memcontrol.c 145.6 KB
Newer Older
B
Balbir Singh 已提交
1 2 3 4 5
/* memcontrol.c - Memory Controller
 *
 * Copyright IBM Corporation, 2007
 * Author Balbir Singh <balbir@linux.vnet.ibm.com>
 *
6 7 8
 * Copyright 2007 OpenVZ SWsoft Inc
 * Author: Pavel Emelianov <xemul@openvz.org>
 *
9 10 11 12
 * Memory thresholds
 * Copyright (C) 2009 Nokia Corporation
 * Author: Kirill A. Shutemov
 *
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		count[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;
147 148
	struct mem_cgroup	*mem;		/* Back pointer, we cannot */
						/* use container_of	   */
149 150 151 152 153 154 155 156 157 158 159 160
};
/* Macro for accessing counter */
#define MEM_CGROUP_ZSTAT(mz, idx)	((mz)->count[(idx)])

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

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

161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180
/*
 * 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;

181 182 183 184 185
struct mem_cgroup_threshold {
	struct eventfd_ctx *eventfd;
	u64 threshold;
};

K
KAMEZAWA Hiroyuki 已提交
186
/* For threshold */
187 188
struct mem_cgroup_threshold_ary {
	/* An array index points to threshold just below usage. */
189
	int current_threshold;
190 191 192 193 194
	/* Size of entries[] */
	unsigned int size;
	/* Array of thresholds */
	struct mem_cgroup_threshold entries[0];
};
195 196 197 198 199 200 201 202 203 204 205 206

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
207 208 209 210 211
/* for OOM */
struct mem_cgroup_eventfd_list {
	struct list_head list;
	struct eventfd_ctx *eventfd;
};
212

213 214
static void mem_cgroup_threshold(struct mem_cgroup *memcg);
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg);
215

B
Balbir Singh 已提交
216 217 218 219 220 221 222
/*
 * 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
223 224 225
 * 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
226 227 228 229 230 231 232
 */
struct mem_cgroup {
	struct cgroup_subsys_state css;
	/*
	 * the counter to account for memory usage
	 */
	struct res_counter res;
233 234 235 236
	/*
	 * the counter to account for mem+swap usage.
	 */
	struct res_counter memsw;
237 238 239 240
	/*
	 * Per cgroup active and inactive list, similar to the
	 * per zone LRU lists.
	 */
241
	struct mem_cgroup_lru_info info;
242 243 244
	int last_scanned_node;
#if MAX_NUMNODES > 1
	nodemask_t	scan_nodes;
245 246
	atomic_t	numainfo_events;
	atomic_t	numainfo_updating;
247
#endif
248 249 250 251
	/*
	 * Should the accounting and control be hierarchical, per subtree?
	 */
	bool use_hierarchy;
252 253 254 255

	bool		oom_lock;
	atomic_t	under_oom;

256
	atomic_t	refcnt;
257

258
	int	swappiness;
259 260
	/* OOM-Killer disable */
	int		oom_kill_disable;
K
KOSAKI Motohiro 已提交
261

262 263 264
	/* set when res.limit == memsw.limit */
	bool		memsw_is_minimum;

265 266 267 268
	/* protect arrays of thresholds */
	struct mutex thresholds_lock;

	/* thresholds for memory usage. RCU-protected */
269
	struct mem_cgroup_thresholds thresholds;
270

271
	/* thresholds for mem+swap usage. RCU-protected */
272
	struct mem_cgroup_thresholds memsw_thresholds;
273

K
KAMEZAWA Hiroyuki 已提交
274 275
	/* For oom notifier event fd */
	struct list_head oom_notify;
276

277 278 279 280 281
	/*
	 * 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;
282
	/*
283
	 * percpu counter.
284
	 */
285
	struct mem_cgroup_stat_cpu *stat;
286 287 288 289 290 291
	/*
	 * 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
292 293 294 295

#ifdef CONFIG_INET
	struct tcp_memcontrol tcp_mem;
#endif
B
Balbir Singh 已提交
296 297
};

298 299 300 301 302 303
/* 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 {
304
	MOVE_CHARGE_TYPE_ANON,	/* private anonymous page and swap of it */
305
	MOVE_CHARGE_TYPE_FILE,	/* file page(including tmpfs) and swap of it */
306 307 308
	NR_MOVE_TYPE,
};

309 310
/* "mc" and its members are protected by cgroup_mutex */
static struct move_charge_struct {
311
	spinlock_t	  lock; /* for from, to */
312 313 314
	struct mem_cgroup *from;
	struct mem_cgroup *to;
	unsigned long precharge;
315
	unsigned long moved_charge;
316
	unsigned long moved_swap;
317 318 319
	struct task_struct *moving_task;	/* a task moving charges */
	wait_queue_head_t waitq;		/* a waitq for other context */
} mc = {
320
	.lock = __SPIN_LOCK_UNLOCKED(mc.lock),
321 322
	.waitq = __WAIT_QUEUE_HEAD_INITIALIZER(mc.waitq),
};
323

D
Daisuke Nishimura 已提交
324 325 326 327 328 329
static bool move_anon(void)
{
	return test_bit(MOVE_CHARGE_TYPE_ANON,
					&mc.to->move_charge_at_immigrate);
}

330 331 332 333 334 335
static bool move_file(void)
{
	return test_bit(MOVE_CHARGE_TYPE_FILE,
					&mc.to->move_charge_at_immigrate);
}

336 337 338 339 340 341 342
/*
 * 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)

343 344 345
enum charge_type {
	MEM_CGROUP_CHARGE_TYPE_CACHE = 0,
	MEM_CGROUP_CHARGE_TYPE_MAPPED,
346
	MEM_CGROUP_CHARGE_TYPE_SHMEM,	/* used by page migration of shmem */
347
	MEM_CGROUP_CHARGE_TYPE_FORCE,	/* used by force_empty */
K
KAMEZAWA Hiroyuki 已提交
348
	MEM_CGROUP_CHARGE_TYPE_SWAPOUT,	/* for accounting swapcache */
K
KAMEZAWA Hiroyuki 已提交
349
	MEM_CGROUP_CHARGE_TYPE_DROP,	/* a page was unused swap cache */
350 351 352
	NR_CHARGE_TYPE,
};

353
/* for encoding cft->private value on file */
354 355 356
#define _MEM			(0)
#define _MEMSWAP		(1)
#define _OOM_TYPE		(2)
357 358 359
#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
360 361
/* Used for OOM nofiier */
#define OOM_CONTROL		(0)
362

363 364 365 366 367 368 369 370
/*
 * 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)

371 372
static void mem_cgroup_get(struct mem_cgroup *memcg);
static void mem_cgroup_put(struct mem_cgroup *memcg);
G
Glauber Costa 已提交
373 374 375 376 377

/* Writing them here to avoid exposing memcg's inner layout */
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
#ifdef CONFIG_INET
#include <net/sock.h>
G
Glauber Costa 已提交
378
#include <net/ip.h>
G
Glauber Costa 已提交
379 380 381 382 383 384 385 386 387

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

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

388 389 390 391 392 393 394 395 396 397 398 399 400 401
		/* 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
402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421
		rcu_read_lock();
		memcg = mem_cgroup_from_task(current);
		if (!mem_cgroup_is_root(memcg)) {
			mem_cgroup_get(memcg);
			sk->sk_cgrp = sk->sk_prot->proto_cgroup(memcg);
		}
		rcu_read_unlock();
	}
}
EXPORT_SYMBOL(sock_update_memcg);

void sock_release_memcg(struct sock *sk)
{
	if (static_branch(&memcg_socket_limit_enabled) && sk->sk_cgrp) {
		struct mem_cgroup *memcg;
		WARN_ON(!sk->sk_cgrp->memcg);
		memcg = sk->sk_cgrp->memcg;
		mem_cgroup_put(memcg);
	}
}
G
Glauber Costa 已提交
422 423 424 425 426 427 428 429 430

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
431 432 433
#endif /* CONFIG_INET */
#endif /* CONFIG_CGROUP_MEM_RES_CTLR_KMEM */

434
static void drain_all_stock_async(struct mem_cgroup *memcg);
435

436
static struct mem_cgroup_per_zone *
437
mem_cgroup_zoneinfo(struct mem_cgroup *memcg, int nid, int zid)
438
{
439
	return &memcg->info.nodeinfo[nid]->zoneinfo[zid];
440 441
}

442
struct cgroup_subsys_state *mem_cgroup_css(struct mem_cgroup *memcg)
443
{
444
	return &memcg->css;
445 446
}

447
static struct mem_cgroup_per_zone *
448
page_cgroup_zoneinfo(struct mem_cgroup *memcg, struct page *page)
449
{
450 451
	int nid = page_to_nid(page);
	int zid = page_zonenum(page);
452

453
	return mem_cgroup_zoneinfo(memcg, nid, zid);
454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471
}

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
472
__mem_cgroup_insert_exceeded(struct mem_cgroup *memcg,
473
				struct mem_cgroup_per_zone *mz,
474 475
				struct mem_cgroup_tree_per_zone *mctz,
				unsigned long long new_usage_in_excess)
476 477 478 479 480 481 482 483
{
	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;

484 485 486
	mz->usage_in_excess = new_usage_in_excess;
	if (!mz->usage_in_excess)
		return;
487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502
	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;
503 504 505
}

static void
506
__mem_cgroup_remove_exceeded(struct mem_cgroup *memcg,
507 508 509 510 511 512 513 514 515
				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;
}

516
static void
517
mem_cgroup_remove_exceeded(struct mem_cgroup *memcg,
518 519 520 521
				struct mem_cgroup_per_zone *mz,
				struct mem_cgroup_tree_per_zone *mctz)
{
	spin_lock(&mctz->lock);
522
	__mem_cgroup_remove_exceeded(memcg, mz, mctz);
523 524 525 526
	spin_unlock(&mctz->lock);
}


527
static void mem_cgroup_update_tree(struct mem_cgroup *memcg, struct page *page)
528
{
529
	unsigned long long excess;
530 531
	struct mem_cgroup_per_zone *mz;
	struct mem_cgroup_tree_per_zone *mctz;
532 533
	int nid = page_to_nid(page);
	int zid = page_zonenum(page);
534 535 536
	mctz = soft_limit_tree_from_page(page);

	/*
537 538
	 * Necessary to update all ancestors when hierarchy is used.
	 * because their event counter is not touched.
539
	 */
540 541 542
	for (; memcg; memcg = parent_mem_cgroup(memcg)) {
		mz = mem_cgroup_zoneinfo(memcg, nid, zid);
		excess = res_counter_soft_limit_excess(&memcg->res);
543 544 545 546
		/*
		 * We have to update the tree if mz is on RB-tree or
		 * mem is over its softlimit.
		 */
547
		if (excess || mz->on_tree) {
548 549 550
			spin_lock(&mctz->lock);
			/* if on-tree, remove it */
			if (mz->on_tree)
551
				__mem_cgroup_remove_exceeded(memcg, mz, mctz);
552
			/*
553 554
			 * Insert again. mz->usage_in_excess will be updated.
			 * If excess is 0, no tree ops.
555
			 */
556
			__mem_cgroup_insert_exceeded(memcg, mz, mctz, excess);
557 558
			spin_unlock(&mctz->lock);
		}
559 560 561
	}
}

562
static void mem_cgroup_remove_from_trees(struct mem_cgroup *memcg)
563 564 565 566 567 568 569
{
	int node, zone;
	struct mem_cgroup_per_zone *mz;
	struct mem_cgroup_tree_per_zone *mctz;

	for_each_node_state(node, N_POSSIBLE) {
		for (zone = 0; zone < MAX_NR_ZONES; zone++) {
570
			mz = mem_cgroup_zoneinfo(memcg, node, zone);
571
			mctz = soft_limit_tree_node_zone(node, zone);
572
			mem_cgroup_remove_exceeded(memcg, mz, mctz);
573 574 575 576
		}
	}
}

577 578 579 580
static struct mem_cgroup_per_zone *
__mem_cgroup_largest_soft_limit_node(struct mem_cgroup_tree_per_zone *mctz)
{
	struct rb_node *rightmost = NULL;
581
	struct mem_cgroup_per_zone *mz;
582 583

retry:
584
	mz = NULL;
585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613
	rightmost = rb_last(&mctz->rb_root);
	if (!rightmost)
		goto done;		/* Nothing to reclaim from */

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

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

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

614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632
/*
 * 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.
 */
633
static long mem_cgroup_read_stat(struct mem_cgroup *memcg,
634
				 enum mem_cgroup_stat_index idx)
635
{
636
	long val = 0;
637 638
	int cpu;

639 640
	get_online_cpus();
	for_each_online_cpu(cpu)
641
		val += per_cpu(memcg->stat->count[idx], cpu);
642
#ifdef CONFIG_HOTPLUG_CPU
643 644 645
	spin_lock(&memcg->pcp_counter_lock);
	val += memcg->nocpu_base.count[idx];
	spin_unlock(&memcg->pcp_counter_lock);
646 647
#endif
	put_online_cpus();
648 649 650
	return val;
}

651
static void mem_cgroup_swap_statistics(struct mem_cgroup *memcg,
652 653 654
					 bool charge)
{
	int val = (charge) ? 1 : -1;
655
	this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_SWAPOUT], val);
656 657
}

658
void mem_cgroup_pgfault(struct mem_cgroup *memcg, int val)
659
{
660
	this_cpu_add(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT], val);
661 662
}

663
void mem_cgroup_pgmajfault(struct mem_cgroup *memcg, int val)
664
{
665
	this_cpu_add(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT], val);
666 667
}

668
static unsigned long mem_cgroup_read_events(struct mem_cgroup *memcg,
669 670 671 672 673 674
					    enum mem_cgroup_events_index idx)
{
	unsigned long val = 0;
	int cpu;

	for_each_online_cpu(cpu)
675
		val += per_cpu(memcg->stat->events[idx], cpu);
676
#ifdef CONFIG_HOTPLUG_CPU
677 678 679
	spin_lock(&memcg->pcp_counter_lock);
	val += memcg->nocpu_base.events[idx];
	spin_unlock(&memcg->pcp_counter_lock);
680 681 682 683
#endif
	return val;
}

684
static void mem_cgroup_charge_statistics(struct mem_cgroup *memcg,
685
					 bool file, int nr_pages)
686
{
687 688
	preempt_disable();

689
	if (file)
690 691
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_CACHE],
				nr_pages);
692
	else
693 694
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_RSS],
				nr_pages);
695

696 697
	/* pagein of a big page is an event. So, ignore page size */
	if (nr_pages > 0)
698
		__this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGPGIN]);
699
	else {
700
		__this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGPGOUT]);
701 702
		nr_pages = -nr_pages; /* for event */
	}
703

704
	__this_cpu_add(memcg->stat->events[MEM_CGROUP_EVENTS_COUNT], nr_pages);
705

706
	preempt_enable();
707 708
}

709
unsigned long
710
mem_cgroup_zone_nr_lru_pages(struct mem_cgroup *memcg, int nid, int zid,
711
			unsigned int lru_mask)
712 713
{
	struct mem_cgroup_per_zone *mz;
714 715 716
	enum lru_list l;
	unsigned long ret = 0;

717
	mz = mem_cgroup_zoneinfo(memcg, nid, zid);
718 719 720 721 722 723 724 725 726

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

static unsigned long
727
mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
728 729
			int nid, unsigned int lru_mask)
{
730 731 732
	u64 total = 0;
	int zid;

733
	for (zid = 0; zid < MAX_NR_ZONES; zid++)
734 735
		total += mem_cgroup_zone_nr_lru_pages(memcg,
						nid, zid, lru_mask);
736

737 738
	return total;
}
739

740
static unsigned long mem_cgroup_nr_lru_pages(struct mem_cgroup *memcg,
741
			unsigned int lru_mask)
742
{
743
	int nid;
744 745
	u64 total = 0;

746
	for_each_node_state(nid, N_HIGH_MEMORY)
747
		total += mem_cgroup_node_nr_lru_pages(memcg, nid, lru_mask);
748
	return total;
749 750
}

751
static bool __memcg_event_check(struct mem_cgroup *memcg, int target)
752 753 754
{
	unsigned long val, next;

755 756
	val = __this_cpu_read(memcg->stat->events[MEM_CGROUP_EVENTS_COUNT]);
	next = __this_cpu_read(memcg->stat->targets[target]);
757 758 759 760
	/* from time_after() in jiffies.h */
	return ((long)next - (long)val < 0);
}

761
static void __mem_cgroup_target_update(struct mem_cgroup *memcg, int target)
762
{
763
	unsigned long val, next;
764

765
	val = __this_cpu_read(memcg->stat->events[MEM_CGROUP_EVENTS_COUNT]);
766

767 768 769 770 771 772 773
	switch (target) {
	case MEM_CGROUP_TARGET_THRESH:
		next = val + THRESHOLDS_EVENTS_TARGET;
		break;
	case MEM_CGROUP_TARGET_SOFTLIMIT:
		next = val + SOFTLIMIT_EVENTS_TARGET;
		break;
774 775 776
	case MEM_CGROUP_TARGET_NUMAINFO:
		next = val + NUMAINFO_EVENTS_TARGET;
		break;
777 778 779 780
	default:
		return;
	}

781
	__this_cpu_write(memcg->stat->targets[target], next);
782 783 784 785 786 787
}

/*
 * Check events in order.
 *
 */
788
static void memcg_check_events(struct mem_cgroup *memcg, struct page *page)
789
{
790
	preempt_disable();
791
	/* threshold event is triggered in finer grain than soft limit */
792 793 794 795
	if (unlikely(__memcg_event_check(memcg, MEM_CGROUP_TARGET_THRESH))) {
		mem_cgroup_threshold(memcg);
		__mem_cgroup_target_update(memcg, MEM_CGROUP_TARGET_THRESH);
		if (unlikely(__memcg_event_check(memcg,
796
			     MEM_CGROUP_TARGET_SOFTLIMIT))) {
797 798
			mem_cgroup_update_tree(memcg, page);
			__mem_cgroup_target_update(memcg,
799 800 801
						   MEM_CGROUP_TARGET_SOFTLIMIT);
		}
#if MAX_NUMNODES > 1
802
		if (unlikely(__memcg_event_check(memcg,
803
			MEM_CGROUP_TARGET_NUMAINFO))) {
804 805
			atomic_inc(&memcg->numainfo_events);
			__mem_cgroup_target_update(memcg,
806
				MEM_CGROUP_TARGET_NUMAINFO);
807
		}
808
#endif
809
	}
810
	preempt_enable();
811 812
}

G
Glauber Costa 已提交
813
struct mem_cgroup *mem_cgroup_from_cont(struct cgroup *cont)
B
Balbir Singh 已提交
814 815 816 817 818 819
{
	return container_of(cgroup_subsys_state(cont,
				mem_cgroup_subsys_id), struct mem_cgroup,
				css);
}

820
struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p)
821
{
822 823 824 825 826 827 828 829
	/*
	 * 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;

830 831 832 833
	return container_of(task_subsys_state(p, mem_cgroup_subsys_id),
				struct mem_cgroup, css);
}

834
struct mem_cgroup *try_get_mem_cgroup_from_mm(struct mm_struct *mm)
835
{
836
	struct mem_cgroup *memcg = NULL;
837 838 839

	if (!mm)
		return NULL;
840 841 842 843 844 845 846
	/*
	 * 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 {
847 848
		memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
		if (unlikely(!memcg))
849
			break;
850
	} while (!css_tryget(&memcg->css));
851
	rcu_read_unlock();
852
	return memcg;
853 854
}

855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874
/**
 * 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
875
{
876 877
	struct mem_cgroup *memcg = NULL;
	int id = 0;
878

879 880 881
	if (mem_cgroup_disabled())
		return NULL;

882 883
	if (!root)
		root = root_mem_cgroup;
K
KAMEZAWA Hiroyuki 已提交
884

885 886
	if (prev && !reclaim)
		id = css_id(&prev->css);
K
KAMEZAWA Hiroyuki 已提交
887

