memcontrol.c 146.8 KB
Newer Older
B
Balbir Singh 已提交
1 2 3 4 5
/* memcontrol.c - Memory Controller
 *
 * Copyright IBM Corporation, 2007
 * Author Balbir Singh <balbir@linux.vnet.ibm.com>
 *
6 7 8
 * Copyright 2007 OpenVZ SWsoft Inc
 * Author: Pavel Emelianov <xemul@openvz.org>
 *
9 10 11 12
 * Memory thresholds
 * Copyright (C) 2009 Nokia Corporation
 * Author: Kirill A. Shutemov
 *
B
Balbir Singh 已提交
13 14 15 16 17 18 19 20 21 22 23 24 25 26
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 */

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

56 57
#include <asm/uaccess.h>

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

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

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

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

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


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

95 96 97 98
enum mem_cgroup_events_index {
	MEM_CGROUP_EVENTS_PGPGIN,	/* # of pages paged in */
	MEM_CGROUP_EVENTS_PGPGOUT,	/* # of pages paged out */
	MEM_CGROUP_EVENTS_COUNT,	/* # of pages paged in/out */
99 100
	MEM_CGROUP_EVENTS_PGFAULT,	/* # of page-faults */
	MEM_CGROUP_EVENTS_PGMAJFAULT,	/* # of major page-faults */
101 102
	MEM_CGROUP_EVENTS_NSTATS,
};
103 104 105 106 107 108 109 110 111
/*
 * Per memcg event counter is incremented at every pagein/pageout. With THP,
 * it will be incremated by the number of pages. This counter is used for
 * for trigger some periodic events. This is straightforward and better
 * than using jiffies etc. to handle periodic memcg event.
 */
enum mem_cgroup_events_target {
	MEM_CGROUP_TARGET_THRESH,
	MEM_CGROUP_TARGET_SOFTLIMIT,
112
	MEM_CGROUP_TARGET_NUMAINFO,
113 114 115 116
	MEM_CGROUP_NTARGETS,
};
#define THRESHOLDS_EVENTS_TARGET (128)
#define SOFTLIMIT_EVENTS_TARGET (1024)
117
#define NUMAINFO_EVENTS_TARGET	(1024)
118

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

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

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

139 140
	struct mem_cgroup_reclaim_iter reclaim_iter[DEF_PRIORITY + 1];

K
KOSAKI Motohiro 已提交
141
	struct zone_reclaim_stat reclaim_stat;
142 143 144 145
	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;
146
	struct mem_cgroup	*memcg;		/* Back pointer, we cannot */
147
						/* use container_of	   */
148 149 150 151 152 153 154 155 156 157
};

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];
};

158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177
/*
 * 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;

178 179 180 181 182
struct mem_cgroup_threshold {
	struct eventfd_ctx *eventfd;
	u64 threshold;
};

K
KAMEZAWA Hiroyuki 已提交
183
/* For threshold */
184 185
struct mem_cgroup_threshold_ary {
	/* An array index points to threshold just below usage. */
186
	int current_threshold;
187 188 189 190 191
	/* Size of entries[] */
	unsigned int size;
	/* Array of thresholds */
	struct mem_cgroup_threshold entries[0];
};
192 193 194 195 196 197 198 199 200 201 202 203

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
204 205 206 207 208
/* for OOM */
struct mem_cgroup_eventfd_list {
	struct list_head list;
	struct eventfd_ctx *eventfd;
};
209

210 211
static void mem_cgroup_threshold(struct mem_cgroup *memcg);
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg);
212

B
Balbir Singh 已提交
213 214 215 216 217 218 219
/*
 * 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
220 221 222
 * 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
223 224 225 226 227 228 229
 */
struct mem_cgroup {
	struct cgroup_subsys_state css;
	/*
	 * the counter to account for memory usage
	 */
	struct res_counter res;
230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253

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

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

254 255 256 257
	/*
	 * Per cgroup active and inactive list, similar to the
	 * per zone LRU lists.
	 */
258
	struct mem_cgroup_lru_info info;
259 260 261
	int last_scanned_node;
#if MAX_NUMNODES > 1
	nodemask_t	scan_nodes;
262 263
	atomic_t	numainfo_events;
	atomic_t	numainfo_updating;
264
#endif
265 266 267 268
	/*
	 * Should the accounting and control be hierarchical, per subtree?
	 */
	bool use_hierarchy;
269 270 271 272

	bool		oom_lock;
	atomic_t	under_oom;

273
	atomic_t	refcnt;
274

275
	int	swappiness;
276 277
	/* OOM-Killer disable */
	int		oom_kill_disable;
K
KOSAKI Motohiro 已提交
278

279 280 281
	/* set when res.limit == memsw.limit */
	bool		memsw_is_minimum;

282 283 284 285
	/* protect arrays of thresholds */
	struct mutex thresholds_lock;

	/* thresholds for memory usage. RCU-protected */
286
	struct mem_cgroup_thresholds thresholds;
287

288
	/* thresholds for mem+swap usage. RCU-protected */
289
	struct mem_cgroup_thresholds memsw_thresholds;
290

K
KAMEZAWA Hiroyuki 已提交
291 292
	/* For oom notifier event fd */
	struct list_head oom_notify;
293

294 295 296 297 298
	/*
	 * 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;
299 300 301 302
	/*
	 * set > 0 if pages under this cgroup are moving to other cgroup.
	 */
	atomic_t	moving_account;
303 304
	/* taken only while moving_account > 0 */
	spinlock_t	move_lock;
305
	/*
306
	 * percpu counter.
307
	 */
308
	struct mem_cgroup_stat_cpu *stat;
309 310 311 312 313 314
	/*
	 * 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
315 316 317 318

#ifdef CONFIG_INET
	struct tcp_memcontrol tcp_mem;
#endif
B
Balbir Singh 已提交
319 320
};

321 322 323 324 325 326
/* 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 {
327
	MOVE_CHARGE_TYPE_ANON,	/* private anonymous page and swap of it */
328
	MOVE_CHARGE_TYPE_FILE,	/* file page(including tmpfs) and swap of it */
329 330 331
	NR_MOVE_TYPE,
};

332 333
/* "mc" and its members are protected by cgroup_mutex */
static struct move_charge_struct {
334
	spinlock_t	  lock; /* for from, to */
335 336 337
	struct mem_cgroup *from;
	struct mem_cgroup *to;
	unsigned long precharge;
338
	unsigned long moved_charge;
339
	unsigned long moved_swap;
340 341 342
	struct task_struct *moving_task;	/* a task moving charges */
	wait_queue_head_t waitq;		/* a waitq for other context */
} mc = {
343
	.lock = __SPIN_LOCK_UNLOCKED(mc.lock),
344 345
	.waitq = __WAIT_QUEUE_HEAD_INITIALIZER(mc.waitq),
};
346

D
Daisuke Nishimura 已提交
347 348 349 350 351 352
static bool move_anon(void)
{
	return test_bit(MOVE_CHARGE_TYPE_ANON,
					&mc.to->move_charge_at_immigrate);
}

353 354 355 356 357 358
static bool move_file(void)
{
	return test_bit(MOVE_CHARGE_TYPE_FILE,
					&mc.to->move_charge_at_immigrate);
}

359 360 361 362 363 364 365
/*
 * 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)

366 367 368
enum charge_type {
	MEM_CGROUP_CHARGE_TYPE_CACHE = 0,
	MEM_CGROUP_CHARGE_TYPE_MAPPED,
369
	MEM_CGROUP_CHARGE_TYPE_SHMEM,	/* used by page migration of shmem */
370
	MEM_CGROUP_CHARGE_TYPE_FORCE,	/* used by force_empty */
K
KAMEZAWA Hiroyuki 已提交
371
	MEM_CGROUP_CHARGE_TYPE_SWAPOUT,	/* for accounting swapcache */
K
KAMEZAWA Hiroyuki 已提交
372
	MEM_CGROUP_CHARGE_TYPE_DROP,	/* a page was unused swap cache */
373 374 375
	NR_CHARGE_TYPE,
};

376
/* for encoding cft->private value on file */
377 378 379
#define _MEM			(0)
#define _MEMSWAP		(1)
#define _OOM_TYPE		(2)
380 381 382
#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
383 384
/* Used for OOM nofiier */
#define OOM_CONTROL		(0)
385

386 387 388 389 390 391 392 393
/*
 * 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)

394 395
static void mem_cgroup_get(struct mem_cgroup *memcg);
static void mem_cgroup_put(struct mem_cgroup *memcg);
G
Glauber Costa 已提交
396 397 398 399

/* Writing them here to avoid exposing memcg's inner layout */
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
#include <net/sock.h>
G
Glauber Costa 已提交
400
#include <net/ip.h>
G
Glauber Costa 已提交
401 402 403 404

static bool mem_cgroup_is_root(struct mem_cgroup *memcg);
void sock_update_memcg(struct sock *sk)
{
405
	if (mem_cgroup_sockets_enabled) {
G
Glauber Costa 已提交
406 407 408 409
		struct mem_cgroup *memcg;

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

410 411 412 413 414 415 416 417 418 419 420 421 422 423
		/* 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
424 425 426 427 428 429 430 431 432 433 434 435 436
		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)
{
437
	if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
G
Glauber Costa 已提交
438 439 440 441 442 443
		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
444

445
#ifdef CONFIG_INET
G
Glauber Costa 已提交
446 447 448 449 450 451 452 453
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
454 455 456
#endif /* CONFIG_INET */
#endif /* CONFIG_CGROUP_MEM_RES_CTLR_KMEM */

457
static void drain_all_stock_async(struct mem_cgroup *memcg);
458

459
static struct mem_cgroup_per_zone *
460
mem_cgroup_zoneinfo(struct mem_cgroup *memcg, int nid, int zid)
461
{
462
	return &memcg->info.nodeinfo[nid]->zoneinfo[zid];
463 464
}

465
struct cgroup_subsys_state *mem_cgroup_css(struct mem_cgroup *memcg)
466
{
467
	return &memcg->css;
468 469
}

470
static struct mem_cgroup_per_zone *
471
page_cgroup_zoneinfo(struct mem_cgroup *memcg, struct page *page)
472
{
473 474
	int nid = page_to_nid(page);
	int zid = page_zonenum(page);
475

476
	return mem_cgroup_zoneinfo(memcg, nid, zid);
477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494
}

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
495
__mem_cgroup_insert_exceeded(struct mem_cgroup *memcg,
496
				struct mem_cgroup_per_zone *mz,
497 498
				struct mem_cgroup_tree_per_zone *mctz,
				unsigned long long new_usage_in_excess)
499 500 501 502 503 504 505 506
{
	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;

507 508 509
	mz->usage_in_excess = new_usage_in_excess;
	if (!mz->usage_in_excess)
		return;
510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525
	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;
526 527 528
}

static void
529
__mem_cgroup_remove_exceeded(struct mem_cgroup *memcg,
530 531 532 533 534 535 536 537 538
				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;
}

539
static void
540
mem_cgroup_remove_exceeded(struct mem_cgroup *memcg,
541 542 543 544
				struct mem_cgroup_per_zone *mz,
				struct mem_cgroup_tree_per_zone *mctz)
{
	spin_lock(&mctz->lock);
545
	__mem_cgroup_remove_exceeded(memcg, mz, mctz);
546 547 548 549
	spin_unlock(&mctz->lock);
}


550
static void mem_cgroup_update_tree(struct mem_cgroup *memcg, struct page *page)
551
{
552
	unsigned long long excess;
553 554
	struct mem_cgroup_per_zone *mz;
	struct mem_cgroup_tree_per_zone *mctz;
555 556
	int nid = page_to_nid(page);
	int zid = page_zonenum(page);
557 558 559
	mctz = soft_limit_tree_from_page(page);

	/*
560 561
	 * Necessary to update all ancestors when hierarchy is used.
	 * because their event counter is not touched.
562
	 */
563 564 565
	for (; memcg; memcg = parent_mem_cgroup(memcg)) {
		mz = mem_cgroup_zoneinfo(memcg, nid, zid);
		excess = res_counter_soft_limit_excess(&memcg->res);
566 567 568 569
		/*
		 * We have to update the tree if mz is on RB-tree or
		 * mem is over its softlimit.
		 */
570
		if (excess || mz->on_tree) {
571 572 573
			spin_lock(&mctz->lock);
			/* if on-tree, remove it */
			if (mz->on_tree)
574
				__mem_cgroup_remove_exceeded(memcg, mz, mctz);
575
			/*
576 577
			 * Insert again. mz->usage_in_excess will be updated.
			 * If excess is 0, no tree ops.
578
			 */
579
			__mem_cgroup_insert_exceeded(memcg, mz, mctz, excess);
580 581
			spin_unlock(&mctz->lock);
		}
582 583 584
	}
}

585
static void mem_cgroup_remove_from_trees(struct mem_cgroup *memcg)
586 587 588 589 590
{
	int node, zone;
	struct mem_cgroup_per_zone *mz;
	struct mem_cgroup_tree_per_zone *mctz;

B
Bob Liu 已提交
591
	for_each_node(node) {
592
		for (zone = 0; zone < MAX_NR_ZONES; zone++) {
593
			mz = mem_cgroup_zoneinfo(memcg, node, zone);
594
			mctz = soft_limit_tree_node_zone(node, zone);
595
			mem_cgroup_remove_exceeded(memcg, mz, mctz);
596 597 598 599
		}
	}
}

600 601 602 603
static struct mem_cgroup_per_zone *
__mem_cgroup_largest_soft_limit_node(struct mem_cgroup_tree_per_zone *mctz)
{
	struct rb_node *rightmost = NULL;
604
	struct mem_cgroup_per_zone *mz;
605 606

retry:
607
	mz = NULL;
608 609 610 611 612 613 614 615 616 617
	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.
	 */
618 619 620
	__mem_cgroup_remove_exceeded(mz->memcg, mz, mctz);
	if (!res_counter_soft_limit_excess(&mz->memcg->res) ||
		!css_tryget(&mz->memcg->css))
621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636
		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;
}

637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655
/*
 * 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.
 */
656
static long mem_cgroup_read_stat(struct mem_cgroup *memcg,
657
				 enum mem_cgroup_stat_index idx)
658
{
659
	long val = 0;
660 661
	int cpu;

662 663
	get_online_cpus();
	for_each_online_cpu(cpu)
664
		val += per_cpu(memcg->stat->count[idx], cpu);
665
#ifdef CONFIG_HOTPLUG_CPU
666 667 668
	spin_lock(&memcg->pcp_counter_lock);
	val += memcg->nocpu_base.count[idx];
	spin_unlock(&memcg->pcp_counter_lock);
669 670
#endif
	put_online_cpus();
671 672 673
	return val;
}

674
static void mem_cgroup_swap_statistics(struct mem_cgroup *memcg,
675 676 677
					 bool charge)
{
	int val = (charge) ? 1 : -1;
678
	this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_SWAPOUT], val);
679 680
}

681
static unsigned long mem_cgroup_read_events(struct mem_cgroup *memcg,
682 683 684 685 686 687
					    enum mem_cgroup_events_index idx)
{
	unsigned long val = 0;
	int cpu;

	for_each_online_cpu(cpu)
688
		val += per_cpu(memcg->stat->events[idx], cpu);
689
#ifdef CONFIG_HOTPLUG_CPU
690 691 692
	spin_lock(&memcg->pcp_counter_lock);
	val += memcg->nocpu_base.events[idx];
	spin_unlock(&memcg->pcp_counter_lock);
693 694 695 696
#endif
	return val;
}

697
static void mem_cgroup_charge_statistics(struct mem_cgroup *memcg,
698
					 bool anon, int nr_pages)
699
{
700 701
	preempt_disable();

702 703 704 705 706 707
	/*
	 * Here, RSS means 'mapped anon' and anon's SwapCache. Shmem/tmpfs is
	 * counted as CACHE even if it's on ANON LRU.
	 */
	if (anon)
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_RSS],
708
				nr_pages);
709
	else
710
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_CACHE],
711
				nr_pages);
712

713 714
	/* pagein of a big page is an event. So, ignore page size */
	if (nr_pages > 0)
715
		__this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGPGIN]);
716
	else {
717
		__this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGPGOUT]);
718 719
		nr_pages = -nr_pages; /* for event */
	}
720

721
	__this_cpu_add(memcg->stat->events[MEM_CGROUP_EVENTS_COUNT], nr_pages);
722

723
	preempt_enable();
724 725
}

726
unsigned long
727
mem_cgroup_zone_nr_lru_pages(struct mem_cgroup *memcg, int nid, int zid,
728
			unsigned int lru_mask)
729 730
{
	struct mem_cgroup_per_zone *mz;
H
Hugh Dickins 已提交
731
	enum lru_list lru;
732 733
	unsigned long ret = 0;

734
	mz = mem_cgroup_zoneinfo(memcg, nid, zid);
735

H
Hugh Dickins 已提交
736 737 738
	for_each_lru(lru) {
		if (BIT(lru) & lru_mask)
			ret += mz->lru_size[lru];
739 740 741 742 743
	}
	return ret;
}

static unsigned long
744
mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
745 746
			int nid, unsigned int lru_mask)
{
747 748 749
	u64 total = 0;
	int zid;

750
	for (zid = 0; zid < MAX_NR_ZONES; zid++)
751 752
		total += mem_cgroup_zone_nr_lru_pages(memcg,
						nid, zid, lru_mask);
753

754 755
	return total;
}
756

757
static unsigned long mem_cgroup_nr_lru_pages(struct mem_cgroup *memcg,
758
			unsigned int lru_mask)
759
{
760
	int nid;
761 762
	u64 total = 0;

763
	for_each_node_state(nid, N_HIGH_MEMORY)
764
		total += mem_cgroup_node_nr_lru_pages(memcg, nid, lru_mask);
765
	return total;
766 767
}

768 769
static bool mem_cgroup_event_ratelimit(struct mem_cgroup *memcg,
				       enum mem_cgroup_events_target target)
770 771 772
{
	unsigned long val, next;

773 774
	val = __this_cpu_read(memcg->stat->events[MEM_CGROUP_EVENTS_COUNT]);
	next = __this_cpu_read(memcg->stat->targets[target]);
775
	/* from time_after() in jiffies.h */
776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791
	if ((long)next - (long)val < 0) {
		switch (target) {
		case MEM_CGROUP_TARGET_THRESH:
			next = val + THRESHOLDS_EVENTS_TARGET;
			break;
		case MEM_CGROUP_TARGET_SOFTLIMIT:
			next = val + SOFTLIMIT_EVENTS_TARGET;
			break;
		case MEM_CGROUP_TARGET_NUMAINFO:
			next = val + NUMAINFO_EVENTS_TARGET;
			break;
		default:
			break;
		}
		__this_cpu_write(memcg->stat->targets[target], next);
		return true;
792
	}
793
	return false;
794 795 796 797 798 799
}

/*
 * Check events in order.
 *
 */
800
static void memcg_check_events(struct mem_cgroup *memcg, struct page *page)
801
{
802
	preempt_disable();
803
	/* threshold event is triggered in finer grain than soft limit */
804 805
	if (unlikely(mem_cgroup_event_ratelimit(memcg,
						MEM_CGROUP_TARGET_THRESH))) {
806 807
		bool do_softlimit;
		bool do_numainfo __maybe_unused;
808 809 810 811 812 813 814 815 816

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

817
		mem_cgroup_threshold(memcg);
818
		if (unlikely(do_softlimit))
819
			mem_cgroup_update_tree(memcg, page);
820
#if MAX_NUMNODES > 1
821
		if (unlikely(do_numainfo))
822
			atomic_inc(&memcg->numainfo_events);
823
#endif
824 825
	} else
		preempt_enable();
826 827
}

G
Glauber Costa 已提交
828
struct mem_cgroup *mem_cgroup_from_cont(struct cgroup *cont)
B
Balbir Singh 已提交
829 830 831 832 833 834
{
	return container_of(cgroup_subsys_state(cont,
				mem_cgroup_subsys_id), struct mem_cgroup,
				css);
}

835
struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p)
836
{
837 838 839 840 841 842 843 844
	/*
	 * 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;

845 846 847 848
	return container_of(task_subsys_state(p, mem_cgroup_subsys_id),
				struct mem_cgroup, css);
}

