[PATCH 00/12] sched: cleanup set_task_cpu() issues

[PATCH 00/12] sched: cleanup set_task_cpu() issues

[Peter Zijlstra]

A lot more invasive than I would have liked, but at least it seems to cover all
races I spotted.

[PATCH 01/12] sched: Mark boot-cpu active before smp_init()

[Peter Zijlstra]

[-- Attachment #1: foo12.patch --]
[-- Type: text/plain, Size: 994 bytes --]

A UP machine has 1 active cpu, not having the boot-cpu in the active
map when starting the scheduler confuses things.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
Index: linux-2.6/init/main.c
===================================================================
--- linux-2.6.orig/init/main.c
+++ linux-2.6/init/main.c
@@ -369,12 +369,6 @@ static void __init smp_init(void)
 {
 	unsigned int cpu;
 
-	/*
-	 * Set up the current CPU as possible to migrate to.
-	 * The other ones will be done by cpu_up/cpu_down()
-	 */
-	set_cpu_active(smp_processor_id(), true);
-
 	/* FIXME: This should be done in userspace --RR */
 	for_each_present_cpu(cpu) {
 		if (num_online_cpus() >= setup_max_cpus)
@@ -486,6 +480,7 @@ static void __init boot_cpu_init(void)
 	int cpu = smp_processor_id();
 	/* Mark the boot cpu "present", "online" etc for SMP and UP case */
 	set_cpu_online(cpu, true);
+	set_cpu_active(cpu, true);
 	set_cpu_present(cpu, true);
 	set_cpu_possible(cpu, true);
 }

-- 

[PATCH 02/12] sched: Fix set_cpu_active() in cpu_down()

[Peter Zijlstra]

[-- Attachment #1: foo13.patch --]
[-- Type: text/plain, Size: 2249 bytes --]

Sachin found cpu hotplug test failures on powerpc, which made the
kernel hang on his POWER box. 

The problem is that we fail to re-activate a cpu when a hot-unplug
fails. Fix this by moving the de-activation into _cpu_down after doing
the initial checks.

Remove the synchronize_sched() calls and rely on those implied by
rebuilding the sched domains using the new mask.

Reported-by: Sachin Sant <sachinp@in.ibm.com>
Signed-off-by: Xiaotian Feng <dfeng@redhat.com>
Tested-by: Sachin Sant <sachinp@in.ibm.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
 kernel/cpu.c |   24 +++---------------------
 1 file changed, 3 insertions(+), 21 deletions(-)

Index: linux-2.6/kernel/cpu.c
===================================================================
--- linux-2.6.orig/kernel/cpu.c
+++ linux-2.6/kernel/cpu.c
@@ -209,6 +209,7 @@ static int __ref _cpu_down(unsigned int 
 		return -ENOMEM;
 
 	cpu_hotplug_begin();
+	set_cpu_active(cpu, false);
 	err = __raw_notifier_call_chain(&cpu_chain, CPU_DOWN_PREPARE | mod,
 					hcpu, -1, &nr_calls);
 	if (err == NOTIFY_BAD) {
@@ -280,18 +281,6 @@ int __ref cpu_down(unsigned int cpu)
 		goto out;
 	}
 
-	set_cpu_active(cpu, false);
-
-	/*
-	 * Make sure the all cpus did the reschedule and are not
-	 * using stale version of the cpu_active_mask.
-	 * This is not strictly necessary becuase stop_machine()
-	 * that we run down the line already provides the required
-	 * synchronization. But it's really a side effect and we do not
-	 * want to depend on the innards of the stop_machine here.
-	 */
-	synchronize_sched();
-
 	err = _cpu_down(cpu, 0);
 
 out:
@@ -382,19 +371,12 @@ int disable_nonboot_cpus(void)
 		return error;
 	cpu_maps_update_begin();
 	first_cpu = cpumask_first(cpu_online_mask);
-	/* We take down all of the non-boot CPUs in one shot to avoid races
+	/*
+	 * We take down all of the non-boot CPUs in one shot to avoid races
 	 * with the userspace trying to use the CPU hotplug at the same time
 	 */
 	cpumask_clear(frozen_cpus);
 
-	for_each_online_cpu(cpu) {
-		if (cpu == first_cpu)
-			continue;
-		set_cpu_active(cpu, false);
-	}
-
-	synchronize_sched();
-
 	printk("Disabling non-boot CPUs ...\n");
 	for_each_online_cpu(cpu) {
 		if (cpu == first_cpu)

-- 

[PATCH 03/12] sched: Fix task_hot() test order

[Peter Zijlstra]

[-- Attachment #1: foo10.patch --]
[-- Type: text/plain, Size: 880 bytes --]

Make sure not to access sched_fair fields before verifying it is
indeed a sched_fair task.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
CC: stable@kernel.org
---
 kernel/sched.c |    6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

Index: linux-2.6/kernel/sched.c
===================================================================
--- linux-2.6.orig/kernel/sched.c
+++ linux-2.6/kernel/sched.c
@@ -2044,6 +2044,9 @@ task_hot(struct task_struct *p, u64 now,
 {
 	s64 delta;
 
+	if (p->sched_class != &fair_sched_class)
+		return 0;
+
 	/*
 	 * Buddy candidates are cache hot:
 	 */
@@ -2052,9 +2055,6 @@ task_hot(struct task_struct *p, u64 now,
 			 &p->se == cfs_rq_of(&p->se)->last))
 		return 1;
 
-	if (p->sched_class != &fair_sched_class)
-		return 0;
-
 	if (sysctl_sched_migration_cost == -1)
 		return 1;
 	if (sysctl_sched_migration_cost == 0)

-- 

[PATCH 04/12] sched: select_task_rq_fair() must honour SD_LOAD_BALANCE

[Peter Zijlstra]

[-- Attachment #1: foo7.patch --]
[-- Type: text/plain, Size: 652 bytes --]

We should skip !SD_LOAD_BALANCE domains.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
CC: stable@kernel.org
---
 kernel/sched_fair.c |    3 +++
 1 file changed, 3 insertions(+)

Index: linux-2.6/kernel/sched_fair.c
===================================================================
--- linux-2.6.orig/kernel/sched_fair.c
+++ linux-2.6/kernel/sched_fair.c
@@ -1429,6 +1429,9 @@ static int select_task_rq_fair(struct ta
 	}
 
 	for_each_domain(cpu, tmp) {
+		if (!(tmp->flags & SD_LOAD_BALANCE))
+			continue;
+
 		/*
 		 * If power savings logic is enabled for a domain, see if we
 		 * are not overloaded, if so, don't balance wider.

-- 

[PATCH 05/12] sched: Use TASK_WAKING for fork wakups

[Peter Zijlstra]

[-- Attachment #1: foo11.patch --]
[-- Type: text/plain, Size: 1789 bytes --]

For later convenience use TASK_WAKING for fresh tasks.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
 kernel/sched.c |   18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)

Index: linux-2.6/kernel/sched.c
===================================================================
--- linux-2.6.orig/kernel/sched.c
+++ linux-2.6/kernel/sched.c
@@ -2544,14 +2544,6 @@ static void __sched_fork(struct task_str
 #ifdef CONFIG_PREEMPT_NOTIFIERS
 	INIT_HLIST_HEAD(&p->preempt_notifiers);
 #endif
-
-	/*
-	 * We mark the process as running here, but have not actually
-	 * inserted it onto the runqueue yet. This guarantees that
-	 * nobody will actually run it, and a signal or other external
-	 * event cannot wake it up and insert it on the runqueue either.
-	 */
-	p->state = TASK_RUNNING;
 }
 
 /*
@@ -2562,6 +2554,12 @@ void sched_fork(struct task_struct *p, i
 	int cpu = get_cpu();
 
 	__sched_fork(p);
+	/*
+	 * We mark the process as waking here. This guarantees that
+	 * nobody will actually run it, and a signal or other external
+	 * event cannot wake it up and insert it on the runqueue either.
+	 */
+	p->state = TASK_WAKING;
 
 	/*
 	 * Revert to default priority/policy on fork if requested.
@@ -2630,7 +2628,8 @@ void wake_up_new_task(struct task_struct
 	struct rq *rq;
 
 	rq = task_rq_lock(p, &flags);
-	BUG_ON(p->state != TASK_RUNNING);
+	BUG_ON(p->state != TASK_WAKING);
+	p->state = TASK_RUNNING;
 	update_rq_clock(rq);
 	activate_task(rq, p, 0);
 	trace_sched_wakeup_new(rq, p, 1);
@@ -6992,6 +6991,7 @@ void __cpuinit init_idle(struct task_str
 	raw_spin_lock_irqsave(&rq->lock, flags);
 
 	__sched_fork(idle);
+	idle->state = TASK_RUNNING;
 	idle->se.exec_start = sched_clock();
 
 	cpumask_copy(&idle->cpus_allowed, cpumask_of(cpu));

-- 

[PATCH 06/12] sched: Ensure set_task_cpu() is never called on blocked tasks

[Peter Zijlstra]

[-- Attachment #1: foo4.patch --]
[-- Type: text/plain, Size: 5156 bytes --]

In order to clean up the set_task_cpu() rq dependencies we need to
ensure it is never called on blocked tasks because such usage does not
pair with consistent rq->lock usage.

This puts the migration burden on ttwu().

Furthermore we need to close a race against changing ->cpus_allowed,
since select_task_rq() runs with only preemption disabled.

For sched_fork() this is safe because the child isn't in the tasklist
yet, for wakeup we fix this by synchronizing set_cpus_allowed_ptr()
against TASK_WAKING, which leaves sched_exec to be a problem

This also closes a hole in (6ad4c1888 sched: Fix balance vs hotplug
race) where ->select_task_rq() doesn't validate the result against the
sched_domain/root_domain.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
 kernel/sched.c |   85 ++++++++++++++++++++++++++++++++++++++++++++-------------
 1 file changed, 66 insertions(+), 19 deletions(-)

Index: linux-2.6/kernel/sched.c
===================================================================
--- linux-2.6.orig/kernel/sched.c
+++ linux-2.6/kernel/sched.c
@@ -2018,22 +2018,15 @@ static inline void check_class_changed(s
  */
 void kthread_bind(struct task_struct *p, unsigned int cpu)
 {
-	struct rq *rq = cpu_rq(cpu);
-	unsigned long flags;
-
 	/* Must have done schedule() in kthread() before we set_task_cpu */
 	if (!wait_task_inactive(p, TASK_UNINTERRUPTIBLE)) {
 		WARN_ON(1);
 		return;
 	}
 
-	raw_spin_lock_irqsave(&rq->lock, flags);
-	update_rq_clock(rq);
-	set_task_cpu(p, cpu);
 	p->cpus_allowed = cpumask_of_cpu(cpu);
 	p->rt.nr_cpus_allowed = 1;
 	p->flags |= PF_THREAD_BOUND;
-	raw_spin_unlock_irqrestore(&rq->lock, flags);
 }
 EXPORT_SYMBOL(kthread_bind);
 
@@ -2074,6 +2067,14 @@ void set_task_cpu(struct task_struct *p,
 	struct cfs_rq *old_cfsrq = task_cfs_rq(p),
 		      *new_cfsrq = cpu_cfs_rq(old_cfsrq, new_cpu);
 
