memcontrol.c 185.6 KB
Newer Older
B
Balbir Singh 已提交
1 2 3 4 5
/* memcontrol.c - Memory Controller
 *
 * Copyright IBM Corporation, 2007
 * Author Balbir Singh <balbir@linux.vnet.ibm.com>
 *
6 7 8
 * Copyright 2007 OpenVZ SWsoft Inc
 * Author: Pavel Emelianov <xemul@openvz.org>
 *
9 10 11 12
 * Memory thresholds
 * Copyright (C) 2009 Nokia Corporation
 * Author: Kirill A. Shutemov
 *
13 14 15 16
 * Kernel Memory Controller
 * Copyright (C) 2012 Parallels Inc. and Google Inc.
 * Authors: Glauber Costa and Suleiman Souhlal
 *
B
Balbir Singh 已提交
17 18 19 20 21 22 23 24 25 26 27 28 29 30
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 */

#include <linux/res_counter.h>
#include <linux/memcontrol.h>
#include <linux/cgroup.h>
31
#include <linux/mm.h>
32
#include <linux/hugetlb.h>
K
KAMEZAWA Hiroyuki 已提交
33
#include <linux/pagemap.h>
34
#include <linux/smp.h>
35
#include <linux/page-flags.h>
36
#include <linux/backing-dev.h>
37 38
#include <linux/bit_spinlock.h>
#include <linux/rcupdate.h>
39
#include <linux/limits.h>
40
#include <linux/export.h>
41
#include <linux/mutex.h>
42
#include <linux/rbtree.h>
43
#include <linux/slab.h>
44
#include <linux/swap.h>
45
#include <linux/swapops.h>
46
#include <linux/spinlock.h>
47 48
#include <linux/eventfd.h>
#include <linux/sort.h>
49
#include <linux/fs.h>
50
#include <linux/seq_file.h>
51
#include <linux/vmalloc.h>
52
#include <linux/vmpressure.h>
53
#include <linux/mm_inline.h>
54
#include <linux/page_cgroup.h>
55
#include <linux/cpu.h>
56
#include <linux/oom.h>
K
KAMEZAWA Hiroyuki 已提交
57
#include "internal.h"
G
Glauber Costa 已提交
58
#include <net/sock.h>
M
Michal Hocko 已提交
59
#include <net/ip.h>
G
Glauber Costa 已提交
60
#include <net/tcp_memcontrol.h>
B
Balbir Singh 已提交
61

62 63
#include <asm/uaccess.h>

64 65
#include <trace/events/vmscan.h>

66
struct cgroup_subsys mem_cgroup_subsys __read_mostly;
67 68
EXPORT_SYMBOL(mem_cgroup_subsys);

69
#define MEM_CGROUP_RECLAIM_RETRIES	5
70
static struct mem_cgroup *root_mem_cgroup __read_mostly;
B
Balbir Singh 已提交
71

A
Andrew Morton 已提交
72
#ifdef CONFIG_MEMCG_SWAP
L
Li Zefan 已提交
73
/* Turned on only when memory cgroup is enabled && really_do_swap_account = 1 */
74
int do_swap_account __read_mostly;
75 76

/* for remember boot option*/
A
Andrew Morton 已提交
77
#ifdef CONFIG_MEMCG_SWAP_ENABLED
78 79 80 81 82
static int really_do_swap_account __initdata = 1;
#else
static int really_do_swap_account __initdata = 0;
#endif

83
#else
84
#define do_swap_account		0
85 86 87
#endif


88 89 90 91 92 93 94
/*
 * Statistics for memory cgroup.
 */
enum mem_cgroup_stat_index {
	/*
	 * For MEM_CONTAINER_TYPE_ALL, usage = pagecache + rss.
	 */
95 96 97 98 99
	MEM_CGROUP_STAT_CACHE,		/* # of pages charged as cache */
	MEM_CGROUP_STAT_RSS,		/* # of pages charged as anon rss */
	MEM_CGROUP_STAT_RSS_HUGE,	/* # of pages charged as anon huge */
	MEM_CGROUP_STAT_FILE_MAPPED,	/* # of pages charged as file rss */
	MEM_CGROUP_STAT_SWAP,		/* # of pages, swapped out */
100 101 102
	MEM_CGROUP_STAT_NSTATS,
};

103 104 105
static const char * const mem_cgroup_stat_names[] = {
	"cache",
	"rss",
106
	"rss_huge",
107 108 109 110
	"mapped_file",
	"swap",
};

111 112 113
enum mem_cgroup_events_index {
	MEM_CGROUP_EVENTS_PGPGIN,	/* # of pages paged in */
	MEM_CGROUP_EVENTS_PGPGOUT,	/* # of pages paged out */
114 115
	MEM_CGROUP_EVENTS_PGFAULT,	/* # of page-faults */
	MEM_CGROUP_EVENTS_PGMAJFAULT,	/* # of major page-faults */
116 117
	MEM_CGROUP_EVENTS_NSTATS,
};
118 119 120 121 122 123 124 125

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

126 127 128 129 130 131 132 133
static const char * const mem_cgroup_lru_names[] = {
	"inactive_anon",
	"active_anon",
	"inactive_file",
	"active_file",
	"unevictable",
};

134 135 136 137 138 139 140 141 142
/*
 * 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,
	MEM_CGROUP_TARGET_SOFTLIMIT,
143
	MEM_CGROUP_TARGET_NUMAINFO,
144 145
	MEM_CGROUP_NTARGETS,
};
146 147 148
#define THRESHOLDS_EVENTS_TARGET 128
#define SOFTLIMIT_EVENTS_TARGET 1024
#define NUMAINFO_EVENTS_TARGET	1024
149

150
struct mem_cgroup_stat_cpu {
151
	long count[MEM_CGROUP_STAT_NSTATS];
152
	unsigned long events[MEM_CGROUP_EVENTS_NSTATS];
153
	unsigned long nr_page_events;
154
	unsigned long targets[MEM_CGROUP_NTARGETS];
155 156
};

157
struct mem_cgroup_reclaim_iter {
M
Michal Hocko 已提交
158 159 160 161
	/*
	 * last scanned hierarchy member. Valid only if last_dead_count
	 * matches memcg->dead_count of the hierarchy root group.
	 */
162
	struct mem_cgroup *last_visited;
M
Michal Hocko 已提交
163 164
	unsigned long last_dead_count;

165 166 167 168
	/* scan generation, increased every round-trip */
	unsigned int generation;
};

169 170 171 172
/*
 * per-zone information in memory controller.
 */
struct mem_cgroup_per_zone {
173
	struct lruvec		lruvec;
174
	unsigned long		lru_size[NR_LRU_LISTS];
K
KOSAKI Motohiro 已提交
175

176 177
	struct mem_cgroup_reclaim_iter reclaim_iter[DEF_PRIORITY + 1];

178 179 180 181
	struct rb_node		tree_node;	/* RB tree node */
	unsigned long long	usage_in_excess;/* Set to the value by which */
						/* the soft limit is exceeded*/
	bool			on_tree;
182
	struct mem_cgroup	*memcg;		/* Back pointer, we cannot */
183
						/* use container_of	   */
184 185 186 187 188 189
};

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

190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209
/*
 * 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;

210 211 212 213 214
struct mem_cgroup_threshold {
	struct eventfd_ctx *eventfd;
	u64 threshold;
};

K
KAMEZAWA Hiroyuki 已提交
215
/* For threshold */
216
struct mem_cgroup_threshold_ary {
217
	/* An array index points to threshold just below or equal to usage. */
218
	int current_threshold;
219 220 221 222 223
	/* Size of entries[] */
	unsigned int size;
	/* Array of thresholds */
	struct mem_cgroup_threshold entries[0];
};
224 225 226 227 228 229 230 231 232 233 234 235

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
236 237 238 239 240
/* for OOM */
struct mem_cgroup_eventfd_list {
	struct list_head list;
	struct eventfd_ctx *eventfd;
};
241

242 243
static void mem_cgroup_threshold(struct mem_cgroup *memcg);
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg);
244

B
Balbir Singh 已提交
245 246 247 248 249 250 251
/*
 * 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
252 253 254
 * 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
255 256 257 258 259 260 261
 */
struct mem_cgroup {
	struct cgroup_subsys_state css;
	/*
	 * the counter to account for memory usage
	 */
	struct res_counter res;
262

263 264 265
	/* vmpressure notifications */
	struct vmpressure vmpressure;

266 267 268 269
	/*
	 * the counter to account for mem+swap usage.
	 */
	struct res_counter memsw;
270

271 272 273 274
	/*
	 * the counter to account for kernel memory usage.
	 */
	struct res_counter kmem;
275 276 277 278
	/*
	 * Should the accounting and control be hierarchical, per subtree?
	 */
	bool use_hierarchy;
279
	unsigned long kmem_account_flags; /* See KMEM_ACCOUNTED_*, below */
280 281 282 283

	bool		oom_lock;
	atomic_t	under_oom;

284
	int	swappiness;
285 286
	/* OOM-Killer disable */
	int		oom_kill_disable;
K
KOSAKI Motohiro 已提交
287

288 289 290
	/* set when res.limit == memsw.limit */
	bool		memsw_is_minimum;

291 292 293 294
	/* protect arrays of thresholds */
	struct mutex thresholds_lock;

	/* thresholds for memory usage. RCU-protected */
295
	struct mem_cgroup_thresholds thresholds;
296

297
	/* thresholds for mem+swap usage. RCU-protected */
298
	struct mem_cgroup_thresholds memsw_thresholds;
299

K
KAMEZAWA Hiroyuki 已提交
300 301
	/* For oom notifier event fd */
	struct list_head oom_notify;
302

303 304 305 306 307
	/*
	 * Should we move charges of a task when a task is moved into this
	 * mem_cgroup ? And what type of charges should we move ?
	 */
	unsigned long 	move_charge_at_immigrate;
308 309 310 311
	/*
	 * set > 0 if pages under this cgroup are moving to other cgroup.
	 */
	atomic_t	moving_account;
312 313
	/* taken only while moving_account > 0 */
	spinlock_t	move_lock;
314
	/*
315
	 * percpu counter.
316
	 */
317
	struct mem_cgroup_stat_cpu __percpu *stat;
318 319 320 321 322 323
	/*
	 * 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
324

M
Michal Hocko 已提交
325
	atomic_t	dead_count;
M
Michal Hocko 已提交
326
#if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_INET)
G
Glauber Costa 已提交
327 328
	struct tcp_memcontrol tcp_mem;
#endif
329 330 331 332 333 334 335 336
#if defined(CONFIG_MEMCG_KMEM)
	/* analogous to slab_common's slab_caches list. per-memcg */
	struct list_head memcg_slab_caches;
	/* Not a spinlock, we can take a lot of time walking the list */
	struct mutex slab_caches_mutex;
        /* Index in the kmem_cache->memcg_params->memcg_caches array */
	int kmemcg_id;
#endif
337 338 339 340 341 342 343

	int last_scanned_node;
#if MAX_NUMNODES > 1
	nodemask_t	scan_nodes;
	atomic_t	numainfo_events;
	atomic_t	numainfo_updating;
#endif
344

345 346
	struct mem_cgroup_per_node *nodeinfo[0];
	/* WARNING: nodeinfo must be the last member here */
B
Balbir Singh 已提交
347 348
};

349 350 351 352 353 354
static size_t memcg_size(void)
{
	return sizeof(struct mem_cgroup) +
		nr_node_ids * sizeof(struct mem_cgroup_per_node);
}

355 356 357
/* internal only representation about the status of kmem accounting. */
enum {
	KMEM_ACCOUNTED_ACTIVE = 0, /* accounted by this cgroup itself */
358
	KMEM_ACCOUNTED_ACTIVATED, /* static key enabled. */
359
	KMEM_ACCOUNTED_DEAD, /* dead memcg with pending kmem charges */
360 361
};

362 363 364
/* We account when limit is on, but only after call sites are patched */
#define KMEM_ACCOUNTED_MASK \
		((1 << KMEM_ACCOUNTED_ACTIVE) | (1 << KMEM_ACCOUNTED_ACTIVATED))
365 366 367 368 369 370

#ifdef CONFIG_MEMCG_KMEM
static inline void memcg_kmem_set_active(struct mem_cgroup *memcg)
{
	set_bit(KMEM_ACCOUNTED_ACTIVE, &memcg->kmem_account_flags);
}
371 372 373 374 375 376

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

377 378 379 380 381
static void memcg_kmem_set_activated(struct mem_cgroup *memcg)
{
	set_bit(KMEM_ACCOUNTED_ACTIVATED, &memcg->kmem_account_flags);
}

382 383 384 385 386
static void memcg_kmem_clear_activated(struct mem_cgroup *memcg)
{
	clear_bit(KMEM_ACCOUNTED_ACTIVATED, &memcg->kmem_account_flags);
}

387 388
static void memcg_kmem_mark_dead(struct mem_cgroup *memcg)
{
389 390 391 392 393
	/*
	 * Our caller must use css_get() first, because memcg_uncharge_kmem()
	 * will call css_put() if it sees the memcg is dead.
	 */
	smp_wmb();
394 395 396 397 398 399 400 401 402
	if (test_bit(KMEM_ACCOUNTED_ACTIVE, &memcg->kmem_account_flags))
		set_bit(KMEM_ACCOUNTED_DEAD, &memcg->kmem_account_flags);
}

static bool memcg_kmem_test_and_clear_dead(struct mem_cgroup *memcg)
{
	return test_and_clear_bit(KMEM_ACCOUNTED_DEAD,
				  &memcg->kmem_account_flags);
}
403 404
#endif

405 406
/* Stuffs for move charges at task migration. */
/*
407 408
 * Types of charges to be moved. "move_charge_at_immitgrate" and
 * "immigrate_flags" are treated as a left-shifted bitmap of these types.
409 410
 */
enum move_type {
411
	MOVE_CHARGE_TYPE_ANON,	/* private anonymous page and swap of it */
412
	MOVE_CHARGE_TYPE_FILE,	/* file page(including tmpfs) and swap of it */
413 414 415
	NR_MOVE_TYPE,
};

416 417
/* "mc" and its members are protected by cgroup_mutex */
static struct move_charge_struct {
418
	spinlock_t	  lock; /* for from, to */
419 420
	struct mem_cgroup *from;
	struct mem_cgroup *to;
421
	unsigned long immigrate_flags;
422
	unsigned long precharge;
423
	unsigned long moved_charge;
424
	unsigned long moved_swap;
425 426 427
	struct task_struct *moving_task;	/* a task moving charges */
	wait_queue_head_t waitq;		/* a waitq for other context */
} mc = {
428
	.lock = __SPIN_LOCK_UNLOCKED(mc.lock),
429 430
	.waitq = __WAIT_QUEUE_HEAD_INITIALIZER(mc.waitq),
};
431

D
Daisuke Nishimura 已提交
432 433
static bool move_anon(void)
{
434
	return test_bit(MOVE_CHARGE_TYPE_ANON, &mc.immigrate_flags);
D
Daisuke Nishimura 已提交
435 436
}

437 438
static bool move_file(void)
{
439
	return test_bit(MOVE_CHARGE_TYPE_FILE, &mc.immigrate_flags);
440 441
}

442 443 444 445
/*
 * Maximum loops in mem_cgroup_hierarchical_reclaim(), used for soft
 * limit reclaim to prevent infinite loops, if they ever occur.
 */
446 447
#define	MEM_CGROUP_MAX_RECLAIM_LOOPS		100
#define	MEM_CGROUP_MAX_SOFT_LIMIT_RECLAIM_LOOPS	2
448

449 450
enum charge_type {
	MEM_CGROUP_CHARGE_TYPE_CACHE = 0,
451
	MEM_CGROUP_CHARGE_TYPE_ANON,
K
KAMEZAWA Hiroyuki 已提交
452
	MEM_CGROUP_CHARGE_TYPE_SWAPOUT,	/* for accounting swapcache */
K
KAMEZAWA Hiroyuki 已提交
453
	MEM_CGROUP_CHARGE_TYPE_DROP,	/* a page was unused swap cache */
454 455 456
	NR_CHARGE_TYPE,
};

457
/* for encoding cft->private value on file */
G
Glauber Costa 已提交
458 459 460 461
enum res_type {
	_MEM,
	_MEMSWAP,
	_OOM_TYPE,
462
	_KMEM,
G
Glauber Costa 已提交
463 464
};

465 466
#define MEMFILE_PRIVATE(x, val)	((x) << 16 | (val))
#define MEMFILE_TYPE(val)	((val) >> 16 & 0xffff)
467
#define MEMFILE_ATTR(val)	((val) & 0xffff)
K
KAMEZAWA Hiroyuki 已提交
468 469
/* Used for OOM nofiier */
#define OOM_CONTROL		(0)
470

471 472 473 474 475 476 477 478
/*
 * Reclaim flags for mem_cgroup_hierarchical_reclaim
 */
#define MEM_CGROUP_RECLAIM_NOSWAP_BIT	0x0
#define MEM_CGROUP_RECLAIM_NOSWAP	(1 << MEM_CGROUP_RECLAIM_NOSWAP_BIT)
#define MEM_CGROUP_RECLAIM_SHRINK_BIT	0x1
#define MEM_CGROUP_RECLAIM_SHRINK	(1 << MEM_CGROUP_RECLAIM_SHRINK_BIT)

479 480 481 482 483 484 485
/*
 * 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);

486 487 488 489 490 491
static inline
struct mem_cgroup *mem_cgroup_from_css(struct cgroup_subsys_state *s)
{
	return container_of(s, struct mem_cgroup, css);
}

492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509
/* 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;
}

struct vmpressure *css_to_vmpressure(struct cgroup_subsys_state *css)
{
	return &mem_cgroup_from_css(css)->vmpressure;
}

510 511 512 513 514
static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
{
	return (memcg == root_mem_cgroup);
}

G
Glauber Costa 已提交
515
/* Writing them here to avoid exposing memcg's inner layout */
M
Michal Hocko 已提交
516
#if defined(CONFIG_INET) && defined(CONFIG_MEMCG_KMEM)
G
Glauber Costa 已提交
517 518 519

void sock_update_memcg(struct sock *sk)
{
520
	if (mem_cgroup_sockets_enabled) {
G
Glauber Costa 已提交
521
		struct mem_cgroup *memcg;
522
		struct cg_proto *cg_proto;
G
Glauber Costa 已提交
523 524 525

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

526 527 528 529 530 531 532 533 534 535
		/* 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));
536
			css_get(&sk->sk_cgrp->memcg->css);
537 538 539
			return;
		}

G
Glauber Costa 已提交
540 541
		rcu_read_lock();
		memcg = mem_cgroup_from_task(current);
542
		cg_proto = sk->sk_prot->proto_cgroup(memcg);
543 544
		if (!mem_cgroup_is_root(memcg) &&
		    memcg_proto_active(cg_proto) && css_tryget(&memcg->css)) {
545
			sk->sk_cgrp = cg_proto;
G
Glauber Costa 已提交
546 547 548 549 550 551 552 553
		}
		rcu_read_unlock();
	}
}
EXPORT_SYMBOL(sock_update_memcg);

void sock_release_memcg(struct sock *sk)
{
554
	if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
G
Glauber Costa 已提交
555 556 557
		struct mem_cgroup *memcg;
		WARN_ON(!sk->sk_cgrp->memcg);
		memcg = sk->sk_cgrp->memcg;
558
		css_put(&sk->sk_cgrp->memcg->css);
G
Glauber Costa 已提交
559 560
	}
}
G
Glauber Costa 已提交
561 562 563 564 565 566 567 568 569

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

	return &memcg->tcp_mem.cg_proto;
}
EXPORT_SYMBOL(tcp_proto_cgroup);
G
Glauber Costa 已提交
570

571 572 573 574 575 576 577 578 579 580 581 582
static void disarm_sock_keys(struct mem_cgroup *memcg)
{
	if (!memcg_proto_activated(&memcg->tcp_mem.cg_proto))
		return;
	static_key_slow_dec(&memcg_socket_limit_enabled);
}
#else
static void disarm_sock_keys(struct mem_cgroup *memcg)
{
}
#endif

583
#ifdef CONFIG_MEMCG_KMEM
584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601
/*
 * This will be the memcg's index in each cache's ->memcg_params->memcg_caches.
 * There are two main reasons for not using the css_id for this:
 *  1) 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.
 *
 *  2) In order not to violate the cgroup API, we would like to do all memory
 *     allocation in ->create(). At that point, we haven't yet allocated the
 *     css_id. Having a separate index prevents us from messing with the cgroup
 *     core for this
 *
 * 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);
602 603
int memcg_limited_groups_array_size;

604 605 606 607 608 609 610 611 612 613 614 615 616 617 618
/*
 * 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.
 *
 * MAX_SIZE should be as large as the number of css_ids. Ideally, we could get
 * 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
 * css_id space is not getting any smaller, and we don't have to necessarily
 * increase ours as well if it increases.
 */
#define MEMCG_CACHES_MIN_SIZE 4
#define MEMCG_CACHES_MAX_SIZE 65535

619 620 621 622 623 624
/*
 * 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
 */
625
struct static_key memcg_kmem_enabled_key;
626
EXPORT_SYMBOL(memcg_kmem_enabled_key);
627 628 629

static void disarm_kmem_keys(struct mem_cgroup *memcg)
{
630
	if (memcg_kmem_is_active(memcg)) {
631
		static_key_slow_dec(&memcg_kmem_enabled_key);
632 633
		ida_simple_remove(&kmem_limited_groups, memcg->kmemcg_id);
	}
634 635 636 637 638
	/*
	 * This check can't live in kmem destruction function,
	 * since the charges will outlive the cgroup
	 */
	WARN_ON(res_counter_read_u64(&memcg->kmem, RES_USAGE) != 0);
639 640 641 642 643 644 645 646 647 648 649 650 651
}
#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);
}

652
static void drain_all_stock_async(struct mem_cgroup *memcg);
653

654
static struct mem_cgroup_per_zone *
655
mem_cgroup_zoneinfo(struct mem_cgroup *memcg, int nid, int zid)
656
{
657
	VM_BUG_ON((unsigned)nid >= nr_node_ids);
658
	return &memcg->nodeinfo[nid]->zoneinfo[zid];
659 660
}

661
struct cgroup_subsys_state *mem_cgroup_css(struct mem_cgroup *memcg)
662
{
663
	return &memcg->css;
664 665
}

666
static struct mem_cgroup_per_zone *
667
page_cgroup_zoneinfo(struct mem_cgroup *memcg, struct page *page)
668
{
669 670
	int nid = page_to_nid(page);
	int zid = page_zonenum(page);
671

672
	return mem_cgroup_zoneinfo(memcg, nid, zid);
673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690
}

static struct mem_cgroup_tree_per_zone *
soft_limit_tree_node_zone(int nid, int zid)
{
	return &soft_limit_tree.rb_tree_per_node[nid]->rb_tree_per_zone[zid];
}

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

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

static void
691
__mem_cgroup_insert_exceeded(struct mem_cgroup *memcg,
692
				struct mem_cgroup_per_zone *mz,
693 694
				struct mem_cgroup_tree_per_zone *mctz,
				unsigned long long new_usage_in_excess)
695 696 697 698 699 700 701 702
{
	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;

703 704 705
	mz->usage_in_excess = new_usage_in_excess;
	if (!mz->usage_in_excess)
		return;
706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721
	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;
722 723 724
}

static void
725
__mem_cgroup_remove_exceeded(struct mem_cgroup *memcg,
726 727 728 729 730 731 732 733 734
				struct mem_cgroup_per_zone *mz,
				struct mem_cgroup_tree_per_zone *mctz)
{
	if (!mz->on_tree)
		return;
	rb_erase(&mz->tree_node, &mctz->rb_root);
	mz->on_tree = false;
}

735
static void
736
mem_cgroup_remove_exceeded(struct mem_cgroup *memcg,
737 738 739 740
				struct mem_cgroup_per_zone *mz,
				struct mem_cgroup_tree_per_zone *mctz)
{
	spin_lock(&mctz->lock);
741
	__mem_cgroup_remove_exceeded(memcg, mz, mctz);
742 743 744 745
	spin_unlock(&mctz->lock);
}


746
static void mem_cgroup_update_tree(struct mem_cgroup *memcg, struct page *page)
747
{
748
	unsigned long long excess;
749 750
	struct mem_cgroup_per_zone *mz;
	struct mem_cgroup_tree_per_zone *mctz;
751 752
	int nid = page_to_nid(page);
	int zid = page_zonenum(page);
753 754 755
	mctz = soft_limit_tree_from_page(page);

	/*
756 757
	 * Necessary to update all ancestors when hierarchy is used.
	 * because their event counter is not touched.
758
	 */
759 760 761
	for (; memcg; memcg = parent_mem_cgroup(memcg)) {
		mz = mem_cgroup_zoneinfo(memcg, nid, zid);
		excess = res_counter_soft_limit_excess(&memcg->res);
762 763 764 765
		/*
		 * We have to update the tree if mz is on RB-tree or
		 * mem is over its softlimit.
		 */
766
		if (excess || mz->on_tree) {
767 768 769
			spin_lock(&mctz->lock);
			/* if on-tree, remove it */
			if (mz->on_tree)
770
				__mem_cgroup_remove_exceeded(memcg, mz, mctz);
771
			/*
772 773
			 * Insert again. mz->usage_in_excess will be updated.
			 * If excess is 0, no tree ops.
774
			 */
775
			__mem_cgroup_insert_exceeded(memcg, mz, mctz, excess);
776 777
			spin_unlock(&mctz->lock);
		}
778 779 780
	}
}

781
static void mem_cgroup_remove_from_trees(struct mem_cgroup *memcg)
782 783 784 785 786
{
	int node, zone;
	struct mem_cgroup_per_zone *mz;
	struct mem_cgroup_tree_per_zone *mctz;

B
Bob Liu 已提交
787
	for_each_node(node) {
788
		for (zone = 0; zone < MAX_NR_ZONES; zone++) {
789
			mz = mem_cgroup_zoneinfo(memcg, node, zone);
790
			mctz = soft_limit_tree_node_zone(node, zone);
791
			mem_cgroup_remove_exceeded(memcg, mz, mctz);
792 793 794 795
		}
	}
}

796 797 798 799
static struct mem_cgroup_per_zone *
__mem_cgroup_largest_soft_limit_node(struct mem_cgroup_tree_per_zone *mctz)
{
	struct rb_node *rightmost = NULL;
800
	struct mem_cgroup_per_zone *mz;
801 802

retry:
803
	mz = NULL;
804 805 806 807 808 809 810 811 812 813
	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.
	 */
814 815 816
	__mem_cgroup_remove_exceeded(mz->memcg, mz, mctz);
	if (!res_counter_soft_limit_excess(&mz->memcg->res) ||
		!css_tryget(&mz->memcg->css))
817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832
		goto retry;
done:
	return mz;
}

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

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

833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851
/*
 * 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.
 */
852
static long mem_cgroup_read_stat(struct mem_cgroup *memcg,
853
				 enum mem_cgroup_stat_index idx)
854
{
855
	long val = 0;
856 857
	int cpu;

858 859
	get_online_cpus();
	for_each_online_cpu(cpu)
860
		val += per_cpu(memcg->stat->count[idx], cpu);
861
#ifdef CONFIG_HOTPLUG_CPU
862 863 864
	spin_lock(&memcg->pcp_counter_lock);
	val += memcg->nocpu_base.count[idx];
	spin_unlock(&memcg->pcp_counter_lock);
865 866
#endif
	put_online_cpus();
867 868 869
	return val;
}

870
static void mem_cgroup_swap_statistics(struct mem_cgroup *memcg,
871 872 873
					 bool charge)
{
	int val = (charge) ? 1 : -1;
874
	this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_SWAP], val);
875 876
}

877
static unsigned long mem_cgroup_read_events(struct mem_cgroup *memcg,
878 879 880 881 882 883
					    enum mem_cgroup_events_index idx)
{
	unsigned long val = 0;
	int cpu;

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

893
static void mem_cgroup_charge_statistics(struct mem_cgroup *memcg,
894
					 struct page *page,
895
					 bool anon, int nr_pages)
896
{
897 898
	preempt_disable();

899 900 901 902 903 904
	/*
	 * Here, RSS means 'mapped anon' and anon's SwapCache. Shmem/tmpfs is
	 * counted as CACHE even if it's on ANON LRU.
	 */
	if (anon)
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_RSS],
905
				nr_pages);
906
	else
907
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_CACHE],
908
				nr_pages);
909

910 911 912 913
	if (PageTransHuge(page))
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_RSS_HUGE],
				nr_pages);

914 915
	/* pagein of a big page is an event. So, ignore page size */
	if (nr_pages > 0)
916
		__this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGPGIN]);
917
	else {
918
		__this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGPGOUT]);
919 920
		nr_pages = -nr_pages; /* for event */
	}
921

922
	__this_cpu_add(memcg->stat->nr_page_events, nr_pages);
923

924
	preempt_enable();
925 926
}

927
unsigned long
928
mem_cgroup_get_lru_size(struct lruvec *lruvec, enum lru_list lru)
929 930 931 932 933 934 935 936
{
	struct mem_cgroup_per_zone *mz;

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

static unsigned long
937
mem_cgroup_zone_nr_lru_pages(struct mem_cgroup *memcg, int nid, int zid,
938
			unsigned int lru_mask)
939 940
{
	struct mem_cgroup_per_zone *mz;
H
Hugh Dickins 已提交
941
	enum lru_list lru;
942 943
	unsigned long ret = 0;