849
struct mem_cgroup *try_get_mem_cgroup_from_mm(struct mm_struct *mm)
850
{
851
	struct mem_cgroup *memcg = NULL;
852 853 854

	if (!mm)
		return NULL;
855 856 857 858 859 860 861
	/*
	 * 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 {
862 863
		memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
		if (unlikely(!memcg))
864
			break;
865
	} while (!css_tryget(&memcg->css));
866
	rcu_read_unlock();
867
	return memcg;
868 869
}

870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889
/**
 * 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
890
{
891 892
	struct mem_cgroup *memcg = NULL;
	int id = 0;
893

894 895 896
	if (mem_cgroup_disabled())
		return NULL;

897 898
	if (!root)
		root = root_mem_cgroup;
K
KAMEZAWA Hiroyuki 已提交
899

900 901
	if (prev && !reclaim)
		id = css_id(&prev->css);
K
KAMEZAWA Hiroyuki 已提交
902

903 904
	if (prev && prev != root)
		css_put(&prev->css);
K
KAMEZAWA Hiroyuki 已提交
905

906 907 908 909 910
	if (!root->use_hierarchy && root != root_mem_cgroup) {
		if (prev)
			return NULL;
		return root;
	}
K
KAMEZAWA Hiroyuki 已提交
911

912
	while (!memcg) {
913
		struct mem_cgroup_reclaim_iter *uninitialized_var(iter);
914
		struct cgroup_subsys_state *css;
915

916 917 918 919 920 921 922 923 924 925 926
		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
927

928 929 930 931 932 933 934 935
		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
936 937
		rcu_read_unlock();

938 939 940 941 942 943 944
		if (reclaim) {
			iter->position = id;
			if (!css)
				iter->generation++;
			else if (!prev && memcg)
				reclaim->generation = iter->generation;
		}
945 946 947 948 949

		if (prev && !css)
			return NULL;
	}
	return memcg;
K
KAMEZAWA Hiroyuki 已提交
950
}
K
KAMEZAWA Hiroyuki 已提交
951

952 953 954 955 956 957 958
/**
 * 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)
959 960 961 962 963 964
{
	if (!root)
		root = root_mem_cgroup;
	if (prev && prev != root)
		css_put(&prev->css);
}
K
KAMEZAWA Hiroyuki 已提交
965

966 967 968 969 970 971
/*
 * 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)		\
972
	for (iter = mem_cgroup_iter(root, NULL, NULL);	\
973
	     iter != NULL;				\
974
	     iter = mem_cgroup_iter(root, iter, NULL))
975

976
#define for_each_mem_cgroup(iter)			\
977
	for (iter = mem_cgroup_iter(NULL, NULL, NULL);	\
978
	     iter != NULL;				\
979
	     iter = mem_cgroup_iter(NULL, iter, NULL))
K
KAMEZAWA Hiroyuki 已提交
980

981
static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
982
{
983
	return (memcg == root_mem_cgroup);
984 985
}

986 987
void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
{
988
	struct mem_cgroup *memcg;
989 990 991 992 993

	if (!mm)
		return;

	rcu_read_lock();
994 995
	memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
	if (unlikely(!memcg))
996 997 998 999
		goto out;

	switch (idx) {
	case PGFAULT:
1000 1001 1002 1003
		this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGFAULT]);
		break;
	case PGMAJFAULT:
		this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT]);
1004 1005 1006 1007 1008 1009 1010 1011 1012
		break;
	default:
		BUG();
	}
out:
	rcu_read_unlock();
}
EXPORT_SYMBOL(mem_cgroup_count_vm_event);

1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033
/**
 * 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
1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046
/*
 * 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.
 */
1047

1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061
/**
 * 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
1062 1063
{
	struct mem_cgroup_per_zone *mz;
1064 1065
	struct mem_cgroup *memcg;
	struct page_cgroup *pc;
1066

1067
	if (mem_cgroup_disabled())
1068 1069
		return &zone->lruvec;

K
KAMEZAWA Hiroyuki 已提交
1070
	pc = lookup_page_cgroup(page);
1071
	memcg = pc->mem_cgroup;
1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084

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

1085 1086
	mz = page_cgroup_zoneinfo(memcg, page);
	/* compound_order() is stabilized through lru_lock */
1087
	mz->lru_size[lru] += 1 << compound_order(page);
1088
	return &mz->lruvec;
K
KAMEZAWA Hiroyuki 已提交
1089
}
1090

1091 1092 1093 1094 1095 1096 1097 1098 1099
/**
 * 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.
1100
 */
1101
void mem_cgroup_lru_del_list(struct page *page, enum lru_list lru)
1102 1103
{
	struct mem_cgroup_per_zone *mz;
1104
	struct mem_cgroup *memcg;
1105 1106 1107 1108 1109 1110
	struct page_cgroup *pc;

	if (mem_cgroup_disabled())
		return;

	pc = lookup_page_cgroup(page);
1111 1112
	memcg = pc->mem_cgroup;
	VM_BUG_ON(!memcg);
1113 1114
	mz = page_cgroup_zoneinfo(memcg, page);
	/* huge page split is done under lru_lock. so, we have no races. */
1115 1116
	VM_BUG_ON(mz->lru_size[lru] < (1 << compound_order(page)));
	mz->lru_size[lru] -= 1 << compound_order(page);
1117 1118
}

1119
void mem_cgroup_lru_del(struct page *page)
K
KAMEZAWA Hiroyuki 已提交
1120
{
1121
	mem_cgroup_lru_del_list(page, page_lru(page));
1122 1123
}

1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141
/**
 * 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)
1142
{
1143 1144 1145
	/* 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
1146
}
1147

1148
/*
1149
 * Checks whether given mem is same or in the root_mem_cgroup's
1150 1151
 * hierarchy subtree
 */
1152 1153
bool __mem_cgroup_same_or_subtree(const struct mem_cgroup *root_memcg,
				  struct mem_cgroup *memcg)
1154
{
1155 1156 1157 1158
	if (root_memcg == memcg)
		return true;
	if (!root_memcg->use_hierarchy)
		return false;
1159 1160 1161 1162 1163 1164 1165 1166
	return css_is_ancestor(&memcg->css, &root_memcg->css);
}

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

1167
	rcu_read_lock();
1168
	ret = __mem_cgroup_same_or_subtree(root_memcg, memcg);
1169 1170
	rcu_read_unlock();
	return ret;
1171 1172
}

1173
int task_in_mem_cgroup(struct task_struct *task, const struct mem_cgroup *memcg)
1174 1175
{
	int ret;
1176
	struct mem_cgroup *curr = NULL;
1177
	struct task_struct *p;
1178

1179
	p = find_lock_task_mm(task);
1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194
	if (p) {
		curr = try_get_mem_cgroup_from_mm(p->mm);
		task_unlock(p);
	} else {
		/*
		 * All threads may have already detached their mm's, but the oom
		 * killer still needs to detect if they have already been oom
		 * killed to prevent needlessly killing additional tasks.
		 */
		task_lock(task);
		curr = mem_cgroup_from_task(task);
		if (curr)
			css_get(&curr->css);
		task_unlock(task);
	}
1195 1196
	if (!curr)
		return 0;
1197
	/*
1198
	 * We should check use_hierarchy of "memcg" not "curr". Because checking
1199
	 * use_hierarchy of "curr" here make this function true if hierarchy is
1200 1201
	 * enabled in "curr" and "curr" is a child of "memcg" in *cgroup*
	 * hierarchy(even if use_hierarchy is disabled in "memcg").
1202
	 */
1203
	ret = mem_cgroup_same_or_subtree(memcg, curr);
1204
	css_put(&curr->css);
1205 1206 1207
	return ret;
}

1208
int mem_cgroup_inactive_anon_is_low(struct mem_cgroup *memcg, struct zone *zone)
1209
{
1210 1211 1212
	unsigned long inactive_ratio;
	int nid = zone_to_nid(zone);
	int zid = zone_idx(zone);
1213
	unsigned long inactive;
1214
	unsigned long active;
1215
	unsigned long gb;
1216

1217 1218 1219 1220
	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));
1221

1222 1223 1224 1225 1226 1227
	gb = (inactive + active) >> (30 - PAGE_SHIFT);
	if (gb)
		inactive_ratio = int_sqrt(10 * gb);
	else
		inactive_ratio = 1;

1228
	return inactive * inactive_ratio < active;
1229 1230
}

1231
int mem_cgroup_inactive_file_is_low(struct mem_cgroup *memcg, struct zone *zone)
1232 1233 1234
{
	unsigned long active;
	unsigned long inactive;
1235 1236
	int zid = zone_idx(zone);
	int nid = zone_to_nid(zone);
1237

1238 1239 1240 1241
	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));
1242 1243 1244 1245

	return (active > inactive);
}

K
KOSAKI Motohiro 已提交
1246 1247 1248
struct zone_reclaim_stat *mem_cgroup_get_reclaim_stat(struct mem_cgroup *memcg,
						      struct zone *zone)
{
1249
	int nid = zone_to_nid(zone);
K
KOSAKI Motohiro 已提交
1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265
	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);
1266 1267
	if (!PageCgroupUsed(pc))
		return NULL;
1268 1269
	/* Ensure pc->mem_cgroup is visible after reading PCG_USED. */
	smp_rmb();
1270
	mz = page_cgroup_zoneinfo(pc->mem_cgroup, page);
K
KOSAKI Motohiro 已提交
1271 1272 1273
	return &mz->reclaim_stat;
}

1274 1275 1276
#define mem_cgroup_from_res_counter(counter, member)	\
	container_of(counter, struct mem_cgroup, member)

1277
/**
1278 1279
 * mem_cgroup_margin - calculate chargeable space of a memory cgroup
 * @mem: the memory cgroup
1280
 *
1281
 * Returns the maximum amount of memory @mem can be charged with, in
1282
 * pages.
1283
 */
1284
static unsigned long mem_cgroup_margin(struct mem_cgroup *memcg)
1285
{
1286 1287
	unsigned long long margin;

1288
	margin = res_counter_margin(&memcg->res);
1289
	if (do_swap_account)
1290
		margin = min(margin, res_counter_margin(&memcg->memsw));
1291
	return margin >> PAGE_SHIFT;
1292 1293
}

1294
int mem_cgroup_swappiness(struct mem_cgroup *memcg)
K
KOSAKI Motohiro 已提交
1295 1296 1297 1298 1299 1300 1301
{
	struct cgroup *cgrp = memcg->css.cgroup;

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

1302
	return memcg->swappiness;
K
KOSAKI Motohiro 已提交
1303 1304
}

1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318
/*
 * memcg->moving_account is used for checking possibility that some thread is
 * calling move_account(). When a thread on CPU-A starts moving pages under
 * a memcg, other threads should check memcg->moving_account under
 * rcu_read_lock(), like this:
 *
 *         CPU-A                                    CPU-B
 *                                              rcu_read_lock()
 *         memcg->moving_account+1              if (memcg->mocing_account)
 *                                                   take heavy locks.
 *         synchronize_rcu()                    update something.
 *                                              rcu_read_unlock()
 *         start move here.
 */
1319 1320 1321 1322

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

1323
static void mem_cgroup_start_move(struct mem_cgroup *memcg)
1324
{
1325
	atomic_inc(&memcg_moving);
1326
	atomic_inc(&memcg->moving_account);
1327 1328 1329
	synchronize_rcu();
}

1330
static void mem_cgroup_end_move(struct mem_cgroup *memcg)
1331
{
1332 1333 1334 1335
	/*
	 * Now, mem_cgroup_clear_mc() may call this function with NULL.
	 * We check NULL in callee rather than caller.
	 */
1336 1337
	if (memcg) {
		atomic_dec(&memcg_moving);
1338
		atomic_dec(&memcg->moving_account);
1339
	}
1340
}
1341

1342 1343 1344
/*
 * 2 routines for checking "mem" is under move_account() or not.
 *
1345 1346
 * mem_cgroup_stolen() -  checking whether a cgroup is mc.from or not. This
 *			  is used for avoiding races in accounting.  If true,
1347 1348 1349 1350 1351 1352 1353
 *			  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".
 */

1354
static bool mem_cgroup_stolen(struct mem_cgroup *memcg)
1355 1356
{
	VM_BUG_ON(!rcu_read_lock_held());
1357
	return atomic_read(&memcg->moving_account) > 0;
1358
}
1359

1360
static bool mem_cgroup_under_move(struct mem_cgroup *memcg)
1361
{
1362 1363
	struct mem_cgroup *from;
	struct mem_cgroup *to;
1364
	bool ret = false;
1365 1366 1367 1368 1369 1370 1371 1372 1373
	/*
	 * 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;
1374

1375 1376
	ret = mem_cgroup_same_or_subtree(memcg, from)
		|| mem_cgroup_same_or_subtree(memcg, to);
1377 1378
unlock:
	spin_unlock(&mc.lock);
1379 1380 1381
	return ret;
}

1382
static bool mem_cgroup_wait_acct_move(struct mem_cgroup *memcg)
1383 1384
{
	if (mc.moving_task && current != mc.moving_task) {
1385
		if (mem_cgroup_under_move(memcg)) {
1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397
			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;
}

1398 1399 1400 1401
/*
 * Take this lock when
 * - a code tries to modify page's memcg while it's USED.
 * - a code tries to modify page state accounting in a memcg.
1402
 * see mem_cgroup_stolen(), too.
1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415
 */
static void move_lock_mem_cgroup(struct mem_cgroup *memcg,
				  unsigned long *flags)
{
	spin_lock_irqsave(&memcg->move_lock, *flags);
}

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

1416
/**
1417
 * mem_cgroup_print_oom_info: Called from OOM with tasklist_lock held in read mode.
1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435
 * @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;

1436
	if (!memcg || !p)
1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481
		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));
}

1482 1483 1484 1485
/*
 * This function returns the number of memcg under hierarchy tree. Returns
 * 1(self count) if no children.
 */
1486
static int mem_cgroup_count_children(struct mem_cgroup *memcg)
1487 1488
{
	int num = 0;
K
KAMEZAWA Hiroyuki 已提交
1489 1490
	struct mem_cgroup *iter;

1491
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
1492
		num++;
1493 1494 1495
	return num;
}

D
David Rientjes 已提交
1496 1497 1498 1499 1500 1501 1502 1503
/*
 * Return the memory (and swap, if configured) limit for a memcg.
 */
u64 mem_cgroup_get_limit(struct mem_cgroup *memcg)
{
	u64 limit;
	u64 memsw;

1504 1505 1506
	limit = res_counter_read_u64(&memcg->res, RES_LIMIT);
	limit += total_swap_pages << PAGE_SHIFT;

D
David Rientjes 已提交
1507 1508 1509 1510 1511 1512 1513 1514
	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);
}

1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550
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;
}

1551 1552 1553 1554 1555 1556 1557 1558 1559 1560
/**
 * 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.
 */
1561
static bool test_mem_cgroup_node_reclaimable(struct mem_cgroup *memcg,
1562 1563
		int nid, bool noswap)
{
1564
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_FILE))
1565 1566 1567
		return true;
	if (noswap || !total_swap_pages)
		return false;
1568
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_ANON))
1569 1570 1571 1572
		return true;
	return false;

}
1573 1574 1575 1576 1577 1578 1579 1580
#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.
 *
 */
1581
static void mem_cgroup_may_update_nodemask(struct mem_cgroup *memcg)
1582 1583
{
	int nid;
1584 1585 1586 1587
	/*
	 * numainfo_events > 0 means there was at least NUMAINFO_EVENTS_TARGET
	 * pagein/pageout changes since the last update.
	 */
1588
	if (!atomic_read(&memcg->numainfo_events))
1589
		return;
1590
	if (atomic_inc_return(&memcg->numainfo_updating) > 1)
1591 1592 1593
		return;

	/* make a nodemask where this memcg uses memory from */
1594
	memcg->scan_nodes = node_states[N_HIGH_MEMORY];
1595 1596 1597

	for_each_node_mask(nid, node_states[N_HIGH_MEMORY]) {

1598 1599
		if (!test_mem_cgroup_node_reclaimable(memcg, nid, false))
			node_clear(nid, memcg->scan_nodes);
1600
	}
1601

1602 1603
	atomic_set(&memcg->numainfo_events, 0);
	atomic_set(&memcg->numainfo_updating, 0);
1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617
}

/*
 * 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.
 */
1618
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
1619 1620 1621
{
	int node;

1622 1623
	mem_cgroup_may_update_nodemask(memcg);
	node = memcg->last_scanned_node;
1624

1625
	node = next_node(node, memcg->scan_nodes);
1626
	if (node == MAX_NUMNODES)
1627
		node = first_node(memcg->scan_nodes);
1628 1629 1630 1631 1632 1633 1634 1635 1636
	/*
	 * 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();

1637
	memcg->last_scanned_node = node;
1638 1639 1640
	return node;
}

1641 1642 1643 1644 1645 1646
/*
 * 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.
 */
1647
bool mem_cgroup_reclaimable(struct mem_cgroup *memcg, bool noswap)
1648 1649 1650 1651 1652 1653 1654
{
	int nid;

	/*
	 * quick check...making use of scan_node.
	 * We can skip unused nodes.
	 */
1655 1656
	if (!nodes_empty(memcg->scan_nodes)) {
		for (nid = first_node(memcg->scan_nodes);
1657
		     nid < MAX_NUMNODES;
1658
		     nid = next_node(nid, memcg->scan_nodes)) {
1659

1660
			if (test_mem_cgroup_node_reclaimable(memcg, nid, noswap))
1661 1662 1663 1664 1665 1666 1667
				return true;
		}
	}
	/*
	 * Check rest of nodes.
	 */
	for_each_node_state(nid, N_HIGH_MEMORY) {
1668
		if (node_isset(nid, memcg->scan_nodes))
1669
			continue;
1670
		if (test_mem_cgroup_node_reclaimable(memcg, nid, noswap))
1671 1672 1673 1674 1675
			return true;
	}
	return false;
}

1676
#else
1677
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
1678 1679 1680
{
	return 0;
}
1681

1682
bool mem_cgroup_reclaimable(struct mem_cgroup *memcg, bool noswap)
1683
{
1684
	return test_mem_cgroup_node_reclaimable(memcg, 0, noswap);
1685
}
1686 1687
#endif

1688 1689 1690 1691
static int mem_cgroup_soft_reclaim(struct mem_cgroup *root_memcg,
				   struct zone *zone,
				   gfp_t gfp_mask,
				   unsigned long *total_scanned)
1692
{
1693
	struct mem_cgroup *victim = NULL;
1694
	int total = 0;
K
KAMEZAWA Hiroyuki 已提交
1695
	int loop = 0;
1696
	unsigned long excess;
1697
	unsigned long nr_scanned;
1698 1699 1700 1701
	struct mem_cgroup_reclaim_cookie reclaim = {
		.zone = zone,
		.priority = 0,
	};
1702

1703
	excess = res_counter_soft_limit_excess(&root_memcg->res) >> PAGE_SHIFT;
K
KAMEZAWA Hiroyuki 已提交
1704

1705
	while (1) {
1706
		victim = mem_cgroup_iter(root_memcg, victim, &reclaim);
1707
		if (!victim) {
K
KAMEZAWA Hiroyuki 已提交
1708
			loop++;
1709 1710 1711 1712 1713 1714
			if (loop >= 2) {
				/*
				 * If we have not been able to reclaim
				 * anything, it might because there are
				 * no reclaimable pages under this hierarchy
				 */
1715
				if (!total)
1716 1717
					break;
				/*
L
Lucas De Marchi 已提交
1718
				 * We want to do more targeted reclaim.
1719 1720 1721 1722 1723
				 * 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) ||
1724
					(loop > MEM_CGROUP_MAX_RECLAIM_LOOPS))
1725 1726
					break;
			}
1727
			continue;
1728
		}
1729
		if (!mem_cgroup_reclaimable(victim, false))
1730
			continue;
1731 1732 1733 1734
		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))
1735
			break;
1736
	}
1737
	mem_cgroup_iter_break(root_memcg, victim);
K
KAMEZAWA Hiroyuki 已提交
1738
	return total;
1739 1740
}

