2008-02-07 09:13:50 +01:00
|
|
|
/* memcontrol.c - Memory Controller
|
|
|
|
*
|
|
|
|
* Copyright IBM Corporation, 2007
|
|
|
|
* Author Balbir Singh <balbir@linux.vnet.ibm.com>
|
|
|
|
*
|
2008-02-07 09:13:51 +01:00
|
|
|
* Copyright 2007 OpenVZ SWsoft Inc
|
|
|
|
* Author: Pavel Emelianov <xemul@openvz.org>
|
|
|
|
*
|
2008-02-07 09:13:50 +01:00
|
|
|
* This program is free software; you can redistribute it and/or modify
|
|
|
|
* it under the terms of the GNU General Public License as published by
|
|
|
|
* the Free Software Foundation; either version 2 of the License, or
|
|
|
|
* (at your option) any later version.
|
|
|
|
*
|
|
|
|
* This program is distributed in the hope that it will be useful,
|
|
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
|
|
* GNU General Public License for more details.
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include <linux/res_counter.h>
|
|
|
|
#include <linux/memcontrol.h>
|
|
|
|
#include <linux/cgroup.h>
|
2008-02-07 09:13:51 +01:00
|
|
|
#include <linux/mm.h>
|
2009-01-08 03:07:56 +01:00
|
|
|
#include <linux/pagemap.h>
|
2008-02-07 09:14:24 +01:00
|
|
|
#include <linux/smp.h>
|
2008-02-07 09:13:53 +01:00
|
|
|
#include <linux/page-flags.h>
|
2008-02-07 09:13:56 +01:00
|
|
|
#include <linux/backing-dev.h>
|
2008-02-07 09:13:53 +01:00
|
|
|
#include <linux/bit_spinlock.h>
|
|
|
|
#include <linux/rcupdate.h>
|
2009-01-08 03:08:00 +01:00
|
|
|
#include <linux/mutex.h>
|
2008-04-29 10:00:19 +02:00
|
|
|
#include <linux/slab.h>
|
2008-02-07 09:13:56 +01:00
|
|
|
#include <linux/swap.h>
|
|
|
|
#include <linux/spinlock.h>
|
|
|
|
#include <linux/fs.h>
|
2008-02-07 09:14:25 +01:00
|
|
|
#include <linux/seq_file.h>
|
2008-04-29 10:00:24 +02:00
|
|
|
#include <linux/vmalloc.h>
|
2008-10-19 05:26:14 +02:00
|
|
|
#include <linux/mm_inline.h>
|
2008-10-19 05:28:16 +02:00
|
|
|
#include <linux/page_cgroup.h>
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
#include "internal.h"
|
2008-02-07 09:13:50 +01:00
|
|
|
|
2008-02-07 09:13:59 +01:00
|
|
|
#include <asm/uaccess.h>
|
|
|
|
|
2008-07-25 10:47:08 +02:00
|
|
|
struct cgroup_subsys mem_cgroup_subsys __read_mostly;
|
|
|
|
#define MEM_CGROUP_RECLAIM_RETRIES 5
|
2008-02-07 09:13:50 +01:00
|
|
|
|
2009-01-08 03:07:57 +01:00
|
|
|
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
|
|
|
|
/* Turned on only when memory cgroup is enabled && really_do_swap_account = 0 */
|
|
|
|
int do_swap_account __read_mostly;
|
|
|
|
static int really_do_swap_account __initdata = 1; /* for remember boot option*/
|
|
|
|
#else
|
|
|
|
#define do_swap_account (0)
|
|
|
|
#endif
|
|
|
|
|
|
|
|
|
2008-02-07 09:14:24 +01:00
|
|
|
/*
|
|
|
|
* Statistics for memory cgroup.
|
|
|
|
*/
|
|
|
|
enum mem_cgroup_stat_index {
|
|
|
|
/*
|
|
|
|
* For MEM_CONTAINER_TYPE_ALL, usage = pagecache + rss.
|
|
|
|
*/
|
|
|
|
MEM_CGROUP_STAT_CACHE, /* # of pages charged as cache */
|
|
|
|
MEM_CGROUP_STAT_RSS, /* # of pages charged as rss */
|
2008-05-01 13:35:12 +02:00
|
|
|
MEM_CGROUP_STAT_PGPGIN_COUNT, /* # of pages paged in */
|
|
|
|
MEM_CGROUP_STAT_PGPGOUT_COUNT, /* # of pages paged out */
|
2008-02-07 09:14:24 +01:00
|
|
|
|
|
|
|
MEM_CGROUP_STAT_NSTATS,
|
|
|
|
};
|
|
|
|
|
|
|
|
struct mem_cgroup_stat_cpu {
|
|
|
|
s64 count[MEM_CGROUP_STAT_NSTATS];
|
|
|
|
} ____cacheline_aligned_in_smp;
|
|
|
|
|
|
|
|
struct mem_cgroup_stat {
|
2009-01-08 03:07:53 +01:00
|
|
|
struct mem_cgroup_stat_cpu cpustat[0];
|
2008-02-07 09:14:24 +01:00
|
|
|
};
|
|
|
|
|
|
|
|
/*
|
|
|
|
* For accounting under irq disable, no need for increment preempt count.
|
|
|
|
*/
|
2008-10-19 05:28:10 +02:00
|
|
|
static inline void __mem_cgroup_stat_add_safe(struct mem_cgroup_stat_cpu *stat,
|
2008-02-07 09:14:24 +01:00
|
|
|
enum mem_cgroup_stat_index idx, int val)
|
|
|
|
{
|
2008-10-19 05:28:10 +02:00
|
|
|
stat->count[idx] += val;
|
2008-02-07 09:14:24 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
static s64 mem_cgroup_read_stat(struct mem_cgroup_stat *stat,
|
|
|
|
enum mem_cgroup_stat_index idx)
|
|
|
|
{
|
|
|
|
int cpu;
|
|
|
|
s64 ret = 0;
|
|
|
|
for_each_possible_cpu(cpu)
|
|
|
|
ret += stat->cpustat[cpu].count[idx];
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-02-07 09:14:31 +01:00
|
|
|
/*
|
|
|
|
* per-zone information in memory controller.
|
|
|
|
*/
|
|
|
|
struct mem_cgroup_per_zone {
|
2008-02-07 09:14:39 +01:00
|
|
|
/*
|
|
|
|
* spin_lock to protect the per cgroup LRU
|
|
|
|
*/
|
2008-10-19 05:26:14 +02:00
|
|
|
struct list_head lists[NR_LRU_LISTS];
|
|
|
|
unsigned long count[NR_LRU_LISTS];
|
2008-02-07 09:14:31 +01:00
|
|
|
};
|
|
|
|
/* Macro for accessing counter */
|
|
|
|
#define MEM_CGROUP_ZSTAT(mz, idx) ((mz)->count[(idx)])
|
|
|
|
|
|
|
|
struct mem_cgroup_per_node {
|
|
|
|
struct mem_cgroup_per_zone zoneinfo[MAX_NR_ZONES];
|
|
|
|
};
|
|
|
|
|
|
|
|
struct mem_cgroup_lru_info {
|
|
|
|
struct mem_cgroup_per_node *nodeinfo[MAX_NUMNODES];
|
|
|
|
};
|
|
|
|
|
2008-02-07 09:13:50 +01:00
|
|
|
/*
|
|
|
|
* The memory controller data structure. The memory controller controls both
|
|
|
|
* page cache and RSS per cgroup. We would eventually like to provide
|
|
|
|
* statistics based on the statistics developed by Rik Van Riel for clock-pro,
|
|
|
|
* to help the administrator determine what knobs to tune.
|
|
|
|
*
|
|
|
|
* TODO: Add a water mark for the memory controller. Reclaim will begin when
|
2008-02-07 09:13:53 +01:00
|
|
|
* we hit the water mark. May be even add a low water mark, such that
|
|
|
|
* no reclaim occurs from a cgroup at it's low water mark, this is
|
|
|
|
* a feature that will be implemented much later in the future.
|
2008-02-07 09:13:50 +01:00
|
|
|
*/
|
|
|
|
struct mem_cgroup {
|
|
|
|
struct cgroup_subsys_state css;
|
|
|
|
/*
|
|
|
|
* the counter to account for memory usage
|
|
|
|
*/
|
|
|
|
struct res_counter res;
|
2009-01-08 03:08:00 +01:00
|
|
|
/*
|
|
|
|
* the counter to account for mem+swap usage.
|
|
|
|
*/
|
|
|
|
struct res_counter memsw;
|
2008-02-07 09:13:51 +01:00
|
|
|
/*
|
|
|
|
* Per cgroup active and inactive list, similar to the
|
|
|
|
* per zone LRU lists.
|
|
|
|
*/
|
2008-02-07 09:14:31 +01:00
|
|
|
struct mem_cgroup_lru_info info;
|
2008-02-07 09:14:39 +01:00
|
|
|
|
2008-02-07 09:14:34 +01:00
|
|
|
int prev_priority; /* for recording reclaim priority */
|
2009-01-08 03:08:06 +01:00
|
|
|
|
|
|
|
/*
|
|
|
|
* While reclaiming in a hiearchy, we cache the last child we
|
|
|
|
* reclaimed from. Protected by cgroup_lock()
|
|
|
|
*/
|
|
|
|
struct mem_cgroup *last_scanned_child;
|
2009-01-08 03:08:07 +01:00
|
|
|
/*
|
|
|
|
* Should the accounting and control be hierarchical, per subtree?
|
|
|
|
*/
|
|
|
|
bool use_hierarchy;
|
2009-01-08 03:08:08 +01:00
|
|
|
unsigned long last_oom_jiffies;
|
2009-01-08 03:08:00 +01:00
|
|
|
int obsolete;
|
|
|
|
atomic_t refcnt;
|
2008-02-07 09:14:24 +01:00
|
|
|
/*
|
2009-01-08 03:07:53 +01:00
|
|
|
* statistics. This must be placed at the end of memcg.
|
2008-02-07 09:14:24 +01:00
|
|
|
*/
|
|
|
|
struct mem_cgroup_stat stat;
|
2008-02-07 09:13:50 +01:00
|
|
|
};
|
|
|
|
|
2008-02-07 09:14:17 +01:00
|
|
|
enum charge_type {
|
|
|
|
MEM_CGROUP_CHARGE_TYPE_CACHE = 0,
|
|
|
|
MEM_CGROUP_CHARGE_TYPE_MAPPED,
|
2008-10-19 05:26:32 +02:00
|
|
|
MEM_CGROUP_CHARGE_TYPE_SHMEM, /* used by page migration of shmem */
|
2008-10-19 05:28:11 +02:00
|
|
|
MEM_CGROUP_CHARGE_TYPE_FORCE, /* used by force_empty */
|
2009-01-08 03:07:56 +01:00
|
|
|
MEM_CGROUP_CHARGE_TYPE_SWAPOUT, /* for accounting swapcache */
|
2008-10-19 05:28:11 +02:00
|
|
|
NR_CHARGE_TYPE,
|
|
|
|
};
|
|
|
|
|
2008-10-19 05:28:16 +02:00
|
|
|
/* only for here (for easy reading.) */
|
|
|
|
#define PCGF_CACHE (1UL << PCG_CACHE)
|
|
|
|
#define PCGF_USED (1UL << PCG_USED)
|
|
|
|
#define PCGF_LOCK (1UL << PCG_LOCK)
|
2008-10-19 05:28:11 +02:00
|
|
|
static const unsigned long
|
|
|
|
pcg_default_flags[NR_CHARGE_TYPE] = {
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
PCGF_CACHE | PCGF_USED | PCGF_LOCK, /* File Cache */
|
|
|
|
PCGF_USED | PCGF_LOCK, /* Anon */
|
|
|
|
PCGF_CACHE | PCGF_USED | PCGF_LOCK, /* Shmem */
|
2008-10-19 05:28:16 +02:00
|
|
|
0, /* FORCE */
|
2008-02-07 09:14:17 +01:00
|
|
|
};
|
|
|
|
|
2009-01-08 03:08:00 +01:00
|
|
|
|
|
|
|
/* for encoding cft->private value on file */
|
|
|
|
#define _MEM (0)
|
|
|
|
#define _MEMSWAP (1)
|
|
|
|
#define MEMFILE_PRIVATE(x, val) (((x) << 16) | (val))
|
|
|
|
#define MEMFILE_TYPE(val) (((val) >> 16) & 0xffff)
|
|
|
|
#define MEMFILE_ATTR(val) ((val) & 0xffff)
|
|
|
|
|
|
|
|
static void mem_cgroup_get(struct mem_cgroup *mem);
|
|
|
|
static void mem_cgroup_put(struct mem_cgroup *mem);
|
|
|
|
|
2008-10-19 05:28:11 +02:00
|
|
|
static void mem_cgroup_charge_statistics(struct mem_cgroup *mem,
|
|
|
|
struct page_cgroup *pc,
|
|
|
|
bool charge)
|
2008-02-07 09:14:24 +01:00
|
|
|
{
|
|
|
|
int val = (charge)? 1 : -1;
|
|
|
|
struct mem_cgroup_stat *stat = &mem->stat;
|
2008-10-19 05:28:10 +02:00
|
|
|
struct mem_cgroup_stat_cpu *cpustat;
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
int cpu = get_cpu();
|
2008-02-07 09:14:24 +01:00
|
|
|
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
cpustat = &stat->cpustat[cpu];
|
2008-10-19 05:28:11 +02:00
|
|
|
if (PageCgroupCache(pc))
|
2008-10-19 05:28:10 +02:00
|
|
|
__mem_cgroup_stat_add_safe(cpustat, MEM_CGROUP_STAT_CACHE, val);
|
2008-02-07 09:14:24 +01:00
|
|
|
else
|
2008-10-19 05:28:10 +02:00
|
|
|
__mem_cgroup_stat_add_safe(cpustat, MEM_CGROUP_STAT_RSS, val);
|
2008-05-01 13:35:12 +02:00
|
|
|
|
|
|
|
if (charge)
|
2008-10-19 05:28:10 +02:00
|
|
|
__mem_cgroup_stat_add_safe(cpustat,
|
2008-05-01 13:35:12 +02:00
|
|
|
MEM_CGROUP_STAT_PGPGIN_COUNT, 1);
|
|
|
|
else
|
2008-10-19 05:28:10 +02:00
|
|
|
__mem_cgroup_stat_add_safe(cpustat,
|
2008-05-01 13:35:12 +02:00
|
|
|
MEM_CGROUP_STAT_PGPGOUT_COUNT, 1);
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
put_cpu();
|
2008-02-07 09:14:31 +01:00
|
|
|
}
|
|
|
|
|
2008-03-04 23:29:10 +01:00
|
|
|
static struct mem_cgroup_per_zone *
|
2008-02-07 09:14:31 +01:00
|
|
|
mem_cgroup_zoneinfo(struct mem_cgroup *mem, int nid, int zid)
|
|
|
|
{
|
|
|
|
return &mem->info.nodeinfo[nid]->zoneinfo[zid];
|
|
|
|
}
|
|
|
|
|
2008-03-04 23:29:10 +01:00
|
|
|
static struct mem_cgroup_per_zone *
|
2008-02-07 09:14:31 +01:00
|
|
|
page_cgroup_zoneinfo(struct page_cgroup *pc)
|
|
|
|
{
|
|
|
|
struct mem_cgroup *mem = pc->mem_cgroup;
|
|
|
|
int nid = page_cgroup_nid(pc);
|
|
|
|
int zid = page_cgroup_zid(pc);
|
2008-02-07 09:14:24 +01:00
|
|
|
|
2008-02-07 09:14:31 +01:00
|
|
|
return mem_cgroup_zoneinfo(mem, nid, zid);
|
|
|
|
}
|
|
|
|
|
|
|
|
static unsigned long mem_cgroup_get_all_zonestat(struct mem_cgroup *mem,
|
2008-10-19 05:26:14 +02:00
|
|
|
enum lru_list idx)
|
2008-02-07 09:14:31 +01:00
|
|
|
{
|
|
|
|
int nid, zid;
|
|
|
|
struct mem_cgroup_per_zone *mz;
|
|
|
|
u64 total = 0;
|
|
|
|
|
|
|
|
for_each_online_node(nid)
|
|
|
|
for (zid = 0; zid < MAX_NR_ZONES; zid++) {
|
|
|
|
mz = mem_cgroup_zoneinfo(mem, nid, zid);
|
|
|
|
total += MEM_CGROUP_ZSTAT(mz, idx);
|
|
|
|
}
|
|
|
|
return total;
|
2008-02-07 09:14:24 +01:00
|
|
|
}
|
|
|
|
|
2008-03-04 23:29:10 +01:00
|
|
|
static struct mem_cgroup *mem_cgroup_from_cont(struct cgroup *cont)
|
2008-02-07 09:13:50 +01:00
|
|
|
{
|
|
|
|
return container_of(cgroup_subsys_state(cont,
|
|
|
|
mem_cgroup_subsys_id), struct mem_cgroup,
|
|
|
|
css);
|
|
|
|
}
|
|
|
|
|
cgroups: add an owner to the mm_struct
Remove the mem_cgroup member from mm_struct and instead adds an owner.
This approach was suggested by Paul Menage. The advantage of this approach
is that, once the mm->owner is known, using the subsystem id, the cgroup
can be determined. It also allows several control groups that are
virtually grouped by mm_struct, to exist independent of the memory
controller i.e., without adding mem_cgroup's for each controller, to
mm_struct.
A new config option CONFIG_MM_OWNER is added and the memory resource
controller selects this config option.
This patch also adds cgroup callbacks to notify subsystems when mm->owner
changes. The mm_cgroup_changed callback is called with the task_lock() of
the new task held and is called just prior to changing the mm->owner.
I am indebted to Paul Menage for the several reviews of this patchset and
helping me make it lighter and simpler.
This patch was tested on a powerpc box, it was compiled with both the
MM_OWNER config turned on and off.
After the thread group leader exits, it's moved to init_css_state by
cgroup_exit(), thus all future charges from runnings threads would be
redirected to the init_css_set's subsystem.
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Pavel Emelianov <xemul@openvz.org>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Sudhir Kumar <skumar@linux.vnet.ibm.com>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Hirokazu Takahashi <taka@valinux.co.jp>
Cc: David Rientjes <rientjes@google.com>,
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Pekka Enberg <penberg@cs.helsinki.fi>
Reviewed-by: Paul Menage <menage@google.com>
Cc: Oleg Nesterov <oleg@tv-sign.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-29 10:00:16 +02:00
|
|
|
struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p)
|
2008-02-07 09:13:51 +01:00
|
|
|
{
|
mm owner: fix race between swapoff and exit
There's a race between mm->owner assignment and swapoff, more easily
seen when task slab poisoning is turned on. The condition occurs when
try_to_unuse() runs in parallel with an exiting task. A similar race
can occur with callers of get_task_mm(), such as /proc/<pid>/<mmstats>
or ptrace or page migration.
CPU0 CPU1
try_to_unuse
looks at mm = task0->mm
increments mm->mm_users
task 0 exits
mm->owner needs to be updated, but no
new owner is found (mm_users > 1, but
no other task has task->mm = task0->mm)
mm_update_next_owner() leaves
mmput(mm) decrements mm->mm_users
task0 freed
dereferencing mm->owner fails
The fix is to notify the subsystem via mm_owner_changed callback(),
if no new owner is found, by specifying the new task as NULL.
Jiri Slaby:
mm->owner was set to NULL prior to calling cgroup_mm_owner_callbacks(), but
must be set after that, so as not to pass NULL as old owner causing oops.
Daisuke Nishimura:
mm_update_next_owner() may set mm->owner to NULL, but mem_cgroup_from_task()
and its callers need to take account of this situation to avoid oops.
Hugh Dickins:
Lockdep warning and hang below exec_mmap() when testing these patches.
exit_mm() up_reads mmap_sem before calling mm_update_next_owner(),
so exec_mmap() now needs to do the same. And with that repositioning,
there's now no point in mm_need_new_owner() allowing for NULL mm.
Reported-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-09-29 00:09:31 +02:00
|
|
|
/*
|
|
|
|
* mm_update_next_owner() may clear mm->owner to NULL
|
|
|
|
* if it races with swapoff, page migration, etc.
|
|
|
|
* So this can be called with p == NULL.
|
|
|
|
*/
|
|
|
|
if (unlikely(!p))
|
|
|
|
return NULL;
|
|
|
|
|
2008-02-07 09:13:51 +01:00
|
|
|
return container_of(task_subsys_state(p, mem_cgroup_subsys_id),
|
|
|
|
struct mem_cgroup, css);
|
|
|
|
}
|
|
|
|
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
/*
|
|
|
|
* Following LRU functions are allowed to be used without PCG_LOCK.
|
|
|
|
* Operations are called by routine of global LRU independently from memcg.
|
|
|
|
* What we have to take care of here is validness of pc->mem_cgroup.
|
|
|
|
*
|
|
|
|
* Changes to pc->mem_cgroup happens when
|
|
|
|
* 1. charge
|
|
|
|
* 2. moving account
|
|
|
|
* In typical case, "charge" is done before add-to-lru. Exception is SwapCache.
|
|
|
|
* It is added to LRU before charge.
|
|
|
|
* If PCG_USED bit is not set, page_cgroup is not added to this private LRU.
|
|
|
|
* When moving account, the page is not on LRU. It's isolated.
|
|
|
|
*/
|
2008-10-19 05:26:32 +02:00
|
|
|
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
void mem_cgroup_del_lru_list(struct page *page, enum lru_list lru)
|
|
|
|
{
|
|
|
|
struct page_cgroup *pc;
|
|
|
|
struct mem_cgroup *mem;
|
|
|
|
struct mem_cgroup_per_zone *mz;
|
2008-02-07 09:14:31 +01:00
|
|
|
|
2009-01-08 03:08:02 +01:00
|
|
|
if (mem_cgroup_disabled())
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
return;
|
|
|
|
pc = lookup_page_cgroup(page);
|
|
|
|
/* can happen while we handle swapcache. */
|
|
|
|
if (list_empty(&pc->lru))
|
|
|
|
return;
|
|
|
|
mz = page_cgroup_zoneinfo(pc);
|
|
|
|
mem = pc->mem_cgroup;
|
2008-10-19 05:26:14 +02:00
|
|
|
MEM_CGROUP_ZSTAT(mz, lru) -= 1;
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
list_del_init(&pc->lru);
|
|
|
|
return;
|
2008-02-07 09:14:31 +01:00
|
|
|
}
|
|
|
|
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
void mem_cgroup_del_lru(struct page *page)
|
2008-02-07 09:14:31 +01:00
|
|
|
{
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
mem_cgroup_del_lru_list(page, page_lru(page));
|
|
|
|
}
|
2008-10-19 05:26:14 +02:00
|
|
|
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
void mem_cgroup_rotate_lru_list(struct page *page, enum lru_list lru)
|
|
|
|
{
|
|
|
|
struct mem_cgroup_per_zone *mz;
|
|
|
|
struct page_cgroup *pc;
|
2008-10-19 05:26:14 +02:00
|
|
|
|
2009-01-08 03:08:02 +01:00
|
|
|
if (mem_cgroup_disabled())
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
return;
|
2008-02-07 09:14:31 +01:00
|
|
|
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
pc = lookup_page_cgroup(page);
|
|
|
|
smp_rmb();
|
|
|
|
/* unused page is not rotated. */
|
|
|
|
if (!PageCgroupUsed(pc))
|
|
|
|
return;
|
|
|
|
mz = page_cgroup_zoneinfo(pc);
|
|
|
|
list_move(&pc->lru, &mz->lists[lru]);
|
2008-02-07 09:14:31 +01:00
|
|
|
}
|
|
|
|
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
void mem_cgroup_add_lru_list(struct page *page, enum lru_list lru)
|
2008-02-07 09:13:56 +01:00
|
|
|
{
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
struct page_cgroup *pc;
|
|
|
|
struct mem_cgroup_per_zone *mz;
|
2008-02-07 09:14:31 +01:00
|
|
|
|
2009-01-08 03:08:02 +01:00
|
|
|
if (mem_cgroup_disabled())
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
return;
|
|
|
|
pc = lookup_page_cgroup(page);
|
|
|
|
/* barrier to sync with "charge" */
|
|
|
|
smp_rmb();
|
|
|
|
if (!PageCgroupUsed(pc))
|
Unevictable LRU Infrastructure
When the system contains lots of mlocked or otherwise unevictable pages,
the pageout code (kswapd) can spend lots of time scanning over these
pages. Worse still, the presence of lots of unevictable pages can confuse
kswapd into thinking that more aggressive pageout modes are required,
resulting in all kinds of bad behaviour.
