2007-07-09 18:51:58 +02:00
|
|
|
/*
|
|
|
|
* Completely Fair Scheduling (CFS) Class (SCHED_NORMAL/SCHED_BATCH)
|
|
|
|
*
|
|
|
|
* Copyright (C) 2007 Red Hat, Inc., Ingo Molnar <mingo@redhat.com>
|
|
|
|
*
|
|
|
|
* Interactivity improvements by Mike Galbraith
|
|
|
|
* (C) 2007 Mike Galbraith <efault@gmx.de>
|
|
|
|
*
|
|
|
|
* Various enhancements by Dmitry Adamushko.
|
|
|
|
* (C) 2007 Dmitry Adamushko <dmitry.adamushko@gmail.com>
|
|
|
|
*
|
|
|
|
* Group scheduling enhancements by Srivatsa Vaddagiri
|
|
|
|
* Copyright IBM Corporation, 2007
|
|
|
|
* Author: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
|
|
|
|
*
|
|
|
|
* Scaled math optimizations by Thomas Gleixner
|
|
|
|
* Copyright (C) 2007, Thomas Gleixner <tglx@linutronix.de>
|
2007-08-25 18:41:53 +02:00
|
|
|
*
|
|
|
|
* Adaptive scheduling granularity, math enhancements by Peter Zijlstra
|
|
|
|
* Copyright (C) 2007 Red Hat, Inc., Peter Zijlstra <pzijlstr@redhat.com>
|
2007-07-09 18:51:58 +02:00
|
|
|
*/
|
|
|
|
|
2008-01-25 21:08:34 +01:00
|
|
|
#include <linux/latencytop.h>
|
|
|
|
|
2007-07-09 18:51:58 +02:00
|
|
|
/*
|
2007-08-25 18:41:53 +02:00
|
|
|
* Targeted preemption latency for CPU-bound tasks:
|
2007-11-26 21:21:49 +01:00
|
|
|
* (default: 20ms * (1 + ilog(ncpus)), units: nanoseconds)
|
2007-07-09 18:51:58 +02:00
|
|
|
*
|
2007-08-25 18:41:53 +02:00
|
|
|
* NOTE: this latency value is not the same as the concept of
|
2007-10-15 17:00:14 +02:00
|
|
|
* 'timeslice length' - timeslices in CFS are of variable length
|
|
|
|
* and have no persistent notion like in traditional, time-slice
|
|
|
|
* based scheduling concepts.
|
2007-07-09 18:51:58 +02:00
|
|
|
*
|
2007-10-15 17:00:14 +02:00
|
|
|
* (to see the precise effective timeslice length of your workload,
|
|
|
|
* run vmstat and monitor the context-switches (cs) field)
|
2007-07-09 18:51:58 +02:00
|
|
|
*/
|
2007-11-09 22:39:38 +01:00
|
|
|
unsigned int sysctl_sched_latency = 20000000ULL;
|
2007-10-15 17:00:02 +02:00
|
|
|
|
|
|
|
/*
|
2007-11-09 22:39:37 +01:00
|
|
|
* Minimal preemption granularity for CPU-bound tasks:
|
2007-11-26 21:21:49 +01:00
|
|
|
* (default: 4 msec * (1 + ilog(ncpus)), units: nanoseconds)
|
2007-10-15 17:00:02 +02:00
|
|
|
*/
|
2007-11-26 21:21:49 +01:00
|
|
|
unsigned int sysctl_sched_min_granularity = 4000000ULL;
|
2007-08-25 18:41:53 +02:00
|
|
|
|
|
|
|
/*
|
2007-11-09 22:39:37 +01:00
|
|
|
* is kept at sysctl_sched_latency / sysctl_sched_min_granularity
|
|
|
|
*/
|
2007-11-26 21:21:49 +01:00
|
|
|
static unsigned int sched_nr_latency = 5;
|
2007-11-09 22:39:37 +01:00
|
|
|
|
|
|
|
/*
|
|
|
|
* After fork, child runs first. (default) If set to 0 then
|
|
|
|
* parent will (try to) run first.
|
2007-08-25 18:41:53 +02:00
|
|
|
*/
|
2007-11-09 22:39:37 +01:00
|
|
|
const_debug unsigned int sysctl_sched_child_runs_first = 1;
|
2007-07-09 18:51:58 +02:00
|
|
|
|
2007-09-19 23:34:46 +02:00
|
|
|
/*
|
|
|
|
* sys_sched_yield() compat mode
|
|
|
|
*
|
|
|
|
* This option switches the agressive yield implementation of the
|
|
|
|
* old scheduler back on.
|
|
|
|
*/
|
|
|
|
unsigned int __read_mostly sysctl_sched_compat_yield;
|
|
|
|
|
2007-07-09 18:51:58 +02:00
|
|
|
/*
|
|
|
|
* SCHED_OTHER wake-up granularity.
|
2008-06-27 13:41:16 +02:00
|
|
|
* (default: 5 msec * (1 + ilog(ncpus)), units: nanoseconds)
|
2007-07-09 18:51:58 +02:00
|
|
|
*
|
|
|
|
* This option delays the preemption effects of decoupled workloads
|
|
|
|
* and reduces their over-scheduling. Synchronous workloads will still
|
|
|
|
* have immediate wakeup/sleep latencies.
|
|
|
|
*/
|
2008-06-27 13:41:16 +02:00
|
|
|
unsigned int sysctl_sched_wakeup_granularity = 5000000UL;
|
2007-07-09 18:51:58 +02:00
|
|
|
|
2007-10-15 17:00:18 +02:00
|
|
|
const_debug unsigned int sysctl_sched_migration_cost = 500000UL;
|
|
|
|
|
2007-07-09 18:51:58 +02:00
|
|
|
/**************************************************************
|
|
|
|
* CFS operations on generic schedulable entities:
|
|
|
|
*/
|
|
|
|
|
2008-04-19 19:45:00 +02:00
|
|
|
static inline struct task_struct *task_of(struct sched_entity *se)
|
|
|
|
{
|
|
|
|
return container_of(se, struct task_struct, se);
|
|
|
|
}
|
|
|
|
|
2007-10-15 17:00:03 +02:00
|
|
|
#ifdef CONFIG_FAIR_GROUP_SCHED
|
2007-07-09 18:51:58 +02:00
|
|
|
|
2007-10-15 17:00:03 +02:00
|
|
|
/* cpu runqueue to which this cfs_rq is attached */
|
2007-07-09 18:51:58 +02:00
|
|
|
static inline struct rq *rq_of(struct cfs_rq *cfs_rq)
|
|
|
|
{
|
2007-10-15 17:00:03 +02:00
|
|
|
return cfs_rq->rq;
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
|
2007-10-15 17:00:03 +02:00
|
|
|
/* An entity is a task if it doesn't "own" a runqueue */
|
|
|
|
#define entity_is_task(se) (!se->my_q)
|
2007-07-09 18:51:58 +02:00
|
|
|
|
2008-04-19 19:45:00 +02:00
|
|
|
/* Walk up scheduling entities hierarchy */
|
|
|
|
#define for_each_sched_entity(se) \
|
|
|
|
for (; se; se = se->parent)
|
|
|
|
|
|
|
|
static inline struct cfs_rq *task_cfs_rq(struct task_struct *p)
|
|
|
|
{
|
|
|
|
return p->se.cfs_rq;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* runqueue on which this entity is (to be) queued */
|
|
|
|
static inline struct cfs_rq *cfs_rq_of(struct sched_entity *se)
|
|
|
|
{
|
|
|
|
return se->cfs_rq;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* runqueue "owned" by this group */
|
|
|
|
static inline struct cfs_rq *group_cfs_rq(struct sched_entity *grp)
|
|
|
|
{
|
|
|
|
return grp->my_q;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Given a group's cfs_rq on one cpu, return its corresponding cfs_rq on
|
|
|
|
* another cpu ('this_cpu')
|
|
|
|
*/
|
|
|
|
static inline struct cfs_rq *cpu_cfs_rq(struct cfs_rq *cfs_rq, int this_cpu)
|
|
|
|
{
|
|
|
|
return cfs_rq->tg->cfs_rq[this_cpu];
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Iterate thr' all leaf cfs_rq's on a runqueue */
|
|
|
|
#define for_each_leaf_cfs_rq(rq, cfs_rq) \
|
|
|
|
list_for_each_entry_rcu(cfs_rq, &rq->leaf_cfs_rq_list, leaf_cfs_rq_list)
|
|
|
|
|
|
|
|
/* Do the two (enqueued) entities belong to the same group ? */
|
|
|
|
static inline int
|
|
|
|
is_same_group(struct sched_entity *se, struct sched_entity *pse)
|
|
|
|
{
|
|
|
|
if (se->cfs_rq == pse->cfs_rq)
|
|
|
|
return 1;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline struct sched_entity *parent_entity(struct sched_entity *se)
|
|
|
|
{
|
|
|
|
return se->parent;
|
|
|
|
}
|
|
|
|
|
2007-10-15 17:00:03 +02:00
|
|
|
#else /* CONFIG_FAIR_GROUP_SCHED */
|
2007-07-09 18:51:58 +02:00
|
|
|
|
2007-10-15 17:00:03 +02:00
|
|
|
static inline struct rq *rq_of(struct cfs_rq *cfs_rq)
|
|
|
|
{
|
|
|
|
return container_of(cfs_rq, struct rq, cfs);
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
#define entity_is_task(se) 1
|
|
|
|
|
2008-04-19 19:45:00 +02:00
|
|
|
#define for_each_sched_entity(se) \
|
|
|
|
for (; se; se = NULL)
|
2007-07-09 18:51:58 +02:00
|
|
|
|
2008-04-19 19:45:00 +02:00
|
|
|
static inline struct cfs_rq *task_cfs_rq(struct task_struct *p)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
2008-04-19 19:45:00 +02:00
|
|
|
return &task_rq(p)->cfs;
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
|
2008-04-19 19:45:00 +02:00
|
|
|
static inline struct cfs_rq *cfs_rq_of(struct sched_entity *se)
|
|
|
|
{
|
|
|
|
struct task_struct *p = task_of(se);
|
|
|
|
struct rq *rq = task_rq(p);
|
|
|
|
|
|
|
|
return &rq->cfs;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* runqueue "owned" by this group */
|
|
|
|
static inline struct cfs_rq *group_cfs_rq(struct sched_entity *grp)
|
|
|
|
{
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline struct cfs_rq *cpu_cfs_rq(struct cfs_rq *cfs_rq, int this_cpu)
|
|
|
|
{
|
|
|
|
return &cpu_rq(this_cpu)->cfs;
|
|
|
|
}
|
|
|
|
|
|
|
|
#define for_each_leaf_cfs_rq(rq, cfs_rq) \
|
|
|
|
for (cfs_rq = &rq->cfs; cfs_rq; cfs_rq = NULL)
|
|
|
|
|
|
|
|
static inline int
|
|
|
|
is_same_group(struct sched_entity *se, struct sched_entity *pse)
|
|
|
|
{
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline struct sched_entity *parent_entity(struct sched_entity *se)
|
|
|
|
{
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
#endif /* CONFIG_FAIR_GROUP_SCHED */
|
|
|
|
|
2007-07-09 18:51:58 +02:00
|
|
|
|
|
|
|
/**************************************************************
|
|
|
|
* Scheduling class tree data structure manipulation methods:
|
|
|
|
*/
|
|
|
|
|
2007-10-15 17:00:14 +02:00
|
|
|
static inline u64 max_vruntime(u64 min_vruntime, u64 vruntime)
|
2007-10-15 17:00:07 +02:00
|
|
|
{
|
2007-10-15 17:00:11 +02:00
|
|
|
s64 delta = (s64)(vruntime - min_vruntime);
|
|
|
|
if (delta > 0)
|
2007-10-15 17:00:07 +02:00
|
|
|
min_vruntime = vruntime;
|
|
|
|
|
|
|
|
return min_vruntime;
|
|
|
|
}
|
|
|
|
|
2007-10-15 17:00:14 +02:00
|
|
|
static inline u64 min_vruntime(u64 min_vruntime, u64 vruntime)
|
2007-10-15 17:00:12 +02:00
|
|
|
{
|
|
|
|
s64 delta = (s64)(vruntime - min_vruntime);
|
|
|
|
if (delta < 0)
|
|
|
|
min_vruntime = vruntime;
|
|
|
|
|
|
|
|
return min_vruntime;
|
|
|
|
}
|
|
|
|
|
2007-10-15 17:00:14 +02:00
|
|
|
static inline s64 entity_key(struct cfs_rq *cfs_rq, struct sched_entity *se)
|
2007-10-15 17:00:05 +02:00
|
|
|
{
|
2007-10-15 17:00:07 +02:00
|
|
|
return se->vruntime - cfs_rq->min_vruntime;
|
2007-10-15 17:00:05 +02:00
|
|
|
}
|
|
|
|
|
2007-07-09 18:51:58 +02:00
|
|
|
/*
|
|
|
|
* Enqueue an entity into the rb-tree:
|
|
|
|
*/
|
2007-10-15 17:00:14 +02:00
|
|
|
static void __enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
|
|
|
struct rb_node **link = &cfs_rq->tasks_timeline.rb_node;
|
|
|
|
struct rb_node *parent = NULL;
|
|
|
|
struct sched_entity *entry;
|
2007-10-15 17:00:05 +02:00
|
|
|
s64 key = entity_key(cfs_rq, se);
|
2007-07-09 18:51:58 +02:00
|
|
|
int leftmost = 1;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Find the right place in the rbtree:
|
|
|
|
*/
|
|
|
|
while (*link) {
|
|
|
|
parent = *link;
|
|
|
|
entry = rb_entry(parent, struct sched_entity, run_node);
|
|
|
|
/*
|
|
|
|
* We dont care about collisions. Nodes with
|
|
|
|
* the same key stay together.