K
KAMEZAWA Hiroyuki 已提交
1741 1742 1743
/*
 * Check OOM-Killer is already running under our hierarchy.
 * If someone is running, return false.
1744
 * Has to be called with memcg_oom_lock
K
KAMEZAWA Hiroyuki 已提交
1745
 */
1746
static bool mem_cgroup_oom_lock(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
1747
{
1748
	struct mem_cgroup *iter, *failed = NULL;
1749

1750
	for_each_mem_cgroup_tree(iter, memcg) {
1751
		if (iter->oom_lock) {
1752 1753 1754 1755 1756
			/*
			 * this subtree of our hierarchy is already locked
			 * so we cannot give a lock.
			 */
			failed = iter;
1757 1758
			mem_cgroup_iter_break(memcg, iter);
			break;
1759 1760
		} else
			iter->oom_lock = true;
K
KAMEZAWA Hiroyuki 已提交
1761
	}
K
KAMEZAWA Hiroyuki 已提交
1762

1763
	if (!failed)
1764
		return true;
1765 1766 1767 1768 1769

	/*
	 * OK, we failed to lock the whole subtree so we have to clean up
	 * what we set up to the failing subtree
	 */
1770
	for_each_mem_cgroup_tree(iter, memcg) {
1771
		if (iter == failed) {
1772 1773
			mem_cgroup_iter_break(memcg, iter);
			break;
1774 1775 1776
		}
		iter->oom_lock = false;
	}
1777
	return false;
1778
}
1779

1780
/*
1781
 * Has to be called with memcg_oom_lock
1782
 */
1783
static int mem_cgroup_oom_unlock(struct mem_cgroup *memcg)
1784
{
K
KAMEZAWA Hiroyuki 已提交
1785 1786
	struct mem_cgroup *iter;

1787
	for_each_mem_cgroup_tree(iter, memcg)
1788 1789 1790 1791
		iter->oom_lock = false;
	return 0;
}

1792
static void mem_cgroup_mark_under_oom(struct mem_cgroup *memcg)
1793 1794 1795
{
	struct mem_cgroup *iter;

1796
	for_each_mem_cgroup_tree(iter, memcg)
1797 1798 1799
		atomic_inc(&iter->under_oom);
}

1800
static void mem_cgroup_unmark_under_oom(struct mem_cgroup *memcg)
1801 1802 1803
{
	struct mem_cgroup *iter;

K
KAMEZAWA Hiroyuki 已提交
1804 1805 1806 1807 1808
	/*
	 * 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.
	 */
1809
	for_each_mem_cgroup_tree(iter, memcg)
1810
		atomic_add_unless(&iter->under_oom, -1, 0);
1811 1812
}

1813
static DEFINE_SPINLOCK(memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
1814 1815
static DECLARE_WAIT_QUEUE_HEAD(memcg_oom_waitq);

K
KAMEZAWA Hiroyuki 已提交
1816
struct oom_wait_info {
1817
	struct mem_cgroup *memcg;
K
KAMEZAWA Hiroyuki 已提交
1818 1819 1820 1821 1822 1823
	wait_queue_t	wait;
};

static int memcg_oom_wake_function(wait_queue_t *wait,
	unsigned mode, int sync, void *arg)
{
1824 1825
	struct mem_cgroup *wake_memcg = (struct mem_cgroup *)arg;
	struct mem_cgroup *oom_wait_memcg;
K
KAMEZAWA Hiroyuki 已提交
1826 1827 1828
	struct oom_wait_info *oom_wait_info;

	oom_wait_info = container_of(wait, struct oom_wait_info, wait);
1829
	oom_wait_memcg = oom_wait_info->memcg;
K
KAMEZAWA Hiroyuki 已提交
1830 1831

	/*
1832
	 * Both of oom_wait_info->memcg and wake_memcg are stable under us.
K
KAMEZAWA Hiroyuki 已提交
1833 1834
	 * Then we can use css_is_ancestor without taking care of RCU.
	 */
1835 1836
	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
1837 1838 1839 1840
		return 0;
	return autoremove_wake_function(wait, mode, sync, arg);
}

1841
static void memcg_wakeup_oom(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
1842
{
1843 1844
	/* for filtering, pass "memcg" as argument. */
	__wake_up(&memcg_oom_waitq, TASK_NORMAL, 0, memcg);
K
KAMEZAWA Hiroyuki 已提交
1845 1846
}

1847
static void memcg_oom_recover(struct mem_cgroup *memcg)
1848
{
1849 1850
	if (memcg && atomic_read(&memcg->under_oom))
		memcg_wakeup_oom(memcg);
1851 1852
}

K
KAMEZAWA Hiroyuki 已提交
1853 1854 1855
/*
 * try to call OOM killer. returns false if we should exit memory-reclaim loop.
 */
1856
bool mem_cgroup_handle_oom(struct mem_cgroup *memcg, gfp_t mask, int order)
1857
{
K
KAMEZAWA Hiroyuki 已提交
1858
	struct oom_wait_info owait;
1859
	bool locked, need_to_kill;
K
KAMEZAWA Hiroyuki 已提交
1860

1861
	owait.memcg = memcg;
K
KAMEZAWA Hiroyuki 已提交
1862 1863 1864 1865
	owait.wait.flags = 0;
	owait.wait.func = memcg_oom_wake_function;
	owait.wait.private = current;
	INIT_LIST_HEAD(&owait.wait.task_list);
1866
	need_to_kill = true;
1867
	mem_cgroup_mark_under_oom(memcg);
1868

1869
	/* At first, try to OOM lock hierarchy under memcg.*/
1870
	spin_lock(&memcg_oom_lock);
1871
	locked = mem_cgroup_oom_lock(memcg);
K
KAMEZAWA Hiroyuki 已提交
1872 1873 1874 1875 1876
	/*
	 * 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.
	 */
1877
	prepare_to_wait(&memcg_oom_waitq, &owait.wait, TASK_KILLABLE);
1878
	if (!locked || memcg->oom_kill_disable)
1879 1880
		need_to_kill = false;
	if (locked)
1881
		mem_cgroup_oom_notify(memcg);
1882
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
1883

1884 1885
	if (need_to_kill) {
		finish_wait(&memcg_oom_waitq, &owait.wait);
1886
		mem_cgroup_out_of_memory(memcg, mask, order);
1887
	} else {
K
KAMEZAWA Hiroyuki 已提交
1888
		schedule();
K
KAMEZAWA Hiroyuki 已提交
1889
		finish_wait(&memcg_oom_waitq, &owait.wait);
K
KAMEZAWA Hiroyuki 已提交
1890
	}
1891
	spin_lock(&memcg_oom_lock);
1892
	if (locked)
1893 1894
		mem_cgroup_oom_unlock(memcg);
	memcg_wakeup_oom(memcg);
1895
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
1896

1897
	mem_cgroup_unmark_under_oom(memcg);
1898

K
KAMEZAWA Hiroyuki 已提交
1899 1900 1901
	if (test_thread_flag(TIF_MEMDIE) || fatal_signal_pending(current))
		return false;
	/* Give chance to dying process */
1902
	schedule_timeout_uninterruptible(1);
K
KAMEZAWA Hiroyuki 已提交
1903
	return true;
1904 1905
}

1906 1907 1908
/*
 * Currently used to update mapped file statistics, but the routine can be
 * generalized to update other statistics as well.
1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925
 *
 * 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
1926 1927
 * small, we check mm->moving_account and detect there are possibility of race
 * If there is, we take a lock.
1928
 */
1929

1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944
void __mem_cgroup_begin_update_page_stat(struct page *page,
				bool *locked, unsigned long *flags)
{
	struct mem_cgroup *memcg;
	struct page_cgroup *pc;

	pc = lookup_page_cgroup(page);
again:
	memcg = pc->mem_cgroup;
	if (unlikely(!memcg || !PageCgroupUsed(pc)))
		return;
	/*
	 * If this memory cgroup is not under account moving, we don't
	 * need to take move_lock_page_cgroup(). Because we already hold
	 * rcu_read_lock(), any calls to move_account will be delayed until
1945
	 * rcu_read_unlock() if mem_cgroup_stolen() == true.
1946
	 */
1947
	if (!mem_cgroup_stolen(memcg))
1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969
		return;

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

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

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

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

1977
	if (mem_cgroup_disabled())
1978
		return;
1979

1980 1981
	memcg = pc->mem_cgroup;
	if (unlikely(!memcg || !PageCgroupUsed(pc)))
1982
		return;
1983 1984

	switch (idx) {
1985 1986
	case MEMCG_NR_FILE_MAPPED:
		idx = MEM_CGROUP_STAT_FILE_MAPPED;
1987 1988 1989
		break;
	default:
		BUG();
1990
	}
1991

1992
	this_cpu_add(memcg->stat->count[idx], val);
1993
}
1994

1995 1996 1997 1998
/*
 * size of first charge trial. "32" comes from vmscan.c's magic value.
 * TODO: maybe necessary to use big numbers in big irons.
 */
1999
#define CHARGE_BATCH	32U
2000 2001
struct memcg_stock_pcp {
	struct mem_cgroup *cached; /* this never be root cgroup */
2002
	unsigned int nr_pages;
2003
	struct work_struct work;
2004 2005
	unsigned long flags;
#define FLUSHING_CACHED_CHARGE	(0)
2006 2007
};
static DEFINE_PER_CPU(struct memcg_stock_pcp, memcg_stock);
2008
static DEFINE_MUTEX(percpu_charge_mutex);
2009 2010

/*
2011
 * Try to consume stocked charge on this cpu. If success, one page is consumed
2012 2013 2014 2015
 * 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.
 */
2016
static bool consume_stock(struct mem_cgroup *memcg)
2017 2018 2019 2020 2021
{
	struct memcg_stock_pcp *stock;
	bool ret = true;

	stock = &get_cpu_var(memcg_stock);
2022
	if (memcg == stock->cached && stock->nr_pages)
2023
		stock->nr_pages--;
2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036
	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;

2037 2038 2039 2040
	if (stock->nr_pages) {
		unsigned long bytes = stock->nr_pages * PAGE_SIZE;

		res_counter_uncharge(&old->res, bytes);
2041
		if (do_swap_account)
2042 2043
			res_counter_uncharge(&old->memsw, bytes);
		stock->nr_pages = 0;
2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055
	}
	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);
2056
	clear_bit(FLUSHING_CACHED_CHARGE, &stock->flags);
2057 2058 2059 2060
}

/*
 * Cache charges(val) which is from res_counter, to local per_cpu area.
2061
 * This will be consumed by consume_stock() function, later.
2062
 */
2063
static void refill_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
2064 2065 2066
{
	struct memcg_stock_pcp *stock = &get_cpu_var(memcg_stock);

2067
	if (stock->cached != memcg) { /* reset if necessary */
2068
		drain_stock(stock);
2069
		stock->cached = memcg;
2070
	}
2071
	stock->nr_pages += nr_pages;
2072 2073 2074 2075
	put_cpu_var(memcg_stock);
}

/*
2076
 * Drains all per-CPU charge caches for given root_memcg resp. subtree
2077 2078
 * of the hierarchy under it. sync flag says whether we should block
 * until the work is done.
2079
 */
2080
static void drain_all_stock(struct mem_cgroup *root_memcg, bool sync)
2081
{
2082
	int cpu, curcpu;
2083

2084 2085
	/* Notify other cpus that system-wide "drain" is running */
	get_online_cpus();
2086
	curcpu = get_cpu();
2087 2088
	for_each_online_cpu(cpu) {
		struct memcg_stock_pcp *stock = &per_cpu(memcg_stock, cpu);
2089
		struct mem_cgroup *memcg;
2090

2091 2092
		memcg = stock->cached;
		if (!memcg || !stock->nr_pages)
2093
			continue;
2094
		if (!mem_cgroup_same_or_subtree(root_memcg, memcg))
2095
			continue;
2096 2097 2098 2099 2100 2101
		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);
		}
2102
	}
2103
	put_cpu();
2104 2105 2106 2107 2108 2109

	if (!sync)
		goto out;

	for_each_online_cpu(cpu) {
		struct memcg_stock_pcp *stock = &per_cpu(memcg_stock, cpu);
2110
		if (test_bit(FLUSHING_CACHED_CHARGE, &stock->flags))
2111 2112 2113
			flush_work(&stock->work);
	}
out:
2114
 	put_online_cpus();
2115 2116 2117 2118 2119 2120 2121 2122
}

/*
 * 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.
 */
2123
static void drain_all_stock_async(struct mem_cgroup *root_memcg)
2124
{
2125 2126 2127 2128 2129
	/*
	 * If someone calls draining, avoid adding more kworker runs.
	 */
	if (!mutex_trylock(&percpu_charge_mutex))
		return;
2130
	drain_all_stock(root_memcg, false);
2131
	mutex_unlock(&percpu_charge_mutex);
2132 2133 2134
}

/* This is a synchronous drain interface. */
2135
static void drain_all_stock_sync(struct mem_cgroup *root_memcg)
2136 2137
{
	/* called when force_empty is called */
2138
	mutex_lock(&percpu_charge_mutex);
2139
	drain_all_stock(root_memcg, true);
2140
	mutex_unlock(&percpu_charge_mutex);
2141 2142
}

2143 2144 2145 2146
/*
 * This function drains percpu counter value from DEAD cpu and
 * move it to local cpu. Note that this function can be preempted.
 */
2147
static void mem_cgroup_drain_pcp_counter(struct mem_cgroup *memcg, int cpu)
2148 2149 2150
{
	int i;

2151
	spin_lock(&memcg->pcp_counter_lock);
2152
	for (i = 0; i < MEM_CGROUP_STAT_DATA; i++) {
2153
		long x = per_cpu(memcg->stat->count[i], cpu);
2154

2155 2156
		per_cpu(memcg->stat->count[i], cpu) = 0;
		memcg->nocpu_base.count[i] += x;
2157
	}
2158
	for (i = 0; i < MEM_CGROUP_EVENTS_NSTATS; i++) {
2159
		unsigned long x = per_cpu(memcg->stat->events[i], cpu);
2160

2161 2162
		per_cpu(memcg->stat->events[i], cpu) = 0;
		memcg->nocpu_base.events[i] += x;
2163
	}
2164
	spin_unlock(&memcg->pcp_counter_lock);
2165 2166 2167
}

static int __cpuinit memcg_cpu_hotplug_callback(struct notifier_block *nb,
2168 2169 2170 2171 2172
					unsigned long action,
					void *hcpu)
{
	int cpu = (unsigned long)hcpu;
	struct memcg_stock_pcp *stock;
2173
	struct mem_cgroup *iter;
2174

2175
	if (action == CPU_ONLINE)
2176 2177
		return NOTIFY_OK;

2178
	if (action != CPU_DEAD && action != CPU_DEAD_FROZEN)
2179
		return NOTIFY_OK;
2180

2181
	for_each_mem_cgroup(iter)
2182 2183
		mem_cgroup_drain_pcp_counter(iter, cpu);

2184 2185 2186 2187 2188
	stock = &per_cpu(memcg_stock, cpu);
	drain_stock(stock);
	return NOTIFY_OK;
}

2189 2190 2191 2192 2193 2194 2195 2196 2197 2198

/* 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 */
};

2199
static int mem_cgroup_do_charge(struct mem_cgroup *memcg, gfp_t gfp_mask,
2200
				unsigned int nr_pages, bool oom_check)
2201
{
2202
	unsigned long csize = nr_pages * PAGE_SIZE;
2203 2204 2205 2206 2207
	struct mem_cgroup *mem_over_limit;
	struct res_counter *fail_res;
	unsigned long flags = 0;
	int ret;

2208
	ret = res_counter_charge(&memcg->res, csize, &fail_res);
2209 2210 2211 2212

	if (likely(!ret)) {
		if (!do_swap_account)
			return CHARGE_OK;
2213
		ret = res_counter_charge(&memcg->memsw, csize, &fail_res);
2214 2215 2216
		if (likely(!ret))
			return CHARGE_OK;

2217
		res_counter_uncharge(&memcg->res, csize);
2218 2219 2220 2221
		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);
2222
	/*
2223 2224
	 * nr_pages can be either a huge page (HPAGE_PMD_NR), a batch
	 * of regular pages (CHARGE_BATCH), or a single regular page (1).
2225 2226 2227 2228
	 *
	 * Never reclaim on behalf of optional batching, retry with a
	 * single page instead.
	 */
2229
	if (nr_pages == CHARGE_BATCH)
2230 2231 2232 2233 2234
		return CHARGE_RETRY;

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

2235
	ret = mem_cgroup_reclaim(mem_over_limit, gfp_mask, flags);
2236
	if (mem_cgroup_margin(mem_over_limit) >= nr_pages)
2237
		return CHARGE_RETRY;
2238
	/*
2239 2240 2241 2242 2243 2244 2245
	 * 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.
2246
	 */
2247
	if (nr_pages == 1 && ret)
2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260
		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 */
2261
	if (!mem_cgroup_handle_oom(mem_over_limit, gfp_mask, get_order(csize)))
2262 2263 2264 2265 2266
		return CHARGE_OOM_DIE;

	return CHARGE_RETRY;
}

2267
/*
2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286
 * __mem_cgroup_try_charge() does
 * 1. detect memcg to be charged against from passed *mm and *ptr,
 * 2. update res_counter
 * 3. call memory reclaim if necessary.
 *
 * In some special case, if the task is fatal, fatal_signal_pending() or
 * has TIF_MEMDIE, this function returns -EINTR while writing root_mem_cgroup
 * to *ptr. There are two reasons for this. 1: fatal threads should quit as soon
 * as possible without any hazards. 2: all pages should have a valid
 * pc->mem_cgroup. If mm is NULL and the caller doesn't pass a valid memcg
 * pointer, that is treated as a charge to root_mem_cgroup.
 *
 * So __mem_cgroup_try_charge() will return
 *  0       ...  on success, filling *ptr with a valid memcg pointer.
 *  -ENOMEM ...  charge failure because of resource limits.
 *  -EINTR  ...  if thread is fatal. *ptr is filled with root_mem_cgroup.
 *
 * Unlike the exported interface, an "oom" parameter is added. if oom==true,
 * the oom-killer can be invoked.
2287
 */
2288
static int __mem_cgroup_try_charge(struct mm_struct *mm,
A
Andrea Arcangeli 已提交
2289
				   gfp_t gfp_mask,
2290
				   unsigned int nr_pages,
2291
				   struct mem_cgroup **ptr,
2292
				   bool oom)
2293
{
2294
	unsigned int batch = max(CHARGE_BATCH, nr_pages);
2295
	int nr_oom_retries = MEM_CGROUP_RECLAIM_RETRIES;
2296
	struct mem_cgroup *memcg = NULL;
2297
	int ret;
2298

K
KAMEZAWA Hiroyuki 已提交
2299 2300 2301 2302 2303 2304 2305 2306
	/*
	 * 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;
2307

2308
	/*
2309 2310
	 * We always charge the cgroup the mm_struct belongs to.
	 * The mm_struct's mem_cgroup changes on task migration if the
2311 2312 2313
	 * thread group leader migrates. It's possible that mm is not
	 * set, if so charge the init_mm (happens for pagecache usage).
	 */
2314
	if (!*ptr && !mm)
2315
		*ptr = root_mem_cgroup;
K
KAMEZAWA Hiroyuki 已提交
2316
again:
2317 2318 2319 2320
	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
2321
			goto done;
2322
		if (nr_pages == 1 && consume_stock(memcg))
K
KAMEZAWA Hiroyuki 已提交
2323
			goto done;
2324
		css_get(&memcg->css);
2325
	} else {
K
KAMEZAWA Hiroyuki 已提交
2326
		struct task_struct *p;
2327

