memcontrol.c 149.0 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/swap_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
};

146 147
struct reclaim_iter {
	struct mem_cgroup *position;
148 149 150 151
	/* scan generation, increased every round-trip */
	unsigned int generation;
};

152 153 154 155
/*
 * per-zone information in memory controller.
 */
struct mem_cgroup_per_zone {
156
	struct lruvec		lruvec;
157
	unsigned long		lru_size[NR_LRU_LISTS];
K
KOSAKI Motohiro 已提交
158

159
	struct reclaim_iter	iter[DEF_PRIORITY + 1];
160

161
	struct rb_node		tree_node;	/* RB tree node */
162
	unsigned long		usage_in_excess;/* Set to the value by which */
163 164
						/* the soft limit is exceeded*/
	bool			on_tree;
165
	struct mem_cgroup	*memcg;		/* Back pointer, we cannot */
166
						/* use container_of	   */
167 168 169 170 171 172
};

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

173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192
/*
 * 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;

193 194
struct mem_cgroup_threshold {
	struct eventfd_ctx *eventfd;
195
	unsigned long threshold;
196 197
};

K
KAMEZAWA Hiroyuki 已提交
198
/* For threshold */
199
struct mem_cgroup_threshold_ary {
200
	/* An array index points to threshold just below or equal to usage. */
201
	int current_threshold;
202 203 204 205 206
	/* Size of entries[] */
	unsigned int size;
	/* Array of thresholds */
	struct mem_cgroup_threshold entries[0];
};
207 208 209 210 211 212 213 214 215 216 217 218

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
219 220 221 222 223
/* for OOM */
struct mem_cgroup_eventfd_list {
	struct list_head list;
	struct eventfd_ctx *eventfd;
};
224

225 226 227
/*
 * cgroup_event represents events which userspace want to receive.
 */
228
struct mem_cgroup_event {
229
	/*
230
	 * memcg which the event belongs to.
231
	 */
232
	struct mem_cgroup *memcg;
233 234 235 236 237 238 239 240
	/*
	 * 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;
241 242 243 244 245
	/*
	 * 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.
	 */
246
	int (*register_event)(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
247
			      struct eventfd_ctx *eventfd, const char *args);
248 249 250 251 252
	/*
	 * 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.
	 */
253
	void (*unregister_event)(struct mem_cgroup *memcg,
254
				 struct eventfd_ctx *eventfd);
255 256 257 258 259 260 261 262 263 264
	/*
	 * 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;
};

265 266
static void mem_cgroup_threshold(struct mem_cgroup *memcg);
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg);
267

B
Balbir Singh 已提交
268 269 270 271 272 273 274
/*
 * 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
275 276 277
 * 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
278 279 280
 */
struct mem_cgroup {
	struct cgroup_subsys_state css;
281 282 283 284 285 286 287

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

	unsigned long soft_limit;
288

289 290 291
	/* vmpressure notifications */
	struct vmpressure vmpressure;

292 293 294
	/* css_online() has been completed */
	int initialized;

295 296 297 298
	/*
	 * Should the accounting and control be hierarchical, per subtree?
	 */
	bool use_hierarchy;
299 300 301

	bool		oom_lock;
	atomic_t	under_oom;
302
	atomic_t	oom_wakeups;
303

304
	int	swappiness;
305 306
	/* OOM-Killer disable */
	int		oom_kill_disable;
K
KOSAKI Motohiro 已提交
307

308 309 310 311
	/* protect arrays of thresholds */
	struct mutex thresholds_lock;

	/* thresholds for memory usage. RCU-protected */
312
	struct mem_cgroup_thresholds thresholds;
313

314
	/* thresholds for mem+swap usage. RCU-protected */
315
	struct mem_cgroup_thresholds memsw_thresholds;
316

K
KAMEZAWA Hiroyuki 已提交
317 318
	/* For oom notifier event fd */
	struct list_head oom_notify;
319

320 321 322 323
	/*
	 * 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
324
	unsigned long move_charge_at_immigrate;
325 326 327 328
	/*
	 * set > 0 if pages under this cgroup are moving to other cgroup.
	 */
	atomic_t	moving_account;
329 330
	/* taken only while moving_account > 0 */
	spinlock_t	move_lock;
331
	/*
332
	 * percpu counter.
333
	 */
334
	struct mem_cgroup_stat_cpu __percpu *stat;
335 336 337 338 339 340
	/*
	 * 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
341

M
Michal Hocko 已提交
342
#if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_INET)
343
	struct cg_proto tcp_mem;
G
Glauber Costa 已提交
344
#endif
345
#if defined(CONFIG_MEMCG_KMEM)
346 347
	/* analogous to slab_common's slab_caches list, but per-memcg;
	 * protected by memcg_slab_mutex */
348 349 350 351
	struct list_head memcg_slab_caches;
        /* Index in the kmem_cache->memcg_params->memcg_caches array */
	int kmemcg_id;
#endif
352 353 354 355 356 357 358

	int last_scanned_node;
#if MAX_NUMNODES > 1
	nodemask_t	scan_nodes;
	atomic_t	numainfo_events;
	atomic_t	numainfo_updating;
#endif
359

360 361 362 363
	/* List of events which userspace want to receive */
	struct list_head event_list;
	spinlock_t event_list_lock;

364 365
	struct mem_cgroup_per_node *nodeinfo[0];
	/* WARNING: nodeinfo must be the last member here */
B
Balbir Singh 已提交
366 367
};

368
#ifdef CONFIG_MEMCG_KMEM
369 370
static bool memcg_kmem_is_active(struct mem_cgroup *memcg)
{
V
Vladimir Davydov 已提交
371
	return memcg->kmemcg_id >= 0;
372
}
373 374
#endif

375 376
/* Stuffs for move charges at task migration. */
/*
377 378
 * Types of charges to be moved. "move_charge_at_immitgrate" and
 * "immigrate_flags" are treated as a left-shifted bitmap of these types.
379 380
 */
enum move_type {
381
	MOVE_CHARGE_TYPE_ANON,	/* private anonymous page and swap of it */
382
	MOVE_CHARGE_TYPE_FILE,	/* file page(including tmpfs) and swap of it */
383 384 385
	NR_MOVE_TYPE,
};

386 387
/* "mc" and its members are protected by cgroup_mutex */
static struct move_charge_struct {
388
	spinlock_t	  lock; /* for from, to */
389 390
	struct mem_cgroup *from;
	struct mem_cgroup *to;
391
	unsigned long immigrate_flags;
392
	unsigned long precharge;
393
	unsigned long moved_charge;
394
	unsigned long moved_swap;
395 396 397
	struct task_struct *moving_task;	/* a task moving charges */
	wait_queue_head_t waitq;		/* a waitq for other context */
} mc = {
398
	.lock = __SPIN_LOCK_UNLOCKED(mc.lock),
399 400
	.waitq = __WAIT_QUEUE_HEAD_INITIALIZER(mc.waitq),
};
401

D
Daisuke Nishimura 已提交
402 403
static bool move_anon(void)
{
404
	return test_bit(MOVE_CHARGE_TYPE_ANON, &mc.immigrate_flags);
D
Daisuke Nishimura 已提交
405 406
}

407 408
static bool move_file(void)
{
409
	return test_bit(MOVE_CHARGE_TYPE_FILE, &mc.immigrate_flags);
410 411
}

412 413 414 415
/*
 * Maximum loops in mem_cgroup_hierarchical_reclaim(), used for soft
 * limit reclaim to prevent infinite loops, if they ever occur.
 */
416
#define	MEM_CGROUP_MAX_RECLAIM_LOOPS		100
417
#define	MEM_CGROUP_MAX_SOFT_LIMIT_RECLAIM_LOOPS	2
418

419 420
enum charge_type {
	MEM_CGROUP_CHARGE_TYPE_CACHE = 0,
421
	MEM_CGROUP_CHARGE_TYPE_ANON,
K
KAMEZAWA Hiroyuki 已提交
422
	MEM_CGROUP_CHARGE_TYPE_SWAPOUT,	/* for accounting swapcache */
K
KAMEZAWA Hiroyuki 已提交
423
	MEM_CGROUP_CHARGE_TYPE_DROP,	/* a page was unused swap cache */
424 425 426
	NR_CHARGE_TYPE,
};

427
/* for encoding cft->private value on file */
G
Glauber Costa 已提交
428 429 430 431
enum res_type {
	_MEM,
	_MEMSWAP,
	_OOM_TYPE,
432
	_KMEM,
G
Glauber Costa 已提交
433 434
};

435 436
#define MEMFILE_PRIVATE(x, val)	((x) << 16 | (val))
#define MEMFILE_TYPE(val)	((val) >> 16 & 0xffff)
437
#define MEMFILE_ATTR(val)	((val) & 0xffff)
K
KAMEZAWA Hiroyuki 已提交
438 439
/* Used for OOM nofiier */
#define OOM_CONTROL		(0)
440

441 442 443 444 445 446 447
/*
 * 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);

448 449
struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *s)
{
450
	return s ? container_of(s, struct mem_cgroup, css) : NULL;
451 452
}

453 454 455 456 457 458 459 460 461 462 463 464 465
/* 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;
}

466 467 468 469 470
static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
{
	return (memcg == root_mem_cgroup);
}

471 472 473 474 475 476
/*
 * 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
477 478
static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
{
479
	return memcg->css.id;
L
Li Zefan 已提交
480 481 482 483 484 485
}

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

486
	css = css_from_id(id, &memory_cgrp_subsys);
L
Li Zefan 已提交
487 488 489
	return mem_cgroup_from_css(css);
}

G
Glauber Costa 已提交
490
/* Writing them here to avoid exposing memcg's inner layout */
M
Michal Hocko 已提交
491
#if defined(CONFIG_INET) && defined(CONFIG_MEMCG_KMEM)
G
Glauber Costa 已提交
492 493 494

void sock_update_memcg(struct sock *sk)
{
495
	if (mem_cgroup_sockets_enabled) {
G
Glauber Costa 已提交
496
		struct mem_cgroup *memcg;
497
		struct cg_proto *cg_proto;
G
Glauber Costa 已提交
498 499 500

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

501 502 503 504 505 506 507 508 509 510
		/* 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));
511
			css_get(&sk->sk_cgrp->memcg->css);
512 513 514
			return;
		}

G
Glauber Costa 已提交
515 516
		rcu_read_lock();
		memcg = mem_cgroup_from_task(current);
517
		cg_proto = sk->sk_prot->proto_cgroup(memcg);
518
		if (!mem_cgroup_is_root(memcg) &&
519 520
		    memcg_proto_active(cg_proto) &&
		    css_tryget_online(&memcg->css)) {
521
			sk->sk_cgrp = cg_proto;
G
Glauber Costa 已提交
522 523 524 525 526 527 528 529
		}
		rcu_read_unlock();
	}
}
EXPORT_SYMBOL(sock_update_memcg);

void sock_release_memcg(struct sock *sk)
{
530
	if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
G
Glauber Costa 已提交
531 532 533
		struct mem_cgroup *memcg;
		WARN_ON(!sk->sk_cgrp->memcg);
		memcg = sk->sk_cgrp->memcg;
534
		css_put(&sk->sk_cgrp->memcg->css);
G
Glauber Costa 已提交
535 536
	}
}
G
Glauber Costa 已提交
537 538 539 540 541 542

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

543
	return &memcg->tcp_mem;
G
Glauber Costa 已提交
544 545
}
EXPORT_SYMBOL(tcp_proto_cgroup);
G
Glauber Costa 已提交
546

547 548
static void disarm_sock_keys(struct mem_cgroup *memcg)
{
549
	if (!memcg_proto_activated(&memcg->tcp_mem))
550 551 552 553 554 555 556 557 558
		return;
	static_key_slow_dec(&memcg_socket_limit_enabled);
}
#else
static void disarm_sock_keys(struct mem_cgroup *memcg)
{
}
#endif

559
#ifdef CONFIG_MEMCG_KMEM
560 561
/*
 * This will be the memcg's index in each cache's ->memcg_params->memcg_caches.
L
Li Zefan 已提交
562 563 564 565 566
 * 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.
567 568 569 570 571 572
 *
 * 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);
573 574
int memcg_limited_groups_array_size;

575 576 577 578 579 580
/*
 * 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
581
 * MAX_SIZE should be as large as the number of cgrp_ids. Ideally, we could get
582 583
 * 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
584
 * cgrp_id space is not getting any smaller, and we don't have to necessarily
585 586 587
 * increase ours as well if it increases.
 */
#define MEMCG_CACHES_MIN_SIZE 4
L
Li Zefan 已提交
588
#define MEMCG_CACHES_MAX_SIZE MEM_CGROUP_ID_MAX
589

590 591 592 593 594 595
/*
 * 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
 */
596
struct static_key memcg_kmem_enabled_key;
597
EXPORT_SYMBOL(memcg_kmem_enabled_key);
598

599 600
static void memcg_free_cache_id(int id);

601 602
static void disarm_kmem_keys(struct mem_cgroup *memcg)
{
603
	if (memcg_kmem_is_active(memcg)) {
604
		static_key_slow_dec(&memcg_kmem_enabled_key);
605
		memcg_free_cache_id(memcg->kmemcg_id);
606
	}
607 608 609 610
	/*
	 * This check can't live in kmem destruction function,
	 * since the charges will outlive the cgroup
	 */
611
	WARN_ON(page_counter_read(&memcg->kmem));
612 613 614 615 616 617 618 619 620 621 622 623 624
}
#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);
}

625
static struct mem_cgroup_per_zone *
626
mem_cgroup_zone_zoneinfo(struct mem_cgroup *memcg, struct zone *zone)
627
{
628 629 630
	int nid = zone_to_nid(zone);
	int zid = zone_idx(zone);

631
	return &memcg->nodeinfo[nid]->zoneinfo[zid];
632 633
}

634
struct cgroup_subsys_state *mem_cgroup_css(struct mem_cgroup *memcg)
635
{
636
	return &memcg->css;
637 638
}

639
static struct mem_cgroup_per_zone *
640
mem_cgroup_page_zoneinfo(struct mem_cgroup *memcg, struct page *page)
641
{
642 643
	int nid = page_to_nid(page);
	int zid = page_zonenum(page);
644

645
	return &memcg->nodeinfo[nid]->zoneinfo[zid];
646 647
}

648 649 650 651 652 653 654 655 656 657 658 659 660 661 662
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];
}

663 664
static void __mem_cgroup_insert_exceeded(struct mem_cgroup_per_zone *mz,
					 struct mem_cgroup_tree_per_zone *mctz,
665
					 unsigned long new_usage_in_excess)
666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694
{
	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;
}

695 696
static void __mem_cgroup_remove_exceeded(struct mem_cgroup_per_zone *mz,
					 struct mem_cgroup_tree_per_zone *mctz)
697 698 699 700 701 702 703
{
	if (!mz->on_tree)
		return;
	rb_erase(&mz->tree_node, &mctz->rb_root);
	mz->on_tree = false;
}

704 705
static void mem_cgroup_remove_exceeded(struct mem_cgroup_per_zone *mz,
				       struct mem_cgroup_tree_per_zone *mctz)
706
{
707 708 709
	unsigned long flags;

	spin_lock_irqsave(&mctz->lock, flags);
710
	__mem_cgroup_remove_exceeded(mz, mctz);
711
	spin_unlock_irqrestore(&mctz->lock, flags);
712 713
}

714 715 716 717 718 719 720 721 722 723 724
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;
}
725 726 727

static void mem_cgroup_update_tree(struct mem_cgroup *memcg, struct page *page)
{
728
	unsigned long excess;
729 730 731
	struct mem_cgroup_per_zone *mz;
	struct mem_cgroup_tree_per_zone *mctz;

732
	mctz = soft_limit_tree_from_page(page);
733 734 735 736 737
	/*
	 * Necessary to update all ancestors when hierarchy is used.
	 * because their event counter is not touched.
	 */
	for (; memcg; memcg = parent_mem_cgroup(memcg)) {
738
		mz = mem_cgroup_page_zoneinfo(memcg, page);
739
		excess = soft_limit_excess(memcg);
740 741 742 743 744
		/*
		 * We have to update the tree if mz is on RB-tree or
		 * mem is over its softlimit.
		 */
		if (excess || mz->on_tree) {
745 746 747
			unsigned long flags;

			spin_lock_irqsave(&mctz->lock, flags);
748 749
			/* if on-tree, remove it */
			if (mz->on_tree)
750
				__mem_cgroup_remove_exceeded(mz, mctz);
751 752 753 754
			/*
			 * Insert again. mz->usage_in_excess will be updated.
			 * If excess is 0, no tree ops.
			 */
755
			__mem_cgroup_insert_exceeded(mz, mctz, excess);
756
			spin_unlock_irqrestore(&mctz->lock, flags);
757 758 759 760 761 762 763
		}
	}
}

static void mem_cgroup_remove_from_trees(struct mem_cgroup *memcg)
{
	struct mem_cgroup_tree_per_zone *mctz;
764 765
	struct mem_cgroup_per_zone *mz;
	int nid, zid;
766

767 768 769 770
	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);
771
			mem_cgroup_remove_exceeded(mz, mctz);
772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793
		}
	}
}

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.
	 */
794
	__mem_cgroup_remove_exceeded(mz, mctz);
795
	if (!soft_limit_excess(mz->memcg) ||
796
	    !css_tryget_online(&mz->memcg->css))
797 798 799 800 801 802 803 804 805 806
		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;

807
	spin_lock_irq(&mctz->lock);
808
	mz = __mem_cgroup_largest_soft_limit_node(mctz);
809
	spin_unlock_irq(&mctz->lock);
810 811 812
	return mz;
}

813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831
/*
 * 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.
 */
832
static long mem_cgroup_read_stat(struct mem_cgroup *memcg,
833
				 enum mem_cgroup_stat_index idx)
834
{
835
	long val = 0;
836 837
	int cpu;

838 839
	get_online_cpus();
	for_each_online_cpu(cpu)
840
		val += per_cpu(memcg->stat->count[idx], cpu);
841
#ifdef CONFIG_HOTPLUG_CPU
842 843 844
	spin_lock(&memcg->pcp_counter_lock);
	val += memcg->nocpu_base.count[idx];
	spin_unlock(&memcg->pcp_counter_lock);
845 846
#endif
	put_online_cpus();
847 848 849
	return val;
}

850
static unsigned long mem_cgroup_read_events(struct mem_cgroup *memcg,
851 852 853 854 855
					    enum mem_cgroup_events_index idx)
{
	unsigned long val = 0;
	int cpu;

856
	get_online_cpus();
857
	for_each_online_cpu(cpu)
858
		val += per_cpu(memcg->stat->events[idx], cpu);
859
#ifdef CONFIG_HOTPLUG_CPU
860 861 862
	spin_lock(&memcg->pcp_counter_lock);
	val += memcg->nocpu_base.events[idx];
	spin_unlock(&memcg->pcp_counter_lock);
863
#endif
864
	put_online_cpus();
865 866 867
	return val;
}

