2012-07-07 00:25:10 +04:00
/*
* Slab allocator functions that are independent of the allocator strategy
*
* ( C ) 2012 Christoph Lameter < cl @ linux . com >
*/
# include <linux/slab.h>
# include <linux/mm.h>
# include <linux/poison.h>
# include <linux/interrupt.h>
# include <linux/memory.h>
# include <linux/compiler.h>
# include <linux/module.h>
2012-07-07 00:25:13 +04:00
# include <linux/cpu.h>
# include <linux/uaccess.h>
2012-10-19 18:20:25 +04:00
# include <linux/seq_file.h>
# include <linux/proc_fs.h>
2012-07-07 00:25:10 +04:00
# include <asm/cacheflush.h>
# include <asm/tlbflush.h>
# include <asm/page.h>
2012-12-19 02:22:34 +04:00
# include <linux/memcontrol.h>
2014-08-07 03:04:44 +04:00
# define CREATE_TRACE_POINTS
2013-09-04 20:35:34 +04:00
# include <trace/events/kmem.h>
2012-07-07 00:25:10 +04:00
2012-07-07 00:25:11 +04:00
# include "slab.h"
enum slab_state slab_state ;
2012-07-07 00:25:12 +04:00
LIST_HEAD ( slab_caches ) ;
DEFINE_MUTEX ( slab_mutex ) ;
2012-09-05 04:20:33 +04:00
struct kmem_cache * kmem_cache ;
2012-07-07 00:25:11 +04:00
2012-08-16 11:09:46 +04:00
# ifdef CONFIG_DEBUG_VM
2014-04-08 02:39:26 +04:00
static int kmem_cache_sanity_check ( const char * name , size_t size )
2012-07-07 00:25:10 +04:00
{
struct kmem_cache * s = NULL ;
if ( ! name | | in_interrupt ( ) | | size < sizeof ( void * ) | |
size > KMALLOC_MAX_SIZE ) {
2012-08-16 11:09:46 +04:00
pr_err ( " kmem_cache_create(%s) integrity check failed \n " , name ) ;
return - EINVAL ;
2012-07-07 00:25:10 +04:00
}
2012-08-16 11:12:18 +04:00
2012-07-07 00:25:13 +04:00
list_for_each_entry ( s , & slab_caches , list ) {
char tmp ;
int res ;
/*
* This happens when the module gets unloaded and doesn ' t
* destroy its slab cache and no - one else reuses the vmalloc
* area of the module . Print a warning .
*/
res = probe_kernel_address ( s - > name , tmp ) ;
if ( res ) {
2012-08-16 11:09:46 +04:00
pr_err ( " Slab cache with size %d has lost its name \n " ,
2012-07-07 00:25:13 +04:00
s - > object_size ) ;
continue ;
}
2014-03-05 02:13:47 +04:00
# if !defined(CONFIG_SLUB)
2014-04-08 02:39:26 +04:00
if ( ! strcmp ( s - > name , name ) ) {
2012-08-16 11:09:46 +04:00
pr_err ( " %s (%s): Cache name already exists. \n " ,
__func__ , name ) ;
2012-07-07 00:25:13 +04:00
dump_stack ( ) ;
s = NULL ;
2012-08-16 11:09:46 +04:00
return - EINVAL ;
2012-07-07 00:25:13 +04:00
}
2013-09-22 01:56:34 +04:00
# endif
2012-07-07 00:25:13 +04:00
}
WARN_ON ( strchr ( name , ' ' ) ) ; /* It confuses parsers */
2012-08-16 11:09:46 +04:00
return 0 ;
}
# else
2014-04-08 02:39:26 +04:00
static inline int kmem_cache_sanity_check ( const char * name , size_t size )
2012-08-16 11:09:46 +04:00
{
return 0 ;
}
2012-07-07 00:25:13 +04:00
# endif
memcg: allocate memory for memcg caches whenever a new memcg appears
Every cache that is considered a root cache (basically the "original"
caches, tied to the root memcg/no-memcg) will have an array that should be
large enough to store a cache pointer per each memcg in the system.
Theoreticaly, this is as high as 1 << sizeof(css_id), which is currently
in the 64k pointers range. Most of the time, we won't be using that much.
What goes in this patch, is a simple scheme to dynamically allocate such
an array, in order to minimize memory usage for memcg caches. Because we
would also like to avoid allocations all the time, at least for now, the
array will only grow. It will tend to be big enough to hold the maximum
number of kmem-limited memcgs ever achieved.
We'll allocate it to be a minimum of 64 kmem-limited memcgs. When we have
more than that, we'll start doubling the size of this array every time the
limit is reached.
Because we are only considering kmem limited memcgs, a natural point for
this to happen is when we write to the limit. At that point, we already
have set_limit_mutex held, so that will become our natural synchronization
mechanism.
Signed-off-by: Glauber Costa <glommer@parallels.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Frederic Weisbecker <fweisbec@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: JoonSoo Kim <js1304@gmail.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Rik van Riel <riel@redhat.com>
Cc: Suleiman Souhlal <suleiman@google.com>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2012-12-19 02:22:38 +04:00
# ifdef CONFIG_MEMCG_KMEM
int memcg_update_all_caches ( int num_memcgs )
{
struct kmem_cache * s ;
int ret = 0 ;
mutex_lock ( & slab_mutex ) ;
list_for_each_entry ( s , & slab_caches , list ) {
if ( ! is_root_cache ( s ) )
continue ;
ret = memcg_update_cache_size ( s , num_memcgs ) ;
/*
* See comment in memcontrol . c , memcg_update_cache_size :
* Instead of freeing the memory , we ' ll just leave the caches
* up to this point in an updated state .
*/
if ( ret )
goto out ;
}
memcg_update_array_size ( num_memcgs ) ;
out :
mutex_unlock ( & slab_mutex ) ;
return ret ;
}
# endif
2012-11-28 20:23:16 +04:00
/*
* Figure out what the alignment of the objects will be given a set of
* flags , a user specified alignment and the size of the objects .
*/
unsigned long calculate_alignment ( unsigned long flags ,
unsigned long align , unsigned long size )
{
/*
* If the user wants hardware cache aligned objects then follow that
* suggestion if the object is sufficiently large .
*
* The hardware cache alignment cannot override the specified
* alignment though . If that is greater then use it .
*/
if ( flags & SLAB_HWCACHE_ALIGN ) {
unsigned long ralign = cache_line_size ( ) ;
while ( size < = ralign / 2 )
ralign / = 2 ;
align = max ( align , ralign ) ;
}
if ( align < ARCH_SLAB_MINALIGN )
align = ARCH_SLAB_MINALIGN ;
return ALIGN ( align , sizeof ( void * ) ) ;
}
2014-04-08 02:39:26 +04:00
static struct kmem_cache *
do_kmem_cache_create ( char * name , size_t object_size , size_t size , size_t align ,
unsigned long flags , void ( * ctor ) ( void * ) ,
struct mem_cgroup * memcg , struct kmem_cache * root_cache )
{
struct kmem_cache * s ;
int err ;
err = - ENOMEM ;
s = kmem_cache_zalloc ( kmem_cache , GFP_KERNEL ) ;
if ( ! s )
goto out ;
s - > name = name ;
s - > object_size = object_size ;
s - > size = size ;
s - > align = align ;
s - > ctor = ctor ;
err = memcg_alloc_cache_params ( memcg , s , root_cache ) ;
if ( err )
goto out_free_cache ;
err = __kmem_cache_create ( s , flags ) ;
if ( err )
goto out_free_cache ;
s - > refcount = 1 ;
list_add ( & s - > list , & slab_caches ) ;
out :
if ( err )
return ERR_PTR ( err ) ;
return s ;
out_free_cache :
memcg_free_cache_params ( s ) ;
kfree ( s ) ;
goto out ;
}
2012-11-28 20:23:16 +04:00
2012-08-16 11:09:46 +04:00
/*
* kmem_cache_create - Create a cache .
* @ name : A string which is used in / proc / slabinfo to identify this cache .
* @ size : The size of objects to be created in this cache .
* @ align : The required alignment for the objects .
* @ flags : SLAB flags
* @ ctor : A constructor for the objects .
*
* Returns a ptr to the cache on success , NULL on failure .