Infrastructure to manage pages excluded from reclaim--i.e., hidden from
vmscan. Based on a patch by Larry Woodman of Red Hat. Reworked to
maintain "unevictable" pages on a separate per-zone LRU list, to "hide"
them from vmscan.
Kosaki Motohiro added the support for the memory controller unevictable
lru list.
Pages on the unevictable list have both PG_unevictable and PG_lru set.
Thus, PG_unevictable is analogous to and mutually exclusive with
PG_active--it specifies which LRU list the page is on.
The unevictable infrastructure is enabled by a new mm Kconfig option
[CONFIG_]UNEVICTABLE_LRU.
A new function 'page_evictable(page, vma)' in vmscan.c tests whether or
not a page may be evictable. Subsequent patches will add the various
!evictable tests. We'll want to keep these tests light-weight for use in
shrink_active_list() and, possibly, the fault path.
To avoid races between tasks putting pages [back] onto an LRU list and
tasks that might be moving the page from non-evictable to evictable state,
the new function 'putback_lru_page()' -- inverse to 'isolate_lru_page()'
-- tests the "evictability" of a page after placing it on the LRU, before
dropping the reference. If the page has become unevictable,
putback_lru_page() will redo the 'putback', thus moving the page to the
unevictable list. This way, we avoid "stranding" evictable pages on the
unevictable list.
[akpm@linux-foundation.org: fix fallout from out-of-order merge]
[riel@redhat.com: fix UNEVICTABLE_LRU and !PROC_PAGE_MONITOR build]
[nishimura@mxp.nes.nec.co.jp: remove redundant mapping check]
[kosaki.motohiro@jp.fujitsu.com: unevictable-lru-infrastructure: putback_lru_page()/unevictable page handling rework]
[kosaki.motohiro@jp.fujitsu.com: kill unnecessary lock_page() in vmscan.c]
[kosaki.motohiro@jp.fujitsu.com: revert migration change of unevictable lru infrastructure]
[kosaki.motohiro@jp.fujitsu.com: revert to unevictable-lru-infrastructure-kconfig-fix.patch]
[kosaki.motohiro@jp.fujitsu.com: restore patch failure of vmstat-unevictable-and-mlocked-pages-vm-events.patch]
Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Debugged-by: Benjamin Kidwell <benjkidwell@yahoo.com>
Signed-off-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-10-19 05:26:39 +02:00
|
|
|
return;
|
2008-10-19 05:26:14 +02:00
|
|
|
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
mz = page_cgroup_zoneinfo(pc);
|
2008-10-19 05:26:14 +02:00
|
|
|
MEM_CGROUP_ZSTAT(mz, lru) += 1;
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
list_add(&pc->lru, &mz->lists[lru]);
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* To add swapcache into LRU. Be careful to all this function.
|
|
|
|
* zone->lru_lock shouldn't be held and irq must not be disabled.
|
|
|
|
*/
|
|
|
|
static void mem_cgroup_lru_fixup(struct page *page)
|
|
|
|
{
|
|
|
|
if (!isolate_lru_page(page))
|
|
|
|
putback_lru_page(page);
|
|
|
|
}
|
|
|
|
|
|
|
|
void mem_cgroup_move_lists(struct page *page,
|
|
|
|
enum lru_list from, enum lru_list to)
|
|
|
|
{
|
2009-01-08 03:08:02 +01:00
|
|
|
if (mem_cgroup_disabled())
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
return;
|
|
|
|
mem_cgroup_del_lru_list(page, from);
|
|
|
|
mem_cgroup_add_lru_list(page, to);
|
2008-02-07 09:13:56 +01:00
|
|
|
}
|
|
|
|
|
2008-02-07 09:14:06 +01:00
|
|
|
int task_in_mem_cgroup(struct task_struct *task, const struct mem_cgroup *mem)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
task_lock(task);
|
2008-03-04 23:29:01 +01:00
|
|
|
ret = task->mm && mm_match_cgroup(task->mm, mem);
|
2008-02-07 09:14:06 +01:00
|
|
|
task_unlock(task);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-02-07 09:14:32 +01:00
|
|
|
/*
|
|
|
|
* Calculate mapped_ratio under memory controller. This will be used in
|
|
|
|
* vmscan.c for deteremining we have to reclaim mapped pages.
|
|
|
|
*/
|
|
|
|
int mem_cgroup_calc_mapped_ratio(struct mem_cgroup *mem)
|
|
|
|
{
|
|
|
|
long total, rss;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* usage is recorded in bytes. But, here, we assume the number of
|
|
|
|
* physical pages can be represented by "long" on any arch.
|
|
|
|
*/
|
|
|
|
total = (long) (mem->res.usage >> PAGE_SHIFT) + 1L;
|
|
|
|
rss = (long)mem_cgroup_read_stat(&mem->stat, MEM_CGROUP_STAT_RSS);
|
|
|
|
return (int)((rss * 100L) / total);
|
|
|
|
}
|
2008-03-04 23:29:09 +01:00
|
|
|
|
2008-02-07 09:14:34 +01:00
|
|
|
/*
|
|
|
|
* prev_priority control...this will be used in memory reclaim path.
|
|
|
|
*/
|
|
|
|
int mem_cgroup_get_reclaim_priority(struct mem_cgroup *mem)
|
|
|
|
{
|
|
|
|
return mem->prev_priority;
|
|
|
|
}
|
|
|
|
|
|
|
|
void mem_cgroup_note_reclaim_priority(struct mem_cgroup *mem, int priority)
|
|
|
|
{
|
|
|
|
if (priority < mem->prev_priority)
|
|
|
|
mem->prev_priority = priority;
|
|
|
|
}
|
|
|
|
|
|
|
|
void mem_cgroup_record_reclaim_priority(struct mem_cgroup *mem, int priority)
|
|
|
|
{
|
|
|
|
mem->prev_priority = priority;
|
|
|
|
}
|
|
|
|
|
2008-02-07 09:14:35 +01:00
|
|
|
/*
|
|
|
|
* Calculate # of pages to be scanned in this priority/zone.
|
|
|
|
* See also vmscan.c
|
|
|
|
*
|
|
|
|
* priority starts from "DEF_PRIORITY" and decremented in each loop.
|
|
|
|
* (see include/linux/mmzone.h)
|
|
|
|
*/
|
|
|
|
|
2008-10-19 05:26:14 +02:00
|
|
|
long mem_cgroup_calc_reclaim(struct mem_cgroup *mem, struct zone *zone,
|
|
|
|
int priority, enum lru_list lru)
|
2008-02-07 09:14:35 +01:00
|
|
|
{
|
2008-10-19 05:26:14 +02:00
|
|
|
long nr_pages;
|
2008-02-07 09:14:35 +01:00
|
|
|
int nid = zone->zone_pgdat->node_id;
|
|
|
|
int zid = zone_idx(zone);
|
|
|
|
struct mem_cgroup_per_zone *mz = mem_cgroup_zoneinfo(mem, nid, zid);
|
|
|
|
|
2008-10-19 05:26:14 +02:00
|
|
|
nr_pages = MEM_CGROUP_ZSTAT(mz, lru);
|
2008-02-07 09:14:35 +01:00
|
|
|
|
2008-10-19 05:26:14 +02:00
|
|
|
return (nr_pages >> priority);
|
2008-02-07 09:14:35 +01:00
|
|
|
}
|
|
|
|
|
2008-02-07 09:13:56 +01:00
|
|
|
unsigned long mem_cgroup_isolate_pages(unsigned long nr_to_scan,
|
|
|
|
struct list_head *dst,
|
|
|
|
unsigned long *scanned, int order,
|
|
|
|
int mode, struct zone *z,
|
|
|
|
struct mem_cgroup *mem_cont,
|
2008-10-19 05:26:32 +02:00
|
|
|
int active, int file)
|
2008-02-07 09:13:56 +01:00
|
|
|
{
|
|
|
|
unsigned long nr_taken = 0;
|
|
|
|
struct page *page;
|
|
|
|
unsigned long scan;
|
|
|
|
LIST_HEAD(pc_list);
|
|
|
|
struct list_head *src;
|
2008-02-07 09:14:11 +01:00
|
|
|
struct page_cgroup *pc, *tmp;
|
2008-02-07 09:14:38 +01:00
|
|
|
int nid = z->zone_pgdat->node_id;
|
|
|
|
int zid = zone_idx(z);
|
|
|
|
struct mem_cgroup_per_zone *mz;
|
2008-10-19 05:26:32 +02:00
|
|
|
int lru = LRU_FILE * !!file + !!active;
|
2008-02-07 09:13:56 +01:00
|
|
|
|
cgroups: add an owner to the mm_struct
Remove the mem_cgroup member from mm_struct and instead adds an owner.
This approach was suggested by Paul Menage. The advantage of this approach
is that, once the mm->owner is known, using the subsystem id, the cgroup
can be determined. It also allows several control groups that are
virtually grouped by mm_struct, to exist independent of the memory
controller i.e., without adding mem_cgroup's for each controller, to
mm_struct.
A new config option CONFIG_MM_OWNER is added and the memory resource
controller selects this config option.
This patch also adds cgroup callbacks to notify subsystems when mm->owner
changes. The mm_cgroup_changed callback is called with the task_lock() of
the new task held and is called just prior to changing the mm->owner.
I am indebted to Paul Menage for the several reviews of this patchset and
helping me make it lighter and simpler.
This patch was tested on a powerpc box, it was compiled with both the
MM_OWNER config turned on and off.
After the thread group leader exits, it's moved to init_css_state by
cgroup_exit(), thus all future charges from runnings threads would be
redirected to the init_css_set's subsystem.
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Pavel Emelianov <xemul@openvz.org>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Sudhir Kumar <skumar@linux.vnet.ibm.com>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Hirokazu Takahashi <taka@valinux.co.jp>
Cc: David Rientjes <rientjes@google.com>,
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Pekka Enberg <penberg@cs.helsinki.fi>
Reviewed-by: Paul Menage <menage@google.com>
Cc: Oleg Nesterov <oleg@tv-sign.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-29 10:00:16 +02:00
|
|
|
BUG_ON(!mem_cont);
|
2008-02-07 09:14:38 +01:00
|
|
|
mz = mem_cgroup_zoneinfo(mem_cont, nid, zid);
|
2008-10-19 05:26:14 +02:00
|
|
|
src = &mz->lists[lru];
|
2008-02-07 09:13:56 +01:00
|
|
|
|
2008-02-07 09:14:11 +01:00
|
|
|
scan = 0;
|
|
|
|
list_for_each_entry_safe_reverse(pc, tmp, src, lru) {
|
2008-02-07 09:14:12 +01:00
|
|
|
if (scan >= nr_to_scan)
|
2008-02-07 09:14:11 +01:00
|
|
|
break;
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
|
|
|
|
page = pc->page;
|
2008-10-19 05:28:16 +02:00
|
|
|
if (unlikely(!PageCgroupUsed(pc)))
|
|
|
|
continue;
|
2008-02-07 09:14:12 +01:00
|
|
|
if (unlikely(!PageLRU(page)))
|
2008-02-07 09:14:11 +01:00
|
|
|
continue;
|
|
|
|
|
2008-02-07 09:14:12 +01:00
|
|
|
scan++;
|
2008-10-19 05:26:32 +02:00
|
|
|
if (__isolate_lru_page(page, mode, file) == 0) {
|
2008-02-07 09:13:56 +01:00
|
|
|
list_move(&page->lru, dst);
|
|
|
|
nr_taken++;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
*scanned = scan;
|
|
|
|
return nr_taken;
|
|
|
|
}
|
|
|
|
|
2009-01-08 03:08:06 +01:00
|
|
|
#define mem_cgroup_from_res_counter(counter, member) \
|
|
|
|
container_of(counter, struct mem_cgroup, member)
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This routine finds the DFS walk successor. This routine should be
|
|
|
|
* called with cgroup_mutex held
|
|
|
|
*/
|
|
|
|
static struct mem_cgroup *
|
|
|
|
mem_cgroup_get_next_node(struct mem_cgroup *curr, struct mem_cgroup *root_mem)
|
|
|
|
{
|
|
|
|
struct cgroup *cgroup, *curr_cgroup, *root_cgroup;
|
|
|
|
|
|
|
|
curr_cgroup = curr->css.cgroup;
|
|
|
|
root_cgroup = root_mem->css.cgroup;
|
|
|
|
|
|
|
|
if (!list_empty(&curr_cgroup->children)) {
|
|
|
|
/*
|
|
|
|
* Walk down to children
|
|
|
|
*/
|
|
|
|
mem_cgroup_put(curr);
|
|
|
|
cgroup = list_entry(curr_cgroup->children.next,
|
|
|
|
struct cgroup, sibling);
|
|
|
|
curr = mem_cgroup_from_cont(cgroup);
|
|
|
|
mem_cgroup_get(curr);
|
|
|
|
goto done;
|
|
|
|
}
|
|
|
|
|
|
|
|
visit_parent:
|
|
|
|
if (curr_cgroup == root_cgroup) {
|
|
|
|
mem_cgroup_put(curr);
|
|
|
|
curr = root_mem;
|
|
|
|
mem_cgroup_get(curr);
|
|
|
|
goto done;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Goto next sibling
|
|
|
|
*/
|
|
|
|
if (curr_cgroup->sibling.next != &curr_cgroup->parent->children) {
|
|
|
|
mem_cgroup_put(curr);
|
|
|
|
cgroup = list_entry(curr_cgroup->sibling.next, struct cgroup,
|
|
|
|
sibling);
|
|
|
|
curr = mem_cgroup_from_cont(cgroup);
|
|
|
|
mem_cgroup_get(curr);
|
|
|
|
goto done;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Go up to next parent and next parent's sibling if need be
|
|
|
|
*/
|
|
|
|
curr_cgroup = curr_cgroup->parent;
|
|
|
|
goto visit_parent;
|
|
|
|
|
|
|
|
done:
|
|
|
|
root_mem->last_scanned_child = curr;
|
|
|
|
return curr;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Visit the first child (need not be the first child as per the ordering
|
|
|
|
* of the cgroup list, since we track last_scanned_child) of @mem and use
|
|
|
|
* that to reclaim free pages from.
|
|
|
|
*/
|
|
|
|
static struct mem_cgroup *
|
|
|
|
mem_cgroup_get_first_node(struct mem_cgroup *root_mem)
|
|
|
|
{
|
|
|
|
struct cgroup *cgroup;
|
|
|
|
struct mem_cgroup *ret;
|
|
|
|
bool obsolete = (root_mem->last_scanned_child &&
|
|
|
|
root_mem->last_scanned_child->obsolete);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Scan all children under the mem_cgroup mem
|
|
|
|
*/
|
|
|
|
cgroup_lock();
|
|
|
|
if (list_empty(&root_mem->css.cgroup->children)) {
|
|
|
|
ret = root_mem;
|
|
|
|
goto done;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!root_mem->last_scanned_child || obsolete) {
|
|
|
|
|
|
|
|
if (obsolete)
|
|
|
|
mem_cgroup_put(root_mem->last_scanned_child);
|
|
|
|
|
|
|
|
cgroup = list_first_entry(&root_mem->css.cgroup->children,
|
|
|
|
struct cgroup, sibling);
|
|
|
|
ret = mem_cgroup_from_cont(cgroup);
|
|
|
|
mem_cgroup_get(ret);
|
|
|
|
} else
|
|
|
|
ret = mem_cgroup_get_next_node(root_mem->last_scanned_child,
|
|
|
|
root_mem);
|
|
|
|
|
|
|
|
done:
|
|
|
|
root_mem->last_scanned_child = ret;
|
|
|
|
cgroup_unlock();
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2009-01-08 03:08:12 +01:00
|
|
|
static bool mem_cgroup_check_under_limit(struct mem_cgroup *mem)
|
|
|
|
{
|
|
|
|
if (do_swap_account) {
|
|
|
|
if (res_counter_check_under_limit(&mem->res) &&
|
|
|
|
res_counter_check_under_limit(&mem->memsw))
|
|
|
|
return true;
|
|
|
|
} else
|
|
|
|
if (res_counter_check_under_limit(&mem->res))
|
|
|
|
return true;
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2009-01-08 03:08:06 +01:00
|
|
|
/*
|
|
|
|
* Dance down the hierarchy if needed to reclaim memory. We remember the
|
|
|
|
* last child we reclaimed from, so that we don't end up penalizing
|
|
|
|
* one child extensively based on its position in the children list.
|
|
|
|
*
|
|
|
|
* root_mem is the original ancestor that we've been reclaim from.
|
|
|
|
*/
|
|
|
|
static int mem_cgroup_hierarchical_reclaim(struct mem_cgroup *root_mem,
|
|
|
|
gfp_t gfp_mask, bool noswap)
|
|
|
|
{
|
|
|
|
struct mem_cgroup *next_mem;
|
|
|
|
int ret = 0;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Reclaim unconditionally and don't check for return value.
|
|
|
|
* We need to reclaim in the current group and down the tree.
|
|
|
|
* One might think about checking for children before reclaiming,
|
|
|
|
* but there might be left over accounting, even after children
|
|
|
|
* have left.
|
|
|
|
*/
|
|
|
|
ret = try_to_free_mem_cgroup_pages(root_mem, gfp_mask, noswap);
|
2009-01-08 03:08:12 +01:00
|
|
|
if (mem_cgroup_check_under_limit(root_mem))
|
2009-01-08 03:08:06 +01:00
|
|
|
return 0;
|
2009-01-08 03:08:13 +01:00
|
|
|
if (!root_mem->use_hierarchy)
|
|
|
|
return ret;
|
2009-01-08 03:08:06 +01:00
|
|
|
|
|
|
|
next_mem = mem_cgroup_get_first_node(root_mem);
|
|
|
|
|
|
|
|
while (next_mem != root_mem) {
|
|
|
|
if (next_mem->obsolete) {
|
|
|
|
mem_cgroup_put(next_mem);
|
|
|
|
cgroup_lock();
|
|
|
|
next_mem = mem_cgroup_get_first_node(root_mem);
|
|
|
|
cgroup_unlock();
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
ret = try_to_free_mem_cgroup_pages(next_mem, gfp_mask, noswap);
|
2009-01-08 03:08:12 +01:00
|
|
|
if (mem_cgroup_check_under_limit(root_mem))
|
2009-01-08 03:08:06 +01:00
|
|
|
return 0;
|
|
|
|
cgroup_lock();
|
|
|
|
next_mem = mem_cgroup_get_next_node(next_mem, root_mem);
|
|
|
|
cgroup_unlock();
|
|
|
|
}
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2009-01-08 03:08:08 +01:00
|
|
|
bool mem_cgroup_oom_called(struct task_struct *task)
|
|
|
|
{
|
|
|
|
bool ret = false;
|
|
|
|
struct mem_cgroup *mem;
|
|
|
|
struct mm_struct *mm;
|
|
|
|
|
|
|
|
rcu_read_lock();
|
|
|
|
mm = task->mm;
|
|
|
|
if (!mm)
|
|
|
|
mm = &init_mm;
|
|
|
|
mem = mem_cgroup_from_task(rcu_dereference(mm->owner));
|
|
|
|
if (mem && time_before(jiffies, mem->last_oom_jiffies + HZ/10))
|
|
|
|
ret = true;
|
|
|
|
rcu_read_unlock();
|
|
|
|
return ret;
|
|
|
|
}
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
/*
|
|
|
|
* Unlike exported interface, "oom" parameter is added. if oom==true,
|
|
|
|
* oom-killer can be invoked.
|
2008-02-07 09:13:53 +01:00
|
|
|
*/
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
static int __mem_cgroup_try_charge(struct mm_struct *mm,
|
2009-01-08 03:08:00 +01:00
|
|
|
gfp_t gfp_mask, struct mem_cgroup **memcg,
|
|
|
|
bool oom)
|
2008-02-07 09:13:53 +01:00
|
|
|
{
|
2009-01-08 03:08:06 +01:00
|
|
|
struct mem_cgroup *mem, *mem_over_limit;
|
2009-01-08 03:07:48 +01:00
|
|
|
int nr_retries = MEM_CGROUP_RECLAIM_RETRIES;
|
2009-01-08 03:08:05 +01:00
|
|
|
struct res_counter *fail_res;
|
2009-01-08 03:08:08 +01:00
|
|
|
|
|
|
|
if (unlikely(test_thread_flag(TIF_MEMDIE))) {
|
|
|
|
/* Don't account this! */
|
|
|
|
*memcg = NULL;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-02-07 09:13:53 +01:00
|
|
|
/*
|
2008-02-07 09:14:19 +01:00
|
|
|
* We always charge the cgroup the mm_struct belongs to.
|
|
|
|
* The mm_struct's mem_cgroup changes on task migration if the
|
2008-02-07 09:13:53 +01:00
|
|
|
* thread group leader migrates. It's possible that mm is not
|
|
|
|
* set, if so charge the init_mm (happens for pagecache usage).
|
|
|
|
*/
|
2009-01-08 03:07:48 +01:00
|
|
|
if (likely(!*memcg)) {
|
2008-07-25 10:47:10 +02:00
|
|
|
rcu_read_lock();
|
|
|
|
mem = mem_cgroup_from_task(rcu_dereference(mm->owner));
|
mm owner: fix race between swapoff and exit
There's a race between mm->owner assignment and swapoff, more easily
seen when task slab poisoning is turned on. The condition occurs when
try_to_unuse() runs in parallel with an exiting task. A similar race
can occur with callers of get_task_mm(), such as /proc/<pid>/<mmstats>
or ptrace or page migration.
CPU0 CPU1
try_to_unuse
looks at mm = task0->mm
increments mm->mm_users
task 0 exits
mm->owner needs to be updated, but no
new owner is found (mm_users > 1, but
no other task has task->mm = task0->mm)
mm_update_next_owner() leaves
mmput(mm) decrements mm->mm_users
task0 freed
dereferencing mm->owner fails
The fix is to notify the subsystem via mm_owner_changed callback(),
if no new owner is found, by specifying the new task as NULL.
Jiri Slaby:
mm->owner was set to NULL prior to calling cgroup_mm_owner_callbacks(), but
must be set after that, so as not to pass NULL as old owner causing oops.
