cgroup: Add generic cgroup subsystem permission checks.
San Mehat [Thu, 21 May 2009 21:10:06 +0000 (14:10 -0700)]
    Rather than using explicit euid == 0 checks when trying to move
tasks into a cgroup via CFS, move permission checks into each
specific cgroup subsystem. If a subsystem does not specify a
'can_attach' handler, then we fall back to doing our checks the old way.

    This way non-root processes can add arbitrary processes to
a cgroup if all the registered subsystems on that cgroup agree.

    Also change explicit euid == 0 check to CAP_SYS_ADMIN

Change-Id: Ife77a6dca056cce4eb922d447a809189234a1549
Signed-off-by: San Mehat <san@google.com>
Reviewed-on: http://git-master/r/111033
Reviewed-by: Prashant Malani <pmalani@nvidia.com>
Tested-by: Prashant Malani <pmalani@nvidia.com>
Reviewed-by: Automatic_Commit_Validation_User
Reviewed-by: Bo Yan <byan@nvidia.com>

kernel/cgroup.c
kernel/cgroup_freezer.c
kernel/cpuset.c
kernel/sched.c [new file with mode: 0644]

index 15db92d..145deeb 100644 (file)
@@ -60,6 +60,7 @@
 #include <linux/eventfd.h>
 #include <linux/poll.h>
 #include <linux/flex_array.h> /* used in cgroup_attach_proc */
+#include <linux/capability.h>
 
 #include <linux/atomic.h>
 
@@ -1925,6 +1926,15 @@ int cgroup_attach_task(struct cgroup *cgrp, struct task_struct *tsk)
                                failed_ss = ss;
                                goto out;
                        }
+               } else if (!capable(CAP_SYS_ADMIN)) {
+                       const struct cred *cred = current_cred(), *tcred;
+
+                       /* No can_attach() - check perms generically */
+                       tcred = __task_cred(tsk);
+                       if (cred->euid != tcred->uid &&
+                           cred->euid != tcred->suid) {
+                               return -EACCES;
+                       }
                }
        }
 
index f86e939..5c248e5 100644 (file)
@@ -168,6 +168,14 @@ static int freezer_can_attach(struct cgroup *new_cgroup,
        struct freezer *freezer;
        struct task_struct *task;
 
+       if ((current != task) && (!capable(CAP_SYS_ADMIN))) {
+               const struct cred *cred = current_cred(), *tcred;
+
+               tcred = __task_cred(task);
+               if (cred->euid != tcred->uid && cred->euid != tcred->suid)
+                       return -EPERM;
+       }
+
        /*
         * Anything frozen can't move or be moved to/from.
         */
index 14f7070..48b90d3 100644 (file)
@@ -1362,6 +1362,41 @@ static int fmeter_getrate(struct fmeter *fmp)
        return val;
 }
 
+/* Called by cgroups to determine if a cpuset is usable; cgroup_mutex held */
+static int cpuset_can_attach(struct cgroup_subsys *ss, struct cgroup *cont,
+                            struct task_struct *tsk)
+{
+       struct cpuset *cs = cgroup_cs(cont);
+
+       if ((current != task) && (!capable(CAP_SYS_ADMIN))) {
+               const struct cred *cred = current_cred(), *tcred;
+
+               if (cred->euid != tcred->uid && cred->euid != tcred->suid)
+                       return -EPERM;
+       }
+       if (cpumask_empty(cs->cpus_allowed) || nodes_empty(cs->mems_allowed))
+               return -ENOSPC;
+
+       /*
+        * Kthreads bound to specific cpus cannot be moved to a new cpuset; we
+        * cannot change their cpu affinity and isolating such threads by their
+        * set of allowed nodes is unnecessary.  Thus, cpusets are not
+        * applicable for such threads.  This prevents checking for success of
+        * set_cpus_allowed_ptr() on all attached tasks before cpus_allowed may
+        * be changed.
+        */
+       if (tsk->flags & PF_THREAD_BOUND)
+               return -EINVAL;
+
+       return 0;
+}
+
+static int cpuset_can_attach_task(struct cgroup *cgrp, struct task_struct *task)
+{
+       return security_task_setscheduler(task);
+}
+
 /*
  * Protected by cgroup_lock. The nodemasks must be stored globally because
  * dynamically allocating them is not allowed in can_attach, and they must
diff --git a/kernel/sched.c b/kernel/sched.c
new file mode 100644 (file)
index 0000000..279d255
--- /dev/null
@@ -0,0 +1,9367 @@
+/*
+ *  kernel/sched.c
+ *
+ *  Kernel scheduler and related syscalls
+ *
+ *  Copyright (C) 1991-2002  Linus Torvalds
+ *
+ *  1996-12-23  Modified by Dave Grothe to fix bugs in semaphores and
+ *             make semaphores SMP safe
+ *  1998-11-19 Implemented schedule_timeout() and related stuff
+ *             by Andrea Arcangeli
+ *  2002-01-04 New ultra-scalable O(1) scheduler by Ingo Molnar:
+ *             hybrid priority-list and round-robin design with
+ *             an array-switch method of distributing timeslices
+ *             and per-CPU runqueues.  Cleanups and useful suggestions
+ *             by Davide Libenzi, preemptible kernel bits by Robert Love.
+ *  2003-09-03 Interactivity tuning by Con Kolivas.
+ *  2004-04-02 Scheduler domains code by Nick Piggin
+ *  2007-04-15  Work begun on replacing all interactivity tuning with a
+ *              fair scheduling design by Con Kolivas.
+ *  2007-05-05  Load balancing (smp-nice) and other improvements
+ *              by Peter Williams
+ *  2007-05-06  Interactivity improvements to CFS by Mike Galbraith
+ *  2007-07-01  Group scheduling enhancements by Srivatsa Vaddagiri
+ *  2007-11-29  RT balancing improvements by Steven Rostedt, Gregory Haskins,
+ *              Thomas Gleixner, Mike Kravetz
+ */
+
+#include <linux/mm.h>
+#include <linux/module.h>
+#include <linux/nmi.h>
+#include <linux/init.h>
+#include <linux/uaccess.h>
+#include <linux/highmem.h>
+#include <asm/mmu_context.h>
+#include <linux/interrupt.h>
+#include <linux/capability.h>
+#include <linux/completion.h>
+#include <linux/kernel_stat.h>
+#include <linux/debug_locks.h>
+#include <linux/perf_event.h>
+#include <linux/security.h>
+#include <linux/notifier.h>
+#include <linux/profile.h>
+#include <linux/freezer.h>
+#include <linux/vmalloc.h>
+#include <linux/blkdev.h>
+#include <linux/delay.h>
+#include <linux/pid_namespace.h>
+#include <linux/smp.h>
+#include <linux/threads.h>
+#include <linux/timer.h>
+#include <linux/rcupdate.h>
+#include <linux/cpu.h>
+#include <linux/cpuset.h>
+#include <linux/percpu.h>
+#include <linux/proc_fs.h>
+#include <linux/seq_file.h>
+#include <linux/stop_machine.h>
+#include <linux/sysctl.h>
+#include <linux/syscalls.h>
+#include <linux/times.h>
+#include <linux/tsacct_kern.h>
+#include <linux/kprobes.h>
+#include <linux/delayacct.h>
+#include <linux/unistd.h>
+#include <linux/pagemap.h>
+#include <linux/hrtimer.h>
+#include <linux/tick.h>
+#include <linux/debugfs.h>
+#include <linux/ctype.h>
+#include <linux/ftrace.h>
+#include <linux/slab.h>
+
+#include <asm/tlb.h>
+#include <asm/irq_regs.h>
+#include <asm/mutex.h>
+#ifdef CONFIG_PARAVIRT
+#include <asm/paravirt.h>
+#endif
+
+#include "sched_cpupri.h"
+#include "workqueue_sched.h"
+#include "sched_autogroup.h"
+
+#define CREATE_TRACE_POINTS
+#include <trace/events/sched.h>
+
+/*
+ * Convert user-nice values [ -20 ... 0 ... 19 ]
+ * to static priority [ MAX_RT_PRIO..MAX_PRIO-1 ],
+ * and back.
+ */
+#define NICE_TO_PRIO(nice)     (MAX_RT_PRIO + (nice) + 20)
+#define PRIO_TO_NICE(prio)     ((prio) - MAX_RT_PRIO - 20)
+#define TASK_NICE(p)           PRIO_TO_NICE((p)->static_prio)
+
+/*
+ * 'User priority' is the nice value converted to something we
+ * can work with better when scaling various scheduler parameters,
+ * it's a [ 0 ... 39 ] range.
+ */
+#define USER_PRIO(p)           ((p)-MAX_RT_PRIO)
+#define TASK_USER_PRIO(p)      USER_PRIO((p)->static_prio)
+#define MAX_USER_PRIO          (USER_PRIO(MAX_PRIO))
+
+/*
+ * Helpers for converting nanosecond timing to jiffy resolution
+ */
+#define NS_TO_JIFFIES(TIME)    ((unsigned long)(TIME) / (NSEC_PER_SEC / HZ))
+
+#define NICE_0_LOAD            SCHED_LOAD_SCALE
+#define NICE_0_SHIFT           SCHED_LOAD_SHIFT
+
+/*
+ * These are the 'tuning knobs' of the scheduler:
+ *
+ * default timeslice is 100 msecs (used only for SCHED_RR tasks).
+ * Timeslices get refilled after they expire.
+ */
+#define DEF_TIMESLICE          (100 * HZ / 1000)
+
+/*
+ * single value that denotes runtime == period, ie unlimited time.
+ */
+#define RUNTIME_INF    ((u64)~0ULL)
+
+static inline int rt_policy(int policy)
+{
+       if (policy == SCHED_FIFO || policy == SCHED_RR)
+               return 1;
+       return 0;
+}
+
+static inline int task_has_rt_policy(struct task_struct *p)
+{
+       return rt_policy(p->policy);
+}
+
+/*
+ * This is the priority-queue data structure of the RT scheduling class:
+ */
+struct rt_prio_array {
+       DECLARE_BITMAP(bitmap, MAX_RT_PRIO+1); /* include 1 bit for delimiter */
+       struct list_head queue[MAX_RT_PRIO];
+};
+
+struct rt_bandwidth {
+       /* nests inside the rq lock: */
+       raw_spinlock_t          rt_runtime_lock;
+       ktime_t                 rt_period;
+       u64                     rt_runtime;
+       struct hrtimer          rt_period_timer;
+};
+
+static struct rt_bandwidth def_rt_bandwidth;
+
+static int do_sched_rt_period_timer(struct rt_bandwidth *rt_b, int overrun);
+
+static enum hrtimer_restart sched_rt_period_timer(struct hrtimer *timer)
+{
+       struct rt_bandwidth *rt_b =
+               container_of(timer, struct rt_bandwidth, rt_period_timer);
+       ktime_t now;
+       int overrun;
+       int idle = 0;
+
+       for (;;) {
+               now = hrtimer_cb_get_time(timer);
+               overrun = hrtimer_forward(timer, now, rt_b->rt_period);
+
+               if (!overrun)
+                       break;
+
+               idle = do_sched_rt_period_timer(rt_b, overrun);
+       }
+
+       return idle ? HRTIMER_NORESTART : HRTIMER_RESTART;
+}
+
+static
+void init_rt_bandwidth(struct rt_bandwidth *rt_b, u64 period, u64 runtime)
+{
+       rt_b->rt_period = ns_to_ktime(period);
+       rt_b->rt_runtime = runtime;
+
+       raw_spin_lock_init(&rt_b->rt_runtime_lock);
+
+       hrtimer_init(&rt_b->rt_period_timer,
+                       CLOCK_MONOTONIC, HRTIMER_MODE_REL);
+       rt_b->rt_period_timer.function = sched_rt_period_timer;
+}
+
+static inline int rt_bandwidth_enabled(void)
+{
+       return sysctl_sched_rt_runtime >= 0;
+}
+
+static void start_rt_bandwidth(struct rt_bandwidth *rt_b)
+{
+       ktime_t now;
+
+       if (!rt_bandwidth_enabled() || rt_b->rt_runtime == RUNTIME_INF)
+               return;
+
+       if (hrtimer_active(&rt_b->rt_period_timer))
+               return;
+
+       raw_spin_lock(&rt_b->rt_runtime_lock);
+       for (;;) {
+               unsigned long delta;
+               ktime_t soft, hard;
+
+               if (hrtimer_active(&rt_b->rt_period_timer))
+                       break;
+
+               now = hrtimer_cb_get_time(&rt_b->rt_period_timer);
+               hrtimer_forward(&rt_b->rt_period_timer, now, rt_b->rt_period);
+
+               soft = hrtimer_get_softexpires(&rt_b->rt_period_timer);
+               hard = hrtimer_get_expires(&rt_b->rt_period_timer);
+               delta = ktime_to_ns(ktime_sub(hard, soft));
+               __hrtimer_start_range_ns(&rt_b->rt_period_timer, soft, delta,
+                               HRTIMER_MODE_ABS_PINNED, 0);
+       }
+       raw_spin_unlock(&rt_b->rt_runtime_lock);
+}
+
+#ifdef CONFIG_RT_GROUP_SCHED
+static void destroy_rt_bandwidth(struct rt_bandwidth *rt_b)
+{
+       hrtimer_cancel(&rt_b->rt_period_timer);
+}
+#endif
+
+/*
+ * sched_domains_mutex serializes calls to init_sched_domains,
+ * detach_destroy_domains and partition_sched_domains.
+ */
+static DEFINE_MUTEX(sched_domains_mutex);
+
+#ifdef CONFIG_CGROUP_SCHED
+
+#include <linux/cgroup.h>
+
+struct cfs_rq;
+
+static LIST_HEAD(task_groups);
+
+/* task group related information */
+struct task_group {
+       struct cgroup_subsys_state css;
+
+#ifdef CONFIG_FAIR_GROUP_SCHED
+       /* schedulable entities of this group on each cpu */
+       struct sched_entity **se;
+       /* runqueue "owned" by this group on each cpu */
+       struct cfs_rq **cfs_rq;
+       unsigned long shares;
+
+       atomic_t load_weight;
+#endif
+
+#ifdef CONFIG_RT_GROUP_SCHED
+       struct sched_rt_entity **rt_se;
+       struct rt_rq **rt_rq;
+
+       struct rt_bandwidth rt_bandwidth;
+#endif
+
+       struct rcu_head rcu;
+       struct list_head list;
+
+       struct task_group *parent;
+       struct list_head siblings;
+       struct list_head children;
+
+#ifdef CONFIG_SCHED_AUTOGROUP
+       struct autogroup *autogroup;
+#endif
+};
+
+/* task_group_lock serializes the addition/removal of task groups */
+static DEFINE_SPINLOCK(task_group_lock);
+
+#ifdef CONFIG_FAIR_GROUP_SCHED
+
+# define ROOT_TASK_GROUP_LOAD  NICE_0_LOAD
+
+/*
+ * A weight of 0 or 1 can cause arithmetics problems.
+ * A weight of a cfs_rq is the sum of weights of which entities
+ * are queued on this cfs_rq, so a weight of a entity should not be
+ * too large, so as the shares value of a task group.
+ * (The default weight is 1024 - so there's no practical
+ *  limitation from this.)
+ */
+#define MIN_SHARES     (1UL <<  1)
+#define MAX_SHARES     (1UL << 18)
+
+static int root_task_group_load = ROOT_TASK_GROUP_LOAD;
+#endif
+
+/* Default task group.
+ *     Every task in system belong to this group at bootup.
+ */
+struct task_group root_task_group;
+
+#endif /* CONFIG_CGROUP_SCHED */
+
+/* CFS-related fields in a runqueue */
+struct cfs_rq {
+       struct load_weight load;
+       unsigned long nr_running;
+
+       u64 exec_clock;
+       u64 min_vruntime;
+#ifndef CONFIG_64BIT
+       u64 min_vruntime_copy;
+#endif
+
+       struct rb_root tasks_timeline;
+       struct rb_node *rb_leftmost;
+
+       struct list_head tasks;
+       struct list_head *balance_iterator;
+
+       /*
+        * 'curr' points to currently running entity on this cfs_rq.
+        * It is set to NULL otherwise (i.e when none are currently running).
+        */
+       struct sched_entity *curr, *next, *last, *skip;
+
+#ifdef CONFIG_SCHED_DEBUG
+       unsigned int nr_spread_over;
+#endif
+
+#ifdef CONFIG_FAIR_GROUP_SCHED
+       struct rq *rq;  /* cpu runqueue to which this cfs_rq is attached */
+
+       /*
+        * leaf cfs_rqs are those that hold tasks (lowest schedulable entity in
+        * a hierarchy). Non-leaf lrqs hold other higher schedulable entities
+        * (like users, containers etc.)
+        *
+        * leaf_cfs_rq_list ties together list of leaf cfs_rq's in a cpu. This
+        * list is used during load balance.
+        */
+       int on_list;
+       struct list_head leaf_cfs_rq_list;
+       struct task_group *tg;  /* group that "owns" this runqueue */
+
+#ifdef CONFIG_SMP
+       /*
+        * the part of load.weight contributed by tasks
+        */
+       unsigned long task_weight;
+
+       /*
+        *   h_load = weight * f(tg)
+        *
+        * Where f(tg) is the recursive weight fraction assigned to
+        * this group.
+        */
+       unsigned long h_load;
+
+       /*
+        * Maintaining per-cpu shares distribution for group scheduling
+        *
+        * load_stamp is the last time we updated the load average
+        * load_last is the last time we updated the load average and saw load
+        * load_unacc_exec_time is currently unaccounted execution time
+        */
+       u64 load_avg;
+       u64 load_period;
+       u64 load_stamp, load_last, load_unacc_exec_time;
+
+       unsigned long load_contribution;
+#endif
+#endif
+};
+
+/* Real-Time classes' related field in a runqueue: */
+struct rt_rq {
+       struct rt_prio_array active;
+       unsigned long rt_nr_running;
+#if defined CONFIG_SMP || defined CONFIG_RT_GROUP_SCHED
+       struct {
+               int curr; /* highest queued rt task prio */
+#ifdef CONFIG_SMP
+               int next; /* next highest */
+#endif
+       } highest_prio;
+#endif
+#ifdef CONFIG_SMP
+       unsigned long rt_nr_migratory;
+       unsigned long rt_nr_total;
+       int overloaded;
+       struct plist_head pushable_tasks;
+#endif
+       int rt_throttled;
+       u64 rt_time;
+       u64 rt_runtime;
+       /* Nests inside the rq lock: */
+       raw_spinlock_t rt_runtime_lock;
+
+#ifdef CONFIG_RT_GROUP_SCHED
+       unsigned long rt_nr_boosted;
+
+       struct rq *rq;
+       struct list_head leaf_rt_rq_list;
+       struct task_group *tg;
+#endif
+};
+
+#ifdef CONFIG_SMP
+
+/*
+ * We add the notion of a root-domain which will be used to define per-domain
+ * variables. Each exclusive cpuset essentially defines an island domain by
+ * fully partitioning the member cpus from any other cpuset. Whenever a new
+ * exclusive cpuset is created, we also create and attach a new root-domain
+ * object.
+ *
+ */
+struct root_domain {
+       atomic_t refcount;
+       atomic_t rto_count;
+       struct rcu_head rcu;
+       cpumask_var_t span;
+       cpumask_var_t online;
+
+       /*
+        * The "RT overload" flag: it gets set if a CPU has more than
+        * one runnable RT task.
+        */
+       cpumask_var_t rto_mask;
+       struct cpupri cpupri;
+};
+
+/*
+ * By default the system creates a single root-domain with all cpus as
+ * members (mimicking the global state we have today).
+ */
+static struct root_domain def_root_domain;
+
+#endif /* CONFIG_SMP */
+
+/*
+ * This is the main, per-CPU runqueue data structure.
+ *
+ * Locking rule: those places that want to lock multiple runqueues
+ * (such as the load balancing or the thread migration code), lock
+ * acquire operations must be ordered by ascending &runqueue.
+ */
+struct rq {
+       /* runqueue lock: */
+       raw_spinlock_t lock;
+
+       /*
+        * nr_running and cpu_load should be in the same cacheline because
+        * remote CPUs use both these fields when doing load calculation.
+        */
+       unsigned long nr_running;
+       #define CPU_LOAD_IDX_MAX 5
+       unsigned long cpu_load[CPU_LOAD_IDX_MAX];
+       unsigned long last_load_update_tick;
+#ifdef CONFIG_NO_HZ
+       u64 nohz_stamp;
+       unsigned char nohz_balance_kick;
+#endif
+       int skip_clock_update;
+
+       /* capture load from *all* tasks on this cpu: */
+       struct load_weight load;
+       unsigned long nr_load_updates;
+       u64 nr_switches;
+
+       struct cfs_rq cfs;
+       struct rt_rq rt;
+
+#ifdef CONFIG_FAIR_GROUP_SCHED
+       /* list of leaf cfs_rq on this cpu: */
+       struct list_head leaf_cfs_rq_list;
+#endif
+#ifdef CONFIG_RT_GROUP_SCHED
+       struct list_head leaf_rt_rq_list;
+#endif
+
+       /*
+        * This is part of a global counter where only the total sum
+        * over all CPUs matters. A task can increase this counter on
+        * one CPU and if it got migrated afterwards it may decrease
+        * it on another CPU. Always updated under the runqueue lock:
+        */
+       unsigned long nr_uninterruptible;
+
+       struct task_struct *curr, *idle, *stop;
+       unsigned long next_balance;
+       struct mm_struct *prev_mm;
+
+       u64 clock;
+       u64 clock_task;
+
+       atomic_t nr_iowait;
+
+#ifdef CONFIG_SMP
+       struct root_domain *rd;
+       struct sched_domain *sd;
+
+       unsigned long cpu_power;
+
+       unsigned char idle_at_tick;
+       /* For active balancing */
+       int post_schedule;
+       int active_balance;
+       int push_cpu;
+       struct cpu_stop_work active_balance_work;
+       /* cpu of this runqueue: */
+       int cpu;
+       int online;
+
+       unsigned long avg_load_per_task;
+
+       u64 rt_avg;
+       u64 age_stamp;
+       u64 idle_stamp;
+       u64 avg_idle;
+#endif
+
+#ifdef CONFIG_IRQ_TIME_ACCOUNTING
+       u64 prev_irq_time;
+#endif
+#ifdef CONFIG_PARAVIRT
+       u64 prev_steal_time;
+#endif
+#ifdef CONFIG_PARAVIRT_TIME_ACCOUNTING
+       u64 prev_steal_time_rq;
+#endif
+
+       /* calc_load related fields */
+       unsigned long calc_load_update;
+       long calc_load_active;
+
+#ifdef CONFIG_SCHED_HRTICK
+#ifdef CONFIG_SMP
+       int hrtick_csd_pending;
+       struct call_single_data hrtick_csd;
+#endif
+       struct hrtimer hrtick_timer;
+#endif
+
+#ifdef CONFIG_SCHEDSTATS
+       /* latency stats */
+       struct sched_info rq_sched_info;
+       unsigned long long rq_cpu_time;
+       /* could above be rq->cfs_rq.exec_clock + rq->rt_rq.rt_runtime ? */
+
+       /* sys_sched_yield() stats */
+       unsigned int yld_count;
+
+       /* schedule() stats */
+       unsigned int sched_switch;
+       unsigned int sched_count;
+       unsigned int sched_goidle;
+
+       /* try_to_wake_up() stats */
+       unsigned int ttwu_count;
+       unsigned int ttwu_local;
+#endif
+
+#ifdef CONFIG_SMP
+       struct task_struct *wake_list;
+#endif
+};
+
+static DEFINE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
+
+
+static void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags);
+
+static inline int cpu_of(struct rq *rq)
+{
+#ifdef CONFIG_SMP
+       return rq->cpu;
+#else
+       return 0;
+#endif
+}
+
+#define rcu_dereference_check_sched_domain(p) \
+       rcu_dereference_check((p), \
+                             lockdep_is_held(&sched_domains_mutex))
+
+/*
+ * The domain tree (rq->sd) is protected by RCU's quiescent state transition.
+ * See detach_destroy_domains: synchronize_sched for details.
+ *
+ * The domain tree of any CPU may only be accessed from within
+ * preempt-disabled sections.
+ */
+#define for_each_domain(cpu, __sd) \
+       for (__sd = rcu_dereference_check_sched_domain(cpu_rq(cpu)->sd); __sd; __sd = __sd->parent)
+
+#define cpu_rq(cpu)            (&per_cpu(runqueues, (cpu)))
+#define this_rq()              (&__get_cpu_var(runqueues))
+#define task_rq(p)             cpu_rq(task_cpu(p))
+#define cpu_curr(cpu)          (cpu_rq(cpu)->curr)
+#define raw_rq()               (&__raw_get_cpu_var(runqueues))
+
+#ifdef CONFIG_CGROUP_SCHED
+
+/*
+ * Return the group to which this tasks belongs.
+ *
+ * We use task_subsys_state_check() and extend the RCU verification with
+ * pi->lock and rq->lock because cpu_cgroup_attach() holds those locks for each
+ * task it moves into the cgroup. Therefore by holding either of those locks,
+ * we pin the task to the current cgroup.
+ */
+static inline struct task_group *task_group(struct task_struct *p)
+{
+       struct task_group *tg;
+       struct cgroup_subsys_state *css;
+
+       css = task_subsys_state_check(p, cpu_cgroup_subsys_id,
+                       lockdep_is_held(&p->pi_lock) ||
+                       lockdep_is_held(&task_rq(p)->lock));
+       tg = container_of(css, struct task_group, css);
+
+       return autogroup_task_group(p, tg);
+}
+
+/* Change a task's cfs_rq and parent entity if it moves across CPUs/groups */
+static inline void set_task_rq(struct task_struct *p, unsigned int cpu)
+{
+#ifdef CONFIG_FAIR_GROUP_SCHED
+       p->se.cfs_rq = task_group(p)->cfs_rq[cpu];
+       p->se.parent = task_group(p)->se[cpu];
+#endif
+
+#ifdef CONFIG_RT_GROUP_SCHED
+       p->rt.rt_rq  = task_group(p)->rt_rq[cpu];
+       p->rt.parent = task_group(p)->rt_se[cpu];
+#endif
+}
+
+#else /* CONFIG_CGROUP_SCHED */
+
+static inline void set_task_rq(struct task_struct *p, unsigned int cpu) { }
+static inline struct task_group *task_group(struct task_struct *p)
+{
+       return NULL;
+}
+
+#endif /* CONFIG_CGROUP_SCHED */
+
+static void update_rq_clock_task(struct rq *rq, s64 delta);
+
+static void update_rq_clock(struct rq *rq)
+{
+       s64 delta;
+
+       if (rq->skip_clock_update > 0)
+               return;
+
+       delta = sched_clock_cpu(cpu_of(rq)) - rq->clock;
+       rq->clock += delta;
+       update_rq_clock_task(rq, delta);
+}
+
+/*
+ * Tunables that become constants when CONFIG_SCHED_DEBUG is off:
+ */
+#ifdef CONFIG_SCHED_DEBUG
+# define const_debug __read_mostly
+#else
+# define const_debug static const
+#endif
+
+/**
+ * runqueue_is_locked - Returns true if the current cpu runqueue is locked
+ * @cpu: the processor in question.
+ *
+ * This interface allows printk to be called with the runqueue lock
+ * held and know whether or not it is OK to wake up the klogd.
+ */
+int runqueue_is_locked(int cpu)
+{
+       return raw_spin_is_locked(&cpu_rq(cpu)->lock);
+}
+
+/*
+ * Debugging: various feature bits
+ */
+
+#define SCHED_FEAT(name, enabled)      \
+       __SCHED_FEAT_##name ,
+
+enum {
+#include "sched_features.h"
+};
+
+#undef SCHED_FEAT
+
+#define SCHED_FEAT(name, enabled)      \
+       (1UL << __SCHED_FEAT_##name) * enabled |
+
+const_debug unsigned int sysctl_sched_features =
+#include "sched_features.h"
+       0;
+
+#undef SCHED_FEAT
+
+#ifdef CONFIG_SCHED_DEBUG
+#define SCHED_FEAT(name, enabled)      \
+       #name ,
+
+static __read_mostly char *sched_feat_names[] = {
+#include "sched_features.h"
+       NULL
+};
+
+#undef SCHED_FEAT
+
+static int sched_feat_show(struct seq_file *m, void *v)
+{
+       int i;
+
+       for (i = 0; sched_feat_names[i]; i++) {
+               if (!(sysctl_sched_features & (1UL << i)))
+                       seq_puts(m, "NO_");
+               seq_printf(m, "%s ", sched_feat_names[i]);
+       }
+       seq_puts(m, "\n");
+
+       return 0;
+}
+
+static ssize_t
+sched_feat_write(struct file *filp, const char __user *ubuf,
+               size_t cnt, loff_t *ppos)
+{
+       char buf[64];
+       char *cmp;
+       int neg = 0;
+       int i;
+
+       if (cnt > 63)
+               cnt = 63;
+
+       if (copy_from_user(&buf, ubuf, cnt))
+               return -EFAULT;
+
+       buf[cnt] = 0;
+       cmp = strstrip(buf);
+
+       if (strncmp(cmp, "NO_", 3) == 0) {
+               neg = 1;
+               cmp += 3;
+       }
+
+       for (i = 0; sched_feat_names[i]; i++) {
+               if (strcmp(cmp, sched_feat_names[i]) == 0) {
+                       if (neg)
+                               sysctl_sched_features &= ~(1UL << i);
+                       else
+                               sysctl_sched_features |= (1UL << i);
+                       break;
+               }
+       }
+
+       if (!sched_feat_names[i])
+               return -EINVAL;
+
+       *ppos += cnt;
+
+       return cnt;
+}
+
+static int sched_feat_open(struct inode *inode, struct file *filp)
+{
+       return single_open(filp, sched_feat_show, NULL);
+}
+
+static const struct file_operations sched_feat_fops = {
+       .open           = sched_feat_open,
+       .write          = sched_feat_write,
+       .read           = seq_read,
+       .llseek         = seq_lseek,
+       .release        = single_release,
+};
+
+static __init int sched_init_debug(void)
+{
+       debugfs_create_file("sched_features", 0644, NULL, NULL,
+                       &sched_feat_fops);
+
+       return 0;
+}
+late_initcall(sched_init_debug);
+
+#endif
+
+#define sched_feat(x) (sysctl_sched_features & (1UL << __SCHED_FEAT_##x))
+
+/*
+ * Number of tasks to iterate in a single balance run.
+ * Limited because this is done with IRQs disabled.
+ */
+const_debug unsigned int sysctl_sched_nr_migrate = 32;
+
+/*
+ * period over which we average the RT time consumption, measured
+ * in ms.
+ *
+ * default: 1s
+ */
+const_debug unsigned int sysctl_sched_time_avg = MSEC_PER_SEC;
+
+/*
+ * period over which we measure -rt task cpu usage in us.
+ * default: 1s
+ */
+unsigned int sysctl_sched_rt_period = 1000000;
+
+static __read_mostly int scheduler_running;
+
+/*
+ * part of the period that we allow rt tasks to run in us.
+ * default: 0.95s
+ */
+int sysctl_sched_rt_runtime = 950000;
+
+static inline u64 global_rt_period(void)
+{
+       return (u64)sysctl_sched_rt_period * NSEC_PER_USEC;
+}
+
+static inline u64 global_rt_runtime(void)
+{
+       if (sysctl_sched_rt_runtime < 0)
+               return RUNTIME_INF;
+
+       return (u64)sysctl_sched_rt_runtime * NSEC_PER_USEC;
+}
+
+#ifndef prepare_arch_switch
+# define prepare_arch_switch(next)     do { } while (0)
+#endif
+#ifndef finish_arch_switch
+# define finish_arch_switch(prev)      do { } while (0)
+#endif
+
+static inline int task_current(struct rq *rq, struct task_struct *p)
+{
+       return rq->curr == p;
+}
+
+static inline int task_running(struct rq *rq, struct task_struct *p)
+{
+#ifdef CONFIG_SMP
+       return p->on_cpu;
+#else
+       return task_current(rq, p);
+#endif
+}
+
+#ifndef __ARCH_WANT_UNLOCKED_CTXSW
+static inline void prepare_lock_switch(struct rq *rq, struct task_struct *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->on_cpu = 1;
+#endif
+}
+
+static inline void finish_lock_switch(struct rq *rq, struct task_struct *prev)
+{
+#ifdef CONFIG_SMP
+       /*
+        * After ->on_cpu 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->on_cpu = 0;
+#endif
+#ifdef CONFIG_DEBUG_SPINLOCK
+       /* this is a valid case when another task releases the spinlock */
+       rq->lock.owner = current;
+#endif
+       /*
+        * If we are tracking spinlock dependencies then we have to
+        * fix up the runqueue lock - which gets 'carried over' from
+        * prev into current:
+        */
+       spin_acquire(&rq->lock.dep_map, 0, 0, _THIS_IP_);
+
+       raw_spin_unlock_irq(&rq->lock);
+}
+
+#else /* __ARCH_WANT_UNLOCKED_CTXSW */
+static inline void prepare_lock_switch(struct rq *rq, struct task_struct *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->on_cpu = 1;
+#endif
+#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
+       raw_spin_unlock_irq(&rq->lock);
+#else
+       raw_spin_unlock(&rq->lock);
+#endif
+}
+
+static inline void finish_lock_switch(struct rq *rq, struct task_struct *prev)
+{
+#ifdef CONFIG_SMP
+       /*
+        * After ->on_cpu 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->on_cpu = 0;
+#endif
+#ifndef __ARCH_WANT_INTERRUPTS_ON_CTXSW
+       local_irq_enable();
+#endif
+}
+#endif /* __ARCH_WANT_UNLOCKED_CTXSW */
+
+/*
+ * __task_rq_lock - lock the rq @p resides on.
+ */
+static inline struct rq *__task_rq_lock(struct task_struct *p)
+       __acquires(rq->lock)
+{
+       struct rq *rq;
+
+       lockdep_assert_held(&p->pi_lock);
+
+       for (;;) {
+               rq = task_rq(p);
+               raw_spin_lock(&rq->lock);
+               if (likely(rq == task_rq(p)))
+                       return rq;
+               raw_spin_unlock(&rq->lock);
+       }
+}
+
+/*
+ * task_rq_lock - lock p->pi_lock and lock the rq @p resides on.
+ */
+static struct rq *task_rq_lock(struct task_struct *p, unsigned long *flags)
+       __acquires(p->pi_lock)
+       __acquires(rq->lock)
+{
+       struct rq *rq;
+
+       for (;;) {
+               raw_spin_lock_irqsave(&p->pi_lock, *flags);
+               rq = task_rq(p);
+               raw_spin_lock(&rq->lock);
+               if (likely(rq == task_rq(p)))
+                       return rq;
+               raw_spin_unlock(&rq->lock);
+               raw_spin_unlock_irqrestore(&p->pi_lock, *flags);
+       }
+}
+
+static void __task_rq_unlock(struct rq *rq)
+       __releases(rq->lock)
+{
+       raw_spin_unlock(&rq->lock);
+}
+
+static inline void
+task_rq_unlock(struct rq *rq, struct task_struct *p, unsigned long *flags)
+       __releases(rq->lock)
+       __releases(p->pi_lock)
+{
+       raw_spin_unlock(&rq->lock);
+       raw_spin_unlock_irqrestore(&p->pi_lock, *flags);
+}
+
+/*
+ * this_rq_lock - lock this runqueue and disable interrupts.
+ */
+static struct rq *this_rq_lock(void)
+       __acquires(rq->lock)
+{
+       struct rq *rq;
+
+       local_irq_disable();
+       rq = this_rq();
+       raw_spin_lock(&rq->lock);
+
+       return rq;
+}
+
+#ifdef CONFIG_SCHED_HRTICK
+/*
+ * Use HR-timers to deliver accurate preemption points.
+ *
+ * Its all a bit involved since we cannot program an hrt while holding the
+ * rq->lock. So what we do is store a state in in rq->hrtick_* and ask for a
+ * reschedule event.
+ *
+ * When we get rescheduled we reprogram the hrtick_timer outside of the
+ * rq->lock.
+ */
+
+/*
+ * Use hrtick when:
+ *  - enabled by features
+ *  - hrtimer is actually high res
+ */
+static inline int hrtick_enabled(struct rq *rq)
+{
+       if (!sched_feat(HRTICK))
+               return 0;
+       if (!cpu_active(cpu_of(rq)))
+               return 0;
+       return hrtimer_is_hres_active(&rq->hrtick_timer);
+}
+
+static void hrtick_clear(struct rq *rq)
+{
+       if (hrtimer_active(&rq->hrtick_timer))
+               hrtimer_cancel(&rq->hrtick_timer);
+}
+
+/*
+ * High-resolution timer tick.
+ * Runs from hardirq context with interrupts disabled.
+ */
+static enum hrtimer_restart hrtick(struct hrtimer *timer)
+{
+       struct rq *rq = container_of(timer, struct rq, hrtick_timer);
+
+       WARN_ON_ONCE(cpu_of(rq) != smp_processor_id());
+
+       raw_spin_lock(&rq->lock);
+       update_rq_clock(rq);
+       rq->curr->sched_class->task_tick(rq, rq->curr, 1);
+       raw_spin_unlock(&rq->lock);
+
+       return HRTIMER_NORESTART;
+}
+
+#ifdef CONFIG_SMP
+/*
+ * called from hardirq (IPI) context
+ */
+static void __hrtick_start(void *arg)
+{
+       struct rq *rq = arg;
+
+       raw_spin_lock(&rq->lock);
+       hrtimer_restart(&rq->hrtick_timer);
+       rq->hrtick_csd_pending = 0;
+       raw_spin_unlock(&rq->lock);
+}
+
+/*
+ * Called to set the hrtick timer state.
