[net-next-2.6.git] / kernel / sched_cpupri.c

/*
 *  kernel/sched_cpupri.c
 *
 *  CPU priority management
 *
 *  Copyright (C) 2007-2008 Novell
 *
 *  Author: Gregory Haskins <ghaskins@novell.com>
 *
 *  This code tracks the priority of each CPU so that global migration
 *  decisions are easy to calculate.  Each CPU can be in a state as follows:
 *
 *                 (INVALID), IDLE, NORMAL, RT1, ... RT99
 *
 *  going from the lowest priority to the highest.  CPUs in the INVALID state
 *  are not eligible for routing.  The system maintains this state with
 *  a 2 dimensional bitmap (the first for priority class, the second for cpus
 *  in that class).  Therefore a typical application without affinity
 *  restrictions can find a suitable CPU with O(1) complexity (e.g. two bit
 *  searches).  For tasks with affinity restrictions, the algorithm has a
 *  worst case complexity of O(min(102, nr_domcpus)), though the scenario that
 *  yields the worst case search is fairly contrived.
 *
 *  This program is free software; you can redistribute it and/or
 *  modify it under the terms of the GNU General Public License
 *  as published by the Free Software Foundation; version 2
 *  of the License.
 */

#include "sched_cpupri.h"

/* Convert between a 140 based task->prio, and our 102 based cpupri */
static int convert_prio(int prio)
{
	int cpupri;

	if (prio == CPUPRI_INVALID)
		cpupri = CPUPRI_INVALID;
	else if (prio == MAX_PRIO)
		cpupri = CPUPRI_IDLE;
	else if (prio >= MAX_RT_PRIO)
		cpupri = CPUPRI_NORMAL;
	else
		cpupri = MAX_RT_PRIO - prio + 1;

	return cpupri;
}

#define for_each_cpupri_active(array, idx)                    \
  for (idx = find_first_bit(array, CPUPRI_NR_PRIORITIES);     \
       idx < CPUPRI_NR_PRIORITIES;                            \
       idx = find_next_bit(array, CPUPRI_NR_PRIORITIES, idx+1))

/**
 * cpupri_find - find the best (lowest-pri) CPU in the system
 * @cp: The cpupri context
 * @p: The task
 * @lowest_mask: A mask to fill in with selected CPUs
 *
 * Note: This function returns the recommended CPUs as calculated during the
 * current invokation.  By the time the call returns, the CPUs may have in
 * fact changed priorities any number of times.  While not ideal, it is not
 * an issue of correctness since the normal rebalancer logic will correct
 * any discrepancies created by racing against the uncertainty of the current
 * priority configuration.
 *
 * Returns: (int)bool - CPUs were found
 */
int cpupri_find(struct cpupri *cp, struct task_struct *p,
		struct cpumask *lowest_mask)
{
	int                  idx      = 0;
	int                  task_pri = convert_prio(p->prio);

	for_each_cpupri_active(cp->pri_active, idx) {
		struct cpupri_vec *vec  = &cp->pri_to_cpu[idx];

		if (idx >= task_pri)
			break;

		if (cpumask_any_and(&p->cpus_allowed, vec->mask) >= nr_cpu_ids)
			continue;

		cpumask_and(lowest_mask, &p->cpus_allowed, vec->mask);
		return 1;
	}

	return 0;
}

/**
 * cpupri_set - update the cpu priority setting
 * @cp: The cpupri context
 * @cpu: The target cpu
 * @pri: The priority (INVALID-RT99) to assign to this CPU
 *
 * Note: Assumes cpu_rq(cpu)->lock is locked
 *
 * Returns: (void)
 */
void cpupri_set(struct cpupri *cp, int cpu, int newpri)
{
	int                 *currpri = &cp->cpu_to_pri[cpu];
	int                  oldpri  = *currpri;
	unsigned long        flags;

	newpri = convert_prio(newpri);

	BUG_ON(newpri >= CPUPRI_NR_PRIORITIES);

	if (newpri == oldpri)
		return;

	/*
	 * If the cpu was currently mapped to a different value, we
	 * first need to unmap the old value
	 */
	if (likely(oldpri != CPUPRI_INVALID)) {
		struct cpupri_vec *vec  = &cp->pri_to_cpu[oldpri];

		spin_lock_irqsave(&vec->lock, flags);

		vec->count--;
		if (!vec->count)
			clear_bit(oldpri, cp->pri_active);
		cpumask_clear_cpu(cpu, vec->mask);

		spin_unlock_irqrestore(&vec->lock, flags);
	}

	if (likely(newpri != CPUPRI_INVALID)) {
		struct cpupri_vec *vec = &cp->pri_to_cpu[newpri];

		spin_lock_irqsave(&vec->lock, flags);