944
	mz = mem_cgroup_zoneinfo(memcg, nid, zid);
945

H
Hugh Dickins 已提交
946 947 948
	for_each_lru(lru) {
		if (BIT(lru) & lru_mask)
			ret += mz->lru_size[lru];
949 950 951 952 953
	}
	return ret;
}

static unsigned long
954
mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
955 956
			int nid, unsigned int lru_mask)
{
957 958 959
	u64 total = 0;
	int zid;

960
	for (zid = 0; zid < MAX_NR_ZONES; zid++)
961 962
		total += mem_cgroup_zone_nr_lru_pages(memcg,
						nid, zid, lru_mask);
963

964 965
	return total;
}
966

967
static unsigned long mem_cgroup_nr_lru_pages(struct mem_cgroup *memcg,
968
			unsigned int lru_mask)
969
{
970
	int nid;
971 972
	u64 total = 0;

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

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

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

/*
 * Check events in order.
 *
 */
1010
static void memcg_check_events(struct mem_cgroup *memcg, struct page *page)
1011
{
1012
	preempt_disable();
1013
	/* threshold event is triggered in finer grain than soft limit */
1014 1015
	if (unlikely(mem_cgroup_event_ratelimit(memcg,
						MEM_CGROUP_TARGET_THRESH))) {
1016 1017
		bool do_softlimit;
		bool do_numainfo __maybe_unused;
1018 1019 1020 1021 1022 1023 1024 1025 1026

		do_softlimit = mem_cgroup_event_ratelimit(memcg,
						MEM_CGROUP_TARGET_SOFTLIMIT);
#if MAX_NUMNODES > 1
		do_numainfo = mem_cgroup_event_ratelimit(memcg,
						MEM_CGROUP_TARGET_NUMAINFO);
#endif
		preempt_enable();

1027
		mem_cgroup_threshold(memcg);
1028
		if (unlikely(do_softlimit))
1029
			mem_cgroup_update_tree(memcg, page);
1030
#if MAX_NUMNODES > 1
1031
		if (unlikely(do_numainfo))
1032
			atomic_inc(&memcg->numainfo_events);
1033
#endif
1034 1035
	} else
		preempt_enable();
1036 1037
}

G
Glauber Costa 已提交
1038
struct mem_cgroup *mem_cgroup_from_cont(struct cgroup *cont)
B
Balbir Singh 已提交
1039
{
1040 1041
	return mem_cgroup_from_css(
		cgroup_subsys_state(cont, mem_cgroup_subsys_id));
B
Balbir Singh 已提交
1042 1043
}

1044
struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p)
1045
{
1046 1047 1048 1049 1050 1051 1052 1053
	/*
	 * 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;

1054
	return mem_cgroup_from_css(task_subsys_state(p, mem_cgroup_subsys_id));
1055 1056
}

1057
struct mem_cgroup *try_get_mem_cgroup_from_mm(struct mm_struct *mm)
1058
{
1059
	struct mem_cgroup *memcg = NULL;
1060 1061 1062

	if (!mm)
		return NULL;
1063 1064 1065 1066 1067 1068 1069
	/*
	 * Because we have no locks, mm->owner's may be being moved to other
	 * cgroup. We use css_tryget() here even if this looks
	 * pessimistic (rather than adding locks here).
	 */
	rcu_read_lock();
	do {
1070 1071
		memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
		if (unlikely(!memcg))
1072
			break;
1073
	} while (!css_tryget(&memcg->css));
1074
	rcu_read_unlock();
1075
	return memcg;
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 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122
/*
 * Returns a next (in a pre-order walk) alive memcg (with elevated css
 * ref. count) or NULL if the whole root's subtree has been visited.
 *
 * helper function to be used by mem_cgroup_iter
 */
static struct mem_cgroup *__mem_cgroup_iter_next(struct mem_cgroup *root,
		struct mem_cgroup *last_visited)
{
	struct cgroup *prev_cgroup, *next_cgroup;

	/*
	 * Root is not visited by cgroup iterators so it needs an
	 * explicit visit.
	 */
	if (!last_visited)
		return root;

	prev_cgroup = (last_visited == root) ? NULL
		: last_visited->css.cgroup;
skip_node:
	next_cgroup = cgroup_next_descendant_pre(
			prev_cgroup, root->css.cgroup);

	/*
	 * Even if we found a group we have to make sure it is
	 * alive. css && !memcg means that the groups should be
	 * skipped and we should continue the tree walk.
	 * last_visited css is safe to use because it is
	 * protected by css_get and the tree walk is rcu safe.
	 */
	if (next_cgroup) {
		struct mem_cgroup *mem = mem_cgroup_from_cont(
				next_cgroup);
		if (css_tryget(&mem->css))
			return mem;
		else {
			prev_cgroup = next_cgroup;
			goto skip_node;
		}
	}

	return NULL;
}

1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174
static void mem_cgroup_iter_invalidate(struct mem_cgroup *root)
{
	/*
	 * When a group in the hierarchy below root is destroyed, the
	 * hierarchy iterator can no longer be trusted since it might
	 * have pointed to the destroyed group.  Invalidate it.
	 */
	atomic_inc(&root->dead_count);
}

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

static void mem_cgroup_iter_update(struct mem_cgroup_reclaim_iter *iter,
				   struct mem_cgroup *last_visited,
				   struct mem_cgroup *new_position,
				   int sequence)
{
	if (last_visited)
		css_put(&last_visited->css);
	/*
	 * We store the sequence count from the time @last_visited was
	 * loaded successfully instead of rereading it here so that we
	 * don't lose destruction events in between.  We could have
	 * raced with the destruction of @new_position after all.
	 */
	iter->last_visited = new_position;
	smp_wmb();
	iter->last_dead_count = sequence;
}

1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194
/**
 * 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.
 */
struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *root,
				   struct mem_cgroup *prev,
				   struct mem_cgroup_reclaim_cookie *reclaim)
K
KAMEZAWA Hiroyuki 已提交
1195
{
1196
	struct mem_cgroup *memcg = NULL;
1197
	struct mem_cgroup *last_visited = NULL;
1198

1199 1200 1201
	if (mem_cgroup_disabled())
		return NULL;

1202 1203
	if (!root)
		root = root_mem_cgroup;
K
KAMEZAWA Hiroyuki 已提交
1204

1205
	if (prev && !reclaim)
1206
		last_visited = prev;
K
KAMEZAWA Hiroyuki 已提交
1207

1208 1209
	if (!root->use_hierarchy && root != root_mem_cgroup) {
		if (prev)
1210
			goto out_css_put;
1211 1212
		return root;
	}
K
KAMEZAWA Hiroyuki 已提交
1213

1214
	rcu_read_lock();
1215
	while (!memcg) {
1216
		struct mem_cgroup_reclaim_iter *uninitialized_var(iter);
1217
		int uninitialized_var(seq);
1218

1219 1220 1221 1222 1223 1224 1225
		if (reclaim) {
			int nid = zone_to_nid(reclaim->zone);
			int zid = zone_idx(reclaim->zone);
			struct mem_cgroup_per_zone *mz;

			mz = mem_cgroup_zoneinfo(root, nid, zid);
			iter = &mz->reclaim_iter[reclaim->priority];
1226
			if (prev && reclaim->generation != iter->generation) {
M
Michal Hocko 已提交
1227
				iter->last_visited = NULL;
1228 1229
				goto out_unlock;
			}
M
Michal Hocko 已提交
1230

1231
			last_visited = mem_cgroup_iter_load(iter, root, &seq);
1232
		}
K
KAMEZAWA Hiroyuki 已提交
1233

1234
		memcg = __mem_cgroup_iter_next(root, last_visited);
K
KAMEZAWA Hiroyuki 已提交
1235

1236
		if (reclaim) {
1237
			mem_cgroup_iter_update(iter, last_visited, memcg, seq);
1238

M
Michal Hocko 已提交
1239
			if (!memcg)
1240 1241 1242 1243
				iter->generation++;
			else if (!prev && memcg)
				reclaim->generation = iter->generation;
		}
1244

M
Michal Hocko 已提交
1245
		if (prev && !memcg)
1246
			goto out_unlock;
1247
	}
1248 1249
out_unlock:
	rcu_read_unlock();
1250 1251 1252 1253
out_css_put:
	if (prev && prev != root)
		css_put(&prev->css);

1254
	return memcg;
K
KAMEZAWA Hiroyuki 已提交
1255
}
K
KAMEZAWA Hiroyuki 已提交
1256

1257 1258 1259 1260 1261 1262 1263
/**
 * 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)
1264 1265 1266 1267 1268 1269
{
	if (!root)
		root = root_mem_cgroup;
	if (prev && prev != root)
		css_put(&prev->css);
}
K
KAMEZAWA Hiroyuki 已提交
1270

1271 1272 1273 1274 1275 1276
/*
 * 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)		\
1277
	for (iter = mem_cgroup_iter(root, NULL, NULL);	\
1278
	     iter != NULL;				\
1279
	     iter = mem_cgroup_iter(root, iter, NULL))
1280

1281
#define for_each_mem_cgroup(iter)			\
1282
	for (iter = mem_cgroup_iter(NULL, NULL, NULL);	\
1283
	     iter != NULL;				\
1284
	     iter = mem_cgroup_iter(NULL, iter, NULL))
K
KAMEZAWA Hiroyuki 已提交
1285

1286
void __mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
1287
{
1288
	struct mem_cgroup *memcg;
1289 1290

	rcu_read_lock();
1291 1292
	memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
	if (unlikely(!memcg))
1293 1294 1295 1296
		goto out;

	switch (idx) {
	case PGFAULT:
1297 1298 1299 1300
		this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]);
		break;
	case PGMAJFAULT:
		this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]);
1301 1302 1303 1304 1305 1306 1307
		break;
	default:
		BUG();
	}
out:
	rcu_read_unlock();
}
1308
EXPORT_SYMBOL(__mem_cgroup_count_vm_event);
1309

1310 1311 1312
/**
 * mem_cgroup_zone_lruvec - get the lru list vector for a zone and memcg
 * @zone: zone of the wanted lruvec
1313
 * @memcg: memcg of the wanted lruvec
1314 1315 1316 1317 1318 1319 1320 1321 1322
 *
 * 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;
1323
	struct lruvec *lruvec;
1324

1325 1326 1327 1328
	if (mem_cgroup_disabled()) {
		lruvec = &zone->lruvec;
		goto out;
	}
1329 1330

	mz = mem_cgroup_zoneinfo(memcg, zone_to_nid(zone), zone_idx(zone));
1331 1332 1333 1334 1335 1336 1337 1338 1339 1340
	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;
1341 1342
}

K
KAMEZAWA Hiroyuki 已提交
1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355
/*
 * Following LRU functions are allowed to be used without PCG_LOCK.
 * Operations are called by routine of global LRU independently from memcg.
 * What we have to take care of here is validness of pc->mem_cgroup.
 *
 * Changes to pc->mem_cgroup happens when
 * 1. charge
 * 2. moving account
 * In typical case, "charge" is done before add-to-lru. Exception is SwapCache.
 * It is added to LRU before charge.
 * If PCG_USED bit is not set, page_cgroup is not added to this private LRU.
 * When moving account, the page is not on LRU. It's isolated.
 */
1356

1357
/**
1358
 * mem_cgroup_page_lruvec - return lruvec for adding an lru page
1359
 * @page: the page
1360
 * @zone: zone of the page
1361
 */
1362
struct lruvec *mem_cgroup_page_lruvec(struct page *page, struct zone *zone)
K
KAMEZAWA Hiroyuki 已提交
1363 1364
{
	struct mem_cgroup_per_zone *mz;
1365 1366
	struct mem_cgroup *memcg;
	struct page_cgroup *pc;
1367
	struct lruvec *lruvec;
1368

1369 1370 1371 1372
	if (mem_cgroup_disabled()) {
		lruvec = &zone->lruvec;
		goto out;
	}
1373

K
KAMEZAWA Hiroyuki 已提交
1374
	pc = lookup_page_cgroup(page);
1375
	memcg = pc->mem_cgroup;
1376 1377

	/*
1378
	 * Surreptitiously switch any uncharged offlist page to root:
1379 1380 1381 1382 1383 1384 1385
	 * an uncharged page off lru does nothing to secure
	 * its former mem_cgroup from sudden removal.
	 *
	 * Our caller holds lru_lock, and PageCgroupUsed is updated
	 * under page_cgroup lock: between them, they make all uses
	 * of pc->mem_cgroup safe.
	 */
1386
	if (!PageLRU(page) && !PageCgroupUsed(pc) && memcg != root_mem_cgroup)
1387 1388
		pc->mem_cgroup = memcg = root_mem_cgroup;

1389
	mz = page_cgroup_zoneinfo(memcg, page);
1390 1391 1392 1393 1394 1395 1396 1397 1398 1399
	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
1400
}
1401

1402
/**
1403 1404 1405 1406
 * 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
1407
 *
1408 1409
 * This function must be called when a page is added to or removed from an
 * lru list.
1410
 */
1411 1412
void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
				int nr_pages)
1413 1414
{
	struct mem_cgroup_per_zone *mz;
1415
	unsigned long *lru_size;
1416 1417 1418 1419

	if (mem_cgroup_disabled())
		return;

1420 1421 1422 1423
	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
1424
}
1425

1426
/*
1427
 * Checks whether given mem is same or in the root_mem_cgroup's
1428 1429
 * hierarchy subtree
 */
1430 1431
bool __mem_cgroup_same_or_subtree(const struct mem_cgroup *root_memcg,
				  struct mem_cgroup *memcg)
1432
{
1433 1434
	if (root_memcg == memcg)
		return true;
1435
	if (!root_memcg->use_hierarchy || !memcg)
1436
		return false;
1437 1438 1439 1440 1441 1442 1443 1444
	return css_is_ancestor(&memcg->css, &root_memcg->css);
}

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

1445
	rcu_read_lock();
1446
	ret = __mem_cgroup_same_or_subtree(root_memcg, memcg);
1447 1448
	rcu_read_unlock();
	return ret;
1449 1450
}

1451 1452
bool task_in_mem_cgroup(struct task_struct *task,
			const struct mem_cgroup *memcg)
1453
{
1454
	struct mem_cgroup *curr = NULL;
1455
	struct task_struct *p;
1456
	bool ret;
1457

1458
	p = find_lock_task_mm(task);
1459 1460 1461 1462 1463 1464 1465 1466 1467
	if (p) {
		curr = try_get_mem_cgroup_from_mm(p->mm);
		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.
		 */
1468
		rcu_read_lock();
1469 1470 1471
		curr = mem_cgroup_from_task(task);
		if (curr)
			css_get(&curr->css);
1472
		rcu_read_unlock();
1473
	}
1474
	if (!curr)
1475
		return false;
1476
	/*
1477
	 * We should check use_hierarchy of "memcg" not "curr". Because checking
1478
	 * use_hierarchy of "curr" here make this function true if hierarchy is
1479 1480
	 * enabled in "curr" and "curr" is a child of "memcg" in *cgroup*
	 * hierarchy(even if use_hierarchy is disabled in "memcg").
1481
	 */
1482
	ret = mem_cgroup_same_or_subtree(memcg, curr);
1483
	css_put(&curr->css);
1484 1485 1486
	return ret;
}

1487
int mem_cgroup_inactive_anon_is_low(struct lruvec *lruvec)
1488
{
1489
	unsigned long inactive_ratio;
1490
	unsigned long inactive;
1491
	unsigned long active;
1492
	unsigned long gb;
1493

1494 1495
	inactive = mem_cgroup_get_lru_size(lruvec, LRU_INACTIVE_ANON);
	active = mem_cgroup_get_lru_size(lruvec, LRU_ACTIVE_ANON);
1496

1497 1498 1499 1500 1501 1502
	gb = (inactive + active) >> (30 - PAGE_SHIFT);
	if (gb)
		inactive_ratio = int_sqrt(10 * gb);
	else
		inactive_ratio = 1;

1503
	return inactive * inactive_ratio < active;
1504 1505
}

1506 1507 1508
#define mem_cgroup_from_res_counter(counter, member)	\
	container_of(counter, struct mem_cgroup, member)

1509
/**
1510
 * mem_cgroup_margin - calculate chargeable space of a memory cgroup
W
Wanpeng Li 已提交
1511
 * @memcg: the memory cgroup
1512
 *
1513
 * Returns the maximum amount of memory @mem can be charged with, in
1514
 * pages.
1515
 */
1516
static unsigned long mem_cgroup_margin(struct mem_cgroup *memcg)
1517
{
1518 1519
	unsigned long long margin;

1520
	margin = res_counter_margin(&memcg->res);
1521
	if (do_swap_account)
1522
		margin = min(margin, res_counter_margin(&memcg->memsw));
1523
	return margin >> PAGE_SHIFT;
1524 1525
}

1526
int mem_cgroup_swappiness(struct mem_cgroup *memcg)
K
KOSAKI Motohiro 已提交
1527 1528 1529 1530 1531 1532 1533
{
	struct cgroup *cgrp = memcg->css.cgroup;

	/* root ? */
	if (cgrp->parent == NULL)
		return vm_swappiness;

1534
	return memcg->swappiness;
K
KOSAKI Motohiro 已提交
1535 1536
}

1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550
/*
 * memcg->moving_account is used for checking possibility that some thread is
 * calling move_account(). When a thread on CPU-A starts moving pages under
 * a memcg, other threads should check memcg->moving_account under
 * rcu_read_lock(), like this:
 *
 *         CPU-A                                    CPU-B
 *                                              rcu_read_lock()
 *         memcg->moving_account+1              if (memcg->mocing_account)
 *                                                   take heavy locks.
 *         synchronize_rcu()                    update something.
 *                                              rcu_read_unlock()
 *         start move here.
 */
1551 1552 1553 1554

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

1555
static void mem_cgroup_start_move(struct mem_cgroup *memcg)
1556
{
1557
	atomic_inc(&memcg_moving);
1558
	atomic_inc(&memcg->moving_account);
1559 1560 1561
	synchronize_rcu();
}

1562
static void mem_cgroup_end_move(struct mem_cgroup *memcg)
1563
{
1564 1565 1566 1567
	/*
	 * Now, mem_cgroup_clear_mc() may call this function with NULL.
	 * We check NULL in callee rather than caller.
	 */
1568 1569
	if (memcg) {
		atomic_dec(&memcg_moving);
1570
		atomic_dec(&memcg->moving_account);
1571
	}
1572
}
1573

1574 1575 1576
/*
 * 2 routines for checking "mem" is under move_account() or not.
 *
1577 1578
 * mem_cgroup_stolen() -  checking whether a cgroup is mc.from or not. This
 *			  is used for avoiding races in accounting.  If true,
1579 1580 1581 1582 1583 1584 1585
 *			  pc->mem_cgroup may be overwritten.
 *
 * mem_cgroup_under_move() - checking a cgroup is mc.from or mc.to or
 *			  under hierarchy of moving cgroups. This is for
 *			  waiting at hith-memory prressure caused by "move".
 */

1586
static bool mem_cgroup_stolen(struct mem_cgroup *memcg)
1587 1588
{
	VM_BUG_ON(!rcu_read_lock_held());
1589
	return atomic_read(&memcg->moving_account) > 0;
1590
}
1591

1592
static bool mem_cgroup_under_move(struct mem_cgroup *memcg)
1593
{
1594 1595
	struct mem_cgroup *from;
	struct mem_cgroup *to;
1596
	bool ret = false;
1597 1598 1599 1600 1601 1602 1603 1604 1605
	/*
	 * 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;
1606

1607 1608
	ret = mem_cgroup_same_or_subtree(memcg, from)
		|| mem_cgroup_same_or_subtree(memcg, to);
1609 1610
unlock:
	spin_unlock(&mc.lock);
1611 1612 1613
	return ret;
}

1614
static bool mem_cgroup_wait_acct_move(struct mem_cgroup *memcg)
1615 1616
{
	if (mc.moving_task && current != mc.moving_task) {
1617
		if (mem_cgroup_under_move(memcg)) {
1618 1619 1620 1621 1622 1623 1624 1625 1626 1627 1628 1629
			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;
}

1630 1631 1632 1633
/*
 * Take this lock when
 * - a code tries to modify page's memcg while it's USED.
 * - a code tries to modify page state accounting in a memcg.
1634
 * see mem_cgroup_stolen(), too.
1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647
 */
static void move_lock_mem_cgroup(struct mem_cgroup *memcg,
				  unsigned long *flags)
{
	spin_lock_irqsave(&memcg->move_lock, *flags);
}

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

1648
#define K(x) ((x) << (PAGE_SHIFT-10))
1649
/**
1650
 * mem_cgroup_print_oom_info: Print OOM information relevant to memory controller.
1651 1652 1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667
 * @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)
{
	struct cgroup *task_cgrp;
	struct cgroup *mem_cgrp;
	/*
	 * Need a buffer in BSS, can't rely on allocations. The code relies
	 * on the assumption that OOM is serialized for memory controller.
	 * If this assumption is broken, revisit this code.
	 */
	static char memcg_name[PATH_MAX];
	int ret;
1668 1669
	struct mem_cgroup *iter;
	unsigned int i;
1670

1671
	if (!p)
1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689
		return;

	rcu_read_lock();

	mem_cgrp = memcg->css.cgroup;
	task_cgrp = task_cgroup(p, mem_cgroup_subsys_id);

	ret = cgroup_path(task_cgrp, memcg_name, PATH_MAX);
	if (ret < 0) {
		/*
		 * Unfortunately, we are unable to convert to a useful name
		 * But we'll still print out the usage information
		 */
		rcu_read_unlock();
		goto done;
	}
	rcu_read_unlock();

1690
	pr_info("Task in %s killed", memcg_name);
1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702

	rcu_read_lock();
	ret = cgroup_path(mem_cgrp, memcg_name, PATH_MAX);
	if (ret < 0) {
		rcu_read_unlock();
		goto done;
	}
	rcu_read_unlock();

	/*
	 * Continues from above, so we don't need an KERN_ level
	 */
1703
	pr_cont(" as a result of limit of %s\n", memcg_name);
1704 1705
done:

1706
	pr_info("memory: usage %llukB, limit %llukB, failcnt %llu\n",
1707 1708 1709
		res_counter_read_u64(&memcg->res, RES_USAGE) >> 10,
		res_counter_read_u64(&memcg->res, RES_LIMIT) >> 10,
		res_counter_read_u64(&memcg->res, RES_FAILCNT));
1710
	pr_info("memory+swap: usage %llukB, limit %llukB, failcnt %llu\n",
1711 1712 1713
		res_counter_read_u64(&memcg->memsw, RES_USAGE) >> 10,
		res_counter_read_u64(&memcg->memsw, RES_LIMIT) >> 10,
		res_counter_read_u64(&memcg->memsw, RES_FAILCNT));
1714
	pr_info("kmem: usage %llukB, limit %llukB, failcnt %llu\n",
1715 1716 1717
		res_counter_read_u64(&memcg->kmem, RES_USAGE) >> 10,
		res_counter_read_u64(&memcg->kmem, RES_LIMIT) >> 10,
		res_counter_read_u64(&memcg->kmem, RES_FAILCNT));
1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741

	for_each_mem_cgroup_tree(iter, memcg) {
		pr_info("Memory cgroup stats");

		rcu_read_lock();
		ret = cgroup_path(iter->css.cgroup, memcg_name, PATH_MAX);
		if (!ret)
			pr_cont(" for %s", memcg_name);
		rcu_read_unlock();
		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");
	}
1742 1743
}

1744 1745 1746 1747
/*
 * This function returns the number of memcg under hierarchy tree. Returns
 * 1(self count) if no children.
 */
1748
static int mem_cgroup_count_children(struct mem_cgroup *memcg)
1749 1750
{
	int num = 0;
K
KAMEZAWA Hiroyuki 已提交
1751 1752
	struct mem_cgroup *iter;

1753
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
1754
		num++;
1755 1756 1757
	return num;
}

D
David Rientjes 已提交
1758 1759 1760
/*
 * Return the memory (and swap, if configured) limit for a memcg.
 */
1761
static u64 mem_cgroup_get_limit(struct mem_cgroup *memcg)
D
David Rientjes 已提交
1762 1763 1764
{
	u64 limit;

1765 1766
	limit = res_counter_read_u64(&memcg->res, RES_LIMIT);

D
David Rientjes 已提交
1767
	/*
1768
	 * Do not consider swap space if we cannot swap due to swappiness
D
David Rientjes 已提交
1769
	 */
1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783
	if (mem_cgroup_swappiness(memcg)) {
		u64 memsw;

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

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

	return limit;
D
David Rientjes 已提交
1784 1785
}

1786 1787
static void mem_cgroup_out_of_memory(struct mem_cgroup *memcg, gfp_t gfp_mask,
				     int order)
1788 1789 1790 1791 1792 1793 1794
{
	struct mem_cgroup *iter;
	unsigned long chosen_points = 0;
	unsigned long totalpages;
	unsigned int points = 0;
	struct task_struct *chosen = NULL;

1795
	/*
1796 1797 1798
	 * 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.
1799
	 */
1800
	if (fatal_signal_pending(current) || current->flags & PF_EXITING) {
1801 1802 1803 1804 1805
		set_thread_flag(TIF_MEMDIE);
		return;
	}

	check_panic_on_oom(CONSTRAINT_MEMCG, gfp_mask, order, NULL);
1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852
	totalpages = mem_cgroup_get_limit(memcg) >> PAGE_SHIFT ? : 1;
	for_each_mem_cgroup_tree(iter, memcg) {
		struct cgroup *cgroup = iter->css.cgroup;
		struct cgroup_iter it;
		struct task_struct *task;

		cgroup_iter_start(cgroup, &it);
		while ((task = cgroup_iter_next(cgroup, &it))) {
			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:
				cgroup_iter_end(cgroup, &it);
				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);
			if (points > chosen_points) {
				if (chosen)
					put_task_struct(chosen);
				chosen = task;
				chosen_points = points;
				get_task_struct(chosen);
			}
		}
		cgroup_iter_end(cgroup, &it);
	}

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

1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888
static unsigned long mem_cgroup_reclaim(struct mem_cgroup *memcg,
					gfp_t gfp_mask,
					unsigned long flags)
{
	unsigned long total = 0;
	bool noswap = false;
	int loop;

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

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

1889 1890
/**
 * test_mem_cgroup_node_reclaimable
W
Wanpeng Li 已提交
1891
 * @memcg: the target memcg
1892 1893 1894 1895 1896 1897 1898
 * @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.
 */
1899
static bool test_mem_cgroup_node_reclaimable(struct mem_cgroup *memcg,
1900 1901
		int nid, bool noswap)
{
1902
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_FILE))
1903 1904 1905
		return true;
	if (noswap || !total_swap_pages)
		return false;
1906
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_ANON))
1907 1908 1909 1910
		return true;
	return false;

}
1911 1912 1913 1914 1915 1916 1917 1918
#if MAX_NUMNODES > 1

/*
 * 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.
 *
 */
1919
static void mem_cgroup_may_update_nodemask(struct mem_cgroup *memcg)
1920 1921
{
	int nid;
1922 1923 1924 1925
	/*
	 * numainfo_events > 0 means there was at least NUMAINFO_EVENTS_TARGET
	 * pagein/pageout changes since the last update.
	 */
1926
	if (!atomic_read(&memcg->numainfo_events))
1927
		return;
1928
	if (atomic_inc_return(&memcg->numainfo_updating) > 1)
1929 1930 1931
		return;

	/* make a nodemask where this memcg uses memory from */
1932
	memcg->scan_nodes = node_states[N_MEMORY];
1933

1934
	for_each_node_mask(nid, node_states[N_MEMORY]) {
1935

1936 1937
		if (!test_mem_cgroup_node_reclaimable(memcg, nid, false))
			node_clear(nid, memcg->scan_nodes);
1938
	}
1939

1940 1941
	atomic_set(&memcg->numainfo_events, 0);
	atomic_set(&memcg->numainfo_updating, 0);
1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955
}

/*
 * 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.
 */
1956
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
1957 1958 1959
{
	int node;

1960 1961
	mem_cgroup_may_update_nodemask(memcg);
	node = memcg->last_scanned_node;
1962

1963
	node = next_node(node, memcg->scan_nodes);
1964
	if (node == MAX_NUMNODES)
1965
		node = first_node(memcg->scan_nodes);
1966 1967 1968 1969 1970 1971 1972 1973 1974
	/*
	 * 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();

1975
	memcg->last_scanned_node = node;
1976 1977 1978
	return node;
}

1979 1980 1981 1982 1983 1984
/*
 * Check all nodes whether it contains reclaimable pages or not.
 * For quick scan, we make use of scan_nodes. This will allow us to skip
 * unused nodes. But scan_nodes is lazily updated and may not cotain
 * enough new information. We need to do double check.
 */
1985
static bool mem_cgroup_reclaimable(struct mem_cgroup *memcg, bool noswap)
1986 1987 1988 1989 1990 1991 1992
{
	int nid;

	/*
	 * quick check...making use of scan_node.
	 * We can skip unused nodes.
	 */
1993 1994
	if (!nodes_empty(memcg->scan_nodes)) {
		for (nid = first_node(memcg->scan_nodes);
1995
		     nid < MAX_NUMNODES;
1996
		     nid = next_node(nid, memcg->scan_nodes)) {
1997

