From: Thomas Gleixner <tglx@linutronix.de>
Date: Mon, 6 Jun 2011 12:20:33 +0200
Subject: sched: Move mmdrop to RCU on RT
Git-repo: git://git.kernel.org/pub/scm/linux/kernel/git/rt/linux-rt-devel.git
Git-commit: 80197b77daaf5079c7d85692ff8a4a31a477c273
Patch-mainline: Queued in subsystem maintainer repository
References: SLE Realtime Extension

Takes sleeping locks and calls into the memory allocator, so nothing
we want to do in task switch and oder atomic contexts.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Mike Galbraith <mgalbraith@suse.de>
---
 include/linux/mm_types.h |    4 ++++
 include/linux/sched/mm.h |   11 +++++++++++
 kernel/fork.c            |   13 +++++++++++++
 kernel/sched/core.c      |   18 ++++++++++++++++--
 4 files changed, 44 insertions(+), 2 deletions(-)

--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h
@@ -11,6 +11,7 @@
 #include <linux/completion.h>
 #include <linux/cpumask.h>
 #include <linux/uprobes.h>
+#include <linux/rcupdate.h>
 #include <linux/page-flags-layout.h>
 #include <linux/workqueue.h>
 
@@ -499,6 +500,9 @@ struct mm_struct {
 	bool tlb_flush_batched;
 #endif
 	struct uprobes_state uprobes_state;
+#ifdef CONFIG_PREEMPT_RT_BASE
+	struct rcu_head delayed_drop;
+#endif
 #ifdef CONFIG_HUGETLB_PAGE
 	atomic_long_t hugetlb_usage;
 #endif
--- a/include/linux/sched/mm.h
+++ b/include/linux/sched/mm.h
@@ -42,6 +42,17 @@ static inline void mmdrop(struct mm_stru
 		__mmdrop(mm);
 }
 
+#ifdef CONFIG_PREEMPT_RT_BASE
+extern void __mmdrop_delayed(struct rcu_head *rhp);
+static inline void mmdrop_delayed(struct mm_struct *mm)
+{
+	if (atomic_dec_and_test(&mm->mm_count))
+		call_rcu(&mm->delayed_drop, __mmdrop_delayed);
+}
+#else
+# define mmdrop_delayed(mm)	mmdrop(mm)
+#endif
+
 static inline void mmdrop_async_fn(struct work_struct *work)
 {
 	struct mm_struct *mm = container_of(work, struct mm_struct, async_put_work);
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -912,6 +912,19 @@ void __mmdrop(struct mm_struct *mm)
 }
 EXPORT_SYMBOL_GPL(__mmdrop);
 
+#ifdef CONFIG_PREEMPT_RT_BASE
+/*
+ * RCU callback for delayed mm drop. Not strictly rcu, but we don't
+ * want another facility to make this work.
+ */
+void __mmdrop_delayed(struct rcu_head *rhp)
+{
+	struct mm_struct *mm = container_of(rhp, struct mm_struct, delayed_drop);
+
+	__mmdrop(mm);
+}
+#endif
+
 static inline void __mmput(struct mm_struct *mm)
 {
 	VM_BUG_ON(atomic_read(&mm->mm_users));
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -2794,8 +2794,12 @@ static struct rq *finish_task_switch(str
 	finish_arch_post_lock_switch();
 
 	fire_sched_in_preempt_notifiers(current);
+	/*
+	 * We use mmdrop_delayed() here so we don't have to do the
+	 * full __mmdrop() when we are the last user.
+	 */
 	if (mm)
-		mmdrop(mm);
+		mmdrop_delayed(mm);
 	if (unlikely(prev_state == TASK_DEAD)) {
 		if (prev->sched_class->task_dead)
 			prev->sched_class->task_dead(prev);
@@ -5650,6 +5654,8 @@ void sched_setnuma(struct task_struct *p
 #endif /* CONFIG_NUMA_BALANCING */
 
 #ifdef CONFIG_HOTPLUG_CPU
+static DEFINE_PER_CPU(struct mm_struct *, idle_last_mm);
+
 /*
  * Ensure that the idle task is using init_mm right before its CPU goes
  * offline.
@@ -5664,7 +5670,11 @@ void idle_task_exit(void)
 		switch_mm(mm, &init_mm, current);
 		finish_arch_post_lock_switch();
 	}
-	mmdrop(mm);
+	/*
+	 * Defer the cleanup to an alive cpu. On RT we can neither
+	 * call mmdrop() nor mmdrop_delayed() from here.
+	 */
+	per_cpu(idle_last_mm, smp_processor_id()) = mm;
 }
 
 /*
@@ -5989,6 +5999,10 @@ int sched_cpu_dying(unsigned int cpu)
 	update_max_interval();
 	nohz_balance_exit_idle(cpu);
 	hrtick_clear(rq);
+	if (per_cpu(idle_last_mm, cpu)) {
+		mmdrop_delayed(per_cpu(idle_last_mm, cpu));
+		per_cpu(idle_last_mm, cpu) = NULL;
+	}
 	return 0;
 }
 #endif