memcontrol.c 144.8 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
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
	MEM_CGROUP_ON_MOVE,	/* someone is moving account between groups */
93 94 95
	MEM_CGROUP_STAT_NSTATS,
};

96 97 98 99
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 */
100 101
	MEM_CGROUP_EVENTS_PGFAULT,	/* # of page-faults */
	MEM_CGROUP_EVENTS_PGMAJFAULT,	/* # of major page-faults */
102 103
	MEM_CGROUP_EVENTS_NSTATS,
};
104 105 106 107 108 109 110 111 112
/*
 * 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,
113
	MEM_CGROUP_TARGET_NUMAINFO,
114 115 116 117
	MEM_CGROUP_NTARGETS,
};
#define THRESHOLDS_EVENTS_TARGET (128)
#define SOFTLIMIT_EVENTS_TARGET (1024)
118
#define NUMAINFO_EVENTS_TARGET	(1024)
119

120
struct mem_cgroup_stat_cpu {
121
	long count[MEM_CGROUP_STAT_NSTATS];
122
	unsigned long events[MEM_CGROUP_EVENTS_NSTATS];
123
	unsigned long targets[MEM_CGROUP_NTARGETS];
124 125
};

126 127 128 129
/*
 * per-zone information in memory controller.
 */
struct mem_cgroup_per_zone {
130 131 132
	/*
	 * spin_lock to protect the per cgroup LRU
	 */
133 134
	struct list_head	lists[NR_LRU_LISTS];
	unsigned long		count[NR_LRU_LISTS];
K
KOSAKI Motohiro 已提交
135 136

	struct zone_reclaim_stat reclaim_stat;
137 138 139 140
	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;
141 142
	struct mem_cgroup	*mem;		/* Back pointer, we cannot */
						/* use container_of	   */
143 144 145 146 147 148 149 150 151 152 153 154
};
/* Macro for accessing counter */
#define MEM_CGROUP_ZSTAT(mz, idx)	((mz)->count[(idx)])

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];
};

155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174
/*
 * 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;

175 176 177 178 179
struct mem_cgroup_threshold {
	struct eventfd_ctx *eventfd;
	u64 threshold;
};

K
KAMEZAWA Hiroyuki 已提交
180
/* For threshold */
181 182
struct mem_cgroup_threshold_ary {
	/* An array index points to threshold just below usage. */
183
	int current_threshold;
184 185 186 187 188
	/* Size of entries[] */
	unsigned int size;
	/* Array of thresholds */
	struct mem_cgroup_threshold entries[0];
};
189 190 191 192 193 194 195 196 197 198 199 200

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
201 202 203 204 205
/* for OOM */
struct mem_cgroup_eventfd_list {
	struct list_head list;
	struct eventfd_ctx *eventfd;
};
206

207 208
static void mem_cgroup_threshold(struct mem_cgroup *memcg);
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg);
209

B
Balbir Singh 已提交
210 211 212 213 214 215 216
/*
 * 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
217 218 219
 * 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
220 221 222 223 224 225 226
 */
struct mem_cgroup {
	struct cgroup_subsys_state css;
	/*
	 * the counter to account for memory usage
	 */
	struct res_counter res;
227 228 229 230
	/*
	 * the counter to account for mem+swap usage.
	 */
	struct res_counter memsw;
231 232 233 234
	/*
	 * Per cgroup active and inactive list, similar to the
	 * per zone LRU lists.
	 */
235
	struct mem_cgroup_lru_info info;
236
	/*
237
	 * While reclaiming in a hierarchy, we cache the last child we
K
KAMEZAWA Hiroyuki 已提交
238
	 * reclaimed from.
239
	 */
K
KAMEZAWA Hiroyuki 已提交
240
	int last_scanned_child;
241 242 243
	int last_scanned_node;
#if MAX_NUMNODES > 1
	nodemask_t	scan_nodes;
244 245
	atomic_t	numainfo_events;
	atomic_t	numainfo_updating;
246
#endif
247 248 249 250
	/*
	 * Should the accounting and control be hierarchical, per subtree?
	 */
	bool use_hierarchy;
251 252 253 254

	bool		oom_lock;
	atomic_t	under_oom;

255
	atomic_t	refcnt;
256

257
	int	swappiness;
258 259
	/* OOM-Killer disable */
	int		oom_kill_disable;
K
KOSAKI Motohiro 已提交
260

261 262 263
	/* set when res.limit == memsw.limit */
	bool		memsw_is_minimum;

264 265 266 267
	/* protect arrays of thresholds */
	struct mutex thresholds_lock;

	/* thresholds for memory usage. RCU-protected */
268
	struct mem_cgroup_thresholds thresholds;
269

270
	/* thresholds for mem+swap usage. RCU-protected */
271
	struct mem_cgroup_thresholds memsw_thresholds;
272

K
KAMEZAWA Hiroyuki 已提交
273 274
	/* For oom notifier event fd */
	struct list_head oom_notify;
275

276 277 278 279 280
	/*
	 * 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;
281
	/*
282
	 * percpu counter.
283
	 */
284
	struct mem_cgroup_stat_cpu *stat;
285 286 287 288 289 290
	/*
	 * 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
291 292 293 294

#ifdef CONFIG_INET
	struct tcp_memcontrol tcp_mem;
#endif
B
Balbir Singh 已提交
295 296
};

297 298 299 300 301 302
/* 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 {
303
	MOVE_CHARGE_TYPE_ANON,	/* private anonymous page and swap of it */
304
	MOVE_CHARGE_TYPE_FILE,	/* file page(including tmpfs) and swap of it */
305 306 307
	NR_MOVE_TYPE,
};

308 309
/* "mc" and its members are protected by cgroup_mutex */
static struct move_charge_struct {
310
	spinlock_t	  lock; /* for from, to */
311 312 313
	struct mem_cgroup *from;
	struct mem_cgroup *to;
	unsigned long precharge;
314
	unsigned long moved_charge;
315
	unsigned long moved_swap;
316 317 318
	struct task_struct *moving_task;	/* a task moving charges */
	wait_queue_head_t waitq;		/* a waitq for other context */
} mc = {
319
	.lock = __SPIN_LOCK_UNLOCKED(mc.lock),
320 321
	.waitq = __WAIT_QUEUE_HEAD_INITIALIZER(mc.waitq),
};
322

D
Daisuke Nishimura 已提交
323 324 325 326 327 328
static bool move_anon(void)
{
	return test_bit(MOVE_CHARGE_TYPE_ANON,
					&mc.to->move_charge_at_immigrate);
}

329 330 331 332 333 334
static bool move_file(void)
{
	return test_bit(MOVE_CHARGE_TYPE_FILE,
					&mc.to->move_charge_at_immigrate);
}

335 336 337 338 339 340 341
/*
 * 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)

342 343 344
enum charge_type {
	MEM_CGROUP_CHARGE_TYPE_CACHE = 0,
	MEM_CGROUP_CHARGE_TYPE_MAPPED,
345
	MEM_CGROUP_CHARGE_TYPE_SHMEM,	/* used by page migration of shmem */
346
	MEM_CGROUP_CHARGE_TYPE_FORCE,	/* used by force_empty */
K
KAMEZAWA Hiroyuki 已提交
347
	MEM_CGROUP_CHARGE_TYPE_SWAPOUT,	/* for accounting swapcache */
K
KAMEZAWA Hiroyuki 已提交
348
	MEM_CGROUP_CHARGE_TYPE_DROP,	/* a page was unused swap cache */
349 350 351
	NR_CHARGE_TYPE,
};

352
/* for encoding cft->private value on file */
353 354 355
#define _MEM			(0)
#define _MEMSWAP		(1)
#define _OOM_TYPE		(2)
356 357 358
#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
359 360
/* Used for OOM nofiier */
#define OOM_CONTROL		(0)
361

362 363 364 365 366 367 368
/*
 * 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)
369 370
#define MEM_CGROUP_RECLAIM_SOFT_BIT	0x2
#define MEM_CGROUP_RECLAIM_SOFT		(1 << MEM_CGROUP_RECLAIM_SOFT_BIT)
371

372 373
static void mem_cgroup_get(struct mem_cgroup *memcg);
static void mem_cgroup_put(struct mem_cgroup *memcg);
G
Glauber Costa 已提交
374 375 376 377 378

/* Writing them here to avoid exposing memcg's inner layout */
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
#ifdef CONFIG_INET
#include <net/sock.h>
G
Glauber Costa 已提交
379
#include <net/ip.h>
G
Glauber Costa 已提交
380 381 382 383 384 385 386 387 388

static bool mem_cgroup_is_root(struct mem_cgroup *memcg);
void sock_update_memcg(struct sock *sk)
{
	if (static_branch(&memcg_socket_limit_enabled)) {
		struct mem_cgroup *memcg;

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

389 390 391 392 393 394 395 396 397 398 399 400 401 402
		/* 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
403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422
		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)
{
	if (static_branch(&memcg_socket_limit_enabled) && sk->sk_cgrp) {
		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
423 424 425 426 427 428 429 430 431

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
432 433 434
#endif /* CONFIG_INET */
#endif /* CONFIG_CGROUP_MEM_RES_CTLR_KMEM */

435
static void drain_all_stock_async(struct mem_cgroup *memcg);
436

437
static struct mem_cgroup_per_zone *
438
mem_cgroup_zoneinfo(struct mem_cgroup *memcg, int nid, int zid)
439
{
440
	return &memcg->info.nodeinfo[nid]->zoneinfo[zid];
441 442
}

443
struct cgroup_subsys_state *mem_cgroup_css(struct mem_cgroup *memcg)
444
{
445
	return &memcg->css;
446 447
}

448
static struct mem_cgroup_per_zone *
449
page_cgroup_zoneinfo(struct mem_cgroup *memcg, struct page *page)
450
{
451 452
	int nid = page_to_nid(page);
	int zid = page_zonenum(page);
453

454
	return mem_cgroup_zoneinfo(memcg, nid, zid);
455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472
}

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
473
__mem_cgroup_insert_exceeded(struct mem_cgroup *memcg,
474
				struct mem_cgroup_per_zone *mz,
475 476
				struct mem_cgroup_tree_per_zone *mctz,
				unsigned long long new_usage_in_excess)
477 478 479 480 481 482 483 484
{
	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;

485 486 487
	mz->usage_in_excess = new_usage_in_excess;
	if (!mz->usage_in_excess)
		return;
488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503
	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;
504 505 506
}

static void
507
__mem_cgroup_remove_exceeded(struct mem_cgroup *memcg,
508 509 510 511 512 513 514 515 516
				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;
}

517
static void
518
mem_cgroup_remove_exceeded(struct mem_cgroup *memcg,
519 520 521 522
				struct mem_cgroup_per_zone *mz,
				struct mem_cgroup_tree_per_zone *mctz)
{
	spin_lock(&mctz->lock);
523
	__mem_cgroup_remove_exceeded(memcg, mz, mctz);
524 525 526 527
	spin_unlock(&mctz->lock);
}


528
static void mem_cgroup_update_tree(struct mem_cgroup *memcg, struct page *page)
529
{
530
	unsigned long long excess;
531 532
	struct mem_cgroup_per_zone *mz;
	struct mem_cgroup_tree_per_zone *mctz;
533 534
	int nid = page_to_nid(page);
	int zid = page_zonenum(page);
535 536 537
	mctz = soft_limit_tree_from_page(page);

	/*
538 539
	 * Necessary to update all ancestors when hierarchy is used.
	 * because their event counter is not touched.
540
	 */
541 542 543
	for (; memcg; memcg = parent_mem_cgroup(memcg)) {
		mz = mem_cgroup_zoneinfo(memcg, nid, zid);
		excess = res_counter_soft_limit_excess(&memcg->res);
544 545 546 547
		/*
		 * We have to update the tree if mz is on RB-tree or
		 * mem is over its softlimit.
		 */
548
		if (excess || mz->on_tree) {
549 550 551
			spin_lock(&mctz->lock);
			/* if on-tree, remove it */
			if (mz->on_tree)
552
				__mem_cgroup_remove_exceeded(memcg, mz, mctz);
553
			/*
554 555
			 * Insert again. mz->usage_in_excess will be updated.
			 * If excess is 0, no tree ops.
556
			 */
557
			__mem_cgroup_insert_exceeded(memcg, mz, mctz, excess);
558 559
			spin_unlock(&mctz->lock);
		}
560 561 562
	}
}

563
static void mem_cgroup_remove_from_trees(struct mem_cgroup *memcg)
564 565 566 567 568 569 570
{
	int node, zone;
	struct mem_cgroup_per_zone *mz;
	struct mem_cgroup_tree_per_zone *mctz;

	for_each_node_state(node, N_POSSIBLE) {
		for (zone = 0; zone < MAX_NR_ZONES; zone++) {
571
			mz = mem_cgroup_zoneinfo(memcg, node, zone);
572
			mctz = soft_limit_tree_node_zone(node, zone);
573
			mem_cgroup_remove_exceeded(memcg, mz, mctz);
574 575 576 577
		}
	}
}

578 579 580 581
static struct mem_cgroup_per_zone *
__mem_cgroup_largest_soft_limit_node(struct mem_cgroup_tree_per_zone *mctz)
{
	struct rb_node *rightmost = NULL;
582
	struct mem_cgroup_per_zone *mz;
583 584

retry:
585
	mz = NULL;
586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614
	rightmost = rb_last(&mctz->rb_root);
	if (!rightmost)
		goto done;		/* Nothing to reclaim from */

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

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

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

615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633
/*
 * 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.
 */
634
static long mem_cgroup_read_stat(struct mem_cgroup *memcg,
635
				 enum mem_cgroup_stat_index idx)
636
{
637
	long val = 0;
638 639
	int cpu;

640 641
	get_online_cpus();
	for_each_online_cpu(cpu)
642
		val += per_cpu(memcg->stat->count[idx], cpu);
643
#ifdef CONFIG_HOTPLUG_CPU
644 645 646
	spin_lock(&memcg->pcp_counter_lock);
	val += memcg->nocpu_base.count[idx];
	spin_unlock(&memcg->pcp_counter_lock);
647 648
#endif
	put_online_cpus();
649 650 651
	return val;
}

652
static void mem_cgroup_swap_statistics(struct mem_cgroup *memcg,
653 654 655
					 bool charge)
{
	int val = (charge) ? 1 : -1;
656
	this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_SWAPOUT], val);
657 658
}

659
void mem_cgroup_pgfault(struct mem_cgroup *memcg, int val)
660
{
661
	this_cpu_add(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT], val);
662 663
}

664
void mem_cgroup_pgmajfault(struct mem_cgroup *memcg, int val)
665
{
666
	this_cpu_add(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT], val);
667 668
}

669
static unsigned long mem_cgroup_read_events(struct mem_cgroup *memcg,
670 671 672 673 674 675
					    enum mem_cgroup_events_index idx)
{
	unsigned long val = 0;
	int cpu;

	for_each_online_cpu(cpu)
676
		val += per_cpu(memcg->stat->events[idx], cpu);
677
#ifdef CONFIG_HOTPLUG_CPU
678 679 680
	spin_lock(&memcg->pcp_counter_lock);
	val += memcg->nocpu_base.events[idx];
	spin_unlock(&memcg->pcp_counter_lock);
681 682 683 684
#endif
	return val;
}

685
static void mem_cgroup_charge_statistics(struct mem_cgroup *memcg,
686
					 bool file, int nr_pages)
687
{
688 689
	preempt_disable();

690
	if (file)
691 692
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_CACHE],
				nr_pages);
693
	else
694 695
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_RSS],
				nr_pages);
696

697 698
	/* pagein of a big page is an event. So, ignore page size */
	if (nr_pages > 0)
699
		__this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGPGIN]);
700
	else {
701
		__this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGPGOUT]);
702 703
		nr_pages = -nr_pages; /* for event */
	}
704

705
	__this_cpu_add(memcg->stat->events[MEM_CGROUP_EVENTS_COUNT], nr_pages);
706

707
	preempt_enable();
708 709
}

710
unsigned long
711
mem_cgroup_zone_nr_lru_pages(struct mem_cgroup *memcg, int nid, int zid,
712
			unsigned int lru_mask)
713 714
{
	struct mem_cgroup_per_zone *mz;
715 716 717
	enum lru_list l;
	unsigned long ret = 0;

718
	mz = mem_cgroup_zoneinfo(memcg, nid, zid);
719 720 721 722 723 724 725 726 727

	for_each_lru(l) {
		if (BIT(l) & lru_mask)
			ret += MEM_CGROUP_ZSTAT(mz, l);
	}
	return ret;
}

static unsigned long
728
mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
729 730
			int nid, unsigned int lru_mask)
{
731 732 733
	u64 total = 0;
	int zid;

734
	for (zid = 0; zid < MAX_NR_ZONES; zid++)
735 736
		total += mem_cgroup_zone_nr_lru_pages(memcg,
						nid, zid, lru_mask);
737

738 739
	return total;
}
740

741
static unsigned long mem_cgroup_nr_lru_pages(struct mem_cgroup *memcg,
742
			unsigned int lru_mask)
743
{
744
	int nid;
745 746
	u64 total = 0;

747
	for_each_node_state(nid, N_HIGH_MEMORY)
748
		total += mem_cgroup_node_nr_lru_pages(memcg, nid, lru_mask);
749
	return total;
750 751
}

752
static bool __memcg_event_check(struct mem_cgroup *memcg, int target)
753 754 755
{
	unsigned long val, next;

756 757
	val = __this_cpu_read(memcg->stat->events[MEM_CGROUP_EVENTS_COUNT]);
	next = __this_cpu_read(memcg->stat->targets[target]);
758 759 760 761
	/* from time_after() in jiffies.h */
	return ((long)next - (long)val < 0);
}

762
static void __mem_cgroup_target_update(struct mem_cgroup *memcg, int target)
763
{
764
	unsigned long val, next;
765

766
	val = __this_cpu_read(memcg->stat->events[MEM_CGROUP_EVENTS_COUNT]);
767

768 769 770 771 772 773 774
	switch (target) {
	case MEM_CGROUP_TARGET_THRESH:
		next = val + THRESHOLDS_EVENTS_TARGET;
		break;
	case MEM_CGROUP_TARGET_SOFTLIMIT:
		next = val + SOFTLIMIT_EVENTS_TARGET;
		break;
775 776 777
	case MEM_CGROUP_TARGET_NUMAINFO:
		next = val + NUMAINFO_EVENTS_TARGET;
		break;
778 779 780 781
	default:
		return;
	}

782
	__this_cpu_write(memcg->stat->targets[target], next);
783 784 785 786 787 788
}

/*
 * Check events in order.
 *
 */
789
static void memcg_check_events(struct mem_cgroup *memcg, struct page *page)
790
{
791
	preempt_disable();
792
	/* threshold event is triggered in finer grain than soft limit */
793 794 795 796
	if (unlikely(__memcg_event_check(memcg, MEM_CGROUP_TARGET_THRESH))) {
		mem_cgroup_threshold(memcg);
		__mem_cgroup_target_update(memcg, MEM_CGROUP_TARGET_THRESH);
		if (unlikely(__memcg_event_check(memcg,
797
			     MEM_CGROUP_TARGET_SOFTLIMIT))) {
798 799
			mem_cgroup_update_tree(memcg, page);
			__mem_cgroup_target_update(memcg,
800 801 802
						   MEM_CGROUP_TARGET_SOFTLIMIT);
		}
#if MAX_NUMNODES > 1
803
		if (unlikely(__memcg_event_check(memcg,
804
			MEM_CGROUP_TARGET_NUMAINFO))) {
805 806
			atomic_inc(&memcg->numainfo_events);
			__mem_cgroup_target_update(memcg,
807
				MEM_CGROUP_TARGET_NUMAINFO);
808
		}
809
#endif
810
	}
811
	preempt_enable();
812 813
}

G
Glauber Costa 已提交
814
struct mem_cgroup *mem_cgroup_from_cont(struct cgroup *cont)
B
Balbir Singh 已提交
815 816 817 818 819 820
{
	return container_of(cgroup_subsys_state(cont,
				mem_cgroup_subsys_id), struct mem_cgroup,
				css);
}

821
struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p)
822
{
823 824 825 826 827 828 829 830
	/*
	 * 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;

831 832 833 834
	return container_of(task_subsys_state(p, mem_cgroup_subsys_id),
				struct mem_cgroup, css);
}

835
struct mem_cgroup *try_get_mem_cgroup_from_mm(struct mm_struct *mm)
836
{
837
	struct mem_cgroup *memcg = NULL;
838 839 840

	if (!mm)
		return NULL;
841 842 843 844 845 846 847
	/*
	 * 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 {
848 849
		memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
		if (unlikely(!memcg))
850
			break;
851
	} while (!css_tryget(&memcg->css));
852
	rcu_read_unlock();
853
	return memcg;
854 855
}

856 857 858
static struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *root,
					  struct mem_cgroup *prev,
					  bool reclaim)
K
KAMEZAWA Hiroyuki 已提交
859
{
860 861
	struct mem_cgroup *memcg = NULL;
	int id = 0;
862

863 864
	if (!root)
		root = root_mem_cgroup;
K
KAMEZAWA Hiroyuki 已提交
865

866 867
	if (prev && !reclaim)
		id = css_id(&prev->css);
K
KAMEZAWA Hiroyuki 已提交
868

869 870
	if (prev && prev != root)
		css_put(&prev->css);
K
KAMEZAWA Hiroyuki 已提交
871

872 873 874 875 876
	if (!root->use_hierarchy && root != root_mem_cgroup) {
		if (prev)
			return NULL;
		return root;
	}
K
KAMEZAWA Hiroyuki 已提交
877