+ *
+ * called with rq->lock held and irqs disabled
+ */
+static void hrtick_start(struct rq *rq, u64 delay)
+{
+       struct hrtimer *timer = &rq->hrtick_timer;
+       ktime_t time = ktime_add_ns(timer->base->get_time(), delay);
+
+       hrtimer_set_expires(timer, time);
+
+       if (rq == this_rq()) {
+               hrtimer_restart(timer);
+       } else if (!rq->hrtick_csd_pending) {
+               __smp_call_function_single(cpu_of(rq), &rq->hrtick_csd, 0);
+               rq->hrtick_csd_pending = 1;
+       }
+}
+
+static int
+hotplug_hrtick(struct notifier_block *nfb, unsigned long action, void *hcpu)
+{
+       int cpu = (int)(long)hcpu;
+
+       switch (action) {
+       case CPU_UP_CANCELED:
+       case CPU_UP_CANCELED_FROZEN:
+       case CPU_DOWN_PREPARE:
+       case CPU_DOWN_PREPARE_FROZEN:
+       case CPU_DEAD:
+       case CPU_DEAD_FROZEN:
+               hrtick_clear(cpu_rq(cpu));
+               return NOTIFY_OK;
+       }
+
+       return NOTIFY_DONE;
+}
+
+static __init void init_hrtick(void)
+{
+       hotcpu_notifier(hotplug_hrtick, 0);
+}
+#else
+/*
+ * Called to set the hrtick timer state.
+ *
+ * called with rq->lock held and irqs disabled
+ */
+static void hrtick_start(struct rq *rq, u64 delay)
+{
+       __hrtimer_start_range_ns(&rq->hrtick_timer, ns_to_ktime(delay), 0,
+                       HRTIMER_MODE_REL_PINNED, 0);
+}
+
+static inline void init_hrtick(void)
+{
+}
+#endif /* CONFIG_SMP */
+
+static void init_rq_hrtick(struct rq *rq)
+{
+#ifdef CONFIG_SMP
+       rq->hrtick_csd_pending = 0;
+
+       rq->hrtick_csd.flags = 0;
+       rq->hrtick_csd.func = __hrtick_start;
+       rq->hrtick_csd.info = rq;
+#endif
+
+       hrtimer_init(&rq->hrtick_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
+       rq->hrtick_timer.function = hrtick;
+}
+#else  /* CONFIG_SCHED_HRTICK */
+static inline void hrtick_clear(struct rq *rq)
+{
+}
+
+static inline void init_rq_hrtick(struct rq *rq)
+{
+}
+
+static inline void init_hrtick(void)
+{
+}
+#endif /* CONFIG_SCHED_HRTICK */
+
+/*
+ * resched_task - mark a task 'to be rescheduled now'.
+ *
+ * On UP this means the setting of the need_resched flag, on SMP it
+ * might also involve a cross-CPU call to trigger the scheduler on
+ * the target CPU.
+ */
+#ifdef CONFIG_SMP
+
+#ifndef tsk_is_polling
+#define tsk_is_polling(t) test_tsk_thread_flag(t, TIF_POLLING_NRFLAG)
+#endif
+
+static void resched_task(struct task_struct *p)
+{
+       int cpu;
+
+       assert_raw_spin_locked(&task_rq(p)->lock);
+
+       if (test_tsk_need_resched(p))
+               return;
+
+       set_tsk_need_resched(p);
+
+       cpu = task_cpu(p);
+       if (cpu == smp_processor_id())
+               return;
+
+       /* NEED_RESCHED must be visible before we test polling */
+       smp_mb();
+       if (!tsk_is_polling(p))
+               smp_send_reschedule(cpu);
+}
+
+static void resched_cpu(int cpu)
+{
+       struct rq *rq = cpu_rq(cpu);
+       unsigned long flags;
+
+       if (!raw_spin_trylock_irqsave(&rq->lock, flags))
+               return;
+       resched_task(cpu_curr(cpu));
+       raw_spin_unlock_irqrestore(&rq->lock, flags);
+}
+
+#ifdef CONFIG_NO_HZ
+/*
+ * In the semi idle case, use the nearest busy cpu for migrating timers
+ * from an idle cpu.  This is good for power-savings.
+ *
+ * We don't do similar optimization for completely idle system, as
+ * selecting an idle cpu will add more delays to the timers than intended
+ * (as that cpu's timer base may not be uptodate wrt jiffies etc).
+ */
+int get_nohz_timer_target(void)
+{
+       int cpu = smp_processor_id();
+       int i;
+       struct sched_domain *sd;
+
+       rcu_read_lock();
+       for_each_domain(cpu, sd) {
+               for_each_cpu(i, sched_domain_span(sd)) {
+                       if (!idle_cpu(i)) {
+                               cpu = i;
+                               goto unlock;
+                       }
+               }
+       }
+unlock:
+       rcu_read_unlock();
+       return cpu;
+}
+/*
+ * When add_timer_on() enqueues a timer into the timer wheel of an
+ * idle CPU then this timer might expire before the next timer event
+ * which is scheduled to wake up that CPU. In case of a completely
+ * idle system the next event might even be infinite time into the
+ * future. wake_up_idle_cpu() ensures that the CPU is woken up and
+ * leaves the inner idle loop so the newly added timer is taken into
+ * account when the CPU goes back to idle and evaluates the timer
+ * wheel for the next timer event.
+ */
+void wake_up_idle_cpu(int cpu)
+{
+       struct rq *rq = cpu_rq(cpu);
+
+       if (cpu == smp_processor_id())
+               return;
+
+       /*
+        * This is safe, as this function is called with the timer
+        * wheel base lock of (cpu) held. When the CPU is on the way
+        * to idle and has not yet set rq->curr to idle then it will
+        * be serialized on the timer wheel base lock and take the new
+        * timer into account automatically.
+        */
+       if (rq->curr != rq->idle)
+               return;
+
+       /*
+        * We can set TIF_RESCHED on the idle task of the other CPU
+        * lockless. The worst case is that the other CPU runs the
+        * idle task through an additional NOOP schedule()
+        */
+       set_tsk_need_resched(rq->idle);
+
+       /* NEED_RESCHED must be visible before we test polling */
+       smp_mb();
+       if (!tsk_is_polling(rq->idle))
+               smp_send_reschedule(cpu);
+}
+
+#endif /* CONFIG_NO_HZ */
+
+static u64 sched_avg_period(void)
+{
+       return (u64)sysctl_sched_time_avg * NSEC_PER_MSEC / 2;
+}
+
+static void sched_avg_update(struct rq *rq)
+{
+       s64 period = sched_avg_period();
+
+       while ((s64)(rq->clock - rq->age_stamp) > period) {
+               /*
+                * Inline assembly required to prevent the compiler
+                * optimising this loop into a divmod call.
+                * See __iter_div_u64_rem() for another example of this.
+                */
+               asm("" : "+rm" (rq->age_stamp));
+               rq->age_stamp += period;
+               rq->rt_avg /= 2;
+       }
+}
+
+static void sched_rt_avg_update(struct rq *rq, u64 rt_delta)
+{
+       rq->rt_avg += rt_delta;
+       sched_avg_update(rq);
+}
+
+#else /* !CONFIG_SMP */
+static void resched_task(struct task_struct *p)
+{
+       assert_raw_spin_locked(&task_rq(p)->lock);
+       set_tsk_need_resched(p);
+}
+
+static void sched_rt_avg_update(struct rq *rq, u64 rt_delta)
+{
+}
+
+static void sched_avg_update(struct rq *rq)
+{
+}
+#endif /* CONFIG_SMP */
+
+#if BITS_PER_LONG == 32
+# define WMULT_CONST   (~0UL)
+#else
+# define WMULT_CONST   (1UL << 32)
+#endif
+
+#define WMULT_SHIFT    32
+
+/*
+ * Shift right and round:
+ */
+#define SRR(x, y) (((x) + (1UL << ((y) - 1))) >> (y))
+
+/*
+ * delta *= weight / lw
+ */
+static unsigned long
+calc_delta_mine(unsigned long delta_exec, unsigned long weight,
+               struct load_weight *lw)
+{
+       u64 tmp;
+
+       /*
+        * weight can be less than 2^SCHED_LOAD_RESOLUTION for task group sched
+        * entities since MIN_SHARES = 2. Treat weight as 1 if less than
+        * 2^SCHED_LOAD_RESOLUTION.
+        */
+       if (likely(weight > (1UL << SCHED_LOAD_RESOLUTION)))
+               tmp = (u64)delta_exec * scale_load_down(weight);
+       else
+               tmp = (u64)delta_exec;
+
+       if (!lw->inv_weight) {
+               unsigned long w = scale_load_down(lw->weight);
+
+               if (BITS_PER_LONG > 32 && unlikely(w >= WMULT_CONST))
+                       lw->inv_weight = 1;
+               else if (unlikely(!w))
+                       lw->inv_weight = WMULT_CONST;
+               else
+                       lw->inv_weight = WMULT_CONST / w;
+       }
+
+       /*
+        * Check whether we'd overflow the 64-bit multiplication:
+        */
+       if (unlikely(tmp > WMULT_CONST))
+               tmp = SRR(SRR(tmp, WMULT_SHIFT/2) * lw->inv_weight,
+                       WMULT_SHIFT/2);
+       else
+               tmp = SRR(tmp * lw->inv_weight, WMULT_SHIFT);
+
+       return (unsigned long)min(tmp, (u64)(unsigned long)LONG_MAX);
+}
+
+static inline void update_load_add(struct load_weight *lw, unsigned long inc)
+{
+       lw->weight += inc;
+       lw->inv_weight = 0;
+}
+
+static inline void update_load_sub(struct load_weight *lw, unsigned long dec)
+{
+       lw->weight -= dec;
+       lw->inv_weight = 0;
+}
+
+static inline void update_load_set(struct load_weight *lw, unsigned long w)
+{
+       lw->weight = w;
+       lw->inv_weight = 0;
+}
+
+/*
+ * To aid in avoiding the subversion of "niceness" due to uneven distribution
+ * of tasks with abnormal "nice" values across CPUs the contribution that
+ * each task makes to its run queue's load is weighted according to its
+ * scheduling class and "nice" value. For SCHED_NORMAL tasks this is just a
+ * scaled version of the new time slice allocation that they receive on time
+ * slice expiry etc.
+ */
+
+#define WEIGHT_IDLEPRIO                3
+#define WMULT_IDLEPRIO         1431655765
+
+/*
+ * Nice levels are multiplicative, with a gentle 10% change for every
+ * nice level changed. I.e. when a CPU-bound task goes from nice 0 to
+ * nice 1, it will get ~10% less CPU time than another CPU-bound task
+ * that remained on nice 0.
+ *
+ * The "10% effect" is relative and cumulative: from _any_ nice level,
+ * if you go up 1 level, it's -10% CPU usage, if you go down 1 level
+ * it's +10% CPU usage. (to achieve that we use a multiplier of 1.25.
+ * If a task goes up by ~10% and another task goes down by ~10% then
+ * the relative distance between them is ~25%.)
+ */
+static const int prio_to_weight[40] = {
+ /* -20 */     88761,     71755,     56483,     46273,     36291,
+ /* -15 */     29154,     23254,     18705,     14949,     11916,
+ /* -10 */      9548,      7620,      6100,      4904,      3906,
+ /*  -5 */      3121,      2501,      1991,      1586,      1277,
+ /*   0 */      1024,       820,       655,       526,       423,
+ /*   5 */       335,       272,       215,       172,       137,
+ /*  10 */       110,        87,        70,        56,        45,
+ /*  15 */        36,        29,        23,        18,        15,
+};
+
+/*
+ * Inverse (2^32/x) values of the prio_to_weight[] array, precalculated.
+ *
+ * In cases where the weight does not change often, we can use the
+ * precalculated inverse to speed up arithmetics by turning divisions
+ * into multiplications:
+ */
+static const u32 prio_to_wmult[40] = {
+ /* -20 */     48388,     59856,     76040,     92818,    118348,
+ /* -15 */    147320,    184698,    229616,    287308,    360437,
+ /* -10 */    449829,    563644,    704093,    875809,   1099582,
+ /*  -5 */   1376151,   1717300,   2157191,   2708050,   3363326,
+ /*   0 */   4194304,   5237765,   6557202,   8165337,  10153587,
+ /*   5 */  12820798,  15790321,  19976592,  24970740,  31350126,
+ /*  10 */  39045157,  49367440,  61356676,  76695844,  95443717,
+ /*  15 */ 119304647, 148102320, 186737708, 238609294, 286331153,
+};
+
+/* Time spent by the tasks of the cpu accounting group executing in ... */
+enum cpuacct_stat_index {
+       CPUACCT_STAT_USER,      /* ... user mode */
+       CPUACCT_STAT_SYSTEM,    /* ... kernel mode */
+
+       CPUACCT_STAT_NSTATS,
+};
+
+#ifdef CONFIG_CGROUP_CPUACCT
+static void cpuacct_charge(struct task_struct *tsk, u64 cputime);
+static void cpuacct_update_stats(struct task_struct *tsk,
+               enum cpuacct_stat_index idx, cputime_t val);
+#else
+static inline void cpuacct_charge(struct task_struct *tsk, u64 cputime) {}
+static inline void cpuacct_update_stats(struct task_struct *tsk,
+               enum cpuacct_stat_index idx, cputime_t val) {}
+#endif
+
+static inline void inc_cpu_load(struct rq *rq, unsigned long load)
+{
+       update_load_add(&rq->load, load);
+}
+
+static inline void dec_cpu_load(struct rq *rq, unsigned long load)
+{
+       update_load_sub(&rq->load, load);
+}
+
+#if (defined(CONFIG_SMP) && defined(CONFIG_FAIR_GROUP_SCHED)) || defined(CONFIG_RT_GROUP_SCHED)
+typedef int (*tg_visitor)(struct task_group *, void *);
+
+/*
+ * Iterate the full tree, calling @down when first entering a node and @up when
+ * leaving it for the final time.
+ */
+static int walk_tg_tree(tg_visitor down, tg_visitor up, void *data)
+{
+       struct task_group *parent, *child;
+       int ret;
+
+       rcu_read_lock();
+       parent = &root_task_group;
+down:
+       ret = (*down)(parent, data);
+       if (ret)
+               goto out_unlock;
+       list_for_each_entry_rcu(child, &parent->children, siblings) {
+               parent = child;
+               goto down;
+
+up:
+               continue;
+       }
+       ret = (*up)(parent, data);
+       if (ret)
+               goto out_unlock;
+
+       child = parent;
+       parent = parent->parent;
+       if (parent)
+               goto up;
+out_unlock:
+       rcu_read_unlock();
+
+       return ret;
+}
+
+static int tg_nop(struct task_group *tg, void *data)
+{
+       return 0;
+}
+#endif
+
+#ifdef CONFIG_SMP
+/* Used instead of source_load when we know the type == 0 */
+static unsigned long weighted_cpuload(const int cpu)
+{
+       return cpu_rq(cpu)->load.weight;
+}
+
+/*
+ * Return a low guess at the load of a migration-source cpu weighted
+ * according to the scheduling class and "nice" value.
+ *
+ * We want to under-estimate the load of migration sources, to
+ * balance conservatively.
+ */
+static unsigned long source_load(int cpu, int type)
+{
+       struct rq *rq = cpu_rq(cpu);
+       unsigned long total = weighted_cpuload(cpu);
+
+       if (type == 0 || !sched_feat(LB_BIAS))
+               return total;
+
+       return min(rq->cpu_load[type-1], total);
+}
+
+/*
+ * Return a high guess at the load of a migration-target cpu weighted
+ * according to the scheduling class and "nice" value.
+ */
+static unsigned long target_load(int cpu, int type)
+{
+       struct rq *rq = cpu_rq(cpu);
+       unsigned long total = weighted_cpuload(cpu);
+
+       if (type == 0 || !sched_feat(LB_BIAS))
+               return total;
+
+       return max(rq->cpu_load[type-1], total);
+}
+
+static unsigned long power_of(int cpu)
+{
+       return cpu_rq(cpu)->cpu_power;
+}
+
+static int task_hot(struct task_struct *p, u64 now, struct sched_domain *sd);
+
+static unsigned long cpu_avg_load_per_task(int cpu)
+{
+       struct rq *rq = cpu_rq(cpu);
+       unsigned long nr_running = ACCESS_ONCE(rq->nr_running);
+
+       if (nr_running)
+               rq->avg_load_per_task = rq->load.weight / nr_running;
+       else
+               rq->avg_load_per_task = 0;
+
+       return rq->avg_load_per_task;
+}
+
+#ifdef CONFIG_PREEMPT
+
+static void double_rq_lock(struct rq *rq1, struct rq *rq2);
+
+/*
+ * fair double_lock_balance: Safely acquires both rq->locks in a fair
+ * way at the expense of forcing extra atomic operations in all
+ * invocations.  This assures that the double_lock is acquired using the
+ * same underlying policy as the spinlock_t on this architecture, which
+ * reduces latency compared to the unfair variant below.  However, it
+ * also adds more overhead and therefore may reduce throughput.
+ */
+static inline int _double_lock_balance(struct rq *this_rq, struct rq *busiest)
+       __releases(this_rq->lock)
+       __acquires(busiest->lock)
+       __acquires(this_rq->lock)
+{
+       raw_spin_unlock(&this_rq->lock);
+       double_rq_lock(this_rq, busiest);
+
+       return 1;
+}
+
+#else
+/*
+ * Unfair double_lock_balance: Optimizes throughput at the expense of
+ * latency by eliminating extra atomic operations when the locks are
+ * already in proper order on entry.  This favors lower cpu-ids and will
+ * grant the double lock to lower cpus over higher ids under contention,
+ * regardless of entry order into the function.
+ */
+static int _double_lock_balance(struct rq *this_rq, struct rq *busiest)
+       __releases(this_rq->lock)
+       __acquires(busiest->lock)
+       __acquires(this_rq->lock)
+{
+       int ret = 0;
+
+       if (unlikely(!raw_spin_trylock(&busiest->lock))) {
+               if (busiest < this_rq) {
+                       raw_spin_unlock(&this_rq->lock);
+                       raw_spin_lock(&busiest->lock);
+                       raw_spin_lock_nested(&this_rq->lock,
+                                             SINGLE_DEPTH_NESTING);
+                       ret = 1;
+               } else
+                       raw_spin_lock_nested(&busiest->lock,
+                                             SINGLE_DEPTH_NESTING);
+       }
+       return ret;
+}
+
+#endif /* CONFIG_PREEMPT */
+
+/*
+ * double_lock_balance - lock the busiest runqueue, this_rq is locked already.
+ */
+static int double_lock_balance(struct rq *this_rq, struct rq *busiest)
+{
+       if (unlikely(!irqs_disabled())) {
+               /* printk() doesn't work good under rq->lock */
+               raw_spin_unlock(&this_rq->lock);
+               BUG_ON(1);
+       }
+
+       return _double_lock_balance(this_rq, busiest);
+}
+
+static inline void double_unlock_balance(struct rq *this_rq, struct rq *busiest)
+       __releases(busiest->lock)
+{
+       raw_spin_unlock(&busiest->lock);
+       lock_set_subclass(&this_rq->lock.dep_map, 0, _RET_IP_);
+}
+
+/*
+ * double_rq_lock - safely lock two runqueues
+ *
+ * Note this does not disable interrupts like task_rq_lock,
+ * you need to do so manually before calling.
+ */
+static void double_rq_lock(struct rq *rq1, struct rq *rq2)
+       __acquires(rq1->lock)
+       __acquires(rq2->lock)
+{
+       BUG_ON(!irqs_disabled());
+       if (rq1 == rq2) {
+               raw_spin_lock(&rq1->lock);
+               __acquire(rq2->lock);   /* Fake it out ;) */
+       } else {
+               if (rq1 < rq2) {
+                       raw_spin_lock(&rq1->lock);
+                       raw_spin_lock_nested(&rq2->lock, SINGLE_DEPTH_NESTING);
+               } else {
+                       raw_spin_lock(&rq2->lock);
+                       raw_spin_lock_nested(&rq1->lock, SINGLE_DEPTH_NESTING);
+               }
+       }
+}
+
+/*
+ * double_rq_unlock - safely unlock two runqueues
+ *
+ * Note this does not restore interrupts like task_rq_unlock,
+ * you need to do so manually after calling.
+ */
+static void double_rq_unlock(struct rq *rq1, struct rq *rq2)
+       __releases(rq1->lock)
+       __releases(rq2->lock)
+{
+       raw_spin_unlock(&rq1->lock);
+       if (rq1 != rq2)
+               raw_spin_unlock(&rq2->lock);
+       else
+               __release(rq2->lock);
+}
+
+#else /* CONFIG_SMP */
+
+/*
+ * double_rq_lock - safely lock two runqueues
+ *
+ * Note this does not disable interrupts like task_rq_lock,
+ * you need to do so manually before calling.
+ */
+static void double_rq_lock(struct rq *rq1, struct rq *rq2)
+       __acquires(rq1->lock)
+       __acquires(rq2->lock)
+{
+       BUG_ON(!irqs_disabled());
+       BUG_ON(rq1 != rq2);
+       raw_spin_lock(&rq1->lock);
+       __acquire(rq2->lock);   /* Fake it out ;) */
+}
+
+/*
+ * double_rq_unlock - safely unlock two runqueues
+ *
+ * Note this does not restore interrupts like task_rq_unlock,
+ * you need to do so manually after calling.
+ */
+static void double_rq_unlock(struct rq *rq1, struct rq *rq2)
+       __releases(rq1->lock)
+       __releases(rq2->lock)
+{
+       BUG_ON(rq1 != rq2);
+       raw_spin_unlock(&rq1->lock);
+       __release(rq2->lock);
+}
+
+#endif
+
+static void calc_load_account_idle(struct rq *this_rq);
+static void update_sysctl(void);
+static int get_update_sysctl_factor(void);
+static void update_cpu_load(struct rq *this_rq);
+
+static inline void __set_task_cpu(struct task_struct *p, unsigned int cpu)
+{
+       set_task_rq(p, cpu);
+#ifdef CONFIG_SMP
+       /*
+        * After ->cpu is set up to a new value, task_rq_lock(p, ...) can be
+        * successfuly executed on another CPU. We must ensure that updates of
+        * per-task data have been completed by this moment.
+        */
+       smp_wmb();
+       task_thread_info(p)->cpu = cpu;
+#endif
+}
+
+static const struct sched_class rt_sched_class;
+
+#define sched_class_highest (&stop_sched_class)
+#define for_each_class(class) \
+   for (class = sched_class_highest; class; class = class->next)
+
+#include "sched_stats.h"
+
+static void inc_nr_running(struct rq *rq)
+{
+       rq->nr_running++;
+}
+
+static void dec_nr_running(struct rq *rq)
+{
+       rq->nr_running--;
+}
+
+static void set_load_weight(struct task_struct *p)
+{
+       int prio = p->static_prio - MAX_RT_PRIO;
+       struct load_weight *load = &p->se.load;
+
+       /*
+        * SCHED_IDLE tasks get minimal weight:
+        */
+       if (p->policy == SCHED_IDLE) {
+               load->weight = scale_load(WEIGHT_IDLEPRIO);
+               load->inv_weight = WMULT_IDLEPRIO;
+               return;
+       }
+
+       load->weight = scale_load(prio_to_weight[prio]);
+       load->inv_weight = prio_to_wmult[prio];
+}
+
+static void enqueue_task(struct rq *rq, struct task_struct *p, int flags)
+{
+       update_rq_clock(rq);
+       sched_info_queued(p);
+       p->sched_class->enqueue_task(rq, p, flags);
+}
+
+static void dequeue_task(struct rq *rq, struct task_struct *p, int flags)
+{
+       update_rq_clock(rq);
+       sched_info_dequeued(p);
+       p->sched_class->dequeue_task(rq, p, flags);
+}
+
+/*
+ * activate_task - move a task to the runqueue.
+ */
+static void activate_task(struct rq *rq, struct task_struct *p, int flags)
+{
+       if (task_contributes_to_load(p))
+               rq->nr_uninterruptible--;
+
+       enqueue_task(rq, p, flags);
+       inc_nr_running(rq);
+}
+
+/*
+ * deactivate_task - remove a task from the runqueue.
+ */
+static void deactivate_task(struct rq *rq, struct task_struct *p, int flags)
+{
+       if (task_contributes_to_load(p))
+               rq->nr_uninterruptible++;
+
+       dequeue_task(rq, p, flags);
+       dec_nr_running(rq);
+}
+
+#ifdef CONFIG_IRQ_TIME_ACCOUNTING
+
+/*
+ * There are no locks covering percpu hardirq/softirq time.
+ * They are only modified in account_system_vtime, on corresponding CPU
+ * with interrupts disabled. So, writes are safe.
+ * They are read and saved off onto struct rq in update_rq_clock().
+ * This may result in other CPU reading this CPU's irq time and can
+ * race with irq/account_system_vtime on this CPU. We would either get old
+ * or new value with a side effect of accounting a slice of irq time to wrong
+ * task when irq is in progress while we read rq->clock. That is a worthy
+ * compromise in place of having locks on each irq in account_system_time.
+ */
+static DEFINE_PER_CPU(u64, cpu_hardirq_time);
+static DEFINE_PER_CPU(u64, cpu_softirq_time);
+
+static DEFINE_PER_CPU(u64, irq_start_time);
+static int sched_clock_irqtime;
+
+void enable_sched_clock_irqtime(void)
+{
+       sched_clock_irqtime = 1;
+}
+
+void disable_sched_clock_irqtime(void)
+{
+       sched_clock_irqtime = 0;
+}
+
+#ifndef CONFIG_64BIT
+static DEFINE_PER_CPU(seqcount_t, irq_time_seq);
+
+static inline void irq_time_write_begin(void)
+{
+       __this_cpu_inc(irq_time_seq.sequence);
+       smp_wmb();
+}
+
+static inline void irq_time_write_end(void)
+{
+       smp_wmb();
+       __this_cpu_inc(irq_time_seq.sequence);
+}
+
+static inline u64 irq_time_read(int cpu)
+{
+       u64 irq_time;
+       unsigned seq;
+
+       do {
+               seq = read_seqcount_begin(&per_cpu(irq_time_seq, cpu));
+               irq_time = per_cpu(cpu_softirq_time, cpu) +
+                          per_cpu(cpu_hardirq_time, cpu);
+       } while (read_seqcount_retry(&per_cpu(irq_time_seq, cpu), seq));
+
+       return irq_time;
+}
+#else /* CONFIG_64BIT */
+static inline void irq_time_write_begin(void)
+{
+}
+
+static inline void irq_time_write_end(void)
+{
+}
+
+static inline u64 irq_time_read(int cpu)
+{
+       return per_cpu(cpu_softirq_time, cpu) + per_cpu(cpu_hardirq_time, cpu);
+}
+#endif /* CONFIG_64BIT */
+
+/*
+ * Called before incrementing preempt_count on {soft,}irq_enter
+ * and before decrementing preempt_count on {soft,}irq_exit.
+ */
+void account_system_vtime(struct task_struct *curr)
+{
+       unsigned long flags;
+       s64 delta;
+       int cpu;
+
+       if (!sched_clock_irqtime)
+               return;
+
+       local_irq_save(flags);
+
+       cpu = smp_processor_id();
+       delta = sched_clock_cpu(cpu) - __this_cpu_read(irq_start_time);
+       __this_cpu_add(irq_start_time, delta);
+
+       irq_time_write_begin();
+       /*
+        * We do not account for softirq time from ksoftirqd here.
+        * We want to continue accounting softirq time to ksoftirqd thread
+        * in that case, so as not to confuse scheduler with a special task
+        * that do not consume any time, but still wants to run.
+        */
+       if (hardirq_count())
+               __this_cpu_add(cpu_hardirq_time, delta);
+       else if (in_serving_softirq() && curr != this_cpu_ksoftirqd())
+               __this_cpu_add(cpu_softirq_time, delta);
+
+       irq_time_write_end();
+       local_irq_restore(flags);
+}
+EXPORT_SYMBOL_GPL(account_system_vtime);
+
+#endif /* CONFIG_IRQ_TIME_ACCOUNTING */
+
+#ifdef CONFIG_PARAVIRT
+static inline u64 steal_ticks(u64 steal)
+{
+       if (unlikely(steal > NSEC_PER_SEC))
+               return div_u64(steal, TICK_NSEC);
+
+       return __iter_div_u64_rem(steal, TICK_NSEC, &steal);
+}
+#endif
+
+static void update_rq_clock_task(struct rq *rq, s64 delta)
+{
+/*
+ * In theory, the compile should just see 0 here, and optimize out the call
+ * to sched_rt_avg_update. But I don't trust it...
+ */
+#if defined(CONFIG_IRQ_TIME_ACCOUNTING) || defined(CONFIG_PARAVIRT_TIME_ACCOUNTING)
+       s64 steal = 0, irq_delta = 0;
+#endif
+#ifdef CONFIG_IRQ_TIME_ACCOUNTING
+       irq_delta = irq_time_read(cpu_of(rq)) - rq->prev_irq_time;
+
+       /*
+        * Since irq_time is only updated on {soft,}irq_exit, we might run into
+        * this case when a previous update_rq_clock() happened inside a
+        * {soft,}irq region.
+        *
+        * When this happens, we stop ->clock_task and only update the
+        * prev_irq_time stamp to account for the part that fit, so that a next
+        * update will consume the rest. This ensures ->clock_task is
+        * monotonic.
+        *
+        * It does however cause some slight miss-attribution of {soft,}irq
+        * time, a more accurate solution would be to update the irq_time using
+        * the current rq->clock timestamp, except that would require using
+        * atomic ops.
+        */
+       if (irq_delta > delta)
+               irq_delta = delta;
+
+       rq->prev_irq_time += irq_delta;
+       delta -= irq_delta;
+#endif
+#ifdef CONFIG_PARAVIRT_TIME_ACCOUNTING
+       if (static_branch((&paravirt_steal_rq_enabled))) {
+               u64 st;
+
+               steal = paravirt_steal_clock(cpu_of(rq));
+               steal -= rq->prev_steal_time_rq;
+
+               if (unlikely(steal > delta))
+                       steal = delta;
+
+               st = steal_ticks(steal);
+               steal = st * TICK_NSEC;
+
+               rq->prev_steal_time_rq += steal;
+
+               delta -= steal;
+       }
+#endif
+
+       rq->clock_task += delta;
+
+#if defined(CONFIG_IRQ_TIME_ACCOUNTING) || defined(CONFIG_PARAVIRT_TIME_ACCOUNTING)
+       if ((irq_delta + steal) && sched_feat(NONTASK_POWER))
+               sched_rt_avg_update(rq, irq_delta + steal);
+#endif
+}
+
+#ifdef CONFIG_IRQ_TIME_ACCOUNTING
+static int irqtime_account_hi_update(void)
+{
+       struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;
+       unsigned long flags;
+       u64 latest_ns;
+       int ret = 0;
+
+       local_irq_save(flags);
+       latest_ns = this_cpu_read(cpu_hardirq_time);
+       if (cputime64_gt(nsecs_to_cputime64(latest_ns), cpustat->irq))
+               ret = 1;
+       local_irq_restore(flags);
+       return ret;
+}
+
+static int irqtime_account_si_update(void)
+{
+       struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;
+       unsigned long flags;
+       u64 latest_ns;
+       int ret = 0;
+
+       local_irq_save(flags);
+       latest_ns = this_cpu_read(cpu_softirq_time);
+       if (cputime64_gt(nsecs_to_cputime64(latest_ns), cpustat->softirq))
+               ret = 1;
+       local_irq_restore(flags);
+       return ret;
+}
+
+#else /* CONFIG_IRQ_TIME_ACCOUNTING */
+
+#define sched_clock_irqtime    (0)
+
+#endif
+
+#include "sched_idletask.c"
+#include "sched_fair.c"
+#include "sched_rt.c"
+#include "sched_autogroup.c"
+#include "sched_stoptask.c"
+#ifdef CONFIG_SCHED_DEBUG
+# include "sched_debug.c"
+#endif
+
+void sched_set_stop_task(int cpu, struct task_struct *stop)
+{
+       struct sched_param param = { .sched_priority = MAX_RT_PRIO - 1 };
+       struct task_struct *old_stop = cpu_rq(cpu)->stop;
+
+       if (stop) {
+               /*
+                * Make it appear like a SCHED_FIFO task, its something
+                * userspace knows about and won't get confused about.
+                *
+                * Also, it will make PI more or less work without too
+                * much confusion -- but then, stop work should not
+                * rely on PI working anyway.
+                */
+               sched_setscheduler_nocheck(stop, SCHED_FIFO, &param);
+
+               stop->sched_class = &stop_sched_class;
+       }
+
+       cpu_rq(cpu)->stop = stop;
+
+       if (old_stop) {
+               /*
+                * Reset it back to a normal scheduling class so that
+                * it can die in pieces.
+                */
+               old_stop->sched_class = &rt_sched_class;
+       }
+}
+
+/*
+ * __normal_prio - return the priority that is based on the static prio
+ */
+static inline int __normal_prio(struct task_struct *p)
+{
+       return p->static_prio;
+}
+
+/*
+ * Calculate the expected normal priority: i.e. priority
+ * without taking RT-inheritance into account. Might be
+ * boosted by interactivity modifiers. Changes upon fork,
+ * setprio syscalls, and whenever the interactivity
+ * estimator recalculates.
+ */
+static inline int normal_prio(struct task_struct *p)
+{
+       int prio;
+
+       if (task_has_rt_policy(p))
+               prio = MAX_RT_PRIO-1 - p->rt_priority;
+       else
+               prio = __normal_prio(p);
+       return prio;
+}
+
+/*
+ * Calculate the current priority, i.e. the priority
+ * taken into account by the scheduler. This value might
+ * be boosted by RT tasks, or might be boosted by
+ * interactivity modifiers. Will be RT if the task got
+ * RT-boosted. If not then it returns p->normal_prio.
+ */
+static int effective_prio(struct task_struct *p)
+{
+       p->normal_prio = normal_prio(p);
+       /*
+        * If we are RT tasks or we were boosted to RT priority,
+        * keep the priority unchanged. Otherwise, update priority
+        * to the normal priority:
+        */
+       if (!rt_prio(p->prio))
+               return p->normal_prio;
+       return p->prio;
+}
+
+/**
+ * task_curr - is this task currently executing on a CPU?
+ * @p: the task in question.
+ */
+inline int task_curr(const struct task_struct *p)
+{
+       return cpu_curr(task_cpu(p)) == p;
+}
+
+static inline void check_class_changed(struct rq *rq, struct task_struct *p,
+                                      const struct sched_class *prev_class,
+                                      int oldprio)
+{
+       if (prev_class != p->sched_class) {
+               if (prev_class->switched_from)
+                       prev_class->switched_from(rq, p);
+               p->sched_class->switched_to(rq, p);
+       } else if (oldprio != p->prio)
+               p->sched_class->prio_changed(rq, p, oldprio);
+}
+
+static void check_preempt_curr(struct rq *rq, struct task_struct *p, int flags)
+{
+       const struct sched_class *class;
+
+       if (p->sched_class == rq->curr->sched_class) {
+               rq->curr->sched_class->check_preempt_curr(rq, p, flags);
+       } else {
+               for_each_class(class) {
+                       if (class == rq->curr->sched_class)
+                               break;
+                       if (class == p->sched_class) {
+                               resched_task(rq->curr);
+                               break;
+                       }
+               }
+       }
+
+       /*
+        * A queue event has occurred, and we're going to schedule.  In
+        * this case, we can save a useless back to back clock update.
+        */
+       if (rq->curr->on_rq && test_tsk_need_resched(rq->curr))
+               rq->skip_clock_update = 1;
+}
+
+#ifdef CONFIG_SMP
+/*
+ * Is this task likely cache-hot:
+ */
+static int
+task_hot(struct task_struct *p, u64 now, struct sched_domain *sd)
+{
+       s64 delta;
+
+       if (p->sched_class != &fair_sched_class)
+               return 0;
+
+       if (unlikely(p->policy == SCHED_IDLE))
+               return 0;
+
+       /*
+        * Buddy candidates are cache hot:
+        */
+       if (sched_feat(CACHE_HOT_BUDDY) && this_rq()->nr_running &&
+                       (&p->se == cfs_rq_of(&p->se)->next ||
+                        &p->se == cfs_rq_of(&p->se)->last))
+               return 1;
+
+       if (sysctl_sched_migration_cost == -1)
+               return 1;
+       if (sysctl_sched_migration_cost == 0)
+               return 0;
+
+       delta = now - p->se.exec_start;
+
+       return delta < (s64)sysctl_sched_migration_cost;
+}
+
+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_ONCE(p->state != TASK_RUNNING && p->state != TASK_WAKING &&
+                       !(task_thread_info(p)->preempt_count & PREEMPT_ACTIVE));
+
+#ifdef CONFIG_LOCKDEP
+       /*
+        * The caller should hold either p->pi_lock or rq->lock, when changing
+        * a task's CPU. ->pi_lock for waking tasks, rq->lock for runnable tasks.
+        *
+        * sched_move_task() holds both and thus holding either pins the cgroup,
+        * see set_task_rq().
+        *
+        * Furthermore, all task_rq users should acquire both locks, see
+        * task_rq_lock().
+        */
+       WARN_ON_ONCE(debug_locks && !(lockdep_is_held(&p->pi_lock) ||
+                                     lockdep_is_held(&task_rq(p)->lock)));
+#endif
+#endif
+
+       trace_sched_migrate_task(p, new_cpu);
+
+       if (task_cpu(p) != new_cpu) {
+               p->se.nr_migrations++;
+               perf_sw_event(PERF_COUNT_SW_CPU_MIGRATIONS, 1, NULL, 0);
+       }
+
+       __set_task_cpu(p, new_cpu);
+}
+
+struct migration_arg {
+       struct task_struct *task;
+       int dest_cpu;
+};
+
+static int migration_cpu_stop(void *data);
+
+/*
+ * wait_task_inactive - wait for a thread to unschedule.