+#ifdef CONFIG_SCHED_DEBUG
+	/*
+	 * We should never call set_task_cpu() on a blocked task,
+	 * ttwu() will sort out the placement.
+	 */
+	WARN_ON(p->state != TASK_RUNNING && p->state != TASK_WAKING);
+#endif
+
 	trace_sched_migrate_task(p, new_cpu);
 
 	if (old_cpu != new_cpu) {
@@ -2107,13 +2108,10 @@ migrate_task(struct task_struct *p, int 
 
 	/*
 	 * If the task is not on a runqueue (and not running), then
-	 * it is sufficient to simply update the task's cpu field.
+	 * the next wake-up will properly place the task.
 	 */
-	if (!p->se.on_rq && !task_running(rq, p)) {
-		update_rq_clock(rq);
-		set_task_cpu(p, dest_cpu);
+	if (!p->se.on_rq && !task_running(rq, p))
 		return 0;
-	}
 
 	init_completion(&req->done);
 	req->task = p;
@@ -2319,10 +2317,42 @@ void task_oncpu_function_call(struct tas
 }
 
 #ifdef CONFIG_SMP
+/*
+ * Called from:
+ *
+ *  - fork, @p is stable because it isn't on the tasklist yet
+ *
+ *  - exec, @p is unstable XXX
+ *
+ *  - wake-up, we serialize ->cpus_allowed against TASK_WAKING so
+ *             we should be good.
+ */
 static inline
 int select_task_rq(struct task_struct *p, int sd_flags, int wake_flags)
 {
-	return p->sched_class->select_task_rq(p, sd_flags, wake_flags);
+	int cpu = p->sched_class->select_task_rq(p, sd_flags, wake_flags);
+
+	/*
+	 * In order not to call set_task_cpu() on a blocking task we need
+	 * to rely on ttwu() to place the task on a valid ->cpus_allowed
+	 * cpu.
+	 *
+	 * Since this is common to all placement strategies, this lives here.
+	 *
+	 * [ this allows ->select_task() to simply return task_cpu(p) and
+	 *   not worry about this generic constraint ]
+	 */
+	if (unlikely(!cpumask_test_cpu(cpu, &p->cpus_allowed) ||
+		     !cpu_active(cpu))) {
+
+		cpu = cpumask_any_and(&p->cpus_allowed, cpu_active_mask);
+		/*
+		 * XXX: race against hot-plug modifying cpu_active_mask
+		 */
+		BUG_ON(cpu >= nr_cpu_ids);
+	}
+
+	return cpu;
 }
 #endif
 
@@ -7102,7 +7132,23 @@ int set_cpus_allowed_ptr(struct task_str
 	struct rq *rq;
 	int ret = 0;
 
+	/*
+	 * Since we rely on wake-ups to migrate sleeping tasks, don't change
+	 * the ->cpus_allowed mask from under waking tasks, which would be
+	 * possible when we change rq->lock in ttwu(), so synchronize against
+	 * TASK_WAKING to avoid that.
+	 */
+again:
+	while (p->state == TASK_WAKING)
+		cpu_relax();
+
 	rq = task_rq_lock(p, &flags);
+
+	if (p->state == TASK_WAKING) {
+		task_rq_unlock(rq, &flags);
+		goto again;
+	}
+
 	if (!cpumask_intersects(new_mask, cpu_active_mask)) {
 		ret = -EINVAL;
 		goto out;
@@ -7158,7 +7204,7 @@ EXPORT_SYMBOL_GPL(set_cpus_allowed_ptr);
 static int __migrate_task(struct task_struct *p, int src_cpu, int dest_cpu)
 {
 	struct rq *rq_dest, *rq_src;
-	int ret = 0, on_rq;
+	int ret = 0;
 
 	if (unlikely(!cpu_active(dest_cpu)))
 		return ret;
@@ -7174,12 +7220,13 @@ static int __migrate_task(struct task_st
 	if (!cpumask_test_cpu(dest_cpu, &p->cpus_allowed))
 		goto fail;
 
-	on_rq = p->se.on_rq;
-	if (on_rq)
+	/*
+	 * If we're not on a rq, the next wake-up will ensure we're
+	 * placed properly.
+	 */
+	if (p->se.on_rq) {
 		deactivate_task(rq_src, p, 0);
-
-	set_task_cpu(p, dest_cpu);
-	if (on_rq) {
+		set_task_cpu(p, dest_cpu);
 		activate_task(rq_dest, p, 0);
 		check_preempt_curr(rq_dest, p, 0);
 	}

-- 

[PATCH 07/12] sched: Fix sched_exec() balancing

[Peter Zijlstra]

[-- Attachment #1: foo6.patch --]
[-- Type: text/plain, Size: 2737 bytes --]

Since we access ->cpus_allowed without holding rq->lock we need a
retry loop to validate the result, this comes for near free when we
merge sched_migrate_task() into sched_exec() since that already does
the needed check.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
 kernel/sched.c |   45 +++++++++++++++++++++++----------------------
 1 file changed, 23 insertions(+), 22 deletions(-)

Index: linux-2.6/kernel/sched.c
===================================================================
--- linux-2.6.orig/kernel/sched.c
+++ linux-2.6/kernel/sched.c
@@ -2322,7 +2322,7 @@ void task_oncpu_function_call(struct tas
  *
  *  - fork, @p is stable because it isn't on the tasklist yet
  *
- *  - exec, @p is unstable XXX
+ *  - exec, @p is unstable, retry loop
  *
  *  - wake-up, we serialize ->cpus_allowed against TASK_WAKING so
  *             we should be good.
@@ -3132,21 +3132,36 @@ static void double_rq_unlock(struct rq *
 }
 
 /*
- * If dest_cpu is allowed for this process, migrate the task to it.
- * This is accomplished by forcing the cpu_allowed mask to only
- * allow dest_cpu, which will force the cpu onto dest_cpu. Then
- * the cpu_allowed mask is restored.
+ * sched_exec - execve() is a valuable balancing opportunity, because at
+ * this point the task has the smallest effective memory and cache footprint.
  */
-static void sched_migrate_task(struct task_struct *p, int dest_cpu)
+void sched_exec(void)
 {
+	struct task_struct *p = current;
 	struct migration_req req;
+	int dest_cpu, this_cpu;
 	unsigned long flags;
 	struct rq *rq;
 
+again:
+	this_cpu = get_cpu();
+	dest_cpu = select_task_rq(p, SD_BALANCE_EXEC, 0);
+	if (dest_cpu == this_cpu) {
+		put_cpu();
+		return;
+	}
+
 	rq = task_rq_lock(p, &flags);
+	put_cpu();
+
+	/*
+	 * select_task_rq() can race against ->cpus_allowed
+	 */
 	if (!cpumask_test_cpu(dest_cpu, &p->cpus_allowed)
-	    || unlikely(!cpu_active(dest_cpu)))
-		goto out;
+	    || unlikely(!cpu_active(dest_cpu))) {
+		task_rq_unlock(rq, &flags);
+		goto again;
+	}
 
 	/* force the process onto the specified CPU */
 	if (migrate_task(p, dest_cpu, &req)) {
@@ -3161,24 +3176,10 @@ static void sched_migrate_task(struct ta
 
 		return;
 	}
-out:
 	task_rq_unlock(rq, &flags);
 }
 
 /*
- * sched_exec - execve() is a valuable balancing opportunity, because at
- * this point the task has the smallest effective memory and cache footprint.
- */
-void sched_exec(void)
-{
-	int new_cpu, this_cpu = get_cpu();
-	new_cpu = select_task_rq(current, SD_BALANCE_EXEC, 0);
-	put_cpu();
-	if (new_cpu != this_cpu)
-		sched_migrate_task(current, new_cpu);
-}
-
-/*
  * pull_task - move a task from a remote runqueue to the local runqueue.
  * Both runqueues must be locked.
  */

-- 

[PATCH 08/12] sched: Fix select_task_rq() vs hotplug issues

[Peter Zijlstra]

[-- Attachment #1: foo8.patch --]
[-- Type: text/plain, Size: 3519 bytes --]

Since select_task_rq() is now responsible for guaranteeing
->cpus_allowed and cpu_active_mask, we need to verify this.

select_task_rq_rt() can blindly return smp_processor_id()/task_cpu()
without checking the valid masks, select_task_rq_fair() can do the
same in the rare case that all SD_flags are disabled.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
 kernel/sched.c |   75 ++++++++++++++++++++++++++++++---------------------------
 1 file changed, 40 insertions(+), 35 deletions(-)

Index: linux-2.6/kernel/sched.c
===================================================================
--- linux-2.6.orig/kernel/sched.c
+++ linux-2.6/kernel/sched.c
@@ -2317,6 +2317,43 @@ void task_oncpu_function_call(struct tas
 }
 
 #ifdef CONFIG_SMP
+static int select_fallback_rq(int cpu, struct task_struct *p)
+{
+	int dest_cpu;
+	const struct cpumask *nodemask = cpumask_of_node(cpu_to_node(cpu));
+
+	/* Look for allowed, online CPU in same node. */
+	for_each_cpu_and(dest_cpu, nodemask, cpu_active_mask)
+		if (cpumask_test_cpu(dest_cpu, &p->cpus_allowed))
+			return dest_cpu;
+
+	/* Any allowed, online CPU? */
+	dest_cpu = cpumask_any_and(&p->cpus_allowed, cpu_active_mask);
+	if (dest_cpu < nr_cpu_ids)
+		return dest_cpu;
+
+	/* No more Mr. Nice Guy. */
+	if (dest_cpu >= nr_cpu_ids) {
+		rcu_read_lock();
+		cpuset_cpus_allowed_locked(p, &p->cpus_allowed);
+		rcu_read_unlock();
+		dest_cpu = cpumask_any_and(cpu_active_mask, &p->cpus_allowed);
+
+		/*
+		 * Don't tell them about moving exiting tasks or
+		 * kernel threads (both mm NULL), since they never
+		 * leave kernel.
+		 */
+		if (p->mm && printk_ratelimit()) {
+			printk(KERN_INFO "process %d (%s) no "
+			       "longer affine to cpu%d\n",
+			       task_pid_nr(p), p->comm, cpu);
+		}
+	}
+
+	return dest_cpu;
+}
+
 /*
  * Called from:
  *
@@ -2343,14 +2380,8 @@ int select_task_rq(struct task_struct *p
 	 *   not worry about this generic constraint ]
 	 */
 	if (unlikely(!cpumask_test_cpu(cpu, &p->cpus_allowed) ||
-		     !cpu_active(cpu))) {
-
-		cpu = cpumask_any_and(&p->cpus_allowed, cpu_active_mask);
-		/*
-		 * XXX: race against hot-plug modifying cpu_active_mask
-		 */
-		BUG_ON(cpu >= nr_cpu_ids);
-	}
+		     !cpu_active(cpu)))
+		cpu = select_fallback_rq(task_cpu(p), p);
 
 	return cpu;
 }
@@ -7319,36 +7350,10 @@ static int __migrate_task_irq(struct tas
 static void move_task_off_dead_cpu(int dead_cpu, struct task_struct *p)
 {
 	int dest_cpu;
-	const struct cpumask *nodemask = cpumask_of_node(cpu_to_node(dead_cpu));
 
 again:
-	/* Look for allowed, online CPU in same node. */
-	for_each_cpu_and(dest_cpu, nodemask, cpu_active_mask)
-		if (cpumask_test_cpu(dest_cpu, &p->cpus_allowed))
-			goto move;
-
-	/* Any allowed, online CPU? */
-	dest_cpu = cpumask_any_and(&p->cpus_allowed, cpu_active_mask);
-	if (dest_cpu < nr_cpu_ids)
-		goto move;
-
-	/* No more Mr. Nice Guy. */
-	if (dest_cpu >= nr_cpu_ids) {
-		cpuset_cpus_allowed_locked(p, &p->cpus_allowed);
-		dest_cpu = cpumask_any_and(cpu_active_mask, &p->cpus_allowed);
-
-		/*
-		 * Don't tell them about moving exiting tasks or
-		 * kernel threads (both mm NULL), since they never
-		 * leave kernel.
-		 */
-		if (p->mm && printk_ratelimit()) {
-			pr_info("process %d (%s) no longer affine to cpu%d\n",
-				task_pid_nr(p), p->comm, dead_cpu);
-		}
-	}
+	dest_cpu = select_fallback_rq(dead_cpu, p);
 