868
static void mem_cgroup_charge_statistics(struct mem_cgroup *memcg,
869
					 struct page *page,
870
					 int nr_pages)
871
{
872 873 874 875
	/*
	 * Here, RSS means 'mapped anon' and anon's SwapCache. Shmem/tmpfs is
	 * counted as CACHE even if it's on ANON LRU.
	 */
876
	if (PageAnon(page))
877
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_RSS],
878
				nr_pages);
879
	else
880
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_CACHE],
881
				nr_pages);
882

883 884 885 886
	if (PageTransHuge(page))
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_RSS_HUGE],
				nr_pages);

887 888
	/* pagein of a big page is an event. So, ignore page size */
	if (nr_pages > 0)
889
		__this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGPGIN]);
890
	else {
891
		__this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGPGOUT]);
892 893
		nr_pages = -nr_pages; /* for event */
	}
894

895
	__this_cpu_add(memcg->stat->nr_page_events, nr_pages);
896 897
}

898
unsigned long mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
899 900 901 902 903 904 905
{
	struct mem_cgroup_per_zone *mz;

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

906 907 908
static unsigned long mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
						  int nid,
						  unsigned int lru_mask)
909
{
910
	unsigned long nr = 0;
911 912
	int zid;

913
	VM_BUG_ON((unsigned)nid >= nr_node_ids);
914

915 916 917 918 919 920 921 922 923 924 925 926
	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;
927
}
928

929
static unsigned long mem_cgroup_nr_lru_pages(struct mem_cgroup *memcg,
930
			unsigned int lru_mask)
931
{
932
	unsigned long nr = 0;
933
	int nid;
934

935
	for_each_node_state(nid, N_MEMORY)
936 937
		nr += mem_cgroup_node_nr_lru_pages(memcg, nid, lru_mask);
	return nr;
938 939
}

940 941
static bool mem_cgroup_event_ratelimit(struct mem_cgroup *memcg,
				       enum mem_cgroup_events_target target)
942 943 944
{
	unsigned long val, next;

945
	val = __this_cpu_read(memcg->stat->nr_page_events);
946
	next = __this_cpu_read(memcg->stat->targets[target]);
947
	/* from time_after() in jiffies.h */
948 949 950 951 952
	if ((long)next - (long)val < 0) {
		switch (target) {
		case MEM_CGROUP_TARGET_THRESH:
			next = val + THRESHOLDS_EVENTS_TARGET;
			break;
953 954 955
		case MEM_CGROUP_TARGET_SOFTLIMIT:
			next = val + SOFTLIMIT_EVENTS_TARGET;
			break;
956 957 958 959 960 961 962 963
		case MEM_CGROUP_TARGET_NUMAINFO:
			next = val + NUMAINFO_EVENTS_TARGET;
			break;
		default:
			break;
		}
		__this_cpu_write(memcg->stat->targets[target], next);
		return true;
964
	}
965
	return false;
966 967 968 969 970 971
}

/*
 * Check events in order.
 *
 */
972
static void memcg_check_events(struct mem_cgroup *memcg, struct page *page)
973 974
{
	/* threshold event is triggered in finer grain than soft limit */
975 976
	if (unlikely(mem_cgroup_event_ratelimit(memcg,
						MEM_CGROUP_TARGET_THRESH))) {
977
		bool do_softlimit;
978
		bool do_numainfo __maybe_unused;
979

980 981
		do_softlimit = mem_cgroup_event_ratelimit(memcg,
						MEM_CGROUP_TARGET_SOFTLIMIT);
982 983 984 985
#if MAX_NUMNODES > 1
		do_numainfo = mem_cgroup_event_ratelimit(memcg,
						MEM_CGROUP_TARGET_NUMAINFO);
#endif
986
		mem_cgroup_threshold(memcg);
987 988
		if (unlikely(do_softlimit))
			mem_cgroup_update_tree(memcg, page);
989
#if MAX_NUMNODES > 1
990
		if (unlikely(do_numainfo))
991
			atomic_inc(&memcg->numainfo_events);
992
#endif
993
	}
994 995
}

996
struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p)
997
{
998 999 1000 1001 1002 1003 1004 1005
	/*
	 * 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;

1006
	return mem_cgroup_from_css(task_css(p, memory_cgrp_id));
1007 1008
}

1009
static struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
1010
{
1011
	struct mem_cgroup *memcg = NULL;
1012

1013 1014
	rcu_read_lock();
	do {
1015 1016 1017 1018 1019 1020
		/*
		 * 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))
1021
			memcg = root_mem_cgroup;
1022 1023 1024 1025 1026
		else {
			memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
			if (unlikely(!memcg))
				memcg = root_mem_cgroup;
		}
1027
	} while (!css_tryget_online(&memcg->css));
1028
	rcu_read_unlock();
1029
	return memcg;
1030 1031
}

1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048
/**
 * 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.
 */
1049
struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *root,
1050
				   struct mem_cgroup *prev,
1051
				   struct mem_cgroup_reclaim_cookie *reclaim)
K
KAMEZAWA Hiroyuki 已提交
1052
{
1053 1054
	struct reclaim_iter *uninitialized_var(iter);
	struct cgroup_subsys_state *css = NULL;
1055
	struct mem_cgroup *memcg = NULL;
1056
	struct mem_cgroup *pos = NULL;
1057

1058 1059
	if (mem_cgroup_disabled())
		return NULL;
1060

1061 1062
	if (!root)
		root = root_mem_cgroup;
K
KAMEZAWA Hiroyuki 已提交
1063

1064
	if (prev && !reclaim)
1065
		pos = prev;
K
KAMEZAWA Hiroyuki 已提交
1066

1067 1068
	if (!root->use_hierarchy && root != root_mem_cgroup) {
		if (prev)
1069
			goto out;
1070
		return root;
1071
	}
K
KAMEZAWA Hiroyuki 已提交
1072

1073
	rcu_read_lock();
M
Michal Hocko 已提交
1074

1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108
	if (reclaim) {
		struct mem_cgroup_per_zone *mz;

		mz = mem_cgroup_zone_zoneinfo(root, reclaim->zone);
		iter = &mz->iter[reclaim->priority];

		if (prev && reclaim->generation != iter->generation)
			goto out_unlock;

		do {
			pos = ACCESS_ONCE(iter->position);
			/*
			 * A racing update may change the position and
			 * put the last reference, hence css_tryget(),
			 * or retry to see the updated position.
			 */
		} while (pos && !css_tryget(&pos->css));
	}

	if (pos)
		css = &pos->css;

	for (;;) {
		css = css_next_descendant_pre(css, &root->css);
		if (!css) {
			/*
			 * Reclaimers share the hierarchy walk, and a
			 * new one might jump in right at the end of
			 * the hierarchy - make sure they see at least
			 * one group and restart from the beginning.
			 */
			if (!prev)
				continue;
			break;
1109
		}
K
KAMEZAWA Hiroyuki 已提交
1110

1111 1112 1113 1114 1115 1116
		/*
		 * Verify the css and acquire a reference.  The root
		 * is provided by the caller, so we know it's alive
		 * and kicking, and don't take an extra reference.
		 */
		memcg = mem_cgroup_from_css(css);
K
KAMEZAWA Hiroyuki 已提交
1117

1118 1119
		if (css == &root->css)
			break;
K
KAMEZAWA Hiroyuki 已提交
1120

1121
		if (css_tryget(css)) {
1122 1123 1124 1125 1126 1127 1128
			/*
			 * Make sure the memcg is initialized:
			 * mem_cgroup_css_online() orders the the
			 * initialization against setting the flag.
			 */
			if (smp_load_acquire(&memcg->initialized))
				break;
1129

1130
			css_put(css);
1131
		}
1132

1133
		memcg = NULL;
1134
	}
1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154

	if (reclaim) {
		if (cmpxchg(&iter->position, pos, memcg) == pos) {
			if (memcg)
				css_get(&memcg->css);
			if (pos)
				css_put(&pos->css);
		}

		/*
		 * pairs with css_tryget when dereferencing iter->position
		 * above.
		 */
		if (pos)
			css_put(&pos->css);

		if (!memcg)
			iter->generation++;
		else if (!prev)
			reclaim->generation = iter->generation;
1155
	}
1156

1157 1158
out_unlock:
	rcu_read_unlock();
1159
out:
1160 1161 1162
	if (prev && prev != root)
		css_put(&prev->css);

1163
	return memcg;
K
KAMEZAWA Hiroyuki 已提交
1164
}
K
KAMEZAWA Hiroyuki 已提交
1165

1166 1167 1168 1169 1170 1171 1172
/**
 * 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)
1173 1174 1175 1176 1177 1178
{
	if (!root)
		root = root_mem_cgroup;
	if (prev && prev != root)
		css_put(&prev->css);
}
K
KAMEZAWA Hiroyuki 已提交
1179

1180 1181 1182 1183 1184 1185
/*
 * 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)		\
1186
	for (iter = mem_cgroup_iter(root, NULL, NULL);	\
1187
	     iter != NULL;				\
1188
	     iter = mem_cgroup_iter(root, iter, NULL))
1189

1190
#define for_each_mem_cgroup(iter)			\
1191
	for (iter = mem_cgroup_iter(NULL, NULL, NULL);	\
1192
	     iter != NULL;				\
1193
	     iter = mem_cgroup_iter(NULL, iter, NULL))
K
KAMEZAWA Hiroyuki 已提交
1194

1195
void __mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
1196
{
1197
	struct mem_cgroup *memcg;
1198 1199

	rcu_read_lock();
1200 1201
	memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
	if (unlikely(!memcg))
1202 1203 1204 1205
		goto out;

	switch (idx) {
	case PGFAULT:
1206 1207 1208 1209
		this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]);
		break;
	case PGMAJFAULT:
		this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]);
1210 1211 1212 1213 1214 1215 1216
		break;
	default:
		BUG();
	}
out:
	rcu_read_unlock();
}
1217
EXPORT_SYMBOL(__mem_cgroup_count_vm_event);
1218

1219 1220 1221
/**
 * mem_cgroup_zone_lruvec - get the lru list vector for a zone and memcg
 * @zone: zone of the wanted lruvec
1222
 * @memcg: memcg of the wanted lruvec
1223 1224 1225 1226 1227 1228 1229 1230 1231
 *
 * 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;
1232
	struct lruvec *lruvec;
1233

1234 1235 1236 1237
	if (mem_cgroup_disabled()) {
		lruvec = &zone->lruvec;
		goto out;
	}
1238

1239
	mz = mem_cgroup_zone_zoneinfo(memcg, zone);
1240 1241 1242 1243 1244 1245 1246 1247 1248 1249
	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;
1250 1251 1252
}

/**
1253
 * mem_cgroup_page_lruvec - return lruvec for isolating/putting an LRU page
1254
 * @page: the page
1255
 * @zone: zone of the page
1256 1257 1258 1259
 *
 * This function is only safe when following the LRU page isolation
 * and putback protocol: the LRU lock must be held, and the page must
 * either be PageLRU() or the caller must have isolated/allocated it.
1260
 */
1261
struct lruvec *mem_cgroup_page_lruvec(struct page *page, struct zone *zone)
K
KAMEZAWA Hiroyuki 已提交
1262 1263
{
	struct mem_cgroup_per_zone *mz;
1264
	struct mem_cgroup *memcg;
1265
	struct lruvec *lruvec;
1266

1267 1268 1269 1270
	if (mem_cgroup_disabled()) {
		lruvec = &zone->lruvec;
		goto out;
	}
1271

1272
	memcg = page->mem_cgroup;
1273
	/*
1274
	 * Swapcache readahead pages are added to the LRU - and
1275
	 * possibly migrated - before they are charged.
1276
	 */
1277 1278
	if (!memcg)
		memcg = root_mem_cgroup;
1279

1280
	mz = mem_cgroup_page_zoneinfo(memcg, page);
1281 1282 1283 1284 1285 1286 1287 1288 1289 1290
	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
1291
}
1292

1293
/**
1294 1295 1296 1297
 * 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
1298
 *
1299 1300
 * This function must be called when a page is added to or removed from an
 * lru list.
1301
 */
1302 1303
void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
				int nr_pages)
1304 1305
{
	struct mem_cgroup_per_zone *mz;
1306
	unsigned long *lru_size;
1307 1308 1309 1310

	if (mem_cgroup_disabled())
		return;

1311 1312 1313 1314
	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
1315
}
1316

1317
bool mem_cgroup_is_descendant(struct mem_cgroup *memcg, struct mem_cgroup *root)
1318
{
1319
	if (root == memcg)
1320
		return true;
1321
	if (!root->use_hierarchy)
1322
		return false;
1323
	return cgroup_is_descendant(memcg->css.cgroup, root->css.cgroup);
1324 1325
}

1326
bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg)
1327
{
1328
	struct mem_cgroup *task_memcg;
1329
	struct task_struct *p;
1330
	bool ret;
1331

1332
	p = find_lock_task_mm(task);
1333
	if (p) {
1334
		task_memcg = get_mem_cgroup_from_mm(p->mm);
1335 1336 1337 1338 1339 1340 1341
		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.
		 */
1342
		rcu_read_lock();
1343 1344
		task_memcg = mem_cgroup_from_task(task);
		css_get(&task_memcg->css);
1345
		rcu_read_unlock();
1346
	}
1347 1348
	ret = mem_cgroup_is_descendant(task_memcg, memcg);
	css_put(&task_memcg->css);
1349 1350 1351
	return ret;
}

1352
int mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
1353
{
1354
	unsigned long inactive_ratio;
1355
	unsigned long inactive;
1356
	unsigned long active;
1357
	unsigned long gb;
1358

1359 1360
	inactive = mem_cgroup_get_lru_size(lruvec, LRU_INACTIVE_ANON);
	active = mem_cgroup_get_lru_size(lruvec, LRU_ACTIVE_ANON);
1361

1362 1363 1364 1365 1366 1367
	gb = (inactive + active) >> (30 - PAGE_SHIFT);
	if (gb)
		inactive_ratio = int_sqrt(10 * gb);
	else
		inactive_ratio = 1;

1368
	return inactive * inactive_ratio < active;
1369 1370
}

1371
#define mem_cgroup_from_counter(counter, member)	\
1372 1373
	container_of(counter, struct mem_cgroup, member)

1374
/**
1375
 * mem_cgroup_margin - calculate chargeable space of a memory cgroup
W
Wanpeng Li 已提交
1376
 * @memcg: the memory cgroup
1377
 *
1378
 * Returns the maximum amount of memory @mem can be charged with, in
1379
 * pages.
1380
 */
1381
static unsigned long mem_cgroup_margin(struct mem_cgroup *memcg)
1382
{
1383 1384 1385
	unsigned long margin = 0;
	unsigned long count;
	unsigned long limit;
1386

1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399
	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;
1400 1401
}

1402
int mem_cgroup_swappiness(struct mem_cgroup *memcg)
K
KOSAKI Motohiro 已提交
1403 1404
{
	/* root ? */
1405
	if (mem_cgroup_disabled() || !memcg->css.parent)
K
KOSAKI Motohiro 已提交
1406 1407
		return vm_swappiness;

1408
	return memcg->swappiness;
K
KOSAKI Motohiro 已提交
1409 1410
}

1411
/*
Q
Qiang Huang 已提交
1412
 * A routine for checking "mem" is under move_account() or not.
1413
 *
Q
Qiang Huang 已提交
1414 1415 1416
 * 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".
1417
 */
1418
static bool mem_cgroup_under_move(struct mem_cgroup *memcg)
1419
{
1420 1421
	struct mem_cgroup *from;
	struct mem_cgroup *to;
1422
	bool ret = false;
1423 1424 1425 1426 1427 1428 1429 1430 1431
	/*
	 * 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;
1432

1433 1434
	ret = mem_cgroup_is_descendant(from, memcg) ||
		mem_cgroup_is_descendant(to, memcg);
1435 1436
unlock:
	spin_unlock(&mc.lock);
1437 1438 1439
	return ret;
}

1440
static bool mem_cgroup_wait_acct_move(struct mem_cgroup *memcg)
1441 1442
{
	if (mc.moving_task && current != mc.moving_task) {
1443
		if (mem_cgroup_under_move(memcg)) {
1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455
			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;
}

1456
#define K(x) ((x) << (PAGE_SHIFT-10))
1457
/**
1458
 * mem_cgroup_print_oom_info: Print OOM information relevant to memory controller.
1459 1460 1461 1462 1463 1464 1465 1466
 * @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
1467
	/* oom_info_lock ensures that parallel ooms do not interleave */
1468
	static DEFINE_MUTEX(oom_info_lock);
1469 1470
	struct mem_cgroup *iter;
	unsigned int i;
1471

1472
	if (!p)
1473 1474
		return;

1475
	mutex_lock(&oom_info_lock);
1476 1477
	rcu_read_lock();

T
Tejun Heo 已提交
1478 1479 1480 1481 1482
	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");
1483 1484 1485

	rcu_read_unlock();

1486 1487 1488 1489 1490 1491 1492 1493 1494
	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);
1495 1496

	for_each_mem_cgroup_tree(iter, memcg) {
T
Tejun Heo 已提交
1497 1498
		pr_info("Memory cgroup stats for ");
		pr_cont_cgroup_path(iter->css.cgroup);
1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513
		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");
	}
1514
	mutex_unlock(&oom_info_lock);
1515 1516
}

1517 1518 1519 1520
/*
 * This function returns the number of memcg under hierarchy tree. Returns
 * 1(self count) if no children.
 */
1521
static int mem_cgroup_count_children(struct mem_cgroup *memcg)
1522 1523
{
	int num = 0;
K
KAMEZAWA Hiroyuki 已提交
1524 1525
	struct mem_cgroup *iter;

1526
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
1527
		num++;
1528 1529 1530
	return num;
}

D
David Rientjes 已提交
1531 1532 1533
/*
 * Return the memory (and swap, if configured) limit for a memcg.
 */
1534
static unsigned long mem_cgroup_get_limit(struct mem_cgroup *memcg)
D
David Rientjes 已提交
1535
{
1536
	unsigned long limit;
1537

1538
	limit = memcg->memory.limit;
1539
	if (mem_cgroup_swappiness(memcg)) {
1540
		unsigned long memsw_limit;
1541

1542 1543
		memsw_limit = memcg->memsw.limit;
		limit = min(limit + total_swap_pages, memsw_limit);
1544 1545
	}
	return limit;
D
David Rientjes 已提交
1546 1547
}

1548 1549
static void mem_cgroup_out_of_memory(struct mem_cgroup *memcg, gfp_t gfp_mask,
				     int order)
1550 1551 1552 1553 1554 1555 1556
{
	struct mem_cgroup *iter;
	unsigned long chosen_points = 0;
	unsigned long totalpages;
	unsigned int points = 0;
	struct task_struct *chosen = NULL;

1557
	/*
1558 1559 1560
	 * 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.
1561
	 */
1562
	if (fatal_signal_pending(current) || task_will_free_mem(current)) {
1563 1564 1565 1566 1567
		set_thread_flag(TIF_MEMDIE);
		return;
	}

	check_panic_on_oom(CONSTRAINT_MEMCG, gfp_mask, order, NULL);
1568
	totalpages = mem_cgroup_get_limit(memcg) ? : 1;
1569
	for_each_mem_cgroup_tree(iter, memcg) {
1570
		struct css_task_iter it;
1571 1572
		struct task_struct *task;

