memcontrol.c 187.2 KB
Newer Older
B
Balbir Singh 已提交
1 2 3 4 5
/* memcontrol.c - Memory Controller
 *
 * Copyright IBM Corporation, 2007
 * Author Balbir Singh <balbir@linux.vnet.ibm.com>
 *
6 7 8
 * Copyright 2007 OpenVZ SWsoft Inc
 * Author: Pavel Emelianov <xemul@openvz.org>
 *
9 10 11 12
 * Memory thresholds
 * Copyright (C) 2009 Nokia Corporation
 * Author: Kirill A. Shutemov
 *
13 14 15 16
 * Kernel Memory Controller
 * Copyright (C) 2012 Parallels Inc. and Google Inc.
 * Authors: Glauber Costa and Suleiman Souhlal
 *
B
Balbir Singh 已提交
17 18 19 20 21 22 23 24 25 26 27 28 29 30
 * 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>
31
#include <linux/mm.h>
32
#include <linux/hugetlb.h>
K
KAMEZAWA Hiroyuki 已提交
33
#include <linux/pagemap.h>
34
#include <linux/smp.h>
35
#include <linux/page-flags.h>
36
#include <linux/backing-dev.h>
37 38
#include <linux/bit_spinlock.h>
#include <linux/rcupdate.h>
39
#include <linux/limits.h>
40
#include <linux/export.h>
41
#include <linux/mutex.h>
42
#include <linux/rbtree.h>
43
#include <linux/slab.h>
44
#include <linux/swap.h>
45
#include <linux/swapops.h>
46
#include <linux/spinlock.h>
47
#include <linux/eventfd.h>
48
#include <linux/poll.h>
49
#include <linux/sort.h>
50
#include <linux/fs.h>
51
#include <linux/seq_file.h>
52
#include <linux/vmpressure.h>
53
#include <linux/mm_inline.h>
54
#include <linux/page_cgroup.h>
55
#include <linux/cpu.h>
56
#include <linux/oom.h>
57
#include <linux/lockdep.h>
58
#include <linux/file.h>
K
KAMEZAWA Hiroyuki 已提交
59
#include "internal.h"
G
Glauber Costa 已提交
60
#include <net/sock.h>
M
Michal Hocko 已提交
61
#include <net/ip.h>
G
Glauber Costa 已提交
62
#include <net/tcp_memcontrol.h>
63
#include "slab.h"
B
Balbir Singh 已提交
64

65 66
#include <asm/uaccess.h>

67 68
#include <trace/events/vmscan.h>

69 70
struct cgroup_subsys memory_cgrp_subsys __read_mostly;
EXPORT_SYMBOL(memory_cgrp_subsys);
71

72
#define MEM_CGROUP_RECLAIM_RETRIES	5
73
static struct mem_cgroup *root_mem_cgroup __read_mostly;
B
Balbir Singh 已提交
74

A
Andrew Morton 已提交
75
#ifdef CONFIG_MEMCG_SWAP
L
Li Zefan 已提交
76
/* Turned on only when memory cgroup is enabled && really_do_swap_account = 1 */
77
int do_swap_account __read_mostly;
78 79

/* for remember boot option*/
A
Andrew Morton 已提交
80
#ifdef CONFIG_MEMCG_SWAP_ENABLED
81 82 83 84 85
static int really_do_swap_account __initdata = 1;
#else
static int really_do_swap_account __initdata = 0;
#endif

86
#else
87
#define do_swap_account		0
88 89 90
#endif


91 92 93
static const char * const mem_cgroup_stat_names[] = {
	"cache",
	"rss",
94
	"rss_huge",
95
	"mapped_file",
96
	"writeback",
97 98 99
	"swap",
};

100 101 102
enum mem_cgroup_events_index {
	MEM_CGROUP_EVENTS_PGPGIN,	/* # of pages paged in */
	MEM_CGROUP_EVENTS_PGPGOUT,	/* # of pages paged out */
103 104
	MEM_CGROUP_EVENTS_PGFAULT,	/* # of page-faults */
	MEM_CGROUP_EVENTS_PGMAJFAULT,	/* # of major page-faults */
105 106
	MEM_CGROUP_EVENTS_NSTATS,
};
107 108 109 110 111 112 113 114

static const char * const mem_cgroup_events_names[] = {
	"pgpgin",
	"pgpgout",
	"pgfault",
	"pgmajfault",
};

115 116 117 118 119 120 121 122
static const char * const mem_cgroup_lru_names[] = {
	"inactive_anon",
	"active_anon",
	"inactive_file",
	"active_file",
	"unevictable",
};

123 124 125 126 127 128 129 130
/*
 * Per memcg event counter is incremented at every pagein/pageout. With THP,
 * it will be incremated by the number of pages. This counter is used for
 * for trigger some periodic events. This is straightforward and better
 * than using jiffies etc. to handle periodic memcg event.
 */
enum mem_cgroup_events_target {
	MEM_CGROUP_TARGET_THRESH,
131
	MEM_CGROUP_TARGET_SOFTLIMIT,
132
	MEM_CGROUP_TARGET_NUMAINFO,
133 134
	MEM_CGROUP_NTARGETS,
};
135 136 137
#define THRESHOLDS_EVENTS_TARGET 128
#define SOFTLIMIT_EVENTS_TARGET 1024
#define NUMAINFO_EVENTS_TARGET	1024
138

139
struct mem_cgroup_stat_cpu {
140
	long count[MEM_CGROUP_STAT_NSTATS];
141
	unsigned long events[MEM_CGROUP_EVENTS_NSTATS];
142
	unsigned long nr_page_events;
143
	unsigned long targets[MEM_CGROUP_NTARGETS];
144 145
};

146
struct mem_cgroup_reclaim_iter {
M
Michal Hocko 已提交
147 148 149 150
	/*
	 * last scanned hierarchy member. Valid only if last_dead_count
	 * matches memcg->dead_count of the hierarchy root group.
	 */
151
	struct mem_cgroup *last_visited;
152
	int last_dead_count;
M
Michal Hocko 已提交
153

154 155 156 157
	/* scan generation, increased every round-trip */
	unsigned int generation;
};

158 159 160 161
/*
 * per-zone information in memory controller.
 */
struct mem_cgroup_per_zone {
162
	struct lruvec		lruvec;
163
	unsigned long		lru_size[NR_LRU_LISTS];
K
KOSAKI Motohiro 已提交
164

165 166
	struct mem_cgroup_reclaim_iter reclaim_iter[DEF_PRIORITY + 1];

167 168 169 170
	struct rb_node		tree_node;	/* RB tree node */
	unsigned long long	usage_in_excess;/* Set to the value by which */
						/* the soft limit is exceeded*/
	bool			on_tree;
171
	struct mem_cgroup	*memcg;		/* Back pointer, we cannot */
172
						/* use container_of	   */
173 174 175 176 177 178
};

struct mem_cgroup_per_node {
	struct mem_cgroup_per_zone zoneinfo[MAX_NR_ZONES];
};

179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198
/*
 * Cgroups above their limits are maintained in a RB-Tree, independent of
 * their hierarchy representation
 */

struct mem_cgroup_tree_per_zone {
	struct rb_root rb_root;
	spinlock_t lock;
};

struct mem_cgroup_tree_per_node {
	struct mem_cgroup_tree_per_zone rb_tree_per_zone[MAX_NR_ZONES];
};

struct mem_cgroup_tree {
	struct mem_cgroup_tree_per_node *rb_tree_per_node[MAX_NUMNODES];
};

static struct mem_cgroup_tree soft_limit_tree __read_mostly;

199 200 201 202 203
struct mem_cgroup_threshold {
	struct eventfd_ctx *eventfd;
	u64 threshold;
};

K
KAMEZAWA Hiroyuki 已提交
204
/* For threshold */
205
struct mem_cgroup_threshold_ary {
206
	/* An array index points to threshold just below or equal to usage. */
207
	int current_threshold;
208 209 210 211 212
	/* Size of entries[] */
	unsigned int size;
	/* Array of thresholds */
	struct mem_cgroup_threshold entries[0];
};
213 214 215 216 217 218 219 220 221 222 223 224

struct mem_cgroup_thresholds {
	/* Primary thresholds array */
	struct mem_cgroup_threshold_ary *primary;
	/*
	 * Spare threshold array.
	 * This is needed to make mem_cgroup_unregister_event() "never fail".
	 * It must be able to store at least primary->size - 1 entries.
	 */
	struct mem_cgroup_threshold_ary *spare;
};

K
KAMEZAWA Hiroyuki 已提交
225 226 227 228 229
/* for OOM */
struct mem_cgroup_eventfd_list {
	struct list_head list;
	struct eventfd_ctx *eventfd;
};
230

231 232 233
/*
 * cgroup_event represents events which userspace want to receive.
 */
234
struct mem_cgroup_event {
235
	/*
236
	 * memcg which the event belongs to.
237
	 */
238
	struct mem_cgroup *memcg;
239 240 241 242 243 244 245 246
	/*
	 * eventfd to signal userspace about the event.
	 */
	struct eventfd_ctx *eventfd;
	/*
	 * Each of these stored in a list by the cgroup.
	 */
	struct list_head list;
247 248 249 250 251
	/*
	 * register_event() callback will be used to add new userspace
	 * waiter for changes related to this event.  Use eventfd_signal()
	 * on eventfd to send notification to userspace.
	 */
252
	int (*register_event)(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
253
			      struct eventfd_ctx *eventfd, const char *args);
254 255 256 257 258
	/*
	 * unregister_event() callback will be called when userspace closes
	 * the eventfd or on cgroup removing.  This callback must be set,
	 * if you want provide notification functionality.
	 */
259
	void (*unregister_event)(struct mem_cgroup *memcg,
260
				 struct eventfd_ctx *eventfd);
261 262 263 264 265 266 267 268 269 270
	/*
	 * All fields below needed to unregister event when
	 * userspace closes eventfd.
	 */
	poll_table pt;
	wait_queue_head_t *wqh;
	wait_queue_t wait;
	struct work_struct remove;
};

271 272
static void mem_cgroup_threshold(struct mem_cgroup *memcg);
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg);
273

B
Balbir Singh 已提交
274 275 276 277 278 279 280
/*
 * 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
281 282 283
 * 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.
B
Balbir Singh 已提交
284 285 286 287 288 289 290
 */
struct mem_cgroup {
	struct cgroup_subsys_state css;
	/*
	 * the counter to account for memory usage
	 */
	struct res_counter res;
291

292 293 294
	/* vmpressure notifications */
	struct vmpressure vmpressure;

295 296 297 298
	/*
	 * the counter to account for mem+swap usage.
	 */
	struct res_counter memsw;
299

300 301 302 303
	/*
	 * the counter to account for kernel memory usage.
	 */
	struct res_counter kmem;
304 305 306 307
	/*
	 * Should the accounting and control be hierarchical, per subtree?
	 */
	bool use_hierarchy;
308
	unsigned long kmem_account_flags; /* See KMEM_ACCOUNTED_*, below */
309 310 311

	bool		oom_lock;
	atomic_t	under_oom;
312
	atomic_t	oom_wakeups;
313

314
	int	swappiness;
315 316
	/* OOM-Killer disable */
	int		oom_kill_disable;
K
KOSAKI Motohiro 已提交
317

318 319 320
	/* set when res.limit == memsw.limit */
	bool		memsw_is_minimum;

321 322 323 324
	/* protect arrays of thresholds */
	struct mutex thresholds_lock;

	/* thresholds for memory usage. RCU-protected */
325
	struct mem_cgroup_thresholds thresholds;
326

327
	/* thresholds for mem+swap usage. RCU-protected */
328
	struct mem_cgroup_thresholds memsw_thresholds;
329

K
KAMEZAWA Hiroyuki 已提交
330 331
	/* For oom notifier event fd */
	struct list_head oom_notify;
332

333 334 335 336
	/*
	 * Should we move charges of a task when a task is moved into this
	 * mem_cgroup ? And what type of charges should we move ?
	 */
A
Andrew Morton 已提交
337
	unsigned long move_charge_at_immigrate;
338 339 340 341
	/*
	 * set > 0 if pages under this cgroup are moving to other cgroup.
	 */
	atomic_t	moving_account;
342 343
	/* taken only while moving_account > 0 */
	spinlock_t	move_lock;
344
	/*
345
	 * percpu counter.
346
	 */
347
	struct mem_cgroup_stat_cpu __percpu *stat;
348 349 350 351 352 353
	/*
	 * used when a cpu is offlined or other synchronizations
	 * See mem_cgroup_read_stat().
	 */
	struct mem_cgroup_stat_cpu nocpu_base;
	spinlock_t pcp_counter_lock;
G
Glauber Costa 已提交
354

M
Michal Hocko 已提交
355
	atomic_t	dead_count;
M
Michal Hocko 已提交
356
#if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_INET)
357
	struct cg_proto tcp_mem;
G
Glauber Costa 已提交
358
#endif
359
#if defined(CONFIG_MEMCG_KMEM)
360 361
	/* analogous to slab_common's slab_caches list, but per-memcg;
	 * protected by memcg_slab_mutex */
362 363 364 365
	struct list_head memcg_slab_caches;
        /* Index in the kmem_cache->memcg_params->memcg_caches array */
	int kmemcg_id;
#endif
366 367 368 369 370 371 372

	int last_scanned_node;
#if MAX_NUMNODES > 1
	nodemask_t	scan_nodes;
	atomic_t	numainfo_events;
	atomic_t	numainfo_updating;
#endif
373

374 375 376 377
	/* List of events which userspace want to receive */
	struct list_head event_list;
	spinlock_t event_list_lock;

378 379
	struct mem_cgroup_per_node *nodeinfo[0];
	/* WARNING: nodeinfo must be the last member here */
B
Balbir Singh 已提交
380 381
};

382 383
/* internal only representation about the status of kmem accounting. */
enum {
384
	KMEM_ACCOUNTED_ACTIVE, /* accounted by this cgroup itself */
385
	KMEM_ACCOUNTED_DEAD, /* dead memcg with pending kmem charges */
386 387 388 389 390 391 392
};

#ifdef CONFIG_MEMCG_KMEM
static inline void memcg_kmem_set_active(struct mem_cgroup *memcg)
{
	set_bit(KMEM_ACCOUNTED_ACTIVE, &memcg->kmem_account_flags);
}
393 394 395 396 397 398 399 400

static bool memcg_kmem_is_active(struct mem_cgroup *memcg)
{
	return test_bit(KMEM_ACCOUNTED_ACTIVE, &memcg->kmem_account_flags);
}

static void memcg_kmem_mark_dead(struct mem_cgroup *memcg)
{
401 402 403 404 405
	/*
	 * Our caller must use css_get() first, because memcg_uncharge_kmem()
	 * will call css_put() if it sees the memcg is dead.
	 */
	smp_wmb();
406 407 408 409 410 411 412 413 414
	if (test_bit(KMEM_ACCOUNTED_ACTIVE, &memcg->kmem_account_flags))
		set_bit(KMEM_ACCOUNTED_DEAD, &memcg->kmem_account_flags);
}

static bool memcg_kmem_test_and_clear_dead(struct mem_cgroup *memcg)
{
	return test_and_clear_bit(KMEM_ACCOUNTED_DEAD,
				  &memcg->kmem_account_flags);
}
415 416
#endif

417 418
/* Stuffs for move charges at task migration. */
/*
419 420
 * Types of charges to be moved. "move_charge_at_immitgrate" and
 * "immigrate_flags" are treated as a left-shifted bitmap of these types.
421 422
 */
enum move_type {
423
	MOVE_CHARGE_TYPE_ANON,	/* private anonymous page and swap of it */
424
	MOVE_CHARGE_TYPE_FILE,	/* file page(including tmpfs) and swap of it */
425 426 427
	NR_MOVE_TYPE,
};

428 429
/* "mc" and its members are protected by cgroup_mutex */
static struct move_charge_struct {
430
	spinlock_t	  lock; /* for from, to */
431 432
	struct mem_cgroup *from;
	struct mem_cgroup *to;
433
	unsigned long immigrate_flags;
434
	unsigned long precharge;
435
	unsigned long moved_charge;
436
	unsigned long moved_swap;
437 438 439
	struct task_struct *moving_task;	/* a task moving charges */
	wait_queue_head_t waitq;		/* a waitq for other context */
} mc = {
440
	.lock = __SPIN_LOCK_UNLOCKED(mc.lock),
441 442
	.waitq = __WAIT_QUEUE_HEAD_INITIALIZER(mc.waitq),
};
443

D
Daisuke Nishimura 已提交
444 445
static bool move_anon(void)
{
446
	return test_bit(MOVE_CHARGE_TYPE_ANON, &mc.immigrate_flags);
D
Daisuke Nishimura 已提交
447 448
}

449 450
static bool move_file(void)
{
451
	return test_bit(MOVE_CHARGE_TYPE_FILE, &mc.immigrate_flags);
452 453
}

454 455 456 457
/*
 * Maximum loops in mem_cgroup_hierarchical_reclaim(), used for soft
 * limit reclaim to prevent infinite loops, if they ever occur.
 */
458
#define	MEM_CGROUP_MAX_RECLAIM_LOOPS		100
459
#define	MEM_CGROUP_MAX_SOFT_LIMIT_RECLAIM_LOOPS	2
460

461 462
enum charge_type {
	MEM_CGROUP_CHARGE_TYPE_CACHE = 0,
463
	MEM_CGROUP_CHARGE_TYPE_ANON,
K
KAMEZAWA Hiroyuki 已提交
464
	MEM_CGROUP_CHARGE_TYPE_SWAPOUT,	/* for accounting swapcache */
K
KAMEZAWA Hiroyuki 已提交
465
	MEM_CGROUP_CHARGE_TYPE_DROP,	/* a page was unused swap cache */
466 467 468
	NR_CHARGE_TYPE,
};

469
/* for encoding cft->private value on file */
G
Glauber Costa 已提交
470 471 472 473
enum res_type {
	_MEM,
	_MEMSWAP,
	_OOM_TYPE,
474
	_KMEM,
G
Glauber Costa 已提交
475 476
};

477 478
#define MEMFILE_PRIVATE(x, val)	((x) << 16 | (val))
#define MEMFILE_TYPE(val)	((val) >> 16 & 0xffff)
479
#define MEMFILE_ATTR(val)	((val) & 0xffff)
K
KAMEZAWA Hiroyuki 已提交
480 481
/* Used for OOM nofiier */
#define OOM_CONTROL		(0)
482

483 484 485 486 487 488 489 490
/*
 * Reclaim flags for mem_cgroup_hierarchical_reclaim
 */
#define MEM_CGROUP_RECLAIM_NOSWAP_BIT	0x0
#define MEM_CGROUP_RECLAIM_NOSWAP	(1 << MEM_CGROUP_RECLAIM_NOSWAP_BIT)
#define MEM_CGROUP_RECLAIM_SHRINK_BIT	0x1
#define MEM_CGROUP_RECLAIM_SHRINK	(1 << MEM_CGROUP_RECLAIM_SHRINK_BIT)

491 492 493 494 495 496 497
/*
 * The memcg_create_mutex will be held whenever a new cgroup is created.
 * As a consequence, any change that needs to protect against new child cgroups
 * appearing has to hold it as well.
 */
static DEFINE_MUTEX(memcg_create_mutex);

498 499
struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *s)
{
500
	return s ? container_of(s, struct mem_cgroup, css) : NULL;
501 502
}

503 504 505 506 507 508 509 510 511 512 513 514 515
/* Some nice accessors for the vmpressure. */
struct vmpressure *memcg_to_vmpressure(struct mem_cgroup *memcg)
{
	if (!memcg)
		memcg = root_mem_cgroup;
	return &memcg->vmpressure;
}

struct cgroup_subsys_state *vmpressure_to_css(struct vmpressure *vmpr)
{
	return &container_of(vmpr, struct mem_cgroup, vmpressure)->css;
}

516 517 518 519 520
static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
{
	return (memcg == root_mem_cgroup);
}

521 522 523 524 525 526
/*
 * We restrict the id in the range of [1, 65535], so it can fit into
 * an unsigned short.
 */
#define MEM_CGROUP_ID_MAX	USHRT_MAX

L
Li Zefan 已提交
527 528 529 530 531 532 533 534 535 536 537 538 539
static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
{
	/*
	 * The ID of the root cgroup is 0, but memcg treat 0 as an
	 * invalid ID, so we return (cgroup_id + 1).
	 */
	return memcg->css.cgroup->id + 1;
}

static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
{
	struct cgroup_subsys_state *css;

540
	css = css_from_id(id - 1, &memory_cgrp_subsys);
L
Li Zefan 已提交
541 542 543
	return mem_cgroup_from_css(css);
}

G
Glauber Costa 已提交
544
/* Writing them here to avoid exposing memcg's inner layout */
M
Michal Hocko 已提交
545
#if defined(CONFIG_INET) && defined(CONFIG_MEMCG_KMEM)
G
Glauber Costa 已提交
546 547 548

void sock_update_memcg(struct sock *sk)
{
549
	if (mem_cgroup_sockets_enabled) {
G
Glauber Costa 已提交
550
		struct mem_cgroup *memcg;
551
		struct cg_proto *cg_proto;
G
Glauber Costa 已提交
552 553 554

		BUG_ON(!sk->sk_prot->proto_cgroup);

555 556 557 558 559 560 561 562 563 564
		/* Socket cloning can throw us here with sk_cgrp already
		 * filled. It won't however, necessarily happen from
		 * process context. So the test for root memcg given
		 * the current task's memcg won't help us in this case.
		 *
		 * Respecting the original socket's memcg is a better
		 * decision in this case.
		 */
		if (sk->sk_cgrp) {
			BUG_ON(mem_cgroup_is_root(sk->sk_cgrp->memcg));
565
			css_get(&sk->sk_cgrp->memcg->css);
566 567 568
			return;
		}

G
Glauber Costa 已提交
569 570
		rcu_read_lock();
		memcg = mem_cgroup_from_task(current);
571
		cg_proto = sk->sk_prot->proto_cgroup(memcg);
572 573
		if (!mem_cgroup_is_root(memcg) &&
		    memcg_proto_active(cg_proto) && css_tryget(&memcg->css)) {
574
			sk->sk_cgrp = cg_proto;
G
Glauber Costa 已提交
575 576 577 578 579 580 581 582
		}
		rcu_read_unlock();
	}
}
EXPORT_SYMBOL(sock_update_memcg);

void sock_release_memcg(struct sock *sk)
{
583
	if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
G
Glauber Costa 已提交
584 585 586
		struct mem_cgroup *memcg;
		WARN_ON(!sk->sk_cgrp->memcg);
		memcg = sk->sk_cgrp->memcg;
587
		css_put(&sk->sk_cgrp->memcg->css);
G
Glauber Costa 已提交
588 589
	}
}
G
Glauber Costa 已提交
590 591 592 593 594 595

struct cg_proto *tcp_proto_cgroup(struct mem_cgroup *memcg)
{
	if (!memcg || mem_cgroup_is_root(memcg))
		return NULL;

596
	return &memcg->tcp_mem;
G
Glauber Costa 已提交
597 598
}
EXPORT_SYMBOL(tcp_proto_cgroup);
G
Glauber Costa 已提交
599

600 601
static void disarm_sock_keys(struct mem_cgroup *memcg)
{
602
	if (!memcg_proto_activated(&memcg->tcp_mem))
603 604 605 606 607 608 609 610 611
		return;
	static_key_slow_dec(&memcg_socket_limit_enabled);
}
#else
static void disarm_sock_keys(struct mem_cgroup *memcg)
{
}
#endif

612
#ifdef CONFIG_MEMCG_KMEM
613 614
/*
 * This will be the memcg's index in each cache's ->memcg_params->memcg_caches.
L
Li Zefan 已提交
615 616 617 618 619
 * The main reason for not using cgroup id for this:
 *  this works better in sparse environments, where we have a lot of memcgs,
 *  but only a few kmem-limited. Or also, if we have, for instance, 200
 *  memcgs, and none but the 200th is kmem-limited, we'd have to have a
 *  200 entry array for that.
620 621 622 623 624 625
 *
 * The current size of the caches array is stored in
 * memcg_limited_groups_array_size.  It will double each time we have to
 * increase it.
 */
static DEFINE_IDA(kmem_limited_groups);
626 627
int memcg_limited_groups_array_size;

628 629 630 631 632 633
/*
 * MIN_SIZE is different than 1, because we would like to avoid going through
 * the alloc/free process all the time. In a small machine, 4 kmem-limited
 * cgroups is a reasonable guess. In the future, it could be a parameter or
 * tunable, but that is strictly not necessary.
 *
L
Li Zefan 已提交
634
 * MAX_SIZE should be as large as the number of cgrp_ids. Ideally, we could get
635 636
 * this constant directly from cgroup, but it is understandable that this is
 * better kept as an internal representation in cgroup.c. In any case, the
L
Li Zefan 已提交
637
 * cgrp_id space is not getting any smaller, and we don't have to necessarily
638 639 640
 * increase ours as well if it increases.
 */
#define MEMCG_CACHES_MIN_SIZE 4
L
Li Zefan 已提交
641
#define MEMCG_CACHES_MAX_SIZE MEM_CGROUP_ID_MAX
642

643 644 645 646 647 648
/*
 * A lot of the calls to the cache allocation functions are expected to be
 * inlined by the compiler. Since the calls to memcg_kmem_get_cache are
 * conditional to this static branch, we'll have to allow modules that does
 * kmem_cache_alloc and the such to see this symbol as well
 */
649
struct static_key memcg_kmem_enabled_key;
650
EXPORT_SYMBOL(memcg_kmem_enabled_key);
651 652 653

static void disarm_kmem_keys(struct mem_cgroup *memcg)
{
654
	if (memcg_kmem_is_active(memcg)) {
655
		static_key_slow_dec(&memcg_kmem_enabled_key);
656 657
		ida_simple_remove(&kmem_limited_groups, memcg->kmemcg_id);
	}
658 659 660 661 662
	/*
	 * This check can't live in kmem destruction function,
	 * since the charges will outlive the cgroup
	 */
	WARN_ON(res_counter_read_u64(&memcg->kmem, RES_USAGE) != 0);
663 664 665 666 667 668 669 670 671 672 673 674 675
}
#else
static void disarm_kmem_keys(struct mem_cgroup *memcg)
{
}
#endif /* CONFIG_MEMCG_KMEM */

static void disarm_static_keys(struct mem_cgroup *memcg)
{
	disarm_sock_keys(memcg);
	disarm_kmem_keys(memcg);
}

676
static void drain_all_stock_async(struct mem_cgroup *memcg);
677

678
static struct mem_cgroup_per_zone *
679
mem_cgroup_zoneinfo(struct mem_cgroup *memcg, int nid, int zid)
680
{
681
	VM_BUG_ON((unsigned)nid >= nr_node_ids);
682
	return &memcg->nodeinfo[nid]->zoneinfo[zid];
683 684
}

685
struct cgroup_subsys_state *mem_cgroup_css(struct mem_cgroup *memcg)
686
{
687
	return &memcg->css;
688 689
}

690
static struct mem_cgroup_per_zone *
691
page_cgroup_zoneinfo(struct mem_cgroup *memcg, struct page *page)
692
{
693 694
	int nid = page_to_nid(page);
	int zid = page_zonenum(page);
695

696
	return mem_cgroup_zoneinfo(memcg, nid, zid);
697 698
}

699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856
static struct mem_cgroup_tree_per_zone *
soft_limit_tree_node_zone(int nid, int zid)
{
	return &soft_limit_tree.rb_tree_per_node[nid]->rb_tree_per_zone[zid];
}

static struct mem_cgroup_tree_per_zone *
soft_limit_tree_from_page(struct page *page)
{
	int nid = page_to_nid(page);
	int zid = page_zonenum(page);

	return &soft_limit_tree.rb_tree_per_node[nid]->rb_tree_per_zone[zid];
}

static void
__mem_cgroup_insert_exceeded(struct mem_cgroup *memcg,
				struct mem_cgroup_per_zone *mz,
				struct mem_cgroup_tree_per_zone *mctz,
				unsigned long long new_usage_in_excess)
{
	struct rb_node **p = &mctz->rb_root.rb_node;
	struct rb_node *parent = NULL;
	struct mem_cgroup_per_zone *mz_node;

	if (mz->on_tree)
		return;

	mz->usage_in_excess = new_usage_in_excess;
	if (!mz->usage_in_excess)
		return;
	while (*p) {
		parent = *p;
		mz_node = rb_entry(parent, struct mem_cgroup_per_zone,
					tree_node);
		if (mz->usage_in_excess < mz_node->usage_in_excess)
			p = &(*p)->rb_left;
		/*
		 * We can't avoid mem cgroups that are over their soft
		 * limit by the same amount
		 */
		else if (mz->usage_in_excess >= mz_node->usage_in_excess)
			p = &(*p)->rb_right;
	}
	rb_link_node(&mz->tree_node, parent, p);
	rb_insert_color(&mz->tree_node, &mctz->rb_root);
	mz->on_tree = true;
}

static void
__mem_cgroup_remove_exceeded(struct mem_cgroup *memcg,
				struct mem_cgroup_per_zone *mz,
				struct mem_cgroup_tree_per_zone *mctz)
{
	if (!mz->on_tree)
		return;
	rb_erase(&mz->tree_node, &mctz->rb_root);
	mz->on_tree = false;
}

static void
mem_cgroup_remove_exceeded(struct mem_cgroup *memcg,
				struct mem_cgroup_per_zone *mz,
				struct mem_cgroup_tree_per_zone *mctz)
{
	spin_lock(&mctz->lock);
	__mem_cgroup_remove_exceeded(memcg, mz, mctz);
	spin_unlock(&mctz->lock);
}


static void mem_cgroup_update_tree(struct mem_cgroup *memcg, struct page *page)
{
	unsigned long long excess;
	struct mem_cgroup_per_zone *mz;
	struct mem_cgroup_tree_per_zone *mctz;
	int nid = page_to_nid(page);
	int zid = page_zonenum(page);
	mctz = soft_limit_tree_from_page(page);

	/*
	 * Necessary to update all ancestors when hierarchy is used.
	 * because their event counter is not touched.
	 */
	for (; memcg; memcg = parent_mem_cgroup(memcg)) {
		mz = mem_cgroup_zoneinfo(memcg, nid, zid);
		excess = res_counter_soft_limit_excess(&memcg->res);
		/*
		 * We have to update the tree if mz is on RB-tree or
		 * mem is over its softlimit.
		 */
		if (excess || mz->on_tree) {
			spin_lock(&mctz->lock);
			/* if on-tree, remove it */
			if (mz->on_tree)
				__mem_cgroup_remove_exceeded(memcg, mz, mctz);
			/*
			 * Insert again. mz->usage_in_excess will be updated.
			 * If excess is 0, no tree ops.
			 */
			__mem_cgroup_insert_exceeded(memcg, mz, mctz, excess);
			spin_unlock(&mctz->lock);
		}
	}
}

static void mem_cgroup_remove_from_trees(struct mem_cgroup *memcg)
{
	int node, zone;
	struct mem_cgroup_per_zone *mz;
	struct mem_cgroup_tree_per_zone *mctz;

	for_each_node(node) {
		for (zone = 0; zone < MAX_NR_ZONES; zone++) {
			mz = mem_cgroup_zoneinfo(memcg, node, zone);
			mctz = soft_limit_tree_node_zone(node, zone);
			mem_cgroup_remove_exceeded(memcg, mz, mctz);
		}
	}
}

static struct mem_cgroup_per_zone *
__mem_cgroup_largest_soft_limit_node(struct mem_cgroup_tree_per_zone *mctz)
{
	struct rb_node *rightmost = NULL;
	struct mem_cgroup_per_zone *mz;

retry:
	mz = NULL;
	rightmost = rb_last(&mctz->rb_root);
	if (!rightmost)
		goto done;		/* Nothing to reclaim from */

	mz = rb_entry(rightmost, struct mem_cgroup_per_zone, tree_node);
	/*
	 * Remove the node now but someone else can add it back,
	 * we will to add it back at the end of reclaim to its correct
	 * position in the tree.
	 */
	__mem_cgroup_remove_exceeded(mz->memcg, mz, mctz);
	if (!res_counter_soft_limit_excess(&mz->memcg->res) ||
		!css_tryget(&mz->memcg->css))
		goto retry;
done:
	return mz;
}

static struct mem_cgroup_per_zone *
mem_cgroup_largest_soft_limit_node(struct mem_cgroup_tree_per_zone *mctz)
{
	struct mem_cgroup_per_zone *mz;

	spin_lock(&mctz->lock);
	mz = __mem_cgroup_largest_soft_limit_node(mctz);
	spin_unlock(&mctz->lock);
	return mz;
}

857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875
/*
 * Implementation Note: reading percpu statistics for memcg.
 *
 * Both of vmstat[] and percpu_counter has threshold and do periodic
 * synchronization to implement "quick" read. There are trade-off between
 * reading cost and precision of value. Then, we may have a chance to implement
 * a periodic synchronizion of counter in memcg's counter.
 *
 * But this _read() function is used for user interface now. The user accounts
 * memory usage by memory cgroup and he _always_ requires exact value because
 * he accounts memory. Even if we provide quick-and-fuzzy read, we always
 * have to visit all online cpus and make sum. So, for now, unnecessary
 * synchronization is not implemented. (just implemented for cpu hotplug)
 *
 * If there are kernel internal actions which can make use of some not-exact
 * value, and reading all cpu value can be performance bottleneck in some
 * common workload, threashold and synchonization as vmstat[] should be
 * implemented.
 */
876
static long mem_cgroup_read_stat(struct mem_cgroup *memcg,
877
				 enum mem_cgroup_stat_index idx)
878
{
879
	long val = 0;
880 881
	int cpu;

882 883
	get_online_cpus();
	for_each_online_cpu(cpu)
884
		val += per_cpu(memcg->stat->count[idx], cpu);
885
#ifdef CONFIG_HOTPLUG_CPU
886 887 888
	spin_lock(&memcg->pcp_counter_lock);
	val += memcg->nocpu_base.count[idx];
	spin_unlock(&memcg->pcp_counter_lock);
889 890
#endif
	put_online_cpus();
891 892 893
	return val;
}

894
static void mem_cgroup_swap_statistics(struct mem_cgroup *memcg,
895 896 897
					 bool charge)
{
	int val = (charge) ? 1 : -1;
898
	this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_SWAP], val);
899 900
}

