18 files changed, 33 insertions, 154 deletions

diff --git a/arch/ia64/kernel/setup.c b/arch/ia64/kernel/setup.c
index 4d9864cc92c9..cf06fe799041 100644
--- a/arch/ia64/kernel/setup.c
+++ b/arch/ia64/kernel/setup.c
@@ -980,15 +980,6 @@ cpu_init (void)
         pm_idle = default_idle;
 }
 
-/*
- * On SMP systems, when the scheduler does migration-cost autodetection,
- * it needs a way to flush as much of the CPU's caches as possible.
- */
-void sched_cacheflush(void)
-{
-        ia64_sal_cache_flush(3);
-}
-
 void __init
 check_bugs (void)
 {
diff --git a/include/asm-alpha/system.h b/include/asm-alpha/system.h
index cf1021a97b2e..620c4d86cbf4 100644
--- a/include/asm-alpha/system.h
+++ b/include/asm-alpha/system.h
@@ -139,16 +139,6 @@ extern void halt(void) __attribute__((noreturn));
 struct task_struct;
 extern struct task_struct *alpha_switch_to(unsigned long, struct task_struct*);
 
-/*
- * On SMP systems, when the scheduler does migration-cost autodetection,
- * it needs a way to flush as much of the CPU's caches as possible.
- *
- * TODO: fill this in!
- */
-static inline void sched_cacheflush(void)
-{
-}
-
 #define imb() \
 __asm__ __volatile__ ("call_pal %0 #imb" : : "i" (PAL_imb) : "memory")
 
diff --git a/include/asm-arm/system.h b/include/asm-arm/system.h
index 6f8e6a69dc5f..94ea8c6dc1a4 100644
--- a/include/asm-arm/system.h
+++ b/include/asm-arm/system.h
@@ -254,16 +254,6 @@ do {									\
         last = __switch_to(prev,task_thread_info(prev), task_thread_info(next));        \
 } while (0)
 
-/*
- * On SMP systems, when the scheduler does migration-cost autodetection,
- * it needs a way to flush as much of the CPU's caches as possible.
- *
- * TODO: fill this in!
- */
-static inline void sched_cacheflush(void)
-{
-}
-
 #if defined(CONFIG_CPU_SA1100) || defined(CONFIG_CPU_SA110)
 /*
  * On the StrongARM, "swp" is terminally broken since it bypasses the
diff --git a/include/asm-arm26/system.h b/include/asm-arm26/system.h
index 4703593b3bb5..e09da5ff1f54 100644
--- a/include/asm-arm26/system.h
+++ b/include/asm-arm26/system.h
@@ -110,16 +110,6 @@ do {									\
 } while (0)
 
 /*
- * On SMP systems, when the scheduler does migration-cost autodetection,
- * it needs a way to flush as much of the CPU's caches as possible.
- *
- * TODO: fill this in!
- */
-static inline void sched_cacheflush(void)
-{
-}
-
-/*
  * Save the current interrupt enable state & disable IRQs
  */
 #define local_irq_save(x)                               \
diff --git a/include/asm-i386/system.h b/include/asm-i386/system.h
index 94ed3686a5f3..609756c61676 100644
--- a/include/asm-i386/system.h
+++ b/include/asm-i386/system.h
@@ -310,15 +310,6 @@ void enable_hlt(void);
 extern int es7000_plat;
 void cpu_idle_wait(void);
 
-/*
- * On SMP systems, when the scheduler does migration-cost autodetection,
- * it needs a way to flush as much of the CPU's caches as possible:
- */
-static inline void sched_cacheflush(void)
-{
-        wbinvd();
-}
-
 extern unsigned long arch_align_stack(unsigned long sp);
 extern void free_init_pages(char *what, unsigned long begin, unsigned long end);
 
diff --git a/include/asm-ia64/system.h b/include/asm-ia64/system.h
index 384fbf7f2a0f..91bb8e00066c 100644
--- a/include/asm-ia64/system.h
+++ b/include/asm-ia64/system.h
@@ -259,7 +259,6 @@ extern void ia64_load_extra (struct task_struct *task);
 #define ia64_platform_is(x) (strcmp(x, platform_name) == 0)
 
 void cpu_idle_wait(void);
-void sched_cacheflush(void);
 
