memcontrol.c 147.1 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
 *
13 14 15 16
 * Kernel Memory Controller
 * Copyright (C) 2012 Parallels Inc. and Google Inc.
 * Authors: Glauber Costa and Suleiman Souhlal
 *
17 18 19 20 21 22
 * Native page reclaim
 * Charge lifetime sanitation
 * Lockless page tracking & accounting
 * Unified hierarchy configuration model
 * Copyright (C) 2015 Red Hat, Inc., Johannes Weiner
 *
B
Balbir Singh 已提交
23 24 25 26 27 28 29 30 31 32 33
 * 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.
 */

34
#include <linux/page_counter.h>
B
Balbir Singh 已提交
35 36
#include <linux/memcontrol.h>
#include <linux/cgroup.h>
37
#include <linux/mm.h>
38
#include <linux/hugetlb.h>
K
KAMEZAWA Hiroyuki 已提交
39
#include <linux/pagemap.h>
40
#include <linux/smp.h>
41
#include <linux/page-flags.h>
42
#include <linux/backing-dev.h>
43 44
#include <linux/bit_spinlock.h>
#include <linux/rcupdate.h>
45
#include <linux/limits.h>
46
#include <linux/export.h>
47
#include <linux/mutex.h>
48
#include <linux/rbtree.h>
49
#include <linux/slab.h>
50
#include <linux/swap.h>
51
#include <linux/swapops.h>
52
#include <linux/spinlock.h>
53
#include <linux/eventfd.h>
54
#include <linux/poll.h>
55
#include <linux/sort.h>
56
#include <linux/fs.h>
57
#include <linux/seq_file.h>
58
#include <linux/vmpressure.h>
59
#include <linux/mm_inline.h>
60
#include <linux/swap_cgroup.h>
61
#include <linux/cpu.h>
62
#include <linux/oom.h>
63
#include <linux/lockdep.h>
64
#include <linux/file.h>
65
#include <linux/tracehook.h>
K
KAMEZAWA Hiroyuki 已提交
66
#include "internal.h"
G
Glauber Costa 已提交
67
#include <net/sock.h>
M
Michal Hocko 已提交
68
#include <net/ip.h>
G
Glauber Costa 已提交
69
#include <net/tcp_memcontrol.h>
70
#include "slab.h"
B
Balbir Singh 已提交
71

72 73
#include <asm/uaccess.h>

74 75
#include <trace/events/vmscan.h>

76 77
struct cgroup_subsys memory_cgrp_subsys __read_mostly;
EXPORT_SYMBOL(memory_cgrp_subsys);
78

79 80
struct mem_cgroup *root_mem_cgroup __read_mostly;

81
#define MEM_CGROUP_RECLAIM_RETRIES	5
B
Balbir Singh 已提交
82

83 84 85
/* Socket memory accounting disabled? */
static bool cgroup_memory_nosocket;

86
/* Whether the swap controller is active */
A
Andrew Morton 已提交
87
#ifdef CONFIG_MEMCG_SWAP
88 89
int do_swap_account __read_mostly;
#else
90
#define do_swap_account		0
91 92
#endif

93 94 95 96 97 98
/* Whether legacy memory+swap accounting is active */
static bool do_memsw_account(void)
{
	return !cgroup_subsys_on_dfl(memory_cgrp_subsys) && do_swap_account;
}

99 100 101
static const char * const mem_cgroup_stat_names[] = {
	"cache",
	"rss",
102
	"rss_huge",
103
	"mapped_file",
104
	"dirty",
105
	"writeback",
106 107 108 109 110 111 112 113 114 115
	"swap",
};

static const char * const mem_cgroup_events_names[] = {
	"pgpgin",
	"pgpgout",
	"pgfault",
	"pgmajfault",
};

116 117 118 119 120 121 122 123
static const char * const mem_cgroup_lru_names[] = {
	"inactive_anon",
	"active_anon",
	"inactive_file",
	"active_file",
	"unevictable",
};

124 125 126
#define THRESHOLDS_EVENTS_TARGET 128
#define SOFTLIMIT_EVENTS_TARGET 1024
#define NUMAINFO_EVENTS_TARGET	1024
127

128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147
/*
 * 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;

K
KAMEZAWA Hiroyuki 已提交
148 149 150 151 152
/* for OOM */
struct mem_cgroup_eventfd_list {
	struct list_head list;
	struct eventfd_ctx *eventfd;
};
153

154 155 156
/*
 * cgroup_event represents events which userspace want to receive.
 */
157
struct mem_cgroup_event {
158
	/*
159
	 * memcg which the event belongs to.
160
	 */
161
	struct mem_cgroup *memcg;
162 163 164 165 166 167 168 169
	/*
	 * eventfd to signal userspace about the event.
	 */
	struct eventfd_ctx *eventfd;
	/*
	 * Each of these stored in a list by the cgroup.
	 */
	struct list_head list;
170 171 172 173 174
	/*
	 * register_event() callback will be used to add new userspace
	 * waiter for changes related to this event.  Use eventfd_signal()
	 * on eventfd to send notification to userspace.
	 */
175
	int (*register_event)(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
176
			      struct eventfd_ctx *eventfd, const char *args);
177 178 179 180 181
	/*
	 * unregister_event() callback will be called when userspace closes
	 * the eventfd or on cgroup removing.  This callback must be set,
	 * if you want provide notification functionality.
	 */
182
	void (*unregister_event)(struct mem_cgroup *memcg,
183
				 struct eventfd_ctx *eventfd);
184 185 186 187 188 189 190 191 192 193
	/*
	 * All fields below needed to unregister event when
	 * userspace closes eventfd.
	 */
	poll_table pt;
	wait_queue_head_t *wqh;
	wait_queue_t wait;
	struct work_struct remove;
};

194 195
static void mem_cgroup_threshold(struct mem_cgroup *memcg);
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg);
196

197 198
/* Stuffs for move charges at task migration. */
/*
199
 * Types of charges to be moved.
200
 */
201 202 203
#define MOVE_ANON	0x1U
#define MOVE_FILE	0x2U
#define MOVE_MASK	(MOVE_ANON | MOVE_FILE)
204

205 206
/* "mc" and its members are protected by cgroup_mutex */
static struct move_charge_struct {
207
	spinlock_t	  lock; /* for from, to */
208 209
	struct mem_cgroup *from;
	struct mem_cgroup *to;
210
	unsigned long flags;
211
	unsigned long precharge;
212
	unsigned long moved_charge;
213
	unsigned long moved_swap;
214 215 216
	struct task_struct *moving_task;	/* a task moving charges */
	wait_queue_head_t waitq;		/* a waitq for other context */
} mc = {
217
	.lock = __SPIN_LOCK_UNLOCKED(mc.lock),
218 219
	.waitq = __WAIT_QUEUE_HEAD_INITIALIZER(mc.waitq),
};
220

221 222 223 224
/*
 * Maximum loops in mem_cgroup_hierarchical_reclaim(), used for soft
 * limit reclaim to prevent infinite loops, if they ever occur.
 */
225
#define	MEM_CGROUP_MAX_RECLAIM_LOOPS		100
226
#define	MEM_CGROUP_MAX_SOFT_LIMIT_RECLAIM_LOOPS	2
227

228 229
enum charge_type {
	MEM_CGROUP_CHARGE_TYPE_CACHE = 0,
230
	MEM_CGROUP_CHARGE_TYPE_ANON,
K
KAMEZAWA Hiroyuki 已提交
231
	MEM_CGROUP_CHARGE_TYPE_SWAPOUT,	/* for accounting swapcache */
K
KAMEZAWA Hiroyuki 已提交
232
	MEM_CGROUP_CHARGE_TYPE_DROP,	/* a page was unused swap cache */
233 234 235
	NR_CHARGE_TYPE,
};

236
/* for encoding cft->private value on file */
G
Glauber Costa 已提交
237 238 239 240
enum res_type {
	_MEM,
	_MEMSWAP,
	_OOM_TYPE,
241
	_KMEM,
G
Glauber Costa 已提交
242 243
};

244 245
#define MEMFILE_PRIVATE(x, val)	((x) << 16 | (val))
#define MEMFILE_TYPE(val)	((val) >> 16 & 0xffff)
246
#define MEMFILE_ATTR(val)	((val) & 0xffff)
K
KAMEZAWA Hiroyuki 已提交
247 248
/* Used for OOM nofiier */
#define OOM_CONTROL		(0)
249

250 251 252 253 254 255 256
/*
 * The memcg_create_mutex will be held whenever a new cgroup is created.
 * As a consequence, any change that needs to protect against new child cgroups
 * appearing has to hold it as well.
 */
static DEFINE_MUTEX(memcg_create_mutex);

257 258 259 260 261 262 263 264 265 266 267 268 269
/* Some nice accessors for the vmpressure. */
struct vmpressure *memcg_to_vmpressure(struct mem_cgroup *memcg)
{
	if (!memcg)
		memcg = root_mem_cgroup;
	return &memcg->vmpressure;
}

struct cgroup_subsys_state *vmpressure_to_css(struct vmpressure *vmpr)
{
	return &container_of(vmpr, struct mem_cgroup, vmpressure)->css;
}

270 271 272 273 274
static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
{
	return (memcg == root_mem_cgroup);
}

275 276 277 278 279 280
/*
 * We restrict the id in the range of [1, 65535], so it can fit into
 * an unsigned short.
 */
#define MEM_CGROUP_ID_MAX	USHRT_MAX

L
Li Zefan 已提交
281 282
static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
{
283
	return memcg->css.id;
L
Li Zefan 已提交
284 285
}

286 287 288 289 290 291
/*
 * A helper function to get mem_cgroup from ID. must be called under
 * rcu_read_lock().  The caller is responsible for calling
 * css_tryget_online() if the mem_cgroup is used for charging. (dropping
 * refcnt from swap can be called against removed memcg.)
 */
L
Li Zefan 已提交
292 293 294 295
static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
{
	struct cgroup_subsys_state *css;

296
	css = css_from_id(id, &memory_cgrp_subsys);
L
Li Zefan 已提交
297 298 299
	return mem_cgroup_from_css(css);
}

300
#ifndef CONFIG_SLOB
301
/*
302
 * This will be the memcg's index in each cache's ->memcg_params.memcg_caches.
L
Li Zefan 已提交
303 304 305 306 307
 * The main reason for not using cgroup id for this:
 *  this works better in sparse environments, where we have a lot of memcgs,
 *  but only a few kmem-limited. Or also, if we have, for instance, 200
 *  memcgs, and none but the 200th is kmem-limited, we'd have to have a
 *  200 entry array for that.
308
 *
309 310
 * The current size of the caches array is stored in memcg_nr_cache_ids. It
 * will double each time we have to increase it.
311
 */
312 313
static DEFINE_IDA(memcg_cache_ida);
int memcg_nr_cache_ids;
314

315 316 317 318 319 320 321 322 323 324 325 326 327
/* Protects memcg_nr_cache_ids */
static DECLARE_RWSEM(memcg_cache_ids_sem);

void memcg_get_cache_ids(void)
{
	down_read(&memcg_cache_ids_sem);
}

void memcg_put_cache_ids(void)
{
	up_read(&memcg_cache_ids_sem);
}

328 329 330 331 332 333
/*
 * MIN_SIZE is different than 1, because we would like to avoid going through
 * the alloc/free process all the time. In a small machine, 4 kmem-limited
 * cgroups is a reasonable guess. In the future, it could be a parameter or
 * tunable, but that is strictly not necessary.
 *
L
Li Zefan 已提交
334
 * MAX_SIZE should be as large as the number of cgrp_ids. Ideally, we could get
335 336
 * this constant directly from cgroup, but it is understandable that this is
 * better kept as an internal representation in cgroup.c. In any case, the
L
Li Zefan 已提交
337
 * cgrp_id space is not getting any smaller, and we don't have to necessarily
338 339 340
 * increase ours as well if it increases.
 */
#define MEMCG_CACHES_MIN_SIZE 4
L
Li Zefan 已提交
341
#define MEMCG_CACHES_MAX_SIZE MEM_CGROUP_ID_MAX
342

343 344 345 346 347 348
/*
 * A lot of the calls to the cache allocation functions are expected to be
 * inlined by the compiler. Since the calls to memcg_kmem_get_cache are
 * conditional to this static branch, we'll have to allow modules that does
 * kmem_cache_alloc and the such to see this symbol as well
 */
349
DEFINE_STATIC_KEY_FALSE(memcg_kmem_enabled_key);
350
EXPORT_SYMBOL(memcg_kmem_enabled_key);
351

352
#endif /* !CONFIG_SLOB */
353

354
static struct mem_cgroup_per_zone *
355
mem_cgroup_zone_zoneinfo(struct mem_cgroup *memcg, struct zone *zone)
356
{
357 358 359
	int nid = zone_to_nid(zone);
	int zid = zone_idx(zone);

360
	return &memcg->nodeinfo[nid]->zoneinfo[zid];
361 362
}

363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386
/**
 * mem_cgroup_css_from_page - css of the memcg associated with a page
 * @page: page of interest
 *
 * If memcg is bound to the default hierarchy, css of the memcg associated
 * with @page is returned.  The returned css remains associated with @page
 * until it is released.
 *
 * If memcg is bound to a traditional hierarchy, the css of root_mem_cgroup
 * is returned.
 *
 * XXX: The above description of behavior on the default hierarchy isn't
 * strictly true yet as replace_page_cache_page() can modify the
 * association before @page is released even on the default hierarchy;
 * however, the current and planned usages don't mix the the two functions
 * and replace_page_cache_page() will soon be updated to make the invariant
 * actually true.
 */
struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page)
{
	struct mem_cgroup *memcg;

	memcg = page->mem_cgroup;

387
	if (!memcg || !cgroup_subsys_on_dfl(memory_cgrp_subsys))
388 389 390 391 392
		memcg = root_mem_cgroup;

	return &memcg->css;
}

393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420
/**
 * page_cgroup_ino - return inode number of the memcg a page is charged to
 * @page: the page
 *
 * Look up the closest online ancestor of the memory cgroup @page is charged to
 * and return its inode number or 0 if @page is not charged to any cgroup. It
 * is safe to call this function without holding a reference to @page.
 *
 * Note, this function is inherently racy, because there is nothing to prevent
 * the cgroup inode from getting torn down and potentially reallocated a moment
 * after page_cgroup_ino() returns, so it only should be used by callers that
 * do not care (such as procfs interfaces).
 */
ino_t page_cgroup_ino(struct page *page)
{
	struct mem_cgroup *memcg;
	unsigned long ino = 0;

	rcu_read_lock();
	memcg = READ_ONCE(page->mem_cgroup);
	while (memcg && !(memcg->css.flags & CSS_ONLINE))
		memcg = parent_mem_cgroup(memcg);
	if (memcg)
		ino = cgroup_ino(memcg->css.cgroup);
	rcu_read_unlock();
	return ino;
}

421
static struct mem_cgroup_per_zone *
422
mem_cgroup_page_zoneinfo(struct mem_cgroup *memcg, struct page *page)
423
{
424 425
	int nid = page_to_nid(page);
	int zid = page_zonenum(page);
426

427
	return &memcg->nodeinfo[nid]->zoneinfo[zid];
428 429
}

430 431 432 433 434 435 436 437 438 439 440 441 442 443 444
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];
}

445 446
static void __mem_cgroup_insert_exceeded(struct mem_cgroup_per_zone *mz,
					 struct mem_cgroup_tree_per_zone *mctz,
447
					 unsigned long new_usage_in_excess)
448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476
{
	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;

	mz->usage_in_excess = new_usage_in_excess;
	if (!mz->usage_in_excess)
		return;
	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;
}

477 478
static void __mem_cgroup_remove_exceeded(struct mem_cgroup_per_zone *mz,
					 struct mem_cgroup_tree_per_zone *mctz)
479 480 481 482 483 484 485
{
	if (!mz->on_tree)
		return;
	rb_erase(&mz->tree_node, &mctz->rb_root);
	mz->on_tree = false;
}

486 487
static void mem_cgroup_remove_exceeded(struct mem_cgroup_per_zone *mz,
				       struct mem_cgroup_tree_per_zone *mctz)
488
{
489 490 491
	unsigned long flags;

	spin_lock_irqsave(&mctz->lock, flags);
492
	__mem_cgroup_remove_exceeded(mz, mctz);
493
	spin_unlock_irqrestore(&mctz->lock, flags);
494 495
}

496 497 498
static unsigned long soft_limit_excess(struct mem_cgroup *memcg)
{
	unsigned long nr_pages = page_counter_read(&memcg->memory);
499
	unsigned long soft_limit = READ_ONCE(memcg->soft_limit);
500 501 502 503 504 505 506
	unsigned long excess = 0;

	if (nr_pages > soft_limit)
		excess = nr_pages - soft_limit;

	return excess;
}
507 508 509

static void mem_cgroup_update_tree(struct mem_cgroup *memcg, struct page *page)
{
510
	unsigned long excess;
511 512 513
	struct mem_cgroup_per_zone *mz;
	struct mem_cgroup_tree_per_zone *mctz;

514
	mctz = soft_limit_tree_from_page(page);
515 516 517 518 519
	/*
	 * Necessary to update all ancestors when hierarchy is used.
	 * because their event counter is not touched.
	 */
	for (; memcg; memcg = parent_mem_cgroup(memcg)) {
520
		mz = mem_cgroup_page_zoneinfo(memcg, page);
521
		excess = soft_limit_excess(memcg);
522 523 524 525 526
		/*
		 * We have to update the tree if mz is on RB-tree or
		 * mem is over its softlimit.
		 */
		if (excess || mz->on_tree) {
527 528 529
			unsigned long flags;

			spin_lock_irqsave(&mctz->lock, flags);
530 531
			/* if on-tree, remove it */
			if (mz->on_tree)
532
				__mem_cgroup_remove_exceeded(mz, mctz);
533 534 535 536
			/*
			 * Insert again. mz->usage_in_excess will be updated.
			 * If excess is 0, no tree ops.
			 */
537
			__mem_cgroup_insert_exceeded(mz, mctz, excess);
538
			spin_unlock_irqrestore(&mctz->lock, flags);
539 540 541 542 543 544 545
		}
	}
}

static void mem_cgroup_remove_from_trees(struct mem_cgroup *memcg)
{
	struct mem_cgroup_tree_per_zone *mctz;
546 547
	struct mem_cgroup_per_zone *mz;
	int nid, zid;
548

549 550 551 552
	for_each_node(nid) {
		for (zid = 0; zid < MAX_NR_ZONES; zid++) {
			mz = &memcg->nodeinfo[nid]->zoneinfo[zid];
			mctz = soft_limit_tree_node_zone(nid, zid);
553
			mem_cgroup_remove_exceeded(mz, mctz);
554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575
		}
	}
}

static struct mem_cgroup_per_zone *
__mem_cgroup_largest_soft_limit_node(struct mem_cgroup_tree_per_zone *mctz)
{
	struct rb_node *rightmost = NULL;
	struct mem_cgroup_per_zone *mz;

retry:
	mz = NULL;
	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.
	 */
576
	__mem_cgroup_remove_exceeded(mz, mctz);
577
	if (!soft_limit_excess(mz->memcg) ||
578
	    !css_tryget_online(&mz->memcg->css))
579 580 581 582 583 584 585 586 587 588
		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;

589
	spin_lock_irq(&mctz->lock);
590
	mz = __mem_cgroup_largest_soft_limit_node(mctz);
591
	spin_unlock_irq(&mctz->lock);
592 593 594
	return mz;
}

595
/*
596 597
 * Return page count for single (non recursive) @memcg.
 *
598 599 600 601 602
 * 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
603
 * a periodic synchronization of counter in memcg's counter.
604 605 606 607 608 609 610 611 612
 *
 * 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
613
 * common workload, threshold and synchronization as vmstat[] should be
614 615
 * implemented.
 */
616 617
static unsigned long
mem_cgroup_read_stat(struct mem_cgroup *memcg, enum mem_cgroup_stat_index idx)
618
{
619
	long val = 0;
620 621
	int cpu;

622
	/* Per-cpu values can be negative, use a signed accumulator */
623
	for_each_possible_cpu(cpu)
624
		val += per_cpu(memcg->stat->count[idx], cpu);
625 626 627 628 629 630
	/*
	 * Summing races with updates, so val may be negative.  Avoid exposing
	 * transient negative values.
	 */
	if (val < 0)
		val = 0;
631 632 633
	return val;
}

634
static unsigned long mem_cgroup_read_events(struct mem_cgroup *memcg,
635 636 637 638 639
					    enum mem_cgroup_events_index idx)
{
	unsigned long val = 0;
	int cpu;

640
	for_each_possible_cpu(cpu)
641
		val += per_cpu(memcg->stat->events[idx], cpu);
642 643 644
	return val;
}

645
static void mem_cgroup_charge_statistics(struct mem_cgroup *memcg,
646
					 struct page *page,
647
					 bool compound, int nr_pages)
648
{
649 650 651 652
	/*
	 * Here, RSS means 'mapped anon' and anon's SwapCache. Shmem/tmpfs is
	 * counted as CACHE even if it's on ANON LRU.
	 */
653
	if (PageAnon(page))
654
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_RSS],
655
				nr_pages);
656
	else
657
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_CACHE],
658
				nr_pages);
659

660 661
	if (compound) {
		VM_BUG_ON_PAGE(!PageTransHuge(page), page);
662 663
		__this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_RSS_HUGE],
				nr_pages);
664
	}
665

666 667
	/* pagein of a big page is an event. So, ignore page size */
	if (nr_pages > 0)
668
		__this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGPGIN]);
669
	else {
670
		__this_cpu_inc(memcg->stat->events[MEM_CGROUP_EVENTS_PGPGOUT]);
671 672
		nr_pages = -nr_pages; /* for event */
	}
673

674
	__this_cpu_add(memcg->stat->nr_page_events, nr_pages);
675 676
}

677 678 679
static unsigned long mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
						  int nid,
						  unsigned int lru_mask)
680
{
681
	unsigned long nr = 0;
682 683
	int zid;

684
	VM_BUG_ON((unsigned)nid >= nr_node_ids);
685

686 687 688 689 690 691 692 693 694 695 696 697
	for (zid = 0; zid < MAX_NR_ZONES; zid++) {
		struct mem_cgroup_per_zone *mz;
		enum lru_list lru;

		for_each_lru(lru) {
			if (!(BIT(lru) & lru_mask))
				continue;
			mz = &memcg->nodeinfo[nid]->zoneinfo[zid];
			nr += mz->lru_size[lru];
		}
	}
	return nr;
698
}
699

700
static unsigned long mem_cgroup_nr_lru_pages(struct mem_cgroup *memcg,
701
			unsigned int lru_mask)
702
{
703
	unsigned long nr = 0;
704
	int nid;
705

706
	for_each_node_state(nid, N_MEMORY)
707 708
		nr += mem_cgroup_node_nr_lru_pages(memcg, nid, lru_mask);
	return nr;
709 710
}

711 712
static bool mem_cgroup_event_ratelimit(struct mem_cgroup *memcg,
				       enum mem_cgroup_events_target target)
713 714 715
{
	unsigned long val, next;

716
	val = __this_cpu_read(memcg->stat->nr_page_events);
717
	next = __this_cpu_read(memcg->stat->targets[target]);
718
	/* from time_after() in jiffies.h */
719 720 721 722 723
	if ((long)next - (long)val < 0) {
		switch (target) {
		case MEM_CGROUP_TARGET_THRESH:
			next = val + THRESHOLDS_EVENTS_TARGET;
			break;
724 725 726
		case MEM_CGROUP_TARGET_SOFTLIMIT:
			next = val + SOFTLIMIT_EVENTS_TARGET;
			break;
727 728 729 730 731 732 733 734
		case MEM_CGROUP_TARGET_NUMAINFO:
			next = val + NUMAINFO_EVENTS_TARGET;
			break;
		default:
			break;
		}
		__this_cpu_write(memcg->stat->targets[target], next);
		return true;
735
	}
736
	return false;
737 738 739 740 741 742
}

/*
 * Check events in order.
 *
 */
743
static void memcg_check_events(struct mem_cgroup *memcg, struct page *page)
744 745
{
	/* threshold event is triggered in finer grain than soft limit */
746 747
	if (unlikely(mem_cgroup_event_ratelimit(memcg,
						MEM_CGROUP_TARGET_THRESH))) {
748
		bool do_softlimit;
749
		bool do_numainfo __maybe_unused;
750

751 752
		do_softlimit = mem_cgroup_event_ratelimit(memcg,
						MEM_CGROUP_TARGET_SOFTLIMIT);
753 754 755 756
#if MAX_NUMNODES > 1
		do_numainfo = mem_cgroup_event_ratelimit(memcg,
						MEM_CGROUP_TARGET_NUMAINFO);
#endif
757
		mem_cgroup_threshold(memcg);
758 759
		if (unlikely(do_softlimit))
			mem_cgroup_update_tree(memcg, page);
760
#if MAX_NUMNODES > 1
761
		if (unlikely(do_numainfo))
762
			atomic_inc(&memcg->numainfo_events);
763
#endif
764
	}
765 766
}

767
struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p)
768
{
769 770 771 772 773 774 775 776
	/*
	 * 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;

777
	return mem_cgroup_from_css(task_css(p, memory_cgrp_id));
778
}
M
Michal Hocko 已提交
779
EXPORT_SYMBOL(mem_cgroup_from_task);
780

781
static struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
782
{
783
	struct mem_cgroup *memcg = NULL;
784

785 786
	rcu_read_lock();
	do {
787 788 789 790 791 792
		/*
		 * Page cache insertions can happen withou an
		 * actual mm context, e.g. during disk probing
		 * on boot, loopback IO, acct() writes etc.
		 */
		if (unlikely(!mm))
793
			memcg = root_mem_cgroup;
794 795 796 797 798
		else {
			memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
			if (unlikely(!memcg))
				memcg = root_mem_cgroup;
		}
799
	} while (!css_tryget_online(&memcg->css));
800
	rcu_read_unlock();
801
	return memcg;
802 803
}