+ *
+ * If @match_state is nonzero, it's the @p->state value just checked and
+ * not expected to change.  If it changes, i.e. @p might have woken up,
+ * then return zero.  When we succeed in waiting for @p to be off its CPU,
+ * we return a positive number (its total switch count).  If a second call
+ * a short while later returns the same number, the caller can be sure that
+ * @p has remained unscheduled the whole time.
+ *
+ * The caller must ensure that the task *will* unschedule sometime soon,
+ * else this function might spin for a *long* time. This function can't
+ * be called with interrupts off, or it may introduce deadlock with
+ * smp_call_function() if an IPI is sent by the same process we are
+ * waiting to become inactive.
+ */
+unsigned long wait_task_inactive(struct task_struct *p, long match_state)
+{
+       unsigned long flags;
+       int running, on_rq;
+       unsigned long ncsw;
+       struct rq *rq;
+
+       for (;;) {
+               /*
+                * We do the initial early heuristics without holding
+                * any task-queue locks at all. We'll only try to get
+                * the runqueue lock when things look like they will
+                * work out!
+                */
+               rq = task_rq(p);
+
+               /*
+                * If the task is actively running on another CPU
+                * still, just relax and busy-wait without holding
+                * any locks.
+                *
+                * NOTE! Since we don't hold any locks, it's not
+                * even sure that "rq" stays as the right runqueue!
+                * But we don't care, since "task_running()" will
+                * return false if the runqueue has changed and p
+                * is actually now running somewhere else!
+                */
+               while (task_running(rq, p)) {
+                       if (match_state && unlikely(p->state != match_state))
+                               return 0;
+                       cpu_relax();
+               }
+
+               /*
+                * Ok, time to look more closely! We need the rq
+                * lock now, to be *sure*. If we're wrong, we'll
+                * just go back and repeat.
+                */
+               rq = task_rq_lock(p, &flags);
+               trace_sched_wait_task(p);
+               running = task_running(rq, p);
+               on_rq = p->on_rq;
+               ncsw = 0;
+               if (!match_state || p->state == match_state)
+                       ncsw = p->nvcsw | LONG_MIN; /* sets MSB */
+               task_rq_unlock(rq, p, &flags);
+
+               /*
+                * If it changed from the expected state, bail out now.
+                */
+               if (unlikely(!ncsw))
+                       break;
+
+               /*
+                * Was it really running after all now that we
+                * checked with the proper locks actually held?
+                *
+                * Oops. Go back and try again..
+                */
+               if (unlikely(running)) {
+                       cpu_relax();
+                       continue;
+               }
+
+               /*
+                * It's not enough that it's not actively running,
+                * it must be off the runqueue _entirely_, and not
+                * preempted!
+                *
+                * So if it was still runnable (but just not actively
+                * running right now), it's preempted, and we should
+                * yield - it could be a while.
+                */
+               if (unlikely(on_rq)) {
+                       ktime_t to = ktime_set(0, NSEC_PER_SEC/HZ);
+
+                       set_current_state(TASK_UNINTERRUPTIBLE);
+                       schedule_hrtimeout(&to, HRTIMER_MODE_REL);
+                       continue;
+               }
+
+               /*
+                * Ahh, all good. It wasn't running, and it wasn't
+                * runnable, which means that it will never become
+                * running in the future either. We're all done!
+                */
+               break;
+       }
+
+       return ncsw;
+}
+
+/***
+ * kick_process - kick a running thread to enter/exit the kernel
+ * @p: the to-be-kicked thread
+ *
+ * Cause a process which is running on another CPU to enter
+ * kernel-mode, without any delay. (to get signals handled.)
+ *
+ * NOTE: this function doesn't have to take the runqueue lock,
+ * because all it wants to ensure is that the remote task enters
+ * the kernel. If the IPI races and the task has been migrated
+ * to another CPU then no harm is done and the purpose has been
+ * achieved as well.
+ */
+void kick_process(struct task_struct *p)
+{
+       int cpu;
+
+       preempt_disable();
+       cpu = task_cpu(p);
+       if ((cpu != smp_processor_id()) && task_curr(p))
+               smp_send_reschedule(cpu);
+       preempt_enable();
+}
+EXPORT_SYMBOL_GPL(kick_process);
+#endif /* CONFIG_SMP */
+
+#ifdef CONFIG_SMP
+/*
+ * ->cpus_allowed is protected by both rq->lock and p->pi_lock
+ */
+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. */
+       dest_cpu = cpuset_cpus_allowed_fallback(p);
+       /*
+        * 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;
+}
+
+/*
+ * The caller (fork, wakeup) owns p->pi_lock, ->cpus_allowed is stable.
+ */
+static inline
+int select_task_rq(struct task_struct *p, int sd_flags, int 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_online(cpu)))
+               cpu = select_fallback_rq(task_cpu(p), p);
+
+       return cpu;
+}
+
+static void update_avg(u64 *avg, u64 sample)
+{
+       s64 diff = sample - *avg;
+       *avg += diff >> 3;
+}
+#endif
+
+static void
+ttwu_stat(struct task_struct *p, int cpu, int wake_flags)
+{
+#ifdef CONFIG_SCHEDSTATS
+       struct rq *rq = this_rq();
+
+#ifdef CONFIG_SMP
+       int this_cpu = smp_processor_id();
+
+       if (cpu == this_cpu) {
+               schedstat_inc(rq, ttwu_local);
+               schedstat_inc(p, se.statistics.nr_wakeups_local);
+       } else {
+               struct sched_domain *sd;
+
+               schedstat_inc(p, se.statistics.nr_wakeups_remote);
+               rcu_read_lock();
+               for_each_domain(this_cpu, sd) {
+                       if (cpumask_test_cpu(cpu, sched_domain_span(sd))) {
+                               schedstat_inc(sd, ttwu_wake_remote);
+                               break;
+                       }
+               }
+               rcu_read_unlock();
+       }
+
+       if (wake_flags & WF_MIGRATED)
+               schedstat_inc(p, se.statistics.nr_wakeups_migrate);
+
+#endif /* CONFIG_SMP */
+
+       schedstat_inc(rq, ttwu_count);
+       schedstat_inc(p, se.statistics.nr_wakeups);
+
+       if (wake_flags & WF_SYNC)
+               schedstat_inc(p, se.statistics.nr_wakeups_sync);
+
+#endif /* CONFIG_SCHEDSTATS */
+}
+
+static void ttwu_activate(struct rq *rq, struct task_struct *p, int en_flags)
+{
+       activate_task(rq, p, en_flags);
+       p->on_rq = 1;
+
+       /* if a worker is waking up, notify workqueue */
+       if (p->flags & PF_WQ_WORKER)
+               wq_worker_waking_up(p, cpu_of(rq));
+}
+
+/*
+ * Mark the task runnable and perform wakeup-preemption.
+ */
+static void
+ttwu_do_wakeup(struct rq *rq, struct task_struct *p, int wake_flags)
+{
+       trace_sched_wakeup(p, true);
+       check_preempt_curr(rq, p, wake_flags);
+
+       p->state = TASK_RUNNING;
+#ifdef CONFIG_SMP
+       if (p->sched_class->task_woken)
+               p->sched_class->task_woken(rq, p);
+
+       if (rq->idle_stamp) {
+               u64 delta = rq->clock - rq->idle_stamp;
+               u64 max = 2*sysctl_sched_migration_cost;
+
+               if (delta > max)
+                       rq->avg_idle = max;
+               else
+                       update_avg(&rq->avg_idle, delta);
+               rq->idle_stamp = 0;
+       }
+#endif
+}
+
+static void
+ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags)
+{
+#ifdef CONFIG_SMP
+       if (p->sched_contributes_to_load)
+               rq->nr_uninterruptible--;
+#endif
+
+       ttwu_activate(rq, p, ENQUEUE_WAKEUP | ENQUEUE_WAKING);
+       ttwu_do_wakeup(rq, p, wake_flags);
+}
+
+/*
+ * Called in case the task @p isn't fully descheduled from its runqueue,
+ * in this case we must do a remote wakeup. Its a 'light' wakeup though,
+ * since all we need to do is flip p->state to TASK_RUNNING, since
+ * the task is still ->on_rq.
+ */
+static int ttwu_remote(struct task_struct *p, int wake_flags)
+{
+       struct rq *rq;
+       int ret = 0;
+
+       rq = __task_rq_lock(p);
+       if (p->on_rq) {
+               ttwu_do_wakeup(rq, p, wake_flags);
+               ret = 1;
+       }
+       __task_rq_unlock(rq);
+
+       return ret;
+}
+
+#ifdef CONFIG_SMP
+static void sched_ttwu_do_pending(struct task_struct *list)
+{
+       struct rq *rq = this_rq();
+
+       raw_spin_lock(&rq->lock);
+
+       while (list) {
+               struct task_struct *p = list;
+               list = list->wake_entry;
+               ttwu_do_activate(rq, p, 0);
+       }
+
+       raw_spin_unlock(&rq->lock);
+}
+
+#ifdef CONFIG_HOTPLUG_CPU
+
+static void sched_ttwu_pending(void)
+{
+       struct rq *rq = this_rq();
+       struct task_struct *list = xchg(&rq->wake_list, NULL);
+
+       if (!list)
+               return;
+
+       sched_ttwu_do_pending(list);
+}
+
+#endif /* CONFIG_HOTPLUG_CPU */
+
+void scheduler_ipi(void)
+{
+       struct rq *rq = this_rq();
+       struct task_struct *list = xchg(&rq->wake_list, NULL);
+
+       if (!list)
+               return;
+
+       /*
+        * Not all reschedule IPI handlers call irq_enter/irq_exit, since
+        * traditionally all their work was done from the interrupt return
+        * path. Now that we actually do some work, we need to make sure
+        * we do call them.
+        *
+        * Some archs already do call them, luckily irq_enter/exit nest
+        * properly.
+        *
+        * Arguably we should visit all archs and update all handlers,
+        * however a fair share of IPIs are still resched only so this would
+        * somewhat pessimize the simple resched case.
+        */
+       irq_enter();
+       sched_ttwu_do_pending(list);
+       irq_exit();
+}
+
+static void ttwu_queue_remote(struct task_struct *p, int cpu)
+{
+       struct rq *rq = cpu_rq(cpu);
+       struct task_struct *next = rq->wake_list;
+
+       for (;;) {
+               struct task_struct *old = next;
+
+               p->wake_entry = next;
+               next = cmpxchg(&rq->wake_list, old, p);
+               if (next == old)
+                       break;
+       }
+
+       if (!next)
+               smp_send_reschedule(cpu);
+}
+
+#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
+static int ttwu_activate_remote(struct task_struct *p, int wake_flags)
+{
+       struct rq *rq;
+       int ret = 0;
+
+       rq = __task_rq_lock(p);
+       if (p->on_cpu) {
+               ttwu_activate(rq, p, ENQUEUE_WAKEUP);
+               ttwu_do_wakeup(rq, p, wake_flags);
+               ret = 1;
+       }
+       __task_rq_unlock(rq);
+
+       return ret;
+
+}
+#endif /* __ARCH_WANT_INTERRUPTS_ON_CTXSW */
+#endif /* CONFIG_SMP */
+
+static void ttwu_queue(struct task_struct *p, int cpu)
+{
+       struct rq *rq = cpu_rq(cpu);
+
+#if defined(CONFIG_SMP)
+       if (sched_feat(TTWU_QUEUE) && cpu != smp_processor_id()) {
+               sched_clock_cpu(cpu); /* sync clocks x-cpu */
+               ttwu_queue_remote(p, cpu);
+               return;
+       }
+#endif
+
+       raw_spin_lock(&rq->lock);
+       ttwu_do_activate(rq, p, 0);
+       raw_spin_unlock(&rq->lock);
+}
+
+/**
+ * try_to_wake_up - wake up a thread
+ * @p: the thread to be awakened
+ * @state: the mask of task states that can be woken
+ * @wake_flags: wake modifier flags (WF_*)
+ *
+ * Put it on the run-queue if it's not already there. The "current"
+ * thread is always on the run-queue (except when the actual
+ * re-schedule is in progress), and as such you're allowed to do
+ * the simpler "current->state = TASK_RUNNING" to mark yourself
+ * runnable without the overhead of this.
+ *
+ * Returns %true if @p was woken up, %false if it was already running
+ * or @state didn't match @p's state.
+ */
+static int
+try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
+{
+       unsigned long flags;
+       int cpu, success = 0;
+
+       smp_wmb();
+       raw_spin_lock_irqsave(&p->pi_lock, flags);
+       if (!(p->state & state))
+               goto out;
+
+       success = 1; /* we're going to change ->state */
+       cpu = task_cpu(p);
+
+       if (p->on_rq && ttwu_remote(p, wake_flags))
+               goto stat;
+
+#ifdef CONFIG_SMP
+       /*
+        * If the owning (remote) cpu is still in the middle of schedule() with
+        * this task as prev, wait until its done referencing the task.
+        */
+       while (p->on_cpu) {
+#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
+               /*
+                * In case the architecture enables interrupts in
+                * context_switch(), we cannot busy wait, since that
+                * would lead to deadlocks when an interrupt hits and
+                * tries to wake up @prev. So bail and do a complete
+                * remote wakeup.
+                */
+               if (ttwu_activate_remote(p, wake_flags))
+                       goto stat;
+#else
+               cpu_relax();
+#endif
+       }
+       /*
+        * Pairs with the smp_wmb() in finish_lock_switch().
+        */
+       smp_rmb();
+
+       p->sched_contributes_to_load = !!task_contributes_to_load(p);
+       p->state = TASK_WAKING;
+
+       if (p->sched_class->task_waking)
+               p->sched_class->task_waking(p);
+
+       cpu = select_task_rq(p, SD_BALANCE_WAKE, wake_flags);
+       if (task_cpu(p) != cpu) {
+               wake_flags |= WF_MIGRATED;
+               set_task_cpu(p, cpu);
+       }
+#endif /* CONFIG_SMP */
+
+       ttwu_queue(p, cpu);
+stat:
+       ttwu_stat(p, cpu, wake_flags);
+out:
+       raw_spin_unlock_irqrestore(&p->pi_lock, flags);
+
+       return success;
+}
+
+/**
+ * try_to_wake_up_local - try to wake up a local task with rq lock held
+ * @p: the thread to be awakened
+ *
+ * Put @p on the run-queue if it's not already there. The caller must
+ * ensure that this_rq() is locked, @p is bound to this_rq() and not
+ * the current task.
+ */
+static void try_to_wake_up_local(struct task_struct *p)
+{
+       struct rq *rq = task_rq(p);
+
+       BUG_ON(rq != this_rq());
+       BUG_ON(p == current);
+       lockdep_assert_held(&rq->lock);
+
+       if (!raw_spin_trylock(&p->pi_lock)) {
+               raw_spin_unlock(&rq->lock);
+               raw_spin_lock(&p->pi_lock);
+               raw_spin_lock(&rq->lock);
+       }
+
+       if (!(p->state & TASK_NORMAL))
+               goto out;
+
+       if (!p->on_rq)
+               ttwu_activate(rq, p, ENQUEUE_WAKEUP);
+
+       ttwu_do_wakeup(rq, p, 0);
+       ttwu_stat(p, smp_processor_id(), 0);
+out:
+       raw_spin_unlock(&p->pi_lock);
+}
+
+/**
+ * wake_up_process - Wake up a specific process
+ * @p: The process to be woken up.
+ *
+ * Attempt to wake up the nominated process and move it to the set of runnable
+ * processes.  Returns 1 if the process was woken up, 0 if it was already
+ * running.
+ *
+ * It may be assumed that this function implies a write memory barrier before
+ * changing the task state if and only if any tasks are woken up.
+ */
+int wake_up_process(struct task_struct *p)
+{
+       return try_to_wake_up(p, TASK_ALL, 0);
+}
+EXPORT_SYMBOL(wake_up_process);
+
+int wake_up_state(struct task_struct *p, unsigned int state)
+{
+       return try_to_wake_up(p, state, 0);
+}
+
+/*
+ * Perform scheduler related setup for a newly forked process p.
+ * p is forked by current.
+ *
+ * __sched_fork() is basic setup used by init_idle() too:
+ */
+static void __sched_fork(struct task_struct *p)
+{
+       p->on_rq                        = 0;
+
+       p->se.on_rq                     = 0;
+       p->se.exec_start                = 0;
+       p->se.sum_exec_runtime          = 0;
+       p->se.prev_sum_exec_runtime     = 0;
+       p->se.nr_migrations             = 0;
+       p->se.vruntime                  = 0;
+       INIT_LIST_HEAD(&p->se.group_node);
+
+#ifdef CONFIG_SCHEDSTATS
+       memset(&p->se.statistics, 0, sizeof(p->se.statistics));
+#endif
+
+       INIT_LIST_HEAD(&p->rt.run_list);
+
+#ifdef CONFIG_PREEMPT_NOTIFIERS
+       INIT_HLIST_HEAD(&p->preempt_notifiers);
+#endif
+}
+
+/*
+ * fork()/clone()-time setup:
+ */
+void sched_fork(struct task_struct *p)
+{
+       unsigned long flags;
+       int cpu = get_cpu();
+
+       __sched_fork(p);
+       /*
+        * We mark the process as running 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_RUNNING;
+
+       /*
+        * Revert to default priority/policy on fork if requested.
+        */
+       if (unlikely(p->sched_reset_on_fork)) {
+               if (p->policy == SCHED_FIFO || p->policy == SCHED_RR) {
+                       p->policy = SCHED_NORMAL;
+                       p->normal_prio = p->static_prio;
+               }
+
+               if (PRIO_TO_NICE(p->static_prio) < 0) {
+                       p->static_prio = NICE_TO_PRIO(0);
+                       p->normal_prio = p->static_prio;
+                       set_load_weight(p);
+               }
+
+               /*
+                * We don't need the reset flag anymore after the fork. It has
+                * fulfilled its duty:
+                */
+               p->sched_reset_on_fork = 0;
+       }
+
+       /*
+        * Make sure we do not leak PI boosting priority to the child.
+        */
+       p->prio = current->normal_prio;
+
+       if (!rt_prio(p->prio))
+               p->sched_class = &fair_sched_class;
+
+       if (p->sched_class->task_fork)
+               p->sched_class->task_fork(p);
+
+       /*
+        * The child is not yet in the pid-hash so no cgroup attach races,
+        * and the cgroup is pinned to this child due to cgroup_fork()
+        * is ran before sched_fork().
+        *
+        * Silence PROVE_RCU.
+        */
+       raw_spin_lock_irqsave(&p->pi_lock, flags);
+       set_task_cpu(p, cpu);
+       raw_spin_unlock_irqrestore(&p->pi_lock, flags);
+
+#if defined(CONFIG_SCHEDSTATS) || defined(CONFIG_TASK_DELAY_ACCT)
+       if (likely(sched_info_on()))
+               memset(&p->sched_info, 0, sizeof(p->sched_info));
+#endif
+#if defined(CONFIG_SMP)
+       p->on_cpu = 0;
+#endif
+#ifdef CONFIG_PREEMPT_COUNT
+       /* Want to start with kernel preemption disabled. */
+       task_thread_info(p)->preempt_count = 1;
+#endif
+#ifdef CONFIG_SMP
+       plist_node_init(&p->pushable_tasks, MAX_PRIO);
+#endif
+
+       put_cpu();
+}
+
+/*
+ * wake_up_new_task - wake up a newly created task for the first time.
+ *
+ * This function will do some initial scheduler statistics housekeeping
+ * that must be done for every newly created context, then puts the task
+ * on the runqueue and wakes it.
+ */
+void wake_up_new_task(struct task_struct *p)
+{
+       unsigned long flags;
+       struct rq *rq;
+
+       raw_spin_lock_irqsave(&p->pi_lock, flags);
+#ifdef CONFIG_SMP
+       /*
+        * Fork balancing, do it here and not earlier because:
+        *  - cpus_allowed can change in the fork path
+        *  - any previously selected cpu might disappear through hotplug
+        */
+       set_task_cpu(p, select_task_rq(p, SD_BALANCE_FORK, 0));
+#endif
+
+       rq = __task_rq_lock(p);
+       activate_task(rq, p, 0);
+       p->on_rq = 1;
+       trace_sched_wakeup_new(p, true);
+       check_preempt_curr(rq, p, WF_FORK);
+#ifdef CONFIG_SMP
+       if (p->sched_class->task_woken)
+               p->sched_class->task_woken(rq, p);
+#endif
+       task_rq_unlock(rq, p, &flags);
+}
+
+#ifdef CONFIG_PREEMPT_NOTIFIERS
+
+/**
+ * preempt_notifier_register - tell me when current is being preempted & rescheduled
+ * @notifier: notifier struct to register
+ */
+void preempt_notifier_register(struct preempt_notifier *notifier)
+{
+       hlist_add_head(&notifier->link, &current->preempt_notifiers);
+}
+EXPORT_SYMBOL_GPL(preempt_notifier_register);
+
+/**
+ * preempt_notifier_unregister - no longer interested in preemption notifications
+ * @notifier: notifier struct to unregister
+ *
+ * This is safe to call from within a preemption notifier.
+ */
+void preempt_notifier_unregister(struct preempt_notifier *notifier)
+{
+       hlist_del(&notifier->link);
+}
+EXPORT_SYMBOL_GPL(preempt_notifier_unregister);
+
+static void fire_sched_in_preempt_notifiers(struct task_struct *curr)
+{
+       struct preempt_notifier *notifier;
+       struct hlist_node *node;
+
+       hlist_for_each_entry(notifier, node, &curr->preempt_notifiers, link)
+               notifier->ops->sched_in(notifier, raw_smp_processor_id());
+}
+
+static void
+fire_sched_out_preempt_notifiers(struct task_struct *curr,
+                                struct task_struct *next)
+{
+       struct preempt_notifier *notifier;
+       struct hlist_node *node;
+
+       hlist_for_each_entry(notifier, node, &curr->preempt_notifiers, link)
+               notifier->ops->sched_out(notifier, next);
+}
+
+#else /* !CONFIG_PREEMPT_NOTIFIERS */
+
+static void fire_sched_in_preempt_notifiers(struct task_struct *curr)
+{
+}
+
+static void
+fire_sched_out_preempt_notifiers(struct task_struct *curr,
+                                struct task_struct *next)
+{
+}
+
+#endif /* CONFIG_PREEMPT_NOTIFIERS */
+
+/**
+ * prepare_task_switch - prepare to switch tasks
+ * @rq: the runqueue preparing to switch
+ * @prev: the current task that is being switched out
+ * @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(struct rq *rq, struct task_struct *prev,
+                   struct task_struct *next)
+{
+       sched_info_switch(prev, next);
+       perf_event_task_sched_out(prev, next);
+       fire_sched_out_preempt_notifiers(prev, next);
+       prepare_lock_switch(rq, next);
+       prepare_arch_switch(next);
+       trace_sched_switch(prev, next);
+}
+
+/**
+ * finish_task_switch - clean up after a task-switch
+ * @rq: runqueue associated with task-switch
+ * @prev: the thread we just switched away from.
+ *
+ * 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 void finish_task_switch(struct rq *rq, struct task_struct *prev)
+       __releases(rq->lock)
+{
+       struct mm_struct *mm = rq->prev_mm;
+       long prev_state;
+
+       rq->prev_mm = NULL;
+
+       /*
+        * A task struct has one reference for the use as "current".
+        * If a task dies, then it sets TASK_DEAD in tsk->state and calls
+        * schedule one last time. The schedule call will never return, and
+        * the scheduled task must drop that reference.
+        * The test for TASK_DEAD must occur while the runqueue locks are
+        * still held, otherwise prev could be scheduled on another cpu, die
+        * there before we look at prev->state, and then the reference would
+        * be dropped twice.
+        *              Manfred Spraul <manfred@colorfullife.com>
+        */
+       prev_state = prev->state;
+       finish_arch_switch(prev);
+#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
+       local_irq_disable();
+#endif /* __ARCH_WANT_INTERRUPTS_ON_CTXSW */
+       perf_event_task_sched_in(prev, current);
+#ifdef __ARCH_WANT_INTERRUPTS_ON_CTXSW
+       local_irq_enable();
+#endif /* __ARCH_WANT_INTERRUPTS_ON_CTXSW */
+       finish_lock_switch(rq, prev);
+
+       fire_sched_in_preempt_notifiers(current);
+       if (mm)
+               mmdrop(mm);
+       if (unlikely(prev_state == TASK_DEAD)) {
+               /*
+                * Remove function-return probe instances associated with this
+                * task and put them back on the free list.
+                */
+               kprobe_flush_task(prev);
+               put_task_struct(prev);
+       }
+}
+
+#ifdef CONFIG_SMP
+
+/* assumes rq->lock is held */
+static inline void pre_schedule(struct rq *rq, struct task_struct *prev)
+{
+       if (prev->sched_class->pre_schedule)
+               prev->sched_class->pre_schedule(rq, prev);
+}
+
+/* rq->lock is NOT held, but preemption is disabled */
+static inline void post_schedule(struct rq *rq)
+{
+       if (rq->post_schedule) {
+               unsigned long flags;
+
+               raw_spin_lock_irqsave(&rq->lock, flags);
+               if (rq->curr->sched_class->post_schedule)
+                       rq->curr->sched_class->post_schedule(rq);
+               raw_spin_unlock_irqrestore(&rq->lock, flags);
+
+               rq->post_schedule = 0;
+       }
+}
+
+#else
+
+static inline void pre_schedule(struct rq *rq, struct task_struct *p)
+{
+}
+
+static inline void post_schedule(struct rq *rq)
+{
+}
+
+#endif
+
+/**
+ * schedule_tail - first thing a freshly forked thread must call.
+ * @prev: the thread we just switched away from.
+ */
+asmlinkage void schedule_tail(struct task_struct *prev)
+       __releases(rq->lock)
+{
+       struct rq *rq = this_rq();
+
+       finish_task_switch(rq, prev);
+
+       /*
+        * FIXME: do we need to worry about rq being invalidated by the
+        * task_switch?
+        */
+       post_schedule(rq);
+
+#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(task_pid_vnr(current), current->set_child_tid);
+}
+
+/*
+ * context_switch - switch to the new MM and the new
+ * thread's register state.
+ */
+static inline void
+context_switch(struct rq *rq, struct task_struct *prev,
+              struct task_struct *next)
+{
+       struct mm_struct *mm, *oldmm;
+
+       prepare_task_switch(rq, prev, next);
+
+       mm = next->mm;
+       oldmm = prev->active_mm;
+       /*
+        * For paravirt, this is coupled with an exit in switch_to to
+        * combine the page table reload and the switch backend into
+        * one hypercall.
+        */
+       arch_start_context_switch(prev);
+
+       if (!mm) {
+               next->active_mm = oldmm;
+               atomic_inc(&oldmm->mm_count);
+               enter_lazy_tlb(oldmm, next);
+       } else
+               switch_mm(oldmm, mm, next);
+
+       if (!prev->mm) {
+               prev->active_mm = NULL;
+               rq->prev_mm = oldmm;
+       }
+       /*
+        * Since the runqueue lock will be released by the next
+        * task (which is an invalid locking op but in the case
+        * of the scheduler it's an obvious special-case), so we
+        * do an early lockdep release here:
+        */
+#ifndef __ARCH_WANT_UNLOCKED_CTXSW
+       spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
+#endif
+
+       /* Here we just switch the register state and the stack. */
+       switch_to(prev, next, prev);
+
+       barrier();
+       /*
+        * 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);
+}
+
+/*
+ * nr_running, nr_uninterruptible and nr_context_switches:
+ *
+ * externally visible scheduler statistics: current number of runnable
+ * threads, current number of uninterruptible-sleeping threads, total
+ * number of context switches performed since bootup.
+ */
+unsigned long nr_running(void)
+{
+       unsigned long i, sum = 0;
+
+       for_each_online_cpu(i)
+               sum += cpu_rq(i)->nr_running;
+
+       return sum;
+}
+
+unsigned long nr_uninterruptible(void)
+{
+       unsigned long i, sum = 0;
+
+       for_each_possible_cpu(i)
+               sum += cpu_rq(i)->nr_uninterruptible;
+
+       /*
+        * Since we read the counters lockless, it might be slightly
+        * inaccurate. Do not allow it to go below zero though:
+        */
+       if (unlikely((long)sum < 0))
+               sum = 0;
+
+       return sum;
+}
+
+unsigned long long nr_context_switches(void)
+{
+       int i;
+       unsigned long long sum = 0;
+
+       for_each_possible_cpu(i)
+               sum += cpu_rq(i)->nr_switches;
+
+       return sum;
+}
+
+unsigned long nr_iowait(void)
+{
+       unsigned long i, sum = 0;
+
+       for_each_possible_cpu(i)
+               sum += atomic_read(&cpu_rq(i)->nr_iowait);
+
+       return sum;
+}
+
+unsigned long nr_iowait_cpu(int cpu)
+{
+       struct rq *this = cpu_rq(cpu);
+       return atomic_read(&this->nr_iowait);
+}
+
+unsigned long this_cpu_load(void)
+{
+       struct rq *this = this_rq();
+       return this->cpu_load[0];
+}
+
+
+/* Variables and functions for calc_load */
+static atomic_long_t calc_load_tasks;
+static unsigned long calc_load_update;
+unsigned long avenrun[3];
+EXPORT_SYMBOL(avenrun);
+
+static long calc_load_fold_active(struct rq *this_rq)
+{
+       long nr_active, delta = 0;
+
+       nr_active = this_rq->nr_running;
+       nr_active += (long) this_rq->nr_uninterruptible;
+
+       if (nr_active != this_rq->calc_load_active) {
+               delta = nr_active - this_rq->calc_load_active;
+               this_rq->calc_load_active = nr_active;
+       }
+
+       return delta;
+}
+
+static unsigned long
+calc_load(unsigned long load, unsigned long exp, unsigned long active)
+{
+       load *= exp;
+       load += active * (FIXED_1 - exp);
+       load += 1UL << (FSHIFT - 1);
+       return load >> FSHIFT;
+}
+
+#ifdef CONFIG_NO_HZ
+/*
+ * For NO_HZ we delay the active fold to the next LOAD_FREQ update.
+ *
+ * When making the ILB scale, we should try to pull this in as well.
+ */
+static atomic_long_t calc_load_tasks_idle;
+
+static void calc_load_account_idle(struct rq *this_rq)
+{
+       long delta;
+
+       delta = calc_load_fold_active(this_rq);
+       if (delta)
+               atomic_long_add(delta, &calc_load_tasks_idle);
+}
+
+static long calc_load_fold_idle(void)
+{
+       long delta = 0;
+
+       /*
+        * Its got a race, we don't care...
+        */
+       if (atomic_long_read(&calc_load_tasks_idle))
+               delta = atomic_long_xchg(&calc_load_tasks_idle, 0);
+
+       return delta;
+}
+
+/**
+ * fixed_power_int - compute: x^n, in O(log n) time
+ *
+ * @x:         base of the power
+ * @frac_bits: fractional bits of @x
+ * @n:         power to raise @x to.
+ *
+ * By exploiting the relation between the definition of the natural power
+ * function: x^n := x*x*...*x (x multiplied by itself for n times), and
+ * the binary encoding of numbers used by computers: n := \Sum n_i * 2^i,
+ * (where: n_i \elem {0, 1}, the binary vector representing n),
+ * we find: x^n := x^(\Sum n_i * 2^i) := \Prod x^(n_i * 2^i), which is
+ * of course trivially computable in O(log_2 n), the length of our binary
+ * vector.
+ */
+static unsigned long
+fixed_power_int(unsigned long x, unsigned int frac_bits, unsigned int n)
+{
+       unsigned long result = 1UL << frac_bits;
+
+       if (n) for (;;) {
+               if (n & 1) {
+                       result *= x;
+                       result += 1UL << (frac_bits - 1);
+                       result >>= frac_bits;
+               }
+               n >>= 1;
+               if (!n)
+                       break;
+               x *= x;
+               x += 1UL << (frac_bits - 1);
+               x >>= frac_bits;
+       }
+
+       return result;
+}
+
+/*
+ * a1 = a0 * e + a * (1 - e)
+ *
+ * a2 = a1 * e + a * (1 - e)
+ *    = (a0 * e + a * (1 - e)) * e + a * (1 - e)
+ *    = a0 * e^2 + a * (1 - e) * (1 + e)
+ *
+ * a3 = a2 * e + a * (1 - e)
+ *    = (a0 * e^2 + a * (1 - e) * (1 + e)) * e + a * (1 - e)
+ *    = a0 * e^3 + a * (1 - e) * (1 + e + e^2)
+ *
+ *  ...
+ *
+ * an = a0 * e^n + a * (1 - e) * (1 + e + ... + e^n-1) [1]
+ *    = a0 * e^n + a * (1 - e) * (1 - e^n)/(1 - e)
+ *    = a0 * e^n + a * (1 - e^n)
+ *
+ * [1] application of the geometric series:
+ *
+ *              n         1 - x^(n+1)
+ *     S_n := \Sum x^i = -------------
+ *             i=0          1 - x
+ */
+static unsigned long
+calc_load_n(unsigned long load, unsigned long exp,
+           unsigned long active, unsigned int n)
+{
+
+       return calc_load(load, fixed_power_int(exp, FSHIFT, n), active);
+}
+
+/*
+ * NO_HZ can leave us missing all per-cpu ticks calling
+ * calc_load_account_active(), but since an idle CPU folds its delta into
+ * calc_load_tasks_idle per calc_load_account_idle(), all we need to do is fold
+ * in the pending idle delta if our idle period crossed a load cycle boundary.
+ *
+ * Once we've updated the global active value, we need to apply the exponential
+ * weights adjusted to the number of cycles missed.
+ */
+static void calc_global_nohz(unsigned long ticks)
+{
+       long delta, active, n;
+
+       if (time_before(jiffies, calc_load_update))
+               return;
+
+       /*
+        * If we crossed a calc_load_update boundary, make sure to fold
+        * any pending idle changes, the respective CPUs might have
+        * missed the tick driven calc_load_account_active() update
+        * due to NO_HZ.
+        */
+       delta = calc_load_fold_idle();
+       if (delta)
+               atomic_long_add(delta, &calc_load_tasks);
+
+       /*
+        * If we were idle for multiple load cycles, apply them.
+        */
+       if (ticks >= LOAD_FREQ) {
+               n = ticks / LOAD_FREQ;
+
+               active = atomic_long_read(&calc_load_tasks);
+               active = active > 0 ? active * FIXED_1 : 0;
+
+               avenrun[0] = calc_load_n(avenrun[0], EXP_1, active, n);
+               avenrun[1] = calc_load_n(avenrun[1], EXP_5, active, n);
+               avenrun[2] = calc_load_n(avenrun[2], EXP_15, active, n);
+
+               calc_load_update += n * LOAD_FREQ;
+       }
+
+       /*
+        * Its possible the remainder of the above division also crosses
+        * a LOAD_FREQ period, the regular check in calc_global_load()
+        * which comes after this will take care of that.
+        *
+        * Consider us being 11 ticks before a cycle completion, and us
+        * sleeping for 4*LOAD_FREQ + 22 ticks, then the above code will
+        * age us 4 cycles, and the test in calc_global_load() will
+        * pick up the final one.
+        */
+}
+#else
+static void calc_load_account_idle(struct rq *this_rq)
+{
+}
+
+static inline long calc_load_fold_idle(void)
+{
+       return 0;
+}
+
+static void calc_global_nohz(unsigned long ticks)
+{
+}
+#endif
+
+/**
+ * get_avenrun - get the load average array
+ * @loads:     pointer to dest load array
+ * @offset:    offset to add
+ * @shift:     shift count to shift the result left
+ *
+ * These values are estimates at best, so no need for locking.
+ */
+void get_avenrun(unsigned long *loads, unsigned long offset, int shift)
+{
+       loads[0] = (avenrun[0] + offset) << shift;
+       loads[1] = (avenrun[1] + offset) << shift;
+       loads[2] = (avenrun[2] + offset) << shift;
+}
+
+/*
+ * calc_load - update the avenrun load estimates 10 ticks after the
+ * CPUs have updated calc_load_tasks.
+ */
+void calc_global_load(unsigned long ticks)
+{
+       long active;
+
+       calc_global_nohz(ticks);
+
+       if (time_before(jiffies, calc_load_update + 10))
+               return;
+
+       active = atomic_long_read(&calc_load_tasks);
+       active = active > 0 ? active * FIXED_1 : 0;
+
+       avenrun[0] = calc_load(avenrun[0], EXP_1, active);
+       avenrun[1] = calc_load(avenrun[1], EXP_5, active);
+       avenrun[2] = calc_load(avenrun[2], EXP_15, active);
+
+       calc_load_update += LOAD_FREQ;
+}
+
+/*
+ * Called from update_cpu_load() to periodically update this CPU's
+ * active count.
+ */
+static void calc_load_account_active(struct rq *this_rq)
+{
+       long delta;
+
+       if (time_before(jiffies, this_rq->calc_load_update))
+               return;
+
+       delta  = calc_load_fold_active(this_rq);
+       delta += calc_load_fold_idle();
+       if (delta)
+               atomic_long_add(delta, &calc_load_tasks);
+
+       this_rq->calc_load_update += LOAD_FREQ;
+}
+
+/*
+ * The exact cpuload at various idx values, calculated at every tick would be
+ * load = (2^idx - 1) / 2^idx * load + 1 / 2^idx * cur_load
+ *
+ * If a cpu misses updates for n-1 ticks (as it was idle) and update gets called
+ * on nth tick when cpu may be busy, then we have:
+ * load = ((2^idx - 1) / 2^idx)^(n-1) * load
+ * load = (2^idx - 1) / 2^idx) * load + 1 / 2^idx * cur_load
+ *
+ * decay_load_missed() below does efficient calculation of
+ * load = ((2^idx - 1) / 2^idx)^(n-1) * load
+ * avoiding 0..n-1 loop doing load = ((2^idx - 1) / 2^idx) * load
+ *
+ * The calculation is approximated on a 128 point scale.
