memcontrol.c 171.5 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
 * 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.
 */

28
#include <linux/page_counter.h>
B
Balbir Singh 已提交
29 30
#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
static int really_do_swap_account __initdata = 1;
#else
83
static int really_do_swap_account __initdata;
84 85
#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
	struct rb_node		tree_node;	/* RB tree node */
168
	unsigned long		usage_in_excess;/* Set to the value by which */
169 170
						/* 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
struct mem_cgroup_threshold {
	struct eventfd_ctx *eventfd;
201
	unsigned long threshold;
202 203
};

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
 */
struct mem_cgroup {
	struct cgroup_subsys_state css;
287 288 289 290 291 292 293

	/* Accounted resources */
	struct page_counter memory;
	struct page_counter memsw;
	struct page_counter kmem;

	unsigned long soft_limit;
294

295 296 297
	/* vmpressure notifications */
	struct vmpressure vmpressure;

298 299 300
	/* css_online() has been completed */
	int initialized;

301 302 303 304
	/*
	 * Should the accounting and control be hierarchical, per subtree?
	 */
	bool use_hierarchy;
305
	unsigned long kmem_account_flags; /* See KMEM_ACCOUNTED_*, below */
306 307 308

	bool		oom_lock;
	atomic_t	under_oom;
309
	atomic_t	oom_wakeups;
310

311
	int	swappiness;
312 313
	/* OOM-Killer disable */
	int		oom_kill_disable;
K
KOSAKI Motohiro 已提交
314

315 316 317 318
	/* protect arrays of thresholds */
	struct mutex thresholds_lock;

	/* thresholds for memory usage. RCU-protected */
319
	struct mem_cgroup_thresholds thresholds;
320

321
	/* thresholds for mem+swap usage. RCU-protected */
322
	struct mem_cgroup_thresholds memsw_thresholds;
323

K
KAMEZAWA Hiroyuki 已提交
324 325
	/* For oom notifier event fd */
	struct list_head oom_notify;
326

327 328 329 330
	/*
	 * 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
331
	unsigned long move_charge_at_immigrate;
332 333 334 335
	/*
	 * set > 0 if pages under this cgroup are moving to other cgroup.
	 */
	atomic_t	moving_account;
336 337
	/* taken only while moving_account > 0 */
	spinlock_t	move_lock;
338
	/*
339
	 * percpu counter.
340
	 */
341
	struct mem_cgroup_stat_cpu __percpu *stat;
342 343 344 345 346 347
	/*
	 * 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
348

M
Michal Hocko 已提交
349
	atomic_t	dead_count;
M
Michal Hocko 已提交
350
#if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_INET)
351
	struct cg_proto tcp_mem;
G
Glauber Costa 已提交
352
#endif
353
#if defined(CONFIG_MEMCG_KMEM)
354 355
	/* analogous to slab_common's slab_caches list, but per-memcg;
	 * protected by memcg_slab_mutex */
356 357 358 359
	struct list_head memcg_slab_caches;
        /* Index in the kmem_cache->memcg_params->memcg_caches array */
	int kmemcg_id;
#endif
360 361 362 363 364 365 366

	int last_scanned_node;
#if MAX_NUMNODES > 1
	nodemask_t	scan_nodes;
	atomic_t	numainfo_events;
	atomic_t	numainfo_updating;
#endif
367

368 369 370 371
	/* List of events which userspace want to receive */
	struct list_head event_list;
	spinlock_t event_list_lock;

372 373
	struct mem_cgroup_per_node *nodeinfo[0];
	/* WARNING: nodeinfo must be the last member here */
B
Balbir Singh 已提交
374 375
};

376 377
/* internal only representation about the status of kmem accounting. */
enum {
378
	KMEM_ACCOUNTED_ACTIVE, /* accounted by this cgroup itself */
379
	KMEM_ACCOUNTED_DEAD, /* dead memcg with pending kmem charges */
380 381 382 383 384 385 386
};

#ifdef CONFIG_MEMCG_KMEM
static inline void memcg_kmem_set_active(struct mem_cgroup *memcg)
{
	set_bit(KMEM_ACCOUNTED_ACTIVE, &memcg->kmem_account_flags);
}
387 388 389 390 391 392 393 394

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)
{
395 396 397 398 399
	/*
	 * Our caller must use css_get() first, because memcg_uncharge_kmem()
	 * will call css_put() if it sees the memcg is dead.
	 */
	smp_wmb();
400 401 402 403 404 405 406 407 408
	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);
}
409 410
#endif

411 412
/* Stuffs for move charges at task migration. */
/*
413 414
 * Types of charges to be moved. "move_charge_at_immitgrate" and
 * "immigrate_flags" are treated as a left-shifted bitmap of these types.
415 416
 */
enum move_type {
417
	MOVE_CHARGE_TYPE_ANON,	/* private anonymous page and swap of it */
418
	MOVE_CHARGE_TYPE_FILE,	/* file page(including tmpfs) and swap of it */
419 420 421
	NR_MOVE_TYPE,
};

422 423
/* "mc" and its members are protected by cgroup_mutex */
static struct move_charge_struct {
424
	spinlock_t	  lock; /* for from, to */
425 426
	struct mem_cgroup *from;
	struct mem_cgroup *to;
427
	unsigned long immigrate_flags;
428
	unsigned long precharge;
429
	unsigned long moved_charge;
430
	unsigned long moved_swap;
431 432 433
	struct task_struct *moving_task;	/* a task moving charges */
	wait_queue_head_t waitq;		/* a waitq for other context */
} mc = {
434
	.lock = __SPIN_LOCK_UNLOCKED(mc.lock),
435 436
	.waitq = __WAIT_QUEUE_HEAD_INITIALIZER(mc.waitq),
};
437

D
Daisuke Nishimura 已提交
438 439
static bool move_anon(void)
{
440
	return test_bit(MOVE_CHARGE_TYPE_ANON, &mc.immigrate_flags);
D
Daisuke Nishimura 已提交
441 442
}

443 444
static bool move_file(void)
{
445
	return test_bit(MOVE_CHARGE_TYPE_FILE, &mc.immigrate_flags);
446 447
}

448 449 450 451
/*
 * Maximum loops in mem_cgroup_hierarchical_reclaim(), used for soft
 * limit reclaim to prevent infinite loops, if they ever occur.
 */
452
#define	MEM_CGROUP_MAX_RECLAIM_LOOPS		100
453
#define	MEM_CGROUP_MAX_SOFT_LIMIT_RECLAIM_LOOPS	2
454

455 456
enum charge_type {
	MEM_CGROUP_CHARGE_TYPE_CACHE = 0,
457
	MEM_CGROUP_CHARGE_TYPE_ANON,
K
KAMEZAWA Hiroyuki 已提交
458
	MEM_CGROUP_CHARGE_TYPE_SWAPOUT,	/* for accounting swapcache */
K
KAMEZAWA Hiroyuki 已提交
459
	MEM_CGROUP_CHARGE_TYPE_DROP,	/* a page was unused swap cache */
460 461 462
	NR_CHARGE_TYPE,
};

463
/* for encoding cft->private value on file */
G
Glauber Costa 已提交
464 465 466 467
enum res_type {
	_MEM,
	_MEMSWAP,
	_OOM_TYPE,
468
	_KMEM,
G
Glauber Costa 已提交
469 470
};

471 472
#define MEMFILE_PRIVATE(x, val)	((x) << 16 | (val))
#define MEMFILE_TYPE(val)	((val) >> 16 & 0xffff)
473
#define MEMFILE_ATTR(val)	((val) & 0xffff)
K
KAMEZAWA Hiroyuki 已提交
474 475
/* Used for OOM nofiier */
#define OOM_CONTROL		(0)
476

477 478 479 480 481 482 483
/*
 * 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);

484 485
struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *s)
{
486
	return s ? container_of(s, struct mem_cgroup, css) : NULL;
487 488
}

489 490 491 492 493 494 495 496 497 498 499 500 501
/* 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;
}

502 503 504 505 506
static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
{
	return (memcg == root_mem_cgroup);
}

507 508 509 510 511 512
/*
 * 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
513 514
static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
{
515
	return memcg->css.id;
L
Li Zefan 已提交
516 517 518 519 520 521
}

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

522
	css = css_from_id(id, &memory_cgrp_subsys);
L
Li Zefan 已提交
523 524 525
	return mem_cgroup_from_css(css);
}

G
Glauber Costa 已提交
526
/* Writing them here to avoid exposing memcg's inner layout */
M
Michal Hocko 已提交
527
#if defined(CONFIG_INET) && defined(CONFIG_MEMCG_KMEM)
G
Glauber Costa 已提交
528 529 530

void sock_update_memcg(struct sock *sk)
{
531
	if (mem_cgroup_sockets_enabled) {
G
Glauber Costa 已提交
532
		struct mem_cgroup *memcg;
533
		struct cg_proto *cg_proto;
G
Glauber Costa 已提交
534 535 536

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

537 538 539 540 541 542 543 544 545 546
		/* 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));
547
			css_get(&sk->sk_cgrp->memcg->css);
548 549 550
			return;
		}

G
Glauber Costa 已提交
551 552
		rcu_read_lock();
		memcg = mem_cgroup_from_task(current);
553
		cg_proto = sk->sk_prot->proto_cgroup(memcg);
554
		if (!mem_cgroup_is_root(memcg) &&
555 556
		    memcg_proto_active(cg_proto) &&
		    css_tryget_online(&memcg->css)) {
557
			sk->sk_cgrp = cg_proto;
G
Glauber Costa 已提交
558 559 560 561 562 563 564 565
		}
		rcu_read_unlock();
	}
}
EXPORT_SYMBOL(sock_update_memcg);

void sock_release_memcg(struct sock *sk)
{
566
	if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
G
Glauber Costa 已提交
567 568 569
		struct mem_cgroup *memcg;
		WARN_ON(!sk->sk_cgrp->memcg);
		memcg = sk->sk_cgrp->memcg;
570
		css_put(&sk->sk_cgrp->memcg->css);
G
Glauber Costa 已提交
571 572
	}
}
G
Glauber Costa 已提交
573 574 575 576 577 578

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

579
	return &memcg->tcp_mem;
G
Glauber Costa 已提交
580 581
}
EXPORT_SYMBOL(tcp_proto_cgroup);
G
Glauber Costa 已提交
582

583 584
static void disarm_sock_keys(struct mem_cgroup *memcg)
{
585
	if (!memcg_proto_activated(&memcg->tcp_mem))
586 587 588 589 590 591 592 593 594
		return;
	static_key_slow_dec(&memcg_socket_limit_enabled);
}
#else
static void disarm_sock_keys(struct mem_cgroup *memcg)
{
}
#endif

595
#ifdef CONFIG_MEMCG_KMEM
596 597
/*
 * This will be the memcg's index in each cache's ->memcg_params->memcg_caches.
L
Li Zefan 已提交
598 599 600 601 602
 * 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.
603 604 605 606 607 608
 *
 * 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);
609 610
int memcg_limited_groups_array_size;

611 612 613 614 615 616
/*
 * 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
617
 * MAX_SIZE should be as large as the number of cgrp_ids. Ideally, we could get
618 619
 * 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
620
 * cgrp_id space is not getting any smaller, and we don't have to necessarily
621 622 623
 * increase ours as well if it increases.
 */
#define MEMCG_CACHES_MIN_SIZE 4
L
Li Zefan 已提交
624
#define MEMCG_CACHES_MAX_SIZE MEM_CGROUP_ID_MAX
625

626 627 628 629 630 631
/*
 * 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
 */
632
struct static_key memcg_kmem_enabled_key;
633
EXPORT_SYMBOL(memcg_kmem_enabled_key);
634

635 636
static void memcg_free_cache_id(int id);

637 638
static void disarm_kmem_keys(struct mem_cgroup *memcg)
{
639
	if (memcg_kmem_is_active(memcg)) {
640
		static_key_slow_dec(&memcg_kmem_enabled_key);
641
		memcg_free_cache_id(memcg->kmemcg_id);
642
	}
643 644 645 646
	/*
	 * This check can't live in kmem destruction function,
	 * since the charges will outlive the cgroup
	 */
647
	WARN_ON(page_counter_read(&memcg->kmem));
648 649 650 651 652 653 654 655 656 657 658 659 660
}
#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);
}

661
static void drain_all_stock_async(struct mem_cgroup *memcg);
662

663
static struct mem_cgroup_per_zone *
664
mem_cgroup_zone_zoneinfo(struct mem_cgroup *memcg, struct zone *zone)
665
{
666 667 668
	int nid = zone_to_nid(zone);
	int zid = zone_idx(zone);

669
	return &memcg->nodeinfo[nid]->zoneinfo[zid];
670 671
}

672
struct cgroup_subsys_state *mem_cgroup_css(struct mem_cgroup *memcg)
673
{
674
	return &memcg->css;
675 676
}

677
static struct mem_cgroup_per_zone *
678
mem_cgroup_page_zoneinfo(struct mem_cgroup *memcg, struct page *page)
679
{
680 681
	int nid = page_to_nid(page);
	int zid = page_zonenum(page);
682

683
	return &memcg->nodeinfo[nid]->zoneinfo[zid];
684 685
}

686 687 688 689 690 691 692 693 694 695 696 697 698 699 700
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];
}

701 702
static void __mem_cgroup_insert_exceeded(struct mem_cgroup_per_zone *mz,
					 struct mem_cgroup_tree_per_zone *mctz,
703
					 unsigned long new_usage_in_excess)
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
{
	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;
}

733 734
static void __mem_cgroup_remove_exceeded(struct mem_cgroup_per_zone *mz,
					 struct mem_cgroup_tree_per_zone *mctz)
735 736 737 738 739 740 741
{
	if (!mz->on_tree)
		return;
	rb_erase(&mz->tree_node, &mctz->rb_root);
	mz->on_tree = false;
}

742 743
static void mem_cgroup_remove_exceeded(struct mem_cgroup_per_zone *mz,
				       struct mem_cgroup_tree_per_zone *mctz)
744
{
745 746 747
	unsigned long flags;

	spin_lock_irqsave(&mctz->lock, flags);
748
	__mem_cgroup_remove_exceeded(mz, mctz);
749
	spin_unlock_irqrestore(&mctz->lock, flags);
750 751
}

752 753 754 755 756 757 758 759 760 761 762
static unsigned long soft_limit_excess(struct mem_cgroup *memcg)
{
	unsigned long nr_pages = page_counter_read(&memcg->memory);
	unsigned long soft_limit = ACCESS_ONCE(memcg->soft_limit);
	unsigned long excess = 0;

	if (nr_pages > soft_limit)
		excess = nr_pages - soft_limit;

	return excess;
}
763 764 765

static void mem_cgroup_update_tree(struct mem_cgroup *memcg, struct page *page)
{
766
	unsigned long excess;
767 768 769
	struct mem_cgroup_per_zone *mz;
	struct mem_cgroup_tree_per_zone *mctz;

770
	mctz = soft_limit_tree_from_page(page);
771 772 773 774 775
	/*
	 * Necessary to update all ancestors when hierarchy is used.
	 * because their event counter is not touched.
	 */
	for (; memcg; memcg = parent_mem_cgroup(memcg)) {
776
		mz = mem_cgroup_page_zoneinfo(memcg, page);
777
		excess = soft_limit_excess(memcg);
778 779 780 781 782
		/*
		 * We have to update the tree if mz is on RB-tree or
		 * mem is over its softlimit.
		 */
		if (excess || mz->on_tree) {
783 784 785
			unsigned long flags;

			spin_lock_irqsave(&mctz->lock, flags);
786 787
			/* if on-tree, remove it */
			if (mz->on_tree)
788
				__mem_cgroup_remove_exceeded(mz, mctz);
789 790 791 792
			/*
			 * Insert again. mz->usage_in_excess will be updated.
			 * If excess is 0, no tree ops.
			 */
793
			__mem_cgroup_insert_exceeded(mz, mctz, excess);
794
			spin_unlock_irqrestore(&mctz->lock, flags);
795 796 797 798 799 800 801
		}
	}
}

static void mem_cgroup_remove_from_trees(struct mem_cgroup *memcg)
{
	struct mem_cgroup_tree_per_zone *mctz;
802 803
	struct mem_cgroup_per_zone *mz;
	int nid, zid;
804

805 806 807 808
	for_each_node(nid) {
		for (zid = 0; zid < MAX_NR_ZONES; zid++) {
			mz = &memcg->nodeinfo[nid]->zoneinfo[zid];
			mctz = soft_limit_tree_node_zone(nid, zid);
809
			mem_cgroup_remove_exceeded(mz, mctz);
810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831
		}
	}
}

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.
	 */
832
	__mem_cgroup_remove_exceeded(mz, mctz);
833
	if (!soft_limit_excess(mz->memcg) ||
834
	    !css_tryget_online(&mz->memcg->css))
835 836 837 838 839 840 841 842 843 844
		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;

845
	spin_lock_irq(&mctz->lock);
846
	mz = __mem_cgroup_largest_soft_limit_node(mctz);
847
	spin_unlock_irq(&mctz->lock);
848 849 850
	return mz;
}

851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869
/*
 * 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.
 */
870
static long mem_cgroup_read_stat(struct mem_cgroup *memcg,
871
				 enum mem_cgroup_stat_index idx)
872
{
873
	long val = 0;
874 875
	int cpu;

876 877
	get_online_cpus();
	for_each_online_cpu(cpu)
878
		val += per_cpu(memcg->stat->count[idx], cpu);
879
#ifdef CONFIG_HOTPLUG_CPU
880 881 882
	spin_lock(&memcg->pcp_counter_lock);
	val += memcg->nocpu_base.count[idx];
	spin_unlock(&memcg->pcp_counter_lock);
883 884
#endif
	put_online_cpus();
885 886 887
	return val;
}

888
static unsigned long mem_cgroup_read_events(struct mem_cgroup *memcg,
889 890 891 892 893
					    enum mem_cgroup_events_index idx)
{
	unsigned long val = 0;
	int cpu;

894
	get_online_cpus();
895
	for_each_online_cpu(cpu)
896
		val += per_cpu(memcg->stat->events[idx], cpu);
897
#ifdef CONFIG_HOTPLUG_CPU
898 899 900
	spin_lock(&memcg->pcp_counter_lock);
	val += memcg->nocpu_base.events[idx];
	spin_unlock(&memcg->pcp_counter_lock);
901
#endif
902
	put_online_cpus();
903 904 905
	return val;
}

906
static void mem_cgroup_charge_statistics(struct mem_cgroup *memcg,
907
					 struct page *page,
908
					 int nr_pages)
909
{
910 911 912 913
	/*
	 * Here, RSS means 'mapped anon' and anon's SwapCache. Shmem/tmpfs is
	 * counted as CACHE even if it's on ANON LRU.
	 */
914
	if (PageAnon(page))
915
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_RSS],
916
				nr_pages);
917
	else
918
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_CACHE],
919
				nr_pages);
920

921 922 923 924
	if (PageTransHuge(page))
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_RSS_HUGE],
				nr_pages);

925 926
	/* pagein of a big page is an event. So, ignore page size */
	if (nr_pages > 0)
927
		__this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGPGIN]);
928
	else {
929
		__this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGPGOUT]);
930 931
		nr_pages = -nr_pages; /* for event */
	}
932

933
	__this_cpu_add(memcg->stat->nr_page_events, nr_pages);
934 935
}

936
unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
937 938 939 940 941 942 943
{
	struct mem_cgroup_per_zone *mz;

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

944 945 946
static unsigned long mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
						  int nid,
						  unsigned int lru_mask)
947
{
948
	unsigned long nr = 0;
949 950
	int zid;

951
	VM_BUG_ON((unsigned)nid >= nr_node_ids);
952

953 954 955 956 957 958 959 960 961 962 963 964
	for (zid = 0; zid < MAX_NR_ZONES; zid++) {
		struct mem_cgroup_per_zone *mz;
		enum lru_list lru;

		for_each_lru(lru) {
			if (!(BIT(lru) & lru_mask))
				continue;
			mz = &memcg->nodeinfo[nid]->zoneinfo[zid];
			nr += mz->lru_size[lru];
		}
	}
	return nr;
965
}
966

967
static unsigned long mem_cgroup_nr_lru_pages(struct mem_cgroup *memcg,
968
			unsigned int lru_mask)
969
{
970
	unsigned long nr = 0;
971
	int nid;
972

973
	for_each_node_state(nid, N_MEMORY)
974 975
		nr += mem_cgroup_node_nr_lru_pages(memcg, nid, lru_mask);
	return nr;
976 977
}

978 979
static bool mem_cgroup_event_ratelimit(struct mem_cgroup *memcg,
				       enum mem_cgroup_events_target target)
980 981 982
{
	unsigned long val, next;

983
	val = __this_cpu_read(memcg->stat->nr_page_events);
984
	next = __this_cpu_read(memcg->stat->targets[target]);
985
	/* from time_after() in jiffies.h */
986 987 988 989 990
	if ((long)next - (long)val < 0) {
		switch (target) {
		case MEM_CGROUP_TARGET_THRESH:
			next = val + THRESHOLDS_EVENTS_TARGET;
			break;
991 992 993
		case MEM_CGROUP_TARGET_SOFTLIMIT:
			next = val + SOFTLIMIT_EVENTS_TARGET;
			break;
994 995 996 997 998 999 1000 1001
		case MEM_CGROUP_TARGET_NUMAINFO:
			next = val + NUMAINFO_EVENTS_TARGET;
			break;
		default:
			break;
		}
		__this_cpu_write(memcg->stat->targets[target], next);
		return true;
1002
	}
1003
	return false;
1004 1005 1006 1007 1008 1009
}

/*
 * Check events in order.
 *
 */
1010
static void memcg_check_events(struct mem_cgroup *memcg, struct page *page)
1011 1012
{
	/* threshold event is triggered in finer grain than soft limit */
1013 1014
	if (unlikely(mem_cgroup_event_ratelimit(memcg,
						MEM_CGROUP_TARGET_THRESH))) {
1015
		bool do_softlimit;
1016
		bool do_numainfo __maybe_unused;
1017

1018 1019
		do_softlimit = mem_cgroup_event_ratelimit(memcg,
						MEM_CGROUP_TARGET_SOFTLIMIT);
1020 1021 1022 1023
#if MAX_NUMNODES > 1
		do_numainfo = mem_cgroup_event_ratelimit(memcg,
						MEM_CGROUP_TARGET_NUMAINFO);
#endif
1024
		mem_cgroup_threshold(memcg);
1025 1026
		if (unlikely(do_softlimit))
			mem_cgroup_update_tree(memcg, page);
1027
#if MAX_NUMNODES > 1
1028
		if (unlikely(do_numainfo))
1029
			atomic_inc(&memcg->numainfo_events);
1030
#endif
1031
	}
1032 1033
}

1034
struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p)
1035
{
1036 1037 1038 1039 1040 1041 1042 1043
	/*
	 * 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;

