memcontrol.c 145.5 KB
Newer Older
B
Balbir Singh 已提交
1 2 3 4 5
/* memcontrol.c - Memory Controller
 *
 * Copyright IBM Corporation, 2007
 * Author Balbir Singh <balbir@linux.vnet.ibm.com>
 *
6 7 8
 * Copyright 2007 OpenVZ SWsoft Inc
 * Author: Pavel Emelianov <xemul@openvz.org>
 *
9 10 11 12
 * Memory thresholds
 * Copyright (C) 2009 Nokia Corporation
 * Author: Kirill A. Shutemov
 *
B
Balbir Singh 已提交
13 14 15 16 17 18 19 20 21 22 23 24 25 26
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 */

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

56 57
#include <asm/uaccess.h>

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

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

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

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

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


80 81 82 83 84 85 86 87
/*
 * Statistics for memory cgroup.
 */
enum mem_cgroup_stat_index {
	/*
	 * For MEM_CONTAINER_TYPE_ALL, usage = pagecache + rss.
	 */
	MEM_CGROUP_STAT_CACHE, 	   /* # of pages charged as cache */
88
	MEM_CGROUP_STAT_RSS,	   /* # of pages charged as anon rss */
89
	MEM_CGROUP_STAT_FILE_MAPPED,  /* # of pages charged as file rss */
90
	MEM_CGROUP_STAT_SWAPOUT, /* # of pages, swapped out */
91
	MEM_CGROUP_STAT_DATA, /* end of data requires synchronization */
92
	MEM_CGROUP_ON_MOVE,	/* someone is moving account between groups */
93 94 95
	MEM_CGROUP_STAT_NSTATS,
};

96 97 98 99
enum mem_cgroup_events_index {
	MEM_CGROUP_EVENTS_PGPGIN,	/* # of pages paged in */
	MEM_CGROUP_EVENTS_PGPGOUT,	/* # of pages paged out */
	MEM_CGROUP_EVENTS_COUNT,	/* # of pages paged in/out */
100 101
	MEM_CGROUP_EVENTS_PGFAULT,	/* # of page-faults */
	MEM_CGROUP_EVENTS_PGMAJFAULT,	/* # of major page-faults */
102 103
	MEM_CGROUP_EVENTS_NSTATS,
};
104 105 106 107 108 109 110 111 112
/*
 * Per memcg event counter is incremented at every pagein/pageout. With THP,
 * it will be incremated by the number of pages. This counter is used for
 * for trigger some periodic events. This is straightforward and better
 * than using jiffies etc. to handle periodic memcg event.
 */
enum mem_cgroup_events_target {
	MEM_CGROUP_TARGET_THRESH,
	MEM_CGROUP_TARGET_SOFTLIMIT,
113
	MEM_CGROUP_TARGET_NUMAINFO,
114 115 116 117
	MEM_CGROUP_NTARGETS,
};
#define THRESHOLDS_EVENTS_TARGET (128)
#define SOFTLIMIT_EVENTS_TARGET (1024)
118
#define NUMAINFO_EVENTS_TARGET	(1024)
119

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

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

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

140 141
	struct mem_cgroup_reclaim_iter reclaim_iter[DEF_PRIORITY + 1];

K
KOSAKI Motohiro 已提交
142
	struct zone_reclaim_stat reclaim_stat;
143 144 145 146
	struct rb_node		tree_node;	/* RB tree node */
	unsigned long long	usage_in_excess;/* Set to the value by which */
						/* the soft limit is exceeded*/
	bool			on_tree;
147 148
	struct mem_cgroup	*mem;		/* Back pointer, we cannot */
						/* use container_of	   */
149 150 151 152 153 154 155 156 157 158 159 160
};
/* Macro for accessing counter */
#define MEM_CGROUP_ZSTAT(mz, idx)	((mz)->count[(idx)])

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

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

161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180
/*
 * Cgroups above their limits are maintained in a RB-Tree, independent of
 * their hierarchy representation
 */

struct mem_cgroup_tree_per_zone {
	struct rb_root rb_root;
	spinlock_t lock;
};

struct mem_cgroup_tree_per_node {
	struct mem_cgroup_tree_per_zone rb_tree_per_zone[MAX_NR_ZONES];
};

struct mem_cgroup_tree {
	struct mem_cgroup_tree_per_node *rb_tree_per_node[MAX_NUMNODES];
};

static struct mem_cgroup_tree soft_limit_tree __read_mostly;

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

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

struct mem_cgroup_thresholds {
	/* Primary thresholds array */
	struct mem_cgroup_threshold_ary *primary;
	/*
	 * Spare threshold array.
	 * This is needed to make mem_cgroup_unregister_event() "never fail".
	 * It must be able to store at least primary->size - 1 entries.
	 */
	struct mem_cgroup_threshold_ary *spare;
};

K
KAMEZAWA Hiroyuki 已提交
207 208 209 210 211
/* for OOM */
struct mem_cgroup_eventfd_list {
	struct list_head list;
	struct eventfd_ctx *eventfd;
};
212

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

B
Balbir Singh 已提交
216 217 218 219 220 221 222
/*
 * The memory controller data structure. The memory controller controls both
 * page cache and RSS per cgroup. We would eventually like to provide
 * statistics based on the statistics developed by Rik Van Riel for clock-pro,
 * to help the administrator determine what knobs to tune.
 *
 * TODO: Add a water mark for the memory controller. Reclaim will begin when
223 224 225
 * we hit the water mark. May be even add a low water mark, such that
 * no reclaim occurs from a cgroup at it's low water mark, this is
 * a feature that will be implemented much later in the future.
B
Balbir Singh 已提交
226 227 228 229 230 231 232
 */
struct mem_cgroup {
	struct cgroup_subsys_state css;
	/*
	 * the counter to account for memory usage
	 */
	struct res_counter res;
233 234 235 236
	/*
	 * the counter to account for mem+swap usage.
	 */
	struct res_counter memsw;
237 238 239 240
	/*
	 * Per cgroup active and inactive list, similar to the
	 * per zone LRU lists.
	 */
241
	struct mem_cgroup_lru_info info;
242 243 244
	int last_scanned_node;
#if MAX_NUMNODES > 1
	nodemask_t	scan_nodes;
245 246
	atomic_t	numainfo_events;
	atomic_t	numainfo_updating;
247
#endif
248 249 250 251
	/*
	 * Should the accounting and control be hierarchical, per subtree?
	 */
	bool use_hierarchy;
252 253 254 255

	bool		oom_lock;
	atomic_t	under_oom;

256
	atomic_t	refcnt;
257

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

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

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

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

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

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

277 278 279 280 281
	/*
	 * Should we move charges of a task when a task is moved into this
	 * mem_cgroup ? And what type of charges should we move ?
	 */
	unsigned long 	move_charge_at_immigrate;
282
	/*
283
	 * percpu counter.
284
	 */
285
	struct mem_cgroup_stat_cpu *stat;
286 287 288 289 290 291
	/*
	 * used when a cpu is offlined or other synchronizations
	 * See mem_cgroup_read_stat().
	 */
	struct mem_cgroup_stat_cpu nocpu_base;
	spinlock_t pcp_counter_lock;
G
Glauber Costa 已提交
292 293 294 295

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

298 299 300 301 302 303
/* Stuffs for move charges at task migration. */
/*
 * Types of charges to be moved. "move_charge_at_immitgrate" is treated as a
 * left-shifted bitmap of these types.
 */
enum move_type {
304
	MOVE_CHARGE_TYPE_ANON,	/* private anonymous page and swap of it */
305
	MOVE_CHARGE_TYPE_FILE,	/* file page(including tmpfs) and swap of it */
306 307 308
	NR_MOVE_TYPE,
};

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

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

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

336 337 338 339 340 341 342
/*
 * Maximum loops in mem_cgroup_hierarchical_reclaim(), used for soft
 * limit reclaim to prevent infinite loops, if they ever occur.
 */
#define	MEM_CGROUP_MAX_RECLAIM_LOOPS		(100)
#define	MEM_CGROUP_MAX_SOFT_LIMIT_RECLAIM_LOOPS	(2)

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

353
/* for encoding cft->private value on file */
354 355 356
#define _MEM			(0)
#define _MEMSWAP		(1)
#define _OOM_TYPE		(2)
357 358 359
#define MEMFILE_PRIVATE(x, val)	(((x) << 16) | (val))
#define MEMFILE_TYPE(val)	(((val) >> 16) & 0xffff)
#define MEMFILE_ATTR(val)	((val) & 0xffff)
K
KAMEZAWA Hiroyuki 已提交
360 361
/* Used for OOM nofiier */
#define OOM_CONTROL		(0)
362

363 364 365 366 367 368 369 370
/*
 * Reclaim flags for mem_cgroup_hierarchical_reclaim
 */
#define MEM_CGROUP_RECLAIM_NOSWAP_BIT	0x0
#define MEM_CGROUP_RECLAIM_NOSWAP	(1 << MEM_CGROUP_RECLAIM_NOSWAP_BIT)
#define MEM_CGROUP_RECLAIM_SHRINK_BIT	0x1
#define MEM_CGROUP_RECLAIM_SHRINK	(1 << MEM_CGROUP_RECLAIM_SHRINK_BIT)

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

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

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

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

388 389 390 391 392 393 394 395 396 397 398 399 400 401
		/* Socket cloning can throw us here with sk_cgrp already
		 * filled. It won't however, necessarily happen from
		 * process context. So the test for root memcg given
		 * the current task's memcg won't help us in this case.
		 *
		 * Respecting the original socket's memcg is a better
		 * decision in this case.
		 */
		if (sk->sk_cgrp) {
			BUG_ON(mem_cgroup_is_root(sk->sk_cgrp->memcg));
			mem_cgroup_get(sk->sk_cgrp->memcg);
			return;
		}

G
Glauber Costa 已提交
402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421
		rcu_read_lock();
		memcg = mem_cgroup_from_task(current);
		if (!mem_cgroup_is_root(memcg)) {
			mem_cgroup_get(memcg);
			sk->sk_cgrp = sk->sk_prot->proto_cgroup(memcg);
		}
		rcu_read_unlock();
	}
}
EXPORT_SYMBOL(sock_update_memcg);

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

struct cg_proto *tcp_proto_cgroup(struct mem_cgroup *memcg)
{
	if (!memcg || mem_cgroup_is_root(memcg))
		return NULL;

	return &memcg->tcp_mem.cg_proto;
}
EXPORT_SYMBOL(tcp_proto_cgroup);
G
Glauber Costa 已提交
431 432 433
#endif /* CONFIG_INET */
#endif /* CONFIG_CGROUP_MEM_RES_CTLR_KMEM */

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

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

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

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

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

static struct mem_cgroup_tree_per_zone *
soft_limit_tree_node_zone(int nid, int zid)
{
	return &soft_limit_tree.rb_tree_per_node[nid]->rb_tree_per_zone[zid];
}

static struct mem_cgroup_tree_per_zone *
soft_limit_tree_from_page(struct page *page)
{
	int nid = page_to_nid(page);
	int zid = page_zonenum(page);

	return &soft_limit_tree.rb_tree_per_node[nid]->rb_tree_per_zone[zid];
}

static void
472
__mem_cgroup_insert_exceeded(struct mem_cgroup *memcg,
473
				struct mem_cgroup_per_zone *mz,
474 475
				struct mem_cgroup_tree_per_zone *mctz,
				unsigned long long new_usage_in_excess)
476 477 478 479 480 481 482 483
{
	struct rb_node **p = &mctz->rb_root.rb_node;
	struct rb_node *parent = NULL;
	struct mem_cgroup_per_zone *mz_node;

	if (mz->on_tree)
		return;

484 485 486
	mz->usage_in_excess = new_usage_in_excess;
	if (!mz->usage_in_excess)
		return;
487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502
	while (*p) {
		parent = *p;
		mz_node = rb_entry(parent, struct mem_cgroup_per_zone,
					tree_node);
		if (mz->usage_in_excess < mz_node->usage_in_excess)
			p = &(*p)->rb_left;
		/*
		 * We can't avoid mem cgroups that are over their soft
		 * limit by the same amount
		 */
		else if (mz->usage_in_excess >= mz_node->usage_in_excess)
			p = &(*p)->rb_right;
	}
	rb_link_node(&mz->tree_node, parent, p);
	rb_insert_color(&mz->tree_node, &mctz->rb_root);
	mz->on_tree = true;
503 504 505
}

static void
506
__mem_cgroup_remove_exceeded(struct mem_cgroup *memcg,
507 508 509 510 511 512 513 514 515
				struct mem_cgroup_per_zone *mz,
				struct mem_cgroup_tree_per_zone *mctz)
{
	if (!mz->on_tree)
		return;
	rb_erase(&mz->tree_node, &mctz->rb_root);
	mz->on_tree = false;
}

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


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

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

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

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

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

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

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

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

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

614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632
/*
 * Implementation Note: reading percpu statistics for memcg.
 *
 * Both of vmstat[] and percpu_counter has threshold and do periodic
 * synchronization to implement "quick" read. There are trade-off between
 * reading cost and precision of value. Then, we may have a chance to implement
 * a periodic synchronizion of counter in memcg's counter.
 *
 * But this _read() function is used for user interface now. The user accounts
 * memory usage by memory cgroup and he _always_ requires exact value because
 * he accounts memory. Even if we provide quick-and-fuzzy read, we always
 * have to visit all online cpus and make sum. So, for now, unnecessary
 * synchronization is not implemented. (just implemented for cpu hotplug)
 *
 * If there are kernel internal actions which can make use of some not-exact
 * value, and reading all cpu value can be performance bottleneck in some
 * common workload, threashold and synchonization as vmstat[] should be
 * implemented.
 */
633
static long mem_cgroup_read_stat(struct mem_cgroup *memcg,
634
				 enum mem_cgroup_stat_index idx)
635
{
636
	long val = 0;
637 638
	int cpu;

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

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

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

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

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

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

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

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

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

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

706
	preempt_enable();
707 708
}

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

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

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

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

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

737 738
	return total;
}
739

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

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

751 752
static bool mem_cgroup_event_ratelimit(struct mem_cgroup *memcg,
				       enum mem_cgroup_events_target target)
753 754 755
{
	unsigned long val, next;

756 757
	val = __this_cpu_read(memcg->stat->events[MEM_CGROUP_EVENTS_COUNT]);
	next = __this_cpu_read(memcg->stat->targets[target]);
758
	/* from time_after() in jiffies.h */
759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774
	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;
775
	}
776
	return false;
777 778 779 780 781 782
}

/*
 * Check events in order.
 *
 */
783
static void memcg_check_events(struct mem_cgroup *memcg, struct page *page)
784
{
785
	preempt_disable();
786
	/* threshold event is triggered in finer grain than soft limit */
787 788 789 790 791 792 793 794 795 796 797 798
	if (unlikely(mem_cgroup_event_ratelimit(memcg,
						MEM_CGROUP_TARGET_THRESH))) {
		bool do_softlimit, do_numainfo;

		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();

799
		mem_cgroup_threshold(memcg);
800
		if (unlikely(do_softlimit))
801
			mem_cgroup_update_tree(memcg, page);
802
#if MAX_NUMNODES > 1
803
		if (unlikely(do_numainfo))
804
			atomic_inc(&memcg->numainfo_events);
805
#endif
806 807
	} else
		preempt_enable();
808 809
}

G
Glauber Costa 已提交
810
struct mem_cgroup *mem_cgroup_from_cont(struct cgroup *cont)
B
Balbir Singh 已提交
811 812 813 814 815 816
{
	return container_of(cgroup_subsys_state(cont,
				mem_cgroup_subsys_id), struct mem_cgroup,
				css);
}

817
struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p)
818
{
819 820 821 822 823 824 825 826
	/*
	 * 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;

827 828 829 830
	return container_of(task_subsys_state(p, mem_cgroup_subsys_id),
				struct mem_cgroup, css);
}

831
struct mem_cgroup *try_get_mem_cgroup_from_mm(struct mm_struct *mm)
832
{
833
	struct mem_cgroup *memcg = NULL;
834 835 836

	if (!mm)
		return NULL;
837 838 839 840 841 842 843
	/*
	 * 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 {
844 845
		memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
		if (unlikely(!memcg))
846
			break;
847
	} while (!css_tryget(&memcg->css));
848
	rcu_read_unlock();
849
	return memcg;
850 851
}

852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871
/**
 * 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
872
{
873 874
	struct mem_cgroup *memcg = NULL;
	int id = 0;
875

876 877 878
	if (mem_cgroup_disabled())
		return NULL;

879 880
	if (!root)
		root = root_mem_cgroup;
K
KAMEZAWA Hiroyuki 已提交
881

882 883
	if (prev && !reclaim)
		id = css_id(&prev->css);
K
KAMEZAWA Hiroyuki 已提交
884

885 886
	if (prev && prev != root)
		css_put(&prev->css);
K
KAMEZAWA Hiroyuki 已提交
887

888 889 890 891 892
	if (!root->use_hierarchy && root != root_mem_cgroup) {
		if (prev)
			return NULL;
		return root;
	}
K
KAMEZAWA Hiroyuki 已提交
893

894
	while (!memcg) {
895
		struct mem_cgroup_reclaim_iter *uninitialized_var(iter);
896
		struct cgroup_subsys_state *css;
897

898 899 900 901 902 903 904 905 906 907 908
		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
909

910 911 912 913 914 915 916 917
		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
918 919
		rcu_read_unlock();

920 921 922 923 924 925 926
		if (reclaim) {
			iter->position = id;
			if (!css)
				iter->generation++;
			else if (!prev && memcg)
				reclaim->generation = iter->generation;
		}
927 928 929 930 931

		if (prev && !css)
			return NULL;
	}
	return memcg;
K
KAMEZAWA Hiroyuki 已提交
932
}
K
KAMEZAWA Hiroyuki 已提交
933

934 935 936 937 938 939 940
/**
 * 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)
941 942 943 944 945 946
{
	if (!root)
		root = root_mem_cgroup;
	if (prev && prev != root)
		css_put(&prev->css);
}
K
KAMEZAWA Hiroyuki 已提交
947

948 949 950 951 952 953
/*
 * 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)		\
954
	for (iter = mem_cgroup_iter(root, NULL, NULL);	\
955
	     iter != NULL;				\
956
	     iter = mem_cgroup_iter(root, iter, NULL))
957

958
#define for_each_mem_cgroup(iter)			\
959
	for (iter = mem_cgroup_iter(NULL, NULL, NULL);	\
960
	     iter != NULL;				\
961
	     iter = mem_cgroup_iter(NULL, iter, NULL))
K
KAMEZAWA Hiroyuki 已提交
962

963
static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
964
{
965
	return (memcg == root_mem_cgroup);
966 967
}

968 969
void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
{
970
	struct mem_cgroup *memcg;
971 972 973 974 975

	if (!mm)
		return;

	rcu_read_lock();
976 977
	memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
	if (unlikely(!memcg))
978 979 980 981
		goto out;

	switch (idx) {
	case PGMAJFAULT:
982
		mem_cgroup_pgmajfault(memcg, 1);
983 984
		break;
	case PGFAULT:
985
		mem_cgroup_pgfault(memcg, 1);
986 987 988 989 990 991 992 993 994
		break;
	default:
		BUG();
	}
out:
	rcu_read_unlock();
}
EXPORT_SYMBOL(mem_cgroup_count_vm_event);

995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015
/**
 * 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
1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028
/*
 * 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.
 */
1029

1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043
/**
 * 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
1044 1045
{
	struct mem_cgroup_per_zone *mz;
1046 1047
	struct mem_cgroup *memcg;
	struct page_cgroup *pc;
1048

1049
	if (mem_cgroup_disabled())
1050 1051
		return &zone->lruvec;

