memcontrol.c 143.2 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
};

126 127 128 129 130 131 132
struct mem_cgroup_reclaim_iter {
	/* css_id of the last scanned hierarchy member */
	int position;
	/* scan generation, increased every round-trip */
	unsigned int generation;
};

133 134 135 136
/*
 * per-zone information in memory controller.
 */
struct mem_cgroup_per_zone {
137
	struct lruvec		lruvec;
138
	unsigned long		lru_size[NR_LRU_LISTS];
K
KOSAKI Motohiro 已提交
139

140 141
	struct mem_cgroup_reclaim_iter reclaim_iter[DEF_PRIORITY + 1];

K
KOSAKI Motohiro 已提交
142
	struct zone_reclaim_stat reclaim_stat;
143 144 145 146
	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;
147
	struct mem_cgroup	*memcg;		/* Back pointer, we cannot */
148
						/* use container_of	   */
149 150 151 152 153 154 155 156 157 158
};

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];
};

159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178
/*
 * 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;

179 180 181 182 183
struct mem_cgroup_threshold {
	struct eventfd_ctx *eventfd;
	u64 threshold;
};

K
KAMEZAWA Hiroyuki 已提交
184
/* For threshold */
185 186
struct mem_cgroup_threshold_ary {
	/* An array index points to threshold just below usage. */
187
	int current_threshold;
188 189 190 191 192
	/* Size of entries[] */
	unsigned int size;
	/* Array of thresholds */
	struct mem_cgroup_threshold entries[0];
};
193 194 195 196 197 198 199 200 201 202 203 204

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
205 206 207 208 209
/* for OOM */
struct mem_cgroup_eventfd_list {
	struct list_head list;
	struct eventfd_ctx *eventfd;
};
210

211 212
static void mem_cgroup_threshold(struct mem_cgroup *memcg);
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg);
213

B
Balbir Singh 已提交
214 215 216 217 218 219 220
/*
 * 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
221 222 223
 * 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
224 225 226 227 228 229 230
 */
struct mem_cgroup {
	struct cgroup_subsys_state css;
	/*
	 * the counter to account for memory usage
	 */
	struct res_counter res;
231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254

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

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

255 256 257 258
	/*
	 * Per cgroup active and inactive list, similar to the
	 * per zone LRU lists.
	 */
259
	struct mem_cgroup_lru_info info;
260 261 262
	int last_scanned_node;
#if MAX_NUMNODES > 1
	nodemask_t	scan_nodes;
263 264
	atomic_t	numainfo_events;
	atomic_t	numainfo_updating;
265
#endif
266 267 268 269
	/*
	 * Should the accounting and control be hierarchical, per subtree?
	 */
	bool use_hierarchy;
270 271 272 273

	bool		oom_lock;
	atomic_t	under_oom;

274
	atomic_t	refcnt;
275

276
	int	swappiness;
277 278
	/* OOM-Killer disable */
	int		oom_kill_disable;
K
KOSAKI Motohiro 已提交
279

280 281 282
	/* set when res.limit == memsw.limit */
	bool		memsw_is_minimum;

283 284 285 286
	/* protect arrays of thresholds */
	struct mutex thresholds_lock;

	/* thresholds for memory usage. RCU-protected */
287
	struct mem_cgroup_thresholds thresholds;
288

289
	/* thresholds for mem+swap usage. RCU-protected */
290
	struct mem_cgroup_thresholds memsw_thresholds;
291

K
KAMEZAWA Hiroyuki 已提交
292 293
	/* For oom notifier event fd */
	struct list_head oom_notify;
294

295 296 297 298 299
	/*
	 * 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;
300
	/*
301
	 * percpu counter.
302
	 */
303
	struct mem_cgroup_stat_cpu *stat;
304 305 306 307 308 309
	/*
	 * 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
310 311 312 313

#ifdef CONFIG_INET
	struct tcp_memcontrol tcp_mem;
#endif
B
Balbir Singh 已提交
314 315
};

316 317 318 319 320 321
/* 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 {
322
	MOVE_CHARGE_TYPE_ANON,	/* private anonymous page and swap of it */
323
	MOVE_CHARGE_TYPE_FILE,	/* file page(including tmpfs) and swap of it */
324 325 326
	NR_MOVE_TYPE,
};

327 328
/* "mc" and its members are protected by cgroup_mutex */
static struct move_charge_struct {
329
	spinlock_t	  lock; /* for from, to */
330 331 332
	struct mem_cgroup *from;
	struct mem_cgroup *to;
	unsigned long precharge;
333
	unsigned long moved_charge;
334
	unsigned long moved_swap;
335 336 337
	struct task_struct *moving_task;	/* a task moving charges */
	wait_queue_head_t waitq;		/* a waitq for other context */
} mc = {
338
	.lock = __SPIN_LOCK_UNLOCKED(mc.lock),
339 340
	.waitq = __WAIT_QUEUE_HEAD_INITIALIZER(mc.waitq),
};
341

D
Daisuke Nishimura 已提交
342 343 344 345 346 347
static bool move_anon(void)
{
	return test_bit(MOVE_CHARGE_TYPE_ANON,
					&mc.to->move_charge_at_immigrate);
}

348 349 350 351 352 353
static bool move_file(void)
{
	return test_bit(MOVE_CHARGE_TYPE_FILE,
					&mc.to->move_charge_at_immigrate);
}

354 355 356 357 358 359 360
/*
 * 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)

361 362 363
enum charge_type {
	MEM_CGROUP_CHARGE_TYPE_CACHE = 0,
	MEM_CGROUP_CHARGE_TYPE_MAPPED,
364
	MEM_CGROUP_CHARGE_TYPE_SHMEM,	/* used by page migration of shmem */
365
	MEM_CGROUP_CHARGE_TYPE_FORCE,	/* used by force_empty */
K
KAMEZAWA Hiroyuki 已提交
366
	MEM_CGROUP_CHARGE_TYPE_SWAPOUT,	/* for accounting swapcache */
K
KAMEZAWA Hiroyuki 已提交
367
	MEM_CGROUP_CHARGE_TYPE_DROP,	/* a page was unused swap cache */
368 369 370
	NR_CHARGE_TYPE,
};

371
/* for encoding cft->private value on file */
372 373 374
#define _MEM			(0)
#define _MEMSWAP		(1)
#define _OOM_TYPE		(2)
375 376 377
#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
378 379
/* Used for OOM nofiier */
#define OOM_CONTROL		(0)
380

381 382 383 384 385 386 387 388
/*
 * 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)

389 390
static void mem_cgroup_get(struct mem_cgroup *memcg);
static void mem_cgroup_put(struct mem_cgroup *memcg);
G
Glauber Costa 已提交
391 392 393 394

/* Writing them here to avoid exposing memcg's inner layout */
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
#include <net/sock.h>
G
Glauber Costa 已提交
395
#include <net/ip.h>
G
Glauber Costa 已提交
396 397 398 399

static bool mem_cgroup_is_root(struct mem_cgroup *memcg);
void sock_update_memcg(struct sock *sk)
{
400
	if (mem_cgroup_sockets_enabled) {
G
Glauber Costa 已提交
401 402 403 404
		struct mem_cgroup *memcg;

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

405 406 407 408 409 410 411 412 413 414 415 416 417 418
		/* 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
419 420 421 422 423 424 425 426 427 428 429 430 431
		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)
{
432
	if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
G
Glauber Costa 已提交
433 434 435 436 437 438
		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
439

440
#ifdef CONFIG_INET
G
Glauber Costa 已提交
441 442 443 444 445 446 447 448
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
449 450 451
#endif /* CONFIG_INET */
#endif /* CONFIG_CGROUP_MEM_RES_CTLR_KMEM */

452
static void drain_all_stock_async(struct mem_cgroup *memcg);
453

454
static struct mem_cgroup_per_zone *
455
mem_cgroup_zoneinfo(struct mem_cgroup *memcg, int nid, int zid)
456
{
457
	return &memcg->info.nodeinfo[nid]->zoneinfo[zid];
458 459
}

460
struct cgroup_subsys_state *mem_cgroup_css(struct mem_cgroup *memcg)
461
{
462
	return &memcg->css;
463 464
}

465
static struct mem_cgroup_per_zone *
466
page_cgroup_zoneinfo(struct mem_cgroup *memcg, struct page *page)
467
{
468 469
	int nid = page_to_nid(page);
	int zid = page_zonenum(page);
470

471
	return mem_cgroup_zoneinfo(memcg, nid, zid);
472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489
}

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
490
__mem_cgroup_insert_exceeded(struct mem_cgroup *memcg,
491
				struct mem_cgroup_per_zone *mz,
492 493
				struct mem_cgroup_tree_per_zone *mctz,
				unsigned long long new_usage_in_excess)
494 495 496 497 498 499 500 501
{
	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;

502 503 504
	mz->usage_in_excess = new_usage_in_excess;
	if (!mz->usage_in_excess)
		return;
505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520
	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;
521 522 523
}

static void
524
__mem_cgroup_remove_exceeded(struct mem_cgroup *memcg,
525 526 527 528 529 530 531 532 533
				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;
}

534
static void
535
mem_cgroup_remove_exceeded(struct mem_cgroup *memcg,
536 537 538 539
				struct mem_cgroup_per_zone *mz,
				struct mem_cgroup_tree_per_zone *mctz)
{
	spin_lock(&mctz->lock);
540
	__mem_cgroup_remove_exceeded(memcg, mz, mctz);
541 542 543 544
	spin_unlock(&mctz->lock);
}


545
static void mem_cgroup_update_tree(struct mem_cgroup *memcg, struct page *page)
546
{
547
	unsigned long long excess;
548 549
	struct mem_cgroup_per_zone *mz;
	struct mem_cgroup_tree_per_zone *mctz;
550 551
	int nid = page_to_nid(page);
	int zid = page_zonenum(page);
552 553 554
	mctz = soft_limit_tree_from_page(page);

	/*
555 556
	 * Necessary to update all ancestors when hierarchy is used.
	 * because their event counter is not touched.
557
	 */
558 559 560
	for (; memcg; memcg = parent_mem_cgroup(memcg)) {
		mz = mem_cgroup_zoneinfo(memcg, nid, zid);
		excess = res_counter_soft_limit_excess(&memcg->res);
561 562 563 564
		/*
		 * We have to update the tree if mz is on RB-tree or
		 * mem is over its softlimit.
		 */
565
		if (excess || mz->on_tree) {
566 567 568
			spin_lock(&mctz->lock);
			/* if on-tree, remove it */
			if (mz->on_tree)
569
				__mem_cgroup_remove_exceeded(memcg, mz, mctz);
570
			/*
571 572
			 * Insert again. mz->usage_in_excess will be updated.
			 * If excess is 0, no tree ops.
573
			 */
574
			__mem_cgroup_insert_exceeded(memcg, mz, mctz, excess);
575 576
			spin_unlock(&mctz->lock);
		}
577 578 579
	}
}

580
static void mem_cgroup_remove_from_trees(struct mem_cgroup *memcg)
581 582 583 584 585
{
	int node, zone;
	struct mem_cgroup_per_zone *mz;
	struct mem_cgroup_tree_per_zone *mctz;

B
Bob Liu 已提交
586
	for_each_node(node) {
587
		for (zone = 0; zone < MAX_NR_ZONES; zone++) {
588
			mz = mem_cgroup_zoneinfo(memcg, node, zone);
589
			mctz = soft_limit_tree_node_zone(node, zone);
590
			mem_cgroup_remove_exceeded(memcg, mz, mctz);
591 592 593 594
		}
	}
}

595 596 597 598
static struct mem_cgroup_per_zone *
__mem_cgroup_largest_soft_limit_node(struct mem_cgroup_tree_per_zone *mctz)
{
	struct rb_node *rightmost = NULL;
599
	struct mem_cgroup_per_zone *mz;
600 601

retry:
602
	mz = NULL;
603 604 605 606 607 608 609 610 611 612
	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.
	 */
613 614 615
	__mem_cgroup_remove_exceeded(mz->memcg, mz, mctz);
	if (!res_counter_soft_limit_excess(&mz->memcg->res) ||
		!css_tryget(&mz->memcg->css))
616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631
		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;
}

632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650
/*
 * 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.
 */
651
static long mem_cgroup_read_stat(struct mem_cgroup *memcg,
652
				 enum mem_cgroup_stat_index idx)
653
{
654
	long val = 0;
655 656
	int cpu;

657 658
	get_online_cpus();
	for_each_online_cpu(cpu)
659
		val += per_cpu(memcg->stat->count[idx], cpu);
660
#ifdef CONFIG_HOTPLUG_CPU
661 662 663
	spin_lock(&memcg->pcp_counter_lock);
	val += memcg->nocpu_base.count[idx];
	spin_unlock(&memcg->pcp_counter_lock);
664 665
#endif
	put_online_cpus();
666 667 668
	return val;
}

669
static void mem_cgroup_swap_statistics(struct mem_cgroup *memcg,
670 671 672
					 bool charge)
{
	int val = (charge) ? 1 : -1;
673
	this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_SWAPOUT], val);
674 675
}

676
static unsigned long mem_cgroup_read_events(struct mem_cgroup *memcg,
677 678 679 680 681 682
					    enum mem_cgroup_events_index idx)
{
	unsigned long val = 0;
	int cpu;

	for_each_online_cpu(cpu)
683
		val += per_cpu(memcg->stat->events[idx], cpu);
684
#ifdef CONFIG_HOTPLUG_CPU
685 686 687
	spin_lock(&memcg->pcp_counter_lock);
	val += memcg->nocpu_base.events[idx];
	spin_unlock(&memcg->pcp_counter_lock);
688 689 690 691
#endif
	return val;
}

692
static void mem_cgroup_charge_statistics(struct mem_cgroup *memcg,
693
					 bool file, int nr_pages)
694
{
695 696
	preempt_disable();

697
	if (file)
698 699
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_CACHE],
				nr_pages);
700
	else
701 702
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_RSS],
				nr_pages);
703

704 705
	/* pagein of a big page is an event. So, ignore page size */
	if (nr_pages > 0)
706
		__this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGPGIN]);
707
	else {
708
		__this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGPGOUT]);
709 710
		nr_pages = -nr_pages; /* for event */
	}
711

712
	__this_cpu_add(memcg->stat->events[MEM_CGROUP_EVENTS_COUNT], nr_pages);
713

714
	preempt_enable();
715 716
}

717
unsigned long
718
mem_cgroup_zone_nr_lru_pages(struct mem_cgroup *memcg, int nid, int zid,
719
			unsigned int lru_mask)
720 721
{
	struct mem_cgroup_per_zone *mz;
H
Hugh Dickins 已提交
722
	enum lru_list lru;
723 724
	unsigned long ret = 0;

725
	mz = mem_cgroup_zoneinfo(memcg, nid, zid);
726

H
Hugh Dickins 已提交
727 728 729
	for_each_lru(lru) {
		if (BIT(lru) & lru_mask)
			ret += mz->lru_size[lru];
730 731 732 733 734
	}
	return ret;
}

static unsigned long
735
mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
736 737
			int nid, unsigned int lru_mask)
{
738 739 740
	u64 total = 0;
	int zid;

741
	for (zid = 0; zid < MAX_NR_ZONES; zid++)
742 743
		total += mem_cgroup_zone_nr_lru_pages(memcg,
						nid, zid, lru_mask);
744

745 746
	return total;
}
747

748
static unsigned long mem_cgroup_nr_lru_pages(struct mem_cgroup *memcg,
749
			unsigned int lru_mask)
750
{
751
	int nid;
752 753
	u64 total = 0;

754
	for_each_node_state(nid, N_HIGH_MEMORY)
755
		total += mem_cgroup_node_nr_lru_pages(memcg, nid, lru_mask);
756
	return total;
757 758
}

759 760
static bool mem_cgroup_event_ratelimit(struct mem_cgroup *memcg,
				       enum mem_cgroup_events_target target)
761 762 763
{
	unsigned long val, next;

764 765
	val = __this_cpu_read(memcg->stat->events[MEM_CGROUP_EVENTS_COUNT]);
	next = __this_cpu_read(memcg->stat->targets[target]);
766
	/* from time_after() in jiffies.h */
767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782
	if ((long)next - (long)val < 0) {
		switch (target) {
		case MEM_CGROUP_TARGET_THRESH:
			next = val + THRESHOLDS_EVENTS_TARGET;
			break;
		case MEM_CGROUP_TARGET_SOFTLIMIT:
			next = val + SOFTLIMIT_EVENTS_TARGET;
			break;
		case MEM_CGROUP_TARGET_NUMAINFO:
			next = val + NUMAINFO_EVENTS_TARGET;
			break;
		default:
			break;
		}
		__this_cpu_write(memcg->stat->targets[target], next);
		return true;
783
	}
784
	return false;
785 786 787 788 789 790
}

/*
 * Check events in order.
 *
 */
791
static void memcg_check_events(struct mem_cgroup *memcg, struct page *page)
792
{
793
	preempt_disable();
794
	/* threshold event is triggered in finer grain than soft limit */
795 796
	if (unlikely(mem_cgroup_event_ratelimit(memcg,
						MEM_CGROUP_TARGET_THRESH))) {
797 798
		bool do_softlimit;
		bool do_numainfo __maybe_unused;
799 800 801 802 803 804 805 806 807

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

808
		mem_cgroup_threshold(memcg);
809
		if (unlikely(do_softlimit))
810
			mem_cgroup_update_tree(memcg, page);
811
#if MAX_NUMNODES > 1
812
		if (unlikely(do_numainfo))
813
			atomic_inc(&memcg->numainfo_events);
814
#endif
815 816
	} else
		preempt_enable();
817 818
}

G
Glauber Costa 已提交
819
struct mem_cgroup *mem_cgroup_from_cont(struct cgroup *cont)
B
Balbir Singh 已提交
820 821 822 823 824 825
{
	return container_of(cgroup_subsys_state(cont,
				mem_cgroup_subsys_id), struct mem_cgroup,
				css);
}

826
struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p)
827
{
828 829 830 831 832 833 834 835
	/*
	 * 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;

836 837 838 839
	return container_of(task_subsys_state(p, mem_cgroup_subsys_id),
				struct mem_cgroup, css);
}

840
struct mem_cgroup *try_get_mem_cgroup_from_mm(struct mm_struct *mm)
841
{
842
	struct mem_cgroup *memcg = NULL;
843 844 845

	if (!mm)
		return NULL;
846 847 848 849 850 851 852
	/*
	 * 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 {
853 854
		memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
		if (unlikely(!memcg))
855
			break;
856
	} while (!css_tryget(&memcg->css));
857
	rcu_read_unlock();
858
	return memcg;
859 860
}

861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880
/**
 * mem_cgroup_iter - iterate over memory cgroup hierarchy
 * @root: hierarchy root
 * @prev: previously returned memcg, NULL on first invocation
 * @reclaim: cookie for shared reclaim walks, NULL for full walks
 *
 * Returns references to children of the hierarchy below @root, or
 * @root itself, or %NULL after a full round-trip.
 *
 * Caller must pass the return value in @prev on subsequent
 * invocations for reference counting, or use mem_cgroup_iter_break()
 * to cancel a hierarchy walk before the round-trip is complete.
 *
 * Reclaimers can specify a zone and a priority level in @reclaim to
 * divide up the memcgs in the hierarchy among all concurrent
 * reclaimers operating on the same zone and priority.
 */
struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *root,
				   struct mem_cgroup *prev,
				   struct mem_cgroup_reclaim_cookie *reclaim)
K
KAMEZAWA Hiroyuki 已提交
881
{
882 883
	struct mem_cgroup *memcg = NULL;
	int id = 0;
884