* Cannot be called within a interrupt , but can be interrupted .
* The @ ctor is run when new pages are allocated by the cache .
*
* The flags are
*
* % SLAB_POISON - Poison the slab with a known test pattern ( a5a5a5a5 )
* to catch references to uninitialised memory .
*
* % SLAB_RED_ZONE - Insert ` Red ' zones around the allocated memory to check
* for buffer overruns .
*
* % SLAB_HWCACHE_ALIGN - Align the objects in this cache to a hardware
* cacheline . This can be beneficial if you ' re counting cycles as closely
* as davem .
*/
2012-12-19 02:22:34 +04:00
struct kmem_cache *
2014-04-08 02:39:26 +04:00
kmem_cache_create ( const char * name , size_t size , size_t align ,
unsigned long flags , void ( * ctor ) ( void * ) )
2012-08-16 11:09:46 +04:00
{
2014-04-08 02:39:26 +04:00
struct kmem_cache * s ;
char * cache_name ;
2014-01-24 03:52:55 +04:00
int err ;
2012-07-07 00:25:10 +04:00
2012-08-16 11:09:46 +04:00
get_online_cpus ( ) ;
slab: get_online_mems for kmem_cache_{create,destroy,shrink}
When we create a sl[au]b cache, we allocate kmem_cache_node structures
for each online NUMA node. To handle nodes taken online/offline, we
register memory hotplug notifier and allocate/free kmem_cache_node
corresponding to the node that changes its state for each kmem cache.
To synchronize between the two paths we hold the slab_mutex during both
the cache creationg/destruction path and while tuning per-node parts of
kmem caches in memory hotplug handler, but that's not quite right,
because it does not guarantee that a newly created cache will have all
kmem_cache_nodes initialized in case it races with memory hotplug. For
instance, in case of slub:
CPU0 CPU1
---- ----
kmem_cache_create: online_pages:
__kmem_cache_create: slab_memory_callback:
slab_mem_going_online_callback:
lock slab_mutex
for each slab_caches list entry
allocate kmem_cache node
unlock slab_mutex
lock slab_mutex
init_kmem_cache_nodes:
for_each_node_state(node, N_NORMAL_MEMORY)
allocate kmem_cache node
add kmem_cache to slab_caches list
unlock slab_mutex
online_pages (continued):
node_states_set_node
As a result we'll get a kmem cache with not all kmem_cache_nodes
allocated.
To avoid issues like that we should hold get/put_online_mems() during
the whole kmem cache creation/destruction/shrink paths, just like we
deal with cpu hotplug. This patch does the trick.
Note, that after it's applied, there is no need in taking the slab_mutex
for kmem_cache_shrink any more, so it is removed from there.
Signed-off-by: Vladimir Davydov <vdavydov@parallels.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Tang Chen <tangchen@cn.fujitsu.com>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Toshi Kani <toshi.kani@hp.com>
Cc: Xishi Qiu <qiuxishi@huawei.com>
Cc: Jiang Liu <liuj97@gmail.com>
Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Wen Congyang <wency@cn.fujitsu.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-06-05 03:07:20 +04:00
get_online_mems ( ) ;
2012-08-16 11:09:46 +04:00
mutex_lock ( & slab_mutex ) ;
2012-09-05 04:20:33 +04:00
2014-04-08 02:39:26 +04:00
err = kmem_cache_sanity_check ( name , size ) ;
2014-10-10 02:25:58 +04:00
if ( err ) {
s = NULL ; /* suppress uninit var warning */
2014-01-24 03:52:55 +04:00
goto out_unlock ;
2014-10-10 02:25:58 +04:00
}
2012-09-05 04:20:33 +04:00
2012-10-17 15:36:51 +04:00
/*
* Some allocators will constraint the set of valid flags to a subset
* of all flags . We expect them to define CACHE_CREATE_MASK in this
* case , and we ' ll just provide them with a sanitized version of the
* passed flags .
*/
flags & = CACHE_CREATE_MASK ;
2012-09-05 04:20:33 +04:00
2014-04-08 02:39:26 +04:00
s = __kmem_cache_alias ( name , size , align , flags , ctor ) ;
if ( s )
2014-01-24 03:52:55 +04:00
goto out_unlock ;
2012-12-19 02:22:34 +04:00
2014-04-08 02:39:26 +04:00
cache_name = kstrdup ( name , GFP_KERNEL ) ;
if ( ! cache_name ) {
err = - ENOMEM ;
goto out_unlock ;
}
2012-09-05 03:38:33 +04:00
2014-04-08 02:39:26 +04:00
s = do_kmem_cache_create ( cache_name , size , size ,
calculate_alignment ( flags , align , size ) ,
flags , ctor , NULL , NULL ) ;
if ( IS_ERR ( s ) ) {
err = PTR_ERR ( s ) ;
kfree ( cache_name ) ;
}
2014-01-24 03:52:55 +04:00
out_unlock :
2012-07-07 00:25:13 +04:00
mutex_unlock ( & slab_mutex ) ;
slab: get_online_mems for kmem_cache_{create,destroy,shrink}
When we create a sl[au]b cache, we allocate kmem_cache_node structures
for each online NUMA node. To handle nodes taken online/offline, we
register memory hotplug notifier and allocate/free kmem_cache_node
corresponding to the node that changes its state for each kmem cache.
To synchronize between the two paths we hold the slab_mutex during both
the cache creationg/destruction path and while tuning per-node parts of
kmem caches in memory hotplug handler, but that's not quite right,
because it does not guarantee that a newly created cache will have all
kmem_cache_nodes initialized in case it races with memory hotplug. For
instance, in case of slub:
CPU0 CPU1
---- ----
kmem_cache_create: online_pages:
__kmem_cache_create: slab_memory_callback:
slab_mem_going_online_callback:
lock slab_mutex
for each slab_caches list entry
allocate kmem_cache node
unlock slab_mutex
lock slab_mutex
init_kmem_cache_nodes:
for_each_node_state(node, N_NORMAL_MEMORY)
allocate kmem_cache node
add kmem_cache to slab_caches list
unlock slab_mutex
online_pages (continued):
node_states_set_node
As a result we'll get a kmem cache with not all kmem_cache_nodes
allocated.
To avoid issues like that we should hold get/put_online_mems() during
the whole kmem cache creation/destruction/shrink paths, just like we
deal with cpu hotplug. This patch does the trick.
Note, that after it's applied, there is no need in taking the slab_mutex
for kmem_cache_shrink any more, so it is removed from there.
Signed-off-by: Vladimir Davydov <vdavydov@parallels.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Tang Chen <tangchen@cn.fujitsu.com>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Toshi Kani <toshi.kani@hp.com>
Cc: Xishi Qiu <qiuxishi@huawei.com>
Cc: Jiang Liu <liuj97@gmail.com>
Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Wen Congyang <wency@cn.fujitsu.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-06-05 03:07:20 +04:00
put_online_mems ( ) ;
2012-07-07 00:25:13 +04:00
put_online_cpus ( ) ;
slab: fix wrong retval on kmem_cache_create_memcg error path
On kmem_cache_create_memcg() error path we set 'err', but leave 's' (the
new cache ptr) undefined. The latter can be NULL if we could not
allocate the cache, or pointing to a freed area if we failed somewhere
later while trying to initialize it. Initially we checked 'err'
immediately before exiting the function and returned NULL if it was set
ignoring the value of 's':
out_unlock:
...
if (err) {
/* report error */
return NULL;
}
return s;
Recently this check was, in fact, broken by commit f717eb3abb5e ("slab:
do not panic if we fail to create memcg cache"), which turned it to:
out_unlock:
...
if (err && !memcg) {
/* report error */
return NULL;
}
return s;
As a result, if we are failing creating a cache for a memcg, we will
skip the check and return 's' that can contain crap. Obviously, commit
f717eb3abb5e intended not to return crap on error allocating a cache for
a memcg, but only to remove the error reporting in this case, so the
check should look like this:
out_unlock:
...
if (err) {
if (!memcg)
return NULL;
/* report error */
return NULL;
}
return s;
[rientjes@google.com: despaghettification]
[vdavydov@parallels.com: patch monkeying]
Signed-off-by: David Rientjes <rientjes@google.com>
Signed-off-by: Vladimir Davydov <vdavydov@parallels.com>
Signed-off-by: Dave Jones <davej@redhat.com>
Reported-by: Dave Jones <davej@redhat.com>
Acked-by: Pekka Enberg <penberg@kernel.org>
Cc: Christoph Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-01-30 02:05:48 +04:00
if ( err ) {
2012-09-05 04:20:33 +04:00
if ( flags & SLAB_PANIC )
panic ( " kmem_cache_create: Failed to create slab '%s'. Error %d \n " ,
name , err ) ;
else {
printk ( KERN_WARNING " kmem_cache_create(%s) failed with error %d " ,
name , err ) ;
dump_stack ( ) ;
}
return NULL ;
}
2012-07-07 00:25:10 +04:00
return s ;
}
2014-04-08 02:39:26 +04:00
EXPORT_SYMBOL ( kmem_cache_create ) ;
2012-12-19 02:22:34 +04:00
2014-04-08 02:39:26 +04:00
# ifdef CONFIG_MEMCG_KMEM
/*
2014-06-05 03:10:02 +04:00
* memcg_create_kmem_cache - Create a cache for a memory cgroup .