901
static unsigned long mem_cgroup_read_events(struct mem_cgroup *memcg,
902 903 904 905 906
					    enum mem_cgroup_events_index idx)
{
	unsigned long val = 0;
	int cpu;

907
	get_online_cpus();
908
	for_each_online_cpu(cpu)
909
		val += per_cpu(memcg->stat->events[idx], cpu);
910
#ifdef CONFIG_HOTPLUG_CPU
911 912 913
	spin_lock(&memcg->pcp_counter_lock);
	val += memcg->nocpu_base.events[idx];
	spin_unlock(&memcg->pcp_counter_lock);
914
#endif
915
	put_online_cpus();
916 917 918
	return val;
}

919
static void mem_cgroup_charge_statistics(struct mem_cgroup *memcg,
920
					 struct page *page,
921
					 bool anon, int nr_pages)
922
{
923 924 925 926 927 928
	/*
	 * Here, RSS means 'mapped anon' and anon's SwapCache. Shmem/tmpfs is
	 * counted as CACHE even if it's on ANON LRU.
	 */
	if (anon)
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_RSS],
929
				nr_pages);
930
	else
931
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_CACHE],
932
				nr_pages);
933

934 935 936 937
	if (PageTransHuge(page))
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_RSS_HUGE],
				nr_pages);

938 939
	/* pagein of a big page is an event. So, ignore page size */
	if (nr_pages > 0)
940
		__this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGPGIN]);
941
	else {
942
		__this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGPGOUT]);
943 944
		nr_pages = -nr_pages; /* for event */
	}
945

946
	__this_cpu_add(memcg->stat->nr_page_events, nr_pages);
947 948
}

949
unsigned long
950
mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
951 952 953 954 955 956 957 958
{
	struct mem_cgroup_per_zone *mz;

	mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec);
	return mz->lru_size[lru];
}

static unsigned long
959
mem_cgroup_zone_nr_lru_pages(struct mem_cgroup *memcg, int nid, int zid,
960
			unsigned int lru_mask)
961 962
{
	struct mem_cgroup_per_zone *mz;
H
Hugh Dickins 已提交
963
	enum lru_list lru;
964 965
	unsigned long ret = 0;

966
	mz = mem_cgroup_zoneinfo(memcg, nid, zid);
967

H
Hugh Dickins 已提交
968 969 970
	for_each_lru(lru) {
		if (BIT(lru) & lru_mask)
			ret += mz->lru_size[lru];
971 972 973 974 975
	}
	return ret;
}

static unsigned long
976
mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
977 978
			int nid, unsigned int lru_mask)
{
979 980 981
	u64 total = 0;
	int zid;

982
	for (zid = 0; zid < MAX_NR_ZONES; zid++)
983 984
		total += mem_cgroup_zone_nr_lru_pages(memcg,
						nid, zid, lru_mask);
985

986 987
	return total;
}
988

989
static unsigned long mem_cgroup_nr_lru_pages(struct mem_cgroup *memcg,
990
			unsigned int lru_mask)
991
{
992
	int nid;
993 994
	u64 total = 0;

995
	for_each_node_state(nid, N_MEMORY)
996
		total += mem_cgroup_node_nr_lru_pages(memcg, nid, lru_mask);
997
	return total;
998 999
}

1000 1001
static bool mem_cgroup_event_ratelimit(struct mem_cgroup *memcg,
				       enum mem_cgroup_events_target target)
1002 1003 1004
{
	unsigned long val, next;

1005
	val = __this_cpu_read(memcg->stat->nr_page_events);
1006
	next = __this_cpu_read(memcg->stat->targets[target]);
1007
	/* from time_after() in jiffies.h */
1008 1009 1010 1011 1012
	if ((long)next - (long)val < 0) {
		switch (target) {
		case MEM_CGROUP_TARGET_THRESH:
			next = val + THRESHOLDS_EVENTS_TARGET;
			break;
1013 1014 1015
		case MEM_CGROUP_TARGET_SOFTLIMIT:
			next = val + SOFTLIMIT_EVENTS_TARGET;
			break;
1016 1017 1018 1019 1020 1021 1022 1023
		case MEM_CGROUP_TARGET_NUMAINFO:
			next = val + NUMAINFO_EVENTS_TARGET;
			break;
		default:
			break;
		}
		__this_cpu_write(memcg->stat->targets[target], next);
		return true;
1024
	}
1025
	return false;
1026 1027 1028 1029 1030 1031
}

/*
 * Check events in order.
 *
 */
1032
static void memcg_check_events(struct mem_cgroup *memcg, struct page *page)
1033
{
1034
	preempt_disable();
1035
	/* threshold event is triggered in finer grain than soft limit */
1036 1037
	if (unlikely(mem_cgroup_event_ratelimit(memcg,
						MEM_CGROUP_TARGET_THRESH))) {
1038
		bool do_softlimit;
1039
		bool do_numainfo __maybe_unused;
1040

1041 1042
		do_softlimit = mem_cgroup_event_ratelimit(memcg,
						MEM_CGROUP_TARGET_SOFTLIMIT);
1043 1044 1045 1046 1047 1048
#if MAX_NUMNODES > 1
		do_numainfo = mem_cgroup_event_ratelimit(memcg,
						MEM_CGROUP_TARGET_NUMAINFO);
#endif
		preempt_enable();

1049
		mem_cgroup_threshold(memcg);
1050 1051
		if (unlikely(do_softlimit))
			mem_cgroup_update_tree(memcg, page);
1052
#if MAX_NUMNODES > 1
1053
		if (unlikely(do_numainfo))
1054
			atomic_inc(&memcg->numainfo_events);
1055
#endif
1056 1057
	} else
		preempt_enable();
1058 1059
}

1060
struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p)
1061
{
1062 1063 1064 1065 1066 1067 1068 1069
	/*
	 * 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;

1070
	return mem_cgroup_from_css(task_css(p, memory_cgrp_id));
1071 1072
}

1073
static struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
1074
{
1075
	struct mem_cgroup *memcg = NULL;
1076

1077 1078
	rcu_read_lock();
	do {
1079 1080 1081 1082 1083 1084
		/*
		 * Page cache insertions can happen withou an
		 * actual mm context, e.g. during disk probing
		 * on boot, loopback IO, acct() writes etc.
		 */
		if (unlikely(!mm))
1085
			memcg = root_mem_cgroup;
1086 1087 1088 1089 1090
		else {
			memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
			if (unlikely(!memcg))
				memcg = root_mem_cgroup;
		}
1091
	} while (!css_tryget(&memcg->css));
1092
	rcu_read_unlock();
1093
	return memcg;
1094 1095
}

1096 1097 1098 1099 1100 1101 1102
/*
 * Returns a next (in a pre-order walk) alive memcg (with elevated css
 * ref. count) or NULL if the whole root's subtree has been visited.
 *
 * helper function to be used by mem_cgroup_iter
 */
static struct mem_cgroup *__mem_cgroup_iter_next(struct mem_cgroup *root,
1103
		struct mem_cgroup *last_visited)
1104
{
1105
	struct cgroup_subsys_state *prev_css, *next_css;
1106

1107
	prev_css = last_visited ? &last_visited->css : NULL;
1108
skip_node:
1109
	next_css = css_next_descendant_pre(prev_css, &root->css);
1110 1111 1112 1113 1114 1115 1116

	/*
	 * Even if we found a group we have to make sure it is
	 * alive. css && !memcg means that the groups should be
	 * skipped and we should continue the tree walk.
	 * last_visited css is safe to use because it is
	 * protected by css_get and the tree walk is rcu safe.
1117 1118 1119 1120 1121 1122 1123 1124
	 *
	 * We do not take a reference on the root of the tree walk
	 * because we might race with the root removal when it would
	 * be the only node in the iterated hierarchy and mem_cgroup_iter
	 * would end up in an endless loop because it expects that at
	 * least one valid node will be returned. Root cannot disappear
	 * because caller of the iterator should hold it already so
	 * skipping css reference should be safe.
1125
	 */
1126
	if (next_css) {
1127 1128
		if ((next_css == &root->css) ||
		    ((next_css->flags & CSS_ONLINE) && css_tryget(next_css)))
1129
			return mem_cgroup_from_css(next_css);
1130 1131 1132

		prev_css = next_css;
		goto skip_node;
1133 1134 1135 1136 1137
	}

	return NULL;
}

1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165
static void mem_cgroup_iter_invalidate(struct mem_cgroup *root)
{
	/*
	 * When a group in the hierarchy below root is destroyed, the
	 * hierarchy iterator can no longer be trusted since it might
	 * have pointed to the destroyed group.  Invalidate it.
	 */
	atomic_inc(&root->dead_count);
}

static struct mem_cgroup *
mem_cgroup_iter_load(struct mem_cgroup_reclaim_iter *iter,
		     struct mem_cgroup *root,
		     int *sequence)
{
	struct mem_cgroup *position = NULL;
	/*
	 * A cgroup destruction happens in two stages: offlining and
	 * release.  They are separated by a RCU grace period.
	 *
	 * If the iterator is valid, we may still race with an
	 * offlining.  The RCU lock ensures the object won't be
	 * released, tryget will fail if we lost the race.
	 */
	*sequence = atomic_read(&root->dead_count);
	if (iter->last_dead_count == *sequence) {
		smp_rmb();
		position = iter->last_visited;
1166 1167 1168 1169 1170 1171 1172 1173 1174

		/*
		 * We cannot take a reference to root because we might race
		 * with root removal and returning NULL would end up in
		 * an endless loop on the iterator user level when root
		 * would be returned all the time.
		 */
		if (position && position != root &&
				!css_tryget(&position->css))
1175 1176 1177 1178 1179 1180 1181 1182
			position = NULL;
	}
	return position;
}

static void mem_cgroup_iter_update(struct mem_cgroup_reclaim_iter *iter,
				   struct mem_cgroup *last_visited,
				   struct mem_cgroup *new_position,
1183
				   struct mem_cgroup *root,
1184 1185
				   int sequence)
{
1186 1187
	/* root reference counting symmetric to mem_cgroup_iter_load */
	if (last_visited && last_visited != root)
1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199
		css_put(&last_visited->css);
	/*
	 * We store the sequence count from the time @last_visited was
	 * loaded successfully instead of rereading it here so that we
	 * don't lose destruction events in between.  We could have
	 * raced with the destruction of @new_position after all.
	 */
	iter->last_visited = new_position;
	smp_wmb();
	iter->last_dead_count = sequence;
}

1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216
/**
 * mem_cgroup_iter - iterate over memory cgroup hierarchy
 * @root: hierarchy root
 * @prev: previously returned memcg, NULL on first invocation
 * @reclaim: cookie for shared reclaim walks, NULL for full walks
 *
 * Returns references to children of the hierarchy below @root, or
 * @root itself, or %NULL after a full round-trip.
 *
 * Caller must pass the return value in @prev on subsequent
 * invocations for reference counting, or use mem_cgroup_iter_break()
 * to cancel a hierarchy walk before the round-trip is complete.
 *
 * Reclaimers can specify a zone and a priority level in @reclaim to
 * divide up the memcgs in the hierarchy among all concurrent
 * reclaimers operating on the same zone and priority.
 */
1217
struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *root,
1218
				   struct mem_cgroup *prev,
1219
				   struct mem_cgroup_reclaim_cookie *reclaim)
K
KAMEZAWA Hiroyuki 已提交
1220
{
1221
	struct mem_cgroup *memcg = NULL;
1222
	struct mem_cgroup *last_visited = NULL;
1223

1224 1225
	if (mem_cgroup_disabled())
		return NULL;
1226

1227 1228
	if (!root)
		root = root_mem_cgroup;
K
KAMEZAWA Hiroyuki 已提交
1229

1230
	if (prev && !reclaim)
1231
		last_visited = prev;
K
KAMEZAWA Hiroyuki 已提交
1232

1233 1234
	if (!root->use_hierarchy && root != root_mem_cgroup) {
		if (prev)
1235
			goto out_css_put;
1236
		return root;
1237
	}
K
KAMEZAWA Hiroyuki 已提交
1238

1239
	rcu_read_lock();
1240
	while (!memcg) {
1241
		struct mem_cgroup_reclaim_iter *uninitialized_var(iter);
1242
		int uninitialized_var(seq);
1243

1244 1245 1246 1247 1248 1249 1250
		if (reclaim) {
			int nid = zone_to_nid(reclaim->zone);
			int zid = zone_idx(reclaim->zone);
			struct mem_cgroup_per_zone *mz;

			mz = mem_cgroup_zoneinfo(root, nid, zid);
			iter = &mz->reclaim_iter[reclaim->priority];
1251
			if (prev && reclaim->generation != iter->generation) {
M
Michal Hocko 已提交
1252
				iter->last_visited = NULL;
1253 1254
				goto out_unlock;
			}
M
Michal Hocko 已提交
1255

1256
			last_visited = mem_cgroup_iter_load(iter, root, &seq);
1257
		}
K
KAMEZAWA Hiroyuki 已提交
1258

1259
		memcg = __mem_cgroup_iter_next(root, last_visited);
K
KAMEZAWA Hiroyuki 已提交
1260

1261
		if (reclaim) {
1262 1263
			mem_cgroup_iter_update(iter, last_visited, memcg, root,
					seq);
1264

M
Michal Hocko 已提交
1265
			if (!memcg)
1266 1267 1268 1269
				iter->generation++;
			else if (!prev && memcg)
				reclaim->generation = iter->generation;
		}
1270

1271
		if (prev && !memcg)
1272
			goto out_unlock;
1273
	}
1274 1275
out_unlock:
	rcu_read_unlock();
1276 1277 1278 1279
out_css_put:
	if (prev && prev != root)
		css_put(&prev->css);

1280
	return memcg;
K
KAMEZAWA Hiroyuki 已提交
1281
}
K
KAMEZAWA Hiroyuki 已提交
1282

1283 1284 1285 1286 1287 1288 1289
/**
 * mem_cgroup_iter_break - abort a hierarchy walk prematurely
 * @root: hierarchy root
 * @prev: last visited hierarchy member as returned by mem_cgroup_iter()
 */
void mem_cgroup_iter_break(struct mem_cgroup *root,
			   struct mem_cgroup *prev)
1290 1291 1292 1293 1294 1295
{
	if (!root)
		root = root_mem_cgroup;
	if (prev && prev != root)
		css_put(&prev->css);
}
K
KAMEZAWA Hiroyuki 已提交
1296

1297 1298 1299 1300 1301 1302
/*
 * Iteration constructs for visiting all cgroups (under a tree).  If
 * loops are exited prematurely (break), mem_cgroup_iter_break() must
 * be used for reference counting.
 */
#define for_each_mem_cgroup_tree(iter, root)		\
1303
	for (iter = mem_cgroup_iter(root, NULL, NULL);	\
1304
	     iter != NULL;				\
1305
	     iter = mem_cgroup_iter(root, iter, NULL))
1306

1307
#define for_each_mem_cgroup(iter)			\
1308
	for (iter = mem_cgroup_iter(NULL, NULL, NULL);	\
1309
	     iter != NULL;				\
1310
	     iter = mem_cgroup_iter(NULL, iter, NULL))
K
KAMEZAWA Hiroyuki 已提交
1311

1312
void __mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
1313
{
1314
	struct mem_cgroup *memcg;
1315 1316

	rcu_read_lock();
1317 1318
	memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
	if (unlikely(!memcg))
1319 1320 1321 1322
		goto out;

	switch (idx) {
	case PGFAULT:
1323 1324 1325 1326
		this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]);
		break;
	case PGMAJFAULT:
		this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]);
1327 1328 1329 1330 1331 1332 1333
		break;
	default:
		BUG();
	}
out:
	rcu_read_unlock();
}
1334
EXPORT_SYMBOL(__mem_cgroup_count_vm_event);
1335

1336 1337 1338
/**
 * mem_cgroup_zone_lruvec - get the lru list vector for a zone and memcg
 * @zone: zone of the wanted lruvec
1339
 * @memcg: memcg of the wanted lruvec
1340 1341 1342 1343 1344 1345 1346 1347 1348
 *
 * Returns the lru list vector holding pages for the given @zone and
 * @mem.  This can be the global zone lruvec, if the memory controller
 * is disabled.
 */
struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
				      struct mem_cgroup *memcg)
{
	struct mem_cgroup_per_zone *mz;
1349
	struct lruvec *lruvec;
1350

1351 1352 1353 1354
	if (mem_cgroup_disabled()) {
		lruvec = &zone->lruvec;
		goto out;
	}
1355 1356

	mz = mem_cgroup_zoneinfo(memcg, zone_to_nid(zone), zone_idx(zone));
1357 1358 1359 1360 1361 1362 1363 1364 1365 1366
	lruvec = &mz->lruvec;
out:
	/*
	 * Since a node can be onlined after the mem_cgroup was created,
	 * we have to be prepared to initialize lruvec->zone here;
	 * and if offlined then reonlined, we need to reinitialize it.
	 */
	if (unlikely(lruvec->zone != zone))
		lruvec->zone = zone;
	return lruvec;
1367 1368
}

K
KAMEZAWA Hiroyuki 已提交
1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381
/*
 * 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.
 */
1382

1383
/**
1384
 * mem_cgroup_page_lruvec - return lruvec for adding an lru page
1385
 * @page: the page
1386
 * @zone: zone of the page
1387
 */
1388
struct lruvec *mem_cgroup_page_lruvec(struct page *page, struct zone *zone)
K
KAMEZAWA Hiroyuki 已提交
1389 1390
{
	struct mem_cgroup_per_zone *mz;
1391 1392
	struct mem_cgroup *memcg;
	struct page_cgroup *pc;
1393
	struct lruvec *lruvec;
1394

1395 1396 1397 1398
	if (mem_cgroup_disabled()) {
		lruvec = &zone->lruvec;
		goto out;
	}
1399

K
KAMEZAWA Hiroyuki 已提交
1400
	pc = lookup_page_cgroup(page);
1401
	memcg = pc->mem_cgroup;
1402 1403

	/*
1404
	 * Surreptitiously switch any uncharged offlist page to root:
1405 1406 1407 1408 1409 1410 1411
	 * an uncharged page off lru does nothing to secure
	 * its former mem_cgroup from sudden removal.
	 *
	 * Our caller holds lru_lock, and PageCgroupUsed is updated
	 * under page_cgroup lock: between them, they make all uses
	 * of pc->mem_cgroup safe.
	 */
1412
	if (!PageLRU(page) && !PageCgroupUsed(pc) && memcg != root_mem_cgroup)
1413 1414
		pc->mem_cgroup = memcg = root_mem_cgroup;

1415
	mz = page_cgroup_zoneinfo(memcg, page);
1416 1417 1418 1419 1420 1421 1422 1423 1424 1425
	lruvec = &mz->lruvec;
out:
	/*
	 * Since a node can be onlined after the mem_cgroup was created,
	 * we have to be prepared to initialize lruvec->zone here;
	 * and if offlined then reonlined, we need to reinitialize it.
	 */
	if (unlikely(lruvec->zone != zone))
		lruvec->zone = zone;
	return lruvec;
K
KAMEZAWA Hiroyuki 已提交
1426
}
1427

1428
/**
1429 1430 1431 1432
 * mem_cgroup_update_lru_size - account for adding or removing an lru page
 * @lruvec: mem_cgroup per zone lru vector
 * @lru: index of lru list the page is sitting on
 * @nr_pages: positive when adding or negative when removing
1433
 *
1434 1435
 * This function must be called when a page is added to or removed from an
 * lru list.
1436
 */
1437 1438
void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
				int nr_pages)
1439 1440
{
	struct mem_cgroup_per_zone *mz;
1441
	unsigned long *lru_size;
1442 1443 1444 1445

	if (mem_cgroup_disabled())
		return;

1446 1447 1448 1449
	mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec);
	lru_size = mz->lru_size + lru;
	*lru_size += nr_pages;
	VM_BUG_ON((long)(*lru_size) < 0);
K
KAMEZAWA Hiroyuki 已提交
1450
}
1451

1452
/*
1453
 * Checks whether given mem is same or in the root_mem_cgroup's
1454 1455
 * hierarchy subtree
 */
1456 1457
bool __mem_cgroup_same_or_subtree(const struct mem_cgroup *root_memcg,
				  struct mem_cgroup *memcg)
1458
{
1459 1460
	if (root_memcg == memcg)
		return true;
1461
	if (!root_memcg->use_hierarchy || !memcg)
1462
		return false;
1463
	return cgroup_is_descendant(memcg->css.cgroup, root_memcg->css.cgroup);
1464 1465 1466 1467 1468 1469 1470
}

static bool mem_cgroup_same_or_subtree(const struct mem_cgroup *root_memcg,
				       struct mem_cgroup *memcg)
{
	bool ret;

1471
	rcu_read_lock();
1472
	ret = __mem_cgroup_same_or_subtree(root_memcg, memcg);
1473 1474
	rcu_read_unlock();
	return ret;
1475 1476
}

1477 1478
bool task_in_mem_cgroup(struct task_struct *task,
			const struct mem_cgroup *memcg)
1479
{
1480
	struct mem_cgroup *curr = NULL;
1481
	struct task_struct *p;
1482
	bool ret;
1483

1484
	p = find_lock_task_mm(task);
1485
	if (p) {
1486
		curr = get_mem_cgroup_from_mm(p->mm);
1487 1488 1489 1490 1491 1492 1493
		task_unlock(p);
	} else {
		/*
		 * All threads may have already detached their mm's, but the oom
		 * killer still needs to detect if they have already been oom
		 * killed to prevent needlessly killing additional tasks.
		 */
1494
		rcu_read_lock();
1495 1496 1497
		curr = mem_cgroup_from_task(task);
		if (curr)
			css_get(&curr->css);
1498
		rcu_read_unlock();
1499
	}
1500
	/*
1501
	 * We should check use_hierarchy of "memcg" not "curr". Because checking
1502
	 * use_hierarchy of "curr" here make this function true if hierarchy is
1503 1504
	 * enabled in "curr" and "curr" is a child of "memcg" in *cgroup*
	 * hierarchy(even if use_hierarchy is disabled in "memcg").
1505
	 */
1506
	ret = mem_cgroup_same_or_subtree(memcg, curr);
1507
	css_put(&curr->css);
1508 1509 1510
	return ret;
}

1511
int mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
1512
{
1513
	unsigned long inactive_ratio;
1514
	unsigned long inactive;
1515
	unsigned long active;
1516
	unsigned long gb;
1517

1518 1519
	inactive = mem_cgroup_get_lru_size(lruvec, LRU_INACTIVE_ANON);
	active = mem_cgroup_get_lru_size(lruvec, LRU_ACTIVE_ANON);
1520

1521 1522 1523 1524 1525 1526
	gb = (inactive + active) >> (30 - PAGE_SHIFT);
	if (gb)
		inactive_ratio = int_sqrt(10 * gb);
	else
		inactive_ratio = 1;

1527
	return inactive * inactive_ratio < active;
1528 1529
}

1530 1531 1532
#define mem_cgroup_from_res_counter(counter, member)	\
	container_of(counter, struct mem_cgroup, member)

1533
/**
1534
 * mem_cgroup_margin - calculate chargeable space of a memory cgroup
W
Wanpeng Li 已提交
1535
 * @memcg: the memory cgroup
1536
 *
1537
 * Returns the maximum amount of memory @mem can be charged with, in
1538
 * pages.
1539
 */
1540
static unsigned long mem_cgroup_margin(struct mem_cgroup *memcg)
1541
{
1542 1543
	unsigned long long margin;

1544
	margin = res_counter_margin(&memcg->res);
1545
	if (do_swap_account)
1546
		margin = min(margin, res_counter_margin(&memcg->memsw));
1547
	return margin >> PAGE_SHIFT;
1548 1549
}

1550
int mem_cgroup_swappiness(struct mem_cgroup *memcg)
K
KOSAKI Motohiro 已提交
1551 1552
{
	/* root ? */
T
Tejun Heo 已提交
1553
	if (!css_parent(&memcg->css))
K
KOSAKI Motohiro 已提交
1554 1555
		return vm_swappiness;

1556
	return memcg->swappiness;
K
KOSAKI Motohiro 已提交
1557 1558
}

1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572
/*
 * memcg->moving_account is used for checking possibility that some thread is
 * calling move_account(). When a thread on CPU-A starts moving pages under
 * a memcg, other threads should check memcg->moving_account under
 * rcu_read_lock(), like this:
 *
 *         CPU-A                                    CPU-B
 *                                              rcu_read_lock()
 *         memcg->moving_account+1              if (memcg->mocing_account)
 *                                                   take heavy locks.
 *         synchronize_rcu()                    update something.
 *                                              rcu_read_unlock()
 *         start move here.
 */
1573 1574 1575 1576

/* for quick checking without looking up memcg */
atomic_t memcg_moving __read_mostly;

1577
static void mem_cgroup_start_move(struct mem_cgroup *memcg)
1578
{
1579
	atomic_inc(&memcg_moving);
1580
	atomic_inc(&memcg->moving_account);
1581 1582 1583
	synchronize_rcu();
}

1584
static void mem_cgroup_end_move(struct mem_cgroup *memcg)
1585
{
1586 1587 1588 1589
	/*
	 * Now, mem_cgroup_clear_mc() may call this function with NULL.
	 * We check NULL in callee rather than caller.
	 */
1590 1591
	if (memcg) {
		atomic_dec(&memcg_moving);
1592
		atomic_dec(&memcg->moving_account);
1593
	}
1594
}
1595

1596 1597 1598
/*
 * 2 routines for checking "mem" is under move_account() or not.
 *
1599 1600
 * mem_cgroup_stolen() -  checking whether a cgroup is mc.from or not. This
 *			  is used for avoiding races in accounting.  If true,
1601 1602 1603 1604 1605 1606 1607
 *			  pc->mem_cgroup may be overwritten.
 *
 * mem_cgroup_under_move() - checking a cgroup is mc.from or mc.to or
 *			  under hierarchy of moving cgroups. This is for
 *			  waiting at hith-memory prressure caused by "move".
 */

1608
static bool mem_cgroup_stolen(struct mem_cgroup *memcg)
1609 1610
{
	VM_BUG_ON(!rcu_read_lock_held());
1611
	return atomic_read(&memcg->moving_account) > 0;
1612
}
1613

1614
static bool mem_cgroup_under_move(struct mem_cgroup *memcg)
1615
{
1616 1617
	struct mem_cgroup *from;
	struct mem_cgroup *to;
1618
	bool ret = false;
1619 1620 1621 1622 1623 1624 1625 1626 1627
	/*
	 * Unlike task_move routines, we access mc.to, mc.from not under
	 * mutual exclusion by cgroup_mutex. Here, we take spinlock instead.
	 */
	spin_lock(&mc.lock);
	from = mc.from;
	to = mc.to;
	if (!from)
		goto unlock;
1628

1629 1630
	ret = mem_cgroup_same_or_subtree(memcg, from)
		|| mem_cgroup_same_or_subtree(memcg, to);
1631 1632
unlock:
	spin_unlock(&mc.lock);
1633 1634 1635
	return ret;
}

1636
static bool mem_cgroup_wait_acct_move(struct mem_cgroup *memcg)
1637 1638
{
	if (mc.moving_task && current != mc.moving_task) {
1639
		if (mem_cgroup_under_move(memcg)) {
1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651
			DEFINE_WAIT(wait);
			prepare_to_wait(&mc.waitq, &wait, TASK_INTERRUPTIBLE);
			/* moving charge context might have finished. */
			if (mc.moving_task)
				schedule();
			finish_wait(&mc.waitq, &wait);
			return true;
		}
	}
	return false;
}

1652 1653 1654 1655
/*
 * Take this lock when
 * - a code tries to modify page's memcg while it's USED.
 * - a code tries to modify page state accounting in a memcg.
1656
 * see mem_cgroup_stolen(), too.
1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669
 */
static void move_lock_mem_cgroup(struct mem_cgroup *memcg,
				  unsigned long *flags)
{
	spin_lock_irqsave(&memcg->move_lock, *flags);
}

static void move_unlock_mem_cgroup(struct mem_cgroup *memcg,
				unsigned long *flags)
{
	spin_unlock_irqrestore(&memcg->move_lock, *flags);
}

1670
#define K(x) ((x) << (PAGE_SHIFT-10))
1671
/**
1672
 * mem_cgroup_print_oom_info: Print OOM information relevant to memory controller.
1673 1674 1675 1676 1677 1678 1679 1680
 * @memcg: The memory cgroup that went over limit
 * @p: Task that is going to be killed
 *
 * NOTE: @memcg and @p's mem_cgroup can be different when hierarchy is
 * enabled
 */
void mem_cgroup_print_oom_info(struct mem_cgroup *memcg, struct task_struct *p)
{
T
Tejun Heo 已提交
1681
	/* oom_info_lock ensures that parallel ooms do not interleave */
1682
	static DEFINE_MUTEX(oom_info_lock);
1683 1684
	struct mem_cgroup *iter;
	unsigned int i;
1685

1686
	if (!p)
1687 1688
		return;

1689
	mutex_lock(&oom_info_lock);
1690 1691
	rcu_read_lock();

T
Tejun Heo 已提交
1692 1693 1694 1695 1696
	pr_info("Task in ");
	pr_cont_cgroup_path(task_cgroup(p, memory_cgrp_id));
	pr_info(" killed as a result of limit of ");
	pr_cont_cgroup_path(memcg->css.cgroup);
	pr_info("\n");
1697 1698 1699

	rcu_read_unlock();

1700
	pr_info("memory: usage %llukB, limit %llukB, failcnt %llu\n",
1701 1702 1703
		res_counter_read_u64(&memcg->res, RES_USAGE) >> 10,
		res_counter_read_u64(&memcg->res, RES_LIMIT) >> 10,
		res_counter_read_u64(&memcg->res, RES_FAILCNT));
1704
	pr_info("memory+swap: usage %llukB, limit %llukB, failcnt %llu\n",
1705 1706 1707
		res_counter_read_u64(&memcg->memsw, RES_USAGE) >> 10,
		res_counter_read_u64(&memcg->memsw, RES_LIMIT) >> 10,
		res_counter_read_u64(&memcg->memsw, RES_FAILCNT));
1708
	pr_info("kmem: usage %llukB, limit %llukB, failcnt %llu\n",
1709 1710 1711
		res_counter_read_u64(&memcg->kmem, RES_USAGE) >> 10,
		res_counter_read_u64(&memcg->kmem, RES_LIMIT) >> 10,
		res_counter_read_u64(&memcg->kmem, RES_FAILCNT));
1712 1713

	for_each_mem_cgroup_tree(iter, memcg) {
T
Tejun Heo 已提交
1714 1715
		pr_info("Memory cgroup stats for ");
		pr_cont_cgroup_path(iter->css.cgroup);
1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730
		pr_cont(":");

		for (i = 0; i < MEM_CGROUP_STAT_NSTATS; i++) {
			if (i == MEM_CGROUP_STAT_SWAP && !do_swap_account)
				continue;
			pr_cont(" %s:%ldKB", mem_cgroup_stat_names[i],
				K(mem_cgroup_read_stat(iter, i)));
		}

		for (i = 0; i < NR_LRU_LISTS; i++)
			pr_cont(" %s:%luKB", mem_cgroup_lru_names[i],
				K(mem_cgroup_nr_lru_pages(iter, BIT(i))));

		pr_cont("\n");
	}
1731
	mutex_unlock(&oom_info_lock);
1732 1733
}

1734 1735 1736 1737
/*
 * This function returns the number of memcg under hierarchy tree. Returns
 * 1(self count) if no children.
 */
1738
static int mem_cgroup_count_children(struct mem_cgroup *memcg)
1739 1740
{
	int num = 0;
K
KAMEZAWA Hiroyuki 已提交
1741 1742
	struct mem_cgroup *iter;

1743
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
1744
		num++;
1745 1746 1747
	return num;
}

D
David Rientjes 已提交
1748 1749 1750
/*
 * Return the memory (and swap, if configured) limit for a memcg.
 */
1751
static u64 mem_cgroup_get_limit(struct mem_cgroup *memcg)
D
David Rientjes 已提交
1752 1753 1754
{
	u64 limit;

1755 1756
	limit = res_counter_read_u64(&memcg->res, RES_LIMIT);

D
David Rientjes 已提交
1757
	/*
1758
	 * Do not consider swap space if we cannot swap due to swappiness
D
David Rientjes 已提交
1759
	 */
1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773
	if (mem_cgroup_swappiness(memcg)) {
		u64 memsw;

		limit += total_swap_pages << PAGE_SHIFT;
		memsw = res_counter_read_u64(&memcg->memsw, RES_LIMIT);

		/*
		 * If memsw is finite and limits the amount of swap space
		 * available to this memcg, return that limit.
		 */
		limit = min(limit, memsw);
	}

	return limit;
D
David Rientjes 已提交
1774 1775
}

1776 1777
static void mem_cgroup_out_of_memory(struct mem_cgroup *memcg, gfp_t gfp_mask,
				     int order)
1778 1779 1780 1781 1782 1783 1784
{
	struct mem_cgroup *iter;
	unsigned long chosen_points = 0;
	unsigned long totalpages;
	unsigned int points = 0;
	struct task_struct *chosen = NULL;

1785
	/*
1786 1787 1788
	 * If current has a pending SIGKILL or is exiting, then automatically
	 * select it.  The goal is to allow it to allocate so that it may
	 * quickly exit and free its memory.
1789
	 */
1790
	if (fatal_signal_pending(current) || current->flags & PF_EXITING) {
1791 1792 1793 1794 1795
		set_thread_flag(TIF_MEMDIE);
		return;
	}

	check_panic_on_oom(CONSTRAINT_MEMCG, gfp_mask, order, NULL);
1796 1797
	totalpages = mem_cgroup_get_limit(memcg) >> PAGE_SHIFT ? : 1;
	for_each_mem_cgroup_tree(iter, memcg) {
1798
		struct css_task_iter it;
1799 1800
		struct task_struct *task;

