[linux.git] / block / blk-mq-cpumap.c

/*
 * CPU <-> hardware queue mapping helpers
 *
 * Copyright (C) 2013-2014 Jens Axboe
 */
#include <linux/kernel.h>
#include <linux/threads.h>
#include <linux/module.h>
#include <linux/mm.h>
#include <linux/smp.h>
#include <linux/cpu.h>

#include <linux/blk-mq.h>
#include "blk.h"
#include "blk-mq.h"

static int cpu_to_queue_index(unsigned int nr_cpus, unsigned int nr_queues,
			      const int cpu)
{
	return cpu * nr_queues / nr_cpus;
}

static int get_first_sibling(unsigned int cpu)
{
	unsigned int ret;

	ret = cpumask_first(topology_sibling_cpumask(cpu));
	if (ret < nr_cpu_ids)
		return ret;

	return cpu;
}

int blk_mq_map_queues(struct blk_mq_tag_set *set)
{
	unsigned int *map = set->mq_map;
	unsigned int nr_queues = set->nr_hw_queues;
	const struct cpumask *online_mask = cpu_online_mask;
	unsigned int i, nr_cpus, nr_uniq_cpus, queue, first_sibling;
	cpumask_var_t cpus;

	if (!alloc_cpumask_var(&cpus, GFP_ATOMIC))
		return -ENOMEM;

	cpumask_clear(cpus);
	nr_cpus = nr_uniq_cpus = 0;
	for_each_cpu(i, online_mask) {
		nr_cpus++;
		first_sibling = get_first_sibling(i);
		if (!cpumask_test_cpu(first_sibling, cpus))
			nr_uniq_cpus++;
		cpumask_set_cpu(i, cpus);
	}

	queue = 0;
	for_each_possible_cpu(i) {
		if (!cpumask_test_cpu(i, online_mask)) {
			map[i] = 0;
			continue;
		}

		/*
		 * Easy case - we have equal or more hardware queues. Or
		 * there are no thread siblings to take into account. Do
		 * 1:1 if enough, or sequential mapping if less.
		 */
		if (nr_queues >= nr_cpus || nr_cpus == nr_uniq_cpus) {
			map[i] = cpu_to_queue_index(nr_cpus, nr_queues, queue);
			queue++;
			continue;
		}

		/*
		 * Less then nr_cpus queues, and we have some number of
		 * threads per cores. Map sibling threads to the same
		 * queue.
		 */
		first_sibling = get_first_sibling(i);
		if (first_sibling == i) {
			map[i] = cpu_to_queue_index(nr_uniq_cpus, nr_queues,
							queue);
			queue++;
		} else
			map[i] = map[first_sibling];
	}

	free_cpumask_var(cpus);
	return 0;
}
EXPORT_SYMBOL_GPL(blk_mq_map_queues);

/*
 * We have no quick way of doing reverse lookups. This is only used at
 * queue init time, so runtime isn't important.
 */
int blk_mq_hw_queue_to_node(unsigned int *mq_map, unsigned int index)
{
	int i;

	for_each_possible_cpu(i) {
		if (index == mq_map[i])
			return local_memory_node(cpu_to_node(i));
	}

	return NUMA_NO_NODE;
}
Commit	Line	Data
75bb4625 JA	1	/*
	2	* CPU <-> hardware queue mapping helpers
	3	*
	4	* Copyright (C) 2013-2014 Jens Axboe
	5	*/
320ae51f JA	6	#include <linux/kernel.h>
	7	#include <linux/threads.h>
	8	#include <linux/module.h>
	9	#include <linux/mm.h>
	10	#include <linux/smp.h>
	11	#include <linux/cpu.h>
	12
	13	#include <linux/blk-mq.h>
	14	#include "blk.h"
	15	#include "blk-mq.h"
	16
320ae51f JA	17	static int cpu_to_queue_index(unsigned int nr_cpus, unsigned int nr_queues,
	18	const int cpu)
	19	{
959f5f5b	20	return cpu * nr_queues / nr_cpus;
320ae51f JA	21	}
	22
	23	static int get_first_sibling(unsigned int cpu)
	24	{
	25	unsigned int ret;
	26
06931e62	27	ret = cpumask_first(topology_sibling_cpumask(cpu));
320ae51f JA	28	if (ret < nr_cpu_ids)
	29	return ret;
	30
	31	return cpu;
	32	}
	33
da695ba2	34	int blk_mq_map_queues(struct blk_mq_tag_set *set)
320ae51f	35	{
da695ba2 CH	36	unsigned int *map = set->mq_map;
	37	unsigned int nr_queues = set->nr_hw_queues;
	38	const struct cpumask *online_mask = cpu_online_mask;
320ae51f JA	39	unsigned int i, nr_cpus, nr_uniq_cpus, queue, first_sibling;
	40	cpumask_var_t cpus;
	41
	42	if (!alloc_cpumask_var(&cpus, GFP_ATOMIC))
da695ba2	43	return -ENOMEM;
320ae51f JA	44
	45	cpumask_clear(cpus);
	46	nr_cpus = nr_uniq_cpus = 0;
5778322e	47	for_each_cpu(i, online_mask) {
320ae51f JA	48	nr_cpus++;
	49	first_sibling = get_first_sibling(i);
	50	if (!cpumask_test_cpu(first_sibling, cpus))
	51	nr_uniq_cpus++;
	52	cpumask_set_cpu(i, cpus);
	53	}
	54
	55	queue = 0;
	56	for_each_possible_cpu(i) {
5778322e	57	if (!cpumask_test_cpu(i, online_mask)) {
320ae51f JA	58	map[i] = 0;
	59	continue;
	60	}
	61
	62	/*
	63	* Easy case - we have equal or more hardware queues. Or
	64	* there are no thread siblings to take into account. Do
	65	* 1:1 if enough, or sequential mapping if less.
	66	*/
	67	if (nr_queues >= nr_cpus \|\| nr_cpus == nr_uniq_cpus) {
	68	map[i] = cpu_to_queue_index(nr_cpus, nr_queues, queue);
	69	queue++;
	70	continue;
	71	}
	72
	73	/*
	74	* Less then nr_cpus queues, and we have some number of
	75	* threads per cores. Map sibling threads to the same
	76	* queue.
	77	*/
	78	first_sibling = get_first_sibling(i);
	79	if (first_sibling == i) {
	80	map[i] = cpu_to_queue_index(nr_uniq_cpus, nr_queues,
	81	queue);
	82	queue++;
	83	} else
	84	map[i] = map[first_sibling];
	85	}
	86
320ae51f JA	87	free_cpumask_var(cpus);
	88	return 0;
	89	}
9e5a7e22	90	EXPORT_SYMBOL_GPL(blk_mq_map_queues);
320ae51f	91
f14bbe77 JA	92	/*
	93	* We have no quick way of doing reverse lookups. This is only used at
	94	* queue init time, so runtime isn't important.
	95	*/
	96	int blk_mq_hw_queue_to_node(unsigned int *mq_map, unsigned int index)
	97	{
	98	int i;
	99
	100	for_each_possible_cpu(i) {
	101	if (index == mq_map[i])
bffed457	102	return local_memory_node(cpu_to_node(i));
f14bbe77 JA	103	}
	104
	105	return NUMA_NO_NODE;
	106	}