2014-04-08 02:39:26 +04:00
* @ memcg : The memory cgroup the new cache is for .
* @ root_cache : The parent of the new cache .
2014-06-05 03:08:23 +04:00
* @ memcg_name : The name of the memory cgroup ( used for naming the new cache ) .
2014-04-08 02:39:26 +04:00
*
* This function attempts to create a kmem cache that will serve allocation
* requests going from @ memcg to @ root_cache . The new cache inherits properties
* from its parent .
*/
2014-06-05 03:10:02 +04:00
struct kmem_cache * memcg_create_kmem_cache ( struct mem_cgroup * memcg ,
2014-06-05 03:08:23 +04:00
struct kmem_cache * root_cache ,
const char * memcg_name )
2012-12-19 02:22:34 +04:00
{
memcg, slab: simplify synchronization scheme
At present, we have the following mutexes protecting data related to per
memcg kmem caches:
- slab_mutex. This one is held during the whole kmem cache creation
and destruction paths. We also take it when updating per root cache
memcg_caches arrays (see memcg_update_all_caches). As a result, taking
it guarantees there will be no changes to any kmem cache (including per
memcg). Why do we need something else then? The point is it is
private to slab implementation and has some internal dependencies with
other mutexes (get_online_cpus). So we just don't want to rely upon it
and prefer to introduce additional mutexes instead.
- activate_kmem_mutex. Initially it was added to synchronize
initializing kmem limit (memcg_activate_kmem). However, since we can
grow per root cache memcg_caches arrays only on kmem limit
initialization (see memcg_update_all_caches), we also employ it to
protect against memcg_caches arrays relocation (e.g. see
__kmem_cache_destroy_memcg_children).
- We have a convention not to take slab_mutex in memcontrol.c, but we
want to walk over per memcg memcg_slab_caches lists there (e.g. for
destroying all memcg caches on offline). So we have per memcg
slab_caches_mutex's protecting those lists.
The mutexes are taken in the following order:
activate_kmem_mutex -> slab_mutex -> memcg::slab_caches_mutex
Such a syncrhonization scheme has a number of flaws, for instance:
- We can't call kmem_cache_{destroy,shrink} while walking over a
memcg::memcg_slab_caches list due to locking order. As a result, in
mem_cgroup_destroy_all_caches we schedule the
memcg_cache_params::destroy work shrinking and destroying the cache.
- We don't have a mutex to synchronize per memcg caches destruction
between memcg offline (mem_cgroup_destroy_all_caches) and root cache
destruction (__kmem_cache_destroy_memcg_children). Currently we just
don't bother about it.
This patch simplifies it by substituting per memcg slab_caches_mutex's
with the global memcg_slab_mutex. It will be held whenever a new per
memcg cache is created or destroyed, so it protects per root cache
memcg_caches arrays and per memcg memcg_slab_caches lists. The locking
order is following:
activate_kmem_mutex -> memcg_slab_mutex -> slab_mutex
This allows us to call kmem_cache_{create,shrink,destroy} under the
memcg_slab_mutex. As a result, we don't need memcg_cache_params::destroy
work any more - we can simply destroy caches while iterating over a per
memcg slab caches list.
Also using the global mutex simplifies synchronization between concurrent
per memcg caches creation/destruction, e.g. mem_cgroup_destroy_all_caches
vs __kmem_cache_destroy_memcg_children.
The downside of this is that we substitute per-memcg slab_caches_mutex's
with a hummer-like global mutex, but since we already take either the
slab_mutex or the cgroup_mutex along with a memcg::slab_caches_mutex, it
shouldn't hurt concurrency a lot.
Signed-off-by: Vladimir Davydov <vdavydov@parallels.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Glauber Costa <glommer@gmail.com>
Cc: Pekka Enberg <penberg@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-06-05 03:07:40 +04:00
struct kmem_cache * s = NULL ;
2014-04-08 02:39:26 +04:00
char * cache_name ;
get_online_cpus ( ) ;
slab: get_online_mems for kmem_cache_{create,destroy,shrink}
When we create a sl[au]b cache, we allocate kmem_cache_node structures
for each online NUMA node. To handle nodes taken online/offline, we
register memory hotplug notifier and allocate/free kmem_cache_node
corresponding to the node that changes its state for each kmem cache.
To synchronize between the two paths we hold the slab_mutex during both
the cache creationg/destruction path and while tuning per-node parts of
kmem caches in memory hotplug handler, but that's not quite right,
because it does not guarantee that a newly created cache will have all
kmem_cache_nodes initialized in case it races with memory hotplug. For
instance, in case of slub:
CPU0 CPU1
---- ----
kmem_cache_create: online_pages:
__kmem_cache_create: slab_memory_callback:
slab_mem_going_online_callback:
lock slab_mutex
for each slab_caches list entry
allocate kmem_cache node
unlock slab_mutex
lock slab_mutex
init_kmem_cache_nodes:
for_each_node_state(node, N_NORMAL_MEMORY)
allocate kmem_cache node
add kmem_cache to slab_caches list
unlock slab_mutex
online_pages (continued):
node_states_set_node
As a result we'll get a kmem cache with not all kmem_cache_nodes
allocated.
To avoid issues like that we should hold get/put_online_mems() during
the whole kmem cache creation/destruction/shrink paths, just like we
deal with cpu hotplug. This patch does the trick.
Note, that after it's applied, there is no need in taking the slab_mutex
for kmem_cache_shrink any more, so it is removed from there.
Signed-off-by: Vladimir Davydov <vdavydov@parallels.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Tang Chen <tangchen@cn.fujitsu.com>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Toshi Kani <toshi.kani@hp.com>
Cc: Xishi Qiu <qiuxishi@huawei.com>
Cc: Jiang Liu <liuj97@gmail.com>
Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Wen Congyang <wency@cn.fujitsu.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-06-05 03:07:20 +04:00
get_online_mems ( ) ;
2014-04-08 02:39:26 +04:00
mutex_lock ( & slab_mutex ) ;
2014-06-05 03:08:23 +04:00
cache_name = kasprintf ( GFP_KERNEL , " %s(%d:%s) " , root_cache - > name ,
memcg_cache_id ( memcg ) , memcg_name ) ;
2014-04-08 02:39:26 +04:00
if ( ! cache_name )
goto out_unlock ;
s = do_kmem_cache_create ( cache_name , root_cache - > object_size ,
root_cache - > size , root_cache - > align ,
root_cache - > flags , root_cache - > ctor ,
memcg , root_cache ) ;
memcg, slab: simplify synchronization scheme
At present, we have the following mutexes protecting data related to per
memcg kmem caches:
- slab_mutex. This one is held during the whole kmem cache creation
and destruction paths. We also take it when updating per root cache
memcg_caches arrays (see memcg_update_all_caches). As a result, taking
it guarantees there will be no changes to any kmem cache (including per
memcg). Why do we need something else then? The point is it is
private to slab implementation and has some internal dependencies with
other mutexes (get_online_cpus). So we just don't want to rely upon it
and prefer to introduce additional mutexes instead.