-move:
 	/* It can have affinity changed while we were choosing. */
 	if (unlikely(!__migrate_task_irq(p, dead_cpu, dest_cpu)))
 		goto again;

-- 

[PATCH 09/12] sched: Move kthread_bind() back to kthread.c

[Peter Zijlstra]

[-- Attachment #1: foo5.patch --]
[-- Type: text/plain, Size: 2631 bytes --]

Since kthread_bind() lost its dependencies on sched.c, move it back
where it came from.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
 kernel/kthread.c |   23 +++++++++++++++++++++++
 kernel/sched.c   |   26 --------------------------
 2 files changed, 23 insertions(+), 26 deletions(-)

Index: linux-2.6/kernel/kthread.c
===================================================================
--- linux-2.6.orig/kernel/kthread.c
+++ linux-2.6/kernel/kthread.c
@@ -150,6 +150,29 @@ struct task_struct *kthread_create(int (
 EXPORT_SYMBOL(kthread_create);
 
 /**
+ * kthread_bind - bind a just-created kthread to a cpu.
+ * @p: thread created by kthread_create().
+ * @cpu: cpu (might not be online, must be possible) for @k to run on.
+ *
+ * Description: This function is equivalent to set_cpus_allowed(),
+ * except that @cpu doesn't need to be online, and the thread must be
+ * stopped (i.e., just returned from kthread_create()).
+ */
+void kthread_bind(struct task_struct *p, unsigned int cpu)
+{
+	/* Must have done schedule() in kthread() before we set_task_cpu */
+	if (!wait_task_inactive(p, TASK_UNINTERRUPTIBLE)) {
+		WARN_ON(1);
+		return;
+	}
+
+	p->cpus_allowed = cpumask_of_cpu(cpu);
+	p->rt.nr_cpus_allowed = 1;
+	p->flags |= PF_THREAD_BOUND;
+}
+EXPORT_SYMBOL(kthread_bind);
+
+/**
  * kthread_stop - stop a thread created by kthread_create().
  * @k: thread created by kthread_create().
  *
Index: linux-2.6/kernel/sched.c
===================================================================
--- linux-2.6.orig/kernel/sched.c
+++ linux-2.6/kernel/sched.c
@@ -2002,32 +2002,6 @@ static inline void check_class_changed(s
 		p->sched_class->prio_changed(rq, p, oldprio, running);
 }
 
-/**
- * kthread_bind - bind a just-created kthread to a cpu.
- * @p: thread created by kthread_create().
- * @cpu: cpu (might not be online, must be possible) for @k to run on.
- *
- * Description: This function is equivalent to set_cpus_allowed(),
- * except that @cpu doesn't need to be online, and the thread must be
- * stopped (i.e., just returned from kthread_create()).
- *
- * Function lives here instead of kthread.c because it messes with
- * scheduler internals which require locking.
- */
-void kthread_bind(struct task_struct *p, unsigned int cpu)
-{
-	/* Must have done schedule() in kthread() before we set_task_cpu */
-	if (!wait_task_inactive(p, TASK_UNINTERRUPTIBLE)) {
-		WARN_ON(1);
-		return;
-	}
-
-	p->cpus_allowed = cpumask_of_cpu(cpu);
-	p->rt.nr_cpus_allowed = 1;
-	p->flags |= PF_THREAD_BOUND;
-}
-EXPORT_SYMBOL(kthread_bind);
-
 #ifdef CONFIG_SMP
 /*
  * Is this task likely cache-hot:

-- 

[PATCH 10/12] sched: Add pre and post wakeup hooks

[Peter Zijlstra]

[-- Attachment #1: foo0.patch --]
[-- Type: text/plain, Size: 3043 bytes --]

As will be apparent in the next patch, we need a pre wakeup hook for
sched_fair task migration, hence rename the post wakeup hook and one pre
wakeup.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
 include/linux/sched.h |    3 ++-
 kernel/sched.c        |   12 ++++++++----
 kernel/sched_rt.c     |    4 ++--
 3 files changed, 12 insertions(+), 7 deletions(-)

Index: linux-2.6/include/linux/sched.h
===================================================================
--- linux-2.6.orig/include/linux/sched.h
+++ linux-2.6/include/linux/sched.h
@@ -1091,7 +1091,8 @@ struct sched_class {
 			      enum cpu_idle_type idle);
 	void (*pre_schedule) (struct rq *this_rq, struct task_struct *task);
 	void (*post_schedule) (struct rq *this_rq);
-	void (*task_wake_up) (struct rq *this_rq, struct task_struct *task);
+	void (*task_waking) (struct rq *this_rq, struct task_struct *task);
+	void (*task_woken) (struct rq *this_rq, struct task_struct *task);
 
 	void (*set_cpus_allowed)(struct task_struct *p,
 				 const struct cpumask *newmask);
Index: linux-2.6/kernel/sched.c
===================================================================
--- linux-2.6.orig/kernel/sched.c
+++ linux-2.6/kernel/sched.c
@@ -2380,6 +2380,10 @@ static int try_to_wake_up(struct task_st
 	if (task_contributes_to_load(p))
 		rq->nr_uninterruptible--;
 	p->state = TASK_WAKING;
+
+	if (p->sched_class->task_waking)
+		p->sched_class->task_waking(rq, p);
+
 	__task_rq_unlock(rq);
 
 	cpu = select_task_rq(p, SD_BALANCE_WAKE, wake_flags);
@@ -2443,8 +2447,8 @@ out_running:
 
 	p->state = TASK_RUNNING;
 #ifdef CONFIG_SMP
-	if (p->sched_class->task_wake_up)
-		p->sched_class->task_wake_up(rq, p);
+	if (p->sched_class->task_woken)
+		p->sched_class->task_woken(rq, p);
 
 	if (unlikely(rq->idle_stamp)) {
 		u64 delta = rq->clock - rq->idle_stamp;
@@ -2636,8 +2640,8 @@ void wake_up_new_task(struct task_struct
 	trace_sched_wakeup_new(rq, p, 1);
 	check_preempt_curr(rq, p, WF_FORK);
 #ifdef CONFIG_SMP
-	if (p->sched_class->task_wake_up)
-		p->sched_class->task_wake_up(rq, p);
+	if (p->sched_class->task_woken)
+		p->sched_class->task_woken(rq, p);
 #endif
 	task_rq_unlock(rq, &flags);
 }
Index: linux-2.6/kernel/sched_rt.c
===================================================================
--- linux-2.6.orig/kernel/sched_rt.c
+++ linux-2.6/kernel/sched_rt.c
@@ -1472,7 +1472,7 @@ static void post_schedule_rt(struct rq *
  * If we are not running and we are not going to reschedule soon, we should
  * try to push tasks away now
  */
-static void task_wake_up_rt(struct rq *rq, struct task_struct *p)
+static void task_woken_rt(struct rq *rq, struct task_struct *p)
 {
 	if (!task_running(rq, p) &&
 	    !test_tsk_need_resched(rq->curr) &&
@@ -1753,7 +1753,7 @@ static const struct sched_class rt_sched
 	.rq_offline             = rq_offline_rt,
 	.pre_schedule		= pre_schedule_rt,
 	.post_schedule		= post_schedule_rt,
-	.task_wake_up		= task_wake_up_rt,
+	.task_woken		= task_woken_rt,
 	.switched_from		= switched_from_rt,
 #endif
 

-- 

[PATCH 11/12] sched: Remove the cfs_rq dependency from set_task_cpu()

[Peter Zijlstra]

[-- Attachment #1: foo1.patch --]
[-- Type: text/plain, Size: 6308 bytes --]

In order to remove the cfs_rq dependency from set_task_cpu() we need
to ensure the task is cfs_rq invariant for all callsites.

The simple approach is to substract cfs_rq->min_vruntime from
se->vruntime on dequeue, and add cfs_rq->min_vruntime on enqueue.

However, this has the downside of breaking FAIR_SLEEPERS since we
loose the old vruntime as we only maintain the relative position.

To solve this, we observe that we only migrate runnable tasks, we do
this using deactivate_task(.sleep=0) and activate_task(.wakeup=0),
therefore we can restrain the min_vruntime invariance to that state.

The only other case is wakeup balancing, since we want to maintain the
old vruntime we cannot make it relative on dequeue, but since we don't
migrate inactive tasks, we can do so right before we activate it
again.

This is where we need the new pre-wakeup hook, we need to call this
while still holding the old rq->lock. We could fold it into
->select_task_rq(), but since that has multiple callsites and would
obfuscate the locking requirements, that seems like a fudge.

This leaves the fork() case, simply make sure that ->task_fork()
leaves the ->vruntime in a relative state.