1801 1802
		css_task_iter_start(&iter->css, &it);
		while ((task = css_task_iter_next(&it))) {
1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814
			switch (oom_scan_process_thread(task, totalpages, NULL,
							false)) {
			case OOM_SCAN_SELECT:
				if (chosen)
					put_task_struct(chosen);
				chosen = task;
				chosen_points = ULONG_MAX;
				get_task_struct(chosen);
				/* fall through */
			case OOM_SCAN_CONTINUE:
				continue;
			case OOM_SCAN_ABORT:
1815
				css_task_iter_end(&it);
1816 1817 1818 1819 1820 1821 1822 1823
				mem_cgroup_iter_break(memcg, iter);
				if (chosen)
					put_task_struct(chosen);
				return;
			case OOM_SCAN_OK:
				break;
			};
			points = oom_badness(task, memcg, NULL, totalpages);
1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835
			if (!points || points < chosen_points)
				continue;
			/* Prefer thread group leaders for display purposes */
			if (points == chosen_points &&
			    thread_group_leader(chosen))
				continue;

			if (chosen)
				put_task_struct(chosen);
			chosen = task;
			chosen_points = points;
			get_task_struct(chosen);
1836
		}
1837
		css_task_iter_end(&it);
1838 1839 1840 1841 1842 1843 1844 1845 1846
	}

	if (!chosen)
		return;
	points = chosen_points * 1000 / totalpages;
	oom_kill_process(chosen, gfp_mask, order, points, totalpages, memcg,
			 NULL, "Memory cgroup out of memory");
}

1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882
static unsigned long mem_cgroup_reclaim(struct mem_cgroup *memcg,
					gfp_t gfp_mask,
					unsigned long flags)
{
	unsigned long total = 0;
	bool noswap = false;
	int loop;

	if (flags & MEM_CGROUP_RECLAIM_NOSWAP)
		noswap = true;
	if (!(flags & MEM_CGROUP_RECLAIM_SHRINK) && memcg->memsw_is_minimum)
		noswap = true;

	for (loop = 0; loop < MEM_CGROUP_MAX_RECLAIM_LOOPS; loop++) {
		if (loop)
			drain_all_stock_async(memcg);
		total += try_to_free_mem_cgroup_pages(memcg, gfp_mask, noswap);
		/*
		 * Allow limit shrinkers, which are triggered directly
		 * by userspace, to catch signals and stop reclaim
		 * after minimal progress, regardless of the margin.
		 */
		if (total && (flags & MEM_CGROUP_RECLAIM_SHRINK))
			break;
		if (mem_cgroup_margin(memcg))
			break;
		/*
		 * If nothing was reclaimed after two attempts, there
		 * may be no reclaimable pages in this hierarchy.
		 */
		if (loop && !total)
			break;
	}
	return total;
}

1883 1884
/**
 * test_mem_cgroup_node_reclaimable
W
Wanpeng Li 已提交
1885
 * @memcg: the target memcg
1886 1887 1888 1889 1890 1891 1892
 * @nid: the node ID to be checked.
 * @noswap : specify true here if the user wants flle only information.
 *
 * This function returns whether the specified memcg contains any
 * reclaimable pages on a node. Returns true if there are any reclaimable
 * pages in the node.
 */
1893
static bool test_mem_cgroup_node_reclaimable(struct mem_cgroup *memcg,
1894 1895
		int nid, bool noswap)
{
1896
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_FILE))
1897 1898 1899
		return true;
	if (noswap || !total_swap_pages)
		return false;
1900
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_ANON))
1901 1902 1903 1904
		return true;
	return false;

}
1905
#if MAX_NUMNODES > 1
1906 1907 1908 1909 1910 1911 1912

/*
 * Always updating the nodemask is not very good - even if we have an empty
 * list or the wrong list here, we can start from some node and traverse all
 * nodes based on the zonelist. So update the list loosely once per 10 secs.
 *
 */
1913
static void mem_cgroup_may_update_nodemask(struct mem_cgroup *memcg)
1914 1915
{
	int nid;
1916 1917 1918 1919
	/*
	 * numainfo_events > 0 means there was at least NUMAINFO_EVENTS_TARGET
	 * pagein/pageout changes since the last update.
	 */
1920
	if (!atomic_read(&memcg->numainfo_events))
1921
		return;
1922
	if (atomic_inc_return(&memcg->numainfo_updating) > 1)
1923 1924 1925
		return;

	/* make a nodemask where this memcg uses memory from */
1926
	memcg->scan_nodes = node_states[N_MEMORY];
1927

1928
	for_each_node_mask(nid, node_states[N_MEMORY]) {
1929

1930 1931
		if (!test_mem_cgroup_node_reclaimable(memcg, nid, false))
			node_clear(nid, memcg->scan_nodes);
1932
	}
1933

1934 1935
	atomic_set(&memcg->numainfo_events, 0);
	atomic_set(&memcg->numainfo_updating, 0);
1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949
}

/*
 * Selecting a node where we start reclaim from. Because what we need is just
 * reducing usage counter, start from anywhere is O,K. Considering
 * memory reclaim from current node, there are pros. and cons.
 *
 * Freeing memory from current node means freeing memory from a node which
 * we'll use or we've used. So, it may make LRU bad. And if several threads
 * hit limits, it will see a contention on a node. But freeing from remote
 * node means more costs for memory reclaim because of memory latency.
 *
 * Now, we use round-robin. Better algorithm is welcomed.
 */
1950
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
1951 1952 1953
{
	int node;

1954 1955
	mem_cgroup_may_update_nodemask(memcg);
	node = memcg->last_scanned_node;
1956

1957
	node = next_node(node, memcg->scan_nodes);
1958
	if (node == MAX_NUMNODES)
1959
		node = first_node(memcg->scan_nodes);
1960 1961 1962 1963 1964 1965 1966 1967 1968
	/*
	 * We call this when we hit limit, not when pages are added to LRU.
	 * No LRU may hold pages because all pages are UNEVICTABLE or
	 * memcg is too small and all pages are not on LRU. In that case,
	 * we use curret node.
	 */
	if (unlikely(node == MAX_NUMNODES))
		node = numa_node_id();

1969
	memcg->last_scanned_node = node;
1970 1971 1972
	return node;
}

1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
/*
 * Check all nodes whether it contains reclaimable pages or not.
 * For quick scan, we make use of scan_nodes. This will allow us to skip
 * unused nodes. But scan_nodes is lazily updated and may not cotain
 * enough new information. We need to do double check.
 */
static bool mem_cgroup_reclaimable(struct mem_cgroup *memcg, bool noswap)
{
	int nid;

	/*
	 * quick check...making use of scan_node.
	 * We can skip unused nodes.
	 */
	if (!nodes_empty(memcg->scan_nodes)) {
		for (nid = first_node(memcg->scan_nodes);
		     nid < MAX_NUMNODES;
		     nid = next_node(nid, memcg->scan_nodes)) {

			if (test_mem_cgroup_node_reclaimable(memcg, nid, noswap))
				return true;
		}
	}
	/*
	 * Check rest of nodes.
	 */
	for_each_node_state(nid, N_MEMORY) {
		if (node_isset(nid, memcg->scan_nodes))
			continue;
		if (test_mem_cgroup_node_reclaimable(memcg, nid, noswap))
			return true;
	}
	return false;
}

2008
#else
2009
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
2010 2011 2012
{
	return 0;
}
2013

2014 2015 2016 2017
static bool mem_cgroup_reclaimable(struct mem_cgroup *memcg, bool noswap)
{
	return test_mem_cgroup_node_reclaimable(memcg, 0, noswap);
}
2018 2019
#endif

2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067
static int mem_cgroup_soft_reclaim(struct mem_cgroup *root_memcg,
				   struct zone *zone,
				   gfp_t gfp_mask,
				   unsigned long *total_scanned)
{
	struct mem_cgroup *victim = NULL;
	int total = 0;
	int loop = 0;
	unsigned long excess;
	unsigned long nr_scanned;
	struct mem_cgroup_reclaim_cookie reclaim = {
		.zone = zone,
		.priority = 0,
	};

	excess = res_counter_soft_limit_excess(&root_memcg->res) >> PAGE_SHIFT;

	while (1) {
		victim = mem_cgroup_iter(root_memcg, victim, &reclaim);
		if (!victim) {
			loop++;
			if (loop >= 2) {
				/*
				 * If we have not been able to reclaim
				 * anything, it might because there are
				 * no reclaimable pages under this hierarchy
				 */
				if (!total)
					break;
				/*
				 * We want to do more targeted reclaim.
				 * excess >> 2 is not to excessive so as to
				 * reclaim too much, nor too less that we keep
				 * coming back to reclaim from this cgroup
				 */
				if (total >= (excess >> 2) ||
					(loop > MEM_CGROUP_MAX_RECLAIM_LOOPS))
					break;
			}
			continue;
		}
		if (!mem_cgroup_reclaimable(victim, false))
			continue;
		total += mem_cgroup_shrink_node_zone(victim, gfp_mask, false,
						     zone, &nr_scanned);
		*total_scanned += nr_scanned;
		if (!res_counter_soft_limit_excess(&root_memcg->res))
			break;
2068
	}
2069 2070
	mem_cgroup_iter_break(root_memcg, victim);
	return total;
2071 2072
}

2073 2074 2075 2076 2077 2078
#ifdef CONFIG_LOCKDEP
static struct lockdep_map memcg_oom_lock_dep_map = {
	.name = "memcg_oom_lock",
};
#endif

2079 2080
static DEFINE_SPINLOCK(memcg_oom_lock);

K
KAMEZAWA Hiroyuki 已提交
2081 2082 2083 2084
/*
 * Check OOM-Killer is already running under our hierarchy.
 * If someone is running, return false.
 */
2085
static bool mem_cgroup_oom_trylock(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
2086
{
2087
	struct mem_cgroup *iter, *failed = NULL;
2088

2089 2090
	spin_lock(&memcg_oom_lock);

2091
	for_each_mem_cgroup_tree(iter, memcg) {
2092
		if (iter->oom_lock) {
2093 2094 2095 2096 2097
			/*
			 * this subtree of our hierarchy is already locked
			 * so we cannot give a lock.
			 */
			failed = iter;
2098 2099
			mem_cgroup_iter_break(memcg, iter);
			break;
2100 2101
		} else
			iter->oom_lock = true;
K
KAMEZAWA Hiroyuki 已提交
2102
	}
K
KAMEZAWA Hiroyuki 已提交
2103

2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114
	if (failed) {
		/*
		 * OK, we failed to lock the whole subtree so we have
		 * to clean up what we set up to the failing subtree
		 */
		for_each_mem_cgroup_tree(iter, memcg) {
			if (iter == failed) {
				mem_cgroup_iter_break(memcg, iter);
				break;
			}
			iter->oom_lock = false;
2115
		}
2116 2117
	} else
		mutex_acquire(&memcg_oom_lock_dep_map, 0, 1, _RET_IP_);
2118 2119 2120 2121

	spin_unlock(&memcg_oom_lock);

	return !failed;
2122
}
2123

2124
static void mem_cgroup_oom_unlock(struct mem_cgroup *memcg)
2125
{
K
KAMEZAWA Hiroyuki 已提交
2126 2127
	struct mem_cgroup *iter;

2128
	spin_lock(&memcg_oom_lock);
2129
	mutex_release(&memcg_oom_lock_dep_map, 1, _RET_IP_);
2130
	for_each_mem_cgroup_tree(iter, memcg)
2131
		iter->oom_lock = false;
2132
	spin_unlock(&memcg_oom_lock);
2133 2134
}

2135
static void mem_cgroup_mark_under_oom(struct mem_cgroup *memcg)
2136 2137 2138
{
	struct mem_cgroup *iter;

2139
	for_each_mem_cgroup_tree(iter, memcg)
2140 2141 2142
		atomic_inc(&iter->under_oom);
}

2143
static void mem_cgroup_unmark_under_oom(struct mem_cgroup *memcg)
2144 2145 2146
{
	struct mem_cgroup *iter;

K
KAMEZAWA Hiroyuki 已提交
2147 2148 2149 2150 2151
	/*
	 * When a new child is created while the hierarchy is under oom,
	 * mem_cgroup_oom_lock() may not be called. We have to use
	 * atomic_add_unless() here.
	 */
2152
	for_each_mem_cgroup_tree(iter, memcg)
2153
		atomic_add_unless(&iter->under_oom, -1, 0);
2154 2155
}

K
KAMEZAWA Hiroyuki 已提交
2156 2157
static DECLARE_WAIT_QUEUE_HEAD(memcg_oom_waitq);

K
KAMEZAWA Hiroyuki 已提交
2158
struct oom_wait_info {
2159
	struct mem_cgroup *memcg;
K
KAMEZAWA Hiroyuki 已提交
2160 2161 2162 2163 2164 2165
	wait_queue_t	wait;
};

static int memcg_oom_wake_function(wait_queue_t *wait,
	unsigned mode, int sync, void *arg)
{
2166 2167
	struct mem_cgroup *wake_memcg = (struct mem_cgroup *)arg;
	struct mem_cgroup *oom_wait_memcg;
K
KAMEZAWA Hiroyuki 已提交
2168 2169 2170
	struct oom_wait_info *oom_wait_info;

	oom_wait_info = container_of(wait, struct oom_wait_info, wait);
2171
	oom_wait_memcg = oom_wait_info->memcg;
K
KAMEZAWA Hiroyuki 已提交
2172 2173

	/*
2174
	 * Both of oom_wait_info->memcg and wake_memcg are stable under us.
K
KAMEZAWA Hiroyuki 已提交
2175 2176
	 * Then we can use css_is_ancestor without taking care of RCU.
	 */
2177 2178
	if (!mem_cgroup_same_or_subtree(oom_wait_memcg, wake_memcg)
		&& !mem_cgroup_same_or_subtree(wake_memcg, oom_wait_memcg))
K
KAMEZAWA Hiroyuki 已提交
2179 2180 2181 2182
		return 0;
	return autoremove_wake_function(wait, mode, sync, arg);
}

2183
static void memcg_wakeup_oom(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
2184
{
2185
	atomic_inc(&memcg->oom_wakeups);
2186 2187
	/* for filtering, pass "memcg" as argument. */
	__wake_up(&memcg_oom_waitq, TASK_NORMAL, 0, memcg);
K
KAMEZAWA Hiroyuki 已提交
2188 2189
}

2190
static void memcg_oom_recover(struct mem_cgroup *memcg)
2191
{
2192 2193
	if (memcg && atomic_read(&memcg->under_oom))
		memcg_wakeup_oom(memcg);
2194 2195
}

2196
static void mem_cgroup_oom(struct mem_cgroup *memcg, gfp_t mask, int order)
2197
{
2198 2199
	if (!current->memcg_oom.may_oom)
		return;
K
KAMEZAWA Hiroyuki 已提交
2200
	/*
2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212
	 * We are in the middle of the charge context here, so we
	 * don't want to block when potentially sitting on a callstack
	 * that holds all kinds of filesystem and mm locks.
	 *
	 * Also, the caller may handle a failed allocation gracefully
	 * (like optional page cache readahead) and so an OOM killer
	 * invocation might not even be necessary.
	 *
	 * That's why we don't do anything here except remember the
	 * OOM context and then deal with it at the end of the page
	 * fault when the stack is unwound, the locks are released,
	 * and when we know whether the fault was overall successful.
K
KAMEZAWA Hiroyuki 已提交
2213
	 */
2214 2215 2216 2217
	css_get(&memcg->css);
	current->memcg_oom.memcg = memcg;
	current->memcg_oom.gfp_mask = mask;
	current->memcg_oom.order = order;
2218 2219 2220 2221
}

/**
 * mem_cgroup_oom_synchronize - complete memcg OOM handling
2222
 * @handle: actually kill/wait or just clean up the OOM state
2223
 *
2224 2225
 * This has to be called at the end of a page fault if the memcg OOM
 * handler was enabled.
2226
 *
2227
 * Memcg supports userspace OOM handling where failed allocations must
2228 2229 2230 2231
 * sleep on a waitqueue until the userspace task resolves the
 * situation.  Sleeping directly in the charge context with all kinds
 * of locks held is not a good idea, instead we remember an OOM state
 * in the task and mem_cgroup_oom_synchronize() has to be called at
2232
 * the end of the page fault to complete the OOM handling.
2233 2234
 *
 * Returns %true if an ongoing memcg OOM situation was detected and
2235
 * completed, %false otherwise.
2236
 */
2237
bool mem_cgroup_oom_synchronize(bool handle)
2238
{
2239
	struct mem_cgroup *memcg = current->memcg_oom.memcg;
2240
	struct oom_wait_info owait;
2241
	bool locked;
2242 2243 2244

	/* OOM is global, do not handle */
	if (!memcg)
2245
		return false;
2246

2247 2248
	if (!handle)
		goto cleanup;
2249 2250 2251 2252 2253 2254

	owait.memcg = memcg;
	owait.wait.flags = 0;
	owait.wait.func = memcg_oom_wake_function;
	owait.wait.private = current;
	INIT_LIST_HEAD(&owait.wait.task_list);
K
KAMEZAWA Hiroyuki 已提交
2255

2256
	prepare_to_wait(&memcg_oom_waitq, &owait.wait, TASK_KILLABLE);
2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269
	mem_cgroup_mark_under_oom(memcg);

	locked = mem_cgroup_oom_trylock(memcg);

	if (locked)
		mem_cgroup_oom_notify(memcg);

	if (locked && !memcg->oom_kill_disable) {
		mem_cgroup_unmark_under_oom(memcg);
		finish_wait(&memcg_oom_waitq, &owait.wait);
		mem_cgroup_out_of_memory(memcg, current->memcg_oom.gfp_mask,
					 current->memcg_oom.order);
	} else {
2270
		schedule();
2271 2272 2273 2274 2275
		mem_cgroup_unmark_under_oom(memcg);
		finish_wait(&memcg_oom_waitq, &owait.wait);
	}

	if (locked) {
2276 2277 2278 2279 2280 2281 2282 2283
		mem_cgroup_oom_unlock(memcg);
		/*
		 * There is no guarantee that an OOM-lock contender
		 * sees the wakeups triggered by the OOM kill
		 * uncharges.  Wake any sleepers explicitely.
		 */
		memcg_oom_recover(memcg);
	}
2284 2285
cleanup:
	current->memcg_oom.memcg = NULL;
2286
	css_put(&memcg->css);
K
KAMEZAWA Hiroyuki 已提交
2287
	return true;
2288 2289
}

2290 2291 2292
/*
 * Currently used to update mapped file statistics, but the routine can be
 * generalized to update other statistics as well.
2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309
 *
 * Notes: Race condition
 *
 * We usually use page_cgroup_lock() for accessing page_cgroup member but
 * it tends to be costly. But considering some conditions, we doesn't need
 * to do so _always_.
 *
 * Considering "charge", lock_page_cgroup() is not required because all
 * file-stat operations happen after a page is attached to radix-tree. There
 * are no race with "charge".
 *
 * Considering "uncharge", we know that memcg doesn't clear pc->mem_cgroup
 * at "uncharge" intentionally. So, we always see valid pc->mem_cgroup even
 * if there are race with "uncharge". Statistics itself is properly handled
 * by flags.
 *
 * Considering "move", this is an only case we see a race. To make the race
2310 2311
 * small, we check mm->moving_account and detect there are possibility of race
 * If there is, we take a lock.
2312
 */
2313

2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326
void __mem_cgroup_begin_update_page_stat(struct page *page,
				bool *locked, unsigned long *flags)
{
	struct mem_cgroup *memcg;
	struct page_cgroup *pc;

	pc = lookup_page_cgroup(page);
again:
	memcg = pc->mem_cgroup;
	if (unlikely(!memcg || !PageCgroupUsed(pc)))
		return;
	/*
	 * If this memory cgroup is not under account moving, we don't
2327
	 * need to take move_lock_mem_cgroup(). Because we already hold
2328
	 * rcu_read_lock(), any calls to move_account will be delayed until
2329
	 * rcu_read_unlock() if mem_cgroup_stolen() == true.
2330
	 */
2331
	if (!mem_cgroup_stolen(memcg))
2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348
		return;

	move_lock_mem_cgroup(memcg, flags);
	if (memcg != pc->mem_cgroup || !PageCgroupUsed(pc)) {
		move_unlock_mem_cgroup(memcg, flags);
		goto again;
	}
	*locked = true;
}

void __mem_cgroup_end_update_page_stat(struct page *page, unsigned long *flags)
{
	struct page_cgroup *pc = lookup_page_cgroup(page);

	/*
	 * It's guaranteed that pc->mem_cgroup never changes while
	 * lock is held because a routine modifies pc->mem_cgroup
2349
	 * should take move_lock_mem_cgroup().
2350 2351 2352 2353
	 */
	move_unlock_mem_cgroup(pc->mem_cgroup, flags);
}

2354
void mem_cgroup_update_page_stat(struct page *page,
S
Sha Zhengju 已提交
2355
				 enum mem_cgroup_stat_index idx, int val)
2356
{
2357
	struct mem_cgroup *memcg;
2358
	struct page_cgroup *pc = lookup_page_cgroup(page);
2359
	unsigned long uninitialized_var(flags);
2360

2361
	if (mem_cgroup_disabled())
2362
		return;
2363

2364
	VM_BUG_ON(!rcu_read_lock_held());
2365 2366
	memcg = pc->mem_cgroup;
	if (unlikely(!memcg || !PageCgroupUsed(pc)))
2367
		return;
2368

2369
	this_cpu_add(memcg->stat->count[idx], val);
2370
}
2371

2372 2373 2374 2375
/*
 * size of first charge trial. "32" comes from vmscan.c's magic value.
 * TODO: maybe necessary to use big numbers in big irons.
 */
2376
#define CHARGE_BATCH	32U
2377 2378
struct memcg_stock_pcp {
	struct mem_cgroup *cached; /* this never be root cgroup */
2379
	unsigned int nr_pages;
2380
	struct work_struct work;
2381
	unsigned long flags;
2382
#define FLUSHING_CACHED_CHARGE	0
2383 2384
};
static DEFINE_PER_CPU(struct memcg_stock_pcp, memcg_stock);
2385
static DEFINE_MUTEX(percpu_charge_mutex);
2386

2387 2388 2389 2390 2391 2392 2393 2394 2395 2396
/**
 * consume_stock: Try to consume stocked charge on this cpu.
 * @memcg: memcg to consume from.
 * @nr_pages: how many pages to charge.
 *
 * The charges will only happen if @memcg matches the current cpu's memcg
 * stock, and at least @nr_pages are available in that stock.  Failure to
 * service an allocation will refill the stock.
 *
 * returns true if successful, false otherwise.
2397
 */
2398
static bool consume_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
2399 2400 2401 2402
{
	struct memcg_stock_pcp *stock;
	bool ret = true;

2403 2404 2405
	if (nr_pages > CHARGE_BATCH)
		return false;

2406
	stock = &get_cpu_var(memcg_stock);
2407 2408
	if (memcg == stock->cached && stock->nr_pages >= nr_pages)
		stock->nr_pages -= nr_pages;
2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421
	else /* need to call res_counter_charge */
		ret = false;
	put_cpu_var(memcg_stock);
	return ret;
}

/*
 * Returns stocks cached in percpu to res_counter and reset cached information.
 */
static void drain_stock(struct memcg_stock_pcp *stock)
{
	struct mem_cgroup *old = stock->cached;

2422 2423 2424 2425
	if (stock->nr_pages) {
		unsigned long bytes = stock->nr_pages * PAGE_SIZE;

		res_counter_uncharge(&old->res, bytes);
2426
		if (do_swap_account)
2427 2428
			res_counter_uncharge(&old->memsw, bytes);
		stock->nr_pages = 0;
2429 2430 2431 2432 2433 2434 2435 2436 2437 2438
	}
	stock->cached = NULL;
}

/*
 * This must be called under preempt disabled or must be called by
 * a thread which is pinned to local cpu.
 */
static void drain_local_stock(struct work_struct *dummy)
{
2439
	struct memcg_stock_pcp *stock = this_cpu_ptr(&memcg_stock);
2440
	drain_stock(stock);
2441
	clear_bit(FLUSHING_CACHED_CHARGE, &stock->flags);
2442 2443
}

2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454
static void __init memcg_stock_init(void)
{
	int cpu;

	for_each_possible_cpu(cpu) {
		struct memcg_stock_pcp *stock =
					&per_cpu(memcg_stock, cpu);
		INIT_WORK(&stock->work, drain_local_stock);
	}
}

2455 2456
/*
 * Cache charges(val) which is from res_counter, to local per_cpu area.
2457
 * This will be consumed by consume_stock() function, later.
2458
 */
2459
static void refill_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
2460 2461 2462
{
	struct memcg_stock_pcp *stock = &get_cpu_var(memcg_stock);

2463
	if (stock->cached != memcg) { /* reset if necessary */
2464
		drain_stock(stock);
2465
		stock->cached = memcg;
2466
	}
2467
	stock->nr_pages += nr_pages;
2468 2469 2470 2471
	put_cpu_var(memcg_stock);
}

/*
2472
 * Drains all per-CPU charge caches for given root_memcg resp. subtree
2473 2474
 * of the hierarchy under it. sync flag says whether we should block
 * until the work is done.
2475
 */
2476
static void drain_all_stock(struct mem_cgroup *root_memcg, bool sync)
2477
{
2478
	int cpu, curcpu;
2479

2480 2481
	/* Notify other cpus that system-wide "drain" is running */
	get_online_cpus();
2482
	curcpu = get_cpu();
2483 2484
	for_each_online_cpu(cpu) {
		struct memcg_stock_pcp *stock = &per_cpu(memcg_stock, cpu);
2485
		struct mem_cgroup *memcg;
2486

2487 2488
		memcg = stock->cached;
		if (!memcg || !stock->nr_pages)
2489
			continue;
2490
		if (!mem_cgroup_same_or_subtree(root_memcg, memcg))
2491
			continue;
2492 2493 2494 2495 2496 2497
		if (!test_and_set_bit(FLUSHING_CACHED_CHARGE, &stock->flags)) {
			if (cpu == curcpu)
				drain_local_stock(&stock->work);
			else
				schedule_work_on(cpu, &stock->work);
		}
2498
	}
2499
	put_cpu();
2500 2501 2502 2503 2504 2505

	if (!sync)
		goto out;

	for_each_online_cpu(cpu) {
		struct memcg_stock_pcp *stock = &per_cpu(memcg_stock, cpu);
2506
		if (test_bit(FLUSHING_CACHED_CHARGE, &stock->flags))
2507 2508 2509
			flush_work(&stock->work);
	}
out:
A
Andrew Morton 已提交
2510
	put_online_cpus();
2511 2512 2513 2514 2515 2516 2517 2518
}

/*
 * Tries to drain stocked charges in other cpus. This function is asynchronous
 * and just put a work per cpu for draining localy on each cpu. Caller can
 * expects some charges will be back to res_counter later but cannot wait for
 * it.
 */
2519
static void drain_all_stock_async(struct mem_cgroup *root_memcg)
2520
{
2521 2522 2523 2524 2525
	/*
	 * If someone calls draining, avoid adding more kworker runs.
	 */
	if (!mutex_trylock(&percpu_charge_mutex))
		return;
2526
	drain_all_stock(root_memcg, false);
2527
	mutex_unlock(&percpu_charge_mutex);
2528 2529 2530
}

/* This is a synchronous drain interface. */
2531
static void drain_all_stock_sync(struct mem_cgroup *root_memcg)
2532 2533
{
	/* called when force_empty is called */
2534
	mutex_lock(&percpu_charge_mutex);
2535
	drain_all_stock(root_memcg, true);
2536
	mutex_unlock(&percpu_charge_mutex);
2537 2538
}

2539 2540 2541 2542
/*
 * This function drains percpu counter value from DEAD cpu and
 * move it to local cpu. Note that this function can be preempted.
 */
2543
static void mem_cgroup_drain_pcp_counter(struct mem_cgroup *memcg, int cpu)
2544 2545 2546
{
	int i;

2547
	spin_lock(&memcg->pcp_counter_lock);
2548
	for (i = 0; i < MEM_CGROUP_STAT_NSTATS; i++) {
2549
		long x = per_cpu(memcg->stat->count[i], cpu);
2550

2551 2552
		per_cpu(memcg->stat->count[i], cpu) = 0;
		memcg->nocpu_base.count[i] += x;
2553
	}
2554
	for (i = 0; i < MEM_CGROUP_EVENTS_NSTATS; i++) {
2555
		unsigned long x = per_cpu(memcg->stat->events[i], cpu);
2556

2557 2558
		per_cpu(memcg->stat->events[i], cpu) = 0;
		memcg->nocpu_base.events[i] += x;
2559
	}
2560
	spin_unlock(&memcg->pcp_counter_lock);
2561 2562
}

2563
static int memcg_cpu_hotplug_callback(struct notifier_block *nb,
2564 2565 2566 2567 2568
					unsigned long action,
					void *hcpu)
{
	int cpu = (unsigned long)hcpu;
	struct memcg_stock_pcp *stock;
2569
	struct mem_cgroup *iter;
2570

2571
	if (action == CPU_ONLINE)
2572 2573
		return NOTIFY_OK;

2574
	if (action != CPU_DEAD && action != CPU_DEAD_FROZEN)
2575
		return NOTIFY_OK;
2576

2577
	for_each_mem_cgroup(iter)
2578 2579
		mem_cgroup_drain_pcp_counter(iter, cpu);

2580 2581 2582 2583 2584
	stock = &per_cpu(memcg_stock, cpu);
	drain_stock(stock);
	return NOTIFY_OK;
}

2585

2586
/* See mem_cgroup_try_charge() for details */
2587 2588 2589 2590 2591 2592 2593
enum {
	CHARGE_OK,		/* success */
	CHARGE_RETRY,		/* need to retry but retry is not bad */
	CHARGE_NOMEM,		/* we can't do more. return -ENOMEM */
	CHARGE_WOULDBLOCK,	/* GFP_WAIT wasn't set and no enough res. */
};

2594
static int mem_cgroup_do_charge(struct mem_cgroup *memcg, gfp_t gfp_mask,
2595
				unsigned int nr_pages, unsigned int min_pages,
2596
				bool invoke_oom)
2597
{
2598
	unsigned long csize = nr_pages * PAGE_SIZE;
2599 2600 2601 2602 2603
	struct mem_cgroup *mem_over_limit;
	struct res_counter *fail_res;
	unsigned long flags = 0;
	int ret;

2604
	ret = res_counter_charge(&memcg->res, csize, &fail_res);
2605 2606 2607 2608

	if (likely(!ret)) {
		if (!do_swap_account)
			return CHARGE_OK;
2609
		ret = res_counter_charge(&memcg->memsw, csize, &fail_res);
2610 2611 2612
		if (likely(!ret))
			return CHARGE_OK;

2613
		res_counter_uncharge(&memcg->res, csize);
2614 2615 2616 2617
		mem_over_limit = mem_cgroup_from_res_counter(fail_res, memsw);
		flags |= MEM_CGROUP_RECLAIM_NOSWAP;
	} else
		mem_over_limit = mem_cgroup_from_res_counter(fail_res, res);
2618 2619 2620 2621
	/*
	 * Never reclaim on behalf of optional batching, retry with a
	 * single page instead.
	 */
2622
	if (nr_pages > min_pages)
2623 2624 2625 2626 2627
		return CHARGE_RETRY;

	if (!(gfp_mask & __GFP_WAIT))
		return CHARGE_WOULDBLOCK;

2628 2629 2630
	if (gfp_mask & __GFP_NORETRY)
		return CHARGE_NOMEM;

2631
	ret = mem_cgroup_reclaim(mem_over_limit, gfp_mask, flags);
2632
	if (mem_cgroup_margin(mem_over_limit) >= nr_pages)
2633
		return CHARGE_RETRY;
2634
	/*
2635 2636 2637 2638 2639 2640 2641
	 * Even though the limit is exceeded at this point, reclaim
	 * may have been able to free some pages.  Retry the charge
	 * before killing the task.
	 *
	 * Only for regular pages, though: huge pages are rather
	 * unlikely to succeed so close to the limit, and we fall back
	 * to regular pages anyway in case of failure.
2642
	 */
2643
	if (nr_pages <= (1 << PAGE_ALLOC_COSTLY_ORDER) && ret)
2644 2645 2646 2647 2648 2649 2650 2651 2652
		return CHARGE_RETRY;

	/*
	 * At task move, charge accounts can be doubly counted. So, it's
	 * better to wait until the end of task_move if something is going on.
	 */
	if (mem_cgroup_wait_acct_move(mem_over_limit))
		return CHARGE_RETRY;

2653 2654
	if (invoke_oom)
		mem_cgroup_oom(mem_over_limit, gfp_mask, get_order(csize));
2655

2656
	return CHARGE_NOMEM;
2657 2658
}

2659 2660 2661 2662 2663
/**
 * mem_cgroup_try_charge - try charging a memcg
 * @memcg: memcg to charge
 * @nr_pages: number of pages to charge
 * @oom: trigger OOM if reclaim fails
2664
 *
2665 2666
 * Returns 0 if @memcg was charged successfully, -EINTR if the charge
 * was bypassed to root_mem_cgroup, and -ENOMEM if the charge failed.
2667
 */
2668 2669 2670 2671
static int mem_cgroup_try_charge(struct mem_cgroup *memcg,
				 gfp_t gfp_mask,
				 unsigned int nr_pages,
				 bool oom)
2672
{
2673
	unsigned int batch = max(CHARGE_BATCH, nr_pages);
2674 2675
	int nr_oom_retries = MEM_CGROUP_RECLAIM_RETRIES;
	int ret;
2676

2677 2678
	if (mem_cgroup_is_root(memcg))
		goto done;
K
KAMEZAWA Hiroyuki 已提交
2679
	/*
2680 2681 2682 2683
	 * Unlike in global OOM situations, memcg is not in a physical
	 * memory shortage.  Allow dying and OOM-killed tasks to
	 * bypass the last charges so that they can exit quickly and
	 * free their memory.
K
KAMEZAWA Hiroyuki 已提交
2684
	 */
2685
	if (unlikely(test_thread_flag(TIF_MEMDIE) ||
2686 2687
		     fatal_signal_pending(current) ||
		     current->flags & PF_EXITING))
K
KAMEZAWA Hiroyuki 已提交
2688
		goto bypass;
2689

2690
	if (unlikely(task_in_memcg_oom(current)))
2691
		goto nomem;
2692

2693 2694
	if (gfp_mask & __GFP_NOFAIL)
		oom = false;
K
KAMEZAWA Hiroyuki 已提交
2695
again:
2696 2697
	if (consume_stock(memcg, nr_pages))
		goto done;
2698

2699
	do {
2700
		bool invoke_oom = oom && !nr_oom_retries;
2701

2702
		/* If killed, bypass charge */
2703
		if (fatal_signal_pending(current))
2704
			goto bypass;
2705