K
KAMEZAWA Hiroyuki 已提交
1052
	pc = lookup_page_cgroup(page);
1053
	VM_BUG_ON(PageCgroupAcctLRU(pc));
1054
	/*
1055 1056 1057 1058 1059 1060 1061
	 * putback:				charge:
	 * SetPageLRU				SetPageCgroupUsed
	 * smp_mb				smp_mb
	 * PageCgroupUsed && add to memcg LRU	PageLRU && add to memcg LRU
	 *
	 * Ensure that one of the two sides adds the page to the memcg
	 * LRU during a race.
1062
	 */
1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080
	smp_mb();
	/*
	 * If the page is uncharged, it may be freed soon, but it
	 * could also be swap cache (readahead, swapoff) that needs to
	 * be reclaimable in the future.  root_mem_cgroup will babysit
	 * it for the time being.
	 */
	if (PageCgroupUsed(pc)) {
		/* Ensure pc->mem_cgroup is visible after reading PCG_USED. */
		smp_rmb();
		memcg = pc->mem_cgroup;
		SetPageCgroupAcctLRU(pc);
	} else
		memcg = root_mem_cgroup;
	mz = page_cgroup_zoneinfo(memcg, page);
	/* compound_order() is stabilized through lru_lock */
	MEM_CGROUP_ZSTAT(mz, lru) += 1 << compound_order(page);
	return &mz->lruvec;
K
KAMEZAWA Hiroyuki 已提交
1081
}
1082

1083 1084 1085 1086 1087 1088 1089 1090 1091
/**
 * 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.
1092
 */
1093
void mem_cgroup_lru_del_list(struct page *page, enum lru_list lru)
1094 1095
{
	struct mem_cgroup_per_zone *mz;
1096
	struct mem_cgroup *memcg;
1097 1098 1099 1100 1101 1102
	struct page_cgroup *pc;

	if (mem_cgroup_disabled())
		return;

	pc = lookup_page_cgroup(page);
1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117
	/*
	 * root_mem_cgroup babysits uncharged LRU pages, but
	 * PageCgroupUsed is cleared when the page is about to get
	 * freed.  PageCgroupAcctLRU remembers whether the
	 * LRU-accounting happened against pc->mem_cgroup or
	 * root_mem_cgroup.
	 */
	if (TestClearPageCgroupAcctLRU(pc)) {
		VM_BUG_ON(!pc->mem_cgroup);
		memcg = pc->mem_cgroup;
	} else
		memcg = root_mem_cgroup;
	mz = page_cgroup_zoneinfo(memcg, page);
	/* huge page split is done under lru_lock. so, we have no races. */
	MEM_CGROUP_ZSTAT(mz, lru) -= 1 << compound_order(page);
1118 1119
}

1120
void mem_cgroup_lru_del(struct page *page)
K
KAMEZAWA Hiroyuki 已提交
1121
{
1122
	mem_cgroup_lru_del_list(page, page_lru(page));
1123 1124
}

1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142
/**
 * 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)
1143
{
1144 1145 1146
	/* 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
1147
}
1148

K
KAMEZAWA Hiroyuki 已提交
1149
/*
1150 1151 1152 1153
 * At handling SwapCache and other FUSE stuff, pc->mem_cgroup may be changed
 * while it's linked to lru because the page may be reused after it's fully
 * uncharged. To handle that, unlink page_cgroup from LRU when charge it again.
 * It's done under lock_page and expected that zone->lru_lock isnever held.
K
KAMEZAWA Hiroyuki 已提交
1154
 */
1155
static void mem_cgroup_lru_del_before_commit(struct page *page)
K
KAMEZAWA Hiroyuki 已提交
1156
{
1157
	enum lru_list lru;
1158 1159 1160 1161
	unsigned long flags;
	struct zone *zone = page_zone(page);
	struct page_cgroup *pc = lookup_page_cgroup(page);

1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172
	/*
	 * Doing this check without taking ->lru_lock seems wrong but this
	 * is safe. Because if page_cgroup's USED bit is unset, the page
	 * will not be added to any memcg's LRU. If page_cgroup's USED bit is
	 * set, the commit after this will fail, anyway.
	 * This all charge/uncharge is done under some mutual execustion.
	 * So, we don't need to taking care of changes in USED bit.
	 */
	if (likely(!PageLRU(page)))
		return;

1173
	spin_lock_irqsave(&zone->lru_lock, flags);
1174
	lru = page_lru(page);
1175
	/*
1176 1177 1178 1179 1180 1181 1182 1183 1184
	 * The uncharged page could still be registered to the LRU of
	 * the stale pc->mem_cgroup.
	 *
	 * As pc->mem_cgroup is about to get overwritten, the old LRU
	 * accounting needs to be taken care of.  Let root_mem_cgroup
	 * babysit the page until the new memcg is responsible for it.
	 *
	 * The PCG_USED bit is guarded by lock_page() as the page is
	 * swapcache/pagecache.
1185
	 */
1186 1187 1188 1189
	if (PageLRU(page) && PageCgroupAcctLRU(pc) && !PageCgroupUsed(pc)) {
		del_page_from_lru_list(zone, page, lru);
		add_page_to_lru_list(zone, page, lru);
	}
1190
	spin_unlock_irqrestore(&zone->lru_lock, flags);
K
KAMEZAWA Hiroyuki 已提交
1191 1192
}

1193
static void mem_cgroup_lru_add_after_commit(struct page *page)
1194
{
1195
	enum lru_list lru;
1196 1197 1198
	unsigned long flags;
	struct zone *zone = page_zone(page);
	struct page_cgroup *pc = lookup_page_cgroup(page);
1199 1200 1201 1202 1203 1204 1205 1206 1207 1208
	/*
	 * putback:				charge:
	 * SetPageLRU				SetPageCgroupUsed
	 * smp_mb				smp_mb
	 * PageCgroupUsed && add to memcg LRU	PageLRU && add to memcg LRU
	 *
	 * Ensure that one of the two sides adds the page to the memcg
	 * LRU during a race.
	 */
	smp_mb();
1209 1210 1211
	/* taking care of that the page is added to LRU while we commit it */
	if (likely(!PageLRU(page)))
		return;
1212
	spin_lock_irqsave(&zone->lru_lock, flags);
1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225
	lru = page_lru(page);
	/*
	 * If the page is not on the LRU, someone will soon put it
	 * there.  If it is, and also already accounted for on the
	 * memcg-side, it must be on the right lruvec as setting
	 * pc->mem_cgroup and PageCgroupUsed is properly ordered.
	 * Otherwise, root_mem_cgroup has been babysitting the page
	 * during the charge.  Move it to the new memcg now.
	 */
	if (PageLRU(page) && !PageCgroupAcctLRU(pc)) {
		del_page_from_lru_list(zone, page, lru);
		add_page_to_lru_list(zone, page, lru);
	}
1226 1227 1228
	spin_unlock_irqrestore(&zone->lru_lock, flags);
}

1229
/*
1230
 * Checks whether given mem is same or in the root_mem_cgroup's
1231 1232
 * hierarchy subtree
 */
1233 1234
static bool mem_cgroup_same_or_subtree(const struct mem_cgroup *root_memcg,
		struct mem_cgroup *memcg)
1235
{
1236 1237 1238
	if (root_memcg != memcg) {
		return (root_memcg->use_hierarchy &&
			css_is_ancestor(&memcg->css, &root_memcg->css));
1239 1240 1241 1242 1243
	}

	return true;
}

1244
int task_in_mem_cgroup(struct task_struct *task, const struct mem_cgroup *memcg)
1245 1246
{
	int ret;
1247
	struct mem_cgroup *curr = NULL;
1248
	struct task_struct *p;
1249

1250 1251 1252 1253 1254
	p = find_lock_task_mm(task);
	if (!p)
		return 0;
	curr = try_get_mem_cgroup_from_mm(p->mm);
	task_unlock(p);
1255 1256
	if (!curr)
		return 0;
1257
	/*
1258
	 * We should check use_hierarchy of "memcg" not "curr". Because checking
1259
	 * use_hierarchy of "curr" here make this function true if hierarchy is
1260 1261
	 * enabled in "curr" and "curr" is a child of "memcg" in *cgroup*
	 * hierarchy(even if use_hierarchy is disabled in "memcg").
1262
	 */
1263
	ret = mem_cgroup_same_or_subtree(memcg, curr);
1264
	css_put(&curr->css);
1265 1266 1267
	return ret;
}

1268
int mem_cgroup_inactive_anon_is_low(struct mem_cgroup *memcg, struct zone *zone)
1269
{
1270 1271 1272
	unsigned long inactive_ratio;
	int nid = zone_to_nid(zone);
	int zid = zone_idx(zone);
1273
	unsigned long inactive;
1274
	unsigned long active;
1275
	unsigned long gb;
1276

1277 1278 1279 1280
	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));
1281

1282 1283 1284 1285 1286 1287
	gb = (inactive + active) >> (30 - PAGE_SHIFT);
	if (gb)
		inactive_ratio = int_sqrt(10 * gb);
	else
		inactive_ratio = 1;

1288
	return inactive * inactive_ratio < active;
1289 1290
}

1291
int mem_cgroup_inactive_file_is_low(struct mem_cgroup *memcg, struct zone *zone)
1292 1293 1294
{
	unsigned long active;
	unsigned long inactive;
1295 1296
	int zid = zone_idx(zone);
	int nid = zone_to_nid(zone);
1297

1298 1299 1300 1301
	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));
1302 1303 1304 1305

	return (active > inactive);
}

K
KOSAKI Motohiro 已提交
1306 1307 1308
struct zone_reclaim_stat *mem_cgroup_get_reclaim_stat(struct mem_cgroup *memcg,
						      struct zone *zone)
{
1309
	int nid = zone_to_nid(zone);
K
KOSAKI Motohiro 已提交
1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325
	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);
1326 1327
	if (!PageCgroupUsed(pc))
		return NULL;
1328 1329
	/* Ensure pc->mem_cgroup is visible after reading PCG_USED. */
	smp_rmb();
1330
	mz = page_cgroup_zoneinfo(pc->mem_cgroup, page);
K
KOSAKI Motohiro 已提交
1331 1332 1333
	return &mz->reclaim_stat;
}

1334 1335 1336
#define mem_cgroup_from_res_counter(counter, member)	\
	container_of(counter, struct mem_cgroup, member)

1337
/**
1338 1339
 * mem_cgroup_margin - calculate chargeable space of a memory cgroup
 * @mem: the memory cgroup
1340
 *
1341
 * Returns the maximum amount of memory @mem can be charged with, in
1342
 * pages.
1343
 */
1344
static unsigned long mem_cgroup_margin(struct mem_cgroup *memcg)
1345
{
1346 1347
	unsigned long long margin;

1348
	margin = res_counter_margin(&memcg->res);
1349
	if (do_swap_account)
1350
		margin = min(margin, res_counter_margin(&memcg->memsw));
1351
	return margin >> PAGE_SHIFT;
1352 1353
}

1354
int mem_cgroup_swappiness(struct mem_cgroup *memcg)
K
KOSAKI Motohiro 已提交
1355 1356 1357 1358 1359 1360 1361
{
	struct cgroup *cgrp = memcg->css.cgroup;

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

1362
	return memcg->swappiness;
K
KOSAKI Motohiro 已提交
1363 1364
}

1365
static void mem_cgroup_start_move(struct mem_cgroup *memcg)
1366 1367
{
	int cpu;
1368 1369

	get_online_cpus();
1370
	spin_lock(&memcg->pcp_counter_lock);
1371
	for_each_online_cpu(cpu)
1372 1373 1374
		per_cpu(memcg->stat->count[MEM_CGROUP_ON_MOVE], cpu) += 1;
	memcg->nocpu_base.count[MEM_CGROUP_ON_MOVE] += 1;
	spin_unlock(&memcg->pcp_counter_lock);
1375
	put_online_cpus();
1376 1377 1378 1379

	synchronize_rcu();
}

1380
static void mem_cgroup_end_move(struct mem_cgroup *memcg)
1381 1382 1383
{
	int cpu;

1384
	if (!memcg)
1385
		return;
1386
	get_online_cpus();
1387
	spin_lock(&memcg->pcp_counter_lock);
1388
	for_each_online_cpu(cpu)
1389 1390 1391
		per_cpu(memcg->stat->count[MEM_CGROUP_ON_MOVE], cpu) -= 1;
	memcg->nocpu_base.count[MEM_CGROUP_ON_MOVE] -= 1;
	spin_unlock(&memcg->pcp_counter_lock);
1392
	put_online_cpus();
1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405
}
/*
 * 2 routines for checking "mem" is under move_account() or not.
 *
 * mem_cgroup_stealed() - checking a cgroup is mc.from or not. This is used
 *			  for avoiding race in accounting. If true,
 *			  pc->mem_cgroup may be overwritten.
 *
 * mem_cgroup_under_move() - checking a cgroup is mc.from or mc.to or
 *			  under hierarchy of moving cgroups. This is for
 *			  waiting at hith-memory prressure caused by "move".
 */

1406
static bool mem_cgroup_stealed(struct mem_cgroup *memcg)
1407 1408
{
	VM_BUG_ON(!rcu_read_lock_held());
1409
	return this_cpu_read(memcg->stat->count[MEM_CGROUP_ON_MOVE]) > 0;
1410
}
1411

1412
static bool mem_cgroup_under_move(struct mem_cgroup *memcg)
1413
{
1414 1415
	struct mem_cgroup *from;
	struct mem_cgroup *to;
1416
	bool ret = false;
1417 1418 1419 1420 1421 1422 1423 1424 1425
	/*
	 * 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;
1426

1427 1428
	ret = mem_cgroup_same_or_subtree(memcg, from)
		|| mem_cgroup_same_or_subtree(memcg, to);
1429 1430
unlock:
	spin_unlock(&mc.lock);
1431 1432 1433
	return ret;
}

1434
static bool mem_cgroup_wait_acct_move(struct mem_cgroup *memcg)
1435 1436
{
	if (mc.moving_task && current != mc.moving_task) {
1437
		if (mem_cgroup_under_move(memcg)) {
1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448 1449
			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;
}

1450
/**
1451
 * mem_cgroup_print_oom_info: Called from OOM with tasklist_lock held in read mode.
1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469
 * @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;

1470
	if (!memcg || !p)
1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516
		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));
}

1517 1518 1519 1520
/*
 * This function returns the number of memcg under hierarchy tree. Returns
 * 1(self count) if no children.
 */
1521
static int mem_cgroup_count_children(struct mem_cgroup *memcg)
1522 1523
{
	int num = 0;
K
KAMEZAWA Hiroyuki 已提交
1524 1525
	struct mem_cgroup *iter;

1526
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
1527
		num++;
1528 1529 1530
	return num;
}

D
David Rientjes 已提交
1531 1532 1533 1534 1535 1536 1537 1538
/*
 * Return the memory (and swap, if configured) limit for a memcg.
 */
u64 mem_cgroup_get_limit(struct mem_cgroup *memcg)
{
	u64 limit;
	u64 memsw;

1539 1540 1541
	limit = res_counter_read_u64(&memcg->res, RES_LIMIT);
	limit += total_swap_pages << PAGE_SHIFT;

D
David Rientjes 已提交
1542 1543 1544 1545 1546 1547 1548 1549
	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);
}

1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585
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;
}

1586 1587 1588 1589 1590 1591 1592 1593 1594 1595
/**
 * 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.
 */
1596
static bool test_mem_cgroup_node_reclaimable(struct mem_cgroup *memcg,
1597 1598
		int nid, bool noswap)
{
1599
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_FILE))
1600 1601 1602
		return true;
	if (noswap || !total_swap_pages)
		return false;
1603
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_ANON))
1604 1605 1606 1607
		return true;
	return false;

}
1608 1609 1610 1611 1612 1613 1614 1615
#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.
 *
 */
1616
static void mem_cgroup_may_update_nodemask(struct mem_cgroup *memcg)
1617 1618
{
	int nid;
1619 1620 1621 1622
	/*
	 * numainfo_events > 0 means there was at least NUMAINFO_EVENTS_TARGET
	 * pagein/pageout changes since the last update.
	 */
1623
	if (!atomic_read(&memcg->numainfo_events))
1624
		return;
1625
	if (atomic_inc_return(&memcg->numainfo_updating) > 1)
1626 1627 1628
		return;

	/* make a nodemask where this memcg uses memory from */
1629
	memcg->scan_nodes = node_states[N_HIGH_MEMORY];
1630 1631 1632

	for_each_node_mask(nid, node_states[N_HIGH_MEMORY]) {

1633 1634
		if (!test_mem_cgroup_node_reclaimable(memcg, nid, false))
			node_clear(nid, memcg->scan_nodes);
1635
	}
1636

1637 1638
	atomic_set(&memcg->numainfo_events, 0);
	atomic_set(&memcg->numainfo_updating, 0);
1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652
}

/*
 * 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.
 */
1653
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
1654 1655 1656
{
	int node;

1657 1658
	mem_cgroup_may_update_nodemask(memcg);
	node = memcg->last_scanned_node;
1659

1660
	node = next_node(node, memcg->scan_nodes);
1661
	if (node == MAX_NUMNODES)
1662
		node = first_node(memcg->scan_nodes);
1663 1664 1665 1666 1667 1668 1669 1670 1671
	/*
	 * 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();

1672
	memcg->last_scanned_node = node;
1673 1674 1675
	return node;
}

1676 1677 1678 1679 1680 1681
/*
 * 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.
 */
1682
bool mem_cgroup_reclaimable(struct mem_cgroup *memcg, bool noswap)
1683 1684 1685 1686 1687 1688 1689
{
	int nid;

	/*
	 * quick check...making use of scan_node.
	 * We can skip unused nodes.
	 */
1690 1691
	if (!nodes_empty(memcg->scan_nodes)) {
		for (nid = first_node(memcg->scan_nodes);
1692
		     nid < MAX_NUMNODES;
1693
		     nid = next_node(nid, memcg->scan_nodes)) {
1694

1695
			if (test_mem_cgroup_node_reclaimable(memcg, nid, noswap))
1696 1697 1698 1699 1700 1701 1702
				return true;
		}
	}
	/*
	 * Check rest of nodes.
	 */
	for_each_node_state(nid, N_HIGH_MEMORY) {
1703
		if (node_isset(nid, memcg->scan_nodes))
1704
			continue;
1705
		if (test_mem_cgroup_node_reclaimable(memcg, nid, noswap))
1706 1707 1708 1709 1710
			return true;
	}
	return false;
}

1711
#else
1712
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
1713 1714 1715
{
	return 0;
}
1716

1717
bool mem_cgroup_reclaimable(struct mem_cgroup *memcg, bool noswap)
1718
{
1719
	return test_mem_cgroup_node_reclaimable(memcg, 0, noswap);
1720
}
1721 1722
#endif

1723 1724 1725 1726
static int mem_cgroup_soft_reclaim(struct mem_cgroup *root_memcg,
				   struct zone *zone,
				   gfp_t gfp_mask,
				   unsigned long *total_scanned)
1727
{
1728
	struct mem_cgroup *victim = NULL;
1729
	int total = 0;
K
KAMEZAWA Hiroyuki 已提交
1730
	int loop = 0;
1731
	unsigned long excess;
1732
	unsigned long nr_scanned;
1733 1734 1735 1736
	struct mem_cgroup_reclaim_cookie reclaim = {
		.zone = zone,
		.priority = 0,
	};
1737

1738
	excess = res_counter_soft_limit_excess(&root_memcg->res) >> PAGE_SHIFT;
K
KAMEZAWA Hiroyuki 已提交
1739

1740
	while (1) {
1741
		victim = mem_cgroup_iter(root_memcg, victim, &reclaim);
1742
		if (!victim) {
K
KAMEZAWA Hiroyuki 已提交
1743
			loop++;
1744 1745 1746 1747 1748 1749
			if (loop >= 2) {
				/*
				 * If we have not been able to reclaim
				 * anything, it might because there are
				 * no reclaimable pages under this hierarchy
				 */
1750
				if (!total)
1751 1752
					break;
				/*
L
Lucas De Marchi 已提交
1753
				 * We want to do more targeted reclaim.
1754 1755 1756 1757 1758
				 * 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) ||
1759
					(loop > MEM_CGROUP_MAX_RECLAIM_LOOPS))
1760 1761
					break;
			}
1762
			continue;
1763
		}
1764
		if (!mem_cgroup_reclaimable(victim, false))
1765
			continue;
1766 1767 1768 1769
		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))
1770
			break;
1771
	}
1772
	mem_cgroup_iter_break(root_memcg, victim);
K
KAMEZAWA Hiroyuki 已提交
1773
	return total;
1774 1775
}