804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820
/**
 * mem_cgroup_iter - iterate over memory cgroup hierarchy
 * @root: hierarchy root
 * @prev: previously returned memcg, NULL on first invocation
 * @reclaim: cookie for shared reclaim walks, NULL for full walks
 *
 * Returns references to children of the hierarchy below @root, or
 * @root itself, or %NULL after a full round-trip.
 *
 * Caller must pass the return value in @prev on subsequent
 * invocations for reference counting, or use mem_cgroup_iter_break()
 * to cancel a hierarchy walk before the round-trip is complete.
 *
 * Reclaimers can specify a zone and a priority level in @reclaim to
 * divide up the memcgs in the hierarchy among all concurrent
 * reclaimers operating on the same zone and priority.
 */
821
struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *root,
822
				   struct mem_cgroup *prev,
823
				   struct mem_cgroup_reclaim_cookie *reclaim)
K
KAMEZAWA Hiroyuki 已提交
824
{
M
Michal Hocko 已提交
825
	struct mem_cgroup_reclaim_iter *uninitialized_var(iter);
826
	struct cgroup_subsys_state *css = NULL;
827
	struct mem_cgroup *memcg = NULL;
828
	struct mem_cgroup *pos = NULL;
829

830 831
	if (mem_cgroup_disabled())
		return NULL;
832

833 834
	if (!root)
		root = root_mem_cgroup;
K
KAMEZAWA Hiroyuki 已提交
835

836
	if (prev && !reclaim)
837
		pos = prev;
K
KAMEZAWA Hiroyuki 已提交
838

839 840
	if (!root->use_hierarchy && root != root_mem_cgroup) {
		if (prev)
841
			goto out;
842
		return root;
843
	}
K
KAMEZAWA Hiroyuki 已提交
844

845
	rcu_read_lock();
M
Michal Hocko 已提交
846

847 848 849 850 851 852 853 854 855
	if (reclaim) {
		struct mem_cgroup_per_zone *mz;

		mz = mem_cgroup_zone_zoneinfo(root, reclaim->zone);
		iter = &mz->iter[reclaim->priority];

		if (prev && reclaim->generation != iter->generation)
			goto out_unlock;

856
		while (1) {
857
			pos = READ_ONCE(iter->position);
858 859
			if (!pos || css_tryget(&pos->css))
				break;
860
			/*
861 862 863 864 865 866
			 * css reference reached zero, so iter->position will
			 * be cleared by ->css_released. However, we should not
			 * rely on this happening soon, because ->css_released
			 * is called from a work queue, and by busy-waiting we
			 * might block it. So we clear iter->position right
			 * away.
867
			 */
868 869
			(void)cmpxchg(&iter->position, pos, NULL);
		}
870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886
	}

	if (pos)
		css = &pos->css;

	for (;;) {
		css = css_next_descendant_pre(css, &root->css);
		if (!css) {
			/*
			 * Reclaimers share the hierarchy walk, and a
			 * new one might jump in right at the end of
			 * the hierarchy - make sure they see at least
			 * one group and restart from the beginning.
			 */
			if (!prev)
				continue;
			break;
887
		}
K
KAMEZAWA Hiroyuki 已提交
888

889 890 891 892 893 894
		/*
		 * Verify the css and acquire a reference.  The root
		 * is provided by the caller, so we know it's alive
		 * and kicking, and don't take an extra reference.
		 */
		memcg = mem_cgroup_from_css(css);
K
KAMEZAWA Hiroyuki 已提交
895

896 897
		if (css == &root->css)
			break;
K
KAMEZAWA Hiroyuki 已提交
898

899
		if (css_tryget(css)) {
900 901 902 903 904 905 906
			/*
			 * Make sure the memcg is initialized:
			 * mem_cgroup_css_online() orders the the
			 * initialization against setting the flag.
			 */
			if (smp_load_acquire(&memcg->initialized))
				break;
907

908
			css_put(css);
909
		}
910

911
		memcg = NULL;
912
	}
913 914 915

	if (reclaim) {
		/*
916 917 918
		 * The position could have already been updated by a competing
		 * thread, so check that the value hasn't changed since we read
		 * it to avoid reclaiming from the same cgroup twice.
919
		 */
920 921
		(void)cmpxchg(&iter->position, pos, memcg);

922 923 924 925 926 927 928
		if (pos)
			css_put(&pos->css);

		if (!memcg)
			iter->generation++;
		else if (!prev)
			reclaim->generation = iter->generation;
929
	}
930

931 932
out_unlock:
	rcu_read_unlock();
933
out:
934 935 936
	if (prev && prev != root)
		css_put(&prev->css);

937
	return memcg;
K
KAMEZAWA Hiroyuki 已提交
938
}
K
KAMEZAWA Hiroyuki 已提交
939

940 941 942 943 944 945 946
/**
 * mem_cgroup_iter_break - abort a hierarchy walk prematurely
 * @root: hierarchy root
 * @prev: last visited hierarchy member as returned by mem_cgroup_iter()
 */
void mem_cgroup_iter_break(struct mem_cgroup *root,
			   struct mem_cgroup *prev)
947 948 949 950 951 952
{
	if (!root)
		root = root_mem_cgroup;
	if (prev && prev != root)
		css_put(&prev->css);
}
K
KAMEZAWA Hiroyuki 已提交
953

954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975
static void invalidate_reclaim_iterators(struct mem_cgroup *dead_memcg)
{
	struct mem_cgroup *memcg = dead_memcg;
	struct mem_cgroup_reclaim_iter *iter;
	struct mem_cgroup_per_zone *mz;
	int nid, zid;
	int i;

	while ((memcg = parent_mem_cgroup(memcg))) {
		for_each_node(nid) {
			for (zid = 0; zid < MAX_NR_ZONES; zid++) {
				mz = &memcg->nodeinfo[nid]->zoneinfo[zid];
				for (i = 0; i <= DEF_PRIORITY; i++) {
					iter = &mz->iter[i];
					cmpxchg(&iter->position,
						dead_memcg, NULL);
				}
			}
		}
	}
}

976 977 978 979 980 981
/*
 * 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)		\
982
	for (iter = mem_cgroup_iter(root, NULL, NULL);	\
983
	     iter != NULL;				\
984
	     iter = mem_cgroup_iter(root, iter, NULL))
985

986
#define for_each_mem_cgroup(iter)			\
987
	for (iter = mem_cgroup_iter(NULL, NULL, NULL);	\
988
	     iter != NULL;				\
989
	     iter = mem_cgroup_iter(NULL, iter, NULL))
K
KAMEZAWA Hiroyuki 已提交
990

991 992 993
/**
 * mem_cgroup_zone_lruvec - get the lru list vector for a zone and memcg
 * @zone: zone of the wanted lruvec
994
 * @memcg: memcg of the wanted lruvec
995 996 997 998 999 1000 1001 1002 1003
 *
 * Returns the lru list vector holding pages for the given @zone and
 * @mem.  This can be the global zone lruvec, if the memory controller
 * is disabled.
 */
struct lruvec *mem_cgroup_zone_lruvec(struct zone *zone,
				      struct mem_cgroup *memcg)
{
	struct mem_cgroup_per_zone *mz;
1004
	struct lruvec *lruvec;
1005

1006 1007 1008 1009
	if (mem_cgroup_disabled()) {
		lruvec = &zone->lruvec;
		goto out;
	}
1010

1011
	mz = mem_cgroup_zone_zoneinfo(memcg, zone);
1012 1013 1014 1015 1016 1017 1018 1019 1020 1021
	lruvec = &mz->lruvec;
out:
	/*
	 * Since a node can be onlined after the mem_cgroup was created,
	 * we have to be prepared to initialize lruvec->zone here;
	 * and if offlined then reonlined, we need to reinitialize it.
	 */
	if (unlikely(lruvec->zone != zone))
		lruvec->zone = zone;
	return lruvec;
1022 1023 1024
}

/**
1025
 * mem_cgroup_page_lruvec - return lruvec for isolating/putting an LRU page
1026
 * @page: the page
1027
 * @zone: zone of the page
1028 1029 1030 1031
 *
 * This function is only safe when following the LRU page isolation
 * and putback protocol: the LRU lock must be held, and the page must
 * either be PageLRU() or the caller must have isolated/allocated it.
1032
 */
1033
struct lruvec *mem_cgroup_page_lruvec(struct page *page, struct zone *zone)
K
KAMEZAWA Hiroyuki 已提交
1034 1035
{
	struct mem_cgroup_per_zone *mz;
1036
	struct mem_cgroup *memcg;
1037
	struct lruvec *lruvec;
1038

1039 1040 1041 1042
	if (mem_cgroup_disabled()) {
		lruvec = &zone->lruvec;
		goto out;
	}
1043

1044
	memcg = page->mem_cgroup;
1045
	/*
1046
	 * Swapcache readahead pages are added to the LRU - and
1047
	 * possibly migrated - before they are charged.
1048
	 */
1049 1050
	if (!memcg)
		memcg = root_mem_cgroup;
1051

1052
	mz = mem_cgroup_page_zoneinfo(memcg, page);
1053 1054 1055 1056 1057 1058 1059 1060 1061 1062
	lruvec = &mz->lruvec;
out:
	/*
	 * Since a node can be onlined after the mem_cgroup was created,
	 * we have to be prepared to initialize lruvec->zone here;
	 * and if offlined then reonlined, we need to reinitialize it.
	 */
	if (unlikely(lruvec->zone != zone))
		lruvec->zone = zone;
	return lruvec;
K
KAMEZAWA Hiroyuki 已提交
1063
}
1064

1065
/**
1066 1067 1068 1069
 * mem_cgroup_update_lru_size - account for adding or removing an lru page
 * @lruvec: mem_cgroup per zone lru vector
 * @lru: index of lru list the page is sitting on
 * @nr_pages: positive when adding or negative when removing
1070
 *
1071 1072
 * This function must be called when a page is added to or removed from an
 * lru list.
1073
 */
1074 1075
void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
				int nr_pages)
1076 1077
{
	struct mem_cgroup_per_zone *mz;
1078
	unsigned long *lru_size;
1079 1080 1081 1082

	if (mem_cgroup_disabled())
		return;

1083 1084 1085 1086
	mz = container_of(lruvec, struct mem_cgroup_per_zone, lruvec);
	lru_size = mz->lru_size + lru;
	*lru_size += nr_pages;
	VM_BUG_ON((long)(*lru_size) < 0);
K
KAMEZAWA Hiroyuki 已提交
1087
}
1088

1089
bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg)
1090
{
1091
	struct mem_cgroup *task_memcg;
1092
	struct task_struct *p;
1093
	bool ret;
1094

1095
	p = find_lock_task_mm(task);
1096
	if (p) {
1097
		task_memcg = get_mem_cgroup_from_mm(p->mm);
1098 1099 1100 1101 1102 1103 1104
		task_unlock(p);
	} else {
		/*
		 * All threads may have already detached their mm's, but the oom
		 * killer still needs to detect if they have already been oom
		 * killed to prevent needlessly killing additional tasks.
		 */
1105
		rcu_read_lock();
1106 1107
		task_memcg = mem_cgroup_from_task(task);
		css_get(&task_memcg->css);
1108
		rcu_read_unlock();
1109
	}
1110 1111
	ret = mem_cgroup_is_descendant(task_memcg, memcg);
	css_put(&task_memcg->css);
1112 1113 1114
	return ret;
}

1115
/**
1116
 * mem_cgroup_margin - calculate chargeable space of a memory cgroup
W
Wanpeng Li 已提交
1117
 * @memcg: the memory cgroup
1118
 *
1119
 * Returns the maximum amount of memory @mem can be charged with, in
1120
 * pages.
1121
 */
1122
static unsigned long mem_cgroup_margin(struct mem_cgroup *memcg)
1123
{
1124 1125 1126
	unsigned long margin = 0;
	unsigned long count;
	unsigned long limit;
1127

1128
	count = page_counter_read(&memcg->memory);
1129
	limit = READ_ONCE(memcg->memory.limit);
1130 1131 1132
	if (count < limit)
		margin = limit - count;

1133
	if (do_memsw_account()) {
1134
		count = page_counter_read(&memcg->memsw);
1135
		limit = READ_ONCE(memcg->memsw.limit);
1136 1137 1138 1139 1140
		if (count <= limit)
			margin = min(margin, limit - count);
	}

	return margin;
1141 1142
}

1143
/*
Q
Qiang Huang 已提交
1144
 * A routine for checking "mem" is under move_account() or not.
1145
 *
Q
Qiang Huang 已提交
1146 1147 1148
 * Checking a cgroup is mc.from or mc.to or under hierarchy of
 * moving cgroups. This is for waiting at high-memory pressure
 * caused by "move".
1149
 */
1150
static bool mem_cgroup_under_move(struct mem_cgroup *memcg)
1151
{
1152 1153
	struct mem_cgroup *from;
	struct mem_cgroup *to;
1154
	bool ret = false;
1155 1156 1157 1158 1159 1160 1161 1162 1163
	/*
	 * 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;
1164

1165 1166
	ret = mem_cgroup_is_descendant(from, memcg) ||
		mem_cgroup_is_descendant(to, memcg);
1167 1168
unlock:
	spin_unlock(&mc.lock);
1169 1170 1171
	return ret;
}

1172
static bool mem_cgroup_wait_acct_move(struct mem_cgroup *memcg)
1173 1174
{
	if (mc.moving_task && current != mc.moving_task) {
1175
		if (mem_cgroup_under_move(memcg)) {
1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187
			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;
}

1188
#define K(x) ((x) << (PAGE_SHIFT-10))
1189
/**
1190
 * mem_cgroup_print_oom_info: Print OOM information relevant to memory controller.
1191 1192 1193 1194 1195 1196 1197 1198
 * @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)
{
T
Tejun Heo 已提交
1199
	/* oom_info_lock ensures that parallel ooms do not interleave */
1200
	static DEFINE_MUTEX(oom_info_lock);
1201 1202
	struct mem_cgroup *iter;
	unsigned int i;
1203

1204
	mutex_lock(&oom_info_lock);
1205 1206
	rcu_read_lock();

1207 1208 1209 1210 1211 1212 1213 1214
	if (p) {
		pr_info("Task in ");
		pr_cont_cgroup_path(task_cgroup(p, memory_cgrp_id));
		pr_cont(" killed as a result of limit of ");
	} else {
		pr_info("Memory limit reached of cgroup ");
	}

T
Tejun Heo 已提交
1215
	pr_cont_cgroup_path(memcg->css.cgroup);
1216
	pr_cont("\n");
1217 1218 1219

	rcu_read_unlock();

1220 1221 1222 1223 1224 1225 1226 1227 1228
	pr_info("memory: usage %llukB, limit %llukB, failcnt %lu\n",
		K((u64)page_counter_read(&memcg->memory)),
		K((u64)memcg->memory.limit), memcg->memory.failcnt);
	pr_info("memory+swap: usage %llukB, limit %llukB, failcnt %lu\n",
		K((u64)page_counter_read(&memcg->memsw)),
		K((u64)memcg->memsw.limit), memcg->memsw.failcnt);
	pr_info("kmem: usage %llukB, limit %llukB, failcnt %lu\n",
		K((u64)page_counter_read(&memcg->kmem)),
		K((u64)memcg->kmem.limit), memcg->kmem.failcnt);
1229 1230

	for_each_mem_cgroup_tree(iter, memcg) {
T
Tejun Heo 已提交
1231 1232
		pr_info("Memory cgroup stats for ");
		pr_cont_cgroup_path(iter->css.cgroup);
1233 1234 1235
		pr_cont(":");

		for (i = 0; i < MEM_CGROUP_STAT_NSTATS; i++) {
1236
			if (i == MEM_CGROUP_STAT_SWAP && !do_memsw_account())
1237
				continue;
1238
			pr_cont(" %s:%luKB", mem_cgroup_stat_names[i],
1239 1240 1241 1242 1243 1244 1245 1246 1247
				K(mem_cgroup_read_stat(iter, i)));
		}

		for (i = 0; i < NR_LRU_LISTS; i++)
			pr_cont(" %s:%luKB", mem_cgroup_lru_names[i],
				K(mem_cgroup_nr_lru_pages(iter, BIT(i))));

		pr_cont("\n");
	}
1248
	mutex_unlock(&oom_info_lock);
1249 1250
}

1251 1252 1253 1254
/*
 * This function returns the number of memcg under hierarchy tree. Returns
 * 1(self count) if no children.
 */
1255
static int mem_cgroup_count_children(struct mem_cgroup *memcg)
1256 1257
{
	int num = 0;
K
KAMEZAWA Hiroyuki 已提交
1258 1259
	struct mem_cgroup *iter;

1260
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
1261
		num++;
1262 1263 1264
	return num;
}

D
David Rientjes 已提交
1265 1266 1267
/*
 * Return the memory (and swap, if configured) limit for a memcg.
 */
1268
static unsigned long mem_cgroup_get_limit(struct mem_cgroup *memcg)
D
David Rientjes 已提交
1269
{
1270
	unsigned long limit;
1271

1272
	limit = memcg->memory.limit;
1273
	if (mem_cgroup_swappiness(memcg)) {
1274
		unsigned long memsw_limit;
1275

1276 1277
		memsw_limit = memcg->memsw.limit;
		limit = min(limit + total_swap_pages, memsw_limit);
1278 1279
	}
	return limit;
D
David Rientjes 已提交
1280 1281
}

1282 1283
static void mem_cgroup_out_of_memory(struct mem_cgroup *memcg, gfp_t gfp_mask,
				     int order)
1284
{
1285 1286 1287 1288 1289 1290
	struct oom_control oc = {
		.zonelist = NULL,
		.nodemask = NULL,
		.gfp_mask = gfp_mask,
		.order = order,
	};
1291 1292 1293 1294 1295 1296
	struct mem_cgroup *iter;
	unsigned long chosen_points = 0;
	unsigned long totalpages;
	unsigned int points = 0;
	struct task_struct *chosen = NULL;

1297 1298
	mutex_lock(&oom_lock);

1299
	/*
1300 1301 1302
	 * If current has a pending SIGKILL or is exiting, then automatically
	 * select it.  The goal is to allow it to allocate so that it may
	 * quickly exit and free its memory.
1303
	 */
1304
	if (fatal_signal_pending(current) || task_will_free_mem(current)) {
1305
		mark_oom_victim(current);
1306
		goto unlock;
1307 1308
	}

1309
	check_panic_on_oom(&oc, CONSTRAINT_MEMCG, memcg);
1310
	totalpages = mem_cgroup_get_limit(memcg) ? : 1;
1311
	for_each_mem_cgroup_tree(iter, memcg) {
1312
		struct css_task_iter it;
1313 1314
		struct task_struct *task;

1315 1316
		css_task_iter_start(&iter->css, &it);
		while ((task = css_task_iter_next(&it))) {
1317
			switch (oom_scan_process_thread(&oc, task, totalpages)) {
1318 1319 1320 1321 1322 1323 1324 1325 1326 1327
			case OOM_SCAN_SELECT:
				if (chosen)
					put_task_struct(chosen);
				chosen = task;
				chosen_points = ULONG_MAX;
				get_task_struct(chosen);
				/* fall through */
			case OOM_SCAN_CONTINUE:
				continue;
			case OOM_SCAN_ABORT:
1328
				css_task_iter_end(&it);
1329 1330 1331
				mem_cgroup_iter_break(memcg, iter);
				if (chosen)
					put_task_struct(chosen);
1332
				goto unlock;
1333 1334 1335 1336
			case OOM_SCAN_OK:
				break;
			};
			points = oom_badness(task, memcg, NULL, totalpages);
1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348
			if (!points || points < chosen_points)
				continue;
			/* Prefer thread group leaders for display purposes */
			if (points == chosen_points &&
			    thread_group_leader(chosen))
				continue;

			if (chosen)
				put_task_struct(chosen);
			chosen = task;
			chosen_points = points;
			get_task_struct(chosen);
1349
		}
1350
		css_task_iter_end(&it);
1351 1352
	}

1353 1354
	if (chosen) {
		points = chosen_points * 1000 / totalpages;
1355 1356
		oom_kill_process(&oc, chosen, points, totalpages, memcg,
				 "Memory cgroup out of memory");
1357 1358 1359
	}
unlock:
	mutex_unlock(&oom_lock);
1360 1361
}

1362 1363
#if MAX_NUMNODES > 1

1364 1365
/**
 * test_mem_cgroup_node_reclaimable
W
Wanpeng Li 已提交
1366
 * @memcg: the target memcg
1367 1368 1369 1370 1371 1372 1373
 * @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.
 */
1374
static bool test_mem_cgroup_node_reclaimable(struct mem_cgroup *memcg,
1375 1376
		int nid, bool noswap)
{
1377
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_FILE))
1378 1379 1380
		return true;
	if (noswap || !total_swap_pages)
		return false;
1381
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_ANON))
1382 1383 1384 1385
		return true;
	return false;

}
1386 1387 1388 1389 1390 1391 1392

/*
 * 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.
 *
 */
1393
static void mem_cgroup_may_update_nodemask(struct mem_cgroup *memcg)
1394 1395
{
	int nid;
1396 1397 1398 1399
	/*
	 * numainfo_events > 0 means there was at least NUMAINFO_EVENTS_TARGET
	 * pagein/pageout changes since the last update.
	 */
1400
	if (!atomic_read(&memcg->numainfo_events))
1401
		return;
1402
	if (atomic_inc_return(&memcg->numainfo_updating) > 1)
1403 1404 1405
		return;

	/* make a nodemask where this memcg uses memory from */
1406
	memcg->scan_nodes = node_states[N_MEMORY];
1407

1408
	for_each_node_mask(nid, node_states[N_MEMORY]) {
1409

1410 1411
		if (!test_mem_cgroup_node_reclaimable(memcg, nid, false))
			node_clear(nid, memcg->scan_nodes);
1412
	}
1413

1414 1415
	atomic_set(&memcg->numainfo_events, 0);
	atomic_set(&memcg->numainfo_updating, 0);
1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429
}

/*
 * 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.
 */
1430
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
1431 1432 1433
{
	int node;

1434 1435
	mem_cgroup_may_update_nodemask(memcg);
	node = memcg->last_scanned_node;
1436

1437
	node = next_node(node, memcg->scan_nodes);
1438
	if (node == MAX_NUMNODES)
1439
		node = first_node(memcg->scan_nodes);
1440 1441 1442 1443 1444 1445 1446 1447 1448
	/*
	 * 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();

1449
	memcg->last_scanned_node = node;
1450 1451 1452
	return node;
}
#else
1453
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
1454 1455 1456 1457 1458
{
	return 0;
}
#endif

1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473
static int mem_cgroup_soft_reclaim(struct mem_cgroup *root_memcg,
				   struct zone *zone,
				   gfp_t gfp_mask,
				   unsigned long *total_scanned)
{
	struct mem_cgroup *victim = NULL;
	int total = 0;
	int loop = 0;
	unsigned long excess;
	unsigned long nr_scanned;
	struct mem_cgroup_reclaim_cookie reclaim = {
		.zone = zone,
		.priority = 0,
	};

1474
	excess = soft_limit_excess(root_memcg);
1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502

	while (1) {
		victim = mem_cgroup_iter(root_memcg, victim, &reclaim);
		if (!victim) {
			loop++;
			if (loop >= 2) {
				/*
				 * If we have not been able to reclaim
				 * anything, it might because there are
				 * no reclaimable pages under this hierarchy
				 */
				if (!total)
					break;
				/*
				 * We want to do more targeted reclaim.
				 * 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) ||
					(loop > MEM_CGROUP_MAX_RECLAIM_LOOPS))
					break;
			}
			continue;
		}
		total += mem_cgroup_shrink_node_zone(victim, gfp_mask, false,
						     zone, &nr_scanned);
		*total_scanned += nr_scanned;
1503
		if (!soft_limit_excess(root_memcg))
1504
			break;
1505
	}
1506 1507
	mem_cgroup_iter_break(root_memcg, victim);
	return total;
1508 1509
}

1510 1511 1512 1513 1514 1515
#ifdef CONFIG_LOCKDEP
static struct lockdep_map memcg_oom_lock_dep_map = {
	.name = "memcg_oom_lock",
};
#endif

1516 1517
static DEFINE_SPINLOCK(memcg_oom_lock);

K
KAMEZAWA Hiroyuki 已提交
1518 1519 1520 1521
/*
 * Check OOM-Killer is already running under our hierarchy.
 * If someone is running, return false.
 */
1522
static bool mem_cgroup_oom_trylock(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
1523
{
1524
	struct mem_cgroup *iter, *failed = NULL;
1525

1526 1527
	spin_lock(&memcg_oom_lock);

1528
	for_each_mem_cgroup_tree(iter, memcg) {
1529
		if (iter->oom_lock) {
1530 1531 1532 1533 1534
			/*
			 * this subtree of our hierarchy is already locked
			 * so we cannot give a lock.
			 */
			failed = iter;
1535 1536
			mem_cgroup_iter_break(memcg, iter);
			break;
1537 1538
		} else
			iter->oom_lock = true;
K
KAMEZAWA Hiroyuki 已提交
1539
	}
K
KAMEZAWA Hiroyuki 已提交
1540

1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551
	if (failed) {
		/*
		 * OK, we failed to lock the whole subtree so we have
		 * to clean up what we set up to the failing subtree
		 */
		for_each_mem_cgroup_tree(iter, memcg) {
			if (iter == failed) {
				mem_cgroup_iter_break(memcg, iter);
				break;
			}
			iter->oom_lock = false;
1552
		}
1553 1554
	} else
		mutex_acquire(&memcg_oom_lock_dep_map, 0, 1, _RET_IP_);
1555 1556 1557 1558

	spin_unlock(&memcg_oom_lock);

	return !failed;
1559
}
1560

1561
static void mem_cgroup_oom_unlock(struct mem_cgroup *memcg)
1562
{
K
KAMEZAWA Hiroyuki 已提交
1563 1564
	struct mem_cgroup *iter;

