memcontrol.c 145.7 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
 *
B
Balbir Singh 已提交
13 14 15 16 17 18 19 20 21 22 23 24 25 26
 * 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>
27
#include <linux/mm.h>
28
#include <linux/hugetlb.h>
K
KAMEZAWA Hiroyuki 已提交
29
#include <linux/pagemap.h>
30
#include <linux/smp.h>
31
#include <linux/page-flags.h>
32
#include <linux/backing-dev.h>
33 34
#include <linux/bit_spinlock.h>
#include <linux/rcupdate.h>
35
#include <linux/limits.h>
36
#include <linux/export.h>
37
#include <linux/mutex.h>
38
#include <linux/rbtree.h>
39
#include <linux/slab.h>
40
#include <linux/swap.h>
41
#include <linux/swapops.h>
42
#include <linux/spinlock.h>
43 44
#include <linux/eventfd.h>
#include <linux/sort.h>
45
#include <linux/fs.h>
46
#include <linux/seq_file.h>
47
#include <linux/vmalloc.h>
48
#include <linux/mm_inline.h>
49
#include <linux/page_cgroup.h>
50
#include <linux/cpu.h>
51
#include <linux/oom.h>
K
KAMEZAWA Hiroyuki 已提交
52
#include "internal.h"
G
Glauber Costa 已提交
53 54
#include <net/sock.h>
#include <net/tcp_memcontrol.h>
B
Balbir Singh 已提交
55

56 57
#include <asm/uaccess.h>

58 59
#include <trace/events/vmscan.h>

60 61
struct cgroup_subsys mem_cgroup_subsys __read_mostly;
#define MEM_CGROUP_RECLAIM_RETRIES	5
62
static struct mem_cgroup *root_mem_cgroup __read_mostly;
B
Balbir Singh 已提交
63

64
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
L
Li Zefan 已提交
65
/* Turned on only when memory cgroup is enabled && really_do_swap_account = 1 */
66
int do_swap_account __read_mostly;
67 68 69 70 71 72 73 74

/* for remember boot option*/
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP_ENABLED
static int really_do_swap_account __initdata = 1;
#else
static int really_do_swap_account __initdata = 0;
#endif

75 76 77 78 79
#else
#define do_swap_account		(0)
#endif


80 81 82 83 84 85 86 87
/*
 * Statistics for memory cgroup.
 */
enum mem_cgroup_stat_index {
	/*
	 * For MEM_CONTAINER_TYPE_ALL, usage = pagecache + rss.
	 */
	MEM_CGROUP_STAT_CACHE, 	   /* # of pages charged as cache */
88
	MEM_CGROUP_STAT_RSS,	   /* # of pages charged as anon rss */
89
	MEM_CGROUP_STAT_FILE_MAPPED,  /* # of pages charged as file rss */
90
	MEM_CGROUP_STAT_SWAPOUT, /* # of pages, swapped out */
91
	MEM_CGROUP_STAT_DATA, /* end of data requires synchronization */
92 93 94
	MEM_CGROUP_STAT_NSTATS,
};

95 96 97 98
enum mem_cgroup_events_index {
	MEM_CGROUP_EVENTS_PGPGIN,	/* # of pages paged in */
	MEM_CGROUP_EVENTS_PGPGOUT,	/* # of pages paged out */
	MEM_CGROUP_EVENTS_COUNT,	/* # of pages paged in/out */
99 100
	MEM_CGROUP_EVENTS_PGFAULT,	/* # of page-faults */
	MEM_CGROUP_EVENTS_PGMAJFAULT,	/* # of major page-faults */
101 102
	MEM_CGROUP_EVENTS_NSTATS,
};
103 104 105 106 107 108 109 110 111
/*
 * 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,
112
	MEM_CGROUP_TARGET_NUMAINFO,
113 114 115 116
	MEM_CGROUP_NTARGETS,
};
#define THRESHOLDS_EVENTS_TARGET (128)
#define SOFTLIMIT_EVENTS_TARGET (1024)
117
#define NUMAINFO_EVENTS_TARGET	(1024)
118

119
struct mem_cgroup_stat_cpu {
120
	long count[MEM_CGROUP_STAT_NSTATS];
121
	unsigned long events[MEM_CGROUP_EVENTS_NSTATS];
122
	unsigned long targets[MEM_CGROUP_NTARGETS];
123 124
};

125 126 127 128 129 130 131
struct mem_cgroup_reclaim_iter {
	/* css_id of the last scanned hierarchy member */
	int position;
	/* scan generation, increased every round-trip */
	unsigned int generation;
};

132 133 134 135
/*
 * per-zone information in memory controller.
 */
struct mem_cgroup_per_zone {
136
	struct lruvec		lruvec;
137
	unsigned long		lru_size[NR_LRU_LISTS];
K
KOSAKI Motohiro 已提交
138

139 140
	struct mem_cgroup_reclaim_iter reclaim_iter[DEF_PRIORITY + 1];

141 142 143 144
	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;
145
	struct mem_cgroup	*memcg;		/* Back pointer, we cannot */
146
						/* use container_of	   */
147 148 149 150 151 152 153 154 155 156
};

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

struct mem_cgroup_lru_info {
	struct mem_cgroup_per_node *nodeinfo[MAX_NUMNODES];
};

157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176
/*
 * 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;

177 178 179 180 181
struct mem_cgroup_threshold {
	struct eventfd_ctx *eventfd;
	u64 threshold;
};

K
KAMEZAWA Hiroyuki 已提交
182
/* For threshold */
183 184
struct mem_cgroup_threshold_ary {
	/* An array index points to threshold just below usage. */
185
	int current_threshold;
186 187 188 189 190
	/* Size of entries[] */
	unsigned int size;
	/* Array of thresholds */
	struct mem_cgroup_threshold entries[0];
};
191 192 193 194 195 196 197 198 199 200 201 202

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
203 204 205 206 207
/* for OOM */
struct mem_cgroup_eventfd_list {
	struct list_head list;
	struct eventfd_ctx *eventfd;
};
208

209 210
static void mem_cgroup_threshold(struct mem_cgroup *memcg);
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg);
211

B
Balbir Singh 已提交
212 213 214 215 216 217 218
/*
 * 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
219 220 221
 * 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
222 223 224 225 226 227 228
 */
struct mem_cgroup {
	struct cgroup_subsys_state css;
	/*
	 * the counter to account for memory usage
	 */
	struct res_counter res;
229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252

	union {
		/*
		 * the counter to account for mem+swap usage.
		 */
		struct res_counter memsw;

		/*
		 * rcu_freeing is used only when freeing struct mem_cgroup,
		 * so put it into a union to avoid wasting more memory.
		 * It must be disjoint from the css field.  It could be
		 * in a union with the res field, but res plays a much
		 * larger part in mem_cgroup life than memsw, and might
		 * be of interest, even at time of free, when debugging.
		 * So share rcu_head with the less interesting memsw.
		 */
		struct rcu_head rcu_freeing;
		/*
		 * But when using vfree(), that cannot be done at
		 * interrupt time, so we must then queue the work.
		 */
		struct work_struct work_freeing;
	};

253 254 255 256
	/*
	 * Per cgroup active and inactive list, similar to the
	 * per zone LRU lists.
	 */
257
	struct mem_cgroup_lru_info info;
258 259 260
	int last_scanned_node;
#if MAX_NUMNODES > 1
	nodemask_t	scan_nodes;
261 262
	atomic_t	numainfo_events;
	atomic_t	numainfo_updating;
263
#endif
264 265 266 267
	/*
	 * Should the accounting and control be hierarchical, per subtree?
	 */
	bool use_hierarchy;
268 269 270 271

	bool		oom_lock;
	atomic_t	under_oom;

272
	atomic_t	refcnt;
273

274
	int	swappiness;
275 276
	/* OOM-Killer disable */
	int		oom_kill_disable;
K
KOSAKI Motohiro 已提交
277

278 279 280
	/* set when res.limit == memsw.limit */
	bool		memsw_is_minimum;

281 282 283 284
	/* protect arrays of thresholds */
	struct mutex thresholds_lock;

	/* thresholds for memory usage. RCU-protected */
285
	struct mem_cgroup_thresholds thresholds;
286

287
	/* thresholds for mem+swap usage. RCU-protected */
288
	struct mem_cgroup_thresholds memsw_thresholds;
289

K
KAMEZAWA Hiroyuki 已提交
290 291
	/* For oom notifier event fd */
	struct list_head oom_notify;
292

293 294 295 296 297
	/*
	 * 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;
298 299 300 301
	/*
	 * set > 0 if pages under this cgroup are moving to other cgroup.
	 */
	atomic_t	moving_account;
302 303
	/* taken only while moving_account > 0 */
	spinlock_t	move_lock;
304
	/*
305
	 * percpu counter.
306
	 */
307
	struct mem_cgroup_stat_cpu *stat;
308 309 310 311 312 313
	/*
	 * 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
314 315 316 317

#ifdef CONFIG_INET
	struct tcp_memcontrol tcp_mem;
#endif
B
Balbir Singh 已提交
318 319
};

320 321 322 323 324 325
/* Stuffs for move charges at task migration. */
/*
 * Types of charges to be moved. "move_charge_at_immitgrate" is treated as a
 * left-shifted bitmap of these types.
 */
enum move_type {
326
	MOVE_CHARGE_TYPE_ANON,	/* private anonymous page and swap of it */
327
	MOVE_CHARGE_TYPE_FILE,	/* file page(including tmpfs) and swap of it */
328 329 330
	NR_MOVE_TYPE,
};

331 332
/* "mc" and its members are protected by cgroup_mutex */
static struct move_charge_struct {
333
	spinlock_t	  lock; /* for from, to */
334 335 336
	struct mem_cgroup *from;
	struct mem_cgroup *to;
	unsigned long precharge;
337
	unsigned long moved_charge;
338
	unsigned long moved_swap;
339 340 341
	struct task_struct *moving_task;	/* a task moving charges */
	wait_queue_head_t waitq;		/* a waitq for other context */
} mc = {
342
	.lock = __SPIN_LOCK_UNLOCKED(mc.lock),
343 344
	.waitq = __WAIT_QUEUE_HEAD_INITIALIZER(mc.waitq),
};
345

D
Daisuke Nishimura 已提交
346 347 348 349 350 351
static bool move_anon(void)
{
	return test_bit(MOVE_CHARGE_TYPE_ANON,
					&mc.to->move_charge_at_immigrate);
}

352 353 354 355 356 357
static bool move_file(void)
{
	return test_bit(MOVE_CHARGE_TYPE_FILE,
					&mc.to->move_charge_at_immigrate);
}

358 359 360 361 362 363 364
/*
 * Maximum loops in mem_cgroup_hierarchical_reclaim(), used for soft
 * limit reclaim to prevent infinite loops, if they ever occur.
 */
#define	MEM_CGROUP_MAX_RECLAIM_LOOPS		(100)
#define	MEM_CGROUP_MAX_SOFT_LIMIT_RECLAIM_LOOPS	(2)

365 366 367
enum charge_type {
	MEM_CGROUP_CHARGE_TYPE_CACHE = 0,
	MEM_CGROUP_CHARGE_TYPE_MAPPED,
368
	MEM_CGROUP_CHARGE_TYPE_SHMEM,	/* used by page migration of shmem */
369
	MEM_CGROUP_CHARGE_TYPE_FORCE,	/* used by force_empty */
K
KAMEZAWA Hiroyuki 已提交
370
	MEM_CGROUP_CHARGE_TYPE_SWAPOUT,	/* for accounting swapcache */
K
KAMEZAWA Hiroyuki 已提交
371
	MEM_CGROUP_CHARGE_TYPE_DROP,	/* a page was unused swap cache */
372 373 374
	NR_CHARGE_TYPE,
};

375
/* for encoding cft->private value on file */
376 377 378
#define _MEM			(0)
#define _MEMSWAP		(1)
#define _OOM_TYPE		(2)
379 380 381
#define MEMFILE_PRIVATE(x, val)	(((x) << 16) | (val))
#define MEMFILE_TYPE(val)	(((val) >> 16) & 0xffff)
#define MEMFILE_ATTR(val)	((val) & 0xffff)
K
KAMEZAWA Hiroyuki 已提交
382 383
/* Used for OOM nofiier */
#define OOM_CONTROL		(0)
384

385 386 387 388 389 390 391 392
/*
 * 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)

393 394
static void mem_cgroup_get(struct mem_cgroup *memcg);
static void mem_cgroup_put(struct mem_cgroup *memcg);
G
Glauber Costa 已提交
395 396 397 398

/* Writing them here to avoid exposing memcg's inner layout */
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
#include <net/sock.h>
G
Glauber Costa 已提交
399
#include <net/ip.h>
G
Glauber Costa 已提交
400 401 402 403

static bool mem_cgroup_is_root(struct mem_cgroup *memcg);
void sock_update_memcg(struct sock *sk)
{
404
	if (mem_cgroup_sockets_enabled) {
G
Glauber Costa 已提交
405 406 407 408
		struct mem_cgroup *memcg;

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

409 410 411 412 413 414 415 416 417 418 419 420 421 422
		/* 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));
			mem_cgroup_get(sk->sk_cgrp->memcg);
			return;
		}

G
Glauber Costa 已提交
423 424 425 426 427 428 429 430 431 432 433 434 435
		rcu_read_lock();
		memcg = mem_cgroup_from_task(current);
		if (!mem_cgroup_is_root(memcg)) {
			mem_cgroup_get(memcg);
			sk->sk_cgrp = sk->sk_prot->proto_cgroup(memcg);
		}
		rcu_read_unlock();
	}
}
EXPORT_SYMBOL(sock_update_memcg);

void sock_release_memcg(struct sock *sk)
{
436
	if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
G
Glauber Costa 已提交
437 438 439 440 441 442
		struct mem_cgroup *memcg;
		WARN_ON(!sk->sk_cgrp->memcg);
		memcg = sk->sk_cgrp->memcg;
		mem_cgroup_put(memcg);
	}
}
G
Glauber Costa 已提交
443

444
#ifdef CONFIG_INET
G
Glauber Costa 已提交
445 446 447 448 449 450 451 452
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
453 454 455
#endif /* CONFIG_INET */
#endif /* CONFIG_CGROUP_MEM_RES_CTLR_KMEM */

456
static void drain_all_stock_async(struct mem_cgroup *memcg);
457

458
static struct mem_cgroup_per_zone *
459
mem_cgroup_zoneinfo(struct mem_cgroup *memcg, int nid, int zid)
460
{
461
	return &memcg->info.nodeinfo[nid]->zoneinfo[zid];
462 463
}

464
struct cgroup_subsys_state *mem_cgroup_css(struct mem_cgroup *memcg)
465
{
466
	return &memcg->css;
467 468
}

469
static struct mem_cgroup_per_zone *
470
page_cgroup_zoneinfo(struct mem_cgroup *memcg, struct page *page)
471
{
472 473
	int nid = page_to_nid(page);
	int zid = page_zonenum(page);
474

475
	return mem_cgroup_zoneinfo(memcg, nid, zid);
476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493
}

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
494
__mem_cgroup_insert_exceeded(struct mem_cgroup *memcg,
495
				struct mem_cgroup_per_zone *mz,
496 497
				struct mem_cgroup_tree_per_zone *mctz,
				unsigned long long new_usage_in_excess)
498 499 500 501 502 503 504 505
{
	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;

506 507 508
	mz->usage_in_excess = new_usage_in_excess;
	if (!mz->usage_in_excess)
		return;
509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524
	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;
525 526 527
}

static void
528
__mem_cgroup_remove_exceeded(struct mem_cgroup *memcg,
529 530 531 532 533 534 535 536 537
				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;
}

538
static void
539
mem_cgroup_remove_exceeded(struct mem_cgroup *memcg,
540 541 542 543
				struct mem_cgroup_per_zone *mz,
				struct mem_cgroup_tree_per_zone *mctz)
{
	spin_lock(&mctz->lock);
544
	__mem_cgroup_remove_exceeded(memcg, mz, mctz);
545 546 547 548
	spin_unlock(&mctz->lock);
}


549
static void mem_cgroup_update_tree(struct mem_cgroup *memcg, struct page *page)
550
{
551
	unsigned long long excess;
552 553
	struct mem_cgroup_per_zone *mz;
	struct mem_cgroup_tree_per_zone *mctz;
554 555
	int nid = page_to_nid(page);
	int zid = page_zonenum(page);
556 557 558
	mctz = soft_limit_tree_from_page(page);

	/*
559 560
	 * Necessary to update all ancestors when hierarchy is used.
	 * because their event counter is not touched.
561
	 */
562 563 564
	for (; memcg; memcg = parent_mem_cgroup(memcg)) {
		mz = mem_cgroup_zoneinfo(memcg, nid, zid);
		excess = res_counter_soft_limit_excess(&memcg->res);
565 566 567 568
		/*
		 * We have to update the tree if mz is on RB-tree or
		 * mem is over its softlimit.
		 */
569
		if (excess || mz->on_tree) {
570 571 572
			spin_lock(&mctz->lock);
			/* if on-tree, remove it */
			if (mz->on_tree)
573
				__mem_cgroup_remove_exceeded(memcg, mz, mctz);
574
			/*
575 576
			 * Insert again. mz->usage_in_excess will be updated.
			 * If excess is 0, no tree ops.
577
			 */
578
			__mem_cgroup_insert_exceeded(memcg, mz, mctz, excess);
579 580
			spin_unlock(&mctz->lock);
		}
581 582 583
	}
}

584
static void mem_cgroup_remove_from_trees(struct mem_cgroup *memcg)
585 586 587 588 589
{
	int node, zone;
	struct mem_cgroup_per_zone *mz;
	struct mem_cgroup_tree_per_zone *mctz;

B
Bob Liu 已提交
590
	for_each_node(node) {
591
		for (zone = 0; zone < MAX_NR_ZONES; zone++) {
592
			mz = mem_cgroup_zoneinfo(memcg, node, zone);
593
			mctz = soft_limit_tree_node_zone(node, zone);
594
			mem_cgroup_remove_exceeded(memcg, mz, mctz);
595 596 597 598
		}
	}
}

599 600 601 602
static struct mem_cgroup_per_zone *
__mem_cgroup_largest_soft_limit_node(struct mem_cgroup_tree_per_zone *mctz)
{
	struct rb_node *rightmost = NULL;
603
	struct mem_cgroup_per_zone *mz;
604 605

retry:
606
	mz = NULL;
607 608 609 610 611 612 613 614 615 616
	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.
	 */
617 618 619
	__mem_cgroup_remove_exceeded(mz->memcg, mz, mctz);
	if (!res_counter_soft_limit_excess(&mz->memcg->res) ||
		!css_tryget(&mz->memcg->css))
620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635
		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;
}

636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654
/*
 * 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.
 */
655
static long mem_cgroup_read_stat(struct mem_cgroup *memcg,
656
				 enum mem_cgroup_stat_index idx)
657
{
658
	long val = 0;
659 660
	int cpu;

661 662
	get_online_cpus();
	for_each_online_cpu(cpu)
663
		val += per_cpu(memcg->stat->count[idx], cpu);
664
#ifdef CONFIG_HOTPLUG_CPU
665 666 667
	spin_lock(&memcg->pcp_counter_lock);
	val += memcg->nocpu_base.count[idx];
	spin_unlock(&memcg->pcp_counter_lock);
668 669
#endif
	put_online_cpus();
670 671 672
	return val;
}

673
static void mem_cgroup_swap_statistics(struct mem_cgroup *memcg,
674 675 676
					 bool charge)
{
	int val = (charge) ? 1 : -1;
677
	this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_SWAPOUT], val);
678 679
}

680
static unsigned long mem_cgroup_read_events(struct mem_cgroup *memcg,
681 682 683 684 685 686
					    enum mem_cgroup_events_index idx)
{
	unsigned long val = 0;
	int cpu;

	for_each_online_cpu(cpu)
687
		val += per_cpu(memcg->stat->events[idx], cpu);
688
#ifdef CONFIG_HOTPLUG_CPU
689 690 691
	spin_lock(&memcg->pcp_counter_lock);
	val += memcg->nocpu_base.events[idx];
	spin_unlock(&memcg->pcp_counter_lock);
692 693 694 695
#endif
	return val;
}

696
static void mem_cgroup_charge_statistics(struct mem_cgroup *memcg,
697
					 bool anon, int nr_pages)
698
{
699 700
	preempt_disable();

701 702 703 704 705 706
	/*
	 * 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],
707
				nr_pages);
708
	else
709
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_CACHE],
710
				nr_pages);
711

712 713
	/* pagein of a big page is an event. So, ignore page size */
	if (nr_pages > 0)
714
		__this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGPGIN]);
715
	else {
716
		__this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGPGOUT]);
717 718
		nr_pages = -nr_pages; /* for event */
	}
719

720
	__this_cpu_add(memcg->stat->events[MEM_CGROUP_EVENTS_COUNT], nr_pages);
721

722
	preempt_enable();
723 724
}

725
unsigned long
726
mem_cgroup_zone_nr_lru_pages(struct mem_cgroup *memcg, int nid, int zid,
727
			unsigned int lru_mask)
728 729
{
	struct mem_cgroup_per_zone *mz;
H
Hugh Dickins 已提交
730
	enum lru_list lru;
731 732
	unsigned long ret = 0;

