--- linux-2.6.1/include/linux/sched.h	2004-01-27 17:02:14.440238704 +1100
+++ linux-2.6.1-ck1/include/linux/sched.h	2004-01-27 17:01:45.778595936 +1100
@@ -126,7 +126,16 @@ extern unsigned long nr_iowait(void);
 #define SCHED_NORMAL		0
 #define SCHED_FIFO		1
 #define SCHED_RR		2
+#define SCHED_BATCH		3
 
+#define SCHED_MIN		0
+#define SCHED_MAX		3
+
+#define SCHED_RANGE(policy)	((policy) >= SCHED_MIN && \
+					(policy) <= SCHED_MAX)
+#define SCHED_RT(policy)	((policy) == SCHED_FIFO || \
+					(policy) == SCHED_RR)
+					
 struct sched_param {
 	int sched_priority;
 };
@@ -284,6 +293,7 @@ struct signal_struct {
 #define MAX_PRIO		(MAX_RT_PRIO + 40)
 
 #define rt_task(p)		((p)->prio < MAX_RT_PRIO)
+#define batch_task(p)		((p)->policy == SCHED_BATCH)
 
 /*
  * Some day this will be a full-fledged user tracking system..
--- linux-2.6.1/kernel/sched.c	2004-01-27 17:02:14.452236880 +1100
+++ linux-2.6.1-ck1/kernel/sched.c	2004-01-27 17:03:20.418208528 +1100
@@ -141,7 +141,7 @@
 		INTERACTIVE_DELTA)
 
 #define TASK_INTERACTIVE(p) \
-	((p)->prio <= (p)->static_prio - DELTA(p))
+	((p)->prio <= (p)->static_prio - DELTA(p) && !batch_task(p))
 
 #define JUST_INTERACTIVE_SLEEP(p) \
 	(JIFFIES_TO_NS(MAX_SLEEP_AVG * \
@@ -153,8 +153,13 @@
 #define LOW_CREDIT(p) \
 	((p)->interactive_credit < -CREDIT_LIMIT)
 
+/*
+ * Batch tasks are preempted by any priority normal tasks and never
+ * preempt other tasks.
+ */
 #define TASK_PREEMPTS_CURR(p, rq) \
-	((p)->prio < (rq)->curr->prio)
+	((((p)->prio < (rq)->curr->prio) || (batch_task((rq)->curr))) && \
+		(!batch_task(p)))
 
 /*
  * BASE_TIMESLICE scales user-nice values [ -20 ... 19 ]
@@ -172,6 +177,13 @@
 
 static inline unsigned int task_timeslice(task_t *p)
 {
+	/*
+	 * Batch tasks get much longer timeslices to optimise cpu throughput.
+	 * Since they yield to any other tasks this is not a problem.
+	 */
+	if (unlikely(batch_task(p)))
+		return BASE_TIMESLICE(p) * 10;
+
 	return BASE_TIMESLICE(p);
 }
 
@@ -199,10 +211,10 @@ struct prio_array {
 struct runqueue {
 	spinlock_t lock;
 	unsigned long nr_running, nr_switches, expired_timestamp,
-			nr_uninterruptible;
+			nr_uninterruptible, nr_batch;
 	task_t *curr, *idle;
 	struct mm_struct *prev_mm;
-	prio_array_t *active, *expired, arrays[2];
+	prio_array_t *active, *expired, *batch, arrays[3];
 	int best_expired_prio, prev_cpu_load[NR_CPUS];
 #ifdef CONFIG_NUMA
 	atomic_t *node_nr_running;
@@ -274,9 +286,11 @@ __init void node_nr_running_init(void)
 
 #else /* !CONFIG_NUMA */
 
-# define nr_running_init(rq)   do { } while (0)
-# define nr_running_inc(rq)    do { (rq)->nr_running++; } while (0)
-# define nr_running_dec(rq)    do { (rq)->nr_running--; } while (0)
+# define nr_running_init(rq)	do { } while (0)
+# define nr_running_inc(rq)	do { (rq)->nr_running++; } while (0)
+# define nr_running_dec(rq)	do { (rq)->nr_running--; } while (0)
+# define nr_batch_inc(rq)	do { (rq)->nr_batch++; } while (0)
+# define nr_batch_dec(rq)	do { (rq)->nr_batch--; } while (0)
 