1573 1574
		css_task_iter_start(&iter->css, &it);
		while ((task = css_task_iter_next(&it))) {
1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586
			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:
1587
				css_task_iter_end(&it);
1588 1589 1590 1591 1592 1593 1594 1595
				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);
1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607
			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);
1608
		}
1609
		css_task_iter_end(&it);
1610 1611 1612 1613 1614 1615 1616 1617 1618
	}

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

1619 1620
/**
 * test_mem_cgroup_node_reclaimable
W
Wanpeng Li 已提交
1621
 * @memcg: the target memcg
1622 1623 1624 1625 1626 1627 1628
 * @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.
 */
1629
static bool test_mem_cgroup_node_reclaimable(struct mem_cgroup *memcg,
1630 1631
		int nid, bool noswap)
{
1632
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_FILE))
1633 1634 1635
		return true;
	if (noswap || !total_swap_pages)
		return false;
1636
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_ANON))
1637 1638 1639 1640
		return true;
	return false;

}
1641
#if MAX_NUMNODES > 1
1642 1643 1644 1645 1646 1647 1648

/*
 * 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.
 *
 */
1649
static void mem_cgroup_may_update_nodemask(struct mem_cgroup *memcg)
1650 1651
{
	int nid;
1652 1653 1654 1655
	/*
	 * numainfo_events > 0 means there was at least NUMAINFO_EVENTS_TARGET
	 * pagein/pageout changes since the last update.
	 */
1656
	if (!atomic_read(&memcg->numainfo_events))
1657
		return;
1658
	if (atomic_inc_return(&memcg->numainfo_updating) > 1)
1659 1660 1661
		return;

	/* make a nodemask where this memcg uses memory from */
1662
	memcg->scan_nodes = node_states[N_MEMORY];
1663

1664
	for_each_node_mask(nid, node_states[N_MEMORY]) {
1665

1666 1667
		if (!test_mem_cgroup_node_reclaimable(memcg, nid, false))
			node_clear(nid, memcg->scan_nodes);
1668
	}
1669

1670 1671
	atomic_set(&memcg->numainfo_events, 0);
	atomic_set(&memcg->numainfo_updating, 0);
1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685
}

/*
 * 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.
 */
1686
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
1687 1688 1689
{
	int node;

1690 1691
	mem_cgroup_may_update_nodemask(memcg);
	node = memcg->last_scanned_node;
1692

1693
	node = next_node(node, memcg->scan_nodes);
1694
	if (node == MAX_NUMNODES)
1695
		node = first_node(memcg->scan_nodes);
1696 1697 1698 1699 1700 1701 1702 1703 1704
	/*
	 * 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();

1705
	memcg->last_scanned_node = node;
1706 1707 1708
	return node;
}
#else
1709
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
1710 1711 1712 1713 1714
{
	return 0;
}
#endif

1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729
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,
	};

1730
	excess = soft_limit_excess(root_memcg);
1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758

	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;
		}
		total += mem_cgroup_shrink_node_zone(victim, gfp_mask, false,
						     zone, &nr_scanned);
		*total_scanned += nr_scanned;
1759
		if (!soft_limit_excess(root_memcg))
1760
			break;
1761
	}
1762 1763
	mem_cgroup_iter_break(root_memcg, victim);
	return total;
1764 1765
}

1766 1767 1768 1769 1770 1771
#ifdef CONFIG_LOCKDEP
static struct lockdep_map memcg_oom_lock_dep_map = {
	.name = "memcg_oom_lock",
};
#endif

1772 1773
static DEFINE_SPINLOCK(memcg_oom_lock);

K
KAMEZAWA Hiroyuki 已提交
1774 1775 1776 1777
/*
 * Check OOM-Killer is already running under our hierarchy.
 * If someone is running, return false.
 */
1778
static bool mem_cgroup_oom_trylock(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
1779
{
1780
	struct mem_cgroup *iter, *failed = NULL;
1781

1782 1783
	spin_lock(&memcg_oom_lock);

1784
	for_each_mem_cgroup_tree(iter, memcg) {
1785
		if (iter->oom_lock) {
1786 1787 1788 1789 1790
			/*
			 * this subtree of our hierarchy is already locked
			 * so we cannot give a lock.
			 */
			failed = iter;
1791 1792
			mem_cgroup_iter_break(memcg, iter);
			break;
1793 1794
		} else
			iter->oom_lock = true;
K
KAMEZAWA Hiroyuki 已提交
1795
	}
K
KAMEZAWA Hiroyuki 已提交
1796

1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807
	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;
1808
		}
1809 1810
	} else
		mutex_acquire(&memcg_oom_lock_dep_map, 0, 1, _RET_IP_);
1811 1812 1813 1814

	spin_unlock(&memcg_oom_lock);

	return !failed;
1815
}
1816

1817
static void mem_cgroup_oom_unlock(struct mem_cgroup *memcg)
1818
{
K
KAMEZAWA Hiroyuki 已提交
1819 1820
	struct mem_cgroup *iter;

1821
	spin_lock(&memcg_oom_lock);
1822
	mutex_release(&memcg_oom_lock_dep_map, 1, _RET_IP_);
1823
	for_each_mem_cgroup_tree(iter, memcg)
1824
		iter->oom_lock = false;
1825
	spin_unlock(&memcg_oom_lock);
1826 1827
}

1828
static void mem_cgroup_mark_under_oom(struct mem_cgroup *memcg)
1829 1830 1831
{
	struct mem_cgroup *iter;

1832
	for_each_mem_cgroup_tree(iter, memcg)
1833 1834 1835
		atomic_inc(&iter->under_oom);
}

1836
static void mem_cgroup_unmark_under_oom(struct mem_cgroup *memcg)
1837 1838 1839
{
	struct mem_cgroup *iter;

K
KAMEZAWA Hiroyuki 已提交
1840 1841 1842 1843 1844
	/*
	 * 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.
	 */
1845
	for_each_mem_cgroup_tree(iter, memcg)
1846
		atomic_add_unless(&iter->under_oom, -1, 0);
1847 1848
}

K
KAMEZAWA Hiroyuki 已提交
1849 1850
static DECLARE_WAIT_QUEUE_HEAD(memcg_oom_waitq);

K
KAMEZAWA Hiroyuki 已提交
1851
struct oom_wait_info {
1852
	struct mem_cgroup *memcg;
K
KAMEZAWA Hiroyuki 已提交
1853 1854 1855 1856 1857 1858
	wait_queue_t	wait;
};

static int memcg_oom_wake_function(wait_queue_t *wait,
	unsigned mode, int sync, void *arg)
{
1859 1860
	struct mem_cgroup *wake_memcg = (struct mem_cgroup *)arg;
	struct mem_cgroup *oom_wait_memcg;
K
KAMEZAWA Hiroyuki 已提交
1861 1862 1863
	struct oom_wait_info *oom_wait_info;

	oom_wait_info = container_of(wait, struct oom_wait_info, wait);
1864
	oom_wait_memcg = oom_wait_info->memcg;
K
KAMEZAWA Hiroyuki 已提交
1865

1866 1867
	if (!mem_cgroup_is_descendant(wake_memcg, oom_wait_memcg) &&
	    !mem_cgroup_is_descendant(oom_wait_memcg, wake_memcg))
K
KAMEZAWA Hiroyuki 已提交
1868 1869 1870 1871
		return 0;
	return autoremove_wake_function(wait, mode, sync, arg);
}

1872
static void memcg_wakeup_oom(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
1873
{
1874
	atomic_inc(&memcg->oom_wakeups);
1875 1876
	/* for filtering, pass "memcg" as argument. */
	__wake_up(&memcg_oom_waitq, TASK_NORMAL, 0, memcg);
K
KAMEZAWA Hiroyuki 已提交
1877 1878
}

1879
static void memcg_oom_recover(struct mem_cgroup *memcg)
1880
{
1881 1882
	if (memcg && atomic_read(&memcg->under_oom))
		memcg_wakeup_oom(memcg);
1883 1884
}

1885
static void mem_cgroup_oom(struct mem_cgroup *memcg, gfp_t mask, int order)
1886
{
1887 1888
	if (!current->memcg_oom.may_oom)
		return;
K
KAMEZAWA Hiroyuki 已提交
1889
	/*
1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901
	 * 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
1902
	 */
1903 1904 1905 1906
	css_get(&memcg->css);
	current->memcg_oom.memcg = memcg;
	current->memcg_oom.gfp_mask = mask;
	current->memcg_oom.order = order;
1907 1908 1909 1910
}

/**
 * mem_cgroup_oom_synchronize - complete memcg OOM handling
1911
 * @handle: actually kill/wait or just clean up the OOM state
1912
 *
1913 1914
 * This has to be called at the end of a page fault if the memcg OOM
 * handler was enabled.
1915
 *
1916
 * Memcg supports userspace OOM handling where failed allocations must
1917 1918 1919 1920
 * 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
1921
 * the end of the page fault to complete the OOM handling.
1922 1923
 *
 * Returns %true if an ongoing memcg OOM situation was detected and
1924
 * completed, %false otherwise.
1925
 */
1926
bool mem_cgroup_oom_synchronize(bool handle)
1927
{
1928
	struct mem_cgroup *memcg = current->memcg_oom.memcg;
1929
	struct oom_wait_info owait;
1930
	bool locked;
1931 1932 1933

	/* OOM is global, do not handle */
	if (!memcg)
1934
		return false;
1935

1936 1937
	if (!handle)
		goto cleanup;
1938 1939 1940 1941 1942 1943

	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
1944

1945
	prepare_to_wait(&memcg_oom_waitq, &owait.wait, TASK_KILLABLE);
1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958
	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 {
1959
		schedule();
1960 1961 1962 1963 1964
		mem_cgroup_unmark_under_oom(memcg);
		finish_wait(&memcg_oom_waitq, &owait.wait);
	}

	if (locked) {
1965 1966 1967 1968 1969 1970 1971 1972
		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);
	}
1973 1974
cleanup:
	current->memcg_oom.memcg = NULL;
1975
	css_put(&memcg->css);
K
KAMEZAWA Hiroyuki 已提交
1976
	return true;
1977 1978
}

1979 1980 1981 1982 1983
/**
 * 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
1984
 *
1985 1986 1987
 * 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:
1988
 *
1989 1990 1991 1992
 *   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);
1993
 *
1994 1995 1996
 * 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.
1997
 *
1998 1999 2000 2001 2002
 * 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.
2003
 */
2004 2005 2006
struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page,
					      bool *locked,
					      unsigned long *flags)
2007 2008 2009
{
	struct mem_cgroup *memcg;

2010 2011 2012 2013
	rcu_read_lock();

	if (mem_cgroup_disabled())
		return NULL;
2014
again:
2015
	memcg = page->mem_cgroup;
2016
	if (unlikely(!memcg))
2017 2018 2019
		return NULL;

	*locked = false;
Q
Qiang Huang 已提交
2020
	if (atomic_read(&memcg->moving_account) <= 0)
2021
		return memcg;
2022

2023
	spin_lock_irqsave(&memcg->move_lock, *flags);
2024
	if (memcg != page->mem_cgroup) {
2025
		spin_unlock_irqrestore(&memcg->move_lock, *flags);
2026 2027 2028
		goto again;
	}
	*locked = true;
2029 2030

	return memcg;
2031 2032
}

2033 2034 2035 2036 2037 2038
/**
 * 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()
 */
2039 2040
void mem_cgroup_end_page_stat(struct mem_cgroup *memcg, bool *locked,
			      unsigned long *flags)
2041
{
2042 2043
	if (memcg && *locked)
		spin_unlock_irqrestore(&memcg->move_lock, *flags);
2044

2045
	rcu_read_unlock();
2046 2047
}

2048 2049 2050 2051 2052 2053 2054 2055 2056
/**
 * 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
2057
				 enum mem_cgroup_stat_index idx, int val)
2058
{
2059
	VM_BUG_ON(!rcu_read_lock_held());
2060

2061 2062
	if (memcg)
		this_cpu_add(memcg->stat->count[idx], val);
2063
}
2064

2065 2066 2067 2068
/*
 * size of first charge trial. "32" comes from vmscan.c's magic value.
 * TODO: maybe necessary to use big numbers in big irons.
 */
2069
#define CHARGE_BATCH	32U
2070 2071
struct memcg_stock_pcp {
	struct mem_cgroup *cached; /* this never be root cgroup */
2072
	unsigned int nr_pages;
2073
	struct work_struct work;
2074
	unsigned long flags;
2075
#define FLUSHING_CACHED_CHARGE	0
2076 2077
};
static DEFINE_PER_CPU(struct memcg_stock_pcp, memcg_stock);
2078
static DEFINE_MUTEX(percpu_charge_mutex);
2079

2080 2081 2082 2083 2084 2085 2086 2087 2088 2089
/**
 * 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.
2090
 */
2091
static bool consume_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
2092 2093
{
	struct memcg_stock_pcp *stock;
2094
	bool ret = false;
2095

2096
	if (nr_pages > CHARGE_BATCH)
2097
		return ret;
2098

2099
	stock = &get_cpu_var(memcg_stock);
2100
	if (memcg == stock->cached && stock->nr_pages >= nr_pages) {
2101
		stock->nr_pages -= nr_pages;
2102 2103
		ret = true;
	}
2104 2105 2106 2107 2108
	put_cpu_var(memcg_stock);
	return ret;
}

/*
2109
 * Returns stocks cached in percpu and reset cached information.
2110 2111 2112 2113 2114
 */
static void drain_stock(struct memcg_stock_pcp *stock)
{
	struct mem_cgroup *old = stock->cached;

2115
	if (stock->nr_pages) {
2116
		page_counter_uncharge(&old->memory, stock->nr_pages);
2117
		if (do_swap_account)
2118
			page_counter_uncharge(&old->memsw, stock->nr_pages);
2119
		css_put_many(&old->css, stock->nr_pages);
2120
		stock->nr_pages = 0;
2121 2122 2123 2124 2125 2126 2127 2128 2129 2130
	}
	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)
{
2131
	struct memcg_stock_pcp *stock = this_cpu_ptr(&memcg_stock);
2132
	drain_stock(stock);
2133
	clear_bit(FLUSHING_CACHED_CHARGE, &stock->flags);
2134 2135
}

2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146
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);
	}
}

2147
/*
2148
 * Cache charges(val) to local per_cpu area.
2149
 * This will be consumed by consume_stock() function, later.
2150
 */
2151
static void refill_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
2152 2153 2154
{
	struct memcg_stock_pcp *stock = &get_cpu_var(memcg_stock);

2155
	if (stock->cached != memcg) { /* reset if necessary */
2156
		drain_stock(stock);
2157
		stock->cached = memcg;
2158
	}
2159
	stock->nr_pages += nr_pages;
2160 2161 2162 2163
	put_cpu_var(memcg_stock);
}

/*
2164
 * Drains all per-CPU charge caches for given root_memcg resp. subtree
2165
 * of the hierarchy under it.
2166
 */
2167
static void drain_all_stock(struct mem_cgroup *root_memcg)
2168
{
2169
	int cpu, curcpu;
2170

2171 2172 2173
	/* If someone's already draining, avoid adding running more workers. */
	if (!mutex_trylock(&percpu_charge_mutex))
		return;
2174 2175
	/* Notify other cpus that system-wide "drain" is running */
	get_online_cpus();
2176
	curcpu = get_cpu();
2177 2178
	for_each_online_cpu(cpu) {
		struct memcg_stock_pcp *stock = &per_cpu(memcg_stock, cpu);
2179
		struct mem_cgroup *memcg;
2180

2181 2182
		memcg = stock->cached;
		if (!memcg || !stock->nr_pages)
2183
			continue;
2184
		if (!mem_cgroup_is_descendant(memcg, root_memcg))
2185
			continue;
2186 2187 2188 2189 2190 2191
		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);
		}
2192
	}
2193
	put_cpu();
A
Andrew Morton 已提交
2194
	put_online_cpus();
2195
	mutex_unlock(&percpu_charge_mutex);
2196 2197
}

2198 2199 2200 2201
/*
 * This function drains percpu counter value from DEAD cpu and
 * move it to local cpu. Note that this function can be preempted.
 */
2202
static void mem_cgroup_drain_pcp_counter(struct mem_cgroup *memcg, int cpu)
2203 2204 2205
{
	int i;

2206
	spin_lock(&memcg->pcp_counter_lock);
2207
	for (i = 0; i < MEM_CGROUP_STAT_NSTATS; i++) {
2208
		long x = per_cpu(memcg->stat->count[i], cpu);
2209

2210 2211
		per_cpu(memcg->stat->count[i], cpu) = 0;
		memcg->nocpu_base.count[i] += x;
2212
	}
2213
	for (i = 0; i < MEM_CGROUP_EVENTS_NSTATS; i++) {
2214
		unsigned long x = per_cpu(memcg->stat->events[i], cpu);
2215

2216 2217
		per_cpu(memcg->stat->events[i], cpu) = 0;
		memcg->nocpu_base.events[i] += x;
2218
	}
2219
	spin_unlock(&memcg->pcp_counter_lock);
2220 2221
}

2222
static int memcg_cpu_hotplug_callback(struct notifier_block *nb,
2223 2224 2225 2226 2227
					unsigned long action,
					void *hcpu)
{
	int cpu = (unsigned long)hcpu;
	struct memcg_stock_pcp *stock;
2228
	struct mem_cgroup *iter;
2229

2230
	if (action == CPU_ONLINE)
2231 2232
		return NOTIFY_OK;

2233
	if (action != CPU_DEAD && action != CPU_DEAD_FROZEN)
2234
		return NOTIFY_OK;
2235

2236
	for_each_mem_cgroup(iter)
2237 2238
		mem_cgroup_drain_pcp_counter(iter, cpu);

2239 2240 2241 2242 2243
	stock = &per_cpu(memcg_stock, cpu);
	drain_stock(stock);
	return NOTIFY_OK;
}

2244 2245
static int try_charge(struct mem_cgroup *memcg, gfp_t gfp_mask,
		      unsigned int nr_pages)
2246
{
2247
	unsigned int batch = max(CHARGE_BATCH, nr_pages);
2248
	int nr_retries = MEM_CGROUP_RECLAIM_RETRIES;
2249
	struct mem_cgroup *mem_over_limit;
2250
	struct page_counter *counter;
2251
	unsigned long nr_reclaimed;
2252 2253
	bool may_swap = true;
	bool drained = false;
2254
	int ret = 0;
2255

2256 2257
	if (mem_cgroup_is_root(memcg))
		goto done;
2258
retry:
2259 2260
	if (consume_stock(memcg, nr_pages))
		goto done;
2261

2262
	if (!do_swap_account ||
2263 2264
	    !page_counter_try_charge(&memcg->memsw, batch, &counter)) {
		if (!page_counter_try_charge(&memcg->memory, batch, &counter))
2265
			goto done_restock;
2266
		if (do_swap_account)
2267 2268
			page_counter_uncharge(&memcg->memsw, batch);
		mem_over_limit = mem_cgroup_from_counter(counter, memory);
2269
	} else {
2270
		mem_over_limit = mem_cgroup_from_counter(counter, memsw);
2271
		may_swap = false;
2272
	}
2273

2274 2275 2276 2277
	if (batch > nr_pages) {
		batch = nr_pages;
		goto retry;
	}
2278

