sched: remove scheduler locking from arch code

Index: linux-2.6/kernel/sched.c
===================================================================
@@ -269,14 +269,71 @@ static DEFINE_PER_CPU(struct runqueue, r
 #define task_rq(p)		cpu_rq(task_cpu(p))
 #define cpu_curr(cpu)		(cpu_rq(cpu)->curr)
 
-/*
- * Default context-switch locking:
- */
 #ifndef prepare_arch_switch
-# define prepare_arch_switch(rq, next)	do { } while (0)
-# define finish_arch_switch(rq, next)	spin_unlock_irq(&(rq)->lock)
-# define task_running(rq, p)		((rq)->curr == (p))
+# define prepare_arch_switch(next)	do { } while (0)
+#endif
+#ifndef finish_arch_switch
+# define finish_arch_switch(prev)	do { } while (0)
+#endif
+
+#ifndef __ARCH_WANT_UNLOCKED_CTXSW
+static inline int task_running(runqueue_t *rq, task_t *p)
+{
+	return rq->curr == p;
+}
+
+static inline void prepare_lock_switch(runqueue_t *rq, task_t *next)
+{
+}
+
+static inline void finish_lock_switch(runqueue_t *rq, task_t *prev)
+{
+	spin_unlock_irq(&rq->lock);
+}
+
+#else /* __ARCH_WANT_UNLOCKED_CTXSW */
+static inline int task_running(runqueue_t *rq, task_t *p)
+{
+#ifdef CONFIG_SMP
+	return p->oncpu;
+#else
+	return rq->curr == p;
+#endif
+}
+
+static inline void prepare_lock_switch(runqueue_t *rq, task_t *next)
+{
+#ifdef CONFIG_SMP
+	/*
+	 * We can optimise this out completely for !SMP, because the
+	 * SMP rebalancing from interrupt is the only thing that cares
+	 * here.
+	 */
+	next->oncpu = 1;
+#endif
+#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
+	spin_unlock_irq(&rq->lock);
+#else
+	spin_unlock(&rq->lock);
 #endif
+}
+
+static inline void finish_lock_switch(runqueue_t *rq, task_t *prev)
+{
+#ifdef CONFIG_SMP
+	/*
+	 * After ->oncpu is cleared, the task can be moved to a different CPU.
+	 * We must ensure this doesn't happen until the switch is completely
+	 * finished.
+	 */
+	smp_wmb();
+	prev->oncpu = 0;
+#endif
+#ifndef __ARCH_WANT_INTERRUPTS_ON_CTXSW
+	local_irq_enable();
+#endif
+}
+#endif /* __ARCH_WANT_UNLOCKED_CTXSW */
 
 /*
  * task_rq_lock - lock the runqueue a given task resides on and disable

@@ -1197,17 +1254,14 @@ void fastcall sched_fork(task_t *p)
 	p->state = TASK_RUNNING;
 	INIT_LIST_HEAD(&p->run_list);
 	p->array = NULL;
-	spin_lock_init(&p->switch_lock);
 #ifdef CONFIG_SCHEDSTATS
 	memset(&p->sched_info, 0, sizeof(p->sched_info));
 #endif
+#if defined(CONFIG_SMP) && defined(__ARCH_WANT_UNLOCKED_CTXSW)
+	p->oncpu = 0;
+#endif
 #ifdef CONFIG_PREEMPT
-	/*
-	 * During context-switch we hold precisely one spinlock, which
-	 * schedule_tail drops. (in the common case it's this_rq()->lock,
-	 * but it also can be p->switch_lock.) So we compensate with a count
-	 * of 1. Also, we want to start with kernel preemption disabled.
-	 */
+	/* Want to start with kernel preemption disabled. */
 	p->thread_info->preempt_count = 1;
 #endif
 	/*