878 879
	while (!memcg) {
		struct cgroup_subsys_state *css;
880

881 882
		if (reclaim)
			id = root->last_scanned_child;
K
KAMEZAWA Hiroyuki 已提交
883

884 885 886 887 888 889 890 891
		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
892 893
		rcu_read_unlock();

894 895 896 897 898 899 900
		if (reclaim)
			root->last_scanned_child = id;

		if (prev && !css)
			return NULL;
	}
	return memcg;
K
KAMEZAWA Hiroyuki 已提交
901
}
K
KAMEZAWA Hiroyuki 已提交
902

903 904 905 906 907 908 909 910
static void mem_cgroup_iter_break(struct mem_cgroup *root,
				  struct mem_cgroup *prev)
{
	if (!root)
		root = root_mem_cgroup;
	if (prev && prev != root)
		css_put(&prev->css);
}
K
KAMEZAWA Hiroyuki 已提交
911

912 913 914 915 916 917 918 919 920
/*
 * 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)		\
	for (iter = mem_cgroup_iter(root, NULL, false);	\
	     iter != NULL;				\
	     iter = mem_cgroup_iter(root, iter, false))
921

922 923 924 925
#define for_each_mem_cgroup(iter)			\
	for (iter = mem_cgroup_iter(NULL, NULL, false);	\
	     iter != NULL;				\
	     iter = mem_cgroup_iter(NULL, iter, false))
K
KAMEZAWA Hiroyuki 已提交
926

927
static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
928
{
929
	return (memcg == root_mem_cgroup);
930 931
}

932 933
void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
{
934
	struct mem_cgroup *memcg;
935 936 937 938 939

	if (!mm)
		return;

	rcu_read_lock();
940 941
	memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
	if (unlikely(!memcg))
942 943 944 945
		goto out;

	switch (idx) {
	case PGMAJFAULT:
946
		mem_cgroup_pgmajfault(memcg, 1);
947 948
		break;
	case PGFAULT:
949
		mem_cgroup_pgfault(memcg, 1);
950 951 952 953 954 955 956 957 958
		break;
	default:
		BUG();
	}
out:
	rcu_read_unlock();
}
EXPORT_SYMBOL(mem_cgroup_count_vm_event);

K
KAMEZAWA Hiroyuki 已提交
959 960 961 962 963 964 965 966 967 968 969 970 971
/*
 * 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.
 */
972

K
KAMEZAWA Hiroyuki 已提交
973 974 975 976
void mem_cgroup_del_lru_list(struct page *page, enum lru_list lru)
{
	struct page_cgroup *pc;
	struct mem_cgroup_per_zone *mz;
977

978
	if (mem_cgroup_disabled())
K
KAMEZAWA Hiroyuki 已提交
979 980 981
		return;
	pc = lookup_page_cgroup(page);
	/* can happen while we handle swapcache. */
982
	if (!TestClearPageCgroupAcctLRU(pc))
K
KAMEZAWA Hiroyuki 已提交
983
		return;
984
	VM_BUG_ON(!pc->mem_cgroup);
985 986 987 988
	/*
	 * We don't check PCG_USED bit. It's cleared when the "page" is finally
	 * removed from global LRU.
	 */
989
	mz = page_cgroup_zoneinfo(pc->mem_cgroup, page);
990 991
	/* huge page split is done under lru_lock. so, we have no races. */
	MEM_CGROUP_ZSTAT(mz, lru) -= 1 << compound_order(page);
992 993 994
	if (mem_cgroup_is_root(pc->mem_cgroup))
		return;
	VM_BUG_ON(list_empty(&pc->lru));
K
KAMEZAWA Hiroyuki 已提交
995
	list_del_init(&pc->lru);
996 997
}

K
KAMEZAWA Hiroyuki 已提交
998
void mem_cgroup_del_lru(struct page *page)
999
{
K
KAMEZAWA Hiroyuki 已提交
1000 1001
	mem_cgroup_del_lru_list(page, page_lru(page));
}
1002

1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024
/*
 * Writeback is about to end against a page which has been marked for immediate
 * reclaim.  If it still appears to be reclaimable, move it to the tail of the
 * inactive list.
 */
void mem_cgroup_rotate_reclaimable_page(struct page *page)
{
	struct mem_cgroup_per_zone *mz;
	struct page_cgroup *pc;
	enum lru_list lru = page_lru(page);

	if (mem_cgroup_disabled())
		return;

	pc = lookup_page_cgroup(page);
	/* unused or root page is not rotated. */
	if (!PageCgroupUsed(pc))
		return;
	/* Ensure pc->mem_cgroup is visible after reading PCG_USED. */
	smp_rmb();
	if (mem_cgroup_is_root(pc->mem_cgroup))
		return;
1025
	mz = page_cgroup_zoneinfo(pc->mem_cgroup, page);
1026 1027 1028
	list_move_tail(&pc->lru, &mz->lists[lru]);
}

K
KAMEZAWA Hiroyuki 已提交
1029 1030 1031 1032
void mem_cgroup_rotate_lru_list(struct page *page, enum lru_list lru)
{
	struct mem_cgroup_per_zone *mz;
	struct page_cgroup *pc;
1033

1034
	if (mem_cgroup_disabled())
K
KAMEZAWA Hiroyuki 已提交
1035
		return;
1036

K
KAMEZAWA Hiroyuki 已提交
1037
	pc = lookup_page_cgroup(page);
1038
	/* unused or root page is not rotated. */
1039 1040 1041 1042 1043
	if (!PageCgroupUsed(pc))
		return;
	/* Ensure pc->mem_cgroup is visible after reading PCG_USED. */
	smp_rmb();
	if (mem_cgroup_is_root(pc->mem_cgroup))
K
KAMEZAWA Hiroyuki 已提交
1044
		return;
1045
	mz = page_cgroup_zoneinfo(pc->mem_cgroup, page);
K
KAMEZAWA Hiroyuki 已提交
1046
	list_move(&pc->lru, &mz->lists[lru]);
1047 1048
}

K
KAMEZAWA Hiroyuki 已提交
1049
void mem_cgroup_add_lru_list(struct page *page, enum lru_list lru)
1050
{
K
KAMEZAWA Hiroyuki 已提交
1051 1052
	struct page_cgroup *pc;
	struct mem_cgroup_per_zone *mz;
1053

1054
	if (mem_cgroup_disabled())
K
KAMEZAWA Hiroyuki 已提交
1055 1056
		return;
	pc = lookup_page_cgroup(page);
1057
	VM_BUG_ON(PageCgroupAcctLRU(pc));
1058 1059 1060 1061 1062 1063 1064 1065 1066 1067
	/*
	 * putback:				charge:
	 * SetPageLRU				SetPageCgroupUsed
	 * smp_mb				smp_mb
	 * PageCgroupUsed && add to memcg LRU	PageLRU && add to memcg LRU
	 *
	 * Ensure that one of the two sides adds the page to the memcg
	 * LRU during a race.
	 */
	smp_mb();
K
KAMEZAWA Hiroyuki 已提交
1068
	if (!PageCgroupUsed(pc))
L
Lee Schermerhorn 已提交
1069
		return;
1070 1071
	/* Ensure pc->mem_cgroup is visible after reading PCG_USED. */
	smp_rmb();
1072
	mz = page_cgroup_zoneinfo(pc->mem_cgroup, page);
1073 1074
	/* huge page split is done under lru_lock. so, we have no races. */
	MEM_CGROUP_ZSTAT(mz, lru) += 1 << compound_order(page);
1075 1076 1077
	SetPageCgroupAcctLRU(pc);
	if (mem_cgroup_is_root(pc->mem_cgroup))
		return;
K
KAMEZAWA Hiroyuki 已提交
1078 1079
	list_add(&pc->lru, &mz->lists[lru]);
}
1080

K
KAMEZAWA Hiroyuki 已提交
1081
/*
1082 1083 1084 1085
 * At handling SwapCache and other FUSE stuff, pc->mem_cgroup may be changed
 * while it's linked to lru because the page may be reused after it's fully
 * uncharged. To handle that, unlink page_cgroup from LRU when charge it again.
 * It's done under lock_page and expected that zone->lru_lock isnever held.
K
KAMEZAWA Hiroyuki 已提交
1086
 */
1087
static void mem_cgroup_lru_del_before_commit(struct page *page)
K
KAMEZAWA Hiroyuki 已提交
1088
{
1089 1090 1091 1092
	unsigned long flags;
	struct zone *zone = page_zone(page);
	struct page_cgroup *pc = lookup_page_cgroup(page);

1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103
	/*
	 * Doing this check without taking ->lru_lock seems wrong but this
	 * is safe. Because if page_cgroup's USED bit is unset, the page
	 * will not be added to any memcg's LRU. If page_cgroup's USED bit is
	 * set, the commit after this will fail, anyway.
	 * This all charge/uncharge is done under some mutual execustion.
	 * So, we don't need to taking care of changes in USED bit.
	 */
	if (likely(!PageLRU(page)))
		return;

1104 1105 1106 1107 1108 1109 1110 1111
	spin_lock_irqsave(&zone->lru_lock, flags);
	/*
	 * Forget old LRU when this page_cgroup is *not* used. This Used bit
	 * is guarded by lock_page() because the page is SwapCache.
	 */
	if (!PageCgroupUsed(pc))
		mem_cgroup_del_lru_list(page, page_lru(page));
	spin_unlock_irqrestore(&zone->lru_lock, flags);
K
KAMEZAWA Hiroyuki 已提交
1112 1113
}

1114
static void mem_cgroup_lru_add_after_commit(struct page *page)
1115 1116 1117 1118
{
	unsigned long flags;
	struct zone *zone = page_zone(page);
	struct page_cgroup *pc = lookup_page_cgroup(page);
1119 1120 1121 1122 1123 1124 1125 1126 1127 1128
	/*
	 * putback:				charge:
	 * SetPageLRU				SetPageCgroupUsed
	 * smp_mb				smp_mb
	 * PageCgroupUsed && add to memcg LRU	PageLRU && add to memcg LRU
	 *
	 * Ensure that one of the two sides adds the page to the memcg
	 * LRU during a race.
	 */
	smp_mb();
1129 1130 1131
	/* taking care of that the page is added to LRU while we commit it */
	if (likely(!PageLRU(page)))
		return;
1132 1133
	spin_lock_irqsave(&zone->lru_lock, flags);
	/* link when the page is linked to LRU but page_cgroup isn't */
1134
	if (PageLRU(page) && !PageCgroupAcctLRU(pc))
1135 1136 1137 1138 1139
		mem_cgroup_add_lru_list(page, page_lru(page));
	spin_unlock_irqrestore(&zone->lru_lock, flags);
}


K
KAMEZAWA Hiroyuki 已提交
1140 1141 1142
void mem_cgroup_move_lists(struct page *page,
			   enum lru_list from, enum lru_list to)
{
1143
	if (mem_cgroup_disabled())
K
KAMEZAWA Hiroyuki 已提交
1144 1145 1146
		return;
	mem_cgroup_del_lru_list(page, from);
	mem_cgroup_add_lru_list(page, to);
1147 1148
}

1149
/*
1150
 * Checks whether given mem is same or in the root_mem_cgroup's
1151 1152
 * hierarchy subtree
 */
1153 1154
static bool mem_cgroup_same_or_subtree(const struct mem_cgroup *root_memcg,
		struct mem_cgroup *memcg)
1155
{
1156 1157 1158
	if (root_memcg != memcg) {
		return (root_memcg->use_hierarchy &&
			css_is_ancestor(&memcg->css, &root_memcg->css));
1159 1160 1161 1162 1163
	}

	return true;
}

1164
int task_in_mem_cgroup(struct task_struct *task, const struct mem_cgroup *memcg)
1165 1166
{
	int ret;
1167
	struct mem_cgroup *curr = NULL;
1168
	struct task_struct *p;
1169

1170 1171 1172 1173 1174
	p = find_lock_task_mm(task);
	if (!p)
		return 0;
	curr = try_get_mem_cgroup_from_mm(p->mm);
	task_unlock(p);
1175 1176
	if (!curr)
		return 0;
1177
	/*
1178
	 * We should check use_hierarchy of "memcg" not "curr". Because checking
1179
	 * use_hierarchy of "curr" here make this function true if hierarchy is
1180 1181
	 * enabled in "curr" and "curr" is a child of "memcg" in *cgroup*
	 * hierarchy(even if use_hierarchy is disabled in "memcg").
1182
	 */
1183
	ret = mem_cgroup_same_or_subtree(memcg, curr);
1184
	css_put(&curr->css);
1185 1186 1187
	return ret;
}

1188
int mem_cgroup_inactive_anon_is_low(struct mem_cgroup *memcg, struct zone *zone)
1189
{
1190 1191 1192
	unsigned long inactive_ratio;
	int nid = zone_to_nid(zone);
	int zid = zone_idx(zone);
1193
	unsigned long inactive;
1194
	unsigned long active;
1195
	unsigned long gb;
1196

1197 1198 1199 1200
	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));
1201

1202 1203 1204 1205 1206 1207
	gb = (inactive + active) >> (30 - PAGE_SHIFT);
	if (gb)
		inactive_ratio = int_sqrt(10 * gb);
	else
		inactive_ratio = 1;

1208
	return inactive * inactive_ratio < active;
1209 1210
}

1211
int mem_cgroup_inactive_file_is_low(struct mem_cgroup *memcg, struct zone *zone)
1212 1213 1214
{
	unsigned long active;
	unsigned long inactive;
1215 1216
	int zid = zone_idx(zone);
	int nid = zone_to_nid(zone);
1217

1218 1219 1220 1221
	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));
1222 1223 1224 1225

	return (active > inactive);
}

K
KOSAKI Motohiro 已提交
1226 1227 1228
struct zone_reclaim_stat *mem_cgroup_get_reclaim_stat(struct mem_cgroup *memcg,
						      struct zone *zone)
{
1229
	int nid = zone_to_nid(zone);
K
KOSAKI Motohiro 已提交
1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245
	int zid = zone_idx(zone);
	struct mem_cgroup_per_zone *mz = mem_cgroup_zoneinfo(memcg, nid, zid);

	return &mz->reclaim_stat;
}

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);
1246 1247
	if (!PageCgroupUsed(pc))
		return NULL;
1248 1249
	/* Ensure pc->mem_cgroup is visible after reading PCG_USED. */
	smp_rmb();
1250
	mz = page_cgroup_zoneinfo(pc->mem_cgroup, page);
K
KOSAKI Motohiro 已提交
1251 1252 1253
	return &mz->reclaim_stat;
}

1254 1255 1256
unsigned long mem_cgroup_isolate_pages(unsigned long nr_to_scan,
					struct list_head *dst,
					unsigned long *scanned, int order,
1257 1258
					isolate_mode_t mode,
					struct zone *z,
1259
					struct mem_cgroup *mem_cont,
1260
					int active, int file)
1261 1262 1263 1264 1265 1266
{
	unsigned long nr_taken = 0;
	struct page *page;
	unsigned long scan;
	LIST_HEAD(pc_list);
	struct list_head *src;
1267
	struct page_cgroup *pc, *tmp;
1268
	int nid = zone_to_nid(z);
1269 1270
	int zid = zone_idx(z);
	struct mem_cgroup_per_zone *mz;
1271
	int lru = LRU_FILE * file + active;
1272
	int ret;
1273

1274
	BUG_ON(!mem_cont);
1275
	mz = mem_cgroup_zoneinfo(mem_cont, nid, zid);
1276
	src = &mz->lists[lru];
1277

1278 1279
	scan = 0;
	list_for_each_entry_safe_reverse(pc, tmp, src, lru) {
H
Hugh Dickins 已提交
1280
		if (scan >= nr_to_scan)
1281
			break;
K
KAMEZAWA Hiroyuki 已提交
1282

1283 1284
		if (unlikely(!PageCgroupUsed(pc)))
			continue;
1285

1286
		page = lookup_cgroup_page(pc);
1287

H
Hugh Dickins 已提交
1288
		if (unlikely(!PageLRU(page)))
1289 1290
			continue;

H
Hugh Dickins 已提交
1291
		scan++;
1292 1293 1294
		ret = __isolate_lru_page(page, mode, file);
		switch (ret) {
		case 0:
1295
			list_move(&page->lru, dst);
1296
			mem_cgroup_del_lru(page);
1297
			nr_taken += hpage_nr_pages(page);
1298 1299 1300 1301 1302 1303 1304
			break;
		case -EBUSY:
			/* we don't affect global LRU but rotate in our LRU */
			mem_cgroup_rotate_lru_list(page, page_lru(page));
			break;
		default:
			break;
1305 1306 1307 1308
		}
	}

	*scanned = scan;
1309 1310 1311 1312

	trace_mm_vmscan_memcg_isolate(0, nr_to_scan, scan, nr_taken,
				      0, 0, 0, mode);

1313 1314 1315
	return nr_taken;
}

1316 1317 1318
#define mem_cgroup_from_res_counter(counter, member)	\
	container_of(counter, struct mem_cgroup, member)

1319
/**
1320 1321
 * mem_cgroup_margin - calculate chargeable space of a memory cgroup
 * @mem: the memory cgroup
1322
 *
1323
 * Returns the maximum amount of memory @mem can be charged with, in
1324
 * pages.
1325
 */
1326
static unsigned long mem_cgroup_margin(struct mem_cgroup *memcg)
1327
{
1328 1329
	unsigned long long margin;

1330
	margin = res_counter_margin(&memcg->res);
1331
	if (do_swap_account)
1332
		margin = min(margin, res_counter_margin(&memcg->memsw));
1333
	return margin >> PAGE_SHIFT;
1334 1335
}

1336
int mem_cgroup_swappiness(struct mem_cgroup *memcg)
K
KOSAKI Motohiro 已提交
1337 1338 1339 1340 1341 1342 1343
{
	struct cgroup *cgrp = memcg->css.cgroup;

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

1344
	return memcg->swappiness;
K
KOSAKI Motohiro 已提交
1345 1346
}

1347
static void mem_cgroup_start_move(struct mem_cgroup *memcg)
1348 1349
{
	int cpu;
1350 1351

	get_online_cpus();
1352
	spin_lock(&memcg->pcp_counter_lock);
1353
	for_each_online_cpu(cpu)
1354 1355 1356
		per_cpu(memcg->stat->count[MEM_CGROUP_ON_MOVE], cpu) += 1;
	memcg->nocpu_base.count[MEM_CGROUP_ON_MOVE] += 1;
	spin_unlock(&memcg->pcp_counter_lock);
1357
	put_online_cpus();
1358 1359 1360 1361

	synchronize_rcu();
}

1362
static void mem_cgroup_end_move(struct mem_cgroup *memcg)
1363 1364 1365
{
	int cpu;

1366
	if (!memcg)
1367
		return;
1368
	get_online_cpus();
1369
	spin_lock(&memcg->pcp_counter_lock);
1370
	for_each_online_cpu(cpu)
1371 1372 1373
		per_cpu(memcg->stat->count[MEM_CGROUP_ON_MOVE], cpu) -= 1;
	memcg->nocpu_base.count[MEM_CGROUP_ON_MOVE] -= 1;
	spin_unlock(&memcg->pcp_counter_lock);
1374
	put_online_cpus();
1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387
}
/*
 * 2 routines for checking "mem" is under move_account() or not.
 *
 * mem_cgroup_stealed() - checking a cgroup is mc.from or not. This is used
 *			  for avoiding race in accounting. If true,
 *			  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".
 */

1388
static bool mem_cgroup_stealed(struct mem_cgroup *memcg)
1389 1390
{
	VM_BUG_ON(!rcu_read_lock_held());
1391
	return this_cpu_read(memcg->stat->count[MEM_CGROUP_ON_MOVE]) > 0;
1392
}
1393

1394
static bool mem_cgroup_under_move(struct mem_cgroup *memcg)
1395
{
1396 1397
	struct mem_cgroup *from;
	struct mem_cgroup *to;
1398
	bool ret = false;
1399 1400 1401 1402 1403 1404 1405 1406 1407
	/*
	 * 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;
1408

1409 1410
	ret = mem_cgroup_same_or_subtree(memcg, from)
		|| mem_cgroup_same_or_subtree(memcg, to);
1411 1412
unlock:
	spin_unlock(&mc.lock);
1413 1414 1415
	return ret;
}

1416
static bool mem_cgroup_wait_acct_move(struct mem_cgroup *memcg)
1417 1418
{
	if (mc.moving_task && current != mc.moving_task) {
1419
		if (mem_cgroup_under_move(memcg)) {
1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431
			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;
}

1432
/**
1433
 * mem_cgroup_print_oom_info: Called from OOM with tasklist_lock held in read mode.
1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451
 * @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;

1452
	if (!memcg || !p)
1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498
		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));
}

1499 1500 1501 1502
/*
 * This function returns the number of memcg under hierarchy tree. Returns
 * 1(self count) if no children.
 */
1503
static int mem_cgroup_count_children(struct mem_cgroup *memcg)
1504 1505
{
	int num = 0;
K
KAMEZAWA Hiroyuki 已提交
1506 1507
	struct mem_cgroup *iter;

1508
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
1509
		num++;
1510 1511 1512
	return num;
}

D
David Rientjes 已提交
1513 1514 1515 1516 1517 1518 1519 1520
/*
 * Return the memory (and swap, if configured) limit for a memcg.
 */
u64 mem_cgroup_get_limit(struct mem_cgroup *memcg)
{
	u64 limit;
	u64 memsw;

1521 1522 1523
	limit = res_counter_read_u64(&memcg->res, RES_LIMIT);
	limit += total_swap_pages << PAGE_SHIFT;

D
David Rientjes 已提交
1524 1525 1526 1527 1528 1529 1530 1531
	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);
}