|
|
|
|
*/
|
2007-10-15 17:00:05 +02:00
|
|
|
if (key < entity_key(cfs_rq, entry)) {
|
2007-07-09 18:51:58 +02:00
|
|
|
link = &parent->rb_left;
|
|
|
|
} else {
|
|
|
|
link = &parent->rb_right;
|
|
|
|
leftmost = 0;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Maintain a cache of leftmost tree entries (it is frequently
|
|
|
|
* used):
|
|
|
|
*/
|
2008-03-14 20:55:51 +01:00
|
|
|
if (leftmost) {
|
2007-10-15 17:00:11 +02:00
|
|
|
cfs_rq->rb_leftmost = &se->run_node;
|
2008-03-14 20:55:51 +01:00
|
|
|
/*
|
|
|
|
* maintain cfs_rq->min_vruntime to be a monotonic increasing
|
|
|
|
* value tracking the leftmost vruntime in the tree.
|
|
|
|
*/
|
|
|
|
cfs_rq->min_vruntime =
|
|
|
|
max_vruntime(cfs_rq->min_vruntime, se->vruntime);
|
|
|
|
}
|
2007-07-09 18:51:58 +02:00
|
|
|
|
|
|
|
rb_link_node(&se->run_node, parent, link);
|
|
|
|
rb_insert_color(&se->run_node, &cfs_rq->tasks_timeline);
|
|
|
|
}
|
|
|
|
|
2007-10-15 17:00:14 +02:00
|
|
|
static void __dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
2008-03-14 20:55:51 +01:00
|
|
|
if (cfs_rq->rb_leftmost == &se->run_node) {
|
|
|
|
struct rb_node *next_node;
|
|
|
|
struct sched_entity *next;
|
|
|
|
|
|
|
|
next_node = rb_next(&se->run_node);
|
|
|
|
cfs_rq->rb_leftmost = next_node;
|
|
|
|
|
|
|
|
if (next_node) {
|
|
|
|
next = rb_entry(next_node,
|
|
|
|
struct sched_entity, run_node);
|
|
|
|
cfs_rq->min_vruntime =
|
|
|
|
max_vruntime(cfs_rq->min_vruntime,
|
|
|
|
next->vruntime);
|
|
|
|
}
|
|
|
|
}
|
2007-10-15 17:00:04 +02:00
|
|
|
|
2008-03-14 21:12:12 +01:00
|
|
|
if (cfs_rq->next == se)
|
|
|
|
cfs_rq->next = NULL;
|
|
|
|
|
2007-07-09 18:51:58 +02:00
|
|
|
rb_erase(&se->run_node, &cfs_rq->tasks_timeline);
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline struct rb_node *first_fair(struct cfs_rq *cfs_rq)
|
|
|
|
{
|
|
|
|
return cfs_rq->rb_leftmost;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct sched_entity *__pick_next_entity(struct cfs_rq *cfs_rq)
|
|
|
|
{
|
|
|
|
return rb_entry(first_fair(cfs_rq), struct sched_entity, run_node);
|
|
|
|
}
|
|
|
|
|
2007-10-15 17:00:05 +02:00
|
|
|
static inline struct sched_entity *__pick_last_entity(struct cfs_rq *cfs_rq)
|
|
|
|
{
|
2008-02-22 10:32:21 +01:00
|
|
|
struct rb_node *last = rb_last(&cfs_rq->tasks_timeline);
|
2007-10-15 17:00:05 +02:00
|
|
|
|
2008-02-22 08:55:53 +01:00
|
|
|
if (!last)
|
|
|
|
return NULL;
|
2008-02-22 10:32:21 +01:00
|
|
|
|
|
|
|
return rb_entry(last, struct sched_entity, run_node);
|
2007-10-15 17:00:05 +02:00
|
|
|
}
|
|
|
|
|
2007-07-09 18:51:58 +02:00
|
|
|
/**************************************************************
|
|
|
|
* Scheduling class statistics methods:
|
|
|
|
*/
|
|
|
|
|
2007-11-09 22:39:37 +01:00
|
|
|
#ifdef CONFIG_SCHED_DEBUG
|
|
|
|
int sched_nr_latency_handler(struct ctl_table *table, int write,
|
|
|
|
struct file *filp, void __user *buffer, size_t *lenp,
|
|
|
|
loff_t *ppos)
|
|
|
|
{
|
|
|
|
int ret = proc_dointvec_minmax(table, write, filp, buffer, lenp, ppos);
|
|
|
|
|
|
|
|
if (ret || !write)
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
sched_nr_latency = DIV_ROUND_UP(sysctl_sched_latency,
|
|
|
|
sysctl_sched_min_granularity);
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
#endif
|
2007-10-15 17:00:13 +02:00
|
|
|
|
2008-06-27 13:41:11 +02:00
|
|
|
/*
|
|
|
|
* delta *= w / rw
|
|
|
|
*/
|
|
|
|
static inline unsigned long
|
|
|
|
calc_delta_weight(unsigned long delta, struct sched_entity *se)
|
|
|
|
{
|
|
|
|
for_each_sched_entity(se) {
|
|
|
|
delta = calc_delta_mine(delta,
|
|
|
|
se->load.weight, &cfs_rq_of(se)->load);
|
|
|
|
}
|
|
|
|
|
|
|
|
return delta;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* delta *= rw / w
|
|
|
|
*/
|
|
|
|
static inline unsigned long
|
|
|
|
calc_delta_fair(unsigned long delta, struct sched_entity *se)
|
|
|
|
{
|
|
|
|
for_each_sched_entity(se) {
|
|
|
|
delta = calc_delta_mine(delta,
|
|
|
|
cfs_rq_of(se)->load.weight, &se->load);
|
|
|
|
}
|
|
|
|
|
|
|
|
return delta;
|
|
|
|
}
|
|
|
|
|
2007-10-15 17:00:13 +02:00
|
|
|
/*
|
|
|
|
* The idea is to set a period in which each task runs once.
|
|
|
|
*
|
|
|
|
* When there are too many tasks (sysctl_sched_nr_latency) we have to stretch
|
|
|
|
* this period because otherwise the slices get too small.
|
|
|
|
*
|
|
|
|
* p = (nr <= nl) ? l : l*nr/nl
|
|
|
|
*/
|
2007-10-15 17:00:04 +02:00
|
|
|
static u64 __sched_period(unsigned long nr_running)
|
|
|
|
{
|
|
|
|
u64 period = sysctl_sched_latency;
|
2007-11-09 22:39:37 +01:00
|
|
|
unsigned long nr_latency = sched_nr_latency;
|
2007-10-15 17:00:04 +02:00
|
|
|
|
|
|
|
if (unlikely(nr_running > nr_latency)) {
|
2008-01-25 21:08:21 +01:00
|
|
|
period = sysctl_sched_min_granularity;
|
2007-10-15 17:00:04 +02:00
|
|
|
period *= nr_running;
|
|
|
|
}
|
|
|
|
|
|
|
|
return period;
|
|
|
|
}
|
|
|
|
|
2007-10-15 17:00:13 +02:00
|
|
|
/*
|
|
|
|
* We calculate the wall-time slice from the period by taking a part
|
|
|
|
* proportional to the weight.
|
|
|
|
*
|
|
|
|
* s = p*w/rw
|
|
|
|
*/
|
2007-10-15 17:00:05 +02:00
|
|
|
static u64 sched_slice(struct cfs_rq *cfs_rq, struct sched_entity *se)
|
2007-08-25 18:41:53 +02:00
|
|
|
{
|
2008-06-27 13:41:11 +02:00
|
|
|
return calc_delta_weight(__sched_period(cfs_rq->nr_running), se);
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
|
2007-10-15 17:00:13 +02:00
|
|
|
/*
|
2008-04-19 19:45:00 +02:00
|
|
|
* We calculate the vruntime slice of a to be inserted task
|
2007-10-15 17:00:13 +02:00
|
|
|
*
|
2008-06-27 13:41:11 +02:00
|
|
|
* vs = s*rw/w = p
|
2007-10-15 17:00:13 +02:00
|
|
|
*/
|
2008-04-19 19:45:00 +02:00
|
|
|
static u64 sched_vslice_add(struct cfs_rq *cfs_rq, struct sched_entity *se)
|
2007-10-15 17:00:10 +02:00
|
|
|
{
|
2008-04-19 19:45:00 +02:00
|
|
|
unsigned long nr_running = cfs_rq->nr_running;
|
2007-10-15 17:00:10 +02:00
|
|
|
|
2008-04-19 19:45:00 +02:00
|
|
|
if (!se->on_rq)
|
|
|
|
nr_running++;
|
2007-10-15 17:00:10 +02:00
|
|
|
|
2008-06-27 13:41:11 +02:00
|
|
|
return __sched_period(nr_running);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The goal of calc_delta_asym() is to be asymmetrically around NICE_0_LOAD, in
|
|
|
|
* that it favours >=0 over <0.
|
|
|
|
*
|
|
|
|
* -20 |
|
|
|
|
* |
|
|
|
|
* 0 --------+-------
|
|
|
|
* .'
|
|
|
|
* 19 .'
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
static unsigned long
|
|
|
|
calc_delta_asym(unsigned long delta, struct sched_entity *se)
|
|
|
|
{
|
|
|
|
struct load_weight lw = {
|
|
|
|
.weight = NICE_0_LOAD,
|
|
|
|
.inv_weight = 1UL << (WMULT_SHIFT-NICE_0_SHIFT)
|
|
|
|
};
|
2007-10-15 17:00:12 +02:00
|
|
|
|
2008-04-19 19:45:00 +02:00
|
|
|
for_each_sched_entity(se) {
|
2008-06-27 13:41:11 +02:00
|
|
|
struct load_weight *se_lw = &se->load;
|
2008-06-27 13:41:13 +02:00
|
|
|
unsigned long rw = cfs_rq_of(se)->load.weight;
|
2008-04-19 19:45:00 +02:00
|
|
|
|
2008-06-27 13:41:12 +02:00
|
|
|
#ifdef CONFIG_FAIR_SCHED_GROUP
|
|
|
|
struct cfs_rq *cfs_rq = se->my_q;
|
|
|
|
struct task_group *tg = NULL
|
|
|
|
|
|
|
|
if (cfs_rq)
|
|
|
|
tg = cfs_rq->tg;
|
|
|
|
|
|
|
|
if (tg && tg->shares < NICE_0_LOAD) {
|
|
|
|
/*
|
|
|
|
* scale shares to what it would have been had
|
|
|
|
* tg->weight been NICE_0_LOAD:
|
|
|
|
*
|
|
|
|
* weight = 1024 * shares / tg->weight
|
|
|
|
*/
|
|
|
|
lw.weight *= se->load.weight;
|
|
|
|
lw.weight /= tg->shares;
|
|
|
|
|
|
|
|
lw.inv_weight = 0;
|
|
|
|
|
|
|
|
se_lw = &lw;
|
2008-06-27 13:41:13 +02:00
|
|
|
rw += lw.weight - se->load.weight;
|
2008-06-27 13:41:12 +02:00
|
|
|
} else
|
|
|
|
#endif
|
2008-04-19 19:45:00 +02:00
|
|
|
|
2008-06-27 13:41:13 +02:00
|
|
|
if (se->load.weight < NICE_0_LOAD) {
|
2008-06-27 13:41:11 +02:00
|
|
|
se_lw = &lw;
|
2008-06-27 13:41:13 +02:00
|
|
|
rw += NICE_0_LOAD - se->load.weight;
|
|
|
|
}
|
2008-04-19 19:45:00 +02:00
|
|
|
|
2008-06-27 13:41:13 +02:00
|
|
|
delta = calc_delta_mine(delta, rw, se_lw);
|
2008-04-19 19:45:00 +02:00
|
|
|
}
|
|
|
|
|
2008-06-27 13:41:11 +02:00
|
|
|
return delta;
|
2007-10-15 17:00:10 +02:00
|
|
|
}
|
|
|
|
|
2007-07-09 18:51:58 +02:00
|
|
|
/*
|
|
|
|
* Update the current task's runtime statistics. Skip current tasks that
|
|
|
|
* are not in our scheduling class.