1565
	spin_lock(&memcg_oom_lock);
1566
	mutex_release(&memcg_oom_lock_dep_map, 1, _RET_IP_);
1567
	for_each_mem_cgroup_tree(iter, memcg)
1568
		iter->oom_lock = false;
1569
	spin_unlock(&memcg_oom_lock);
1570 1571
}

1572
static void mem_cgroup_mark_under_oom(struct mem_cgroup *memcg)
1573 1574 1575
{
	struct mem_cgroup *iter;

1576
	spin_lock(&memcg_oom_lock);
1577
	for_each_mem_cgroup_tree(iter, memcg)
1578 1579
		iter->under_oom++;
	spin_unlock(&memcg_oom_lock);
1580 1581
}

1582
static void mem_cgroup_unmark_under_oom(struct mem_cgroup *memcg)
1583 1584 1585
{
	struct mem_cgroup *iter;

K
KAMEZAWA Hiroyuki 已提交
1586 1587
	/*
	 * When a new child is created while the hierarchy is under oom,
1588
	 * mem_cgroup_oom_lock() may not be called. Watch for underflow.
K
KAMEZAWA Hiroyuki 已提交
1589
	 */
1590
	spin_lock(&memcg_oom_lock);
1591
	for_each_mem_cgroup_tree(iter, memcg)
1592 1593 1594
		if (iter->under_oom > 0)
			iter->under_oom--;
	spin_unlock(&memcg_oom_lock);
1595 1596
}

K
KAMEZAWA Hiroyuki 已提交
1597 1598
static DECLARE_WAIT_QUEUE_HEAD(memcg_oom_waitq);

K
KAMEZAWA Hiroyuki 已提交
1599
struct oom_wait_info {
1600
	struct mem_cgroup *memcg;
K
KAMEZAWA Hiroyuki 已提交
1601 1602 1603 1604 1605 1606
	wait_queue_t	wait;
};

static int memcg_oom_wake_function(wait_queue_t *wait,
	unsigned mode, int sync, void *arg)
{
1607 1608
	struct mem_cgroup *wake_memcg = (struct mem_cgroup *)arg;
	struct mem_cgroup *oom_wait_memcg;
K
KAMEZAWA Hiroyuki 已提交
1609 1610 1611
	struct oom_wait_info *oom_wait_info;

	oom_wait_info = container_of(wait, struct oom_wait_info, wait);
1612
	oom_wait_memcg = oom_wait_info->memcg;
K
KAMEZAWA Hiroyuki 已提交
1613

1614 1615
	if (!mem_cgroup_is_descendant(wake_memcg, oom_wait_memcg) &&
	    !mem_cgroup_is_descendant(oom_wait_memcg, wake_memcg))
K
KAMEZAWA Hiroyuki 已提交
1616 1617 1618 1619
		return 0;
	return autoremove_wake_function(wait, mode, sync, arg);
}

1620
static void memcg_oom_recover(struct mem_cgroup *memcg)
1621
{
1622 1623 1624 1625 1626 1627 1628 1629 1630
	/*
	 * For the following lockless ->under_oom test, the only required
	 * guarantee is that it must see the state asserted by an OOM when
	 * this function is called as a result of userland actions
	 * triggered by the notification of the OOM.  This is trivially
	 * achieved by invoking mem_cgroup_mark_under_oom() before
	 * triggering notification.
	 */
	if (memcg && memcg->under_oom)
1631
		__wake_up(&memcg_oom_waitq, TASK_NORMAL, 0, memcg);
1632 1633
}

1634
static void mem_cgroup_oom(struct mem_cgroup *memcg, gfp_t mask, int order)
1635
{
T
Tejun Heo 已提交
1636
	if (!current->memcg_may_oom)
1637
		return;
K
KAMEZAWA Hiroyuki 已提交
1638
	/*
1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650
	 * We are in the middle of the charge context here, so we
	 * don't want to block when potentially sitting on a callstack
	 * that holds all kinds of filesystem and mm locks.
	 *
	 * Also, the caller may handle a failed allocation gracefully
	 * (like optional page cache readahead) and so an OOM killer
	 * invocation might not even be necessary.
	 *
	 * That's why we don't do anything here except remember the
	 * OOM context and then deal with it at the end of the page
	 * fault when the stack is unwound, the locks are released,
	 * and when we know whether the fault was overall successful.
K
KAMEZAWA Hiroyuki 已提交
1651
	 */
1652
	css_get(&memcg->css);
T
Tejun Heo 已提交
1653 1654 1655
	current->memcg_in_oom = memcg;
	current->memcg_oom_gfp_mask = mask;
	current->memcg_oom_order = order;
1656 1657 1658 1659
}

/**
 * mem_cgroup_oom_synchronize - complete memcg OOM handling
1660
 * @handle: actually kill/wait or just clean up the OOM state
1661
 *
1662 1663
 * This has to be called at the end of a page fault if the memcg OOM
 * handler was enabled.
1664
 *
1665
 * Memcg supports userspace OOM handling where failed allocations must
1666 1667 1668 1669
 * sleep on a waitqueue until the userspace task resolves the
 * situation.  Sleeping directly in the charge context with all kinds
 * of locks held is not a good idea, instead we remember an OOM state
 * in the task and mem_cgroup_oom_synchronize() has to be called at
1670
 * the end of the page fault to complete the OOM handling.
1671 1672
 *
 * Returns %true if an ongoing memcg OOM situation was detected and
1673
 * completed, %false otherwise.
1674
 */
1675
bool mem_cgroup_oom_synchronize(bool handle)
1676
{
T
Tejun Heo 已提交
1677
	struct mem_cgroup *memcg = current->memcg_in_oom;
1678
	struct oom_wait_info owait;
1679
	bool locked;
1680 1681 1682

	/* OOM is global, do not handle */
	if (!memcg)
1683
		return false;
1684

1685
	if (!handle || oom_killer_disabled)
1686
		goto cleanup;
1687 1688 1689 1690 1691 1692

	owait.memcg = memcg;
	owait.wait.flags = 0;
	owait.wait.func = memcg_oom_wake_function;
	owait.wait.private = current;
	INIT_LIST_HEAD(&owait.wait.task_list);
K
KAMEZAWA Hiroyuki 已提交
1693

1694
	prepare_to_wait(&memcg_oom_waitq, &owait.wait, TASK_KILLABLE);
1695 1696 1697 1698 1699 1700 1701 1702 1703 1704
	mem_cgroup_mark_under_oom(memcg);

	locked = mem_cgroup_oom_trylock(memcg);

	if (locked)
		mem_cgroup_oom_notify(memcg);

	if (locked && !memcg->oom_kill_disable) {
		mem_cgroup_unmark_under_oom(memcg);
		finish_wait(&memcg_oom_waitq, &owait.wait);
T
Tejun Heo 已提交
1705 1706
		mem_cgroup_out_of_memory(memcg, current->memcg_oom_gfp_mask,
					 current->memcg_oom_order);
1707
	} else {
1708
		schedule();
1709 1710 1711 1712 1713
		mem_cgroup_unmark_under_oom(memcg);
		finish_wait(&memcg_oom_waitq, &owait.wait);
	}

	if (locked) {
1714 1715 1716 1717 1718 1719 1720 1721
		mem_cgroup_oom_unlock(memcg);
		/*
		 * There is no guarantee that an OOM-lock contender
		 * sees the wakeups triggered by the OOM kill
		 * uncharges.  Wake any sleepers explicitely.
		 */
		memcg_oom_recover(memcg);
	}
1722
cleanup:
T
Tejun Heo 已提交
1723
	current->memcg_in_oom = NULL;
1724
	css_put(&memcg->css);
K
KAMEZAWA Hiroyuki 已提交
1725
	return true;
1726 1727
}

1728 1729 1730
/**
 * mem_cgroup_begin_page_stat - begin a page state statistics transaction
 * @page: page that is going to change accounted state
1731
 *
1732 1733 1734
 * This function must mark the beginning of an accounted page state
 * change to prevent double accounting when the page is concurrently
 * being moved to another memcg:
1735
 *
1736
 *   memcg = mem_cgroup_begin_page_stat(page);
1737 1738
 *   if (TestClearPageState(page))
 *     mem_cgroup_update_page_stat(memcg, state, -1);
1739
 *   mem_cgroup_end_page_stat(memcg);
1740
 */
1741
struct mem_cgroup *mem_cgroup_begin_page_stat(struct page *page)
1742 1743
{
	struct mem_cgroup *memcg;
1744
	unsigned long flags;
1745

1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757
	/*
	 * The RCU lock is held throughout the transaction.  The fast
	 * path can get away without acquiring the memcg->move_lock
	 * because page moving starts with an RCU grace period.
	 *
	 * The RCU lock also protects the memcg from being freed when
	 * the page state that is going to change is the only thing
	 * preventing the page from being uncharged.
	 * E.g. end-writeback clearing PageWriteback(), which allows
	 * migration to go ahead and uncharge the page before the
	 * account transaction might be complete.
	 */
1758 1759 1760 1761
	rcu_read_lock();

	if (mem_cgroup_disabled())
		return NULL;
1762
again:
1763
	memcg = page->mem_cgroup;
1764
	if (unlikely(!memcg))
1765 1766
		return NULL;

Q
Qiang Huang 已提交
1767
	if (atomic_read(&memcg->moving_account) <= 0)
1768
		return memcg;
1769

1770
	spin_lock_irqsave(&memcg->move_lock, flags);
1771
	if (memcg != page->mem_cgroup) {
1772
		spin_unlock_irqrestore(&memcg->move_lock, flags);
1773 1774
		goto again;
	}
1775 1776 1777 1778 1779 1780 1781 1782

	/*
	 * When charge migration first begins, we can have locked and
	 * unlocked page stat updates happening concurrently.  Track
	 * the task who has the lock for mem_cgroup_end_page_stat().
	 */
	memcg->move_lock_task = current;
	memcg->move_lock_flags = flags;
1783 1784

	return memcg;
1785
}
1786
EXPORT_SYMBOL(mem_cgroup_begin_page_stat);
1787

1788 1789 1790 1791
/**
 * mem_cgroup_end_page_stat - finish a page state statistics transaction
 * @memcg: the memcg that was accounted against
 */
1792
void mem_cgroup_end_page_stat(struct mem_cgroup *memcg)
1793
{
1794 1795 1796 1797 1798 1799 1800 1801
	if (memcg && memcg->move_lock_task == current) {
		unsigned long flags = memcg->move_lock_flags;

		memcg->move_lock_task = NULL;
		memcg->move_lock_flags = 0;

		spin_unlock_irqrestore(&memcg->move_lock, flags);
	}
1802

1803
	rcu_read_unlock();
1804
}
1805
EXPORT_SYMBOL(mem_cgroup_end_page_stat);
1806

1807 1808 1809 1810
/*
 * size of first charge trial. "32" comes from vmscan.c's magic value.
 * TODO: maybe necessary to use big numbers in big irons.
 */
1811
#define CHARGE_BATCH	32U
1812 1813
struct memcg_stock_pcp {
	struct mem_cgroup *cached; /* this never be root cgroup */
1814
	unsigned int nr_pages;
1815
	struct work_struct work;
1816
	unsigned long flags;
1817
#define FLUSHING_CACHED_CHARGE	0
1818 1819
};
static DEFINE_PER_CPU(struct memcg_stock_pcp, memcg_stock);
1820
static DEFINE_MUTEX(percpu_charge_mutex);
1821

1822 1823 1824 1825 1826 1827 1828 1829 1830 1831
/**
 * consume_stock: Try to consume stocked charge on this cpu.
 * @memcg: memcg to consume from.
 * @nr_pages: how many pages to charge.
 *
 * The charges will only happen if @memcg matches the current cpu's memcg
 * stock, and at least @nr_pages are available in that stock.  Failure to
 * service an allocation will refill the stock.
 *
 * returns true if successful, false otherwise.
1832
 */
1833
static bool consume_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
1834 1835
{
	struct memcg_stock_pcp *stock;
1836
	bool ret = false;
1837

1838
	if (nr_pages > CHARGE_BATCH)
1839
		return ret;
1840

1841
	stock = &get_cpu_var(memcg_stock);
1842
	if (memcg == stock->cached && stock->nr_pages >= nr_pages) {
1843
		stock->nr_pages -= nr_pages;
1844 1845
		ret = true;
	}
1846 1847 1848 1849 1850
	put_cpu_var(memcg_stock);
	return ret;
}

/*
1851
 * Returns stocks cached in percpu and reset cached information.
1852 1853 1854 1855 1856
 */
static void drain_stock(struct memcg_stock_pcp *stock)
{
	struct mem_cgroup *old = stock->cached;

1857
	if (stock->nr_pages) {
1858
		page_counter_uncharge(&old->memory, stock->nr_pages);
1859
		if (do_memsw_account())
1860
			page_counter_uncharge(&old->memsw, stock->nr_pages);
1861
		css_put_many(&old->css, stock->nr_pages);
1862
		stock->nr_pages = 0;
1863 1864 1865 1866 1867 1868 1869 1870 1871 1872
	}
	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)
{
1873
	struct memcg_stock_pcp *stock = this_cpu_ptr(&memcg_stock);
1874
	drain_stock(stock);
1875
	clear_bit(FLUSHING_CACHED_CHARGE, &stock->flags);
1876 1877 1878
}

/*
1879
 * Cache charges(val) to local per_cpu area.
1880
 * This will be consumed by consume_stock() function, later.
1881
 */
1882
static void refill_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
1883 1884 1885
{
	struct memcg_stock_pcp *stock = &get_cpu_var(memcg_stock);

1886
	if (stock->cached != memcg) { /* reset if necessary */
1887
		drain_stock(stock);
1888
		stock->cached = memcg;
1889
	}
1890
	stock->nr_pages += nr_pages;
1891 1892 1893 1894
	put_cpu_var(memcg_stock);
}

/*
1895
 * Drains all per-CPU charge caches for given root_memcg resp. subtree
1896
 * of the hierarchy under it.
1897
 */
1898
static void drain_all_stock(struct mem_cgroup *root_memcg)
1899
{
1900
	int cpu, curcpu;
1901

1902 1903 1904
	/* If someone's already draining, avoid adding running more workers. */
	if (!mutex_trylock(&percpu_charge_mutex))
		return;
1905 1906
	/* Notify other cpus that system-wide "drain" is running */
	get_online_cpus();
1907
	curcpu = get_cpu();
1908 1909
	for_each_online_cpu(cpu) {
		struct memcg_stock_pcp *stock = &per_cpu(memcg_stock, cpu);
1910
		struct mem_cgroup *memcg;
1911

1912 1913
		memcg = stock->cached;
		if (!memcg || !stock->nr_pages)
1914
			continue;
1915
		if (!mem_cgroup_is_descendant(memcg, root_memcg))
1916
			continue;
1917 1918 1919 1920 1921 1922
		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);
		}
1923
	}
1924
	put_cpu();
A
Andrew Morton 已提交
1925
	put_online_cpus();
1926
	mutex_unlock(&percpu_charge_mutex);
1927 1928
}

1929
static int memcg_cpu_hotplug_callback(struct notifier_block *nb,
1930 1931 1932 1933 1934 1935
					unsigned long action,
					void *hcpu)
{
	int cpu = (unsigned long)hcpu;
	struct memcg_stock_pcp *stock;

1936
	if (action == CPU_ONLINE)
1937 1938
		return NOTIFY_OK;

1939
	if (action != CPU_DEAD && action != CPU_DEAD_FROZEN)
1940
		return NOTIFY_OK;
1941

1942 1943 1944 1945 1946
	stock = &per_cpu(memcg_stock, cpu);
	drain_stock(stock);
	return NOTIFY_OK;
}

1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966
static void reclaim_high(struct mem_cgroup *memcg,
			 unsigned int nr_pages,
			 gfp_t gfp_mask)
{
	do {
		if (page_counter_read(&memcg->memory) <= memcg->high)
			continue;
		mem_cgroup_events(memcg, MEMCG_HIGH, 1);
		try_to_free_mem_cgroup_pages(memcg, nr_pages, gfp_mask, true);
	} while ((memcg = parent_mem_cgroup(memcg)));
}

static void high_work_func(struct work_struct *work)
{
	struct mem_cgroup *memcg;

	memcg = container_of(work, struct mem_cgroup, high_work);
	reclaim_high(memcg, CHARGE_BATCH, GFP_KERNEL);
}

1967 1968 1969 1970 1971 1972 1973
/*
 * Scheduled by try_charge() to be executed from the userland return path
 * and reclaims memory over the high limit.
 */
void mem_cgroup_handle_over_high(void)
{
	unsigned int nr_pages = current->memcg_nr_pages_over_high;
1974
	struct mem_cgroup *memcg;
1975 1976 1977 1978

	if (likely(!nr_pages))
		return;

1979 1980
	memcg = get_mem_cgroup_from_mm(current->mm);
	reclaim_high(memcg, nr_pages, GFP_KERNEL);
1981 1982 1983 1984
	css_put(&memcg->css);
	current->memcg_nr_pages_over_high = 0;
}

1985 1986
static int try_charge(struct mem_cgroup *memcg, gfp_t gfp_mask,
		      unsigned int nr_pages)
1987
{
1988
	unsigned int batch = max(CHARGE_BATCH, nr_pages);
1989
	int nr_retries = MEM_CGROUP_RECLAIM_RETRIES;
1990
	struct mem_cgroup *mem_over_limit;
1991
	struct page_counter *counter;
1992
	unsigned long nr_reclaimed;
1993 1994
	bool may_swap = true;
	bool drained = false;
1995

1996
	if (mem_cgroup_is_root(memcg))
1997
		return 0;
1998
retry:
1999
	if (consume_stock(memcg, nr_pages))
2000
		return 0;
2001

2002
	if (!do_memsw_account() ||
2003 2004
	    page_counter_try_charge(&memcg->memsw, batch, &counter)) {
		if (page_counter_try_charge(&memcg->memory, batch, &counter))
2005
			goto done_restock;
2006
		if (do_memsw_account())
2007 2008
			page_counter_uncharge(&memcg->memsw, batch);
		mem_over_limit = mem_cgroup_from_counter(counter, memory);
2009
	} else {
2010
		mem_over_limit = mem_cgroup_from_counter(counter, memsw);
2011
		may_swap = false;
2012
	}
2013

2014 2015 2016 2017
	if (batch > nr_pages) {
		batch = nr_pages;
		goto retry;
	}
2018

2019 2020 2021 2022 2023 2024 2025 2026 2027
	/*
	 * Unlike in global OOM situations, memcg is not in a physical
	 * memory shortage.  Allow dying and OOM-killed tasks to
	 * bypass the last charges so that they can exit quickly and
	 * free their memory.
	 */
	if (unlikely(test_thread_flag(TIF_MEMDIE) ||
		     fatal_signal_pending(current) ||
		     current->flags & PF_EXITING))
2028
		goto force;
2029 2030 2031 2032

	if (unlikely(task_in_memcg_oom(current)))
		goto nomem;

2033
	if (!gfpflags_allow_blocking(gfp_mask))
2034
		goto nomem;
2035

2036 2037
	mem_cgroup_events(mem_over_limit, MEMCG_MAX, 1);

2038 2039
	nr_reclaimed = try_to_free_mem_cgroup_pages(mem_over_limit, nr_pages,
						    gfp_mask, may_swap);
2040

2041
	if (mem_cgroup_margin(mem_over_limit) >= nr_pages)
2042
		goto retry;
2043

2044
	if (!drained) {
2045
		drain_all_stock(mem_over_limit);
2046 2047 2048 2049
		drained = true;
		goto retry;
	}

2050 2051
	if (gfp_mask & __GFP_NORETRY)
		goto nomem;
2052 2053 2054 2055 2056 2057 2058 2059 2060
	/*
	 * 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.
	 */
2061
	if (nr_reclaimed && nr_pages <= (1 << PAGE_ALLOC_COSTLY_ORDER))
2062 2063 2064 2065 2066 2067 2068 2069
		goto 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))
		goto retry;

2070 2071 2072
	if (nr_retries--)
		goto retry;

2073
	if (gfp_mask & __GFP_NOFAIL)
2074
		goto force;
2075

2076
	if (fatal_signal_pending(current))
2077
		goto force;
2078

2079 2080
	mem_cgroup_events(mem_over_limit, MEMCG_OOM, 1);

2081 2082
	mem_cgroup_oom(mem_over_limit, gfp_mask,
		       get_order(nr_pages * PAGE_SIZE));
2083
nomem:
2084
	if (!(gfp_mask & __GFP_NOFAIL))
2085
		return -ENOMEM;
2086 2087 2088 2089 2090 2091 2092
force:
	/*
	 * The allocation either can't fail or will lead to more memory
	 * being freed very soon.  Allow memory usage go over the limit
	 * temporarily by force charging it.
	 */
	page_counter_charge(&memcg->memory, nr_pages);
2093
	if (do_memsw_account())
2094 2095 2096 2097
		page_counter_charge(&memcg->memsw, nr_pages);
	css_get_many(&memcg->css, nr_pages);

	return 0;
2098 2099

done_restock:
2100
	css_get_many(&memcg->css, batch);
2101 2102
	if (batch > nr_pages)
		refill_stock(memcg, batch - nr_pages);
2103

2104
	/*
2105 2106
	 * If the hierarchy is above the normal consumption range, schedule
	 * reclaim on returning to userland.  We can perform reclaim here
2107
	 * if __GFP_RECLAIM but let's always punt for simplicity and so that
2108 2109 2110 2111
	 * GFP_KERNEL can consistently be used during reclaim.  @memcg is
	 * not recorded as it most likely matches current's and won't
	 * change in the meantime.  As high limit is checked again before
	 * reclaim, the cost of mismatch is negligible.
2112 2113
	 */
	do {
2114
		if (page_counter_read(&memcg->memory) > memcg->high) {
2115 2116 2117 2118 2119
			/* Don't bother a random interrupted task */
			if (in_interrupt()) {
				schedule_work(&memcg->high_work);
				break;
			}
V
Vladimir Davydov 已提交
2120
			current->memcg_nr_pages_over_high += batch;
2121 2122 2123
			set_notify_resume(current);
			break;
		}
2124
	} while ((memcg = parent_mem_cgroup(memcg)));
2125 2126

	return 0;
2127
}
2128

2129
static void cancel_charge(struct mem_cgroup *memcg, unsigned int nr_pages)
2130
{
2131 2132 2133
	if (mem_cgroup_is_root(memcg))
		return;

2134
	page_counter_uncharge(&memcg->memory, nr_pages);
2135
	if (do_memsw_account())
2136
		page_counter_uncharge(&memcg->memsw, nr_pages);
2137

2138
	css_put_many(&memcg->css, nr_pages);
2139 2140
}

2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171
static void lock_page_lru(struct page *page, int *isolated)
{
	struct zone *zone = page_zone(page);

	spin_lock_irq(&zone->lru_lock);
	if (PageLRU(page)) {
		struct lruvec *lruvec;

		lruvec = mem_cgroup_page_lruvec(page, zone);
		ClearPageLRU(page);
		del_page_from_lru_list(page, lruvec, page_lru(page));
		*isolated = 1;
	} else
		*isolated = 0;
}

static void unlock_page_lru(struct page *page, int isolated)
{
	struct zone *zone = page_zone(page);

	if (isolated) {
		struct lruvec *lruvec;

		lruvec = mem_cgroup_page_lruvec(page, zone);
		VM_BUG_ON_PAGE(PageLRU(page), page);
		SetPageLRU(page);
		add_page_to_lru_list(page, lruvec, page_lru(page));
	}
	spin_unlock_irq(&zone->lru_lock);
}

2172
static void commit_charge(struct page *page, struct mem_cgroup *memcg,
2173
			  bool lrucare)
2174
{
2175
	int isolated;
2176

2177
	VM_BUG_ON_PAGE(page->mem_cgroup, page);
2178 2179 2180 2181 2182

	/*
	 * In some cases, SwapCache and FUSE(splice_buf->radixtree), the page
	 * may already be on some other mem_cgroup's LRU.  Take care of it.
	 */
2183 2184
	if (lrucare)
		lock_page_lru(page, &isolated);
2185

2186 2187
	/*
	 * Nobody should be changing or seriously looking at
2188
	 * page->mem_cgroup at this point:
2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199
	 *
	 * - the page is uncharged
	 *
	 * - the page is off-LRU
	 *
	 * - an anonymous fault has exclusive page access, except for
	 *   a locked page table
	 *
	 * - a page cache insertion, a swapin fault, or a migration
	 *   have the page locked
	 */
2200
	page->mem_cgroup = memcg;
2201

2202 2203
	if (lrucare)
		unlock_page_lru(page, isolated);
2204
}
2205

2206
#ifndef CONFIG_SLOB
2207
static int memcg_alloc_cache_id(void)
2208
{
2209 2210 2211
	int id, size;
	int err;

2212
	id = ida_simple_get(&memcg_cache_ida,
2213 2214 2215
			    0, MEMCG_CACHES_MAX_SIZE, GFP_KERNEL);
	if (id < 0)
		return id;
2216

2217
	if (id < memcg_nr_cache_ids)
2218 2219 2220 2221 2222 2223
		return id;

	/*
	 * There's no space for the new id in memcg_caches arrays,
	 * so we have to grow them.
	 */
2224
	down_write(&memcg_cache_ids_sem);
2225 2226

	size = 2 * (id + 1);
2227 2228 2229 2230 2231
	if (size < MEMCG_CACHES_MIN_SIZE)
		size = MEMCG_CACHES_MIN_SIZE;
	else if (size > MEMCG_CACHES_MAX_SIZE)
		size = MEMCG_CACHES_MAX_SIZE;

2232
	err = memcg_update_all_caches(size);
2233 2234
	if (!err)
		err = memcg_update_all_list_lrus(size);
2235 2236 2237 2238 2239
	if (!err)
		memcg_nr_cache_ids = size;

	up_write(&memcg_cache_ids_sem);

2240
	if (err) {
2241
		ida_simple_remove(&memcg_cache_ida, id);
2242 2243 2244 2245 2246 2247 2248
		return err;
	}
	return id;
}

static void memcg_free_cache_id(int id)
{
2249
	ida_simple_remove(&memcg_cache_ida, id);
2250 2251
}

