2006-09-26 10:31:50 +04:00
|
|
|
/*
|
|
|
|
* linux/mm/allocpercpu.c
|
|
|
|
*
|
2008-07-04 20:59:22 +04:00
|
|
|
* Separated from slab.c August 11, 2006 Christoph Lameter
|
2006-09-26 10:31:50 +04:00
|
|
|
*/
|
|
|
|
#include <linux/mm.h>
|
|
|
|
#include <linux/module.h>
|
2009-03-30 14:07:44 +04:00
|
|
|
#include <linux/bootmem.h>
|
|
|
|
#include <asm/sections.h>
|
2006-09-26 10:31:50 +04:00
|
|
|
|
2008-03-05 01:28:35 +03:00
|
|
|
#ifndef cache_line_size
|
|
|
|
#define cache_line_size() L1_CACHE_BYTES
|
|
|
|
#endif
|
|
|
|
|
2006-09-26 10:31:50 +04:00
|
|
|
/**
|
|
|
|
* percpu_depopulate - depopulate per-cpu data for given cpu
|
|
|
|
* @__pdata: per-cpu data to depopulate
|
|
|
|
* @cpu: depopulate per-cpu data for this cpu
|
|
|
|
*
|
|
|
|
* Depopulating per-cpu data for a cpu going offline would be a typical
|
|
|
|
* use case. You need to register a cpu hotplug handler for that purpose.
|
|
|
|
*/
|
2008-07-26 06:46:23 +04:00
|
|
|
static void percpu_depopulate(void *__pdata, int cpu)
|
2006-09-26 10:31:50 +04:00
|
|
|
{
|
|
|
|
struct percpu_data *pdata = __percpu_disguise(__pdata);
|
2006-12-07 07:32:37 +03:00
|
|
|
|
|
|
|
kfree(pdata->ptrs[cpu]);
|
|
|
|
pdata->ptrs[cpu] = NULL;
|
2006-09-26 10:31:50 +04:00
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* percpu_depopulate_mask - depopulate per-cpu data for some cpu's
|
|
|
|
* @__pdata: per-cpu data to depopulate
|
|
|
|
* @mask: depopulate per-cpu data for cpu's selected through mask bits
|
|
|
|
*/
|
2009-04-06 09:08:29 +04:00
|
|
|
static void __percpu_depopulate_mask(void *__pdata, const cpumask_t *mask)
|
2006-09-26 10:31:50 +04:00
|
|
|
{
|
|
|
|
int cpu;
|
2008-05-12 23:21:13 +04:00
|
|
|
for_each_cpu_mask_nr(cpu, *mask)
|
2006-09-26 10:31:50 +04:00
|
|
|
percpu_depopulate(__pdata, cpu);
|
|
|
|
}
|
2008-07-26 06:46:23 +04:00
|
|
|
|
|
|
|
#define percpu_depopulate_mask(__pdata, mask) \
|
|
|
|
__percpu_depopulate_mask((__pdata), &(mask))
|
2006-09-26 10:31:50 +04:00
|
|
|
|
|
|
|
/**
|
|
|
|
* percpu_populate - populate per-cpu data for given cpu
|
|
|
|
* @__pdata: per-cpu data to populate further
|
|
|
|
* @size: size of per-cpu object
|
|
|
|
* @gfp: may sleep or not etc.
|
|
|
|
* @cpu: populate per-data for this cpu
|
|
|
|
*
|
|
|
|
* Populating per-cpu data for a cpu coming online would be a typical
|
|
|
|
* use case. You need to register a cpu hotplug handler for that purpose.
|
|
|
|
* Per-cpu object is populated with zeroed buffer.