1044
	return mem_cgroup_from_css(task_css(p, memory_cgrp_id));
1045 1046
}

1047
static struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
1048
{
1049
	struct mem_cgroup *memcg = NULL;
1050

1051 1052
	rcu_read_lock();
	do {
1053 1054 1055 1056 1057 1058
		/*
		 * 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))
1059
			memcg = root_mem_cgroup;
1060 1061 1062 1063 1064
		else {
			memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
			if (unlikely(!memcg))
				memcg = root_mem_cgroup;
		}
1065
	} while (!css_tryget_online(&memcg->css));
1066
	rcu_read_unlock();
1067
	return memcg;
1068 1069
}

1070 1071 1072 1073 1074 1075 1076
/*
 * 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,
1077
		struct mem_cgroup *last_visited)
1078
{
1079
	struct cgroup_subsys_state *prev_css, *next_css;
1080

1081
	prev_css = last_visited ? &last_visited->css : NULL;
1082
skip_node:
1083
	next_css = css_next_descendant_pre(prev_css, &root->css);
1084 1085 1086 1087 1088 1089 1090

	/*
	 * 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.
1091 1092 1093 1094 1095 1096 1097 1098
	 *
	 * 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.
1099
	 */
1100
	if (next_css) {
1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115
		struct mem_cgroup *memcg = mem_cgroup_from_css(next_css);

		if (next_css == &root->css)
			return memcg;

		if (css_tryget_online(next_css)) {
			/*
			 * Make sure the memcg is initialized:
			 * mem_cgroup_css_online() orders the the
			 * initialization against setting the flag.
			 */
			if (smp_load_acquire(&memcg->initialized))
				return memcg;
			css_put(next_css);
		}
1116 1117 1118

		prev_css = next_css;
		goto skip_node;
1119 1120 1121 1122 1123
	}

	return NULL;
}

1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151
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;
1152 1153 1154 1155 1156 1157 1158 1159

		/*
		 * 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 &&
1160
		    !css_tryget_online(&position->css))
1161 1162 1163 1164 1165 1166 1167 1168
			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,
1169
				   struct mem_cgroup *root,
1170 1171
				   int sequence)
{
1172 1173
	/* root reference counting symmetric to mem_cgroup_iter_load */
	if (last_visited && last_visited != root)
1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185
		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;
}

1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202
/**
 * 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.
 */
1203
struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *root,
1204
				   struct mem_cgroup *prev,
1205
				   struct mem_cgroup_reclaim_cookie *reclaim)
K
KAMEZAWA Hiroyuki 已提交
1206
{
1207
	struct mem_cgroup *memcg = NULL;
1208
	struct mem_cgroup *last_visited = NULL;
1209

1210 1211
	if (mem_cgroup_disabled())
		return NULL;
1212

1213 1214
	if (!root)
		root = root_mem_cgroup;
K
KAMEZAWA Hiroyuki 已提交
1215

1216
	if (prev && !reclaim)
1217
		last_visited = prev;
K
KAMEZAWA Hiroyuki 已提交
1218

1219 1220
	if (!root->use_hierarchy && root != root_mem_cgroup) {
		if (prev)
1221
			goto out_css_put;
1222
		return root;
1223
	}
K
KAMEZAWA Hiroyuki 已提交
1224

1225
	rcu_read_lock();
1226
	while (!memcg) {
1227
		struct mem_cgroup_reclaim_iter *uninitialized_var(iter);
1228
		int uninitialized_var(seq);
1229

1230 1231 1232
		if (reclaim) {
			struct mem_cgroup_per_zone *mz;

1233
			mz = mem_cgroup_zone_zoneinfo(root, reclaim->zone);
1234
			iter = &mz->reclaim_iter[reclaim->priority];
1235
			if (prev && reclaim->generation != iter->generation) {
M
Michal Hocko 已提交
1236
				iter->last_visited = NULL;
1237 1238
				goto out_unlock;
			}
M
Michal Hocko 已提交
1239

1240
			last_visited = mem_cgroup_iter_load(iter, root, &seq);
1241
		}
K
KAMEZAWA Hiroyuki 已提交
1242

1243
		memcg = __mem_cgroup_iter_next(root, last_visited);
K
KAMEZAWA Hiroyuki 已提交
1244

1245
		if (reclaim) {
1246 1247
			mem_cgroup_iter_update(iter, last_visited, memcg, root,
					seq);
1248

M
Michal Hocko 已提交
1249
			if (!memcg)
1250 1251 1252 1253
				iter->generation++;
			else if (!prev && memcg)
				reclaim->generation = iter->generation;
		}
1254

1255
		if (prev && !memcg)
1256
			goto out_unlock;
1257
	}
1258 1259
out_unlock:
	rcu_read_unlock();
1260 1261 1262 1263
out_css_put:
	if (prev && prev != root)
		css_put(&prev->css);

1264
	return memcg;
K
KAMEZAWA Hiroyuki 已提交
1265
}
K
KAMEZAWA Hiroyuki 已提交
1266

1267 1268 1269 1270 1271 1272 1273
/**
 * 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)
1274 1275 1276 1277 1278 1279
{
	if (!root)
		root = root_mem_cgroup;
	if (prev && prev != root)
		css_put(&prev->css);
}
K
KAMEZAWA Hiroyuki 已提交
1280

1281 1282 1283 1284 1285 1286
/*
 * 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)		\
1287
	for (iter = mem_cgroup_iter(root, NULL, NULL);	\
1288
	     iter != NULL;				\
1289
	     iter = mem_cgroup_iter(root, iter, NULL))
1290

1291
#define for_each_mem_cgroup(iter)			\
1292
	for (iter = mem_cgroup_iter(NULL, NULL, NULL);	\
1293
	     iter != NULL;				\
1294
	     iter = mem_cgroup_iter(NULL, iter, NULL))
K
KAMEZAWA Hiroyuki 已提交
1295

1296
void __mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
1297
{
1298
	struct mem_cgroup *memcg;
1299 1300

	rcu_read_lock();
1301 1302
	memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
	if (unlikely(!memcg))
1303 1304 1305 1306
		goto out;

	switch (idx) {
	case PGFAULT:
1307 1308 1309 1310
		this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]);
		break;
	case PGMAJFAULT:
		this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]);
1311 1312 1313 1314 1315 1316 1317
		break;
	default:
		BUG();
	}
out:
	rcu_read_unlock();
}
1318
EXPORT_SYMBOL(__mem_cgroup_count_vm_event);
1319

1320 1321 1322
/**
 * mem_cgroup_zone_lruvec - get the lru list vector for a zone and memcg
 * @zone: zone of the wanted lruvec
1323
 * @memcg: memcg of the wanted lruvec
1324 1325 1326 1327 1328 1329 1330 1331 1332
 *
 * 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;
1333
	struct lruvec *lruvec;
1334

1335 1336 1337 1338
	if (mem_cgroup_disabled()) {
		lruvec = &zone->lruvec;
		goto out;
	}
1339

1340
	mz = mem_cgroup_zone_zoneinfo(memcg, zone);
1341 1342 1343 1344 1345 1346 1347 1348 1349 1350
	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;
1351 1352 1353
}

/**
1354
 * mem_cgroup_page_lruvec - return lruvec for adding an lru page
1355
 * @page: the page
1356
 * @zone: zone of the page
1357
 */
1358
struct lruvec *mem_cgroup_page_lruvec(struct page *page, struct zone *zone)
K
KAMEZAWA Hiroyuki 已提交
1359 1360
{
	struct mem_cgroup_per_zone *mz;
1361 1362
	struct mem_cgroup *memcg;
	struct page_cgroup *pc;
1363
	struct lruvec *lruvec;
1364

1365 1366 1367 1368
	if (mem_cgroup_disabled()) {
		lruvec = &zone->lruvec;
		goto out;
	}
1369

K
KAMEZAWA Hiroyuki 已提交
1370
	pc = lookup_page_cgroup(page);
1371
	memcg = pc->mem_cgroup;
1372 1373

	/*
1374
	 * Surreptitiously switch any uncharged offlist page to root:
1375 1376 1377 1378 1379 1380 1381
	 * 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.
	 */
1382
	if (!PageLRU(page) && !PageCgroupUsed(pc) && memcg != root_mem_cgroup)
1383 1384
		pc->mem_cgroup = memcg = root_mem_cgroup;

1385
	mz = mem_cgroup_page_zoneinfo(memcg, page);
1386 1387 1388 1389 1390 1391 1392 1393 1394 1395
	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
1396
}
1397

1398
/**
1399 1400 1401 1402
 * 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
1403
 *
1404 1405
 * This function must be called when a page is added to or removed from an
 * lru list.
1406
 */
1407 1408
void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
				int nr_pages)
1409 1410
{
	struct mem_cgroup_per_zone *mz;
1411
	unsigned long *lru_size;
1412 1413 1414 1415

	if (mem_cgroup_disabled())
		return;

1416 1417 1418 1419
	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
1420
}
1421

1422
/*
1423
 * Checks whether given mem is same or in the root_mem_cgroup's
1424 1425
 * hierarchy subtree
 */
1426 1427
bool __mem_cgroup_same_or_subtree(const struct mem_cgroup *root_memcg,
				  struct mem_cgroup *memcg)
1428
{
1429 1430
	if (root_memcg == memcg)
		return true;
1431
	if (!root_memcg->use_hierarchy || !memcg)
1432
		return false;
1433
	return cgroup_is_descendant(memcg->css.cgroup, root_memcg->css.cgroup);
1434 1435 1436 1437 1438 1439 1440
}

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

1441
	rcu_read_lock();
1442
	ret = __mem_cgroup_same_or_subtree(root_memcg, memcg);
1443 1444
	rcu_read_unlock();
	return ret;
1445 1446
}

1447 1448
bool task_in_mem_cgroup(struct task_struct *task,
			const struct mem_cgroup *memcg)
1449
{
1450
	struct mem_cgroup *curr = NULL;
1451
	struct task_struct *p;
1452
	bool ret;
1453

1454
	p = find_lock_task_mm(task);
1455
	if (p) {
1456
		curr = get_mem_cgroup_from_mm(p->mm);
1457 1458 1459 1460 1461 1462 1463
		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.
		 */
1464
		rcu_read_lock();
1465 1466 1467
		curr = mem_cgroup_from_task(task);
		if (curr)
			css_get(&curr->css);
1468
		rcu_read_unlock();
1469
	}
1470
	/*
1471
	 * We should check use_hierarchy of "memcg" not "curr". Because checking
1472
	 * use_hierarchy of "curr" here make this function true if hierarchy is
1473 1474
	 * enabled in "curr" and "curr" is a child of "memcg" in *cgroup*
	 * hierarchy(even if use_hierarchy is disabled in "memcg").
1475
	 */
1476
	ret = mem_cgroup_same_or_subtree(memcg, curr);
1477
	css_put(&curr->css);
1478 1479 1480
	return ret;
}

1481
int mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
1482
{
1483
	unsigned long inactive_ratio;
1484
	unsigned long inactive;
1485
	unsigned long active;
1486
	unsigned long gb;
1487

1488 1489
	inactive = mem_cgroup_get_lru_size(lruvec, LRU_INACTIVE_ANON);
	active = mem_cgroup_get_lru_size(lruvec, LRU_ACTIVE_ANON);
1490

1491 1492 1493 1494 1495 1496
	gb = (inactive + active) >> (30 - PAGE_SHIFT);
	if (gb)
		inactive_ratio = int_sqrt(10 * gb);
	else
		inactive_ratio = 1;

1497
	return inactive * inactive_ratio < active;
1498 1499
}

1500
#define mem_cgroup_from_counter(counter, member)	\
1501 1502
	container_of(counter, struct mem_cgroup, member)

1503
/**
1504
 * mem_cgroup_margin - calculate chargeable space of a memory cgroup
W
Wanpeng Li 已提交
1505
 * @memcg: the memory cgroup
1506
 *
1507
 * Returns the maximum amount of memory @mem can be charged with, in
1508
 * pages.
1509
 */
1510
static unsigned long mem_cgroup_margin(struct mem_cgroup *memcg)
1511
{
1512 1513 1514
	unsigned long margin = 0;
	unsigned long count;
	unsigned long limit;
1515

1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528
	count = page_counter_read(&memcg->memory);
	limit = ACCESS_ONCE(memcg->memory.limit);
	if (count < limit)
		margin = limit - count;

	if (do_swap_account) {
		count = page_counter_read(&memcg->memsw);
		limit = ACCESS_ONCE(memcg->memsw.limit);
		if (count <= limit)
			margin = min(margin, limit - count);
	}

	return margin;
1529 1530
}

1531
int mem_cgroup_swappiness(struct mem_cgroup *memcg)
K
KOSAKI Motohiro 已提交
1532 1533
{
	/* root ? */
1534
	if (mem_cgroup_disabled() || !memcg->css.parent)
K
KOSAKI Motohiro 已提交
1535 1536
		return vm_swappiness;

1537
	return memcg->swappiness;
K
KOSAKI Motohiro 已提交
1538 1539
}

1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553
/*
 * 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.
 */
1554

1555
static void mem_cgroup_start_move(struct mem_cgroup *memcg)
1556
{
1557
	atomic_inc(&memcg->moving_account);
1558 1559 1560
	synchronize_rcu();
}

1561
static void mem_cgroup_end_move(struct mem_cgroup *memcg)
1562
{
1563 1564 1565 1566
	/*
	 * Now, mem_cgroup_clear_mc() may call this function with NULL.
	 * We check NULL in callee rather than caller.
	 */
1567
	if (memcg)
1568
		atomic_dec(&memcg->moving_account);
1569
}
1570

1571
/*
Q
Qiang Huang 已提交
1572
 * A routine for checking "mem" is under move_account() or not.
1573
 *
Q
Qiang Huang 已提交
1574 1575 1576
 * Checking a cgroup is mc.from or mc.to or under hierarchy of
 * moving cgroups. This is for waiting at high-memory pressure
 * caused by "move".
1577
 */
1578
static bool mem_cgroup_under_move(struct mem_cgroup *memcg)
1579
{
1580 1581
	struct mem_cgroup *from;
	struct mem_cgroup *to;
1582
	bool ret = false;
1583 1584 1585 1586 1587 1588 1589 1590 1591
	/*
	 * 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;
1592

1593 1594
	ret = mem_cgroup_same_or_subtree(memcg, from)
		|| mem_cgroup_same_or_subtree(memcg, to);
1595 1596
unlock:
	spin_unlock(&mc.lock);
1597 1598 1599
	return ret;
}

1600
static bool mem_cgroup_wait_acct_move(struct mem_cgroup *memcg)
1601 1602
{
	if (mc.moving_task && current != mc.moving_task) {
1603
		if (mem_cgroup_under_move(memcg)) {
1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615
			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;
}

1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632
/*
 * 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.
 */
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);
}

1633
#define K(x) ((x) << (PAGE_SHIFT-10))
1634
/**
1635
 * mem_cgroup_print_oom_info: Print OOM information relevant to memory controller.
1636 1637 1638 1639 1640 1641 1642 1643
 * @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
1644
	/* oom_info_lock ensures that parallel ooms do not interleave */
1645
	static DEFINE_MUTEX(oom_info_lock);
1646 1647
	struct mem_cgroup *iter;
	unsigned int i;
1648

1649
	if (!p)
1650 1651
		return;

1652
	mutex_lock(&oom_info_lock);
1653 1654
	rcu_read_lock();

T
Tejun Heo 已提交
1655 1656 1657 1658 1659
	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");
1660 1661 1662

	rcu_read_unlock();

1663 1664 1665 1666 1667 1668 1669 1670 1671
	pr_info("memory: usage %llukB, limit %llukB, failcnt %lu\n",
		K((u64)page_counter_read(&memcg->memory)),
		K((u64)memcg->memory.limit), memcg->memory.failcnt);
	pr_info("memory+swap: usage %llukB, limit %llukB, failcnt %lu\n",
		K((u64)page_counter_read(&memcg->memsw)),
		K((u64)memcg->memsw.limit), memcg->memsw.failcnt);
	pr_info("kmem: usage %llukB, limit %llukB, failcnt %lu\n",
		K((u64)page_counter_read(&memcg->kmem)),
		K((u64)memcg->kmem.limit), memcg->kmem.failcnt);
1672 1673

	for_each_mem_cgroup_tree(iter, memcg) {
T
Tejun Heo 已提交
1674 1675
		pr_info("Memory cgroup stats for ");
		pr_cont_cgroup_path(iter->css.cgroup);
1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690
		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");
	}
1691
	mutex_unlock(&oom_info_lock);
1692 1693
}

1694 1695 1696 1697
/*
 * This function returns the number of memcg under hierarchy tree. Returns
 * 1(self count) if no children.
 */
1698
static int mem_cgroup_count_children(struct mem_cgroup *memcg)
1699 1700
{
	int num = 0;
K
KAMEZAWA Hiroyuki 已提交
1701 1702
	struct mem_cgroup *iter;

1703
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
1704
		num++;
1705 1706 1707
	return num;
}

D
David Rientjes 已提交
1708 1709 1710
/*
 * Return the memory (and swap, if configured) limit for a memcg.
 */
1711
static unsigned long mem_cgroup_get_limit(struct mem_cgroup *memcg)
D
David Rientjes 已提交
1712
{
1713
	unsigned long limit;
D
David Rientjes 已提交
1714

1715
	limit = memcg->memory.limit;
1716
	if (mem_cgroup_swappiness(memcg)) {
1717
		unsigned long memsw_limit;
1718

1719 1720
		memsw_limit = memcg->memsw.limit;
		limit = min(limit + total_swap_pages, memsw_limit);
1721 1722
	}
	return limit;
D
David Rientjes 已提交
1723 1724
}

1725 1726
static void mem_cgroup_out_of_memory(struct mem_cgroup *memcg, gfp_t gfp_mask,
				     int order)
1727 1728 1729 1730 1731 1732 1733
{
	struct mem_cgroup *iter;
	unsigned long chosen_points = 0;
	unsigned long totalpages;
	unsigned int points = 0;
	struct task_struct *chosen = NULL;

1734
	/*
1735 1736 1737
	 * 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.
1738
	 */
1739
	if (fatal_signal_pending(current) || current->flags & PF_EXITING) {
1740 1741 1742 1743 1744
		set_thread_flag(TIF_MEMDIE);
		return;
	}

	check_panic_on_oom(CONSTRAINT_MEMCG, gfp_mask, order, NULL);
1745
	totalpages = mem_cgroup_get_limit(memcg) ? : 1;
1746
	for_each_mem_cgroup_tree(iter, memcg) {
1747
		struct css_task_iter it;
1748 1749
		struct task_struct *task;

1750 1751
		css_task_iter_start(&iter->css, &it);
		while ((task = css_task_iter_next(&it))) {
1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763
			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:
1764
				css_task_iter_end(&it);
1765 1766 1767 1768 1769 1770 1771 1772
				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);
1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784
			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);
1785
		}
1786
		css_task_iter_end(&it);
1787 1788 1789 1790 1791 1792 1793 1794 1795
	}

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

1796 1797
/**
 * test_mem_cgroup_node_reclaimable
W
Wanpeng Li 已提交
1798
 * @memcg: the target memcg
1799 1800 1801 1802 1803 1804 1805
 * @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.
 */
1806
static bool test_mem_cgroup_node_reclaimable(struct mem_cgroup *memcg,
1807 1808
		int nid, bool noswap)
{
1809
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_FILE))
1810 1811 1812
		return true;
	if (noswap || !total_swap_pages)
		return false;
1813
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_ANON))
1814 1815 1816 1817
		return true;
	return false;

}
1818
#if MAX_NUMNODES > 1
1819 1820 1821 1822 1823 1824 1825

/*
 * 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.
 *
 */
1826
static void mem_cgroup_may_update_nodemask(struct mem_cgroup *memcg)
1827 1828
{
	int nid;
1829 1830 1831 1832
	/*
	 * numainfo_events > 0 means there was at least NUMAINFO_EVENTS_TARGET
	 * pagein/pageout changes since the last update.
	 */
1833
	if (!atomic_read(&memcg->numainfo_events))
1834
		return;
1835
	if (atomic_inc_return(&memcg->numainfo_updating) > 1)
1836 1837 1838
		return;

	/* make a nodemask where this memcg uses memory from */
1839
	memcg->scan_nodes = node_states[N_MEMORY];
1840

1841
	for_each_node_mask(nid, node_states[N_MEMORY]) {
1842

1843 1844
		if (!test_mem_cgroup_node_reclaimable(memcg, nid, false))
			node_clear(nid, memcg->scan_nodes);
1845
	}
1846

1847 1848
	atomic_set(&memcg->numainfo_events, 0);
	atomic_set(&memcg->numainfo_updating, 0);
1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862
}

/*
 * 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.
 */
1863
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
1864 1865 1866
{
	int node;

1867 1868
	mem_cgroup_may_update_nodemask(memcg);
	node = memcg->last_scanned_node;
1869

1870
	node = next_node(node, memcg->scan_nodes);
1871
	if (node == MAX_NUMNODES)
1872
		node = first_node(memcg->scan_nodes);
1873 1874 1875 1876 1877 1878 1879 1880 1881
	/*
	 * 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();

1882
	memcg->last_scanned_node = node;
1883 1884 1885
	return node;
}

1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920
/*
 * 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;
}

1921
#else
1922
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
1923 1924 1925
{
	return 0;
}
1926

1927 1928 1929 1930
static bool mem_cgroup_reclaimable(struct mem_cgroup *memcg, bool noswap)
{
	return test_mem_cgroup_node_reclaimable(memcg, 0, noswap);
}
1931 1932
#endif

1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947
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,
	};

1948
	excess = soft_limit_excess(root_memcg);
1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978

	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;
1979
		if (!soft_limit_excess(root_memcg))
1980
			break;
1981
	}
1982 1983
	mem_cgroup_iter_break(root_memcg, victim);
	return total;
1984 1985
}

1986 1987 1988 1989 1990 1991
#ifdef CONFIG_LOCKDEP
static struct lockdep_map memcg_oom_lock_dep_map = {
	.name = "memcg_oom_lock",
};
#endif

1992 1993
static DEFINE_SPINLOCK(memcg_oom_lock);

K
KAMEZAWA Hiroyuki 已提交
1994 1995 1996 1997
/*
 * Check OOM-Killer is already running under our hierarchy.
 * If someone is running, return false.
 */
1998
static bool mem_cgroup_oom_trylock(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
1999
{
2000
	struct mem_cgroup *iter, *failed = NULL;
2001

2002 2003
	spin_lock(&memcg_oom_lock);

2004
	for_each_mem_cgroup_tree(iter, memcg) {
2005
		if (iter->oom_lock) {
2006 2007 2008 2009 2010
			/*
			 * this subtree of our hierarchy is already locked
			 * so we cannot give a lock.
			 */
			failed = iter;
2011 2012
			mem_cgroup_iter_break(memcg, iter);
			break;
2013 2014
		} else
			iter->oom_lock = true;
K
KAMEZAWA Hiroyuki 已提交
2015
	}
K
KAMEZAWA Hiroyuki 已提交
2016