1532 1533 1534 1535 1536 1537 1538 1539 1540 1541
/**
 * 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.
 */
1542
static bool test_mem_cgroup_node_reclaimable(struct mem_cgroup *memcg,
1543 1544
		int nid, bool noswap)
{
1545
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_FILE))
1546 1547 1548
		return true;
	if (noswap || !total_swap_pages)
		return false;
1549
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_ANON))
1550 1551 1552 1553
		return true;
	return false;

}
1554 1555 1556 1557 1558 1559 1560 1561
#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.
 *
 */
1562
static void mem_cgroup_may_update_nodemask(struct mem_cgroup *memcg)
1563 1564
{
	int nid;
1565 1566 1567 1568
	/*
	 * numainfo_events > 0 means there was at least NUMAINFO_EVENTS_TARGET
	 * pagein/pageout changes since the last update.
	 */
1569
	if (!atomic_read(&memcg->numainfo_events))
1570
		return;
1571
	if (atomic_inc_return(&memcg->numainfo_updating) > 1)
1572 1573 1574
		return;

	/* make a nodemask where this memcg uses memory from */
1575
	memcg->scan_nodes = node_states[N_HIGH_MEMORY];
1576 1577 1578

	for_each_node_mask(nid, node_states[N_HIGH_MEMORY]) {

1579 1580
		if (!test_mem_cgroup_node_reclaimable(memcg, nid, false))
			node_clear(nid, memcg->scan_nodes);
1581
	}
1582

1583 1584
	atomic_set(&memcg->numainfo_events, 0);
	atomic_set(&memcg->numainfo_updating, 0);
1585 1586 1587 1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598
}

/*
 * 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.
 */
1599
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
1600 1601 1602
{
	int node;

1603 1604
	mem_cgroup_may_update_nodemask(memcg);
	node = memcg->last_scanned_node;
1605

1606
	node = next_node(node, memcg->scan_nodes);
1607
	if (node == MAX_NUMNODES)
1608
		node = first_node(memcg->scan_nodes);
1609 1610 1611 1612 1613 1614 1615 1616 1617
	/*
	 * 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();

1618
	memcg->last_scanned_node = node;
1619 1620 1621
	return node;
}

1622 1623 1624 1625 1626 1627
/*
 * 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.
 */
1628
bool mem_cgroup_reclaimable(struct mem_cgroup *memcg, bool noswap)
1629 1630 1631 1632 1633 1634 1635
{
	int nid;

	/*
	 * quick check...making use of scan_node.
	 * We can skip unused nodes.
	 */
1636 1637
	if (!nodes_empty(memcg->scan_nodes)) {
		for (nid = first_node(memcg->scan_nodes);
1638
		     nid < MAX_NUMNODES;
1639
		     nid = next_node(nid, memcg->scan_nodes)) {
1640

1641
			if (test_mem_cgroup_node_reclaimable(memcg, nid, noswap))
1642 1643 1644 1645 1646 1647 1648
				return true;
		}
	}
	/*
	 * Check rest of nodes.
	 */
	for_each_node_state(nid, N_HIGH_MEMORY) {
1649
		if (node_isset(nid, memcg->scan_nodes))
1650
			continue;
1651
		if (test_mem_cgroup_node_reclaimable(memcg, nid, noswap))
1652 1653 1654 1655 1656
			return true;
	}
	return false;
}

1657
#else
1658
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
1659 1660 1661
{
	return 0;
}
1662

1663
bool mem_cgroup_reclaimable(struct mem_cgroup *memcg, bool noswap)
1664
{
1665
	return test_mem_cgroup_node_reclaimable(memcg, 0, noswap);
1666
}
1667 1668
#endif

K
KAMEZAWA Hiroyuki 已提交
1669 1670 1671 1672
/*
 * Scan the hierarchy if needed to reclaim memory. We remember the last child
 * we reclaimed from, so that we don't end up penalizing one child extensively
 * based on its position in the children list.
1673
 *
1674
 * root_memcg is the original ancestor that we've been reclaim from.
K
KAMEZAWA Hiroyuki 已提交
1675
 *
1676
 * We give up and return to the caller when we visit root_memcg twice.
K
KAMEZAWA Hiroyuki 已提交
1677
 * (other groups can be removed while we're walking....)
1678 1679
 *
 * If shrink==true, for avoiding to free too much, this returns immedieately.
1680
 */
1681
static int mem_cgroup_hierarchical_reclaim(struct mem_cgroup *root_memcg,
1682
						struct zone *zone,
1683
						gfp_t gfp_mask,
1684 1685
						unsigned long reclaim_options,
						unsigned long *total_scanned)
1686
{
1687
	struct mem_cgroup *victim = NULL;
K
KAMEZAWA Hiroyuki 已提交
1688 1689
	int ret, total = 0;
	int loop = 0;
1690 1691
	bool noswap = reclaim_options & MEM_CGROUP_RECLAIM_NOSWAP;
	bool shrink = reclaim_options & MEM_CGROUP_RECLAIM_SHRINK;
1692
	bool check_soft = reclaim_options & MEM_CGROUP_RECLAIM_SOFT;
1693
	unsigned long excess;
1694
	unsigned long nr_scanned;
1695

1696
	excess = res_counter_soft_limit_excess(&root_memcg->res) >> PAGE_SHIFT;
K
KAMEZAWA Hiroyuki 已提交
1697

1698
	/* If memsw_is_minimum==1, swap-out is of-no-use. */
1699
	if (!check_soft && !shrink && root_memcg->memsw_is_minimum)
1700 1701
		noswap = true;

1702
	while (1) {
1703 1704
		victim = mem_cgroup_iter(root_memcg, victim, true);
		if (!victim) {
K
KAMEZAWA Hiroyuki 已提交
1705
			loop++;
1706 1707 1708 1709 1710 1711 1712
			/*
			 * We are not draining per cpu cached charges during
			 * soft limit reclaim  because global reclaim doesn't
			 * care about charges. It tries to free some memory and
			 * charges will not give any.
			 */
			if (!check_soft && loop >= 1)
1713
				drain_all_stock_async(root_memcg);
1714 1715 1716 1717 1718 1719
			if (loop >= 2) {
				/*
				 * If we have not been able to reclaim
				 * anything, it might because there are
				 * no reclaimable pages under this hierarchy
				 */
1720
				if (!check_soft || !total)
1721 1722
					break;
				/*
L
Lucas De Marchi 已提交
1723
				 * We want to do more targeted reclaim.
1724 1725 1726 1727 1728
				 * 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) ||
1729
					(loop > MEM_CGROUP_MAX_RECLAIM_LOOPS))
1730 1731
					break;
			}
1732
			continue;
1733
		}
1734
		if (!mem_cgroup_reclaimable(victim, noswap)) {
K
KAMEZAWA Hiroyuki 已提交
1735
			/* this cgroup's local usage == 0 */
1736 1737
			continue;
		}
K
KAMEZAWA Hiroyuki 已提交
1738
		/* we use swappiness of local cgroup */
1739
		if (check_soft) {
1740
			ret = mem_cgroup_shrink_node_zone(victim, gfp_mask,
1741 1742
				noswap, zone, &nr_scanned);
			*total_scanned += nr_scanned;
1743
		} else
1744
			ret = try_to_free_mem_cgroup_pages(victim, gfp_mask,
1745
						noswap);
1746
		total += ret;
1747 1748 1749 1750 1751 1752
		/*
		 * At shrinking usage, we can't check we should stop here or
		 * reclaim more. It's depends on callers. last_scanned_child
		 * will work enough for keeping fairness under tree.
		 */
		if (shrink)
1753
			break;
1754
		if (check_soft) {
1755
			if (!res_counter_soft_limit_excess(&root_memcg->res))
1756
				break;
1757
		} else if (mem_cgroup_margin(root_memcg))
1758
			break;
1759
	}
1760
	mem_cgroup_iter_break(root_memcg, victim);
K
KAMEZAWA Hiroyuki 已提交
1761
	return total;
1762 1763
}

K
KAMEZAWA Hiroyuki 已提交
1764 1765 1766
/*
 * Check OOM-Killer is already running under our hierarchy.
 * If someone is running, return false.
1767
 * Has to be called with memcg_oom_lock
K
KAMEZAWA Hiroyuki 已提交
1768
 */
1769
static bool mem_cgroup_oom_lock(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
1770
{
1771
	struct mem_cgroup *iter, *failed = NULL;
1772

1773
	for_each_mem_cgroup_tree(iter, memcg) {
1774
		if (iter->oom_lock) {
1775 1776 1777 1778 1779
			/*
			 * this subtree of our hierarchy is already locked
			 * so we cannot give a lock.
			 */
			failed = iter;
1780 1781
			mem_cgroup_iter_break(memcg, iter);
			break;
1782 1783
		} else
			iter->oom_lock = true;
K
KAMEZAWA Hiroyuki 已提交
1784
	}
K
KAMEZAWA Hiroyuki 已提交
1785

1786
	if (!failed)
1787
		return true;
1788 1789 1790 1791 1792

	/*
	 * OK, we failed to lock the whole subtree so we have to clean up
	 * what we set up to the failing subtree
	 */
1793
	for_each_mem_cgroup_tree(iter, memcg) {
1794
		if (iter == failed) {
1795 1796
			mem_cgroup_iter_break(memcg, iter);
			break;
1797 1798 1799
		}
		iter->oom_lock = false;
	}
1800
	return false;
1801
}
1802

1803
/*
1804
 * Has to be called with memcg_oom_lock
1805
 */
1806
static int mem_cgroup_oom_unlock(struct mem_cgroup *memcg)
1807
{
K
KAMEZAWA Hiroyuki 已提交
1808 1809
	struct mem_cgroup *iter;

1810
	for_each_mem_cgroup_tree(iter, memcg)
1811 1812 1813 1814
		iter->oom_lock = false;
	return 0;
}

1815
static void mem_cgroup_mark_under_oom(struct mem_cgroup *memcg)
1816 1817 1818
{
	struct mem_cgroup *iter;

1819
	for_each_mem_cgroup_tree(iter, memcg)
1820 1821 1822
		atomic_inc(&iter->under_oom);
}

1823
static void mem_cgroup_unmark_under_oom(struct mem_cgroup *memcg)
1824 1825 1826
{
	struct mem_cgroup *iter;

K
KAMEZAWA Hiroyuki 已提交
1827 1828 1829 1830 1831
	/*
	 * 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.
	 */
1832
	for_each_mem_cgroup_tree(iter, memcg)
1833
		atomic_add_unless(&iter->under_oom, -1, 0);
1834 1835
}

1836
static DEFINE_SPINLOCK(memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
1837 1838
static DECLARE_WAIT_QUEUE_HEAD(memcg_oom_waitq);

K
KAMEZAWA Hiroyuki 已提交
1839 1840 1841 1842 1843 1844 1845 1846
struct oom_wait_info {
	struct mem_cgroup *mem;
	wait_queue_t	wait;
};

static int memcg_oom_wake_function(wait_queue_t *wait,
	unsigned mode, int sync, void *arg)
{
1847 1848
	struct mem_cgroup *wake_memcg = (struct mem_cgroup *)arg,
			  *oom_wait_memcg;
K
KAMEZAWA Hiroyuki 已提交
1849 1850 1851
	struct oom_wait_info *oom_wait_info;

	oom_wait_info = container_of(wait, struct oom_wait_info, wait);
1852
	oom_wait_memcg = oom_wait_info->mem;
K
KAMEZAWA Hiroyuki 已提交
1853 1854 1855 1856 1857

	/*
	 * Both of oom_wait_info->mem and wake_mem are stable under us.
	 * Then we can use css_is_ancestor without taking care of RCU.
	 */
1858 1859
	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
1860 1861 1862 1863
		return 0;
	return autoremove_wake_function(wait, mode, sync, arg);
}

1864
static void memcg_wakeup_oom(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
1865
{
1866 1867
	/* for filtering, pass "memcg" as argument. */
	__wake_up(&memcg_oom_waitq, TASK_NORMAL, 0, memcg);
K
KAMEZAWA Hiroyuki 已提交
1868 1869
}

1870
static void memcg_oom_recover(struct mem_cgroup *memcg)
1871
{
1872 1873
	if (memcg && atomic_read(&memcg->under_oom))
		memcg_wakeup_oom(memcg);
1874 1875
}

K
KAMEZAWA Hiroyuki 已提交
1876 1877 1878
/*
 * try to call OOM killer. returns false if we should exit memory-reclaim loop.
 */
1879
bool mem_cgroup_handle_oom(struct mem_cgroup *memcg, gfp_t mask)
1880
{
K
KAMEZAWA Hiroyuki 已提交
1881
	struct oom_wait_info owait;
1882
	bool locked, need_to_kill;
K
KAMEZAWA Hiroyuki 已提交
1883

1884
	owait.mem = memcg;
K
KAMEZAWA Hiroyuki 已提交
1885 1886 1887 1888
	owait.wait.flags = 0;
	owait.wait.func = memcg_oom_wake_function;
	owait.wait.private = current;
	INIT_LIST_HEAD(&owait.wait.task_list);
1889
	need_to_kill = true;
1890
	mem_cgroup_mark_under_oom(memcg);
1891

1892
	/* At first, try to OOM lock hierarchy under memcg.*/
1893
	spin_lock(&memcg_oom_lock);
1894
	locked = mem_cgroup_oom_lock(memcg);
K
KAMEZAWA Hiroyuki 已提交
1895 1896 1897 1898 1899
	/*
	 * 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.
	 */
1900
	prepare_to_wait(&memcg_oom_waitq, &owait.wait, TASK_KILLABLE);
1901
	if (!locked || memcg->oom_kill_disable)
1902 1903
		need_to_kill = false;
	if (locked)
1904
		mem_cgroup_oom_notify(memcg);
1905
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
1906

1907 1908
	if (need_to_kill) {
		finish_wait(&memcg_oom_waitq, &owait.wait);
1909
		mem_cgroup_out_of_memory(memcg, mask);
1910
	} else {
K
KAMEZAWA Hiroyuki 已提交
1911
		schedule();
K
KAMEZAWA Hiroyuki 已提交
1912
		finish_wait(&memcg_oom_waitq, &owait.wait);
K
KAMEZAWA Hiroyuki 已提交
1913
	}
1914
	spin_lock(&memcg_oom_lock);
1915
	if (locked)
1916 1917
		mem_cgroup_oom_unlock(memcg);
	memcg_wakeup_oom(memcg);
1918
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
1919

1920
	mem_cgroup_unmark_under_oom(memcg);
1921

K
KAMEZAWA Hiroyuki 已提交
1922 1923 1924
	if (test_thread_flag(TIF_MEMDIE) || fatal_signal_pending(current))
		return false;
	/* Give chance to dying process */
1925
	schedule_timeout_uninterruptible(1);
K
KAMEZAWA Hiroyuki 已提交
1926
	return true;
1927 1928
}

1929 1930 1931
/*
 * Currently used to update mapped file statistics, but the routine can be
 * generalized to update other statistics as well.
1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950
 *
 * 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
 * small, we check MEM_CGROUP_ON_MOVE percpu value and detect there are
 * possibility of race condition. If there is, we take a lock.
1951
 */
1952

1953 1954
void mem_cgroup_update_page_stat(struct page *page,
				 enum mem_cgroup_page_stat_item idx, int val)
1955
{
1956
	struct mem_cgroup *memcg;
1957 1958
	struct page_cgroup *pc = lookup_page_cgroup(page);
	bool need_unlock = false;
1959
	unsigned long uninitialized_var(flags);
1960 1961 1962 1963

	if (unlikely(!pc))
		return;

1964
	rcu_read_lock();
1965 1966
	memcg = pc->mem_cgroup;
	if (unlikely(!memcg || !PageCgroupUsed(pc)))
1967 1968
		goto out;
	/* pc->mem_cgroup is unstable ? */
1969
	if (unlikely(mem_cgroup_stealed(memcg)) || PageTransHuge(page)) {
1970
		/* take a lock against to access pc->mem_cgroup */
1971
		move_lock_page_cgroup(pc, &flags);
1972
		need_unlock = true;
1973 1974
		memcg = pc->mem_cgroup;
		if (!memcg || !PageCgroupUsed(pc))
1975 1976
			goto out;
	}
1977 1978

	switch (idx) {
1979
	case MEMCG_NR_FILE_MAPPED:
1980 1981 1982
		if (val > 0)
			SetPageCgroupFileMapped(pc);
		else if (!page_mapped(page))
1983
			ClearPageCgroupFileMapped(pc);
1984
		idx = MEM_CGROUP_STAT_FILE_MAPPED;
1985 1986 1987
		break;
	default:
		BUG();
1988
	}
1989

1990
	this_cpu_add(memcg->stat->count[idx], val);
1991

1992 1993
out:
	if (unlikely(need_unlock))
1994
		move_unlock_page_cgroup(pc, &flags);
1995 1996
	rcu_read_unlock();
	return;
1997
}
1998
EXPORT_SYMBOL(mem_cgroup_update_page_stat);
1999

2000 2001 2002 2003
/*
 * size of first charge trial. "32" comes from vmscan.c's magic value.
 * TODO: maybe necessary to use big numbers in big irons.
 */
2004
#define CHARGE_BATCH	32U
2005 2006
struct memcg_stock_pcp {
	struct mem_cgroup *cached; /* this never be root cgroup */
2007
	unsigned int nr_pages;
2008
	struct work_struct work;
2009 2010
	unsigned long flags;
#define FLUSHING_CACHED_CHARGE	(0)
2011 2012
};
static DEFINE_PER_CPU(struct memcg_stock_pcp, memcg_stock);
2013
static DEFINE_MUTEX(percpu_charge_mutex);
2014 2015

/*
2016
 * Try to consume stocked charge on this cpu. If success, one page is consumed
2017 2018 2019 2020
 * 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.
 */
2021
static bool consume_stock(struct mem_cgroup *memcg)
2022 2023 2024 2025 2026
{
	struct memcg_stock_pcp *stock;
	bool ret = true;

	stock = &get_cpu_var(memcg_stock);
2027
	if (memcg == stock->cached && stock->nr_pages)
2028
		stock->nr_pages--;
2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041
	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;

2042 2043 2044 2045
	if (stock->nr_pages) {
		unsigned long bytes = stock->nr_pages * PAGE_SIZE;

		res_counter_uncharge(&old->res, bytes);
2046
		if (do_swap_account)
2047 2048
			res_counter_uncharge(&old->memsw, bytes);
		stock->nr_pages = 0;
2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060
	}
	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);
2061
	clear_bit(FLUSHING_CACHED_CHARGE, &stock->flags);
2062 2063 2064 2065
}

/*
 * Cache charges(val) which is from res_counter, to local per_cpu area.
2066
 * This will be consumed by consume_stock() function, later.
2067
 */
2068
static void refill_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
2069 2070 2071
{
	struct memcg_stock_pcp *stock = &get_cpu_var(memcg_stock);

2072
	if (stock->cached != memcg) { /* reset if necessary */
2073
		drain_stock(stock);
2074
		stock->cached = memcg;
2075
	}
2076
	stock->nr_pages += nr_pages;
2077 2078 2079 2080
	put_cpu_var(memcg_stock);
}

/*
2081
 * Drains all per-CPU charge caches for given root_memcg resp. subtree
2082 2083
 * of the hierarchy under it. sync flag says whether we should block
 * until the work is done.
2084
 */
2085
static void drain_all_stock(struct mem_cgroup *root_memcg, bool sync)
2086
{
2087
	int cpu, curcpu;
2088

2089 2090
	/* Notify other cpus that system-wide "drain" is running */
	get_online_cpus();
2091
	curcpu = get_cpu();
2092 2093
	for_each_online_cpu(cpu) {
		struct memcg_stock_pcp *stock = &per_cpu(memcg_stock, cpu);
2094
		struct mem_cgroup *memcg;
2095

2096 2097
		memcg = stock->cached;
		if (!memcg || !stock->nr_pages)
2098
			continue;
2099
		if (!mem_cgroup_same_or_subtree(root_memcg, memcg))
2100
			continue;
2101 2102 2103 2104 2105 2106
		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);
		}
2107
	}
2108
	put_cpu();
2109 2110 2111 2112 2113 2114

	if (!sync)
		goto out;

	for_each_online_cpu(cpu) {
		struct memcg_stock_pcp *stock = &per_cpu(memcg_stock, cpu);
2115
		if (test_bit(FLUSHING_CACHED_CHARGE, &stock->flags))
2116 2117 2118
			flush_work(&stock->work);
	}
out:
2119
 	put_online_cpus();
2120 2121 2122 2123 2124 2125 2126 2127
}

/*
 * 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.
 */
2128
static void drain_all_stock_async(struct mem_cgroup *root_memcg)
2129
{
2130 2131 2132 2133 2134
	/*
	 * If someone calls draining, avoid adding more kworker runs.
	 */
	if (!mutex_trylock(&percpu_charge_mutex))
		return;
2135
	drain_all_stock(root_memcg, false);
2136
	mutex_unlock(&percpu_charge_mutex);
2137 2138 2139
}

/* This is a synchronous drain interface. */
2140
static void drain_all_stock_sync(struct mem_cgroup *root_memcg)
2141 2142
{
	/* called when force_empty is called */
2143
	mutex_lock(&percpu_charge_mutex);
2144
	drain_all_stock(root_memcg, true);
2145
	mutex_unlock(&percpu_charge_mutex);
2146 2147
}

2148 2149 2150 2151
/*
 * This function drains percpu counter value from DEAD cpu and
 * move it to local cpu. Note that this function can be preempted.
 */
2152
static void mem_cgroup_drain_pcp_counter(struct mem_cgroup *memcg, int cpu)
2153 2154 2155
{
	int i;