885 886 887
	if (mem_cgroup_disabled())
		return NULL;

888 889
	if (!root)
		root = root_mem_cgroup;
K
KAMEZAWA Hiroyuki 已提交
890

891 892
	if (prev && !reclaim)
		id = css_id(&prev->css);
K
KAMEZAWA Hiroyuki 已提交
893

894 895
	if (prev && prev != root)
		css_put(&prev->css);
K
KAMEZAWA Hiroyuki 已提交
896

897 898 899 900 901
	if (!root->use_hierarchy && root != root_mem_cgroup) {
		if (prev)
			return NULL;
		return root;
	}
K
KAMEZAWA Hiroyuki 已提交
902

903
	while (!memcg) {
904
		struct mem_cgroup_reclaim_iter *uninitialized_var(iter);
905
		struct cgroup_subsys_state *css;
906

907 908 909 910 911 912 913 914 915 916 917
		if (reclaim) {
			int nid = zone_to_nid(reclaim->zone);
			int zid = zone_idx(reclaim->zone);
			struct mem_cgroup_per_zone *mz;

			mz = mem_cgroup_zoneinfo(root, nid, zid);
			iter = &mz->reclaim_iter[reclaim->priority];
			if (prev && reclaim->generation != iter->generation)
				return NULL;
			id = iter->position;
		}
K
KAMEZAWA Hiroyuki 已提交
918

919 920 921 922 923 924 925 926
		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
927 928
		rcu_read_unlock();

929 930 931 932 933 934 935
		if (reclaim) {
			iter->position = id;
			if (!css)
				iter->generation++;
			else if (!prev && memcg)
				reclaim->generation = iter->generation;
		}
936 937 938 939 940

		if (prev && !css)
			return NULL;
	}
	return memcg;
K
KAMEZAWA Hiroyuki 已提交
941
}
K
KAMEZAWA Hiroyuki 已提交
942

943 944 945 946 947 948 949
/**
 * mem_cgroup_iter_break - abort a hierarchy walk prematurely
 * @root: hierarchy root
 * @prev: last visited hierarchy member as returned by mem_cgroup_iter()
 */
void mem_cgroup_iter_break(struct mem_cgroup *root,
			   struct mem_cgroup *prev)
950 951 952 953 954 955
{
	if (!root)
		root = root_mem_cgroup;
	if (prev && prev != root)
		css_put(&prev->css);
}
K
KAMEZAWA Hiroyuki 已提交
956

957 958 959 960 961 962
/*
 * 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)		\
963
	for (iter = mem_cgroup_iter(root, NULL, NULL);	\
964
	     iter != NULL;				\
965
	     iter = mem_cgroup_iter(root, iter, NULL))
966

967
#define for_each_mem_cgroup(iter)			\
968
	for (iter = mem_cgroup_iter(NULL, NULL, NULL);	\
969
	     iter != NULL;				\
970
	     iter = mem_cgroup_iter(NULL, iter, NULL))
K
KAMEZAWA Hiroyuki 已提交
971

972
static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
973
{
974
	return (memcg == root_mem_cgroup);
975 976
}

977 978
void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
{
979
	struct mem_cgroup *memcg;
980 981 982 983 984

	if (!mm)
		return;

	rcu_read_lock();
985 986
	memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
	if (unlikely(!memcg))
987 988 989 990
		goto out;

	switch (idx) {
	case PGFAULT:
991 992 993 994
		this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]);
		break;
	case PGMAJFAULT:
		this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]);
995 996 997 998 999 1000 1001 1002 1003
		break;
	default:
		BUG();
	}
out:
	rcu_read_unlock();
}
EXPORT_SYMBOL(mem_cgroup_count_vm_event);

1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024
/**
 * mem_cgroup_zone_lruvec - get the lru list vector for a zone and memcg
 * @zone: zone of the wanted lruvec
 * @mem: memcg of the wanted lruvec
 *
 * Returns the lru list vector holding pages for the given @zone and
 * @mem.  This can be the global zone lruvec, if the memory controller
 * is disabled.
 */
struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
				      struct mem_cgroup *memcg)
{
	struct mem_cgroup_per_zone *mz;

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

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

K
KAMEZAWA Hiroyuki 已提交
1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037
/*
 * 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.
 */
1038

1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052
/**
 * mem_cgroup_lru_add_list - account for adding an lru page and return lruvec
 * @zone: zone of the page
 * @page: the page
 * @lru: current lru
 *
 * This function accounts for @page being added to @lru, and returns
 * the lruvec for the given @zone and the memcg @page is charged to.
 *
 * The callsite is then responsible for physically linking the page to
 * the returned lruvec->lists[@lru].
 */
struct lruvec *mem_cgroup_lru_add_list(struct zone *zone, struct page *page,
				       enum lru_list lru)
K
KAMEZAWA Hiroyuki 已提交
1053 1054
{
	struct mem_cgroup_per_zone *mz;
1055 1056
	struct mem_cgroup *memcg;
	struct page_cgroup *pc;
1057

1058
	if (mem_cgroup_disabled())
1059 1060
		return &zone->lruvec;

K
KAMEZAWA Hiroyuki 已提交
1061
	pc = lookup_page_cgroup(page);
1062
	memcg = pc->mem_cgroup;
1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075

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

1076 1077
	mz = page_cgroup_zoneinfo(memcg, page);
	/* compound_order() is stabilized through lru_lock */
1078
	mz->lru_size[lru] += 1 << compound_order(page);
1079
	return &mz->lruvec;
K
KAMEZAWA Hiroyuki 已提交
1080
}
1081

1082 1083 1084 1085 1086 1087 1088 1089 1090
/**
 * mem_cgroup_lru_del_list - account for removing an lru page
 * @page: the page
 * @lru: target lru
 *
 * This function accounts for @page being removed from @lru.
 *
 * The callsite is then responsible for physically unlinking
 * @page->lru.
1091
 */
1092
void mem_cgroup_lru_del_list(struct page *page, enum lru_list lru)
1093 1094
{
	struct mem_cgroup_per_zone *mz;
1095
	struct mem_cgroup *memcg;
1096 1097 1098 1099 1100 1101
	struct page_cgroup *pc;

	if (mem_cgroup_disabled())
		return;

	pc = lookup_page_cgroup(page);
1102 1103
	memcg = pc->mem_cgroup;
	VM_BUG_ON(!memcg);
1104 1105
	mz = page_cgroup_zoneinfo(memcg, page);
	/* huge page split is done under lru_lock. so, we have no races. */
1106 1107
	VM_BUG_ON(mz->lru_size[lru] < (1 << compound_order(page)));
	mz->lru_size[lru] -= 1 << compound_order(page);
1108 1109
}

1110
void mem_cgroup_lru_del(struct page *page)
K
KAMEZAWA Hiroyuki 已提交
1111
{
1112
	mem_cgroup_lru_del_list(page, page_lru(page));
1113 1114
}

1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132
/**
 * mem_cgroup_lru_move_lists - account for moving a page between lrus
 * @zone: zone of the page
 * @page: the page
 * @from: current lru
 * @to: target lru
 *
 * This function accounts for @page being moved between the lrus @from
 * and @to, and returns the lruvec for the given @zone and the memcg
 * @page is charged to.
 *
 * The callsite is then responsible for physically relinking
 * @page->lru to the returned lruvec->lists[@to].
 */
struct lruvec *mem_cgroup_lru_move_lists(struct zone *zone,
					 struct page *page,
					 enum lru_list from,
					 enum lru_list to)
1133
{
1134 1135 1136
	/* XXX: Optimize this, especially for @from == @to */
	mem_cgroup_lru_del_list(page, from);
	return mem_cgroup_lru_add_list(zone, page, to);
K
KAMEZAWA Hiroyuki 已提交
1137
}
1138

1139
/*
1140
 * Checks whether given mem is same or in the root_mem_cgroup's
1141 1142
 * hierarchy subtree
 */
1143 1144
static bool mem_cgroup_same_or_subtree(const struct mem_cgroup *root_memcg,
		struct mem_cgroup *memcg)
1145
{
1146 1147 1148
	if (root_memcg != memcg) {
		return (root_memcg->use_hierarchy &&
			css_is_ancestor(&memcg->css, &root_memcg->css));
1149 1150 1151 1152 1153
	}

	return true;
}

1154
int task_in_mem_cgroup(struct task_struct *task, const struct mem_cgroup *memcg)
1155 1156
{
	int ret;
1157
	struct mem_cgroup *curr = NULL;
1158
	struct task_struct *p;
1159

1160
	p = find_lock_task_mm(task);
1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175
	if (p) {
		curr = try_get_mem_cgroup_from_mm(p->mm);
		task_unlock(p);
	} else {
		/*
		 * All threads may have already detached their mm's, but the oom
		 * killer still needs to detect if they have already been oom
		 * killed to prevent needlessly killing additional tasks.
		 */
		task_lock(task);
		curr = mem_cgroup_from_task(task);
		if (curr)
			css_get(&curr->css);
		task_unlock(task);
	}
1176 1177
	if (!curr)
		return 0;
1178
	/*
1179
	 * We should check use_hierarchy of "memcg" not "curr". Because checking
1180
	 * use_hierarchy of "curr" here make this function true if hierarchy is
1181 1182
	 * enabled in "curr" and "curr" is a child of "memcg" in *cgroup*
	 * hierarchy(even if use_hierarchy is disabled in "memcg").
1183
	 */
1184
	ret = mem_cgroup_same_or_subtree(memcg, curr);
1185
	css_put(&curr->css);
1186 1187 1188
	return ret;
}

1189
int mem_cgroup_inactive_anon_is_low(struct mem_cgroup *memcg, struct zone *zone)
1190
{
1191 1192 1193
	unsigned long inactive_ratio;
	int nid = zone_to_nid(zone);
	int zid = zone_idx(zone);
1194
	unsigned long inactive;
1195
	unsigned long active;
1196
	unsigned long gb;
1197

1198 1199 1200 1201
	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));
1202

1203 1204 1205 1206 1207 1208
	gb = (inactive + active) >> (30 - PAGE_SHIFT);
	if (gb)
		inactive_ratio = int_sqrt(10 * gb);
	else
		inactive_ratio = 1;

1209
	return inactive * inactive_ratio < active;
1210 1211
}

1212
int mem_cgroup_inactive_file_is_low(struct mem_cgroup *memcg, struct zone *zone)
1213 1214 1215
{
	unsigned long active;
	unsigned long inactive;
1216 1217
	int zid = zone_idx(zone);
	int nid = zone_to_nid(zone);
1218

1219 1220 1221 1222
	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));
1223 1224 1225 1226

	return (active > inactive);
}

K
KOSAKI Motohiro 已提交
1227 1228 1229
struct zone_reclaim_stat *mem_cgroup_get_reclaim_stat(struct mem_cgroup *memcg,
						      struct zone *zone)
{
1230
	int nid = zone_to_nid(zone);
K
KOSAKI Motohiro 已提交
1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246
	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);
1247 1248
	if (!PageCgroupUsed(pc))
		return NULL;
1249 1250
	/* Ensure pc->mem_cgroup is visible after reading PCG_USED. */
	smp_rmb();
1251
	mz = page_cgroup_zoneinfo(pc->mem_cgroup, page);
K
KOSAKI Motohiro 已提交
1252 1253 1254
	return &mz->reclaim_stat;
}

1255 1256 1257
#define mem_cgroup_from_res_counter(counter, member)	\
	container_of(counter, struct mem_cgroup, member)

1258
/**
1259 1260
 * mem_cgroup_margin - calculate chargeable space of a memory cgroup
 * @mem: the memory cgroup
1261
 *
1262
 * Returns the maximum amount of memory @mem can be charged with, in
1263
 * pages.
1264
 */
1265
static unsigned long mem_cgroup_margin(struct mem_cgroup *memcg)
1266
{
1267 1268
	unsigned long long margin;

1269
	margin = res_counter_margin(&memcg->res);
1270
	if (do_swap_account)
1271
		margin = min(margin, res_counter_margin(&memcg->memsw));
1272
	return margin >> PAGE_SHIFT;
1273 1274
}

1275
int mem_cgroup_swappiness(struct mem_cgroup *memcg)
K
KOSAKI Motohiro 已提交
1276 1277 1278 1279 1280 1281 1282
{
	struct cgroup *cgrp = memcg->css.cgroup;

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

1283
	return memcg->swappiness;
K
KOSAKI Motohiro 已提交
1284 1285
}

1286
static void mem_cgroup_start_move(struct mem_cgroup *memcg)
1287 1288
{
	int cpu;
1289 1290

	get_online_cpus();
1291
	spin_lock(&memcg->pcp_counter_lock);
1292
	for_each_online_cpu(cpu)
1293 1294 1295
		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);
1296
	put_online_cpus();
1297 1298 1299 1300

	synchronize_rcu();
}

1301
static void mem_cgroup_end_move(struct mem_cgroup *memcg)
1302 1303 1304
{
	int cpu;

1305
	if (!memcg)
1306
		return;
1307
	get_online_cpus();
1308
	spin_lock(&memcg->pcp_counter_lock);
1309
	for_each_online_cpu(cpu)
1310 1311 1312
		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);
1313
	put_online_cpus();
1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326
}
/*
 * 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".
 */

1327
static bool mem_cgroup_stealed(struct mem_cgroup *memcg)
1328 1329
{
	VM_BUG_ON(!rcu_read_lock_held());
1330
	return this_cpu_read(memcg->stat->count[MEM_CGROUP_ON_MOVE]) > 0;
1331
}
1332

1333
static bool mem_cgroup_under_move(struct mem_cgroup *memcg)
1334
{
1335 1336
	struct mem_cgroup *from;
	struct mem_cgroup *to;
1337
	bool ret = false;
1338 1339 1340 1341 1342 1343 1344 1345 1346
	/*
	 * 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;
1347

1348 1349
	ret = mem_cgroup_same_or_subtree(memcg, from)
		|| mem_cgroup_same_or_subtree(memcg, to);
1350 1351
unlock:
	spin_unlock(&mc.lock);
1352 1353 1354
	return ret;
}

1355
static bool mem_cgroup_wait_acct_move(struct mem_cgroup *memcg)
1356 1357
{
	if (mc.moving_task && current != mc.moving_task) {
1358
		if (mem_cgroup_under_move(memcg)) {
1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370
			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;
}

1371
/**
1372
 * mem_cgroup_print_oom_info: Called from OOM with tasklist_lock held in read mode.
1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390
 * @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;

1391
	if (!memcg || !p)
1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436
		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));
}

1437 1438 1439 1440
/*
 * This function returns the number of memcg under hierarchy tree. Returns
 * 1(self count) if no children.
 */
1441
static int mem_cgroup_count_children(struct mem_cgroup *memcg)
1442 1443
{
	int num = 0;
K
KAMEZAWA Hiroyuki 已提交
1444 1445
	struct mem_cgroup *iter;

1446
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
1447
		num++;
1448 1449 1450
	return num;
}

D
David Rientjes 已提交
1451 1452 1453 1454 1455 1456 1457 1458
/*
 * Return the memory (and swap, if configured) limit for a memcg.
 */
u64 mem_cgroup_get_limit(struct mem_cgroup *memcg)
{
	u64 limit;
	u64 memsw;

1459 1460 1461
	limit = res_counter_read_u64(&memcg->res, RES_LIMIT);
	limit += total_swap_pages << PAGE_SHIFT;

D
David Rientjes 已提交
1462 1463 1464 1465 1466 1467 1468 1469
	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);
}

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 1499 1500 1501 1502 1503 1504 1505
static unsigned long mem_cgroup_reclaim(struct mem_cgroup *memcg,
					gfp_t gfp_mask,
					unsigned long flags)
{
	unsigned long total = 0;
	bool noswap = false;
	int loop;

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

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

1506 1507 1508 1509 1510 1511 1512 1513 1514 1515
/**
 * 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.
 */
1516
static bool test_mem_cgroup_node_reclaimable(struct mem_cgroup *memcg,
1517 1518
		int nid, bool noswap)
{
1519
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_FILE))
1520 1521 1522
		return true;
	if (noswap || !total_swap_pages)
		return false;
1523
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_ANON))
1524 1525 1526 1527
		return true;
	return false;

}
1528 1529 1530 1531 1532 1533 1534 1535
#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.
 *
 */
1536
static void mem_cgroup_may_update_nodemask(struct mem_cgroup *memcg)
1537 1538
{
	int nid;
1539 1540 1541 1542
	/*
	 * numainfo_events > 0 means there was at least NUMAINFO_EVENTS_TARGET
	 * pagein/pageout changes since the last update.
	 */
1543
	if (!atomic_read(&memcg->numainfo_events))
1544
		return;
1545
	if (atomic_inc_return(&memcg->numainfo_updating) > 1)
1546 1547 1548
		return;

	/* make a nodemask where this memcg uses memory from */
1549
	memcg->scan_nodes = node_states[N_HIGH_MEMORY];
1550 1551 1552

	for_each_node_mask(nid, node_states[N_HIGH_MEMORY]) {

1553 1554
		if (!test_mem_cgroup_node_reclaimable(memcg, nid, false))
			node_clear(nid, memcg->scan_nodes);
1555
	}
1556

1557 1558
	atomic_set(&memcg->numainfo_events, 0);
	atomic_set(&memcg->numainfo_updating, 0);
1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572
}

/*
 * 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.
 */
1573
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
1574 1575 1576
{
	int node;

1577 1578
	mem_cgroup_may_update_nodemask(memcg);
	node = memcg->last_scanned_node;
1579

1580
	node = next_node(node, memcg->scan_nodes);
1581
	if (node == MAX_NUMNODES)
1582
		node = first_node(memcg->scan_nodes);
1583 1584 1585 1586 1587 1588 1589 1590 1591
	/*
	 * 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();

1592
	memcg->last_scanned_node = node;
1593 1594 1595
	return node;
}

1596 1597 1598 1599 1600 1601
/*
 * 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.
 */
1602
bool mem_cgroup_reclaimable(struct mem_cgroup *memcg, bool noswap)
1603 1604 1605 1606 1607 1608 1609
{
	int nid;

	/*
	 * quick check...making use of scan_node.
	 * We can skip unused nodes.
	 */
1610 1611
	if (!nodes_empty(memcg->scan_nodes)) {
		for (nid = first_node(memcg->scan_nodes);
1612
		     nid < MAX_NUMNODES;
1613
		     nid = next_node(nid, memcg->scan_nodes)) {
1614

1615
			if (test_mem_cgroup_node_reclaimable(memcg, nid, noswap))
1616 1617 1618 1619 1620 1621 1622
				return true;
		}
	}
	/*
	 * Check rest of nodes.
	 */
	for_each_node_state(nid, N_HIGH_MEMORY) {
1623
		if (node_isset(nid, memcg->scan_nodes))
1624
			continue;
1625
		if (test_mem_cgroup_node_reclaimable(memcg, nid, noswap))
1626 1627 1628 1629 1630
			return true;
	}
	return false;
}

1631
#else
1632
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
1633 1634 1635
{
	return 0;
}
1636

1637
bool mem_cgroup_reclaimable(struct mem_cgroup *memcg, bool noswap)
1638
{
1639
	return test_mem_cgroup_node_reclaimable(memcg, 0, noswap);
1640
}
1641 1642
#endif