|
|
|
|
*/
|
|
|
|
static inline void
|
2007-10-15 17:00:03 +02:00
|
|
|
__update_curr(struct cfs_rq *cfs_rq, struct sched_entity *curr,
|
|
|
|
unsigned long delta_exec)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
2007-10-15 17:00:06 +02:00
|
|
|
unsigned long delta_exec_weighted;
|
2007-07-09 18:51:58 +02:00
|
|
|
|
2007-08-02 17:41:40 +02:00
|
|
|
schedstat_set(curr->exec_max, max((u64)delta_exec, curr->exec_max));
|
2007-07-09 18:51:58 +02:00
|
|
|
|
|
|
|
curr->sum_exec_runtime += delta_exec;
|
2007-10-15 17:00:06 +02:00
|
|
|
schedstat_add(cfs_rq, exec_clock, delta_exec);
|
2008-06-27 13:41:11 +02:00
|
|
|
delta_exec_weighted = calc_delta_fair(delta_exec, curr);
|
2007-10-15 17:00:04 +02:00
|
|
|
curr->vruntime += delta_exec_weighted;
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
|
2007-08-09 11:16:47 +02:00
|
|
|
static void update_curr(struct cfs_rq *cfs_rq)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
2007-10-15 17:00:03 +02:00
|
|
|
struct sched_entity *curr = cfs_rq->curr;
|
2007-10-15 17:00:03 +02:00
|
|
|
u64 now = rq_of(cfs_rq)->clock;
|
2007-07-09 18:51:58 +02:00
|
|
|
unsigned long delta_exec;
|
|
|
|
|
|
|
|
if (unlikely(!curr))
|
|
|
|
return;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Get the amount of time the current task was running
|
|
|
|
* since the last time we changed load (this cannot
|
|
|
|
* overflow on 32 bits):
|
|
|
|
*/
|
2007-10-15 17:00:03 +02:00
|
|
|
delta_exec = (unsigned long)(now - curr->exec_start);
|
2007-07-09 18:51:58 +02:00
|
|
|
|
2007-10-15 17:00:03 +02:00
|
|
|
__update_curr(cfs_rq, curr, delta_exec);
|
|
|
|
curr->exec_start = now;
|
2007-12-02 20:04:49 +01:00
|
|
|
|
|
|
|
if (entity_is_task(curr)) {
|
|
|
|
struct task_struct *curtask = task_of(curr);
|
|
|
|
|
|
|
|
cpuacct_charge(curtask, delta_exec);
|
|
|
|
}
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static inline void
|
2007-08-09 11:16:47 +02:00
|
|
|
update_stats_wait_start(struct cfs_rq *cfs_rq, struct sched_entity *se)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
2007-08-09 11:16:47 +02:00
|
|
|
schedstat_set(se->wait_start, rq_of(cfs_rq)->clock);
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Task is being enqueued - update stats:
|
|
|
|
*/
|
2007-08-09 11:16:47 +02:00
|
|
|
static void update_stats_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
|
|
|
/*
|
|
|
|
* Are we enqueueing a waiting task? (for current tasks
|
|
|
|
* a dequeue/enqueue event is a NOP)
|
|
|
|
*/
|
2007-10-15 17:00:03 +02:00
|
|
|
if (se != cfs_rq->curr)
|
2007-08-09 11:16:47 +02:00
|
|
|
update_stats_wait_start(cfs_rq, se);
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
2007-08-09 11:16:47 +02:00
|
|
|
update_stats_wait_end(struct cfs_rq *cfs_rq, struct sched_entity *se)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
2007-10-15 17:00:06 +02:00
|
|
|
schedstat_set(se->wait_max, max(se->wait_max,
|
|
|
|
rq_of(cfs_rq)->clock - se->wait_start));
|
2008-01-25 21:08:35 +01:00
|
|
|
schedstat_set(se->wait_count, se->wait_count + 1);
|
|
|
|
schedstat_set(se->wait_sum, se->wait_sum +
|
|
|
|
rq_of(cfs_rq)->clock - se->wait_start);
|
2007-08-02 17:41:40 +02:00
|
|
|
schedstat_set(se->wait_start, 0);
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static inline void
|
2007-08-09 11:16:48 +02:00
|
|
|
update_stats_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
|
|
|
/*
|
|
|
|
* Mark the end of the wait period if dequeueing a
|
|
|
|
* waiting task:
|
|
|
|
*/
|
2007-10-15 17:00:03 +02:00
|
|
|
if (se != cfs_rq->curr)
|
2007-08-09 11:16:47 +02:00
|
|
|
update_stats_wait_end(cfs_rq, se);
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We are picking a new current task - update its stats:
|
|
|
|
*/
|
|
|
|
static inline void
|
2007-08-09 11:16:47 +02:00
|
|
|
update_stats_curr_start(struct cfs_rq *cfs_rq, struct sched_entity *se)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
|
|
|
/*
|
|
|
|
* We are starting a new run period:
|
|
|
|
*/
|
2007-08-09 11:16:47 +02:00
|
|
|
se->exec_start = rq_of(cfs_rq)->clock;
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/**************************************************
|
|
|
|
* Scheduling class queueing methods:
|
|
|
|
*/
|
|
|
|
|
2008-06-27 13:41:14 +02:00
|
|
|
#if defined CONFIG_SMP && defined CONFIG_FAIR_GROUP_SCHED
|
|
|
|
static void
|
|
|
|
add_cfs_task_weight(struct cfs_rq *cfs_rq, unsigned long weight)
|
|
|
|
{
|
|
|
|
cfs_rq->task_weight += weight;
|
|
|
|
}
|
|
|
|
#else
|
|
|
|
static inline void
|
|
|
|
add_cfs_task_weight(struct cfs_rq *cfs_rq, unsigned long weight)
|
|
|
|
{
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2007-10-15 17:00:07 +02:00
|
|
|
static void
|
|
|
|
account_entity_enqueue(struct cfs_rq *cfs_rq, struct sched_entity *se)
|
|
|
|
{
|
|
|
|
update_load_add(&cfs_rq->load, se->load.weight);
|
2008-06-27 13:41:14 +02:00
|
|
|
if (!parent_entity(se))
|
|
|
|
inc_cpu_load(rq_of(cfs_rq), se->load.weight);
|
|
|
|
if (entity_is_task(se))
|
|
|
|
add_cfs_task_weight(cfs_rq, se->load.weight);
|
2007-10-15 17:00:07 +02:00
|
|
|
cfs_rq->nr_running++;
|
|
|
|
se->on_rq = 1;
|
2008-04-19 19:45:00 +02:00
|
|
|
list_add(&se->group_node, &cfs_rq->tasks);
|
2007-10-15 17:00:07 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
account_entity_dequeue(struct cfs_rq *cfs_rq, struct sched_entity *se)
|
|
|
|
{
|
|
|
|
update_load_sub(&cfs_rq->load, se->load.weight);
|
2008-06-27 13:41:14 +02:00
|
|
|
if (!parent_entity(se))
|
|
|
|
dec_cpu_load(rq_of(cfs_rq), se->load.weight);
|
|
|
|
if (entity_is_task(se))
|
|
|
|
add_cfs_task_weight(cfs_rq, -se->load.weight);
|
2007-10-15 17:00:07 +02:00
|
|
|
cfs_rq->nr_running--;
|
|
|
|
se->on_rq = 0;
|
2008-04-19 19:45:00 +02:00
|
|
|
list_del_init(&se->group_node);
|
2007-10-15 17:00:07 +02:00
|
|
|
}
|
|
|
|
|
2007-08-09 11:16:48 +02:00
|
|
|
static void enqueue_sleeper(struct cfs_rq *cfs_rq, struct sched_entity *se)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
|
|
|
#ifdef CONFIG_SCHEDSTATS
|
|
|
|
if (se->sleep_start) {
|
2007-08-09 11:16:47 +02:00
|
|
|
u64 delta = rq_of(cfs_rq)->clock - se->sleep_start;
|
2008-01-25 21:08:34 +01:00
|
|
|
struct task_struct *tsk = task_of(se);
|
2007-07-09 18:51:58 +02:00
|
|
|
|
|
|
|
if ((s64)delta < 0)
|
|
|
|
delta = 0;
|
|
|
|
|
|
|
|
if (unlikely(delta > se->sleep_max))
|
|
|
|
se->sleep_max = delta;
|
|
|
|
|
|
|
|
se->sleep_start = 0;
|
|
|
|
se->sum_sleep_runtime += delta;
|
2008-01-25 21:08:34 +01:00
|
|
|
|
|
|
|
account_scheduler_latency(tsk, delta >> 10, 1);
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
if (se->block_start) {
|
2007-08-09 11:16:47 +02:00
|
|
|
u64 delta = rq_of(cfs_rq)->clock - se->block_start;
|
2008-01-25 21:08:34 +01:00
|
|
|
struct task_struct *tsk = task_of(se);
|
2007-07-09 18:51:58 +02:00
|
|
|
|
|
|
|
if ((s64)delta < 0)
|
|
|
|
delta = 0;
|
|
|
|
|
|
|
|
if (unlikely(delta > se->block_max))
|
|
|
|
se->block_max = delta;
|
|
|
|
|
|
|
|
se->block_start = 0;
|
|
|
|
se->sum_sleep_runtime += delta;
|
2007-10-02 14:13:08 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Blocking time is in units of nanosecs, so shift by 20 to
|
|
|
|
* get a milliseconds-range estimation of the amount of
|
|
|
|
* time that the task spent sleeping:
|
|
|
|
*/
|
|
|
|
if (unlikely(prof_on == SLEEP_PROFILING)) {
|
2007-10-15 17:00:06 +02:00
|
|
|
|
2007-10-02 14:13:08 +02:00
|
|
|
profile_hits(SLEEP_PROFILING, (void *)get_wchan(tsk),
|
|
|
|
delta >> 20);
|
|
|
|
}
|
2008-01-25 21:08:34 +01:00
|
|
|
account_scheduler_latency(tsk, delta >> 10, 0);
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
2007-10-15 17:00:10 +02:00
|
|
|
static void check_spread(struct cfs_rq *cfs_rq, struct sched_entity *se)
|
|
|
|
{
|
|
|
|
#ifdef CONFIG_SCHED_DEBUG
|
|
|
|
s64 d = se->vruntime - cfs_rq->min_vruntime;
|
|
|
|
|
|
|
|
if (d < 0)
|
|
|
|
d = -d;
|
|
|
|
|
|
|
|
if (d > 3*sysctl_sched_latency)
|
|
|
|
schedstat_inc(cfs_rq, nr_spread_over);
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
2007-10-15 17:00:05 +02:00
|
|
|
static void
|
|
|
|
place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial)
|
|
|
|
{
|
2007-10-15 17:00:10 +02:00
|
|
|
u64 vruntime;
|
2007-10-15 17:00:05 +02:00
|
|
|
|
2008-03-14 20:55:51 +01:00
|
|
|
if (first_fair(cfs_rq)) {
|
|
|
|
vruntime = min_vruntime(cfs_rq->min_vruntime,
|
|
|
|
__pick_next_entity(cfs_rq)->vruntime);
|
|
|
|
} else
|
|
|
|
vruntime = cfs_rq->min_vruntime;
|
2007-10-15 17:00:05 +02:00
|
|
|
|
2007-11-09 22:39:37 +01:00
|
|
|
/*
|
|
|
|
* The 'current' period is already promised to the current tasks,
|
|
|
|
* however the extra weight of the new task will slow them down a
|
|
|
|
* little, place the new task so that it fits in the slot that
|
|
|
|
* stays open at the end.
|
|
|
|
*/
|
2007-10-15 17:00:05 +02:00
|
|
|
if (initial && sched_feat(START_DEBIT))
|
2007-10-15 17:00:13 +02:00
|
|
|
vruntime += sched_vslice_add(cfs_rq, se);
|
2007-10-15 17:00:05 +02:00
|
|
|
|
2007-10-15 17:00:11 +02:00
|
|
|
if (!initial) {
|
2007-11-09 22:39:37 +01:00
|
|
|
/* sleeps upto a single latency don't count. */
|
2008-06-27 13:41:11 +02:00
|
|
|
if (sched_feat(NEW_FAIR_SLEEPERS)) {
|
|
|
|
unsigned long thresh = sysctl_sched_latency;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* convert the sleeper threshold into virtual time
|
|
|
|
*/
|
|
|
|
if (sched_feat(NORMALIZED_SLEEPER))
|
|
|
|
thresh = calc_delta_fair(thresh, se);
|
|
|
|
|
|
|
|
vruntime -= thresh;
|
|
|
|
}
|
2007-10-15 17:00:11 +02:00
|
|
|
|
2007-11-09 22:39:37 +01:00
|
|
|
/* ensure we never gain time by being placed backwards. */
|
|
|
|
vruntime = max_vruntime(se->vruntime, vruntime);
|
2007-10-15 17:00:05 +02:00
|
|
|
}
|
|
|
|
|
2007-10-15 17:00:10 +02:00
|
|
|
se->vruntime = vruntime;
|
2007-10-15 17:00:05 +02:00
|
|
|
}
|
|
|
|
|
2007-07-09 18:51:58 +02:00
|
|
|
static void
|
2007-10-15 17:00:08 +02:00
|
|
|
enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int wakeup)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
|
|
|
/*
|
2007-10-15 17:00:13 +02:00
|
|
|
* Update run-time statistics of the 'current'.
|
2007-07-09 18:51:58 +02:00
|
|
|
*/
|
2007-08-09 11:16:47 +02:00
|
|
|
update_curr(cfs_rq);
|
2008-05-05 23:56:17 +02:00
|
|
|
account_entity_enqueue(cfs_rq, se);
|
2007-07-09 18:51:58 +02:00
|
|
|
|
2007-10-15 17:00:04 +02:00
|
|
|
if (wakeup) {
|
2007-10-15 17:00:05 +02:00
|
|
|
place_entity(cfs_rq, se, 0);
|
2007-08-09 11:16:48 +02:00
|
|
|
enqueue_sleeper(cfs_rq, se);
|
2007-10-15 17:00:04 +02:00
|
|
|
}
|
2007-07-09 18:51:58 +02:00
|
|
|
|
2007-08-09 11:16:47 +02:00
|
|
|
update_stats_enqueue(cfs_rq, se);
|
2007-10-15 17:00:10 +02:00
|
|
|
check_spread(cfs_rq, se);
|
2007-10-15 17:00:08 +02:00
|
|
|
if (se != cfs_rq->curr)
|
|
|
|
__enqueue_entity(cfs_rq, se);
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
2007-08-09 11:16:48 +02:00
|
|
|
dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int sleep)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
2007-10-15 17:00:13 +02:00
|
|
|
/*
|
|
|
|
* Update run-time statistics of the 'current'.
|
|
|
|
*/
|
|
|
|
update_curr(cfs_rq);
|
|
|
|
|
2007-08-09 11:16:48 +02:00
|
|
|
update_stats_dequeue(cfs_rq, se);
|
2007-10-15 17:00:06 +02:00
|
|
|
if (sleep) {
|
2007-10-15 17:00:10 +02:00
|
|
|
#ifdef CONFIG_SCHEDSTATS
|
2007-07-09 18:51:58 +02:00
|
|
|
if (entity_is_task(se)) {
|
|
|
|
struct task_struct *tsk = task_of(se);
|
|
|
|
|
|
|
|
if (tsk->state & TASK_INTERRUPTIBLE)
|
2007-08-09 11:16:47 +02:00
|
|
|
se->sleep_start = rq_of(cfs_rq)->clock;
|
2007-07-09 18:51:58 +02:00
|
|
|
if (tsk->state & TASK_UNINTERRUPTIBLE)
|
2007-08-09 11:16:47 +02:00
|
|
|
se->block_start = rq_of(cfs_rq)->clock;
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
2007-10-15 17:00:06 +02:00
|
|
|
#endif
|
2007-10-15 17:00:10 +02:00
|
|
|
}
|
|
|
|
|
2007-10-15 17:00:08 +02:00
|
|
|
if (se != cfs_rq->curr)
|
2007-10-15 17:00:07 +02:00
|
|
|
__dequeue_entity(cfs_rq, se);
|
|
|
|
account_entity_dequeue(cfs_rq, se);
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Preempt the current task with a newly woken task if needed:
|
|
|
|
*/
|
2007-09-05 14:32:49 +02:00
|
|
|
static void
|
2007-10-15 17:00:05 +02:00
|
|
|
check_preempt_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
2007-09-05 14:32:49 +02:00
|
|
|
unsigned long ideal_runtime, delta_exec;
|
|
|
|
|
2007-10-15 17:00:05 +02:00
|
|
|
ideal_runtime = sched_slice(cfs_rq, curr);
|
2007-09-05 14:32:49 +02:00
|
|
|
delta_exec = curr->sum_exec_runtime - curr->prev_sum_exec_runtime;
|
2007-11-09 22:39:39 +01:00
|
|
|
if (delta_exec > ideal_runtime)
|
2007-07-09 18:51:58 +02:00
|
|
|
resched_task(rq_of(cfs_rq)->curr);
|
|
|
|
}
|
|
|
|
|
2007-10-15 17:00:08 +02:00
|
|
|
static void
|
2007-08-09 11:16:48 +02:00
|
|
|
set_next_entity(struct cfs_rq *cfs_rq, struct sched_entity *se)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
2007-10-15 17:00:08 +02:00
|
|
|
/* 'current' is not kept within the tree. */
|
|
|
|
if (se->on_rq) {
|
|
|
|
/*
|
|
|
|
* Any task has to be enqueued before it get to execute on
|
|
|
|
* a CPU. So account for the time it spent waiting on the
|
|
|
|
* runqueue.