888 889
	if (prev && prev != root)
		css_put(&prev->css);
K
KAMEZAWA Hiroyuki 已提交
890

891 892 893 894 895
	if (!root->use_hierarchy && root != root_mem_cgroup) {
		if (prev)
			return NULL;
		return root;
	}
K
KAMEZAWA Hiroyuki 已提交
896

897
	while (!memcg) {
898
		struct mem_cgroup_reclaim_iter *uninitialized_var(iter);
899
		struct cgroup_subsys_state *css;
900

901 902 903 904 905 906 907 908 909 910 911
		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
912

913 914 915 916 917 918 919 920
		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
921 922
		rcu_read_unlock();

923 924 925 926 927 928 929
		if (reclaim) {
			iter->position = id;
			if (!css)
				iter->generation++;
			else if (!prev && memcg)
				reclaim->generation = iter->generation;
		}
930 931 932 933 934

		if (prev && !css)
			return NULL;
	}
	return memcg;
K
KAMEZAWA Hiroyuki 已提交
935
}
K
KAMEZAWA Hiroyuki 已提交
936

937 938 939 940 941 942 943
/**
 * 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)
944 945 946 947 948 949
{
	if (!root)
		root = root_mem_cgroup;
	if (prev && prev != root)
		css_put(&prev->css);
}
K
KAMEZAWA Hiroyuki 已提交
950

951 952 953 954 955 956
/*
 * 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)		\
957
	for (iter = mem_cgroup_iter(root, NULL, NULL);	\
958
	     iter != NULL;				\
959
	     iter = mem_cgroup_iter(root, iter, NULL))
960

961
#define for_each_mem_cgroup(iter)			\
962
	for (iter = mem_cgroup_iter(NULL, NULL, NULL);	\
963
	     iter != NULL;				\
964
	     iter = mem_cgroup_iter(NULL, iter, NULL))
K
KAMEZAWA Hiroyuki 已提交
965

966
static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
967
{
968
	return (memcg == root_mem_cgroup);
969 970
}

971 972
void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
{
973
	struct mem_cgroup *memcg;
974 975 976 977 978

	if (!mm)
		return;

	rcu_read_lock();
979 980
	memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
	if (unlikely(!memcg))
981 982 983 984
		goto out;

	switch (idx) {
	case PGMAJFAULT:
985
		mem_cgroup_pgmajfault(memcg, 1);
986 987
		break;
	case PGFAULT:
988
		mem_cgroup_pgfault(memcg, 1);
989 990 991 992 993 994 995 996 997
		break;
	default:
		BUG();
	}
out:
	rcu_read_unlock();
}
EXPORT_SYMBOL(mem_cgroup_count_vm_event);

998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018
/**
 * 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
1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031
/*
 * 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.
 */
1032

1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046
/**
 * 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
1047 1048
{
	struct mem_cgroup_per_zone *mz;
1049 1050
	struct mem_cgroup *memcg;
	struct page_cgroup *pc;
1051

1052
	if (mem_cgroup_disabled())
1053 1054
		return &zone->lruvec;

K
KAMEZAWA Hiroyuki 已提交
1055
	pc = lookup_page_cgroup(page);
1056
	VM_BUG_ON(PageCgroupAcctLRU(pc));
1057
	/*
1058 1059 1060 1061 1062 1063 1064
	 * putback:				charge:
	 * SetPageLRU				SetPageCgroupUsed
	 * smp_mb				smp_mb
	 * PageCgroupUsed && add to memcg LRU	PageLRU && add to memcg LRU
	 *
	 * Ensure that one of the two sides adds the page to the memcg
	 * LRU during a race.
1065
	 */
1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083
	smp_mb();
	/*
	 * If the page is uncharged, it may be freed soon, but it
	 * could also be swap cache (readahead, swapoff) that needs to
	 * be reclaimable in the future.  root_mem_cgroup will babysit
	 * it for the time being.
	 */
	if (PageCgroupUsed(pc)) {
		/* Ensure pc->mem_cgroup is visible after reading PCG_USED. */
		smp_rmb();
		memcg = pc->mem_cgroup;
		SetPageCgroupAcctLRU(pc);
	} else
		memcg = root_mem_cgroup;
	mz = page_cgroup_zoneinfo(memcg, page);
	/* compound_order() is stabilized through lru_lock */
	MEM_CGROUP_ZSTAT(mz, lru) += 1 << compound_order(page);
	return &mz->lruvec;
K
KAMEZAWA Hiroyuki 已提交
1084
}
1085

1086 1087 1088 1089 1090 1091 1092 1093 1094
/**
 * 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.
1095
 */
1096
void mem_cgroup_lru_del_list(struct page *page, enum lru_list lru)
1097 1098
{
	struct mem_cgroup_per_zone *mz;
1099
	struct mem_cgroup *memcg;
1100 1101 1102 1103 1104 1105
	struct page_cgroup *pc;

	if (mem_cgroup_disabled())
		return;

	pc = lookup_page_cgroup(page);
1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120
	/*
	 * root_mem_cgroup babysits uncharged LRU pages, but
	 * PageCgroupUsed is cleared when the page is about to get
	 * freed.  PageCgroupAcctLRU remembers whether the
	 * LRU-accounting happened against pc->mem_cgroup or
	 * root_mem_cgroup.
	 */
	if (TestClearPageCgroupAcctLRU(pc)) {
		VM_BUG_ON(!pc->mem_cgroup);
		memcg = pc->mem_cgroup;
	} else
		memcg = root_mem_cgroup;
	mz = page_cgroup_zoneinfo(memcg, page);
	/* huge page split is done under lru_lock. so, we have no races. */
	MEM_CGROUP_ZSTAT(mz, lru) -= 1 << compound_order(page);
1121 1122
}

1123
void mem_cgroup_lru_del(struct page *page)
K
KAMEZAWA Hiroyuki 已提交
1124
{
1125
	mem_cgroup_lru_del_list(page, page_lru(page));
1126 1127
}

1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145
/**
 * 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)
1146
{
1147 1148 1149
	/* 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
1150
}
1151

K
KAMEZAWA Hiroyuki 已提交
1152
/*
1153 1154 1155 1156
 * At handling SwapCache and other FUSE stuff, pc->mem_cgroup may be changed
 * while it's linked to lru because the page may be reused after it's fully
 * uncharged. To handle that, unlink page_cgroup from LRU when charge it again.
 * It's done under lock_page and expected that zone->lru_lock isnever held.
K
KAMEZAWA Hiroyuki 已提交
1157
 */
1158
static void mem_cgroup_lru_del_before_commit(struct page *page)
K
KAMEZAWA Hiroyuki 已提交
1159
{
1160
	enum lru_list lru;
1161 1162 1163 1164
	unsigned long flags;
	struct zone *zone = page_zone(page);
	struct page_cgroup *pc = lookup_page_cgroup(page);

1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175
	/*
	 * Doing this check without taking ->lru_lock seems wrong but this
	 * is safe. Because if page_cgroup's USED bit is unset, the page
	 * will not be added to any memcg's LRU. If page_cgroup's USED bit is
	 * set, the commit after this will fail, anyway.
	 * This all charge/uncharge is done under some mutual execustion.
	 * So, we don't need to taking care of changes in USED bit.
	 */
	if (likely(!PageLRU(page)))
		return;

1176
	spin_lock_irqsave(&zone->lru_lock, flags);
1177
	lru = page_lru(page);
1178
	/*
1179 1180 1181 1182 1183 1184 1185 1186 1187
	 * The uncharged page could still be registered to the LRU of
	 * the stale pc->mem_cgroup.
	 *
	 * As pc->mem_cgroup is about to get overwritten, the old LRU
	 * accounting needs to be taken care of.  Let root_mem_cgroup
	 * babysit the page until the new memcg is responsible for it.
	 *
	 * The PCG_USED bit is guarded by lock_page() as the page is
	 * swapcache/pagecache.
1188
	 */
1189 1190 1191 1192
	if (PageLRU(page) && PageCgroupAcctLRU(pc) && !PageCgroupUsed(pc)) {
		del_page_from_lru_list(zone, page, lru);
		add_page_to_lru_list(zone, page, lru);
	}
1193
	spin_unlock_irqrestore(&zone->lru_lock, flags);
K
KAMEZAWA Hiroyuki 已提交
1194 1195
}

1196
static void mem_cgroup_lru_add_after_commit(struct page *page)
1197
{
1198
	enum lru_list lru;
1199 1200 1201
	unsigned long flags;
	struct zone *zone = page_zone(page);
	struct page_cgroup *pc = lookup_page_cgroup(page);
1202 1203 1204 1205 1206 1207 1208 1209 1210 1211
	/*
	 * putback:				charge:
	 * SetPageLRU				SetPageCgroupUsed
	 * smp_mb				smp_mb
	 * PageCgroupUsed && add to memcg LRU	PageLRU && add to memcg LRU
	 *
	 * Ensure that one of the two sides adds the page to the memcg
	 * LRU during a race.
	 */
	smp_mb();
1212 1213 1214
	/* taking care of that the page is added to LRU while we commit it */
	if (likely(!PageLRU(page)))
		return;
1215
	spin_lock_irqsave(&zone->lru_lock, flags);
1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228
	lru = page_lru(page);
	/*
	 * If the page is not on the LRU, someone will soon put it
	 * there.  If it is, and also already accounted for on the
	 * memcg-side, it must be on the right lruvec as setting
	 * pc->mem_cgroup and PageCgroupUsed is properly ordered.
	 * Otherwise, root_mem_cgroup has been babysitting the page
	 * during the charge.  Move it to the new memcg now.
	 */
	if (PageLRU(page) && !PageCgroupAcctLRU(pc)) {
		del_page_from_lru_list(zone, page, lru);
		add_page_to_lru_list(zone, page, lru);
	}
1229 1230 1231
	spin_unlock_irqrestore(&zone->lru_lock, flags);
}

1232
/*
1233
 * Checks whether given mem is same or in the root_mem_cgroup's
1234 1235
 * hierarchy subtree
 */
1236 1237
static bool mem_cgroup_same_or_subtree(const struct mem_cgroup *root_memcg,
		struct mem_cgroup *memcg)
1238
{
1239 1240 1241
	if (root_memcg != memcg) {
		return (root_memcg->use_hierarchy &&
			css_is_ancestor(&memcg->css, &root_memcg->css));
1242 1243 1244 1245 1246
	}

	return true;
}

1247
int task_in_mem_cgroup(struct task_struct *task, const struct mem_cgroup *memcg)
1248 1249
{
	int ret;
1250
	struct mem_cgroup *curr = NULL;
1251
	struct task_struct *p;
1252

1253 1254 1255 1256 1257
	p = find_lock_task_mm(task);
	if (!p)
		return 0;
	curr = try_get_mem_cgroup_from_mm(p->mm);
	task_unlock(p);
1258 1259
	if (!curr)
		return 0;
1260
	/*
1261
	 * We should check use_hierarchy of "memcg" not "curr". Because checking
1262
	 * use_hierarchy of "curr" here make this function true if hierarchy is
1263 1264
	 * enabled in "curr" and "curr" is a child of "memcg" in *cgroup*
	 * hierarchy(even if use_hierarchy is disabled in "memcg").
1265
	 */
1266
	ret = mem_cgroup_same_or_subtree(memcg, curr);
1267
	css_put(&curr->css);
1268 1269 1270
	return ret;
}

1271
int mem_cgroup_inactive_anon_is_low(struct mem_cgroup *memcg, struct zone *zone)
1272
{
1273 1274 1275
	unsigned long inactive_ratio;
	int nid = zone_to_nid(zone);
	int zid = zone_idx(zone);
1276
	unsigned long inactive;
1277
	unsigned long active;
1278
	unsigned long gb;
1279

1280 1281 1282 1283
	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));
1284

1285 1286 1287 1288 1289 1290
	gb = (inactive + active) >> (30 - PAGE_SHIFT);
	if (gb)
		inactive_ratio = int_sqrt(10 * gb);
	else
		inactive_ratio = 1;

1291
	return inactive * inactive_ratio < active;
1292 1293
}

1294
int mem_cgroup_inactive_file_is_low(struct mem_cgroup *memcg, struct zone *zone)
1295 1296 1297
{
	unsigned long active;
	unsigned long inactive;
1298 1299
	int zid = zone_idx(zone);
	int nid = zone_to_nid(zone);
1300

1301 1302 1303 1304
	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));
1305 1306 1307 1308

	return (active > inactive);
}

K
KOSAKI Motohiro 已提交
1309 1310 1311
struct zone_reclaim_stat *mem_cgroup_get_reclaim_stat(struct mem_cgroup *memcg,
						      struct zone *zone)
{
1312
	int nid = zone_to_nid(zone);
K
KOSAKI Motohiro 已提交
1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328
	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);
1329 1330
	if (!PageCgroupUsed(pc))
		return NULL;
1331 1332
	/* Ensure pc->mem_cgroup is visible after reading PCG_USED. */
	smp_rmb();
1333
	mz = page_cgroup_zoneinfo(pc->mem_cgroup, page);
K
KOSAKI Motohiro 已提交
1334 1335 1336
	return &mz->reclaim_stat;
}

1337 1338 1339
#define mem_cgroup_from_res_counter(counter, member)	\
	container_of(counter, struct mem_cgroup, member)

1340
/**
1341 1342
 * mem_cgroup_margin - calculate chargeable space of a memory cgroup
 * @mem: the memory cgroup
1343
 *
1344
 * Returns the maximum amount of memory @mem can be charged with, in
1345
 * pages.
1346
 */
1347
static unsigned long mem_cgroup_margin(struct mem_cgroup *memcg)
1348
{
1349 1350
	unsigned long long margin;

1351
	margin = res_counter_margin(&memcg->res);
1352
	if (do_swap_account)
1353
		margin = min(margin, res_counter_margin(&memcg->memsw));
1354
	return margin >> PAGE_SHIFT;
1355 1356
}

1357
int mem_cgroup_swappiness(struct mem_cgroup *memcg)
K
KOSAKI Motohiro 已提交
1358 1359 1360 1361 1362 1363 1364
{
	struct cgroup *cgrp = memcg->css.cgroup;

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

1365
	return memcg->swappiness;
K
KOSAKI Motohiro 已提交
1366 1367
}

1368
static void mem_cgroup_start_move(struct mem_cgroup *memcg)
1369 1370
{
	int cpu;
1371 1372

	get_online_cpus();
1373
	spin_lock(&memcg->pcp_counter_lock);
1374
	for_each_online_cpu(cpu)
1375 1376 1377
		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);
1378
	put_online_cpus();
1379 1380 1381 1382

	synchronize_rcu();
}

1383
static void mem_cgroup_end_move(struct mem_cgroup *memcg)
1384 1385 1386
{
	int cpu;

1387
	if (!memcg)
1388
		return;
1389
	get_online_cpus();
1390
	spin_lock(&memcg->pcp_counter_lock);
1391
	for_each_online_cpu(cpu)
1392 1393 1394
		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);
1395
	put_online_cpus();
1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408
}
/*
 * 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".
 */

1409
static bool mem_cgroup_stealed(struct mem_cgroup *memcg)
1410 1411
{
	VM_BUG_ON(!rcu_read_lock_held());
1412
	return this_cpu_read(memcg->stat->count[MEM_CGROUP_ON_MOVE]) > 0;
1413
}
1414

1415
static bool mem_cgroup_under_move(struct mem_cgroup *memcg)
1416
{
1417 1418
	struct mem_cgroup *from;
	struct mem_cgroup *to;
1419
	bool ret = false;
1420 1421 1422 1423 1424 1425 1426 1427 1428
	/*
	 * 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;
1429

1430 1431
	ret = mem_cgroup_same_or_subtree(memcg, from)
		|| mem_cgroup_same_or_subtree(memcg, to);
1432 1433
unlock:
	spin_unlock(&mc.lock);
1434 1435 1436
	return ret;
}

1437
static bool mem_cgroup_wait_acct_move(struct mem_cgroup *memcg)
1438 1439
{
	if (mc.moving_task && current != mc.moving_task) {
1440
		if (mem_cgroup_under_move(memcg)) {
1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452
			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;
}

1453
/**
1454
 * mem_cgroup_print_oom_info: Called from OOM with tasklist_lock held in read mode.
1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472
 * @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;

1473
	if (!memcg || !p)
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 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519
		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));
}

1520 1521 1522 1523
/*
 * This function returns the number of memcg under hierarchy tree. Returns
 * 1(self count) if no children.
 */
1524
static int mem_cgroup_count_children(struct mem_cgroup *memcg)
1525 1526
{
	int num = 0;
K
KAMEZAWA Hiroyuki 已提交
1527 1528
	struct mem_cgroup *iter;

1529
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
1530
		num++;
1531 1532 1533
	return num;
}

D
David Rientjes 已提交
1534 1535 1536 1537 1538 1539 1540 1541
/*
 * Return the memory (and swap, if configured) limit for a memcg.
 */
u64 mem_cgroup_get_limit(struct mem_cgroup *memcg)
{
	u64 limit;
	u64 memsw;

1542 1543 1544
	limit = res_counter_read_u64(&memcg->res, RES_LIMIT);
	limit += total_swap_pages << PAGE_SHIFT;

D
David Rientjes 已提交
1545 1546 1547 1548 1549 1550 1551 1552
	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);
}

1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588
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;
}

1589 1590 1591 1592 1593 1594 1595 1596 1597 1598
/**
 * 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.
 */
1599
static bool test_mem_cgroup_node_reclaimable(struct mem_cgroup *memcg,
1600 1601
		int nid, bool noswap)
{
1602
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_FILE))
1603 1604 1605
		return true;
	if (noswap || !total_swap_pages)
		return false;
1606
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_ANON))
1607 1608 1609 1610
		return true;
	return false;

}
1611 1612 1613 1614 1615 1616 1617 1618
#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.
 *
 */
1619
static void mem_cgroup_may_update_nodemask(struct mem_cgroup *memcg)
1620 1621
{
	int nid;
1622 1623 1624 1625
	/*
	 * numainfo_events > 0 means there was at least NUMAINFO_EVENTS_TARGET
	 * pagein/pageout changes since the last update.
	 */
1626
	if (!atomic_read(&memcg->numainfo_events))
1627
		return;
1628
	if (atomic_inc_return(&memcg->numainfo_updating) > 1)
1629 1630 1631
		return;

	/* make a nodemask where this memcg uses memory from */
1632
	memcg->scan_nodes = node_states[N_HIGH_MEMORY];
1633 1634 1635

	for_each_node_mask(nid, node_states[N_HIGH_MEMORY]) {

1636 1637
		if (!test_mem_cgroup_node_reclaimable(memcg, nid, false))
			node_clear(nid, memcg->scan_nodes);
1638
	}
1639

1640 1641
	atomic_set(&memcg->numainfo_events, 0);
	atomic_set(&memcg->numainfo_updating, 0);
1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655
}

/*
 * 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.
 */
1656
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
1657 1658 1659
{
	int node;

1660 1661
	mem_cgroup_may_update_nodemask(memcg);
	node = memcg->last_scanned_node;
1662

1663
	node = next_node(node, memcg->scan_nodes);
1664
	if (node == MAX_NUMNODES)
1665
		node = first_node(memcg->scan_nodes);
1666 1667 1668 1669 1670 1671 1672 1673 1674
	/*
	 * 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();

1675
	memcg->last_scanned_node = node;
1676 1677 1678
	return node;
}

1679 1680 1681 1682 1683 1684
/*
 * 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.
 */
1685
bool mem_cgroup_reclaimable(struct mem_cgroup *memcg, bool noswap)
1686 1687 1688 1689 1690 1691 1692
{
	int nid;