733
	mz = mem_cgroup_zoneinfo(memcg, nid, zid);
734

H
Hugh Dickins 已提交
735 736 737
	for_each_lru(lru) {
		if (BIT(lru) & lru_mask)
			ret += mz->lru_size[lru];
738 739 740 741 742
	}
	return ret;
}

static unsigned long
743
mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
744 745
			int nid, unsigned int lru_mask)
{
746 747 748
	u64 total = 0;
	int zid;

749
	for (zid = 0; zid < MAX_NR_ZONES; zid++)
750 751
		total += mem_cgroup_zone_nr_lru_pages(memcg,
						nid, zid, lru_mask);
752

753 754
	return total;
}
755

756
static unsigned long mem_cgroup_nr_lru_pages(struct mem_cgroup *memcg,
757
			unsigned int lru_mask)
758
{
759
	int nid;
760 761
	u64 total = 0;

762
	for_each_node_state(nid, N_HIGH_MEMORY)
763
		total += mem_cgroup_node_nr_lru_pages(memcg, nid, lru_mask);
764
	return total;
765 766
}

767 768
static bool mem_cgroup_event_ratelimit(struct mem_cgroup *memcg,
				       enum mem_cgroup_events_target target)
769 770 771
{
	unsigned long val, next;

772 773
	val = __this_cpu_read(memcg->stat->events[MEM_CGROUP_EVENTS_COUNT]);
	next = __this_cpu_read(memcg->stat->targets[target]);
774
	/* from time_after() in jiffies.h */
775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790
	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;
791
	}
792
	return false;
793 794 795 796 797 798
}

/*
 * Check events in order.
 *
 */
799
static void memcg_check_events(struct mem_cgroup *memcg, struct page *page)
800
{
801
	preempt_disable();
802
	/* threshold event is triggered in finer grain than soft limit */
803 804
	if (unlikely(mem_cgroup_event_ratelimit(memcg,
						MEM_CGROUP_TARGET_THRESH))) {
805 806
		bool do_softlimit;
		bool do_numainfo __maybe_unused;
807 808 809 810 811 812 813 814 815

		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();

816
		mem_cgroup_threshold(memcg);
817
		if (unlikely(do_softlimit))
818
			mem_cgroup_update_tree(memcg, page);
819
#if MAX_NUMNODES > 1
820
		if (unlikely(do_numainfo))
821
			atomic_inc(&memcg->numainfo_events);
822
#endif
823 824
	} else
		preempt_enable();
825 826
}

G
Glauber Costa 已提交
827
struct mem_cgroup *mem_cgroup_from_cont(struct cgroup *cont)
B
Balbir Singh 已提交
828 829 830 831 832 833
{
	return container_of(cgroup_subsys_state(cont,
				mem_cgroup_subsys_id), struct mem_cgroup,
				css);
}

834
struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p)
835
{
836 837 838 839 840 841 842 843
	/*
	 * 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;

844 845 846 847
	return container_of(task_subsys_state(p, mem_cgroup_subsys_id),
				struct mem_cgroup, css);
}

848
struct mem_cgroup *try_get_mem_cgroup_from_mm(struct mm_struct *mm)
849
{
850
	struct mem_cgroup *memcg = NULL;
851 852 853

	if (!mm)
		return NULL;
854 855 856 857 858 859 860
	/*
	 * 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 {
861 862
		memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
		if (unlikely(!memcg))
863
			break;
864
	} while (!css_tryget(&memcg->css));
865
	rcu_read_unlock();
866
	return memcg;
867 868
}

869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888
/**
 * 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
889
{
890 891
	struct mem_cgroup *memcg = NULL;
	int id = 0;
892

893 894 895
	if (mem_cgroup_disabled())
		return NULL;

896 897
	if (!root)
		root = root_mem_cgroup;
K
KAMEZAWA Hiroyuki 已提交
898

899 900
	if (prev && !reclaim)
		id = css_id(&prev->css);
K
KAMEZAWA Hiroyuki 已提交
901

902 903
	if (prev && prev != root)
		css_put(&prev->css);
K
KAMEZAWA Hiroyuki 已提交
904

905 906 907 908 909
	if (!root->use_hierarchy && root != root_mem_cgroup) {
		if (prev)
			return NULL;
		return root;
	}
K
KAMEZAWA Hiroyuki 已提交
910

911
	while (!memcg) {
912
		struct mem_cgroup_reclaim_iter *uninitialized_var(iter);
913
		struct cgroup_subsys_state *css;
914

915 916 917 918 919 920 921 922 923 924 925
		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];
			if (prev && reclaim->generation != iter->generation)
				return NULL;
			id = iter->position;
		}
K
KAMEZAWA Hiroyuki 已提交
926

927 928 929 930 931 932 933 934
		rcu_read_lock();
		css = css_get_next(&mem_cgroup_subsys, id + 1, &root->css, &id);
		if (css) {
			if (css == &root->css || css_tryget(css))
				memcg = container_of(css,
						     struct mem_cgroup, css);
		} else
			id = 0;
K
KAMEZAWA Hiroyuki 已提交
935 936
		rcu_read_unlock();

937 938 939 940 941 942 943
		if (reclaim) {
			iter->position = id;
			if (!css)
				iter->generation++;
			else if (!prev && memcg)
				reclaim->generation = iter->generation;
		}
944 945 946 947 948

		if (prev && !css)
			return NULL;
	}
	return memcg;
K
KAMEZAWA Hiroyuki 已提交
949
}
K
KAMEZAWA Hiroyuki 已提交
950

951 952 953 954 955 956 957
/**
 * 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)
958 959 960 961 962 963
{
	if (!root)
		root = root_mem_cgroup;
	if (prev && prev != root)
		css_put(&prev->css);
}
K
KAMEZAWA Hiroyuki 已提交
964

965 966 967 968 969 970
/*
 * 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)		\
971
	for (iter = mem_cgroup_iter(root, NULL, NULL);	\
972
	     iter != NULL;				\
973
	     iter = mem_cgroup_iter(root, iter, NULL))
974

975
#define for_each_mem_cgroup(iter)			\
976
	for (iter = mem_cgroup_iter(NULL, NULL, NULL);	\
977
	     iter != NULL;				\
978
	     iter = mem_cgroup_iter(NULL, iter, NULL))
K
KAMEZAWA Hiroyuki 已提交
979

980
static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
981
{
982
	return (memcg == root_mem_cgroup);
983 984
}

985 986
void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
{
987
	struct mem_cgroup *memcg;
988 989 990 991 992

	if (!mm)
		return;

	rcu_read_lock();
993 994
	memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
	if (unlikely(!memcg))
995 996 997 998
		goto out;

	switch (idx) {
	case PGFAULT:
999 1000 1001 1002
		this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]);
		break;
	case PGMAJFAULT:
		this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]);
1003 1004 1005 1006 1007 1008 1009 1010 1011
		break;
	default:
		BUG();
	}
out:
	rcu_read_unlock();
}
EXPORT_SYMBOL(mem_cgroup_count_vm_event);

1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032
/**
 * mem_cgroup_zone_lruvec - get the lru list vector for a zone and memcg
 * @zone: zone of the wanted lruvec
 * @mem: memcg of the wanted lruvec
 *
 * 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;

	if (mem_cgroup_disabled())
		return &zone->lruvec;

	mz = mem_cgroup_zoneinfo(memcg, zone_to_nid(zone), zone_idx(zone));
	return &mz->lruvec;
}

K
KAMEZAWA Hiroyuki 已提交
1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045
/*
 * 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.
 */
1046

1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060
/**
 * mem_cgroup_lru_add_list - account for adding an lru page and return lruvec
 * @zone: zone of the page
 * @page: the page
 * @lru: current lru
 *
 * This function accounts for @page being added to @lru, and returns
 * the lruvec for the given @zone and the memcg @page is charged to.
 *
 * The callsite is then responsible for physically linking the page to
 * the returned lruvec->lists[@lru].
 */
struct lruvec *mem_cgroup_lru_add_list(struct zone *zone, struct page *page,
				       enum lru_list lru)
K
KAMEZAWA Hiroyuki 已提交
1061 1062
{
	struct mem_cgroup_per_zone *mz;
1063 1064
	struct mem_cgroup *memcg;
	struct page_cgroup *pc;
1065

1066
	if (mem_cgroup_disabled())
1067 1068
		return &zone->lruvec;

K
KAMEZAWA Hiroyuki 已提交
1069
	pc = lookup_page_cgroup(page);
1070
	memcg = pc->mem_cgroup;
1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083

	/*
	 * Surreptitiously switch any uncharged page to root:
	 * 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.
	 */
	if (!PageCgroupUsed(pc) && memcg != root_mem_cgroup)
		pc->mem_cgroup = memcg = root_mem_cgroup;

1084 1085
	mz = page_cgroup_zoneinfo(memcg, page);
	/* compound_order() is stabilized through lru_lock */
1086
	mz->lru_size[lru] += 1 << compound_order(page);
1087
	return &mz->lruvec;
K
KAMEZAWA Hiroyuki 已提交
1088
}
1089

1090 1091 1092 1093 1094 1095 1096 1097 1098
/**
 * mem_cgroup_lru_del_list - account for removing an lru page
 * @page: the page
 * @lru: target lru
 *
 * This function accounts for @page being removed from @lru.
 *
 * The callsite is then responsible for physically unlinking
 * @page->lru.
1099
 */
1100
void mem_cgroup_lru_del_list(struct page *page, enum lru_list lru)
1101 1102
{
	struct mem_cgroup_per_zone *mz;
1103
	struct mem_cgroup *memcg;
1104 1105 1106 1107 1108 1109
	struct page_cgroup *pc;

	if (mem_cgroup_disabled())
		return;

	pc = lookup_page_cgroup(page);
1110 1111
	memcg = pc->mem_cgroup;
	VM_BUG_ON(!memcg);
1112 1113
	mz = page_cgroup_zoneinfo(memcg, page);
	/* huge page split is done under lru_lock. so, we have no races. */
1114 1115
	VM_BUG_ON(mz->lru_size[lru] < (1 << compound_order(page)));
	mz->lru_size[lru] -= 1 << compound_order(page);
1116 1117
}

1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135
/**
 * mem_cgroup_lru_move_lists - account for moving a page between lrus
 * @zone: zone of the page
 * @page: the page
 * @from: current lru
 * @to: target lru
 *
 * This function accounts for @page being moved between the lrus @from
 * and @to, and returns the lruvec for the given @zone and the memcg
 * @page is charged to.
 *
 * The callsite is then responsible for physically relinking
 * @page->lru to the returned lruvec->lists[@to].
 */
struct lruvec *mem_cgroup_lru_move_lists(struct zone *zone,
					 struct page *page,
					 enum lru_list from,
					 enum lru_list to)
1136
{
1137 1138 1139
	/* XXX: Optimize this, especially for @from == @to */
	mem_cgroup_lru_del_list(page, from);
	return mem_cgroup_lru_add_list(zone, page, to);
K
KAMEZAWA Hiroyuki 已提交
1140
}
1141

1142
/*
1143
 * Checks whether given mem is same or in the root_mem_cgroup's
1144 1145
 * hierarchy subtree
 */
1146 1147
bool __mem_cgroup_same_or_subtree(const struct mem_cgroup *root_memcg,
				  struct mem_cgroup *memcg)
1148
{
1149 1150 1151 1152
	if (root_memcg == memcg)
		return true;
	if (!root_memcg->use_hierarchy)
		return false;
1153 1154 1155 1156 1157 1158 1159 1160
	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;

1161
	rcu_read_lock();
1162
	ret = __mem_cgroup_same_or_subtree(root_memcg, memcg);
1163 1164
	rcu_read_unlock();
	return ret;
1165 1166
}

1167
int task_in_mem_cgroup(struct task_struct *task, const struct mem_cgroup *memcg)
1168 1169
{
	int ret;
1170
	struct mem_cgroup *curr = NULL;
1171
	struct task_struct *p;
1172

1173
	p = find_lock_task_mm(task);
1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188
	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.
		 */
		task_lock(task);
		curr = mem_cgroup_from_task(task);
		if (curr)
			css_get(&curr->css);
		task_unlock(task);
	}
1189 1190
	if (!curr)
		return 0;
1191
	/*
1192
	 * We should check use_hierarchy of "memcg" not "curr". Because checking
1193
	 * use_hierarchy of "curr" here make this function true if hierarchy is
1194 1195
	 * enabled in "curr" and "curr" is a child of "memcg" in *cgroup*
	 * hierarchy(even if use_hierarchy is disabled in "memcg").
1196
	 */
1197
	ret = mem_cgroup_same_or_subtree(memcg, curr);
1198
	css_put(&curr->css);
1199 1200 1201
	return ret;
}

1202
int mem_cgroup_inactive_anon_is_low(struct mem_cgroup *memcg, struct zone *zone)
1203
{
1204 1205 1206
	unsigned long inactive_ratio;
	int nid = zone_to_nid(zone);
	int zid = zone_idx(zone);
1207
	unsigned long inactive;
1208
	unsigned long active;
1209
	unsigned long gb;
1210

1211 1212 1213 1214
	inactive = mem_cgroup_zone_nr_lru_pages(memcg, nid, zid,
						BIT(LRU_INACTIVE_ANON));
	active = mem_cgroup_zone_nr_lru_pages(memcg, nid, zid,
					      BIT(LRU_ACTIVE_ANON));
1215

1216 1217 1218 1219 1220 1221
	gb = (inactive + active) >> (30 - PAGE_SHIFT);
	if (gb)
		inactive_ratio = int_sqrt(10 * gb);
	else
		inactive_ratio = 1;

1222
	return inactive * inactive_ratio < active;
1223 1224
}

1225
int mem_cgroup_inactive_file_is_low(struct mem_cgroup *memcg, struct zone *zone)
1226 1227 1228
{
	unsigned long active;
	unsigned long inactive;
1229 1230
	int zid = zone_idx(zone);
	int nid = zone_to_nid(zone);
1231

1232 1233 1234 1235
	inactive = mem_cgroup_zone_nr_lru_pages(memcg, nid, zid,
						BIT(LRU_INACTIVE_FILE));
	active = mem_cgroup_zone_nr_lru_pages(memcg, nid, zid,
					      BIT(LRU_ACTIVE_FILE));
1236 1237 1238 1239

	return (active > inactive);
}

K
KOSAKI Motohiro 已提交
1240 1241 1242 1243 1244 1245 1246 1247 1248 1249
struct zone_reclaim_stat *
mem_cgroup_get_reclaim_stat_from_page(struct page *page)
{
	struct page_cgroup *pc;
	struct mem_cgroup_per_zone *mz;

	if (mem_cgroup_disabled())
		return NULL;

	pc = lookup_page_cgroup(page);
1250 1251
	if (!PageCgroupUsed(pc))
		return NULL;
1252 1253
	/* Ensure pc->mem_cgroup is visible after reading PCG_USED. */
	smp_rmb();
1254
	mz = page_cgroup_zoneinfo(pc->mem_cgroup, page);
1255
	return &mz->lruvec.reclaim_stat;
K
KOSAKI Motohiro 已提交
1256 1257
}

1258 1259 1260
#define mem_cgroup_from_res_counter(counter, member)	\
	container_of(counter, struct mem_cgroup, member)

1261
/**
1262 1263
 * mem_cgroup_margin - calculate chargeable space of a memory cgroup
 * @mem: the memory cgroup
1264
 *
1265
 * Returns the maximum amount of memory @mem can be charged with, in
1266
 * pages.
1267
 */
1268
static unsigned long mem_cgroup_margin(struct mem_cgroup *memcg)
1269
{
1270 1271
	unsigned long long margin;

1272
	margin = res_counter_margin(&memcg->res);
1273
	if (do_swap_account)
1274
		margin = min(margin, res_counter_margin(&memcg->memsw));
1275
	return margin >> PAGE_SHIFT;
1276 1277
}

1278
int mem_cgroup_swappiness(struct mem_cgroup *memcg)
K
KOSAKI Motohiro 已提交
1279 1280 1281 1282 1283 1284 1285
{
	struct cgroup *cgrp = memcg->css.cgroup;

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

1286
	return memcg->swappiness;
K
KOSAKI Motohiro 已提交
1287 1288
}

1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302
/*
 * 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.
 */
1303 1304 1305 1306

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

1307
static void mem_cgroup_start_move(struct mem_cgroup *memcg)
1308
{
1309
	atomic_inc(&memcg_moving);
1310
	atomic_inc(&memcg->moving_account);
1311 1312 1313
	synchronize_rcu();
}

1314
static void mem_cgroup_end_move(struct mem_cgroup *memcg)
1315
{
1316 1317 1318 1319
	/*
	 * Now, mem_cgroup_clear_mc() may call this function with NULL.
	 * We check NULL in callee rather than caller.
	 */
1320 1321
	if (memcg) {
		atomic_dec(&memcg_moving);
1322
		atomic_dec(&memcg->moving_account);
1323
	}
1324
}
1325

1326 1327 1328
/*
 * 2 routines for checking "mem" is under move_account() or not.
 *
1329 1330
 * mem_cgroup_stolen() -  checking whether a cgroup is mc.from or not. This
 *			  is used for avoiding races in accounting.  If true,
1331 1332 1333 1334 1335 1336 1337
 *			  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".
 */

1338
static bool mem_cgroup_stolen(struct mem_cgroup *memcg)
1339 1340
{
	VM_BUG_ON(!rcu_read_lock_held());
1341
	return atomic_read(&memcg->moving_account) > 0;
1342
}
1343

1344
static bool mem_cgroup_under_move(struct mem_cgroup *memcg)
1345
{
1346 1347
	struct mem_cgroup *from;
	struct mem_cgroup *to;
1348
	bool ret = false;
1349 1350 1351 1352 1353 1354 1355 1356 1357
	/*
	 * 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;
1358

1359 1360
	ret = mem_cgroup_same_or_subtree(memcg, from)
		|| mem_cgroup_same_or_subtree(memcg, to);
1361 1362
unlock:
	spin_unlock(&mc.lock);
1363 1364 1365
	return ret;
}

1366
static bool mem_cgroup_wait_acct_move(struct mem_cgroup *memcg)
1367 1368
{
	if (mc.moving_task && current != mc.moving_task) {
1369
		if (mem_cgroup_under_move(memcg)) {
1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381
			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;
}

1382 1383 1384 1385
/*
 * 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.
1386
 * see mem_cgroup_stolen(), too.
1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399
 */
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);
}

1400
/**
1401
 * mem_cgroup_print_oom_info: Called from OOM with tasklist_lock held in read mode.
1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419
 * @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;

1420
	if (!memcg || !p)
1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465
		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();

	printk(KERN_INFO "Task in %s killed", memcg_name);

	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
	 */
	printk(KERN_CONT " as a result of limit of %s\n", memcg_name);
done:

	printk(KERN_INFO "memory: usage %llukB, limit %llukB, failcnt %llu\n",
		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));
	printk(KERN_INFO "memory+swap: usage %llukB, limit %llukB, "
		"failcnt %llu\n",
		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));
}

1466 1467 1468 1469
/*
 * This function returns the number of memcg under hierarchy tree. Returns
 * 1(self count) if no children.
 */
1470
static int mem_cgroup_count_children(struct mem_cgroup *memcg)
1471 1472
{
	int num = 0;
K
KAMEZAWA Hiroyuki 已提交
1473 1474
	struct mem_cgroup *iter;

1475
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
1476
		num++;
1477 1478 1479
	return num;
}

D
David Rientjes 已提交
1480 1481 1482 1483 1484 1485 1486 1487
/*
 * Return the memory (and swap, if configured) limit for a memcg.
 */
u64 mem_cgroup_get_limit(struct mem_cgroup *memcg)
{
	u64 limit;
	u64 memsw;

1488 1489 1490
	limit = res_counter_read_u64(&memcg->res, RES_LIMIT);
	limit += total_swap_pages << PAGE_SHIFT;

D
David Rientjes 已提交
1491 1492 1493 1494 1495 1496 1497 1498
	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.
	 */
	return min(limit, memsw);
}

1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534
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;
}

1535 1536 1537 1538 1539 1540 1541 1542 1543 1544
/**
 * test_mem_cgroup_node_reclaimable
 * @mem: the target memcg
 * @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.
 */
1545
static bool test_mem_cgroup_node_reclaimable(struct mem_cgroup *memcg,
1546 1547
		int nid, bool noswap)
{
1548
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_FILE))
1549 1550 1551
		return true;
	if (noswap || !total_swap_pages)
		return false;
1552
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_ANON))
1553 1554 1555 1556
		return true;
	return false;

}
1557 1558 1559 1560 1561 1562 1563 1564
#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.
 *
 */
1565
static void mem_cgroup_may_update_nodemask(struct mem_cgroup *memcg)
1566 1567
{
	int nid;
1568 1569 1570 1571
	/*
	 * numainfo_events > 0 means there was at least NUMAINFO_EVENTS_TARGET
	 * pagein/pageout changes since the last update.
	 */
1572
	if (!atomic_read(&memcg->numainfo_events))
1573
		return;
1574
	if (atomic_inc_return(&memcg->numainfo_updating) > 1)
1575 1576 1577
		return;