|
|
|
|
*/
|
2008-07-26 06:46:23 +04:00
|
|
|
static void *percpu_populate(void *__pdata, size_t size, gfp_t gfp, int cpu)
|
2006-09-26 10:31:50 +04:00
|
|
|
{
|
|
|
|
struct percpu_data *pdata = __percpu_disguise(__pdata);
|
|
|
|
int node = cpu_to_node(cpu);
|
|
|
|
|
2008-03-05 01:28:35 +03:00
|
|
|
/*
|
|
|
|
* We should make sure each CPU gets private memory.
|
|
|
|
*/
|
|
|
|
size = roundup(size, cache_line_size());
|
|
|
|
|
2006-09-26 10:31:50 +04:00
|
|
|
BUG_ON(pdata->ptrs[cpu]);
|
2007-07-17 15:03:29 +04:00
|
|
|
if (node_online(node))
|
|
|
|
pdata->ptrs[cpu] = kmalloc_node(size, gfp|__GFP_ZERO, node);
|
|
|
|
else
|
2006-09-26 10:31:50 +04:00
|
|
|
pdata->ptrs[cpu] = kzalloc(size, gfp);
|
|
|
|
return pdata->ptrs[cpu];
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* percpu_populate_mask - populate per-cpu data for more cpu's
|
|
|
|
* @__pdata: per-cpu data to populate further
|
|
|
|
* @size: size of per-cpu object
|
|
|
|
* @gfp: may sleep or not etc.
|
|
|
|
* @mask: populate per-cpu data for cpu's selected through mask bits
|
|
|
|
*
|
|
|
|
* Per-cpu objects are populated with zeroed buffers.
|
|
|
|
*/
|
2008-07-26 06:46:23 +04:00
|
|
|
static int __percpu_populate_mask(void *__pdata, size_t size, gfp_t gfp,
|
|
|
|
cpumask_t *mask)
|
2006-09-26 10:31:50 +04:00
|
|
|
{
|
2008-04-05 05:11:12 +04:00
|
|
|
cpumask_t populated;
|
2006-09-26 10:31:50 +04:00
|
|
|
int cpu;
|
|
|
|
|
2008-04-05 05:11:12 +04:00
|
|
|
cpus_clear(populated);
|
2008-05-12 23:21:13 +04:00
|
|
|
for_each_cpu_mask_nr(cpu, *mask)
|
2006-09-26 10:31:50 +04:00
|
|
|
if (unlikely(!percpu_populate(__pdata, size, gfp, cpu))) {
|
|
|
|
__percpu_depopulate_mask(__pdata, &populated);
|
|
|
|
return -ENOMEM;
|
|
|
|
} else
|
|
|
|
cpu_set(cpu, populated);
|
|
|
|
return 0;
|
|
|
|
}
|
2008-07-26 06:46:23 +04:00
|
|
|
|
|
|
|
#define percpu_populate_mask(__pdata, size, gfp, mask) \
|
|
|
|
__percpu_populate_mask((__pdata), (size), (gfp), &(mask))
|
2006-09-26 10:31:50 +04:00
|
|
|
|
|
|
|
/**
|
2009-02-20 10:29:08 +03:00
|
|
|
* alloc_percpu - initial setup of per-cpu data
|
2006-09-26 10:31:50 +04:00
|
|
|
* @size: size of per-cpu object
|
2009-02-20 10:29:08 +03:00
|
|
|
* @align: alignment
|
2006-09-26 10:31:50 +04:00
|
|
|
*
|
2009-02-20 10:29:08 +03:00
|
|
|
* Allocate dynamic percpu area. Percpu objects are populated with
|
|
|
|
* zeroed buffers.