1643 1644 1645 1646
static int mem_cgroup_soft_reclaim(struct mem_cgroup *root_memcg,
				   struct zone *zone,
				   gfp_t gfp_mask,
				   unsigned long *total_scanned)
1647
{
1648
	struct mem_cgroup *victim = NULL;
1649
	int total = 0;
K
KAMEZAWA Hiroyuki 已提交
1650
	int loop = 0;
1651
	unsigned long excess;
1652
	unsigned long nr_scanned;
1653 1654 1655 1656
	struct mem_cgroup_reclaim_cookie reclaim = {
		.zone = zone,
		.priority = 0,
	};
1657

1658
	excess = res_counter_soft_limit_excess(&root_memcg->res) >> PAGE_SHIFT;
K
KAMEZAWA Hiroyuki 已提交
1659

1660
	while (1) {
1661
		victim = mem_cgroup_iter(root_memcg, victim, &reclaim);
1662
		if (!victim) {
K
KAMEZAWA Hiroyuki 已提交
1663
			loop++;
1664 1665 1666 1667 1668 1669
			if (loop >= 2) {
				/*
				 * If we have not been able to reclaim
				 * anything, it might because there are
				 * no reclaimable pages under this hierarchy
				 */
1670
				if (!total)
1671 1672
					break;
				/*
L
Lucas De Marchi 已提交
1673
				 * We want to do more targeted reclaim.
1674 1675 1676 1677 1678
				 * 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) ||
1679
					(loop > MEM_CGROUP_MAX_RECLAIM_LOOPS))
1680 1681
					break;
			}
1682
			continue;
1683
		}
1684
		if (!mem_cgroup_reclaimable(victim, false))
1685
			continue;
1686 1687 1688 1689
		total += mem_cgroup_shrink_node_zone(victim, gfp_mask, false,
						     zone, &nr_scanned);
		*total_scanned += nr_scanned;
		if (!res_counter_soft_limit_excess(&root_memcg->res))
1690
			break;
1691
	}
1692
	mem_cgroup_iter_break(root_memcg, victim);
K
KAMEZAWA Hiroyuki 已提交
1693
	return total;
1694 1695
}

K
KAMEZAWA Hiroyuki 已提交
1696 1697 1698
/*
 * Check OOM-Killer is already running under our hierarchy.
 * If someone is running, return false.
1699
 * Has to be called with memcg_oom_lock
K
KAMEZAWA Hiroyuki 已提交
1700
 */
1701
static bool mem_cgroup_oom_lock(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
1702
{
1703
	struct mem_cgroup *iter, *failed = NULL;
1704

1705
	for_each_mem_cgroup_tree(iter, memcg) {
1706
		if (iter->oom_lock) {
1707 1708 1709 1710 1711
			/*
			 * this subtree of our hierarchy is already locked
			 * so we cannot give a lock.
			 */
			failed = iter;
1712 1713
			mem_cgroup_iter_break(memcg, iter);
			break;
1714 1715
		} else
			iter->oom_lock = true;
K
KAMEZAWA Hiroyuki 已提交
1716
	}
K
KAMEZAWA Hiroyuki 已提交
1717

1718
	if (!failed)
1719
		return true;
1720 1721 1722 1723 1724

	/*
	 * OK, we failed to lock the whole subtree so we have to clean up
	 * what we set up to the failing subtree
	 */
1725
	for_each_mem_cgroup_tree(iter, memcg) {
1726
		if (iter == failed) {
1727 1728
			mem_cgroup_iter_break(memcg, iter);
			break;
1729 1730 1731
		}
		iter->oom_lock = false;
	}
1732
	return false;
1733
}
1734

1735
/*
1736
 * Has to be called with memcg_oom_lock
1737
 */
1738
static int mem_cgroup_oom_unlock(struct mem_cgroup *memcg)
1739
{
K
KAMEZAWA Hiroyuki 已提交
1740 1741
	struct mem_cgroup *iter;

1742
	for_each_mem_cgroup_tree(iter, memcg)
1743 1744 1745 1746
		iter->oom_lock = false;
	return 0;
}

1747
static void mem_cgroup_mark_under_oom(struct mem_cgroup *memcg)
1748 1749 1750
{
	struct mem_cgroup *iter;

1751
	for_each_mem_cgroup_tree(iter, memcg)
1752 1753 1754
		atomic_inc(&iter->under_oom);
}

1755
static void mem_cgroup_unmark_under_oom(struct mem_cgroup *memcg)
1756 1757 1758
{
	struct mem_cgroup *iter;

K
KAMEZAWA Hiroyuki 已提交
1759 1760 1761 1762 1763
	/*
	 * 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.
	 */
1764
	for_each_mem_cgroup_tree(iter, memcg)
1765
		atomic_add_unless(&iter->under_oom, -1, 0);
1766 1767
}

1768
static DEFINE_SPINLOCK(memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
1769 1770
static DECLARE_WAIT_QUEUE_HEAD(memcg_oom_waitq);

K
KAMEZAWA Hiroyuki 已提交
1771
struct oom_wait_info {
1772
	struct mem_cgroup *memcg;
K
KAMEZAWA Hiroyuki 已提交
1773 1774 1775 1776 1777 1778
	wait_queue_t	wait;
};

static int memcg_oom_wake_function(wait_queue_t *wait,
	unsigned mode, int sync, void *arg)
{
1779 1780
	struct mem_cgroup *wake_memcg = (struct mem_cgroup *)arg;
	struct mem_cgroup *oom_wait_memcg;
K
KAMEZAWA Hiroyuki 已提交
1781 1782 1783
	struct oom_wait_info *oom_wait_info;

	oom_wait_info = container_of(wait, struct oom_wait_info, wait);
1784
	oom_wait_memcg = oom_wait_info->memcg;
K
KAMEZAWA Hiroyuki 已提交
1785 1786

	/*
1787
	 * Both of oom_wait_info->memcg and wake_memcg are stable under us.
K
KAMEZAWA Hiroyuki 已提交
1788 1789
	 * Then we can use css_is_ancestor without taking care of RCU.
	 */
1790 1791
	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
1792 1793 1794 1795
		return 0;
	return autoremove_wake_function(wait, mode, sync, arg);
}

1796
static void memcg_wakeup_oom(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
1797
{
1798 1799
	/* for filtering, pass "memcg" as argument. */
	__wake_up(&memcg_oom_waitq, TASK_NORMAL, 0, memcg);
K
KAMEZAWA Hiroyuki 已提交
1800 1801
}

1802
static void memcg_oom_recover(struct mem_cgroup *memcg)
1803
{
1804 1805
	if (memcg && atomic_read(&memcg->under_oom))
		memcg_wakeup_oom(memcg);
1806 1807
}

K
KAMEZAWA Hiroyuki 已提交
1808 1809 1810
/*
 * try to call OOM killer. returns false if we should exit memory-reclaim loop.
 */
1811
bool mem_cgroup_handle_oom(struct mem_cgroup *memcg, gfp_t mask, int order)
1812
{
K
KAMEZAWA Hiroyuki 已提交
1813
	struct oom_wait_info owait;
1814
	bool locked, need_to_kill;
K
KAMEZAWA Hiroyuki 已提交
1815

1816
	owait.memcg = memcg;
K
KAMEZAWA Hiroyuki 已提交
1817 1818 1819 1820
	owait.wait.flags = 0;
	owait.wait.func = memcg_oom_wake_function;
	owait.wait.private = current;
	INIT_LIST_HEAD(&owait.wait.task_list);
1821
	need_to_kill = true;
1822
	mem_cgroup_mark_under_oom(memcg);
1823

1824
	/* At first, try to OOM lock hierarchy under memcg.*/
1825
	spin_lock(&memcg_oom_lock);
1826
	locked = mem_cgroup_oom_lock(memcg);
K
KAMEZAWA Hiroyuki 已提交
1827 1828 1829 1830 1831
	/*
	 * 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.
	 */
1832
	prepare_to_wait(&memcg_oom_waitq, &owait.wait, TASK_KILLABLE);
1833
	if (!locked || memcg->oom_kill_disable)
1834 1835
		need_to_kill = false;
	if (locked)
1836
		mem_cgroup_oom_notify(memcg);
1837
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
1838

1839 1840
	if (need_to_kill) {
		finish_wait(&memcg_oom_waitq, &owait.wait);
1841
		mem_cgroup_out_of_memory(memcg, mask, order);
1842
	} else {
K
KAMEZAWA Hiroyuki 已提交
1843
		schedule();
K
KAMEZAWA Hiroyuki 已提交
1844
		finish_wait(&memcg_oom_waitq, &owait.wait);
K
KAMEZAWA Hiroyuki 已提交
1845
	}
1846
	spin_lock(&memcg_oom_lock);
1847
	if (locked)
1848 1849
		mem_cgroup_oom_unlock(memcg);
	memcg_wakeup_oom(memcg);
1850
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
1851

1852
	mem_cgroup_unmark_under_oom(memcg);
1853

K
KAMEZAWA Hiroyuki 已提交
1854 1855 1856
	if (test_thread_flag(TIF_MEMDIE) || fatal_signal_pending(current))
		return false;
	/* Give chance to dying process */
1857
	schedule_timeout_uninterruptible(1);
K
KAMEZAWA Hiroyuki 已提交
1858
	return true;
1859 1860
}

1861 1862 1863
/*
 * Currently used to update mapped file statistics, but the routine can be
 * generalized to update other statistics as well.
1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882
 *
 * 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.
1883
 */
1884

1885 1886
void mem_cgroup_update_page_stat(struct page *page,
				 enum mem_cgroup_page_stat_item idx, int val)
1887
{
1888
	struct mem_cgroup *memcg;
1889 1890
	struct page_cgroup *pc = lookup_page_cgroup(page);
	bool need_unlock = false;
1891
	unsigned long uninitialized_var(flags);
1892

1893
	if (mem_cgroup_disabled())
1894 1895
		return;

1896
	rcu_read_lock();
1897 1898
	memcg = pc->mem_cgroup;
	if (unlikely(!memcg || !PageCgroupUsed(pc)))
1899 1900
		goto out;
	/* pc->mem_cgroup is unstable ? */
1901
	if (unlikely(mem_cgroup_stealed(memcg))) {
1902
		/* take a lock against to access pc->mem_cgroup */
1903
		move_lock_page_cgroup(pc, &flags);
1904
		need_unlock = true;
1905 1906
		memcg = pc->mem_cgroup;
		if (!memcg || !PageCgroupUsed(pc))
1907 1908
			goto out;
	}
1909 1910

	switch (idx) {
1911
	case MEMCG_NR_FILE_MAPPED:
1912 1913 1914
		if (val > 0)
			SetPageCgroupFileMapped(pc);
		else if (!page_mapped(page))
1915
			ClearPageCgroupFileMapped(pc);
1916
		idx = MEM_CGROUP_STAT_FILE_MAPPED;
1917 1918 1919
		break;
	default:
		BUG();
1920
	}
1921

1922
	this_cpu_add(memcg->stat->count[idx], val);
1923

1924 1925
out:
	if (unlikely(need_unlock))
1926
		move_unlock_page_cgroup(pc, &flags);
1927
	rcu_read_unlock();
1928
}
1929
EXPORT_SYMBOL(mem_cgroup_update_page_stat);
1930

1931 1932 1933 1934
/*
 * size of first charge trial. "32" comes from vmscan.c's magic value.
 * TODO: maybe necessary to use big numbers in big irons.
 */
1935
#define CHARGE_BATCH	32U
1936 1937
struct memcg_stock_pcp {
	struct mem_cgroup *cached; /* this never be root cgroup */
1938
	unsigned int nr_pages;
1939
	struct work_struct work;
1940 1941
	unsigned long flags;
#define FLUSHING_CACHED_CHARGE	(0)
1942 1943
};
static DEFINE_PER_CPU(struct memcg_stock_pcp, memcg_stock);
1944
static DEFINE_MUTEX(percpu_charge_mutex);
1945 1946

/*
1947
 * Try to consume stocked charge on this cpu. If success, one page is consumed
1948 1949 1950 1951
 * 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.
 */
1952
static bool consume_stock(struct mem_cgroup *memcg)
1953 1954 1955 1956 1957
{
	struct memcg_stock_pcp *stock;
	bool ret = true;

	stock = &get_cpu_var(memcg_stock);
1958
	if (memcg == stock->cached && stock->nr_pages)
1959
		stock->nr_pages--;
1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972
	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;

1973 1974 1975 1976
	if (stock->nr_pages) {
		unsigned long bytes = stock->nr_pages * PAGE_SIZE;

		res_counter_uncharge(&old->res, bytes);
1977
		if (do_swap_account)
1978 1979
			res_counter_uncharge(&old->memsw, bytes);
		stock->nr_pages = 0;
1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991
	}
	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);
1992
	clear_bit(FLUSHING_CACHED_CHARGE, &stock->flags);
1993 1994 1995 1996
}

/*
 * Cache charges(val) which is from res_counter, to local per_cpu area.
1997
 * This will be consumed by consume_stock() function, later.
1998
 */
1999
static void refill_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
2000 2001 2002
{
	struct memcg_stock_pcp *stock = &get_cpu_var(memcg_stock);

2003
	if (stock->cached != memcg) { /* reset if necessary */
2004
		drain_stock(stock);
2005
		stock->cached = memcg;
2006
	}
2007
	stock->nr_pages += nr_pages;
2008 2009 2010 2011
	put_cpu_var(memcg_stock);
}

/*
2012
 * Drains all per-CPU charge caches for given root_memcg resp. subtree
2013 2014
 * of the hierarchy under it. sync flag says whether we should block
 * until the work is done.
2015
 */
2016
static void drain_all_stock(struct mem_cgroup *root_memcg, bool sync)
2017
{
2018
	int cpu, curcpu;
2019

2020 2021
	/* Notify other cpus that system-wide "drain" is running */
	get_online_cpus();
2022
	curcpu = get_cpu();
2023 2024
	for_each_online_cpu(cpu) {
		struct memcg_stock_pcp *stock = &per_cpu(memcg_stock, cpu);
2025
		struct mem_cgroup *memcg;
2026

2027 2028
		memcg = stock->cached;
		if (!memcg || !stock->nr_pages)
2029
			continue;
2030
		if (!mem_cgroup_same_or_subtree(root_memcg, memcg))
2031
			continue;
2032 2033 2034 2035 2036 2037
		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);
		}
2038
	}
2039
	put_cpu();
2040 2041 2042 2043 2044 2045

	if (!sync)
		goto out;

	for_each_online_cpu(cpu) {
		struct memcg_stock_pcp *stock = &per_cpu(memcg_stock, cpu);
2046
		if (test_bit(FLUSHING_CACHED_CHARGE, &stock->flags))
2047 2048 2049
			flush_work(&stock->work);
	}
out:
2050
 	put_online_cpus();
2051 2052 2053 2054 2055 2056 2057 2058
}

/*
 * 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.
 */
2059
static void drain_all_stock_async(struct mem_cgroup *root_memcg)
2060
{
2061 2062 2063 2064 2065
	/*
	 * If someone calls draining, avoid adding more kworker runs.
	 */
	if (!mutex_trylock(&percpu_charge_mutex))
		return;
2066
	drain_all_stock(root_memcg, false);
2067
	mutex_unlock(&percpu_charge_mutex);
2068 2069 2070
}

/* This is a synchronous drain interface. */
2071
static void drain_all_stock_sync(struct mem_cgroup *root_memcg)
2072 2073
{
	/* called when force_empty is called */
2074
	mutex_lock(&percpu_charge_mutex);
2075
	drain_all_stock(root_memcg, true);
2076
	mutex_unlock(&percpu_charge_mutex);
2077 2078
}

2079 2080 2081 2082
/*
 * This function drains percpu counter value from DEAD cpu and
 * move it to local cpu. Note that this function can be preempted.
 */
2083
static void mem_cgroup_drain_pcp_counter(struct mem_cgroup *memcg, int cpu)
2084 2085 2086
{
	int i;

2087
	spin_lock(&memcg->pcp_counter_lock);
2088
	for (i = 0; i < MEM_CGROUP_STAT_DATA; i++) {
2089
		long x = per_cpu(memcg->stat->count[i], cpu);
2090

2091 2092
		per_cpu(memcg->stat->count[i], cpu) = 0;
		memcg->nocpu_base.count[i] += x;
2093
	}
2094
	for (i = 0; i < MEM_CGROUP_EVENTS_NSTATS; i++) {
2095
		unsigned long x = per_cpu(memcg->stat->events[i], cpu);
2096

2097 2098
		per_cpu(memcg->stat->events[i], cpu) = 0;
		memcg->nocpu_base.events[i] += x;
2099
	}
2100
	/* need to clear ON_MOVE value, works as a kind of lock. */
2101 2102
	per_cpu(memcg->stat->count[MEM_CGROUP_ON_MOVE], cpu) = 0;
	spin_unlock(&memcg->pcp_counter_lock);
2103 2104
}

2105
static void synchronize_mem_cgroup_on_move(struct mem_cgroup *memcg, int cpu)
2106 2107 2108
{
	int idx = MEM_CGROUP_ON_MOVE;

2109 2110 2111
	spin_lock(&memcg->pcp_counter_lock);
	per_cpu(memcg->stat->count[idx], cpu) = memcg->nocpu_base.count[idx];
	spin_unlock(&memcg->pcp_counter_lock);
2112 2113 2114
}

static int __cpuinit memcg_cpu_hotplug_callback(struct notifier_block *nb,
2115 2116 2117 2118 2119
					unsigned long action,
					void *hcpu)
{
	int cpu = (unsigned long)hcpu;
	struct memcg_stock_pcp *stock;
2120
	struct mem_cgroup *iter;
2121

2122
	if ((action == CPU_ONLINE)) {
2123
		for_each_mem_cgroup(iter)
2124 2125 2126 2127
			synchronize_mem_cgroup_on_move(iter, cpu);
		return NOTIFY_OK;
	}

2128
	if ((action != CPU_DEAD) || action != CPU_DEAD_FROZEN)
2129
		return NOTIFY_OK;
2130

2131
	for_each_mem_cgroup(iter)
2132 2133
		mem_cgroup_drain_pcp_counter(iter, cpu);

2134 2135 2136 2137 2138
	stock = &per_cpu(memcg_stock, cpu);
	drain_stock(stock);
	return NOTIFY_OK;
}

2139 2140 2141 2142 2143 2144 2145 2146 2147 2148

/* 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 */
};

2149
static int mem_cgroup_do_charge(struct mem_cgroup *memcg, gfp_t gfp_mask,
2150
				unsigned int nr_pages, bool oom_check)
2151
{
2152
	unsigned long csize = nr_pages * PAGE_SIZE;
2153 2154 2155 2156 2157
	struct mem_cgroup *mem_over_limit;
	struct res_counter *fail_res;
	unsigned long flags = 0;
	int ret;

2158
	ret = res_counter_charge(&memcg->res, csize, &fail_res);
2159 2160 2161 2162

	if (likely(!ret)) {
		if (!do_swap_account)
			return CHARGE_OK;
2163
		ret = res_counter_charge(&memcg->memsw, csize, &fail_res);
2164 2165 2166
		if (likely(!ret))
			return CHARGE_OK;

2167
		res_counter_uncharge(&memcg->res, csize);
2168 2169 2170 2171
		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);
2172
	/*
2173 2174
	 * nr_pages can be either a huge page (HPAGE_PMD_NR), a batch
	 * of regular pages (CHARGE_BATCH), or a single regular page (1).
2175 2176 2177 2178
	 *
	 * Never reclaim on behalf of optional batching, retry with a
	 * single page instead.
	 */
2179
	if (nr_pages == CHARGE_BATCH)
2180 2181 2182 2183 2184
		return CHARGE_RETRY;