- activate_kmem_mutex. Initially it was added to synchronize
initializing kmem limit (memcg_activate_kmem). However, since we can
grow per root cache memcg_caches arrays only on kmem limit
initialization (see memcg_update_all_caches), we also employ it to
protect against memcg_caches arrays relocation (e.g. see
__kmem_cache_destroy_memcg_children).
- We have a convention not to take slab_mutex in memcontrol.c, but we
want to walk over per memcg memcg_slab_caches lists there (e.g. for
destroying all memcg caches on offline). So we have per memcg
slab_caches_mutex's protecting those lists.
The mutexes are taken in the following order:
activate_kmem_mutex -> slab_mutex -> memcg::slab_caches_mutex
Such a syncrhonization scheme has a number of flaws, for instance:
- We can't call kmem_cache_{destroy,shrink} while walking over a
memcg::memcg_slab_caches list due to locking order. As a result, in
mem_cgroup_destroy_all_caches we schedule the
memcg_cache_params::destroy work shrinking and destroying the cache.
- We don't have a mutex to synchronize per memcg caches destruction
between memcg offline (mem_cgroup_destroy_all_caches) and root cache
destruction (__kmem_cache_destroy_memcg_children). Currently we just
don't bother about it.
This patch simplifies it by substituting per memcg slab_caches_mutex's
with the global memcg_slab_mutex. It will be held whenever a new per
memcg cache is created or destroyed, so it protects per root cache
memcg_caches arrays and per memcg memcg_slab_caches lists. The locking
order is following:
activate_kmem_mutex -> memcg_slab_mutex -> slab_mutex
This allows us to call kmem_cache_{create,shrink,destroy} under the
memcg_slab_mutex. As a result, we don't need memcg_cache_params::destroy
work any more - we can simply destroy caches while iterating over a per
memcg slab caches list.
Also using the global mutex simplifies synchronization between concurrent
per memcg caches creation/destruction, e.g. mem_cgroup_destroy_all_caches
vs __kmem_cache_destroy_memcg_children.
The downside of this is that we substitute per-memcg slab_caches_mutex's
with a hummer-like global mutex, but since we already take either the
slab_mutex or the cgroup_mutex along with a memcg::slab_caches_mutex, it
shouldn't hurt concurrency a lot.
Signed-off-by: Vladimir Davydov <vdavydov@parallels.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Glauber Costa <glommer@gmail.com>
Cc: Pekka Enberg <penberg@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-06-05 03:07:40 +04:00
if ( IS_ERR ( s ) ) {
2014-04-08 02:39:26 +04:00
kfree ( cache_name ) ;
memcg, slab: simplify synchronization scheme
At present, we have the following mutexes protecting data related to per
memcg kmem caches:
- slab_mutex. This one is held during the whole kmem cache creation
and destruction paths. We also take it when updating per root cache
memcg_caches arrays (see memcg_update_all_caches). As a result, taking
it guarantees there will be no changes to any kmem cache (including per
memcg). Why do we need something else then? The point is it is
private to slab implementation and has some internal dependencies with
other mutexes (get_online_cpus). So we just don't want to rely upon it
and prefer to introduce additional mutexes instead.
- activate_kmem_mutex. Initially it was added to synchronize
initializing kmem limit (memcg_activate_kmem). However, since we can
grow per root cache memcg_caches arrays only on kmem limit
initialization (see memcg_update_all_caches), we also employ it to
protect against memcg_caches arrays relocation (e.g. see
__kmem_cache_destroy_memcg_children).
- We have a convention not to take slab_mutex in memcontrol.c, but we
want to walk over per memcg memcg_slab_caches lists there (e.g. for
destroying all memcg caches on offline). So we have per memcg
slab_caches_mutex's protecting those lists.
The mutexes are taken in the following order:
activate_kmem_mutex -> slab_mutex -> memcg::slab_caches_mutex
Such a syncrhonization scheme has a number of flaws, for instance:
- We can't call kmem_cache_{destroy,shrink} while walking over a
memcg::memcg_slab_caches list due to locking order. As a result, in
mem_cgroup_destroy_all_caches we schedule the
memcg_cache_params::destroy work shrinking and destroying the cache.
- We don't have a mutex to synchronize per memcg caches destruction
between memcg offline (mem_cgroup_destroy_all_caches) and root cache
destruction (__kmem_cache_destroy_memcg_children). Currently we just
don't bother about it.
This patch simplifies it by substituting per memcg slab_caches_mutex's
with the global memcg_slab_mutex. It will be held whenever a new per
memcg cache is created or destroyed, so it protects per root cache
memcg_caches arrays and per memcg memcg_slab_caches lists. The locking
order is following:
activate_kmem_mutex -> memcg_slab_mutex -> slab_mutex
This allows us to call kmem_cache_{create,shrink,destroy} under the
memcg_slab_mutex. As a result, we don't need memcg_cache_params::destroy
work any more - we can simply destroy caches while iterating over a per
memcg slab caches list.
Also using the global mutex simplifies synchronization between concurrent
per memcg caches creation/destruction, e.g. mem_cgroup_destroy_all_caches
vs __kmem_cache_destroy_memcg_children.
The downside of this is that we substitute per-memcg slab_caches_mutex's
with a hummer-like global mutex, but since we already take either the
slab_mutex or the cgroup_mutex along with a memcg::slab_caches_mutex, it
shouldn't hurt concurrency a lot.
Signed-off-by: Vladimir Davydov <vdavydov@parallels.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Glauber Costa <glommer@gmail.com>
Cc: Pekka Enberg <penberg@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-06-05 03:07:40 +04:00
s = NULL ;
}
2014-04-08 02:39:26 +04:00
out_unlock :
mutex_unlock ( & slab_mutex ) ;
slab: get_online_mems for kmem_cache_{create,destroy,shrink}
When we create a sl[au]b cache, we allocate kmem_cache_node structures
for each online NUMA node. To handle nodes taken online/offline, we
register memory hotplug notifier and allocate/free kmem_cache_node
corresponding to the node that changes its state for each kmem cache.
To synchronize between the two paths we hold the slab_mutex during both
the cache creationg/destruction path and while tuning per-node parts of
kmem caches in memory hotplug handler, but that's not quite right,
because it does not guarantee that a newly created cache will have all
kmem_cache_nodes initialized in case it races with memory hotplug. For
instance, in case of slub:
CPU0 CPU1
---- ----
kmem_cache_create: online_pages:
__kmem_cache_create: slab_memory_callback:
slab_mem_going_online_callback:
lock slab_mutex
for each slab_caches list entry
allocate kmem_cache node
unlock slab_mutex
lock slab_mutex
init_kmem_cache_nodes:
for_each_node_state(node, N_NORMAL_MEMORY)
allocate kmem_cache node
add kmem_cache to slab_caches list
unlock slab_mutex
online_pages (continued):
node_states_set_node
As a result we'll get a kmem cache with not all kmem_cache_nodes
allocated.
To avoid issues like that we should hold get/put_online_mems() during
the whole kmem cache creation/destruction/shrink paths, just like we
deal with cpu hotplug. This patch does the trick.
Note, that after it's applied, there is no need in taking the slab_mutex
for kmem_cache_shrink any more, so it is removed from there.
Signed-off-by: Vladimir Davydov <vdavydov@parallels.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Tang Chen <tangchen@cn.fujitsu.com>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Toshi Kani <toshi.kani@hp.com>
Cc: Xishi Qiu <qiuxishi@huawei.com>
Cc: Jiang Liu <liuj97@gmail.com>
Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Wen Congyang <wency@cn.fujitsu.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-06-05 03:07:20 +04:00
put_online_mems ( ) ;
2014-04-08 02:39:26 +04:00
put_online_cpus ( ) ;
memcg, slab: simplify synchronization scheme
At present, we have the following mutexes protecting data related to per
memcg kmem caches:
- slab_mutex. This one is held during the whole kmem cache creation
and destruction paths. We also take it when updating per root cache
memcg_caches arrays (see memcg_update_all_caches). As a result, taking
it guarantees there will be no changes to any kmem cache (including per
memcg). Why do we need something else then? The point is it is
private to slab implementation and has some internal dependencies with
other mutexes (get_online_cpus). So we just don't want to rely upon it
and prefer to introduce additional mutexes instead.