K
KAMEZAWA Hiroyuki 已提交
1776 1777 1778
/*
 * Check OOM-Killer is already running under our hierarchy.
 * If someone is running, return false.
1779
 * Has to be called with memcg_oom_lock
K
KAMEZAWA Hiroyuki 已提交
1780
 */
1781
static bool mem_cgroup_oom_lock(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
1782
{
1783
	struct mem_cgroup *iter, *failed = NULL;
1784

1785
	for_each_mem_cgroup_tree(iter, memcg) {
1786
		if (iter->oom_lock) {
1787 1788 1789 1790 1791
			/*
			 * this subtree of our hierarchy is already locked
			 * so we cannot give a lock.
			 */
			failed = iter;
1792 1793
			mem_cgroup_iter_break(memcg, iter);
			break;
1794 1795
		} else
			iter->oom_lock = true;
K
KAMEZAWA Hiroyuki 已提交
1796
	}
K
KAMEZAWA Hiroyuki 已提交
1797

1798
	if (!failed)
1799
		return true;
1800 1801 1802 1803 1804

	/*
	 * OK, we failed to lock the whole subtree so we have to clean up
	 * what we set up to the failing subtree
	 */
1805
	for_each_mem_cgroup_tree(iter, memcg) {
1806
		if (iter == failed) {
1807 1808
			mem_cgroup_iter_break(memcg, iter);
			break;
1809 1810 1811
		}
		iter->oom_lock = false;
	}
1812
	return false;
1813
}
1814

1815
/*
1816
 * Has to be called with memcg_oom_lock
1817
 */
1818
static int mem_cgroup_oom_unlock(struct mem_cgroup *memcg)
1819
{
K
KAMEZAWA Hiroyuki 已提交
1820 1821
	struct mem_cgroup *iter;

1822
	for_each_mem_cgroup_tree(iter, memcg)
1823 1824 1825 1826
		iter->oom_lock = false;
	return 0;
}

1827
static void mem_cgroup_mark_under_oom(struct mem_cgroup *memcg)
1828 1829 1830
{
	struct mem_cgroup *iter;

1831
	for_each_mem_cgroup_tree(iter, memcg)
1832 1833 1834
		atomic_inc(&iter->under_oom);
}

1835
static void mem_cgroup_unmark_under_oom(struct mem_cgroup *memcg)
1836 1837 1838
{
	struct mem_cgroup *iter;

K
KAMEZAWA Hiroyuki 已提交
1839 1840 1841 1842 1843
	/*
	 * 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.
	 */
1844
	for_each_mem_cgroup_tree(iter, memcg)
1845
		atomic_add_unless(&iter->under_oom, -1, 0);
1846 1847
}

1848
static DEFINE_SPINLOCK(memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
1849 1850
static DECLARE_WAIT_QUEUE_HEAD(memcg_oom_waitq);

K
KAMEZAWA Hiroyuki 已提交
1851 1852 1853 1854 1855 1856 1857 1858
struct oom_wait_info {
	struct mem_cgroup *mem;
	wait_queue_t	wait;
};

static int memcg_oom_wake_function(wait_queue_t *wait,
	unsigned mode, int sync, void *arg)
{
1859 1860
	struct mem_cgroup *wake_memcg = (struct mem_cgroup *)arg,
			  *oom_wait_memcg;
K
KAMEZAWA Hiroyuki 已提交
1861 1862 1863
	struct oom_wait_info *oom_wait_info;

	oom_wait_info = container_of(wait, struct oom_wait_info, wait);
1864
	oom_wait_memcg = oom_wait_info->mem;
K
KAMEZAWA Hiroyuki 已提交
1865 1866 1867 1868 1869

	/*
	 * Both of oom_wait_info->mem and wake_mem are stable under us.
	 * Then we can use css_is_ancestor without taking care of RCU.
	 */
1870 1871
	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
1872 1873 1874 1875
		return 0;
	return autoremove_wake_function(wait, mode, sync, arg);
}

1876
static void memcg_wakeup_oom(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
1877
{
1878 1879
	/* for filtering, pass "memcg" as argument. */
	__wake_up(&memcg_oom_waitq, TASK_NORMAL, 0, memcg);
K
KAMEZAWA Hiroyuki 已提交
1880 1881
}

1882
static void memcg_oom_recover(struct mem_cgroup *memcg)
1883
{
1884 1885
	if (memcg && atomic_read(&memcg->under_oom))
		memcg_wakeup_oom(memcg);
1886 1887
}

K
KAMEZAWA Hiroyuki 已提交
1888 1889 1890
/*
 * try to call OOM killer. returns false if we should exit memory-reclaim loop.
 */
1891
bool mem_cgroup_handle_oom(struct mem_cgroup *memcg, gfp_t mask)
1892
{
K
KAMEZAWA Hiroyuki 已提交
1893
	struct oom_wait_info owait;
1894
	bool locked, need_to_kill;
K
KAMEZAWA Hiroyuki 已提交
1895

1896
	owait.mem = memcg;
K
KAMEZAWA Hiroyuki 已提交
1897 1898 1899 1900
	owait.wait.flags = 0;
	owait.wait.func = memcg_oom_wake_function;
	owait.wait.private = current;
	INIT_LIST_HEAD(&owait.wait.task_list);
1901
	need_to_kill = true;
1902
	mem_cgroup_mark_under_oom(memcg);
1903

1904
	/* At first, try to OOM lock hierarchy under memcg.*/
1905
	spin_lock(&memcg_oom_lock);
1906
	locked = mem_cgroup_oom_lock(memcg);
K
KAMEZAWA Hiroyuki 已提交
1907 1908 1909 1910 1911
	/*
	 * 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.
	 */
1912
	prepare_to_wait(&memcg_oom_waitq, &owait.wait, TASK_KILLABLE);
1913
	if (!locked || memcg->oom_kill_disable)
1914 1915
		need_to_kill = false;
	if (locked)
1916
		mem_cgroup_oom_notify(memcg);
1917
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
1918

1919 1920
	if (need_to_kill) {
		finish_wait(&memcg_oom_waitq, &owait.wait);
1921
		mem_cgroup_out_of_memory(memcg, mask);
1922
	} else {
K
KAMEZAWA Hiroyuki 已提交
1923
		schedule();
K
KAMEZAWA Hiroyuki 已提交
1924
		finish_wait(&memcg_oom_waitq, &owait.wait);
K
KAMEZAWA Hiroyuki 已提交
1925
	}
1926
	spin_lock(&memcg_oom_lock);
1927
	if (locked)
1928 1929
		mem_cgroup_oom_unlock(memcg);
	memcg_wakeup_oom(memcg);
1930
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
1931

1932
	mem_cgroup_unmark_under_oom(memcg);
1933

K
KAMEZAWA Hiroyuki 已提交
1934 1935 1936
	if (test_thread_flag(TIF_MEMDIE) || fatal_signal_pending(current))
		return false;
	/* Give chance to dying process */
1937
	schedule_timeout_uninterruptible(1);
K
KAMEZAWA Hiroyuki 已提交
1938
	return true;
1939 1940
}

1941 1942 1943
/*
 * Currently used to update mapped file statistics, but the routine can be
 * generalized to update other statistics as well.
1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962
 *
 * Notes: Race condition
 *
 * We usually use page_cgroup_lock() for accessing page_cgroup member but
 * it tends to be costly. But considering some conditions, we doesn't need
 * to do so _always_.
 *
 * Considering "charge", lock_page_cgroup() is not required because all
 * file-stat operations happen after a page is attached to radix-tree. There
 * are no race with "charge".
 *
 * Considering "uncharge", we know that memcg doesn't clear pc->mem_cgroup
 * at "uncharge" intentionally. So, we always see valid pc->mem_cgroup even
 * if there are race with "uncharge". Statistics itself is properly handled
 * by flags.
 *
 * Considering "move", this is an only case we see a race. To make the race
 * small, we check MEM_CGROUP_ON_MOVE percpu value and detect there are
 * possibility of race condition. If there is, we take a lock.
1963
 */
1964

1965 1966
void mem_cgroup_update_page_stat(struct page *page,
				 enum mem_cgroup_page_stat_item idx, int val)
1967
{
1968
	struct mem_cgroup *memcg;
1969 1970
	struct page_cgroup *pc = lookup_page_cgroup(page);
	bool need_unlock = false;
1971
	unsigned long uninitialized_var(flags);
1972 1973 1974 1975

	if (unlikely(!pc))
		return;

1976
	rcu_read_lock();
1977 1978
	memcg = pc->mem_cgroup;
	if (unlikely(!memcg || !PageCgroupUsed(pc)))
1979 1980
		goto out;
	/* pc->mem_cgroup is unstable ? */
1981
	if (unlikely(mem_cgroup_stealed(memcg)) || PageTransHuge(page)) {
1982
		/* take a lock against to access pc->mem_cgroup */
1983
		move_lock_page_cgroup(pc, &flags);
1984
		need_unlock = true;
1985 1986
		memcg = pc->mem_cgroup;
		if (!memcg || !PageCgroupUsed(pc))
1987 1988
			goto out;
	}
1989 1990

	switch (idx) {
1991
	case MEMCG_NR_FILE_MAPPED:
1992 1993 1994
		if (val > 0)
			SetPageCgroupFileMapped(pc);
		else if (!page_mapped(page))
1995
			ClearPageCgroupFileMapped(pc);
1996
		idx = MEM_CGROUP_STAT_FILE_MAPPED;
1997 1998 1999
		break;
	default:
		BUG();
2000
	}
2001

2002
	this_cpu_add(memcg->stat->count[idx], val);
2003

2004 2005
out:
	if (unlikely(need_unlock))
2006
		move_unlock_page_cgroup(pc, &flags);
2007 2008
	rcu_read_unlock();
	return;
2009
}
2010
EXPORT_SYMBOL(mem_cgroup_update_page_stat);
2011

2012 2013 2014 2015
/*
 * size of first charge trial. "32" comes from vmscan.c's magic value.
 * TODO: maybe necessary to use big numbers in big irons.
 */
2016
#define CHARGE_BATCH	32U
2017 2018
struct memcg_stock_pcp {
	struct mem_cgroup *cached; /* this never be root cgroup */
2019
	unsigned int nr_pages;
2020
	struct work_struct work;
2021 2022
	unsigned long flags;
#define FLUSHING_CACHED_CHARGE	(0)
2023 2024
};
static DEFINE_PER_CPU(struct memcg_stock_pcp, memcg_stock);
2025
static DEFINE_MUTEX(percpu_charge_mutex);
2026 2027

/*
2028
 * Try to consume stocked charge on this cpu. If success, one page is consumed
2029 2030 2031 2032
 * 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.
 */
2033
static bool consume_stock(struct mem_cgroup *memcg)
2034 2035 2036 2037 2038
{
	struct memcg_stock_pcp *stock;
	bool ret = true;

	stock = &get_cpu_var(memcg_stock);
2039
	if (memcg == stock->cached && stock->nr_pages)
2040
		stock->nr_pages--;
2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053
	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;

2054 2055 2056 2057
	if (stock->nr_pages) {
		unsigned long bytes = stock->nr_pages * PAGE_SIZE;

		res_counter_uncharge(&old->res, bytes);
2058
		if (do_swap_account)
2059 2060
			res_counter_uncharge(&old->memsw, bytes);
		stock->nr_pages = 0;
2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072
	}
	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);
2073
	clear_bit(FLUSHING_CACHED_CHARGE, &stock->flags);
2074 2075 2076 2077
}

/*
 * Cache charges(val) which is from res_counter, to local per_cpu area.
2078
 * This will be consumed by consume_stock() function, later.
2079
 */
2080
static void refill_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
2081 2082 2083
{
	struct memcg_stock_pcp *stock = &get_cpu_var(memcg_stock);

2084
	if (stock->cached != memcg) { /* reset if necessary */
2085
		drain_stock(stock);
2086
		stock->cached = memcg;
2087
	}
2088
	stock->nr_pages += nr_pages;
2089 2090 2091 2092
	put_cpu_var(memcg_stock);
}

/*
2093
 * Drains all per-CPU charge caches for given root_memcg resp. subtree
2094 2095
 * of the hierarchy under it. sync flag says whether we should block
 * until the work is done.
2096
 */
2097
static void drain_all_stock(struct mem_cgroup *root_memcg, bool sync)
2098
{
2099
	int cpu, curcpu;
2100

2101 2102
	/* Notify other cpus that system-wide "drain" is running */
	get_online_cpus();
2103
	curcpu = get_cpu();
2104 2105
	for_each_online_cpu(cpu) {
		struct memcg_stock_pcp *stock = &per_cpu(memcg_stock, cpu);
2106
		struct mem_cgroup *memcg;
2107

2108 2109
		memcg = stock->cached;
		if (!memcg || !stock->nr_pages)
2110
			continue;
2111
		if (!mem_cgroup_same_or_subtree(root_memcg, memcg))
2112
			continue;
2113 2114 2115 2116 2117 2118
		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);
		}
2119
	}
2120
	put_cpu();
2121 2122 2123 2124 2125 2126

	if (!sync)
		goto out;

	for_each_online_cpu(cpu) {
		struct memcg_stock_pcp *stock = &per_cpu(memcg_stock, cpu);
2127
		if (test_bit(FLUSHING_CACHED_CHARGE, &stock->flags))
2128 2129 2130
			flush_work(&stock->work);
	}
out:
2131
 	put_online_cpus();
2132 2133 2134 2135 2136 2137 2138 2139
}

/*
 * 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.
 */
2140
static void drain_all_stock_async(struct mem_cgroup *root_memcg)
2141
{
2142 2143 2144 2145 2146
	/*
	 * If someone calls draining, avoid adding more kworker runs.
	 */
	if (!mutex_trylock(&percpu_charge_mutex))
		return;
2147
	drain_all_stock(root_memcg, false);
2148
	mutex_unlock(&percpu_charge_mutex);
2149 2150 2151
}

/* This is a synchronous drain interface. */
2152
static void drain_all_stock_sync(struct mem_cgroup *root_memcg)
2153 2154
{
	/* called when force_empty is called */
2155
	mutex_lock(&percpu_charge_mutex);
2156
	drain_all_stock(root_memcg, true);
2157
	mutex_unlock(&percpu_charge_mutex);
2158 2159
}

2160 2161 2162 2163
/*
 * This function drains percpu counter value from DEAD cpu and
 * move it to local cpu. Note that this function can be preempted.
 */
2164
static void mem_cgroup_drain_pcp_counter(struct mem_cgroup *memcg, int cpu)
2165 2166 2167
{
	int i;

2168
	spin_lock(&memcg->pcp_counter_lock);
2169
	for (i = 0; i < MEM_CGROUP_STAT_DATA; i++) {
2170
		long x = per_cpu(memcg->stat->count[i], cpu);
2171

2172 2173
		per_cpu(memcg->stat->count[i], cpu) = 0;
		memcg->nocpu_base.count[i] += x;
2174
	}
2175
	for (i = 0; i < MEM_CGROUP_EVENTS_NSTATS; i++) {
2176
		unsigned long x = per_cpu(memcg->stat->events[i], cpu);
2177

2178 2179
		per_cpu(memcg->stat->events[i], cpu) = 0;
		memcg->nocpu_base.events[i] += x;
2180
	}
2181
	/* need to clear ON_MOVE value, works as a kind of lock. */
2182 2183
	per_cpu(memcg->stat->count[MEM_CGROUP_ON_MOVE], cpu) = 0;
	spin_unlock(&memcg->pcp_counter_lock);
2184 2185
}

2186
static void synchronize_mem_cgroup_on_move(struct mem_cgroup *memcg, int cpu)
2187 2188 2189
{
	int idx = MEM_CGROUP_ON_MOVE;

2190 2191 2192
	spin_lock(&memcg->pcp_counter_lock);
	per_cpu(memcg->stat->count[idx], cpu) = memcg->nocpu_base.count[idx];
	spin_unlock(&memcg->pcp_counter_lock);
2193 2194 2195
}

static int __cpuinit memcg_cpu_hotplug_callback(struct notifier_block *nb,
2196 2197 2198 2199 2200
					unsigned long action,
					void *hcpu)
{
	int cpu = (unsigned long)hcpu;
	struct memcg_stock_pcp *stock;
2201
	struct mem_cgroup *iter;
2202

2203
	if ((action == CPU_ONLINE)) {
2204
		for_each_mem_cgroup(iter)
2205 2206 2207 2208
			synchronize_mem_cgroup_on_move(iter, cpu);
		return NOTIFY_OK;
	}

2209
	if ((action != CPU_DEAD) || action != CPU_DEAD_FROZEN)
2210
		return NOTIFY_OK;
2211

2212
	for_each_mem_cgroup(iter)
2213 2214
		mem_cgroup_drain_pcp_counter(iter, cpu);

2215 2216 2217 2218 2219
	stock = &per_cpu(memcg_stock, cpu);
	drain_stock(stock);
	return NOTIFY_OK;
}

2220 2221 2222 2223 2224 2225 2226 2227 2228 2229

/* 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 */
};

2230
static int mem_cgroup_do_charge(struct mem_cgroup *memcg, gfp_t gfp_mask,
2231
				unsigned int nr_pages, bool oom_check)
2232
{
2233
	unsigned long csize = nr_pages * PAGE_SIZE;
2234 2235 2236 2237 2238
	struct mem_cgroup *mem_over_limit;
	struct res_counter *fail_res;
	unsigned long flags = 0;
	int ret;

2239
	ret = res_counter_charge(&memcg->res, csize, &fail_res);
2240 2241 2242 2243

	if (likely(!ret)) {
		if (!do_swap_account)
			return CHARGE_OK;
2244
		ret = res_counter_charge(&memcg->memsw, csize, &fail_res);
2245 2246 2247
		if (likely(!ret))
			return CHARGE_OK;

2248
		res_counter_uncharge(&memcg->res, csize);
2249 2250 2251 2252
		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);
2253
	/*
2254 2255
	 * nr_pages can be either a huge page (HPAGE_PMD_NR), a batch
	 * of regular pages (CHARGE_BATCH), or a single regular page (1).
2256 2257 2258 2259
	 *
	 * Never reclaim on behalf of optional batching, retry with a
	 * single page instead.
	 */
2260
	if (nr_pages == CHARGE_BATCH)
2261 2262 2263 2264 2265
		return CHARGE_RETRY;

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

2266
	ret = mem_cgroup_reclaim(mem_over_limit, gfp_mask, flags);
2267
	if (mem_cgroup_margin(mem_over_limit) >= nr_pages)
2268
		return CHARGE_RETRY;
2269
	/*
2270 2271 2272 2273 2274 2275 2276
	 * 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.
2277
	 */
2278
	if (nr_pages == 1 && ret)
2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297
		return CHARGE_RETRY;

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

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

	return CHARGE_RETRY;
}

2298 2299 2300
/*
 * Unlike exported interface, "oom" parameter is added. if oom==true,
 * oom-killer can be invoked.
2301
 */
2302
static int __mem_cgroup_try_charge(struct mm_struct *mm,
A
Andrea Arcangeli 已提交
2303
				   gfp_t gfp_mask,
2304
				   unsigned int nr_pages,
2305
				   struct mem_cgroup **ptr,
2306
				   bool oom)
2307
{
2308
	unsigned int batch = max(CHARGE_BATCH, nr_pages);
2309
	int nr_oom_retries = MEM_CGROUP_RECLAIM_RETRIES;
2310
	struct mem_cgroup *memcg = NULL;
2311
	int ret;
2312

K
KAMEZAWA Hiroyuki 已提交
2313 2314 2315 2316 2317 2318 2319 2320
	/*
	 * 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;
2321

2322
	/*
2323 2324
	 * We always charge the cgroup the mm_struct belongs to.
	 * The mm_struct's mem_cgroup changes on task migration if the
2325 2326 2327
	 * thread group leader migrates. It's possible that mm is not
	 * set, if so charge the init_mm (happens for pagecache usage).
	 */
2328
	if (!*ptr && !mm)
K
KAMEZAWA Hiroyuki 已提交
2329 2330
		goto bypass;
again:
2331 2332 2333 2334
	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
2335
			goto done;
2336
		if (nr_pages == 1 && consume_stock(memcg))
K
KAMEZAWA Hiroyuki 已提交
2337
			goto done;
2338
		css_get(&memcg->css);
2339
	} else {
K
KAMEZAWA Hiroyuki 已提交
2340
		struct task_struct *p;
2341