2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292
	/*
	 * 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;

2293 2294
	if (!(gfp_mask & __GFP_WAIT))
		goto nomem;
2295

2296 2297
	nr_reclaimed = try_to_free_mem_cgroup_pages(mem_over_limit, nr_pages,
						    gfp_mask, may_swap);
2298

2299
	if (mem_cgroup_margin(mem_over_limit) >= nr_pages)
2300
		goto retry;
2301

2302
	if (!drained) {
2303
		drain_all_stock(mem_over_limit);
2304 2305 2306 2307
		drained = true;
		goto retry;
	}

2308 2309
	if (gfp_mask & __GFP_NORETRY)
		goto nomem;
2310 2311 2312 2313 2314 2315 2316 2317 2318
	/*
	 * 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.
	 */
2319
	if (nr_reclaimed && nr_pages <= (1 << PAGE_ALLOC_COSTLY_ORDER))
2320 2321 2322 2323 2324 2325 2326 2327
		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;

2328 2329 2330
	if (nr_retries--)
		goto retry;

2331 2332 2333
	if (gfp_mask & __GFP_NOFAIL)
		goto bypass;

2334 2335 2336
	if (fatal_signal_pending(current))
		goto bypass;

2337
	mem_cgroup_oom(mem_over_limit, gfp_mask, get_order(nr_pages));
2338
nomem:
2339
	if (!(gfp_mask & __GFP_NOFAIL))
2340
		return -ENOMEM;
K
KAMEZAWA Hiroyuki 已提交
2341
bypass:
2342
	return -EINTR;
2343 2344

done_restock:
2345
	css_get_many(&memcg->css, batch);
2346 2347 2348
	if (batch > nr_pages)
		refill_stock(memcg, batch - nr_pages);
done:
2349
	return ret;
2350
}
2351

2352
static void cancel_charge(struct mem_cgroup *memcg, unsigned int nr_pages)
2353
{
2354 2355 2356
	if (mem_cgroup_is_root(memcg))
		return;

2357
	page_counter_uncharge(&memcg->memory, nr_pages);
2358
	if (do_swap_account)
2359
		page_counter_uncharge(&memcg->memsw, nr_pages);
2360

2361
	css_put_many(&memcg->css, nr_pages);
2362 2363
}

2364 2365
/*
 * A helper function to get mem_cgroup from ID. must be called under
2366 2367 2368
 * 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.)
2369 2370 2371 2372 2373 2374
 */
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
2375
	return mem_cgroup_from_id(id);
2376 2377
}

2378 2379 2380 2381 2382 2383 2384 2385 2386 2387
/*
 * 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.
 */
2388
struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page)
2389
{
2390
	struct mem_cgroup *memcg;
2391
	unsigned short id;
2392 2393
	swp_entry_t ent;

2394
	VM_BUG_ON_PAGE(!PageLocked(page), page);
2395

2396
	memcg = page->mem_cgroup;
2397 2398
	if (memcg) {
		if (!css_tryget_online(&memcg->css))
2399
			memcg = NULL;
2400
	} else if (PageSwapCache(page)) {
2401
		ent.val = page_private(page);
2402
		id = lookup_swap_cgroup_id(ent);
2403
		rcu_read_lock();
2404
		memcg = mem_cgroup_lookup(id);
2405
		if (memcg && !css_tryget_online(&memcg->css))
2406
			memcg = NULL;
2407
		rcu_read_unlock();
2408
	}
2409
	return memcg;
2410 2411
}

2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442
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);
}

2443
static void commit_charge(struct page *page, struct mem_cgroup *memcg,
2444
			  bool lrucare)
2445
{
2446
	int isolated;
2447

2448
	VM_BUG_ON_PAGE(page->mem_cgroup, page);
2449 2450 2451 2452 2453

	/*
	 * 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.
	 */
2454 2455
	if (lrucare)
		lock_page_lru(page, &isolated);
2456

2457 2458
	/*
	 * Nobody should be changing or seriously looking at
2459
	 * page->mem_cgroup at this point:
2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470
	 *
	 * - 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
	 */
2471
	page->mem_cgroup = memcg;
2472

2473 2474
	if (lrucare)
		unlock_page_lru(page, isolated);
2475
}
2476

2477
#ifdef CONFIG_MEMCG_KMEM
2478 2479 2480 2481 2482 2483
/*
 * 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);

G
Glauber Costa 已提交
2484 2485 2486 2487 2488 2489 2490 2491 2492 2493
/*
 * 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;
2494
	return cache_from_memcg_idx(cachep, memcg_cache_id(p->memcg));
G
Glauber Costa 已提交
2495 2496
}

2497 2498
static int memcg_charge_kmem(struct mem_cgroup *memcg, gfp_t gfp,
			     unsigned long nr_pages)
2499
{
2500
	struct page_counter *counter;
2501 2502
	int ret = 0;

2503 2504
	ret = page_counter_try_charge(&memcg->kmem, nr_pages, &counter);
	if (ret < 0)
2505 2506
		return ret;

2507
	ret = try_charge(memcg, gfp, nr_pages);
2508 2509
	if (ret == -EINTR)  {
		/*
2510 2511 2512 2513 2514 2515
		 * 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
2516 2517 2518
		 * our minds.
		 *
		 * This condition will only trigger if the task entered
2519 2520 2521
		 * 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
2522 2523
		 * directed to the root cgroup in memcontrol.h
		 */
2524
		page_counter_charge(&memcg->memory, nr_pages);
2525
		if (do_swap_account)
2526
			page_counter_charge(&memcg->memsw, nr_pages);
2527
		css_get_many(&memcg->css, nr_pages);
2528 2529
		ret = 0;
	} else if (ret)
2530
		page_counter_uncharge(&memcg->kmem, nr_pages);
2531 2532 2533 2534

	return ret;
}

2535 2536
static void memcg_uncharge_kmem(struct mem_cgroup *memcg,
				unsigned long nr_pages)
2537
{
2538
	page_counter_uncharge(&memcg->memory, nr_pages);
2539
	if (do_swap_account)
2540
		page_counter_uncharge(&memcg->memsw, nr_pages);
2541

2542
	page_counter_uncharge(&memcg->kmem, nr_pages);
2543

2544
	css_put_many(&memcg->css, nr_pages);
2545 2546
}

2547 2548 2549 2550 2551 2552 2553 2554 2555 2556
/*
 * 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;
}

2557
static int memcg_alloc_cache_id(void)
2558
{
2559 2560 2561 2562 2563 2564 2565
	int id, size;
	int err;

	id = ida_simple_get(&kmem_limited_groups,
			    0, MEMCG_CACHES_MAX_SIZE, GFP_KERNEL);
	if (id < 0)
		return id;
2566

2567 2568 2569 2570 2571 2572 2573 2574 2575
	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);
2576 2577 2578 2579 2580
	if (size < MEMCG_CACHES_MIN_SIZE)
		size = MEMCG_CACHES_MIN_SIZE;
	else if (size > MEMCG_CACHES_MAX_SIZE)
		size = MEMCG_CACHES_MAX_SIZE;

2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594
	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);
2595 2596 2597 2598 2599 2600 2601 2602 2603
}

/*
 * 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)
{
2604
	memcg_limited_groups_array_size = num;
2605 2606
}

2607 2608
static void memcg_register_cache(struct mem_cgroup *memcg,
				 struct kmem_cache *root_cache)
2609
{
2610 2611
	static char memcg_name_buf[NAME_MAX + 1]; /* protected by
						     memcg_slab_mutex */
2612
	struct kmem_cache *cachep;
2613 2614
	int id;

2615 2616 2617 2618 2619 2620 2621 2622 2623 2624
	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))
2625 2626
		return;

2627
	cgroup_name(memcg->css.cgroup, memcg_name_buf, NAME_MAX + 1);
2628
	cachep = memcg_create_kmem_cache(memcg, root_cache, memcg_name_buf);
2629
	/*
2630 2631 2632
	 * 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.
2633
	 */
2634 2635
	if (!cachep)
		return;
2636

2637
	css_get(&memcg->css);
2638
	list_add(&cachep->memcg_params->list, &memcg->memcg_slab_caches);
2639

2640
	/*
2641 2642 2643
	 * 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.
2644
	 */
2645 2646
	smp_wmb();

2647 2648
	BUG_ON(root_cache->memcg_params->memcg_caches[id]);
	root_cache->memcg_params->memcg_caches[id] = cachep;
2649
}
2650

2651
static void memcg_unregister_cache(struct kmem_cache *cachep)
2652
{
2653
	struct kmem_cache *root_cache;
2654 2655 2656
	struct mem_cgroup *memcg;
	int id;

2657
	lockdep_assert_held(&memcg_slab_mutex);
2658

2659
	BUG_ON(is_root_cache(cachep));
2660

2661 2662
	root_cache = cachep->memcg_params->root_cache;
	memcg = cachep->memcg_params->memcg;
2663
	id = memcg_cache_id(memcg);
2664

2665 2666
	BUG_ON(root_cache->memcg_params->memcg_caches[id] != cachep);
	root_cache->memcg_params->memcg_caches[id] = NULL;
2667

2668 2669 2670
	list_del(&cachep->memcg_params->list);

	kmem_cache_destroy(cachep);
2671 2672 2673

	/* drop the reference taken in memcg_register_cache */
	css_put(&memcg->css);
2674 2675
}

2676
int __memcg_cleanup_cache_params(struct kmem_cache *s)
2677 2678
{
	struct kmem_cache *c;
2679
	int i, failed = 0;
2680

2681
	mutex_lock(&memcg_slab_mutex);
2682 2683
	for_each_memcg_cache_index(i) {
		c = cache_from_memcg_idx(s, i);
2684 2685 2686
		if (!c)
			continue;

2687
		memcg_unregister_cache(c);
2688 2689 2690

		if (cache_from_memcg_idx(s, i))
			failed++;
2691
	}
2692
	mutex_unlock(&memcg_slab_mutex);
2693
	return failed;
2694 2695
}

2696
static void memcg_unregister_all_caches(struct mem_cgroup *memcg)
G
Glauber Costa 已提交
2697 2698
{
	struct kmem_cache *cachep;
2699
	struct memcg_cache_params *params, *tmp;
G
Glauber Costa 已提交
2700 2701 2702 2703

	if (!memcg_kmem_is_active(memcg))
		return;

2704 2705
	mutex_lock(&memcg_slab_mutex);
	list_for_each_entry_safe(params, tmp, &memcg->memcg_slab_caches, list) {
G
Glauber Costa 已提交
2706
		cachep = memcg_params_to_cache(params);
2707 2708
		kmem_cache_shrink(cachep);
		if (atomic_read(&cachep->memcg_params->nr_pages) == 0)
2709
			memcg_unregister_cache(cachep);
G
Glauber Costa 已提交
2710
	}
2711
	mutex_unlock(&memcg_slab_mutex);
G
Glauber Costa 已提交
2712 2713
}

2714
struct memcg_register_cache_work {
2715 2716 2717 2718 2719
	struct mem_cgroup *memcg;
	struct kmem_cache *cachep;
	struct work_struct work;
};

2720
static void memcg_register_cache_func(struct work_struct *w)
2721
{
2722 2723
	struct memcg_register_cache_work *cw =
		container_of(w, struct memcg_register_cache_work, work);
2724 2725
	struct mem_cgroup *memcg = cw->memcg;
	struct kmem_cache *cachep = cw->cachep;
2726

2727
	mutex_lock(&memcg_slab_mutex);
2728
	memcg_register_cache(memcg, cachep);
2729 2730
	mutex_unlock(&memcg_slab_mutex);

2731
	css_put(&memcg->css);
2732 2733 2734 2735 2736 2737
	kfree(cw);
}

/*
 * Enqueue the creation of a per-memcg kmem_cache.
 */
2738 2739
static void __memcg_schedule_register_cache(struct mem_cgroup *memcg,
					    struct kmem_cache *cachep)
2740
{
2741
	struct memcg_register_cache_work *cw;
2742

2743
	cw = kmalloc(sizeof(*cw), GFP_NOWAIT);
2744 2745
	if (cw == NULL) {
		css_put(&memcg->css);
2746 2747 2748 2749 2750 2751
		return;
	}

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

2752
	INIT_WORK(&cw->work, memcg_register_cache_func);
2753 2754 2755
	schedule_work(&cw->work);
}

2756 2757
static void memcg_schedule_register_cache(struct mem_cgroup *memcg,
					  struct kmem_cache *cachep)
2758 2759 2760 2761
{
	/*
	 * We need to stop accounting when we kmalloc, because if the
	 * corresponding kmalloc cache is not yet created, the first allocation
2762
	 * in __memcg_schedule_register_cache will recurse.
2763 2764 2765 2766 2767 2768 2769
	 *
	 * 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.
	 */
2770
	current->memcg_kmem_skip_account = 1;
2771
	__memcg_schedule_register_cache(memcg, cachep);
2772
	current->memcg_kmem_skip_account = 0;
2773
}
2774 2775 2776

int __memcg_charge_slab(struct kmem_cache *cachep, gfp_t gfp, int order)
{
2777
	unsigned int nr_pages = 1 << order;
2778 2779
	int res;

2780
	res = memcg_charge_kmem(cachep->memcg_params->memcg, gfp, nr_pages);
2781
	if (!res)
2782
		atomic_add(nr_pages, &cachep->memcg_params->nr_pages);
2783 2784 2785 2786 2787
	return res;
}

void __memcg_uncharge_slab(struct kmem_cache *cachep, int order)
{
2788 2789 2790 2791
	unsigned int nr_pages = 1 << order;

	memcg_uncharge_kmem(cachep->memcg_params->memcg, nr_pages);
	atomic_sub(nr_pages, &cachep->memcg_params->nr_pages);
2792 2793
}

2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806
/*
 * 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.
 */
2807
struct kmem_cache *__memcg_kmem_get_cache(struct kmem_cache *cachep)
2808 2809
{
	struct mem_cgroup *memcg;
2810
	struct kmem_cache *memcg_cachep;
2811 2812 2813 2814

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

2815
	if (current->memcg_kmem_skip_account)
2816 2817
		return cachep;

2818 2819 2820
	rcu_read_lock();
	memcg = mem_cgroup_from_task(rcu_dereference(current->mm->owner));

2821
	if (!memcg_kmem_is_active(memcg))
2822
		goto out;
2823

2824 2825 2826
	memcg_cachep = cache_from_memcg_idx(cachep, memcg_cache_id(memcg));
	if (likely(memcg_cachep)) {
		cachep = memcg_cachep;
2827
		goto out;
2828 2829
	}

2830
	/* The corresponding put will be done in the workqueue. */
2831
	if (!css_tryget_online(&memcg->css))
2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842
		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
2843 2844 2845
	 * memcg_create_kmem_cache, this means no further allocation
	 * could happen with the slab_mutex held. So it's better to
	 * defer everything.
2846
	 */
2847
	memcg_schedule_register_cache(memcg, cachep);
2848 2849 2850 2851
	return cachep;
out:
	rcu_read_unlock();
	return cachep;
2852 2853
}

2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874
/*
 * 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;
2875

2876
	memcg = get_mem_cgroup_from_mm(current->mm);
2877

2878
	if (!memcg_kmem_is_active(memcg)) {
2879 2880 2881 2882
		css_put(&memcg->css);
		return true;
	}

2883
	ret = memcg_charge_kmem(memcg, gfp, 1 << order);
2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897
	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)
{
	VM_BUG_ON(mem_cgroup_is_root(memcg));

	/* The page allocation failed. Revert */
	if (!page) {
2898
		memcg_uncharge_kmem(memcg, 1 << order);
2899 2900
		return;
	}
2901
	page->mem_cgroup = memcg;
2902 2903 2904 2905
}

void __memcg_kmem_uncharge_pages(struct page *page, int order)
{
2906
	struct mem_cgroup *memcg = page->mem_cgroup;
2907 2908 2909 2910

	if (!memcg)
		return;

2911
	VM_BUG_ON_PAGE(mem_cgroup_is_root(memcg), page);
2912

2913
	memcg_uncharge_kmem(memcg, 1 << order);
2914
	page->mem_cgroup = NULL;
2915
}
G
Glauber Costa 已提交
2916
#else
2917
static inline void memcg_unregister_all_caches(struct mem_cgroup *memcg)
G
Glauber Costa 已提交
2918 2919
{
}
2920 2921
#endif /* CONFIG_MEMCG_KMEM */

2922 2923 2924 2925
#ifdef CONFIG_TRANSPARENT_HUGEPAGE

/*
 * Because tail pages are not marked as "used", set it. We're under
2926 2927 2928
 * 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.
2929
 */
2930
void mem_cgroup_split_huge_fixup(struct page *head)
2931
{
2932
	int i;
2933

2934 2935
	if (mem_cgroup_disabled())
		return;
2936

2937
	for (i = 1; i < HPAGE_PMD_NR; i++)
2938
		head[i].mem_cgroup = head->mem_cgroup;
2939

2940
	__this_cpu_sub(head->mem_cgroup->stat->count[MEM_CGROUP_STAT_RSS_HUGE],
2941
		       HPAGE_PMD_NR);
2942
}
2943
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
2944

2945
/**
2946
 * mem_cgroup_move_account - move account of the page
2947
 * @page: the page
2948
 * @nr_pages: number of regular pages (>1 for huge pages)
2949 2950 2951 2952
 * @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
2953
 * - page is not on LRU (isolate_page() is useful.)
2954
 * - compound_lock is held when nr_pages > 1
2955
 *
2956 2957
 * This function doesn't do "charge" to new cgroup and doesn't do "uncharge"
 * from old cgroup.
2958
 */
2959 2960 2961
static int mem_cgroup_move_account(struct page *page,
				   unsigned int nr_pages,
				   struct mem_cgroup *from,
2962
				   struct mem_cgroup *to)
2963
{
2964 2965
	unsigned long flags;
	int ret;
2966

2967
	VM_BUG_ON(from == to);
2968
	VM_BUG_ON_PAGE(PageLRU(page), page);
2969 2970 2971 2972 2973 2974 2975
	/*
	 * 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;
2976
	if (nr_pages > 1 && !PageTransHuge(page))
2977 2978
		goto out;

2979
	/*
2980
	 * Prevent mem_cgroup_migrate() from looking at page->mem_cgroup
2981 2982 2983 2984 2985
	 * 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;
2986 2987

	ret = -EINVAL;
2988
	if (page->mem_cgroup != from)
2989
		goto out_unlock;
2990

2991
	spin_lock_irqsave(&from->move_lock, flags);
2992

2993
	if (!PageAnon(page) && page_mapped(page)) {
2994 2995 2996 2997 2998
		__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);
	}
2999

3000 3001 3002 3003 3004 3005
	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);
	}
3006

3007
	/*
3008
	 * It is safe to change page->mem_cgroup here because the page
3009 3010 3011
	 * is referenced, charged, and isolated - we can't race with
	 * uncharging, charging, migration, or LRU putback.
	 */
3012

3013
	/* caller should have done css_get */
3014
	page->mem_cgroup = to;
3015 3016
	spin_unlock_irqrestore(&from->move_lock, flags);