K
KAMEZAWA Hiroyuki 已提交
2328 2329 2330
		rcu_read_lock();
		p = rcu_dereference(mm->owner);
		/*
2331
		 * Because we don't have task_lock(), "p" can exit.
2332
		 * In that case, "memcg" can point to root or p can be NULL with
2333 2334 2335 2336 2337 2338
		 * 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
2339
		 */
2340
		memcg = mem_cgroup_from_task(p);
2341 2342 2343
		if (!memcg)
			memcg = root_mem_cgroup;
		if (mem_cgroup_is_root(memcg)) {
K
KAMEZAWA Hiroyuki 已提交
2344 2345 2346
			rcu_read_unlock();
			goto done;
		}
2347
		if (nr_pages == 1 && consume_stock(memcg)) {
K
KAMEZAWA Hiroyuki 已提交
2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359
			/*
			 * 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 */
2360
		if (!css_tryget(&memcg->css)) {
K
KAMEZAWA Hiroyuki 已提交
2361 2362 2363 2364 2365
			rcu_read_unlock();
			goto again;
		}
		rcu_read_unlock();
	}
2366

2367 2368
	do {
		bool oom_check;
2369

2370
		/* If killed, bypass charge */
K
KAMEZAWA Hiroyuki 已提交
2371
		if (fatal_signal_pending(current)) {
2372
			css_put(&memcg->css);
2373
			goto bypass;
K
KAMEZAWA Hiroyuki 已提交
2374
		}
2375

2376 2377 2378 2379
		oom_check = false;
		if (oom && !nr_oom_retries) {
			oom_check = true;
			nr_oom_retries = MEM_CGROUP_RECLAIM_RETRIES;
2380
		}
2381

2382
		ret = mem_cgroup_do_charge(memcg, gfp_mask, batch, oom_check);
2383 2384 2385 2386
		switch (ret) {
		case CHARGE_OK:
			break;
		case CHARGE_RETRY: /* not in OOM situation but retry */
2387
			batch = nr_pages;
2388 2389
			css_put(&memcg->css);
			memcg = NULL;
K
KAMEZAWA Hiroyuki 已提交
2390
			goto again;
2391
		case CHARGE_WOULDBLOCK: /* !__GFP_WAIT */
2392
			css_put(&memcg->css);
2393 2394
			goto nomem;
		case CHARGE_NOMEM: /* OOM routine works */
K
KAMEZAWA Hiroyuki 已提交
2395
			if (!oom) {
2396
				css_put(&memcg->css);
K
KAMEZAWA Hiroyuki 已提交
2397
				goto nomem;
K
KAMEZAWA Hiroyuki 已提交
2398
			}
2399 2400 2401 2402
			/* If oom, we never return -ENOMEM */
			nr_oom_retries--;
			break;
		case CHARGE_OOM_DIE: /* Killed by OOM Killer */
2403
			css_put(&memcg->css);
K
KAMEZAWA Hiroyuki 已提交
2404
			goto bypass;
2405
		}
2406 2407
	} while (ret != CHARGE_OK);

2408
	if (batch > nr_pages)
2409 2410
		refill_stock(memcg, batch - nr_pages);
	css_put(&memcg->css);
2411
done:
2412
	*ptr = memcg;
2413 2414
	return 0;
nomem:
2415
	*ptr = NULL;
2416
	return -ENOMEM;
K
KAMEZAWA Hiroyuki 已提交
2417
bypass:
2418 2419
	*ptr = root_mem_cgroup;
	return -EINTR;
2420
}
2421

2422 2423 2424 2425 2426
/*
 * 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().
 */
2427
static void __mem_cgroup_cancel_charge(struct mem_cgroup *memcg,
2428
				       unsigned int nr_pages)
2429
{
2430
	if (!mem_cgroup_is_root(memcg)) {
2431 2432
		unsigned long bytes = nr_pages * PAGE_SIZE;

2433
		res_counter_uncharge(&memcg->res, bytes);
2434
		if (do_swap_account)
2435
			res_counter_uncharge(&memcg->memsw, bytes);
2436
	}
2437 2438
}

2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457
/*
 * 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);
}

2458
struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page)
2459
{
2460
	struct mem_cgroup *memcg = NULL;
2461
	struct page_cgroup *pc;
2462
	unsigned short id;
2463 2464
	swp_entry_t ent;

2465 2466 2467
	VM_BUG_ON(!PageLocked(page));

	pc = lookup_page_cgroup(page);
2468
	lock_page_cgroup(pc);
2469
	if (PageCgroupUsed(pc)) {
2470 2471 2472
		memcg = pc->mem_cgroup;
		if (memcg && !css_tryget(&memcg->css))
			memcg = NULL;
2473
	} else if (PageSwapCache(page)) {
2474
		ent.val = page_private(page);
2475
		id = lookup_swap_cgroup_id(ent);
2476
		rcu_read_lock();
2477 2478 2479
		memcg = mem_cgroup_lookup(id);
		if (memcg && !css_tryget(&memcg->css))
			memcg = NULL;
2480
		rcu_read_unlock();
2481
	}
2482
	unlock_page_cgroup(pc);
2483
	return memcg;
2484 2485
}

2486
static void __mem_cgroup_commit_charge(struct mem_cgroup *memcg,
2487
				       struct page *page,
2488
				       unsigned int nr_pages,
2489 2490
				       enum charge_type ctype,
				       bool lrucare)
2491
{
2492
	struct page_cgroup *pc = lookup_page_cgroup(page);
2493 2494
	struct zone *uninitialized_var(zone);
	bool was_on_lru = false;
2495
	bool anon;
2496

2497 2498 2499
	lock_page_cgroup(pc);
	if (unlikely(PageCgroupUsed(pc))) {
		unlock_page_cgroup(pc);
2500
		__mem_cgroup_cancel_charge(memcg, nr_pages);
2501 2502 2503 2504 2505 2506
		return;
	}
	/*
	 * we don't need page_cgroup_lock about tail pages, becase they are not
	 * accessed by any other context at this point.
	 */
2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521

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

2522
	pc->mem_cgroup = memcg;
2523 2524 2525 2526 2527 2528 2529
	/*
	 * 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
2530
	smp_wmb();
2531
	SetPageCgroupUsed(pc);
2532

2533 2534 2535 2536 2537 2538 2539 2540 2541
	if (lrucare) {
		if (was_on_lru) {
			VM_BUG_ON(PageLRU(page));
			SetPageLRU(page);
			add_page_to_lru_list(zone, page, page_lru(page));
		}
		spin_unlock_irq(&zone->lru_lock);
	}

2542 2543 2544 2545 2546 2547
	if (ctype == MEM_CGROUP_CHARGE_TYPE_MAPPED)
		anon = true;
	else
		anon = false;

	mem_cgroup_charge_statistics(memcg, anon, nr_pages);
2548
	unlock_page_cgroup(pc);
2549

2550 2551 2552 2553 2554
	/*
	 * "charge_statistics" updated event counter. Then, check it.
	 * Insert ancestor (and ancestor's ancestors), to softlimit RB-tree.
	 * if they exceeds softlimit.
	 */
2555
	memcg_check_events(memcg, page);
2556
}
2557

2558 2559
#ifdef CONFIG_TRANSPARENT_HUGEPAGE

2560
#define PCGF_NOCOPY_AT_SPLIT ((1 << PCG_LOCK) | (1 << PCG_MIGRATION))
2561 2562
/*
 * Because tail pages are not marked as "used", set it. We're under
2563 2564 2565
 * 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.
2566
 */
2567
void mem_cgroup_split_huge_fixup(struct page *head)
2568 2569
{
	struct page_cgroup *head_pc = lookup_page_cgroup(head);
2570 2571
	struct page_cgroup *pc;
	int i;
2572

2573 2574
	if (mem_cgroup_disabled())
		return;
2575 2576 2577 2578 2579 2580
	for (i = 1; i < HPAGE_PMD_NR; i++) {
		pc = head_pc + i;
		pc->mem_cgroup = head_pc->mem_cgroup;
		smp_wmb();/* see __commit_charge() */
		pc->flags = head_pc->flags & ~PCGF_NOCOPY_AT_SPLIT;
	}
2581
}
2582
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
2583

2584
/**
2585
 * mem_cgroup_move_account - move account of the page
2586
 * @page: the page
2587
 * @nr_pages: number of regular pages (>1 for huge pages)
2588 2589 2590
 * @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.
2591
 * @uncharge: whether we should call uncharge and css_put against @from.
2592 2593
 *
 * The caller must confirm following.
K
KAMEZAWA Hiroyuki 已提交
2594
 * - page is not on LRU (isolate_page() is useful.)
2595
 * - compound_lock is held when nr_pages > 1
2596
 *
2597
 * This function doesn't do "charge" nor css_get to new cgroup. It should be
L
Lucas De Marchi 已提交
2598
 * done by a caller(__mem_cgroup_try_charge would be useful). If @uncharge is
2599 2600
 * true, this function does "uncharge" from old cgroup, but it doesn't if
 * @uncharge is false, so a caller should do "uncharge".
2601
 */
2602 2603 2604 2605 2606 2607
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)
2608
{
2609 2610
	unsigned long flags;
	int ret;
2611
	bool anon = PageAnon(page);
2612

2613
	VM_BUG_ON(from == to);
2614
	VM_BUG_ON(PageLRU(page));
2615 2616 2617 2618 2619 2620 2621
	/*
	 * 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;
2622
	if (nr_pages > 1 && !PageTransHuge(page))
2623 2624 2625 2626 2627 2628 2629 2630
		goto out;

	lock_page_cgroup(pc);

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

2631
	move_lock_mem_cgroup(from, &flags);
2632

2633
	if (!anon && page_mapped(page)) {
2634 2635 2636 2637 2638
		/* 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();
2639
	}
2640
	mem_cgroup_charge_statistics(from, anon, -nr_pages);
2641 2642
	if (uncharge)
		/* This is not "cancel", but cancel_charge does all we need. */
2643
		__mem_cgroup_cancel_charge(from, nr_pages);
2644

2645
	/* caller should have done css_get */
K
KAMEZAWA Hiroyuki 已提交
2646
	pc->mem_cgroup = to;
2647
	mem_cgroup_charge_statistics(to, anon, nr_pages);
2648 2649 2650
	/*
	 * We charges against "to" which may not have any tasks. Then, "to"
	 * can be under rmdir(). But in current implementation, caller of
2651
	 * this function is just force_empty() and move charge, so it's
L
Lucas De Marchi 已提交
2652
	 * guaranteed that "to" is never removed. So, we don't check rmdir
2653
	 * status here.
2654
	 */
2655
	move_unlock_mem_cgroup(from, &flags);
2656 2657
	ret = 0;
unlock:
2658
	unlock_page_cgroup(pc);
2659 2660 2661
	/*
	 * check events
	 */
2662 2663
	memcg_check_events(to, page);
	memcg_check_events(from, page);
2664
out:
2665 2666 2667 2668 2669 2670 2671
	return ret;
}

/*
 * move charges to its parent.
 */

2672 2673
static int mem_cgroup_move_parent(struct page *page,
				  struct page_cgroup *pc,
2674 2675 2676 2677 2678 2679
				  struct mem_cgroup *child,
				  gfp_t gfp_mask)
{
	struct cgroup *cg = child->css.cgroup;
	struct cgroup *pcg = cg->parent;
	struct mem_cgroup *parent;
2680
	unsigned int nr_pages;
2681
	unsigned long uninitialized_var(flags);
2682 2683 2684 2685 2686 2687
	int ret;

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

2688 2689 2690 2691 2692
	ret = -EBUSY;
	if (!get_page_unless_zero(page))
		goto out;
	if (isolate_lru_page(page))
		goto put;
2693

2694
	nr_pages = hpage_nr_pages(page);
K
KAMEZAWA Hiroyuki 已提交
2695

2696
	parent = mem_cgroup_from_cont(pcg);
2697
	ret = __mem_cgroup_try_charge(NULL, gfp_mask, nr_pages, &parent, false);
2698
	if (ret)
2699
		goto put_back;
2700

2701
	if (nr_pages > 1)
2702 2703
		flags = compound_lock_irqsave(page);

2704
	ret = mem_cgroup_move_account(page, nr_pages, pc, child, parent, true);
2705
	if (ret)
2706
		__mem_cgroup_cancel_charge(parent, nr_pages);
2707

2708
	if (nr_pages > 1)
2709
		compound_unlock_irqrestore(page, flags);
2710
put_back:
K
KAMEZAWA Hiroyuki 已提交
2711
	putback_lru_page(page);
2712
put:
2713
	put_page(page);
2714
out:
2715 2716 2717
	return ret;
}

2718 2719 2720 2721 2722 2723 2724
/*
 * 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,
2725
				gfp_t gfp_mask, enum charge_type ctype)
2726
{
2727
	struct mem_cgroup *memcg = NULL;
2728
	unsigned int nr_pages = 1;
2729
	bool oom = true;
2730
	int ret;
A
Andrea Arcangeli 已提交
2731

A
Andrea Arcangeli 已提交
2732
	if (PageTransHuge(page)) {
2733
		nr_pages <<= compound_order(page);
A
Andrea Arcangeli 已提交
2734
		VM_BUG_ON(!PageTransHuge(page));
2735 2736 2737 2738 2739
		/*
		 * 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
2740
	}
2741

2742
	ret = __mem_cgroup_try_charge(mm, gfp_mask, nr_pages, &memcg, oom);
2743
	if (ret == -ENOMEM)
2744
		return ret;
2745
	__mem_cgroup_commit_charge(memcg, page, nr_pages, ctype, false);
2746 2747 2748
	return 0;
}

2749 2750
int mem_cgroup_newpage_charge(struct page *page,
			      struct mm_struct *mm, gfp_t gfp_mask)
2751
{
2752
	if (mem_cgroup_disabled())
2753
		return 0;
2754 2755 2756
	VM_BUG_ON(page_mapped(page));
	VM_BUG_ON(page->mapping && !PageAnon(page));
	VM_BUG_ON(!mm);
2757
	return mem_cgroup_charge_common(page, mm, gfp_mask,
2758
					MEM_CGROUP_CHARGE_TYPE_MAPPED);
2759 2760
}

D
Daisuke Nishimura 已提交
2761 2762 2763 2764
static void
__mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr,
					enum charge_type ctype);

2765 2766
int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm,
				gfp_t gfp_mask)
2767
{
2768
	struct mem_cgroup *memcg = NULL;
2769
	enum charge_type type = MEM_CGROUP_CHARGE_TYPE_CACHE;
2770 2771
	int ret;

2772
	if (mem_cgroup_disabled())
2773
		return 0;
2774 2775
	if (PageCompound(page))
		return 0;
2776

2777
	if (unlikely(!mm))
2778
		mm = &init_mm;
2779 2780
	if (!page_is_file_cache(page))
		type = MEM_CGROUP_CHARGE_TYPE_SHMEM;
2781

2782
	if (!PageSwapCache(page))
2783
		ret = mem_cgroup_charge_common(page, mm, gfp_mask, type);
2784
	else { /* page is swapcache/shmem */
2785
		ret = mem_cgroup_try_charge_swapin(mm, page, gfp_mask, &memcg);
D
Daisuke Nishimura 已提交
2786
		if (!ret)
2787 2788
			__mem_cgroup_commit_charge_swapin(page, memcg, type);
	}
2789
	return ret;
2790 2791
}

2792 2793 2794
/*
 * While swap-in, try_charge -> commit or cancel, the page is locked.
 * And when try_charge() successfully returns, one refcnt to memcg without
2795
 * struct page_cgroup is acquired. This refcnt will be consumed by
2796 2797
 * "commit()" or removed by "cancel()"
 */
2798 2799
int mem_cgroup_try_charge_swapin(struct mm_struct *mm,
				 struct page *page,
2800
				 gfp_t mask, struct mem_cgroup **memcgp)
2801
{
2802
	struct mem_cgroup *memcg;
2803
	int ret;
2804

2805
	*memcgp = NULL;
2806

2807
	if (mem_cgroup_disabled())
2808 2809 2810 2811 2812 2813
		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
2814 2815 2816
	 * 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.
2817 2818
	 */
	if (!PageSwapCache(page))
H
Hugh Dickins 已提交
2819
		goto charge_cur_mm;
2820 2821
	memcg = try_get_mem_cgroup_from_page(page);
	if (!memcg)
2822
		goto charge_cur_mm;
2823 2824
	*memcgp = memcg;
	ret = __mem_cgroup_try_charge(NULL, mask, 1, memcgp, true);
2825
	css_put(&memcg->css);
2826 2827
	if (ret == -EINTR)
		ret = 0;
2828
	return ret;
2829 2830 2831
charge_cur_mm:
	if (unlikely(!mm))
		mm = &init_mm;
2832 2833 2834 2835
	ret = __mem_cgroup_try_charge(mm, mask, 1, memcgp, true);
	if (ret == -EINTR)
		ret = 0;
	return ret;
2836 2837
}

D
Daisuke Nishimura 已提交
2838
static void
2839
__mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *memcg,
D
Daisuke Nishimura 已提交
2840
					enum charge_type ctype)
2841
{
2842
	if (mem_cgroup_disabled())
2843
		return;
2844
	if (!memcg)
2845
		return;
2846
	cgroup_exclude_rmdir(&memcg->css);
2847

2848
	__mem_cgroup_commit_charge(memcg, page, 1, ctype, true);
2849 2850 2851
	/*
	 * Now swap is on-memory. This means this page may be
	 * counted both as mem and swap....double count.
2852 2853 2854
	 * 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.
2855
	 */
2856
	if (do_swap_account && PageSwapCache(page)) {
2857
		swp_entry_t ent = {.val = page_private(page)};
2858
		struct mem_cgroup *swap_memcg;
2859 2860 2861 2862
		unsigned short id;

		id = swap_cgroup_record(ent, 0);
		rcu_read_lock();
2863 2864
		swap_memcg = mem_cgroup_lookup(id);
		if (swap_memcg) {
2865 2866 2867 2868
			/*
			 * This recorded memcg can be obsolete one. So, avoid
			 * calling css_tryget
			 */
2869 2870 2871 2872 2873
			if (!mem_cgroup_is_root(swap_memcg))
				res_counter_uncharge(&swap_memcg->memsw,
						     PAGE_SIZE);
			mem_cgroup_swap_statistics(swap_memcg, false);
			mem_cgroup_put(swap_memcg);
2874
		}
2875
		rcu_read_unlock();
2876
	}
2877 2878 2879 2880 2881
	/*
	 * 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.
	 */
2882
	cgroup_release_and_wakeup_rmdir(&memcg->css);
2883 2884
}

2885 2886
void mem_cgroup_commit_charge_swapin(struct page *page,
				     struct mem_cgroup *memcg)
D
Daisuke Nishimura 已提交
2887
{
2888 2889
	__mem_cgroup_commit_charge_swapin(page, memcg,
					  MEM_CGROUP_CHARGE_TYPE_MAPPED);
D
Daisuke Nishimura 已提交
2890 2891
}

2892
void mem_cgroup_cancel_charge_swapin(struct mem_cgroup *memcg)
2893
{
2894
	if (mem_cgroup_disabled())
2895
		return;
2896
	if (!memcg)
2897
		return;
2898
	__mem_cgroup_cancel_charge(memcg, 1);
2899 2900
}

2901
static void mem_cgroup_do_uncharge(struct mem_cgroup *memcg,
2902 2903
				   unsigned int nr_pages,
				   const enum charge_type ctype)
2904 2905 2906
{
	struct memcg_batch_info *batch = NULL;
	bool uncharge_memsw = true;
2907

2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918
	/* 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)
2919
		batch->memcg = memcg;
2920 2921
	/*
	 * do_batch > 0 when unmapping pages or inode invalidate/truncate.
L
Lucas De Marchi 已提交
2922
	 * In those cases, all pages freed continuously can be expected to be in
2923 2924 2925 2926 2927 2928 2929 2930
	 * 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;

2931
	if (nr_pages > 1)
A
Andrea Arcangeli 已提交
2932 2933
		goto direct_uncharge;

2934 2935 2936 2937 2938
	/*
	 * 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.
	 */
2939
	if (batch->memcg != memcg)
2940 2941
		goto direct_uncharge;
	/* remember freed charge and uncharge it later */
2942
	batch->nr_pages++;
2943
	if (uncharge_memsw)
2944
		batch->memsw_nr_pages++;
2945 2946
	return;
direct_uncharge:
2947
	res_counter_uncharge(&memcg->res, nr_pages * PAGE_SIZE);
2948
	if (uncharge_memsw)
2949 2950 2951
		res_counter_uncharge(&memcg->memsw, nr_pages * PAGE_SIZE);
	if (unlikely(batch->memcg != memcg))
		memcg_oom_recover(memcg);
2952
}
2953