2252
struct memcg_kmem_cache_create_work {
2253 2254 2255 2256 2257
	struct mem_cgroup *memcg;
	struct kmem_cache *cachep;
	struct work_struct work;
};

2258
static void memcg_kmem_cache_create_func(struct work_struct *w)
2259
{
2260 2261
	struct memcg_kmem_cache_create_work *cw =
		container_of(w, struct memcg_kmem_cache_create_work, work);
2262 2263
	struct mem_cgroup *memcg = cw->memcg;
	struct kmem_cache *cachep = cw->cachep;
2264

2265
	memcg_create_kmem_cache(memcg, cachep);
2266

2267
	css_put(&memcg->css);
2268 2269 2270 2271 2272 2273
	kfree(cw);
}

/*
 * Enqueue the creation of a per-memcg kmem_cache.
 */
2274 2275
static void __memcg_schedule_kmem_cache_create(struct mem_cgroup *memcg,
					       struct kmem_cache *cachep)
2276
{
2277
	struct memcg_kmem_cache_create_work *cw;
2278

2279
	cw = kmalloc(sizeof(*cw), GFP_NOWAIT);
2280
	if (!cw)
2281
		return;
2282 2283

	css_get(&memcg->css);
2284 2285 2286

	cw->memcg = memcg;
	cw->cachep = cachep;
2287
	INIT_WORK(&cw->work, memcg_kmem_cache_create_func);
2288 2289 2290 2291

	schedule_work(&cw->work);
}

2292 2293
static void memcg_schedule_kmem_cache_create(struct mem_cgroup *memcg,
					     struct kmem_cache *cachep)
2294 2295 2296 2297
{
	/*
	 * We need to stop accounting when we kmalloc, because if the
	 * corresponding kmalloc cache is not yet created, the first allocation
2298
	 * in __memcg_schedule_kmem_cache_create will recurse.
2299 2300 2301 2302 2303 2304 2305
	 *
	 * However, it is better to enclose the whole function. Depending on
	 * the debugging options enabled, INIT_WORK(), for instance, can
	 * trigger an allocation. This too, will make us recurse. Because at
	 * this point we can't allow ourselves back into memcg_kmem_get_cache,
	 * the safest choice is to do it like this, wrapping the whole function.
	 */
2306
	current->memcg_kmem_skip_account = 1;
2307
	__memcg_schedule_kmem_cache_create(memcg, cachep);
2308
	current->memcg_kmem_skip_account = 0;
2309
}
2310

2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323
/*
 * Return the kmem_cache we're supposed to use for a slab allocation.
 * We try to use the current memcg's version of the cache.
 *
 * If the cache does not exist yet, if we are the first user of it,
 * we either create it immediately, if possible, or create it asynchronously
 * in a workqueue.
 * In the latter case, we will let the current allocation go through with
 * the original cache.
 *
 * Can't be called in interrupt context or from kernel threads.
 * This function needs to be called with rcu_read_lock() held.
 */
V
Vladimir Davydov 已提交
2324
struct kmem_cache *__memcg_kmem_get_cache(struct kmem_cache *cachep, gfp_t gfp)
2325 2326
{
	struct mem_cgroup *memcg;
2327
	struct kmem_cache *memcg_cachep;
2328
	int kmemcg_id;
2329

2330
	VM_BUG_ON(!is_root_cache(cachep));
2331

V
Vladimir Davydov 已提交
2332 2333 2334 2335 2336 2337
	if (cachep->flags & SLAB_ACCOUNT)
		gfp |= __GFP_ACCOUNT;

	if (!(gfp & __GFP_ACCOUNT))
		return cachep;

2338
	if (current->memcg_kmem_skip_account)
2339 2340
		return cachep;

2341
	memcg = get_mem_cgroup_from_mm(current->mm);
2342
	kmemcg_id = READ_ONCE(memcg->kmemcg_id);
2343
	if (kmemcg_id < 0)
2344
		goto out;
2345

2346
	memcg_cachep = cache_from_memcg_idx(cachep, kmemcg_id);
2347 2348
	if (likely(memcg_cachep))
		return memcg_cachep;
2349 2350 2351 2352 2353 2354 2355 2356 2357

	/*
	 * If we are in a safe context (can wait, and not in interrupt
	 * context), we could be be predictable and return right away.
	 * This would guarantee that the allocation being performed
	 * already belongs in the new cache.
	 *
	 * However, there are some clashes that can arrive from locking.
	 * For instance, because we acquire the slab_mutex while doing
2358 2359 2360
	 * memcg_create_kmem_cache, this means no further allocation
	 * could happen with the slab_mutex held. So it's better to
	 * defer everything.
2361
	 */
2362
	memcg_schedule_kmem_cache_create(memcg, cachep);
2363
out:
2364
	css_put(&memcg->css);
2365
	return cachep;
2366 2367
}

2368 2369 2370
void __memcg_kmem_put_cache(struct kmem_cache *cachep)
{
	if (!is_root_cache(cachep))
2371
		css_put(&cachep->memcg_params.memcg->css);
2372 2373
}

2374 2375
int __memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
			      struct mem_cgroup *memcg)
2376
{
2377 2378
	unsigned int nr_pages = 1 << order;
	struct page_counter *counter;
2379 2380
	int ret;

2381
	if (!memcg_kmem_online(memcg))
2382
		return 0;
2383

2384 2385
	if (!page_counter_try_charge(&memcg->kmem, nr_pages, &counter))
		return -ENOMEM;
2386

2387 2388 2389 2390
	ret = try_charge(memcg, gfp, nr_pages);
	if (ret) {
		page_counter_uncharge(&memcg->kmem, nr_pages);
		return ret;
2391 2392
	}

2393
	page->mem_cgroup = memcg;
2394

2395
	return 0;
2396 2397
}

2398
int __memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
2399
{
2400 2401
	struct mem_cgroup *memcg;
	int ret;
2402

2403 2404
	memcg = get_mem_cgroup_from_mm(current->mm);
	ret = __memcg_kmem_charge_memcg(page, gfp, order, memcg);
2405
	css_put(&memcg->css);
2406
	return ret;
2407 2408
}

2409
void __memcg_kmem_uncharge(struct page *page, int order)
2410
{
2411
	struct mem_cgroup *memcg = page->mem_cgroup;
2412
	unsigned int nr_pages = 1 << order;
2413 2414 2415 2416

	if (!memcg)
		return;

2417
	VM_BUG_ON_PAGE(mem_cgroup_is_root(memcg), page);
2418

2419 2420
	page_counter_uncharge(&memcg->kmem, nr_pages);
	page_counter_uncharge(&memcg->memory, nr_pages);
2421
	if (do_memsw_account())
2422
		page_counter_uncharge(&memcg->memsw, nr_pages);
2423

2424
	page->mem_cgroup = NULL;
2425
	css_put_many(&memcg->css, nr_pages);
2426
}
2427
#endif /* !CONFIG_SLOB */
2428

2429 2430 2431 2432
#ifdef CONFIG_TRANSPARENT_HUGEPAGE

/*
 * Because tail pages are not marked as "used", set it. We're under
2433
 * zone->lru_lock and migration entries setup in all page mappings.
2434
 */
2435
void mem_cgroup_split_huge_fixup(struct page *head)
2436
{
2437
	int i;
2438

2439 2440
	if (mem_cgroup_disabled())
		return;
2441

2442
	for (i = 1; i < HPAGE_PMD_NR; i++)
2443
		head[i].mem_cgroup = head->mem_cgroup;
2444

2445
	__this_cpu_sub(head->mem_cgroup->stat->count[MEM_CGROUP_STAT_RSS_HUGE],
2446
		       HPAGE_PMD_NR);
2447
}
2448
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
2449

A
Andrew Morton 已提交
2450
#ifdef CONFIG_MEMCG_SWAP
2451 2452
static void mem_cgroup_swap_statistics(struct mem_cgroup *memcg,
					 bool charge)
K
KAMEZAWA Hiroyuki 已提交
2453
{
2454 2455
	int val = (charge) ? 1 : -1;
	this_cpu_add(memcg->stat->count[MEM_CGROUP_STAT_SWAP], val);
K
KAMEZAWA Hiroyuki 已提交
2456
}
2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468

/**
 * 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
 *
 * 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.
 *
2469
 * The caller must have charged to @to, IOW, called page_counter_charge() about
2470 2471 2472
 * both res and memsw, and called css_get().
 */
static int mem_cgroup_move_swap_account(swp_entry_t entry,
2473
				struct mem_cgroup *from, struct mem_cgroup *to)
2474 2475 2476
{
	unsigned short old_id, new_id;

L
Li Zefan 已提交
2477 2478
	old_id = mem_cgroup_id(from);
	new_id = mem_cgroup_id(to);
2479 2480 2481

	if (swap_cgroup_cmpxchg(entry, old_id, new_id) == old_id) {
		mem_cgroup_swap_statistics(from, false);
2482
		mem_cgroup_swap_statistics(to, true);
2483 2484 2485 2486 2487 2488
		return 0;
	}
	return -EINVAL;
}
#else
static inline int mem_cgroup_move_swap_account(swp_entry_t entry,
2489
				struct mem_cgroup *from, struct mem_cgroup *to)
2490 2491 2492
{
	return -EINVAL;
}
2493
#endif
K
KAMEZAWA Hiroyuki 已提交
2494

2495
static DEFINE_MUTEX(memcg_limit_mutex);
2496

2497
static int mem_cgroup_resize_limit(struct mem_cgroup *memcg,
2498
				   unsigned long limit)
2499
{
2500 2501 2502
	unsigned long curusage;
	unsigned long oldusage;
	bool enlarge = false;
2503
	int retry_count;
2504
	int ret;
2505 2506 2507 2508 2509 2510

	/*
	 * 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.
	 */
2511 2512
	retry_count = MEM_CGROUP_RECLAIM_RETRIES *
		      mem_cgroup_count_children(memcg);
2513

2514
	oldusage = page_counter_read(&memcg->memory);
2515

2516
	do {
2517 2518 2519 2520
		if (signal_pending(current)) {
			ret = -EINTR;
			break;
		}
2521 2522 2523 2524

		mutex_lock(&memcg_limit_mutex);
		if (limit > memcg->memsw.limit) {
			mutex_unlock(&memcg_limit_mutex);
2525
			ret = -EINVAL;
2526 2527
			break;
		}
2528 2529 2530 2531
		if (limit > memcg->memory.limit)
			enlarge = true;
		ret = page_counter_limit(&memcg->memory, limit);
		mutex_unlock(&memcg_limit_mutex);
2532 2533 2534 2535

		if (!ret)
			break;

2536 2537
		try_to_free_mem_cgroup_pages(memcg, 1, GFP_KERNEL, true);

2538
		curusage = page_counter_read(&memcg->memory);
2539
		/* Usage is reduced ? */
A
Andrew Morton 已提交
2540
		if (curusage >= oldusage)
2541 2542 2543
			retry_count--;
		else
			oldusage = curusage;
2544 2545
	} while (retry_count);

2546 2547
	if (!ret && enlarge)
		memcg_oom_recover(memcg);
2548

2549 2550 2551
	return ret;
}

L
Li Zefan 已提交
2552
static int mem_cgroup_resize_memsw_limit(struct mem_cgroup *memcg,
2553
					 unsigned long limit)
2554
{
2555 2556 2557
	unsigned long curusage;
	unsigned long oldusage;
	bool enlarge = false;
2558
	int retry_count;
2559
	int ret;
2560

2561
	/* see mem_cgroup_resize_res_limit */
2562 2563 2564 2565 2566 2567
	retry_count = MEM_CGROUP_RECLAIM_RETRIES *
		      mem_cgroup_count_children(memcg);

	oldusage = page_counter_read(&memcg->memsw);

	do {
2568 2569 2570 2571
		if (signal_pending(current)) {
			ret = -EINTR;
			break;
		}
2572 2573 2574 2575

		mutex_lock(&memcg_limit_mutex);
		if (limit < memcg->memory.limit) {
			mutex_unlock(&memcg_limit_mutex);
2576 2577 2578
			ret = -EINVAL;
			break;
		}
2579 2580 2581 2582
		if (limit > memcg->memsw.limit)
			enlarge = true;
		ret = page_counter_limit(&memcg->memsw, limit);
		mutex_unlock(&memcg_limit_mutex);
2583 2584 2585 2586

		if (!ret)
			break;

2587 2588
		try_to_free_mem_cgroup_pages(memcg, 1, GFP_KERNEL, false);

2589
		curusage = page_counter_read(&memcg->memsw);
2590
		/* Usage is reduced ? */
2591
		if (curusage >= oldusage)
2592
			retry_count--;
2593 2594
		else
			oldusage = curusage;
2595 2596
	} while (retry_count);

2597 2598
	if (!ret && enlarge)
		memcg_oom_recover(memcg);
2599

2600 2601 2602
	return ret;
}

2603 2604 2605 2606 2607 2608 2609 2610 2611
unsigned long mem_cgroup_soft_limit_reclaim(struct zone *zone, int order,
					    gfp_t gfp_mask,
					    unsigned long *total_scanned)
{
	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;
2612
	unsigned long excess;
2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636
	unsigned long nr_scanned;

	if (order > 0)
		return 0;

	mctz = soft_limit_tree_node_zone(zone_to_nid(zone), zone_idx(zone));
	/*
	 * 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;

		nr_scanned = 0;
		reclaimed = mem_cgroup_soft_reclaim(mz->memcg, zone,
						    gfp_mask, &nr_scanned);
		nr_reclaimed += reclaimed;
		*total_scanned += nr_scanned;
2637
		spin_lock_irq(&mctz->lock);
2638
		__mem_cgroup_remove_exceeded(mz, mctz);
2639 2640 2641 2642 2643 2644

		/*
		 * If we failed to reclaim anything from this memory cgroup
		 * it is time to move on to the next cgroup
		 */
		next_mz = NULL;
2645 2646 2647
		if (!reclaimed)
			next_mz = __mem_cgroup_largest_soft_limit_node(mctz);

2648
		excess = soft_limit_excess(mz->memcg);
2649 2650 2651 2652 2653 2654 2655 2656 2657
		/*
		 * 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.
		 */
		/* If excess == 0, no tree ops */
2658
		__mem_cgroup_insert_exceeded(mz, mctz, excess);
2659
		spin_unlock_irq(&mctz->lock);
2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676
		css_put(&mz->memcg->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->memcg->css);
	return nr_reclaimed;
}

2677 2678 2679 2680 2681 2682
/*
 * Test whether @memcg has children, dead or alive.  Note that this
 * function doesn't care whether @memcg has use_hierarchy enabled and
 * returns %true if there are child csses according to the cgroup
 * hierarchy.  Testing use_hierarchy is the caller's responsiblity.
 */
2683 2684
static inline bool memcg_has_children(struct mem_cgroup *memcg)
{
2685 2686
	bool ret;

2687
	/*
2688 2689 2690 2691
	 * The lock does not prevent addition or deletion of children, but
	 * it prevents a new child from being initialized based on this
	 * parent in css_online(), so it's enough to decide whether
	 * hierarchically inherited attributes can still be changed or not.
2692
	 */
2693 2694 2695 2696 2697 2698
	lockdep_assert_held(&memcg_create_mutex);

	rcu_read_lock();
	ret = css_next_child(NULL, &memcg->css);
	rcu_read_unlock();
	return ret;
2699 2700
}

2701 2702 2703 2704 2705 2706 2707 2708 2709 2710
/*
 * Reclaims as many pages from the given memcg as possible and moves
 * the rest to the parent.
 *
 * Caller is responsible for holding css reference for memcg.
 */
static int mem_cgroup_force_empty(struct mem_cgroup *memcg)
{
	int nr_retries = MEM_CGROUP_RECLAIM_RETRIES;

2711 2712
	/* we call try-to-free pages for make this cgroup empty */
	lru_add_drain_all();
2713
	/* try to free all pages in this cgroup */
2714
	while (nr_retries && page_counter_read(&memcg->memory)) {
2715
		int progress;
2716

2717 2718 2719
		if (signal_pending(current))
			return -EINTR;

2720 2721
		progress = try_to_free_mem_cgroup_pages(memcg, 1,
							GFP_KERNEL, true);
2722
		if (!progress) {
2723
			nr_retries--;
2724
			/* maybe some writeback is necessary */
2725
			congestion_wait(BLK_RW_ASYNC, HZ/10);
2726
		}
2727 2728

	}
2729 2730

	return 0;
2731 2732
}

2733 2734 2735
static ssize_t mem_cgroup_force_empty_write(struct kernfs_open_file *of,
					    char *buf, size_t nbytes,
					    loff_t off)
2736
{
2737
	struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of));
2738

2739 2740
	if (mem_cgroup_is_root(memcg))
		return -EINVAL;
2741
	return mem_cgroup_force_empty(memcg) ?: nbytes;
2742 2743
}

2744 2745
static u64 mem_cgroup_hierarchy_read(struct cgroup_subsys_state *css,
				     struct cftype *cft)
2746
{
2747
	return mem_cgroup_from_css(css)->use_hierarchy;
2748 2749
}

2750 2751
static int mem_cgroup_hierarchy_write(struct cgroup_subsys_state *css,
				      struct cftype *cft, u64 val)
2752 2753
{
	int retval = 0;
2754
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
T
Tejun Heo 已提交
2755
	struct mem_cgroup *parent_memcg = mem_cgroup_from_css(memcg->css.parent);
2756

2757
	mutex_lock(&memcg_create_mutex);
2758 2759 2760 2761

	if (memcg->use_hierarchy == val)
		goto out;

2762
	/*
2763
	 * If parent's use_hierarchy is set, we can't make any modifications
2764 2765 2766 2767 2768 2769
	 * 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.
	 */
2770
	if ((!parent_memcg || !parent_memcg->use_hierarchy) &&
2771
				(val == 1 || val == 0)) {
2772
		if (!memcg_has_children(memcg))
2773
			memcg->use_hierarchy = val;
2774 2775 2776 2777
		else
			retval = -EBUSY;
	} else
		retval = -EINVAL;
2778 2779

out:
2780
	mutex_unlock(&memcg_create_mutex);
2781 2782 2783 2784

	return retval;
}

2785 2786
static unsigned long tree_stat(struct mem_cgroup *memcg,
			       enum mem_cgroup_stat_index idx)
2787 2788
{
	struct mem_cgroup *iter;
2789
	unsigned long val = 0;
2790 2791 2792 2793 2794 2795 2796

	for_each_mem_cgroup_tree(iter, memcg)
		val += mem_cgroup_read_stat(iter, idx);

	return val;
}

2797
static unsigned long mem_cgroup_usage(struct mem_cgroup *memcg, bool swap)
2798
{
2799
	unsigned long val;
2800

2801 2802 2803 2804 2805 2806
	if (mem_cgroup_is_root(memcg)) {
		val = tree_stat(memcg, MEM_CGROUP_STAT_CACHE);
		val += tree_stat(memcg, MEM_CGROUP_STAT_RSS);
		if (swap)
			val += tree_stat(memcg, MEM_CGROUP_STAT_SWAP);
	} else {
2807
		if (!swap)
2808
			val = page_counter_read(&memcg->memory);
2809
		else
2810
			val = page_counter_read(&memcg->memsw);
2811
	}
2812
	return val;
2813 2814
}

2815 2816 2817 2818 2819 2820 2821
enum {
	RES_USAGE,
	RES_LIMIT,
	RES_MAX_USAGE,
	RES_FAILCNT,
	RES_SOFT_LIMIT,
};
2822

2823
static u64 mem_cgroup_read_u64(struct cgroup_subsys_state *css,
2824
			       struct cftype *cft)
B
Balbir Singh 已提交
2825
{
2826
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
2827
	struct page_counter *counter;
2828

2829
	switch (MEMFILE_TYPE(cft->private)) {
2830
	case _MEM:
2831 2832
		counter = &memcg->memory;
		break;
2833
	case _MEMSWAP:
2834 2835
		counter = &memcg->memsw;
		break;
2836
	case _KMEM:
2837
		counter = &memcg->kmem;
2838
		break;
2839 2840 2841
	default:
		BUG();
	}
2842 2843 2844 2845

	switch (MEMFILE_ATTR(cft->private)) {
	case RES_USAGE:
		if (counter == &memcg->memory)
2846
			return (u64)mem_cgroup_usage(memcg, false) * PAGE_SIZE;
2847
		if (counter == &memcg->memsw)
2848
			return (u64)mem_cgroup_usage(memcg, true) * PAGE_SIZE;
2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860
		return (u64)page_counter_read(counter) * PAGE_SIZE;
	case RES_LIMIT:
		return (u64)counter->limit * PAGE_SIZE;
	case RES_MAX_USAGE:
		return (u64)counter->watermark * PAGE_SIZE;
	case RES_FAILCNT:
		return counter->failcnt;
	case RES_SOFT_LIMIT:
		return (u64)memcg->soft_limit * PAGE_SIZE;
	default:
		BUG();
	}
B
Balbir Singh 已提交
2861
}
2862

2863
#ifndef CONFIG_SLOB
2864
static int memcg_online_kmem(struct mem_cgroup *memcg)
2865 2866 2867 2868
{
	int err = 0;
	int memcg_id;

2869
	BUG_ON(memcg->kmemcg_id >= 0);
2870
	BUG_ON(memcg->kmem_state);
2871

2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883
	/*
	 * For simplicity, we won't allow this to be disabled.  It also can't
	 * be changed if the cgroup has children already, or if tasks had
	 * already joined.
	 *
	 * If tasks join before we set the limit, a person looking at
	 * kmem.usage_in_bytes will have no way to determine when it took
	 * place, which makes the value quite meaningless.
	 *
	 * After it first became limited, changes in the value of the limit are
	 * of course permitted.
	 */
2884
	mutex_lock(&memcg_create_mutex);
2885
	if (cgroup_is_populated(memcg->css.cgroup) ||
2886
	    (memcg->use_hierarchy && memcg_has_children(memcg)))
2887 2888 2889 2890
		err = -EBUSY;
	mutex_unlock(&memcg_create_mutex);
	if (err)
		goto out;
2891

2892
	memcg_id = memcg_alloc_cache_id();
2893 2894 2895 2896 2897
	if (memcg_id < 0) {
		err = memcg_id;
		goto out;
	}

2898
	static_branch_inc(&memcg_kmem_enabled_key);
2899
	/*
2900
	 * A memory cgroup is considered kmem-online as soon as it gets
V
Vladimir Davydov 已提交
2901
	 * kmemcg_id. Setting the id after enabling static branching will
2902 2903 2904
	 * guarantee no one starts accounting before all call sites are
	 * patched.
	 */
V
Vladimir Davydov 已提交
2905
	memcg->kmemcg_id = memcg_id;
2906
	memcg->kmem_state = KMEM_ONLINE;
2907
out:
2908 2909 2910
	return err;
}

2911
static int memcg_propagate_kmem(struct mem_cgroup *memcg)
2912
{
2913
	int ret = 0;
2914
	struct mem_cgroup *parent = parent_mem_cgroup(memcg);
2915

2916 2917
	if (!parent)
		return 0;
2918

2919
	mutex_lock(&memcg_limit_mutex);
2920
	/*
2921 2922 2923
	 * If the parent cgroup is not kmem-online now, it cannot be
	 * onlined after this point, because it has at least one child
	 * already.
2924
	 */
2925 2926
	if (memcg_kmem_online(parent))
		ret = memcg_online_kmem(memcg);
2927
	mutex_unlock(&memcg_limit_mutex);
2928
	return ret;
2929
}
2930 2931 2932 2933 2934 2935 2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983

static void memcg_offline_kmem(struct mem_cgroup *memcg)
{
	struct cgroup_subsys_state *css;
	struct mem_cgroup *parent, *child;
	int kmemcg_id;

	if (memcg->kmem_state != KMEM_ONLINE)
		return;
	/*
	 * Clear the online state before clearing memcg_caches array
	 * entries. The slab_mutex in memcg_deactivate_kmem_caches()
	 * guarantees that no cache will be created for this cgroup
	 * after we are done (see memcg_create_kmem_cache()).
	 */
	memcg->kmem_state = KMEM_ALLOCATED;

	memcg_deactivate_kmem_caches(memcg);

	kmemcg_id = memcg->kmemcg_id;
	BUG_ON(kmemcg_id < 0);

	parent = parent_mem_cgroup(memcg);
	if (!parent)
		parent = root_mem_cgroup;

	/*
	 * Change kmemcg_id of this cgroup and all its descendants to the
	 * parent's id, and then move all entries from this cgroup's list_lrus
	 * to ones of the parent. After we have finished, all list_lrus
	 * corresponding to this cgroup are guaranteed to remain empty. The
	 * ordering is imposed by list_lru_node->lock taken by
	 * memcg_drain_all_list_lrus().
	 */
	css_for_each_descendant_pre(css, &memcg->css) {
		child = mem_cgroup_from_css(css);
		BUG_ON(child->kmemcg_id != kmemcg_id);
		child->kmemcg_id = parent->kmemcg_id;
		if (!memcg->use_hierarchy)
			break;
	}
	memcg_drain_all_list_lrus(kmemcg_id, parent->kmemcg_id);

	memcg_free_cache_id(kmemcg_id);
}

static void memcg_free_kmem(struct mem_cgroup *memcg)
{
	if (memcg->kmem_state == KMEM_ALLOCATED) {
		memcg_destroy_kmem_caches(memcg);
		static_branch_dec(&memcg_kmem_enabled_key);
		WARN_ON(page_counter_read(&memcg->kmem));
	}
}
2984
#else
2985 2986 2987 2988 2989 2990 2991 2992 2993 2994 2995 2996 2997
static int memcg_propagate_kmem(struct mem_cgroup *memcg)
{
	return 0;
}
static void memcg_offline_kmem(struct mem_cgroup *memcg)
{
}
static void memcg_free_kmem(struct mem_cgroup *memcg)
{
}
#endif /* !CONFIG_SLOB */

