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 {
137 138 139
	/*
	 * spin_lock to protect the per cgroup LRU
	 */
140 141
	struct list_head	lists[NR_LRU_LISTS];
	unsigned long		count[NR_LRU_LISTS];
K
KOSAKI Motohiro 已提交
142

143 144
	struct mem_cgroup_reclaim_iter reclaim_iter[DEF_PRIORITY + 1];

K
KOSAKI Motohiro 已提交
145
	struct zone_reclaim_stat reclaim_stat;
146 147 148 149
	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;
150 151
	struct mem_cgroup	*mem;		/* Back pointer, we cannot */
						/* use container_of	   */
152 153 154 155 156 157 158 159 160 161 162 163
};
/* 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];
};

164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183
/*
 * 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;

184 185 186 187 188
struct mem_cgroup_threshold {
	struct eventfd_ctx *eventfd;
	u64 threshold;
};

K
KAMEZAWA Hiroyuki 已提交
189
/* For threshold */
190 191
struct mem_cgroup_threshold_ary {
	/* An array index points to threshold just below usage. */
192
	int current_threshold;
193 194 195 196 197
	/* Size of entries[] */
	unsigned int size;
	/* Array of thresholds */
	struct mem_cgroup_threshold entries[0];
};
198 199 200 201 202 203 204 205 206 207 208 209

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
210 211 212 213 214
/* for OOM */
struct mem_cgroup_eventfd_list {
	struct list_head list;
	struct eventfd_ctx *eventfd;
};
215

216 217
static void mem_cgroup_threshold(struct mem_cgroup *memcg);
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg);
218

B
Balbir Singh 已提交
219 220 221 222 223 224 225
/*
 * 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
226 227 228
 * 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
229 230 231 232 233 234 235
 */
struct mem_cgroup {
	struct cgroup_subsys_state css;
	/*
	 * the counter to account for memory usage
	 */
	struct res_counter res;
236 237 238 239
	/*
	 * the counter to account for mem+swap usage.
	 */
	struct res_counter memsw;
240 241 242 243
	/*
	 * Per cgroup active and inactive list, similar to the
	 * per zone LRU lists.
	 */
244
	struct mem_cgroup_lru_info info;
245 246 247
	int last_scanned_node;
#if MAX_NUMNODES > 1
	nodemask_t	scan_nodes;
248 249
	atomic_t	numainfo_events;
	atomic_t	numainfo_updating;
250
#endif
251 252 253 254
	/*
	 * Should the accounting and control be hierarchical, per subtree?
	 */
	bool use_hierarchy;
255 256 257 258

	bool		oom_lock;
	atomic_t	under_oom;

259
	atomic_t	refcnt;
260

261
	int	swappiness;
262 263
	/* OOM-Killer disable */
	int		oom_kill_disable;
K
KOSAKI Motohiro 已提交
264

265 266 267
	/* set when res.limit == memsw.limit */
	bool		memsw_is_minimum;

268 269 270 271
	/* protect arrays of thresholds */
	struct mutex thresholds_lock;

	/* thresholds for memory usage. RCU-protected */
272
	struct mem_cgroup_thresholds thresholds;
273

274
	/* thresholds for mem+swap usage. RCU-protected */
275
	struct mem_cgroup_thresholds memsw_thresholds;
276

K
KAMEZAWA Hiroyuki 已提交
277 278
	/* For oom notifier event fd */
	struct list_head oom_notify;
279

280 281 282 283 284
	/*
	 * 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;
285
	/*
286
	 * percpu counter.
287
	 */
288
	struct mem_cgroup_stat_cpu *stat;
289 290 291 292 293 294
	/*
	 * 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
295 296 297 298

#ifdef CONFIG_INET
	struct tcp_memcontrol tcp_mem;
#endif
B
Balbir Singh 已提交
299 300
};

301 302 303 304 305 306
/* 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 {
307
	MOVE_CHARGE_TYPE_ANON,	/* private anonymous page and swap of it */
308
	MOVE_CHARGE_TYPE_FILE,	/* file page(including tmpfs) and swap of it */
309 310 311
	NR_MOVE_TYPE,
};

312 313
/* "mc" and its members are protected by cgroup_mutex */
static struct move_charge_struct {
314
	spinlock_t	  lock; /* for from, to */
315 316 317
	struct mem_cgroup *from;
	struct mem_cgroup *to;
	unsigned long precharge;
318
	unsigned long moved_charge;
319
	unsigned long moved_swap;
320 321 322
	struct task_struct *moving_task;	/* a task moving charges */
	wait_queue_head_t waitq;		/* a waitq for other context */
} mc = {
323
	.lock = __SPIN_LOCK_UNLOCKED(mc.lock),
324 325
	.waitq = __WAIT_QUEUE_HEAD_INITIALIZER(mc.waitq),
};
326

D
Daisuke Nishimura 已提交
327 328 329 330 331 332
static bool move_anon(void)
{
	return test_bit(MOVE_CHARGE_TYPE_ANON,
					&mc.to->move_charge_at_immigrate);
}

333 334 335 336 337 338
static bool move_file(void)
{
	return test_bit(MOVE_CHARGE_TYPE_FILE,
					&mc.to->move_charge_at_immigrate);
}

339 340 341 342 343 344 345
/*
 * 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)

346 347 348
enum charge_type {
	MEM_CGROUP_CHARGE_TYPE_CACHE = 0,
	MEM_CGROUP_CHARGE_TYPE_MAPPED,
349
	MEM_CGROUP_CHARGE_TYPE_SHMEM,	/* used by page migration of shmem */
350
	MEM_CGROUP_CHARGE_TYPE_FORCE,	/* used by force_empty */
K
KAMEZAWA Hiroyuki 已提交
351
	MEM_CGROUP_CHARGE_TYPE_SWAPOUT,	/* for accounting swapcache */
K
KAMEZAWA Hiroyuki 已提交
352
	MEM_CGROUP_CHARGE_TYPE_DROP,	/* a page was unused swap cache */
353 354 355
	NR_CHARGE_TYPE,
};

356
/* for encoding cft->private value on file */
357 358 359
#define _MEM			(0)
#define _MEMSWAP		(1)
#define _OOM_TYPE		(2)
360 361 362
#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
363 364
/* Used for OOM nofiier */
#define OOM_CONTROL		(0)
365

366 367 368 369 370 371 372
/*
 * 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)
373 374
#define MEM_CGROUP_RECLAIM_SOFT_BIT	0x2
#define MEM_CGROUP_RECLAIM_SOFT		(1 << MEM_CGROUP_RECLAIM_SOFT_BIT)
375

376 377
static void mem_cgroup_get(struct mem_cgroup *memcg);
static void mem_cgroup_put(struct mem_cgroup *memcg);
G
Glauber Costa 已提交
378 379 380 381 382

/* 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
383
#include <net/ip.h>
G
Glauber Costa 已提交
384 385 386 387 388 389 390 391 392

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);

393 394 395 396 397 398 399 400 401 402 403 404 405 406
		/* 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
407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426
		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
427 428 429 430 431 432 433 434 435

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
436 437 438
#endif /* CONFIG_INET */
#endif /* CONFIG_CGROUP_MEM_RES_CTLR_KMEM */

439
static void drain_all_stock_async(struct mem_cgroup *memcg);
440

441
static struct mem_cgroup_per_zone *
442
mem_cgroup_zoneinfo(struct mem_cgroup *memcg, int nid, int zid)
443
{
444
	return &memcg->info.nodeinfo[nid]->zoneinfo[zid];
445 446
}

447
struct cgroup_subsys_state *mem_cgroup_css(struct mem_cgroup *memcg)
448
{
449
	return &memcg->css;
450 451
}

452
static struct mem_cgroup_per_zone *
453
page_cgroup_zoneinfo(struct mem_cgroup *memcg, struct page *page)
454
{
455 456
	int nid = page_to_nid(page);
	int zid = page_zonenum(page);
457

458
	return mem_cgroup_zoneinfo(memcg, nid, zid);
459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476
}

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
477
__mem_cgroup_insert_exceeded(struct mem_cgroup *memcg,
478
				struct mem_cgroup_per_zone *mz,
479 480
				struct mem_cgroup_tree_per_zone *mctz,
				unsigned long long new_usage_in_excess)
481 482 483 484 485 486 487 488
{
	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;

489 490 491
	mz->usage_in_excess = new_usage_in_excess;
	if (!mz->usage_in_excess)
		return;
492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507
	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;
508 509 510
}

static void
511
__mem_cgroup_remove_exceeded(struct mem_cgroup *memcg,
512 513 514 515 516 517 518 519 520
				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;
}

521
static void
522
mem_cgroup_remove_exceeded(struct mem_cgroup *memcg,
523 524 525 526
				struct mem_cgroup_per_zone *mz,
				struct mem_cgroup_tree_per_zone *mctz)
{
	spin_lock(&mctz->lock);
527
	__mem_cgroup_remove_exceeded(memcg, mz, mctz);
528 529 530 531
	spin_unlock(&mctz->lock);
}


532
static void mem_cgroup_update_tree(struct mem_cgroup *memcg, struct page *page)
533
{
534
	unsigned long long excess;
535 536
	struct mem_cgroup_per_zone *mz;
	struct mem_cgroup_tree_per_zone *mctz;
537 538
	int nid = page_to_nid(page);
	int zid = page_zonenum(page);
539 540 541
	mctz = soft_limit_tree_from_page(page);

	/*
542 543
	 * Necessary to update all ancestors when hierarchy is used.
	 * because their event counter is not touched.
544
	 */
545 546 547
	for (; memcg; memcg = parent_mem_cgroup(memcg)) {
		mz = mem_cgroup_zoneinfo(memcg, nid, zid);
		excess = res_counter_soft_limit_excess(&memcg->res);
548 549 550 551
		/*
		 * We have to update the tree if mz is on RB-tree or
		 * mem is over its softlimit.
		 */
552
		if (excess || mz->on_tree) {
553 554 555
			spin_lock(&mctz->lock);
			/* if on-tree, remove it */
			if (mz->on_tree)
556
				__mem_cgroup_remove_exceeded(memcg, mz, mctz);
557
			/*
558 559
			 * Insert again. mz->usage_in_excess will be updated.
			 * If excess is 0, no tree ops.
560
			 */
561
			__mem_cgroup_insert_exceeded(memcg, mz, mctz, excess);
562 563
			spin_unlock(&mctz->lock);
		}
564 565 566
	}
}

567
static void mem_cgroup_remove_from_trees(struct mem_cgroup *memcg)
568 569 570 571 572 573 574
{
	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++) {
575
			mz = mem_cgroup_zoneinfo(memcg, node, zone);
576
			mctz = soft_limit_tree_node_zone(node, zone);
577
			mem_cgroup_remove_exceeded(memcg, mz, mctz);
578 579 580 581
		}
	}
}

582 583 584 585
static struct mem_cgroup_per_zone *
__mem_cgroup_largest_soft_limit_node(struct mem_cgroup_tree_per_zone *mctz)
{
	struct rb_node *rightmost = NULL;
586
	struct mem_cgroup_per_zone *mz;
587 588

retry:
589
	mz = NULL;
590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618
	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;
}

619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637
/*
 * 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.
 */
638
static long mem_cgroup_read_stat(struct mem_cgroup *memcg,
639
				 enum mem_cgroup_stat_index idx)
640
{
641
	long val = 0;
642 643
	int cpu;

644 645
	get_online_cpus();
	for_each_online_cpu(cpu)
646
		val += per_cpu(memcg->stat->count[idx], cpu);
647
#ifdef CONFIG_HOTPLUG_CPU
648 649 650
	spin_lock(&memcg->pcp_counter_lock);
	val += memcg->nocpu_base.count[idx];
	spin_unlock(&memcg->pcp_counter_lock);
651 652
#endif
	put_online_cpus();
653 654 655
	return val;
}

656
static void mem_cgroup_swap_statistics(struct mem_cgroup *memcg,
657 658 659
					 bool charge)
{
	int val = (charge) ? 1 : -1;
660
	this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_SWAPOUT], val);
661 662
}

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

668
void mem_cgroup_pgmajfault(struct mem_cgroup *memcg, int val)
669
{
670
	this_cpu_add(memcg->stat->events[MEM_CGROUP_EVENTS_PGMAJFAULT], val);
671 672
}

673
static unsigned long mem_cgroup_read_events(struct mem_cgroup *memcg,
674 675 676 677 678 679
					    enum mem_cgroup_events_index idx)
{
	unsigned long val = 0;
	int cpu;

	for_each_online_cpu(cpu)
680
		val += per_cpu(memcg->stat->events[idx], cpu);
681
#ifdef CONFIG_HOTPLUG_CPU
682 683 684
	spin_lock(&memcg->pcp_counter_lock);
	val += memcg->nocpu_base.events[idx];
	spin_unlock(&memcg->pcp_counter_lock);
685 686 687 688
#endif
	return val;
}

689
static void mem_cgroup_charge_statistics(struct mem_cgroup *memcg,
690
					 bool file, int nr_pages)
691
{
692 693
	preempt_disable();

694
	if (file)
695 696
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_CACHE],
				nr_pages);
697
	else
698 699
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_RSS],
				nr_pages);
700

701 702
	/* pagein of a big page is an event. So, ignore page size */
	if (nr_pages > 0)
703
		__this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGPGIN]);
704
	else {
705
		__this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGPGOUT]);
706 707
		nr_pages = -nr_pages; /* for event */
	}
708

709
	__this_cpu_add(memcg->stat->events[MEM_CGROUP_EVENTS_COUNT], nr_pages);
710

711
	preempt_enable();
712 713
}

714
unsigned long
715
mem_cgroup_zone_nr_lru_pages(struct mem_cgroup *memcg, int nid, int zid,
716
			unsigned int lru_mask)
717 718
{
	struct mem_cgroup_per_zone *mz;
719 720 721
	enum lru_list l;
	unsigned long ret = 0;

722
	mz = mem_cgroup_zoneinfo(memcg, nid, zid);
723 724 725 726 727 728 729 730 731

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

static unsigned long
732
mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
733 734
			int nid, unsigned int lru_mask)
{
735 736 737
	u64 total = 0;
	int zid;

738
	for (zid = 0; zid < MAX_NR_ZONES; zid++)
739 740
		total += mem_cgroup_zone_nr_lru_pages(memcg,
						nid, zid, lru_mask);
741

742 743
	return total;
}
744

745
static unsigned long mem_cgroup_nr_lru_pages(struct mem_cgroup *memcg,
746
			unsigned int lru_mask)
747
{
748
	int nid;
749 750
	u64 total = 0;

751
	for_each_node_state(nid, N_HIGH_MEMORY)
752
		total += mem_cgroup_node_nr_lru_pages(memcg, nid, lru_mask);
753
	return total;
754 755
}

756
static bool __memcg_event_check(struct mem_cgroup *memcg, int target)
757 758 759
{
	unsigned long val, next;

760 761
	val = __this_cpu_read(memcg->stat->events[MEM_CGROUP_EVENTS_COUNT]);
	next = __this_cpu_read(memcg->stat->targets[target]);
762 763 764 765
	/* from time_after() in jiffies.h */
	return ((long)next - (long)val < 0);
}

766
static void __mem_cgroup_target_update(struct mem_cgroup *memcg, int target)
767
{
768
	unsigned long val, next;
769

770
	val = __this_cpu_read(memcg->stat->events[MEM_CGROUP_EVENTS_COUNT]);
771

772 773 774 775 776 777 778
	switch (target) {
	case MEM_CGROUP_TARGET_THRESH:
		next = val + THRESHOLDS_EVENTS_TARGET;
		break;
	case MEM_CGROUP_TARGET_SOFTLIMIT:
		next = val + SOFTLIMIT_EVENTS_TARGET;
		break;
779 780 781
	case MEM_CGROUP_TARGET_NUMAINFO:
		next = val + NUMAINFO_EVENTS_TARGET;
		break;
782 783 784 785
	default:
		return;
	}

786
	__this_cpu_write(memcg->stat->targets[target], next);
787 788 789 790 791 792
}

/*
 * Check events in order.
 *
 */
793
static void memcg_check_events(struct mem_cgroup *memcg, struct page *page)
794
{
795
	preempt_disable();
796
	/* threshold event is triggered in finer grain than soft limit */
797 798 799 800
	if (unlikely(__memcg_event_check(memcg, MEM_CGROUP_TARGET_THRESH))) {
		mem_cgroup_threshold(memcg);
		__mem_cgroup_target_update(memcg, MEM_CGROUP_TARGET_THRESH);
		if (unlikely(__memcg_event_check(memcg,
801
			     MEM_CGROUP_TARGET_SOFTLIMIT))) {
802 803
			mem_cgroup_update_tree(memcg, page);
			__mem_cgroup_target_update(memcg,
804 805 806
						   MEM_CGROUP_TARGET_SOFTLIMIT);
		}
#if MAX_NUMNODES > 1
807
		if (unlikely(__memcg_event_check(memcg,
808
			MEM_CGROUP_TARGET_NUMAINFO))) {
809 810
			atomic_inc(&memcg->numainfo_events);
			__mem_cgroup_target_update(memcg,
811
				MEM_CGROUP_TARGET_NUMAINFO);
812
		}
813
#endif
814
	}
815
	preempt_enable();
816 817
}

G
Glauber Costa 已提交
818
struct mem_cgroup *mem_cgroup_from_cont(struct cgroup *cont)
B
Balbir Singh 已提交
819 820 821 822 823 824
{
	return container_of(cgroup_subsys_state(cont,
				mem_cgroup_subsys_id), struct mem_cgroup,
				css);
}

825
struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p)
826
{
827 828 829 830 831 832 833 834
	/*
	 * 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;

835 836 837 838
	return container_of(task_subsys_state(p, mem_cgroup_subsys_id),
				struct mem_cgroup, css);
}

839
struct mem_cgroup *try_get_mem_cgroup_from_mm(struct mm_struct *mm)
840
{
841
	struct mem_cgroup *memcg = NULL;
842 843 844

	if (!mm)
		return NULL;
845 846 847 848 849 850 851
	/*
	 * 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 {
852 853
		memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
		if (unlikely(!memcg))
854
			break;
855
	} while (!css_tryget(&memcg->css));
856
	rcu_read_unlock();
857
	return memcg;
858 859
}

860 861 862 863 864 865 866 867 868 869
struct mem_cgroup_reclaim_cookie {
	struct zone *zone;
	int priority;
	unsigned int generation;
};

static 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
870
{
871 872
	struct mem_cgroup *memcg = NULL;
	int id = 0;
873

874 875
	if (!root)
		root = root_mem_cgroup;
K
KAMEZAWA Hiroyuki 已提交
876

877 878
	if (prev && !reclaim)
		id = css_id(&prev->css);
K
KAMEZAWA Hiroyuki 已提交
879

880 881
	if (prev && prev != root)
		css_put(&prev->css);
K
KAMEZAWA Hiroyuki 已提交
882

883 884 885 886 887
	if (!root->use_hierarchy && root != root_mem_cgroup) {
		if (prev)
			return NULL;
		return root;
	}
K
KAMEZAWA Hiroyuki 已提交
888

889
	while (!memcg) {
890
		struct mem_cgroup_reclaim_iter *uninitialized_var(iter);
891
		struct cgroup_subsys_state *css;
892

893 894 895 896 897 898 899 900 901 902 903
		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
904

905 906 907 908 909 910 911 912
		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
913 914
		rcu_read_unlock();

915 916 917 918 919 920 921
		if (reclaim) {
			iter->position = id;
			if (!css)
				iter->generation++;
			else if (!prev && memcg)
				reclaim->generation = iter->generation;
		}
922 923 924 925 926

		if (prev && !css)
			return NULL;
	}
	return memcg;
K
KAMEZAWA Hiroyuki 已提交
927
}
K
KAMEZAWA Hiroyuki 已提交
928