K
KAMEZAWA Hiroyuki 已提交
2342 2343 2344
		rcu_read_lock();
		p = rcu_dereference(mm->owner);
		/*
2345
		 * Because we don't have task_lock(), "p" can exit.
2346
		 * In that case, "memcg" can point to root or p can be NULL with
2347 2348 2349 2350 2351 2352
		 * 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
2353
		 */
2354 2355
		memcg = mem_cgroup_from_task(p);
		if (!memcg || mem_cgroup_is_root(memcg)) {
K
KAMEZAWA Hiroyuki 已提交
2356 2357 2358
			rcu_read_unlock();
			goto done;
		}
2359
		if (nr_pages == 1 && consume_stock(memcg)) {
K
KAMEZAWA Hiroyuki 已提交
2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371
			/*
			 * 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 */
2372
		if (!css_tryget(&memcg->css)) {
K
KAMEZAWA Hiroyuki 已提交
2373 2374 2375 2376 2377
			rcu_read_unlock();
			goto again;
		}
		rcu_read_unlock();
	}
2378

2379 2380
	do {
		bool oom_check;
2381

2382
		/* If killed, bypass charge */
K
KAMEZAWA Hiroyuki 已提交
2383
		if (fatal_signal_pending(current)) {
2384
			css_put(&memcg->css);
2385
			goto bypass;
K
KAMEZAWA Hiroyuki 已提交
2386
		}
2387

2388 2389 2390 2391
		oom_check = false;
		if (oom && !nr_oom_retries) {
			oom_check = true;
			nr_oom_retries = MEM_CGROUP_RECLAIM_RETRIES;
2392
		}
2393

2394
		ret = mem_cgroup_do_charge(memcg, gfp_mask, batch, oom_check);
2395 2396 2397 2398
		switch (ret) {
		case CHARGE_OK:
			break;
		case CHARGE_RETRY: /* not in OOM situation but retry */
2399
			batch = nr_pages;
2400 2401
			css_put(&memcg->css);
			memcg = NULL;
K
KAMEZAWA Hiroyuki 已提交
2402
			goto again;
2403
		case CHARGE_WOULDBLOCK: /* !__GFP_WAIT */
2404
			css_put(&memcg->css);
2405 2406
			goto nomem;
		case CHARGE_NOMEM: /* OOM routine works */
K
KAMEZAWA Hiroyuki 已提交
2407
			if (!oom) {
2408
				css_put(&memcg->css);
K
KAMEZAWA Hiroyuki 已提交
2409
				goto nomem;
K
KAMEZAWA Hiroyuki 已提交
2410
			}
2411 2412 2413 2414
			/* If oom, we never return -ENOMEM */
			nr_oom_retries--;
			break;
		case CHARGE_OOM_DIE: /* Killed by OOM Killer */
2415
			css_put(&memcg->css);
K
KAMEZAWA Hiroyuki 已提交
2416
			goto bypass;
2417
		}
2418 2419
	} while (ret != CHARGE_OK);

2420
	if (batch > nr_pages)
2421 2422
		refill_stock(memcg, batch - nr_pages);
	css_put(&memcg->css);
2423
done:
2424
	*ptr = memcg;
2425 2426
	return 0;
nomem:
2427
	*ptr = NULL;
2428
	return -ENOMEM;
K
KAMEZAWA Hiroyuki 已提交
2429
bypass:
2430
	*ptr = NULL;
K
KAMEZAWA Hiroyuki 已提交
2431
	return 0;
2432
}
2433

2434 2435 2436 2437 2438
/*
 * 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().
 */
2439
static void __mem_cgroup_cancel_charge(struct mem_cgroup *memcg,
2440
				       unsigned int nr_pages)
2441
{
2442
	if (!mem_cgroup_is_root(memcg)) {
2443 2444
		unsigned long bytes = nr_pages * PAGE_SIZE;

2445
		res_counter_uncharge(&memcg->res, bytes);
2446
		if (do_swap_account)
2447
			res_counter_uncharge(&memcg->memsw, bytes);
2448
	}
2449 2450
}

2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469
/*
 * 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);
}

2470
struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page)
2471
{
2472
	struct mem_cgroup *memcg = NULL;
2473
	struct page_cgroup *pc;
2474
	unsigned short id;
2475 2476
	swp_entry_t ent;

2477 2478 2479
	VM_BUG_ON(!PageLocked(page));

	pc = lookup_page_cgroup(page);
2480
	lock_page_cgroup(pc);
2481
	if (PageCgroupUsed(pc)) {
2482 2483 2484
		memcg = pc->mem_cgroup;
		if (memcg && !css_tryget(&memcg->css))
			memcg = NULL;
2485
	} else if (PageSwapCache(page)) {
2486
		ent.val = page_private(page);
2487 2488
		id = lookup_swap_cgroup(ent);
		rcu_read_lock();
2489 2490 2491
		memcg = mem_cgroup_lookup(id);
		if (memcg && !css_tryget(&memcg->css))
			memcg = NULL;
2492
		rcu_read_unlock();
2493
	}
2494
	unlock_page_cgroup(pc);
2495
	return memcg;
2496 2497
}

2498
static void __mem_cgroup_commit_charge(struct mem_cgroup *memcg,
2499
				       struct page *page,
2500
				       unsigned int nr_pages,
2501
				       struct page_cgroup *pc,
2502
				       enum charge_type ctype)
2503
{
2504 2505 2506
	lock_page_cgroup(pc);
	if (unlikely(PageCgroupUsed(pc))) {
		unlock_page_cgroup(pc);
2507
		__mem_cgroup_cancel_charge(memcg, nr_pages);
2508 2509 2510 2511 2512 2513
		return;
	}
	/*
	 * we don't need page_cgroup_lock about tail pages, becase they are not
	 * accessed by any other context at this point.
	 */
2514
	pc->mem_cgroup = memcg;
2515 2516 2517 2518 2519 2520 2521
	/*
	 * 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
2522
	smp_wmb();
2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535
	switch (ctype) {
	case MEM_CGROUP_CHARGE_TYPE_CACHE:
	case MEM_CGROUP_CHARGE_TYPE_SHMEM:
		SetPageCgroupCache(pc);
		SetPageCgroupUsed(pc);
		break;
	case MEM_CGROUP_CHARGE_TYPE_MAPPED:
		ClearPageCgroupCache(pc);
		SetPageCgroupUsed(pc);
		break;
	default:
		break;
	}
2536

2537
	mem_cgroup_charge_statistics(memcg, PageCgroupCache(pc), nr_pages);
2538
	unlock_page_cgroup(pc);
2539 2540 2541 2542 2543
	/*
	 * "charge_statistics" updated event counter. Then, check it.
	 * Insert ancestor (and ancestor's ancestors), to softlimit RB-tree.
	 * if they exceeds softlimit.
	 */
2544
	memcg_check_events(memcg, page);
2545
}
2546

2547 2548 2549 2550 2551 2552
#ifdef CONFIG_TRANSPARENT_HUGEPAGE

#define PCGF_NOCOPY_AT_SPLIT ((1 << PCG_LOCK) | (1 << PCG_MOVE_LOCK) |\
			(1 << PCG_ACCT_LRU) | (1 << PCG_MIGRATION))
/*
 * Because tail pages are not marked as "used", set it. We're under
2553 2554 2555
 * 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.
2556
 */
2557
void mem_cgroup_split_huge_fixup(struct page *head)
2558 2559
{
	struct page_cgroup *head_pc = lookup_page_cgroup(head);
2560 2561
	struct page_cgroup *pc;
	int i;
2562

2563 2564
	if (mem_cgroup_disabled())
		return;
2565 2566 2567 2568 2569 2570 2571 2572 2573 2574
	for (i = 1; i < HPAGE_PMD_NR; i++) {
		pc = head_pc + i;
		pc->mem_cgroup = head_pc->mem_cgroup;
		smp_wmb();/* see __commit_charge() */
		/*
		 * LRU flags cannot be copied because we need to add tail
		 * page to LRU by generic call and our hooks will be called.
		 */
		pc->flags = head_pc->flags & ~PCGF_NOCOPY_AT_SPLIT;
	}
2575

2576 2577 2578 2579 2580 2581 2582
	if (PageCgroupAcctLRU(head_pc)) {
		enum lru_list lru;
		struct mem_cgroup_per_zone *mz;
		/*
		 * We hold lru_lock, then, reduce counter directly.
		 */
		lru = page_lru(head);
2583
		mz = page_cgroup_zoneinfo(head_pc->mem_cgroup, head);
2584
		MEM_CGROUP_ZSTAT(mz, lru) -= HPAGE_PMD_NR - 1;
2585
	}
2586 2587 2588
}
#endif

2589
/**
2590
 * mem_cgroup_move_account - move account of the page
2591
 * @page: the page
2592
 * @nr_pages: number of regular pages (>1 for huge pages)
2593 2594 2595
 * @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.
2596
 * @uncharge: whether we should call uncharge and css_put against @from.
2597 2598
 *
 * The caller must confirm following.
K
KAMEZAWA Hiroyuki 已提交
2599
 * - page is not on LRU (isolate_page() is useful.)
2600
 * - compound_lock is held when nr_pages > 1
2601
 *
2602
 * This function doesn't do "charge" nor css_get to new cgroup. It should be
L
Lucas De Marchi 已提交
2603
 * done by a caller(__mem_cgroup_try_charge would be useful). If @uncharge is
2604 2605
 * true, this function does "uncharge" from old cgroup, but it doesn't if
 * @uncharge is false, so a caller should do "uncharge".
2606
 */
2607 2608 2609 2610 2611 2612
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)
2613
{
2614 2615
	unsigned long flags;
	int ret;
2616

2617
	VM_BUG_ON(from == to);
2618
	VM_BUG_ON(PageLRU(page));
2619 2620 2621 2622 2623 2624 2625
	/*
	 * 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;
2626
	if (nr_pages > 1 && !PageTransHuge(page))
2627 2628 2629 2630 2631 2632 2633 2634 2635
		goto out;

	lock_page_cgroup(pc);

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

	move_lock_page_cgroup(pc, &flags);
2636

2637
	if (PageCgroupFileMapped(pc)) {
2638 2639 2640 2641 2642
		/* 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();
2643
	}
2644
	mem_cgroup_charge_statistics(from, PageCgroupCache(pc), -nr_pages);
2645 2646
	if (uncharge)
		/* This is not "cancel", but cancel_charge does all we need. */
2647
		__mem_cgroup_cancel_charge(from, nr_pages);
2648

2649
	/* caller should have done css_get */
K
KAMEZAWA Hiroyuki 已提交
2650
	pc->mem_cgroup = to;
2651
	mem_cgroup_charge_statistics(to, PageCgroupCache(pc), nr_pages);
2652 2653 2654
	/*
	 * We charges against "to" which may not have any tasks. Then, "to"
	 * can be under rmdir(). But in current implementation, caller of
2655
	 * this function is just force_empty() and move charge, so it's
L
Lucas De Marchi 已提交
2656
	 * guaranteed that "to" is never removed. So, we don't check rmdir
2657
	 * status here.
2658
	 */
2659 2660 2661
	move_unlock_page_cgroup(pc, &flags);
	ret = 0;
unlock:
2662
	unlock_page_cgroup(pc);
2663 2664 2665
	/*
	 * check events
	 */
2666 2667
	memcg_check_events(to, page);
	memcg_check_events(from, page);
2668
out:
2669 2670 2671 2672 2673 2674 2675
	return ret;
}

/*
 * move charges to its parent.
 */

2676 2677
static int mem_cgroup_move_parent(struct page *page,
				  struct page_cgroup *pc,
2678 2679 2680 2681 2682 2683
				  struct mem_cgroup *child,
				  gfp_t gfp_mask)
{
	struct cgroup *cg = child->css.cgroup;
	struct cgroup *pcg = cg->parent;
	struct mem_cgroup *parent;
2684
	unsigned int nr_pages;
2685
	unsigned long uninitialized_var(flags);
2686 2687 2688 2689 2690 2691
	int ret;

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

2692 2693 2694 2695 2696
	ret = -EBUSY;
	if (!get_page_unless_zero(page))
		goto out;
	if (isolate_lru_page(page))
		goto put;
2697

2698
	nr_pages = hpage_nr_pages(page);
K
KAMEZAWA Hiroyuki 已提交
2699

2700
	parent = mem_cgroup_from_cont(pcg);
2701
	ret = __mem_cgroup_try_charge(NULL, gfp_mask, nr_pages, &parent, false);
2702
	if (ret || !parent)
2703
		goto put_back;
2704

2705
	if (nr_pages > 1)
2706 2707
		flags = compound_lock_irqsave(page);

2708
	ret = mem_cgroup_move_account(page, nr_pages, pc, child, parent, true);
2709
	if (ret)
2710
		__mem_cgroup_cancel_charge(parent, nr_pages);
2711

2712
	if (nr_pages > 1)
2713
		compound_unlock_irqrestore(page, flags);
2714
put_back:
K
KAMEZAWA Hiroyuki 已提交
2715
	putback_lru_page(page);
2716
put:
2717
	put_page(page);
2718
out:
2719 2720 2721
	return ret;
}

2722 2723 2724 2725 2726 2727 2728
/*
 * 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,
2729
				gfp_t gfp_mask, enum charge_type ctype)
2730
{
2731
	struct mem_cgroup *memcg = NULL;
2732
	unsigned int nr_pages = 1;
2733
	struct page_cgroup *pc;
2734
	bool oom = true;
2735
	int ret;
A
Andrea Arcangeli 已提交
2736

A
Andrea Arcangeli 已提交
2737
	if (PageTransHuge(page)) {
2738
		nr_pages <<= compound_order(page);
A
Andrea Arcangeli 已提交
2739
		VM_BUG_ON(!PageTransHuge(page));
2740 2741 2742 2743 2744
		/*
		 * 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
2745
	}
2746 2747

	pc = lookup_page_cgroup(page);
2748
	BUG_ON(!pc); /* XXX: remove this and move pc lookup into commit */
2749

2750 2751
	ret = __mem_cgroup_try_charge(mm, gfp_mask, nr_pages, &memcg, oom);
	if (ret || !memcg)
2752 2753
		return ret;

2754
	__mem_cgroup_commit_charge(memcg, page, nr_pages, pc, ctype);
2755 2756 2757
	return 0;
}

2758 2759
int mem_cgroup_newpage_charge(struct page *page,
			      struct mm_struct *mm, gfp_t gfp_mask)
2760
{
2761
	if (mem_cgroup_disabled())
2762
		return 0;
2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773
	/*
	 * If already mapped, we don't have to account.
	 * If page cache, page->mapping has address_space.
	 * But page->mapping may have out-of-use anon_vma pointer,
	 * detecit it by PageAnon() check. newly-mapped-anon's page->mapping
	 * is NULL.
  	 */
	if (page_mapped(page) || (page->mapping && !PageAnon(page)))
		return 0;
	if (unlikely(!mm))
		mm = &init_mm;
2774
	return mem_cgroup_charge_common(page, mm, gfp_mask,
2775
				MEM_CGROUP_CHARGE_TYPE_MAPPED);
2776 2777
}

D
Daisuke Nishimura 已提交
2778 2779 2780 2781
static void
__mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr,
					enum charge_type ctype);

2782
static void
2783
__mem_cgroup_commit_charge_lrucare(struct page *page, struct mem_cgroup *memcg,
2784 2785 2786 2787 2788 2789 2790 2791 2792
					enum charge_type ctype)
{
	struct page_cgroup *pc = lookup_page_cgroup(page);
	/*
	 * In some case, SwapCache, FUSE(splice_buf->radixtree), the page
	 * is already on LRU. It means the page may on some other page_cgroup's
	 * LRU. Take care of it.
	 */
	mem_cgroup_lru_del_before_commit(page);
2793
	__mem_cgroup_commit_charge(memcg, page, 1, pc, ctype);
2794 2795 2796 2797
	mem_cgroup_lru_add_after_commit(page);
	return;
}

2798 2799
int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm,
				gfp_t gfp_mask)
2800
{
2801
	struct mem_cgroup *memcg = NULL;
2802 2803
	int ret;

2804
	if (mem_cgroup_disabled())
2805
		return 0;
2806 2807
	if (PageCompound(page))
		return 0;
2808

2809
	if (unlikely(!mm))
2810
		mm = &init_mm;
2811

2812
	if (page_is_file_cache(page)) {
2813 2814
		ret = __mem_cgroup_try_charge(mm, gfp_mask, 1, &memcg, true);
		if (ret || !memcg)
2815
			return ret;
2816

2817 2818 2819 2820 2821
		/*
		 * FUSE reuses pages without going through the final
		 * put that would remove them from the LRU list, make
		 * sure that they get relinked properly.
		 */
2822
		__mem_cgroup_commit_charge_lrucare(page, memcg,
2823 2824 2825
					MEM_CGROUP_CHARGE_TYPE_CACHE);
		return ret;
	}
D
Daisuke Nishimura 已提交
2826 2827
	/* shmem */
	if (PageSwapCache(page)) {
2828
		ret = mem_cgroup_try_charge_swapin(mm, page, gfp_mask, &memcg);
D
Daisuke Nishimura 已提交
2829
		if (!ret)
2830
			__mem_cgroup_commit_charge_swapin(page, memcg,
D
Daisuke Nishimura 已提交
2831 2832 2833
					MEM_CGROUP_CHARGE_TYPE_SHMEM);
	} else
		ret = mem_cgroup_charge_common(page, mm, gfp_mask,
2834
					MEM_CGROUP_CHARGE_TYPE_SHMEM);
2835 2836

	return ret;
2837 2838
}

2839 2840 2841
/*
 * While swap-in, try_charge -> commit or cancel, the page is locked.
 * And when try_charge() successfully returns, one refcnt to memcg without
2842
 * struct page_cgroup is acquired. This refcnt will be consumed by
2843 2844
 * "commit()" or removed by "cancel()"
 */
2845 2846 2847 2848
int mem_cgroup_try_charge_swapin(struct mm_struct *mm,
				 struct page *page,
				 gfp_t mask, struct mem_cgroup **ptr)
{
2849
	struct mem_cgroup *memcg;
2850
	int ret;
2851

2852 2853
	*ptr = NULL;

2854
	if (mem_cgroup_disabled())
2855 2856 2857 2858 2859 2860
		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
2861 2862 2863
	 * 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.
2864 2865
	 */
	if (!PageSwapCache(page))
H
Hugh Dickins 已提交
2866
		goto charge_cur_mm;
2867 2868
	memcg = try_get_mem_cgroup_from_page(page);
	if (!memcg)
2869
		goto charge_cur_mm;
2870
	*ptr = memcg;
2871
	ret = __mem_cgroup_try_charge(NULL, mask, 1, ptr, true);
2872
	css_put(&memcg->css);
2873
	return ret;
2874 2875 2876
charge_cur_mm:
	if (unlikely(!mm))
		mm = &init_mm;
2877
	return __mem_cgroup_try_charge(mm, mask, 1, ptr, true);
2878 2879
}

D
Daisuke Nishimura 已提交
2880 2881 2882
static void
__mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr,
					enum charge_type ctype)
2883
{
2884
	if (mem_cgroup_disabled())
2885 2886 2887
		return;
	if (!ptr)
		return;
2888
	cgroup_exclude_rmdir(&ptr->css);
2889 2890

	__mem_cgroup_commit_charge_lrucare(page, ptr, ctype);
2891 2892 2893
	/*
	 * Now swap is on-memory. This means this page may be
	 * counted both as mem and swap....double count.
2894 2895 2896
	 * 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.
2897
	 */
2898
	if (do_swap_account && PageSwapCache(page)) {
2899
		swp_entry_t ent = {.val = page_private(page)};
2900
		unsigned short id;
2901
		struct mem_cgroup *memcg;
2902 2903 2904 2905

		id = swap_cgroup_record(ent, 0);
		rcu_read_lock();
		memcg = mem_cgroup_lookup(id);
2906
		if (memcg) {
2907 2908 2909 2910
			/*
			 * This recorded memcg can be obsolete one. So, avoid
			 * calling css_tryget
			 */
2911
			if (!mem_cgroup_is_root(memcg))
2912
				res_counter_uncharge(&memcg->memsw, PAGE_SIZE);
2913
			mem_cgroup_swap_statistics(memcg, false);
2914 2915
			mem_cgroup_put(memcg);
		}
2916
		rcu_read_unlock();
2917
	}
2918 2919 2920 2921 2922 2923
	/*
	 * At swapin, we may charge account against cgroup which has no tasks.
	 * So, rmdir()->pre_destroy() can be called while we do this charge.
	 * In that case, we need to call pre_destroy() again. check it here.
	 */
	cgroup_release_and_wakeup_rmdir(&ptr->css);
2924 2925
}