2954
/*
2955
 * uncharge if !page_mapped(page)
2956
 */
2957
static struct mem_cgroup *
2958
__mem_cgroup_uncharge_common(struct page *page, enum charge_type ctype)
2959
{
2960
	struct mem_cgroup *memcg = NULL;
2961 2962
	unsigned int nr_pages = 1;
	struct page_cgroup *pc;
2963
	bool anon;
2964

2965
	if (mem_cgroup_disabled())
2966
		return NULL;
2967

K
KAMEZAWA Hiroyuki 已提交
2968
	if (PageSwapCache(page))
2969
		return NULL;
K
KAMEZAWA Hiroyuki 已提交
2970

A
Andrea Arcangeli 已提交
2971
	if (PageTransHuge(page)) {
2972
		nr_pages <<= compound_order(page);
A
Andrea Arcangeli 已提交
2973 2974
		VM_BUG_ON(!PageTransHuge(page));
	}
2975
	/*
2976
	 * Check if our page_cgroup is valid
2977
	 */
2978
	pc = lookup_page_cgroup(page);
2979
	if (unlikely(!PageCgroupUsed(pc)))
2980
		return NULL;
2981

2982
	lock_page_cgroup(pc);
K
KAMEZAWA Hiroyuki 已提交
2983

2984
	memcg = pc->mem_cgroup;
2985

K
KAMEZAWA Hiroyuki 已提交
2986 2987 2988
	if (!PageCgroupUsed(pc))
		goto unlock_out;

2989 2990
	anon = PageAnon(page);

K
KAMEZAWA Hiroyuki 已提交
2991 2992
	switch (ctype) {
	case MEM_CGROUP_CHARGE_TYPE_MAPPED:
2993 2994 2995 2996 2997
		/*
		 * Generally PageAnon tells if it's the anon statistics to be
		 * updated; but sometimes e.g. mem_cgroup_uncharge_page() is
		 * used before page reached the stage of being marked PageAnon.
		 */
2998 2999
		anon = true;
		/* fallthrough */
K
KAMEZAWA Hiroyuki 已提交
3000
	case MEM_CGROUP_CHARGE_TYPE_DROP:
3001 3002
		/* See mem_cgroup_prepare_migration() */
		if (page_mapped(page) || PageCgroupMigration(pc))
K
KAMEZAWA Hiroyuki 已提交
3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013
			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;
3014
	}
K
KAMEZAWA Hiroyuki 已提交
3015

3016
	mem_cgroup_charge_statistics(memcg, anon, -nr_pages);
K
KAMEZAWA Hiroyuki 已提交
3017

3018
	ClearPageCgroupUsed(pc);
3019 3020 3021 3022 3023 3024
	/*
	 * 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.
	 */
3025

3026
	unlock_page_cgroup(pc);
K
KAMEZAWA Hiroyuki 已提交
3027
	/*
3028
	 * even after unlock, we have memcg->res.usage here and this memcg
K
KAMEZAWA Hiroyuki 已提交
3029 3030
	 * will never be freed.
	 */
3031
	memcg_check_events(memcg, page);
K
KAMEZAWA Hiroyuki 已提交
3032
	if (do_swap_account && ctype == MEM_CGROUP_CHARGE_TYPE_SWAPOUT) {
3033 3034
		mem_cgroup_swap_statistics(memcg, true);
		mem_cgroup_get(memcg);
K
KAMEZAWA Hiroyuki 已提交
3035
	}
3036 3037
	if (!mem_cgroup_is_root(memcg))
		mem_cgroup_do_uncharge(memcg, nr_pages, ctype);
3038

3039
	return memcg;
K
KAMEZAWA Hiroyuki 已提交
3040 3041 3042

unlock_out:
	unlock_page_cgroup(pc);
3043
	return NULL;
3044 3045
}

3046 3047
void mem_cgroup_uncharge_page(struct page *page)
{
3048 3049 3050
	/* early check. */
	if (page_mapped(page))
		return;
3051
	VM_BUG_ON(page->mapping && !PageAnon(page));
3052 3053 3054 3055 3056 3057
	__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));
3058
	VM_BUG_ON(page->mapping);
3059 3060 3061
	__mem_cgroup_uncharge_common(page, MEM_CGROUP_CHARGE_TYPE_CACHE);
}

3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075
/*
 * 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;
3076 3077
		current->memcg_batch.nr_pages = 0;
		current->memcg_batch.memsw_nr_pages = 0;
3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097
	}
}

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.
	 */
3098 3099 3100 3101 3102 3103
	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);
3104
	memcg_oom_recover(batch->memcg);
3105 3106 3107 3108
	/* forget this pointer (for sanity check) */
	batch->memcg = NULL;
}

3109
#ifdef CONFIG_SWAP
3110
/*
3111
 * called after __delete_from_swap_cache() and drop "page" account.
3112 3113
 * memcg information is recorded to swap_cgroup of "ent"
 */
K
KAMEZAWA Hiroyuki 已提交
3114 3115
void
mem_cgroup_uncharge_swapcache(struct page *page, swp_entry_t ent, bool swapout)
3116 3117
{
	struct mem_cgroup *memcg;
K
KAMEZAWA Hiroyuki 已提交
3118 3119 3120 3121 3122 3123
	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);
3124

K
KAMEZAWA Hiroyuki 已提交
3125 3126 3127 3128 3129
	/*
	 * record memcg information,  if swapout && memcg != NULL,
	 * mem_cgroup_get() was called in uncharge().
	 */
	if (do_swap_account && swapout && memcg)
3130
		swap_cgroup_record(ent, css_id(&memcg->css));
3131
}
3132
#endif
3133 3134 3135 3136 3137 3138 3139

#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
3140
{
3141
	struct mem_cgroup *memcg;
3142
	unsigned short id;
3143 3144 3145 3146

	if (!do_swap_account)
		return;

3147 3148 3149
	id = swap_cgroup_record(ent, 0);
	rcu_read_lock();
	memcg = mem_cgroup_lookup(id);
3150
	if (memcg) {
3151 3152 3153 3154
		/*
		 * We uncharge this because swap is freed.
		 * This memcg can be obsolete one. We avoid calling css_tryget
		 */
3155
		if (!mem_cgroup_is_root(memcg))
3156
			res_counter_uncharge(&memcg->memsw, PAGE_SIZE);
3157
		mem_cgroup_swap_statistics(memcg, false);
3158 3159
		mem_cgroup_put(memcg);
	}
3160
	rcu_read_unlock();
K
KAMEZAWA Hiroyuki 已提交
3161
}
3162 3163 3164 3165 3166 3167

/**
 * 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
3168
 * @need_fixup: whether we should fixup res_counters and refcounts.
3169 3170 3171 3172 3173 3174 3175 3176 3177 3178
 *
 * 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,
3179
		struct mem_cgroup *from, struct mem_cgroup *to, bool need_fixup)
3180 3181 3182 3183 3184 3185 3186 3187
{
	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);
3188
		mem_cgroup_swap_statistics(to, true);
3189
		/*
3190 3191 3192 3193 3194 3195
		 * 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.
3196 3197
		 */
		mem_cgroup_get(to);
3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208
		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);
		}
3209 3210 3211 3212 3213 3214
		return 0;
	}
	return -EINVAL;
}
#else
static inline int mem_cgroup_move_swap_account(swp_entry_t entry,
3215
		struct mem_cgroup *from, struct mem_cgroup *to, bool need_fixup)
3216 3217 3218
{
	return -EINVAL;
}
3219
#endif
K
KAMEZAWA Hiroyuki 已提交
3220

3221
/*
3222 3223
 * Before starting migration, account PAGE_SIZE to mem_cgroup that the old
 * page belongs to.
3224
 */
3225
int mem_cgroup_prepare_migration(struct page *page,
3226
	struct page *newpage, struct mem_cgroup **memcgp, gfp_t gfp_mask)
3227
{
3228
	struct mem_cgroup *memcg = NULL;
3229
	struct page_cgroup *pc;
3230
	enum charge_type ctype;
3231
	int ret = 0;
3232

3233
	*memcgp = NULL;
3234

A
Andrea Arcangeli 已提交
3235
	VM_BUG_ON(PageTransHuge(page));
3236
	if (mem_cgroup_disabled())
3237 3238
		return 0;

3239 3240 3241
	pc = lookup_page_cgroup(page);
	lock_page_cgroup(pc);
	if (PageCgroupUsed(pc)) {
3242 3243
		memcg = pc->mem_cgroup;
		css_get(&memcg->css);
3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274
		/*
		 * 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);
3275
	}
3276
	unlock_page_cgroup(pc);
3277 3278 3279 3280
	/*
	 * If the page is not charged at this point,
	 * we return here.
	 */
3281
	if (!memcg)
3282
		return 0;
3283

3284 3285
	*memcgp = memcg;
	ret = __mem_cgroup_try_charge(NULL, gfp_mask, 1, memcgp, false);
3286
	css_put(&memcg->css);/* drop extra refcnt */
3287
	if (ret) {
3288 3289 3290 3291 3292 3293 3294 3295 3296
		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);
		}
3297
		/* we'll need to revisit this error code (we have -EINTR) */
3298
		return -ENOMEM;
3299
	}
3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311
	/*
	 * We charge new page before it's used/mapped. So, even if unlock_page()
	 * is called before end_migration, we can catch all events on this new
	 * page. In the case new page is migrated but not remapped, new page's
	 * mapcount will be finally 0 and we call uncharge in end_migration().
	 */
	if (PageAnon(page))
		ctype = MEM_CGROUP_CHARGE_TYPE_MAPPED;
	else if (page_is_file_cache(page))
		ctype = MEM_CGROUP_CHARGE_TYPE_CACHE;
	else
		ctype = MEM_CGROUP_CHARGE_TYPE_SHMEM;
3312
	__mem_cgroup_commit_charge(memcg, newpage, 1, ctype, false);
3313
	return ret;
3314
}
3315

3316
/* remove redundant charge if migration failed*/
3317
void mem_cgroup_end_migration(struct mem_cgroup *memcg,
3318
	struct page *oldpage, struct page *newpage, bool migration_ok)
3319
{
3320
	struct page *used, *unused;
3321
	struct page_cgroup *pc;
3322
	bool anon;
3323

3324
	if (!memcg)
3325
		return;
3326
	/* blocks rmdir() */
3327
	cgroup_exclude_rmdir(&memcg->css);
3328
	if (!migration_ok) {
3329 3330
		used = oldpage;
		unused = newpage;
3331
	} else {
3332
		used = newpage;
3333 3334
		unused = oldpage;
	}
3335
	/*
3336 3337 3338
	 * 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.
3339
	 */
3340 3341 3342 3343
	pc = lookup_page_cgroup(oldpage);
	lock_page_cgroup(pc);
	ClearPageCgroupMigration(pc);
	unlock_page_cgroup(pc);
3344 3345 3346 3347
	anon = PageAnon(used);
	__mem_cgroup_uncharge_common(unused,
		anon ? MEM_CGROUP_CHARGE_TYPE_MAPPED
		     : MEM_CGROUP_CHARGE_TYPE_CACHE);
3348

3349
	/*
3350 3351 3352 3353 3354 3355
	 * 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)
3356
	 */
3357
	if (anon)
3358
		mem_cgroup_uncharge_page(used);
3359
	/*
3360 3361
	 * At migration, we may charge account against cgroup which has no
	 * tasks.
3362 3363 3364
	 * 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.
	 */
3365
	cgroup_release_and_wakeup_rmdir(&memcg->css);
3366
}
3367

3368 3369 3370 3371 3372 3373 3374 3375
/*
 * 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)
{
3376
	struct mem_cgroup *memcg = NULL;
3377 3378 3379 3380 3381 3382 3383 3384 3385
	struct page_cgroup *pc;
	enum charge_type type = MEM_CGROUP_CHARGE_TYPE_CACHE;

	if (mem_cgroup_disabled())
		return;

	pc = lookup_page_cgroup(oldpage);
	/* fix accounting on old pages */
	lock_page_cgroup(pc);
3386 3387 3388 3389 3390
	if (PageCgroupUsed(pc)) {
		memcg = pc->mem_cgroup;
		mem_cgroup_charge_statistics(memcg, false, -1);
		ClearPageCgroupUsed(pc);
	}
3391 3392
	unlock_page_cgroup(pc);

3393 3394 3395 3396 3397 3398 3399
	/*
	 * When called from shmem_replace_page(), in some cases the
	 * oldpage has already been charged, and in some cases not.
	 */
	if (!memcg)
		return;

3400 3401 3402 3403 3404 3405 3406 3407
	if (PageSwapBacked(oldpage))
		type = MEM_CGROUP_CHARGE_TYPE_SHMEM;

	/*
	 * Even if newpage->mapping was NULL before starting replacement,
	 * the newpage may be on LRU(or pagevec for LRU) already. We lock
	 * LRU while we overwrite pc->mem_cgroup.
	 */
3408
	__mem_cgroup_commit_charge(memcg, newpage, 1, type, true);
3409 3410
}

3411 3412 3413 3414 3415 3416
#ifdef CONFIG_DEBUG_VM
static struct page_cgroup *lookup_page_cgroup_used(struct page *page)
{
	struct page_cgroup *pc;

	pc = lookup_page_cgroup(page);
3417 3418 3419 3420 3421
	/*
	 * Can be NULL while feeding pages into the page allocator for
	 * the first time, i.e. during boot or memory hotplug;
	 * or when mem_cgroup_disabled().
	 */
3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440
	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) {
3441
		printk(KERN_ALERT "pc:%p pc->flags:%lx pc->mem_cgroup:%p\n",
3442 3443 3444 3445 3446
		       pc, pc->flags, pc->mem_cgroup);
	}
}
#endif

3447 3448
static DEFINE_MUTEX(set_limit_mutex);

3449
static int mem_cgroup_resize_limit(struct mem_cgroup *memcg,
3450
				unsigned long long val)
3451
{
3452
	int retry_count;
3453
	u64 memswlimit, memlimit;
3454
	int ret = 0;
3455 3456
	int children = mem_cgroup_count_children(memcg);
	u64 curusage, oldusage;
3457
	int enlarge;
3458 3459 3460 3461 3462 3463 3464 3465 3466

	/*
	 * 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);
3467

3468
	enlarge = 0;
3469
	while (retry_count) {
3470 3471 3472 3473
		if (signal_pending(current)) {
			ret = -EINTR;
			break;
		}
3474 3475 3476
		/*
		 * Rather than hide all in some function, I do this in
		 * open coded manner. You see what this really does.
3477
		 * We have to guarantee memcg->res.limit < memcg->memsw.limit.
3478 3479 3480 3481 3482 3483
		 */
		mutex_lock(&set_limit_mutex);
		memswlimit = res_counter_read_u64(&memcg->memsw, RES_LIMIT);
		if (memswlimit < val) {
			ret = -EINVAL;
			mutex_unlock(&set_limit_mutex);
3484 3485
			break;
		}
3486 3487 3488 3489 3490

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

3491
		ret = res_counter_set_limit(&memcg->res, val);
3492 3493 3494 3495 3496 3497
		if (!ret) {
			if (memswlimit == val)
				memcg->memsw_is_minimum = true;
			else
				memcg->memsw_is_minimum = false;
		}
3498 3499 3500 3501 3502
		mutex_unlock(&set_limit_mutex);

		if (!ret)
			break;

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

3515 3516 3517
	return ret;
}

L
Li Zefan 已提交
3518 3519
static int mem_cgroup_resize_memsw_limit(struct mem_cgroup *memcg,
					unsigned long long val)
3520
{
3521
	int retry_count;
3522
	u64 memlimit, memswlimit, oldusage, curusage;
3523 3524
	int children = mem_cgroup_count_children(memcg);
	int ret = -EBUSY;
3525
	int enlarge = 0;
3526

3527 3528 3529
	/* see mem_cgroup_resize_res_limit */
 	retry_count = children * MEM_CGROUP_RECLAIM_RETRIES;
	oldusage = res_counter_read_u64(&memcg->memsw, RES_USAGE);
3530 3531 3532 3533 3534 3535 3536 3537
	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.
3538
		 * We have to guarantee memcg->res.limit < memcg->memsw.limit.
3539 3540 3541 3542 3543 3544 3545 3546
		 */
		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;
		}
3547 3548 3549
		memswlimit = res_counter_read_u64(&memcg->memsw, RES_LIMIT);
		if (memswlimit < val)
			enlarge = 1;
3550
		ret = res_counter_set_limit(&memcg->memsw, val);
3551 3552 3553 3554 3555 3556
		if (!ret) {
			if (memlimit == val)
				memcg->memsw_is_minimum = true;
			else
				memcg->memsw_is_minimum = false;
		}
3557 3558 3559 3560 3561
		mutex_unlock(&set_limit_mutex);

		if (!ret)
			break;

3562 3563 3564
		mem_cgroup_reclaim(memcg, GFP_KERNEL,
				   MEM_CGROUP_RECLAIM_NOSWAP |
				   MEM_CGROUP_RECLAIM_SHRINK);
3565
		curusage = res_counter_read_u64(&memcg->memsw, RES_USAGE);
3566
		/* Usage is reduced ? */
3567
		if (curusage >= oldusage)
3568
			retry_count--;
3569 3570
		else
			oldusage = curusage;
3571
	}
3572 3573
	if (!ret && enlarge)
		memcg_oom_recover(memcg);
3574 3575 3576
	return ret;
}

3577
unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
3578 3579
					    gfp_t gfp_mask,
					    unsigned long *total_scanned)
3580 3581 3582 3583 3584 3585
{
	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;
3586
	unsigned long long excess;
3587
	unsigned long nr_scanned;
3588 3589 3590 3591

	if (order > 0)
		return 0;

3592
	mctz = soft_limit_tree_node_zone(zone_to_nid(zone), zone_idx(zone));
3593 3594 3595 3596 3597 3598 3599 3600 3601 3602 3603 3604 3605
	/*
	 * 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;

3606
		nr_scanned = 0;
3607
		reclaimed = mem_cgroup_soft_reclaim(mz->memcg, zone,
3608
						    gfp_mask, &nr_scanned);
3609
		nr_reclaimed += reclaimed;
3610
		*total_scanned += nr_scanned;
3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628 3629 3630 3631 3632
		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);
3633
				if (next_mz == mz)
3634
					css_put(&next_mz->memcg->css);
3635
				else /* next_mz == NULL or other memcg */
3636 3637 3638
					break;
			} while (1);
		}
3639 3640
		__mem_cgroup_remove_exceeded(mz->memcg, mz, mctz);
		excess = res_counter_soft_limit_excess(&mz->memcg->res);
3641 3642 3643 3644 3645 3646 3647 3648
		/*
		 * 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.
		 */
3649
		/* If excess == 0, no tree ops */
3650
		__mem_cgroup_insert_exceeded(mz->memcg, mz, mctz, excess);
3651
		spin_unlock(&mctz->lock);
3652
		css_put(&mz->memcg->css);
3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664
		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)
3665
		css_put(&next_mz->memcg->css);
3666 3667 3668
	return nr_reclaimed;
}

3669 3670 3671 3672
/*
 * This routine traverse page_cgroup in given list and drop them all.
 * *And* this routine doesn't reclaim page itself, just removes page_cgroup.
 */
3673
static int mem_cgroup_force_empty_list(struct mem_cgroup *memcg,
K
KAMEZAWA Hiroyuki 已提交
3674
				int node, int zid, enum lru_list lru)
3675
{
K
KAMEZAWA Hiroyuki 已提交
3676 3677
	struct mem_cgroup_per_zone *mz;
	unsigned long flags, loop;
3678
	struct list_head *list;
3679 3680
	struct page *busy;
	struct zone *zone;
3681
	int ret = 0;
3682

K
KAMEZAWA Hiroyuki 已提交
3683
	zone = &NODE_DATA(node)->node_zones[zid];
3684
	mz = mem_cgroup_zoneinfo(memcg, node, zid);
3685
	list = &mz->lruvec.lists[lru];
3686

3687
	loop = mz->lru_size[lru];
3688 3689 3690 3691
	/* give some margin against EBUSY etc...*/
	loop += 256;
	busy = NULL;
	while (loop--) {
3692
		struct page_cgroup *pc;
3693 3694
		struct page *page;