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

2185
	ret = mem_cgroup_reclaim(mem_over_limit, gfp_mask, flags);
2186
	if (mem_cgroup_margin(mem_over_limit) >= nr_pages)
2187
		return CHARGE_RETRY;
2188
	/*
2189 2190 2191 2192 2193 2194 2195
	 * 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.
2196
	 */
2197
	if (nr_pages == 1 && ret)
2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210
		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 */
2211
	if (!mem_cgroup_handle_oom(mem_over_limit, gfp_mask, get_order(csize)))
2212 2213 2214 2215 2216
		return CHARGE_OOM_DIE;

	return CHARGE_RETRY;
}

2217
/*
2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236
 * __mem_cgroup_try_charge() does
 * 1. detect memcg to be charged against from passed *mm and *ptr,
 * 2. update res_counter
 * 3. call memory reclaim if necessary.
 *
 * In some special case, if the task is fatal, fatal_signal_pending() or
 * has TIF_MEMDIE, this function returns -EINTR while writing root_mem_cgroup
 * to *ptr. There are two reasons for this. 1: fatal threads should quit as soon
 * as possible without any hazards. 2: all pages should have a valid
 * pc->mem_cgroup. If mm is NULL and the caller doesn't pass a valid memcg
 * pointer, that is treated as a charge to root_mem_cgroup.
 *
 * So __mem_cgroup_try_charge() will return
 *  0       ...  on success, filling *ptr with a valid memcg pointer.
 *  -ENOMEM ...  charge failure because of resource limits.
 *  -EINTR  ...  if thread is fatal. *ptr is filled with root_mem_cgroup.
 *
 * Unlike the exported interface, an "oom" parameter is added. if oom==true,
 * the oom-killer can be invoked.
2237
 */
2238
static int __mem_cgroup_try_charge(struct mm_struct *mm,
A
Andrea Arcangeli 已提交
2239
				   gfp_t gfp_mask,
2240
				   unsigned int nr_pages,
2241
				   struct mem_cgroup **ptr,
2242
				   bool oom)
2243
{
2244
	unsigned int batch = max(CHARGE_BATCH, nr_pages);
2245
	int nr_oom_retries = MEM_CGROUP_RECLAIM_RETRIES;
2246
	struct mem_cgroup *memcg = NULL;
2247
	int ret;
2248

K
KAMEZAWA Hiroyuki 已提交
2249 2250 2251 2252 2253 2254 2255 2256
	/*
	 * 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;
2257

2258
	/*
2259 2260
	 * We always charge the cgroup the mm_struct belongs to.
	 * The mm_struct's mem_cgroup changes on task migration if the
2261 2262 2263
	 * thread group leader migrates. It's possible that mm is not
	 * set, if so charge the init_mm (happens for pagecache usage).
	 */
2264
	if (!*ptr && !mm)
2265
		*ptr = root_mem_cgroup;
K
KAMEZAWA Hiroyuki 已提交
2266
again:
2267 2268 2269 2270
	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
2271
			goto done;
2272
		if (nr_pages == 1 && consume_stock(memcg))
K
KAMEZAWA Hiroyuki 已提交
2273
			goto done;
2274
		css_get(&memcg->css);
2275
	} else {
K
KAMEZAWA Hiroyuki 已提交
2276
		struct task_struct *p;
2277

K
KAMEZAWA Hiroyuki 已提交
2278 2279 2280
		rcu_read_lock();
		p = rcu_dereference(mm->owner);
		/*
2281
		 * Because we don't have task_lock(), "p" can exit.
2282
		 * In that case, "memcg" can point to root or p can be NULL with
2283 2284 2285 2286 2287 2288
		 * 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
2289
		 */
2290
		memcg = mem_cgroup_from_task(p);
2291 2292 2293
		if (!memcg)
			memcg = root_mem_cgroup;
		if (mem_cgroup_is_root(memcg)) {
K
KAMEZAWA Hiroyuki 已提交
2294 2295 2296
			rcu_read_unlock();
			goto done;
		}
2297
		if (nr_pages == 1 && consume_stock(memcg)) {
K
KAMEZAWA Hiroyuki 已提交
2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309
			/*
			 * 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 */
2310
		if (!css_tryget(&memcg->css)) {
K
KAMEZAWA Hiroyuki 已提交
2311 2312 2313 2314 2315
			rcu_read_unlock();
			goto again;
		}
		rcu_read_unlock();
	}
2316

2317 2318
	do {
		bool oom_check;
2319

2320
		/* If killed, bypass charge */
K
KAMEZAWA Hiroyuki 已提交
2321
		if (fatal_signal_pending(current)) {
2322
			css_put(&memcg->css);
2323
			goto bypass;
K
KAMEZAWA Hiroyuki 已提交
2324
		}
2325

2326 2327 2328 2329
		oom_check = false;
		if (oom && !nr_oom_retries) {
			oom_check = true;
			nr_oom_retries = MEM_CGROUP_RECLAIM_RETRIES;
2330
		}
2331

2332
		ret = mem_cgroup_do_charge(memcg, gfp_mask, batch, oom_check);
2333 2334 2335 2336
		switch (ret) {
		case CHARGE_OK:
			break;
		case CHARGE_RETRY: /* not in OOM situation but retry */
2337
			batch = nr_pages;
2338 2339
			css_put(&memcg->css);
			memcg = NULL;
K
KAMEZAWA Hiroyuki 已提交
2340
			goto again;
2341
		case CHARGE_WOULDBLOCK: /* !__GFP_WAIT */
2342
			css_put(&memcg->css);
2343 2344
			goto nomem;
		case CHARGE_NOMEM: /* OOM routine works */
K
KAMEZAWA Hiroyuki 已提交
2345
			if (!oom) {
2346
				css_put(&memcg->css);
K
KAMEZAWA Hiroyuki 已提交
2347
				goto nomem;
K
KAMEZAWA Hiroyuki 已提交
2348
			}
2349 2350 2351 2352
			/* If oom, we never return -ENOMEM */
			nr_oom_retries--;
			break;
		case CHARGE_OOM_DIE: /* Killed by OOM Killer */
2353
			css_put(&memcg->css);
K
KAMEZAWA Hiroyuki 已提交
2354
			goto bypass;
2355
		}
2356 2357
	} while (ret != CHARGE_OK);

2358
	if (batch > nr_pages)
2359 2360
		refill_stock(memcg, batch - nr_pages);
	css_put(&memcg->css);
2361
done:
2362
	*ptr = memcg;
2363 2364
	return 0;
nomem:
2365
	*ptr = NULL;
2366
	return -ENOMEM;
K
KAMEZAWA Hiroyuki 已提交
2367
bypass:
2368 2369
	*ptr = root_mem_cgroup;
	return -EINTR;
2370
}
2371

2372 2373 2374 2375 2376
/*
 * 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().
 */
2377
static void __mem_cgroup_cancel_charge(struct mem_cgroup *memcg,
2378
				       unsigned int nr_pages)
2379
{
2380
	if (!mem_cgroup_is_root(memcg)) {
2381 2382
		unsigned long bytes = nr_pages * PAGE_SIZE;

2383
		res_counter_uncharge(&memcg->res, bytes);
2384
		if (do_swap_account)
2385
			res_counter_uncharge(&memcg->memsw, bytes);
2386
	}
2387 2388
}

2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407
/*
 * 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);
}

2408
struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page)
2409
{
2410
	struct mem_cgroup *memcg = NULL;
2411
	struct page_cgroup *pc;
2412
	unsigned short id;
2413 2414
	swp_entry_t ent;

2415 2416 2417
	VM_BUG_ON(!PageLocked(page));

	pc = lookup_page_cgroup(page);
2418
	lock_page_cgroup(pc);
2419
	if (PageCgroupUsed(pc)) {
2420 2421 2422
		memcg = pc->mem_cgroup;
		if (memcg && !css_tryget(&memcg->css))
			memcg = NULL;
2423
	} else if (PageSwapCache(page)) {
2424
		ent.val = page_private(page);
2425
		id = lookup_swap_cgroup_id(ent);
2426
		rcu_read_lock();
2427 2428 2429
		memcg = mem_cgroup_lookup(id);
		if (memcg && !css_tryget(&memcg->css))
			memcg = NULL;
2430
		rcu_read_unlock();
2431
	}
2432
	unlock_page_cgroup(pc);
2433
	return memcg;
2434 2435
}

2436
static void __mem_cgroup_commit_charge(struct mem_cgroup *memcg,
2437
				       struct page *page,
2438
				       unsigned int nr_pages,
2439
				       struct page_cgroup *pc,
2440 2441
				       enum charge_type ctype,
				       bool lrucare)
2442
{
2443 2444 2445
	struct zone *uninitialized_var(zone);
	bool was_on_lru = false;

2446 2447 2448
	lock_page_cgroup(pc);
	if (unlikely(PageCgroupUsed(pc))) {
		unlock_page_cgroup(pc);
2449
		__mem_cgroup_cancel_charge(memcg, nr_pages);
2450 2451 2452 2453 2454 2455
		return;
	}
	/*
	 * we don't need page_cgroup_lock about tail pages, becase they are not
	 * accessed by any other context at this point.
	 */
2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470

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

2471
	pc->mem_cgroup = memcg;
2472 2473 2474 2475 2476 2477 2478
	/*
	 * 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
2479
	smp_wmb();
2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492
	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;
	}
2493

2494 2495 2496 2497 2498 2499 2500 2501 2502
	if (lrucare) {
		if (was_on_lru) {
			VM_BUG_ON(PageLRU(page));
			SetPageLRU(page);
			add_page_to_lru_list(zone, page, page_lru(page));
		}
		spin_unlock_irq(&zone->lru_lock);
	}

2503
	mem_cgroup_charge_statistics(memcg, PageCgroupCache(pc), nr_pages);
2504
	unlock_page_cgroup(pc);
2505

2506 2507 2508 2509 2510
	/*
	 * "charge_statistics" updated event counter. Then, check it.
	 * Insert ancestor (and ancestor's ancestors), to softlimit RB-tree.
	 * if they exceeds softlimit.
	 */
2511
	memcg_check_events(memcg, page);
2512
}
2513

2514 2515 2516
#ifdef CONFIG_TRANSPARENT_HUGEPAGE

#define PCGF_NOCOPY_AT_SPLIT ((1 << PCG_LOCK) | (1 << PCG_MOVE_LOCK) |\
2517
			(1 << PCG_MIGRATION))
2518 2519
/*
 * Because tail pages are not marked as "used", set it. We're under
2520 2521 2522
 * zone->lru_lock, 'splitting on pmd' and compound_lock.
 * charge/uncharge will be never happen and move_account() is done under
 * compound_lock(), so we don't have to take care of races.
2523
 */
2524
void mem_cgroup_split_huge_fixup(struct page *head)
2525 2526
{
	struct page_cgroup *head_pc = lookup_page_cgroup(head);
2527 2528
	struct page_cgroup *pc;
	int i;
2529

2530 2531
	if (mem_cgroup_disabled())
		return;
2532 2533 2534 2535 2536 2537
	for (i = 1; i < HPAGE_PMD_NR; i++) {
		pc = head_pc + i;
		pc->mem_cgroup = head_pc->mem_cgroup;
		smp_wmb();/* see __commit_charge() */
		pc->flags = head_pc->flags & ~PCGF_NOCOPY_AT_SPLIT;
	}
2538
}
2539
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
2540

2541
/**
2542
 * mem_cgroup_move_account - move account of the page
2543
 * @page: the page
2544
 * @nr_pages: number of regular pages (>1 for huge pages)
2545 2546 2547
 * @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.
2548
 * @uncharge: whether we should call uncharge and css_put against @from.
2549 2550
 *
 * The caller must confirm following.
K
KAMEZAWA Hiroyuki 已提交
2551
 * - page is not on LRU (isolate_page() is useful.)
2552
 * - compound_lock is held when nr_pages > 1
2553
 *
2554
 * This function doesn't do "charge" nor css_get to new cgroup. It should be
L
Lucas De Marchi 已提交
2555
 * done by a caller(__mem_cgroup_try_charge would be useful). If @uncharge is
2556 2557
 * true, this function does "uncharge" from old cgroup, but it doesn't if
 * @uncharge is false, so a caller should do "uncharge".
2558
 */
2559 2560 2561 2562 2563 2564
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)
2565
{
2566 2567
	unsigned long flags;
	int ret;
2568

2569
	VM_BUG_ON(from == to);
2570
	VM_BUG_ON(PageLRU(page));
2571 2572 2573 2574 2575 2576 2577
	/*
	 * 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;
2578
	if (nr_pages > 1 && !PageTransHuge(page))
2579 2580 2581 2582 2583 2584 2585 2586 2587
		goto out;

	lock_page_cgroup(pc);

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

	move_lock_page_cgroup(pc, &flags);
2588

2589
	if (PageCgroupFileMapped(pc)) {
2590 2591 2592 2593 2594
		/* 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();
2595
	}
2596
	mem_cgroup_charge_statistics(from, PageCgroupCache(pc), -nr_pages);
2597 2598
	if (uncharge)
		/* This is not "cancel", but cancel_charge does all we need. */
2599
		__mem_cgroup_cancel_charge(from, nr_pages);
2600

2601
	/* caller should have done css_get */
K
KAMEZAWA Hiroyuki 已提交
2602
	pc->mem_cgroup = to;
2603
	mem_cgroup_charge_statistics(to, PageCgroupCache(pc), nr_pages);
2604 2605 2606
	/*
	 * We charges against "to" which may not have any tasks. Then, "to"
	 * can be under rmdir(). But in current implementation, caller of
2607
	 * this function is just force_empty() and move charge, so it's
L
Lucas De Marchi 已提交
2608
	 * guaranteed that "to" is never removed. So, we don't check rmdir
2609
	 * status here.
2610
	 */
2611 2612 2613
	move_unlock_page_cgroup(pc, &flags);
	ret = 0;
unlock:
2614
	unlock_page_cgroup(pc);
2615 2616 2617
	/*
	 * check events
	 */
2618 2619
	memcg_check_events(to, page);
	memcg_check_events(from, page);
2620
out:
2621 2622 2623 2624 2625 2626 2627
	return ret;
}

/*
 * move charges to its parent.
 */

2628 2629
static int mem_cgroup_move_parent(struct page *page,
				  struct page_cgroup *pc,
2630 2631 2632 2633 2634 2635
				  struct mem_cgroup *child,
				  gfp_t gfp_mask)
{
	struct cgroup *cg = child->css.cgroup;
	struct cgroup *pcg = cg->parent;
	struct mem_cgroup *parent;
2636
	unsigned int nr_pages;
2637
	unsigned long uninitialized_var(flags);
2638 2639 2640 2641 2642 2643
	int ret;

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

2644 2645 2646 2647 2648
	ret = -EBUSY;
	if (!get_page_unless_zero(page))
		goto out;
	if (isolate_lru_page(page))
		goto put;
2649

2650
	nr_pages = hpage_nr_pages(page);
K
KAMEZAWA Hiroyuki 已提交
2651

2652
	parent = mem_cgroup_from_cont(pcg);
2653
	ret = __mem_cgroup_try_charge(NULL, gfp_mask, nr_pages, &parent, false);
2654
	if (ret)
2655
		goto put_back;
2656

2657
	if (nr_pages > 1)
2658 2659
		flags = compound_lock_irqsave(page);

2660
	ret = mem_cgroup_move_account(page, nr_pages, pc, child, parent, true);
2661
	if (ret)
2662
		__mem_cgroup_cancel_charge(parent, nr_pages);
2663

2664
	if (nr_pages > 1)
2665
		compound_unlock_irqrestore(page, flags);
2666
put_back:
K
KAMEZAWA Hiroyuki 已提交
2667
	putback_lru_page(page);
2668
put:
2669
	put_page(page);
2670
out:
2671 2672 2673
	return ret;
}

2674 2675 2676 2677 2678 2679 2680
/*
 * 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,
2681
				gfp_t gfp_mask, enum charge_type ctype)
2682
{
2683
	struct mem_cgroup *memcg = NULL;
2684
	unsigned int nr_pages = 1;
2685
	struct page_cgroup *pc;
2686
	bool oom = true;
2687
	int ret;
A
Andrea Arcangeli 已提交
2688

A
Andrea Arcangeli 已提交
2689
	if (PageTransHuge(page)) {
2690
		nr_pages <<= compound_order(page);
A
Andrea Arcangeli 已提交
2691
		VM_BUG_ON(!PageTransHuge(page));
2692 2693 2694 2695 2696
		/*
		 * 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
2697
	}
2698 2699

	pc = lookup_page_cgroup(page);
2700
	ret = __mem_cgroup_try_charge(mm, gfp_mask, nr_pages, &memcg, oom);
2701
	if (ret == -ENOMEM)
2702
		return ret;
2703
	__mem_cgroup_commit_charge(memcg, page, nr_pages, pc, ctype, false);
2704 2705 2706
	return 0;
}

2707 2708
int mem_cgroup_newpage_charge(struct page *page,
			      struct mm_struct *mm, gfp_t gfp_mask)
2709
{
2710
	if (mem_cgroup_disabled())
2711
		return 0;
2712 2713 2714
	VM_BUG_ON(page_mapped(page));
	VM_BUG_ON(page->mapping && !PageAnon(page));
	VM_BUG_ON(!mm);
2715
	return mem_cgroup_charge_common(page, mm, gfp_mask,
2716
					MEM_CGROUP_CHARGE_TYPE_MAPPED);
2717 2718
}

D
Daisuke Nishimura 已提交
2719 2720 2721 2722
static void
__mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr,
					enum charge_type ctype);

2723 2724
int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm,
				gfp_t gfp_mask)
2725
{
2726
	struct mem_cgroup *memcg = NULL;
2727
	enum charge_type type = MEM_CGROUP_CHARGE_TYPE_CACHE;
2728 2729
	int ret;

2730
	if (mem_cgroup_disabled())
2731
		return 0;
2732 2733
	if (PageCompound(page))
		return 0;
2734

2735
	if (unlikely(!mm))
2736
		mm = &init_mm;
2737 2738
	if (!page_is_file_cache(page))
		type = MEM_CGROUP_CHARGE_TYPE_SHMEM;
2739

2740
	if (!PageSwapCache(page))
2741
		ret = mem_cgroup_charge_common(page, mm, gfp_mask, type);
2742
	else { /* page is swapcache/shmem */
2743
		ret = mem_cgroup_try_charge_swapin(mm, page, gfp_mask, &memcg);
D
Daisuke Nishimura 已提交
2744
		if (!ret)
2745 2746
			__mem_cgroup_commit_charge_swapin(page, memcg, type);
	}
2747
	return ret;
2748 2749
}

2750 2751 2752
/*
 * While swap-in, try_charge -> commit or cancel, the page is locked.
 * And when try_charge() successfully returns, one refcnt to memcg without
2753
 * struct page_cgroup is acquired. This refcnt will be consumed by
2754 2755
 * "commit()" or removed by "cancel()"
 */
2756 2757
int mem_cgroup_try_charge_swapin(struct mm_struct *mm,
				 struct page *page,
2758
				 gfp_t mask, struct mem_cgroup **memcgp)
2759
{
2760
	struct mem_cgroup *memcg;
2761
	int ret;
2762