	/* make a nodemask where this memcg uses memory from */
1578
	memcg->scan_nodes = node_states[N_HIGH_MEMORY];
1579 1580 1581

	for_each_node_mask(nid, node_states[N_HIGH_MEMORY]) {

1582 1583
		if (!test_mem_cgroup_node_reclaimable(memcg, nid, false))
			node_clear(nid, memcg->scan_nodes);
1584
	}
1585

1586 1587
	atomic_set(&memcg->numainfo_events, 0);
	atomic_set(&memcg->numainfo_updating, 0);
1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601
}

/*
 * 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.
 */
1602
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
1603 1604 1605
{
	int node;

1606 1607
	mem_cgroup_may_update_nodemask(memcg);
	node = memcg->last_scanned_node;
1608

1609
	node = next_node(node, memcg->scan_nodes);
1610
	if (node == MAX_NUMNODES)
1611
		node = first_node(memcg->scan_nodes);
1612 1613 1614 1615 1616 1617 1618 1619 1620
	/*
	 * 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();

1621
	memcg->last_scanned_node = node;
1622 1623 1624
	return node;
}

1625 1626 1627 1628 1629 1630
/*
 * 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.
 */
1631
static bool mem_cgroup_reclaimable(struct mem_cgroup *memcg, bool noswap)
1632 1633 1634 1635 1636 1637 1638
{
	int nid;

	/*
	 * quick check...making use of scan_node.
	 * We can skip unused nodes.
	 */
1639 1640
	if (!nodes_empty(memcg->scan_nodes)) {
		for (nid = first_node(memcg->scan_nodes);
1641
		     nid < MAX_NUMNODES;
1642
		     nid = next_node(nid, memcg->scan_nodes)) {
1643

1644
			if (test_mem_cgroup_node_reclaimable(memcg, nid, noswap))
1645 1646 1647 1648 1649 1650 1651
				return true;
		}
	}
	/*
	 * Check rest of nodes.
	 */
	for_each_node_state(nid, N_HIGH_MEMORY) {
1652
		if (node_isset(nid, memcg->scan_nodes))
1653
			continue;
1654
		if (test_mem_cgroup_node_reclaimable(memcg, nid, noswap))
1655 1656 1657 1658 1659
			return true;
	}
	return false;
}

1660
#else
1661
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
1662 1663 1664
{
	return 0;
}
1665

1666
static bool mem_cgroup_reclaimable(struct mem_cgroup *memcg, bool noswap)
1667
{
1668
	return test_mem_cgroup_node_reclaimable(memcg, 0, noswap);
1669
}
1670 1671
#endif

1672 1673 1674 1675
static int mem_cgroup_soft_reclaim(struct mem_cgroup *root_memcg,
				   struct zone *zone,
				   gfp_t gfp_mask,
				   unsigned long *total_scanned)
1676
{
1677
	struct mem_cgroup *victim = NULL;
1678
	int total = 0;
K
KAMEZAWA Hiroyuki 已提交
1679
	int loop = 0;
1680
	unsigned long excess;
1681
	unsigned long nr_scanned;
1682 1683 1684 1685
	struct mem_cgroup_reclaim_cookie reclaim = {
		.zone = zone,
		.priority = 0,
	};
1686

1687
	excess = res_counter_soft_limit_excess(&root_memcg->res) >> PAGE_SHIFT;
K
KAMEZAWA Hiroyuki 已提交
1688

1689
	while (1) {
1690
		victim = mem_cgroup_iter(root_memcg, victim, &reclaim);
1691
		if (!victim) {
K
KAMEZAWA Hiroyuki 已提交
1692
			loop++;
1693 1694 1695 1696 1697 1698
			if (loop >= 2) {
				/*
				 * If we have not been able to reclaim
				 * anything, it might because there are
				 * no reclaimable pages under this hierarchy
				 */
1699
				if (!total)
1700 1701
					break;
				/*
L
Lucas De Marchi 已提交
1702
				 * We want to do more targeted reclaim.
1703 1704 1705 1706 1707
				 * 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) ||
1708
					(loop > MEM_CGROUP_MAX_RECLAIM_LOOPS))
1709 1710
					break;
			}
1711
			continue;
1712
		}
1713
		if (!mem_cgroup_reclaimable(victim, false))
1714
			continue;
1715 1716 1717 1718
		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))
1719
			break;
1720
	}
1721
	mem_cgroup_iter_break(root_memcg, victim);
K
KAMEZAWA Hiroyuki 已提交
1722
	return total;
1723 1724
}

K
KAMEZAWA Hiroyuki 已提交
1725 1726 1727
/*
 * Check OOM-Killer is already running under our hierarchy.
 * If someone is running, return false.
1728
 * Has to be called with memcg_oom_lock
K
KAMEZAWA Hiroyuki 已提交
1729
 */
1730
static bool mem_cgroup_oom_lock(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
1731
{
1732
	struct mem_cgroup *iter, *failed = NULL;
1733

1734
	for_each_mem_cgroup_tree(iter, memcg) {
1735
		if (iter->oom_lock) {
1736 1737 1738 1739 1740
			/*
			 * this subtree of our hierarchy is already locked
			 * so we cannot give a lock.
			 */
			failed = iter;
1741 1742
			mem_cgroup_iter_break(memcg, iter);
			break;
1743 1744
		} else
			iter->oom_lock = true;
K
KAMEZAWA Hiroyuki 已提交
1745
	}
K
KAMEZAWA Hiroyuki 已提交
1746

1747
	if (!failed)
1748
		return true;
1749 1750 1751 1752 1753

	/*
	 * OK, we failed to lock the whole subtree so we have to clean up
	 * what we set up to the failing subtree
	 */
1754
	for_each_mem_cgroup_tree(iter, memcg) {
1755
		if (iter == failed) {
1756 1757
			mem_cgroup_iter_break(memcg, iter);
			break;
1758 1759 1760
		}
		iter->oom_lock = false;
	}
1761
	return false;
1762
}
1763

1764
/*
1765
 * Has to be called with memcg_oom_lock
1766
 */
1767
static int mem_cgroup_oom_unlock(struct mem_cgroup *memcg)
1768
{
K
KAMEZAWA Hiroyuki 已提交
1769 1770
	struct mem_cgroup *iter;

1771
	for_each_mem_cgroup_tree(iter, memcg)
1772 1773 1774 1775
		iter->oom_lock = false;
	return 0;
}

1776
static void mem_cgroup_mark_under_oom(struct mem_cgroup *memcg)
1777 1778 1779
{
	struct mem_cgroup *iter;

1780
	for_each_mem_cgroup_tree(iter, memcg)
1781 1782 1783
		atomic_inc(&iter->under_oom);
}

1784
static void mem_cgroup_unmark_under_oom(struct mem_cgroup *memcg)
1785 1786 1787
{
	struct mem_cgroup *iter;

K
KAMEZAWA Hiroyuki 已提交
1788 1789 1790 1791 1792
	/*
	 * 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.
	 */
1793
	for_each_mem_cgroup_tree(iter, memcg)
1794
		atomic_add_unless(&iter->under_oom, -1, 0);
1795 1796
}

1797
static DEFINE_SPINLOCK(memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
1798 1799
static DECLARE_WAIT_QUEUE_HEAD(memcg_oom_waitq);

K
KAMEZAWA Hiroyuki 已提交
1800
struct oom_wait_info {
1801
	struct mem_cgroup *memcg;
K
KAMEZAWA Hiroyuki 已提交
1802 1803 1804 1805 1806 1807
	wait_queue_t	wait;
};

static int memcg_oom_wake_function(wait_queue_t *wait,
	unsigned mode, int sync, void *arg)
{
1808 1809
	struct mem_cgroup *wake_memcg = (struct mem_cgroup *)arg;
	struct mem_cgroup *oom_wait_memcg;
K
KAMEZAWA Hiroyuki 已提交
1810 1811 1812
	struct oom_wait_info *oom_wait_info;

	oom_wait_info = container_of(wait, struct oom_wait_info, wait);
1813
	oom_wait_memcg = oom_wait_info->memcg;
K
KAMEZAWA Hiroyuki 已提交
1814 1815

	/*
1816
	 * Both of oom_wait_info->memcg and wake_memcg are stable under us.
K
KAMEZAWA Hiroyuki 已提交
1817 1818
	 * Then we can use css_is_ancestor without taking care of RCU.
	 */
1819 1820
	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
1821 1822 1823 1824
		return 0;
	return autoremove_wake_function(wait, mode, sync, arg);
}

1825
static void memcg_wakeup_oom(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
1826
{
1827 1828
	/* for filtering, pass "memcg" as argument. */
	__wake_up(&memcg_oom_waitq, TASK_NORMAL, 0, memcg);
K
KAMEZAWA Hiroyuki 已提交
1829 1830
}

1831
static void memcg_oom_recover(struct mem_cgroup *memcg)
1832
{
1833 1834
	if (memcg && atomic_read(&memcg->under_oom))
		memcg_wakeup_oom(memcg);
1835 1836
}

K
KAMEZAWA Hiroyuki 已提交
1837 1838 1839
/*
 * try to call OOM killer. returns false if we should exit memory-reclaim loop.
 */
1840 1841
static bool mem_cgroup_handle_oom(struct mem_cgroup *memcg, gfp_t mask,
				  int order)
1842
{
K
KAMEZAWA Hiroyuki 已提交
1843
	struct oom_wait_info owait;
1844
	bool locked, need_to_kill;
K
KAMEZAWA Hiroyuki 已提交
1845

1846
	owait.memcg = memcg;
K
KAMEZAWA Hiroyuki 已提交
1847 1848 1849 1850
	owait.wait.flags = 0;
	owait.wait.func = memcg_oom_wake_function;
	owait.wait.private = current;
	INIT_LIST_HEAD(&owait.wait.task_list);
1851
	need_to_kill = true;
1852
	mem_cgroup_mark_under_oom(memcg);
1853

1854
	/* At first, try to OOM lock hierarchy under memcg.*/
1855
	spin_lock(&memcg_oom_lock);
1856
	locked = mem_cgroup_oom_lock(memcg);
K
KAMEZAWA Hiroyuki 已提交
1857 1858 1859 1860 1861
	/*
	 * 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.
	 */
1862
	prepare_to_wait(&memcg_oom_waitq, &owait.wait, TASK_KILLABLE);
1863
	if (!locked || memcg->oom_kill_disable)
1864 1865
		need_to_kill = false;
	if (locked)
1866
		mem_cgroup_oom_notify(memcg);
1867
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
1868

1869 1870
	if (need_to_kill) {
		finish_wait(&memcg_oom_waitq, &owait.wait);
1871
		mem_cgroup_out_of_memory(memcg, mask, order);
1872
	} else {
K
KAMEZAWA Hiroyuki 已提交
1873
		schedule();
K
KAMEZAWA Hiroyuki 已提交
1874
		finish_wait(&memcg_oom_waitq, &owait.wait);
K
KAMEZAWA Hiroyuki 已提交
1875
	}
1876
	spin_lock(&memcg_oom_lock);
1877
	if (locked)
1878 1879
		mem_cgroup_oom_unlock(memcg);
	memcg_wakeup_oom(memcg);
1880
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
1881

1882
	mem_cgroup_unmark_under_oom(memcg);
1883

K
KAMEZAWA Hiroyuki 已提交
1884 1885 1886
	if (test_thread_flag(TIF_MEMDIE) || fatal_signal_pending(current))
		return false;
	/* Give chance to dying process */
1887
	schedule_timeout_uninterruptible(1);
K
KAMEZAWA Hiroyuki 已提交
1888
	return true;
1889 1890
}

1891 1892 1893
/*
 * Currently used to update mapped file statistics, but the routine can be
 * generalized to update other statistics as well.
1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910
 *
 * 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
1911 1912
 * small, we check mm->moving_account and detect there are possibility of race
 * If there is, we take a lock.
1913
 */
1914

1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929
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
	 * need to take move_lock_page_cgroup(). Because we already hold
	 * rcu_read_lock(), any calls to move_account will be delayed until
1930
	 * rcu_read_unlock() if mem_cgroup_stolen() == true.
1931
	 */
1932
	if (!mem_cgroup_stolen(memcg))
1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954
		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
	 * should take move_lock_page_cgroup().
	 */
	move_unlock_mem_cgroup(pc->mem_cgroup, flags);
}

1955 1956
void mem_cgroup_update_page_stat(struct page *page,
				 enum mem_cgroup_page_stat_item idx, int val)
1957
{
1958
	struct mem_cgroup *memcg;
1959
	struct page_cgroup *pc = lookup_page_cgroup(page);
1960
	unsigned long uninitialized_var(flags);
1961

1962
	if (mem_cgroup_disabled())
1963
		return;
1964

1965 1966
	memcg = pc->mem_cgroup;
	if (unlikely(!memcg || !PageCgroupUsed(pc)))
1967
		return;
1968 1969

	switch (idx) {
1970 1971
	case MEMCG_NR_FILE_MAPPED:
		idx = MEM_CGROUP_STAT_FILE_MAPPED;
1972 1973 1974
		break;
	default:
		BUG();
1975
	}
1976

1977
	this_cpu_add(memcg->stat->count[idx], val);
1978
}
1979

1980 1981 1982 1983
/*
 * size of first charge trial. "32" comes from vmscan.c's magic value.
 * TODO: maybe necessary to use big numbers in big irons.
 */
1984
#define CHARGE_BATCH	32U
1985 1986
struct memcg_stock_pcp {
	struct mem_cgroup *cached; /* this never be root cgroup */
1987
	unsigned int nr_pages;
1988
	struct work_struct work;
1989 1990
	unsigned long flags;
#define FLUSHING_CACHED_CHARGE	(0)
1991 1992
};
static DEFINE_PER_CPU(struct memcg_stock_pcp, memcg_stock);
1993
static DEFINE_MUTEX(percpu_charge_mutex);
1994 1995

/*
1996
 * Try to consume stocked charge on this cpu. If success, one page is consumed
1997 1998 1999 2000
 * from local stock and true is returned. If the stock is 0 or charges from a
 * cgroup which is not current target, returns false. This stock will be
 * refilled.
 */
2001
static bool consume_stock(struct mem_cgroup *memcg)
2002 2003 2004 2005 2006
{
	struct memcg_stock_pcp *stock;
	bool ret = true;

	stock = &get_cpu_var(memcg_stock);
2007
	if (memcg == stock->cached && stock->nr_pages)
2008
		stock->nr_pages--;
2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021
	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;

2022 2023 2024 2025
	if (stock->nr_pages) {
		unsigned long bytes = stock->nr_pages * PAGE_SIZE;

		res_counter_uncharge(&old->res, bytes);
2026
		if (do_swap_account)
2027 2028
			res_counter_uncharge(&old->memsw, bytes);
		stock->nr_pages = 0;
2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040
	}
	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);
2041
	clear_bit(FLUSHING_CACHED_CHARGE, &stock->flags);
2042 2043 2044 2045
}

/*
 * Cache charges(val) which is from res_counter, to local per_cpu area.
2046
 * This will be consumed by consume_stock() function, later.
2047
 */
2048
static void refill_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
2049 2050 2051
{
	struct memcg_stock_pcp *stock = &get_cpu_var(memcg_stock);

2052
	if (stock->cached != memcg) { /* reset if necessary */
2053
		drain_stock(stock);
2054
		stock->cached = memcg;
2055
	}
2056
	stock->nr_pages += nr_pages;
2057 2058 2059 2060
	put_cpu_var(memcg_stock);
}

/*
2061
 * Drains all per-CPU charge caches for given root_memcg resp. subtree
2062 2063
 * of the hierarchy under it. sync flag says whether we should block
 * until the work is done.
2064
 */
2065
static void drain_all_stock(struct mem_cgroup *root_memcg, bool sync)
2066
{
2067
	int cpu, curcpu;
2068

2069 2070
	/* Notify other cpus that system-wide "drain" is running */
	get_online_cpus();
2071
	curcpu = get_cpu();
2072 2073
	for_each_online_cpu(cpu) {
		struct memcg_stock_pcp *stock = &per_cpu(memcg_stock, cpu);
2074
		struct mem_cgroup *memcg;
2075

2076 2077
		memcg = stock->cached;
		if (!memcg || !stock->nr_pages)
2078
			continue;
2079
		if (!mem_cgroup_same_or_subtree(root_memcg, memcg))
2080
			continue;
2081 2082 2083 2084 2085 2086
		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);
		}
2087
	}
2088
	put_cpu();
2089 2090 2091 2092 2093 2094

	if (!sync)
		goto out;

	for_each_online_cpu(cpu) {
		struct memcg_stock_pcp *stock = &per_cpu(memcg_stock, cpu);
2095
		if (test_bit(FLUSHING_CACHED_CHARGE, &stock->flags))
2096 2097 2098
			flush_work(&stock->work);
	}
out:
2099
 	put_online_cpus();
2100 2101 2102 2103 2104 2105 2106 2107
}

/*
 * 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.
 */
2108
static void drain_all_stock_async(struct mem_cgroup *root_memcg)
2109
{
2110 2111 2112 2113 2114
	/*
	 * If someone calls draining, avoid adding more kworker runs.
	 */
	if (!mutex_trylock(&percpu_charge_mutex))
		return;
2115
	drain_all_stock(root_memcg, false);
2116
	mutex_unlock(&percpu_charge_mutex);
2117 2118 2119
}

/* This is a synchronous drain interface. */
2120
static void drain_all_stock_sync(struct mem_cgroup *root_memcg)
2121 2122
{
	/* called when force_empty is called */
2123
	mutex_lock(&percpu_charge_mutex);
2124
	drain_all_stock(root_memcg, true);
2125
	mutex_unlock(&percpu_charge_mutex);
2126 2127
}

2128 2129 2130 2131
/*
 * This function drains percpu counter value from DEAD cpu and
 * move it to local cpu. Note that this function can be preempted.
 */
2132
static void mem_cgroup_drain_pcp_counter(struct mem_cgroup *memcg, int cpu)
2133 2134 2135
{
	int i;

2136
	spin_lock(&memcg->pcp_counter_lock);
2137
	for (i = 0; i < MEM_CGROUP_STAT_DATA; i++) {
2138
		long x = per_cpu(memcg->stat->count[i], cpu);
2139

2140 2141
		per_cpu(memcg->stat->count[i], cpu) = 0;
		memcg->nocpu_base.count[i] += x;
2142
	}
2143
	for (i = 0; i < MEM_CGROUP_EVENTS_NSTATS; i++) {
2144
		unsigned long x = per_cpu(memcg->stat->events[i], cpu);
2145

2146 2147
		per_cpu(memcg->stat->events[i], cpu) = 0;
		memcg->nocpu_base.events[i] += x;
2148
	}
2149
	spin_unlock(&memcg->pcp_counter_lock);
2150 2151 2152
}

static int __cpuinit memcg_cpu_hotplug_callback(struct notifier_block *nb,
2153 2154 2155 2156 2157
					unsigned long action,
					void *hcpu)
{
	int cpu = (unsigned long)hcpu;
	struct memcg_stock_pcp *stock;
2158
	struct mem_cgroup *iter;
2159

2160
	if (action == CPU_ONLINE)
2161 2162
		return NOTIFY_OK;

2163
	if (action != CPU_DEAD && action != CPU_DEAD_FROZEN)
2164
		return NOTIFY_OK;
2165

2166
	for_each_mem_cgroup(iter)
2167 2168
		mem_cgroup_drain_pcp_counter(iter, cpu);

2169 2170 2171 2172 2173
	stock = &per_cpu(memcg_stock, cpu);
	drain_stock(stock);
	return NOTIFY_OK;
}

2174 2175 2176 2177 2178 2179 2180 2181 2182 2183

/* 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 */
};

2184
static int mem_cgroup_do_charge(struct mem_cgroup *memcg, gfp_t gfp_mask,
2185
				unsigned int nr_pages, bool oom_check)
2186
{
2187
	unsigned long csize = nr_pages * PAGE_SIZE;
2188 2189 2190 2191 2192
	struct mem_cgroup *mem_over_limit;
	struct res_counter *fail_res;
	unsigned long flags = 0;
	int ret;

2193
	ret = res_counter_charge(&memcg->res, csize, &fail_res);
2194 2195 2196 2197

	if (likely(!ret)) {
		if (!do_swap_account)
			return CHARGE_OK;
2198
		ret = res_counter_charge(&memcg->memsw, csize, &fail_res);
2199 2200 2201
		if (likely(!ret))
			return CHARGE_OK;

2202
		res_counter_uncharge(&memcg->res, csize);
2203 2204 2205 2206
		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);
2207
	/*
2208 2209
	 * nr_pages can be either a huge page (HPAGE_PMD_NR), a batch
	 * of regular pages (CHARGE_BATCH), or a single regular page (1).
2210 2211 2212 2213
	 *
	 * Never reclaim on behalf of optional batching, retry with a
	 * single page instead.
	 */
2214
	if (nr_pages == CHARGE_BATCH)
2215 2216 2217 2218 2219
		return CHARGE_RETRY;