2156
	spin_lock(&memcg->pcp_counter_lock);
2157
	for (i = 0; i < MEM_CGROUP_STAT_DATA; i++) {
2158
		long x = per_cpu(memcg->stat->count[i], cpu);
2159

2160 2161
		per_cpu(memcg->stat->count[i], cpu) = 0;
		memcg->nocpu_base.count[i] += x;
2162
	}
2163
	for (i = 0; i < MEM_CGROUP_EVENTS_NSTATS; i++) {
2164
		unsigned long x = per_cpu(memcg->stat->events[i], cpu);
2165

2166 2167
		per_cpu(memcg->stat->events[i], cpu) = 0;
		memcg->nocpu_base.events[i] += x;
2168
	}
2169
	/* need to clear ON_MOVE value, works as a kind of lock. */
2170 2171
	per_cpu(memcg->stat->count[MEM_CGROUP_ON_MOVE], cpu) = 0;
	spin_unlock(&memcg->pcp_counter_lock);
2172 2173
}

2174
static void synchronize_mem_cgroup_on_move(struct mem_cgroup *memcg, int cpu)
2175 2176 2177
{
	int idx = MEM_CGROUP_ON_MOVE;

2178 2179 2180
	spin_lock(&memcg->pcp_counter_lock);
	per_cpu(memcg->stat->count[idx], cpu) = memcg->nocpu_base.count[idx];
	spin_unlock(&memcg->pcp_counter_lock);
2181 2182 2183
}

static int __cpuinit memcg_cpu_hotplug_callback(struct notifier_block *nb,
2184 2185 2186 2187 2188
					unsigned long action,
					void *hcpu)
{
	int cpu = (unsigned long)hcpu;
	struct memcg_stock_pcp *stock;
2189
	struct mem_cgroup *iter;
2190

2191
	if ((action == CPU_ONLINE)) {
2192
		for_each_mem_cgroup(iter)
2193 2194 2195 2196
			synchronize_mem_cgroup_on_move(iter, cpu);
		return NOTIFY_OK;
	}

2197
	if ((action != CPU_DEAD) || action != CPU_DEAD_FROZEN)
2198
		return NOTIFY_OK;
2199

2200
	for_each_mem_cgroup(iter)
2201 2202
		mem_cgroup_drain_pcp_counter(iter, cpu);

2203 2204 2205 2206 2207
	stock = &per_cpu(memcg_stock, cpu);
	drain_stock(stock);
	return NOTIFY_OK;
}

2208 2209 2210 2211 2212 2213 2214 2215 2216 2217

/* 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 */
};

2218
static int mem_cgroup_do_charge(struct mem_cgroup *memcg, gfp_t gfp_mask,
2219
				unsigned int nr_pages, bool oom_check)
2220
{
2221
	unsigned long csize = nr_pages * PAGE_SIZE;
2222 2223 2224 2225 2226
	struct mem_cgroup *mem_over_limit;
	struct res_counter *fail_res;
	unsigned long flags = 0;
	int ret;

2227
	ret = res_counter_charge(&memcg->res, csize, &fail_res);
2228 2229 2230 2231

	if (likely(!ret)) {
		if (!do_swap_account)
			return CHARGE_OK;
2232
		ret = res_counter_charge(&memcg->memsw, csize, &fail_res);
2233 2234 2235
		if (likely(!ret))
			return CHARGE_OK;

2236
		res_counter_uncharge(&memcg->res, csize);
2237 2238 2239 2240
		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);
2241
	/*
2242 2243
	 * nr_pages can be either a huge page (HPAGE_PMD_NR), a batch
	 * of regular pages (CHARGE_BATCH), or a single regular page (1).
2244 2245 2246 2247
	 *
	 * Never reclaim on behalf of optional batching, retry with a
	 * single page instead.
	 */
2248
	if (nr_pages == CHARGE_BATCH)
2249 2250 2251 2252 2253 2254
		return CHARGE_RETRY;

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

	ret = mem_cgroup_hierarchical_reclaim(mem_over_limit, NULL,
2255
					      gfp_mask, flags, NULL);
2256
	if (mem_cgroup_margin(mem_over_limit) >= nr_pages)
2257
		return CHARGE_RETRY;
2258
	/*
2259 2260 2261 2262 2263 2264 2265
	 * 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.
2266
	 */
2267
	if (nr_pages == 1 && ret)
2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286
		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 */
	if (!mem_cgroup_handle_oom(mem_over_limit, gfp_mask))
		return CHARGE_OOM_DIE;

	return CHARGE_RETRY;
}

2287 2288 2289
/*
 * Unlike exported interface, "oom" parameter is added. if oom==true,
 * oom-killer can be invoked.
2290
 */
2291
static int __mem_cgroup_try_charge(struct mm_struct *mm,
A
Andrea Arcangeli 已提交
2292
				   gfp_t gfp_mask,
2293
				   unsigned int nr_pages,
2294
				   struct mem_cgroup **ptr,
2295
				   bool oom)
2296
{
2297
	unsigned int batch = max(CHARGE_BATCH, nr_pages);
2298
	int nr_oom_retries = MEM_CGROUP_RECLAIM_RETRIES;
2299
	struct mem_cgroup *memcg = NULL;
2300
	int ret;
2301

K
KAMEZAWA Hiroyuki 已提交
2302 2303 2304 2305 2306 2307 2308 2309
	/*
	 * 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;
2310

2311
	/*
2312 2313
	 * We always charge the cgroup the mm_struct belongs to.
	 * The mm_struct's mem_cgroup changes on task migration if the
2314 2315 2316
	 * thread group leader migrates. It's possible that mm is not
	 * set, if so charge the init_mm (happens for pagecache usage).
	 */
2317
	if (!*ptr && !mm)
K
KAMEZAWA Hiroyuki 已提交
2318 2319
		goto bypass;
again:
2320 2321 2322 2323
	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
2324
			goto done;
2325
		if (nr_pages == 1 && consume_stock(memcg))
K
KAMEZAWA Hiroyuki 已提交
2326
			goto done;
2327
		css_get(&memcg->css);
2328
	} else {
K
KAMEZAWA Hiroyuki 已提交
2329
		struct task_struct *p;
2330

K
KAMEZAWA Hiroyuki 已提交
2331 2332 2333
		rcu_read_lock();
		p = rcu_dereference(mm->owner);
		/*
2334
		 * Because we don't have task_lock(), "p" can exit.
2335
		 * In that case, "memcg" can point to root or p can be NULL with
2336 2337 2338 2339 2340 2341
		 * 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
2342
		 */
2343 2344
		memcg = mem_cgroup_from_task(p);
		if (!memcg || mem_cgroup_is_root(memcg)) {
K
KAMEZAWA Hiroyuki 已提交
2345 2346 2347
			rcu_read_unlock();
			goto done;
		}
2348
		if (nr_pages == 1 && consume_stock(memcg)) {
K
KAMEZAWA Hiroyuki 已提交
2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360
			/*
			 * 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 */
2361
		if (!css_tryget(&memcg->css)) {
K
KAMEZAWA Hiroyuki 已提交
2362 2363 2364 2365 2366
			rcu_read_unlock();
			goto again;
		}
		rcu_read_unlock();
	}
2367

2368 2369
	do {
		bool oom_check;
2370

2371
		/* If killed, bypass charge */
K
KAMEZAWA Hiroyuki 已提交
2372
		if (fatal_signal_pending(current)) {
2373
			css_put(&memcg->css);
2374
			goto bypass;
K
KAMEZAWA Hiroyuki 已提交
2375
		}
2376

2377 2378 2379 2380
		oom_check = false;
		if (oom && !nr_oom_retries) {
			oom_check = true;
			nr_oom_retries = MEM_CGROUP_RECLAIM_RETRIES;
2381
		}
2382

2383
		ret = mem_cgroup_do_charge(memcg, gfp_mask, batch, oom_check);
2384 2385 2386 2387
		switch (ret) {
		case CHARGE_OK:
			break;
		case CHARGE_RETRY: /* not in OOM situation but retry */
2388
			batch = nr_pages;
2389 2390
			css_put(&memcg->css);
			memcg = NULL;
K
KAMEZAWA Hiroyuki 已提交
2391
			goto again;
2392
		case CHARGE_WOULDBLOCK: /* !__GFP_WAIT */
2393
			css_put(&memcg->css);
2394 2395
			goto nomem;
		case CHARGE_NOMEM: /* OOM routine works */
K
KAMEZAWA Hiroyuki 已提交
2396
			if (!oom) {
2397
				css_put(&memcg->css);
K
KAMEZAWA Hiroyuki 已提交
2398
				goto nomem;
K
KAMEZAWA Hiroyuki 已提交
2399
			}
2400 2401 2402 2403
			/* If oom, we never return -ENOMEM */
			nr_oom_retries--;
			break;
		case CHARGE_OOM_DIE: /* Killed by OOM Killer */
2404
			css_put(&memcg->css);
K
KAMEZAWA Hiroyuki 已提交
2405
			goto bypass;
2406
		}
2407 2408
	} while (ret != CHARGE_OK);

2409
	if (batch > nr_pages)
2410 2411
		refill_stock(memcg, batch - nr_pages);
	css_put(&memcg->css);
2412
done:
2413
	*ptr = memcg;
2414 2415
	return 0;
nomem:
2416
	*ptr = NULL;
2417
	return -ENOMEM;
K
KAMEZAWA Hiroyuki 已提交
2418
bypass:
2419
	*ptr = NULL;
K
KAMEZAWA Hiroyuki 已提交
2420
	return 0;
2421
}
2422

2423 2424 2425 2426 2427
/*
 * 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().
 */
2428
static void __mem_cgroup_cancel_charge(struct mem_cgroup *memcg,
2429
				       unsigned int nr_pages)
2430
{
2431
	if (!mem_cgroup_is_root(memcg)) {
2432 2433
		unsigned long bytes = nr_pages * PAGE_SIZE;

2434
		res_counter_uncharge(&memcg->res, bytes);
2435
		if (do_swap_account)
2436
			res_counter_uncharge(&memcg->memsw, bytes);
2437
	}
2438 2439
}

2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458
/*
 * 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);
}

2459
struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page)
2460
{
2461
	struct mem_cgroup *memcg = NULL;
2462
	struct page_cgroup *pc;
2463
	unsigned short id;
2464 2465
	swp_entry_t ent;

2466 2467 2468
	VM_BUG_ON(!PageLocked(page));

	pc = lookup_page_cgroup(page);
2469
	lock_page_cgroup(pc);
2470
	if (PageCgroupUsed(pc)) {
2471 2472 2473
		memcg = pc->mem_cgroup;
		if (memcg && !css_tryget(&memcg->css))
			memcg = NULL;
2474
	} else if (PageSwapCache(page)) {
2475
		ent.val = page_private(page);
2476 2477
		id = lookup_swap_cgroup(ent);
		rcu_read_lock();
2478 2479 2480
		memcg = mem_cgroup_lookup(id);
		if (memcg && !css_tryget(&memcg->css))
			memcg = NULL;
2481
		rcu_read_unlock();
2482
	}
2483
	unlock_page_cgroup(pc);
2484
	return memcg;
2485 2486
}

2487
static void __mem_cgroup_commit_charge(struct mem_cgroup *memcg,
2488
				       struct page *page,
2489
				       unsigned int nr_pages,
2490
				       struct page_cgroup *pc,
2491
				       enum charge_type ctype)
2492
{
2493 2494 2495
	lock_page_cgroup(pc);
	if (unlikely(PageCgroupUsed(pc))) {
		unlock_page_cgroup(pc);
2496
		__mem_cgroup_cancel_charge(memcg, nr_pages);
2497 2498 2499 2500 2501 2502
		return;
	}
	/*
	 * we don't need page_cgroup_lock about tail pages, becase they are not
	 * accessed by any other context at this point.
	 */
2503
	pc->mem_cgroup = memcg;
2504 2505 2506 2507 2508 2509 2510
	/*
	 * 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
2511
	smp_wmb();
2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524
	switch (ctype) {
	case MEM_CGROUP_CHARGE_TYPE_CACHE:
	case MEM_CGROUP_CHARGE_TYPE_SHMEM:
		SetPageCgroupCache(pc);
		SetPageCgroupUsed(pc);
		break;
	case MEM_CGROUP_CHARGE_TYPE_MAPPED:
		ClearPageCgroupCache(pc);
		SetPageCgroupUsed(pc);
		break;
	default:
		break;
	}
2525

2526
	mem_cgroup_charge_statistics(memcg, PageCgroupCache(pc), nr_pages);
2527
	unlock_page_cgroup(pc);
2528 2529 2530 2531 2532
	/*
	 * "charge_statistics" updated event counter. Then, check it.
	 * Insert ancestor (and ancestor's ancestors), to softlimit RB-tree.
	 * if they exceeds softlimit.
	 */
2533
	memcg_check_events(memcg, page);
2534
}
2535

2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549
#ifdef CONFIG_TRANSPARENT_HUGEPAGE

#define PCGF_NOCOPY_AT_SPLIT ((1 << PCG_LOCK) | (1 << PCG_MOVE_LOCK) |\
			(1 << PCG_ACCT_LRU) | (1 << PCG_MIGRATION))
/*
 * Because tail pages are not marked as "used", set it. We're under
 * zone->lru_lock, 'splitting on pmd' and compund_lock.
 */
void mem_cgroup_split_huge_fixup(struct page *head, struct page *tail)
{
	struct page_cgroup *head_pc = lookup_page_cgroup(head);
	struct page_cgroup *tail_pc = lookup_page_cgroup(tail);
	unsigned long flags;

2550 2551
	if (mem_cgroup_disabled())
		return;
2552
	/*
2553
	 * We have no races with charge/uncharge but will have races with
2554 2555 2556 2557 2558 2559
	 * page state accounting.
	 */
	move_lock_page_cgroup(head_pc, &flags);

	tail_pc->mem_cgroup = head_pc->mem_cgroup;
	smp_wmb(); /* see __commit_charge() */
2560 2561 2562 2563 2564 2565 2566 2567 2568 2569
	if (PageCgroupAcctLRU(head_pc)) {
		enum lru_list lru;
		struct mem_cgroup_per_zone *mz;

		/*
		 * LRU flags cannot be copied because we need to add tail
		 *.page to LRU by generic call and our hook will be called.
		 * We hold lru_lock, then, reduce counter directly.
		 */
		lru = page_lru(head);
2570
		mz = page_cgroup_zoneinfo(head_pc->mem_cgroup, head);
2571 2572
		MEM_CGROUP_ZSTAT(mz, lru) -= 1;
	}
2573 2574 2575 2576 2577
	tail_pc->flags = head_pc->flags & ~PCGF_NOCOPY_AT_SPLIT;
	move_unlock_page_cgroup(head_pc, &flags);
}
#endif

2578
/**
2579
 * mem_cgroup_move_account - move account of the page
2580
 * @page: the page
2581
 * @nr_pages: number of regular pages (>1 for huge pages)
2582 2583 2584
 * @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.
2585
 * @uncharge: whether we should call uncharge and css_put against @from.
2586 2587
 *
 * The caller must confirm following.
K
KAMEZAWA Hiroyuki 已提交
2588
 * - page is not on LRU (isolate_page() is useful.)
2589
 * - compound_lock is held when nr_pages > 1
2590
 *
2591
 * This function doesn't do "charge" nor css_get to new cgroup. It should be
L
Lucas De Marchi 已提交
2592
 * done by a caller(__mem_cgroup_try_charge would be useful). If @uncharge is
2593 2594
 * true, this function does "uncharge" from old cgroup, but it doesn't if
 * @uncharge is false, so a caller should do "uncharge".
2595
 */
2596 2597 2598 2599 2600 2601
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)
2602
{
2603 2604
	unsigned long flags;
	int ret;
2605

2606
	VM_BUG_ON(from == to);
2607
	VM_BUG_ON(PageLRU(page));
2608 2609 2610 2611 2612 2613 2614
	/*
	 * 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;
2615
	if (nr_pages > 1 && !PageTransHuge(page))
2616 2617 2618 2619 2620 2621 2622 2623 2624
		goto out;

	lock_page_cgroup(pc);

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

	move_lock_page_cgroup(pc, &flags);
2625

2626
	if (PageCgroupFileMapped(pc)) {
2627 2628 2629 2630 2631
		/* 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();
2632
	}
2633
	mem_cgroup_charge_statistics(from, PageCgroupCache(pc), -nr_pages);
2634 2635
	if (uncharge)
		/* This is not "cancel", but cancel_charge does all we need. */
2636
		__mem_cgroup_cancel_charge(from, nr_pages);
2637

2638
	/* caller should have done css_get */
K
KAMEZAWA Hiroyuki 已提交
2639
	pc->mem_cgroup = to;
2640
	mem_cgroup_charge_statistics(to, PageCgroupCache(pc), nr_pages);
2641 2642 2643
	/*
	 * We charges against "to" which may not have any tasks. Then, "to"
	 * can be under rmdir(). But in current implementation, caller of
2644
	 * this function is just force_empty() and move charge, so it's
L
Lucas De Marchi 已提交
2645
	 * guaranteed that "to" is never removed. So, we don't check rmdir
2646
	 * status here.
2647
	 */
2648 2649 2650
	move_unlock_page_cgroup(pc, &flags);
	ret = 0;
unlock:
2651
	unlock_page_cgroup(pc);
2652 2653 2654
	/*
	 * check events
	 */
2655 2656
	memcg_check_events(to, page);
	memcg_check_events(from, page);
2657
out:
2658 2659 2660 2661 2662 2663 2664
	return ret;
}

/*
 * move charges to its parent.
 */

2665 2666
static int mem_cgroup_move_parent(struct page *page,
				  struct page_cgroup *pc,
2667 2668 2669 2670 2671 2672
				  struct mem_cgroup *child,
				  gfp_t gfp_mask)
{
	struct cgroup *cg = child->css.cgroup;
	struct cgroup *pcg = cg->parent;
	struct mem_cgroup *parent;
2673
	unsigned int nr_pages;
2674
	unsigned long uninitialized_var(flags);
2675 2676 2677 2678 2679 2680
	int ret;

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

2681 2682 2683 2684 2685
	ret = -EBUSY;
	if (!get_page_unless_zero(page))
		goto out;
	if (isolate_lru_page(page))
		goto put;
2686

2687
	nr_pages = hpage_nr_pages(page);
K
KAMEZAWA Hiroyuki 已提交
2688

2689
	parent = mem_cgroup_from_cont(pcg);
2690
	ret = __mem_cgroup_try_charge(NULL, gfp_mask, nr_pages, &parent, false);
2691
	if (ret || !parent)
2692
		goto put_back;
2693

2694
	if (nr_pages > 1)
2695 2696
		flags = compound_lock_irqsave(page);

2697
	ret = mem_cgroup_move_account(page, nr_pages, pc, child, parent, true);
2698
	if (ret)
2699
		__mem_cgroup_cancel_charge(parent, nr_pages);
2700

2701
	if (nr_pages > 1)
2702
		compound_unlock_irqrestore(page, flags);
2703
put_back:
K
KAMEZAWA Hiroyuki 已提交
2704
	putback_lru_page(page);
2705
put:
2706
	put_page(page);
2707
out:
2708 2709 2710
	return ret;
}

2711 2712 2713 2714 2715 2716 2717
/*
 * 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,
2718
				gfp_t gfp_mask, enum charge_type ctype)
2719
{
2720
	struct mem_cgroup *memcg = NULL;
2721
	unsigned int nr_pages = 1;
2722
	struct page_cgroup *pc;
2723
	bool oom = true;
2724
	int ret;
A
Andrea Arcangeli 已提交
2725

A
Andrea Arcangeli 已提交
2726
	if (PageTransHuge(page)) {
2727
		nr_pages <<= compound_order(page);
A
Andrea Arcangeli 已提交
2728
		VM_BUG_ON(!PageTransHuge(page));
2729 2730 2731 2732 2733
		/*
		 * 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
2734
	}
2735 2736

	pc = lookup_page_cgroup(page);
2737
	BUG_ON(!pc); /* XXX: remove this and move pc lookup into commit */
2738

2739 2740
	ret = __mem_cgroup_try_charge(mm, gfp_mask, nr_pages, &memcg, oom);
	if (ret || !memcg)
2741 2742
		return ret;

2743
	__mem_cgroup_commit_charge(memcg, page, nr_pages, pc, ctype);
2744 2745 2746
	return 0;
}

2747 2748
int mem_cgroup_newpage_charge(struct page *page,
			      struct mm_struct *mm, gfp_t gfp_mask)
2749
{
2750
	if (mem_cgroup_disabled())
2751
		return 0;
2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762
	/*
	 * If already mapped, we don't have to account.
	 * If page cache, page->mapping has address_space.
	 * But page->mapping may have out-of-use anon_vma pointer,
	 * detecit it by PageAnon() check. newly-mapped-anon's page->mapping
	 * is NULL.
  	 */
	if (page_mapped(page) || (page->mapping && !PageAnon(page)))
		return 0;
	if (unlikely(!mm))
		mm = &init_mm;
2763
	return mem_cgroup_charge_common(page, mm, gfp_mask,
2764
				MEM_CGROUP_CHARGE_TYPE_MAPPED);
2765 2766
}

D
Daisuke Nishimura 已提交
2767 2768 2769 2770
static void
__mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr,
					enum charge_type ctype);