@@ -1389,22 +1443,40 @@ void fastcall sched_exit(task_t * p)
 }
 
 /**
+ * prepare_task_switch - prepare to switch tasks
+ * @rq: the runqueue preparing to switch
+ * @next: the task we are going to switch to.
+ *
+ * This is called with the rq lock held and interrupts off. It must
+ * be paired with a subsequent finish_task_switch after the context
+ * switch.
+ *
+ * prepare_task_switch sets up locking and calls architecture specific
+ * hooks.
+ */
+static inline void prepare_task_switch(runqueue_t *rq, task_t *next)
+{
+	prepare_lock_switch(rq, next);
+	prepare_arch_switch(next);
+}
+
+/**
  * finish_task_switch - clean up after a task-switch
  * @prev: the thread we just switched away from.
  *
- * We enter this with the runqueue still locked, and finish_arch_switch()
- * will unlock it along with doing any other architecture-specific cleanup
- * actions.
+ * finish_task_switch must be called after the context switch, paired
+ * with a prepare_task_switch call before the context switch.
+ * finish_task_switch will reconcile locking set up by prepare_task_switch,
+ * and do any other architecture-specific cleanup actions.
  *
  * Note that we may have delayed dropping an mm in context_switch(). If
  * so, we finish that here outside of the runqueue lock.  (Doing it
  * with the lock held can cause deadlocks; see schedule() for
  * details.)
  */
-static inline void finish_task_switch(task_t *prev)
+static inline void finish_task_switch(runqueue_t *rq, task_t *prev)
 	__releases(rq->lock)
 {
-	runqueue_t *rq = this_rq();
 	struct mm_struct *mm = rq->prev_mm;
 	unsigned long prev_task_flags;
 

@@ -1422,7 +1494,8 @@ static inline void finish_task_switch(ta
 	 *		Manfred Spraul <[email protected]>
 	 */
 	prev_task_flags = prev->flags;
-	finish_arch_switch(rq, prev);
+	finish_arch_switch(prev);
+	finish_lock_switch(rq, prev);
 	if (mm)
 		mmdrop(mm);
 	if (unlikely(prev_task_flags & PF_DEAD))

@@ -1436,8 +1509,12 @@ static inline void finish_task_switch(ta
 asmlinkage void schedule_tail(task_t *prev)
 	__releases(rq->lock)
 {
-	finish_task_switch(prev);
-
+	runqueue_t *rq = this_rq();
+	finish_task_switch(rq, prev);
+#ifdef __ARCH_WANT_UNLOCKED_CTXSW
+	/* In this case, finish_task_switch does not reenable preemption */
+	preempt_enable();
+#endif
 	if (current->set_child_tid)
 		put_user(current->pid, current->set_child_tid);
 }

@@ -2820,11 +2897,15 @@ switch_tasks:
 		rq->curr = next;
 		++*switch_count;
 
-		prepare_arch_switch(rq, next);
+		prepare_task_switch(rq, next);
 		prev = context_switch(rq, prev, next);
 		barrier();
-
-		finish_task_switch(prev);
+		/*
+		 * this_rq must be evaluated again because prev may have moved
+		 * CPUs since it called schedule(), thus the 'rq' on its stack
+		 * frame will be invalid.
+		 */
+		finish_task_switch(this_rq(), prev);
 	} else
 		spin_unlock_irq(&rq->lock);
 

@@ -4075,6 +4156,9 @@ void __devinit init_idle(task_t *idle, i
 
 	spin_lock_irqsave(&rq->lock, flags);
 	rq->curr = rq->idle = idle;
+#if defined(CONFIG_SMP) && defined(__ARCH_WANT_UNLOCKED_CTXSW)
+	idle->oncpu = 1;
+#endif
 	set_tsk_need_resched(idle);
 	spin_unlock_irqrestore(&rq->lock, flags);
 

Index: linux-2.6/include/linux/sched.h
===================================================================
@@ -357,6 +357,11 @@ struct signal_struct {
 #endif
 };
 
+/* Context switch must be unlocked if interrupts are to be enabled */
+#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
+# define __ARCH_WANT_UNLOCKED_CTXSW
+#endif
+
 /*
  * Bits in flags field of signal_struct.
  */