3017
	ret = 0;
3018 3019 3020

	local_irq_disable();
	mem_cgroup_charge_statistics(to, page, nr_pages);
3021
	memcg_check_events(to, page);
3022
	mem_cgroup_charge_statistics(from, page, -nr_pages);
3023
	memcg_check_events(from, page);
3024 3025 3026
	local_irq_enable();
out_unlock:
	unlock_page(page);
3027
out:
3028 3029 3030
	return ret;
}

A
Andrew Morton 已提交
3031
#ifdef CONFIG_MEMCG_SWAP
3032 3033
static void mem_cgroup_swap_statistics(struct mem_cgroup *memcg,
					 bool charge)
K
KAMEZAWA Hiroyuki 已提交
3034
{
3035 3036
	int val = (charge) ? 1 : -1;
	this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_SWAP], val);
K
KAMEZAWA Hiroyuki 已提交
3037
}
3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049

/**
 * 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.
 *
3050
 * The caller must have charged to @to, IOW, called page_counter_charge() about
3051 3052 3053
 * both res and memsw, and called css_get().
 */
static int mem_cgroup_move_swap_account(swp_entry_t entry,
3054
				struct mem_cgroup *from, struct mem_cgroup *to)
3055 3056 3057
{
	unsigned short old_id, new_id;

L
Li Zefan 已提交
3058 3059
	old_id = mem_cgroup_id(from);
	new_id = mem_cgroup_id(to);
3060 3061 3062

	if (swap_cgroup_cmpxchg(entry, old_id, new_id) == old_id) {
		mem_cgroup_swap_statistics(from, false);
3063
		mem_cgroup_swap_statistics(to, true);
3064
		/*
3065
		 * This function is only called from task migration context now.
3066
		 * It postpones page_counter and refcount handling till the end
3067
		 * of task migration(mem_cgroup_clear_mc()) for performance
L
Li Zefan 已提交
3068 3069 3070 3071 3072 3073
		 * 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().
3074
		 */
L
Li Zefan 已提交
3075
		css_get(&to->css);
3076 3077 3078 3079 3080 3081
		return 0;
	}
	return -EINVAL;
}
#else
static inline int mem_cgroup_move_swap_account(swp_entry_t entry,
3082
				struct mem_cgroup *from, struct mem_cgroup *to)
3083 3084 3085
{
	return -EINVAL;
}
3086
#endif
K
KAMEZAWA Hiroyuki 已提交
3087

3088
static DEFINE_MUTEX(memcg_limit_mutex);
3089

3090
static int mem_cgroup_resize_limit(struct mem_cgroup *memcg,
3091
				   unsigned long limit)
3092
{
3093 3094 3095
	unsigned long curusage;
	unsigned long oldusage;
	bool enlarge = false;
3096
	int retry_count;
3097
	int ret;
3098 3099 3100 3101 3102 3103

	/*
	 * 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.
	 */
3104 3105
	retry_count = MEM_CGROUP_RECLAIM_RETRIES *
		      mem_cgroup_count_children(memcg);
3106

3107
	oldusage = page_counter_read(&memcg->memory);
3108

3109
	do {
3110 3111 3112 3113
		if (signal_pending(current)) {
			ret = -EINTR;
			break;
		}
3114 3115 3116 3117

		mutex_lock(&memcg_limit_mutex);
		if (limit > memcg->memsw.limit) {
			mutex_unlock(&memcg_limit_mutex);
3118
			ret = -EINVAL;
3119 3120
			break;
		}
3121 3122 3123 3124
		if (limit > memcg->memory.limit)
			enlarge = true;
		ret = page_counter_limit(&memcg->memory, limit);
		mutex_unlock(&memcg_limit_mutex);
3125 3126 3127 3128

		if (!ret)
			break;

3129 3130
		try_to_free_mem_cgroup_pages(memcg, 1, GFP_KERNEL, true);

3131
		curusage = page_counter_read(&memcg->memory);
3132
		/* Usage is reduced ? */
A
Andrew Morton 已提交
3133
		if (curusage >= oldusage)
3134 3135 3136
			retry_count--;
		else
			oldusage = curusage;
3137 3138
	} while (retry_count);

3139 3140
	if (!ret && enlarge)
		memcg_oom_recover(memcg);
3141

3142 3143 3144
	return ret;
}

L
Li Zefan 已提交
3145
static int mem_cgroup_resize_memsw_limit(struct mem_cgroup *memcg,
3146
					 unsigned long limit)
3147
{
3148 3149 3150
	unsigned long curusage;
	unsigned long oldusage;
	bool enlarge = false;
3151
	int retry_count;
3152
	int ret;
3153

3154
	/* see mem_cgroup_resize_res_limit */
3155 3156 3157 3158 3159 3160
	retry_count = MEM_CGROUP_RECLAIM_RETRIES *
		      mem_cgroup_count_children(memcg);

	oldusage = page_counter_read(&memcg->memsw);

	do {
3161 3162 3163 3164
		if (signal_pending(current)) {
			ret = -EINTR;
			break;
		}
3165 3166 3167 3168

		mutex_lock(&memcg_limit_mutex);
		if (limit < memcg->memory.limit) {
			mutex_unlock(&memcg_limit_mutex);
3169 3170 3171
			ret = -EINVAL;
			break;
		}
3172 3173 3174 3175
		if (limit > memcg->memsw.limit)
			enlarge = true;
		ret = page_counter_limit(&memcg->memsw, limit);
		mutex_unlock(&memcg_limit_mutex);
3176 3177 3178 3179

		if (!ret)
			break;

3180 3181
		try_to_free_mem_cgroup_pages(memcg, 1, GFP_KERNEL, false);

3182
		curusage = page_counter_read(&memcg->memsw);
3183
		/* Usage is reduced ? */
3184
		if (curusage >= oldusage)
3185
			retry_count--;
3186 3187
		else
			oldusage = curusage;
3188 3189
	} while (retry_count);

3190 3191
	if (!ret && enlarge)
		memcg_oom_recover(memcg);
3192

3193 3194 3195
	return ret;
}

3196 3197 3198 3199 3200 3201 3202 3203 3204
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;
3205
	unsigned long excess;
3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229
	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;
3230
		spin_lock_irq(&mctz->lock);
3231
		__mem_cgroup_remove_exceeded(mz, mctz);
3232 3233 3234 3235 3236 3237

		/*
		 * If we failed to reclaim anything from this memory cgroup
		 * it is time to move on to the next cgroup
		 */
		next_mz = NULL;
3238 3239 3240
		if (!reclaimed)
			next_mz = __mem_cgroup_largest_soft_limit_node(mctz);

3241
		excess = soft_limit_excess(mz->memcg);
3242 3243 3244 3245 3246 3247 3248 3249 3250
		/*
		 * 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 */
3251
		__mem_cgroup_insert_exceeded(mz, mctz, excess);
3252
		spin_unlock_irq(&mctz->lock);
3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269
		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;
}

3270 3271 3272 3273 3274 3275
/*
 * 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.
 */
3276 3277
static inline bool memcg_has_children(struct mem_cgroup *memcg)
{
3278 3279
	bool ret;

3280
	/*
3281 3282 3283 3284
	 * 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.
3285
	 */
3286 3287 3288 3289 3290 3291
	lockdep_assert_held(&memcg_create_mutex);

	rcu_read_lock();
	ret = css_next_child(NULL, &memcg->css);
	rcu_read_unlock();
	return ret;
3292 3293
}

3294 3295 3296 3297 3298 3299 3300 3301 3302 3303
/*
 * 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;

3304 3305
	/* we call try-to-free pages for make this cgroup empty */
	lru_add_drain_all();
3306
	/* try to free all pages in this cgroup */
3307
	while (nr_retries && page_counter_read(&memcg->memory)) {
3308
		int progress;
3309

3310 3311 3312
		if (signal_pending(current))
			return -EINTR;

3313 3314
		progress = try_to_free_mem_cgroup_pages(memcg, 1,
							GFP_KERNEL, true);
3315
		if (!progress) {
3316
			nr_retries--;
3317
			/* maybe some writeback is necessary */
3318
			congestion_wait(BLK_RW_ASYNC, HZ/10);
3319
		}
3320 3321

	}
3322 3323

	return 0;
3324 3325
}

3326 3327 3328
static ssize_t mem_cgroup_force_empty_write(struct kernfs_open_file *of,
					    char *buf, size_t nbytes,
					    loff_t off)
3329
{
3330
	struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of));
3331

3332 3333
	if (mem_cgroup_is_root(memcg))
		return -EINVAL;
3334
	return mem_cgroup_force_empty(memcg) ?: nbytes;
3335 3336
}

3337 3338
static u64 mem_cgroup_hierarchy_read(struct cgroup_subsys_state *css,
				     struct cftype *cft)
3339
{
3340
	return mem_cgroup_from_css(css)->use_hierarchy;
3341 3342
}

3343 3344
static int mem_cgroup_hierarchy_write(struct cgroup_subsys_state *css,
				      struct cftype *cft, u64 val)
3345 3346
{
	int retval = 0;
3347
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
T
Tejun Heo 已提交
3348
	struct mem_cgroup *parent_memcg = mem_cgroup_from_css(memcg->css.parent);
3349

3350
	mutex_lock(&memcg_create_mutex);
3351 3352 3353 3354

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

3355
	/*
3356
	 * If parent's use_hierarchy is set, we can't make any modifications
3357 3358 3359 3360 3361 3362
	 * 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.
	 */
3363
	if ((!parent_memcg || !parent_memcg->use_hierarchy) &&
3364
				(val == 1 || val == 0)) {
3365
		if (!memcg_has_children(memcg))
3366
			memcg->use_hierarchy = val;
3367 3368 3369 3370
		else
			retval = -EBUSY;
	} else
		retval = -EINVAL;
3371 3372

out:
3373
	mutex_unlock(&memcg_create_mutex);
3374 3375 3376 3377

	return retval;
}

3378 3379
static unsigned long tree_stat(struct mem_cgroup *memcg,
			       enum mem_cgroup_stat_index idx)
3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396
{
	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;

3397 3398 3399 3400 3401 3402
	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 {
3403
		if (!swap)
3404
			val = page_counter_read(&memcg->memory);
3405
		else
3406
			val = page_counter_read(&memcg->memsw);
3407 3408 3409 3410
	}
	return val << PAGE_SHIFT;
}

3411 3412 3413 3414 3415 3416 3417
enum {
	RES_USAGE,
	RES_LIMIT,
	RES_MAX_USAGE,
	RES_FAILCNT,
	RES_SOFT_LIMIT,
};
3418

3419
static u64 mem_cgroup_read_u64(struct cgroup_subsys_state *css,
3420
			       struct cftype *cft)
B
Balbir Singh 已提交
3421
{
3422
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
3423
	struct page_counter *counter;
3424

3425
	switch (MEMFILE_TYPE(cft->private)) {
3426
	case _MEM:
3427 3428
		counter = &memcg->memory;
		break;
3429
	case _MEMSWAP:
3430 3431
		counter = &memcg->memsw;
		break;
3432
	case _KMEM:
3433
		counter = &memcg->kmem;
3434
		break;
3435 3436 3437
	default:
		BUG();
	}
3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456

	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
3457
}
3458 3459

#ifdef CONFIG_MEMCG_KMEM
3460 3461
static int memcg_activate_kmem(struct mem_cgroup *memcg,
			       unsigned long nr_pages)
3462 3463 3464 3465 3466 3467 3468
{
	int err = 0;
	int memcg_id;

	if (memcg_kmem_is_active(memcg))
		return 0;

3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480
	/*
	 * 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.
	 */
3481
	mutex_lock(&memcg_create_mutex);
3482 3483
	if (cgroup_has_tasks(memcg->css.cgroup) ||
	    (memcg->use_hierarchy && memcg_has_children(memcg)))
3484 3485 3486 3487
		err = -EBUSY;
	mutex_unlock(&memcg_create_mutex);
	if (err)
		goto out;
3488

3489
	memcg_id = memcg_alloc_cache_id();
3490 3491 3492 3493 3494 3495
	if (memcg_id < 0) {
		err = memcg_id;
		goto out;
	}

	/*
V
Vladimir Davydov 已提交
3496 3497
	 * We couldn't have accounted to this cgroup, because it hasn't got
	 * activated yet, so this should succeed.
3498
	 */
3499
	err = page_counter_limit(&memcg->kmem, nr_pages);
3500 3501 3502 3503
	VM_BUG_ON(err);

	static_key_slow_inc(&memcg_kmem_enabled_key);
	/*
V
Vladimir Davydov 已提交
3504 3505
	 * A memory cgroup is considered kmem-active as soon as it gets
	 * kmemcg_id. Setting the id after enabling static branching will
3506 3507 3508
	 * guarantee no one starts accounting before all call sites are
	 * patched.
	 */
V
Vladimir Davydov 已提交
3509
	memcg->kmemcg_id = memcg_id;
3510
out:
3511 3512 3513 3514
	return err;
}

static int memcg_update_kmem_limit(struct mem_cgroup *memcg,
3515
				   unsigned long limit)
3516 3517 3518
{
	int ret;

3519
	mutex_lock(&memcg_limit_mutex);
3520
	if (!memcg_kmem_is_active(memcg))
3521
		ret = memcg_activate_kmem(memcg, limit);
3522
	else
3523 3524
		ret = page_counter_limit(&memcg->kmem, limit);
	mutex_unlock(&memcg_limit_mutex);
3525 3526 3527
	return ret;
}

3528
static int memcg_propagate_kmem(struct mem_cgroup *memcg)
3529
{
3530
	int ret = 0;
3531
	struct mem_cgroup *parent = parent_mem_cgroup(memcg);
3532

3533 3534
	if (!parent)
		return 0;
3535

3536
	mutex_lock(&memcg_limit_mutex);
3537
	/*
3538 3539
	 * If the parent cgroup is not kmem-active now, it cannot be activated
	 * after this point, because it has at least one child already.
3540
	 */
3541
	if (memcg_kmem_is_active(parent))
3542 3543
		ret = memcg_activate_kmem(memcg, PAGE_COUNTER_MAX);
	mutex_unlock(&memcg_limit_mutex);
3544
	return ret;
3545
}
3546 3547
#else
static int memcg_update_kmem_limit(struct mem_cgroup *memcg,
3548
				   unsigned long limit)
3549 3550 3551
{
	return -EINVAL;
}
3552
#endif /* CONFIG_MEMCG_KMEM */
3553

3554 3555 3556 3557
/*
 * The user of this function is...
 * RES_LIMIT.
 */
3558 3559
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
3560
{
3561
	struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of));
3562
	unsigned long nr_pages;
3563 3564
	int ret;

3565
	buf = strstrip(buf);
3566 3567 3568
	ret = page_counter_memparse(buf, &nr_pages);
	if (ret)
		return ret;
3569

3570
	switch (MEMFILE_ATTR(of_cft(of)->private)) {
3571
	case RES_LIMIT:
3572 3573 3574 3575
		if (mem_cgroup_is_root(memcg)) { /* Can't set limit on root */
			ret = -EINVAL;
			break;
		}
3576 3577 3578
		switch (MEMFILE_TYPE(of_cft(of)->private)) {
		case _MEM:
			ret = mem_cgroup_resize_limit(memcg, nr_pages);
3579
			break;
3580 3581
		case _MEMSWAP:
			ret = mem_cgroup_resize_memsw_limit(memcg, nr_pages);
3582
			break;
3583 3584 3585 3586
		case _KMEM:
			ret = memcg_update_kmem_limit(memcg, nr_pages);
			break;
		}
3587
		break;
3588 3589 3590
	case RES_SOFT_LIMIT:
		memcg->soft_limit = nr_pages;
		ret = 0;
3591 3592
		break;
	}
3593
	return ret ?: nbytes;
B
Balbir Singh 已提交
3594 3595
}

3596 3597
static ssize_t mem_cgroup_reset(struct kernfs_open_file *of, char *buf,
				size_t nbytes, loff_t off)
3598
{
3599
	struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of));
3600
	struct page_counter *counter;
3601

3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614
	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();
	}
3615

3616
	switch (MEMFILE_ATTR(of_cft(of)->private)) {
3617
	case RES_MAX_USAGE:
3618
		page_counter_reset_watermark(counter);
3619 3620
		break;
	case RES_FAILCNT:
3621
		counter->failcnt = 0;
3622
		break;
3623 3624
	default:
		BUG();
3625
	}
3626

3627
	return nbytes;
3628 3629
}

3630
static u64 mem_cgroup_move_charge_read(struct cgroup_subsys_state *css,
3631 3632
					struct cftype *cft)
{
3633
	return mem_cgroup_from_css(css)->move_charge_at_immigrate;
3634 3635
}

3636
#ifdef CONFIG_MMU
3637
static int mem_cgroup_move_charge_write(struct cgroup_subsys_state *css,
3638 3639
					struct cftype *cft, u64 val)
{
3640
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
3641 3642 3643

	if (val >= (1 << NR_MOVE_TYPE))
		return -EINVAL;
3644

3645
	/*
3646 3647 3648 3649
	 * 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.
3650
	 */
3651
	memcg->move_charge_at_immigrate = val;
3652 3653
	return 0;
}
3654
#else
3655
static int mem_cgroup_move_charge_write(struct cgroup_subsys_state *css,
3656 3657 3658 3659 3660
					struct cftype *cft, u64 val)
{
	return -ENOSYS;
}
#endif
3661

3662
#ifdef CONFIG_NUMA
3663
static int memcg_numa_stat_show(struct seq_file *m, void *v)
3664
{
3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676
	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;
3677
	int nid;
3678
	unsigned long nr;
3679
	struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(m));
3680

3681 3682 3683 3684 3685 3686 3687 3688 3689
	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');
3690 3691
	}

3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706
	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');
3707 3708 3709 3710 3711 3712
	}

	return 0;
}
#endif /* CONFIG_NUMA */

3713 3714 3715 3716 3717
static inline void mem_cgroup_lru_names_not_uptodate(void)
{
	BUILD_BUG_ON(ARRAY_SIZE(mem_cgroup_lru_names) != NR_LRU_LISTS);
}

3718
static int memcg_stat_show(struct seq_file *m, void *v)
3719
{
3720
	struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(m));
3721
	unsigned long memory, memsw;
3722 3723
	struct mem_cgroup *mi;
	unsigned int i;
3724

3725
	for (i = 0; i < MEM_CGROUP_STAT_NSTATS; i++) {
3726
		if (i == MEM_CGROUP_STAT_SWAP && !do_swap_account)
3727
			continue;
3728 3729
		seq_printf(m, "%s %ld\n", mem_cgroup_stat_names[i],
			   mem_cgroup_read_stat(memcg, i) * PAGE_SIZE);
3730
	}
L
Lee Schermerhorn 已提交
3731

3732 3733 3734 3735 3736 3737 3738 3739
	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
3740
	/* Hierarchical information */
3741 3742 3743 3744
	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);
3745
	}
3746 3747 3748 3749 3750
	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
3751

3752 3753 3754
	for (i = 0; i < MEM_CGROUP_STAT_NSTATS; i++) {
		long long val = 0;

3755
		if (i == MEM_CGROUP_STAT_SWAP && !do_swap_account)
3756
			continue;
3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776
		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);
3777
	}
K
KAMEZAWA Hiroyuki 已提交
3778