2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027
	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;
2028
		}
2029 2030
	} else
		mutex_acquire(&memcg_oom_lock_dep_map, 0, 1, _RET_IP_);
2031 2032 2033 2034

	spin_unlock(&memcg_oom_lock);

	return !failed;
2035
}
2036

2037
static void mem_cgroup_oom_unlock(struct mem_cgroup *memcg)
2038
{
K
KAMEZAWA Hiroyuki 已提交
2039 2040
	struct mem_cgroup *iter;

2041
	spin_lock(&memcg_oom_lock);
2042
	mutex_release(&memcg_oom_lock_dep_map, 1, _RET_IP_);
2043
	for_each_mem_cgroup_tree(iter, memcg)
2044
		iter->oom_lock = false;
2045
	spin_unlock(&memcg_oom_lock);
2046 2047
}

2048
static void mem_cgroup_mark_under_oom(struct mem_cgroup *memcg)
2049 2050 2051
{
	struct mem_cgroup *iter;

2052
	for_each_mem_cgroup_tree(iter, memcg)
2053 2054 2055
		atomic_inc(&iter->under_oom);
}

2056
static void mem_cgroup_unmark_under_oom(struct mem_cgroup *memcg)
2057 2058 2059
{
	struct mem_cgroup *iter;

K
KAMEZAWA Hiroyuki 已提交
2060 2061 2062 2063 2064
	/*
	 * 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.
	 */
2065
	for_each_mem_cgroup_tree(iter, memcg)
2066
		atomic_add_unless(&iter->under_oom, -1, 0);
2067 2068
}

K
KAMEZAWA Hiroyuki 已提交
2069 2070
static DECLARE_WAIT_QUEUE_HEAD(memcg_oom_waitq);

K
KAMEZAWA Hiroyuki 已提交
2071
struct oom_wait_info {
2072
	struct mem_cgroup *memcg;
K
KAMEZAWA Hiroyuki 已提交
2073 2074 2075 2076 2077 2078
	wait_queue_t	wait;
};

static int memcg_oom_wake_function(wait_queue_t *wait,
	unsigned mode, int sync, void *arg)
{
2079 2080
	struct mem_cgroup *wake_memcg = (struct mem_cgroup *)arg;
	struct mem_cgroup *oom_wait_memcg;
K
KAMEZAWA Hiroyuki 已提交
2081 2082 2083
	struct oom_wait_info *oom_wait_info;

	oom_wait_info = container_of(wait, struct oom_wait_info, wait);
2084
	oom_wait_memcg = oom_wait_info->memcg;
K
KAMEZAWA Hiroyuki 已提交
2085 2086

	/*
2087
	 * Both of oom_wait_info->memcg and wake_memcg are stable under us.
K
KAMEZAWA Hiroyuki 已提交
2088 2089
	 * Then we can use css_is_ancestor without taking care of RCU.
	 */
2090 2091
	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
2092 2093 2094 2095
		return 0;
	return autoremove_wake_function(wait, mode, sync, arg);
}

2096
static void memcg_wakeup_oom(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
2097
{
2098
	atomic_inc(&memcg->oom_wakeups);
2099 2100
	/* for filtering, pass "memcg" as argument. */
	__wake_up(&memcg_oom_waitq, TASK_NORMAL, 0, memcg);
K
KAMEZAWA Hiroyuki 已提交
2101 2102
}

2103
static void memcg_oom_recover(struct mem_cgroup *memcg)
2104
{
2105 2106
	if (memcg && atomic_read(&memcg->under_oom))
		memcg_wakeup_oom(memcg);
2107 2108
}

2109
static void mem_cgroup_oom(struct mem_cgroup *memcg, gfp_t mask, int order)
2110
{
2111 2112
	if (!current->memcg_oom.may_oom)
		return;
K
KAMEZAWA Hiroyuki 已提交
2113
	/*
2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125
	 * 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
2126
	 */
2127 2128 2129 2130
	css_get(&memcg->css);
	current->memcg_oom.memcg = memcg;
	current->memcg_oom.gfp_mask = mask;
	current->memcg_oom.order = order;
2131 2132 2133 2134
}

/**
 * mem_cgroup_oom_synchronize - complete memcg OOM handling
2135
 * @handle: actually kill/wait or just clean up the OOM state
2136
 *
2137 2138
 * This has to be called at the end of a page fault if the memcg OOM
 * handler was enabled.
2139
 *
2140
 * Memcg supports userspace OOM handling where failed allocations must
2141 2142 2143 2144
 * 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
2145
 * the end of the page fault to complete the OOM handling.
2146 2147
 *
 * Returns %true if an ongoing memcg OOM situation was detected and
2148
 * completed, %false otherwise.
2149
 */
2150
bool mem_cgroup_oom_synchronize(bool handle)
2151
{
2152
	struct mem_cgroup *memcg = current->memcg_oom.memcg;
2153
	struct oom_wait_info owait;
2154
	bool locked;
2155 2156 2157

	/* OOM is global, do not handle */
	if (!memcg)
2158
		return false;
2159

2160 2161
	if (!handle)
		goto cleanup;
2162 2163 2164 2165 2166 2167

	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
2168

2169
	prepare_to_wait(&memcg_oom_waitq, &owait.wait, TASK_KILLABLE);
2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182
	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 {
2183
		schedule();
2184 2185 2186 2187 2188
		mem_cgroup_unmark_under_oom(memcg);
		finish_wait(&memcg_oom_waitq, &owait.wait);
	}

	if (locked) {
2189 2190 2191 2192 2193 2194 2195 2196
		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);
	}
2197 2198
cleanup:
	current->memcg_oom.memcg = NULL;
2199
	css_put(&memcg->css);
K
KAMEZAWA Hiroyuki 已提交
2200
	return true;
2201 2202
}

2203 2204 2205 2206 2207
/**
 * mem_cgroup_begin_page_stat - begin a page state statistics transaction
 * @page: page that is going to change accounted state
 * @locked: &memcg->move_lock slowpath was taken
 * @flags: IRQ-state flags for &memcg->move_lock
2208
 *
2209 2210 2211
 * This function must mark the beginning of an accounted page state
 * change to prevent double accounting when the page is concurrently
 * being moved to another memcg:
2212
 *
2213 2214 2215 2216
 *   memcg = mem_cgroup_begin_page_stat(page, &locked, &flags);
 *   if (TestClearPageState(page))
 *     mem_cgroup_update_page_stat(memcg, state, -1);
 *   mem_cgroup_end_page_stat(memcg, locked, flags);
2217
 *
2218 2219 2220
 * The RCU lock is held throughout the transaction.  The fast path can
 * get away without acquiring the memcg->move_lock (@locked is false)
 * because page moving starts with an RCU grace period.
2221
 *
2222 2223 2224 2225 2226
 * The RCU lock also protects the memcg from being freed when the page
 * state that is going to change is the only thing preventing the page
 * from being uncharged.  E.g. end-writeback clearing PageWriteback(),
 * which allows migration to go ahead and uncharge the page before the
 * account transaction might be complete.
2227
 */
2228 2229 2230
struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page,
					      bool *locked,
					      unsigned long *flags)
2231 2232 2233 2234
{
	struct mem_cgroup *memcg;
	struct page_cgroup *pc;

2235 2236 2237 2238 2239
	rcu_read_lock();

	if (mem_cgroup_disabled())
		return NULL;

2240 2241 2242 2243
	pc = lookup_page_cgroup(page);
again:
	memcg = pc->mem_cgroup;
	if (unlikely(!memcg || !PageCgroupUsed(pc)))
2244 2245 2246
		return NULL;

	*locked = false;
Q
Qiang Huang 已提交
2247
	if (atomic_read(&memcg->moving_account) <= 0)
2248
		return memcg;
2249 2250 2251 2252 2253 2254 2255

	move_lock_mem_cgroup(memcg, flags);
	if (memcg != pc->mem_cgroup || !PageCgroupUsed(pc)) {
		move_unlock_mem_cgroup(memcg, flags);
		goto again;
	}
	*locked = true;
2256 2257

	return memcg;
2258 2259
}

2260 2261 2262 2263 2264 2265 2266 2267
/**
 * mem_cgroup_end_page_stat - finish a page state statistics transaction
 * @memcg: the memcg that was accounted against
 * @locked: value received from mem_cgroup_begin_page_stat()
 * @flags: value received from mem_cgroup_begin_page_stat()
 */
void mem_cgroup_end_page_stat(struct mem_cgroup *memcg, bool locked,
			      unsigned long flags)
2268
{
2269 2270
	if (memcg && locked)
		move_unlock_mem_cgroup(memcg, &flags);
2271

2272
	rcu_read_unlock();
2273 2274
}

2275 2276 2277 2278 2279 2280 2281 2282 2283
/**
 * mem_cgroup_update_page_stat - update page state statistics
 * @memcg: memcg to account against
 * @idx: page state item to account
 * @val: number of pages (positive or negative)
 *
 * See mem_cgroup_begin_page_stat() for locking requirements.
 */
void mem_cgroup_update_page_stat(struct mem_cgroup *memcg,
S
Sha Zhengju 已提交
2284
				 enum mem_cgroup_stat_index idx, int val)
2285
{
2286
	VM_BUG_ON(!rcu_read_lock_held());
2287

2288 2289
	if (memcg)
		this_cpu_add(memcg->stat->count[idx], val);
2290
}
2291

2292 2293 2294 2295
/*
 * size of first charge trial. "32" comes from vmscan.c's magic value.
 * TODO: maybe necessary to use big numbers in big irons.
 */
2296
#define CHARGE_BATCH	32U
2297 2298
struct memcg_stock_pcp {
	struct mem_cgroup *cached; /* this never be root cgroup */
2299
	unsigned int nr_pages;
2300
	struct work_struct work;
2301
	unsigned long flags;
2302
#define FLUSHING_CACHED_CHARGE	0
2303 2304
};
static DEFINE_PER_CPU(struct memcg_stock_pcp, memcg_stock);
2305
static DEFINE_MUTEX(percpu_charge_mutex);
2306

2307 2308 2309 2310 2311 2312 2313 2314 2315 2316
/**
 * 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.
2317
 */
2318
static bool consume_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
2319 2320
{
	struct memcg_stock_pcp *stock;
2321
	bool ret = false;
2322

2323
	if (nr_pages > CHARGE_BATCH)
2324
		return ret;
2325

2326
	stock = &get_cpu_var(memcg_stock);
2327
	if (memcg == stock->cached && stock->nr_pages >= nr_pages) {
2328
		stock->nr_pages -= nr_pages;
2329 2330
		ret = true;
	}
2331 2332 2333 2334 2335
	put_cpu_var(memcg_stock);
	return ret;
}

/*
2336
 * Returns stocks cached in percpu and reset cached information.
2337 2338 2339 2340 2341
 */
static void drain_stock(struct memcg_stock_pcp *stock)
{
	struct mem_cgroup *old = stock->cached;

2342
	if (stock->nr_pages) {
2343
		page_counter_uncharge(&old->memory, stock->nr_pages);
2344
		if (do_swap_account)
2345
			page_counter_uncharge(&old->memsw, stock->nr_pages);
2346
		stock->nr_pages = 0;
2347 2348 2349 2350 2351 2352 2353 2354 2355 2356
	}
	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)
{
2357
	struct memcg_stock_pcp *stock = this_cpu_ptr(&memcg_stock);
2358
	drain_stock(stock);
2359
	clear_bit(FLUSHING_CACHED_CHARGE, &stock->flags);
2360 2361
}

2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372
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);
	}
}

2373
/*
2374
 * Cache charges(val) to local per_cpu area.
2375
 * This will be consumed by consume_stock() function, later.
2376
 */
2377
static void refill_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
2378 2379 2380
{
	struct memcg_stock_pcp *stock = &get_cpu_var(memcg_stock);

2381
	if (stock->cached != memcg) { /* reset if necessary */
2382
		drain_stock(stock);
2383
		stock->cached = memcg;
2384
	}
2385
	stock->nr_pages += nr_pages;
2386 2387 2388 2389
	put_cpu_var(memcg_stock);
}

/*
2390
 * Drains all per-CPU charge caches for given root_memcg resp. subtree
2391 2392
 * of the hierarchy under it. sync flag says whether we should block
 * until the work is done.
2393
 */
2394
static void drain_all_stock(struct mem_cgroup *root_memcg, bool sync)
2395
{
2396
	int cpu, curcpu;
2397

2398 2399
	/* Notify other cpus that system-wide "drain" is running */
	get_online_cpus();
2400
	curcpu = get_cpu();
2401 2402
	for_each_online_cpu(cpu) {
		struct memcg_stock_pcp *stock = &per_cpu(memcg_stock, cpu);
2403
		struct mem_cgroup *memcg;
2404

2405 2406
		memcg = stock->cached;
		if (!memcg || !stock->nr_pages)
2407
			continue;
2408
		if (!mem_cgroup_same_or_subtree(root_memcg, memcg))
2409
			continue;
2410 2411 2412 2413 2414 2415
		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);
		}
2416
	}
2417
	put_cpu();
2418 2419 2420 2421 2422 2423

	if (!sync)
		goto out;

	for_each_online_cpu(cpu) {
		struct memcg_stock_pcp *stock = &per_cpu(memcg_stock, cpu);
2424
		if (test_bit(FLUSHING_CACHED_CHARGE, &stock->flags))
2425 2426 2427
			flush_work(&stock->work);
	}
out:
A
Andrew Morton 已提交
2428
	put_online_cpus();
2429 2430 2431 2432 2433
}

/*
 * 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
2434
 * expects some charges will be back later but cannot wait for it.
2435
 */
2436
static void drain_all_stock_async(struct mem_cgroup *root_memcg)
2437
{
2438 2439 2440 2441 2442
	/*
	 * If someone calls draining, avoid adding more kworker runs.
	 */
	if (!mutex_trylock(&percpu_charge_mutex))
		return;
2443
	drain_all_stock(root_memcg, false);
2444
	mutex_unlock(&percpu_charge_mutex);
2445 2446 2447
}

/* This is a synchronous drain interface. */
2448
static void drain_all_stock_sync(struct mem_cgroup *root_memcg)
2449 2450
{
	/* called when force_empty is called */
2451
	mutex_lock(&percpu_charge_mutex);
2452
	drain_all_stock(root_memcg, true);
2453
	mutex_unlock(&percpu_charge_mutex);
2454 2455
}

2456 2457 2458 2459
/*
 * This function drains percpu counter value from DEAD cpu and
 * move it to local cpu. Note that this function can be preempted.
 */
2460
static void mem_cgroup_drain_pcp_counter(struct mem_cgroup *memcg, int cpu)
2461 2462 2463
{
	int i;

2464
	spin_lock(&memcg->pcp_counter_lock);
2465
	for (i = 0; i < MEM_CGROUP_STAT_NSTATS; i++) {
2466
		long x = per_cpu(memcg->stat->count[i], cpu);
2467

2468 2469
		per_cpu(memcg->stat->count[i], cpu) = 0;
		memcg->nocpu_base.count[i] += x;
2470
	}
2471
	for (i = 0; i < MEM_CGROUP_EVENTS_NSTATS; i++) {
2472
		unsigned long x = per_cpu(memcg->stat->events[i], cpu);
2473

2474 2475
		per_cpu(memcg->stat->events[i], cpu) = 0;
		memcg->nocpu_base.events[i] += x;
2476
	}
2477
	spin_unlock(&memcg->pcp_counter_lock);
2478 2479
}

2480
static int memcg_cpu_hotplug_callback(struct notifier_block *nb,
2481 2482 2483 2484 2485
					unsigned long action,
					void *hcpu)
{
	int cpu = (unsigned long)hcpu;
	struct memcg_stock_pcp *stock;
2486
	struct mem_cgroup *iter;
2487

2488
	if (action == CPU_ONLINE)
2489 2490
		return NOTIFY_OK;

2491
	if (action != CPU_DEAD && action != CPU_DEAD_FROZEN)
2492
		return NOTIFY_OK;
2493

2494
	for_each_mem_cgroup(iter)
2495 2496
		mem_cgroup_drain_pcp_counter(iter, cpu);

2497 2498 2499 2500 2501
	stock = &per_cpu(memcg_stock, cpu);
	drain_stock(stock);
	return NOTIFY_OK;
}

2502 2503
static int try_charge(struct mem_cgroup *memcg, gfp_t gfp_mask,
		      unsigned int nr_pages)
2504
{
2505
	unsigned int batch = max(CHARGE_BATCH, nr_pages);
2506
	int nr_retries = MEM_CGROUP_RECLAIM_RETRIES;
2507
	struct mem_cgroup *mem_over_limit;
2508
	struct page_counter *counter;
2509
	unsigned long nr_reclaimed;
2510 2511
	bool may_swap = true;
	bool drained = false;
2512
	int ret = 0;
2513

2514 2515
	if (mem_cgroup_is_root(memcg))
		goto done;
2516
retry:
2517 2518
	if (consume_stock(memcg, nr_pages))
		goto done;
2519

2520
	if (!do_swap_account ||
2521 2522
	    !page_counter_try_charge(&memcg->memsw, batch, &counter)) {
		if (!page_counter_try_charge(&memcg->memory, batch, &counter))
2523
			goto done_restock;
2524
		if (do_swap_account)
2525 2526
			page_counter_uncharge(&memcg->memsw, batch);
		mem_over_limit = mem_cgroup_from_counter(counter, memory);
2527
	} else {
2528
		mem_over_limit = mem_cgroup_from_counter(counter, memsw);
2529
		may_swap = false;
2530
	}
2531

2532 2533 2534 2535
	if (batch > nr_pages) {
		batch = nr_pages;
		goto retry;
	}
2536

2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550
	/*
	 * 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.
	 */
	if (unlikely(test_thread_flag(TIF_MEMDIE) ||
		     fatal_signal_pending(current) ||
		     current->flags & PF_EXITING))
		goto bypass;

	if (unlikely(task_in_memcg_oom(current)))
		goto nomem;

2551 2552
	if (!(gfp_mask & __GFP_WAIT))
		goto nomem;
2553

2554 2555
	nr_reclaimed = try_to_free_mem_cgroup_pages(mem_over_limit, nr_pages,
						    gfp_mask, may_swap);
2556

2557
	if (mem_cgroup_margin(mem_over_limit) >= nr_pages)
2558
		goto retry;
2559

2560 2561 2562 2563 2564 2565
	if (!drained) {
		drain_all_stock_async(mem_over_limit);
		drained = true;
		goto retry;
	}

2566 2567
	if (gfp_mask & __GFP_NORETRY)
		goto nomem;
2568 2569 2570 2571 2572 2573 2574 2575 2576
	/*
	 * 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.
	 */
2577
	if (nr_reclaimed && nr_pages <= (1 << PAGE_ALLOC_COSTLY_ORDER))
2578 2579 2580 2581 2582 2583 2584 2585
		goto 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))
		goto retry;

2586 2587 2588
	if (nr_retries--)
		goto retry;

2589 2590 2591
	if (gfp_mask & __GFP_NOFAIL)
		goto bypass;

2592 2593 2594
	if (fatal_signal_pending(current))
		goto bypass;

2595
	mem_cgroup_oom(mem_over_limit, gfp_mask, get_order(nr_pages));
2596
nomem:
2597
	if (!(gfp_mask & __GFP_NOFAIL))
2598
		return -ENOMEM;
K
KAMEZAWA Hiroyuki 已提交
2599
bypass:
2600
	return -EINTR;
2601 2602 2603 2604 2605

done_restock:
	if (batch > nr_pages)
		refill_stock(memcg, batch - nr_pages);
done:
2606
	return ret;
2607
}
2608

2609
static void cancel_charge(struct mem_cgroup *memcg, unsigned int nr_pages)
2610
{
2611 2612 2613
	if (mem_cgroup_is_root(memcg))
		return;

2614
	page_counter_uncharge(&memcg->memory, nr_pages);
2615
	if (do_swap_account)
2616
		page_counter_uncharge(&memcg->memsw, nr_pages);
2617 2618
}

2619 2620
/*
 * A helper function to get mem_cgroup from ID. must be called under
2621 2622 2623
 * rcu_read_lock().  The caller is responsible for calling
 * css_tryget_online() if the mem_cgroup is used for charging. (dropping
 * refcnt from swap can be called against removed memcg.)
2624 2625 2626 2627 2628 2629
 */
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
2630
	return mem_cgroup_from_id(id);
2631 2632
}

2633 2634 2635 2636 2637 2638 2639 2640 2641 2642
/*
 * try_get_mem_cgroup_from_page - look up page's memcg association
 * @page: the page
 *
 * Look up, get a css reference, and return the memcg that owns @page.
 *
 * The page must be locked to prevent racing with swap-in and page
 * cache charges.  If coming from an unlocked page table, the caller
 * must ensure the page is on the LRU or this can race with charging.
 */
2643
struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page)
2644
{
2645
	struct mem_cgroup *memcg = NULL;
2646
	struct page_cgroup *pc;
2647
	unsigned short id;
2648 2649
	swp_entry_t ent;

2650
	VM_BUG_ON_PAGE(!PageLocked(page), page);
2651 2652

	pc = lookup_page_cgroup(page);
2653
	if (PageCgroupUsed(pc)) {
2654
		memcg = pc->mem_cgroup;
2655
		if (memcg && !css_tryget_online(&memcg->css))
2656
			memcg = NULL;
2657
	} else if (PageSwapCache(page)) {
2658
		ent.val = page_private(page);
2659
		id = lookup_swap_cgroup_id(ent);
2660
		rcu_read_lock();
2661
		memcg = mem_cgroup_lookup(id);
2662
		if (memcg && !css_tryget_online(&memcg->css))
2663
			memcg = NULL;
2664
		rcu_read_unlock();
2665
	}
2666
	return memcg;
2667 2668
}

2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699
static void lock_page_lru(struct page *page, int *isolated)
{
	struct zone *zone = page_zone(page);

	spin_lock_irq(&zone->lru_lock);
	if (PageLRU(page)) {
		struct lruvec *lruvec;

		lruvec = mem_cgroup_page_lruvec(page, zone);
		ClearPageLRU(page);
		del_page_from_lru_list(page, lruvec, page_lru(page));
		*isolated = 1;
	} else
		*isolated = 0;
}

static void unlock_page_lru(struct page *page, int isolated)
{
	struct zone *zone = page_zone(page);

	if (isolated) {
		struct lruvec *lruvec;

		lruvec = mem_cgroup_page_lruvec(page, zone);
		VM_BUG_ON_PAGE(PageLRU(page), page);
		SetPageLRU(page);
		add_page_to_lru_list(page, lruvec, page_lru(page));
	}
	spin_unlock_irq(&zone->lru_lock);
}