- activate_kmem_mutex. Initially it was added to synchronize
initializing kmem limit (memcg_activate_kmem). However, since we can
grow per root cache memcg_caches arrays only on kmem limit
initialization (see memcg_update_all_caches), we also employ it to
protect against memcg_caches arrays relocation (e.g. see
__kmem_cache_destroy_memcg_children).
- We have a convention not to take slab_mutex in memcontrol.c, but we
want to walk over per memcg memcg_slab_caches lists there (e.g. for
destroying all memcg caches on offline). So we have per memcg
slab_caches_mutex's protecting those lists.
The mutexes are taken in the following order:
activate_kmem_mutex -> slab_mutex -> memcg::slab_caches_mutex
Such a syncrhonization scheme has a number of flaws, for instance:
- We can't call kmem_cache_{destroy,shrink} while walking over a
memcg::memcg_slab_caches list due to locking order. As a result, in
mem_cgroup_destroy_all_caches we schedule the
memcg_cache_params::destroy work shrinking and destroying the cache.
- We don't have a mutex to synchronize per memcg caches destruction
between memcg offline (mem_cgroup_destroy_all_caches) and root cache
destruction (__kmem_cache_destroy_memcg_children). Currently we just
don't bother about it.
This patch simplifies it by substituting per memcg slab_caches_mutex's
with the global memcg_slab_mutex. It will be held whenever a new per
memcg cache is created or destroyed, so it protects per root cache
memcg_caches arrays and per memcg memcg_slab_caches lists. The locking
order is following:
activate_kmem_mutex -> memcg_slab_mutex -> slab_mutex
This allows us to call kmem_cache_{create,shrink,destroy} under the
memcg_slab_mutex. As a result, we don't need memcg_cache_params::destroy
work any more - we can simply destroy caches while iterating over a per
memcg slab caches list.
Also using the global mutex simplifies synchronization between concurrent
per memcg caches creation/destruction, e.g. mem_cgroup_destroy_all_caches
vs __kmem_cache_destroy_memcg_children.
The downside of this is that we substitute per-memcg slab_caches_mutex's
with a hummer-like global mutex, but since we already take either the
slab_mutex or the cgroup_mutex along with a memcg::slab_caches_mutex, it
shouldn't hurt concurrency a lot.
Signed-off-by: Vladimir Davydov <vdavydov@parallels.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Glauber Costa <glommer@gmail.com>
Cc: Pekka Enberg <penberg@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-06-05 03:07:40 +04:00
return s ;
2012-12-19 02:22:34 +04:00
}
2014-04-08 02:39:28 +04:00
2014-06-05 03:10:02 +04:00
static int memcg_cleanup_cache_params ( struct kmem_cache * s )
2014-04-08 02:39:28 +04:00
{
int rc ;
if ( ! s - > memcg_params | |
! s - > memcg_params - > is_root_cache )
return 0 ;
mutex_unlock ( & slab_mutex ) ;
2014-06-05 03:10:02 +04:00
rc = __memcg_cleanup_cache_params ( s ) ;
2014-04-08 02:39:28 +04:00
mutex_lock ( & slab_mutex ) ;
return rc ;
}
# else
2014-06-05 03:10:02 +04:00
static int memcg_cleanup_cache_params ( struct kmem_cache * s )
2014-04-08 02:39:28 +04:00
{
return 0 ;
}
2014-04-08 02:39:26 +04:00
# endif /* CONFIG_MEMCG_KMEM */
2012-07-07 00:25:11 +04:00
2014-05-06 23:50:08 +04:00
void slab_kmem_cache_release ( struct kmem_cache * s )
{
kfree ( s - > name ) ;
kmem_cache_free ( kmem_cache , s ) ;
}
2012-09-05 03:18:33 +04:00
void kmem_cache_destroy ( struct kmem_cache * s )
{
get_online_cpus ( ) ;
slab: get_online_mems for kmem_cache_{create,destroy,shrink}
When we create a sl[au]b cache, we allocate kmem_cache_node structures
for each online NUMA node. To handle nodes taken online/offline, we
register memory hotplug notifier and allocate/free kmem_cache_node
corresponding to the node that changes its state for each kmem cache.
To synchronize between the two paths we hold the slab_mutex during both
the cache creationg/destruction path and while tuning per-node parts of
kmem caches in memory hotplug handler, but that's not quite right,
because it does not guarantee that a newly created cache will have all
kmem_cache_nodes initialized in case it races with memory hotplug. For
instance, in case of slub:
CPU0 CPU1
---- ----
kmem_cache_create: online_pages:
__kmem_cache_create: slab_memory_callback:
slab_mem_going_online_callback:
lock slab_mutex
for each slab_caches list entry
allocate kmem_cache node
unlock slab_mutex
lock slab_mutex
init_kmem_cache_nodes:
for_each_node_state(node, N_NORMAL_MEMORY)
allocate kmem_cache node
add kmem_cache to slab_caches list
unlock slab_mutex
online_pages (continued):
node_states_set_node
As a result we'll get a kmem cache with not all kmem_cache_nodes
allocated.
To avoid issues like that we should hold get/put_online_mems() during
the whole kmem cache creation/destruction/shrink paths, just like we
deal with cpu hotplug. This patch does the trick.
Note, that after it's applied, there is no need in taking the slab_mutex
for kmem_cache_shrink any more, so it is removed from there.
Signed-off-by: Vladimir Davydov <vdavydov@parallels.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Tang Chen <tangchen@cn.fujitsu.com>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Toshi Kani <toshi.kani@hp.com>
Cc: Xishi Qiu <qiuxishi@huawei.com>
Cc: Jiang Liu <liuj97@gmail.com>
Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Wen Congyang <wency@cn.fujitsu.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-06-05 03:07:20 +04:00
get_online_mems ( ) ;
2012-09-05 03:18:33 +04:00
mutex_lock ( & slab_mutex ) ;
2014-04-08 02:39:28 +04:00
2012-09-05 03:18:33 +04:00
s - > refcount - - ;
2014-04-08 02:39:28 +04:00
if ( s - > refcount )
goto out_unlock ;
2014-06-05 03:10:02 +04:00
if ( memcg_cleanup_cache_params ( s ) ! = 0 )
2014-04-08 02:39:28 +04:00
goto out_unlock ;
if ( __kmem_cache_shutdown ( s ) ! = 0 ) {
printk ( KERN_ERR " kmem_cache_destroy %s: "
" Slab cache still has objects \n " , s - > name ) ;
dump_stack ( ) ;
goto out_unlock ;
2012-09-05 03:18:33 +04:00
}
2014-04-08 02:39:28 +04:00
2014-06-05 03:10:03 +04:00
list_del ( & s - > list ) ;
2014-04-08 02:39:28 +04:00
mutex_unlock ( & slab_mutex ) ;
if ( s - > flags & SLAB_DESTROY_BY_RCU )
rcu_barrier ( ) ;
memcg_free_cache_params ( s ) ;
2014-05-06 23:50:08 +04:00
# ifdef SLAB_SUPPORTS_SYSFS
sysfs_slab_remove ( s ) ;
# else
slab_kmem_cache_release ( s ) ;
# endif
slab: get_online_mems for kmem_cache_{create,destroy,shrink}
When we create a sl[au]b cache, we allocate kmem_cache_node structures
for each online NUMA node. To handle nodes taken online/offline, we
register memory hotplug notifier and allocate/free kmem_cache_node
corresponding to the node that changes its state for each kmem cache.
To synchronize between the two paths we hold the slab_mutex during both
the cache creationg/destruction path and while tuning per-node parts of
kmem caches in memory hotplug handler, but that's not quite right,
because it does not guarantee that a newly created cache will have all
kmem_cache_nodes initialized in case it races with memory hotplug. For
instance, in case of slub:
CPU0 CPU1
---- ----
kmem_cache_create: online_pages:
__kmem_cache_create: slab_memory_callback:
slab_mem_going_online_callback:
lock slab_mutex
for each slab_caches list entry
allocate kmem_cache node
unlock slab_mutex
lock slab_mutex
init_kmem_cache_nodes:
for_each_node_state(node, N_NORMAL_MEMORY)
allocate kmem_cache node
add kmem_cache to slab_caches list
unlock slab_mutex
online_pages (continued):
node_states_set_node
As a result we'll get a kmem cache with not all kmem_cache_nodes
allocated.