1998
			if (test_mem_cgroup_node_reclaimable(memcg, nid, noswap))
1999 2000 2001 2002 2003 2004
				return true;
		}
	}
	/*
	 * Check rest of nodes.
	 */
2005
	for_each_node_state(nid, N_MEMORY) {
2006
		if (node_isset(nid, memcg->scan_nodes))
2007
			continue;
2008
		if (test_mem_cgroup_node_reclaimable(memcg, nid, noswap))
2009 2010 2011 2012 2013
			return true;
	}
	return false;
}

2014
#else
2015
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
2016 2017 2018
{
	return 0;
}
2019

2020
static bool mem_cgroup_reclaimable(struct mem_cgroup *memcg, bool noswap)
2021
{
2022
	return test_mem_cgroup_node_reclaimable(memcg, 0, noswap);
2023
}
2024 2025
#endif

2026 2027 2028 2029
static int mem_cgroup_soft_reclaim(struct mem_cgroup *root_memcg,
				   struct zone *zone,
				   gfp_t gfp_mask,
				   unsigned long *total_scanned)
2030
{
2031
	struct mem_cgroup *victim = NULL;
2032
	int total = 0;
K
KAMEZAWA Hiroyuki 已提交
2033
	int loop = 0;
2034
	unsigned long excess;
2035
	unsigned long nr_scanned;
2036 2037 2038 2039
	struct mem_cgroup_reclaim_cookie reclaim = {
		.zone = zone,
		.priority = 0,
	};
2040

2041
	excess = res_counter_soft_limit_excess(&root_memcg->res) >> PAGE_SHIFT;
K
KAMEZAWA Hiroyuki 已提交
2042

2043
	while (1) {
2044
		victim = mem_cgroup_iter(root_memcg, victim, &reclaim);
2045
		if (!victim) {
K
KAMEZAWA Hiroyuki 已提交
2046
			loop++;
2047 2048 2049 2050 2051 2052
			if (loop >= 2) {
				/*
				 * If we have not been able to reclaim
				 * anything, it might because there are
				 * no reclaimable pages under this hierarchy
				 */
2053
				if (!total)
2054 2055
					break;
				/*
L
Lucas De Marchi 已提交
2056
				 * We want to do more targeted reclaim.
2057 2058 2059 2060 2061
				 * 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) ||
2062
					(loop > MEM_CGROUP_MAX_RECLAIM_LOOPS))
2063 2064
					break;
			}
2065
			continue;
2066
		}
2067
		if (!mem_cgroup_reclaimable(victim, false))
2068
			continue;
2069 2070 2071 2072
		total += mem_cgroup_shrink_node_zone(victim, gfp_mask, false,
						     zone, &nr_scanned);
		*total_scanned += nr_scanned;
		if (!res_counter_soft_limit_excess(&root_memcg->res))
2073
			break;
2074
	}
2075
	mem_cgroup_iter_break(root_memcg, victim);
K
KAMEZAWA Hiroyuki 已提交
2076
	return total;
2077 2078
}

K
KAMEZAWA Hiroyuki 已提交
2079 2080 2081
/*
 * Check OOM-Killer is already running under our hierarchy.
 * If someone is running, return false.
2082
 * Has to be called with memcg_oom_lock
K
KAMEZAWA Hiroyuki 已提交
2083
 */
2084
static bool mem_cgroup_oom_lock(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
2085
{
2086
	struct mem_cgroup *iter, *failed = NULL;
2087

2088
	for_each_mem_cgroup_tree(iter, memcg) {
2089
		if (iter->oom_lock) {
2090 2091 2092 2093 2094
			/*
			 * this subtree of our hierarchy is already locked
			 * so we cannot give a lock.
			 */
			failed = iter;
2095 2096
			mem_cgroup_iter_break(memcg, iter);
			break;
2097 2098
		} else
			iter->oom_lock = true;
K
KAMEZAWA Hiroyuki 已提交
2099
	}
K
KAMEZAWA Hiroyuki 已提交
2100

2101
	if (!failed)
2102
		return true;
2103 2104 2105 2106 2107

	/*
	 * OK, we failed to lock the whole subtree so we have to clean up
	 * what we set up to the failing subtree
	 */
2108
	for_each_mem_cgroup_tree(iter, memcg) {
2109
		if (iter == failed) {
2110 2111
			mem_cgroup_iter_break(memcg, iter);
			break;
2112 2113 2114
		}
		iter->oom_lock = false;
	}
2115
	return false;
2116
}
2117

2118
/*
2119
 * Has to be called with memcg_oom_lock
2120
 */
2121
static int mem_cgroup_oom_unlock(struct mem_cgroup *memcg)
2122
{
K
KAMEZAWA Hiroyuki 已提交
2123 2124
	struct mem_cgroup *iter;

2125
	for_each_mem_cgroup_tree(iter, memcg)
2126 2127 2128 2129
		iter->oom_lock = false;
	return 0;
}

2130
static void mem_cgroup_mark_under_oom(struct mem_cgroup *memcg)
2131 2132 2133
{
	struct mem_cgroup *iter;

2134
	for_each_mem_cgroup_tree(iter, memcg)
2135 2136 2137
		atomic_inc(&iter->under_oom);
}

2138
static void mem_cgroup_unmark_under_oom(struct mem_cgroup *memcg)
2139 2140 2141
{
	struct mem_cgroup *iter;

K
KAMEZAWA Hiroyuki 已提交
2142 2143 2144 2145 2146
	/*
	 * 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.
	 */
2147
	for_each_mem_cgroup_tree(iter, memcg)
2148
		atomic_add_unless(&iter->under_oom, -1, 0);
2149 2150
}

2151
static DEFINE_SPINLOCK(memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
2152 2153
static DECLARE_WAIT_QUEUE_HEAD(memcg_oom_waitq);

K
KAMEZAWA Hiroyuki 已提交
2154
struct oom_wait_info {
2155
	struct mem_cgroup *memcg;
K
KAMEZAWA Hiroyuki 已提交
2156 2157 2158 2159 2160 2161
	wait_queue_t	wait;
};

static int memcg_oom_wake_function(wait_queue_t *wait,
	unsigned mode, int sync, void *arg)
{
2162 2163
	struct mem_cgroup *wake_memcg = (struct mem_cgroup *)arg;
	struct mem_cgroup *oom_wait_memcg;
K
KAMEZAWA Hiroyuki 已提交
2164 2165 2166
	struct oom_wait_info *oom_wait_info;

	oom_wait_info = container_of(wait, struct oom_wait_info, wait);
2167
	oom_wait_memcg = oom_wait_info->memcg;
K
KAMEZAWA Hiroyuki 已提交
2168 2169

	/*
2170
	 * Both of oom_wait_info->memcg and wake_memcg are stable under us.
K
KAMEZAWA Hiroyuki 已提交
2171 2172
	 * Then we can use css_is_ancestor without taking care of RCU.
	 */
2173 2174
	if (!mem_cgroup_same_or_subtree(oom_wait_memcg, wake_memcg)
		&& !mem_cgroup_same_or_subtree(wake_memcg, oom_wait_memcg))
K
KAMEZAWA Hiroyuki 已提交
2175 2176 2177 2178
		return 0;
	return autoremove_wake_function(wait, mode, sync, arg);
}

2179
static void memcg_wakeup_oom(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
2180
{
2181 2182
	/* for filtering, pass "memcg" as argument. */
	__wake_up(&memcg_oom_waitq, TASK_NORMAL, 0, memcg);
K
KAMEZAWA Hiroyuki 已提交
2183 2184
}

2185
static void memcg_oom_recover(struct mem_cgroup *memcg)
2186
{
2187 2188
	if (memcg && atomic_read(&memcg->under_oom))
		memcg_wakeup_oom(memcg);
2189 2190
}

K
KAMEZAWA Hiroyuki 已提交
2191 2192 2193
/*
 * try to call OOM killer. returns false if we should exit memory-reclaim loop.
 */
2194 2195
static bool mem_cgroup_handle_oom(struct mem_cgroup *memcg, gfp_t mask,
				  int order)
2196
{
K
KAMEZAWA Hiroyuki 已提交
2197
	struct oom_wait_info owait;
2198
	bool locked, need_to_kill;
K
KAMEZAWA Hiroyuki 已提交
2199

2200
	owait.memcg = memcg;
K
KAMEZAWA Hiroyuki 已提交
2201 2202 2203 2204
	owait.wait.flags = 0;
	owait.wait.func = memcg_oom_wake_function;
	owait.wait.private = current;
	INIT_LIST_HEAD(&owait.wait.task_list);
2205
	need_to_kill = true;
2206
	mem_cgroup_mark_under_oom(memcg);
2207

2208
	/* At first, try to OOM lock hierarchy under memcg.*/
2209
	spin_lock(&memcg_oom_lock);
2210
	locked = mem_cgroup_oom_lock(memcg);
K
KAMEZAWA Hiroyuki 已提交
2211 2212 2213 2214 2215
	/*
	 * Even if signal_pending(), we can't quit charge() loop without
	 * accounting. So, UNINTERRUPTIBLE is appropriate. But SIGKILL
	 * under OOM is always welcomed, use TASK_KILLABLE here.
	 */
2216
	prepare_to_wait(&memcg_oom_waitq, &owait.wait, TASK_KILLABLE);
2217
	if (!locked || memcg->oom_kill_disable)
2218 2219
		need_to_kill = false;
	if (locked)
2220
		mem_cgroup_oom_notify(memcg);
2221
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
2222

2223 2224
	if (need_to_kill) {
		finish_wait(&memcg_oom_waitq, &owait.wait);
2225
		mem_cgroup_out_of_memory(memcg, mask, order);
2226
	} else {
K
KAMEZAWA Hiroyuki 已提交
2227
		schedule();
K
KAMEZAWA Hiroyuki 已提交
2228
		finish_wait(&memcg_oom_waitq, &owait.wait);
K
KAMEZAWA Hiroyuki 已提交
2229
	}
2230
	spin_lock(&memcg_oom_lock);
2231
	if (locked)
2232 2233
		mem_cgroup_oom_unlock(memcg);
	memcg_wakeup_oom(memcg);
2234
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
2235

2236
	mem_cgroup_unmark_under_oom(memcg);
2237

K
KAMEZAWA Hiroyuki 已提交
2238 2239 2240
	if (test_thread_flag(TIF_MEMDIE) || fatal_signal_pending(current))
		return false;
	/* Give chance to dying process */
2241
	schedule_timeout_uninterruptible(1);
K
KAMEZAWA Hiroyuki 已提交
2242
	return true;
2243 2244
}

2245 2246 2247
/*
 * Currently used to update mapped file statistics, but the routine can be
 * generalized to update other statistics as well.
2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264
 *
 * Notes: Race condition
 *
 * We usually use page_cgroup_lock() for accessing page_cgroup member but
 * it tends to be costly. But considering some conditions, we doesn't need
 * to do so _always_.
 *
 * Considering "charge", lock_page_cgroup() is not required because all
 * file-stat operations happen after a page is attached to radix-tree. There
 * are no race with "charge".
 *
 * Considering "uncharge", we know that memcg doesn't clear pc->mem_cgroup
 * at "uncharge" intentionally. So, we always see valid pc->mem_cgroup even
 * if there are race with "uncharge". Statistics itself is properly handled
 * by flags.
 *
 * Considering "move", this is an only case we see a race. To make the race
2265 2266
 * small, we check mm->moving_account and detect there are possibility of race
 * If there is, we take a lock.
2267
 */
2268

2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281
void __mem_cgroup_begin_update_page_stat(struct page *page,
				bool *locked, unsigned long *flags)
{
	struct mem_cgroup *memcg;
	struct page_cgroup *pc;

	pc = lookup_page_cgroup(page);
again:
	memcg = pc->mem_cgroup;
	if (unlikely(!memcg || !PageCgroupUsed(pc)))
		return;
	/*
	 * If this memory cgroup is not under account moving, we don't
2282
	 * need to take move_lock_mem_cgroup(). Because we already hold
2283
	 * rcu_read_lock(), any calls to move_account will be delayed until
2284
	 * rcu_read_unlock() if mem_cgroup_stolen() == true.
2285
	 */
2286
	if (!mem_cgroup_stolen(memcg))
2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303
		return;

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

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

	/*
	 * It's guaranteed that pc->mem_cgroup never changes while
	 * lock is held because a routine modifies pc->mem_cgroup
2304
	 * should take move_lock_mem_cgroup().
2305 2306 2307 2308
	 */
	move_unlock_mem_cgroup(pc->mem_cgroup, flags);
}

2309 2310
void mem_cgroup_update_page_stat(struct page *page,
				 enum mem_cgroup_page_stat_item idx, int val)
2311
{
2312
	struct mem_cgroup *memcg;
2313
	struct page_cgroup *pc = lookup_page_cgroup(page);
2314
	unsigned long uninitialized_var(flags);
2315

2316
	if (mem_cgroup_disabled())
2317
		return;
2318

2319 2320
	memcg = pc->mem_cgroup;
	if (unlikely(!memcg || !PageCgroupUsed(pc)))
2321
		return;
2322 2323

	switch (idx) {
2324 2325
	case MEMCG_NR_FILE_MAPPED:
		idx = MEM_CGROUP_STAT_FILE_MAPPED;
2326 2327 2328
		break;
	default:
		BUG();
2329
	}
2330

2331
	this_cpu_add(memcg->stat->count[idx], val);
2332
}
2333

2334 2335 2336 2337
/*
 * size of first charge trial. "32" comes from vmscan.c's magic value.
 * TODO: maybe necessary to use big numbers in big irons.
 */
2338
#define CHARGE_BATCH	32U
2339 2340
struct memcg_stock_pcp {
	struct mem_cgroup *cached; /* this never be root cgroup */
2341
	unsigned int nr_pages;
2342
	struct work_struct work;
2343
	unsigned long flags;
2344
#define FLUSHING_CACHED_CHARGE	0
2345 2346
};
static DEFINE_PER_CPU(struct memcg_stock_pcp, memcg_stock);
2347
static DEFINE_MUTEX(percpu_charge_mutex);
2348

2349 2350 2351 2352 2353 2354 2355 2356 2357 2358
/**
 * 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.
2359
 */
2360
static bool consume_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
2361 2362 2363 2364
{
	struct memcg_stock_pcp *stock;
	bool ret = true;

2365 2366 2367
	if (nr_pages > CHARGE_BATCH)
		return false;

2368
	stock = &get_cpu_var(memcg_stock);
2369 2370
	if (memcg == stock->cached && stock->nr_pages >= nr_pages)
		stock->nr_pages -= nr_pages;
2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383
	else /* need to call res_counter_charge */
		ret = false;
	put_cpu_var(memcg_stock);
	return ret;
}

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

2384 2385 2386 2387
	if (stock->nr_pages) {
		unsigned long bytes = stock->nr_pages * PAGE_SIZE;

		res_counter_uncharge(&old->res, bytes);
2388
		if (do_swap_account)
2389 2390
			res_counter_uncharge(&old->memsw, bytes);
		stock->nr_pages = 0;
2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402
	}
	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)
{
	struct memcg_stock_pcp *stock = &__get_cpu_var(memcg_stock);
	drain_stock(stock);
2403
	clear_bit(FLUSHING_CACHED_CHARGE, &stock->flags);
2404 2405
}

2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416
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);
	}
}

2417 2418
/*
 * Cache charges(val) which is from res_counter, to local per_cpu area.
2419
 * This will be consumed by consume_stock() function, later.
2420
 */
2421
static void refill_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
2422 2423 2424
{
	struct memcg_stock_pcp *stock = &get_cpu_var(memcg_stock);

2425
	if (stock->cached != memcg) { /* reset if necessary */
2426
		drain_stock(stock);
2427
		stock->cached = memcg;
2428
	}
2429
	stock->nr_pages += nr_pages;
2430 2431 2432 2433
	put_cpu_var(memcg_stock);
}

/*
2434
 * Drains all per-CPU charge caches for given root_memcg resp. subtree
2435 2436
 * of the hierarchy under it. sync flag says whether we should block
 * until the work is done.
2437
 */
2438
static void drain_all_stock(struct mem_cgroup *root_memcg, bool sync)
2439
{
2440
	int cpu, curcpu;
2441

2442 2443
	/* Notify other cpus that system-wide "drain" is running */
	get_online_cpus();
2444
	curcpu = get_cpu();
2445 2446
	for_each_online_cpu(cpu) {
		struct memcg_stock_pcp *stock = &per_cpu(memcg_stock, cpu);
2447
		struct mem_cgroup *memcg;
2448

2449 2450
		memcg = stock->cached;
		if (!memcg || !stock->nr_pages)
2451
			continue;
2452
		if (!mem_cgroup_same_or_subtree(root_memcg, memcg))
2453
			continue;
2454 2455 2456 2457 2458 2459
		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);
		}
2460
	}
2461
	put_cpu();
2462 2463 2464 2465 2466 2467

	if (!sync)
		goto out;

	for_each_online_cpu(cpu) {
		struct memcg_stock_pcp *stock = &per_cpu(memcg_stock, cpu);
2468
		if (test_bit(FLUSHING_CACHED_CHARGE, &stock->flags))
2469 2470 2471
			flush_work(&stock->work);
	}
out:
2472
 	put_online_cpus();
2473 2474 2475 2476 2477 2478 2479 2480
}

/*
 * Tries to drain stocked charges in other cpus. This function is asynchronous
 * and just put a work per cpu for draining localy on each cpu. Caller can
 * expects some charges will be back to res_counter later but cannot wait for
 * it.
 */
2481
static void drain_all_stock_async(struct mem_cgroup *root_memcg)
2482
{
2483 2484 2485 2486 2487
	/*
	 * If someone calls draining, avoid adding more kworker runs.
	 */
	if (!mutex_trylock(&percpu_charge_mutex))
		return;
2488
	drain_all_stock(root_memcg, false);
2489
	mutex_unlock(&percpu_charge_mutex);
2490 2491 2492
}

/* This is a synchronous drain interface. */
2493
static void drain_all_stock_sync(struct mem_cgroup *root_memcg)
2494 2495
{
	/* called when force_empty is called */
2496
	mutex_lock(&percpu_charge_mutex);
2497
	drain_all_stock(root_memcg, true);
2498
	mutex_unlock(&percpu_charge_mutex);
2499 2500
}

2501 2502 2503 2504
/*
 * This function drains percpu counter value from DEAD cpu and
 * move it to local cpu. Note that this function can be preempted.
 */
2505
static void mem_cgroup_drain_pcp_counter(struct mem_cgroup *memcg, int cpu)
2506 2507 2508
{
	int i;

2509
	spin_lock(&memcg->pcp_counter_lock);
2510
	for (i = 0; i < MEM_CGROUP_STAT_NSTATS; i++) {
2511
		long x = per_cpu(memcg->stat->count[i], cpu);
2512

2513 2514
		per_cpu(memcg->stat->count[i], cpu) = 0;
		memcg->nocpu_base.count[i] += x;
2515
	}
2516
	for (i = 0; i < MEM_CGROUP_EVENTS_NSTATS; i++) {
2517
		unsigned long x = per_cpu(memcg->stat->events[i], cpu);
2518

2519 2520
		per_cpu(memcg->stat->events[i], cpu) = 0;
		memcg->nocpu_base.events[i] += x;
2521
	}
2522
	spin_unlock(&memcg->pcp_counter_lock);
2523 2524 2525
}

static int __cpuinit memcg_cpu_hotplug_callback(struct notifier_block *nb,
2526 2527 2528 2529 2530
					unsigned long action,
					void *hcpu)
{
	int cpu = (unsigned long)hcpu;
	struct memcg_stock_pcp *stock;
2531
	struct mem_cgroup *iter;
2532

2533
	if (action == CPU_ONLINE)
2534 2535
		return NOTIFY_OK;

2536
	if (action != CPU_DEAD && action != CPU_DEAD_FROZEN)
2537
		return NOTIFY_OK;
2538

2539
	for_each_mem_cgroup(iter)
2540 2541
		mem_cgroup_drain_pcp_counter(iter, cpu);

2542 2543 2544 2545 2546
	stock = &per_cpu(memcg_stock, cpu);
	drain_stock(stock);
	return NOTIFY_OK;
}

2547 2548 2549 2550 2551 2552 2553 2554 2555 2556

/* See __mem_cgroup_try_charge() for details */
enum {
	CHARGE_OK,		/* success */
	CHARGE_RETRY,		/* need to retry but retry is not bad */
	CHARGE_NOMEM,		/* we can't do more. return -ENOMEM */
	CHARGE_WOULDBLOCK,	/* GFP_WAIT wasn't set and no enough res. */
	CHARGE_OOM_DIE,		/* the current is killed because of OOM */
};

2557
static int mem_cgroup_do_charge(struct mem_cgroup *memcg, gfp_t gfp_mask,
2558 2559
				unsigned int nr_pages, unsigned int min_pages,
				bool oom_check)
2560
{
2561
	unsigned long csize = nr_pages * PAGE_SIZE;
2562 2563 2564 2565 2566
	struct mem_cgroup *mem_over_limit;
	struct res_counter *fail_res;
	unsigned long flags = 0;
	int ret;

2567
	ret = res_counter_charge(&memcg->res, csize, &fail_res);
2568 2569 2570 2571

	if (likely(!ret)) {
		if (!do_swap_account)
			return CHARGE_OK;
2572
		ret = res_counter_charge(&memcg->memsw, csize, &fail_res);
2573 2574 2575
		if (likely(!ret))
			return CHARGE_OK;

2576
		res_counter_uncharge(&memcg->res, csize);
2577 2578 2579 2580
		mem_over_limit = mem_cgroup_from_res_counter(fail_res, memsw);
		flags |= MEM_CGROUP_RECLAIM_NOSWAP;
	} else
		mem_over_limit = mem_cgroup_from_res_counter(fail_res, res);
2581 2582 2583 2584
	/*
	 * Never reclaim on behalf of optional batching, retry with a
	 * single page instead.
	 */
2585
	if (nr_pages > min_pages)
2586 2587 2588 2589 2590
		return CHARGE_RETRY;

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

2591 2592 2593
	if (gfp_mask & __GFP_NORETRY)
		return CHARGE_NOMEM;

2594
	ret = mem_cgroup_reclaim(mem_over_limit, gfp_mask, flags);
2595
	if (mem_cgroup_margin(mem_over_limit) >= nr_pages)
2596
		return CHARGE_RETRY;
2597
	/*
2598 2599 2600 2601 2602 2603 2604
	 * 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.
2605
	 */
2606
	if (nr_pages <= (1 << PAGE_ALLOC_COSTLY_ORDER) && ret)
2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619
		return CHARGE_RETRY;

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

	/* If we don't need to call oom-killer at el, return immediately */
	if (!oom_check)
		return CHARGE_NOMEM;
	/* check OOM */
2620
	if (!mem_cgroup_handle_oom(mem_over_limit, gfp_mask, get_order(csize)))
2621 2622 2623 2624 2625
		return CHARGE_OOM_DIE;

	return CHARGE_RETRY;
}

2626
/*
2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645
 * __mem_cgroup_try_charge() does
 * 1. detect memcg to be charged against from passed *mm and *ptr,
 * 2. update res_counter
 * 3. call memory reclaim if necessary.
 *
 * In some special case, if the task is fatal, fatal_signal_pending() or
 * has TIF_MEMDIE, this function returns -EINTR while writing root_mem_cgroup
 * to *ptr. There are two reasons for this. 1: fatal threads should quit as soon
 * as possible without any hazards. 2: all pages should have a valid
 * pc->mem_cgroup. If mm is NULL and the caller doesn't pass a valid memcg
 * pointer, that is treated as a charge to root_mem_cgroup.
 *
 * So __mem_cgroup_try_charge() will return
 *  0       ...  on success, filling *ptr with a valid memcg pointer.
 *  -ENOMEM ...  charge failure because of resource limits.
 *  -EINTR  ...  if thread is fatal. *ptr is filled with root_mem_cgroup.
 *
 * Unlike the exported interface, an "oom" parameter is added. if oom==true,
 * the oom-killer can be invoked.
2646
 */
2647
static int __mem_cgroup_try_charge(struct mm_struct *mm,
A
Andrea Arcangeli 已提交
2648
				   gfp_t gfp_mask,
2649
				   unsigned int nr_pages,
2650
				   struct mem_cgroup **ptr,
2651
				   bool oom)
2652
{
2653
	unsigned int batch = max(CHARGE_BATCH, nr_pages);
2654
	int nr_oom_retries = MEM_CGROUP_RECLAIM_RETRIES;
2655
	struct mem_cgroup *memcg = NULL;
2656
	int ret;
2657

K
KAMEZAWA Hiroyuki 已提交
2658 2659 2660 2661 2662 2663 2664 2665
	/*
	 * Unlike gloval-vm's OOM-kill, we're not in memory shortage
	 * in system level. So, allow to go ahead dying process in addition to
	 * MEMDIE process.
	 */
	if (unlikely(test_thread_flag(TIF_MEMDIE)
		     || fatal_signal_pending(current)))
		goto bypass;
2666

2667
	/*
2668 2669
	 * We always charge the cgroup the mm_struct belongs to.
	 * The mm_struct's mem_cgroup changes on task migration if the
2670
	 * thread group leader migrates. It's possible that mm is not
2671
	 * set, if so charge the root memcg (happens for pagecache usage).
2672
	 */
2673
	if (!*ptr && !mm)
2674
		*ptr = root_mem_cgroup;
K
KAMEZAWA Hiroyuki 已提交
2675
again:
2676 2677 2678
	if (*ptr) { /* css should be a valid one */
		memcg = *ptr;
		if (mem_cgroup_is_root(memcg))
K
KAMEZAWA Hiroyuki 已提交
2679
			goto done;
2680
		if (consume_stock(memcg, nr_pages))
K
KAMEZAWA Hiroyuki 已提交
2681
			goto done;
2682
		css_get(&memcg->css);
2683
	} else {
K
KAMEZAWA Hiroyuki 已提交
2684
		struct task_struct *p;
2685

K
KAMEZAWA Hiroyuki 已提交
2686 2687 2688
		rcu_read_lock();
		p = rcu_dereference(mm->owner);
		/*
2689
		 * Because we don't have task_lock(), "p" can exit.
2690
		 * In that case, "memcg" can point to root or p can be NULL with
2691 2692 2693 2694 2695 2696
		 * race with swapoff. Then, we have small risk of mis-accouning.
		 * But such kind of mis-account by race always happens because
		 * we don't have cgroup_mutex(). It's overkill and we allo that
		 * small race, here.
		 * (*) swapoff at el will charge against mm-struct not against
		 * task-struct. So, mm->owner can be NULL.
K
KAMEZAWA Hiroyuki 已提交
2697
		 */
2698
		memcg = mem_cgroup_from_task(p);
2699 2700 2701
		if (!memcg)
			memcg = root_mem_cgroup;
		if (mem_cgroup_is_root(memcg)) {
K
KAMEZAWA Hiroyuki 已提交
2702 2703 2704
			rcu_read_unlock();
			goto done;
		}
2705
		if (consume_stock(memcg, nr_pages)) {
K
KAMEZAWA Hiroyuki 已提交
2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717
			/*
			 * It seems dagerous to access memcg without css_get().
			 * But considering how consume_stok works, it's not
			 * necessary. If consume_stock success, some charges
			 * from this memcg are cached on this cpu. So, we
			 * don't need to call css_get()/css_tryget() before
			 * calling consume_stock().
			 */
			rcu_read_unlock();
			goto done;
		}
		/* after here, we may be blocked. we need to get refcnt */
2718
		if (!css_tryget(&memcg->css)) {
K
KAMEZAWA Hiroyuki 已提交
2719 2720 2721 2722 2723
			rcu_read_unlock();
			goto again;
		}
		rcu_read_unlock();
	}
2724

2725 2726
	do {
		bool oom_check;
2727

2728
		/* If killed, bypass charge */
K
KAMEZAWA Hiroyuki 已提交
2729
		if (fatal_signal_pending(current)) {
2730
			css_put(&memcg->css);
2731
			goto bypass;
K
KAMEZAWA Hiroyuki 已提交
2732
		}
2733

2734 2735 2736 2737
		oom_check = false;
		if (oom && !nr_oom_retries) {
			oom_check = true;
			nr_oom_retries = MEM_CGROUP_RECLAIM_RETRIES;
2738
		}
2739

2740 2741
		ret = mem_cgroup_do_charge(memcg, gfp_mask, batch, nr_pages,
		    oom_check);
2742 2743 2744 2745
		switch (ret) {
		case CHARGE_OK:
			break;
		case CHARGE_RETRY: /* not in OOM situation but retry */
2746
			batch = nr_pages;
2747 2748
			css_put(&memcg->css);
			memcg = NULL;
K
KAMEZAWA Hiroyuki 已提交
2749
			goto again;
2750
		case CHARGE_WOULDBLOCK: /* !__GFP_WAIT */
2751
			css_put(&memcg->css);
2752 2753
			goto nomem;
		case CHARGE_NOMEM: /* OOM routine works */
K
KAMEZAWA Hiroyuki 已提交
2754
			if (!oom) {
2755
				css_put(&memcg->css);
K
KAMEZAWA Hiroyuki 已提交
2756
				goto nomem;
K
KAMEZAWA Hiroyuki 已提交
2757
			}
2758 2759 2760 2761
			/* If oom, we never return -ENOMEM */
			nr_oom_retries--;
			break;
		case CHARGE_OOM_DIE: /* Killed by OOM Killer */
2762
			css_put(&memcg->css);
K
KAMEZAWA Hiroyuki 已提交
2763
			goto bypass;
2764
		}
2765 2766
	} while (ret != CHARGE_OK);