2700
static void commit_charge(struct page *page, struct mem_cgroup *memcg,
2701
			  bool lrucare)
2702
{
2703
	struct page_cgroup *pc = lookup_page_cgroup(page);
2704
	int isolated;
2705

2706
	VM_BUG_ON_PAGE(PageCgroupUsed(pc), page);
2707 2708 2709 2710
	/*
	 * we don't need page_cgroup_lock about tail pages, becase they are not
	 * accessed by any other context at this point.
	 */
2711 2712 2713 2714 2715

	/*
	 * 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.
	 */
2716 2717
	if (lrucare)
		lock_page_lru(page, &isolated);
2718

2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732
	/*
	 * Nobody should be changing or seriously looking at
	 * pc->mem_cgroup and pc->flags at this point:
	 *
	 * - the page is uncharged
	 *
	 * - the page is off-LRU
	 *
	 * - an anonymous fault has exclusive page access, except for
	 *   a locked page table
	 *
	 * - a page cache insertion, a swapin fault, or a migration
	 *   have the page locked
	 */
2733
	pc->mem_cgroup = memcg;
2734
	pc->flags = PCG_USED | PCG_MEM | (do_swap_account ? PCG_MEMSW : 0);
2735

2736 2737
	if (lrucare)
		unlock_page_lru(page, isolated);
2738
}
2739

2740
#ifdef CONFIG_MEMCG_KMEM
2741 2742 2743 2744 2745 2746
/*
 * 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);

2747 2748
static DEFINE_MUTEX(activate_kmem_mutex);

G
Glauber Costa 已提交
2749 2750 2751 2752 2753 2754 2755 2756 2757 2758
/*
 * 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;
2759
	return cache_from_memcg_idx(cachep, memcg_cache_id(p->memcg));
G
Glauber Costa 已提交
2760 2761
}

2762
#ifdef CONFIG_SLABINFO
2763
static int mem_cgroup_slabinfo_read(struct seq_file *m, void *v)
2764
{
2765
	struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(m));
2766 2767
	struct memcg_cache_params *params;

2768
	if (!memcg_kmem_is_active(memcg))
2769 2770 2771 2772
		return -EIO;

	print_slabinfo_header(m);

2773
	mutex_lock(&memcg_slab_mutex);
2774 2775
	list_for_each_entry(params, &memcg->memcg_slab_caches, list)
		cache_show(memcg_params_to_cache(params), m);
2776
	mutex_unlock(&memcg_slab_mutex);
2777 2778 2779 2780 2781

	return 0;
}
#endif

2782 2783
static int memcg_charge_kmem(struct mem_cgroup *memcg, gfp_t gfp,
			     unsigned long nr_pages)
2784
{
2785
	struct page_counter *counter;
2786 2787
	int ret = 0;

2788 2789
	ret = page_counter_try_charge(&memcg->kmem, nr_pages, &counter);
	if (ret < 0)
2790 2791
		return ret;

2792
	ret = try_charge(memcg, gfp, nr_pages);
2793 2794
	if (ret == -EINTR)  {
		/*
2795 2796 2797 2798 2799 2800
		 * try_charge() chose to bypass to root due to 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
2801 2802 2803
		 * our minds.
		 *
		 * This condition will only trigger if the task entered
2804 2805 2806
		 * memcg_charge_kmem in a sane state, but was OOM-killed
		 * during try_charge() above. Tasks that were already dying
		 * when the allocation triggers should have been already
2807 2808
		 * directed to the root cgroup in memcontrol.h
		 */
2809
		page_counter_charge(&memcg->memory, nr_pages);
2810
		if (do_swap_account)
2811
			page_counter_charge(&memcg->memsw, nr_pages);
2812 2813
		ret = 0;
	} else if (ret)
2814
		page_counter_uncharge(&memcg->kmem, nr_pages);
2815 2816 2817 2818

	return ret;
}

2819 2820
static void memcg_uncharge_kmem(struct mem_cgroup *memcg,
				unsigned long nr_pages)
2821
{
2822
	page_counter_uncharge(&memcg->memory, nr_pages);
2823
	if (do_swap_account)
2824
		page_counter_uncharge(&memcg->memsw, nr_pages);
2825 2826

	/* Not down to 0 */
2827
	if (page_counter_uncharge(&memcg->kmem, nr_pages))
2828 2829
		return;

2830 2831 2832 2833 2834 2835 2836 2837
	/*
	 * 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().
	 */
2838
	if (memcg_kmem_test_and_clear_dead(memcg))
2839
		css_put(&memcg->css);
2840 2841
}

2842 2843 2844 2845 2846 2847 2848 2849 2850 2851
/*
 * 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;
}

2852
static int memcg_alloc_cache_id(void)
2853
{
2854 2855 2856 2857 2858 2859 2860
	int id, size;
	int err;

	id = ida_simple_get(&kmem_limited_groups,
			    0, MEMCG_CACHES_MAX_SIZE, GFP_KERNEL);
	if (id < 0)
		return id;
2861

2862 2863 2864 2865 2866 2867 2868 2869 2870
	if (id < memcg_limited_groups_array_size)
		return id;

	/*
	 * There's no space for the new id in memcg_caches arrays,
	 * so we have to grow them.
	 */

	size = 2 * (id + 1);
2871 2872 2873 2874 2875
	if (size < MEMCG_CACHES_MIN_SIZE)
		size = MEMCG_CACHES_MIN_SIZE;
	else if (size > MEMCG_CACHES_MAX_SIZE)
		size = MEMCG_CACHES_MAX_SIZE;

2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889
	mutex_lock(&memcg_slab_mutex);
	err = memcg_update_all_caches(size);
	mutex_unlock(&memcg_slab_mutex);

	if (err) {
		ida_simple_remove(&kmem_limited_groups, id);
		return err;
	}
	return id;
}

static void memcg_free_cache_id(int id)
{
	ida_simple_remove(&kmem_limited_groups, id);
2890 2891 2892 2893 2894 2895 2896 2897 2898
}

/*
 * 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)
{
2899
	memcg_limited_groups_array_size = num;
2900 2901
}

2902 2903
static void memcg_register_cache(struct mem_cgroup *memcg,
				 struct kmem_cache *root_cache)
2904
{
2905 2906
	static char memcg_name_buf[NAME_MAX + 1]; /* protected by
						     memcg_slab_mutex */
2907
	struct kmem_cache *cachep;
2908 2909
	int id;

2910 2911 2912 2913 2914 2915 2916 2917 2918 2919
	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))
2920 2921
		return;

2922
	cgroup_name(memcg->css.cgroup, memcg_name_buf, NAME_MAX + 1);
2923
	cachep = memcg_create_kmem_cache(memcg, root_cache, memcg_name_buf);
2924
	/*
2925 2926 2927
	 * 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.
2928
	 */
2929 2930
	if (!cachep)
		return;
2931

2932
	css_get(&memcg->css);
2933
	list_add(&cachep->memcg_params->list, &memcg->memcg_slab_caches);
2934

2935
	/*
2936 2937 2938
	 * 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.
2939
	 */
2940 2941
	smp_wmb();

2942 2943
	BUG_ON(root_cache->memcg_params->memcg_caches[id]);
	root_cache->memcg_params->memcg_caches[id] = cachep;
2944
}
2945

2946
static void memcg_unregister_cache(struct kmem_cache *cachep)
2947
{
2948
	struct kmem_cache *root_cache;
2949 2950 2951
	struct mem_cgroup *memcg;
	int id;

2952
	lockdep_assert_held(&memcg_slab_mutex);
2953

2954
	BUG_ON(is_root_cache(cachep));
2955

2956 2957
	root_cache = cachep->memcg_params->root_cache;
	memcg = cachep->memcg_params->memcg;
2958
	id = memcg_cache_id(memcg);
2959

2960 2961
	BUG_ON(root_cache->memcg_params->memcg_caches[id] != cachep);
	root_cache->memcg_params->memcg_caches[id] = NULL;
2962

2963 2964 2965
	list_del(&cachep->memcg_params->list);

	kmem_cache_destroy(cachep);
2966 2967 2968

	/* drop the reference taken in memcg_register_cache */
	css_put(&memcg->css);
2969 2970
}

2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001
/*
 * 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--;
}

3002
int __memcg_cleanup_cache_params(struct kmem_cache *s)
3003 3004
{
	struct kmem_cache *c;
3005
	int i, failed = 0;
3006

3007
	mutex_lock(&memcg_slab_mutex);
3008 3009
	for_each_memcg_cache_index(i) {
		c = cache_from_memcg_idx(s, i);
3010 3011 3012
		if (!c)
			continue;

3013
		memcg_unregister_cache(c);
3014 3015 3016

		if (cache_from_memcg_idx(s, i))
			failed++;
3017
	}
3018
	mutex_unlock(&memcg_slab_mutex);
3019
	return failed;
3020 3021
}

3022
static void memcg_unregister_all_caches(struct mem_cgroup *memcg)
G
Glauber Costa 已提交
3023 3024
{
	struct kmem_cache *cachep;
3025
	struct memcg_cache_params *params, *tmp;
G
Glauber Costa 已提交
3026 3027 3028 3029

	if (!memcg_kmem_is_active(memcg))
		return;

3030 3031
	mutex_lock(&memcg_slab_mutex);
	list_for_each_entry_safe(params, tmp, &memcg->memcg_slab_caches, list) {
G
Glauber Costa 已提交
3032
		cachep = memcg_params_to_cache(params);
3033 3034
		kmem_cache_shrink(cachep);
		if (atomic_read(&cachep->memcg_params->nr_pages) == 0)
3035
			memcg_unregister_cache(cachep);
G
Glauber Costa 已提交
3036
	}
3037
	mutex_unlock(&memcg_slab_mutex);
G
Glauber Costa 已提交
3038 3039
}

3040
struct memcg_register_cache_work {
3041 3042 3043 3044 3045
	struct mem_cgroup *memcg;
	struct kmem_cache *cachep;
	struct work_struct work;
};

3046
static void memcg_register_cache_func(struct work_struct *w)
3047
{
3048 3049
	struct memcg_register_cache_work *cw =
		container_of(w, struct memcg_register_cache_work, work);
3050 3051
	struct mem_cgroup *memcg = cw->memcg;
	struct kmem_cache *cachep = cw->cachep;
3052

3053
	mutex_lock(&memcg_slab_mutex);
3054
	memcg_register_cache(memcg, cachep);
3055 3056
	mutex_unlock(&memcg_slab_mutex);

3057
	css_put(&memcg->css);
3058 3059 3060 3061 3062 3063
	kfree(cw);
}

/*
 * Enqueue the creation of a per-memcg kmem_cache.
 */
3064 3065
static void __memcg_schedule_register_cache(struct mem_cgroup *memcg,
					    struct kmem_cache *cachep)
3066
{
3067
	struct memcg_register_cache_work *cw;
3068

3069
	cw = kmalloc(sizeof(*cw), GFP_NOWAIT);
3070 3071
	if (cw == NULL) {
		css_put(&memcg->css);
3072 3073 3074 3075 3076 3077
		return;
	}

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

3078
	INIT_WORK(&cw->work, memcg_register_cache_func);
3079 3080 3081
	schedule_work(&cw->work);
}

3082 3083
static void memcg_schedule_register_cache(struct mem_cgroup *memcg,
					  struct kmem_cache *cachep)
3084 3085 3086 3087
{
	/*
	 * We need to stop accounting when we kmalloc, because if the
	 * corresponding kmalloc cache is not yet created, the first allocation
3088
	 * in __memcg_schedule_register_cache will recurse.
3089 3090 3091 3092 3093 3094 3095 3096
	 *
	 * 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();
3097
	__memcg_schedule_register_cache(memcg, cachep);
3098 3099
	memcg_resume_kmem_account();
}
3100 3101 3102

int __memcg_charge_slab(struct kmem_cache *cachep, gfp_t gfp, int order)
{
3103
	unsigned int nr_pages = 1 << order;
3104 3105
	int res;

3106
	res = memcg_charge_kmem(cachep->memcg_params->memcg, gfp, nr_pages);
3107
	if (!res)
3108
		atomic_add(nr_pages, &cachep->memcg_params->nr_pages);
3109 3110 3111 3112 3113
	return res;
}

void __memcg_uncharge_slab(struct kmem_cache *cachep, int order)
{
3114 3115 3116 3117
	unsigned int nr_pages = 1 << order;

	memcg_uncharge_kmem(cachep->memcg_params->memcg, nr_pages);
	atomic_sub(nr_pages, &cachep->memcg_params->nr_pages);
3118 3119
}

3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132 3133 3134 3135 3136
/*
 * 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;
3137
	struct kmem_cache *memcg_cachep;
3138 3139 3140 3141

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

3142 3143 3144
	if (!current->mm || current->memcg_kmem_skip_account)
		return cachep;

3145 3146 3147
	rcu_read_lock();
	memcg = mem_cgroup_from_task(rcu_dereference(current->mm->owner));

3148
	if (!memcg_kmem_is_active(memcg))
3149
		goto out;
3150

3151 3152 3153
	memcg_cachep = cache_from_memcg_idx(cachep, memcg_cache_id(memcg));
	if (likely(memcg_cachep)) {
		cachep = memcg_cachep;
3154
		goto out;
3155 3156
	}

3157
	/* The corresponding put will be done in the workqueue. */
3158
	if (!css_tryget_online(&memcg->css))
3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169
		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
3170 3171 3172
	 * memcg_create_kmem_cache, this means no further allocation
	 * could happen with the slab_mutex held. So it's better to
	 * defer everything.
3173
	 */
3174
	memcg_schedule_register_cache(memcg, cachep);
3175 3176 3177 3178
	return cachep;
out:
	rcu_read_unlock();
	return cachep;
3179 3180
}

3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201
/*
 * 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;
3202 3203 3204 3205

	/*
	 * Disabling accounting is only relevant for some specific memcg
	 * internal allocations. Therefore we would initially not have such
V
Vladimir Davydov 已提交
3206 3207 3208 3209 3210 3211
	 * 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.
3212 3213 3214 3215 3216 3217
	 *
	 * 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
3218 3219 3220
	 *	memcg_stop_kmem_account();
	 *	kmalloc(<large_number>)
	 *	memcg_resume_kmem_account();
3221 3222 3223 3224 3225 3226 3227 3228 3229 3230
	 *
	 * 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;

3231
	memcg = get_mem_cgroup_from_mm(current->mm);
3232

3233
	if (!memcg_kmem_is_active(memcg)) {
3234 3235 3236 3237
		css_put(&memcg->css);
		return true;
	}

3238
	ret = memcg_charge_kmem(memcg, gfp, 1 << order);
3239 3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254
	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) {
3255
		memcg_uncharge_kmem(memcg, 1 << order);
3256 3257
		return;
	}
3258 3259 3260 3261
	/*
	 * The page is freshly allocated and not visible to any
	 * outside callers yet.  Set up pc non-atomically.
	 */
3262 3263
	pc = lookup_page_cgroup(page);
	pc->mem_cgroup = memcg;
3264
	pc->flags = PCG_USED;
3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276
}

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


	pc = lookup_page_cgroup(page);
	if (!PageCgroupUsed(pc))
		return;

3277 3278
	memcg = pc->mem_cgroup;
	pc->flags = 0;
3279 3280 3281 3282 3283 3284 3285 3286

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

3287
	VM_BUG_ON_PAGE(mem_cgroup_is_root(memcg), page);
3288
	memcg_uncharge_kmem(memcg, 1 << order);
3289
}
G
Glauber Costa 已提交
3290
#else
3291
static inline void memcg_unregister_all_caches(struct mem_cgroup *memcg)
G
Glauber Costa 已提交
3292 3293
{
}
3294 3295
#endif /* CONFIG_MEMCG_KMEM */

3296 3297 3298 3299
#ifdef CONFIG_TRANSPARENT_HUGEPAGE

/*
 * Because tail pages are not marked as "used", set it. We're under
3300 3301 3302
 * 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.
3303
 */
3304
void mem_cgroup_split_huge_fixup(struct page *head)
3305 3306
{
	struct page_cgroup *head_pc = lookup_page_cgroup(head);
3307
	struct page_cgroup *pc;
3308
	struct mem_cgroup *memcg;
3309
	int i;
3310

3311 3312
	if (mem_cgroup_disabled())
		return;
3313 3314

	memcg = head_pc->mem_cgroup;
3315 3316
	for (i = 1; i < HPAGE_PMD_NR; i++) {
		pc = head_pc + i;
3317
		pc->mem_cgroup = memcg;
3318
		pc->flags = head_pc->flags;
3319
	}
3320 3321
	__this_cpu_sub(memcg->stat->count[MEM_CGROUP_STAT_RSS_HUGE],
		       HPAGE_PMD_NR);
3322
}
3323
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
3324

3325
/**
3326
 * mem_cgroup_move_account - move account of the page
3327
 * @page: the page
3328
 * @nr_pages: number of regular pages (>1 for huge pages)
3329 3330 3331 3332 3333
 * @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
3334
 * - page is not on LRU (isolate_page() is useful.)
3335
 * - compound_lock is held when nr_pages > 1
3336
 *
3337 3338
 * This function doesn't do "charge" to new cgroup and doesn't do "uncharge"
 * from old cgroup.
3339
 */
3340 3341 3342 3343
static int mem_cgroup_move_account(struct page *page,
				   unsigned int nr_pages,
				   struct page_cgroup *pc,
				   struct mem_cgroup *from,
3344
				   struct mem_cgroup *to)
3345
{
3346 3347
	unsigned long flags;
	int ret;
3348

3349
	VM_BUG_ON(from == to);
3350
	VM_BUG_ON_PAGE(PageLRU(page), page);
3351 3352 3353 3354 3355 3356 3357
	/*
	 * 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;
3358
	if (nr_pages > 1 && !PageTransHuge(page))
3359 3360
		goto out;

3361 3362 3363 3364 3365 3366 3367
	/*
	 * Prevent mem_cgroup_migrate() from looking at pc->mem_cgroup
	 * of its source page while we change it: page migration takes
	 * both pages off the LRU, but page cache replacement doesn't.
	 */
	if (!trylock_page(page))
		goto out;
3368 3369 3370

	ret = -EINVAL;
	if (!PageCgroupUsed(pc) || pc->mem_cgroup != from)
3371
		goto out_unlock;
3372

3373
	move_lock_mem_cgroup(from, &flags);
3374

3375
	if (!PageAnon(page) && page_mapped(page)) {
3376 3377 3378 3379 3380
		__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);
	}
3381

3382 3383 3384 3385 3386 3387
	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);
	}
3388

3389 3390 3391 3392 3393
	/*
	 * It is safe to change pc->mem_cgroup here because the page
	 * is referenced, charged, and isolated - we can't race with
	 * uncharging, charging, migration, or LRU putback.
	 */
3394

3395
	/* caller should have done css_get */
K
KAMEZAWA Hiroyuki 已提交
3396
	pc->mem_cgroup = to;
3397
	move_unlock_mem_cgroup(from, &flags);
3398
	ret = 0;
3399 3400 3401

	local_irq_disable();
	mem_cgroup_charge_statistics(to, page, nr_pages);
3402
	memcg_check_events(to, page);
3403
	mem_cgroup_charge_statistics(from, page, -nr_pages);
3404
	memcg_check_events(from, page);
3405 3406 3407
	local_irq_enable();
out_unlock:
	unlock_page(page);
3408
out:
3409 3410 3411
	return ret;
}

3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431
/**
 * 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.
3432
 */
3433 3434
static int mem_cgroup_move_parent(struct page *page,
				  struct page_cgroup *pc,
3435
				  struct mem_cgroup *child)
3436 3437
{
	struct mem_cgroup *parent;
3438
	unsigned int nr_pages;
3439
	unsigned long uninitialized_var(flags);
3440 3441
	int ret;

3442
	VM_BUG_ON(mem_cgroup_is_root(child));
3443

3444 3445 3446 3447 3448
	ret = -EBUSY;
	if (!get_page_unless_zero(page))
		goto out;
	if (isolate_lru_page(page))
		goto put;
3449

3450
	nr_pages = hpage_nr_pages(page);
K
KAMEZAWA Hiroyuki 已提交
3451

3452 3453 3454 3455 3456 3457
	parent = parent_mem_cgroup(child);
	/*
	 * If no parent, move charges to root cgroup.
	 */
	if (!parent)
		parent = root_mem_cgroup;
3458

3459
	if (nr_pages > 1) {
3460
		VM_BUG_ON_PAGE(!PageTransHuge(page), page);
3461
		flags = compound_lock_irqsave(page);
3462
	}
3463

3464
	ret = mem_cgroup_move_account(page, nr_pages,
3465
				pc, child, parent);
3466 3467 3468 3469 3470 3471
	if (!ret) {
		/* Take charge off the local counters */
		page_counter_cancel(&child->memory, nr_pages);
		if (do_swap_account)
			page_counter_cancel(&child->memsw, nr_pages);
	}
3472

3473
	if (nr_pages > 1)
3474
		compound_unlock_irqrestore(page, flags);
K
KAMEZAWA Hiroyuki 已提交
3475
	putback_lru_page(page);
3476
put:
3477
	put_page(page);
3478
out:
3479 3480 3481
	return ret;
}

A
Andrew Morton 已提交
3482
#ifdef CONFIG_MEMCG_SWAP
3483 3484
static void mem_cgroup_swap_statistics(struct mem_cgroup *memcg,
					 bool charge)
K
KAMEZAWA Hiroyuki 已提交
3485
{
3486 3487
	int val = (charge) ? 1 : -1;
	this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_SWAP], val);
K
KAMEZAWA Hiroyuki 已提交
3488
}
3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500

/**
 * 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.
 *
3501
 * The caller must have charged to @to, IOW, called page_counter_charge() about
3502 3503 3504
 * both res and memsw, and called css_get().
 */
static int mem_cgroup_move_swap_account(swp_entry_t entry,
3505
				struct mem_cgroup *from, struct mem_cgroup *to)
3506 3507 3508
{
	unsigned short old_id, new_id;

L
Li Zefan 已提交
3509 3510
	old_id = mem_cgroup_id(from);
	new_id = mem_cgroup_id(to);
3511 3512 3513

	if (swap_cgroup_cmpxchg(entry, old_id, new_id) == old_id) {
		mem_cgroup_swap_statistics(from, false);
3514
		mem_cgroup_swap_statistics(to, true);
3515
		/*
3516
		 * This function is only called from task migration context now.
3517
		 * It postpones page_counter and refcount handling till the end
3518
		 * of task migration(mem_cgroup_clear_mc()) for performance
L
Li Zefan 已提交
3519 3520 3521 3522 3523 3524
		 * 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().
3525
		 */
L
Li Zefan 已提交
3526
		css_get(&to->css);
3527 3528 3529 3530 3531 3532
		return 0;
	}
	return -EINVAL;
}
#else
static inline int mem_cgroup_move_swap_account(swp_entry_t entry,
3533
				struct mem_cgroup *from, struct mem_cgroup *to)
3534 3535 3536
{
	return -EINVAL;
}
3537
#endif
K
KAMEZAWA Hiroyuki 已提交
3538