2706 2707
		ret = mem_cgroup_do_charge(memcg, gfp_mask, batch,
					   nr_pages, invoke_oom);
2708 2709 2710 2711
		switch (ret) {
		case CHARGE_OK:
			break;
		case CHARGE_RETRY: /* not in OOM situation but retry */
2712
			batch = nr_pages;
K
KAMEZAWA Hiroyuki 已提交
2713
			goto again;
2714 2715 2716
		case CHARGE_WOULDBLOCK: /* !__GFP_WAIT */
			goto nomem;
		case CHARGE_NOMEM: /* OOM routine works */
2717
			if (!oom || invoke_oom)
K
KAMEZAWA Hiroyuki 已提交
2718
				goto nomem;
2719 2720
			nr_oom_retries--;
			break;
2721
		}
2722 2723
	} while (ret != CHARGE_OK);

2724
	if (batch > nr_pages)
2725
		refill_stock(memcg, batch - nr_pages);
2726
done:
2727 2728
	return 0;
nomem:
2729
	if (!(gfp_mask & __GFP_NOFAIL))
2730
		return -ENOMEM;
K
KAMEZAWA Hiroyuki 已提交
2731
bypass:
2732
	return -EINTR;
2733
}
2734

2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763
/**
 * mem_cgroup_try_charge_mm - try charging a mm
 * @mm: mm_struct to charge
 * @nr_pages: number of pages to charge
 * @oom: trigger OOM if reclaim fails
 *
 * Returns the charged mem_cgroup associated with the given mm_struct or
 * NULL the charge failed.
 */
static struct mem_cgroup *mem_cgroup_try_charge_mm(struct mm_struct *mm,
				 gfp_t gfp_mask,
				 unsigned int nr_pages,
				 bool oom)

{
	struct mem_cgroup *memcg;
	int ret;

	memcg = get_mem_cgroup_from_mm(mm);
	ret = mem_cgroup_try_charge(memcg, gfp_mask, nr_pages, oom);
	css_put(&memcg->css);
	if (ret == -EINTR)
		memcg = root_mem_cgroup;
	else if (ret)
		memcg = NULL;

	return memcg;
}

2764 2765 2766 2767 2768
/*
 * Somemtimes we have to undo a charge we got by try_charge().
 * This function is for that and do uncharge, put css's refcnt.
 * gotten by try_charge().
 */
2769
static void __mem_cgroup_cancel_charge(struct mem_cgroup *memcg,
2770
				       unsigned int nr_pages)
2771
{
2772
	if (!mem_cgroup_is_root(memcg)) {
2773 2774
		unsigned long bytes = nr_pages * PAGE_SIZE;

2775
		res_counter_uncharge(&memcg->res, bytes);
2776
		if (do_swap_account)
2777
			res_counter_uncharge(&memcg->memsw, bytes);
2778
	}
2779 2780
}

2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798
/*
 * Cancel chrages in this cgroup....doesn't propagate to parent cgroup.
 * This is useful when moving usage to parent cgroup.
 */
static void __mem_cgroup_cancel_local_charge(struct mem_cgroup *memcg,
					unsigned int nr_pages)
{
	unsigned long bytes = nr_pages * PAGE_SIZE;

	if (mem_cgroup_is_root(memcg))
		return;

	res_counter_uncharge_until(&memcg->res, memcg->res.parent, bytes);
	if (do_swap_account)
		res_counter_uncharge_until(&memcg->memsw,
						memcg->memsw.parent, bytes);
}

2799 2800
/*
 * A helper function to get mem_cgroup from ID. must be called under
T
Tejun Heo 已提交
2801 2802 2803
 * rcu_read_lock().  The caller is responsible for calling css_tryget if
 * the mem_cgroup is used for charging. (dropping refcnt from swap can be
 * called against removed memcg.)
2804 2805 2806 2807 2808 2809
 */
static struct mem_cgroup *mem_cgroup_lookup(unsigned short id)
{
	/* ID 0 is unused ID */
	if (!id)
		return NULL;
L
Li Zefan 已提交
2810
	return mem_cgroup_from_id(id);
2811 2812
}

2813
struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page)
2814
{
2815
	struct mem_cgroup *memcg = NULL;
2816
	struct page_cgroup *pc;
2817
	unsigned short id;
2818 2819
	swp_entry_t ent;

2820
	VM_BUG_ON_PAGE(!PageLocked(page), page);
2821 2822

	pc = lookup_page_cgroup(page);
2823
	lock_page_cgroup(pc);
2824
	if (PageCgroupUsed(pc)) {
2825 2826 2827
		memcg = pc->mem_cgroup;
		if (memcg && !css_tryget(&memcg->css))
			memcg = NULL;
2828
	} else if (PageSwapCache(page)) {
2829
		ent.val = page_private(page);
2830
		id = lookup_swap_cgroup_id(ent);
2831
		rcu_read_lock();
2832 2833 2834
		memcg = mem_cgroup_lookup(id);
		if (memcg && !css_tryget(&memcg->css))
			memcg = NULL;
2835
		rcu_read_unlock();
2836
	}
2837
	unlock_page_cgroup(pc);
2838
	return memcg;
2839 2840
}

2841
static void __mem_cgroup_commit_charge(struct mem_cgroup *memcg,
2842
				       struct page *page,
2843
				       unsigned int nr_pages,
2844 2845
				       enum charge_type ctype,
				       bool lrucare)
2846
{
2847
	struct page_cgroup *pc = lookup_page_cgroup(page);
2848
	struct zone *uninitialized_var(zone);
2849
	struct lruvec *lruvec;
2850
	bool was_on_lru = false;
2851
	bool anon;
2852

2853
	lock_page_cgroup(pc);
2854
	VM_BUG_ON_PAGE(PageCgroupUsed(pc), page);
2855 2856 2857 2858
	/*
	 * we don't need page_cgroup_lock about tail pages, becase they are not
	 * accessed by any other context at this point.
	 */
2859 2860 2861 2862 2863 2864 2865 2866 2867

	/*
	 * In some cases, SwapCache and FUSE(splice_buf->radixtree), the page
	 * may already be on some other mem_cgroup's LRU.  Take care of it.
	 */
	if (lrucare) {
		zone = page_zone(page);
		spin_lock_irq(&zone->lru_lock);
		if (PageLRU(page)) {
2868
			lruvec = mem_cgroup_zone_lruvec(zone, pc->mem_cgroup);
2869
			ClearPageLRU(page);
2870
			del_page_from_lru_list(page, lruvec, page_lru(page));
2871 2872 2873 2874
			was_on_lru = true;
		}
	}

2875
	pc->mem_cgroup = memcg;
2876 2877 2878 2879 2880 2881
	/*
	 * We access a page_cgroup asynchronously without lock_page_cgroup().
	 * Especially when a page_cgroup is taken from a page, pc->mem_cgroup
	 * is accessed after testing USED bit. To make pc->mem_cgroup visible
	 * before USED bit, we need memory barrier here.
	 * See mem_cgroup_add_lru_list(), etc.
A
Andrew Morton 已提交
2882
	 */
K
KAMEZAWA Hiroyuki 已提交
2883
	smp_wmb();
2884
	SetPageCgroupUsed(pc);
2885

2886 2887
	if (lrucare) {
		if (was_on_lru) {
2888
			lruvec = mem_cgroup_zone_lruvec(zone, pc->mem_cgroup);
2889
			VM_BUG_ON_PAGE(PageLRU(page), page);
2890
			SetPageLRU(page);
2891
			add_page_to_lru_list(page, lruvec, page_lru(page));
2892 2893 2894 2895
		}
		spin_unlock_irq(&zone->lru_lock);
	}

2896
	if (ctype == MEM_CGROUP_CHARGE_TYPE_ANON)
2897 2898 2899 2900
		anon = true;
	else
		anon = false;

2901
	mem_cgroup_charge_statistics(memcg, page, anon, nr_pages);
2902
	unlock_page_cgroup(pc);
2903

2904
	/*
2905 2906 2907
	 * "charge_statistics" updated event counter. Then, check it.
	 * Insert ancestor (and ancestor's ancestors), to softlimit RB-tree.
	 * if they exceeds softlimit.
2908
	 */
2909
	memcg_check_events(memcg, page);
2910
}
2911

2912 2913
static DEFINE_MUTEX(set_limit_mutex);

2914
#ifdef CONFIG_MEMCG_KMEM
2915 2916 2917 2918 2919 2920
/*
 * The memcg_slab_mutex is held whenever a per memcg kmem cache is created or
 * destroyed. It protects memcg_caches arrays and memcg_slab_caches lists.
 */
static DEFINE_MUTEX(memcg_slab_mutex);

2921 2922
static DEFINE_MUTEX(activate_kmem_mutex);

2923 2924 2925
static inline bool memcg_can_account_kmem(struct mem_cgroup *memcg)
{
	return !mem_cgroup_disabled() && !mem_cgroup_is_root(memcg) &&
2926
		memcg_kmem_is_active(memcg);
2927 2928
}

G
Glauber Costa 已提交
2929 2930 2931 2932 2933 2934 2935 2936 2937 2938
/*
 * This is a bit cumbersome, but it is rarely used and avoids a backpointer
 * in the memcg_cache_params struct.
 */
static struct kmem_cache *memcg_params_to_cache(struct memcg_cache_params *p)
{
	struct kmem_cache *cachep;

	VM_BUG_ON(p->is_root_cache);
	cachep = p->root_cache;
2939
	return cache_from_memcg_idx(cachep, memcg_cache_id(p->memcg));
G
Glauber Costa 已提交
2940 2941
}

2942
#ifdef CONFIG_SLABINFO
2943
static int mem_cgroup_slabinfo_read(struct seq_file *m, void *v)
2944
{
2945
	struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(m));
2946 2947 2948 2949 2950 2951 2952
	struct memcg_cache_params *params;

	if (!memcg_can_account_kmem(memcg))
		return -EIO;

	print_slabinfo_header(m);

2953
	mutex_lock(&memcg_slab_mutex);
2954 2955
	list_for_each_entry(params, &memcg->memcg_slab_caches, list)
		cache_show(memcg_params_to_cache(params), m);
2956
	mutex_unlock(&memcg_slab_mutex);
2957 2958 2959 2960 2961

	return 0;
}
#endif

2962
static int memcg_charge_kmem(struct mem_cgroup *memcg, gfp_t gfp, u64 size)
2963 2964 2965 2966 2967 2968 2969 2970
{
	struct res_counter *fail_res;
	int ret = 0;

	ret = res_counter_charge(&memcg->kmem, size, &fail_res);
	if (ret)
		return ret;

2971 2972
	ret = mem_cgroup_try_charge(memcg, gfp, size >> PAGE_SHIFT,
				    oom_gfp_allowed(gfp));
2973 2974
	if (ret == -EINTR)  {
		/*
2975
		 * mem_cgroup_try_charge() chosed to bypass to root due to
2976 2977 2978 2979 2980 2981 2982 2983 2984
		 * OOM kill or fatal signal.  Since our only options are to
		 * either fail the allocation or charge it to this cgroup, do
		 * it as a temporary condition. But we can't fail. From a
		 * kmem/slab perspective, the cache has already been selected,
		 * by mem_cgroup_kmem_get_cache(), so it is too late to change
		 * our minds.
		 *
		 * This condition will only trigger if the task entered
		 * memcg_charge_kmem in a sane state, but was OOM-killed during
2985
		 * mem_cgroup_try_charge() above. Tasks that were already
2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999
		 * dying when the allocation triggers should have been already
		 * directed to the root cgroup in memcontrol.h
		 */
		res_counter_charge_nofail(&memcg->res, size, &fail_res);
		if (do_swap_account)
			res_counter_charge_nofail(&memcg->memsw, size,
						  &fail_res);
		ret = 0;
	} else if (ret)
		res_counter_uncharge(&memcg->kmem, size);

	return ret;
}

3000
static void memcg_uncharge_kmem(struct mem_cgroup *memcg, u64 size)
3001 3002 3003 3004
{
	res_counter_uncharge(&memcg->res, size);
	if (do_swap_account)
		res_counter_uncharge(&memcg->memsw, size);
3005 3006 3007 3008 3009

	/* Not down to 0 */
	if (res_counter_uncharge(&memcg->kmem, size))
		return;

3010 3011 3012 3013 3014 3015 3016 3017
	/*
	 * Releases a reference taken in kmem_cgroup_css_offline in case
	 * this last uncharge is racing with the offlining code or it is
	 * outliving the memcg existence.
	 *
	 * The memory barrier imposed by test&clear is paired with the
	 * explicit one in memcg_kmem_mark_dead().
	 */
3018
	if (memcg_kmem_test_and_clear_dead(memcg))
3019
		css_put(&memcg->css);
3020 3021
}

3022 3023 3024 3025 3026 3027 3028 3029 3030 3031
/*
 * helper for acessing a memcg's index. It will be used as an index in the
 * child cache array in kmem_cache, and also to derive its name. This function
 * will return -1 when this is not a kmem-limited memcg.
 */
int memcg_cache_id(struct mem_cgroup *memcg)
{
	return memcg ? memcg->kmemcg_id : -1;
}

3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061
static size_t memcg_caches_array_size(int num_groups)
{
	ssize_t size;
	if (num_groups <= 0)
		return 0;

	size = 2 * num_groups;
	if (size < MEMCG_CACHES_MIN_SIZE)
		size = MEMCG_CACHES_MIN_SIZE;
	else if (size > MEMCG_CACHES_MAX_SIZE)
		size = MEMCG_CACHES_MAX_SIZE;

	return size;
}

/*
 * We should update the current array size iff all caches updates succeed. This
 * can only be done from the slab side. The slab mutex needs to be held when
 * calling this.
 */
void memcg_update_array_size(int num)
{
	if (num > memcg_limited_groups_array_size)
		memcg_limited_groups_array_size = memcg_caches_array_size(num);
}

int memcg_update_cache_size(struct kmem_cache *s, int num_groups)
{
	struct memcg_cache_params *cur_params = s->memcg_params;

3062
	VM_BUG_ON(!is_root_cache(s));
3063 3064 3065

	if (num_groups > memcg_limited_groups_array_size) {
		int i;
3066
		struct memcg_cache_params *new_params;
3067 3068 3069
		ssize_t size = memcg_caches_array_size(num_groups);

		size *= sizeof(void *);
3070
		size += offsetof(struct memcg_cache_params, memcg_caches);
3071

3072 3073
		new_params = kzalloc(size, GFP_KERNEL);
		if (!new_params)
3074 3075
			return -ENOMEM;

3076
		new_params->is_root_cache = true;
3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089

		/*
		 * There is the chance it will be bigger than
		 * memcg_limited_groups_array_size, if we failed an allocation
		 * in a cache, in which case all caches updated before it, will
		 * have a bigger array.
		 *
		 * But if that is the case, the data after
		 * memcg_limited_groups_array_size is certainly unused
		 */
		for (i = 0; i < memcg_limited_groups_array_size; i++) {
			if (!cur_params->memcg_caches[i])
				continue;
3090
			new_params->memcg_caches[i] =
3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102
						cur_params->memcg_caches[i];
		}

		/*
		 * Ideally, we would wait until all caches succeed, and only
		 * then free the old one. But this is not worth the extra
		 * pointer per-cache we'd have to have for this.
		 *
		 * It is not a big deal if some caches are left with a size
		 * bigger than the others. And all updates will reset this
		 * anyway.
		 */
3103 3104 3105
		rcu_assign_pointer(s->memcg_params, new_params);
		if (cur_params)
			kfree_rcu(cur_params, rcu_head);
3106 3107 3108 3109
	}
	return 0;
}

3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132
char *memcg_create_cache_name(struct mem_cgroup *memcg,
			      struct kmem_cache *root_cache)
{
	static char *buf = NULL;

	/*
	 * We need a mutex here to protect the shared buffer. Since this is
	 * expected to be called only on cache creation, we can employ the
	 * slab_mutex for that purpose.
	 */
	lockdep_assert_held(&slab_mutex);

	if (!buf) {
		buf = kmalloc(NAME_MAX + 1, GFP_KERNEL);
		if (!buf)
			return NULL;
	}

	cgroup_name(memcg->css.cgroup, buf, NAME_MAX + 1);
	return kasprintf(GFP_KERNEL, "%s(%d:%s)", root_cache->name,
			 memcg_cache_id(memcg), buf);
}

3133 3134
int memcg_alloc_cache_params(struct mem_cgroup *memcg, struct kmem_cache *s,
			     struct kmem_cache *root_cache)
3135
{
3136
	size_t size;
3137 3138 3139 3140

	if (!memcg_kmem_enabled())
		return 0;

3141 3142
	if (!memcg) {
		size = offsetof(struct memcg_cache_params, memcg_caches);
3143
		size += memcg_limited_groups_array_size * sizeof(void *);
3144 3145
	} else
		size = sizeof(struct memcg_cache_params);
3146

3147 3148 3149 3150
	s->memcg_params = kzalloc(size, GFP_KERNEL);
	if (!s->memcg_params)
		return -ENOMEM;

G
Glauber Costa 已提交
3151
	if (memcg) {
3152
		s->memcg_params->memcg = memcg;
G
Glauber Costa 已提交
3153
		s->memcg_params->root_cache = root_cache;
3154
		css_get(&memcg->css);
3155 3156 3157
	} else
		s->memcg_params->is_root_cache = true;

3158 3159 3160
	return 0;
}

3161 3162
void memcg_free_cache_params(struct kmem_cache *s)
{
3163 3164 3165 3166
	if (!s->memcg_params)
		return;
	if (!s->memcg_params->is_root_cache)
		css_put(&s->memcg_params->memcg->css);
3167 3168 3169
	kfree(s->memcg_params);
}

3170 3171
static void memcg_kmem_create_cache(struct mem_cgroup *memcg,
				    struct kmem_cache *root_cache)
3172
{
3173
	struct kmem_cache *cachep;
3174 3175
	int id;

3176 3177 3178 3179 3180 3181 3182 3183 3184 3185
	lockdep_assert_held(&memcg_slab_mutex);

	id = memcg_cache_id(memcg);

	/*
	 * Since per-memcg caches are created asynchronously on first
	 * allocation (see memcg_kmem_get_cache()), several threads can try to
	 * create the same cache, but only one of them may succeed.
	 */
	if (cache_from_memcg_idx(root_cache, id))
3186 3187
		return;

3188
	cachep = kmem_cache_create_memcg(memcg, root_cache);
3189
	/*
3190 3191 3192
	 * If we could not create a memcg cache, do not complain, because
	 * that's not critical at all as we can always proceed with the root
	 * cache.
3193
	 */
3194 3195
	if (!cachep)
		return;
3196

3197
	list_add(&cachep->memcg_params->list, &memcg->memcg_slab_caches);
3198

3199
	/*
3200 3201 3202
	 * Since readers won't lock (see cache_from_memcg_idx()), we need a
	 * barrier here to ensure nobody will see the kmem_cache partially
	 * initialized.
3203
	 */
3204 3205
	smp_wmb();

3206 3207
	BUG_ON(root_cache->memcg_params->memcg_caches[id]);
	root_cache->memcg_params->memcg_caches[id] = cachep;
3208
}
3209

3210
static void memcg_kmem_destroy_cache(struct kmem_cache *cachep)
3211
{
3212
	struct kmem_cache *root_cache;
3213 3214 3215
	struct mem_cgroup *memcg;
	int id;

3216
	lockdep_assert_held(&memcg_slab_mutex);
3217

3218
	BUG_ON(is_root_cache(cachep));
3219

3220 3221
	root_cache = cachep->memcg_params->root_cache;
	memcg = cachep->memcg_params->memcg;
3222
	id = memcg_cache_id(memcg);
3223

3224 3225
	BUG_ON(root_cache->memcg_params->memcg_caches[id] != cachep);
	root_cache->memcg_params->memcg_caches[id] = NULL;
3226

3227 3228 3229
	list_del(&cachep->memcg_params->list);

	kmem_cache_destroy(cachep);
3230 3231
}

3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262
/*
 * During the creation a new cache, we need to disable our accounting mechanism
 * altogether. This is true even if we are not creating, but rather just
 * enqueing new caches to be created.
 *
 * This is because that process will trigger allocations; some visible, like
 * explicit kmallocs to auxiliary data structures, name strings and internal
 * cache structures; some well concealed, like INIT_WORK() that can allocate
 * objects during debug.
 *
 * If any allocation happens during memcg_kmem_get_cache, we will recurse back
 * to it. This may not be a bounded recursion: since the first cache creation
 * failed to complete (waiting on the allocation), we'll just try to create the
 * cache again, failing at the same point.
 *
 * memcg_kmem_get_cache is prepared to abort after seeing a positive count of
 * memcg_kmem_skip_account. So we enclose anything that might allocate memory
 * inside the following two functions.
 */
static inline void memcg_stop_kmem_account(void)
{
	VM_BUG_ON(!current->mm);
	current->memcg_kmem_skip_account++;
}

static inline void memcg_resume_kmem_account(void)
{
	VM_BUG_ON(!current->mm);
	current->memcg_kmem_skip_account--;
}

3263
int __kmem_cache_destroy_memcg_children(struct kmem_cache *s)
3264 3265
{
	struct kmem_cache *c;
3266
	int i, failed = 0;
3267

3268
	mutex_lock(&memcg_slab_mutex);
3269 3270
	for_each_memcg_cache_index(i) {
		c = cache_from_memcg_idx(s, i);
3271 3272 3273
		if (!c)
			continue;

3274
		memcg_kmem_destroy_cache(c);
3275 3276 3277

		if (cache_from_memcg_idx(s, i))
			failed++;
3278
	}
3279
	mutex_unlock(&memcg_slab_mutex);
3280
	return failed;
3281 3282
}

G
Glauber Costa 已提交
3283 3284 3285
static void mem_cgroup_destroy_all_caches(struct mem_cgroup *memcg)
{
	struct kmem_cache *cachep;
3286
	struct memcg_cache_params *params, *tmp;
G
Glauber Costa 已提交
3287 3288 3289 3290

	if (!memcg_kmem_is_active(memcg))
		return;

3291 3292
	mutex_lock(&memcg_slab_mutex);
	list_for_each_entry_safe(params, tmp, &memcg->memcg_slab_caches, list) {
G
Glauber Costa 已提交
3293
		cachep = memcg_params_to_cache(params);
3294 3295 3296
		kmem_cache_shrink(cachep);
		if (atomic_read(&cachep->memcg_params->nr_pages) == 0)
			memcg_kmem_destroy_cache(cachep);
G
Glauber Costa 已提交
3297
	}
3298
	mutex_unlock(&memcg_slab_mutex);
G
Glauber Costa 已提交
3299 3300
}

3301 3302 3303 3304 3305 3306
struct create_work {
	struct mem_cgroup *memcg;
	struct kmem_cache *cachep;
	struct work_struct work;
};

3307 3308
static void memcg_create_cache_work_func(struct work_struct *w)
{
3309 3310 3311
	struct create_work *cw = container_of(w, struct create_work, work);
	struct mem_cgroup *memcg = cw->memcg;
	struct kmem_cache *cachep = cw->cachep;
3312

3313 3314 3315 3316
	mutex_lock(&memcg_slab_mutex);
	memcg_kmem_create_cache(memcg, cachep);
	mutex_unlock(&memcg_slab_mutex);

3317
	css_put(&memcg->css);
3318 3319 3320 3321 3322 3323
	kfree(cw);
}

/*
 * Enqueue the creation of a per-memcg kmem_cache.
 */
3324 3325
static void __memcg_create_cache_enqueue(struct mem_cgroup *memcg,
					 struct kmem_cache *cachep)
3326 3327 3328 3329
{
	struct create_work *cw;

	cw = kmalloc(sizeof(struct create_work), GFP_NOWAIT);
3330 3331
	if (cw == NULL) {
		css_put(&memcg->css);
3332 3333 3334 3335 3336 3337 3338 3339 3340 3341
		return;
	}

	cw->memcg = memcg;
	cw->cachep = cachep;

	INIT_WORK(&cw->work, memcg_create_cache_work_func);
	schedule_work(&cw->work);
}

3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359
static void memcg_create_cache_enqueue(struct mem_cgroup *memcg,
				       struct kmem_cache *cachep)
{
	/*
	 * We need to stop accounting when we kmalloc, because if the
	 * corresponding kmalloc cache is not yet created, the first allocation
	 * in __memcg_create_cache_enqueue will recurse.
	 *
	 * However, it is better to enclose the whole function. Depending on
	 * the debugging options enabled, INIT_WORK(), for instance, can
	 * trigger an allocation. This too, will make us recurse. Because at
	 * this point we can't allow ourselves back into memcg_kmem_get_cache,
	 * the safest choice is to do it like this, wrapping the whole function.
	 */
	memcg_stop_kmem_account();
	__memcg_create_cache_enqueue(memcg, cachep);
	memcg_resume_kmem_account();
}
3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377

int __memcg_charge_slab(struct kmem_cache *cachep, gfp_t gfp, int order)
{
	int res;

	res = memcg_charge_kmem(cachep->memcg_params->memcg, gfp,
				PAGE_SIZE << order);
	if (!res)
		atomic_add(1 << order, &cachep->memcg_params->nr_pages);
	return res;
}

void __memcg_uncharge_slab(struct kmem_cache *cachep, int order)
{
	memcg_uncharge_kmem(cachep->memcg_params->memcg, PAGE_SIZE << order);
	atomic_sub(1 << order, &cachep->memcg_params->nr_pages);
}

3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394
/*
 * Return the kmem_cache we're supposed to use for a slab allocation.
 * We try to use the current memcg's version of the cache.
 *
 * If the cache does not exist yet, if we are the first user of it,
 * we either create it immediately, if possible, or create it asynchronously
 * in a workqueue.
 * In the latter case, we will let the current allocation go through with
 * the original cache.
 *
 * Can't be called in interrupt context or from kernel threads.
 * This function needs to be called with rcu_read_lock() held.
 */
struct kmem_cache *__memcg_kmem_get_cache(struct kmem_cache *cachep,
					  gfp_t gfp)
{
	struct mem_cgroup *memcg;
3395
	struct kmem_cache *memcg_cachep;
3396 3397 3398 3399

	VM_BUG_ON(!cachep->memcg_params);
	VM_BUG_ON(!cachep->memcg_params->is_root_cache);

3400 3401 3402
	if (!current->mm || current->memcg_kmem_skip_account)
		return cachep;

3403 3404 3405 3406
	rcu_read_lock();
	memcg = mem_cgroup_from_task(rcu_dereference(current->mm->owner));

	if (!memcg_can_account_kmem(memcg))
3407
		goto out;
3408

3409 3410 3411
	memcg_cachep = cache_from_memcg_idx(cachep, memcg_cache_id(memcg));
	if (likely(memcg_cachep)) {
		cachep = memcg_cachep;
3412
		goto out;
3413 3414
	}

3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441
	/* The corresponding put will be done in the workqueue. */
	if (!css_tryget(&memcg->css))
		goto out;
	rcu_read_unlock();

	/*
	 * If we are in a safe context (can wait, and not in interrupt
	 * context), we could be be predictable and return right away.
	 * This would guarantee that the allocation being performed
	 * already belongs in the new cache.
	 *
	 * However, there are some clashes that can arrive from locking.
	 * For instance, because we acquire the slab_mutex while doing
	 * kmem_cache_dup, this means no further allocation could happen
	 * with the slab_mutex held.
	 *
	 * Also, because cache creation issue get_online_cpus(), this
	 * creates a lock chain: memcg_slab_mutex -> cpu_hotplug_mutex,
	 * that ends up reversed during cpu hotplug. (cpuset allocates
	 * a bunch of GFP_KERNEL memory during cpuup). Due to all that,
	 * better to defer everything.
	 */
	memcg_create_cache_enqueue(memcg, cachep);
	return cachep;
out:
	rcu_read_unlock();
	return cachep;
3442 3443
}

3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464
/*
 * We need to verify if the allocation against current->mm->owner's memcg is
 * possible for the given order. But the page is not allocated yet, so we'll
 * need a further commit step to do the final arrangements.
 *
 * It is possible for the task to switch cgroups in this mean time, so at
 * commit time, we can't rely on task conversion any longer.  We'll then use
 * the handle argument to return to the caller which cgroup we should commit
 * against. We could also return the memcg directly and avoid the pointer
 * passing, but a boolean return value gives better semantics considering
 * the compiled-out case as well.
 *
 * Returning true means the allocation is possible.
 */
bool
__memcg_kmem_newpage_charge(gfp_t gfp, struct mem_cgroup **_memcg, int order)
{
	struct mem_cgroup *memcg;
	int ret;

	*_memcg = NULL;
3465 3466 3467 3468

	/*
	 * Disabling accounting is only relevant for some specific memcg
	 * internal allocations. Therefore we would initially not have such
V
Vladimir Davydov 已提交
3469 3470 3471 3472 3473 3474
	 * check here, since direct calls to the page allocator that are
	 * accounted to kmemcg (alloc_kmem_pages and friends) only happen
	 * outside memcg core. We are mostly concerned with cache allocations,
	 * and by having this test at memcg_kmem_get_cache, we are already able
	 * to relay the allocation to the root cache and bypass the memcg cache
	 * altogether.
3475 3476 3477 3478 3479 3480
	 *
	 * There is one exception, though: the SLUB allocator does not create
	 * large order caches, but rather service large kmallocs directly from
	 * the page allocator. Therefore, the following sequence when backed by
	 * the SLUB allocator:
	 *
A
Andrew Morton 已提交
3481 3482 3483
	 *	memcg_stop_kmem_account();
	 *	kmalloc(<large_number>)
	 *	memcg_resume_kmem_account();
3484 3485 3486 3487 3488 3489 3490 3491 3492 3493
	 *
	 * would effectively ignore the fact that we should skip accounting,
	 * since it will drive us directly to this function without passing
	 * through the cache selector memcg_kmem_get_cache. Such large
	 * allocations are extremely rare but can happen, for instance, for the
	 * cache arrays. We bring this test here.
	 */
	if (!current->mm || current->memcg_kmem_skip_account)
		return true;

3494
	memcg = get_mem_cgroup_from_mm(current->mm);
3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513 3514 3515 3516 3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556

	if (!memcg_can_account_kmem(memcg)) {
		css_put(&memcg->css);
		return true;
	}

	ret = memcg_charge_kmem(memcg, gfp, PAGE_SIZE << order);
	if (!ret)
		*_memcg = memcg;

	css_put(&memcg->css);
	return (ret == 0);
}

void __memcg_kmem_commit_charge(struct page *page, struct mem_cgroup *memcg,
			      int order)
{
	struct page_cgroup *pc;

	VM_BUG_ON(mem_cgroup_is_root(memcg));

	/* The page allocation failed. Revert */
	if (!page) {
		memcg_uncharge_kmem(memcg, PAGE_SIZE << order);
		return;
	}

	pc = lookup_page_cgroup(page);
	lock_page_cgroup(pc);
	pc->mem_cgroup = memcg;
	SetPageCgroupUsed(pc);
	unlock_page_cgroup(pc);
}

void __memcg_kmem_uncharge_pages(struct page *page, int order)
{
	struct mem_cgroup *memcg = NULL;
	struct page_cgroup *pc;


	pc = lookup_page_cgroup(page);
	/*
	 * Fast unlocked return. Theoretically might have changed, have to
	 * check again after locking.
	 */
	if (!PageCgroupUsed(pc))
		return;

	lock_page_cgroup(pc);
	if (PageCgroupUsed(pc)) {
		memcg = pc->mem_cgroup;
		ClearPageCgroupUsed(pc);
	}
	unlock_page_cgroup(pc);

	/*
	 * We trust that only if there is a memcg associated with the page, it
	 * is a valid allocation
	 */
	if (!memcg)
		return;

3557
	VM_BUG_ON_PAGE(mem_cgroup_is_root(memcg), page);
3558 3559
	memcg_uncharge_kmem(memcg, PAGE_SIZE << order);
}
G
Glauber Costa 已提交
3560 3561 3562 3563
#else
static inline void mem_cgroup_destroy_all_caches(struct mem_cgroup *memcg)
{
}
3564 3565
#endif /* CONFIG_MEMCG_KMEM */

3566 3567
#ifdef CONFIG_TRANSPARENT_HUGEPAGE

3568
#define PCGF_NOCOPY_AT_SPLIT (1 << PCG_LOCK | 1 << PCG_MIGRATION)
3569 3570
/*
 * Because tail pages are not marked as "used", set it. We're under
3571 3572 3573
 * zone->lru_lock, 'splitting on pmd' and compound_lock.
 * charge/uncharge will be never happen and move_account() is done under
 * compound_lock(), so we don't have to take care of races.
3574
 */
3575
void mem_cgroup_split_huge_fixup(struct page *head)
3576 3577
{
	struct page_cgroup *head_pc = lookup_page_cgroup(head);
3578
	struct page_cgroup *pc;
3579
	struct mem_cgroup *memcg;
3580
	int i;
3581

3582 3583
	if (mem_cgroup_disabled())
		return;
3584 3585

	memcg = head_pc->mem_cgroup;
3586 3587
	for (i = 1; i < HPAGE_PMD_NR; i++) {
		pc = head_pc + i;
3588
		pc->mem_cgroup = memcg;
3589 3590 3591
		smp_wmb();/* see __commit_charge() */
		pc->flags = head_pc->flags & ~PCGF_NOCOPY_AT_SPLIT;
	}
3592 3593
	__this_cpu_sub(memcg->stat->count[MEM_CGROUP_STAT_RSS_HUGE],
		       HPAGE_PMD_NR);
3594
}
3595
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
3596

3597
/**
3598
 * mem_cgroup_move_account - move account of the page
3599
 * @page: the page
3600
 * @nr_pages: number of regular pages (>1 for huge pages)
3601 3602 3603 3604 3605
 * @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.
K
KAMEZAWA Hiroyuki 已提交
3606
 * - page is not on LRU (isolate_page() is useful.)
3607
 * - compound_lock is held when nr_pages > 1
3608
 *
3609 3610
 * This function doesn't do "charge" to new cgroup and doesn't do "uncharge"
 * from old cgroup.
3611
 */
3612 3613 3614 3615
static int mem_cgroup_move_account(struct page *page,
				   unsigned int nr_pages,
				   struct page_cgroup *pc,
				   struct mem_cgroup *from,
3616
				   struct mem_cgroup *to)
3617
{
3618 3619
	unsigned long flags;
	int ret;
3620
	bool anon = PageAnon(page);
3621