K
KOSAKI Motohiro 已提交
3779 3780 3781 3782
#ifdef CONFIG_DEBUG_VM
	{
		int nid, zid;
		struct mem_cgroup_per_zone *mz;
3783
		struct zone_reclaim_stat *rstat;
K
KOSAKI Motohiro 已提交
3784 3785 3786 3787 3788
		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++) {
3789
				mz = &memcg->nodeinfo[nid]->zoneinfo[zid];
3790
				rstat = &mz->lruvec.reclaim_stat;
K
KOSAKI Motohiro 已提交
3791

3792 3793 3794 3795
				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
3796
			}
3797 3798 3799 3800
		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
3801 3802 3803
	}
#endif

3804 3805 3806
	return 0;
}

3807 3808
static u64 mem_cgroup_swappiness_read(struct cgroup_subsys_state *css,
				      struct cftype *cft)
K
KOSAKI Motohiro 已提交
3809
{
3810
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
K
KOSAKI Motohiro 已提交
3811

3812
	return mem_cgroup_swappiness(memcg);
K
KOSAKI Motohiro 已提交
3813 3814
}

3815 3816
static int mem_cgroup_swappiness_write(struct cgroup_subsys_state *css,
				       struct cftype *cft, u64 val)
K
KOSAKI Motohiro 已提交
3817
{
3818
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
K
KOSAKI Motohiro 已提交
3819

3820
	if (val > 100)
K
KOSAKI Motohiro 已提交
3821 3822
		return -EINVAL;

3823
	if (css->parent)
3824 3825 3826
		memcg->swappiness = val;
	else
		vm_swappiness = val;
3827

K
KOSAKI Motohiro 已提交
3828 3829 3830
	return 0;
}

3831 3832 3833
static void __mem_cgroup_threshold(struct mem_cgroup *memcg, bool swap)
{
	struct mem_cgroup_threshold_ary *t;
3834
	unsigned long usage;
3835 3836 3837 3838
	int i;

	rcu_read_lock();
	if (!swap)
3839
		t = rcu_dereference(memcg->thresholds.primary);
3840
	else
3841
		t = rcu_dereference(memcg->memsw_thresholds.primary);
3842 3843 3844 3845

	if (!t)
		goto unlock;

3846
	usage = mem_cgroup_usage(memcg, swap);
3847 3848

	/*
3849
	 * current_threshold points to threshold just below or equal to usage.
3850 3851 3852
	 * If it's not true, a threshold was crossed after last
	 * call of __mem_cgroup_threshold().
	 */
3853
	i = t->current_threshold;
3854 3855 3856 3857 3858 3859 3860 3861 3862 3863 3864 3865 3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876

	/*
	 * 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 */
3877
	t->current_threshold = i - 1;
3878 3879 3880 3881 3882 3883
unlock:
	rcu_read_unlock();
}

static void mem_cgroup_threshold(struct mem_cgroup *memcg)
{
3884 3885 3886 3887 3888 3889 3890
	while (memcg) {
		__mem_cgroup_threshold(memcg, false);
		if (do_swap_account)
			__mem_cgroup_threshold(memcg, true);

		memcg = parent_mem_cgroup(memcg);
	}
3891 3892 3893 3894 3895 3896 3897
}

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

3898 3899 3900 3901 3902 3903 3904
	if (_a->threshold > _b->threshold)
		return 1;

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

	return 0;
3905 3906
}

3907
static int mem_cgroup_oom_notify_cb(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
3908 3909 3910
{
	struct mem_cgroup_eventfd_list *ev;

3911 3912
	spin_lock(&memcg_oom_lock);

3913
	list_for_each_entry(ev, &memcg->oom_notify, list)
K
KAMEZAWA Hiroyuki 已提交
3914
		eventfd_signal(ev->eventfd, 1);
3915 3916

	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
3917 3918 3919
	return 0;
}

3920
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
3921
{
K
KAMEZAWA Hiroyuki 已提交
3922 3923
	struct mem_cgroup *iter;

3924
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
3925
		mem_cgroup_oom_notify_cb(iter);
K
KAMEZAWA Hiroyuki 已提交
3926 3927
}

3928
static int __mem_cgroup_usage_register_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
3929
	struct eventfd_ctx *eventfd, const char *args, enum res_type type)
3930
{
3931 3932
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
3933 3934
	unsigned long threshold;
	unsigned long usage;
3935
	int i, size, ret;
3936

3937
	ret = page_counter_memparse(args, &threshold);
3938 3939 3940 3941
	if (ret)
		return ret;

	mutex_lock(&memcg->thresholds_lock);
3942

3943
	if (type == _MEM) {
3944
		thresholds = &memcg->thresholds;
3945
		usage = mem_cgroup_usage(memcg, false);
3946
	} else if (type == _MEMSWAP) {
3947
		thresholds = &memcg->memsw_thresholds;
3948
		usage = mem_cgroup_usage(memcg, true);
3949
	} else
3950 3951 3952
		BUG();

	/* Check if a threshold crossed before adding a new one */
3953
	if (thresholds->primary)
3954 3955
		__mem_cgroup_threshold(memcg, type == _MEMSWAP);

3956
	size = thresholds->primary ? thresholds->primary->size + 1 : 1;
3957 3958

	/* Allocate memory for new array of thresholds */
3959
	new = kmalloc(sizeof(*new) + size * sizeof(struct mem_cgroup_threshold),
3960
			GFP_KERNEL);
3961
	if (!new) {
3962 3963 3964
		ret = -ENOMEM;
		goto unlock;
	}
3965
	new->size = size;
3966 3967

	/* Copy thresholds (if any) to new array */
3968 3969
	if (thresholds->primary) {
		memcpy(new->entries, thresholds->primary->entries, (size - 1) *
3970
				sizeof(struct mem_cgroup_threshold));
3971 3972
	}

3973
	/* Add new threshold */
3974 3975
	new->entries[size - 1].eventfd = eventfd;
	new->entries[size - 1].threshold = threshold;
3976 3977

	/* Sort thresholds. Registering of new threshold isn't time-critical */
3978
	sort(new->entries, size, sizeof(struct mem_cgroup_threshold),
3979 3980 3981
			compare_thresholds, NULL);

	/* Find current threshold */
3982
	new->current_threshold = -1;
3983
	for (i = 0; i < size; i++) {
3984
		if (new->entries[i].threshold <= usage) {
3985
			/*
3986 3987
			 * new->current_threshold will not be used until
			 * rcu_assign_pointer(), so it's safe to increment
3988 3989
			 * it here.
			 */
3990
			++new->current_threshold;
3991 3992
		} else
			break;
3993 3994
	}

3995 3996 3997 3998 3999
	/* Free old spare buffer and save old primary buffer as spare */
	kfree(thresholds->spare);
	thresholds->spare = thresholds->primary;

	rcu_assign_pointer(thresholds->primary, new);
4000

4001
	/* To be sure that nobody uses thresholds */
4002 4003 4004 4005 4006 4007 4008 4009
	synchronize_rcu();

unlock:
	mutex_unlock(&memcg->thresholds_lock);

	return ret;
}

4010
static int mem_cgroup_usage_register_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
4011 4012
	struct eventfd_ctx *eventfd, const char *args)
{
4013
	return __mem_cgroup_usage_register_event(memcg, eventfd, args, _MEM);
T
Tejun Heo 已提交
4014 4015
}

4016
static int memsw_cgroup_usage_register_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
4017 4018
	struct eventfd_ctx *eventfd, const char *args)
{
4019
	return __mem_cgroup_usage_register_event(memcg, eventfd, args, _MEMSWAP);
T
Tejun Heo 已提交
4020 4021
}

4022
static void __mem_cgroup_usage_unregister_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
4023
	struct eventfd_ctx *eventfd, enum res_type type)
4024
{
4025 4026
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
4027
	unsigned long usage;
4028
	int i, j, size;
4029 4030

	mutex_lock(&memcg->thresholds_lock);
4031 4032

	if (type == _MEM) {
4033
		thresholds = &memcg->thresholds;
4034
		usage = mem_cgroup_usage(memcg, false);
4035
	} else if (type == _MEMSWAP) {
4036
		thresholds = &memcg->memsw_thresholds;
4037
		usage = mem_cgroup_usage(memcg, true);
4038
	} else
4039 4040
		BUG();

4041 4042 4043
	if (!thresholds->primary)
		goto unlock;

4044 4045 4046 4047
	/* Check if a threshold crossed before removing */
	__mem_cgroup_threshold(memcg, type == _MEMSWAP);

	/* Calculate new number of threshold */
4048 4049 4050
	size = 0;
	for (i = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd != eventfd)
4051 4052 4053
			size++;
	}

4054
	new = thresholds->spare;
4055

4056 4057
	/* Set thresholds array to NULL if we don't have thresholds */
	if (!size) {
4058 4059
		kfree(new);
		new = NULL;
4060
		goto swap_buffers;
4061 4062
	}

4063
	new->size = size;
4064 4065

	/* Copy thresholds and find current threshold */
4066 4067 4068
	new->current_threshold = -1;
	for (i = 0, j = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd == eventfd)
4069 4070
			continue;

4071
		new->entries[j] = thresholds->primary->entries[i];
4072
		if (new->entries[j].threshold <= usage) {
4073
			/*
4074
			 * new->current_threshold will not be used
4075 4076 4077
			 * until rcu_assign_pointer(), so it's safe to increment
			 * it here.
			 */
4078
			++new->current_threshold;
4079 4080 4081 4082
		}
		j++;
	}

4083
swap_buffers:
4084 4085
	/* Swap primary and spare array */
	thresholds->spare = thresholds->primary;
4086 4087 4088 4089 4090 4091
	/* If all events are unregistered, free the spare array */
	if (!new) {
		kfree(thresholds->spare);
		thresholds->spare = NULL;
	}

4092
	rcu_assign_pointer(thresholds->primary, new);
4093

4094
	/* To be sure that nobody uses thresholds */
4095
	synchronize_rcu();
4096
unlock:
4097 4098
	mutex_unlock(&memcg->thresholds_lock);
}
4099

4100
static void mem_cgroup_usage_unregister_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
4101 4102
	struct eventfd_ctx *eventfd)
{
4103
	return __mem_cgroup_usage_unregister_event(memcg, eventfd, _MEM);
T
Tejun Heo 已提交
4104 4105
}

4106
static void memsw_cgroup_usage_unregister_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
4107 4108
	struct eventfd_ctx *eventfd)
{
4109
	return __mem_cgroup_usage_unregister_event(memcg, eventfd, _MEMSWAP);
T
Tejun Heo 已提交
4110 4111
}

4112
static int mem_cgroup_oom_register_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
4113
	struct eventfd_ctx *eventfd, const char *args)
K
KAMEZAWA Hiroyuki 已提交
4114 4115 4116 4117 4118 4119 4120
{
	struct mem_cgroup_eventfd_list *event;

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

4121
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4122 4123 4124 4125 4126

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

	/* already in OOM ? */
4127
	if (atomic_read(&memcg->under_oom))
K
KAMEZAWA Hiroyuki 已提交
4128
		eventfd_signal(eventfd, 1);
4129
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4130 4131 4132 4133

	return 0;
}

4134
static void mem_cgroup_oom_unregister_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
4135
	struct eventfd_ctx *eventfd)
K
KAMEZAWA Hiroyuki 已提交
4136 4137 4138
{
	struct mem_cgroup_eventfd_list *ev, *tmp;

4139
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4140

4141
	list_for_each_entry_safe(ev, tmp, &memcg->oom_notify, list) {
K
KAMEZAWA Hiroyuki 已提交
4142 4143 4144 4145 4146 4147
		if (ev->eventfd == eventfd) {
			list_del(&ev->list);
			kfree(ev);
		}
	}

4148
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4149 4150
}

4151
static int mem_cgroup_oom_control_read(struct seq_file *sf, void *v)
4152
{
4153
	struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(sf));
4154

4155 4156
	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));
4157 4158 4159
	return 0;
}

4160
static int mem_cgroup_oom_control_write(struct cgroup_subsys_state *css,
4161 4162
	struct cftype *cft, u64 val)
{
4163
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
4164 4165

	/* cannot set to root cgroup and only 0 and 1 are allowed */
4166
	if (!css->parent || !((val == 0) || (val == 1)))
4167 4168
		return -EINVAL;

4169
	memcg->oom_kill_disable = val;
4170
	if (!val)
4171
		memcg_oom_recover(memcg);
4172

4173 4174 4175
	return 0;
}

A
Andrew Morton 已提交
4176
#ifdef CONFIG_MEMCG_KMEM
4177
static int memcg_init_kmem(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
4178
{
4179 4180 4181 4182 4183
	int ret;

	ret = memcg_propagate_kmem(memcg);
	if (ret)
		return ret;
4184

4185
	return mem_cgroup_sockets_init(memcg, ss);
4186
}
4187

4188
static void memcg_destroy_kmem(struct mem_cgroup *memcg)
G
Glauber Costa 已提交
4189
{
4190
	mem_cgroup_sockets_destroy(memcg);
4191
}
4192
#else
4193
static int memcg_init_kmem(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
4194 4195 4196
{
	return 0;
}
G
Glauber Costa 已提交
4197

4198 4199 4200
static void memcg_destroy_kmem(struct mem_cgroup *memcg)
{
}
4201 4202
#endif

4203 4204 4205 4206 4207 4208 4209 4210 4211 4212 4213 4214 4215
/*
 * 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.
 */

4216 4217 4218 4219 4220
/*
 * Unregister event and free resources.
 *
 * Gets called from workqueue.
 */
4221
static void memcg_event_remove(struct work_struct *work)
4222
{
4223 4224
	struct mem_cgroup_event *event =
		container_of(work, struct mem_cgroup_event, remove);
4225
	struct mem_cgroup *memcg = event->memcg;
4226 4227 4228

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

4229
	event->unregister_event(memcg, event->eventfd);
4230 4231 4232 4233 4234 4235

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

	eventfd_ctx_put(event->eventfd);
	kfree(event);
4236
	css_put(&memcg->css);
4237 4238 4239 4240 4241 4242 4243
}

/*
 * Gets called on POLLHUP on eventfd when user closes it.
 *
 * Called with wqh->lock held and interrupts disabled.
 */
4244 4245
static int memcg_event_wake(wait_queue_t *wait, unsigned mode,
			    int sync, void *key)
4246
{
4247 4248
	struct mem_cgroup_event *event =
		container_of(wait, struct mem_cgroup_event, wait);
4249
	struct mem_cgroup *memcg = event->memcg;
4250 4251 4252 4253 4254 4255 4256 4257 4258 4259 4260 4261
	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.
		 */
4262
		spin_lock(&memcg->event_list_lock);
4263 4264 4265 4266 4267 4268 4269 4270
		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);
		}
4271
		spin_unlock(&memcg->event_list_lock);
4272 4273 4274 4275 4276
	}

	return 0;
}

4277
static void memcg_event_ptable_queue_proc(struct file *file,
4278 4279
		wait_queue_head_t *wqh, poll_table *pt)
{
4280 4281
	struct mem_cgroup_event *event =
		container_of(pt, struct mem_cgroup_event, pt);
4282 4283 4284 4285 4286 4287

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

/*
4288 4289
 * DO NOT USE IN NEW FILES.
 *
4290 4291 4292 4293 4294
 * 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.
 */
4295 4296
static ssize_t memcg_write_event_control(struct kernfs_open_file *of,
					 char *buf, size_t nbytes, loff_t off)
4297
{
4298
	struct cgroup_subsys_state *css = of_css(of);
4299
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
4300
	struct mem_cgroup_event *event;
4301 4302 4303 4304
	struct cgroup_subsys_state *cfile_css;
	unsigned int efd, cfd;
	struct fd efile;
	struct fd cfile;
4305
	const char *name;
4306 4307 4308
	char *endp;
	int ret;

4309 4310 4311
	buf = strstrip(buf);

	efd = simple_strtoul(buf, &endp, 10);
4312 4313
	if (*endp != ' ')
		return -EINVAL;
4314
	buf = endp + 1;
4315

4316
	cfd = simple_strtoul(buf, &endp, 10);
4317 4318
	if ((*endp != ' ') && (*endp != '\0'))
		return -EINVAL;
4319
	buf = endp + 1;
4320 4321 4322 4323 4324

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

4325
	event->memcg = memcg;
4326
	INIT_LIST_HEAD(&event->list);
4327 4328 4329
	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);
4330 4331 4332 4333 4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352 4353 4354

	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;

4355 4356 4357 4358 4359
	/*
	 * 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.
4360 4361
	 *
	 * DO NOT ADD NEW FILES.
4362
	 */
A
Al Viro 已提交
4363
	name = cfile.file->f_path.dentry->d_name.name;
4364 4365 4366 4367 4368 4369 4370 4371 4372 4373 4374

	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
4375 4376
		event->register_event = memsw_cgroup_usage_register_event;
		event->unregister_event = memsw_cgroup_usage_unregister_event;
4377 4378 4379 4380 4381
	} else {
		ret = -EINVAL;
		goto out_put_cfile;
	}

4382
	/*
4383 4384 4385
	 * 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.
4386
	 */
A
Al Viro 已提交
4387
	cfile_css = css_tryget_online_from_dir(cfile.file->f_path.dentry->d_parent,
4388
					       &memory_cgrp_subsys);
4389
	ret = -EINVAL;
4390
	if (IS_ERR(cfile_css))
4391
		goto out_put_cfile;
4392 4393
	if (cfile_css != css) {
		css_put(cfile_css);
4394
		goto out_put_cfile;
4395
	}
4396

4397
	ret = event->register_event(memcg, event->eventfd, buf);
4398 4399 4400 4401 4402
	if (ret)
		goto out_put_css;

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

4403 4404 4405
	spin_lock(&memcg->event_list_lock);
	list_add(&event->list, &memcg->event_list);
	spin_unlock(&memcg->event_list_lock);
4406 4407 4408 4409

	fdput(cfile);
	fdput(efile);