	/*
	 * quick check...making use of scan_node.
	 * We can skip unused nodes.
	 */
1693 1694
	if (!nodes_empty(memcg->scan_nodes)) {
		for (nid = first_node(memcg->scan_nodes);
1695
		     nid < MAX_NUMNODES;
1696
		     nid = next_node(nid, memcg->scan_nodes)) {
1697

1698
			if (test_mem_cgroup_node_reclaimable(memcg, nid, noswap))
1699 1700 1701 1702 1703 1704 1705
				return true;
		}
	}
	/*
	 * Check rest of nodes.
	 */
	for_each_node_state(nid, N_HIGH_MEMORY) {
1706
		if (node_isset(nid, memcg->scan_nodes))
1707
			continue;
1708
		if (test_mem_cgroup_node_reclaimable(memcg, nid, noswap))
1709 1710 1711 1712 1713
			return true;
	}
	return false;
}

1714
#else
1715
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
1716 1717 1718
{
	return 0;
}
1719

1720
bool mem_cgroup_reclaimable(struct mem_cgroup *memcg, bool noswap)
1721
{
1722
	return test_mem_cgroup_node_reclaimable(memcg, 0, noswap);
1723
}
1724 1725
#endif

1726 1727 1728 1729
static int mem_cgroup_soft_reclaim(struct mem_cgroup *root_memcg,
				   struct zone *zone,
				   gfp_t gfp_mask,
				   unsigned long *total_scanned)
1730
{
1731
	struct mem_cgroup *victim = NULL;
1732
	int total = 0;
K
KAMEZAWA Hiroyuki 已提交
1733
	int loop = 0;
1734
	unsigned long excess;
1735
	unsigned long nr_scanned;
1736 1737 1738 1739
	struct mem_cgroup_reclaim_cookie reclaim = {
		.zone = zone,
		.priority = 0,
	};
1740

1741
	excess = res_counter_soft_limit_excess(&root_memcg->res) >> PAGE_SHIFT;
K
KAMEZAWA Hiroyuki 已提交
1742

1743
	while (1) {
1744
		victim = mem_cgroup_iter(root_memcg, victim, &reclaim);
1745
		if (!victim) {
K
KAMEZAWA Hiroyuki 已提交
1746
			loop++;
1747 1748 1749 1750 1751 1752
			if (loop >= 2) {
				/*
				 * If we have not been able to reclaim
				 * anything, it might because there are
				 * no reclaimable pages under this hierarchy
				 */
1753
				if (!total)
1754 1755
					break;
				/*
L
Lucas De Marchi 已提交
1756
				 * We want to do more targeted reclaim.
1757 1758 1759 1760 1761
				 * 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) ||
1762
					(loop > MEM_CGROUP_MAX_RECLAIM_LOOPS))
1763 1764
					break;
			}
1765
			continue;
1766
		}
1767
		if (!mem_cgroup_reclaimable(victim, false))
1768
			continue;
1769 1770 1771 1772
		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))
1773
			break;
1774
	}
1775
	mem_cgroup_iter_break(root_memcg, victim);
K
KAMEZAWA Hiroyuki 已提交
1776
	return total;
1777 1778
}

K
KAMEZAWA Hiroyuki 已提交
1779 1780 1781
/*
 * Check OOM-Killer is already running under our hierarchy.
 * If someone is running, return false.
1782
 * Has to be called with memcg_oom_lock
K
KAMEZAWA Hiroyuki 已提交
1783
 */
1784
static bool mem_cgroup_oom_lock(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
1785
{
1786
	struct mem_cgroup *iter, *failed = NULL;
1787

1788
	for_each_mem_cgroup_tree(iter, memcg) {
1789
		if (iter->oom_lock) {
1790 1791 1792 1793 1794
			/*
			 * this subtree of our hierarchy is already locked
			 * so we cannot give a lock.
			 */
			failed = iter;
1795 1796
			mem_cgroup_iter_break(memcg, iter);
			break;
1797 1798
		} else
			iter->oom_lock = true;
K
KAMEZAWA Hiroyuki 已提交
1799
	}
K
KAMEZAWA Hiroyuki 已提交
1800

1801
	if (!failed)
1802
		return true;
1803 1804 1805 1806 1807

	/*
	 * OK, we failed to lock the whole subtree so we have to clean up
	 * what we set up to the failing subtree
	 */
1808
	for_each_mem_cgroup_tree(iter, memcg) {
1809
		if (iter == failed) {
1810 1811
			mem_cgroup_iter_break(memcg, iter);
			break;
1812 1813 1814
		}
		iter->oom_lock = false;
	}
1815
	return false;
1816
}
1817

1818
/*
1819
 * Has to be called with memcg_oom_lock
1820
 */
1821
static int mem_cgroup_oom_unlock(struct mem_cgroup *memcg)
1822
{
K
KAMEZAWA Hiroyuki 已提交
1823 1824
	struct mem_cgroup *iter;

1825
	for_each_mem_cgroup_tree(iter, memcg)
1826 1827 1828 1829
		iter->oom_lock = false;
	return 0;
}

1830
static void mem_cgroup_mark_under_oom(struct mem_cgroup *memcg)
1831 1832 1833
{
	struct mem_cgroup *iter;

1834
	for_each_mem_cgroup_tree(iter, memcg)
1835 1836 1837
		atomic_inc(&iter->under_oom);
}

1838
static void mem_cgroup_unmark_under_oom(struct mem_cgroup *memcg)
1839 1840 1841
{
	struct mem_cgroup *iter;

K
KAMEZAWA Hiroyuki 已提交
1842 1843 1844 1845 1846
	/*
	 * 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.
	 */
1847
	for_each_mem_cgroup_tree(iter, memcg)
1848
		atomic_add_unless(&iter->under_oom, -1, 0);
1849 1850
}

1851
static DEFINE_SPINLOCK(memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
1852 1853
static DECLARE_WAIT_QUEUE_HEAD(memcg_oom_waitq);

K
KAMEZAWA Hiroyuki 已提交
1854 1855 1856 1857 1858 1859 1860 1861
struct oom_wait_info {
	struct mem_cgroup *mem;
	wait_queue_t	wait;
};

static int memcg_oom_wake_function(wait_queue_t *wait,
	unsigned mode, int sync, void *arg)
{
1862 1863
	struct mem_cgroup *wake_memcg = (struct mem_cgroup *)arg,
			  *oom_wait_memcg;
K
KAMEZAWA Hiroyuki 已提交
1864 1865 1866
	struct oom_wait_info *oom_wait_info;

	oom_wait_info = container_of(wait, struct oom_wait_info, wait);
1867
	oom_wait_memcg = oom_wait_info->mem;
K
KAMEZAWA Hiroyuki 已提交
1868 1869 1870 1871 1872

	/*
	 * Both of oom_wait_info->mem and wake_mem are stable under us.
	 * Then we can use css_is_ancestor without taking care of RCU.
	 */
1873 1874
	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
1875 1876 1877 1878
		return 0;
	return autoremove_wake_function(wait, mode, sync, arg);
}

1879
static void memcg_wakeup_oom(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
1880
{
1881 1882
	/* for filtering, pass "memcg" as argument. */
	__wake_up(&memcg_oom_waitq, TASK_NORMAL, 0, memcg);
K
KAMEZAWA Hiroyuki 已提交
1883 1884
}

1885
static void memcg_oom_recover(struct mem_cgroup *memcg)
1886
{
1887 1888
	if (memcg && atomic_read(&memcg->under_oom))
		memcg_wakeup_oom(memcg);
1889 1890
}

K
KAMEZAWA Hiroyuki 已提交
1891 1892 1893
/*
 * try to call OOM killer. returns false if we should exit memory-reclaim loop.
 */
1894
bool mem_cgroup_handle_oom(struct mem_cgroup *memcg, gfp_t mask)
1895
{
K
KAMEZAWA Hiroyuki 已提交
1896
	struct oom_wait_info owait;
1897
	bool locked, need_to_kill;
K
KAMEZAWA Hiroyuki 已提交
1898

1899
	owait.mem = memcg;
K
KAMEZAWA Hiroyuki 已提交
1900 1901 1902 1903
	owait.wait.flags = 0;
	owait.wait.func = memcg_oom_wake_function;
	owait.wait.private = current;
	INIT_LIST_HEAD(&owait.wait.task_list);
1904
	need_to_kill = true;
1905
	mem_cgroup_mark_under_oom(memcg);
1906

1907
	/* At first, try to OOM lock hierarchy under memcg.*/
1908
	spin_lock(&memcg_oom_lock);
1909
	locked = mem_cgroup_oom_lock(memcg);
K
KAMEZAWA Hiroyuki 已提交
1910 1911 1912 1913 1914
	/*
	 * 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.
	 */
1915
	prepare_to_wait(&memcg_oom_waitq, &owait.wait, TASK_KILLABLE);
1916
	if (!locked || memcg->oom_kill_disable)
1917 1918
		need_to_kill = false;
	if (locked)
1919
		mem_cgroup_oom_notify(memcg);
1920
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
1921

1922 1923
	if (need_to_kill) {
		finish_wait(&memcg_oom_waitq, &owait.wait);
1924
		mem_cgroup_out_of_memory(memcg, mask);
1925
	} else {
K
KAMEZAWA Hiroyuki 已提交
1926
		schedule();
K
KAMEZAWA Hiroyuki 已提交
1927
		finish_wait(&memcg_oom_waitq, &owait.wait);
K
KAMEZAWA Hiroyuki 已提交
1928
	}
1929
	spin_lock(&memcg_oom_lock);
1930
	if (locked)
1931 1932
		mem_cgroup_oom_unlock(memcg);
	memcg_wakeup_oom(memcg);
1933
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
1934

1935
	mem_cgroup_unmark_under_oom(memcg);
1936

K
KAMEZAWA Hiroyuki 已提交
1937 1938 1939
	if (test_thread_flag(TIF_MEMDIE) || fatal_signal_pending(current))
		return false;
	/* Give chance to dying process */
1940
	schedule_timeout_uninterruptible(1);
K
KAMEZAWA Hiroyuki 已提交
1941
	return true;
1942 1943
}

1944 1945 1946
/*
 * Currently used to update mapped file statistics, but the routine can be
 * generalized to update other statistics as well.
1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965
 *
 * 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.
1966
 */
1967

1968 1969
void mem_cgroup_update_page_stat(struct page *page,
				 enum mem_cgroup_page_stat_item idx, int val)
1970
{
1971
	struct mem_cgroup *memcg;
1972 1973
	struct page_cgroup *pc = lookup_page_cgroup(page);
	bool need_unlock = false;
1974
	unsigned long uninitialized_var(flags);
1975 1976 1977 1978

	if (unlikely(!pc))
		return;

1979
	rcu_read_lock();
1980 1981
	memcg = pc->mem_cgroup;
	if (unlikely(!memcg || !PageCgroupUsed(pc)))
1982 1983
		goto out;
	/* pc->mem_cgroup is unstable ? */
1984
	if (unlikely(mem_cgroup_stealed(memcg)) || PageTransHuge(page)) {
1985
		/* take a lock against to access pc->mem_cgroup */
1986
		move_lock_page_cgroup(pc, &flags);
1987
		need_unlock = true;
1988 1989
		memcg = pc->mem_cgroup;
		if (!memcg || !PageCgroupUsed(pc))
1990 1991
			goto out;
	}
1992 1993

	switch (idx) {
1994
	case MEMCG_NR_FILE_MAPPED:
1995 1996 1997
		if (val > 0)
			SetPageCgroupFileMapped(pc);
		else if (!page_mapped(page))
1998
			ClearPageCgroupFileMapped(pc);
1999
		idx = MEM_CGROUP_STAT_FILE_MAPPED;
2000 2001 2002
		break;
	default:
		BUG();
2003
	}
2004

2005
	this_cpu_add(memcg->stat->count[idx], val);
2006

2007 2008
out:
	if (unlikely(need_unlock))
2009
		move_unlock_page_cgroup(pc, &flags);
2010 2011
	rcu_read_unlock();
	return;
2012
}
2013
EXPORT_SYMBOL(mem_cgroup_update_page_stat);
2014

2015 2016 2017 2018
/*
 * size of first charge trial. "32" comes from vmscan.c's magic value.
 * TODO: maybe necessary to use big numbers in big irons.
 */
2019
#define CHARGE_BATCH	32U
2020 2021
struct memcg_stock_pcp {
	struct mem_cgroup *cached; /* this never be root cgroup */
2022
	unsigned int nr_pages;
2023
	struct work_struct work;
2024 2025
	unsigned long flags;
#define FLUSHING_CACHED_CHARGE	(0)
2026 2027
};
static DEFINE_PER_CPU(struct memcg_stock_pcp, memcg_stock);
2028
static DEFINE_MUTEX(percpu_charge_mutex);
2029 2030

/*
2031
 * Try to consume stocked charge on this cpu. If success, one page is consumed
2032 2033 2034 2035
 * 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.
 */
2036
static bool consume_stock(struct mem_cgroup *memcg)
2037 2038 2039 2040 2041
{
	struct memcg_stock_pcp *stock;
	bool ret = true;

	stock = &get_cpu_var(memcg_stock);
2042
	if (memcg == stock->cached && stock->nr_pages)
2043
		stock->nr_pages--;
2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056
	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;

2057 2058 2059 2060
	if (stock->nr_pages) {
		unsigned long bytes = stock->nr_pages * PAGE_SIZE;

		res_counter_uncharge(&old->res, bytes);
2061
		if (do_swap_account)
2062 2063
			res_counter_uncharge(&old->memsw, bytes);
		stock->nr_pages = 0;
2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075
	}
	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);
2076
	clear_bit(FLUSHING_CACHED_CHARGE, &stock->flags);
2077 2078 2079 2080
}

/*
 * Cache charges(val) which is from res_counter, to local per_cpu area.
2081
 * This will be consumed by consume_stock() function, later.
2082
 */
2083
static void refill_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
2084 2085 2086
{
	struct memcg_stock_pcp *stock = &get_cpu_var(memcg_stock);

2087
	if (stock->cached != memcg) { /* reset if necessary */
2088
		drain_stock(stock);
2089
		stock->cached = memcg;
2090
	}
2091
	stock->nr_pages += nr_pages;
2092 2093 2094 2095
	put_cpu_var(memcg_stock);
}

/*
2096
 * Drains all per-CPU charge caches for given root_memcg resp. subtree
2097 2098
 * of the hierarchy under it. sync flag says whether we should block
 * until the work is done.
2099
 */
2100
static void drain_all_stock(struct mem_cgroup *root_memcg, bool sync)
2101
{
2102
	int cpu, curcpu;
2103

2104 2105
	/* Notify other cpus that system-wide "drain" is running */
	get_online_cpus();
2106
	curcpu = get_cpu();
2107 2108
	for_each_online_cpu(cpu) {
		struct memcg_stock_pcp *stock = &per_cpu(memcg_stock, cpu);
2109
		struct mem_cgroup *memcg;
2110

2111 2112
		memcg = stock->cached;
		if (!memcg || !stock->nr_pages)
2113
			continue;
2114
		if (!mem_cgroup_same_or_subtree(root_memcg, memcg))
2115
			continue;
2116 2117 2118 2119 2120 2121
		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);
		}
2122
	}
2123
	put_cpu();
2124 2125 2126 2127 2128 2129

	if (!sync)
		goto out;

	for_each_online_cpu(cpu) {
		struct memcg_stock_pcp *stock = &per_cpu(memcg_stock, cpu);
2130
		if (test_bit(FLUSHING_CACHED_CHARGE, &stock->flags))
2131 2132 2133
			flush_work(&stock->work);
	}
out:
2134
 	put_online_cpus();
2135 2136 2137 2138 2139 2140 2141 2142
}

/*
 * 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.
 */
2143
static void drain_all_stock_async(struct mem_cgroup *root_memcg)
2144
{
2145 2146 2147 2148 2149
	/*
	 * If someone calls draining, avoid adding more kworker runs.
	 */
	if (!mutex_trylock(&percpu_charge_mutex))
		return;
2150
	drain_all_stock(root_memcg, false);
2151
	mutex_unlock(&percpu_charge_mutex);
2152 2153 2154
}

/* This is a synchronous drain interface. */
2155
static void drain_all_stock_sync(struct mem_cgroup *root_memcg)
2156 2157
{
	/* called when force_empty is called */
2158
	mutex_lock(&percpu_charge_mutex);
2159
	drain_all_stock(root_memcg, true);
2160
	mutex_unlock(&percpu_charge_mutex);
2161 2162
}

2163 2164 2165 2166
/*
 * This function drains percpu counter value from DEAD cpu and
 * move it to local cpu. Note that this function can be preempted.
 */
2167
static void mem_cgroup_drain_pcp_counter(struct mem_cgroup *memcg, int cpu)
2168 2169 2170
{
	int i;

2171
	spin_lock(&memcg->pcp_counter_lock);
2172
	for (i = 0; i < MEM_CGROUP_STAT_DATA; i++) {
2173
		long x = per_cpu(memcg->stat->count[i], cpu);
2174

2175 2176
		per_cpu(memcg->stat->count[i], cpu) = 0;
		memcg->nocpu_base.count[i] += x;
2177
	}
2178
	for (i = 0; i < MEM_CGROUP_EVENTS_NSTATS; i++) {
2179
		unsigned long x = per_cpu(memcg->stat->events[i], cpu);
2180

2181 2182
		per_cpu(memcg->stat->events[i], cpu) = 0;
		memcg->nocpu_base.events[i] += x;
2183
	}
2184
	/* need to clear ON_MOVE value, works as a kind of lock. */
2185 2186
	per_cpu(memcg->stat->count[MEM_CGROUP_ON_MOVE], cpu) = 0;
	spin_unlock(&memcg->pcp_counter_lock);
2187 2188
}

2189
static void synchronize_mem_cgroup_on_move(struct mem_cgroup *memcg, int cpu)
2190 2191 2192
{
	int idx = MEM_CGROUP_ON_MOVE;

2193 2194 2195
	spin_lock(&memcg->pcp_counter_lock);
	per_cpu(memcg->stat->count[idx], cpu) = memcg->nocpu_base.count[idx];
	spin_unlock(&memcg->pcp_counter_lock);
2196 2197 2198
}

static int __cpuinit memcg_cpu_hotplug_callback(struct notifier_block *nb,
2199 2200 2201 2202 2203
					unsigned long action,
					void *hcpu)
{
	int cpu = (unsigned long)hcpu;
	struct memcg_stock_pcp *stock;
2204
	struct mem_cgroup *iter;
2205

2206
	if ((action == CPU_ONLINE)) {
2207
		for_each_mem_cgroup(iter)
2208 2209 2210 2211
			synchronize_mem_cgroup_on_move(iter, cpu);
		return NOTIFY_OK;
	}

2212
	if ((action != CPU_DEAD) || action != CPU_DEAD_FROZEN)
2213
		return NOTIFY_OK;
2214

2215
	for_each_mem_cgroup(iter)
2216 2217
		mem_cgroup_drain_pcp_counter(iter, cpu);

2218 2219 2220 2221 2222
	stock = &per_cpu(memcg_stock, cpu);
	drain_stock(stock);
	return NOTIFY_OK;
}

2223 2224 2225 2226 2227 2228 2229 2230 2231 2232

/* 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 */
};

2233
static int mem_cgroup_do_charge(struct mem_cgroup *memcg, gfp_t gfp_mask,
2234
				unsigned int nr_pages, bool oom_check)
2235
{
2236
	unsigned long csize = nr_pages * PAGE_SIZE;
2237 2238 2239 2240 2241
	struct mem_cgroup *mem_over_limit;
	struct res_counter *fail_res;
	unsigned long flags = 0;
	int ret;

2242
	ret = res_counter_charge(&memcg->res, csize, &fail_res);
2243 2244 2245 2246

	if (likely(!ret)) {
		if (!do_swap_account)
			return CHARGE_OK;
2247
		ret = res_counter_charge(&memcg->memsw, csize, &fail_res);
2248 2249 2250
		if (likely(!ret))
			return CHARGE_OK;