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

2220
	ret = mem_cgroup_reclaim(mem_over_limit, gfp_mask, flags);
2221
	if (mem_cgroup_margin(mem_over_limit) >= nr_pages)
2222
		return CHARGE_RETRY;
2223
	/*
2224 2225 2226 2227 2228 2229 2230
	 * 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.
2231
	 */
2232
	if (nr_pages == 1 && ret)
2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245
		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 */
2246
	if (!mem_cgroup_handle_oom(mem_over_limit, gfp_mask, get_order(csize)))
2247 2248 2249 2250 2251
		return CHARGE_OOM_DIE;

	return CHARGE_RETRY;
}

2252
/*
2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271
 * __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.
2272
 */
2273
static int __mem_cgroup_try_charge(struct mm_struct *mm,
A
Andrea Arcangeli 已提交
2274
				   gfp_t gfp_mask,
2275
				   unsigned int nr_pages,
2276
				   struct mem_cgroup **ptr,
2277
				   bool oom)
2278
{
2279
	unsigned int batch = max(CHARGE_BATCH, nr_pages);
2280
	int nr_oom_retries = MEM_CGROUP_RECLAIM_RETRIES;
2281
	struct mem_cgroup *memcg = NULL;
2282
	int ret;
2283

K
KAMEZAWA Hiroyuki 已提交
2284 2285 2286 2287 2288 2289 2290 2291
	/*
	 * 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;
2292

2293
	/*
2294 2295
	 * We always charge the cgroup the mm_struct belongs to.
	 * The mm_struct's mem_cgroup changes on task migration if the
2296 2297 2298
	 * thread group leader migrates. It's possible that mm is not
	 * set, if so charge the init_mm (happens for pagecache usage).
	 */
2299
	if (!*ptr && !mm)
2300
		*ptr = root_mem_cgroup;
K
KAMEZAWA Hiroyuki 已提交
2301
again:
2302 2303 2304 2305
	if (*ptr) { /* css should be a valid one */
		memcg = *ptr;
		VM_BUG_ON(css_is_removed(&memcg->css));
		if (mem_cgroup_is_root(memcg))
K
KAMEZAWA Hiroyuki 已提交
2306
			goto done;
2307
		if (nr_pages == 1 && consume_stock(memcg))
K
KAMEZAWA Hiroyuki 已提交
2308
			goto done;
2309
		css_get(&memcg->css);
2310
	} else {
K
KAMEZAWA Hiroyuki 已提交
2311
		struct task_struct *p;
2312

K
KAMEZAWA Hiroyuki 已提交
2313 2314 2315
		rcu_read_lock();
		p = rcu_dereference(mm->owner);
		/*
2316
		 * Because we don't have task_lock(), "p" can exit.
2317
		 * In that case, "memcg" can point to root or p can be NULL with
2318 2319 2320 2321 2322 2323
		 * 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
2324
		 */
2325
		memcg = mem_cgroup_from_task(p);
2326 2327 2328
		if (!memcg)
			memcg = root_mem_cgroup;
		if (mem_cgroup_is_root(memcg)) {
K
KAMEZAWA Hiroyuki 已提交
2329 2330 2331
			rcu_read_unlock();
			goto done;
		}
2332
		if (nr_pages == 1 && consume_stock(memcg)) {
K
KAMEZAWA Hiroyuki 已提交
2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344
			/*
			 * 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 */
2345
		if (!css_tryget(&memcg->css)) {
K
KAMEZAWA Hiroyuki 已提交
2346 2347 2348 2349 2350
			rcu_read_unlock();
			goto again;
		}
		rcu_read_unlock();
	}
2351

2352 2353
	do {
		bool oom_check;
2354

2355
		/* If killed, bypass charge */
K
KAMEZAWA Hiroyuki 已提交
2356
		if (fatal_signal_pending(current)) {
2357
			css_put(&memcg->css);
2358
			goto bypass;
K
KAMEZAWA Hiroyuki 已提交
2359
		}
2360

2361 2362 2363 2364
		oom_check = false;
		if (oom && !nr_oom_retries) {
			oom_check = true;
			nr_oom_retries = MEM_CGROUP_RECLAIM_RETRIES;
2365
		}
2366

2367
		ret = mem_cgroup_do_charge(memcg, gfp_mask, batch, oom_check);
2368 2369 2370 2371
		switch (ret) {
		case CHARGE_OK:
			break;
		case CHARGE_RETRY: /* not in OOM situation but retry */
2372
			batch = nr_pages;
2373 2374
			css_put(&memcg->css);
			memcg = NULL;
K
KAMEZAWA Hiroyuki 已提交
2375
			goto again;
2376
		case CHARGE_WOULDBLOCK: /* !__GFP_WAIT */
2377
			css_put(&memcg->css);
2378 2379
			goto nomem;
		case CHARGE_NOMEM: /* OOM routine works */
K
KAMEZAWA Hiroyuki 已提交
2380
			if (!oom) {
2381
				css_put(&memcg->css);
K
KAMEZAWA Hiroyuki 已提交
2382
				goto nomem;
K
KAMEZAWA Hiroyuki 已提交
2383
			}
2384 2385 2386 2387
			/* If oom, we never return -ENOMEM */
			nr_oom_retries--;
			break;
		case CHARGE_OOM_DIE: /* Killed by OOM Killer */
2388
			css_put(&memcg->css);
K
KAMEZAWA Hiroyuki 已提交
2389
			goto bypass;
2390
		}
2391 2392
	} while (ret != CHARGE_OK);

2393
	if (batch > nr_pages)
2394 2395
		refill_stock(memcg, batch - nr_pages);
	css_put(&memcg->css);
2396
done:
2397
	*ptr = memcg;
2398 2399
	return 0;
nomem:
2400
	*ptr = NULL;
2401
	return -ENOMEM;
K
KAMEZAWA Hiroyuki 已提交
2402
bypass:
2403 2404
	*ptr = root_mem_cgroup;
	return -EINTR;
2405
}
2406

2407 2408 2409 2410 2411
/*
 * 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().
 */
2412
static void __mem_cgroup_cancel_charge(struct mem_cgroup *memcg,
2413
				       unsigned int nr_pages)
2414
{
2415
	if (!mem_cgroup_is_root(memcg)) {
2416 2417
		unsigned long bytes = nr_pages * PAGE_SIZE;

2418
		res_counter_uncharge(&memcg->res, bytes);
2419
		if (do_swap_account)
2420
			res_counter_uncharge(&memcg->memsw, bytes);
2421
	}
2422 2423
}

2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442
/*
 * A helper function to get mem_cgroup from ID. must be called under
 * rcu_read_lock(). The caller must check css_is_removed() or some if
 * it's concern. (dropping refcnt from swap can be called against removed
 * memcg.)
 */
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;
	return container_of(css, struct mem_cgroup, css);
}

2443
struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page)
2444
{
2445
	struct mem_cgroup *memcg = NULL;
2446
	struct page_cgroup *pc;
2447
	unsigned short id;
2448 2449
	swp_entry_t ent;

2450 2451 2452
	VM_BUG_ON(!PageLocked(page));

	pc = lookup_page_cgroup(page);
2453
	lock_page_cgroup(pc);
2454
	if (PageCgroupUsed(pc)) {
2455 2456 2457
		memcg = pc->mem_cgroup;
		if (memcg && !css_tryget(&memcg->css))
			memcg = NULL;
2458
	} else if (PageSwapCache(page)) {
2459
		ent.val = page_private(page);
2460
		id = lookup_swap_cgroup_id(ent);
2461
		rcu_read_lock();
2462 2463 2464
		memcg = mem_cgroup_lookup(id);
		if (memcg && !css_tryget(&memcg->css))
			memcg = NULL;
2465
		rcu_read_unlock();
2466
	}
2467
	unlock_page_cgroup(pc);
2468
	return memcg;
2469 2470
}

2471
static void __mem_cgroup_commit_charge(struct mem_cgroup *memcg,
2472
				       struct page *page,
2473
				       unsigned int nr_pages,
2474 2475
				       enum charge_type ctype,
				       bool lrucare)
2476
{
2477
	struct page_cgroup *pc = lookup_page_cgroup(page);
2478 2479
	struct zone *uninitialized_var(zone);
	bool was_on_lru = false;
2480
	bool anon;
2481

2482 2483 2484
	lock_page_cgroup(pc);
	if (unlikely(PageCgroupUsed(pc))) {
		unlock_page_cgroup(pc);
2485
		__mem_cgroup_cancel_charge(memcg, nr_pages);
2486 2487 2488 2489 2490 2491
		return;
	}
	/*
	 * we don't need page_cgroup_lock about tail pages, becase they are not
	 * accessed by any other context at this point.
	 */
2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506

	/*
	 * 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)) {
			ClearPageLRU(page);
			del_page_from_lru_list(zone, page, page_lru(page));
			was_on_lru = true;
		}
	}

2507
	pc->mem_cgroup = memcg;
2508 2509 2510 2511 2512 2513 2514
	/*
	 * 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
2515
	smp_wmb();
2516
	SetPageCgroupUsed(pc);
2517

2518 2519 2520 2521 2522 2523 2524 2525 2526
	if (lrucare) {
		if (was_on_lru) {
			VM_BUG_ON(PageLRU(page));
			SetPageLRU(page);
			add_page_to_lru_list(zone, page, page_lru(page));
		}
		spin_unlock_irq(&zone->lru_lock);
	}

2527 2528 2529 2530 2531 2532
	if (ctype == MEM_CGROUP_CHARGE_TYPE_MAPPED)
		anon = true;
	else
		anon = false;

	mem_cgroup_charge_statistics(memcg, anon, nr_pages);
2533
	unlock_page_cgroup(pc);
2534

2535 2536 2537 2538 2539
	/*
	 * "charge_statistics" updated event counter. Then, check it.
	 * Insert ancestor (and ancestor's ancestors), to softlimit RB-tree.
	 * if they exceeds softlimit.
	 */
2540
	memcg_check_events(memcg, page);
2541
}
2542

2543 2544
#ifdef CONFIG_TRANSPARENT_HUGEPAGE

2545
#define PCGF_NOCOPY_AT_SPLIT ((1 << PCG_LOCK) | (1 << PCG_MIGRATION))
2546 2547
/*
 * Because tail pages are not marked as "used", set it. We're under
2548 2549 2550
 * 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.
2551
 */
2552
void mem_cgroup_split_huge_fixup(struct page *head)
2553 2554
{
	struct page_cgroup *head_pc = lookup_page_cgroup(head);
2555 2556
	struct page_cgroup *pc;
	int i;
2557

2558 2559
	if (mem_cgroup_disabled())
		return;
2560 2561 2562 2563 2564 2565
	for (i = 1; i < HPAGE_PMD_NR; i++) {
		pc = head_pc + i;
		pc->mem_cgroup = head_pc->mem_cgroup;
		smp_wmb();/* see __commit_charge() */
		pc->flags = head_pc->flags & ~PCGF_NOCOPY_AT_SPLIT;
	}
2566
}
2567
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
2568

2569
/**
2570
 * mem_cgroup_move_account - move account of the page
2571
 * @page: the page
2572
 * @nr_pages: number of regular pages (>1 for huge pages)
2573 2574 2575
 * @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.
2576
 * @uncharge: whether we should call uncharge and css_put against @from.
2577 2578
 *
 * The caller must confirm following.
K
KAMEZAWA Hiroyuki 已提交
2579
 * - page is not on LRU (isolate_page() is useful.)
2580
 * - compound_lock is held when nr_pages > 1
2581
 *
2582
 * This function doesn't do "charge" nor css_get to new cgroup. It should be
L
Lucas De Marchi 已提交
2583
 * done by a caller(__mem_cgroup_try_charge would be useful). If @uncharge is
2584 2585
 * true, this function does "uncharge" from old cgroup, but it doesn't if
 * @uncharge is false, so a caller should do "uncharge".
2586
 */
2587 2588 2589 2590 2591 2592
static int mem_cgroup_move_account(struct page *page,
				   unsigned int nr_pages,
				   struct page_cgroup *pc,
				   struct mem_cgroup *from,
				   struct mem_cgroup *to,
				   bool uncharge)
2593
{
2594 2595
	unsigned long flags;
	int ret;
2596
	bool anon = PageAnon(page);
2597

2598
	VM_BUG_ON(from == to);
2599
	VM_BUG_ON(PageLRU(page));
2600 2601 2602 2603 2604 2605 2606
	/*
	 * 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;
2607
	if (nr_pages > 1 && !PageTransHuge(page))
2608 2609 2610 2611 2612 2613 2614 2615
		goto out;

	lock_page_cgroup(pc);

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

2616
	move_lock_mem_cgroup(from, &flags);
2617

2618
	if (!anon && page_mapped(page)) {
2619 2620 2621 2622 2623
		/* 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();
2624
	}
2625
	mem_cgroup_charge_statistics(from, anon, -nr_pages);
2626 2627
	if (uncharge)
		/* This is not "cancel", but cancel_charge does all we need. */
2628
		__mem_cgroup_cancel_charge(from, nr_pages);
2629

2630
	/* caller should have done css_get */
K
KAMEZAWA Hiroyuki 已提交
2631
	pc->mem_cgroup = to;
2632
	mem_cgroup_charge_statistics(to, anon, nr_pages);
2633 2634 2635
	/*
	 * We charges against "to" which may not have any tasks. Then, "to"
	 * can be under rmdir(). But in current implementation, caller of
2636
	 * this function is just force_empty() and move charge, so it's
L
Lucas De Marchi 已提交
2637
	 * guaranteed that "to" is never removed. So, we don't check rmdir
2638
	 * status here.
2639
	 */
2640
	move_unlock_mem_cgroup(from, &flags);
2641 2642
	ret = 0;
unlock:
2643
	unlock_page_cgroup(pc);
2644 2645 2646
	/*
	 * check events
	 */
2647 2648
	memcg_check_events(to, page);
	memcg_check_events(from, page);
2649
out:
2650 2651 2652 2653 2654 2655 2656
	return ret;
}

/*
 * move charges to its parent.
 */

2657 2658
static int mem_cgroup_move_parent(struct page *page,
				  struct page_cgroup *pc,
2659 2660 2661 2662 2663 2664
				  struct mem_cgroup *child,
				  gfp_t gfp_mask)
{
	struct cgroup *cg = child->css.cgroup;
	struct cgroup *pcg = cg->parent;
	struct mem_cgroup *parent;
2665
	unsigned int nr_pages;
2666
	unsigned long uninitialized_var(flags);
2667 2668 2669 2670 2671 2672
	int ret;

	/* Is ROOT ? */
	if (!pcg)
		return -EINVAL;

2673 2674 2675 2676 2677
	ret = -EBUSY;
	if (!get_page_unless_zero(page))
		goto out;
	if (isolate_lru_page(page))
		goto put;
2678

2679
	nr_pages = hpage_nr_pages(page);
K
KAMEZAWA Hiroyuki 已提交
2680

2681
	parent = mem_cgroup_from_cont(pcg);
2682
	ret = __mem_cgroup_try_charge(NULL, gfp_mask, nr_pages, &parent, false);
2683
	if (ret)
2684
		goto put_back;
2685

2686
	if (nr_pages > 1)
2687 2688
		flags = compound_lock_irqsave(page);

2689
	ret = mem_cgroup_move_account(page, nr_pages, pc, child, parent, true);
2690
	if (ret)
2691
		__mem_cgroup_cancel_charge(parent, nr_pages);
2692

2693
	if (nr_pages > 1)
2694
		compound_unlock_irqrestore(page, flags);
2695
put_back:
K
KAMEZAWA Hiroyuki 已提交
2696
	putback_lru_page(page);
2697
put:
2698
	put_page(page);
2699
out:
2700 2701 2702
	return ret;
}

2703 2704 2705 2706 2707 2708 2709
/*
 * 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,
2710
				gfp_t gfp_mask, enum charge_type ctype)
2711
{
2712
	struct mem_cgroup *memcg = NULL;
2713
	unsigned int nr_pages = 1;
2714
	bool oom = true;
2715
	int ret;
A
Andrea Arcangeli 已提交
2716

A
Andrea Arcangeli 已提交
2717
	if (PageTransHuge(page)) {
2718
		nr_pages <<= compound_order(page);
A
Andrea Arcangeli 已提交
2719
		VM_BUG_ON(!PageTransHuge(page));
2720 2721 2722 2723 2724
		/*
		 * 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
2725
	}
2726

2727
	ret = __mem_cgroup_try_charge(mm, gfp_mask, nr_pages, &memcg, oom);
2728
	if (ret == -ENOMEM)
2729
		return ret;
2730
	__mem_cgroup_commit_charge(memcg, page, nr_pages, ctype, false);
2731 2732 2733
	return 0;
}

2734 2735
int mem_cgroup_newpage_charge(struct page *page,
			      struct mm_struct *mm, gfp_t gfp_mask)
2736
{
2737
	if (mem_cgroup_disabled())
2738
		return 0;
2739 2740 2741
	VM_BUG_ON(page_mapped(page));
	VM_BUG_ON(page->mapping && !PageAnon(page));
	VM_BUG_ON(!mm);
2742
	return mem_cgroup_charge_common(page, mm, gfp_mask,
2743
					MEM_CGROUP_CHARGE_TYPE_MAPPED);
2744 2745
}

D
Daisuke Nishimura 已提交
2746 2747 2748 2749
static void
__mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr,
					enum charge_type ctype);

2750 2751
int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm,
				gfp_t gfp_mask)
2752
{
2753
	struct mem_cgroup *memcg = NULL;
2754
	enum charge_type type = MEM_CGROUP_CHARGE_TYPE_CACHE;
2755 2756
	int ret;

2757
	if (mem_cgroup_disabled())
2758
		return 0;
2759 2760
	if (PageCompound(page))
		return 0;
2761

2762
	if (unlikely(!mm))
2763
		mm = &init_mm;
2764 2765
	if (!page_is_file_cache(page))
		type = MEM_CGROUP_CHARGE_TYPE_SHMEM;
2766

2767
	if (!PageSwapCache(page))
2768
		ret = mem_cgroup_charge_common(page, mm, gfp_mask, type);
2769
	else { /* page is swapcache/shmem */
2770
		ret = mem_cgroup_try_charge_swapin(mm, page, gfp_mask, &memcg);
D
Daisuke Nishimura 已提交
2771
		if (!ret)
2772 2773
			__mem_cgroup_commit_charge_swapin(page, memcg, type);
	}
2774
	return ret;
2775 2776
}

2777 2778 2779
/*
 * While swap-in, try_charge -> commit or cancel, the page is locked.
 * And when try_charge() successfully returns, one refcnt to memcg without
2780
 * struct page_cgroup is acquired. This refcnt will be consumed by
2781 2782
 * "commit()" or removed by "cancel()"
 */
2783 2784
int mem_cgroup_try_charge_swapin(struct mm_struct *mm,
				 struct page *page,
2785
				 gfp_t mask, struct mem_cgroup **memcgp)
2786
{
2787
	struct mem_cgroup *memcg;
2788
	int ret;
2789

2790
	*memcgp = NULL;
2791

2792
	if (mem_cgroup_disabled())
2793 2794 2795 2796 2797 2798
		return 0;

	if (!do_swap_account)
		goto charge_cur_mm;
	/*
	 * A racing thread's fault, or swapoff, may have already updated
H
Hugh Dickins 已提交
2799 2800 2801
	 * the pte, and even removed page from swap cache: in those cases
	 * do_swap_page()'s pte_same() test will fail; but there's also a
	 * KSM case which does need to charge the page.
2802 2803
	 */
	if (!PageSwapCache(page))
H
Hugh Dickins 已提交
2804
		goto charge_cur_mm;
2805 2806
	memcg = try_get_mem_cgroup_from_page(page);
	if (!memcg)
2807
		goto charge_cur_mm;
2808 2809
	*memcgp = memcg;
	ret = __mem_cgroup_try_charge(NULL, mask, 1, memcgp, true);
2810
	css_put(&memcg->css);
2811 2812
	if (ret == -EINTR)
		ret = 0;
2813
	return ret;
2814 2815 2816
charge_cur_mm:
	if (unlikely(!mm))
		mm = &init_mm;
2817 2818 2819 2820
	ret = __mem_cgroup_try_charge(mm, mask, 1, memcgp, true);
	if (ret == -EINTR)
		ret = 0;
	return ret;
2821 2822
}

D
Daisuke Nishimura 已提交
2823
static void
2824
__mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *memcg,
D
Daisuke Nishimura 已提交
2825
					enum charge_type ctype)
2826
{
2827
	if (mem_cgroup_disabled())
2828
		return;
2829
	if (!memcg)
2830
		return;
2831
	cgroup_exclude_rmdir(&memcg->css);
2832

2833
	__mem_cgroup_commit_charge(memcg, page, 1, ctype, true);
2834 2835 2836
	/*
	 * Now swap is on-memory. This means this page may be
	 * counted both as mem and swap....double count.
2837 2838 2839
	 * 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.
2840
	 */
2841
	if (do_swap_account && PageSwapCache(page)) {
2842
		swp_entry_t ent = {.val = page_private(page)};
2843
		mem_cgroup_uncharge_swap(ent);
2844
	}
2845 2846 2847 2848 2849
	/*
	 * At swapin, we may charge account against cgroup which has no tasks.
	 * So, rmdir()->pre_destroy() can be called while we do this charge.
	 * In that case, we need to call pre_destroy() again. check it here.
	 */
2850
	cgroup_release_and_wakeup_rmdir(&memcg->css);
2851 2852
}