4410
	return nbytes;
4411 4412

out_put_css:
4413
	css_put(css);
4414 4415 4416 4417 4418 4419 4420 4421 4422 4423 4424 4425
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
4426 4427
static struct cftype mem_cgroup_files[] = {
	{
4428
		.name = "usage_in_bytes",
4429
		.private = MEMFILE_PRIVATE(_MEM, RES_USAGE),
4430
		.read_u64 = mem_cgroup_read_u64,
B
Balbir Singh 已提交
4431
	},
4432 4433
	{
		.name = "max_usage_in_bytes",
4434
		.private = MEMFILE_PRIVATE(_MEM, RES_MAX_USAGE),
4435
		.write = mem_cgroup_reset,
4436
		.read_u64 = mem_cgroup_read_u64,
4437
	},
B
Balbir Singh 已提交
4438
	{
4439
		.name = "limit_in_bytes",
4440
		.private = MEMFILE_PRIVATE(_MEM, RES_LIMIT),
4441
		.write = mem_cgroup_write,
4442
		.read_u64 = mem_cgroup_read_u64,
B
Balbir Singh 已提交
4443
	},
4444 4445 4446
	{
		.name = "soft_limit_in_bytes",
		.private = MEMFILE_PRIVATE(_MEM, RES_SOFT_LIMIT),
4447
		.write = mem_cgroup_write,
4448
		.read_u64 = mem_cgroup_read_u64,
4449
	},
B
Balbir Singh 已提交
4450 4451
	{
		.name = "failcnt",
4452
		.private = MEMFILE_PRIVATE(_MEM, RES_FAILCNT),
4453
		.write = mem_cgroup_reset,
4454
		.read_u64 = mem_cgroup_read_u64,
B
Balbir Singh 已提交
4455
	},
4456 4457
	{
		.name = "stat",
4458
		.seq_show = memcg_stat_show,
4459
	},
4460 4461
	{
		.name = "force_empty",
4462
		.write = mem_cgroup_force_empty_write,
4463
	},
4464 4465 4466 4467 4468
	{
		.name = "use_hierarchy",
		.write_u64 = mem_cgroup_hierarchy_write,
		.read_u64 = mem_cgroup_hierarchy_read,
	},
4469
	{
4470
		.name = "cgroup.event_control",		/* XXX: for compat */
4471
		.write = memcg_write_event_control,
4472 4473 4474
		.flags = CFTYPE_NO_PREFIX,
		.mode = S_IWUGO,
	},
K
KOSAKI Motohiro 已提交
4475 4476 4477 4478 4479
	{
		.name = "swappiness",
		.read_u64 = mem_cgroup_swappiness_read,
		.write_u64 = mem_cgroup_swappiness_write,
	},
4480 4481 4482 4483 4484
	{
		.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
4485 4486
	{
		.name = "oom_control",
4487
		.seq_show = mem_cgroup_oom_control_read,
4488
		.write_u64 = mem_cgroup_oom_control_write,
K
KAMEZAWA Hiroyuki 已提交
4489 4490
		.private = MEMFILE_PRIVATE(_OOM_TYPE, OOM_CONTROL),
	},
4491 4492 4493
	{
		.name = "pressure_level",
	},
4494 4495 4496
#ifdef CONFIG_NUMA
	{
		.name = "numa_stat",
4497
		.seq_show = memcg_numa_stat_show,
4498 4499
	},
#endif
4500 4501 4502 4503
#ifdef CONFIG_MEMCG_KMEM
	{
		.name = "kmem.limit_in_bytes",
		.private = MEMFILE_PRIVATE(_KMEM, RES_LIMIT),
4504
		.write = mem_cgroup_write,
4505
		.read_u64 = mem_cgroup_read_u64,
4506 4507 4508 4509
	},
	{
		.name = "kmem.usage_in_bytes",
		.private = MEMFILE_PRIVATE(_KMEM, RES_USAGE),
4510
		.read_u64 = mem_cgroup_read_u64,
4511 4512 4513 4514
	},
	{
		.name = "kmem.failcnt",
		.private = MEMFILE_PRIVATE(_KMEM, RES_FAILCNT),
4515
		.write = mem_cgroup_reset,
4516
		.read_u64 = mem_cgroup_read_u64,
4517 4518 4519 4520
	},
	{
		.name = "kmem.max_usage_in_bytes",
		.private = MEMFILE_PRIVATE(_KMEM, RES_MAX_USAGE),
4521
		.write = mem_cgroup_reset,
4522
		.read_u64 = mem_cgroup_read_u64,
4523
	},
4524 4525 4526
#ifdef CONFIG_SLABINFO
	{
		.name = "kmem.slabinfo",
4527 4528 4529 4530
		.seq_start = slab_start,
		.seq_next = slab_next,
		.seq_stop = slab_stop,
		.seq_show = memcg_slab_show,
4531 4532
	},
#endif
4533
#endif
4534
	{ },	/* terminate */
4535
};
4536

4537 4538 4539 4540 4541
#ifdef CONFIG_MEMCG_SWAP
static struct cftype memsw_cgroup_files[] = {
	{
		.name = "memsw.usage_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_USAGE),
4542
		.read_u64 = mem_cgroup_read_u64,
4543 4544 4545 4546
	},
	{
		.name = "memsw.max_usage_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_MAX_USAGE),
4547
		.write = mem_cgroup_reset,
4548
		.read_u64 = mem_cgroup_read_u64,
4549 4550 4551 4552
	},
	{
		.name = "memsw.limit_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_LIMIT),
4553
		.write = mem_cgroup_write,
4554
		.read_u64 = mem_cgroup_read_u64,
4555 4556 4557 4558
	},
	{
		.name = "memsw.failcnt",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_FAILCNT),
4559
		.write = mem_cgroup_reset,
4560
		.read_u64 = mem_cgroup_read_u64,
4561 4562 4563 4564
	},
	{ },	/* terminate */
};
#endif
4565
static int alloc_mem_cgroup_per_zone_info(struct mem_cgroup *memcg, int node)
4566 4567
{
	struct mem_cgroup_per_node *pn;
4568
	struct mem_cgroup_per_zone *mz;
4569
	int zone, tmp = node;
4570 4571 4572 4573 4574 4575 4576 4577
	/*
	 * 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.
	 */
4578 4579
	if (!node_state(node, N_NORMAL_MEMORY))
		tmp = -1;
4580
	pn = kzalloc_node(sizeof(*pn), GFP_KERNEL, tmp);
4581 4582
	if (!pn)
		return 1;
4583 4584 4585

	for (zone = 0; zone < MAX_NR_ZONES; zone++) {
		mz = &pn->zoneinfo[zone];
4586
		lruvec_init(&mz->lruvec);
4587 4588
		mz->usage_in_excess = 0;
		mz->on_tree = false;
4589
		mz->memcg = memcg;
4590
	}
4591
	memcg->nodeinfo[node] = pn;
4592 4593 4594
	return 0;
}

4595
static void free_mem_cgroup_per_zone_info(struct mem_cgroup *memcg, int node)
4596
{
4597
	kfree(memcg->nodeinfo[node]);
4598 4599
}

4600 4601
static struct mem_cgroup *mem_cgroup_alloc(void)
{
4602
	struct mem_cgroup *memcg;
4603
	size_t size;
4604

4605 4606
	size = sizeof(struct mem_cgroup);
	size += nr_node_ids * sizeof(struct mem_cgroup_per_node *);
4607

4608
	memcg = kzalloc(size, GFP_KERNEL);
4609
	if (!memcg)
4610 4611
		return NULL;

4612 4613
	memcg->stat = alloc_percpu(struct mem_cgroup_stat_cpu);
	if (!memcg->stat)
4614
		goto out_free;
4615 4616
	spin_lock_init(&memcg->pcp_counter_lock);
	return memcg;
4617 4618

out_free:
4619
	kfree(memcg);
4620
	return NULL;
4621 4622
}

4623
/*
4624 4625 4626 4627 4628 4629 4630 4631
 * 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.
4632
 */
4633 4634

static void __mem_cgroup_free(struct mem_cgroup *memcg)
4635
{
4636
	int node;
4637

4638
	mem_cgroup_remove_from_trees(memcg);
4639 4640 4641 4642 4643 4644

	for_each_node(node)
		free_mem_cgroup_per_zone_info(memcg, node);

	free_percpu(memcg->stat);

4645
	disarm_static_keys(memcg);
4646
	kfree(memcg);
4647
}
4648

4649 4650 4651
/*
 * Returns the parent mem_cgroup in memcgroup hierarchy with hierarchy enabled.
 */
G
Glauber Costa 已提交
4652
struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
4653
{
4654
	if (!memcg->memory.parent)
4655
		return NULL;
4656
	return mem_cgroup_from_counter(memcg->memory.parent, memory);
4657
}
G
Glauber Costa 已提交
4658
EXPORT_SYMBOL(parent_mem_cgroup);
4659

4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672 4673 4674 4675 4676 4677 4678 4679 4680 4681 4682
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
4683
static struct cgroup_subsys_state * __ref
4684
mem_cgroup_css_alloc(struct cgroup_subsys_state *parent_css)
B
Balbir Singh 已提交
4685
{
4686
	struct mem_cgroup *memcg;
K
KAMEZAWA Hiroyuki 已提交
4687
	long error = -ENOMEM;
4688
	int node;
B
Balbir Singh 已提交
4689

4690 4691
	memcg = mem_cgroup_alloc();
	if (!memcg)
K
KAMEZAWA Hiroyuki 已提交
4692
		return ERR_PTR(error);
4693

B
Bob Liu 已提交
4694
	for_each_node(node)
4695
		if (alloc_mem_cgroup_per_zone_info(memcg, node))
4696
			goto free_out;
4697

4698
	/* root ? */
4699
	if (parent_css == NULL) {
4700
		root_mem_cgroup = memcg;
4701 4702 4703
		page_counter_init(&memcg->memory, NULL);
		page_counter_init(&memcg->memsw, NULL);
		page_counter_init(&memcg->kmem, NULL);
4704
	}
4705

4706 4707 4708 4709 4710
	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);
4711
	vmpressure_init(&memcg->vmpressure);
4712 4713
	INIT_LIST_HEAD(&memcg->event_list);
	spin_lock_init(&memcg->event_list_lock);
V
Vladimir Davydov 已提交
4714 4715 4716 4717
#ifdef CONFIG_MEMCG_KMEM
	memcg->kmemcg_id = -1;
	INIT_LIST_HEAD(&memcg->memcg_slab_caches);
#endif
4718 4719 4720 4721 4722 4723 4724 4725 4726

	return &memcg->css;

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

static int
4727
mem_cgroup_css_online(struct cgroup_subsys_state *css)
4728
{
4729
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
T
Tejun Heo 已提交
4730
	struct mem_cgroup *parent = mem_cgroup_from_css(css->parent);
4731
	int ret;
4732

4733
	if (css->id > MEM_CGROUP_ID_MAX)
4734 4735
		return -ENOSPC;

T
Tejun Heo 已提交
4736
	if (!parent)
4737 4738
		return 0;

4739
	mutex_lock(&memcg_create_mutex);
4740 4741 4742 4743 4744 4745

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

	if (parent->use_hierarchy) {
4746 4747 4748
		page_counter_init(&memcg->memory, &parent->memory);
		page_counter_init(&memcg->memsw, &parent->memsw);
		page_counter_init(&memcg->kmem, &parent->kmem);
4749

4750
		/*
4751 4752
		 * No need to take a reference to the parent because cgroup
		 * core guarantees its existence.
4753
		 */
4754
	} else {
4755 4756 4757
		page_counter_init(&memcg->memory, NULL);
		page_counter_init(&memcg->memsw, NULL);
		page_counter_init(&memcg->kmem, NULL);
4758 4759 4760 4761 4762
		/*
		 * Deeper hierachy with use_hierarchy == false doesn't make
		 * much sense so let cgroup subsystem know about this
		 * unfortunate state in our controller.
		 */
4763
		if (parent != root_mem_cgroup)
4764
			memory_cgrp_subsys.broken_hierarchy = true;
4765
	}
4766
	mutex_unlock(&memcg_create_mutex);
4767

4768 4769 4770 4771 4772 4773 4774 4775 4776 4777 4778 4779
	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
4780 4781
}

4782
static void mem_cgroup_css_offline(struct cgroup_subsys_state *css)
4783
{
4784
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
4785
	struct mem_cgroup_event *event, *tmp;
4786 4787 4788 4789 4790 4791

	/*
	 * Unregister events and notify userspace.
	 * Notify userspace about cgroup removing only after rmdir of cgroup
	 * directory to avoid race between userspace and kernelspace.
	 */
4792 4793
	spin_lock(&memcg->event_list_lock);
	list_for_each_entry_safe(event, tmp, &memcg->event_list, list) {
4794 4795 4796
		list_del_init(&event->list);
		schedule_work(&event->remove);
	}
4797
	spin_unlock(&memcg->event_list_lock);
4798

4799
	memcg_unregister_all_caches(memcg);
4800
	vmpressure_cleanup(&memcg->vmpressure);
4801 4802
}

4803
static void mem_cgroup_css_free(struct cgroup_subsys_state *css)
B
Balbir Singh 已提交
4804
{
4805
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
4806

4807
	memcg_destroy_kmem(memcg);
4808
	__mem_cgroup_free(memcg);
B
Balbir Singh 已提交
4809 4810
}

4811 4812 4813 4814 4815 4816 4817 4818 4819 4820 4821 4822 4823 4824 4825 4826 4827
/**
 * 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);

4828 4829 4830 4831
	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;
4832 4833
}

4834
#ifdef CONFIG_MMU
4835
/* Handlers for move charge at task migration. */
4836
static int mem_cgroup_do_precharge(unsigned long count)
4837
{
4838
	int ret;
4839 4840

	/* Try a single bulk charge without reclaim first */
4841
	ret = try_charge(mc.to, GFP_KERNEL & ~__GFP_WAIT, count);
4842
	if (!ret) {
4843 4844 4845
		mc.precharge += count;
		return ret;
	}
4846
	if (ret == -EINTR) {
4847
		cancel_charge(root_mem_cgroup, count);
4848 4849
		return ret;
	}
4850 4851

	/* Try charges one by one with reclaim */
4852
	while (count--) {
4853
		ret = try_charge(mc.to, GFP_KERNEL & ~__GFP_NORETRY, 1);
4854 4855 4856
		/*
		 * In case of failure, any residual charges against
		 * mc.to will be dropped by mem_cgroup_clear_mc()
4857 4858
		 * later on.  However, cancel any charges that are
		 * bypassed to root right away or they'll be lost.
4859
		 */
4860
		if (ret == -EINTR)
4861
			cancel_charge(root_mem_cgroup, 1);
4862 4863
		if (ret)
			return ret;
4864
		mc.precharge++;
4865
		cond_resched();
4866
	}
4867
	return 0;
4868 4869 4870
}

/**
4871
 * get_mctgt_type - get target type of moving charge
4872 4873 4874
 * @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
4875
 * @target: the pointer the target page or swap ent will be stored(can be NULL)
4876 4877 4878 4879 4880 4881
 *
 * 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).
4882 4883 4884
 *   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.
4885 4886 4887 4888 4889
 *
 * Called with pte lock held.
 */
union mc_target {
	struct page	*page;
4890
	swp_entry_t	ent;
4891 4892 4893
};

enum mc_target_type {
4894
	MC_TARGET_NONE = 0,
4895
	MC_TARGET_PAGE,
4896
	MC_TARGET_SWAP,
4897 4898
};

D
Daisuke Nishimura 已提交
4899 4900
static struct page *mc_handle_present_pte(struct vm_area_struct *vma,
						unsigned long addr, pte_t ptent)
4901
{
D
Daisuke Nishimura 已提交
4902
	struct page *page = vm_normal_page(vma, addr, ptent);
4903

D
Daisuke Nishimura 已提交
4904 4905 4906 4907
	if (!page || !page_mapped(page))
		return NULL;
	if (PageAnon(page)) {
		/* we don't move shared anon */
4908
		if (!move_anon())
D
Daisuke Nishimura 已提交
4909
			return NULL;
4910 4911
	} else if (!move_file())
		/* we ignore mapcount for file pages */
D
Daisuke Nishimura 已提交
4912 4913 4914 4915 4916 4917 4918
		return NULL;
	if (!get_page_unless_zero(page))
		return NULL;

	return page;
}

4919
#ifdef CONFIG_SWAP
D
Daisuke Nishimura 已提交
4920 4921 4922 4923 4924 4925 4926 4927
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;
4928 4929 4930 4931
	/*
	 * Because lookup_swap_cache() updates some statistics counter,
	 * we call find_get_page() with swapper_space directly.
	 */
4932
	page = find_get_page(swap_address_space(ent), ent.val);
D
Daisuke Nishimura 已提交
4933 4934 4935 4936 4937
	if (do_swap_account)
		entry->val = ent.val;

	return page;
}
4938 4939 4940 4941 4942 4943 4944
#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
4945

4946 4947 4948 4949 4950 4951 4952 4953 4954 4955 4956 4957 4958 4959 4960 4961 4962 4963 4964
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). */
4965 4966
#ifdef CONFIG_SWAP
	/* shmem/tmpfs may report page out on swap: account for that too. */
4967 4968 4969 4970 4971 4972 4973 4974 4975 4976 4977 4978
	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);
4979
#endif
4980 4981 4982
	return page;
}

4983
static enum mc_target_type get_mctgt_type(struct vm_area_struct *vma,
D
Daisuke Nishimura 已提交
4984 4985 4986
		unsigned long addr, pte_t ptent, union mc_target *target)
{
	struct page *page = NULL;
4987
	enum mc_target_type ret = MC_TARGET_NONE;
D
Daisuke Nishimura 已提交
4988 4989 4990 4991 4992 4993
	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);
4994 4995
	else if (pte_none(ptent) || pte_file(ptent))
		page = mc_handle_file_pte(vma, addr, ptent, &ent);
D
Daisuke Nishimura 已提交
4996 4997

	if (!page && !ent.val)
4998
		return ret;
4999 5000
	if (page) {
		/*
5001
		 * Do only loose check w/o serialization.
5002
		 * mem_cgroup_move_account() checks the page is valid or
5003
		 * not under LRU exclusion.
5004
		 */
5005
		if (page->mem_cgroup == mc.from) {
5006 5007 5008 5009 5010 5011 5012
			ret = MC_TARGET_PAGE;
			if (target)
				target->page = page;
		}
		if (!ret || !target)
			put_page(page);
	}
D
Daisuke Nishimura 已提交
5013 5014
	/* 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
5015
	    mem_cgroup_id(mc.from) == lookup_swap_cgroup_id(ent)) {
5016 5017 5018
		ret = MC_TARGET_SWAP;
		if (target)
			target->ent = ent;
5019 5020 5021 5022
	}
	return ret;
}

5023 5024 5025 5026 5027 5028 5029 5030 5031 5032 5033 5034 5035
#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;
	enum mc_target_type ret = MC_TARGET_NONE;

	page = pmd_page(pmd);
5036
	VM_BUG_ON_PAGE(!page || !PageHead(page), page);
5037 5038
	if (!move_anon())
		return ret;
5039
	if (page->mem_cgroup == mc.from) {
5040 5041 5042 5043 5044 5045 5046 5047 5048 5049 5050 5051 5052 5053 5054 5055
		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

5056 5057 5058 5059 5060 5061 5062 5063
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;

5064
	if (pmd_trans_huge_lock(pmd, vma, &ptl) == 1) {
5065 5066
		if (get_mctgt_type_thp(vma, addr, *pmd, NULL) == MC_TARGET_PAGE)
			mc.precharge += HPAGE_PMD_NR;
5067
		spin_unlock(ptl);
5068
		return 0;
5069
	}
5070

5071 5072
	if (pmd_trans_unstable(pmd))
		return 0;
5073 5074
	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
	for (; addr != end; pte++, addr += PAGE_SIZE)
5075
		if (get_mctgt_type(vma, addr, *pte, NULL))
5076 5077 5078 5079
			mc.precharge++;	/* increment precharge temporarily */
	pte_unmap_unlock(pte - 1, ptl);
	cond_resched();

5080 5081 5082
	return 0;
}

5083 5084 5085 5086 5087
static unsigned long mem_cgroup_count_precharge(struct mm_struct *mm)
{
	unsigned long precharge;
	struct vm_area_struct *vma;