3622
	VM_BUG_ON(from == to);
3623
	VM_BUG_ON_PAGE(PageLRU(page), page);
3624 3625 3626 3627 3628 3629 3630
	/*
	 * The page is isolated from LRU. So, collapse function
	 * will not handle this page. But page splitting can happen.
	 * Do this check under compound_page_lock(). The caller should
	 * hold it.
	 */
	ret = -EBUSY;
3631
	if (nr_pages > 1 && !PageTransHuge(page))
3632 3633 3634 3635 3636 3637 3638 3639
		goto out;

	lock_page_cgroup(pc);

	ret = -EINVAL;
	if (!PageCgroupUsed(pc) || pc->mem_cgroup != from)
		goto unlock;

3640
	move_lock_mem_cgroup(from, &flags);
3641

3642 3643 3644 3645 3646 3647
	if (!anon && page_mapped(page)) {
		__this_cpu_sub(from->stat->count[MEM_CGROUP_STAT_FILE_MAPPED],
			       nr_pages);
		__this_cpu_add(to->stat->count[MEM_CGROUP_STAT_FILE_MAPPED],
			       nr_pages);
	}
3648

3649 3650 3651 3652 3653 3654
	if (PageWriteback(page)) {
		__this_cpu_sub(from->stat->count[MEM_CGROUP_STAT_WRITEBACK],
			       nr_pages);
		__this_cpu_add(to->stat->count[MEM_CGROUP_STAT_WRITEBACK],
			       nr_pages);
	}
3655

3656
	mem_cgroup_charge_statistics(from, page, anon, -nr_pages);
3657

3658
	/* caller should have done css_get */
K
KAMEZAWA Hiroyuki 已提交
3659
	pc->mem_cgroup = to;
3660
	mem_cgroup_charge_statistics(to, page, anon, nr_pages);
3661
	move_unlock_mem_cgroup(from, &flags);
3662 3663
	ret = 0;
unlock:
3664
	unlock_page_cgroup(pc);
3665 3666 3667
	/*
	 * check events
	 */
3668 3669
	memcg_check_events(to, page);
	memcg_check_events(from, page);
3670
out:
3671 3672 3673
	return ret;
}

3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693
/**
 * mem_cgroup_move_parent - moves page to the parent group
 * @page: the page to move
 * @pc: page_cgroup of the page
 * @child: page's cgroup
 *
 * move charges to its parent or the root cgroup if the group has no
 * parent (aka use_hierarchy==0).
 * Although this might fail (get_page_unless_zero, isolate_lru_page or
 * mem_cgroup_move_account fails) the failure is always temporary and
 * it signals a race with a page removal/uncharge or migration. In the
 * first case the page is on the way out and it will vanish from the LRU
 * on the next attempt and the call should be retried later.
 * Isolation from the LRU fails only if page has been isolated from
 * the LRU since we looked at it and that usually means either global
 * reclaim or migration going on. The page will either get back to the
 * LRU or vanish.
 * Finaly mem_cgroup_move_account fails only if the page got uncharged
 * (!PageCgroupUsed) or moved to a different group. The page will
 * disappear in the next attempt.
3694
 */
3695 3696
static int mem_cgroup_move_parent(struct page *page,
				  struct page_cgroup *pc,
3697
				  struct mem_cgroup *child)
3698 3699
{
	struct mem_cgroup *parent;
3700
	unsigned int nr_pages;
3701
	unsigned long uninitialized_var(flags);
3702 3703
	int ret;

3704
	VM_BUG_ON(mem_cgroup_is_root(child));
3705

3706 3707 3708 3709 3710
	ret = -EBUSY;
	if (!get_page_unless_zero(page))
		goto out;
	if (isolate_lru_page(page))
		goto put;
3711

3712
	nr_pages = hpage_nr_pages(page);
K
KAMEZAWA Hiroyuki 已提交
3713

3714 3715 3716 3717 3718 3719
	parent = parent_mem_cgroup(child);
	/*
	 * If no parent, move charges to root cgroup.
	 */
	if (!parent)
		parent = root_mem_cgroup;
3720

3721
	if (nr_pages > 1) {
3722
		VM_BUG_ON_PAGE(!PageTransHuge(page), page);
3723
		flags = compound_lock_irqsave(page);
3724
	}
3725

3726
	ret = mem_cgroup_move_account(page, nr_pages,
3727
				pc, child, parent);
3728 3729
	if (!ret)
		__mem_cgroup_cancel_local_charge(child, nr_pages);
3730

3731
	if (nr_pages > 1)
3732
		compound_unlock_irqrestore(page, flags);
K
KAMEZAWA Hiroyuki 已提交
3733
	putback_lru_page(page);
3734
put:
3735
	put_page(page);
3736
out:
3737 3738 3739
	return ret;
}

3740
int mem_cgroup_charge_anon(struct page *page,
3741
			      struct mm_struct *mm, gfp_t gfp_mask)
3742
{
3743
	unsigned int nr_pages = 1;
3744
	struct mem_cgroup *memcg;
3745
	bool oom = true;
A
Andrea Arcangeli 已提交
3746

3747 3748 3749 3750 3751 3752 3753
	if (mem_cgroup_disabled())
		return 0;

	VM_BUG_ON_PAGE(page_mapped(page), page);
	VM_BUG_ON_PAGE(page->mapping && !PageAnon(page), page);
	VM_BUG_ON(!mm);

A
Andrea Arcangeli 已提交
3754
	if (PageTransHuge(page)) {
3755
		nr_pages <<= compound_order(page);
3756
		VM_BUG_ON_PAGE(!PageTransHuge(page), page);
3757 3758 3759 3760 3761
		/*
		 * Never OOM-kill a process for a huge page.  The
		 * fault handler will fall back to regular pages.
		 */
		oom = false;
A
Andrea Arcangeli 已提交
3762
	}
3763

3764 3765 3766
	memcg = mem_cgroup_try_charge_mm(mm, gfp_mask, nr_pages, oom);
	if (!memcg)
		return -ENOMEM;
3767 3768
	__mem_cgroup_commit_charge(memcg, page, nr_pages,
				   MEM_CGROUP_CHARGE_TYPE_ANON, false);
3769 3770 3771
	return 0;
}

3772 3773 3774
/*
 * While swap-in, try_charge -> commit or cancel, the page is locked.
 * And when try_charge() successfully returns, one refcnt to memcg without
3775
 * struct page_cgroup is acquired. This refcnt will be consumed by
3776 3777
 * "commit()" or removed by "cancel()"
 */
3778 3779 3780 3781
static int __mem_cgroup_try_charge_swapin(struct mm_struct *mm,
					  struct page *page,
					  gfp_t mask,
					  struct mem_cgroup **memcgp)
3782
{
3783
	struct mem_cgroup *memcg = NULL;
3784
	struct page_cgroup *pc;
3785
	int ret;
3786

3787 3788 3789 3790 3791 3792 3793 3794 3795
	pc = lookup_page_cgroup(page);
	/*
	 * Every swap fault against a single page tries to charge the
	 * page, bail as early as possible.  shmem_unuse() encounters
	 * already charged pages, too.  The USED bit is protected by
	 * the page lock, which serializes swap cache removal, which
	 * in turn serializes uncharging.
	 */
	if (PageCgroupUsed(pc))
3796 3797 3798
		goto out;
	if (do_swap_account)
		memcg = try_get_mem_cgroup_from_page(page);
3799
	if (!memcg)
3800 3801
		memcg = get_mem_cgroup_from_mm(mm);
	ret = mem_cgroup_try_charge(memcg, mask, 1, true);
3802
	css_put(&memcg->css);
3803
	if (ret == -EINTR)
3804 3805 3806 3807 3808 3809
		memcg = root_mem_cgroup;
	else if (ret)
		return ret;
out:
	*memcgp = memcg;
	return 0;
3810 3811
}

3812 3813 3814
int mem_cgroup_try_charge_swapin(struct mm_struct *mm, struct page *page,
				 gfp_t gfp_mask, struct mem_cgroup **memcgp)
{
3815 3816
	if (mem_cgroup_disabled()) {
		*memcgp = NULL;
3817
		return 0;
3818
	}
3819 3820 3821 3822 3823 3824 3825
	/*
	 * A racing thread's fault, or swapoff, may have already
	 * updated the pte, and even removed page from swap cache: in
	 * those cases unuse_pte()'s pte_same() test will fail; but
	 * there's also a KSM case which does need to charge the page.
	 */
	if (!PageSwapCache(page)) {
3826
		struct mem_cgroup *memcg;
3827

3828 3829 3830 3831 3832
		memcg = mem_cgroup_try_charge_mm(mm, gfp_mask, 1, true);
		if (!memcg)
			return -ENOMEM;
		*memcgp = memcg;
		return 0;
3833
	}
3834 3835 3836
	return __mem_cgroup_try_charge_swapin(mm, page, gfp_mask, memcgp);
}

3837 3838 3839 3840 3841 3842 3843 3844 3845
void mem_cgroup_cancel_charge_swapin(struct mem_cgroup *memcg)
{
	if (mem_cgroup_disabled())
		return;
	if (!memcg)
		return;
	__mem_cgroup_cancel_charge(memcg, 1);
}

D
Daisuke Nishimura 已提交
3846
static void
3847
__mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *memcg,
D
Daisuke Nishimura 已提交
3848
					enum charge_type ctype)
3849
{
3850
	if (mem_cgroup_disabled())
3851
		return;
3852
	if (!memcg)
3853
		return;
3854

3855
	__mem_cgroup_commit_charge(memcg, page, 1, ctype, true);
3856 3857 3858
	/*
	 * Now swap is on-memory. This means this page may be
	 * counted both as mem and swap....double count.
3859 3860 3861
	 * Fix it by uncharging from memsw. Basically, this SwapCache is stable
	 * under lock_page(). But in do_swap_page()::memory.c, reuse_swap_page()
	 * may call delete_from_swap_cache() before reach here.
3862
	 */
3863
	if (do_swap_account && PageSwapCache(page)) {
3864
		swp_entry_t ent = {.val = page_private(page)};
3865
		mem_cgroup_uncharge_swap(ent);
3866
	}
3867 3868
}

3869 3870
void mem_cgroup_commit_charge_swapin(struct page *page,
				     struct mem_cgroup *memcg)
D
Daisuke Nishimura 已提交
3871
{
3872
	__mem_cgroup_commit_charge_swapin(page, memcg,
3873
					  MEM_CGROUP_CHARGE_TYPE_ANON);
D
Daisuke Nishimura 已提交
3874 3875
}

3876
int mem_cgroup_charge_file(struct page *page, struct mm_struct *mm,
3877
				gfp_t gfp_mask)
3878
{
3879
	enum charge_type type = MEM_CGROUP_CHARGE_TYPE_CACHE;
3880
	struct mem_cgroup *memcg;
3881 3882
	int ret;

3883
	if (mem_cgroup_disabled())
3884 3885 3886 3887
		return 0;
	if (PageCompound(page))
		return 0;

3888
	if (PageSwapCache(page)) { /* shmem */
3889 3890
		ret = __mem_cgroup_try_charge_swapin(mm, page,
						     gfp_mask, &memcg);
3891 3892 3893 3894
		if (ret)
			return ret;
		__mem_cgroup_commit_charge_swapin(page, memcg, type);
		return 0;
3895
	}
3896

3897 3898 3899
	memcg = mem_cgroup_try_charge_mm(mm, gfp_mask, 1, true);
	if (!memcg)
		return -ENOMEM;
3900 3901
	__mem_cgroup_commit_charge(memcg, page, 1, type, false);
	return 0;
3902 3903
}

3904
static void mem_cgroup_do_uncharge(struct mem_cgroup *memcg,
3905 3906
				   unsigned int nr_pages,
				   const enum charge_type ctype)
3907 3908 3909
{
	struct memcg_batch_info *batch = NULL;
	bool uncharge_memsw = true;
3910

3911 3912 3913 3914 3915 3916 3917 3918 3919 3920 3921
	/* If swapout, usage of swap doesn't decrease */
	if (!do_swap_account || ctype == MEM_CGROUP_CHARGE_TYPE_SWAPOUT)
		uncharge_memsw = false;

	batch = &current->memcg_batch;
	/*
	 * In usual, we do css_get() when we remember memcg pointer.
	 * But in this case, we keep res->usage until end of a series of
	 * uncharges. Then, it's ok to ignore memcg's refcnt.
	 */
	if (!batch->memcg)
3922
		batch->memcg = memcg;
3923 3924
	/*
	 * do_batch > 0 when unmapping pages or inode invalidate/truncate.
L
Lucas De Marchi 已提交
3925
	 * In those cases, all pages freed continuously can be expected to be in
3926 3927 3928 3929 3930 3931 3932 3933
	 * the same cgroup and we have chance to coalesce uncharges.
	 * But we do uncharge one by one if this is killed by OOM(TIF_MEMDIE)
	 * because we want to do uncharge as soon as possible.
	 */

	if (!batch->do_batch || test_thread_flag(TIF_MEMDIE))
		goto direct_uncharge;

3934
	if (nr_pages > 1)
A
Andrea Arcangeli 已提交
3935 3936
		goto direct_uncharge;

3937 3938 3939 3940 3941
	/*
	 * In typical case, batch->memcg == mem. This means we can
	 * merge a series of uncharges to an uncharge of res_counter.
	 * If not, we uncharge res_counter ony by one.
	 */
3942
	if (batch->memcg != memcg)
3943 3944
		goto direct_uncharge;
	/* remember freed charge and uncharge it later */
3945
	batch->nr_pages++;
3946
	if (uncharge_memsw)
3947
		batch->memsw_nr_pages++;
3948 3949
	return;
direct_uncharge:
3950
	res_counter_uncharge(&memcg->res, nr_pages * PAGE_SIZE);
3951
	if (uncharge_memsw)
3952 3953 3954
		res_counter_uncharge(&memcg->memsw, nr_pages * PAGE_SIZE);
	if (unlikely(batch->memcg != memcg))
		memcg_oom_recover(memcg);
3955
}
3956

3957
/*
3958
 * uncharge if !page_mapped(page)
3959
 */
3960
static struct mem_cgroup *
3961 3962
__mem_cgroup_uncharge_common(struct page *page, enum charge_type ctype,
			     bool end_migration)
3963
{
3964
	struct mem_cgroup *memcg = NULL;
3965 3966
	unsigned int nr_pages = 1;
	struct page_cgroup *pc;
3967
	bool anon;
3968

3969
	if (mem_cgroup_disabled())
3970
		return NULL;
3971

A
Andrea Arcangeli 已提交
3972
	if (PageTransHuge(page)) {
3973
		nr_pages <<= compound_order(page);
3974
		VM_BUG_ON_PAGE(!PageTransHuge(page), page);
A
Andrea Arcangeli 已提交
3975
	}
3976
	/*
3977
	 * Check if our page_cgroup is valid
3978
	 */
3979
	pc = lookup_page_cgroup(page);
3980
	if (unlikely(!PageCgroupUsed(pc)))
3981
		return NULL;
3982

3983
	lock_page_cgroup(pc);
K
KAMEZAWA Hiroyuki 已提交
3984

3985
	memcg = pc->mem_cgroup;
3986

K
KAMEZAWA Hiroyuki 已提交
3987 3988 3989
	if (!PageCgroupUsed(pc))
		goto unlock_out;

3990 3991
	anon = PageAnon(page);

K
KAMEZAWA Hiroyuki 已提交
3992
	switch (ctype) {
3993
	case MEM_CGROUP_CHARGE_TYPE_ANON:
3994 3995 3996 3997 3998
		/*
		 * Generally PageAnon tells if it's the anon statistics to be
		 * updated; but sometimes e.g. mem_cgroup_uncharge_page() is
		 * used before page reached the stage of being marked PageAnon.
		 */
3999 4000
		anon = true;
		/* fallthrough */
K
KAMEZAWA Hiroyuki 已提交
4001
	case MEM_CGROUP_CHARGE_TYPE_DROP:
4002
		/* See mem_cgroup_prepare_migration() */
4003 4004 4005 4006 4007 4008 4009 4010 4011 4012
		if (page_mapped(page))
			goto unlock_out;
		/*
		 * Pages under migration may not be uncharged.  But
		 * end_migration() /must/ be the one uncharging the
		 * unused post-migration page and so it has to call
		 * here with the migration bit still set.  See the
		 * res_counter handling below.
		 */
		if (!end_migration && PageCgroupMigration(pc))
K
KAMEZAWA Hiroyuki 已提交
4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023
			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;
4024
	}
K
KAMEZAWA Hiroyuki 已提交
4025

4026
	mem_cgroup_charge_statistics(memcg, page, anon, -nr_pages);
K
KAMEZAWA Hiroyuki 已提交
4027

4028
	ClearPageCgroupUsed(pc);
4029 4030 4031 4032 4033 4034
	/*
	 * pc->mem_cgroup is not cleared here. It will be accessed when it's
	 * freed from LRU. This is safe because uncharged page is expected not
	 * to be reused (freed soon). Exception is SwapCache, it's handled by
	 * special functions.
	 */
4035

4036
	unlock_page_cgroup(pc);
K
KAMEZAWA Hiroyuki 已提交
4037
	/*
4038
	 * even after unlock, we have memcg->res.usage here and this memcg
L
Li Zefan 已提交
4039
	 * will never be freed, so it's safe to call css_get().
K
KAMEZAWA Hiroyuki 已提交
4040
	 */
4041
	memcg_check_events(memcg, page);
K
KAMEZAWA Hiroyuki 已提交
4042
	if (do_swap_account && ctype == MEM_CGROUP_CHARGE_TYPE_SWAPOUT) {
4043
		mem_cgroup_swap_statistics(memcg, true);
L
Li Zefan 已提交
4044
		css_get(&memcg->css);
K
KAMEZAWA Hiroyuki 已提交
4045
	}
4046 4047 4048 4049 4050 4051
	/*
	 * Migration does not charge the res_counter for the
	 * replacement page, so leave it alone when phasing out the
	 * page that is unused after the migration.
	 */
	if (!end_migration && !mem_cgroup_is_root(memcg))
4052
		mem_cgroup_do_uncharge(memcg, nr_pages, ctype);
4053

4054
	return memcg;
K
KAMEZAWA Hiroyuki 已提交
4055 4056 4057

unlock_out:
	unlock_page_cgroup(pc);
4058
	return NULL;
4059 4060
}

4061 4062
void mem_cgroup_uncharge_page(struct page *page)
{
4063 4064 4065
	/* early check. */
	if (page_mapped(page))
		return;
4066
	VM_BUG_ON_PAGE(page->mapping && !PageAnon(page), page);
4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078
	/*
	 * If the page is in swap cache, uncharge should be deferred
	 * to the swap path, which also properly accounts swap usage
	 * and handles memcg lifetime.
	 *
	 * Note that this check is not stable and reclaim may add the
	 * page to swap cache at any time after this.  However, if the
	 * page is not in swap cache by the time page->mapcount hits
	 * 0, there won't be any page table references to the swap
	 * slot, and reclaim will free it and not actually write the
	 * page to disk.
	 */
4079 4080
	if (PageSwapCache(page))
		return;
4081
	__mem_cgroup_uncharge_common(page, MEM_CGROUP_CHARGE_TYPE_ANON, false);
4082 4083 4084 4085
}

void mem_cgroup_uncharge_cache_page(struct page *page)
{
4086 4087
	VM_BUG_ON_PAGE(page_mapped(page), page);
	VM_BUG_ON_PAGE(page->mapping, page);
4088
	__mem_cgroup_uncharge_common(page, MEM_CGROUP_CHARGE_TYPE_CACHE, false);
4089 4090
}

4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103 4104
/*
 * Batch_start/batch_end is called in unmap_page_range/invlidate/trucate.
 * In that cases, pages are freed continuously and we can expect pages
 * are in the same memcg. All these calls itself limits the number of
 * pages freed at once, then uncharge_start/end() is called properly.
 * This may be called prural(2) times in a context,
 */

void mem_cgroup_uncharge_start(void)
{
	current->memcg_batch.do_batch++;
	/* We can do nest. */
	if (current->memcg_batch.do_batch == 1) {
		current->memcg_batch.memcg = NULL;
4105 4106
		current->memcg_batch.nr_pages = 0;
		current->memcg_batch.memsw_nr_pages = 0;
4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126
	}
}

void mem_cgroup_uncharge_end(void)
{
	struct memcg_batch_info *batch = &current->memcg_batch;

	if (!batch->do_batch)
		return;

	batch->do_batch--;
	if (batch->do_batch) /* If stacked, do nothing. */
		return;

	if (!batch->memcg)
		return;
	/*
	 * This "batch->memcg" is valid without any css_get/put etc...
	 * bacause we hide charges behind us.
	 */
4127 4128 4129 4130 4131 4132
	if (batch->nr_pages)
		res_counter_uncharge(&batch->memcg->res,
				     batch->nr_pages * PAGE_SIZE);
	if (batch->memsw_nr_pages)
		res_counter_uncharge(&batch->memcg->memsw,
				     batch->memsw_nr_pages * PAGE_SIZE);
4133
	memcg_oom_recover(batch->memcg);
4134 4135 4136 4137
	/* forget this pointer (for sanity check) */
	batch->memcg = NULL;
}

4138
#ifdef CONFIG_SWAP
4139
/*
4140
 * called after __delete_from_swap_cache() and drop "page" account.
4141 4142
 * memcg information is recorded to swap_cgroup of "ent"
 */
K
KAMEZAWA Hiroyuki 已提交
4143 4144
void
mem_cgroup_uncharge_swapcache(struct page *page, swp_entry_t ent, bool swapout)
4145 4146
{
	struct mem_cgroup *memcg;
K
KAMEZAWA Hiroyuki 已提交
4147 4148 4149 4150 4151
	int ctype = MEM_CGROUP_CHARGE_TYPE_SWAPOUT;

	if (!swapout) /* this was a swap cache but the swap is unused ! */
		ctype = MEM_CGROUP_CHARGE_TYPE_DROP;

4152
	memcg = __mem_cgroup_uncharge_common(page, ctype, false);
4153

K
KAMEZAWA Hiroyuki 已提交
4154 4155
	/*
	 * record memcg information,  if swapout && memcg != NULL,
L
Li Zefan 已提交
4156
	 * css_get() was called in uncharge().
K
KAMEZAWA Hiroyuki 已提交
4157 4158
	 */
	if (do_swap_account && swapout && memcg)
L
Li Zefan 已提交
4159
		swap_cgroup_record(ent, mem_cgroup_id(memcg));
4160
}
4161
#endif
4162

A
Andrew Morton 已提交
4163
#ifdef CONFIG_MEMCG_SWAP
4164 4165 4166 4167 4168
/*
 * called from swap_entry_free(). remove record in swap_cgroup and
 * uncharge "memsw" account.
 */
void mem_cgroup_uncharge_swap(swp_entry_t ent)
K
KAMEZAWA Hiroyuki 已提交
4169
{
4170
	struct mem_cgroup *memcg;
4171
	unsigned short id;
4172 4173 4174 4175

	if (!do_swap_account)
		return;

4176 4177 4178
	id = swap_cgroup_record(ent, 0);
	rcu_read_lock();
	memcg = mem_cgroup_lookup(id);
4179
	if (memcg) {
4180 4181 4182 4183
		/*
		 * We uncharge this because swap is freed.
		 * This memcg can be obsolete one. We avoid calling css_tryget
		 */
4184
		if (!mem_cgroup_is_root(memcg))
4185
			res_counter_uncharge(&memcg->memsw, PAGE_SIZE);
4186
		mem_cgroup_swap_statistics(memcg, false);
L
Li Zefan 已提交
4187
		css_put(&memcg->css);
4188
	}
4189
	rcu_read_unlock();
K
KAMEZAWA Hiroyuki 已提交
4190
}
4191 4192 4193 4194 4195 4196 4197 4198 4199 4200 4201 4202 4203 4204 4205 4206

/**
 * mem_cgroup_move_swap_account - move swap charge and swap_cgroup's record.
 * @entry: swap entry to be moved
 * @from:  mem_cgroup which the entry is moved from
 * @to:  mem_cgroup which the entry is moved to
 *
 * It succeeds only when the swap_cgroup's record for this entry is the same
 * as the mem_cgroup's id of @from.
 *
 * Returns 0 on success, -EINVAL on failure.
 *
 * The caller must have charged to @to, IOW, called res_counter_charge() about
 * both res and memsw, and called css_get().
 */
static int mem_cgroup_move_swap_account(swp_entry_t entry,
4207
				struct mem_cgroup *from, struct mem_cgroup *to)
4208 4209 4210
{
	unsigned short old_id, new_id;

L
Li Zefan 已提交
4211 4212
	old_id = mem_cgroup_id(from);
	new_id = mem_cgroup_id(to);
4213 4214 4215

	if (swap_cgroup_cmpxchg(entry, old_id, new_id) == old_id) {
		mem_cgroup_swap_statistics(from, false);
4216
		mem_cgroup_swap_statistics(to, true);
4217
		/*
4218 4219 4220
		 * This function is only called from task migration context now.
		 * It postpones res_counter and refcount handling till the end
		 * of task migration(mem_cgroup_clear_mc()) for performance
L
Li Zefan 已提交
4221 4222 4223 4224 4225 4226
		 * improvement. But we cannot postpone css_get(to)  because if
		 * the process that has been moved to @to does swap-in, the
		 * refcount of @to might be decreased to 0.
		 *
		 * We are in attach() phase, so the cgroup is guaranteed to be
		 * alive, so we can just call css_get().
4227
		 */
L
Li Zefan 已提交
4228
		css_get(&to->css);
4229 4230 4231 4232 4233 4234
		return 0;
	}
	return -EINVAL;
}
#else
static inline int mem_cgroup_move_swap_account(swp_entry_t entry,
4235
				struct mem_cgroup *from, struct mem_cgroup *to)
4236 4237 4238
{
	return -EINVAL;
}
4239
#endif
K
KAMEZAWA Hiroyuki 已提交
4240

4241
/*
4242 4243
 * Before starting migration, account PAGE_SIZE to mem_cgroup that the old
 * page belongs to.
4244
 */
4245 4246
void mem_cgroup_prepare_migration(struct page *page, struct page *newpage,
				  struct mem_cgroup **memcgp)
4247
{
4248
	struct mem_cgroup *memcg = NULL;
4249
	unsigned int nr_pages = 1;
4250
	struct page_cgroup *pc;
4251
	enum charge_type ctype;
4252

4253
	*memcgp = NULL;
4254

4255
	if (mem_cgroup_disabled())
4256
		return;
4257

4258 4259 4260
	if (PageTransHuge(page))
		nr_pages <<= compound_order(page);

4261 4262 4263
	pc = lookup_page_cgroup(page);
	lock_page_cgroup(pc);
	if (PageCgroupUsed(pc)) {
4264 4265
		memcg = pc->mem_cgroup;
		css_get(&memcg->css);
4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296
		/*
		 * At migrating an anonymous page, its mapcount goes down
		 * to 0 and uncharge() will be called. But, even if it's fully
		 * unmapped, migration may fail and this page has to be
		 * charged again. We set MIGRATION flag here and delay uncharge
		 * until end_migration() is called
		 *
		 * Corner Case Thinking
		 * A)
		 * When the old page was mapped as Anon and it's unmap-and-freed
		 * while migration was ongoing.
		 * If unmap finds the old page, uncharge() of it will be delayed
		 * until end_migration(). If unmap finds a new page, it's
		 * uncharged when it make mapcount to be 1->0. If unmap code
		 * finds swap_migration_entry, the new page will not be mapped
		 * and end_migration() will find it(mapcount==0).
		 *
		 * B)
		 * When the old page was mapped but migraion fails, the kernel
		 * remaps it. A charge for it is kept by MIGRATION flag even
		 * if mapcount goes down to 0. We can do remap successfully
		 * without charging it again.
		 *
		 * C)
		 * The "old" page is under lock_page() until the end of
		 * migration, so, the old page itself will not be swapped-out.
		 * If the new page is swapped out before end_migraton, our
		 * hook to usual swap-out path will catch the event.
		 */
		if (PageAnon(page))
			SetPageCgroupMigration(pc);
4297
	}
4298
	unlock_page_cgroup(pc);
4299 4300 4301 4302
	/*
	 * If the page is not charged at this point,
	 * we return here.
	 */
4303
	if (!memcg)
4304
		return;
4305

4306
	*memcgp = memcg;
4307 4308 4309 4310 4311 4312 4313
	/*
	 * We charge new page before it's used/mapped. So, even if unlock_page()
	 * is called before end_migration, we can catch all events on this new
	 * page. In the case new page is migrated but not remapped, new page's
	 * mapcount will be finally 0 and we call uncharge in end_migration().
	 */
	if (PageAnon(page))
4314
		ctype = MEM_CGROUP_CHARGE_TYPE_ANON;
4315
	else
4316
		ctype = MEM_CGROUP_CHARGE_TYPE_CACHE;
4317 4318 4319 4320 4321
	/*
	 * The page is committed to the memcg, but it's not actually
	 * charged to the res_counter since we plan on replacing the
	 * old one and only one page is going to be left afterwards.
	 */
4322
	__mem_cgroup_commit_charge(memcg, newpage, nr_pages, ctype, false);
4323
}
4324

4325
/* remove redundant charge if migration failed*/
4326
void mem_cgroup_end_migration(struct mem_cgroup *memcg,
4327
	struct page *oldpage, struct page *newpage, bool migration_ok)
4328
{
4329
	struct page *used, *unused;
4330
	struct page_cgroup *pc;
4331
	bool anon;
4332

4333
	if (!memcg)
4334
		return;
4335

4336
	if (!migration_ok) {
4337 4338
		used = oldpage;
		unused = newpage;
4339
	} else {
4340
		used = newpage;
4341 4342
		unused = oldpage;
	}
4343
	anon = PageAnon(used);
4344 4345 4346 4347
	__mem_cgroup_uncharge_common(unused,
				     anon ? MEM_CGROUP_CHARGE_TYPE_ANON
				     : MEM_CGROUP_CHARGE_TYPE_CACHE,
				     true);
4348
	css_put(&memcg->css);
4349
	/*
4350 4351 4352
	 * We disallowed uncharge of pages under migration because mapcount
	 * of the page goes down to zero, temporarly.
	 * Clear the flag and check the page should be charged.
4353
	 */
4354 4355 4356 4357 4358
	pc = lookup_page_cgroup(oldpage);
	lock_page_cgroup(pc);
	ClearPageCgroupMigration(pc);
	unlock_page_cgroup(pc);

4359
	/*
4360 4361 4362 4363 4364 4365
	 * If a page is a file cache, radix-tree replacement is very atomic
	 * and we can skip this check. When it was an Anon page, its mapcount
	 * goes down to 0. But because we added MIGRATION flage, it's not
	 * uncharged yet. There are several case but page->mapcount check
	 * and USED bit check in mem_cgroup_uncharge_page() will do enough
	 * check. (see prepare_charge() also)
4366
	 */
4367
	if (anon)
4368
		mem_cgroup_uncharge_page(used);
4369
}
4370

4371 4372 4373 4374 4375 4376 4377 4378
/*
 * At replace page cache, newpage is not under any memcg but it's on
 * LRU. So, this function doesn't touch res_counter but handles LRU
 * in correct way. Both pages are locked so we cannot race with uncharge.
 */
void mem_cgroup_replace_page_cache(struct page *oldpage,
				  struct page *newpage)
{
4379
	struct mem_cgroup *memcg = NULL;
4380 4381 4382 4383 4384 4385 4386 4387 4388
	struct page_cgroup *pc;
	enum charge_type type = MEM_CGROUP_CHARGE_TYPE_CACHE;

	if (mem_cgroup_disabled())
		return;

	pc = lookup_page_cgroup(oldpage);
	/* fix accounting on old pages */
	lock_page_cgroup(pc);
4389 4390
	if (PageCgroupUsed(pc)) {
		memcg = pc->mem_cgroup;
4391
		mem_cgroup_charge_statistics(memcg, oldpage, false, -1);
4392 4393
		ClearPageCgroupUsed(pc);
	}
4394 4395
	unlock_page_cgroup(pc);

4396 4397 4398 4399 4400 4401
	/*
	 * When called from shmem_replace_page(), in some cases the
	 * oldpage has already been charged, and in some cases not.
	 */
	if (!memcg)
		return;
4402 4403 4404 4405 4406
	/*
	 * Even if newpage->mapping was NULL before starting replacement,
	 * the newpage may be on LRU(or pagevec for LRU) already. We lock
	 * LRU while we overwrite pc->mem_cgroup.
	 */
4407
	__mem_cgroup_commit_charge(memcg, newpage, 1, type, true);
4408 4409
}

4410 4411 4412 4413 4414 4415
#ifdef CONFIG_DEBUG_VM
static struct page_cgroup *lookup_page_cgroup_used(struct page *page)
{
	struct page_cgroup *pc;

	pc = lookup_page_cgroup(page);
4416 4417 4418 4419 4420
	/*
	 * Can be NULL while feeding pages into the page allocator for
	 * the first time, i.e. during boot or memory hotplug;
	 * or when mem_cgroup_disabled().
	 */
4421 4422 4423 4424 4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439
	if (likely(pc) && PageCgroupUsed(pc))
		return pc;
	return NULL;
}

bool mem_cgroup_bad_page_check(struct page *page)
{
	if (mem_cgroup_disabled())
		return false;

	return lookup_page_cgroup_used(page) != NULL;
}

void mem_cgroup_print_bad_page(struct page *page)
{
	struct page_cgroup *pc;

	pc = lookup_page_cgroup_used(page);
	if (pc) {
4440 4441
		pr_alert("pc:%p pc->flags:%lx pc->mem_cgroup:%p\n",
			 pc, pc->flags, pc->mem_cgroup);
4442 4443 4444 4445
	}
}
#endif

4446
static int mem_cgroup_resize_limit(struct mem_cgroup *memcg,
4447
				unsigned long long val)
4448
{
4449
	int retry_count;
4450
	u64 memswlimit, memlimit;
4451
	int ret = 0;
4452 4453
	int children = mem_cgroup_count_children(memcg);
	u64 curusage, oldusage;
4454
	int enlarge;
4455 4456 4457 4458 4459 4460 4461 4462 4463