 #define arch_align_stack(x) (x)
 
diff --git a/include/asm-m32r/system.h b/include/asm-m32r/system.h
index 8ee73d3f316d..2365de5c2955 100644
--- a/include/asm-m32r/system.h
+++ b/include/asm-m32r/system.h
@@ -54,16 +54,6 @@
         ); \
 } while(0)
 
-/*
- * On SMP systems, when the scheduler does migration-cost autodetection,
- * it needs a way to flush as much of the CPU's caches as possible.
- *
- * TODO: fill this in!
- */
-static inline void sched_cacheflush(void)
-{
-}
-
 /* Interrupt Control */
 #if !defined(CONFIG_CHIP_M32102) && !defined(CONFIG_CHIP_M32104)
 #define local_irq_enable() \
diff --git a/include/asm-mips/system.h b/include/asm-mips/system.h
index 46bdb3f566f9..76339165bc20 100644
--- a/include/asm-mips/system.h
+++ b/include/asm-mips/system.h
@@ -71,16 +71,6 @@ do {									\
                 write_c0_userlocal(task_thread_info(current)->tp_value);\
 } while(0)
 
-/*
- * On SMP systems, when the scheduler does migration-cost autodetection,
- * it needs a way to flush as much of the CPU's caches as possible.
- *
- * TODO: fill this in!
- */
-static inline void sched_cacheflush(void)
-{
-}
-
 static inline unsigned long __xchg_u32(volatile int * m, unsigned int val)
 {
         __u32 retval;
diff --git a/include/asm-parisc/system.h b/include/asm-parisc/system.h
index 21fbfc5afd02..ee80c920b464 100644
--- a/include/asm-parisc/system.h
+++ b/include/asm-parisc/system.h
@@ -48,17 +48,6 @@ extern struct task_struct *_switch_to(struct task_struct *, struct task_struct *
         (last) = _switch_to(prev, next);                        \
 } while(0)
 
-/*
- * On SMP systems, when the scheduler does migration-cost autodetection,
- * it needs a way to flush as much of the CPU's caches as possible.
- *
- * TODO: fill this in!
- */
-static inline void sched_cacheflush(void)
-{
-}
-
-
 /* interrupt control */
 #define local_save_flags(x)     __asm__ __volatile__("ssm 0, %0" : "=r" (x) : : "memory")
 #define local_irq_disable()     __asm__ __volatile__("rsm %0,%%r0\n" : : "i" (PSW_I) : "memory" )
diff --git a/include/asm-powerpc/system.h b/include/asm-powerpc/system.h
index 32aa42b748be..41520b7a7b76 100644
--- a/include/asm-powerpc/system.h
+++ b/include/asm-powerpc/system.h
@@ -184,16 +184,6 @@ struct thread_struct;
 extern struct task_struct *_switch(struct thread_struct *prev,
                                    struct thread_struct *next);
 
-/*
- * On SMP systems, when the scheduler does migration-cost autodetection,
- * it needs a way to flush as much of the CPU's caches as possible.
- *
- * TODO: fill this in!
- */
-static inline void sched_cacheflush(void)
-{
-}
-
 extern unsigned int rtas_data;
 extern int mem_init_done;       /* set on boot once kmalloc can be called */
 extern unsigned long memory_limit;
diff --git a/include/asm-ppc/system.h b/include/asm-ppc/system.h
index d84a3cf4d033..f1311a8f310f 100644
--- a/include/asm-ppc/system.h
+++ b/include/asm-ppc/system.h
@@ -129,16 +129,6 @@ extern struct task_struct *__switch_to(struct task_struct *,
         struct task_struct *);
 #define switch_to(prev, next, last)     ((last) = __switch_to((prev), (next)))
 
