[POWERPC] spu sched: static timeslicing for SCHED_RR contexts

For SCHED_RR tasks we can do some really trivial timeslicing.  Basically
we fire up a time for every scheduler tick that searches for a higher
or same priority thread that is on the runqueue and if there is one
context switches to it.  Because we can't lock spus from timer context
we actually run this from a delayed runqueue instead of a timer.

A nice optimization would be to skip the actual priority bitmap search
when there are less contexts than physical spus available.  To implement
this I need a so far unpublished patch from Andre, and it will be added
after we have that patch in.

Note that right now we only do the time slicing for SCHED_RR tasks.
The code would work for SCHED_OTHER tasks aswell, but their prio
value is defered from the one the PPU thread has at time of spu_run,
and using this for spu scheduling decisions would make the code very
unfair.  SCHED_OTHER support will be enabled once we the spu scheduler
knows how to calculcate cpu_context.prio (very soon)

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>

4 files changed, 56 insertions, 3 deletions

diff --git a/arch/powerpc/platforms/cell/spufs/context.c b/arch/powerpc/platforms/cell/spufs/context.c
index d581f4ec99ba..04ad2e364e97 100644
--- a/arch/powerpc/platforms/cell/spufs/context.c
+++ b/arch/powerpc/platforms/cell/spufs/context.c
@@ -54,7 +54,9 @@ struct spu_context *alloc_spu_context(struct spu_gang *gang)
 	if (gang)
 		spu_gang_add_ctx(gang, ctx);
 	ctx->rt_priority = current->rt_priority;
+	ctx->policy = current->policy;
 	ctx->prio = current->prio;
+	INIT_DELAYED_WORK(&ctx->sched_work, spu_sched_tick);
 	goto out;
 out_free:
 	kfree(ctx);
diff --git a/arch/powerpc/platforms/cell/spufs/run.c b/arch/powerpc/platforms/cell/spufs/run.c
index a973e79e9fdc..353a8fa07ab8 100644
--- a/arch/powerpc/platforms/cell/spufs/run.c
+++ b/arch/powerpc/platforms/cell/spufs/run.c
@@ -164,8 +164,10 @@ static inline int spu_run_init(struct spu_context *ctx, u32 * npc)
 			(SPU_RUNCNTL_RUNNABLE | SPU_RUNCNTL_ISOLATE);
 		if (runcntl == 0)
 			runcntl = SPU_RUNCNTL_RUNNABLE;
-	} else
+	} else {
+		spu_start_tick(ctx);
 		ctx->ops->npc_write(ctx, *npc);
+	}
 
 	ctx->ops->runcntl_write(ctx, runcntl);
 	return ret;
@@ -176,6 +178,7 @@ static inline int spu_run_fini(struct spu_context *ctx, u32 * npc,
 {
 	int ret = 0;
 
+	spu_stop_tick(ctx);
 	*status = ctx->ops->status_read(ctx);
 	*npc = ctx->ops->npc_read(ctx);
 	spu_release(ctx);
@@ -329,8 +332,10 @@ long spufs_run_spu(struct file *file, struct spu_context *ctx,
 		}
 		if (unlikely(ctx->state != SPU_STATE_RUNNABLE)) {
 			ret = spu_reacquire_runnable(ctx, npc, &status);
-			if (ret)
+			if (ret) {
+				spu_stop_tick(ctx);
 				goto out2;
+			}
 			continue;
 		}
 		ret = spu_process_events(ctx);
diff --git a/arch/powerpc/platforms/cell/spufs/sched.c b/arch/powerpc/platforms/cell/spufs/sched.c
index ba4b01e01ace..2f25e68b4bac 100644
--- a/arch/powerpc/platforms/cell/spufs/sched.c
+++ b/arch/powerpc/platforms/cell/spufs/sched.c
@@ -44,7 +44,7 @@
 #include <asm/spu_priv1.h>
 #include "spufs.h"
 
-#define SPU_MIN_TIMESLICE 	(100 * HZ / 1000)
+#define SPU_TIMESLICE	(HZ)
 
 struct spu_prio_array {
 	DECLARE_BITMAP(bitmap, MAX_PRIO);
@@ -55,6 +55,7 @@ struct spu_prio_array {
 };
 
 static struct spu_prio_array *spu_prio;
+static struct workqueue_struct *spu_sched_wq;
 
 static inline int node_allowed(int node)
 {
@@ -68,6 +69,40 @@ static inline int node_allowed(int node)
 	return 1;
 }
 
+void spu_start_tick(struct spu_context *ctx)
+{
+	if (ctx->policy == SCHED_RR)
+		queue_delayed_work(spu_sched_wq, &ctx->sched_work, SPU_TIMESLICE);
+}
+
+void spu_stop_tick(struct spu_context *ctx)
+{
+	if (ctx->policy == SCHED_RR)
+		cancel_delayed_work(&ctx->sched_work);
+}
+
+void spu_sched_tick(struct work_struct *work)
+{
+	struct spu_context *ctx =
+		container_of(work, struct spu_context, sched_work.work);
+	struct spu *spu;
+	int rearm = 1;
+
+	mutex_lock(&ctx->state_mutex);
+	spu = ctx->spu;
+	if (spu) {
+		int best = sched_find_first_bit(spu_prio->bitmap);
+		if (best <= ctx->prio) {
+			spu_deactivate(ctx);
+			rearm = 0;
+		}
+	}
+	mutex_unlock(&ctx->state_mutex);
+
+	if (rearm)
+		spu_start_tick(ctx);
+}
+
 /**
  * spu_add_to_active_list - add spu to active list
  * @spu:	spu to add to the active list
@@ -437,10 +472,15 @@ int __init spu_sched_init(void)
 {
 	int i;
 
+	spu_sched_wq = create_singlethread_workqueue("spusched");
+	if (!spu_sched_wq)
+		return 1;
+
 	spu_prio = kzalloc(sizeof(struct spu_prio_array), GFP_KERNEL);
 	if (!spu_prio) {
 		printk(KERN_WARNING "%s: Unable to allocate priority queue.\n",
 		       __FUNCTION__);
+		       destroy_workqueue(spu_sched_wq);
 		return 1;
 	}
 	for (i = 0; i < MAX_PRIO; i++) {
@@ -471,4 +511,5 @@ void __exit spu_sched_exit(void)
 		mutex_unlock(&spu_prio->active_mutex[node]);
 	}
 	kfree(spu_prio);
+	destroy_workqueue(spu_sched_wq);
 }
diff --git a/arch/powerpc/platforms/cell/spufs/spufs.h b/arch/powerpc/platforms/cell/spufs/spufs.h
index 85b182d16464..0c437891dfd5 100644
--- a/arch/powerpc/platforms/cell/spufs/spufs.h
+++ b/arch/powerpc/platforms/cell/spufs/spufs.h
@@ -82,8 +82,10 @@ struct spu_context {
 
 	/* scheduler fields */
  	struct list_head rq;
+	struct delayed_work sched_work;
 	unsigned long sched_flags;
 	unsigned long rt_priority;
+	int policy;
 	int prio;
 };
 
@@ -195,6 +197,9 @@ enum {
 int spu_activate(struct spu_context *ctx, unsigned long flags);
 void spu_deactivate(struct spu_context *ctx);
 void spu_yield(struct spu_context *ctx);
+void spu_start_tick(struct spu_context *ctx);
+void spu_stop_tick(struct spu_context *ctx);
+void spu_sched_tick(struct work_struct *work);
 int __init spu_sched_init(void);
 void __exit spu_sched_exit(void);