|
|
|
|
*/
|
|
|
|
update_stats_wait_end(cfs_rq, se);
|
|
|
|
__dequeue_entity(cfs_rq, se);
|
|
|
|
}
|
|
|
|
|
2007-08-09 11:16:47 +02:00
|
|
|
update_stats_curr_start(cfs_rq, se);
|
2007-10-15 17:00:03 +02:00
|
|
|
cfs_rq->curr = se;
|
2007-10-15 17:00:02 +02:00
|
|
|
#ifdef CONFIG_SCHEDSTATS
|
|
|
|
/*
|
|
|
|
* Track our maximum slice length, if the CPU's load is at
|
|
|
|
* least twice that of our own weight (i.e. dont track it
|
|
|
|
* when there are only lesser-weight tasks around):
|
|
|
|
*/
|
2007-10-15 17:00:06 +02:00
|
|
|
if (rq_of(cfs_rq)->load.weight >= 2*se->load.weight) {
|
2007-10-15 17:00:02 +02:00
|
|
|
se->slice_max = max(se->slice_max,
|
|
|
|
se->sum_exec_runtime - se->prev_sum_exec_runtime);
|
|
|
|
}
|
|
|
|
#endif
|
2007-09-05 14:32:49 +02:00
|
|
|
se->prev_sum_exec_runtime = se->sum_exec_runtime;
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
|
2008-03-14 21:12:12 +01:00
|
|
|
static struct sched_entity *
|
|
|
|
pick_next(struct cfs_rq *cfs_rq, struct sched_entity *se)
|
|
|
|
{
|
2008-06-27 13:41:16 +02:00
|
|
|
struct rq *rq = rq_of(cfs_rq);
|
|
|
|
u64 pair_slice = rq->clock - cfs_rq->pair_start;
|
2008-03-14 21:12:12 +01:00
|
|
|
|
2008-06-27 13:41:16 +02:00
|
|
|
if (!cfs_rq->next || pair_slice > sched_slice(cfs_rq, cfs_rq->next)) {
|
|
|
|
cfs_rq->pair_start = rq->clock;
|
2008-03-14 21:12:12 +01:00
|
|
|
return se;
|
2008-06-27 13:41:16 +02:00
|
|
|
}
|
2008-03-14 21:12:12 +01:00
|
|
|
|
|
|
|
return cfs_rq->next;
|
|
|
|
}
|
|
|
|
|
2007-08-09 11:16:48 +02:00
|
|
|
static struct sched_entity *pick_next_entity(struct cfs_rq *cfs_rq)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
2007-10-15 17:00:13 +02:00
|
|
|
struct sched_entity *se = NULL;
|
2007-07-09 18:51:58 +02:00
|
|
|
|
2007-10-15 17:00:13 +02:00
|
|
|
if (first_fair(cfs_rq)) {
|
|
|
|
se = __pick_next_entity(cfs_rq);
|
2008-03-14 21:12:12 +01:00
|
|
|
se = pick_next(cfs_rq, se);
|
2007-10-15 17:00:13 +02:00
|
|
|
set_next_entity(cfs_rq, se);
|
|
|
|
}
|
2007-07-09 18:51:58 +02:00
|
|
|
|
|
|
|
return se;
|
|
|
|
}
|
|
|
|
|
2007-08-09 11:16:48 +02:00
|
|
|
static void put_prev_entity(struct cfs_rq *cfs_rq, struct sched_entity *prev)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
|
|
|
/*
|
|
|
|
* If still on the runqueue then deactivate_task()
|
|
|
|
* was not called and update_curr() has to be done:
|
|
|
|
*/
|
|
|
|
if (prev->on_rq)
|
2007-08-09 11:16:47 +02:00
|
|
|
update_curr(cfs_rq);
|
2007-07-09 18:51:58 +02:00
|
|
|
|
2007-10-15 17:00:10 +02:00
|
|
|
check_spread(cfs_rq, prev);
|
2007-10-15 17:00:07 +02:00
|
|
|
if (prev->on_rq) {
|
2007-08-09 11:16:47 +02:00
|
|
|
update_stats_wait_start(cfs_rq, prev);
|
2007-10-15 17:00:07 +02:00
|
|
|
/* Put 'current' back into the tree. */
|
|
|
|
__enqueue_entity(cfs_rq, prev);
|
|
|
|
}
|
2007-10-15 17:00:03 +02:00
|
|
|
cfs_rq->curr = NULL;
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
|
2008-01-25 21:08:29 +01:00
|
|
|
static void
|
|
|
|
entity_tick(struct cfs_rq *cfs_rq, struct sched_entity *curr, int queued)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
|
|
|
/*
|
2007-10-15 17:00:07 +02:00
|
|
|
* Update run-time statistics of the 'current'.
|
2007-07-09 18:51:58 +02:00
|
|
|
*/
|
2007-10-15 17:00:07 +02:00
|
|
|
update_curr(cfs_rq);
|
2007-07-09 18:51:58 +02:00
|
|
|
|
2008-01-25 21:08:29 +01:00
|
|
|
#ifdef CONFIG_SCHED_HRTICK
|
|
|
|
/*
|
|
|
|
* queued ticks are scheduled to match the slice, so don't bother
|
|
|
|
* validating it and just reschedule.
|
|
|
|
*/
|
2008-04-25 03:17:55 +02:00
|
|
|
if (queued) {
|
|
|
|
resched_task(rq_of(cfs_rq)->curr);
|
|
|
|
return;
|
|
|
|
}
|
2008-01-25 21:08:29 +01:00
|
|
|
/*
|
|
|
|
* don't let the period tick interfere with the hrtick preemption
|
|
|
|
*/
|
|
|
|
if (!sched_feat(DOUBLE_TICK) &&
|
|
|
|
hrtimer_active(&rq_of(cfs_rq)->hrtick_timer))
|
|
|
|
return;
|
|
|
|
#endif
|
|
|
|
|
2007-10-15 17:00:14 +02:00
|
|
|
if (cfs_rq->nr_running > 1 || !sched_feat(WAKEUP_PREEMPT))
|
2007-10-15 17:00:05 +02:00
|
|
|
check_preempt_tick(cfs_rq, curr);
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/**************************************************
|
|
|
|
* CFS operations on tasks:
|
|
|
|
*/
|
|
|
|
|
2008-01-25 21:08:29 +01:00
|
|
|
#ifdef CONFIG_SCHED_HRTICK
|
|
|
|
static void hrtick_start_fair(struct rq *rq, struct task_struct *p)
|
|
|
|
{
|
|
|
|
struct sched_entity *se = &p->se;
|
|
|
|
struct cfs_rq *cfs_rq = cfs_rq_of(se);
|
|
|
|
|
|
|
|
WARN_ON(task_rq(p) != rq);
|
|
|
|
|
|
|
|
if (hrtick_enabled(rq) && cfs_rq->nr_running > 1) {
|
|
|
|
u64 slice = sched_slice(cfs_rq, se);
|
|
|
|
u64 ran = se->sum_exec_runtime - se->prev_sum_exec_runtime;
|
|
|
|
s64 delta = slice - ran;
|
|
|
|
|
|
|
|
if (delta < 0) {
|
|
|
|
if (rq->curr == p)
|
|
|
|
resched_task(p);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Don't schedule slices shorter than 10000ns, that just
|
|
|
|
* doesn't make sense. Rely on vruntime for fairness.
|
|
|
|
*/
|
2008-07-18 18:01:23 +02:00
|
|
|
if (rq->curr != p)
|
2008-07-28 11:53:11 +02:00
|
|
|
delta = max_t(s64, 10000LL, delta);
|
2008-01-25 21:08:29 +01:00
|
|
|
|
2008-07-18 18:01:23 +02:00
|
|
|
hrtick_start(rq, delta);
|
2008-01-25 21:08:29 +01:00
|
|
|
}
|
|
|
|
}
|
2008-06-24 20:09:43 +02:00
|
|
|
#else /* !CONFIG_SCHED_HRTICK */
|
2008-01-25 21:08:29 +01:00
|
|
|
static inline void
|
|
|
|
hrtick_start_fair(struct rq *rq, struct task_struct *p)
|
|
|
|
{
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2007-07-09 18:51:58 +02:00
|
|
|
/*
|
|
|
|
* The enqueue_task method is called before nr_running is
|
|
|
|
* increased. Here we update the fair scheduling stats and
|
|
|
|
* then put the task into the rbtree:
|
|
|
|
*/
|
2007-08-09 11:16:48 +02:00
|
|
|
static void enqueue_task_fair(struct rq *rq, struct task_struct *p, int wakeup)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
|
|
|
struct cfs_rq *cfs_rq;
|
2008-02-25 17:34:02 +01:00
|
|
|
struct sched_entity *se = &p->se;
|
2007-07-09 18:51:58 +02:00
|
|
|
|
|
|
|
for_each_sched_entity(se) {
|
2008-02-25 17:34:02 +01:00
|
|
|
if (se->on_rq)
|
2007-07-09 18:51:58 +02:00
|
|
|
break;
|
|
|
|
cfs_rq = cfs_rq_of(se);
|
2007-10-15 17:00:08 +02:00
|
|
|
enqueue_entity(cfs_rq, se, wakeup);
|
2007-10-15 17:00:12 +02:00
|
|
|
wakeup = 1;
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
2008-01-25 21:08:29 +01:00
|
|
|
|
|
|
|
hrtick_start_fair(rq, rq->curr);
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* The dequeue_task method is called before nr_running is
|
|
|
|
* decreased. We remove the task from the rbtree and
|
|
|
|
* update the fair scheduling stats:
|
|
|
|
*/
|
2007-08-09 11:16:48 +02:00
|
|
|
static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int sleep)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
|
|
|
struct cfs_rq *cfs_rq;
|
2008-02-25 17:34:02 +01:00
|
|
|
struct sched_entity *se = &p->se;
|
2007-07-09 18:51:58 +02:00
|
|
|
|
|
|
|
for_each_sched_entity(se) {
|
|
|
|
cfs_rq = cfs_rq_of(se);
|
2007-08-09 11:16:48 +02:00
|
|
|
dequeue_entity(cfs_rq, se, sleep);
|
2007-07-09 18:51:58 +02:00
|
|
|
/* Don't dequeue parent if it has other entities besides us */
|
2008-02-25 17:34:02 +01:00
|
|
|
if (cfs_rq->load.weight)
|
2007-07-09 18:51:58 +02:00
|
|
|
break;
|
2007-10-15 17:00:12 +02:00
|
|
|
sleep = 1;
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
2008-01-25 21:08:29 +01:00
|
|
|
|
|
|
|
hrtick_start_fair(rq, rq->curr);
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2007-09-19 23:34:46 +02:00
|
|
|
* sched_yield() support is very simple - we dequeue and enqueue.
|
|
|
|
*
|
|
|
|
* If compat_yield is turned on then we requeue to the end of the tree.
|
2007-07-09 18:51:58 +02:00
|
|
|
*/
|
2007-10-15 17:00:08 +02:00
|
|
|
static void yield_task_fair(struct rq *rq)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
2007-12-04 17:04:39 +01:00
|
|
|
struct task_struct *curr = rq->curr;
|
|
|
|
struct cfs_rq *cfs_rq = task_cfs_rq(curr);
|
|
|
|
struct sched_entity *rightmost, *se = &curr->se;
|
2007-07-09 18:51:58 +02:00
|
|
|
|
|
|
|
/*
|
2007-09-19 23:34:46 +02:00
|
|
|
* Are we the only task in the tree?
|
|
|
|
*/
|
|
|
|
if (unlikely(cfs_rq->nr_running == 1))
|
|
|
|
return;
|
|
|
|
|
2007-12-04 17:04:39 +01:00
|
|
|
if (likely(!sysctl_sched_compat_yield) && curr->policy != SCHED_BATCH) {
|
2008-05-03 18:29:28 +02:00
|
|
|
update_rq_clock(rq);
|
2007-09-19 23:34:46 +02:00
|
|
|
/*
|
2007-10-15 17:00:13 +02:00
|
|
|
* Update run-time statistics of the 'current'.
|
2007-09-19 23:34:46 +02:00
|
|
|
*/
|
2007-10-15 17:00:12 +02:00
|
|
|
update_curr(cfs_rq);
|
2007-09-19 23:34:46 +02:00
|
|
|
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* Find the rightmost entry in the rbtree:
|
2007-07-09 18:51:58 +02:00
|
|
|
*/
|
2007-10-15 17:00:12 +02:00
|
|
|
rightmost = __pick_last_entity(cfs_rq);
|
2007-09-19 23:34:46 +02:00
|
|
|
/*
|
|
|
|
* Already in the rightmost position?
|
|
|
|
*/
|
2008-02-18 13:39:37 +01:00
|
|
|
if (unlikely(!rightmost || rightmost->vruntime < se->vruntime))
|
2007-09-19 23:34:46 +02:00
|
|
|
return;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Minimally necessary key value to be last in the tree:
|
2007-10-15 17:00:12 +02:00
|
|
|
* Upon rescheduling, sched_class::put_prev_task() will place
|
|
|
|
* 'current' within the tree based on its new key value.