3539 3540 3541 3542 3543 3544
#ifdef CONFIG_DEBUG_VM
static struct page_cgroup *lookup_page_cgroup_used(struct page *page)
{
	struct page_cgroup *pc;

	pc = lookup_page_cgroup(page);
3545 3546 3547 3548 3549
	/*
	 * 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().
	 */
3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568
	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) {
3569 3570
		pr_alert("pc:%p pc->flags:%lx pc->mem_cgroup:%p\n",
			 pc, pc->flags, pc->mem_cgroup);
3571 3572 3573 3574
	}
}
#endif

3575 3576
static DEFINE_MUTEX(memcg_limit_mutex);

3577
static int mem_cgroup_resize_limit(struct mem_cgroup *memcg,
3578
				   unsigned long limit)
3579
{
3580 3581 3582
	unsigned long curusage;
	unsigned long oldusage;
	bool enlarge = false;
3583
	int retry_count;
3584
	int ret;
3585 3586 3587 3588 3589 3590

	/*
	 * 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.
	 */
3591 3592
	retry_count = MEM_CGROUP_RECLAIM_RETRIES *
		      mem_cgroup_count_children(memcg);
3593

3594
	oldusage = page_counter_read(&memcg->memory);
3595

3596
	do {
3597 3598 3599 3600
		if (signal_pending(current)) {
			ret = -EINTR;
			break;
		}
3601 3602 3603 3604

		mutex_lock(&memcg_limit_mutex);
		if (limit > memcg->memsw.limit) {
			mutex_unlock(&memcg_limit_mutex);
3605
			ret = -EINVAL;
3606 3607
			break;
		}
3608 3609 3610 3611
		if (limit > memcg->memory.limit)
			enlarge = true;
		ret = page_counter_limit(&memcg->memory, limit);
		mutex_unlock(&memcg_limit_mutex);
3612 3613 3614 3615

		if (!ret)
			break;

3616 3617
		try_to_free_mem_cgroup_pages(memcg, 1, GFP_KERNEL, true);

3618
		curusage = page_counter_read(&memcg->memory);
3619
		/* Usage is reduced ? */
A
Andrew Morton 已提交
3620
		if (curusage >= oldusage)
3621 3622 3623
			retry_count--;
		else
			oldusage = curusage;
3624 3625
	} while (retry_count);

3626 3627
	if (!ret && enlarge)
		memcg_oom_recover(memcg);
3628

3629 3630 3631
	return ret;
}

L
Li Zefan 已提交
3632
static int mem_cgroup_resize_memsw_limit(struct mem_cgroup *memcg,
3633
					 unsigned long limit)
3634
{
3635 3636 3637
	unsigned long curusage;
	unsigned long oldusage;
	bool enlarge = false;
3638
	int retry_count;
3639
	int ret;
3640

3641
	/* see mem_cgroup_resize_res_limit */
3642 3643 3644 3645 3646 3647
	retry_count = MEM_CGROUP_RECLAIM_RETRIES *
		      mem_cgroup_count_children(memcg);

	oldusage = page_counter_read(&memcg->memsw);

	do {
3648 3649 3650 3651
		if (signal_pending(current)) {
			ret = -EINTR;
			break;
		}
3652 3653 3654 3655

		mutex_lock(&memcg_limit_mutex);
		if (limit < memcg->memory.limit) {
			mutex_unlock(&memcg_limit_mutex);
3656 3657 3658
			ret = -EINVAL;
			break;
		}
3659 3660 3661 3662
		if (limit > memcg->memsw.limit)
			enlarge = true;
		ret = page_counter_limit(&memcg->memsw, limit);
		mutex_unlock(&memcg_limit_mutex);
3663 3664 3665 3666

		if (!ret)
			break;

3667 3668
		try_to_free_mem_cgroup_pages(memcg, 1, GFP_KERNEL, false);

3669
		curusage = page_counter_read(&memcg->memsw);
3670
		/* Usage is reduced ? */
3671
		if (curusage >= oldusage)
3672
			retry_count--;
3673 3674
		else
			oldusage = curusage;
3675 3676
	} while (retry_count);

3677 3678
	if (!ret && enlarge)
		memcg_oom_recover(memcg);
3679

3680 3681 3682
	return ret;
}

3683 3684 3685 3686 3687 3688 3689 3690 3691
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;
3692
	unsigned long excess;
3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716
	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;
3717
		spin_lock_irq(&mctz->lock);
3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744

		/*
		 * 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);
		}
3745
		__mem_cgroup_remove_exceeded(mz, mctz);
3746
		excess = soft_limit_excess(mz->memcg);
3747 3748 3749 3750 3751 3752 3753 3754 3755
		/*
		 * 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 */
3756
		__mem_cgroup_insert_exceeded(mz, mctz, excess);
3757
		spin_unlock_irq(&mctz->lock);
3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774
		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;
}

3775 3776 3777 3778 3779 3780 3781
/**
 * 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
 *
3782
 * Traverse a specified page_cgroup list and try to drop them all.  This doesn't
3783 3784
 * reclaim the pages page themselves - pages are moved to the parent (or root)
 * group.
3785
 */
3786
static void mem_cgroup_force_empty_list(struct mem_cgroup *memcg,
K
KAMEZAWA Hiroyuki 已提交
3787
				int node, int zid, enum lru_list lru)
3788
{
3789
	struct lruvec *lruvec;
3790
	unsigned long flags;
3791
	struct list_head *list;
3792 3793
	struct page *busy;
	struct zone *zone;
3794

K
KAMEZAWA Hiroyuki 已提交
3795
	zone = &NODE_DATA(node)->node_zones[zid];
3796 3797
	lruvec = mem_cgroup_zone_lruvec(zone, memcg);
	list = &lruvec->lists[lru];
3798

3799
	busy = NULL;
3800
	do {
3801
		struct page_cgroup *pc;
3802 3803
		struct page *page;

K
KAMEZAWA Hiroyuki 已提交
3804
		spin_lock_irqsave(&zone->lru_lock, flags);
3805
		if (list_empty(list)) {
K
KAMEZAWA Hiroyuki 已提交
3806
			spin_unlock_irqrestore(&zone->lru_lock, flags);
3807
			break;
3808
		}
3809 3810 3811
		page = list_entry(list->prev, struct page, lru);
		if (busy == page) {
			list_move(&page->lru, list);
3812
			busy = NULL;
K
KAMEZAWA Hiroyuki 已提交
3813
			spin_unlock_irqrestore(&zone->lru_lock, flags);
3814 3815
			continue;
		}
K
KAMEZAWA Hiroyuki 已提交
3816
		spin_unlock_irqrestore(&zone->lru_lock, flags);
3817

3818
		pc = lookup_page_cgroup(page);
3819

3820
		if (mem_cgroup_move_parent(page, pc, memcg)) {
3821
			/* found lock contention or "pc" is obsolete. */
3822
			busy = page;
3823 3824
		} else
			busy = NULL;
3825
		cond_resched();
3826
	} while (!list_empty(list));
3827 3828 3829
}

/*
3830 3831
 * make mem_cgroup's charge to be 0 if there is no task by moving
 * all the charges and pages to the parent.
3832
 * This enables deleting this mem_cgroup.
3833 3834
 *
 * Caller is responsible for holding css reference on the memcg.
3835
 */
3836
static void mem_cgroup_reparent_charges(struct mem_cgroup *memcg)
3837
{
3838
	int node, zid;
3839

3840
	do {
3841 3842
		/* This is for making all *used* pages to be on LRU. */
		lru_add_drain_all();
3843 3844
		drain_all_stock_sync(memcg);
		mem_cgroup_start_move(memcg);
3845
		for_each_node_state(node, N_MEMORY) {
3846
			for (zid = 0; zid < MAX_NR_ZONES; zid++) {
H
Hugh Dickins 已提交
3847 3848
				enum lru_list lru;
				for_each_lru(lru) {
3849
					mem_cgroup_force_empty_list(memcg,
H
Hugh Dickins 已提交
3850
							node, zid, lru);
3851
				}
3852
			}
3853
		}
3854 3855
		mem_cgroup_end_move(memcg);
		memcg_oom_recover(memcg);
3856
		cond_resched();
3857

3858
		/*
3859 3860 3861 3862 3863
		 * 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.
		 *
3864 3865 3866 3867 3868 3869
		 * 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.
		 */
3870 3871
	} while (page_counter_read(&memcg->memory) -
		 page_counter_read(&memcg->kmem) > 0);
3872 3873
}

3874 3875 3876 3877 3878 3879
/*
 * Test whether @memcg has children, dead or alive.  Note that this
 * function doesn't care whether @memcg has use_hierarchy enabled and
 * returns %true if there are child csses according to the cgroup
 * hierarchy.  Testing use_hierarchy is the caller's responsiblity.
 */
3880 3881
static inline bool memcg_has_children(struct mem_cgroup *memcg)
{
3882 3883
	bool ret;

3884
	/*
3885 3886 3887 3888
	 * The lock does not prevent addition or deletion 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.
3889
	 */
3890 3891 3892 3893 3894 3895
	lockdep_assert_held(&memcg_create_mutex);

	rcu_read_lock();
	ret = css_next_child(NULL, &memcg->css);
	rcu_read_unlock();
	return ret;
3896 3897
}

3898 3899 3900 3901 3902 3903 3904 3905 3906 3907
/*
 * 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;

3908 3909
	/* we call try-to-free pages for make this cgroup empty */
	lru_add_drain_all();
3910
	/* try to free all pages in this cgroup */
3911
	while (nr_retries && page_counter_read(&memcg->memory)) {
3912
		int progress;
3913

3914 3915 3916
		if (signal_pending(current))
			return -EINTR;

3917 3918
		progress = try_to_free_mem_cgroup_pages(memcg, 1,
							GFP_KERNEL, true);
3919
		if (!progress) {
3920
			nr_retries--;
3921
			/* maybe some writeback is necessary */
3922
			congestion_wait(BLK_RW_ASYNC, HZ/10);
3923
		}
3924 3925

	}
3926 3927

	return 0;
3928 3929
}

3930 3931 3932
static ssize_t mem_cgroup_force_empty_write(struct kernfs_open_file *of,
					    char *buf, size_t nbytes,
					    loff_t off)
3933
{
3934
	struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of));
3935

3936 3937
	if (mem_cgroup_is_root(memcg))
		return -EINVAL;
3938
	return mem_cgroup_force_empty(memcg) ?: nbytes;
3939 3940
}

3941 3942
static u64 mem_cgroup_hierarchy_read(struct cgroup_subsys_state *css,
				     struct cftype *cft)
3943
{
3944
	return mem_cgroup_from_css(css)->use_hierarchy;
3945 3946
}

3947 3948
static int mem_cgroup_hierarchy_write(struct cgroup_subsys_state *css,
				      struct cftype *cft, u64 val)
3949 3950
{
	int retval = 0;
3951
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
T
Tejun Heo 已提交
3952
	struct mem_cgroup *parent_memcg = mem_cgroup_from_css(memcg->css.parent);
3953

3954
	mutex_lock(&memcg_create_mutex);
3955 3956 3957 3958

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

3959
	/*
3960
	 * If parent's use_hierarchy is set, we can't make any modifications
3961 3962 3963 3964 3965 3966
	 * 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.
	 */
3967
	if ((!parent_memcg || !parent_memcg->use_hierarchy) &&
3968
				(val == 1 || val == 0)) {
3969
		if (!memcg_has_children(memcg))
3970
			memcg->use_hierarchy = val;
3971 3972 3973 3974
		else
			retval = -EBUSY;
	} else
		retval = -EINVAL;
3975 3976

out:
3977
	mutex_unlock(&memcg_create_mutex);
3978 3979 3980 3981

	return retval;
}

3982 3983
static unsigned long tree_stat(struct mem_cgroup *memcg,
			       enum mem_cgroup_stat_index idx)
3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000
{
	struct mem_cgroup *iter;
	long val = 0;

	/* Per-cpu values can be negative, use a signed accumulator */
	for_each_mem_cgroup_tree(iter, memcg)
		val += mem_cgroup_read_stat(iter, idx);

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

static inline u64 mem_cgroup_usage(struct mem_cgroup *memcg, bool swap)
{
	u64 val;

4001 4002 4003 4004 4005 4006
	if (mem_cgroup_is_root(memcg)) {
		val = tree_stat(memcg, MEM_CGROUP_STAT_CACHE);
		val += tree_stat(memcg, MEM_CGROUP_STAT_RSS);
		if (swap)
			val += tree_stat(memcg, MEM_CGROUP_STAT_SWAP);
	} else {
4007
		if (!swap)
4008
			val = page_counter_read(&memcg->memory);
4009
		else
4010
			val = page_counter_read(&memcg->memsw);
4011 4012 4013 4014
	}
	return val << PAGE_SHIFT;
}

4015 4016 4017 4018 4019 4020 4021
enum {
	RES_USAGE,
	RES_LIMIT,
	RES_MAX_USAGE,
	RES_FAILCNT,
	RES_SOFT_LIMIT,
};
4022

4023
static u64 mem_cgroup_read_u64(struct cgroup_subsys_state *css,
4024
			       struct cftype *cft)
B
Balbir Singh 已提交
4025
{
4026
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
4027
	struct page_counter *counter;
4028

4029
	switch (MEMFILE_TYPE(cft->private)) {
4030
	case _MEM:
4031 4032
		counter = &memcg->memory;
		break;
4033
	case _MEMSWAP:
4034 4035
		counter = &memcg->memsw;
		break;
4036
	case _KMEM:
4037
		counter = &memcg->kmem;
4038
		break;
4039 4040 4041
	default:
		BUG();
	}
4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060

	switch (MEMFILE_ATTR(cft->private)) {
	case RES_USAGE:
		if (counter == &memcg->memory)
			return mem_cgroup_usage(memcg, false);
		if (counter == &memcg->memsw)
			return mem_cgroup_usage(memcg, true);
		return (u64)page_counter_read(counter) * PAGE_SIZE;
	case RES_LIMIT:
		return (u64)counter->limit * PAGE_SIZE;
	case RES_MAX_USAGE:
		return (u64)counter->watermark * PAGE_SIZE;
	case RES_FAILCNT:
		return counter->failcnt;
	case RES_SOFT_LIMIT:
		return (u64)memcg->soft_limit * PAGE_SIZE;
	default:
		BUG();
	}
B
Balbir Singh 已提交
4061
}
4062 4063

#ifdef CONFIG_MEMCG_KMEM
4064 4065
/* should be called with activate_kmem_mutex held */
static int __memcg_activate_kmem(struct mem_cgroup *memcg,
4066
				 unsigned long nr_pages)
4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079
{
	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();

4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091
	/*
	 * 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.
	 */
4092
	mutex_lock(&memcg_create_mutex);
4093 4094
	if (cgroup_has_tasks(memcg->css.cgroup) ||
	    (memcg->use_hierarchy && memcg_has_children(memcg)))
4095 4096 4097 4098
		err = -EBUSY;
	mutex_unlock(&memcg_create_mutex);
	if (err)
		goto out;
4099

4100
	memcg_id = memcg_alloc_cache_id();
4101 4102 4103 4104 4105 4106 4107 4108 4109 4110 4111 4112
	if (memcg_id < 0) {
		err = memcg_id;
		goto out;
	}

	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.
	 */
4113
	err = page_counter_limit(&memcg->kmem, nr_pages);
4114 4115 4116 4117 4118 4119 4120 4121 4122
	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);
4123
out:
4124 4125 4126 4127 4128
	memcg_resume_kmem_account();
	return err;
}

static int memcg_activate_kmem(struct mem_cgroup *memcg,
4129
			       unsigned long nr_pages)
4130 4131 4132 4133
{
	int ret;

	mutex_lock(&activate_kmem_mutex);
4134
	ret = __memcg_activate_kmem(memcg, nr_pages);
4135 4136 4137 4138 4139
	mutex_unlock(&activate_kmem_mutex);
	return ret;
}

static int memcg_update_kmem_limit(struct mem_cgroup *memcg,
4140
				   unsigned long limit)
4141 4142 4143
{
	int ret;

4144
	mutex_lock(&memcg_limit_mutex);
4145
	if (!memcg_kmem_is_active(memcg))
4146
		ret = memcg_activate_kmem(memcg, limit);
4147
	else
4148 4149
		ret = page_counter_limit(&memcg->kmem, limit);
	mutex_unlock(&memcg_limit_mutex);
4150 4151 4152
	return ret;
}

4153
static int memcg_propagate_kmem(struct mem_cgroup *memcg)
4154
{
4155
	int ret = 0;
4156
	struct mem_cgroup *parent = parent_mem_cgroup(memcg);
4157

4158 4159
	if (!parent)
		return 0;
4160

4161
	mutex_lock(&activate_kmem_mutex);
4162
	/*
4163 4164
	 * If the parent cgroup is not kmem-active now, it cannot be activated
	 * after this point, because it has at least one child already.
4165
	 */
4166
	if (memcg_kmem_is_active(parent))
4167
		ret = __memcg_activate_kmem(memcg, PAGE_COUNTER_MAX);
4168
	mutex_unlock(&activate_kmem_mutex);
4169
	return ret;
4170
}
4171 4172
#else
static int memcg_update_kmem_limit(struct mem_cgroup *memcg,
4173
				   unsigned long limit)
4174 4175 4176
{
	return -EINVAL;
}
4177
#endif /* CONFIG_MEMCG_KMEM */
4178

4179 4180 4181 4182
/*
 * The user of this function is...
 * RES_LIMIT.
 */
4183 4184
static ssize_t mem_cgroup_write(struct kernfs_open_file *of,
				char *buf, size_t nbytes, loff_t off)
B
Balbir Singh 已提交
4185
{
4186
	struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of));
4187
	unsigned long nr_pages;
4188 4189
	int ret;

4190
	buf = strstrip(buf);
4191 4192 4193
	ret = page_counter_memparse(buf, &nr_pages);
	if (ret)
		return ret;
4194

4195
	switch (MEMFILE_ATTR(of_cft(of)->private)) {
4196
	case RES_LIMIT:
4197 4198 4199 4200
		if (mem_cgroup_is_root(memcg)) { /* Can't set limit on root */
			ret = -EINVAL;
			break;
		}
4201 4202 4203
		switch (MEMFILE_TYPE(of_cft(of)->private)) {
		case _MEM:
			ret = mem_cgroup_resize_limit(memcg, nr_pages);
4204
			break;
4205 4206
		case _MEMSWAP:
			ret = mem_cgroup_resize_memsw_limit(memcg, nr_pages);
4207
			break;
4208 4209 4210 4211
		case _KMEM:
			ret = memcg_update_kmem_limit(memcg, nr_pages);
			break;
		}
4212
		break;
4213 4214 4215
	case RES_SOFT_LIMIT:
		memcg->soft_limit = nr_pages;
		ret = 0;
4216 4217
		break;
	}
4218
	return ret ?: nbytes;
B
Balbir Singh 已提交
4219 4220
}

4221 4222
static ssize_t mem_cgroup_reset(struct kernfs_open_file *of, char *buf,
				size_t nbytes, loff_t off)
4223
{
4224
	struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of));
4225
	struct page_counter *counter;
4226

4227 4228 4229 4230 4231 4232 4233 4234 4235 4236 4237 4238 4239
	switch (MEMFILE_TYPE(of_cft(of)->private)) {
	case _MEM:
		counter = &memcg->memory;
		break;
	case _MEMSWAP:
		counter = &memcg->memsw;
		break;
	case _KMEM:
		counter = &memcg->kmem;
		break;
	default:
		BUG();
	}
4240

4241
	switch (MEMFILE_ATTR(of_cft(of)->private)) {
4242
	case RES_MAX_USAGE:
4243
		page_counter_reset_watermark(counter);
4244 4245
		break;
	case RES_FAILCNT:
4246
		counter->failcnt = 0;
4247
		break;
4248 4249
	default:
		BUG();
4250
	}
4251

4252
	return nbytes;
4253 4254
}

4255
static u64 mem_cgroup_move_charge_read(struct cgroup_subsys_state *css,
4256 4257
					struct cftype *cft)
{
4258
	return mem_cgroup_from_css(css)->move_charge_at_immigrate;
4259 4260
}

4261
#ifdef CONFIG_MMU
4262
static int mem_cgroup_move_charge_write(struct cgroup_subsys_state *css,
4263 4264
					struct cftype *cft, u64 val)
{
4265
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
4266 4267 4268

	if (val >= (1 << NR_MOVE_TYPE))
		return -EINVAL;
4269

4270
	/*
4271 4272 4273 4274
	 * 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.
4275
	 */
4276
	memcg->move_charge_at_immigrate = val;
4277 4278
	return 0;
}
4279
#else
4280
static int mem_cgroup_move_charge_write(struct cgroup_subsys_state *css,
4281 4282 4283 4284 4285
					struct cftype *cft, u64 val)
{
	return -ENOSYS;
}
#endif
4286

4287
#ifdef CONFIG_NUMA
4288
static int memcg_numa_stat_show(struct seq_file *m, void *v)
4289
{
4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301
	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;
4302
	int nid;
4303
	unsigned long nr;
4304
	struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(m));
4305

4306 4307 4308 4309 4310 4311 4312 4313 4314
	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');
4315 4316
	}

4317 4318 4319 4320 4321 4322 4323 4324 4325 4326 4327 4328 4329 4330 4331
	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');
4332 4333 4334 4335 4336 4337
	}

	return 0;
}
#endif /* CONFIG_NUMA */

4338 4339 4340 4341 4342
static inline void mem_cgroup_lru_names_not_uptodate(void)
{
	BUILD_BUG_ON(ARRAY_SIZE(mem_cgroup_lru_names) != NR_LRU_LISTS);
}

4343
static int memcg_stat_show(struct seq_file *m, void *v)
4344
{
4345
	struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(m));
4346
	unsigned long memory, memsw;
4347 4348
	struct mem_cgroup *mi;
	unsigned int i;
4349

4350
	for (i = 0; i < MEM_CGROUP_STAT_NSTATS; i++) {
4351
		if (i == MEM_CGROUP_STAT_SWAP && !do_swap_account)
4352
			continue;
4353 4354
		seq_printf(m, "%s %ld\n", mem_cgroup_stat_names[i],
			   mem_cgroup_read_stat(memcg, i) * PAGE_SIZE);
4355
	}
L
Lee Schermerhorn 已提交
4356

4357 4358 4359 4360 4361 4362 4363 4364
	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
4365
	/* Hierarchical information */
4366 4367 4368 4369
	memory = memsw = PAGE_COUNTER_MAX;
	for (mi = memcg; mi; mi = parent_mem_cgroup(mi)) {
		memory = min(memory, mi->memory.limit);
		memsw = min(memsw, mi->memsw.limit);
4370
	}
4371 4372 4373 4374 4375
	seq_printf(m, "hierarchical_memory_limit %llu\n",
		   (u64)memory * PAGE_SIZE);
	if (do_swap_account)
		seq_printf(m, "hierarchical_memsw_limit %llu\n",
			   (u64)memsw * PAGE_SIZE);
K
KOSAKI Motohiro 已提交
4376