929 930 931 932 933 934 935 936
static void mem_cgroup_iter_break(struct mem_cgroup *root,
				  struct mem_cgroup *prev)
{
	if (!root)
		root = root_mem_cgroup;
	if (prev && prev != root)
		css_put(&prev->css);
}
K
KAMEZAWA Hiroyuki 已提交
937

938 939 940 941 942 943
/*
 * 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)		\
944
	for (iter = mem_cgroup_iter(root, NULL, NULL);	\
945
	     iter != NULL;				\
946
	     iter = mem_cgroup_iter(root, iter, NULL))
947

948
#define for_each_mem_cgroup(iter)			\
949
	for (iter = mem_cgroup_iter(NULL, NULL, NULL);	\
950
	     iter != NULL;				\
951
	     iter = mem_cgroup_iter(NULL, iter, NULL))
K
KAMEZAWA Hiroyuki 已提交
952

953
static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
954
{
955
	return (memcg == root_mem_cgroup);
956 957
}

958 959
void mem_cgroup_count_vm_event(struct mm_struct *mm, enum vm_event_item idx)
{
960
	struct mem_cgroup *memcg;
961 962 963 964 965

	if (!mm)
		return;

	rcu_read_lock();
966 967
	memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
	if (unlikely(!memcg))
968 969 970 971
		goto out;

	switch (idx) {
	case PGMAJFAULT:
972
		mem_cgroup_pgmajfault(memcg, 1);
973 974
		break;
	case PGFAULT:
975
		mem_cgroup_pgfault(memcg, 1);
976 977 978 979 980 981 982 983 984
		break;
	default:
		BUG();
	}
out:
	rcu_read_unlock();
}
EXPORT_SYMBOL(mem_cgroup_count_vm_event);

K
KAMEZAWA Hiroyuki 已提交
985 986 987 988 989 990 991 992 993 994 995 996 997
/*
 * 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.
 */
998

K
KAMEZAWA Hiroyuki 已提交
999 1000 1001 1002
void mem_cgroup_del_lru_list(struct page *page, enum lru_list lru)
{
	struct page_cgroup *pc;
	struct mem_cgroup_per_zone *mz;
1003

1004
	if (mem_cgroup_disabled())
K
KAMEZAWA Hiroyuki 已提交
1005 1006 1007
		return;
	pc = lookup_page_cgroup(page);
	/* can happen while we handle swapcache. */
1008
	if (!TestClearPageCgroupAcctLRU(pc))
K
KAMEZAWA Hiroyuki 已提交
1009
		return;
1010
	VM_BUG_ON(!pc->mem_cgroup);
1011 1012 1013 1014
	/*
	 * We don't check PCG_USED bit. It's cleared when the "page" is finally
	 * removed from global LRU.
	 */
1015
	mz = page_cgroup_zoneinfo(pc->mem_cgroup, page);
1016 1017
	/* huge page split is done under lru_lock. so, we have no races. */
	MEM_CGROUP_ZSTAT(mz, lru) -= 1 << compound_order(page);
1018 1019 1020
	if (mem_cgroup_is_root(pc->mem_cgroup))
		return;
	VM_BUG_ON(list_empty(&pc->lru));
K
KAMEZAWA Hiroyuki 已提交
1021
	list_del_init(&pc->lru);
1022 1023
}

K
KAMEZAWA Hiroyuki 已提交
1024
void mem_cgroup_del_lru(struct page *page)
1025
{
K
KAMEZAWA Hiroyuki 已提交
1026 1027
	mem_cgroup_del_lru_list(page, page_lru(page));
}
1028

1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050
/*
 * Writeback is about to end against a page which has been marked for immediate
 * reclaim.  If it still appears to be reclaimable, move it to the tail of the
 * inactive list.
 */
void mem_cgroup_rotate_reclaimable_page(struct page *page)
{
	struct mem_cgroup_per_zone *mz;
	struct page_cgroup *pc;
	enum lru_list lru = page_lru(page);

	if (mem_cgroup_disabled())
		return;

	pc = lookup_page_cgroup(page);
	/* unused or root page is not rotated. */
	if (!PageCgroupUsed(pc))
		return;
	/* Ensure pc->mem_cgroup is visible after reading PCG_USED. */
	smp_rmb();
	if (mem_cgroup_is_root(pc->mem_cgroup))
		return;
1051
	mz = page_cgroup_zoneinfo(pc->mem_cgroup, page);
1052 1053 1054
	list_move_tail(&pc->lru, &mz->lists[lru]);
}

K
KAMEZAWA Hiroyuki 已提交
1055 1056 1057 1058
void mem_cgroup_rotate_lru_list(struct page *page, enum lru_list lru)
{
	struct mem_cgroup_per_zone *mz;
	struct page_cgroup *pc;
1059

1060
	if (mem_cgroup_disabled())
K
KAMEZAWA Hiroyuki 已提交
1061
		return;
1062

K
KAMEZAWA Hiroyuki 已提交
1063
	pc = lookup_page_cgroup(page);
1064
	/* unused or root page is not rotated. */
1065 1066 1067 1068 1069
	if (!PageCgroupUsed(pc))
		return;
	/* Ensure pc->mem_cgroup is visible after reading PCG_USED. */
	smp_rmb();
	if (mem_cgroup_is_root(pc->mem_cgroup))
K
KAMEZAWA Hiroyuki 已提交
1070
		return;
1071
	mz = page_cgroup_zoneinfo(pc->mem_cgroup, page);
K
KAMEZAWA Hiroyuki 已提交
1072
	list_move(&pc->lru, &mz->lists[lru]);
1073 1074
}

K
KAMEZAWA Hiroyuki 已提交
1075
void mem_cgroup_add_lru_list(struct page *page, enum lru_list lru)
1076
{
K
KAMEZAWA Hiroyuki 已提交
1077 1078
	struct page_cgroup *pc;
	struct mem_cgroup_per_zone *mz;
1079

1080
	if (mem_cgroup_disabled())
K
KAMEZAWA Hiroyuki 已提交
1081 1082
		return;
	pc = lookup_page_cgroup(page);
1083
	VM_BUG_ON(PageCgroupAcctLRU(pc));
1084 1085 1086 1087 1088 1089 1090 1091 1092 1093
	/*
	 * putback:				charge:
	 * SetPageLRU				SetPageCgroupUsed
	 * smp_mb				smp_mb
	 * PageCgroupUsed && add to memcg LRU	PageLRU && add to memcg LRU
	 *
	 * Ensure that one of the two sides adds the page to the memcg
	 * LRU during a race.
	 */
	smp_mb();
K
KAMEZAWA Hiroyuki 已提交
1094
	if (!PageCgroupUsed(pc))
L
Lee Schermerhorn 已提交
1095
		return;
1096 1097
	/* Ensure pc->mem_cgroup is visible after reading PCG_USED. */
	smp_rmb();
1098
	mz = page_cgroup_zoneinfo(pc->mem_cgroup, page);
1099 1100
	/* huge page split is done under lru_lock. so, we have no races. */
	MEM_CGROUP_ZSTAT(mz, lru) += 1 << compound_order(page);
1101 1102 1103
	SetPageCgroupAcctLRU(pc);
	if (mem_cgroup_is_root(pc->mem_cgroup))
		return;
K
KAMEZAWA Hiroyuki 已提交
1104 1105
	list_add(&pc->lru, &mz->lists[lru]);
}
1106

K
KAMEZAWA Hiroyuki 已提交
1107
/*
1108 1109 1110 1111
 * 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
1112
 */
1113
static void mem_cgroup_lru_del_before_commit(struct page *page)
K
KAMEZAWA Hiroyuki 已提交
1114
{
1115 1116 1117 1118
	unsigned long flags;
	struct zone *zone = page_zone(page);
	struct page_cgroup *pc = lookup_page_cgroup(page);

1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129
	/*
	 * 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;

1130 1131 1132 1133 1134 1135 1136 1137
	spin_lock_irqsave(&zone->lru_lock, flags);
	/*
	 * Forget old LRU when this page_cgroup is *not* used. This Used bit
	 * is guarded by lock_page() because the page is SwapCache.
	 */
	if (!PageCgroupUsed(pc))
		mem_cgroup_del_lru_list(page, page_lru(page));
	spin_unlock_irqrestore(&zone->lru_lock, flags);
K
KAMEZAWA Hiroyuki 已提交
1138 1139
}

1140
static void mem_cgroup_lru_add_after_commit(struct page *page)
1141 1142 1143 1144
{
	unsigned long flags;
	struct zone *zone = page_zone(page);
	struct page_cgroup *pc = lookup_page_cgroup(page);
1145 1146 1147 1148 1149 1150 1151 1152 1153 1154
	/*
	 * 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();
1155 1156 1157
	/* taking care of that the page is added to LRU while we commit it */
	if (likely(!PageLRU(page)))
		return;
1158 1159
	spin_lock_irqsave(&zone->lru_lock, flags);
	/* link when the page is linked to LRU but page_cgroup isn't */
1160
	if (PageLRU(page) && !PageCgroupAcctLRU(pc))
1161 1162 1163 1164 1165
		mem_cgroup_add_lru_list(page, page_lru(page));
	spin_unlock_irqrestore(&zone->lru_lock, flags);
}


K
KAMEZAWA Hiroyuki 已提交
1166 1167 1168
void mem_cgroup_move_lists(struct page *page,
			   enum lru_list from, enum lru_list to)
{
1169
	if (mem_cgroup_disabled())
K
KAMEZAWA Hiroyuki 已提交
1170 1171 1172
		return;
	mem_cgroup_del_lru_list(page, from);
	mem_cgroup_add_lru_list(page, to);
1173 1174
}

1175
/*
1176
 * Checks whether given mem is same or in the root_mem_cgroup's
1177 1178
 * hierarchy subtree
 */
1179 1180
static bool mem_cgroup_same_or_subtree(const struct mem_cgroup *root_memcg,
		struct mem_cgroup *memcg)
1181
{
1182 1183 1184
	if (root_memcg != memcg) {
		return (root_memcg->use_hierarchy &&
			css_is_ancestor(&memcg->css, &root_memcg->css));
1185 1186 1187 1188 1189
	}

	return true;
}

1190
int task_in_mem_cgroup(struct task_struct *task, const struct mem_cgroup *memcg)
1191 1192
{
	int ret;
1193
	struct mem_cgroup *curr = NULL;
1194
	struct task_struct *p;
1195

1196 1197 1198 1199 1200
	p = find_lock_task_mm(task);
	if (!p)
		return 0;
	curr = try_get_mem_cgroup_from_mm(p->mm);
	task_unlock(p);
1201 1202
	if (!curr)
		return 0;
1203
	/*
1204
	 * We should check use_hierarchy of "memcg" not "curr". Because checking
1205
	 * use_hierarchy of "curr" here make this function true if hierarchy is
1206 1207
	 * enabled in "curr" and "curr" is a child of "memcg" in *cgroup*
	 * hierarchy(even if use_hierarchy is disabled in "memcg").
1208
	 */
1209
	ret = mem_cgroup_same_or_subtree(memcg, curr);
1210
	css_put(&curr->css);
1211 1212 1213
	return ret;
}

1214
int mem_cgroup_inactive_anon_is_low(struct mem_cgroup *memcg, struct zone *zone)
1215
{
1216 1217 1218
	unsigned long inactive_ratio;
	int nid = zone_to_nid(zone);
	int zid = zone_idx(zone);
1219
	unsigned long inactive;
1220
	unsigned long active;
1221
	unsigned long gb;
1222

1223 1224 1225 1226
	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));
1227

1228 1229 1230 1231 1232 1233
	gb = (inactive + active) >> (30 - PAGE_SHIFT);
	if (gb)
		inactive_ratio = int_sqrt(10 * gb);
	else
		inactive_ratio = 1;

1234
	return inactive * inactive_ratio < active;
1235 1236
}

1237
int mem_cgroup_inactive_file_is_low(struct mem_cgroup *memcg, struct zone *zone)
1238 1239 1240
{
	unsigned long active;
	unsigned long inactive;
1241 1242
	int zid = zone_idx(zone);
	int nid = zone_to_nid(zone);
1243

1244 1245 1246 1247
	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));
1248 1249 1250 1251

	return (active > inactive);
}

K
KOSAKI Motohiro 已提交
1252 1253 1254
struct zone_reclaim_stat *mem_cgroup_get_reclaim_stat(struct mem_cgroup *memcg,
						      struct zone *zone)
{
1255
	int nid = zone_to_nid(zone);
K
KOSAKI Motohiro 已提交
1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271
	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);
1272 1273
	if (!PageCgroupUsed(pc))
		return NULL;
1274 1275
	/* Ensure pc->mem_cgroup is visible after reading PCG_USED. */
	smp_rmb();
1276
	mz = page_cgroup_zoneinfo(pc->mem_cgroup, page);
K
KOSAKI Motohiro 已提交
1277 1278 1279
	return &mz->reclaim_stat;
}

1280 1281 1282
unsigned long mem_cgroup_isolate_pages(unsigned long nr_to_scan,
					struct list_head *dst,
					unsigned long *scanned, int order,
1283 1284
					isolate_mode_t mode,
					struct zone *z,
1285
					struct mem_cgroup *mem_cont,
1286
					int active, int file)
1287 1288 1289 1290 1291 1292
{
	unsigned long nr_taken = 0;
	struct page *page;
	unsigned long scan;
	LIST_HEAD(pc_list);
	struct list_head *src;
1293
	struct page_cgroup *pc, *tmp;
1294
	int nid = zone_to_nid(z);
1295 1296
	int zid = zone_idx(z);
	struct mem_cgroup_per_zone *mz;
1297
	int lru = LRU_FILE * file + active;
1298
	int ret;
1299

1300
	BUG_ON(!mem_cont);
1301
	mz = mem_cgroup_zoneinfo(mem_cont, nid, zid);
1302
	src = &mz->lists[lru];
1303

1304 1305
	scan = 0;
	list_for_each_entry_safe_reverse(pc, tmp, src, lru) {
H
Hugh Dickins 已提交
1306
		if (scan >= nr_to_scan)
1307
			break;
K
KAMEZAWA Hiroyuki 已提交
1308

1309 1310
		if (unlikely(!PageCgroupUsed(pc)))
			continue;
1311

1312
		page = lookup_cgroup_page(pc);
1313

H
Hugh Dickins 已提交
1314
		if (unlikely(!PageLRU(page)))
1315 1316
			continue;

H
Hugh Dickins 已提交
1317
		scan++;
1318 1319 1320
		ret = __isolate_lru_page(page, mode, file);
		switch (ret) {
		case 0:
1321
			list_move(&page->lru, dst);
1322
			mem_cgroup_del_lru(page);
1323
			nr_taken += hpage_nr_pages(page);
1324 1325 1326 1327 1328 1329 1330
			break;
		case -EBUSY:
			/* we don't affect global LRU but rotate in our LRU */
			mem_cgroup_rotate_lru_list(page, page_lru(page));
			break;
		default:
			break;
1331 1332 1333 1334
		}
	}

	*scanned = scan;
1335 1336 1337 1338

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

1339 1340 1341
	return nr_taken;
}

1342 1343 1344
#define mem_cgroup_from_res_counter(counter, member)	\
	container_of(counter, struct mem_cgroup, member)

1345
/**
1346 1347
 * mem_cgroup_margin - calculate chargeable space of a memory cgroup
 * @mem: the memory cgroup
1348
 *
1349
 * Returns the maximum amount of memory @mem can be charged with, in
1350
 * pages.
1351
 */
1352
static unsigned long mem_cgroup_margin(struct mem_cgroup *memcg)
1353
{
1354 1355
	unsigned long long margin;

1356
	margin = res_counter_margin(&memcg->res);
1357
	if (do_swap_account)
1358
		margin = min(margin, res_counter_margin(&memcg->memsw));
1359
	return margin >> PAGE_SHIFT;
1360 1361
}

1362
int mem_cgroup_swappiness(struct mem_cgroup *memcg)
K
KOSAKI Motohiro 已提交
1363 1364 1365 1366 1367 1368 1369
{
	struct cgroup *cgrp = memcg->css.cgroup;

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

1370
	return memcg->swappiness;
K
KOSAKI Motohiro 已提交
1371 1372
}

1373
static void mem_cgroup_start_move(struct mem_cgroup *memcg)
1374 1375
{
	int cpu;
1376 1377

	get_online_cpus();
1378
	spin_lock(&memcg->pcp_counter_lock);
1379
	for_each_online_cpu(cpu)
1380 1381 1382
		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);
1383
	put_online_cpus();
1384 1385 1386 1387

	synchronize_rcu();
}

1388
static void mem_cgroup_end_move(struct mem_cgroup *memcg)
1389 1390 1391
{
	int cpu;

1392
	if (!memcg)
1393
		return;
1394
	get_online_cpus();
1395
	spin_lock(&memcg->pcp_counter_lock);
1396
	for_each_online_cpu(cpu)
1397 1398 1399
		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);
1400
	put_online_cpus();
1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413
}
/*
 * 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".
 */

1414
static bool mem_cgroup_stealed(struct mem_cgroup *memcg)
1415 1416
{
	VM_BUG_ON(!rcu_read_lock_held());
1417
	return this_cpu_read(memcg->stat->count[MEM_CGROUP_ON_MOVE]) > 0;
1418
}
1419

1420
static bool mem_cgroup_under_move(struct mem_cgroup *memcg)
1421
{
1422 1423
	struct mem_cgroup *from;
	struct mem_cgroup *to;
1424
	bool ret = false;
1425 1426 1427 1428 1429 1430 1431 1432 1433
	/*
	 * 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;
1434

1435 1436
	ret = mem_cgroup_same_or_subtree(memcg, from)
		|| mem_cgroup_same_or_subtree(memcg, to);
1437 1438
unlock:
	spin_unlock(&mc.lock);
1439 1440 1441
	return ret;
}

1442
static bool mem_cgroup_wait_acct_move(struct mem_cgroup *memcg)
1443 1444
{
	if (mc.moving_task && current != mc.moving_task) {
1445
		if (mem_cgroup_under_move(memcg)) {
1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457
			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;
}

1458
/**
1459
 * mem_cgroup_print_oom_info: Called from OOM with tasklist_lock held in read mode.
1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477
 * @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;

1478
	if (!memcg || !p)
1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524
		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));
}

1525 1526 1527 1528
/*
 * This function returns the number of memcg under hierarchy tree. Returns
 * 1(self count) if no children.
 */
1529
static int mem_cgroup_count_children(struct mem_cgroup *memcg)
1530 1531
{
	int num = 0;
K
KAMEZAWA Hiroyuki 已提交
1532 1533
	struct mem_cgroup *iter;

1534
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
1535
		num++;
1536 1537 1538
	return num;
}

D
David Rientjes 已提交
1539 1540 1541 1542 1543 1544 1545 1546
/*
 * Return the memory (and swap, if configured) limit for a memcg.
 */
u64 mem_cgroup_get_limit(struct mem_cgroup *memcg)
{
	u64 limit;
	u64 memsw;

1547 1548 1549
	limit = res_counter_read_u64(&memcg->res, RES_LIMIT);
	limit += total_swap_pages << PAGE_SHIFT;

D
David Rientjes 已提交
1550 1551 1552 1553 1554 1555 1556 1557
	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);
}

1558 1559 1560 1561 1562 1563 1564 1565 1566 1567
/**
 * 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.
 */
1568
static bool test_mem_cgroup_node_reclaimable(struct mem_cgroup *memcg,
1569 1570
		int nid, bool noswap)
{
1571
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_FILE))
1572 1573 1574
		return true;
	if (noswap || !total_swap_pages)
		return false;
1575
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_ANON))
1576 1577 1578 1579
		return true;
	return false;