This covers all cases where set_task_cpu() gets called, and ensures it
sees a relative vruntime.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
 include/linux/sched.h |    2 +-
 kernel/sched.c        |    6 +-----
 kernel/sched_fair.c   |   50 ++++++++++++++++++++++++++++++++++++++++++++------
 3 files changed, 46 insertions(+), 12 deletions(-)

Index: linux-2.6/kernel/sched.c
===================================================================
--- linux-2.6.orig/kernel/sched.c
+++ linux-2.6/kernel/sched.c
@@ -2038,8 +2038,6 @@ task_hot(struct task_struct *p, u64 now,
 void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
 {
 	int old_cpu = task_cpu(p);
-	struct cfs_rq *old_cfsrq = task_cfs_rq(p),
-		      *new_cfsrq = cpu_cfs_rq(old_cfsrq, new_cpu);
 
 #ifdef CONFIG_SCHED_DEBUG
 	/*
@@ -2056,8 +2054,6 @@ void set_task_cpu(struct task_struct *p,
 		perf_sw_event(PERF_COUNT_SW_CPU_MIGRATIONS,
 				     1, 1, NULL, 0);
 	}
-	p->se.vruntime -= old_cfsrq->min_vruntime -
-					 new_cfsrq->min_vruntime;
 
 	__set_task_cpu(p, new_cpu);
 }
@@ -10109,7 +10105,7 @@ void sched_move_task(struct task_struct 
 
 #ifdef CONFIG_FAIR_GROUP_SCHED
 	if (tsk->sched_class->moved_group)
-		tsk->sched_class->moved_group(tsk);
+		tsk->sched_class->moved_group(tsk, on_rq);
 #endif
 
 	if (unlikely(running))
Index: linux-2.6/kernel/sched_fair.c
===================================================================
--- linux-2.6.orig/kernel/sched_fair.c
+++ linux-2.6/kernel/sched_fair.c
@@ -510,6 +510,7 @@ __update_curr(struct cfs_rq *cfs_rq, str
 	curr->sum_exec_runtime += delta_exec;
 	schedstat_add(cfs_rq, exec_clock, delta_exec);
 	delta_exec_weighted = calc_delta_fair(delta_exec, curr);
+
 	curr->vruntime += delta_exec_weighted;
 	update_min_vruntime(cfs_rq);
 }
@@ -765,16 +766,26 @@ place_entity(struct cfs_rq *cfs_rq, stru
 	se->vruntime = vruntime;
 }
 
+#define ENQUEUE_WAKEUP	1
+#define ENQUEUE_MIGRATE 2
+
 static void
-enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int wakeup)
+enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 {
 	/*
+	 * Update the normalized vruntime before updating min_vruntime
+	 * through callig update_curr().
+	 */
+	if (!(flags & ENQUEUE_WAKEUP) || (flags & ENQUEUE_MIGRATE))
+		se->vruntime += cfs_rq->min_vruntime;
+
+	/*
 	 * Update run-time statistics of the 'current'.
 	 */
 	update_curr(cfs_rq);
 	account_entity_enqueue(cfs_rq, se);
 
-	if (wakeup) {
+	if (flags & ENQUEUE_WAKEUP) {
 		place_entity(cfs_rq, se, 0);
 		enqueue_sleeper(cfs_rq, se);
 	}
@@ -828,6 +839,14 @@ dequeue_entity(struct cfs_rq *cfs_rq, st
 		__dequeue_entity(cfs_rq, se);
 	account_entity_dequeue(cfs_rq, se);
 	update_min_vruntime(cfs_rq);
+
+	/*
+	 * Normalize the entity after updating the min_vruntime because the
+	 * update can refer to the ->curr item and we need to reflect this
+	 * movement in our normalized position.
+	 */
+	if (!sleep)
+		se->vruntime -= cfs_rq->min_vruntime;
 }
 
 /*
@@ -1038,13 +1057,19 @@ static void enqueue_task_fair(struct rq 
 {
 	struct cfs_rq *cfs_rq;
 	struct sched_entity *se = &p->se;
+	int flags = 0;
+
+	if (wakeup)
+		flags |= ENQUEUE_WAKEUP;
+	if (p->state == TASK_WAKING)
+		flags |= ENQUEUE_MIGRATE;
 
 	for_each_sched_entity(se) {
 		if (se->on_rq)
 			break;
 		cfs_rq = cfs_rq_of(se);
-		enqueue_entity(cfs_rq, se, wakeup);
-		wakeup = 1;
+		enqueue_entity(cfs_rq, se, flags);
+		flags = ENQUEUE_WAKEUP;
 	}
 
 	hrtick_update(rq);
@@ -1120,6 +1145,14 @@ static void yield_task_fair(struct rq *r
 
 #ifdef CONFIG_SMP
 
+static void task_waking_fair(struct rq *rq, struct task_struct *p)
+{
+	struct sched_entity *se = &p->se;
+	struct cfs_rq *cfs_rq = cfs_rq_of(se);
+
+	se->vruntime -= cfs_rq->min_vruntime;
+}
+
 #ifdef CONFIG_FAIR_GROUP_SCHED
 /*
  * effective_load() calculates the load change as seen from the root_task_group
@@ -1978,6 +2011,8 @@ static void task_fork_fair(struct task_s
 		resched_task(rq->curr);
 	}
 
+	se->vruntime -= cfs_rq->min_vruntime;
+
 	raw_spin_unlock_irqrestore(&rq->lock, flags);
 }
 
@@ -2031,12 +2066,13 @@ static void set_curr_task_fair(struct rq
 }
 
 #ifdef CONFIG_FAIR_GROUP_SCHED
-static void moved_group_fair(struct task_struct *p)
+static void moved_group_fair(struct task_struct *p, int on_rq)
 {
 	struct cfs_rq *cfs_rq = task_cfs_rq(p);
 
 	update_curr(cfs_rq);
-	place_entity(cfs_rq, &p->se, 1);
+	if (!on_rq)
+		place_entity(cfs_rq, &p->se, 1);
 }
 #endif
 
@@ -2076,6 +2112,8 @@ static const struct sched_class fair_sch
 	.move_one_task		= move_one_task_fair,
 	.rq_online		= rq_online_fair,
 	.rq_offline		= rq_offline_fair,
+
+	.task_waking		= task_waking_fair,
 #endif
 
 	.set_curr_task          = set_curr_task_fair,
Index: linux-2.6/include/linux/sched.h
===================================================================
--- linux-2.6.orig/include/linux/sched.h
+++ linux-2.6/include/linux/sched.h
@@ -1116,7 +1116,7 @@ struct sched_class {
 					 struct task_struct *task);
 
 #ifdef CONFIG_FAIR_GROUP_SCHED
-	void (*moved_group) (struct task_struct *p);
+	void (*moved_group) (struct task_struct *p, int on_rq);
 #endif
 };
 

-- 

[PATCH 12/12] sched: Simplify set_task_cpu()

[Peter Zijlstra]

[-- Attachment #1: foo2.patch --]
[-- Type: text/plain, Size: 1087 bytes --]

Rearrange code a bit now that its a simpler function.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
---
 kernel/sched.c |   13 +++++--------
 1 file changed, 5 insertions(+), 8 deletions(-)

Index: linux-2.6/kernel/sched.c
===================================================================
--- linux-2.6.orig/kernel/sched.c
+++ linux-2.6/kernel/sched.c
@@ -2034,11 +2034,8 @@ task_hot(struct task_struct *p, u64 now,
 	return delta < (s64)sysctl_sched_migration_cost;
 }
 
-
 void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
 {
-	int old_cpu = task_cpu(p);
-
 #ifdef CONFIG_SCHED_DEBUG
 	/*
 	 * We should never call set_task_cpu() on a blocked task,
@@ -2049,11 +2046,11 @@ void set_task_cpu(struct task_struct *p,
 
 	trace_sched_migrate_task(p, new_cpu);
 
-	if (old_cpu != new_cpu) {
-		p->se.nr_migrations++;
-		perf_sw_event(PERF_COUNT_SW_CPU_MIGRATIONS,
-				     1, 1, NULL, 0);
-	}
+	if (task_cpu(p) == new_cpu)
+		return;
+
+	p->se.nr_migrations++;
+	perf_sw_event(PERF_COUNT_SW_CPU_MIGRATIONS, 1, 1, NULL, 0);
 
 	__set_task_cpu(p, new_cpu);
 }

-- 

Re: [PATCH 02/12] sched: Fix set_cpu_active() in cpu_down()

[Peter Zijlstra]

On Wed, 2009-12-16 at 18:04 +0100, Peter Zijlstra wrote:
> plain text document attachment (foo13.patch)

still need to fix quilt...

From: Xiaotian Feng <dfeng@redhat.com>

> Sachin found cpu hotplug test failures on powerpc, which made the
> kernel hang on his POWER box. 
> 
> The problem is that we fail to re-activate a cpu when a hot-unplug
> fails. Fix this by moving the de-activation into _cpu_down after doing
> the initial checks.
> 
> Remove the synchronize_sched() calls and rely on those implied by
> rebuilding the sched domains using the new mask.
> 
> Reported-by: Sachin Sant <sachinp@in.ibm.com>
> Signed-off-by: Xiaotian Feng <dfeng@redhat.com>
> Tested-by: Sachin Sant <sachinp@in.ibm.com>
> Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>

[tip:sched/urgent] sched: Mark boot-cpu active before smp_init()

[tip-bot for Peter Zijlstra]

Commit-ID:  933b0618d8b2a59c7a0742e43836544e02f1e9bd
Gitweb:     http://git.kernel.org/tip/933b0618d8b2a59c7a0742e43836544e02f1e9bd
Author:     Peter Zijlstra <a.p.zijlstra@chello.nl>
AuthorDate: Wed, 16 Dec 2009 18:04:31 +0100
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Wed, 16 Dec 2009 19:01:53 +0100

sched: Mark boot-cpu active before smp_init()

A UP machine has 1 active cpu, not having the boot-cpu in the
active map when starting the scheduler confuses things.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
LKML-Reference: <20091216170517.423469527@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
---
 init/main.c |    7 +------
 1 files changed, 1 insertions(+), 6 deletions(-)

diff --git a/init/main.c b/init/main.c
index c3db4a9..dac44a9 100644
--- a/init/main.c
+++ b/init/main.c
@@ -369,12 +369,6 @@ static void __init smp_init(void)
 {
 	unsigned int cpu;
 
-	/*
-	 * Set up the current CPU as possible to migrate to.
-	 * The other ones will be done by cpu_up/cpu_down()
-	 */
-	set_cpu_active(smp_processor_id(), true);
-
 	/* FIXME: This should be done in userspace --RR */
 	for_each_present_cpu(cpu) {
 		if (num_online_cpus() >= setup_max_cpus)
@@ -486,6 +480,7 @@ static void __init boot_cpu_init(void)
 	int cpu = smp_processor_id();
 	/* Mark the boot cpu "present", "online" etc for SMP and UP case */
 	set_cpu_online(cpu, true);
+	set_cpu_active(cpu, true);
 	set_cpu_present(cpu, true);
 	set_cpu_possible(cpu, true);
 }

[tip:sched/urgent] sched: Fix set_cpu_active() in cpu_down()

[tip-bot for Xiaotian Feng]

Commit-ID:  9ee349ad6d326df3633d43f54202427295999c47
Gitweb:     http://git.kernel.org/tip/9ee349ad6d326df3633d43f54202427295999c47
Author:     Xiaotian Feng <dfeng@redhat.com>
AuthorDate: Wed, 16 Dec 2009 18:04:32 +0100
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Wed, 16 Dec 2009 19:01:53 +0100

sched: Fix set_cpu_active() in cpu_down()

Sachin found cpu hotplug test failures on powerpc, which made
the kernel hang on his POWER box.

The problem is that we fail to re-activate a cpu when a
hot-unplug fails. Fix this by moving the de-activation into
_cpu_down after doing the initial checks.

Remove the synchronize_sched() calls and rely on those implied
by rebuilding the sched domains using the new mask.