 #endif /* CONFIG_NUMA */
 
@@ -369,7 +383,10 @@ static int effective_prio(task_t *p)
 	prio = p->static_prio - bonus;
 	if (prio < MAX_RT_PRIO)
 		prio = MAX_RT_PRIO;
-	if (prio > MAX_PRIO-1)
+	if (prio > MAX_PRIO-1 || batch_task(p))
+		/*
+		 * Batch tasks are seen as lowest priority.
+		 */
 		prio = MAX_PRIO-1;
 	return prio;
 }
@@ -379,7 +396,12 @@ static int effective_prio(task_t *p)
  */
 static inline void __activate_task(task_t *p, runqueue_t *rq)
 {
-	enqueue_task(p, rq->active);
+	if (unlikely(batch_task(p))) {
+		enqueue_task(p, rq->batch);
+		nr_batch_inc(rq);
+	} else
+		enqueue_task(p, rq->active);
+
 	nr_running_inc(rq);
 }
 
@@ -503,6 +525,8 @@ static inline void activate_task(task_t 
 static inline void deactivate_task(struct task_struct *p, runqueue_t *rq)
 {
 	nr_running_dec(rq);
+	if (unlikely(batch_task(p)))
+		nr_batch_dec(rq);
 	if (p->state == TASK_UNINTERRUPTIBLE)
 		rq->nr_uninterruptible++;
 	dequeue_task(p, p->array);
@@ -712,8 +736,14 @@ void wake_up_forked_process(task_t * p)
 	else {
 		p->prio = current->prio;
 		list_add_tail(&p->run_list, &current->run_list);
-		p->array = current->array;
-		p->array->nr_active++;
+		if (unlikely(batch_task(p))) {
+			p->array = task_rq(current)->batch;
+			p->array->nr_active++;
+			nr_batch_inc(rq);
+		} else {
+			p->array = current->array;
+			p->array->nr_active++;
+		}
 		nr_running_inc(rq);
 	}
 	task_rq_unlock(rq, &flags);
@@ -1141,7 +1171,12 @@ static inline void pull_task(runqueue_t 
 	nr_running_dec(src_rq);
 	set_task_cpu(p, this_cpu);
 	nr_running_inc(this_rq);
-	enqueue_task(p, this_rq->active);
+	if (unlikely(batch_task(p))) {
+		nr_batch_dec(src_rq);
+		nr_batch_inc(this_rq);
+		enqueue_task(p, this_rq->batch);
+	} else
+		enqueue_task(p, this_rq->active);
 	/*
 	 * Note that idle threads have a prio of MAX_PRIO, for this test
 	 * to be always true for them.
@@ -1403,10 +1438,16 @@ void scheduler_tick(int user_ticks, int 
 			cpustat->iowait += sys_ticks;
 		else
 			cpustat->idle += sys_ticks;
+
+		if (rq->nr_running) {
+			resched_task(p);
+			goto out;
+		}
 		rebalance_tick(rq, 1);
 		return;
 	}
-	if (TASK_NICE(p) > 0)
+
+	if (TASK_NICE(p) > 0 || batch_task(p))
 		cpustat->nice += user_ticks;
 	else
 		cpustat->user += user_ticks;
@@ -1448,14 +1489,21 @@ void scheduler_tick(int user_ticks, int 
 		p->time_slice = task_timeslice(p);
 		p->first_time_slice = 0;
 
-		if (!rq->expired_timestamp)
+		if (!rq->expired_timestamp && !batch_task(p))
 			rq->expired_timestamp = jiffies;
-		if (!TASK_INTERACTIVE(p) || EXPIRED_STARVING(rq)) {
-			enqueue_task(p, rq->expired);
-			if (p->static_prio < rq->best_expired_prio)
-				rq->best_expired_prio = p->static_prio;
-		} else
-			enqueue_task(p, rq->active);
+		if (unlikely(batch_task(p)))
+			/*
+			 * Batch tasks expire to the batch array.
+			 */
+			enqueue_task(p, rq->batch);
+		else {
+			if (!TASK_INTERACTIVE(p) || EXPIRED_STARVING(rq)) {
+				enqueue_task(p, rq->expired);
+				if (p->static_prio < rq->best_expired_prio)
+					rq->best_expired_prio = p->static_prio;
+			} else
+				enqueue_task(p, rq->active);
+		}
 	} else {
 		/*
 		 * Prevent a too long timeslice allowing a task to monopolize
@@ -1553,7 +1601,6 @@ need_resched:
 		else
 			deactivate_task(prev, rq);
 	}
-
 	if (unlikely(!rq->nr_running)) {
 #ifdef CONFIG_SMP
 		load_balance(rq, 1, cpu_to_node_mask(smp_processor_id()));
@@ -1584,10 +1631,20 @@ need_resched:
 		/*
 		 * Switch the active and expired arrays.
 		 */
-		rq->active = rq->expired;
-		rq->expired = array;
+		if (likely(rq->expired->nr_active)) {
+			rq->active = rq->expired;
+			rq->expired = array;
+			rq->expired_timestamp = 0;
+		} else {
+			/*
+			 * Switch to the batch array if there are no
+			 * normal tasks left waiting.
+			 */
+			rq->active = rq->batch;
+			rq->batch = array;
+		}
+
 		array = rq->active;
-		rq->expired_timestamp = 0;
 		rq->best_expired_prio = MAX_PRIO;
 	}
 