|
2007-09-19 23:34:46 +02:00
|
|
|
*/
|
2007-10-15 17:00:07 +02:00
|
|
|
se->vruntime = rightmost->vruntime + 1;
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
|
2008-01-25 21:08:09 +01:00
|
|
|
/*
|
|
|
|
* wake_idle() will wake a task on an idle cpu if task->cpu is
|
|
|
|
* not idle and an idle cpu is available. The span of cpus to
|
|
|
|
* search starts with cpus closest then further out as needed,
|
|
|
|
* so we always favor a closer, idle cpu.
|
2008-07-15 13:43:49 +02:00
|
|
|
* Domains may include CPUs that are not usable for migration,
|
|
|
|
* hence we need to mask them out (cpu_active_map)
|
2008-01-25 21:08:09 +01:00
|
|
|
*
|
|
|
|
* Returns the CPU we should wake onto.
|
|
|
|
*/
|
|
|
|
#if defined(ARCH_HAS_SCHED_WAKE_IDLE)
|
|
|
|
static int wake_idle(int cpu, struct task_struct *p)
|
|
|
|
{
|
|
|
|
cpumask_t tmp;
|
|
|
|
struct sched_domain *sd;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If it is idle, then it is the best cpu to run this task.
|
|
|
|
*
|
|
|
|
* This cpu is also the best, if it has more than one task already.
|
|
|
|
* Siblings must be also busy(in most cases) as they didn't already
|
|
|
|
* pickup the extra load from this cpu and hence we need not check
|
|
|
|
* sibling runqueue info. This will avoid the checks and cache miss
|
|
|
|
* penalities associated with that.
|
|
|
|
*/
|
2008-04-28 18:40:01 +02:00
|
|
|
if (idle_cpu(cpu) || cpu_rq(cpu)->cfs.nr_running > 1)
|
2008-01-25 21:08:09 +01:00
|
|
|
return cpu;
|
|
|
|
|
|
|
|
for_each_domain(cpu, sd) {
|
2008-04-15 07:04:23 +02:00
|
|
|
if ((sd->flags & SD_WAKE_IDLE)
|
|
|
|
|| ((sd->flags & SD_WAKE_IDLE_FAR)
|
|
|
|
&& !task_hot(p, task_rq(p)->clock, sd))) {
|
2008-01-25 21:08:09 +01:00
|
|
|
cpus_and(tmp, sd->span, p->cpus_allowed);
|
2008-07-15 13:43:49 +02:00
|
|
|
cpus_and(tmp, tmp, cpu_active_map);
|
2008-05-12 21:21:13 +02:00
|
|
|
for_each_cpu_mask_nr(i, tmp) {
|
2008-01-25 21:08:09 +01:00
|
|
|
if (idle_cpu(i)) {
|
|
|
|
if (i != task_cpu(p)) {
|
|
|
|
schedstat_inc(p,
|
|
|
|
se.nr_wakeups_idle);
|
|
|
|
}
|
|
|
|
return i;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return cpu;
|
|
|
|
}
|
2008-06-24 20:09:43 +02:00
|
|
|
#else /* !ARCH_HAS_SCHED_WAKE_IDLE*/
|
2008-01-25 21:08:09 +01:00
|
|
|
static inline int wake_idle(int cpu, struct task_struct *p)
|
|
|
|
{
|
|
|
|
return cpu;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
|
|
|
#ifdef CONFIG_SMP
|
2008-03-16 20:36:10 +01:00
|
|
|
|
2008-03-19 01:42:00 +01:00
|
|
|
static const struct sched_class fair_sched_class;
|
|
|
|
|
2008-06-27 13:41:27 +02:00
|
|
|
#ifdef CONFIG_FAIR_GROUP_SCHED
|
2008-06-27 13:41:39 +02:00
|
|
|
/*
|
|
|
|
* effective_load() calculates the load change as seen from the root_task_group
|
|
|
|
*
|
|
|
|
* Adding load to a group doesn't make a group heavier, but can cause movement
|
|
|
|
* of group shares between cpus. Assuming the shares were perfectly aligned one
|
|
|
|
* can calculate the shift in shares.
|
|
|
|
*
|
|
|
|
* The problem is that perfectly aligning the shares is rather expensive, hence
|
|
|
|
* we try to avoid doing that too often - see update_shares(), which ratelimits
|
|
|
|
* this change.
|
|
|
|
*
|
|
|
|
* We compensate this by not only taking the current delta into account, but
|
|
|
|
* also considering the delta between when the shares were last adjusted and
|
|
|
|
* now.
|
|
|
|
*
|
|
|
|
* We still saw a performance dip, some tracing learned us that between
|
|
|
|
* cgroup:/ and cgroup:/foo balancing the number of affine wakeups increased
|
|
|
|
* significantly. Therefore try to bias the error in direction of failing
|
|
|
|
* the affine wakeup.
|
|
|
|
*
|
|
|
|
*/
|
2008-06-27 13:41:38 +02:00
|
|
|
static long effective_load(struct task_group *tg, int cpu,
|
|
|
|
long wl, long wg)
|
2008-06-27 13:41:27 +02:00
|
|
|
{
|
2008-06-27 13:41:30 +02:00
|
|
|
struct sched_entity *se = tg->se[cpu];
|
2008-06-27 13:41:38 +02:00
|
|
|
long more_w;
|
|
|
|
|
|
|
|
if (!tg->parent)
|
|
|
|
return wl;
|
|
|
|
|
2008-06-27 13:41:39 +02:00
|
|
|
/*
|
|
|
|
* By not taking the decrease of shares on the other cpu into
|
|
|
|
* account our error leans towards reducing the affine wakeups.
|
|
|
|
*/
|
|
|
|
if (!wl && sched_feat(ASYM_EFF_LOAD))
|
|
|
|
return wl;
|
|
|
|
|
2008-06-27 13:41:38 +02:00
|
|
|
/*
|
|
|
|
* Instead of using this increment, also add the difference
|
|
|
|
* between when the shares were last updated and now.
|
|
|
|
*/
|
|
|
|
more_w = se->my_q->load.weight - se->my_q->rq_weight;
|
|
|
|
wl += more_w;
|
|
|
|
wg += more_w;
|
2008-06-27 13:41:27 +02:00
|
|
|
|
2008-06-27 13:41:30 +02:00
|
|
|
for_each_sched_entity(se) {
|
|
|
|
#define D(n) (likely(n) ? (n) : 1)
|
|
|
|
|
2008-06-27 13:41:32 +02:00
|
|
|
long S, rw, s, a, b;
|
2008-06-27 13:41:30 +02:00
|
|
|
|
|
|
|
S = se->my_q->tg->shares;
|
|
|
|
s = se->my_q->shares;
|
2008-06-27 13:41:38 +02:00
|
|
|
rw = se->my_q->rq_weight;
|
2008-06-27 13:41:27 +02:00
|
|
|
|
2008-06-27 13:41:32 +02:00
|
|
|
a = S*(rw + wl);
|
|
|
|
b = S*rw + s*wg;
|
2008-06-27 13:41:30 +02:00
|
|
|
|
2008-06-27 13:41:32 +02:00
|
|
|
wl = s*(a-b)/D(b);
|
sched: correct wakeup weight calculations
rw_i = {2, 4, 1, 0}
s_i = {2/7, 4/7, 1/7, 0}
wakeup on cpu0, weight=1
rw'_i = {3, 4, 1, 0}
s'_i = {3/8, 4/8, 1/8, 0}
s_0 = S * rw_0 / \Sum rw_j ->
\Sum rw_j = S*rw_0/s_0 = 1*2*7/2 = 7 (correct)
s'_0 = S * (rw_0 + 1) / (\Sum rw_j + 1) =
1 * (2+1) / (7+1) = 3/8 (correct
so we find that adding 1 to cpu0 gains 5/56 in weight
if say the other cpu were, cpu1, we'd also have to calculate its 4/56 loss
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-06-27 13:41:37 +02:00
|
|
|
/*
|
|
|
|
* Assume the group is already running and will
|
|
|
|
* thus already be accounted for in the weight.
|
|
|
|
*
|
|
|
|
* That is, moving shares between CPUs, does not
|
|
|
|
* alter the group weight.
|
|
|
|
*/
|
2008-06-27 13:41:30 +02:00
|
|
|
wg = 0;
|
|
|
|
#undef D
|
|
|
|
}
|
2008-06-27 13:41:27 +02:00
|
|
|
|
2008-06-27 13:41:30 +02:00
|
|
|
return wl;
|
2008-06-27 13:41:27 +02:00
|
|
|
}
|
2008-06-27 13:41:30 +02:00
|
|
|
|
2008-06-27 13:41:27 +02:00
|
|
|
#else
|
2008-06-27 13:41:30 +02:00
|
|
|
|
sched: correct wakeup weight calculations
rw_i = {2, 4, 1, 0}
s_i = {2/7, 4/7, 1/7, 0}
wakeup on cpu0, weight=1
rw'_i = {3, 4, 1, 0}
s'_i = {3/8, 4/8, 1/8, 0}
s_0 = S * rw_0 / \Sum rw_j ->
\Sum rw_j = S*rw_0/s_0 = 1*2*7/2 = 7 (correct)
s'_0 = S * (rw_0 + 1) / (\Sum rw_j + 1) =
1 * (2+1) / (7+1) = 3/8 (correct
so we find that adding 1 to cpu0 gains 5/56 in weight
if say the other cpu were, cpu1, we'd also have to calculate its 4/56 loss
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-06-27 13:41:37 +02:00
|
|
|
static inline unsigned long effective_load(struct task_group *tg, int cpu,
|
|
|
|
unsigned long wl, unsigned long wg)
|
2008-06-27 13:41:30 +02:00
|
|
|
{
|
sched: correct wakeup weight calculations
rw_i = {2, 4, 1, 0}
s_i = {2/7, 4/7, 1/7, 0}
wakeup on cpu0, weight=1
rw'_i = {3, 4, 1, 0}
s'_i = {3/8, 4/8, 1/8, 0}
s_0 = S * rw_0 / \Sum rw_j ->
\Sum rw_j = S*rw_0/s_0 = 1*2*7/2 = 7 (correct)
s'_0 = S * (rw_0 + 1) / (\Sum rw_j + 1) =
1 * (2+1) / (7+1) = 3/8 (correct
so we find that adding 1 to cpu0 gains 5/56 in weight
if say the other cpu were, cpu1, we'd also have to calculate its 4/56 loss
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-06-27 13:41:37 +02:00
|
|
|
return wl;
|
2008-06-27 13:41:27 +02:00
|
|
|
}
|
2008-06-27 13:41:30 +02:00
|
|
|
|
2008-06-27 13:41:27 +02:00
|
|
|
#endif
|
|
|
|
|
2008-03-16 20:36:10 +01:00
|
|
|
static int
|
2008-03-19 01:42:00 +01:00
|
|
|
wake_affine(struct rq *rq, struct sched_domain *this_sd, struct rq *this_rq,
|
|
|
|
struct task_struct *p, int prev_cpu, int this_cpu, int sync,
|
|
|
|
int idx, unsigned long load, unsigned long this_load,
|
2008-03-16 20:36:10 +01:00
|
|
|
unsigned int imbalance)
|
|
|
|
{
|
2008-03-19 01:42:00 +01:00
|
|
|
struct task_struct *curr = this_rq->curr;
|
sched: correct wakeup weight calculations
rw_i = {2, 4, 1, 0}
s_i = {2/7, 4/7, 1/7, 0}
wakeup on cpu0, weight=1
rw'_i = {3, 4, 1, 0}
s'_i = {3/8, 4/8, 1/8, 0}
s_0 = S * rw_0 / \Sum rw_j ->
\Sum rw_j = S*rw_0/s_0 = 1*2*7/2 = 7 (correct)
s'_0 = S * (rw_0 + 1) / (\Sum rw_j + 1) =
1 * (2+1) / (7+1) = 3/8 (correct
so we find that adding 1 to cpu0 gains 5/56 in weight
if say the other cpu were, cpu1, we'd also have to calculate its 4/56 loss
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-06-27 13:41:37 +02:00
|
|
|
struct task_group *tg;
|
2008-03-16 20:36:10 +01:00
|
|
|
unsigned long tl = this_load;
|
|
|
|
unsigned long tl_per_task;
|
sched: correct wakeup weight calculations
rw_i = {2, 4, 1, 0}
s_i = {2/7, 4/7, 1/7, 0}
wakeup on cpu0, weight=1
rw'_i = {3, 4, 1, 0}
s'_i = {3/8, 4/8, 1/8, 0}
s_0 = S * rw_0 / \Sum rw_j ->
\Sum rw_j = S*rw_0/s_0 = 1*2*7/2 = 7 (correct)
s'_0 = S * (rw_0 + 1) / (\Sum rw_j + 1) =
1 * (2+1) / (7+1) = 3/8 (correct
so we find that adding 1 to cpu0 gains 5/56 in weight
if say the other cpu were, cpu1, we'd also have to calculate its 4/56 loss
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-06-27 13:41:37 +02:00
|
|
|
unsigned long weight;
|
2008-05-29 11:11:41 +02:00
|
|
|
int balanced;
|
2008-03-16 20:36:10 +01:00
|
|
|
|
2008-05-29 11:11:41 +02:00
|
|
|
if (!(this_sd->flags & SD_WAKE_AFFINE) || !sched_feat(AFFINE_WAKEUPS))
|
2008-03-16 20:36:10 +01:00
|
|
|
return 0;
|
|
|
|
|
2008-05-29 11:11:41 +02:00
|
|
|
/*
|
|
|
|
* If sync wakeup then subtract the (maximum possible)
|
|
|
|
* effect of the currently running task from the load
|
|
|
|
* of the current CPU:
|
|
|
|
*/
|
sched: correct wakeup weight calculations
rw_i = {2, 4, 1, 0}
s_i = {2/7, 4/7, 1/7, 0}
wakeup on cpu0, weight=1
rw'_i = {3, 4, 1, 0}
s'_i = {3/8, 4/8, 1/8, 0}
s_0 = S * rw_0 / \Sum rw_j ->
\Sum rw_j = S*rw_0/s_0 = 1*2*7/2 = 7 (correct)
s'_0 = S * (rw_0 + 1) / (\Sum rw_j + 1) =
1 * (2+1) / (7+1) = 3/8 (correct
so we find that adding 1 to cpu0 gains 5/56 in weight
if say the other cpu were, cpu1, we'd also have to calculate its 4/56 loss
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-06-27 13:41:37 +02:00
|
|
|
if (sync) {
|
|
|
|
tg = task_group(current);
|
|
|
|
weight = current->se.load.weight;
|
|
|
|
|
|
|
|
tl += effective_load(tg, this_cpu, -weight, -weight);
|
|
|
|
load += effective_load(tg, prev_cpu, 0, -weight);
|
|
|
|
}
|
2008-05-29 11:11:41 +02:00
|
|
|
|
sched: correct wakeup weight calculations
rw_i = {2, 4, 1, 0}
s_i = {2/7, 4/7, 1/7, 0}
wakeup on cpu0, weight=1
rw'_i = {3, 4, 1, 0}
s'_i = {3/8, 4/8, 1/8, 0}
s_0 = S * rw_0 / \Sum rw_j ->
\Sum rw_j = S*rw_0/s_0 = 1*2*7/2 = 7 (correct)
s'_0 = S * (rw_0 + 1) / (\Sum rw_j + 1) =
1 * (2+1) / (7+1) = 3/8 (correct
so we find that adding 1 to cpu0 gains 5/56 in weight
if say the other cpu were, cpu1, we'd also have to calculate its 4/56 loss
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-06-27 13:41:37 +02:00
|
|
|
tg = task_group(p);
|
|
|
|
weight = p->se.load.weight;
|
2008-05-29 11:11:41 +02:00
|
|
|
|
sched: correct wakeup weight calculations
rw_i = {2, 4, 1, 0}
s_i = {2/7, 4/7, 1/7, 0}
wakeup on cpu0, weight=1
rw'_i = {3, 4, 1, 0}
s'_i = {3/8, 4/8, 1/8, 0}
s_0 = S * rw_0 / \Sum rw_j ->
\Sum rw_j = S*rw_0/s_0 = 1*2*7/2 = 7 (correct)
s'_0 = S * (rw_0 + 1) / (\Sum rw_j + 1) =
1 * (2+1) / (7+1) = 3/8 (correct
so we find that adding 1 to cpu0 gains 5/56 in weight
if say the other cpu were, cpu1, we'd also have to calculate its 4/56 loss
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-06-27 13:41:37 +02:00
|
|
|
balanced = 100*(tl + effective_load(tg, this_cpu, weight, weight)) <=
|
|
|
|
imbalance*(load + effective_load(tg, prev_cpu, 0, weight));
|
2008-05-29 11:11:41 +02:00
|
|
|
|
2008-03-16 20:36:10 +01:00
|
|
|
/*
|
2008-03-19 01:42:00 +01:00
|
|
|
* If the currently running task will sleep within
|
|
|
|
* a reasonable amount of time then attract this newly
|
|
|
|
* woken task:
|
2008-03-16 20:36:10 +01:00
|
|
|
*/
|
2008-06-27 22:30:00 +02:00
|
|
|
if (sync && balanced) {
|
2008-03-19 01:42:00 +01:00
|
|
|
if (curr->se.avg_overlap < sysctl_sched_migration_cost &&
|
2008-06-27 22:30:00 +02:00
|
|
|
p->se.avg_overlap < sysctl_sched_migration_cost)
|
2008-03-19 01:42:00 +01:00
|
|
|
return 1;
|
|
|
|
}
|
2008-03-16 20:36:10 +01:00
|
|
|
|
|
|
|
schedstat_inc(p, se.nr_wakeups_affine_attempts);
|
|
|
|
tl_per_task = cpu_avg_load_per_task(this_cpu);
|
|
|
|
|
2008-03-16 20:56:26 +01:00
|
|
|
if ((tl <= load && tl + target_load(prev_cpu, idx) <= tl_per_task) ||
|
2008-05-29 11:11:41 +02:00
|
|
|
balanced) {
|
2008-03-16 20:36:10 +01:00
|
|
|
/*
|
|
|
|
* This domain has SD_WAKE_AFFINE and
|
|
|
|
* p is cache cold in this domain, and
|
|
|
|
* there is no bad imbalance.