2251
		res_counter_uncharge(&memcg->res, csize);
2252 2253 2254 2255
		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);
2256
	/*
2257 2258
	 * nr_pages can be either a huge page (HPAGE_PMD_NR), a batch
	 * of regular pages (CHARGE_BATCH), or a single regular page (1).
2259 2260 2261 2262
	 *
	 * Never reclaim on behalf of optional batching, retry with a
	 * single page instead.
	 */
2263
	if (nr_pages == CHARGE_BATCH)
2264 2265 2266 2267 2268
		return CHARGE_RETRY;

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

2269
	ret = mem_cgroup_reclaim(mem_over_limit, gfp_mask, flags);
2270
	if (mem_cgroup_margin(mem_over_limit) >= nr_pages)
2271
		return CHARGE_RETRY;
2272
	/*
2273 2274 2275 2276 2277 2278 2279
	 * 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.
2280
	 */
2281
	if (nr_pages == 1 && ret)
2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300
		return CHARGE_RETRY;

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

	/* If we don't need to call oom-killer at el, return immediately */
	if (!oom_check)
		return CHARGE_NOMEM;
	/* check OOM */
	if (!mem_cgroup_handle_oom(mem_over_limit, gfp_mask))
		return CHARGE_OOM_DIE;

	return CHARGE_RETRY;
}

2301 2302 2303
/*
 * Unlike exported interface, "oom" parameter is added. if oom==true,
 * oom-killer can be invoked.
2304
 */
2305
static int __mem_cgroup_try_charge(struct mm_struct *mm,
A
Andrea Arcangeli 已提交
2306
				   gfp_t gfp_mask,
2307
				   unsigned int nr_pages,
2308
				   struct mem_cgroup **ptr,
2309
				   bool oom)
2310
{
2311
	unsigned int batch = max(CHARGE_BATCH, nr_pages);
2312
	int nr_oom_retries = MEM_CGROUP_RECLAIM_RETRIES;
2313
	struct mem_cgroup *memcg = NULL;
2314
	int ret;
2315

K
KAMEZAWA Hiroyuki 已提交
2316 2317 2318 2319 2320 2321 2322 2323
	/*
	 * 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;
2324

2325
	/*
2326 2327
	 * We always charge the cgroup the mm_struct belongs to.
	 * The mm_struct's mem_cgroup changes on task migration if the
2328 2329 2330
	 * thread group leader migrates. It's possible that mm is not
	 * set, if so charge the init_mm (happens for pagecache usage).
	 */
2331
	if (!*ptr && !mm)
K
KAMEZAWA Hiroyuki 已提交
2332 2333
		goto bypass;
again:
2334 2335 2336 2337
	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
2338
			goto done;
2339
		if (nr_pages == 1 && consume_stock(memcg))
K
KAMEZAWA Hiroyuki 已提交
2340
			goto done;
2341
		css_get(&memcg->css);
2342
	} else {
K
KAMEZAWA Hiroyuki 已提交
2343
		struct task_struct *p;
2344

K
KAMEZAWA Hiroyuki 已提交
2345 2346 2347
		rcu_read_lock();
		p = rcu_dereference(mm->owner);
		/*
2348
		 * Because we don't have task_lock(), "p" can exit.
2349
		 * In that case, "memcg" can point to root or p can be NULL with
2350 2351 2352 2353 2354 2355
		 * 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
2356
		 */
2357 2358
		memcg = mem_cgroup_from_task(p);
		if (!memcg || mem_cgroup_is_root(memcg)) {
K
KAMEZAWA Hiroyuki 已提交
2359 2360 2361
			rcu_read_unlock();
			goto done;
		}
2362
		if (nr_pages == 1 && consume_stock(memcg)) {
K
KAMEZAWA Hiroyuki 已提交
2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374
			/*
			 * 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 */
2375
		if (!css_tryget(&memcg->css)) {
K
KAMEZAWA Hiroyuki 已提交
2376 2377 2378 2379 2380
			rcu_read_unlock();
			goto again;
		}
		rcu_read_unlock();
	}
2381

2382 2383
	do {
		bool oom_check;
2384

2385
		/* If killed, bypass charge */
K
KAMEZAWA Hiroyuki 已提交
2386
		if (fatal_signal_pending(current)) {
2387
			css_put(&memcg->css);
2388
			goto bypass;
K
KAMEZAWA Hiroyuki 已提交
2389
		}
2390

2391 2392 2393 2394
		oom_check = false;
		if (oom && !nr_oom_retries) {
			oom_check = true;
			nr_oom_retries = MEM_CGROUP_RECLAIM_RETRIES;
2395
		}
2396

2397
		ret = mem_cgroup_do_charge(memcg, gfp_mask, batch, oom_check);
2398 2399 2400 2401
		switch (ret) {
		case CHARGE_OK:
			break;
		case CHARGE_RETRY: /* not in OOM situation but retry */
2402
			batch = nr_pages;
2403 2404
			css_put(&memcg->css);
			memcg = NULL;
K
KAMEZAWA Hiroyuki 已提交
2405
			goto again;
2406
		case CHARGE_WOULDBLOCK: /* !__GFP_WAIT */
2407
			css_put(&memcg->css);
2408 2409
			goto nomem;
		case CHARGE_NOMEM: /* OOM routine works */
K
KAMEZAWA Hiroyuki 已提交
2410
			if (!oom) {
2411
				css_put(&memcg->css);
K
KAMEZAWA Hiroyuki 已提交
2412
				goto nomem;
K
KAMEZAWA Hiroyuki 已提交
2413
			}
2414 2415 2416 2417
			/* If oom, we never return -ENOMEM */
			nr_oom_retries--;
			break;
		case CHARGE_OOM_DIE: /* Killed by OOM Killer */
2418
			css_put(&memcg->css);
K
KAMEZAWA Hiroyuki 已提交
2419
			goto bypass;
2420
		}
2421 2422
	} while (ret != CHARGE_OK);

2423
	if (batch > nr_pages)
2424 2425
		refill_stock(memcg, batch - nr_pages);
	css_put(&memcg->css);
2426
done:
2427
	*ptr = memcg;
2428 2429
	return 0;
nomem:
2430
	*ptr = NULL;
2431
	return -ENOMEM;
K
KAMEZAWA Hiroyuki 已提交
2432
bypass:
2433
	*ptr = NULL;
K
KAMEZAWA Hiroyuki 已提交
2434
	return 0;
2435
}
2436

2437 2438 2439 2440 2441
/*
 * 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().
 */
2442
static void __mem_cgroup_cancel_charge(struct mem_cgroup *memcg,
2443
				       unsigned int nr_pages)
2444
{
2445
	if (!mem_cgroup_is_root(memcg)) {
2446 2447
		unsigned long bytes = nr_pages * PAGE_SIZE;

2448
		res_counter_uncharge(&memcg->res, bytes);
2449
		if (do_swap_account)
2450
			res_counter_uncharge(&memcg->memsw, bytes);
2451
	}
2452 2453
}

2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472
/*
 * 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);
}

2473
struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page)
2474
{
2475
	struct mem_cgroup *memcg = NULL;
2476
	struct page_cgroup *pc;
2477
	unsigned short id;
2478 2479
	swp_entry_t ent;

2480 2481 2482
	VM_BUG_ON(!PageLocked(page));

	pc = lookup_page_cgroup(page);
2483
	lock_page_cgroup(pc);
2484
	if (PageCgroupUsed(pc)) {
2485 2486 2487
		memcg = pc->mem_cgroup;
		if (memcg && !css_tryget(&memcg->css))
			memcg = NULL;
2488
	} else if (PageSwapCache(page)) {
2489
		ent.val = page_private(page);
2490 2491
		id = lookup_swap_cgroup(ent);
		rcu_read_lock();
2492 2493 2494
		memcg = mem_cgroup_lookup(id);
		if (memcg && !css_tryget(&memcg->css))
			memcg = NULL;
2495
		rcu_read_unlock();
2496
	}
2497
	unlock_page_cgroup(pc);
2498
	return memcg;
2499 2500
}

2501
static void __mem_cgroup_commit_charge(struct mem_cgroup *memcg,
2502
				       struct page *page,
2503
				       unsigned int nr_pages,
2504
				       struct page_cgroup *pc,
2505
				       enum charge_type ctype)
2506
{
2507 2508 2509
	lock_page_cgroup(pc);
	if (unlikely(PageCgroupUsed(pc))) {
		unlock_page_cgroup(pc);
2510
		__mem_cgroup_cancel_charge(memcg, nr_pages);
2511 2512 2513 2514 2515 2516
		return;
	}
	/*
	 * we don't need page_cgroup_lock about tail pages, becase they are not
	 * accessed by any other context at this point.
	 */
2517
	pc->mem_cgroup = memcg;
2518 2519 2520 2521 2522 2523 2524
	/*
	 * 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
2525
	smp_wmb();
2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538
	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;
	}
2539

2540
	mem_cgroup_charge_statistics(memcg, PageCgroupCache(pc), nr_pages);
2541
	unlock_page_cgroup(pc);
2542 2543 2544 2545 2546
	/*
	 * "charge_statistics" updated event counter. Then, check it.
	 * Insert ancestor (and ancestor's ancestors), to softlimit RB-tree.
	 * if they exceeds softlimit.
	 */
2547
	memcg_check_events(memcg, page);
2548
}
2549

2550 2551 2552 2553 2554 2555
#ifdef CONFIG_TRANSPARENT_HUGEPAGE

#define PCGF_NOCOPY_AT_SPLIT ((1 << PCG_LOCK) | (1 << PCG_MOVE_LOCK) |\
			(1 << PCG_ACCT_LRU) | (1 << PCG_MIGRATION))
/*
 * Because tail pages are not marked as "used", set it. We're under
2556 2557 2558
 * 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.
2559
 */
2560
void mem_cgroup_split_huge_fixup(struct page *head)
2561 2562
{
	struct page_cgroup *head_pc = lookup_page_cgroup(head);
2563 2564
	struct page_cgroup *pc;
	int i;
2565

2566 2567
	if (mem_cgroup_disabled())
		return;
2568 2569 2570 2571 2572 2573 2574 2575 2576 2577
	for (i = 1; i < HPAGE_PMD_NR; i++) {
		pc = head_pc + i;
		pc->mem_cgroup = head_pc->mem_cgroup;
		smp_wmb();/* see __commit_charge() */
		/*
		 * LRU flags cannot be copied because we need to add tail
		 * page to LRU by generic call and our hooks will be called.
		 */
		pc->flags = head_pc->flags & ~PCGF_NOCOPY_AT_SPLIT;
	}
2578

2579 2580 2581 2582 2583 2584 2585
	if (PageCgroupAcctLRU(head_pc)) {
		enum lru_list lru;
		struct mem_cgroup_per_zone *mz;
		/*
		 * We hold lru_lock, then, reduce counter directly.
		 */
		lru = page_lru(head);
2586
		mz = page_cgroup_zoneinfo(head_pc->mem_cgroup, head);
2587
		MEM_CGROUP_ZSTAT(mz, lru) -= HPAGE_PMD_NR - 1;
2588
	}
2589 2590 2591
}
#endif

2592
/**
2593
 * mem_cgroup_move_account - move account of the page
2594
 * @page: the page
2595
 * @nr_pages: number of regular pages (>1 for huge pages)
2596 2597 2598
 * @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.
2599
 * @uncharge: whether we should call uncharge and css_put against @from.
2600 2601
 *
 * The caller must confirm following.
K
KAMEZAWA Hiroyuki 已提交
2602
 * - page is not on LRU (isolate_page() is useful.)
2603
 * - compound_lock is held when nr_pages > 1
2604
 *
2605
 * This function doesn't do "charge" nor css_get to new cgroup. It should be
L
Lucas De Marchi 已提交
2606
 * done by a caller(__mem_cgroup_try_charge would be useful). If @uncharge is
2607 2608
 * true, this function does "uncharge" from old cgroup, but it doesn't if
 * @uncharge is false, so a caller should do "uncharge".
2609
 */
2610 2611 2612 2613 2614 2615
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)
2616
{
2617 2618
	unsigned long flags;
	int ret;
2619

2620
	VM_BUG_ON(from == to);
2621
	VM_BUG_ON(PageLRU(page));
2622 2623 2624 2625 2626 2627 2628
	/*
	 * 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;
2629
	if (nr_pages > 1 && !PageTransHuge(page))
2630 2631 2632 2633 2634 2635 2636 2637 2638
		goto out;

	lock_page_cgroup(pc);

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

	move_lock_page_cgroup(pc, &flags);
2639

2640
	if (PageCgroupFileMapped(pc)) {
2641 2642 2643 2644 2645
		/* 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();
2646
	}
2647
	mem_cgroup_charge_statistics(from, PageCgroupCache(pc), -nr_pages);
2648 2649
	if (uncharge)
		/* This is not "cancel", but cancel_charge does all we need. */
2650
		__mem_cgroup_cancel_charge(from, nr_pages);
2651

2652
	/* caller should have done css_get */
K
KAMEZAWA Hiroyuki 已提交
2653
	pc->mem_cgroup = to;
2654
	mem_cgroup_charge_statistics(to, PageCgroupCache(pc), nr_pages);
2655 2656 2657
	/*
	 * We charges against "to" which may not have any tasks. Then, "to"
	 * can be under rmdir(). But in current implementation, caller of
2658
	 * this function is just force_empty() and move charge, so it's
L
Lucas De Marchi 已提交
2659
	 * guaranteed that "to" is never removed. So, we don't check rmdir
2660
	 * status here.
2661
	 */
2662 2663 2664
	move_unlock_page_cgroup(pc, &flags);
	ret = 0;
unlock:
2665
	unlock_page_cgroup(pc);
2666 2667 2668
	/*
	 * check events
	 */
2669 2670
	memcg_check_events(to, page);
	memcg_check_events(from, page);
2671
out:
2672 2673 2674 2675 2676 2677 2678
	return ret;
}

/*
 * move charges to its parent.
 */

2679 2680
static int mem_cgroup_move_parent(struct page *page,
				  struct page_cgroup *pc,
2681 2682 2683 2684 2685 2686
				  struct mem_cgroup *child,
				  gfp_t gfp_mask)
{
	struct cgroup *cg = child->css.cgroup;
	struct cgroup *pcg = cg->parent;
	struct mem_cgroup *parent;
2687
	unsigned int nr_pages;
2688
	unsigned long uninitialized_var(flags);
2689 2690 2691 2692 2693 2694
	int ret;

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

2695 2696 2697 2698 2699
	ret = -EBUSY;
	if (!get_page_unless_zero(page))
		goto out;
	if (isolate_lru_page(page))
		goto put;
2700

2701
	nr_pages = hpage_nr_pages(page);
K
KAMEZAWA Hiroyuki 已提交
2702

2703
	parent = mem_cgroup_from_cont(pcg);
2704
	ret = __mem_cgroup_try_charge(NULL, gfp_mask, nr_pages, &parent, false);
2705
	if (ret || !parent)
2706
		goto put_back;
2707

2708
	if (nr_pages > 1)
2709 2710
		flags = compound_lock_irqsave(page);

2711
	ret = mem_cgroup_move_account(page, nr_pages, pc, child, parent, true);
2712
	if (ret)
2713
		__mem_cgroup_cancel_charge(parent, nr_pages);
2714

2715
	if (nr_pages > 1)
2716
		compound_unlock_irqrestore(page, flags);
2717
put_back:
K
KAMEZAWA Hiroyuki 已提交
2718
	putback_lru_page(page);
2719
put:
2720
	put_page(page);
2721
out:
2722 2723 2724
	return ret;
}

2725 2726 2727 2728 2729 2730 2731
/*
 * 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,
2732
				gfp_t gfp_mask, enum charge_type ctype)
2733
{
2734
	struct mem_cgroup *memcg = NULL;
2735
	unsigned int nr_pages = 1;
2736
	struct page_cgroup *pc;
2737
	bool oom = true;
2738
	int ret;
A
Andrea Arcangeli 已提交
2739

A
Andrea Arcangeli 已提交
2740
	if (PageTransHuge(page)) {
2741
		nr_pages <<= compound_order(page);
A
Andrea Arcangeli 已提交
2742
		VM_BUG_ON(!PageTransHuge(page));
2743 2744 2745 2746 2747
		/*
		 * 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
2748
	}
2749 2750

	pc = lookup_page_cgroup(page);
2751
	BUG_ON(!pc); /* XXX: remove this and move pc lookup into commit */
2752

2753 2754
	ret = __mem_cgroup_try_charge(mm, gfp_mask, nr_pages, &memcg, oom);
	if (ret || !memcg)
2755 2756
		return ret;

2757
	__mem_cgroup_commit_charge(memcg, page, nr_pages, pc, ctype);
2758 2759 2760
	return 0;
}

2761 2762
int mem_cgroup_newpage_charge(struct page *page,
			      struct mm_struct *mm, gfp_t gfp_mask)
2763
{
2764
	if (mem_cgroup_disabled())
2765
		return 0;
2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776
	/*
	 * If already mapped, we don't have to account.
	 * If page cache, page->mapping has address_space.
	 * But page->mapping may have out-of-use anon_vma pointer,
	 * detecit it by PageAnon() check. newly-mapped-anon's page->mapping
	 * is NULL.
  	 */
	if (page_mapped(page) || (page->mapping && !PageAnon(page)))
		return 0;
	if (unlikely(!mm))
		mm = &init_mm;
2777
	return mem_cgroup_charge_common(page, mm, gfp_mask,
2778
				MEM_CGROUP_CHARGE_TYPE_MAPPED);
2779 2780
}

D
Daisuke Nishimura 已提交
2781 2782 2783 2784
static void
__mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr,
					enum charge_type ctype);

2785
static void
2786
__mem_cgroup_commit_charge_lrucare(struct page *page, struct mem_cgroup *memcg,
2787 2788 2789 2790 2791 2792 2793 2794 2795
					enum charge_type ctype)
{
	struct page_cgroup *pc = lookup_page_cgroup(page);
	/*
	 * In some case, SwapCache, FUSE(splice_buf->radixtree), the page
	 * is already on LRU. It means the page may on some other page_cgroup's
	 * LRU. Take care of it.
	 */
	mem_cgroup_lru_del_before_commit(page);
2796
	__mem_cgroup_commit_charge(memcg, page, 1, pc, ctype);
2797 2798 2799 2800
	mem_cgroup_lru_add_after_commit(page);
	return;
}

2801 2802
int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm,
				gfp_t gfp_mask)
2803
{
2804
	struct mem_cgroup *memcg = NULL;
2805 2806
	int ret;

2807
	if (mem_cgroup_disabled())
2808
		return 0;
2809 2810
	if (PageCompound(page))
		return 0;
2811

2812
	if (unlikely(!mm))
2813
		mm = &init_mm;
2814

2815
	if (page_is_file_cache(page)) {
2816 2817
		ret = __mem_cgroup_try_charge(mm, gfp_mask, 1, &memcg, true);
		if (ret || !memcg)
2818
			return ret;
2819

2820 2821 2822 2823 2824
		/*
		 * FUSE reuses pages without going through the final
		 * put that would remove them from the LRU list, make
		 * sure that they get relinked properly.
		 */
2825
		__mem_cgroup_commit_charge_lrucare(page, memcg,
2826 2827 2828
					MEM_CGROUP_CHARGE_TYPE_CACHE);
		return ret;
	}
D
Daisuke Nishimura 已提交
2829 2830
	/* shmem */
	if (PageSwapCache(page)) {
2831
		ret = mem_cgroup_try_charge_swapin(mm, page, gfp_mask, &memcg);
D
Daisuke Nishimura 已提交
2832
		if (!ret)
2833
			__mem_cgroup_commit_charge_swapin(page, memcg,
D
Daisuke Nishimura 已提交
2834 2835 2836
					MEM_CGROUP_CHARGE_TYPE_SHMEM);
	} else
		ret = mem_cgroup_charge_common(page, mm, gfp_mask,
2837
					MEM_CGROUP_CHARGE_TYPE_SHMEM);
2838 2839

	return ret;
2840 2841
}

2842 2843 2844
/*
 * While swap-in, try_charge -> commit or cancel, the page is locked.
 * And when try_charge() successfully returns, one refcnt to memcg without
2845
 * struct page_cgroup is acquired. This refcnt will be consumed by
2846 2847
 * "commit()" or removed by "cancel()"
 */
2848 2849 2850 2851
int mem_cgroup_try_charge_swapin(struct mm_struct *mm,
				 struct page *page,
				 gfp_t mask, struct mem_cgroup **ptr)
{
2852
	struct mem_cgroup *memcg;
2853
	int ret;
2854