Reported-by: Sachin Sant <sachinp@in.ibm.com>
Signed-off-by: Xiaotian Feng <dfeng@redhat.com>
Tested-by: Sachin Sant <sachinp@in.ibm.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
LKML-Reference: <20091216170517.500272612@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
---
 kernel/cpu.c |   24 +++---------------------
 1 files changed, 3 insertions(+), 21 deletions(-)

diff --git a/kernel/cpu.c b/kernel/cpu.c
index 291ac58..1c8ddd6 100644
--- a/kernel/cpu.c
+++ b/kernel/cpu.c
@@ -209,6 +209,7 @@ static int __ref _cpu_down(unsigned int cpu, int tasks_frozen)
 		return -ENOMEM;
 
 	cpu_hotplug_begin();
+	set_cpu_active(cpu, false);
 	err = __raw_notifier_call_chain(&cpu_chain, CPU_DOWN_PREPARE | mod,
 					hcpu, -1, &nr_calls);
 	if (err == NOTIFY_BAD) {
@@ -280,18 +281,6 @@ int __ref cpu_down(unsigned int cpu)
 		goto out;
 	}
 
-	set_cpu_active(cpu, false);
-
-	/*
-	 * Make sure the all cpus did the reschedule and are not
-	 * using stale version of the cpu_active_mask.
-	 * This is not strictly necessary becuase stop_machine()
-	 * that we run down the line already provides the required
-	 * synchronization. But it's really a side effect and we do not
-	 * want to depend on the innards of the stop_machine here.
-	 */
-	synchronize_sched();
-
 	err = _cpu_down(cpu, 0);
 
 out:
@@ -382,19 +371,12 @@ int disable_nonboot_cpus(void)
 		return error;
 	cpu_maps_update_begin();
 	first_cpu = cpumask_first(cpu_online_mask);
-	/* We take down all of the non-boot CPUs in one shot to avoid races
+	/*
+	 * We take down all of the non-boot CPUs in one shot to avoid races
 	 * with the userspace trying to use the CPU hotplug at the same time
 	 */
 	cpumask_clear(frozen_cpus);
 
-	for_each_online_cpu(cpu) {
-		if (cpu == first_cpu)
-			continue;
-		set_cpu_active(cpu, false);
-	}
-
-	synchronize_sched();
-
 	printk("Disabling non-boot CPUs ...\n");
 	for_each_online_cpu(cpu) {
 		if (cpu == first_cpu)

[tip:sched/urgent] sched: Fix task_hot() test order

[tip-bot for Peter Zijlstra]

Commit-ID:  e6c8fba7771563b2f3dfb96a78f36ec17e15bdf0
Gitweb:     http://git.kernel.org/tip/e6c8fba7771563b2f3dfb96a78f36ec17e15bdf0
Author:     Peter Zijlstra <a.p.zijlstra@chello.nl>
AuthorDate: Wed, 16 Dec 2009 18:04:33 +0100
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Wed, 16 Dec 2009 19:01:54 +0100

sched: Fix task_hot() test order

Make sure not to access sched_fair fields before verifying it is
indeed a sched_fair task.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
CC: stable@kernel.org
LKML-Reference: <20091216170517.577998058@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
---
 kernel/sched.c |    6 +++---
 1 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/kernel/sched.c b/kernel/sched.c
index 9c30858..1d8ca25 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -2046,6 +2046,9 @@ task_hot(struct task_struct *p, u64 now, struct sched_domain *sd)
 {
 	s64 delta;
 
+	if (p->sched_class != &fair_sched_class)
+		return 0;
+
 	/*
 	 * Buddy candidates are cache hot:
 	 */
@@ -2054,9 +2057,6 @@ task_hot(struct task_struct *p, u64 now, struct sched_domain *sd)
 			 &p->se == cfs_rq_of(&p->se)->last))
 		return 1;
 
-	if (p->sched_class != &fair_sched_class)
-		return 0;
-
 	if (sysctl_sched_migration_cost == -1)
 		return 1;
 	if (sysctl_sched_migration_cost == 0)

[tip:sched/urgent] sched: Select_task_rq_fair() must honour SD_LOAD_BALANCE

[tip-bot for Peter Zijlstra]

Commit-ID:  e4f4288842ee12747e10c354d72be7d424c0b627
Gitweb:     http://git.kernel.org/tip/e4f4288842ee12747e10c354d72be7d424c0b627
Author:     Peter Zijlstra <a.p.zijlstra@chello.nl>
AuthorDate: Wed, 16 Dec 2009 18:04:34 +0100
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Wed, 16 Dec 2009 19:01:55 +0100

sched: Select_task_rq_fair() must honour SD_LOAD_BALANCE

We should skip !SD_LOAD_BALANCE domains.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
LKML-Reference: <20091216170517.653578430@chello.nl>
CC: stable@kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>
---
 kernel/sched_fair.c |    3 +++
 1 files changed, 3 insertions(+), 0 deletions(-)

diff --git a/kernel/sched_fair.c b/kernel/sched_fair.c
index 5bedf6e..ec1d271 100644
--- a/kernel/sched_fair.c
+++ b/kernel/sched_fair.c
@@ -1429,6 +1429,9 @@ static int select_task_rq_fair(struct task_struct *p, int sd_flag, int wake_flag
 	}
 
 	for_each_domain(cpu, tmp) {
+		if (!(tmp->flags & SD_LOAD_BALANCE))
+			continue;
+
 		/*
 		 * If power savings logic is enabled for a domain, see if we
 		 * are not overloaded, if so, don't balance wider.

[tip:sched/urgent] sched: Use TASK_WAKING for fork wakups

[tip-bot for Peter Zijlstra]

Commit-ID:  06b83b5fbea273672822b6ee93e16781046553ec
Gitweb:     http://git.kernel.org/tip/06b83b5fbea273672822b6ee93e16781046553ec
Author:     Peter Zijlstra <a.p.zijlstra@chello.nl>
AuthorDate: Wed, 16 Dec 2009 18:04:35 +0100
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Wed, 16 Dec 2009 19:01:55 +0100

sched: Use TASK_WAKING for fork wakups

For later convenience use TASK_WAKING for fresh tasks.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
LKML-Reference: <20091216170517.732561278@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
---
 kernel/sched.c |   18 +++++++++---------
 1 files changed, 9 insertions(+), 9 deletions(-)

diff --git a/kernel/sched.c b/kernel/sched.c
index 1d8ca25..1672823 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -2540,14 +2540,6 @@ static void __sched_fork(struct task_struct *p)
 #ifdef CONFIG_PREEMPT_NOTIFIERS
 	INIT_HLIST_HEAD(&p->preempt_notifiers);
 #endif
-
-	/*
-	 * We mark the process as running here, but have not actually
-	 * inserted it onto the runqueue yet. This guarantees that
-	 * nobody will actually run it, and a signal or other external
-	 * event cannot wake it up and insert it on the runqueue either.
-	 */
-	p->state = TASK_RUNNING;
 }
 
 /*
@@ -2558,6 +2550,12 @@ void sched_fork(struct task_struct *p, int clone_flags)
 	int cpu = get_cpu();
 
 	__sched_fork(p);
+	/*
+	 * We mark the process as waking here. This guarantees that
+	 * nobody will actually run it, and a signal or other external
+	 * event cannot wake it up and insert it on the runqueue either.
+	 */
+	p->state = TASK_WAKING;
 
 	/*
 	 * Revert to default priority/policy on fork if requested.
@@ -2626,7 +2624,8 @@ void wake_up_new_task(struct task_struct *p, unsigned long clone_flags)
 	struct rq *rq;
 
 	rq = task_rq_lock(p, &flags);
-	BUG_ON(p->state != TASK_RUNNING);
+	BUG_ON(p->state != TASK_WAKING);
+	p->state = TASK_RUNNING;
 	update_rq_clock(rq);
 	activate_task(rq, p, 0);
 	trace_sched_wakeup_new(rq, p, 1);
@@ -6984,6 +6983,7 @@ void __cpuinit init_idle(struct task_struct *idle, int cpu)
 	raw_spin_lock_irqsave(&rq->lock, flags);
 
 	__sched_fork(idle);
+	idle->state = TASK_RUNNING;
 	idle->se.exec_start = sched_clock();
 
 	cpumask_copy(&idle->cpus_allowed, cpumask_of(cpu));

[tip:sched/urgent] sched: Ensure set_task_cpu() is never called on blocked tasks

[tip-bot for Peter Zijlstra]

Commit-ID:  e2912009fb7b715728311b0d8fe327a1432b3f79
Gitweb:     http://git.kernel.org/tip/e2912009fb7b715728311b0d8fe327a1432b3f79
Author:     Peter Zijlstra <a.p.zijlstra@chello.nl>
AuthorDate: Wed, 16 Dec 2009 18:04:36 +0100
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Wed, 16 Dec 2009 19:01:56 +0100

sched: Ensure set_task_cpu() is never called on blocked tasks

In order to clean up the set_task_cpu() rq dependencies we need
to ensure it is never called on blocked tasks because such usage
does not pair with consistent rq->lock usage.

This puts the migration burden on ttwu().

Furthermore we need to close a race against changing
->cpus_allowed, since select_task_rq() runs with only preemption
disabled.

For sched_fork() this is safe because the child isn't in the
tasklist yet, for wakeup we fix this by synchronizing
set_cpus_allowed_ptr() against TASK_WAKING, which leaves
sched_exec to be a problem

This also closes a hole in (6ad4c1888 sched: Fix balance vs
hotplug race) where ->select_task_rq() doesn't validate the
result against the sched_domain/root_domain.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
LKML-Reference: <20091216170517.807938893@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
---
 kernel/sched.c |   85 +++++++++++++++++++++++++++++++++++++++++++------------
 1 files changed, 66 insertions(+), 19 deletions(-)

diff --git a/kernel/sched.c b/kernel/sched.c
index 1672823..33d7965 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -2018,22 +2018,15 @@ static inline void check_class_changed(struct rq *rq, struct task_struct *p,
  */
 void kthread_bind(struct task_struct *p, unsigned int cpu)
 {
-	struct rq *rq = cpu_rq(cpu);
-	unsigned long flags;
-
 	/* Must have done schedule() in kthread() before we set_task_cpu */
 	if (!wait_task_inactive(p, TASK_UNINTERRUPTIBLE)) {
 		WARN_ON(1);
 		return;
 	}
 
-	raw_spin_lock_irqsave(&rq->lock, flags);
-	update_rq_clock(rq);
-	set_task_cpu(p, cpu);
 	p->cpus_allowed = cpumask_of_cpu(cpu);
 	p->rt.nr_cpus_allowed = 1;
 	p->flags |= PF_THREAD_BOUND;
-	raw_spin_unlock_irqrestore(&rq->lock, flags);
 }
 EXPORT_SYMBOL(kthread_bind);
 
@@ -2074,6 +2067,14 @@ void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
 	struct cfs_rq *old_cfsrq = task_cfs_rq(p),
 		      *new_cfsrq = cpu_cfs_rq(old_cfsrq, new_cpu);
 
+#ifdef CONFIG_SCHED_DEBUG
+	/*
+	 * We should never call set_task_cpu() on a blocked task,
+	 * ttwu() will sort out the placement.
+	 */
+	WARN_ON(p->state != TASK_RUNNING && p->state != TASK_WAKING);
+#endif
+
 	trace_sched_migrate_task(p, new_cpu);
 
 	if (old_cpu != new_cpu) {
@@ -2107,13 +2108,10 @@ migrate_task(struct task_struct *p, int dest_cpu, struct migration_req *req)
 