|
|
|
|
*/
|
|
|
|
schedstat_inc(this_sd, ttwu_move_affine);
|
|
|
|
schedstat_inc(p, se.nr_wakeups_affine);
|
|
|
|
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-01-25 21:08:09 +01:00
|
|
|
static int select_task_rq_fair(struct task_struct *p, int sync)
|
|
|
|
{
|
|
|
|
struct sched_domain *sd, *this_sd = NULL;
|
2008-03-16 20:56:26 +01:00
|
|
|
int prev_cpu, this_cpu, new_cpu;
|
2008-03-16 20:36:10 +01:00
|
|
|
unsigned long load, this_load;
|
2008-03-19 01:42:00 +01:00
|
|
|
struct rq *rq, *this_rq;
|
2008-03-16 20:36:10 +01:00
|
|
|
unsigned int imbalance;
|
|
|
|
int idx;
|
2008-01-25 21:08:09 +01:00
|
|
|
|
2008-03-16 20:56:26 +01:00
|
|
|
prev_cpu = task_cpu(p);
|
|
|
|
rq = task_rq(p);
|
|
|
|
this_cpu = smp_processor_id();
|
2008-03-19 01:42:00 +01:00
|
|
|
this_rq = cpu_rq(this_cpu);
|
2008-03-16 20:56:26 +01:00
|
|
|
new_cpu = prev_cpu;
|
2008-01-25 21:08:09 +01:00
|
|
|
|
2008-03-16 20:56:26 +01:00
|
|
|
/*
|
|
|
|
* 'this_sd' is the first domain that both
|
|
|
|
* this_cpu and prev_cpu are present in:
|
|
|
|
*/
|
2008-01-25 21:08:09 +01:00
|
|
|
for_each_domain(this_cpu, sd) {
|
2008-03-16 20:56:26 +01:00
|
|
|
if (cpu_isset(prev_cpu, sd->span)) {
|
2008-01-25 21:08:09 +01:00
|
|
|
this_sd = sd;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (unlikely(!cpu_isset(this_cpu, p->cpus_allowed)))
|
2008-03-16 21:21:47 +01:00
|
|
|
goto out;
|
2008-01-25 21:08:09 +01:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Check for affine wakeup and passive balancing possibilities.
|
|
|
|
*/
|
2008-03-16 20:36:10 +01:00
|
|
|
if (!this_sd)
|
2008-03-16 21:21:47 +01:00
|
|
|
goto out;
|
2008-01-25 21:08:09 +01:00
|
|
|
|
2008-03-16 20:36:10 +01:00
|
|
|
idx = this_sd->wake_idx;
|
|
|
|
|
|
|
|
imbalance = 100 + (this_sd->imbalance_pct - 100) / 2;
|
|
|
|
|
2008-03-16 20:56:26 +01:00
|
|
|
load = source_load(prev_cpu, idx);
|
2008-03-16 20:36:10 +01:00
|
|
|
this_load = target_load(this_cpu, idx);
|
|
|
|
|
2008-03-19 01:42:00 +01:00
|
|
|
if (wake_affine(rq, this_sd, this_rq, p, prev_cpu, this_cpu, sync, idx,
|
|
|
|
load, this_load, imbalance))
|
|
|
|
return this_cpu;
|
|
|
|
|
|
|
|
if (prev_cpu == this_cpu)
|
2008-03-16 21:21:47 +01:00
|
|
|
goto out;
|
2008-03-16 20:36:10 +01:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Start passive balancing when half the imbalance_pct
|
|
|
|
* limit is reached.
|
|
|
|
*/
|
|
|
|
if (this_sd->flags & SD_WAKE_BALANCE) {
|
|
|
|
if (imbalance*this_load <= 100*load) {
|
|
|
|
schedstat_inc(this_sd, ttwu_move_balance);
|
|
|
|
schedstat_inc(p, se.nr_wakeups_passive);
|
2008-03-19 01:42:00 +01:00
|
|
|
return this_cpu;
|
2008-01-25 21:08:09 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-03-16 21:21:47 +01:00
|
|
|
out:
|
2008-01-25 21:08:09 +01:00
|
|
|
return wake_idle(new_cpu, p);
|
|
|
|
}
|
|
|
|
#endif /* CONFIG_SMP */
|
|
|
|
|
2008-04-19 19:44:57 +02:00
|
|
|
static unsigned long wakeup_gran(struct sched_entity *se)
|
|
|
|
{
|
|
|
|
unsigned long gran = sysctl_sched_wakeup_granularity;
|
|
|
|
|
|
|
|
/*
|
2008-06-27 13:41:11 +02:00
|
|
|
* More easily preempt - nice tasks, while not making it harder for
|
|
|
|
* + nice tasks.
|
2008-04-19 19:44:57 +02:00
|
|
|
*/
|
2008-06-27 13:41:12 +02:00
|
|
|
if (sched_feat(ASYM_GRAN))
|
|
|
|
gran = calc_delta_asym(sysctl_sched_wakeup_granularity, se);
|
|
|
|
else
|
|
|
|
gran = calc_delta_fair(sysctl_sched_wakeup_granularity, se);
|
2008-04-19 19:44:57 +02:00
|
|
|
|
|
|
|
return gran;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Should 'se' preempt 'curr'.
|
|
|
|
*
|
|
|
|
* |s1
|
|
|
|
* |s2
|
|
|
|
* |s3
|
|
|
|
* g
|
|
|
|
* |<--->|c
|
|
|
|
*
|
|
|
|
* w(c, s1) = -1
|
|
|
|
* w(c, s2) = 0
|
|
|
|
* w(c, s3) = 1
|
|
|
|
*
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
wakeup_preempt_entity(struct sched_entity *curr, struct sched_entity *se)
|
|
|
|
{
|
|
|
|
s64 gran, vdiff = curr->vruntime - se->vruntime;
|
|
|
|
|
|
|
|
if (vdiff < 0)
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
gran = wakeup_gran(curr);
|
|
|
|
if (vdiff > gran)
|
|
|
|
return 1;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
2008-01-25 21:08:09 +01:00
|
|
|
|
2008-04-19 19:44:59 +02:00
|
|
|
/* return depth at which a sched entity is present in the hierarchy */
|
|
|
|
static inline int depth_se(struct sched_entity *se)
|
|
|
|
{
|
|
|
|
int depth = 0;
|
|
|
|
|
|
|
|
for_each_sched_entity(se)
|
|
|
|
depth++;
|
|
|
|
|
|
|
|
return depth;
|
|
|
|
}
|
|
|
|
|
2007-07-09 18:51:58 +02:00
|
|
|
/*
|
|
|
|
* Preempt the current task with a newly woken task if needed:
|
|
|
|
*/
|
2008-09-20 23:38:02 +02:00
|
|
|
static void check_preempt_wakeup(struct rq *rq, struct task_struct *p, int sync)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
|
|
|
struct task_struct *curr = rq->curr;
|
2007-10-15 17:00:12 +02:00
|
|
|
struct cfs_rq *cfs_rq = task_cfs_rq(curr);
|
2007-10-15 17:00:12 +02:00
|
|
|
struct sched_entity *se = &curr->se, *pse = &p->se;
|
2008-04-19 19:44:59 +02:00
|
|
|
int se_depth, pse_depth;
|
2007-07-09 18:51:58 +02:00
|
|
|
|
|
|
|
if (unlikely(rt_prio(p->prio))) {
|
2007-08-09 11:16:47 +02:00
|
|
|
update_rq_clock(rq);
|
2007-08-09 11:16:47 +02:00
|
|
|
update_curr(cfs_rq);
|
2007-07-09 18:51:58 +02:00
|
|
|
resched_task(curr);
|
|
|
|
return;
|
|
|
|
}
|
2008-03-14 21:12:12 +01:00
|
|
|
|
2008-03-19 01:42:00 +01:00
|
|
|
if (unlikely(se == pse))
|
|
|
|
return;
|
|
|
|
|
2008-08-28 11:12:49 +02:00
|
|
|
/*
|
|
|
|
* We can come here with TIF_NEED_RESCHED already set from new task
|
|
|
|
* wake up path.
|
|
|
|
*/
|
|
|
|
if (test_tsk_need_resched(curr))
|
|
|
|
return;
|
|
|
|
|
2008-03-14 21:12:12 +01:00
|
|
|
cfs_rq_of(pse)->next = pse;
|
|
|
|
|
2007-10-15 17:00:18 +02:00
|
|
|
/*
|
|
|
|
* Batch tasks do not preempt (their preemption is driven by
|
|
|
|
* the tick):
|
|
|
|
*/
|
|
|
|
if (unlikely(p->policy == SCHED_BATCH))
|
|
|
|
return;
|
2007-07-09 18:51:58 +02:00
|
|
|
|
2007-11-09 22:39:39 +01:00
|
|
|
if (!sched_feat(WAKEUP_PREEMPT))
|
|
|
|
return;
|
2007-10-15 17:00:12 +02:00
|
|
|
|
2008-09-20 23:38:02 +02:00
|
|
|
if (sched_feat(WAKEUP_OVERLAP) && sync &&
|
|
|
|
se->avg_overlap < sysctl_sched_migration_cost &&
|
|
|
|
pse->avg_overlap < sysctl_sched_migration_cost) {
|
|
|
|
resched_task(curr);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2008-04-19 19:44:59 +02:00
|
|
|
/*
|
|
|
|
* preemption test can be made between sibling entities who are in the
|
|
|
|
* same cfs_rq i.e who have a common parent. Walk up the hierarchy of
|
|
|
|
* both tasks until we find their ancestors who are siblings of common
|
|
|
|
* parent.