2853 2854
void mem_cgroup_commit_charge_swapin(struct page *page,
				     struct mem_cgroup *memcg)
D
Daisuke Nishimura 已提交
2855
{
2856 2857
	__mem_cgroup_commit_charge_swapin(page, memcg,
					  MEM_CGROUP_CHARGE_TYPE_MAPPED);
D
Daisuke Nishimura 已提交
2858 2859
}

2860
void mem_cgroup_cancel_charge_swapin(struct mem_cgroup *memcg)
2861
{
2862
	if (mem_cgroup_disabled())
2863
		return;
2864
	if (!memcg)
2865
		return;
2866
	__mem_cgroup_cancel_charge(memcg, 1);
2867 2868
}

2869
static void mem_cgroup_do_uncharge(struct mem_cgroup *memcg,
2870 2871
				   unsigned int nr_pages,
				   const enum charge_type ctype)
2872 2873 2874
{
	struct memcg_batch_info *batch = NULL;
	bool uncharge_memsw = true;
2875

2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886
	/* 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)
2887
		batch->memcg = memcg;
2888 2889
	/*
	 * do_batch > 0 when unmapping pages or inode invalidate/truncate.
L
Lucas De Marchi 已提交
2890
	 * In those cases, all pages freed continuously can be expected to be in
2891 2892 2893 2894 2895 2896 2897 2898
	 * 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;

2899
	if (nr_pages > 1)
A
Andrea Arcangeli 已提交
2900 2901
		goto direct_uncharge;

2902 2903 2904 2905 2906
	/*
	 * 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.
	 */
2907
	if (batch->memcg != memcg)
2908 2909
		goto direct_uncharge;
	/* remember freed charge and uncharge it later */
2910
	batch->nr_pages++;
2911
	if (uncharge_memsw)
2912
		batch->memsw_nr_pages++;
2913 2914
	return;
direct_uncharge:
2915
	res_counter_uncharge(&memcg->res, nr_pages * PAGE_SIZE);
2916
	if (uncharge_memsw)
2917 2918 2919
		res_counter_uncharge(&memcg->memsw, nr_pages * PAGE_SIZE);
	if (unlikely(batch->memcg != memcg))
		memcg_oom_recover(memcg);
2920
}
2921

2922
/*
2923
 * uncharge if !page_mapped(page)
2924
 */
2925
static struct mem_cgroup *
2926
__mem_cgroup_uncharge_common(struct page *page, enum charge_type ctype)
2927
{
2928
	struct mem_cgroup *memcg = NULL;
2929 2930
	unsigned int nr_pages = 1;
	struct page_cgroup *pc;
2931
	bool anon;
2932

2933
	if (mem_cgroup_disabled())
2934
		return NULL;
2935

K
KAMEZAWA Hiroyuki 已提交
2936
	if (PageSwapCache(page))
2937
		return NULL;
K
KAMEZAWA Hiroyuki 已提交
2938

A
Andrea Arcangeli 已提交
2939
	if (PageTransHuge(page)) {
2940
		nr_pages <<= compound_order(page);
A
Andrea Arcangeli 已提交
2941 2942
		VM_BUG_ON(!PageTransHuge(page));
	}
2943
	/*
2944
	 * Check if our page_cgroup is valid
2945
	 */
2946
	pc = lookup_page_cgroup(page);
2947
	if (unlikely(!PageCgroupUsed(pc)))
2948
		return NULL;
2949

2950
	lock_page_cgroup(pc);
K
KAMEZAWA Hiroyuki 已提交
2951

2952
	memcg = pc->mem_cgroup;
2953

K
KAMEZAWA Hiroyuki 已提交
2954 2955 2956
	if (!PageCgroupUsed(pc))
		goto unlock_out;

2957 2958
	anon = PageAnon(page);

K
KAMEZAWA Hiroyuki 已提交
2959 2960
	switch (ctype) {
	case MEM_CGROUP_CHARGE_TYPE_MAPPED:
2961 2962 2963 2964 2965
		/*
		 * 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.
		 */
2966 2967
		anon = true;
		/* fallthrough */
K
KAMEZAWA Hiroyuki 已提交
2968
	case MEM_CGROUP_CHARGE_TYPE_DROP:
2969 2970
		/* See mem_cgroup_prepare_migration() */
		if (page_mapped(page) || PageCgroupMigration(pc))
K
KAMEZAWA Hiroyuki 已提交
2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981
			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;
2982
	}
K
KAMEZAWA Hiroyuki 已提交
2983

2984
	mem_cgroup_charge_statistics(memcg, anon, -nr_pages);
K
KAMEZAWA Hiroyuki 已提交
2985

2986
	ClearPageCgroupUsed(pc);
2987 2988 2989 2990 2991 2992
	/*
	 * 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.
	 */
2993

2994
	unlock_page_cgroup(pc);
K
KAMEZAWA Hiroyuki 已提交
2995
	/*
2996
	 * even after unlock, we have memcg->res.usage here and this memcg
K
KAMEZAWA Hiroyuki 已提交
2997 2998
	 * will never be freed.
	 */
2999
	memcg_check_events(memcg, page);
K
KAMEZAWA Hiroyuki 已提交
3000
	if (do_swap_account && ctype == MEM_CGROUP_CHARGE_TYPE_SWAPOUT) {
3001 3002
		mem_cgroup_swap_statistics(memcg, true);
		mem_cgroup_get(memcg);
K
KAMEZAWA Hiroyuki 已提交
3003
	}
3004 3005
	if (!mem_cgroup_is_root(memcg))
		mem_cgroup_do_uncharge(memcg, nr_pages, ctype);
3006

3007
	return memcg;
K
KAMEZAWA Hiroyuki 已提交
3008 3009 3010

unlock_out:
	unlock_page_cgroup(pc);
3011
	return NULL;
3012 3013
}

3014 3015
void mem_cgroup_uncharge_page(struct page *page)
{
3016 3017 3018
	/* early check. */
	if (page_mapped(page))
		return;
3019
	VM_BUG_ON(page->mapping && !PageAnon(page));
3020 3021 3022 3023 3024 3025
	__mem_cgroup_uncharge_common(page, MEM_CGROUP_CHARGE_TYPE_MAPPED);
}

void mem_cgroup_uncharge_cache_page(struct page *page)
{
	VM_BUG_ON(page_mapped(page));
3026
	VM_BUG_ON(page->mapping);
3027 3028 3029
	__mem_cgroup_uncharge_common(page, MEM_CGROUP_CHARGE_TYPE_CACHE);
}

3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043
/*
 * 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;
3044 3045
		current->memcg_batch.nr_pages = 0;
		current->memcg_batch.memsw_nr_pages = 0;
3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065
	}
}

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.
	 */
3066 3067 3068 3069 3070 3071
	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);
3072
	memcg_oom_recover(batch->memcg);
3073 3074 3075 3076
	/* forget this pointer (for sanity check) */
	batch->memcg = NULL;
}

3077
#ifdef CONFIG_SWAP
3078
/*
3079
 * called after __delete_from_swap_cache() and drop "page" account.
3080 3081
 * memcg information is recorded to swap_cgroup of "ent"
 */
K
KAMEZAWA Hiroyuki 已提交
3082 3083
void
mem_cgroup_uncharge_swapcache(struct page *page, swp_entry_t ent, bool swapout)
3084 3085
{
	struct mem_cgroup *memcg;
K
KAMEZAWA Hiroyuki 已提交
3086 3087 3088 3089 3090 3091
	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;

	memcg = __mem_cgroup_uncharge_common(page, ctype);
3092

K
KAMEZAWA Hiroyuki 已提交
3093 3094 3095 3096 3097
	/*
	 * record memcg information,  if swapout && memcg != NULL,
	 * mem_cgroup_get() was called in uncharge().
	 */
	if (do_swap_account && swapout && memcg)
3098
		swap_cgroup_record(ent, css_id(&memcg->css));
3099
}
3100
#endif
3101 3102 3103 3104 3105 3106 3107

#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
/*
 * called from swap_entry_free(). remove record in swap_cgroup and
 * uncharge "memsw" account.
 */
void mem_cgroup_uncharge_swap(swp_entry_t ent)
K
KAMEZAWA Hiroyuki 已提交
3108
{
3109
	struct mem_cgroup *memcg;
3110
	unsigned short id;
3111 3112 3113 3114

	if (!do_swap_account)
		return;

3115 3116 3117
	id = swap_cgroup_record(ent, 0);
	rcu_read_lock();
	memcg = mem_cgroup_lookup(id);
3118
	if (memcg) {
3119 3120 3121 3122
		/*
		 * We uncharge this because swap is freed.
		 * This memcg can be obsolete one. We avoid calling css_tryget
		 */
3123
		if (!mem_cgroup_is_root(memcg))
3124
			res_counter_uncharge(&memcg->memsw, PAGE_SIZE);
3125
		mem_cgroup_swap_statistics(memcg, false);
3126 3127
		mem_cgroup_put(memcg);
	}
3128
	rcu_read_unlock();
K
KAMEZAWA Hiroyuki 已提交
3129
}
3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145

/**
 * 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,
3146
				struct mem_cgroup *from, struct mem_cgroup *to)
3147 3148 3149 3150 3151 3152 3153 3154
{
	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);
3155
		mem_cgroup_swap_statistics(to, true);
3156
		/*
3157 3158 3159 3160 3161 3162
		 * 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
		 * improvement. But we cannot postpone mem_cgroup_get(to)
		 * because if the process that has been moved to @to does
		 * swap-in, the refcount of @to might be decreased to 0.
3163 3164 3165 3166 3167 3168 3169 3170
		 */
		mem_cgroup_get(to);
		return 0;
	}
	return -EINVAL;
}
#else
static inline int mem_cgroup_move_swap_account(swp_entry_t entry,
3171
				struct mem_cgroup *from, struct mem_cgroup *to)
3172 3173 3174
{
	return -EINVAL;
}
3175
#endif
K
KAMEZAWA Hiroyuki 已提交
3176

3177
/*
3178 3179
 * Before starting migration, account PAGE_SIZE to mem_cgroup that the old
 * page belongs to.
3180
 */
3181
int mem_cgroup_prepare_migration(struct page *page,
3182
	struct page *newpage, struct mem_cgroup **memcgp, gfp_t gfp_mask)
3183
{
3184
	struct mem_cgroup *memcg = NULL;
3185
	struct page_cgroup *pc;
3186
	enum charge_type ctype;
3187
	int ret = 0;
3188

3189
	*memcgp = NULL;
3190

A
Andrea Arcangeli 已提交
3191
	VM_BUG_ON(PageTransHuge(page));
3192
	if (mem_cgroup_disabled())
3193 3194
		return 0;

3195 3196 3197
	pc = lookup_page_cgroup(page);
	lock_page_cgroup(pc);
	if (PageCgroupUsed(pc)) {
3198 3199
		memcg = pc->mem_cgroup;
		css_get(&memcg->css);
3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230
		/*
		 * 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);
3231
	}
3232
	unlock_page_cgroup(pc);
3233 3234 3235 3236
	/*
	 * If the page is not charged at this point,
	 * we return here.
	 */
3237
	if (!memcg)
3238
		return 0;
3239

3240 3241
	*memcgp = memcg;
	ret = __mem_cgroup_try_charge(NULL, gfp_mask, 1, memcgp, false);
3242
	css_put(&memcg->css);/* drop extra refcnt */
3243
	if (ret) {
3244 3245 3246 3247 3248 3249 3250 3251 3252
		if (PageAnon(page)) {
			lock_page_cgroup(pc);
			ClearPageCgroupMigration(pc);
			unlock_page_cgroup(pc);
			/*
			 * The old page may be fully unmapped while we kept it.
			 */
			mem_cgroup_uncharge_page(page);
		}
3253
		/* we'll need to revisit this error code (we have -EINTR) */
3254
		return -ENOMEM;
3255
	}
3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267
	/*
	 * 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))
		ctype = MEM_CGROUP_CHARGE_TYPE_MAPPED;
	else if (page_is_file_cache(page))
		ctype = MEM_CGROUP_CHARGE_TYPE_CACHE;
	else
		ctype = MEM_CGROUP_CHARGE_TYPE_SHMEM;
3268
	__mem_cgroup_commit_charge(memcg, newpage, 1, ctype, false);
3269
	return ret;
3270
}
3271

3272
/* remove redundant charge if migration failed*/
3273
void mem_cgroup_end_migration(struct mem_cgroup *memcg,
3274
	struct page *oldpage, struct page *newpage, bool migration_ok)
3275
{
3276
	struct page *used, *unused;
3277
	struct page_cgroup *pc;
3278
	bool anon;
3279

3280
	if (!memcg)
3281
		return;
3282
	/* blocks rmdir() */
3283
	cgroup_exclude_rmdir(&memcg->css);
3284
	if (!migration_ok) {
3285 3286
		used = oldpage;
		unused = newpage;
3287
	} else {
3288
		used = newpage;
3289 3290
		unused = oldpage;
	}
3291
	/*
3292 3293 3294
	 * 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.
3295
	 */
3296 3297 3298 3299
	pc = lookup_page_cgroup(oldpage);
	lock_page_cgroup(pc);
	ClearPageCgroupMigration(pc);
	unlock_page_cgroup(pc);
3300 3301 3302 3303
	anon = PageAnon(used);
	__mem_cgroup_uncharge_common(unused,
		anon ? MEM_CGROUP_CHARGE_TYPE_MAPPED
		     : MEM_CGROUP_CHARGE_TYPE_CACHE);
3304

3305
	/*
3306 3307 3308 3309 3310 3311
	 * 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)
3312
	 */
3313
	if (anon)
3314
		mem_cgroup_uncharge_page(used);
3315
	/*
3316 3317
	 * At migration, we may charge account against cgroup which has no
	 * tasks.
3318 3319 3320
	 * So, rmdir()->pre_destroy() can be called while we do this charge.
	 * In that case, we need to call pre_destroy() again. check it here.
	 */
3321
	cgroup_release_and_wakeup_rmdir(&memcg->css);
3322
}
3323

3324 3325 3326 3327 3328 3329 3330 3331
/*
 * 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)
{
3332
	struct mem_cgroup *memcg = NULL;
3333 3334 3335 3336 3337 3338 3339 3340 3341
	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);
3342 3343 3344 3345 3346
	if (PageCgroupUsed(pc)) {
		memcg = pc->mem_cgroup;
		mem_cgroup_charge_statistics(memcg, false, -1);
		ClearPageCgroupUsed(pc);
	}
3347 3348
	unlock_page_cgroup(pc);

3349 3350 3351 3352 3353 3354 3355
	/*
	 * When called from shmem_replace_page(), in some cases the
	 * oldpage has already been charged, and in some cases not.
	 */
	if (!memcg)
		return;

3356 3357 3358 3359 3360 3361 3362 3363
	if (PageSwapBacked(oldpage))
		type = MEM_CGROUP_CHARGE_TYPE_SHMEM;

	/*
	 * 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.
	 */
3364
	__mem_cgroup_commit_charge(memcg, newpage, 1, type, true);
3365 3366
}

3367 3368 3369 3370 3371 3372
#ifdef CONFIG_DEBUG_VM
static struct page_cgroup *lookup_page_cgroup_used(struct page *page)
{
	struct page_cgroup *pc;

	pc = lookup_page_cgroup(page);
3373 3374 3375 3376 3377
	/*
	 * 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().
	 */
3378 3379 3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396
	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) {
3397
		printk(KERN_ALERT "pc:%p pc->flags:%lx pc->mem_cgroup:%p\n",
3398 3399 3400 3401 3402
		       pc, pc->flags, pc->mem_cgroup);
	}
}
#endif

3403 3404
static DEFINE_MUTEX(set_limit_mutex);

3405
static int mem_cgroup_resize_limit(struct mem_cgroup *memcg,
3406
				unsigned long long val)
3407
{
3408
	int retry_count;
3409
	u64 memswlimit, memlimit;
3410
	int ret = 0;
3411 3412
	int children = mem_cgroup_count_children(memcg);
	u64 curusage, oldusage;
3413
	int enlarge;
3414 3415 3416 3417 3418 3419 3420 3421 3422

	/*
	 * 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);
3423

3424
	enlarge = 0;
3425
	while (retry_count) {
3426 3427 3428 3429
		if (signal_pending(current)) {
			ret = -EINTR;
			break;
		}
3430 3431 3432
		/*
		 * Rather than hide all in some function, I do this in
		 * open coded manner. You see what this really does.
3433
		 * We have to guarantee memcg->res.limit < memcg->memsw.limit.
3434 3435 3436 3437 3438 3439
		 */
		mutex_lock(&set_limit_mutex);
		memswlimit = res_counter_read_u64(&memcg->memsw, RES_LIMIT);
		if (memswlimit < val) {
			ret = -EINVAL;
			mutex_unlock(&set_limit_mutex);
3440 3441
			break;
		}
3442 3443 3444 3445 3446

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

3447
		ret = res_counter_set_limit(&memcg->res, val);
3448 3449 3450 3451 3452 3453
		if (!ret) {
			if (memswlimit == val)
				memcg->memsw_is_minimum = true;
			else
				memcg->memsw_is_minimum = false;
		}
3454 3455 3456 3457 3458
		mutex_unlock(&set_limit_mutex);

		if (!ret)
			break;

3459 3460
		mem_cgroup_reclaim(memcg, GFP_KERNEL,
				   MEM_CGROUP_RECLAIM_SHRINK);
3461 3462 3463 3464 3465 3466
		curusage = res_counter_read_u64(&memcg->res, RES_USAGE);
		/* Usage is reduced ? */
  		if (curusage >= oldusage)
			retry_count--;
		else
			oldusage = curusage;
3467
	}
3468 3469
	if (!ret && enlarge)
		memcg_oom_recover(memcg);
3470

3471 3472 3473
	return ret;
}

L
Li Zefan 已提交
3474 3475
static int mem_cgroup_resize_memsw_limit(struct mem_cgroup *memcg,
					unsigned long long val)
3476
{
3477
	int retry_count;
3478
	u64 memlimit, memswlimit, oldusage, curusage;
3479 3480
	int children = mem_cgroup_count_children(memcg);
	int ret = -EBUSY;
3481
	int enlarge = 0;
3482

3483 3484 3485
	/* see mem_cgroup_resize_res_limit */
 	retry_count = children * MEM_CGROUP_RECLAIM_RETRIES;
	oldusage = res_counter_read_u64(&memcg->memsw, RES_USAGE);
3486 3487 3488 3489 3490 3491 3492 3493
	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.
3494
		 * We have to guarantee memcg->res.limit < memcg->memsw.limit.
3495 3496 3497 3498 3499 3500 3501 3502
		 */
		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;
		}
3503 3504 3505
		memswlimit = res_counter_read_u64(&memcg->memsw, RES_LIMIT);
		if (memswlimit < val)
			enlarge = 1;
3506
		ret = res_counter_set_limit(&memcg->memsw, val);
3507 3508 3509 3510 3511 3512
		if (!ret) {
			if (memlimit == val)
				memcg->memsw_is_minimum = true;
			else
				memcg->memsw_is_minimum = false;
		}
3513 3514 3515 3516 3517
		mutex_unlock(&set_limit_mutex);

		if (!ret)
			break;

3518 3519 3520
		mem_cgroup_reclaim(memcg, GFP_KERNEL,
				   MEM_CGROUP_RECLAIM_NOSWAP |
				   MEM_CGROUP_RECLAIM_SHRINK);
3521
		curusage = res_counter_read_u64(&memcg->memsw, RES_USAGE);
3522
		/* Usage is reduced ? */
3523
		if (curusage >= oldusage)
3524
			retry_count--;
3525 3526
		else
			oldusage = curusage;
3527
	}
3528 3529
	if (!ret && enlarge)
		memcg_oom_recover(memcg);
3530 3531 3532
	return ret;
}

3533
unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
3534 3535
					    gfp_t gfp_mask,
					    unsigned long *total_scanned)
3536 3537 3538 3539 3540 3541
{
	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;
3542
	unsigned long long excess;
3543
	unsigned long nr_scanned;
3544 3545 3546 3547

	if (order > 0)
		return 0;

3548
	mctz = soft_limit_tree_node_zone(zone_to_nid(zone), zone_idx(zone));
3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561
	/*
	 * 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;

3562
		nr_scanned = 0;
3563
		reclaimed = mem_cgroup_soft_reclaim(mz->memcg, zone,
3564
						    gfp_mask, &nr_scanned);
3565
		nr_reclaimed += reclaimed;
3566
		*total_scanned += nr_scanned;
3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588
		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);
3589
				if (next_mz == mz)
3590
					css_put(&next_mz->memcg->css);
3591
				else /* next_mz == NULL or other memcg */
3592 3593 3594
					break;
			} while (1);
		}
3595 3596
		__mem_cgroup_remove_exceeded(mz->memcg, mz, mctz);
		excess = res_counter_soft_limit_excess(&mz->memcg->res);
3597 3598 3599 3600 3601 3602 3603 3604
		/*
		 * 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.
		 */
3605
		/* If excess == 0, no tree ops */
3606
		__mem_cgroup_insert_exceeded(mz->memcg, mz, mctz, excess);
3607
		spin_unlock(&mctz->lock);
3608
		css_put(&mz->memcg->css);
3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620
		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)
3621
		css_put(&next_mz->memcg->css);
3622 3623 3624
	return nr_reclaimed;
}