2767
	if (batch > nr_pages)
2768 2769
		refill_stock(memcg, batch - nr_pages);
	css_put(&memcg->css);
2770
done:
2771
	*ptr = memcg;
2772 2773
	return 0;
nomem:
2774
	*ptr = NULL;
2775
	return -ENOMEM;
K
KAMEZAWA Hiroyuki 已提交
2776
bypass:
2777 2778
	*ptr = root_mem_cgroup;
	return -EINTR;
2779
}
2780

2781 2782 2783 2784 2785
/*
 * Somemtimes we have to undo a charge we got by try_charge().
 * This function is for that and do uncharge, put css's refcnt.
 * gotten by try_charge().
 */
2786
static void __mem_cgroup_cancel_charge(struct mem_cgroup *memcg,
2787
				       unsigned int nr_pages)
2788
{
2789
	if (!mem_cgroup_is_root(memcg)) {
2790 2791
		unsigned long bytes = nr_pages * PAGE_SIZE;

2792
		res_counter_uncharge(&memcg->res, bytes);
2793
		if (do_swap_account)
2794
			res_counter_uncharge(&memcg->memsw, bytes);
2795
	}
2796 2797
}

2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815
/*
 * Cancel chrages in this cgroup....doesn't propagate to parent cgroup.
 * This is useful when moving usage to parent cgroup.
 */
static void __mem_cgroup_cancel_local_charge(struct mem_cgroup *memcg,
					unsigned int nr_pages)
{
	unsigned long bytes = nr_pages * PAGE_SIZE;

	if (mem_cgroup_is_root(memcg))
		return;

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

2816 2817
/*
 * A helper function to get mem_cgroup from ID. must be called under
T
Tejun Heo 已提交
2818 2819 2820
 * rcu_read_lock().  The caller is responsible for calling css_tryget if
 * the mem_cgroup is used for charging. (dropping refcnt from swap can be
 * called against removed memcg.)
2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831
 */
static struct mem_cgroup *mem_cgroup_lookup(unsigned short id)
{
	struct cgroup_subsys_state *css;

	/* ID 0 is unused ID */
	if (!id)
		return NULL;
	css = css_lookup(&mem_cgroup_subsys, id);
	if (!css)
		return NULL;
2832
	return mem_cgroup_from_css(css);
2833 2834
}

2835
struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page)
2836
{
2837
	struct mem_cgroup *memcg = NULL;
2838
	struct page_cgroup *pc;
2839
	unsigned short id;
2840 2841
	swp_entry_t ent;

2842 2843 2844
	VM_BUG_ON(!PageLocked(page));

	pc = lookup_page_cgroup(page);
2845
	lock_page_cgroup(pc);
2846
	if (PageCgroupUsed(pc)) {
2847 2848 2849
		memcg = pc->mem_cgroup;
		if (memcg && !css_tryget(&memcg->css))
			memcg = NULL;
2850
	} else if (PageSwapCache(page)) {
2851
		ent.val = page_private(page);
2852
		id = lookup_swap_cgroup_id(ent);
2853
		rcu_read_lock();
2854 2855 2856
		memcg = mem_cgroup_lookup(id);
		if (memcg && !css_tryget(&memcg->css))
			memcg = NULL;
2857
		rcu_read_unlock();
2858
	}
2859
	unlock_page_cgroup(pc);
2860
	return memcg;
2861 2862
}

2863
static void __mem_cgroup_commit_charge(struct mem_cgroup *memcg,
2864
				       struct page *page,
2865
				       unsigned int nr_pages,
2866 2867
				       enum charge_type ctype,
				       bool lrucare)
2868
{
2869
	struct page_cgroup *pc = lookup_page_cgroup(page);
2870
	struct zone *uninitialized_var(zone);
2871
	struct lruvec *lruvec;
2872
	bool was_on_lru = false;
2873
	bool anon;
2874

2875
	lock_page_cgroup(pc);
2876
	VM_BUG_ON(PageCgroupUsed(pc));
2877 2878 2879 2880
	/*
	 * we don't need page_cgroup_lock about tail pages, becase they are not
	 * accessed by any other context at this point.
	 */
2881 2882 2883 2884 2885 2886 2887 2888 2889

	/*
	 * In some cases, SwapCache and FUSE(splice_buf->radixtree), the page
	 * may already be on some other mem_cgroup's LRU.  Take care of it.
	 */
	if (lrucare) {
		zone = page_zone(page);
		spin_lock_irq(&zone->lru_lock);
		if (PageLRU(page)) {
2890
			lruvec = mem_cgroup_zone_lruvec(zone, pc->mem_cgroup);
2891
			ClearPageLRU(page);
2892
			del_page_from_lru_list(page, lruvec, page_lru(page));
2893 2894 2895 2896
			was_on_lru = true;
		}
	}

2897
	pc->mem_cgroup = memcg;
2898 2899 2900 2901 2902 2903 2904
	/*
	 * We access a page_cgroup asynchronously without lock_page_cgroup().
	 * Especially when a page_cgroup is taken from a page, pc->mem_cgroup
	 * is accessed after testing USED bit. To make pc->mem_cgroup visible
	 * before USED bit, we need memory barrier here.
	 * See mem_cgroup_add_lru_list(), etc.
 	 */
K
KAMEZAWA Hiroyuki 已提交
2905
	smp_wmb();
2906
	SetPageCgroupUsed(pc);
2907

2908 2909
	if (lrucare) {
		if (was_on_lru) {
2910
			lruvec = mem_cgroup_zone_lruvec(zone, pc->mem_cgroup);
2911 2912
			VM_BUG_ON(PageLRU(page));
			SetPageLRU(page);
2913
			add_page_to_lru_list(page, lruvec, page_lru(page));
2914 2915 2916 2917
		}
		spin_unlock_irq(&zone->lru_lock);
	}

2918
	if (ctype == MEM_CGROUP_CHARGE_TYPE_ANON)
2919 2920 2921 2922
		anon = true;
	else
		anon = false;

2923
	mem_cgroup_charge_statistics(memcg, page, anon, nr_pages);
2924
	unlock_page_cgroup(pc);
2925

2926 2927 2928 2929 2930
	/*
	 * "charge_statistics" updated event counter. Then, check it.
	 * Insert ancestor (and ancestor's ancestors), to softlimit RB-tree.
	 * if they exceeds softlimit.
	 */
2931
	memcg_check_events(memcg, page);
2932
}
2933

2934 2935
static DEFINE_MUTEX(set_limit_mutex);

2936 2937 2938 2939 2940 2941 2942
#ifdef CONFIG_MEMCG_KMEM
static inline bool memcg_can_account_kmem(struct mem_cgroup *memcg)
{
	return !mem_cgroup_disabled() && !mem_cgroup_is_root(memcg) &&
		(memcg->kmem_account_flags & KMEM_ACCOUNTED_MASK);
}

G
Glauber Costa 已提交
2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955
/*
 * 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;
	return cachep->memcg_params->memcg_caches[memcg_cache_id(p->memcg)];
}

2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976
#ifdef CONFIG_SLABINFO
static int mem_cgroup_slabinfo_read(struct cgroup *cont, struct cftype *cft,
					struct seq_file *m)
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
	struct memcg_cache_params *params;

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

	print_slabinfo_header(m);

	mutex_lock(&memcg->slab_caches_mutex);
	list_for_each_entry(params, &memcg->memcg_slab_caches, list)
		cache_show(memcg_params_to_cache(params), m);
	mutex_unlock(&memcg->slab_caches_mutex);

	return 0;
}
#endif

2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029
static int memcg_charge_kmem(struct mem_cgroup *memcg, gfp_t gfp, u64 size)
{
	struct res_counter *fail_res;
	struct mem_cgroup *_memcg;
	int ret = 0;
	bool may_oom;

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

	/*
	 * Conditions under which we can wait for the oom_killer. Those are
	 * the same conditions tested by the core page allocator
	 */
	may_oom = (gfp & __GFP_FS) && !(gfp & __GFP_NORETRY);

	_memcg = memcg;
	ret = __mem_cgroup_try_charge(NULL, gfp, size >> PAGE_SHIFT,
				      &_memcg, may_oom);

	if (ret == -EINTR)  {
		/*
		 * __mem_cgroup_try_charge() chosed 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
		 * our minds.
		 *
		 * This condition will only trigger if the task entered
		 * memcg_charge_kmem in a sane state, but was OOM-killed during
		 * __mem_cgroup_try_charge() above. Tasks that were already
		 * dying when the allocation triggers should have been already
		 * directed to the root cgroup in memcontrol.h
		 */
		res_counter_charge_nofail(&memcg->res, size, &fail_res);
		if (do_swap_account)
			res_counter_charge_nofail(&memcg->memsw, size,
						  &fail_res);
		ret = 0;
	} else if (ret)
		res_counter_uncharge(&memcg->kmem, size);

	return ret;
}

static void memcg_uncharge_kmem(struct mem_cgroup *memcg, u64 size)
{
	res_counter_uncharge(&memcg->res, size);
	if (do_swap_account)
		res_counter_uncharge(&memcg->memsw, size);
3030 3031 3032 3033 3034

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

3035 3036 3037 3038 3039 3040 3041 3042
	/*
	 * Releases a reference taken in kmem_cgroup_css_offline in case
	 * this last uncharge is racing with the offlining code or it is
	 * outliving the memcg existence.
	 *
	 * The memory barrier imposed by test&clear is paired with the
	 * explicit one in memcg_kmem_mark_dead().
	 */
3043
	if (memcg_kmem_test_and_clear_dead(memcg))
3044
		css_put(&memcg->css);
3045 3046
}

3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066
void memcg_cache_list_add(struct mem_cgroup *memcg, struct kmem_cache *cachep)
{
	if (!memcg)
		return;

	mutex_lock(&memcg->slab_caches_mutex);
	list_add(&cachep->memcg_params->list, &memcg->memcg_slab_caches);
	mutex_unlock(&memcg->slab_caches_mutex);
}

/*
 * 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;
}

3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129
/*
 * This ends up being protected by the set_limit mutex, during normal
 * operation, because that is its main call site.
 *
 * But when we create a new cache, we can call this as well if its parent
 * is kmem-limited. That will have to hold set_limit_mutex as well.
 */
int memcg_update_cache_sizes(struct mem_cgroup *memcg)
{
	int num, ret;

	num = ida_simple_get(&kmem_limited_groups,
				0, MEMCG_CACHES_MAX_SIZE, GFP_KERNEL);
	if (num < 0)
		return num;
	/*
	 * After this point, kmem_accounted (that we test atomically in
	 * the beginning of this conditional), is no longer 0. This
	 * guarantees only one process will set the following boolean
	 * to true. We don't need test_and_set because we're protected
	 * by the set_limit_mutex anyway.
	 */
	memcg_kmem_set_activated(memcg);

	ret = memcg_update_all_caches(num+1);
	if (ret) {
		ida_simple_remove(&kmem_limited_groups, num);
		memcg_kmem_clear_activated(memcg);
		return ret;
	}

	memcg->kmemcg_id = num;
	INIT_LIST_HEAD(&memcg->memcg_slab_caches);
	mutex_init(&memcg->slab_caches_mutex);
	return 0;
}

static size_t memcg_caches_array_size(int num_groups)
{
	ssize_t size;
	if (num_groups <= 0)
		return 0;

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

	return size;
}

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

3130 3131
static void kmem_cache_destroy_work_func(struct work_struct *w);

3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182
int memcg_update_cache_size(struct kmem_cache *s, int num_groups)
{
	struct memcg_cache_params *cur_params = s->memcg_params;

	VM_BUG_ON(s->memcg_params && !s->memcg_params->is_root_cache);

	if (num_groups > memcg_limited_groups_array_size) {
		int i;
		ssize_t size = memcg_caches_array_size(num_groups);

		size *= sizeof(void *);
		size += sizeof(struct memcg_cache_params);

		s->memcg_params = kzalloc(size, GFP_KERNEL);
		if (!s->memcg_params) {
			s->memcg_params = cur_params;
			return -ENOMEM;
		}

		s->memcg_params->is_root_cache = true;

		/*
		 * There is the chance it will be bigger than
		 * memcg_limited_groups_array_size, if we failed an allocation
		 * in a cache, in which case all caches updated before it, will
		 * have a bigger array.
		 *
		 * But if that is the case, the data after
		 * memcg_limited_groups_array_size is certainly unused
		 */
		for (i = 0; i < memcg_limited_groups_array_size; i++) {
			if (!cur_params->memcg_caches[i])
				continue;
			s->memcg_params->memcg_caches[i] =
						cur_params->memcg_caches[i];
		}

		/*
		 * Ideally, we would wait until all caches succeed, and only
		 * then free the old one. But this is not worth the extra
		 * pointer per-cache we'd have to have for this.
		 *
		 * It is not a big deal if some caches are left with a size
		 * bigger than the others. And all updates will reset this
		 * anyway.
		 */
		kfree(cur_params);
	}
	return 0;
}

G
Glauber Costa 已提交
3183 3184
int memcg_register_cache(struct mem_cgroup *memcg, struct kmem_cache *s,
			 struct kmem_cache *root_cache)
3185 3186 3187 3188 3189 3190
{
	size_t size = sizeof(struct memcg_cache_params);

	if (!memcg_kmem_enabled())
		return 0;

3191 3192 3193
	if (!memcg)
		size += memcg_limited_groups_array_size * sizeof(void *);

3194 3195 3196 3197
	s->memcg_params = kzalloc(size, GFP_KERNEL);
	if (!s->memcg_params)
		return -ENOMEM;

3198 3199
	INIT_WORK(&s->memcg_params->destroy,
			kmem_cache_destroy_work_func);
G
Glauber Costa 已提交
3200
	if (memcg) {
3201
		s->memcg_params->memcg = memcg;
G
Glauber Costa 已提交
3202
		s->memcg_params->root_cache = root_cache;
3203 3204 3205
	} else
		s->memcg_params->is_root_cache = true;

3206 3207 3208 3209 3210
	return 0;
}

void memcg_release_cache(struct kmem_cache *s)
{
3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234
	struct kmem_cache *root;
	struct mem_cgroup *memcg;
	int id;

	/*
	 * This happens, for instance, when a root cache goes away before we
	 * add any memcg.
	 */
	if (!s->memcg_params)
		return;

	if (s->memcg_params->is_root_cache)
		goto out;

	memcg = s->memcg_params->memcg;
	id  = memcg_cache_id(memcg);

	root = s->memcg_params->root_cache;
	root->memcg_params->memcg_caches[id] = NULL;

	mutex_lock(&memcg->slab_caches_mutex);
	list_del(&s->memcg_params->list);
	mutex_unlock(&memcg->slab_caches_mutex);

3235
	css_put(&memcg->css);
3236
out:
3237 3238 3239
	kfree(s->memcg_params);
}

3240 3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270
/*
 * During the creation a new cache, we need to disable our accounting mechanism
 * altogether. This is true even if we are not creating, but rather just
 * enqueing new caches to be created.
 *
 * This is because that process will trigger allocations; some visible, like
 * explicit kmallocs to auxiliary data structures, name strings and internal
 * cache structures; some well concealed, like INIT_WORK() that can allocate
 * objects during debug.
 *
 * If any allocation happens during memcg_kmem_get_cache, we will recurse back
 * to it. This may not be a bounded recursion: since the first cache creation
 * failed to complete (waiting on the allocation), we'll just try to create the
 * cache again, failing at the same point.
 *
 * memcg_kmem_get_cache is prepared to abort after seeing a positive count of
 * memcg_kmem_skip_account. So we enclose anything that might allocate memory
 * inside the following two functions.
 */
static inline void memcg_stop_kmem_account(void)
{
	VM_BUG_ON(!current->mm);
	current->memcg_kmem_skip_account++;
}

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

G
Glauber Costa 已提交
3271 3272 3273 3274 3275 3276 3277 3278 3279
static void kmem_cache_destroy_work_func(struct work_struct *w)
{
	struct kmem_cache *cachep;
	struct memcg_cache_params *p;

	p = container_of(w, struct memcg_cache_params, destroy);

	cachep = memcg_params_to_cache(p);

G
Glauber Costa 已提交
3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300
	/*
	 * If we get down to 0 after shrink, we could delete right away.
	 * However, memcg_release_pages() already puts us back in the workqueue
	 * in that case. If we proceed deleting, we'll get a dangling
	 * reference, and removing the object from the workqueue in that case
	 * is unnecessary complication. We are not a fast path.
	 *
	 * Note that this case is fundamentally different from racing with
	 * shrink_slab(): if memcg_cgroup_destroy_cache() is called in
	 * kmem_cache_shrink, not only we would be reinserting a dead cache
	 * into the queue, but doing so from inside the worker racing to
	 * destroy it.
	 *
	 * So if we aren't down to zero, we'll just schedule a worker and try
	 * again
	 */
	if (atomic_read(&cachep->memcg_params->nr_pages) != 0) {
		kmem_cache_shrink(cachep);
		if (atomic_read(&cachep->memcg_params->nr_pages) == 0)
			return;
	} else
G
Glauber Costa 已提交
3301 3302 3303 3304 3305 3306 3307 3308
		kmem_cache_destroy(cachep);
}

void mem_cgroup_destroy_cache(struct kmem_cache *cachep)
{
	if (!cachep->memcg_params->dead)
		return;

G
Glauber Costa 已提交
3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328
	/*
	 * There are many ways in which we can get here.
	 *
	 * We can get to a memory-pressure situation while the delayed work is
	 * still pending to run. The vmscan shrinkers can then release all
	 * cache memory and get us to destruction. If this is the case, we'll
	 * be executed twice, which is a bug (the second time will execute over
	 * bogus data). In this case, cancelling the work should be fine.
	 *
	 * But we can also get here from the worker itself, if
	 * kmem_cache_shrink is enough to shake all the remaining objects and
	 * get the page count to 0. In this case, we'll deadlock if we try to
	 * cancel the work (the worker runs with an internal lock held, which
	 * is the same lock we would hold for cancel_work_sync().)
	 *
	 * Since we can't possibly know who got us here, just refrain from
	 * running if there is already work pending
	 */
	if (work_pending(&cachep->memcg_params->destroy))
		return;
G
Glauber Costa 已提交
3329 3330 3331 3332 3333 3334 3335
	/*
	 * We have to defer the actual destroying to a workqueue, because
	 * we might currently be in a context that cannot sleep.
	 */
	schedule_work(&cachep->memcg_params->destroy);
}

3336 3337 3338 3339 3340 3341 3342 3343 3344
/*
 * This lock protects updaters, not readers. We want readers to be as fast as
 * they can, and they will either see NULL or a valid cache value. Our model
 * allow them to see NULL, in which case the root memcg will be selected.
 *
 * We need this lock because multiple allocations to the same cache from a non
 * will span more than one worker. Only one of them can create the cache.
 */
static DEFINE_MUTEX(memcg_cache_mutex);
3345

3346 3347 3348
/*
 * Called with memcg_cache_mutex held
 */
3349 3350 3351 3352
static struct kmem_cache *kmem_cache_dup(struct mem_cgroup *memcg,
					 struct kmem_cache *s)
{
	struct kmem_cache *new;
3353
	static char *tmp_name = NULL;
3354

3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372
	lockdep_assert_held(&memcg_cache_mutex);

	/*
	 * kmem_cache_create_memcg duplicates the given name and
	 * cgroup_name for this name requires RCU context.
	 * This static temporary buffer is used to prevent from
	 * pointless shortliving allocation.
	 */
	if (!tmp_name) {
		tmp_name = kmalloc(PATH_MAX, GFP_KERNEL);
		if (!tmp_name)
			return NULL;
	}

	rcu_read_lock();
	snprintf(tmp_name, PATH_MAX, "%s(%d:%s)", s->name,
			 memcg_cache_id(memcg), cgroup_name(memcg->css.cgroup));
	rcu_read_unlock();
3373

3374
	new = kmem_cache_create_memcg(memcg, tmp_name, s->object_size, s->align,
G
Glauber Costa 已提交
3375
				      (s->flags & ~SLAB_PANIC), s->ctor, s);
3376

3377 3378 3379
	if (new)
		new->allocflags |= __GFP_KMEMCG;

3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394
	return new;
}

static struct kmem_cache *memcg_create_kmem_cache(struct mem_cgroup *memcg,
						  struct kmem_cache *cachep)
{
	struct kmem_cache *new_cachep;
	int idx;

	BUG_ON(!memcg_can_account_kmem(memcg));

	idx = memcg_cache_id(memcg);

	mutex_lock(&memcg_cache_mutex);
	new_cachep = cachep->memcg_params->memcg_caches[idx];
3395 3396
	if (new_cachep) {
		css_put(&memcg->css);
3397
		goto out;
3398
	}
3399 3400 3401 3402

	new_cachep = kmem_cache_dup(memcg, cachep);
	if (new_cachep == NULL) {
		new_cachep = cachep;
3403
		css_put(&memcg->css);
3404 3405 3406
		goto out;
	}

G
Glauber Costa 已提交
3407
	atomic_set(&new_cachep->memcg_params->nr_pages , 0);
3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419

	cachep->memcg_params->memcg_caches[idx] = new_cachep;
	/*
	 * the readers won't lock, make sure everybody sees the updated value,
	 * so they won't put stuff in the queue again for no reason
	 */
	wmb();
out:
	mutex_unlock(&memcg_cache_mutex);
	return new_cachep;
}

3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458
void kmem_cache_destroy_memcg_children(struct kmem_cache *s)
{
	struct kmem_cache *c;
	int i;

	if (!s->memcg_params)
		return;
	if (!s->memcg_params->is_root_cache)
		return;

	/*
	 * If the cache is being destroyed, we trust that there is no one else
	 * requesting objects from it. Even if there are, the sanity checks in
	 * kmem_cache_destroy should caught this ill-case.
	 *
	 * Still, we don't want anyone else freeing memcg_caches under our
	 * noses, which can happen if a new memcg comes to life. As usual,
	 * we'll take the set_limit_mutex to protect ourselves against this.
	 */
	mutex_lock(&set_limit_mutex);
	for (i = 0; i < memcg_limited_groups_array_size; i++) {
		c = s->memcg_params->memcg_caches[i];
		if (!c)
			continue;

		/*
		 * We will now manually delete the caches, so to avoid races
		 * we need to cancel all pending destruction workers and
		 * proceed with destruction ourselves.
		 *
		 * kmem_cache_destroy() will call kmem_cache_shrink internally,
		 * and that could spawn the workers again: it is likely that
		 * the cache still have active pages until this very moment.
		 * This would lead us back to mem_cgroup_destroy_cache.
		 *
		 * But that will not execute at all if the "dead" flag is not
		 * set, so flip it down to guarantee we are in control.
		 */
		c->memcg_params->dead = false;
G
Glauber Costa 已提交
3459
		cancel_work_sync(&c->memcg_params->destroy);
3460 3461 3462 3463 3464
		kmem_cache_destroy(c);
	}
	mutex_unlock(&set_limit_mutex);
}

3465 3466 3467 3468 3469 3470
struct create_work {
	struct mem_cgroup *memcg;
	struct kmem_cache *cachep;
	struct work_struct work;
};

G
Glauber Costa 已提交
3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487
static void mem_cgroup_destroy_all_caches(struct mem_cgroup *memcg)
{
	struct kmem_cache *cachep;
	struct memcg_cache_params *params;

	if (!memcg_kmem_is_active(memcg))
		return;

	mutex_lock(&memcg->slab_caches_mutex);
	list_for_each_entry(params, &memcg->memcg_slab_caches, list) {
		cachep = memcg_params_to_cache(params);
		cachep->memcg_params->dead = true;
		schedule_work(&cachep->memcg_params->destroy);
	}
	mutex_unlock(&memcg->slab_caches_mutex);
}

3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499
static void memcg_create_cache_work_func(struct work_struct *w)
{
	struct create_work *cw;

	cw = container_of(w, struct create_work, work);
	memcg_create_kmem_cache(cw->memcg, cw->cachep);
	kfree(cw);
}

/*
 * Enqueue the creation of a per-memcg kmem_cache.
 */
3500 3501
static void __memcg_create_cache_enqueue(struct mem_cgroup *memcg,
					 struct kmem_cache *cachep)
3502 3503 3504 3505
{
	struct create_work *cw;

	cw = kmalloc(sizeof(struct create_work), GFP_NOWAIT);
3506 3507
	if (cw == NULL) {
		css_put(&memcg->css);
3508 3509 3510 3511 3512 3513 3514 3515 3516 3517
		return;
	}

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

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

3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535
static void memcg_create_cache_enqueue(struct mem_cgroup *memcg,
				       struct kmem_cache *cachep)
{
	/*
	 * We need to stop accounting when we kmalloc, because if the
	 * corresponding kmalloc cache is not yet created, the first allocation
	 * in __memcg_create_cache_enqueue will recurse.
	 *
	 * However, it is better to enclose the whole function. Depending on
	 * the debugging options enabled, INIT_WORK(), for instance, can
	 * trigger an allocation. This too, will make us recurse. Because at
	 * this point we can't allow ourselves back into memcg_kmem_get_cache,
	 * the safest choice is to do it like this, wrapping the whole function.
	 */
	memcg_stop_kmem_account();
	__memcg_create_cache_enqueue(memcg, cachep);
	memcg_resume_kmem_account();
}
3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557
/*
 * Return the kmem_cache we're supposed to use for a slab allocation.
 * We try to use the current memcg's version of the cache.
 *
 * If the cache does not exist yet, if we are the first user of it,
 * we either create it immediately, if possible, or create it asynchronously
 * in a workqueue.
 * In the latter case, we will let the current allocation go through with
 * the original cache.
 *
 * Can't be called in interrupt context or from kernel threads.
 * This function needs to be called with rcu_read_lock() held.
 */
struct kmem_cache *__memcg_kmem_get_cache(struct kmem_cache *cachep,
					  gfp_t gfp)
{
	struct mem_cgroup *memcg;
	int idx;

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

3558 3559 3560
	if (!current->mm || current->memcg_kmem_skip_account)
		return cachep;

3561 3562 3563 3564
	rcu_read_lock();
	memcg = mem_cgroup_from_task(rcu_dereference(current->mm->owner));

	if (!memcg_can_account_kmem(memcg))
3565
		goto out;
3566 3567 3568 3569 3570 3571 3572 3573

	idx = memcg_cache_id(memcg);

	/*
	 * barrier to mare sure we're always seeing the up to date value.  The
	 * code updating memcg_caches will issue a write barrier to match this.
	 */
	read_barrier_depends();
3574 3575 3576
	if (likely(cachep->memcg_params->memcg_caches[idx])) {
		cachep = cachep->memcg_params->memcg_caches[idx];
		goto out;
3577 3578
	}

3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605
	/* The corresponding put will be done in the workqueue. */
	if (!css_tryget(&memcg->css))
		goto out;
	rcu_read_unlock();

	/*
	 * If we are in a safe context (can wait, and not in interrupt
	 * context), we could be be predictable and return right away.
	 * This would guarantee that the allocation being performed
	 * already belongs in the new cache.
	 *
	 * However, there are some clashes that can arrive from locking.
	 * For instance, because we acquire the slab_mutex while doing
	 * kmem_cache_dup, this means no further allocation could happen
	 * with the slab_mutex held.
	 *
	 * Also, because cache creation issue get_online_cpus(), this
	 * creates a lock chain: memcg_slab_mutex -> cpu_hotplug_mutex,
	 * that ends up reversed during cpu hotplug. (cpuset allocates
	 * a bunch of GFP_KERNEL memory during cpuup). Due to all that,
	 * better to defer everything.
	 */
	memcg_create_cache_enqueue(memcg, cachep);
	return cachep;
out:
	rcu_read_unlock();
	return cachep;
3606 3607 3608
}
EXPORT_SYMBOL(__memcg_kmem_get_cache);

3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629
/*
 * 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;
3630 3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657

	/*
	 * Disabling accounting is only relevant for some specific memcg
	 * internal allocations. Therefore we would initially not have such
	 * check here, since direct calls to the page allocator that are marked
	 * with GFP_KMEMCG only happen outside memcg core. We are mostly
	 * concerned with cache allocations, and by having this test at
	 * memcg_kmem_get_cache, we are already able to relay the allocation to
	 * the root cache and bypass the memcg cache altogether.
	 *
	 * There is one exception, though: the SLUB allocator does not create
	 * large order caches, but rather service large kmallocs directly from
	 * the page allocator. Therefore, the following sequence when backed by
	 * the SLUB allocator:
	 *
	 * 	memcg_stop_kmem_account();
	 * 	kmalloc(<large_number>)
	 * 	memcg_resume_kmem_account();
	 *
	 * would effectively ignore the fact that we should skip accounting,
	 * since it will drive us directly to this function without passing
	 * through the cache selector memcg_kmem_get_cache. Such large
	 * allocations are extremely rare but can happen, for instance, for the
	 * cache arrays. We bring this test here.
	 */
	if (!current->mm || current->memcg_kmem_skip_account)
		return true;

3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731
	memcg = try_get_mem_cgroup_from_mm(current->mm);