|
|
|
|
*/
|
|
|
|
|
|
|
|
/* First walk up until both entities are at same depth */
|
|
|
|
se_depth = depth_se(se);
|
|
|
|
pse_depth = depth_se(pse);
|
|
|
|
|
|
|
|
while (se_depth > pse_depth) {
|
|
|
|
se_depth--;
|
|
|
|
se = parent_entity(se);
|
|
|
|
}
|
|
|
|
|
|
|
|
while (pse_depth > se_depth) {
|
|
|
|
pse_depth--;
|
|
|
|
pse = parent_entity(pse);
|
|
|
|
}
|
|
|
|
|
2007-11-09 22:39:39 +01:00
|
|
|
while (!is_same_group(se, pse)) {
|
|
|
|
se = parent_entity(se);
|
|
|
|
pse = parent_entity(pse);
|
2007-10-15 17:00:14 +02:00
|
|
|
}
|
2007-11-09 22:39:39 +01:00
|
|
|
|
2008-04-19 19:44:57 +02:00
|
|
|
if (wakeup_preempt_entity(se, pse) == 1)
|
2007-11-09 22:39:39 +01:00
|
|
|
resched_task(curr);
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
|
2007-08-09 11:16:48 +02:00
|
|
|
static struct task_struct *pick_next_task_fair(struct rq *rq)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
2008-01-25 21:08:29 +01:00
|
|
|
struct task_struct *p;
|
2007-07-09 18:51:58 +02:00
|
|
|
struct cfs_rq *cfs_rq = &rq->cfs;
|
|
|
|
struct sched_entity *se;
|
|
|
|
|
|
|
|
if (unlikely(!cfs_rq->nr_running))
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
do {
|
2007-08-09 11:16:48 +02:00
|
|
|
se = pick_next_entity(cfs_rq);
|
2007-07-09 18:51:58 +02:00
|
|
|
cfs_rq = group_cfs_rq(se);
|
|
|
|
} while (cfs_rq);
|
|
|
|
|
2008-01-25 21:08:29 +01:00
|
|
|
p = task_of(se);
|
|
|
|
hrtick_start_fair(rq, p);
|
|
|
|
|
|
|
|
return p;
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Account for a descheduled task:
|
|
|
|
*/
|
2007-08-09 11:16:49 +02:00
|
|
|
static void put_prev_task_fair(struct rq *rq, struct task_struct *prev)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
|
|
|
struct sched_entity *se = &prev->se;
|
|
|
|
struct cfs_rq *cfs_rq;
|
|
|
|
|
|
|
|
for_each_sched_entity(se) {
|
|
|
|
cfs_rq = cfs_rq_of(se);
|
2007-08-09 11:16:48 +02:00
|
|
|
put_prev_entity(cfs_rq, se);
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2007-10-24 18:23:51 +02:00
|
|
|
#ifdef CONFIG_SMP
|
2007-07-09 18:51:58 +02:00
|
|
|
/**************************************************
|
|
|
|
* Fair scheduling class load-balancing methods:
|
|
|
|
*/
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Load-balancing iterator. Note: while the runqueue stays locked
|
|
|
|
* during the whole iteration, the current task might be
|
|
|
|
* dequeued so the iterator has to be dequeue-safe. Here we
|
|
|
|
* achieve that by always pre-iterating before returning
|
|
|
|
* the current task:
|
|
|
|
*/
|
2007-10-15 17:00:13 +02:00
|
|
|
static struct task_struct *
|
2008-04-19 19:45:00 +02:00
|
|
|
__load_balance_iterator(struct cfs_rq *cfs_rq, struct list_head *next)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
2008-04-19 19:44:59 +02:00
|
|
|
struct task_struct *p = NULL;
|
|
|
|
struct sched_entity *se;
|
2007-07-09 18:51:58 +02:00
|
|
|
|
2008-08-11 13:32:02 +02:00
|
|
|
if (next == &cfs_rq->tasks)
|
|
|
|
return NULL;
|
|
|
|
|
|
|
|
/* Skip over entities that are not tasks */
|
|
|
|
do {
|
2008-04-19 19:45:00 +02:00
|
|
|
se = list_entry(next, struct sched_entity, group_node);
|
|
|
|
next = next->next;
|
2008-08-11 13:32:02 +02:00
|
|
|
} while (next != &cfs_rq->tasks && !entity_is_task(se));
|
2008-04-19 19:44:59 +02:00
|
|
|
|
2008-09-06 11:20:23 +02:00
|
|
|
if (next == &cfs_rq->tasks && !entity_is_task(se))
|
2008-08-11 13:32:02 +02:00
|
|
|
return NULL;
|
2008-04-19 19:45:00 +02:00
|
|
|
|
|
|
|
cfs_rq->balance_iterator = next;
|
2008-08-11 13:32:02 +02:00
|
|
|
|
|
|
|
if (entity_is_task(se))
|
|
|
|
p = task_of(se);
|
|
|
|
|
2007-07-09 18:51:58 +02:00
|
|
|
return p;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct task_struct *load_balance_start_fair(void *arg)
|
|
|
|
{
|
|
|
|
struct cfs_rq *cfs_rq = arg;
|
|
|
|
|
2008-04-19 19:45:00 +02:00
|
|
|
return __load_balance_iterator(cfs_rq, cfs_rq->tasks.next);
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static struct task_struct *load_balance_next_fair(void *arg)
|
|
|
|
{
|
|
|
|
struct cfs_rq *cfs_rq = arg;
|
|
|
|
|
2008-04-19 19:45:00 +02:00
|
|
|
return __load_balance_iterator(cfs_rq, cfs_rq->balance_iterator);
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
|
2008-06-27 13:41:14 +02:00
|
|
|
static unsigned long
|
|
|
|
__load_balance_fair(struct rq *this_rq, int this_cpu, struct rq *busiest,
|
|
|
|
unsigned long max_load_move, struct sched_domain *sd,
|
|
|
|
enum cpu_idle_type idle, int *all_pinned, int *this_best_prio,
|
|
|
|
struct cfs_rq *cfs_rq)
|
2008-02-25 17:34:02 +01:00
|
|
|
{
|
2008-06-27 13:41:14 +02:00
|
|
|
struct rq_iterator cfs_rq_iterator;
|
2008-02-25 17:34:02 +01:00
|
|
|
|
2008-06-27 13:41:14 +02:00
|
|
|
cfs_rq_iterator.start = load_balance_start_fair;
|
|
|
|
cfs_rq_iterator.next = load_balance_next_fair;
|
|
|
|
cfs_rq_iterator.arg = cfs_rq;
|
2008-02-25 17:34:02 +01:00
|
|
|
|
2008-06-27 13:41:14 +02:00
|
|
|
return balance_tasks(this_rq, this_cpu, busiest,
|
|
|
|
max_load_move, sd, idle, all_pinned,
|
|
|
|
this_best_prio, &cfs_rq_iterator);
|
2008-02-25 17:34:02 +01:00
|
|
|
}
|
|
|
|
|
2008-06-27 13:41:14 +02:00
|
|
|
#ifdef CONFIG_FAIR_GROUP_SCHED
|
sched: simplify move_tasks()
The move_tasks() function is currently multiplexed with two distinct
capabilities:
1. attempt to move a specified amount of weighted load from one run
queue to another; and
2. attempt to move a specified number of tasks from one run queue to
another.
The first of these capabilities is used in two places, load_balance()
and load_balance_idle(), and in both of these cases the return value of
move_tasks() is used purely to decide if tasks/load were moved and no
notice of the actual number of tasks moved is taken.
The second capability is used in exactly one place,
active_load_balance(), to attempt to move exactly one task and, as
before, the return value is only used as an indicator of success or failure.
This multiplexing of sched_task() was introduced, by me, as part of the
smpnice patches and was motivated by the fact that the alternative, one
function to move specified load and one to move a single task, would
have led to two functions of roughly the same complexity as the old
move_tasks() (or the new balance_tasks()). However, the new modular
design of the new CFS scheduler allows a simpler solution to be adopted
and this patch addresses that solution by:
1. adding a new function, move_one_task(), to be used by
active_load_balance(); and
2. making move_tasks() a single purpose function that tries to move a
specified weighted load and returns 1 for success and 0 for failure.
One of the consequences of these changes is that neither move_one_task()
or the new move_tasks() care how many tasks sched_class.load_balance()
moves and this enables its interface to be simplified by returning the
amount of load moved as its result and removing the load_moved pointer
from the argument list. This helps simplify the new move_tasks() and
slightly reduces the amount of work done in each of
sched_class.load_balance()'s implementations.
Further simplification, e.g. changes to balance_tasks(), are possible
but (slightly) complicated by the special needs of load_balance_fair()
so I've left them to a later patch (if this one gets accepted).
NB Since move_tasks() gets called with two run queue locks held even
small reductions in overhead are worthwhile.
[ mingo@elte.hu ]
this change also reduces code size nicely:
text data bss dec hex filename
39216 3618 24 42858 a76a sched.o.before
39173 3618 24 42815 a73f sched.o.after
Signed-off-by: Peter Williams <pwil3058@bigpond.net.au>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2007-08-09 11:16:46 +02:00
|
|
|
static unsigned long
|
2007-07-09 18:51:58 +02:00
|
|
|
load_balance_fair(struct rq *this_rq, int this_cpu, struct rq *busiest,
|
2007-10-24 18:23:51 +02:00
|
|
|
unsigned long max_load_move,
|
2007-08-09 11:16:46 +02:00
|
|
|
struct sched_domain *sd, enum cpu_idle_type idle,
|
|
|
|
int *all_pinned, int *this_best_prio)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
|
|
|
long rem_load_move = max_load_move;
|
2008-06-27 13:41:14 +02:00
|
|
|
int busiest_cpu = cpu_of(busiest);
|
|
|
|
struct task_group *tg;
|
2008-04-19 19:45:00 +02:00
|
|
|
|
2008-06-27 13:41:14 +02:00
|
|
|
rcu_read_lock();
|
2008-06-27 13:41:23 +02:00
|
|
|
update_h_load(busiest_cpu);
|
2008-04-19 19:45:00 +02:00
|
|
|
|
2008-09-22 19:06:09 +02:00
|
|
|
list_for_each_entry_rcu(tg, &task_groups, list) {
|
2008-06-27 13:41:23 +02:00
|
|
|
struct cfs_rq *busiest_cfs_rq = tg->cfs_rq[busiest_cpu];
|
2008-06-27 13:41:29 +02:00
|
|
|
unsigned long busiest_h_load = busiest_cfs_rq->h_load;
|
|
|
|
unsigned long busiest_weight = busiest_cfs_rq->load.weight;
|
2008-06-27 13:41:36 +02:00
|
|
|
u64 rem_load, moved_load;
|
2008-04-19 19:45:00 +02:00
|
|
|
|
2008-06-27 13:41:14 +02:00
|
|
|
/*
|
|
|
|
* empty group
|
|
|
|
*/
|
2008-06-27 13:41:23 +02:00
|
|
|
if (!busiest_cfs_rq->task_weight)
|
2007-07-09 18:51:58 +02:00
|
|
|
continue;
|
|
|
|
|
2008-06-27 13:41:36 +02:00
|
|
|
rem_load = (u64)rem_load_move * busiest_weight;
|
|
|
|
rem_load = div_u64(rem_load, busiest_h_load + 1);
|
2007-07-09 18:51:58 +02:00
|
|
|
|
2008-06-27 13:41:14 +02:00
|
|
|
moved_load = __load_balance_fair(this_rq, this_cpu, busiest,
|
2008-06-27 13:41:20 +02:00
|
|
|
rem_load, sd, idle, all_pinned, this_best_prio,
|
2008-06-27 13:41:14 +02:00
|
|
|
tg->cfs_rq[busiest_cpu]);
|
2007-07-09 18:51:58 +02:00
|
|
|
|
2008-06-27 13:41:14 +02:00
|
|
|
if (!moved_load)
|
2007-07-09 18:51:58 +02:00
|
|
|
continue;
|
|
|
|
|
2008-06-27 13:41:29 +02:00
|
|
|
moved_load *= busiest_h_load;
|
2008-06-27 13:41:36 +02:00
|
|
|
moved_load = div_u64(moved_load, busiest_weight + 1);
|
2007-07-09 18:51:58 +02:00
|
|
|
|
2008-06-27 13:41:14 +02:00
|
|
|
rem_load_move -= moved_load;
|
|
|
|
if (rem_load_move < 0)
|
2007-07-09 18:51:58 +02:00
|
|
|
break;
|
|
|
|
}
|
2008-06-27 13:41:14 +02:00
|
|
|
rcu_read_unlock();
|
2007-07-09 18:51:58 +02:00
|
|
|
|
sched: simplify move_tasks()
The move_tasks() function is currently multiplexed with two distinct
capabilities:
1. attempt to move a specified amount of weighted load from one run
queue to another; and
2. attempt to move a specified number of tasks from one run queue to
another.
The first of these capabilities is used in two places, load_balance()
and load_balance_idle(), and in both of these cases the return value of
move_tasks() is used purely to decide if tasks/load were moved and no
notice of the actual number of tasks moved is taken.
The second capability is used in exactly one place,
active_load_balance(), to attempt to move exactly one task and, as
before, the return value is only used as an indicator of success or failure.
This multiplexing of sched_task() was introduced, by me, as part of the
smpnice patches and was motivated by the fact that the alternative, one
function to move specified load and one to move a single task, would
have led to two functions of roughly the same complexity as the old
move_tasks() (or the new balance_tasks()). However, the new modular
design of the new CFS scheduler allows a simpler solution to be adopted
and this patch addresses that solution by:
1. adding a new function, move_one_task(), to be used by
active_load_balance(); and
2. making move_tasks() a single purpose function that tries to move a
specified weighted load and returns 1 for success and 0 for failure.
One of the consequences of these changes is that neither move_one_task()
or the new move_tasks() care how many tasks sched_class.load_balance()
moves and this enables its interface to be simplified by returning the
amount of load moved as its result and removing the load_moved pointer
from the argument list. This helps simplify the new move_tasks() and
slightly reduces the amount of work done in each of
sched_class.load_balance()'s implementations.
Further simplification, e.g. changes to balance_tasks(), are possible
but (slightly) complicated by the special needs of load_balance_fair()
so I've left them to a later patch (if this one gets accepted).
NB Since move_tasks() gets called with two run queue locks held even
small reductions in overhead are worthwhile.