}
1580 1581 1582 1583 1584 1585 1586 1587
#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.
 *
 */
1588
static void mem_cgroup_may_update_nodemask(struct mem_cgroup *memcg)
1589 1590
{
	int nid;
1591 1592 1593 1594
	/*
	 * numainfo_events > 0 means there was at least NUMAINFO_EVENTS_TARGET
	 * pagein/pageout changes since the last update.
	 */
1595
	if (!atomic_read(&memcg->numainfo_events))
1596
		return;
1597
	if (atomic_inc_return(&memcg->numainfo_updating) > 1)
1598 1599 1600
		return;

	/* make a nodemask where this memcg uses memory from */
1601
	memcg->scan_nodes = node_states[N_HIGH_MEMORY];
1602 1603 1604

	for_each_node_mask(nid, node_states[N_HIGH_MEMORY]) {

1605 1606
		if (!test_mem_cgroup_node_reclaimable(memcg, nid, false))
			node_clear(nid, memcg->scan_nodes);
1607
	}
1608

1609 1610
	atomic_set(&memcg->numainfo_events, 0);
	atomic_set(&memcg->numainfo_updating, 0);
1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624
}

/*
 * 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.
 */
1625
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
1626 1627 1628
{
	int node;

1629 1630
	mem_cgroup_may_update_nodemask(memcg);
	node = memcg->last_scanned_node;
1631

1632
	node = next_node(node, memcg->scan_nodes);
1633
	if (node == MAX_NUMNODES)
1634
		node = first_node(memcg->scan_nodes);
1635 1636 1637 1638 1639 1640 1641 1642 1643
	/*
	 * 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();

1644
	memcg->last_scanned_node = node;
1645 1646 1647
	return node;
}

1648 1649 1650 1651 1652 1653
/*
 * 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.
 */
1654
bool mem_cgroup_reclaimable(struct mem_cgroup *memcg, bool noswap)
1655 1656 1657 1658 1659 1660 1661
{
	int nid;

	/*
	 * quick check...making use of scan_node.
	 * We can skip unused nodes.
	 */
1662 1663
	if (!nodes_empty(memcg->scan_nodes)) {
		for (nid = first_node(memcg->scan_nodes);
1664
		     nid < MAX_NUMNODES;
1665
		     nid = next_node(nid, memcg->scan_nodes)) {
1666

1667
			if (test_mem_cgroup_node_reclaimable(memcg, nid, noswap))
1668 1669 1670 1671 1672 1673 1674
				return true;
		}
	}
	/*
	 * Check rest of nodes.
	 */
	for_each_node_state(nid, N_HIGH_MEMORY) {
1675
		if (node_isset(nid, memcg->scan_nodes))
1676
			continue;
1677
		if (test_mem_cgroup_node_reclaimable(memcg, nid, noswap))
1678 1679 1680 1681 1682
			return true;
	}
	return false;
}

1683
#else
1684
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
1685 1686 1687
{
	return 0;
}
1688

1689
bool mem_cgroup_reclaimable(struct mem_cgroup *memcg, bool noswap)
1690
{
1691
	return test_mem_cgroup_node_reclaimable(memcg, 0, noswap);
1692
}
1693 1694
#endif

K
KAMEZAWA Hiroyuki 已提交
1695 1696 1697 1698
/*
 * Scan the hierarchy if needed to reclaim memory. We remember the last child
 * we reclaimed from, so that we don't end up penalizing one child extensively
 * based on its position in the children list.
1699
 *
1700
 * root_memcg is the original ancestor that we've been reclaim from.
K
KAMEZAWA Hiroyuki 已提交
1701
 *
1702
 * We give up and return to the caller when we visit root_memcg twice.
K
KAMEZAWA Hiroyuki 已提交
1703
 * (other groups can be removed while we're walking....)
1704 1705
 *
 * If shrink==true, for avoiding to free too much, this returns immedieately.
1706
 */
1707
static int mem_cgroup_hierarchical_reclaim(struct mem_cgroup *root_memcg,
1708
						struct zone *zone,
1709
						gfp_t gfp_mask,
1710 1711
						unsigned long reclaim_options,
						unsigned long *total_scanned)
1712
{
1713
	struct mem_cgroup *victim = NULL;
K
KAMEZAWA Hiroyuki 已提交
1714 1715
	int ret, total = 0;
	int loop = 0;
1716 1717
	bool noswap = reclaim_options & MEM_CGROUP_RECLAIM_NOSWAP;
	bool shrink = reclaim_options & MEM_CGROUP_RECLAIM_SHRINK;
1718
	bool check_soft = reclaim_options & MEM_CGROUP_RECLAIM_SOFT;
1719
	unsigned long excess;
1720
	unsigned long nr_scanned;
1721 1722 1723 1724
	struct mem_cgroup_reclaim_cookie reclaim = {
		.zone = zone,
		.priority = 0,
	};
1725

1726
	excess = res_counter_soft_limit_excess(&root_memcg->res) >> PAGE_SHIFT;
K
KAMEZAWA Hiroyuki 已提交
1727

1728
	/* If memsw_is_minimum==1, swap-out is of-no-use. */
1729
	if (!check_soft && !shrink && root_memcg->memsw_is_minimum)
1730 1731
		noswap = true;

1732
	while (1) {
1733
		victim = mem_cgroup_iter(root_memcg, victim, &reclaim);
1734
		if (!victim) {
K
KAMEZAWA Hiroyuki 已提交
1735
			loop++;
1736 1737 1738 1739 1740 1741 1742
			/*
			 * We are not draining per cpu cached charges during
			 * soft limit reclaim  because global reclaim doesn't
			 * care about charges. It tries to free some memory and
			 * charges will not give any.
			 */
			if (!check_soft && loop >= 1)
1743
				drain_all_stock_async(root_memcg);
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 (!check_soft || !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, noswap)) {
K
KAMEZAWA Hiroyuki 已提交
1765
			/* this cgroup's local usage == 0 */
1766 1767
			continue;
		}
K
KAMEZAWA Hiroyuki 已提交
1768
		/* we use swappiness of local cgroup */
1769
		if (check_soft) {
1770
			ret = mem_cgroup_shrink_node_zone(victim, gfp_mask,
1771 1772
				noswap, zone, &nr_scanned);
			*total_scanned += nr_scanned;
1773
		} else
1774
			ret = try_to_free_mem_cgroup_pages(victim, gfp_mask,
1775
						noswap);
1776
		total += ret;
1777 1778 1779 1780 1781 1782
		/*
		 * At shrinking usage, we can't check we should stop here or
		 * reclaim more. It's depends on callers. last_scanned_child
		 * will work enough for keeping fairness under tree.
		 */
		if (shrink)
1783
			break;
1784
		if (check_soft) {
1785
			if (!res_counter_soft_limit_excess(&root_memcg->res))
1786
				break;
1787
		} else if (mem_cgroup_margin(root_memcg))
1788
			break;
1789
	}
1790
	mem_cgroup_iter_break(root_memcg, victim);
K
KAMEZAWA Hiroyuki 已提交
1791
	return total;
1792 1793
}

K
KAMEZAWA Hiroyuki 已提交
1794 1795 1796
/*
 * Check OOM-Killer is already running under our hierarchy.
 * If someone is running, return false.
1797
 * Has to be called with memcg_oom_lock
K
KAMEZAWA Hiroyuki 已提交
1798
 */
1799
static bool mem_cgroup_oom_lock(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
1800
{
1801
	struct mem_cgroup *iter, *failed = NULL;
1802

1803
	for_each_mem_cgroup_tree(iter, memcg) {
1804
		if (iter->oom_lock) {
1805 1806 1807 1808 1809
			/*
			 * this subtree of our hierarchy is already locked
			 * so we cannot give a lock.
			 */
			failed = iter;
1810 1811
			mem_cgroup_iter_break(memcg, iter);
			break;
1812 1813
		} else
			iter->oom_lock = true;
K
KAMEZAWA Hiroyuki 已提交
1814
	}
K
KAMEZAWA Hiroyuki 已提交
1815

1816
	if (!failed)
1817
		return true;
1818 1819 1820 1821 1822

	/*
	 * OK, we failed to lock the whole subtree so we have to clean up
	 * what we set up to the failing subtree
	 */
1823
	for_each_mem_cgroup_tree(iter, memcg) {
1824
		if (iter == failed) {
1825 1826
			mem_cgroup_iter_break(memcg, iter);
			break;
1827 1828 1829
		}
		iter->oom_lock = false;
	}
1830
	return false;
1831
}
1832

1833
/*
1834
 * Has to be called with memcg_oom_lock
1835
 */
1836
static int mem_cgroup_oom_unlock(struct mem_cgroup *memcg)
1837
{
K
KAMEZAWA Hiroyuki 已提交
1838 1839
	struct mem_cgroup *iter;

1840
	for_each_mem_cgroup_tree(iter, memcg)
1841 1842 1843 1844
		iter->oom_lock = false;
	return 0;
}

1845
static void mem_cgroup_mark_under_oom(struct mem_cgroup *memcg)
1846 1847 1848
{
	struct mem_cgroup *iter;

1849
	for_each_mem_cgroup_tree(iter, memcg)
1850 1851 1852
		atomic_inc(&iter->under_oom);
}

1853
static void mem_cgroup_unmark_under_oom(struct mem_cgroup *memcg)
1854 1855 1856
{
	struct mem_cgroup *iter;

K
KAMEZAWA Hiroyuki 已提交
1857 1858 1859 1860 1861
	/*
	 * 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.
	 */
1862
	for_each_mem_cgroup_tree(iter, memcg)
1863
		atomic_add_unless(&iter->under_oom, -1, 0);
1864 1865
}

1866
static DEFINE_SPINLOCK(memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
1867 1868
static DECLARE_WAIT_QUEUE_HEAD(memcg_oom_waitq);

K
KAMEZAWA Hiroyuki 已提交
1869 1870 1871 1872 1873 1874 1875 1876
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)
{
1877 1878
	struct mem_cgroup *wake_memcg = (struct mem_cgroup *)arg,
			  *oom_wait_memcg;
K
KAMEZAWA Hiroyuki 已提交
1879 1880 1881
	struct oom_wait_info *oom_wait_info;

	oom_wait_info = container_of(wait, struct oom_wait_info, wait);
1882
	oom_wait_memcg = oom_wait_info->mem;
K
KAMEZAWA Hiroyuki 已提交
1883 1884 1885 1886 1887

	/*
	 * 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.
	 */
1888 1889
	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
1890 1891 1892 1893
		return 0;
	return autoremove_wake_function(wait, mode, sync, arg);
}

1894
static void memcg_wakeup_oom(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
1895
{
1896 1897
	/* for filtering, pass "memcg" as argument. */
	__wake_up(&memcg_oom_waitq, TASK_NORMAL, 0, memcg);
K
KAMEZAWA Hiroyuki 已提交
1898 1899
}

1900
static void memcg_oom_recover(struct mem_cgroup *memcg)
1901
{
1902 1903
	if (memcg && atomic_read(&memcg->under_oom))
		memcg_wakeup_oom(memcg);
1904 1905
}

K
KAMEZAWA Hiroyuki 已提交
1906 1907 1908
/*
 * try to call OOM killer. returns false if we should exit memory-reclaim loop.
 */
1909
bool mem_cgroup_handle_oom(struct mem_cgroup *memcg, gfp_t mask)
1910
{
K
KAMEZAWA Hiroyuki 已提交
1911
	struct oom_wait_info owait;
1912
	bool locked, need_to_kill;
K
KAMEZAWA Hiroyuki 已提交
1913

1914
	owait.mem = memcg;
K
KAMEZAWA Hiroyuki 已提交
1915 1916 1917 1918
	owait.wait.flags = 0;
	owait.wait.func = memcg_oom_wake_function;
	owait.wait.private = current;
	INIT_LIST_HEAD(&owait.wait.task_list);
1919
	need_to_kill = true;
1920
	mem_cgroup_mark_under_oom(memcg);
1921

1922
	/* At first, try to OOM lock hierarchy under memcg.*/
1923
	spin_lock(&memcg_oom_lock);
1924
	locked = mem_cgroup_oom_lock(memcg);
K
KAMEZAWA Hiroyuki 已提交
1925 1926 1927 1928 1929
	/*
	 * 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.
	 */
1930
	prepare_to_wait(&memcg_oom_waitq, &owait.wait, TASK_KILLABLE);
1931
	if (!locked || memcg->oom_kill_disable)
1932 1933
		need_to_kill = false;
	if (locked)
1934
		mem_cgroup_oom_notify(memcg);
1935
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
1936

1937 1938
	if (need_to_kill) {
		finish_wait(&memcg_oom_waitq, &owait.wait);
1939
		mem_cgroup_out_of_memory(memcg, mask);
1940
	} else {
K
KAMEZAWA Hiroyuki 已提交
1941
		schedule();
K
KAMEZAWA Hiroyuki 已提交
1942
		finish_wait(&memcg_oom_waitq, &owait.wait);
K
KAMEZAWA Hiroyuki 已提交
1943
	}
1944
	spin_lock(&memcg_oom_lock);
1945
	if (locked)
1946 1947
		mem_cgroup_oom_unlock(memcg);
	memcg_wakeup_oom(memcg);
1948
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
1949

1950
	mem_cgroup_unmark_under_oom(memcg);
1951

K
KAMEZAWA Hiroyuki 已提交
1952 1953 1954
	if (test_thread_flag(TIF_MEMDIE) || fatal_signal_pending(current))
		return false;
	/* Give chance to dying process */
1955
	schedule_timeout_uninterruptible(1);
K
KAMEZAWA Hiroyuki 已提交
1956
	return true;
1957 1958
}

1959 1960 1961
/*
 * Currently used to update mapped file statistics, but the routine can be
 * generalized to update other statistics as well.
1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980
 *
 * 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.
1981
 */
1982

1983 1984
void mem_cgroup_update_page_stat(struct page *page,
				 enum mem_cgroup_page_stat_item idx, int val)
1985
{
1986
	struct mem_cgroup *memcg;
1987 1988
	struct page_cgroup *pc = lookup_page_cgroup(page);
	bool need_unlock = false;
1989
	unsigned long uninitialized_var(flags);
1990 1991 1992 1993

	if (unlikely(!pc))
		return;

1994
	rcu_read_lock();
1995 1996
	memcg = pc->mem_cgroup;
	if (unlikely(!memcg || !PageCgroupUsed(pc)))
1997 1998
		goto out;
	/* pc->mem_cgroup is unstable ? */
1999
	if (unlikely(mem_cgroup_stealed(memcg)) || PageTransHuge(page)) {
2000
		/* take a lock against to access pc->mem_cgroup */
2001
		move_lock_page_cgroup(pc, &flags);
2002
		need_unlock = true;
2003 2004
		memcg = pc->mem_cgroup;
		if (!memcg || !PageCgroupUsed(pc))
2005 2006
			goto out;
	}
2007 2008

	switch (idx) {
2009
	case MEMCG_NR_FILE_MAPPED:
2010 2011 2012
		if (val > 0)
			SetPageCgroupFileMapped(pc);
		else if (!page_mapped(page))
2013
			ClearPageCgroupFileMapped(pc);
2014
		idx = MEM_CGROUP_STAT_FILE_MAPPED;
2015 2016 2017
		break;
	default:
		BUG();
2018
	}
2019

2020
	this_cpu_add(memcg->stat->count[idx], val);
2021

2022 2023
out:
	if (unlikely(need_unlock))
2024
		move_unlock_page_cgroup(pc, &flags);
2025 2026
	rcu_read_unlock();
	return;
2027
}
2028
EXPORT_SYMBOL(mem_cgroup_update_page_stat);
2029

2030 2031 2032 2033
/*
 * size of first charge trial. "32" comes from vmscan.c's magic value.
 * TODO: maybe necessary to use big numbers in big irons.
 */
2034
#define CHARGE_BATCH	32U
2035 2036
struct memcg_stock_pcp {
	struct mem_cgroup *cached; /* this never be root cgroup */
2037
	unsigned int nr_pages;
2038
	struct work_struct work;
2039 2040
	unsigned long flags;
#define FLUSHING_CACHED_CHARGE	(0)
2041 2042
};
static DEFINE_PER_CPU(struct memcg_stock_pcp, memcg_stock);
2043
static DEFINE_MUTEX(percpu_charge_mutex);
2044 2045

/*
2046
 * Try to consume stocked charge on this cpu. If success, one page is consumed
2047 2048 2049 2050
 * 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.
 */
2051
static bool consume_stock(struct mem_cgroup *memcg)
2052 2053 2054 2055 2056
{
	struct memcg_stock_pcp *stock;
	bool ret = true;

	stock = &get_cpu_var(memcg_stock);
2057
	if (memcg == stock->cached && stock->nr_pages)
2058
		stock->nr_pages--;
2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071
	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;

2072 2073 2074 2075
	if (stock->nr_pages) {
		unsigned long bytes = stock->nr_pages * PAGE_SIZE;

		res_counter_uncharge(&old->res, bytes);
2076
		if (do_swap_account)
2077 2078
			res_counter_uncharge(&old->memsw, bytes);
		stock->nr_pages = 0;
2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090
	}
	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);
2091
	clear_bit(FLUSHING_CACHED_CHARGE, &stock->flags);
2092 2093 2094 2095
}

/*
 * Cache charges(val) which is from res_counter, to local per_cpu area.
2096
 * This will be consumed by consume_stock() function, later.
2097
 */
2098
static void refill_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
2099 2100 2101
{
	struct memcg_stock_pcp *stock = &get_cpu_var(memcg_stock);

2102
	if (stock->cached != memcg) { /* reset if necessary */
2103
		drain_stock(stock);
2104
		stock->cached = memcg;
2105
	}
2106
	stock->nr_pages += nr_pages;
2107 2108 2109 2110
	put_cpu_var(memcg_stock);
}

/*
2111
 * Drains all per-CPU charge caches for given root_memcg resp. subtree
2112 2113
 * of the hierarchy under it. sync flag says whether we should block
 * until the work is done.
2114
 */
2115
static void drain_all_stock(struct mem_cgroup *root_memcg, bool sync)
2116
{
2117
	int cpu, curcpu;
2118

2119 2120
	/* Notify other cpus that system-wide "drain" is running */
	get_online_cpus();
2121
	curcpu = get_cpu();
2122 2123
	for_each_online_cpu(cpu) {
		struct memcg_stock_pcp *stock = &per_cpu(memcg_stock, cpu);
2124
		struct mem_cgroup *memcg;
2125

2126 2127
		memcg = stock->cached;
		if (!memcg || !stock->nr_pages)
2128
			continue;
2129
		if (!mem_cgroup_same_or_subtree(root_memcg, memcg))
2130
			continue;
2131 2132 2133 2134 2135 2136
		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);
		}
2137
	}
2138
	put_cpu();
2139 2140 2141 2142 2143 2144

	if (!sync)
		goto out;

	for_each_online_cpu(cpu) {
		struct memcg_stock_pcp *stock = &per_cpu(memcg_stock, cpu);
2145
		if (test_bit(FLUSHING_CACHED_CHARGE, &stock->flags))
2146 2147 2148
			flush_work(&stock->work);
	}
out:
2149
 	put_online_cpus();
2150 2151 2152 2153 2154 2155 2156 2157
}

/*
 * 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.
 */
2158
static void drain_all_stock_async(struct mem_cgroup *root_memcg)
2159
{
2160 2161 2162 2163 2164
	/*
	 * If someone calls draining, avoid adding more kworker runs.
	 */
	if (!mutex_trylock(&percpu_charge_mutex))
		return;
2165
	drain_all_stock(root_memcg, false);
2166
	mutex_unlock(&percpu_charge_mutex);
2167 2168 2169
}