		cpumask_set_cpu(cpu, vec->mask);
		vec->count++;
		if (vec->count == 1)
			set_bit(newpri, cp->pri_active);

		spin_unlock_irqrestore(&vec->lock, flags);
	}

	*currpri = newpri;
}

/**
 * cpupri_init - initialize the cpupri structure
 * @cp: The cpupri context
 * @bootmem: true if allocations need to use bootmem
 *
 * Returns: -ENOMEM if memory fails.
 */
int __init_refok cpupri_init(struct cpupri *cp, bool bootmem)
{
	int i;

	memset(cp, 0, sizeof(*cp));

	for (i = 0; i < CPUPRI_NR_PRIORITIES; i++) {
		struct cpupri_vec *vec = &cp->pri_to_cpu[i];

		spin_lock_init(&vec->lock);
		vec->count = 0;
		if (bootmem)
			alloc_bootmem_cpumask_var(&vec->mask);
		else if (!alloc_cpumask_var(&vec->mask, GFP_KERNEL))
			goto cleanup;
	}

	for_each_possible_cpu(i)
		cp->cpu_to_pri[i] = CPUPRI_INVALID;
	return 0;

cleanup:
	for (i--; i >= 0; i--)
		free_cpumask_var(cp->pri_to_cpu[i].mask);
	return -ENOMEM;
}

/**
 * cpupri_cleanup - clean up the cpupri structure
 * @cp: The cpupri context
 */
void cpupri_cleanup(struct cpupri *cp)
{
	int i;