2771
static void
2772
__mem_cgroup_commit_charge_lrucare(struct page *page, struct mem_cgroup *memcg,
2773 2774 2775 2776 2777 2778 2779 2780 2781
					enum charge_type ctype)
{
	struct page_cgroup *pc = lookup_page_cgroup(page);
	/*
	 * In some case, SwapCache, FUSE(splice_buf->radixtree), the page
	 * is already on LRU. It means the page may on some other page_cgroup's
	 * LRU. Take care of it.
	 */
	mem_cgroup_lru_del_before_commit(page);
2782
	__mem_cgroup_commit_charge(memcg, page, 1, pc, ctype);
2783 2784 2785 2786
	mem_cgroup_lru_add_after_commit(page);
	return;
}

2787 2788
int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm,
				gfp_t gfp_mask)
2789
{
2790
	struct mem_cgroup *memcg = NULL;
2791 2792
	int ret;

2793
	if (mem_cgroup_disabled())
2794
		return 0;
2795 2796
	if (PageCompound(page))
		return 0;
2797

2798
	if (unlikely(!mm))
2799
		mm = &init_mm;
2800

2801
	if (page_is_file_cache(page)) {
2802 2803
		ret = __mem_cgroup_try_charge(mm, gfp_mask, 1, &memcg, true);
		if (ret || !memcg)
2804
			return ret;
2805

2806 2807 2808 2809 2810
		/*
		 * FUSE reuses pages without going through the final
		 * put that would remove them from the LRU list, make
		 * sure that they get relinked properly.
		 */
2811
		__mem_cgroup_commit_charge_lrucare(page, memcg,
2812 2813 2814
					MEM_CGROUP_CHARGE_TYPE_CACHE);
		return ret;
	}
D
Daisuke Nishimura 已提交
2815 2816
	/* shmem */
	if (PageSwapCache(page)) {
2817
		ret = mem_cgroup_try_charge_swapin(mm, page, gfp_mask, &memcg);
D
Daisuke Nishimura 已提交
2818
		if (!ret)
2819
			__mem_cgroup_commit_charge_swapin(page, memcg,
D
Daisuke Nishimura 已提交
2820 2821 2822
					MEM_CGROUP_CHARGE_TYPE_SHMEM);
	} else
		ret = mem_cgroup_charge_common(page, mm, gfp_mask,
2823
					MEM_CGROUP_CHARGE_TYPE_SHMEM);
2824 2825

	return ret;
2826 2827
}

2828 2829 2830
/*
 * While swap-in, try_charge -> commit or cancel, the page is locked.
 * And when try_charge() successfully returns, one refcnt to memcg without
2831
 * struct page_cgroup is acquired. This refcnt will be consumed by
2832 2833
 * "commit()" or removed by "cancel()"
 */
2834 2835 2836 2837
int mem_cgroup_try_charge_swapin(struct mm_struct *mm,
				 struct page *page,
				 gfp_t mask, struct mem_cgroup **ptr)
{
2838
	struct mem_cgroup *memcg;
2839
	int ret;
2840

2841 2842
	*ptr = NULL;

2843
	if (mem_cgroup_disabled())
2844 2845 2846 2847 2848 2849
		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
2850 2851 2852
	 * 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.
2853 2854
	 */
	if (!PageSwapCache(page))
H
Hugh Dickins 已提交
2855
		goto charge_cur_mm;
2856 2857
	memcg = try_get_mem_cgroup_from_page(page);
	if (!memcg)
2858
		goto charge_cur_mm;
2859
	*ptr = memcg;
2860
	ret = __mem_cgroup_try_charge(NULL, mask, 1, ptr, true);
2861
	css_put(&memcg->css);
2862
	return ret;
2863 2864 2865
charge_cur_mm:
	if (unlikely(!mm))
		mm = &init_mm;
2866
	return __mem_cgroup_try_charge(mm, mask, 1, ptr, true);
2867 2868
}

D
Daisuke Nishimura 已提交
2869 2870 2871
static void
__mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr,
					enum charge_type ctype)
2872
{
2873
	if (mem_cgroup_disabled())
2874 2875 2876
		return;
	if (!ptr)
		return;
2877
	cgroup_exclude_rmdir(&ptr->css);
2878 2879

	__mem_cgroup_commit_charge_lrucare(page, ptr, ctype);
2880 2881 2882
	/*
	 * Now swap is on-memory. This means this page may be
	 * counted both as mem and swap....double count.
2883 2884 2885
	 * 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.
2886
	 */
2887
	if (do_swap_account && PageSwapCache(page)) {
2888
		swp_entry_t ent = {.val = page_private(page)};
2889
		unsigned short id;
2890
		struct mem_cgroup *memcg;
2891 2892 2893 2894

		id = swap_cgroup_record(ent, 0);
		rcu_read_lock();
		memcg = mem_cgroup_lookup(id);
2895
		if (memcg) {
2896 2897 2898 2899
			/*
			 * This recorded memcg can be obsolete one. So, avoid
			 * calling css_tryget
			 */
2900
			if (!mem_cgroup_is_root(memcg))
2901
				res_counter_uncharge(&memcg->memsw, PAGE_SIZE);
2902
			mem_cgroup_swap_statistics(memcg, false);
2903 2904
			mem_cgroup_put(memcg);
		}
2905
		rcu_read_unlock();
2906
	}
2907 2908 2909 2910 2911 2912
	/*
	 * 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.
	 */
	cgroup_release_and_wakeup_rmdir(&ptr->css);
2913 2914
}

D
Daisuke Nishimura 已提交
2915 2916 2917 2918 2919 2920
void mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr)
{
	__mem_cgroup_commit_charge_swapin(page, ptr,
					MEM_CGROUP_CHARGE_TYPE_MAPPED);
}

2921
void mem_cgroup_cancel_charge_swapin(struct mem_cgroup *memcg)
2922
{
2923
	if (mem_cgroup_disabled())
2924
		return;
2925
	if (!memcg)
2926
		return;
2927
	__mem_cgroup_cancel_charge(memcg, 1);
2928 2929
}

2930
static void mem_cgroup_do_uncharge(struct mem_cgroup *memcg,
2931 2932
				   unsigned int nr_pages,
				   const enum charge_type ctype)
2933 2934 2935
{
	struct memcg_batch_info *batch = NULL;
	bool uncharge_memsw = true;
2936

2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947
	/* 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)
2948
		batch->memcg = memcg;
2949 2950
	/*
	 * do_batch > 0 when unmapping pages or inode invalidate/truncate.
L
Lucas De Marchi 已提交
2951
	 * In those cases, all pages freed continuously can be expected to be in
2952 2953 2954 2955 2956 2957 2958 2959
	 * 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;

2960
	if (nr_pages > 1)
A
Andrea Arcangeli 已提交
2961 2962
		goto direct_uncharge;

2963 2964 2965 2966 2967
	/*
	 * 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.
	 */
2968
	if (batch->memcg != memcg)
2969 2970
		goto direct_uncharge;
	/* remember freed charge and uncharge it later */
2971
	batch->nr_pages++;
2972
	if (uncharge_memsw)
2973
		batch->memsw_nr_pages++;
2974 2975
	return;
direct_uncharge:
2976
	res_counter_uncharge(&memcg->res, nr_pages * PAGE_SIZE);
2977
	if (uncharge_memsw)
2978 2979 2980
		res_counter_uncharge(&memcg->memsw, nr_pages * PAGE_SIZE);
	if (unlikely(batch->memcg != memcg))
		memcg_oom_recover(memcg);
2981 2982
	return;
}
2983

2984
/*
2985
 * uncharge if !page_mapped(page)
2986
 */
2987
static struct mem_cgroup *
2988
__mem_cgroup_uncharge_common(struct page *page, enum charge_type ctype)
2989
{
2990
	struct mem_cgroup *memcg = NULL;
2991 2992
	unsigned int nr_pages = 1;
	struct page_cgroup *pc;
2993

2994
	if (mem_cgroup_disabled())
2995
		return NULL;
2996

K
KAMEZAWA Hiroyuki 已提交
2997
	if (PageSwapCache(page))
2998
		return NULL;
K
KAMEZAWA Hiroyuki 已提交
2999

A
Andrea Arcangeli 已提交
3000
	if (PageTransHuge(page)) {
3001
		nr_pages <<= compound_order(page);
A
Andrea Arcangeli 已提交
3002 3003
		VM_BUG_ON(!PageTransHuge(page));
	}
3004
	/*
3005
	 * Check if our page_cgroup is valid
3006
	 */
3007 3008
	pc = lookup_page_cgroup(page);
	if (unlikely(!pc || !PageCgroupUsed(pc)))
3009
		return NULL;
3010

3011
	lock_page_cgroup(pc);
K
KAMEZAWA Hiroyuki 已提交
3012

3013
	memcg = pc->mem_cgroup;
3014

K
KAMEZAWA Hiroyuki 已提交
3015 3016 3017 3018 3019
	if (!PageCgroupUsed(pc))
		goto unlock_out;

	switch (ctype) {
	case MEM_CGROUP_CHARGE_TYPE_MAPPED:
K
KAMEZAWA Hiroyuki 已提交
3020
	case MEM_CGROUP_CHARGE_TYPE_DROP:
3021 3022
		/* See mem_cgroup_prepare_migration() */
		if (page_mapped(page) || PageCgroupMigration(pc))
K
KAMEZAWA Hiroyuki 已提交
3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033
			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;
3034
	}
K
KAMEZAWA Hiroyuki 已提交
3035

3036
	mem_cgroup_charge_statistics(memcg, PageCgroupCache(pc), -nr_pages);
K
KAMEZAWA Hiroyuki 已提交
3037

3038
	ClearPageCgroupUsed(pc);
3039 3040 3041 3042 3043 3044
	/*
	 * 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.
	 */
3045

3046
	unlock_page_cgroup(pc);
K
KAMEZAWA Hiroyuki 已提交
3047
	/*
3048
	 * even after unlock, we have memcg->res.usage here and this memcg
K
KAMEZAWA Hiroyuki 已提交
3049 3050
	 * will never be freed.
	 */
3051
	memcg_check_events(memcg, page);
K
KAMEZAWA Hiroyuki 已提交
3052
	if (do_swap_account && ctype == MEM_CGROUP_CHARGE_TYPE_SWAPOUT) {
3053 3054
		mem_cgroup_swap_statistics(memcg, true);
		mem_cgroup_get(memcg);
K
KAMEZAWA Hiroyuki 已提交
3055
	}
3056 3057
	if (!mem_cgroup_is_root(memcg))
		mem_cgroup_do_uncharge(memcg, nr_pages, ctype);
3058

3059
	return memcg;
K
KAMEZAWA Hiroyuki 已提交
3060 3061 3062

unlock_out:
	unlock_page_cgroup(pc);
3063
	return NULL;
3064 3065
}

3066 3067
void mem_cgroup_uncharge_page(struct page *page)
{
3068 3069 3070 3071 3072
	/* early check. */
	if (page_mapped(page))
		return;
	if (page->mapping && !PageAnon(page))
		return;
3073 3074 3075 3076 3077 3078
	__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));
3079
	VM_BUG_ON(page->mapping);
3080 3081 3082
	__mem_cgroup_uncharge_common(page, MEM_CGROUP_CHARGE_TYPE_CACHE);
}

3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096
/*
 * 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;
3097 3098
		current->memcg_batch.nr_pages = 0;
		current->memcg_batch.memsw_nr_pages = 0;
3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118
	}
}

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.
	 */
3119 3120 3121 3122 3123 3124
	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);
3125
	memcg_oom_recover(batch->memcg);
3126 3127 3128 3129
	/* forget this pointer (for sanity check) */
	batch->memcg = NULL;
}

3130
#ifdef CONFIG_SWAP
3131
/*
3132
 * called after __delete_from_swap_cache() and drop "page" account.
3133 3134
 * memcg information is recorded to swap_cgroup of "ent"
 */
K
KAMEZAWA Hiroyuki 已提交
3135 3136
void
mem_cgroup_uncharge_swapcache(struct page *page, swp_entry_t ent, bool swapout)
3137 3138
{
	struct mem_cgroup *memcg;
K
KAMEZAWA Hiroyuki 已提交
3139 3140 3141 3142 3143 3144
	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);
3145

K
KAMEZAWA Hiroyuki 已提交
3146 3147 3148 3149 3150
	/*
	 * record memcg information,  if swapout && memcg != NULL,
	 * mem_cgroup_get() was called in uncharge().
	 */
	if (do_swap_account && swapout && memcg)
3151
		swap_cgroup_record(ent, css_id(&memcg->css));
3152
}
3153
#endif
3154 3155 3156 3157 3158 3159 3160

#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
3161
{
3162
	struct mem_cgroup *memcg;
3163
	unsigned short id;
3164 3165 3166 3167

	if (!do_swap_account)
		return;

3168 3169 3170
	id = swap_cgroup_record(ent, 0);
	rcu_read_lock();
	memcg = mem_cgroup_lookup(id);
3171
	if (memcg) {
3172 3173 3174 3175
		/*
		 * We uncharge this because swap is freed.
		 * This memcg can be obsolete one. We avoid calling css_tryget
		 */
3176
		if (!mem_cgroup_is_root(memcg))
3177
			res_counter_uncharge(&memcg->memsw, PAGE_SIZE);
3178
		mem_cgroup_swap_statistics(memcg, false);
3179 3180
		mem_cgroup_put(memcg);
	}
3181
	rcu_read_unlock();
K
KAMEZAWA Hiroyuki 已提交
3182
}
3183 3184 3185 3186 3187 3188

/**
 * 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
3189
 * @need_fixup: whether we should fixup res_counters and refcounts.
3190 3191 3192 3193 3194 3195 3196 3197 3198 3199
 *
 * 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,
3200
		struct mem_cgroup *from, struct mem_cgroup *to, bool need_fixup)
3201 3202 3203 3204 3205 3206 3207 3208
{
	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);
3209
		mem_cgroup_swap_statistics(to, true);
3210
		/*
3211 3212 3213 3214 3215 3216
		 * 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.
3217 3218
		 */
		mem_cgroup_get(to);
3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229
		if (need_fixup) {
			if (!mem_cgroup_is_root(from))
				res_counter_uncharge(&from->memsw, PAGE_SIZE);
			mem_cgroup_put(from);
			/*
			 * we charged both to->res and to->memsw, so we should
			 * uncharge to->res.
			 */
			if (!mem_cgroup_is_root(to))
				res_counter_uncharge(&to->res, PAGE_SIZE);
		}
3230 3231 3232 3233 3234 3235
		return 0;
	}
	return -EINVAL;
}
#else
static inline int mem_cgroup_move_swap_account(swp_entry_t entry,
3236
		struct mem_cgroup *from, struct mem_cgroup *to, bool need_fixup)
3237 3238 3239
{
	return -EINVAL;
}
3240
#endif
K
KAMEZAWA Hiroyuki 已提交
3241

3242
/*
3243 3244
 * Before starting migration, account PAGE_SIZE to mem_cgroup that the old
 * page belongs to.
3245
 */
3246
int mem_cgroup_prepare_migration(struct page *page,
3247
	struct page *newpage, struct mem_cgroup **ptr, gfp_t gfp_mask)
3248
{
3249
	struct mem_cgroup *memcg = NULL;
3250
	struct page_cgroup *pc;
3251
	enum charge_type ctype;
3252
	int ret = 0;
3253

3254 3255
	*ptr = NULL;

A
Andrea Arcangeli 已提交
3256
	VM_BUG_ON(PageTransHuge(page));
3257
	if (mem_cgroup_disabled())
3258 3259
		return 0;

3260 3261 3262
	pc = lookup_page_cgroup(page);
	lock_page_cgroup(pc);
	if (PageCgroupUsed(pc)) {
3263 3264
		memcg = pc->mem_cgroup;
		css_get(&memcg->css);
3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295
		/*
		 * 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);
3296
	}
3297
	unlock_page_cgroup(pc);
3298 3299 3300 3301
	/*
	 * If the page is not charged at this point,
	 * we return here.
	 */
3302
	if (!memcg)
3303
		return 0;
3304

3305
	*ptr = memcg;
3306
	ret = __mem_cgroup_try_charge(NULL, gfp_mask, 1, ptr, false);
3307
	css_put(&memcg->css);/* drop extra refcnt */
3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318
	if (ret || *ptr == NULL) {
		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);
		}
		return -ENOMEM;
3319
	}
3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332
	/*
	 * 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().
	 */
	pc = lookup_page_cgroup(newpage);
	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;
3333
	__mem_cgroup_commit_charge(memcg, page, 1, pc, ctype);
3334
	return ret;
3335
}
3336

3337
/* remove redundant charge if migration failed*/
3338
void mem_cgroup_end_migration(struct mem_cgroup *memcg,
3339
	struct page *oldpage, struct page *newpage, bool migration_ok)
3340
{
3341
	struct page *used, *unused;
3342 3343
	struct page_cgroup *pc;

3344
	if (!memcg)
3345
		return;
3346
	/* blocks rmdir() */
3347
	cgroup_exclude_rmdir(&memcg->css);
3348
	if (!migration_ok) {
3349 3350
		used = oldpage;
		unused = newpage;
3351
	} else {
3352
		used = newpage;
3353 3354
		unused = oldpage;
	}
3355
	/*
3356 3357 3358
	 * 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.
3359
	 */
3360 3361 3362 3363
	pc = lookup_page_cgroup(oldpage);
	lock_page_cgroup(pc);
	ClearPageCgroupMigration(pc);
	unlock_page_cgroup(pc);
3364

3365 3366
	__mem_cgroup_uncharge_common(unused, MEM_CGROUP_CHARGE_TYPE_FORCE);

3367
	/*
3368 3369 3370 3371 3372 3373
	 * 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)
3374
	 */
3375 3376
	if (PageAnon(used))
		mem_cgroup_uncharge_page(used);
3377
	/*
3378 3379
	 * At migration, we may charge account against cgroup which has no
	 * tasks.
3380 3381 3382
	 * 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.
	 */
3383
	cgroup_release_and_wakeup_rmdir(&memcg->css);
3384
}
3385

3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429
/*
 * 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)
{
	struct mem_cgroup *memcg;
	struct page_cgroup *pc;
	struct zone *zone;
	enum charge_type type = MEM_CGROUP_CHARGE_TYPE_CACHE;
	unsigned long flags;

	if (mem_cgroup_disabled())
		return;

	pc = lookup_page_cgroup(oldpage);
	/* fix accounting on old pages */
	lock_page_cgroup(pc);
	memcg = pc->mem_cgroup;
	mem_cgroup_charge_statistics(memcg, PageCgroupCache(pc), -1);
	ClearPageCgroupUsed(pc);
	unlock_page_cgroup(pc);

	if (PageSwapBacked(oldpage))
		type = MEM_CGROUP_CHARGE_TYPE_SHMEM;

	zone = page_zone(newpage);
	pc = lookup_page_cgroup(newpage);
	/*
	 * 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.
	 */
	spin_lock_irqsave(&zone->lru_lock, flags);
	if (PageLRU(newpage))
		del_page_from_lru_list(zone, newpage, page_lru(newpage));
	__mem_cgroup_commit_charge(memcg, newpage, 1, pc, type);
	if (PageLRU(newpage))
		add_page_to_lru_list(zone, newpage, page_lru(newpage));
	spin_unlock_irqrestore(&zone->lru_lock, flags);
}

3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475
#ifdef CONFIG_DEBUG_VM
static struct page_cgroup *lookup_page_cgroup_used(struct page *page)
{
	struct page_cgroup *pc;

	pc = lookup_page_cgroup(page);
	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) {
		int ret = -1;
		char *path;

		printk(KERN_ALERT "pc:%p pc->flags:%lx pc->mem_cgroup:%p",
		       pc, pc->flags, pc->mem_cgroup);

		path = kmalloc(PATH_MAX, GFP_KERNEL);
		if (path) {
			rcu_read_lock();
			ret = cgroup_path(pc->mem_cgroup->css.cgroup,
							path, PATH_MAX);
			rcu_read_unlock();
		}

		printk(KERN_CONT "(%s)\n",
				(ret < 0) ? "cannot get the path" : path);
		kfree(path);
	}
}
#endif

3476 3477
static DEFINE_MUTEX(set_limit_mutex);

3478
static int mem_cgroup_resize_limit(struct mem_cgroup *memcg,
3479
				unsigned long long val)
3480
{
3481
	int retry_count;
3482
	u64 memswlimit, memlimit;
3483
	int ret = 0;
3484 3485
	int children = mem_cgroup_count_children(memcg);
	u64 curusage, oldusage;
3486
	int enlarge;
3487 3488 3489 3490 3491 3492 3493 3494 3495

	/*
	 * 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);
3496

3497
	enlarge = 0;
3498
	while (retry_count) {
3499 3500 3501 3502
		if (signal_pending(current)) {
			ret = -EINTR;
			break;
		}
3503 3504 3505
		/*
		 * Rather than hide all in some function, I do this in
		 * open coded manner. You see what this really does.
3506
		 * We have to guarantee memcg->res.limit < memcg->memsw.limit.
3507 3508 3509 3510 3511 3512
		 */
		mutex_lock(&set_limit_mutex);
		memswlimit = res_counter_read_u64(&memcg->memsw, RES_LIMIT);
		if (memswlimit < val) {
			ret = -EINVAL;
			mutex_unlock(&set_limit_mutex);
3513 3514
			break;
		}
3515 3516 3517 3518 3519

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

3520
		ret = res_counter_set_limit(&memcg->res, val);
3521 3522 3523 3524 3525 3526
		if (!ret) {
			if (memswlimit == val)
				memcg->memsw_is_minimum = true;
			else
				memcg->memsw_is_minimum = false;
		}
3527 3528 3529 3530 3531
		mutex_unlock(&set_limit_mutex);

		if (!ret)
			break;