2855 2856
	*ptr = NULL;

2857
	if (mem_cgroup_disabled())
2858 2859 2860 2861 2862 2863
		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
2864 2865 2866
	 * 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.
2867 2868
	 */
	if (!PageSwapCache(page))
H
Hugh Dickins 已提交
2869
		goto charge_cur_mm;
2870 2871
	memcg = try_get_mem_cgroup_from_page(page);
	if (!memcg)
2872
		goto charge_cur_mm;
2873
	*ptr = memcg;
2874
	ret = __mem_cgroup_try_charge(NULL, mask, 1, ptr, true);
2875
	css_put(&memcg->css);
2876
	return ret;
2877 2878 2879
charge_cur_mm:
	if (unlikely(!mm))
		mm = &init_mm;
2880
	return __mem_cgroup_try_charge(mm, mask, 1, ptr, true);
2881 2882
}

D
Daisuke Nishimura 已提交
2883 2884 2885
static void
__mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr,
					enum charge_type ctype)
2886
{
2887
	if (mem_cgroup_disabled())
2888 2889 2890
		return;
	if (!ptr)
		return;
2891
	cgroup_exclude_rmdir(&ptr->css);
2892 2893

	__mem_cgroup_commit_charge_lrucare(page, ptr, ctype);
2894 2895 2896
	/*
	 * Now swap is on-memory. This means this page may be
	 * counted both as mem and swap....double count.
2897 2898 2899
	 * 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.
2900
	 */
2901
	if (do_swap_account && PageSwapCache(page)) {
2902
		swp_entry_t ent = {.val = page_private(page)};
2903
		unsigned short id;
2904
		struct mem_cgroup *memcg;
2905 2906 2907 2908

		id = swap_cgroup_record(ent, 0);
		rcu_read_lock();
		memcg = mem_cgroup_lookup(id);
2909
		if (memcg) {
2910 2911 2912 2913
			/*
			 * This recorded memcg can be obsolete one. So, avoid
			 * calling css_tryget
			 */
2914
			if (!mem_cgroup_is_root(memcg))
2915
				res_counter_uncharge(&memcg->memsw, PAGE_SIZE);
2916
			mem_cgroup_swap_statistics(memcg, false);
2917 2918
			mem_cgroup_put(memcg);
		}
2919
		rcu_read_unlock();
2920
	}
2921 2922 2923 2924 2925 2926
	/*
	 * At swapin, we may charge account against cgroup which has no tasks.
	 * So, rmdir()->pre_destroy() can be called while we do this charge.
	 * In that case, we need to call pre_destroy() again. check it here.
	 */
	cgroup_release_and_wakeup_rmdir(&ptr->css);
2927 2928
}

D
Daisuke Nishimura 已提交
2929 2930 2931 2932 2933 2934
void mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr)
{
	__mem_cgroup_commit_charge_swapin(page, ptr,
					MEM_CGROUP_CHARGE_TYPE_MAPPED);
}

2935
void mem_cgroup_cancel_charge_swapin(struct mem_cgroup *memcg)
2936
{
2937
	if (mem_cgroup_disabled())
2938
		return;
2939
	if (!memcg)
2940
		return;
2941
	__mem_cgroup_cancel_charge(memcg, 1);
2942 2943
}

2944
static void mem_cgroup_do_uncharge(struct mem_cgroup *memcg,
2945 2946
				   unsigned int nr_pages,
				   const enum charge_type ctype)
2947 2948 2949
{
	struct memcg_batch_info *batch = NULL;
	bool uncharge_memsw = true;
2950

2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961
	/* 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)
2962
		batch->memcg = memcg;
2963 2964
	/*
	 * do_batch > 0 when unmapping pages or inode invalidate/truncate.
L
Lucas De Marchi 已提交
2965
	 * In those cases, all pages freed continuously can be expected to be in
2966 2967 2968 2969 2970 2971 2972 2973
	 * 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;

2974
	if (nr_pages > 1)
A
Andrea Arcangeli 已提交
2975 2976
		goto direct_uncharge;

2977 2978 2979 2980 2981
	/*
	 * 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.
	 */
2982
	if (batch->memcg != memcg)
2983 2984
		goto direct_uncharge;
	/* remember freed charge and uncharge it later */
2985
	batch->nr_pages++;
2986
	if (uncharge_memsw)
2987
		batch->memsw_nr_pages++;
2988 2989
	return;
direct_uncharge:
2990
	res_counter_uncharge(&memcg->res, nr_pages * PAGE_SIZE);
2991
	if (uncharge_memsw)
2992 2993 2994
		res_counter_uncharge(&memcg->memsw, nr_pages * PAGE_SIZE);
	if (unlikely(batch->memcg != memcg))
		memcg_oom_recover(memcg);
2995 2996
	return;
}
2997

2998
/*
2999
 * uncharge if !page_mapped(page)
3000
 */
3001
static struct mem_cgroup *
3002
__mem_cgroup_uncharge_common(struct page *page, enum charge_type ctype)
3003
{
3004
	struct mem_cgroup *memcg = NULL;
3005 3006
	unsigned int nr_pages = 1;
	struct page_cgroup *pc;
3007

3008
	if (mem_cgroup_disabled())
3009
		return NULL;
3010

K
KAMEZAWA Hiroyuki 已提交
3011
	if (PageSwapCache(page))
3012
		return NULL;
K
KAMEZAWA Hiroyuki 已提交
3013

A
Andrea Arcangeli 已提交
3014
	if (PageTransHuge(page)) {
3015
		nr_pages <<= compound_order(page);
A
Andrea Arcangeli 已提交
3016 3017
		VM_BUG_ON(!PageTransHuge(page));
	}
3018
	/*
3019
	 * Check if our page_cgroup is valid
3020
	 */
3021 3022
	pc = lookup_page_cgroup(page);
	if (unlikely(!pc || !PageCgroupUsed(pc)))
3023
		return NULL;
3024

3025
	lock_page_cgroup(pc);
K
KAMEZAWA Hiroyuki 已提交
3026

3027
	memcg = pc->mem_cgroup;
3028

K
KAMEZAWA Hiroyuki 已提交
3029 3030 3031 3032 3033
	if (!PageCgroupUsed(pc))
		goto unlock_out;

	switch (ctype) {
	case MEM_CGROUP_CHARGE_TYPE_MAPPED:
K
KAMEZAWA Hiroyuki 已提交
3034
	case MEM_CGROUP_CHARGE_TYPE_DROP:
3035 3036
		/* See mem_cgroup_prepare_migration() */
		if (page_mapped(page) || PageCgroupMigration(pc))
K
KAMEZAWA Hiroyuki 已提交
3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047
			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;
3048
	}
K
KAMEZAWA Hiroyuki 已提交
3049

3050
	mem_cgroup_charge_statistics(memcg, PageCgroupCache(pc), -nr_pages);
K
KAMEZAWA Hiroyuki 已提交
3051

3052
	ClearPageCgroupUsed(pc);
3053 3054 3055 3056 3057 3058
	/*
	 * 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.
	 */
3059

3060
	unlock_page_cgroup(pc);
K
KAMEZAWA Hiroyuki 已提交
3061
	/*
3062
	 * even after unlock, we have memcg->res.usage here and this memcg
K
KAMEZAWA Hiroyuki 已提交
3063 3064
	 * will never be freed.
	 */
3065
	memcg_check_events(memcg, page);
K
KAMEZAWA Hiroyuki 已提交
3066
	if (do_swap_account && ctype == MEM_CGROUP_CHARGE_TYPE_SWAPOUT) {
3067 3068
		mem_cgroup_swap_statistics(memcg, true);
		mem_cgroup_get(memcg);
K
KAMEZAWA Hiroyuki 已提交
3069
	}
3070 3071
	if (!mem_cgroup_is_root(memcg))
		mem_cgroup_do_uncharge(memcg, nr_pages, ctype);
3072

3073
	return memcg;
K
KAMEZAWA Hiroyuki 已提交
3074 3075 3076

unlock_out:
	unlock_page_cgroup(pc);
3077
	return NULL;
3078 3079
}

3080 3081
void mem_cgroup_uncharge_page(struct page *page)
{
3082 3083 3084 3085 3086
	/* early check. */
	if (page_mapped(page))
		return;
	if (page->mapping && !PageAnon(page))
		return;
3087 3088 3089 3090 3091 3092
	__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));
3093
	VM_BUG_ON(page->mapping);
3094 3095 3096
	__mem_cgroup_uncharge_common(page, MEM_CGROUP_CHARGE_TYPE_CACHE);
}

3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110
/*
 * 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;
3111 3112
		current->memcg_batch.nr_pages = 0;
		current->memcg_batch.memsw_nr_pages = 0;
3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132
	}
}

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.
	 */
3133 3134 3135 3136 3137 3138
	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);
3139
	memcg_oom_recover(batch->memcg);
3140 3141 3142 3143
	/* forget this pointer (for sanity check) */
	batch->memcg = NULL;
}

3144
#ifdef CONFIG_SWAP
3145
/*
3146
 * called after __delete_from_swap_cache() and drop "page" account.
3147 3148
 * memcg information is recorded to swap_cgroup of "ent"
 */
K
KAMEZAWA Hiroyuki 已提交
3149 3150
void
mem_cgroup_uncharge_swapcache(struct page *page, swp_entry_t ent, bool swapout)
3151 3152
{
	struct mem_cgroup *memcg;
K
KAMEZAWA Hiroyuki 已提交
3153 3154 3155 3156 3157 3158
	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);
3159

K
KAMEZAWA Hiroyuki 已提交
3160 3161 3162 3163 3164
	/*
	 * record memcg information,  if swapout && memcg != NULL,
	 * mem_cgroup_get() was called in uncharge().
	 */
	if (do_swap_account && swapout && memcg)
3165
		swap_cgroup_record(ent, css_id(&memcg->css));
3166
}
3167
#endif
3168 3169 3170 3171 3172 3173 3174

#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
3175
{
3176
	struct mem_cgroup *memcg;
3177
	unsigned short id;
3178 3179 3180 3181

	if (!do_swap_account)
		return;

3182 3183 3184
	id = swap_cgroup_record(ent, 0);
	rcu_read_lock();
	memcg = mem_cgroup_lookup(id);
3185
	if (memcg) {
3186 3187 3188 3189
		/*
		 * We uncharge this because swap is freed.
		 * This memcg can be obsolete one. We avoid calling css_tryget
		 */
3190
		if (!mem_cgroup_is_root(memcg))
3191
			res_counter_uncharge(&memcg->memsw, PAGE_SIZE);
3192
		mem_cgroup_swap_statistics(memcg, false);
3193 3194
		mem_cgroup_put(memcg);
	}
3195
	rcu_read_unlock();
K
KAMEZAWA Hiroyuki 已提交
3196
}
3197 3198 3199 3200 3201 3202

/**
 * 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
3203
 * @need_fixup: whether we should fixup res_counters and refcounts.
3204 3205 3206 3207 3208 3209 3210 3211 3212 3213
 *
 * 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,
3214
		struct mem_cgroup *from, struct mem_cgroup *to, bool need_fixup)
3215 3216 3217 3218 3219 3220 3221 3222
{
	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);
3223
		mem_cgroup_swap_statistics(to, true);
3224
		/*
3225 3226 3227 3228 3229 3230
		 * 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.
3231 3232
		 */
		mem_cgroup_get(to);
3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243
		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);
		}
3244 3245 3246 3247 3248 3249
		return 0;
	}
	return -EINVAL;
}
#else
static inline int mem_cgroup_move_swap_account(swp_entry_t entry,
3250
		struct mem_cgroup *from, struct mem_cgroup *to, bool need_fixup)
3251 3252 3253
{
	return -EINVAL;
}
3254
#endif
K
KAMEZAWA Hiroyuki 已提交
3255

3256
/*
3257 3258
 * Before starting migration, account PAGE_SIZE to mem_cgroup that the old
 * page belongs to.
3259
 */
3260
int mem_cgroup_prepare_migration(struct page *page,
3261
	struct page *newpage, struct mem_cgroup **ptr, gfp_t gfp_mask)
3262
{
3263
	struct mem_cgroup *memcg = NULL;
3264
	struct page_cgroup *pc;
3265
	enum charge_type ctype;
3266
	int ret = 0;
3267

3268 3269
	*ptr = NULL;

A
Andrea Arcangeli 已提交
3270
	VM_BUG_ON(PageTransHuge(page));
3271
	if (mem_cgroup_disabled())
3272 3273
		return 0;

3274 3275 3276
	pc = lookup_page_cgroup(page);
	lock_page_cgroup(pc);
	if (PageCgroupUsed(pc)) {
3277 3278
		memcg = pc->mem_cgroup;
		css_get(&memcg->css);
3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309
		/*
		 * 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);
3310
	}
3311
	unlock_page_cgroup(pc);
3312 3313 3314 3315
	/*
	 * If the page is not charged at this point,
	 * we return here.
	 */
3316
	if (!memcg)
3317
		return 0;
3318

3319
	*ptr = memcg;
3320
	ret = __mem_cgroup_try_charge(NULL, gfp_mask, 1, ptr, false);
3321
	css_put(&memcg->css);/* drop extra refcnt */
3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332
	if (ret || *ptr == NULL) {
		if (PageAnon(page)) {
			lock_page_cgroup(pc);
			ClearPageCgroupMigration(pc);
			unlock_page_cgroup(pc);
			/*
			 * The old page may be fully unmapped while we kept it.
			 */
			mem_cgroup_uncharge_page(page);
		}
		return -ENOMEM;
3333
	}
3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346
	/*
	 * 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;
3347
	__mem_cgroup_commit_charge(memcg, page, 1, pc, ctype);
3348
	return ret;
3349
}
3350

3351
/* remove redundant charge if migration failed*/
3352
void mem_cgroup_end_migration(struct mem_cgroup *memcg,
3353
	struct page *oldpage, struct page *newpage, bool migration_ok)
3354
{
3355
	struct page *used, *unused;
3356 3357
	struct page_cgroup *pc;

3358
	if (!memcg)
3359
		return;
3360
	/* blocks rmdir() */
3361
	cgroup_exclude_rmdir(&memcg->css);
3362
	if (!migration_ok) {
3363 3364
		used = oldpage;
		unused = newpage;
3365
	} else {
3366
		used = newpage;
3367 3368
		unused = oldpage;
	}
3369
	/*
3370 3371 3372
	 * 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.
3373
	 */
3374 3375 3376 3377
	pc = lookup_page_cgroup(oldpage);
	lock_page_cgroup(pc);
	ClearPageCgroupMigration(pc);
	unlock_page_cgroup(pc);
3378

3379 3380
	__mem_cgroup_uncharge_common(unused, MEM_CGROUP_CHARGE_TYPE_FORCE);

3381
	/*
3382 3383 3384 3385 3386 3387
	 * 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)
3388
	 */
3389 3390
	if (PageAnon(used))
		mem_cgroup_uncharge_page(used);
3391
	/*
3392 3393
	 * At migration, we may charge account against cgroup which has no
	 * tasks.
3394 3395 3396
	 * 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.
	 */
3397
	cgroup_release_and_wakeup_rmdir(&memcg->css);
3398
}
3399

3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443
/*
 * At replace page cache, newpage is not under any memcg but it's on
 * LRU. So, this function doesn't touch res_counter but handles LRU
 * in correct way. Both pages are locked so we cannot race with uncharge.
 */
void mem_cgroup_replace_page_cache(struct page *oldpage,
				  struct page *newpage)
{
	struct mem_cgroup *memcg;
	struct page_cgroup *pc;
	struct zone *zone;
	enum charge_type type = MEM_CGROUP_CHARGE_TYPE_CACHE;
	unsigned long flags;

	if (mem_cgroup_disabled())
		return;

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

	if (PageSwapBacked(oldpage))
		type = MEM_CGROUP_CHARGE_TYPE_SHMEM;

	zone = page_zone(newpage);
	pc = lookup_page_cgroup(newpage);
	/*
	 * Even if newpage->mapping was NULL before starting replacement,
	 * the newpage may be on LRU(or pagevec for LRU) already. We lock
	 * LRU while we overwrite pc->mem_cgroup.
	 */
	spin_lock_irqsave(&zone->lru_lock, flags);
	if (PageLRU(newpage))
		del_page_from_lru_list(zone, newpage, page_lru(newpage));
	__mem_cgroup_commit_charge(memcg, newpage, 1, pc, type);
	if (PageLRU(newpage))
		add_page_to_lru_list(zone, newpage, page_lru(newpage));
	spin_unlock_irqrestore(&zone->lru_lock, flags);
}

3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489
#ifdef CONFIG_DEBUG_VM
static struct page_cgroup *lookup_page_cgroup_used(struct page *page)
{
	struct page_cgroup *pc;

	pc = lookup_page_cgroup(page);
	if (likely(pc) && PageCgroupUsed(pc))
		return pc;
	return NULL;
}

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

	return lookup_page_cgroup_used(page) != NULL;
}

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

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

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

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

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

3490 3491
static DEFINE_MUTEX(set_limit_mutex);

3492
static int mem_cgroup_resize_limit(struct mem_cgroup *memcg,
3493
				unsigned long long val)
3494
{
3495
	int retry_count;
3496
	u64 memswlimit, memlimit;
3497
	int ret = 0;
3498 3499
	int children = mem_cgroup_count_children(memcg);
	u64 curusage, oldusage;
3500
	int enlarge;
3501 3502 3503 3504 3505 3506 3507 3508 3509

	/*
	 * 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);
3510

3511
	enlarge = 0;
3512
	while (retry_count) {
3513 3514 3515 3516
		if (signal_pending(current)) {
			ret = -EINTR;
			break;
		}
3517 3518 3519
		/*
		 * Rather than hide all in some function, I do this in
		 * open coded manner. You see what this really does.
3520
		 * We have to guarantee memcg->res.limit < memcg->memsw.limit.
3521 3522 3523 3524 3525 3526
		 */
		mutex_lock(&set_limit_mutex);
		memswlimit = res_counter_read_u64(&memcg->memsw, RES_LIMIT);
		if (memswlimit < val) {
			ret = -EINVAL;
			mutex_unlock(&set_limit_mutex);
3527 3528
			break;
		}
3529 3530 3531 3532 3533

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

3534
		ret = res_counter_set_limit(&memcg->res, val);
3535 3536 3537 3538 3539 3540
		if (!ret) {
			if (memswlimit == val)
				memcg->memsw_is_minimum = true;
			else
				memcg->memsw_is_minimum = false;
		}
3541 3542 3543 3544 3545
		mutex_unlock(&set_limit_mutex);

		if (!ret)
			break;

3546 3547
		mem_cgroup_reclaim(memcg, GFP_KERNEL,
				   MEM_CGROUP_RECLAIM_SHRINK);
3548 3549 3550 3551 3552 3553
		curusage = res_counter_read_u64(&memcg->res, RES_USAGE);
		/* Usage is reduced ? */
  		if (curusage >= oldusage)
			retry_count--;
		else
			oldusage = curusage;
3554
	}
3555 3556
	if (!ret && enlarge)
		memcg_oom_recover(memcg);
3557