|
2006-09-26 10:31:50 +04:00
|
|
|
*/
|
2009-02-20 10:29:08 +03:00
|
|
|
void *__alloc_percpu(size_t size, size_t align)
|
2006-09-26 10:31:50 +04:00
|
|
|
{
|
2008-03-05 01:28:35 +03:00
|
|
|
/*
|
|
|
|
* We allocate whole cache lines to avoid false sharing
|
|
|
|
*/
|
|
|
|
size_t sz = roundup(nr_cpu_ids * sizeof(void *), cache_line_size());
|
2009-02-20 10:29:08 +03:00
|
|
|
void *pdata = kzalloc(sz, GFP_KERNEL);
|
2006-09-26 10:31:50 +04:00
|
|
|
void *__pdata = __percpu_disguise(pdata);
|
|
|
|
|
2009-02-20 10:29:08 +03:00
|
|
|
/*
|
|
|
|
* Can't easily make larger alignment work with kmalloc. WARN
|
|
|
|
* on it. Larger alignment should only be used for module
|
|
|
|
* percpu sections on SMP for which this path isn't used.
|
|
|
|
*/
|
2009-03-11 08:36:54 +03:00
|
|
|
WARN_ON_ONCE(align > SMP_CACHE_BYTES);
|
2009-02-20 10:29:08 +03:00
|
|
|
|
2006-09-26 10:31:50 +04:00
|
|
|
if (unlikely(!pdata))
|
|
|
|
return NULL;
|
2009-02-20 10:29:08 +03:00
|
|
|
if (likely(!__percpu_populate_mask(__pdata, size, GFP_KERNEL,
|
|
|
|
&cpu_possible_map)))
|
2006-09-26 10:31:50 +04:00
|
|
|
return __pdata;
|
|
|
|
kfree(pdata);
|
|
|
|
return NULL;
|
|
|
|
}
|
2009-02-20 10:29:08 +03:00
|
|
|
EXPORT_SYMBOL_GPL(__alloc_percpu);
|
2006-09-26 10:31:50 +04:00
|
|
|
|
|
|
|
/**
|
2009-02-20 10:29:08 +03:00
|
|
|
* free_percpu - final cleanup of per-cpu data
|
2006-09-26 10:31:50 +04:00
|
|
|
* @__pdata: object to clean up
|
|
|
|
*
|
|
|
|
* We simply clean up any per-cpu object left. No need for the client to
|
|
|
|
* track and specify through a bis mask which per-cpu objects are to free.
|
|
|
|
*/
|
2009-02-20 10:29:08 +03:00
|
|
|
void free_percpu(void *__pdata)
|
2006-09-26 10:31:50 +04:00
|
|
|
{
|
2006-12-07 07:32:37 +03:00
|
|
|
if (unlikely(!__pdata))
|
|
|
|
return;
|
2009-03-31 08:05:15 +04:00
|
|
|
__percpu_depopulate_mask(__pdata, cpu_possible_mask);
|
2006-09-26 10:31:50 +04:00
|
|
|
kfree(__percpu_disguise(__pdata));
|
|
|
|
}
|
2009-02-20 10:29:08 +03:00
|
|
|
EXPORT_SYMBOL_GPL(free_percpu);
|
2009-03-30 14:07:44 +04:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Generic percpu area setup.
|
|
|
|
*/
|
|
|
|
#ifndef CONFIG_HAVE_SETUP_PER_CPU_AREA
|
|
|
|
unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
|
|
|
|
|
|
|
|
EXPORT_SYMBOL(__per_cpu_offset);
|
|
|
|
|
|
|
|
void __init setup_per_cpu_areas(void)
|
|
|
|
{
|
|
|
|
unsigned long size, i;
|
|
|
|
char *ptr;
|
|
|
|
unsigned long nr_possible_cpus = num_possible_cpus();
|
|
|
|
|
|
|
|
/* Copy section for each CPU (we discard the original) */
|
|
|
|
size = ALIGN(PERCPU_ENOUGH_ROOM, PAGE_SIZE);
|
|
|
|
ptr = alloc_bootmem_pages(size * nr_possible_cpus);
|
|
|
|
|
|
|
|
for_each_possible_cpu(i) {
|
|
|
|
__per_cpu_offset[i] = ptr - __per_cpu_start;
|
|
|
|
memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
|
|
|
|
ptr += size;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
#endif /* CONFIG_HAVE_SETUP_PER_CPU_AREA */
|