	/*
	 * very rare case described in mem_cgroup_from_task. Unfortunately there
	 * isn't much we can do without complicating this too much, and it would
	 * be gfp-dependent anyway. Just let it go
	 */
	if (unlikely(!memcg))
		return true;

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

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

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

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

	VM_BUG_ON(mem_cgroup_is_root(memcg));

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

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

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


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

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

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

	VM_BUG_ON(mem_cgroup_is_root(memcg));
	memcg_uncharge_kmem(memcg, PAGE_SIZE << order);
}
G
Glauber Costa 已提交
3732 3733 3734 3735
#else
static inline void mem_cgroup_destroy_all_caches(struct mem_cgroup *memcg)
{
}
3736 3737
#endif /* CONFIG_MEMCG_KMEM */

3738 3739
#ifdef CONFIG_TRANSPARENT_HUGEPAGE

3740
#define PCGF_NOCOPY_AT_SPLIT (1 << PCG_LOCK | 1 << PCG_MIGRATION)
3741 3742
/*
 * Because tail pages are not marked as "used", set it. We're under
3743 3744 3745
 * 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.
3746
 */
3747
void mem_cgroup_split_huge_fixup(struct page *head)
3748 3749
{
	struct page_cgroup *head_pc = lookup_page_cgroup(head);
3750
	struct page_cgroup *pc;
3751
	struct mem_cgroup *memcg;
3752
	int i;
3753

3754 3755
	if (mem_cgroup_disabled())
		return;
3756 3757

	memcg = head_pc->mem_cgroup;
3758 3759
	for (i = 1; i < HPAGE_PMD_NR; i++) {
		pc = head_pc + i;
3760
		pc->mem_cgroup = memcg;
3761 3762 3763
		smp_wmb();/* see __commit_charge() */
		pc->flags = head_pc->flags & ~PCGF_NOCOPY_AT_SPLIT;
	}
3764 3765
	__this_cpu_sub(memcg->stat->count[MEM_CGROUP_STAT_RSS_HUGE],
		       HPAGE_PMD_NR);
3766
}
3767
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
3768

3769
/**
3770
 * mem_cgroup_move_account - move account of the page
3771
 * @page: the page
3772
 * @nr_pages: number of regular pages (>1 for huge pages)
3773 3774 3775 3776 3777
 * @pc:	page_cgroup of the page.
 * @from: mem_cgroup which the page is moved from.
 * @to:	mem_cgroup which the page is moved to. @from != @to.
 *
 * The caller must confirm following.
K
KAMEZAWA Hiroyuki 已提交
3778
 * - page is not on LRU (isolate_page() is useful.)
3779
 * - compound_lock is held when nr_pages > 1
3780
 *
3781 3782
 * This function doesn't do "charge" to new cgroup and doesn't do "uncharge"
 * from old cgroup.
3783
 */
3784 3785 3786 3787
static int mem_cgroup_move_account(struct page *page,
				   unsigned int nr_pages,
				   struct page_cgroup *pc,
				   struct mem_cgroup *from,
3788
				   struct mem_cgroup *to)
3789
{
3790 3791
	unsigned long flags;
	int ret;
3792
	bool anon = PageAnon(page);
3793

3794
	VM_BUG_ON(from == to);
3795
	VM_BUG_ON(PageLRU(page));
3796 3797 3798 3799 3800 3801 3802
	/*
	 * 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;
3803
	if (nr_pages > 1 && !PageTransHuge(page))
3804 3805 3806 3807 3808 3809 3810 3811
		goto out;

	lock_page_cgroup(pc);

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

3812
	move_lock_mem_cgroup(from, &flags);
3813

3814
	if (!anon && page_mapped(page)) {
3815 3816 3817 3818 3819
		/* Update mapped_file data for mem_cgroup */
		preempt_disable();
		__this_cpu_dec(from->stat->count[MEM_CGROUP_STAT_FILE_MAPPED]);
		__this_cpu_inc(to->stat->count[MEM_CGROUP_STAT_FILE_MAPPED]);
		preempt_enable();
3820
	}
3821
	mem_cgroup_charge_statistics(from, page, anon, -nr_pages);
3822

3823
	/* caller should have done css_get */
K
KAMEZAWA Hiroyuki 已提交
3824
	pc->mem_cgroup = to;
3825
	mem_cgroup_charge_statistics(to, page, anon, nr_pages);
3826
	move_unlock_mem_cgroup(from, &flags);
3827 3828
	ret = 0;
unlock:
3829
	unlock_page_cgroup(pc);
3830 3831 3832
	/*
	 * check events
	 */
3833 3834
	memcg_check_events(to, page);
	memcg_check_events(from, page);
3835
out:
3836 3837 3838
	return ret;
}

3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858
/**
 * mem_cgroup_move_parent - moves page to the parent group
 * @page: the page to move
 * @pc: page_cgroup of the page
 * @child: page's cgroup
 *
 * move charges to its parent or the root cgroup if the group has no
 * parent (aka use_hierarchy==0).
 * Although this might fail (get_page_unless_zero, isolate_lru_page or
 * mem_cgroup_move_account fails) the failure is always temporary and
 * it signals a race with a page removal/uncharge or migration. In the
 * first case the page is on the way out and it will vanish from the LRU
 * on the next attempt and the call should be retried later.
 * Isolation from the LRU fails only if page has been isolated from
 * the LRU since we looked at it and that usually means either global
 * reclaim or migration going on. The page will either get back to the
 * LRU or vanish.
 * Finaly mem_cgroup_move_account fails only if the page got uncharged
 * (!PageCgroupUsed) or moved to a different group. The page will
 * disappear in the next attempt.
3859
 */
3860 3861
static int mem_cgroup_move_parent(struct page *page,
				  struct page_cgroup *pc,
3862
				  struct mem_cgroup *child)
3863 3864
{
	struct mem_cgroup *parent;
3865
	unsigned int nr_pages;
3866
	unsigned long uninitialized_var(flags);
3867 3868
	int ret;

3869
	VM_BUG_ON(mem_cgroup_is_root(child));
3870

3871 3872 3873 3874 3875
	ret = -EBUSY;
	if (!get_page_unless_zero(page))
		goto out;
	if (isolate_lru_page(page))
		goto put;
3876

3877
	nr_pages = hpage_nr_pages(page);
K
KAMEZAWA Hiroyuki 已提交
3878

3879 3880 3881 3882 3883 3884
	parent = parent_mem_cgroup(child);
	/*
	 * If no parent, move charges to root cgroup.
	 */
	if (!parent)
		parent = root_mem_cgroup;
3885

3886 3887
	if (nr_pages > 1) {
		VM_BUG_ON(!PageTransHuge(page));
3888
		flags = compound_lock_irqsave(page);
3889
	}
3890

3891
	ret = mem_cgroup_move_account(page, nr_pages,
3892
				pc, child, parent);
3893 3894
	if (!ret)
		__mem_cgroup_cancel_local_charge(child, nr_pages);
3895

3896
	if (nr_pages > 1)
3897
		compound_unlock_irqrestore(page, flags);
K
KAMEZAWA Hiroyuki 已提交
3898
	putback_lru_page(page);
3899
put:
3900
	put_page(page);
3901
out:
3902 3903 3904
	return ret;
}

3905 3906 3907 3908 3909 3910 3911
/*
 * Charge the memory controller for page usage.
 * Return
 * 0 if the charge was successful
 * < 0 if the cgroup is over its limit
 */
static int mem_cgroup_charge_common(struct page *page, struct mm_struct *mm,
3912
				gfp_t gfp_mask, enum charge_type ctype)
3913
{
3914
	struct mem_cgroup *memcg = NULL;
3915
	unsigned int nr_pages = 1;
3916
	bool oom = true;
3917
	int ret;
A
Andrea Arcangeli 已提交
3918

A
Andrea Arcangeli 已提交
3919
	if (PageTransHuge(page)) {
3920
		nr_pages <<= compound_order(page);
A
Andrea Arcangeli 已提交
3921
		VM_BUG_ON(!PageTransHuge(page));
3922 3923 3924 3925 3926
		/*
		 * Never OOM-kill a process for a huge page.  The
		 * fault handler will fall back to regular pages.
		 */
		oom = false;
A
Andrea Arcangeli 已提交
3927
	}
3928

3929
	ret = __mem_cgroup_try_charge(mm, gfp_mask, nr_pages, &memcg, oom);
3930
	if (ret == -ENOMEM)
3931
		return ret;
3932
	__mem_cgroup_commit_charge(memcg, page, nr_pages, ctype, false);
3933 3934 3935
	return 0;
}

3936 3937
int mem_cgroup_newpage_charge(struct page *page,
			      struct mm_struct *mm, gfp_t gfp_mask)
3938
{
3939
	if (mem_cgroup_disabled())
3940
		return 0;
3941 3942 3943
	VM_BUG_ON(page_mapped(page));
	VM_BUG_ON(page->mapping && !PageAnon(page));
	VM_BUG_ON(!mm);
3944
	return mem_cgroup_charge_common(page, mm, gfp_mask,
3945
					MEM_CGROUP_CHARGE_TYPE_ANON);
3946 3947
}

3948 3949 3950
/*
 * While swap-in, try_charge -> commit or cancel, the page is locked.
 * And when try_charge() successfully returns, one refcnt to memcg without
3951
 * struct page_cgroup is acquired. This refcnt will be consumed by
3952 3953
 * "commit()" or removed by "cancel()"
 */
3954 3955 3956 3957
static int __mem_cgroup_try_charge_swapin(struct mm_struct *mm,
					  struct page *page,
					  gfp_t mask,
					  struct mem_cgroup **memcgp)
3958
{
3959
	struct mem_cgroup *memcg;
3960
	struct page_cgroup *pc;
3961
	int ret;
3962

3963 3964 3965 3966 3967 3968 3969 3970 3971 3972
	pc = lookup_page_cgroup(page);
	/*
	 * Every swap fault against a single page tries to charge the
	 * page, bail as early as possible.  shmem_unuse() encounters
	 * already charged pages, too.  The USED bit is protected by
	 * the page lock, which serializes swap cache removal, which
	 * in turn serializes uncharging.
	 */
	if (PageCgroupUsed(pc))
		return 0;
3973 3974
	if (!do_swap_account)
		goto charge_cur_mm;
3975 3976
	memcg = try_get_mem_cgroup_from_page(page);
	if (!memcg)
3977
		goto charge_cur_mm;
3978 3979
	*memcgp = memcg;
	ret = __mem_cgroup_try_charge(NULL, mask, 1, memcgp, true);
3980
	css_put(&memcg->css);
3981 3982
	if (ret == -EINTR)
		ret = 0;
3983
	return ret;
3984
charge_cur_mm:
3985 3986 3987 3988
	ret = __mem_cgroup_try_charge(mm, mask, 1, memcgp, true);
	if (ret == -EINTR)
		ret = 0;
	return ret;
3989 3990
}

3991 3992 3993 3994 3995 3996
int mem_cgroup_try_charge_swapin(struct mm_struct *mm, struct page *page,
				 gfp_t gfp_mask, struct mem_cgroup **memcgp)
{
	*memcgp = NULL;
	if (mem_cgroup_disabled())
		return 0;
3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010
	/*
	 * A racing thread's fault, or swapoff, may have already
	 * updated the pte, and even removed page from swap cache: in
	 * those cases unuse_pte()'s pte_same() test will fail; but
	 * there's also a KSM case which does need to charge the page.
	 */
	if (!PageSwapCache(page)) {
		int ret;

		ret = __mem_cgroup_try_charge(mm, gfp_mask, 1, memcgp, true);
		if (ret == -EINTR)
			ret = 0;
		return ret;
	}
4011 4012 4013
	return __mem_cgroup_try_charge_swapin(mm, page, gfp_mask, memcgp);
}

4014 4015 4016 4017 4018 4019 4020 4021 4022
void mem_cgroup_cancel_charge_swapin(struct mem_cgroup *memcg)
{
	if (mem_cgroup_disabled())
		return;
	if (!memcg)
		return;
	__mem_cgroup_cancel_charge(memcg, 1);
}

D
Daisuke Nishimura 已提交
4023
static void
4024
__mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *memcg,
D
Daisuke Nishimura 已提交
4025
					enum charge_type ctype)
4026
{
4027
	if (mem_cgroup_disabled())
4028
		return;
4029
	if (!memcg)
4030
		return;
4031

4032
	__mem_cgroup_commit_charge(memcg, page, 1, ctype, true);
4033 4034 4035
	/*
	 * Now swap is on-memory. This means this page may be
	 * counted both as mem and swap....double count.
4036 4037 4038
	 * Fix it by uncharging from memsw. Basically, this SwapCache is stable
	 * under lock_page(). But in do_swap_page()::memory.c, reuse_swap_page()
	 * may call delete_from_swap_cache() before reach here.
4039
	 */
4040
	if (do_swap_account && PageSwapCache(page)) {
4041
		swp_entry_t ent = {.val = page_private(page)};
4042
		mem_cgroup_uncharge_swap(ent);
4043
	}
4044 4045
}

4046 4047
void mem_cgroup_commit_charge_swapin(struct page *page,
				     struct mem_cgroup *memcg)
D
Daisuke Nishimura 已提交
4048
{
4049
	__mem_cgroup_commit_charge_swapin(page, memcg,
4050
					  MEM_CGROUP_CHARGE_TYPE_ANON);
D
Daisuke Nishimura 已提交
4051 4052
}

4053 4054
int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm,
				gfp_t gfp_mask)
4055
{
4056 4057 4058 4059
	struct mem_cgroup *memcg = NULL;
	enum charge_type type = MEM_CGROUP_CHARGE_TYPE_CACHE;
	int ret;

4060
	if (mem_cgroup_disabled())
4061 4062 4063 4064 4065 4066 4067
		return 0;
	if (PageCompound(page))
		return 0;

	if (!PageSwapCache(page))
		ret = mem_cgroup_charge_common(page, mm, gfp_mask, type);
	else { /* page is swapcache/shmem */
4068 4069
		ret = __mem_cgroup_try_charge_swapin(mm, page,
						     gfp_mask, &memcg);
4070 4071 4072 4073
		if (!ret)
			__mem_cgroup_commit_charge_swapin(page, memcg, type);
	}
	return ret;
4074 4075
}

4076
static void mem_cgroup_do_uncharge(struct mem_cgroup *memcg,
4077 4078
				   unsigned int nr_pages,
				   const enum charge_type ctype)
4079 4080 4081
{
	struct memcg_batch_info *batch = NULL;
	bool uncharge_memsw = true;
4082

4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093
	/* If swapout, usage of swap doesn't decrease */
	if (!do_swap_account || ctype == MEM_CGROUP_CHARGE_TYPE_SWAPOUT)
		uncharge_memsw = false;

	batch = &current->memcg_batch;
	/*
	 * In usual, we do css_get() when we remember memcg pointer.
	 * But in this case, we keep res->usage until end of a series of
	 * uncharges. Then, it's ok to ignore memcg's refcnt.
	 */
	if (!batch->memcg)
4094
		batch->memcg = memcg;
4095 4096
	/*
	 * do_batch > 0 when unmapping pages or inode invalidate/truncate.
L
Lucas De Marchi 已提交
4097
	 * In those cases, all pages freed continuously can be expected to be in
4098 4099 4100 4101 4102 4103 4104 4105
	 * the same cgroup and we have chance to coalesce uncharges.
	 * But we do uncharge one by one if this is killed by OOM(TIF_MEMDIE)
	 * because we want to do uncharge as soon as possible.
	 */

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

4106
	if (nr_pages > 1)
A
Andrea Arcangeli 已提交
4107 4108
		goto direct_uncharge;

4109 4110 4111 4112 4113
	/*
	 * In typical case, batch->memcg == mem. This means we can
	 * merge a series of uncharges to an uncharge of res_counter.
	 * If not, we uncharge res_counter ony by one.
	 */
4114
	if (batch->memcg != memcg)
4115 4116
		goto direct_uncharge;
	/* remember freed charge and uncharge it later */
4117
	batch->nr_pages++;
4118
	if (uncharge_memsw)
4119
		batch->memsw_nr_pages++;
4120 4121
	return;
direct_uncharge:
4122
	res_counter_uncharge(&memcg->res, nr_pages * PAGE_SIZE);
4123
	if (uncharge_memsw)
4124 4125 4126
		res_counter_uncharge(&memcg->memsw, nr_pages * PAGE_SIZE);
	if (unlikely(batch->memcg != memcg))
		memcg_oom_recover(memcg);
4127
}
4128

4129
/*
4130
 * uncharge if !page_mapped(page)
4131
 */
4132
static struct mem_cgroup *
4133 4134
__mem_cgroup_uncharge_common(struct page *page, enum charge_type ctype,
			     bool end_migration)
4135
{
4136
	struct mem_cgroup *memcg = NULL;
4137 4138
	unsigned int nr_pages = 1;
	struct page_cgroup *pc;
4139
	bool anon;
4140

4141
	if (mem_cgroup_disabled())
4142
		return NULL;
4143

A
Andrea Arcangeli 已提交
4144
	if (PageTransHuge(page)) {
4145
		nr_pages <<= compound_order(page);
A
Andrea Arcangeli 已提交
4146 4147
		VM_BUG_ON(!PageTransHuge(page));
	}
4148
	/*
4149
	 * Check if our page_cgroup is valid
4150
	 */
4151
	pc = lookup_page_cgroup(page);
4152
	if (unlikely(!PageCgroupUsed(pc)))
4153
		return NULL;
4154

4155
	lock_page_cgroup(pc);
K
KAMEZAWA Hiroyuki 已提交
4156

4157
	memcg = pc->mem_cgroup;
4158

K
KAMEZAWA Hiroyuki 已提交
4159 4160 4161
	if (!PageCgroupUsed(pc))
		goto unlock_out;

4162 4163
	anon = PageAnon(page);

K
KAMEZAWA Hiroyuki 已提交
4164
	switch (ctype) {
4165
	case MEM_CGROUP_CHARGE_TYPE_ANON:
4166 4167 4168 4169 4170
		/*
		 * Generally PageAnon tells if it's the anon statistics to be
		 * updated; but sometimes e.g. mem_cgroup_uncharge_page() is
		 * used before page reached the stage of being marked PageAnon.
		 */
4171 4172
		anon = true;
		/* fallthrough */
K
KAMEZAWA Hiroyuki 已提交
4173
	case MEM_CGROUP_CHARGE_TYPE_DROP:
4174
		/* See mem_cgroup_prepare_migration() */
4175 4176 4177 4178 4179 4180 4181 4182 4183 4184
		if (page_mapped(page))
			goto unlock_out;
		/*
		 * Pages under migration may not be uncharged.  But
		 * end_migration() /must/ be the one uncharging the
		 * unused post-migration page and so it has to call
		 * here with the migration bit still set.  See the
		 * res_counter handling below.
		 */
		if (!end_migration && PageCgroupMigration(pc))
K
KAMEZAWA Hiroyuki 已提交
4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195
			goto unlock_out;
		break;
	case MEM_CGROUP_CHARGE_TYPE_SWAPOUT:
		if (!PageAnon(page)) {	/* Shared memory */
			if (page->mapping && !page_is_file_cache(page))
				goto unlock_out;
		} else if (page_mapped(page)) /* Anon */
				goto unlock_out;
		break;
	default:
		break;
4196
	}
K
KAMEZAWA Hiroyuki 已提交
4197

4198
	mem_cgroup_charge_statistics(memcg, page, anon, -nr_pages);
K
KAMEZAWA Hiroyuki 已提交
4199

4200
	ClearPageCgroupUsed(pc);
4201 4202 4203 4204 4205 4206
	/*
	 * pc->mem_cgroup is not cleared here. It will be accessed when it's
	 * freed from LRU. This is safe because uncharged page is expected not
	 * to be reused (freed soon). Exception is SwapCache, it's handled by
	 * special functions.
	 */
4207

4208
	unlock_page_cgroup(pc);
K
KAMEZAWA Hiroyuki 已提交
4209
	/*
4210
	 * even after unlock, we have memcg->res.usage here and this memcg
L
Li Zefan 已提交
4211
	 * will never be freed, so it's safe to call css_get().
K
KAMEZAWA Hiroyuki 已提交
4212
	 */
4213
	memcg_check_events(memcg, page);
K
KAMEZAWA Hiroyuki 已提交
4214
	if (do_swap_account && ctype == MEM_CGROUP_CHARGE_TYPE_SWAPOUT) {
4215
		mem_cgroup_swap_statistics(memcg, true);
L
Li Zefan 已提交
4216
		css_get(&memcg->css);
K
KAMEZAWA Hiroyuki 已提交
4217
	}
4218 4219 4220 4221 4222 4223
	/*
	 * Migration does not charge the res_counter for the
	 * replacement page, so leave it alone when phasing out the
	 * page that is unused after the migration.
	 */
	if (!end_migration && !mem_cgroup_is_root(memcg))
4224
		mem_cgroup_do_uncharge(memcg, nr_pages, ctype);
4225

4226
	return memcg;
K
KAMEZAWA Hiroyuki 已提交
4227 4228 4229

unlock_out:
	unlock_page_cgroup(pc);
4230
	return NULL;
4231 4232
}

4233 4234
void mem_cgroup_uncharge_page(struct page *page)
{
4235 4236 4237
	/* early check. */
	if (page_mapped(page))
		return;
4238
	VM_BUG_ON(page->mapping && !PageAnon(page));
4239 4240 4241 4242 4243 4244 4245 4246 4247 4248 4249 4250
	/*
	 * If the page is in swap cache, uncharge should be deferred
	 * to the swap path, which also properly accounts swap usage
	 * and handles memcg lifetime.
	 *
	 * Note that this check is not stable and reclaim may add the
	 * page to swap cache at any time after this.  However, if the
	 * page is not in swap cache by the time page->mapcount hits
	 * 0, there won't be any page table references to the swap
	 * slot, and reclaim will free it and not actually write the
	 * page to disk.
	 */
4251 4252
	if (PageSwapCache(page))
		return;
4253
	__mem_cgroup_uncharge_common(page, MEM_CGROUP_CHARGE_TYPE_ANON, false);
4254 4255 4256 4257 4258
}

void mem_cgroup_uncharge_cache_page(struct page *page)
{
	VM_BUG_ON(page_mapped(page));
4259
	VM_BUG_ON(page->mapping);
4260
	__mem_cgroup_uncharge_common(page, MEM_CGROUP_CHARGE_TYPE_CACHE, false);
4261 4262
}

4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276
/*
 * Batch_start/batch_end is called in unmap_page_range/invlidate/trucate.
 * In that cases, pages are freed continuously and we can expect pages
 * are in the same memcg. All these calls itself limits the number of
 * pages freed at once, then uncharge_start/end() is called properly.
 * This may be called prural(2) times in a context,
 */

void mem_cgroup_uncharge_start(void)
{
	current->memcg_batch.do_batch++;
	/* We can do nest. */
	if (current->memcg_batch.do_batch == 1) {
		current->memcg_batch.memcg = NULL;
4277 4278
		current->memcg_batch.nr_pages = 0;
		current->memcg_batch.memsw_nr_pages = 0;
4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298
	}
}

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

	if (!batch->do_batch)
		return;

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

	if (!batch->memcg)
		return;
	/*
	 * This "batch->memcg" is valid without any css_get/put etc...
	 * bacause we hide charges behind us.
	 */
4299 4300 4301 4302 4303 4304
	if (batch->nr_pages)
		res_counter_uncharge(&batch->memcg->res,
				     batch->nr_pages * PAGE_SIZE);
	if (batch->memsw_nr_pages)
		res_counter_uncharge(&batch->memcg->memsw,
				     batch->memsw_nr_pages * PAGE_SIZE);
4305
	memcg_oom_recover(batch->memcg);
4306 4307 4308 4309
	/* forget this pointer (for sanity check) */
	batch->memcg = NULL;
}

4310
#ifdef CONFIG_SWAP
4311
/*
4312
 * called after __delete_from_swap_cache() and drop "page" account.
4313 4314
 * memcg information is recorded to swap_cgroup of "ent"
 */
K
KAMEZAWA Hiroyuki 已提交
4315 4316
void
mem_cgroup_uncharge_swapcache(struct page *page, swp_entry_t ent, bool swapout)
4317 4318
{
	struct mem_cgroup *memcg;
K
KAMEZAWA Hiroyuki 已提交
4319 4320 4321 4322 4323
	int ctype = MEM_CGROUP_CHARGE_TYPE_SWAPOUT;

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

4324
	memcg = __mem_cgroup_uncharge_common(page, ctype, false);
4325

K
KAMEZAWA Hiroyuki 已提交
4326 4327
	/*
	 * record memcg information,  if swapout && memcg != NULL,
L
Li Zefan 已提交
4328
	 * css_get() was called in uncharge().
K
KAMEZAWA Hiroyuki 已提交
4329 4330
	 */
	if (do_swap_account && swapout && memcg)
4331
		swap_cgroup_record(ent, css_id(&memcg->css));
4332
}
4333
#endif
4334

A
Andrew Morton 已提交
4335
#ifdef CONFIG_MEMCG_SWAP
4336 4337 4338 4339 4340
/*
 * called from swap_entry_free(). remove record in swap_cgroup and
 * uncharge "memsw" account.
 */
void mem_cgroup_uncharge_swap(swp_entry_t ent)
K
KAMEZAWA Hiroyuki 已提交
4341
{
4342
	struct mem_cgroup *memcg;
4343
	unsigned short id;
4344 4345 4346 4347

	if (!do_swap_account)
		return;

4348 4349 4350
	id = swap_cgroup_record(ent, 0);
	rcu_read_lock();
	memcg = mem_cgroup_lookup(id);
4351
	if (memcg) {
4352 4353 4354 4355
		/*
		 * We uncharge this because swap is freed.
		 * This memcg can be obsolete one. We avoid calling css_tryget
		 */
4356
		if (!mem_cgroup_is_root(memcg))
4357
			res_counter_uncharge(&memcg->memsw, PAGE_SIZE);
4358
		mem_cgroup_swap_statistics(memcg, false);
L
Li Zefan 已提交
4359
		css_put(&memcg->css);
4360
	}
4361
	rcu_read_unlock();
K
KAMEZAWA Hiroyuki 已提交
4362
}
4363 4364 4365 4366 4367 4368 4369 4370 4371 4372 4373 4374 4375 4376 4377 4378

/**
 * mem_cgroup_move_swap_account - move swap charge and swap_cgroup's record.
 * @entry: swap entry to be moved
 * @from:  mem_cgroup which the entry is moved from
 * @to:  mem_cgroup which the entry is moved to
 *
 * It succeeds only when the swap_cgroup's record for this entry is the same
 * as the mem_cgroup's id of @from.
 *
 * Returns 0 on success, -EINVAL on failure.
 *
 * The caller must have charged to @to, IOW, called res_counter_charge() about
 * both res and memsw, and called css_get().
 */
static int mem_cgroup_move_swap_account(swp_entry_t entry,
4379
				struct mem_cgroup *from, struct mem_cgroup *to)
4380 4381 4382 4383 4384 4385 4386 4387
{
	unsigned short old_id, new_id;

	old_id = css_id(&from->css);
	new_id = css_id(&to->css);

	if (swap_cgroup_cmpxchg(entry, old_id, new_id) == old_id) {
		mem_cgroup_swap_statistics(from, false);
4388
		mem_cgroup_swap_statistics(to, true);
4389
		/*
4390 4391 4392
		 * This function is only called from task migration context now.
		 * It postpones res_counter and refcount handling till the end
		 * of task migration(mem_cgroup_clear_mc()) for performance
L
Li Zefan 已提交
4393 4394 4395 4396 4397 4398
		 * 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().
4399
		 */
L
Li Zefan 已提交
4400
		css_get(&to->css);
4401 4402 4403 4404 4405 4406
		return 0;
	}
	return -EINVAL;
}
#else
static inline int mem_cgroup_move_swap_account(swp_entry_t entry,
4407
				struct mem_cgroup *from, struct mem_cgroup *to)
4408 4409 4410
{
	return -EINVAL;
}
4411
#endif
K
KAMEZAWA Hiroyuki 已提交
4412

4413
/*
4414 4415
 * Before starting migration, account PAGE_SIZE to mem_cgroup that the old
 * page belongs to.
4416
 */
4417 4418
void mem_cgroup_prepare_migration(struct page *page, struct page *newpage,
				  struct mem_cgroup **memcgp)
4419
{
4420
	struct mem_cgroup *memcg = NULL;
4421
	unsigned int nr_pages = 1;
4422
	struct page_cgroup *pc;
4423
	enum charge_type ctype;
4424

4425
	*memcgp = NULL;
4426

4427
	if (mem_cgroup_disabled())
4428
		return;
4429

4430 4431 4432
	if (PageTransHuge(page))
		nr_pages <<= compound_order(page);