Daisuke Nishimura:
mm_update_next_owner() may set mm->owner to NULL, but mem_cgroup_from_task()
and its callers need to take account of this situation to avoid oops.
Hugh Dickins:
Lockdep warning and hang below exec_mmap() when testing these patches.
exit_mm() up_reads mmap_sem before calling mm_update_next_owner(),
so exec_mmap() now needs to do the same. And with that repositioning,
there's now no point in mm_need_new_owner() allowing for NULL mm.
Reported-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-09-29 00:09:31 +02:00
|
|
|
if (unlikely(!mem)) {
|
|
|
|
rcu_read_unlock();
|
|
|
|
return 0;
|
|
|
|
}
|
2008-07-25 10:47:10 +02:00
|
|
|
/*
|
|
|
|
* For every charge from the cgroup, increment reference count
|
|
|
|
*/
|
|
|
|
css_get(&mem->css);
|
2009-01-08 03:07:48 +01:00
|
|
|
*memcg = mem;
|
2008-07-25 10:47:10 +02:00
|
|
|
rcu_read_unlock();
|
|
|
|
} else {
|
2009-01-08 03:07:48 +01:00
|
|
|
mem = *memcg;
|
|
|
|
css_get(&mem->css);
|
2008-07-25 10:47:10 +02:00
|
|
|
}
|
2008-02-07 09:13:53 +01:00
|
|
|
|
2009-01-08 03:08:00 +01:00
|
|
|
while (1) {
|
|
|
|
int ret;
|
|
|
|
bool noswap = false;
|
2009-01-08 03:07:48 +01:00
|
|
|
|
2009-01-08 03:08:05 +01:00
|
|
|
ret = res_counter_charge(&mem->res, PAGE_SIZE, &fail_res);
|
2009-01-08 03:08:00 +01:00
|
|
|
if (likely(!ret)) {
|
|
|
|
if (!do_swap_account)
|
|
|
|
break;
|
2009-01-08 03:08:05 +01:00
|
|
|
ret = res_counter_charge(&mem->memsw, PAGE_SIZE,
|
|
|
|
&fail_res);
|
2009-01-08 03:08:00 +01:00
|
|
|
if (likely(!ret))
|
|
|
|
break;
|
|
|
|
/* mem+swap counter fails */
|
|
|
|
res_counter_uncharge(&mem->res, PAGE_SIZE);
|
|
|
|
noswap = true;
|
2009-01-08 03:08:06 +01:00
|
|
|
mem_over_limit = mem_cgroup_from_res_counter(fail_res,
|
|
|
|
memsw);
|
|
|
|
} else
|
|
|
|
/* mem counter fails */
|
|
|
|
mem_over_limit = mem_cgroup_from_res_counter(fail_res,
|
|
|
|
res);
|
|
|
|
|
2008-02-07 09:14:19 +01:00
|
|
|
if (!(gfp_mask & __GFP_WAIT))
|
2009-01-08 03:07:48 +01:00
|
|
|
goto nomem;
|
2008-02-07 09:14:02 +01:00
|
|
|
|
2009-01-08 03:08:06 +01:00
|
|
|
ret = mem_cgroup_hierarchical_reclaim(mem_over_limit, gfp_mask,
|
|
|
|
noswap);
|
2008-02-07 09:13:56 +01:00
|
|
|
|
|
|
|
/*
|
2008-03-04 23:29:09 +01:00
|
|
|
* try_to_free_mem_cgroup_pages() might not give us a full
|
|
|
|
* picture of reclaim. Some pages are reclaimed and might be
|
|
|
|
* moved to swap cache or just unmapped from the cgroup.
|
|
|
|
* Check the limit again to see if the reclaim reduced the
|
|
|
|
* current usage of the cgroup before giving up
|
2009-01-08 03:08:00 +01:00
|
|
|
*
|
2008-03-04 23:29:09 +01:00
|
|
|
*/
|
2009-01-08 03:08:12 +01:00
|
|
|
if (mem_cgroup_check_under_limit(mem_over_limit))
|
|
|
|
continue;
|
2008-02-07 09:14:19 +01:00
|
|
|
|
|
|
|
if (!nr_retries--) {
|
2009-01-08 03:08:08 +01:00
|
|
|
if (oom) {
|
2009-01-08 03:08:09 +01:00
|
|
|
mem_cgroup_out_of_memory(mem_over_limit, gfp_mask);
|
|
|
|
mem_over_limit->last_oom_jiffies = jiffies;
|
2009-01-08 03:08:08 +01:00
|
|
|
}
|
2009-01-08 03:07:48 +01:00
|
|
|
goto nomem;
|
2008-02-07 09:13:56 +01:00
|
|
|
}
|
2008-02-07 09:13:53 +01:00
|
|
|
}
|
2009-01-08 03:07:48 +01:00
|
|
|
return 0;
|
|
|
|
nomem:
|
|
|
|
css_put(&mem->css);
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
2008-02-07 09:13:53 +01:00
|
|
|
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
/**
|
|
|
|
* mem_cgroup_try_charge - get charge of PAGE_SIZE.
|
|
|
|
* @mm: an mm_struct which is charged against. (when *memcg is NULL)
|
|
|
|
* @gfp_mask: gfp_mask for reclaim.
|
|
|
|
* @memcg: a pointer to memory cgroup which is charged against.
|
|
|
|
*
|
|
|
|
* charge against memory cgroup pointed by *memcg. if *memcg == NULL, estimated
|
|
|
|
* memory cgroup from @mm is got and stored in *memcg.
|
|
|
|
*
|
|
|
|
* Returns 0 if success. -ENOMEM at failure.
|
|
|
|
* This call can invoke OOM-Killer.
|
|
|
|
*/
|
|
|
|
|
|
|
|
int mem_cgroup_try_charge(struct mm_struct *mm,
|
|
|
|
gfp_t mask, struct mem_cgroup **memcg)
|
|
|
|
{
|
|
|
|
return __mem_cgroup_try_charge(mm, mask, memcg, true);
|
|
|
|
}
|
|
|
|
|
2009-01-08 03:07:48 +01:00
|
|
|
/*
|
|
|
|
* commit a charge got by mem_cgroup_try_charge() and makes page_cgroup to be
|
|
|
|
* USED state. If already USED, uncharge and return.
|
|
|
|
*/
|
|
|
|
|
|
|
|
static void __mem_cgroup_commit_charge(struct mem_cgroup *mem,
|
|
|
|
struct page_cgroup *pc,
|
|
|
|
enum charge_type ctype)
|
|
|
|
{
|
|
|
|
/* try_charge() can return NULL to *memcg, taking care of it. */
|
|
|
|
if (!mem)
|
|
|
|
return;
|
2008-10-19 05:28:16 +02:00
|
|
|
|
|
|
|
lock_page_cgroup(pc);
|
|
|
|
if (unlikely(PageCgroupUsed(pc))) {
|
|
|
|
unlock_page_cgroup(pc);
|
|
|
|
res_counter_uncharge(&mem->res, PAGE_SIZE);
|
2009-01-08 03:08:00 +01:00
|
|
|
if (do_swap_account)
|
|
|
|
res_counter_uncharge(&mem->memsw, PAGE_SIZE);
|
2008-10-19 05:28:16 +02:00
|
|
|
css_put(&mem->css);
|
2009-01-08 03:07:48 +01:00
|
|
|
return;
|
2008-10-19 05:28:16 +02:00
|
|
|
}
|
2008-02-07 09:13:53 +01:00
|
|
|
pc->mem_cgroup = mem;
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
smp_wmb();
|
2008-10-19 05:28:11 +02:00
|
|
|
pc->flags = pcg_default_flags[ctype];
|
2008-02-07 09:14:19 +01:00
|
|
|
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
mem_cgroup_charge_statistics(mem, pc, true);
|
2008-10-19 05:28:16 +02:00
|
|
|
|
|
|
|
unlock_page_cgroup(pc);
|
2009-01-08 03:07:48 +01:00
|
|
|
}
|
2008-02-07 09:13:56 +01:00
|
|
|
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
/**
|
|
|
|
* mem_cgroup_move_account - move account of the page
|
|
|
|
* @pc: page_cgroup of the page.
|
|
|
|
* @from: mem_cgroup which the page is moved from.
|
|
|
|
* @to: mem_cgroup which the page is moved to. @from != @to.
|
|
|
|
*
|
|
|
|
* The caller must confirm following.
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
* - page is not on LRU (isolate_page() is useful.)
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
*
|
|
|
|
* returns 0 at success,
|
|
|
|
* returns -EBUSY when lock is busy or "pc" is unstable.
|
|
|
|
*
|
|
|
|
* This function does "uncharge" from old cgroup but doesn't do "charge" to
|
|
|
|
* new cgroup. It should be done by a caller.
|
|
|
|
*/
|
|
|
|
|
|
|
|
static int mem_cgroup_move_account(struct page_cgroup *pc,
|
|
|
|
struct mem_cgroup *from, struct mem_cgroup *to)
|
|
|
|
{
|
|
|
|
struct mem_cgroup_per_zone *from_mz, *to_mz;
|
|
|
|
int nid, zid;
|
|
|
|
int ret = -EBUSY;
|
|
|
|
|
|
|
|
VM_BUG_ON(from == to);
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
VM_BUG_ON(PageLRU(pc->page));
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
|
|
|
|
nid = page_cgroup_nid(pc);
|
|
|
|
zid = page_cgroup_zid(pc);
|
|
|
|
from_mz = mem_cgroup_zoneinfo(from, nid, zid);
|
|
|
|
to_mz = mem_cgroup_zoneinfo(to, nid, zid);
|
|
|
|
|
|
|
|
if (!trylock_page_cgroup(pc))
|
|
|
|
return ret;
|
|
|
|
|
|
|
|
if (!PageCgroupUsed(pc))
|
|
|
|
goto out;
|
|
|
|
|
|
|
|
if (pc->mem_cgroup != from)
|
|
|
|
goto out;
|
|
|
|
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
css_put(&from->css);
|
|
|
|
res_counter_uncharge(&from->res, PAGE_SIZE);
|
|
|
|
mem_cgroup_charge_statistics(from, pc, false);
|
|
|
|
if (do_swap_account)
|
|
|
|
res_counter_uncharge(&from->memsw, PAGE_SIZE);
|
|
|
|
pc->mem_cgroup = to;
|
|
|
|
mem_cgroup_charge_statistics(to, pc, true);
|
|
|
|
css_get(&to->css);
|
|
|
|
ret = 0;
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
out:
|
|
|
|
unlock_page_cgroup(pc);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* move charges to its parent.
|
|
|
|
*/
|
|
|
|
|
|
|
|
static int mem_cgroup_move_parent(struct page_cgroup *pc,
|
|
|
|
struct mem_cgroup *child,
|
|
|
|
gfp_t gfp_mask)
|
|
|
|
{
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
struct page *page = pc->page;
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
struct cgroup *cg = child->css.cgroup;
|
|
|
|
struct cgroup *pcg = cg->parent;
|
|
|
|
struct mem_cgroup *parent;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
/* Is ROOT ? */
|
|
|
|
if (!pcg)
|
|
|
|
return -EINVAL;
|
|
|
|
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
parent = mem_cgroup_from_cont(pcg);
|
|
|
|
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
ret = __mem_cgroup_try_charge(NULL, gfp_mask, &parent, false);
|
2009-01-08 03:08:08 +01:00
|
|
|
if (ret || !parent)
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
return ret;
|
|
|
|
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
if (!get_page_unless_zero(page))
|
|
|
|
return -EBUSY;
|
|
|
|
|
|
|
|
ret = isolate_lru_page(page);
|
|
|
|
|
|
|
|
if (ret)
|
|
|
|
goto cancel;
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
|
|
|
|
ret = mem_cgroup_move_account(pc, child, parent);
|
|
|
|
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
/* drop extra refcnt by try_charge() (move_account increment one) */
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
css_put(&parent->css);
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
putback_lru_page(page);
|
|
|
|
if (!ret) {
|
|
|
|
put_page(page);
|
|
|
|
return 0;
|
2009-01-08 03:08:00 +01:00
|
|
|
}
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
/* uncharge if move fails */
|
|
|
|
cancel:
|
|
|
|
res_counter_uncharge(&parent->res, PAGE_SIZE);
|
|
|
|
if (do_swap_account)
|
|
|
|
res_counter_uncharge(&parent->memsw, PAGE_SIZE);
|
|
|
|
put_page(page);
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2009-01-08 03:07:48 +01:00
|
|
|
/*
|
|
|
|
* Charge the memory controller for page usage.
|
|
|
|
* Return
|
|
|
|
* 0 if the charge was successful
|
|
|
|
* < 0 if the cgroup is over its limit
|
|
|
|
*/
|
|
|
|
static int mem_cgroup_charge_common(struct page *page, struct mm_struct *mm,
|
|
|
|
gfp_t gfp_mask, enum charge_type ctype,
|
|
|
|
struct mem_cgroup *memcg)
|
|
|
|
{
|
|
|
|
struct mem_cgroup *mem;
|
|
|
|
struct page_cgroup *pc;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
pc = lookup_page_cgroup(page);
|
|
|
|
/* can happen at boot */
|
|
|
|
if (unlikely(!pc))
|
|
|
|
return 0;
|
|
|
|
prefetchw(pc);
|
|
|
|
|
|
|
|
mem = memcg;
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
ret = __mem_cgroup_try_charge(mm, gfp_mask, &mem, true);
|
2009-01-08 03:08:08 +01:00
|
|
|
if (ret || !mem)
|
2009-01-08 03:07:48 +01:00
|
|
|
return ret;
|
|
|
|
|
|
|
|
__mem_cgroup_commit_charge(mem, pc, ctype);
|
2008-02-07 09:13:53 +01:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2009-01-08 03:07:48 +01:00
|
|
|
int mem_cgroup_newpage_charge(struct page *page,
|
|
|
|
struct mm_struct *mm, gfp_t gfp_mask)
|
2008-02-07 09:14:17 +01:00
|
|
|
{
|
2009-01-08 03:08:02 +01:00
|
|
|
if (mem_cgroup_disabled())
|
2008-07-25 10:47:18 +02:00
|
|
|
return 0;
|
2008-10-19 05:28:16 +02:00
|
|
|
if (PageCompound(page))
|
|
|
|
return 0;
|
memcg: remove refcnt from page_cgroup
memcg: performance improvements
Patch Description
1/5 ... remove refcnt fron page_cgroup patch (shmem handling is fixed)
2/5 ... swapcache handling patch
3/5 ... add helper function for shmem's memory reclaim patch
4/5 ... optimize by likely/unlikely ppatch
5/5 ... remove redundunt check patch (shmem handling is fixed.)
Unix bench result.
== 2.6.26-rc2-mm1 + memory resource controller
Execl Throughput 2915.4 lps (29.6 secs, 3 samples)
C Compiler Throughput 1019.3 lpm (60.0 secs, 3 samples)
Shell Scripts (1 concurrent) 5796.0 lpm (60.0 secs, 3 samples)
Shell Scripts (8 concurrent) 1097.7 lpm (60.0 secs, 3 samples)
Shell Scripts (16 concurrent) 565.3 lpm (60.0 secs, 3 samples)
File Read 1024 bufsize 2000 maxblocks 1022128.0 KBps (30.0 secs, 3 samples)
File Write 1024 bufsize 2000 maxblocks 544057.0 KBps (30.0 secs, 3 samples)
File Copy 1024 bufsize 2000 maxblocks 346481.0 KBps (30.0 secs, 3 samples)
File Read 256 bufsize 500 maxblocks 319325.0 KBps (30.0 secs, 3 samples)
File Write 256 bufsize 500 maxblocks 148788.0 KBps (30.0 secs, 3 samples)
File Copy 256 bufsize 500 maxblocks 99051.0 KBps (30.0 secs, 3 samples)
File Read 4096 bufsize 8000 maxblocks 2058917.0 KBps (30.0 secs, 3 samples)
File Write 4096 bufsize 8000 maxblocks 1606109.0 KBps (30.0 secs, 3 samples)
File Copy 4096 bufsize 8000 maxblocks 854789.0 KBps (30.0 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places 126145.2 lpm (30.0 secs, 3 samples)
INDEX VALUES
TEST BASELINE RESULT INDEX
Execl Throughput 43.0 2915.4 678.0
File Copy 1024 bufsize 2000 maxblocks 3960.0 346481.0 875.0
File Copy 256 bufsize 500 maxblocks 1655.0 99051.0 598.5
File Copy 4096 bufsize 8000 maxblocks 5800.0 854789.0 1473.8
Shell Scripts (8 concurrent) 6.0 1097.7 1829.5
=========
FINAL SCORE 991.3
== 2.6.26-rc2-mm1 + this set ==
Execl Throughput 3012.9 lps (29.9 secs, 3 samples)
C Compiler Throughput 981.0 lpm (60.0 secs, 3 samples)
Shell Scripts (1 concurrent) 5872.0 lpm (60.0 secs, 3 samples)
Shell Scripts (8 concurrent) 1120.3 lpm (60.0 secs, 3 samples)
Shell Scripts (16 concurrent) 578.0 lpm (60.0 secs, 3 samples)
File Read 1024 bufsize 2000 maxblocks 1003993.0 KBps (30.0 secs, 3 samples)
File Write 1024 bufsize 2000 maxblocks 550452.0 KBps (30.0 secs, 3 samples)
File Copy 1024 bufsize 2000 maxblocks 347159.0 KBps (30.0 secs, 3 samples)
File Read 256 bufsize 500 maxblocks 314644.0 KBps (30.0 secs, 3 samples)
File Write 256 bufsize 500 maxblocks 151852.0 KBps (30.0 secs, 3 samples)
File Copy 256 bufsize 500 maxblocks 101000.0 KBps (30.0 secs, 3 samples)
File Read 4096 bufsize 8000 maxblocks 2033256.0 KBps (30.0 secs, 3 samples)
File Write 4096 bufsize 8000 maxblocks 1611814.0 KBps (30.0 secs, 3 samples)
File Copy 4096 bufsize 8000 maxblocks 847979.0 KBps (30.0 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places 128148.7 lpm (30.0 secs, 3 samples)
INDEX VALUES
TEST BASELINE RESULT INDEX
Execl Throughput 43.0 3012.9 700.7
File Copy 1024 bufsize 2000 maxblocks 3960.0 347159.0 876.7
File Copy 256 bufsize 500 maxblocks 1655.0 101000.0 610.3
File Copy 4096 bufsize 8000 maxblocks 5800.0 847979.0 1462.0
Shell Scripts (8 concurrent) 6.0 1120.3 1867.2
=========
FINAL SCORE 1004.6
This patch:
Remove refcnt from page_cgroup().
After this,
* A page is charged only when !page_mapped() && no page_cgroup is assigned.
* Anon page is newly mapped.
* File page is added to mapping->tree.
* A page is uncharged only when
* Anon page is fully unmapped.
* File page is removed from LRU.
There is no change in behavior from user's view.
This patch also removes unnecessary calls in rmap.c which was used only for
refcnt mangement.
[akpm@linux-foundation.org: fix warning]
[hugh@veritas.com: fix shmem_unuse_inode charging]
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-25 10:47:14 +02:00
|
|
|
/*
|
|
|
|
* If already mapped, we don't have to account.
|
|
|
|
* If page cache, page->mapping has address_space.
|
|
|
|
* But page->mapping may have out-of-use anon_vma pointer,
|
|
|
|
* detecit it by PageAnon() check. newly-mapped-anon's page->mapping
|
|
|
|
* is NULL.
|
|
|
|
*/
|
|
|
|
if (page_mapped(page) || (page->mapping && !PageAnon(page)))
|
|
|
|
return 0;
|
|
|
|
if (unlikely(!mm))
|
|
|
|
mm = &init_mm;
|
2008-02-07 09:14:17 +01:00
|
|
|
return mem_cgroup_charge_common(page, mm, gfp_mask,
|
2008-07-25 10:47:10 +02:00
|
|
|
MEM_CGROUP_CHARGE_TYPE_MAPPED, NULL);
|
2008-02-07 09:14:17 +01:00
|
|
|
}
|
|
|
|
|
2008-02-07 09:14:02 +01:00
|
|
|
int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm,
|
|
|
|
gfp_t gfp_mask)
|
2008-02-07 09:13:59 +01:00
|
|
|
{
|
2009-01-08 03:08:02 +01:00
|
|
|
if (mem_cgroup_disabled())
|
2008-07-25 10:47:18 +02:00
|
|
|
return 0;
|
2008-10-19 05:28:16 +02:00
|
|
|
if (PageCompound(page))
|
|
|
|
return 0;
|
2008-07-25 10:47:17 +02:00
|
|
|
/*
|
|
|
|
* Corner case handling. This is called from add_to_page_cache()
|
|
|
|
* in usual. But some FS (shmem) precharges this page before calling it
|
|
|
|
* and call add_to_page_cache() with GFP_NOWAIT.
|
|
|
|
*
|
|
|
|
* For GFP_NOWAIT case, the page may be pre-charged before calling
|
|
|
|
* add_to_page_cache(). (See shmem.c) check it here and avoid to call
|
|
|
|
* charge twice. (It works but has to pay a bit larger cost.)
|
|
|
|
*/
|
|
|
|
if (!(gfp_mask & __GFP_WAIT)) {
|
|
|
|
struct page_cgroup *pc;
|
|
|
|
|
2008-10-19 05:28:16 +02:00
|
|
|
|
|
|
|
pc = lookup_page_cgroup(page);
|
|
|
|
if (!pc)
|
|
|
|
return 0;
|
|
|
|
lock_page_cgroup(pc);
|
|
|
|
if (PageCgroupUsed(pc)) {
|
|
|
|
unlock_page_cgroup(pc);
|
2008-07-25 10:47:17 +02:00
|
|
|
return 0;
|
|
|
|
}
|
2008-10-19 05:28:16 +02:00
|
|
|
unlock_page_cgroup(pc);
|
2008-07-25 10:47:17 +02:00
|
|
|
}
|
|
|
|
|
memcg: remove refcnt from page_cgroup
memcg: performance improvements
Patch Description
1/5 ... remove refcnt fron page_cgroup patch (shmem handling is fixed)
2/5 ... swapcache handling patch
3/5 ... add helper function for shmem's memory reclaim patch
4/5 ... optimize by likely/unlikely ppatch
5/5 ... remove redundunt check patch (shmem handling is fixed.)
Unix bench result.