2763
	*memcgp = NULL;
2764

2765
	if (mem_cgroup_disabled())
2766 2767 2768 2769 2770 2771
		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
2772 2773 2774
	 * 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.
2775 2776
	 */
	if (!PageSwapCache(page))
H
Hugh Dickins 已提交
2777
		goto charge_cur_mm;
2778 2779
	memcg = try_get_mem_cgroup_from_page(page);
	if (!memcg)
2780
		goto charge_cur_mm;
2781 2782
	*memcgp = memcg;
	ret = __mem_cgroup_try_charge(NULL, mask, 1, memcgp, true);
2783
	css_put(&memcg->css);
2784 2785
	if (ret == -EINTR)
		ret = 0;
2786
	return ret;
2787 2788 2789
charge_cur_mm:
	if (unlikely(!mm))
		mm = &init_mm;
2790 2791 2792 2793
	ret = __mem_cgroup_try_charge(mm, mask, 1, memcgp, true);
	if (ret == -EINTR)
		ret = 0;
	return ret;
2794 2795
}

D
Daisuke Nishimura 已提交
2796
static void
2797
__mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *memcg,
D
Daisuke Nishimura 已提交
2798
					enum charge_type ctype)
2799
{
2800 2801
	struct page_cgroup *pc;

2802
	if (mem_cgroup_disabled())
2803
		return;
2804
	if (!memcg)
2805
		return;
2806
	cgroup_exclude_rmdir(&memcg->css);
2807

2808 2809
	pc = lookup_page_cgroup(page);
	__mem_cgroup_commit_charge(memcg, page, 1, pc, ctype, true);
2810 2811 2812
	/*
	 * Now swap is on-memory. This means this page may be
	 * counted both as mem and swap....double count.
2813 2814 2815
	 * 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.
2816
	 */
2817
	if (do_swap_account && PageSwapCache(page)) {
2818
		swp_entry_t ent = {.val = page_private(page)};
2819
		struct mem_cgroup *swap_memcg;
2820 2821 2822 2823
		unsigned short id;

		id = swap_cgroup_record(ent, 0);
		rcu_read_lock();
2824 2825
		swap_memcg = mem_cgroup_lookup(id);
		if (swap_memcg) {
2826 2827 2828 2829
			/*
			 * This recorded memcg can be obsolete one. So, avoid
			 * calling css_tryget
			 */
2830 2831 2832 2833 2834
			if (!mem_cgroup_is_root(swap_memcg))
				res_counter_uncharge(&swap_memcg->memsw,
						     PAGE_SIZE);
			mem_cgroup_swap_statistics(swap_memcg, false);
			mem_cgroup_put(swap_memcg);
2835
		}
2836
		rcu_read_unlock();
2837
	}
2838 2839 2840 2841 2842
	/*
	 * 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.
	 */
2843
	cgroup_release_and_wakeup_rmdir(&memcg->css);
2844 2845
}

2846 2847
void mem_cgroup_commit_charge_swapin(struct page *page,
				     struct mem_cgroup *memcg)
D
Daisuke Nishimura 已提交
2848
{
2849 2850
	__mem_cgroup_commit_charge_swapin(page, memcg,
					  MEM_CGROUP_CHARGE_TYPE_MAPPED);
D
Daisuke Nishimura 已提交
2851 2852
}

2853
void mem_cgroup_cancel_charge_swapin(struct mem_cgroup *memcg)
2854
{
2855
	if (mem_cgroup_disabled())
2856
		return;
2857
	if (!memcg)
2858
		return;
2859
	__mem_cgroup_cancel_charge(memcg, 1);
2860 2861
}

2862
static void mem_cgroup_do_uncharge(struct mem_cgroup *memcg,
2863 2864
				   unsigned int nr_pages,
				   const enum charge_type ctype)
2865 2866 2867
{
	struct memcg_batch_info *batch = NULL;
	bool uncharge_memsw = true;
2868

2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879
	/* 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)
2880
		batch->memcg = memcg;
2881 2882
	/*
	 * do_batch > 0 when unmapping pages or inode invalidate/truncate.
L
Lucas De Marchi 已提交
2883
	 * In those cases, all pages freed continuously can be expected to be in
2884 2885 2886 2887 2888 2889 2890 2891
	 * 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;

2892
	if (nr_pages > 1)
A
Andrea Arcangeli 已提交
2893 2894
		goto direct_uncharge;

2895 2896 2897 2898 2899
	/*
	 * 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.
	 */
2900
	if (batch->memcg != memcg)
2901 2902
		goto direct_uncharge;
	/* remember freed charge and uncharge it later */
2903
	batch->nr_pages++;
2904
	if (uncharge_memsw)
2905
		batch->memsw_nr_pages++;
2906 2907
	return;
direct_uncharge:
2908
	res_counter_uncharge(&memcg->res, nr_pages * PAGE_SIZE);
2909
	if (uncharge_memsw)
2910 2911 2912
		res_counter_uncharge(&memcg->memsw, nr_pages * PAGE_SIZE);
	if (unlikely(batch->memcg != memcg))
		memcg_oom_recover(memcg);
2913
}
2914

2915
/*
2916
 * uncharge if !page_mapped(page)
2917
 */
2918
static struct mem_cgroup *
2919
__mem_cgroup_uncharge_common(struct page *page, enum charge_type ctype)
2920
{
2921
	struct mem_cgroup *memcg = NULL;
2922 2923
	unsigned int nr_pages = 1;
	struct page_cgroup *pc;
2924

2925
	if (mem_cgroup_disabled())
2926
		return NULL;
2927

K
KAMEZAWA Hiroyuki 已提交
2928
	if (PageSwapCache(page))
2929
		return NULL;
K
KAMEZAWA Hiroyuki 已提交
2930

A
Andrea Arcangeli 已提交
2931
	if (PageTransHuge(page)) {
2932
		nr_pages <<= compound_order(page);
A
Andrea Arcangeli 已提交
2933 2934
		VM_BUG_ON(!PageTransHuge(page));
	}
2935
	/*
2936
	 * Check if our page_cgroup is valid
2937
	 */
2938
	pc = lookup_page_cgroup(page);
2939
	if (unlikely(!PageCgroupUsed(pc)))
2940
		return NULL;
2941

2942
	lock_page_cgroup(pc);
K
KAMEZAWA Hiroyuki 已提交
2943

2944
	memcg = pc->mem_cgroup;
2945

K
KAMEZAWA Hiroyuki 已提交
2946 2947 2948 2949 2950
	if (!PageCgroupUsed(pc))
		goto unlock_out;

	switch (ctype) {
	case MEM_CGROUP_CHARGE_TYPE_MAPPED:
K
KAMEZAWA Hiroyuki 已提交
2951
	case MEM_CGROUP_CHARGE_TYPE_DROP:
2952 2953
		/* See mem_cgroup_prepare_migration() */
		if (page_mapped(page) || PageCgroupMigration(pc))
K
KAMEZAWA Hiroyuki 已提交
2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964
			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;
2965
	}
K
KAMEZAWA Hiroyuki 已提交
2966

2967
	mem_cgroup_charge_statistics(memcg, PageCgroupCache(pc), -nr_pages);
K
KAMEZAWA Hiroyuki 已提交
2968

2969
	ClearPageCgroupUsed(pc);
2970 2971 2972 2973 2974 2975
	/*
	 * 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.
	 */
2976

2977
	unlock_page_cgroup(pc);
K
KAMEZAWA Hiroyuki 已提交
2978
	/*
2979
	 * even after unlock, we have memcg->res.usage here and this memcg
K
KAMEZAWA Hiroyuki 已提交
2980 2981
	 * will never be freed.
	 */
2982
	memcg_check_events(memcg, page);
K
KAMEZAWA Hiroyuki 已提交
2983
	if (do_swap_account && ctype == MEM_CGROUP_CHARGE_TYPE_SWAPOUT) {
2984 2985
		mem_cgroup_swap_statistics(memcg, true);
		mem_cgroup_get(memcg);
K
KAMEZAWA Hiroyuki 已提交
2986
	}
2987 2988
	if (!mem_cgroup_is_root(memcg))
		mem_cgroup_do_uncharge(memcg, nr_pages, ctype);
2989

2990
	return memcg;
K
KAMEZAWA Hiroyuki 已提交
2991 2992 2993

unlock_out:
	unlock_page_cgroup(pc);
2994
	return NULL;
2995 2996
}

2997 2998
void mem_cgroup_uncharge_page(struct page *page)
{
2999 3000 3001
	/* early check. */
	if (page_mapped(page))
		return;
3002
	VM_BUG_ON(page->mapping && !PageAnon(page));
3003 3004 3005 3006 3007 3008
	__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));
3009
	VM_BUG_ON(page->mapping);
3010 3011 3012
	__mem_cgroup_uncharge_common(page, MEM_CGROUP_CHARGE_TYPE_CACHE);
}

3013 3014 3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026
/*
 * 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;
3027 3028
		current->memcg_batch.nr_pages = 0;
		current->memcg_batch.memsw_nr_pages = 0;
3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048
	}
}

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.
	 */
3049 3050 3051 3052 3053 3054
	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);
3055
	memcg_oom_recover(batch->memcg);
3056 3057 3058 3059
	/* forget this pointer (for sanity check) */
	batch->memcg = NULL;
}

3060
#ifdef CONFIG_SWAP
3061
/*
3062
 * called after __delete_from_swap_cache() and drop "page" account.
3063 3064
 * memcg information is recorded to swap_cgroup of "ent"
 */
K
KAMEZAWA Hiroyuki 已提交
3065 3066
void
mem_cgroup_uncharge_swapcache(struct page *page, swp_entry_t ent, bool swapout)
3067 3068
{
	struct mem_cgroup *memcg;
K
KAMEZAWA Hiroyuki 已提交
3069 3070 3071 3072 3073 3074
	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);
3075

K
KAMEZAWA Hiroyuki 已提交
3076 3077 3078 3079 3080
	/*
	 * record memcg information,  if swapout && memcg != NULL,
	 * mem_cgroup_get() was called in uncharge().
	 */
	if (do_swap_account && swapout && memcg)
3081
		swap_cgroup_record(ent, css_id(&memcg->css));
3082
}
3083
#endif
3084 3085 3086 3087 3088 3089 3090

#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
3091
{
3092
	struct mem_cgroup *memcg;
3093
	unsigned short id;
3094 3095 3096 3097

	if (!do_swap_account)
		return;

3098 3099 3100
	id = swap_cgroup_record(ent, 0);
	rcu_read_lock();
	memcg = mem_cgroup_lookup(id);
3101
	if (memcg) {
3102 3103 3104 3105
		/*
		 * We uncharge this because swap is freed.
		 * This memcg can be obsolete one. We avoid calling css_tryget
		 */
3106
		if (!mem_cgroup_is_root(memcg))
3107
			res_counter_uncharge(&memcg->memsw, PAGE_SIZE);
3108
		mem_cgroup_swap_statistics(memcg, false);
3109 3110
		mem_cgroup_put(memcg);
	}
3111
	rcu_read_unlock();
K
KAMEZAWA Hiroyuki 已提交
3112
}
3113 3114 3115 3116 3117 3118

/**
 * 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
3119
 * @need_fixup: whether we should fixup res_counters and refcounts.
3120 3121 3122 3123 3124 3125 3126 3127 3128 3129
 *
 * 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,
3130
		struct mem_cgroup *from, struct mem_cgroup *to, bool need_fixup)
3131 3132 3133 3134 3135 3136 3137 3138
{
	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);
3139
		mem_cgroup_swap_statistics(to, true);
3140
		/*
3141 3142 3143 3144 3145 3146
		 * 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.
3147 3148
		 */
		mem_cgroup_get(to);
3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159
		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);
		}
3160 3161 3162 3163 3164 3165
		return 0;
	}
	return -EINVAL;
}
#else
static inline int mem_cgroup_move_swap_account(swp_entry_t entry,
3166
		struct mem_cgroup *from, struct mem_cgroup *to, bool need_fixup)
3167 3168 3169
{
	return -EINVAL;
}
3170
#endif
K
KAMEZAWA Hiroyuki 已提交
3171

3172
/*
3173 3174
 * Before starting migration, account PAGE_SIZE to mem_cgroup that the old
 * page belongs to.
3175
 */
3176
int mem_cgroup_prepare_migration(struct page *page,
3177
	struct page *newpage, struct mem_cgroup **memcgp, gfp_t gfp_mask)
3178
{
3179
	struct mem_cgroup *memcg = NULL;
3180
	struct page_cgroup *pc;
3181
	enum charge_type ctype;
3182
	int ret = 0;
3183

3184
	*memcgp = NULL;
3185

A
Andrea Arcangeli 已提交
3186
	VM_BUG_ON(PageTransHuge(page));
3187
	if (mem_cgroup_disabled())
3188 3189
		return 0;

3190 3191 3192
	pc = lookup_page_cgroup(page);
	lock_page_cgroup(pc);
	if (PageCgroupUsed(pc)) {
3193 3194
		memcg = pc->mem_cgroup;
		css_get(&memcg->css);
3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225
		/*
		 * 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);
3226
	}
3227
	unlock_page_cgroup(pc);
3228 3229 3230 3231
	/*
	 * If the page is not charged at this point,
	 * we return here.
	 */
3232
	if (!memcg)
3233
		return 0;
3234

3235 3236
	*memcgp = memcg;
	ret = __mem_cgroup_try_charge(NULL, gfp_mask, 1, memcgp, false);
3237
	css_put(&memcg->css);/* drop extra refcnt */
3238
	if (ret) {
3239 3240 3241 3242 3243 3244 3245 3246 3247
		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);
		}
3248
		/* we'll need to revisit this error code (we have -EINTR) */
3249
		return -ENOMEM;
3250
	}
3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263
	/*
	 * 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;
3264
	__mem_cgroup_commit_charge(memcg, newpage, 1, pc, ctype, false);
3265
	return ret;
3266
}
3267

3268
/* remove redundant charge if migration failed*/
3269
void mem_cgroup_end_migration(struct mem_cgroup *memcg,
3270
	struct page *oldpage, struct page *newpage, bool migration_ok)
3271
{
3272
	struct page *used, *unused;
3273 3274
	struct page_cgroup *pc;

3275
	if (!memcg)
3276
		return;
3277
	/* blocks rmdir() */
3278
	cgroup_exclude_rmdir(&memcg->css);
3279
	if (!migration_ok) {
3280 3281
		used = oldpage;
		unused = newpage;
3282
	} else {
3283
		used = newpage;
3284 3285
		unused = oldpage;
	}
3286
	/*
3287 3288 3289
	 * 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.
3290
	 */
3291 3292 3293 3294
	pc = lookup_page_cgroup(oldpage);
	lock_page_cgroup(pc);
	ClearPageCgroupMigration(pc);
	unlock_page_cgroup(pc);
3295

3296 3297
	__mem_cgroup_uncharge_common(unused, MEM_CGROUP_CHARGE_TYPE_FORCE);

3298
	/*
3299 3300 3301 3302 3303 3304
	 * 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)
3305
	 */
3306 3307
	if (PageAnon(used))
		mem_cgroup_uncharge_page(used);
3308
	/*
3309 3310
	 * At migration, we may charge account against cgroup which has no
	 * tasks.
3311 3312 3313
	 * 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.
	 */
3314
	cgroup_release_and_wakeup_rmdir(&memcg->css);
3315
}
3316

3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347
/*
 * 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;
	enum charge_type type = MEM_CGROUP_CHARGE_TYPE_CACHE;

	if (mem_cgroup_disabled())
		return;

	pc = lookup_page_cgroup(oldpage);
	/* fix accounting on old pages */
	lock_page_cgroup(pc);
	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;

	/*
	 * 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.
	 */
3348
	__mem_cgroup_commit_charge(memcg, newpage, 1, pc, type, true);
3349 3350
}

3351 3352 3353 3354 3355 3356
#ifdef CONFIG_DEBUG_VM
static struct page_cgroup *lookup_page_cgroup_used(struct page *page)
{
	struct page_cgroup *pc;

	pc = lookup_page_cgroup(page);
3357 3358 3359 3360 3361
	/*
	 * Can be NULL while feeding pages into the page allocator for
	 * the first time, i.e. during boot or memory hotplug;
	 * or when mem_cgroup_disabled().
	 */
3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378 3379 3380
	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) {
3381
		printk(KERN_ALERT "pc:%p pc->flags:%lx pc->mem_cgroup:%p\n",
3382 3383 3384 3385 3386
		       pc, pc->flags, pc->mem_cgroup);
	}
}
#endif

3387 3388
static DEFINE_MUTEX(set_limit_mutex);

3389
static int mem_cgroup_resize_limit(struct mem_cgroup *memcg,
3390
				unsigned long long val)
3391
{
3392
	int retry_count;
3393
	u64 memswlimit, memlimit;
3394
	int ret = 0;
3395 3396
	int children = mem_cgroup_count_children(memcg);
	u64 curusage, oldusage;
3397
	int enlarge;
3398 3399 3400 3401 3402 3403 3404 3405 3406

	/*
	 * 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);
3407

3408
	enlarge = 0;
3409
	while (retry_count) {
3410 3411 3412 3413
		if (signal_pending(current)) {
			ret = -EINTR;
			break;
		}
3414 3415 3416
		/*
		 * Rather than hide all in some function, I do this in
		 * open coded manner. You see what this really does.
3417
		 * We have to guarantee memcg->res.limit < memcg->memsw.limit.
3418 3419 3420 3421 3422 3423
		 */
		mutex_lock(&set_limit_mutex);
		memswlimit = res_counter_read_u64(&memcg->memsw, RES_LIMIT);
		if (memswlimit < val) {
			ret = -EINVAL;
			mutex_unlock(&set_limit_mutex);
3424 3425
			break;
		}
3426 3427 3428 3429 3430

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

3431
		ret = res_counter_set_limit(&memcg->res, val);
3432 3433 3434 3435 3436 3437
		if (!ret) {
			if (memswlimit == val)
				memcg->memsw_is_minimum = true;
			else
				memcg->memsw_is_minimum = false;
		}
3438 3439 3440 3441 3442
		mutex_unlock(&set_limit_mutex);

		if (!ret)
			break;

3443 3444
		mem_cgroup_reclaim(memcg, GFP_KERNEL,
				   MEM_CGROUP_RECLAIM_SHRINK);
3445 3446 3447 3448 3449 3450
		curusage = res_counter_read_u64(&memcg->res, RES_USAGE);
		/* Usage is reduced ? */
  		if (curusage >= oldusage)
			retry_count--;
		else
			oldusage = curusage;
3451
	}
3452 3453
	if (!ret && enlarge)
		memcg_oom_recover(memcg);
3454

3455 3456 3457
	return ret;
}

L
Li Zefan 已提交
3458 3459
static int mem_cgroup_resize_memsw_limit(struct mem_cgroup *memcg,
					unsigned long long val)
3460
{
3461
	int retry_count;
3462
	u64 memlimit, memswlimit, oldusage, curusage;
3463 3464
	int children = mem_cgroup_count_children(memcg);
	int ret = -EBUSY;
3465
	int enlarge = 0;
3466

3467 3468 3469
	/* see mem_cgroup_resize_res_limit */
 	retry_count = children * MEM_CGROUP_RECLAIM_RETRIES;
	oldusage = res_counter_read_u64(&memcg->memsw, RES_USAGE);
3470 3471 3472 3473 3474 3475 3476 3477
	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.
3478
		 * We have to guarantee memcg->res.limit < memcg->memsw.limit.
3479 3480 3481 3482 3483 3484 3485 3486
		 */
		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;
		}
3487 3488 3489
		memswlimit = res_counter_read_u64(&memcg->memsw, RES_LIMIT);
		if (memswlimit < val)
			enlarge = 1;
3490
		ret = res_counter_set_limit(&memcg->memsw, val);
3491 3492 3493 3494 3495 3496
		if (!ret) {
			if (memlimit == val)
				memcg->memsw_is_minimum = true;
			else
				memcg->memsw_is_minimum = false;
		}
3497 3498 3499 3500 3501
		mutex_unlock(&set_limit_mutex);

		if (!ret)
			break;