#ifdef CONFIG_MEMCG_KMEM
2998
static int memcg_update_kmem_limit(struct mem_cgroup *memcg,
2999
				   unsigned long limit)
3000
{
3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013
	int ret;

	mutex_lock(&memcg_limit_mutex);
	/* Top-level cgroup doesn't propagate from root */
	if (!memcg_kmem_online(memcg)) {
		ret = memcg_online_kmem(memcg);
		if (ret)
			goto out;
	}
	ret = page_counter_limit(&memcg->kmem, limit);
out:
	mutex_unlock(&memcg_limit_mutex);
	return ret;
3014
}
3015 3016 3017
#else
static int memcg_update_kmem_limit(struct mem_cgroup *memcg,
				   unsigned long limit)
3018
{
3019
	return -EINVAL;
3020
}
3021
#endif /* CONFIG_MEMCG_KMEM */
3022

3023

3024 3025 3026 3027
/*
 * The user of this function is...
 * RES_LIMIT.
 */
3028 3029
static ssize_t mem_cgroup_write(struct kernfs_open_file *of,
				char *buf, size_t nbytes, loff_t off)
B
Balbir Singh 已提交
3030
{
3031
	struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of));
3032
	unsigned long nr_pages;
3033 3034
	int ret;

3035
	buf = strstrip(buf);
3036
	ret = page_counter_memparse(buf, "-1", &nr_pages);
3037 3038
	if (ret)
		return ret;
3039

3040
	switch (MEMFILE_ATTR(of_cft(of)->private)) {
3041
	case RES_LIMIT:
3042 3043 3044 3045
		if (mem_cgroup_is_root(memcg)) { /* Can't set limit on root */
			ret = -EINVAL;
			break;
		}
3046 3047 3048
		switch (MEMFILE_TYPE(of_cft(of)->private)) {
		case _MEM:
			ret = mem_cgroup_resize_limit(memcg, nr_pages);
3049
			break;
3050 3051
		case _MEMSWAP:
			ret = mem_cgroup_resize_memsw_limit(memcg, nr_pages);
3052
			break;
3053 3054 3055 3056
		case _KMEM:
			ret = memcg_update_kmem_limit(memcg, nr_pages);
			break;
		}
3057
		break;
3058 3059 3060
	case RES_SOFT_LIMIT:
		memcg->soft_limit = nr_pages;
		ret = 0;
3061 3062
		break;
	}
3063
	return ret ?: nbytes;
B
Balbir Singh 已提交
3064 3065
}

3066 3067
static ssize_t mem_cgroup_reset(struct kernfs_open_file *of, char *buf,
				size_t nbytes, loff_t off)
3068
{
3069
	struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of));
3070
	struct page_counter *counter;
3071

3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084
	switch (MEMFILE_TYPE(of_cft(of)->private)) {
	case _MEM:
		counter = &memcg->memory;
		break;
	case _MEMSWAP:
		counter = &memcg->memsw;
		break;
	case _KMEM:
		counter = &memcg->kmem;
		break;
	default:
		BUG();
	}
3085

3086
	switch (MEMFILE_ATTR(of_cft(of)->private)) {
3087
	case RES_MAX_USAGE:
3088
		page_counter_reset_watermark(counter);
3089 3090
		break;
	case RES_FAILCNT:
3091
		counter->failcnt = 0;
3092
		break;
3093 3094
	default:
		BUG();
3095
	}
3096

3097
	return nbytes;
3098 3099
}

3100
static u64 mem_cgroup_move_charge_read(struct cgroup_subsys_state *css,
3101 3102
					struct cftype *cft)
{
3103
	return mem_cgroup_from_css(css)->move_charge_at_immigrate;
3104 3105
}

3106
#ifdef CONFIG_MMU
3107
static int mem_cgroup_move_charge_write(struct cgroup_subsys_state *css,
3108 3109
					struct cftype *cft, u64 val)
{
3110
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
3111

3112
	if (val & ~MOVE_MASK)
3113
		return -EINVAL;
3114

3115
	/*
3116 3117 3118 3119
	 * No kind of locking is needed in here, because ->can_attach() will
	 * check this value once in the beginning of the process, and then carry
	 * on with stale data. This means that changes to this value will only
	 * affect task migrations starting after the change.
3120
	 */
3121
	memcg->move_charge_at_immigrate = val;
3122 3123
	return 0;
}
3124
#else
3125
static int mem_cgroup_move_charge_write(struct cgroup_subsys_state *css,
3126 3127 3128 3129 3130
					struct cftype *cft, u64 val)
{
	return -ENOSYS;
}
#endif
3131

3132
#ifdef CONFIG_NUMA
3133
static int memcg_numa_stat_show(struct seq_file *m, void *v)
3134
{
3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146
	struct numa_stat {
		const char *name;
		unsigned int lru_mask;
	};

	static const struct numa_stat stats[] = {
		{ "total", LRU_ALL },
		{ "file", LRU_ALL_FILE },
		{ "anon", LRU_ALL_ANON },
		{ "unevictable", BIT(LRU_UNEVICTABLE) },
	};
	const struct numa_stat *stat;
3147
	int nid;
3148
	unsigned long nr;
3149
	struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(m));
3150

3151 3152 3153 3154 3155 3156 3157 3158 3159
	for (stat = stats; stat < stats + ARRAY_SIZE(stats); stat++) {
		nr = mem_cgroup_nr_lru_pages(memcg, stat->lru_mask);
		seq_printf(m, "%s=%lu", stat->name, nr);
		for_each_node_state(nid, N_MEMORY) {
			nr = mem_cgroup_node_nr_lru_pages(memcg, nid,
							  stat->lru_mask);
			seq_printf(m, " N%d=%lu", nid, nr);
		}
		seq_putc(m, '\n');
3160 3161
	}

3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176
	for (stat = stats; stat < stats + ARRAY_SIZE(stats); stat++) {
		struct mem_cgroup *iter;

		nr = 0;
		for_each_mem_cgroup_tree(iter, memcg)
			nr += mem_cgroup_nr_lru_pages(iter, stat->lru_mask);
		seq_printf(m, "hierarchical_%s=%lu", stat->name, nr);
		for_each_node_state(nid, N_MEMORY) {
			nr = 0;
			for_each_mem_cgroup_tree(iter, memcg)
				nr += mem_cgroup_node_nr_lru_pages(
					iter, nid, stat->lru_mask);
			seq_printf(m, " N%d=%lu", nid, nr);
		}
		seq_putc(m, '\n');
3177 3178 3179 3180 3181 3182
	}

	return 0;
}
#endif /* CONFIG_NUMA */

3183
static int memcg_stat_show(struct seq_file *m, void *v)
3184
{
3185
	struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(m));
3186
	unsigned long memory, memsw;
3187 3188
	struct mem_cgroup *mi;
	unsigned int i;
3189

3190 3191 3192 3193
	BUILD_BUG_ON(ARRAY_SIZE(mem_cgroup_stat_names) !=
		     MEM_CGROUP_STAT_NSTATS);
	BUILD_BUG_ON(ARRAY_SIZE(mem_cgroup_events_names) !=
		     MEM_CGROUP_EVENTS_NSTATS);
3194 3195
	BUILD_BUG_ON(ARRAY_SIZE(mem_cgroup_lru_names) != NR_LRU_LISTS);

3196
	for (i = 0; i < MEM_CGROUP_STAT_NSTATS; i++) {
3197
		if (i == MEM_CGROUP_STAT_SWAP && !do_memsw_account())
3198
			continue;
3199
		seq_printf(m, "%s %lu\n", mem_cgroup_stat_names[i],
3200
			   mem_cgroup_read_stat(memcg, i) * PAGE_SIZE);
3201
	}
L
Lee Schermerhorn 已提交
3202

3203 3204 3205 3206 3207 3208 3209 3210
	for (i = 0; i < MEM_CGROUP_EVENTS_NSTATS; i++)
		seq_printf(m, "%s %lu\n", mem_cgroup_events_names[i],
			   mem_cgroup_read_events(memcg, i));

	for (i = 0; i < NR_LRU_LISTS; i++)
		seq_printf(m, "%s %lu\n", mem_cgroup_lru_names[i],
			   mem_cgroup_nr_lru_pages(memcg, BIT(i)) * PAGE_SIZE);

K
KAMEZAWA Hiroyuki 已提交
3211
	/* Hierarchical information */
3212 3213 3214 3215
	memory = memsw = PAGE_COUNTER_MAX;
	for (mi = memcg; mi; mi = parent_mem_cgroup(mi)) {
		memory = min(memory, mi->memory.limit);
		memsw = min(memsw, mi->memsw.limit);
3216
	}
3217 3218
	seq_printf(m, "hierarchical_memory_limit %llu\n",
		   (u64)memory * PAGE_SIZE);
3219
	if (do_memsw_account())
3220 3221
		seq_printf(m, "hierarchical_memsw_limit %llu\n",
			   (u64)memsw * PAGE_SIZE);
K
KOSAKI Motohiro 已提交
3222

3223
	for (i = 0; i < MEM_CGROUP_STAT_NSTATS; i++) {
3224
		unsigned long long val = 0;
3225

3226
		if (i == MEM_CGROUP_STAT_SWAP && !do_memsw_account())
3227
			continue;
3228 3229
		for_each_mem_cgroup_tree(mi, memcg)
			val += mem_cgroup_read_stat(mi, i) * PAGE_SIZE;
3230
		seq_printf(m, "total_%s %llu\n", mem_cgroup_stat_names[i], val);
3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243 3244 3245 3246 3247
	}

	for (i = 0; i < MEM_CGROUP_EVENTS_NSTATS; i++) {
		unsigned long long val = 0;

		for_each_mem_cgroup_tree(mi, memcg)
			val += mem_cgroup_read_events(mi, i);
		seq_printf(m, "total_%s %llu\n",
			   mem_cgroup_events_names[i], val);
	}

	for (i = 0; i < NR_LRU_LISTS; i++) {
		unsigned long long val = 0;

		for_each_mem_cgroup_tree(mi, memcg)
			val += mem_cgroup_nr_lru_pages(mi, BIT(i)) * PAGE_SIZE;
		seq_printf(m, "total_%s %llu\n", mem_cgroup_lru_names[i], val);
3248
	}
K
KAMEZAWA Hiroyuki 已提交
3249

K
KOSAKI Motohiro 已提交
3250 3251 3252 3253
#ifdef CONFIG_DEBUG_VM
	{
		int nid, zid;
		struct mem_cgroup_per_zone *mz;
3254
		struct zone_reclaim_stat *rstat;
K
KOSAKI Motohiro 已提交
3255 3256 3257 3258 3259
		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++) {
3260
				mz = &memcg->nodeinfo[nid]->zoneinfo[zid];
3261
				rstat = &mz->lruvec.reclaim_stat;
K
KOSAKI Motohiro 已提交
3262

3263 3264 3265 3266
				recent_rotated[0] += rstat->recent_rotated[0];
				recent_rotated[1] += rstat->recent_rotated[1];
				recent_scanned[0] += rstat->recent_scanned[0];
				recent_scanned[1] += rstat->recent_scanned[1];
K
KOSAKI Motohiro 已提交
3267
			}
3268 3269 3270 3271
		seq_printf(m, "recent_rotated_anon %lu\n", recent_rotated[0]);
		seq_printf(m, "recent_rotated_file %lu\n", recent_rotated[1]);
		seq_printf(m, "recent_scanned_anon %lu\n", recent_scanned[0]);
		seq_printf(m, "recent_scanned_file %lu\n", recent_scanned[1]);
K
KOSAKI Motohiro 已提交
3272 3273 3274
	}
#endif

3275 3276 3277
	return 0;
}

3278 3279
static u64 mem_cgroup_swappiness_read(struct cgroup_subsys_state *css,
				      struct cftype *cft)
K
KOSAKI Motohiro 已提交
3280
{
3281
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
K
KOSAKI Motohiro 已提交
3282

3283
	return mem_cgroup_swappiness(memcg);
K
KOSAKI Motohiro 已提交
3284 3285
}

3286 3287
static int mem_cgroup_swappiness_write(struct cgroup_subsys_state *css,
				       struct cftype *cft, u64 val)
K
KOSAKI Motohiro 已提交
3288
{
3289
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
K
KOSAKI Motohiro 已提交
3290

3291
	if (val > 100)
K
KOSAKI Motohiro 已提交
3292 3293
		return -EINVAL;

3294
	if (css->parent)
3295 3296 3297
		memcg->swappiness = val;
	else
		vm_swappiness = val;
3298

K
KOSAKI Motohiro 已提交
3299 3300 3301
	return 0;
}

3302 3303 3304
static void __mem_cgroup_threshold(struct mem_cgroup *memcg, bool swap)
{
	struct mem_cgroup_threshold_ary *t;
3305
	unsigned long usage;
3306 3307 3308 3309
	int i;

	rcu_read_lock();
	if (!swap)
3310
		t = rcu_dereference(memcg->thresholds.primary);
3311
	else
3312
		t = rcu_dereference(memcg->memsw_thresholds.primary);
3313 3314 3315 3316

	if (!t)
		goto unlock;

3317
	usage = mem_cgroup_usage(memcg, swap);
3318 3319

	/*
3320
	 * current_threshold points to threshold just below or equal to usage.
3321 3322 3323
	 * If it's not true, a threshold was crossed after last
	 * call of __mem_cgroup_threshold().
	 */
3324
	i = t->current_threshold;
3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347

	/*
	 * 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 */
3348
	t->current_threshold = i - 1;
3349 3350 3351 3352 3353 3354
unlock:
	rcu_read_unlock();
}

static void mem_cgroup_threshold(struct mem_cgroup *memcg)
{
3355 3356
	while (memcg) {
		__mem_cgroup_threshold(memcg, false);
3357
		if (do_memsw_account())
3358 3359 3360 3361
			__mem_cgroup_threshold(memcg, true);

		memcg = parent_mem_cgroup(memcg);
	}
3362 3363 3364 3365 3366 3367 3368
}

static int compare_thresholds(const void *a, const void *b)
{
	const struct mem_cgroup_threshold *_a = a;
	const struct mem_cgroup_threshold *_b = b;

3369 3370 3371 3372 3373 3374 3375
	if (_a->threshold > _b->threshold)
		return 1;

	if (_a->threshold < _b->threshold)
		return -1;

	return 0;
3376 3377
}

3378
static int mem_cgroup_oom_notify_cb(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
3379 3380 3381
{
	struct mem_cgroup_eventfd_list *ev;

3382 3383
	spin_lock(&memcg_oom_lock);

3384
	list_for_each_entry(ev, &memcg->oom_notify, list)
K
KAMEZAWA Hiroyuki 已提交
3385
		eventfd_signal(ev->eventfd, 1);
3386 3387

	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
3388 3389 3390
	return 0;
}

3391
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
3392
{
K
KAMEZAWA Hiroyuki 已提交
3393 3394
	struct mem_cgroup *iter;

3395
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
3396
		mem_cgroup_oom_notify_cb(iter);
K
KAMEZAWA Hiroyuki 已提交
3397 3398
}

3399
static int __mem_cgroup_usage_register_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
3400
	struct eventfd_ctx *eventfd, const char *args, enum res_type type)
3401
{
3402 3403
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
3404 3405
	unsigned long threshold;
	unsigned long usage;
3406
	int i, size, ret;
3407

3408
	ret = page_counter_memparse(args, "-1", &threshold);
3409 3410 3411 3412
	if (ret)
		return ret;

	mutex_lock(&memcg->thresholds_lock);
3413

3414
	if (type == _MEM) {
3415
		thresholds = &memcg->thresholds;
3416
		usage = mem_cgroup_usage(memcg, false);
3417
	} else if (type == _MEMSWAP) {
3418
		thresholds = &memcg->memsw_thresholds;
3419
		usage = mem_cgroup_usage(memcg, true);
3420
	} else
3421 3422 3423
		BUG();

	/* Check if a threshold crossed before adding a new one */
3424
	if (thresholds->primary)
3425 3426
		__mem_cgroup_threshold(memcg, type == _MEMSWAP);

3427
	size = thresholds->primary ? thresholds->primary->size + 1 : 1;
3428 3429

	/* Allocate memory for new array of thresholds */
3430
	new = kmalloc(sizeof(*new) + size * sizeof(struct mem_cgroup_threshold),
3431
			GFP_KERNEL);
3432
	if (!new) {
3433 3434 3435
		ret = -ENOMEM;
		goto unlock;
	}
3436
	new->size = size;
3437 3438

	/* Copy thresholds (if any) to new array */
3439 3440
	if (thresholds->primary) {
		memcpy(new->entries, thresholds->primary->entries, (size - 1) *
3441
				sizeof(struct mem_cgroup_threshold));
3442 3443
	}

3444
	/* Add new threshold */
3445 3446
	new->entries[size - 1].eventfd = eventfd;
	new->entries[size - 1].threshold = threshold;
3447 3448

	/* Sort thresholds. Registering of new threshold isn't time-critical */
3449
	sort(new->entries, size, sizeof(struct mem_cgroup_threshold),
3450 3451 3452
			compare_thresholds, NULL);

	/* Find current threshold */
3453
	new->current_threshold = -1;
3454
	for (i = 0; i < size; i++) {
3455
		if (new->entries[i].threshold <= usage) {
3456
			/*
3457 3458
			 * new->current_threshold will not be used until
			 * rcu_assign_pointer(), so it's safe to increment
3459 3460
			 * it here.
			 */
3461
			++new->current_threshold;
3462 3463
		} else
			break;
3464 3465
	}

3466 3467 3468 3469 3470
	/* Free old spare buffer and save old primary buffer as spare */
	kfree(thresholds->spare);
	thresholds->spare = thresholds->primary;

	rcu_assign_pointer(thresholds->primary, new);
3471

3472
	/* To be sure that nobody uses thresholds */
3473 3474 3475 3476 3477 3478 3479 3480
	synchronize_rcu();

unlock:
	mutex_unlock(&memcg->thresholds_lock);

	return ret;
}

3481
static int mem_cgroup_usage_register_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
3482 3483
	struct eventfd_ctx *eventfd, const char *args)
{
3484
	return __mem_cgroup_usage_register_event(memcg, eventfd, args, _MEM);
T
Tejun Heo 已提交
3485 3486
}

3487
static int memsw_cgroup_usage_register_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
3488 3489
	struct eventfd_ctx *eventfd, const char *args)
{
3490
	return __mem_cgroup_usage_register_event(memcg, eventfd, args, _MEMSWAP);
T
Tejun Heo 已提交
3491 3492
}

3493
static void __mem_cgroup_usage_unregister_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
3494
	struct eventfd_ctx *eventfd, enum res_type type)
3495
{
3496 3497
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
3498
	unsigned long usage;
3499
	int i, j, size;
3500 3501

	mutex_lock(&memcg->thresholds_lock);
3502 3503

	if (type == _MEM) {
3504
		thresholds = &memcg->thresholds;
3505
		usage = mem_cgroup_usage(memcg, false);
3506
	} else if (type == _MEMSWAP) {
3507
		thresholds = &memcg->memsw_thresholds;
3508
		usage = mem_cgroup_usage(memcg, true);
3509
	} else
3510 3511
		BUG();

3512 3513 3514
	if (!thresholds->primary)
		goto unlock;

3515 3516 3517 3518
	/* Check if a threshold crossed before removing */
	__mem_cgroup_threshold(memcg, type == _MEMSWAP);

	/* Calculate new number of threshold */
3519 3520 3521
	size = 0;
	for (i = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd != eventfd)
3522 3523 3524
			size++;
	}

3525
	new = thresholds->spare;
3526

3527 3528
	/* Set thresholds array to NULL if we don't have thresholds */
	if (!size) {
3529 3530
		kfree(new);
		new = NULL;
3531
		goto swap_buffers;
3532 3533
	}

3534
	new->size = size;
3535 3536

	/* Copy thresholds and find current threshold */
3537 3538 3539
	new->current_threshold = -1;
	for (i = 0, j = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd == eventfd)
3540 3541
			continue;

3542
		new->entries[j] = thresholds->primary->entries[i];
3543
		if (new->entries[j].threshold <= usage) {
3544
			/*
3545
			 * new->current_threshold will not be used
3546 3547 3548
			 * until rcu_assign_pointer(), so it's safe to increment
			 * it here.
			 */
3549
			++new->current_threshold;
3550 3551 3552 3553
		}
		j++;
	}

3554
swap_buffers:
3555 3556
	/* Swap primary and spare array */
	thresholds->spare = thresholds->primary;
3557

3558
	rcu_assign_pointer(thresholds->primary, new);
3559

3560
	/* To be sure that nobody uses thresholds */
3561
	synchronize_rcu();
3562 3563 3564 3565 3566 3567

	/* If all events are unregistered, free the spare array */
	if (!new) {
		kfree(thresholds->spare);
		thresholds->spare = NULL;
	}
3568
unlock:
3569 3570
	mutex_unlock(&memcg->thresholds_lock);
}
3571

3572
static void mem_cgroup_usage_unregister_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
3573 3574
	struct eventfd_ctx *eventfd)
{
3575
	return __mem_cgroup_usage_unregister_event(memcg, eventfd, _MEM);
T
Tejun Heo 已提交
3576 3577
}

3578
static void memsw_cgroup_usage_unregister_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
3579 3580
	struct eventfd_ctx *eventfd)
{
3581
	return __mem_cgroup_usage_unregister_event(memcg, eventfd, _MEMSWAP);
T
Tejun Heo 已提交
3582 3583
}

3584
static int mem_cgroup_oom_register_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
3585
	struct eventfd_ctx *eventfd, const char *args)
K
KAMEZAWA Hiroyuki 已提交
3586 3587 3588 3589 3590 3591 3592
{
	struct mem_cgroup_eventfd_list *event;

	event = kmalloc(sizeof(*event),	GFP_KERNEL);
	if (!event)
		return -ENOMEM;

3593
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
3594 3595 3596 3597 3598

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

	/* already in OOM ? */
3599
	if (memcg->under_oom)
K
KAMEZAWA Hiroyuki 已提交
3600
		eventfd_signal(eventfd, 1);
3601
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
3602 3603 3604 3605

	return 0;
}

3606
static void mem_cgroup_oom_unregister_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
3607
	struct eventfd_ctx *eventfd)
K
KAMEZAWA Hiroyuki 已提交
3608 3609 3610
{
	struct mem_cgroup_eventfd_list *ev, *tmp;

3611
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
3612

3613
	list_for_each_entry_safe(ev, tmp, &memcg->oom_notify, list) {
K
KAMEZAWA Hiroyuki 已提交
3614 3615 3616 3617 3618 3619
		if (ev->eventfd == eventfd) {
			list_del(&ev->list);
			kfree(ev);
		}
	}

3620
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
3621 3622
}

3623
static int mem_cgroup_oom_control_read(struct seq_file *sf, void *v)
3624
{
3625
	struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(sf));
3626

3627
	seq_printf(sf, "oom_kill_disable %d\n", memcg->oom_kill_disable);
3628
	seq_printf(sf, "under_oom %d\n", (bool)memcg->under_oom);
3629 3630 3631
	return 0;
}

3632
static int mem_cgroup_oom_control_write(struct cgroup_subsys_state *css,
3633 3634
	struct cftype *cft, u64 val)
{
3635
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
3636 3637

	/* cannot set to root cgroup and only 0 and 1 are allowed */
3638
	if (!css->parent || !((val == 0) || (val == 1)))
3639 3640
		return -EINVAL;

3641
	memcg->oom_kill_disable = val;
3642
	if (!val)
3643
		memcg_oom_recover(memcg);
3644

3645 3646 3647
	return 0;
}

3648 3649 3650 3651 3652 3653 3654
#ifdef CONFIG_CGROUP_WRITEBACK

struct list_head *mem_cgroup_cgwb_list(struct mem_cgroup *memcg)
{
	return &memcg->cgwb_list;
}

T
Tejun Heo 已提交
3655 3656 3657 3658 3659 3660 3661 3662 3663 3664
static int memcg_wb_domain_init(struct mem_cgroup *memcg, gfp_t gfp)
{
	return wb_domain_init(&memcg->cgwb_domain, gfp);
}

static void memcg_wb_domain_exit(struct mem_cgroup *memcg)
{
	wb_domain_exit(&memcg->cgwb_domain);
}

3665 3666 3667 3668 3669
static void memcg_wb_domain_size_changed(struct mem_cgroup *memcg)
{
	wb_domain_size_changed(&memcg->cgwb_domain);
}

T
Tejun Heo 已提交
3670 3671 3672 3673 3674 3675 3676 3677 3678 3679
struct wb_domain *mem_cgroup_wb_domain(struct bdi_writeback *wb)
{
	struct mem_cgroup *memcg = mem_cgroup_from_css(wb->memcg_css);

	if (!memcg->css.parent)
		return NULL;

	return &memcg->cgwb_domain;
}

3680 3681 3682
/**
 * mem_cgroup_wb_stats - retrieve writeback related stats from its memcg
 * @wb: bdi_writeback in question
3683 3684
 * @pfilepages: out parameter for number of file pages
 * @pheadroom: out parameter for number of allocatable pages according to memcg
3685 3686 3687
 * @pdirty: out parameter for number of dirty pages
 * @pwriteback: out parameter for number of pages under writeback
 *
3688 3689 3690
 * Determine the numbers of file, headroom, dirty, and writeback pages in
 * @wb's memcg.  File, dirty and writeback are self-explanatory.  Headroom
 * is a bit more involved.
3691
 *
3692 3693 3694 3695 3696
 * A memcg's headroom is "min(max, high) - used".  In the hierarchy, the
 * headroom is calculated as the lowest headroom of itself and the
 * ancestors.  Note that this doesn't consider the actual amount of
 * available memory in the system.  The caller should further cap
 * *@pheadroom accordingly.
3697
 */
3698 3699 3700
void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
			 unsigned long *pheadroom, unsigned long *pdirty,
			 unsigned long *pwriteback)
3701 3702 3703 3704 3705 3706 3707 3708
{
	struct mem_cgroup *memcg = mem_cgroup_from_css(wb->memcg_css);
	struct mem_cgroup *parent;

	*pdirty = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_DIRTY);