	/*
	 * For keeping hierarchical_reclaim simple, how long we should retry
	 * is depends on callers. We set our retry-count to be function
	 * of # of children which we should visit in this loop.
	 */
	retry_count = MEM_CGROUP_RECLAIM_RETRIES * children;

	oldusage = res_counter_read_u64(&memcg->res, RES_USAGE);
4464

4465
	enlarge = 0;
4466
	while (retry_count) {
4467 4468 4469 4470
		if (signal_pending(current)) {
			ret = -EINTR;
			break;
		}
4471 4472 4473
		/*
		 * Rather than hide all in some function, I do this in
		 * open coded manner. You see what this really does.
4474
		 * We have to guarantee memcg->res.limit <= memcg->memsw.limit.
4475 4476 4477 4478 4479 4480
		 */
		mutex_lock(&set_limit_mutex);
		memswlimit = res_counter_read_u64(&memcg->memsw, RES_LIMIT);
		if (memswlimit < val) {
			ret = -EINVAL;
			mutex_unlock(&set_limit_mutex);
4481 4482
			break;
		}
4483 4484 4485 4486 4487

		memlimit = res_counter_read_u64(&memcg->res, RES_LIMIT);
		if (memlimit < val)
			enlarge = 1;

4488
		ret = res_counter_set_limit(&memcg->res, val);
4489 4490 4491 4492 4493 4494
		if (!ret) {
			if (memswlimit == val)
				memcg->memsw_is_minimum = true;
			else
				memcg->memsw_is_minimum = false;
		}
4495 4496 4497 4498 4499
		mutex_unlock(&set_limit_mutex);

		if (!ret)
			break;

4500 4501
		mem_cgroup_reclaim(memcg, GFP_KERNEL,
				   MEM_CGROUP_RECLAIM_SHRINK);
4502 4503
		curusage = res_counter_read_u64(&memcg->res, RES_USAGE);
		/* Usage is reduced ? */
A
Andrew Morton 已提交
4504
		if (curusage >= oldusage)
4505 4506 4507
			retry_count--;
		else
			oldusage = curusage;
4508
	}
4509 4510
	if (!ret && enlarge)
		memcg_oom_recover(memcg);
4511

4512 4513 4514
	return ret;
}

L
Li Zefan 已提交
4515 4516
static int mem_cgroup_resize_memsw_limit(struct mem_cgroup *memcg,
					unsigned long long val)
4517
{
4518
	int retry_count;
4519
	u64 memlimit, memswlimit, oldusage, curusage;
4520 4521
	int children = mem_cgroup_count_children(memcg);
	int ret = -EBUSY;
4522
	int enlarge = 0;
4523

4524
	/* see mem_cgroup_resize_res_limit */
A
Andrew Morton 已提交
4525
	retry_count = children * MEM_CGROUP_RECLAIM_RETRIES;
4526
	oldusage = res_counter_read_u64(&memcg->memsw, RES_USAGE);
4527 4528 4529 4530 4531 4532 4533 4534
	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.
4535
		 * We have to guarantee memcg->res.limit <= memcg->memsw.limit.
4536 4537 4538 4539 4540 4541 4542 4543
		 */
		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;
		}
4544 4545 4546
		memswlimit = res_counter_read_u64(&memcg->memsw, RES_LIMIT);
		if (memswlimit < val)
			enlarge = 1;
4547
		ret = res_counter_set_limit(&memcg->memsw, val);
4548 4549 4550 4551 4552 4553
		if (!ret) {
			if (memlimit == val)
				memcg->memsw_is_minimum = true;
			else
				memcg->memsw_is_minimum = false;
		}
4554 4555 4556 4557 4558
		mutex_unlock(&set_limit_mutex);

		if (!ret)
			break;

4559 4560 4561
		mem_cgroup_reclaim(memcg, GFP_KERNEL,
				   MEM_CGROUP_RECLAIM_NOSWAP |
				   MEM_CGROUP_RECLAIM_SHRINK);
4562
		curusage = res_counter_read_u64(&memcg->memsw, RES_USAGE);
4563
		/* Usage is reduced ? */
4564
		if (curusage >= oldusage)
4565
			retry_count--;
4566 4567
		else
			oldusage = curusage;
4568
	}
4569 4570
	if (!ret && enlarge)
		memcg_oom_recover(memcg);
4571 4572 4573
	return ret;
}

4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665
unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
					    gfp_t gfp_mask,
					    unsigned long *total_scanned)
{
	unsigned long nr_reclaimed = 0;
	struct mem_cgroup_per_zone *mz, *next_mz = NULL;
	unsigned long reclaimed;
	int loop = 0;
	struct mem_cgroup_tree_per_zone *mctz;
	unsigned long long excess;
	unsigned long nr_scanned;

	if (order > 0)
		return 0;

	mctz = soft_limit_tree_node_zone(zone_to_nid(zone), zone_idx(zone));
	/*
	 * This loop can run a while, specially if mem_cgroup's continuously
	 * keep exceeding their soft limit and putting the system under
	 * pressure
	 */
	do {
		if (next_mz)
			mz = next_mz;
		else
			mz = mem_cgroup_largest_soft_limit_node(mctz);
		if (!mz)
			break;

		nr_scanned = 0;
		reclaimed = mem_cgroup_soft_reclaim(mz->memcg, zone,
						    gfp_mask, &nr_scanned);
		nr_reclaimed += reclaimed;
		*total_scanned += nr_scanned;
		spin_lock(&mctz->lock);

		/*
		 * If we failed to reclaim anything from this memory cgroup
		 * it is time to move on to the next cgroup
		 */
		next_mz = NULL;
		if (!reclaimed) {
			do {
				/*
				 * Loop until we find yet another one.
				 *
				 * By the time we get the soft_limit lock
				 * again, someone might have aded the
				 * group back on the RB tree. Iterate to
				 * make sure we get a different mem.
				 * mem_cgroup_largest_soft_limit_node returns
				 * NULL if no other cgroup is present on
				 * the tree
				 */
				next_mz =
				__mem_cgroup_largest_soft_limit_node(mctz);
				if (next_mz == mz)
					css_put(&next_mz->memcg->css);
				else /* next_mz == NULL or other memcg */
					break;
			} while (1);
		}
		__mem_cgroup_remove_exceeded(mz->memcg, mz, mctz);
		excess = res_counter_soft_limit_excess(&mz->memcg->res);
		/*
		 * One school of thought says that we should not add
		 * back the node to the tree if reclaim returns 0.
		 * But our reclaim could return 0, simply because due
		 * to priority we are exposing a smaller subset of
		 * memory to reclaim from. Consider this as a longer
		 * term TODO.
		 */
		/* If excess == 0, no tree ops */
		__mem_cgroup_insert_exceeded(mz->memcg, mz, mctz, excess);
		spin_unlock(&mctz->lock);
		css_put(&mz->memcg->css);
		loop++;
		/*
		 * Could not reclaim anything and there are no more
		 * mem cgroups to try or we seem to be looping without
		 * reclaiming anything.
		 */
		if (!nr_reclaimed &&
			(next_mz == NULL ||
			loop > MEM_CGROUP_MAX_SOFT_LIMIT_RECLAIM_LOOPS))
			break;
	} while (!nr_reclaimed);
	if (next_mz)
		css_put(&next_mz->memcg->css);
	return nr_reclaimed;
}

4666 4667 4668 4669 4670 4671 4672
/**
 * mem_cgroup_force_empty_list - clears LRU of a group
 * @memcg: group to clear
 * @node: NUMA node
 * @zid: zone id
 * @lru: lru to to clear
 *
4673
 * Traverse a specified page_cgroup list and try to drop them all.  This doesn't
4674 4675
 * reclaim the pages page themselves - pages are moved to the parent (or root)
 * group.
4676
 */
4677
static void mem_cgroup_force_empty_list(struct mem_cgroup *memcg,
K
KAMEZAWA Hiroyuki 已提交
4678
				int node, int zid, enum lru_list lru)
4679
{
4680
	struct lruvec *lruvec;
4681
	unsigned long flags;
4682
	struct list_head *list;
4683 4684
	struct page *busy;
	struct zone *zone;
4685

K
KAMEZAWA Hiroyuki 已提交
4686
	zone = &NODE_DATA(node)->node_zones[zid];
4687 4688
	lruvec = mem_cgroup_zone_lruvec(zone, memcg);
	list = &lruvec->lists[lru];
4689

4690
	busy = NULL;
4691
	do {
4692
		struct page_cgroup *pc;
4693 4694
		struct page *page;

K
KAMEZAWA Hiroyuki 已提交
4695
		spin_lock_irqsave(&zone->lru_lock, flags);
4696
		if (list_empty(list)) {
K
KAMEZAWA Hiroyuki 已提交
4697
			spin_unlock_irqrestore(&zone->lru_lock, flags);
4698
			break;
4699
		}
4700 4701 4702
		page = list_entry(list->prev, struct page, lru);
		if (busy == page) {
			list_move(&page->lru, list);
4703
			busy = NULL;
K
KAMEZAWA Hiroyuki 已提交
4704
			spin_unlock_irqrestore(&zone->lru_lock, flags);
4705 4706
			continue;
		}
K
KAMEZAWA Hiroyuki 已提交
4707
		spin_unlock_irqrestore(&zone->lru_lock, flags);
4708

4709
		pc = lookup_page_cgroup(page);
4710

4711
		if (mem_cgroup_move_parent(page, pc, memcg)) {
4712
			/* found lock contention or "pc" is obsolete. */
4713
			busy = page;
4714 4715 4716
			cond_resched();
		} else
			busy = NULL;
4717
	} while (!list_empty(list));
4718 4719 4720
}

/*
4721 4722
 * make mem_cgroup's charge to be 0 if there is no task by moving
 * all the charges and pages to the parent.
4723
 * This enables deleting this mem_cgroup.
4724 4725
 *
 * Caller is responsible for holding css reference on the memcg.
4726
 */
4727
static void mem_cgroup_reparent_charges(struct mem_cgroup *memcg)
4728
{
4729
	int node, zid;
4730
	u64 usage;
4731

4732
	do {
4733 4734
		/* This is for making all *used* pages to be on LRU. */
		lru_add_drain_all();
4735 4736
		drain_all_stock_sync(memcg);
		mem_cgroup_start_move(memcg);
4737
		for_each_node_state(node, N_MEMORY) {
4738
			for (zid = 0; zid < MAX_NR_ZONES; zid++) {
H
Hugh Dickins 已提交
4739 4740
				enum lru_list lru;
				for_each_lru(lru) {
4741
					mem_cgroup_force_empty_list(memcg,
H
Hugh Dickins 已提交
4742
							node, zid, lru);
4743
				}
4744
			}
4745
		}
4746 4747
		mem_cgroup_end_move(memcg);
		memcg_oom_recover(memcg);
4748
		cond_resched();
4749

4750
		/*
4751 4752 4753 4754 4755
		 * Kernel memory may not necessarily be trackable to a specific
		 * process. So they are not migrated, and therefore we can't
		 * expect their value to drop to 0 here.
		 * Having res filled up with kmem only is enough.
		 *
4756 4757 4758 4759 4760 4761
		 * This is a safety check because mem_cgroup_force_empty_list
		 * could have raced with mem_cgroup_replace_page_cache callers
		 * so the lru seemed empty but the page could have been added
		 * right after the check. RES_USAGE should be safe as we always
		 * charge before adding to the LRU.
		 */
4762 4763 4764
		usage = res_counter_read_u64(&memcg->res, RES_USAGE) -
			res_counter_read_u64(&memcg->kmem, RES_USAGE);
	} while (usage > 0);
4765 4766
}

4767 4768
static inline bool memcg_has_children(struct mem_cgroup *memcg)
{
4769 4770 4771 4772 4773 4774 4775 4776 4777 4778
	lockdep_assert_held(&memcg_create_mutex);
	/*
	 * The lock does not prevent addition or deletion to the list
	 * of children, but it prevents a new child from being
	 * initialized based on this parent in css_online(), so it's
	 * enough to decide whether hierarchically inherited
	 * attributes can still be changed or not.
	 */
	return memcg->use_hierarchy &&
		!list_empty(&memcg->css.cgroup->children);
4779 4780
}

4781 4782 4783 4784 4785 4786 4787 4788 4789 4790
/*
 * Reclaims as many pages from the given memcg as possible and moves
 * the rest to the parent.
 *
 * Caller is responsible for holding css reference for memcg.
 */
static int mem_cgroup_force_empty(struct mem_cgroup *memcg)
{
	int nr_retries = MEM_CGROUP_RECLAIM_RETRIES;
	struct cgroup *cgrp = memcg->css.cgroup;
4791

4792
	/* returns EBUSY if there is a task or if we come here twice. */
4793
	if (cgroup_has_tasks(cgrp) || !list_empty(&cgrp->children))
4794 4795
		return -EBUSY;

4796 4797
	/* we call try-to-free pages for make this cgroup empty */
	lru_add_drain_all();
4798
	/* try to free all pages in this cgroup */
4799
	while (nr_retries && res_counter_read_u64(&memcg->res, RES_USAGE) > 0) {
4800
		int progress;
4801

4802 4803 4804
		if (signal_pending(current))
			return -EINTR;

4805
		progress = try_to_free_mem_cgroup_pages(memcg, GFP_KERNEL,
4806
						false);
4807
		if (!progress) {
4808
			nr_retries--;
4809
			/* maybe some writeback is necessary */
4810
			congestion_wait(BLK_RW_ASYNC, HZ/10);
4811
		}
4812 4813

	}
K
KAMEZAWA Hiroyuki 已提交
4814
	lru_add_drain();
4815 4816 4817
	mem_cgroup_reparent_charges(memcg);

	return 0;
4818 4819
}

4820 4821
static int mem_cgroup_force_empty_write(struct cgroup_subsys_state *css,
					unsigned int event)
4822
{
4823
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
4824

4825 4826
	if (mem_cgroup_is_root(memcg))
		return -EINVAL;
4827
	return mem_cgroup_force_empty(memcg);
4828 4829
}

4830 4831
static u64 mem_cgroup_hierarchy_read(struct cgroup_subsys_state *css,
				     struct cftype *cft)
4832
{
4833
	return mem_cgroup_from_css(css)->use_hierarchy;
4834 4835
}

4836 4837
static int mem_cgroup_hierarchy_write(struct cgroup_subsys_state *css,
				      struct cftype *cft, u64 val)
4838 4839
{
	int retval = 0;
4840
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
T
Tejun Heo 已提交
4841
	struct mem_cgroup *parent_memcg = mem_cgroup_from_css(css_parent(&memcg->css));
4842

4843
	mutex_lock(&memcg_create_mutex);
4844 4845 4846 4847

	if (memcg->use_hierarchy == val)
		goto out;

4848
	/*
4849
	 * If parent's use_hierarchy is set, we can't make any modifications
4850 4851 4852 4853 4854 4855
	 * 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.
	 */
4856
	if ((!parent_memcg || !parent_memcg->use_hierarchy) &&
4857
				(val == 1 || val == 0)) {
4858
		if (list_empty(&memcg->css.cgroup->children))
4859
			memcg->use_hierarchy = val;
4860 4861 4862 4863
		else
			retval = -EBUSY;
	} else
		retval = -EINVAL;
4864 4865

out:
4866
	mutex_unlock(&memcg_create_mutex);
4867 4868 4869 4870

	return retval;
}

4871

4872
static unsigned long mem_cgroup_recursive_stat(struct mem_cgroup *memcg,
4873
					       enum mem_cgroup_stat_index idx)
4874
{
K
KAMEZAWA Hiroyuki 已提交
4875
	struct mem_cgroup *iter;
4876
	long val = 0;
4877

4878
	/* Per-cpu values can be negative, use a signed accumulator */
4879
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
4880 4881 4882 4883 4884
		val += mem_cgroup_read_stat(iter, idx);

	if (val < 0) /* race ? */
		val = 0;
	return val;
4885 4886
}

4887
static inline u64 mem_cgroup_usage(struct mem_cgroup *memcg, bool swap)
4888
{
K
KAMEZAWA Hiroyuki 已提交
4889
	u64 val;
4890

4891
	if (!mem_cgroup_is_root(memcg)) {
4892
		if (!swap)
4893
			return res_counter_read_u64(&memcg->res, RES_USAGE);
4894
		else
4895
			return res_counter_read_u64(&memcg->memsw, RES_USAGE);
4896 4897
	}

4898 4899 4900 4901
	/*
	 * Transparent hugepages are still accounted for in MEM_CGROUP_STAT_RSS
	 * as well as in MEM_CGROUP_STAT_RSS_HUGE.
	 */
4902 4903
	val = mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_CACHE);
	val += mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_RSS);
4904

K
KAMEZAWA Hiroyuki 已提交
4905
	if (swap)
4906
		val += mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_SWAP);
4907 4908 4909 4910

	return val << PAGE_SHIFT;
}

4911 4912
static u64 mem_cgroup_read_u64(struct cgroup_subsys_state *css,
				   struct cftype *cft)
B
Balbir Singh 已提交
4913
{
4914
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
4915
	u64 val;
4916
	int name;
G
Glauber Costa 已提交
4917
	enum res_type type;
4918 4919 4920

	type = MEMFILE_TYPE(cft->private);
	name = MEMFILE_ATTR(cft->private);
4921

4922 4923
	switch (type) {
	case _MEM:
4924
		if (name == RES_USAGE)
4925
			val = mem_cgroup_usage(memcg, false);
4926
		else
4927
			val = res_counter_read_u64(&memcg->res, name);
4928 4929
		break;
	case _MEMSWAP:
4930
		if (name == RES_USAGE)
4931
			val = mem_cgroup_usage(memcg, true);
4932
		else
4933
			val = res_counter_read_u64(&memcg->memsw, name);
4934
		break;
4935 4936 4937
	case _KMEM:
		val = res_counter_read_u64(&memcg->kmem, name);
		break;
4938 4939 4940
	default:
		BUG();
	}
4941

4942
	return val;
B
Balbir Singh 已提交
4943
}
4944 4945

#ifdef CONFIG_MEMCG_KMEM
4946 4947 4948 4949 4950 4951 4952 4953 4954 4955 4956 4957 4958 4959 4960 4961
/* should be called with activate_kmem_mutex held */
static int __memcg_activate_kmem(struct mem_cgroup *memcg,
				 unsigned long long limit)
{
	int err = 0;
	int memcg_id;

	if (memcg_kmem_is_active(memcg))
		return 0;

	/*
	 * We are going to allocate memory for data shared by all memory
	 * cgroups so let's stop accounting here.
	 */
	memcg_stop_kmem_account();

4962 4963 4964 4965 4966 4967 4968 4969 4970 4971 4972 4973
	/*
	 * For simplicity, we won't allow this to be disabled.  It also can't
	 * be changed if the cgroup has children already, or if tasks had
	 * already joined.
	 *
	 * If tasks join before we set the limit, a person looking at
	 * kmem.usage_in_bytes will have no way to determine when it took
	 * place, which makes the value quite meaningless.
	 *
	 * After it first became limited, changes in the value of the limit are
	 * of course permitted.
	 */
4974
	mutex_lock(&memcg_create_mutex);
4975
	if (cgroup_has_tasks(memcg->css.cgroup) || memcg_has_children(memcg))
4976 4977 4978 4979
		err = -EBUSY;
	mutex_unlock(&memcg_create_mutex);
	if (err)
		goto out;
4980

4981 4982 4983 4984 4985 4986 4987 4988 4989 4990 4991
	memcg_id = ida_simple_get(&kmem_limited_groups,
				  0, MEMCG_CACHES_MAX_SIZE, GFP_KERNEL);
	if (memcg_id < 0) {
		err = memcg_id;
		goto out;
	}

	/*
	 * Make sure we have enough space for this cgroup in each root cache's
	 * memcg_params.
	 */
4992
	mutex_lock(&memcg_slab_mutex);
4993
	err = memcg_update_all_caches(memcg_id + 1);
4994
	mutex_unlock(&memcg_slab_mutex);
4995 4996 4997 4998 4999 5000 5001 5002 5003 5004 5005 5006 5007 5008 5009 5010 5011 5012 5013 5014
	if (err)
		goto out_rmid;

	memcg->kmemcg_id = memcg_id;
	INIT_LIST_HEAD(&memcg->memcg_slab_caches);

	/*
	 * We couldn't have accounted to this cgroup, because it hasn't got the
	 * active bit set yet, so this should succeed.
	 */
	err = res_counter_set_limit(&memcg->kmem, limit);
	VM_BUG_ON(err);

	static_key_slow_inc(&memcg_kmem_enabled_key);
	/*
	 * Setting the active bit after enabling static branching will
	 * guarantee no one starts accounting before all call sites are
	 * patched.
	 */
	memcg_kmem_set_active(memcg);
5015
out:
5016 5017 5018 5019 5020 5021 5022 5023 5024 5025 5026 5027 5028 5029 5030 5031 5032 5033 5034 5035 5036 5037 5038 5039 5040 5041 5042 5043
	memcg_resume_kmem_account();
	return err;

out_rmid:
	ida_simple_remove(&kmem_limited_groups, memcg_id);
	goto out;
}

static int memcg_activate_kmem(struct mem_cgroup *memcg,
			       unsigned long long limit)
{
	int ret;

	mutex_lock(&activate_kmem_mutex);
	ret = __memcg_activate_kmem(memcg, limit);
	mutex_unlock(&activate_kmem_mutex);
	return ret;
}

static int memcg_update_kmem_limit(struct mem_cgroup *memcg,
				   unsigned long long val)
{
	int ret;

	if (!memcg_kmem_is_active(memcg))
		ret = memcg_activate_kmem(memcg, val);
	else
		ret = res_counter_set_limit(&memcg->kmem, val);
5044 5045 5046
	return ret;
}

5047
static int memcg_propagate_kmem(struct mem_cgroup *memcg)
5048
{
5049
	int ret = 0;
5050
	struct mem_cgroup *parent = parent_mem_cgroup(memcg);
5051

5052 5053
	if (!parent)
		return 0;
5054

5055
	mutex_lock(&activate_kmem_mutex);
5056
	/*
5057 5058
	 * If the parent cgroup is not kmem-active now, it cannot be activated
	 * after this point, because it has at least one child already.
5059
	 */
5060 5061 5062
	if (memcg_kmem_is_active(parent))
		ret = __memcg_activate_kmem(memcg, RES_COUNTER_MAX);
	mutex_unlock(&activate_kmem_mutex);
5063
	return ret;
5064
}
5065 5066 5067 5068 5069 5070
#else
static int memcg_update_kmem_limit(struct mem_cgroup *memcg,
				   unsigned long long val)
{
	return -EINVAL;
}
5071
#endif /* CONFIG_MEMCG_KMEM */
5072

5073 5074 5075 5076
/*
 * The user of this function is...
 * RES_LIMIT.
 */
5077
static int mem_cgroup_write(struct cgroup_subsys_state *css, struct cftype *cft,
5078
			    char *buffer)
B
Balbir Singh 已提交
5079
{
5080
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
G
Glauber Costa 已提交
5081 5082
	enum res_type type;
	int name;
5083 5084 5085
	unsigned long long val;
	int ret;

5086 5087
	type = MEMFILE_TYPE(cft->private);
	name = MEMFILE_ATTR(cft->private);
5088

5089
	switch (name) {
5090
	case RES_LIMIT:
5091 5092 5093 5094
		if (mem_cgroup_is_root(memcg)) { /* Can't set limit on root */
			ret = -EINVAL;
			break;
		}
5095 5096
		/* This function does all necessary parse...reuse it */
		ret = res_counter_memparse_write_strategy(buffer, &val);
5097 5098 5099
		if (ret)
			break;
		if (type == _MEM)
5100
			ret = mem_cgroup_resize_limit(memcg, val);
5101
		else if (type == _MEMSWAP)
5102
			ret = mem_cgroup_resize_memsw_limit(memcg, val);
5103
		else if (type == _KMEM)
5104
			ret = memcg_update_kmem_limit(memcg, val);
5105 5106
		else
			return -EINVAL;
5107
		break;
5108 5109 5110 5111 5112 5113 5114 5115 5116 5117 5118 5119 5120 5121
	case RES_SOFT_LIMIT:
		ret = res_counter_memparse_write_strategy(buffer, &val);
		if (ret)
			break;
		/*
		 * For memsw, soft limits are hard to implement in terms
		 * of semantics, for now, we support soft limits for
		 * control without swap
		 */
		if (type == _MEM)
			ret = res_counter_set_soft_limit(&memcg->res, val);
		else
			ret = -EINVAL;
		break;
5122 5123 5124 5125 5126
	default:
		ret = -EINVAL; /* should be BUG() ? */
		break;
	}
	return ret;
B
Balbir Singh 已提交
5127 5128
}

5129 5130 5131 5132 5133 5134 5135 5136 5137 5138
static void memcg_get_hierarchical_limit(struct mem_cgroup *memcg,
		unsigned long long *mem_limit, unsigned long long *memsw_limit)
{
	unsigned long long min_limit, min_memsw_limit, tmp;

	min_limit = res_counter_read_u64(&memcg->res, RES_LIMIT);
	min_memsw_limit = res_counter_read_u64(&memcg->memsw, RES_LIMIT);
	if (!memcg->use_hierarchy)
		goto out;

T
Tejun Heo 已提交
5139 5140
	while (css_parent(&memcg->css)) {
		memcg = mem_cgroup_from_css(css_parent(&memcg->css));
5141 5142 5143 5144 5145 5146 5147 5148 5149 5150 5151 5152
		if (!memcg->use_hierarchy)
			break;
		tmp = res_counter_read_u64(&memcg->res, RES_LIMIT);
		min_limit = min(min_limit, tmp);
		tmp = res_counter_read_u64(&memcg->memsw, RES_LIMIT);
		min_memsw_limit = min(min_memsw_limit, tmp);
	}
out:
	*mem_limit = min_limit;
	*memsw_limit = min_memsw_limit;
}

5153
static int mem_cgroup_reset(struct cgroup_subsys_state *css, unsigned int event)
5154
{
5155
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
G
Glauber Costa 已提交
5156 5157
	int name;
	enum res_type type;
5158

5159 5160
	type = MEMFILE_TYPE(event);
	name = MEMFILE_ATTR(event);
5161

5162
	switch (name) {
5163
	case RES_MAX_USAGE:
5164
		if (type == _MEM)
5165
			res_counter_reset_max(&memcg->res);
5166
		else if (type == _MEMSWAP)
5167
			res_counter_reset_max(&memcg->memsw);
5168 5169 5170 5171
		else if (type == _KMEM)
			res_counter_reset_max(&memcg->kmem);
		else
			return -EINVAL;
5172 5173
		break;
	case RES_FAILCNT:
5174
		if (type == _MEM)
5175
			res_counter_reset_failcnt(&memcg->res);
5176
		else if (type == _MEMSWAP)
5177
			res_counter_reset_failcnt(&memcg->memsw);
5178 5179 5180 5181
		else if (type == _KMEM)
			res_counter_reset_failcnt(&memcg->kmem);
		else
			return -EINVAL;
5182 5183
		break;
	}
5184

5185
	return 0;
5186 5187
}

5188
static u64 mem_cgroup_move_charge_read(struct cgroup_subsys_state *css,
5189 5190
					struct cftype *cft)
{
5191
	return mem_cgroup_from_css(css)->move_charge_at_immigrate;
5192 5193
}

5194
#ifdef CONFIG_MMU
5195
static int mem_cgroup_move_charge_write(struct cgroup_subsys_state *css,
5196 5197
					struct cftype *cft, u64 val)
{
5198
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
5199 5200 5201

	if (val >= (1 << NR_MOVE_TYPE))
		return -EINVAL;
5202

5203
	/*
5204 5205 5206 5207
	 * No kind of locking is needed in here, because ->can_attach() will
	 * check this value once in the beginning of the process, and then carry
	 * on with stale data. This means that changes to this value will only
	 * affect task migrations starting after the change.
5208
	 */
5209
	memcg->move_charge_at_immigrate = val;
5210 5211
	return 0;
}
5212
#else
5213
static int mem_cgroup_move_charge_write(struct cgroup_subsys_state *css,
5214 5215 5216 5217 5218
					struct cftype *cft, u64 val)
{
	return -ENOSYS;
}
#endif
5219

5220
#ifdef CONFIG_NUMA
5221
static int memcg_numa_stat_show(struct seq_file *m, void *v)
5222
{
5223 5224 5225 5226 5227 5228 5229 5230 5231 5232 5233 5234
	struct numa_stat {
		const char *name;
		unsigned int lru_mask;
	};

	static const struct numa_stat stats[] = {
		{ "total", LRU_ALL },
		{ "file", LRU_ALL_FILE },
		{ "anon", LRU_ALL_ANON },
		{ "unevictable", BIT(LRU_UNEVICTABLE) },
	};
	const struct numa_stat *stat;
5235
	int nid;
5236
	unsigned long nr;
5237
	struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(m));
5238

5239 5240 5241 5242 5243 5244 5245 5246 5247
	for (stat = stats; stat < stats + ARRAY_SIZE(stats); stat++) {
		nr = mem_cgroup_nr_lru_pages(memcg, stat->lru_mask);
		seq_printf(m, "%s=%lu", stat->name, nr);
		for_each_node_state(nid, N_MEMORY) {
			nr = mem_cgroup_node_nr_lru_pages(memcg, nid,
							  stat->lru_mask);
			seq_printf(m, " N%d=%lu", nid, nr);
		}
		seq_putc(m, '\n');
5248 5249
	}

5250 5251 5252 5253 5254 5255 5256 5257 5258 5259 5260 5261 5262 5263 5264
	for (stat = stats; stat < stats + ARRAY_SIZE(stats); stat++) {
		struct mem_cgroup *iter;

		nr = 0;
		for_each_mem_cgroup_tree(iter, memcg)
			nr += mem_cgroup_nr_lru_pages(iter, stat->lru_mask);
		seq_printf(m, "hierarchical_%s=%lu", stat->name, nr);
		for_each_node_state(nid, N_MEMORY) {
			nr = 0;
			for_each_mem_cgroup_tree(iter, memcg)
				nr += mem_cgroup_node_nr_lru_pages(
					iter, nid, stat->lru_mask);
			seq_printf(m, " N%d=%lu", nid, nr);
		}
		seq_putc(m, '\n');
5265 5266 5267 5268 5269 5270
	}

	return 0;
}
#endif /* CONFIG_NUMA */

5271 5272 5273 5274 5275
static inline void mem_cgroup_lru_names_not_uptodate(void)
{
	BUILD_BUG_ON(ARRAY_SIZE(mem_cgroup_lru_names) != NR_LRU_LISTS);
}

5276
static int memcg_stat_show(struct seq_file *m, void *v)
5277
{
5278
	struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(m));
5279 5280
	struct mem_cgroup *mi;
	unsigned int i;
5281

5282
	for (i = 0; i < MEM_CGROUP_STAT_NSTATS; i++) {
5283
		if (i == MEM_CGROUP_STAT_SWAP && !do_swap_account)
5284
			continue;
5285 5286
		seq_printf(m, "%s %ld\n", mem_cgroup_stat_names[i],
			   mem_cgroup_read_stat(memcg, i) * PAGE_SIZE);
5287
	}
L
Lee Schermerhorn 已提交
5288

5289 5290 5291 5292 5293 5294 5295 5296
	for (i = 0; i < MEM_CGROUP_EVENTS_NSTATS; i++)
		seq_printf(m, "%s %lu\n", mem_cgroup_events_names[i],
			   mem_cgroup_read_events(memcg, i));

	for (i = 0; i < NR_LRU_LISTS; i++)
		seq_printf(m, "%s %lu\n", mem_cgroup_lru_names[i],
			   mem_cgroup_nr_lru_pages(memcg, BIT(i)) * PAGE_SIZE);

K
KAMEZAWA Hiroyuki 已提交
5297
	/* Hierarchical information */
5298 5299
	{
		unsigned long long limit, memsw_limit;
5300
		memcg_get_hierarchical_limit(memcg, &limit, &memsw_limit);
5301
		seq_printf(m, "hierarchical_memory_limit %llu\n", limit);
5302
		if (do_swap_account)
5303 5304
			seq_printf(m, "hierarchical_memsw_limit %llu\n",
				   memsw_limit);
5305
	}
K
KOSAKI Motohiro 已提交
5306

5307 5308 5309
	for (i = 0; i < MEM_CGROUP_STAT_NSTATS; i++) {
		long long val = 0;

5310
		if (i == MEM_CGROUP_STAT_SWAP && !do_swap_account)
5311
			continue;
5312 5313 5314 5315 5316 5317 5318 5319 5320 5321 5322 5323 5324 5325 5326 5327 5328 5329 5330 5331
		for_each_mem_cgroup_tree(mi, memcg)
			val += mem_cgroup_read_stat(mi, i) * PAGE_SIZE;
		seq_printf(m, "total_%s %lld\n", mem_cgroup_stat_names[i], val);
	}

	for (i = 0; i < MEM_CGROUP_EVENTS_NSTATS; i++) {
		unsigned long long val = 0;

		for_each_mem_cgroup_tree(mi, memcg)
			val += mem_cgroup_read_events(mi, i);
		seq_printf(m, "total_%s %llu\n",
			   mem_cgroup_events_names[i], val);
	}

	for (i = 0; i < NR_LRU_LISTS; i++) {
		unsigned long long val = 0;

		for_each_mem_cgroup_tree(mi, memcg)
			val += mem_cgroup_nr_lru_pages(mi, BIT(i)) * PAGE_SIZE;
		seq_printf(m, "total_%s %llu\n", mem_cgroup_lru_names[i], val);
5332
	}
K
KAMEZAWA Hiroyuki 已提交
5333

K
KOSAKI Motohiro 已提交
5334 5335 5336 5337
#ifdef CONFIG_DEBUG_VM
	{
		int nid, zid;
		struct mem_cgroup_per_zone *mz;
5338
		struct zone_reclaim_stat *rstat;
K
KOSAKI Motohiro 已提交
5339 5340 5341 5342 5343
		unsigned long recent_rotated[2] = {0, 0};
		unsigned long recent_scanned[2] = {0, 0};

		for_each_online_node(nid)
			for (zid = 0; zid < MAX_NR_ZONES; zid++) {
5344
				mz = mem_cgroup_zoneinfo(memcg, nid, zid);
5345
				rstat = &mz->lruvec.reclaim_stat;
K
KOSAKI Motohiro 已提交
5346