3625 3626 3627 3628
/*
 * This routine traverse page_cgroup in given list and drop them all.
 * *And* this routine doesn't reclaim page itself, just removes page_cgroup.
 */
3629
static int mem_cgroup_force_empty_list(struct mem_cgroup *memcg,
K
KAMEZAWA Hiroyuki 已提交
3630
				int node, int zid, enum lru_list lru)
3631
{
K
KAMEZAWA Hiroyuki 已提交
3632 3633
	struct mem_cgroup_per_zone *mz;
	unsigned long flags, loop;
3634
	struct list_head *list;
3635 3636
	struct page *busy;
	struct zone *zone;
3637
	int ret = 0;
3638

K
KAMEZAWA Hiroyuki 已提交
3639
	zone = &NODE_DATA(node)->node_zones[zid];
3640
	mz = mem_cgroup_zoneinfo(memcg, node, zid);
3641
	list = &mz->lruvec.lists[lru];
3642

3643
	loop = mz->lru_size[lru];
3644 3645 3646 3647
	/* give some margin against EBUSY etc...*/
	loop += 256;
	busy = NULL;
	while (loop--) {
3648
		struct page_cgroup *pc;
3649 3650
		struct page *page;

3651
		ret = 0;
K
KAMEZAWA Hiroyuki 已提交
3652
		spin_lock_irqsave(&zone->lru_lock, flags);
3653
		if (list_empty(list)) {
K
KAMEZAWA Hiroyuki 已提交
3654
			spin_unlock_irqrestore(&zone->lru_lock, flags);
3655
			break;
3656
		}
3657 3658 3659
		page = list_entry(list->prev, struct page, lru);
		if (busy == page) {
			list_move(&page->lru, list);
3660
			busy = NULL;
K
KAMEZAWA Hiroyuki 已提交
3661
			spin_unlock_irqrestore(&zone->lru_lock, flags);
3662 3663
			continue;
		}
K
KAMEZAWA Hiroyuki 已提交
3664
		spin_unlock_irqrestore(&zone->lru_lock, flags);
3665

3666
		pc = lookup_page_cgroup(page);
3667

3668
		ret = mem_cgroup_move_parent(page, pc, memcg, GFP_KERNEL);
3669
		if (ret == -ENOMEM || ret == -EINTR)
3670
			break;
3671 3672 3673

		if (ret == -EBUSY || ret == -EINVAL) {
			/* found lock contention or "pc" is obsolete. */
3674
			busy = page;
3675 3676 3677
			cond_resched();
		} else
			busy = NULL;
3678
	}
K
KAMEZAWA Hiroyuki 已提交
3679

3680 3681 3682
	if (!ret && !list_empty(list))
		return -EBUSY;
	return ret;
3683 3684 3685 3686 3687 3688
}

/*
 * make mem_cgroup's charge to be 0 if there is no task.
 * This enables deleting this mem_cgroup.
 */
3689
static int mem_cgroup_force_empty(struct mem_cgroup *memcg, bool free_all)
3690
{
3691 3692 3693
	int ret;
	int node, zid, shrink;
	int nr_retries = MEM_CGROUP_RECLAIM_RETRIES;
3694
	struct cgroup *cgrp = memcg->css.cgroup;
3695

3696
	css_get(&memcg->css);
3697 3698

	shrink = 0;
3699 3700 3701
	/* should free all ? */
	if (free_all)
		goto try_to_free;
3702
move_account:
3703
	do {
3704
		ret = -EBUSY;
3705 3706 3707 3708
		if (cgroup_task_count(cgrp) || !list_empty(&cgrp->children))
			goto out;
		ret = -EINTR;
		if (signal_pending(current))
3709
			goto out;
3710 3711
		/* This is for making all *used* pages to be on LRU. */
		lru_add_drain_all();
3712
		drain_all_stock_sync(memcg);
3713
		ret = 0;
3714
		mem_cgroup_start_move(memcg);
3715
		for_each_node_state(node, N_HIGH_MEMORY) {
3716
			for (zid = 0; !ret && zid < MAX_NR_ZONES; zid++) {
H
Hugh Dickins 已提交
3717 3718
				enum lru_list lru;
				for_each_lru(lru) {
3719
					ret = mem_cgroup_force_empty_list(memcg,
H
Hugh Dickins 已提交
3720
							node, zid, lru);
3721 3722 3723
					if (ret)
						break;
				}
3724
			}
3725 3726 3727
			if (ret)
				break;
		}
3728 3729
		mem_cgroup_end_move(memcg);
		memcg_oom_recover(memcg);
3730 3731 3732
		/* it seems parent cgroup doesn't have enough mem */
		if (ret == -ENOMEM)
			goto try_to_free;
3733
		cond_resched();
3734
	/* "ret" should also be checked to ensure all lists are empty. */
3735
	} while (res_counter_read_u64(&memcg->res, RES_USAGE) > 0 || ret);
3736
out:
3737
	css_put(&memcg->css);
3738
	return ret;
3739 3740

try_to_free:
3741 3742
	/* returns EBUSY if there is a task or if we come here twice. */
	if (cgroup_task_count(cgrp) || !list_empty(&cgrp->children) || shrink) {
3743 3744 3745
		ret = -EBUSY;
		goto out;
	}
3746 3747
	/* we call try-to-free pages for make this cgroup empty */
	lru_add_drain_all();
3748 3749
	/* try to free all pages in this cgroup */
	shrink = 1;
3750
	while (nr_retries && res_counter_read_u64(&memcg->res, RES_USAGE) > 0) {
3751
		int progress;
3752 3753 3754 3755 3756

		if (signal_pending(current)) {
			ret = -EINTR;
			goto out;
		}
3757
		progress = try_to_free_mem_cgroup_pages(memcg, GFP_KERNEL,
3758
						false);
3759
		if (!progress) {
3760
			nr_retries--;
3761
			/* maybe some writeback is necessary */
3762
			congestion_wait(BLK_RW_ASYNC, HZ/10);
3763
		}
3764 3765

	}
K
KAMEZAWA Hiroyuki 已提交
3766
	lru_add_drain();
3767
	/* try move_account...there may be some *locked* pages. */
3768
	goto move_account;
3769 3770
}

3771
static int mem_cgroup_force_empty_write(struct cgroup *cont, unsigned int event)
3772 3773 3774 3775 3776
{
	return mem_cgroup_force_empty(mem_cgroup_from_cont(cont), true);
}


3777 3778 3779 3780 3781 3782 3783 3784 3785
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;
3786
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
3787
	struct cgroup *parent = cont->parent;
3788
	struct mem_cgroup *parent_memcg = NULL;
3789 3790

	if (parent)
3791
		parent_memcg = mem_cgroup_from_cont(parent);
3792 3793 3794

	cgroup_lock();
	/*
3795
	 * If parent's use_hierarchy is set, we can't make any modifications
3796 3797 3798 3799 3800 3801
	 * 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.
	 */
3802
	if ((!parent_memcg || !parent_memcg->use_hierarchy) &&
3803 3804
				(val == 1 || val == 0)) {
		if (list_empty(&cont->children))
3805
			memcg->use_hierarchy = val;
3806 3807 3808 3809 3810 3811 3812 3813 3814
		else
			retval = -EBUSY;
	} else
		retval = -EINVAL;
	cgroup_unlock();

	return retval;
}

3815

3816
static unsigned long mem_cgroup_recursive_stat(struct mem_cgroup *memcg,
3817
					       enum mem_cgroup_stat_index idx)
3818
{
K
KAMEZAWA Hiroyuki 已提交
3819
	struct mem_cgroup *iter;
3820
	long val = 0;
3821

3822
	/* Per-cpu values can be negative, use a signed accumulator */
3823
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
3824 3825 3826 3827 3828
		val += mem_cgroup_read_stat(iter, idx);

	if (val < 0) /* race ? */
		val = 0;
	return val;
3829 3830
}

3831
static inline u64 mem_cgroup_usage(struct mem_cgroup *memcg, bool swap)
3832
{
K
KAMEZAWA Hiroyuki 已提交
3833
	u64 val;
3834

3835
	if (!mem_cgroup_is_root(memcg)) {
3836
		if (!swap)
3837
			return res_counter_read_u64(&memcg->res, RES_USAGE);
3838
		else
3839
			return res_counter_read_u64(&memcg->memsw, RES_USAGE);
3840 3841
	}

3842 3843
	val = mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_CACHE);
	val += mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_RSS);
3844

K
KAMEZAWA Hiroyuki 已提交
3845
	if (swap)
3846
		val += mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_SWAPOUT);
3847 3848 3849 3850

	return val << PAGE_SHIFT;
}

3851 3852 3853
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
3854
{
3855
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
3856
	char str[64];
3857
	u64 val;
3858
	int type, name, len;
3859 3860 3861

	type = MEMFILE_TYPE(cft->private);
	name = MEMFILE_ATTR(cft->private);
3862 3863 3864 3865

	if (!do_swap_account && type == _MEMSWAP)
		return -EOPNOTSUPP;

3866 3867
	switch (type) {
	case _MEM:
3868
		if (name == RES_USAGE)
3869
			val = mem_cgroup_usage(memcg, false);
3870
		else
3871
			val = res_counter_read_u64(&memcg->res, name);
3872 3873
		break;
	case _MEMSWAP:
3874
		if (name == RES_USAGE)
3875
			val = mem_cgroup_usage(memcg, true);
3876
		else
3877
			val = res_counter_read_u64(&memcg->memsw, name);
3878 3879 3880 3881
		break;
	default:
		BUG();
	}
3882 3883 3884

	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
3885
}
3886 3887 3888 3889
/*
 * The user of this function is...
 * RES_LIMIT.
 */
3890 3891
static int mem_cgroup_write(struct cgroup *cont, struct cftype *cft,
			    const char *buffer)
B
Balbir Singh 已提交
3892
{
3893
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
3894
	int type, name;
3895 3896 3897
	unsigned long long val;
	int ret;

3898 3899
	type = MEMFILE_TYPE(cft->private);
	name = MEMFILE_ATTR(cft->private);
3900 3901 3902 3903

	if (!do_swap_account && type == _MEMSWAP)
		return -EOPNOTSUPP;

3904
	switch (name) {
3905
	case RES_LIMIT:
3906 3907 3908 3909
		if (mem_cgroup_is_root(memcg)) { /* Can't set limit on root */
			ret = -EINVAL;
			break;
		}
3910 3911
		/* This function does all necessary parse...reuse it */
		ret = res_counter_memparse_write_strategy(buffer, &val);
3912 3913 3914
		if (ret)
			break;
		if (type == _MEM)
3915
			ret = mem_cgroup_resize_limit(memcg, val);
3916 3917
		else
			ret = mem_cgroup_resize_memsw_limit(memcg, val);
3918
		break;
3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932
	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;
3933 3934 3935 3936 3937
	default:
		ret = -EINVAL; /* should be BUG() ? */
		break;
	}
	return ret;
B
Balbir Singh 已提交
3938 3939
}

3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960 3961 3962 3963 3964 3965 3966
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;
}

3967
static int mem_cgroup_reset(struct cgroup *cont, unsigned int event)
3968
{
3969
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
3970
	int type, name;
3971

3972 3973
	type = MEMFILE_TYPE(event);
	name = MEMFILE_ATTR(event);
3974 3975 3976 3977

	if (!do_swap_account && type == _MEMSWAP)
		return -EOPNOTSUPP;

3978
	switch (name) {
3979
	case RES_MAX_USAGE:
3980
		if (type == _MEM)
3981
			res_counter_reset_max(&memcg->res);
3982
		else
3983
			res_counter_reset_max(&memcg->memsw);
3984 3985
		break;
	case RES_FAILCNT:
3986
		if (type == _MEM)
3987
			res_counter_reset_failcnt(&memcg->res);
3988
		else
3989
			res_counter_reset_failcnt(&memcg->memsw);
3990 3991
		break;
	}
3992

3993
	return 0;
3994 3995
}

3996 3997 3998 3999 4000 4001
static u64 mem_cgroup_move_charge_read(struct cgroup *cgrp,
					struct cftype *cft)
{
	return mem_cgroup_from_cont(cgrp)->move_charge_at_immigrate;
}

4002
#ifdef CONFIG_MMU
4003 4004 4005
static int mem_cgroup_move_charge_write(struct cgroup *cgrp,
					struct cftype *cft, u64 val)
{
4006
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4007 4008 4009 4010 4011 4012 4013 4014 4015

	if (val >= (1 << NR_MOVE_TYPE))
		return -EINVAL;
	/*
	 * We check this value several times in both in can_attach() and
	 * attach(), so we need cgroup lock to prevent this value from being
	 * inconsistent.
	 */
	cgroup_lock();
4016
	memcg->move_charge_at_immigrate = val;
4017 4018 4019 4020
	cgroup_unlock();

	return 0;
}
4021 4022 4023 4024 4025 4026 4027
#else
static int mem_cgroup_move_charge_write(struct cgroup *cgrp,
					struct cftype *cft, u64 val)
{
	return -ENOSYS;
}
#endif
4028

K
KAMEZAWA Hiroyuki 已提交
4029 4030 4031 4032 4033

/* For read statistics */
enum {
	MCS_CACHE,
	MCS_RSS,
4034
	MCS_FILE_MAPPED,
K
KAMEZAWA Hiroyuki 已提交
4035 4036
	MCS_PGPGIN,
	MCS_PGPGOUT,
4037
	MCS_SWAP,
4038 4039
	MCS_PGFAULT,
	MCS_PGMAJFAULT,
K
KAMEZAWA Hiroyuki 已提交
4040 4041 4042 4043 4044 4045 4046 4047 4048 4049
	MCS_INACTIVE_ANON,
	MCS_ACTIVE_ANON,
	MCS_INACTIVE_FILE,
	MCS_ACTIVE_FILE,
	MCS_UNEVICTABLE,
	NR_MCS_STAT,
};

struct mcs_total_stat {
	s64 stat[NR_MCS_STAT];
4050 4051
};

4052
static struct {
K
KAMEZAWA Hiroyuki 已提交
4053 4054 4055 4056 4057
	char *local_name;
	char *total_name;
} memcg_stat_strings[NR_MCS_STAT] = {
	{"cache", "total_cache"},
	{"rss", "total_rss"},
4058
	{"mapped_file", "total_mapped_file"},
K
KAMEZAWA Hiroyuki 已提交
4059 4060
	{"pgpgin", "total_pgpgin"},
	{"pgpgout", "total_pgpgout"},
4061
	{"swap", "total_swap"},
4062 4063
	{"pgfault", "total_pgfault"},
	{"pgmajfault", "total_pgmajfault"},
K
KAMEZAWA Hiroyuki 已提交
4064 4065 4066 4067 4068 4069 4070 4071
	{"inactive_anon", "total_inactive_anon"},
	{"active_anon", "total_active_anon"},
	{"inactive_file", "total_inactive_file"},
	{"active_file", "total_active_file"},
	{"unevictable", "total_unevictable"}
};


K
KAMEZAWA Hiroyuki 已提交
4072
static void
4073
mem_cgroup_get_local_stat(struct mem_cgroup *memcg, struct mcs_total_stat *s)
K
KAMEZAWA Hiroyuki 已提交
4074 4075 4076 4077
{
	s64 val;

	/* per cpu stat */
4078
	val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_CACHE);
K
KAMEZAWA Hiroyuki 已提交
4079
	s->stat[MCS_CACHE] += val * PAGE_SIZE;
4080
	val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_RSS);
K
KAMEZAWA Hiroyuki 已提交
4081
	s->stat[MCS_RSS] += val * PAGE_SIZE;
4082
	val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_FILE_MAPPED);
4083
	s->stat[MCS_FILE_MAPPED] += val * PAGE_SIZE;
4084
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGPGIN);
K
KAMEZAWA Hiroyuki 已提交
4085
	s->stat[MCS_PGPGIN] += val;
4086
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGPGOUT);
K
KAMEZAWA Hiroyuki 已提交
4087
	s->stat[MCS_PGPGOUT] += val;
4088
	if (do_swap_account) {
4089
		val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_SWAPOUT);
4090 4091
		s->stat[MCS_SWAP] += val * PAGE_SIZE;
	}
4092
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGFAULT);
4093
	s->stat[MCS_PGFAULT] += val;
4094
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGMAJFAULT);
4095
	s->stat[MCS_PGMAJFAULT] += val;
K
KAMEZAWA Hiroyuki 已提交
4096 4097

	/* per zone stat */
4098
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_INACTIVE_ANON));
K
KAMEZAWA Hiroyuki 已提交
4099
	s->stat[MCS_INACTIVE_ANON] += val * PAGE_SIZE;
4100
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_ACTIVE_ANON));
K
KAMEZAWA Hiroyuki 已提交
4101
	s->stat[MCS_ACTIVE_ANON] += val * PAGE_SIZE;
4102
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_INACTIVE_FILE));
K
KAMEZAWA Hiroyuki 已提交
4103
	s->stat[MCS_INACTIVE_FILE] += val * PAGE_SIZE;
4104
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_ACTIVE_FILE));
K
KAMEZAWA Hiroyuki 已提交
4105
	s->stat[MCS_ACTIVE_FILE] += val * PAGE_SIZE;
4106
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_UNEVICTABLE));
K
KAMEZAWA Hiroyuki 已提交
4107 4108 4109 4110
	s->stat[MCS_UNEVICTABLE] += val * PAGE_SIZE;
}

static void
4111
mem_cgroup_get_total_stat(struct mem_cgroup *memcg, struct mcs_total_stat *s)
K
KAMEZAWA Hiroyuki 已提交
4112
{
K
KAMEZAWA Hiroyuki 已提交
4113 4114
	struct mem_cgroup *iter;

4115
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
4116
		mem_cgroup_get_local_stat(iter, s);
K
KAMEZAWA Hiroyuki 已提交
4117 4118
}

4119 4120 4121 4122 4123 4124 4125
#ifdef CONFIG_NUMA
static int mem_control_numa_stat_show(struct seq_file *m, void *arg)
{
	int nid;
	unsigned long total_nr, file_nr, anon_nr, unevictable_nr;
	unsigned long node_nr;
	struct cgroup *cont = m->private;
4126
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
4127

4128
	total_nr = mem_cgroup_nr_lru_pages(memcg, LRU_ALL);
4129 4130
	seq_printf(m, "total=%lu", total_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4131
		node_nr = mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL);
4132 4133 4134 4135
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');

4136
	file_nr = mem_cgroup_nr_lru_pages(memcg, LRU_ALL_FILE);
4137 4138
	seq_printf(m, "file=%lu", file_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4139
		node_nr = mem_cgroup_node_nr_lru_pages(memcg, nid,
4140
				LRU_ALL_FILE);
4141 4142 4143 4144
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');

4145
	anon_nr = mem_cgroup_nr_lru_pages(memcg, LRU_ALL_ANON);
4146 4147
	seq_printf(m, "anon=%lu", anon_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4148
		node_nr = mem_cgroup_node_nr_lru_pages(memcg, nid,
4149
				LRU_ALL_ANON);
4150 4151 4152 4153
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');

4154
	unevictable_nr = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_UNEVICTABLE));
4155 4156
	seq_printf(m, "unevictable=%lu", unevictable_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4157
		node_nr = mem_cgroup_node_nr_lru_pages(memcg, nid,
4158
				BIT(LRU_UNEVICTABLE));
4159 4160 4161 4162 4163 4164 4165
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');
	return 0;
}
#endif /* CONFIG_NUMA */

4166 4167
static int mem_control_stat_show(struct cgroup *cont, struct cftype *cft,
				 struct cgroup_map_cb *cb)
4168
{
4169
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
K
KAMEZAWA Hiroyuki 已提交
4170
	struct mcs_total_stat mystat;
4171 4172
	int i;

K
KAMEZAWA Hiroyuki 已提交
4173
	memset(&mystat, 0, sizeof(mystat));
4174
	mem_cgroup_get_local_stat(memcg, &mystat);
4175

4176

4177 4178 4179
	for (i = 0; i < NR_MCS_STAT; i++) {
		if (i == MCS_SWAP && !do_swap_account)
			continue;
K
KAMEZAWA Hiroyuki 已提交
4180
		cb->fill(cb, memcg_stat_strings[i].local_name, mystat.stat[i]);
4181
	}
L
Lee Schermerhorn 已提交
4182

K
KAMEZAWA Hiroyuki 已提交
4183
	/* Hierarchical information */
4184 4185
	{
		unsigned long long limit, memsw_limit;
4186
		memcg_get_hierarchical_limit(memcg, &limit, &memsw_limit);
4187 4188 4189 4190
		cb->fill(cb, "hierarchical_memory_limit", limit);
		if (do_swap_account)
			cb->fill(cb, "hierarchical_memsw_limit", memsw_limit);
	}
K
KOSAKI Motohiro 已提交
4191

K
KAMEZAWA Hiroyuki 已提交
4192
	memset(&mystat, 0, sizeof(mystat));
4193
	mem_cgroup_get_total_stat(memcg, &mystat);
4194 4195 4196
	for (i = 0; i < NR_MCS_STAT; i++) {
		if (i == MCS_SWAP && !do_swap_account)
			continue;
K
KAMEZAWA Hiroyuki 已提交
4197
		cb->fill(cb, memcg_stat_strings[i].total_name, mystat.stat[i]);
4198
	}
K
KAMEZAWA Hiroyuki 已提交
4199

