8e3a77b408
blk-mq layer performs a remapping between s/w and h/w contexts and also between h/w contexts and CPUs, whenever a CPU hotplug event happens. This remapping has to wait for queue freezing which may take tens of miliseconds, resulting in a high latency in CPU hotplug path. This patch makes the above mentioned mappings static so that we can avoid remapping when CPU hotplug event happens and this results in improved CPU hotplug latency of up to 90 percent for CPU up path and of up to 50 percent for CPU down path. Change-Id: Idf38cb6c4e78c91fda3c86608c6d0441f01ab435 Signed-off-by: Imran Khan <kimran@codeaurora.org> Signed-off-by: Kyle Yan <kyan@codeaurora.org>
87 lines
1.8 KiB
C
87 lines
1.8 KiB
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;
|
|
unsigned int i, queue, first_sibling;
|
|
cpumask_var_t cpus;
|
|
|
|
queue = 0;
|
|
for_each_possible_cpu(i) {
|
|
/*
|
|
* 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_cpu_ids) {
|
|
map[i] = cpu_to_queue_index(nr_cpu_ids, 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_cpu_ids, nr_queues,
|
|
queue);
|
|
queue++;
|
|
} else
|
|
map[i] = map[first_sibling];
|
|
}
|
|
|
|
free_cpumask_var(cpus);
|
|
return 0;
|
|
}
|
|
|
|
/*
|
|
* 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;
|
|
}
|