/* This is a synchronous drain interface. */
2170
static void drain_all_stock_sync(struct mem_cgroup *root_memcg)
2171 2172
{
	/* called when force_empty is called */
2173
	mutex_lock(&percpu_charge_mutex);
2174
	drain_all_stock(root_memcg, true);
2175
	mutex_unlock(&percpu_charge_mutex);
2176 2177
}

2178 2179 2180 2181
/*
 * This function drains percpu counter value from DEAD cpu and
 * move it to local cpu. Note that this function can be preempted.
 */
2182
static void mem_cgroup_drain_pcp_counter(struct mem_cgroup *memcg, int cpu)
2183 2184 2185
{
	int i;

2186
	spin_lock(&memcg->pcp_counter_lock);
2187
	for (i = 0; i < MEM_CGROUP_STAT_DATA; i++) {
2188
		long x = per_cpu(memcg->stat->count[i], cpu);
2189

2190 2191
		per_cpu(memcg->stat->count[i], cpu) = 0;
		memcg->nocpu_base.count[i] += x;
2192
	}
2193
	for (i = 0; i < MEM_CGROUP_EVENTS_NSTATS; i++) {
2194
		unsigned long x = per_cpu(memcg->stat->events[i], cpu);
2195

2196 2197
		per_cpu(memcg->stat->events[i], cpu) = 0;
		memcg->nocpu_base.events[i] += x;
2198
	}
2199
	/* need to clear ON_MOVE value, works as a kind of lock. */
2200 2201
	per_cpu(memcg->stat->count[MEM_CGROUP_ON_MOVE], cpu) = 0;
	spin_unlock(&memcg->pcp_counter_lock);
2202 2203
}

2204
static void synchronize_mem_cgroup_on_move(struct mem_cgroup *memcg, int cpu)
2205 2206 2207
{
	int idx = MEM_CGROUP_ON_MOVE;

2208 2209 2210
	spin_lock(&memcg->pcp_counter_lock);
	per_cpu(memcg->stat->count[idx], cpu) = memcg->nocpu_base.count[idx];
	spin_unlock(&memcg->pcp_counter_lock);
2211 2212 2213
}

static int __cpuinit memcg_cpu_hotplug_callback(struct notifier_block *nb,
2214 2215 2216 2217 2218
					unsigned long action,
					void *hcpu)
{
	int cpu = (unsigned long)hcpu;
	struct memcg_stock_pcp *stock;
2219
	struct mem_cgroup *iter;
2220

2221
	if ((action == CPU_ONLINE)) {
2222
		for_each_mem_cgroup(iter)
2223 2224 2225 2226
			synchronize_mem_cgroup_on_move(iter, cpu);
		return NOTIFY_OK;
	}

2227
	if ((action != CPU_DEAD) || action != CPU_DEAD_FROZEN)
2228
		return NOTIFY_OK;
2229

2230
	for_each_mem_cgroup(iter)
2231 2232
		mem_cgroup_drain_pcp_counter(iter, cpu);

2233 2234 2235 2236 2237
	stock = &per_cpu(memcg_stock, cpu);
	drain_stock(stock);
	return NOTIFY_OK;
}

2238 2239 2240 2241 2242 2243 2244 2245 2246 2247

/* 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 */
};

2248
static int mem_cgroup_do_charge(struct mem_cgroup *memcg, gfp_t gfp_mask,
2249
				unsigned int nr_pages, bool oom_check)
2250
{
2251
	unsigned long csize = nr_pages * PAGE_SIZE;
2252 2253 2254 2255 2256
	struct mem_cgroup *mem_over_limit;
	struct res_counter *fail_res;
	unsigned long flags = 0;
	int ret;

2257
	ret = res_counter_charge(&memcg->res, csize, &fail_res);
2258 2259 2260 2261

	if (likely(!ret)) {
		if (!do_swap_account)
			return CHARGE_OK;
2262
		ret = res_counter_charge(&memcg->memsw, csize, &fail_res);
2263 2264 2265
		if (likely(!ret))
			return CHARGE_OK;

2266
		res_counter_uncharge(&memcg->res, csize);
2267 2268 2269 2270
		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);
2271
	/*
2272 2273
	 * nr_pages can be either a huge page (HPAGE_PMD_NR), a batch
	 * of regular pages (CHARGE_BATCH), or a single regular page (1).
2274 2275 2276 2277
	 *
	 * Never reclaim on behalf of optional batching, retry with a
	 * single page instead.
	 */
2278
	if (nr_pages == CHARGE_BATCH)
2279 2280 2281 2282 2283 2284
		return CHARGE_RETRY;

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

	ret = mem_cgroup_hierarchical_reclaim(mem_over_limit, NULL,
2285
					      gfp_mask, flags, NULL);
2286
	if (mem_cgroup_margin(mem_over_limit) >= nr_pages)
2287
		return CHARGE_RETRY;
2288
	/*
2289 2290 2291 2292 2293 2294 2295
	 * 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.
2296
	 */
2297
	if (nr_pages == 1 && ret)
2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316
		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;
}

2317 2318 2319
/*
 * Unlike exported interface, "oom" parameter is added. if oom==true,
 * oom-killer can be invoked.
2320
 */
2321
static int __mem_cgroup_try_charge(struct mm_struct *mm,
A
Andrea Arcangeli 已提交
2322
				   gfp_t gfp_mask,
2323
				   unsigned int nr_pages,
2324
				   struct mem_cgroup **ptr,
2325
				   bool oom)
2326
{
2327
	unsigned int batch = max(CHARGE_BATCH, nr_pages);
2328
	int nr_oom_retries = MEM_CGROUP_RECLAIM_RETRIES;
2329
	struct mem_cgroup *memcg = NULL;
2330
	int ret;
2331

K
KAMEZAWA Hiroyuki 已提交
2332 2333 2334 2335 2336 2337 2338 2339
	/*
	 * 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;
2340

2341
	/*
2342 2343
	 * We always charge the cgroup the mm_struct belongs to.
	 * The mm_struct's mem_cgroup changes on task migration if the
2344 2345 2346
	 * thread group leader migrates. It's possible that mm is not
	 * set, if so charge the init_mm (happens for pagecache usage).
	 */
2347
	if (!*ptr && !mm)
K
KAMEZAWA Hiroyuki 已提交
2348 2349
		goto bypass;
again:
2350 2351 2352 2353
	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
2354
			goto done;
2355
		if (nr_pages == 1 && consume_stock(memcg))
K
KAMEZAWA Hiroyuki 已提交
2356
			goto done;
2357
		css_get(&memcg->css);
2358
	} else {
K
KAMEZAWA Hiroyuki 已提交
2359
		struct task_struct *p;
2360

K
KAMEZAWA Hiroyuki 已提交
2361 2362 2363
		rcu_read_lock();
		p = rcu_dereference(mm->owner);
		/*
2364
		 * Because we don't have task_lock(), "p" can exit.
2365
		 * In that case, "memcg" can point to root or p can be NULL with
2366 2367 2368 2369 2370 2371
		 * 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
2372
		 */
2373 2374
		memcg = mem_cgroup_from_task(p);
		if (!memcg || mem_cgroup_is_root(memcg)) {
K
KAMEZAWA Hiroyuki 已提交
2375 2376 2377
			rcu_read_unlock();
			goto done;
		}
2378
		if (nr_pages == 1 && consume_stock(memcg)) {
K
KAMEZAWA Hiroyuki 已提交
2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390
			/*
			 * 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 */
2391
		if (!css_tryget(&memcg->css)) {
K
KAMEZAWA Hiroyuki 已提交
2392 2393 2394 2395 2396
			rcu_read_unlock();
			goto again;
		}
		rcu_read_unlock();
	}
2397

2398 2399
	do {
		bool oom_check;
2400

2401
		/* If killed, bypass charge */
K
KAMEZAWA Hiroyuki 已提交
2402
		if (fatal_signal_pending(current)) {
2403
			css_put(&memcg->css);
2404
			goto bypass;
K
KAMEZAWA Hiroyuki 已提交
2405
		}
2406

2407 2408 2409 2410
		oom_check = false;
		if (oom && !nr_oom_retries) {
			oom_check = true;
			nr_oom_retries = MEM_CGROUP_RECLAIM_RETRIES;
2411
		}
2412

2413
		ret = mem_cgroup_do_charge(memcg, gfp_mask, batch, oom_check);
2414 2415 2416 2417
		switch (ret) {
		case CHARGE_OK:
			break;
		case CHARGE_RETRY: /* not in OOM situation but retry */
2418
			batch = nr_pages;
2419 2420
			css_put(&memcg->css);
			memcg = NULL;
K
KAMEZAWA Hiroyuki 已提交
2421
			goto again;
2422
		case CHARGE_WOULDBLOCK: /* !__GFP_WAIT */
2423
			css_put(&memcg->css);
2424 2425
			goto nomem;
		case CHARGE_NOMEM: /* OOM routine works */
K
KAMEZAWA Hiroyuki 已提交
2426
			if (!oom) {
2427
				css_put(&memcg->css);
K
KAMEZAWA Hiroyuki 已提交
2428
				goto nomem;
K
KAMEZAWA Hiroyuki 已提交
2429
			}
2430 2431 2432 2433
			/* If oom, we never return -ENOMEM */
			nr_oom_retries--;
			break;
		case CHARGE_OOM_DIE: /* Killed by OOM Killer */
2434
			css_put(&memcg->css);
K
KAMEZAWA Hiroyuki 已提交
2435
			goto bypass;
2436
		}
2437 2438
	} while (ret != CHARGE_OK);

2439
	if (batch > nr_pages)
2440 2441
		refill_stock(memcg, batch - nr_pages);
	css_put(&memcg->css);
2442
done:
2443
	*ptr = memcg;
2444 2445
	return 0;
nomem:
2446
	*ptr = NULL;
2447
	return -ENOMEM;
K
KAMEZAWA Hiroyuki 已提交
2448
bypass:
2449
	*ptr = NULL;
K
KAMEZAWA Hiroyuki 已提交
2450
	return 0;
2451
}
2452

2453 2454 2455 2456 2457
/*
 * 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().
 */
2458
static void __mem_cgroup_cancel_charge(struct mem_cgroup *memcg,
2459
				       unsigned int nr_pages)
2460
{
2461
	if (!mem_cgroup_is_root(memcg)) {
2462 2463
		unsigned long bytes = nr_pages * PAGE_SIZE;

2464
		res_counter_uncharge(&memcg->res, bytes);
2465
		if (do_swap_account)
2466
			res_counter_uncharge(&memcg->memsw, bytes);
2467
	}
2468 2469
}

2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488
/*
 * 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);
}

2489
struct mem_cgroup *try_get_mem_cgroup_from_page(struct page *page)
2490
{
2491
	struct mem_cgroup *memcg = NULL;
2492
	struct page_cgroup *pc;
2493
	unsigned short id;
2494 2495
	swp_entry_t ent;

2496 2497 2498
	VM_BUG_ON(!PageLocked(page));

	pc = lookup_page_cgroup(page);
2499
	lock_page_cgroup(pc);
2500
	if (PageCgroupUsed(pc)) {
2501 2502 2503
		memcg = pc->mem_cgroup;
		if (memcg && !css_tryget(&memcg->css))
			memcg = NULL;
2504
	} else if (PageSwapCache(page)) {
2505
		ent.val = page_private(page);
2506 2507
		id = lookup_swap_cgroup(ent);
		rcu_read_lock();
2508 2509 2510
		memcg = mem_cgroup_lookup(id);
		if (memcg && !css_tryget(&memcg->css))
			memcg = NULL;
2511
		rcu_read_unlock();
2512
	}
2513
	unlock_page_cgroup(pc);
2514
	return memcg;
2515 2516
}

2517
static void __mem_cgroup_commit_charge(struct mem_cgroup *memcg,
2518
				       struct page *page,
2519
				       unsigned int nr_pages,
2520
				       struct page_cgroup *pc,
2521
				       enum charge_type ctype)
2522
{
2523 2524 2525
	lock_page_cgroup(pc);
	if (unlikely(PageCgroupUsed(pc))) {
		unlock_page_cgroup(pc);
2526
		__mem_cgroup_cancel_charge(memcg, nr_pages);
2527 2528 2529 2530 2531 2532
		return;
	}
	/*
	 * we don't need page_cgroup_lock about tail pages, becase they are not
	 * accessed by any other context at this point.
	 */
2533
	pc->mem_cgroup = memcg;
2534 2535 2536 2537 2538 2539 2540
	/*
	 * 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
2541
	smp_wmb();
2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554
	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;
	}
2555

2556
	mem_cgroup_charge_statistics(memcg, PageCgroupCache(pc), nr_pages);
2557
	unlock_page_cgroup(pc);
2558 2559 2560 2561 2562
	/*
	 * "charge_statistics" updated event counter. Then, check it.
	 * Insert ancestor (and ancestor's ancestors), to softlimit RB-tree.
	 * if they exceeds softlimit.
	 */
2563
	memcg_check_events(memcg, page);
2564
}
2565

2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579
#ifdef CONFIG_TRANSPARENT_HUGEPAGE

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

2580 2581
	if (mem_cgroup_disabled())
		return;
2582
	/*
2583
	 * We have no races with charge/uncharge but will have races with
2584 2585 2586 2587 2588 2589
	 * page state accounting.
	 */
	move_lock_page_cgroup(head_pc, &flags);

	tail_pc->mem_cgroup = head_pc->mem_cgroup;
	smp_wmb(); /* see __commit_charge() */
2590 2591 2592 2593 2594 2595 2596 2597 2598 2599
	if (PageCgroupAcctLRU(head_pc)) {
		enum lru_list lru;
		struct mem_cgroup_per_zone *mz;

		/*
		 * LRU flags cannot be copied because we need to add tail
		 *.page to LRU by generic call and our hook will be called.
		 * We hold lru_lock, then, reduce counter directly.
		 */
		lru = page_lru(head);
2600
		mz = page_cgroup_zoneinfo(head_pc->mem_cgroup, head);
2601 2602
		MEM_CGROUP_ZSTAT(mz, lru) -= 1;
	}
2603 2604 2605 2606 2607
	tail_pc->flags = head_pc->flags & ~PCGF_NOCOPY_AT_SPLIT;
	move_unlock_page_cgroup(head_pc, &flags);
}
#endif

2608
/**
2609
 * mem_cgroup_move_account - move account of the page
2610
 * @page: the page
2611
 * @nr_pages: number of regular pages (>1 for huge pages)
2612 2613 2614
 * @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.
2615
 * @uncharge: whether we should call uncharge and css_put against @from.
2616 2617
 *
 * The caller must confirm following.
K
KAMEZAWA Hiroyuki 已提交
2618
 * - page is not on LRU (isolate_page() is useful.)
2619
 * - compound_lock is held when nr_pages > 1
2620
 *
2621
 * This function doesn't do "charge" nor css_get to new cgroup. It should be
L
Lucas De Marchi 已提交
2622
 * done by a caller(__mem_cgroup_try_charge would be useful). If @uncharge is
2623 2624
 * true, this function does "uncharge" from old cgroup, but it doesn't if
 * @uncharge is false, so a caller should do "uncharge".
2625
 */
2626 2627 2628 2629 2630 2631
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)
2632
{
2633 2634
	unsigned long flags;
	int ret;
2635

2636
	VM_BUG_ON(from == to);
2637
	VM_BUG_ON(PageLRU(page));
2638 2639 2640 2641 2642 2643 2644
	/*
	 * 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;
2645
	if (nr_pages > 1 && !PageTransHuge(page))
2646 2647 2648 2649 2650 2651 2652 2653 2654
		goto out;

	lock_page_cgroup(pc);

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

	move_lock_page_cgroup(pc, &flags);
2655

2656
	if (PageCgroupFileMapped(pc)) {
2657 2658 2659 2660 2661
		/* 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();
2662
	}
2663
	mem_cgroup_charge_statistics(from, PageCgroupCache(pc), -nr_pages);
2664 2665
	if (uncharge)
		/* This is not "cancel", but cancel_charge does all we need. */
2666
		__mem_cgroup_cancel_charge(from, nr_pages);
2667

2668
	/* caller should have done css_get */
K
KAMEZAWA Hiroyuki 已提交
2669
	pc->mem_cgroup = to;
2670
	mem_cgroup_charge_statistics(to, PageCgroupCache(pc), nr_pages);
2671 2672 2673
	/*
	 * We charges against "to" which may not have any tasks. Then, "to"
	 * can be under rmdir(). But in current implementation, caller of
2674
	 * this function is just force_empty() and move charge, so it's
L
Lucas De Marchi 已提交
2675
	 * guaranteed that "to" is never removed. So, we don't check rmdir
2676
	 * status here.
2677
	 */
2678 2679 2680
	move_unlock_page_cgroup(pc, &flags);
	ret = 0;
unlock:
2681
	unlock_page_cgroup(pc);
2682 2683 2684
	/*
	 * check events
	 */
2685 2686
	memcg_check_events(to, page);
	memcg_check_events(from, page);
2687
out:
2688 2689 2690 2691 2692 2693 2694
	return ret;
}

/*
 * move charges to its parent.
 */

2695 2696
static int mem_cgroup_move_parent(struct page *page,
				  struct page_cgroup *pc,
2697 2698 2699 2700 2701 2702
				  struct mem_cgroup *child,
				  gfp_t gfp_mask)
{
	struct cgroup *cg = child->css.cgroup;
	struct cgroup *pcg = cg->parent;
	struct mem_cgroup *parent;
2703
	unsigned int nr_pages;
2704
	unsigned long uninitialized_var(flags);
2705 2706 2707 2708 2709 2710
	int ret;

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

2711 2712 2713 2714 2715
	ret = -EBUSY;
	if (!get_page_unless_zero(page))
		goto out;
	if (isolate_lru_page(page))
		goto put;
2716

2717
	nr_pages = hpage_nr_pages(page);
K
KAMEZAWA Hiroyuki 已提交
2718

2719
	parent = mem_cgroup_from_cont(pcg);
2720
	ret = __mem_cgroup_try_charge(NULL, gfp_mask, nr_pages, &parent, false);
2721
	if (ret || !parent)
2722
		goto put_back;
2723

2724
	if (nr_pages > 1)
2725 2726
		flags = compound_lock_irqsave(page);

2727
	ret = mem_cgroup_move_account(page, nr_pages, pc, child, parent, true);
2728
	if (ret)
2729
		__mem_cgroup_cancel_charge(parent, nr_pages);
2730

2731
	if (nr_pages > 1)
2732
		compound_unlock_irqrestore(page, flags);
2733
put_back:
K
KAMEZAWA Hiroyuki 已提交
2734
	putback_lru_page(page);
2735
put:
2736
	put_page(page);
2737
out:
2738 2739 2740
	return ret;
}

2741 2742 2743 2744 2745 2746 2747
/*
 * 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,
2748
				gfp_t gfp_mask, enum charge_type ctype)
2749
{
2750
	struct mem_cgroup *memcg = NULL;
2751
	unsigned int nr_pages = 1;
2752
	struct page_cgroup *pc;
2753
	bool oom = true;
2754
	int ret;
A
Andrea Arcangeli 已提交
2755

A
Andrea Arcangeli 已提交
2756
	if (PageTransHuge(page)) {
2757
		nr_pages <<= compound_order(page);
A
Andrea Arcangeli 已提交
2758
		VM_BUG_ON(!PageTransHuge(page));
2759 2760 2761 2762 2763
		/*
		 * 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
2764
	}
2765 2766

	pc = lookup_page_cgroup(page);
2767
	BUG_ON(!pc); /* XXX: remove this and move pc lookup into commit */
2768

2769 2770
	ret = __mem_cgroup_try_charge(mm, gfp_mask, nr_pages, &memcg, oom);
	if (ret || !memcg)
2771 2772
		return ret;