4433 4434 4435
	pc = lookup_page_cgroup(page);
	lock_page_cgroup(pc);
	if (PageCgroupUsed(pc)) {
4436 4437
		memcg = pc->mem_cgroup;
		css_get(&memcg->css);
4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468
		/*
		 * At migrating an anonymous page, its mapcount goes down
		 * to 0 and uncharge() will be called. But, even if it's fully
		 * unmapped, migration may fail and this page has to be
		 * charged again. We set MIGRATION flag here and delay uncharge
		 * until end_migration() is called
		 *
		 * Corner Case Thinking
		 * A)
		 * When the old page was mapped as Anon and it's unmap-and-freed
		 * while migration was ongoing.
		 * If unmap finds the old page, uncharge() of it will be delayed
		 * until end_migration(). If unmap finds a new page, it's
		 * uncharged when it make mapcount to be 1->0. If unmap code
		 * finds swap_migration_entry, the new page will not be mapped
		 * and end_migration() will find it(mapcount==0).
		 *
		 * B)
		 * When the old page was mapped but migraion fails, the kernel
		 * remaps it. A charge for it is kept by MIGRATION flag even
		 * if mapcount goes down to 0. We can do remap successfully
		 * without charging it again.
		 *
		 * C)
		 * The "old" page is under lock_page() until the end of
		 * migration, so, the old page itself will not be swapped-out.
		 * If the new page is swapped out before end_migraton, our
		 * hook to usual swap-out path will catch the event.
		 */
		if (PageAnon(page))
			SetPageCgroupMigration(pc);
4469
	}
4470
	unlock_page_cgroup(pc);
4471 4472 4473 4474
	/*
	 * If the page is not charged at this point,
	 * we return here.
	 */
4475
	if (!memcg)
4476
		return;
4477

4478
	*memcgp = memcg;
4479 4480 4481 4482 4483 4484 4485
	/*
	 * We charge new page before it's used/mapped. So, even if unlock_page()
	 * is called before end_migration, we can catch all events on this new
	 * page. In the case new page is migrated but not remapped, new page's
	 * mapcount will be finally 0 and we call uncharge in end_migration().
	 */
	if (PageAnon(page))
4486
		ctype = MEM_CGROUP_CHARGE_TYPE_ANON;
4487
	else
4488
		ctype = MEM_CGROUP_CHARGE_TYPE_CACHE;
4489 4490 4491 4492 4493
	/*
	 * The page is committed to the memcg, but it's not actually
	 * charged to the res_counter since we plan on replacing the
	 * old one and only one page is going to be left afterwards.
	 */
4494
	__mem_cgroup_commit_charge(memcg, newpage, nr_pages, ctype, false);
4495
}
4496

4497
/* remove redundant charge if migration failed*/
4498
void mem_cgroup_end_migration(struct mem_cgroup *memcg,
4499
	struct page *oldpage, struct page *newpage, bool migration_ok)
4500
{
4501
	struct page *used, *unused;
4502
	struct page_cgroup *pc;
4503
	bool anon;
4504

4505
	if (!memcg)
4506
		return;
4507

4508
	if (!migration_ok) {
4509 4510
		used = oldpage;
		unused = newpage;
4511
	} else {
4512
		used = newpage;
4513 4514
		unused = oldpage;
	}
4515
	anon = PageAnon(used);
4516 4517 4518 4519
	__mem_cgroup_uncharge_common(unused,
				     anon ? MEM_CGROUP_CHARGE_TYPE_ANON
				     : MEM_CGROUP_CHARGE_TYPE_CACHE,
				     true);
4520
	css_put(&memcg->css);
4521
	/*
4522 4523 4524
	 * We disallowed uncharge of pages under migration because mapcount
	 * of the page goes down to zero, temporarly.
	 * Clear the flag and check the page should be charged.
4525
	 */
4526 4527 4528 4529 4530
	pc = lookup_page_cgroup(oldpage);
	lock_page_cgroup(pc);
	ClearPageCgroupMigration(pc);
	unlock_page_cgroup(pc);

4531
	/*
4532 4533 4534 4535 4536 4537
	 * If a page is a file cache, radix-tree replacement is very atomic
	 * and we can skip this check. When it was an Anon page, its mapcount
	 * goes down to 0. But because we added MIGRATION flage, it's not
	 * uncharged yet. There are several case but page->mapcount check
	 * and USED bit check in mem_cgroup_uncharge_page() will do enough
	 * check. (see prepare_charge() also)
4538
	 */
4539
	if (anon)
4540
		mem_cgroup_uncharge_page(used);
4541
}
4542

4543 4544 4545 4546 4547 4548 4549 4550
/*
 * At replace page cache, newpage is not under any memcg but it's on
 * LRU. So, this function doesn't touch res_counter but handles LRU
 * in correct way. Both pages are locked so we cannot race with uncharge.
 */
void mem_cgroup_replace_page_cache(struct page *oldpage,
				  struct page *newpage)
{
4551
	struct mem_cgroup *memcg = NULL;
4552 4553 4554 4555 4556 4557 4558 4559 4560
	struct page_cgroup *pc;
	enum charge_type type = MEM_CGROUP_CHARGE_TYPE_CACHE;

	if (mem_cgroup_disabled())
		return;

	pc = lookup_page_cgroup(oldpage);
	/* fix accounting on old pages */
	lock_page_cgroup(pc);
4561 4562
	if (PageCgroupUsed(pc)) {
		memcg = pc->mem_cgroup;
4563
		mem_cgroup_charge_statistics(memcg, oldpage, false, -1);
4564 4565
		ClearPageCgroupUsed(pc);
	}
4566 4567
	unlock_page_cgroup(pc);

4568 4569 4570 4571 4572 4573
	/*
	 * When called from shmem_replace_page(), in some cases the
	 * oldpage has already been charged, and in some cases not.
	 */
	if (!memcg)
		return;
4574 4575 4576 4577 4578
	/*
	 * Even if newpage->mapping was NULL before starting replacement,
	 * the newpage may be on LRU(or pagevec for LRU) already. We lock
	 * LRU while we overwrite pc->mem_cgroup.
	 */
4579
	__mem_cgroup_commit_charge(memcg, newpage, 1, type, true);
4580 4581
}

4582 4583 4584 4585 4586 4587
#ifdef CONFIG_DEBUG_VM
static struct page_cgroup *lookup_page_cgroup_used(struct page *page)
{
	struct page_cgroup *pc;

	pc = lookup_page_cgroup(page);
4588 4589 4590 4591 4592
	/*
	 * Can be NULL while feeding pages into the page allocator for
	 * the first time, i.e. during boot or memory hotplug;
	 * or when mem_cgroup_disabled().
	 */
4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611
	if (likely(pc) && PageCgroupUsed(pc))
		return pc;
	return NULL;
}

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

	return lookup_page_cgroup_used(page) != NULL;
}

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

	pc = lookup_page_cgroup_used(page);
	if (pc) {
4612 4613
		pr_alert("pc:%p pc->flags:%lx pc->mem_cgroup:%p\n",
			 pc, pc->flags, pc->mem_cgroup);
4614 4615 4616 4617
	}
}
#endif

4618
static int mem_cgroup_resize_limit(struct mem_cgroup *memcg,
4619
				unsigned long long val)
4620
{
4621
	int retry_count;
4622
	u64 memswlimit, memlimit;
4623
	int ret = 0;
4624 4625
	int children = mem_cgroup_count_children(memcg);
	u64 curusage, oldusage;
4626
	int enlarge;
4627 4628 4629 4630 4631 4632 4633 4634 4635

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

	oldusage = res_counter_read_u64(&memcg->res, RES_USAGE);
4636

4637
	enlarge = 0;
4638
	while (retry_count) {
4639 4640 4641 4642
		if (signal_pending(current)) {
			ret = -EINTR;
			break;
		}
4643 4644 4645
		/*
		 * Rather than hide all in some function, I do this in
		 * open coded manner. You see what this really does.
4646
		 * We have to guarantee memcg->res.limit <= memcg->memsw.limit.
4647 4648 4649 4650 4651 4652
		 */
		mutex_lock(&set_limit_mutex);
		memswlimit = res_counter_read_u64(&memcg->memsw, RES_LIMIT);
		if (memswlimit < val) {
			ret = -EINVAL;
			mutex_unlock(&set_limit_mutex);
4653 4654
			break;
		}
4655 4656 4657 4658 4659

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

4660
		ret = res_counter_set_limit(&memcg->res, val);
4661 4662 4663 4664 4665 4666
		if (!ret) {
			if (memswlimit == val)
				memcg->memsw_is_minimum = true;
			else
				memcg->memsw_is_minimum = false;
		}
4667 4668 4669 4670 4671
		mutex_unlock(&set_limit_mutex);

		if (!ret)
			break;

4672 4673
		mem_cgroup_reclaim(memcg, GFP_KERNEL,
				   MEM_CGROUP_RECLAIM_SHRINK);
4674 4675 4676 4677 4678 4679
		curusage = res_counter_read_u64(&memcg->res, RES_USAGE);
		/* Usage is reduced ? */
  		if (curusage >= oldusage)
			retry_count--;
		else
			oldusage = curusage;
4680
	}
4681 4682
	if (!ret && enlarge)
		memcg_oom_recover(memcg);
4683

4684 4685 4686
	return ret;
}

L
Li Zefan 已提交
4687 4688
static int mem_cgroup_resize_memsw_limit(struct mem_cgroup *memcg,
					unsigned long long val)
4689
{
4690
	int retry_count;
4691
	u64 memlimit, memswlimit, oldusage, curusage;
4692 4693
	int children = mem_cgroup_count_children(memcg);
	int ret = -EBUSY;
4694
	int enlarge = 0;
4695

4696 4697 4698
	/* see mem_cgroup_resize_res_limit */
 	retry_count = children * MEM_CGROUP_RECLAIM_RETRIES;
	oldusage = res_counter_read_u64(&memcg->memsw, RES_USAGE);
4699 4700 4701 4702 4703 4704 4705 4706
	while (retry_count) {
		if (signal_pending(current)) {
			ret = -EINTR;
			break;
		}
		/*
		 * Rather than hide all in some function, I do this in
		 * open coded manner. You see what this really does.
4707
		 * We have to guarantee memcg->res.limit <= memcg->memsw.limit.
4708 4709 4710 4711 4712 4713 4714 4715
		 */
		mutex_lock(&set_limit_mutex);
		memlimit = res_counter_read_u64(&memcg->res, RES_LIMIT);
		if (memlimit > val) {
			ret = -EINVAL;
			mutex_unlock(&set_limit_mutex);
			break;
		}
4716 4717 4718
		memswlimit = res_counter_read_u64(&memcg->memsw, RES_LIMIT);
		if (memswlimit < val)
			enlarge = 1;
4719
		ret = res_counter_set_limit(&memcg->memsw, val);
4720 4721 4722 4723 4724 4725
		if (!ret) {
			if (memlimit == val)
				memcg->memsw_is_minimum = true;
			else
				memcg->memsw_is_minimum = false;
		}
4726 4727 4728 4729 4730
		mutex_unlock(&set_limit_mutex);

		if (!ret)
			break;

4731 4732 4733
		mem_cgroup_reclaim(memcg, GFP_KERNEL,
				   MEM_CGROUP_RECLAIM_NOSWAP |
				   MEM_CGROUP_RECLAIM_SHRINK);
4734
		curusage = res_counter_read_u64(&memcg->memsw, RES_USAGE);
4735
		/* Usage is reduced ? */
4736
		if (curusage >= oldusage)
4737
			retry_count--;
4738 4739
		else
			oldusage = curusage;
4740
	}
4741 4742
	if (!ret && enlarge)
		memcg_oom_recover(memcg);
4743 4744 4745
	return ret;
}

4746
unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
4747 4748
					    gfp_t gfp_mask,
					    unsigned long *total_scanned)
4749 4750 4751 4752 4753 4754
{
	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;
4755
	unsigned long long excess;
4756
	unsigned long nr_scanned;
4757 4758 4759 4760

	if (order > 0)
		return 0;

4761
	mctz = soft_limit_tree_node_zone(zone_to_nid(zone), zone_idx(zone));
4762 4763 4764 4765 4766 4767 4768 4769 4770 4771 4772 4773 4774
	/*
	 * 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;

4775
		nr_scanned = 0;
4776
		reclaimed = mem_cgroup_soft_reclaim(mz->memcg, zone,
4777
						    gfp_mask, &nr_scanned);
4778
		nr_reclaimed += reclaimed;
4779
		*total_scanned += nr_scanned;
4780 4781 4782 4783 4784 4785 4786 4787 4788 4789 4790 4791 4792 4793 4794 4795 4796 4797 4798 4799 4800 4801
		spin_lock(&mctz->lock);

		/*
		 * If we failed to reclaim anything from this memory cgroup
		 * it is time to move on to the next cgroup
		 */
		next_mz = NULL;
		if (!reclaimed) {
			do {
				/*
				 * Loop until we find yet another one.
				 *
				 * By the time we get the soft_limit lock
				 * again, someone might have aded the
				 * group back on the RB tree. Iterate to
				 * make sure we get a different mem.
				 * mem_cgroup_largest_soft_limit_node returns
				 * NULL if no other cgroup is present on
				 * the tree
				 */
				next_mz =
				__mem_cgroup_largest_soft_limit_node(mctz);
4802
				if (next_mz == mz)
4803
					css_put(&next_mz->memcg->css);
4804
				else /* next_mz == NULL or other memcg */
4805 4806 4807
					break;
			} while (1);
		}
4808 4809
		__mem_cgroup_remove_exceeded(mz->memcg, mz, mctz);
		excess = res_counter_soft_limit_excess(&mz->memcg->res);
4810 4811 4812 4813 4814 4815 4816 4817
		/*
		 * 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.
		 */
4818
		/* If excess == 0, no tree ops */
4819
		__mem_cgroup_insert_exceeded(mz->memcg, mz, mctz, excess);
4820
		spin_unlock(&mctz->lock);
4821
		css_put(&mz->memcg->css);
4822 4823 4824 4825 4826 4827 4828 4829 4830 4831 4832 4833
		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)
4834
		css_put(&next_mz->memcg->css);
4835 4836 4837
	return nr_reclaimed;
}

4838 4839 4840 4841 4842 4843 4844
/**
 * mem_cgroup_force_empty_list - clears LRU of a group
 * @memcg: group to clear
 * @node: NUMA node
 * @zid: zone id
 * @lru: lru to to clear
 *
4845
 * Traverse a specified page_cgroup list and try to drop them all.  This doesn't
4846 4847
 * reclaim the pages page themselves - pages are moved to the parent (or root)
 * group.
4848
 */
4849
static void mem_cgroup_force_empty_list(struct mem_cgroup *memcg,
K
KAMEZAWA Hiroyuki 已提交
4850
				int node, int zid, enum lru_list lru)
4851
{
4852
	struct lruvec *lruvec;
4853
	unsigned long flags;
4854
	struct list_head *list;
4855 4856
	struct page *busy;
	struct zone *zone;
4857

K
KAMEZAWA Hiroyuki 已提交
4858
	zone = &NODE_DATA(node)->node_zones[zid];
4859 4860
	lruvec = mem_cgroup_zone_lruvec(zone, memcg);
	list = &lruvec->lists[lru];
4861

4862
	busy = NULL;
4863
	do {
4864
		struct page_cgroup *pc;
4865 4866
		struct page *page;

K
KAMEZAWA Hiroyuki 已提交
4867
		spin_lock_irqsave(&zone->lru_lock, flags);
4868
		if (list_empty(list)) {
K
KAMEZAWA Hiroyuki 已提交
4869
			spin_unlock_irqrestore(&zone->lru_lock, flags);
4870
			break;
4871
		}
4872 4873 4874
		page = list_entry(list->prev, struct page, lru);
		if (busy == page) {
			list_move(&page->lru, list);
4875
			busy = NULL;
K
KAMEZAWA Hiroyuki 已提交
4876
			spin_unlock_irqrestore(&zone->lru_lock, flags);
4877 4878
			continue;
		}
K
KAMEZAWA Hiroyuki 已提交
4879
		spin_unlock_irqrestore(&zone->lru_lock, flags);
4880

4881
		pc = lookup_page_cgroup(page);
4882

4883
		if (mem_cgroup_move_parent(page, pc, memcg)) {
4884
			/* found lock contention or "pc" is obsolete. */
4885
			busy = page;
4886 4887 4888
			cond_resched();
		} else
			busy = NULL;
4889
	} while (!list_empty(list));
4890 4891 4892
}

/*
4893 4894
 * make mem_cgroup's charge to be 0 if there is no task by moving
 * all the charges and pages to the parent.
4895
 * This enables deleting this mem_cgroup.
4896 4897
 *
 * Caller is responsible for holding css reference on the memcg.
4898
 */
4899
static void mem_cgroup_reparent_charges(struct mem_cgroup *memcg)
4900
{
4901
	int node, zid;
4902
	u64 usage;
4903

4904
	do {
4905 4906
		/* This is for making all *used* pages to be on LRU. */
		lru_add_drain_all();
4907 4908
		drain_all_stock_sync(memcg);
		mem_cgroup_start_move(memcg);
4909
		for_each_node_state(node, N_MEMORY) {
4910
			for (zid = 0; zid < MAX_NR_ZONES; zid++) {
H
Hugh Dickins 已提交
4911 4912
				enum lru_list lru;
				for_each_lru(lru) {
4913
					mem_cgroup_force_empty_list(memcg,
H
Hugh Dickins 已提交
4914
							node, zid, lru);
4915
				}
4916
			}
4917
		}
4918 4919
		mem_cgroup_end_move(memcg);
		memcg_oom_recover(memcg);
4920
		cond_resched();
4921

4922
		/*
4923 4924 4925 4926 4927
		 * Kernel memory may not necessarily be trackable to a specific
		 * process. So they are not migrated, and therefore we can't
		 * expect their value to drop to 0 here.
		 * Having res filled up with kmem only is enough.
		 *
4928 4929 4930 4931 4932 4933
		 * This is a safety check because mem_cgroup_force_empty_list
		 * could have raced with mem_cgroup_replace_page_cache callers
		 * so the lru seemed empty but the page could have been added
		 * right after the check. RES_USAGE should be safe as we always
		 * charge before adding to the LRU.
		 */
4934 4935 4936
		usage = res_counter_read_u64(&memcg->res, RES_USAGE) -
			res_counter_read_u64(&memcg->kmem, RES_USAGE);
	} while (usage > 0);
4937 4938
}

4939 4940 4941 4942 4943 4944 4945 4946 4947 4948 4949 4950 4951 4952 4953 4954
/*
 * This mainly exists for tests during the setting of set of use_hierarchy.
 * Since this is the very setting we are changing, the current hierarchy value
 * is meaningless
 */
static inline bool __memcg_has_children(struct mem_cgroup *memcg)
{
	struct cgroup *pos;

	/* bounce at first found */
	cgroup_for_each_child(pos, memcg->css.cgroup)
		return true;
	return false;
}

/*
4955 4956
 * Must be called with memcg_create_mutex held, unless the cgroup is guaranteed
 * to be already dead (as in mem_cgroup_force_empty, for instance).  This is
4957 4958 4959 4960 4961 4962 4963 4964 4965
 * from mem_cgroup_count_children(), in the sense that we don't really care how
 * many children we have; we only need to know if we have any.  It also counts
 * any memcg without hierarchy as infertile.
 */
static inline bool memcg_has_children(struct mem_cgroup *memcg)
{
	return memcg->use_hierarchy && __memcg_has_children(memcg);
}

4966 4967 4968 4969 4970 4971 4972 4973 4974 4975
/*
 * Reclaims as many pages from the given memcg as possible and moves
 * the rest to the parent.
 *
 * Caller is responsible for holding css reference for memcg.
 */
static int mem_cgroup_force_empty(struct mem_cgroup *memcg)
{
	int nr_retries = MEM_CGROUP_RECLAIM_RETRIES;
	struct cgroup *cgrp = memcg->css.cgroup;
4976

4977
	/* returns EBUSY if there is a task or if we come here twice. */
4978 4979 4980
	if (cgroup_task_count(cgrp) || !list_empty(&cgrp->children))
		return -EBUSY;

4981 4982
	/* we call try-to-free pages for make this cgroup empty */
	lru_add_drain_all();
4983
	/* try to free all pages in this cgroup */
4984
	while (nr_retries && res_counter_read_u64(&memcg->res, RES_USAGE) > 0) {
4985
		int progress;
4986

4987 4988 4989
		if (signal_pending(current))
			return -EINTR;

4990
		progress = try_to_free_mem_cgroup_pages(memcg, GFP_KERNEL,
4991
						false);
4992
		if (!progress) {
4993
			nr_retries--;
4994
			/* maybe some writeback is necessary */
4995
			congestion_wait(BLK_RW_ASYNC, HZ/10);
4996
		}
4997 4998

	}
K
KAMEZAWA Hiroyuki 已提交
4999
	lru_add_drain();
5000 5001 5002
	mem_cgroup_reparent_charges(memcg);

	return 0;
5003 5004
}

5005
static int mem_cgroup_force_empty_write(struct cgroup *cont, unsigned int event)
5006
{
5007 5008 5009
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
	int ret;

5010 5011
	if (mem_cgroup_is_root(memcg))
		return -EINVAL;
5012 5013 5014 5015 5016
	css_get(&memcg->css);
	ret = mem_cgroup_force_empty(memcg);
	css_put(&memcg->css);

	return ret;
5017 5018 5019
}


5020 5021 5022 5023 5024 5025 5026 5027 5028
static u64 mem_cgroup_hierarchy_read(struct cgroup *cont, struct cftype *cft)
{
	return mem_cgroup_from_cont(cont)->use_hierarchy;
}

static int mem_cgroup_hierarchy_write(struct cgroup *cont, struct cftype *cft,
					u64 val)
{
	int retval = 0;
5029
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
5030
	struct cgroup *parent = cont->parent;
5031
	struct mem_cgroup *parent_memcg = NULL;
5032 5033

	if (parent)
5034
		parent_memcg = mem_cgroup_from_cont(parent);
5035

5036
	mutex_lock(&memcg_create_mutex);
5037 5038 5039 5040

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

5041
	/*
5042
	 * If parent's use_hierarchy is set, we can't make any modifications
5043 5044 5045 5046 5047 5048
	 * 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.
	 */
5049
	if ((!parent_memcg || !parent_memcg->use_hierarchy) &&
5050
				(val == 1 || val == 0)) {
5051
		if (!__memcg_has_children(memcg))
5052
			memcg->use_hierarchy = val;
5053 5054 5055 5056
		else
			retval = -EBUSY;
	} else
		retval = -EINVAL;
5057 5058

out:
5059
	mutex_unlock(&memcg_create_mutex);
5060 5061 5062 5063

	return retval;
}

5064

5065
static unsigned long mem_cgroup_recursive_stat(struct mem_cgroup *memcg,
5066
					       enum mem_cgroup_stat_index idx)
5067
{
K
KAMEZAWA Hiroyuki 已提交
5068
	struct mem_cgroup *iter;
5069
	long val = 0;
5070

5071
	/* Per-cpu values can be negative, use a signed accumulator */
5072
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
5073 5074 5075 5076 5077
		val += mem_cgroup_read_stat(iter, idx);

	if (val < 0) /* race ? */
		val = 0;
	return val;
5078 5079
}

5080
static inline u64 mem_cgroup_usage(struct mem_cgroup *memcg, bool swap)
5081
{
K
KAMEZAWA Hiroyuki 已提交
5082
	u64 val;
5083

5084
	if (!mem_cgroup_is_root(memcg)) {
5085
		if (!swap)
5086
			return res_counter_read_u64(&memcg->res, RES_USAGE);
5087
		else
5088
			return res_counter_read_u64(&memcg->memsw, RES_USAGE);
5089 5090
	}

5091 5092 5093 5094
	/*
	 * Transparent hugepages are still accounted for in MEM_CGROUP_STAT_RSS
	 * as well as in MEM_CGROUP_STAT_RSS_HUGE.
	 */
5095 5096
	val = mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_CACHE);
	val += mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_RSS);
5097

K
KAMEZAWA Hiroyuki 已提交
5098
	if (swap)
5099
		val += mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_SWAP);
5100 5101 5102 5103

	return val << PAGE_SHIFT;
}

5104 5105 5106
static ssize_t mem_cgroup_read(struct cgroup *cont, struct cftype *cft,
			       struct file *file, char __user *buf,
			       size_t nbytes, loff_t *ppos)
B
Balbir Singh 已提交
5107
{
5108
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
5109
	char str[64];
5110
	u64 val;
G
Glauber Costa 已提交
5111 5112
	int name, len;
	enum res_type type;
5113 5114 5115

	type = MEMFILE_TYPE(cft->private);
	name = MEMFILE_ATTR(cft->private);
5116

5117 5118
	switch (type) {
	case _MEM:
5119
		if (name == RES_USAGE)
5120
			val = mem_cgroup_usage(memcg, false);
5121
		else
5122
			val = res_counter_read_u64(&memcg->res, name);
5123 5124
		break;
	case _MEMSWAP:
5125
		if (name == RES_USAGE)
5126
			val = mem_cgroup_usage(memcg, true);
5127
		else
5128
			val = res_counter_read_u64(&memcg->memsw, name);
5129
		break;
5130 5131 5132
	case _KMEM:
		val = res_counter_read_u64(&memcg->kmem, name);
		break;
5133 5134 5135
	default:
		BUG();
	}
5136 5137 5138

	len = scnprintf(str, sizeof(str), "%llu\n", (unsigned long long)val);
	return simple_read_from_buffer(buf, nbytes, ppos, str, len);
B
Balbir Singh 已提交
5139
}
5140 5141 5142 5143 5144 5145 5146 5147 5148 5149 5150 5151 5152 5153 5154 5155 5156 5157

static int memcg_update_kmem_limit(struct cgroup *cont, u64 val)
{
	int ret = -EINVAL;
#ifdef CONFIG_MEMCG_KMEM
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
	/*
	 * 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.
	 */
5158
	mutex_lock(&memcg_create_mutex);
5159 5160
	mutex_lock(&set_limit_mutex);
	if (!memcg->kmem_account_flags && val != RESOURCE_MAX) {
5161
		if (cgroup_task_count(cont) || memcg_has_children(memcg)) {
5162 5163 5164 5165 5166 5167
			ret = -EBUSY;
			goto out;
		}
		ret = res_counter_set_limit(&memcg->kmem, val);
		VM_BUG_ON(ret);

5168 5169 5170 5171 5172
		ret = memcg_update_cache_sizes(memcg);
		if (ret) {
			res_counter_set_limit(&memcg->kmem, RESOURCE_MAX);
			goto out;
		}
5173 5174 5175 5176 5177 5178
		static_key_slow_inc(&memcg_kmem_enabled_key);
		/*
		 * setting the active bit after the inc will guarantee no one
		 * starts accounting before all call sites are patched
		 */
		memcg_kmem_set_active(memcg);
5179 5180 5181 5182
	} else
		ret = res_counter_set_limit(&memcg->kmem, val);
out:
	mutex_unlock(&set_limit_mutex);
5183
	mutex_unlock(&memcg_create_mutex);
5184 5185 5186 5187
#endif
	return ret;
}

5188
#ifdef CONFIG_MEMCG_KMEM
5189
static int memcg_propagate_kmem(struct mem_cgroup *memcg)
5190
{
5191
	int ret = 0;
5192 5193
	struct mem_cgroup *parent = parent_mem_cgroup(memcg);
	if (!parent)
5194 5195
		goto out;

5196
	memcg->kmem_account_flags = parent->kmem_account_flags;
5197 5198 5199 5200 5201 5202 5203 5204 5205 5206
	/*
	 * When that happen, we need to disable the static branch only on those
	 * memcgs that enabled it. To achieve this, we would be forced to
	 * complicate the code by keeping track of which memcgs were the ones
	 * that actually enabled limits, and which ones got it from its
	 * parents.
	 *
	 * It is a lot simpler just to do static_key_slow_inc() on every child
	 * that is accounted.
	 */
5207 5208 5209 5210
	if (!memcg_kmem_is_active(memcg))
		goto out;

	/*
5211 5212 5213
	 * __mem_cgroup_free() will issue static_key_slow_dec() because this
	 * memcg is active already. If the later initialization fails then the
	 * cgroup core triggers the cleanup so we do not have to do it here.
5214 5215 5216 5217
	 */
	static_key_slow_inc(&memcg_kmem_enabled_key);

	mutex_lock(&set_limit_mutex);
5218
	memcg_stop_kmem_account();
5219
	ret = memcg_update_cache_sizes(memcg);
5220
	memcg_resume_kmem_account();
5221 5222 5223
	mutex_unlock(&set_limit_mutex);
out:
	return ret;
5224
}
5225
#endif /* CONFIG_MEMCG_KMEM */
5226

5227 5228 5229 5230
/*
 * The user of this function is...
 * RES_LIMIT.
 */
5231 5232
static int mem_cgroup_write(struct cgroup *cont, struct cftype *cft,
			    const char *buffer)
B
Balbir Singh 已提交
5233
{
5234
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
G
Glauber Costa 已提交
5235 5236
	enum res_type type;
	int name;
5237 5238 5239
	unsigned long long val;
	int ret;

5240 5241
	type = MEMFILE_TYPE(cft->private);
	name = MEMFILE_ATTR(cft->private);
5242