4377 4378 4379
	for (i = 0; i < MEM_CGROUP_STAT_NSTATS; i++) {
		long long val = 0;

4380
		if (i == MEM_CGROUP_STAT_SWAP && !do_swap_account)
4381
			continue;
4382 4383 4384 4385 4386 4387 4388 4389 4390 4391 4392 4393 4394 4395 4396 4397 4398 4399 4400 4401
		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);
4402
	}
K
KAMEZAWA Hiroyuki 已提交
4403

K
KOSAKI Motohiro 已提交
4404 4405 4406 4407
#ifdef CONFIG_DEBUG_VM
	{
		int nid, zid;
		struct mem_cgroup_per_zone *mz;
4408
		struct zone_reclaim_stat *rstat;
K
KOSAKI Motohiro 已提交
4409 4410 4411 4412 4413
		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++) {
4414
				mz = &memcg->nodeinfo[nid]->zoneinfo[zid];
4415
				rstat = &mz->lruvec.reclaim_stat;
K
KOSAKI Motohiro 已提交
4416

4417 4418 4419 4420
				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
4421
			}
4422 4423 4424 4425
		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
4426 4427 4428
	}
#endif

4429 4430 4431
	return 0;
}

4432 4433
static u64 mem_cgroup_swappiness_read(struct cgroup_subsys_state *css,
				      struct cftype *cft)
K
KOSAKI Motohiro 已提交
4434
{
4435
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
K
KOSAKI Motohiro 已提交
4436

4437
	return mem_cgroup_swappiness(memcg);
K
KOSAKI Motohiro 已提交
4438 4439
}

4440 4441
static int mem_cgroup_swappiness_write(struct cgroup_subsys_state *css,
				       struct cftype *cft, u64 val)
K
KOSAKI Motohiro 已提交
4442
{
4443
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
K
KOSAKI Motohiro 已提交
4444

4445
	if (val > 100)
K
KOSAKI Motohiro 已提交
4446 4447
		return -EINVAL;

4448
	if (css->parent)
4449 4450 4451
		memcg->swappiness = val;
	else
		vm_swappiness = val;
4452

K
KOSAKI Motohiro 已提交
4453 4454 4455
	return 0;
}

4456 4457 4458
static void __mem_cgroup_threshold(struct mem_cgroup *memcg, bool swap)
{
	struct mem_cgroup_threshold_ary *t;
4459
	unsigned long usage;
4460 4461 4462 4463
	int i;

	rcu_read_lock();
	if (!swap)
4464
		t = rcu_dereference(memcg->thresholds.primary);
4465
	else
4466
		t = rcu_dereference(memcg->memsw_thresholds.primary);
4467 4468 4469 4470

	if (!t)
		goto unlock;

4471
	usage = mem_cgroup_usage(memcg, swap);
4472 4473

	/*
4474
	 * current_threshold points to threshold just below or equal to usage.
4475 4476 4477
	 * If it's not true, a threshold was crossed after last
	 * call of __mem_cgroup_threshold().
	 */
4478
	i = t->current_threshold;
4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495 4496 4497 4498 4499 4500 4501

	/*
	 * 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 */
4502
	t->current_threshold = i - 1;
4503 4504 4505 4506 4507 4508
unlock:
	rcu_read_unlock();
}

static void mem_cgroup_threshold(struct mem_cgroup *memcg)
{
4509 4510 4511 4512 4513 4514 4515
	while (memcg) {
		__mem_cgroup_threshold(memcg, false);
		if (do_swap_account)
			__mem_cgroup_threshold(memcg, true);

		memcg = parent_mem_cgroup(memcg);
	}
4516 4517 4518 4519 4520 4521 4522
}

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

4523 4524 4525 4526 4527 4528 4529
	if (_a->threshold > _b->threshold)
		return 1;

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

	return 0;
4530 4531
}

4532
static int mem_cgroup_oom_notify_cb(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
4533 4534 4535
{
	struct mem_cgroup_eventfd_list *ev;

4536 4537
	spin_lock(&memcg_oom_lock);

4538
	list_for_each_entry(ev, &memcg->oom_notify, list)
K
KAMEZAWA Hiroyuki 已提交
4539
		eventfd_signal(ev->eventfd, 1);
4540 4541

	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4542 4543 4544
	return 0;
}

4545
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
4546
{
K
KAMEZAWA Hiroyuki 已提交
4547 4548
	struct mem_cgroup *iter;

4549
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
4550
		mem_cgroup_oom_notify_cb(iter);
K
KAMEZAWA Hiroyuki 已提交
4551 4552
}

4553
static int __mem_cgroup_usage_register_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
4554
	struct eventfd_ctx *eventfd, const char *args, enum res_type type)
4555
{
4556 4557
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
4558 4559
	unsigned long threshold;
	unsigned long usage;
4560
	int i, size, ret;
4561

4562
	ret = page_counter_memparse(args, &threshold);
4563 4564 4565 4566
	if (ret)
		return ret;

	mutex_lock(&memcg->thresholds_lock);
4567

4568
	if (type == _MEM) {
4569
		thresholds = &memcg->thresholds;
4570
		usage = mem_cgroup_usage(memcg, false);
4571
	} else if (type == _MEMSWAP) {
4572
		thresholds = &memcg->memsw_thresholds;
4573
		usage = mem_cgroup_usage(memcg, true);
4574
	} else
4575 4576 4577
		BUG();

	/* Check if a threshold crossed before adding a new one */
4578
	if (thresholds->primary)
4579 4580
		__mem_cgroup_threshold(memcg, type == _MEMSWAP);

4581
	size = thresholds->primary ? thresholds->primary->size + 1 : 1;
4582 4583

	/* Allocate memory for new array of thresholds */
4584
	new = kmalloc(sizeof(*new) + size * sizeof(struct mem_cgroup_threshold),
4585
			GFP_KERNEL);
4586
	if (!new) {
4587 4588 4589
		ret = -ENOMEM;
		goto unlock;
	}
4590
	new->size = size;
4591 4592

	/* Copy thresholds (if any) to new array */
4593 4594
	if (thresholds->primary) {
		memcpy(new->entries, thresholds->primary->entries, (size - 1) *
4595
				sizeof(struct mem_cgroup_threshold));
4596 4597
	}

4598
	/* Add new threshold */
4599 4600
	new->entries[size - 1].eventfd = eventfd;
	new->entries[size - 1].threshold = threshold;
4601 4602

	/* Sort thresholds. Registering of new threshold isn't time-critical */
4603
	sort(new->entries, size, sizeof(struct mem_cgroup_threshold),
4604 4605 4606
			compare_thresholds, NULL);

	/* Find current threshold */
4607
	new->current_threshold = -1;
4608
	for (i = 0; i < size; i++) {
4609
		if (new->entries[i].threshold <= usage) {
4610
			/*
4611 4612
			 * new->current_threshold will not be used until
			 * rcu_assign_pointer(), so it's safe to increment
4613 4614
			 * it here.
			 */
4615
			++new->current_threshold;
4616 4617
		} else
			break;
4618 4619
	}

4620 4621 4622 4623 4624
	/* Free old spare buffer and save old primary buffer as spare */
	kfree(thresholds->spare);
	thresholds->spare = thresholds->primary;

	rcu_assign_pointer(thresholds->primary, new);
4625

4626
	/* To be sure that nobody uses thresholds */
4627 4628 4629 4630 4631 4632 4633 4634
	synchronize_rcu();

unlock:
	mutex_unlock(&memcg->thresholds_lock);

	return ret;
}

4635
static int mem_cgroup_usage_register_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
4636 4637
	struct eventfd_ctx *eventfd, const char *args)
{
4638
	return __mem_cgroup_usage_register_event(memcg, eventfd, args, _MEM);
T
Tejun Heo 已提交
4639 4640
}

4641
static int memsw_cgroup_usage_register_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
4642 4643
	struct eventfd_ctx *eventfd, const char *args)
{
4644
	return __mem_cgroup_usage_register_event(memcg, eventfd, args, _MEMSWAP);
T
Tejun Heo 已提交
4645 4646
}

4647
static void __mem_cgroup_usage_unregister_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
4648
	struct eventfd_ctx *eventfd, enum res_type type)
4649
{
4650 4651
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
4652
	unsigned long usage;
4653
	int i, j, size;
4654 4655

	mutex_lock(&memcg->thresholds_lock);
4656 4657

	if (type == _MEM) {
4658
		thresholds = &memcg->thresholds;
4659
		usage = mem_cgroup_usage(memcg, false);
4660
	} else if (type == _MEMSWAP) {
4661
		thresholds = &memcg->memsw_thresholds;
4662
		usage = mem_cgroup_usage(memcg, true);
4663
	} else
4664 4665
		BUG();

4666 4667 4668
	if (!thresholds->primary)
		goto unlock;

4669 4670 4671 4672
	/* Check if a threshold crossed before removing */
	__mem_cgroup_threshold(memcg, type == _MEMSWAP);

	/* Calculate new number of threshold */
4673 4674 4675
	size = 0;
	for (i = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd != eventfd)
4676 4677 4678
			size++;
	}

4679
	new = thresholds->spare;
4680

4681 4682
	/* Set thresholds array to NULL if we don't have thresholds */
	if (!size) {
4683 4684
		kfree(new);
		new = NULL;
4685
		goto swap_buffers;
4686 4687
	}

4688
	new->size = size;
4689 4690

	/* Copy thresholds and find current threshold */
4691 4692 4693
	new->current_threshold = -1;
	for (i = 0, j = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd == eventfd)
4694 4695
			continue;

4696
		new->entries[j] = thresholds->primary->entries[i];
4697
		if (new->entries[j].threshold <= usage) {
4698
			/*
4699
			 * new->current_threshold will not be used
4700 4701 4702
			 * until rcu_assign_pointer(), so it's safe to increment
			 * it here.
			 */
4703
			++new->current_threshold;
4704 4705 4706 4707
		}
		j++;
	}

4708
swap_buffers:
4709 4710
	/* Swap primary and spare array */
	thresholds->spare = thresholds->primary;
4711 4712 4713 4714 4715 4716
	/* If all events are unregistered, free the spare array */
	if (!new) {
		kfree(thresholds->spare);
		thresholds->spare = NULL;
	}

4717
	rcu_assign_pointer(thresholds->primary, new);
4718

4719
	/* To be sure that nobody uses thresholds */
4720
	synchronize_rcu();
4721
unlock:
4722 4723
	mutex_unlock(&memcg->thresholds_lock);
}
4724

4725
static void mem_cgroup_usage_unregister_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
4726 4727
	struct eventfd_ctx *eventfd)
{
4728
	return __mem_cgroup_usage_unregister_event(memcg, eventfd, _MEM);
T
Tejun Heo 已提交
4729 4730
}

4731
static void memsw_cgroup_usage_unregister_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
4732 4733
	struct eventfd_ctx *eventfd)
{
4734
	return __mem_cgroup_usage_unregister_event(memcg, eventfd, _MEMSWAP);
T
Tejun Heo 已提交
4735 4736
}

4737
static int mem_cgroup_oom_register_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
4738
	struct eventfd_ctx *eventfd, const char *args)
K
KAMEZAWA Hiroyuki 已提交
4739 4740 4741 4742 4743 4744 4745
{
	struct mem_cgroup_eventfd_list *event;

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

4746
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4747 4748 4749 4750 4751

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

	/* already in OOM ? */
4752
	if (atomic_read(&memcg->under_oom))
K
KAMEZAWA Hiroyuki 已提交
4753
		eventfd_signal(eventfd, 1);
4754
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4755 4756 4757 4758

	return 0;
}

4759
static void mem_cgroup_oom_unregister_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
4760
	struct eventfd_ctx *eventfd)
K
KAMEZAWA Hiroyuki 已提交
4761 4762 4763
{
	struct mem_cgroup_eventfd_list *ev, *tmp;

4764
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4765

4766
	list_for_each_entry_safe(ev, tmp, &memcg->oom_notify, list) {
K
KAMEZAWA Hiroyuki 已提交
4767 4768 4769 4770 4771 4772
		if (ev->eventfd == eventfd) {
			list_del(&ev->list);
			kfree(ev);
		}
	}

4773
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4774 4775
}

4776
static int mem_cgroup_oom_control_read(struct seq_file *sf, void *v)
4777
{
4778
	struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(sf));
4779

4780 4781
	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));
4782 4783 4784
	return 0;
}

4785
static int mem_cgroup_oom_control_write(struct cgroup_subsys_state *css,
4786 4787
	struct cftype *cft, u64 val)
{
4788
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
4789 4790

	/* cannot set to root cgroup and only 0 and 1 are allowed */
4791
	if (!css->parent || !((val == 0) || (val == 1)))
4792 4793
		return -EINVAL;

4794
	memcg->oom_kill_disable = val;
4795
	if (!val)
4796
		memcg_oom_recover(memcg);
4797

4798 4799 4800
	return 0;
}

A
Andrew Morton 已提交
4801
#ifdef CONFIG_MEMCG_KMEM
4802
static int memcg_init_kmem(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
4803
{
4804 4805
	int ret;

4806
	memcg->kmemcg_id = -1;
4807 4808 4809
	ret = memcg_propagate_kmem(memcg);
	if (ret)
		return ret;
4810

4811
	return mem_cgroup_sockets_init(memcg, ss);
4812
}
4813

4814
static void memcg_destroy_kmem(struct mem_cgroup *memcg)
G
Glauber Costa 已提交
4815
{
4816
	mem_cgroup_sockets_destroy(memcg);
4817 4818 4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836
}

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
4837 4838 4839 4840
	 * css_offline() when the referencemight have dropped down to 0 and
	 * shouldn't be incremented anymore (css_tryget_online() would
	 * fail) we do not have other options because of the kmem
	 * allocations lifetime.
4841 4842
	 */
	css_get(&memcg->css);
4843 4844 4845

	memcg_kmem_mark_dead(memcg);

4846
	if (page_counter_read(&memcg->kmem))
4847 4848 4849
		return;

	if (memcg_kmem_test_and_clear_dead(memcg))
4850
		css_put(&memcg->css);
G
Glauber Costa 已提交
4851
}
4852
#else
4853
static int memcg_init_kmem(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
4854 4855 4856
{
	return 0;
}
G
Glauber Costa 已提交
4857

4858 4859 4860 4861 4862
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
4863 4864
{
}
4865 4866
#endif

4867 4868 4869 4870 4871 4872 4873 4874 4875 4876 4877 4878 4879
/*
 * 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.
 */

4880 4881 4882 4883 4884
/*
 * Unregister event and free resources.
 *
 * Gets called from workqueue.
 */
4885
static void memcg_event_remove(struct work_struct *work)
4886
{
4887 4888
	struct mem_cgroup_event *event =
		container_of(work, struct mem_cgroup_event, remove);
4889
	struct mem_cgroup *memcg = event->memcg;
4890 4891 4892

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

4893
	event->unregister_event(memcg, event->eventfd);
4894 4895 4896 4897 4898 4899

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

	eventfd_ctx_put(event->eventfd);
	kfree(event);
4900
	css_put(&memcg->css);
4901 4902 4903 4904 4905 4906 4907
}

/*
 * Gets called on POLLHUP on eventfd when user closes it.
 *
 * Called with wqh->lock held and interrupts disabled.
 */
4908 4909
static int memcg_event_wake(wait_queue_t *wait, unsigned mode,
			    int sync, void *key)
4910
{
4911 4912
	struct mem_cgroup_event *event =
		container_of(wait, struct mem_cgroup_event, wait);
4913
	struct mem_cgroup *memcg = event->memcg;
4914 4915 4916 4917 4918 4919 4920 4921 4922 4923 4924 4925
	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.
		 */
4926
		spin_lock(&memcg->event_list_lock);
4927 4928 4929 4930 4931 4932 4933 4934
		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);
		}
4935
		spin_unlock(&memcg->event_list_lock);
4936 4937 4938 4939 4940
	}

	return 0;
}

4941
static void memcg_event_ptable_queue_proc(struct file *file,
4942 4943
		wait_queue_head_t *wqh, poll_table *pt)
{
4944 4945
	struct mem_cgroup_event *event =
		container_of(pt, struct mem_cgroup_event, pt);
4946 4947 4948 4949 4950 4951

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

/*
4952 4953
 * DO NOT USE IN NEW FILES.
 *
4954 4955 4956 4957 4958
 * 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.
 */
4959 4960
static ssize_t memcg_write_event_control(struct kernfs_open_file *of,
					 char *buf, size_t nbytes, loff_t off)
4961
{
4962
	struct cgroup_subsys_state *css = of_css(of);
4963
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
4964
	struct mem_cgroup_event *event;
4965 4966 4967 4968
	struct cgroup_subsys_state *cfile_css;
	unsigned int efd, cfd;
	struct fd efile;
	struct fd cfile;
4969
	const char *name;
4970 4971 4972
	char *endp;
	int ret;

4973 4974 4975
	buf = strstrip(buf);

	efd = simple_strtoul(buf, &endp, 10);
4976 4977
	if (*endp != ' ')
		return -EINVAL;
4978
	buf = endp + 1;
4979

4980
	cfd = simple_strtoul(buf, &endp, 10);
4981 4982
	if ((*endp != ' ') && (*endp != '\0'))
		return -EINVAL;
4983
	buf = endp + 1;
4984 4985 4986 4987 4988

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

4989
	event->memcg = memcg;
4990
	INIT_LIST_HEAD(&event->list);
4991 4992 4993
	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);
4994 4995 4996 4997 4998 4999 5000 5001 5002 5003 5004 5005 5006 5007 5008 5009 5010 5011 5012 5013 5014 5015 5016 5017 5018

	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;

5019 5020 5021 5022 5023
	/*
	 * 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.
5024 5025
	 *
	 * DO NOT ADD NEW FILES.
5026 5027 5028 5029 5030 5031 5032 5033 5034 5035 5036 5037 5038
	 */
	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
5039 5040
		event->register_event = memsw_cgroup_usage_register_event;
		event->unregister_event = memsw_cgroup_usage_unregister_event;
5041 5042 5043 5044 5045
	} else {
		ret = -EINVAL;
		goto out_put_cfile;
	}

5046
	/*
5047 5048 5049
	 * 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.
5050
	 */
5051 5052
	cfile_css = css_tryget_online_from_dir(cfile.file->f_dentry->d_parent,
					       &memory_cgrp_subsys);
5053
	ret = -EINVAL;
5054
	if (IS_ERR(cfile_css))
5055
		goto out_put_cfile;
5056 5057
	if (cfile_css != css) {
		css_put(cfile_css);
5058
		goto out_put_cfile;
5059
	}
5060

5061
	ret = event->register_event(memcg, event->eventfd, buf);
5062 5063 5064 5065 5066
	if (ret)
		goto out_put_css;

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

5067 5068 5069
	spin_lock(&memcg->event_list_lock);
	list_add(&event->list, &memcg->event_list);
	spin_unlock(&memcg->event_list_lock);
5070 5071 5072 5073

	fdput(cfile);
	fdput(efile);

5074
	return nbytes;
5075 5076

out_put_css:
5077
	css_put(css);
5078 5079 5080 5081 5082 5083 5084 5085 5086 5087 5088 5089
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
5090 5091
static struct cftype mem_cgroup_files[] = {
	{
5092
		.name = "usage_in_bytes",
5093
		.private = MEMFILE_PRIVATE(_MEM, RES_USAGE),
5094
		.read_u64 = mem_cgroup_read_u64,
B
Balbir Singh 已提交
5095
	},
5096 5097
	{
		.name = "max_usage_in_bytes",
5098
		.private = MEMFILE_PRIVATE(_MEM, RES_MAX_USAGE),
5099
		.write = mem_cgroup_reset,
5100
		.read_u64 = mem_cgroup_read_u64,
5101
	},
B
Balbir Singh 已提交
5102
	{
5103
		.name = "limit_in_bytes",
5104
		.private = MEMFILE_PRIVATE(_MEM, RES_LIMIT),
5105
		.write = mem_cgroup_write,
5106
		.read_u64 = mem_cgroup_read_u64,
B
Balbir Singh 已提交
5107
	},
5108 5109 5110
	{
		.name = "soft_limit_in_bytes",
		.private = MEMFILE_PRIVATE(_MEM, RES_SOFT_LIMIT),
5111
		.write = mem_cgroup_write,
5112
		.read_u64 = mem_cgroup_read_u64,
5113
	},
B
Balbir Singh 已提交
5114 5115
	{
		.name = "failcnt",
5116
		.private = MEMFILE_PRIVATE(_MEM, RES_FAILCNT),
5117
		.write = mem_cgroup_reset,
5118
		.read_u64 = mem_cgroup_read_u64,
B
Balbir Singh 已提交
5119
	},
5120 5121
	{
		.name = "stat",
5122
		.seq_show = memcg_stat_show,
5123
	},
5124 5125
	{
		.name = "force_empty",
5126
		.write = mem_cgroup_force_empty_write,
5127
	},
5128 5129 5130 5131 5132
	{
		.name = "use_hierarchy",
		.write_u64 = mem_cgroup_hierarchy_write,
		.read_u64 = mem_cgroup_hierarchy_read,
	},
5133
	{
5134
		.name = "cgroup.event_control",		/* XXX: for compat */
5135
		.write = memcg_write_event_control,
5136 5137 5138
		.flags = CFTYPE_NO_PREFIX,
		.mode = S_IWUGO,
	},
K
KOSAKI Motohiro 已提交
5139 5140 5141 5142 5143
	{
		.name = "swappiness",
		.read_u64 = mem_cgroup_swappiness_read,
		.write_u64 = mem_cgroup_swappiness_write,
	},
5144 5145 5146 5147 5148
	{
		.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
5149 5150
	{
		.name = "oom_control",
5151
		.seq_show = mem_cgroup_oom_control_read,
5152
		.write_u64 = mem_cgroup_oom_control_write,
K
KAMEZAWA Hiroyuki 已提交
5153 5154
		.private = MEMFILE_PRIVATE(_OOM_TYPE, OOM_CONTROL),
	},
5155 5156 5157
	{
		.name = "pressure_level",
	},
5158 5159 5160
#ifdef CONFIG_NUMA
	{
		.name = "numa_stat",
5161
		.seq_show = memcg_numa_stat_show,
5162 5163
	},
#endif
5164 5165 5166 5167
#ifdef CONFIG_MEMCG_KMEM
	{
		.name = "kmem.limit_in_bytes",
		.private = MEMFILE_PRIVATE(_KMEM, RES_LIMIT),
5168
		.write = mem_cgroup_write,
5169
		.read_u64 = mem_cgroup_read_u64,
5170 5171 5172 5173
	},
	{
		.name = "kmem.usage_in_bytes",
		.private = MEMFILE_PRIVATE(_KMEM, RES_USAGE),
5174
		.read_u64 = mem_cgroup_read_u64,
5175 5176 5177 5178
	},
	{
		.name = "kmem.failcnt",
		.private = MEMFILE_PRIVATE(_KMEM, RES_FAILCNT),
5179
		.write = mem_cgroup_reset,
5180
		.read_u64 = mem_cgroup_read_u64,
5181 5182 5183 5184
	},
	{
		.name = "kmem.max_usage_in_bytes",
		.private = MEMFILE_PRIVATE(_KMEM, RES_MAX_USAGE),
5185
		.write = mem_cgroup_reset,
5186
		.read_u64 = mem_cgroup_read_u64,
5187
	},
5188 5189 5190
#ifdef CONFIG_SLABINFO
	{
		.name = "kmem.slabinfo",
5191
		.seq_show = mem_cgroup_slabinfo_read,
5192 5193
	},
#endif
5194
#endif
5195
	{ },	/* terminate */
5196
};
5197

