned int sysctl_sched_migration_cost = 500000UL;
 
@@ -546,7 +546,6 @@ dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int sleep)
 
        update_stats_dequeue(cfs_rq, se);
        if (sleep) {
-               se->peer_preempt = 0;
 #ifdef CONFIG_SCHEDSTATS
                if (entity_is_task(se)) {
                        struct task_struct *tsk = task_of(se);
@@ -574,10 +573,8 @@ check_preempt_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr)
 
        ideal_runtime = sched_slice(cfs_rq, curr);
        delta_exec = curr->sum_exec_runtime - curr->prev_sum_exec_runtime;
-       if (delta_exec > ideal_runtime ||
-                       (sched_feat(PREEMPT_RESTRICT) && curr->peer_preempt))
+       if (delta_exec > ideal_runtime)
                resched_task(rq_of(cfs_rq)->curr);
-       curr->peer_preempt = 0;
 }
 
 static void
@@ -866,12 +863,8 @@ static void check_preempt_wakeup(struct rq *rq, struct task_struct *p)
                if (unlikely(se->load.weight != NICE_0_LOAD))
                        gran = calc_delta_fair(gran, &se->load);
 
-               if (delta > gran) {
-                       int now = !sched_feat(PREEMPT_RESTRICT);
-
-                       if (now || p->prio < curr->prio || !se->peer_preempt++)
-                               resched_task(curr);
-               }
+               if (delta > gran)
+                       resched_task(curr);
        }
 }
 
@@ -1083,7 +1076,6 @@ static void task_new_fair(struct rq *rq, struct task_struct *p)
                swap(curr->vruntime, se->vruntime);
        }
 
-       se->peer_preempt = 0;
        enqueue_task_fair(rq, p, 0);
        resched_task(rq->curr);
 }