3532
		mem_cgroup_hierarchical_reclaim(memcg, NULL, GFP_KERNEL,
3533 3534
						MEM_CGROUP_RECLAIM_SHRINK,
						NULL);
3535 3536 3537 3538 3539 3540
		curusage = res_counter_read_u64(&memcg->res, RES_USAGE);
		/* Usage is reduced ? */
  		if (curusage >= oldusage)
			retry_count--;
		else
			oldusage = curusage;
3541
	}
3542 3543
	if (!ret && enlarge)
		memcg_oom_recover(memcg);
3544

3545 3546 3547
	return ret;
}

L
Li Zefan 已提交
3548 3549
static int mem_cgroup_resize_memsw_limit(struct mem_cgroup *memcg,
					unsigned long long val)
3550
{
3551
	int retry_count;
3552
	u64 memlimit, memswlimit, oldusage, curusage;
3553 3554
	int children = mem_cgroup_count_children(memcg);
	int ret = -EBUSY;
3555
	int enlarge = 0;
3556

3557 3558 3559
	/* see mem_cgroup_resize_res_limit */
 	retry_count = children * MEM_CGROUP_RECLAIM_RETRIES;
	oldusage = res_counter_read_u64(&memcg->memsw, RES_USAGE);
3560 3561 3562 3563 3564 3565 3566 3567
	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.
3568
		 * We have to guarantee memcg->res.limit < memcg->memsw.limit.
3569 3570 3571 3572 3573 3574 3575 3576
		 */
		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;
		}
3577 3578 3579
		memswlimit = res_counter_read_u64(&memcg->memsw, RES_LIMIT);
		if (memswlimit < val)
			enlarge = 1;
3580
		ret = res_counter_set_limit(&memcg->memsw, val);
3581 3582 3583 3584 3585 3586
		if (!ret) {
			if (memlimit == val)
				memcg->memsw_is_minimum = true;
			else
				memcg->memsw_is_minimum = false;
		}
3587 3588 3589 3590 3591
		mutex_unlock(&set_limit_mutex);

		if (!ret)
			break;

3592
		mem_cgroup_hierarchical_reclaim(memcg, NULL, GFP_KERNEL,
3593
						MEM_CGROUP_RECLAIM_NOSWAP |
3594 3595
						MEM_CGROUP_RECLAIM_SHRINK,
						NULL);
3596
		curusage = res_counter_read_u64(&memcg->memsw, RES_USAGE);
3597
		/* Usage is reduced ? */
3598
		if (curusage >= oldusage)
3599
			retry_count--;
3600 3601
		else
			oldusage = curusage;
3602
	}
3603 3604
	if (!ret && enlarge)
		memcg_oom_recover(memcg);
3605 3606 3607
	return ret;
}

3608
unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
3609 3610
					    gfp_t gfp_mask,
					    unsigned long *total_scanned)
3611 3612 3613 3614 3615 3616
{
	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;
3617
	unsigned long long excess;
3618
	unsigned long nr_scanned;
3619 3620 3621 3622

	if (order > 0)
		return 0;

3623
	mctz = soft_limit_tree_node_zone(zone_to_nid(zone), zone_idx(zone));
3624 3625 3626 3627 3628 3629 3630 3631 3632 3633 3634 3635 3636
	/*
	 * 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;

3637
		nr_scanned = 0;
3638 3639
		reclaimed = mem_cgroup_hierarchical_reclaim(mz->mem, zone,
						gfp_mask,
3640 3641
						MEM_CGROUP_RECLAIM_SOFT,
						&nr_scanned);
3642
		nr_reclaimed += reclaimed;
3643
		*total_scanned += nr_scanned;
3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665
		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);
3666
				if (next_mz == mz)
3667
					css_put(&next_mz->mem->css);
3668
				else /* next_mz == NULL or other memcg */
3669 3670 3671 3672
					break;
			} while (1);
		}
		__mem_cgroup_remove_exceeded(mz->mem, mz, mctz);
3673
		excess = res_counter_soft_limit_excess(&mz->mem->res);
3674 3675 3676 3677 3678 3679 3680 3681
		/*
		 * 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.
		 */
3682 3683
		/* If excess == 0, no tree ops */
		__mem_cgroup_insert_exceeded(mz->mem, mz, mctz, excess);
3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701
		spin_unlock(&mctz->lock);
		css_put(&mz->mem->css);
		loop++;
		/*
		 * Could not reclaim anything and there are no more
		 * mem cgroups to try or we seem to be looping without
		 * reclaiming anything.
		 */
		if (!nr_reclaimed &&
			(next_mz == NULL ||
			loop > MEM_CGROUP_MAX_SOFT_LIMIT_RECLAIM_LOOPS))
			break;
	} while (!nr_reclaimed);
	if (next_mz)
		css_put(&next_mz->mem->css);
	return nr_reclaimed;
}

3702 3703 3704 3705
/*
 * This routine traverse page_cgroup in given list and drop them all.
 * *And* this routine doesn't reclaim page itself, just removes page_cgroup.
 */
3706
static int mem_cgroup_force_empty_list(struct mem_cgroup *memcg,
K
KAMEZAWA Hiroyuki 已提交
3707
				int node, int zid, enum lru_list lru)
3708
{
K
KAMEZAWA Hiroyuki 已提交
3709 3710
	struct zone *zone;
	struct mem_cgroup_per_zone *mz;
3711
	struct page_cgroup *pc, *busy;
K
KAMEZAWA Hiroyuki 已提交
3712
	unsigned long flags, loop;
3713
	struct list_head *list;
3714
	int ret = 0;
3715

K
KAMEZAWA Hiroyuki 已提交
3716
	zone = &NODE_DATA(node)->node_zones[zid];
3717
	mz = mem_cgroup_zoneinfo(memcg, node, zid);
3718
	list = &mz->lists[lru];
3719

3720 3721 3722 3723 3724
	loop = MEM_CGROUP_ZSTAT(mz, lru);
	/* give some margin against EBUSY etc...*/
	loop += 256;
	busy = NULL;
	while (loop--) {
3725 3726
		struct page *page;

3727
		ret = 0;
K
KAMEZAWA Hiroyuki 已提交
3728
		spin_lock_irqsave(&zone->lru_lock, flags);
3729
		if (list_empty(list)) {
K
KAMEZAWA Hiroyuki 已提交
3730
			spin_unlock_irqrestore(&zone->lru_lock, flags);
3731
			break;
3732 3733 3734 3735
		}
		pc = list_entry(list->prev, struct page_cgroup, lru);
		if (busy == pc) {
			list_move(&pc->lru, list);
3736
			busy = NULL;
K
KAMEZAWA Hiroyuki 已提交
3737
			spin_unlock_irqrestore(&zone->lru_lock, flags);
3738 3739
			continue;
		}
K
KAMEZAWA Hiroyuki 已提交
3740
		spin_unlock_irqrestore(&zone->lru_lock, flags);
3741

3742
		page = lookup_cgroup_page(pc);
3743

3744
		ret = mem_cgroup_move_parent(page, pc, memcg, GFP_KERNEL);
3745
		if (ret == -ENOMEM)
3746
			break;
3747 3748 3749 3750 3751 3752 3753

		if (ret == -EBUSY || ret == -EINVAL) {
			/* found lock contention or "pc" is obsolete. */
			busy = pc;
			cond_resched();
		} else
			busy = NULL;
3754
	}
K
KAMEZAWA Hiroyuki 已提交
3755

3756 3757 3758
	if (!ret && !list_empty(list))
		return -EBUSY;
	return ret;
3759 3760 3761 3762 3763 3764
}

/*
 * make mem_cgroup's charge to be 0 if there is no task.
 * This enables deleting this mem_cgroup.
 */
3765
static int mem_cgroup_force_empty(struct mem_cgroup *memcg, bool free_all)
3766
{
3767 3768 3769
	int ret;
	int node, zid, shrink;
	int nr_retries = MEM_CGROUP_RECLAIM_RETRIES;
3770
	struct cgroup *cgrp = memcg->css.cgroup;
3771

3772
	css_get(&memcg->css);
3773 3774

	shrink = 0;
3775 3776 3777
	/* should free all ? */
	if (free_all)
		goto try_to_free;
3778
move_account:
3779
	do {
3780
		ret = -EBUSY;
3781 3782 3783 3784
		if (cgroup_task_count(cgrp) || !list_empty(&cgrp->children))
			goto out;
		ret = -EINTR;
		if (signal_pending(current))
3785
			goto out;
3786 3787
		/* This is for making all *used* pages to be on LRU. */
		lru_add_drain_all();
3788
		drain_all_stock_sync(memcg);
3789
		ret = 0;
3790
		mem_cgroup_start_move(memcg);
3791
		for_each_node_state(node, N_HIGH_MEMORY) {
3792
			for (zid = 0; !ret && zid < MAX_NR_ZONES; zid++) {
3793
				enum lru_list l;
3794
				for_each_lru(l) {
3795
					ret = mem_cgroup_force_empty_list(memcg,
K
KAMEZAWA Hiroyuki 已提交
3796
							node, zid, l);
3797 3798 3799
					if (ret)
						break;
				}
3800
			}
3801 3802 3803
			if (ret)
				break;
		}
3804 3805
		mem_cgroup_end_move(memcg);
		memcg_oom_recover(memcg);
3806 3807 3808
		/* it seems parent cgroup doesn't have enough mem */
		if (ret == -ENOMEM)
			goto try_to_free;
3809
		cond_resched();
3810
	/* "ret" should also be checked to ensure all lists are empty. */
3811
	} while (memcg->res.usage > 0 || ret);
3812
out:
3813
	css_put(&memcg->css);
3814
	return ret;
3815 3816

try_to_free:
3817 3818
	/* returns EBUSY if there is a task or if we come here twice. */
	if (cgroup_task_count(cgrp) || !list_empty(&cgrp->children) || shrink) {
3819 3820 3821
		ret = -EBUSY;
		goto out;
	}
3822 3823
	/* we call try-to-free pages for make this cgroup empty */
	lru_add_drain_all();
3824 3825
	/* try to free all pages in this cgroup */
	shrink = 1;
3826
	while (nr_retries && memcg->res.usage > 0) {
3827
		int progress;
3828 3829 3830 3831 3832

		if (signal_pending(current)) {
			ret = -EINTR;
			goto out;
		}
3833
		progress = try_to_free_mem_cgroup_pages(memcg, GFP_KERNEL,
3834
						false);
3835
		if (!progress) {
3836
			nr_retries--;
3837
			/* maybe some writeback is necessary */
3838
			congestion_wait(BLK_RW_ASYNC, HZ/10);
3839
		}
3840 3841

	}
K
KAMEZAWA Hiroyuki 已提交
3842
	lru_add_drain();
3843
	/* try move_account...there may be some *locked* pages. */
3844
	goto move_account;
3845 3846
}

3847 3848 3849 3850 3851 3852
int mem_cgroup_force_empty_write(struct cgroup *cont, unsigned int event)
{
	return mem_cgroup_force_empty(mem_cgroup_from_cont(cont), true);
}


3853 3854 3855 3856 3857 3858 3859 3860 3861
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;
3862
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
3863
	struct cgroup *parent = cont->parent;
3864
	struct mem_cgroup *parent_memcg = NULL;
3865 3866

	if (parent)
3867
		parent_memcg = mem_cgroup_from_cont(parent);
3868 3869 3870

	cgroup_lock();
	/*
3871
	 * If parent's use_hierarchy is set, we can't make any modifications
3872 3873 3874 3875 3876 3877
	 * 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.
	 */
3878
	if ((!parent_memcg || !parent_memcg->use_hierarchy) &&
3879 3880
				(val == 1 || val == 0)) {
		if (list_empty(&cont->children))
3881
			memcg->use_hierarchy = val;
3882 3883 3884 3885 3886 3887 3888 3889 3890
		else
			retval = -EBUSY;
	} else
		retval = -EINVAL;
	cgroup_unlock();

	return retval;
}

3891

3892
static unsigned long mem_cgroup_recursive_stat(struct mem_cgroup *memcg,
3893
					       enum mem_cgroup_stat_index idx)
3894
{
K
KAMEZAWA Hiroyuki 已提交
3895
	struct mem_cgroup *iter;
3896
	long val = 0;
3897

3898
	/* Per-cpu values can be negative, use a signed accumulator */
3899
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
3900 3901 3902 3903 3904
		val += mem_cgroup_read_stat(iter, idx);

	if (val < 0) /* race ? */
		val = 0;
	return val;
3905 3906
}

3907
static inline u64 mem_cgroup_usage(struct mem_cgroup *memcg, bool swap)
3908
{
K
KAMEZAWA Hiroyuki 已提交
3909
	u64 val;
3910

3911
	if (!mem_cgroup_is_root(memcg)) {
3912
		if (!swap)
3913
			return res_counter_read_u64(&memcg->res, RES_USAGE);
3914
		else
3915
			return res_counter_read_u64(&memcg->memsw, RES_USAGE);
3916 3917
	}

3918 3919
	val = mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_CACHE);
	val += mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_RSS);
3920

K
KAMEZAWA Hiroyuki 已提交
3921
	if (swap)
3922
		val += mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_SWAPOUT);
3923 3924 3925 3926

	return val << PAGE_SHIFT;
}

3927
static u64 mem_cgroup_read(struct cgroup *cont, struct cftype *cft)
B
Balbir Singh 已提交
3928
{
3929
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
3930
	u64 val;
3931 3932 3933 3934 3935 3936
	int type, name;

	type = MEMFILE_TYPE(cft->private);
	name = MEMFILE_ATTR(cft->private);
	switch (type) {
	case _MEM:
3937
		if (name == RES_USAGE)
3938
			val = mem_cgroup_usage(memcg, false);
3939
		else
3940
			val = res_counter_read_u64(&memcg->res, name);
3941 3942
		break;
	case _MEMSWAP:
3943
		if (name == RES_USAGE)
3944
			val = mem_cgroup_usage(memcg, true);
3945
		else
3946
			val = res_counter_read_u64(&memcg->memsw, name);
3947 3948 3949 3950 3951 3952
		break;
	default:
		BUG();
		break;
	}
	return val;
B
Balbir Singh 已提交
3953
}
3954 3955 3956 3957
/*
 * The user of this function is...
 * RES_LIMIT.
 */
3958 3959
static int mem_cgroup_write(struct cgroup *cont, struct cftype *cft,
			    const char *buffer)
B
Balbir Singh 已提交
3960
{
3961
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
3962
	int type, name;
3963 3964 3965
	unsigned long long val;
	int ret;

3966 3967 3968
	type = MEMFILE_TYPE(cft->private);
	name = MEMFILE_ATTR(cft->private);
	switch (name) {
3969
	case RES_LIMIT:
3970 3971 3972 3973
		if (mem_cgroup_is_root(memcg)) { /* Can't set limit on root */
			ret = -EINVAL;
			break;
		}
3974 3975
		/* This function does all necessary parse...reuse it */
		ret = res_counter_memparse_write_strategy(buffer, &val);
3976 3977 3978
		if (ret)
			break;
		if (type == _MEM)
3979
			ret = mem_cgroup_resize_limit(memcg, val);
3980 3981
		else
			ret = mem_cgroup_resize_memsw_limit(memcg, val);
3982
		break;
3983 3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996
	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;
3997 3998 3999 4000 4001
	default:
		ret = -EINVAL; /* should be BUG() ? */
		break;
	}
	return ret;
B
Balbir Singh 已提交
4002 4003
}

4004 4005 4006 4007 4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031
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;
	return;
}

4032
static int mem_cgroup_reset(struct cgroup *cont, unsigned int event)
4033
{
4034
	struct mem_cgroup *memcg;
4035
	int type, name;
4036

4037
	memcg = mem_cgroup_from_cont(cont);
4038 4039 4040
	type = MEMFILE_TYPE(event);
	name = MEMFILE_ATTR(event);
	switch (name) {
4041
	case RES_MAX_USAGE:
4042
		if (type == _MEM)
4043
			res_counter_reset_max(&memcg->res);
4044
		else
4045
			res_counter_reset_max(&memcg->memsw);
4046 4047
		break;
	case RES_FAILCNT:
4048
		if (type == _MEM)
4049
			res_counter_reset_failcnt(&memcg->res);
4050
		else
4051
			res_counter_reset_failcnt(&memcg->memsw);
4052 4053
		break;
	}
4054

4055
	return 0;
4056 4057
}

4058 4059 4060 4061 4062 4063
static u64 mem_cgroup_move_charge_read(struct cgroup *cgrp,
					struct cftype *cft)
{
	return mem_cgroup_from_cont(cgrp)->move_charge_at_immigrate;
}

4064
#ifdef CONFIG_MMU
4065 4066 4067
static int mem_cgroup_move_charge_write(struct cgroup *cgrp,
					struct cftype *cft, u64 val)
{
4068
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4069 4070 4071 4072 4073 4074 4075 4076 4077

	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();
4078
	memcg->move_charge_at_immigrate = val;
4079 4080 4081 4082
	cgroup_unlock();

	return 0;
}
4083 4084 4085 4086 4087 4088 4089
#else
static int mem_cgroup_move_charge_write(struct cgroup *cgrp,
					struct cftype *cft, u64 val)
{
	return -ENOSYS;
}
#endif
4090

K
KAMEZAWA Hiroyuki 已提交
4091 4092 4093 4094 4095

/* For read statistics */
enum {
	MCS_CACHE,
	MCS_RSS,
4096
	MCS_FILE_MAPPED,
K
KAMEZAWA Hiroyuki 已提交
4097 4098
	MCS_PGPGIN,
	MCS_PGPGOUT,
4099
	MCS_SWAP,
4100 4101
	MCS_PGFAULT,
	MCS_PGMAJFAULT,
K
KAMEZAWA Hiroyuki 已提交
4102 4103 4104 4105 4106 4107 4108 4109 4110 4111
	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];
4112 4113
};

K
KAMEZAWA Hiroyuki 已提交
4114 4115 4116 4117 4118 4119
struct {
	char *local_name;
	char *total_name;
} memcg_stat_strings[NR_MCS_STAT] = {
	{"cache", "total_cache"},
	{"rss", "total_rss"},
4120
	{"mapped_file", "total_mapped_file"},
K
KAMEZAWA Hiroyuki 已提交
4121 4122
	{"pgpgin", "total_pgpgin"},
	{"pgpgout", "total_pgpgout"},
4123
	{"swap", "total_swap"},
4124 4125
	{"pgfault", "total_pgfault"},
	{"pgmajfault", "total_pgmajfault"},
K
KAMEZAWA Hiroyuki 已提交
4126 4127 4128 4129 4130 4131 4132 4133
	{"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
4134
static void
4135
mem_cgroup_get_local_stat(struct mem_cgroup *memcg, struct mcs_total_stat *s)
K
KAMEZAWA Hiroyuki 已提交
4136 4137 4138 4139
{
	s64 val;

	/* per cpu stat */
4140
	val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_CACHE);
K
KAMEZAWA Hiroyuki 已提交
4141
	s->stat[MCS_CACHE] += val * PAGE_SIZE;
4142
	val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_RSS);
K
KAMEZAWA Hiroyuki 已提交
4143
	s->stat[MCS_RSS] += val * PAGE_SIZE;
4144
	val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_FILE_MAPPED);
4145
	s->stat[MCS_FILE_MAPPED] += val * PAGE_SIZE;
4146
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGPGIN);
K
KAMEZAWA Hiroyuki 已提交
4147
	s->stat[MCS_PGPGIN] += val;
4148
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGPGOUT);
K
KAMEZAWA Hiroyuki 已提交
4149
	s->stat[MCS_PGPGOUT] += val;
4150
	if (do_swap_account) {
4151
		val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_SWAPOUT);
4152 4153
		s->stat[MCS_SWAP] += val * PAGE_SIZE;
	}
4154
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGFAULT);
4155
	s->stat[MCS_PGFAULT] += val;
4156
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGMAJFAULT);
4157
	s->stat[MCS_PGMAJFAULT] += val;
K
KAMEZAWA Hiroyuki 已提交
4158 4159

	/* per zone stat */
4160
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_INACTIVE_ANON));
K
KAMEZAWA Hiroyuki 已提交
4161
	s->stat[MCS_INACTIVE_ANON] += val * PAGE_SIZE;
4162
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_ACTIVE_ANON));
K
KAMEZAWA Hiroyuki 已提交
4163
	s->stat[MCS_ACTIVE_ANON] += val * PAGE_SIZE;
4164
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_INACTIVE_FILE));
K
KAMEZAWA Hiroyuki 已提交
4165
	s->stat[MCS_INACTIVE_FILE] += val * PAGE_SIZE;
4166
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_ACTIVE_FILE));
K
KAMEZAWA Hiroyuki 已提交
4167
	s->stat[MCS_ACTIVE_FILE] += val * PAGE_SIZE;
4168
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_UNEVICTABLE));
K
KAMEZAWA Hiroyuki 已提交
4169 4170 4171 4172
	s->stat[MCS_UNEVICTABLE] += val * PAGE_SIZE;
}

static void
4173
mem_cgroup_get_total_stat(struct mem_cgroup *memcg, struct mcs_total_stat *s)
K
KAMEZAWA Hiroyuki 已提交
4174
{
K
KAMEZAWA Hiroyuki 已提交
4175 4176
	struct mem_cgroup *iter;

4177
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
4178
		mem_cgroup_get_local_stat(iter, s);
K
KAMEZAWA Hiroyuki 已提交
4179 4180
}

4181 4182 4183 4184 4185 4186 4187 4188 4189
#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;
	struct mem_cgroup *mem_cont = mem_cgroup_from_cont(cont);