D
Daisuke Nishimura 已提交
2926 2927 2928 2929 2930 2931
void mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr)
{
	__mem_cgroup_commit_charge_swapin(page, ptr,
					MEM_CGROUP_CHARGE_TYPE_MAPPED);
}

2932
void mem_cgroup_cancel_charge_swapin(struct mem_cgroup *memcg)
2933
{
2934
	if (mem_cgroup_disabled())
2935
		return;
2936
	if (!memcg)
2937
		return;
2938
	__mem_cgroup_cancel_charge(memcg, 1);
2939 2940
}

2941
static void mem_cgroup_do_uncharge(struct mem_cgroup *memcg,
2942 2943
				   unsigned int nr_pages,
				   const enum charge_type ctype)
2944 2945 2946
{
	struct memcg_batch_info *batch = NULL;
	bool uncharge_memsw = true;
2947

2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958
	/* 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)
2959
		batch->memcg = memcg;
2960 2961
	/*
	 * do_batch > 0 when unmapping pages or inode invalidate/truncate.
L
Lucas De Marchi 已提交
2962
	 * In those cases, all pages freed continuously can be expected to be in
2963 2964 2965 2966 2967 2968 2969 2970
	 * 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;

2971
	if (nr_pages > 1)
A
Andrea Arcangeli 已提交
2972 2973
		goto direct_uncharge;

2974 2975 2976 2977 2978
	/*
	 * 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.
	 */
2979
	if (batch->memcg != memcg)
2980 2981
		goto direct_uncharge;
	/* remember freed charge and uncharge it later */
2982
	batch->nr_pages++;
2983
	if (uncharge_memsw)
2984
		batch->memsw_nr_pages++;
2985 2986
	return;
direct_uncharge:
2987
	res_counter_uncharge(&memcg->res, nr_pages * PAGE_SIZE);
2988
	if (uncharge_memsw)
2989 2990 2991
		res_counter_uncharge(&memcg->memsw, nr_pages * PAGE_SIZE);
	if (unlikely(batch->memcg != memcg))
		memcg_oom_recover(memcg);
2992 2993
	return;
}
2994

2995
/*
2996
 * uncharge if !page_mapped(page)
2997
 */
2998
static struct mem_cgroup *
2999
__mem_cgroup_uncharge_common(struct page *page, enum charge_type ctype)
3000
{
3001
	struct mem_cgroup *memcg = NULL;
3002 3003
	unsigned int nr_pages = 1;
	struct page_cgroup *pc;
3004

3005
	if (mem_cgroup_disabled())
3006
		return NULL;
3007

K
KAMEZAWA Hiroyuki 已提交
3008
	if (PageSwapCache(page))
3009
		return NULL;
K
KAMEZAWA Hiroyuki 已提交
3010

A
Andrea Arcangeli 已提交
3011
	if (PageTransHuge(page)) {
3012
		nr_pages <<= compound_order(page);
A
Andrea Arcangeli 已提交
3013 3014
		VM_BUG_ON(!PageTransHuge(page));
	}
3015
	/*
3016
	 * Check if our page_cgroup is valid
3017
	 */
3018 3019
	pc = lookup_page_cgroup(page);
	if (unlikely(!pc || !PageCgroupUsed(pc)))
3020
		return NULL;
3021

3022
	lock_page_cgroup(pc);
K
KAMEZAWA Hiroyuki 已提交
3023

3024
	memcg = pc->mem_cgroup;
3025

K
KAMEZAWA Hiroyuki 已提交
3026 3027 3028 3029 3030
	if (!PageCgroupUsed(pc))
		goto unlock_out;

	switch (ctype) {
	case MEM_CGROUP_CHARGE_TYPE_MAPPED:
K
KAMEZAWA Hiroyuki 已提交
3031
	case MEM_CGROUP_CHARGE_TYPE_DROP:
3032 3033
		/* See mem_cgroup_prepare_migration() */
		if (page_mapped(page) || PageCgroupMigration(pc))
K
KAMEZAWA Hiroyuki 已提交
3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044
			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;
3045
	}
K
KAMEZAWA Hiroyuki 已提交
3046

3047
	mem_cgroup_charge_statistics(memcg, PageCgroupCache(pc), -nr_pages);
K
KAMEZAWA Hiroyuki 已提交
3048

3049
	ClearPageCgroupUsed(pc);
3050 3051 3052 3053 3054 3055
	/*
	 * 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.
	 */
3056

3057
	unlock_page_cgroup(pc);
K
KAMEZAWA Hiroyuki 已提交
3058
	/*
3059
	 * even after unlock, we have memcg->res.usage here and this memcg
K
KAMEZAWA Hiroyuki 已提交
3060 3061
	 * will never be freed.
	 */
3062
	memcg_check_events(memcg, page);
K
KAMEZAWA Hiroyuki 已提交
3063
	if (do_swap_account && ctype == MEM_CGROUP_CHARGE_TYPE_SWAPOUT) {
3064 3065
		mem_cgroup_swap_statistics(memcg, true);
		mem_cgroup_get(memcg);
K
KAMEZAWA Hiroyuki 已提交
3066
	}
3067 3068
	if (!mem_cgroup_is_root(memcg))
		mem_cgroup_do_uncharge(memcg, nr_pages, ctype);
3069

3070
	return memcg;
K
KAMEZAWA Hiroyuki 已提交
3071 3072 3073

unlock_out:
	unlock_page_cgroup(pc);
3074
	return NULL;
3075 3076
}

3077 3078
void mem_cgroup_uncharge_page(struct page *page)
{
3079 3080 3081 3082 3083
	/* early check. */
	if (page_mapped(page))
		return;
	if (page->mapping && !PageAnon(page))
		return;
3084 3085 3086 3087 3088 3089
	__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));
3090
	VM_BUG_ON(page->mapping);
3091 3092 3093
	__mem_cgroup_uncharge_common(page, MEM_CGROUP_CHARGE_TYPE_CACHE);
}

3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107
/*
 * 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;
3108 3109
		current->memcg_batch.nr_pages = 0;
		current->memcg_batch.memsw_nr_pages = 0;
3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129
	}
}

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.
	 */
3130 3131 3132 3133 3134 3135
	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);
3136
	memcg_oom_recover(batch->memcg);
3137 3138 3139 3140
	/* forget this pointer (for sanity check) */
	batch->memcg = NULL;
}

3141
#ifdef CONFIG_SWAP
3142
/*
3143
 * called after __delete_from_swap_cache() and drop "page" account.
3144 3145
 * memcg information is recorded to swap_cgroup of "ent"
 */
K
KAMEZAWA Hiroyuki 已提交
3146 3147
void
mem_cgroup_uncharge_swapcache(struct page *page, swp_entry_t ent, bool swapout)
3148 3149
{
	struct mem_cgroup *memcg;
K
KAMEZAWA Hiroyuki 已提交
3150 3151 3152 3153 3154 3155
	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);
3156

K
KAMEZAWA Hiroyuki 已提交
3157 3158 3159 3160 3161
	/*
	 * record memcg information,  if swapout && memcg != NULL,
	 * mem_cgroup_get() was called in uncharge().
	 */
	if (do_swap_account && swapout && memcg)
3162
		swap_cgroup_record(ent, css_id(&memcg->css));
3163
}
3164
#endif
3165 3166 3167 3168 3169 3170 3171

#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
3172
{
3173
	struct mem_cgroup *memcg;
3174
	unsigned short id;
3175 3176 3177 3178

	if (!do_swap_account)
		return;

3179 3180 3181
	id = swap_cgroup_record(ent, 0);
	rcu_read_lock();
	memcg = mem_cgroup_lookup(id);
3182
	if (memcg) {
3183 3184 3185 3186
		/*
		 * We uncharge this because swap is freed.
		 * This memcg can be obsolete one. We avoid calling css_tryget
		 */
3187
		if (!mem_cgroup_is_root(memcg))
3188
			res_counter_uncharge(&memcg->memsw, PAGE_SIZE);
3189
		mem_cgroup_swap_statistics(memcg, false);
3190 3191
		mem_cgroup_put(memcg);
	}
3192
	rcu_read_unlock();
K
KAMEZAWA Hiroyuki 已提交
3193
}
3194 3195 3196 3197 3198 3199

/**
 * 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
3200
 * @need_fixup: whether we should fixup res_counters and refcounts.
3201 3202 3203 3204 3205 3206 3207 3208 3209 3210
 *
 * 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,
3211
		struct mem_cgroup *from, struct mem_cgroup *to, bool need_fixup)
3212 3213 3214 3215 3216 3217 3218 3219
{
	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);
3220
		mem_cgroup_swap_statistics(to, true);
3221
		/*
3222 3223 3224 3225 3226 3227
		 * 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.
3228 3229
		 */
		mem_cgroup_get(to);
3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240
		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);
		}
3241 3242 3243 3244 3245 3246
		return 0;
	}
	return -EINVAL;
}
#else
static inline int mem_cgroup_move_swap_account(swp_entry_t entry,
3247
		struct mem_cgroup *from, struct mem_cgroup *to, bool need_fixup)
3248 3249 3250
{
	return -EINVAL;
}
3251
#endif
K
KAMEZAWA Hiroyuki 已提交
3252

3253
/*
3254 3255
 * Before starting migration, account PAGE_SIZE to mem_cgroup that the old
 * page belongs to.
3256
 */
3257
int mem_cgroup_prepare_migration(struct page *page,
3258
	struct page *newpage, struct mem_cgroup **ptr, gfp_t gfp_mask)
3259
{
3260
	struct mem_cgroup *memcg = NULL;
3261
	struct page_cgroup *pc;
3262
	enum charge_type ctype;
3263
	int ret = 0;
3264

3265 3266
	*ptr = NULL;

A
Andrea Arcangeli 已提交
3267
	VM_BUG_ON(PageTransHuge(page));
3268
	if (mem_cgroup_disabled())
3269 3270
		return 0;

3271 3272 3273
	pc = lookup_page_cgroup(page);
	lock_page_cgroup(pc);
	if (PageCgroupUsed(pc)) {
3274 3275
		memcg = pc->mem_cgroup;
		css_get(&memcg->css);
3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306
		/*
		 * 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);
3307
	}
3308
	unlock_page_cgroup(pc);
3309 3310 3311 3312
	/*
	 * If the page is not charged at this point,
	 * we return here.
	 */
3313
	if (!memcg)
3314
		return 0;
3315

3316
	*ptr = memcg;
3317
	ret = __mem_cgroup_try_charge(NULL, gfp_mask, 1, ptr, false);
3318
	css_put(&memcg->css);/* drop extra refcnt */
3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329
	if (ret || *ptr == NULL) {
		if (PageAnon(page)) {
			lock_page_cgroup(pc);
			ClearPageCgroupMigration(pc);
			unlock_page_cgroup(pc);
			/*
			 * The old page may be fully unmapped while we kept it.
			 */
			mem_cgroup_uncharge_page(page);
		}
		return -ENOMEM;
3330
	}
3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343
	/*
	 * We charge new page before it's used/mapped. So, even if unlock_page()
	 * is called before end_migration, we can catch all events on this new
	 * page. In the case new page is migrated but not remapped, new page's
	 * mapcount will be finally 0 and we call uncharge in end_migration().
	 */
	pc = lookup_page_cgroup(newpage);
	if (PageAnon(page))
		ctype = MEM_CGROUP_CHARGE_TYPE_MAPPED;
	else if (page_is_file_cache(page))
		ctype = MEM_CGROUP_CHARGE_TYPE_CACHE;
	else
		ctype = MEM_CGROUP_CHARGE_TYPE_SHMEM;
3344
	__mem_cgroup_commit_charge(memcg, page, 1, pc, ctype);
3345
	return ret;
3346
}
3347

3348
/* remove redundant charge if migration failed*/
3349
void mem_cgroup_end_migration(struct mem_cgroup *memcg,
3350
	struct page *oldpage, struct page *newpage, bool migration_ok)
3351
{
3352
	struct page *used, *unused;
3353 3354
	struct page_cgroup *pc;

3355
	if (!memcg)
3356
		return;
3357
	/* blocks rmdir() */
3358
	cgroup_exclude_rmdir(&memcg->css);
3359
	if (!migration_ok) {
3360 3361
		used = oldpage;
		unused = newpage;
3362
	} else {
3363
		used = newpage;
3364 3365
		unused = oldpage;
	}
3366
	/*
3367 3368 3369
	 * 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.
3370
	 */
3371 3372 3373 3374
	pc = lookup_page_cgroup(oldpage);
	lock_page_cgroup(pc);
	ClearPageCgroupMigration(pc);
	unlock_page_cgroup(pc);
3375

3376 3377
	__mem_cgroup_uncharge_common(unused, MEM_CGROUP_CHARGE_TYPE_FORCE);

3378
	/*
3379 3380 3381 3382 3383 3384
	 * 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)
3385
	 */
3386 3387
	if (PageAnon(used))
		mem_cgroup_uncharge_page(used);
3388
	/*
3389 3390
	 * At migration, we may charge account against cgroup which has no
	 * tasks.
3391 3392 3393
	 * 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.
	 */
3394
	cgroup_release_and_wakeup_rmdir(&memcg->css);
3395
}
3396

3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440
/*
 * At replace page cache, newpage is not under any memcg but it's on
 * LRU. So, this function doesn't touch res_counter but handles LRU
 * in correct way. Both pages are locked so we cannot race with uncharge.
 */
void mem_cgroup_replace_page_cache(struct page *oldpage,
				  struct page *newpage)
{
	struct mem_cgroup *memcg;
	struct page_cgroup *pc;
	struct zone *zone;
	enum charge_type type = MEM_CGROUP_CHARGE_TYPE_CACHE;
	unsigned long flags;

	if (mem_cgroup_disabled())
		return;

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

	if (PageSwapBacked(oldpage))
		type = MEM_CGROUP_CHARGE_TYPE_SHMEM;

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

3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486
#ifdef CONFIG_DEBUG_VM
static struct page_cgroup *lookup_page_cgroup_used(struct page *page)
{
	struct page_cgroup *pc;

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

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

	return lookup_page_cgroup_used(page) != NULL;
}

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

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

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

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

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

3487 3488
static DEFINE_MUTEX(set_limit_mutex);

3489
static int mem_cgroup_resize_limit(struct mem_cgroup *memcg,
3490
				unsigned long long val)
3491
{
3492
	int retry_count;
3493
	u64 memswlimit, memlimit;
3494
	int ret = 0;
3495 3496
	int children = mem_cgroup_count_children(memcg);
	u64 curusage, oldusage;
3497
	int enlarge;
3498 3499 3500 3501 3502 3503 3504 3505 3506

	/*
	 * 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);
3507

3508
	enlarge = 0;
3509
	while (retry_count) {
3510 3511 3512 3513
		if (signal_pending(current)) {
			ret = -EINTR;
			break;
		}
3514 3515 3516
		/*
		 * Rather than hide all in some function, I do this in
		 * open coded manner. You see what this really does.
3517
		 * We have to guarantee memcg->res.limit < memcg->memsw.limit.
3518 3519 3520 3521 3522 3523
		 */
		mutex_lock(&set_limit_mutex);
		memswlimit = res_counter_read_u64(&memcg->memsw, RES_LIMIT);
		if (memswlimit < val) {
			ret = -EINVAL;
			mutex_unlock(&set_limit_mutex);
3524 3525
			break;
		}
3526 3527 3528 3529 3530

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

3531
		ret = res_counter_set_limit(&memcg->res, val);
3532 3533 3534 3535 3536 3537
		if (!ret) {
			if (memswlimit == val)
				memcg->memsw_is_minimum = true;
			else
				memcg->memsw_is_minimum = false;
		}
3538 3539 3540 3541 3542
		mutex_unlock(&set_limit_mutex);

		if (!ret)
			break;

3543 3544
		mem_cgroup_reclaim(memcg, GFP_KERNEL,
				   MEM_CGROUP_RECLAIM_SHRINK);
3545 3546 3547 3548 3549 3550
		curusage = res_counter_read_u64(&memcg->res, RES_USAGE);
		/* Usage is reduced ? */
  		if (curusage >= oldusage)
			retry_count--;
		else
			oldusage = curusage;
3551
	}
3552 3553
	if (!ret && enlarge)
		memcg_oom_recover(memcg);
3554

3555 3556 3557
	return ret;
}

L
Li Zefan 已提交
3558 3559
static int mem_cgroup_resize_memsw_limit(struct mem_cgroup *memcg,
					unsigned long long val)
3560
{
3561
	int retry_count;
3562
	u64 memlimit, memswlimit, oldusage, curusage;
3563 3564
	int children = mem_cgroup_count_children(memcg);
	int ret = -EBUSY;
3565
	int enlarge = 0;
3566

3567 3568 3569
	/* see mem_cgroup_resize_res_limit */
 	retry_count = children * MEM_CGROUP_RECLAIM_RETRIES;
	oldusage = res_counter_read_u64(&memcg->memsw, RES_USAGE);
3570 3571 3572 3573 3574 3575 3576 3577
	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.
3578
		 * We have to guarantee memcg->res.limit < memcg->memsw.limit.
3579 3580 3581 3582 3583 3584 3585 3586
		 */
		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;
		}
3587 3588 3589
		memswlimit = res_counter_read_u64(&memcg->memsw, RES_LIMIT);
		if (memswlimit < val)
			enlarge = 1;
3590
		ret = res_counter_set_limit(&memcg->memsw, val);
3591 3592 3593 3594 3595 3596
		if (!ret) {
			if (memlimit == val)
				memcg->memsw_is_minimum = true;
			else
				memcg->memsw_is_minimum = false;
		}
3597 3598 3599 3600 3601
		mutex_unlock(&set_limit_mutex);

		if (!ret)
			break;

3602 3603 3604
		mem_cgroup_reclaim(memcg, GFP_KERNEL,
				   MEM_CGROUP_RECLAIM_NOSWAP |
				   MEM_CGROUP_RECLAIM_SHRINK);
3605
		curusage = res_counter_read_u64(&memcg->memsw, RES_USAGE);
3606
		/* Usage is reduced ? */
3607
		if (curusage >= oldusage)
3608
			retry_count--;
3609 3610
		else
			oldusage = curusage;
3611
	}
3612 3613
	if (!ret && enlarge)
		memcg_oom_recover(memcg);
3614 3615 3616
	return ret;
}

3617
unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
3618 3619
					    gfp_t gfp_mask,
					    unsigned long *total_scanned)
3620 3621 3622 3623 3624 3625
{
	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;
3626
	unsigned long long excess;
3627
	unsigned long nr_scanned;
3628 3629 3630 3631

	if (order > 0)
		return 0;

3632
	mctz = soft_limit_tree_node_zone(zone_to_nid(zone), zone_idx(zone));
3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645
	/*
	 * 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;

3646
		nr_scanned = 0;
3647 3648
		reclaimed = mem_cgroup_soft_reclaim(mz->mem, zone,
						    gfp_mask, &nr_scanned);
3649
		nr_reclaimed += reclaimed;
3650
		*total_scanned += nr_scanned;
3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666 3667 3668 3669 3670 3671 3672
		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);
3673
				if (next_mz == mz)
3674
					css_put(&next_mz->mem->css);
3675
				else /* next_mz == NULL or other memcg */
3676 3677 3678 3679
					break;
			} while (1);
		}
		__mem_cgroup_remove_exceeded(mz->mem, mz, mctz);
3680
		excess = res_counter_soft_limit_excess(&mz->mem->res);
3681 3682 3683 3684 3685 3686 3687 3688
		/*
		 * 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.
		 */
3689 3690
		/* If excess == 0, no tree ops */
		__mem_cgroup_insert_exceeded(mz->mem, mz, mctz, excess);
3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708
		spin_unlock(&mctz->lock);
		css_put(&mz->mem->css);
		loop++;
		/*
		 * Could not reclaim anything and there are no more
		 * mem cgroups to try or we seem to be looping without
		 * reclaiming anything.
		 */
		if (!nr_reclaimed &&
			(next_mz == NULL ||
			loop > MEM_CGROUP_MAX_SOFT_LIMIT_RECLAIM_LOOPS))
			break;
	} while (!nr_reclaimed);
	if (next_mz)
		css_put(&next_mz->mem->css);
	return nr_reclaimed;
}