	for (i = 0; i < CPUPRI_NR_PRIORITIES; i++)
		free_cpumask_var(cp->pri_to_cpu[i].mask);
}
Commit	Line	Data
6e0534f2 GH	1	/*
	2	* kernel/sched_cpupri.c
	3	*
	4	* CPU priority management
	5	*
	6	* Copyright (C) 2007-2008 Novell
	7	*
	8	* Author: Gregory Haskins <ghaskins@novell.com>
	9	*
	10	* This code tracks the priority of each CPU so that global migration
	11	* decisions are easy to calculate. Each CPU can be in a state as follows:
	12	*
	13	* (INVALID), IDLE, NORMAL, RT1, ... RT99
	14	*
	15	* going from the lowest priority to the highest. CPUs in the INVALID state
	16	* are not eligible for routing. The system maintains this state with
	17	* a 2 dimensional bitmap (the first for priority class, the second for cpus
	18	* in that class). Therefore a typical application without affinity
	19	* restrictions can find a suitable CPU with O(1) complexity (e.g. two bit
	20	* searches). For tasks with affinity restrictions, the algorithm has a
	21	* worst case complexity of O(min(102, nr_domcpus)), though the scenario that
	22	* yields the worst case search is fairly contrived.
	23	*
	24	* This program is free software; you can redistribute it and/or
	25	* modify it under the terms of the GNU General Public License
	26	* as published by the Free Software Foundation; version 2
	27	* of the License.
	28	*/
	29
	30	#include "sched_cpupri.h"
	31
	32	/* Convert between a 140 based task->prio, and our 102 based cpupri */
	33	static int convert_prio(int prio)
	34	{
	35	int cpupri;
	36
	37	if (prio == CPUPRI_INVALID)
	38	cpupri = CPUPRI_INVALID;
	39	else if (prio == MAX_PRIO)
	40	cpupri = CPUPRI_IDLE;
	41	else if (prio >= MAX_RT_PRIO)
	42	cpupri = CPUPRI_NORMAL;
	43	else
	44	cpupri = MAX_RT_PRIO - prio + 1;
	45
	46	return cpupri;
	47	}
	48
	49	#define for_each_cpupri_active(array, idx) \
	50	for (idx = find_first_bit(array, CPUPRI_NR_PRIORITIES); \
	51	idx < CPUPRI_NR_PRIORITIES; \
	52	idx = find_next_bit(array, CPUPRI_NR_PRIORITIES, idx+1))
	53
	54	/**
	55	* cpupri_find - find the best (lowest-pri) CPU in the system
	56	* @cp: The cpupri context
	57	* @p: The task
	58	* @lowest_mask: A mask to fill in with selected CPUs
	59	*
	60	* Note: This function returns the recommended CPUs as calculated during the
	61	* current invokation. By the time the call returns, the CPUs may have in
	62	* fact changed priorities any number of times. While not ideal, it is not
	63	* an issue of correctness since the normal rebalancer logic will correct
	64	* any discrepancies created by racing against the uncertainty of the current
65	* priority configuration.
66	*
67	* Returns: (int)bool - CPUs were found
68	*/
69	int cpupri_find(struct cpupri cp, struct task_struct p,
68e74568	70	struct cpumask *lowest_mask)
6e0534f2 GH	71	{
	72	int idx = 0;
	73	int task_pri = convert_prio(p->prio);
	74
	75	for_each_cpupri_active(cp->pri_active, idx) {
	76	struct cpupri_vec *vec = &cp->pri_to_cpu[idx];
6e0534f2 GH	77
	78	if (idx >= task_pri)
	79	break;
	80
68e74568	81	if (cpumask_any_and(&p->cpus_allowed, vec->mask) >= nr_cpu_ids)
6e0534f2 GH	82	continue;
6e0534f2 GH	83
68e74568	84	cpumask_and(lowest_mask, &p->cpus_allowed, vec->mask);
6e0534f2 GH	85	return 1;
	86	}
	87
	88	return 0;
	89	}
	90
	91	/**
	92	* cpupri_set - update the cpu priority setting
	93	* @cp: The cpupri context
	94	* @cpu: The target cpu
	95	* @pri: The priority (INVALID-RT99) to assign to this CPU
	96	*
	97	* Note: Assumes cpu_rq(cpu)->lock is locked
	98	*
	99	* Returns: (void)
	100	*/
	101	void cpupri_set(struct cpupri *cp, int cpu, int newpri)
	102	{
	103	int *currpri = &cp->cpu_to_pri[cpu];
	104	int oldpri = *currpri;
	105	unsigned long flags;
	106
	107	newpri = convert_prio(newpri);
	108
	109	BUG_ON(newpri >= CPUPRI_NR_PRIORITIES);
	110
	111	if (newpri == oldpri)
	112	return;
	113
	114	/*
	115	* If the cpu was currently mapped to a different value, we
	116	* first need to unmap the old value
	117	*/
	118	if (likely(oldpri != CPUPRI_INVALID)) {
	119	struct cpupri_vec *vec = &cp->pri_to_cpu[oldpri];
	120
	121	spin_lock_irqsave(&vec->lock, flags);
	122
	123	vec->count--;
	124	if (!vec->count)
	125	clear_bit(oldpri, cp->pri_active);
68e74568	126	cpumask_clear_cpu(cpu, vec->mask);
6e0534f2 GH	127
	128	spin_unlock_irqrestore(&vec->lock, flags);
	129	}
	130
	131	if (likely(newpri != CPUPRI_INVALID)) {
	132	struct cpupri_vec *vec = &cp->pri_to_cpu[newpri];
	133
	134	spin_lock_irqsave(&vec->lock, flags);
	135
68e74568	136	cpumask_set_cpu(cpu, vec->mask);
6e0534f2 GH	137	vec->count++;
	138	if (vec->count == 1)
	139	set_bit(newpri, cp->pri_active);
	140
	141	spin_unlock_irqrestore(&vec->lock, flags);
	142	}
	143
	144	*currpri = newpri;
	145	}
	146
	147	/**
	148	* cpupri_init - initialize the cpupri structure
	149	* @cp: The cpupri context
68e74568	150	* @bootmem: true if allocations need to use bootmem
6e0534f2	151	*
68e74568	152	* Returns: -ENOMEM if memory fails.
6e0534f2	153	*/
db2f59c8	154	int __init_refok cpupri_init(struct cpupri *cp, bool bootmem)
6e0534f2 GH	155	{
	156	int i;
	157
	158	memset(cp, 0, sizeof(*cp));
	159
	160	for (i = 0; i < CPUPRI_NR_PRIORITIES; i++) {
	161	struct cpupri_vec *vec = &cp->pri_to_cpu[i];
	162
	163	spin_lock_init(&vec->lock);
	164	vec->count = 0;
68e74568 RR	165	if (bootmem)
	166	alloc_bootmem_cpumask_var(&vec->mask);
	167	else if (!alloc_cpumask_var(&vec->mask, GFP_KERNEL))
	168	goto cleanup;
6e0534f2 GH	169	}
	170
	171	for_each_possible_cpu(i)
	172	cp->cpu_to_pri[i] = CPUPRI_INVALID;
68e74568 RR	173	return 0;
	174
	175	cleanup:
	176	for (i--; i >= 0; i--)
	177	free_cpumask_var(cp->pri_to_cpu[i].mask);
	178	return -ENOMEM;
6e0534f2 GH	179	}
6e0534f2 GH	180
68e74568 RR	181	/**
	182	* cpupri_cleanup - clean up the cpupri structure
	183	* @cp: The cpupri context
	184	*/
	185	void cpupri_cleanup(struct cpupri *cp)
	186	{
	187	int i;
6e0534f2	188
68e74568 RR	189	for (i = 0; i < CPUPRI_NR_PRIORITIES; i++)
	190	free_cpumask_var(cp->pri_to_cpu[i].mask);
	191	}