+ * degrade_zero_ticks is the number of ticks after which load at any
+ * particular idx is approximated to be zero.
+ * degrade_factor is a precomputed table, a row for each load idx.
+ * Each column corresponds to degradation factor for a power of two ticks,
+ * based on 128 point scale.
+ * Example:
+ * row 2, col 3 (=12) says that the degradation at load idx 2 after
+ * 8 ticks is 12/128 (which is an approximation of exact factor 3^8/4^8).
+ *
+ * With this power of 2 load factors, we can degrade the load n times
+ * by looking at 1 bits in n and doing as many mult/shift instead of
+ * n mult/shifts needed by the exact degradation.
+ */
+#define DEGRADE_SHIFT          7
+static const unsigned char
+               degrade_zero_ticks[CPU_LOAD_IDX_MAX] = {0, 8, 32, 64, 128};
+static const unsigned char
+               degrade_factor[CPU_LOAD_IDX_MAX][DEGRADE_SHIFT + 1] = {
+                                       {0, 0, 0, 0, 0, 0, 0, 0},
+                                       {64, 32, 8, 0, 0, 0, 0, 0},
+                                       {96, 72, 40, 12, 1, 0, 0},
+                                       {112, 98, 75, 43, 15, 1, 0},
+                                       {120, 112, 98, 76, 45, 16, 2} };
+
+/*
+ * Update cpu_load for any missed ticks, due to tickless idle. The backlog
+ * would be when CPU is idle and so we just decay the old load without
+ * adding any new load.
+ */
+static unsigned long
+decay_load_missed(unsigned long load, unsigned long missed_updates, int idx)
+{
+       int j = 0;
+
+       if (!missed_updates)
+               return load;
+
+       if (missed_updates >= degrade_zero_ticks[idx])
+               return 0;
+
+       if (idx == 1)
+               return load >> missed_updates;
+
+       while (missed_updates) {
+               if (missed_updates % 2)
+                       load = (load * degrade_factor[idx][j]) >> DEGRADE_SHIFT;
+
+               missed_updates >>= 1;
+               j++;
+       }
+       return load;
+}
+
+/*
+ * Update rq->cpu_load[] statistics. This function is usually called every
+ * scheduler tick (TICK_NSEC). With tickless idle this will not be called
+ * every tick. We fix it up based on jiffies.
+ */
+static void update_cpu_load(struct rq *this_rq)
+{
+       unsigned long this_load = this_rq->load.weight;
+       unsigned long curr_jiffies = jiffies;
+       unsigned long pending_updates;
+       int i, scale;
+
+       this_rq->nr_load_updates++;
+
+       /* Avoid repeated calls on same jiffy, when moving in and out of idle */
+       if (curr_jiffies == this_rq->last_load_update_tick)
+               return;
+
+       pending_updates = curr_jiffies - this_rq->last_load_update_tick;
+       this_rq->last_load_update_tick = curr_jiffies;
+
+       /* Update our load: */
+       this_rq->cpu_load[0] = this_load; /* Fasttrack for idx 0 */
+       for (i = 1, scale = 2; i < CPU_LOAD_IDX_MAX; i++, scale += scale) {
+               unsigned long old_load, new_load;
+
+               /* scale is effectively 1 << i now, and >> i divides by scale */
+
+               old_load = this_rq->cpu_load[i];
+               old_load = decay_load_missed(old_load, pending_updates - 1, i);
+               new_load = this_load;
+               /*
+                * Round up the averaging division if load is increasing. This
+                * prevents us from getting stuck on 9 if the load is 10, for
+                * example.
+                */
+               if (new_load > old_load)
+                       new_load += scale - 1;
+
+               this_rq->cpu_load[i] = (old_load * (scale - 1) + new_load) >> i;
+       }
+
+       sched_avg_update(this_rq);
+}
+
+static void update_cpu_load_active(struct rq *this_rq)
+{
+       update_cpu_load(this_rq);
+
+       calc_load_account_active(this_rq);
+}
+
+#ifdef CONFIG_SMP
+
+/*
+ * 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)
+{
+       struct task_struct *p = current;
+       unsigned long flags;
+       int dest_cpu;
+
+       raw_spin_lock_irqsave(&p->pi_lock, flags);
+       dest_cpu = p->sched_class->select_task_rq(p, SD_BALANCE_EXEC, 0);
+       if (dest_cpu == smp_processor_id())
+               goto unlock;
+
+       if (likely(cpu_active(dest_cpu))) {
+               struct migration_arg arg = { p, dest_cpu };
+
+               raw_spin_unlock_irqrestore(&p->pi_lock, flags);
+               stop_one_cpu(task_cpu(p), migration_cpu_stop, &arg);
+               return;
+       }
+unlock:
+       raw_spin_unlock_irqrestore(&p->pi_lock, flags);
+}
+
+#endif
+
+DEFINE_PER_CPU(struct kernel_stat, kstat);
+
+EXPORT_PER_CPU_SYMBOL(kstat);
+
+/*
+ * Return any ns on the sched_clock that have not yet been accounted in
+ * @p in case that task is currently running.
+ *
+ * Called with task_rq_lock() held on @rq.
+ */
+static u64 do_task_delta_exec(struct task_struct *p, struct rq *rq)
+{
+       u64 ns = 0;
+
+       if (task_current(rq, p)) {
+               update_rq_clock(rq);
+               ns = rq->clock_task - p->se.exec_start;
+               if ((s64)ns < 0)
+                       ns = 0;
+       }
+
+       return ns;
+}
+
+unsigned long long task_delta_exec(struct task_struct *p)
+{
+       unsigned long flags;
+       struct rq *rq;
+       u64 ns = 0;
+
+       rq = task_rq_lock(p, &flags);
+       ns = do_task_delta_exec(p, rq);
+       task_rq_unlock(rq, p, &flags);
+
+       return ns;
+}
+
+/*
+ * Return accounted runtime for the task.
+ * In case the task is currently running, return the runtime plus current's
+ * pending runtime that have not been accounted yet.
+ */
+unsigned long long task_sched_runtime(struct task_struct *p)
+{
+       unsigned long flags;
+       struct rq *rq;
+       u64 ns = 0;
+
+       rq = task_rq_lock(p, &flags);
+       ns = p->se.sum_exec_runtime + do_task_delta_exec(p, rq);
+       task_rq_unlock(rq, p, &flags);
+
+       return ns;
+}
+
+/*
+ * Account user cpu time to a process.
+ * @p: the process that the cpu time gets accounted to
+ * @cputime: the cpu time spent in user space since the last update
+ * @cputime_scaled: cputime scaled by cpu frequency
+ */
+void account_user_time(struct task_struct *p, cputime_t cputime,
+                      cputime_t cputime_scaled)
+{
+       struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;
+       cputime64_t tmp;
+
+       /* Add user time to process. */
+       p->utime = cputime_add(p->utime, cputime);
+       p->utimescaled = cputime_add(p->utimescaled, cputime_scaled);
+       account_group_user_time(p, cputime);
+
+       /* Add user time to cpustat. */
+       tmp = cputime_to_cputime64(cputime);
+       if (TASK_NICE(p) > 0)
+               cpustat->nice = cputime64_add(cpustat->nice, tmp);
+       else
+               cpustat->user = cputime64_add(cpustat->user, tmp);
+
+       cpuacct_update_stats(p, CPUACCT_STAT_USER, cputime);
+       /* Account for user time used */
+       acct_update_integrals(p);
+}
+
+/*
+ * Account guest cpu time to a process.
+ * @p: the process that the cpu time gets accounted to
+ * @cputime: the cpu time spent in virtual machine since the last update
+ * @cputime_scaled: cputime scaled by cpu frequency
+ */
+static void account_guest_time(struct task_struct *p, cputime_t cputime,
+                              cputime_t cputime_scaled)
+{
+       cputime64_t tmp;
+       struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;
+
+       tmp = cputime_to_cputime64(cputime);
+
+       /* Add guest time to process. */
+       p->utime = cputime_add(p->utime, cputime);
+       p->utimescaled = cputime_add(p->utimescaled, cputime_scaled);
+       account_group_user_time(p, cputime);
+       p->gtime = cputime_add(p->gtime, cputime);
+
+       /* Add guest time to cpustat. */
+       if (TASK_NICE(p) > 0) {
+               cpustat->nice = cputime64_add(cpustat->nice, tmp);
+               cpustat->guest_nice = cputime64_add(cpustat->guest_nice, tmp);
+       } else {
+               cpustat->user = cputime64_add(cpustat->user, tmp);
+               cpustat->guest = cputime64_add(cpustat->guest, tmp);
+       }
+}
+
+/*
+ * Account system cpu time to a process and desired cpustat field
+ * @p: the process that the cpu time gets accounted to
+ * @cputime: the cpu time spent in kernel space since the last update
+ * @cputime_scaled: cputime scaled by cpu frequency
+ * @target_cputime64: pointer to cpustat field that has to be updated
+ */
+static inline
+void __account_system_time(struct task_struct *p, cputime_t cputime,
+                       cputime_t cputime_scaled, cputime64_t *target_cputime64)
+{
+       cputime64_t tmp = cputime_to_cputime64(cputime);
+
+       /* Add system time to process. */
+       p->stime = cputime_add(p->stime, cputime);
+       p->stimescaled = cputime_add(p->stimescaled, cputime_scaled);
+       account_group_system_time(p, cputime);
+
+       /* Add system time to cpustat. */
+       *target_cputime64 = cputime64_add(*target_cputime64, tmp);
+       cpuacct_update_stats(p, CPUACCT_STAT_SYSTEM, cputime);
+
+       /* Account for system time used */
+       acct_update_integrals(p);
+}
+
+/*
+ * Account system cpu time to a process.
+ * @p: the process that the cpu time gets accounted to
+ * @hardirq_offset: the offset to subtract from hardirq_count()
+ * @cputime: the cpu time spent in kernel space since the last update
+ * @cputime_scaled: cputime scaled by cpu frequency
+ */
+void account_system_time(struct task_struct *p, int hardirq_offset,
+                        cputime_t cputime, cputime_t cputime_scaled)
+{
+       struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;
+       cputime64_t *target_cputime64;
+
+       if ((p->flags & PF_VCPU) && (irq_count() - hardirq_offset == 0)) {
+               account_guest_time(p, cputime, cputime_scaled);
+               return;
+       }
+
+       if (hardirq_count() - hardirq_offset)
+               target_cputime64 = &cpustat->irq;
+       else if (in_serving_softirq())
+               target_cputime64 = &cpustat->softirq;
+       else
+               target_cputime64 = &cpustat->system;
+
+       __account_system_time(p, cputime, cputime_scaled, target_cputime64);
+}
+
+/*
+ * Account for involuntary wait time.
+ * @cputime: the cpu time spent in involuntary wait
+ */
+void account_steal_time(cputime_t cputime)
+{
+       struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;
+       cputime64_t cputime64 = cputime_to_cputime64(cputime);
+
+       cpustat->steal = cputime64_add(cpustat->steal, cputime64);
+}
+
+/*
+ * Account for idle time.
+ * @cputime: the cpu time spent in idle wait
+ */
+void account_idle_time(cputime_t cputime)
+{
+       struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;
+       cputime64_t cputime64 = cputime_to_cputime64(cputime);
+       struct rq *rq = this_rq();
+
+       if (atomic_read(&rq->nr_iowait) > 0)
+               cpustat->iowait = cputime64_add(cpustat->iowait, cputime64);
+       else
+               cpustat->idle = cputime64_add(cpustat->idle, cputime64);
+}
+
+static __always_inline bool steal_account_process_tick(void)
+{
+#ifdef CONFIG_PARAVIRT
+       if (static_branch(&paravirt_steal_enabled)) {
+               u64 steal, st = 0;
+
+               steal = paravirt_steal_clock(smp_processor_id());
+               steal -= this_rq()->prev_steal_time;
+
+               st = steal_ticks(steal);
+               this_rq()->prev_steal_time += st * TICK_NSEC;
+
+               account_steal_time(st);
+               return st;
+       }
+#endif
+       return false;
+}
+
+#ifndef CONFIG_VIRT_CPU_ACCOUNTING
+
+#ifdef CONFIG_IRQ_TIME_ACCOUNTING
+/*
+ * Account a tick to a process and cpustat
+ * @p: the process that the cpu time gets accounted to
+ * @user_tick: is the tick from userspace
+ * @rq: the pointer to rq
+ *
+ * Tick demultiplexing follows the order
+ * - pending hardirq update
+ * - pending softirq update
+ * - user_time
+ * - idle_time
+ * - system time
+ *   - check for guest_time
+ *   - else account as system_time
+ *
+ * Check for hardirq is done both for system and user time as there is
+ * no timer going off while we are on hardirq and hence we may never get an
+ * opportunity to update it solely in system time.
+ * p->stime and friends are only updated on system time and not on irq
+ * softirq as those do not count in task exec_runtime any more.
+ */
+static void irqtime_account_process_tick(struct task_struct *p, int user_tick,
+                                               struct rq *rq)
+{
+       cputime_t one_jiffy_scaled = cputime_to_scaled(cputime_one_jiffy);
+       cputime64_t tmp = cputime_to_cputime64(cputime_one_jiffy);
+       struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;
+
+       if (steal_account_process_tick())
+               return;
+
+       if (irqtime_account_hi_update()) {
+               cpustat->irq = cputime64_add(cpustat->irq, tmp);
+       } else if (irqtime_account_si_update()) {
+               cpustat->softirq = cputime64_add(cpustat->softirq, tmp);
+       } else if (this_cpu_ksoftirqd() == p) {
+               /*
+                * ksoftirqd time do not get accounted in cpu_softirq_time.
+                * So, we have to handle it separately here.
+                * Also, p->stime needs to be updated for ksoftirqd.
+                */
+               __account_system_time(p, cputime_one_jiffy, one_jiffy_scaled,
+                                       &cpustat->softirq);
+       } else if (user_tick) {
+               account_user_time(p, cputime_one_jiffy, one_jiffy_scaled);
+       } else if (p == rq->idle) {
+               account_idle_time(cputime_one_jiffy);
+       } else if (p->flags & PF_VCPU) { /* System time or guest time */
+               account_guest_time(p, cputime_one_jiffy, one_jiffy_scaled);
+       } else {
+               __account_system_time(p, cputime_one_jiffy, one_jiffy_scaled,
+                                       &cpustat->system);
+       }
+}
+
+static void irqtime_account_idle_ticks(int ticks)
+{
+       int i;
+       struct rq *rq = this_rq();
+
+       for (i = 0; i < ticks; i++)
+               irqtime_account_process_tick(current, 0, rq);
+}
+#else /* CONFIG_IRQ_TIME_ACCOUNTING */
+static void irqtime_account_idle_ticks(int ticks) {}
+static void irqtime_account_process_tick(struct task_struct *p, int user_tick,
+                                               struct rq *rq) {}
+#endif /* CONFIG_IRQ_TIME_ACCOUNTING */
+
+/*
+ * Account a single tick of cpu time.
+ * @p: the process that the cpu time gets accounted to
+ * @user_tick: indicates if the tick is a user or a system tick
+ */
+void account_process_tick(struct task_struct *p, int user_tick)
+{
+       cputime_t one_jiffy_scaled = cputime_to_scaled(cputime_one_jiffy);
+       struct rq *rq = this_rq();
+
+       if (sched_clock_irqtime) {
+               irqtime_account_process_tick(p, user_tick, rq);
+               return;
+       }
+
+       if (steal_account_process_tick())
+               return;
+
+       if (user_tick)
+               account_user_time(p, cputime_one_jiffy, one_jiffy_scaled);
+       else if ((p != rq->idle) || (irq_count() != HARDIRQ_OFFSET))
+               account_system_time(p, HARDIRQ_OFFSET, cputime_one_jiffy,
+                                   one_jiffy_scaled);
+       else
+               account_idle_time(cputime_one_jiffy);
+}
+
+/*
+ * Account multiple ticks of steal time.
+ * @p: the process from which the cpu time has been stolen
+ * @ticks: number of stolen ticks
+ */
+void account_steal_ticks(unsigned long ticks)
+{
+       account_steal_time(jiffies_to_cputime(ticks));
+}
+
+/*
+ * Account multiple ticks of idle time.
+ * @ticks: number of stolen ticks
+ */
+void account_idle_ticks(unsigned long ticks)
+{
+
+       if (sched_clock_irqtime) {
+               irqtime_account_idle_ticks(ticks);
+               return;
+       }
+
+       account_idle_time(jiffies_to_cputime(ticks));
+}
+
+#endif
+
+/*
+ * Use precise platform statistics if available:
+ */
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+void task_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
+{
+       *ut = p->utime;
+       *st = p->stime;
+}
+
+void thread_group_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
+{
+       struct task_cputime cputime;
+
+       thread_group_cputime(p, &cputime);
+
+       *ut = cputime.utime;
+       *st = cputime.stime;
+}
+#else
+
+#ifndef nsecs_to_cputime
+# define nsecs_to_cputime(__nsecs)     nsecs_to_jiffies(__nsecs)
+#endif
+
+void task_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
+{
+       cputime_t rtime, utime = p->utime, total = cputime_add(utime, p->stime);
+
+       /*
+        * Use CFS's precise accounting:
+        */
+       rtime = nsecs_to_cputime(p->se.sum_exec_runtime);
+
+       if (total) {
+               u64 temp = rtime;
+
+               temp *= utime;
+               do_div(temp, total);
+               utime = (cputime_t)temp;
+       } else
+               utime = rtime;
+
+       /*
+        * Compare with previous values, to keep monotonicity:
+        */
+       p->prev_utime = max(p->prev_utime, utime);
+       p->prev_stime = max(p->prev_stime, cputime_sub(rtime, p->prev_utime));
+
+       *ut = p->prev_utime;
+       *st = p->prev_stime;
+}
+
+/*
+ * Must be called with siglock held.
+ */
+void thread_group_times(struct task_struct *p, cputime_t *ut, cputime_t *st)
+{
+       struct signal_struct *sig = p->signal;
+       struct task_cputime cputime;
+       cputime_t rtime, utime, total;
+
+       thread_group_cputime(p, &cputime);
+
+       total = cputime_add(cputime.utime, cputime.stime);
+       rtime = nsecs_to_cputime(cputime.sum_exec_runtime);
+
+       if (total) {
+               u64 temp = rtime;
+
+               temp *= cputime.utime;
+               do_div(temp, total);
+               utime = (cputime_t)temp;
+       } else
+               utime = rtime;
+
+       sig->prev_utime = max(sig->prev_utime, utime);
+       sig->prev_stime = max(sig->prev_stime,
+                             cputime_sub(rtime, sig->prev_utime));
+
+       *ut = sig->prev_utime;
+       *st = sig->prev_stime;
+}
+#endif
+
+/*
+ * This function gets called by the timer code, with HZ frequency.
+ * We call it with interrupts disabled.
+ */
+void scheduler_tick(void)
+{
+       int cpu = smp_processor_id();
+       struct rq *rq = cpu_rq(cpu);
+       struct task_struct *curr = rq->curr;
+
+       sched_clock_tick();
+
+       raw_spin_lock(&rq->lock);
+       update_rq_clock(rq);
+       update_cpu_load_active(rq);
+       curr->sched_class->task_tick(rq, curr, 0);
+       raw_spin_unlock(&rq->lock);
+
+       perf_event_task_tick();
+
+#ifdef CONFIG_SMP
+       rq->idle_at_tick = idle_cpu(cpu);
+       trigger_load_balance(rq, cpu);
+#endif
+}
+
+notrace unsigned long get_parent_ip(unsigned long addr)
+{
+       if (in_lock_functions(addr)) {
+               addr = CALLER_ADDR2;
+               if (in_lock_functions(addr))
+                       addr = CALLER_ADDR3;
+       }
+       return addr;
+}
+
+#if defined(CONFIG_PREEMPT) && (defined(CONFIG_DEBUG_PREEMPT) || \
+                               defined(CONFIG_PREEMPT_TRACER))
+
+void __kprobes add_preempt_count(int val)
+{
+#ifdef CONFIG_DEBUG_PREEMPT
+       /*
+        * Underflow?
+        */
+       if (DEBUG_LOCKS_WARN_ON((preempt_count() < 0)))
+               return;
+#endif
+       preempt_count() += val;
+#ifdef CONFIG_DEBUG_PREEMPT
+       /*
+        * Spinlock count overflowing soon?
+        */
+       DEBUG_LOCKS_WARN_ON((preempt_count() & PREEMPT_MASK) >=
+                               PREEMPT_MASK - 10);
+#endif
+       if (preempt_count() == val)
+               trace_preempt_off(CALLER_ADDR0, get_parent_ip(CALLER_ADDR1));
+}
+EXPORT_SYMBOL(add_preempt_count);
+
+void __kprobes sub_preempt_count(int val)
+{
+#ifdef CONFIG_DEBUG_PREEMPT
+       /*
+        * Underflow?
+        */
+       if (DEBUG_LOCKS_WARN_ON(val > preempt_count()))
+               return;
+       /*
+        * Is the spinlock portion underflowing?
+        */
+       if (DEBUG_LOCKS_WARN_ON((val < PREEMPT_MASK) &&
+                       !(preempt_count() & PREEMPT_MASK)))
+               return;
+#endif
+
+       if (preempt_count() == val)
+               trace_preempt_on(CALLER_ADDR0, get_parent_ip(CALLER_ADDR1));
+       preempt_count() -= val;
+}
+EXPORT_SYMBOL(sub_preempt_count);
+
+#endif
+
+/*
+ * Print scheduling while atomic bug:
+ */
+static noinline void __schedule_bug(struct task_struct *prev)
+{
+       struct pt_regs *regs = get_irq_regs();
+
+       printk(KERN_ERR "BUG: scheduling while atomic: %s/%d/0x%08x\n",
+               prev->comm, prev->pid, preempt_count());
+
+       debug_show_held_locks(prev);
+       print_modules();
+       if (irqs_disabled())
+               print_irqtrace_events(prev);
+
+       if (regs)
+               show_regs(regs);
+       else
+               dump_stack();
+}
+
+/*
+ * Various schedule()-time debugging checks and statistics:
+ */
+static inline void schedule_debug(struct task_struct *prev)
+{
+       /*
+        * Test if we are atomic. Since do_exit() needs to call into
+        * schedule() atomically, we ignore that path for now.
+        * Otherwise, whine if we are scheduling when we should not be.
+        */
+       if (unlikely(in_atomic_preempt_off() && !prev->exit_state))
+               __schedule_bug(prev);
+
+       profile_hit(SCHED_PROFILING, __builtin_return_address(0));
+
+       schedstat_inc(this_rq(), sched_count);
+}
+
+static void put_prev_task(struct rq *rq, struct task_struct *prev)
+{
+       if (prev->on_rq || rq->skip_clock_update < 0)
+               update_rq_clock(rq);
+       prev->sched_class->put_prev_task(rq, prev);
+}
+
+/*
+ * Pick up the highest-prio task:
+ */
+static inline struct task_struct *
+pick_next_task(struct rq *rq)
+{
+       const struct sched_class *class;
+       struct task_struct *p;
+
+       /*
+        * Optimization: we know that if all tasks are in
+        * the fair class we can call that function directly:
+        */
+       if (likely(rq->nr_running == rq->cfs.nr_running)) {
+               p = fair_sched_class.pick_next_task(rq);
+               if (likely(p))
+                       return p;
+       }
+
+       for_each_class(class) {
+               p = class->pick_next_task(rq);
+               if (p)
+                       return p;
+       }
+
+       BUG(); /* the idle class will always have a runnable task */
+}
+
+/*
+ * __schedule() is the main scheduler function.
+ */
+static void __sched __schedule(void)
+{
+       struct task_struct *prev, *next;
+       unsigned long *switch_count;
+       struct rq *rq;
+       int cpu;
+
+need_resched:
+       preempt_disable();
+       cpu = smp_processor_id();
+       rq = cpu_rq(cpu);
+       rcu_note_context_switch(cpu);
+       prev = rq->curr;
+
+       schedule_debug(prev);
+
+       if (sched_feat(HRTICK))
+               hrtick_clear(rq);
+
+       raw_spin_lock_irq(&rq->lock);
+
+       switch_count = &prev->nivcsw;
+       if (prev->state && !(preempt_count() & PREEMPT_ACTIVE)) {
+               if (unlikely(signal_pending_state(prev->state, prev))) {
+                       prev->state = TASK_RUNNING;
+               } else {
+                       deactivate_task(rq, prev, DEQUEUE_SLEEP);
+                       prev->on_rq = 0;
+
+                       /*
+                        * If a worker went to sleep, notify and ask workqueue
+                        * whether it wants to wake up a task to maintain
+                        * concurrency.
+                        */
+                       if (prev->flags & PF_WQ_WORKER) {
+                               struct task_struct *to_wakeup;
+
+                               to_wakeup = wq_worker_sleeping(prev, cpu);
+                               if (to_wakeup)
+                                       try_to_wake_up_local(to_wakeup);
+                       }
+               }
+               switch_count = &prev->nvcsw;
+       }
+
+       pre_schedule(rq, prev);
+
+       if (unlikely(!rq->nr_running))
+               idle_balance(cpu, rq);
+
+       put_prev_task(rq, prev);
+       next = pick_next_task(rq);
+       clear_tsk_need_resched(prev);
+       rq->skip_clock_update = 0;
+
+       if (likely(prev != next)) {
+               rq->nr_switches++;
+               rq->curr = next;
+               ++*switch_count;
+
+               context_switch(rq, prev, next); /* unlocks the rq */
+               /*
+                * The context switch have flipped the stack from under us
+                * and restored the local variables which were saved when
+                * this task called schedule() in the past. prev == current
+                * is still correct, but it can be moved to another cpu/rq.
+                */
+               cpu = smp_processor_id();
+               rq = cpu_rq(cpu);
+       } else
+               raw_spin_unlock_irq(&rq->lock);
+
+       post_schedule(rq);
+
+       preempt_enable_no_resched();
+       if (need_resched())
+               goto need_resched;
+}
+
+static inline void sched_submit_work(struct task_struct *tsk)
+{
+       if (!tsk->state)
+               return;
+       /*
+        * If we are going to sleep and we have plugged IO queued,
+        * make sure to submit it to avoid deadlocks.
+        */
+       if (blk_needs_flush_plug(tsk))
+               blk_schedule_flush_plug(tsk);
+}
+
+asmlinkage void __sched schedule(void)
+{
+       struct task_struct *tsk = current;
+
+       sched_submit_work(tsk);
+       __schedule();
+}
+EXPORT_SYMBOL(schedule);
+
+#ifdef CONFIG_MUTEX_SPIN_ON_OWNER
+
+static inline bool owner_running(struct mutex *lock, struct task_struct *owner)
+{
+       if (lock->owner != owner)
+               return false;
+
+       /*
+        * Ensure we emit the owner->on_cpu, dereference _after_ checking
+        * lock->owner still matches owner, if that fails, owner might
+        * point to free()d memory, if it still matches, the rcu_read_lock()
+        * ensures the memory stays valid.
+        */
+       barrier();
+
+       return owner->on_cpu;
+}
+
+/*
+ * Look out! "owner" is an entirely speculative pointer
+ * access and not reliable.
+ */
+int mutex_spin_on_owner(struct mutex *lock, struct task_struct *owner)
+{
+       if (!sched_feat(OWNER_SPIN))
+               return 0;
+
+       rcu_read_lock();
+       while (owner_running(lock, owner)) {
+               if (need_resched())
+                       break;
+
+               arch_mutex_cpu_relax();
+       }
+       rcu_read_unlock();
+
+       /*
+        * We break out the loop above on need_resched() and when the
+        * owner changed, which is a sign for heavy contention. Return
+        * success only when lock->owner is NULL.
+        */
+       return lock->owner == NULL;
+}
+#endif
+
+#ifdef CONFIG_PREEMPT
+/*
+ * this is the entry point to schedule() from in-kernel preemption
+ * off of preempt_enable. Kernel preemptions off return from interrupt
+ * occur there and call schedule directly.
+ */
+asmlinkage void __sched notrace preempt_schedule(void)
+{
+       struct thread_info *ti = current_thread_info();
+
+       /*
+        * If there is a non-zero preempt_count or interrupts are disabled,
+        * we do not want to preempt the current task. Just return..
+        */
+       if (likely(ti->preempt_count || irqs_disabled()))
+               return;
+
+       do {
+               add_preempt_count_notrace(PREEMPT_ACTIVE);
+               __schedule();
+               sub_preempt_count_notrace(PREEMPT_ACTIVE);
+
+               /*
+                * Check again in case we missed a preemption opportunity
+                * between schedule and now.
+                */
+               barrier();
+       } while (need_resched());
+}
+EXPORT_SYMBOL(preempt_schedule);
+
+/*
+ * this is the entry point to schedule() from kernel preemption
+ * off of irq context.
+ * Note, that this is called and return with irqs disabled. This will
+ * protect us against recursive calling from irq.
+ */
+asmlinkage void __sched preempt_schedule_irq(void)
+{
+       struct thread_info *ti = current_thread_info();
+
+       /* Catch callers which need to be fixed */
+       BUG_ON(ti->preempt_count || !irqs_disabled());
+
+       do {
+               add_preempt_count(PREEMPT_ACTIVE);
+               local_irq_enable();
+               __schedule();
+               local_irq_disable();
+               sub_preempt_count(PREEMPT_ACTIVE);
+
+               /*
+                * Check again in case we missed a preemption opportunity
+                * between schedule and now.
+                */
+               barrier();
+       } while (need_resched());
+}
+
+#endif /* CONFIG_PREEMPT */
+
+int default_wake_function(wait_queue_t *curr, unsigned mode, int wake_flags,
+                         void *key)
+{
+       return try_to_wake_up(curr->private, mode, wake_flags);
+}
+EXPORT_SYMBOL(default_wake_function);
+
+/*
+ * The core wakeup function. Non-exclusive wakeups (nr_exclusive == 0) just
+ * wake everything up. If it's an exclusive wakeup (nr_exclusive == small +ve
+ * number) then we wake all the non-exclusive tasks and one exclusive task.
+ *
+ * There are circumstances in which we can try to wake a task which has already
+ * started to run but is not in state TASK_RUNNING. try_to_wake_up() returns
+ * zero in this (rare) case, and we handle it by continuing to scan the queue.
+ */
+static void __wake_up_common(wait_queue_head_t *q, unsigned int mode,
+                       int nr_exclusive, int wake_flags, void *key)
+{
+       wait_queue_t *curr, *next;
+
+       list_for_each_entry_safe(curr, next, &q->task_list, task_list) {
+               unsigned flags = curr->flags;
+
+               if (curr->func(curr, mode, wake_flags, key) &&
+                               (flags & WQ_FLAG_EXCLUSIVE) && !--nr_exclusive)
+                       break;
+       }
+}
+
+/**
+ * __wake_up - wake up threads blocked on a waitqueue.
+ * @q: the waitqueue
+ * @mode: which threads
+ * @nr_exclusive: how many wake-one or wake-many threads to wake up
+ * @key: is directly passed to the wakeup function
+ *
+ * It may be assumed that this function implies a write memory barrier before
+ * changing the task state if and only if any tasks are woken up.
+ */
+void __wake_up(wait_queue_head_t *q, unsigned int mode,
+                       int nr_exclusive, void *key)
+{
+       unsigned long flags;
+
+       spin_lock_irqsave(&q->lock, flags);
+       __wake_up_common(q, mode, nr_exclusive, 0, key);
+       spin_unlock_irqrestore(&q->lock, flags);
+}
+EXPORT_SYMBOL(__wake_up);
+
+/*
+ * Same as __wake_up but called with the spinlock in wait_queue_head_t held.
+ */
+void __wake_up_locked(wait_queue_head_t *q, unsigned int mode)
+{
+       __wake_up_common(q, mode, 1, 0, NULL);
+}
+EXPORT_SYMBOL_GPL(__wake_up_locked);
+
+void __wake_up_locked_key(wait_queue_head_t *q, unsigned int mode, void *key)
+{
+       __wake_up_common(q, mode, 1, 0, key);
+}
+EXPORT_SYMBOL_GPL(__wake_up_locked_key);
+
+/**
+ * __wake_up_sync_key - wake up threads blocked on a waitqueue.
+ * @q: the waitqueue
+ * @mode: which threads
+ * @nr_exclusive: how many wake-one or wake-many threads to wake up
+ * @key: opaque value to be passed to wakeup targets
+ *
+ * The sync wakeup differs that the waker knows that it will schedule
+ * away soon, so while the target thread will be woken up, it will not
+ * be migrated to another CPU - ie. the two threads are 'synchronized'
+ * with each other. This can prevent needless bouncing between CPUs.
+ *
+ * On UP it can prevent extra preemption.
+ *
+ * It may be assumed that this function implies a write memory barrier before
+ * changing the task state if and only if any tasks are woken up.
+ */
+void __wake_up_sync_key(wait_queue_head_t *q, unsigned int mode,
+                       int nr_exclusive, void *key)
+{
+       unsigned long flags;
+       int wake_flags = WF_SYNC;
+
+       if (unlikely(!q))
+               return;
+
+       if (unlikely(!nr_exclusive))
+               wake_flags = 0;
+
+       spin_lock_irqsave(&q->lock, flags);
+       __wake_up_common(q, mode, nr_exclusive, wake_flags, key);
+       spin_unlock_irqrestore(&q->lock, flags);
+}
+EXPORT_SYMBOL_GPL(__wake_up_sync_key);
+
+/*
+ * __wake_up_sync - see __wake_up_sync_key()
+ */
+void __wake_up_sync(wait_queue_head_t *q, unsigned int mode, int nr_exclusive)
+{
+       __wake_up_sync_key(q, mode, nr_exclusive, NULL);
+}
+EXPORT_SYMBOL_GPL(__wake_up_sync);     /* For internal use only */
+
+/**
+ * complete: - signals a single thread waiting on this completion
+ * @x:  holds the state of this particular completion
+ *
+ * This will wake up a single thread waiting on this completion. Threads will be
+ * awakened in the same order in which they were queued.
+ *
+ * See also complete_all(), wait_for_completion() and related routines.
+ *
+ * It may be assumed that this function implies a write memory barrier before
+ * changing the task state if and only if any tasks are woken up.
+ */
+void complete(struct completion *x)
+{
+       unsigned long flags;
+
+       spin_lock_irqsave(&x->wait.lock, flags);
+       x->done++;
+       __wake_up_common(&x->wait, TASK_NORMAL, 1, 0, NULL);
+       spin_unlock_irqrestore(&x->wait.lock, flags);
+}
+EXPORT_SYMBOL(complete);
+
+/**
+ * complete_all: - signals all threads waiting on this completion
+ * @x:  holds the state of this particular completion
+ *
+ * This will wake up all threads waiting on this particular completion event.
+ *
+ * It may be assumed that this function implies a write memory barrier before
+ * changing the task state if and only if any tasks are woken up.
+ */
+void complete_all(struct completion *x)
+{
+       unsigned long flags;
+
+       spin_lock_irqsave(&x->wait.lock, flags);
+       x->done += UINT_MAX/2;
+       __wake_up_common(&x->wait, TASK_NORMAL, 0, 0, NULL);
+       spin_unlock_irqrestore(&x->wait.lock, flags);
+}
+EXPORT_SYMBOL(complete_all);
+
+static inline long __sched
+do_wait_for_common(struct completion *x, long timeout, int state)
+{
+       if (!x->done) {
+               DECLARE_WAITQUEUE(wait, current);
+
+               __add_wait_queue_tail_exclusive(&x->wait, &wait);
+               do {
+                       if (signal_pending_state(state, current)) {
+                               timeout = -ERESTARTSYS;
+                               break;
+                       }
+                       __set_current_state(state);
+                       spin_unlock_irq(&x->wait.lock);
+                       timeout = schedule_timeout(timeout);
+                       spin_lock_irq(&x->wait.lock);
+               } while (!x->done && timeout);
+               __remove_wait_queue(&x->wait, &wait);
+               if (!x->done)
+                       return timeout;
+       }
+       x->done--;
+       return timeout ?: 1;
+}
+
+static long __sched
+wait_for_common(struct completion *x, long timeout, int state)
+{
+       might_sleep();
+
+       spin_lock_irq(&x->wait.lock);
+       timeout = do_wait_for_common(x, timeout, state);
+       spin_unlock_irq(&x->wait.lock);
+       return timeout;
+}
+
+/**
+ * wait_for_completion: - waits for completion of a task
+ * @x:  holds the state of this particular completion
+ *
+ * This waits to be signaled for completion of a specific task. It is NOT
+ * interruptible and there is no timeout.
+ *
+ * See also similar routines (i.e. wait_for_completion_timeout()) with timeout
+ * and interrupt capability. Also see complete().
+ */
+void __sched wait_for_completion(struct completion *x)
+{
+       wait_for_common(x, MAX_SCHEDULE_TIMEOUT, TASK_UNINTERRUPTIBLE);
+}
+EXPORT_SYMBOL(wait_for_completion);
+
+/**
+ * wait_for_completion_timeout: - waits for completion of a task (w/timeout)
+ * @x:  holds the state of this particular completion
+ * @timeout:  timeout value in jiffies
+ *
+ * This waits for either a completion of a specific task to be signaled or for a
+ * specified timeout to expire. The timeout is in jiffies. It is not
+ * interruptible.
+ */
+unsigned long __sched
+wait_for_completion_timeout(struct completion *x, unsigned long timeout)
+{
+       return wait_for_common(x, timeout, TASK_UNINTERRUPTIBLE);
+}
+EXPORT_SYMBOL(wait_for_completion_timeout);
+
+/**
+ * wait_for_completion_interruptible: - waits for completion of a task (w/intr)
+ * @x:  holds the state of this particular completion
+ *
+ * This waits for completion of a specific task to be signaled. It is
+ * interruptible.