5198 5199 5200 5201 5202
#ifdef CONFIG_MEMCG_SWAP
static struct cftype memsw_cgroup_files[] = {
	{
		.name = "memsw.usage_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_USAGE),
5203
		.read_u64 = mem_cgroup_read_u64,
5204 5205 5206 5207
	},
	{
		.name = "memsw.max_usage_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_MAX_USAGE),
5208
		.write = mem_cgroup_reset,
5209
		.read_u64 = mem_cgroup_read_u64,
5210 5211 5212 5213
	},
	{
		.name = "memsw.limit_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_LIMIT),
5214
		.write = mem_cgroup_write,
5215
		.read_u64 = mem_cgroup_read_u64,
5216 5217 5218 5219
	},
	{
		.name = "memsw.failcnt",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_FAILCNT),
5220
		.write = mem_cgroup_reset,
5221
		.read_u64 = mem_cgroup_read_u64,
5222 5223 5224 5225
	},
	{ },	/* terminate */
};
#endif
5226
static int alloc_mem_cgroup_per_zone_info(struct mem_cgroup *memcg, int node)
5227 5228
{
	struct mem_cgroup_per_node *pn;
5229
	struct mem_cgroup_per_zone *mz;
5230
	int zone, tmp = node;
5231 5232 5233 5234 5235 5236 5237 5238
	/*
	 * 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.
	 */
5239 5240
	if (!node_state(node, N_NORMAL_MEMORY))
		tmp = -1;
5241
	pn = kzalloc_node(sizeof(*pn), GFP_KERNEL, tmp);
5242 5243
	if (!pn)
		return 1;
5244 5245 5246

	for (zone = 0; zone < MAX_NR_ZONES; zone++) {
		mz = &pn->zoneinfo[zone];
5247
		lruvec_init(&mz->lruvec);
5248 5249
		mz->usage_in_excess = 0;
		mz->on_tree = false;
5250
		mz->memcg = memcg;
5251
	}
5252
	memcg->nodeinfo[node] = pn;
5253 5254 5255
	return 0;
}

5256
static void free_mem_cgroup_per_zone_info(struct mem_cgroup *memcg, int node)
5257
{
5258
	kfree(memcg->nodeinfo[node]);
5259 5260
}

5261 5262
static struct mem_cgroup *mem_cgroup_alloc(void)
{
5263
	struct mem_cgroup *memcg;
5264
	size_t size;
5265

5266 5267
	size = sizeof(struct mem_cgroup);
	size += nr_node_ids * sizeof(struct mem_cgroup_per_node *);
5268

5269
	memcg = kzalloc(size, GFP_KERNEL);
5270
	if (!memcg)
5271 5272
		return NULL;

5273 5274
	memcg->stat = alloc_percpu(struct mem_cgroup_stat_cpu);
	if (!memcg->stat)
5275
		goto out_free;
5276 5277
	spin_lock_init(&memcg->pcp_counter_lock);
	return memcg;
5278 5279

out_free:
5280
	kfree(memcg);
5281
	return NULL;
5282 5283
}

5284
/*
5285 5286 5287 5288 5289 5290 5291 5292
 * 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.
5293
 */
5294 5295

static void __mem_cgroup_free(struct mem_cgroup *memcg)
5296
{
5297
	int node;
5298

5299
	mem_cgroup_remove_from_trees(memcg);
5300 5301 5302 5303 5304 5305

	for_each_node(node)
		free_mem_cgroup_per_zone_info(memcg, node);

	free_percpu(memcg->stat);

5306 5307 5308 5309 5310 5311 5312 5313 5314 5315 5316
	/*
	 * 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.
	 */
5317
	disarm_static_keys(memcg);
5318
	kfree(memcg);
5319
}
5320

5321 5322 5323
/*
 * Returns the parent mem_cgroup in memcgroup hierarchy with hierarchy enabled.
 */
G
Glauber Costa 已提交
5324
struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
5325
{
5326
	if (!memcg->memory.parent)
5327
		return NULL;
5328
	return mem_cgroup_from_counter(memcg->memory.parent, memory);
5329
}
G
Glauber Costa 已提交
5330
EXPORT_SYMBOL(parent_mem_cgroup);
5331

5332 5333 5334 5335 5336 5337 5338 5339 5340 5341 5342 5343 5344 5345 5346 5347 5348 5349 5350 5351 5352 5353 5354
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
5355
static struct cgroup_subsys_state * __ref
5356
mem_cgroup_css_alloc(struct cgroup_subsys_state *parent_css)
B
Balbir Singh 已提交
5357
{
5358
	struct mem_cgroup *memcg;
K
KAMEZAWA Hiroyuki 已提交
5359
	long error = -ENOMEM;
5360
	int node;
B
Balbir Singh 已提交
5361

5362 5363
	memcg = mem_cgroup_alloc();
	if (!memcg)
K
KAMEZAWA Hiroyuki 已提交
5364
		return ERR_PTR(error);
5365

B
Bob Liu 已提交
5366
	for_each_node(node)
5367
		if (alloc_mem_cgroup_per_zone_info(memcg, node))
5368
			goto free_out;
5369

5370
	/* root ? */
5371
	if (parent_css == NULL) {
5372
		root_mem_cgroup = memcg;
5373 5374 5375
		page_counter_init(&memcg->memory, NULL);
		page_counter_init(&memcg->memsw, NULL);
		page_counter_init(&memcg->kmem, NULL);
5376
	}
5377

5378 5379 5380 5381 5382
	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);
5383
	vmpressure_init(&memcg->vmpressure);
5384 5385
	INIT_LIST_HEAD(&memcg->event_list);
	spin_lock_init(&memcg->event_list_lock);
5386 5387 5388 5389 5390 5391 5392 5393 5394

	return &memcg->css;

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

static int
5395
mem_cgroup_css_online(struct cgroup_subsys_state *css)
5396
{
5397
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
T
Tejun Heo 已提交
5398
	struct mem_cgroup *parent = mem_cgroup_from_css(css->parent);
5399
	int ret;
5400

5401
	if (css->id > MEM_CGROUP_ID_MAX)
5402 5403
		return -ENOSPC;

T
Tejun Heo 已提交
5404
	if (!parent)
5405 5406
		return 0;

5407
	mutex_lock(&memcg_create_mutex);
5408 5409 5410 5411 5412 5413

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

	if (parent->use_hierarchy) {
5414 5415 5416
		page_counter_init(&memcg->memory, &parent->memory);
		page_counter_init(&memcg->memsw, &parent->memsw);
		page_counter_init(&memcg->kmem, &parent->kmem);
5417

5418
		/*
5419 5420
		 * No need to take a reference to the parent because cgroup
		 * core guarantees its existence.
5421
		 */
5422
	} else {
5423 5424 5425
		page_counter_init(&memcg->memory, NULL);
		page_counter_init(&memcg->memsw, NULL);
		page_counter_init(&memcg->kmem, NULL);
5426 5427 5428 5429 5430
		/*
		 * Deeper hierachy with use_hierarchy == false doesn't make
		 * much sense so let cgroup subsystem know about this
		 * unfortunate state in our controller.
		 */
5431
		if (parent != root_mem_cgroup)
5432
			memory_cgrp_subsys.broken_hierarchy = true;
5433
	}
5434
	mutex_unlock(&memcg_create_mutex);
5435

5436 5437 5438 5439 5440 5441 5442 5443 5444 5445 5446 5447
	ret = memcg_init_kmem(memcg, &memory_cgrp_subsys);
	if (ret)
		return ret;

	/*
	 * Make sure the memcg is initialized: mem_cgroup_iter()
	 * orders reading memcg->initialized against its callers
	 * reading the memcg members.
	 */
	smp_store_release(&memcg->initialized, 1);

	return 0;
B
Balbir Singh 已提交
5448 5449
}

M
Michal Hocko 已提交
5450 5451 5452 5453 5454 5455 5456 5457
/*
 * 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)))
5458
		mem_cgroup_iter_invalidate(parent);
M
Michal Hocko 已提交
5459 5460 5461 5462 5463 5464

	/*
	 * if the root memcg is not hierarchical we have to check it
	 * explicitely.
	 */
	if (!root_mem_cgroup->use_hierarchy)
5465
		mem_cgroup_iter_invalidate(root_mem_cgroup);
M
Michal Hocko 已提交
5466 5467
}

5468
static void mem_cgroup_css_offline(struct cgroup_subsys_state *css)
5469
{
5470
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
5471
	struct mem_cgroup_event *event, *tmp;
5472
	struct cgroup_subsys_state *iter;
5473 5474 5475 5476 5477 5478

	/*
	 * Unregister events and notify userspace.
	 * Notify userspace about cgroup removing only after rmdir of cgroup
	 * directory to avoid race between userspace and kernelspace.
	 */
5479 5480
	spin_lock(&memcg->event_list_lock);
	list_for_each_entry_safe(event, tmp, &memcg->event_list, list) {
5481 5482 5483
		list_del_init(&event->list);
		schedule_work(&event->remove);
	}
5484
	spin_unlock(&memcg->event_list_lock);
5485

5486 5487
	kmem_cgroup_css_offline(memcg);

M
Michal Hocko 已提交
5488
	mem_cgroup_invalidate_reclaim_iterators(memcg);
5489 5490 5491 5492 5493 5494 5495 5496

	/*
	 * 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));

5497
	memcg_unregister_all_caches(memcg);
5498
	vmpressure_cleanup(&memcg->vmpressure);
5499 5500
}

5501
static void mem_cgroup_css_free(struct cgroup_subsys_state *css)
B
Balbir Singh 已提交
5502
{
5503
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
5504 5505 5506
	/*
	 * XXX: css_offline() would be where we should reparent all
	 * memory to prepare the cgroup for destruction.  However,
5507
	 * memcg does not do css_tryget_online() and page_counter charging
5508 5509 5510 5511 5512 5513 5514 5515 5516 5517 5518 5519 5520
	 * 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()
5521
	 *                           css_tryget_online()
5522
	 *                           rcu_read_unlock()
5523
	 * disable css_tryget_online()
5524 5525 5526
	 * call_rcu()
	 *   offline_css()
	 *     reparent_charges()
5527
	 *                           page_counter_try_charge()
5528 5529 5530 5531 5532 5533 5534 5535 5536 5537 5538 5539
	 *                           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);
5540

5541
	memcg_destroy_kmem(memcg);
5542
	__mem_cgroup_free(memcg);
B
Balbir Singh 已提交
5543 5544
}

5545 5546 5547 5548 5549 5550 5551 5552 5553 5554 5555 5556 5557 5558 5559 5560 5561
/**
 * mem_cgroup_css_reset - reset the states of a mem_cgroup
 * @css: the target css
 *
 * Reset the states of the mem_cgroup associated with @css.  This is
 * invoked when the userland requests disabling on the default hierarchy
 * but the memcg is pinned through dependency.  The memcg should stop
 * applying policies and should revert to the vanilla state as it may be
 * made visible again.
 *
 * The current implementation only resets the essential configurations.
 * This needs to be expanded to cover all the visible parts.
 */
static void mem_cgroup_css_reset(struct cgroup_subsys_state *css)
{
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);

5562 5563 5564 5565
	mem_cgroup_resize_limit(memcg, PAGE_COUNTER_MAX);
	mem_cgroup_resize_memsw_limit(memcg, PAGE_COUNTER_MAX);
	memcg_update_kmem_limit(memcg, PAGE_COUNTER_MAX);
	memcg->soft_limit = 0;
5566 5567
}

5568
#ifdef CONFIG_MMU
5569
/* Handlers for move charge at task migration. */
5570
static int mem_cgroup_do_precharge(unsigned long count)
5571
{
5572
	int ret;
5573 5574

	/* Try a single bulk charge without reclaim first */
5575
	ret = try_charge(mc.to, GFP_KERNEL & ~__GFP_WAIT, count);
5576
	if (!ret) {
5577 5578 5579
		mc.precharge += count;
		return ret;
	}
5580
	if (ret == -EINTR) {
5581
		cancel_charge(root_mem_cgroup, count);
5582 5583
		return ret;
	}
5584 5585

	/* Try charges one by one with reclaim */
5586
	while (count--) {
5587
		ret = try_charge(mc.to, GFP_KERNEL & ~__GFP_NORETRY, 1);
5588 5589 5590
		/*
		 * In case of failure, any residual charges against
		 * mc.to will be dropped by mem_cgroup_clear_mc()
5591 5592
		 * later on.  However, cancel any charges that are
		 * bypassed to root right away or they'll be lost.
5593
		 */
5594
		if (ret == -EINTR)
5595
			cancel_charge(root_mem_cgroup, 1);
5596 5597
		if (ret)
			return ret;
5598
		mc.precharge++;
5599
		cond_resched();
5600
	}
5601
	return 0;
5602 5603 5604
}

/**
5605
 * get_mctgt_type - get target type of moving charge
5606 5607 5608
 * @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
5609
 * @target: the pointer the target page or swap ent will be stored(can be NULL)
5610 5611 5612 5613 5614 5615
 *
 * 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).
5616 5617 5618
 *   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.
5619 5620 5621 5622 5623
 *
 * Called with pte lock held.
 */
union mc_target {
	struct page	*page;
5624
	swp_entry_t	ent;
5625 5626 5627
};

enum mc_target_type {
5628
	MC_TARGET_NONE = 0,
5629
	MC_TARGET_PAGE,
5630
	MC_TARGET_SWAP,
5631 5632
};

D
Daisuke Nishimura 已提交
5633 5634
static struct page *mc_handle_present_pte(struct vm_area_struct *vma,
						unsigned long addr, pte_t ptent)
5635
{
D
Daisuke Nishimura 已提交
5636
	struct page *page = vm_normal_page(vma, addr, ptent);
5637

D
Daisuke Nishimura 已提交
5638 5639 5640 5641
	if (!page || !page_mapped(page))
		return NULL;
	if (PageAnon(page)) {
		/* we don't move shared anon */
5642
		if (!move_anon())
D
Daisuke Nishimura 已提交
5643
			return NULL;
5644 5645
	} else if (!move_file())
		/* we ignore mapcount for file pages */
D
Daisuke Nishimura 已提交
5646 5647 5648 5649 5650 5651 5652
		return NULL;
	if (!get_page_unless_zero(page))
		return NULL;

	return page;
}

5653
#ifdef CONFIG_SWAP
D
Daisuke Nishimura 已提交
5654 5655 5656 5657 5658 5659 5660 5661
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;
5662 5663 5664 5665
	/*
	 * Because lookup_swap_cache() updates some statistics counter,
	 * we call find_get_page() with swapper_space directly.
	 */
5666
	page = find_get_page(swap_address_space(ent), ent.val);
D
Daisuke Nishimura 已提交
5667 5668 5669 5670 5671
	if (do_swap_account)
		entry->val = ent.val;

	return page;
}
5672 5673 5674 5675 5676 5677 5678
#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
5679

5680 5681 5682 5683 5684 5685 5686 5687 5688 5689 5690 5691 5692 5693 5694 5695 5696 5697 5698
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). */
5699 5700
#ifdef CONFIG_SWAP
	/* shmem/tmpfs may report page out on swap: account for that too. */
5701 5702 5703 5704 5705 5706 5707 5708 5709 5710 5711 5712
	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);
5713
#endif
5714 5715 5716
	return page;
}

5717
static enum mc_target_type get_mctgt_type(struct vm_area_struct *vma,
D
Daisuke Nishimura 已提交
5718 5719 5720 5721
		unsigned long addr, pte_t ptent, union mc_target *target)
{
	struct page *page = NULL;
	struct page_cgroup *pc;
5722
	enum mc_target_type ret = MC_TARGET_NONE;
D
Daisuke Nishimura 已提交
5723 5724 5725 5726 5727 5728
	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);
5729 5730
	else if (pte_none(ptent) || pte_file(ptent))
		page = mc_handle_file_pte(vma, addr, ptent, &ent);
D
Daisuke Nishimura 已提交
5731 5732

	if (!page && !ent.val)
5733
		return ret;
5734 5735 5736
	if (page) {
		pc = lookup_page_cgroup(page);
		/*
5737 5738 5739
		 * Do only loose check w/o serialization.
		 * mem_cgroup_move_account() checks the pc is valid or
		 * not under LRU exclusion.
5740 5741 5742 5743 5744 5745 5746 5747 5748
		 */
		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
5749 5750
	/* 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
5751
	    mem_cgroup_id(mc.from) == lookup_swap_cgroup_id(ent)) {
5752 5753 5754
		ret = MC_TARGET_SWAP;
		if (target)
			target->ent = ent;
5755 5756 5757 5758
	}
	return ret;
}

5759 5760 5761 5762 5763 5764 5765 5766 5767 5768 5769 5770 5771 5772
#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);
5773
	VM_BUG_ON_PAGE(!page || !PageHead(page), page);
5774 5775 5776 5777 5778 5779 5780 5781 5782 5783 5784 5785 5786 5787 5788 5789 5790 5791 5792 5793
	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

5794 5795 5796 5797 5798 5799 5800 5801
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;

5802
	if (pmd_trans_huge_lock(pmd, vma, &ptl) == 1) {
5803 5804
		if (get_mctgt_type_thp(vma, addr, *pmd, NULL) == MC_TARGET_PAGE)
			mc.precharge += HPAGE_PMD_NR;
5805
		spin_unlock(ptl);
5806
		return 0;
5807
	}
5808

5809 5810
	if (pmd_trans_unstable(pmd))
		return 0;
5811 5812
	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
	for (; addr != end; pte++, addr += PAGE_SIZE)
5813
		if (get_mctgt_type(vma, addr, *pte, NULL))
5814 5815 5816 5817
			mc.precharge++;	/* increment precharge temporarily */
	pte_unmap_unlock(pte - 1, ptl);
	cond_resched();

5818 5819 5820
	return 0;
}

5821 5822 5823 5824 5825
static unsigned long mem_cgroup_count_precharge(struct mm_struct *mm)
{
	unsigned long precharge;
	struct vm_area_struct *vma;

5826
	down_read(&mm->mmap_sem);
5827 5828 5829 5830 5831 5832 5833 5834 5835 5836 5837
	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);
	}
5838
	up_read(&mm->mmap_sem);
5839 5840 5841 5842 5843 5844 5845 5846 5847

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

	return precharge;
}

static int mem_cgroup_precharge_mc(struct mm_struct *mm)
{
5848 5849 5850 5851 5852
	unsigned long precharge = mem_cgroup_count_precharge(mm);

	VM_BUG_ON(mc.moving_task);
	mc.moving_task = current;
	return mem_cgroup_do_precharge(precharge);
5853 5854
}

5855 5856
/* cancels all extra charges on mc.from and mc.to, and wakes up all waiters. */
static void __mem_cgroup_clear_mc(void)
5857
{
5858 5859
	struct mem_cgroup *from = mc.from;
	struct mem_cgroup *to = mc.to;
L
Li Zefan 已提交
5860
	int i;
5861

5862
	/* we must uncharge all the leftover precharges from mc.to */
5863
	if (mc.precharge) {
5864
		cancel_charge(mc.to, mc.precharge);
5865 5866 5867 5868 5869 5870 5871
		mc.precharge = 0;
	}
	/*
	 * we didn't uncharge from mc.from at mem_cgroup_move_account(), so
	 * we must uncharge here.
	 */
	if (mc.moved_charge) {
5872
		cancel_charge(mc.from, mc.moved_charge);
5873
		mc.moved_charge = 0;
5874
	}
5875 5876 5877
	/* we must fixup refcnts and charges */
	if (mc.moved_swap) {
		/* uncharge swap account from the old cgroup */
5878
		if (!mem_cgroup_is_root(mc.from))
5879
			page_counter_uncharge(&mc.from->memsw, mc.moved_swap);
5880

5881
		/*
5882 5883
		 * we charged both to->memory and to->memsw, so we
		 * should uncharge to->memory.
5884
		 */
5885
		if (!mem_cgroup_is_root(mc.to))
5886 5887 5888 5889 5890
			page_counter_uncharge(&mc.to->memory, mc.moved_swap);

		for (i = 0; i < mc.moved_swap; i++)
			css_put(&mc.from->css);

L
Li Zefan 已提交
5891
		/* we've already done css_get(mc.to) */
5892 5893
		mc.moved_swap = 0;
	}
5894 5895 5896 5897 5898 5899 5900 5901 5902 5903 5904 5905 5906 5907 5908
	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();
5909
	spin_lock(&mc.lock);
5910 5911
	mc.from = NULL;
	mc.to = NULL;
5912
	spin_unlock(&mc.lock);
5913
	mem_cgroup_end_move(from);
5914 5915
}

5916
static int mem_cgroup_can_attach(struct cgroup_subsys_state *css,
5917
				 struct cgroup_taskset *tset)
5918
{
5919
	struct task_struct *p = cgroup_taskset_first(tset);
5920
	int ret = 0;
5921
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
5922
	unsigned long move_charge_at_immigrate;
5923

5924 5925 5926 5927 5928 5929 5930
	/*
	 * 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) {
5931 5932 5933
		struct mm_struct *mm;
		struct mem_cgroup *from = mem_cgroup_from_task(p);

5934
		VM_BUG_ON(from == memcg);
5935 5936 5937 5938 5939

		mm = get_task_mm(p);
		if (!mm)
			return 0;
		/* We move charges only when we move a owner of the mm */
5940 5941 5942 5943
		if (mm->owner == p) {
			VM_BUG_ON(mc.from);
			VM_BUG_ON(mc.to);
			VM_BUG_ON(mc.precharge);
5944
			VM_BUG_ON(mc.moved_charge);
5945
			VM_BUG_ON(mc.moved_swap);
5946
			mem_cgroup_start_move(from);
5947
			spin_lock(&mc.lock);
5948
			mc.from = from;
5949
			mc.to = memcg;
5950
			mc.immigrate_flags = move_charge_at_immigrate;
5951
			spin_unlock(&mc.lock);
5952
			/* We set mc.moving_task later */
5953 5954 5955 5956

			ret = mem_cgroup_precharge_mc(mm);
			if (ret)
				mem_cgroup_clear_mc();
5957 5958
		}
		mmput(mm);
5959 5960 5961 5962
	}
	return ret;
}