@@ -582,6 +587,9 @@ struct task_struct {
 
 	int lock_depth;		/* Lock depth */
 
+#if defined(CONFIG_SMP) && defined(__ARCH_WANT_UNLOCKED_CTXSW)
+	int oncpu;
+#endif
 	int prio, static_prio;
 	struct list_head run_list;
 	prio_array_t *array;

@@ -700,8 +708,6 @@ struct task_struct {
 	spinlock_t alloc_lock;
 /* Protection of proc_dentry: nesting proc_lock, dcache_lock, write_lock_irq(&tasklist_lock); */
 	spinlock_t proc_lock;
-/* context-switch lock */
-	spinlock_t switch_lock;
 
 /* journalling filesystem info */
 	void *journal_info;

Index: linux-2.6/include/linux/init_task.h
===================================================================
@@ -74,14 +74,13 @@ extern struct group_info init_groups;
 	.usage		= ATOMIC_INIT(2),				\
 	.flags		= 0,						\
 	.lock_depth	= -1,						\
-	.prio		= MAX_PRIO-20,					\
-	.static_prio	= MAX_PRIO-20,					\
+	.prio		= MAX_PRIO-29,					\
+	.static_prio	= MAX_PRIO-29,					\
 	.policy		= SCHED_NORMAL,					\
 	.cpus_allowed	= CPU_MASK_ALL,					\
 	.mm		= NULL,						\
 	.active_mm	= &init_mm,					\
 	.run_list	= LIST_HEAD_INIT(tsk.run_list),			\
-	.time_slice	= HZ,						\
 	.tasks		= LIST_HEAD_INIT(tsk.tasks),			\
 	.ptrace_children= LIST_HEAD_INIT(tsk.ptrace_children),		\
 	.ptrace_list	= LIST_HEAD_INIT(tsk.ptrace_list),		\

@@ -108,7 +107,6 @@ extern struct group_info init_groups;
 	.blocked	= {{0}},					\
 	.alloc_lock	= SPIN_LOCK_UNLOCKED,				\
 	.proc_lock	= SPIN_LOCK_UNLOCKED,				\
-	.switch_lock	= SPIN_LOCK_UNLOCKED,				\
 	.journal_info	= NULL,						\
 	.cpu_timers	= INIT_CPU_TIMERS(tsk.cpu_timers),		\
 }

Index: linux-2.6/include/asm-ia64/system.h
===================================================================
@@ -183,8 +183,6 @@ do {								\
 
 #ifdef __KERNEL__
 
-#define prepare_to_switch()    do { } while(0)
-
 #ifdef CONFIG_IA32_SUPPORT
 # define IS_IA32_PROCESS(regs)	(ia64_psr(regs)->is != 0)
 #else

@@ -274,13 +272,7 @@ extern void ia64_load_extra (struct task
  * of that CPU which will not be released, because there we wait for the
  * tasklist_lock to become available.
  */
-#define prepare_arch_switch(rq, next)		\
-do {						\
-	spin_lock(&(next)->switch_lock);	\
-	spin_unlock(&(rq)->lock);		\
-} while (0)
-#define finish_arch_switch(rq, prev)	spin_unlock_irq(&(prev)->switch_lock)
-#define task_running(rq, p) 		((rq)->curr == (p) || spin_is_locked(&(p)->switch_lock))
+#define __ARCH_WANT_UNLOCKED_CTXSW
 
 #define ia64_platform_is(x) (strcmp(x, platform_name) == 0)
 

Index: linux-2.6/include/asm-mips/system.h
===================================================================
@@ -422,16 +422,10 @@ extern void __die_if_kernel(const char *
 extern int stop_a_enabled;
 