	/* this should eventually include NR_UNSTABLE_NFS */
	*pwriteback = mem_cgroup_read_stat(memcg, MEM_CGROUP_STAT_WRITEBACK);
3709 3710 3711
	*pfilepages = mem_cgroup_nr_lru_pages(memcg, (1 << LRU_INACTIVE_FILE) |
						     (1 << LRU_ACTIVE_FILE));
	*pheadroom = PAGE_COUNTER_MAX;
3712 3713 3714 3715 3716

	while ((parent = parent_mem_cgroup(memcg))) {
		unsigned long ceiling = min(memcg->memory.limit, memcg->high);
		unsigned long used = page_counter_read(&memcg->memory);

3717
		*pheadroom = min(*pheadroom, ceiling - min(ceiling, used));
3718 3719 3720 3721
		memcg = parent;
	}
}

T
Tejun Heo 已提交
3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732
#else	/* CONFIG_CGROUP_WRITEBACK */

static int memcg_wb_domain_init(struct mem_cgroup *memcg, gfp_t gfp)
{
	return 0;
}

static void memcg_wb_domain_exit(struct mem_cgroup *memcg)
{
}

3733 3734 3735 3736
static void memcg_wb_domain_size_changed(struct mem_cgroup *memcg)
{
}

3737 3738
#endif	/* CONFIG_CGROUP_WRITEBACK */

3739 3740 3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751
/*
 * DO NOT USE IN NEW FILES.
 *
 * "cgroup.event_control" implementation.
 *
 * This is way over-engineered.  It tries to support fully configurable
 * events for each user.  Such level of flexibility is completely
 * unnecessary especially in the light of the planned unified hierarchy.
 *
 * Please deprecate this and replace with something simpler if at all
 * possible.
 */

3752 3753 3754 3755 3756
/*
 * Unregister event and free resources.
 *
 * Gets called from workqueue.
 */
3757
static void memcg_event_remove(struct work_struct *work)
3758
{
3759 3760
	struct mem_cgroup_event *event =
		container_of(work, struct mem_cgroup_event, remove);
3761
	struct mem_cgroup *memcg = event->memcg;
3762 3763 3764

	remove_wait_queue(event->wqh, &event->wait);

3765
	event->unregister_event(memcg, event->eventfd);
3766 3767 3768 3769 3770 3771

	/* Notify userspace the event is going away. */
	eventfd_signal(event->eventfd, 1);

	eventfd_ctx_put(event->eventfd);
	kfree(event);
3772
	css_put(&memcg->css);
3773 3774 3775 3776 3777 3778 3779
}

/*
 * Gets called on POLLHUP on eventfd when user closes it.
 *
 * Called with wqh->lock held and interrupts disabled.
 */
3780 3781
static int memcg_event_wake(wait_queue_t *wait, unsigned mode,
			    int sync, void *key)
3782
{
3783 3784
	struct mem_cgroup_event *event =
		container_of(wait, struct mem_cgroup_event, wait);
3785
	struct mem_cgroup *memcg = event->memcg;
3786 3787 3788 3789 3790 3791 3792 3793 3794 3795 3796 3797
	unsigned long flags = (unsigned long)key;

	if (flags & POLLHUP) {
		/*
		 * If the event has been detached at cgroup removal, we
		 * can simply return knowing the other side will cleanup
		 * for us.
		 *
		 * We can't race against event freeing since the other
		 * side will require wqh->lock via remove_wait_queue(),
		 * which we hold.
		 */
3798
		spin_lock(&memcg->event_list_lock);
3799 3800 3801 3802 3803 3804 3805 3806
		if (!list_empty(&event->list)) {
			list_del_init(&event->list);
			/*
			 * We are in atomic context, but cgroup_event_remove()
			 * may sleep, so we have to call it in workqueue.
			 */
			schedule_work(&event->remove);
		}
3807
		spin_unlock(&memcg->event_list_lock);
3808 3809 3810 3811 3812
	}

	return 0;
}

3813
static void memcg_event_ptable_queue_proc(struct file *file,
3814 3815
		wait_queue_head_t *wqh, poll_table *pt)
{
3816 3817
	struct mem_cgroup_event *event =
		container_of(pt, struct mem_cgroup_event, pt);
3818 3819 3820 3821 3822 3823

	event->wqh = wqh;
	add_wait_queue(wqh, &event->wait);
}

/*
3824 3825
 * DO NOT USE IN NEW FILES.
 *
3826 3827 3828 3829 3830
 * Parse input and register new cgroup event handler.
 *
 * Input must be in format '<event_fd> <control_fd> <args>'.
 * Interpretation of args is defined by control file implementation.
 */
3831 3832
static ssize_t memcg_write_event_control(struct kernfs_open_file *of,
					 char *buf, size_t nbytes, loff_t off)
3833
{
3834
	struct cgroup_subsys_state *css = of_css(of);
3835
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
3836
	struct mem_cgroup_event *event;
3837 3838 3839 3840
	struct cgroup_subsys_state *cfile_css;
	unsigned int efd, cfd;
	struct fd efile;
	struct fd cfile;
3841
	const char *name;
3842 3843 3844
	char *endp;
	int ret;

3845 3846 3847
	buf = strstrip(buf);

	efd = simple_strtoul(buf, &endp, 10);
3848 3849
	if (*endp != ' ')
		return -EINVAL;
3850
	buf = endp + 1;
3851

3852
	cfd = simple_strtoul(buf, &endp, 10);
3853 3854
	if ((*endp != ' ') && (*endp != '\0'))
		return -EINVAL;
3855
	buf = endp + 1;
3856 3857 3858 3859 3860

	event = kzalloc(sizeof(*event), GFP_KERNEL);
	if (!event)
		return -ENOMEM;

3861
	event->memcg = memcg;
3862
	INIT_LIST_HEAD(&event->list);
3863 3864 3865
	init_poll_funcptr(&event->pt, memcg_event_ptable_queue_proc);
	init_waitqueue_func_entry(&event->wait, memcg_event_wake);
	INIT_WORK(&event->remove, memcg_event_remove);
3866 3867 3868 3869 3870 3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890

	efile = fdget(efd);
	if (!efile.file) {
		ret = -EBADF;
		goto out_kfree;
	}

	event->eventfd = eventfd_ctx_fileget(efile.file);
	if (IS_ERR(event->eventfd)) {
		ret = PTR_ERR(event->eventfd);
		goto out_put_efile;
	}

	cfile = fdget(cfd);
	if (!cfile.file) {
		ret = -EBADF;
		goto out_put_eventfd;
	}

	/* the process need read permission on control file */
	/* AV: shouldn't we check that it's been opened for read instead? */
	ret = inode_permission(file_inode(cfile.file), MAY_READ);
	if (ret < 0)
		goto out_put_cfile;

3891 3892 3893 3894 3895
	/*
	 * Determine the event callbacks and set them in @event.  This used
	 * to be done via struct cftype but cgroup core no longer knows
	 * about these events.  The following is crude but the whole thing
	 * is for compatibility anyway.
3896 3897
	 *
	 * DO NOT ADD NEW FILES.
3898
	 */
A
Al Viro 已提交
3899
	name = cfile.file->f_path.dentry->d_name.name;
3900 3901 3902 3903 3904 3905 3906 3907 3908 3909 3910

	if (!strcmp(name, "memory.usage_in_bytes")) {
		event->register_event = mem_cgroup_usage_register_event;
		event->unregister_event = mem_cgroup_usage_unregister_event;
	} else if (!strcmp(name, "memory.oom_control")) {
		event->register_event = mem_cgroup_oom_register_event;
		event->unregister_event = mem_cgroup_oom_unregister_event;
	} else if (!strcmp(name, "memory.pressure_level")) {
		event->register_event = vmpressure_register_event;
		event->unregister_event = vmpressure_unregister_event;
	} else if (!strcmp(name, "memory.memsw.usage_in_bytes")) {
T
Tejun Heo 已提交
3911 3912
		event->register_event = memsw_cgroup_usage_register_event;
		event->unregister_event = memsw_cgroup_usage_unregister_event;
3913 3914 3915 3916 3917
	} else {
		ret = -EINVAL;
		goto out_put_cfile;
	}

3918
	/*
3919 3920 3921
	 * Verify @cfile should belong to @css.  Also, remaining events are
	 * automatically removed on cgroup destruction but the removal is
	 * asynchronous, so take an extra ref on @css.
3922
	 */
A
Al Viro 已提交
3923
	cfile_css = css_tryget_online_from_dir(cfile.file->f_path.dentry->d_parent,
3924
					       &memory_cgrp_subsys);
3925
	ret = -EINVAL;
3926
	if (IS_ERR(cfile_css))
3927
		goto out_put_cfile;
3928 3929
	if (cfile_css != css) {
		css_put(cfile_css);
3930
		goto out_put_cfile;
3931
	}
3932

3933
	ret = event->register_event(memcg, event->eventfd, buf);
3934 3935 3936 3937 3938
	if (ret)
		goto out_put_css;

	efile.file->f_op->poll(efile.file, &event->pt);

3939 3940 3941
	spin_lock(&memcg->event_list_lock);
	list_add(&event->list, &memcg->event_list);
	spin_unlock(&memcg->event_list_lock);
3942 3943 3944 3945

	fdput(cfile);
	fdput(efile);

3946
	return nbytes;
3947 3948

out_put_css:
3949
	css_put(css);
3950 3951 3952 3953 3954 3955 3956 3957 3958 3959 3960 3961
out_put_cfile:
	fdput(cfile);
out_put_eventfd:
	eventfd_ctx_put(event->eventfd);
out_put_efile:
	fdput(efile);
out_kfree:
	kfree(event);

	return ret;
}

3962
static struct cftype mem_cgroup_legacy_files[] = {
B
Balbir Singh 已提交
3963
	{
3964
		.name = "usage_in_bytes",
3965
		.private = MEMFILE_PRIVATE(_MEM, RES_USAGE),
3966
		.read_u64 = mem_cgroup_read_u64,
B
Balbir Singh 已提交
3967
	},
3968 3969
	{
		.name = "max_usage_in_bytes",
3970
		.private = MEMFILE_PRIVATE(_MEM, RES_MAX_USAGE),
3971
		.write = mem_cgroup_reset,
3972
		.read_u64 = mem_cgroup_read_u64,
3973
	},
B
Balbir Singh 已提交
3974
	{
3975
		.name = "limit_in_bytes",
3976
		.private = MEMFILE_PRIVATE(_MEM, RES_LIMIT),
3977
		.write = mem_cgroup_write,
3978
		.read_u64 = mem_cgroup_read_u64,
B
Balbir Singh 已提交
3979
	},
3980 3981 3982
	{
		.name = "soft_limit_in_bytes",
		.private = MEMFILE_PRIVATE(_MEM, RES_SOFT_LIMIT),
3983
		.write = mem_cgroup_write,
3984
		.read_u64 = mem_cgroup_read_u64,
3985
	},
B
Balbir Singh 已提交
3986 3987
	{
		.name = "failcnt",
3988
		.private = MEMFILE_PRIVATE(_MEM, RES_FAILCNT),
3989
		.write = mem_cgroup_reset,
3990
		.read_u64 = mem_cgroup_read_u64,
B
Balbir Singh 已提交
3991
	},
3992 3993
	{
		.name = "stat",
3994
		.seq_show = memcg_stat_show,
3995
	},
3996 3997
	{
		.name = "force_empty",
3998
		.write = mem_cgroup_force_empty_write,
3999
	},
4000 4001 4002 4003 4004
	{
		.name = "use_hierarchy",
		.write_u64 = mem_cgroup_hierarchy_write,
		.read_u64 = mem_cgroup_hierarchy_read,
	},
4005
	{
4006
		.name = "cgroup.event_control",		/* XXX: for compat */
4007
		.write = memcg_write_event_control,
4008
		.flags = CFTYPE_NO_PREFIX | CFTYPE_WORLD_WRITABLE,
4009
	},
K
KOSAKI Motohiro 已提交
4010 4011 4012 4013 4014
	{
		.name = "swappiness",
		.read_u64 = mem_cgroup_swappiness_read,
		.write_u64 = mem_cgroup_swappiness_write,
	},
4015 4016 4017 4018 4019
	{
		.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
4020 4021
	{
		.name = "oom_control",
4022
		.seq_show = mem_cgroup_oom_control_read,
4023
		.write_u64 = mem_cgroup_oom_control_write,
K
KAMEZAWA Hiroyuki 已提交
4024 4025
		.private = MEMFILE_PRIVATE(_OOM_TYPE, OOM_CONTROL),
	},
4026 4027 4028
	{
		.name = "pressure_level",
	},
4029 4030 4031
#ifdef CONFIG_NUMA
	{
		.name = "numa_stat",
4032
		.seq_show = memcg_numa_stat_show,
4033 4034
	},
#endif
4035 4036 4037 4038
#ifdef CONFIG_MEMCG_KMEM
	{
		.name = "kmem.limit_in_bytes",
		.private = MEMFILE_PRIVATE(_KMEM, RES_LIMIT),
4039
		.write = mem_cgroup_write,
4040
		.read_u64 = mem_cgroup_read_u64,
4041 4042 4043 4044
	},
	{
		.name = "kmem.usage_in_bytes",
		.private = MEMFILE_PRIVATE(_KMEM, RES_USAGE),
4045
		.read_u64 = mem_cgroup_read_u64,
4046 4047 4048 4049
	},
	{
		.name = "kmem.failcnt",
		.private = MEMFILE_PRIVATE(_KMEM, RES_FAILCNT),
4050
		.write = mem_cgroup_reset,
4051
		.read_u64 = mem_cgroup_read_u64,
4052 4053 4054 4055
	},
	{
		.name = "kmem.max_usage_in_bytes",
		.private = MEMFILE_PRIVATE(_KMEM, RES_MAX_USAGE),
4056
		.write = mem_cgroup_reset,
4057
		.read_u64 = mem_cgroup_read_u64,
4058
	},
4059 4060 4061
#ifdef CONFIG_SLABINFO
	{
		.name = "kmem.slabinfo",
4062 4063 4064 4065
		.seq_start = slab_start,
		.seq_next = slab_next,
		.seq_stop = slab_stop,
		.seq_show = memcg_slab_show,
4066 4067
	},
#endif
4068
#endif
4069
	{ },	/* terminate */
4070
};
4071

4072
static int alloc_mem_cgroup_per_zone_info(struct mem_cgroup *memcg, int node)
4073 4074
{
	struct mem_cgroup_per_node *pn;
4075
	struct mem_cgroup_per_zone *mz;
4076
	int zone, tmp = node;
4077 4078 4079 4080 4081 4082 4083 4084
	/*
	 * 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.
	 */
4085 4086
	if (!node_state(node, N_NORMAL_MEMORY))
		tmp = -1;
4087
	pn = kzalloc_node(sizeof(*pn), GFP_KERNEL, tmp);
4088 4089
	if (!pn)
		return 1;
4090 4091 4092

	for (zone = 0; zone < MAX_NR_ZONES; zone++) {
		mz = &pn->zoneinfo[zone];
4093
		lruvec_init(&mz->lruvec);
4094 4095
		mz->usage_in_excess = 0;
		mz->on_tree = false;
4096
		mz->memcg = memcg;
4097
	}
4098
	memcg->nodeinfo[node] = pn;
4099 4100 4101
	return 0;
}

4102
static void free_mem_cgroup_per_zone_info(struct mem_cgroup *memcg, int node)
4103
{
4104
	kfree(memcg->nodeinfo[node]);
4105 4106
}

4107 4108
static struct mem_cgroup *mem_cgroup_alloc(void)
{
4109
	struct mem_cgroup *memcg;
4110
	size_t size;
4111

4112 4113
	size = sizeof(struct mem_cgroup);
	size += nr_node_ids * sizeof(struct mem_cgroup_per_node *);
4114

4115
	memcg = kzalloc(size, GFP_KERNEL);
4116
	if (!memcg)
4117 4118
		return NULL;

4119 4120
	memcg->stat = alloc_percpu(struct mem_cgroup_stat_cpu);
	if (!memcg->stat)
4121
		goto out_free;
T
Tejun Heo 已提交
4122 4123 4124 4125

	if (memcg_wb_domain_init(memcg, GFP_KERNEL))
		goto out_free_stat;

4126
	return memcg;
4127

T
Tejun Heo 已提交
4128 4129
out_free_stat:
	free_percpu(memcg->stat);
4130
out_free:
4131
	kfree(memcg);
4132
	return NULL;
4133 4134
}

4135
/*
4136 4137 4138 4139 4140 4141 4142 4143
 * 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.
4144
 */
4145 4146

static void __mem_cgroup_free(struct mem_cgroup *memcg)
4147
{
4148
	int node;
4149

4150 4151
	cancel_work_sync(&memcg->high_work);

4152
	mem_cgroup_remove_from_trees(memcg);
4153 4154 4155 4156 4157

	for_each_node(node)
		free_mem_cgroup_per_zone_info(memcg, node);

	free_percpu(memcg->stat);
T
Tejun Heo 已提交
4158
	memcg_wb_domain_exit(memcg);
4159
	kfree(memcg);
4160
}
4161

L
Li Zefan 已提交
4162
static struct cgroup_subsys_state * __ref
4163
mem_cgroup_css_alloc(struct cgroup_subsys_state *parent_css)
B
Balbir Singh 已提交
4164
{
4165
	struct mem_cgroup *memcg;
K
KAMEZAWA Hiroyuki 已提交
4166
	long error = -ENOMEM;
4167
	int node;
B
Balbir Singh 已提交
4168

4169 4170
	memcg = mem_cgroup_alloc();
	if (!memcg)
K
KAMEZAWA Hiroyuki 已提交
4171
		return ERR_PTR(error);
4172

B
Bob Liu 已提交
4173
	for_each_node(node)
4174
		if (alloc_mem_cgroup_per_zone_info(memcg, node))
4175
			goto free_out;
4176

4177
	/* root ? */
4178
	if (parent_css == NULL) {
4179
		root_mem_cgroup = memcg;
4180
		page_counter_init(&memcg->memory, NULL);
4181
		memcg->high = PAGE_COUNTER_MAX;
4182
		memcg->soft_limit = PAGE_COUNTER_MAX;
4183 4184
		page_counter_init(&memcg->memsw, NULL);
		page_counter_init(&memcg->kmem, NULL);
4185
	}
4186

4187
	INIT_WORK(&memcg->high_work, high_work_func);
4188 4189 4190 4191 4192
	memcg->last_scanned_node = MAX_NUMNODES;
	INIT_LIST_HEAD(&memcg->oom_notify);
	memcg->move_charge_at_immigrate = 0;
	mutex_init(&memcg->thresholds_lock);
	spin_lock_init(&memcg->move_lock);
4193
	vmpressure_init(&memcg->vmpressure);
4194 4195
	INIT_LIST_HEAD(&memcg->event_list);
	spin_lock_init(&memcg->event_list_lock);
4196
#ifndef CONFIG_SLOB
V
Vladimir Davydov 已提交
4197 4198
	memcg->kmemcg_id = -1;
#endif
4199 4200
#ifdef CONFIG_CGROUP_WRITEBACK
	INIT_LIST_HEAD(&memcg->cgwb_list);
4201 4202 4203
#endif
#ifdef CONFIG_INET
	memcg->socket_pressure = jiffies;
4204
#endif
4205 4206 4207 4208 4209 4210 4211 4212
	return &memcg->css;

free_out:
	__mem_cgroup_free(memcg);
	return ERR_PTR(error);
}

static int
4213
mem_cgroup_css_online(struct cgroup_subsys_state *css)
4214
{
4215
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
T
Tejun Heo 已提交
4216
	struct mem_cgroup *parent = mem_cgroup_from_css(css->parent);
4217
	int ret;
4218

4219
	if (css->id > MEM_CGROUP_ID_MAX)
4220 4221
		return -ENOSPC;

T
Tejun Heo 已提交
4222
	if (!parent)
4223 4224
		return 0;

4225
	mutex_lock(&memcg_create_mutex);
4226 4227 4228 4229 4230 4231

	memcg->use_hierarchy = parent->use_hierarchy;
	memcg->oom_kill_disable = parent->oom_kill_disable;
	memcg->swappiness = mem_cgroup_swappiness(parent);

	if (parent->use_hierarchy) {
4232
		page_counter_init(&memcg->memory, &parent->memory);
4233
		memcg->high = PAGE_COUNTER_MAX;
4234
		memcg->soft_limit = PAGE_COUNTER_MAX;
4235 4236
		page_counter_init(&memcg->memsw, &parent->memsw);
		page_counter_init(&memcg->kmem, &parent->kmem);
4237

4238
		/*
4239 4240
		 * No need to take a reference to the parent because cgroup
		 * core guarantees its existence.
4241
		 */
4242
	} else {
4243
		page_counter_init(&memcg->memory, NULL);
4244
		memcg->high = PAGE_COUNTER_MAX;
4245
		memcg->soft_limit = PAGE_COUNTER_MAX;
4246 4247
		page_counter_init(&memcg->memsw, NULL);
		page_counter_init(&memcg->kmem, NULL);
4248 4249 4250 4251 4252
		/*
		 * Deeper hierachy with use_hierarchy == false doesn't make
		 * much sense so let cgroup subsystem know about this
		 * unfortunate state in our controller.
		 */
4253
		if (parent != root_mem_cgroup)
4254
			memory_cgrp_subsys.broken_hierarchy = true;
4255
	}
4256
	mutex_unlock(&memcg_create_mutex);
4257

4258
	ret = memcg_propagate_kmem(memcg);
4259 4260
	if (ret)
		return ret;
4261 4262

#ifdef CONFIG_MEMCG_KMEM
4263 4264 4265 4266
	ret = tcp_init_cgroup(memcg);
	if (ret)
		return ret;
#endif
4267

4268 4269
#ifdef CONFIG_INET
	if (cgroup_subsys_on_dfl(memory_cgrp_subsys) && !cgroup_memory_nosocket)
4270
		static_branch_inc(&memcg_sockets_enabled_key);
4271 4272
#endif

4273 4274 4275 4276 4277 4278 4279 4280
	/*
	 * Make sure the memcg is initialized: mem_cgroup_iter()
	 * orders reading memcg->initialized against its callers
	 * reading the memcg members.
	 */
	smp_store_release(&memcg->initialized, 1);

	return 0;
B
Balbir Singh 已提交
4281 4282
}

4283
static void mem_cgroup_css_offline(struct cgroup_subsys_state *css)
4284
{
4285
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
4286
	struct mem_cgroup_event *event, *tmp;
4287 4288 4289 4290 4291 4292

	/*
	 * Unregister events and notify userspace.
	 * Notify userspace about cgroup removing only after rmdir of cgroup
	 * directory to avoid race between userspace and kernelspace.
	 */
4293 4294
	spin_lock(&memcg->event_list_lock);
	list_for_each_entry_safe(event, tmp, &memcg->event_list, list) {
4295 4296 4297
		list_del_init(&event->list);
		schedule_work(&event->remove);
	}
4298
	spin_unlock(&memcg->event_list_lock);
4299

4300
	vmpressure_cleanup(&memcg->vmpressure);
4301

4302
	memcg_offline_kmem(memcg);
4303 4304

	wb_memcg_offline(memcg);
4305 4306
}

4307 4308 4309 4310 4311 4312 4313
static void mem_cgroup_css_released(struct cgroup_subsys_state *css)
{
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);

	invalidate_reclaim_iterators(memcg);
}

4314
static void mem_cgroup_css_free(struct cgroup_subsys_state *css)
B
Balbir Singh 已提交
4315
{
4316
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
4317

4318 4319
#ifdef CONFIG_INET
	if (cgroup_subsys_on_dfl(memory_cgrp_subsys) && !cgroup_memory_nosocket)
4320
		static_branch_dec(&memcg_sockets_enabled_key);
4321
#endif
4322 4323

	memcg_free_kmem(memcg);
4324 4325

#ifdef CONFIG_MEMCG_KMEM
4326 4327 4328
	tcp_destroy_cgroup(memcg);
#endif

4329
	__mem_cgroup_free(memcg);
B
Balbir Singh 已提交
4330 4331
}

4332 4333 4334 4335 4336 4337 4338 4339 4340 4341 4342 4343 4344 4345 4346 4347 4348
/**
 * mem_cgroup_css_reset - reset the states of a mem_cgroup
 * @css: the target css
 *
 * Reset the states of the mem_cgroup associated with @css.  This is
 * invoked when the userland requests disabling on the default hierarchy
 * but the memcg is pinned through dependency.  The memcg should stop
 * applying policies and should revert to the vanilla state as it may be
 * made visible again.
 *
 * The current implementation only resets the essential configurations.
 * This needs to be expanded to cover all the visible parts.
 */
static void mem_cgroup_css_reset(struct cgroup_subsys_state *css)
{
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);

4349 4350 4351
	mem_cgroup_resize_limit(memcg, PAGE_COUNTER_MAX);
	mem_cgroup_resize_memsw_limit(memcg, PAGE_COUNTER_MAX);
	memcg_update_kmem_limit(memcg, PAGE_COUNTER_MAX);
4352 4353
	memcg->low = 0;
	memcg->high = PAGE_COUNTER_MAX;
4354
	memcg->soft_limit = PAGE_COUNTER_MAX;
4355
	memcg_wb_domain_size_changed(memcg);
4356 4357
}

4358
#ifdef CONFIG_MMU
4359
/* Handlers for move charge at task migration. */
4360
static int mem_cgroup_do_precharge(unsigned long count)
4361
{
4362
	int ret;
4363

4364 4365
	/* Try a single bulk charge without reclaim first, kswapd may wake */
	ret = try_charge(mc.to, GFP_KERNEL & ~__GFP_DIRECT_RECLAIM, count);
4366
	if (!ret) {
4367 4368 4369
		mc.precharge += count;
		return ret;
	}
4370 4371

	/* Try charges one by one with reclaim */
4372
	while (count--) {
4373
		ret = try_charge(mc.to, GFP_KERNEL & ~__GFP_NORETRY, 1);
4374 4375
		if (ret)
			return ret;
4376
		mc.precharge++;
4377
		cond_resched();
4378
	}
4379
	return 0;
4380 4381 4382
}