K
KOSAKI Motohiro 已提交
4200 4201 4202 4203
#ifdef CONFIG_DEBUG_VM
	{
		int nid, zid;
		struct mem_cgroup_per_zone *mz;
4204
		struct zone_reclaim_stat *rstat;
K
KOSAKI Motohiro 已提交
4205 4206 4207 4208 4209
		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++) {
4210
				mz = mem_cgroup_zoneinfo(memcg, nid, zid);
4211
				rstat = &mz->lruvec.reclaim_stat;
K
KOSAKI Motohiro 已提交
4212

4213 4214 4215 4216
				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
4217 4218 4219 4220 4221 4222 4223 4224
			}
		cb->fill(cb, "recent_rotated_anon", recent_rotated[0]);
		cb->fill(cb, "recent_rotated_file", recent_rotated[1]);
		cb->fill(cb, "recent_scanned_anon", recent_scanned[0]);
		cb->fill(cb, "recent_scanned_file", recent_scanned[1]);
	}
#endif

4225 4226 4227
	return 0;
}

K
KOSAKI Motohiro 已提交
4228 4229 4230 4231
static u64 mem_cgroup_swappiness_read(struct cgroup *cgrp, struct cftype *cft)
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);

4232
	return mem_cgroup_swappiness(memcg);
K
KOSAKI Motohiro 已提交
4233 4234 4235 4236 4237 4238 4239
}

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;
4240

K
KOSAKI Motohiro 已提交
4241 4242 4243 4244 4245 4246 4247
	if (val > 100)
		return -EINVAL;

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

	parent = mem_cgroup_from_cont(cgrp->parent);
4248 4249 4250

	cgroup_lock();

K
KOSAKI Motohiro 已提交
4251 4252
	/* If under hierarchy, only empty-root can set this value */
	if ((parent->use_hierarchy) ||
4253 4254
	    (memcg->use_hierarchy && !list_empty(&cgrp->children))) {
		cgroup_unlock();
K
KOSAKI Motohiro 已提交
4255
		return -EINVAL;
4256
	}
K
KOSAKI Motohiro 已提交
4257 4258 4259

	memcg->swappiness = val;

4260 4261
	cgroup_unlock();

K
KOSAKI Motohiro 已提交
4262 4263 4264
	return 0;
}

4265 4266 4267 4268 4269 4270 4271 4272
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)
4273
		t = rcu_dereference(memcg->thresholds.primary);
4274
	else
4275
		t = rcu_dereference(memcg->memsw_thresholds.primary);
4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286

	if (!t)
		goto unlock;

	usage = mem_cgroup_usage(memcg, swap);

	/*
	 * current_threshold points to threshold just below usage.
	 * If it's not true, a threshold was crossed after last
	 * call of __mem_cgroup_threshold().
	 */
4287
	i = t->current_threshold;
4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310

	/*
	 * 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 */
4311
	t->current_threshold = i - 1;
4312 4313 4314 4315 4316 4317
unlock:
	rcu_read_unlock();
}

static void mem_cgroup_threshold(struct mem_cgroup *memcg)
{
4318 4319 4320 4321 4322 4323 4324
	while (memcg) {
		__mem_cgroup_threshold(memcg, false);
		if (do_swap_account)
			__mem_cgroup_threshold(memcg, true);

		memcg = parent_mem_cgroup(memcg);
	}
4325 4326 4327 4328 4329 4330 4331 4332 4333 4334
}

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;
}

4335
static int mem_cgroup_oom_notify_cb(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
4336 4337 4338
{
	struct mem_cgroup_eventfd_list *ev;

4339
	list_for_each_entry(ev, &memcg->oom_notify, list)
K
KAMEZAWA Hiroyuki 已提交
4340 4341 4342 4343
		eventfd_signal(ev->eventfd, 1);
	return 0;
}

4344
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
4345
{
K
KAMEZAWA Hiroyuki 已提交
4346 4347
	struct mem_cgroup *iter;

4348
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
4349
		mem_cgroup_oom_notify_cb(iter);
K
KAMEZAWA Hiroyuki 已提交
4350 4351 4352 4353
}

static int mem_cgroup_usage_register_event(struct cgroup *cgrp,
	struct cftype *cft, struct eventfd_ctx *eventfd, const char *args)
4354 4355
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4356 4357
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
4358 4359
	int type = MEMFILE_TYPE(cft->private);
	u64 threshold, usage;
4360
	int i, size, ret;
4361 4362 4363 4364 4365 4366

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

	mutex_lock(&memcg->thresholds_lock);
4367

4368
	if (type == _MEM)
4369
		thresholds = &memcg->thresholds;
4370
	else if (type == _MEMSWAP)
4371
		thresholds = &memcg->memsw_thresholds;
4372 4373 4374 4375 4376 4377
	else
		BUG();

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

	/* Check if a threshold crossed before adding a new one */
4378
	if (thresholds->primary)
4379 4380
		__mem_cgroup_threshold(memcg, type == _MEMSWAP);

4381
	size = thresholds->primary ? thresholds->primary->size + 1 : 1;
4382 4383

	/* Allocate memory for new array of thresholds */
4384
	new = kmalloc(sizeof(*new) + size * sizeof(struct mem_cgroup_threshold),
4385
			GFP_KERNEL);
4386
	if (!new) {
4387 4388 4389
		ret = -ENOMEM;
		goto unlock;
	}
4390
	new->size = size;
4391 4392

	/* Copy thresholds (if any) to new array */
4393 4394
	if (thresholds->primary) {
		memcpy(new->entries, thresholds->primary->entries, (size - 1) *
4395
				sizeof(struct mem_cgroup_threshold));
4396 4397
	}

4398
	/* Add new threshold */
4399 4400
	new->entries[size - 1].eventfd = eventfd;
	new->entries[size - 1].threshold = threshold;
4401 4402

	/* Sort thresholds. Registering of new threshold isn't time-critical */
4403
	sort(new->entries, size, sizeof(struct mem_cgroup_threshold),
4404 4405 4406
			compare_thresholds, NULL);

	/* Find current threshold */
4407
	new->current_threshold = -1;
4408
	for (i = 0; i < size; i++) {
4409
		if (new->entries[i].threshold < usage) {
4410
			/*
4411 4412
			 * new->current_threshold will not be used until
			 * rcu_assign_pointer(), so it's safe to increment
4413 4414
			 * it here.
			 */
4415
			++new->current_threshold;
4416 4417 4418
		}
	}

4419 4420 4421 4422 4423
	/* Free old spare buffer and save old primary buffer as spare */
	kfree(thresholds->spare);
	thresholds->spare = thresholds->primary;

	rcu_assign_pointer(thresholds->primary, new);
4424

4425
	/* To be sure that nobody uses thresholds */
4426 4427 4428 4429 4430 4431 4432 4433
	synchronize_rcu();

unlock:
	mutex_unlock(&memcg->thresholds_lock);

	return ret;
}

4434
static void mem_cgroup_usage_unregister_event(struct cgroup *cgrp,
K
KAMEZAWA Hiroyuki 已提交
4435
	struct cftype *cft, struct eventfd_ctx *eventfd)
4436 4437
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4438 4439
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
4440 4441
	int type = MEMFILE_TYPE(cft->private);
	u64 usage;
4442
	int i, j, size;
4443 4444 4445

	mutex_lock(&memcg->thresholds_lock);
	if (type == _MEM)
4446
		thresholds = &memcg->thresholds;
4447
	else if (type == _MEMSWAP)
4448
		thresholds = &memcg->memsw_thresholds;
4449 4450 4451
	else
		BUG();

4452 4453 4454
	if (!thresholds->primary)
		goto unlock;

4455 4456 4457 4458 4459 4460
	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 */
4461 4462 4463
	size = 0;
	for (i = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd != eventfd)
4464 4465 4466
			size++;
	}

4467
	new = thresholds->spare;
4468

4469 4470
	/* Set thresholds array to NULL if we don't have thresholds */
	if (!size) {
4471 4472
		kfree(new);
		new = NULL;
4473
		goto swap_buffers;
4474 4475
	}

4476
	new->size = size;
4477 4478

	/* Copy thresholds and find current threshold */
4479 4480 4481
	new->current_threshold = -1;
	for (i = 0, j = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd == eventfd)
4482 4483
			continue;

4484 4485
		new->entries[j] = thresholds->primary->entries[i];
		if (new->entries[j].threshold < usage) {
4486
			/*
4487
			 * new->current_threshold will not be used
4488 4489 4490
			 * until rcu_assign_pointer(), so it's safe to increment
			 * it here.
			 */
4491
			++new->current_threshold;
4492 4493 4494 4495
		}
		j++;
	}

4496
swap_buffers:
4497 4498
	/* Swap primary and spare array */
	thresholds->spare = thresholds->primary;
4499 4500 4501 4502 4503 4504
	/* If all events are unregistered, free the spare array */
	if (!new) {
		kfree(thresholds->spare);
		thresholds->spare = NULL;
	}

4505
	rcu_assign_pointer(thresholds->primary, new);
4506

4507
	/* To be sure that nobody uses thresholds */
4508
	synchronize_rcu();
4509
unlock:
4510 4511
	mutex_unlock(&memcg->thresholds_lock);
}
4512

K
KAMEZAWA Hiroyuki 已提交
4513 4514 4515 4516 4517 4518 4519 4520 4521 4522 4523 4524
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;
	int type = MEMFILE_TYPE(cft->private);

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

4525
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4526 4527 4528 4529 4530

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

	/* already in OOM ? */
4531
	if (atomic_read(&memcg->under_oom))
K
KAMEZAWA Hiroyuki 已提交
4532
		eventfd_signal(eventfd, 1);
4533
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4534 4535 4536 4537

	return 0;
}

4538
static void mem_cgroup_oom_unregister_event(struct cgroup *cgrp,
K
KAMEZAWA Hiroyuki 已提交
4539 4540
	struct cftype *cft, struct eventfd_ctx *eventfd)
{
4541
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
K
KAMEZAWA Hiroyuki 已提交
4542 4543 4544 4545 4546
	struct mem_cgroup_eventfd_list *ev, *tmp;
	int type = MEMFILE_TYPE(cft->private);

	BUG_ON(type != _OOM_TYPE);

4547
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4548

4549
	list_for_each_entry_safe(ev, tmp, &memcg->oom_notify, list) {
K
KAMEZAWA Hiroyuki 已提交
4550 4551 4552 4553 4554 4555
		if (ev->eventfd == eventfd) {
			list_del(&ev->list);
			kfree(ev);
		}
	}

4556
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4557 4558
}

4559 4560 4561
static int mem_cgroup_oom_control_read(struct cgroup *cgrp,
	struct cftype *cft,  struct cgroup_map_cb *cb)
{
4562
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4563

4564
	cb->fill(cb, "oom_kill_disable", memcg->oom_kill_disable);
4565

4566
	if (atomic_read(&memcg->under_oom))
4567 4568 4569 4570 4571 4572 4573 4574 4575
		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)
{
4576
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4577 4578 4579 4580 4581 4582 4583 4584 4585 4586 4587
	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);

	cgroup_lock();
	/* oom-kill-disable is a flag for subhierarchy. */
	if ((parent->use_hierarchy) ||
4588
	    (memcg->use_hierarchy && !list_empty(&cgrp->children))) {
4589 4590 4591
		cgroup_unlock();
		return -EINVAL;
	}
4592
	memcg->oom_kill_disable = val;
4593
	if (!val)
4594
		memcg_oom_recover(memcg);
4595 4596 4597 4598
	cgroup_unlock();
	return 0;
}

4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614
#ifdef CONFIG_NUMA
static const struct file_operations mem_control_numa_stat_file_operations = {
	.read = seq_read,
	.llseek = seq_lseek,
	.release = single_release,
};

static int mem_control_numa_stat_open(struct inode *unused, struct file *file)
{
	struct cgroup *cont = file->f_dentry->d_parent->d_fsdata;

	file->f_op = &mem_control_numa_stat_file_operations;
	return single_open(file, mem_control_numa_stat_show, cont);
}
#endif /* CONFIG_NUMA */

4615
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
4616
static int memcg_init_kmem(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
4617
{
4618
	return mem_cgroup_sockets_init(memcg, ss);
4619 4620
};

4621
static void kmem_cgroup_destroy(struct mem_cgroup *memcg)
G
Glauber Costa 已提交
4622
{
4623
	mem_cgroup_sockets_destroy(memcg);
G
Glauber Costa 已提交
4624
}
4625
#else
4626
static int memcg_init_kmem(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
4627 4628 4629
{
	return 0;
}
G
Glauber Costa 已提交
4630

4631
static void kmem_cgroup_destroy(struct mem_cgroup *memcg)
G
Glauber Costa 已提交
4632 4633
{
}
4634 4635
#endif

B
Balbir Singh 已提交
4636 4637
static struct cftype mem_cgroup_files[] = {
	{
4638
		.name = "usage_in_bytes",
4639
		.private = MEMFILE_PRIVATE(_MEM, RES_USAGE),
4640
		.read = mem_cgroup_read,
K
KAMEZAWA Hiroyuki 已提交
4641 4642
		.register_event = mem_cgroup_usage_register_event,
		.unregister_event = mem_cgroup_usage_unregister_event,
B
Balbir Singh 已提交
4643
	},
4644 4645
	{
		.name = "max_usage_in_bytes",
4646
		.private = MEMFILE_PRIVATE(_MEM, RES_MAX_USAGE),
4647
		.trigger = mem_cgroup_reset,
4648
		.read = mem_cgroup_read,
4649
	},
B
Balbir Singh 已提交
4650
	{
4651
		.name = "limit_in_bytes",
4652
		.private = MEMFILE_PRIVATE(_MEM, RES_LIMIT),
4653
		.write_string = mem_cgroup_write,
4654
		.read = mem_cgroup_read,
B
Balbir Singh 已提交
4655
	},
4656 4657 4658 4659
	{
		.name = "soft_limit_in_bytes",
		.private = MEMFILE_PRIVATE(_MEM, RES_SOFT_LIMIT),
		.write_string = mem_cgroup_write,
4660
		.read = mem_cgroup_read,
4661
	},
B
Balbir Singh 已提交
4662 4663
	{
		.name = "failcnt",
4664
		.private = MEMFILE_PRIVATE(_MEM, RES_FAILCNT),
4665
		.trigger = mem_cgroup_reset,
4666
		.read = mem_cgroup_read,
B
Balbir Singh 已提交
4667
	},
4668 4669
	{
		.name = "stat",
4670
		.read_map = mem_control_stat_show,
4671
	},
4672 4673 4674 4675
	{
		.name = "force_empty",
		.trigger = mem_cgroup_force_empty_write,
	},
4676 4677 4678 4679 4680
	{
		.name = "use_hierarchy",
		.write_u64 = mem_cgroup_hierarchy_write,
		.read_u64 = mem_cgroup_hierarchy_read,
	},
K
KOSAKI Motohiro 已提交
4681 4682 4683 4684 4685
	{
		.name = "swappiness",
		.read_u64 = mem_cgroup_swappiness_read,
		.write_u64 = mem_cgroup_swappiness_write,
	},
4686 4687 4688 4689 4690
	{
		.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
4691 4692
	{
		.name = "oom_control",
4693 4694
		.read_map = mem_cgroup_oom_control_read,
		.write_u64 = mem_cgroup_oom_control_write,
K
KAMEZAWA Hiroyuki 已提交
4695 4696 4697 4698
		.register_event = mem_cgroup_oom_register_event,
		.unregister_event = mem_cgroup_oom_unregister_event,
		.private = MEMFILE_PRIVATE(_OOM_TYPE, OOM_CONTROL),
	},
4699 4700 4701 4702
#ifdef CONFIG_NUMA
	{
		.name = "numa_stat",
		.open = mem_control_numa_stat_open,
4703
		.mode = S_IRUGO,
4704 4705
	},
#endif
4706 4707 4708 4709
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
	{
		.name = "memsw.usage_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_USAGE),
4710
		.read = mem_cgroup_read,
K
KAMEZAWA Hiroyuki 已提交
4711 4712
		.register_event = mem_cgroup_usage_register_event,
		.unregister_event = mem_cgroup_usage_unregister_event,
4713 4714 4715 4716 4717
	},
	{
		.name = "memsw.max_usage_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_MAX_USAGE),
		.trigger = mem_cgroup_reset,
4718
		.read = mem_cgroup_read,
4719 4720 4721 4722 4723
	},
	{
		.name = "memsw.limit_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_LIMIT),
		.write_string = mem_cgroup_write,
4724
		.read = mem_cgroup_read,
4725 4726 4727 4728 4729
	},
	{
		.name = "memsw.failcnt",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_FAILCNT),
		.trigger = mem_cgroup_reset,
4730
		.read = mem_cgroup_read,
4731 4732
	},
#endif
4733
	{ },	/* terminate */
4734
};
4735

4736
static int alloc_mem_cgroup_per_zone_info(struct mem_cgroup *memcg, int node)
4737 4738
{
	struct mem_cgroup_per_node *pn;
4739
	struct mem_cgroup_per_zone *mz;
H
Hugh Dickins 已提交
4740
	enum lru_list lru;
4741
	int zone, tmp = node;
4742 4743 4744 4745 4746 4747 4748 4749
	/*
	 * 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.
	 */
4750 4751
	if (!node_state(node, N_NORMAL_MEMORY))
		tmp = -1;
4752
	pn = kzalloc_node(sizeof(*pn), GFP_KERNEL, tmp);
4753 4754
	if (!pn)
		return 1;
4755 4756 4757

	for (zone = 0; zone < MAX_NR_ZONES; zone++) {
		mz = &pn->zoneinfo[zone];
H
Hugh Dickins 已提交
4758 4759
		for_each_lru(lru)
			INIT_LIST_HEAD(&mz->lruvec.lists[lru]);
4760
		mz->usage_in_excess = 0;
4761
		mz->on_tree = false;
4762
		mz->memcg = memcg;
4763
	}
4764
	memcg->info.nodeinfo[node] = pn;
4765 4766 4767
	return 0;
}

4768
static void free_mem_cgroup_per_zone_info(struct mem_cgroup *memcg, int node)
4769
{
4770
	kfree(memcg->info.nodeinfo[node]);
4771 4772
}

4773 4774
static struct mem_cgroup *mem_cgroup_alloc(void)
{
4775
	struct mem_cgroup *memcg;
4776
	int size = sizeof(struct mem_cgroup);
4777

4778
	/* Can be very big if MAX_NUMNODES is very big */
4779
	if (size < PAGE_SIZE)
4780
		memcg = kzalloc(size, GFP_KERNEL);
4781
	else
4782
		memcg = vzalloc(size);
4783

4784
	if (!memcg)
4785 4786
		return NULL;

4787 4788
	memcg->stat = alloc_percpu(struct mem_cgroup_stat_cpu);
	if (!memcg->stat)
4789
		goto out_free;
4790 4791
	spin_lock_init(&memcg->pcp_counter_lock);
	return memcg;
4792 4793 4794

out_free:
	if (size < PAGE_SIZE)
4795
		kfree(memcg);
4796
	else
4797
		vfree(memcg);
4798
	return NULL;
4799 4800
}

4801 4802 4803 4804 4805 4806 4807 4808 4809 4810 4811 4812 4813 4814 4815 4816 4817 4818 4819 4820 4821
/*
 * Helpers for freeing a vzalloc()ed mem_cgroup by RCU,
 * but in process context.  The work_freeing structure is overlaid
 * on the rcu_freeing structure, which itself is overlaid on memsw.
 */
static void vfree_work(struct work_struct *work)
{
	struct mem_cgroup *memcg;

	memcg = container_of(work, struct mem_cgroup, work_freeing);
	vfree(memcg);
}
static void vfree_rcu(struct rcu_head *rcu_head)
{
	struct mem_cgroup *memcg;

	memcg = container_of(rcu_head, struct mem_cgroup, rcu_freeing);
	INIT_WORK(&memcg->work_freeing, vfree_work);
	schedule_work(&memcg->work_freeing);
}

4822 4823 4824 4825 4826 4827 4828 4829 4830 4831 4832
/*
 * 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.
 */

4833
static void __mem_cgroup_free(struct mem_cgroup *memcg)
4834
{
K
KAMEZAWA Hiroyuki 已提交
4835 4836
	int node;

4837 4838
	mem_cgroup_remove_from_trees(memcg);
	free_css_id(&mem_cgroup_subsys, &memcg->css);
K
KAMEZAWA Hiroyuki 已提交
4839

B
Bob Liu 已提交
4840
	for_each_node(node)
4841
		free_mem_cgroup_per_zone_info(memcg, node);
K
KAMEZAWA Hiroyuki 已提交
4842

4843
	free_percpu(memcg->stat);
4844
	if (sizeof(struct mem_cgroup) < PAGE_SIZE)
4845
		kfree_rcu(memcg, rcu_freeing);
4846
	else
4847
		call_rcu(&memcg->rcu_freeing, vfree_rcu);
4848 4849
}

4850
static void mem_cgroup_get(struct mem_cgroup *memcg)
4851
{
4852
	atomic_inc(&memcg->refcnt);
4853 4854
}