+ */
+int __sched wait_for_completion_interruptible(struct completion *x)
+{
+       long t = wait_for_common(x, MAX_SCHEDULE_TIMEOUT, TASK_INTERRUPTIBLE);
+       if (t == -ERESTARTSYS)
+               return t;
+       return 0;
+}
+EXPORT_SYMBOL(wait_for_completion_interruptible);
+
+/**
+ * wait_for_completion_interruptible_timeout: - waits for completion (w/(to,intr))
+ * @x:  holds the state of this particular completion
+ * @timeout:  timeout value in jiffies
+ *
+ * This waits for either a completion of a specific task to be signaled or for a
+ * specified timeout to expire. It is interruptible. The timeout is in jiffies.
+ */
+long __sched
+wait_for_completion_interruptible_timeout(struct completion *x,
+                                         unsigned long timeout)
+{
+       return wait_for_common(x, timeout, TASK_INTERRUPTIBLE);
+}
+EXPORT_SYMBOL(wait_for_completion_interruptible_timeout);
+
+/**
+ * wait_for_completion_killable: - waits for completion of a task (killable)
+ * @x:  holds the state of this particular completion
+ *
+ * This waits to be signaled for completion of a specific task. It can be
+ * interrupted by a kill signal.
+ */
+int __sched wait_for_completion_killable(struct completion *x)
+{
+       long t = wait_for_common(x, MAX_SCHEDULE_TIMEOUT, TASK_KILLABLE);
+       if (t == -ERESTARTSYS)
+               return t;
+       return 0;
+}
+EXPORT_SYMBOL(wait_for_completion_killable);
+
+/**
+ * wait_for_completion_killable_timeout: - waits for completion of a task (w/(to,killable))
+ * @x:  holds the state of this particular completion
+ * @timeout:  timeout value in jiffies
+ *
+ * This waits for either a completion of a specific task to be
+ * signaled or for a specified timeout to expire. It can be
+ * interrupted by a kill signal. The timeout is in jiffies.
+ */
+long __sched
+wait_for_completion_killable_timeout(struct completion *x,
+                                    unsigned long timeout)
+{
+       return wait_for_common(x, timeout, TASK_KILLABLE);
+}
+EXPORT_SYMBOL(wait_for_completion_killable_timeout);
+
+/**
+ *     try_wait_for_completion - try to decrement a completion without blocking
+ *     @x:     completion structure
+ *
+ *     Returns: 0 if a decrement cannot be done without blocking
+ *              1 if a decrement succeeded.
+ *
+ *     If a completion is being used as a counting completion,
+ *     attempt to decrement the counter without blocking. This
+ *     enables us to avoid waiting if the resource the completion
+ *     is protecting is not available.
+ */
+bool try_wait_for_completion(struct completion *x)
+{
+       unsigned long flags;
+       int ret = 1;
+
+       spin_lock_irqsave(&x->wait.lock, flags);
+       if (!x->done)
+               ret = 0;
+       else
+               x->done--;
+       spin_unlock_irqrestore(&x->wait.lock, flags);
+       return ret;
+}
+EXPORT_SYMBOL(try_wait_for_completion);
+
+/**
+ *     completion_done - Test to see if a completion has any waiters
+ *     @x:     completion structure
+ *
+ *     Returns: 0 if there are waiters (wait_for_completion() in progress)
+ *              1 if there are no waiters.
+ *
+ */
+bool completion_done(struct completion *x)
+{
+       unsigned long flags;
+       int ret = 1;
+
+       spin_lock_irqsave(&x->wait.lock, flags);
+       if (!x->done)
+               ret = 0;
+       spin_unlock_irqrestore(&x->wait.lock, flags);
+       return ret;
+}
+EXPORT_SYMBOL(completion_done);
+
+static long __sched
+sleep_on_common(wait_queue_head_t *q, int state, long timeout)
+{
+       unsigned long flags;
+       wait_queue_t wait;
+
+       init_waitqueue_entry(&wait, current);
+
+       __set_current_state(state);
+
+       spin_lock_irqsave(&q->lock, flags);
+       __add_wait_queue(q, &wait);
+       spin_unlock(&q->lock);
+       timeout = schedule_timeout(timeout);
+       spin_lock_irq(&q->lock);
+       __remove_wait_queue(q, &wait);
+       spin_unlock_irqrestore(&q->lock, flags);
+
+       return timeout;
+}
+
+void __sched interruptible_sleep_on(wait_queue_head_t *q)
+{
+       sleep_on_common(q, TASK_INTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
+}
+EXPORT_SYMBOL(interruptible_sleep_on);
+
+long __sched
+interruptible_sleep_on_timeout(wait_queue_head_t *q, long timeout)
+{
+       return sleep_on_common(q, TASK_INTERRUPTIBLE, timeout);
+}
+EXPORT_SYMBOL(interruptible_sleep_on_timeout);
+
+void __sched sleep_on(wait_queue_head_t *q)
+{
+       sleep_on_common(q, TASK_UNINTERRUPTIBLE, MAX_SCHEDULE_TIMEOUT);
+}
+EXPORT_SYMBOL(sleep_on);
+
+long __sched sleep_on_timeout(wait_queue_head_t *q, long timeout)
+{
+       return sleep_on_common(q, TASK_UNINTERRUPTIBLE, timeout);
+}
+EXPORT_SYMBOL(sleep_on_timeout);
+
+#ifdef CONFIG_RT_MUTEXES
+
+/*
+ * rt_mutex_setprio - set the current priority of a task
+ * @p: task
+ * @prio: prio value (kernel-internal form)
+ *
+ * This function changes the 'effective' priority of a task. It does
+ * not touch ->normal_prio like __setscheduler().
+ *
+ * Used by the rt_mutex code to implement priority inheritance logic.
+ */
+void rt_mutex_setprio(struct task_struct *p, int prio)
+{
+       int oldprio, on_rq, running;
+       struct rq *rq;
+       const struct sched_class *prev_class;
+
+       BUG_ON(prio < 0 || prio > MAX_PRIO);
+
+       rq = __task_rq_lock(p);
+
+       trace_sched_pi_setprio(p, prio);
+       oldprio = p->prio;
+       prev_class = p->sched_class;
+       on_rq = p->on_rq;
+       running = task_current(rq, p);
+       if (on_rq)
+               dequeue_task(rq, p, 0);
+       if (running)
+               p->sched_class->put_prev_task(rq, p);
+
+       if (rt_prio(prio))
+               p->sched_class = &rt_sched_class;
+       else
+               p->sched_class = &fair_sched_class;
+
+       p->prio = prio;
+
+       if (running)
+               p->sched_class->set_curr_task(rq);
+       if (on_rq)
+               enqueue_task(rq, p, oldprio < prio ? ENQUEUE_HEAD : 0);
+
+       check_class_changed(rq, p, prev_class, oldprio);
+       __task_rq_unlock(rq);
+}
+
+#endif
+
+void set_user_nice(struct task_struct *p, long nice)
+{
+       int old_prio, delta, on_rq;
+       unsigned long flags;
+       struct rq *rq;
+
+       if (TASK_NICE(p) == nice || nice < -20 || nice > 19)
+               return;
+       /*
+        * We have to be careful, if called from sys_setpriority(),
+        * the task might be in the middle of scheduling on another CPU.
+        */
+       rq = task_rq_lock(p, &flags);
+       /*
+        * The RT priorities are set via sched_setscheduler(), but we still
+        * allow the 'normal' nice value to be set - but as expected
+        * it wont have any effect on scheduling until the task is
+        * SCHED_FIFO/SCHED_RR:
+        */
+       if (task_has_rt_policy(p)) {
+               p->static_prio = NICE_TO_PRIO(nice);
+               goto out_unlock;
+       }
+       on_rq = p->on_rq;
+       if (on_rq)
+               dequeue_task(rq, p, 0);
+
+       p->static_prio = NICE_TO_PRIO(nice);
+       set_load_weight(p);
+       old_prio = p->prio;
+       p->prio = effective_prio(p);
+       delta = p->prio - old_prio;
+
+       if (on_rq) {
+               enqueue_task(rq, p, 0);
+               /*
+                * If the task increased its priority or is running and
+                * lowered its priority, then reschedule its CPU:
+                */
+               if (delta < 0 || (delta > 0 && task_running(rq, p)))
+                       resched_task(rq->curr);
+       }
+out_unlock:
+       task_rq_unlock(rq, p, &flags);
+}
+EXPORT_SYMBOL(set_user_nice);
+
+/*
+ * can_nice - check if a task can reduce its nice value
+ * @p: task
+ * @nice: nice value
+ */
+int can_nice(const struct task_struct *p, const int nice)
+{
+       /* convert nice value [19,-20] to rlimit style value [1,40] */
+       int nice_rlim = 20 - nice;
+
+       return (nice_rlim <= task_rlimit(p, RLIMIT_NICE) ||
+               capable(CAP_SYS_NICE));
+}
+
+#ifdef __ARCH_WANT_SYS_NICE
+
+/*
+ * sys_nice - change the priority of the current process.
+ * @increment: priority increment
+ *
+ * sys_setpriority is a more generic, but much slower function that
+ * does similar things.
+ */
+SYSCALL_DEFINE1(nice, int, increment)
+{
+       long nice, retval;
+
+       /*
+        * Setpriority might change our priority at the same moment.
+        * We don't have to worry. Conceptually one call occurs first
+        * and we have a single winner.
+        */
+       if (increment < -40)
+               increment = -40;
+       if (increment > 40)
+               increment = 40;
+
+       nice = TASK_NICE(current) + increment;
+       if (nice < -20)
+               nice = -20;
+       if (nice > 19)
+               nice = 19;
+
+       if (increment < 0 && !can_nice(current, nice))
+               return -EPERM;
+
+       retval = security_task_setnice(current, nice);
+       if (retval)
+               return retval;
+
+       set_user_nice(current, nice);
+       return 0;
+}
+
+#endif
+
+/**
+ * task_prio - return the priority value of a given task.
+ * @p: the task in question.
+ *
+ * This is the priority value as seen by users in /proc.
+ * RT tasks are offset by -200. Normal tasks are centered
+ * around 0, value goes from -16 to +15.
+ */
+int task_prio(const struct task_struct *p)
+{
+       return p->prio - MAX_RT_PRIO;
+}
+
+/**
+ * task_nice - return the nice value of a given task.
+ * @p: the task in question.
+ */
+int task_nice(const struct task_struct *p)
+{
+       return TASK_NICE(p);
+}
+EXPORT_SYMBOL(task_nice);
+
+/**
+ * idle_cpu - is a given cpu idle currently?
+ * @cpu: the processor in question.
+ */
+int idle_cpu(int cpu)
+{
+       return cpu_curr(cpu) == cpu_rq(cpu)->idle;
+}
+
+/**
+ * idle_task - return the idle task for a given cpu.
+ * @cpu: the processor in question.
+ */
+struct task_struct *idle_task(int cpu)
+{
+       return cpu_rq(cpu)->idle;
+}
+
+/**
+ * find_process_by_pid - find a process with a matching PID value.
+ * @pid: the pid in question.
+ */
+static struct task_struct *find_process_by_pid(pid_t pid)
+{
+       return pid ? find_task_by_vpid(pid) : current;
+}
+
+/* Actually do priority change: must hold rq lock. */
+static void
+__setscheduler(struct rq *rq, struct task_struct *p, int policy, int prio)
+{
+       p->policy = policy;
+       p->rt_priority = prio;
+       p->normal_prio = normal_prio(p);
+       /* we are holding p->pi_lock already */
+       p->prio = rt_mutex_getprio(p);
+       if (rt_prio(p->prio))
+               p->sched_class = &rt_sched_class;
+       else
+               p->sched_class = &fair_sched_class;
+       set_load_weight(p);
+}
+
+/*
+ * check the target process has a UID that matches the current process's
+ */
+static bool check_same_owner(struct task_struct *p)
+{
+       const struct cred *cred = current_cred(), *pcred;
+       bool match;
+
+       rcu_read_lock();
+       pcred = __task_cred(p);
+       if (cred->user->user_ns == pcred->user->user_ns)
+               match = (cred->euid == pcred->euid ||
+                        cred->euid == pcred->uid);
+       else
+               match = false;
+       rcu_read_unlock();
+       return match;
+}
+
+static int __sched_setscheduler(struct task_struct *p, int policy,
+                               const struct sched_param *param, bool user)
+{
+       int retval, oldprio, oldpolicy = -1, on_rq, running;
+       unsigned long flags;
+       const struct sched_class *prev_class;
+       struct rq *rq;
+       int reset_on_fork;
+
+       /* may grab non-irq protected spin_locks */
+       BUG_ON(in_interrupt());
+recheck:
+       /* double check policy once rq lock held */
+       if (policy < 0) {
+               reset_on_fork = p->sched_reset_on_fork;
+               policy = oldpolicy = p->policy;
+       } else {
+               reset_on_fork = !!(policy & SCHED_RESET_ON_FORK);
+               policy &= ~SCHED_RESET_ON_FORK;
+
+               if (policy != SCHED_FIFO && policy != SCHED_RR &&
+                               policy != SCHED_NORMAL && policy != SCHED_BATCH &&
+                               policy != SCHED_IDLE)
+                       return -EINVAL;
+       }
+
+       /*
+        * Valid priorities for SCHED_FIFO and SCHED_RR are
+        * 1..MAX_USER_RT_PRIO-1, valid priority for SCHED_NORMAL,
+        * SCHED_BATCH and SCHED_IDLE is 0.
+        */
+       if (param->sched_priority < 0 ||
+           (p->mm && param->sched_priority > MAX_USER_RT_PRIO-1) ||
+           (!p->mm && param->sched_priority > MAX_RT_PRIO-1))
+               return -EINVAL;
+       if (rt_policy(policy) != (param->sched_priority != 0))
+               return -EINVAL;
+
+       /*
+        * Allow unprivileged RT tasks to decrease priority:
+        */
+       if (user && !capable(CAP_SYS_NICE)) {
+               if (rt_policy(policy)) {
+                       unsigned long rlim_rtprio =
+                                       task_rlimit(p, RLIMIT_RTPRIO);
+
+                       /* can't set/change the rt policy */
+                       if (policy != p->policy && !rlim_rtprio)
+                               return -EPERM;
+
+                       /* can't increase priority */
+                       if (param->sched_priority > p->rt_priority &&
+                           param->sched_priority > rlim_rtprio)
+                               return -EPERM;
+               }
+
+               /*
+                * Treat SCHED_IDLE as nice 20. Only allow a switch to
+                * SCHED_NORMAL if the RLIMIT_NICE would normally permit it.
+                */
+               if (p->policy == SCHED_IDLE && policy != SCHED_IDLE) {
+                       if (!can_nice(p, TASK_NICE(p)))
+                               return -EPERM;
+               }
+
+               /* can't change other user's priorities */
+               if (!check_same_owner(p))
+                       return -EPERM;
+
+               /* Normal users shall not reset the sched_reset_on_fork flag */
+               if (p->sched_reset_on_fork && !reset_on_fork)
+                       return -EPERM;
+       }
+
+       if (user) {
+               retval = security_task_setscheduler(p);
+               if (retval)
+                       return retval;
+       }
+
+       /*
+        * make sure no PI-waiters arrive (or leave) while we are
+        * changing the priority of the task:
+        *
+        * To be able to change p->policy safely, the appropriate
+        * runqueue lock must be held.
+        */
+       rq = task_rq_lock(p, &flags);
+
+       /*
+        * Changing the policy of the stop threads its a very bad idea
+        */
+       if (p == rq->stop) {
+               task_rq_unlock(rq, p, &flags);
+               return -EINVAL;
+       }
+
+       /*
+        * If not changing anything there's no need to proceed further:
+        */
+       if (unlikely(policy == p->policy && (!rt_policy(policy) ||
+                       param->sched_priority == p->rt_priority))) {
+
+               __task_rq_unlock(rq);
+               raw_spin_unlock_irqrestore(&p->pi_lock, flags);
+               return 0;
+       }
+
+#ifdef CONFIG_RT_GROUP_SCHED
+       if (user) {
+               /*
+                * Do not allow realtime tasks into groups that have no runtime
+                * assigned.
+                */
+               if (rt_bandwidth_enabled() && rt_policy(policy) &&
+                               task_group(p)->rt_bandwidth.rt_runtime == 0 &&
+                               !task_group_is_autogroup(task_group(p))) {
+                       task_rq_unlock(rq, p, &flags);
+                       return -EPERM;
+               }
+       }
+#endif
+
+       /* recheck policy now with rq lock held */
+       if (unlikely(oldpolicy != -1 && oldpolicy != p->policy)) {
+               policy = oldpolicy = -1;
+               task_rq_unlock(rq, p, &flags);
+               goto recheck;
+       }
+       on_rq = p->on_rq;
+       running = task_current(rq, p);
+       if (on_rq)
+               deactivate_task(rq, p, 0);
+       if (running)
+               p->sched_class->put_prev_task(rq, p);
+
+       p->sched_reset_on_fork = reset_on_fork;
+
+       oldprio = p->prio;
+       prev_class = p->sched_class;
+       __setscheduler(rq, p, policy, param->sched_priority);
+
+       if (running)
+               p->sched_class->set_curr_task(rq);
+       if (on_rq)
+               activate_task(rq, p, 0);
+
+       check_class_changed(rq, p, prev_class, oldprio);
+       task_rq_unlock(rq, p, &flags);
+
+       rt_mutex_adjust_pi(p);
+
+       return 0;
+}
+
+/**
+ * sched_setscheduler - change the scheduling policy and/or RT priority of a thread.
+ * @p: the task in question.
+ * @policy: new policy.
+ * @param: structure containing the new RT priority.
+ *
+ * NOTE that the task may be already dead.
+ */
+int sched_setscheduler(struct task_struct *p, int policy,
+                      const struct sched_param *param)
+{
+       return __sched_setscheduler(p, policy, param, true);
+}
+EXPORT_SYMBOL_GPL(sched_setscheduler);
+
+/**
+ * sched_setscheduler_nocheck - change the scheduling policy and/or RT priority of a thread from kernelspace.
+ * @p: the task in question.
+ * @policy: new policy.
+ * @param: structure containing the new RT priority.
+ *
+ * Just like sched_setscheduler, only don't bother checking if the
+ * current context has permission.  For example, this is needed in
+ * stop_machine(): we create temporary high priority worker threads,
+ * but our caller might not have that capability.
+ */
+int sched_setscheduler_nocheck(struct task_struct *p, int policy,
+                              const struct sched_param *param)
+{
+       return __sched_setscheduler(p, policy, param, false);
+}
+
+static int
+do_sched_setscheduler(pid_t pid, int policy, struct sched_param __user *param)
+{
+       struct sched_param lparam;
+       struct task_struct *p;
+       int retval;
+
+       if (!param || pid < 0)
+               return -EINVAL;
+       if (copy_from_user(&lparam, param, sizeof(struct sched_param)))
+               return -EFAULT;
+
+       rcu_read_lock();
+       retval = -ESRCH;
+       p = find_process_by_pid(pid);
+       if (p != NULL)
+               retval = sched_setscheduler(p, policy, &lparam);
+       rcu_read_unlock();
+
+       return retval;
+}
+
+/**
+ * sys_sched_setscheduler - set/change the scheduler policy and RT priority
+ * @pid: the pid in question.
+ * @policy: new policy.
+ * @param: structure containing the new RT priority.
+ */
+SYSCALL_DEFINE3(sched_setscheduler, pid_t, pid, int, policy,
+               struct sched_param __user *, param)
+{
+       /* negative values for policy are not valid */
+       if (policy < 0)
+               return -EINVAL;
+
+       return do_sched_setscheduler(pid, policy, param);
+}
+
+/**
+ * sys_sched_setparam - set/change the RT priority of a thread
+ * @pid: the pid in question.
+ * @param: structure containing the new RT priority.
+ */
+SYSCALL_DEFINE2(sched_setparam, pid_t, pid, struct sched_param __user *, param)
+{
+       return do_sched_setscheduler(pid, -1, param);
+}
+
+/**
+ * sys_sched_getscheduler - get the policy (scheduling class) of a thread
+ * @pid: the pid in question.
+ */
+SYSCALL_DEFINE1(sched_getscheduler, pid_t, pid)
+{
+       struct task_struct *p;
+       int retval;
+
+       if (pid < 0)
+               return -EINVAL;
+
+       retval = -ESRCH;
+       rcu_read_lock();
+       p = find_process_by_pid(pid);
+       if (p) {
+               retval = security_task_getscheduler(p);
+               if (!retval)
+                       retval = p->policy
+                               | (p->sched_reset_on_fork ? SCHED_RESET_ON_FORK : 0);
+       }
+       rcu_read_unlock();
+       return retval;
+}
+
+/**
+ * sys_sched_getparam - get the RT priority of a thread
+ * @pid: the pid in question.
+ * @param: structure containing the RT priority.
+ */
+SYSCALL_DEFINE2(sched_getparam, pid_t, pid, struct sched_param __user *, param)
+{
+       struct sched_param lp;
+       struct task_struct *p;
+       int retval;
+
+       if (!param || pid < 0)
+               return -EINVAL;
+
+       rcu_read_lock();
+       p = find_process_by_pid(pid);
+       retval = -ESRCH;
+       if (!p)
+               goto out_unlock;
+
+       retval = security_task_getscheduler(p);
+       if (retval)
+               goto out_unlock;
+
+       lp.sched_priority = p->rt_priority;
+       rcu_read_unlock();
+
+       /*
+        * This one might sleep, we cannot do it with a spinlock held ...
+        */
+       retval = copy_to_user(param, &lp, sizeof(*param)) ? -EFAULT : 0;
+
+       return retval;
+
+out_unlock:
+       rcu_read_unlock();
+       return retval;
+}
+
+long sched_setaffinity(pid_t pid, const struct cpumask *in_mask)
+{
+       cpumask_var_t cpus_allowed, new_mask;
+       struct task_struct *p;
+       int retval;
+
+       get_online_cpus();
+       rcu_read_lock();
+
+       p = find_process_by_pid(pid);
+       if (!p) {
+               rcu_read_unlock();
+               put_online_cpus();
+               return -ESRCH;
+       }
+
+       /* Prevent p going away */
+       get_task_struct(p);
+       rcu_read_unlock();
+
+       if (!alloc_cpumask_var(&cpus_allowed, GFP_KERNEL)) {
+               retval = -ENOMEM;
+               goto out_put_task;
+       }
+       if (!alloc_cpumask_var(&new_mask, GFP_KERNEL)) {
+               retval = -ENOMEM;
+               goto out_free_cpus_allowed;
+       }
+       retval = -EPERM;
+       if (!check_same_owner(p) && !task_ns_capable(p, CAP_SYS_NICE))
+               goto out_unlock;
+
+       retval = security_task_setscheduler(p);
+       if (retval)
+               goto out_unlock;
+
+       cpuset_cpus_allowed(p, cpus_allowed);
+       cpumask_and(new_mask, in_mask, cpus_allowed);
+again:
+       retval = set_cpus_allowed_ptr(p, new_mask);
+
+       if (!retval) {
+               cpuset_cpus_allowed(p, cpus_allowed);
+               if (!cpumask_subset(new_mask, cpus_allowed)) {
+                       /*
+                        * We must have raced with a concurrent cpuset
+                        * update. Just reset the cpus_allowed to the
+                        * cpuset's cpus_allowed
+                        */
+                       cpumask_copy(new_mask, cpus_allowed);
+                       goto again;
+               }
+       }
+out_unlock:
+       free_cpumask_var(new_mask);
+out_free_cpus_allowed:
+       free_cpumask_var(cpus_allowed);
+out_put_task:
+       put_task_struct(p);
+       put_online_cpus();
+       return retval;
+}
+
+static int get_user_cpu_mask(unsigned long __user *user_mask_ptr, unsigned len,
+                            struct cpumask *new_mask)
+{
+       if (len < cpumask_size())
+               cpumask_clear(new_mask);
+       else if (len > cpumask_size())
+               len = cpumask_size();
+
+       return copy_from_user(new_mask, user_mask_ptr, len) ? -EFAULT : 0;
+}
+
+/**
+ * sys_sched_setaffinity - set the cpu affinity of a process
+ * @pid: pid of the process
+ * @len: length in bytes of the bitmask pointed to by user_mask_ptr
+ * @user_mask_ptr: user-space pointer to the new cpu mask
+ */
+SYSCALL_DEFINE3(sched_setaffinity, pid_t, pid, unsigned int, len,
+               unsigned long __user *, user_mask_ptr)
+{
+       cpumask_var_t new_mask;
+       int retval;
+
+       if (!alloc_cpumask_var(&new_mask, GFP_KERNEL))
+               return -ENOMEM;
+
+       retval = get_user_cpu_mask(user_mask_ptr, len, new_mask);
+       if (retval == 0)
+               retval = sched_setaffinity(pid, new_mask);
+       free_cpumask_var(new_mask);
+       return retval;
+}
+
+long sched_getaffinity(pid_t pid, struct cpumask *mask)
+{
+       struct task_struct *p;
+       unsigned long flags;
+       int retval;
+
+       get_online_cpus();
+       rcu_read_lock();
+
+       retval = -ESRCH;
+       p = find_process_by_pid(pid);
+       if (!p)
+               goto out_unlock;
+
+       retval = security_task_getscheduler(p);
+       if (retval)
+               goto out_unlock;
+
+       raw_spin_lock_irqsave(&p->pi_lock, flags);
+       cpumask_and(mask, &p->cpus_allowed, cpu_online_mask);
+       raw_spin_unlock_irqrestore(&p->pi_lock, flags);
+
+out_unlock:
+       rcu_read_unlock();
+       put_online_cpus();
+
+       return retval;
+}
+
+/**
+ * sys_sched_getaffinity - get the cpu affinity of a process
+ * @pid: pid of the process
+ * @len: length in bytes of the bitmask pointed to by user_mask_ptr
+ * @user_mask_ptr: user-space pointer to hold the current cpu mask
+ */
+SYSCALL_DEFINE3(sched_getaffinity, pid_t, pid, unsigned int, len,
+               unsigned long __user *, user_mask_ptr)
+{
+       int ret;
+       cpumask_var_t mask;
+
+       if ((len * BITS_PER_BYTE) < nr_cpu_ids)
+               return -EINVAL;
+       if (len & (sizeof(unsigned long)-1))
+               return -EINVAL;
+
+       if (!alloc_cpumask_var(&mask, GFP_KERNEL))
+               return -ENOMEM;
+
+       ret = sched_getaffinity(pid, mask);
+       if (ret == 0) {
+               size_t retlen = min_t(size_t, len, cpumask_size());
+
+               if (copy_to_user(user_mask_ptr, mask, retlen))
+                       ret = -EFAULT;
+               else
+                       ret = retlen;
+       }
+       free_cpumask_var(mask);
+
+       return ret;
+}
+
+/**
+ * sys_sched_yield - yield the current processor to other threads.
+ *
+ * This function yields the current CPU to other tasks. If there are no
+ * other threads running on this CPU then this function will return.
+ */
+SYSCALL_DEFINE0(sched_yield)
+{
+       struct rq *rq = this_rq_lock();
+
+       schedstat_inc(rq, yld_count);
+       current->sched_class->yield_task(rq);
+
+       /*
+        * Since we are going to call schedule() anyway, there's
+        * no need to preempt or enable interrupts:
+        */
+       __release(rq->lock);
+       spin_release(&rq->lock.dep_map, 1, _THIS_IP_);
+       do_raw_spin_unlock(&rq->lock);
+       preempt_enable_no_resched();
+
+       schedule();
+
+       return 0;
+}
+
+static inline int should_resched(void)
+{
+       return need_resched() && !(preempt_count() & PREEMPT_ACTIVE);
+}
+
+static void __cond_resched(void)
+{
+       add_preempt_count(PREEMPT_ACTIVE);
+       __schedule();
+       sub_preempt_count(PREEMPT_ACTIVE);
+}
+
+int __sched _cond_resched(void)
+{
+       if (should_resched()) {
+               __cond_resched();
+               return 1;
+       }
+       return 0;
+}
+EXPORT_SYMBOL(_cond_resched);
+
+/*
+ * __cond_resched_lock() - if a reschedule is pending, drop the given lock,
+ * call schedule, and on return reacquire the lock.
+ *
+ * This works OK both with and without CONFIG_PREEMPT. We do strange low-level
+ * operations here to prevent schedule() from being called twice (once via
+ * spin_unlock(), once by hand).
+ */
+int __cond_resched_lock(spinlock_t *lock)
+{
+       int resched = should_resched();
+       int ret = 0;
+
+       lockdep_assert_held(lock);
+
+       if (spin_needbreak(lock) || resched) {
+               spin_unlock(lock);
+               if (resched)
+                       __cond_resched();
+               else
+                       cpu_relax();
+               ret = 1;
+               spin_lock(lock);
+       }
+       return ret;
+}
+EXPORT_SYMBOL(__cond_resched_lock);
+
+int __sched __cond_resched_softirq(void)
+{
+       BUG_ON(!in_softirq());
+
+       if (should_resched()) {
+               local_bh_enable();
+               __cond_resched();
+               local_bh_disable();
+               return 1;
+       }
+       return 0;
+}
+EXPORT_SYMBOL(__cond_resched_softirq);
+
+/**
+ * yield - yield the current processor to other threads.
+ *
+ * This is a shortcut for kernel-space yielding - it marks the
+ * thread runnable and calls sys_sched_yield().
+ */
+void __sched yield(void)
+{
+       set_current_state(TASK_RUNNING);
+       sys_sched_yield();
+}
+EXPORT_SYMBOL(yield);
+
+/**
+ * yield_to - yield the current processor to another thread in
+ * your thread group, or accelerate that thread toward the
+ * processor it's on.
+ * @p: target task
+ * @preempt: whether task preemption is allowed or not
+ *
+ * It's the caller's job to ensure that the target task struct
+ * can't go away on us before we can do any checks.
+ *
+ * Returns true if we indeed boosted the target task.
+ */
+bool __sched yield_to(struct task_struct *p, bool preempt)
+{
+       struct task_struct *curr = current;
+       struct rq *rq, *p_rq;
+       unsigned long flags;
+       bool yielded = 0;
+
+       local_irq_save(flags);
+       rq = this_rq();
+
+again:
+       p_rq = task_rq(p);
+       double_rq_lock(rq, p_rq);
+       while (task_rq(p) != p_rq) {
+               double_rq_unlock(rq, p_rq);
+               goto again;
+       }
+
+       if (!curr->sched_class->yield_to_task)
+               goto out;
+
+       if (curr->sched_class != p->sched_class)
+               goto out;
+
+       if (task_running(p_rq, p) || p->state)
+               goto out;
+
+       yielded = curr->sched_class->yield_to_task(rq, p, preempt);
+       if (yielded) {
+               schedstat_inc(rq, yld_count);
+               /*
+                * Make p's CPU reschedule; pick_next_entity takes care of
+                * fairness.
+                */
+               if (preempt && rq != p_rq)
+                       resched_task(p_rq->curr);
+       }
+
+out:
+       double_rq_unlock(rq, p_rq);
+       local_irq_restore(flags);
+
+       if (yielded)
+               schedule();
+
+       return yielded;
+}
+EXPORT_SYMBOL_GPL(yield_to);
+
+/*
+ * This task is about to go to sleep on IO. Increment rq->nr_iowait so
+ * that process accounting knows that this is a task in IO wait state.
+ */
+void __sched io_schedule(void)
+{
+       struct rq *rq = raw_rq();
+
+       delayacct_blkio_start();
+       atomic_inc(&rq->nr_iowait);
+       blk_flush_plug(current);
+       current->in_iowait = 1;
+       schedule();
+       current->in_iowait = 0;
+       atomic_dec(&rq->nr_iowait);
+       delayacct_blkio_end();
+}
+EXPORT_SYMBOL(io_schedule);
+
+long __sched io_schedule_timeout(long timeout)
+{
+       struct rq *rq = raw_rq();
+       long ret;
+
+       delayacct_blkio_start();
+       atomic_inc(&rq->nr_iowait);
+       blk_flush_plug(current);
+       current->in_iowait = 1;
+       ret = schedule_timeout(timeout);
+       current->in_iowait = 0;
+       atomic_dec(&rq->nr_iowait);
+       delayacct_blkio_end();
+       return ret;
+}
+
+/**
+ * sys_sched_get_priority_max - return maximum RT priority.
+ * @policy: scheduling class.
+ *
+ * this syscall returns the maximum rt_priority that can be used
+ * by a given scheduling class.
+ */
+SYSCALL_DEFINE1(sched_get_priority_max, int, policy)
+{
+       int ret = -EINVAL;
+
+       switch (policy) {
+       case SCHED_FIFO:
+       case SCHED_RR:
+               ret = MAX_USER_RT_PRIO-1;
+               break;
+       case SCHED_NORMAL:
+       case SCHED_BATCH:
+       case SCHED_IDLE:
+               ret = 0;
+               break;
+       }
+       return ret;
+}
+
+/**
+ * sys_sched_get_priority_min - return minimum RT priority.
+ * @policy: scheduling class.
+ *
+ * this syscall returns the minimum rt_priority that can be used
+ * by a given scheduling class.
+ */
+SYSCALL_DEFINE1(sched_get_priority_min, int, policy)
+{
+       int ret = -EINVAL;
+
+       switch (policy) {
+       case SCHED_FIFO:
+       case SCHED_RR:
+               ret = 1;
+               break;
+       case SCHED_NORMAL:
+       case SCHED_BATCH:
+       case SCHED_IDLE:
+               ret = 0;
+       }
+       return ret;
+}
+
+/**
+ * sys_sched_rr_get_interval - return the default timeslice of a process.
+ * @pid: pid of the process.
+ * @interval: userspace pointer to the timeslice value.
+ *
+ * this syscall writes the default timeslice value of a given process
+ * into the user-space timespec buffer. A value of '0' means infinity.
+ */
+SYSCALL_DEFINE2(sched_rr_get_interval, pid_t, pid,
+               struct timespec __user *, interval)
+{
+       struct task_struct *p;
+       unsigned int time_slice;
+       unsigned long flags;
+       struct rq *rq;
+       int retval;
+       struct timespec t;
+
+       if (pid < 0)
+               return -EINVAL;
+
+       retval = -ESRCH;
+       rcu_read_lock();
+       p = find_process_by_pid(pid);
+       if (!p)
+               goto out_unlock;
+
+       retval = security_task_getscheduler(p);
+       if (retval)
+               goto out_unlock;
+
+       rq = task_rq_lock(p, &flags);
+       time_slice = p->sched_class->get_rr_interval(rq, p);
+       task_rq_unlock(rq, p, &flags);
+
+       rcu_read_unlock();
+       jiffies_to_timespec(time_slice, &t);
+       retval = copy_to_user(interval, &t, sizeof(t)) ? -EFAULT : 0;
+       return retval;
+
+out_unlock:
+       rcu_read_unlock();
+       return retval;
+}
+
+static const char stat_nam[] = TASK_STATE_TO_CHAR_STR;
+
+void sched_show_task(struct task_struct *p)
+{
+       unsigned long free = 0;
+       unsigned state;
+
+       state = p->state ? __ffs(p->state) + 1 : 0;
+       printk(KERN_INFO "%-15.15s %c", p->comm,
+               state < sizeof(stat_nam) - 1 ? stat_nam[state] : '?');
+#if BITS_PER_LONG == 32
+       if (state == TASK_RUNNING)
+               printk(KERN_CONT " running  ");
+       else
+               printk(KERN_CONT " %08lx ", thread_saved_pc(p));
+#else
+       if (state == TASK_RUNNING)
+               printk(KERN_CONT "  running task    ");
+       else
+               printk(KERN_CONT " %016lx ", thread_saved_pc(p));
+#endif
+#ifdef CONFIG_DEBUG_STACK_USAGE
+       free = stack_not_used(p);
+#endif
+       printk(KERN_CONT "%5lu %5d %6d 0x%08lx\n", free,
+               task_pid_nr(p), task_pid_nr(p->real_parent),
+               (unsigned long)task_thread_info(p)->flags);
+
+       show_stack(p, NULL);
+}
+
+void show_state_filter(unsigned long state_filter)
+{
+       struct task_struct *g, *p;
+
+#if BITS_PER_LONG == 32
+       printk(KERN_INFO
+               "  task                PC stack   pid father\n");
+#else
+       printk(KERN_INFO
+               "  task                        PC stack   pid father\n");
+#endif
+       read_lock(&tasklist_lock);
+       do_each_thread(g, p) {
+               /*
+                * reset the NMI-timeout, listing all files on a slow
+                * console might take a lot of time:
+                */
+               touch_nmi_watchdog();
+               if (!state_filter || (p->state & state_filter))
+                       sched_show_task(p);
+       } while_each_thread(g, p);
+
+       touch_all_softlockup_watchdogs();
+
+#ifdef CONFIG_SCHED_DEBUG
+       sysrq_sched_debug_show();
+#endif
+       read_unlock(&tasklist_lock);
+       /*
+        * Only show locks if all tasks are dumped:
+        */
+       if (!state_filter)
+               debug_show_all_locks();
+}
+
+void __cpuinit init_idle_bootup_task(struct task_struct *idle)
+{
+       idle->sched_class = &idle_sched_class;
+}
+
+/**
+ * init_idle - set up an idle thread for a given CPU
+ * @idle: task in question
+ * @cpu: cpu the idle task belongs to
+ *
+ * NOTE: this function does not set the idle thread's NEED_RESCHED
+ * flag, to make booting more robust.
+ */
+void __cpuinit init_idle(struct task_struct *idle, int cpu)
+{
+       struct rq *rq = cpu_rq(cpu);
+       unsigned long flags;
+
+       raw_spin_lock_irqsave(&rq->lock, flags);
+
+       __sched_fork(idle);
+       idle->state = TASK_RUNNING;
+       idle->se.exec_start = sched_clock();
+
+       do_set_cpus_allowed(idle, cpumask_of(cpu));
+       /*
+        * We're having a chicken and egg problem, even though we are
+        * holding rq->lock, the cpu isn't yet set to this cpu so the
+        * lockdep check in task_group() will fail.
+        *
+        * Similar case to sched_fork(). / Alternatively we could
+        * use task_rq_lock() here and obtain the other rq->lock.