3695
		ret = 0;
K
KAMEZAWA Hiroyuki 已提交
3696
		spin_lock_irqsave(&zone->lru_lock, flags);
3697
		if (list_empty(list)) {
K
KAMEZAWA Hiroyuki 已提交
3698
			spin_unlock_irqrestore(&zone->lru_lock, flags);
3699
			break;
3700
		}
3701 3702 3703
		page = list_entry(list->prev, struct page, lru);
		if (busy == page) {
			list_move(&page->lru, list);
3704
			busy = NULL;
K
KAMEZAWA Hiroyuki 已提交
3705
			spin_unlock_irqrestore(&zone->lru_lock, flags);
3706 3707
			continue;
		}
K
KAMEZAWA Hiroyuki 已提交
3708
		spin_unlock_irqrestore(&zone->lru_lock, flags);
3709

3710
		pc = lookup_page_cgroup(page);
3711

3712
		ret = mem_cgroup_move_parent(page, pc, memcg, GFP_KERNEL);
3713
		if (ret == -ENOMEM || ret == -EINTR)
3714
			break;
3715 3716 3717

		if (ret == -EBUSY || ret == -EINVAL) {
			/* found lock contention or "pc" is obsolete. */
3718
			busy = page;
3719 3720 3721
			cond_resched();
		} else
			busy = NULL;
3722
	}
K
KAMEZAWA Hiroyuki 已提交
3723

3724 3725 3726
	if (!ret && !list_empty(list))
		return -EBUSY;
	return ret;
3727 3728 3729 3730 3731 3732
}

/*
 * make mem_cgroup's charge to be 0 if there is no task.
 * This enables deleting this mem_cgroup.
 */
3733
static int mem_cgroup_force_empty(struct mem_cgroup *memcg, bool free_all)
3734
{
3735 3736 3737
	int ret;
	int node, zid, shrink;
	int nr_retries = MEM_CGROUP_RECLAIM_RETRIES;
3738
	struct cgroup *cgrp = memcg->css.cgroup;
3739

3740
	css_get(&memcg->css);
3741 3742

	shrink = 0;
3743 3744 3745
	/* should free all ? */
	if (free_all)
		goto try_to_free;
3746
move_account:
3747
	do {
3748
		ret = -EBUSY;
3749 3750 3751 3752
		if (cgroup_task_count(cgrp) || !list_empty(&cgrp->children))
			goto out;
		ret = -EINTR;
		if (signal_pending(current))
3753
			goto out;
3754 3755
		/* This is for making all *used* pages to be on LRU. */
		lru_add_drain_all();
3756
		drain_all_stock_sync(memcg);
3757
		ret = 0;
3758
		mem_cgroup_start_move(memcg);
3759
		for_each_node_state(node, N_HIGH_MEMORY) {
3760
			for (zid = 0; !ret && zid < MAX_NR_ZONES; zid++) {
H
Hugh Dickins 已提交
3761 3762
				enum lru_list lru;
				for_each_lru(lru) {
3763
					ret = mem_cgroup_force_empty_list(memcg,
H
Hugh Dickins 已提交
3764
							node, zid, lru);
3765 3766 3767
					if (ret)
						break;
				}
3768
			}
3769 3770 3771
			if (ret)
				break;
		}
3772 3773
		mem_cgroup_end_move(memcg);
		memcg_oom_recover(memcg);
3774 3775 3776
		/* it seems parent cgroup doesn't have enough mem */
		if (ret == -ENOMEM)
			goto try_to_free;
3777
		cond_resched();
3778
	/* "ret" should also be checked to ensure all lists are empty. */
3779
	} while (res_counter_read_u64(&memcg->res, RES_USAGE) > 0 || ret);
3780
out:
3781
	css_put(&memcg->css);
3782
	return ret;
3783 3784

try_to_free:
3785 3786
	/* returns EBUSY if there is a task or if we come here twice. */
	if (cgroup_task_count(cgrp) || !list_empty(&cgrp->children) || shrink) {
3787 3788 3789
		ret = -EBUSY;
		goto out;
	}
3790 3791
	/* we call try-to-free pages for make this cgroup empty */
	lru_add_drain_all();
3792 3793
	/* try to free all pages in this cgroup */
	shrink = 1;
3794
	while (nr_retries && res_counter_read_u64(&memcg->res, RES_USAGE) > 0) {
3795
		int progress;
3796 3797 3798 3799 3800

		if (signal_pending(current)) {
			ret = -EINTR;
			goto out;
		}
3801
		progress = try_to_free_mem_cgroup_pages(memcg, GFP_KERNEL,
3802
						false);
3803
		if (!progress) {
3804
			nr_retries--;
3805
			/* maybe some writeback is necessary */
3806
			congestion_wait(BLK_RW_ASYNC, HZ/10);
3807
		}
3808 3809

	}
K
KAMEZAWA Hiroyuki 已提交
3810
	lru_add_drain();
3811
	/* try move_account...there may be some *locked* pages. */
3812
	goto move_account;
3813 3814
}

3815 3816 3817 3818 3819 3820
int mem_cgroup_force_empty_write(struct cgroup *cont, unsigned int event)
{
	return mem_cgroup_force_empty(mem_cgroup_from_cont(cont), true);
}


3821 3822 3823 3824 3825 3826 3827 3828 3829
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;
3830
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
3831
	struct cgroup *parent = cont->parent;
3832
	struct mem_cgroup *parent_memcg = NULL;
3833 3834

	if (parent)
3835
		parent_memcg = mem_cgroup_from_cont(parent);
3836 3837 3838

	cgroup_lock();
	/*
3839
	 * If parent's use_hierarchy is set, we can't make any modifications
3840 3841 3842 3843 3844 3845
	 * 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.
	 */
3846
	if ((!parent_memcg || !parent_memcg->use_hierarchy) &&
3847 3848
				(val == 1 || val == 0)) {
		if (list_empty(&cont->children))
3849
			memcg->use_hierarchy = val;
3850 3851 3852 3853 3854 3855 3856 3857 3858
		else
			retval = -EBUSY;
	} else
		retval = -EINVAL;
	cgroup_unlock();

	return retval;
}

3859

3860
static unsigned long mem_cgroup_recursive_stat(struct mem_cgroup *memcg,
3861
					       enum mem_cgroup_stat_index idx)
3862
{
K
KAMEZAWA Hiroyuki 已提交
3863
	struct mem_cgroup *iter;
3864
	long val = 0;
3865

3866
	/* Per-cpu values can be negative, use a signed accumulator */
3867
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
3868 3869 3870 3871 3872
		val += mem_cgroup_read_stat(iter, idx);

	if (val < 0) /* race ? */
		val = 0;
	return val;
3873 3874
}

3875
static inline u64 mem_cgroup_usage(struct mem_cgroup *memcg, bool swap)
3876
{
K
KAMEZAWA Hiroyuki 已提交
3877
	u64 val;
3878

3879
	if (!mem_cgroup_is_root(memcg)) {
3880
		if (!swap)
3881
			return res_counter_read_u64(&memcg->res, RES_USAGE);
3882
		else
3883
			return res_counter_read_u64(&memcg->memsw, RES_USAGE);
3884 3885
	}

3886 3887
	val = mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_CACHE);
	val += mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_RSS);
3888

K
KAMEZAWA Hiroyuki 已提交
3889
	if (swap)
3890
		val += mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_SWAPOUT);
3891 3892 3893 3894

	return val << PAGE_SHIFT;
}

3895 3896 3897
static ssize_t mem_cgroup_read(struct cgroup *cont, struct cftype *cft,
			       struct file *file, char __user *buf,
			       size_t nbytes, loff_t *ppos)
B
Balbir Singh 已提交
3898
{
3899
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
3900
	char str[64];
3901
	u64 val;
3902
	int type, name, len;
3903 3904 3905

	type = MEMFILE_TYPE(cft->private);
	name = MEMFILE_ATTR(cft->private);
3906 3907 3908 3909

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

3910 3911
	switch (type) {
	case _MEM:
3912
		if (name == RES_USAGE)
3913
			val = mem_cgroup_usage(memcg, false);
3914
		else
3915
			val = res_counter_read_u64(&memcg->res, name);
3916 3917
		break;
	case _MEMSWAP:
3918
		if (name == RES_USAGE)
3919
			val = mem_cgroup_usage(memcg, true);
3920
		else
3921
			val = res_counter_read_u64(&memcg->memsw, name);
3922 3923 3924 3925
		break;
	default:
		BUG();
	}
3926 3927 3928

	len = scnprintf(str, sizeof(str), "%llu\n", (unsigned long long)val);
	return simple_read_from_buffer(buf, nbytes, ppos, str, len);
B
Balbir Singh 已提交
3929
}
3930 3931 3932 3933
/*
 * The user of this function is...
 * RES_LIMIT.
 */
3934 3935
static int mem_cgroup_write(struct cgroup *cont, struct cftype *cft,
			    const char *buffer)
B
Balbir Singh 已提交
3936
{
3937
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
3938
	int type, name;
3939 3940 3941
	unsigned long long val;
	int ret;

3942 3943
	type = MEMFILE_TYPE(cft->private);
	name = MEMFILE_ATTR(cft->private);
3944 3945 3946 3947

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

3948
	switch (name) {
3949
	case RES_LIMIT:
3950 3951 3952 3953
		if (mem_cgroup_is_root(memcg)) { /* Can't set limit on root */
			ret = -EINVAL;
			break;
		}
3954 3955
		/* This function does all necessary parse...reuse it */
		ret = res_counter_memparse_write_strategy(buffer, &val);
3956 3957 3958
		if (ret)
			break;
		if (type == _MEM)
3959
			ret = mem_cgroup_resize_limit(memcg, val);
3960 3961
		else
			ret = mem_cgroup_resize_memsw_limit(memcg, val);
3962
		break;
3963 3964 3965 3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976
	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;
3977 3978 3979 3980 3981
	default:
		ret = -EINVAL; /* should be BUG() ? */
		break;
	}
	return ret;
B
Balbir Singh 已提交
3982 3983
}

3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006 4007 4008 4009 4010
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;
}

4011
static int mem_cgroup_reset(struct cgroup *cont, unsigned int event)
4012
{
4013
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
4014
	int type, name;
4015

4016 4017
	type = MEMFILE_TYPE(event);
	name = MEMFILE_ATTR(event);
4018 4019 4020 4021

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

4022
	switch (name) {
4023
	case RES_MAX_USAGE:
4024
		if (type == _MEM)
4025
			res_counter_reset_max(&memcg->res);
4026
		else
4027
			res_counter_reset_max(&memcg->memsw);
4028 4029
		break;
	case RES_FAILCNT:
4030
		if (type == _MEM)
4031
			res_counter_reset_failcnt(&memcg->res);
4032
		else
4033
			res_counter_reset_failcnt(&memcg->memsw);
4034 4035
		break;
	}
4036

4037
	return 0;
4038 4039
}

4040 4041 4042 4043 4044 4045
static u64 mem_cgroup_move_charge_read(struct cgroup *cgrp,
					struct cftype *cft)
{
	return mem_cgroup_from_cont(cgrp)->move_charge_at_immigrate;
}

4046
#ifdef CONFIG_MMU
4047 4048 4049
static int mem_cgroup_move_charge_write(struct cgroup *cgrp,
					struct cftype *cft, u64 val)
{
4050
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4051 4052 4053 4054 4055 4056 4057 4058 4059

	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();
4060
	memcg->move_charge_at_immigrate = val;
4061 4062 4063 4064
	cgroup_unlock();

	return 0;
}
4065 4066 4067 4068 4069 4070 4071
#else
static int mem_cgroup_move_charge_write(struct cgroup *cgrp,
					struct cftype *cft, u64 val)
{
	return -ENOSYS;
}
#endif
4072

K
KAMEZAWA Hiroyuki 已提交
4073 4074 4075 4076 4077

/* For read statistics */
enum {
	MCS_CACHE,
	MCS_RSS,
4078
	MCS_FILE_MAPPED,
K
KAMEZAWA Hiroyuki 已提交
4079 4080
	MCS_PGPGIN,
	MCS_PGPGOUT,
4081
	MCS_SWAP,
4082 4083
	MCS_PGFAULT,
	MCS_PGMAJFAULT,
K
KAMEZAWA Hiroyuki 已提交
4084 4085 4086 4087 4088 4089 4090 4091 4092 4093
	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];
4094 4095
};

K
KAMEZAWA Hiroyuki 已提交
4096 4097 4098 4099 4100 4101
struct {
	char *local_name;
	char *total_name;
} memcg_stat_strings[NR_MCS_STAT] = {
	{"cache", "total_cache"},
	{"rss", "total_rss"},
4102
	{"mapped_file", "total_mapped_file"},
K
KAMEZAWA Hiroyuki 已提交
4103 4104
	{"pgpgin", "total_pgpgin"},
	{"pgpgout", "total_pgpgout"},
4105
	{"swap", "total_swap"},
4106 4107
	{"pgfault", "total_pgfault"},
	{"pgmajfault", "total_pgmajfault"},
K
KAMEZAWA Hiroyuki 已提交
4108 4109 4110 4111 4112 4113 4114 4115
	{"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
4116
static void
4117
mem_cgroup_get_local_stat(struct mem_cgroup *memcg, struct mcs_total_stat *s)
K
KAMEZAWA Hiroyuki 已提交
4118 4119 4120 4121
{
	s64 val;

	/* per cpu stat */
4122
	val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_CACHE);
K
KAMEZAWA Hiroyuki 已提交
4123
	s->stat[MCS_CACHE] += val * PAGE_SIZE;
4124
	val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_RSS);
K
KAMEZAWA Hiroyuki 已提交
4125
	s->stat[MCS_RSS] += val * PAGE_SIZE;
4126
	val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_FILE_MAPPED);
4127
	s->stat[MCS_FILE_MAPPED] += val * PAGE_SIZE;
4128
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGPGIN);
K
KAMEZAWA Hiroyuki 已提交
4129
	s->stat[MCS_PGPGIN] += val;
4130
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGPGOUT);
K
KAMEZAWA Hiroyuki 已提交
4131
	s->stat[MCS_PGPGOUT] += val;
4132
	if (do_swap_account) {
4133
		val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_SWAPOUT);
4134 4135
		s->stat[MCS_SWAP] += val * PAGE_SIZE;
	}
4136
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGFAULT);
4137
	s->stat[MCS_PGFAULT] += val;
4138
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGMAJFAULT);
4139
	s->stat[MCS_PGMAJFAULT] += val;
K
KAMEZAWA Hiroyuki 已提交
4140 4141

	/* per zone stat */
4142
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_INACTIVE_ANON));
K
KAMEZAWA Hiroyuki 已提交
4143
	s->stat[MCS_INACTIVE_ANON] += val * PAGE_SIZE;
4144
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_ACTIVE_ANON));
K
KAMEZAWA Hiroyuki 已提交
4145
	s->stat[MCS_ACTIVE_ANON] += val * PAGE_SIZE;
4146
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_INACTIVE_FILE));
K
KAMEZAWA Hiroyuki 已提交
4147
	s->stat[MCS_INACTIVE_FILE] += val * PAGE_SIZE;
4148
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_ACTIVE_FILE));
K
KAMEZAWA Hiroyuki 已提交
4149
	s->stat[MCS_ACTIVE_FILE] += val * PAGE_SIZE;
4150
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_UNEVICTABLE));
K
KAMEZAWA Hiroyuki 已提交
4151 4152 4153 4154
	s->stat[MCS_UNEVICTABLE] += val * PAGE_SIZE;
}

static void
4155
mem_cgroup_get_total_stat(struct mem_cgroup *memcg, struct mcs_total_stat *s)
K
KAMEZAWA Hiroyuki 已提交
4156
{
K
KAMEZAWA Hiroyuki 已提交
4157 4158
	struct mem_cgroup *iter;

4159
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
4160
		mem_cgroup_get_local_stat(iter, s);
K
KAMEZAWA Hiroyuki 已提交
4161 4162
}

4163 4164 4165 4166 4167 4168 4169
#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;
4170
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
4171

4172
	total_nr = mem_cgroup_nr_lru_pages(memcg, LRU_ALL);
4173 4174
	seq_printf(m, "total=%lu", total_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4175
		node_nr = mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL);
4176 4177 4178 4179
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');

4180
	file_nr = mem_cgroup_nr_lru_pages(memcg, LRU_ALL_FILE);
4181 4182
	seq_printf(m, "file=%lu", file_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4183
		node_nr = mem_cgroup_node_nr_lru_pages(memcg, nid,
4184
				LRU_ALL_FILE);
4185 4186 4187 4188
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');

4189
	anon_nr = mem_cgroup_nr_lru_pages(memcg, LRU_ALL_ANON);
4190 4191
	seq_printf(m, "anon=%lu", anon_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4192
		node_nr = mem_cgroup_node_nr_lru_pages(memcg, nid,
4193
				LRU_ALL_ANON);
4194 4195 4196 4197
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');

4198
	unevictable_nr = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_UNEVICTABLE));
4199 4200
	seq_printf(m, "unevictable=%lu", unevictable_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4201
		node_nr = mem_cgroup_node_nr_lru_pages(memcg, nid,
4202
				BIT(LRU_UNEVICTABLE));
4203 4204 4205 4206 4207 4208 4209
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');
	return 0;
}
#endif /* CONFIG_NUMA */

4210 4211
static int mem_control_stat_show(struct cgroup *cont, struct cftype *cft,
				 struct cgroup_map_cb *cb)
4212
{
4213
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
K
KAMEZAWA Hiroyuki 已提交
4214
	struct mcs_total_stat mystat;
4215 4216
	int i;

K
KAMEZAWA Hiroyuki 已提交
4217
	memset(&mystat, 0, sizeof(mystat));
4218
	mem_cgroup_get_local_stat(memcg, &mystat);
4219

4220

4221 4222 4223
	for (i = 0; i < NR_MCS_STAT; i++) {
		if (i == MCS_SWAP && !do_swap_account)
			continue;
K
KAMEZAWA Hiroyuki 已提交
4224
		cb->fill(cb, memcg_stat_strings[i].local_name, mystat.stat[i]);
4225
	}
L
Lee Schermerhorn 已提交
4226

K
KAMEZAWA Hiroyuki 已提交
4227
	/* Hierarchical information */
4228 4229
	{
		unsigned long long limit, memsw_limit;
4230
		memcg_get_hierarchical_limit(memcg, &limit, &memsw_limit);
4231 4232 4233 4234
		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
4235

K
KAMEZAWA Hiroyuki 已提交
4236
	memset(&mystat, 0, sizeof(mystat));
4237
	mem_cgroup_get_total_stat(memcg, &mystat);
4238 4239 4240
	for (i = 0; i < NR_MCS_STAT; i++) {
		if (i == MCS_SWAP && !do_swap_account)
			continue;
K
KAMEZAWA Hiroyuki 已提交
4241
		cb->fill(cb, memcg_stat_strings[i].total_name, mystat.stat[i]);
4242
	}
K
KAMEZAWA Hiroyuki 已提交
4243

K
KOSAKI Motohiro 已提交
4244 4245 4246 4247 4248 4249 4250 4251 4252
#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++) {
4253
				mz = mem_cgroup_zoneinfo(memcg, nid, zid);
K
KOSAKI Motohiro 已提交
4254 4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270

				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

4271 4272 4273
	return 0;
}

K
KOSAKI Motohiro 已提交
4274 4275 4276 4277
static u64 mem_cgroup_swappiness_read(struct cgroup *cgrp, struct cftype *cft)
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);

4278
	return mem_cgroup_swappiness(memcg);
K
KOSAKI Motohiro 已提交
4279 4280 4281 4282 4283 4284 4285
}

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;
4286

K
KOSAKI Motohiro 已提交
4287 4288 4289 4290 4291 4292 4293
	if (val > 100)
		return -EINVAL;

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

	parent = mem_cgroup_from_cont(cgrp->parent);
4294 4295 4296

	cgroup_lock();