2773
	__mem_cgroup_commit_charge(memcg, page, nr_pages, pc, ctype);
2774 2775 2776
	return 0;
}

2777 2778
int mem_cgroup_newpage_charge(struct page *page,
			      struct mm_struct *mm, gfp_t gfp_mask)
2779
{
2780
	if (mem_cgroup_disabled())
2781
		return 0;
2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792
	/*
	 * 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;
2793
	return mem_cgroup_charge_common(page, mm, gfp_mask,
2794
				MEM_CGROUP_CHARGE_TYPE_MAPPED);
2795 2796
}

D
Daisuke Nishimura 已提交
2797 2798 2799 2800
static void
__mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr,
					enum charge_type ctype);

2801
static void
2802
__mem_cgroup_commit_charge_lrucare(struct page *page, struct mem_cgroup *memcg,
2803 2804 2805 2806 2807 2808 2809 2810 2811
					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);
2812
	__mem_cgroup_commit_charge(memcg, page, 1, pc, ctype);
2813 2814 2815 2816
	mem_cgroup_lru_add_after_commit(page);
	return;
}

2817 2818
int mem_cgroup_cache_charge(struct page *page, struct mm_struct *mm,
				gfp_t gfp_mask)
2819
{
2820
	struct mem_cgroup *memcg = NULL;
2821 2822
	int ret;

2823
	if (mem_cgroup_disabled())
2824
		return 0;
2825 2826
	if (PageCompound(page))
		return 0;
2827

2828
	if (unlikely(!mm))
2829
		mm = &init_mm;
2830

2831
	if (page_is_file_cache(page)) {
2832 2833
		ret = __mem_cgroup_try_charge(mm, gfp_mask, 1, &memcg, true);
		if (ret || !memcg)
2834
			return ret;
2835

2836 2837 2838 2839 2840
		/*
		 * FUSE reuses pages without going through the final
		 * put that would remove them from the LRU list, make
		 * sure that they get relinked properly.
		 */
2841
		__mem_cgroup_commit_charge_lrucare(page, memcg,
2842 2843 2844
					MEM_CGROUP_CHARGE_TYPE_CACHE);
		return ret;
	}
D
Daisuke Nishimura 已提交
2845 2846
	/* shmem */
	if (PageSwapCache(page)) {
2847
		ret = mem_cgroup_try_charge_swapin(mm, page, gfp_mask, &memcg);
D
Daisuke Nishimura 已提交
2848
		if (!ret)
2849
			__mem_cgroup_commit_charge_swapin(page, memcg,
D
Daisuke Nishimura 已提交
2850 2851 2852
					MEM_CGROUP_CHARGE_TYPE_SHMEM);
	} else
		ret = mem_cgroup_charge_common(page, mm, gfp_mask,
2853
					MEM_CGROUP_CHARGE_TYPE_SHMEM);
2854 2855

	return ret;
2856 2857
}

2858 2859 2860
/*
 * While swap-in, try_charge -> commit or cancel, the page is locked.
 * And when try_charge() successfully returns, one refcnt to memcg without
2861
 * struct page_cgroup is acquired. This refcnt will be consumed by
2862 2863
 * "commit()" or removed by "cancel()"
 */
2864 2865 2866 2867
int mem_cgroup_try_charge_swapin(struct mm_struct *mm,
				 struct page *page,
				 gfp_t mask, struct mem_cgroup **ptr)
{
2868
	struct mem_cgroup *memcg;
2869
	int ret;
2870

2871 2872
	*ptr = NULL;

2873
	if (mem_cgroup_disabled())
2874 2875 2876 2877 2878 2879
		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
2880 2881 2882
	 * 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.
2883 2884
	 */
	if (!PageSwapCache(page))
H
Hugh Dickins 已提交
2885
		goto charge_cur_mm;
2886 2887
	memcg = try_get_mem_cgroup_from_page(page);
	if (!memcg)
2888
		goto charge_cur_mm;
2889
	*ptr = memcg;
2890
	ret = __mem_cgroup_try_charge(NULL, mask, 1, ptr, true);
2891
	css_put(&memcg->css);
2892
	return ret;
2893 2894 2895
charge_cur_mm:
	if (unlikely(!mm))
		mm = &init_mm;
2896
	return __mem_cgroup_try_charge(mm, mask, 1, ptr, true);
2897 2898
}

D
Daisuke Nishimura 已提交
2899 2900 2901
static void
__mem_cgroup_commit_charge_swapin(struct page *page, struct mem_cgroup *ptr,
					enum charge_type ctype)
2902
{
2903
	if (mem_cgroup_disabled())
2904 2905 2906
		return;
	if (!ptr)
		return;
2907
	cgroup_exclude_rmdir(&ptr->css);
2908 2909

	__mem_cgroup_commit_charge_lrucare(page, ptr, ctype);
2910 2911 2912
	/*
	 * Now swap is on-memory. This means this page may be
	 * counted both as mem and swap....double count.
2913 2914 2915
	 * 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.
2916
	 */
2917
	if (do_swap_account && PageSwapCache(page)) {
2918
		swp_entry_t ent = {.val = page_private(page)};
2919
		unsigned short id;
2920
		struct mem_cgroup *memcg;
2921 2922 2923 2924

		id = swap_cgroup_record(ent, 0);
		rcu_read_lock();
		memcg = mem_cgroup_lookup(id);
2925
		if (memcg) {
2926 2927 2928 2929
			/*
			 * This recorded memcg can be obsolete one. So, avoid
			 * calling css_tryget
			 */
2930
			if (!mem_cgroup_is_root(memcg))
2931
				res_counter_uncharge(&memcg->memsw, PAGE_SIZE);
2932
			mem_cgroup_swap_statistics(memcg, false);
2933 2934
			mem_cgroup_put(memcg);
		}
2935
		rcu_read_unlock();
2936
	}
2937 2938 2939 2940 2941 2942
	/*
	 * 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);
2943 2944
}

D
Daisuke Nishimura 已提交
2945 2946 2947 2948 2949 2950
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);
}

2951
void mem_cgroup_cancel_charge_swapin(struct mem_cgroup *memcg)
2952
{
2953
	if (mem_cgroup_disabled())
2954
		return;
2955
	if (!memcg)
2956
		return;
2957
	__mem_cgroup_cancel_charge(memcg, 1);
2958 2959
}

2960
static void mem_cgroup_do_uncharge(struct mem_cgroup *memcg,
2961 2962
				   unsigned int nr_pages,
				   const enum charge_type ctype)
2963 2964 2965
{
	struct memcg_batch_info *batch = NULL;
	bool uncharge_memsw = true;
2966

2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977
	/* 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)
2978
		batch->memcg = memcg;
2979 2980
	/*
	 * do_batch > 0 when unmapping pages or inode invalidate/truncate.
L
Lucas De Marchi 已提交
2981
	 * In those cases, all pages freed continuously can be expected to be in
2982 2983 2984 2985 2986 2987 2988 2989
	 * 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;

2990
	if (nr_pages > 1)
A
Andrea Arcangeli 已提交
2991 2992
		goto direct_uncharge;

2993 2994 2995 2996 2997
	/*
	 * 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.
	 */
2998
	if (batch->memcg != memcg)
2999 3000
		goto direct_uncharge;
	/* remember freed charge and uncharge it later */
3001
	batch->nr_pages++;
3002
	if (uncharge_memsw)
3003
		batch->memsw_nr_pages++;
3004 3005
	return;
direct_uncharge:
3006
	res_counter_uncharge(&memcg->res, nr_pages * PAGE_SIZE);
3007
	if (uncharge_memsw)
3008 3009 3010
		res_counter_uncharge(&memcg->memsw, nr_pages * PAGE_SIZE);
	if (unlikely(batch->memcg != memcg))
		memcg_oom_recover(memcg);
3011 3012
	return;
}
3013

3014
/*
3015
 * uncharge if !page_mapped(page)
3016
 */
3017
static struct mem_cgroup *
3018
__mem_cgroup_uncharge_common(struct page *page, enum charge_type ctype)
3019
{
3020
	struct mem_cgroup *memcg = NULL;
3021 3022
	unsigned int nr_pages = 1;
	struct page_cgroup *pc;
3023

3024
	if (mem_cgroup_disabled())
3025
		return NULL;
3026

K
KAMEZAWA Hiroyuki 已提交
3027
	if (PageSwapCache(page))
3028
		return NULL;
K
KAMEZAWA Hiroyuki 已提交
3029

A
Andrea Arcangeli 已提交
3030
	if (PageTransHuge(page)) {
3031
		nr_pages <<= compound_order(page);
A
Andrea Arcangeli 已提交
3032 3033
		VM_BUG_ON(!PageTransHuge(page));
	}
3034
	/*
3035
	 * Check if our page_cgroup is valid
3036
	 */
3037 3038
	pc = lookup_page_cgroup(page);
	if (unlikely(!pc || !PageCgroupUsed(pc)))
3039
		return NULL;
3040

3041
	lock_page_cgroup(pc);
K
KAMEZAWA Hiroyuki 已提交
3042

3043
	memcg = pc->mem_cgroup;
3044

K
KAMEZAWA Hiroyuki 已提交
3045 3046 3047 3048 3049
	if (!PageCgroupUsed(pc))
		goto unlock_out;

	switch (ctype) {
	case MEM_CGROUP_CHARGE_TYPE_MAPPED:
K
KAMEZAWA Hiroyuki 已提交
3050
	case MEM_CGROUP_CHARGE_TYPE_DROP:
3051 3052
		/* See mem_cgroup_prepare_migration() */
		if (page_mapped(page) || PageCgroupMigration(pc))
K
KAMEZAWA Hiroyuki 已提交
3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063
			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;
3064
	}
K
KAMEZAWA Hiroyuki 已提交
3065

3066
	mem_cgroup_charge_statistics(memcg, PageCgroupCache(pc), -nr_pages);
K
KAMEZAWA Hiroyuki 已提交
3067

3068
	ClearPageCgroupUsed(pc);
3069 3070 3071 3072 3073 3074
	/*
	 * 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.
	 */
3075

3076
	unlock_page_cgroup(pc);
K
KAMEZAWA Hiroyuki 已提交
3077
	/*
3078
	 * even after unlock, we have memcg->res.usage here and this memcg
K
KAMEZAWA Hiroyuki 已提交
3079 3080
	 * will never be freed.
	 */
3081
	memcg_check_events(memcg, page);
K
KAMEZAWA Hiroyuki 已提交
3082
	if (do_swap_account && ctype == MEM_CGROUP_CHARGE_TYPE_SWAPOUT) {
3083 3084
		mem_cgroup_swap_statistics(memcg, true);
		mem_cgroup_get(memcg);
K
KAMEZAWA Hiroyuki 已提交
3085
	}
3086 3087
	if (!mem_cgroup_is_root(memcg))
		mem_cgroup_do_uncharge(memcg, nr_pages, ctype);
3088

3089
	return memcg;
K
KAMEZAWA Hiroyuki 已提交
3090 3091 3092

unlock_out:
	unlock_page_cgroup(pc);
3093
	return NULL;
3094 3095
}

3096 3097
void mem_cgroup_uncharge_page(struct page *page)
{
3098 3099 3100 3101 3102
	/* early check. */
	if (page_mapped(page))
		return;
	if (page->mapping && !PageAnon(page))
		return;
3103 3104 3105 3106 3107 3108
	__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));
3109
	VM_BUG_ON(page->mapping);
3110 3111 3112
	__mem_cgroup_uncharge_common(page, MEM_CGROUP_CHARGE_TYPE_CACHE);
}

3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126
/*
 * 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;
3127 3128
		current->memcg_batch.nr_pages = 0;
		current->memcg_batch.memsw_nr_pages = 0;
3129 3130 3131 3132 3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148
	}
}

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.
	 */
3149 3150 3151 3152 3153 3154
	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);
3155
	memcg_oom_recover(batch->memcg);
3156 3157 3158 3159
	/* forget this pointer (for sanity check) */
	batch->memcg = NULL;
}

3160
#ifdef CONFIG_SWAP
3161
/*
3162
 * called after __delete_from_swap_cache() and drop "page" account.
3163 3164
 * memcg information is recorded to swap_cgroup of "ent"
 */
K
KAMEZAWA Hiroyuki 已提交
3165 3166
void
mem_cgroup_uncharge_swapcache(struct page *page, swp_entry_t ent, bool swapout)
3167 3168
{
	struct mem_cgroup *memcg;
K
KAMEZAWA Hiroyuki 已提交
3169 3170 3171 3172 3173 3174
	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);
3175

K
KAMEZAWA Hiroyuki 已提交
3176 3177 3178 3179 3180
	/*
	 * record memcg information,  if swapout && memcg != NULL,
	 * mem_cgroup_get() was called in uncharge().
	 */
	if (do_swap_account && swapout && memcg)
3181
		swap_cgroup_record(ent, css_id(&memcg->css));
3182
}
3183
#endif
3184 3185 3186 3187 3188 3189 3190

#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
3191
{
3192
	struct mem_cgroup *memcg;
3193
	unsigned short id;
3194 3195 3196 3197

	if (!do_swap_account)
		return;

3198 3199 3200
	id = swap_cgroup_record(ent, 0);
	rcu_read_lock();
	memcg = mem_cgroup_lookup(id);
3201
	if (memcg) {
3202 3203 3204 3205
		/*
		 * We uncharge this because swap is freed.
		 * This memcg can be obsolete one. We avoid calling css_tryget
		 */
3206
		if (!mem_cgroup_is_root(memcg))
3207
			res_counter_uncharge(&memcg->memsw, PAGE_SIZE);
3208
		mem_cgroup_swap_statistics(memcg, false);
3209 3210
		mem_cgroup_put(memcg);
	}
3211
	rcu_read_unlock();
K
KAMEZAWA Hiroyuki 已提交
3212
}
3213 3214 3215 3216 3217 3218

/**
 * 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
3219
 * @need_fixup: whether we should fixup res_counters and refcounts.
3220 3221 3222 3223 3224 3225 3226 3227 3228 3229
 *
 * 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,
3230
		struct mem_cgroup *from, struct mem_cgroup *to, bool need_fixup)
3231 3232 3233 3234 3235 3236 3237 3238
{
	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);
3239
		mem_cgroup_swap_statistics(to, true);
3240
		/*
3241 3242 3243 3244 3245 3246
		 * 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.
3247 3248
		 */
		mem_cgroup_get(to);
3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259
		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);
		}
3260 3261 3262 3263 3264 3265
		return 0;
	}
	return -EINVAL;
}
#else
static inline int mem_cgroup_move_swap_account(swp_entry_t entry,
3266
		struct mem_cgroup *from, struct mem_cgroup *to, bool need_fixup)
3267 3268 3269
{
	return -EINVAL;
}
3270
#endif
K
KAMEZAWA Hiroyuki 已提交
3271

3272
/*
3273 3274
 * Before starting migration, account PAGE_SIZE to mem_cgroup that the old
 * page belongs to.
3275
 */
3276
int mem_cgroup_prepare_migration(struct page *page,
3277
	struct page *newpage, struct mem_cgroup **ptr, gfp_t gfp_mask)
3278
{
3279
	struct mem_cgroup *memcg = NULL;
3280
	struct page_cgroup *pc;
3281
	enum charge_type ctype;
3282
	int ret = 0;
3283

3284 3285
	*ptr = NULL;

A
Andrea Arcangeli 已提交
3286
	VM_BUG_ON(PageTransHuge(page));
3287
	if (mem_cgroup_disabled())
3288 3289
		return 0;

3290 3291 3292
	pc = lookup_page_cgroup(page);
	lock_page_cgroup(pc);
	if (PageCgroupUsed(pc)) {
3293 3294
		memcg = pc->mem_cgroup;
		css_get(&memcg->css);
3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325
		/*
		 * 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);
3326
	}
3327
	unlock_page_cgroup(pc);
3328 3329 3330 3331
	/*
	 * If the page is not charged at this point,
	 * we return here.
	 */
3332
	if (!memcg)
3333
		return 0;
3334

3335
	*ptr = memcg;
3336
	ret = __mem_cgroup_try_charge(NULL, gfp_mask, 1, ptr, false);
3337
	css_put(&memcg->css);/* drop extra refcnt */
3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348
	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;
3349
	}
3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362
	/*
	 * 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;
3363
	__mem_cgroup_commit_charge(memcg, page, 1, pc, ctype);
3364
	return ret;
3365
}
3366

3367
/* remove redundant charge if migration failed*/
3368
void mem_cgroup_end_migration(struct mem_cgroup *memcg,
3369
	struct page *oldpage, struct page *newpage, bool migration_ok)
3370
{
3371
	struct page *used, *unused;
3372 3373
	struct page_cgroup *pc;

3374
	if (!memcg)
3375
		return;
3376
	/* blocks rmdir() */
3377
	cgroup_exclude_rmdir(&memcg->css);
3378
	if (!migration_ok) {
3379 3380
		used = oldpage;
		unused = newpage;
3381
	} else {
3382
		used = newpage;
3383 3384
		unused = oldpage;
	}
3385
	/*
3386 3387 3388
	 * 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.
3389
	 */
3390 3391 3392 3393
	pc = lookup_page_cgroup(oldpage);
	lock_page_cgroup(pc);
	ClearPageCgroupMigration(pc);
	unlock_page_cgroup(pc);
3394

3395 3396
	__mem_cgroup_uncharge_common(unused, MEM_CGROUP_CHARGE_TYPE_FORCE);

3397
	/*
3398 3399 3400 3401 3402 3403
	 * 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)
3404
	 */
3405 3406
	if (PageAnon(used))
		mem_cgroup_uncharge_page(used);
3407
	/*
3408 3409
	 * At migration, we may charge account against cgroup which has no
	 * tasks.
3410 3411 3412
	 * 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.
	 */
3413
	cgroup_release_and_wakeup_rmdir(&memcg->css);
3414
}
3415

3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459
/*
 * 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);
}

3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505
#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

3506 3507
static DEFINE_MUTEX(set_limit_mutex);

3508
static int mem_cgroup_resize_limit(struct mem_cgroup *memcg,
3509
				unsigned long long val)
3510
{
3511
	int retry_count;
3512
	u64 memswlimit, memlimit;
3513
	int ret = 0;
3514 3515
	int children = mem_cgroup_count_children(memcg);
	u64 curusage, oldusage;
3516
	int enlarge;
3517 3518 3519 3520 3521 3522 3523 3524 3525

	/*
	 * 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);
3526

3527
	enlarge = 0;
3528
	while (retry_count) {
3529 3530 3531 3532
		if (signal_pending(current)) {
			ret = -EINTR;
			break;
		}
3533 3534 3535
		/*
		 * Rather than hide all in some function, I do this in
		 * open coded manner. You see what this really does.
3536
		 * We have to guarantee memcg->res.limit < memcg->memsw.limit.
3537 3538 3539 3540 3541 3542
		 */
		mutex_lock(&set_limit_mutex);
		memswlimit = res_counter_read_u64(&memcg->memsw, RES_LIMIT);
		if (memswlimit < val) {
			ret = -EINVAL;
			mutex_unlock(&set_limit_mutex);
3543 3544
			break;
		}
3545 3546 3547 3548 3549

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

3550
		ret = res_counter_set_limit(&memcg->res, val);
3551 3552 3553 3554 3555 3556
		if (!ret) {
			if (memswlimit == val)
				memcg->memsw_is_minimum = true;
			else
				memcg->memsw_is_minimum = false;
		}
3557 3558 3559 3560 3561
		mutex_unlock(&set_limit_mutex);

		if (!ret)
			break;

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

3575 3576 3577
	return ret;
}

L
Li Zefan 已提交
3578 3579
static int mem_cgroup_resize_memsw_limit(struct mem_cgroup *memcg,
					unsigned long long val)
3580
{
3581
	int retry_count;
3582
	u64 memlimit, memswlimit, oldusage, curusage;
3583 3584
	int children = mem_cgroup_count_children(memcg);
	int ret = -EBUSY;
3585
	int enlarge = 0;
3586