-/*
- * On SMP systems, when the scheduler does migration-cost autodetection,
- * it needs a way to flush as much of the CPU's caches as possible.
- *
- * TODO: fill this in!
- */
-static inline void sched_cacheflush(void)
-{
-}
-
 struct thread_struct;
 extern struct task_struct *_switch(struct thread_struct *prev,
                                    struct thread_struct *next);
diff --git a/include/asm-s390/system.h b/include/asm-s390/system.h
index bbe137c3ed69..64a3cd05cae1 100644
--- a/include/asm-s390/system.h
+++ b/include/asm-s390/system.h
@@ -97,16 +97,6 @@ static inline void restore_access_regs(unsigned int *acrs)
         prev = __switch_to(prev,next);                                       \
 } while (0)
 
-/*
- * On SMP systems, when the scheduler does migration-cost autodetection,
- * it needs a way to flush as much of the CPU's caches as possible.
- *
- * TODO: fill this in!
- */
-static inline void sched_cacheflush(void)
-{
-}
-
 #ifdef CONFIG_VIRT_CPU_ACCOUNTING
 extern void account_vtime(struct task_struct *);
 extern void account_tick_vtime(struct task_struct *);
diff --git a/include/asm-sh/system.h b/include/asm-sh/system.h
index 7c75045ae22b..245042537205 100644
--- a/include/asm-sh/system.h
+++ b/include/asm-sh/system.h
@@ -64,16 +64,6 @@ struct task_struct *__switch_to(struct task_struct *prev,
         last = __last;                                                  \
 } while (0)
 
-/*
- * On SMP systems, when the scheduler does migration-cost autodetection,
- * it needs a way to flush as much of the CPU's caches as possible.
- *
- * TODO: fill this in!
- */
-static inline void sched_cacheflush(void)
-{
-}
-
 #ifdef CONFIG_CPU_SH4A
 #define __icbi()                        \
 {                                       \
diff --git a/include/asm-sparc/system.h b/include/asm-sparc/system.h
index 8b4e23b3bb38..d1a2572e3f55 100644
--- a/include/asm-sparc/system.h
+++ b/include/asm-sparc/system.h
@@ -165,16 +165,6 @@ extern void fpsave(unsigned long *fpregs, unsigned long *fsr,
         } while(0)
 
 /*
- * On SMP systems, when the scheduler does migration-cost autodetection,
- * it needs a way to flush as much of the CPU's caches as possible.
- *
- * TODO: fill this in!
- */
-static inline void sched_cacheflush(void)
-{
-}
-
-/*
  * Changing the IRQ level on the Sparc.
  */
 extern void local_irq_restore(unsigned long);
diff --git a/include/asm-sparc64/system.h b/include/asm-sparc64/system.h
index 8ba380ec6daa..409067408eec 100644
--- a/include/asm-sparc64/system.h
+++ b/include/asm-sparc64/system.h
@@ -204,16 +204,6 @@ do {	if (test_thread_flag(TIF_PERFCTR)) {				\
         }                                                               \
 } while(0)
 
-/*
- * On SMP systems, when the scheduler does migration-cost autodetection,
- * it needs a way to flush as much of the CPU's caches as possible.
- *
- * TODO: fill this in!
- */
-static inline void sched_cacheflush(void)
-{
-}
-
 static inline unsigned long xchg32(__volatile__ unsigned int *m, unsigned int val)
 {
         unsigned long tmp1, tmp2;
diff --git a/include/asm-x86_64/system.h b/include/asm-x86_64/system.h
index ead9f9a56234..e4f246d62c46 100644
--- a/include/asm-x86_64/system.h
+++ b/include/asm-x86_64/system.h
@@ -111,15 +111,6 @@ static inline void write_cr4(unsigned long val)
 #define wbinvd() \
         __asm__ __volatile__ ("wbinvd": : :"memory");
 