/**
4383
 * get_mctgt_type - get target type of moving charge
4384 4385 4386
 * @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
4387
 * @target: the pointer the target page or swap ent will be stored(can be NULL)
4388 4389 4390 4391 4392 4393
 *
 * 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).
4394 4395 4396
 *   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.
4397 4398 4399 4400 4401
 *
 * Called with pte lock held.
 */
union mc_target {
	struct page	*page;
4402
	swp_entry_t	ent;
4403 4404 4405
};

enum mc_target_type {
4406
	MC_TARGET_NONE = 0,
4407
	MC_TARGET_PAGE,
4408
	MC_TARGET_SWAP,
4409 4410
};

D
Daisuke Nishimura 已提交
4411 4412
static struct page *mc_handle_present_pte(struct vm_area_struct *vma,
						unsigned long addr, pte_t ptent)
4413
{
D
Daisuke Nishimura 已提交
4414
	struct page *page = vm_normal_page(vma, addr, ptent);
4415

D
Daisuke Nishimura 已提交
4416 4417 4418
	if (!page || !page_mapped(page))
		return NULL;
	if (PageAnon(page)) {
4419
		if (!(mc.flags & MOVE_ANON))
D
Daisuke Nishimura 已提交
4420
			return NULL;
4421 4422 4423 4424
	} else {
		if (!(mc.flags & MOVE_FILE))
			return NULL;
	}
D
Daisuke Nishimura 已提交
4425 4426 4427 4428 4429 4430
	if (!get_page_unless_zero(page))
		return NULL;

	return page;
}

4431
#ifdef CONFIG_SWAP
D
Daisuke Nishimura 已提交
4432 4433 4434 4435 4436 4437
static struct page *mc_handle_swap_pte(struct vm_area_struct *vma,
			unsigned long addr, pte_t ptent, swp_entry_t *entry)
{
	struct page *page = NULL;
	swp_entry_t ent = pte_to_swp_entry(ptent);

4438
	if (!(mc.flags & MOVE_ANON) || non_swap_entry(ent))
D
Daisuke Nishimura 已提交
4439
		return NULL;
4440 4441 4442 4443
	/*
	 * Because lookup_swap_cache() updates some statistics counter,
	 * we call find_get_page() with swapper_space directly.
	 */
4444
	page = find_get_page(swap_address_space(ent), ent.val);
4445
	if (do_memsw_account())
D
Daisuke Nishimura 已提交
4446 4447 4448 4449
		entry->val = ent.val;

	return page;
}
4450 4451 4452 4453 4454 4455 4456
#else
static struct page *mc_handle_swap_pte(struct vm_area_struct *vma,
			unsigned long addr, pte_t ptent, swp_entry_t *entry)
{
	return NULL;
}
#endif
D
Daisuke Nishimura 已提交
4457

4458 4459 4460 4461 4462 4463 4464 4465 4466
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 address_space *mapping;
	pgoff_t pgoff;

	if (!vma->vm_file) /* anonymous vma */
		return NULL;
4467
	if (!(mc.flags & MOVE_FILE))
4468 4469 4470
		return NULL;

	mapping = vma->vm_file->f_mapping;
4471
	pgoff = linear_page_index(vma, addr);
4472 4473

	/* page is moved even if it's not RSS of this task(page-faulted). */
4474 4475
#ifdef CONFIG_SWAP
	/* shmem/tmpfs may report page out on swap: account for that too. */
4476 4477 4478 4479
	if (shmem_mapping(mapping)) {
		page = find_get_entry(mapping, pgoff);
		if (radix_tree_exceptional_entry(page)) {
			swp_entry_t swp = radix_to_swp_entry(page);
4480
			if (do_memsw_account())
4481 4482 4483 4484 4485 4486 4487
				*entry = swp;
			page = find_get_page(swap_address_space(swp), swp.val);
		}
	} else
		page = find_get_page(mapping, pgoff);
#else
	page = find_get_page(mapping, pgoff);
4488
#endif
4489 4490 4491
	return page;
}

4492 4493 4494 4495 4496 4497 4498
/**
 * mem_cgroup_move_account - move account of the page
 * @page: the page
 * @nr_pages: number of regular pages (>1 for huge pages)
 * @from: mem_cgroup which the page is moved from.
 * @to:	mem_cgroup which the page is moved to. @from != @to.
 *
4499
 * The caller must make sure the page is not on LRU (isolate_page() is useful.)
4500 4501 4502 4503 4504
 *
 * This function doesn't do "charge" to new cgroup and doesn't do "uncharge"
 * from old cgroup.
 */
static int mem_cgroup_move_account(struct page *page,
4505
				   bool compound,
4506 4507 4508 4509
				   struct mem_cgroup *from,
				   struct mem_cgroup *to)
{
	unsigned long flags;
4510
	unsigned int nr_pages = compound ? hpage_nr_pages(page) : 1;
4511
	int ret;
4512
	bool anon;
4513 4514 4515

	VM_BUG_ON(from == to);
	VM_BUG_ON_PAGE(PageLRU(page), page);
4516
	VM_BUG_ON(compound && !PageTransHuge(page));
4517 4518

	/*
4519 4520
	 * Prevent mem_cgroup_replace_page() from looking at
	 * page->mem_cgroup of its source page while we change it.
4521
	 */
4522
	ret = -EBUSY;
4523 4524 4525 4526 4527 4528 4529
	if (!trylock_page(page))
		goto out;

	ret = -EINVAL;
	if (page->mem_cgroup != from)
		goto out_unlock;

4530 4531
	anon = PageAnon(page);

4532 4533
	spin_lock_irqsave(&from->move_lock, flags);

4534
	if (!anon && page_mapped(page)) {
4535 4536 4537 4538 4539 4540
		__this_cpu_sub(from->stat->count[MEM_CGROUP_STAT_FILE_MAPPED],
			       nr_pages);
		__this_cpu_add(to->stat->count[MEM_CGROUP_STAT_FILE_MAPPED],
			       nr_pages);
	}

4541 4542 4543 4544 4545 4546 4547 4548 4549 4550 4551 4552 4553 4554 4555 4556
	/*
	 * move_lock grabbed above and caller set from->moving_account, so
	 * mem_cgroup_update_page_stat() will serialize updates to PageDirty.
	 * So mapping should be stable for dirty pages.
	 */
	if (!anon && PageDirty(page)) {
		struct address_space *mapping = page_mapping(page);

		if (mapping_cap_account_dirty(mapping)) {
			__this_cpu_sub(from->stat->count[MEM_CGROUP_STAT_DIRTY],
				       nr_pages);
			__this_cpu_add(to->stat->count[MEM_CGROUP_STAT_DIRTY],
				       nr_pages);
		}
	}

4557 4558 4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576
	if (PageWriteback(page)) {
		__this_cpu_sub(from->stat->count[MEM_CGROUP_STAT_WRITEBACK],
			       nr_pages);
		__this_cpu_add(to->stat->count[MEM_CGROUP_STAT_WRITEBACK],
			       nr_pages);
	}

	/*
	 * It is safe to change page->mem_cgroup here because the page
	 * is referenced, charged, and isolated - we can't race with
	 * uncharging, charging, migration, or LRU putback.
	 */

	/* caller should have done css_get */
	page->mem_cgroup = to;
	spin_unlock_irqrestore(&from->move_lock, flags);

	ret = 0;

	local_irq_disable();
4577
	mem_cgroup_charge_statistics(to, page, compound, nr_pages);
4578
	memcg_check_events(to, page);
4579
	mem_cgroup_charge_statistics(from, page, compound, -nr_pages);
4580 4581 4582 4583 4584 4585 4586 4587
	memcg_check_events(from, page);
	local_irq_enable();
out_unlock:
	unlock_page(page);
out:
	return ret;
}

4588
static enum mc_target_type get_mctgt_type(struct vm_area_struct *vma,
D
Daisuke Nishimura 已提交
4589 4590 4591
		unsigned long addr, pte_t ptent, union mc_target *target)
{
	struct page *page = NULL;
4592
	enum mc_target_type ret = MC_TARGET_NONE;
D
Daisuke Nishimura 已提交
4593 4594 4595 4596 4597 4598
	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);
4599
	else if (pte_none(ptent))
4600
		page = mc_handle_file_pte(vma, addr, ptent, &ent);
D
Daisuke Nishimura 已提交
4601 4602

	if (!page && !ent.val)
4603
		return ret;
4604 4605
	if (page) {
		/*
4606
		 * Do only loose check w/o serialization.
4607
		 * mem_cgroup_move_account() checks the page is valid or
4608
		 * not under LRU exclusion.
4609
		 */
4610
		if (page->mem_cgroup == mc.from) {
4611 4612 4613 4614 4615 4616 4617
			ret = MC_TARGET_PAGE;
			if (target)
				target->page = page;
		}
		if (!ret || !target)
			put_page(page);
	}
D
Daisuke Nishimura 已提交
4618 4619
	/* There is a swap entry and a page doesn't exist or isn't charged */
	if (ent.val && !ret &&
L
Li Zefan 已提交
4620
	    mem_cgroup_id(mc.from) == lookup_swap_cgroup_id(ent)) {
4621 4622 4623
		ret = MC_TARGET_SWAP;
		if (target)
			target->ent = ent;
4624 4625 4626 4627
	}
	return ret;
}

4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
/*
 * We don't consider swapping or file mapped pages because THP does not
 * support them for now.
 * Caller should make sure that pmd_trans_huge(pmd) is true.
 */
static enum mc_target_type get_mctgt_type_thp(struct vm_area_struct *vma,
		unsigned long addr, pmd_t pmd, union mc_target *target)
{
	struct page *page = NULL;
	enum mc_target_type ret = MC_TARGET_NONE;

	page = pmd_page(pmd);
4641
	VM_BUG_ON_PAGE(!page || !PageHead(page), page);
4642
	if (!(mc.flags & MOVE_ANON))
4643
		return ret;
4644
	if (page->mem_cgroup == mc.from) {
4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660
		ret = MC_TARGET_PAGE;
		if (target) {
			get_page(page);
			target->page = page;
		}
	}
	return ret;
}
#else
static inline enum mc_target_type get_mctgt_type_thp(struct vm_area_struct *vma,
		unsigned long addr, pmd_t pmd, union mc_target *target)
{
	return MC_TARGET_NONE;
}
#endif

4661 4662 4663 4664
static int mem_cgroup_count_precharge_pte_range(pmd_t *pmd,
					unsigned long addr, unsigned long end,
					struct mm_walk *walk)
{
4665
	struct vm_area_struct *vma = walk->vma;
4666 4667 4668
	pte_t *pte;
	spinlock_t *ptl;

4669
	if (pmd_trans_huge_lock(pmd, vma, &ptl)) {
4670 4671
		if (get_mctgt_type_thp(vma, addr, *pmd, NULL) == MC_TARGET_PAGE)
			mc.precharge += HPAGE_PMD_NR;
4672
		spin_unlock(ptl);
4673
		return 0;
4674
	}
4675

4676 4677
	if (pmd_trans_unstable(pmd))
		return 0;
4678 4679
	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
	for (; addr != end; pte++, addr += PAGE_SIZE)
4680
		if (get_mctgt_type(vma, addr, *pte, NULL))
4681 4682 4683 4684
			mc.precharge++;	/* increment precharge temporarily */
	pte_unmap_unlock(pte - 1, ptl);
	cond_resched();

4685 4686 4687
	return 0;
}

4688 4689 4690 4691
static unsigned long mem_cgroup_count_precharge(struct mm_struct *mm)
{
	unsigned long precharge;

4692 4693 4694 4695
	struct mm_walk mem_cgroup_count_precharge_walk = {
		.pmd_entry = mem_cgroup_count_precharge_pte_range,
		.mm = mm,
	};
4696
	down_read(&mm->mmap_sem);
4697
	walk_page_range(0, ~0UL, &mem_cgroup_count_precharge_walk);
4698
	up_read(&mm->mmap_sem);
4699 4700 4701 4702 4703 4704 4705 4706 4707

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

	return precharge;
}

static int mem_cgroup_precharge_mc(struct mm_struct *mm)
{
4708 4709 4710 4711 4712
	unsigned long precharge = mem_cgroup_count_precharge(mm);

	VM_BUG_ON(mc.moving_task);
	mc.moving_task = current;
	return mem_cgroup_do_precharge(precharge);
4713 4714
}

4715 4716
/* cancels all extra charges on mc.from and mc.to, and wakes up all waiters. */
static void __mem_cgroup_clear_mc(void)
4717
{
4718 4719 4720
	struct mem_cgroup *from = mc.from;
	struct mem_cgroup *to = mc.to;

4721
	/* we must uncharge all the leftover precharges from mc.to */
4722
	if (mc.precharge) {
4723
		cancel_charge(mc.to, mc.precharge);
4724 4725 4726 4727 4728 4729 4730
		mc.precharge = 0;
	}
	/*
	 * we didn't uncharge from mc.from at mem_cgroup_move_account(), so
	 * we must uncharge here.
	 */
	if (mc.moved_charge) {
4731
		cancel_charge(mc.from, mc.moved_charge);
4732
		mc.moved_charge = 0;
4733
	}
4734 4735 4736
	/* we must fixup refcnts and charges */
	if (mc.moved_swap) {
		/* uncharge swap account from the old cgroup */
4737
		if (!mem_cgroup_is_root(mc.from))
4738
			page_counter_uncharge(&mc.from->memsw, mc.moved_swap);
4739

4740
		/*
4741 4742
		 * we charged both to->memory and to->memsw, so we
		 * should uncharge to->memory.
4743
		 */
4744
		if (!mem_cgroup_is_root(mc.to))
4745 4746
			page_counter_uncharge(&mc.to->memory, mc.moved_swap);

4747
		css_put_many(&mc.from->css, mc.moved_swap);
4748

L
Li Zefan 已提交
4749
		/* we've already done css_get(mc.to) */
4750 4751
		mc.moved_swap = 0;
	}
4752 4753 4754 4755 4756 4757 4758 4759 4760 4761 4762 4763 4764
	memcg_oom_recover(from);
	memcg_oom_recover(to);
	wake_up_all(&mc.waitq);
}

static void mem_cgroup_clear_mc(void)
{
	/*
	 * we must clear moving_task before waking up waiters at the end of
	 * task migration.
	 */
	mc.moving_task = NULL;
	__mem_cgroup_clear_mc();
4765
	spin_lock(&mc.lock);
4766 4767
	mc.from = NULL;
	mc.to = NULL;
4768
	spin_unlock(&mc.lock);
4769 4770
}

4771
static int mem_cgroup_can_attach(struct cgroup_taskset *tset)
4772
{
4773
	struct cgroup_subsys_state *css;
4774
	struct mem_cgroup *memcg = NULL; /* unneeded init to make gcc happy */
4775
	struct mem_cgroup *from;
4776
	struct task_struct *leader, *p;
4777
	struct mm_struct *mm;
4778
	unsigned long move_flags;
4779
	int ret = 0;
4780

4781 4782
	/* charge immigration isn't supported on the default hierarchy */
	if (cgroup_subsys_on_dfl(memory_cgrp_subsys))
4783 4784
		return 0;

4785 4786 4787 4788 4789 4790 4791
	/*
	 * Multi-process migrations only happen on the default hierarchy
	 * where charge immigration is not used.  Perform charge
	 * immigration if @tset contains a leader and whine if there are
	 * multiple.
	 */
	p = NULL;
4792
	cgroup_taskset_for_each_leader(leader, css, tset) {
4793 4794
		WARN_ON_ONCE(p);
		p = leader;
4795
		memcg = mem_cgroup_from_css(css);
4796 4797 4798 4799
	}
	if (!p)
		return 0;

4800 4801 4802 4803 4804 4805 4806 4807 4808
	/*
	 * We are now commited to this value whatever it is. Changes in this
	 * tunable will only affect upcoming migrations, not the current one.
	 * So we need to save it, and keep it going.
	 */
	move_flags = READ_ONCE(memcg->move_charge_at_immigrate);
	if (!move_flags)
		return 0;

4809 4810 4811 4812 4813 4814 4815 4816 4817 4818 4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830 4831 4832 4833
	from = mem_cgroup_from_task(p);

	VM_BUG_ON(from == memcg);

	mm = get_task_mm(p);
	if (!mm)
		return 0;
	/* We move charges only when we move a owner of the mm */
	if (mm->owner == p) {
		VM_BUG_ON(mc.from);
		VM_BUG_ON(mc.to);
		VM_BUG_ON(mc.precharge);
		VM_BUG_ON(mc.moved_charge);
		VM_BUG_ON(mc.moved_swap);

		spin_lock(&mc.lock);
		mc.from = from;
		mc.to = memcg;
		mc.flags = move_flags;
		spin_unlock(&mc.lock);
		/* We set mc.moving_task later */

		ret = mem_cgroup_precharge_mc(mm);
		if (ret)
			mem_cgroup_clear_mc();
4834
	}
4835
	mmput(mm);
4836 4837 4838
	return ret;
}

4839
static void mem_cgroup_cancel_attach(struct cgroup_taskset *tset)
4840
{
4841 4842
	if (mc.to)
		mem_cgroup_clear_mc();
4843 4844
}

4845 4846 4847
static int mem_cgroup_move_charge_pte_range(pmd_t *pmd,
				unsigned long addr, unsigned long end,
				struct mm_walk *walk)
4848
{
4849
	int ret = 0;
4850
	struct vm_area_struct *vma = walk->vma;
4851 4852
	pte_t *pte;
	spinlock_t *ptl;
4853 4854 4855
	enum mc_target_type target_type;
	union mc_target target;
	struct page *page;
4856

4857
	if (pmd_trans_huge_lock(pmd, vma, &ptl)) {
4858
		if (mc.precharge < HPAGE_PMD_NR) {
4859
			spin_unlock(ptl);
4860 4861 4862 4863 4864 4865
			return 0;
		}
		target_type = get_mctgt_type_thp(vma, addr, *pmd, &target);
		if (target_type == MC_TARGET_PAGE) {
			page = target.page;
			if (!isolate_lru_page(page)) {
4866
				if (!mem_cgroup_move_account(page, true,
4867
							     mc.from, mc.to)) {
4868 4869 4870 4871 4872 4873 4874
					mc.precharge -= HPAGE_PMD_NR;
					mc.moved_charge += HPAGE_PMD_NR;
				}
				putback_lru_page(page);
			}
			put_page(page);
		}
4875
		spin_unlock(ptl);
4876
		return 0;
4877 4878
	}

4879 4880
	if (pmd_trans_unstable(pmd))
		return 0;
4881 4882 4883 4884
retry:
	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
	for (; addr != end; addr += PAGE_SIZE) {
		pte_t ptent = *(pte++);
4885
		swp_entry_t ent;
4886 4887 4888 4889

		if (!mc.precharge)
			break;

4890
		switch (get_mctgt_type(vma, addr, ptent, &target)) {
4891 4892
		case MC_TARGET_PAGE:
			page = target.page;
4893 4894 4895 4896 4897 4898 4899 4900
			/*
			 * We can have a part of the split pmd here. Moving it
			 * can be done but it would be too convoluted so simply
			 * ignore such a partial THP and keep it in original
			 * memcg. There should be somebody mapping the head.
			 */
			if (PageTransCompound(page))
				goto put;
4901 4902
			if (isolate_lru_page(page))
				goto put;
4903 4904
			if (!mem_cgroup_move_account(page, false,
						mc.from, mc.to)) {
4905
				mc.precharge--;
4906 4907
				/* we uncharge from mc.from later. */
				mc.moved_charge++;
4908 4909
			}
			putback_lru_page(page);
4910
put:			/* get_mctgt_type() gets the page */
4911 4912
			put_page(page);
			break;
4913 4914
		case MC_TARGET_SWAP:
			ent = target.ent;
4915
			if (!mem_cgroup_move_swap_account(ent, mc.from, mc.to)) {
4916
				mc.precharge--;
4917 4918 4919
				/* we fixup refcnts and charges later. */
				mc.moved_swap++;
			}
4920
			break;
4921 4922 4923 4924 4925 4926 4927 4928 4929 4930 4931 4932 4933 4934
		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.
		 */
4935
		ret = mem_cgroup_do_precharge(1);
4936 4937 4938 4939 4940 4941 4942 4943 4944
		if (!ret)
			goto retry;
	}

	return ret;
}

static void mem_cgroup_move_charge(struct mm_struct *mm)
{
4945 4946 4947 4948
	struct mm_walk mem_cgroup_move_charge_walk = {
		.pmd_entry = mem_cgroup_move_charge_pte_range,
		.mm = mm,
	};
4949 4950

	lru_add_drain_all();
4951 4952 4953 4954 4955 4956 4957
	/*
	 * Signal mem_cgroup_begin_page_stat() to take the memcg's
	 * move_lock while we're moving its pages to another memcg.
	 * Then wait for already started RCU-only updates to finish.
	 */
	atomic_inc(&mc.from->moving_account);
	synchronize_rcu();
4958 4959 4960 4961 4962 4963 4964 4965 4966 4967 4968 4969 4970
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;
	}
4971 4972 4973 4974 4975
	/*
	 * When we have consumed all precharges and failed in doing
	 * additional charge, the page walk just aborts.
	 */
	walk_page_range(0, ~0UL, &mem_cgroup_move_charge_walk);
4976
	up_read(&mm->mmap_sem);
4977
	atomic_dec(&mc.from->moving_account);
4978 4979
}

4980
static void mem_cgroup_move_task(struct cgroup_taskset *tset)
B
Balbir Singh 已提交
4981
{
4982 4983
	struct cgroup_subsys_state *css;
	struct task_struct *p = cgroup_taskset_first(tset, &css);
4984
	struct mm_struct *mm = get_task_mm(p);
4985 4986

	if (mm) {
4987 4988
		if (mc.to)
			mem_cgroup_move_charge(mm);
4989 4990
		mmput(mm);
	}
4991 4992
	if (mc.to)
		mem_cgroup_clear_mc();
B
Balbir Singh 已提交
4993
}
4994
#else	/* !CONFIG_MMU */
4995
static int mem_cgroup_can_attach(struct cgroup_taskset *tset)
4996 4997 4998
{
	return 0;
}
4999
static void mem_cgroup_cancel_attach(struct cgroup_taskset *tset)
5000 5001
{
}
5002
static void mem_cgroup_move_task(struct cgroup_taskset *tset)
5003 5004 5005
{
}
#endif
B
Balbir Singh 已提交
5006

5007 5008
/*
 * Cgroup retains root cgroups across [un]mount cycles making it necessary
5009 5010
 * to verify whether we're attached to the default hierarchy on each mount
 * attempt.
5011
 */
5012
static void mem_cgroup_bind(struct cgroup_subsys_state *root_css)
5013 5014
{
	/*
5015
	 * use_hierarchy is forced on the default hierarchy.  cgroup core
5016 5017 5018
	 * guarantees that @root doesn't have any children, so turning it
	 * on for the root memcg is enough.
	 */
5019
	if (cgroup_subsys_on_dfl(memory_cgrp_subsys))
5020 5021 5022
		root_mem_cgroup->use_hierarchy = true;
	else
		root_mem_cgroup->use_hierarchy = false;
5023 5024
}

5025 5026 5027
static u64 memory_current_read(struct cgroup_subsys_state *css,
			       struct cftype *cft)
{
5028 5029 5030
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);

	return (u64)page_counter_read(&memcg->memory) * PAGE_SIZE;
5031 5032 5033 5034 5035
}

static int memory_low_show(struct seq_file *m, void *v)
{
	struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(m));
5036
	unsigned long low = READ_ONCE(memcg->low);
5037 5038

	if (low == PAGE_COUNTER_MAX)
5039
		seq_puts(m, "max\n");
5040 5041 5042 5043 5044 5045 5046 5047 5048 5049 5050 5051 5052 5053
	else
		seq_printf(m, "%llu\n", (u64)low * PAGE_SIZE);

	return 0;
}

static ssize_t memory_low_write(struct kernfs_open_file *of,
				char *buf, size_t nbytes, loff_t off)
{
	struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of));
	unsigned long low;
	int err;

	buf = strstrip(buf);
5054
	err = page_counter_memparse(buf, "max", &low);
5055 5056 5057 5058 5059 5060 5061 5062 5063 5064 5065
	if (err)
		return err;

	memcg->low = low;

	return nbytes;
}

static int memory_high_show(struct seq_file *m, void *v)
{
	struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(m));
5066
	unsigned long high = READ_ONCE(memcg->high);
5067 5068

	if (high == PAGE_COUNTER_MAX)
5069
		seq_puts(m, "max\n");
5070 5071 5072 5073 5074 5075 5076 5077 5078 5079 5080 5081 5082 5083
	else
		seq_printf(m, "%llu\n", (u64)high * PAGE_SIZE);

	return 0;
}

static ssize_t memory_high_write(struct kernfs_open_file *of,
				 char *buf, size_t nbytes, loff_t off)
{
	struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of));
	unsigned long high;
	int err;

	buf = strstrip(buf);
5084
	err = page_counter_memparse(buf, "max", &high);
5085 5086 5087 5088 5089
	if (err)
		return err;

	memcg->high = high;

5090
	memcg_wb_domain_size_changed(memcg);
5091 5092 5093 5094 5095 5096
	return nbytes;
}

static int memory_max_show(struct seq_file *m, void *v)
{
	struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(m));
5097
	unsigned long max = READ_ONCE(memcg->memory.limit);
5098 5099

	if (max == PAGE_COUNTER_MAX)
5100
		seq_puts(m, "max\n");
5101 5102 5103 5104 5105 5106 5107 5108 5109 5110 5111 5112 5113 5114
	else
		seq_printf(m, "%llu\n", (u64)max * PAGE_SIZE);

	return 0;
}

static ssize_t memory_max_write(struct kernfs_open_file *of,
				char *buf, size_t nbytes, loff_t off)
{
	struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of));
	unsigned long max;
	int err;

	buf = strstrip(buf);
5115
	err = page_counter_memparse(buf, "max", &max);
5116 5117 5118 5119 5120 5121 5122
	if (err)
		return err;

	err = mem_cgroup_resize_limit(memcg, max);
	if (err)
		return err;