5243
	switch (name) {
5244
	case RES_LIMIT:
5245 5246 5247 5248
		if (mem_cgroup_is_root(memcg)) { /* Can't set limit on root */
			ret = -EINVAL;
			break;
		}
5249 5250
		/* This function does all necessary parse...reuse it */
		ret = res_counter_memparse_write_strategy(buffer, &val);
5251 5252 5253
		if (ret)
			break;
		if (type == _MEM)
5254
			ret = mem_cgroup_resize_limit(memcg, val);
5255
		else if (type == _MEMSWAP)
5256
			ret = mem_cgroup_resize_memsw_limit(memcg, val);
5257 5258 5259 5260
		else if (type == _KMEM)
			ret = memcg_update_kmem_limit(cont, val);
		else
			return -EINVAL;
5261
		break;
5262 5263 5264 5265 5266 5267 5268 5269 5270 5271 5272 5273 5274 5275
	case RES_SOFT_LIMIT:
		ret = res_counter_memparse_write_strategy(buffer, &val);
		if (ret)
			break;
		/*
		 * For memsw, soft limits are hard to implement in terms
		 * of semantics, for now, we support soft limits for
		 * control without swap
		 */
		if (type == _MEM)
			ret = res_counter_set_soft_limit(&memcg->res, val);
		else
			ret = -EINVAL;
		break;
5276 5277 5278 5279 5280
	default:
		ret = -EINVAL; /* should be BUG() ? */
		break;
	}
	return ret;
B
Balbir Singh 已提交
5281 5282
}

5283 5284 5285 5286 5287 5288 5289 5290 5291 5292 5293 5294 5295 5296 5297 5298 5299 5300 5301 5302 5303 5304 5305 5306 5307 5308 5309
static void memcg_get_hierarchical_limit(struct mem_cgroup *memcg,
		unsigned long long *mem_limit, unsigned long long *memsw_limit)
{
	struct cgroup *cgroup;
	unsigned long long min_limit, min_memsw_limit, tmp;

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

	while (cgroup->parent) {
		cgroup = cgroup->parent;
		memcg = mem_cgroup_from_cont(cgroup);
		if (!memcg->use_hierarchy)
			break;
		tmp = res_counter_read_u64(&memcg->res, RES_LIMIT);
		min_limit = min(min_limit, tmp);
		tmp = res_counter_read_u64(&memcg->memsw, RES_LIMIT);
		min_memsw_limit = min(min_memsw_limit, tmp);
	}
out:
	*mem_limit = min_limit;
	*memsw_limit = min_memsw_limit;
}

5310
static int mem_cgroup_reset(struct cgroup *cont, unsigned int event)
5311
{
5312
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
G
Glauber Costa 已提交
5313 5314
	int name;
	enum res_type type;
5315

5316 5317
	type = MEMFILE_TYPE(event);
	name = MEMFILE_ATTR(event);
5318

5319
	switch (name) {
5320
	case RES_MAX_USAGE:
5321
		if (type == _MEM)
5322
			res_counter_reset_max(&memcg->res);
5323
		else if (type == _MEMSWAP)
5324
			res_counter_reset_max(&memcg->memsw);
5325 5326 5327 5328
		else if (type == _KMEM)
			res_counter_reset_max(&memcg->kmem);
		else
			return -EINVAL;
5329 5330
		break;
	case RES_FAILCNT:
5331
		if (type == _MEM)
5332
			res_counter_reset_failcnt(&memcg->res);
5333
		else if (type == _MEMSWAP)
5334
			res_counter_reset_failcnt(&memcg->memsw);
5335 5336 5337 5338
		else if (type == _KMEM)
			res_counter_reset_failcnt(&memcg->kmem);
		else
			return -EINVAL;
5339 5340
		break;
	}
5341

5342
	return 0;
5343 5344
}

5345 5346 5347 5348 5349 5350
static u64 mem_cgroup_move_charge_read(struct cgroup *cgrp,
					struct cftype *cft)
{
	return mem_cgroup_from_cont(cgrp)->move_charge_at_immigrate;
}

5351
#ifdef CONFIG_MMU
5352 5353 5354
static int mem_cgroup_move_charge_write(struct cgroup *cgrp,
					struct cftype *cft, u64 val)
{
5355
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
5356 5357 5358

	if (val >= (1 << NR_MOVE_TYPE))
		return -EINVAL;
5359

5360
	/*
5361 5362 5363 5364
	 * 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.
5365
	 */
5366
	memcg->move_charge_at_immigrate = val;
5367 5368
	return 0;
}
5369 5370 5371 5372 5373 5374 5375
#else
static int mem_cgroup_move_charge_write(struct cgroup *cgrp,
					struct cftype *cft, u64 val)
{
	return -ENOSYS;
}
#endif
5376

5377
#ifdef CONFIG_NUMA
5378
static int memcg_numa_stat_show(struct cgroup *cont, struct cftype *cft,
5379
				      struct seq_file *m)
5380 5381 5382 5383
{
	int nid;
	unsigned long total_nr, file_nr, anon_nr, unevictable_nr;
	unsigned long node_nr;
5384
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
5385

5386
	total_nr = mem_cgroup_nr_lru_pages(memcg, LRU_ALL);
5387
	seq_printf(m, "total=%lu", total_nr);
5388
	for_each_node_state(nid, N_MEMORY) {
5389
		node_nr = mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL);
5390 5391 5392 5393
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');

5394
	file_nr = mem_cgroup_nr_lru_pages(memcg, LRU_ALL_FILE);
5395
	seq_printf(m, "file=%lu", file_nr);
5396
	for_each_node_state(nid, N_MEMORY) {
5397
		node_nr = mem_cgroup_node_nr_lru_pages(memcg, nid,
5398
				LRU_ALL_FILE);
5399 5400 5401 5402
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');

5403
	anon_nr = mem_cgroup_nr_lru_pages(memcg, LRU_ALL_ANON);
5404
	seq_printf(m, "anon=%lu", anon_nr);
5405
	for_each_node_state(nid, N_MEMORY) {
5406
		node_nr = mem_cgroup_node_nr_lru_pages(memcg, nid,
5407
				LRU_ALL_ANON);
5408 5409 5410 5411
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');

5412
	unevictable_nr = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_UNEVICTABLE));
5413
	seq_printf(m, "unevictable=%lu", unevictable_nr);
5414
	for_each_node_state(nid, N_MEMORY) {
5415
		node_nr = mem_cgroup_node_nr_lru_pages(memcg, nid,
5416
				BIT(LRU_UNEVICTABLE));
5417 5418 5419 5420 5421 5422 5423
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');
	return 0;
}
#endif /* CONFIG_NUMA */

5424 5425 5426 5427 5428
static inline void mem_cgroup_lru_names_not_uptodate(void)
{
	BUILD_BUG_ON(ARRAY_SIZE(mem_cgroup_lru_names) != NR_LRU_LISTS);
}

5429
static int memcg_stat_show(struct cgroup *cont, struct cftype *cft,
5430
				 struct seq_file *m)
5431
{
5432
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
5433 5434
	struct mem_cgroup *mi;
	unsigned int i;
5435

5436
	for (i = 0; i < MEM_CGROUP_STAT_NSTATS; i++) {
5437
		if (i == MEM_CGROUP_STAT_SWAP && !do_swap_account)
5438
			continue;
5439 5440
		seq_printf(m, "%s %ld\n", mem_cgroup_stat_names[i],
			   mem_cgroup_read_stat(memcg, i) * PAGE_SIZE);
5441
	}
L
Lee Schermerhorn 已提交
5442

5443 5444 5445 5446 5447 5448 5449 5450
	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
5451
	/* Hierarchical information */
5452 5453
	{
		unsigned long long limit, memsw_limit;
5454
		memcg_get_hierarchical_limit(memcg, &limit, &memsw_limit);
5455
		seq_printf(m, "hierarchical_memory_limit %llu\n", limit);
5456
		if (do_swap_account)
5457 5458
			seq_printf(m, "hierarchical_memsw_limit %llu\n",
				   memsw_limit);
5459
	}
K
KOSAKI Motohiro 已提交
5460

5461 5462 5463
	for (i = 0; i < MEM_CGROUP_STAT_NSTATS; i++) {
		long long val = 0;

5464
		if (i == MEM_CGROUP_STAT_SWAP && !do_swap_account)
5465
			continue;
5466 5467 5468 5469 5470 5471 5472 5473 5474 5475 5476 5477 5478 5479 5480 5481 5482 5483 5484 5485
		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);
5486
	}
K
KAMEZAWA Hiroyuki 已提交
5487

K
KOSAKI Motohiro 已提交
5488 5489 5490 5491
#ifdef CONFIG_DEBUG_VM
	{
		int nid, zid;
		struct mem_cgroup_per_zone *mz;
5492
		struct zone_reclaim_stat *rstat;
K
KOSAKI Motohiro 已提交
5493 5494 5495 5496 5497
		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++) {
5498
				mz = mem_cgroup_zoneinfo(memcg, nid, zid);
5499
				rstat = &mz->lruvec.reclaim_stat;
K
KOSAKI Motohiro 已提交
5500

5501 5502 5503 5504
				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
5505
			}
5506 5507 5508 5509
		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
5510 5511 5512
	}
#endif

5513 5514 5515
	return 0;
}

K
KOSAKI Motohiro 已提交
5516 5517 5518 5519
static u64 mem_cgroup_swappiness_read(struct cgroup *cgrp, struct cftype *cft)
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);

5520
	return mem_cgroup_swappiness(memcg);
K
KOSAKI Motohiro 已提交
5521 5522 5523 5524 5525 5526 5527
}

static int mem_cgroup_swappiness_write(struct cgroup *cgrp, struct cftype *cft,
				       u64 val)
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
	struct mem_cgroup *parent;
5528

K
KOSAKI Motohiro 已提交
5529 5530 5531 5532 5533 5534 5535
	if (val > 100)
		return -EINVAL;

	if (cgrp->parent == NULL)
		return -EINVAL;

	parent = mem_cgroup_from_cont(cgrp->parent);
5536

5537
	mutex_lock(&memcg_create_mutex);
5538

K
KOSAKI Motohiro 已提交
5539
	/* If under hierarchy, only empty-root can set this value */
5540
	if ((parent->use_hierarchy) || memcg_has_children(memcg)) {
5541
		mutex_unlock(&memcg_create_mutex);
K
KOSAKI Motohiro 已提交
5542
		return -EINVAL;
5543
	}
K
KOSAKI Motohiro 已提交
5544 5545 5546

	memcg->swappiness = val;

5547
	mutex_unlock(&memcg_create_mutex);
5548

K
KOSAKI Motohiro 已提交
5549 5550 5551
	return 0;
}

5552 5553 5554 5555 5556 5557 5558 5559
static void __mem_cgroup_threshold(struct mem_cgroup *memcg, bool swap)
{
	struct mem_cgroup_threshold_ary *t;
	u64 usage;
	int i;

	rcu_read_lock();
	if (!swap)
5560
		t = rcu_dereference(memcg->thresholds.primary);
5561
	else
5562
		t = rcu_dereference(memcg->memsw_thresholds.primary);
5563 5564 5565 5566 5567 5568 5569

	if (!t)
		goto unlock;

	usage = mem_cgroup_usage(memcg, swap);

	/*
5570
	 * current_threshold points to threshold just below or equal to usage.
5571 5572 5573
	 * If it's not true, a threshold was crossed after last
	 * call of __mem_cgroup_threshold().
	 */
5574
	i = t->current_threshold;
5575 5576 5577 5578 5579 5580 5581 5582 5583 5584 5585 5586 5587 5588 5589 5590 5591 5592 5593 5594 5595 5596 5597

	/*
	 * 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 */
5598
	t->current_threshold = i - 1;
5599 5600 5601 5602 5603 5604
unlock:
	rcu_read_unlock();
}

static void mem_cgroup_threshold(struct mem_cgroup *memcg)
{
5605 5606 5607 5608 5609 5610 5611
	while (memcg) {
		__mem_cgroup_threshold(memcg, false);
		if (do_swap_account)
			__mem_cgroup_threshold(memcg, true);

		memcg = parent_mem_cgroup(memcg);
	}
5612 5613 5614 5615 5616 5617 5618 5619 5620 5621
}

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

	return _a->threshold - _b->threshold;
}

5622
static int mem_cgroup_oom_notify_cb(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
5623 5624 5625
{
	struct mem_cgroup_eventfd_list *ev;

5626
	list_for_each_entry(ev, &memcg->oom_notify, list)
K
KAMEZAWA Hiroyuki 已提交
5627 5628 5629 5630
		eventfd_signal(ev->eventfd, 1);
	return 0;
}

5631
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
5632
{
K
KAMEZAWA Hiroyuki 已提交
5633 5634
	struct mem_cgroup *iter;

5635
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
5636
		mem_cgroup_oom_notify_cb(iter);
K
KAMEZAWA Hiroyuki 已提交
5637 5638 5639 5640
}

static int mem_cgroup_usage_register_event(struct cgroup *cgrp,
	struct cftype *cft, struct eventfd_ctx *eventfd, const char *args)
5641 5642
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
5643 5644
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
G
Glauber Costa 已提交
5645
	enum res_type type = MEMFILE_TYPE(cft->private);
5646
	u64 threshold, usage;
5647
	int i, size, ret;
5648 5649 5650 5651 5652 5653

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

	mutex_lock(&memcg->thresholds_lock);
5654

5655
	if (type == _MEM)
5656
		thresholds = &memcg->thresholds;
5657
	else if (type == _MEMSWAP)
5658
		thresholds = &memcg->memsw_thresholds;
5659 5660 5661 5662 5663 5664
	else
		BUG();

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

	/* Check if a threshold crossed before adding a new one */
5665
	if (thresholds->primary)
5666 5667
		__mem_cgroup_threshold(memcg, type == _MEMSWAP);

5668
	size = thresholds->primary ? thresholds->primary->size + 1 : 1;
5669 5670

	/* Allocate memory for new array of thresholds */
5671
	new = kmalloc(sizeof(*new) + size * sizeof(struct mem_cgroup_threshold),
5672
			GFP_KERNEL);
5673
	if (!new) {
5674 5675 5676
		ret = -ENOMEM;
		goto unlock;
	}
5677
	new->size = size;
5678 5679

	/* Copy thresholds (if any) to new array */
5680 5681
	if (thresholds->primary) {
		memcpy(new->entries, thresholds->primary->entries, (size - 1) *
5682
				sizeof(struct mem_cgroup_threshold));
5683 5684
	}

5685
	/* Add new threshold */
5686 5687
	new->entries[size - 1].eventfd = eventfd;
	new->entries[size - 1].threshold = threshold;
5688 5689

	/* Sort thresholds. Registering of new threshold isn't time-critical */
5690
	sort(new->entries, size, sizeof(struct mem_cgroup_threshold),
5691 5692 5693
			compare_thresholds, NULL);

	/* Find current threshold */
5694
	new->current_threshold = -1;
5695
	for (i = 0; i < size; i++) {
5696
		if (new->entries[i].threshold <= usage) {
5697
			/*
5698 5699
			 * new->current_threshold will not be used until
			 * rcu_assign_pointer(), so it's safe to increment
5700 5701
			 * it here.
			 */
5702
			++new->current_threshold;
5703 5704
		} else
			break;
5705 5706
	}

5707 5708 5709 5710 5711
	/* Free old spare buffer and save old primary buffer as spare */
	kfree(thresholds->spare);
	thresholds->spare = thresholds->primary;

	rcu_assign_pointer(thresholds->primary, new);
5712

5713
	/* To be sure that nobody uses thresholds */
5714 5715 5716 5717 5718 5719 5720 5721
	synchronize_rcu();

unlock:
	mutex_unlock(&memcg->thresholds_lock);

	return ret;
}

5722
static void mem_cgroup_usage_unregister_event(struct cgroup *cgrp,
K
KAMEZAWA Hiroyuki 已提交
5723
	struct cftype *cft, struct eventfd_ctx *eventfd)
5724 5725
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
5726 5727
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
G
Glauber Costa 已提交
5728
	enum res_type type = MEMFILE_TYPE(cft->private);
5729
	u64 usage;
5730
	int i, j, size;
5731 5732 5733

	mutex_lock(&memcg->thresholds_lock);
	if (type == _MEM)
5734
		thresholds = &memcg->thresholds;
5735
	else if (type == _MEMSWAP)
5736
		thresholds = &memcg->memsw_thresholds;
5737 5738 5739
	else
		BUG();

5740 5741 5742
	if (!thresholds->primary)
		goto unlock;

5743 5744 5745 5746 5747 5748
	usage = mem_cgroup_usage(memcg, type == _MEMSWAP);

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

	/* Calculate new number of threshold */
5749 5750 5751
	size = 0;
	for (i = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd != eventfd)
5752 5753 5754
			size++;
	}

5755
	new = thresholds->spare;
5756

5757 5758
	/* Set thresholds array to NULL if we don't have thresholds */
	if (!size) {
5759 5760
		kfree(new);
		new = NULL;
5761
		goto swap_buffers;
5762 5763
	}

5764
	new->size = size;
5765 5766

	/* Copy thresholds and find current threshold */
5767 5768 5769
	new->current_threshold = -1;
	for (i = 0, j = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd == eventfd)
5770 5771
			continue;

5772
		new->entries[j] = thresholds->primary->entries[i];
5773
		if (new->entries[j].threshold <= usage) {
5774
			/*
5775
			 * new->current_threshold will not be used
5776 5777 5778
			 * until rcu_assign_pointer(), so it's safe to increment
			 * it here.
			 */
5779
			++new->current_threshold;
5780 5781 5782 5783
		}
		j++;
	}

5784
swap_buffers:
5785 5786
	/* Swap primary and spare array */
	thresholds->spare = thresholds->primary;
5787 5788 5789 5790 5791 5792
	/* If all events are unregistered, free the spare array */
	if (!new) {
		kfree(thresholds->spare);
		thresholds->spare = NULL;
	}

5793
	rcu_assign_pointer(thresholds->primary, new);
5794

5795
	/* To be sure that nobody uses thresholds */
5796
	synchronize_rcu();
5797
unlock:
5798 5799
	mutex_unlock(&memcg->thresholds_lock);
}
5800

K
KAMEZAWA Hiroyuki 已提交
5801 5802 5803 5804 5805
static int mem_cgroup_oom_register_event(struct cgroup *cgrp,
	struct cftype *cft, struct eventfd_ctx *eventfd, const char *args)
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
	struct mem_cgroup_eventfd_list *event;
G
Glauber Costa 已提交
5806
	enum res_type type = MEMFILE_TYPE(cft->private);
K
KAMEZAWA Hiroyuki 已提交
5807 5808 5809 5810 5811 5812

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

5813
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
5814 5815 5816 5817 5818

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

	/* already in OOM ? */
5819
	if (atomic_read(&memcg->under_oom))
K
KAMEZAWA Hiroyuki 已提交
5820
		eventfd_signal(eventfd, 1);
5821
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
5822 5823 5824 5825

	return 0;
}

5826
static void mem_cgroup_oom_unregister_event(struct cgroup *cgrp,
K
KAMEZAWA Hiroyuki 已提交
5827 5828
	struct cftype *cft, struct eventfd_ctx *eventfd)
{
5829
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
K
KAMEZAWA Hiroyuki 已提交
5830
	struct mem_cgroup_eventfd_list *ev, *tmp;
G
Glauber Costa 已提交
5831
	enum res_type type = MEMFILE_TYPE(cft->private);
K
KAMEZAWA Hiroyuki 已提交
5832 5833 5834

	BUG_ON(type != _OOM_TYPE);

5835
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
5836

5837
	list_for_each_entry_safe(ev, tmp, &memcg->oom_notify, list) {
K
KAMEZAWA Hiroyuki 已提交
5838 5839 5840 5841 5842 5843
		if (ev->eventfd == eventfd) {
			list_del(&ev->list);
			kfree(ev);
		}
	}

5844
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
5845 5846
}

5847 5848 5849
static int mem_cgroup_oom_control_read(struct cgroup *cgrp,
	struct cftype *cft,  struct cgroup_map_cb *cb)
{
5850
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
5851

5852
	cb->fill(cb, "oom_kill_disable", memcg->oom_kill_disable);
5853

5854
	if (atomic_read(&memcg->under_oom))
5855 5856 5857 5858 5859 5860 5861 5862 5863
		cb->fill(cb, "under_oom", 1);
	else
		cb->fill(cb, "under_oom", 0);
	return 0;
}

static int mem_cgroup_oom_control_write(struct cgroup *cgrp,
	struct cftype *cft, u64 val)
{
5864
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
5865 5866 5867 5868 5869 5870 5871 5872
	struct mem_cgroup *parent;

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

	parent = mem_cgroup_from_cont(cgrp->parent);

5873
	mutex_lock(&memcg_create_mutex);
5874
	/* oom-kill-disable is a flag for subhierarchy. */
5875
	if ((parent->use_hierarchy) || memcg_has_children(memcg)) {
5876
		mutex_unlock(&memcg_create_mutex);
5877 5878
		return -EINVAL;
	}
5879
	memcg->oom_kill_disable = val;
5880
	if (!val)
5881
		memcg_oom_recover(memcg);
5882
	mutex_unlock(&memcg_create_mutex);
5883 5884 5885
	return 0;
}

A
Andrew Morton 已提交
5886
#ifdef CONFIG_MEMCG_KMEM
5887
static int memcg_init_kmem(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
5888
{
5889 5890
	int ret;

5891
	memcg->kmemcg_id = -1;
5892 5893 5894
	ret = memcg_propagate_kmem(memcg);
	if (ret)
		return ret;
5895

5896
	return mem_cgroup_sockets_init(memcg, ss);
5897
}
5898

5899
static void memcg_destroy_kmem(struct mem_cgroup *memcg)
G
Glauber Costa 已提交
5900
{
5901
	mem_cgroup_sockets_destroy(memcg);
5902 5903 5904 5905 5906 5907 5908 5909 5910 5911 5912 5913 5914 5915 5916 5917 5918 5919 5920 5921 5922 5923 5924 5925 5926 5927
}

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

	/*
	 * kmem charges can outlive the cgroup. In the case of slab
	 * pages, for instance, a page contain objects from various
	 * processes. As we prevent from taking a reference for every
	 * such allocation we have to be careful when doing uncharge
	 * (see memcg_uncharge_kmem) and here during offlining.
	 *
	 * The idea is that that only the _last_ uncharge which sees
	 * the dead memcg will drop the last reference. An additional
	 * reference is taken here before the group is marked dead
	 * which is then paired with css_put during uncharge resp. here.
	 *
	 * Although this might sound strange as this path is called from
	 * css_offline() when the referencemight have dropped down to 0
	 * and shouldn't be incremented anymore (css_tryget would fail)
	 * we do not have other options because of the kmem allocations
	 * lifetime.
	 */
	css_get(&memcg->css);
5928 5929 5930 5931 5932 5933 5934

	memcg_kmem_mark_dead(memcg);

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

	if (memcg_kmem_test_and_clear_dead(memcg))
5935
		css_put(&memcg->css);
G
Glauber Costa 已提交
5936
}
5937
#else
5938
static int memcg_init_kmem(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
5939 5940 5941
{
	return 0;
}
G
Glauber Costa 已提交
5942

5943 5944 5945 5946 5947
static void memcg_destroy_kmem(struct mem_cgroup *memcg)
{
}

static void kmem_cgroup_css_offline(struct mem_cgroup *memcg)
G
Glauber Costa 已提交
5948 5949
{
}
5950 5951
#endif

B
Balbir Singh 已提交
5952 5953
static struct cftype mem_cgroup_files[] = {
	{
5954
		.name = "usage_in_bytes",
5955
		.private = MEMFILE_PRIVATE(_MEM, RES_USAGE),
5956
		.read = mem_cgroup_read,
K
KAMEZAWA Hiroyuki 已提交
5957 5958
		.register_event = mem_cgroup_usage_register_event,
		.unregister_event = mem_cgroup_usage_unregister_event,
B
Balbir Singh 已提交
5959
	},
5960 5961
	{
		.name = "max_usage_in_bytes",
5962
		.private = MEMFILE_PRIVATE(_MEM, RES_MAX_USAGE),
5963
		.trigger = mem_cgroup_reset,
5964
		.read = mem_cgroup_read,
5965
	},
B
Balbir Singh 已提交
5966
	{
5967
		.name = "limit_in_bytes",
5968
		.private = MEMFILE_PRIVATE(_MEM, RES_LIMIT),
5969
		.write_string = mem_cgroup_write,
5970
		.read = mem_cgroup_read,
B
Balbir Singh 已提交
5971
	},
5972 5973 5974 5975
	{
		.name = "soft_limit_in_bytes",
		.private = MEMFILE_PRIVATE(_MEM, RES_SOFT_LIMIT),
		.write_string = mem_cgroup_write,
5976
		.read = mem_cgroup_read,
5977
	},
B
Balbir Singh 已提交
5978 5979
	{
		.name = "failcnt",
5980
		.private = MEMFILE_PRIVATE(_MEM, RES_FAILCNT),
5981
		.trigger = mem_cgroup_reset,
5982
		.read = mem_cgroup_read,
B
Balbir Singh 已提交
5983
	},
5984 5985
	{
		.name = "stat",
5986
		.read_seq_string = memcg_stat_show,
5987
	},
5988 5989 5990 5991
	{
		.name = "force_empty",
		.trigger = mem_cgroup_force_empty_write,
	},
5992 5993
	{
		.name = "use_hierarchy",
5994
		.flags = CFTYPE_INSANE,
5995 5996 5997
		.write_u64 = mem_cgroup_hierarchy_write,
		.read_u64 = mem_cgroup_hierarchy_read,
	},
K
KOSAKI Motohiro 已提交
5998 5999 6000 6001 6002
	{
		.name = "swappiness",
		.read_u64 = mem_cgroup_swappiness_read,
		.write_u64 = mem_cgroup_swappiness_write,
	},
6003 6004 6005 6006 6007
	{
		.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
6008 6009
	{
		.name = "oom_control",
6010 6011
		.read_map = mem_cgroup_oom_control_read,
		.write_u64 = mem_cgroup_oom_control_write,
K
KAMEZAWA Hiroyuki 已提交
6012 6013 6014 6015
		.register_event = mem_cgroup_oom_register_event,
		.unregister_event = mem_cgroup_oom_unregister_event,
		.private = MEMFILE_PRIVATE(_OOM_TYPE, OOM_CONTROL),
	},
6016 6017 6018 6019 6020
	{
		.name = "pressure_level",
		.register_event = vmpressure_register_event,
		.unregister_event = vmpressure_unregister_event,
	},
6021 6022 6023
#ifdef CONFIG_NUMA
	{
		.name = "numa_stat",
6024
		.read_seq_string = memcg_numa_stat_show,
6025 6026
	},
#endif
6027 6028 6029 6030 6031 6032 6033 6034 6035 6036 6037 6038 6039 6040 6041 6042 6043 6044 6045 6046 6047 6048 6049 6050
#ifdef CONFIG_MEMCG_KMEM
	{
		.name = "kmem.limit_in_bytes",
		.private = MEMFILE_PRIVATE(_KMEM, RES_LIMIT),
		.write_string = mem_cgroup_write,
		.read = mem_cgroup_read,
	},
	{
		.name = "kmem.usage_in_bytes",
		.private = MEMFILE_PRIVATE(_KMEM, RES_USAGE),
		.read = mem_cgroup_read,
	},
	{
		.name = "kmem.failcnt",
		.private = MEMFILE_PRIVATE(_KMEM, RES_FAILCNT),
		.trigger = mem_cgroup_reset,
		.read = mem_cgroup_read,
	},
	{
		.name = "kmem.max_usage_in_bytes",
		.private = MEMFILE_PRIVATE(_KMEM, RES_MAX_USAGE),
		.trigger = mem_cgroup_reset,
		.read = mem_cgroup_read,
	},
6051 6052 6053 6054 6055 6056
#ifdef CONFIG_SLABINFO
	{
		.name = "kmem.slabinfo",
		.read_seq_string = mem_cgroup_slabinfo_read,
	},
#endif
6057
#endif
6058
	{ },	/* terminate */
6059
};
6060

6061 6062 6063 6064 6065 6066 6067 6068 6069 6070 6071 6072 6073 6074 6075 6076 6077 6078 6079 6080 6081 6082 6083 6084 6085 6086 6087 6088 6089 6090
#ifdef CONFIG_MEMCG_SWAP
static struct cftype memsw_cgroup_files[] = {
	{
		.name = "memsw.usage_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_USAGE),
		.read = mem_cgroup_read,
		.register_event = mem_cgroup_usage_register_event,
		.unregister_event = mem_cgroup_usage_unregister_event,
	},
	{
		.name = "memsw.max_usage_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_MAX_USAGE),
		.trigger = mem_cgroup_reset,
		.read = mem_cgroup_read,
	},
	{
		.name = "memsw.limit_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_LIMIT),
		.write_string = mem_cgroup_write,
		.read = mem_cgroup_read,
	},
	{
		.name = "memsw.failcnt",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_FAILCNT),
		.trigger = mem_cgroup_reset,
		.read = mem_cgroup_read,
	},
	{ },	/* terminate */
};
#endif
6091
static int alloc_mem_cgroup_per_zone_info(struct mem_cgroup *memcg, int node)
6092 6093
{
	struct mem_cgroup_per_node *pn;
6094
	struct mem_cgroup_per_zone *mz;
6095
	int zone, tmp = node;
6096 6097 6098 6099 6100 6101 6102 6103
	/*
	 * 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.
	 */
6104 6105
	if (!node_state(node, N_NORMAL_MEMORY))
		tmp = -1;
6106
	pn = kzalloc_node(sizeof(*pn), GFP_KERNEL, tmp);
6107 6108
	if (!pn)
		return 1;
6109 6110 6111

	for (zone = 0; zone < MAX_NR_ZONES; zone++) {
		mz = &pn->zoneinfo[zone];
6112
		lruvec_init(&mz->lruvec);
6113
		mz->usage_in_excess = 0;
6114
		mz->on_tree = false;
6115
		mz->memcg = memcg;
6116
	}
6117
	memcg->nodeinfo[node] = pn;
6118 6119 6120
	return 0;
}