5088
	down_read(&mm->mmap_sem);
5089 5090 5091 5092 5093 5094 5095 5096 5097 5098 5099
	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);
	}
5100
	up_read(&mm->mmap_sem);
5101 5102 5103 5104 5105 5106 5107 5108 5109

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

	return precharge;
}

static int mem_cgroup_precharge_mc(struct mm_struct *mm)
{
5110 5111 5112 5113 5114
	unsigned long precharge = mem_cgroup_count_precharge(mm);

	VM_BUG_ON(mc.moving_task);
	mc.moving_task = current;
	return mem_cgroup_do_precharge(precharge);
5115 5116
}

5117 5118
/* cancels all extra charges on mc.from and mc.to, and wakes up all waiters. */
static void __mem_cgroup_clear_mc(void)
5119
{
5120 5121 5122
	struct mem_cgroup *from = mc.from;
	struct mem_cgroup *to = mc.to;

5123
	/* we must uncharge all the leftover precharges from mc.to */
5124
	if (mc.precharge) {
5125
		cancel_charge(mc.to, mc.precharge);
5126 5127 5128 5129 5130 5131 5132
		mc.precharge = 0;
	}
	/*
	 * we didn't uncharge from mc.from at mem_cgroup_move_account(), so
	 * we must uncharge here.
	 */
	if (mc.moved_charge) {
5133
		cancel_charge(mc.from, mc.moved_charge);
5134
		mc.moved_charge = 0;
5135
	}
5136 5137 5138
	/* we must fixup refcnts and charges */
	if (mc.moved_swap) {
		/* uncharge swap account from the old cgroup */
5139
		if (!mem_cgroup_is_root(mc.from))
5140
			page_counter_uncharge(&mc.from->memsw, mc.moved_swap);
5141

5142
		/*
5143 5144
		 * we charged both to->memory and to->memsw, so we
		 * should uncharge to->memory.
5145
		 */
5146
		if (!mem_cgroup_is_root(mc.to))
5147 5148
			page_counter_uncharge(&mc.to->memory, mc.moved_swap);

5149
		css_put_many(&mc.from->css, mc.moved_swap);
5150

L
Li Zefan 已提交
5151
		/* we've already done css_get(mc.to) */
5152 5153
		mc.moved_swap = 0;
	}
5154 5155 5156 5157 5158 5159 5160 5161 5162 5163 5164 5165 5166
	memcg_oom_recover(from);
	memcg_oom_recover(to);
	wake_up_all(&mc.waitq);
}

static void mem_cgroup_clear_mc(void)
{
	/*
	 * we must clear moving_task before waking up waiters at the end of
	 * task migration.
	 */
	mc.moving_task = NULL;
	__mem_cgroup_clear_mc();
5167
	spin_lock(&mc.lock);
5168 5169
	mc.from = NULL;
	mc.to = NULL;
5170
	spin_unlock(&mc.lock);
5171 5172
}

5173
static int mem_cgroup_can_attach(struct cgroup_subsys_state *css,
5174
				 struct cgroup_taskset *tset)
5175
{
5176
	struct task_struct *p = cgroup_taskset_first(tset);
5177
	int ret = 0;
5178
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
5179
	unsigned long move_charge_at_immigrate;
5180

5181 5182 5183 5184 5185 5186 5187
	/*
	 * 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) {
5188 5189 5190
		struct mm_struct *mm;
		struct mem_cgroup *from = mem_cgroup_from_task(p);

5191
		VM_BUG_ON(from == memcg);
5192 5193 5194 5195 5196

		mm = get_task_mm(p);
		if (!mm)
			return 0;
		/* We move charges only when we move a owner of the mm */
5197 5198 5199 5200
		if (mm->owner == p) {
			VM_BUG_ON(mc.from);
			VM_BUG_ON(mc.to);
			VM_BUG_ON(mc.precharge);
5201
			VM_BUG_ON(mc.moved_charge);
5202
			VM_BUG_ON(mc.moved_swap);
5203

5204
			spin_lock(&mc.lock);
5205
			mc.from = from;
5206
			mc.to = memcg;
5207
			mc.immigrate_flags = move_charge_at_immigrate;
5208
			spin_unlock(&mc.lock);
5209
			/* We set mc.moving_task later */
5210 5211 5212 5213

			ret = mem_cgroup_precharge_mc(mm);
			if (ret)
				mem_cgroup_clear_mc();
5214 5215
		}
		mmput(mm);
5216 5217 5218 5219
	}
	return ret;
}

5220
static void mem_cgroup_cancel_attach(struct cgroup_subsys_state *css,
5221
				     struct cgroup_taskset *tset)
5222
{
5223 5224
	if (mc.to)
		mem_cgroup_clear_mc();
5225 5226
}

5227 5228 5229
static int mem_cgroup_move_charge_pte_range(pmd_t *pmd,
				unsigned long addr, unsigned long end,
				struct mm_walk *walk)
5230
{
5231 5232 5233 5234
	int ret = 0;
	struct vm_area_struct *vma = walk->private;
	pte_t *pte;
	spinlock_t *ptl;
5235 5236 5237
	enum mc_target_type target_type;
	union mc_target target;
	struct page *page;
5238

5239 5240 5241 5242 5243 5244 5245 5246 5247 5248
	/*
	 * 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.
	 */
5249
	if (pmd_trans_huge_lock(pmd, vma, &ptl) == 1) {
5250
		if (mc.precharge < HPAGE_PMD_NR) {
5251
			spin_unlock(ptl);
5252 5253 5254 5255 5256 5257 5258
			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)) {
				if (!mem_cgroup_move_account(page, HPAGE_PMD_NR,
5259
							     mc.from, mc.to)) {
5260 5261 5262 5263 5264 5265 5266
					mc.precharge -= HPAGE_PMD_NR;
					mc.moved_charge += HPAGE_PMD_NR;
				}
				putback_lru_page(page);
			}
			put_page(page);
		}
5267
		spin_unlock(ptl);
5268
		return 0;
5269 5270
	}

5271 5272
	if (pmd_trans_unstable(pmd))
		return 0;
5273 5274 5275 5276
retry:
	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
	for (; addr != end; addr += PAGE_SIZE) {
		pte_t ptent = *(pte++);
5277
		swp_entry_t ent;
5278 5279 5280 5281

		if (!mc.precharge)
			break;

5282
		switch (get_mctgt_type(vma, addr, ptent, &target)) {
5283 5284 5285 5286
		case MC_TARGET_PAGE:
			page = target.page;
			if (isolate_lru_page(page))
				goto put;
5287
			if (!mem_cgroup_move_account(page, 1, mc.from, mc.to)) {
5288
				mc.precharge--;
5289 5290
				/* we uncharge from mc.from later. */
				mc.moved_charge++;
5291 5292
			}
			putback_lru_page(page);
5293
put:			/* get_mctgt_type() gets the page */
5294 5295
			put_page(page);
			break;
5296 5297
		case MC_TARGET_SWAP:
			ent = target.ent;
5298
			if (!mem_cgroup_move_swap_account(ent, mc.from, mc.to)) {
5299
				mc.precharge--;
5300 5301 5302
				/* we fixup refcnts and charges later. */
				mc.moved_swap++;
			}
5303
			break;
5304 5305 5306 5307 5308 5309 5310 5311 5312 5313 5314 5315 5316 5317
		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.
		 */
5318
		ret = mem_cgroup_do_precharge(1);
5319 5320 5321 5322 5323 5324 5325 5326 5327 5328 5329 5330
		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();
5331 5332 5333 5334 5335 5336 5337
	/*
	 * Signal mem_cgroup_begin_page_stat() to take the memcg's
	 * move_lock while we're moving its pages to another memcg.
	 * Then wait for already started RCU-only updates to finish.
	 */
	atomic_inc(&mc.from->moving_account);
	synchronize_rcu();
5338 5339 5340 5341 5342 5343 5344 5345 5346 5347 5348 5349 5350
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;
	}
5351 5352 5353 5354 5355 5356 5357 5358 5359 5360 5361 5362 5363 5364 5365 5366 5367 5368
	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;
	}
5369
	up_read(&mm->mmap_sem);
5370
	atomic_dec(&mc.from->moving_account);
5371 5372
}

5373
static void mem_cgroup_move_task(struct cgroup_subsys_state *css,
5374
				 struct cgroup_taskset *tset)
B
Balbir Singh 已提交
5375
{
5376
	struct task_struct *p = cgroup_taskset_first(tset);
5377
	struct mm_struct *mm = get_task_mm(p);
5378 5379

	if (mm) {
5380 5381
		if (mc.to)
			mem_cgroup_move_charge(mm);
5382 5383
		mmput(mm);
	}
5384 5385
	if (mc.to)
		mem_cgroup_clear_mc();
B
Balbir Singh 已提交
5386
}
5387
#else	/* !CONFIG_MMU */
5388
static int mem_cgroup_can_attach(struct cgroup_subsys_state *css,
5389
				 struct cgroup_taskset *tset)
5390 5391 5392
{
	return 0;
}
5393
static void mem_cgroup_cancel_attach(struct cgroup_subsys_state *css,
5394
				     struct cgroup_taskset *tset)
5395 5396
{
}
5397
static void mem_cgroup_move_task(struct cgroup_subsys_state *css,
5398
				 struct cgroup_taskset *tset)
5399 5400 5401
{
}
#endif
B
Balbir Singh 已提交
5402

5403 5404
/*
 * Cgroup retains root cgroups across [un]mount cycles making it necessary
5405 5406
 * to verify whether we're attached to the default hierarchy on each mount
 * attempt.
5407
 */
5408
static void mem_cgroup_bind(struct cgroup_subsys_state *root_css)
5409 5410
{
	/*
5411
	 * use_hierarchy is forced on the default hierarchy.  cgroup core
5412 5413 5414
	 * guarantees that @root doesn't have any children, so turning it
	 * on for the root memcg is enough.
	 */
5415
	if (cgroup_on_dfl(root_css->cgroup))
5416
		mem_cgroup_from_css(root_css)->use_hierarchy = true;
5417 5418
}

5419
struct cgroup_subsys memory_cgrp_subsys = {
5420
	.css_alloc = mem_cgroup_css_alloc,
5421
	.css_online = mem_cgroup_css_online,
5422 5423
	.css_offline = mem_cgroup_css_offline,
	.css_free = mem_cgroup_css_free,
5424
	.css_reset = mem_cgroup_css_reset,
5425 5426
	.can_attach = mem_cgroup_can_attach,
	.cancel_attach = mem_cgroup_cancel_attach,
B
Balbir Singh 已提交
5427
	.attach = mem_cgroup_move_task,
5428
	.bind = mem_cgroup_bind,
5429
	.legacy_cftypes = mem_cgroup_files,
5430
	.early_init = 0,
B
Balbir Singh 已提交
5431
};
5432

A
Andrew Morton 已提交
5433
#ifdef CONFIG_MEMCG_SWAP
5434 5435
static int __init enable_swap_account(char *s)
{
5436
	if (!strcmp(s, "1"))
5437
		really_do_swap_account = 1;
5438
	else if (!strcmp(s, "0"))
5439 5440 5441
		really_do_swap_account = 0;
	return 1;
}
5442
__setup("swapaccount=", enable_swap_account);
5443

5444 5445
static void __init memsw_file_init(void)
{
5446 5447
	WARN_ON(cgroup_add_legacy_cftypes(&memory_cgrp_subsys,
					  memsw_cgroup_files));
5448 5449 5450 5451 5452 5453 5454 5455
}

static void __init enable_swap_cgroup(void)
{
	if (!mem_cgroup_disabled() && really_do_swap_account) {
		do_swap_account = 1;
		memsw_file_init();
	}
5456
}
5457

5458
#else
5459
static void __init enable_swap_cgroup(void)
5460 5461
{
}
5462
#endif
5463

5464 5465 5466 5467 5468 5469 5470 5471 5472 5473
#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)
{
5474
	struct mem_cgroup *memcg;
5475 5476 5477 5478 5479 5480 5481 5482
	unsigned short oldid;

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

	if (!do_swap_account)
		return;

5483
	memcg = page->mem_cgroup;
5484 5485

	/* Readahead page, never charged */
5486
	if (!memcg)
5487 5488
		return;

5489
	oldid = swap_cgroup_record(entry, mem_cgroup_id(memcg));
5490
	VM_BUG_ON_PAGE(oldid, page);
5491 5492
	mem_cgroup_swap_statistics(memcg, true);

5493
	page->mem_cgroup = NULL;
5494

5495 5496 5497 5498 5499
	if (!mem_cgroup_is_root(memcg))
		page_counter_uncharge(&memcg->memory, 1);

	/* XXX: caller holds IRQ-safe mapping->tree_lock */
	VM_BUG_ON(!irqs_disabled());
5500

5501 5502
	mem_cgroup_charge_statistics(memcg, page, -1);
	memcg_check_events(memcg, page);
5503 5504 5505 5506 5507 5508 5509 5510 5511 5512 5513 5514 5515 5516 5517 5518 5519 5520 5521 5522
}

/**
 * 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) {
5523
		if (!mem_cgroup_is_root(memcg))
5524
			page_counter_uncharge(&memcg->memsw, 1);
5525 5526 5527 5528 5529 5530 5531
		mem_cgroup_swap_statistics(memcg, false);
		css_put(&memcg->css);
	}
	rcu_read_unlock();
}
#endif

5532 5533 5534 5535 5536 5537 5538 5539 5540 5541 5542 5543 5544 5545 5546 5547 5548 5549 5550 5551 5552 5553 5554 5555 5556 5557 5558 5559 5560 5561 5562 5563 5564 5565 5566
/**
 * 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)) {
		/*
		 * 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.
		 */
5567
		if (page->mem_cgroup)
5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578 5579 5580 5581 5582 5583 5584 5585 5586 5587 5588 5589 5590 5591 5592 5593 5594 5595 5596 5597 5598 5599 5600 5601 5602 5603 5604 5605 5606 5607 5608 5609 5610 5611 5612 5613 5614 5615 5616 5617 5618 5619 5620 5621 5622 5623 5624 5625 5626 5627
			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;

5628 5629
	commit_charge(page, memcg, lrucare);

5630 5631 5632 5633 5634
	if (PageTransHuge(page)) {
		nr_pages <<= compound_order(page);
		VM_BUG_ON_PAGE(!PageTransHuge(page), page);
	}

5635 5636 5637 5638
	local_irq_disable();
	mem_cgroup_charge_statistics(memcg, page, nr_pages);
	memcg_check_events(memcg, page);
	local_irq_enable();
5639 5640 5641 5642 5643 5644 5645 5646 5647 5648 5649 5650 5651 5652 5653 5654 5655 5656 5657 5658 5659 5660 5661 5662 5663 5664 5665 5666 5667 5668 5669 5670 5671 5672 5673 5674 5675 5676 5677 5678 5679

	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);
}

5680 5681 5682 5683
static void uncharge_batch(struct mem_cgroup *memcg, unsigned long pgpgout,
			   unsigned long nr_anon, unsigned long nr_file,
			   unsigned long nr_huge, struct page *dummy_page)
{
5684
	unsigned long nr_pages = nr_anon + nr_file;
5685 5686
	unsigned long flags;

5687
	if (!mem_cgroup_is_root(memcg)) {
5688 5689 5690
		page_counter_uncharge(&memcg->memory, nr_pages);
		if (do_swap_account)
			page_counter_uncharge(&memcg->memsw, nr_pages);
5691 5692
		memcg_oom_recover(memcg);
	}
5693 5694 5695 5696 5697 5698

	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);
5699
	__this_cpu_add(memcg->stat->nr_page_events, nr_pages);
5700 5701
	memcg_check_events(memcg, dummy_page);
	local_irq_restore(flags);
5702 5703

	if (!mem_cgroup_is_root(memcg))
5704
		css_put_many(&memcg->css, nr_pages);
5705 5706 5707 5708 5709 5710 5711 5712 5713 5714 5715 5716 5717 5718 5719 5720 5721 5722 5723 5724 5725 5726
}

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

	next = page_list->next;
	do {
		unsigned int nr_pages = 1;

		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);

5727
		if (!page->mem_cgroup)
5728 5729 5730 5731
			continue;

		/*
		 * Nobody should be changing or seriously looking at
5732
		 * page->mem_cgroup at this point, we have fully
5733
		 * exclusive access to the page.
5734 5735
		 */

5736
		if (memcg != page->mem_cgroup) {
5737
			if (memcg) {
5738 5739 5740
				uncharge_batch(memcg, pgpgout, nr_anon, nr_file,
					       nr_huge, page);
				pgpgout = nr_anon = nr_file = nr_huge = 0;
5741
			}
5742
			memcg = page->mem_cgroup;
5743 5744 5745 5746 5747 5748 5749 5750 5751 5752 5753 5754 5755
		}

		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;

5756
		page->mem_cgroup = NULL;
5757 5758 5759 5760 5761

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

	if (memcg)
5762 5763
		uncharge_batch(memcg, pgpgout, nr_anon, nr_file,
			       nr_huge, page);
5764 5765
}

5766 5767 5768 5769 5770 5771 5772 5773 5774 5775 5776 5777
/**
 * 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)
{
	if (mem_cgroup_disabled())
		return;

5778
	/* Don't touch page->lru of any random page, pre-check: */
5779
	if (!page->mem_cgroup)
5780 5781
		return;

5782 5783 5784
	INIT_LIST_HEAD(&page->lru);
	uncharge_list(&page->lru);
}
5785

5786 5787 5788 5789 5790 5791 5792 5793 5794 5795 5796
/**
 * 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;
5797

5798 5799
	if (!list_empty(page_list))
		uncharge_list(page_list);
5800 5801 5802 5803 5804 5805 5806 5807 5808 5809 5810 5811 5812 5813 5814
}

/**
 * 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)
{
5815
	struct mem_cgroup *memcg;
5816 5817 5818 5819 5820 5821 5822
	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);
5823 5824
	VM_BUG_ON_PAGE(PageTransHuge(oldpage) != PageTransHuge(newpage),
		       newpage);
5825 5826 5827 5828 5829

	if (mem_cgroup_disabled())
		return;

	/* Page cache replacement: new page already charged? */
5830
	if (newpage->mem_cgroup)
5831 5832
		return;

5833 5834 5835 5836 5837 5838
	/*
	 * Swapcache readahead pages can get migrated before being
	 * charged, and migration from compaction can happen to an
	 * uncharged page when the PFN walker finds a page that
	 * reclaim just put back on the LRU but has not released yet.
	 */
5839
	memcg = oldpage->mem_cgroup;
5840
	if (!memcg)
5841 5842 5843 5844 5845
		return;

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

5846
	oldpage->mem_cgroup = NULL;
5847 5848 5849 5850

	if (lrucare)
		unlock_page_lru(oldpage, isolated);

5851
	commit_charge(newpage, memcg, lrucare);
5852 5853
}

5854
/*
5855 5856 5857 5858 5859 5860
 * 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.
5861 5862 5863 5864
 */
static int __init mem_cgroup_init(void)
{
	hotcpu_notifier(memcg_cpu_hotplug_callback, 0);
5865
	enable_swap_cgroup();
5866
	mem_cgroup_soft_limit_tree_init();
5867
	memcg_stock_init();
5868 5869 5870
	return 0;
}
subsys_initcall(mem_cgroup_init);