To avoid issues like that we should hold get/put_online_mems() during
the whole kmem cache creation/destruction/shrink paths, just like we
deal with cpu hotplug. This patch does the trick.
Note, that after it's applied, there is no need in taking the slab_mutex
for kmem_cache_shrink any more, so it is removed from there.
Signed-off-by: Vladimir Davydov <vdavydov@parallels.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Tang Chen <tangchen@cn.fujitsu.com>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Toshi Kani <toshi.kani@hp.com>
Cc: Xishi Qiu <qiuxishi@huawei.com>
Cc: Jiang Liu <liuj97@gmail.com>
Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Wen Congyang <wency@cn.fujitsu.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-06-05 03:07:20 +04:00
goto out ;
2014-04-08 02:39:28 +04:00
out_unlock :
mutex_unlock ( & slab_mutex ) ;
slab: get_online_mems for kmem_cache_{create,destroy,shrink}
When we create a sl[au]b cache, we allocate kmem_cache_node structures
for each online NUMA node. To handle nodes taken online/offline, we
register memory hotplug notifier and allocate/free kmem_cache_node
corresponding to the node that changes its state for each kmem cache.
To synchronize between the two paths we hold the slab_mutex during both
the cache creationg/destruction path and while tuning per-node parts of
kmem caches in memory hotplug handler, but that's not quite right,
because it does not guarantee that a newly created cache will have all
kmem_cache_nodes initialized in case it races with memory hotplug. For
instance, in case of slub:
CPU0 CPU1
---- ----
kmem_cache_create: online_pages:
__kmem_cache_create: slab_memory_callback:
slab_mem_going_online_callback:
lock slab_mutex
for each slab_caches list entry
allocate kmem_cache node
unlock slab_mutex
lock slab_mutex
init_kmem_cache_nodes:
for_each_node_state(node, N_NORMAL_MEMORY)
allocate kmem_cache node
add kmem_cache to slab_caches list
unlock slab_mutex
online_pages (continued):
node_states_set_node
As a result we'll get a kmem cache with not all kmem_cache_nodes
allocated.
To avoid issues like that we should hold get/put_online_mems() during
the whole kmem cache creation/destruction/shrink paths, just like we
deal with cpu hotplug. This patch does the trick.
Note, that after it's applied, there is no need in taking the slab_mutex
for kmem_cache_shrink any more, so it is removed from there.
Signed-off-by: Vladimir Davydov <vdavydov@parallels.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Tang Chen <tangchen@cn.fujitsu.com>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Toshi Kani <toshi.kani@hp.com>
Cc: Xishi Qiu <qiuxishi@huawei.com>
Cc: Jiang Liu <liuj97@gmail.com>
Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Wen Congyang <wency@cn.fujitsu.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-06-05 03:07:20 +04:00
out :
put_online_mems ( ) ;
2012-09-05 03:18:33 +04:00
put_online_cpus ( ) ;
}
EXPORT_SYMBOL ( kmem_cache_destroy ) ;
slab: get_online_mems for kmem_cache_{create,destroy,shrink}
When we create a sl[au]b cache, we allocate kmem_cache_node structures
for each online NUMA node. To handle nodes taken online/offline, we
register memory hotplug notifier and allocate/free kmem_cache_node
corresponding to the node that changes its state for each kmem cache.
To synchronize between the two paths we hold the slab_mutex during both
the cache creationg/destruction path and while tuning per-node parts of
kmem caches in memory hotplug handler, but that's not quite right,
because it does not guarantee that a newly created cache will have all
kmem_cache_nodes initialized in case it races with memory hotplug. For
instance, in case of slub:
CPU0 CPU1
---- ----
kmem_cache_create: online_pages:
__kmem_cache_create: slab_memory_callback:
slab_mem_going_online_callback:
lock slab_mutex
for each slab_caches list entry
allocate kmem_cache node
unlock slab_mutex
lock slab_mutex
init_kmem_cache_nodes:
for_each_node_state(node, N_NORMAL_MEMORY)
allocate kmem_cache node
add kmem_cache to slab_caches list
unlock slab_mutex
online_pages (continued):
node_states_set_node
As a result we'll get a kmem cache with not all kmem_cache_nodes
allocated.
To avoid issues like that we should hold get/put_online_mems() during
the whole kmem cache creation/destruction/shrink paths, just like we
deal with cpu hotplug. This patch does the trick.
Note, that after it's applied, there is no need in taking the slab_mutex
for kmem_cache_shrink any more, so it is removed from there.
Signed-off-by: Vladimir Davydov <vdavydov@parallels.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Tang Chen <tangchen@cn.fujitsu.com>
Cc: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Toshi Kani <toshi.kani@hp.com>
Cc: Xishi Qiu <qiuxishi@huawei.com>
Cc: Jiang Liu <liuj97@gmail.com>
Cc: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Wen Congyang <wency@cn.fujitsu.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2014-06-05 03:07:20 +04:00
/**
* kmem_cache_shrink - Shrink a cache .
* @ cachep : The cache to shrink .
*
* Releases as many slabs as possible for a cache .
* To help debugging , a zero exit status indicates all slabs were released .
*/
int kmem_cache_shrink ( struct kmem_cache * cachep )
{
int ret ;
get_online_cpus ( ) ;
get_online_mems ( ) ;
ret = __kmem_cache_shrink ( cachep ) ;
put_online_mems ( ) ;
put_online_cpus ( ) ;
return ret ;
}
EXPORT_SYMBOL ( kmem_cache_shrink ) ;
2012-07-07 00:25:11 +04:00
int slab_is_available ( void )
{
return slab_state > = UP ;
}
2012-10-19 18:20:25 +04:00
2012-11-28 20:23:07 +04:00
# ifndef CONFIG_SLOB
/* Create a cache during boot when no slab services are available yet */
void __init create_boot_cache ( struct kmem_cache * s , const char * name , size_t size ,
unsigned long flags )
{
int err ;
s - > name = name ;
s - > size = s - > object_size = size ;
2012-11-28 20:23:16 +04:00
s - > align = calculate_alignment ( flags , ARCH_KMALLOC_MINALIGN , size ) ;
2012-11-28 20:23:07 +04:00
err = __kmem_cache_create ( s , flags ) ;
if ( err )
2013-01-10 23:00:53 +04:00
panic ( " Creation of kmalloc slab %s size=%zu failed. Reason %d \n " ,
2012-11-28 20:23:07 +04:00
name , size , err ) ;
s - > refcount = - 1 ; /* Exempt from merging for now */
}
struct kmem_cache * __init create_kmalloc_cache ( const char * name , size_t size ,
unsigned long flags )
{
struct kmem_cache * s = kmem_cache_zalloc ( kmem_cache , GFP_NOWAIT ) ;
if ( ! s )
panic ( " Out of memory when creating slab %s \n " , name ) ;
create_boot_cache ( s , name , size , flags ) ;
list_add ( & s - > list , & slab_caches ) ;
s - > refcount = 1 ;
return s ;
}
2013-01-10 23:12:17 +04:00
struct kmem_cache * kmalloc_caches [ KMALLOC_SHIFT_HIGH + 1 ] ;
EXPORT_SYMBOL ( kmalloc_caches ) ;
# ifdef CONFIG_ZONE_DMA
struct kmem_cache * kmalloc_dma_caches [ KMALLOC_SHIFT_HIGH + 1 ] ;
EXPORT_SYMBOL ( kmalloc_dma_caches ) ;
# endif
2013-01-10 23:14:19 +04:00
/*
* Conversion table for small slabs sizes / 8 to the index in the
* kmalloc array . This is necessary for slabs < 192 since we have non power
* of two cache sizes there . The size of larger slabs can be determined using
* fls .