4190
	total_nr = mem_cgroup_nr_lru_pages(mem_cont, LRU_ALL);
4191 4192
	seq_printf(m, "total=%lu", total_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4193
		node_nr = mem_cgroup_node_nr_lru_pages(mem_cont, nid, LRU_ALL);
4194 4195 4196 4197
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');

4198
	file_nr = mem_cgroup_nr_lru_pages(mem_cont, LRU_ALL_FILE);
4199 4200
	seq_printf(m, "file=%lu", file_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4201 4202
		node_nr = mem_cgroup_node_nr_lru_pages(mem_cont, nid,
				LRU_ALL_FILE);
4203 4204 4205 4206
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');

4207
	anon_nr = mem_cgroup_nr_lru_pages(mem_cont, LRU_ALL_ANON);
4208 4209
	seq_printf(m, "anon=%lu", anon_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4210 4211
		node_nr = mem_cgroup_node_nr_lru_pages(mem_cont, nid,
				LRU_ALL_ANON);
4212 4213 4214 4215
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');

4216
	unevictable_nr = mem_cgroup_nr_lru_pages(mem_cont, BIT(LRU_UNEVICTABLE));
4217 4218
	seq_printf(m, "unevictable=%lu", unevictable_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4219 4220
		node_nr = mem_cgroup_node_nr_lru_pages(mem_cont, nid,
				BIT(LRU_UNEVICTABLE));
4221 4222 4223 4224 4225 4226 4227
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');
	return 0;
}
#endif /* CONFIG_NUMA */

4228 4229
static int mem_control_stat_show(struct cgroup *cont, struct cftype *cft,
				 struct cgroup_map_cb *cb)
4230 4231
{
	struct mem_cgroup *mem_cont = mem_cgroup_from_cont(cont);
K
KAMEZAWA Hiroyuki 已提交
4232
	struct mcs_total_stat mystat;
4233 4234
	int i;

K
KAMEZAWA Hiroyuki 已提交
4235 4236
	memset(&mystat, 0, sizeof(mystat));
	mem_cgroup_get_local_stat(mem_cont, &mystat);
4237

4238

4239 4240 4241
	for (i = 0; i < NR_MCS_STAT; i++) {
		if (i == MCS_SWAP && !do_swap_account)
			continue;
K
KAMEZAWA Hiroyuki 已提交
4242
		cb->fill(cb, memcg_stat_strings[i].local_name, mystat.stat[i]);
4243
	}
L
Lee Schermerhorn 已提交
4244

K
KAMEZAWA Hiroyuki 已提交
4245
	/* Hierarchical information */
4246 4247 4248 4249 4250 4251 4252
	{
		unsigned long long limit, memsw_limit;
		memcg_get_hierarchical_limit(mem_cont, &limit, &memsw_limit);
		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
4253

K
KAMEZAWA Hiroyuki 已提交
4254 4255
	memset(&mystat, 0, sizeof(mystat));
	mem_cgroup_get_total_stat(mem_cont, &mystat);
4256 4257 4258
	for (i = 0; i < NR_MCS_STAT; i++) {
		if (i == MCS_SWAP && !do_swap_account)
			continue;
K
KAMEZAWA Hiroyuki 已提交
4259
		cb->fill(cb, memcg_stat_strings[i].total_name, mystat.stat[i]);
4260
	}
K
KAMEZAWA Hiroyuki 已提交
4261

K
KOSAKI Motohiro 已提交
4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288
#ifdef CONFIG_DEBUG_VM
	{
		int nid, zid;
		struct mem_cgroup_per_zone *mz;
		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++) {
				mz = mem_cgroup_zoneinfo(mem_cont, nid, zid);

				recent_rotated[0] +=
					mz->reclaim_stat.recent_rotated[0];
				recent_rotated[1] +=
					mz->reclaim_stat.recent_rotated[1];
				recent_scanned[0] +=
					mz->reclaim_stat.recent_scanned[0];
				recent_scanned[1] +=
					mz->reclaim_stat.recent_scanned[1];
			}
		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

4289 4290 4291
	return 0;
}

K
KOSAKI Motohiro 已提交
4292 4293 4294 4295
static u64 mem_cgroup_swappiness_read(struct cgroup *cgrp, struct cftype *cft)
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);

4296
	return mem_cgroup_swappiness(memcg);
K
KOSAKI Motohiro 已提交
4297 4298 4299 4300 4301 4302 4303
}

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;
4304

K
KOSAKI Motohiro 已提交
4305 4306 4307 4308 4309 4310 4311
	if (val > 100)
		return -EINVAL;

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

	parent = mem_cgroup_from_cont(cgrp->parent);
4312 4313 4314

	cgroup_lock();

K
KOSAKI Motohiro 已提交
4315 4316
	/* If under hierarchy, only empty-root can set this value */
	if ((parent->use_hierarchy) ||
4317 4318
	    (memcg->use_hierarchy && !list_empty(&cgrp->children))) {
		cgroup_unlock();
K
KOSAKI Motohiro 已提交
4319
		return -EINVAL;
4320
	}
K
KOSAKI Motohiro 已提交
4321 4322 4323

	memcg->swappiness = val;

4324 4325
	cgroup_unlock();

K
KOSAKI Motohiro 已提交
4326 4327 4328
	return 0;
}

4329 4330 4331 4332 4333 4334 4335 4336
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)
4337
		t = rcu_dereference(memcg->thresholds.primary);
4338
	else
4339
		t = rcu_dereference(memcg->memsw_thresholds.primary);
4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350

	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().
	 */
4351
	i = t->current_threshold;
4352 4353 4354 4355 4356 4357 4358 4359 4360 4361 4362 4363 4364 4365 4366 4367 4368 4369 4370 4371 4372 4373 4374

	/*
	 * 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 */
4375
	t->current_threshold = i - 1;
4376 4377 4378 4379 4380 4381
unlock:
	rcu_read_unlock();
}

static void mem_cgroup_threshold(struct mem_cgroup *memcg)
{
4382 4383 4384 4385 4386 4387 4388
	while (memcg) {
		__mem_cgroup_threshold(memcg, false);
		if (do_swap_account)
			__mem_cgroup_threshold(memcg, true);

		memcg = parent_mem_cgroup(memcg);
	}
4389 4390 4391 4392 4393 4394 4395 4396 4397 4398
}

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;
}

4399
static int mem_cgroup_oom_notify_cb(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
4400 4401 4402
{
	struct mem_cgroup_eventfd_list *ev;

4403
	list_for_each_entry(ev, &memcg->oom_notify, list)
K
KAMEZAWA Hiroyuki 已提交
4404 4405 4406 4407
		eventfd_signal(ev->eventfd, 1);
	return 0;
}

4408
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
4409
{
K
KAMEZAWA Hiroyuki 已提交
4410 4411
	struct mem_cgroup *iter;

4412
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
4413
		mem_cgroup_oom_notify_cb(iter);
K
KAMEZAWA Hiroyuki 已提交
4414 4415 4416 4417
}

static int mem_cgroup_usage_register_event(struct cgroup *cgrp,
	struct cftype *cft, struct eventfd_ctx *eventfd, const char *args)
4418 4419
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4420 4421
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
4422 4423
	int type = MEMFILE_TYPE(cft->private);
	u64 threshold, usage;
4424
	int i, size, ret;
4425 4426 4427 4428 4429 4430

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

	mutex_lock(&memcg->thresholds_lock);
4431

4432
	if (type == _MEM)
4433
		thresholds = &memcg->thresholds;
4434
	else if (type == _MEMSWAP)
4435
		thresholds = &memcg->memsw_thresholds;
4436 4437 4438 4439 4440 4441
	else
		BUG();

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

	/* Check if a threshold crossed before adding a new one */
4442
	if (thresholds->primary)
4443 4444
		__mem_cgroup_threshold(memcg, type == _MEMSWAP);

4445
	size = thresholds->primary ? thresholds->primary->size + 1 : 1;
4446 4447

	/* Allocate memory for new array of thresholds */
4448
	new = kmalloc(sizeof(*new) + size * sizeof(struct mem_cgroup_threshold),
4449
			GFP_KERNEL);
4450
	if (!new) {
4451 4452 4453
		ret = -ENOMEM;
		goto unlock;
	}
4454
	new->size = size;
4455 4456

	/* Copy thresholds (if any) to new array */
4457 4458
	if (thresholds->primary) {
		memcpy(new->entries, thresholds->primary->entries, (size - 1) *
4459
				sizeof(struct mem_cgroup_threshold));
4460 4461
	}

4462
	/* Add new threshold */
4463 4464
	new->entries[size - 1].eventfd = eventfd;
	new->entries[size - 1].threshold = threshold;
4465 4466

	/* Sort thresholds. Registering of new threshold isn't time-critical */
4467
	sort(new->entries, size, sizeof(struct mem_cgroup_threshold),
4468 4469 4470
			compare_thresholds, NULL);

	/* Find current threshold */
4471
	new->current_threshold = -1;
4472
	for (i = 0; i < size; i++) {
4473
		if (new->entries[i].threshold < usage) {
4474
			/*
4475 4476
			 * new->current_threshold will not be used until
			 * rcu_assign_pointer(), so it's safe to increment
4477 4478
			 * it here.
			 */
4479
			++new->current_threshold;
4480 4481 4482
		}
	}

4483 4484 4485 4486 4487
	/* Free old spare buffer and save old primary buffer as spare */
	kfree(thresholds->spare);
	thresholds->spare = thresholds->primary;

	rcu_assign_pointer(thresholds->primary, new);
4488

4489
	/* To be sure that nobody uses thresholds */
4490 4491 4492 4493 4494 4495 4496 4497
	synchronize_rcu();

unlock:
	mutex_unlock(&memcg->thresholds_lock);

	return ret;
}

4498
static void mem_cgroup_usage_unregister_event(struct cgroup *cgrp,
K
KAMEZAWA Hiroyuki 已提交
4499
	struct cftype *cft, struct eventfd_ctx *eventfd)
4500 4501
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4502 4503
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
4504 4505
	int type = MEMFILE_TYPE(cft->private);
	u64 usage;
4506
	int i, j, size;
4507 4508 4509

	mutex_lock(&memcg->thresholds_lock);
	if (type == _MEM)
4510
		thresholds = &memcg->thresholds;
4511
	else if (type == _MEMSWAP)
4512
		thresholds = &memcg->memsw_thresholds;
4513 4514 4515 4516 4517 4518 4519 4520 4521 4522 4523 4524 4525 4526 4527
	else
		BUG();

	/*
	 * Something went wrong if we trying to unregister a threshold
	 * if we don't have thresholds
	 */
	BUG_ON(!thresholds);

	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 */
4528 4529 4530
	size = 0;
	for (i = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd != eventfd)
4531 4532 4533
			size++;
	}

4534
	new = thresholds->spare;
4535

4536 4537
	/* Set thresholds array to NULL if we don't have thresholds */
	if (!size) {
4538 4539
		kfree(new);
		new = NULL;
4540
		goto swap_buffers;
4541 4542
	}

4543
	new->size = size;
4544 4545

	/* Copy thresholds and find current threshold */
4546 4547 4548
	new->current_threshold = -1;
	for (i = 0, j = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd == eventfd)
4549 4550
			continue;

4551 4552
		new->entries[j] = thresholds->primary->entries[i];
		if (new->entries[j].threshold < usage) {
4553
			/*
4554
			 * new->current_threshold will not be used
4555 4556 4557
			 * until rcu_assign_pointer(), so it's safe to increment
			 * it here.
			 */
4558
			++new->current_threshold;
4559 4560 4561 4562
		}
		j++;
	}

4563
swap_buffers:
4564 4565 4566
	/* Swap primary and spare array */
	thresholds->spare = thresholds->primary;
	rcu_assign_pointer(thresholds->primary, new);
4567

4568
	/* To be sure that nobody uses thresholds */
4569 4570 4571 4572
	synchronize_rcu();

	mutex_unlock(&memcg->thresholds_lock);
}
4573

K
KAMEZAWA Hiroyuki 已提交
4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584 4585
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;

4586
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4587 4588 4589 4590 4591

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

	/* already in OOM ? */
4592
	if (atomic_read(&memcg->under_oom))
K
KAMEZAWA Hiroyuki 已提交
4593
		eventfd_signal(eventfd, 1);
4594
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4595 4596 4597 4598

	return 0;
}

4599
static void mem_cgroup_oom_unregister_event(struct cgroup *cgrp,
K
KAMEZAWA Hiroyuki 已提交
4600 4601
	struct cftype *cft, struct eventfd_ctx *eventfd)
{
4602
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
K
KAMEZAWA Hiroyuki 已提交
4603 4604 4605 4606 4607
	struct mem_cgroup_eventfd_list *ev, *tmp;
	int type = MEMFILE_TYPE(cft->private);

	BUG_ON(type != _OOM_TYPE);

4608
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4609

4610
	list_for_each_entry_safe(ev, tmp, &memcg->oom_notify, list) {
K
KAMEZAWA Hiroyuki 已提交
4611 4612 4613 4614 4615 4616
		if (ev->eventfd == eventfd) {
			list_del(&ev->list);
			kfree(ev);
		}
	}

4617
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4618 4619
}

4620 4621 4622
static int mem_cgroup_oom_control_read(struct cgroup *cgrp,
	struct cftype *cft,  struct cgroup_map_cb *cb)
{
4623
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4624

4625
	cb->fill(cb, "oom_kill_disable", memcg->oom_kill_disable);
4626

4627
	if (atomic_read(&memcg->under_oom))
4628 4629 4630 4631 4632 4633 4634 4635 4636
		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)
{
4637
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648
	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) ||
4649
	    (memcg->use_hierarchy && !list_empty(&cgrp->children))) {
4650 4651 4652
		cgroup_unlock();
		return -EINVAL;
	}
4653
	memcg->oom_kill_disable = val;
4654
	if (!val)
4655
		memcg_oom_recover(memcg);
4656 4657 4658 4659
	cgroup_unlock();
	return 0;
}

4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672 4673 4674 4675
#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 */

4676 4677 4678
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
static int register_kmem_files(struct cgroup *cont, struct cgroup_subsys *ss)
{
G
Glauber Costa 已提交
4679 4680 4681 4682 4683 4684 4685
	/*
	 * Part of this would be better living in a separate allocation
	 * function, leaving us with just the cgroup tree population work.
	 * We, however, depend on state such as network's proto_list that
	 * is only initialized after cgroup creation. I found the less
	 * cumbersome way to deal with it to defer it all to populate time
	 */
4686
	return mem_cgroup_sockets_init(cont, ss);
4687 4688
};

G
Glauber Costa 已提交
4689 4690 4691 4692 4693
static void kmem_cgroup_destroy(struct cgroup_subsys *ss,
				struct cgroup *cont)
{
	mem_cgroup_sockets_destroy(cont, ss);
}
4694 4695 4696 4697 4698
#else
static int register_kmem_files(struct cgroup *cont, struct cgroup_subsys *ss)
{
	return 0;
}
G
Glauber Costa 已提交
4699 4700 4701 4702 4703

static void kmem_cgroup_destroy(struct cgroup_subsys *ss,
				struct cgroup *cont)
{
}
4704 4705
#endif

B
Balbir Singh 已提交
4706 4707
static struct cftype mem_cgroup_files[] = {
	{
4708
		.name = "usage_in_bytes",
4709
		.private = MEMFILE_PRIVATE(_MEM, RES_USAGE),
4710
		.read_u64 = 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,
B
Balbir Singh 已提交
4713
	},
4714 4715
	{
		.name = "max_usage_in_bytes",
4716
		.private = MEMFILE_PRIVATE(_MEM, RES_MAX_USAGE),
4717
		.trigger = mem_cgroup_reset,
4718 4719
		.read_u64 = mem_cgroup_read,
	},
B
Balbir Singh 已提交
4720
	{
4721
		.name = "limit_in_bytes",
4722
		.private = MEMFILE_PRIVATE(_MEM, RES_LIMIT),
4723
		.write_string = mem_cgroup_write,
4724
		.read_u64 = mem_cgroup_read,
B
Balbir Singh 已提交
4725
	},
4726 4727 4728 4729 4730 4731
	{
		.name = "soft_limit_in_bytes",
		.private = MEMFILE_PRIVATE(_MEM, RES_SOFT_LIMIT),
		.write_string = mem_cgroup_write,
		.read_u64 = mem_cgroup_read,
	},
B
Balbir Singh 已提交
4732 4733
	{
		.name = "failcnt",
4734
		.private = MEMFILE_PRIVATE(_MEM, RES_FAILCNT),
4735
		.trigger = mem_cgroup_reset,
4736
		.read_u64 = mem_cgroup_read,
B
Balbir Singh 已提交
4737
	},
4738 4739
	{
		.name = "stat",
4740
		.read_map = mem_control_stat_show,
4741
	},
4742 4743 4744 4745
	{
		.name = "force_empty",
		.trigger = mem_cgroup_force_empty_write,
	},
4746 4747 4748 4749 4750
	{
		.name = "use_hierarchy",
		.write_u64 = mem_cgroup_hierarchy_write,
		.read_u64 = mem_cgroup_hierarchy_read,
	},
K
KOSAKI Motohiro 已提交
4751 4752 4753 4754 4755
	{
		.name = "swappiness",
		.read_u64 = mem_cgroup_swappiness_read,
		.write_u64 = mem_cgroup_swappiness_write,
	},
4756 4757 4758 4759 4760
	{
		.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
4761 4762
	{
		.name = "oom_control",
4763 4764
		.read_map = mem_cgroup_oom_control_read,
		.write_u64 = mem_cgroup_oom_control_write,
K
KAMEZAWA Hiroyuki 已提交
4765 4766 4767 4768
		.register_event = mem_cgroup_oom_register_event,
		.unregister_event = mem_cgroup_oom_unregister_event,
		.private = MEMFILE_PRIVATE(_OOM_TYPE, OOM_CONTROL),
	},
4769 4770 4771 4772
#ifdef CONFIG_NUMA
	{
		.name = "numa_stat",
		.open = mem_control_numa_stat_open,
4773
		.mode = S_IRUGO,
4774 4775
	},
#endif
B
Balbir Singh 已提交
4776 4777
};

4778 4779 4780 4781 4782 4783
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
static struct cftype memsw_cgroup_files[] = {
	{
		.name = "memsw.usage_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_USAGE),
		.read_u64 = mem_cgroup_read,
K
KAMEZAWA Hiroyuki 已提交
4784 4785
		.register_event = mem_cgroup_usage_register_event,
		.unregister_event = mem_cgroup_usage_unregister_event,
4786 4787 4788 4789 4790 4791 4792 4793 4794 4795 4796 4797 4798 4799 4800 4801 4802 4803 4804 4805 4806 4807 4808 4809 4810 4811 4812 4813 4814 4815 4816 4817 4818 4819 4820
	},
	{
		.name = "memsw.max_usage_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_MAX_USAGE),
		.trigger = mem_cgroup_reset,
		.read_u64 = mem_cgroup_read,
	},
	{
		.name = "memsw.limit_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_LIMIT),
		.write_string = mem_cgroup_write,
		.read_u64 = mem_cgroup_read,
	},
	{
		.name = "memsw.failcnt",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_FAILCNT),
		.trigger = mem_cgroup_reset,
		.read_u64 = mem_cgroup_read,
	},
};

static int register_memsw_files(struct cgroup *cont, struct cgroup_subsys *ss)
{
	if (!do_swap_account)
		return 0;
	return cgroup_add_files(cont, ss, memsw_cgroup_files,
				ARRAY_SIZE(memsw_cgroup_files));
};
#else
static int register_memsw_files(struct cgroup *cont, struct cgroup_subsys *ss)
{
	return 0;
}
#endif

4821
static int alloc_mem_cgroup_per_zone_info(struct mem_cgroup *memcg, int node)
4822 4823
{
	struct mem_cgroup_per_node *pn;
4824
	struct mem_cgroup_per_zone *mz;
4825
	enum lru_list l;
4826
	int zone, tmp = node;
4827 4828 4829 4830 4831 4832 4833 4834
	/*
	 * 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.
	 */
4835 4836
	if (!node_state(node, N_NORMAL_MEMORY))
		tmp = -1;
4837
	pn = kzalloc_node(sizeof(*pn), GFP_KERNEL, tmp);
4838 4839
	if (!pn)
		return 1;
4840 4841 4842

	for (zone = 0; zone < MAX_NR_ZONES; zone++) {
		mz = &pn->zoneinfo[zone];
4843 4844
		for_each_lru(l)
			INIT_LIST_HEAD(&mz->lists[l]);
4845
		mz->usage_in_excess = 0;
4846
		mz->on_tree = false;
4847
		mz->mem = memcg;
4848
	}
4849
	memcg->info.nodeinfo[node] = pn;
4850 4851 4852
	return 0;
}

4853
static void free_mem_cgroup_per_zone_info(struct mem_cgroup *memcg, int node)
4854
{
4855
	kfree(memcg->info.nodeinfo[node]);
4856 4857
}

4858 4859 4860
static struct mem_cgroup *mem_cgroup_alloc(void)
{
	struct mem_cgroup *mem;
4861
	int size = sizeof(struct mem_cgroup);
4862

4863
	/* Can be very big if MAX_NUMNODES is very big */
4864
	if (size < PAGE_SIZE)
4865
		mem = kzalloc(size, GFP_KERNEL);
4866
	else
4867
		mem = vzalloc(size);
4868

4869 4870 4871
	if (!mem)
		return NULL;

4872
	mem->stat = alloc_percpu(struct mem_cgroup_stat_cpu);
4873 4874
	if (!mem->stat)
		goto out_free;
4875
	spin_lock_init(&mem->pcp_counter_lock);
4876
	return mem;
4877 4878 4879 4880 4881 4882 4883

out_free:
	if (size < PAGE_SIZE)
		kfree(mem);
	else
		vfree(mem);
	return NULL;
4884 4885
}

