Revision aa76042a016474775ccd187c068669148c30c3bb authored by James Hogan on 27 May 2016, 21:25:23 UTC, committed by Ralf Baechle on 28 May 2016, 10:35:11 UTC
The Hardware page Table Walker (HTW) is being misconfigured on 64-bit kernels. The PWSize.PS (pointer size) bit determines whether pointers within directories are loaded as 32-bit or 64-bit addresses, but was never being set to 1 for 64-bit kernels where the unsigned long in pgd_t is 64-bits wide. This actually reduces rather than improves performance when the HTW is enabled on P6600 since the HTW is initiated lots, but walks are all aborted due I think to bad intermediate pointers. Since we were already taking the width of the PTEs into account by setting PWSize.PTEW, which is the left shift applied to the page table index *in addition to* the native pointer size, we also need to reduce PTEW by 1 when PS=1. This is done by calculating PTEW based on the relative size of pte_t compared to pgd_t. Finally in order for the HTW to be used when PS=1, the appropriate XK/XS/XU bits corresponding to the different 64-bit segments need to be set in PWCtl. We enable only XU for now to enable walking for XUSeg. Supporting walking for XKSeg would be a bit more involved so is left for a future patch. It would either require the use of a per-CPU top level base directory if supported by the HTW (a bit like pgd_current but with a second entry pointing at swapper_pg_dir), or the HTW would prepend bit 63 of the address to the global directory index which doesn't really match how we split user and kernel page directories. Fixes: cab25bc7537b ("MIPS: Extend hardware table walking support to MIPS64") Signed-off-by: James Hogan <james.hogan@imgtec.com> Cc: Paul Burton <paul.burton@imgtec.com> Cc: linux-mips@linux-mips.org Patchwork: https://patchwork.linux-mips.org/patch/13364/ Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
1 parent 6446e6c
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_update_queue_map(unsigned int *map, unsigned int nr_queues,
const struct cpumask *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 1;
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;
}
unsigned int *blk_mq_make_queue_map(struct blk_mq_tag_set *set)
{
unsigned int *map;
/* If cpus are offline, map them to first hctx */
map = kzalloc_node(sizeof(*map) * nr_cpu_ids, GFP_KERNEL,
set->numa_node);
if (!map)
return NULL;
if (!blk_mq_update_queue_map(map, set->nr_hw_queues, cpu_online_mask))
return map;
kfree(map);
return NULL;
}
/*
* 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;
}
![swh spinner](/static/img/swh-spinner.gif)
Computing file changes ...