3558 3559 3560
	return ret;
}

L
Li Zefan 已提交
3561 3562
static int mem_cgroup_resize_memsw_limit(struct mem_cgroup *memcg,
					unsigned long long val)
3563
{
3564
	int retry_count;
3565
	u64 memlimit, memswlimit, oldusage, curusage;
3566 3567
	int children = mem_cgroup_count_children(memcg);
	int ret = -EBUSY;
3568
	int enlarge = 0;
3569

3570 3571 3572
	/* see mem_cgroup_resize_res_limit */
 	retry_count = children * MEM_CGROUP_RECLAIM_RETRIES;
	oldusage = res_counter_read_u64(&memcg->memsw, RES_USAGE);
3573 3574 3575 3576 3577 3578 3579 3580
	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.
3581
		 * We have to guarantee memcg->res.limit < memcg->memsw.limit.
3582 3583 3584 3585 3586 3587 3588 3589
		 */
		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;
		}
3590 3591 3592
		memswlimit = res_counter_read_u64(&memcg->memsw, RES_LIMIT);
		if (memswlimit < val)
			enlarge = 1;
3593
		ret = res_counter_set_limit(&memcg->memsw, val);
3594 3595 3596 3597 3598 3599
		if (!ret) {
			if (memlimit == val)
				memcg->memsw_is_minimum = true;
			else
				memcg->memsw_is_minimum = false;
		}
3600 3601 3602 3603 3604
		mutex_unlock(&set_limit_mutex);

		if (!ret)
			break;

3605 3606 3607
		mem_cgroup_reclaim(memcg, GFP_KERNEL,
				   MEM_CGROUP_RECLAIM_NOSWAP |
				   MEM_CGROUP_RECLAIM_SHRINK);
3608
		curusage = res_counter_read_u64(&memcg->memsw, RES_USAGE);
3609
		/* Usage is reduced ? */
3610
		if (curusage >= oldusage)
3611
			retry_count--;
3612 3613
		else
			oldusage = curusage;
3614
	}
3615 3616
	if (!ret && enlarge)
		memcg_oom_recover(memcg);
3617 3618 3619
	return ret;
}

3620
unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
3621 3622
					    gfp_t gfp_mask,
					    unsigned long *total_scanned)
3623 3624 3625 3626 3627 3628
{
	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;
3629
	unsigned long long excess;
3630
	unsigned long nr_scanned;
3631 3632 3633 3634

	if (order > 0)
		return 0;

3635
	mctz = soft_limit_tree_node_zone(zone_to_nid(zone), zone_idx(zone));
3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648
	/*
	 * 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;

3649
		nr_scanned = 0;
3650 3651
		reclaimed = mem_cgroup_soft_reclaim(mz->mem, zone,
						    gfp_mask, &nr_scanned);
3652
		nr_reclaimed += reclaimed;
3653
		*total_scanned += nr_scanned;
3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675
		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);
3676
				if (next_mz == mz)
3677
					css_put(&next_mz->mem->css);
3678
				else /* next_mz == NULL or other memcg */
3679 3680 3681 3682
					break;
			} while (1);
		}
		__mem_cgroup_remove_exceeded(mz->mem, mz, mctz);
3683
		excess = res_counter_soft_limit_excess(&mz->mem->res);
3684 3685 3686 3687 3688 3689 3690 3691
		/*
		 * 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.
		 */
3692 3693
		/* If excess == 0, no tree ops */
		__mem_cgroup_insert_exceeded(mz->mem, mz, mctz, excess);
3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711
		spin_unlock(&mctz->lock);
		css_put(&mz->mem->css);
		loop++;
		/*
		 * Could not reclaim anything and there are no more
		 * mem cgroups to try or we seem to be looping without
		 * reclaiming anything.
		 */
		if (!nr_reclaimed &&
			(next_mz == NULL ||
			loop > MEM_CGROUP_MAX_SOFT_LIMIT_RECLAIM_LOOPS))
			break;
	} while (!nr_reclaimed);
	if (next_mz)
		css_put(&next_mz->mem->css);
	return nr_reclaimed;
}

3712 3713 3714 3715
/*
 * This routine traverse page_cgroup in given list and drop them all.
 * *And* this routine doesn't reclaim page itself, just removes page_cgroup.
 */
3716
static int mem_cgroup_force_empty_list(struct mem_cgroup *memcg,
K
KAMEZAWA Hiroyuki 已提交
3717
				int node, int zid, enum lru_list lru)
3718
{
K
KAMEZAWA Hiroyuki 已提交
3719 3720
	struct mem_cgroup_per_zone *mz;
	unsigned long flags, loop;
3721
	struct list_head *list;
3722 3723
	struct page *busy;
	struct zone *zone;
3724
	int ret = 0;
3725

K
KAMEZAWA Hiroyuki 已提交
3726
	zone = &NODE_DATA(node)->node_zones[zid];
3727
	mz = mem_cgroup_zoneinfo(memcg, node, zid);
3728
	list = &mz->lruvec.lists[lru];
3729

3730 3731 3732 3733 3734
	loop = MEM_CGROUP_ZSTAT(mz, lru);
	/* give some margin against EBUSY etc...*/
	loop += 256;
	busy = NULL;
	while (loop--) {
3735
		struct page_cgroup *pc;
3736 3737
		struct page *page;

3738
		ret = 0;
K
KAMEZAWA Hiroyuki 已提交
3739
		spin_lock_irqsave(&zone->lru_lock, flags);
3740
		if (list_empty(list)) {
K
KAMEZAWA Hiroyuki 已提交
3741
			spin_unlock_irqrestore(&zone->lru_lock, flags);
3742
			break;
3743
		}
3744 3745 3746
		page = list_entry(list->prev, struct page, lru);
		if (busy == page) {
			list_move(&page->lru, list);
3747
			busy = NULL;
K
KAMEZAWA Hiroyuki 已提交
3748
			spin_unlock_irqrestore(&zone->lru_lock, flags);
3749 3750
			continue;
		}
K
KAMEZAWA Hiroyuki 已提交
3751
		spin_unlock_irqrestore(&zone->lru_lock, flags);
3752

3753
		pc = lookup_page_cgroup(page);
3754

3755
		ret = mem_cgroup_move_parent(page, pc, memcg, GFP_KERNEL);
3756
		if (ret == -ENOMEM)
3757
			break;
3758 3759 3760

		if (ret == -EBUSY || ret == -EINVAL) {
			/* found lock contention or "pc" is obsolete. */
3761
			busy = page;
3762 3763 3764
			cond_resched();
		} else
			busy = NULL;
3765
	}
K
KAMEZAWA Hiroyuki 已提交
3766

3767 3768 3769
	if (!ret && !list_empty(list))
		return -EBUSY;
	return ret;
3770 3771 3772 3773 3774 3775
}

/*
 * make mem_cgroup's charge to be 0 if there is no task.
 * This enables deleting this mem_cgroup.
 */
3776
static int mem_cgroup_force_empty(struct mem_cgroup *memcg, bool free_all)
3777
{
3778 3779 3780
	int ret;
	int node, zid, shrink;
	int nr_retries = MEM_CGROUP_RECLAIM_RETRIES;
3781
	struct cgroup *cgrp = memcg->css.cgroup;
3782

3783
	css_get(&memcg->css);
3784 3785

	shrink = 0;
3786 3787 3788
	/* should free all ? */
	if (free_all)
		goto try_to_free;
3789
move_account:
3790
	do {
3791
		ret = -EBUSY;
3792 3793 3794 3795
		if (cgroup_task_count(cgrp) || !list_empty(&cgrp->children))
			goto out;
		ret = -EINTR;
		if (signal_pending(current))
3796
			goto out;
3797 3798
		/* This is for making all *used* pages to be on LRU. */
		lru_add_drain_all();
3799
		drain_all_stock_sync(memcg);
3800
		ret = 0;
3801
		mem_cgroup_start_move(memcg);
3802
		for_each_node_state(node, N_HIGH_MEMORY) {
3803
			for (zid = 0; !ret && zid < MAX_NR_ZONES; zid++) {
3804
				enum lru_list l;
3805
				for_each_lru(l) {
3806
					ret = mem_cgroup_force_empty_list(memcg,
K
KAMEZAWA Hiroyuki 已提交
3807
							node, zid, l);
3808 3809 3810
					if (ret)
						break;
				}
3811
			}
3812 3813 3814
			if (ret)
				break;
		}
3815 3816
		mem_cgroup_end_move(memcg);
		memcg_oom_recover(memcg);
3817 3818 3819
		/* it seems parent cgroup doesn't have enough mem */
		if (ret == -ENOMEM)
			goto try_to_free;
3820
		cond_resched();
3821
	/* "ret" should also be checked to ensure all lists are empty. */
3822
	} while (memcg->res.usage > 0 || ret);
3823
out:
3824
	css_put(&memcg->css);
3825
	return ret;
3826 3827

try_to_free:
3828 3829
	/* returns EBUSY if there is a task or if we come here twice. */
	if (cgroup_task_count(cgrp) || !list_empty(&cgrp->children) || shrink) {
3830 3831 3832
		ret = -EBUSY;
		goto out;
	}
3833 3834
	/* we call try-to-free pages for make this cgroup empty */
	lru_add_drain_all();
3835 3836
	/* try to free all pages in this cgroup */
	shrink = 1;
3837
	while (nr_retries && memcg->res.usage > 0) {
3838
		int progress;
3839 3840 3841 3842 3843

		if (signal_pending(current)) {
			ret = -EINTR;
			goto out;
		}
3844
		progress = try_to_free_mem_cgroup_pages(memcg, GFP_KERNEL,
3845
						false);
3846
		if (!progress) {
3847
			nr_retries--;
3848
			/* maybe some writeback is necessary */
3849
			congestion_wait(BLK_RW_ASYNC, HZ/10);
3850
		}
3851 3852

	}
K
KAMEZAWA Hiroyuki 已提交
3853
	lru_add_drain();
3854
	/* try move_account...there may be some *locked* pages. */
3855
	goto move_account;
3856 3857
}

3858 3859 3860 3861 3862 3863
int mem_cgroup_force_empty_write(struct cgroup *cont, unsigned int event)
{
	return mem_cgroup_force_empty(mem_cgroup_from_cont(cont), true);
}


3864 3865 3866 3867 3868 3869 3870 3871 3872
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;
3873
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
3874
	struct cgroup *parent = cont->parent;
3875
	struct mem_cgroup *parent_memcg = NULL;
3876 3877

	if (parent)
3878
		parent_memcg = mem_cgroup_from_cont(parent);
3879 3880 3881

	cgroup_lock();
	/*
3882
	 * If parent's use_hierarchy is set, we can't make any modifications
3883 3884 3885 3886 3887 3888
	 * 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.
	 */
3889
	if ((!parent_memcg || !parent_memcg->use_hierarchy) &&
3890 3891
				(val == 1 || val == 0)) {
		if (list_empty(&cont->children))
3892
			memcg->use_hierarchy = val;
3893 3894 3895 3896 3897 3898 3899 3900 3901
		else
			retval = -EBUSY;
	} else
		retval = -EINVAL;
	cgroup_unlock();

	return retval;
}

3902

3903
static unsigned long mem_cgroup_recursive_stat(struct mem_cgroup *memcg,
3904
					       enum mem_cgroup_stat_index idx)
3905
{
K
KAMEZAWA Hiroyuki 已提交
3906
	struct mem_cgroup *iter;
3907
	long val = 0;
3908

3909
	/* Per-cpu values can be negative, use a signed accumulator */
3910
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
3911 3912 3913 3914 3915
		val += mem_cgroup_read_stat(iter, idx);

	if (val < 0) /* race ? */
		val = 0;
	return val;
3916 3917
}

3918
static inline u64 mem_cgroup_usage(struct mem_cgroup *memcg, bool swap)
3919
{
K
KAMEZAWA Hiroyuki 已提交
3920
	u64 val;
3921

3922
	if (!mem_cgroup_is_root(memcg)) {
3923
		if (!swap)
3924
			return res_counter_read_u64(&memcg->res, RES_USAGE);
3925
		else
3926
			return res_counter_read_u64(&memcg->memsw, RES_USAGE);
3927 3928
	}

3929 3930
	val = mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_CACHE);
	val += mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_RSS);
3931

K
KAMEZAWA Hiroyuki 已提交
3932
	if (swap)
3933
		val += mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_SWAPOUT);
3934 3935 3936 3937

	return val << PAGE_SHIFT;
}

3938
static u64 mem_cgroup_read(struct cgroup *cont, struct cftype *cft)
B
Balbir Singh 已提交
3939
{
3940
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
3941
	u64 val;
3942 3943 3944 3945 3946 3947
	int type, name;

	type = MEMFILE_TYPE(cft->private);
	name = MEMFILE_ATTR(cft->private);
	switch (type) {
	case _MEM:
3948
		if (name == RES_USAGE)
3949
			val = mem_cgroup_usage(memcg, false);
3950
		else
3951
			val = res_counter_read_u64(&memcg->res, name);
3952 3953
		break;
	case _MEMSWAP:
3954
		if (name == RES_USAGE)
3955
			val = mem_cgroup_usage(memcg, true);
3956
		else
3957
			val = res_counter_read_u64(&memcg->memsw, name);
3958 3959 3960 3961 3962 3963
		break;
	default:
		BUG();
		break;
	}
	return val;
B
Balbir Singh 已提交
3964
}
3965 3966 3967 3968
/*
 * The user of this function is...
 * RES_LIMIT.
 */
3969 3970
static int mem_cgroup_write(struct cgroup *cont, struct cftype *cft,
			    const char *buffer)
B
Balbir Singh 已提交
3971
{
3972
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
3973
	int type, name;
3974 3975 3976
	unsigned long long val;
	int ret;

3977 3978 3979
	type = MEMFILE_TYPE(cft->private);
	name = MEMFILE_ATTR(cft->private);
	switch (name) {
3980
	case RES_LIMIT:
3981 3982 3983 3984
		if (mem_cgroup_is_root(memcg)) { /* Can't set limit on root */
			ret = -EINVAL;
			break;
		}
3985 3986
		/* This function does all necessary parse...reuse it */
		ret = res_counter_memparse_write_strategy(buffer, &val);
3987 3988 3989
		if (ret)
			break;
		if (type == _MEM)
3990
			ret = mem_cgroup_resize_limit(memcg, val);
3991 3992
		else
			ret = mem_cgroup_resize_memsw_limit(memcg, val);
3993
		break;
3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007
	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;
4008 4009 4010 4011 4012
	default:
		ret = -EINVAL; /* should be BUG() ? */
		break;
	}
	return ret;
B
Balbir Singh 已提交
4013 4014
}

4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042
static void memcg_get_hierarchical_limit(struct mem_cgroup *memcg,
		unsigned long long *mem_limit, unsigned long long *memsw_limit)
{
	struct cgroup *cgroup;
	unsigned long long min_limit, min_memsw_limit, tmp;

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

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

4043
static int mem_cgroup_reset(struct cgroup *cont, unsigned int event)
4044
{
4045
	struct mem_cgroup *memcg;
4046
	int type, name;
4047

4048
	memcg = mem_cgroup_from_cont(cont);
4049 4050 4051
	type = MEMFILE_TYPE(event);
	name = MEMFILE_ATTR(event);
	switch (name) {
4052
	case RES_MAX_USAGE:
4053
		if (type == _MEM)
4054
			res_counter_reset_max(&memcg->res);
4055
		else
4056
			res_counter_reset_max(&memcg->memsw);
4057 4058
		break;
	case RES_FAILCNT:
4059
		if (type == _MEM)
4060
			res_counter_reset_failcnt(&memcg->res);
4061
		else
4062
			res_counter_reset_failcnt(&memcg->memsw);
4063 4064
		break;
	}
4065

4066
	return 0;
4067 4068
}

4069 4070 4071 4072 4073 4074
static u64 mem_cgroup_move_charge_read(struct cgroup *cgrp,
					struct cftype *cft)
{
	return mem_cgroup_from_cont(cgrp)->move_charge_at_immigrate;
}

4075
#ifdef CONFIG_MMU
4076 4077 4078
static int mem_cgroup_move_charge_write(struct cgroup *cgrp,
					struct cftype *cft, u64 val)
{
4079
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4080 4081 4082 4083 4084 4085 4086 4087 4088

	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();
4089
	memcg->move_charge_at_immigrate = val;
4090 4091 4092 4093
	cgroup_unlock();

	return 0;
}
4094 4095 4096 4097 4098 4099 4100
#else
static int mem_cgroup_move_charge_write(struct cgroup *cgrp,
					struct cftype *cft, u64 val)
{
	return -ENOSYS;
}
#endif
4101

K
KAMEZAWA Hiroyuki 已提交
4102 4103 4104 4105 4106

/* For read statistics */
enum {
	MCS_CACHE,
	MCS_RSS,
4107
	MCS_FILE_MAPPED,
K
KAMEZAWA Hiroyuki 已提交
4108 4109
	MCS_PGPGIN,
	MCS_PGPGOUT,
4110
	MCS_SWAP,
4111 4112
	MCS_PGFAULT,
	MCS_PGMAJFAULT,
K
KAMEZAWA Hiroyuki 已提交
4113 4114 4115 4116 4117 4118 4119 4120 4121 4122
	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];
4123 4124
};

K
KAMEZAWA Hiroyuki 已提交
4125 4126 4127 4128 4129 4130
struct {
	char *local_name;
	char *total_name;
} memcg_stat_strings[NR_MCS_STAT] = {
	{"cache", "total_cache"},
	{"rss", "total_rss"},
4131
	{"mapped_file", "total_mapped_file"},
K
KAMEZAWA Hiroyuki 已提交
4132 4133
	{"pgpgin", "total_pgpgin"},
	{"pgpgout", "total_pgpgout"},
4134
	{"swap", "total_swap"},
4135 4136
	{"pgfault", "total_pgfault"},
	{"pgmajfault", "total_pgmajfault"},
K
KAMEZAWA Hiroyuki 已提交
4137 4138 4139 4140 4141 4142 4143 4144
	{"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
4145
static void
4146
mem_cgroup_get_local_stat(struct mem_cgroup *memcg, struct mcs_total_stat *s)
K
KAMEZAWA Hiroyuki 已提交
4147 4148 4149 4150
{
	s64 val;

	/* per cpu stat */
4151
	val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_CACHE);
K
KAMEZAWA Hiroyuki 已提交
4152
	s->stat[MCS_CACHE] += val * PAGE_SIZE;
4153
	val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_RSS);
K
KAMEZAWA Hiroyuki 已提交
4154
	s->stat[MCS_RSS] += val * PAGE_SIZE;
4155
	val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_FILE_MAPPED);
4156
	s->stat[MCS_FILE_MAPPED] += val * PAGE_SIZE;
4157
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGPGIN);
K
KAMEZAWA Hiroyuki 已提交
4158
	s->stat[MCS_PGPGIN] += val;
4159
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGPGOUT);
K
KAMEZAWA Hiroyuki 已提交
4160
	s->stat[MCS_PGPGOUT] += val;
4161
	if (do_swap_account) {
4162
		val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_SWAPOUT);
4163 4164
		s->stat[MCS_SWAP] += val * PAGE_SIZE;
	}
4165
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGFAULT);
4166
	s->stat[MCS_PGFAULT] += val;
4167
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGMAJFAULT);
4168
	s->stat[MCS_PGMAJFAULT] += val;
K
KAMEZAWA Hiroyuki 已提交
4169 4170