 	/*
 	 * If the task is not on a runqueue (and not running), then
-	 * it is sufficient to simply update the task's cpu field.
+	 * the next wake-up will properly place the task.
 	 */
-	if (!p->se.on_rq && !task_running(rq, p)) {
-		update_rq_clock(rq);
-		set_task_cpu(p, dest_cpu);
+	if (!p->se.on_rq && !task_running(rq, p))
 		return 0;
-	}
 
 	init_completion(&req->done);
 	req->task = p;
@@ -2319,10 +2317,42 @@ void task_oncpu_function_call(struct task_struct *p,
 }
 
 #ifdef CONFIG_SMP
+/*
+ * Called from:
+ *
+ *  - fork, @p is stable because it isn't on the tasklist yet
+ *
+ *  - exec, @p is unstable XXX
+ *
+ *  - wake-up, we serialize ->cpus_allowed against TASK_WAKING so
+ *             we should be good.
+ */
 static inline
 int select_task_rq(struct task_struct *p, int sd_flags, int wake_flags)
 {
-	return p->sched_class->select_task_rq(p, sd_flags, wake_flags);
+	int cpu = p->sched_class->select_task_rq(p, sd_flags, wake_flags);
+
+	/*
+	 * In order not to call set_task_cpu() on a blocking task we need
+	 * to rely on ttwu() to place the task on a valid ->cpus_allowed
+	 * cpu.
+	 *
+	 * Since this is common to all placement strategies, this lives here.
+	 *
+	 * [ this allows ->select_task() to simply return task_cpu(p) and
+	 *   not worry about this generic constraint ]
+	 */
+	if (unlikely(!cpumask_test_cpu(cpu, &p->cpus_allowed) ||
+		     !cpu_active(cpu))) {
+
+		cpu = cpumask_any_and(&p->cpus_allowed, cpu_active_mask);
+		/*
+		 * XXX: race against hot-plug modifying cpu_active_mask
+		 */
+		BUG_ON(cpu >= nr_cpu_ids);
+	}
+
+	return cpu;
 }
 #endif
 
@@ -7098,7 +7128,23 @@ int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
 	struct rq *rq;
 	int ret = 0;
 
+	/*
+	 * Since we rely on wake-ups to migrate sleeping tasks, don't change
+	 * the ->cpus_allowed mask from under waking tasks, which would be
+	 * possible when we change rq->lock in ttwu(), so synchronize against
+	 * TASK_WAKING to avoid that.
+	 */
+again:
+	while (p->state == TASK_WAKING)
+		cpu_relax();
+
 	rq = task_rq_lock(p, &flags);
+
+	if (p->state == TASK_WAKING) {
+		task_rq_unlock(rq, &flags);
+		goto again;
+	}
+
 	if (!cpumask_intersects(new_mask, cpu_active_mask)) {
 		ret = -EINVAL;
 		goto out;
@@ -7154,7 +7200,7 @@ EXPORT_SYMBOL_GPL(set_cpus_allowed_ptr);
 static int __migrate_task(struct task_struct *p, int src_cpu, int dest_cpu)
 {
 	struct rq *rq_dest, *rq_src;
-	int ret = 0, on_rq;
+	int ret = 0;
 
 	if (unlikely(!cpu_active(dest_cpu)))
 		return ret;
@@ -7170,12 +7216,13 @@ static int __migrate_task(struct task_struct *p, int src_cpu, int dest_cpu)
 	if (!cpumask_test_cpu(dest_cpu, &p->cpus_allowed))
 		goto fail;
 
-	on_rq = p->se.on_rq;
-	if (on_rq)
+	/*
+	 * If we're not on a rq, the next wake-up will ensure we're
+	 * placed properly.
+	 */
+	if (p->se.on_rq) {
 		deactivate_task(rq_src, p, 0);
-
-	set_task_cpu(p, dest_cpu);
-	if (on_rq) {
+		set_task_cpu(p, dest_cpu);
 		activate_task(rq_dest, p, 0);
 		check_preempt_curr(rq_dest, p, 0);
 	}

[tip:sched/urgent] sched: Fix sched_exec() balancing

[tip-bot for Peter Zijlstra]

Commit-ID:  3802290628348674985d14914f9bfee7b9084548
Gitweb:     http://git.kernel.org/tip/3802290628348674985d14914f9bfee7b9084548
Author:     Peter Zijlstra <a.p.zijlstra@chello.nl>
AuthorDate: Wed, 16 Dec 2009 18:04:37 +0100
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Wed, 16 Dec 2009 19:01:56 +0100

sched: Fix sched_exec() balancing

Since we access ->cpus_allowed without holding rq->lock we need
a retry loop to validate the result, this comes for near free
when we merge sched_migrate_task() into sched_exec() since that
already does the needed check.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
LKML-Reference: <20091216170517.884743662@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
---
 kernel/sched.c |   45 +++++++++++++++++++++++----------------------
 1 files changed, 23 insertions(+), 22 deletions(-)

diff --git a/kernel/sched.c b/kernel/sched.c
index 33d7965..63e55ac 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -2322,7 +2322,7 @@ void task_oncpu_function_call(struct task_struct *p,
  *
  *  - fork, @p is stable because it isn't on the tasklist yet
  *
- *  - exec, @p is unstable XXX
+ *  - exec, @p is unstable, retry loop
  *
  *  - wake-up, we serialize ->cpus_allowed against TASK_WAKING so
  *             we should be good.
@@ -3132,21 +3132,36 @@ static void double_rq_unlock(struct rq *rq1, struct rq *rq2)
 }
 
 /*
- * If dest_cpu is allowed for this process, migrate the task to it.
- * This is accomplished by forcing the cpu_allowed mask to only
- * allow dest_cpu, which will force the cpu onto dest_cpu. Then
- * the cpu_allowed mask is restored.
+ * sched_exec - execve() is a valuable balancing opportunity, because at
+ * this point the task has the smallest effective memory and cache footprint.
  */
-static void sched_migrate_task(struct task_struct *p, int dest_cpu)
+void sched_exec(void)
 {
+	struct task_struct *p = current;
 	struct migration_req req;
+	int dest_cpu, this_cpu;
 	unsigned long flags;
 	struct rq *rq;
 
+again:
+	this_cpu = get_cpu();
+	dest_cpu = select_task_rq(p, SD_BALANCE_EXEC, 0);
+	if (dest_cpu == this_cpu) {
+		put_cpu();
+		return;
+	}
+
 	rq = task_rq_lock(p, &flags);
+	put_cpu();
+
+	/*
+	 * select_task_rq() can race against ->cpus_allowed
+	 */
 	if (!cpumask_test_cpu(dest_cpu, &p->cpus_allowed)
-	    || unlikely(!cpu_active(dest_cpu)))
-		goto out;
+	    || unlikely(!cpu_active(dest_cpu))) {
+		task_rq_unlock(rq, &flags);
+		goto again;
+	}
 
 	/* force the process onto the specified CPU */
 	if (migrate_task(p, dest_cpu, &req)) {
@@ -3161,24 +3176,10 @@ static void sched_migrate_task(struct task_struct *p, int dest_cpu)
 
 		return;
 	}
-out:
 	task_rq_unlock(rq, &flags);
 }
 
 /*
- * sched_exec - execve() is a valuable balancing opportunity, because at
- * this point the task has the smallest effective memory and cache footprint.
- */
-void sched_exec(void)
-{
-	int new_cpu, this_cpu = get_cpu();
-	new_cpu = select_task_rq(current, SD_BALANCE_EXEC, 0);
-	put_cpu();
-	if (new_cpu != this_cpu)
-		sched_migrate_task(current, new_cpu);
-}
-
-/*
  * pull_task - move a task from a remote runqueue to the local runqueue.
  * Both runqueues must be locked.
  */

[tip:sched/urgent] sched: Fix select_task_rq() vs hotplug issues

[tip-bot for Peter Zijlstra]

Commit-ID:  5da9a0fb673a0ea0a093862f95f6b89b3390c31e
Gitweb:     http://git.kernel.org/tip/5da9a0fb673a0ea0a093862f95f6b89b3390c31e
Author:     Peter Zijlstra <a.p.zijlstra@chello.nl>
AuthorDate: Wed, 16 Dec 2009 18:04:38 +0100
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Wed, 16 Dec 2009 19:01:57 +0100

sched: Fix select_task_rq() vs hotplug issues

Since select_task_rq() is now responsible for guaranteeing
->cpus_allowed and cpu_active_mask, we need to verify this.

select_task_rq_rt() can blindly return
smp_processor_id()/task_cpu() without checking the valid masks,
select_task_rq_fair() can do the same in the rare case that all
SD_flags are disabled.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
LKML-Reference: <20091216170517.961475466@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
---
 kernel/sched.c |   75 +++++++++++++++++++++++++++++--------------------------
 1 files changed, 40 insertions(+), 35 deletions(-)

diff --git a/kernel/sched.c b/kernel/sched.c
index 63e55ac..cc40bda 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -2317,6 +2317,43 @@ void task_oncpu_function_call(struct task_struct *p,
 }
 
 #ifdef CONFIG_SMP
+static int select_fallback_rq(int cpu, struct task_struct *p)
+{
+	int dest_cpu;
+	const struct cpumask *nodemask = cpumask_of_node(cpu_to_node(cpu));
+
+	/* Look for allowed, online CPU in same node. */
+	for_each_cpu_and(dest_cpu, nodemask, cpu_active_mask)
+		if (cpumask_test_cpu(dest_cpu, &p->cpus_allowed))
+			return dest_cpu;
+
+	/* Any allowed, online CPU? */
+	dest_cpu = cpumask_any_and(&p->cpus_allowed, cpu_active_mask);
+	if (dest_cpu < nr_cpu_ids)
+		return dest_cpu;
+
+	/* No more Mr. Nice Guy. */
+	if (dest_cpu >= nr_cpu_ids) {
+		rcu_read_lock();
+		cpuset_cpus_allowed_locked(p, &p->cpus_allowed);
+		rcu_read_unlock();
+		dest_cpu = cpumask_any_and(cpu_active_mask, &p->cpus_allowed);
+
+		/*
+		 * Don't tell them about moving exiting tasks or
+		 * kernel threads (both mm NULL), since they never
+		 * leave kernel.
+		 */
+		if (p->mm && printk_ratelimit()) {
+			printk(KERN_INFO "process %d (%s) no "
+			       "longer affine to cpu%d\n",
+			       task_pid_nr(p), p->comm, cpu);
+		}
+	}
+
+	return dest_cpu;
+}
+
 /*
  * Called from:
  *
@@ -2343,14 +2380,8 @@ int select_task_rq(struct task_struct *p, int sd_flags, int wake_flags)
 	 *   not worry about this generic constraint ]
 	 */
 	if (unlikely(!cpumask_test_cpu(cpu, &p->cpus_allowed) ||
-		     !cpu_active(cpu))) {
-
-		cpu = cpumask_any_and(&p->cpus_allowed, cpu_active_mask);
-		/*
-		 * XXX: race against hot-plug modifying cpu_active_mask
-		 */
-		BUG_ON(cpu >= nr_cpu_ids);
-	}
+		     !cpu_active(cpu)))
+		cpu = select_fallback_rq(task_cpu(p), p);
 
 	return cpu;
 }
@@ -7319,36 +7350,10 @@ static int __migrate_task_irq(struct task_struct *p, int src_cpu, int dest_cpu)
 static void move_task_off_dead_cpu(int dead_cpu, struct task_struct *p)
 {
 	int dest_cpu;
-	const struct cpumask *nodemask = cpumask_of_node(cpu_to_node(dead_cpu));
 
 again:
-	/* Look for allowed, online CPU in same node. */
-	for_each_cpu_and(dest_cpu, nodemask, cpu_active_mask)
-		if (cpumask_test_cpu(dest_cpu, &p->cpus_allowed))
-			goto move;
-
-	/* Any allowed, online CPU? */
-	dest_cpu = cpumask_any_and(&p->cpus_allowed, cpu_active_mask);
-	if (dest_cpu < nr_cpu_ids)
-		goto move;
-
-	/* No more Mr. Nice Guy. */
-	if (dest_cpu >= nr_cpu_ids) {
-		cpuset_cpus_allowed_locked(p, &p->cpus_allowed);
-		dest_cpu = cpumask_any_and(cpu_active_mask, &p->cpus_allowed);
-
-		/*
-		 * Don't tell them about moving exiting tasks or
-		 * kernel threads (both mm NULL), since they never
-		 * leave kernel.
-		 */
-		if (p->mm && printk_ratelimit()) {
-			pr_info("process %d (%s) no longer affine to cpu%d\n",
-				task_pid_nr(p), p->comm, dead_cpu);
-		}
-	}
+	dest_cpu = select_fallback_rq(dead_cpu, p);
 