 /*
- * Taken from include/asm-ia64/system.h; prevents deadlock on SMP
+ * See include/asm-ia64/system.h; prevents deadlock on SMP
  * systems.
  */
-#define prepare_arch_switch(rq, next)		\
-do {						\
-	spin_lock(&(next)->switch_lock);	\
-	spin_unlock(&(rq)->lock);		\
-} while (0)
-#define finish_arch_switch(rq, prev)	spin_unlock_irq(&(prev)->switch_lock)
-#define task_running(rq, p) 		((rq)->curr == (p) || spin_is_locked(&(p)->switch_lock))
+#define __ARCH_WANT_UNLOCKED_CTXSW
 
 #define arch_align_stack(x) (x)
 

Index: linux-2.6/include/asm-s390/system.h
===================================================================
@@ -103,29 +103,18 @@ static inline void restore_access_regs(u
 	prev = __switch_to(prev,next);					     \
 } while (0)
 
-#define prepare_arch_switch(rq, next)	do { } while(0)
-#define task_running(rq, p)		((rq)->curr == (p))
-
 #ifdef CONFIG_VIRT_CPU_ACCOUNTING
 extern void account_user_vtime(struct task_struct *);
 extern void account_system_vtime(struct task_struct *);
-
-#define finish_arch_switch(rq, prev) do {				     \
-	set_fs(current->thread.mm_segment);				     \
-	spin_unlock(&(rq)->lock);					     \
-	account_system_vtime(prev);					     \
-	local_irq_enable();						     \
-} while (0)
-
 #else
+#define account_system_vtime(prev) do { } while (0)
+#endif
 
 #define finish_arch_switch(rq, prev) do {				     \
 	set_fs(current->thread.mm_segment);				     \
-	spin_unlock_irq(&(rq)->lock);					     \
+	account_system_vtime(prev);					     \
 } while (0)
 
-#endif
-
 #define nop() __asm__ __volatile__ ("nop")
 
 #define xchg(ptr,x) \

Index: linux-2.6/include/asm-sparc/system.h
===================================================================
@@ -101,7 +101,7 @@ extern void fpsave(unsigned long *fpregs
  * SWITCH_ENTER and SWITH_DO_LAZY_FPU do not work yet (e.g. SMP does not work)
  * XXX WTF is the above comment? Found in late teen 2.4.x.
  */
-#define prepare_arch_switch(rq, next) do { \
+#define prepare_arch_switch(next) do { \
 	__asm__ __volatile__( \
 	".globl\tflush_patch_switch\nflush_patch_switch:\n\t" \
 	"save %sp, -0x40, %sp; save %sp, -0x40, %sp; save %sp, -0x40, %sp\n\t" \

@@ -109,8 +109,6 @@ extern void fpsave(unsigned long *fpregs
 	"save %sp, -0x40, %sp\n\t" \
 	"restore; restore; restore; restore; restore; restore; restore"); \
 } while(0)
-#define finish_arch_switch(rq, next)	spin_unlock_irq(&(rq)->lock)
-#define task_running(rq, p)		((rq)->curr == (p))
 
 	/* Much care has gone into this code, do not touch it.
 	 *

Index: linux-2.6/include/asm-sparc64/system.h
===================================================================
@@ -139,19 +139,13 @@ extern void __flushw_user(void);
 #define flush_user_windows flushw_user
 #define flush_register_windows flushw_all
 
-#define prepare_arch_switch(rq, next)		\
-do {	spin_lock(&(next)->switch_lock);	\
-	spin_unlock(&(rq)->lock);		\
+/* Don't hold the runqueue lock over context switch */
+#define __ARCH_WANT_UNLOCKED_CTXSW
+#define prepare_arch_switch(next)		\
+do {						\
 	flushw_all();				\
 } while (0)
 
-#define finish_arch_switch(rq, prev)		\
-do {	spin_unlock_irq(&(prev)->switch_lock);	\
-} while (0)
-
-#define task_running(rq, p) \
-	((rq)->curr == (p) || spin_is_locked(&(p)->switch_lock))
-
 	/* See what happens when you design the chip correctly?
 	 *
 	 * We tell gcc we clobber all non-fixed-usage registers except