3587 3588 3589
	/* see mem_cgroup_resize_res_limit */
 	retry_count = children * MEM_CGROUP_RECLAIM_RETRIES;
	oldusage = res_counter_read_u64(&memcg->memsw, RES_USAGE);
3590 3591 3592 3593 3594 3595 3596 3597
	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.
3598
		 * We have to guarantee memcg->res.limit < memcg->memsw.limit.
3599 3600 3601 3602 3603 3604 3605 3606
		 */
		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;
		}
3607 3608 3609
		memswlimit = res_counter_read_u64(&memcg->memsw, RES_LIMIT);
		if (memswlimit < val)
			enlarge = 1;
3610
		ret = res_counter_set_limit(&memcg->memsw, val);
3611 3612 3613 3614 3615 3616
		if (!ret) {
			if (memlimit == val)
				memcg->memsw_is_minimum = true;
			else
				memcg->memsw_is_minimum = false;
		}
3617 3618 3619 3620 3621
		mutex_unlock(&set_limit_mutex);

		if (!ret)
			break;

3622
		mem_cgroup_hierarchical_reclaim(memcg, NULL, GFP_KERNEL,
3623
						MEM_CGROUP_RECLAIM_NOSWAP |
3624 3625
						MEM_CGROUP_RECLAIM_SHRINK,
						NULL);
3626
		curusage = res_counter_read_u64(&memcg->memsw, RES_USAGE);
3627
		/* Usage is reduced ? */
3628
		if (curusage >= oldusage)
3629
			retry_count--;
3630 3631
		else
			oldusage = curusage;
3632
	}
3633 3634
	if (!ret && enlarge)
		memcg_oom_recover(memcg);
3635 3636 3637
	return ret;
}

3638
unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
3639 3640
					    gfp_t gfp_mask,
					    unsigned long *total_scanned)
3641 3642 3643 3644 3645 3646
{
	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;
3647
	unsigned long long excess;
3648
	unsigned long nr_scanned;
3649 3650 3651 3652

	if (order > 0)
		return 0;

3653
	mctz = soft_limit_tree_node_zone(zone_to_nid(zone), zone_idx(zone));
3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664 3665 3666
	/*
	 * 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;

3667
		nr_scanned = 0;
3668 3669
		reclaimed = mem_cgroup_hierarchical_reclaim(mz->mem, zone,
						gfp_mask,
3670 3671
						MEM_CGROUP_RECLAIM_SOFT,
						&nr_scanned);
3672
		nr_reclaimed += reclaimed;
3673
		*total_scanned += nr_scanned;
3674 3675 3676 3677 3678 3679 3680 3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695
		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);
3696
				if (next_mz == mz)
3697
					css_put(&next_mz->mem->css);
3698
				else /* next_mz == NULL or other memcg */
3699 3700 3701 3702
					break;
			} while (1);
		}
		__mem_cgroup_remove_exceeded(mz->mem, mz, mctz);
3703
		excess = res_counter_soft_limit_excess(&mz->mem->res);
3704 3705 3706 3707 3708 3709 3710 3711
		/*
		 * 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.
		 */
3712 3713
		/* If excess == 0, no tree ops */
		__mem_cgroup_insert_exceeded(mz->mem, mz, mctz, excess);
3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731
		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;
}

3732 3733 3734 3735
/*
 * This routine traverse page_cgroup in given list and drop them all.
 * *And* this routine doesn't reclaim page itself, just removes page_cgroup.
 */
3736
static int mem_cgroup_force_empty_list(struct mem_cgroup *memcg,
K
KAMEZAWA Hiroyuki 已提交
3737
				int node, int zid, enum lru_list lru)
3738
{
K
KAMEZAWA Hiroyuki 已提交
3739 3740
	struct zone *zone;
	struct mem_cgroup_per_zone *mz;
3741
	struct page_cgroup *pc, *busy;
K
KAMEZAWA Hiroyuki 已提交
3742
	unsigned long flags, loop;
3743
	struct list_head *list;
3744
	int ret = 0;
3745

K
KAMEZAWA Hiroyuki 已提交
3746
	zone = &NODE_DATA(node)->node_zones[zid];
3747
	mz = mem_cgroup_zoneinfo(memcg, node, zid);
3748
	list = &mz->lists[lru];
3749

3750 3751 3752 3753 3754
	loop = MEM_CGROUP_ZSTAT(mz, lru);
	/* give some margin against EBUSY etc...*/
	loop += 256;
	busy = NULL;
	while (loop--) {
3755 3756
		struct page *page;

3757
		ret = 0;
K
KAMEZAWA Hiroyuki 已提交
3758
		spin_lock_irqsave(&zone->lru_lock, flags);
3759
		if (list_empty(list)) {
K
KAMEZAWA Hiroyuki 已提交
3760
			spin_unlock_irqrestore(&zone->lru_lock, flags);
3761
			break;
3762 3763 3764 3765
		}
		pc = list_entry(list->prev, struct page_cgroup, lru);
		if (busy == pc) {
			list_move(&pc->lru, list);
3766
			busy = NULL;
K
KAMEZAWA Hiroyuki 已提交
3767
			spin_unlock_irqrestore(&zone->lru_lock, flags);
3768 3769
			continue;
		}
K
KAMEZAWA Hiroyuki 已提交
3770
		spin_unlock_irqrestore(&zone->lru_lock, flags);
3771

3772
		page = lookup_cgroup_page(pc);
3773

3774
		ret = mem_cgroup_move_parent(page, pc, memcg, GFP_KERNEL);
3775
		if (ret == -ENOMEM)
3776
			break;
3777 3778 3779 3780 3781 3782 3783

		if (ret == -EBUSY || ret == -EINVAL) {
			/* found lock contention or "pc" is obsolete. */
			busy = pc;
			cond_resched();
		} else
			busy = NULL;
3784
	}
K
KAMEZAWA Hiroyuki 已提交
3785

3786 3787 3788
	if (!ret && !list_empty(list))
		return -EBUSY;
	return ret;
3789 3790 3791 3792 3793 3794
}

/*
 * make mem_cgroup's charge to be 0 if there is no task.
 * This enables deleting this mem_cgroup.
 */
3795
static int mem_cgroup_force_empty(struct mem_cgroup *memcg, bool free_all)
3796
{
3797 3798 3799
	int ret;
	int node, zid, shrink;
	int nr_retries = MEM_CGROUP_RECLAIM_RETRIES;
3800
	struct cgroup *cgrp = memcg->css.cgroup;
3801

3802
	css_get(&memcg->css);
3803 3804

	shrink = 0;
3805 3806 3807
	/* should free all ? */
	if (free_all)
		goto try_to_free;
3808
move_account:
3809
	do {
3810
		ret = -EBUSY;
3811 3812 3813 3814
		if (cgroup_task_count(cgrp) || !list_empty(&cgrp->children))
			goto out;
		ret = -EINTR;
		if (signal_pending(current))
3815
			goto out;
3816 3817
		/* This is for making all *used* pages to be on LRU. */
		lru_add_drain_all();
3818
		drain_all_stock_sync(memcg);
3819
		ret = 0;
3820
		mem_cgroup_start_move(memcg);
3821
		for_each_node_state(node, N_HIGH_MEMORY) {
3822
			for (zid = 0; !ret && zid < MAX_NR_ZONES; zid++) {
3823
				enum lru_list l;
3824
				for_each_lru(l) {
3825
					ret = mem_cgroup_force_empty_list(memcg,
K
KAMEZAWA Hiroyuki 已提交
3826
							node, zid, l);
3827 3828 3829
					if (ret)
						break;
				}
3830
			}
3831 3832 3833
			if (ret)
				break;
		}
3834 3835
		mem_cgroup_end_move(memcg);
		memcg_oom_recover(memcg);
3836 3837 3838
		/* it seems parent cgroup doesn't have enough mem */
		if (ret == -ENOMEM)
			goto try_to_free;
3839
		cond_resched();
3840
	/* "ret" should also be checked to ensure all lists are empty. */
3841
	} while (memcg->res.usage > 0 || ret);
3842
out:
3843
	css_put(&memcg->css);
3844
	return ret;
3845 3846

try_to_free:
3847 3848
	/* returns EBUSY if there is a task or if we come here twice. */
	if (cgroup_task_count(cgrp) || !list_empty(&cgrp->children) || shrink) {
3849 3850 3851
		ret = -EBUSY;
		goto out;
	}
3852 3853
	/* we call try-to-free pages for make this cgroup empty */
	lru_add_drain_all();
3854 3855
	/* try to free all pages in this cgroup */
	shrink = 1;
3856
	while (nr_retries && memcg->res.usage > 0) {
3857
		int progress;
3858 3859 3860 3861 3862

		if (signal_pending(current)) {
			ret = -EINTR;
			goto out;
		}
3863
		progress = try_to_free_mem_cgroup_pages(memcg, GFP_KERNEL,
3864
						false);
3865
		if (!progress) {
3866
			nr_retries--;
3867
			/* maybe some writeback is necessary */
3868
			congestion_wait(BLK_RW_ASYNC, HZ/10);
3869
		}
3870 3871

	}
K
KAMEZAWA Hiroyuki 已提交
3872
	lru_add_drain();
3873
	/* try move_account...there may be some *locked* pages. */
3874
	goto move_account;
3875 3876
}

3877 3878 3879 3880 3881 3882
int mem_cgroup_force_empty_write(struct cgroup *cont, unsigned int event)
{
	return mem_cgroup_force_empty(mem_cgroup_from_cont(cont), true);
}


3883 3884 3885 3886 3887 3888 3889 3890 3891
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;
3892
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
3893
	struct cgroup *parent = cont->parent;
3894
	struct mem_cgroup *parent_memcg = NULL;
3895 3896

	if (parent)
3897
		parent_memcg = mem_cgroup_from_cont(parent);
3898 3899 3900

	cgroup_lock();
	/*
3901
	 * If parent's use_hierarchy is set, we can't make any modifications
3902 3903 3904 3905 3906 3907
	 * 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.
	 */
3908
	if ((!parent_memcg || !parent_memcg->use_hierarchy) &&
3909 3910
				(val == 1 || val == 0)) {
		if (list_empty(&cont->children))
3911
			memcg->use_hierarchy = val;
3912 3913 3914 3915 3916 3917 3918 3919 3920
		else
			retval = -EBUSY;
	} else
		retval = -EINVAL;
	cgroup_unlock();

	return retval;
}

3921

3922
static unsigned long mem_cgroup_recursive_stat(struct mem_cgroup *memcg,
3923
					       enum mem_cgroup_stat_index idx)
3924
{
K
KAMEZAWA Hiroyuki 已提交
3925
	struct mem_cgroup *iter;
3926
	long val = 0;
3927

3928
	/* Per-cpu values can be negative, use a signed accumulator */
3929
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
3930 3931 3932 3933 3934
		val += mem_cgroup_read_stat(iter, idx);

	if (val < 0) /* race ? */
		val = 0;
	return val;
3935 3936
}

3937
static inline u64 mem_cgroup_usage(struct mem_cgroup *memcg, bool swap)
3938
{
K
KAMEZAWA Hiroyuki 已提交
3939
	u64 val;
3940

3941
	if (!mem_cgroup_is_root(memcg)) {
3942
		if (!swap)
3943
			return res_counter_read_u64(&memcg->res, RES_USAGE);
3944
		else
3945
			return res_counter_read_u64(&memcg->memsw, RES_USAGE);
3946 3947
	}

3948 3949
	val = mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_CACHE);
	val += mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_RSS);
3950

K
KAMEZAWA Hiroyuki 已提交
3951
	if (swap)
3952
		val += mem_cgroup_recursive_stat(memcg, MEM_CGROUP_STAT_SWAPOUT);
3953 3954 3955 3956

	return val << PAGE_SHIFT;
}

3957
static u64 mem_cgroup_read(struct cgroup *cont, struct cftype *cft)
B
Balbir Singh 已提交
3958
{
3959
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
3960
	u64 val;
3961 3962 3963 3964 3965 3966
	int type, name;

	type = MEMFILE_TYPE(cft->private);
	name = MEMFILE_ATTR(cft->private);
	switch (type) {
	case _MEM:
3967
		if (name == RES_USAGE)
3968
			val = mem_cgroup_usage(memcg, false);
3969
		else
3970
			val = res_counter_read_u64(&memcg->res, name);
3971 3972
		break;
	case _MEMSWAP:
3973
		if (name == RES_USAGE)
3974
			val = mem_cgroup_usage(memcg, true);
3975
		else
3976
			val = res_counter_read_u64(&memcg->memsw, name);
3977 3978 3979 3980 3981 3982
		break;
	default:
		BUG();
		break;
	}
	return val;
B
Balbir Singh 已提交
3983
}
3984 3985 3986 3987
/*
 * The user of this function is...
 * RES_LIMIT.
 */
3988 3989
static int mem_cgroup_write(struct cgroup *cont, struct cftype *cft,
			    const char *buffer)
B
Balbir Singh 已提交
3990
{
3991
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
3992
	int type, name;
3993 3994 3995
	unsigned long long val;
	int ret;

3996 3997 3998
	type = MEMFILE_TYPE(cft->private);
	name = MEMFILE_ATTR(cft->private);
	switch (name) {
3999
	case RES_LIMIT:
4000 4001 4002 4003
		if (mem_cgroup_is_root(memcg)) { /* Can't set limit on root */
			ret = -EINVAL;
			break;
		}
4004 4005
		/* This function does all necessary parse...reuse it */
		ret = res_counter_memparse_write_strategy(buffer, &val);
4006 4007 4008
		if (ret)
			break;
		if (type == _MEM)
4009
			ret = mem_cgroup_resize_limit(memcg, val);
4010 4011
		else
			ret = mem_cgroup_resize_memsw_limit(memcg, val);
4012
		break;
4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026
	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;
4027 4028 4029 4030 4031
	default:
		ret = -EINVAL; /* should be BUG() ? */
		break;
	}
	return ret;
B
Balbir Singh 已提交
4032 4033
}

4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061
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;
}

4062
static int mem_cgroup_reset(struct cgroup *cont, unsigned int event)
4063
{
4064
	struct mem_cgroup *memcg;
4065
	int type, name;
4066

4067
	memcg = mem_cgroup_from_cont(cont);
4068 4069 4070
	type = MEMFILE_TYPE(event);
	name = MEMFILE_ATTR(event);
	switch (name) {
4071
	case RES_MAX_USAGE:
4072
		if (type == _MEM)
4073
			res_counter_reset_max(&memcg->res);
4074
		else
4075
			res_counter_reset_max(&memcg->memsw);
4076 4077
		break;
	case RES_FAILCNT:
4078
		if (type == _MEM)
4079
			res_counter_reset_failcnt(&memcg->res);
4080
		else
4081
			res_counter_reset_failcnt(&memcg->memsw);
4082 4083
		break;
	}
4084

4085
	return 0;
4086 4087
}

4088 4089 4090 4091 4092 4093
static u64 mem_cgroup_move_charge_read(struct cgroup *cgrp,
					struct cftype *cft)
{
	return mem_cgroup_from_cont(cgrp)->move_charge_at_immigrate;
}

4094
#ifdef CONFIG_MMU
4095 4096 4097
static int mem_cgroup_move_charge_write(struct cgroup *cgrp,
					struct cftype *cft, u64 val)
{
4098
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4099 4100 4101 4102 4103 4104 4105 4106 4107

	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();
4108
	memcg->move_charge_at_immigrate = val;
4109 4110 4111 4112
	cgroup_unlock();

	return 0;
}
4113 4114 4115 4116 4117 4118 4119
#else
static int mem_cgroup_move_charge_write(struct cgroup *cgrp,
					struct cftype *cft, u64 val)
{
	return -ENOSYS;
}
#endif
4120

K
KAMEZAWA Hiroyuki 已提交
4121 4122 4123 4124 4125

/* For read statistics */
enum {
	MCS_CACHE,
	MCS_RSS,
4126
	MCS_FILE_MAPPED,
K
KAMEZAWA Hiroyuki 已提交
4127 4128
	MCS_PGPGIN,
	MCS_PGPGOUT,
4129
	MCS_SWAP,
4130 4131
	MCS_PGFAULT,
	MCS_PGMAJFAULT,
K
KAMEZAWA Hiroyuki 已提交
4132 4133 4134 4135 4136 4137 4138 4139 4140 4141
	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];
4142 4143
};

K
KAMEZAWA Hiroyuki 已提交
4144 4145 4146 4147 4148 4149
struct {
	char *local_name;
	char *total_name;
} memcg_stat_strings[NR_MCS_STAT] = {
	{"cache", "total_cache"},
	{"rss", "total_rss"},
4150
	{"mapped_file", "total_mapped_file"},
K
KAMEZAWA Hiroyuki 已提交
4151 4152
	{"pgpgin", "total_pgpgin"},
	{"pgpgout", "total_pgpgout"},
4153
	{"swap", "total_swap"},
4154 4155
	{"pgfault", "total_pgfault"},
	{"pgmajfault", "total_pgmajfault"},
K
KAMEZAWA Hiroyuki 已提交
4156 4157 4158 4159 4160 4161 4162 4163
	{"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
4164
static void
4165
mem_cgroup_get_local_stat(struct mem_cgroup *memcg, struct mcs_total_stat *s)
K
KAMEZAWA Hiroyuki 已提交
4166 4167 4168 4169
{
	s64 val;

	/* per cpu stat */
4170
	val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_CACHE);
K
KAMEZAWA Hiroyuki 已提交
4171
	s->stat[MCS_CACHE] += val * PAGE_SIZE;
4172
	val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_RSS);
K
KAMEZAWA Hiroyuki 已提交
4173
	s->stat[MCS_RSS] += val * PAGE_SIZE;
4174
	val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_FILE_MAPPED);
4175
	s->stat[MCS_FILE_MAPPED] += val * PAGE_SIZE;
4176
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGPGIN);
K
KAMEZAWA Hiroyuki 已提交
4177
	s->stat[MCS_PGPGIN] += val;
4178
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGPGOUT);
K
KAMEZAWA Hiroyuki 已提交
4179
	s->stat[MCS_PGPGOUT] += val;
4180
	if (do_swap_account) {
4181
		val = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_SWAPOUT);
4182 4183
		s->stat[MCS_SWAP] += val * PAGE_SIZE;
	}
4184
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGFAULT);
4185
	s->stat[MCS_PGFAULT] += val;
4186
	val = mem_cgroup_read_events(memcg, MEM_CGROUP_EVENTS_PGMAJFAULT);
4187
	s->stat[MCS_PGMAJFAULT] += val;
K
KAMEZAWA Hiroyuki 已提交
4188 4189

	/* per zone stat */
4190
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_INACTIVE_ANON));
K
KAMEZAWA Hiroyuki 已提交
4191
	s->stat[MCS_INACTIVE_ANON] += val * PAGE_SIZE;
4192
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_ACTIVE_ANON));
K
KAMEZAWA Hiroyuki 已提交
4193
	s->stat[MCS_ACTIVE_ANON] += val * PAGE_SIZE;
4194
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_INACTIVE_FILE));
K
KAMEZAWA Hiroyuki 已提交
4195
	s->stat[MCS_INACTIVE_FILE] += val * PAGE_SIZE;
4196
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_ACTIVE_FILE));
K
KAMEZAWA Hiroyuki 已提交
4197
	s->stat[MCS_ACTIVE_FILE] += val * PAGE_SIZE;
4198
	val = mem_cgroup_nr_lru_pages(memcg, BIT(LRU_UNEVICTABLE));
K
KAMEZAWA Hiroyuki 已提交
4199 4200 4201 4202
	s->stat[MCS_UNEVICTABLE] += val * PAGE_SIZE;
}

static void
4203
mem_cgroup_get_total_stat(struct mem_cgroup *memcg, struct mcs_total_stat *s)
K
KAMEZAWA Hiroyuki 已提交
4204
{
K
KAMEZAWA Hiroyuki 已提交
4205 4206
	struct mem_cgroup *iter;

