Add option to allow single threading of sched domains

Large sched domains can be very expensive to scan. Add an option
SD_SERIALIZE to the sched domain flags. If that flag is set then
we make sure that no other such domain is being balanced.

Signed-off-by: Christoph Lameter <clameter@sgi.com>

Index: linux-2.6.19-rc5-mm1/include/asm-ia64/topology.h
===================================================================
--- linux-2.6.19-rc5-mm1.orig/include/asm-ia64/topology.h	2006-11-14 14:15:37.293272723 -0600
+++ linux-2.6.19-rc5-mm1/include/asm-ia64/topology.h	2006-11-14 14:15:59.412276984 -0600
@@ -101,6 +101,7 @@ void build_cpu_to_node_map(void);
 	.flags			= SD_LOAD_BALANCE	\
 				| SD_BALANCE_EXEC	\
 				| SD_BALANCE_FORK	\
+				| SD_SERIALIZE		\
 				| SD_WAKE_BALANCE,	\
 	.last_balance		= jiffies,		\
 	.balance_interval	= 64,			\
Index: linux-2.6.19-rc5-mm1/include/linux/sched.h
===================================================================
--- linux-2.6.19-rc5-mm1.orig/include/linux/sched.h	2006-11-14 14:15:37.302062505 -0600
+++ linux-2.6.19-rc5-mm1/include/linux/sched.h	2006-11-14 14:15:59.428879915 -0600
@@ -647,6 +647,7 @@ enum idle_type
 #define SD_SHARE_CPUPOWER	128	/* Domain members share cpu power */
 #define SD_POWERSAVINGS_BALANCE	256	/* Balance for power savings */
 #define SD_SHARE_PKG_RESOURCES	512	/* Domain members share cpu pkg resources */
+#define SD_SERIALIZE		1024	/* Only a single load balancing instance */
 
 #define BALANCE_FOR_MC_POWER	\
 	(sched_smt_power_savings ? SD_POWERSAVINGS_BALANCE : 0)
Index: linux-2.6.19-rc5-mm1/include/linux/topology.h
===================================================================
--- linux-2.6.19-rc5-mm1.orig/include/linux/topology.h	2006-11-14 14:15:37.312805572 -0600
+++ linux-2.6.19-rc5-mm1/include/linux/topology.h	2006-11-14 14:15:59.439622988 -0600
@@ -194,7 +194,8 @@
 	.wake_idx		= 0, /* unused */	\
 	.forkexec_idx		= 0, /* unused */	\
 	.per_cpu_gain		= 100,			\
-	.flags			= SD_LOAD_BALANCE,	\
+	.flags			= SD_LOAD_BALANCE	\
+				| SD_SERIALIZE,	\
 	.last_balance		= jiffies,		\
 	.balance_interval	= 64,			\
 	.nr_balance_failed	= 0,			\
Index: linux-2.6.19-rc5-mm1/kernel/sched.c
===================================================================
--- linux-2.6.19-rc5-mm1.orig/kernel/sched.c	2006-11-14 14:15:58.397544937 -0600
+++ linux-2.6.19-rc5-mm1/kernel/sched.c	2006-11-14 14:15:59.485525208 -0600
@@ -2883,6 +2883,7 @@ static void update_load(struct rq *this_
  *
  * Balancing parameters are set up in arch_init_sched_domains.
  */
+static spinlock_t balancing;
 
 static void run_rebalance_domains(struct softirq_action *h)
 {
@@ -2912,6 +2913,11 @@ static void run_rebalance_domains(struct
 		if (unlikely(!interval))
 			interval = 1;
 
+		if (sd->flags & SD_SERIALIZE) {
+			if (!spin_trylock(&balancing))
+				goto out;
+		}
+
 		if (time_after_eq(jiffies, sd->last_balance + interval)) {
 			if (load_balance(this_cpu, this_rq, sd, idle)) {
 				/*
@@ -2923,6 +2929,9 @@ static void run_rebalance_domains(struct
 			}
 			sd->last_balance = jiffies;
 		}
+		if (sd->flags & SD_SERIALIZE)
+			spin_unlock(&balancing);
+out:
 		if (time_after(next_balance, sd->last_balance + interval))
 			next_balance = sd->last_balance + interval;
 	}
@@ -6080,6 +6089,7 @@ static unsigned long long measure_migrat
 				size_found = size;
 			}
 		}
+
 		/*
 		 * Calculate average fluctuation, we use this to prevent
 		 * noise from triggering an early break out of the loop:
@@ -6987,6 +6997,7 @@ void __init sched_init(void)
 
 #ifdef CONFIG_RT_MUTEXES
 	plist_head_init(&init_task.pi_waiters, &init_task.pi_lock);
+	spin_lock_init(&balancing);
 #endif
 
 	/*
Index: linux-2.6.19-rc5-mm1/include/asm-i386/topology.h
===================================================================
--- linux-2.6.19-rc5-mm1.orig/include/asm-i386/topology.h	2006-11-14 14:15:37.322571997 -0600
+++ linux-2.6.19-rc5-mm1/include/asm-i386/topology.h	2006-11-14 14:15:59.524590928 -0600
@@ -89,6 +89,7 @@ static inline int node_to_first_cpu(int 
 	.flags			= SD_LOAD_BALANCE	\
 				| SD_BALANCE_EXEC	\
 				| SD_BALANCE_FORK	\
+				| SD_SERIALIZE		\
 				| SD_WAKE_BALANCE,	\
 	.last_balance		= jiffies,		\
 	.balance_interval	= 1,			\
Index: linux-2.6.19-rc5-mm1/include/asm-powerpc/topology.h
===================================================================
--- linux-2.6.19-rc5-mm1.orig/include/asm-powerpc/topology.h	2006-11-14 14:15:37.332338421 -0600
+++ linux-2.6.19-rc5-mm1/include/asm-powerpc/topology.h	2006-11-14 14:15:59.537287287 -0600
@@ -59,6 +59,7 @@ extern int pcibus_to_node(struct pci_bus
 				| SD_BALANCE_EXEC	\
 				| SD_BALANCE_NEWIDLE	\
 				| SD_WAKE_IDLE		\
+				| SD_SERIALIZE		\
 				| SD_WAKE_BALANCE,	\
 	.last_balance		= jiffies,		\
 	.balance_interval	= 1,			\
Index: linux-2.6.19-rc5-mm1/include/asm-x86_64/topology.h
===================================================================
--- linux-2.6.19-rc5-mm1.orig/include/asm-x86_64/topology.h	2006-11-14 14:15:37.343081488 -0600
+++ linux-2.6.19-rc5-mm1/include/asm-x86_64/topology.h	2006-11-14 14:15:59.549007002 -0600
@@ -47,6 +47,7 @@ extern int __node_distance(int, int);
 	.flags			= SD_LOAD_BALANCE	\
 				| SD_BALANCE_FORK	\
 				| SD_BALANCE_EXEC	\
+				| SD_SERIALIZE		\
 				| SD_WAKE_BALANCE,	\
 	.last_balance		= jiffies,		\
 	.balance_interval	= 1,			\