K
KOSAKI Motohiro 已提交
4297 4298
	/* If under hierarchy, only empty-root can set this value */
	if ((parent->use_hierarchy) ||
4299 4300
	    (memcg->use_hierarchy && !list_empty(&cgrp->children))) {
		cgroup_unlock();
K
KOSAKI Motohiro 已提交
4301
		return -EINVAL;
4302
	}
K
KOSAKI Motohiro 已提交
4303 4304 4305

	memcg->swappiness = val;

4306 4307
	cgroup_unlock();

K
KOSAKI Motohiro 已提交
4308 4309 4310
	return 0;
}

4311 4312 4313 4314 4315 4316 4317 4318
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)
4319
		t = rcu_dereference(memcg->thresholds.primary);
4320
	else
4321
		t = rcu_dereference(memcg->memsw_thresholds.primary);
4322 4323 4324 4325 4326 4327 4328 4329 4330 4331 4332

	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().
	 */
4333
	i = t->current_threshold;
4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348 4349 4350 4351 4352 4353 4354 4355 4356

	/*
	 * 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 */
4357
	t->current_threshold = i - 1;
4358 4359 4360 4361 4362 4363
unlock:
	rcu_read_unlock();
}

static void mem_cgroup_threshold(struct mem_cgroup *memcg)
{
4364 4365 4366 4367 4368 4369 4370
	while (memcg) {
		__mem_cgroup_threshold(memcg, false);
		if (do_swap_account)
			__mem_cgroup_threshold(memcg, true);

		memcg = parent_mem_cgroup(memcg);
	}
4371 4372 4373 4374 4375 4376 4377 4378 4379 4380
}

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;
}

4381
static int mem_cgroup_oom_notify_cb(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
4382 4383 4384
{
	struct mem_cgroup_eventfd_list *ev;

4385
	list_for_each_entry(ev, &memcg->oom_notify, list)
K
KAMEZAWA Hiroyuki 已提交
4386 4387 4388 4389
		eventfd_signal(ev->eventfd, 1);
	return 0;
}

4390
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
4391
{
K
KAMEZAWA Hiroyuki 已提交
4392 4393
	struct mem_cgroup *iter;

4394
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
4395
		mem_cgroup_oom_notify_cb(iter);
K
KAMEZAWA Hiroyuki 已提交
4396 4397 4398 4399
}

static int mem_cgroup_usage_register_event(struct cgroup *cgrp,
	struct cftype *cft, struct eventfd_ctx *eventfd, const char *args)
4400 4401
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4402 4403
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
4404 4405
	int type = MEMFILE_TYPE(cft->private);
	u64 threshold, usage;
4406
	int i, size, ret;
4407 4408 4409 4410 4411 4412

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

	mutex_lock(&memcg->thresholds_lock);
4413

4414
	if (type == _MEM)
4415
		thresholds = &memcg->thresholds;
4416
	else if (type == _MEMSWAP)
4417
		thresholds = &memcg->memsw_thresholds;
4418 4419 4420 4421 4422 4423
	else
		BUG();

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

	/* Check if a threshold crossed before adding a new one */
4424
	if (thresholds->primary)
4425 4426
		__mem_cgroup_threshold(memcg, type == _MEMSWAP);

4427
	size = thresholds->primary ? thresholds->primary->size + 1 : 1;
4428 4429

	/* Allocate memory for new array of thresholds */
4430
	new = kmalloc(sizeof(*new) + size * sizeof(struct mem_cgroup_threshold),
4431
			GFP_KERNEL);
4432
	if (!new) {
4433 4434 4435
		ret = -ENOMEM;
		goto unlock;
	}
4436
	new->size = size;
4437 4438

	/* Copy thresholds (if any) to new array */
4439 4440
	if (thresholds->primary) {
		memcpy(new->entries, thresholds->primary->entries, (size - 1) *
4441
				sizeof(struct mem_cgroup_threshold));
4442 4443
	}

4444
	/* Add new threshold */
4445 4446
	new->entries[size - 1].eventfd = eventfd;
	new->entries[size - 1].threshold = threshold;
4447 4448

	/* Sort thresholds. Registering of new threshold isn't time-critical */
4449
	sort(new->entries, size, sizeof(struct mem_cgroup_threshold),
4450 4451 4452
			compare_thresholds, NULL);

	/* Find current threshold */
4453
	new->current_threshold = -1;
4454
	for (i = 0; i < size; i++) {
4455
		if (new->entries[i].threshold < usage) {
4456
			/*
4457 4458
			 * new->current_threshold will not be used until
			 * rcu_assign_pointer(), so it's safe to increment
4459 4460
			 * it here.
			 */
4461
			++new->current_threshold;
4462 4463 4464
		}
	}

4465 4466 4467 4468 4469
	/* Free old spare buffer and save old primary buffer as spare */
	kfree(thresholds->spare);
	thresholds->spare = thresholds->primary;

	rcu_assign_pointer(thresholds->primary, new);
4470

4471
	/* To be sure that nobody uses thresholds */
4472 4473 4474 4475 4476 4477 4478 4479
	synchronize_rcu();

unlock:
	mutex_unlock(&memcg->thresholds_lock);

	return ret;
}

4480
static void mem_cgroup_usage_unregister_event(struct cgroup *cgrp,
K
KAMEZAWA Hiroyuki 已提交
4481
	struct cftype *cft, struct eventfd_ctx *eventfd)
4482 4483
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4484 4485
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
4486 4487
	int type = MEMFILE_TYPE(cft->private);
	u64 usage;
4488
	int i, j, size;
4489 4490 4491

	mutex_lock(&memcg->thresholds_lock);
	if (type == _MEM)
4492
		thresholds = &memcg->thresholds;
4493
	else if (type == _MEMSWAP)
4494
		thresholds = &memcg->memsw_thresholds;
4495 4496 4497
	else
		BUG();

4498 4499 4500
	if (!thresholds->primary)
		goto unlock;

4501 4502 4503 4504 4505 4506
	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 */
4507 4508 4509
	size = 0;
	for (i = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd != eventfd)
4510 4511 4512
			size++;
	}

4513
	new = thresholds->spare;
4514

4515 4516
	/* Set thresholds array to NULL if we don't have thresholds */
	if (!size) {
4517 4518
		kfree(new);
		new = NULL;
4519
		goto swap_buffers;
4520 4521
	}

4522
	new->size = size;
4523 4524

	/* Copy thresholds and find current threshold */
4525 4526 4527
	new->current_threshold = -1;
	for (i = 0, j = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd == eventfd)
4528 4529
			continue;

4530 4531
		new->entries[j] = thresholds->primary->entries[i];
		if (new->entries[j].threshold < usage) {
4532
			/*
4533
			 * new->current_threshold will not be used
4534 4535 4536
			 * until rcu_assign_pointer(), so it's safe to increment
			 * it here.
			 */
4537
			++new->current_threshold;
4538 4539 4540 4541
		}
		j++;
	}

4542
swap_buffers:
4543 4544
	/* Swap primary and spare array */
	thresholds->spare = thresholds->primary;
4545 4546 4547 4548 4549 4550
	/* If all events are unregistered, free the spare array */
	if (!new) {
		kfree(thresholds->spare);
		thresholds->spare = NULL;
	}

4551
	rcu_assign_pointer(thresholds->primary, new);
4552

4553
	/* To be sure that nobody uses thresholds */
4554
	synchronize_rcu();
4555
unlock:
4556 4557
	mutex_unlock(&memcg->thresholds_lock);
}
4558

K
KAMEZAWA Hiroyuki 已提交
4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570
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;

4571
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4572 4573 4574 4575 4576

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

	/* already in OOM ? */
4577
	if (atomic_read(&memcg->under_oom))
K
KAMEZAWA Hiroyuki 已提交
4578
		eventfd_signal(eventfd, 1);
4579
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4580 4581 4582 4583

	return 0;
}

4584
static void mem_cgroup_oom_unregister_event(struct cgroup *cgrp,
K
KAMEZAWA Hiroyuki 已提交
4585 4586
	struct cftype *cft, struct eventfd_ctx *eventfd)
{
4587
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
K
KAMEZAWA Hiroyuki 已提交
4588 4589 4590 4591 4592
	struct mem_cgroup_eventfd_list *ev, *tmp;
	int type = MEMFILE_TYPE(cft->private);

	BUG_ON(type != _OOM_TYPE);

4593
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4594

4595
	list_for_each_entry_safe(ev, tmp, &memcg->oom_notify, list) {
K
KAMEZAWA Hiroyuki 已提交
4596 4597 4598 4599 4600 4601
		if (ev->eventfd == eventfd) {
			list_del(&ev->list);
			kfree(ev);
		}
	}

4602
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4603 4604
}

4605 4606 4607
static int mem_cgroup_oom_control_read(struct cgroup *cgrp,
	struct cftype *cft,  struct cgroup_map_cb *cb)
{
4608
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4609

4610
	cb->fill(cb, "oom_kill_disable", memcg->oom_kill_disable);
4611

4612
	if (atomic_read(&memcg->under_oom))
4613 4614 4615 4616 4617 4618 4619 4620 4621
		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)
{
4622
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633
	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) ||
4634
	    (memcg->use_hierarchy && !list_empty(&cgrp->children))) {
4635 4636 4637
		cgroup_unlock();
		return -EINVAL;
	}
4638
	memcg->oom_kill_disable = val;
4639
	if (!val)
4640
		memcg_oom_recover(memcg);
4641 4642 4643 4644
	cgroup_unlock();
	return 0;
}

4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660
#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 */

4661
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
4662
static int memcg_init_kmem(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
4663
{
4664
	return mem_cgroup_sockets_init(memcg, ss);
4665 4666
};

4667
static void kmem_cgroup_destroy(struct mem_cgroup *memcg)
G
Glauber Costa 已提交
4668
{
4669
	mem_cgroup_sockets_destroy(memcg);
G
Glauber Costa 已提交
4670
}
4671
#else
4672
static int memcg_init_kmem(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
4673 4674 4675
{
	return 0;
}
G
Glauber Costa 已提交
4676

4677
static void kmem_cgroup_destroy(struct mem_cgroup *memcg)
G
Glauber Costa 已提交
4678 4679
{
}
4680 4681
#endif

B
Balbir Singh 已提交
4682 4683
static struct cftype mem_cgroup_files[] = {
	{
4684
		.name = "usage_in_bytes",
4685
		.private = MEMFILE_PRIVATE(_MEM, RES_USAGE),
4686
		.read = mem_cgroup_read,
K
KAMEZAWA Hiroyuki 已提交
4687 4688
		.register_event = mem_cgroup_usage_register_event,
		.unregister_event = mem_cgroup_usage_unregister_event,
B
Balbir Singh 已提交
4689
	},
4690 4691
	{
		.name = "max_usage_in_bytes",
4692
		.private = MEMFILE_PRIVATE(_MEM, RES_MAX_USAGE),
4693
		.trigger = mem_cgroup_reset,
4694
		.read = mem_cgroup_read,
4695
	},
B
Balbir Singh 已提交
4696
	{
4697
		.name = "limit_in_bytes",
4698
		.private = MEMFILE_PRIVATE(_MEM, RES_LIMIT),
4699
		.write_string = mem_cgroup_write,
4700
		.read = mem_cgroup_read,
B
Balbir Singh 已提交
4701
	},
4702 4703 4704 4705
	{
		.name = "soft_limit_in_bytes",
		.private = MEMFILE_PRIVATE(_MEM, RES_SOFT_LIMIT),
		.write_string = mem_cgroup_write,
4706
		.read = mem_cgroup_read,
4707
	},
B
Balbir Singh 已提交
4708 4709
	{
		.name = "failcnt",
4710
		.private = MEMFILE_PRIVATE(_MEM, RES_FAILCNT),
4711
		.trigger = mem_cgroup_reset,
4712
		.read = mem_cgroup_read,
B
Balbir Singh 已提交
4713
	},
4714 4715
	{
		.name = "stat",
4716
		.read_map = mem_control_stat_show,
4717
	},
4718 4719 4720 4721
	{
		.name = "force_empty",
		.trigger = mem_cgroup_force_empty_write,
	},
4722 4723 4724 4725 4726
	{
		.name = "use_hierarchy",
		.write_u64 = mem_cgroup_hierarchy_write,
		.read_u64 = mem_cgroup_hierarchy_read,
	},
K
KOSAKI Motohiro 已提交
4727 4728 4729 4730 4731
	{
		.name = "swappiness",
		.read_u64 = mem_cgroup_swappiness_read,
		.write_u64 = mem_cgroup_swappiness_write,
	},
4732 4733 4734 4735 4736
	{
		.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
4737 4738
	{
		.name = "oom_control",
4739 4740
		.read_map = mem_cgroup_oom_control_read,
		.write_u64 = mem_cgroup_oom_control_write,
K
KAMEZAWA Hiroyuki 已提交
4741 4742 4743 4744
		.register_event = mem_cgroup_oom_register_event,
		.unregister_event = mem_cgroup_oom_unregister_event,
		.private = MEMFILE_PRIVATE(_OOM_TYPE, OOM_CONTROL),
	},
4745 4746 4747 4748
#ifdef CONFIG_NUMA
	{
		.name = "numa_stat",
		.open = mem_control_numa_stat_open,
4749
		.mode = S_IRUGO,
4750 4751
	},
#endif
4752 4753 4754 4755
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
	{
		.name = "memsw.usage_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_USAGE),
4756
		.read = mem_cgroup_read,
K
KAMEZAWA Hiroyuki 已提交
4757 4758
		.register_event = mem_cgroup_usage_register_event,
		.unregister_event = mem_cgroup_usage_unregister_event,
4759 4760 4761 4762 4763
	},
	{
		.name = "memsw.max_usage_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_MAX_USAGE),
		.trigger = mem_cgroup_reset,
4764
		.read = mem_cgroup_read,
4765 4766 4767 4768 4769
	},
	{
		.name = "memsw.limit_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_LIMIT),
		.write_string = mem_cgroup_write,
4770
		.read = mem_cgroup_read,
4771 4772 4773 4774 4775
	},
	{
		.name = "memsw.failcnt",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_FAILCNT),
		.trigger = mem_cgroup_reset,
4776
		.read = mem_cgroup_read,
4777 4778
	},
#endif
4779
	{ },	/* terminate */
4780
};
4781

4782
static int alloc_mem_cgroup_per_zone_info(struct mem_cgroup *memcg, int node)
4783 4784
{
	struct mem_cgroup_per_node *pn;
4785
	struct mem_cgroup_per_zone *mz;
H
Hugh Dickins 已提交
4786
	enum lru_list lru;
4787
	int zone, tmp = node;
4788 4789 4790 4791 4792 4793 4794 4795
	/*
	 * 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.
	 */
4796 4797
	if (!node_state(node, N_NORMAL_MEMORY))
		tmp = -1;
4798
	pn = kzalloc_node(sizeof(*pn), GFP_KERNEL, tmp);
4799 4800
	if (!pn)
		return 1;
4801 4802 4803

	for (zone = 0; zone < MAX_NR_ZONES; zone++) {
		mz = &pn->zoneinfo[zone];
H
Hugh Dickins 已提交
4804 4805
		for_each_lru(lru)
			INIT_LIST_HEAD(&mz->lruvec.lists[lru]);
4806
		mz->usage_in_excess = 0;
4807
		mz->on_tree = false;
4808
		mz->memcg = memcg;
4809
	}
4810
	memcg->info.nodeinfo[node] = pn;
4811 4812 4813
	return 0;
}

4814
static void free_mem_cgroup_per_zone_info(struct mem_cgroup *memcg, int node)
4815
{
4816
	kfree(memcg->info.nodeinfo[node]);
4817 4818
}

4819 4820
static struct mem_cgroup *mem_cgroup_alloc(void)
{
4821
	struct mem_cgroup *memcg;
4822
	int size = sizeof(struct mem_cgroup);
4823

4824
	/* Can be very big if MAX_NUMNODES is very big */
4825
	if (size < PAGE_SIZE)
4826
		memcg = kzalloc(size, GFP_KERNEL);
4827
	else
4828
		memcg = vzalloc(size);
4829

4830
	if (!memcg)
4831 4832
		return NULL;

4833 4834
	memcg->stat = alloc_percpu(struct mem_cgroup_stat_cpu);
	if (!memcg->stat)
4835
		goto out_free;
4836 4837
	spin_lock_init(&memcg->pcp_counter_lock);
	return memcg;
4838 4839 4840

out_free:
	if (size < PAGE_SIZE)
4841
		kfree(memcg);
4842
	else
4843
		vfree(memcg);
4844
	return NULL;
4845 4846
}

4847 4848 4849 4850 4851 4852 4853 4854 4855 4856 4857 4858 4859 4860 4861 4862 4863 4864 4865 4866 4867
/*
 * Helpers for freeing a vzalloc()ed mem_cgroup by RCU,
 * but in process context.  The work_freeing structure is overlaid
 * on the rcu_freeing structure, which itself is overlaid on memsw.
 */
static void vfree_work(struct work_struct *work)
{
	struct mem_cgroup *memcg;

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

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

4868 4869 4870 4871 4872 4873 4874 4875 4876 4877 4878
/*
 * 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.
 */

4879
static void __mem_cgroup_free(struct mem_cgroup *memcg)
4880
{
K
KAMEZAWA Hiroyuki 已提交
4881 4882
	int node;

4883 4884
	mem_cgroup_remove_from_trees(memcg);
	free_css_id(&mem_cgroup_subsys, &memcg->css);
K
KAMEZAWA Hiroyuki 已提交
4885

B
Bob Liu 已提交
4886
	for_each_node(node)
4887
		free_mem_cgroup_per_zone_info(memcg, node);
K
KAMEZAWA Hiroyuki 已提交
4888

4889
	free_percpu(memcg->stat);
4890
	if (sizeof(struct mem_cgroup) < PAGE_SIZE)
4891
		kfree_rcu(memcg, rcu_freeing);
4892
	else
4893
		call_rcu(&memcg->rcu_freeing, vfree_rcu);
4894 4895
}

4896
static void mem_cgroup_get(struct mem_cgroup *memcg)
4897
{
4898
	atomic_inc(&memcg->refcnt);
4899 4900
}

4901
static void __mem_cgroup_put(struct mem_cgroup *memcg, int count)
4902
{
4903 4904 4905
	if (atomic_sub_and_test(count, &memcg->refcnt)) {
		struct mem_cgroup *parent = parent_mem_cgroup(memcg);
		__mem_cgroup_free(memcg);
4906 4907 4908
		if (parent)
			mem_cgroup_put(parent);
	}
4909 4910
}

4911
static void mem_cgroup_put(struct mem_cgroup *memcg)
4912
{
4913
	__mem_cgroup_put(memcg, 1);
4914 4915
}

4916 4917 4918
/*
 * Returns the parent mem_cgroup in memcgroup hierarchy with hierarchy enabled.
 */
G
Glauber Costa 已提交
4919
struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
4920
{
4921
	if (!memcg->res.parent)
4922
		return NULL;
4923
	return mem_cgroup_from_res_counter(memcg->res.parent, res);
4924
}
G
Glauber Costa 已提交
4925
EXPORT_SYMBOL(parent_mem_cgroup);
4926

4927 4928 4929
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
static void __init enable_swap_cgroup(void)
{
4930
	if (!mem_cgroup_disabled() && really_do_swap_account)
4931 4932 4933 4934 4935 4936 4937 4938
		do_swap_account = 1;
}
#else
static void __init enable_swap_cgroup(void)
{
}
#endif

4939 4940 4941 4942 4943 4944
static int mem_cgroup_soft_limit_tree_init(void)
{
	struct mem_cgroup_tree_per_node *rtpn;
	struct mem_cgroup_tree_per_zone *rtpz;
	int tmp, node, zone;

B
Bob Liu 已提交
4945
	for_each_node(node) {
4946 4947 4948 4949 4950
		tmp = node;
		if (!node_state(node, N_NORMAL_MEMORY))
			tmp = -1;
		rtpn = kzalloc_node(sizeof(*rtpn), GFP_KERNEL, tmp);
		if (!rtpn)
4951
			goto err_cleanup;
4952 4953 4954 4955 4956 4957 4958 4959 4960 4961