== 2.6.26-rc2-mm1 + memory resource controller
Execl Throughput 2915.4 lps (29.6 secs, 3 samples)
C Compiler Throughput 1019.3 lpm (60.0 secs, 3 samples)
Shell Scripts (1 concurrent) 5796.0 lpm (60.0 secs, 3 samples)
Shell Scripts (8 concurrent) 1097.7 lpm (60.0 secs, 3 samples)
Shell Scripts (16 concurrent) 565.3 lpm (60.0 secs, 3 samples)
File Read 1024 bufsize 2000 maxblocks 1022128.0 KBps (30.0 secs, 3 samples)
File Write 1024 bufsize 2000 maxblocks 544057.0 KBps (30.0 secs, 3 samples)
File Copy 1024 bufsize 2000 maxblocks 346481.0 KBps (30.0 secs, 3 samples)
File Read 256 bufsize 500 maxblocks 319325.0 KBps (30.0 secs, 3 samples)
File Write 256 bufsize 500 maxblocks 148788.0 KBps (30.0 secs, 3 samples)
File Copy 256 bufsize 500 maxblocks 99051.0 KBps (30.0 secs, 3 samples)
File Read 4096 bufsize 8000 maxblocks 2058917.0 KBps (30.0 secs, 3 samples)
File Write 4096 bufsize 8000 maxblocks 1606109.0 KBps (30.0 secs, 3 samples)
File Copy 4096 bufsize 8000 maxblocks 854789.0 KBps (30.0 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places 126145.2 lpm (30.0 secs, 3 samples)
INDEX VALUES
TEST BASELINE RESULT INDEX
Execl Throughput 43.0 2915.4 678.0
File Copy 1024 bufsize 2000 maxblocks 3960.0 346481.0 875.0
File Copy 256 bufsize 500 maxblocks 1655.0 99051.0 598.5
File Copy 4096 bufsize 8000 maxblocks 5800.0 854789.0 1473.8
Shell Scripts (8 concurrent) 6.0 1097.7 1829.5
=========
FINAL SCORE 991.3
== 2.6.26-rc2-mm1 + this set ==
Execl Throughput 3012.9 lps (29.9 secs, 3 samples)
C Compiler Throughput 981.0 lpm (60.0 secs, 3 samples)
Shell Scripts (1 concurrent) 5872.0 lpm (60.0 secs, 3 samples)
Shell Scripts (8 concurrent) 1120.3 lpm (60.0 secs, 3 samples)
Shell Scripts (16 concurrent) 578.0 lpm (60.0 secs, 3 samples)
File Read 1024 bufsize 2000 maxblocks 1003993.0 KBps (30.0 secs, 3 samples)
File Write 1024 bufsize 2000 maxblocks 550452.0 KBps (30.0 secs, 3 samples)
File Copy 1024 bufsize 2000 maxblocks 347159.0 KBps (30.0 secs, 3 samples)
File Read 256 bufsize 500 maxblocks 314644.0 KBps (30.0 secs, 3 samples)
File Write 256 bufsize 500 maxblocks 151852.0 KBps (30.0 secs, 3 samples)
File Copy 256 bufsize 500 maxblocks 101000.0 KBps (30.0 secs, 3 samples)
File Read 4096 bufsize 8000 maxblocks 2033256.0 KBps (30.0 secs, 3 samples)
File Write 4096 bufsize 8000 maxblocks 1611814.0 KBps (30.0 secs, 3 samples)
File Copy 4096 bufsize 8000 maxblocks 847979.0 KBps (30.0 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places 128148.7 lpm (30.0 secs, 3 samples)
INDEX VALUES
TEST BASELINE RESULT INDEX
Execl Throughput 43.0 3012.9 700.7
File Copy 1024 bufsize 2000 maxblocks 3960.0 347159.0 876.7
File Copy 256 bufsize 500 maxblocks 1655.0 101000.0 610.3
File Copy 4096 bufsize 8000 maxblocks 5800.0 847979.0 1462.0
Shell Scripts (8 concurrent) 6.0 1120.3 1867.2
=========
FINAL SCORE 1004.6
This patch:
Remove refcnt from page_cgroup().
After this,
* A page is charged only when !page_mapped() && no page_cgroup is assigned.
* Anon page is newly mapped.
* File page is added to mapping->tree.
* A page is uncharged only when
* Anon page is fully unmapped.
* File page is removed from LRU.
There is no change in behavior from user's view.
This patch also removes unnecessary calls in rmap.c which was used only for
refcnt mangement.
[akpm@linux-foundation.org: fix warning]
[hugh@veritas.com: fix shmem_unuse_inode charging]
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-25 10:47:14 +02:00
|
|
|
if (unlikely(!mm))
|
2008-02-07 09:13:59 +01:00
|
|
|
mm = &init_mm;
|
2008-07-25 10:47:17 +02:00
|
|
|
|
2008-10-19 05:28:11 +02:00
|
|
|
if (page_is_file_cache(page))
|
|
|
|
return mem_cgroup_charge_common(page, mm, gfp_mask,
|
2008-07-25 10:47:10 +02:00
|
|
|
MEM_CGROUP_CHARGE_TYPE_CACHE, NULL);
|
2008-10-19 05:28:11 +02:00
|
|
|
else
|
|
|
|
return mem_cgroup_charge_common(page, mm, gfp_mask,
|
|
|
|
MEM_CGROUP_CHARGE_TYPE_SHMEM, NULL);
|
2008-07-25 10:47:10 +02:00
|
|
|
}
|
|
|
|
|
2009-01-08 03:08:00 +01:00
|
|
|
int mem_cgroup_try_charge_swapin(struct mm_struct *mm,
|
|
|
|
struct page *page,
|
|
|
|
gfp_t mask, struct mem_cgroup **ptr)
|
|
|
|
{
|
|
|
|
struct mem_cgroup *mem;
|
|
|
|
swp_entry_t ent;
|
|
|
|
|
2009-01-08 03:08:02 +01:00
|
|
|
if (mem_cgroup_disabled())
|
2009-01-08 03:08:00 +01:00
|
|
|
return 0;
|
|
|
|
|
|
|
|
if (!do_swap_account)
|
|
|
|
goto charge_cur_mm;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* A racing thread's fault, or swapoff, may have already updated
|
|
|
|
* the pte, and even removed page from swap cache: return success
|
|
|
|
* to go on to do_swap_page()'s pte_same() test, which should fail.
|
|
|
|
*/
|
|
|
|
if (!PageSwapCache(page))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
ent.val = page_private(page);
|
|
|
|
|
|
|
|
mem = lookup_swap_cgroup(ent);
|
|
|
|
if (!mem || mem->obsolete)
|
|
|
|
goto charge_cur_mm;
|
|
|
|
*ptr = mem;
|
|
|
|
return __mem_cgroup_try_charge(NULL, mask, ptr, true);
|
|
|
|
charge_cur_mm:
|
|
|
|
if (unlikely(!mm))
|
|
|
|
mm = &init_mm;
|
|
|
|
return __mem_cgroup_try_charge(mm, mask, ptr, true);
|
|
|
|
}
|
|
|
|
|
2009-01-08 03:07:56 +01:00
|
|
|
#ifdef CONFIG_SWAP
|
2009-01-08 03:08:00 +01:00
|
|
|
|
2009-01-08 03:07:56 +01:00
|
|
|
int mem_cgroup_cache_charge_swapin(struct page *page,
|
|
|
|
struct mm_struct *mm, gfp_t mask, bool locked)
|
|
|
|
{
|
|
|
|
int ret = 0;
|
|
|
|
|
2009-01-08 03:08:02 +01:00
|
|
|
if (mem_cgroup_disabled())
|
2009-01-08 03:07:56 +01:00
|
|
|
return 0;
|
|
|
|
if (unlikely(!mm))
|
|
|
|
mm = &init_mm;
|
|
|
|
if (!locked)
|
|
|
|
lock_page(page);
|
|
|
|
/*
|
|
|
|
* If not locked, the page can be dropped from SwapCache until
|
|
|
|
* we reach here.
|
|
|
|
*/
|
|
|
|
if (PageSwapCache(page)) {
|
2009-01-08 03:08:00 +01:00
|
|
|
struct mem_cgroup *mem = NULL;
|
|
|
|
swp_entry_t ent;
|
|
|
|
|
|
|
|
ent.val = page_private(page);
|
|
|
|
if (do_swap_account) {
|
|
|
|
mem = lookup_swap_cgroup(ent);
|
|
|
|
if (mem && mem->obsolete)
|
|
|
|
mem = NULL;
|
|
|
|
if (mem)
|
|
|
|
mm = NULL;
|
|
|
|
}
|
2009-01-08 03:07:56 +01:00
|
|
|
ret = mem_cgroup_charge_common(page, mm, mask,
|
2009-01-08 03:08:00 +01:00
|
|
|
MEM_CGROUP_CHARGE_TYPE_SHMEM, mem);
|
|
|
|
|
|
|
|
if (!ret && do_swap_account) {
|
|
|
|
/* avoid double counting */
|
|
|
|
mem = swap_cgroup_record(ent, NULL);
|
|
|
|
if (mem) {
|
|
|
|
res_counter_uncharge(&mem->memsw, PAGE_SIZE);
|
|
|
|
mem_cgroup_put(mem);
|
|
|
|
}
|
|
|
|
}
|
2009-01-08 03:07:56 +01:00
|
|
|
}
|
|
|
|
if (!locked)
|
|
|
|
unlock_page(page);
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
/* add this page(page_cgroup) to the LRU we want. */
|
|
|
|
mem_cgroup_lru_fixup(page);
|
2009-01-08 03:07:56 +01:00
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2009-01-08 03:07:48 +01:00
|
|
|
void mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr)
|
|
|
|
{
|
|
|
|
struct page_cgroup *pc;
|
|
|
|
|
2009-01-08 03:08:02 +01:00
|
|
|
if (mem_cgroup_disabled())
|
2009-01-08 03:07:48 +01:00
|
|
|
return;
|
|
|
|
if (!ptr)
|
|
|
|
return;
|
|
|
|
pc = lookup_page_cgroup(page);
|
|
|
|
__mem_cgroup_commit_charge(ptr, pc, MEM_CGROUP_CHARGE_TYPE_MAPPED);
|
2009-01-08 03:08:00 +01:00
|
|
|
/*
|
|
|
|
* Now swap is on-memory. This means this page may be
|
|
|
|
* counted both as mem and swap....double count.
|
|
|
|
* Fix it by uncharging from memsw. This SwapCache is stable
|
|
|
|
* because we're still under lock_page().
|
|
|
|
*/
|
|
|
|
if (do_swap_account) {
|
|
|
|
swp_entry_t ent = {.val = page_private(page)};
|
|
|
|
struct mem_cgroup *memcg;
|
|
|
|
memcg = swap_cgroup_record(ent, NULL);
|
|
|
|
if (memcg) {
|
|
|
|
/* If memcg is obsolete, memcg can be != ptr */
|
|
|
|
res_counter_uncharge(&memcg->memsw, PAGE_SIZE);
|
|
|
|
mem_cgroup_put(memcg);
|
|
|
|
}
|
|
|
|
|
|
|
|
}
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
/* add this page(page_cgroup) to the LRU we want. */
|
|
|
|
mem_cgroup_lru_fixup(page);
|
2009-01-08 03:07:48 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
void mem_cgroup_cancel_charge_swapin(struct mem_cgroup *mem)
|
|
|
|
{
|
2009-01-08 03:08:02 +01:00
|
|
|
if (mem_cgroup_disabled())
|
2009-01-08 03:07:48 +01:00
|
|
|
return;
|
|
|
|
if (!mem)
|
|
|
|
return;
|
|
|
|
res_counter_uncharge(&mem->res, PAGE_SIZE);
|
2009-01-08 03:08:00 +01:00
|
|
|
if (do_swap_account)
|
|
|
|
res_counter_uncharge(&mem->memsw, PAGE_SIZE);
|
2009-01-08 03:07:48 +01:00
|
|
|
css_put(&mem->css);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2008-02-07 09:13:53 +01:00
|
|
|
/*
|
memcg: remove refcnt from page_cgroup
memcg: performance improvements
Patch Description
1/5 ... remove refcnt fron page_cgroup patch (shmem handling is fixed)
2/5 ... swapcache handling patch
3/5 ... add helper function for shmem's memory reclaim patch
4/5 ... optimize by likely/unlikely ppatch
5/5 ... remove redundunt check patch (shmem handling is fixed.)
Unix bench result.
== 2.6.26-rc2-mm1 + memory resource controller
Execl Throughput 2915.4 lps (29.6 secs, 3 samples)
C Compiler Throughput 1019.3 lpm (60.0 secs, 3 samples)
Shell Scripts (1 concurrent) 5796.0 lpm (60.0 secs, 3 samples)
Shell Scripts (8 concurrent) 1097.7 lpm (60.0 secs, 3 samples)
Shell Scripts (16 concurrent) 565.3 lpm (60.0 secs, 3 samples)
File Read 1024 bufsize 2000 maxblocks 1022128.0 KBps (30.0 secs, 3 samples)
File Write 1024 bufsize 2000 maxblocks 544057.0 KBps (30.0 secs, 3 samples)
File Copy 1024 bufsize 2000 maxblocks 346481.0 KBps (30.0 secs, 3 samples)
File Read 256 bufsize 500 maxblocks 319325.0 KBps (30.0 secs, 3 samples)
File Write 256 bufsize 500 maxblocks 148788.0 KBps (30.0 secs, 3 samples)
File Copy 256 bufsize 500 maxblocks 99051.0 KBps (30.0 secs, 3 samples)
File Read 4096 bufsize 8000 maxblocks 2058917.0 KBps (30.0 secs, 3 samples)
File Write 4096 bufsize 8000 maxblocks 1606109.0 KBps (30.0 secs, 3 samples)
File Copy 4096 bufsize 8000 maxblocks 854789.0 KBps (30.0 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places 126145.2 lpm (30.0 secs, 3 samples)
INDEX VALUES
TEST BASELINE RESULT INDEX
Execl Throughput 43.0 2915.4 678.0
File Copy 1024 bufsize 2000 maxblocks 3960.0 346481.0 875.0
File Copy 256 bufsize 500 maxblocks 1655.0 99051.0 598.5
File Copy 4096 bufsize 8000 maxblocks 5800.0 854789.0 1473.8
Shell Scripts (8 concurrent) 6.0 1097.7 1829.5
=========
FINAL SCORE 991.3
== 2.6.26-rc2-mm1 + this set ==
Execl Throughput 3012.9 lps (29.9 secs, 3 samples)
C Compiler Throughput 981.0 lpm (60.0 secs, 3 samples)
Shell Scripts (1 concurrent) 5872.0 lpm (60.0 secs, 3 samples)
Shell Scripts (8 concurrent) 1120.3 lpm (60.0 secs, 3 samples)
Shell Scripts (16 concurrent) 578.0 lpm (60.0 secs, 3 samples)
File Read 1024 bufsize 2000 maxblocks 1003993.0 KBps (30.0 secs, 3 samples)
File Write 1024 bufsize 2000 maxblocks 550452.0 KBps (30.0 secs, 3 samples)
File Copy 1024 bufsize 2000 maxblocks 347159.0 KBps (30.0 secs, 3 samples)
File Read 256 bufsize 500 maxblocks 314644.0 KBps (30.0 secs, 3 samples)
File Write 256 bufsize 500 maxblocks 151852.0 KBps (30.0 secs, 3 samples)
File Copy 256 bufsize 500 maxblocks 101000.0 KBps (30.0 secs, 3 samples)
File Read 4096 bufsize 8000 maxblocks 2033256.0 KBps (30.0 secs, 3 samples)
File Write 4096 bufsize 8000 maxblocks 1611814.0 KBps (30.0 secs, 3 samples)
File Copy 4096 bufsize 8000 maxblocks 847979.0 KBps (30.0 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places 128148.7 lpm (30.0 secs, 3 samples)
INDEX VALUES
TEST BASELINE RESULT INDEX
Execl Throughput 43.0 3012.9 700.7
File Copy 1024 bufsize 2000 maxblocks 3960.0 347159.0 876.7
File Copy 256 bufsize 500 maxblocks 1655.0 101000.0 610.3
File Copy 4096 bufsize 8000 maxblocks 5800.0 847979.0 1462.0
Shell Scripts (8 concurrent) 6.0 1120.3 1867.2
=========
FINAL SCORE 1004.6
This patch:
Remove refcnt from page_cgroup().
After this,
* A page is charged only when !page_mapped() && no page_cgroup is assigned.
* Anon page is newly mapped.
* File page is added to mapping->tree.
* A page is uncharged only when
* Anon page is fully unmapped.
* File page is removed from LRU.
There is no change in behavior from user's view.
This patch also removes unnecessary calls in rmap.c which was used only for
refcnt mangement.
[akpm@linux-foundation.org: fix warning]
[hugh@veritas.com: fix shmem_unuse_inode charging]
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-25 10:47:14 +02:00
|
|
|
* uncharge if !page_mapped(page)
|
2008-02-07 09:13:53 +01:00
|
|
|
*/
|
2009-01-08 03:08:00 +01:00
|
|
|
static struct mem_cgroup *
|
memcg: remove refcnt from page_cgroup
memcg: performance improvements
Patch Description
1/5 ... remove refcnt fron page_cgroup patch (shmem handling is fixed)
2/5 ... swapcache handling patch
3/5 ... add helper function for shmem's memory reclaim patch
4/5 ... optimize by likely/unlikely ppatch
5/5 ... remove redundunt check patch (shmem handling is fixed.)
Unix bench result.
== 2.6.26-rc2-mm1 + memory resource controller
Execl Throughput 2915.4 lps (29.6 secs, 3 samples)
C Compiler Throughput 1019.3 lpm (60.0 secs, 3 samples)
Shell Scripts (1 concurrent) 5796.0 lpm (60.0 secs, 3 samples)
Shell Scripts (8 concurrent) 1097.7 lpm (60.0 secs, 3 samples)
Shell Scripts (16 concurrent) 565.3 lpm (60.0 secs, 3 samples)
File Read 1024 bufsize 2000 maxblocks 1022128.0 KBps (30.0 secs, 3 samples)
File Write 1024 bufsize 2000 maxblocks 544057.0 KBps (30.0 secs, 3 samples)
File Copy 1024 bufsize 2000 maxblocks 346481.0 KBps (30.0 secs, 3 samples)
File Read 256 bufsize 500 maxblocks 319325.0 KBps (30.0 secs, 3 samples)
File Write 256 bufsize 500 maxblocks 148788.0 KBps (30.0 secs, 3 samples)
File Copy 256 bufsize 500 maxblocks 99051.0 KBps (30.0 secs, 3 samples)
File Read 4096 bufsize 8000 maxblocks 2058917.0 KBps (30.0 secs, 3 samples)
File Write 4096 bufsize 8000 maxblocks 1606109.0 KBps (30.0 secs, 3 samples)
File Copy 4096 bufsize 8000 maxblocks 854789.0 KBps (30.0 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places 126145.2 lpm (30.0 secs, 3 samples)
INDEX VALUES
TEST BASELINE RESULT INDEX
Execl Throughput 43.0 2915.4 678.0
File Copy 1024 bufsize 2000 maxblocks 3960.0 346481.0 875.0
File Copy 256 bufsize 500 maxblocks 1655.0 99051.0 598.5
File Copy 4096 bufsize 8000 maxblocks 5800.0 854789.0 1473.8
Shell Scripts (8 concurrent) 6.0 1097.7 1829.5
=========
FINAL SCORE 991.3
== 2.6.26-rc2-mm1 + this set ==
Execl Throughput 3012.9 lps (29.9 secs, 3 samples)
C Compiler Throughput 981.0 lpm (60.0 secs, 3 samples)
Shell Scripts (1 concurrent) 5872.0 lpm (60.0 secs, 3 samples)
Shell Scripts (8 concurrent) 1120.3 lpm (60.0 secs, 3 samples)
Shell Scripts (16 concurrent) 578.0 lpm (60.0 secs, 3 samples)
File Read 1024 bufsize 2000 maxblocks 1003993.0 KBps (30.0 secs, 3 samples)
File Write 1024 bufsize 2000 maxblocks 550452.0 KBps (30.0 secs, 3 samples)
File Copy 1024 bufsize 2000 maxblocks 347159.0 KBps (30.0 secs, 3 samples)
File Read 256 bufsize 500 maxblocks 314644.0 KBps (30.0 secs, 3 samples)
File Write 256 bufsize 500 maxblocks 151852.0 KBps (30.0 secs, 3 samples)
File Copy 256 bufsize 500 maxblocks 101000.0 KBps (30.0 secs, 3 samples)
File Read 4096 bufsize 8000 maxblocks 2033256.0 KBps (30.0 secs, 3 samples)
File Write 4096 bufsize 8000 maxblocks 1611814.0 KBps (30.0 secs, 3 samples)
File Copy 4096 bufsize 8000 maxblocks 847979.0 KBps (30.0 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places 128148.7 lpm (30.0 secs, 3 samples)
INDEX VALUES
TEST BASELINE RESULT INDEX
Execl Throughput 43.0 3012.9 700.7
File Copy 1024 bufsize 2000 maxblocks 3960.0 347159.0 876.7
File Copy 256 bufsize 500 maxblocks 1655.0 101000.0 610.3
File Copy 4096 bufsize 8000 maxblocks 5800.0 847979.0 1462.0
Shell Scripts (8 concurrent) 6.0 1120.3 1867.2
=========
FINAL SCORE 1004.6
This patch:
Remove refcnt from page_cgroup().
After this,
* A page is charged only when !page_mapped() && no page_cgroup is assigned.
* Anon page is newly mapped.
* File page is added to mapping->tree.
* A page is uncharged only when
* Anon page is fully unmapped.
* File page is removed from LRU.
There is no change in behavior from user's view.
This patch also removes unnecessary calls in rmap.c which was used only for
refcnt mangement.
[akpm@linux-foundation.org: fix warning]
[hugh@veritas.com: fix shmem_unuse_inode charging]
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-25 10:47:14 +02:00
|
|
|
__mem_cgroup_uncharge_common(struct page *page, enum charge_type ctype)
|
2008-02-07 09:13:53 +01:00
|
|
|
{
|
2008-03-04 23:29:08 +01:00
|
|
|
struct page_cgroup *pc;
|
2009-01-08 03:08:00 +01:00
|
|
|
struct mem_cgroup *mem = NULL;
|
2008-02-07 09:14:39 +01:00
|
|
|
struct mem_cgroup_per_zone *mz;
|
2008-02-07 09:13:53 +01:00
|
|
|
|
2009-01-08 03:08:02 +01:00
|
|
|
if (mem_cgroup_disabled())
|
2009-01-08 03:08:00 +01:00
|
|
|
return NULL;
|
2008-04-04 23:29:59 +02:00
|
|
|
|
2009-01-08 03:07:56 +01:00
|
|
|
if (PageSwapCache(page))
|
2009-01-08 03:08:00 +01:00
|
|
|
return NULL;
|
2009-01-08 03:07:56 +01:00
|
|
|
|
2008-02-07 09:13:59 +01:00
|
|
|
/*
|
2008-02-07 09:14:41 +01:00
|
|
|
* Check if our page_cgroup is valid
|
2008-02-07 09:13:59 +01:00
|
|
|
*/
|
2008-10-19 05:28:16 +02:00
|
|
|
pc = lookup_page_cgroup(page);
|
|
|
|
if (unlikely(!pc || !PageCgroupUsed(pc)))
|
2009-01-08 03:08:00 +01:00
|
|
|
return NULL;
|
2008-03-04 23:29:11 +01:00
|
|
|
|
2008-10-19 05:28:16 +02:00
|
|
|
lock_page_cgroup(pc);
|
2009-01-08 03:07:56 +01:00
|
|
|
|
2009-01-08 03:08:00 +01:00
|
|
|
mem = pc->mem_cgroup;
|
|
|
|
|
2009-01-08 03:07:56 +01:00
|
|
|
if (!PageCgroupUsed(pc))
|
|
|
|
goto unlock_out;
|
|
|
|
|
|
|
|
switch (ctype) {
|
|
|
|
case MEM_CGROUP_CHARGE_TYPE_MAPPED:
|
|
|
|
if (page_mapped(page))
|
|
|
|
goto unlock_out;
|
|
|
|
break;
|
|
|
|
case MEM_CGROUP_CHARGE_TYPE_SWAPOUT:
|
|
|
|
if (!PageAnon(page)) { /* Shared memory */
|
|
|
|
if (page->mapping && !page_is_file_cache(page))
|
|
|
|
goto unlock_out;
|
|
|
|
} else if (page_mapped(page)) /* Anon */
|
|
|
|
goto unlock_out;
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
break;
|
2008-10-19 05:28:16 +02:00
|
|
|
}
|
2009-01-08 03:07:56 +01:00
|
|
|
|
2009-01-08 03:08:00 +01:00
|
|
|
res_counter_uncharge(&mem->res, PAGE_SIZE);
|
|
|
|
if (do_swap_account && (ctype != MEM_CGROUP_CHARGE_TYPE_SWAPOUT))
|
|
|
|
res_counter_uncharge(&mem->memsw, PAGE_SIZE);
|
|
|
|
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
mem_cgroup_charge_statistics(mem, pc, false);
|
2008-10-19 05:28:16 +02:00
|
|
|
ClearPageCgroupUsed(pc);
|
2008-03-04 23:29:11 +01:00
|
|
|
|
memcg: remove refcnt from page_cgroup
memcg: performance improvements
Patch Description
1/5 ... remove refcnt fron page_cgroup patch (shmem handling is fixed)
2/5 ... swapcache handling patch
3/5 ... add helper function for shmem's memory reclaim patch
4/5 ... optimize by likely/unlikely ppatch
5/5 ... remove redundunt check patch (shmem handling is fixed.)