4207
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
4208
		mem_cgroup_get_local_stat(iter, s);
K
KAMEZAWA Hiroyuki 已提交
4209 4210
}

4211 4212 4213 4214 4215 4216 4217 4218 4219
#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);

4220
	total_nr = mem_cgroup_nr_lru_pages(mem_cont, LRU_ALL);
4221 4222
	seq_printf(m, "total=%lu", total_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4223
		node_nr = mem_cgroup_node_nr_lru_pages(mem_cont, nid, LRU_ALL);
4224 4225 4226 4227
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');

4228
	file_nr = mem_cgroup_nr_lru_pages(mem_cont, LRU_ALL_FILE);
4229 4230
	seq_printf(m, "file=%lu", file_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4231 4232
		node_nr = mem_cgroup_node_nr_lru_pages(mem_cont, nid,
				LRU_ALL_FILE);
4233 4234 4235 4236
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');

4237
	anon_nr = mem_cgroup_nr_lru_pages(mem_cont, LRU_ALL_ANON);
4238 4239
	seq_printf(m, "anon=%lu", anon_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4240 4241
		node_nr = mem_cgroup_node_nr_lru_pages(mem_cont, nid,
				LRU_ALL_ANON);
4242 4243 4244 4245
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');

4246
	unevictable_nr = mem_cgroup_nr_lru_pages(mem_cont, BIT(LRU_UNEVICTABLE));
4247 4248
	seq_printf(m, "unevictable=%lu", unevictable_nr);
	for_each_node_state(nid, N_HIGH_MEMORY) {
4249 4250
		node_nr = mem_cgroup_node_nr_lru_pages(mem_cont, nid,
				BIT(LRU_UNEVICTABLE));
4251 4252 4253 4254 4255 4256 4257
		seq_printf(m, " N%d=%lu", nid, node_nr);
	}
	seq_putc(m, '\n');
	return 0;
}
#endif /* CONFIG_NUMA */

4258 4259
static int mem_control_stat_show(struct cgroup *cont, struct cftype *cft,
				 struct cgroup_map_cb *cb)
4260 4261
{
	struct mem_cgroup *mem_cont = mem_cgroup_from_cont(cont);
K
KAMEZAWA Hiroyuki 已提交
4262
	struct mcs_total_stat mystat;
4263 4264
	int i;

K
KAMEZAWA Hiroyuki 已提交
4265 4266
	memset(&mystat, 0, sizeof(mystat));
	mem_cgroup_get_local_stat(mem_cont, &mystat);
4267

4268

4269 4270 4271
	for (i = 0; i < NR_MCS_STAT; i++) {
		if (i == MCS_SWAP && !do_swap_account)
			continue;
K
KAMEZAWA Hiroyuki 已提交
4272
		cb->fill(cb, memcg_stat_strings[i].local_name, mystat.stat[i]);
4273
	}
L
Lee Schermerhorn 已提交
4274

K
KAMEZAWA Hiroyuki 已提交
4275
	/* Hierarchical information */
4276 4277 4278 4279 4280 4281 4282
	{
		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
4283

K
KAMEZAWA Hiroyuki 已提交
4284 4285
	memset(&mystat, 0, sizeof(mystat));
	mem_cgroup_get_total_stat(mem_cont, &mystat);
4286 4287 4288
	for (i = 0; i < NR_MCS_STAT; i++) {
		if (i == MCS_SWAP && !do_swap_account)
			continue;
K
KAMEZAWA Hiroyuki 已提交
4289
		cb->fill(cb, memcg_stat_strings[i].total_name, mystat.stat[i]);
4290
	}
K
KAMEZAWA Hiroyuki 已提交
4291

K
KOSAKI Motohiro 已提交
4292 4293 4294 4295 4296 4297 4298 4299 4300 4301 4302 4303 4304 4305 4306 4307 4308 4309 4310 4311 4312 4313 4314 4315 4316 4317 4318
#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

4319 4320 4321
	return 0;
}

K
KOSAKI Motohiro 已提交
4322 4323 4324 4325
static u64 mem_cgroup_swappiness_read(struct cgroup *cgrp, struct cftype *cft)
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);

4326
	return mem_cgroup_swappiness(memcg);
K
KOSAKI Motohiro 已提交
4327 4328 4329 4330 4331 4332 4333
}

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;
4334

K
KOSAKI Motohiro 已提交
4335 4336 4337 4338 4339 4340 4341
	if (val > 100)
		return -EINVAL;

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

	parent = mem_cgroup_from_cont(cgrp->parent);
4342 4343 4344

	cgroup_lock();

K
KOSAKI Motohiro 已提交
4345 4346
	/* If under hierarchy, only empty-root can set this value */
	if ((parent->use_hierarchy) ||
4347 4348
	    (memcg->use_hierarchy && !list_empty(&cgrp->children))) {
		cgroup_unlock();
K
KOSAKI Motohiro 已提交
4349
		return -EINVAL;
4350
	}
K
KOSAKI Motohiro 已提交
4351 4352 4353

	memcg->swappiness = val;

4354 4355
	cgroup_unlock();

K
KOSAKI Motohiro 已提交
4356 4357 4358
	return 0;
}

4359 4360 4361 4362 4363 4364 4365 4366
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)
4367
		t = rcu_dereference(memcg->thresholds.primary);
4368
	else
4369
		t = rcu_dereference(memcg->memsw_thresholds.primary);
4370 4371 4372 4373 4374 4375 4376 4377 4378 4379 4380

	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().
	 */
4381
	i = t->current_threshold;
4382 4383 4384 4385 4386 4387 4388 4389 4390 4391 4392 4393 4394 4395 4396 4397 4398 4399 4400 4401 4402 4403 4404

	/*
	 * 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 */
4405
	t->current_threshold = i - 1;
4406 4407 4408 4409 4410 4411
unlock:
	rcu_read_unlock();
}

static void mem_cgroup_threshold(struct mem_cgroup *memcg)
{
4412 4413 4414 4415 4416 4417 4418
	while (memcg) {
		__mem_cgroup_threshold(memcg, false);
		if (do_swap_account)
			__mem_cgroup_threshold(memcg, true);

		memcg = parent_mem_cgroup(memcg);
	}
4419 4420 4421 4422 4423 4424 4425 4426 4427 4428
}

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;
}

4429
static int mem_cgroup_oom_notify_cb(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
4430 4431 4432
{
	struct mem_cgroup_eventfd_list *ev;

4433
	list_for_each_entry(ev, &memcg->oom_notify, list)
K
KAMEZAWA Hiroyuki 已提交
4434 4435 4436 4437
		eventfd_signal(ev->eventfd, 1);
	return 0;
}

4438
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
4439
{
K
KAMEZAWA Hiroyuki 已提交
4440 4441
	struct mem_cgroup *iter;

4442
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
4443
		mem_cgroup_oom_notify_cb(iter);
K
KAMEZAWA Hiroyuki 已提交
4444 4445 4446 4447
}

static int mem_cgroup_usage_register_event(struct cgroup *cgrp,
	struct cftype *cft, struct eventfd_ctx *eventfd, const char *args)
4448 4449
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4450 4451
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
4452 4453
	int type = MEMFILE_TYPE(cft->private);
	u64 threshold, usage;
4454
	int i, size, ret;
4455 4456 4457 4458 4459 4460

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

	mutex_lock(&memcg->thresholds_lock);
4461

4462
	if (type == _MEM)
4463
		thresholds = &memcg->thresholds;
4464
	else if (type == _MEMSWAP)
4465
		thresholds = &memcg->memsw_thresholds;
4466 4467 4468 4469 4470 4471
	else
		BUG();

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

	/* Check if a threshold crossed before adding a new one */
4472
	if (thresholds->primary)
4473 4474
		__mem_cgroup_threshold(memcg, type == _MEMSWAP);

4475
	size = thresholds->primary ? thresholds->primary->size + 1 : 1;
4476 4477

	/* Allocate memory for new array of thresholds */
4478
	new = kmalloc(sizeof(*new) + size * sizeof(struct mem_cgroup_threshold),
4479
			GFP_KERNEL);
4480
	if (!new) {
4481 4482 4483
		ret = -ENOMEM;
		goto unlock;
	}
4484
	new->size = size;
4485 4486

	/* Copy thresholds (if any) to new array */
4487 4488
	if (thresholds->primary) {
		memcpy(new->entries, thresholds->primary->entries, (size - 1) *
4489
				sizeof(struct mem_cgroup_threshold));
4490 4491
	}

4492
	/* Add new threshold */
4493 4494
	new->entries[size - 1].eventfd = eventfd;
	new->entries[size - 1].threshold = threshold;
4495 4496

	/* Sort thresholds. Registering of new threshold isn't time-critical */
4497
	sort(new->entries, size, sizeof(struct mem_cgroup_threshold),
4498 4499 4500
			compare_thresholds, NULL);

	/* Find current threshold */
4501
	new->current_threshold = -1;
4502
	for (i = 0; i < size; i++) {
4503
		if (new->entries[i].threshold < usage) {
4504
			/*
4505 4506
			 * new->current_threshold will not be used until
			 * rcu_assign_pointer(), so it's safe to increment
4507 4508
			 * it here.
			 */
4509
			++new->current_threshold;
4510 4511 4512
		}
	}

4513 4514 4515 4516 4517
	/* Free old spare buffer and save old primary buffer as spare */
	kfree(thresholds->spare);
	thresholds->spare = thresholds->primary;

	rcu_assign_pointer(thresholds->primary, new);
4518

4519
	/* To be sure that nobody uses thresholds */
4520 4521 4522 4523 4524 4525 4526 4527
	synchronize_rcu();

unlock:
	mutex_unlock(&memcg->thresholds_lock);

	return ret;
}

4528
static void mem_cgroup_usage_unregister_event(struct cgroup *cgrp,
K
KAMEZAWA Hiroyuki 已提交
4529
	struct cftype *cft, struct eventfd_ctx *eventfd)
4530 4531
{
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4532 4533
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
4534 4535
	int type = MEMFILE_TYPE(cft->private);
	u64 usage;
4536
	int i, j, size;
4537 4538 4539

	mutex_lock(&memcg->thresholds_lock);
	if (type == _MEM)
4540
		thresholds = &memcg->thresholds;
4541
	else if (type == _MEMSWAP)
4542
		thresholds = &memcg->memsw_thresholds;
4543 4544 4545 4546 4547 4548 4549 4550 4551 4552 4553 4554 4555 4556 4557
	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 */
4558 4559 4560
	size = 0;
	for (i = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd != eventfd)
4561 4562 4563
			size++;
	}

4564
	new = thresholds->spare;
4565

4566 4567
	/* Set thresholds array to NULL if we don't have thresholds */
	if (!size) {
4568 4569
		kfree(new);
		new = NULL;
4570
		goto swap_buffers;
4571 4572
	}

4573
	new->size = size;
4574 4575

	/* Copy thresholds and find current threshold */
4576 4577 4578
	new->current_threshold = -1;
	for (i = 0, j = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd == eventfd)
4579 4580
			continue;

4581 4582
		new->entries[j] = thresholds->primary->entries[i];
		if (new->entries[j].threshold < usage) {
4583
			/*
4584
			 * new->current_threshold will not be used
4585 4586 4587
			 * until rcu_assign_pointer(), so it's safe to increment
			 * it here.
			 */
4588
			++new->current_threshold;
4589 4590 4591 4592
		}
		j++;
	}

4593
swap_buffers:
4594 4595 4596
	/* Swap primary and spare array */
	thresholds->spare = thresholds->primary;
	rcu_assign_pointer(thresholds->primary, new);
4597

4598
	/* To be sure that nobody uses thresholds */
4599 4600 4601 4602
	synchronize_rcu();

	mutex_unlock(&memcg->thresholds_lock);
}
4603

K
KAMEZAWA Hiroyuki 已提交
4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615
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;

4616
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4617 4618 4619 4620 4621

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

	/* already in OOM ? */
4622
	if (atomic_read(&memcg->under_oom))
K
KAMEZAWA Hiroyuki 已提交
4623
		eventfd_signal(eventfd, 1);
4624
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4625 4626 4627 4628

	return 0;
}

4629
static void mem_cgroup_oom_unregister_event(struct cgroup *cgrp,
K
KAMEZAWA Hiroyuki 已提交
4630 4631
	struct cftype *cft, struct eventfd_ctx *eventfd)
{
4632
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
K
KAMEZAWA Hiroyuki 已提交
4633 4634 4635 4636 4637
	struct mem_cgroup_eventfd_list *ev, *tmp;
	int type = MEMFILE_TYPE(cft->private);

	BUG_ON(type != _OOM_TYPE);

4638
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4639

4640
	list_for_each_entry_safe(ev, tmp, &memcg->oom_notify, list) {
K
KAMEZAWA Hiroyuki 已提交
4641 4642 4643 4644 4645 4646
		if (ev->eventfd == eventfd) {
			list_del(&ev->list);
			kfree(ev);
		}
	}

4647
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4648 4649
}

4650 4651 4652
static int mem_cgroup_oom_control_read(struct cgroup *cgrp,
	struct cftype *cft,  struct cgroup_map_cb *cb)
{
4653
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4654

4655
	cb->fill(cb, "oom_kill_disable", memcg->oom_kill_disable);
4656

4657
	if (atomic_read(&memcg->under_oom))
4658 4659 4660 4661 4662 4663 4664 4665 4666
		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)
{
4667
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp);
4668 4669 4670 4671 4672 4673 4674 4675 4676 4677 4678
	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) ||
4679
	    (memcg->use_hierarchy && !list_empty(&cgrp->children))) {
4680 4681 4682
		cgroup_unlock();
		return -EINVAL;
	}
4683
	memcg->oom_kill_disable = val;
4684
	if (!val)
4685
		memcg_oom_recover(memcg);
4686 4687 4688 4689
	cgroup_unlock();
	return 0;
}

4690 4691 4692 4693 4694 4695 4696 4697 4698 4699 4700 4701 4702 4703 4704 4705
#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 */

4706 4707 4708
#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
4709 4710 4711 4712 4713 4714 4715
	/*
	 * 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
	 */
4716
	return mem_cgroup_sockets_init(cont, ss);
4717 4718
};

G
Glauber Costa 已提交
4719 4720 4721 4722 4723
static void kmem_cgroup_destroy(struct cgroup_subsys *ss,
				struct cgroup *cont)
{
	mem_cgroup_sockets_destroy(cont, ss);
}
4724 4725 4726 4727 4728
#else
static int register_kmem_files(struct cgroup *cont, struct cgroup_subsys *ss)
{
	return 0;
}
G
Glauber Costa 已提交
4729 4730 4731 4732 4733

static void kmem_cgroup_destroy(struct cgroup_subsys *ss,
				struct cgroup *cont)
{
}
4734 4735
#endif

B
Balbir Singh 已提交
4736 4737
static struct cftype mem_cgroup_files[] = {
	{
4738
		.name = "usage_in_bytes",
4739
		.private = MEMFILE_PRIVATE(_MEM, RES_USAGE),
4740
		.read_u64 = mem_cgroup_read,
K
KAMEZAWA Hiroyuki 已提交
4741 4742
		.register_event = mem_cgroup_usage_register_event,
		.unregister_event = mem_cgroup_usage_unregister_event,
B
Balbir Singh 已提交
4743
	},
4744 4745
	{
		.name = "max_usage_in_bytes",
4746
		.private = MEMFILE_PRIVATE(_MEM, RES_MAX_USAGE),
4747
		.trigger = mem_cgroup_reset,
4748 4749
		.read_u64 = mem_cgroup_read,
	},
B
Balbir Singh 已提交
4750
	{
4751
		.name = "limit_in_bytes",
4752
		.private = MEMFILE_PRIVATE(_MEM, RES_LIMIT),
4753
		.write_string = mem_cgroup_write,
4754
		.read_u64 = mem_cgroup_read,
B
Balbir Singh 已提交
4755
	},
4756 4757 4758 4759 4760 4761
	{
		.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
4762 4763
	{
		.name = "failcnt",
4764
		.private = MEMFILE_PRIVATE(_MEM, RES_FAILCNT),
4765
		.trigger = mem_cgroup_reset,
4766
		.read_u64 = mem_cgroup_read,
B
Balbir Singh 已提交
4767
	},
4768 4769
	{
		.name = "stat",
4770
		.read_map = mem_control_stat_show,
4771
	},
4772 4773 4774 4775
	{
		.name = "force_empty",
		.trigger = mem_cgroup_force_empty_write,
	},
4776 4777 4778 4779 4780
	{
		.name = "use_hierarchy",
		.write_u64 = mem_cgroup_hierarchy_write,
		.read_u64 = mem_cgroup_hierarchy_read,
	},
K
KOSAKI Motohiro 已提交
4781 4782 4783 4784 4785
	{
		.name = "swappiness",
		.read_u64 = mem_cgroup_swappiness_read,
		.write_u64 = mem_cgroup_swappiness_write,
	},
4786 4787 4788 4789 4790
	{
		.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
4791 4792
	{
		.name = "oom_control",
4793 4794
		.read_map = mem_cgroup_oom_control_read,
		.write_u64 = mem_cgroup_oom_control_write,
K
KAMEZAWA Hiroyuki 已提交
4795 4796 4797 4798
		.register_event = mem_cgroup_oom_register_event,
		.unregister_event = mem_cgroup_oom_unregister_event,
		.private = MEMFILE_PRIVATE(_OOM_TYPE, OOM_CONTROL),
	},
4799 4800 4801 4802
#ifdef CONFIG_NUMA
	{
		.name = "numa_stat",
		.open = mem_control_numa_stat_open,
4803
		.mode = S_IRUGO,
4804 4805
	},
#endif
B
Balbir Singh 已提交
4806 4807
};

4808 4809 4810 4811 4812 4813
#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
4814 4815
		.register_event = mem_cgroup_usage_register_event,
		.unregister_event = mem_cgroup_usage_unregister_event,
4816 4817 4818 4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836 4837 4838 4839 4840 4841 4842 4843 4844 4845 4846 4847 4848 4849 4850
	},
	{
		.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

4851
static int alloc_mem_cgroup_per_zone_info(struct mem_cgroup *memcg, int node)
4852 4853
{
	struct mem_cgroup_per_node *pn;
4854
	struct mem_cgroup_per_zone *mz;
4855
	enum lru_list l;
4856
	int zone, tmp = node;
4857 4858 4859 4860 4861 4862 4863 4864
	/*
	 * 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.
	 */
4865 4866
	if (!node_state(node, N_NORMAL_MEMORY))
		tmp = -1;
4867
	pn = kzalloc_node(sizeof(*pn), GFP_KERNEL, tmp);
4868 4869
	if (!pn)
		return 1;
4870 4871 4872

	for (zone = 0; zone < MAX_NR_ZONES; zone++) {
		mz = &pn->zoneinfo[zone];
4873 4874
		for_each_lru(l)
			INIT_LIST_HEAD(&mz->lists[l]);
4875
		mz->usage_in_excess = 0;
4876
		mz->on_tree = false;
4877
		mz->mem = memcg;
4878
	}
4879
	memcg->info.nodeinfo[node] = pn;
4880 4881 4882
	return 0;
}

4883
static void free_mem_cgroup_per_zone_info(struct mem_cgroup *memcg, int node)
4884
{
4885
	kfree(memcg->info.nodeinfo[node]);
4886 4887
}