4855
static void __mem_cgroup_put(struct mem_cgroup *memcg, int count)
4856
{
4857 4858 4859
	if (atomic_sub_and_test(count, &memcg->refcnt)) {
		struct mem_cgroup *parent = parent_mem_cgroup(memcg);
		__mem_cgroup_free(memcg);
4860 4861 4862
		if (parent)
			mem_cgroup_put(parent);
	}
4863 4864
}

4865
static void mem_cgroup_put(struct mem_cgroup *memcg)
4866
{
4867
	__mem_cgroup_put(memcg, 1);
4868 4869
}

4870 4871 4872
/*
 * Returns the parent mem_cgroup in memcgroup hierarchy with hierarchy enabled.
 */
G
Glauber Costa 已提交
4873
struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
4874
{
4875
	if (!memcg->res.parent)
4876
		return NULL;
4877
	return mem_cgroup_from_res_counter(memcg->res.parent, res);
4878
}
G
Glauber Costa 已提交
4879
EXPORT_SYMBOL(parent_mem_cgroup);
4880

4881 4882 4883
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
static void __init enable_swap_cgroup(void)
{
4884
	if (!mem_cgroup_disabled() && really_do_swap_account)
4885 4886 4887 4888 4889 4890 4891 4892
		do_swap_account = 1;
}
#else
static void __init enable_swap_cgroup(void)
{
}
#endif

4893 4894 4895 4896 4897 4898
static int mem_cgroup_soft_limit_tree_init(void)
{
	struct mem_cgroup_tree_per_node *rtpn;
	struct mem_cgroup_tree_per_zone *rtpz;
	int tmp, node, zone;

B
Bob Liu 已提交
4899
	for_each_node(node) {
4900 4901 4902 4903 4904
		tmp = node;
		if (!node_state(node, N_NORMAL_MEMORY))
			tmp = -1;
		rtpn = kzalloc_node(sizeof(*rtpn), GFP_KERNEL, tmp);
		if (!rtpn)
4905
			goto err_cleanup;
4906 4907 4908 4909 4910 4911 4912 4913 4914 4915

		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);
		}
	}
	return 0;
4916 4917

err_cleanup:
B
Bob Liu 已提交
4918
	for_each_node(node) {
4919 4920 4921 4922 4923 4924 4925
		if (!soft_limit_tree.rb_tree_per_node[node])
			break;
		kfree(soft_limit_tree.rb_tree_per_node[node]);
		soft_limit_tree.rb_tree_per_node[node] = NULL;
	}
	return 1;

4926 4927
}

L
Li Zefan 已提交
4928
static struct cgroup_subsys_state * __ref
4929
mem_cgroup_create(struct cgroup *cont)
B
Balbir Singh 已提交
4930
{
4931
	struct mem_cgroup *memcg, *parent;
K
KAMEZAWA Hiroyuki 已提交
4932
	long error = -ENOMEM;
4933
	int node;
B
Balbir Singh 已提交
4934

4935 4936
	memcg = mem_cgroup_alloc();
	if (!memcg)
K
KAMEZAWA Hiroyuki 已提交
4937
		return ERR_PTR(error);
4938

B
Bob Liu 已提交
4939
	for_each_node(node)
4940
		if (alloc_mem_cgroup_per_zone_info(memcg, node))
4941
			goto free_out;
4942

4943
	/* root ? */
4944
	if (cont->parent == NULL) {
4945
		int cpu;
4946
		enable_swap_cgroup();
4947
		parent = NULL;
4948 4949
		if (mem_cgroup_soft_limit_tree_init())
			goto free_out;
4950
		root_mem_cgroup = memcg;
4951 4952 4953 4954 4955
		for_each_possible_cpu(cpu) {
			struct memcg_stock_pcp *stock =
						&per_cpu(memcg_stock, cpu);
			INIT_WORK(&stock->work, drain_local_stock);
		}
4956
		hotcpu_notifier(memcg_cpu_hotplug_callback, 0);
4957
	} else {
4958
		parent = mem_cgroup_from_cont(cont->parent);
4959 4960
		memcg->use_hierarchy = parent->use_hierarchy;
		memcg->oom_kill_disable = parent->oom_kill_disable;
4961
	}
4962

4963
	if (parent && parent->use_hierarchy) {
4964 4965
		res_counter_init(&memcg->res, &parent->res);
		res_counter_init(&memcg->memsw, &parent->memsw);
4966 4967 4968 4969 4970 4971 4972
		/*
		 * We increment refcnt of the parent to ensure that we can
		 * safely access it on res_counter_charge/uncharge.
		 * This refcnt will be decremented when freeing this
		 * mem_cgroup(see mem_cgroup_put).
		 */
		mem_cgroup_get(parent);
4973
	} else {
4974 4975
		res_counter_init(&memcg->res, NULL);
		res_counter_init(&memcg->memsw, NULL);
4976
	}
4977 4978
	memcg->last_scanned_node = MAX_NUMNODES;
	INIT_LIST_HEAD(&memcg->oom_notify);
4979

K
KOSAKI Motohiro 已提交
4980
	if (parent)
4981 4982 4983 4984
		memcg->swappiness = mem_cgroup_swappiness(parent);
	atomic_set(&memcg->refcnt, 1);
	memcg->move_charge_at_immigrate = 0;
	mutex_init(&memcg->thresholds_lock);
4985
	spin_lock_init(&memcg->move_lock);
4986 4987 4988 4989 4990 4991 4992 4993 4994 4995 4996

	error = memcg_init_kmem(memcg, &mem_cgroup_subsys);
	if (error) {
		/*
		 * We call put now because our (and parent's) refcnts
		 * are already in place. mem_cgroup_put() will internally
		 * call __mem_cgroup_free, so return directly
		 */
		mem_cgroup_put(memcg);
		return ERR_PTR(error);
	}
4997
	return &memcg->css;
4998
free_out:
4999
	__mem_cgroup_free(memcg);
K
KAMEZAWA Hiroyuki 已提交
5000
	return ERR_PTR(error);
B
Balbir Singh 已提交
5001 5002
}

5003
static int mem_cgroup_pre_destroy(struct cgroup *cont)
5004
{
5005
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
5006

5007
	return mem_cgroup_force_empty(memcg, false);
5008 5009
}

5010
static void mem_cgroup_destroy(struct cgroup *cont)
B
Balbir Singh 已提交
5011
{
5012
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
5013

5014
	kmem_cgroup_destroy(memcg);
G
Glauber Costa 已提交
5015

5016
	mem_cgroup_put(memcg);
B
Balbir Singh 已提交
5017 5018
}

5019
#ifdef CONFIG_MMU
5020
/* Handlers for move charge at task migration. */
5021 5022
#define PRECHARGE_COUNT_AT_ONCE	256
static int mem_cgroup_do_precharge(unsigned long count)
5023
{
5024 5025
	int ret = 0;
	int batch_count = PRECHARGE_COUNT_AT_ONCE;
5026
	struct mem_cgroup *memcg = mc.to;
5027

5028
	if (mem_cgroup_is_root(memcg)) {
5029 5030 5031 5032 5033 5034 5035 5036
		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;
		/*
5037
		 * "memcg" cannot be under rmdir() because we've already checked
5038 5039 5040 5041
		 * 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().
		 */
5042
		if (res_counter_charge(&memcg->res, PAGE_SIZE * count, &dummy))
5043
			goto one_by_one;
5044
		if (do_swap_account && res_counter_charge(&memcg->memsw,
5045
						PAGE_SIZE * count, &dummy)) {
5046
			res_counter_uncharge(&memcg->res, PAGE_SIZE * count);
5047 5048 5049 5050 5051 5052 5053 5054 5055 5056 5057 5058 5059 5060 5061 5062
			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();
		}
5063 5064
		ret = __mem_cgroup_try_charge(NULL,
					GFP_KERNEL, 1, &memcg, false);
5065
		if (ret)
5066
			/* mem_cgroup_clear_mc() will do uncharge later */
5067
			return ret;
5068 5069
		mc.precharge++;
	}
5070 5071 5072 5073
	return ret;
}

/**
5074
 * get_mctgt_type - get target type of moving charge
5075 5076 5077
 * @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
5078
 * @target: the pointer the target page or swap ent will be stored(can be NULL)
5079 5080 5081 5082 5083 5084
 *
 * 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).
5085 5086 5087
 *   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.
5088 5089 5090 5091 5092
 *
 * Called with pte lock held.
 */
union mc_target {
	struct page	*page;
5093
	swp_entry_t	ent;
5094 5095 5096
};

enum mc_target_type {
5097
	MC_TARGET_NONE = 0,
5098
	MC_TARGET_PAGE,
5099
	MC_TARGET_SWAP,
5100 5101
};

D
Daisuke Nishimura 已提交
5102 5103
static struct page *mc_handle_present_pte(struct vm_area_struct *vma,
						unsigned long addr, pte_t ptent)
5104
{
D
Daisuke Nishimura 已提交
5105
	struct page *page = vm_normal_page(vma, addr, ptent);
5106

D
Daisuke Nishimura 已提交
5107 5108 5109 5110
	if (!page || !page_mapped(page))
		return NULL;
	if (PageAnon(page)) {
		/* we don't move shared anon */
5111
		if (!move_anon())
D
Daisuke Nishimura 已提交
5112
			return NULL;
5113 5114
	} else if (!move_file())
		/* we ignore mapcount for file pages */
D
Daisuke Nishimura 已提交
5115 5116 5117 5118 5119 5120 5121
		return NULL;
	if (!get_page_unless_zero(page))
		return NULL;

	return page;
}

5122
#ifdef CONFIG_SWAP
D
Daisuke Nishimura 已提交
5123 5124 5125 5126 5127 5128 5129 5130
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;
5131 5132 5133 5134 5135
	/*
	 * Because lookup_swap_cache() updates some statistics counter,
	 * we call find_get_page() with swapper_space directly.
	 */
	page = find_get_page(&swapper_space, ent.val);
D
Daisuke Nishimura 已提交
5136 5137 5138 5139 5140
	if (do_swap_account)
		entry->val = ent.val;

	return page;
}
5141 5142 5143 5144 5145 5146 5147
#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
5148

5149 5150 5151 5152 5153 5154 5155 5156 5157 5158 5159 5160 5161 5162 5163 5164 5165 5166 5167
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). */
5168 5169 5170 5171 5172 5173
	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);
5174
		if (do_swap_account)
5175 5176
			*entry = swap;
		page = find_get_page(&swapper_space, swap.val);
5177
	}
5178
#endif
5179 5180 5181
	return page;
}

5182
static enum mc_target_type get_mctgt_type(struct vm_area_struct *vma,
D
Daisuke Nishimura 已提交
5183 5184 5185 5186
		unsigned long addr, pte_t ptent, union mc_target *target)
{
	struct page *page = NULL;
	struct page_cgroup *pc;
5187
	enum mc_target_type ret = MC_TARGET_NONE;
D
Daisuke Nishimura 已提交
5188 5189 5190 5191 5192 5193
	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);
5194 5195
	else if (pte_none(ptent) || pte_file(ptent))
		page = mc_handle_file_pte(vma, addr, ptent, &ent);
D
Daisuke Nishimura 已提交
5196 5197

	if (!page && !ent.val)
5198
		return ret;
5199 5200 5201 5202 5203 5204 5205 5206 5207 5208 5209 5210 5211 5212 5213
	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
5214 5215
	/* There is a swap entry and a page doesn't exist or isn't charged */
	if (ent.val && !ret &&
5216
			css_id(&mc.from->css) == lookup_swap_cgroup_id(ent)) {
5217 5218 5219
		ret = MC_TARGET_SWAP;
		if (target)
			target->ent = ent;
5220 5221 5222 5223
	}
	return ret;
}

5224 5225 5226 5227 5228 5229 5230 5231 5232 5233 5234 5235 5236 5237 5238 5239 5240 5241 5242 5243 5244 5245 5246 5247 5248 5249 5250 5251 5252 5253 5254 5255 5256 5257 5258
#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

5259 5260 5261 5262 5263 5264 5265 5266
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;

5267 5268 5269 5270
	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);
5271
		return 0;
5272
	}
5273

5274 5275
	if (pmd_trans_unstable(pmd))
		return 0;
5276 5277
	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
	for (; addr != end; pte++, addr += PAGE_SIZE)
5278
		if (get_mctgt_type(vma, addr, *pte, NULL))
5279 5280 5281 5282
			mc.precharge++;	/* increment precharge temporarily */
	pte_unmap_unlock(pte - 1, ptl);
	cond_resched();

5283 5284 5285
	return 0;
}

5286 5287 5288 5289 5290
static unsigned long mem_cgroup_count_precharge(struct mm_struct *mm)
{
	unsigned long precharge;
	struct vm_area_struct *vma;

5291
	down_read(&mm->mmap_sem);
5292 5293 5294 5295 5296 5297 5298 5299 5300 5301 5302
	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);
	}
5303
	up_read(&mm->mmap_sem);
5304 5305 5306 5307 5308 5309 5310 5311 5312

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

	return precharge;
}

static int mem_cgroup_precharge_mc(struct mm_struct *mm)
{
5313 5314 5315 5316 5317
	unsigned long precharge = mem_cgroup_count_precharge(mm);

	VM_BUG_ON(mc.moving_task);
	mc.moving_task = current;
	return mem_cgroup_do_precharge(precharge);
5318 5319
}

5320 5321
/* cancels all extra charges on mc.from and mc.to, and wakes up all waiters. */
static void __mem_cgroup_clear_mc(void)
5322
{
5323 5324 5325
	struct mem_cgroup *from = mc.from;
	struct mem_cgroup *to = mc.to;

5326
	/* we must uncharge all the leftover precharges from mc.to */
5327 5328 5329 5330 5331 5332 5333 5334 5335 5336 5337
	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;
5338
	}
5339 5340 5341 5342 5343 5344 5345 5346 5347 5348 5349 5350 5351 5352 5353 5354 5355 5356 5357
	/* 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);
		__mem_cgroup_put(mc.from, mc.moved_swap);

		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);
		}
		/* we've already done mem_cgroup_get(mc.to) */
		mc.moved_swap = 0;
	}
5358 5359 5360 5361 5362 5363 5364 5365 5366 5367 5368 5369 5370 5371 5372
	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();
5373
	spin_lock(&mc.lock);
5374 5375
	mc.from = NULL;
	mc.to = NULL;
5376
	spin_unlock(&mc.lock);
5377
	mem_cgroup_end_move(from);
5378 5379
}

5380 5381
static int mem_cgroup_can_attach(struct cgroup *cgroup,
				 struct cgroup_taskset *tset)
5382
{
5383
	struct task_struct *p = cgroup_taskset_first(tset);
5384
	int ret = 0;
5385
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgroup);
5386

5387
	if (memcg->move_charge_at_immigrate) {
5388 5389 5390
		struct mm_struct *mm;
		struct mem_cgroup *from = mem_cgroup_from_task(p);

5391
		VM_BUG_ON(from == memcg);
5392 5393 5394 5395 5396

		mm = get_task_mm(p);
		if (!mm)
			return 0;
		/* We move charges only when we move a owner of the mm */
5397 5398 5399 5400
		if (mm->owner == p) {
			VM_BUG_ON(mc.from);
			VM_BUG_ON(mc.to);
			VM_BUG_ON(mc.precharge);
5401
			VM_BUG_ON(mc.moved_charge);
5402
			VM_BUG_ON(mc.moved_swap);
5403
			mem_cgroup_start_move(from);
5404
			spin_lock(&mc.lock);
5405
			mc.from = from;
5406
			mc.to = memcg;
5407
			spin_unlock(&mc.lock);
5408
			/* We set mc.moving_task later */
5409 5410 5411 5412

			ret = mem_cgroup_precharge_mc(mm);
			if (ret)
				mem_cgroup_clear_mc();
5413 5414
		}
		mmput(mm);
5415 5416 5417 5418
	}
	return ret;
}

5419 5420
static void mem_cgroup_cancel_attach(struct cgroup *cgroup,
				     struct cgroup_taskset *tset)
5421
{
5422
	mem_cgroup_clear_mc();
5423 5424
}

5425 5426 5427
static int mem_cgroup_move_charge_pte_range(pmd_t *pmd,
				unsigned long addr, unsigned long end,
				struct mm_walk *walk)
5428
{
5429 5430 5431 5432
	int ret = 0;
	struct vm_area_struct *vma = walk->private;
	pte_t *pte;
	spinlock_t *ptl;
5433 5434 5435 5436
	enum mc_target_type target_type;
	union mc_target target;
	struct page *page;
	struct page_cgroup *pc;
5437

5438 5439 5440 5441 5442 5443 5444 5445 5446 5447 5448
	/*
	 * 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) {
5449
		if (mc.precharge < HPAGE_PMD_NR) {
5450 5451 5452 5453 5454 5455 5456 5457 5458 5459 5460 5461 5462 5463 5464 5465 5466 5467 5468
			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,
							     pc, mc.from, mc.to,
							     false)) {
					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);
5469
		return 0;
5470 5471
	}

5472 5473
	if (pmd_trans_unstable(pmd))
		return 0;
5474 5475 5476 5477
retry:
	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
	for (; addr != end; addr += PAGE_SIZE) {
		pte_t ptent = *(pte++);
5478
		swp_entry_t ent;
5479 5480 5481 5482

		if (!mc.precharge)
			break;

5483
		switch (get_mctgt_type(vma, addr, ptent, &target)) {
5484 5485 5486 5487 5488
		case MC_TARGET_PAGE:
			page = target.page;
			if (isolate_lru_page(page))
				goto put;
			pc = lookup_page_cgroup(page);
5489 5490
			if (!mem_cgroup_move_account(page, 1, pc,
						     mc.from, mc.to, false)) {
5491
				mc.precharge--;
5492 5493
				/* we uncharge from mc.from later. */
				mc.moved_charge++;
5494 5495
			}
			putback_lru_page(page);
5496
put:			/* get_mctgt_type() gets the page */
5497 5498
			put_page(page);
			break;
5499 5500
		case MC_TARGET_SWAP:
			ent = target.ent;
5501
			if (!mem_cgroup_move_swap_account(ent, mc.from, mc.to)) {
5502
				mc.precharge--;
5503 5504 5505
				/* we fixup refcnts and charges later. */
				mc.moved_swap++;
			}
5506
			break;
5507 5508 5509 5510 5511 5512 5513 5514 5515 5516 5517 5518 5519 5520
		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.
		 */
5521
		ret = mem_cgroup_do_precharge(1);
5522 5523 5524 5525 5526 5527 5528 5529 5530 5531 5532 5533
		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();
5534 5535 5536 5537 5538 5539 5540 5541 5542 5543 5544 5545 5546
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;
	}
5547 5548 5549 5550 5551 5552 5553 5554 5555 5556 5557 5558 5559 5560 5561 5562 5563 5564
	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;
	}
5565
	up_read(&mm->mmap_sem);
5566 5567
}

5568 5569
static void mem_cgroup_move_task(struct cgroup *cont,
				 struct cgroup_taskset *tset)
B
Balbir Singh 已提交
5570
{
5571
	struct task_struct *p = cgroup_taskset_first(tset);
5572
	struct mm_struct *mm = get_task_mm(p);
5573 5574

	if (mm) {
5575 5576
		if (mc.to)
			mem_cgroup_move_charge(mm);
5577 5578
		mmput(mm);
	}
5579 5580
	if (mc.to)
		mem_cgroup_clear_mc();
B
Balbir Singh 已提交
5581
}
5582
#else	/* !CONFIG_MMU */
5583 5584
static int mem_cgroup_can_attach(struct cgroup *cgroup,
				 struct cgroup_taskset *tset)
5585 5586 5587
{
	return 0;
}
5588 5589
static void mem_cgroup_cancel_attach(struct cgroup *cgroup,
				     struct cgroup_taskset *tset)
5590 5591
{
}
5592 5593
static void mem_cgroup_move_task(struct cgroup *cont,
				 struct cgroup_taskset *tset)
5594 5595 5596
{
}
#endif
B
Balbir Singh 已提交
5597

B
Balbir Singh 已提交
5598 5599 5600 5601
struct cgroup_subsys mem_cgroup_subsys = {
	.name = "memory",
	.subsys_id = mem_cgroup_subsys_id,
	.create = mem_cgroup_create,
5602
	.pre_destroy = mem_cgroup_pre_destroy,
B
Balbir Singh 已提交
5603
	.destroy = mem_cgroup_destroy,
5604 5605
	.can_attach = mem_cgroup_can_attach,
	.cancel_attach = mem_cgroup_cancel_attach,
B
Balbir Singh 已提交
5606
	.attach = mem_cgroup_move_task,
5607
	.base_cftypes = mem_cgroup_files,
5608
	.early_init = 0,
K
KAMEZAWA Hiroyuki 已提交
5609
	.use_id = 1,
5610
	.__DEPRECATED_clear_css_refs = true,
B
Balbir Singh 已提交
5611
};
5612 5613

#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
5614 5615 5616
static int __init enable_swap_account(char *s)
{
	/* consider enabled if no parameter or 1 is given */
5617
	if (!strcmp(s, "1"))
5618
		really_do_swap_account = 1;
5619
	else if (!strcmp(s, "0"))
5620 5621 5622
		really_do_swap_account = 0;
	return 1;
}
5623
__setup("swapaccount=", enable_swap_account);
5624 5625

#endif