4886 4887 4888 4889 4890 4891 4892 4893 4894 4895 4896
/*
 * 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.
 */

4897
static void __mem_cgroup_free(struct mem_cgroup *memcg)
4898
{
K
KAMEZAWA Hiroyuki 已提交
4899 4900
	int node;

4901 4902
	mem_cgroup_remove_from_trees(memcg);
	free_css_id(&mem_cgroup_subsys, &memcg->css);
K
KAMEZAWA Hiroyuki 已提交
4903

K
KAMEZAWA Hiroyuki 已提交
4904
	for_each_node_state(node, N_POSSIBLE)
4905
		free_mem_cgroup_per_zone_info(memcg, node);
K
KAMEZAWA Hiroyuki 已提交
4906

4907
	free_percpu(memcg->stat);
4908
	if (sizeof(struct mem_cgroup) < PAGE_SIZE)
4909
		kfree(memcg);
4910
	else
4911
		vfree(memcg);
4912 4913
}

4914
static void mem_cgroup_get(struct mem_cgroup *memcg)
4915
{
4916
	atomic_inc(&memcg->refcnt);
4917 4918
}

4919
static void __mem_cgroup_put(struct mem_cgroup *memcg, int count)
4920
{
4921 4922 4923
	if (atomic_sub_and_test(count, &memcg->refcnt)) {
		struct mem_cgroup *parent = parent_mem_cgroup(memcg);
		__mem_cgroup_free(memcg);
4924 4925 4926
		if (parent)
			mem_cgroup_put(parent);
	}
4927 4928
}

4929
static void mem_cgroup_put(struct mem_cgroup *memcg)
4930
{
4931
	__mem_cgroup_put(memcg, 1);
4932 4933
}

4934 4935 4936
/*
 * Returns the parent mem_cgroup in memcgroup hierarchy with hierarchy enabled.
 */
G
Glauber Costa 已提交
4937
struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
4938
{
4939
	if (!memcg->res.parent)
4940
		return NULL;
4941
	return mem_cgroup_from_res_counter(memcg->res.parent, res);
4942
}
G
Glauber Costa 已提交
4943
EXPORT_SYMBOL(parent_mem_cgroup);
4944

4945 4946 4947
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
static void __init enable_swap_cgroup(void)
{
4948
	if (!mem_cgroup_disabled() && really_do_swap_account)
4949 4950 4951 4952 4953 4954 4955 4956
		do_swap_account = 1;
}
#else
static void __init enable_swap_cgroup(void)
{
}
#endif

4957 4958 4959 4960 4961 4962 4963 4964 4965 4966 4967 4968 4969 4970 4971 4972 4973 4974 4975 4976 4977 4978 4979 4980 4981
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;

	for_each_node_state(node, N_POSSIBLE) {
		tmp = node;
		if (!node_state(node, N_NORMAL_MEMORY))
			tmp = -1;
		rtpn = kzalloc_node(sizeof(*rtpn), GFP_KERNEL, tmp);
		if (!rtpn)
			return 1;

		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;
}

L
Li Zefan 已提交
4982
static struct cgroup_subsys_state * __ref
B
Balbir Singh 已提交
4983 4984
mem_cgroup_create(struct cgroup_subsys *ss, struct cgroup *cont)
{
4985
	struct mem_cgroup *memcg, *parent;
K
KAMEZAWA Hiroyuki 已提交
4986
	long error = -ENOMEM;
4987
	int node;
B
Balbir Singh 已提交
4988

4989 4990
	memcg = mem_cgroup_alloc();
	if (!memcg)
K
KAMEZAWA Hiroyuki 已提交
4991
		return ERR_PTR(error);
4992

4993
	for_each_node_state(node, N_POSSIBLE)
4994
		if (alloc_mem_cgroup_per_zone_info(memcg, node))
4995
			goto free_out;
4996

4997
	/* root ? */
4998
	if (cont->parent == NULL) {
4999
		int cpu;
5000
		enable_swap_cgroup();
5001
		parent = NULL;
5002 5003
		if (mem_cgroup_soft_limit_tree_init())
			goto free_out;
5004
		root_mem_cgroup = memcg;
5005 5006 5007 5008 5009
		for_each_possible_cpu(cpu) {
			struct memcg_stock_pcp *stock =
						&per_cpu(memcg_stock, cpu);
			INIT_WORK(&stock->work, drain_local_stock);
		}
5010
		hotcpu_notifier(memcg_cpu_hotplug_callback, 0);
5011
	} else {
5012
		parent = mem_cgroup_from_cont(cont->parent);
5013 5014
		memcg->use_hierarchy = parent->use_hierarchy;
		memcg->oom_kill_disable = parent->oom_kill_disable;
5015
	}
5016

5017
	if (parent && parent->use_hierarchy) {
5018 5019
		res_counter_init(&memcg->res, &parent->res);
		res_counter_init(&memcg->memsw, &parent->memsw);
5020 5021 5022 5023 5024 5025 5026
		/*
		 * 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);
5027
	} else {
5028 5029
		res_counter_init(&memcg->res, NULL);
		res_counter_init(&memcg->memsw, NULL);
5030
	}
5031 5032 5033
	memcg->last_scanned_child = 0;
	memcg->last_scanned_node = MAX_NUMNODES;
	INIT_LIST_HEAD(&memcg->oom_notify);
5034

K
KOSAKI Motohiro 已提交
5035
	if (parent)
5036 5037 5038 5039 5040
		memcg->swappiness = mem_cgroup_swappiness(parent);
	atomic_set(&memcg->refcnt, 1);
	memcg->move_charge_at_immigrate = 0;
	mutex_init(&memcg->thresholds_lock);
	return &memcg->css;
5041
free_out:
5042
	__mem_cgroup_free(memcg);
K
KAMEZAWA Hiroyuki 已提交
5043
	return ERR_PTR(error);
B
Balbir Singh 已提交
5044 5045
}

5046
static int mem_cgroup_pre_destroy(struct cgroup_subsys *ss,
5047 5048
					struct cgroup *cont)
{
5049
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
5050

5051
	return mem_cgroup_force_empty(memcg, false);
5052 5053
}

B
Balbir Singh 已提交
5054 5055 5056
static void mem_cgroup_destroy(struct cgroup_subsys *ss,
				struct cgroup *cont)
{
5057
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
5058

G
Glauber Costa 已提交
5059 5060
	kmem_cgroup_destroy(ss, cont);

5061
	mem_cgroup_put(memcg);
B
Balbir Singh 已提交
5062 5063 5064 5065 5066
}

static int mem_cgroup_populate(struct cgroup_subsys *ss,
				struct cgroup *cont)
{
5067 5068 5069 5070 5071 5072 5073
	int ret;

	ret = cgroup_add_files(cont, ss, mem_cgroup_files,
				ARRAY_SIZE(mem_cgroup_files));

	if (!ret)
		ret = register_memsw_files(cont, ss);
5074 5075 5076 5077

	if (!ret)
		ret = register_kmem_files(cont, ss);

5078
	return ret;
B
Balbir Singh 已提交
5079 5080
}

5081
#ifdef CONFIG_MMU
5082
/* Handlers for move charge at task migration. */
5083 5084
#define PRECHARGE_COUNT_AT_ONCE	256
static int mem_cgroup_do_precharge(unsigned long count)
5085
{
5086 5087
	int ret = 0;
	int batch_count = PRECHARGE_COUNT_AT_ONCE;
5088
	struct mem_cgroup *memcg = mc.to;
5089

5090
	if (mem_cgroup_is_root(memcg)) {
5091 5092 5093 5094 5095 5096 5097 5098
		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;
		/*
5099
		 * "memcg" cannot be under rmdir() because we've already checked
5100 5101 5102 5103
		 * 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().
		 */
5104
		if (res_counter_charge(&memcg->res, PAGE_SIZE * count, &dummy))
5105
			goto one_by_one;
5106
		if (do_swap_account && res_counter_charge(&memcg->memsw,
5107
						PAGE_SIZE * count, &dummy)) {
5108
			res_counter_uncharge(&memcg->res, PAGE_SIZE * count);
5109 5110 5111 5112 5113 5114 5115 5116 5117 5118 5119 5120 5121 5122 5123 5124
			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();
		}
5125 5126 5127
		ret = __mem_cgroup_try_charge(NULL,
					GFP_KERNEL, 1, &memcg, false);
		if (ret || !memcg)
5128 5129 5130 5131
			/* mem_cgroup_clear_mc() will do uncharge later */
			return -ENOMEM;
		mc.precharge++;
	}
5132 5133 5134 5135 5136 5137 5138 5139
	return ret;
}

/**
 * is_target_pte_for_mc - check a pte whether it is valid for move charge
 * @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
5140
 * @target: the pointer the target page or swap ent will be stored(can be NULL)
5141 5142 5143 5144 5145 5146
 *
 * 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).
5147 5148 5149
 *   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.
5150 5151 5152 5153 5154
 *
 * Called with pte lock held.
 */
union mc_target {
	struct page	*page;
5155
	swp_entry_t	ent;
5156 5157 5158 5159 5160
};

enum mc_target_type {
	MC_TARGET_NONE,	/* not used */
	MC_TARGET_PAGE,
5161
	MC_TARGET_SWAP,
5162 5163
};

D
Daisuke Nishimura 已提交
5164 5165
static struct page *mc_handle_present_pte(struct vm_area_struct *vma,
						unsigned long addr, pte_t ptent)
5166
{
D
Daisuke Nishimura 已提交
5167
	struct page *page = vm_normal_page(vma, addr, ptent);
5168

D
Daisuke Nishimura 已提交
5169 5170 5171 5172 5173 5174
	if (!page || !page_mapped(page))
		return NULL;
	if (PageAnon(page)) {
		/* we don't move shared anon */
		if (!move_anon() || page_mapcount(page) > 2)
			return NULL;
5175 5176
	} else if (!move_file())
		/* we ignore mapcount for file pages */
D
Daisuke Nishimura 已提交
5177 5178 5179 5180 5181 5182 5183 5184 5185 5186 5187 5188 5189 5190 5191 5192 5193 5194
		return NULL;
	if (!get_page_unless_zero(page))
		return NULL;

	return page;
}

static struct page *mc_handle_swap_pte(struct vm_area_struct *vma,
			unsigned long addr, pte_t ptent, swp_entry_t *entry)
{
	int usage_count;
	struct page *page = NULL;
	swp_entry_t ent = pte_to_swp_entry(ptent);

	if (!move_anon() || non_swap_entry(ent))
		return NULL;
	usage_count = mem_cgroup_count_swap_user(ent, &page);
	if (usage_count > 1) { /* we don't move shared anon */
5195 5196
		if (page)
			put_page(page);
D
Daisuke Nishimura 已提交
5197
		return NULL;
5198
	}
D
Daisuke Nishimura 已提交
5199 5200 5201 5202 5203 5204
	if (do_swap_account)
		entry->val = ent.val;

	return page;
}

5205 5206 5207 5208 5209 5210 5211 5212 5213 5214 5215 5216 5217 5218 5219 5220 5221 5222 5223 5224 5225
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 inode *inode;
	struct address_space *mapping;
	pgoff_t pgoff;

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

	inode = vma->vm_file->f_path.dentry->d_inode;
	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). */
5226 5227 5228 5229 5230 5231
	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);
5232
		if (do_swap_account)
5233 5234
			*entry = swap;
		page = find_get_page(&swapper_space, swap.val);
5235
	}
5236
#endif
5237 5238 5239
	return page;
}

D
Daisuke Nishimura 已提交
5240 5241 5242 5243 5244 5245 5246 5247 5248 5249 5250 5251
static int is_target_pte_for_mc(struct vm_area_struct *vma,
		unsigned long addr, pte_t ptent, union mc_target *target)
{
	struct page *page = NULL;
	struct page_cgroup *pc;
	int ret = 0;
	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);
5252 5253
	else if (pte_none(ptent) || pte_file(ptent))
		page = mc_handle_file_pte(vma, addr, ptent, &ent);
D
Daisuke Nishimura 已提交
5254 5255 5256

	if (!page && !ent.val)
		return 0;
5257 5258 5259 5260 5261 5262 5263 5264 5265 5266 5267 5268 5269 5270 5271
	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
5272 5273
	/* There is a swap entry and a page doesn't exist or isn't charged */
	if (ent.val && !ret &&
5274 5275 5276 5277
			css_id(&mc.from->css) == lookup_swap_cgroup(ent)) {
		ret = MC_TARGET_SWAP;
		if (target)
			target->ent = ent;
5278 5279 5280 5281 5282 5283 5284 5285 5286 5287 5288 5289
	}
	return ret;
}

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;

5290 5291
	split_huge_page_pmd(walk->mm, pmd);

5292 5293 5294 5295 5296 5297 5298
	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
	for (; addr != end; pte++, addr += PAGE_SIZE)
		if (is_target_pte_for_mc(vma, addr, *pte, NULL))
			mc.precharge++;	/* increment precharge temporarily */
	pte_unmap_unlock(pte - 1, ptl);
	cond_resched();

5299 5300 5301
	return 0;
}

5302 5303 5304 5305 5306
static unsigned long mem_cgroup_count_precharge(struct mm_struct *mm)
{
	unsigned long precharge;
	struct vm_area_struct *vma;

5307
	down_read(&mm->mmap_sem);
5308 5309 5310 5311 5312 5313 5314 5315 5316 5317 5318
	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);
	}
5319
	up_read(&mm->mmap_sem);
5320 5321 5322 5323 5324 5325 5326 5327 5328

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

	return precharge;
}

static int mem_cgroup_precharge_mc(struct mm_struct *mm)
{
5329 5330 5331 5332 5333
	unsigned long precharge = mem_cgroup_count_precharge(mm);

	VM_BUG_ON(mc.moving_task);
	mc.moving_task = current;
	return mem_cgroup_do_precharge(precharge);
5334 5335
}

5336 5337
/* cancels all extra charges on mc.from and mc.to, and wakes up all waiters. */
static void __mem_cgroup_clear_mc(void)
5338
{
5339 5340 5341
	struct mem_cgroup *from = mc.from;
	struct mem_cgroup *to = mc.to;

5342
	/* we must uncharge all the leftover precharges from mc.to */
5343 5344 5345 5346 5347 5348 5349 5350 5351 5352 5353
	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;
5354
	}
5355 5356 5357 5358 5359 5360 5361 5362 5363 5364 5365 5366 5367 5368 5369 5370 5371 5372 5373
	/* 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;
	}
5374 5375 5376 5377 5378 5379 5380 5381 5382 5383 5384 5385 5386 5387 5388
	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();
5389
	spin_lock(&mc.lock);
5390 5391
	mc.from = NULL;
	mc.to = NULL;
5392
	spin_unlock(&mc.lock);
5393
	mem_cgroup_end_move(from);
5394 5395
}

5396 5397
static int mem_cgroup_can_attach(struct cgroup_subsys *ss,
				struct cgroup *cgroup,
5398
				struct cgroup_taskset *tset)
5399
{
5400
	struct task_struct *p = cgroup_taskset_first(tset);
5401
	int ret = 0;
5402
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgroup);
5403

5404
	if (memcg->move_charge_at_immigrate) {
5405 5406 5407
		struct mm_struct *mm;
		struct mem_cgroup *from = mem_cgroup_from_task(p);

5408
		VM_BUG_ON(from == memcg);
5409 5410 5411 5412 5413

		mm = get_task_mm(p);
		if (!mm)
			return 0;
		/* We move charges only when we move a owner of the mm */
5414 5415 5416 5417
		if (mm->owner == p) {
			VM_BUG_ON(mc.from);
			VM_BUG_ON(mc.to);
			VM_BUG_ON(mc.precharge);
5418
			VM_BUG_ON(mc.moved_charge);
5419
			VM_BUG_ON(mc.moved_swap);
5420
			mem_cgroup_start_move(from);
5421
			spin_lock(&mc.lock);
5422
			mc.from = from;
5423
			mc.to = memcg;
5424
			spin_unlock(&mc.lock);
5425
			/* We set mc.moving_task later */
5426 5427 5428 5429

			ret = mem_cgroup_precharge_mc(mm);
			if (ret)
				mem_cgroup_clear_mc();
5430 5431
		}
		mmput(mm);
5432 5433 5434 5435 5436 5437
	}
	return ret;
}

static void mem_cgroup_cancel_attach(struct cgroup_subsys *ss,
				struct cgroup *cgroup,
5438
				struct cgroup_taskset *tset)
5439
{
5440
	mem_cgroup_clear_mc();
5441 5442
}

5443 5444 5445
static int mem_cgroup_move_charge_pte_range(pmd_t *pmd,
				unsigned long addr, unsigned long end,
				struct mm_walk *walk)
5446
{
5447 5448 5449 5450 5451
	int ret = 0;
	struct vm_area_struct *vma = walk->private;
	pte_t *pte;
	spinlock_t *ptl;

5452
	split_huge_page_pmd(walk->mm, pmd);
5453 5454 5455 5456 5457 5458 5459 5460
retry:
	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
	for (; addr != end; addr += PAGE_SIZE) {
		pte_t ptent = *(pte++);
		union mc_target target;
		int type;
		struct page *page;
		struct page_cgroup *pc;
5461
		swp_entry_t ent;
5462 5463 5464 5465 5466 5467 5468 5469 5470 5471 5472

		if (!mc.precharge)
			break;

		type = is_target_pte_for_mc(vma, addr, ptent, &target);
		switch (type) {
		case MC_TARGET_PAGE:
			page = target.page;
			if (isolate_lru_page(page))
				goto put;
			pc = lookup_page_cgroup(page);
5473 5474
			if (!mem_cgroup_move_account(page, 1, pc,
						     mc.from, mc.to, false)) {
5475
				mc.precharge--;
5476 5477
				/* we uncharge from mc.from later. */
				mc.moved_charge++;
5478 5479 5480 5481 5482
			}
			putback_lru_page(page);
put:			/* is_target_pte_for_mc() gets the page */
			put_page(page);
			break;
5483 5484
		case MC_TARGET_SWAP:
			ent = target.ent;
5485 5486
			if (!mem_cgroup_move_swap_account(ent,
						mc.from, mc.to, false)) {
5487
				mc.precharge--;
5488 5489 5490
				/* we fixup refcnts and charges later. */
				mc.moved_swap++;
			}
5491
			break;
5492 5493 5494 5495 5496 5497 5498 5499 5500 5501 5502 5503 5504 5505
		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.
		 */
5506
		ret = mem_cgroup_do_precharge(1);
5507 5508 5509 5510 5511 5512 5513 5514 5515 5516 5517 5518
		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();
5519 5520 5521 5522 5523 5524 5525 5526 5527 5528 5529 5530 5531
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;
	}
5532 5533 5534 5535 5536 5537 5538 5539 5540 5541 5542 5543 5544 5545 5546 5547 5548 5549
	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;
	}
5550
	up_read(&mm->mmap_sem);
5551 5552
}

B
Balbir Singh 已提交
5553 5554
static void mem_cgroup_move_task(struct cgroup_subsys *ss,
				struct cgroup *cont,
5555
				struct cgroup_taskset *tset)
B
Balbir Singh 已提交
5556
{
5557
	struct task_struct *p = cgroup_taskset_first(tset);
5558
	struct mm_struct *mm = get_task_mm(p);
5559 5560

	if (mm) {
5561 5562 5563
		if (mc.to)
			mem_cgroup_move_charge(mm);
		put_swap_token(mm);
5564 5565
		mmput(mm);
	}
5566 5567
	if (mc.to)
		mem_cgroup_clear_mc();
B
Balbir Singh 已提交
5568
}
5569 5570 5571
#else	/* !CONFIG_MMU */
static int mem_cgroup_can_attach(struct cgroup_subsys *ss,
				struct cgroup *cgroup,
5572
				struct cgroup_taskset *tset)
5573 5574 5575 5576 5577
{
	return 0;
}
static void mem_cgroup_cancel_attach(struct cgroup_subsys *ss,
				struct cgroup *cgroup,
5578
				struct cgroup_taskset *tset)
5579 5580 5581 5582
{
}
static void mem_cgroup_move_task(struct cgroup_subsys *ss,
				struct cgroup *cont,
5583
				struct cgroup_taskset *tset)
5584 5585 5586
{
}
#endif
B
Balbir Singh 已提交
5587

B
Balbir Singh 已提交
5588 5589 5590 5591
struct cgroup_subsys mem_cgroup_subsys = {
	.name = "memory",
	.subsys_id = mem_cgroup_subsys_id,
	.create = mem_cgroup_create,
5592
	.pre_destroy = mem_cgroup_pre_destroy,
B
Balbir Singh 已提交
5593 5594
	.destroy = mem_cgroup_destroy,
	.populate = mem_cgroup_populate,
5595 5596
	.can_attach = mem_cgroup_can_attach,
	.cancel_attach = mem_cgroup_cancel_attach,
B
Balbir Singh 已提交
5597
	.attach = mem_cgroup_move_task,
5598
	.early_init = 0,
K
KAMEZAWA Hiroyuki 已提交
5599
	.use_id = 1,
B
Balbir Singh 已提交
5600
};
5601 5602

#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
5603 5604 5605
static int __init enable_swap_account(char *s)
{
	/* consider enabled if no parameter or 1 is given */
5606
	if (!strcmp(s, "1"))
5607
		really_do_swap_account = 1;
5608
	else if (!strcmp(s, "0"))
5609 5610 5611
		really_do_swap_account = 0;
	return 1;
}
5612
__setup("swapaccount=", enable_swap_account);
5613 5614

#endif