*/
static s8 size_index [ 24 ] = {
3 , /* 8 */
4 , /* 16 */
5 , /* 24 */
5 , /* 32 */
6 , /* 40 */
6 , /* 48 */
6 , /* 56 */
6 , /* 64 */
1 , /* 72 */
1 , /* 80 */
1 , /* 88 */
1 , /* 96 */
7 , /* 104 */
7 , /* 112 */
7 , /* 120 */
7 , /* 128 */
2 , /* 136 */
2 , /* 144 */
2 , /* 152 */
2 , /* 160 */
2 , /* 168 */
2 , /* 176 */
2 , /* 184 */
2 /* 192 */
} ;
static inline int size_index_elem ( size_t bytes )
{
return ( bytes - 1 ) / 8 ;
}
/*
* Find the kmem_cache structure that serves a given size of
* allocation
*/
struct kmem_cache * kmalloc_slab ( size_t size , gfp_t flags )
{
int index ;
2013-08-02 06:02:42 +04:00
if ( unlikely ( size > KMALLOC_MAX_SIZE ) ) {
2013-06-10 23:18:00 +04:00
WARN_ON_ONCE ( ! ( flags & __GFP_NOWARN ) ) ;
2013-05-03 19:43:18 +04:00
return NULL ;
2013-06-10 23:18:00 +04:00
}
2013-05-03 19:43:18 +04:00
2013-01-10 23:14:19 +04:00
if ( size < = 192 ) {
if ( ! size )
return ZERO_SIZE_PTR ;
index = size_index [ size_index_elem ( size ) ] ;
} else
index = fls ( size - 1 ) ;
# ifdef CONFIG_ZONE_DMA
2013-02-04 18:46:46 +04:00
if ( unlikely ( ( flags & GFP_DMA ) ) )
2013-01-10 23:14:19 +04:00
return kmalloc_dma_caches [ index ] ;
# endif
return kmalloc_caches [ index ] ;
}
2013-01-10 23:12:17 +04:00
/*
* Create the kmalloc array . Some of the regular kmalloc arrays
* may already have been created because they were needed to
* enable allocations for slab creation .
*/
void __init create_kmalloc_caches ( unsigned long flags )
{
int i ;
2013-01-10 23:14:19 +04:00
/*
* Patch up the size_index table if we have strange large alignment
* requirements for the kmalloc array . This is only the case for
* MIPS it seems . The standard arches will not generate any code here .
*
* Largest permitted alignment is 256 bytes due to the way we
* handle the index determination for the smaller caches .
*
* Make sure that nothing crazy happens if someone starts tinkering
* around with ARCH_KMALLOC_MINALIGN
*/
BUILD_BUG_ON ( KMALLOC_MIN_SIZE > 256 | |
( KMALLOC_MIN_SIZE & ( KMALLOC_MIN_SIZE - 1 ) ) ) ;
for ( i = 8 ; i < KMALLOC_MIN_SIZE ; i + = 8 ) {
int elem = size_index_elem ( i ) ;
if ( elem > = ARRAY_SIZE ( size_index ) )
break ;
size_index [ elem ] = KMALLOC_SHIFT_LOW ;
}
if ( KMALLOC_MIN_SIZE > = 64 ) {
/*
* The 96 byte size cache is not used if the alignment
* is 64 byte .
*/
for ( i = 64 + 8 ; i < = 96 ; i + = 8 )
size_index [ size_index_elem ( i ) ] = 7 ;
}
if ( KMALLOC_MIN_SIZE > = 128 ) {
/*
* The 192 byte sized cache is not used if the alignment
* is 128 byte . Redirect kmalloc to use the 256 byte cache
* instead .
*/
for ( i = 128 + 8 ; i < = 192 ; i + = 8 )
size_index [ size_index_elem ( i ) ] = 8 ;
}
2013-05-03 22:04:18 +04:00
for ( i = KMALLOC_SHIFT_LOW ; i < = KMALLOC_SHIFT_HIGH ; i + + ) {
if ( ! kmalloc_caches [ i ] ) {
2013-01-10 23:12:17 +04:00
kmalloc_caches [ i ] = create_kmalloc_cache ( NULL ,
1 < < i , flags ) ;
2013-05-08 23:56:28 +04:00
}
2013-01-10 23:12:17 +04:00
2013-05-08 23:56:28 +04:00
/*
* Caches that are not of the two - to - the - power - of size .
* These have to be created immediately after the
* earlier power of two caches
*/
if ( KMALLOC_MIN_SIZE < = 32 & & ! kmalloc_caches [ 1 ] & & i = = 6 )
kmalloc_caches [ 1 ] = create_kmalloc_cache ( NULL , 96 , flags ) ;
2013-05-03 22:04:18 +04:00
2013-05-08 23:56:28 +04:00
if ( KMALLOC_MIN_SIZE < = 64 & & ! kmalloc_caches [ 2 ] & & i = = 7 )
kmalloc_caches [ 2 ] = create_kmalloc_cache ( NULL , 192 , flags ) ;
2013-05-03 22:04:18 +04:00
}
2013-01-10 23:12:17 +04:00
/* Kmalloc array is now usable */
slab_state = UP ;
for ( i = 0 ; i < = KMALLOC_SHIFT_HIGH ; i + + ) {
struct kmem_cache * s = kmalloc_caches [ i ] ;
char * n ;
if ( s ) {
n = kasprintf ( GFP_NOWAIT , " kmalloc-%d " , kmalloc_size ( i ) ) ;
BUG_ON ( ! n ) ;
s - > name = n ;
}
}
# ifdef CONFIG_ZONE_DMA
for ( i = 0 ; i < = KMALLOC_SHIFT_HIGH ; i + + ) {
struct kmem_cache * s = kmalloc_caches [ i ] ;
if ( s ) {
int size = kmalloc_size ( i ) ;
char * n = kasprintf ( GFP_NOWAIT ,
" dma-kmalloc-%d " , size ) ;
BUG_ON ( ! n ) ;
kmalloc_dma_caches [ i ] = create_kmalloc_cache ( n ,
size , SLAB_CACHE_DMA | flags ) ;
}
}
# endif
}
2012-11-28 20:23:07 +04:00
# endif /* !CONFIG_SLOB */
2014-06-05 03:07:04 +04:00
/*
* To avoid unnecessary overhead , we pass through large allocation requests
* directly to the page allocator . We use __GFP_COMP , because we will need to
* know the allocation order to free the pages properly in kfree .
*/
2014-06-05 03:06:39 +04:00
void * kmalloc_order ( size_t size , gfp_t flags , unsigned int order )
{
void * ret ;
struct page * page ;
flags | = __GFP_COMP ;
page = alloc_kmem_pages ( flags , order ) ;
ret = page ? page_address ( page ) : NULL ;
kmemleak_alloc ( ret , size , 1 , flags ) ;
return ret ;
}
EXPORT_SYMBOL ( kmalloc_order ) ;
2013-09-04 20:35:34 +04:00
# ifdef CONFIG_TRACING
void * kmalloc_order_trace ( size_t size , gfp_t flags , unsigned int order )
{
void * ret = kmalloc_order ( size , flags , order ) ;
trace_kmalloc ( _RET_IP_ , ret , size , PAGE_SIZE < < order , flags ) ;
return ret ;
}
EXPORT_SYMBOL ( kmalloc_order_trace ) ;
# endif
2012-11-28 20:23:07 +04:00
2012-10-19 18:20:25 +04:00
# ifdef CONFIG_SLABINFO
2013-07-04 04:33:24 +04:00
# ifdef CONFIG_SLAB
# define SLABINFO_RIGHTS (S_IWUSR | S_IRUSR)
# else
# define SLABINFO_RIGHTS S_IRUSR
# endif
2012-12-19 02:23:01 +04:00
void print_slabinfo_header ( struct seq_file * m )
2012-10-19 18:20:26 +04:00
{
/*
* Output format version , so at least we can change it
* without _too_ many complaints .