5347 5348 5349 5350
				recent_rotated[0] += rstat->recent_rotated[0];
				recent_rotated[1] += rstat->recent_rotated[1];
				recent_scanned[0] += rstat->recent_scanned[0];
				recent_scanned[1] += rstat->recent_scanned[1];
K
KOSAKI Motohiro 已提交
5351
			}
5352 5353 5354 5355
		seq_printf(m, "recent_rotated_anon %lu\n", recent_rotated[0]);
		seq_printf(m, "recent_rotated_file %lu\n", recent_rotated[1]);
		seq_printf(m, "recent_scanned_anon %lu\n", recent_scanned[0]);
		seq_printf(m, "recent_scanned_file %lu\n", recent_scanned[1]);
K
KOSAKI Motohiro 已提交
5356 5357 5358
	}
#endif

5359 5360 5361
	return 0;
}

5362 5363
static u64 mem_cgroup_swappiness_read(struct cgroup_subsys_state *css,
				      struct cftype *cft)
K
KOSAKI Motohiro 已提交
5364
{
5365
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
K
KOSAKI Motohiro 已提交
5366

5367
	return mem_cgroup_swappiness(memcg);
K
KOSAKI Motohiro 已提交
5368 5369
}

5370 5371
static int mem_cgroup_swappiness_write(struct cgroup_subsys_state *css,
				       struct cftype *cft, u64 val)
K
KOSAKI Motohiro 已提交
5372
{
5373
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
K
KOSAKI Motohiro 已提交
5374

5375
	if (val > 100)
K
KOSAKI Motohiro 已提交
5376 5377
		return -EINVAL;

5378 5379 5380 5381
	if (css_parent(css))
		memcg->swappiness = val;
	else
		vm_swappiness = val;
5382

K
KOSAKI Motohiro 已提交
5383 5384 5385
	return 0;
}

5386 5387 5388 5389 5390 5391 5392 5393
static void __mem_cgroup_threshold(struct mem_cgroup *memcg, bool swap)
{
	struct mem_cgroup_threshold_ary *t;
	u64 usage;
	int i;

	rcu_read_lock();
	if (!swap)
5394
		t = rcu_dereference(memcg->thresholds.primary);
5395
	else
5396
		t = rcu_dereference(memcg->memsw_thresholds.primary);
5397 5398 5399 5400 5401 5402 5403

	if (!t)
		goto unlock;

	usage = mem_cgroup_usage(memcg, swap);

	/*
5404
	 * current_threshold points to threshold just below or equal to usage.
5405 5406 5407
	 * If it's not true, a threshold was crossed after last
	 * call of __mem_cgroup_threshold().
	 */
5408
	i = t->current_threshold;
5409 5410 5411 5412 5413 5414 5415 5416 5417 5418 5419 5420 5421 5422 5423 5424 5425 5426 5427 5428 5429 5430 5431

	/*
	 * Iterate backward over array of thresholds starting from
	 * current_threshold and check if a threshold is crossed.
	 * If none of thresholds below usage is crossed, we read
	 * only one element of the array here.
	 */
	for (; i >= 0 && unlikely(t->entries[i].threshold > usage); i--)
		eventfd_signal(t->entries[i].eventfd, 1);

	/* i = current_threshold + 1 */
	i++;

	/*
	 * Iterate forward over array of thresholds starting from
	 * current_threshold+1 and check if a threshold is crossed.
	 * If none of thresholds above usage is crossed, we read
	 * only one element of the array here.
	 */
	for (; i < t->size && unlikely(t->entries[i].threshold <= usage); i++)
		eventfd_signal(t->entries[i].eventfd, 1);

	/* Update current_threshold */
5432
	t->current_threshold = i - 1;
5433 5434 5435 5436 5437 5438
unlock:
	rcu_read_unlock();
}

static void mem_cgroup_threshold(struct mem_cgroup *memcg)
{
5439 5440 5441 5442 5443 5444 5445
	while (memcg) {
		__mem_cgroup_threshold(memcg, false);
		if (do_swap_account)
			__mem_cgroup_threshold(memcg, true);

		memcg = parent_mem_cgroup(memcg);
	}
5446 5447 5448 5449 5450 5451 5452
}

static int compare_thresholds(const void *a, const void *b)
{
	const struct mem_cgroup_threshold *_a = a;
	const struct mem_cgroup_threshold *_b = b;

5453 5454 5455 5456 5457 5458 5459
	if (_a->threshold > _b->threshold)
		return 1;

	if (_a->threshold < _b->threshold)
		return -1;

	return 0;
5460 5461
}

5462
static int mem_cgroup_oom_notify_cb(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
5463 5464 5465
{
	struct mem_cgroup_eventfd_list *ev;

5466
	list_for_each_entry(ev, &memcg->oom_notify, list)
K
KAMEZAWA Hiroyuki 已提交
5467 5468 5469 5470
		eventfd_signal(ev->eventfd, 1);
	return 0;
}

5471
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
5472
{
K
KAMEZAWA Hiroyuki 已提交
5473 5474
	struct mem_cgroup *iter;

5475
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
5476
		mem_cgroup_oom_notify_cb(iter);
K
KAMEZAWA Hiroyuki 已提交
5477 5478
}

5479
static int __mem_cgroup_usage_register_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
5480
	struct eventfd_ctx *eventfd, const char *args, enum res_type type)
5481
{
5482 5483
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
5484
	u64 threshold, usage;
5485
	int i, size, ret;
5486 5487 5488 5489 5490 5491

	ret = res_counter_memparse_write_strategy(args, &threshold);
	if (ret)
		return ret;

	mutex_lock(&memcg->thresholds_lock);
5492

5493
	if (type == _MEM)
5494
		thresholds = &memcg->thresholds;
5495
	else if (type == _MEMSWAP)
5496
		thresholds = &memcg->memsw_thresholds;
5497 5498 5499 5500 5501 5502
	else
		BUG();

	usage = mem_cgroup_usage(memcg, type == _MEMSWAP);

	/* Check if a threshold crossed before adding a new one */
5503
	if (thresholds->primary)
5504 5505
		__mem_cgroup_threshold(memcg, type == _MEMSWAP);

5506
	size = thresholds->primary ? thresholds->primary->size + 1 : 1;
5507 5508

	/* Allocate memory for new array of thresholds */
5509
	new = kmalloc(sizeof(*new) + size * sizeof(struct mem_cgroup_threshold),
5510
			GFP_KERNEL);
5511
	if (!new) {
5512 5513 5514
		ret = -ENOMEM;
		goto unlock;
	}
5515
	new->size = size;
5516 5517

	/* Copy thresholds (if any) to new array */
5518 5519
	if (thresholds->primary) {
		memcpy(new->entries, thresholds->primary->entries, (size - 1) *
5520
				sizeof(struct mem_cgroup_threshold));
5521 5522
	}

5523
	/* Add new threshold */
5524 5525
	new->entries[size - 1].eventfd = eventfd;
	new->entries[size - 1].threshold = threshold;
5526 5527

	/* Sort thresholds. Registering of new threshold isn't time-critical */
5528
	sort(new->entries, size, sizeof(struct mem_cgroup_threshold),
5529 5530 5531
			compare_thresholds, NULL);

	/* Find current threshold */
5532
	new->current_threshold = -1;
5533
	for (i = 0; i < size; i++) {
5534
		if (new->entries[i].threshold <= usage) {
5535
			/*
5536 5537
			 * new->current_threshold will not be used until
			 * rcu_assign_pointer(), so it's safe to increment
5538 5539
			 * it here.
			 */
5540
			++new->current_threshold;
5541 5542
		} else
			break;
5543 5544
	}

5545 5546 5547 5548 5549
	/* Free old spare buffer and save old primary buffer as spare */
	kfree(thresholds->spare);
	thresholds->spare = thresholds->primary;

	rcu_assign_pointer(thresholds->primary, new);
5550

5551
	/* To be sure that nobody uses thresholds */
5552 5553 5554 5555 5556 5557 5558 5559
	synchronize_rcu();

unlock:
	mutex_unlock(&memcg->thresholds_lock);

	return ret;
}

5560
static int mem_cgroup_usage_register_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
5561 5562
	struct eventfd_ctx *eventfd, const char *args)
{
5563
	return __mem_cgroup_usage_register_event(memcg, eventfd, args, _MEM);
T
Tejun Heo 已提交
5564 5565
}

5566
static int memsw_cgroup_usage_register_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
5567 5568
	struct eventfd_ctx *eventfd, const char *args)
{
5569
	return __mem_cgroup_usage_register_event(memcg, eventfd, args, _MEMSWAP);
T
Tejun Heo 已提交
5570 5571
}

5572
static void __mem_cgroup_usage_unregister_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
5573
	struct eventfd_ctx *eventfd, enum res_type type)
5574
{
5575 5576
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
5577
	u64 usage;
5578
	int i, j, size;
5579 5580 5581

	mutex_lock(&memcg->thresholds_lock);
	if (type == _MEM)
5582
		thresholds = &memcg->thresholds;
5583
	else if (type == _MEMSWAP)
5584
		thresholds = &memcg->memsw_thresholds;
5585 5586 5587
	else
		BUG();

5588 5589 5590
	if (!thresholds->primary)
		goto unlock;

5591 5592 5593 5594 5595 5596
	usage = mem_cgroup_usage(memcg, type == _MEMSWAP);

	/* Check if a threshold crossed before removing */
	__mem_cgroup_threshold(memcg, type == _MEMSWAP);

	/* Calculate new number of threshold */
5597 5598 5599
	size = 0;
	for (i = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd != eventfd)
5600 5601 5602
			size++;
	}

5603
	new = thresholds->spare;
5604

5605 5606
	/* Set thresholds array to NULL if we don't have thresholds */
	if (!size) {
5607 5608
		kfree(new);
		new = NULL;
5609
		goto swap_buffers;
5610 5611
	}

5612
	new->size = size;
5613 5614

	/* Copy thresholds and find current threshold */
5615 5616 5617
	new->current_threshold = -1;
	for (i = 0, j = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd == eventfd)
5618 5619
			continue;

5620
		new->entries[j] = thresholds->primary->entries[i];
5621
		if (new->entries[j].threshold <= usage) {
5622
			/*
5623
			 * new->current_threshold will not be used
5624 5625 5626
			 * until rcu_assign_pointer(), so it's safe to increment
			 * it here.
			 */
5627
			++new->current_threshold;
5628 5629 5630 5631
		}
		j++;
	}

5632
swap_buffers:
5633 5634
	/* Swap primary and spare array */
	thresholds->spare = thresholds->primary;
5635 5636 5637 5638 5639 5640
	/* If all events are unregistered, free the spare array */
	if (!new) {
		kfree(thresholds->spare);
		thresholds->spare = NULL;
	}

5641
	rcu_assign_pointer(thresholds->primary, new);
5642

5643
	/* To be sure that nobody uses thresholds */
5644
	synchronize_rcu();
5645
unlock:
5646 5647
	mutex_unlock(&memcg->thresholds_lock);
}
5648

5649
static void mem_cgroup_usage_unregister_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
5650 5651
	struct eventfd_ctx *eventfd)
{
5652
	return __mem_cgroup_usage_unregister_event(memcg, eventfd, _MEM);
T
Tejun Heo 已提交
5653 5654
}

5655
static void memsw_cgroup_usage_unregister_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
5656 5657
	struct eventfd_ctx *eventfd)
{
5658
	return __mem_cgroup_usage_unregister_event(memcg, eventfd, _MEMSWAP);
T
Tejun Heo 已提交
5659 5660
}

5661
static int mem_cgroup_oom_register_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
5662
	struct eventfd_ctx *eventfd, const char *args)
K
KAMEZAWA Hiroyuki 已提交
5663 5664 5665 5666 5667 5668 5669
{
	struct mem_cgroup_eventfd_list *event;

	event = kmalloc(sizeof(*event),	GFP_KERNEL);
	if (!event)
		return -ENOMEM;

5670
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
5671 5672 5673 5674 5675

	event->eventfd = eventfd;
	list_add(&event->list, &memcg->oom_notify);

	/* already in OOM ? */
5676
	if (atomic_read(&memcg->under_oom))
K
KAMEZAWA Hiroyuki 已提交
5677
		eventfd_signal(eventfd, 1);
5678
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
5679 5680 5681 5682

	return 0;
}

5683
static void mem_cgroup_oom_unregister_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
5684
	struct eventfd_ctx *eventfd)
K
KAMEZAWA Hiroyuki 已提交
5685 5686 5687
{
	struct mem_cgroup_eventfd_list *ev, *tmp;

5688
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
5689

5690
	list_for_each_entry_safe(ev, tmp, &memcg->oom_notify, list) {
K
KAMEZAWA Hiroyuki 已提交
5691 5692 5693 5694 5695 5696
		if (ev->eventfd == eventfd) {
			list_del(&ev->list);
			kfree(ev);
		}
	}

5697
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
5698 5699
}

5700
static int mem_cgroup_oom_control_read(struct seq_file *sf, void *v)
5701
{
5702
	struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(sf));
5703

5704 5705
	seq_printf(sf, "oom_kill_disable %d\n", memcg->oom_kill_disable);
	seq_printf(sf, "under_oom %d\n", (bool)atomic_read(&memcg->under_oom));
5706 5707 5708
	return 0;
}

5709
static int mem_cgroup_oom_control_write(struct cgroup_subsys_state *css,
5710 5711
	struct cftype *cft, u64 val)
{
5712
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
5713 5714

	/* cannot set to root cgroup and only 0 and 1 are allowed */
5715
	if (!css_parent(css) || !((val == 0) || (val == 1)))
5716 5717
		return -EINVAL;

5718
	memcg->oom_kill_disable = val;
5719
	if (!val)
5720
		memcg_oom_recover(memcg);
5721

5722 5723 5724
	return 0;
}

A
Andrew Morton 已提交
5725
#ifdef CONFIG_MEMCG_KMEM
5726
static int memcg_init_kmem(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
5727
{
5728 5729
	int ret;

5730
	memcg->kmemcg_id = -1;
5731 5732 5733
	ret = memcg_propagate_kmem(memcg);
	if (ret)
		return ret;
5734

5735
	return mem_cgroup_sockets_init(memcg, ss);
5736
}
5737

5738
static void memcg_destroy_kmem(struct mem_cgroup *memcg)
G
Glauber Costa 已提交
5739
{
5740
	mem_cgroup_sockets_destroy(memcg);
5741 5742 5743 5744 5745 5746 5747 5748 5749 5750 5751 5752 5753 5754 5755 5756 5757 5758 5759 5760 5761 5762 5763 5764 5765 5766
}

static void kmem_cgroup_css_offline(struct mem_cgroup *memcg)
{
	if (!memcg_kmem_is_active(memcg))
		return;

	/*
	 * kmem charges can outlive the cgroup. In the case of slab
	 * pages, for instance, a page contain objects from various
	 * processes. As we prevent from taking a reference for every
	 * such allocation we have to be careful when doing uncharge
	 * (see memcg_uncharge_kmem) and here during offlining.
	 *
	 * The idea is that that only the _last_ uncharge which sees
	 * the dead memcg will drop the last reference. An additional
	 * reference is taken here before the group is marked dead
	 * which is then paired with css_put during uncharge resp. here.
	 *
	 * Although this might sound strange as this path is called from
	 * css_offline() when the referencemight have dropped down to 0
	 * and shouldn't be incremented anymore (css_tryget would fail)
	 * we do not have other options because of the kmem allocations
	 * lifetime.
	 */
	css_get(&memcg->css);
5767 5768 5769 5770 5771 5772 5773

	memcg_kmem_mark_dead(memcg);

	if (res_counter_read_u64(&memcg->kmem, RES_USAGE) != 0)
		return;

	if (memcg_kmem_test_and_clear_dead(memcg))
5774
		css_put(&memcg->css);
G
Glauber Costa 已提交
5775
}
5776
#else
5777
static int memcg_init_kmem(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
5778 5779 5780
{
	return 0;
}
G
Glauber Costa 已提交
5781

5782 5783 5784 5785 5786
static void memcg_destroy_kmem(struct mem_cgroup *memcg)
{
}

static void kmem_cgroup_css_offline(struct mem_cgroup *memcg)
G
Glauber Costa 已提交
5787 5788
{
}
5789 5790
#endif

5791 5792 5793 5794 5795 5796 5797 5798 5799 5800 5801 5802 5803
/*
 * DO NOT USE IN NEW FILES.
 *
 * "cgroup.event_control" implementation.
 *
 * This is way over-engineered.  It tries to support fully configurable
 * events for each user.  Such level of flexibility is completely
 * unnecessary especially in the light of the planned unified hierarchy.
 *
 * Please deprecate this and replace with something simpler if at all
 * possible.
 */

5804 5805 5806 5807 5808
/*
 * Unregister event and free resources.
 *
 * Gets called from workqueue.
 */
5809
static void memcg_event_remove(struct work_struct *work)
5810
{
5811 5812
	struct mem_cgroup_event *event =
		container_of(work, struct mem_cgroup_event, remove);
5813
	struct mem_cgroup *memcg = event->memcg;
5814 5815 5816

	remove_wait_queue(event->wqh, &event->wait);

5817
	event->unregister_event(memcg, event->eventfd);
5818 5819 5820 5821 5822 5823

	/* Notify userspace the event is going away. */
	eventfd_signal(event->eventfd, 1);

	eventfd_ctx_put(event->eventfd);
	kfree(event);
5824
	css_put(&memcg->css);
5825 5826 5827 5828 5829 5830 5831
}

/*
 * Gets called on POLLHUP on eventfd when user closes it.
 *
 * Called with wqh->lock held and interrupts disabled.
 */
5832 5833
static int memcg_event_wake(wait_queue_t *wait, unsigned mode,
			    int sync, void *key)
5834
{
5835 5836
	struct mem_cgroup_event *event =
		container_of(wait, struct mem_cgroup_event, wait);
5837
	struct mem_cgroup *memcg = event->memcg;
5838 5839 5840 5841 5842 5843 5844 5845 5846 5847 5848 5849
	unsigned long flags = (unsigned long)key;

	if (flags & POLLHUP) {
		/*
		 * If the event has been detached at cgroup removal, we
		 * can simply return knowing the other side will cleanup
		 * for us.
		 *
		 * We can't race against event freeing since the other
		 * side will require wqh->lock via remove_wait_queue(),
		 * which we hold.
		 */
5850
		spin_lock(&memcg->event_list_lock);
5851 5852 5853 5854 5855 5856 5857 5858
		if (!list_empty(&event->list)) {
			list_del_init(&event->list);
			/*
			 * We are in atomic context, but cgroup_event_remove()
			 * may sleep, so we have to call it in workqueue.
			 */
			schedule_work(&event->remove);
		}
5859
		spin_unlock(&memcg->event_list_lock);
5860 5861 5862 5863 5864
	}

	return 0;
}

5865
static void memcg_event_ptable_queue_proc(struct file *file,
5866 5867
		wait_queue_head_t *wqh, poll_table *pt)
{
5868 5869
	struct mem_cgroup_event *event =
		container_of(pt, struct mem_cgroup_event, pt);
5870 5871 5872 5873 5874 5875

	event->wqh = wqh;
	add_wait_queue(wqh, &event->wait);
}

/*
5876 5877
 * DO NOT USE IN NEW FILES.
 *
5878 5879 5880 5881 5882
 * Parse input and register new cgroup event handler.
 *
 * Input must be in format '<event_fd> <control_fd> <args>'.
 * Interpretation of args is defined by control file implementation.
 */
5883
static int memcg_write_event_control(struct cgroup_subsys_state *css,
5884
				     struct cftype *cft, char *buffer)
5885
{
5886
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
5887
	struct mem_cgroup_event *event;
5888 5889 5890 5891
	struct cgroup_subsys_state *cfile_css;
	unsigned int efd, cfd;
	struct fd efile;
	struct fd cfile;
5892
	const char *name;
5893 5894 5895 5896 5897 5898 5899 5900 5901 5902 5903 5904 5905 5906 5907 5908 5909
	char *endp;
	int ret;

	efd = simple_strtoul(buffer, &endp, 10);
	if (*endp != ' ')
		return -EINVAL;
	buffer = endp + 1;

	cfd = simple_strtoul(buffer, &endp, 10);
	if ((*endp != ' ') && (*endp != '\0'))
		return -EINVAL;
	buffer = endp + 1;

	event = kzalloc(sizeof(*event), GFP_KERNEL);
	if (!event)
		return -ENOMEM;

5910
	event->memcg = memcg;
5911
	INIT_LIST_HEAD(&event->list);
5912 5913 5914
	init_poll_funcptr(&event->pt, memcg_event_ptable_queue_proc);
	init_waitqueue_func_entry(&event->wait, memcg_event_wake);
	INIT_WORK(&event->remove, memcg_event_remove);
5915 5916 5917 5918 5919 5920 5921 5922 5923 5924 5925 5926 5927 5928 5929 5930 5931 5932 5933 5934 5935 5936 5937 5938 5939

	efile = fdget(efd);
	if (!efile.file) {
		ret = -EBADF;
		goto out_kfree;
	}

	event->eventfd = eventfd_ctx_fileget(efile.file);
	if (IS_ERR(event->eventfd)) {
		ret = PTR_ERR(event->eventfd);
		goto out_put_efile;
	}

	cfile = fdget(cfd);
	if (!cfile.file) {
		ret = -EBADF;
		goto out_put_eventfd;
	}

	/* the process need read permission on control file */
	/* AV: shouldn't we check that it's been opened for read instead? */
	ret = inode_permission(file_inode(cfile.file), MAY_READ);
	if (ret < 0)
		goto out_put_cfile;

5940 5941 5942 5943 5944
	/*
	 * Determine the event callbacks and set them in @event.  This used
	 * to be done via struct cftype but cgroup core no longer knows
	 * about these events.  The following is crude but the whole thing
	 * is for compatibility anyway.
5945 5946
	 *
	 * DO NOT ADD NEW FILES.
5947 5948 5949 5950 5951 5952 5953 5954 5955 5956 5957 5958 5959
	 */
	name = cfile.file->f_dentry->d_name.name;

	if (!strcmp(name, "memory.usage_in_bytes")) {
		event->register_event = mem_cgroup_usage_register_event;
		event->unregister_event = mem_cgroup_usage_unregister_event;
	} else if (!strcmp(name, "memory.oom_control")) {
		event->register_event = mem_cgroup_oom_register_event;
		event->unregister_event = mem_cgroup_oom_unregister_event;
	} else if (!strcmp(name, "memory.pressure_level")) {
		event->register_event = vmpressure_register_event;
		event->unregister_event = vmpressure_unregister_event;
	} else if (!strcmp(name, "memory.memsw.usage_in_bytes")) {
T
Tejun Heo 已提交
5960 5961
		event->register_event = memsw_cgroup_usage_register_event;
		event->unregister_event = memsw_cgroup_usage_unregister_event;
5962 5963 5964 5965 5966
	} else {
		ret = -EINVAL;
		goto out_put_cfile;
	}

5967
	/*
5968 5969 5970
	 * Verify @cfile should belong to @css.  Also, remaining events are
	 * automatically removed on cgroup destruction but the removal is
	 * asynchronous, so take an extra ref on @css.
5971
	 */
5972 5973
	cfile_css = css_tryget_from_dir(cfile.file->f_dentry->d_parent,
					&memory_cgrp_subsys);
5974
	ret = -EINVAL;
5975
	if (IS_ERR(cfile_css))
5976
		goto out_put_cfile;
5977 5978
	if (cfile_css != css) {
		css_put(cfile_css);
5979
		goto out_put_cfile;
5980
	}
5981

5982
	ret = event->register_event(memcg, event->eventfd, buffer);
5983 5984 5985 5986 5987
	if (ret)
		goto out_put_css;

	efile.file->f_op->poll(efile.file, &event->pt);

5988 5989 5990
	spin_lock(&memcg->event_list_lock);
	list_add(&event->list, &memcg->event_list);
	spin_unlock(&memcg->event_list_lock);
5991 5992 5993 5994 5995 5996 5997

	fdput(cfile);
	fdput(efile);

	return 0;

out_put_css:
5998
	css_put(css);
5999 6000 6001 6002 6003 6004 6005 6006 6007 6008 6009 6010
out_put_cfile:
	fdput(cfile);
out_put_eventfd:
	eventfd_ctx_put(event->eventfd);
out_put_efile:
	fdput(efile);
out_kfree:
	kfree(event);

	return ret;
}

B
Balbir Singh 已提交
6011 6012
static struct cftype mem_cgroup_files[] = {
	{
6013
		.name = "usage_in_bytes",
6014
		.private = MEMFILE_PRIVATE(_MEM, RES_USAGE),
6015
		.read_u64 = mem_cgroup_read_u64,
B
Balbir Singh 已提交
6016
	},
6017 6018
	{
		.name = "max_usage_in_bytes",
6019
		.private = MEMFILE_PRIVATE(_MEM, RES_MAX_USAGE),
6020
		.trigger = mem_cgroup_reset,
6021
		.read_u64 = mem_cgroup_read_u64,
6022
	},
B
Balbir Singh 已提交
6023
	{
6024
		.name = "limit_in_bytes",
6025
		.private = MEMFILE_PRIVATE(_MEM, RES_LIMIT),
6026
		.write_string = mem_cgroup_write,
6027
		.read_u64 = mem_cgroup_read_u64,
B
Balbir Singh 已提交
6028
	},
6029 6030 6031 6032
	{
		.name = "soft_limit_in_bytes",
		.private = MEMFILE_PRIVATE(_MEM, RES_SOFT_LIMIT),
		.write_string = mem_cgroup_write,
6033
		.read_u64 = mem_cgroup_read_u64,
6034
	},
B
Balbir Singh 已提交
6035 6036
	{
		.name = "failcnt",
6037
		.private = MEMFILE_PRIVATE(_MEM, RES_FAILCNT),
6038
		.trigger = mem_cgroup_reset,
6039
		.read_u64 = mem_cgroup_read_u64,
B
Balbir Singh 已提交
6040
	},
6041 6042
	{
		.name = "stat",
6043
		.seq_show = memcg_stat_show,
6044
	},
6045 6046 6047 6048
	{
		.name = "force_empty",
		.trigger = mem_cgroup_force_empty_write,
	},
6049 6050
	{
		.name = "use_hierarchy",
6051
		.flags = CFTYPE_INSANE,
6052 6053 6054
		.write_u64 = mem_cgroup_hierarchy_write,
		.read_u64 = mem_cgroup_hierarchy_read,
	},
6055
	{
6056 6057
		.name = "cgroup.event_control",		/* XXX: for compat */
		.write_string = memcg_write_event_control,
6058 6059 6060
		.flags = CFTYPE_NO_PREFIX,
		.mode = S_IWUGO,
	},
K
KOSAKI Motohiro 已提交
6061 6062 6063 6064 6065
	{
		.name = "swappiness",
		.read_u64 = mem_cgroup_swappiness_read,
		.write_u64 = mem_cgroup_swappiness_write,
	},
6066 6067 6068 6069 6070
	{
		.name = "move_charge_at_immigrate",
		.read_u64 = mem_cgroup_move_charge_read,
		.write_u64 = mem_cgroup_move_charge_write,
	},
K
KAMEZAWA Hiroyuki 已提交
6071 6072
	{
		.name = "oom_control",
6073
		.seq_show = mem_cgroup_oom_control_read,
6074
		.write_u64 = mem_cgroup_oom_control_write,
K
KAMEZAWA Hiroyuki 已提交
6075 6076
		.private = MEMFILE_PRIVATE(_OOM_TYPE, OOM_CONTROL),
	},
6077 6078 6079
	{
		.name = "pressure_level",
	},
6080 6081 6082
#ifdef CONFIG_NUMA
	{
		.name = "numa_stat",
6083
		.seq_show = memcg_numa_stat_show,
6084 6085
	},
#endif
6086 6087 6088 6089 6090
#ifdef CONFIG_MEMCG_KMEM
	{
		.name = "kmem.limit_in_bytes",
		.private = MEMFILE_PRIVATE(_KMEM, RES_LIMIT),
		.write_string = mem_cgroup_write,
6091
		.read_u64 = mem_cgroup_read_u64,
6092 6093 6094 6095
	},
	{
		.name = "kmem.usage_in_bytes",
		.private = MEMFILE_PRIVATE(_KMEM, RES_USAGE),
6096
		.read_u64 = mem_cgroup_read_u64,
6097 6098 6099 6100 6101
	},
	{
		.name = "kmem.failcnt",
		.private = MEMFILE_PRIVATE(_KMEM, RES_FAILCNT),
		.trigger = mem_cgroup_reset,
6102
		.read_u64 = mem_cgroup_read_u64,
6103 6104 6105 6106 6107
	},
	{
		.name = "kmem.max_usage_in_bytes",
		.private = MEMFILE_PRIVATE(_KMEM, RES_MAX_USAGE),
		.trigger = mem_cgroup_reset,
6108
		.read_u64 = mem_cgroup_read_u64,
6109
	},
6110 6111 6112
#ifdef CONFIG_SLABINFO
	{
		.name = "kmem.slabinfo",
6113
		.seq_show = mem_cgroup_slabinfo_read,
6114 6115
	},
#endif
6116
#endif
6117
	{ },	/* terminate */
6118
};
6119

6120 6121 6122 6123 6124
#ifdef CONFIG_MEMCG_SWAP
static struct cftype memsw_cgroup_files[] = {
	{
		.name = "memsw.usage_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_USAGE),
6125
		.read_u64 = mem_cgroup_read_u64,
6126 6127 6128 6129 6130
	},
	{
		.name = "memsw.max_usage_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_MAX_USAGE),
		.trigger = mem_cgroup_reset,
6131
		.read_u64 = mem_cgroup_read_u64,
6132 6133 6134 6135 6136
	},
	{
		.name = "memsw.limit_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_LIMIT),
		.write_string = mem_cgroup_write,
6137
		.read_u64 = mem_cgroup_read_u64,
6138 6139 6140 6141 6142
	},
	{
		.name = "memsw.failcnt",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_FAILCNT),
		.trigger = mem_cgroup_reset,
6143
		.read_u64 = mem_cgroup_read_u64,
6144 6145 6146 6147
	},
	{ },	/* terminate */
};
#endif
6148
static int alloc_mem_cgroup_per_zone_info(struct mem_cgroup *memcg, int node)
6149 6150
{
	struct mem_cgroup_per_node *pn;
6151
	struct mem_cgroup_per_zone *mz;
6152
	int zone, tmp = node;
6153 6154 6155 6156 6157 6158 6159 6160
	/*
	 * 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.
	 */
6161 6162
	if (!node_state(node, N_NORMAL_MEMORY))
		tmp = -1;
6163
	pn = kzalloc_node(sizeof(*pn), GFP_KERNEL, tmp);
6164 6165
	if (!pn)
		return 1;
6166 6167 6168

	for (zone = 0; zone < MAX_NR_ZONES; zone++) {
		mz = &pn->zoneinfo[zone];
6169
		lruvec_init(&mz->lruvec);
6170 6171
		mz->usage_in_excess = 0;
		mz->on_tree = false;
6172
		mz->memcg = memcg;
6173
	}
6174
	memcg->nodeinfo[node] = pn;
6175 6176 6177
	return 0;
}

6178
static void free_mem_cgroup_per_zone_info(struct mem_cgroup *memcg, int node)
6179
{
6180
	kfree(memcg->nodeinfo[node]);
6181 6182
}

6183 6184
static struct mem_cgroup *mem_cgroup_alloc(void)
{
6185
	struct mem_cgroup *memcg;
6186
	size_t size;
6187

6188 6189
	size = sizeof(struct mem_cgroup);
	size += nr_node_ids * sizeof(struct mem_cgroup_per_node *);
6190

6191
	memcg = kzalloc(size, GFP_KERNEL);
6192
	if (!memcg)
6193 6194
		return NULL;

6195 6196
	memcg->stat = alloc_percpu(struct mem_cgroup_stat_cpu);
	if (!memcg->stat)
6197
		goto out_free;
6198 6199
	spin_lock_init(&memcg->pcp_counter_lock);
	return memcg;
6200 6201

out_free:
6202
	kfree(memcg);
6203
	return NULL;
6204 6205
}

6206
/*
6207 6208 6209 6210 6211 6212 6213 6214
 * 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.
 *
 * Removal of cgroup itself succeeds regardless of refs from swap.
6215
 */
6216 6217

static void __mem_cgroup_free(struct mem_cgroup *memcg)
6218
{
6219
	int node;
6220

6221
	mem_cgroup_remove_from_trees(memcg);
6222 6223 6224 6225 6226 6227

	for_each_node(node)
		free_mem_cgroup_per_zone_info(memcg, node);

	free_percpu(memcg->stat);

6228 6229 6230 6231 6232 6233 6234 6235 6236 6237 6238
	/*
	 * We need to make sure that (at least for now), the jump label
	 * destruction code runs outside of the cgroup lock. This is because
	 * get_online_cpus(), which is called from the static_branch update,
	 * can't be called inside the cgroup_lock. cpusets are the ones
	 * enforcing this dependency, so if they ever change, we might as well.
	 *
	 * schedule_work() will guarantee this happens. Be careful if you need
	 * to move this code around, and make sure it is outside
	 * the cgroup_lock.
	 */
6239
	disarm_static_keys(memcg);
6240
	kfree(memcg);
6241
}
6242

6243 6244 6245
/*
 * Returns the parent mem_cgroup in memcgroup hierarchy with hierarchy enabled.
 */
G
Glauber Costa 已提交
6246
struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
6247
{
6248
	if (!memcg->res.parent)
6249
		return NULL;
6250
	return mem_cgroup_from_res_counter(memcg->res.parent, res);
6251
}
G
Glauber Costa 已提交
6252
EXPORT_SYMBOL(parent_mem_cgroup);
6253

6254 6255 6256 6257 6258 6259 6260 6261 6262 6263 6264 6265 6266 6267 6268 6269 6270 6271 6272 6273 6274 6275 6276
static void __init mem_cgroup_soft_limit_tree_init(void)
{
	struct mem_cgroup_tree_per_node *rtpn;
	struct mem_cgroup_tree_per_zone *rtpz;
	int tmp, node, zone;

	for_each_node(node) {
		tmp = node;
		if (!node_state(node, N_NORMAL_MEMORY))
			tmp = -1;
		rtpn = kzalloc_node(sizeof(*rtpn), GFP_KERNEL, tmp);
		BUG_ON(!rtpn);

		soft_limit_tree.rb_tree_per_node[node] = rtpn;

		for (zone = 0; zone < MAX_NR_ZONES; zone++) {
			rtpz = &rtpn->rb_tree_per_zone[zone];
			rtpz->rb_root = RB_ROOT;
			spin_lock_init(&rtpz->lock);
		}
	}
}