+        *
+        * Silence PROVE_RCU
+        */
+       rcu_read_lock();
+       __set_task_cpu(idle, cpu);
+       rcu_read_unlock();
+
+       rq->curr = rq->idle = idle;
+#if defined(CONFIG_SMP)
+       idle->on_cpu = 1;
+#endif
+       raw_spin_unlock_irqrestore(&rq->lock, flags);
+
+       /* Set the preempt count _outside_ the spinlocks! */
+       task_thread_info(idle)->preempt_count = 0;
+
+       /*
+        * The idle tasks have their own, simple scheduling class:
+        */
+       idle->sched_class = &idle_sched_class;
+       ftrace_graph_init_idle_task(idle, cpu);
+}
+
+/*
+ * In a system that switches off the HZ timer nohz_cpu_mask
+ * indicates which cpus entered this state. This is used
+ * in the rcu update to wait only for active cpus. For system
+ * which do not switch off the HZ timer nohz_cpu_mask should
+ * always be CPU_BITS_NONE.
+ */
+cpumask_var_t nohz_cpu_mask;
+
+/*
+ * Increase the granularity value when there are more CPUs,
+ * because with more CPUs the 'effective latency' as visible
+ * to users decreases. But the relationship is not linear,
+ * so pick a second-best guess by going with the log2 of the
+ * number of CPUs.
+ *
+ * This idea comes from the SD scheduler of Con Kolivas:
+ */
+static int get_update_sysctl_factor(void)
+{
+       unsigned int cpus = min_t(int, num_online_cpus(), 8);
+       unsigned int factor;
+
+       switch (sysctl_sched_tunable_scaling) {
+       case SCHED_TUNABLESCALING_NONE:
+               factor = 1;
+               break;
+       case SCHED_TUNABLESCALING_LINEAR:
+               factor = cpus;
+               break;
+       case SCHED_TUNABLESCALING_LOG:
+       default:
+               factor = 1 + ilog2(cpus);
+               break;
+       }
+
+       return factor;
+}
+
+static void update_sysctl(void)
+{
+       unsigned int factor = get_update_sysctl_factor();
+
+#define SET_SYSCTL(name) \
+       (sysctl_##name = (factor) * normalized_sysctl_##name)
+       SET_SYSCTL(sched_min_granularity);
+       SET_SYSCTL(sched_latency);
+       SET_SYSCTL(sched_wakeup_granularity);
+#undef SET_SYSCTL
+}
+
+static inline void sched_init_granularity(void)
+{
+       update_sysctl();
+}
+
+#ifdef CONFIG_SMP
+void do_set_cpus_allowed(struct task_struct *p, const struct cpumask *new_mask)
+{
+       if (p->sched_class && p->sched_class->set_cpus_allowed)
+               p->sched_class->set_cpus_allowed(p, new_mask);
+       else {
+               cpumask_copy(&p->cpus_allowed, new_mask);
+               p->rt.nr_cpus_allowed = cpumask_weight(new_mask);
+       }
+}
+
+/*
+ * This is how migration works:
+ *
+ * 1) we invoke migration_cpu_stop() on the target CPU using
+ *    stop_one_cpu().
+ * 2) stopper starts to run (implicitly forcing the migrated thread
+ *    off the CPU)
+ * 3) it checks whether the migrated task is still in the wrong runqueue.
+ * 4) if it's in the wrong runqueue then the migration thread removes
+ *    it and puts it into the right queue.
+ * 5) stopper completes and stop_one_cpu() returns and the migration
+ *    is done.
+ */
+
+/*
+ * Change a given task's CPU affinity. Migrate the thread to a
+ * proper CPU and schedule it away if the CPU it's executing on
+ * is removed from the allowed bitmask.
+ *
+ * NOTE: the caller must have a valid reference to the task, the
+ * task must not exit() & deallocate itself prematurely. The
+ * call is not atomic; no spinlocks may be held.
+ */
+int set_cpus_allowed_ptr(struct task_struct *p, const struct cpumask *new_mask)
+{
+       unsigned long flags;
+       struct rq *rq;
+       unsigned int dest_cpu;
+       int ret = 0;
+
+       rq = task_rq_lock(p, &flags);
+
+       if (cpumask_equal(&p->cpus_allowed, new_mask))
+               goto out;
+
+       if (!cpumask_intersects(new_mask, cpu_active_mask)) {
+               ret = -EINVAL;
+               goto out;
+       }
+
+       if (unlikely((p->flags & PF_THREAD_BOUND) && p != current)) {
+               ret = -EINVAL;
+               goto out;
+       }
+
+       do_set_cpus_allowed(p, new_mask);
+
+       /* Can the task run on the task's current CPU? If so, we're done */
+       if (cpumask_test_cpu(task_cpu(p), new_mask))
+               goto out;
+
+       dest_cpu = cpumask_any_and(cpu_active_mask, new_mask);
+       if (p->on_rq) {
+               struct migration_arg arg = { p, dest_cpu };
+               /* Need help from migration thread: drop lock and wait. */
+               task_rq_unlock(rq, p, &flags);
+               stop_one_cpu(cpu_of(rq), migration_cpu_stop, &arg);
+               tlb_migrate_finish(p->mm);
+               return 0;
+       }
+out:
+       task_rq_unlock(rq, p, &flags);
+
+       return ret;
+}
+EXPORT_SYMBOL_GPL(set_cpus_allowed_ptr);
+
+/*
+ * Move (not current) task off this cpu, onto dest cpu. We're doing
+ * this because either it can't run here any more (set_cpus_allowed()
+ * away from this CPU, or CPU going down), or because we're
+ * attempting to rebalance this task on exec (sched_exec).
+ *
+ * So we race with normal scheduler movements, but that's OK, as long
+ * as the task is no longer on this CPU.
+ *
+ * Returns non-zero if task was successfully migrated.
+ */
+static int __migrate_task(struct task_struct *p, int src_cpu, int dest_cpu)
+{
+       struct rq *rq_dest, *rq_src;
+       int ret = 0;
+
+       if (unlikely(!cpu_active(dest_cpu)))
+               return ret;
+
+       rq_src = cpu_rq(src_cpu);
+       rq_dest = cpu_rq(dest_cpu);
+
+       raw_spin_lock(&p->pi_lock);
+       double_rq_lock(rq_src, rq_dest);
+       /* Already moved. */
+       if (task_cpu(p) != src_cpu)
+               goto done;
+       /* Affinity changed (again). */
+       if (!cpumask_test_cpu(dest_cpu, &p->cpus_allowed))
+               goto fail;
+
+       /*
+        * If we're not on a rq, the next wake-up will ensure we're
+        * placed properly.
+        */
+       if (p->on_rq) {
+               deactivate_task(rq_src, p, 0);
+               set_task_cpu(p, dest_cpu);
+               activate_task(rq_dest, p, 0);
+               check_preempt_curr(rq_dest, p, 0);
+       }
+done:
+       ret = 1;
+fail:
+       double_rq_unlock(rq_src, rq_dest);
+       raw_spin_unlock(&p->pi_lock);
+       return ret;
+}
+
+/*
+ * migration_cpu_stop - this will be executed by a highprio stopper thread
+ * and performs thread migration by bumping thread off CPU then
+ * 'pushing' onto another runqueue.
+ */
+static int migration_cpu_stop(void *data)
+{
+       struct migration_arg *arg = data;
+
+       /*
+        * The original target cpu might have gone down and we might
+        * be on another cpu but it doesn't matter.
+        */
+       local_irq_disable();
+       __migrate_task(arg->task, raw_smp_processor_id(), arg->dest_cpu);
+       local_irq_enable();
+       return 0;
+}
+
+#ifdef CONFIG_HOTPLUG_CPU
+
+/*
+ * Ensures that the idle task is using init_mm right before its cpu goes
+ * offline.
+ */
+void idle_task_exit(void)
+{
+       struct mm_struct *mm = current->active_mm;
+
+       BUG_ON(cpu_online(smp_processor_id()));
+
+       if (mm != &init_mm)
+               switch_mm(mm, &init_mm, current);
+       mmdrop(mm);
+}
+
+/*
+ * While a dead CPU has no uninterruptible tasks queued at this point,
+ * it might still have a nonzero ->nr_uninterruptible counter, because
+ * for performance reasons the counter is not stricly tracking tasks to
+ * their home CPUs. So we just add the counter to another CPU's counter,
+ * to keep the global sum constant after CPU-down:
+ */
+static void migrate_nr_uninterruptible(struct rq *rq_src)
+{
+       struct rq *rq_dest = cpu_rq(cpumask_any(cpu_active_mask));
+
+       rq_dest->nr_uninterruptible += rq_src->nr_uninterruptible;
+       rq_src->nr_uninterruptible = 0;
+}
+
+/*
+ * remove the tasks which were accounted by rq from calc_load_tasks.
+ */
+static void calc_global_load_remove(struct rq *rq)
+{
+       atomic_long_sub(rq->calc_load_active, &calc_load_tasks);
+       rq->calc_load_active = 0;
+}
+
+/*
+ * Migrate all tasks from the rq, sleeping tasks will be migrated by
+ * try_to_wake_up()->select_task_rq().
+ *
+ * Called with rq->lock held even though we'er in stop_machine() and
+ * there's no concurrency possible, we hold the required locks anyway
+ * because of lock validation efforts.
+ */
+static void migrate_tasks(unsigned int dead_cpu)
+{
+       struct rq *rq = cpu_rq(dead_cpu);
+       struct task_struct *next, *stop = rq->stop;
+       int dest_cpu;
+
+       /*
+        * Fudge the rq selection such that the below task selection loop
+        * doesn't get stuck on the currently eligible stop task.
+        *
+        * We're currently inside stop_machine() and the rq is either stuck
+        * in the stop_machine_cpu_stop() loop, or we're executing this code,
+        * either way we should never end up calling schedule() until we're
+        * done here.
+        */
+       rq->stop = NULL;
+
+       for ( ; ; ) {
+               /*
+                * There's this thread running, bail when that's the only
+                * remaining thread.
+                */
+               if (rq->nr_running == 1)
+                       break;
+
+               next = pick_next_task(rq);
+               BUG_ON(!next);
+               next->sched_class->put_prev_task(rq, next);
+
+               /* Find suitable destination for @next, with force if needed. */
+               dest_cpu = select_fallback_rq(dead_cpu, next);
+               raw_spin_unlock(&rq->lock);
+
+               __migrate_task(next, dead_cpu, dest_cpu);
+
+               raw_spin_lock(&rq->lock);
+       }
+
+       rq->stop = stop;
+}
+
+#endif /* CONFIG_HOTPLUG_CPU */
+
+#if defined(CONFIG_SCHED_DEBUG) && defined(CONFIG_SYSCTL)
+
+static struct ctl_table sd_ctl_dir[] = {
+       {
+               .procname       = "sched_domain",
+               .mode           = 0555,
+       },
+       {}
+};
+
+static struct ctl_table sd_ctl_root[] = {
+       {
+               .procname       = "kernel",
+               .mode           = 0555,
+               .child          = sd_ctl_dir,
+       },
+       {}
+};
+
+static struct ctl_table *sd_alloc_ctl_entry(int n)
+{
+       struct ctl_table *entry =
+               kcalloc(n, sizeof(struct ctl_table), GFP_KERNEL);
+
+       return entry;
+}
+
+static void sd_free_ctl_entry(struct ctl_table **tablep)
+{
+       struct ctl_table *entry;
+
+       /*
+        * In the intermediate directories, both the child directory and
+        * procname are dynamically allocated and could fail but the mode
+        * will always be set. In the lowest directory the names are
+        * static strings and all have proc handlers.
+        */
+       for (entry = *tablep; entry->mode; entry++) {
+               if (entry->child)
+                       sd_free_ctl_entry(&entry->child);
+               if (entry->proc_handler == NULL)
+                       kfree(entry->procname);
+       }
+
+       kfree(*tablep);
+       *tablep = NULL;
+}
+
+static void
+set_table_entry(struct ctl_table *entry,
+               const char *procname, void *data, int maxlen,
+               mode_t mode, proc_handler *proc_handler)
+{
+       entry->procname = procname;
+       entry->data = data;
+       entry->maxlen = maxlen;
+       entry->mode = mode;
+       entry->proc_handler = proc_handler;
+}
+
+static struct ctl_table *
+sd_alloc_ctl_domain_table(struct sched_domain *sd)
+{
+       struct ctl_table *table = sd_alloc_ctl_entry(13);
+
+       if (table == NULL)
+               return NULL;
+
+       set_table_entry(&table[0], "min_interval", &sd->min_interval,
+               sizeof(long), 0644, proc_doulongvec_minmax);
+       set_table_entry(&table[1], "max_interval", &sd->max_interval,
+               sizeof(long), 0644, proc_doulongvec_minmax);
+       set_table_entry(&table[2], "busy_idx", &sd->busy_idx,
+               sizeof(int), 0644, proc_dointvec_minmax);
+       set_table_entry(&table[3], "idle_idx", &sd->idle_idx,
+               sizeof(int), 0644, proc_dointvec_minmax);
+       set_table_entry(&table[4], "newidle_idx", &sd->newidle_idx,
+               sizeof(int), 0644, proc_dointvec_minmax);
+       set_table_entry(&table[5], "wake_idx", &sd->wake_idx,
+               sizeof(int), 0644, proc_dointvec_minmax);
+       set_table_entry(&table[6], "forkexec_idx", &sd->forkexec_idx,
+               sizeof(int), 0644, proc_dointvec_minmax);
+       set_table_entry(&table[7], "busy_factor", &sd->busy_factor,
+               sizeof(int), 0644, proc_dointvec_minmax);
+       set_table_entry(&table[8], "imbalance_pct", &sd->imbalance_pct,
+               sizeof(int), 0644, proc_dointvec_minmax);
+       set_table_entry(&table[9], "cache_nice_tries",
+               &sd->cache_nice_tries,
+               sizeof(int), 0644, proc_dointvec_minmax);
+       set_table_entry(&table[10], "flags", &sd->flags,
+               sizeof(int), 0644, proc_dointvec_minmax);
+       set_table_entry(&table[11], "name", sd->name,
+               CORENAME_MAX_SIZE, 0444, proc_dostring);
+       /* &table[12] is terminator */
+
+       return table;
+}
+
+static ctl_table *sd_alloc_ctl_cpu_table(int cpu)
+{
+       struct ctl_table *entry, *table;
+       struct sched_domain *sd;
+       int domain_num = 0, i;
+       char buf[32];
+
+       for_each_domain(cpu, sd)
+               domain_num++;
+       entry = table = sd_alloc_ctl_entry(domain_num + 1);
+       if (table == NULL)
+               return NULL;
+
+       i = 0;
+       for_each_domain(cpu, sd) {
+               snprintf(buf, 32, "domain%d", i);
+               entry->procname = kstrdup(buf, GFP_KERNEL);
+               entry->mode = 0555;
+               entry->child = sd_alloc_ctl_domain_table(sd);
+               entry++;
+               i++;
+       }
+       return table;
+}
+
+static struct ctl_table_header *sd_sysctl_header;
+static void register_sched_domain_sysctl(void)
+{
+       int i, cpu_num = num_possible_cpus();
+       struct ctl_table *entry = sd_alloc_ctl_entry(cpu_num + 1);
+       char buf[32];
+
+       WARN_ON(sd_ctl_dir[0].child);
+       sd_ctl_dir[0].child = entry;
+
+       if (entry == NULL)
+               return;
+
+       for_each_possible_cpu(i) {
+               snprintf(buf, 32, "cpu%d", i);
+               entry->procname = kstrdup(buf, GFP_KERNEL);
+               entry->mode = 0555;
+               entry->child = sd_alloc_ctl_cpu_table(i);
+               entry++;
+       }
+
+       WARN_ON(sd_sysctl_header);
+       sd_sysctl_header = register_sysctl_table(sd_ctl_root);
+}
+
+/* may be called multiple times per register */
+static void unregister_sched_domain_sysctl(void)
+{
+       if (sd_sysctl_header)
+               unregister_sysctl_table(sd_sysctl_header);
+       sd_sysctl_header = NULL;
+       if (sd_ctl_dir[0].child)
+               sd_free_ctl_entry(&sd_ctl_dir[0].child);
+}
+#else
+static void register_sched_domain_sysctl(void)
+{
+}
+static void unregister_sched_domain_sysctl(void)
+{
+}
+#endif
+
+static void set_rq_online(struct rq *rq)
+{
+       if (!rq->online) {
+               const struct sched_class *class;
+
+               cpumask_set_cpu(rq->cpu, rq->rd->online);
+               rq->online = 1;
+
+               for_each_class(class) {
+                       if (class->rq_online)
+                               class->rq_online(rq);
+               }
+       }
+}
+
+static void set_rq_offline(struct rq *rq)
+{
+       if (rq->online) {
+               const struct sched_class *class;
+
+               for_each_class(class) {
+                       if (class->rq_offline)
+                               class->rq_offline(rq);
+               }
+
+               cpumask_clear_cpu(rq->cpu, rq->rd->online);
+               rq->online = 0;
+       }
+}
+
+/*
+ * migration_call - callback that gets triggered when a CPU is added.
+ * Here we can start up the necessary migration thread for the new CPU.
+ */
+static int __cpuinit
+migration_call(struct notifier_block *nfb, unsigned long action, void *hcpu)
+{
+       int cpu = (long)hcpu;
+       unsigned long flags;
+       struct rq *rq = cpu_rq(cpu);
+
+       switch (action & ~CPU_TASKS_FROZEN) {
+
+       case CPU_UP_PREPARE:
+               rq->calc_load_update = calc_load_update;
+               break;
+
+       case CPU_ONLINE:
+               /* Update our root-domain */
+               raw_spin_lock_irqsave(&rq->lock, flags);
+               if (rq->rd) {
+                       BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
+
+                       set_rq_online(rq);
+               }
+               raw_spin_unlock_irqrestore(&rq->lock, flags);
+               break;
+
+#ifdef CONFIG_HOTPLUG_CPU
+       case CPU_DYING:
+               sched_ttwu_pending();
+               /* Update our root-domain */
+               raw_spin_lock_irqsave(&rq->lock, flags);
+               if (rq->rd) {
+                       BUG_ON(!cpumask_test_cpu(cpu, rq->rd->span));
+                       set_rq_offline(rq);
+               }
+               migrate_tasks(cpu);
+               BUG_ON(rq->nr_running != 1); /* the migration thread */
+               raw_spin_unlock_irqrestore(&rq->lock, flags);
+
+               migrate_nr_uninterruptible(rq);
+               calc_global_load_remove(rq);
+               break;
+#endif
+       }
+
+       update_max_interval();
+
+       return NOTIFY_OK;
+}
+
+/*
+ * Register at high priority so that task migration (migrate_all_tasks)
+ * happens before everything else.  This has to be lower priority than
+ * the notifier in the perf_event subsystem, though.
+ */
+static struct notifier_block __cpuinitdata migration_notifier = {
+       .notifier_call = migration_call,
+       .priority = CPU_PRI_MIGRATION,
+};
+
+static int __cpuinit sched_cpu_active(struct notifier_block *nfb,
+                                     unsigned long action, void *hcpu)
+{
+       switch (action & ~CPU_TASKS_FROZEN) {
+       case CPU_ONLINE:
+       case CPU_DOWN_FAILED:
+               set_cpu_active((long)hcpu, true);
+               return NOTIFY_OK;
+       default:
+               return NOTIFY_DONE;
+       }
+}
+
+static int __cpuinit sched_cpu_inactive(struct notifier_block *nfb,
+                                       unsigned long action, void *hcpu)
+{
+       switch (action & ~CPU_TASKS_FROZEN) {
+       case CPU_DOWN_PREPARE:
+               set_cpu_active((long)hcpu, false);
+               return NOTIFY_OK;
+       default:
+               return NOTIFY_DONE;
+       }
+}
+
+static int __init migration_init(void)
+{
+       void *cpu = (void *)(long)smp_processor_id();
+       int err;
+
+       /* Initialize migration for the boot CPU */
+       err = migration_call(&migration_notifier, CPU_UP_PREPARE, cpu);
+       BUG_ON(err == NOTIFY_BAD);
+       migration_call(&migration_notifier, CPU_ONLINE, cpu);
+       register_cpu_notifier(&migration_notifier);
+
+       /* Register cpu active notifiers */
+       cpu_notifier(sched_cpu_active, CPU_PRI_SCHED_ACTIVE);
+       cpu_notifier(sched_cpu_inactive, CPU_PRI_SCHED_INACTIVE);
+
+       return 0;
+}
+early_initcall(migration_init);
+#endif
+
+#ifdef CONFIG_SMP
+
+static cpumask_var_t sched_domains_tmpmask; /* sched_domains_mutex */
+
+#ifdef CONFIG_SCHED_DEBUG
+
+static __read_mostly int sched_domain_debug_enabled;
+
+static int __init sched_domain_debug_setup(char *str)
+{
+       sched_domain_debug_enabled = 1;
+
+       return 0;
+}
+early_param("sched_debug", sched_domain_debug_setup);
+
+static int sched_domain_debug_one(struct sched_domain *sd, int cpu, int level,
+                                 struct cpumask *groupmask)
+{
+       struct sched_group *group = sd->groups;
+       char str[256];
+
+       cpulist_scnprintf(str, sizeof(str), sched_domain_span(sd));
+       cpumask_clear(groupmask);
+
+       printk(KERN_DEBUG "%*s domain %d: ", level, "", level);
+
+       if (!(sd->flags & SD_LOAD_BALANCE)) {
+               printk("does not load-balance\n");
+               if (sd->parent)
+                       printk(KERN_ERR "ERROR: !SD_LOAD_BALANCE domain"
+                                       " has parent");
+               return -1;
+       }
+
+       printk(KERN_CONT "span %s level %s\n", str, sd->name);
+
+       if (!cpumask_test_cpu(cpu, sched_domain_span(sd))) {
+               printk(KERN_ERR "ERROR: domain->span does not contain "
+                               "CPU%d\n", cpu);
+       }
+       if (!cpumask_test_cpu(cpu, sched_group_cpus(group))) {
+               printk(KERN_ERR "ERROR: domain->groups does not contain"
+                               " CPU%d\n", cpu);
+       }
+
+       printk(KERN_DEBUG "%*s groups:", level + 1, "");
+       do {
+               if (!group) {
+                       printk("\n");
+                       printk(KERN_ERR "ERROR: group is NULL\n");
+                       break;
+               }
+
+               if (!group->sgp->power) {
+                       printk(KERN_CONT "\n");
+                       printk(KERN_ERR "ERROR: domain->cpu_power not "
+                                       "set\n");
+                       break;
+               }
+
+               if (!cpumask_weight(sched_group_cpus(group))) {
+                       printk(KERN_CONT "\n");
+                       printk(KERN_ERR "ERROR: empty group\n");
+                       break;
+               }
+
+               if (cpumask_intersects(groupmask, sched_group_cpus(group))) {
+                       printk(KERN_CONT "\n");
+                       printk(KERN_ERR "ERROR: repeated CPUs\n");
+                       break;
+               }
+
+               cpumask_or(groupmask, groupmask, sched_group_cpus(group));
+
+               cpulist_scnprintf(str, sizeof(str), sched_group_cpus(group));
+
+               printk(KERN_CONT " %s", str);
+               if (group->sgp->power != SCHED_POWER_SCALE) {
+                       printk(KERN_CONT " (cpu_power = %d)",
+                               group->sgp->power);
+               }
+
+               group = group->next;
+       } while (group != sd->groups);
+       printk(KERN_CONT "\n");
+
+       if (!cpumask_equal(sched_domain_span(sd), groupmask))
+               printk(KERN_ERR "ERROR: groups don't span domain->span\n");
+
+       if (sd->parent &&
+           !cpumask_subset(groupmask, sched_domain_span(sd->parent)))
+               printk(KERN_ERR "ERROR: parent span is not a superset "
+                       "of domain->span\n");
+       return 0;
+}
+
+static void sched_domain_debug(struct sched_domain *sd, int cpu)
+{
+       int level = 0;
+
+       if (!sched_domain_debug_enabled)
+               return;
+
+       if (!sd) {
+               printk(KERN_DEBUG "CPU%d attaching NULL sched-domain.\n", cpu);
+               return;
+       }
+
+       printk(KERN_DEBUG "CPU%d attaching sched-domain:\n", cpu);
+
+       for (;;) {
+               if (sched_domain_debug_one(sd, cpu, level, sched_domains_tmpmask))
+                       break;
+               level++;
+               sd = sd->parent;
+               if (!sd)
+                       break;
+       }
+}
+#else /* !CONFIG_SCHED_DEBUG */
+# define sched_domain_debug(sd, cpu) do { } while (0)
+#endif /* CONFIG_SCHED_DEBUG */
+
+static int sd_degenerate(struct sched_domain *sd)
+{
+       if (cpumask_weight(sched_domain_span(sd)) == 1)
+               return 1;
+
+       /* Following flags need at least 2 groups */
+       if (sd->flags & (SD_LOAD_BALANCE |
+                        SD_BALANCE_NEWIDLE |
+                        SD_BALANCE_FORK |
+                        SD_BALANCE_EXEC |
+                        SD_SHARE_CPUPOWER |
+                        SD_SHARE_PKG_RESOURCES)) {
+               if (sd->groups != sd->groups->next)
+                       return 0;
+       }
+
+       /* Following flags don't use groups */
+       if (sd->flags & (SD_WAKE_AFFINE))
+               return 0;
+
+       return 1;
+}
+
+static int
+sd_parent_degenerate(struct sched_domain *sd, struct sched_domain *parent)
+{
+       unsigned long cflags = sd->flags, pflags = parent->flags;
+
+       if (sd_degenerate(parent))
+               return 1;
+
+       if (!cpumask_equal(sched_domain_span(sd), sched_domain_span(parent)))
+               return 0;
+
+       /* Flags needing groups don't count if only 1 group in parent */
+       if (parent->groups == parent->groups->next) {
+               pflags &= ~(SD_LOAD_BALANCE |
+                               SD_BALANCE_NEWIDLE |
+                               SD_BALANCE_FORK |
+                               SD_BALANCE_EXEC |
+                               SD_SHARE_CPUPOWER |
+                               SD_SHARE_PKG_RESOURCES);
+               if (nr_node_ids == 1)
+                       pflags &= ~SD_SERIALIZE;
+       }
+       if (~cflags & pflags)
+               return 0;
+
+       return 1;
+}
+
+static void free_rootdomain(struct rcu_head *rcu)
+{
+       struct root_domain *rd = container_of(rcu, struct root_domain, rcu);
+
+       cpupri_cleanup(&rd->cpupri);
+       free_cpumask_var(rd->rto_mask);
+       free_cpumask_var(rd->online);
+       free_cpumask_var(rd->span);
+       kfree(rd);
+}
+
+static void rq_attach_root(struct rq *rq, struct root_domain *rd)
+{
+       struct root_domain *old_rd = NULL;
+       unsigned long flags;
+
+       raw_spin_lock_irqsave(&rq->lock, flags);
+
+       if (rq->rd) {
+               old_rd = rq->rd;
+
+               if (cpumask_test_cpu(rq->cpu, old_rd->online))
+                       set_rq_offline(rq);
+
+               cpumask_clear_cpu(rq->cpu, old_rd->span);
+
+               /*
+                * If we dont want to free the old_rt yet then
+                * set old_rd to NULL to skip the freeing later
+                * in this function:
+                */
+               if (!atomic_dec_and_test(&old_rd->refcount))
+                       old_rd = NULL;
+       }
+
+       atomic_inc(&rd->refcount);
+       rq->rd = rd;
+
+       cpumask_set_cpu(rq->cpu, rd->span);
+       if (cpumask_test_cpu(rq->cpu, cpu_active_mask))
+               set_rq_online(rq);
+
+       raw_spin_unlock_irqrestore(&rq->lock, flags);
+
+       if (old_rd)
+               call_rcu_sched(&old_rd->rcu, free_rootdomain);
+}
+
+static int init_rootdomain(struct root_domain *rd)
+{
+       memset(rd, 0, sizeof(*rd));
+
+       if (!alloc_cpumask_var(&rd->span, GFP_KERNEL))
+               goto out;
+       if (!alloc_cpumask_var(&rd->online, GFP_KERNEL))
+               goto free_span;
+       if (!alloc_cpumask_var(&rd->rto_mask, GFP_KERNEL))
+               goto free_online;
+
+       if (cpupri_init(&rd->cpupri) != 0)
+               goto free_rto_mask;
+       return 0;
+
+free_rto_mask:
+       free_cpumask_var(rd->rto_mask);
+free_online:
+       free_cpumask_var(rd->online);
+free_span:
+       free_cpumask_var(rd->span);
+out:
+       return -ENOMEM;
+}
+
+static void init_defrootdomain(void)
+{
+       init_rootdomain(&def_root_domain);
+
+       atomic_set(&def_root_domain.refcount, 1);
+}
+
+static struct root_domain *alloc_rootdomain(void)
+{
+       struct root_domain *rd;
+
+       rd = kmalloc(sizeof(*rd), GFP_KERNEL);
+       if (!rd)
+               return NULL;
+
+       if (init_rootdomain(rd) != 0) {
+               kfree(rd);
+               return NULL;
+       }
+
+       return rd;
+}
+
+static void free_sched_groups(struct sched_group *sg, int free_sgp)
+{
+       struct sched_group *tmp, *first;
+
+       if (!sg)
+               return;
+
+       first = sg;
+       do {
+               tmp = sg->next;
+
+               if (free_sgp && atomic_dec_and_test(&sg->sgp->ref))
+                       kfree(sg->sgp);
+
+               kfree(sg);
+               sg = tmp;
+       } while (sg != first);
+}
+
+static void free_sched_domain(struct rcu_head *rcu)
+{
+       struct sched_domain *sd = container_of(rcu, struct sched_domain, rcu);
+
+       /*
+        * If its an overlapping domain it has private groups, iterate and
+        * nuke them all.
+        */
+       if (sd->flags & SD_OVERLAP) {
+               free_sched_groups(sd->groups, 1);
+       } else if (atomic_dec_and_test(&sd->groups->ref)) {
+               kfree(sd->groups->sgp);
+               kfree(sd->groups);
+       }
+       kfree(sd);
+}
+
+static void destroy_sched_domain(struct sched_domain *sd, int cpu)
+{
+       call_rcu(&sd->rcu, free_sched_domain);
+}
+
+static void destroy_sched_domains(struct sched_domain *sd, int cpu)
+{
+       for (; sd; sd = sd->parent)
+               destroy_sched_domain(sd, cpu);
+}
+
+/*
+ * Attach the domain 'sd' to 'cpu' as its base domain. Callers must
+ * hold the hotplug lock.
+ */
+static void
+cpu_attach_domain(struct sched_domain *sd, struct root_domain *rd, int cpu)
+{
+       struct rq *rq = cpu_rq(cpu);
+       struct sched_domain *tmp;
+
+       /* Remove the sched domains which do not contribute to scheduling. */
+       for (tmp = sd; tmp; ) {
+               struct sched_domain *parent = tmp->parent;
+               if (!parent)
+                       break;
+
+               if (sd_parent_degenerate(tmp, parent)) {
+                       tmp->parent = parent->parent;
+                       if (parent->parent)
+                               parent->parent->child = tmp;
+                       destroy_sched_domain(parent, cpu);
+               } else
+                       tmp = tmp->parent;
+       }
+
+       if (sd && sd_degenerate(sd)) {
+               tmp = sd;
+               sd = sd->parent;
+               destroy_sched_domain(tmp, cpu);
+               if (sd)
+                       sd->child = NULL;
+       }
+
+       sched_domain_debug(sd, cpu);
+
+       rq_attach_root(rq, rd);
+       tmp = rq->sd;
+       rcu_assign_pointer(rq->sd, sd);
+       destroy_sched_domains(tmp, cpu);
+}
+
+/* cpus with isolated domains */
+static cpumask_var_t cpu_isolated_map;
+
+/* Setup the mask of cpus configured for isolated domains */
+static int __init isolated_cpu_setup(char *str)
+{
+       alloc_bootmem_cpumask_var(&cpu_isolated_map);
+       cpulist_parse(str, cpu_isolated_map);
+       return 1;
+}
+
+__setup("isolcpus=", isolated_cpu_setup);
+
+#define SD_NODES_PER_DOMAIN 16
+
+#ifdef CONFIG_NUMA
+
+/**
+ * find_next_best_node - find the next node to include in a sched_domain
+ * @node: node whose sched_domain we're building
+ * @used_nodes: nodes already in the sched_domain
+ *
+ * Find the next node to include in a given scheduling domain. Simply
+ * finds the closest node not already in the @used_nodes map.
+ *
+ * Should use nodemask_t.
+ */
+static int find_next_best_node(int node, nodemask_t *used_nodes)
+{
+       int i, n, val, min_val, best_node = -1;
+
+       min_val = INT_MAX;
+
+       for (i = 0; i < nr_node_ids; i++) {
+               /* Start at @node */
+               n = (node + i) % nr_node_ids;
+
+               if (!nr_cpus_node(n))
+                       continue;
+
+               /* Skip already used nodes */
+               if (node_isset(n, *used_nodes))
+                       continue;
+
+               /* Simple min distance search */
+               val = node_distance(node, n);
+
+               if (val < min_val) {
+                       min_val = val;
+                       best_node = n;
+               }
+       }
+
+       if (best_node != -1)
+               node_set(best_node, *used_nodes);
+       return best_node;
+}
+
+/**
+ * sched_domain_node_span - get a cpumask for a node's sched_domain
+ * @node: node whose cpumask we're constructing
+ * @span: resulting cpumask
+ *
+ * Given a node, construct a good cpumask for its sched_domain to span. It
+ * should be one that prevents unnecessary balancing, but also spreads tasks
+ * out optimally.
+ */
+static void sched_domain_node_span(int node, struct cpumask *span)
+{
+       nodemask_t used_nodes;
+       int i;
+
+       cpumask_clear(span);
+       nodes_clear(used_nodes);
+
+       cpumask_or(span, span, cpumask_of_node(node));
+       node_set(node, used_nodes);
+
+       for (i = 1; i < SD_NODES_PER_DOMAIN; i++) {
+               int next_node = find_next_best_node(node, &used_nodes);
+               if (next_node < 0)
+                       break;
+               cpumask_or(span, span, cpumask_of_node(next_node));
+       }
+}
+
+static const struct cpumask *cpu_node_mask(int cpu)
+{
+       lockdep_assert_held(&sched_domains_mutex);
+
+       sched_domain_node_span(cpu_to_node(cpu), sched_domains_tmpmask);
+
+       return sched_domains_tmpmask;
+}
+
+static const struct cpumask *cpu_allnodes_mask(int cpu)
+{
+       return cpu_possible_mask;
+}
+#endif /* CONFIG_NUMA */
+
+static const struct cpumask *cpu_cpu_mask(int cpu)
+{
+       return cpumask_of_node(cpu_to_node(cpu));
+}
+
+int sched_smt_power_savings = 0, sched_mc_power_savings = 0;
+
+struct sd_data {
+       struct sched_domain **__percpu sd;
+       struct sched_group **__percpu sg;
+       struct sched_group_power **__percpu sgp;
+};
+
+struct s_data {
+       struct sched_domain ** __percpu sd;
+       struct root_domain      *rd;
+};
+
+enum s_alloc {
+       sa_rootdomain,
+       sa_sd,
+       sa_sd_storage,
+       sa_none,
+};
+
+struct sched_domain_topology_level;
+
+typedef struct sched_domain *(*sched_domain_init_f)(struct sched_domain_topology_level *tl, int cpu);
+typedef const struct cpumask *(*sched_domain_mask_f)(int cpu);
+
+#define SDTL_OVERLAP   0x01
+
+struct sched_domain_topology_level {
+       sched_domain_init_f init;
+       sched_domain_mask_f mask;
+       int                 flags;
+       struct sd_data      data;
+};
+
+static int
+build_overlap_sched_groups(struct sched_domain *sd, int cpu)
+{
+       struct sched_group *first = NULL, *last = NULL, *groups = NULL, *sg;
+       const struct cpumask *span = sched_domain_span(sd);
+       struct cpumask *covered = sched_domains_tmpmask;
+       struct sd_data *sdd = sd->private;
+       struct sched_domain *child;
+       int i;
+
+       cpumask_clear(covered);
+
+       for_each_cpu(i, span) {
+               struct cpumask *sg_span;
+
+               if (cpumask_test_cpu(i, covered))
+                       continue;
+
+               sg = kzalloc_node(sizeof(struct sched_group) + cpumask_size(),
+                               GFP_KERNEL, cpu_to_node(i));
+
+               if (!sg)
+                       goto fail;
+
+               sg_span = sched_group_cpus(sg);
+
+               child = *per_cpu_ptr(sdd->sd, i);
+               if (child->child) {
+                       child = child->child;
+                       cpumask_copy(sg_span, sched_domain_span(child));
+               } else
+                       cpumask_set_cpu(i, sg_span);
+
+               cpumask_or(covered, covered, sg_span);
+
+               sg->sgp = *per_cpu_ptr(sdd->sgp, cpumask_first(sg_span));
+               atomic_inc(&sg->sgp->ref);
+
+               if (cpumask_test_cpu(cpu, sg_span))
+                       groups = sg;
+
+               if (!first)
+                       first = sg;
+               if (last)
+                       last->next = sg;
+               last = sg;
+               last->next = first;
+       }
+       sd->groups = groups;
+
+       return 0;
+
+fail:
+       free_sched_groups(first, 0);
+
+       return -ENOMEM;
+}
+
+static int get_group(int cpu, struct sd_data *sdd, struct sched_group **sg)
+{
+       struct sched_domain *sd = *per_cpu_ptr(sdd->sd, cpu);
+       struct sched_domain *child = sd->child;
+
+       if (child)
+               cpu = cpumask_first(sched_domain_span(child));
+
+       if (sg) {
+               *sg = *per_cpu_ptr(sdd->sg, cpu);
+               (*sg)->sgp = *per_cpu_ptr(sdd->sgp, cpu);
+               atomic_set(&(*sg)->sgp->ref, 1); /* for claim_allocations */
+       }
+
+       return cpu;
+}
+
+/*
+ * build_sched_groups will build a circular linked list of the groups
+ * covered by the given span, and will set each group's ->cpumask correctly,
+ * and ->cpu_power to 0.