*/
# ifdef CONFIG_DEBUG_SLAB
seq_puts ( m , " slabinfo - version: 2.1 (statistics) \n " ) ;
# else
seq_puts ( m , " slabinfo - version: 2.1 \n " ) ;
# endif
seq_puts ( m , " # name <active_objs> <num_objs> <objsize> "
" <objperslab> <pagesperslab> " ) ;
seq_puts ( m , " : tunables <limit> <batchcount> <sharedfactor> " ) ;
seq_puts ( m , " : slabdata <active_slabs> <num_slabs> <sharedavail> " ) ;
# ifdef CONFIG_DEBUG_SLAB
seq_puts ( m , " : globalstat <listallocs> <maxobjs> <grown> <reaped> "
" <error> <maxfreeable> <nodeallocs> <remotefrees> <alienoverflow> " ) ;
seq_puts ( m , " : cpustat <allochit> <allocmiss> <freehit> <freemiss> " ) ;
# endif
seq_putc ( m , ' \n ' ) ;
}
2012-10-19 18:20:25 +04:00
static void * s_start ( struct seq_file * m , loff_t * pos )
{
loff_t n = * pos ;
mutex_lock ( & slab_mutex ) ;
if ( ! n )
print_slabinfo_header ( m ) ;
return seq_list_start ( & slab_caches , * pos ) ;
}
2013-07-08 04:08:28 +04:00
void * slab_next ( struct seq_file * m , void * p , loff_t * pos )
2012-10-19 18:20:25 +04:00
{
return seq_list_next ( p , & slab_caches , pos ) ;
}
2013-07-08 04:08:28 +04:00
void slab_stop ( struct seq_file * m , void * p )
2012-10-19 18:20:25 +04:00
{
mutex_unlock ( & slab_mutex ) ;
}
2012-12-19 02:23:01 +04:00
static void
memcg_accumulate_slabinfo ( struct kmem_cache * s , struct slabinfo * info )
{
struct kmem_cache * c ;
struct slabinfo sinfo ;
int i ;
if ( ! is_root_cache ( s ) )
return ;
for_each_memcg_cache_index ( i ) {
2013-11-13 03:08:23 +04:00
c = cache_from_memcg_idx ( s , i ) ;
2012-12-19 02:23:01 +04:00
if ( ! c )
continue ;
memset ( & sinfo , 0 , sizeof ( sinfo ) ) ;
get_slabinfo ( c , & sinfo ) ;
info - > active_slabs + = sinfo . active_slabs ;
info - > num_slabs + = sinfo . num_slabs ;
info - > shared_avail + = sinfo . shared_avail ;
info - > active_objs + = sinfo . active_objs ;
info - > num_objs + = sinfo . num_objs ;
}
}
int cache_show ( struct kmem_cache * s , struct seq_file * m )
2012-10-19 18:20:25 +04:00
{
2012-10-19 18:20:27 +04:00
struct slabinfo sinfo ;
memset ( & sinfo , 0 , sizeof ( sinfo ) ) ;
get_slabinfo ( s , & sinfo ) ;
2012-12-19 02:23:01 +04:00
memcg_accumulate_slabinfo ( s , & sinfo ) ;
2012-10-19 18:20:27 +04:00
seq_printf ( m , " %-17s %6lu %6lu %6u %4u %4d " ,
2012-12-19 02:23:01 +04:00
cache_name ( s ) , sinfo . active_objs , sinfo . num_objs , s - > size ,
2012-10-19 18:20:27 +04:00
sinfo . objects_per_slab , ( 1 < < sinfo . cache_order ) ) ;
seq_printf ( m , " : tunables %4u %4u %4u " ,
sinfo . limit , sinfo . batchcount , sinfo . shared ) ;
seq_printf ( m , " : slabdata %6lu %6lu %6lu " ,
sinfo . active_slabs , sinfo . num_slabs , sinfo . shared_avail ) ;
slabinfo_show_stats ( m , s ) ;
seq_putc ( m , ' \n ' ) ;
return 0 ;
2012-10-19 18:20:25 +04:00
}
2012-12-19 02:23:01 +04:00
static int s_show ( struct seq_file * m , void * p )
{
struct kmem_cache * s = list_entry ( p , struct kmem_cache , list ) ;
if ( ! is_root_cache ( s ) )
return 0 ;
return cache_show ( s , m ) ;
}
2012-10-19 18:20:25 +04:00
/*
* slabinfo_op - iterator that generates / proc / slabinfo
*
* Output layout :
* cache - name
* num - active - objs
* total - objs
* object size
* num - active - slabs
* total - slabs
* num - pages - per - slab
* + further values on SMP and with statistics enabled
*/
static const struct seq_operations slabinfo_op = {
. start = s_start ,
2013-07-08 04:08:28 +04:00
. next = slab_next ,
. stop = slab_stop ,
2012-10-19 18:20:25 +04:00
. show = s_show ,
} ;
static int slabinfo_open ( struct inode * inode , struct file * file )
{
return seq_open ( file , & slabinfo_op ) ;
}
static const struct file_operations proc_slabinfo_operations = {
. open = slabinfo_open ,
. read = seq_read ,
. write = slabinfo_write ,
. llseek = seq_lseek ,
. release = seq_release ,
} ;
static int __init slab_proc_init ( void )
{
2013-07-04 04:33:24 +04:00
proc_create ( " slabinfo " , SLABINFO_RIGHTS , NULL ,
& proc_slabinfo_operations ) ;
2012-10-19 18:20:25 +04:00
return 0 ;
}
module_init ( slab_proc_init ) ;
# endif /* CONFIG_SLABINFO */
2014-08-07 03:04:44 +04:00
static __always_inline void * __do_krealloc ( const void * p , size_t new_size ,
gfp_t flags )
{
void * ret ;
size_t ks = 0 ;
if ( p )
ks = ksize ( p ) ;
if ( ks > = new_size )
return ( void * ) p ;
ret = kmalloc_track_caller ( new_size , flags ) ;
if ( ret & & p )
memcpy ( ret , p , ks ) ;
return ret ;
}
/**
* __krealloc - like krealloc ( ) but don ' t free @ p .
* @ p : object to reallocate memory for .
* @ new_size : how many bytes of memory are required .
* @ flags : the type of memory to allocate .
*
* This function is like krealloc ( ) except it never frees the originally
* allocated buffer . Use this if you don ' t want to free the buffer immediately
* like , for example , with RCU .
*/
void * __krealloc ( const void * p , size_t new_size , gfp_t flags )
{
if ( unlikely ( ! new_size ) )
return ZERO_SIZE_PTR ;
return __do_krealloc ( p , new_size , flags ) ;
}
EXPORT_SYMBOL ( __krealloc ) ;
/**
* krealloc - reallocate memory . The contents will remain unchanged .
* @ p : object to reallocate memory for .
* @ new_size : how many bytes of memory are required .
* @ flags : the type of memory to allocate .
*
* The contents of the object pointed to are preserved up to the
* lesser of the new and old sizes . If @ p is % NULL , krealloc ( )
* behaves exactly like kmalloc ( ) . If @ new_size is 0 and @ p is not a
* % NULL pointer , the object pointed to is freed .
*/
void * krealloc ( const void * p , size_t new_size , gfp_t flags )
{
void * ret ;
if ( unlikely ( ! new_size ) ) {
kfree ( p ) ;
return ZERO_SIZE_PTR ;
}
ret = __do_krealloc ( p , new_size , flags ) ;
if ( ret & & p ! = ret )
kfree ( p ) ;
return ret ;
}
EXPORT_SYMBOL ( krealloc ) ;
/**
* kzfree - like kfree but zero memory
* @ p : object to free memory of
*
* The memory of the object @ p points to is zeroed before freed .
* If @ p is % NULL , kzfree ( ) does nothing .
*
* Note : this function zeroes the whole allocated buffer which can be a good
* deal bigger than the requested buffer size passed to kmalloc ( ) . So be
* careful when using this function in performance sensitive code .
*/
void kzfree ( const void * p )
{
size_t ks ;
void * mem = ( void * ) p ;
if ( unlikely ( ZERO_OR_NULL_PTR ( mem ) ) )
return ;
ks = ksize ( mem ) ;
memset ( mem , 0 , ks ) ;
kfree ( mem ) ;
}
EXPORT_SYMBOL ( kzfree ) ;
/* Tracepoints definitions. */
EXPORT_TRACEPOINT_SYMBOL ( kmalloc ) ;
EXPORT_TRACEPOINT_SYMBOL ( kmem_cache_alloc ) ;
EXPORT_TRACEPOINT_SYMBOL ( kmalloc_node ) ;
EXPORT_TRACEPOINT_SYMBOL ( kmem_cache_alloc_node ) ;
EXPORT_TRACEPOINT_SYMBOL ( kfree ) ;
EXPORT_TRACEPOINT_SYMBOL ( kmem_cache_free ) ;