[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,
		cpumask_t *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];
		cpumask_t mask;

		if (idx >= task_pri)
			break;

		cpus_and(mask, p->cpus_allowed, vec->mask);

		if (cpus_empty(mask))
			continue;

		*lowest_mask = 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);
		cpu_clear(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);

		cpu_set(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
 *
 * Returns: (void)
 */
void cpupri_init(struct cpupri *cp)
{
	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;
		cpus_clear(vec->mask);
	}

	for_each_possible_cpu(i)
		cp->cpu_to_pri[i] = CPUPRI_INVALID;
}
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,
70	cpumask_t *lowest_mask)
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];
77	cpumask_t mask;
78
79	if (idx >= task_pri)
80	break;
81
82	cpus_and(mask, p->cpus_allowed, vec->mask);
83
84	if (cpus_empty(mask))
85	continue;
86
87	*lowest_mask = mask;
88	return 1;
89	}
90
91	return 0;
92	}
93
94	/**
95	* cpupri_set - update the cpu priority setting
96	* @cp: The cpupri context
97	* @cpu: The target cpu
98	* @pri: The priority (INVALID-RT99) to assign to this CPU
99	*
100	* Note: Assumes cpu_rq(cpu)->lock is locked
101	*
102	* Returns: (void)
103	*/
104	void cpupri_set(struct cpupri *cp, int cpu, int newpri)
105	{
106	int *currpri = &cp->cpu_to_pri[cpu];
107	int oldpri = *currpri;
108	unsigned long flags;
109
110	newpri = convert_prio(newpri);
111
112	BUG_ON(newpri >= CPUPRI_NR_PRIORITIES);
113
114	if (newpri == oldpri)
115	return;
116
117	/*
118	* If the cpu was currently mapped to a different value, we
119	* first need to unmap the old value
120	*/
121	if (likely(oldpri != CPUPRI_INVALID)) {
122	struct cpupri_vec *vec = &cp->pri_to_cpu[oldpri];
123
124	spin_lock_irqsave(&vec->lock, flags);
125
126	vec->count--;
127	if (!vec->count)
128	clear_bit(oldpri, cp->pri_active);
129	cpu_clear(cpu, vec->mask);
130
131	spin_unlock_irqrestore(&vec->lock, flags);
132	}
133
134	if (likely(newpri != CPUPRI_INVALID)) {
135	struct cpupri_vec *vec = &cp->pri_to_cpu[newpri];
136
137	spin_lock_irqsave(&vec->lock, flags);
138
139	cpu_set(cpu, vec->mask);
140	vec->count++;
141	if (vec->count == 1)
142	set_bit(newpri, cp->pri_active);
143
144	spin_unlock_irqrestore(&vec->lock, flags);
145	}
146
147	*currpri = newpri;
148	}
149
150	/**
151	* cpupri_init - initialize the cpupri structure
152	* @cp: The cpupri context
153	*
154	* Returns: (void)
155	*/
156	void cpupri_init(struct cpupri *cp)
157	{
158	int i;
159
160	memset(cp, 0, sizeof(*cp));
161
162	for (i = 0; i < CPUPRI_NR_PRIORITIES; i++) {
163	struct cpupri_vec *vec = &cp->pri_to_cpu[i];
164
165	spin_lock_init(&vec->lock);
166	vec->count = 0;
167	cpus_clear(vec->mask);
168	}
169
170	for_each_possible_cpu(i)
171	cp->cpu_to_pri[i] = CPUPRI_INVALID;
172	}
173
174