-move:
 	/* It can have affinity changed while we were choosing. */
 	if (unlikely(!__migrate_task_irq(p, dead_cpu, dest_cpu)))
 		goto again;

[tip:sched/urgent] sched: Move kthread_bind() back to kthread.c

[tip-bot for Peter Zijlstra]

Commit-ID:  881232b70b195768a71cd74ff4b4e8ab9502997b
Gitweb:     http://git.kernel.org/tip/881232b70b195768a71cd74ff4b4e8ab9502997b
Author:     Peter Zijlstra <a.p.zijlstra@chello.nl>
AuthorDate: Wed, 16 Dec 2009 18:04:39 +0100
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Wed, 16 Dec 2009 19:01:57 +0100

sched: Move kthread_bind() back to kthread.c

Since kthread_bind() lost its dependencies on sched.c, move it
back where it came from.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
LKML-Reference: <20091216170518.039524041@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
---
 kernel/kthread.c |   23 +++++++++++++++++++++++
 kernel/sched.c   |   26 --------------------------
 2 files changed, 23 insertions(+), 26 deletions(-)

diff --git a/kernel/kthread.c b/kernel/kthread.c
index ab7ae57..fbb6222 100644
--- a/kernel/kthread.c
+++ b/kernel/kthread.c
@@ -150,6 +150,29 @@ struct task_struct *kthread_create(int (*threadfn)(void *data),
 EXPORT_SYMBOL(kthread_create);
 
 /**
+ * kthread_bind - bind a just-created kthread to a cpu.
+ * @p: thread created by kthread_create().
+ * @cpu: cpu (might not be online, must be possible) for @k to run on.
+ *
+ * Description: This function is equivalent to set_cpus_allowed(),
+ * except that @cpu doesn't need to be online, and the thread must be
+ * stopped (i.e., just returned from kthread_create()).
+ */
+void kthread_bind(struct task_struct *p, unsigned int cpu)
+{
+	/* Must have done schedule() in kthread() before we set_task_cpu */
+	if (!wait_task_inactive(p, TASK_UNINTERRUPTIBLE)) {
+		WARN_ON(1);
+		return;
+	}
+
+	p->cpus_allowed = cpumask_of_cpu(cpu);
+	p->rt.nr_cpus_allowed = 1;
+	p->flags |= PF_THREAD_BOUND;
+}
+EXPORT_SYMBOL(kthread_bind);
+
+/**
  * kthread_stop - stop a thread created by kthread_create().
  * @k: thread created by kthread_create().
  *
diff --git a/kernel/sched.c b/kernel/sched.c
index cc40bda..297dc44 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -2004,32 +2004,6 @@ static inline void check_class_changed(struct rq *rq, struct task_struct *p,
 		p->sched_class->prio_changed(rq, p, oldprio, running);
 }
 
-/**
- * kthread_bind - bind a just-created kthread to a cpu.
- * @p: thread created by kthread_create().
- * @cpu: cpu (might not be online, must be possible) for @k to run on.
- *
- * Description: This function is equivalent to set_cpus_allowed(),
- * except that @cpu doesn't need to be online, and the thread must be
- * stopped (i.e., just returned from kthread_create()).
- *
- * Function lives here instead of kthread.c because it messes with
- * scheduler internals which require locking.
- */
-void kthread_bind(struct task_struct *p, unsigned int cpu)
-{
-	/* Must have done schedule() in kthread() before we set_task_cpu */
-	if (!wait_task_inactive(p, TASK_UNINTERRUPTIBLE)) {
-		WARN_ON(1);
-		return;
-	}
-
-	p->cpus_allowed = cpumask_of_cpu(cpu);
-	p->rt.nr_cpus_allowed = 1;
-	p->flags |= PF_THREAD_BOUND;
-}
-EXPORT_SYMBOL(kthread_bind);
-
 #ifdef CONFIG_SMP
 /*
  * Is this task likely cache-hot:

[tip:sched/urgent] sched: Add pre and post wakeup hooks

[tip-bot for Peter Zijlstra]

Commit-ID:  efbbd05a595343a413964ad85a2ad359b7b7efbd
Gitweb:     http://git.kernel.org/tip/efbbd05a595343a413964ad85a2ad359b7b7efbd
Author:     Peter Zijlstra <a.p.zijlstra@chello.nl>
AuthorDate: Wed, 16 Dec 2009 18:04:40 +0100
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Wed, 16 Dec 2009 19:01:58 +0100

sched: Add pre and post wakeup hooks

As will be apparent in the next patch, we need a pre wakeup hook
for sched_fair task migration, hence rename the post wakeup hook
and one pre wakeup.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
LKML-Reference: <20091216170518.114746117@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
---
 include/linux/sched.h |    3 ++-
 kernel/sched.c        |   12 ++++++++----
 kernel/sched_rt.c     |    4 ++--
 3 files changed, 12 insertions(+), 7 deletions(-)

diff --git a/include/linux/sched.h b/include/linux/sched.h
index 5c858f3..2c9fa1c 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -1091,7 +1091,8 @@ struct sched_class {
 			      enum cpu_idle_type idle);
 	void (*pre_schedule) (struct rq *this_rq, struct task_struct *task);
 	void (*post_schedule) (struct rq *this_rq);
-	void (*task_wake_up) (struct rq *this_rq, struct task_struct *task);
+	void (*task_waking) (struct rq *this_rq, struct task_struct *task);
+	void (*task_woken) (struct rq *this_rq, struct task_struct *task);
 
 	void (*set_cpus_allowed)(struct task_struct *p,
 				 const struct cpumask *newmask);
diff --git a/kernel/sched.c b/kernel/sched.c
index 297dc44..6c571bd 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -2412,6 +2412,10 @@ static int try_to_wake_up(struct task_struct *p, unsigned int state,
 	if (task_contributes_to_load(p))
 		rq->nr_uninterruptible--;
 	p->state = TASK_WAKING;
+
+	if (p->sched_class->task_waking)
+		p->sched_class->task_waking(rq, p);
+
 	__task_rq_unlock(rq);
 
 	cpu = select_task_rq(p, SD_BALANCE_WAKE, wake_flags);
@@ -2475,8 +2479,8 @@ out_running:
 
 	p->state = TASK_RUNNING;
 #ifdef CONFIG_SMP
-	if (p->sched_class->task_wake_up)
-		p->sched_class->task_wake_up(rq, p);
+	if (p->sched_class->task_woken)
+		p->sched_class->task_woken(rq, p);
 
 	if (unlikely(rq->idle_stamp)) {
 		u64 delta = rq->clock - rq->idle_stamp;
@@ -2666,8 +2670,8 @@ void wake_up_new_task(struct task_struct *p, unsigned long clone_flags)
 	trace_sched_wakeup_new(rq, p, 1);
 	check_preempt_curr(rq, p, WF_FORK);
 #ifdef CONFIG_SMP
-	if (p->sched_class->task_wake_up)
-		p->sched_class->task_wake_up(rq, p);
+	if (p->sched_class->task_woken)
+		p->sched_class->task_woken(rq, p);
 #endif
 	task_rq_unlock(rq, &flags);
 }
diff --git a/kernel/sched_rt.c b/kernel/sched_rt.c
index d2ea282..f48328a 100644
--- a/kernel/sched_rt.c
+++ b/kernel/sched_rt.c
@@ -1472,7 +1472,7 @@ static void post_schedule_rt(struct rq *rq)
  * If we are not running and we are not going to reschedule soon, we should
  * try to push tasks away now
  */
-static void task_wake_up_rt(struct rq *rq, struct task_struct *p)
+static void task_woken_rt(struct rq *rq, struct task_struct *p)
 {
 	if (!task_running(rq, p) &&
 	    !test_tsk_need_resched(rq->curr) &&
@@ -1753,7 +1753,7 @@ static const struct sched_class rt_sched_class = {
 	.rq_offline             = rq_offline_rt,
 	.pre_schedule		= pre_schedule_rt,
 	.post_schedule		= post_schedule_rt,
-	.task_wake_up		= task_wake_up_rt,
+	.task_woken		= task_woken_rt,
 	.switched_from		= switched_from_rt,
 #endif
 

[tip:sched/urgent] sched: Remove the cfs_rq dependency from set_task_cpu()

[tip-bot for Peter Zijlstra]

Commit-ID:  88ec22d3edb72b261f8628226cd543589a6d5e1b
Gitweb:     http://git.kernel.org/tip/88ec22d3edb72b261f8628226cd543589a6d5e1b
Author:     Peter Zijlstra <a.p.zijlstra@chello.nl>
AuthorDate: Wed, 16 Dec 2009 18:04:41 +0100
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Wed, 16 Dec 2009 19:01:58 +0100

sched: Remove the cfs_rq dependency from set_task_cpu()

In order to remove the cfs_rq dependency from set_task_cpu() we
need to ensure the task is cfs_rq invariant for all callsites.

The simple approach is to substract cfs_rq->min_vruntime from
se->vruntime on dequeue, and add cfs_rq->min_vruntime on
enqueue.

However, this has the downside of breaking FAIR_SLEEPERS since
we loose the old vruntime as we only maintain the relative
position.

To solve this, we observe that we only migrate runnable tasks,
we do this using deactivate_task(.sleep=0) and
activate_task(.wakeup=0), therefore we can restrain the
min_vruntime invariance to that state.

The only other case is wakeup balancing, since we want to
maintain the old vruntime we cannot make it relative on dequeue,
but since we don't migrate inactive tasks, we can do so right
before we activate it again.