Unix bench result.
== 2.6.26-rc2-mm1 + memory resource controller
Execl Throughput 2915.4 lps (29.6 secs, 3 samples)
C Compiler Throughput 1019.3 lpm (60.0 secs, 3 samples)
Shell Scripts (1 concurrent) 5796.0 lpm (60.0 secs, 3 samples)
Shell Scripts (8 concurrent) 1097.7 lpm (60.0 secs, 3 samples)
Shell Scripts (16 concurrent) 565.3 lpm (60.0 secs, 3 samples)
File Read 1024 bufsize 2000 maxblocks 1022128.0 KBps (30.0 secs, 3 samples)
File Write 1024 bufsize 2000 maxblocks 544057.0 KBps (30.0 secs, 3 samples)
File Copy 1024 bufsize 2000 maxblocks 346481.0 KBps (30.0 secs, 3 samples)
File Read 256 bufsize 500 maxblocks 319325.0 KBps (30.0 secs, 3 samples)
File Write 256 bufsize 500 maxblocks 148788.0 KBps (30.0 secs, 3 samples)
File Copy 256 bufsize 500 maxblocks 99051.0 KBps (30.0 secs, 3 samples)
File Read 4096 bufsize 8000 maxblocks 2058917.0 KBps (30.0 secs, 3 samples)
File Write 4096 bufsize 8000 maxblocks 1606109.0 KBps (30.0 secs, 3 samples)
File Copy 4096 bufsize 8000 maxblocks 854789.0 KBps (30.0 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places 126145.2 lpm (30.0 secs, 3 samples)
INDEX VALUES
TEST BASELINE RESULT INDEX
Execl Throughput 43.0 2915.4 678.0
File Copy 1024 bufsize 2000 maxblocks 3960.0 346481.0 875.0
File Copy 256 bufsize 500 maxblocks 1655.0 99051.0 598.5
File Copy 4096 bufsize 8000 maxblocks 5800.0 854789.0 1473.8
Shell Scripts (8 concurrent) 6.0 1097.7 1829.5
=========
FINAL SCORE 991.3
== 2.6.26-rc2-mm1 + this set ==
Execl Throughput 3012.9 lps (29.9 secs, 3 samples)
C Compiler Throughput 981.0 lpm (60.0 secs, 3 samples)
Shell Scripts (1 concurrent) 5872.0 lpm (60.0 secs, 3 samples)
Shell Scripts (8 concurrent) 1120.3 lpm (60.0 secs, 3 samples)
Shell Scripts (16 concurrent) 578.0 lpm (60.0 secs, 3 samples)
File Read 1024 bufsize 2000 maxblocks 1003993.0 KBps (30.0 secs, 3 samples)
File Write 1024 bufsize 2000 maxblocks 550452.0 KBps (30.0 secs, 3 samples)
File Copy 1024 bufsize 2000 maxblocks 347159.0 KBps (30.0 secs, 3 samples)
File Read 256 bufsize 500 maxblocks 314644.0 KBps (30.0 secs, 3 samples)
File Write 256 bufsize 500 maxblocks 151852.0 KBps (30.0 secs, 3 samples)
File Copy 256 bufsize 500 maxblocks 101000.0 KBps (30.0 secs, 3 samples)
File Read 4096 bufsize 8000 maxblocks 2033256.0 KBps (30.0 secs, 3 samples)
File Write 4096 bufsize 8000 maxblocks 1611814.0 KBps (30.0 secs, 3 samples)
File Copy 4096 bufsize 8000 maxblocks 847979.0 KBps (30.0 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places 128148.7 lpm (30.0 secs, 3 samples)
INDEX VALUES
TEST BASELINE RESULT INDEX
Execl Throughput 43.0 3012.9 700.7
File Copy 1024 bufsize 2000 maxblocks 3960.0 347159.0 876.7
File Copy 256 bufsize 500 maxblocks 1655.0 101000.0 610.3
File Copy 4096 bufsize 8000 maxblocks 5800.0 847979.0 1462.0
Shell Scripts (8 concurrent) 6.0 1120.3 1867.2
=========
FINAL SCORE 1004.6
This patch:
Remove refcnt from page_cgroup().
After this,
* A page is charged only when !page_mapped() && no page_cgroup is assigned.
* Anon page is newly mapped.
* File page is added to mapping->tree.
* A page is uncharged only when
* Anon page is fully unmapped.
* File page is removed from LRU.
There is no change in behavior from user's view.
This patch also removes unnecessary calls in rmap.c which was used only for
refcnt mangement.
[akpm@linux-foundation.org: fix warning]
[hugh@veritas.com: fix shmem_unuse_inode charging]
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-25 10:47:14 +02:00
|
|
|
mz = page_cgroup_zoneinfo(pc);
|
2008-10-19 05:28:16 +02:00
|
|
|
unlock_page_cgroup(pc);
|
memcg: fix oops on NULL lru list
While testing force_empty, during an exit_mmap, __mem_cgroup_remove_list
called from mem_cgroup_uncharge_page oopsed on a NULL pointer in the lru list.
I couldn't see what racing tasks on other cpus were doing, but surmise that
another must have been in mem_cgroup_charge_common on the same page, between
its unlock_page_cgroup and spin_lock_irqsave near done (thanks to that kzalloc
which I'd almost changed to a kmalloc).
Normally such a race cannot happen, the ref_cnt prevents it, the final
uncharge cannot race with the initial charge. But force_empty buggers the
ref_cnt, that's what it's all about; and thereafter forced pages are
vulnerable to races such as this (just think of a shared page also mapped into
an mm of another mem_cgroup than that just emptied). And remain vulnerable
until they're freed indefinitely later.
This patch just fixes the oops by moving the unlock_page_cgroups down below
adding to and removing from the list (only possible given the previous patch);
and while we're at it, we might as well make it an invariant that
page->page_cgroup is always set while pc is on lru.
But this behaviour of force_empty seems highly unsatisfactory to me: why have
a ref_cnt if we always have to cope with it being violated (as in the earlier
page migration patch). We may prefer force_empty to move pages to an orphan
mem_cgroup (could be the root, but better not), from which other cgroups could
recover them; we might need to reverse the locking again; but no time now for
such concerns.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-03-04 23:29:16 +01:00
|
|
|
|
2009-01-08 03:08:13 +01:00
|
|
|
/* at swapout, this memcg will be accessed to record to swap */
|
|
|
|
if (ctype != MEM_CGROUP_CHARGE_TYPE_SWAPOUT)
|
|
|
|
css_put(&mem->css);
|
2008-02-07 09:14:31 +01:00
|
|
|
|
2009-01-08 03:08:00 +01:00
|
|
|
return mem;
|
2009-01-08 03:07:56 +01:00
|
|
|
|
|
|
|
unlock_out:
|
|
|
|
unlock_page_cgroup(pc);
|
2009-01-08 03:08:00 +01:00
|
|
|
return NULL;
|
2008-02-07 09:14:41 +01:00
|
|
|
}
|
|
|
|
|
memcg: remove refcnt from page_cgroup
memcg: performance improvements
Patch Description
1/5 ... remove refcnt fron page_cgroup patch (shmem handling is fixed)
2/5 ... swapcache handling patch
3/5 ... add helper function for shmem's memory reclaim patch
4/5 ... optimize by likely/unlikely ppatch
5/5 ... remove redundunt check patch (shmem handling is fixed.)
Unix bench result.
== 2.6.26-rc2-mm1 + memory resource controller
Execl Throughput 2915.4 lps (29.6 secs, 3 samples)
C Compiler Throughput 1019.3 lpm (60.0 secs, 3 samples)
Shell Scripts (1 concurrent) 5796.0 lpm (60.0 secs, 3 samples)
Shell Scripts (8 concurrent) 1097.7 lpm (60.0 secs, 3 samples)
Shell Scripts (16 concurrent) 565.3 lpm (60.0 secs, 3 samples)
File Read 1024 bufsize 2000 maxblocks 1022128.0 KBps (30.0 secs, 3 samples)
File Write 1024 bufsize 2000 maxblocks 544057.0 KBps (30.0 secs, 3 samples)
File Copy 1024 bufsize 2000 maxblocks 346481.0 KBps (30.0 secs, 3 samples)
File Read 256 bufsize 500 maxblocks 319325.0 KBps (30.0 secs, 3 samples)
File Write 256 bufsize 500 maxblocks 148788.0 KBps (30.0 secs, 3 samples)
File Copy 256 bufsize 500 maxblocks 99051.0 KBps (30.0 secs, 3 samples)
File Read 4096 bufsize 8000 maxblocks 2058917.0 KBps (30.0 secs, 3 samples)
File Write 4096 bufsize 8000 maxblocks 1606109.0 KBps (30.0 secs, 3 samples)
File Copy 4096 bufsize 8000 maxblocks 854789.0 KBps (30.0 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places 126145.2 lpm (30.0 secs, 3 samples)
INDEX VALUES
TEST BASELINE RESULT INDEX
Execl Throughput 43.0 2915.4 678.0
File Copy 1024 bufsize 2000 maxblocks 3960.0 346481.0 875.0
File Copy 256 bufsize 500 maxblocks 1655.0 99051.0 598.5
File Copy 4096 bufsize 8000 maxblocks 5800.0 854789.0 1473.8
Shell Scripts (8 concurrent) 6.0 1097.7 1829.5
=========
FINAL SCORE 991.3
== 2.6.26-rc2-mm1 + this set ==
Execl Throughput 3012.9 lps (29.9 secs, 3 samples)
C Compiler Throughput 981.0 lpm (60.0 secs, 3 samples)
Shell Scripts (1 concurrent) 5872.0 lpm (60.0 secs, 3 samples)
Shell Scripts (8 concurrent) 1120.3 lpm (60.0 secs, 3 samples)
Shell Scripts (16 concurrent) 578.0 lpm (60.0 secs, 3 samples)
File Read 1024 bufsize 2000 maxblocks 1003993.0 KBps (30.0 secs, 3 samples)
File Write 1024 bufsize 2000 maxblocks 550452.0 KBps (30.0 secs, 3 samples)
File Copy 1024 bufsize 2000 maxblocks 347159.0 KBps (30.0 secs, 3 samples)
File Read 256 bufsize 500 maxblocks 314644.0 KBps (30.0 secs, 3 samples)
File Write 256 bufsize 500 maxblocks 151852.0 KBps (30.0 secs, 3 samples)
File Copy 256 bufsize 500 maxblocks 101000.0 KBps (30.0 secs, 3 samples)
File Read 4096 bufsize 8000 maxblocks 2033256.0 KBps (30.0 secs, 3 samples)
File Write 4096 bufsize 8000 maxblocks 1611814.0 KBps (30.0 secs, 3 samples)
File Copy 4096 bufsize 8000 maxblocks 847979.0 KBps (30.0 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places 128148.7 lpm (30.0 secs, 3 samples)
INDEX VALUES
TEST BASELINE RESULT INDEX
Execl Throughput 43.0 3012.9 700.7
File Copy 1024 bufsize 2000 maxblocks 3960.0 347159.0 876.7
File Copy 256 bufsize 500 maxblocks 1655.0 101000.0 610.3
File Copy 4096 bufsize 8000 maxblocks 5800.0 847979.0 1462.0
Shell Scripts (8 concurrent) 6.0 1120.3 1867.2
=========
FINAL SCORE 1004.6
This patch:
Remove refcnt from page_cgroup().
After this,
* A page is charged only when !page_mapped() && no page_cgroup is assigned.
* Anon page is newly mapped.
* File page is added to mapping->tree.
* A page is uncharged only when
* Anon page is fully unmapped.
* File page is removed from LRU.
There is no change in behavior from user's view.
This patch also removes unnecessary calls in rmap.c which was used only for
refcnt mangement.
[akpm@linux-foundation.org: fix warning]
[hugh@veritas.com: fix shmem_unuse_inode charging]
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-25 10:47:14 +02:00
|
|
|
void mem_cgroup_uncharge_page(struct page *page)
|
|
|
|
{
|
2008-10-19 05:28:16 +02:00
|
|
|
/* early check. */
|
|
|
|
if (page_mapped(page))
|
|
|
|
return;
|
|
|
|
if (page->mapping && !PageAnon(page))
|
|
|
|
return;
|
memcg: remove refcnt from page_cgroup
memcg: performance improvements
Patch Description
1/5 ... remove refcnt fron page_cgroup patch (shmem handling is fixed)
2/5 ... swapcache handling patch
3/5 ... add helper function for shmem's memory reclaim patch
4/5 ... optimize by likely/unlikely ppatch
5/5 ... remove redundunt check patch (shmem handling is fixed.)
Unix bench result.
== 2.6.26-rc2-mm1 + memory resource controller
Execl Throughput 2915.4 lps (29.6 secs, 3 samples)
C Compiler Throughput 1019.3 lpm (60.0 secs, 3 samples)
Shell Scripts (1 concurrent) 5796.0 lpm (60.0 secs, 3 samples)
Shell Scripts (8 concurrent) 1097.7 lpm (60.0 secs, 3 samples)
Shell Scripts (16 concurrent) 565.3 lpm (60.0 secs, 3 samples)
File Read 1024 bufsize 2000 maxblocks 1022128.0 KBps (30.0 secs, 3 samples)
File Write 1024 bufsize 2000 maxblocks 544057.0 KBps (30.0 secs, 3 samples)
File Copy 1024 bufsize 2000 maxblocks 346481.0 KBps (30.0 secs, 3 samples)
File Read 256 bufsize 500 maxblocks 319325.0 KBps (30.0 secs, 3 samples)
File Write 256 bufsize 500 maxblocks 148788.0 KBps (30.0 secs, 3 samples)
File Copy 256 bufsize 500 maxblocks 99051.0 KBps (30.0 secs, 3 samples)
File Read 4096 bufsize 8000 maxblocks 2058917.0 KBps (30.0 secs, 3 samples)
File Write 4096 bufsize 8000 maxblocks 1606109.0 KBps (30.0 secs, 3 samples)
File Copy 4096 bufsize 8000 maxblocks 854789.0 KBps (30.0 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places 126145.2 lpm (30.0 secs, 3 samples)
INDEX VALUES
TEST BASELINE RESULT INDEX
Execl Throughput 43.0 2915.4 678.0
File Copy 1024 bufsize 2000 maxblocks 3960.0 346481.0 875.0
File Copy 256 bufsize 500 maxblocks 1655.0 99051.0 598.5
File Copy 4096 bufsize 8000 maxblocks 5800.0 854789.0 1473.8
Shell Scripts (8 concurrent) 6.0 1097.7 1829.5
=========
FINAL SCORE 991.3
== 2.6.26-rc2-mm1 + this set ==
Execl Throughput 3012.9 lps (29.9 secs, 3 samples)
C Compiler Throughput 981.0 lpm (60.0 secs, 3 samples)
Shell Scripts (1 concurrent) 5872.0 lpm (60.0 secs, 3 samples)
Shell Scripts (8 concurrent) 1120.3 lpm (60.0 secs, 3 samples)
Shell Scripts (16 concurrent) 578.0 lpm (60.0 secs, 3 samples)
File Read 1024 bufsize 2000 maxblocks 1003993.0 KBps (30.0 secs, 3 samples)
File Write 1024 bufsize 2000 maxblocks 550452.0 KBps (30.0 secs, 3 samples)
File Copy 1024 bufsize 2000 maxblocks 347159.0 KBps (30.0 secs, 3 samples)
File Read 256 bufsize 500 maxblocks 314644.0 KBps (30.0 secs, 3 samples)
File Write 256 bufsize 500 maxblocks 151852.0 KBps (30.0 secs, 3 samples)
File Copy 256 bufsize 500 maxblocks 101000.0 KBps (30.0 secs, 3 samples)
File Read 4096 bufsize 8000 maxblocks 2033256.0 KBps (30.0 secs, 3 samples)
File Write 4096 bufsize 8000 maxblocks 1611814.0 KBps (30.0 secs, 3 samples)
File Copy 4096 bufsize 8000 maxblocks 847979.0 KBps (30.0 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places 128148.7 lpm (30.0 secs, 3 samples)
INDEX VALUES
TEST BASELINE RESULT INDEX
Execl Throughput 43.0 3012.9 700.7
File Copy 1024 bufsize 2000 maxblocks 3960.0 347159.0 876.7
File Copy 256 bufsize 500 maxblocks 1655.0 101000.0 610.3
File Copy 4096 bufsize 8000 maxblocks 5800.0 847979.0 1462.0
Shell Scripts (8 concurrent) 6.0 1120.3 1867.2
=========
FINAL SCORE 1004.6
This patch:
Remove refcnt from page_cgroup().
After this,
* A page is charged only when !page_mapped() && no page_cgroup is assigned.
* Anon page is newly mapped.
* File page is added to mapping->tree.
* A page is uncharged only when
* Anon page is fully unmapped.
* File page is removed from LRU.
There is no change in behavior from user's view.
This patch also removes unnecessary calls in rmap.c which was used only for
refcnt mangement.
[akpm@linux-foundation.org: fix warning]
[hugh@veritas.com: fix shmem_unuse_inode charging]
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-25 10:47:14 +02:00
|
|
|
__mem_cgroup_uncharge_common(page, MEM_CGROUP_CHARGE_TYPE_MAPPED);
|
|
|
|
}
|
|
|
|
|
|
|
|
void mem_cgroup_uncharge_cache_page(struct page *page)
|
|
|
|
{
|
|
|
|
VM_BUG_ON(page_mapped(page));
|
2008-10-19 05:28:09 +02:00
|
|
|
VM_BUG_ON(page->mapping);
|
memcg: remove refcnt from page_cgroup
memcg: performance improvements
Patch Description
1/5 ... remove refcnt fron page_cgroup patch (shmem handling is fixed)
2/5 ... swapcache handling patch
3/5 ... add helper function for shmem's memory reclaim patch
4/5 ... optimize by likely/unlikely ppatch
5/5 ... remove redundunt check patch (shmem handling is fixed.)
Unix bench result.
== 2.6.26-rc2-mm1 + memory resource controller
Execl Throughput 2915.4 lps (29.6 secs, 3 samples)
C Compiler Throughput 1019.3 lpm (60.0 secs, 3 samples)
Shell Scripts (1 concurrent) 5796.0 lpm (60.0 secs, 3 samples)
Shell Scripts (8 concurrent) 1097.7 lpm (60.0 secs, 3 samples)
Shell Scripts (16 concurrent) 565.3 lpm (60.0 secs, 3 samples)
File Read 1024 bufsize 2000 maxblocks 1022128.0 KBps (30.0 secs, 3 samples)
File Write 1024 bufsize 2000 maxblocks 544057.0 KBps (30.0 secs, 3 samples)
File Copy 1024 bufsize 2000 maxblocks 346481.0 KBps (30.0 secs, 3 samples)
File Read 256 bufsize 500 maxblocks 319325.0 KBps (30.0 secs, 3 samples)
File Write 256 bufsize 500 maxblocks 148788.0 KBps (30.0 secs, 3 samples)
File Copy 256 bufsize 500 maxblocks 99051.0 KBps (30.0 secs, 3 samples)
File Read 4096 bufsize 8000 maxblocks 2058917.0 KBps (30.0 secs, 3 samples)
File Write 4096 bufsize 8000 maxblocks 1606109.0 KBps (30.0 secs, 3 samples)
File Copy 4096 bufsize 8000 maxblocks 854789.0 KBps (30.0 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places 126145.2 lpm (30.0 secs, 3 samples)
INDEX VALUES
TEST BASELINE RESULT INDEX
Execl Throughput 43.0 2915.4 678.0
File Copy 1024 bufsize 2000 maxblocks 3960.0 346481.0 875.0
File Copy 256 bufsize 500 maxblocks 1655.0 99051.0 598.5
File Copy 4096 bufsize 8000 maxblocks 5800.0 854789.0 1473.8
Shell Scripts (8 concurrent) 6.0 1097.7 1829.5
=========
FINAL SCORE 991.3
== 2.6.26-rc2-mm1 + this set ==
Execl Throughput 3012.9 lps (29.9 secs, 3 samples)
C Compiler Throughput 981.0 lpm (60.0 secs, 3 samples)
Shell Scripts (1 concurrent) 5872.0 lpm (60.0 secs, 3 samples)
Shell Scripts (8 concurrent) 1120.3 lpm (60.0 secs, 3 samples)
Shell Scripts (16 concurrent) 578.0 lpm (60.0 secs, 3 samples)
File Read 1024 bufsize 2000 maxblocks 1003993.0 KBps (30.0 secs, 3 samples)
File Write 1024 bufsize 2000 maxblocks 550452.0 KBps (30.0 secs, 3 samples)
File Copy 1024 bufsize 2000 maxblocks 347159.0 KBps (30.0 secs, 3 samples)
File Read 256 bufsize 500 maxblocks 314644.0 KBps (30.0 secs, 3 samples)
File Write 256 bufsize 500 maxblocks 151852.0 KBps (30.0 secs, 3 samples)
File Copy 256 bufsize 500 maxblocks 101000.0 KBps (30.0 secs, 3 samples)
File Read 4096 bufsize 8000 maxblocks 2033256.0 KBps (30.0 secs, 3 samples)
File Write 4096 bufsize 8000 maxblocks 1611814.0 KBps (30.0 secs, 3 samples)
File Copy 4096 bufsize 8000 maxblocks 847979.0 KBps (30.0 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places 128148.7 lpm (30.0 secs, 3 samples)
INDEX VALUES
TEST BASELINE RESULT INDEX
Execl Throughput 43.0 3012.9 700.7
File Copy 1024 bufsize 2000 maxblocks 3960.0 347159.0 876.7
File Copy 256 bufsize 500 maxblocks 1655.0 101000.0 610.3
File Copy 4096 bufsize 8000 maxblocks 5800.0 847979.0 1462.0
Shell Scripts (8 concurrent) 6.0 1120.3 1867.2
=========
FINAL SCORE 1004.6
This patch:
Remove refcnt from page_cgroup().