3502 3503 3504
		mem_cgroup_reclaim(memcg, GFP_KERNEL,
				   MEM_CGROUP_RECLAIM_NOSWAP |
				   MEM_CGROUP_RECLAIM_SHRINK);
3505
		curusage = res_counter_read_u64(&memcg->memsw, RES_USAGE);
3506
		/* Usage is reduced ? */
3507
		if (curusage >= oldusage)
3508
			retry_count--;
3509 3510
		else
			oldusage = curusage;
3511
	}
3512 3513
	if (!ret && enlarge)
		memcg_oom_recover(memcg);
3514 3515 3516
	return ret;
}

3517
unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
3518 3519
					    gfp_t gfp_mask,
					    unsigned long *total_scanned)
3520 3521 3522 3523 3524 3525
{
	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;
3526
	unsigned long long excess;
3527
	unsigned long nr_scanned;
3528 3529 3530 3531

	if (order > 0)
		return 0;

3532
	mctz = soft_limit_tree_node_zone(zone_to_nid(zone), zone_idx(zone));
3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545
	/*
	 * 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;

3546
		nr_scanned = 0;
3547
		reclaimed = mem_cgroup_soft_reclaim(mz->memcg, zone,
3548
						    gfp_mask, &nr_scanned);
3549
		nr_reclaimed += reclaimed;
3550
		*total_scanned += nr_scanned;
3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572
		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);
3573
				if (next_mz == mz)
3574
					css_put(&next_mz->memcg->css);
3575
				else /* next_mz == NULL or other memcg */
3576 3577 3578
					break;
			} while (1);
		}
3579 3580
		__mem_cgroup_remove_exceeded(mz->memcg, mz, mctz);
		excess = res_counter_soft_limit_excess(&mz->memcg->res);
3581 3582 3583 3584 3585 3586 3587 3588
		/*
		 * 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.
		 */
3589
		/* If excess == 0, no tree ops */
3590
		__mem_cgroup_insert_exceeded(mz->memcg, mz, mctz, excess);
3591
		spin_unlock(&mctz->lock);
3592
		css_put(&mz->memcg->css);
3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604
		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)
3605
		css_put(&next_mz->memcg->css);
3606 3607 3608
	return nr_reclaimed;
}

3609 3610 3611 3612
/*
 * This routine traverse page_cgroup in given list and drop them all.
 * *And* this routine doesn't reclaim page itself, just removes page_cgroup.
 */
3613
static int mem_cgroup_force_empty_list(struct mem_cgroup *memcg,
K
KAMEZAWA Hiroyuki 已提交
3614
				int node, int zid, enum lru_list lru)
3615
{
K
KAMEZAWA Hiroyuki 已提交
3616 3617
	struct mem_cgroup_per_zone *mz;
	unsigned long flags, loop;
3618
	struct list_head *list;
3619 3620
	struct page *busy;
	struct zone *zone;
3621
	int ret = 0;
3622

K
KAMEZAWA Hiroyuki 已提交
3623
	zone = &NODE_DATA(node)->node_zones[zid];
3624
	mz = mem_cgroup_zoneinfo(memcg, node, zid);
3625
	list = &mz->lruvec.lists[lru];
3626

3627
	loop = mz->lru_size[lru];
3628 3629 3630 3631
	/* give some margin against EBUSY etc...*/
	loop += 256;
	busy = NULL;
	while (loop--) {
3632
		struct page_cgroup *pc;
3633 3634
		struct page *page;

3635
		ret = 0;
K
KAMEZAWA Hiroyuki 已提交
3636
		spin_lock_irqsave(&zone->lru_lock, flags);
3637
		if (list_empty(list)) {
K
KAMEZAWA Hiroyuki 已提交
3638
			spin_unlock_irqrestore(&zone->lru_lock, flags);
3639
			break;
3640
		}
3641 3642 3643
		page = list_entry(list->prev, struct page, lru);
		if (busy == page) {
			list_move(&page->lru, list);
3644
			busy = NULL;
K
KAMEZAWA Hiroyuki 已提交
3645
			spin_unlock_irqrestore(&zone->lru_lock, flags);
3646 3647
			continue;
		}
K
KAMEZAWA Hiroyuki 已提交
3648
		spin_unlock_irqrestore(&zone->lru_lock, flags);
3649

3650
		pc = lookup_page_cgroup(page);
3651

3652
		ret = mem_cgroup_move_parent(page, pc, memcg, GFP_KERNEL);
3653
		if (ret == -ENOMEM || ret == -EINTR)
3654
			break;
3655 3656 3657

		if (ret == -EBUSY || ret == -EINVAL) {
			/* found lock contention or "pc" is obsolete. */
3658
			busy = page;
3659 3660 3661
			cond_resched();
		} else
			busy = NULL;
3662
	}
K
KAMEZAWA Hiroyuki 已提交
3663

3664 3665 3666
	if (!ret && !list_empty(list))
		return -EBUSY;
	return ret;
3667 3668 3669 3670 3671 3672
}

/*
 * make mem_cgroup's charge to be 0 if there is no task.
 * This enables deleting this mem_cgroup.
 */
3673
static int mem_cgroup_force_empty(struct mem_cgroup *memcg, bool free_all)
3674
{
3675 3676 3677
	int ret;
	int node, zid, shrink;
	int nr_retries = MEM_CGROUP_RECLAIM_RETRIES;
3678
	struct cgroup *cgrp = memcg->css.cgroup;
3679

3680
	css_get(&memcg->css);
3681 3682

	shrink = 0;
3683 3684 3685
	/* should free all ? */
	if (free_all)
		goto try_to_free;
3686
move_account:
3687
	do {
3688
		ret = -EBUSY;
3689 3690 3691 3692
		if (cgroup_task_count(cgrp) || !list_empty(&cgrp->children))
			goto out;
		ret = -EINTR;
		if (signal_pending(current))
3693
			goto out;
3694 3695
		/* This is for making all *used* pages to be on LRU. */
		lru_add_drain_all();
3696
		drain_all_stock_sync(memcg);
3697
		ret = 0;
3698
		mem_cgroup_start_move(memcg);
3699
		for_each_node_state(node, N_HIGH_MEMORY) {
3700
			for (zid = 0; !ret && zid < MAX_NR_ZONES; zid++) {
H
Hugh Dickins 已提交
3701 3702
				enum lru_list lru;
				for_each_lru(lru) {
3703
					ret = mem_cgroup_force_empty_list(memcg,
H
Hugh Dickins 已提交
3704
							node, zid, lru);
3705 3706 3707
					if (ret)
						break;
				}
3708
			}
3709 3710 3711
			if (ret)
				break;
		}
3712 3713
		mem_cgroup_end_move(memcg);
		memcg_oom_recover(memcg);
3714 3715 3716
		/* it seems parent cgroup doesn't have enough mem */
		if (ret == -ENOMEM)
			goto try_to_free;
3717
		cond_resched();
3718
	/* "ret" should also be checked to ensure all lists are empty. */
3719
	} while (memcg->res.usage > 0 || ret);
3720
out:
3721
	css_put(&memcg->css);
3722
	return ret;
3723 3724

try_to_free:
3725 3726
	/* returns EBUSY if there is a task or if we come here twice. */
	if (cgroup_task_count(cgrp) || !list_empty(&cgrp->children) || shrink) {
3727 3728 3729
		ret = -EBUSY;
		goto out;
	}
3730 3731
	/* we call try-to-free pages for make this cgroup empty */
	lru_add_drain_all();
3732 3733
	/* try to free all pages in this cgroup */
	shrink = 1;
3734
	while (nr_retries && memcg->res.usage > 0) {
3735
		int progress;
3736 3737 3738 3739 3740

		if (signal_pending(current)) {
			ret = -EINTR;
			goto out;
		}
3741
		progress = try_to_free_mem_cgroup_pages(memcg, GFP_KERNEL,
3742
						false);
3743
		if (!progress) {
3744
			nr_retries--;
3745
			/* maybe some writeback is necessary */
3746
			congestion_wait(BLK_RW_ASYNC, HZ/10);
3747
		}
3748 3749

	}
K
KAMEZAWA Hiroyuki 已提交
3750
	lru_add_drain();
3751
	/* try move_account...there may be some *locked* pages. */
3752
	goto move_account;
3753 3754
}

3755 3756 3757 3758 3759 3760
int mem_cgroup_force_empty_write(struct cgroup *cont, unsigned int event)
{
	return mem_cgroup_force_empty(mem_cgroup_from_cont(cont), true);
}


3761 3762 3763 3764 3765 3766 3767 3768 3769
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;
3770
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
3771
	struct cgroup *parent = cont->parent;
3772
	struct mem_cgroup *parent_memcg = NULL;
3773 3774

	if (parent)
3775
		parent_memcg = mem_cgroup_from_cont(parent);
3776 3777 3778

	cgroup_lock();
	/*
3779
	 * If parent's use_hierarchy is set, we can't make any modifications
3780 3781 3782 3783 3784 3785
	 * 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.
	 */
3786
	if ((!parent_memcg || !parent_memcg->use_hierarchy) &&
3787 3788
				(val == 1 || val == 0)) {
		if (list_empty(&cont->children))
3789
			memcg->use_hierarchy = val;
3790 3791 3792 3793 3794 3795 3796 3797 3798
		else
			retval = -EBUSY;
	} else
		retval = -EINVAL;
	cgroup_unlock();

	return retval;
}

3799

3800
static unsigned long mem_cgroup_recursive_stat(struct mem_cgroup *memcg,
3801
					       enum mem_cgroup_stat_index idx)
3802
{
K
KAMEZAWA Hiroyuki 已提交
3803
	struct mem_cgroup *iter;
3804
	long val = 0;
3805

3806
	/* Per-cpu values can be negative, use a signed accumulator */
3807
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
3808 3809 3810 3811 3812
		val += mem_cgroup_read_stat(iter, idx);

	if (val < 0) /* race ? */
		val = 0;
	return val;
3813 3814
}

3815
static inline u64 mem_cgroup_usage(struct mem_cgroup *memcg, bool swap)
3816
{
K
KAMEZAWA Hiroyuki 已提交
3817
	u64 val;
3818

3819
	if (!mem_cgroup_is_root(memcg)) {
3820
		if (!swap)
3821
			return res_counter_read_u64(&memcg->res, RES_USAGE);
3822
		else
3823
			return res_counter_read_u64(&memcg->memsw, RES_USAGE);
3824 3825
	}

3826 3827
	val = mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_CACHE);
	val += mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_RSS);
3828

K
KAMEZAWA Hiroyuki 已提交
3829
	if (swap)
3830
		val += mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_SWAPOUT);
3831 3832 3833 3834

	return val << PAGE_SHIFT;
}

3835
static u64 mem_cgroup_read(struct cgroup *cont, struct cftype *cft)
B
Balbir Singh 已提交
3836
{
3837
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
3838
	u64 val;
3839 3840 3841 3842 3843 3844
	int type, name;

	type = MEMFILE_TYPE(cft->private);
	name = MEMFILE_ATTR(cft->private);
	switch (type) {
	case _MEM:
3845
		if (name == RES_USAGE)
3846
			val = mem_cgroup_usage(memcg, false);
3847
		else
3848
			val = res_counter_read_u64(&memcg->res, name);
3849 3850
		break;
	case _MEMSWAP:
3851
		if (name == RES_USAGE)
3852
			val = mem_cgroup_usage(memcg, true);
3853
		else
3854
			val = res_counter_read_u64(&memcg->memsw, name);
3855 3856 3857 3858 3859 3860
		break;
	default:
		BUG();
		break;
	}
	return val;
B
Balbir Singh 已提交
3861
}
3862 3863 3864 3865
/*
 * The user of this function is...
 * RES_LIMIT.
 */
3866 3867
static int mem_cgroup_write(struct cgroup *cont, struct cftype *cft,
			    const char *buffer)
B
Balbir Singh 已提交
3868
{
3869
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
3870
	int type, name;
3871 3872 3873
	unsigned long long val;
	int ret;

3874 3875 3876
	type = MEMFILE_TYPE(cft->private);
	name = MEMFILE_ATTR(cft->private);
	switch (name) {
3877
	case RES_LIMIT:
3878 3879 3880 3881
		if (mem_cgroup_is_root(memcg)) { /* Can't set limit on root */
			ret = -EINVAL;
			break;
		}
3882 3883
		/* This function does all necessary parse...reuse it */
		ret = res_counter_memparse_write_strategy(buffer, &val);
3884 3885 3886
		if (ret)
			break;
		if (type == _MEM)
3887
			ret = mem_cgroup_resize_limit(memcg, val);
3888 3889
		else
			ret = mem_cgroup_resize_memsw_limit(memcg, val);
3890
		break;
3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904
	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;
3905 3906 3907 3908 3909
	default:
		ret = -EINVAL; /* should be BUG() ? */
		break;
	}
	return ret;
B
Balbir Singh 已提交
3910 3911
}

3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938
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;
}

3939
static int mem_cgroup_reset(struct cgroup *cont, unsigned int event)
3940
{
3941
	struct mem_cgroup *memcg;
3942
	int type, name;
3943

3944
	memcg = mem_cgroup_from_cont(cont);
3945 3946 3947
	type = MEMFILE_TYPE(event);
	name = MEMFILE_ATTR(event);
	switch (name) {
3948
	case RES_MAX_USAGE:
3949
		if (type == _MEM)
3950
			res_counter_reset_max(&memcg->res);
3951
		else
3952
			res_counter_reset_max(&memcg->memsw);
3953 3954
		break;
	case RES_FAILCNT:
3955
		if (type == _MEM)
3956
			res_counter_reset_failcnt(&memcg->res);
3957
		else
3958
			res_counter_reset_failcnt(&memcg->memsw);
3959 3960
		break;
	}
3961

3962
	return 0;
3963 3964
}

3965 3966 3967 3968 3969 3970
static u64 mem_cgroup_move_charge_read(struct cgroup *cgrp,
					struct cftype *cft)
{
	return mem_cgroup_from_cont(cgrp)->move_charge_at_immigrate;
}

3971
#ifdef CONFIG_MMU
3972 3973 3974
static int mem_cgroup_move_charge_write(struct cgroup *cgrp,
					struct cftype *cft, u64 val)
{
3975
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
3976 3977 3978 3979 3980 3981 3982 3983 3984

	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();
3985
	memcg->move_charge_at_immigrate = val;
3986 3987 3988 3989
	cgroup_unlock();

	return 0;
}
3990 3991 3992 3993 3994 3995 3996
#else
static int mem_cgroup_move_charge_write(struct cgroup *cgrp,
					struct cftype *cft, u64 val)
{
	return -ENOSYS;
}
#endif
3997

K
KAMEZAWA Hiroyuki 已提交
3998 3999 4000 4001 4002

/* For read statistics */
enum {
	MCS_CACHE,
	MCS_RSS,
4003
	MCS_FILE_MAPPED,
K
KAMEZAWA Hiroyuki 已提交
4004 4005
	MCS_PGPGIN,
	MCS_PGPGOUT,
4006
	MCS_SWAP,
4007 4008
	MCS_PGFAULT,
	MCS_PGMAJFAULT,
K
KAMEZAWA Hiroyuki 已提交
4009 4010 4011 4012 4013 4014 4015 4016 4017 4018
	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];
4019 4020
};

K
KAMEZAWA Hiroyuki 已提交
4021 4022 4023 4024 4025 4026
struct {
	char *local_name;
	char *total_name;
} memcg_stat_strings[NR_MCS_STAT] = {
	{"cache", "total_cache"},
	{"rss", "total_rss"},
4027
	{"mapped_file", "total_mapped_file"},
K
KAMEZAWA Hiroyuki 已提交
4028 4029
	{"pgpgin", "total_pgpgin"},
	{"pgpgout", "total_pgpgout"},
4030
	{"swap", "total_swap"},
4031 4032
	{"pgfault", "total_pgfault"},
	{"pgmajfault", "total_pgmajfault"},
K
KAMEZAWA Hiroyuki 已提交
4033 4034 4035 4036 4037 4038 4039 4040
	{"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
4041
static void
4042
mem_cgroup_get_local_stat(struct mem_cgroup *memcg, struct mcs_total_stat *s)
K
KAMEZAWA Hiroyuki 已提交
4043 4044 4045 4046
{
	s64 val;

	/* per cpu stat */
4047
	val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_CACHE);
K
KAMEZAWA Hiroyuki 已提交
4048
	s->stat[MCS_CACHE] += val * PAGE_SIZE;
4049
	val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_RSS);
K
KAMEZAWA Hiroyuki 已提交
4050
	s->stat[MCS_RSS] += val * PAGE_SIZE;
4051
	val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_FILE_MAPPED);
4052
	s->stat[MCS_FILE_MAPPED] += val * PAGE_SIZE;
4053
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGPGIN);
K
KAMEZAWA Hiroyuki 已提交
4054
	s->stat[MCS_PGPGIN] += val;
4055
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGPGOUT);
K
KAMEZAWA Hiroyuki 已提交
4056
	s->stat[MCS_PGPGOUT] += val;
4057
	if (do_swap_account) {
4058
		val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_SWAPOUT);
4059 4060
		s->stat[MCS_SWAP] += val * PAGE_SIZE;
	}
4061
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGFAULT);
4062
	s->stat[MCS_PGFAULT] += val;
4063
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGMAJFAULT);
4064
	s->stat[MCS_PGMAJFAULT] += val;
K
KAMEZAWA Hiroyuki 已提交
4065 4066

	/* per zone stat */
4067
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_INACTIVE_ANON));
K
KAMEZAWA Hiroyuki 已提交
4068
	s->stat[MCS_INACTIVE_ANON] += val * PAGE_SIZE;
4069
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_ACTIVE_ANON));
K
KAMEZAWA Hiroyuki 已提交
4070
	s->stat[MCS_ACTIVE_ANON] += val * PAGE_SIZE;
4071
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_INACTIVE_FILE));
K
KAMEZAWA Hiroyuki 已提交
4072
	s->stat[MCS_INACTIVE_FILE] += val * PAGE_SIZE;
4073
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_ACTIVE_FILE));
K
KAMEZAWA Hiroyuki 已提交
4074
	s->stat[MCS_ACTIVE_FILE] += val * PAGE_SIZE;
4075
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_UNEVICTABLE));
K
KAMEZAWA Hiroyuki 已提交
4076 4077 4078 4079
	s->stat[MCS_UNEVICTABLE] += val * PAGE_SIZE;
}

static void
4080
mem_cgroup_get_total_stat(struct mem_cgroup *memcg, struct mcs_total_stat *s)
K
KAMEZAWA Hiroyuki 已提交
4081
{
K
KAMEZAWA Hiroyuki 已提交
4082 4083
	struct mem_cgroup *iter;

4084
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
4085
		mem_cgroup_get_local_stat(iter, s);
K
KAMEZAWA Hiroyuki 已提交
4086 4087
}

4088 4089 4090 4091 4092 4093 4094
#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;
4095
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
4096

4097
	total_nr = mem_cgroup_nr_lru_pages(memcg, LRU_ALL);
4098 4099
	seq_printf(m, "total=%lu", total_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4100
		node_nr = mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL);
4101 4102 4103 4104
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');

4105
	file_nr = mem_cgroup_nr_lru_pages(memcg, LRU_ALL_FILE);
4106 4107
	seq_printf(m, "file=%lu", file_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4108
		node_nr = mem_cgroup_node_nr_lru_pages(memcg, nid,
4109
				LRU_ALL_FILE);
4110 4111 4112 4113
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');