5123
	memcg_wb_domain_size_changed(memcg);
5124 5125 5126 5127 5128 5129 5130 5131 5132 5133 5134 5135 5136 5137 5138 5139 5140 5141
	return nbytes;
}

static int memory_events_show(struct seq_file *m, void *v)
{
	struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(m));

	seq_printf(m, "low %lu\n", mem_cgroup_read_events(memcg, MEMCG_LOW));
	seq_printf(m, "high %lu\n", mem_cgroup_read_events(memcg, MEMCG_HIGH));
	seq_printf(m, "max %lu\n", mem_cgroup_read_events(memcg, MEMCG_MAX));
	seq_printf(m, "oom %lu\n", mem_cgroup_read_events(memcg, MEMCG_OOM));

	return 0;
}

static struct cftype memory_files[] = {
	{
		.name = "current",
5142
		.flags = CFTYPE_NOT_ON_ROOT,
5143 5144 5145 5146 5147 5148 5149 5150 5151 5152 5153 5154 5155 5156 5157 5158 5159 5160 5161 5162 5163 5164 5165
		.read_u64 = memory_current_read,
	},
	{
		.name = "low",
		.flags = CFTYPE_NOT_ON_ROOT,
		.seq_show = memory_low_show,
		.write = memory_low_write,
	},
	{
		.name = "high",
		.flags = CFTYPE_NOT_ON_ROOT,
		.seq_show = memory_high_show,
		.write = memory_high_write,
	},
	{
		.name = "max",
		.flags = CFTYPE_NOT_ON_ROOT,
		.seq_show = memory_max_show,
		.write = memory_max_write,
	},
	{
		.name = "events",
		.flags = CFTYPE_NOT_ON_ROOT,
5166
		.file_offset = offsetof(struct mem_cgroup, events_file),
5167 5168 5169 5170 5171
		.seq_show = memory_events_show,
	},
	{ }	/* terminate */
};

5172
struct cgroup_subsys memory_cgrp_subsys = {
5173
	.css_alloc = mem_cgroup_css_alloc,
5174
	.css_online = mem_cgroup_css_online,
5175
	.css_offline = mem_cgroup_css_offline,
5176
	.css_released = mem_cgroup_css_released,
5177
	.css_free = mem_cgroup_css_free,
5178
	.css_reset = mem_cgroup_css_reset,
5179 5180
	.can_attach = mem_cgroup_can_attach,
	.cancel_attach = mem_cgroup_cancel_attach,
B
Balbir Singh 已提交
5181
	.attach = mem_cgroup_move_task,
5182
	.bind = mem_cgroup_bind,
5183 5184
	.dfl_cftypes = memory_files,
	.legacy_cftypes = mem_cgroup_legacy_files,
5185
	.early_init = 0,
B
Balbir Singh 已提交
5186
};
5187

5188 5189 5190 5191 5192 5193 5194 5195 5196 5197 5198 5199 5200 5201 5202 5203 5204 5205 5206 5207 5208 5209
/**
 * mem_cgroup_low - check if memory consumption is below the normal range
 * @root: the highest ancestor to consider
 * @memcg: the memory cgroup to check
 *
 * Returns %true if memory consumption of @memcg, and that of all
 * configurable ancestors up to @root, is below the normal range.
 */
bool mem_cgroup_low(struct mem_cgroup *root, struct mem_cgroup *memcg)
{
	if (mem_cgroup_disabled())
		return false;

	/*
	 * The toplevel group doesn't have a configurable range, so
	 * it's never low when looked at directly, and it is not
	 * considered an ancestor when assessing the hierarchy.
	 */

	if (memcg == root_mem_cgroup)
		return false;

M
Michal Hocko 已提交
5210
	if (page_counter_read(&memcg->memory) >= memcg->low)
5211 5212 5213 5214 5215 5216 5217 5218
		return false;

	while (memcg != root) {
		memcg = parent_mem_cgroup(memcg);

		if (memcg == root_mem_cgroup)
			break;

M
Michal Hocko 已提交
5219
		if (page_counter_read(&memcg->memory) >= memcg->low)
5220 5221 5222 5223 5224
			return false;
	}
	return true;
}

5225 5226 5227 5228 5229 5230 5231 5232 5233 5234 5235 5236 5237 5238 5239 5240 5241 5242
/**
 * mem_cgroup_try_charge - try charging a page
 * @page: page to charge
 * @mm: mm context of the victim
 * @gfp_mask: reclaim mode
 * @memcgp: charged memcg return
 *
 * Try to charge @page to the memcg that @mm belongs to, reclaiming
 * pages according to @gfp_mask if necessary.
 *
 * Returns 0 on success, with *@memcgp pointing to the charged memcg.
 * Otherwise, an error code is returned.
 *
 * After page->mapping has been set up, the caller must finalize the
 * charge with mem_cgroup_commit_charge().  Or abort the transaction
 * with mem_cgroup_cancel_charge() in case page instantiation fails.
 */
int mem_cgroup_try_charge(struct page *page, struct mm_struct *mm,
5243 5244
			  gfp_t gfp_mask, struct mem_cgroup **memcgp,
			  bool compound)
5245 5246
{
	struct mem_cgroup *memcg = NULL;
5247
	unsigned int nr_pages = compound ? hpage_nr_pages(page) : 1;
5248 5249 5250 5251 5252 5253 5254 5255 5256 5257 5258 5259 5260
	int ret = 0;

	if (mem_cgroup_disabled())
		goto out;

	if (PageSwapCache(page)) {
		/*
		 * Every swap fault against a single page tries to charge the
		 * page, bail as early as possible.  shmem_unuse() encounters
		 * already charged pages, too.  The USED bit is protected by
		 * the page lock, which serializes swap cache removal, which
		 * in turn serializes uncharging.
		 */
5261
		VM_BUG_ON_PAGE(!PageLocked(page), page);
5262
		if (page->mem_cgroup)
5263
			goto out;
5264

5265
		if (do_memsw_account()) {
5266 5267 5268 5269 5270 5271 5272 5273 5274
			swp_entry_t ent = { .val = page_private(page), };
			unsigned short id = lookup_swap_cgroup_id(ent);

			rcu_read_lock();
			memcg = mem_cgroup_from_id(id);
			if (memcg && !css_tryget_online(&memcg->css))
				memcg = NULL;
			rcu_read_unlock();
		}
5275 5276 5277 5278 5279 5280 5281 5282 5283 5284 5285 5286 5287 5288 5289 5290 5291 5292 5293 5294 5295 5296 5297 5298 5299 5300 5301 5302 5303 5304
	}

	if (!memcg)
		memcg = get_mem_cgroup_from_mm(mm);

	ret = try_charge(memcg, gfp_mask, nr_pages);

	css_put(&memcg->css);
out:
	*memcgp = memcg;
	return ret;
}

/**
 * mem_cgroup_commit_charge - commit a page charge
 * @page: page to charge
 * @memcg: memcg to charge the page to
 * @lrucare: page might be on LRU already
 *
 * Finalize a charge transaction started by mem_cgroup_try_charge(),
 * after page->mapping has been set up.  This must happen atomically
 * as part of the page instantiation, i.e. under the page table lock
 * for anonymous pages, under the page lock for page and swap cache.
 *
 * In addition, the page must not be on the LRU during the commit, to
 * prevent racing with task migration.  If it might be, use @lrucare.
 *
 * Use mem_cgroup_cancel_charge() to cancel the transaction instead.
 */
void mem_cgroup_commit_charge(struct page *page, struct mem_cgroup *memcg,
5305
			      bool lrucare, bool compound)
5306
{
5307
	unsigned int nr_pages = compound ? hpage_nr_pages(page) : 1;
5308 5309 5310 5311 5312 5313 5314 5315 5316 5317 5318 5319 5320 5321

	VM_BUG_ON_PAGE(!page->mapping, page);
	VM_BUG_ON_PAGE(PageLRU(page) && !lrucare, page);

	if (mem_cgroup_disabled())
		return;
	/*
	 * Swap faults will attempt to charge the same page multiple
	 * times.  But reuse_swap_page() might have removed the page
	 * from swapcache already, so we can't check PageSwapCache().
	 */
	if (!memcg)
		return;

5322 5323 5324
	commit_charge(page, memcg, lrucare);

	local_irq_disable();
5325
	mem_cgroup_charge_statistics(memcg, page, compound, nr_pages);
5326 5327
	memcg_check_events(memcg, page);
	local_irq_enable();
5328

5329
	if (do_memsw_account() && PageSwapCache(page)) {
5330 5331 5332 5333 5334 5335 5336 5337 5338 5339 5340 5341 5342 5343 5344 5345 5346
		swp_entry_t entry = { .val = page_private(page) };
		/*
		 * The swap entry might not get freed for a long time,
		 * let's not wait for it.  The page already received a
		 * memory+swap charge, drop the swap entry duplicate.
		 */
		mem_cgroup_uncharge_swap(entry);
	}
}

/**
 * mem_cgroup_cancel_charge - cancel a page charge
 * @page: page to charge
 * @memcg: memcg to charge the page to
 *
 * Cancel a charge transaction started by mem_cgroup_try_charge().
 */
5347 5348
void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
		bool compound)
5349
{
5350
	unsigned int nr_pages = compound ? hpage_nr_pages(page) : 1;
5351 5352 5353 5354 5355 5356 5357 5358 5359 5360 5361 5362 5363 5364

	if (mem_cgroup_disabled())
		return;
	/*
	 * Swap faults will attempt to charge the same page multiple
	 * times.  But reuse_swap_page() might have removed the page
	 * from swapcache already, so we can't check PageSwapCache().
	 */
	if (!memcg)
		return;

	cancel_charge(memcg, nr_pages);
}

5365 5366 5367 5368
static void uncharge_batch(struct mem_cgroup *memcg, unsigned long pgpgout,
			   unsigned long nr_anon, unsigned long nr_file,
			   unsigned long nr_huge, struct page *dummy_page)
{
5369
	unsigned long nr_pages = nr_anon + nr_file;
5370 5371
	unsigned long flags;

5372
	if (!mem_cgroup_is_root(memcg)) {
5373
		page_counter_uncharge(&memcg->memory, nr_pages);
5374
		if (do_memsw_account())
5375
			page_counter_uncharge(&memcg->memsw, nr_pages);
5376 5377
		memcg_oom_recover(memcg);
	}
5378 5379 5380 5381 5382 5383

	local_irq_save(flags);
	__this_cpu_sub(memcg->stat->count[MEM_CGROUP_STAT_RSS], nr_anon);
	__this_cpu_sub(memcg->stat->count[MEM_CGROUP_STAT_CACHE], nr_file);
	__this_cpu_sub(memcg->stat->count[MEM_CGROUP_STAT_RSS_HUGE], nr_huge);
	__this_cpu_add(memcg->stat->events[MEM_CGROUP_EVENTS_PGPGOUT], pgpgout);
5384
	__this_cpu_add(memcg->stat->nr_page_events, nr_pages);
5385 5386
	memcg_check_events(memcg, dummy_page);
	local_irq_restore(flags);
5387 5388

	if (!mem_cgroup_is_root(memcg))
5389
		css_put_many(&memcg->css, nr_pages);
5390 5391 5392 5393 5394 5395 5396 5397 5398 5399 5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410 5411
}

static void uncharge_list(struct list_head *page_list)
{
	struct mem_cgroup *memcg = NULL;
	unsigned long nr_anon = 0;
	unsigned long nr_file = 0;
	unsigned long nr_huge = 0;
	unsigned long pgpgout = 0;
	struct list_head *next;
	struct page *page;

	next = page_list->next;
	do {
		unsigned int nr_pages = 1;

		page = list_entry(next, struct page, lru);
		next = page->lru.next;

		VM_BUG_ON_PAGE(PageLRU(page), page);
		VM_BUG_ON_PAGE(page_count(page), page);

5412
		if (!page->mem_cgroup)
5413 5414 5415 5416
			continue;

		/*
		 * Nobody should be changing or seriously looking at
5417
		 * page->mem_cgroup at this point, we have fully
5418
		 * exclusive access to the page.
5419 5420
		 */

5421
		if (memcg != page->mem_cgroup) {
5422
			if (memcg) {
5423 5424 5425
				uncharge_batch(memcg, pgpgout, nr_anon, nr_file,
					       nr_huge, page);
				pgpgout = nr_anon = nr_file = nr_huge = 0;
5426
			}
5427
			memcg = page->mem_cgroup;
5428 5429 5430 5431 5432 5433 5434 5435 5436 5437 5438 5439 5440
		}

		if (PageTransHuge(page)) {
			nr_pages <<= compound_order(page);
			VM_BUG_ON_PAGE(!PageTransHuge(page), page);
			nr_huge += nr_pages;
		}

		if (PageAnon(page))
			nr_anon += nr_pages;
		else
			nr_file += nr_pages;

5441
		page->mem_cgroup = NULL;
5442 5443 5444 5445 5446

		pgpgout++;
	} while (next != page_list);

	if (memcg)
5447 5448
		uncharge_batch(memcg, pgpgout, nr_anon, nr_file,
			       nr_huge, page);
5449 5450
}

5451 5452 5453 5454 5455 5456 5457 5458 5459 5460 5461 5462
/**
 * mem_cgroup_uncharge - uncharge a page
 * @page: page to uncharge
 *
 * Uncharge a page previously charged with mem_cgroup_try_charge() and
 * mem_cgroup_commit_charge().
 */
void mem_cgroup_uncharge(struct page *page)
{
	if (mem_cgroup_disabled())
		return;

5463
	/* Don't touch page->lru of any random page, pre-check: */
5464
	if (!page->mem_cgroup)
5465 5466
		return;

5467 5468 5469
	INIT_LIST_HEAD(&page->lru);
	uncharge_list(&page->lru);
}
5470

5471 5472 5473 5474 5475 5476 5477 5478 5479 5480 5481
/**
 * mem_cgroup_uncharge_list - uncharge a list of page
 * @page_list: list of pages to uncharge
 *
 * Uncharge a list of pages previously charged with
 * mem_cgroup_try_charge() and mem_cgroup_commit_charge().
 */
void mem_cgroup_uncharge_list(struct list_head *page_list)
{
	if (mem_cgroup_disabled())
		return;
5482

5483 5484
	if (!list_empty(page_list))
		uncharge_list(page_list);
5485 5486 5487
}

/**
5488
 * mem_cgroup_replace_page - migrate a charge to another page
5489 5490 5491 5492 5493 5494
 * @oldpage: currently charged page
 * @newpage: page to transfer the charge to
 *
 * Migrate the charge from @oldpage to @newpage.
 *
 * Both pages must be locked, @newpage->mapping must be set up.
5495
 * Either or both pages might be on the LRU already.
5496
 */
5497
void mem_cgroup_replace_page(struct page *oldpage, struct page *newpage)
5498
{
5499
	struct mem_cgroup *memcg;
5500 5501 5502 5503 5504
	int isolated;

	VM_BUG_ON_PAGE(!PageLocked(oldpage), oldpage);
	VM_BUG_ON_PAGE(!PageLocked(newpage), newpage);
	VM_BUG_ON_PAGE(PageAnon(oldpage) != PageAnon(newpage), newpage);
5505 5506
	VM_BUG_ON_PAGE(PageTransHuge(oldpage) != PageTransHuge(newpage),
		       newpage);
5507 5508 5509 5510 5511

	if (mem_cgroup_disabled())
		return;

	/* Page cache replacement: new page already charged? */
5512
	if (newpage->mem_cgroup)
5513 5514
		return;

5515
	/* Swapcache readahead pages can get replaced before being charged */
5516
	memcg = oldpage->mem_cgroup;
5517
	if (!memcg)
5518 5519
		return;

5520
	lock_page_lru(oldpage, &isolated);
5521
	oldpage->mem_cgroup = NULL;
5522
	unlock_page_lru(oldpage, isolated);
5523

5524
	commit_charge(newpage, memcg, true);
5525 5526
}

5527
#ifdef CONFIG_INET
5528

5529
DEFINE_STATIC_KEY_FALSE(memcg_sockets_enabled_key);
5530 5531 5532 5533 5534 5535 5536 5537 5538 5539 5540 5541 5542 5543 5544 5545 5546 5547 5548 5549 5550 5551
EXPORT_SYMBOL(memcg_sockets_enabled_key);

void sock_update_memcg(struct sock *sk)
{
	struct mem_cgroup *memcg;

	/* 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_memcg) {
		BUG_ON(mem_cgroup_is_root(sk->sk_memcg));
		css_get(&sk->sk_memcg->css);
		return;
	}

	rcu_read_lock();
	memcg = mem_cgroup_from_task(current);
5552 5553 5554 5555 5556 5557 5558
	if (memcg == root_mem_cgroup)
		goto out;
#ifdef CONFIG_MEMCG_KMEM
	if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && !memcg->tcp_mem.active)
		goto out;
#endif
	if (css_tryget_online(&memcg->css))
5559
		sk->sk_memcg = memcg;
5560
out:
5561 5562 5563 5564 5565 5566 5567 5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578 5579 5580
	rcu_read_unlock();
}
EXPORT_SYMBOL(sock_update_memcg);

void sock_release_memcg(struct sock *sk)
{
	WARN_ON(!sk->sk_memcg);
	css_put(&sk->sk_memcg->css);
}

/**
 * mem_cgroup_charge_skmem - charge socket memory
 * @memcg: memcg to charge
 * @nr_pages: number of pages to charge
 *
 * Charges @nr_pages to @memcg. Returns %true if the charge fit within
 * @memcg's configured limit, %false if the charge had to be forced.
 */
bool mem_cgroup_charge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages)
{
5581
	gfp_t gfp_mask = GFP_KERNEL;
5582

5583 5584 5585 5586 5587 5588 5589 5590 5591 5592 5593 5594
#ifdef CONFIG_MEMCG_KMEM
	if (!cgroup_subsys_on_dfl(memory_cgrp_subsys)) {
		struct page_counter *counter;

		if (page_counter_try_charge(&memcg->tcp_mem.memory_allocated,
					    nr_pages, &counter)) {
			memcg->tcp_mem.memory_pressure = 0;
			return true;
		}
		page_counter_charge(&memcg->tcp_mem.memory_allocated, nr_pages);
		memcg->tcp_mem.memory_pressure = 1;
		return false;
5595
	}
5596 5597 5598 5599 5600 5601 5602 5603 5604
#endif
	/* Don't block in the packet receive path */
	if (in_softirq())
		gfp_mask = GFP_NOWAIT;

	if (try_charge(memcg, gfp_mask, nr_pages) == 0)
		return true;

	try_charge(memcg, gfp_mask|__GFP_NOFAIL, nr_pages);
5605 5606 5607 5608 5609 5610 5611 5612 5613 5614
	return false;
}

/**
 * mem_cgroup_uncharge_skmem - uncharge socket memory
 * @memcg - memcg to uncharge
 * @nr_pages - number of pages to uncharge
 */
void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages)
{
5615 5616 5617 5618 5619 5620 5621 5622 5623
#ifdef CONFIG_MEMCG_KMEM
	if (!cgroup_subsys_on_dfl(memory_cgrp_subsys)) {
		page_counter_uncharge(&memcg->tcp_mem.memory_allocated,
				      nr_pages);
		return;
	}
#endif
	page_counter_uncharge(&memcg->memory, nr_pages);
	css_put_many(&memcg->css, nr_pages);
5624 5625
}

5626 5627 5628 5629 5630 5631 5632 5633 5634 5635 5636 5637 5638 5639 5640
#endif /* CONFIG_INET */

static int __init cgroup_memory(char *s)
{
	char *token;

	while ((token = strsep(&s, ",")) != NULL) {
		if (!*token)
			continue;
		if (!strcmp(token, "nosocket"))
			cgroup_memory_nosocket = true;
	}
	return 0;
}
__setup("cgroup.memory=", cgroup_memory);
5641

5642
/*
5643 5644 5645 5646 5647 5648
 * subsys_initcall() for memory controller.
 *
 * Some parts like hotcpu_notifier() have to be initialized from this context
 * because of lock dependencies (cgroup_lock -> cpu hotplug) but basically
 * everything that doesn't depend on a specific mem_cgroup structure should
 * be initialized from here.
5649 5650 5651
 */
static int __init mem_cgroup_init(void)
{
5652 5653
	int cpu, node;

5654
	hotcpu_notifier(memcg_cpu_hotplug_callback, 0);
5655 5656 5657 5658 5659 5660 5661 5662 5663 5664 5665 5666 5667 5668 5669 5670 5671 5672 5673 5674 5675 5676

	for_each_possible_cpu(cpu)
		INIT_WORK(&per_cpu_ptr(&memcg_stock, cpu)->work,
			  drain_local_stock);

	for_each_node(node) {
		struct mem_cgroup_tree_per_node *rtpn;
		int zone;

		rtpn = kzalloc_node(sizeof(*rtpn), GFP_KERNEL,
				    node_online(node) ? node : NUMA_NO_NODE);

		for (zone = 0; zone < MAX_NR_ZONES; zone++) {
			struct mem_cgroup_tree_per_zone *rtpz;

			rtpz = &rtpn->rb_tree_per_zone[zone];
			rtpz->rb_root = RB_ROOT;
			spin_lock_init(&rtpz->lock);
		}
		soft_limit_tree.rb_tree_per_node[node] = rtpn;
	}

5677 5678 5679
	return 0;
}
subsys_initcall(mem_cgroup_init);
5680 5681 5682 5683 5684 5685 5686 5687 5688 5689 5690 5691 5692 5693 5694 5695 5696

#ifdef CONFIG_MEMCG_SWAP
/**
 * mem_cgroup_swapout - transfer a memsw charge to swap
 * @page: page whose memsw charge to transfer
 * @entry: swap entry to move the charge to
 *
 * Transfer the memsw charge of @page to @entry.
 */
void mem_cgroup_swapout(struct page *page, swp_entry_t entry)
{
	struct mem_cgroup *memcg;
	unsigned short oldid;

	VM_BUG_ON_PAGE(PageLRU(page), page);
	VM_BUG_ON_PAGE(page_count(page), page);

5697
	if (!do_memsw_account())
5698 5699 5700 5701 5702 5703 5704 5705 5706 5707 5708 5709 5710 5711 5712 5713 5714
		return;

	memcg = page->mem_cgroup;

	/* Readahead page, never charged */
	if (!memcg)
		return;

	oldid = swap_cgroup_record(entry, mem_cgroup_id(memcg));
	VM_BUG_ON_PAGE(oldid, page);
	mem_cgroup_swap_statistics(memcg, true);

	page->mem_cgroup = NULL;

	if (!mem_cgroup_is_root(memcg))
		page_counter_uncharge(&memcg->memory, 1);

5715 5716 5717 5718 5719 5720 5721
	/*
	 * Interrupts should be disabled here because the caller holds the
	 * mapping->tree_lock lock which is taken with interrupts-off. It is
	 * important here to have the interrupts disabled because it is the
	 * only synchronisation we have for udpating the per-CPU variables.
	 */
	VM_BUG_ON(!irqs_disabled());
5722
	mem_cgroup_charge_statistics(memcg, page, false, -1);
5723 5724 5725 5726 5727 5728 5729 5730 5731 5732 5733 5734 5735 5736
	memcg_check_events(memcg, page);
}

/**
 * mem_cgroup_uncharge_swap - uncharge a swap entry
 * @entry: swap entry to uncharge
 *
 * Drop the memsw charge associated with @entry.
 */
void mem_cgroup_uncharge_swap(swp_entry_t entry)
{
	struct mem_cgroup *memcg;
	unsigned short id;

5737
	if (!do_memsw_account())
5738 5739 5740 5741
		return;

	id = swap_cgroup_record(entry, 0);
	rcu_read_lock();
5742
	memcg = mem_cgroup_from_id(id);
5743 5744 5745 5746 5747 5748 5749 5750 5751 5752 5753 5754 5755 5756 5757 5758 5759 5760 5761 5762 5763 5764 5765 5766 5767 5768 5769 5770 5771 5772 5773 5774 5775 5776 5777 5778 5779 5780 5781 5782 5783 5784 5785 5786 5787 5788 5789 5790 5791 5792 5793 5794 5795 5796 5797 5798 5799 5800 5801 5802 5803 5804 5805 5806 5807
	if (memcg) {
		if (!mem_cgroup_is_root(memcg))
			page_counter_uncharge(&memcg->memsw, 1);
		mem_cgroup_swap_statistics(memcg, false);
		css_put(&memcg->css);
	}
	rcu_read_unlock();
}

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

static int __init enable_swap_account(char *s)
{
	if (!strcmp(s, "1"))
		really_do_swap_account = 1;
	else if (!strcmp(s, "0"))
		really_do_swap_account = 0;
	return 1;
}
__setup("swapaccount=", enable_swap_account);

static struct cftype memsw_cgroup_files[] = {
	{
		.name = "memsw.usage_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_USAGE),
		.read_u64 = mem_cgroup_read_u64,
	},
	{
		.name = "memsw.max_usage_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_MAX_USAGE),
		.write = mem_cgroup_reset,
		.read_u64 = mem_cgroup_read_u64,
	},
	{
		.name = "memsw.limit_in_bytes",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_LIMIT),
		.write = mem_cgroup_write,
		.read_u64 = mem_cgroup_read_u64,
	},
	{
		.name = "memsw.failcnt",
		.private = MEMFILE_PRIVATE(_MEMSWAP, RES_FAILCNT),
		.write = mem_cgroup_reset,
		.read_u64 = mem_cgroup_read_u64,
	},
	{ },	/* terminate */
};

static int __init mem_cgroup_swap_init(void)
{
	if (!mem_cgroup_disabled() && really_do_swap_account) {
		do_swap_account = 1;
		WARN_ON(cgroup_add_legacy_cftypes(&memory_cgrp_subsys,
						  memsw_cgroup_files));
	}
	return 0;
}
subsys_initcall(mem_cgroup_swap_init);

#endif /* CONFIG_MEMCG_SWAP */