After this,
* A page is charged only when !page_mapped() && no page_cgroup is assigned.
* Anon page is newly mapped.
* File page is added to mapping->tree.
* A page is uncharged only when
* Anon page is fully unmapped.
* File page is removed from LRU.
There is no change in behavior from user's view.
This patch also removes unnecessary calls in rmap.c which was used only for
refcnt mangement.
[akpm@linux-foundation.org: fix warning]
[hugh@veritas.com: fix shmem_unuse_inode charging]
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-25 10:47:14 +02:00
|
|
|
__mem_cgroup_uncharge_common(page, MEM_CGROUP_CHARGE_TYPE_CACHE);
|
|
|
|
}
|
|
|
|
|
2009-01-08 03:08:00 +01:00
|
|
|
/*
|
|
|
|
* called from __delete_from_swap_cache() and drop "page" account.
|
|
|
|
* memcg information is recorded to swap_cgroup of "ent"
|
|
|
|
*/
|
|
|
|
void mem_cgroup_uncharge_swapcache(struct page *page, swp_entry_t ent)
|
|
|
|
{
|
|
|
|
struct mem_cgroup *memcg;
|
|
|
|
|
|
|
|
memcg = __mem_cgroup_uncharge_common(page,
|
|
|
|
MEM_CGROUP_CHARGE_TYPE_SWAPOUT);
|
|
|
|
/* record memcg information */
|
|
|
|
if (do_swap_account && memcg) {
|
|
|
|
swap_cgroup_record(ent, memcg);
|
|
|
|
mem_cgroup_get(memcg);
|
|
|
|
}
|
2009-01-08 03:08:13 +01:00
|
|
|
if (memcg)
|
|
|
|
css_put(&memcg->css);
|
2009-01-08 03:08:00 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
|
|
|
|
/*
|
|
|
|
* called from swap_entry_free(). remove record in swap_cgroup and
|
|
|
|
* uncharge "memsw" account.
|
|
|
|
*/
|
|
|
|
void mem_cgroup_uncharge_swap(swp_entry_t ent)
|
2009-01-08 03:07:56 +01:00
|
|
|
{
|
2009-01-08 03:08:00 +01:00
|
|
|
struct mem_cgroup *memcg;
|
|
|
|
|
|
|
|
if (!do_swap_account)
|
|
|
|
return;
|
|
|
|
|
|
|
|
memcg = swap_cgroup_record(ent, NULL);
|
|
|
|
if (memcg) {
|
|
|
|
res_counter_uncharge(&memcg->memsw, PAGE_SIZE);
|
|
|
|
mem_cgroup_put(memcg);
|
|
|
|
}
|
2009-01-08 03:07:56 +01:00
|
|
|
}
|
2009-01-08 03:08:00 +01:00
|
|
|
#endif
|
2009-01-08 03:07:56 +01:00
|
|
|
|
2008-02-07 09:14:10 +01:00
|
|
|
/*
|
2009-01-08 03:07:50 +01:00
|
|
|
* Before starting migration, account PAGE_SIZE to mem_cgroup that the old
|
|
|
|
* page belongs to.
|
2008-02-07 09:14:10 +01:00
|
|
|
*/
|
2009-01-08 03:07:50 +01:00
|
|
|
int mem_cgroup_prepare_migration(struct page *page, struct mem_cgroup **ptr)
|
2008-02-07 09:14:10 +01:00
|
|
|
{
|
|
|
|
struct page_cgroup *pc;
|
2008-07-25 10:47:10 +02:00
|
|
|
struct mem_cgroup *mem = NULL;
|
|
|
|
int ret = 0;
|
2008-03-04 23:29:09 +01:00
|
|
|
|
2009-01-08 03:08:02 +01:00
|
|
|
if (mem_cgroup_disabled())
|
2008-04-04 23:29:59 +02:00
|
|
|
return 0;
|
|
|
|
|
2008-10-19 05:28:16 +02:00
|
|
|
pc = lookup_page_cgroup(page);
|
|
|
|
lock_page_cgroup(pc);
|
|
|
|
if (PageCgroupUsed(pc)) {
|
2008-07-25 10:47:10 +02:00
|
|
|
mem = pc->mem_cgroup;
|
|
|
|
css_get(&mem->css);
|
|
|
|
}
|
2008-10-19 05:28:16 +02:00
|
|
|
unlock_page_cgroup(pc);
|
2009-01-08 03:07:50 +01:00
|
|
|
|
2008-07-25 10:47:10 +02:00
|
|
|
if (mem) {
|
2009-01-08 03:08:10 +01:00
|
|
|
ret = mem_cgroup_try_charge(NULL, GFP_KERNEL, &mem);
|
2008-07-25 10:47:10 +02:00
|
|
|
css_put(&mem->css);
|
|
|
|
}
|
2009-01-08 03:07:50 +01:00
|
|
|
*ptr = mem;
|
2008-07-25 10:47:10 +02:00
|
|
|
return ret;
|
2008-02-07 09:14:10 +01:00
|
|
|
}
|
2008-03-04 23:29:09 +01:00
|
|
|
|
memcg: remove refcnt from page_cgroup
memcg: performance improvements
Patch Description
1/5 ... remove refcnt fron page_cgroup patch (shmem handling is fixed)
2/5 ... swapcache handling patch
3/5 ... add helper function for shmem's memory reclaim patch
4/5 ... optimize by likely/unlikely ppatch
5/5 ... remove redundunt check patch (shmem handling is fixed.)
Unix bench result.
== 2.6.26-rc2-mm1 + memory resource controller
Execl Throughput 2915.4 lps (29.6 secs, 3 samples)
C Compiler Throughput 1019.3 lpm (60.0 secs, 3 samples)
Shell Scripts (1 concurrent) 5796.0 lpm (60.0 secs, 3 samples)
Shell Scripts (8 concurrent) 1097.7 lpm (60.0 secs, 3 samples)
Shell Scripts (16 concurrent) 565.3 lpm (60.0 secs, 3 samples)
File Read 1024 bufsize 2000 maxblocks 1022128.0 KBps (30.0 secs, 3 samples)
File Write 1024 bufsize 2000 maxblocks 544057.0 KBps (30.0 secs, 3 samples)
File Copy 1024 bufsize 2000 maxblocks 346481.0 KBps (30.0 secs, 3 samples)
File Read 256 bufsize 500 maxblocks 319325.0 KBps (30.0 secs, 3 samples)
File Write 256 bufsize 500 maxblocks 148788.0 KBps (30.0 secs, 3 samples)
File Copy 256 bufsize 500 maxblocks 99051.0 KBps (30.0 secs, 3 samples)
File Read 4096 bufsize 8000 maxblocks 2058917.0 KBps (30.0 secs, 3 samples)
File Write 4096 bufsize 8000 maxblocks 1606109.0 KBps (30.0 secs, 3 samples)
File Copy 4096 bufsize 8000 maxblocks 854789.0 KBps (30.0 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places 126145.2 lpm (30.0 secs, 3 samples)
INDEX VALUES
TEST BASELINE RESULT INDEX
Execl Throughput 43.0 2915.4 678.0
File Copy 1024 bufsize 2000 maxblocks 3960.0 346481.0 875.0
File Copy 256 bufsize 500 maxblocks 1655.0 99051.0 598.5
File Copy 4096 bufsize 8000 maxblocks 5800.0 854789.0 1473.8
Shell Scripts (8 concurrent) 6.0 1097.7 1829.5
=========
FINAL SCORE 991.3
== 2.6.26-rc2-mm1 + this set ==
Execl Throughput 3012.9 lps (29.9 secs, 3 samples)
C Compiler Throughput 981.0 lpm (60.0 secs, 3 samples)
Shell Scripts (1 concurrent) 5872.0 lpm (60.0 secs, 3 samples)
Shell Scripts (8 concurrent) 1120.3 lpm (60.0 secs, 3 samples)
Shell Scripts (16 concurrent) 578.0 lpm (60.0 secs, 3 samples)
File Read 1024 bufsize 2000 maxblocks 1003993.0 KBps (30.0 secs, 3 samples)
File Write 1024 bufsize 2000 maxblocks 550452.0 KBps (30.0 secs, 3 samples)
File Copy 1024 bufsize 2000 maxblocks 347159.0 KBps (30.0 secs, 3 samples)
File Read 256 bufsize 500 maxblocks 314644.0 KBps (30.0 secs, 3 samples)
File Write 256 bufsize 500 maxblocks 151852.0 KBps (30.0 secs, 3 samples)
File Copy 256 bufsize 500 maxblocks 101000.0 KBps (30.0 secs, 3 samples)
File Read 4096 bufsize 8000 maxblocks 2033256.0 KBps (30.0 secs, 3 samples)
File Write 4096 bufsize 8000 maxblocks 1611814.0 KBps (30.0 secs, 3 samples)
File Copy 4096 bufsize 8000 maxblocks 847979.0 KBps (30.0 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places 128148.7 lpm (30.0 secs, 3 samples)
INDEX VALUES
TEST BASELINE RESULT INDEX
Execl Throughput 43.0 3012.9 700.7
File Copy 1024 bufsize 2000 maxblocks 3960.0 347159.0 876.7
File Copy 256 bufsize 500 maxblocks 1655.0 101000.0 610.3
File Copy 4096 bufsize 8000 maxblocks 5800.0 847979.0 1462.0
Shell Scripts (8 concurrent) 6.0 1120.3 1867.2
=========
FINAL SCORE 1004.6
This patch:
Remove refcnt from page_cgroup().
After this,
* A page is charged only when !page_mapped() && no page_cgroup is assigned.
* Anon page is newly mapped.
* File page is added to mapping->tree.
* A page is uncharged only when
* Anon page is fully unmapped.
* File page is removed from LRU.
There is no change in behavior from user's view.
This patch also removes unnecessary calls in rmap.c which was used only for
refcnt mangement.
[akpm@linux-foundation.org: fix warning]
[hugh@veritas.com: fix shmem_unuse_inode charging]
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-25 10:47:14 +02:00
|
|
|
/* remove redundant charge if migration failed*/
|
2009-01-08 03:07:50 +01:00
|
|
|
void mem_cgroup_end_migration(struct mem_cgroup *mem,
|
|
|
|
struct page *oldpage, struct page *newpage)
|
2008-02-07 09:14:10 +01:00
|
|
|
{
|
2009-01-08 03:07:50 +01:00
|
|
|
struct page *target, *unused;
|
|
|
|
struct page_cgroup *pc;
|
|
|
|
enum charge_type ctype;
|
|
|
|
|
|
|
|
if (!mem)
|
|
|
|
return;
|
|
|
|
|
|
|
|
/* at migration success, oldpage->mapping is NULL. */
|
|
|
|
if (oldpage->mapping) {
|
|
|
|
target = oldpage;
|
|
|
|
unused = NULL;
|
|
|
|
} else {
|
|
|
|
target = newpage;
|
|
|
|
unused = oldpage;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (PageAnon(target))
|
|
|
|
ctype = MEM_CGROUP_CHARGE_TYPE_MAPPED;
|
|
|
|
else if (page_is_file_cache(target))
|
|
|
|
ctype = MEM_CGROUP_CHARGE_TYPE_CACHE;
|
|
|
|
else
|
|
|
|
ctype = MEM_CGROUP_CHARGE_TYPE_SHMEM;
|
|
|
|
|
|
|
|
/* unused page is not on radix-tree now. */
|
2009-01-08 03:07:56 +01:00
|
|
|
if (unused)
|
2009-01-08 03:07:50 +01:00
|
|
|
__mem_cgroup_uncharge_common(unused, ctype);
|
|
|
|
|
|
|
|
pc = lookup_page_cgroup(target);
|
memcg: remove refcnt from page_cgroup
memcg: performance improvements
Patch Description
1/5 ... remove refcnt fron page_cgroup patch (shmem handling is fixed)
2/5 ... swapcache handling patch
3/5 ... add helper function for shmem's memory reclaim patch
4/5 ... optimize by likely/unlikely ppatch
5/5 ... remove redundunt check patch (shmem handling is fixed.)
Unix bench result.
== 2.6.26-rc2-mm1 + memory resource controller
Execl Throughput 2915.4 lps (29.6 secs, 3 samples)
C Compiler Throughput 1019.3 lpm (60.0 secs, 3 samples)
Shell Scripts (1 concurrent) 5796.0 lpm (60.0 secs, 3 samples)
Shell Scripts (8 concurrent) 1097.7 lpm (60.0 secs, 3 samples)
Shell Scripts (16 concurrent) 565.3 lpm (60.0 secs, 3 samples)
File Read 1024 bufsize 2000 maxblocks 1022128.0 KBps (30.0 secs, 3 samples)
File Write 1024 bufsize 2000 maxblocks 544057.0 KBps (30.0 secs, 3 samples)
File Copy 1024 bufsize 2000 maxblocks 346481.0 KBps (30.0 secs, 3 samples)
File Read 256 bufsize 500 maxblocks 319325.0 KBps (30.0 secs, 3 samples)
File Write 256 bufsize 500 maxblocks 148788.0 KBps (30.0 secs, 3 samples)
File Copy 256 bufsize 500 maxblocks 99051.0 KBps (30.0 secs, 3 samples)
File Read 4096 bufsize 8000 maxblocks 2058917.0 KBps (30.0 secs, 3 samples)
File Write 4096 bufsize 8000 maxblocks 1606109.0 KBps (30.0 secs, 3 samples)
File Copy 4096 bufsize 8000 maxblocks 854789.0 KBps (30.0 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places 126145.2 lpm (30.0 secs, 3 samples)
INDEX VALUES
TEST BASELINE RESULT INDEX
Execl Throughput 43.0 2915.4 678.0
File Copy 1024 bufsize 2000 maxblocks 3960.0 346481.0 875.0
File Copy 256 bufsize 500 maxblocks 1655.0 99051.0 598.5
File Copy 4096 bufsize 8000 maxblocks 5800.0 854789.0 1473.8
Shell Scripts (8 concurrent) 6.0 1097.7 1829.5
=========
FINAL SCORE 991.3
== 2.6.26-rc2-mm1 + this set ==
Execl Throughput 3012.9 lps (29.9 secs, 3 samples)
C Compiler Throughput 981.0 lpm (60.0 secs, 3 samples)
Shell Scripts (1 concurrent) 5872.0 lpm (60.0 secs, 3 samples)
Shell Scripts (8 concurrent) 1120.3 lpm (60.0 secs, 3 samples)
Shell Scripts (16 concurrent) 578.0 lpm (60.0 secs, 3 samples)
File Read 1024 bufsize 2000 maxblocks 1003993.0 KBps (30.0 secs, 3 samples)
File Write 1024 bufsize 2000 maxblocks 550452.0 KBps (30.0 secs, 3 samples)
File Copy 1024 bufsize 2000 maxblocks 347159.0 KBps (30.0 secs, 3 samples)
File Read 256 bufsize 500 maxblocks 314644.0 KBps (30.0 secs, 3 samples)
File Write 256 bufsize 500 maxblocks 151852.0 KBps (30.0 secs, 3 samples)
File Copy 256 bufsize 500 maxblocks 101000.0 KBps (30.0 secs, 3 samples)
File Read 4096 bufsize 8000 maxblocks 2033256.0 KBps (30.0 secs, 3 samples)
File Write 4096 bufsize 8000 maxblocks 1611814.0 KBps (30.0 secs, 3 samples)
File Copy 4096 bufsize 8000 maxblocks 847979.0 KBps (30.0 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places 128148.7 lpm (30.0 secs, 3 samples)
INDEX VALUES
TEST BASELINE RESULT INDEX
Execl Throughput 43.0 3012.9 700.7
File Copy 1024 bufsize 2000 maxblocks 3960.0 347159.0 876.7
File Copy 256 bufsize 500 maxblocks 1655.0 101000.0 610.3
File Copy 4096 bufsize 8000 maxblocks 5800.0 847979.0 1462.0
Shell Scripts (8 concurrent) 6.0 1120.3 1867.2
=========
FINAL SCORE 1004.6
This patch:
Remove refcnt from page_cgroup().
After this,
* A page is charged only when !page_mapped() && no page_cgroup is assigned.
* Anon page is newly mapped.
* File page is added to mapping->tree.
* A page is uncharged only when
* Anon page is fully unmapped.
* File page is removed from LRU.
There is no change in behavior from user's view.
This patch also removes unnecessary calls in rmap.c which was used only for
refcnt mangement.
[akpm@linux-foundation.org: fix warning]
[hugh@veritas.com: fix shmem_unuse_inode charging]
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-25 10:47:14 +02:00
|
|
|
/*
|
2009-01-08 03:07:50 +01:00
|
|
|
* __mem_cgroup_commit_charge() check PCG_USED bit of page_cgroup.
|
|
|
|
* So, double-counting is effectively avoided.
|
|
|
|
*/
|
|
|
|
__mem_cgroup_commit_charge(mem, pc, ctype);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Both of oldpage and newpage are still under lock_page().
|
|
|
|
* Then, we don't have to care about race in radix-tree.
|
|
|
|
* But we have to be careful that this page is unmapped or not.
|
|
|
|
*
|
|
|
|
* There is a case for !page_mapped(). At the start of
|
|
|
|
* migration, oldpage was mapped. But now, it's zapped.
|
|
|
|
* But we know *target* page is not freed/reused under us.
|
|
|
|
* mem_cgroup_uncharge_page() does all necessary checks.
|
memcg: remove refcnt from page_cgroup
memcg: performance improvements
Patch Description
1/5 ... remove refcnt fron page_cgroup patch (shmem handling is fixed)
2/5 ... swapcache handling patch
3/5 ... add helper function for shmem's memory reclaim patch
4/5 ... optimize by likely/unlikely ppatch
5/5 ... remove redundunt check patch (shmem handling is fixed.)
Unix bench result.
== 2.6.26-rc2-mm1 + memory resource controller
Execl Throughput 2915.4 lps (29.6 secs, 3 samples)
C Compiler Throughput 1019.3 lpm (60.0 secs, 3 samples)
Shell Scripts (1 concurrent) 5796.0 lpm (60.0 secs, 3 samples)
Shell Scripts (8 concurrent) 1097.7 lpm (60.0 secs, 3 samples)
Shell Scripts (16 concurrent) 565.3 lpm (60.0 secs, 3 samples)
File Read 1024 bufsize 2000 maxblocks 1022128.0 KBps (30.0 secs, 3 samples)
File Write 1024 bufsize 2000 maxblocks 544057.0 KBps (30.0 secs, 3 samples)
File Copy 1024 bufsize 2000 maxblocks 346481.0 KBps (30.0 secs, 3 samples)
File Read 256 bufsize 500 maxblocks 319325.0 KBps (30.0 secs, 3 samples)
File Write 256 bufsize 500 maxblocks 148788.0 KBps (30.0 secs, 3 samples)
File Copy 256 bufsize 500 maxblocks 99051.0 KBps (30.0 secs, 3 samples)
File Read 4096 bufsize 8000 maxblocks 2058917.0 KBps (30.0 secs, 3 samples)
File Write 4096 bufsize 8000 maxblocks 1606109.0 KBps (30.0 secs, 3 samples)
File Copy 4096 bufsize 8000 maxblocks 854789.0 KBps (30.0 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places 126145.2 lpm (30.0 secs, 3 samples)
INDEX VALUES
TEST BASELINE RESULT INDEX
Execl Throughput 43.0 2915.4 678.0
File Copy 1024 bufsize 2000 maxblocks 3960.0 346481.0 875.0
File Copy 256 bufsize 500 maxblocks 1655.0 99051.0 598.5
File Copy 4096 bufsize 8000 maxblocks 5800.0 854789.0 1473.8
Shell Scripts (8 concurrent) 6.0 1097.7 1829.5
=========
FINAL SCORE 991.3
== 2.6.26-rc2-mm1 + this set ==
Execl Throughput 3012.9 lps (29.9 secs, 3 samples)
C Compiler Throughput 981.0 lpm (60.0 secs, 3 samples)
Shell Scripts (1 concurrent) 5872.0 lpm (60.0 secs, 3 samples)
Shell Scripts (8 concurrent) 1120.3 lpm (60.0 secs, 3 samples)
Shell Scripts (16 concurrent) 578.0 lpm (60.0 secs, 3 samples)
File Read 1024 bufsize 2000 maxblocks 1003993.0 KBps (30.0 secs, 3 samples)
File Write 1024 bufsize 2000 maxblocks 550452.0 KBps (30.0 secs, 3 samples)
File Copy 1024 bufsize 2000 maxblocks 347159.0 KBps (30.0 secs, 3 samples)
File Read 256 bufsize 500 maxblocks 314644.0 KBps (30.0 secs, 3 samples)
File Write 256 bufsize 500 maxblocks 151852.0 KBps (30.0 secs, 3 samples)
File Copy 256 bufsize 500 maxblocks 101000.0 KBps (30.0 secs, 3 samples)
File Read 4096 bufsize 8000 maxblocks 2033256.0 KBps (30.0 secs, 3 samples)
File Write 4096 bufsize 8000 maxblocks 1611814.0 KBps (30.0 secs, 3 samples)
File Copy 4096 bufsize 8000 maxblocks 847979.0 KBps (30.0 secs, 3 samples)
Dc: sqrt(2) to 99 decimal places 128148.7 lpm (30.0 secs, 3 samples)
INDEX VALUES
TEST BASELINE RESULT INDEX
Execl Throughput 43.0 3012.9 700.7
File Copy 1024 bufsize 2000 maxblocks 3960.0 347159.0 876.7
File Copy 256 bufsize 500 maxblocks 1655.0 101000.0 610.3
File Copy 4096 bufsize 8000 maxblocks 5800.0 847979.0 1462.0
Shell Scripts (8 concurrent) 6.0 1120.3 1867.2
=========
FINAL SCORE 1004.6
This patch:
Remove refcnt from page_cgroup().
After this,
* A page is charged only when !page_mapped() && no page_cgroup is assigned.
* Anon page is newly mapped.
* File page is added to mapping->tree.
* A page is uncharged only when
* Anon page is fully unmapped.
* File page is removed from LRU.
There is no change in behavior from user's view.
This patch also removes unnecessary calls in rmap.c which was used only for
refcnt mangement.
[akpm@linux-foundation.org: fix warning]
[hugh@veritas.com: fix shmem_unuse_inode charging]
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-25 10:47:14 +02:00
|
|
|
*/
|
2009-01-08 03:07:50 +01:00
|
|
|
if (ctype == MEM_CGROUP_CHARGE_TYPE_MAPPED)
|
|
|
|
mem_cgroup_uncharge_page(target);
|
2008-02-07 09:14:10 +01:00
|
|
|
}
|
2008-02-07 09:13:51 +01:00
|
|
|
|
2008-07-25 10:47:15 +02:00
|
|
|
/*
|
|
|
|
* A call to try to shrink memory usage under specified resource controller.
|
|
|
|
* This is typically used for page reclaiming for shmem for reducing side
|
|
|
|
* effect of page allocation from shmem, which is used by some mem_cgroup.
|
|
|
|
*/
|
|
|
|
int mem_cgroup_shrink_usage(struct mm_struct *mm, gfp_t gfp_mask)
|
|
|
|
{
|
|
|
|
struct mem_cgroup *mem;
|
|
|
|
int progress = 0;
|
|
|
|
int retry = MEM_CGROUP_RECLAIM_RETRIES;
|
|
|
|
|
2009-01-08 03:08:02 +01:00
|
|
|
if (mem_cgroup_disabled())
|
2008-07-25 10:47:18 +02:00
|
|
|
return 0;
|
2008-08-13 00:08:41 +02:00
|
|
|
if (!mm)
|
|
|
|
return 0;
|
2008-07-25 10:47:18 +02:00
|
|
|
|
2008-07-25 10:47:15 +02:00
|
|
|
rcu_read_lock();
|
|
|
|
mem = mem_cgroup_from_task(rcu_dereference(mm->owner));
|
mm owner: fix race between swapoff and exit
There's a race between mm->owner assignment and swapoff, more easily
seen when task slab poisoning is turned on. The condition occurs when
try_to_unuse() runs in parallel with an exiting task. A similar race
can occur with callers of get_task_mm(), such as /proc/<pid>/<mmstats>
or ptrace or page migration.