4114
	anon_nr = mem_cgroup_nr_lru_pages(memcg, LRU_ALL_ANON);
4115 4116
	seq_printf(m, "anon=%lu", anon_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4117
		node_nr = mem_cgroup_node_nr_lru_pages(memcg, nid,
4118
				LRU_ALL_ANON);
4119 4120 4121 4122
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');

4123
	unevictable_nr = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_UNEVICTABLE));
4124 4125
	seq_printf(m, "unevictable=%lu", unevictable_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4126
		node_nr = mem_cgroup_node_nr_lru_pages(memcg, nid,
4127
				BIT(LRU_UNEVICTABLE));
4128 4129 4130 4131 4132 4133 4134
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');
	return 0;
}
#endif /* CONFIG_NUMA */

4135 4136
static int mem_control_stat_show(struct cgroup *cont, struct cftype *cft,
				 struct cgroup_map_cb *cb)
4137
{
4138
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
K
KAMEZAWA Hiroyuki 已提交
4139
	struct mcs_total_stat mystat;
4140 4141
	int i;

K
KAMEZAWA Hiroyuki 已提交
4142
	memset(&mystat, 0, sizeof(mystat));
4143
	mem_cgroup_get_local_stat(memcg, &mystat);
4144

4145

4146 4147 4148
	for (i = 0; i < NR_MCS_STAT; i++) {
		if (i == MCS_SWAP && !do_swap_account)
			continue;
K
KAMEZAWA Hiroyuki 已提交
4149
		cb->fill(cb, memcg_stat_strings[i].local_name, mystat.stat[i]);
4150
	}
L
Lee Schermerhorn 已提交
4151

K
KAMEZAWA Hiroyuki 已提交
4152
	/* Hierarchical information */
4153 4154
	{
		unsigned long long limit, memsw_limit;
4155
		memcg_get_hierarchical_limit(memcg, &limit, &memsw_limit);
4156 4157 4158 4159
		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
4160

K
KAMEZAWA Hiroyuki 已提交
4161
	memset(&mystat, 0, sizeof(mystat));
4162
	mem_cgroup_get_total_stat(memcg, &mystat);
4163 4164 4165
	for (i = 0; i < NR_MCS_STAT; i++) {
		if (i == MCS_SWAP && !do_swap_account)
			continue;
K
KAMEZAWA Hiroyuki 已提交
4166
		cb->fill(cb, memcg_stat_strings[i].total_name, mystat.stat[i]);
4167
	}
K
KAMEZAWA Hiroyuki 已提交
4168

K
KOSAKI Motohiro 已提交
4169 4170 4171 4172 4173 4174 4175 4176 4177
#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++) {
4178
				mz = mem_cgroup_zoneinfo(memcg, nid, zid);
K
KOSAKI Motohiro 已提交
4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195

				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

4196 4197 4198
	return 0;
}

K
KOSAKI Motohiro 已提交
4199 4200 4201 4202
static u64 mem_cgroup_swappiness_read(struct cgroup *cgrp, struct cftype *cft)
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);

4203
	return mem_cgroup_swappiness(memcg);
K
KOSAKI Motohiro 已提交
4204 4205 4206 4207 4208 4209 4210
}

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;
4211

K
KOSAKI Motohiro 已提交
4212 4213 4214 4215 4216 4217 4218
	if (val > 100)
		return -EINVAL;

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

	parent = mem_cgroup_from_cont(cgrp->parent);
4219 4220 4221

	cgroup_lock();

K
KOSAKI Motohiro 已提交
4222 4223
	/* If under hierarchy, only empty-root can set this value */
	if ((parent->use_hierarchy) ||
4224 4225
	    (memcg->use_hierarchy && !list_empty(&cgrp->children))) {
		cgroup_unlock();
K
KOSAKI Motohiro 已提交
4226
		return -EINVAL;
4227
	}
K
KOSAKI Motohiro 已提交
4228 4229 4230

	memcg->swappiness = val;

4231 4232
	cgroup_unlock();

K
KOSAKI Motohiro 已提交
4233 4234 4235
	return 0;
}

4236 4237 4238 4239 4240 4241 4242 4243
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)
4244
		t = rcu_dereference(memcg->thresholds.primary);
4245
	else
4246
		t = rcu_dereference(memcg->memsw_thresholds.primary);
4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257

	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().
	 */
4258
	i = t->current_threshold;
4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281

	/*
	 * 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 */
4282
	t->current_threshold = i - 1;
4283 4284 4285 4286 4287 4288
unlock:
	rcu_read_unlock();
}

static void mem_cgroup_threshold(struct mem_cgroup *memcg)
{
4289 4290 4291 4292 4293 4294 4295
	while (memcg) {
		__mem_cgroup_threshold(memcg, false);
		if (do_swap_account)
			__mem_cgroup_threshold(memcg, true);

		memcg = parent_mem_cgroup(memcg);
	}
4296 4297 4298 4299 4300 4301 4302 4303 4304 4305
}

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;
}

4306
static int mem_cgroup_oom_notify_cb(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
4307 4308 4309
{
	struct mem_cgroup_eventfd_list *ev;

4310
	list_for_each_entry(ev, &memcg->oom_notify, list)
K
KAMEZAWA Hiroyuki 已提交
4311 4312 4313 4314
		eventfd_signal(ev->eventfd, 1);
	return 0;
}

4315
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
4316
{
K
KAMEZAWA Hiroyuki 已提交
4317 4318
	struct mem_cgroup *iter;

4319
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
4320
		mem_cgroup_oom_notify_cb(iter);
K
KAMEZAWA Hiroyuki 已提交
4321 4322 4323 4324
}

static int mem_cgroup_usage_register_event(struct cgroup *cgrp,
	struct cftype *cft, struct eventfd_ctx *eventfd, const char *args)
4325 4326
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4327 4328
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
4329 4330
	int type = MEMFILE_TYPE(cft->private);
	u64 threshold, usage;
4331
	int i, size, ret;
4332 4333 4334 4335 4336 4337

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

	mutex_lock(&memcg->thresholds_lock);
4338

4339
	if (type == _MEM)
4340
		thresholds = &memcg->thresholds;
4341
	else if (type == _MEMSWAP)
4342
		thresholds = &memcg->memsw_thresholds;
4343 4344 4345 4346 4347 4348
	else
		BUG();

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

	/* Check if a threshold crossed before adding a new one */
4349
	if (thresholds->primary)
4350 4351
		__mem_cgroup_threshold(memcg, type == _MEMSWAP);

4352
	size = thresholds->primary ? thresholds->primary->size + 1 : 1;
4353 4354

	/* Allocate memory for new array of thresholds */
4355
	new = kmalloc(sizeof(*new) + size * sizeof(struct mem_cgroup_threshold),
4356
			GFP_KERNEL);
4357
	if (!new) {
4358 4359 4360
		ret = -ENOMEM;
		goto unlock;
	}
4361
	new->size = size;
4362 4363

	/* Copy thresholds (if any) to new array */
4364 4365
	if (thresholds->primary) {
		memcpy(new->entries, thresholds->primary->entries, (size - 1) *
4366
				sizeof(struct mem_cgroup_threshold));
4367 4368
	}

4369
	/* Add new threshold */
4370 4371
	new->entries[size - 1].eventfd = eventfd;
	new->entries[size - 1].threshold = threshold;
4372 4373

	/* Sort thresholds. Registering of new threshold isn't time-critical */
4374
	sort(new->entries, size, sizeof(struct mem_cgroup_threshold),
4375 4376 4377
			compare_thresholds, NULL);

	/* Find current threshold */
4378
	new->current_threshold = -1;
4379
	for (i = 0; i < size; i++) {
4380
		if (new->entries[i].threshold < usage) {
4381
			/*
4382 4383
			 * new->current_threshold will not be used until
			 * rcu_assign_pointer(), so it's safe to increment
4384 4385
			 * it here.
			 */
4386
			++new->current_threshold;
4387 4388 4389
		}
	}

4390 4391 4392 4393 4394
	/* Free old spare buffer and save old primary buffer as spare */
	kfree(thresholds->spare);
	thresholds->spare = thresholds->primary;

	rcu_assign_pointer(thresholds->primary, new);
4395

4396
	/* To be sure that nobody uses thresholds */
4397 4398 4399 4400 4401 4402 4403 4404
	synchronize_rcu();

unlock:
	mutex_unlock(&memcg->thresholds_lock);

	return ret;
}

4405
static void mem_cgroup_usage_unregister_event(struct cgroup *cgrp,
K
KAMEZAWA Hiroyuki 已提交
4406
	struct cftype *cft, struct eventfd_ctx *eventfd)
4407 4408
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4409 4410
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
4411 4412
	int type = MEMFILE_TYPE(cft->private);
	u64 usage;
4413
	int i, j, size;
4414 4415 4416

	mutex_lock(&memcg->thresholds_lock);
	if (type == _MEM)
4417
		thresholds = &memcg->thresholds;
4418
	else if (type == _MEMSWAP)
4419
		thresholds = &memcg->memsw_thresholds;
4420 4421 4422 4423 4424 4425 4426 4427 4428
	else
		BUG();

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

4429 4430 4431
	if (!thresholds->primary)
		goto unlock;

4432 4433 4434 4435 4436 4437
	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 */
4438 4439 4440
	size = 0;
	for (i = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd != eventfd)
4441 4442 4443
			size++;
	}

4444
	new = thresholds->spare;
4445

4446 4447
	/* Set thresholds array to NULL if we don't have thresholds */
	if (!size) {
4448 4449
		kfree(new);
		new = NULL;
4450
		goto swap_buffers;
4451 4452
	}

4453
	new->size = size;
4454 4455

	/* Copy thresholds and find current threshold */
4456 4457 4458
	new->current_threshold = -1;
	for (i = 0, j = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd == eventfd)
4459 4460
			continue;

4461 4462
		new->entries[j] = thresholds->primary->entries[i];
		if (new->entries[j].threshold < usage) {
4463
			/*
4464
			 * new->current_threshold will not be used
4465 4466 4467
			 * until rcu_assign_pointer(), so it's safe to increment
			 * it here.
			 */
4468
			++new->current_threshold;
4469 4470 4471 4472
		}
		j++;
	}

4473
swap_buffers:
4474 4475 4476
	/* Swap primary and spare array */
	thresholds->spare = thresholds->primary;
	rcu_assign_pointer(thresholds->primary, new);
4477

4478
	/* To be sure that nobody uses thresholds */
4479
	synchronize_rcu();
4480
unlock:
4481 4482
	mutex_unlock(&memcg->thresholds_lock);
}
4483

K
KAMEZAWA Hiroyuki 已提交
4484 4485 4486 4487 4488 4489 4490 4491 4492 4493 4494 4495
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;

4496
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4497 4498 4499 4500 4501

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

	/* already in OOM ? */
4502
	if (atomic_read(&memcg->under_oom))
K
KAMEZAWA Hiroyuki 已提交
4503
		eventfd_signal(eventfd, 1);
4504
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4505 4506 4507 4508

	return 0;
}

4509
static void mem_cgroup_oom_unregister_event(struct cgroup *cgrp,
K
KAMEZAWA Hiroyuki 已提交
4510 4511
	struct cftype *cft, struct eventfd_ctx *eventfd)
{
4512
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
K
KAMEZAWA Hiroyuki 已提交
4513 4514 4515 4516 4517
	struct mem_cgroup_eventfd_list *ev, *tmp;
	int type = MEMFILE_TYPE(cft->private);

	BUG_ON(type != _OOM_TYPE);

4518
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4519

4520
	list_for_each_entry_safe(ev, tmp, &memcg->oom_notify, list) {
K
KAMEZAWA Hiroyuki 已提交
4521 4522 4523 4524 4525 4526
		if (ev->eventfd == eventfd) {
			list_del(&ev->list);
			kfree(ev);
		}
	}

4527
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4528 4529
}

4530 4531 4532
static int mem_cgroup_oom_control_read(struct cgroup *cgrp,
	struct cftype *cft,  struct cgroup_map_cb *cb)
{
4533
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4534

4535
	cb->fill(cb, "oom_kill_disable", memcg->oom_kill_disable);
4536

4537
	if (atomic_read(&memcg->under_oom))
4538 4539 4540 4541 4542 4543 4544 4545 4546
		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)
{
4547
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558
	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) ||
4559
	    (memcg->use_hierarchy && !list_empty(&cgrp->children))) {
4560 4561 4562
		cgroup_unlock();
		return -EINVAL;
	}
4563
	memcg->oom_kill_disable = val;
4564
	if (!val)
4565
		memcg_oom_recover(memcg);
4566 4567 4568 4569
	cgroup_unlock();
	return 0;
}

4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584 4585
#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 */

4586 4587 4588
#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
4589 4590 4591 4592 4593 4594 4595
	/*
	 * 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
	 */
4596
	return mem_cgroup_sockets_init(cont, ss);
4597 4598
};

4599
static void kmem_cgroup_destroy(struct cgroup *cont)
G
Glauber Costa 已提交
4600
{
4601
	mem_cgroup_sockets_destroy(cont);
G
Glauber Costa 已提交
4602
}
4603 4604 4605 4606 4607
#else
static int register_kmem_files(struct cgroup *cont, struct cgroup_subsys *ss)
{
	return 0;
}
G
Glauber Costa 已提交
4608

4609
static void kmem_cgroup_destroy(struct cgroup *cont)
G
Glauber Costa 已提交
4610 4611
{
}
4612 4613
#endif

B
Balbir Singh 已提交
4614 4615
static struct cftype mem_cgroup_files[] = {
	{
4616
		.name = "usage_in_bytes",
4617
		.private = MEMFILE_PRIVATE(_MEM, RES_USAGE),
4618
		.read_u64 = mem_cgroup_read,
K
KAMEZAWA Hiroyuki 已提交
4619 4620
		.register_event = mem_cgroup_usage_register_event,
		.unregister_event = mem_cgroup_usage_unregister_event,
B
Balbir Singh 已提交
4621
	},
4622 4623
	{
		.name = "max_usage_in_bytes",
4624
		.private = MEMFILE_PRIVATE(_MEM, RES_MAX_USAGE),
4625
		.trigger = mem_cgroup_reset,
4626 4627
		.read_u64 = mem_cgroup_read,
	},
B
Balbir Singh 已提交
4628
	{
4629
		.name = "limit_in_bytes",
4630
		.private = MEMFILE_PRIVATE(_MEM, RES_LIMIT),
4631
		.write_string = mem_cgroup_write,
4632
		.read_u64 = mem_cgroup_read,
B
Balbir Singh 已提交
4633
	},
4634 4635 4636 4637 4638 4639
	{
		.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
4640 4641
	{
		.name = "failcnt",
4642
		.private = MEMFILE_PRIVATE(_MEM, RES_FAILCNT),
4643
		.trigger = mem_cgroup_reset,
4644
		.read_u64 = mem_cgroup_read,
B
Balbir Singh 已提交
4645
	},
4646 4647
	{
		.name = "stat",
4648
		.read_map = mem_control_stat_show,
4649
	},
4650 4651 4652 4653
	{
		.name = "force_empty",
		.trigger = mem_cgroup_force_empty_write,
	},
4654 4655 4656 4657 4658
	{
		.name = "use_hierarchy",
		.write_u64 = mem_cgroup_hierarchy_write,
		.read_u64 = mem_cgroup_hierarchy_read,
	},
K
KOSAKI Motohiro 已提交
4659 4660 4661 4662 4663
	{
		.name = "swappiness",
		.read_u64 = mem_cgroup_swappiness_read,
		.write_u64 = mem_cgroup_swappiness_write,
	},
4664 4665 4666 4667 4668
	{
		.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
4669 4670
	{
		.name = "oom_control",
4671 4672
		.read_map = mem_cgroup_oom_control_read,
		.write_u64 = mem_cgroup_oom_control_write,
K
KAMEZAWA Hiroyuki 已提交
4673 4674 4675 4676
		.register_event = mem_cgroup_oom_register_event,
		.unregister_event = mem_cgroup_oom_unregister_event,
		.private = MEMFILE_PRIVATE(_OOM_TYPE, OOM_CONTROL),
	},
4677 4678 4679 4680
#ifdef CONFIG_NUMA
	{
		.name = "numa_stat",
		.open = mem_control_numa_stat_open,
4681
		.mode = S_IRUGO,
4682 4683
	},
#endif
B
Balbir Singh 已提交
4684 4685
};

4686 4687 4688 4689 4690 4691
#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
4692 4693
		.register_event = mem_cgroup_usage_register_event,
		.unregister_event = mem_cgroup_usage_unregister_event,
4694 4695 4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715 4716 4717 4718 4719 4720 4721 4722 4723 4724 4725 4726 4727 4728
	},
	{
		.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

4729
static int alloc_mem_cgroup_per_zone_info(struct mem_cgroup *memcg, int node)
4730 4731
{
	struct mem_cgroup_per_node *pn;
4732
	struct mem_cgroup_per_zone *mz;
H
Hugh Dickins 已提交
4733
	enum lru_list lru;
4734
	int zone, tmp = node;
4735 4736 4737 4738 4739 4740 4741 4742
	/*
	 * 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.
	 */
4743 4744
	if (!node_state(node, N_NORMAL_MEMORY))
		tmp = -1;
4745
	pn = kzalloc_node(sizeof(*pn), GFP_KERNEL, tmp);
4746 4747
	if (!pn)
		return 1;
4748 4749 4750

	for (zone = 0; zone < MAX_NR_ZONES; zone++) {
		mz = &pn->zoneinfo[zone];
H
Hugh Dickins 已提交
4751 4752
		for_each_lru(lru)
			INIT_LIST_HEAD(&mz->lruvec.lists[lru]);
4753
		mz->usage_in_excess = 0;
4754
		mz->on_tree = false;
4755
		mz->memcg = memcg;
4756
	}
4757
	memcg->info.nodeinfo[node] = pn;
4758 4759 4760
	return 0;
}

4761
static void free_mem_cgroup_per_zone_info(struct mem_cgroup *memcg, int node)
4762
{
4763
	kfree(memcg->info.nodeinfo[node]);
4764 4765
}

4766 4767
static struct mem_cgroup *mem_cgroup_alloc(void)
{
4768
	struct mem_cgroup *memcg;
4769
	int size = sizeof(struct mem_cgroup);
4770

4771
	/* Can be very big if MAX_NUMNODES is very big */
4772
	if (size < PAGE_SIZE)
4773
		memcg = kzalloc(size, GFP_KERNEL);
4774
	else
4775
		memcg = vzalloc(size);
4776

4777
	if (!memcg)
4778 4779
		return NULL;

4780 4781
	memcg->stat = alloc_percpu(struct mem_cgroup_stat_cpu);
	if (!memcg->stat)
4782
		goto out_free;
4783 4784
	spin_lock_init(&memcg->pcp_counter_lock);
	return memcg;
4785 4786 4787

out_free:
	if (size < PAGE_SIZE)
4788
		kfree(memcg);
4789
	else
4790
		vfree(memcg);
4791
	return NULL;
4792 4793
}

4794 4795 4796 4797 4798 4799 4800 4801 4802 4803 4804 4805 4806 4807 4808 4809 4810 4811 4812 4813 4814
/*
 * Helpers for freeing a vzalloc()ed mem_cgroup by RCU,
 * but in process context.  The work_freeing structure is overlaid
 * on the rcu_freeing structure, which itself is overlaid on memsw.
 */
static void vfree_work(struct work_struct *work)
{
	struct mem_cgroup *memcg;