-/*
- * On SMP systems, when the scheduler does migration-cost autodetection,
- * it needs a way to flush as much of the CPU's caches as possible.
- */
-static inline void sched_cacheflush(void)
-{
-        wbinvd();
-}
-
 #endif  /* __KERNEL__ */
 
 #define nop() __asm__ __volatile__ ("nop")
diff --git a/include/linux/sched.h b/include/linux/sched.h
index 94f624aef017..33b9b4841ee7 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -1348,6 +1348,13 @@ static inline int set_cpus_allowed(struct task_struct *p, cpumask_t new_mask)
 #endif
 
 extern unsigned long long sched_clock(void);
+
+/*
+ * For kernel-internal use: high-speed (but slightly incorrect) per-cpu
+ * clock constructed from sched_clock():
+ */
+extern unsigned long long cpu_clock(int cpu);
+
 extern unsigned long long
 task_sched_runtime(struct task_struct *task);
 
diff --git a/kernel/sched.c b/kernel/sched.c
index 645256b228c3..93cf241cfbe9 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -379,6 +379,23 @@ static inline unsigned long long rq_clock(struct rq *rq)
 #define task_rq(p)              cpu_rq(task_cpu(p))
 #define cpu_curr(cpu)           (cpu_rq(cpu)->curr)
 
+/*
+ * For kernel-internal use: high-speed (but slightly incorrect) per-cpu
+ * clock constructed from sched_clock():
+ */
+unsigned long long cpu_clock(int cpu)
+{
+        struct rq *rq = cpu_rq(cpu);
+        unsigned long long now;
+        unsigned long flags;
+
+        spin_lock_irqsave(&rq->lock, flags);
+        now = rq_clock(rq);
+        spin_unlock_irqrestore(&rq->lock, flags);
+
+        return now;
+}
+
 #ifdef CONFIG_FAIR_GROUP_SCHED
 /* Change a task's ->cfs_rq if it moves across CPUs */
 static inline void set_task_cfs_rq(struct task_struct *p)
@@ -2235,7 +2252,7 @@ find_busiest_group(struct sched_domain *sd, int this_cpu,
 
                         rq = cpu_rq(i);
 
-                        if (*sd_idle && !idle_cpu(i))
+                        if (*sd_idle && rq->nr_running)
                                 *sd_idle = 0;
 
                         /* Bias balancing toward cpus of our domain */
@@ -2257,9 +2274,11 @@ find_busiest_group(struct sched_domain *sd, int this_cpu,
                 /*
                  * First idle cpu or the first cpu(busiest) in this sched group
                  * is eligible for doing load balancing at this and above
-                 * domains.
+                 * domains. In the newly idle case, we will allow all the cpu's
+                 * to do the newly idle load balance.
                  */
-                if (local_group && balance_cpu != this_cpu && balance) {
+                if (idle != CPU_NEWLY_IDLE && local_group &&
+                    balance_cpu != this_cpu && balance) {
                         *balance = 0;
                         goto ret;
                 }
@@ -2677,6 +2696,7 @@ load_balance_newidle(int this_cpu, struct rq *this_rq, struct sched_domain *sd)
         unsigned long imbalance;
         int nr_moved = 0;
         int sd_idle = 0;
+        int all_pinned = 0;
         cpumask_t cpus = CPU_MASK_ALL;
 
         /*
@@ -2715,10 +2735,11 @@ redo:
                 double_lock_balance(this_rq, busiest);
                 nr_moved = move_tasks(this_rq, this_cpu, busiest,
                                         minus_1_or_zero(busiest->nr_running),
-                                        imbalance, sd, CPU_NEWLY_IDLE, NULL);
+                                        imbalance, sd, CPU_NEWLY_IDLE,
+                                        &all_pinned);
                 spin_unlock(&busiest->lock);
 
-                if (!nr_moved) {
+                if (unlikely(all_pinned)) {
                         cpu_clear(cpu_of(busiest), cpus);
                         if (!cpus_empty(cpus))
                                 goto redo;