		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;
4962 4963

err_cleanup:
B
Bob Liu 已提交
4964
	for_each_node(node) {
4965 4966 4967 4968 4969 4970 4971
		if (!soft_limit_tree.rb_tree_per_node[node])
			break;
		kfree(soft_limit_tree.rb_tree_per_node[node]);
		soft_limit_tree.rb_tree_per_node[node] = NULL;
	}
	return 1;

4972 4973
}

L
Li Zefan 已提交
4974
static struct cgroup_subsys_state * __ref
4975
mem_cgroup_create(struct cgroup *cont)
B
Balbir Singh 已提交
4976
{
4977
	struct mem_cgroup *memcg, *parent;
K
KAMEZAWA Hiroyuki 已提交
4978
	long error = -ENOMEM;
4979
	int node;
B
Balbir Singh 已提交
4980

4981 4982
	memcg = mem_cgroup_alloc();
	if (!memcg)
K
KAMEZAWA Hiroyuki 已提交
4983
		return ERR_PTR(error);
4984

B
Bob Liu 已提交
4985
	for_each_node(node)
4986
		if (alloc_mem_cgroup_per_zone_info(memcg, node))
4987
			goto free_out;
4988

4989
	/* root ? */
4990
	if (cont->parent == NULL) {
4991
		int cpu;
4992
		enable_swap_cgroup();
4993
		parent = NULL;
4994 4995
		if (mem_cgroup_soft_limit_tree_init())
			goto free_out;
4996
		root_mem_cgroup = memcg;
4997 4998 4999 5000 5001
		for_each_possible_cpu(cpu) {
			struct memcg_stock_pcp *stock =
						&per_cpu(memcg_stock, cpu);
			INIT_WORK(&stock->work, drain_local_stock);
		}
5002
		hotcpu_notifier(memcg_cpu_hotplug_callback, 0);
5003
	} else {
5004
		parent = mem_cgroup_from_cont(cont->parent);
5005 5006
		memcg->use_hierarchy = parent->use_hierarchy;
		memcg->oom_kill_disable = parent->oom_kill_disable;
5007
	}
5008

5009
	if (parent && parent->use_hierarchy) {
5010 5011
		res_counter_init(&memcg->res, &parent->res);
		res_counter_init(&memcg->memsw, &parent->memsw);
5012 5013 5014 5015 5016 5017 5018
		/*
		 * 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);
5019
	} else {
5020 5021
		res_counter_init(&memcg->res, NULL);
		res_counter_init(&memcg->memsw, NULL);
5022
	}
5023 5024
	memcg->last_scanned_node = MAX_NUMNODES;
	INIT_LIST_HEAD(&memcg->oom_notify);
5025

K
KOSAKI Motohiro 已提交
5026
	if (parent)
5027 5028 5029 5030
		memcg->swappiness = mem_cgroup_swappiness(parent);
	atomic_set(&memcg->refcnt, 1);
	memcg->move_charge_at_immigrate = 0;
	mutex_init(&memcg->thresholds_lock);
5031
	spin_lock_init(&memcg->move_lock);
5032 5033 5034 5035 5036 5037 5038 5039 5040 5041 5042

	error = memcg_init_kmem(memcg, &mem_cgroup_subsys);
	if (error) {
		/*
		 * We call put now because our (and parent's) refcnts
		 * are already in place. mem_cgroup_put() will internally
		 * call __mem_cgroup_free, so return directly
		 */
		mem_cgroup_put(memcg);
		return ERR_PTR(error);
	}
5043
	return &memcg->css;
5044
free_out:
5045
	__mem_cgroup_free(memcg);
K
KAMEZAWA Hiroyuki 已提交
5046
	return ERR_PTR(error);
B
Balbir Singh 已提交
5047 5048
}

5049
static int mem_cgroup_pre_destroy(struct cgroup *cont)
5050
{
5051
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
5052

5053
	return mem_cgroup_force_empty(memcg, false);
5054 5055
}

5056
static void mem_cgroup_destroy(struct cgroup *cont)
B
Balbir Singh 已提交
5057
{
5058
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
5059

5060
	kmem_cgroup_destroy(memcg);
G
Glauber Costa 已提交
5061

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

5065
#ifdef CONFIG_MMU
5066
/* Handlers for move charge at task migration. */
5067 5068
#define PRECHARGE_COUNT_AT_ONCE	256
static int mem_cgroup_do_precharge(unsigned long count)
5069
{
5070 5071
	int ret = 0;
	int batch_count = PRECHARGE_COUNT_AT_ONCE;
5072
	struct mem_cgroup *memcg = mc.to;
5073

5074
	if (mem_cgroup_is_root(memcg)) {
5075 5076 5077 5078 5079 5080 5081 5082
		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;
		/*
5083
		 * "memcg" cannot be under rmdir() because we've already checked
5084 5085 5086 5087
		 * 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().
		 */
5088
		if (res_counter_charge(&memcg->res, PAGE_SIZE * count, &dummy))
5089
			goto one_by_one;
5090
		if (do_swap_account && res_counter_charge(&memcg->memsw,
5091
						PAGE_SIZE * count, &dummy)) {
5092
			res_counter_uncharge(&memcg->res, PAGE_SIZE * count);
5093 5094 5095 5096 5097 5098 5099 5100 5101 5102 5103 5104 5105 5106 5107 5108
			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();
		}
5109 5110
		ret = __mem_cgroup_try_charge(NULL,
					GFP_KERNEL, 1, &memcg, false);
5111
		if (ret)
5112
			/* mem_cgroup_clear_mc() will do uncharge later */
5113
			return ret;
5114 5115
		mc.precharge++;
	}
5116 5117 5118 5119
	return ret;
}

/**
5120
 * get_mctgt_type - get target type of moving charge
5121 5122 5123
 * @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
5124
 * @target: the pointer the target page or swap ent will be stored(can be NULL)
5125 5126 5127 5128 5129 5130
 *
 * 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).
5131 5132 5133
 *   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.
5134 5135 5136 5137 5138
 *
 * Called with pte lock held.
 */
union mc_target {
	struct page	*page;
5139
	swp_entry_t	ent;
5140 5141 5142
};

enum mc_target_type {
5143
	MC_TARGET_NONE = 0,
5144
	MC_TARGET_PAGE,
5145
	MC_TARGET_SWAP,
5146 5147
};

D
Daisuke Nishimura 已提交
5148 5149
static struct page *mc_handle_present_pte(struct vm_area_struct *vma,
						unsigned long addr, pte_t ptent)
5150
{
D
Daisuke Nishimura 已提交
5151
	struct page *page = vm_normal_page(vma, addr, ptent);
5152

D
Daisuke Nishimura 已提交
5153 5154 5155 5156
	if (!page || !page_mapped(page))
		return NULL;
	if (PageAnon(page)) {
		/* we don't move shared anon */
5157
		if (!move_anon() || page_mapcount(page) > 2)
D
Daisuke Nishimura 已提交
5158
			return NULL;
5159 5160
	} else if (!move_file())
		/* we ignore mapcount for file pages */
D
Daisuke Nishimura 已提交
5161 5162 5163 5164 5165 5166 5167 5168 5169 5170 5171 5172 5173 5174 5175 5176 5177 5178
		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 */
5179 5180
		if (page)
			put_page(page);
D
Daisuke Nishimura 已提交
5181
		return NULL;
5182
	}
D
Daisuke Nishimura 已提交
5183 5184 5185 5186 5187 5188
	if (do_swap_account)
		entry->val = ent.val;

	return page;
}

5189 5190 5191 5192 5193 5194 5195 5196 5197 5198 5199 5200 5201 5202 5203 5204 5205 5206 5207 5208 5209
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). */
5210 5211 5212 5213 5214 5215
	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);
5216
		if (do_swap_account)
5217 5218
			*entry = swap;
		page = find_get_page(&swapper_space, swap.val);
5219
	}
5220
#endif
5221 5222 5223
	return page;
}

5224
static enum mc_target_type get_mctgt_type(struct vm_area_struct *vma,
D
Daisuke Nishimura 已提交
5225 5226 5227 5228
		unsigned long addr, pte_t ptent, union mc_target *target)
{
	struct page *page = NULL;
	struct page_cgroup *pc;
5229
	enum mc_target_type ret = MC_TARGET_NONE;
D
Daisuke Nishimura 已提交
5230 5231 5232 5233 5234 5235
	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);
5236 5237
	else if (pte_none(ptent) || pte_file(ptent))
		page = mc_handle_file_pte(vma, addr, ptent, &ent);
D
Daisuke Nishimura 已提交
5238 5239

	if (!page && !ent.val)
5240
		return ret;
5241 5242 5243 5244 5245 5246 5247 5248 5249 5250 5251 5252 5253 5254 5255
	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
5256 5257
	/* There is a swap entry and a page doesn't exist or isn't charged */
	if (ent.val && !ret &&
5258
			css_id(&mc.from->css) == lookup_swap_cgroup_id(ent)) {
5259 5260 5261
		ret = MC_TARGET_SWAP;
		if (target)
			target->ent = ent;
5262 5263 5264 5265
	}
	return ret;
}

5266 5267 5268 5269 5270 5271 5272 5273 5274 5275 5276 5277 5278 5279 5280 5281 5282 5283 5284 5285 5286 5287 5288 5289 5290 5291 5292 5293 5294 5295 5296 5297 5298 5299 5300
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
/*
 * We don't consider swapping or file mapped pages because THP does not
 * support them for now.
 * Caller should make sure that pmd_trans_huge(pmd) is true.
 */
static enum mc_target_type get_mctgt_type_thp(struct vm_area_struct *vma,
		unsigned long addr, pmd_t pmd, union mc_target *target)
{
	struct page *page = NULL;
	struct page_cgroup *pc;
	enum mc_target_type ret = MC_TARGET_NONE;

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

5301 5302 5303 5304 5305 5306 5307 5308
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;

5309 5310 5311 5312
	if (pmd_trans_huge_lock(pmd, vma) == 1) {
		if (get_mctgt_type_thp(vma, addr, *pmd, NULL) == MC_TARGET_PAGE)
			mc.precharge += HPAGE_PMD_NR;
		spin_unlock(&vma->vm_mm->page_table_lock);
5313
		return 0;
5314
	}
5315

5316 5317
	if (pmd_trans_unstable(pmd))
		return 0;
5318 5319
	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
	for (; addr != end; pte++, addr += PAGE_SIZE)
5320
		if (get_mctgt_type(vma, addr, *pte, NULL))
5321 5322 5323 5324
			mc.precharge++;	/* increment precharge temporarily */
	pte_unmap_unlock(pte - 1, ptl);
	cond_resched();

5325 5326 5327
	return 0;
}

5328 5329 5330 5331 5332
static unsigned long mem_cgroup_count_precharge(struct mm_struct *mm)
{
	unsigned long precharge;
	struct vm_area_struct *vma;

5333
	down_read(&mm->mmap_sem);
5334 5335 5336 5337 5338 5339 5340 5341 5342 5343 5344
	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);
	}
5345
	up_read(&mm->mmap_sem);
5346 5347 5348 5349 5350 5351 5352 5353 5354

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

	return precharge;
}

static int mem_cgroup_precharge_mc(struct mm_struct *mm)
{
5355 5356 5357 5358 5359
	unsigned long precharge = mem_cgroup_count_precharge(mm);

	VM_BUG_ON(mc.moving_task);
	mc.moving_task = current;
	return mem_cgroup_do_precharge(precharge);
5360 5361
}

5362 5363
/* cancels all extra charges on mc.from and mc.to, and wakes up all waiters. */
static void __mem_cgroup_clear_mc(void)
5364
{
5365 5366 5367
	struct mem_cgroup *from = mc.from;
	struct mem_cgroup *to = mc.to;

5368
	/* we must uncharge all the leftover precharges from mc.to */
5369 5370 5371 5372 5373 5374 5375 5376 5377 5378 5379
	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;
5380
	}
5381 5382 5383 5384 5385 5386 5387 5388 5389 5390 5391 5392 5393 5394 5395 5396 5397 5398 5399
	/* 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;
	}
5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410 5411 5412 5413 5414
	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();
5415
	spin_lock(&mc.lock);
5416 5417
	mc.from = NULL;
	mc.to = NULL;
5418
	spin_unlock(&mc.lock);
5419
	mem_cgroup_end_move(from);
5420 5421
}

5422 5423
static int mem_cgroup_can_attach(struct cgroup *cgroup,
				 struct cgroup_taskset *tset)
5424
{
5425
	struct task_struct *p = cgroup_taskset_first(tset);
5426
	int ret = 0;
5427
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgroup);
5428

5429
	if (memcg->move_charge_at_immigrate) {
5430 5431 5432
		struct mm_struct *mm;
		struct mem_cgroup *from = mem_cgroup_from_task(p);

5433
		VM_BUG_ON(from == memcg);
5434 5435 5436 5437 5438

		mm = get_task_mm(p);
		if (!mm)
			return 0;
		/* We move charges only when we move a owner of the mm */
5439 5440 5441 5442
		if (mm->owner == p) {
			VM_BUG_ON(mc.from);
			VM_BUG_ON(mc.to);
			VM_BUG_ON(mc.precharge);
5443
			VM_BUG_ON(mc.moved_charge);
5444
			VM_BUG_ON(mc.moved_swap);
5445
			mem_cgroup_start_move(from);
5446
			spin_lock(&mc.lock);
5447
			mc.from = from;
5448
			mc.to = memcg;
5449
			spin_unlock(&mc.lock);
5450
			/* We set mc.moving_task later */
5451 5452 5453 5454

			ret = mem_cgroup_precharge_mc(mm);
			if (ret)
				mem_cgroup_clear_mc();
5455 5456
		}
		mmput(mm);
5457 5458 5459 5460
	}
	return ret;
}

5461 5462
static void mem_cgroup_cancel_attach(struct cgroup *cgroup,
				     struct cgroup_taskset *tset)
5463
{
5464
	mem_cgroup_clear_mc();
5465 5466
}

5467 5468 5469
static int mem_cgroup_move_charge_pte_range(pmd_t *pmd,
				unsigned long addr, unsigned long end,
				struct mm_walk *walk)
5470
{
5471 5472 5473 5474
	int ret = 0;
	struct vm_area_struct *vma = walk->private;
	pte_t *pte;
	spinlock_t *ptl;
5475 5476 5477 5478
	enum mc_target_type target_type;
	union mc_target target;
	struct page *page;
	struct page_cgroup *pc;
5479

5480 5481 5482 5483 5484 5485 5486 5487 5488 5489 5490
	/*
	 * We don't take compound_lock() here but no race with splitting thp
	 * happens because:
	 *  - if pmd_trans_huge_lock() returns 1, the relevant thp is not
	 *    under splitting, which means there's no concurrent thp split,
	 *  - if another thread runs into split_huge_page() just after we
	 *    entered this if-block, the thread must wait for page table lock
	 *    to be unlocked in __split_huge_page_splitting(), where the main
	 *    part of thp split is not executed yet.
	 */
	if (pmd_trans_huge_lock(pmd, vma) == 1) {
5491
		if (mc.precharge < HPAGE_PMD_NR) {
5492 5493 5494 5495 5496 5497 5498 5499 5500 5501 5502 5503 5504 5505 5506 5507 5508 5509 5510
			spin_unlock(&vma->vm_mm->page_table_lock);
			return 0;
		}
		target_type = get_mctgt_type_thp(vma, addr, *pmd, &target);
		if (target_type == MC_TARGET_PAGE) {
			page = target.page;
			if (!isolate_lru_page(page)) {
				pc = lookup_page_cgroup(page);
				if (!mem_cgroup_move_account(page, HPAGE_PMD_NR,
							     pc, mc.from, mc.to,
							     false)) {
					mc.precharge -= HPAGE_PMD_NR;
					mc.moved_charge += HPAGE_PMD_NR;
				}
				putback_lru_page(page);
			}
			put_page(page);
		}
		spin_unlock(&vma->vm_mm->page_table_lock);
5511
		return 0;
5512 5513
	}

5514 5515
	if (pmd_trans_unstable(pmd))
		return 0;
5516 5517 5518 5519
retry:
	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
	for (; addr != end; addr += PAGE_SIZE) {
		pte_t ptent = *(pte++);
5520
		swp_entry_t ent;
5521 5522 5523 5524

		if (!mc.precharge)
			break;

5525
		switch (get_mctgt_type(vma, addr, ptent, &target)) {
5526 5527 5528 5529 5530
		case MC_TARGET_PAGE:
			page = target.page;
			if (isolate_lru_page(page))
				goto put;
			pc = lookup_page_cgroup(page);
5531 5532
			if (!mem_cgroup_move_account(page, 1, pc,
						     mc.from, mc.to, false)) {
5533
				mc.precharge--;
5534 5535
				/* we uncharge from mc.from later. */
				mc.moved_charge++;
5536 5537
			}
			putback_lru_page(page);
5538
put:			/* get_mctgt_type() gets the page */
5539 5540
			put_page(page);
			break;
5541 5542
		case MC_TARGET_SWAP:
			ent = target.ent;
5543 5544
			if (!mem_cgroup_move_swap_account(ent,
						mc.from, mc.to, false)) {
5545
				mc.precharge--;
5546 5547 5548
				/* we fixup refcnts and charges later. */
				mc.moved_swap++;
			}
5549
			break;
5550 5551 5552 5553 5554 5555 5556 5557 5558 5559 5560 5561 5562 5563
		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.
		 */
5564
		ret = mem_cgroup_do_precharge(1);
5565 5566 5567 5568 5569 5570 5571 5572 5573 5574 5575 5576
		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();
5577 5578 5579 5580 5581 5582 5583 5584 5585 5586 5587 5588 5589
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;
	}
5590 5591 5592 5593 5594 5595 5596 5597 5598 5599 5600 5601 5602 5603 5604 5605 5606 5607
	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;
	}
5608
	up_read(&mm->mmap_sem);
5609 5610
}

5611 5612
static void mem_cgroup_move_task(struct cgroup *cont,
				 struct cgroup_taskset *tset)
B
Balbir Singh 已提交
5613
{
5614
	struct task_struct *p = cgroup_taskset_first(tset);
5615
	struct mm_struct *mm = get_task_mm(p);
5616 5617

	if (mm) {
5618 5619
		if (mc.to)
			mem_cgroup_move_charge(mm);
5620 5621
		mmput(mm);
	}
5622 5623
	if (mc.to)
		mem_cgroup_clear_mc();
B
Balbir Singh 已提交
5624
}
5625
#else	/* !CONFIG_MMU */
5626 5627
static int mem_cgroup_can_attach(struct cgroup *cgroup,
				 struct cgroup_taskset *tset)
5628 5629 5630
{
	return 0;
}
5631 5632
static void mem_cgroup_cancel_attach(struct cgroup *cgroup,
				     struct cgroup_taskset *tset)
5633 5634
{
}
5635 5636
static void mem_cgroup_move_task(struct cgroup *cont,
				 struct cgroup_taskset *tset)
5637 5638 5639
{
}
#endif
B
Balbir Singh 已提交
5640

B
Balbir Singh 已提交
5641 5642 5643 5644
struct cgroup_subsys mem_cgroup_subsys = {
	.name = "memory",
	.subsys_id = mem_cgroup_subsys_id,
	.create = mem_cgroup_create,
5645
	.pre_destroy = mem_cgroup_pre_destroy,
B
Balbir Singh 已提交
5646
	.destroy = mem_cgroup_destroy,
5647 5648
	.can_attach = mem_cgroup_can_attach,
	.cancel_attach = mem_cgroup_cancel_attach,
B
Balbir Singh 已提交
5649
	.attach = mem_cgroup_move_task,
5650
	.base_cftypes = mem_cgroup_files,
5651
	.early_init = 0,
K
KAMEZAWA Hiroyuki 已提交
5652
	.use_id = 1,
5653
	.__DEPRECATED_clear_css_refs = true,
B
Balbir Singh 已提交
5654
};
5655 5656

#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
5657 5658 5659
static int __init enable_swap_account(char *s)
{
	/* consider enabled if no parameter or 1 is given */
5660
	if (!strcmp(s, "1"))
5661
		really_do_swap_account = 1;
5662
	else if (!strcmp(s, "0"))
5663 5664 5665
		really_do_swap_account = 0;
	return 1;
}
5666
__setup("swapaccount=", enable_swap_account);
5667 5668

#endif