5963
static void mem_cgroup_cancel_attach(struct cgroup_subsys_state *css,
5964
				     struct cgroup_taskset *tset)
5965
{
5966
	mem_cgroup_clear_mc();
5967 5968
}

5969 5970 5971
static int mem_cgroup_move_charge_pte_range(pmd_t *pmd,
				unsigned long addr, unsigned long end,
				struct mm_walk *walk)
5972
{
5973 5974 5975 5976
	int ret = 0;
	struct vm_area_struct *vma = walk->private;
	pte_t *pte;
	spinlock_t *ptl;
5977 5978 5979 5980
	enum mc_target_type target_type;
	union mc_target target;
	struct page *page;
	struct page_cgroup *pc;
5981

5982 5983 5984 5985 5986 5987 5988 5989 5990 5991
	/*
	 * 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.
	 */
5992
	if (pmd_trans_huge_lock(pmd, vma, &ptl) == 1) {
5993
		if (mc.precharge < HPAGE_PMD_NR) {
5994
			spin_unlock(ptl);
5995 5996 5997 5998 5999 6000 6001 6002
			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,
6003
							pc, mc.from, mc.to)) {
6004 6005 6006 6007 6008 6009 6010
					mc.precharge -= HPAGE_PMD_NR;
					mc.moved_charge += HPAGE_PMD_NR;
				}
				putback_lru_page(page);
			}
			put_page(page);
		}
6011
		spin_unlock(ptl);
6012
		return 0;
6013 6014
	}

6015 6016
	if (pmd_trans_unstable(pmd))
		return 0;
6017 6018 6019 6020
retry:
	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
	for (; addr != end; addr += PAGE_SIZE) {
		pte_t ptent = *(pte++);
6021
		swp_entry_t ent;
6022 6023 6024 6025

		if (!mc.precharge)
			break;

6026
		switch (get_mctgt_type(vma, addr, ptent, &target)) {
6027 6028 6029 6030 6031
		case MC_TARGET_PAGE:
			page = target.page;
			if (isolate_lru_page(page))
				goto put;
			pc = lookup_page_cgroup(page);
6032
			if (!mem_cgroup_move_account(page, 1, pc,
6033
						     mc.from, mc.to)) {
6034
				mc.precharge--;
6035 6036
				/* we uncharge from mc.from later. */
				mc.moved_charge++;
6037 6038
			}
			putback_lru_page(page);
6039
put:			/* get_mctgt_type() gets the page */
6040 6041
			put_page(page);
			break;
6042 6043
		case MC_TARGET_SWAP:
			ent = target.ent;
6044
			if (!mem_cgroup_move_swap_account(ent, mc.from, mc.to)) {
6045
				mc.precharge--;
6046 6047 6048
				/* we fixup refcnts and charges later. */
				mc.moved_swap++;
			}
6049
			break;
6050 6051 6052 6053 6054 6055 6056 6057 6058 6059 6060 6061 6062 6063
		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.
		 */
6064
		ret = mem_cgroup_do_precharge(1);
6065 6066 6067 6068 6069 6070 6071 6072 6073 6074 6075 6076
		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();
6077 6078 6079 6080 6081 6082 6083 6084 6085 6086 6087 6088 6089
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;
	}
6090 6091 6092 6093 6094 6095 6096 6097 6098 6099 6100 6101 6102 6103 6104 6105 6106 6107
	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;
	}
6108
	up_read(&mm->mmap_sem);
6109 6110
}

6111
static void mem_cgroup_move_task(struct cgroup_subsys_state *css,
6112
				 struct cgroup_taskset *tset)
B
Balbir Singh 已提交
6113
{
6114
	struct task_struct *p = cgroup_taskset_first(tset);
6115
	struct mm_struct *mm = get_task_mm(p);
6116 6117

	if (mm) {
6118 6119
		if (mc.to)
			mem_cgroup_move_charge(mm);
6120 6121
		mmput(mm);
	}
6122 6123
	if (mc.to)
		mem_cgroup_clear_mc();
B
Balbir Singh 已提交
6124
}
6125
#else	/* !CONFIG_MMU */
6126
static int mem_cgroup_can_attach(struct cgroup_subsys_state *css,
6127
				 struct cgroup_taskset *tset)
6128 6129 6130
{
	return 0;
}
6131
static void mem_cgroup_cancel_attach(struct cgroup_subsys_state *css,
6132
				     struct cgroup_taskset *tset)
6133 6134
{
}
6135
static void mem_cgroup_move_task(struct cgroup_subsys_state *css,
6136
				 struct cgroup_taskset *tset)
6137 6138 6139
{
}
#endif
B
Balbir Singh 已提交
6140

6141 6142
/*
 * Cgroup retains root cgroups across [un]mount cycles making it necessary
6143 6144
 * to verify whether we're attached to the default hierarchy on each mount
 * attempt.
6145
 */
6146
static void mem_cgroup_bind(struct cgroup_subsys_state *root_css)
6147 6148
{
	/*
6149
	 * use_hierarchy is forced on the default hierarchy.  cgroup core
6150 6151 6152
	 * guarantees that @root doesn't have any children, so turning it
	 * on for the root memcg is enough.
	 */
6153
	if (cgroup_on_dfl(root_css->cgroup))
6154
		mem_cgroup_from_css(root_css)->use_hierarchy = true;
6155 6156
}

6157
struct cgroup_subsys memory_cgrp_subsys = {
6158
	.css_alloc = mem_cgroup_css_alloc,
6159
	.css_online = mem_cgroup_css_online,
6160 6161
	.css_offline = mem_cgroup_css_offline,
	.css_free = mem_cgroup_css_free,
6162
	.css_reset = mem_cgroup_css_reset,
6163 6164
	.can_attach = mem_cgroup_can_attach,
	.cancel_attach = mem_cgroup_cancel_attach,
B
Balbir Singh 已提交
6165
	.attach = mem_cgroup_move_task,
6166
	.bind = mem_cgroup_bind,
6167
	.legacy_cftypes = mem_cgroup_files,
6168
	.early_init = 0,
B
Balbir Singh 已提交
6169
};
6170

A
Andrew Morton 已提交
6171
#ifdef CONFIG_MEMCG_SWAP
6172 6173
static int __init enable_swap_account(char *s)
{
6174
	if (!strcmp(s, "1"))
6175
		really_do_swap_account = 1;
6176
	else if (!strcmp(s, "0"))
6177 6178 6179
		really_do_swap_account = 0;
	return 1;
}
6180
__setup("swapaccount=", enable_swap_account);
6181

6182 6183
static void __init memsw_file_init(void)
{
6184 6185
	WARN_ON(cgroup_add_legacy_cftypes(&memory_cgrp_subsys,
					  memsw_cgroup_files));
6186 6187 6188 6189 6190 6191 6192 6193
}

static void __init enable_swap_cgroup(void)
{
	if (!mem_cgroup_disabled() && really_do_swap_account) {
		do_swap_account = 1;
		memsw_file_init();
	}
6194
}
6195

6196
#else
6197
static void __init enable_swap_cgroup(void)
6198 6199
{
}
6200
#endif
6201

6202 6203 6204 6205 6206 6207 6208 6209 6210 6211 6212 6213 6214 6215 6216 6217 6218 6219 6220 6221 6222 6223 6224 6225 6226 6227 6228 6229 6230 6231 6232 6233 6234 6235 6236 6237 6238 6239 6240 6241 6242 6243 6244 6245 6246 6247 6248 6249 6250 6251 6252 6253 6254
#ifdef CONFIG_MEMCG_SWAP
/**
 * mem_cgroup_swapout - transfer a memsw charge to swap
 * @page: page whose memsw charge to transfer
 * @entry: swap entry to move the charge to
 *
 * Transfer the memsw charge of @page to @entry.
 */
void mem_cgroup_swapout(struct page *page, swp_entry_t entry)
{
	struct page_cgroup *pc;
	unsigned short oldid;

	VM_BUG_ON_PAGE(PageLRU(page), page);
	VM_BUG_ON_PAGE(page_count(page), page);

	if (!do_swap_account)
		return;

	pc = lookup_page_cgroup(page);

	/* Readahead page, never charged */
	if (!PageCgroupUsed(pc))
		return;

	VM_BUG_ON_PAGE(!(pc->flags & PCG_MEMSW), page);

	oldid = swap_cgroup_record(entry, mem_cgroup_id(pc->mem_cgroup));
	VM_BUG_ON_PAGE(oldid, page);

	pc->flags &= ~PCG_MEMSW;
	css_get(&pc->mem_cgroup->css);
	mem_cgroup_swap_statistics(pc->mem_cgroup, true);
}

/**
 * mem_cgroup_uncharge_swap - uncharge a swap entry
 * @entry: swap entry to uncharge
 *
 * Drop the memsw charge associated with @entry.
 */
void mem_cgroup_uncharge_swap(swp_entry_t entry)
{
	struct mem_cgroup *memcg;
	unsigned short id;

	if (!do_swap_account)
		return;

	id = swap_cgroup_record(entry, 0);
	rcu_read_lock();
	memcg = mem_cgroup_lookup(id);
	if (memcg) {
6255
		if (!mem_cgroup_is_root(memcg))
6256
			page_counter_uncharge(&memcg->memsw, 1);
6257 6258 6259 6260 6261 6262 6263
		mem_cgroup_swap_statistics(memcg, false);
		css_put(&memcg->css);
	}
	rcu_read_unlock();
}
#endif

6264 6265 6266 6267 6268 6269 6270 6271 6272 6273 6274 6275 6276 6277 6278 6279 6280 6281 6282 6283 6284 6285 6286 6287 6288 6289 6290 6291 6292 6293 6294 6295 6296 6297 6298 6299 6300 6301 6302 6303 6304 6305 6306 6307 6308 6309 6310 6311 6312 6313 6314 6315 6316 6317 6318 6319 6320 6321 6322 6323 6324 6325 6326 6327 6328 6329 6330 6331 6332 6333 6334 6335 6336 6337 6338 6339 6340 6341 6342 6343 6344 6345 6346 6347 6348 6349 6350 6351 6352 6353 6354 6355 6356 6357 6358 6359 6360
/**
 * mem_cgroup_try_charge - try charging a page
 * @page: page to charge
 * @mm: mm context of the victim
 * @gfp_mask: reclaim mode
 * @memcgp: charged memcg return
 *
 * Try to charge @page to the memcg that @mm belongs to, reclaiming
 * pages according to @gfp_mask if necessary.
 *
 * Returns 0 on success, with *@memcgp pointing to the charged memcg.
 * Otherwise, an error code is returned.
 *
 * After page->mapping has been set up, the caller must finalize the
 * charge with mem_cgroup_commit_charge().  Or abort the transaction
 * with mem_cgroup_cancel_charge() in case page instantiation fails.
 */
int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
			  gfp_t gfp_mask, struct mem_cgroup **memcgp)
{
	struct mem_cgroup *memcg = NULL;
	unsigned int nr_pages = 1;
	int ret = 0;

	if (mem_cgroup_disabled())
		goto out;

	if (PageSwapCache(page)) {
		struct page_cgroup *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))
			goto out;
	}

	if (PageTransHuge(page)) {
		nr_pages <<= compound_order(page);
		VM_BUG_ON_PAGE(!PageTransHuge(page), page);
	}

	if (do_swap_account && PageSwapCache(page))
		memcg = try_get_mem_cgroup_from_page(page);
	if (!memcg)
		memcg = get_mem_cgroup_from_mm(mm);

	ret = try_charge(memcg, gfp_mask, nr_pages);

	css_put(&memcg->css);

	if (ret == -EINTR) {
		memcg = root_mem_cgroup;
		ret = 0;
	}
out:
	*memcgp = memcg;
	return ret;
}

/**
 * mem_cgroup_commit_charge - commit a page charge
 * @page: page to charge
 * @memcg: memcg to charge the page to
 * @lrucare: page might be on LRU already
 *
 * Finalize a charge transaction started by mem_cgroup_try_charge(),
 * after page->mapping has been set up.  This must happen atomically
 * as part of the page instantiation, i.e. under the page table lock
 * for anonymous pages, under the page lock for page and swap cache.
 *
 * In addition, the page must not be on the LRU during the commit, to
 * prevent racing with task migration.  If it might be, use @lrucare.
 *
 * Use mem_cgroup_cancel_charge() to cancel the transaction instead.
 */
void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
			      bool lrucare)
{
	unsigned int nr_pages = 1;

	VM_BUG_ON_PAGE(!page->mapping, page);
	VM_BUG_ON_PAGE(PageLRU(page) && !lrucare, page);

	if (mem_cgroup_disabled())
		return;
	/*
	 * Swap faults will attempt to charge the same page multiple
	 * times.  But reuse_swap_page() might have removed the page
	 * from swapcache already, so we can't check PageSwapCache().
	 */
	if (!memcg)
		return;

6361 6362
	commit_charge(page, memcg, lrucare);

6363 6364 6365 6366 6367
	if (PageTransHuge(page)) {
		nr_pages <<= compound_order(page);
		VM_BUG_ON_PAGE(!PageTransHuge(page), page);
	}

6368 6369 6370 6371
	local_irq_disable();
	mem_cgroup_charge_statistics(memcg, page, nr_pages);
	memcg_check_events(memcg, page);
	local_irq_enable();
6372 6373 6374 6375 6376 6377 6378 6379 6380 6381 6382 6383 6384 6385 6386 6387 6388 6389 6390 6391 6392 6393 6394 6395 6396 6397 6398 6399 6400 6401 6402 6403 6404 6405 6406 6407 6408 6409 6410 6411 6412

	if (do_swap_account && PageSwapCache(page)) {
		swp_entry_t entry = { .val = page_private(page) };
		/*
		 * The swap entry might not get freed for a long time,
		 * let's not wait for it.  The page already received a
		 * memory+swap charge, drop the swap entry duplicate.
		 */
		mem_cgroup_uncharge_swap(entry);
	}
}

/**
 * mem_cgroup_cancel_charge - cancel a page charge
 * @page: page to charge
 * @memcg: memcg to charge the page to
 *
 * Cancel a charge transaction started by mem_cgroup_try_charge().
 */
void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg)
{
	unsigned int nr_pages = 1;

	if (mem_cgroup_disabled())
		return;
	/*
	 * Swap faults will attempt to charge the same page multiple
	 * times.  But reuse_swap_page() might have removed the page
	 * from swapcache already, so we can't check PageSwapCache().
	 */
	if (!memcg)
		return;

	if (PageTransHuge(page)) {
		nr_pages <<= compound_order(page);
		VM_BUG_ON_PAGE(!PageTransHuge(page), page);
	}

	cancel_charge(memcg, nr_pages);
}

6413 6414 6415 6416 6417 6418 6419
static void uncharge_batch(struct mem_cgroup *memcg, unsigned long pgpgout,
			   unsigned long nr_mem, unsigned long nr_memsw,
			   unsigned long nr_anon, unsigned long nr_file,
			   unsigned long nr_huge, struct page *dummy_page)
{
	unsigned long flags;

6420 6421
	if (!mem_cgroup_is_root(memcg)) {
		if (nr_mem)
6422
			page_counter_uncharge(&memcg->memory, nr_mem);
6423
		if (nr_memsw)
6424
			page_counter_uncharge(&memcg->memsw, nr_memsw);
6425 6426
		memcg_oom_recover(memcg);
	}
6427 6428 6429 6430 6431 6432 6433 6434 6435 6436 6437 6438 6439 6440 6441 6442 6443 6444 6445 6446 6447 6448 6449 6450 6451 6452 6453 6454 6455 6456 6457 6458 6459 6460 6461 6462 6463 6464 6465 6466 6467 6468 6469 6470 6471 6472 6473 6474 6475 6476 6477 6478 6479 6480 6481 6482 6483 6484 6485 6486 6487 6488 6489 6490 6491 6492 6493 6494 6495 6496 6497 6498 6499 6500 6501 6502 6503 6504 6505

	local_irq_save(flags);
	__this_cpu_sub(memcg->stat->count[MEM_CGROUP_STAT_RSS], nr_anon);
	__this_cpu_sub(memcg->stat->count[MEM_CGROUP_STAT_CACHE], nr_file);
	__this_cpu_sub(memcg->stat->count[MEM_CGROUP_STAT_RSS_HUGE], nr_huge);
	__this_cpu_add(memcg->stat->events[MEM_CGROUP_EVENTS_PGPGOUT], pgpgout);
	__this_cpu_add(memcg->stat->nr_page_events, nr_anon + nr_file);
	memcg_check_events(memcg, dummy_page);
	local_irq_restore(flags);
}

static void uncharge_list(struct list_head *page_list)
{
	struct mem_cgroup *memcg = NULL;
	unsigned long nr_memsw = 0;
	unsigned long nr_anon = 0;
	unsigned long nr_file = 0;
	unsigned long nr_huge = 0;
	unsigned long pgpgout = 0;
	unsigned long nr_mem = 0;
	struct list_head *next;
	struct page *page;

	next = page_list->next;
	do {
		unsigned int nr_pages = 1;
		struct page_cgroup *pc;

		page = list_entry(next, struct page, lru);
		next = page->lru.next;

		VM_BUG_ON_PAGE(PageLRU(page), page);
		VM_BUG_ON_PAGE(page_count(page), page);

		pc = lookup_page_cgroup(page);
		if (!PageCgroupUsed(pc))
			continue;

		/*
		 * Nobody should be changing or seriously looking at
		 * pc->mem_cgroup and pc->flags at this point, we have
		 * fully exclusive access to the page.
		 */

		if (memcg != pc->mem_cgroup) {
			if (memcg) {
				uncharge_batch(memcg, pgpgout, nr_mem, nr_memsw,
					       nr_anon, nr_file, nr_huge, page);
				pgpgout = nr_mem = nr_memsw = 0;
				nr_anon = nr_file = nr_huge = 0;
			}
			memcg = pc->mem_cgroup;
		}

		if (PageTransHuge(page)) {
			nr_pages <<= compound_order(page);
			VM_BUG_ON_PAGE(!PageTransHuge(page), page);
			nr_huge += nr_pages;
		}

		if (PageAnon(page))
			nr_anon += nr_pages;
		else
			nr_file += nr_pages;

		if (pc->flags & PCG_MEM)
			nr_mem += nr_pages;
		if (pc->flags & PCG_MEMSW)
			nr_memsw += nr_pages;
		pc->flags = 0;

		pgpgout++;
	} while (next != page_list);

	if (memcg)
		uncharge_batch(memcg, pgpgout, nr_mem, nr_memsw,
			       nr_anon, nr_file, nr_huge, page);
}

6506 6507 6508 6509 6510 6511 6512 6513 6514 6515 6516 6517 6518 6519
/**
 * mem_cgroup_uncharge - uncharge a page
 * @page: page to uncharge
 *
 * Uncharge a page previously charged with mem_cgroup_try_charge() and
 * mem_cgroup_commit_charge().
 */
void mem_cgroup_uncharge(struct page *page)
{
	struct page_cgroup *pc;

	if (mem_cgroup_disabled())
		return;

6520
	/* Don't touch page->lru of any random page, pre-check: */
6521 6522 6523 6524
	pc = lookup_page_cgroup(page);
	if (!PageCgroupUsed(pc))
		return;

6525 6526 6527
	INIT_LIST_HEAD(&page->lru);
	uncharge_list(&page->lru);
}
6528

6529 6530 6531 6532 6533 6534 6535 6536 6537 6538 6539
/**
 * mem_cgroup_uncharge_list - uncharge a list of page
 * @page_list: list of pages to uncharge
 *
 * Uncharge a list of pages previously charged with
 * mem_cgroup_try_charge() and mem_cgroup_commit_charge().
 */
void mem_cgroup_uncharge_list(struct list_head *page_list)
{
	if (mem_cgroup_disabled())
		return;
6540

6541 6542
	if (!list_empty(page_list))
		uncharge_list(page_list);
6543 6544 6545 6546 6547 6548 6549 6550 6551 6552 6553 6554 6555 6556 6557 6558 6559 6560 6561 6562 6563 6564 6565
}

/**
 * mem_cgroup_migrate - migrate a charge to another page
 * @oldpage: currently charged page
 * @newpage: page to transfer the charge to
 * @lrucare: both pages might be on the LRU already
 *
 * Migrate the charge from @oldpage to @newpage.
 *
 * Both pages must be locked, @newpage->mapping must be set up.
 */
void mem_cgroup_migrate(struct page *oldpage, struct page *newpage,
			bool lrucare)
{
	struct page_cgroup *pc;
	int isolated;

	VM_BUG_ON_PAGE(!PageLocked(oldpage), oldpage);
	VM_BUG_ON_PAGE(!PageLocked(newpage), newpage);
	VM_BUG_ON_PAGE(!lrucare && PageLRU(oldpage), oldpage);
	VM_BUG_ON_PAGE(!lrucare && PageLRU(newpage), newpage);
	VM_BUG_ON_PAGE(PageAnon(oldpage) != PageAnon(newpage), newpage);
6566 6567
	VM_BUG_ON_PAGE(PageTransHuge(oldpage) != PageTransHuge(newpage),
		       newpage);
6568 6569 6570 6571 6572 6573 6574 6575 6576 6577 6578 6579 6580 6581 6582 6583 6584 6585 6586 6587 6588 6589 6590 6591 6592

	if (mem_cgroup_disabled())
		return;

	/* Page cache replacement: new page already charged? */
	pc = lookup_page_cgroup(newpage);
	if (PageCgroupUsed(pc))
		return;

	/* Re-entrant migration: old page already uncharged? */
	pc = lookup_page_cgroup(oldpage);
	if (!PageCgroupUsed(pc))
		return;

	VM_BUG_ON_PAGE(!(pc->flags & PCG_MEM), oldpage);
	VM_BUG_ON_PAGE(do_swap_account && !(pc->flags & PCG_MEMSW), oldpage);

	if (lrucare)
		lock_page_lru(oldpage, &isolated);

	pc->flags = 0;

	if (lrucare)
		unlock_page_lru(oldpage, isolated);

6593
	commit_charge(newpage, pc->mem_cgroup, lrucare);
6594 6595
}

6596
/*
6597 6598 6599 6600 6601 6602
 * 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.
6603 6604 6605 6606
 */
static int __init mem_cgroup_init(void)
{
	hotcpu_notifier(memcg_cpu_hotplug_callback, 0);
6607
	enable_swap_cgroup();
6608
	mem_cgroup_soft_limit_tree_init();
6609
	memcg_stock_init();
6610 6611 6612
	return 0;
}
subsys_initcall(mem_cgroup_init);