Index: linux-2.6/include/asm-arm/system.h
===================================================================
@@ -141,34 +141,12 @@ extern unsigned int user_debug;
 #define set_wmb(var, value) do { var = value; wmb(); } while (0)
 #define nop() __asm__ __volatile__("mov\tr0,r0\t@ nop\n\t");
 
-#ifdef CONFIG_SMP
 /*
- * Define our own context switch locking.  This allows us to enable
- * interrupts over the context switch, otherwise we end up with high
- * interrupt latency.  The real problem area is switch_mm() which may
- * do a full cache flush.
+ * switch_mm() may do a full cache flush over the context switch,
+ * so enable interrupts over the context switch to avoid high
+ * latency.
  */
-#define prepare_arch_switch(rq,next)					\
-do {									\
-	spin_lock(&(next)->switch_lock);				\
-	spin_unlock_irq(&(rq)->lock);					\
-} while (0)
-
-#define finish_arch_switch(rq,prev)					\
-	spin_unlock(&(prev)->switch_lock)
-
-#define task_running(rq,p)						\
-	((rq)->curr == (p) || spin_is_locked(&(p)->switch_lock))
-#else
-/*
- * Our UP-case is more simple, but we assume knowledge of how
- * spin_unlock_irq() and friends are implemented.  This avoids
- * us needlessly decrementing and incrementing the preempt count.
- */
-#define prepare_arch_switch(rq,next)	local_irq_enable()
-#define finish_arch_switch(rq,prev)	spin_unlock(&(rq)->lock)
-#define task_running(rq,p)		((rq)->curr == (p))
-#endif
+#define __ARCH_WANT_INTERRUPTS_ON_CTXSW
 
 /*
  * switch_to(prev, next) should switch from task `prev' to `next'

Index: linux-2.6/kernel/sched.c
===================================================================
@@ -796,21 +796,28 @@ static void deactivate_task(struct task_
 #ifdef CONFIG_SMP
 static void resched_task(task_t *p)
 {
-	int need_resched, nrpolling;
+	int cpu;
 
 	assert_spin_locked(&task_rq(p)->lock);
 
-	/* minimise the chance of sending an interrupt to poll_idle() */
-	nrpolling = test_tsk_thread_flag(p,TIF_POLLING_NRFLAG);
-	need_resched = test_and_set_tsk_thread_flag(p,TIF_NEED_RESCHED);
-	nrpolling |= test_tsk_thread_flag(p,TIF_POLLING_NRFLAG);
+	if (test_tsk_thread_flag(p, TIF_NEED_RESCHED))
+		return;
+	
+	set_tsk_thread_flag(p, TIF_NEED_RESCHED);
 
-	if (!need_resched && !nrpolling && (task_cpu(p) != smp_processor_id()))
-		smp_send_reschedule(task_cpu(p));
+	cpu = task_cpu(p);
+	if (cpu == smp_processor_id())
+		return;
+
+	/* NEED_RESCHED must be visible before we test POLLING_NRFLAG */
+	smp_mb();
+	if (!test_tsk_thread_flag(p, TIF_POLLING_NRFLAG))
+		smp_send_reschedule(cpu);
 }
 #else
 static inline void resched_task(task_t *p)
 {
+	assert_spin_locked(&task_rq(p)->lock);
 	set_tsk_need_resched(p);
 }
 #endif

119 lines not on this page · raw message

Index: linux-2.6/init/main.c
===================================================================
@@ -382,7 +382,7 @@ static void noinline rest_init(void)
 	kernel_thread(init, NULL, CLONE_FS | CLONE_SIGHAND);
 	numa_default_policy();
 	unlock_kernel();
-	preempt_enable_no_resched();
+	/* Don't re-enable preemption */
 	cpu_idle();
 } 
 

Index: linux-2.6/arch/i386/kernel/apm.c
===================================================================
@@ -767,8 +767,20 @@ static int set_system_power_state(u_shor
 static int apm_do_idle(void)
 {
 	u32	eax;
+	u8	ret;
+	int	idled = 0;
 