+ *
+ * Assumes the sched_domain tree is fully constructed
+ */
+static int
+build_sched_groups(struct sched_domain *sd, int cpu)
+{
+       struct sched_group *first = NULL, *last = NULL;
+       struct sd_data *sdd = sd->private;
+       const struct cpumask *span = sched_domain_span(sd);
+       struct cpumask *covered;
+       int i;
+
+       get_group(cpu, sdd, &sd->groups);
+       atomic_inc(&sd->groups->ref);
+
+       if (cpu != cpumask_first(sched_domain_span(sd)))
+               return 0;
+
+       lockdep_assert_held(&sched_domains_mutex);
+       covered = sched_domains_tmpmask;
+
+       cpumask_clear(covered);
+
+       for_each_cpu(i, span) {
+               struct sched_group *sg;
+               int group = get_group(i, sdd, &sg);
+               int j;
+
+               if (cpumask_test_cpu(i, covered))
+                       continue;
+
+               cpumask_clear(sched_group_cpus(sg));
+               sg->sgp->power = 0;
+
+               for_each_cpu(j, span) {
+                       if (get_group(j, sdd, NULL) != group)
+                               continue;
+
+                       cpumask_set_cpu(j, covered);
+                       cpumask_set_cpu(j, sched_group_cpus(sg));
+               }
+
+               if (!first)
+                       first = sg;
+               if (last)
+                       last->next = sg;
+               last = sg;
+       }
+       last->next = first;
+
+       return 0;
+}
+
+/*
+ * Initialize sched groups cpu_power.
+ *
+ * cpu_power indicates the capacity of sched group, which is used while
+ * distributing the load between different sched groups in a sched domain.
+ * Typically cpu_power for all the groups in a sched domain will be same unless
+ * there are asymmetries in the topology. If there are asymmetries, group
+ * having more cpu_power will pickup more load compared to the group having
+ * less cpu_power.
+ */
+static void init_sched_groups_power(int cpu, struct sched_domain *sd)
+{
+       struct sched_group *sg = sd->groups;
+
+       WARN_ON(!sd || !sg);
+
+       do {
+               sg->group_weight = cpumask_weight(sched_group_cpus(sg));
+               sg = sg->next;
+       } while (sg != sd->groups);
+
+       if (cpu != group_first_cpu(sg))
+               return;
+
+       update_group_power(sd, cpu);
+}
+
+/*
+ * Initializers for schedule domains
+ * Non-inlined to reduce accumulated stack pressure in build_sched_domains()
+ */
+
+#ifdef CONFIG_SCHED_DEBUG
+# define SD_INIT_NAME(sd, type)                sd->name = #type
+#else
+# define SD_INIT_NAME(sd, type)                do { } while (0)
+#endif
+
+#define SD_INIT_FUNC(type)                                             \
+static noinline struct sched_domain *                                  \
+sd_init_##type(struct sched_domain_topology_level *tl, int cpu)        \
+{                                                                      \
+       struct sched_domain *sd = *per_cpu_ptr(tl->data.sd, cpu);       \
+       *sd = SD_##type##_INIT;                                         \
+       SD_INIT_NAME(sd, type);                                         \
+       sd->private = &tl->data;                                        \
+       return sd;                                                      \
+}
+
+SD_INIT_FUNC(CPU)
+#ifdef CONFIG_NUMA
+ SD_INIT_FUNC(ALLNODES)
+ SD_INIT_FUNC(NODE)
+#endif
+#ifdef CONFIG_SCHED_SMT
+ SD_INIT_FUNC(SIBLING)
+#endif
+#ifdef CONFIG_SCHED_MC
+ SD_INIT_FUNC(MC)
+#endif
+#ifdef CONFIG_SCHED_BOOK
+ SD_INIT_FUNC(BOOK)
+#endif
+
+static int default_relax_domain_level = -1;
+int sched_domain_level_max;
+
+static int __init setup_relax_domain_level(char *str)
+{
+       unsigned long val;
+
+       val = simple_strtoul(str, NULL, 0);
+       if (val < sched_domain_level_max)
+               default_relax_domain_level = val;
+
+       return 1;
+}
+__setup("relax_domain_level=", setup_relax_domain_level);
+
+static void set_domain_attribute(struct sched_domain *sd,
+                                struct sched_domain_attr *attr)
+{
+       int request;
+
+       if (!attr || attr->relax_domain_level < 0) {
+               if (default_relax_domain_level < 0)
+                       return;
+               else
+                       request = default_relax_domain_level;
+       } else
+               request = attr->relax_domain_level;
+       if (request < sd->level) {
+               /* turn off idle balance on this domain */
+               sd->flags &= ~(SD_BALANCE_WAKE|SD_BALANCE_NEWIDLE);
+       } else {
+               /* turn on idle balance on this domain */
+               sd->flags |= (SD_BALANCE_WAKE|SD_BALANCE_NEWIDLE);
+       }
+}
+
+static void __sdt_free(const struct cpumask *cpu_map);
+static int __sdt_alloc(const struct cpumask *cpu_map);
+
+static void __free_domain_allocs(struct s_data *d, enum s_alloc what,
+                                const struct cpumask *cpu_map)
+{
+       switch (what) {
+       case sa_rootdomain:
+               if (!atomic_read(&d->rd->refcount))
+                       free_rootdomain(&d->rd->rcu); /* fall through */
+       case sa_sd:
+               free_percpu(d->sd); /* fall through */
+       case sa_sd_storage:
+               __sdt_free(cpu_map); /* fall through */
+       case sa_none:
+               break;
+       }
+}
+
+static enum s_alloc __visit_domain_allocation_hell(struct s_data *d,
+                                                  const struct cpumask *cpu_map)
+{
+       memset(d, 0, sizeof(*d));
+
+       if (__sdt_alloc(cpu_map))
+               return sa_sd_storage;
+       d->sd = alloc_percpu(struct sched_domain *);
+       if (!d->sd)
+               return sa_sd_storage;
+       d->rd = alloc_rootdomain();
+       if (!d->rd)
+               return sa_sd;
+       return sa_rootdomain;
+}
+
+/*
+ * NULL the sd_data elements we've used to build the sched_domain and
+ * sched_group structure so that the subsequent __free_domain_allocs()
+ * will not free the data we're using.
+ */
+static void claim_allocations(int cpu, struct sched_domain *sd)
+{
+       struct sd_data *sdd = sd->private;
+
+       WARN_ON_ONCE(*per_cpu_ptr(sdd->sd, cpu) != sd);
+       *per_cpu_ptr(sdd->sd, cpu) = NULL;
+
+       if (atomic_read(&(*per_cpu_ptr(sdd->sg, cpu))->ref))
+               *per_cpu_ptr(sdd->sg, cpu) = NULL;
+
+       if (atomic_read(&(*per_cpu_ptr(sdd->sgp, cpu))->ref))
+               *per_cpu_ptr(sdd->sgp, cpu) = NULL;
+}
+
+#ifdef CONFIG_SCHED_SMT
+static const struct cpumask *cpu_smt_mask(int cpu)
+{
+       return topology_thread_cpumask(cpu);
+}
+#endif
+
+/*
+ * Topology list, bottom-up.
+ */
+static struct sched_domain_topology_level default_topology[] = {
+#ifdef CONFIG_SCHED_SMT
+       { sd_init_SIBLING, cpu_smt_mask, },
+#endif
+#ifdef CONFIG_SCHED_MC
+       { sd_init_MC, cpu_coregroup_mask, },
+#endif
+#ifdef CONFIG_SCHED_BOOK
+       { sd_init_BOOK, cpu_book_mask, },
+#endif
+       { sd_init_CPU, cpu_cpu_mask, },
+#ifdef CONFIG_NUMA
+       { sd_init_NODE, cpu_node_mask, SDTL_OVERLAP, },
+       { sd_init_ALLNODES, cpu_allnodes_mask, },
+#endif
+       { NULL, },
+};
+
+static struct sched_domain_topology_level *sched_domain_topology = default_topology;
+
+static int __sdt_alloc(const struct cpumask *cpu_map)
+{
+       struct sched_domain_topology_level *tl;
+       int j;
+
+       for (tl = sched_domain_topology; tl->init; tl++) {
+               struct sd_data *sdd = &tl->data;
+
+               sdd->sd = alloc_percpu(struct sched_domain *);
+               if (!sdd->sd)
+                       return -ENOMEM;
+
+               sdd->sg = alloc_percpu(struct sched_group *);
+               if (!sdd->sg)
+                       return -ENOMEM;
+
+               sdd->sgp = alloc_percpu(struct sched_group_power *);
+               if (!sdd->sgp)
+                       return -ENOMEM;
+
+               for_each_cpu(j, cpu_map) {
+                       struct sched_domain *sd;
+                       struct sched_group *sg;
+                       struct sched_group_power *sgp;
+
+                       sd = kzalloc_node(sizeof(struct sched_domain) + cpumask_size(),
+                                       GFP_KERNEL, cpu_to_node(j));
+                       if (!sd)
+                               return -ENOMEM;
+
+                       *per_cpu_ptr(sdd->sd, j) = sd;
+
+                       sg = kzalloc_node(sizeof(struct sched_group) + cpumask_size(),
+                                       GFP_KERNEL, cpu_to_node(j));
+                       if (!sg)
+                               return -ENOMEM;
+
+                       *per_cpu_ptr(sdd->sg, j) = sg;
+
+                       sgp = kzalloc_node(sizeof(struct sched_group_power),
+                                       GFP_KERNEL, cpu_to_node(j));
+                       if (!sgp)
+                               return -ENOMEM;
+
+                       *per_cpu_ptr(sdd->sgp, j) = sgp;
+               }
+       }
+
+       return 0;
+}
+
+static void __sdt_free(const struct cpumask *cpu_map)
+{
+       struct sched_domain_topology_level *tl;
+       int j;
+
+       for (tl = sched_domain_topology; tl->init; tl++) {
+               struct sd_data *sdd = &tl->data;
+
+               for_each_cpu(j, cpu_map) {
+                       struct sched_domain *sd = *per_cpu_ptr(sdd->sd, j);
+                       if (sd && (sd->flags & SD_OVERLAP))
+                               free_sched_groups(sd->groups, 0);
+                       kfree(*per_cpu_ptr(sdd->sd, j));
+                       kfree(*per_cpu_ptr(sdd->sg, j));
+                       kfree(*per_cpu_ptr(sdd->sgp, j));
+               }
+               free_percpu(sdd->sd);
+               free_percpu(sdd->sg);
+               free_percpu(sdd->sgp);
+       }
+}
+
+struct sched_domain *build_sched_domain(struct sched_domain_topology_level *tl,
+               struct s_data *d, const struct cpumask *cpu_map,
+               struct sched_domain_attr *attr, struct sched_domain *child,
+               int cpu)
+{
+       struct sched_domain *sd = tl->init(tl, cpu);
+       if (!sd)
+               return child;
+
+       set_domain_attribute(sd, attr);
+       cpumask_and(sched_domain_span(sd), cpu_map, tl->mask(cpu));
+       if (child) {
+               sd->level = child->level + 1;
+               sched_domain_level_max = max(sched_domain_level_max, sd->level);
+               child->parent = sd;
+       }
+       sd->child = child;
+
+       return sd;
+}
+
+/*
+ * Build sched domains for a given set of cpus and attach the sched domains
+ * to the individual cpus
+ */
+static int build_sched_domains(const struct cpumask *cpu_map,
+                              struct sched_domain_attr *attr)
+{
+       enum s_alloc alloc_state = sa_none;
+       struct sched_domain *sd;
+       struct s_data d;
+       int i, ret = -ENOMEM;
+
+       alloc_state = __visit_domain_allocation_hell(&d, cpu_map);
+       if (alloc_state != sa_rootdomain)
+               goto error;
+
+       /* Set up domains for cpus specified by the cpu_map. */
+       for_each_cpu(i, cpu_map) {
+               struct sched_domain_topology_level *tl;
+
+               sd = NULL;
+               for (tl = sched_domain_topology; tl->init; tl++) {
+                       sd = build_sched_domain(tl, &d, cpu_map, attr, sd, i);
+                       if (tl->flags & SDTL_OVERLAP || sched_feat(FORCE_SD_OVERLAP))
+                               sd->flags |= SD_OVERLAP;
+                       if (cpumask_equal(cpu_map, sched_domain_span(sd)))
+                               break;
+               }
+
+               while (sd->child)
+                       sd = sd->child;
+
+               *per_cpu_ptr(d.sd, i) = sd;
+       }
+
+       /* Build the groups for the domains */
+       for_each_cpu(i, cpu_map) {
+               for (sd = *per_cpu_ptr(d.sd, i); sd; sd = sd->parent) {
+                       sd->span_weight = cpumask_weight(sched_domain_span(sd));
+                       if (sd->flags & SD_OVERLAP) {
+                               if (build_overlap_sched_groups(sd, i))
+                                       goto error;
+                       } else {
+                               if (build_sched_groups(sd, i))
+                                       goto error;
+                       }
+               }
+       }
+
+       /* Calculate CPU power for physical packages and nodes */
+       for (i = nr_cpumask_bits-1; i >= 0; i--) {
+               if (!cpumask_test_cpu(i, cpu_map))
+                       continue;
+
+               for (sd = *per_cpu_ptr(d.sd, i); sd; sd = sd->parent) {
+                       claim_allocations(i, sd);
+                       init_sched_groups_power(i, sd);
+               }
+       }
+
+       /* Attach the domains */
+       rcu_read_lock();
+       for_each_cpu(i, cpu_map) {
+               sd = *per_cpu_ptr(d.sd, i);
+               cpu_attach_domain(sd, d.rd, i);
+       }
+       rcu_read_unlock();
+
+       ret = 0;
+error:
+       __free_domain_allocs(&d, alloc_state, cpu_map);
+       return ret;
+}
+
+static cpumask_var_t *doms_cur;        /* current sched domains */
+static int ndoms_cur;          /* number of sched domains in 'doms_cur' */
+static struct sched_domain_attr *dattr_cur;
+                               /* attribues of custom domains in 'doms_cur' */
+
+/*
+ * Special case: If a kmalloc of a doms_cur partition (array of
+ * cpumask) fails, then fallback to a single sched domain,
+ * as determined by the single cpumask fallback_doms.
+ */
+static cpumask_var_t fallback_doms;
+
+/*
+ * arch_update_cpu_topology lets virtualized architectures update the
+ * cpu core maps. It is supposed to return 1 if the topology changed
+ * or 0 if it stayed the same.
+ */
+int __attribute__((weak)) arch_update_cpu_topology(void)
+{
+       return 0;
+}
+
+cpumask_var_t *alloc_sched_domains(unsigned int ndoms)
+{
+       int i;
+       cpumask_var_t *doms;
+
+       doms = kmalloc(sizeof(*doms) * ndoms, GFP_KERNEL);
+       if (!doms)
+               return NULL;
+       for (i = 0; i < ndoms; i++) {
+               if (!alloc_cpumask_var(&doms[i], GFP_KERNEL)) {
+                       free_sched_domains(doms, i);
+                       return NULL;
+               }
+       }
+       return doms;
+}
+
+void free_sched_domains(cpumask_var_t doms[], unsigned int ndoms)
+{
+       unsigned int i;
+       for (i = 0; i < ndoms; i++)
+               free_cpumask_var(doms[i]);
+       kfree(doms);
+}
+
+/*
+ * Set up scheduler domains and groups. Callers must hold the hotplug lock.
+ * For now this just excludes isolated cpus, but could be used to
+ * exclude other special cases in the future.
+ */
+static int init_sched_domains(const struct cpumask *cpu_map)
+{
+       int err;
+
+       arch_update_cpu_topology();
+       ndoms_cur = 1;
+       doms_cur = alloc_sched_domains(ndoms_cur);
+       if (!doms_cur)
+               doms_cur = &fallback_doms;
+       cpumask_andnot(doms_cur[0], cpu_map, cpu_isolated_map);
+       dattr_cur = NULL;
+       err = build_sched_domains(doms_cur[0], NULL);
+       register_sched_domain_sysctl();
+
+       return err;
+}
+
+/*
+ * Detach sched domains from a group of cpus specified in cpu_map
+ * These cpus will now be attached to the NULL domain
+ */
+static void detach_destroy_domains(const struct cpumask *cpu_map)
+{
+       int i;
+
+       rcu_read_lock();
+       for_each_cpu(i, cpu_map)
+               cpu_attach_domain(NULL, &def_root_domain, i);
+       rcu_read_unlock();
+}
+
+/* handle null as "default" */
+static int dattrs_equal(struct sched_domain_attr *cur, int idx_cur,
+                       struct sched_domain_attr *new, int idx_new)
+{
+       struct sched_domain_attr tmp;
+
+       /* fast path */
+       if (!new && !cur)
+               return 1;
+
+       tmp = SD_ATTR_INIT;
+       return !memcmp(cur ? (cur + idx_cur) : &tmp,
+                       new ? (new + idx_new) : &tmp,
+                       sizeof(struct sched_domain_attr));
+}
+
+/*
+ * Partition sched domains as specified by the 'ndoms_new'
+ * cpumasks in the array doms_new[] of cpumasks. This compares
+ * doms_new[] to the current sched domain partitioning, doms_cur[].
+ * It destroys each deleted domain and builds each new domain.
+ *
+ * 'doms_new' is an array of cpumask_var_t's of length 'ndoms_new'.
+ * The masks don't intersect (don't overlap.) We should setup one
+ * sched domain for each mask. CPUs not in any of the cpumasks will
+ * not be load balanced. If the same cpumask appears both in the
+ * current 'doms_cur' domains and in the new 'doms_new', we can leave
+ * it as it is.
+ *
+ * The passed in 'doms_new' should be allocated using
+ * alloc_sched_domains.  This routine takes ownership of it and will
+ * free_sched_domains it when done with it. If the caller failed the
+ * alloc call, then it can pass in doms_new == NULL && ndoms_new == 1,
+ * and partition_sched_domains() will fallback to the single partition
+ * 'fallback_doms', it also forces the domains to be rebuilt.
+ *
+ * If doms_new == NULL it will be replaced with cpu_online_mask.
+ * ndoms_new == 0 is a special case for destroying existing domains,
+ * and it will not create the default domain.
+ *
+ * Call with hotplug lock held
+ */
+void partition_sched_domains(int ndoms_new, cpumask_var_t doms_new[],
+                            struct sched_domain_attr *dattr_new)
+{
+       int i, j, n;
+       int new_topology;
+
+       mutex_lock(&sched_domains_mutex);
+
+       /* always unregister in case we don't destroy any domains */
+       unregister_sched_domain_sysctl();
+
+       /* Let architecture update cpu core mappings. */
+       new_topology = arch_update_cpu_topology();
+
+       n = doms_new ? ndoms_new : 0;
+
+       /* Destroy deleted domains */
+       for (i = 0; i < ndoms_cur; i++) {
+               for (j = 0; j < n && !new_topology; j++) {
+                       if (cpumask_equal(doms_cur[i], doms_new[j])
+                           && dattrs_equal(dattr_cur, i, dattr_new, j))
+                               goto match1;
+               }
+               /* no match - a current sched domain not in new doms_new[] */
+               detach_destroy_domains(doms_cur[i]);
+match1:
+               ;
+       }
+
+       if (doms_new == NULL) {
+               ndoms_cur = 0;
+               doms_new = &fallback_doms;
+               cpumask_andnot(doms_new[0], cpu_active_mask, cpu_isolated_map);
+               WARN_ON_ONCE(dattr_new);
+       }
+
+       /* Build new domains */
+       for (i = 0; i < ndoms_new; i++) {
+               for (j = 0; j < ndoms_cur && !new_topology; j++) {
+                       if (cpumask_equal(doms_new[i], doms_cur[j])
+                           && dattrs_equal(dattr_new, i, dattr_cur, j))
+                               goto match2;
+               }
+               /* no match - add a new doms_new */
+               build_sched_domains(doms_new[i], dattr_new ? dattr_new + i : NULL);
+match2:
+               ;
+       }
+
+       /* Remember the new sched domains */
+       if (doms_cur != &fallback_doms)
+               free_sched_domains(doms_cur, ndoms_cur);
+       kfree(dattr_cur);       /* kfree(NULL) is safe */
+       doms_cur = doms_new;
+       dattr_cur = dattr_new;
+       ndoms_cur = ndoms_new;
+
+       register_sched_domain_sysctl();
+
+       mutex_unlock(&sched_domains_mutex);
+}
+
+#if defined(CONFIG_SCHED_MC) || defined(CONFIG_SCHED_SMT)
+static void reinit_sched_domains(void)
+{
+       get_online_cpus();
+
+       /* Destroy domains first to force the rebuild */
+       partition_sched_domains(0, NULL, NULL);
+
+       rebuild_sched_domains();
+       put_online_cpus();
+}
+
+static ssize_t sched_power_savings_store(const char *buf, size_t count, int smt)
+{
+       unsigned int level = 0;
+
+       if (sscanf(buf, "%u", &level) != 1)
+               return -EINVAL;
+
+       /*
+        * level is always be positive so don't check for
+        * level < POWERSAVINGS_BALANCE_NONE which is 0
+        * What happens on 0 or 1 byte write,
+        * need to check for count as well?
+        */
+
+       if (level >= MAX_POWERSAVINGS_BALANCE_LEVELS)
+               return -EINVAL;
+
+       if (smt)
+               sched_smt_power_savings = level;
+       else
+               sched_mc_power_savings = level;
+
+       reinit_sched_domains();
+
+       return count;
+}
+
+#ifdef CONFIG_SCHED_MC
+static ssize_t sched_mc_power_savings_show(struct sysdev_class *class,
+                                          struct sysdev_class_attribute *attr,
+                                          char *page)
+{
+       return sprintf(page, "%u\n", sched_mc_power_savings);
+}
+static ssize_t sched_mc_power_savings_store(struct sysdev_class *class,
+                                           struct sysdev_class_attribute *attr,
+                                           const char *buf, size_t count)
+{
+       return sched_power_savings_store(buf, count, 0);
+}
+static SYSDEV_CLASS_ATTR(sched_mc_power_savings, 0644,
+                        sched_mc_power_savings_show,
+                        sched_mc_power_savings_store);
+#endif
+
+#ifdef CONFIG_SCHED_SMT
+static ssize_t sched_smt_power_savings_show(struct sysdev_class *dev,
+                                           struct sysdev_class_attribute *attr,
+                                           char *page)
+{
+       return sprintf(page, "%u\n", sched_smt_power_savings);
+}
+static ssize_t sched_smt_power_savings_store(struct sysdev_class *dev,
+                                            struct sysdev_class_attribute *attr,
+                                            const char *buf, size_t count)
+{
+       return sched_power_savings_store(buf, count, 1);
+}
+static SYSDEV_CLASS_ATTR(sched_smt_power_savings, 0644,
+                  sched_smt_power_savings_show,
+                  sched_smt_power_savings_store);
+#endif
+
+int __init sched_create_sysfs_power_savings_entries(struct sysdev_class *cls)
+{
+       int err = 0;
+
+#ifdef CONFIG_SCHED_SMT
+       if (smt_capable())
+               err = sysfs_create_file(&cls->kset.kobj,
+                                       &attr_sched_smt_power_savings.attr);
+#endif
+#ifdef CONFIG_SCHED_MC
+       if (!err && mc_capable())
+               err = sysfs_create_file(&cls->kset.kobj,
+                                       &attr_sched_mc_power_savings.attr);
+#endif
+       return err;
+}
+#endif /* CONFIG_SCHED_MC || CONFIG_SCHED_SMT */
+
+/*
+ * Update cpusets according to cpu_active mask.  If cpusets are
+ * disabled, cpuset_update_active_cpus() becomes a simple wrapper
+ * around partition_sched_domains().
+ */
+static int cpuset_cpu_active(struct notifier_block *nfb, unsigned long action,
+                            void *hcpu)
+{
+       switch (action & ~CPU_TASKS_FROZEN) {
+       case CPU_ONLINE:
+       case CPU_DOWN_FAILED:
+               cpuset_update_active_cpus();
+               return NOTIFY_OK;
+       default:
+               return NOTIFY_DONE;
+       }
+}
+
+static int cpuset_cpu_inactive(struct notifier_block *nfb, unsigned long action,
+                              void *hcpu)
+{
+       switch (action & ~CPU_TASKS_FROZEN) {
+       case CPU_DOWN_PREPARE:
+               cpuset_update_active_cpus();
+               return NOTIFY_OK;
+       default:
+               return NOTIFY_DONE;
+       }
+}
+
+static int update_runtime(struct notifier_block *nfb,
+                               unsigned long action, void *hcpu)
+{
+       int cpu = (int)(long)hcpu;
+
+       switch (action) {
+       case CPU_DOWN_PREPARE:
+       case CPU_DOWN_PREPARE_FROZEN:
+               disable_runtime(cpu_rq(cpu));
+               return NOTIFY_OK;
+
+       case CPU_DOWN_FAILED:
+       case CPU_DOWN_FAILED_FROZEN:
+       case CPU_ONLINE:
+       case CPU_ONLINE_FROZEN:
+               enable_runtime(cpu_rq(cpu));
+               return NOTIFY_OK;
+
+       default:
+               return NOTIFY_DONE;
+       }
+}
+
+void __init sched_init_smp(void)
+{
+       cpumask_var_t non_isolated_cpus;
+
+       alloc_cpumask_var(&non_isolated_cpus, GFP_KERNEL);
+       alloc_cpumask_var(&fallback_doms, GFP_KERNEL);
+
+       get_online_cpus();
+       mutex_lock(&sched_domains_mutex);
+       init_sched_domains(cpu_active_mask);
+       cpumask_andnot(non_isolated_cpus, cpu_possible_mask, cpu_isolated_map);
+       if (cpumask_empty(non_isolated_cpus))
+               cpumask_set_cpu(smp_processor_id(), non_isolated_cpus);
+       mutex_unlock(&sched_domains_mutex);
+       put_online_cpus();
+
+       hotcpu_notifier(cpuset_cpu_active, CPU_PRI_CPUSET_ACTIVE);
+       hotcpu_notifier(cpuset_cpu_inactive, CPU_PRI_CPUSET_INACTIVE);
+
+       /* RT runtime code needs to handle some hotplug events */
+       hotcpu_notifier(update_runtime, 0);
+
+       init_hrtick();
+
+       /* Move init over to a non-isolated CPU */
+       if (set_cpus_allowed_ptr(current, non_isolated_cpus) < 0)
+               BUG();
+       sched_init_granularity();
+       free_cpumask_var(non_isolated_cpus);
+
+       init_sched_rt_class();
+}
+#else
+void __init sched_init_smp(void)
+{
+       sched_init_granularity();
+}
+#endif /* CONFIG_SMP */
+
+const_debug unsigned int sysctl_timer_migration = 1;
+
+int in_sched_functions(unsigned long addr)
+{
+       return in_lock_functions(addr) ||
+               (addr >= (unsigned long)__sched_text_start
+               && addr < (unsigned long)__sched_text_end);
+}
+
+static void init_cfs_rq(struct cfs_rq *cfs_rq)
+{
+       cfs_rq->tasks_timeline = RB_ROOT;
+       INIT_LIST_HEAD(&cfs_rq->tasks);
+       cfs_rq->min_vruntime = (u64)(-(1LL << 20));
+#ifndef CONFIG_64BIT
+       cfs_rq->min_vruntime_copy = cfs_rq->min_vruntime;
+#endif
+}
+
+static void init_rt_rq(struct rt_rq *rt_rq, struct rq *rq)
+{
+       struct rt_prio_array *array;
+       int i;
+
+       array = &rt_rq->active;
+       for (i = 0; i < MAX_RT_PRIO; i++) {
+               INIT_LIST_HEAD(array->queue + i);
+               __clear_bit(i, array->bitmap);
+       }
+       /* delimiter for bitsearch: */
+       __set_bit(MAX_RT_PRIO, array->bitmap);
+
+#if defined CONFIG_SMP
+       rt_rq->highest_prio.curr = MAX_RT_PRIO;
+       rt_rq->highest_prio.next = MAX_RT_PRIO;
+       rt_rq->rt_nr_migratory = 0;
+       rt_rq->overloaded = 0;
+       plist_head_init(&rt_rq->pushable_tasks);
+#endif
+
+       rt_rq->rt_time = 0;
+       rt_rq->rt_throttled = 0;
+       rt_rq->rt_runtime = 0;
+       raw_spin_lock_init(&rt_rq->rt_runtime_lock);
+}
+
+#ifdef CONFIG_FAIR_GROUP_SCHED
+static void init_tg_cfs_entry(struct task_group *tg, struct cfs_rq *cfs_rq,
+                               struct sched_entity *se, int cpu,
+                               struct sched_entity *parent)
+{
+       struct rq *rq = cpu_rq(cpu);
+
+       cfs_rq->tg = tg;
+       cfs_rq->rq = rq;
+#ifdef CONFIG_SMP
+       /* allow initial update_cfs_load() to truncate */
+       cfs_rq->load_stamp = 1;
+#endif
+
+       tg->cfs_rq[cpu] = cfs_rq;
+       tg->se[cpu] = se;
+
+       /* se could be NULL for root_task_group */
+       if (!se)
+               return;
+
+       if (!parent)
+               se->cfs_rq = &rq->cfs;
+       else
+               se->cfs_rq = parent->my_q;
+
+       se->my_q = cfs_rq;
+       update_load_set(&se->load, 0);
+       se->parent = parent;
+}
+#endif
+
+#ifdef CONFIG_RT_GROUP_SCHED
+static void init_tg_rt_entry(struct task_group *tg, struct rt_rq *rt_rq,
+               struct sched_rt_entity *rt_se, int cpu,
+               struct sched_rt_entity *parent)
+{
+       struct rq *rq = cpu_rq(cpu);
+
+       rt_rq->highest_prio.curr = MAX_RT_PRIO;
+       rt_rq->rt_nr_boosted = 0;
+       rt_rq->rq = rq;
+       rt_rq->tg = tg;
+
+       tg->rt_rq[cpu] = rt_rq;
+       tg->rt_se[cpu] = rt_se;
+
+       if (!rt_se)
+               return;
+
+       if (!parent)
+               rt_se->rt_rq = &rq->rt;
+       else
+               rt_se->rt_rq = parent->my_q;
+
+       rt_se->my_q = rt_rq;
+       rt_se->parent = parent;
+       INIT_LIST_HEAD(&rt_se->run_list);
+}
+#endif
+
+void __init sched_init(void)
+{
+       int i, j;
+       unsigned long alloc_size = 0, ptr;
+
+#ifdef CONFIG_FAIR_GROUP_SCHED
+       alloc_size += 2 * nr_cpu_ids * sizeof(void **);
+#endif
+#ifdef CONFIG_RT_GROUP_SCHED
+       alloc_size += 2 * nr_cpu_ids * sizeof(void **);
+#endif
+#ifdef CONFIG_CPUMASK_OFFSTACK
+       alloc_size += num_possible_cpus() * cpumask_size();
+#endif
+       if (alloc_size) {
+               ptr = (unsigned long)kzalloc(alloc_size, GFP_NOWAIT);
+
+#ifdef CONFIG_FAIR_GROUP_SCHED
+               root_task_group.se = (struct sched_entity **)ptr;
+               ptr += nr_cpu_ids * sizeof(void **);
+
+               root_task_group.cfs_rq = (struct cfs_rq **)ptr;
+               ptr += nr_cpu_ids * sizeof(void **);
+
+#endif /* CONFIG_FAIR_GROUP_SCHED */
+#ifdef CONFIG_RT_GROUP_SCHED
+               root_task_group.rt_se = (struct sched_rt_entity **)ptr;
+               ptr += nr_cpu_ids * sizeof(void **);
+
+               root_task_group.rt_rq = (struct rt_rq **)ptr;
+               ptr += nr_cpu_ids * sizeof(void **);
+
+#endif /* CONFIG_RT_GROUP_SCHED */
+#ifdef CONFIG_CPUMASK_OFFSTACK
+               for_each_possible_cpu(i) {
+                       per_cpu(load_balance_tmpmask, i) = (void *)ptr;
+                       ptr += cpumask_size();
+               }
+#endif /* CONFIG_CPUMASK_OFFSTACK */
+       }
+
+#ifdef CONFIG_SMP
+       init_defrootdomain();
+#endif
+
+       init_rt_bandwidth(&def_rt_bandwidth,
+                       global_rt_period(), global_rt_runtime());
+
+#ifdef CONFIG_RT_GROUP_SCHED
+       init_rt_bandwidth(&root_task_group.rt_bandwidth,
+                       global_rt_period(), global_rt_runtime());
+#endif /* CONFIG_RT_GROUP_SCHED */
+
+#ifdef CONFIG_CGROUP_SCHED
+       list_add(&root_task_group.list, &task_groups);
+       INIT_LIST_HEAD(&root_task_group.children);
+       autogroup_init(&init_task);
+#endif /* CONFIG_CGROUP_SCHED */
+
+       for_each_possible_cpu(i) {
+               struct rq *rq;
+
+               rq = cpu_rq(i);
+               raw_spin_lock_init(&rq->lock);
+               rq->nr_running = 0;
+               rq->calc_load_active = 0;
+               rq->calc_load_update = jiffies + LOAD_FREQ;
+               init_cfs_rq(&rq->cfs);
+               init_rt_rq(&rq->rt, rq);
+#ifdef CONFIG_FAIR_GROUP_SCHED
+               root_task_group.shares = root_task_group_load;
+               INIT_LIST_HEAD(&rq->leaf_cfs_rq_list);
+               /*
+                * How much cpu bandwidth does root_task_group get?
+                *
+                * In case of task-groups formed thr' the cgroup filesystem, it
+                * gets 100% of the cpu resources in the system. This overall
+                * system cpu resource is divided among the tasks of
+                * root_task_group and its child task-groups in a fair manner,
+                * based on each entity's (task or task-group's) weight
+                * (se->load.weight).
+                *
+                * In other words, if root_task_group has 10 tasks of weight
+                * 1024) and two child groups A0 and A1 (of weight 1024 each),
+                * then A0's share of the cpu resource is:
+                *
+                *      A0's bandwidth = 1024 / (10*1024 + 1024 + 1024) = 8.33%
+                *
+                * We achieve this by letting root_task_group's tasks sit
+                * directly in rq->cfs (i.e root_task_group->se[] = NULL).
+                */
+               init_tg_cfs_entry(&root_task_group, &rq->cfs, NULL, i, NULL);
+#endif /* CONFIG_FAIR_GROUP_SCHED */
+
+               rq->rt.rt_runtime = def_rt_bandwidth.rt_runtime;
+#ifdef CONFIG_RT_GROUP_SCHED
+               INIT_LIST_HEAD(&rq->leaf_rt_rq_list);
+               init_tg_rt_entry(&root_task_group, &rq->rt, NULL, i, NULL);
+#endif
+
+               for (j = 0; j < CPU_LOAD_IDX_MAX; j++)
+                       rq->cpu_load[j] = 0;
+
+               rq->last_load_update_tick = jiffies;
+
+#ifdef CONFIG_SMP
+               rq->sd = NULL;
+               rq->rd = NULL;
+               rq->cpu_power = SCHED_POWER_SCALE;
+               rq->post_schedule = 0;
+               rq->active_balance = 0;
+               rq->next_balance = jiffies;
+               rq->push_cpu = 0;
+               rq->cpu = i;
+               rq->online = 0;
+               rq->idle_stamp = 0;
+               rq->avg_idle = 2*sysctl_sched_migration_cost;
+               rq_attach_root(rq, &def_root_domain);
+#ifdef CONFIG_NO_HZ
+               rq->nohz_balance_kick = 0;
+               init_sched_softirq_csd(&per_cpu(remote_sched_softirq_cb, i));
+#endif
+#endif
+               init_rq_hrtick(rq);
+               atomic_set(&rq->nr_iowait, 0);
+       }
+
+       set_load_weight(&init_task);
+
+#ifdef CONFIG_PREEMPT_NOTIFIERS
+       INIT_HLIST_HEAD(&init_task.preempt_notifiers);
+#endif
+
+#ifdef CONFIG_SMP
+       open_softirq(SCHED_SOFTIRQ, run_rebalance_domains);
+#endif
+
+#ifdef CONFIG_RT_MUTEXES
+       plist_head_init(&init_task.pi_waiters);
+#endif
+
+       /*
+        * The boot idle thread does lazy MMU switching as well:
+        */
+       atomic_inc(&init_mm.mm_count);
+       enter_lazy_tlb(&init_mm, current);
+
+       /*
+        * Make us the idle thread. Technically, schedule() should not be
+        * called from this thread, however somewhere below it might be,
+        * but because we are the idle thread, we just pick up running again
+        * when this runqueue becomes "idle".