CPU0 CPU1
try_to_unuse
looks at mm = task0->mm
increments mm->mm_users
task 0 exits
mm->owner needs to be updated, but no
new owner is found (mm_users > 1, but
no other task has task->mm = task0->mm)
mm_update_next_owner() leaves
mmput(mm) decrements mm->mm_users
task0 freed
dereferencing mm->owner fails
The fix is to notify the subsystem via mm_owner_changed callback(),
if no new owner is found, by specifying the new task as NULL.
Jiri Slaby:
mm->owner was set to NULL prior to calling cgroup_mm_owner_callbacks(), but
must be set after that, so as not to pass NULL as old owner causing oops.
Daisuke Nishimura:
mm_update_next_owner() may set mm->owner to NULL, but mem_cgroup_from_task()
and its callers need to take account of this situation to avoid oops.
Hugh Dickins:
Lockdep warning and hang below exec_mmap() when testing these patches.
exit_mm() up_reads mmap_sem before calling mm_update_next_owner(),
so exec_mmap() now needs to do the same. And with that repositioning,
there's now no point in mm_need_new_owner() allowing for NULL mm.
Reported-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-09-29 00:09:31 +02:00
|
|
|
if (unlikely(!mem)) {
|
|
|
|
rcu_read_unlock();
|
|
|
|
return 0;
|
|
|
|
}
|
2008-07-25 10:47:15 +02:00
|
|
|
css_get(&mem->css);
|
|
|
|
rcu_read_unlock();
|
|
|
|
|
|
|
|
do {
|
2009-01-08 03:08:00 +01:00
|
|
|
progress = try_to_free_mem_cgroup_pages(mem, gfp_mask, true);
|
2009-01-08 03:08:12 +01:00
|
|
|
progress += mem_cgroup_check_under_limit(mem);
|
2008-07-25 10:47:15 +02:00
|
|
|
} while (!progress && --retry);
|
|
|
|
|
|
|
|
css_put(&mem->css);
|
|
|
|
if (!retry)
|
|
|
|
return -ENOMEM;
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2009-01-08 03:08:00 +01:00
|
|
|
static DEFINE_MUTEX(set_limit_mutex);
|
|
|
|
|
2009-01-06 23:39:44 +01:00
|
|
|
static int mem_cgroup_resize_limit(struct mem_cgroup *memcg,
|
2009-01-08 03:08:00 +01:00
|
|
|
unsigned long long val)
|
2008-07-25 10:47:20 +02:00
|
|
|
{
|
|
|
|
|
|
|
|
int retry_count = MEM_CGROUP_RECLAIM_RETRIES;
|
|
|
|
int progress;
|
2009-01-08 03:08:00 +01:00
|
|
|
u64 memswlimit;
|
2008-07-25 10:47:20 +02:00
|
|
|
int ret = 0;
|
|
|
|
|
2009-01-08 03:08:00 +01:00
|
|
|
while (retry_count) {
|
2008-07-25 10:47:20 +02:00
|
|
|
if (signal_pending(current)) {
|
|
|
|
ret = -EINTR;
|
|
|
|
break;
|
|
|
|
}
|
2009-01-08 03:08:00 +01:00
|
|
|
/*
|
|
|
|
* Rather than hide all in some function, I do this in
|
|
|
|
* open coded manner. You see what this really does.
|
|
|
|
* We have to guarantee mem->res.limit < mem->memsw.limit.
|
|
|
|
*/
|
|
|
|
mutex_lock(&set_limit_mutex);
|
|
|
|
memswlimit = res_counter_read_u64(&memcg->memsw, RES_LIMIT);
|
|
|
|
if (memswlimit < val) {
|
|
|
|
ret = -EINVAL;
|
|
|
|
mutex_unlock(&set_limit_mutex);
|
2008-07-25 10:47:20 +02:00
|
|
|
break;
|
|
|
|
}
|
2009-01-08 03:08:00 +01:00
|
|
|
ret = res_counter_set_limit(&memcg->res, val);
|
|
|
|
mutex_unlock(&set_limit_mutex);
|
|
|
|
|
|
|
|
if (!ret)
|
|
|
|
break;
|
|
|
|
|
2009-01-08 03:07:49 +01:00
|
|
|
progress = try_to_free_mem_cgroup_pages(memcg,
|
2009-01-08 03:08:10 +01:00
|
|
|
GFP_KERNEL, false);
|
2009-01-08 03:08:00 +01:00
|
|
|
if (!progress) retry_count--;
|
|
|
|
}
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
int mem_cgroup_resize_memsw_limit(struct mem_cgroup *memcg,
|
|
|
|
unsigned long long val)
|
|
|
|
{
|
|
|
|
int retry_count = MEM_CGROUP_RECLAIM_RETRIES;
|
|
|
|
u64 memlimit, oldusage, curusage;
|
|
|
|
int ret;
|
|
|
|
|
|
|
|
if (!do_swap_account)
|
|
|
|
return -EINVAL;
|
|
|
|
|
|
|
|
while (retry_count) {
|
|
|
|
if (signal_pending(current)) {
|
|
|
|
ret = -EINTR;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
/*
|
|
|
|
* Rather than hide all in some function, I do this in
|
|
|
|
* open coded manner. You see what this really does.
|
|
|
|
* We have to guarantee mem->res.limit < mem->memsw.limit.
|
|
|
|
*/
|
|
|
|
mutex_lock(&set_limit_mutex);
|
|
|
|
memlimit = res_counter_read_u64(&memcg->res, RES_LIMIT);
|
|
|
|
if (memlimit > val) {
|
|
|
|
ret = -EINVAL;
|
|
|
|
mutex_unlock(&set_limit_mutex);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
ret = res_counter_set_limit(&memcg->memsw, val);
|
|
|
|
mutex_unlock(&set_limit_mutex);
|
|
|
|
|
|
|
|
if (!ret)
|
|
|
|
break;
|
|
|
|
|
|
|
|
oldusage = res_counter_read_u64(&memcg->memsw, RES_USAGE);
|
2009-01-08 03:08:10 +01:00
|
|
|
try_to_free_mem_cgroup_pages(memcg, GFP_KERNEL, true);
|
2009-01-08 03:08:00 +01:00
|
|
|
curusage = res_counter_read_u64(&memcg->memsw, RES_USAGE);
|
|
|
|
if (curusage >= oldusage)
|
2008-07-25 10:47:20 +02:00
|
|
|
retry_count--;
|
|
|
|
}
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2008-02-07 09:14:16 +01:00
|
|
|
/*
|
|
|
|
* This routine traverse page_cgroup in given list and drop them all.
|
|
|
|
* *And* this routine doesn't reclaim page itself, just removes page_cgroup.
|
|
|
|
*/
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
static int mem_cgroup_force_empty_list(struct mem_cgroup *mem,
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
int node, int zid, enum lru_list lru)
|
2008-02-07 09:14:16 +01:00
|
|
|
{
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
struct zone *zone;
|
|
|
|
struct mem_cgroup_per_zone *mz;
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
struct page_cgroup *pc, *busy;
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
unsigned long flags, loop;
|
2008-02-07 09:14:39 +01:00
|
|
|
struct list_head *list;
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
int ret = 0;
|
2008-02-07 09:14:39 +01:00
|
|
|
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
zone = &NODE_DATA(node)->node_zones[zid];
|
|
|
|
mz = mem_cgroup_zoneinfo(mem, node, zid);
|
2008-10-19 05:26:14 +02:00
|
|
|
list = &mz->lists[lru];
|
2008-02-07 09:14:16 +01:00
|
|
|
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
loop = MEM_CGROUP_ZSTAT(mz, lru);
|
|
|
|
/* give some margin against EBUSY etc...*/
|
|
|
|
loop += 256;
|
|
|
|
busy = NULL;
|
|
|
|
while (loop--) {
|
|
|
|
ret = 0;
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
spin_lock_irqsave(&zone->lru_lock, flags);
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
if (list_empty(list)) {
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
spin_unlock_irqrestore(&zone->lru_lock, flags);
|
2008-10-19 05:28:16 +02:00
|
|
|
break;
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
}
|
|
|
|
pc = list_entry(list->prev, struct page_cgroup, lru);
|
|
|
|
if (busy == pc) {
|
|
|
|
list_move(&pc->lru, list);
|
|
|
|
busy = 0;
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
spin_unlock_irqrestore(&zone->lru_lock, flags);
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
continue;
|
|
|
|
}
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
spin_unlock_irqrestore(&zone->lru_lock, flags);
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
|
2009-01-08 03:08:10 +01:00
|
|
|
ret = mem_cgroup_move_parent(pc, mem, GFP_KERNEL);
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
if (ret == -ENOMEM)
|
2008-10-19 05:28:16 +02:00
|
|
|
break;
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
|
|
|
|
if (ret == -EBUSY || ret == -EINVAL) {
|
|
|
|
/* found lock contention or "pc" is obsolete. */
|
|
|
|
busy = pc;
|
|
|
|
cond_resched();
|
|
|
|
} else
|
|
|
|
busy = NULL;
|
2008-02-07 09:14:16 +01:00
|
|
|
}
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
if (!ret && !list_empty(list))
|
|
|
|
return -EBUSY;
|
|
|
|
return ret;
|
2008-02-07 09:14:16 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* make mem_cgroup's charge to be 0 if there is no task.
|
|
|
|
* This enables deleting this mem_cgroup.
|
|
|
|
*/
|
2009-01-08 03:07:55 +01:00
|
|
|
static int mem_cgroup_force_empty(struct mem_cgroup *mem, bool free_all)
|
2008-02-07 09:14:16 +01:00
|
|
|
{
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
int ret;
|
|
|
|
int node, zid, shrink;
|
|
|
|
int nr_retries = MEM_CGROUP_RECLAIM_RETRIES;
|
2009-01-08 03:07:55 +01:00
|
|
|
struct cgroup *cgrp = mem->css.cgroup;
|
2008-03-04 23:29:09 +01:00
|
|
|
|
2008-02-07 09:14:16 +01:00
|
|
|
css_get(&mem->css);
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
|
|
|
|
shrink = 0;
|
2009-01-08 03:07:55 +01:00
|
|
|
/* should free all ? */
|
|
|
|
if (free_all)
|
|
|
|
goto try_to_free;
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
move_account:
|
2008-02-07 09:14:38 +01:00
|
|
|
while (mem->res.usage > 0) {
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
ret = -EBUSY;
|
2009-01-08 03:07:55 +01:00
|
|
|
if (cgroup_task_count(cgrp) || !list_empty(&cgrp->children))
|
|
|
|
goto out;
|
|
|
|
ret = -EINTR;
|
|
|
|
if (signal_pending(current))
|
2008-02-07 09:14:16 +01:00
|
|
|
goto out;
|
2008-10-19 05:28:16 +02:00
|
|
|
/* This is for making all *used* pages to be on LRU. */
|
|
|
|
lru_add_drain_all();
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
ret = 0;
|
|
|
|
for_each_node_state(node, N_POSSIBLE) {
|
|
|
|
for (zid = 0; !ret && zid < MAX_NR_ZONES; zid++) {
|
2008-10-19 05:26:14 +02:00
|
|
|
enum lru_list l;
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
for_each_lru(l) {
|
|
|
|
ret = mem_cgroup_force_empty_list(mem,
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
node, zid, l);
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
if (ret)
|
|
|
|
break;
|
|
|
|
}
|
2008-02-07 09:14:38 +01:00
|
|
|
}
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
if (ret)
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
/* it seems parent cgroup doesn't have enough mem */
|
|
|
|
if (ret == -ENOMEM)
|
|
|
|
goto try_to_free;
|
2008-10-19 05:28:16 +02:00
|
|
|
cond_resched();
|
2008-02-07 09:14:16 +01:00
|
|
|
}
|
|
|
|
ret = 0;
|
|
|
|
out:
|
|
|
|
css_put(&mem->css);
|
|
|
|
return ret;
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
|
|
|
|
try_to_free:
|
2009-01-08 03:07:55 +01:00
|
|
|
/* returns EBUSY if there is a task or if we come here twice. */
|
|
|
|
if (cgroup_task_count(cgrp) || !list_empty(&cgrp->children) || shrink) {
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
ret = -EBUSY;
|
|
|
|
goto out;
|
|
|
|
}
|
2009-01-08 03:07:55 +01:00
|
|
|
/* we call try-to-free pages for make this cgroup empty */
|
|
|
|
lru_add_drain_all();
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
/* try to free all pages in this cgroup */
|
|
|
|
shrink = 1;
|
|
|
|
while (nr_retries && mem->res.usage > 0) {
|
|
|
|
int progress;
|
2009-01-08 03:07:55 +01:00
|
|
|
|
|
|
|
if (signal_pending(current)) {
|
|
|
|
ret = -EINTR;
|
|
|
|
goto out;
|
|
|
|
}
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
progress = try_to_free_mem_cgroup_pages(mem,
|
2009-01-08 03:08:10 +01:00
|
|
|
GFP_KERNEL, false);
|
2009-01-08 03:07:55 +01:00
|
|
|
if (!progress) {
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
nr_retries--;
|
2009-01-08 03:07:55 +01:00
|
|
|
/* maybe some writeback is necessary */
|
|
|
|
congestion_wait(WRITE, HZ/10);
|
|
|
|
}
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
|
|
|
|
}
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
lru_add_drain();
|
memcg: move all acccounting to parent at rmdir()
This patch provides a function to move account information of a page
between mem_cgroups and rewrite force_empty to make use of this.
This moving of page_cgroup is done under
- lru_lock of source/destination mem_cgroup is held.
- lock_page_cgroup() is held.
Then, a routine which touches pc->mem_cgroup without lock_page_cgroup()
should confirm pc->mem_cgroup is still valid or not. Typical code can be
following.
(while page is not under lock_page())
mem = pc->mem_cgroup;
mz = page_cgroup_zoneinfo(pc)
spin_lock_irqsave(&mz->lru_lock);
if (pc->mem_cgroup == mem)
...../* some list handling */
spin_unlock_irqrestore(&mz->lru_lock);
Of course, better way is
lock_page_cgroup(pc);
....
unlock_page_cgroup(pc);
But you should confirm the nest of lock and avoid deadlock.
If you treats page_cgroup from mem_cgroup's LRU under mz->lru_lock,
you don't have to worry about what pc->mem_cgroup points to.
moved pages are added to head of lru, not to tail.
Expected users of this routine is:
- force_empty (rmdir)
- moving tasks between cgroup (for moving account information.)
- hierarchy (maybe useful.)
force_empty(rmdir) uses this move_account and move pages to its parent.
This "move" will not cause OOM (I added "oom" parameter to try_charge().)
If the parent is busy (not enough memory), force_empty calls try_to_free_page()
and reduce usage.
Purpose of this behavior is
- Fix "forget all" behavior of force_empty and avoid leak of accounting.
- By "moving first, free if necessary", keep pages on memory as much as
possible.
Adding a switch to change behavior of force_empty to
- free first, move if necessary
- free all, if there is mlocked/busy pages, return -EBUSY.
is under consideration. (I'll add if someone requtests.)
This patch also removes memory.force_empty file, a brutal debug-only interface.
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Tested-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:07:53 +01:00
|
|
|
/* try move_account...there may be some *locked* pages. */
|
|
|
|
if (mem->res.usage)
|
|
|
|
goto move_account;
|
|
|
|
ret = 0;
|
|
|
|
goto out;
|
2008-02-07 09:14:16 +01:00
|
|
|
}
|
|
|
|
|
2009-01-08 03:07:55 +01:00
|
|
|
int mem_cgroup_force_empty_write(struct cgroup *cont, unsigned int event)
|
|
|
|
{
|
|
|
|
return mem_cgroup_force_empty(mem_cgroup_from_cont(cont), true);
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2009-01-08 03:08:07 +01:00
|
|
|
static u64 mem_cgroup_hierarchy_read(struct cgroup *cont, struct cftype *cft)
|
|
|
|
{
|
|
|
|
return mem_cgroup_from_cont(cont)->use_hierarchy;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int mem_cgroup_hierarchy_write(struct cgroup *cont, struct cftype *cft,
|
|
|
|
u64 val)
|
|
|
|
{
|
|
|
|
int retval = 0;
|
|
|
|
struct mem_cgroup *mem = mem_cgroup_from_cont(cont);
|
|
|
|
struct cgroup *parent = cont->parent;
|
|
|
|
struct mem_cgroup *parent_mem = NULL;
|
|
|
|
|
|
|
|
if (parent)
|
|
|
|
parent_mem = mem_cgroup_from_cont(parent);
|
|
|
|
|
|
|
|
cgroup_lock();
|
|
|
|
/*
|
|
|
|
* If parent's use_hiearchy is set, we can't make any modifications
|
|
|
|
* in the child subtrees. If it is unset, then the change can
|
|
|
|
* occur, provided the current cgroup has no children.
|
|
|
|
*
|
|
|
|
* For the root cgroup, parent_mem is NULL, we allow value to be
|
|
|
|
* set if there are no children.
|
|
|
|
*/
|
|
|
|
if ((!parent_mem || !parent_mem->use_hierarchy) &&
|
|
|
|
(val == 1 || val == 0)) {
|
|
|
|
if (list_empty(&cont->children))
|
|
|
|
mem->use_hierarchy = val;
|
|
|
|
else
|
|
|
|
retval = -EBUSY;
|
|
|
|
} else
|
|
|
|
retval = -EINVAL;
|
|
|
|
cgroup_unlock();
|
|
|
|
|
|
|
|
return retval;
|
|
|
|
}
|
|
|
|
|
2008-04-29 09:59:58 +02:00
|
|
|
static u64 mem_cgroup_read(struct cgroup *cont, struct cftype *cft)
|
2008-02-07 09:13:50 +01:00
|
|
|
{
|
2009-01-08 03:08:00 +01:00
|
|
|
struct mem_cgroup *mem = mem_cgroup_from_cont(cont);
|
|
|
|
u64 val = 0;
|
|
|
|
int type, name;
|
|
|
|
|
|
|
|
type = MEMFILE_TYPE(cft->private);
|
|
|
|
name = MEMFILE_ATTR(cft->private);
|
|
|
|
switch (type) {
|
|
|
|
case _MEM:
|
|
|
|
val = res_counter_read_u64(&mem->res, name);
|
|
|
|
break;
|
|
|
|
case _MEMSWAP:
|
|
|
|
if (do_swap_account)
|
|
|
|
val = res_counter_read_u64(&mem->memsw, name);
|
|
|
|
break;
|
|
|
|
default:
|
|
|
|
BUG();
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
return val;
|
2008-02-07 09:13:50 +01:00
|
|
|
}
|
2008-07-25 10:47:20 +02:00
|
|
|
/*
|
|
|
|
* The user of this function is...
|
|
|
|
* RES_LIMIT.