4888 4889 4890
static struct mem_cgroup *mem_cgroup_alloc(void)
{
	struct mem_cgroup *mem;
4891
	int size = sizeof(struct mem_cgroup);
4892

4893
	/* Can be very big if MAX_NUMNODES is very big */
4894
	if (size < PAGE_SIZE)
4895
		mem = kzalloc(size, GFP_KERNEL);
4896
	else
4897
		mem = vzalloc(size);
4898

4899 4900 4901
	if (!mem)
		return NULL;

4902
	mem->stat = alloc_percpu(struct mem_cgroup_stat_cpu);
4903 4904
	if (!mem->stat)
		goto out_free;
4905
	spin_lock_init(&mem->pcp_counter_lock);
4906
	return mem;
4907 4908 4909 4910 4911 4912 4913

out_free:
	if (size < PAGE_SIZE)
		kfree(mem);
	else
		vfree(mem);
	return NULL;
4914 4915
}

4916 4917 4918 4919 4920 4921 4922 4923 4924 4925 4926
/*
 * 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.
 */

4927
static void __mem_cgroup_free(struct mem_cgroup *memcg)
4928
{
K
KAMEZAWA Hiroyuki 已提交
4929 4930
	int node;

4931 4932
	mem_cgroup_remove_from_trees(memcg);
	free_css_id(&mem_cgroup_subsys, &memcg->css);
K
KAMEZAWA Hiroyuki 已提交
4933

K
KAMEZAWA Hiroyuki 已提交
4934
	for_each_node_state(node, N_POSSIBLE)
4935
		free_mem_cgroup_per_zone_info(memcg, node);
K
KAMEZAWA Hiroyuki 已提交
4936

4937
	free_percpu(memcg->stat);
4938
	if (sizeof(struct mem_cgroup) < PAGE_SIZE)
4939
		kfree(memcg);
4940
	else
4941
		vfree(memcg);
4942 4943
}

4944
static void mem_cgroup_get(struct mem_cgroup *memcg)
4945
{
4946
	atomic_inc(&memcg->refcnt);
4947 4948
}

4949
static void __mem_cgroup_put(struct mem_cgroup *memcg, int count)
4950
{
4951 4952 4953
	if (atomic_sub_and_test(count, &memcg->refcnt)) {
		struct mem_cgroup *parent = parent_mem_cgroup(memcg);
		__mem_cgroup_free(memcg);
4954 4955 4956
		if (parent)
			mem_cgroup_put(parent);
	}
4957 4958
}

4959
static void mem_cgroup_put(struct mem_cgroup *memcg)
4960
{
4961
	__mem_cgroup_put(memcg, 1);
4962 4963
}

4964 4965 4966
/*
 * Returns the parent mem_cgroup in memcgroup hierarchy with hierarchy enabled.
 */
G
Glauber Costa 已提交
4967
struct mem_cgroup *parent_mem_cgroup(struct mem_cgroup *memcg)
4968
{
4969
	if (!memcg->res.parent)
4970
		return NULL;
4971
	return mem_cgroup_from_res_counter(memcg->res.parent, res);
4972
}
G
Glauber Costa 已提交
4973
EXPORT_SYMBOL(parent_mem_cgroup);
4974

4975 4976 4977
#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
static void __init enable_swap_cgroup(void)
{
4978
	if (!mem_cgroup_disabled() && really_do_swap_account)
4979 4980 4981 4982 4983 4984 4985 4986
		do_swap_account = 1;
}
#else
static void __init enable_swap_cgroup(void)
{
}
#endif

4987 4988 4989 4990 4991 4992 4993 4994 4995 4996 4997 4998 4999 5000 5001 5002 5003 5004 5005 5006 5007 5008 5009 5010 5011
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
5012
static struct cgroup_subsys_state * __ref
B
Balbir Singh 已提交
5013 5014
mem_cgroup_create(struct cgroup_subsys *ss, struct cgroup *cont)
{
5015
	struct mem_cgroup *memcg, *parent;
K
KAMEZAWA Hiroyuki 已提交
5016
	long error = -ENOMEM;
5017
	int node;
B
Balbir Singh 已提交
5018

5019 5020
	memcg = mem_cgroup_alloc();
	if (!memcg)
K
KAMEZAWA Hiroyuki 已提交
5021
		return ERR_PTR(error);
5022

5023
	for_each_node_state(node, N_POSSIBLE)
5024
		if (alloc_mem_cgroup_per_zone_info(memcg, node))
5025
			goto free_out;
5026

5027
	/* root ? */
5028
	if (cont->parent == NULL) {
5029
		int cpu;
5030
		enable_swap_cgroup();
5031
		parent = NULL;
5032 5033
		if (mem_cgroup_soft_limit_tree_init())
			goto free_out;
5034
		root_mem_cgroup = memcg;
5035 5036 5037 5038 5039
		for_each_possible_cpu(cpu) {
			struct memcg_stock_pcp *stock =
						&per_cpu(memcg_stock, cpu);
			INIT_WORK(&stock->work, drain_local_stock);
		}
5040
		hotcpu_notifier(memcg_cpu_hotplug_callback, 0);
5041
	} else {
5042
		parent = mem_cgroup_from_cont(cont->parent);
5043 5044
		memcg->use_hierarchy = parent->use_hierarchy;
		memcg->oom_kill_disable = parent->oom_kill_disable;
5045
	}
5046

5047
	if (parent && parent->use_hierarchy) {
5048 5049
		res_counter_init(&memcg->res, &parent->res);
		res_counter_init(&memcg->memsw, &parent->memsw);
5050 5051 5052 5053 5054 5055 5056
		/*
		 * 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);
5057
	} else {
5058 5059
		res_counter_init(&memcg->res, NULL);
		res_counter_init(&memcg->memsw, NULL);
5060
	}
5061 5062
	memcg->last_scanned_node = MAX_NUMNODES;
	INIT_LIST_HEAD(&memcg->oom_notify);
5063

K
KOSAKI Motohiro 已提交
5064
	if (parent)
5065 5066 5067 5068 5069
		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;
5070
free_out:
5071
	__mem_cgroup_free(memcg);
K
KAMEZAWA Hiroyuki 已提交
5072
	return ERR_PTR(error);
B
Balbir Singh 已提交
5073 5074
}

5075
static int mem_cgroup_pre_destroy(struct cgroup_subsys *ss,
5076 5077
					struct cgroup *cont)
{
5078
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
5079

5080
	return mem_cgroup_force_empty(memcg, false);
5081 5082
}

B
Balbir Singh 已提交
5083 5084 5085
static void mem_cgroup_destroy(struct cgroup_subsys *ss,
				struct cgroup *cont)
{
5086
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cont);
5087

G
Glauber Costa 已提交
5088 5089
	kmem_cgroup_destroy(ss, cont);

5090
	mem_cgroup_put(memcg);
B
Balbir Singh 已提交
5091 5092 5093 5094 5095
}

static int mem_cgroup_populate(struct cgroup_subsys *ss,
				struct cgroup *cont)
{
5096 5097 5098 5099 5100 5101 5102
	int ret;

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

	if (!ret)
		ret = register_memsw_files(cont, ss);
5103 5104 5105 5106

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

5107
	return ret;
B
Balbir Singh 已提交
5108 5109
}

5110
#ifdef CONFIG_MMU
5111
/* Handlers for move charge at task migration. */
5112 5113
#define PRECHARGE_COUNT_AT_ONCE	256
static int mem_cgroup_do_precharge(unsigned long count)
5114
{
5115 5116
	int ret = 0;
	int batch_count = PRECHARGE_COUNT_AT_ONCE;
5117
	struct mem_cgroup *memcg = mc.to;
5118

5119
	if (mem_cgroup_is_root(memcg)) {
5120 5121 5122 5123 5124 5125 5126 5127
		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;
		/*
5128
		 * "memcg" cannot be under rmdir() because we've already checked
5129 5130 5131 5132
		 * 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().
		 */
5133
		if (res_counter_charge(&memcg->res, PAGE_SIZE * count, &dummy))
5134
			goto one_by_one;
5135
		if (do_swap_account && res_counter_charge(&memcg->memsw,
5136
						PAGE_SIZE * count, &dummy)) {
5137
			res_counter_uncharge(&memcg->res, PAGE_SIZE * count);
5138 5139 5140 5141 5142 5143 5144 5145 5146 5147 5148 5149 5150 5151 5152 5153
			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();
		}
5154 5155 5156
		ret = __mem_cgroup_try_charge(NULL,
					GFP_KERNEL, 1, &memcg, false);
		if (ret || !memcg)
5157 5158 5159 5160
			/* mem_cgroup_clear_mc() will do uncharge later */
			return -ENOMEM;
		mc.precharge++;
	}
5161 5162 5163 5164 5165 5166 5167 5168
	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
5169
 * @target: the pointer the target page or swap ent will be stored(can be NULL)
5170 5171 5172 5173 5174 5175
 *
 * 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).
5176 5177 5178
 *   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.
5179 5180 5181 5182 5183
 *
 * Called with pte lock held.
 */
union mc_target {
	struct page	*page;
5184
	swp_entry_t	ent;
5185 5186 5187 5188 5189
};

enum mc_target_type {
	MC_TARGET_NONE,	/* not used */
	MC_TARGET_PAGE,
5190
	MC_TARGET_SWAP,
5191 5192
};

D
Daisuke Nishimura 已提交
5193 5194
static struct page *mc_handle_present_pte(struct vm_area_struct *vma,
						unsigned long addr, pte_t ptent)
5195
{
D
Daisuke Nishimura 已提交
5196
	struct page *page = vm_normal_page(vma, addr, ptent);
5197

D
Daisuke Nishimura 已提交
5198 5199 5200 5201 5202 5203
	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;
5204 5205
	} else if (!move_file())
		/* we ignore mapcount for file pages */
D
Daisuke Nishimura 已提交
5206 5207 5208 5209 5210 5211 5212 5213 5214 5215 5216 5217 5218 5219 5220 5221 5222 5223
		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 */
5224 5225
		if (page)
			put_page(page);
D
Daisuke Nishimura 已提交
5226
		return NULL;
5227
	}
D
Daisuke Nishimura 已提交
5228 5229 5230 5231 5232 5233
	if (do_swap_account)
		entry->val = ent.val;

	return page;
}

5234 5235 5236 5237 5238 5239 5240 5241 5242 5243 5244 5245 5246 5247 5248 5249 5250 5251 5252 5253 5254
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). */
5255 5256 5257 5258 5259 5260
	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);
5261
		if (do_swap_account)
5262 5263
			*entry = swap;
		page = find_get_page(&swapper_space, swap.val);
5264
	}
5265
#endif
5266 5267 5268
	return page;
}

D
Daisuke Nishimura 已提交
5269 5270 5271 5272 5273 5274 5275 5276 5277 5278 5279 5280
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);
5281 5282
	else if (pte_none(ptent) || pte_file(ptent))
		page = mc_handle_file_pte(vma, addr, ptent, &ent);
D
Daisuke Nishimura 已提交
5283 5284 5285

	if (!page && !ent.val)
		return 0;
5286 5287 5288 5289 5290 5291 5292 5293 5294 5295 5296 5297 5298 5299 5300
	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
5301 5302
	/* There is a swap entry and a page doesn't exist or isn't charged */
	if (ent.val && !ret &&
5303 5304 5305 5306
			css_id(&mc.from->css) == lookup_swap_cgroup(ent)) {
		ret = MC_TARGET_SWAP;
		if (target)
			target->ent = ent;
5307 5308 5309 5310 5311 5312 5313 5314 5315 5316 5317 5318
	}
	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;

5319 5320
	split_huge_page_pmd(walk->mm, pmd);

5321 5322 5323 5324 5325 5326 5327
	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();

5328 5329 5330
	return 0;
}

5331 5332 5333 5334 5335
static unsigned long mem_cgroup_count_precharge(struct mm_struct *mm)
{
	unsigned long precharge;
	struct vm_area_struct *vma;

5336
	down_read(&mm->mmap_sem);
5337 5338 5339 5340 5341 5342 5343 5344 5345 5346 5347
	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);
	}
5348
	up_read(&mm->mmap_sem);
5349 5350 5351 5352 5353 5354 5355 5356 5357

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

	return precharge;
}

static int mem_cgroup_precharge_mc(struct mm_struct *mm)
{
5358 5359 5360 5361 5362
	unsigned long precharge = mem_cgroup_count_precharge(mm);

	VM_BUG_ON(mc.moving_task);
	mc.moving_task = current;
	return mem_cgroup_do_precharge(precharge);
5363 5364
}

5365 5366
/* cancels all extra charges on mc.from and mc.to, and wakes up all waiters. */
static void __mem_cgroup_clear_mc(void)
5367
{
5368 5369 5370
	struct mem_cgroup *from = mc.from;
	struct mem_cgroup *to = mc.to;

5371
	/* we must uncharge all the leftover precharges from mc.to */
5372 5373 5374 5375 5376 5377 5378 5379 5380 5381 5382
	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;
5383
	}
5384 5385 5386 5387 5388 5389 5390 5391 5392 5393 5394 5395 5396 5397 5398 5399 5400 5401 5402
	/* 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;
	}
5403 5404 5405 5406 5407 5408 5409 5410 5411 5412 5413 5414 5415 5416 5417
	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();
5418
	spin_lock(&mc.lock);
5419 5420
	mc.from = NULL;
	mc.to = NULL;
5421
	spin_unlock(&mc.lock);
5422
	mem_cgroup_end_move(from);
5423 5424
}

5425 5426
static int mem_cgroup_can_attach(struct cgroup_subsys *ss,
				struct cgroup *cgroup,
5427
				struct cgroup_taskset *tset)
5428
{
5429
	struct task_struct *p = cgroup_taskset_first(tset);
5430
	int ret = 0;
5431
	struct mem_cgroup *memcg = mem_cgroup_from_cont(cgroup);
5432

5433
	if (memcg->move_charge_at_immigrate) {
5434 5435 5436
		struct mm_struct *mm;
		struct mem_cgroup *from = mem_cgroup_from_task(p);

5437
		VM_BUG_ON(from == memcg);
5438 5439 5440 5441 5442

		mm = get_task_mm(p);
		if (!mm)
			return 0;
		/* We move charges only when we move a owner of the mm */
5443 5444 5445 5446
		if (mm->owner == p) {
			VM_BUG_ON(mc.from);
			VM_BUG_ON(mc.to);
			VM_BUG_ON(mc.precharge);
5447
			VM_BUG_ON(mc.moved_charge);
5448
			VM_BUG_ON(mc.moved_swap);
5449
			mem_cgroup_start_move(from);
5450
			spin_lock(&mc.lock);
5451
			mc.from = from;
5452
			mc.to = memcg;
5453
			spin_unlock(&mc.lock);
5454
			/* We set mc.moving_task later */
5455 5456 5457 5458

			ret = mem_cgroup_precharge_mc(mm);
			if (ret)
				mem_cgroup_clear_mc();
5459 5460
		}
		mmput(mm);
5461 5462 5463 5464 5465 5466
	}
	return ret;
}

static void mem_cgroup_cancel_attach(struct cgroup_subsys *ss,
				struct cgroup *cgroup,
5467
				struct cgroup_taskset *tset)
5468
{
5469
	mem_cgroup_clear_mc();
5470 5471
}

5472 5473 5474
static int mem_cgroup_move_charge_pte_range(pmd_t *pmd,
				unsigned long addr, unsigned long end,
				struct mm_walk *walk)
5475
{
5476 5477 5478 5479 5480
	int ret = 0;
	struct vm_area_struct *vma = walk->private;
	pte_t *pte;
	spinlock_t *ptl;

5481
	split_huge_page_pmd(walk->mm, pmd);
5482 5483 5484 5485 5486 5487 5488 5489
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;
5490
		swp_entry_t ent;
5491 5492 5493 5494 5495 5496 5497 5498 5499 5500 5501

		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);
5502 5503
			if (!mem_cgroup_move_account(page, 1, pc,
						     mc.from, mc.to, false)) {
5504
				mc.precharge--;
5505 5506
				/* we uncharge from mc.from later. */
				mc.moved_charge++;
5507 5508 5509 5510 5511
			}
			putback_lru_page(page);
put:			/* is_target_pte_for_mc() gets the page */
			put_page(page);
			break;
5512 5513
		case MC_TARGET_SWAP:
			ent = target.ent;
5514 5515
			if (!mem_cgroup_move_swap_account(ent,
						mc.from, mc.to, false)) {
5516
				mc.precharge--;
5517 5518 5519
				/* we fixup refcnts and charges later. */
				mc.moved_swap++;
			}
5520
			break;
5521 5522 5523 5524 5525 5526 5527 5528 5529 5530 5531 5532 5533 5534
		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.
		 */
5535
		ret = mem_cgroup_do_precharge(1);
5536 5537 5538 5539 5540 5541 5542 5543 5544 5545 5546 5547
		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();
5548 5549 5550 5551 5552 5553 5554 5555 5556 5557 5558 5559 5560
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;
	}
5561 5562 5563 5564 5565 5566 5567 5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578
	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;
	}
5579
	up_read(&mm->mmap_sem);
5580 5581
}

B
Balbir Singh 已提交
5582 5583
static void mem_cgroup_move_task(struct cgroup_subsys *ss,
				struct cgroup *cont,
5584
				struct cgroup_taskset *tset)
B
Balbir Singh 已提交
5585
{
5586
	struct task_struct *p = cgroup_taskset_first(tset);
5587
	struct mm_struct *mm = get_task_mm(p);
5588 5589

	if (mm) {
5590 5591 5592
		if (mc.to)
			mem_cgroup_move_charge(mm);
		put_swap_token(mm);
5593 5594
		mmput(mm);
	}
5595 5596
	if (mc.to)
		mem_cgroup_clear_mc();
B
Balbir Singh 已提交
5597
}
5598 5599 5600
#else	/* !CONFIG_MMU */
static int mem_cgroup_can_attach(struct cgroup_subsys *ss,
				struct cgroup *cgroup,
5601
				struct cgroup_taskset *tset)
5602 5603 5604 5605 5606
{
	return 0;
}
static void mem_cgroup_cancel_attach(struct cgroup_subsys *ss,
				struct cgroup *cgroup,
5607
				struct cgroup_taskset *tset)
5608 5609 5610 5611
{
}
static void mem_cgroup_move_task(struct cgroup_subsys *ss,
				struct cgroup *cont,
5612
				struct cgroup_taskset *tset)
5613 5614 5615
{
}
#endif
B
Balbir Singh 已提交
5616

B
Balbir Singh 已提交
5617 5618 5619 5620
struct cgroup_subsys mem_cgroup_subsys = {
	.name = "memory",
	.subsys_id = mem_cgroup_subsys_id,
	.create = mem_cgroup_create,
5621
	.pre_destroy = mem_cgroup_pre_destroy,
B
Balbir Singh 已提交
5622 5623
	.destroy = mem_cgroup_destroy,
	.populate = mem_cgroup_populate,
5624 5625
	.can_attach = mem_cgroup_can_attach,
	.cancel_attach = mem_cgroup_cancel_attach,
B
Balbir Singh 已提交
5626
	.attach = mem_cgroup_move_task,
5627
	.early_init = 0,
K
KAMEZAWA Hiroyuki 已提交
5628
	.use_id = 1,
B
Balbir Singh 已提交
5629
};
5630 5631

#ifdef CONFIG_CGROUP_MEM_RES_CTLR_SWAP
5632 5633 5634
static int __init enable_swap_account(char *s)
{
	/* consider enabled if no parameter or 1 is given */
5635
	if (!strcmp(s, "1"))
5636
		really_do_swap_account = 1;
5637
	else if (!strcmp(s, "0"))
5638 5639 5640
		really_do_swap_account = 0;
	return 1;
}
5641
__setup("swapaccount=", enable_swap_account);
5642 5643

#endif