+        */
+       init_idle(current, smp_processor_id());
+
+       calc_load_update = jiffies + LOAD_FREQ;
+
+       /*
+        * During early bootup we pretend to be a normal task:
+        */
+       current->sched_class = &fair_sched_class;
+
+       /* Allocate the nohz_cpu_mask if CONFIG_CPUMASK_OFFSTACK */
+       zalloc_cpumask_var(&nohz_cpu_mask, GFP_NOWAIT);
+#ifdef CONFIG_SMP
+       zalloc_cpumask_var(&sched_domains_tmpmask, GFP_NOWAIT);
+#ifdef CONFIG_NO_HZ
+       zalloc_cpumask_var(&nohz.idle_cpus_mask, GFP_NOWAIT);
+       alloc_cpumask_var(&nohz.grp_idle_mask, GFP_NOWAIT);
+       atomic_set(&nohz.load_balancer, nr_cpu_ids);
+       atomic_set(&nohz.first_pick_cpu, nr_cpu_ids);
+       atomic_set(&nohz.second_pick_cpu, nr_cpu_ids);
+#endif
+       /* May be allocated at isolcpus cmdline parse time */
+       if (cpu_isolated_map == NULL)
+               zalloc_cpumask_var(&cpu_isolated_map, GFP_NOWAIT);
+#endif /* SMP */
+
+       scheduler_running = 1;
+}
+
+#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
+static inline int preempt_count_equals(int preempt_offset)
+{
+       int nested = (preempt_count() & ~PREEMPT_ACTIVE) + rcu_preempt_depth();
+
+       return (nested == preempt_offset);
+}
+
+static int __might_sleep_init_called;
+int __init __might_sleep_init(void)
+{
+       __might_sleep_init_called = 1;
+       return 0;
+}
+early_initcall(__might_sleep_init);
+
+void __might_sleep(const char *file, int line, int preempt_offset)
+{
+       static unsigned long prev_jiffy;        /* ratelimiting */
+
+       if ((preempt_count_equals(preempt_offset) && !irqs_disabled()) ||
+           oops_in_progress)
+               return;
+       if (system_state != SYSTEM_RUNNING &&
+           (!__might_sleep_init_called || system_state != SYSTEM_BOOTING))
+               return;
+       if (time_before(jiffies, prev_jiffy + HZ) && prev_jiffy)
+               return;
+       prev_jiffy = jiffies;
+
+       printk(KERN_ERR
+               "BUG: sleeping function called from invalid context at %s:%d\n",
+                       file, line);
+       printk(KERN_ERR
+               "in_atomic(): %d, irqs_disabled(): %d, pid: %d, name: %s\n",
+                       in_atomic(), irqs_disabled(),
+                       current->pid, current->comm);
+
+       debug_show_held_locks(current);
+       if (irqs_disabled())
+               print_irqtrace_events(current);
+       dump_stack();
+}
+EXPORT_SYMBOL(__might_sleep);
+#endif
+
+#ifdef CONFIG_MAGIC_SYSRQ
+static void normalize_task(struct rq *rq, struct task_struct *p)
+{
+       const struct sched_class *prev_class = p->sched_class;
+       int old_prio = p->prio;
+       int on_rq;
+
+       on_rq = p->on_rq;
+       if (on_rq)
+               deactivate_task(rq, p, 0);
+       __setscheduler(rq, p, SCHED_NORMAL, 0);
+       if (on_rq) {
+               activate_task(rq, p, 0);
+               resched_task(rq->curr);
+       }
+
+       check_class_changed(rq, p, prev_class, old_prio);
+}
+
+void normalize_rt_tasks(void)
+{
+       struct task_struct *g, *p;
+       unsigned long flags;
+       struct rq *rq;
+
+       read_lock_irqsave(&tasklist_lock, flags);
+       do_each_thread(g, p) {
+               /*
+                * Only normalize user tasks:
+                */
+               if (!p->mm)
+                       continue;
+
+               p->se.exec_start                = 0;
+#ifdef CONFIG_SCHEDSTATS
+               p->se.statistics.wait_start     = 0;
+               p->se.statistics.sleep_start    = 0;
+               p->se.statistics.block_start    = 0;
+#endif
+
+               if (!rt_task(p)) {
+                       /*
+                        * Renice negative nice level userspace
+                        * tasks back to 0:
+                        */
+                       if (TASK_NICE(p) < 0 && p->mm)
+                               set_user_nice(p, 0);
+                       continue;
+               }
+
+               raw_spin_lock(&p->pi_lock);
+               rq = __task_rq_lock(p);
+
+               normalize_task(rq, p);
+
+               __task_rq_unlock(rq);
+               raw_spin_unlock(&p->pi_lock);
+       } while_each_thread(g, p);
+
+       read_unlock_irqrestore(&tasklist_lock, flags);
+}
+
+#endif /* CONFIG_MAGIC_SYSRQ */
+
+#if defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB)
+/*
+ * These functions are only useful for the IA64 MCA handling, or kdb.
+ *
+ * They can only be called when the whole system has been
+ * stopped - every CPU needs to be quiescent, and no scheduling
+ * activity can take place. Using them for anything else would
+ * be a serious bug, and as a result, they aren't even visible
+ * under any other configuration.
+ */
+
+/**
+ * curr_task - return the current task for a given cpu.
+ * @cpu: the processor in question.
+ *
+ * ONLY VALID WHEN THE WHOLE SYSTEM IS STOPPED!
+ */
+struct task_struct *curr_task(int cpu)
+{
+       return cpu_curr(cpu);
+}
+
+#endif /* defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB) */
+
+#ifdef CONFIG_IA64
+/**
+ * set_curr_task - set the current task for a given cpu.
+ * @cpu: the processor in question.
+ * @p: the task pointer to set.
+ *
+ * Description: This function must only be used when non-maskable interrupts
+ * are serviced on a separate stack. It allows the architecture to switch the
+ * notion of the current task on a cpu in a non-blocking manner. This function
+ * must be called with all CPU's synchronized, and interrupts disabled, the
+ * and caller must save the original value of the current task (see
+ * curr_task() above) and restore that value before reenabling interrupts and
+ * re-starting the system.
+ *
+ * ONLY VALID WHEN THE WHOLE SYSTEM IS STOPPED!
+ */
+void set_curr_task(int cpu, struct task_struct *p)
+{
+       cpu_curr(cpu) = p;
+}
+
+#endif
+
+#ifdef CONFIG_FAIR_GROUP_SCHED
+static void free_fair_sched_group(struct task_group *tg)
+{
+       int i;
+
+       for_each_possible_cpu(i) {
+               if (tg->cfs_rq)
+                       kfree(tg->cfs_rq[i]);
+               if (tg->se)
+                       kfree(tg->se[i]);
+       }
+
+       kfree(tg->cfs_rq);
+       kfree(tg->se);
+}
+
+static
+int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent)
+{
+       struct cfs_rq *cfs_rq;
+       struct sched_entity *se;
+       int i;
+
+       tg->cfs_rq = kzalloc(sizeof(cfs_rq) * nr_cpu_ids, GFP_KERNEL);
+       if (!tg->cfs_rq)
+               goto err;
+       tg->se = kzalloc(sizeof(se) * nr_cpu_ids, GFP_KERNEL);
+       if (!tg->se)
+               goto err;
+
+       tg->shares = NICE_0_LOAD;
+
+       for_each_possible_cpu(i) {
+               cfs_rq = kzalloc_node(sizeof(struct cfs_rq),
+                                     GFP_KERNEL, cpu_to_node(i));
+               if (!cfs_rq)
+                       goto err;
+
+               se = kzalloc_node(sizeof(struct sched_entity),
+                                 GFP_KERNEL, cpu_to_node(i));
+               if (!se)
+                       goto err_free_rq;
+
+               init_cfs_rq(cfs_rq);
+               init_tg_cfs_entry(tg, cfs_rq, se, i, parent->se[i]);
+       }
+
+       return 1;
+
+err_free_rq:
+       kfree(cfs_rq);
+err:
+       return 0;
+}
+
+static inline void unregister_fair_sched_group(struct task_group *tg, int cpu)
+{
+       struct rq *rq = cpu_rq(cpu);
+       unsigned long flags;
+
+       /*
+       * Only empty task groups can be destroyed; so we can speculatively
+       * check on_list without danger of it being re-added.
+       */
+       if (!tg->cfs_rq[cpu]->on_list)
+               return;
+
+       raw_spin_lock_irqsave(&rq->lock, flags);
+       list_del_leaf_cfs_rq(tg->cfs_rq[cpu]);
+       raw_spin_unlock_irqrestore(&rq->lock, flags);
+}
+#else /* !CONFIG_FAIR_GROUP_SCHED */
+static inline void free_fair_sched_group(struct task_group *tg)
+{
+}
+
+static inline
+int alloc_fair_sched_group(struct task_group *tg, struct task_group *parent)
+{
+       return 1;
+}
+
+static inline void unregister_fair_sched_group(struct task_group *tg, int cpu)
+{
+}
+#endif /* CONFIG_FAIR_GROUP_SCHED */
+
+#ifdef CONFIG_RT_GROUP_SCHED
+static void free_rt_sched_group(struct task_group *tg)
+{
+       int i;
+
+       if (tg->rt_se)
+               destroy_rt_bandwidth(&tg->rt_bandwidth);
+
+       for_each_possible_cpu(i) {
+               if (tg->rt_rq)
+                       kfree(tg->rt_rq[i]);
+               if (tg->rt_se)
+                       kfree(tg->rt_se[i]);
+       }
+
+       kfree(tg->rt_rq);
+       kfree(tg->rt_se);
+}
+
+static
+int alloc_rt_sched_group(struct task_group *tg, struct task_group *parent)
+{
+       struct rt_rq *rt_rq;
+       struct sched_rt_entity *rt_se;
+       int i;
+
+       tg->rt_rq = kzalloc(sizeof(rt_rq) * nr_cpu_ids, GFP_KERNEL);
+       if (!tg->rt_rq)
+               goto err;
+       tg->rt_se = kzalloc(sizeof(rt_se) * nr_cpu_ids, GFP_KERNEL);
+       if (!tg->rt_se)
+               goto err;
+
+       init_rt_bandwidth(&tg->rt_bandwidth,
+                       ktime_to_ns(def_rt_bandwidth.rt_period), 0);
+
+       for_each_possible_cpu(i) {
+               rt_rq = kzalloc_node(sizeof(struct rt_rq),
+                                    GFP_KERNEL, cpu_to_node(i));
+               if (!rt_rq)
+                       goto err;
+
+               rt_se = kzalloc_node(sizeof(struct sched_rt_entity),
+                                    GFP_KERNEL, cpu_to_node(i));
+               if (!rt_se)
+                       goto err_free_rq;
+
+               init_rt_rq(rt_rq, cpu_rq(i));
+               rt_rq->rt_runtime = tg->rt_bandwidth.rt_runtime;
+               init_tg_rt_entry(tg, rt_rq, rt_se, i, parent->rt_se[i]);
+       }
+
+       return 1;
+
+err_free_rq:
+       kfree(rt_rq);
+err:
+       return 0;
+}
+#else /* !CONFIG_RT_GROUP_SCHED */
+static inline void free_rt_sched_group(struct task_group *tg)
+{
+}
+
+static inline
+int alloc_rt_sched_group(struct task_group *tg, struct task_group *parent)
+{
+       return 1;
+}
+#endif /* CONFIG_RT_GROUP_SCHED */
+
+#ifdef CONFIG_CGROUP_SCHED
+static void free_sched_group(struct task_group *tg)
+{
+       free_fair_sched_group(tg);
+       free_rt_sched_group(tg);
+       autogroup_free(tg);
+       kfree(tg);
+}
+
+/* allocate runqueue etc for a new task group */
+struct task_group *sched_create_group(struct task_group *parent)
+{
+       struct task_group *tg;
+       unsigned long flags;
+
+       tg = kzalloc(sizeof(*tg), GFP_KERNEL);
+       if (!tg)
+               return ERR_PTR(-ENOMEM);
+
+       if (!alloc_fair_sched_group(tg, parent))
+               goto err;
+
+       if (!alloc_rt_sched_group(tg, parent))
+               goto err;
+
+       spin_lock_irqsave(&task_group_lock, flags);
+       list_add_rcu(&tg->list, &task_groups);
+
+       WARN_ON(!parent); /* root should already exist */
+
+       tg->parent = parent;
+       INIT_LIST_HEAD(&tg->children);
+       list_add_rcu(&tg->siblings, &parent->children);
+       spin_unlock_irqrestore(&task_group_lock, flags);
+
+       return tg;
+
+err:
+       free_sched_group(tg);
+       return ERR_PTR(-ENOMEM);
+}
+
+/* rcu callback to free various structures associated with a task group */
+static void free_sched_group_rcu(struct rcu_head *rhp)
+{
+       /* now it should be safe to free those cfs_rqs */
+       free_sched_group(container_of(rhp, struct task_group, rcu));
+}
+
+/* Destroy runqueue etc associated with a task group */
+void sched_destroy_group(struct task_group *tg)
+{
+       unsigned long flags;
+       int i;
+
+       /* end participation in shares distribution */
+       for_each_possible_cpu(i)
+               unregister_fair_sched_group(tg, i);
+
+       spin_lock_irqsave(&task_group_lock, flags);
+       list_del_rcu(&tg->list);
+       list_del_rcu(&tg->siblings);
+       spin_unlock_irqrestore(&task_group_lock, flags);
+
+       /* wait for possible concurrent references to cfs_rqs complete */
+       call_rcu(&tg->rcu, free_sched_group_rcu);
+}
+
+/* change task's runqueue when it moves between groups.
+ *     The caller of this function should have put the task in its new group
+ *     by now. This function just updates tsk->se.cfs_rq and tsk->se.parent to
+ *     reflect its new group.
+ */
+void sched_move_task(struct task_struct *tsk)
+{
+       int on_rq, running;
+       unsigned long flags;
+       struct rq *rq;
+
+       rq = task_rq_lock(tsk, &flags);
+
+       running = task_current(rq, tsk);
+       on_rq = tsk->on_rq;
+
+       if (on_rq)
+               dequeue_task(rq, tsk, 0);
+       if (unlikely(running))
+               tsk->sched_class->put_prev_task(rq, tsk);
+
+#ifdef CONFIG_FAIR_GROUP_SCHED
+       if (tsk->sched_class->task_move_group)
+               tsk->sched_class->task_move_group(tsk, on_rq);
+       else
+#endif
+               set_task_rq(tsk, task_cpu(tsk));
+
+       if (unlikely(running))
+               tsk->sched_class->set_curr_task(rq);
+       if (on_rq)
+               enqueue_task(rq, tsk, 0);
+
+       task_rq_unlock(rq, tsk, &flags);
+}
+#endif /* CONFIG_CGROUP_SCHED */
+
+#ifdef CONFIG_FAIR_GROUP_SCHED
+static DEFINE_MUTEX(shares_mutex);
+
+int sched_group_set_shares(struct task_group *tg, unsigned long shares)
+{
+       int i;
+       unsigned long flags;
+
+       /*
+        * We can't change the weight of the root cgroup.
+        */
+       if (!tg->se[0])
+               return -EINVAL;
+
+       shares = clamp(shares, scale_load(MIN_SHARES), scale_load(MAX_SHARES));
+
+       mutex_lock(&shares_mutex);
+       if (tg->shares == shares)
+               goto done;
+
+       tg->shares = shares;
+       for_each_possible_cpu(i) {
+               struct rq *rq = cpu_rq(i);
+               struct sched_entity *se;
+
+               se = tg->se[i];
+               /* Propagate contribution to hierarchy */
+               raw_spin_lock_irqsave(&rq->lock, flags);
+               for_each_sched_entity(se)
+                       update_cfs_shares(group_cfs_rq(se));
+               raw_spin_unlock_irqrestore(&rq->lock, flags);
+       }
+
+done:
+       mutex_unlock(&shares_mutex);
+       return 0;
+}
+
+unsigned long sched_group_shares(struct task_group *tg)
+{
+       return tg->shares;
+}
+#endif
+
+#ifdef CONFIG_RT_GROUP_SCHED
+/*
+ * Ensure that the real time constraints are schedulable.
+ */
+static DEFINE_MUTEX(rt_constraints_mutex);
+
+static unsigned long to_ratio(u64 period, u64 runtime)
+{
+       if (runtime == RUNTIME_INF)
+               return 1ULL << 20;
+
+       return div64_u64(runtime << 20, period);
+}
+
+/* Must be called with tasklist_lock held */
+static inline int tg_has_rt_tasks(struct task_group *tg)
+{
+       struct task_struct *g, *p;
+
+       do_each_thread(g, p) {
+               if (rt_task(p) && rt_rq_of_se(&p->rt)->tg == tg)
+                       return 1;
+       } while_each_thread(g, p);
+
+       return 0;
+}
+
+struct rt_schedulable_data {
+       struct task_group *tg;
+       u64 rt_period;
+       u64 rt_runtime;
+};
+
+static int tg_schedulable(struct task_group *tg, void *data)
+{
+       struct rt_schedulable_data *d = data;
+       struct task_group *child;
+       unsigned long total, sum = 0;
+       u64 period, runtime;
+
+       period = ktime_to_ns(tg->rt_bandwidth.rt_period);
+       runtime = tg->rt_bandwidth.rt_runtime;
+
+       if (tg == d->tg) {
+               period = d->rt_period;
+               runtime = d->rt_runtime;
+       }
+
+       /*
+        * Cannot have more runtime than the period.
+        */
+       if (runtime > period && runtime != RUNTIME_INF)
+               return -EINVAL;
+
+       /*
+        * Ensure we don't starve existing RT tasks.
+        */
+       if (rt_bandwidth_enabled() && !runtime && tg_has_rt_tasks(tg))
+               return -EBUSY;
+
+       total = to_ratio(period, runtime);
+
+       /*
+        * Nobody can have more than the global setting allows.
+        */
+       if (total > to_ratio(global_rt_period(), global_rt_runtime()))
+               return -EINVAL;
+
+       /*
+        * The sum of our children's runtime should not exceed our own.
+        */
+       list_for_each_entry_rcu(child, &tg->children, siblings) {
+               period = ktime_to_ns(child->rt_bandwidth.rt_period);
+               runtime = child->rt_bandwidth.rt_runtime;
+
+               if (child == d->tg) {
+                       period = d->rt_period;
+                       runtime = d->rt_runtime;
+               }
+
+               sum += to_ratio(period, runtime);
+       }
+
+       if (sum > total)
+               return -EINVAL;
+
+       return 0;
+}
+
+static int __rt_schedulable(struct task_group *tg, u64 period, u64 runtime)
+{
+       struct rt_schedulable_data data = {
+               .tg = tg,
+               .rt_period = period,
+               .rt_runtime = runtime,
+       };
+
+       return walk_tg_tree(tg_schedulable, tg_nop, &data);
+}
+
+static int tg_set_bandwidth(struct task_group *tg,
+               u64 rt_period, u64 rt_runtime)
+{
+       int i, err = 0;
+
+       mutex_lock(&rt_constraints_mutex);
+       read_lock(&tasklist_lock);
+       err = __rt_schedulable(tg, rt_period, rt_runtime);
+       if (err)
+               goto unlock;
+
+       raw_spin_lock_irq(&tg->rt_bandwidth.rt_runtime_lock);
+       tg->rt_bandwidth.rt_period = ns_to_ktime(rt_period);
+       tg->rt_bandwidth.rt_runtime = rt_runtime;
+
+       for_each_possible_cpu(i) {
+               struct rt_rq *rt_rq = tg->rt_rq[i];
+
+               raw_spin_lock(&rt_rq->rt_runtime_lock);
+               rt_rq->rt_runtime = rt_runtime;
+               raw_spin_unlock(&rt_rq->rt_runtime_lock);
+       }
+       raw_spin_unlock_irq(&tg->rt_bandwidth.rt_runtime_lock);
+unlock:
+       read_unlock(&tasklist_lock);
+       mutex_unlock(&rt_constraints_mutex);
+
+       return err;
+}
+
+int sched_group_set_rt_runtime(struct task_group *tg, long rt_runtime_us)
+{
+       u64 rt_runtime, rt_period;
+
+       rt_period = ktime_to_ns(tg->rt_bandwidth.rt_period);
+       rt_runtime = (u64)rt_runtime_us * NSEC_PER_USEC;
+       if (rt_runtime_us < 0)
+               rt_runtime = RUNTIME_INF;
+
+       return tg_set_bandwidth(tg, rt_period, rt_runtime);
+}
+
+long sched_group_rt_runtime(struct task_group *tg)
+{
+       u64 rt_runtime_us;
+
+       if (tg->rt_bandwidth.rt_runtime == RUNTIME_INF)
+               return -1;
+
+       rt_runtime_us = tg->rt_bandwidth.rt_runtime;
+       do_div(rt_runtime_us, NSEC_PER_USEC);
+       return rt_runtime_us;
+}
+
+int sched_group_set_rt_period(struct task_group *tg, long rt_period_us)
+{
+       u64 rt_runtime, rt_period;
+
+       rt_period = (u64)rt_period_us * NSEC_PER_USEC;
+       rt_runtime = tg->rt_bandwidth.rt_runtime;
+
+       if (rt_period == 0)
+               return -EINVAL;
+
+       return tg_set_bandwidth(tg, rt_period, rt_runtime);
+}
+
+long sched_group_rt_period(struct task_group *tg)
+{
+       u64 rt_period_us;
+
+       rt_period_us = ktime_to_ns(tg->rt_bandwidth.rt_period);
+       do_div(rt_period_us, NSEC_PER_USEC);
+       return rt_period_us;
+}
+
+static int sched_rt_global_constraints(void)
+{
+       u64 runtime, period;
+       int ret = 0;
+
+       if (sysctl_sched_rt_period <= 0)
+               return -EINVAL;
+
+       runtime = global_rt_runtime();
+       period = global_rt_period();
+
+       /*
+        * Sanity check on the sysctl variables.
+        */
+       if (runtime > period && runtime != RUNTIME_INF)
+               return -EINVAL;
+
+       mutex_lock(&rt_constraints_mutex);
+       read_lock(&tasklist_lock);
+       ret = __rt_schedulable(NULL, 0, 0);
+       read_unlock(&tasklist_lock);
+       mutex_unlock(&rt_constraints_mutex);
+
+       return ret;
+}
+
+int sched_rt_can_attach(struct task_group *tg, struct task_struct *tsk)
+{
+       /* Don't accept realtime tasks when there is no way for them to run */
+       if (rt_task(tsk) && tg->rt_bandwidth.rt_runtime == 0)
+               return 0;
+
+       return 1;
+}
+
+#else /* !CONFIG_RT_GROUP_SCHED */
+static int sched_rt_global_constraints(void)
+{
+       unsigned long flags;
+       int i;
+
+       if (sysctl_sched_rt_period <= 0)
+               return -EINVAL;
+
+       /*
+        * There's always some RT tasks in the root group
+        * -- migration, kstopmachine etc..
+        */
+       if (sysctl_sched_rt_runtime == 0)
+               return -EBUSY;
+
+       raw_spin_lock_irqsave(&def_rt_bandwidth.rt_runtime_lock, flags);
+       for_each_possible_cpu(i) {
+               struct rt_rq *rt_rq = &cpu_rq(i)->rt;
+
+               raw_spin_lock(&rt_rq->rt_runtime_lock);
+               rt_rq->rt_runtime = global_rt_runtime();
+               raw_spin_unlock(&rt_rq->rt_runtime_lock);
+       }
+       raw_spin_unlock_irqrestore(&def_rt_bandwidth.rt_runtime_lock, flags);
+
+       return 0;
+}
+#endif /* CONFIG_RT_GROUP_SCHED */
+
+int sched_rt_handler(struct ctl_table *table, int write,
+               void __user *buffer, size_t *lenp,
+               loff_t *ppos)
+{
+       int ret;
+       int old_period, old_runtime;
+       static DEFINE_MUTEX(mutex);
+
+       mutex_lock(&mutex);
+       old_period = sysctl_sched_rt_period;
+       old_runtime = sysctl_sched_rt_runtime;
+
+       ret = proc_dointvec(table, write, buffer, lenp, ppos);
+
+       if (!ret && write) {
+               ret = sched_rt_global_constraints();
+               if (ret) {
+                       sysctl_sched_rt_period = old_period;
+                       sysctl_sched_rt_runtime = old_runtime;
+               } else {
+                       def_rt_bandwidth.rt_runtime = global_rt_runtime();
+                       def_rt_bandwidth.rt_period =
+                               ns_to_ktime(global_rt_period());
+               }
+       }
+       mutex_unlock(&mutex);
+
+       return ret;
+}
+
+#ifdef CONFIG_CGROUP_SCHED
+
+/* return corresponding task_group object of a cgroup */
+static inline struct task_group *cgroup_tg(struct cgroup *cgrp)
+{
+       return container_of(cgroup_subsys_state(cgrp, cpu_cgroup_subsys_id),
+                           struct task_group, css);
+}
+
+static struct cgroup_subsys_state *
+cpu_cgroup_create(struct cgroup_subsys *ss, struct cgroup *cgrp)
+{
+       struct task_group *tg, *parent;
+
+       if (!cgrp->parent) {
+               /* This is early initialization for the top cgroup */
+               return &root_task_group.css;
+       }
+
+       parent = cgroup_tg(cgrp->parent);
+       tg = sched_create_group(parent);
+       if (IS_ERR(tg))
+               return ERR_PTR(-ENOMEM);
+
+       return &tg->css;
+}
+
+static void
+cpu_cgroup_destroy(struct cgroup_subsys *ss, struct cgroup *cgrp)
+{
+       struct task_group *tg = cgroup_tg(cgrp);
+
+       sched_destroy_group(tg);
+}
+
+static int
+cpu_cgroup_can_attach_task(struct cgroup *cgrp, struct task_struct *tsk)
+{
+       if ((current != tsk) && (!capable(CAP_SYS_NICE))) {
+               const struct cred *cred = current_cred(), *tcred;
+
+               tcred = __task_cred(tsk);
+
+               if (cred->euid != tcred->uid && cred->euid != tcred->suid)
+                       return -EPERM;
+       }
+
+#ifdef CONFIG_RT_GROUP_SCHED
+       if (!sched_rt_can_attach(cgroup_tg(cgrp), tsk))
+               return -EINVAL;
+#else
+       /* We don't support RT-tasks being in separate groups */
+       if (tsk->sched_class != &fair_sched_class)
+               return -EINVAL;
+#endif
+       return 0;
+}
+
+static void
+cpu_cgroup_attach_task(struct cgroup *cgrp, struct task_struct *tsk)
+{
+       sched_move_task(tsk);
+}
+
+static void
+cpu_cgroup_exit(struct cgroup_subsys *ss, struct cgroup *cgrp,
+               struct cgroup *old_cgrp, struct task_struct *task)
+{
+       /*
+        * cgroup_exit() is called in the copy_process() failure path.
+        * Ignore this case since the task hasn't ran yet, this avoids
+        * trying to poke a half freed task state from generic code.
+        */
+       if (!(task->flags & PF_EXITING))
+               return;
+
+       sched_move_task(task);
+}
+
+#ifdef CONFIG_FAIR_GROUP_SCHED
+static int cpu_shares_write_u64(struct cgroup *cgrp, struct cftype *cftype,
+                               u64 shareval)
+{
+       return sched_group_set_shares(cgroup_tg(cgrp), scale_load(shareval));
+}
+
+static u64 cpu_shares_read_u64(struct cgroup *cgrp, struct cftype *cft)
+{
+       struct task_group *tg = cgroup_tg(cgrp);
+
+       return (u64) scale_load_down(tg->shares);
+}
+#endif /* CONFIG_FAIR_GROUP_SCHED */
+
+#ifdef CONFIG_RT_GROUP_SCHED
+static int cpu_rt_runtime_write(struct cgroup *cgrp, struct cftype *cft,
+                               s64 val)
+{
+       return sched_group_set_rt_runtime(cgroup_tg(cgrp), val);
+}
+
+static s64 cpu_rt_runtime_read(struct cgroup *cgrp, struct cftype *cft)
+{
+       return sched_group_rt_runtime(cgroup_tg(cgrp));
+}
+
+static int cpu_rt_period_write_uint(struct cgroup *cgrp, struct cftype *cftype,
+               u64 rt_period_us)
+{
+       return sched_group_set_rt_period(cgroup_tg(cgrp), rt_period_us);
+}
+
+static u64 cpu_rt_period_read_uint(struct cgroup *cgrp, struct cftype *cft)
+{
+       return sched_group_rt_period(cgroup_tg(cgrp));
+}
+#endif /* CONFIG_RT_GROUP_SCHED */
+
+static struct cftype cpu_files[] = {
+#ifdef CONFIG_FAIR_GROUP_SCHED
+       {
+               .name = "shares",
+               .read_u64 = cpu_shares_read_u64,
+               .write_u64 = cpu_shares_write_u64,
+       },
+#endif
+#ifdef CONFIG_RT_GROUP_SCHED
+       {
+               .name = "rt_runtime_us",
+               .read_s64 = cpu_rt_runtime_read,
+               .write_s64 = cpu_rt_runtime_write,
+       },
+       {
+               .name = "rt_period_us",
+               .read_u64 = cpu_rt_period_read_uint,
+               .write_u64 = cpu_rt_period_write_uint,
+       },
+#endif
+};
+
+static int cpu_cgroup_populate(struct cgroup_subsys *ss, struct cgroup *cont)
+{
+       return cgroup_add_files(cont, ss, cpu_files, ARRAY_SIZE(cpu_files));
+}
+
+struct cgroup_subsys cpu_cgroup_subsys = {
+       .name           = "cpu",
+       .create         = cpu_cgroup_create,
+       .destroy        = cpu_cgroup_destroy,
+       .can_attach_task = cpu_cgroup_can_attach_task,
+       .attach_task    = cpu_cgroup_attach_task,
+       .exit           = cpu_cgroup_exit,
+       .populate       = cpu_cgroup_populate,
+       .subsys_id      = cpu_cgroup_subsys_id,
+       .early_init     = 1,
+};
+
+#endif /* CONFIG_CGROUP_SCHED */
+
+#ifdef CONFIG_CGROUP_CPUACCT
+
+/*
+ * CPU accounting code for task groups.
+ *
+ * Based on the work by Paul Menage (menage@google.com) and Balbir Singh
+ * (balbir@in.ibm.com).
+ */
+
+/* track cpu usage of a group of tasks and its child groups */
+struct cpuacct {
+       struct cgroup_subsys_state css;
+       /* cpuusage holds pointer to a u64-type object on every cpu */
+       u64 __percpu *cpuusage;
+       struct percpu_counter cpustat[CPUACCT_STAT_NSTATS];
+       struct cpuacct *parent;
+};
+
+struct cgroup_subsys cpuacct_subsys;
+
+/* return cpu accounting group corresponding to this container */
+static inline struct cpuacct *cgroup_ca(struct cgroup *cgrp)
+{
+       return container_of(cgroup_subsys_state(cgrp, cpuacct_subsys_id),
+                           struct cpuacct, css);
+}
+
+/* return cpu accounting group to which this task belongs */
+static inline struct cpuacct *task_ca(struct task_struct *tsk)
+{
+       return container_of(task_subsys_state(tsk, cpuacct_subsys_id),
+                           struct cpuacct, css);
+}
+
+/* create a new cpu accounting group */
+static struct cgroup_subsys_state *cpuacct_create(
+       struct cgroup_subsys *ss, struct cgroup *cgrp)
+{
+       struct cpuacct *ca = kzalloc(sizeof(*ca), GFP_KERNEL);
+       int i;
+
+       if (!ca)
+               goto out;
+
+       ca->cpuusage = alloc_percpu(u64);
+       if (!ca->cpuusage)
+               goto out_free_ca;
+
+       for (i = 0; i < CPUACCT_STAT_NSTATS; i++)
+               if (percpu_counter_init(&ca->cpustat[i], 0))
+                       goto out_free_counters;
+
+       if (cgrp->parent)
+               ca->parent = cgroup_ca(cgrp->parent);
+
+       return &ca->css;
+
+out_free_counters:
+       while (--i >= 0)
+               percpu_counter_destroy(&ca->cpustat[i]);
+       free_percpu(ca->cpuusage);
+out_free_ca:
+       kfree(ca);
+out:
+       return ERR_PTR(-ENOMEM);
+}
+
+/* destroy an existing cpu accounting group */
+static void
+cpuacct_destroy(struct cgroup_subsys *ss, struct cgroup *cgrp)
+{
+       struct cpuacct *ca = cgroup_ca(cgrp);
+       int i;
+
+       for (i = 0; i < CPUACCT_STAT_NSTATS; i++)
+               percpu_counter_destroy(&ca->cpustat[i]);
+       free_percpu(ca->cpuusage);
+       kfree(ca);
+}
+
+static u64 cpuacct_cpuusage_read(struct cpuacct *ca, int cpu)
+{
+       u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
+       u64 data;
+
+#ifndef CONFIG_64BIT
+       /*
+        * Take rq->lock to make 64-bit read safe on 32-bit platforms.
+        */
+       raw_spin_lock_irq(&cpu_rq(cpu)->lock);
+       data = *cpuusage;
+       raw_spin_unlock_irq(&cpu_rq(cpu)->lock);
+#else
+       data = *cpuusage;
+#endif
+
+       return data;
+}
+
+static void cpuacct_cpuusage_write(struct cpuacct *ca, int cpu, u64 val)
+{
+       u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
+
+#ifndef CONFIG_64BIT
+       /*
+        * Take rq->lock to make 64-bit write safe on 32-bit platforms.
+        */
+       raw_spin_lock_irq(&cpu_rq(cpu)->lock);
+       *cpuusage = val;
+       raw_spin_unlock_irq(&cpu_rq(cpu)->lock);
+#else
+       *cpuusage = val;
+#endif
+}
+
+/* return total cpu usage (in nanoseconds) of a group */
+static u64 cpuusage_read(struct cgroup *cgrp, struct cftype *cft)
+{
+       struct cpuacct *ca = cgroup_ca(cgrp);
+       u64 totalcpuusage = 0;
+       int i;
+
+       for_each_present_cpu(i)
+               totalcpuusage += cpuacct_cpuusage_read(ca, i);
+
+       return totalcpuusage;
+}
+
+static int cpuusage_write(struct cgroup *cgrp, struct cftype *cftype,
+                                                               u64 reset)
+{
+       struct cpuacct *ca = cgroup_ca(cgrp);
+       int err = 0;
+       int i;
+
+       if (reset) {
+               err = -EINVAL;
+               goto out;
+       }
+
+       for_each_present_cpu(i)
+               cpuacct_cpuusage_write(ca, i, 0);
+
+out:
+       return err;
+}
+
+static int cpuacct_percpu_seq_read(struct cgroup *cgroup, struct cftype *cft,
+                                  struct seq_file *m)
+{
+       struct cpuacct *ca = cgroup_ca(cgroup);
+       u64 percpu;
+       int i;
+
+       for_each_present_cpu(i) {
+               percpu = cpuacct_cpuusage_read(ca, i);
+               seq_printf(m, "%llu ", (unsigned long long) percpu);
+       }
+       seq_printf(m, "\n");
+       return 0;
+}
+
+static const char *cpuacct_stat_desc[] = {
+       [CPUACCT_STAT_USER] = "user",
+       [CPUACCT_STAT_SYSTEM] = "system",
+};
+
+static int cpuacct_stats_show(struct cgroup *cgrp, struct cftype *cft,
+               struct cgroup_map_cb *cb)
+{
+       struct cpuacct *ca = cgroup_ca(cgrp);
+       int i;
+
+       for (i = 0; i < CPUACCT_STAT_NSTATS; i++) {
+               s64 val = percpu_counter_read(&ca->cpustat[i]);
+               val = cputime64_to_clock_t(val);
+               cb->fill(cb, cpuacct_stat_desc[i], val);
+       }
+       return 0;
+}
+
+static struct cftype files[] = {
+       {
+               .name = "usage",
+               .read_u64 = cpuusage_read,
+               .write_u64 = cpuusage_write,
+       },
+       {
+               .name = "usage_percpu",
+               .read_seq_string = cpuacct_percpu_seq_read,
+       },
+       {
+               .name = "stat",
+               .read_map = cpuacct_stats_show,
+       },
+};
+
+static int cpuacct_populate(struct cgroup_subsys *ss, struct cgroup *cgrp)
+{
+       return cgroup_add_files(cgrp, ss, files, ARRAY_SIZE(files));
+}
+
+/*
+ * charge this task's execution time to its accounting group.
+ *
+ * called with rq->lock held.
+ */
+static void cpuacct_charge(struct task_struct *tsk, u64 cputime)
+{
+       struct cpuacct *ca;
+       int cpu;
+
+       if (unlikely(!cpuacct_subsys.active))
+               return;
+
+       cpu = task_cpu(tsk);
+
+       rcu_read_lock();
+
+       ca = task_ca(tsk);
+
+       for (; ca; ca = ca->parent) {
+               u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
+               *cpuusage += cputime;
+       }
+
+       rcu_read_unlock();
+}
+
+/*
+ * When CONFIG_VIRT_CPU_ACCOUNTING is enabled one jiffy can be very large
+ * in cputime_t units. As a result, cpuacct_update_stats calls
+ * percpu_counter_add with values large enough to always overflow the
+ * per cpu batch limit causing bad SMP scalability.
+ *
+ * To fix this we scale percpu_counter_batch by cputime_one_jiffy so we
+ * batch the same amount of time with CONFIG_VIRT_CPU_ACCOUNTING disabled
+ * and enabled. We cap it at INT_MAX which is the largest allowed batch value.
+ */
+#ifdef CONFIG_SMP
+#define CPUACCT_BATCH  \
+       min_t(long, percpu_counter_batch * cputime_one_jiffy, INT_MAX)
+#else
+#define CPUACCT_BATCH  0
+#endif
+
+/*
+ * Charge the system/user time to the task's accounting group.
+ */
+static void cpuacct_update_stats(struct task_struct *tsk,
+               enum cpuacct_stat_index idx, cputime_t val)
+{
+       struct cpuacct *ca;
+       int batch = CPUACCT_BATCH;
+
+       if (unlikely(!cpuacct_subsys.active))
+               return;
+
+       rcu_read_lock();
+       ca = task_ca(tsk);
+
+       do {
+               __percpu_counter_add(&ca->cpustat[idx], val, batch);
+               ca = ca->parent;
+       } while (ca);
+       rcu_read_unlock();
+}
+
+struct cgroup_subsys cpuacct_subsys = {
+       .name = "cpuacct",
+       .create = cpuacct_create,
+       .destroy = cpuacct_destroy,
+       .populate = cpuacct_populate,
+       .subsys_id = cpuacct_subsys_id,
+};
+#endif /* CONFIG_CGROUP_CPUACCT */
+