-	if (apm_bios_call_simple(APM_FUNC_IDLE, 0, 0, &eax)) {
+	clear_thread_flag(TIF_POLLING_NRFLAG);
+	smp_mb__after_clear_bit();
+	if (!need_resched()) {
+		idled = 1;
+		ret = apm_bios_call_simple(APM_FUNC_IDLE, 0, 0, &eax);
+	}
+	set_thread_flag(TIF_POLLING_NRFLAG);
+	if (!idled)
+		return 0;
+
+	if (ret) {
 		static unsigned long t;
 
 		/* This always fails on some SMP boards running UP kernels.

Index: linux-2.6/drivers/acpi/processor_idle.c
===================================================================
@@ -162,6 +162,14 @@ acpi_processor_power_activate (
 	return;
 }
 
+static void acpi_safe_halt (void)
+{
+	clear_thread_flag(TIF_POLLING_NRFLAG);
+	smp_mb__after_clear_bit();
+	while (!need_resched())
+		safe_halt();
+	set_thread_flag(TIF_POLLING_NRFLAG);
+}
 
 static void acpi_processor_idle (void)
 {

@@ -171,7 +179,7 @@ static void acpi_processor_idle (void)
 	int			sleep_ticks = 0;
 	u32			t1, t2 = 0;
 
-	pr = processors[_smp_processor_id()];
+	pr = processors[smp_processor_id()];
 	if (!pr)
 		return;
 

@@ -191,8 +199,13 @@ static void acpi_processor_idle (void)
 	}
 
 	cx = pr->power.state;
-	if (!cx)
-		goto easy_out;
+	if (!cx) {
+		if (pm_idle_save)
+			pm_idle_save();
+		else
+			acpi_safe_halt();
+		return;
+	}
 
 	/*
 	 * Check BM Activity

@@ -272,7 +285,8 @@ static void acpi_processor_idle (void)
 		if (pm_idle_save)
 			pm_idle_save();
 		else
-			safe_halt();
+			acpi_safe_halt();
+
 		/*
                  * TBD: Can't get time duration while in C1, as resumes
 		 *      go to an ISR rather than here.  Need to instrument

@@ -384,16 +398,6 @@ end:
 	 */
 	if (next_state != pr->power.state)
 		acpi_processor_power_activate(pr, next_state);
-
-	return;
-
- easy_out:
-	/* do C1 instead of busy loop */
-	if (pm_idle_save)
-		pm_idle_save();
-	else
-		safe_halt();
-	return;
 }
 
 

Index: linux-2.6/arch/i386/kernel/smpboot.c
===================================================================
@@ -414,6 +414,8 @@ static int cpucount;
  */
 static void __init start_secondary(void *unused)
 {
+	preempt_disable();
+
 	/*
 	 * Dont put anything before smp_callin(), SMP
 	 * booting is too fragile that we want to limit the

114 lines not on this page · raw message

261 lines not on this page · raw message

72 lines not on this page · raw message

Index: linux-2.6/arch/ia64/kernel/smpboot.c
===================================================================
@@ -343,6 +343,8 @@ smp_callin (void)
 int __devinit
 start_secondary (void *unused)
 {
+	preempt_disable();
+
 	/* Early console may use I/O ports */
 	ia64_set_kr(IA64_KR_IO_BASE, __pa(ia64_iobase));
 

Index: linux-2.6/arch/ppc64/kernel/smp.c
===================================================================
@@ -561,7 +561,10 @@ int __devinit __cpu_up(unsigned int cpu)
 /* Activate a secondary processor. */
 int __devinit start_secondary(void *unused)
 {
-	unsigned int cpu = smp_processor_id();
+	unsigned int cpu;
+	
+	preempt_disable();
+	cpu = smp_processor_id();
 
 	atomic_inc(&init_mm.mm_count);
 	current->active_mm = &init_mm;