|
|
|
|
*/
|
2008-07-25 10:47:04 +02:00
|
|
|
static int mem_cgroup_write(struct cgroup *cont, struct cftype *cft,
|
|
|
|
const char *buffer)
|
2008-02-07 09:13:50 +01:00
|
|
|
{
|
2008-07-25 10:47:20 +02:00
|
|
|
struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
|
2009-01-08 03:08:00 +01:00
|
|
|
int type, name;
|
2008-07-25 10:47:20 +02:00
|
|
|
unsigned long long val;
|
|
|
|
int ret;
|
|
|
|
|
2009-01-08 03:08:00 +01:00
|
|
|
type = MEMFILE_TYPE(cft->private);
|
|
|
|
name = MEMFILE_ATTR(cft->private);
|
|
|
|
switch (name) {
|
2008-07-25 10:47:20 +02:00
|
|
|
case RES_LIMIT:
|
|
|
|
/* This function does all necessary parse...reuse it */
|
|
|
|
ret = res_counter_memparse_write_strategy(buffer, &val);
|
2009-01-08 03:08:00 +01:00
|
|
|
if (ret)
|
|
|
|
break;
|
|
|
|
if (type == _MEM)
|
2008-07-25 10:47:20 +02:00
|
|
|
ret = mem_cgroup_resize_limit(memcg, val);
|
2009-01-08 03:08:00 +01:00
|
|
|
else
|
|
|
|
ret = mem_cgroup_resize_memsw_limit(memcg, val);
|
2008-07-25 10:47:20 +02:00
|
|
|
break;
|
|
|
|
default:
|
|
|
|
ret = -EINVAL; /* should be BUG() ? */
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
return ret;
|
2008-02-07 09:13:50 +01:00
|
|
|
}
|
|
|
|
|
2008-04-29 10:00:21 +02:00
|
|
|
static int mem_cgroup_reset(struct cgroup *cont, unsigned int event)
|
2008-04-29 10:00:17 +02:00
|
|
|
{
|
|
|
|
struct mem_cgroup *mem;
|
2009-01-08 03:08:00 +01:00
|
|
|
int type, name;
|
2008-04-29 10:00:17 +02:00
|
|
|
|
|
|
|
mem = mem_cgroup_from_cont(cont);
|
2009-01-08 03:08:00 +01:00
|
|
|
type = MEMFILE_TYPE(event);
|
|
|
|
name = MEMFILE_ATTR(event);
|
|
|
|
switch (name) {
|
2008-04-29 10:00:21 +02:00
|
|
|
case RES_MAX_USAGE:
|
2009-01-08 03:08:00 +01:00
|
|
|
if (type == _MEM)
|
|
|
|
res_counter_reset_max(&mem->res);
|
|
|
|
else
|
|
|
|
res_counter_reset_max(&mem->memsw);
|
2008-04-29 10:00:21 +02:00
|
|
|
break;
|
|
|
|
case RES_FAILCNT:
|
2009-01-08 03:08:00 +01:00
|
|
|
if (type == _MEM)
|
|
|
|
res_counter_reset_failcnt(&mem->res);
|
|
|
|
else
|
|
|
|
res_counter_reset_failcnt(&mem->memsw);
|
2008-04-29 10:00:21 +02:00
|
|
|
break;
|
|
|
|
}
|
2008-04-29 10:00:20 +02:00
|
|
|
return 0;
|
2008-04-29 10:00:17 +02:00
|
|
|
}
|
|
|
|
|
2008-02-07 09:14:25 +01:00
|
|
|
static const struct mem_cgroup_stat_desc {
|
|
|
|
const char *msg;
|
|
|
|
u64 unit;
|
|
|
|
} mem_cgroup_stat_desc[] = {
|
|
|
|
[MEM_CGROUP_STAT_CACHE] = { "cache", PAGE_SIZE, },
|
|
|
|
[MEM_CGROUP_STAT_RSS] = { "rss", PAGE_SIZE, },
|
2008-05-01 13:35:12 +02:00
|
|
|
[MEM_CGROUP_STAT_PGPGIN_COUNT] = {"pgpgin", 1, },
|
|
|
|
[MEM_CGROUP_STAT_PGPGOUT_COUNT] = {"pgpgout", 1, },
|
2008-02-07 09:14:25 +01:00
|
|
|
};
|
|
|
|
|
2008-04-29 10:00:02 +02:00
|
|
|
static int mem_control_stat_show(struct cgroup *cont, struct cftype *cft,
|
|
|
|
struct cgroup_map_cb *cb)
|
2008-02-07 09:14:25 +01:00
|
|
|
{
|
|
|
|
struct mem_cgroup *mem_cont = mem_cgroup_from_cont(cont);
|
|
|
|
struct mem_cgroup_stat *stat = &mem_cont->stat;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
for (i = 0; i < ARRAY_SIZE(stat->cpustat[0].count); i++) {
|
|
|
|
s64 val;
|
|
|
|
|
|
|
|
val = mem_cgroup_read_stat(stat, i);
|
|
|
|
val *= mem_cgroup_stat_desc[i].unit;
|
2008-04-29 10:00:02 +02:00
|
|
|
cb->fill(cb, mem_cgroup_stat_desc[i].msg, val);
|
2008-02-07 09:14:25 +01:00
|
|
|
}
|
2008-02-07 09:14:31 +01:00
|
|
|
/* showing # of active pages */
|
|
|
|
{
|
2008-10-19 05:26:32 +02:00
|
|
|
unsigned long active_anon, inactive_anon;
|
|
|
|
unsigned long active_file, inactive_file;
|
2008-10-19 05:26:40 +02:00
|
|
|
unsigned long unevictable;
|
2008-10-19 05:26:32 +02:00
|
|
|
|
|
|
|
inactive_anon = mem_cgroup_get_all_zonestat(mem_cont,
|
|
|
|
LRU_INACTIVE_ANON);
|
|
|
|
active_anon = mem_cgroup_get_all_zonestat(mem_cont,
|
|
|
|
LRU_ACTIVE_ANON);
|
|
|
|
inactive_file = mem_cgroup_get_all_zonestat(mem_cont,
|
|
|
|
LRU_INACTIVE_FILE);
|
|
|
|
active_file = mem_cgroup_get_all_zonestat(mem_cont,
|
|
|
|
LRU_ACTIVE_FILE);
|
2008-10-19 05:26:40 +02:00
|
|
|
unevictable = mem_cgroup_get_all_zonestat(mem_cont,
|
|
|
|
LRU_UNEVICTABLE);
|
|
|
|
|
2008-10-19 05:26:32 +02:00
|
|
|
cb->fill(cb, "active_anon", (active_anon) * PAGE_SIZE);
|
|
|
|
cb->fill(cb, "inactive_anon", (inactive_anon) * PAGE_SIZE);
|
|
|
|
cb->fill(cb, "active_file", (active_file) * PAGE_SIZE);
|
|
|
|
cb->fill(cb, "inactive_file", (inactive_file) * PAGE_SIZE);
|
2008-10-19 05:26:40 +02:00
|
|
|
cb->fill(cb, "unevictable", unevictable * PAGE_SIZE);
|
|
|
|
|
2008-02-07 09:14:31 +01:00
|
|
|
}
|
2008-02-07 09:14:25 +01:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2009-01-08 03:07:55 +01:00
|
|
|
|
2008-02-07 09:13:50 +01:00
|
|
|
static struct cftype mem_cgroup_files[] = {
|
|
|
|
{
|
2008-02-07 09:13:57 +01:00
|
|
|
.name = "usage_in_bytes",
|
2009-01-08 03:08:00 +01:00
|
|
|
.private = MEMFILE_PRIVATE(_MEM, RES_USAGE),
|
2008-04-29 09:59:58 +02:00
|
|
|
.read_u64 = mem_cgroup_read,
|
2008-02-07 09:13:50 +01:00
|
|
|
},
|
2008-04-29 10:00:17 +02:00
|
|
|
{
|
|
|
|
.name = "max_usage_in_bytes",
|
2009-01-08 03:08:00 +01:00
|
|
|
.private = MEMFILE_PRIVATE(_MEM, RES_MAX_USAGE),
|
2008-04-29 10:00:21 +02:00
|
|
|
.trigger = mem_cgroup_reset,
|
2008-04-29 10:00:17 +02:00
|
|
|
.read_u64 = mem_cgroup_read,
|
|
|
|
},
|
2008-02-07 09:13:50 +01:00
|
|
|
{
|
2008-02-07 09:13:57 +01:00
|
|
|
.name = "limit_in_bytes",
|
2009-01-08 03:08:00 +01:00
|
|
|
.private = MEMFILE_PRIVATE(_MEM, RES_LIMIT),
|
2008-07-25 10:47:04 +02:00
|
|
|
.write_string = mem_cgroup_write,
|
2008-04-29 09:59:58 +02:00
|
|
|
.read_u64 = mem_cgroup_read,
|
2008-02-07 09:13:50 +01:00
|
|
|
},
|
|
|
|
{
|
|
|
|
.name = "failcnt",
|
2009-01-08 03:08:00 +01:00
|
|
|
.private = MEMFILE_PRIVATE(_MEM, RES_FAILCNT),
|
2008-04-29 10:00:21 +02:00
|
|
|
.trigger = mem_cgroup_reset,
|
2008-04-29 09:59:58 +02:00
|
|
|
.read_u64 = mem_cgroup_read,
|
2008-02-07 09:13:50 +01:00
|
|
|
},
|
2008-02-07 09:14:25 +01:00
|
|
|
{
|
|
|
|
.name = "stat",
|
2008-04-29 10:00:02 +02:00
|
|
|
.read_map = mem_control_stat_show,
|
2008-02-07 09:14:25 +01:00
|
|
|
},
|
2009-01-08 03:07:55 +01:00
|
|
|
{
|
|
|
|
.name = "force_empty",
|
|
|
|
.trigger = mem_cgroup_force_empty_write,
|
|
|
|
},
|
2009-01-08 03:08:07 +01:00
|
|
|
{
|
|
|
|
.name = "use_hierarchy",
|
|
|
|
.write_u64 = mem_cgroup_hierarchy_write,
|
|
|
|
.read_u64 = mem_cgroup_hierarchy_read,
|
|
|
|
},
|
2008-02-07 09:13:50 +01:00
|
|
|
};
|
|
|
|
|
2009-01-08 03:08:00 +01:00
|
|
|
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
|
|
|
|
static struct cftype memsw_cgroup_files[] = {
|
|
|
|
{
|
|
|
|
.name = "memsw.usage_in_bytes",
|
|
|
|
.private = MEMFILE_PRIVATE(_MEMSWAP, RES_USAGE),
|
|
|
|
.read_u64 = mem_cgroup_read,
|
|
|
|
},
|
|
|
|
{
|
|
|
|
.name = "memsw.max_usage_in_bytes",
|
|
|
|
.private = MEMFILE_PRIVATE(_MEMSWAP, RES_MAX_USAGE),
|
|
|
|
.trigger = mem_cgroup_reset,
|
|
|
|
.read_u64 = mem_cgroup_read,
|
|
|
|
},
|
|
|
|
{
|
|
|
|
.name = "memsw.limit_in_bytes",
|
|
|
|
.private = MEMFILE_PRIVATE(_MEMSWAP, RES_LIMIT),
|
|
|
|
.write_string = mem_cgroup_write,
|
|
|
|
.read_u64 = mem_cgroup_read,
|
|
|
|
},
|
|
|
|
{
|
|
|
|
.name = "memsw.failcnt",
|
|
|
|
.private = MEMFILE_PRIVATE(_MEMSWAP, RES_FAILCNT),
|
|
|
|
.trigger = mem_cgroup_reset,
|
|
|
|
.read_u64 = mem_cgroup_read,
|
|
|
|
},
|
|
|
|
};
|
|
|
|
|
|
|
|
static int register_memsw_files(struct cgroup *cont, struct cgroup_subsys *ss)
|
|
|
|
{
|
|
|
|
if (!do_swap_account)
|
|
|
|
return 0;
|
|
|
|
return cgroup_add_files(cont, ss, memsw_cgroup_files,
|
|
|
|
ARRAY_SIZE(memsw_cgroup_files));
|
|
|
|
};
|
|
|
|
#else
|
|
|
|
static int register_memsw_files(struct cgroup *cont, struct cgroup_subsys *ss)
|
|
|
|
{
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2008-02-07 09:14:31 +01:00
|
|
|
static int alloc_mem_cgroup_per_zone_info(struct mem_cgroup *mem, int node)
|
|
|
|
{
|
|
|
|
struct mem_cgroup_per_node *pn;
|
2008-02-07 09:14:38 +01:00
|
|
|
struct mem_cgroup_per_zone *mz;
|
2008-10-19 05:26:14 +02:00
|
|
|
enum lru_list l;
|
2008-04-09 02:41:54 +02:00
|
|
|
int zone, tmp = node;
|
2008-02-07 09:14:38 +01:00
|
|
|
/*
|
|
|
|
* This routine is called against possible nodes.
|
|
|
|
* But it's BUG to call kmalloc() against offline node.
|
|
|
|
*
|
|
|
|
* TODO: this routine can waste much memory for nodes which will
|
|
|
|
* never be onlined. It's better to use memory hotplug callback
|
|
|
|
* function.
|
|
|
|
*/
|
2008-04-09 02:41:54 +02:00
|
|
|
if (!node_state(node, N_NORMAL_MEMORY))
|
|
|
|
tmp = -1;
|
|
|
|
pn = kmalloc_node(sizeof(*pn), GFP_KERNEL, tmp);
|
2008-02-07 09:14:31 +01:00
|
|
|
if (!pn)
|
|
|
|
return 1;
|
2008-02-07 09:14:38 +01:00
|
|
|
|
2008-02-07 09:14:31 +01:00
|
|
|
mem->info.nodeinfo[node] = pn;
|
|
|
|
memset(pn, 0, sizeof(*pn));
|
2008-02-07 09:14:38 +01:00
|
|
|
|
|
|
|
for (zone = 0; zone < MAX_NR_ZONES; zone++) {
|
|
|
|
mz = &pn->zoneinfo[zone];
|
2008-10-19 05:26:14 +02:00
|
|
|
for_each_lru(l)
|
|
|
|
INIT_LIST_HEAD(&mz->lists[l]);
|
2008-02-07 09:14:38 +01:00
|
|
|
}
|
2008-02-07 09:14:31 +01:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
2008-02-07 09:14:38 +01:00
|
|
|
static void free_mem_cgroup_per_zone_info(struct mem_cgroup *mem, int node)
|
|
|
|
{
|
|
|
|
kfree(mem->info.nodeinfo[node]);
|
|
|
|
}
|
|
|
|
|
2009-01-08 03:07:53 +01:00
|
|
|
static int mem_cgroup_size(void)
|
|
|
|
{
|
|
|
|
int cpustat_size = nr_cpu_ids * sizeof(struct mem_cgroup_stat_cpu);
|
|
|
|
return sizeof(struct mem_cgroup) + cpustat_size;
|
|
|
|
}
|
|
|
|
|
2008-04-29 10:00:24 +02:00
|
|
|
static struct mem_cgroup *mem_cgroup_alloc(void)
|
|
|
|
{
|
|
|
|
struct mem_cgroup *mem;
|
2009-01-08 03:07:53 +01:00
|
|
|
int size = mem_cgroup_size();
|
2008-04-29 10:00:24 +02:00
|
|
|
|
2009-01-08 03:07:53 +01:00
|
|
|
if (size < PAGE_SIZE)
|
|
|
|
mem = kmalloc(size, GFP_KERNEL);
|
2008-04-29 10:00:24 +02:00
|
|
|
else
|
2009-01-08 03:07:53 +01:00
|
|
|
mem = vmalloc(size);
|
2008-04-29 10:00:24 +02:00
|
|
|
|
|
|
|
if (mem)
|
2009-01-08 03:07:53 +01:00
|
|
|
memset(mem, 0, size);
|
2008-04-29 10:00:24 +02:00
|
|
|
return mem;
|
|
|
|
}
|
|
|
|
|
2009-01-08 03:08:00 +01:00
|
|
|
/*
|
|
|
|
* At destroying mem_cgroup, references from swap_cgroup can remain.
|
|
|
|
* (scanning all at force_empty is too costly...)
|
|
|
|
*
|
|
|
|
* Instead of clearing all references at force_empty, we remember
|
|
|
|
* the number of reference from swap_cgroup and free mem_cgroup when
|
|
|
|
* it goes down to 0.
|
|
|
|
*
|
|
|
|
* When mem_cgroup is destroyed, mem->obsolete will be set to 0 and
|
|
|
|
* entry which points to this memcg will be ignore at swapin.
|
|
|
|
*
|
|
|
|
* Removal of cgroup itself succeeds regardless of refs from swap.
|
|
|
|
*/
|
|
|
|
|
2008-04-29 10:00:24 +02:00
|
|
|
static void mem_cgroup_free(struct mem_cgroup *mem)
|
|
|
|
{
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
int node;
|
|
|
|
|
2009-01-08 03:08:00 +01:00
|
|
|
if (atomic_read(&mem->refcnt) > 0)
|
|
|
|
return;
|
memcg: synchronized LRU
A big patch for changing memcg's LRU semantics.
Now,
- page_cgroup is linked to mem_cgroup's its own LRU (per zone).
- LRU of page_cgroup is not synchronous with global LRU.
- page and page_cgroup is one-to-one and statically allocated.
- To find page_cgroup is on what LRU, you have to check pc->mem_cgroup as
- lru = page_cgroup_zoneinfo(pc, nid_of_pc, zid_of_pc);
- SwapCache is handled.
And, when we handle LRU list of page_cgroup, we do following.
pc = lookup_page_cgroup(page);
lock_page_cgroup(pc); .....................(1)
mz = page_cgroup_zoneinfo(pc);
spin_lock(&mz->lru_lock);
.....add to LRU
spin_unlock(&mz->lru_lock);
unlock_page_cgroup(pc);
But (1) is spin_lock and we have to be afraid of dead-lock with zone->lru_lock.
So, trylock() is used at (1), now. Without (1), we can't trust "mz" is correct.
This is a trial to remove this dirty nesting of locks.
This patch changes mz->lru_lock to be zone->lru_lock.
Then, above sequence will be written as
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
mem_cgroup_add/remove/etc_lru() {
pc = lookup_page_cgroup(page);
mz = page_cgroup_zoneinfo(pc);
if (PageCgroupUsed(pc)) {
....add to LRU
}
spin_lock(&zone->lru_lock); # in vmscan.c or swap.c via global LRU
This is much simpler.
(*) We're safe even if we don't take lock_page_cgroup(pc). Because..
1. When pc->mem_cgroup can be modified.
- at charge.
- at account_move().
2. at charge
the PCG_USED bit is not set before pc->mem_cgroup is fixed.
3. at account_move()
the page is isolated and not on LRU.
Pros.
- easy for maintenance.
- memcg can make use of laziness of pagevec.
- we don't have to duplicated LRU/Active/Unevictable bit in page_cgroup.
- LRU status of memcg will be synchronized with global LRU's one.
- # of locks are reduced.
- account_move() is simplified very much.
Cons.
- may increase cost of LRU rotation.
(no impact if memcg is not configured.)
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 03:08:01 +01:00
|
|
|
|
|
|
|
|
|
|
|
for_each_node_state(node, N_POSSIBLE)
|
|
|
|
free_mem_cgroup_per_zone_info(mem, node);
|
|
|
|
|
2009-01-08 03:07:53 +01:00
|
|
|
if (mem_cgroup_size() < PAGE_SIZE)
|
2008-04-29 10:00:24 +02:00
|
|
|
kfree(mem);
|
|
|
|
else
|
|
|
|
vfree(mem);
|
|
|
|
}
|
|
|
|
|
2009-01-08 03:08:00 +01:00
|
|
|
static void mem_cgroup_get(struct mem_cgroup *mem)
|
|
|
|
{
|
|
|
|
atomic_inc(&mem->refcnt);
|
|
|
|
}
|
|
|
|
|
|
|
|
static void mem_cgroup_put(struct mem_cgroup *mem)
|
|
|
|
{
|
|
|
|
if (atomic_dec_and_test(&mem->refcnt)) {
|
|
|
|
if (!mem->obsolete)
|
|
|
|
return;
|
|
|
|
mem_cgroup_free(mem);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2008-04-29 10:00:24 +02:00
|
|
|
|
2009-01-08 03:07:57 +01:00
|
|
|
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
|
|
|
|
static void __init enable_swap_cgroup(void)
|
|
|
|
{
|
2009-01-08 03:08:02 +01:00
|
|
|
if (!mem_cgroup_disabled() && really_do_swap_account)
|
2009-01-08 03:07:57 +01:00
|
|
|
do_swap_account = 1;
|
|
|
|
}
|
|
|
|
#else
|
|
|
|
static void __init enable_swap_cgroup(void)
|
|
|
|
{
|
|
|
|
}
|
|
|
|
#endif
|
|
|
|
|
2008-02-07 09:13:50 +01:00
|
|
|
static struct cgroup_subsys_state *
|
|
|
|
mem_cgroup_create(struct cgroup_subsys *ss, struct cgroup *cont)
|
|
|
|
{
|
2009-01-08 03:08:05 +01:00
|
|
|
struct mem_cgroup *mem, *parent;
|
2008-02-07 09:14:31 +01:00
|
|
|
int node;
|
2008-02-07 09:13:50 +01:00
|
|
|
|
2009-01-08 03:07:53 +01:00
|
|
|
mem = mem_cgroup_alloc();
|
|
|
|
if (!mem)
|
|
|
|
return ERR_PTR(-ENOMEM);
|
2008-02-07 09:13:51 +01:00
|
|
|
|
2008-02-07 09:14:31 +01:00
|
|
|
for_each_node_state(node, N_POSSIBLE)
|
|
|
|
if (alloc_mem_cgroup_per_zone_info(mem, node))
|
|
|
|
goto free_out;
|
2009-01-08 03:07:57 +01:00
|
|
|
/* root ? */
|
2009-01-08 03:08:05 +01:00
|
|
|
if (cont->parent == NULL) {
|
2009-01-08 03:07:57 +01:00
|
|
|
enable_swap_cgroup();
|
2009-01-08 03:08:05 +01:00
|
|
|
parent = NULL;
|
2009-01-08 03:08:07 +01:00
|
|
|
} else {
|
2009-01-08 03:08:05 +01:00
|
|
|
parent = mem_cgroup_from_cont(cont->parent);
|
2009-01-08 03:08:07 +01:00
|
|
|
mem->use_hierarchy = parent->use_hierarchy;
|
|
|
|
}
|
2009-01-08 03:08:05 +01:00
|
|
|
|
2009-01-08 03:08:07 +01:00
|
|
|
if (parent && parent->use_hierarchy) {
|
|
|
|
res_counter_init(&mem->res, &parent->res);
|
|
|
|
res_counter_init(&mem->memsw, &parent->memsw);
|
|
|
|
} else {
|
|
|
|
res_counter_init(&mem->res, NULL);
|
|
|
|
res_counter_init(&mem->memsw, NULL);
|
|
|
|
}
|
2008-02-07 09:14:31 +01:00
|
|
|
|
2009-01-08 03:08:06 +01:00
|
|
|
mem->last_scanned_child = NULL;
|
|
|
|
|
2008-02-07 09:13:50 +01:00
|
|
|
return &mem->css;
|
2008-02-07 09:14:31 +01:00
|
|
|
free_out:
|
|
|
|
for_each_node_state(node, N_POSSIBLE)
|
2008-02-07 09:14:38 +01:00
|
|
|
free_mem_cgroup_per_zone_info(mem, node);
|
2009-01-08 03:07:53 +01:00
|
|
|
mem_cgroup_free(mem);
|
2008-02-24 00:24:14 +01:00
|
|
|
return ERR_PTR(-ENOMEM);
|
2008-02-07 09:13:50 +01:00
|
|
|
}
|
|
|
|
|
2008-02-07 09:14:28 +01:00
|
|
|
static void mem_cgroup_pre_destroy(struct cgroup_subsys *ss,
|
|
|
|
struct cgroup *cont)
|
|
|
|
{
|
|
|
|
struct mem_cgroup *mem = mem_cgroup_from_cont(cont);
|
2009-01-08 03:08:00 +01:00
|
|
|
mem->obsolete = 1;
|
2009-01-08 03:07:55 +01:00
|
|
|
mem_cgroup_force_empty(mem, false);
|
2008-02-07 09:14:28 +01:00
|
|
|
}
|
|
|
|
|
2008-02-07 09:13:50 +01:00
|
|
|
static void mem_cgroup_destroy(struct cgroup_subsys *ss,
|
|
|
|
struct cgroup *cont)
|
|
|
|
{
|
2008-04-29 10:00:24 +02:00
|
|
|
mem_cgroup_free(mem_cgroup_from_cont(cont));
|
2008-02-07 09:13:50 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
static int mem_cgroup_populate(struct cgroup_subsys *ss,
|
|
|
|
struct cgroup *cont)
|
|
|
|
{
|
2009-01-08 03:08:00 +01:00
|
|
|
int ret;
|
|
|
|
|
|
|
|
ret = cgroup_add_files(cont, ss, mem_cgroup_files,
|
|
|
|
ARRAY_SIZE(mem_cgroup_files));
|
|
|
|
|
|
|
|
if (!ret)
|
|
|
|
ret = register_memsw_files(cont, ss);
|
|
|
|
return ret;
|
2008-02-07 09:13:50 +01:00
|
|
|
}
|
|
|
|
|
2008-02-07 09:13:54 +01:00
|
|
|
static void mem_cgroup_move_task(struct cgroup_subsys *ss,
|
|
|
|
struct cgroup *cont,
|
|
|
|
struct cgroup *old_cont,
|
|
|
|
struct task_struct *p)
|
|
|
|
{
|
|
|
|
/*
|
2009-01-08 03:08:11 +01:00
|
|
|
* FIXME: It's better to move charges of this process from old
|
|
|
|
* memcg to new memcg. But it's just on TODO-List now.
|
2008-02-07 09:13:54 +01:00
|
|
|
*/
|
|
|
|
}
|
|
|
|
|
2008-02-07 09:13:50 +01:00
|
|
|
struct cgroup_subsys mem_cgroup_subsys = {
|
|
|
|
.name = "memory",
|
|
|
|
.subsys_id = mem_cgroup_subsys_id,
|
|
|
|
.create = mem_cgroup_create,
|
2008-02-07 09:14:28 +01:00
|
|
|
.pre_destroy = mem_cgroup_pre_destroy,
|
2008-02-07 09:13:50 +01:00
|
|
|
.destroy = mem_cgroup_destroy,
|
|
|
|
.populate = mem_cgroup_populate,
|
2008-02-07 09:13:54 +01:00
|
|
|
.attach = mem_cgroup_move_task,
|
2008-02-07 09:14:31 +01:00
|
|
|
.early_init = 0,
|
2008-02-07 09:13:50 +01:00
|
|
|
};
|
2009-01-08 03:07:57 +01:00
|
|
|
|
|
|
|
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
|
|
|
|
|
|
|
|
static int __init disable_swap_account(char *s)
|
|
|
|
{
|
|
|
|
really_do_swap_account = 0;
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
__setup("noswapaccount", disable_swap_account);
|
|
|
|
#endif
|