	/* per zone stat */
4171
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_INACTIVE_ANON));
K
KAMEZAWA Hiroyuki 已提交
4172
	s->stat[MCS_INACTIVE_ANON] += val * PAGE_SIZE;
4173
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_ACTIVE_ANON));
K
KAMEZAWA Hiroyuki 已提交
4174
	s->stat[MCS_ACTIVE_ANON] += val * PAGE_SIZE;
4175
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_INACTIVE_FILE));
K
KAMEZAWA Hiroyuki 已提交
4176
	s->stat[MCS_INACTIVE_FILE] += val * PAGE_SIZE;
4177
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_ACTIVE_FILE));
K
KAMEZAWA Hiroyuki 已提交
4178
	s->stat[MCS_ACTIVE_FILE] += val * PAGE_SIZE;
4179
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_UNEVICTABLE));
K
KAMEZAWA Hiroyuki 已提交
4180 4181 4182 4183
	s->stat[MCS_UNEVICTABLE] += val * PAGE_SIZE;
}

static void
4184
mem_cgroup_get_total_stat(struct mem_cgroup *memcg, struct mcs_total_stat *s)
K
KAMEZAWA Hiroyuki 已提交
4185
{
K
KAMEZAWA Hiroyuki 已提交
4186 4187
	struct mem_cgroup *iter;

4188
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
4189
		mem_cgroup_get_local_stat(iter, s);
K
KAMEZAWA Hiroyuki 已提交
4190 4191
}

4192 4193 4194 4195 4196 4197 4198 4199 4200
#ifdef CONFIG_NUMA
static int mem_control_numa_stat_show(struct seq_file *m, void *arg)
{
	int nid;
	unsigned long total_nr, file_nr, anon_nr, unevictable_nr;
	unsigned long node_nr;
	struct cgroup *cont = m->private;
	struct mem_cgroup *mem_cont = mem_cgroup_from_cont(cont);

4201
	total_nr = mem_cgroup_nr_lru_pages(mem_cont, LRU_ALL);
4202 4203
	seq_printf(m, "total=%lu", total_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4204
		node_nr = mem_cgroup_node_nr_lru_pages(mem_cont, nid, LRU_ALL);
4205 4206 4207 4208
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');

4209
	file_nr = mem_cgroup_nr_lru_pages(mem_cont, LRU_ALL_FILE);
4210 4211
	seq_printf(m, "file=%lu", file_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4212 4213
		node_nr = mem_cgroup_node_nr_lru_pages(mem_cont, nid,
				LRU_ALL_FILE);
4214 4215 4216 4217
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');

4218
	anon_nr = mem_cgroup_nr_lru_pages(mem_cont, LRU_ALL_ANON);
4219 4220
	seq_printf(m, "anon=%lu", anon_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4221 4222
		node_nr = mem_cgroup_node_nr_lru_pages(mem_cont, nid,
				LRU_ALL_ANON);
4223 4224 4225 4226
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');

4227
	unevictable_nr = mem_cgroup_nr_lru_pages(mem_cont, BIT(LRU_UNEVICTABLE));
4228 4229
	seq_printf(m, "unevictable=%lu", unevictable_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4230 4231
		node_nr = mem_cgroup_node_nr_lru_pages(mem_cont, nid,
				BIT(LRU_UNEVICTABLE));
4232 4233 4234 4235 4236 4237 4238
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');
	return 0;
}
#endif /* CONFIG_NUMA */

4239 4240
static int mem_control_stat_show(struct cgroup *cont, struct cftype *cft,
				 struct cgroup_map_cb *cb)
4241 4242
{
	struct mem_cgroup *mem_cont = mem_cgroup_from_cont(cont);
K
KAMEZAWA Hiroyuki 已提交
4243
	struct mcs_total_stat mystat;
4244 4245
	int i;

K
KAMEZAWA Hiroyuki 已提交
4246 4247
	memset(&mystat, 0, sizeof(mystat));
	mem_cgroup_get_local_stat(mem_cont, &mystat);
4248

4249

4250 4251 4252
	for (i = 0; i < NR_MCS_STAT; i++) {
		if (i == MCS_SWAP && !do_swap_account)
			continue;
K
KAMEZAWA Hiroyuki 已提交
4253
		cb->fill(cb, memcg_stat_strings[i].local_name, mystat.stat[i]);
4254
	}
L
Lee Schermerhorn 已提交
4255

K
KAMEZAWA Hiroyuki 已提交
4256
	/* Hierarchical information */
4257 4258 4259 4260 4261 4262 4263
	{
		unsigned long long limit, memsw_limit;
		memcg_get_hierarchical_limit(mem_cont, &limit, &memsw_limit);
		cb->fill(cb, "hierarchical_memory_limit", limit);
		if (do_swap_account)
			cb->fill(cb, "hierarchical_memsw_limit", memsw_limit);
	}
K
KOSAKI Motohiro 已提交
4264

K
KAMEZAWA Hiroyuki 已提交
4265 4266
	memset(&mystat, 0, sizeof(mystat));
	mem_cgroup_get_total_stat(mem_cont, &mystat);
4267 4268 4269
	for (i = 0; i < NR_MCS_STAT; i++) {
		if (i == MCS_SWAP && !do_swap_account)
			continue;
K
KAMEZAWA Hiroyuki 已提交
4270
		cb->fill(cb, memcg_stat_strings[i].total_name, mystat.stat[i]);
4271
	}
K
KAMEZAWA Hiroyuki 已提交
4272

K
KOSAKI Motohiro 已提交
4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298 4299
#ifdef CONFIG_DEBUG_VM
	{
		int nid, zid;
		struct mem_cgroup_per_zone *mz;
		unsigned long recent_rotated[2] = {0, 0};
		unsigned long recent_scanned[2] = {0, 0};

		for_each_online_node(nid)
			for (zid = 0; zid < MAX_NR_ZONES; zid++) {
				mz = mem_cgroup_zoneinfo(mem_cont, nid, zid);

				recent_rotated[0] +=
					mz->reclaim_stat.recent_rotated[0];
				recent_rotated[1] +=
					mz->reclaim_stat.recent_rotated[1];
				recent_scanned[0] +=
					mz->reclaim_stat.recent_scanned[0];
				recent_scanned[1] +=
					mz->reclaim_stat.recent_scanned[1];
			}
		cb->fill(cb, "recent_rotated_anon", recent_rotated[0]);
		cb->fill(cb, "recent_rotated_file", recent_rotated[1]);
		cb->fill(cb, "recent_scanned_anon", recent_scanned[0]);
		cb->fill(cb, "recent_scanned_file", recent_scanned[1]);
	}
#endif

4300 4301 4302
	return 0;
}

K
KOSAKI Motohiro 已提交
4303 4304 4305 4306
static u64 mem_cgroup_swappiness_read(struct cgroup *cgrp, struct cftype *cft)
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);

4307
	return mem_cgroup_swappiness(memcg);
K
KOSAKI Motohiro 已提交
4308 4309 4310 4311 4312 4313 4314
}

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;
4315

K
KOSAKI Motohiro 已提交
4316 4317 4318 4319 4320 4321 4322
	if (val > 100)
		return -EINVAL;

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

	parent = mem_cgroup_from_cont(cgrp->parent);
4323 4324 4325

	cgroup_lock();

K
KOSAKI Motohiro 已提交
4326 4327
	/* If under hierarchy, only empty-root can set this value */
	if ((parent->use_hierarchy) ||
4328 4329
	    (memcg->use_hierarchy && !list_empty(&cgrp->children))) {
		cgroup_unlock();
K
KOSAKI Motohiro 已提交
4330
		return -EINVAL;
4331
	}
K
KOSAKI Motohiro 已提交
4332 4333 4334

	memcg->swappiness = val;

4335 4336
	cgroup_unlock();

K
KOSAKI Motohiro 已提交
4337 4338 4339
	return 0;
}

4340 4341 4342 4343 4344 4345 4346 4347
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)
4348
		t = rcu_dereference(memcg->thresholds.primary);
4349
	else
4350
		t = rcu_dereference(memcg->memsw_thresholds.primary);
4351 4352 4353 4354 4355 4356 4357 4358 4359 4360 4361

	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().
	 */
4362
	i = t->current_threshold;
4363 4364 4365 4366 4367 4368 4369 4370 4371 4372 4373 4374 4375 4376 4377 4378 4379 4380 4381 4382 4383 4384 4385

	/*
	 * 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 */
4386
	t->current_threshold = i - 1;
4387 4388 4389 4390 4391 4392
unlock:
	rcu_read_unlock();
}

static void mem_cgroup_threshold(struct mem_cgroup *memcg)
{
4393 4394 4395 4396 4397 4398 4399
	while (memcg) {
		__mem_cgroup_threshold(memcg, false);
		if (do_swap_account)
			__mem_cgroup_threshold(memcg, true);

		memcg = parent_mem_cgroup(memcg);
	}
4400 4401 4402 4403 4404 4405 4406 4407 4408 4409
}

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;
}

4410
static int mem_cgroup_oom_notify_cb(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
4411 4412 4413
{
	struct mem_cgroup_eventfd_list *ev;

4414
	list_for_each_entry(ev, &memcg->oom_notify, list)
K
KAMEZAWA Hiroyuki 已提交
4415 4416 4417 4418
		eventfd_signal(ev->eventfd, 1);
	return 0;
}

4419
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
4420
{
K
KAMEZAWA Hiroyuki 已提交
4421 4422
	struct mem_cgroup *iter;

4423
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
4424
		mem_cgroup_oom_notify_cb(iter);
K
KAMEZAWA Hiroyuki 已提交
4425 4426 4427 4428
}

static int mem_cgroup_usage_register_event(struct cgroup *cgrp,
	struct cftype *cft, struct eventfd_ctx *eventfd, const char *args)
4429 4430
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4431 4432
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
4433 4434
	int type = MEMFILE_TYPE(cft->private);
	u64 threshold, usage;
4435
	int i, size, ret;
4436 4437 4438 4439 4440 4441

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

	mutex_lock(&memcg->thresholds_lock);
4442

4443
	if (type == _MEM)
4444
		thresholds = &memcg->thresholds;
4445
	else if (type == _MEMSWAP)
4446
		thresholds = &memcg->memsw_thresholds;
4447 4448 4449 4450 4451 4452
	else
		BUG();

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

	/* Check if a threshold crossed before adding a new one */
4453
	if (thresholds->primary)
4454 4455
		__mem_cgroup_threshold(memcg, type == _MEMSWAP);

4456
	size = thresholds->primary ? thresholds->primary->size + 1 : 1;
4457 4458

	/* Allocate memory for new array of thresholds */
4459
	new = kmalloc(sizeof(*new) + size * sizeof(struct mem_cgroup_threshold),
4460
			GFP_KERNEL);
4461
	if (!new) {
4462 4463 4464
		ret = -ENOMEM;
		goto unlock;
	}
4465
	new->size = size;
4466 4467

	/* Copy thresholds (if any) to new array */
4468 4469
	if (thresholds->primary) {
		memcpy(new->entries, thresholds->primary->entries, (size - 1) *
4470
				sizeof(struct mem_cgroup_threshold));
4471 4472
	}

4473
	/* Add new threshold */
4474 4475
	new->entries[size - 1].eventfd = eventfd;
	new->entries[size - 1].threshold = threshold;
4476 4477

	/* Sort thresholds. Registering of new threshold isn't time-critical */
4478
	sort(new->entries, size, sizeof(struct mem_cgroup_threshold),
4479 4480 4481
			compare_thresholds, NULL);

	/* Find current threshold */
4482
	new->current_threshold = -1;
4483
	for (i = 0; i < size; i++) {
4484
		if (new->entries[i].threshold < usage) {
4485
			/*
4486 4487
			 * new->current_threshold will not be used until
			 * rcu_assign_pointer(), so it's safe to increment
4488 4489
			 * it here.
			 */
4490
			++new->current_threshold;
4491 4492 4493
		}
	}

4494 4495 4496 4497 4498
	/* Free old spare buffer and save old primary buffer as spare */
	kfree(thresholds->spare);
	thresholds->spare = thresholds->primary;

	rcu_assign_pointer(thresholds->primary, new);
4499

4500
	/* To be sure that nobody uses thresholds */
4501 4502 4503 4504 4505 4506 4507 4508
	synchronize_rcu();

unlock:
	mutex_unlock(&memcg->thresholds_lock);

	return ret;
}

4509
static void mem_cgroup_usage_unregister_event(struct cgroup *cgrp,
K
KAMEZAWA Hiroyuki 已提交
4510
	struct cftype *cft, struct eventfd_ctx *eventfd)
4511 4512
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4513 4514
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
4515 4516
	int type = MEMFILE_TYPE(cft->private);
	u64 usage;
4517
	int i, j, size;
4518 4519 4520

	mutex_lock(&memcg->thresholds_lock);
	if (type == _MEM)
4521
		thresholds = &memcg->thresholds;
4522
	else if (type == _MEMSWAP)
4523
		thresholds = &memcg->memsw_thresholds;
4524 4525 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535 4536 4537 4538
	else
		BUG();

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

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

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

	/* Calculate new number of threshold */
4539 4540 4541
	size = 0;
	for (i = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd != eventfd)
4542 4543 4544
			size++;
	}

4545
	new = thresholds->spare;
4546

4547 4548
	/* Set thresholds array to NULL if we don't have thresholds */
	if (!size) {
4549 4550
		kfree(new);
		new = NULL;
4551
		goto swap_buffers;
4552 4553
	}

4554
	new->size = size;
4555 4556

	/* Copy thresholds and find current threshold */
4557 4558 4559
	new->current_threshold = -1;
	for (i = 0, j = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd == eventfd)
4560 4561
			continue;

4562 4563
		new->entries[j] = thresholds->primary->entries[i];
		if (new->entries[j].threshold < usage) {
4564
			/*
4565
			 * new->current_threshold will not be used
4566 4567 4568
			 * until rcu_assign_pointer(), so it's safe to increment
			 * it here.
			 */
4569
			++new->current_threshold;
4570 4571 4572 4573
		}
		j++;
	}

4574
swap_buffers:
4575 4576 4577
	/* Swap primary and spare array */
	thresholds->spare = thresholds->primary;
	rcu_assign_pointer(thresholds->primary, new);
4578

4579
	/* To be sure that nobody uses thresholds */
4580 4581 4582 4583
	synchronize_rcu();

	mutex_unlock(&memcg->thresholds_lock);
}
4584

K
KAMEZAWA Hiroyuki 已提交
4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596
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;

4597
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4598 4599 4600 4601 4602

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

	/* already in OOM ? */
4603
	if (atomic_read(&memcg->under_oom))
K
KAMEZAWA Hiroyuki 已提交
4604
		eventfd_signal(eventfd, 1);
4605
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4606 4607 4608 4609

	return 0;
}

4610
static void mem_cgroup_oom_unregister_event(struct cgroup *cgrp,
K
KAMEZAWA Hiroyuki 已提交
4611 4612
	struct cftype *cft, struct eventfd_ctx *eventfd)
{
4613
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
K
KAMEZAWA Hiroyuki 已提交
4614 4615 4616 4617 4618
	struct mem_cgroup_eventfd_list *ev, *tmp;
	int type = MEMFILE_TYPE(cft->private);

	BUG_ON(type != _OOM_TYPE);

4619
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4620

4621
	list_for_each_entry_safe(ev, tmp, &memcg->oom_notify, list) {
K
KAMEZAWA Hiroyuki 已提交
4622 4623 4624 4625 4626 4627
		if (ev->eventfd == eventfd) {
			list_del(&ev->list);
			kfree(ev);
		}
	}

4628
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4629 4630
}

4631 4632 4633
static int mem_cgroup_oom_control_read(struct cgroup *cgrp,
	struct cftype *cft,  struct cgroup_map_cb *cb)
{
4634
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4635

4636
	cb->fill(cb, "oom_kill_disable", memcg->oom_kill_disable);
4637

4638
	if (atomic_read(&memcg->under_oom))
4639 4640 4641 4642 4643 4644 4645 4646 4647
		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)
{
4648
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659
	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) ||
4660
	    (memcg->use_hierarchy && !list_empty(&cgrp->children))) {
4661 4662 4663
		cgroup_unlock();
		return -EINVAL;
	}
4664
	memcg->oom_kill_disable = val;
4665
	if (!val)
4666
		memcg_oom_recover(memcg);
4667 4668 4669 4670
	cgroup_unlock();
	return 0;
}

4671 4672 4673 4674 4675 4676 4677 4678 4679 4680 4681 4682 4683 4684 4685 4686
#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 */

4687 4688 4689
#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
4690 4691 4692 4693 4694 4695 4696
	/*
	 * 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
	 */
4697
	return mem_cgroup_sockets_init(cont, ss);
4698 4699
};

G
Glauber Costa 已提交
4700 4701 4702 4703 4704
static void kmem_cgroup_destroy(struct cgroup_subsys *ss,
				struct cgroup *cont)
{
	mem_cgroup_sockets_destroy(cont, ss);
}
4705 4706 4707 4708 4709
#else
static int register_kmem_files(struct cgroup *cont, struct cgroup_subsys *ss)
{
	return 0;
}
G
Glauber Costa 已提交
4710 4711 4712 4713 4714

static void kmem_cgroup_destroy(struct cgroup_subsys *ss,
				struct cgroup *cont)
{
}
4715 4716
#endif

B
Balbir Singh 已提交
4717 4718
static struct cftype mem_cgroup_files[] = {
	{
4719
		.name = "usage_in_bytes",
4720
		.private = MEMFILE_PRIVATE(_MEM, RES_USAGE),
4721
		.read_u64 = mem_cgroup_read,
K
KAMEZAWA Hiroyuki 已提交
4722 4723
		.register_event = mem_cgroup_usage_register_event,
		.unregister_event = mem_cgroup_usage_unregister_event,
B
Balbir Singh 已提交
4724
	},
4725 4726
	{
		.name = "max_usage_in_bytes",
4727
		.private = MEMFILE_PRIVATE(_MEM, RES_MAX_USAGE),
4728
		.trigger = mem_cgroup_reset,
4729 4730
		.read_u64 = mem_cgroup_read,
	},
B
Balbir Singh 已提交
4731
	{
4732
		.name = "limit_in_bytes",
4733
		.private = MEMFILE_PRIVATE(_MEM, RES_LIMIT),
4734
		.write_string = mem_cgroup_write,
4735
		.read_u64 = mem_cgroup_read,
B
Balbir Singh 已提交
4736
	},
4737 4738 4739 4740 4741 4742
	{
		.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
4743 4744
	{
		.name = "failcnt",
4745
		.private = MEMFILE_PRIVATE(_MEM, RES_FAILCNT),
4746
		.trigger = mem_cgroup_reset,
4747
		.read_u64 = mem_cgroup_read,
B
Balbir Singh 已提交
4748
	},
4749 4750
	{
		.name = "stat",
4751
		.read_map = mem_control_stat_show,
4752
	},
4753 4754 4755 4756
	{
		.name = "force_empty",
		.trigger = mem_cgroup_force_empty_write,
	},
4757 4758 4759 4760 4761
	{
		.name = "use_hierarchy",
		.write_u64 = mem_cgroup_hierarchy_write,
		.read_u64 = mem_cgroup_hierarchy_read,
	},
K
KOSAKI Motohiro 已提交
4762 4763 4764 4765 4766
	{
		.name = "swappiness",
		.read_u64 = mem_cgroup_swappiness_read,
		.write_u64 = mem_cgroup_swappiness_write,
	},
4767 4768 4769 4770 4771
	{
		.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
4772 4773
	{
		.name = "oom_control",
4774 4775
		.read_map = mem_cgroup_oom_control_read,
		.write_u64 = mem_cgroup_oom_control_write,
K
KAMEZAWA Hiroyuki 已提交
4776 4777 4778 4779
		.register_event = mem_cgroup_oom_register_event,
		.unregister_event = mem_cgroup_oom_unregister_event,
		.private = MEMFILE_PRIVATE(_OOM_TYPE, OOM_CONTROL),
	},
4780 4781 4782 4783
#ifdef CONFIG_NUMA
	{
		.name = "numa_stat",
		.open = mem_control_numa_stat_open,
4784
		.mode = S_IRUGO,
4785 4786
	},
#endif
B
Balbir Singh 已提交
4787 4788
};