L
Li Zefan 已提交
6277
static struct cgroup_subsys_state * __ref
6278
mem_cgroup_css_alloc(struct cgroup_subsys_state *parent_css)
B
Balbir Singh 已提交
6279
{
6280
	struct mem_cgroup *memcg;
K
KAMEZAWA Hiroyuki 已提交
6281
	long error = -ENOMEM;
6282
	int node;
B
Balbir Singh 已提交
6283

6284 6285
	memcg = mem_cgroup_alloc();
	if (!memcg)
K
KAMEZAWA Hiroyuki 已提交
6286
		return ERR_PTR(error);
6287

B
Bob Liu 已提交
6288
	for_each_node(node)
6289
		if (alloc_mem_cgroup_per_zone_info(memcg, node))
6290
			goto free_out;
6291

6292
	/* root ? */
6293
	if (parent_css == NULL) {
6294
		root_mem_cgroup = memcg;
6295 6296 6297
		res_counter_init(&memcg->res, NULL);
		res_counter_init(&memcg->memsw, NULL);
		res_counter_init(&memcg->kmem, NULL);
6298
	}
6299

6300 6301 6302 6303 6304
	memcg->last_scanned_node = MAX_NUMNODES;
	INIT_LIST_HEAD(&memcg->oom_notify);
	memcg->move_charge_at_immigrate = 0;
	mutex_init(&memcg->thresholds_lock);
	spin_lock_init(&memcg->move_lock);
6305
	vmpressure_init(&memcg->vmpressure);
6306 6307
	INIT_LIST_HEAD(&memcg->event_list);
	spin_lock_init(&memcg->event_list_lock);
6308 6309 6310 6311 6312 6313 6314 6315 6316

	return &memcg->css;

free_out:
	__mem_cgroup_free(memcg);
	return ERR_PTR(error);
}

static int
6317
mem_cgroup_css_online(struct cgroup_subsys_state *css)
6318
{
6319 6320
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
	struct mem_cgroup *parent = mem_cgroup_from_css(css_parent(css));
6321

6322 6323 6324
	if (css->cgroup->id > MEM_CGROUP_ID_MAX)
		return -ENOSPC;

T
Tejun Heo 已提交
6325
	if (!parent)
6326 6327
		return 0;

6328
	mutex_lock(&memcg_create_mutex);
6329 6330 6331 6332 6333 6334

	memcg->use_hierarchy = parent->use_hierarchy;
	memcg->oom_kill_disable = parent->oom_kill_disable;
	memcg->swappiness = mem_cgroup_swappiness(parent);

	if (parent->use_hierarchy) {
6335 6336
		res_counter_init(&memcg->res, &parent->res);
		res_counter_init(&memcg->memsw, &parent->memsw);
6337
		res_counter_init(&memcg->kmem, &parent->kmem);
6338

6339
		/*
6340 6341
		 * No need to take a reference to the parent because cgroup
		 * core guarantees its existence.
6342
		 */
6343
	} else {
6344 6345
		res_counter_init(&memcg->res, NULL);
		res_counter_init(&memcg->memsw, NULL);
6346
		res_counter_init(&memcg->kmem, NULL);
6347 6348 6349 6350 6351
		/*
		 * Deeper hierachy with use_hierarchy == false doesn't make
		 * much sense so let cgroup subsystem know about this
		 * unfortunate state in our controller.
		 */
6352
		if (parent != root_mem_cgroup)
6353
			memory_cgrp_subsys.broken_hierarchy = true;
6354
	}
6355
	mutex_unlock(&memcg_create_mutex);
6356

6357
	return memcg_init_kmem(memcg, &memory_cgrp_subsys);
B
Balbir Singh 已提交
6358 6359
}

M
Michal Hocko 已提交
6360 6361 6362 6363 6364 6365 6366 6367
/*
 * Announce all parents that a group from their hierarchy is gone.
 */
static void mem_cgroup_invalidate_reclaim_iterators(struct mem_cgroup *memcg)
{
	struct mem_cgroup *parent = memcg;

	while ((parent = parent_mem_cgroup(parent)))
6368
		mem_cgroup_iter_invalidate(parent);
M
Michal Hocko 已提交
6369 6370 6371 6372 6373 6374

	/*
	 * if the root memcg is not hierarchical we have to check it
	 * explicitely.
	 */
	if (!root_mem_cgroup->use_hierarchy)
6375
		mem_cgroup_iter_invalidate(root_mem_cgroup);
M
Michal Hocko 已提交
6376 6377
}

6378
static void mem_cgroup_css_offline(struct cgroup_subsys_state *css)
6379
{
6380
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
6381
	struct mem_cgroup_event *event, *tmp;
6382
	struct cgroup_subsys_state *iter;
6383 6384 6385 6386 6387 6388

	/*
	 * Unregister events and notify userspace.
	 * Notify userspace about cgroup removing only after rmdir of cgroup
	 * directory to avoid race between userspace and kernelspace.
	 */
6389 6390
	spin_lock(&memcg->event_list_lock);
	list_for_each_entry_safe(event, tmp, &memcg->event_list, list) {
6391 6392 6393
		list_del_init(&event->list);
		schedule_work(&event->remove);
	}
6394
	spin_unlock(&memcg->event_list_lock);
6395

6396 6397
	kmem_cgroup_css_offline(memcg);

M
Michal Hocko 已提交
6398
	mem_cgroup_invalidate_reclaim_iterators(memcg);
6399 6400 6401 6402 6403 6404 6405 6406

	/*
	 * This requires that offlining is serialized.  Right now that is
	 * guaranteed because css_killed_work_fn() holds the cgroup_mutex.
	 */
	css_for_each_descendant_post(iter, css)
		mem_cgroup_reparent_charges(mem_cgroup_from_css(iter));

G
Glauber Costa 已提交
6407
	mem_cgroup_destroy_all_caches(memcg);
6408
	vmpressure_cleanup(&memcg->vmpressure);
6409 6410
}

6411
static void mem_cgroup_css_free(struct cgroup_subsys_state *css)
B
Balbir Singh 已提交
6412
{
6413
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
6414 6415 6416 6417 6418 6419 6420 6421 6422 6423 6424 6425 6426 6427 6428 6429 6430 6431 6432 6433 6434 6435 6436 6437 6438 6439 6440 6441 6442 6443 6444 6445 6446 6447 6448 6449
	/*
	 * XXX: css_offline() would be where we should reparent all
	 * memory to prepare the cgroup for destruction.  However,
	 * memcg does not do css_tryget() and res_counter charging
	 * under the same RCU lock region, which means that charging
	 * could race with offlining.  Offlining only happens to
	 * cgroups with no tasks in them but charges can show up
	 * without any tasks from the swapin path when the target
	 * memcg is looked up from the swapout record and not from the
	 * current task as it usually is.  A race like this can leak
	 * charges and put pages with stale cgroup pointers into
	 * circulation:
	 *
	 * #0                        #1
	 *                           lookup_swap_cgroup_id()
	 *                           rcu_read_lock()
	 *                           mem_cgroup_lookup()
	 *                           css_tryget()
	 *                           rcu_read_unlock()
	 * disable css_tryget()
	 * call_rcu()
	 *   offline_css()
	 *     reparent_charges()
	 *                           res_counter_charge()
	 *                           css_put()
	 *                             css_free()
	 *                           pc->mem_cgroup = dead memcg
	 *                           add page to lru
	 *
	 * The bulk of the charges are still moved in offline_css() to
	 * avoid pinning a lot of pages in case a long-term reference
	 * like a swapout record is deferring the css_free() to long
	 * after offlining.  But this makes sure we catch any charges
	 * made after offlining:
	 */
	mem_cgroup_reparent_charges(memcg);
6450

6451
	memcg_destroy_kmem(memcg);
6452
	__mem_cgroup_free(memcg);
B
Balbir Singh 已提交
6453 6454
}

6455
#ifdef CONFIG_MMU
6456
/* Handlers for move charge at task migration. */
6457 6458
#define PRECHARGE_COUNT_AT_ONCE	256
static int mem_cgroup_do_precharge(unsigned long count)
6459
{
6460 6461
	int ret = 0;
	int batch_count = PRECHARGE_COUNT_AT_ONCE;
6462
	struct mem_cgroup *memcg = mc.to;
6463

6464
	if (mem_cgroup_is_root(memcg)) {
6465 6466 6467 6468 6469 6470 6471 6472
		mc.precharge += count;
		/* we don't need css_get for root */
		return ret;
	}
	/* try to charge at once */
	if (count > 1) {
		struct res_counter *dummy;
		/*
6473
		 * "memcg" cannot be under rmdir() because we've already checked
6474 6475 6476 6477
		 * by cgroup_lock_live_cgroup() that it is not removed and we
		 * are still under the same cgroup_mutex. So we can postpone
		 * css_get().
		 */
6478
		if (res_counter_charge(&memcg->res, PAGE_SIZE * count, &dummy))
6479
			goto one_by_one;
6480
		if (do_swap_account && res_counter_charge(&memcg->memsw,
6481
						PAGE_SIZE * count, &dummy)) {
6482
			res_counter_uncharge(&memcg->res, PAGE_SIZE * count);
6483 6484 6485 6486 6487 6488 6489 6490 6491 6492 6493 6494 6495 6496 6497 6498
			goto one_by_one;
		}
		mc.precharge += count;
		return ret;
	}
one_by_one:
	/* fall back to one by one charge */
	while (count--) {
		if (signal_pending(current)) {
			ret = -EINTR;
			break;
		}
		if (!batch_count--) {
			batch_count = PRECHARGE_COUNT_AT_ONCE;
			cond_resched();
		}
6499
		ret = mem_cgroup_try_charge(memcg, GFP_KERNEL, 1, false);
6500
		if (ret)
6501
			/* mem_cgroup_clear_mc() will do uncharge later */
6502
			return ret;
6503 6504
		mc.precharge++;
	}
6505 6506 6507 6508
	return ret;
}

/**
6509
 * get_mctgt_type - get target type of moving charge
6510 6511 6512
 * @vma: the vma the pte to be checked belongs
 * @addr: the address corresponding to the pte to be checked
 * @ptent: the pte to be checked
6513
 * @target: the pointer the target page or swap ent will be stored(can be NULL)
6514 6515 6516 6517 6518 6519
 *
 * Returns
 *   0(MC_TARGET_NONE): if the pte is not a target for move charge.
 *   1(MC_TARGET_PAGE): if the page corresponding to this pte is a target for
 *     move charge. if @target is not NULL, the page is stored in target->page
 *     with extra refcnt got(Callers should handle it).
6520 6521 6522
 *   2(MC_TARGET_SWAP): if the swap entry corresponding to this pte is a
 *     target for charge migration. if @target is not NULL, the entry is stored
 *     in target->ent.
6523 6524 6525 6526 6527
 *
 * Called with pte lock held.
 */
union mc_target {
	struct page	*page;
6528
	swp_entry_t	ent;
6529 6530 6531
};

enum mc_target_type {
6532
	MC_TARGET_NONE = 0,
6533
	MC_TARGET_PAGE,
6534
	MC_TARGET_SWAP,
6535 6536
};

D
Daisuke Nishimura 已提交
6537 6538
static struct page *mc_handle_present_pte(struct vm_area_struct *vma,
						unsigned long addr, pte_t ptent)
6539
{
D
Daisuke Nishimura 已提交
6540
	struct page *page = vm_normal_page(vma, addr, ptent);
6541

D
Daisuke Nishimura 已提交
6542 6543 6544 6545
	if (!page || !page_mapped(page))
		return NULL;
	if (PageAnon(page)) {
		/* we don't move shared anon */
6546
		if (!move_anon())
D
Daisuke Nishimura 已提交
6547
			return NULL;
6548 6549
	} else if (!move_file())
		/* we ignore mapcount for file pages */
D
Daisuke Nishimura 已提交
6550 6551 6552 6553 6554 6555 6556
		return NULL;
	if (!get_page_unless_zero(page))
		return NULL;

	return page;
}

6557
#ifdef CONFIG_SWAP
D
Daisuke Nishimura 已提交
6558 6559 6560 6561 6562 6563 6564 6565
static struct page *mc_handle_swap_pte(struct vm_area_struct *vma,
			unsigned long addr, pte_t ptent, swp_entry_t *entry)
{
	struct page *page = NULL;
	swp_entry_t ent = pte_to_swp_entry(ptent);

	if (!move_anon() || non_swap_entry(ent))
		return NULL;
6566 6567 6568 6569
	/*
	 * Because lookup_swap_cache() updates some statistics counter,
	 * we call find_get_page() with swapper_space directly.
	 */
6570
	page = find_get_page(swap_address_space(ent), ent.val);
D
Daisuke Nishimura 已提交
6571 6572 6573 6574 6575
	if (do_swap_account)
		entry->val = ent.val;

	return page;
}
6576 6577 6578 6579 6580 6581 6582
#else
static struct page *mc_handle_swap_pte(struct vm_area_struct *vma,
			unsigned long addr, pte_t ptent, swp_entry_t *entry)
{
	return NULL;
}
#endif
D
Daisuke Nishimura 已提交
6583

6584 6585 6586 6587 6588 6589 6590 6591 6592 6593 6594 6595 6596 6597 6598 6599 6600 6601 6602
static struct page *mc_handle_file_pte(struct vm_area_struct *vma,
			unsigned long addr, pte_t ptent, swp_entry_t *entry)
{
	struct page *page = NULL;
	struct address_space *mapping;
	pgoff_t pgoff;

	if (!vma->vm_file) /* anonymous vma */
		return NULL;
	if (!move_file())
		return NULL;

	mapping = vma->vm_file->f_mapping;
	if (pte_none(ptent))
		pgoff = linear_page_index(vma, addr);
	else /* pte_file(ptent) is true */
		pgoff = pte_to_pgoff(ptent);

	/* page is moved even if it's not RSS of this task(page-faulted). */
6603 6604
#ifdef CONFIG_SWAP
	/* shmem/tmpfs may report page out on swap: account for that too. */
6605 6606 6607 6608 6609 6610 6611 6612 6613 6614 6615 6616
	if (shmem_mapping(mapping)) {
		page = find_get_entry(mapping, pgoff);
		if (radix_tree_exceptional_entry(page)) {
			swp_entry_t swp = radix_to_swp_entry(page);
			if (do_swap_account)
				*entry = swp;
			page = find_get_page(swap_address_space(swp), swp.val);
		}
	} else
		page = find_get_page(mapping, pgoff);
#else
	page = find_get_page(mapping, pgoff);
6617
#endif
6618 6619 6620
	return page;
}

6621
static enum mc_target_type get_mctgt_type(struct vm_area_struct *vma,
D
Daisuke Nishimura 已提交
6622 6623 6624 6625
		unsigned long addr, pte_t ptent, union mc_target *target)
{
	struct page *page = NULL;
	struct page_cgroup *pc;
6626
	enum mc_target_type ret = MC_TARGET_NONE;
D
Daisuke Nishimura 已提交
6627 6628 6629 6630 6631 6632
	swp_entry_t ent = { .val = 0 };

	if (pte_present(ptent))
		page = mc_handle_present_pte(vma, addr, ptent);
	else if (is_swap_pte(ptent))
		page = mc_handle_swap_pte(vma, addr, ptent, &ent);
6633 6634
	else if (pte_none(ptent) || pte_file(ptent))
		page = mc_handle_file_pte(vma, addr, ptent, &ent);
D
Daisuke Nishimura 已提交
6635 6636

	if (!page && !ent.val)
6637
		return ret;
6638 6639 6640 6641 6642 6643 6644 6645 6646 6647 6648 6649 6650 6651 6652
	if (page) {
		pc = lookup_page_cgroup(page);
		/*
		 * Do only loose check w/o page_cgroup lock.
		 * mem_cgroup_move_account() checks the pc is valid or not under
		 * the lock.
		 */
		if (PageCgroupUsed(pc) && pc->mem_cgroup == mc.from) {
			ret = MC_TARGET_PAGE;
			if (target)
				target->page = page;
		}
		if (!ret || !target)
			put_page(page);
	}
D
Daisuke Nishimura 已提交
6653 6654
	/* There is a swap entry and a page doesn't exist or isn't charged */
	if (ent.val && !ret &&
L
Li Zefan 已提交
6655
	    mem_cgroup_id(mc.from) == lookup_swap_cgroup_id(ent)) {
6656 6657 6658
		ret = MC_TARGET_SWAP;
		if (target)
			target->ent = ent;
6659 6660 6661 6662
	}
	return ret;
}

6663 6664 6665 6666 6667 6668 6669 6670 6671 6672 6673 6674 6675 6676
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
/*
 * We don't consider swapping or file mapped pages because THP does not
 * support them for now.
 * Caller should make sure that pmd_trans_huge(pmd) is true.
 */
static enum mc_target_type get_mctgt_type_thp(struct vm_area_struct *vma,
		unsigned long addr, pmd_t pmd, union mc_target *target)
{
	struct page *page = NULL;
	struct page_cgroup *pc;
	enum mc_target_type ret = MC_TARGET_NONE;

	page = pmd_page(pmd);
6677
	VM_BUG_ON_PAGE(!page || !PageHead(page), page);
6678 6679 6680 6681 6682 6683 6684 6685 6686 6687 6688 6689 6690 6691 6692 6693 6694 6695 6696 6697
	if (!move_anon())
		return ret;
	pc = lookup_page_cgroup(page);
	if (PageCgroupUsed(pc) && pc->mem_cgroup == mc.from) {
		ret = MC_TARGET_PAGE;
		if (target) {
			get_page(page);
			target->page = page;
		}
	}
	return ret;
}
#else
static inline enum mc_target_type get_mctgt_type_thp(struct vm_area_struct *vma,
		unsigned long addr, pmd_t pmd, union mc_target *target)
{
	return MC_TARGET_NONE;
}
#endif

6698 6699 6700 6701 6702 6703 6704 6705
static int mem_cgroup_count_precharge_pte_range(pmd_t *pmd,
					unsigned long addr, unsigned long end,
					struct mm_walk *walk)
{
	struct vm_area_struct *vma = walk->private;
	pte_t *pte;
	spinlock_t *ptl;

6706
	if (pmd_trans_huge_lock(pmd, vma, &ptl) == 1) {
6707 6708
		if (get_mctgt_type_thp(vma, addr, *pmd, NULL) == MC_TARGET_PAGE)
			mc.precharge += HPAGE_PMD_NR;
6709
		spin_unlock(ptl);
6710
		return 0;
6711
	}
6712

6713 6714
	if (pmd_trans_unstable(pmd))
		return 0;
6715 6716
	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
	for (; addr != end; pte++, addr += PAGE_SIZE)
6717
		if (get_mctgt_type(vma, addr, *pte, NULL))
6718 6719 6720 6721
			mc.precharge++;	/* increment precharge temporarily */
	pte_unmap_unlock(pte - 1, ptl);
	cond_resched();

6722 6723 6724
	return 0;
}

6725 6726 6727 6728 6729
static unsigned long mem_cgroup_count_precharge(struct mm_struct *mm)
{
	unsigned long precharge;
	struct vm_area_struct *vma;

6730
	down_read(&mm->mmap_sem);
6731 6732 6733 6734 6735 6736 6737 6738 6739 6740 6741
	for (vma = mm->mmap; vma; vma = vma->vm_next) {
		struct mm_walk mem_cgroup_count_precharge_walk = {
			.pmd_entry = mem_cgroup_count_precharge_pte_range,
			.mm = mm,
			.private = vma,
		};
		if (is_vm_hugetlb_page(vma))
			continue;
		walk_page_range(vma->vm_start, vma->vm_end,
					&mem_cgroup_count_precharge_walk);
	}
6742
	up_read(&mm->mmap_sem);
6743 6744 6745 6746 6747 6748 6749 6750 6751

	precharge = mc.precharge;
	mc.precharge = 0;

	return precharge;
}

static int mem_cgroup_precharge_mc(struct mm_struct *mm)
{
6752 6753 6754 6755 6756
	unsigned long precharge = mem_cgroup_count_precharge(mm);

	VM_BUG_ON(mc.moving_task);
	mc.moving_task = current;
	return mem_cgroup_do_precharge(precharge);
6757 6758
}

6759 6760
/* cancels all extra charges on mc.from and mc.to, and wakes up all waiters. */
static void __mem_cgroup_clear_mc(void)
6761
{
6762 6763
	struct mem_cgroup *from = mc.from;
	struct mem_cgroup *to = mc.to;
L
Li Zefan 已提交
6764
	int i;
6765

6766
	/* we must uncharge all the leftover precharges from mc.to */
6767 6768 6769 6770 6771 6772 6773 6774 6775 6776 6777
	if (mc.precharge) {
		__mem_cgroup_cancel_charge(mc.to, mc.precharge);
		mc.precharge = 0;
	}
	/*
	 * we didn't uncharge from mc.from at mem_cgroup_move_account(), so
	 * we must uncharge here.
	 */
	if (mc.moved_charge) {
		__mem_cgroup_cancel_charge(mc.from, mc.moved_charge);
		mc.moved_charge = 0;
6778
	}
6779 6780 6781 6782 6783 6784
	/* we must fixup refcnts and charges */
	if (mc.moved_swap) {
		/* uncharge swap account from the old cgroup */
		if (!mem_cgroup_is_root(mc.from))
			res_counter_uncharge(&mc.from->memsw,
						PAGE_SIZE * mc.moved_swap);
L
Li Zefan 已提交
6785 6786 6787

		for (i = 0; i < mc.moved_swap; i++)
			css_put(&mc.from->css);
6788 6789 6790 6791 6792 6793 6794 6795 6796

		if (!mem_cgroup_is_root(mc.to)) {
			/*
			 * we charged both to->res and to->memsw, so we should
			 * uncharge to->res.
			 */
			res_counter_uncharge(&mc.to->res,
						PAGE_SIZE * mc.moved_swap);
		}
L
Li Zefan 已提交
6797
		/* we've already done css_get(mc.to) */
6798 6799
		mc.moved_swap = 0;
	}
6800 6801 6802 6803 6804 6805 6806 6807 6808 6809 6810 6811 6812 6813 6814
	memcg_oom_recover(from);
	memcg_oom_recover(to);
	wake_up_all(&mc.waitq);
}

static void mem_cgroup_clear_mc(void)
{
	struct mem_cgroup *from = mc.from;

	/*
	 * we must clear moving_task before waking up waiters at the end of
	 * task migration.
	 */
	mc.moving_task = NULL;
	__mem_cgroup_clear_mc();
6815
	spin_lock(&mc.lock);
6816 6817
	mc.from = NULL;
	mc.to = NULL;
6818
	spin_unlock(&mc.lock);
6819
	mem_cgroup_end_move(from);
6820 6821
}

6822
static int mem_cgroup_can_attach(struct cgroup_subsys_state *css,
6823
				 struct cgroup_taskset *tset)
6824
{
6825
	struct task_struct *p = cgroup_taskset_first(tset);
6826
	int ret = 0;
6827
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
6828
	unsigned long move_charge_at_immigrate;
6829

6830 6831 6832 6833 6834 6835 6836
	/*
	 * We are now commited to this value whatever it is. Changes in this
	 * tunable will only affect upcoming migrations, not the current one.
	 * So we need to save it, and keep it going.
	 */
	move_charge_at_immigrate  = memcg->move_charge_at_immigrate;
	if (move_charge_at_immigrate) {
6837 6838 6839
		struct mm_struct *mm;
		struct mem_cgroup *from = mem_cgroup_from_task(p);

6840
		VM_BUG_ON(from == memcg);
6841 6842 6843 6844 6845

		mm = get_task_mm(p);
		if (!mm)
			return 0;
		/* We move charges only when we move a owner of the mm */
6846 6847 6848 6849
		if (mm->owner == p) {
			VM_BUG_ON(mc.from);
			VM_BUG_ON(mc.to);
			VM_BUG_ON(mc.precharge);
6850
			VM_BUG_ON(mc.moved_charge);
6851
			VM_BUG_ON(mc.moved_swap);
6852
			mem_cgroup_start_move(from);
6853
			spin_lock(&mc.lock);
6854
			mc.from = from;
6855
			mc.to = memcg;
6856
			mc.immigrate_flags = move_charge_at_immigrate;
6857
			spin_unlock(&mc.lock);
6858
			/* We set mc.moving_task later */
6859 6860 6861 6862

			ret = mem_cgroup_precharge_mc(mm);
			if (ret)
				mem_cgroup_clear_mc();
6863 6864
		}
		mmput(mm);
6865 6866 6867 6868
	}
	return ret;
}

6869
static void mem_cgroup_cancel_attach(struct cgroup_subsys_state *css,
6870
				     struct cgroup_taskset *tset)
6871
{
6872
	mem_cgroup_clear_mc();
6873 6874
}

6875 6876 6877
static int mem_cgroup_move_charge_pte_range(pmd_t *pmd,
				unsigned long addr, unsigned long end,
				struct mm_walk *walk)
6878
{
6879 6880 6881 6882
	int ret = 0;
	struct vm_area_struct *vma = walk->private;
	pte_t *pte;
	spinlock_t *ptl;
6883 6884 6885 6886
	enum mc_target_type target_type;
	union mc_target target;
	struct page *page;
	struct page_cgroup *pc;
6887

6888 6889 6890 6891 6892 6893 6894 6895 6896 6897
	/*
	 * We don't take compound_lock() here but no race with splitting thp
	 * happens because:
	 *  - if pmd_trans_huge_lock() returns 1, the relevant thp is not
	 *    under splitting, which means there's no concurrent thp split,
	 *  - if another thread runs into split_huge_page() just after we
	 *    entered this if-block, the thread must wait for page table lock
	 *    to be unlocked in __split_huge_page_splitting(), where the main
	 *    part of thp split is not executed yet.
	 */
6898
	if (pmd_trans_huge_lock(pmd, vma, &ptl) == 1) {
6899
		if (mc.precharge < HPAGE_PMD_NR) {
6900
			spin_unlock(ptl);
6901 6902 6903 6904 6905 6906 6907 6908
			return 0;
		}
		target_type = get_mctgt_type_thp(vma, addr, *pmd, &target);
		if (target_type == MC_TARGET_PAGE) {
			page = target.page;
			if (!isolate_lru_page(page)) {
				pc = lookup_page_cgroup(page);
				if (!mem_cgroup_move_account(page, HPAGE_PMD_NR,
6909
							pc, mc.from, mc.to)) {
6910 6911 6912 6913 6914 6915 6916
					mc.precharge -= HPAGE_PMD_NR;
					mc.moved_charge += HPAGE_PMD_NR;
				}
				putback_lru_page(page);
			}
			put_page(page);
		}
6917
		spin_unlock(ptl);
6918
		return 0;
6919 6920
	}

6921 6922
	if (pmd_trans_unstable(pmd))
		return 0;
6923 6924 6925 6926
retry:
	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
	for (; addr != end; addr += PAGE_SIZE) {
		pte_t ptent = *(pte++);
6927
		swp_entry_t ent;
6928 6929 6930 6931

		if (!mc.precharge)
			break;

6932
		switch (get_mctgt_type(vma, addr, ptent, &target)) {
6933 6934 6935 6936 6937
		case MC_TARGET_PAGE:
			page = target.page;
			if (isolate_lru_page(page))
				goto put;
			pc = lookup_page_cgroup(page);
6938
			if (!mem_cgroup_move_account(page, 1, pc,
6939
						     mc.from, mc.to)) {
6940
				mc.precharge--;
6941 6942
				/* we uncharge from mc.from later. */
				mc.moved_charge++;
6943 6944
			}
			putback_lru_page(page);
6945
put:			/* get_mctgt_type() gets the page */
6946 6947
			put_page(page);
			break;
6948 6949
		case MC_TARGET_SWAP:
			ent = target.ent;
6950
			if (!mem_cgroup_move_swap_account(ent, mc.from, mc.to)) {
6951
				mc.precharge--;
6952 6953 6954
				/* we fixup refcnts and charges later. */
				mc.moved_swap++;
			}
6955
			break;
6956 6957 6958 6959 6960 6961 6962 6963 6964 6965 6966 6967 6968 6969
		default:
			break;
		}
	}
	pte_unmap_unlock(pte - 1, ptl);
	cond_resched();

	if (addr != end) {
		/*
		 * We have consumed all precharges we got in can_attach().
		 * We try charge one by one, but don't do any additional
		 * charges to mc.to if we have failed in charge once in attach()
		 * phase.
		 */
6970
		ret = mem_cgroup_do_precharge(1);
6971 6972 6973 6974 6975 6976 6977 6978 6979 6980 6981 6982
		if (!ret)
			goto retry;
	}

	return ret;
}

static void mem_cgroup_move_charge(struct mm_struct *mm)
{
	struct vm_area_struct *vma;

	lru_add_drain_all();
6983 6984 6985 6986 6987 6988 6989 6990 6991 6992 6993 6994 6995
retry:
	if (unlikely(!down_read_trylock(&mm->mmap_sem))) {
		/*
		 * Someone who are holding the mmap_sem might be waiting in
		 * waitq. So we cancel all extra charges, wake up all waiters,
		 * and retry. Because we cancel precharges, we might not be able
		 * to move enough charges, but moving charge is a best-effort
		 * feature anyway, so it wouldn't be a big problem.
		 */
		__mem_cgroup_clear_mc();
		cond_resched();
		goto retry;
	}
6996 6997 6998 6999 7000 7001 7002 7003 7004 7005 7006 7007 7008 7009 7010 7011 7012 7013
	for (vma = mm->mmap; vma; vma = vma->vm_next) {
		int ret;
		struct mm_walk mem_cgroup_move_charge_walk = {
			.pmd_entry = mem_cgroup_move_charge_pte_range,
			.mm = mm,
			.private = vma,
		};
		if (is_vm_hugetlb_page(vma))
			continue;
		ret = walk_page_range(vma->vm_start, vma->vm_end,
						&mem_cgroup_move_charge_walk);
		if (ret)
			/*
			 * means we have consumed all precharges and failed in
			 * doing additional charge. Just abandon here.
			 */
			break;
	}
7014
	up_read(&mm->mmap_sem);
7015 7016
}

7017
static void mem_cgroup_move_task(struct cgroup_subsys_state *css,
7018
				 struct cgroup_taskset *tset)
B
Balbir Singh 已提交
7019
{
7020
	struct task_struct *p = cgroup_taskset_first(tset);
7021
	struct mm_struct *mm = get_task_mm(p);
7022 7023

	if (mm) {
7024 7025
		if (mc.to)
			mem_cgroup_move_charge(mm);
7026 7027
		mmput(mm);
	}
7028 7029
	if (mc.to)
		mem_cgroup_clear_mc();
B
Balbir Singh 已提交
7030
}
7031
#else	/* !CONFIG_MMU */
7032
static int mem_cgroup_can_attach(struct cgroup_subsys_state *css,
7033
				 struct cgroup_taskset *tset)
7034 7035 7036
{
	return 0;
}
7037
static void mem_cgroup_cancel_attach(struct cgroup_subsys_state *css,
7038
				     struct cgroup_taskset *tset)
7039 7040
{
}
7041
static void mem_cgroup_move_task(struct cgroup_subsys_state *css,
7042
				 struct cgroup_taskset *tset)
7043 7044 7045
{
}
#endif
B
Balbir Singh 已提交
7046

7047 7048 7049 7050
/*
 * Cgroup retains root cgroups across [un]mount cycles making it necessary
 * to verify sane_behavior flag on each mount attempt.
 */
7051
static void mem_cgroup_bind(struct cgroup_subsys_state *root_css)
7052 7053 7054 7055 7056 7057
{
	/*
	 * use_hierarchy is forced with sane_behavior.  cgroup core
	 * guarantees that @root doesn't have any children, so turning it
	 * on for the root memcg is enough.
	 */
7058 7059
	if (cgroup_sane_behavior(root_css->cgroup))
		mem_cgroup_from_css(root_css)->use_hierarchy = true;
7060 7061
}

7062
struct cgroup_subsys memory_cgrp_subsys = {
7063
	.css_alloc = mem_cgroup_css_alloc,
7064
	.css_online = mem_cgroup_css_online,
7065 7066
	.css_offline = mem_cgroup_css_offline,
	.css_free = mem_cgroup_css_free,
7067 7068
	.can_attach = mem_cgroup_can_attach,
	.cancel_attach = mem_cgroup_cancel_attach,
B
Balbir Singh 已提交
7069
	.attach = mem_cgroup_move_task,
7070
	.bind = mem_cgroup_bind,
7071
	.base_cftypes = mem_cgroup_files,
7072
	.early_init = 0,
B
Balbir Singh 已提交
7073
};
7074

A
Andrew Morton 已提交
7075
#ifdef CONFIG_MEMCG_SWAP
7076 7077
static int __init enable_swap_account(char *s)
{
7078
	if (!strcmp(s, "1"))
7079
		really_do_swap_account = 1;
7080
	else if (!strcmp(s, "0"))
7081 7082 7083
		really_do_swap_account = 0;
	return 1;
}
7084
__setup("swapaccount=", enable_swap_account);
7085

7086 7087
static void __init memsw_file_init(void)
{
7088
	WARN_ON(cgroup_add_cftypes(&memory_cgrp_subsys, memsw_cgroup_files));
7089 7090 7091 7092 7093 7094 7095 7096
}

static void __init enable_swap_cgroup(void)
{
	if (!mem_cgroup_disabled() && really_do_swap_account) {
		do_swap_account = 1;
		memsw_file_init();
	}
7097
}
7098

7099
#else
7100
static void __init enable_swap_cgroup(void)
7101 7102
{
}
7103
#endif
7104 7105

/*
7106 7107 7108 7109 7110 7111
 * subsys_initcall() for memory controller.
 *
 * Some parts like hotcpu_notifier() have to be initialized from this context
 * because of lock dependencies (cgroup_lock -> cpu hotplug) but basically
 * everything that doesn't depend on a specific mem_cgroup structure should
 * be initialized from here.
7112 7113 7114 7115
 */
static int __init mem_cgroup_init(void)
{
	hotcpu_notifier(memcg_cpu_hotplug_callback, 0);
7116
	enable_swap_cgroup();
7117
	mem_cgroup_soft_limit_tree_init();
7118
	memcg_stock_init();
7119 7120 7121
	return 0;
}
subsys_initcall(mem_cgroup_init);