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

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

4815 4816 4817 4818 4819 4820 4821 4822 4823 4824 4825
/*
 * 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.
 */

4826
static void __mem_cgroup_free(struct mem_cgroup *memcg)
4827
{
K
KAMEZAWA Hiroyuki 已提交
4828 4829
	int node;

4830 4831
	mem_cgroup_remove_from_trees(memcg);
	free_css_id(&mem_cgroup_subsys, &memcg->css);
K
KAMEZAWA Hiroyuki 已提交
4832

B
Bob Liu 已提交
4833
	for_each_node(node)
4834
		free_mem_cgroup_per_zone_info(memcg, node);
K
KAMEZAWA Hiroyuki 已提交
4835

4836
	free_percpu(memcg->stat);
4837
	if (sizeof(struct mem_cgroup) < PAGE_SIZE)
4838
		kfree_rcu(memcg, rcu_freeing);
4839
	else
4840
		call_rcu(&memcg->rcu_freeing, vfree_rcu);
4841 4842
}

4843
static void mem_cgroup_get(struct mem_cgroup *memcg)
4844
{
4845
	atomic_inc(&memcg->refcnt);
4846 4847
}

4848
static void __mem_cgroup_put(struct mem_cgroup *memcg, int count)
4849
{
4850 4851 4852
	if (atomic_sub_and_test(count, &memcg->refcnt)) {
		struct mem_cgroup *parent = parent_mem_cgroup(memcg);
		__mem_cgroup_free(memcg);
4853 4854 4855
		if (parent)
			mem_cgroup_put(parent);
	}
4856 4857
}

4858
static void mem_cgroup_put(struct mem_cgroup *memcg)
4859
{
4860
	__mem_cgroup_put(memcg, 1);
4861 4862
}

4863 4864 4865
/*
 * Returns the parent mem_cgroup in memcgroup hierarchy with hierarchy enabled.
 */
G
Glauber Costa 已提交
4866
struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
4867
{
4868
	if (!memcg->res.parent)
4869
		return NULL;
4870
	return mem_cgroup_from_res_counter(memcg->res.parent, res);
4871
}
G
Glauber Costa 已提交
4872
EXPORT_SYMBOL(parent_mem_cgroup);
4873

4874 4875 4876
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
static void __init enable_swap_cgroup(void)
{
4877
	if (!mem_cgroup_disabled() && really_do_swap_account)
4878 4879 4880 4881 4882 4883 4884 4885
		do_swap_account = 1;
}
#else
static void __init enable_swap_cgroup(void)
{
}
#endif

4886 4887 4888 4889 4890 4891
static int mem_cgroup_soft_limit_tree_init(void)
{
	struct mem_cgroup_tree_per_node *rtpn;
	struct mem_cgroup_tree_per_zone *rtpz;
	int tmp, node, zone;

B
Bob Liu 已提交
4892
	for_each_node(node) {
4893 4894 4895 4896 4897
		tmp = node;
		if (!node_state(node, N_NORMAL_MEMORY))
			tmp = -1;
		rtpn = kzalloc_node(sizeof(*rtpn), GFP_KERNEL, tmp);
		if (!rtpn)
4898
			goto err_cleanup;
4899 4900 4901 4902 4903 4904 4905 4906 4907 4908

		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;
4909 4910

err_cleanup:
B
Bob Liu 已提交
4911
	for_each_node(node) {
4912 4913 4914 4915 4916 4917 4918
		if (!soft_limit_tree.rb_tree_per_node[node])
			break;
		kfree(soft_limit_tree.rb_tree_per_node[node]);
		soft_limit_tree.rb_tree_per_node[node] = NULL;
	}
	return 1;

4919 4920
}

L
Li Zefan 已提交
4921
static struct cgroup_subsys_state * __ref
4922
mem_cgroup_create(struct cgroup *cont)
B
Balbir Singh 已提交
4923
{
4924
	struct mem_cgroup *memcg, *parent;
K
KAMEZAWA Hiroyuki 已提交
4925
	long error = -ENOMEM;
4926
	int node;
B
Balbir Singh 已提交
4927

4928 4929
	memcg = mem_cgroup_alloc();
	if (!memcg)
K
KAMEZAWA Hiroyuki 已提交
4930
		return ERR_PTR(error);
4931

B
Bob Liu 已提交
4932
	for_each_node(node)
4933
		if (alloc_mem_cgroup_per_zone_info(memcg, node))
4934
			goto free_out;
4935

4936
	/* root ? */
4937
	if (cont->parent == NULL) {
4938
		int cpu;
4939
		enable_swap_cgroup();
4940
		parent = NULL;
4941 4942
		if (mem_cgroup_soft_limit_tree_init())
			goto free_out;
4943
		root_mem_cgroup = memcg;
4944 4945 4946 4947 4948
		for_each_possible_cpu(cpu) {
			struct memcg_stock_pcp *stock =
						&per_cpu(memcg_stock, cpu);
			INIT_WORK(&stock->work, drain_local_stock);
		}
4949
		hotcpu_notifier(memcg_cpu_hotplug_callback, 0);
4950
	} else {
4951
		parent = mem_cgroup_from_cont(cont->parent);
4952 4953
		memcg->use_hierarchy = parent->use_hierarchy;
		memcg->oom_kill_disable = parent->oom_kill_disable;
4954
	}
4955

4956
	if (parent && parent->use_hierarchy) {
4957 4958
		res_counter_init(&memcg->res, &parent->res);
		res_counter_init(&memcg->memsw, &parent->memsw);
4959 4960 4961 4962 4963 4964 4965
		/*
		 * 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);
4966
	} else {
4967 4968
		res_counter_init(&memcg->res, NULL);
		res_counter_init(&memcg->memsw, NULL);
4969
	}
4970 4971
	memcg->last_scanned_node = MAX_NUMNODES;
	INIT_LIST_HEAD(&memcg->oom_notify);
4972

K
KOSAKI Motohiro 已提交
4973
	if (parent)
4974 4975 4976 4977 4978
		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;
4979
free_out:
4980
	__mem_cgroup_free(memcg);
K
KAMEZAWA Hiroyuki 已提交
4981
	return ERR_PTR(error);
B
Balbir Singh 已提交
4982 4983
}

4984
static int mem_cgroup_pre_destroy(struct cgroup *cont)
4985
{
4986
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
4987

4988
	return mem_cgroup_force_empty(memcg, false);
4989 4990
}

4991
static void mem_cgroup_destroy(struct cgroup *cont)
B
Balbir Singh 已提交
4992
{
4993
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
4994

4995
	kmem_cgroup_destroy(cont);
G
Glauber Costa 已提交
4996

4997
	mem_cgroup_put(memcg);
B
Balbir Singh 已提交
4998 4999 5000 5001 5002
}

static int mem_cgroup_populate(struct cgroup_subsys *ss,
				struct cgroup *cont)
{
5003 5004 5005 5006 5007 5008 5009
	int ret;

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

	if (!ret)
		ret = register_memsw_files(cont, ss);
5010 5011 5012 5013

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

5014
	return ret;
B
Balbir Singh 已提交
5015 5016
}

5017
#ifdef CONFIG_MMU
5018
/* Handlers for move charge at task migration. */
5019 5020
#define PRECHARGE_COUNT_AT_ONCE	256
static int mem_cgroup_do_precharge(unsigned long count)
5021
{
5022 5023
	int ret = 0;
	int batch_count = PRECHARGE_COUNT_AT_ONCE;
5024
	struct mem_cgroup *memcg = mc.to;
5025

5026
	if (mem_cgroup_is_root(memcg)) {
5027 5028 5029 5030 5031 5032 5033 5034
		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;
		/*
5035
		 * "memcg" cannot be under rmdir() because we've already checked
5036 5037 5038 5039
		 * 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().
		 */
5040
		if (res_counter_charge(&memcg->res, PAGE_SIZE * count, &dummy))
5041
			goto one_by_one;
5042
		if (do_swap_account && res_counter_charge(&memcg->memsw,
5043
						PAGE_SIZE * count, &dummy)) {
5044
			res_counter_uncharge(&memcg->res, PAGE_SIZE * count);
5045 5046 5047 5048 5049 5050 5051 5052 5053 5054 5055 5056 5057 5058 5059 5060
			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();
		}
5061 5062
		ret = __mem_cgroup_try_charge(NULL,
					GFP_KERNEL, 1, &memcg, false);
5063
		if (ret)
5064
			/* mem_cgroup_clear_mc() will do uncharge later */
5065
			return ret;
5066 5067
		mc.precharge++;
	}
5068 5069 5070 5071 5072 5073 5074 5075
	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
5076
 * @target: the pointer the target page or swap ent will be stored(can be NULL)
5077 5078 5079 5080 5081 5082
 *
 * 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).
5083 5084 5085
 *   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.
5086 5087 5088 5089 5090
 *
 * Called with pte lock held.
 */
union mc_target {
	struct page	*page;
5091
	swp_entry_t	ent;
5092 5093 5094 5095 5096
};

enum mc_target_type {
	MC_TARGET_NONE,	/* not used */
	MC_TARGET_PAGE,
5097
	MC_TARGET_SWAP,
5098 5099
};

D
Daisuke Nishimura 已提交
5100 5101
static struct page *mc_handle_present_pte(struct vm_area_struct *vma,
						unsigned long addr, pte_t ptent)
5102
{
D
Daisuke Nishimura 已提交
5103
	struct page *page = vm_normal_page(vma, addr, ptent);
5104

D
Daisuke Nishimura 已提交
5105 5106 5107 5108
	if (!page || !page_mapped(page))
		return NULL;
	if (PageAnon(page)) {
		/* we don't move shared anon */
5109
		if (!move_anon() || page_mapcount(page) > 2)
D
Daisuke Nishimura 已提交
5110
			return NULL;
5111 5112
	} else if (!move_file())
		/* we ignore mapcount for file pages */
D
Daisuke Nishimura 已提交
5113 5114 5115 5116 5117 5118 5119 5120 5121 5122 5123 5124 5125 5126 5127 5128 5129 5130
		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 */
5131 5132
		if (page)
			put_page(page);
D
Daisuke Nishimura 已提交
5133
		return NULL;
5134
	}
D
Daisuke Nishimura 已提交
5135 5136 5137 5138 5139 5140
	if (do_swap_account)
		entry->val = ent.val;

	return page;
}

5141 5142 5143 5144 5145 5146 5147 5148 5149 5150 5151 5152 5153 5154 5155 5156 5157 5158 5159 5160 5161
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). */
5162 5163 5164 5165 5166 5167
	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);
5168
		if (do_swap_account)
5169 5170
			*entry = swap;
		page = find_get_page(&swapper_space, swap.val);
5171
	}
5172
#endif
5173 5174 5175
	return page;
}

D
Daisuke Nishimura 已提交
5176 5177 5178 5179 5180 5181 5182 5183 5184 5185 5186 5187
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);
5188 5189
	else if (pte_none(ptent) || pte_file(ptent))
		page = mc_handle_file_pte(vma, addr, ptent, &ent);
D
Daisuke Nishimura 已提交
5190 5191 5192

	if (!page && !ent.val)
		return 0;
5193 5194 5195 5196 5197 5198 5199 5200 5201 5202 5203 5204 5205 5206 5207
	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
5208 5209
	/* There is a swap entry and a page doesn't exist or isn't charged */
	if (ent.val && !ret &&
5210
			css_id(&mc.from->css) == lookup_swap_cgroup_id(ent)) {
5211 5212 5213
		ret = MC_TARGET_SWAP;
		if (target)
			target->ent = ent;
5214 5215 5216 5217 5218 5219 5220 5221 5222 5223 5224 5225
	}
	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;

5226
	split_huge_page_pmd(walk->mm, pmd);
5227 5228
	if (pmd_trans_unstable(pmd))
		return 0;
5229

5230 5231 5232 5233 5234 5235 5236
	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();

5237 5238 5239
	return 0;
}

5240 5241 5242 5243 5244
static unsigned long mem_cgroup_count_precharge(struct mm_struct *mm)
{
	unsigned long precharge;
	struct vm_area_struct *vma;

5245
	down_read(&mm->mmap_sem);
5246 5247 5248 5249 5250 5251 5252 5253 5254 5255 5256
	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);
	}
5257
	up_read(&mm->mmap_sem);
5258 5259 5260 5261 5262 5263 5264 5265 5266

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

	return precharge;
}

static int mem_cgroup_precharge_mc(struct mm_struct *mm)
{
5267 5268 5269 5270 5271
	unsigned long precharge = mem_cgroup_count_precharge(mm);

	VM_BUG_ON(mc.moving_task);
	mc.moving_task = current;
	return mem_cgroup_do_precharge(precharge);
5272 5273
}

5274 5275
/* cancels all extra charges on mc.from and mc.to, and wakes up all waiters. */
static void __mem_cgroup_clear_mc(void)
5276
{
5277 5278 5279
	struct mem_cgroup *from = mc.from;
	struct mem_cgroup *to = mc.to;

5280
	/* we must uncharge all the leftover precharges from mc.to */
5281 5282 5283 5284 5285 5286 5287 5288 5289 5290 5291
	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;
5292
	}
5293 5294 5295 5296 5297 5298 5299 5300 5301 5302 5303 5304 5305 5306 5307 5308 5309 5310 5311
	/* 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;
	}
5312 5313 5314 5315 5316 5317 5318 5319 5320 5321 5322 5323 5324 5325 5326
	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();
5327
	spin_lock(&mc.lock);
5328 5329
	mc.from = NULL;
	mc.to = NULL;
5330
	spin_unlock(&mc.lock);
5331
	mem_cgroup_end_move(from);
5332 5333
}

5334 5335
static int mem_cgroup_can_attach(struct cgroup *cgroup,
				 struct cgroup_taskset *tset)
5336
{
5337
	struct task_struct *p = cgroup_taskset_first(tset);
5338
	int ret = 0;
5339
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgroup);
5340

5341
	if (memcg->move_charge_at_immigrate) {
5342 5343 5344
		struct mm_struct *mm;
		struct mem_cgroup *from = mem_cgroup_from_task(p);

5345
		VM_BUG_ON(from == memcg);
5346 5347 5348 5349 5350

		mm = get_task_mm(p);
		if (!mm)
			return 0;
		/* We move charges only when we move a owner of the mm */
5351 5352 5353 5354
		if (mm->owner == p) {
			VM_BUG_ON(mc.from);
			VM_BUG_ON(mc.to);
			VM_BUG_ON(mc.precharge);
5355
			VM_BUG_ON(mc.moved_charge);
5356
			VM_BUG_ON(mc.moved_swap);
5357
			mem_cgroup_start_move(from);
5358
			spin_lock(&mc.lock);
5359
			mc.from = from;
5360
			mc.to = memcg;
5361
			spin_unlock(&mc.lock);
5362
			/* We set mc.moving_task later */
5363 5364 5365 5366

			ret = mem_cgroup_precharge_mc(mm);
			if (ret)
				mem_cgroup_clear_mc();
5367 5368
		}
		mmput(mm);
5369 5370 5371 5372
	}
	return ret;
}

5373 5374
static void mem_cgroup_cancel_attach(struct cgroup *cgroup,
				     struct cgroup_taskset *tset)
5375
{
5376
	mem_cgroup_clear_mc();
5377 5378
}

5379 5380 5381
static int mem_cgroup_move_charge_pte_range(pmd_t *pmd,
				unsigned long addr, unsigned long end,
				struct mm_walk *walk)
5382
{
5383 5384 5385 5386 5387
	int ret = 0;
	struct vm_area_struct *vma = walk->private;
	pte_t *pte;
	spinlock_t *ptl;

5388
	split_huge_page_pmd(walk->mm, pmd);
5389 5390
	if (pmd_trans_unstable(pmd))
		return 0;
5391 5392 5393 5394 5395 5396 5397 5398
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;
5399
		swp_entry_t ent;
5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410

		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);
5411 5412
			if (!mem_cgroup_move_account(page, 1, pc,
						     mc.from, mc.to, false)) {
5413
				mc.precharge--;
5414 5415
				/* we uncharge from mc.from later. */
				mc.moved_charge++;
5416 5417 5418 5419 5420
			}
			putback_lru_page(page);
put:			/* is_target_pte_for_mc() gets the page */
			put_page(page);
			break;
5421 5422
		case MC_TARGET_SWAP:
			ent = target.ent;
5423 5424
			if (!mem_cgroup_move_swap_account(ent,
						mc.from, mc.to, false)) {
5425
				mc.precharge--;
5426 5427 5428
				/* we fixup refcnts and charges later. */
				mc.moved_swap++;
			}
5429
			break;
5430 5431 5432 5433 5434 5435 5436 5437 5438 5439 5440 5441 5442 5443
		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.
		 */
5444
		ret = mem_cgroup_do_precharge(1);
5445 5446 5447 5448 5449 5450 5451 5452 5453 5454 5455 5456
		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();
5457 5458 5459 5460 5461 5462 5463 5464 5465 5466 5467 5468 5469
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;
	}
5470 5471 5472 5473 5474 5475 5476 5477 5478 5479 5480 5481 5482 5483 5484 5485 5486 5487
	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;
	}
5488
	up_read(&mm->mmap_sem);
5489 5490
}

5491 5492
static void mem_cgroup_move_task(struct cgroup *cont,
				 struct cgroup_taskset *tset)
B
Balbir Singh 已提交
5493
{
5494
	struct task_struct *p = cgroup_taskset_first(tset);
5495
	struct mm_struct *mm = get_task_mm(p);
5496 5497

	if (mm) {
5498 5499 5500
		if (mc.to)
			mem_cgroup_move_charge(mm);
		put_swap_token(mm);
5501 5502
		mmput(mm);
	}
5503 5504
	if (mc.to)
		mem_cgroup_clear_mc();
B
Balbir Singh 已提交
5505
}
5506
#else	/* !CONFIG_MMU */
5507 5508
static int mem_cgroup_can_attach(struct cgroup *cgroup,
				 struct cgroup_taskset *tset)
5509 5510 5511
{
	return 0;
}
5512 5513
static void mem_cgroup_cancel_attach(struct cgroup *cgroup,
				     struct cgroup_taskset *tset)
5514 5515
{
}
5516 5517
static void mem_cgroup_move_task(struct cgroup *cont,
				 struct cgroup_taskset *tset)
5518 5519 5520
{
}
#endif
B
Balbir Singh 已提交
5521

B
Balbir Singh 已提交
5522 5523 5524 5525
struct cgroup_subsys mem_cgroup_subsys = {
	.name = "memory",
	.subsys_id = mem_cgroup_subsys_id,
	.create = mem_cgroup_create,
5526
	.pre_destroy = mem_cgroup_pre_destroy,
B
Balbir Singh 已提交
5527 5528
	.destroy = mem_cgroup_destroy,
	.populate = mem_cgroup_populate,
5529 5530
	.can_attach = mem_cgroup_can_attach,
	.cancel_attach = mem_cgroup_cancel_attach,
B
Balbir Singh 已提交
5531
	.attach = mem_cgroup_move_task,
5532
	.early_init = 0,
K
KAMEZAWA Hiroyuki 已提交
5533
	.use_id = 1,
B
Balbir Singh 已提交
5534
};
5535 5536

#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
5537 5538 5539
static int __init enable_swap_account(char *s)
{
	/* consider enabled if no parameter or 1 is given */
5540
	if (!strcmp(s, "1"))
5541
		really_do_swap_account = 1;
5542
	else if (!strcmp(s, "0"))
5543 5544 5545
		really_do_swap_account = 0;
	return 1;
}
5546
__setup("swapaccount=", enable_swap_account);
5547 5548

#endif