This is where we need the new pre-wakeup hook, we need to call
this while still holding the old rq->lock. We could fold it into
->select_task_rq(), but since that has multiple callsites and
would obfuscate the locking requirements, that seems like a
fudge.

This leaves the fork() case, simply make sure that ->task_fork()
leaves the ->vruntime in a relative state.

This covers all cases where set_task_cpu() gets called, and
ensures it sees a relative vruntime.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
LKML-Reference: <20091216170518.191697025@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
---
 include/linux/sched.h |    2 +-
 kernel/sched.c        |    6 +----
 kernel/sched_fair.c   |   50 +++++++++++++++++++++++++++++++++++++++++++-----
 3 files changed, 46 insertions(+), 12 deletions(-)

diff --git a/include/linux/sched.h b/include/linux/sched.h
index 2c9fa1c..973b2b8 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -1116,7 +1116,7 @@ struct sched_class {
 					 struct task_struct *task);
 
 #ifdef CONFIG_FAIR_GROUP_SCHED
-	void (*moved_group) (struct task_struct *p);
+	void (*moved_group) (struct task_struct *p, int on_rq);
 #endif
 };
 
diff --git a/kernel/sched.c b/kernel/sched.c
index 6c571bd..f92ce63 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -2038,8 +2038,6 @@ task_hot(struct task_struct *p, u64 now, struct sched_domain *sd)
 void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
 {
 	int old_cpu = task_cpu(p);
-	struct cfs_rq *old_cfsrq = task_cfs_rq(p),
-		      *new_cfsrq = cpu_cfs_rq(old_cfsrq, new_cpu);
 
 #ifdef CONFIG_SCHED_DEBUG
 	/*
@@ -2056,8 +2054,6 @@ void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
 		perf_sw_event(PERF_COUNT_SW_CPU_MIGRATIONS,
 				     1, 1, NULL, 0);
 	}
-	p->se.vruntime -= old_cfsrq->min_vruntime -
-					 new_cfsrq->min_vruntime;
 
 	__set_task_cpu(p, new_cpu);
 }
@@ -10102,7 +10098,7 @@ void sched_move_task(struct task_struct *tsk)
 
 #ifdef CONFIG_FAIR_GROUP_SCHED
 	if (tsk->sched_class->moved_group)
-		tsk->sched_class->moved_group(tsk);
+		tsk->sched_class->moved_group(tsk, on_rq);
 #endif
 
 	if (unlikely(running))
diff --git a/kernel/sched_fair.c b/kernel/sched_fair.c
index ec1d271..42ac3c9 100644
--- a/kernel/sched_fair.c
+++ b/kernel/sched_fair.c
@@ -510,6 +510,7 @@ __update_curr(struct cfs_rq *cfs_rq, struct sched_entity *curr,
 	curr->sum_exec_runtime += delta_exec;
 	schedstat_add(cfs_rq, exec_clock, delta_exec);
 	delta_exec_weighted = calc_delta_fair(delta_exec, curr);
+
 	curr->vruntime += delta_exec_weighted;
 	update_min_vruntime(cfs_rq);
 }
@@ -765,16 +766,26 @@ place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial)
 	se->vruntime = vruntime;
 }
 
+#define ENQUEUE_WAKEUP	1
+#define ENQUEUE_MIGRATE 2
+
 static void
-enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int wakeup)
+enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 {
 	/*
+	 * Update the normalized vruntime before updating min_vruntime
+	 * through callig update_curr().
+	 */
+	if (!(flags & ENQUEUE_WAKEUP) || (flags & ENQUEUE_MIGRATE))
+		se->vruntime += cfs_rq->min_vruntime;
+
+	/*
 	 * Update run-time statistics of the 'current'.
 	 */
 	update_curr(cfs_rq);
 	account_entity_enqueue(cfs_rq, se);
 
-	if (wakeup) {
+	if (flags & ENQUEUE_WAKEUP) {
 		place_entity(cfs_rq, se, 0);
 		enqueue_sleeper(cfs_rq, se);
 	}
@@ -828,6 +839,14 @@ dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int sleep)
 		__dequeue_entity(cfs_rq, se);
 	account_entity_dequeue(cfs_rq, se);
 	update_min_vruntime(cfs_rq);
+
+	/*
+	 * Normalize the entity after updating the min_vruntime because the
+	 * update can refer to the ->curr item and we need to reflect this
+	 * movement in our normalized position.
+	 */
+	if (!sleep)
+		se->vruntime -= cfs_rq->min_vruntime;
 }
 
 /*
@@ -1038,13 +1057,19 @@ static void enqueue_task_fair(struct rq *rq, struct task_struct *p, int wakeup)
 {
 	struct cfs_rq *cfs_rq;
 	struct sched_entity *se = &p->se;
+	int flags = 0;
+
+	if (wakeup)
+		flags |= ENQUEUE_WAKEUP;
+	if (p->state == TASK_WAKING)
+		flags |= ENQUEUE_MIGRATE;
 
 	for_each_sched_entity(se) {
 		if (se->on_rq)
 			break;
 		cfs_rq = cfs_rq_of(se);
-		enqueue_entity(cfs_rq, se, wakeup);
-		wakeup = 1;
+		enqueue_entity(cfs_rq, se, flags);
+		flags = ENQUEUE_WAKEUP;
 	}
 
 	hrtick_update(rq);
@@ -1120,6 +1145,14 @@ static void yield_task_fair(struct rq *rq)
 
 #ifdef CONFIG_SMP
 
+static void task_waking_fair(struct rq *rq, struct task_struct *p)
+{
+	struct sched_entity *se = &p->se;
+	struct cfs_rq *cfs_rq = cfs_rq_of(se);
+
+	se->vruntime -= cfs_rq->min_vruntime;
+}
+
 #ifdef CONFIG_FAIR_GROUP_SCHED
 /*
  * effective_load() calculates the load change as seen from the root_task_group
@@ -1978,6 +2011,8 @@ static void task_fork_fair(struct task_struct *p)
 		resched_task(rq->curr);
 	}
 
+	se->vruntime -= cfs_rq->min_vruntime;
+
 	raw_spin_unlock_irqrestore(&rq->lock, flags);
 }
 
@@ -2031,12 +2066,13 @@ static void set_curr_task_fair(struct rq *rq)
 }
 
 #ifdef CONFIG_FAIR_GROUP_SCHED
-static void moved_group_fair(struct task_struct *p)
+static void moved_group_fair(struct task_struct *p, int on_rq)
 {
 	struct cfs_rq *cfs_rq = task_cfs_rq(p);
 
 	update_curr(cfs_rq);
-	place_entity(cfs_rq, &p->se, 1);
+	if (!on_rq)
+		place_entity(cfs_rq, &p->se, 1);
 }
 #endif
 
@@ -2076,6 +2112,8 @@ static const struct sched_class fair_sched_class = {
 	.move_one_task		= move_one_task_fair,
 	.rq_online		= rq_online_fair,
 	.rq_offline		= rq_offline_fair,
+
+	.task_waking		= task_waking_fair,
 #endif
 
 	.set_curr_task          = set_curr_task_fair,

[tip:sched/urgent] sched: Simplify set_task_cpu()

[tip-bot for Peter Zijlstra]

Commit-ID:  738d2be4301007f054541c5c4bf7fb6a361c9b3a
Gitweb:     http://git.kernel.org/tip/738d2be4301007f054541c5c4bf7fb6a361c9b3a
Author:     Peter Zijlstra <a.p.zijlstra@chello.nl>
AuthorDate: Wed, 16 Dec 2009 18:04:42 +0100
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Wed, 16 Dec 2009 19:01:59 +0100

sched: Simplify set_task_cpu()

Rearrange code a bit now that its a simpler function.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
LKML-Reference: <20091216170518.269101883@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
---
 kernel/sched.c |   13 +++++--------
 1 files changed, 5 insertions(+), 8 deletions(-)

diff --git a/kernel/sched.c b/kernel/sched.c
index f92ce63..8a2bfd3 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -2034,11 +2034,8 @@ task_hot(struct task_struct *p, u64 now, struct sched_domain *sd)
 	return delta < (s64)sysctl_sched_migration_cost;
 }
 
-
 void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
 {
-	int old_cpu = task_cpu(p);
-
 #ifdef CONFIG_SCHED_DEBUG
 	/*
 	 * We should never call set_task_cpu() on a blocked task,
@@ -2049,11 +2046,11 @@ void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
 
 	trace_sched_migrate_task(p, new_cpu);
 
-	if (old_cpu != new_cpu) {
-		p->se.nr_migrations++;
-		perf_sw_event(PERF_COUNT_SW_CPU_MIGRATIONS,
-				     1, 1, NULL, 0);
-	}
+	if (task_cpu(p) == new_cpu)
+		return;
+
+	p->se.nr_migrations++;
+	perf_sw_event(PERF_COUNT_SW_CPU_MIGRATIONS, 1, 1, NULL, 0);
 
 	__set_task_cpu(p, new_cpu);
 }

[PATCH 13/12] sched: remove debug check

[Peter Zijlstra]

Since you're running into this warning and not seeing any obvious ill
side-effects (you'd see some serious starvation issues), please stick
this patch in for now.

I can't seem to convince myself if its a real problem or not, and how it
could possibly happen, but my brain is giving up for today.

---
Index: linux-2.6/kernel/sched.c
===================================================================
--- linux-2.6.orig/kernel/sched.c
+++ linux-2.6/kernel/sched.c
@@ -2036,14 +2036,6 @@ task_hot(struct task_struct *p, u64 now,
 
 void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
 {
-#ifdef CONFIG_SCHED_DEBUG
-	/*
-	 * We should never call set_task_cpu() on a blocked task,
-	 * ttwu() will sort out the placement.
-	 */
-	WARN_ON(p->state != TASK_RUNNING && p->state != TASK_WAKING);
-#endif
-
 	trace_sched_migrate_task(p, new_cpu);
 
 	if (task_cpu(p) == new_cpu)

[tip:sched/urgent] sched: Make warning less noisy

[tip-bot for Ingo Molnar]

Commit-ID:  416eb39556a03d1c7e52b0791e9052ccd71db241
Gitweb:     http://git.kernel.org/tip/416eb39556a03d1c7e52b0791e9052ccd71db241
Author:     Ingo Molnar <mingo@elte.hu>
AuthorDate: Thu, 17 Dec 2009 06:05:49 +0100
Committer:  Ingo Molnar <mingo@elte.hu>
CommitDate: Thu, 17 Dec 2009 06:05:49 +0100

sched: Make warning less noisy

Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
LKML-Reference: <20091216170517.807938893@chello.nl>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
---
 kernel/sched.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/kernel/sched.c b/kernel/sched.c
index 8a2bfd3..af7dfa7 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -2041,7 +2041,7 @@ void set_task_cpu(struct task_struct *p, unsigned int new_cpu)
 	 * We should never call set_task_cpu() on a blocked task,
 	 * ttwu() will sort out the placement.
 	 */
-	WARN_ON(p->state != TASK_RUNNING && p->state != TASK_WAKING);
+	WARN_ON_ONCE(p->state != TASK_RUNNING && p->state != TASK_WAKING);
 #endif
 
 	trace_sched_migrate_task(p, new_cpu);