3709 3710 3711 3712
/*
 * This routine traverse page_cgroup in given list and drop them all.
 * *And* this routine doesn't reclaim page itself, just removes page_cgroup.
 */
3713
static int mem_cgroup_force_empty_list(struct mem_cgroup *memcg,
K
KAMEZAWA Hiroyuki 已提交
3714
				int node, int zid, enum lru_list lru)
3715
{
K
KAMEZAWA Hiroyuki 已提交
3716 3717
	struct mem_cgroup_per_zone *mz;
	unsigned long flags, loop;
3718
	struct list_head *list;
3719 3720
	struct page *busy;
	struct zone *zone;
3721
	int ret = 0;
3722

K
KAMEZAWA Hiroyuki 已提交
3723
	zone = &NODE_DATA(node)->node_zones[zid];
3724
	mz = mem_cgroup_zoneinfo(memcg, node, zid);
3725
	list = &mz->lruvec.lists[lru];
3726

3727 3728 3729 3730 3731
	loop = MEM_CGROUP_ZSTAT(mz, lru);
	/* give some margin against EBUSY etc...*/
	loop += 256;
	busy = NULL;
	while (loop--) {
3732
		struct page_cgroup *pc;
3733 3734
		struct page *page;

3735
		ret = 0;
K
KAMEZAWA Hiroyuki 已提交
3736
		spin_lock_irqsave(&zone->lru_lock, flags);
3737
		if (list_empty(list)) {
K
KAMEZAWA Hiroyuki 已提交
3738
			spin_unlock_irqrestore(&zone->lru_lock, flags);
3739
			break;
3740
		}
3741 3742 3743
		page = list_entry(list->prev, struct page, lru);
		if (busy == page) {
			list_move(&page->lru, list);
3744
			busy = NULL;
K
KAMEZAWA Hiroyuki 已提交
3745
			spin_unlock_irqrestore(&zone->lru_lock, flags);
3746 3747
			continue;
		}
K
KAMEZAWA Hiroyuki 已提交
3748
		spin_unlock_irqrestore(&zone->lru_lock, flags);
3749

3750
		pc = lookup_page_cgroup(page);
3751

3752
		ret = mem_cgroup_move_parent(page, pc, memcg, GFP_KERNEL);
3753
		if (ret == -ENOMEM)
3754
			break;
3755 3756 3757

		if (ret == -EBUSY || ret == -EINVAL) {
			/* found lock contention or "pc" is obsolete. */
3758
			busy = page;
3759 3760 3761
			cond_resched();
		} else
			busy = NULL;
3762
	}
K
KAMEZAWA Hiroyuki 已提交
3763

3764 3765 3766
	if (!ret && !list_empty(list))
		return -EBUSY;
	return ret;
3767 3768 3769 3770 3771 3772
}

/*
 * make mem_cgroup's charge to be 0 if there is no task.
 * This enables deleting this mem_cgroup.
 */
3773
static int mem_cgroup_force_empty(struct mem_cgroup *memcg, bool free_all)
3774
{
3775 3776 3777
	int ret;
	int node, zid, shrink;
	int nr_retries = MEM_CGROUP_RECLAIM_RETRIES;
3778
	struct cgroup *cgrp = memcg->css.cgroup;
3779

3780
	css_get(&memcg->css);
3781 3782

	shrink = 0;
3783 3784 3785
	/* should free all ? */
	if (free_all)
		goto try_to_free;
3786
move_account:
3787
	do {
3788
		ret = -EBUSY;
3789 3790 3791 3792
		if (cgroup_task_count(cgrp) || !list_empty(&cgrp->children))
			goto out;
		ret = -EINTR;
		if (signal_pending(current))
3793
			goto out;
3794 3795
		/* This is for making all *used* pages to be on LRU. */
		lru_add_drain_all();
3796
		drain_all_stock_sync(memcg);
3797
		ret = 0;
3798
		mem_cgroup_start_move(memcg);
3799
		for_each_node_state(node, N_HIGH_MEMORY) {
3800
			for (zid = 0; !ret && zid < MAX_NR_ZONES; zid++) {
3801
				enum lru_list l;
3802
				for_each_lru(l) {
3803
					ret = mem_cgroup_force_empty_list(memcg,
K
KAMEZAWA Hiroyuki 已提交
3804
							node, zid, l);
3805 3806 3807
					if (ret)
						break;
				}
3808
			}
3809 3810 3811
			if (ret)
				break;
		}
3812 3813
		mem_cgroup_end_move(memcg);
		memcg_oom_recover(memcg);
3814 3815 3816
		/* it seems parent cgroup doesn't have enough mem */
		if (ret == -ENOMEM)
			goto try_to_free;
3817
		cond_resched();
3818
	/* "ret" should also be checked to ensure all lists are empty. */
3819
	} while (memcg->res.usage > 0 || ret);
3820
out:
3821
	css_put(&memcg->css);
3822
	return ret;
3823 3824

try_to_free:
3825 3826
	/* returns EBUSY if there is a task or if we come here twice. */
	if (cgroup_task_count(cgrp) || !list_empty(&cgrp->children) || shrink) {
3827 3828 3829
		ret = -EBUSY;
		goto out;
	}
3830 3831
	/* we call try-to-free pages for make this cgroup empty */
	lru_add_drain_all();
3832 3833
	/* try to free all pages in this cgroup */
	shrink = 1;
3834
	while (nr_retries && memcg->res.usage > 0) {
3835
		int progress;
3836 3837 3838 3839 3840

		if (signal_pending(current)) {
			ret = -EINTR;
			goto out;
		}
3841
		progress = try_to_free_mem_cgroup_pages(memcg, GFP_KERNEL,
3842
						false);
3843
		if (!progress) {
3844
			nr_retries--;
3845
			/* maybe some writeback is necessary */
3846
			congestion_wait(BLK_RW_ASYNC, HZ/10);
3847
		}
3848 3849

	}
K
KAMEZAWA Hiroyuki 已提交
3850
	lru_add_drain();
3851
	/* try move_account...there may be some *locked* pages. */
3852
	goto move_account;
3853 3854
}

3855 3856 3857 3858 3859 3860
int mem_cgroup_force_empty_write(struct cgroup *cont, unsigned int event)
{
	return mem_cgroup_force_empty(mem_cgroup_from_cont(cont), true);
}


3861 3862 3863 3864 3865 3866 3867 3868 3869
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;
3870
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
3871
	struct cgroup *parent = cont->parent;
3872
	struct mem_cgroup *parent_memcg = NULL;
3873 3874

	if (parent)
3875
		parent_memcg = mem_cgroup_from_cont(parent);
3876 3877 3878

	cgroup_lock();
	/*
3879
	 * If parent's use_hierarchy is set, we can't make any modifications
3880 3881 3882 3883 3884 3885
	 * 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.
	 */
3886
	if ((!parent_memcg || !parent_memcg->use_hierarchy) &&
3887 3888
				(val == 1 || val == 0)) {
		if (list_empty(&cont->children))
3889
			memcg->use_hierarchy = val;
3890 3891 3892 3893 3894 3895 3896 3897 3898
		else
			retval = -EBUSY;
	} else
		retval = -EINVAL;
	cgroup_unlock();

	return retval;
}

3899

3900
static unsigned long mem_cgroup_recursive_stat(struct mem_cgroup *memcg,
3901
					       enum mem_cgroup_stat_index idx)
3902
{
K
KAMEZAWA Hiroyuki 已提交
3903
	struct mem_cgroup *iter;
3904
	long val = 0;
3905

3906
	/* Per-cpu values can be negative, use a signed accumulator */
3907
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
3908 3909 3910 3911 3912
		val += mem_cgroup_read_stat(iter, idx);

	if (val < 0) /* race ? */
		val = 0;
	return val;
3913 3914
}

3915
static inline u64 mem_cgroup_usage(struct mem_cgroup *memcg, bool swap)
3916
{
K
KAMEZAWA Hiroyuki 已提交
3917
	u64 val;
3918

3919
	if (!mem_cgroup_is_root(memcg)) {
3920
		if (!swap)
3921
			return res_counter_read_u64(&memcg->res, RES_USAGE);
3922
		else
3923
			return res_counter_read_u64(&memcg->memsw, RES_USAGE);
3924 3925
	}

3926 3927
	val = mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_CACHE);
	val += mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_RSS);
3928

K
KAMEZAWA Hiroyuki 已提交
3929
	if (swap)
3930
		val += mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_SWAPOUT);
3931 3932 3933 3934

	return val << PAGE_SHIFT;
}

3935
static u64 mem_cgroup_read(struct cgroup *cont, struct cftype *cft)
B
Balbir Singh 已提交
3936
{
3937
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
3938
	u64 val;
3939 3940 3941 3942 3943 3944
	int type, name;

	type = MEMFILE_TYPE(cft->private);
	name = MEMFILE_ATTR(cft->private);
	switch (type) {
	case _MEM:
3945
		if (name == RES_USAGE)
3946
			val = mem_cgroup_usage(memcg, false);
3947
		else
3948
			val = res_counter_read_u64(&memcg->res, name);
3949 3950
		break;
	case _MEMSWAP:
3951
		if (name == RES_USAGE)
3952
			val = mem_cgroup_usage(memcg, true);
3953
		else
3954
			val = res_counter_read_u64(&memcg->memsw, name);
3955 3956 3957 3958 3959 3960
		break;
	default:
		BUG();
		break;
	}
	return val;
B
Balbir Singh 已提交
3961
}
3962 3963 3964 3965
/*
 * The user of this function is...
 * RES_LIMIT.
 */
3966 3967
static int mem_cgroup_write(struct cgroup *cont, struct cftype *cft,
			    const char *buffer)
B
Balbir Singh 已提交
3968
{
3969
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
3970
	int type, name;
3971 3972 3973
	unsigned long long val;
	int ret;

3974 3975 3976
	type = MEMFILE_TYPE(cft->private);
	name = MEMFILE_ATTR(cft->private);
	switch (name) {
3977
	case RES_LIMIT:
3978 3979 3980 3981
		if (mem_cgroup_is_root(memcg)) { /* Can't set limit on root */
			ret = -EINVAL;
			break;
		}
3982 3983
		/* This function does all necessary parse...reuse it */
		ret = res_counter_memparse_write_strategy(buffer, &val);
3984 3985 3986
		if (ret)
			break;
		if (type == _MEM)
3987
			ret = mem_cgroup_resize_limit(memcg, val);
3988 3989
		else
			ret = mem_cgroup_resize_memsw_limit(memcg, val);
3990
		break;
3991 3992 3993 3994 3995 3996 3997 3998 3999 4000 4001 4002 4003 4004
	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;
4005 4006 4007 4008 4009
	default:
		ret = -EINVAL; /* should be BUG() ? */
		break;
	}
	return ret;
B
Balbir Singh 已提交
4010 4011
}

4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039
static void memcg_get_hierarchical_limit(struct mem_cgroup *memcg,
		unsigned long long *mem_limit, unsigned long long *memsw_limit)
{
	struct cgroup *cgroup;
	unsigned long long min_limit, min_memsw_limit, tmp;

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

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

4040
static int mem_cgroup_reset(struct cgroup *cont, unsigned int event)
4041
{
4042
	struct mem_cgroup *memcg;
4043
	int type, name;
4044

4045
	memcg = mem_cgroup_from_cont(cont);
4046 4047 4048
	type = MEMFILE_TYPE(event);
	name = MEMFILE_ATTR(event);
	switch (name) {
4049
	case RES_MAX_USAGE:
4050
		if (type == _MEM)
4051
			res_counter_reset_max(&memcg->res);
4052
		else
4053
			res_counter_reset_max(&memcg->memsw);
4054 4055
		break;
	case RES_FAILCNT:
4056
		if (type == _MEM)
4057
			res_counter_reset_failcnt(&memcg->res);
4058
		else
4059
			res_counter_reset_failcnt(&memcg->memsw);
4060 4061
		break;
	}
4062

4063
	return 0;
4064 4065
}

4066 4067 4068 4069 4070 4071
static u64 mem_cgroup_move_charge_read(struct cgroup *cgrp,
					struct cftype *cft)
{
	return mem_cgroup_from_cont(cgrp)->move_charge_at_immigrate;
}

4072
#ifdef CONFIG_MMU
4073 4074 4075
static int mem_cgroup_move_charge_write(struct cgroup *cgrp,
					struct cftype *cft, u64 val)
{
4076
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4077 4078 4079 4080 4081 4082 4083 4084 4085

	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();
4086
	memcg->move_charge_at_immigrate = val;
4087 4088 4089 4090
	cgroup_unlock();

	return 0;
}
4091 4092 4093 4094 4095 4096 4097
#else
static int mem_cgroup_move_charge_write(struct cgroup *cgrp,
					struct cftype *cft, u64 val)
{
	return -ENOSYS;
}
#endif
4098

K
KAMEZAWA Hiroyuki 已提交
4099 4100 4101 4102 4103

/* For read statistics */
enum {
	MCS_CACHE,
	MCS_RSS,
4104
	MCS_FILE_MAPPED,
K
KAMEZAWA Hiroyuki 已提交
4105 4106
	MCS_PGPGIN,
	MCS_PGPGOUT,
4107
	MCS_SWAP,
4108 4109
	MCS_PGFAULT,
	MCS_PGMAJFAULT,
K
KAMEZAWA Hiroyuki 已提交
4110 4111 4112 4113 4114 4115 4116 4117 4118 4119
	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];
4120 4121
};

K
KAMEZAWA Hiroyuki 已提交
4122 4123 4124 4125 4126 4127
struct {
	char *local_name;
	char *total_name;
} memcg_stat_strings[NR_MCS_STAT] = {
	{"cache", "total_cache"},
	{"rss", "total_rss"},
4128
	{"mapped_file", "total_mapped_file"},
K
KAMEZAWA Hiroyuki 已提交
4129 4130
	{"pgpgin", "total_pgpgin"},
	{"pgpgout", "total_pgpgout"},
4131
	{"swap", "total_swap"},
4132 4133
	{"pgfault", "total_pgfault"},
	{"pgmajfault", "total_pgmajfault"},
K
KAMEZAWA Hiroyuki 已提交
4134 4135 4136 4137 4138 4139 4140 4141
	{"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
4142
static void
4143
mem_cgroup_get_local_stat(struct mem_cgroup *memcg, struct mcs_total_stat *s)
K
KAMEZAWA Hiroyuki 已提交
4144 4145 4146 4147
{
	s64 val;

	/* per cpu stat */
4148
	val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_CACHE);
K
KAMEZAWA Hiroyuki 已提交
4149
	s->stat[MCS_CACHE] += val * PAGE_SIZE;
4150
	val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_RSS);
K
KAMEZAWA Hiroyuki 已提交
4151
	s->stat[MCS_RSS] += val * PAGE_SIZE;
4152
	val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_FILE_MAPPED);
4153
	s->stat[MCS_FILE_MAPPED] += val * PAGE_SIZE;
4154
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGPGIN);
K
KAMEZAWA Hiroyuki 已提交
4155
	s->stat[MCS_PGPGIN] += val;
4156
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGPGOUT);
K
KAMEZAWA Hiroyuki 已提交
4157
	s->stat[MCS_PGPGOUT] += val;
4158
	if (do_swap_account) {
4159
		val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_SWAPOUT);
4160 4161
		s->stat[MCS_SWAP] += val * PAGE_SIZE;
	}
4162
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGFAULT);
4163
	s->stat[MCS_PGFAULT] += val;
4164
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGMAJFAULT);
4165
	s->stat[MCS_PGMAJFAULT] += val;
K
KAMEZAWA Hiroyuki 已提交
4166 4167

	/* per zone stat */
4168
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_INACTIVE_ANON));
K
KAMEZAWA Hiroyuki 已提交
4169
	s->stat[MCS_INACTIVE_ANON] += val * PAGE_SIZE;
4170
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_ACTIVE_ANON));
K
KAMEZAWA Hiroyuki 已提交
4171
	s->stat[MCS_ACTIVE_ANON] += val * PAGE_SIZE;
4172
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_INACTIVE_FILE));
K
KAMEZAWA Hiroyuki 已提交
4173
	s->stat[MCS_INACTIVE_FILE] += val * PAGE_SIZE;
4174
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_ACTIVE_FILE));
K
KAMEZAWA Hiroyuki 已提交
4175
	s->stat[MCS_ACTIVE_FILE] += val * PAGE_SIZE;
4176
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_UNEVICTABLE));
K
KAMEZAWA Hiroyuki 已提交
4177 4178 4179 4180
	s->stat[MCS_UNEVICTABLE] += val * PAGE_SIZE;
}

static void
4181
mem_cgroup_get_total_stat(struct mem_cgroup *memcg, struct mcs_total_stat *s)
K
KAMEZAWA Hiroyuki 已提交
4182
{
K
KAMEZAWA Hiroyuki 已提交
4183 4184
	struct mem_cgroup *iter;

4185
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
4186
		mem_cgroup_get_local_stat(iter, s);
K
KAMEZAWA Hiroyuki 已提交
4187 4188
}

4189 4190 4191 4192 4193 4194 4195 4196 4197
#ifdef CONFIG_NUMA
static int mem_control_numa_stat_show(struct seq_file *m, void *arg)
{
	int nid;
	unsigned long total_nr, file_nr, anon_nr, unevictable_nr;
	unsigned long node_nr;
	struct cgroup *cont = m->private;
	struct mem_cgroup *mem_cont = mem_cgroup_from_cont(cont);

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

4206
	file_nr = mem_cgroup_nr_lru_pages(mem_cont, LRU_ALL_FILE);
4207 4208
	seq_printf(m, "file=%lu", file_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4209 4210
		node_nr = mem_cgroup_node_nr_lru_pages(mem_cont, nid,
				LRU_ALL_FILE);
4211 4212 4213 4214
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');

4215
	anon_nr = mem_cgroup_nr_lru_pages(mem_cont, LRU_ALL_ANON);
4216 4217
	seq_printf(m, "anon=%lu", anon_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4218 4219
		node_nr = mem_cgroup_node_nr_lru_pages(mem_cont, nid,
				LRU_ALL_ANON);
4220 4221 4222 4223
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');

4224
	unevictable_nr = mem_cgroup_nr_lru_pages(mem_cont, BIT(LRU_UNEVICTABLE));
4225 4226
	seq_printf(m, "unevictable=%lu", unevictable_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4227 4228
		node_nr = mem_cgroup_node_nr_lru_pages(mem_cont, nid,
				BIT(LRU_UNEVICTABLE));
4229 4230 4231 4232 4233 4234 4235
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');
	return 0;
}
#endif /* CONFIG_NUMA */

4236 4237
static int mem_control_stat_show(struct cgroup *cont, struct cftype *cft,
				 struct cgroup_map_cb *cb)
4238 4239
{
	struct mem_cgroup *mem_cont = mem_cgroup_from_cont(cont);
K
KAMEZAWA Hiroyuki 已提交
4240
	struct mcs_total_stat mystat;
4241 4242
	int i;

K
KAMEZAWA Hiroyuki 已提交
4243 4244
	memset(&mystat, 0, sizeof(mystat));
	mem_cgroup_get_local_stat(mem_cont, &mystat);
4245

4246

4247 4248 4249
	for (i = 0; i < NR_MCS_STAT; i++) {
		if (i == MCS_SWAP && !do_swap_account)
			continue;
K
KAMEZAWA Hiroyuki 已提交
4250
		cb->fill(cb, memcg_stat_strings[i].local_name, mystat.stat[i]);
4251
	}
L
Lee Schermerhorn 已提交
4252