[ mingo@elte.hu ]
this change also reduces code size nicely:
text data bss dec hex filename
39216 3618 24 42858 a76a sched.o.before
39173 3618 24 42815 a73f sched.o.after
Signed-off-by: Peter Williams <pwil3058@bigpond.net.au>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2007-08-09 11:16:46 +02:00
|
|
|
return max_load_move - rem_load_move;
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
2008-06-27 13:41:14 +02:00
|
|
|
#else
|
|
|
|
static unsigned long
|
|
|
|
load_balance_fair(struct rq *this_rq, int this_cpu, struct rq *busiest,
|
|
|
|
unsigned long max_load_move,
|
|
|
|
struct sched_domain *sd, enum cpu_idle_type idle,
|
|
|
|
int *all_pinned, int *this_best_prio)
|
|
|
|
{
|
|
|
|
return __load_balance_fair(this_rq, this_cpu, busiest,
|
|
|
|
max_load_move, sd, idle, all_pinned,
|
|
|
|
this_best_prio, &busiest->cfs);
|
|
|
|
}
|
|
|
|
#endif
|
2007-07-09 18:51:58 +02:00
|
|
|
|
2007-10-24 18:23:51 +02:00
|
|
|
static int
|
|
|
|
move_one_task_fair(struct rq *this_rq, int this_cpu, struct rq *busiest,
|
|
|
|
struct sched_domain *sd, enum cpu_idle_type idle)
|
|
|
|
{
|
|
|
|
struct cfs_rq *busy_cfs_rq;
|
|
|
|
struct rq_iterator cfs_rq_iterator;
|
|
|
|
|
|
|
|
cfs_rq_iterator.start = load_balance_start_fair;
|
|
|
|
cfs_rq_iterator.next = load_balance_next_fair;
|
|
|
|
|
|
|
|
for_each_leaf_cfs_rq(busiest, busy_cfs_rq) {
|
|
|
|
/*
|
|
|
|
* pass busy_cfs_rq argument into
|
|
|
|
* load_balance_[start|next]_fair iterators
|
|
|
|
*/
|
|
|
|
cfs_rq_iterator.arg = busy_cfs_rq;
|
|
|
|
if (iter_move_one_task(this_rq, this_cpu, busiest, sd, idle,
|
|
|
|
&cfs_rq_iterator))
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
2008-06-24 20:09:43 +02:00
|
|
|
#endif /* CONFIG_SMP */
|
2007-10-24 18:23:51 +02:00
|
|
|
|
2007-07-09 18:51:58 +02:00
|
|
|
/*
|
|
|
|
* scheduler tick hitting a task of our scheduling class:
|
|
|
|
*/
|
2008-01-25 21:08:29 +01:00
|
|
|
static void task_tick_fair(struct rq *rq, struct task_struct *curr, int queued)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
|
|
|
struct cfs_rq *cfs_rq;
|
|
|
|
struct sched_entity *se = &curr->se;
|
|
|
|
|
|
|
|
for_each_sched_entity(se) {
|
|
|
|
cfs_rq = cfs_rq_of(se);
|
2008-01-25 21:08:29 +01:00
|
|
|
entity_tick(cfs_rq, se, queued);
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2007-10-29 21:18:11 +01:00
|
|
|
#define swap(a, b) do { typeof(a) tmp = (a); (a) = (b); (b) = tmp; } while (0)
|
2007-10-15 17:00:04 +02:00
|
|
|
|
2007-07-09 18:51:58 +02:00
|
|
|
/*
|
|
|
|
* Share the fairness runtime between parent and child, thus the
|
|
|
|
* total amount of pressure for CPU stays equal - new tasks
|
|
|
|
* get a chance to run but frequent forkers are not allowed to
|
|
|
|
* monopolize the CPU. Note: the parent runqueue is locked,
|
|
|
|
* the child is not running yet.
|
|
|
|
*/
|
2007-08-09 11:16:49 +02:00
|
|
|
static void task_new_fair(struct rq *rq, struct task_struct *p)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
|
|
|
struct cfs_rq *cfs_rq = task_cfs_rq(p);
|
2007-10-15 17:00:03 +02:00
|
|
|
struct sched_entity *se = &p->se, *curr = cfs_rq->curr;
|
2007-10-15 17:00:14 +02:00
|
|
|
int this_cpu = smp_processor_id();
|
2007-07-09 18:51:58 +02:00
|
|
|
|
|
|
|
sched_info_queued(p);
|
|
|
|
|
2007-08-28 12:53:24 +02:00
|
|
|
update_curr(cfs_rq);
|
2007-10-15 17:00:05 +02:00
|
|
|
place_entity(cfs_rq, se, 1);
|
2007-10-15 17:00:04 +02:00
|
|
|
|
sched: fix copy_namespace() <-> sched_fork() dependency in do_fork
Sukadev Bhattiprolu reported a kernel crash with control groups.
There are couple of problems discovered by Suka's test:
- The test requires the cgroup filesystem to be mounted with
atleast the cpu and ns options (i.e both namespace and cpu
controllers are active in the same hierarchy).
# mkdir /dev/cpuctl
# mount -t cgroup -ocpu,ns none cpuctl
(or simply)
# mount -t cgroup none cpuctl -> Will activate all controllers
in same hierarchy.
- The test invokes clone() with CLONE_NEWNS set. This causes a a new child
to be created, also a new group (do_fork->copy_namespaces->ns_cgroup_clone->
cgroup_clone) and the child is attached to the new group (cgroup_clone->
attach_task->sched_move_task). At this point in time, the child's scheduler
related fields are uninitialized (including its on_rq field, which it has
inherited from parent). As a result sched_move_task thinks its on
runqueue, when it isn't.
As a solution to this problem, I moved sched_fork() call, which
initializes scheduler related fields on a new task, before
copy_namespaces(). I am not sure though whether moving up will
cause other side-effects. Do you see any issue?
- The second problem exposed by this test is that task_new_fair()
assumes that parent and child will be part of the same group (which
needn't be as this test shows). As a result, cfs_rq->curr can be NULL
for the child.
The solution is to test for curr pointer being NULL in
task_new_fair().
With the patch below, I could run ns_exec() fine w/o a crash.
Reported-by: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Signed-off-by: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2007-11-09 22:39:39 +01:00
|
|
|
/* 'curr' will be NULL if the child belongs to a different group */
|
2007-10-15 17:00:14 +02:00
|
|
|
if (sysctl_sched_child_runs_first && this_cpu == task_cpu(p) &&
|
sched: fix copy_namespace() <-> sched_fork() dependency in do_fork
Sukadev Bhattiprolu reported a kernel crash with control groups.
There are couple of problems discovered by Suka's test:
- The test requires the cgroup filesystem to be mounted with
atleast the cpu and ns options (i.e both namespace and cpu
controllers are active in the same hierarchy).
# mkdir /dev/cpuctl
# mount -t cgroup -ocpu,ns none cpuctl
(or simply)
# mount -t cgroup none cpuctl -> Will activate all controllers
in same hierarchy.
- The test invokes clone() with CLONE_NEWNS set. This causes a a new child
to be created, also a new group (do_fork->copy_namespaces->ns_cgroup_clone->
cgroup_clone) and the child is attached to the new group (cgroup_clone->
attach_task->sched_move_task). At this point in time, the child's scheduler
related fields are uninitialized (including its on_rq field, which it has
inherited from parent). As a result sched_move_task thinks its on
runqueue, when it isn't.
As a solution to this problem, I moved sched_fork() call, which
initializes scheduler related fields on a new task, before
copy_namespaces(). I am not sure though whether moving up will
cause other side-effects. Do you see any issue?
- The second problem exposed by this test is that task_new_fair()
assumes that parent and child will be part of the same group (which
needn't be as this test shows). As a result, cfs_rq->curr can be NULL
for the child.
The solution is to test for curr pointer being NULL in
task_new_fair().
With the patch below, I could run ns_exec() fine w/o a crash.
Reported-by: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Signed-off-by: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2007-11-09 22:39:39 +01:00
|
|
|
curr && curr->vruntime < se->vruntime) {
|
2007-10-15 17:00:08 +02:00
|
|
|
/*
|
2007-10-15 17:00:08 +02:00
|
|
|
* Upon rescheduling, sched_class::put_prev_task() will place
|
|
|
|
* 'current' within the tree based on its new key value.
|
|
|
|
*/
|
2007-10-15 17:00:04 +02:00
|
|
|
swap(curr->vruntime, se->vruntime);
|
2008-08-28 11:12:49 +02:00
|
|
|
resched_task(rq->curr);
|
2007-10-15 17:00:04 +02:00
|
|
|
}
|
2007-07-09 18:51:58 +02:00
|
|
|
|
2007-10-17 16:55:11 +02:00
|
|
|
enqueue_task_fair(rq, p, 0);
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
|
2008-01-25 21:08:22 +01:00
|
|
|
/*
|
|
|
|
* Priority of the task has changed. Check to see if we preempt
|
|
|
|
* the current task.
|
|
|
|
*/
|
|
|
|
static void prio_changed_fair(struct rq *rq, struct task_struct *p,
|
|
|
|
int oldprio, int running)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* Reschedule if we are currently running on this runqueue and
|
|
|
|
* our priority decreased, or if we are not currently running on
|
|
|
|
* this runqueue and our priority is higher than the current's
|
|
|
|
*/
|
|
|
|
if (running) {
|
|
|
|
if (p->prio > oldprio)
|
|
|
|
resched_task(rq->curr);
|
|
|
|
} else
|
2008-09-20 23:38:02 +02:00
|
|
|
check_preempt_curr(rq, p, 0);
|
2008-01-25 21:08:22 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* We switched to the sched_fair class.
|
|
|
|
*/
|
|
|
|
static void switched_to_fair(struct rq *rq, struct task_struct *p,
|
|
|
|
int running)
|
|
|
|
{
|
|
|
|
/*
|
|
|
|
* We were most likely switched from sched_rt, so
|
|
|
|
* kick off the schedule if running, otherwise just see
|
|
|
|
* if we can still preempt the current task.
|
|
|
|
*/
|
|
|
|
if (running)
|
|
|
|
resched_task(rq->curr);
|
|
|
|
else
|
2008-09-20 23:38:02 +02:00
|
|
|
check_preempt_curr(rq, p, 0);
|
2008-01-25 21:08:22 +01:00
|
|
|
}
|
|
|
|
|
2007-10-15 17:00:08 +02:00
|
|
|
/* Account for a task changing its policy or group.
|
|
|
|
*
|
|
|
|
* This routine is mostly called to set cfs_rq->curr field when a task
|
|
|
|
* migrates between groups/classes.
|
|
|
|
*/
|
|
|
|
static void set_curr_task_fair(struct rq *rq)
|
|
|
|
{
|
|
|
|
struct sched_entity *se = &rq->curr->se;
|
|
|
|
|
|
|
|
for_each_sched_entity(se)
|
|
|
|
set_next_entity(cfs_rq_of(se), se);
|
|
|
|
}
|
|
|
|
|
2008-02-29 21:21:01 +01:00
|
|
|
#ifdef CONFIG_FAIR_GROUP_SCHED
|
|
|
|
static void moved_group_fair(struct task_struct *p)
|
|
|
|
{
|
|
|
|
struct cfs_rq *cfs_rq = task_cfs_rq(p);
|
|
|
|
|
|
|
|
update_curr(cfs_rq);
|
|
|
|
place_entity(cfs_rq, &p->se, 1);
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2007-07-09 18:51:58 +02:00
|
|
|
/*
|
|
|
|
* All the scheduling class methods:
|
|
|
|
*/
|
2007-10-15 17:00:12 +02:00
|
|
|
static const struct sched_class fair_sched_class = {
|
|
|
|
.next = &idle_sched_class,
|
2007-07-09 18:51:58 +02:00
|
|
|
.enqueue_task = enqueue_task_fair,
|
|
|
|
.dequeue_task = dequeue_task_fair,
|
|
|
|
.yield_task = yield_task_fair,
|
2008-01-25 21:08:09 +01:00
|
|
|
#ifdef CONFIG_SMP
|
|
|
|
.select_task_rq = select_task_rq_fair,
|
|
|
|
#endif /* CONFIG_SMP */
|
2007-07-09 18:51:58 +02:00
|
|
|
|
2007-10-15 17:00:05 +02:00
|
|
|
.check_preempt_curr = check_preempt_wakeup,
|
2007-07-09 18:51:58 +02:00
|
|
|
|
|
|
|
.pick_next_task = pick_next_task_fair,
|
|
|
|
.put_prev_task = put_prev_task_fair,
|
|
|
|
|
2007-10-24 18:23:51 +02:00
|
|
|
#ifdef CONFIG_SMP
|
2007-07-09 18:51:58 +02:00
|
|
|
.load_balance = load_balance_fair,
|
2007-10-24 18:23:51 +02:00
|
|
|
.move_one_task = move_one_task_fair,
|
2007-10-24 18:23:51 +02:00
|
|
|
#endif
|
2007-07-09 18:51:58 +02:00
|
|
|
|
2007-10-15 17:00:08 +02:00
|
|
|
.set_curr_task = set_curr_task_fair,
|
2007-07-09 18:51:58 +02:00
|
|
|
.task_tick = task_tick_fair,
|
|
|
|
.task_new = task_new_fair,
|
2008-01-25 21:08:22 +01:00
|
|
|
|
|
|
|
.prio_changed = prio_changed_fair,
|
|
|
|
.switched_to = switched_to_fair,
|
2008-02-29 21:21:01 +01:00
|
|
|
|
|
|
|
#ifdef CONFIG_FAIR_GROUP_SCHED
|
|
|
|
.moved_group = moved_group_fair,
|
|
|
|
#endif
|
2007-07-09 18:51:58 +02:00
|
|
|
};
|
|
|
|
|
|
|
|
#ifdef CONFIG_SCHED_DEBUG
|
2007-08-09 11:16:47 +02:00
|
|
|
static void print_cfs_stats(struct seq_file *m, int cpu)
|
2007-07-09 18:51:58 +02:00
|
|
|
{
|
|
|
|
struct cfs_rq *cfs_rq;
|
|
|
|
|
2008-01-25 21:08:34 +01:00
|
|
|
rcu_read_lock();
|
2007-08-09 11:16:51 +02:00
|
|
|
for_each_leaf_cfs_rq(cpu_rq(cpu), cfs_rq)
|
2007-08-09 11:16:47 +02:00
|
|
|
print_cfs_rq(m, cpu, cfs_rq);
|
2008-01-25 21:08:34 +01:00
|
|
|
rcu_read_unlock();
|
2007-07-09 18:51:58 +02:00
|
|
|
}
|
|
|
|
#endif
|