@@ -1958,10 +2015,11 @@ void set_user_nice(task_t *p, long nice)
 		enqueue_task(p, array);
 		/*
 		 * If the task increased its priority or is running and
-		 * lowered its priority, then reschedule its CPU:
+		 * lowered its priority or is batch, then reschedule its CPU:
 		 */
-		if (delta < 0 || (delta > 0 && task_running(rq, p)))
-			resched_task(rq->curr);
+		if (delta < 0 || ((delta > 0 || batch_task(p)) &&
+			task_running(rq, p)))
+				resched_task(rq->curr);
 	}
 out_unlock:
 	task_rq_unlock(rq, &flags);
@@ -2098,9 +2156,8 @@ static int setscheduler(pid_t pid, int p
 		policy = p->policy;
 	else {
 		retval = -EINVAL;
-		if (policy != SCHED_FIFO && policy != SCHED_RR &&
-				policy != SCHED_NORMAL)
-			goto out_unlock;
+		if (!SCHED_RANGE(policy))
+				goto out_unlock;
 	}
 
 	/*
@@ -2110,12 +2167,11 @@ static int setscheduler(pid_t pid, int p
 	retval = -EINVAL;
 	if (lp.sched_priority < 0 || lp.sched_priority > MAX_USER_RT_PRIO-1)
 		goto out_unlock;
-	if ((policy == SCHED_NORMAL) != (lp.sched_priority == 0))
-		goto out_unlock;
+	if (!SCHED_RT(policy) != (lp.sched_priority == 0))
+			goto out_unlock;
 
 	retval = -EPERM;
-	if ((policy == SCHED_FIFO || policy == SCHED_RR) &&
-	    !capable(CAP_SYS_NICE))
+	if (SCHED_RT(policy) && !capable(CAP_SYS_NICE))
 		goto out_unlock;
 	if ((current->euid != p->euid) && (current->euid != p->uid) &&
 	    !capable(CAP_SYS_NICE))
@@ -2132,7 +2188,7 @@ static int setscheduler(pid_t pid, int p
 	p->policy = policy;
 	p->rt_priority = lp.sched_priority;
 	oldprio = p->prio;
-	if (policy != SCHED_NORMAL)
+	if (SCHED_RT(policy))
 		p->prio = MAX_USER_RT_PRIO-1 - p->rt_priority;
 	else
 		p->prio = p->static_prio;
@@ -2146,7 +2202,7 @@ static int setscheduler(pid_t pid, int p
 		if (rq->curr == p) {
 			if (p->prio > oldprio)
 				resched_task(rq->curr);
-		} else if (p->prio < rq->curr->prio)
+		} else if (TASK_PREEMPTS_CURR(p, rq))
 			resched_task(rq->curr);
 	}
 
@@ -2352,7 +2408,10 @@ asmlinkage long sys_sched_yield(void)
 	 */
 	if (likely(!rt_task(current))) {
 		dequeue_task(current, array);
-		enqueue_task(current, rq->expired);
+		if (unlikely(batch_task(current)))
+			enqueue_task(current, rq->batch);
+		else
+			enqueue_task(current, rq->expired);
 	} else {
 		list_del(&current->run_list);
 		list_add_tail(&current->run_list, array->queue + current->prio);
@@ -2437,6 +2496,7 @@ asmlinkage long sys_sched_get_priority_m
 		ret = MAX_USER_RT_PRIO-1;
 		break;
 	case SCHED_NORMAL:
+	case SCHED_BATCH:
 		ret = 0;
 		break;
 	}
@@ -2460,6 +2520,7 @@ asmlinkage long sys_sched_get_priority_m
 		ret = 1;
 		break;
 	case SCHED_NORMAL:
+	case SCHED_BATCH:
 		ret = 0;
 	}
 	return ret;
@@ -2893,6 +2954,7 @@ void __init sched_init(void)
 		rq->cpu = (unsigned long)(i);
 		rq->active = rq->arrays;
 		rq->expired = rq->arrays + 1;
+		rq->batch = rq->arrays + 2;
 		rq->best_expired_prio = MAX_PRIO;
 
 		spin_lock_init(&rq->lock);
@@ -2900,7 +2962,7 @@ void __init sched_init(void)
 		atomic_set(&rq->nr_iowait, 0);
 		nr_running_init(rq);
 
-		for (j = 0; j < 2; j++) {
+		for (j = 0; j < 3; j++) {
 			array = rq->arrays + j;
 			for (k = 0; k < MAX_PRIO; k++) {
 				INIT_LIST_HEAD(array->queue + k);