K
KAMEZAWA Hiroyuki 已提交
4253
	/* Hierarchical information */
4254 4255 4256 4257 4258 4259 4260
	{
		unsigned long long limit, memsw_limit;
		memcg_get_hierarchical_limit(mem_cont, &limit, &memsw_limit);
		cb->fill(cb, "hierarchical_memory_limit", limit);
		if (do_swap_account)
			cb->fill(cb, "hierarchical_memsw_limit", memsw_limit);
	}
K
KOSAKI Motohiro 已提交
4261

K
KAMEZAWA Hiroyuki 已提交
4262 4263
	memset(&mystat, 0, sizeof(mystat));
	mem_cgroup_get_total_stat(mem_cont, &mystat);
4264 4265 4266
	for (i = 0; i < NR_MCS_STAT; i++) {
		if (i == MCS_SWAP && !do_swap_account)
			continue;
K
KAMEZAWA Hiroyuki 已提交
4267
		cb->fill(cb, memcg_stat_strings[i].total_name, mystat.stat[i]);
4268
	}
K
KAMEZAWA Hiroyuki 已提交
4269

K
KOSAKI Motohiro 已提交
4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296
#ifdef CONFIG_DEBUG_VM
	{
		int nid, zid;
		struct mem_cgroup_per_zone *mz;
		unsigned long recent_rotated[2] = {0, 0};
		unsigned long recent_scanned[2] = {0, 0};

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

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

4297 4298 4299
	return 0;
}

K
KOSAKI Motohiro 已提交
4300 4301 4302 4303
static u64 mem_cgroup_swappiness_read(struct cgroup *cgrp, struct cftype *cft)
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);

4304
	return mem_cgroup_swappiness(memcg);
K
KOSAKI Motohiro 已提交
4305 4306 4307 4308 4309 4310 4311
}

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;
4312

K
KOSAKI Motohiro 已提交
4313 4314 4315 4316 4317 4318 4319
	if (val > 100)
		return -EINVAL;

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

	parent = mem_cgroup_from_cont(cgrp->parent);
4320 4321 4322

	cgroup_lock();

K
KOSAKI Motohiro 已提交
4323 4324
	/* If under hierarchy, only empty-root can set this value */
	if ((parent->use_hierarchy) ||
4325 4326
	    (memcg->use_hierarchy && !list_empty(&cgrp->children))) {
		cgroup_unlock();
K
KOSAKI Motohiro 已提交
4327
		return -EINVAL;
4328
	}
K
KOSAKI Motohiro 已提交
4329 4330 4331

	memcg->swappiness = val;

4332 4333
	cgroup_unlock();

K
KOSAKI Motohiro 已提交
4334 4335 4336
	return 0;
}

4337 4338 4339 4340 4341 4342 4343 4344
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)
4345
		t = rcu_dereference(memcg->thresholds.primary);
4346
	else
4347
		t = rcu_dereference(memcg->memsw_thresholds.primary);
4348 4349 4350 4351 4352 4353 4354 4355 4356 4357 4358

	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().
	 */
4359
	i = t->current_threshold;
4360 4361 4362 4363 4364 4365 4366 4367 4368 4369 4370 4371 4372 4373 4374 4375 4376 4377 4378 4379 4380 4381 4382

	/*
	 * 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 */
4383
	t->current_threshold = i - 1;
4384 4385 4386 4387 4388 4389
unlock:
	rcu_read_unlock();
}

static void mem_cgroup_threshold(struct mem_cgroup *memcg)
{
4390 4391 4392 4393 4394 4395 4396
	while (memcg) {
		__mem_cgroup_threshold(memcg, false);
		if (do_swap_account)
			__mem_cgroup_threshold(memcg, true);

		memcg = parent_mem_cgroup(memcg);
	}
4397 4398 4399 4400 4401 4402 4403 4404 4405 4406
}

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;
}

4407
static int mem_cgroup_oom_notify_cb(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
4408 4409 4410
{
	struct mem_cgroup_eventfd_list *ev;

4411
	list_for_each_entry(ev, &memcg->oom_notify, list)
K
KAMEZAWA Hiroyuki 已提交
4412 4413 4414 4415
		eventfd_signal(ev->eventfd, 1);
	return 0;
}

4416
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
4417
{
K
KAMEZAWA Hiroyuki 已提交
4418 4419
	struct mem_cgroup *iter;

4420
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
4421
		mem_cgroup_oom_notify_cb(iter);
K
KAMEZAWA Hiroyuki 已提交
4422 4423 4424 4425
}

static int mem_cgroup_usage_register_event(struct cgroup *cgrp,
	struct cftype *cft, struct eventfd_ctx *eventfd, const char *args)
4426 4427
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4428 4429
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
4430 4431
	int type = MEMFILE_TYPE(cft->private);
	u64 threshold, usage;
4432
	int i, size, ret;
4433 4434 4435 4436 4437 4438

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

	mutex_lock(&memcg->thresholds_lock);
4439

4440
	if (type == _MEM)
4441
		thresholds = &memcg->thresholds;
4442
	else if (type == _MEMSWAP)
4443
		thresholds = &memcg->memsw_thresholds;
4444 4445 4446 4447 4448 4449
	else
		BUG();

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

	/* Check if a threshold crossed before adding a new one */
4450
	if (thresholds->primary)
4451 4452
		__mem_cgroup_threshold(memcg, type == _MEMSWAP);

4453
	size = thresholds->primary ? thresholds->primary->size + 1 : 1;
4454 4455

	/* Allocate memory for new array of thresholds */
4456
	new = kmalloc(sizeof(*new) + size * sizeof(struct mem_cgroup_threshold),
4457
			GFP_KERNEL);
4458
	if (!new) {
4459 4460 4461
		ret = -ENOMEM;
		goto unlock;
	}
4462
	new->size = size;
4463 4464

	/* Copy thresholds (if any) to new array */
4465 4466
	if (thresholds->primary) {
		memcpy(new->entries, thresholds->primary->entries, (size - 1) *
4467
				sizeof(struct mem_cgroup_threshold));
4468 4469
	}

4470
	/* Add new threshold */
4471 4472
	new->entries[size - 1].eventfd = eventfd;
	new->entries[size - 1].threshold = threshold;
4473 4474

	/* Sort thresholds. Registering of new threshold isn't time-critical */
4475
	sort(new->entries, size, sizeof(struct mem_cgroup_threshold),
4476 4477 4478
			compare_thresholds, NULL);

	/* Find current threshold */
4479
	new->current_threshold = -1;
4480
	for (i = 0; i < size; i++) {
4481
		if (new->entries[i].threshold < usage) {
4482
			/*
4483 4484
			 * new->current_threshold will not be used until
			 * rcu_assign_pointer(), so it's safe to increment
4485 4486
			 * it here.
			 */
4487
			++new->current_threshold;
4488 4489 4490
		}
	}

4491 4492 4493 4494 4495
	/* Free old spare buffer and save old primary buffer as spare */
	kfree(thresholds->spare);
	thresholds->spare = thresholds->primary;

	rcu_assign_pointer(thresholds->primary, new);
4496

4497
	/* To be sure that nobody uses thresholds */
4498 4499 4500 4501 4502 4503 4504 4505
	synchronize_rcu();

unlock:
	mutex_unlock(&memcg->thresholds_lock);

	return ret;
}

4506
static void mem_cgroup_usage_unregister_event(struct cgroup *cgrp,
K
KAMEZAWA Hiroyuki 已提交
4507
	struct cftype *cft, struct eventfd_ctx *eventfd)
4508 4509
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4510 4511
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
4512 4513
	int type = MEMFILE_TYPE(cft->private);
	u64 usage;
4514
	int i, j, size;
4515 4516 4517

	mutex_lock(&memcg->thresholds_lock);
	if (type == _MEM)
4518
		thresholds = &memcg->thresholds;
4519
	else if (type == _MEMSWAP)
4520
		thresholds = &memcg->memsw_thresholds;
4521 4522 4523 4524 4525 4526 4527 4528 4529 4530 4531 4532 4533 4534 4535
	else
		BUG();

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

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

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

	/* Calculate new number of threshold */
4536 4537 4538
	size = 0;
	for (i = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd != eventfd)
4539 4540 4541
			size++;
	}

4542
	new = thresholds->spare;
4543

4544 4545
	/* Set thresholds array to NULL if we don't have thresholds */
	if (!size) {
4546 4547
		kfree(new);
		new = NULL;
4548
		goto swap_buffers;
4549 4550
	}

4551
	new->size = size;
4552 4553

	/* Copy thresholds and find current threshold */
4554 4555 4556
	new->current_threshold = -1;
	for (i = 0, j = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd == eventfd)
4557 4558
			continue;

4559 4560
		new->entries[j] = thresholds->primary->entries[i];
		if (new->entries[j].threshold < usage) {
4561
			/*
4562
			 * new->current_threshold will not be used
4563 4564 4565
			 * until rcu_assign_pointer(), so it's safe to increment
			 * it here.
			 */
4566
			++new->current_threshold;
4567 4568 4569 4570
		}
		j++;
	}

4571
swap_buffers:
4572 4573 4574
	/* Swap primary and spare array */
	thresholds->spare = thresholds->primary;
	rcu_assign_pointer(thresholds->primary, new);
4575

4576
	/* To be sure that nobody uses thresholds */
4577 4578 4579 4580
	synchronize_rcu();

	mutex_unlock(&memcg->thresholds_lock);
}
4581

K
KAMEZAWA Hiroyuki 已提交
4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593
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;

4594
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4595 4596 4597 4598 4599

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

	/* already in OOM ? */
4600
	if (atomic_read(&memcg->under_oom))
K
KAMEZAWA Hiroyuki 已提交
4601
		eventfd_signal(eventfd, 1);
4602
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4603 4604 4605 4606

	return 0;
}

4607
static void mem_cgroup_oom_unregister_event(struct cgroup *cgrp,
K
KAMEZAWA Hiroyuki 已提交
4608 4609
	struct cftype *cft, struct eventfd_ctx *eventfd)
{
4610
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
K
KAMEZAWA Hiroyuki 已提交
4611 4612 4613 4614 4615
	struct mem_cgroup_eventfd_list *ev, *tmp;
	int type = MEMFILE_TYPE(cft->private);

	BUG_ON(type != _OOM_TYPE);

4616
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4617

4618
	list_for_each_entry_safe(ev, tmp, &memcg->oom_notify, list) {
K
KAMEZAWA Hiroyuki 已提交
4619 4620 4621 4622 4623 4624
		if (ev->eventfd == eventfd) {
			list_del(&ev->list);
			kfree(ev);
		}
	}

4625
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4626 4627
}

4628 4629 4630
static int mem_cgroup_oom_control_read(struct cgroup *cgrp,
	struct cftype *cft,  struct cgroup_map_cb *cb)
{
4631
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4632

4633
	cb->fill(cb, "oom_kill_disable", memcg->oom_kill_disable);
4634

4635
	if (atomic_read(&memcg->under_oom))
4636 4637 4638 4639 4640 4641 4642 4643 4644
		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)
{
4645
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656
	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) ||
4657
	    (memcg->use_hierarchy && !list_empty(&cgrp->children))) {
4658 4659 4660
		cgroup_unlock();
		return -EINVAL;
	}
4661
	memcg->oom_kill_disable = val;
4662
	if (!val)
4663
		memcg_oom_recover(memcg);
4664 4665 4666 4667
	cgroup_unlock();
	return 0;
}

4668 4669 4670 4671 4672 4673 4674 4675 4676 4677 4678 4679 4680 4681 4682 4683
#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 */

4684 4685 4686
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
static int register_kmem_files(struct cgroup *cont, struct cgroup_subsys *ss)
{
G
Glauber Costa 已提交
4687 4688 4689 4690 4691 4692 4693
	/*
	 * Part of this would be better living in a separate allocation
	 * function, leaving us with just the cgroup tree population work.
	 * We, however, depend on state such as network's proto_list that
	 * is only initialized after cgroup creation. I found the less
	 * cumbersome way to deal with it to defer it all to populate time
	 */
4694
	return mem_cgroup_sockets_init(cont, ss);
4695 4696
};

G
Glauber Costa 已提交
4697 4698 4699 4700 4701
static void kmem_cgroup_destroy(struct cgroup_subsys *ss,
				struct cgroup *cont)
{
	mem_cgroup_sockets_destroy(cont, ss);
}
4702 4703 4704 4705 4706
#else
static int register_kmem_files(struct cgroup *cont, struct cgroup_subsys *ss)
{
	return 0;
}
G
Glauber Costa 已提交
4707 4708 4709 4710 4711

static void kmem_cgroup_destroy(struct cgroup_subsys *ss,
				struct cgroup *cont)
{
}
4712 4713
#endif

B
Balbir Singh 已提交
4714 4715
static struct cftype mem_cgroup_files[] = {
	{
4716
		.name = "usage_in_bytes",
4717
		.private = MEMFILE_PRIVATE(_MEM, RES_USAGE),
4718
		.read_u64 = mem_cgroup_read,
K
KAMEZAWA Hiroyuki 已提交
4719 4720
		.register_event = mem_cgroup_usage_register_event,
		.unregister_event = mem_cgroup_usage_unregister_event,
B
Balbir Singh 已提交
4721
	},
4722 4723
	{
		.name = "max_usage_in_bytes",
4724
		.private = MEMFILE_PRIVATE(_MEM, RES_MAX_USAGE),
4725
		.trigger = mem_cgroup_reset,
4726 4727
		.read_u64 = mem_cgroup_read,
	},
B
Balbir Singh 已提交
4728
	{
4729
		.name = "limit_in_bytes",
4730
		.private = MEMFILE_PRIVATE(_MEM, RES_LIMIT),
4731
		.write_string = mem_cgroup_write,
4732
		.read_u64 = mem_cgroup_read,
B
Balbir Singh 已提交
4733
	},
4734 4735 4736 4737 4738 4739
	{
		.name = "soft_limit_in_bytes",
		.private = MEMFILE_PRIVATE(_MEM, RES_SOFT_LIMIT),
		.write_string = mem_cgroup_write,
		.read_u64 = mem_cgroup_read,
	},
B
Balbir Singh 已提交
4740 4741
	{
		.name = "failcnt",
4742
		.private = MEMFILE_PRIVATE(_MEM, RES_FAILCNT),
4743
		.trigger = mem_cgroup_reset,
4744
		.read_u64 = mem_cgroup_read,
B
Balbir Singh 已提交
4745
	},
4746 4747
	{
		.name = "stat",
4748
		.read_map = mem_control_stat_show,
4749
	},
4750 4751 4752 4753
	{
		.name = "force_empty",
		.trigger = mem_cgroup_force_empty_write,
	},
4754 4755 4756 4757 4758
	{
		.name = "use_hierarchy",
		.write_u64 = mem_cgroup_hierarchy_write,
		.read_u64 = mem_cgroup_hierarchy_read,
	},
K
KOSAKI Motohiro 已提交
4759 4760 4761 4762 4763
	{
		.name = "swappiness",
		.read_u64 = mem_cgroup_swappiness_read,
		.write_u64 = mem_cgroup_swappiness_write,
	},
4764 4765 4766 4767 4768
	{
		.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
4769 4770
	{
		.name = "oom_control",
4771 4772
		.read_map = mem_cgroup_oom_control_read,
		.write_u64 = mem_cgroup_oom_control_write,
K
KAMEZAWA Hiroyuki 已提交
4773 4774 4775 4776
		.register_event = mem_cgroup_oom_register_event,
		.unregister_event = mem_cgroup_oom_unregister_event,
		.private = MEMFILE_PRIVATE(_OOM_TYPE, OOM_CONTROL),
	},
4777 4778 4779 4780
#ifdef CONFIG_NUMA
	{
		.name = "numa_stat",
		.open = mem_control_numa_stat_open,
4781
		.mode = S_IRUGO,
4782 4783
	},
#endif
B
Balbir Singh 已提交
4784 4785
};

4786 4787 4788 4789 4790 4791
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
static struct cftype memsw_cgroup_files[] = {
	{
		.name = "memsw.usage_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_USAGE),
		.read_u64 = mem_cgroup_read,
K
KAMEZAWA Hiroyuki 已提交
4792 4793
		.register_event = mem_cgroup_usage_register_event,
		.unregister_event = mem_cgroup_usage_unregister_event,
4794 4795 4796 4797 4798 4799 4800 4801 4802 4803 4804 4805 4806 4807 4808 4809 4810 4811 4812 4813 4814 4815 4816 4817 4818 4819 4820 4821 4822 4823 4824 4825 4826 4827 4828
	},
	{
		.name = "memsw.max_usage_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_MAX_USAGE),
		.trigger = mem_cgroup_reset,
		.read_u64 = mem_cgroup_read,
	},
	{
		.name = "memsw.limit_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_LIMIT),
		.write_string = mem_cgroup_write,
		.read_u64 = mem_cgroup_read,
	},
	{
		.name = "memsw.failcnt",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_FAILCNT),
		.trigger = mem_cgroup_reset,
		.read_u64 = mem_cgroup_read,
	},
};

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

4829
static int alloc_mem_cgroup_per_zone_info(struct mem_cgroup *memcg, int node)
4830 4831
{
	struct mem_cgroup_per_node *pn;
4832
	struct mem_cgroup_per_zone *mz;
4833
	enum lru_list l;
4834
	int zone, tmp = node;
4835 4836 4837 4838 4839 4840 4841 4842
	/*
	 * 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.
	 */
4843 4844
	if (!node_state(node, N_NORMAL_MEMORY))
		tmp = -1;
4845
	pn = kzalloc_node(sizeof(*pn), GFP_KERNEL, tmp);
4846 4847
	if (!pn)
		return 1;
4848 4849 4850

	for (zone = 0; zone < MAX_NR_ZONES; zone++) {
		mz = &pn->zoneinfo[zone];
4851
		for_each_lru(l)
4852
			INIT_LIST_HEAD(&mz->lruvec.lists[l]);
4853
		mz->usage_in_excess = 0;
4854
		mz->on_tree = false;
4855
		mz->mem = memcg;
4856
	}
4857
	memcg->info.nodeinfo[node] = pn;
4858 4859 4860
	return 0;
}

4861
static void free_mem_cgroup_per_zone_info(struct mem_cgroup *memcg, int node)
4862
{
4863
	kfree(memcg->info.nodeinfo[node]);
4864 4865
}

4866 4867 4868
static struct mem_cgroup *mem_cgroup_alloc(void)
{
	struct mem_cgroup *mem;
4869
	int size = sizeof(struct mem_cgroup);
4870

4871
	/* Can be very big if MAX_NUMNODES is very big */
4872
	if (size < PAGE_SIZE)
4873
		mem = kzalloc(size, GFP_KERNEL);
4874
	else
4875
		mem = vzalloc(size);
4876

4877 4878 4879
	if (!mem)
		return NULL;

4880
	mem->stat = alloc_percpu(struct mem_cgroup_stat_cpu);
4881 4882
	if (!mem->stat)
		goto out_free;
4883
	spin_lock_init(&mem->pcp_counter_lock);
4884
	return mem;
4885 4886 4887 4888 4889 4890 4891

out_free:
	if (size < PAGE_SIZE)
		kfree(mem);
	else
		vfree(mem);
	return NULL;
4892 4893
}

4894 4895 4896 4897 4898 4899 4900 4901 4902 4903 4904
/*
 * 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.
 */

4905
static void __mem_cgroup_free(struct mem_cgroup *memcg)
4906
{
K
KAMEZAWA Hiroyuki 已提交
4907 4908
	int node;

4909 4910
	mem_cgroup_remove_from_trees(memcg);
	free_css_id(&mem_cgroup_subsys, &memcg->css);
K
KAMEZAWA Hiroyuki 已提交
4911

K
KAMEZAWA Hiroyuki 已提交
4912
	for_each_node_state(node, N_POSSIBLE)
4913
		free_mem_cgroup_per_zone_info(memcg, node);
K
KAMEZAWA Hiroyuki 已提交
4914

4915
	free_percpu(memcg->stat);
4916
	if (sizeof(struct mem_cgroup) < PAGE_SIZE)
4917
		kfree(memcg);
4918
	else
4919
		vfree(memcg);
4920 4921
}