4789 4790 4791 4792 4793 4794
#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
4795 4796
		.register_event = mem_cgroup_usage_register_event,
		.unregister_event = mem_cgroup_usage_unregister_event,
4797 4798 4799 4800 4801 4802 4803 4804 4805 4806 4807 4808 4809 4810 4811 4812 4813 4814 4815 4816 4817 4818 4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830 4831
	},
	{
		.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

4832
static int alloc_mem_cgroup_per_zone_info(struct mem_cgroup *memcg, int node)
4833 4834
{
	struct mem_cgroup_per_node *pn;
4835
	struct mem_cgroup_per_zone *mz;
4836
	enum lru_list l;
4837
	int zone, tmp = node;
4838 4839 4840 4841 4842 4843 4844 4845
	/*
	 * 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.
	 */
4846 4847
	if (!node_state(node, N_NORMAL_MEMORY))
		tmp = -1;
4848
	pn = kzalloc_node(sizeof(*pn), GFP_KERNEL, tmp);
4849 4850
	if (!pn)
		return 1;
4851 4852 4853

	for (zone = 0; zone < MAX_NR_ZONES; zone++) {
		mz = &pn->zoneinfo[zone];
4854
		for_each_lru(l)
4855
			INIT_LIST_HEAD(&mz->lruvec.lists[l]);
4856
		mz->usage_in_excess = 0;
4857
		mz->on_tree = false;
4858
		mz->mem = memcg;
4859
	}
4860
	memcg->info.nodeinfo[node] = pn;
4861 4862 4863
	return 0;
}

4864
static void free_mem_cgroup_per_zone_info(struct mem_cgroup *memcg, int node)
4865
{
4866
	kfree(memcg->info.nodeinfo[node]);
4867 4868
}

4869 4870 4871
static struct mem_cgroup *mem_cgroup_alloc(void)
{
	struct mem_cgroup *mem;
4872
	int size = sizeof(struct mem_cgroup);
4873

4874
	/* Can be very big if MAX_NUMNODES is very big */
4875
	if (size < PAGE_SIZE)
4876
		mem = kzalloc(size, GFP_KERNEL);
4877
	else
4878
		mem = vzalloc(size);
4879

4880 4881 4882
	if (!mem)
		return NULL;

4883
	mem->stat = alloc_percpu(struct mem_cgroup_stat_cpu);
4884 4885
	if (!mem->stat)
		goto out_free;
4886
	spin_lock_init(&mem->pcp_counter_lock);
4887
	return mem;
4888 4889 4890 4891 4892 4893 4894

out_free:
	if (size < PAGE_SIZE)
		kfree(mem);
	else
		vfree(mem);
	return NULL;
4895 4896
}

4897 4898 4899 4900 4901 4902 4903 4904 4905 4906 4907
/*
 * 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.
 */

4908
static void __mem_cgroup_free(struct mem_cgroup *memcg)
4909
{
K
KAMEZAWA Hiroyuki 已提交
4910 4911
	int node;

4912 4913
	mem_cgroup_remove_from_trees(memcg);
	free_css_id(&mem_cgroup_subsys, &memcg->css);
K
KAMEZAWA Hiroyuki 已提交
4914

K
KAMEZAWA Hiroyuki 已提交
4915
	for_each_node_state(node, N_POSSIBLE)
4916
		free_mem_cgroup_per_zone_info(memcg, node);
K
KAMEZAWA Hiroyuki 已提交
4917

4918
	free_percpu(memcg->stat);
4919
	if (sizeof(struct mem_cgroup) < PAGE_SIZE)
4920
		kfree(memcg);
4921
	else
4922
		vfree(memcg);
4923 4924
}

4925
static void mem_cgroup_get(struct mem_cgroup *memcg)
4926
{
4927
	atomic_inc(&memcg->refcnt);
4928 4929
}

4930
static void __mem_cgroup_put(struct mem_cgroup *memcg, int count)
4931
{
4932 4933 4934
	if (atomic_sub_and_test(count, &memcg->refcnt)) {
		struct mem_cgroup *parent = parent_mem_cgroup(memcg);
		__mem_cgroup_free(memcg);
4935 4936 4937
		if (parent)
			mem_cgroup_put(parent);
	}
4938 4939
}

4940
static void mem_cgroup_put(struct mem_cgroup *memcg)
4941
{
4942
	__mem_cgroup_put(memcg, 1);
4943 4944
}

4945 4946 4947
/*
 * Returns the parent mem_cgroup in memcgroup hierarchy with hierarchy enabled.
 */
G
Glauber Costa 已提交
4948
struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
4949
{
4950
	if (!memcg->res.parent)
4951
		return NULL;
4952
	return mem_cgroup_from_res_counter(memcg->res.parent, res);
4953
}
G
Glauber Costa 已提交
4954
EXPORT_SYMBOL(parent_mem_cgroup);
4955

4956 4957 4958
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
static void __init enable_swap_cgroup(void)
{
4959
	if (!mem_cgroup_disabled() && really_do_swap_account)
4960 4961 4962 4963 4964 4965 4966 4967
		do_swap_account = 1;
}
#else
static void __init enable_swap_cgroup(void)
{
}
#endif

4968 4969 4970 4971 4972 4973 4974 4975 4976 4977 4978 4979 4980 4981 4982 4983 4984 4985 4986 4987 4988 4989 4990 4991 4992
static int mem_cgroup_soft_limit_tree_init(void)
{
	struct mem_cgroup_tree_per_node *rtpn;
	struct mem_cgroup_tree_per_zone *rtpz;
	int tmp, node, zone;

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

		soft_limit_tree.rb_tree_per_node[node] = rtpn;

		for (zone = 0; zone < MAX_NR_ZONES; zone++) {
			rtpz = &rtpn->rb_tree_per_zone[zone];
			rtpz->rb_root = RB_ROOT;
			spin_lock_init(&rtpz->lock);
		}
	}
	return 0;
}

L
Li Zefan 已提交
4993
static struct cgroup_subsys_state * __ref
B
Balbir Singh 已提交
4994 4995
mem_cgroup_create(struct cgroup_subsys *ss, struct cgroup *cont)
{
4996
	struct mem_cgroup *memcg, *parent;
K
KAMEZAWA Hiroyuki 已提交
4997
	long error = -ENOMEM;
4998
	int node;
B
Balbir Singh 已提交
4999

5000 5001
	memcg = mem_cgroup_alloc();
	if (!memcg)
K
KAMEZAWA Hiroyuki 已提交
5002
		return ERR_PTR(error);
5003

5004
	for_each_node_state(node, N_POSSIBLE)
5005
		if (alloc_mem_cgroup_per_zone_info(memcg, node))
5006
			goto free_out;
5007

5008
	/* root ? */
5009
	if (cont->parent == NULL) {
5010
		int cpu;
5011
		enable_swap_cgroup();
5012
		parent = NULL;
5013 5014
		if (mem_cgroup_soft_limit_tree_init())
			goto free_out;
5015
		root_mem_cgroup = memcg;
5016 5017 5018 5019 5020
		for_each_possible_cpu(cpu) {
			struct memcg_stock_pcp *stock =
						&per_cpu(memcg_stock, cpu);
			INIT_WORK(&stock->work, drain_local_stock);
		}
5021
		hotcpu_notifier(memcg_cpu_hotplug_callback, 0);
5022
	} else {
5023
		parent = mem_cgroup_from_cont(cont->parent);
5024 5025
		memcg->use_hierarchy = parent->use_hierarchy;
		memcg->oom_kill_disable = parent->oom_kill_disable;
5026
	}
5027

5028
	if (parent && parent->use_hierarchy) {
5029 5030
		res_counter_init(&memcg->res, &parent->res);
		res_counter_init(&memcg->memsw, &parent->memsw);
5031 5032 5033 5034 5035 5036 5037
		/*
		 * 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);
5038
	} else {
5039 5040
		res_counter_init(&memcg->res, NULL);
		res_counter_init(&memcg->memsw, NULL);
5041
	}
5042 5043
	memcg->last_scanned_node = MAX_NUMNODES;
	INIT_LIST_HEAD(&memcg->oom_notify);
5044

K
KOSAKI Motohiro 已提交
5045
	if (parent)
5046 5047 5048 5049 5050
		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;
5051
free_out:
5052
	__mem_cgroup_free(memcg);
K
KAMEZAWA Hiroyuki 已提交
5053
	return ERR_PTR(error);
B
Balbir Singh 已提交
5054 5055
}

5056
static int mem_cgroup_pre_destroy(struct cgroup_subsys *ss,
5057 5058
					struct cgroup *cont)
{
5059
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
5060

5061
	return mem_cgroup_force_empty(memcg, false);
5062 5063
}

B
Balbir Singh 已提交
5064 5065 5066
static void mem_cgroup_destroy(struct cgroup_subsys *ss,
				struct cgroup *cont)
{
5067
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
5068

G
Glauber Costa 已提交
5069 5070
	kmem_cgroup_destroy(ss, cont);

5071
	mem_cgroup_put(memcg);
B
Balbir Singh 已提交
5072 5073 5074 5075 5076
}

static int mem_cgroup_populate(struct cgroup_subsys *ss,
				struct cgroup *cont)
{
5077 5078 5079 5080 5081 5082 5083
	int ret;

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

	if (!ret)
		ret = register_memsw_files(cont, ss);
5084 5085 5086 5087

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

5088
	return ret;
B
Balbir Singh 已提交
5089 5090
}

5091
#ifdef CONFIG_MMU
5092
/* Handlers for move charge at task migration. */
5093 5094
#define PRECHARGE_COUNT_AT_ONCE	256
static int mem_cgroup_do_precharge(unsigned long count)
5095
{
5096 5097
	int ret = 0;
	int batch_count = PRECHARGE_COUNT_AT_ONCE;
5098
	struct mem_cgroup *memcg = mc.to;
5099

5100
	if (mem_cgroup_is_root(memcg)) {
5101 5102 5103 5104 5105 5106 5107 5108
		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;
		/*
5109
		 * "memcg" cannot be under rmdir() because we've already checked
5110 5111 5112 5113
		 * 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().
		 */
5114
		if (res_counter_charge(&memcg->res, PAGE_SIZE * count, &dummy))
5115
			goto one_by_one;
5116
		if (do_swap_account && res_counter_charge(&memcg->memsw,
5117
						PAGE_SIZE * count, &dummy)) {
5118
			res_counter_uncharge(&memcg->res, PAGE_SIZE * count);
5119 5120 5121 5122 5123 5124 5125 5126 5127 5128 5129 5130 5131 5132 5133 5134
			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();
		}
5135 5136 5137
		ret = __mem_cgroup_try_charge(NULL,
					GFP_KERNEL, 1, &memcg, false);
		if (ret || !memcg)
5138 5139 5140 5141
			/* mem_cgroup_clear_mc() will do uncharge later */
			return -ENOMEM;
		mc.precharge++;
	}
5142 5143 5144 5145 5146 5147 5148 5149
	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
5150
 * @target: the pointer the target page or swap ent will be stored(can be NULL)
5151 5152 5153 5154 5155 5156
 *
 * 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).
5157 5158 5159
 *   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.
5160 5161 5162 5163 5164
 *
 * Called with pte lock held.
 */
union mc_target {
	struct page	*page;
5165
	swp_entry_t	ent;
5166 5167 5168 5169 5170
};

enum mc_target_type {
	MC_TARGET_NONE,	/* not used */
	MC_TARGET_PAGE,
5171
	MC_TARGET_SWAP,
5172 5173
};

D
Daisuke Nishimura 已提交
5174 5175
static struct page *mc_handle_present_pte(struct vm_area_struct *vma,
						unsigned long addr, pte_t ptent)
5176
{
D
Daisuke Nishimura 已提交
5177
	struct page *page = vm_normal_page(vma, addr, ptent);
5178

D
Daisuke Nishimura 已提交
5179 5180 5181 5182 5183 5184
	if (!page || !page_mapped(page))
		return NULL;
	if (PageAnon(page)) {
		/* we don't move shared anon */
		if (!move_anon() || page_mapcount(page) > 2)
			return NULL;
5185 5186
	} else if (!move_file())
		/* we ignore mapcount for file pages */
D
Daisuke Nishimura 已提交
5187 5188 5189 5190 5191 5192 5193 5194 5195 5196 5197 5198 5199 5200 5201 5202 5203 5204
		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 */
5205 5206
		if (page)
			put_page(page);
D
Daisuke Nishimura 已提交
5207
		return NULL;
5208
	}
D
Daisuke Nishimura 已提交
5209 5210 5211 5212 5213 5214
	if (do_swap_account)
		entry->val = ent.val;

	return page;
}

5215 5216 5217 5218 5219 5220 5221 5222 5223 5224 5225 5226 5227 5228 5229 5230 5231 5232 5233 5234 5235
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). */
5236 5237 5238 5239 5240 5241
	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);
5242
		if (do_swap_account)
5243 5244
			*entry = swap;
		page = find_get_page(&swapper_space, swap.val);
5245
	}
5246
#endif
5247 5248 5249
	return page;
}

D
Daisuke Nishimura 已提交
5250 5251 5252 5253 5254 5255 5256 5257 5258 5259 5260 5261
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);
5262 5263
	else if (pte_none(ptent) || pte_file(ptent))
		page = mc_handle_file_pte(vma, addr, ptent, &ent);
D
Daisuke Nishimura 已提交
5264 5265 5266

	if (!page && !ent.val)
		return 0;
5267 5268 5269 5270 5271 5272 5273 5274 5275 5276 5277 5278 5279 5280 5281
	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
5282 5283
	/* There is a swap entry and a page doesn't exist or isn't charged */
	if (ent.val && !ret &&
5284 5285 5286 5287
			css_id(&mc.from->css) == lookup_swap_cgroup(ent)) {
		ret = MC_TARGET_SWAP;
		if (target)
			target->ent = ent;
5288 5289 5290 5291 5292 5293 5294 5295 5296 5297 5298 5299
	}
	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;

5300 5301
	split_huge_page_pmd(walk->mm, pmd);

5302 5303 5304 5305 5306 5307 5308
	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();

5309 5310 5311
	return 0;
}

5312 5313 5314 5315 5316
static unsigned long mem_cgroup_count_precharge(struct mm_struct *mm)
{
	unsigned long precharge;
	struct vm_area_struct *vma;

5317
	down_read(&mm->mmap_sem);
5318 5319 5320 5321 5322 5323 5324 5325 5326 5327 5328
	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);
	}
5329
	up_read(&mm->mmap_sem);
5330 5331 5332 5333 5334 5335 5336 5337 5338

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

	return precharge;
}

static int mem_cgroup_precharge_mc(struct mm_struct *mm)
{
5339 5340 5341 5342 5343
	unsigned long precharge = mem_cgroup_count_precharge(mm);

	VM_BUG_ON(mc.moving_task);
	mc.moving_task = current;
	return mem_cgroup_do_precharge(precharge);
5344 5345
}

5346 5347
/* cancels all extra charges on mc.from and mc.to, and wakes up all waiters. */
static void __mem_cgroup_clear_mc(void)
5348
{
5349 5350 5351
	struct mem_cgroup *from = mc.from;
	struct mem_cgroup *to = mc.to;

5352
	/* we must uncharge all the leftover precharges from mc.to */
5353 5354 5355 5356 5357 5358 5359 5360 5361 5362 5363
	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;
5364
	}
5365 5366 5367 5368 5369 5370 5371 5372 5373 5374 5375 5376 5377 5378 5379 5380 5381 5382 5383
	/* 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;
	}
5384 5385 5386 5387 5388 5389 5390 5391 5392 5393 5394 5395 5396 5397 5398
	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();
5399
	spin_lock(&mc.lock);
5400 5401
	mc.from = NULL;
	mc.to = NULL;
5402
	spin_unlock(&mc.lock);
5403
	mem_cgroup_end_move(from);
5404 5405
}

5406 5407
static int mem_cgroup_can_attach(struct cgroup_subsys *ss,
				struct cgroup *cgroup,
5408
				struct cgroup_taskset *tset)
5409
{
5410
	struct task_struct *p = cgroup_taskset_first(tset);
5411
	int ret = 0;
5412
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgroup);
5413

5414
	if (memcg->move_charge_at_immigrate) {
5415 5416 5417
		struct mm_struct *mm;
		struct mem_cgroup *from = mem_cgroup_from_task(p);

5418
		VM_BUG_ON(from == memcg);
5419 5420 5421 5422 5423

		mm = get_task_mm(p);
		if (!mm)
			return 0;
		/* We move charges only when we move a owner of the mm */
5424 5425 5426 5427
		if (mm->owner == p) {
			VM_BUG_ON(mc.from);
			VM_BUG_ON(mc.to);
			VM_BUG_ON(mc.precharge);
5428
			VM_BUG_ON(mc.moved_charge);
5429
			VM_BUG_ON(mc.moved_swap);
5430
			mem_cgroup_start_move(from);
5431
			spin_lock(&mc.lock);
5432
			mc.from = from;
5433
			mc.to = memcg;
5434
			spin_unlock(&mc.lock);
5435
			/* We set mc.moving_task later */
5436 5437 5438 5439

			ret = mem_cgroup_precharge_mc(mm);
			if (ret)
				mem_cgroup_clear_mc();
5440 5441
		}
		mmput(mm);
5442 5443 5444 5445 5446 5447
	}
	return ret;
}

static void mem_cgroup_cancel_attach(struct cgroup_subsys *ss,
				struct cgroup *cgroup,
5448
				struct cgroup_taskset *tset)
5449
{
5450
	mem_cgroup_clear_mc();
5451 5452
}

5453 5454 5455
static int mem_cgroup_move_charge_pte_range(pmd_t *pmd,
				unsigned long addr, unsigned long end,
				struct mm_walk *walk)
5456
{
5457 5458 5459 5460 5461
	int ret = 0;
	struct vm_area_struct *vma = walk->private;
	pte_t *pte;
	spinlock_t *ptl;

5462
	split_huge_page_pmd(walk->mm, pmd);
5463 5464 5465 5466 5467 5468 5469 5470
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;
5471
		swp_entry_t ent;
5472 5473 5474 5475 5476 5477 5478 5479 5480 5481 5482

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

B
Balbir Singh 已提交
5563 5564
static void mem_cgroup_move_task(struct cgroup_subsys *ss,
				struct cgroup *cont,
5565
				struct cgroup_taskset *tset)
B
Balbir Singh 已提交
5566
{
5567
	struct task_struct *p = cgroup_taskset_first(tset);
5568
	struct mm_struct *mm = get_task_mm(p);
5569 5570

	if (mm) {
5571 5572 5573
		if (mc.to)
			mem_cgroup_move_charge(mm);
		put_swap_token(mm);
5574 5575
		mmput(mm);
	}
5576 5577
	if (mc.to)
		mem_cgroup_clear_mc();
B
Balbir Singh 已提交
5578
}
5579 5580 5581
#else	/* !CONFIG_MMU */
static int mem_cgroup_can_attach(struct cgroup_subsys *ss,
				struct cgroup *cgroup,
5582
				struct cgroup_taskset *tset)
5583 5584 5585 5586 5587
{
	return 0;
}
static void mem_cgroup_cancel_attach(struct cgroup_subsys *ss,
				struct cgroup *cgroup,
5588
				struct cgroup_taskset *tset)
5589 5590 5591 5592
{
}
static void mem_cgroup_move_task(struct cgroup_subsys *ss,
				struct cgroup *cont,
5593
				struct cgroup_taskset *tset)
5594 5595 5596
{
}
#endif
B
Balbir Singh 已提交
5597

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

#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
5613 5614 5615
static int __init enable_swap_account(char *s)
{
	/* consider enabled if no parameter or 1 is given */
5616
	if (!strcmp(s, "1"))
5617
		really_do_swap_account = 1;
5618
	else if (!strcmp(s, "0"))
5619 5620 5621
		really_do_swap_account = 0;
	return 1;
}
5622
__setup("swapaccount=", enable_swap_account);
5623 5624

#endif