6121
static void free_mem_cgroup_per_zone_info(struct mem_cgroup *memcg, int node)
6122
{
6123
	kfree(memcg->nodeinfo[node]);
6124 6125
}

6126 6127
static struct mem_cgroup *mem_cgroup_alloc(void)
{
6128
	struct mem_cgroup *memcg;
6129
	size_t size = memcg_size();
6130

6131
	/* Can be very big if nr_node_ids is very big */
6132
	if (size < PAGE_SIZE)
6133
		memcg = kzalloc(size, GFP_KERNEL);
6134
	else
6135
		memcg = vzalloc(size);
6136

6137
	if (!memcg)
6138 6139
		return NULL;

6140 6141
	memcg->stat = alloc_percpu(struct mem_cgroup_stat_cpu);
	if (!memcg->stat)
6142
		goto out_free;
6143 6144
	spin_lock_init(&memcg->pcp_counter_lock);
	return memcg;
6145 6146 6147

out_free:
	if (size < PAGE_SIZE)
6148
		kfree(memcg);
6149
	else
6150
		vfree(memcg);
6151
	return NULL;
6152 6153
}

6154
/*
6155 6156 6157 6158 6159 6160 6161 6162
 * 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.
6163
 */
6164 6165

static void __mem_cgroup_free(struct mem_cgroup *memcg)
6166
{
6167
	int node;
6168
	size_t size = memcg_size();
6169

6170 6171 6172 6173 6174 6175 6176 6177
	mem_cgroup_remove_from_trees(memcg);
	free_css_id(&mem_cgroup_subsys, &memcg->css);

	for_each_node(node)
		free_mem_cgroup_per_zone_info(memcg, node);

	free_percpu(memcg->stat);

6178 6179 6180 6181 6182 6183 6184 6185 6186 6187 6188
	/*
	 * We need to make sure that (at least for now), the jump label
	 * destruction code runs outside of the cgroup lock. This is because
	 * get_online_cpus(), which is called from the static_branch update,
	 * can't be called inside the cgroup_lock. cpusets are the ones
	 * enforcing this dependency, so if they ever change, we might as well.
	 *
	 * schedule_work() will guarantee this happens. Be careful if you need
	 * to move this code around, and make sure it is outside
	 * the cgroup_lock.
	 */
6189
	disarm_static_keys(memcg);
6190 6191 6192 6193
	if (size < PAGE_SIZE)
		kfree(memcg);
	else
		vfree(memcg);
6194
}
6195

6196 6197 6198
/*
 * Returns the parent mem_cgroup in memcgroup hierarchy with hierarchy enabled.
 */
G
Glauber Costa 已提交
6199
struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
6200
{
6201
	if (!memcg->res.parent)
6202
		return NULL;
6203
	return mem_cgroup_from_res_counter(memcg->res.parent, res);
6204
}
G
Glauber Costa 已提交
6205
EXPORT_SYMBOL(parent_mem_cgroup);
6206

6207
static void __init mem_cgroup_soft_limit_tree_init(void)
6208 6209 6210 6211 6212
{
	struct mem_cgroup_tree_per_node *rtpn;
	struct mem_cgroup_tree_per_zone *rtpz;
	int tmp, node, zone;

B
Bob Liu 已提交
6213
	for_each_node(node) {
6214 6215 6216 6217
		tmp = node;
		if (!node_state(node, N_NORMAL_MEMORY))
			tmp = -1;
		rtpn = kzalloc_node(sizeof(*rtpn), GFP_KERNEL, tmp);
6218
		BUG_ON(!rtpn);
6219 6220 6221 6222 6223 6224 6225 6226 6227 6228 6229

		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
6230
static struct cgroup_subsys_state * __ref
6231
mem_cgroup_css_alloc(struct cgroup *cont)
B
Balbir Singh 已提交
6232
{
6233
	struct mem_cgroup *memcg;
K
KAMEZAWA Hiroyuki 已提交
6234
	long error = -ENOMEM;
6235
	int node;
B
Balbir Singh 已提交
6236

6237 6238
	memcg = mem_cgroup_alloc();
	if (!memcg)
K
KAMEZAWA Hiroyuki 已提交
6239
		return ERR_PTR(error);
6240

B
Bob Liu 已提交
6241
	for_each_node(node)
6242
		if (alloc_mem_cgroup_per_zone_info(memcg, node))
6243
			goto free_out;
6244

6245
	/* root ? */
6246
	if (cont->parent == NULL) {
6247
		root_mem_cgroup = memcg;
6248 6249 6250
		res_counter_init(&memcg->res, NULL);
		res_counter_init(&memcg->memsw, NULL);
		res_counter_init(&memcg->kmem, NULL);
6251
	}
6252

6253 6254 6255 6256 6257
	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);
6258
	vmpressure_init(&memcg->vmpressure);
6259 6260 6261 6262 6263 6264 6265 6266 6267 6268 6269 6270 6271 6272 6273 6274 6275

	return &memcg->css;

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

static int
mem_cgroup_css_online(struct cgroup *cont)
{
	struct mem_cgroup *memcg, *parent;
	int error = 0;

	if (!cont->parent)
		return 0;

6276
	mutex_lock(&memcg_create_mutex);
6277 6278 6279 6280 6281 6282 6283 6284
	memcg = mem_cgroup_from_cont(cont);
	parent = mem_cgroup_from_cont(cont->parent);

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

	if (parent->use_hierarchy) {
6285 6286
		res_counter_init(&memcg->res, &parent->res);
		res_counter_init(&memcg->memsw, &parent->memsw);
6287
		res_counter_init(&memcg->kmem, &parent->kmem);
6288

6289
		/*
6290 6291
		 * No need to take a reference to the parent because cgroup
		 * core guarantees its existence.
6292
		 */
6293
	} else {
6294 6295
		res_counter_init(&memcg->res, NULL);
		res_counter_init(&memcg->memsw, NULL);
6296
		res_counter_init(&memcg->kmem, NULL);
6297 6298 6299 6300 6301
		/*
		 * Deeper hierachy with use_hierarchy == false doesn't make
		 * much sense so let cgroup subsystem know about this
		 * unfortunate state in our controller.
		 */
6302
		if (parent != root_mem_cgroup)
6303
			mem_cgroup_subsys.broken_hierarchy = true;
6304
	}
6305 6306

	error = memcg_init_kmem(memcg, &mem_cgroup_subsys);
6307
	mutex_unlock(&memcg_create_mutex);
6308
	return error;
B
Balbir Singh 已提交
6309 6310
}

M
Michal Hocko 已提交
6311 6312 6313 6314 6315 6316 6317 6318
/*
 * Announce all parents that a group from their hierarchy is gone.
 */
static void mem_cgroup_invalidate_reclaim_iterators(struct mem_cgroup *memcg)
{
	struct mem_cgroup *parent = memcg;

	while ((parent = parent_mem_cgroup(parent)))
6319
		mem_cgroup_iter_invalidate(parent);
M
Michal Hocko 已提交
6320 6321 6322 6323 6324 6325

	/*
	 * if the root memcg is not hierarchical we have to check it
	 * explicitely.
	 */
	if (!root_mem_cgroup->use_hierarchy)
6326
		mem_cgroup_iter_invalidate(root_mem_cgroup);
M
Michal Hocko 已提交
6327 6328
}

6329
static void mem_cgroup_css_offline(struct cgroup *cont)
6330
{
6331
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
6332

6333 6334
	kmem_cgroup_css_offline(memcg);

M
Michal Hocko 已提交
6335
	mem_cgroup_invalidate_reclaim_iterators(memcg);
6336
	mem_cgroup_reparent_charges(memcg);
G
Glauber Costa 已提交
6337
	mem_cgroup_destroy_all_caches(memcg);
6338 6339
}

6340
static void mem_cgroup_css_free(struct cgroup *cont)
B
Balbir Singh 已提交
6341
{
6342
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
6343

6344
	memcg_destroy_kmem(memcg);
6345
	__mem_cgroup_free(memcg);
B
Balbir Singh 已提交
6346 6347
}

6348
#ifdef CONFIG_MMU
6349
/* Handlers for move charge at task migration. */
6350 6351
#define PRECHARGE_COUNT_AT_ONCE	256
static int mem_cgroup_do_precharge(unsigned long count)
6352
{
6353 6354
	int ret = 0;
	int batch_count = PRECHARGE_COUNT_AT_ONCE;
6355
	struct mem_cgroup *memcg = mc.to;
6356

6357
	if (mem_cgroup_is_root(memcg)) {
6358 6359 6360 6361 6362 6363 6364 6365
		mc.precharge += count;
		/* we don't need css_get for root */
		return ret;
	}
	/* try to charge at once */
	if (count > 1) {
		struct res_counter *dummy;
		/*
6366
		 * "memcg" cannot be under rmdir() because we've already checked
6367 6368 6369 6370
		 * by cgroup_lock_live_cgroup() that it is not removed and we
		 * are still under the same cgroup_mutex. So we can postpone
		 * css_get().
		 */
6371
		if (res_counter_charge(&memcg->res, PAGE_SIZE * count, &dummy))
6372
			goto one_by_one;
6373
		if (do_swap_account && res_counter_charge(&memcg->memsw,
6374
						PAGE_SIZE * count, &dummy)) {
6375
			res_counter_uncharge(&memcg->res, PAGE_SIZE * count);
6376 6377 6378 6379 6380 6381 6382 6383 6384 6385 6386 6387 6388 6389 6390 6391
			goto one_by_one;
		}
		mc.precharge += count;
		return ret;
	}
one_by_one:
	/* fall back to one by one charge */
	while (count--) {
		if (signal_pending(current)) {
			ret = -EINTR;
			break;
		}
		if (!batch_count--) {
			batch_count = PRECHARGE_COUNT_AT_ONCE;
			cond_resched();
		}
6392 6393
		ret = __mem_cgroup_try_charge(NULL,
					GFP_KERNEL, 1, &memcg, false);
6394
		if (ret)
6395
			/* mem_cgroup_clear_mc() will do uncharge later */
6396
			return ret;
6397 6398
		mc.precharge++;
	}
6399 6400 6401 6402
	return ret;
}

/**
6403
 * get_mctgt_type - get target type of moving charge
6404 6405 6406
 * @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
6407
 * @target: the pointer the target page or swap ent will be stored(can be NULL)
6408 6409 6410 6411 6412 6413
 *
 * 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).
6414 6415 6416
 *   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.
6417 6418 6419 6420 6421
 *
 * Called with pte lock held.
 */
union mc_target {
	struct page	*page;
6422
	swp_entry_t	ent;
6423 6424 6425
};

enum mc_target_type {
6426
	MC_TARGET_NONE = 0,
6427
	MC_TARGET_PAGE,
6428
	MC_TARGET_SWAP,
6429 6430
};

D
Daisuke Nishimura 已提交
6431 6432
static struct page *mc_handle_present_pte(struct vm_area_struct *vma,
						unsigned long addr, pte_t ptent)
6433
{
D
Daisuke Nishimura 已提交
6434
	struct page *page = vm_normal_page(vma, addr, ptent);
6435

D
Daisuke Nishimura 已提交
6436 6437 6438 6439
	if (!page || !page_mapped(page))
		return NULL;
	if (PageAnon(page)) {
		/* we don't move shared anon */
6440
		if (!move_anon())
D
Daisuke Nishimura 已提交
6441
			return NULL;
6442 6443
	} else if (!move_file())
		/* we ignore mapcount for file pages */
D
Daisuke Nishimura 已提交
6444 6445 6446 6447 6448 6449 6450
		return NULL;
	if (!get_page_unless_zero(page))
		return NULL;

	return page;
}

6451
#ifdef CONFIG_SWAP
D
Daisuke Nishimura 已提交
6452 6453 6454 6455 6456 6457 6458 6459
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;
6460 6461 6462 6463
	/*
	 * Because lookup_swap_cache() updates some statistics counter,
	 * we call find_get_page() with swapper_space directly.
	 */
6464
	page = find_get_page(swap_address_space(ent), ent.val);
D
Daisuke Nishimura 已提交
6465 6466 6467 6468 6469
	if (do_swap_account)
		entry->val = ent.val;

	return page;
}
6470 6471 6472 6473 6474 6475 6476
#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
6477

6478 6479 6480 6481 6482 6483 6484 6485 6486 6487 6488 6489 6490 6491 6492 6493 6494 6495 6496
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). */
6497 6498 6499 6500 6501 6502
	page = find_get_page(mapping, pgoff);

#ifdef CONFIG_SWAP
	/* shmem/tmpfs may report page out on swap: account for that too. */
	if (radix_tree_exceptional_entry(page)) {
		swp_entry_t swap = radix_to_swp_entry(page);
6503
		if (do_swap_account)
6504
			*entry = swap;
6505
		page = find_get_page(swap_address_space(swap), swap.val);
6506
	}
6507
#endif
6508 6509 6510
	return page;
}

6511
static enum mc_target_type get_mctgt_type(struct vm_area_struct *vma,
D
Daisuke Nishimura 已提交
6512 6513 6514 6515
		unsigned long addr, pte_t ptent, union mc_target *target)
{
	struct page *page = NULL;
	struct page_cgroup *pc;
6516
	enum mc_target_type ret = MC_TARGET_NONE;
D
Daisuke Nishimura 已提交
6517 6518 6519 6520 6521 6522
	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);
6523 6524
	else if (pte_none(ptent) || pte_file(ptent))
		page = mc_handle_file_pte(vma, addr, ptent, &ent);
D
Daisuke Nishimura 已提交
6525 6526

	if (!page && !ent.val)
6527
		return ret;
6528 6529 6530 6531 6532 6533 6534 6535 6536 6537 6538 6539 6540 6541 6542
	if (page) {
		pc = lookup_page_cgroup(page);
		/*
		 * Do only loose check w/o page_cgroup lock.
		 * mem_cgroup_move_account() checks the pc is valid or not under
		 * the lock.
		 */
		if (PageCgroupUsed(pc) && pc->mem_cgroup == mc.from) {
			ret = MC_TARGET_PAGE;
			if (target)
				target->page = page;
		}
		if (!ret || !target)
			put_page(page);
	}
D
Daisuke Nishimura 已提交
6543 6544
	/* There is a swap entry and a page doesn't exist or isn't charged */
	if (ent.val && !ret &&
6545
			css_id(&mc.from->css) == lookup_swap_cgroup_id(ent)) {
6546 6547 6548
		ret = MC_TARGET_SWAP;
		if (target)
			target->ent = ent;
6549 6550 6551 6552
	}
	return ret;
}

6553 6554 6555 6556 6557 6558 6559 6560 6561 6562 6563 6564 6565 6566 6567 6568 6569 6570 6571 6572 6573 6574 6575 6576 6577 6578 6579 6580 6581 6582 6583 6584 6585 6586 6587
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
/*
 * We don't consider swapping or file mapped pages because THP does not
 * support them for now.
 * Caller should make sure that pmd_trans_huge(pmd) is true.
 */
static enum mc_target_type get_mctgt_type_thp(struct vm_area_struct *vma,
		unsigned long addr, pmd_t pmd, union mc_target *target)
{
	struct page *page = NULL;
	struct page_cgroup *pc;
	enum mc_target_type ret = MC_TARGET_NONE;

	page = pmd_page(pmd);
	VM_BUG_ON(!page || !PageHead(page));
	if (!move_anon())
		return ret;
	pc = lookup_page_cgroup(page);
	if (PageCgroupUsed(pc) && pc->mem_cgroup == mc.from) {
		ret = MC_TARGET_PAGE;
		if (target) {
			get_page(page);
			target->page = page;
		}
	}
	return ret;
}
#else
static inline enum mc_target_type get_mctgt_type_thp(struct vm_area_struct *vma,
		unsigned long addr, pmd_t pmd, union mc_target *target)
{
	return MC_TARGET_NONE;
}
#endif

6588 6589 6590 6591 6592 6593 6594 6595
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;

6596 6597 6598 6599
	if (pmd_trans_huge_lock(pmd, vma) == 1) {
		if (get_mctgt_type_thp(vma, addr, *pmd, NULL) == MC_TARGET_PAGE)
			mc.precharge += HPAGE_PMD_NR;
		spin_unlock(&vma->vm_mm->page_table_lock);
6600
		return 0;
6601
	}
6602

6603 6604
	if (pmd_trans_unstable(pmd))
		return 0;
6605 6606
	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
	for (; addr != end; pte++, addr += PAGE_SIZE)
6607
		if (get_mctgt_type(vma, addr, *pte, NULL))
6608 6609 6610 6611
			mc.precharge++;	/* increment precharge temporarily */
	pte_unmap_unlock(pte - 1, ptl);
	cond_resched();

6612 6613 6614
	return 0;
}

6615 6616 6617 6618 6619
static unsigned long mem_cgroup_count_precharge(struct mm_struct *mm)
{
	unsigned long precharge;
	struct vm_area_struct *vma;

6620
	down_read(&mm->mmap_sem);
6621 6622 6623 6624 6625 6626 6627 6628 6629 6630 6631
	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);
	}
6632
	up_read(&mm->mmap_sem);
6633 6634 6635 6636 6637 6638 6639 6640 6641

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

	return precharge;
}

static int mem_cgroup_precharge_mc(struct mm_struct *mm)
{
6642 6643 6644 6645 6646
	unsigned long precharge = mem_cgroup_count_precharge(mm);

	VM_BUG_ON(mc.moving_task);
	mc.moving_task = current;
	return mem_cgroup_do_precharge(precharge);
6647 6648
}

6649 6650
/* cancels all extra charges on mc.from and mc.to, and wakes up all waiters. */
static void __mem_cgroup_clear_mc(void)
6651
{
6652 6653
	struct mem_cgroup *from = mc.from;
	struct mem_cgroup *to = mc.to;
L
Li Zefan 已提交
6654
	int i;
6655

6656
	/* we must uncharge all the leftover precharges from mc.to */
6657 6658 6659 6660 6661 6662 6663 6664 6665 6666 6667
	if (mc.precharge) {
		__mem_cgroup_cancel_charge(mc.to, mc.precharge);
		mc.precharge = 0;
	}
	/*
	 * we didn't uncharge from mc.from at mem_cgroup_move_account(), so
	 * we must uncharge here.
	 */
	if (mc.moved_charge) {
		__mem_cgroup_cancel_charge(mc.from, mc.moved_charge);
		mc.moved_charge = 0;
6668
	}
6669 6670 6671 6672 6673 6674
	/* we must fixup refcnts and charges */
	if (mc.moved_swap) {
		/* uncharge swap account from the old cgroup */
		if (!mem_cgroup_is_root(mc.from))
			res_counter_uncharge(&mc.from->memsw,
						PAGE_SIZE * mc.moved_swap);
L
Li Zefan 已提交
6675 6676 6677

		for (i = 0; i < mc.moved_swap; i++)
			css_put(&mc.from->css);
6678 6679 6680 6681 6682 6683 6684 6685 6686

		if (!mem_cgroup_is_root(mc.to)) {
			/*
			 * we charged both to->res and to->memsw, so we should
			 * uncharge to->res.
			 */
			res_counter_uncharge(&mc.to->res,
						PAGE_SIZE * mc.moved_swap);
		}
L
Li Zefan 已提交
6687
		/* we've already done css_get(mc.to) */
6688 6689
		mc.moved_swap = 0;
	}
6690 6691 6692 6693 6694 6695 6696 6697 6698 6699 6700 6701 6702 6703 6704
	memcg_oom_recover(from);
	memcg_oom_recover(to);
	wake_up_all(&mc.waitq);
}

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

	/*
	 * we must clear moving_task before waking up waiters at the end of
	 * task migration.
	 */
	mc.moving_task = NULL;
	__mem_cgroup_clear_mc();
6705
	spin_lock(&mc.lock);
6706 6707
	mc.from = NULL;
	mc.to = NULL;
6708
	spin_unlock(&mc.lock);
6709
	mem_cgroup_end_move(from);
6710 6711
}

6712 6713
static int mem_cgroup_can_attach(struct cgroup *cgroup,
				 struct cgroup_taskset *tset)
6714
{
6715
	struct task_struct *p = cgroup_taskset_first(tset);
6716
	int ret = 0;
6717
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgroup);
6718
	unsigned long move_charge_at_immigrate;
6719

6720 6721 6722 6723 6724 6725 6726
	/*
	 * 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) {
6727 6728 6729
		struct mm_struct *mm;
		struct mem_cgroup *from = mem_cgroup_from_task(p);

6730
		VM_BUG_ON(from == memcg);
6731 6732 6733 6734 6735

		mm = get_task_mm(p);
		if (!mm)
			return 0;
		/* We move charges only when we move a owner of the mm */
6736 6737 6738 6739
		if (mm->owner == p) {
			VM_BUG_ON(mc.from);
			VM_BUG_ON(mc.to);
			VM_BUG_ON(mc.precharge);
6740
			VM_BUG_ON(mc.moved_charge);
6741
			VM_BUG_ON(mc.moved_swap);
6742
			mem_cgroup_start_move(from);
6743
			spin_lock(&mc.lock);
6744
			mc.from = from;
6745
			mc.to = memcg;
6746
			mc.immigrate_flags = move_charge_at_immigrate;
6747
			spin_unlock(&mc.lock);
6748
			/* We set mc.moving_task later */
6749 6750 6751 6752

			ret = mem_cgroup_precharge_mc(mm);
			if (ret)
				mem_cgroup_clear_mc();
6753 6754
		}
		mmput(mm);
6755 6756 6757 6758
	}
	return ret;
}

6759 6760
static void mem_cgroup_cancel_attach(struct cgroup *cgroup,
				     struct cgroup_taskset *tset)
6761
{
6762
	mem_cgroup_clear_mc();
6763 6764
}

6765 6766 6767
static int mem_cgroup_move_charge_pte_range(pmd_t *pmd,
				unsigned long addr, unsigned long end,
				struct mm_walk *walk)
6768
{
6769 6770 6771 6772
	int ret = 0;
	struct vm_area_struct *vma = walk->private;
	pte_t *pte;
	spinlock_t *ptl;
6773 6774 6775 6776
	enum mc_target_type target_type;
	union mc_target target;
	struct page *page;
	struct page_cgroup *pc;
6777

6778 6779 6780 6781 6782 6783 6784 6785 6786 6787 6788
	/*
	 * 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.
	 */
	if (pmd_trans_huge_lock(pmd, vma) == 1) {
6789
		if (mc.precharge < HPAGE_PMD_NR) {
6790 6791 6792 6793 6794 6795 6796 6797 6798
			spin_unlock(&vma->vm_mm->page_table_lock);
			return 0;
		}
		target_type = get_mctgt_type_thp(vma, addr, *pmd, &target);
		if (target_type == MC_TARGET_PAGE) {
			page = target.page;
			if (!isolate_lru_page(page)) {
				pc = lookup_page_cgroup(page);
				if (!mem_cgroup_move_account(page, HPAGE_PMD_NR,
6799
							pc, mc.from, mc.to)) {
6800 6801 6802 6803 6804 6805 6806 6807
					mc.precharge -= HPAGE_PMD_NR;
					mc.moved_charge += HPAGE_PMD_NR;
				}
				putback_lru_page(page);
			}
			put_page(page);
		}
		spin_unlock(&vma->vm_mm->page_table_lock);
6808
		return 0;
6809 6810
	}

6811 6812
	if (pmd_trans_unstable(pmd))
		return 0;
6813 6814 6815 6816
retry:
	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
	for (; addr != end; addr += PAGE_SIZE) {
		pte_t ptent = *(pte++);
6817
		swp_entry_t ent;
6818 6819 6820 6821

		if (!mc.precharge)
			break;

6822
		switch (get_mctgt_type(vma, addr, ptent, &target)) {
6823 6824 6825 6826 6827
		case MC_TARGET_PAGE:
			page = target.page;
			if (isolate_lru_page(page))
				goto put;
			pc = lookup_page_cgroup(page);
6828
			if (!mem_cgroup_move_account(page, 1, pc,
6829
						     mc.from, mc.to)) {
6830
				mc.precharge--;
6831 6832
				/* we uncharge from mc.from later. */
				mc.moved_charge++;
6833 6834
			}
			putback_lru_page(page);
6835
put:			/* get_mctgt_type() gets the page */
6836 6837
			put_page(page);
			break;
6838 6839
		case MC_TARGET_SWAP:
			ent = target.ent;
6840
			if (!mem_cgroup_move_swap_account(ent, mc.from, mc.to)) {
6841
				mc.precharge--;
6842 6843 6844
				/* we fixup refcnts and charges later. */
				mc.moved_swap++;
			}
6845
			break;
6846 6847 6848 6849 6850 6851 6852 6853 6854 6855 6856 6857 6858 6859
		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.
		 */
6860
		ret = mem_cgroup_do_precharge(1);
6861 6862 6863 6864 6865 6866 6867 6868 6869 6870 6871 6872
		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();
6873 6874 6875 6876 6877 6878 6879 6880 6881 6882 6883 6884 6885
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;
	}
6886 6887 6888 6889 6890 6891 6892 6893 6894 6895 6896 6897 6898 6899 6900 6901 6902 6903
	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;
	}
6904
	up_read(&mm->mmap_sem);
6905 6906
}

6907 6908
static void mem_cgroup_move_task(struct cgroup *cont,
				 struct cgroup_taskset *tset)
B
Balbir Singh 已提交
6909
{
6910
	struct task_struct *p = cgroup_taskset_first(tset);
6911
	struct mm_struct *mm = get_task_mm(p);
6912 6913

	if (mm) {
6914 6915
		if (mc.to)
			mem_cgroup_move_charge(mm);
6916 6917
		mmput(mm);
	}
6918 6919
	if (mc.to)
		mem_cgroup_clear_mc();
B
Balbir Singh 已提交
6920
}
6921
#else	/* !CONFIG_MMU */
6922 6923
static int mem_cgroup_can_attach(struct cgroup *cgroup,
				 struct cgroup_taskset *tset)
6924 6925 6926
{
	return 0;
}
6927 6928
static void mem_cgroup_cancel_attach(struct cgroup *cgroup,
				     struct cgroup_taskset *tset)
6929 6930
{
}
6931 6932
static void mem_cgroup_move_task(struct cgroup *cont,
				 struct cgroup_taskset *tset)
6933 6934 6935
{
}
#endif
B
Balbir Singh 已提交
6936

6937 6938 6939 6940 6941 6942 6943 6944 6945 6946 6947 6948 6949 6950 6951
/*
 * Cgroup retains root cgroups across [un]mount cycles making it necessary
 * to verify sane_behavior flag on each mount attempt.
 */
static void mem_cgroup_bind(struct cgroup *root)
{
	/*
	 * use_hierarchy is forced with sane_behavior.  cgroup core
	 * guarantees that @root doesn't have any children, so turning it
	 * on for the root memcg is enough.
	 */
	if (cgroup_sane_behavior(root))
		mem_cgroup_from_cont(root)->use_hierarchy = true;
}

B
Balbir Singh 已提交
6952 6953 6954
struct cgroup_subsys mem_cgroup_subsys = {
	.name = "memory",
	.subsys_id = mem_cgroup_subsys_id,
6955
	.css_alloc = mem_cgroup_css_alloc,
6956
	.css_online = mem_cgroup_css_online,
6957 6958
	.css_offline = mem_cgroup_css_offline,
	.css_free = mem_cgroup_css_free,
6959 6960
	.can_attach = mem_cgroup_can_attach,
	.cancel_attach = mem_cgroup_cancel_attach,
B
Balbir Singh 已提交
6961
	.attach = mem_cgroup_move_task,
6962
	.bind = mem_cgroup_bind,
6963
	.base_cftypes = mem_cgroup_files,
6964
	.early_init = 0,
K
KAMEZAWA Hiroyuki 已提交
6965
	.use_id = 1,
B
Balbir Singh 已提交
6966
};
6967

A
Andrew Morton 已提交
6968
#ifdef CONFIG_MEMCG_SWAP
6969 6970 6971
static int __init enable_swap_account(char *s)
{
	/* consider enabled if no parameter or 1 is given */
6972
	if (!strcmp(s, "1"))
6973
		really_do_swap_account = 1;
6974
	else if (!strcmp(s, "0"))
6975 6976 6977
		really_do_swap_account = 0;
	return 1;
}
6978
__setup("swapaccount=", enable_swap_account);
6979

6980 6981
static void __init memsw_file_init(void)
{
6982 6983 6984 6985 6986 6987 6988 6989 6990
	WARN_ON(cgroup_add_cftypes(&mem_cgroup_subsys, memsw_cgroup_files));
}

static void __init enable_swap_cgroup(void)
{
	if (!mem_cgroup_disabled() && really_do_swap_account) {
		do_swap_account = 1;
		memsw_file_init();
	}
6991
}
6992

6993
#else
6994
static void __init enable_swap_cgroup(void)
6995 6996
{
}
6997
#endif
6998 6999

/*
7000 7001 7002 7003 7004 7005
 * 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.
7006 7007 7008 7009
 */
static int __init mem_cgroup_init(void)
{
	hotcpu_notifier(memcg_cpu_hotplug_callback, 0);
7010
	enable_swap_cgroup();
7011
	mem_cgroup_soft_limit_tree_init();
7012
	memcg_stock_init();
7013 7014 7015
	return 0;
}
subsys_initcall(mem_cgroup_init);