4922
static void mem_cgroup_get(struct mem_cgroup *memcg)
4923
{
4924
	atomic_inc(&memcg->refcnt);
4925 4926
}

4927
static void __mem_cgroup_put(struct mem_cgroup *memcg, int count)
4928
{
4929 4930 4931
	if (atomic_sub_and_test(count, &memcg->refcnt)) {
		struct mem_cgroup *parent = parent_mem_cgroup(memcg);
		__mem_cgroup_free(memcg);
4932 4933 4934
		if (parent)
			mem_cgroup_put(parent);
	}
4935 4936
}

4937
static void mem_cgroup_put(struct mem_cgroup *memcg)
4938
{
4939
	__mem_cgroup_put(memcg, 1);
4940 4941
}

4942 4943 4944
/*
 * Returns the parent mem_cgroup in memcgroup hierarchy with hierarchy enabled.
 */
G
Glauber Costa 已提交
4945
struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
4946
{
4947
	if (!memcg->res.parent)
4948
		return NULL;
4949
	return mem_cgroup_from_res_counter(memcg->res.parent, res);
4950
}
G
Glauber Costa 已提交
4951
EXPORT_SYMBOL(parent_mem_cgroup);
4952

4953 4954 4955
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
static void __init enable_swap_cgroup(void)
{
4956
	if (!mem_cgroup_disabled() && really_do_swap_account)
4957 4958 4959 4960 4961 4962 4963 4964
		do_swap_account = 1;
}
#else
static void __init enable_swap_cgroup(void)
{
}
#endif

4965 4966 4967 4968 4969 4970 4971 4972 4973 4974 4975 4976 4977 4978 4979 4980 4981 4982 4983 4984 4985 4986 4987 4988 4989
static int mem_cgroup_soft_limit_tree_init(void)
{
	struct mem_cgroup_tree_per_node *rtpn;
	struct mem_cgroup_tree_per_zone *rtpz;
	int tmp, node, zone;

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

		soft_limit_tree.rb_tree_per_node[node] = rtpn;

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

L
Li Zefan 已提交
4990
static struct cgroup_subsys_state * __ref
B
Balbir Singh 已提交
4991 4992
mem_cgroup_create(struct cgroup_subsys *ss, struct cgroup *cont)
{
4993
	struct mem_cgroup *memcg, *parent;
K
KAMEZAWA Hiroyuki 已提交
4994
	long error = -ENOMEM;
4995
	int node;
B
Balbir Singh 已提交
4996

4997 4998
	memcg = mem_cgroup_alloc();
	if (!memcg)
K
KAMEZAWA Hiroyuki 已提交
4999
		return ERR_PTR(error);
5000

5001
	for_each_node_state(node, N_POSSIBLE)
5002
		if (alloc_mem_cgroup_per_zone_info(memcg, node))
5003
			goto free_out;
5004

5005
	/* root ? */
5006
	if (cont->parent == NULL) {
5007
		int cpu;
5008
		enable_swap_cgroup();
5009
		parent = NULL;
5010 5011
		if (mem_cgroup_soft_limit_tree_init())
			goto free_out;
5012
		root_mem_cgroup = memcg;
5013 5014 5015 5016 5017
		for_each_possible_cpu(cpu) {
			struct memcg_stock_pcp *stock =
						&per_cpu(memcg_stock, cpu);
			INIT_WORK(&stock->work, drain_local_stock);
		}
5018
		hotcpu_notifier(memcg_cpu_hotplug_callback, 0);
5019
	} else {
5020
		parent = mem_cgroup_from_cont(cont->parent);
5021 5022
		memcg->use_hierarchy = parent->use_hierarchy;
		memcg->oom_kill_disable = parent->oom_kill_disable;
5023
	}
5024

5025
	if (parent && parent->use_hierarchy) {
5026 5027
		res_counter_init(&memcg->res, &parent->res);
		res_counter_init(&memcg->memsw, &parent->memsw);
5028 5029 5030 5031 5032 5033 5034
		/*
		 * 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);
5035
	} else {
5036 5037
		res_counter_init(&memcg->res, NULL);
		res_counter_init(&memcg->memsw, NULL);
5038
	}
5039 5040
	memcg->last_scanned_node = MAX_NUMNODES;
	INIT_LIST_HEAD(&memcg->oom_notify);
5041

K
KOSAKI Motohiro 已提交
5042
	if (parent)
5043 5044 5045 5046 5047
		memcg->swappiness = mem_cgroup_swappiness(parent);
	atomic_set(&memcg->refcnt, 1);
	memcg->move_charge_at_immigrate = 0;
	mutex_init(&memcg->thresholds_lock);
	return &memcg->css;
5048
free_out:
5049
	__mem_cgroup_free(memcg);
K
KAMEZAWA Hiroyuki 已提交
5050
	return ERR_PTR(error);
B
Balbir Singh 已提交
5051 5052
}

5053
static int mem_cgroup_pre_destroy(struct cgroup_subsys *ss,
5054 5055
					struct cgroup *cont)
{
5056
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
5057

5058
	return mem_cgroup_force_empty(memcg, false);
5059 5060
}

B
Balbir Singh 已提交
5061 5062 5063
static void mem_cgroup_destroy(struct cgroup_subsys *ss,
				struct cgroup *cont)
{
5064
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
5065

G
Glauber Costa 已提交
5066 5067
	kmem_cgroup_destroy(ss, cont);

5068
	mem_cgroup_put(memcg);
B
Balbir Singh 已提交
5069 5070 5071 5072 5073
}

static int mem_cgroup_populate(struct cgroup_subsys *ss,
				struct cgroup *cont)
{
5074 5075 5076 5077 5078 5079 5080
	int ret;

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

	if (!ret)
		ret = register_memsw_files(cont, ss);
5081 5082 5083 5084

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

5085
	return ret;
B
Balbir Singh 已提交
5086 5087
}

5088
#ifdef CONFIG_MMU
5089
/* Handlers for move charge at task migration. */
5090 5091
#define PRECHARGE_COUNT_AT_ONCE	256
static int mem_cgroup_do_precharge(unsigned long count)
5092
{
5093 5094
	int ret = 0;
	int batch_count = PRECHARGE_COUNT_AT_ONCE;
5095
	struct mem_cgroup *memcg = mc.to;
5096

5097
	if (mem_cgroup_is_root(memcg)) {
5098 5099 5100 5101 5102 5103 5104 5105
		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;
		/*
5106
		 * "memcg" cannot be under rmdir() because we've already checked
5107 5108 5109 5110
		 * 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().
		 */
5111
		if (res_counter_charge(&memcg->res, PAGE_SIZE * count, &dummy))
5112
			goto one_by_one;
5113
		if (do_swap_account && res_counter_charge(&memcg->memsw,
5114
						PAGE_SIZE * count, &dummy)) {
5115
			res_counter_uncharge(&memcg->res, PAGE_SIZE * count);
5116 5117 5118 5119 5120 5121 5122 5123 5124 5125 5126 5127 5128 5129 5130 5131
			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();
		}
5132 5133 5134
		ret = __mem_cgroup_try_charge(NULL,
					GFP_KERNEL, 1, &memcg, false);
		if (ret || !memcg)
5135 5136 5137 5138
			/* mem_cgroup_clear_mc() will do uncharge later */
			return -ENOMEM;
		mc.precharge++;
	}
5139 5140 5141 5142 5143 5144 5145 5146
	return ret;
}

/**
 * is_target_pte_for_mc - check a pte whether it is valid for move charge
 * @vma: the vma the pte to be checked belongs
 * @addr: the address corresponding to the pte to be checked
 * @ptent: the pte to be checked
5147
 * @target: the pointer the target page or swap ent will be stored(can be NULL)
5148 5149 5150 5151 5152 5153
 *
 * 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).
5154 5155 5156
 *   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.
5157 5158 5159 5160 5161
 *
 * Called with pte lock held.
 */
union mc_target {
	struct page	*page;
5162
	swp_entry_t	ent;
5163 5164 5165 5166 5167
};

enum mc_target_type {
	MC_TARGET_NONE,	/* not used */
	MC_TARGET_PAGE,
5168
	MC_TARGET_SWAP,
5169 5170
};

D
Daisuke Nishimura 已提交
5171 5172
static struct page *mc_handle_present_pte(struct vm_area_struct *vma,
						unsigned long addr, pte_t ptent)
5173
{
D
Daisuke Nishimura 已提交
5174
	struct page *page = vm_normal_page(vma, addr, ptent);
5175

D
Daisuke Nishimura 已提交
5176 5177 5178 5179 5180 5181
	if (!page || !page_mapped(page))
		return NULL;
	if (PageAnon(page)) {
		/* we don't move shared anon */
		if (!move_anon() || page_mapcount(page) > 2)
			return NULL;
5182 5183
	} else if (!move_file())
		/* we ignore mapcount for file pages */
D
Daisuke Nishimura 已提交
5184 5185 5186 5187 5188 5189 5190 5191 5192 5193 5194 5195 5196 5197 5198 5199 5200 5201
		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 */
5202 5203
		if (page)
			put_page(page);
D
Daisuke Nishimura 已提交
5204
		return NULL;
5205
	}
D
Daisuke Nishimura 已提交
5206 5207 5208 5209 5210 5211
	if (do_swap_account)
		entry->val = ent.val;

	return page;
}

5212 5213 5214 5215 5216 5217 5218 5219 5220 5221 5222 5223 5224 5225 5226 5227 5228 5229 5230 5231 5232
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). */
5233 5234 5235 5236 5237 5238
	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);
5239
		if (do_swap_account)
5240 5241
			*entry = swap;
		page = find_get_page(&swapper_space, swap.val);
5242
	}
5243
#endif
5244 5245 5246
	return page;
}

D
Daisuke Nishimura 已提交
5247 5248 5249 5250 5251 5252 5253 5254 5255 5256 5257 5258
static int is_target_pte_for_mc(struct vm_area_struct *vma,
		unsigned long addr, pte_t ptent, union mc_target *target)
{
	struct page *page = NULL;
	struct page_cgroup *pc;
	int ret = 0;
	swp_entry_t ent = { .val = 0 };

	if (pte_present(ptent))
		page = mc_handle_present_pte(vma, addr, ptent);
	else if (is_swap_pte(ptent))
		page = mc_handle_swap_pte(vma, addr, ptent, &ent);
5259 5260
	else if (pte_none(ptent) || pte_file(ptent))
		page = mc_handle_file_pte(vma, addr, ptent, &ent);
D
Daisuke Nishimura 已提交
5261 5262 5263

	if (!page && !ent.val)
		return 0;
5264 5265 5266 5267 5268 5269 5270 5271 5272 5273 5274 5275 5276 5277 5278
	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
5279 5280
	/* There is a swap entry and a page doesn't exist or isn't charged */
	if (ent.val && !ret &&
5281 5282 5283 5284
			css_id(&mc.from->css) == lookup_swap_cgroup(ent)) {
		ret = MC_TARGET_SWAP;
		if (target)
			target->ent = ent;
5285 5286 5287 5288 5289 5290 5291 5292 5293 5294 5295 5296
	}
	return ret;
}

static int mem_cgroup_count_precharge_pte_range(pmd_t *pmd,
					unsigned long addr, unsigned long end,
					struct mm_walk *walk)
{
	struct vm_area_struct *vma = walk->private;
	pte_t *pte;
	spinlock_t *ptl;

5297 5298
	split_huge_page_pmd(walk->mm, pmd);

5299 5300 5301 5302 5303 5304 5305
	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
	for (; addr != end; pte++, addr += PAGE_SIZE)
		if (is_target_pte_for_mc(vma, addr, *pte, NULL))
			mc.precharge++;	/* increment precharge temporarily */
	pte_unmap_unlock(pte - 1, ptl);
	cond_resched();

5306 5307 5308
	return 0;
}

5309 5310 5311 5312 5313
static unsigned long mem_cgroup_count_precharge(struct mm_struct *mm)
{
	unsigned long precharge;
	struct vm_area_struct *vma;

5314
	down_read(&mm->mmap_sem);
5315 5316 5317 5318 5319 5320 5321 5322 5323 5324 5325
	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);
	}
5326
	up_read(&mm->mmap_sem);
5327 5328 5329 5330 5331 5332 5333 5334 5335

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

	return precharge;
}

static int mem_cgroup_precharge_mc(struct mm_struct *mm)
{
5336 5337 5338 5339 5340
	unsigned long precharge = mem_cgroup_count_precharge(mm);

	VM_BUG_ON(mc.moving_task);
	mc.moving_task = current;
	return mem_cgroup_do_precharge(precharge);
5341 5342
}

5343 5344
/* cancels all extra charges on mc.from and mc.to, and wakes up all waiters. */
static void __mem_cgroup_clear_mc(void)
5345
{
5346 5347 5348
	struct mem_cgroup *from = mc.from;
	struct mem_cgroup *to = mc.to;

5349
	/* we must uncharge all the leftover precharges from mc.to */
5350 5351 5352 5353 5354 5355 5356 5357 5358 5359 5360
	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;
5361
	}
5362 5363 5364 5365 5366 5367 5368 5369 5370 5371 5372 5373 5374 5375 5376 5377 5378 5379 5380
	/* 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;
	}
5381 5382 5383 5384 5385 5386 5387 5388 5389 5390 5391 5392 5393 5394 5395
	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();
5396
	spin_lock(&mc.lock);
5397 5398
	mc.from = NULL;
	mc.to = NULL;
5399
	spin_unlock(&mc.lock);
5400
	mem_cgroup_end_move(from);
5401 5402
}

5403 5404
static int mem_cgroup_can_attach(struct cgroup_subsys *ss,
				struct cgroup *cgroup,
5405
				struct cgroup_taskset *tset)
5406
{
5407
	struct task_struct *p = cgroup_taskset_first(tset);
5408
	int ret = 0;
5409
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgroup);
5410

5411
	if (memcg->move_charge_at_immigrate) {
5412 5413 5414
		struct mm_struct *mm;
		struct mem_cgroup *from = mem_cgroup_from_task(p);

5415
		VM_BUG_ON(from == memcg);
5416 5417 5418 5419 5420

		mm = get_task_mm(p);
		if (!mm)
			return 0;
		/* We move charges only when we move a owner of the mm */
5421 5422 5423 5424
		if (mm->owner == p) {
			VM_BUG_ON(mc.from);
			VM_BUG_ON(mc.to);
			VM_BUG_ON(mc.precharge);
5425
			VM_BUG_ON(mc.moved_charge);
5426
			VM_BUG_ON(mc.moved_swap);
5427
			mem_cgroup_start_move(from);
5428
			spin_lock(&mc.lock);
5429
			mc.from = from;
5430
			mc.to = memcg;
5431
			spin_unlock(&mc.lock);
5432
			/* We set mc.moving_task later */
5433 5434 5435 5436

			ret = mem_cgroup_precharge_mc(mm);
			if (ret)
				mem_cgroup_clear_mc();
5437 5438
		}
		mmput(mm);
5439 5440 5441 5442 5443 5444
	}
	return ret;
}

static void mem_cgroup_cancel_attach(struct cgroup_subsys *ss,
				struct cgroup *cgroup,
5445
				struct cgroup_taskset *tset)
5446
{
5447
	mem_cgroup_clear_mc();
5448 5449
}

5450 5451 5452
static int mem_cgroup_move_charge_pte_range(pmd_t *pmd,
				unsigned long addr, unsigned long end,
				struct mm_walk *walk)
5453
{
5454 5455 5456 5457 5458
	int ret = 0;
	struct vm_area_struct *vma = walk->private;
	pte_t *pte;
	spinlock_t *ptl;

5459
	split_huge_page_pmd(walk->mm, pmd);
5460 5461 5462 5463 5464 5465 5466 5467
retry:
	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
	for (; addr != end; addr += PAGE_SIZE) {
		pte_t ptent = *(pte++);
		union mc_target target;
		int type;
		struct page *page;
		struct page_cgroup *pc;
5468
		swp_entry_t ent;
5469 5470 5471 5472 5473 5474 5475 5476 5477 5478 5479

		if (!mc.precharge)
			break;

		type = is_target_pte_for_mc(vma, addr, ptent, &target);
		switch (type) {
		case MC_TARGET_PAGE:
			page = target.page;
			if (isolate_lru_page(page))
				goto put;
			pc = lookup_page_cgroup(page);
5480 5481
			if (!mem_cgroup_move_account(page, 1, pc,
						     mc.from, mc.to, false)) {
5482
				mc.precharge--;
5483 5484
				/* we uncharge from mc.from later. */
				mc.moved_charge++;
5485 5486 5487 5488 5489
			}
			putback_lru_page(page);
put:			/* is_target_pte_for_mc() gets the page */
			put_page(page);
			break;
5490 5491
		case MC_TARGET_SWAP:
			ent = target.ent;
5492 5493
			if (!mem_cgroup_move_swap_account(ent,
						mc.from, mc.to, false)) {
5494
				mc.precharge--;
5495 5496 5497
				/* we fixup refcnts and charges later. */
				mc.moved_swap++;
			}
5498
			break;
5499 5500 5501 5502 5503 5504 5505 5506 5507 5508 5509 5510 5511 5512
		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.
		 */
5513
		ret = mem_cgroup_do_precharge(1);
5514 5515 5516 5517 5518 5519 5520 5521 5522 5523 5524 5525
		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();
5526 5527 5528 5529 5530 5531 5532 5533 5534 5535 5536 5537 5538
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;
	}
5539 5540 5541 5542 5543 5544 5545 5546 5547 5548 5549 5550 5551 5552 5553 5554 5555 5556
	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;
	}
5557
	up_read(&mm->mmap_sem);
5558 5559
}

B
Balbir Singh 已提交
5560 5561
static void mem_cgroup_move_task(struct cgroup_subsys *ss,
				struct cgroup *cont,
5562
				struct cgroup_taskset *tset)
B
Balbir Singh 已提交
5563
{
5564
	struct task_struct *p = cgroup_taskset_first(tset);
5565
	struct mm_struct *mm = get_task_mm(p);
5566 5567

	if (mm) {
5568 5569 5570
		if (mc.to)
			mem_cgroup_move_charge(mm);
		put_swap_token(mm);
5571 5572
		mmput(mm);
	}
5573 5574
	if (mc.to)
		mem_cgroup_clear_mc();
B
Balbir Singh 已提交
5575
}
5576 5577 5578
#else	/* !CONFIG_MMU */
static int mem_cgroup_can_attach(struct cgroup_subsys *ss,
				struct cgroup *cgroup,
5579
				struct cgroup_taskset *tset)
5580 5581 5582 5583 5584
{
	return 0;
}
static void mem_cgroup_cancel_attach(struct cgroup_subsys *ss,
				struct cgroup *cgroup,
5585
				struct cgroup_taskset *tset)
5586 5587 5588 5589
{
}
static void mem_cgroup_move_task(struct cgroup_subsys *ss,
				struct cgroup *cont,
5590
				struct cgroup_taskset *tset)
5591 5592 5593
{
}
#endif
B
Balbir Singh 已提交
5594

B
Balbir Singh 已提交
5595 5596 5597 5598
struct cgroup_subsys mem_cgroup_subsys = {
	.name = "memory",
	.subsys_id = mem_cgroup_subsys_id,
	.create = mem_cgroup_create,
5599
	.pre_destroy = mem_cgroup_pre_destroy,
B
Balbir Singh 已提交
5600 5601
	.destroy = mem_cgroup_destroy,
	.populate = mem_cgroup_populate,
5602 5603
	.can_attach = mem_cgroup_can_attach,
	.cancel_attach = mem_cgroup_cancel_attach,
B
Balbir Singh 已提交
5604
	.attach = mem_cgroup_move_task,
5605
	.early_init = 0,
K
KAMEZAWA Hiroyuki 已提交
5606
	.use_id = 1,
B
Balbir Singh 已提交
5607
};
5608 5609

#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
5610 5611 5612
static int __init enable_swap_account(char *s)
{
	/* consider enabled if no parameter or 1 is given */
5613
	if (!strcmp(s, "1"))
5614
		really_do_swap_account = 1;
5615
	else if (!strcmp(s, "0"))
5616 5617 5618
		really_do_swap_account = 0;
	return 1;
}
5619
__setup("swapaccount=", enable_swap_account);
5620 5621

#endif