memcontrol.c 155.5 KB
Newer Older
B
Balbir Singh 已提交
1 2 3 4 5
/* memcontrol.c - Memory Controller
 *
 * Copyright IBM Corporation, 2007
 * Author Balbir Singh <balbir@linux.vnet.ibm.com>
 *
6 7 8
 * Copyright 2007 OpenVZ SWsoft Inc
 * Author: Pavel Emelianov <xemul@openvz.org>
 *
9 10 11 12
 * Memory thresholds
 * Copyright (C) 2009 Nokia Corporation
 * Author: Kirill A. Shutemov
 *
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/sched/mm.h>
39
#include <linux/shmem_fs.h>
40
#include <linux/hugetlb.h>
K
KAMEZAWA Hiroyuki 已提交
41
#include <linux/pagemap.h>
42
#include <linux/smp.h>
43
#include <linux/page-flags.h>
44
#include <linux/backing-dev.h>
45 46
#include <linux/bit_spinlock.h>
#include <linux/rcupdate.h>
47
#include <linux/limits.h>
48
#include <linux/export.h>
49
#include <linux/mutex.h>
50
#include <linux/rbtree.h>
51
#include <linux/slab.h>
52
#include <linux/swap.h>
53
#include <linux/swapops.h>
54
#include <linux/spinlock.h>
55
#include <linux/eventfd.h>
56
#include <linux/poll.h>
57
#include <linux/sort.h>
58
#include <linux/fs.h>
59
#include <linux/seq_file.h>
60
#include <linux/vmpressure.h>
61
#include <linux/mm_inline.h>
62
#include <linux/swap_cgroup.h>
63
#include <linux/cpu.h>
64
#include <linux/oom.h>
65
#include <linux/lockdep.h>
66
#include <linux/file.h>
67
#include <linux/tracehook.h>
K
KAMEZAWA Hiroyuki 已提交
68
#include "internal.h"
G
Glauber Costa 已提交
69
#include <net/sock.h>
M
Michal Hocko 已提交
70
#include <net/ip.h>
71
#include "slab.h"
B
Balbir Singh 已提交
72

73
#include <linux/uaccess.h>
74

75 76
#include <trace/events/vmscan.h>

77 78
struct cgroup_subsys memory_cgrp_subsys __read_mostly;
EXPORT_SYMBOL(memory_cgrp_subsys);
79

80 81
struct mem_cgroup *root_mem_cgroup __read_mostly;

82
#define MEM_CGROUP_RECLAIM_RETRIES	5
B
Balbir Singh 已提交
83

84 85 86
/* Socket memory accounting disabled? */
static bool cgroup_memory_nosocket;

87 88 89
/* Kernel memory accounting disabled? */
static bool cgroup_memory_nokmem;

90
/* Whether the swap controller is active */
A
Andrew Morton 已提交
91
#ifdef CONFIG_MEMCG_SWAP
92 93
int do_swap_account __read_mostly;
#else
94
#define do_swap_account		0
95 96
#endif

97 98 99 100 101 102
/* Whether legacy memory+swap accounting is active */
static bool do_memsw_account(void)
{
	return !cgroup_subsys_on_dfl(memory_cgrp_subsys) && do_swap_account;
}

103
static const char *const mem_cgroup_lru_names[] = {
104 105 106 107 108 109 110
	"inactive_anon",
	"active_anon",
	"inactive_file",
	"active_file",
	"unevictable",
};

111 112 113
#define THRESHOLDS_EVENTS_TARGET 128
#define SOFTLIMIT_EVENTS_TARGET 1024
#define NUMAINFO_EVENTS_TARGET	1024
114

115 116 117 118 119
/*
 * Cgroups above their limits are maintained in a RB-Tree, independent of
 * their hierarchy representation
 */

120
struct mem_cgroup_tree_per_node {
121 122 123 124 125 126 127 128 129 130
	struct rb_root rb_root;
	spinlock_t lock;
};

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
131 132 133 134 135
/* for OOM */
struct mem_cgroup_eventfd_list {
	struct list_head list;
	struct eventfd_ctx *eventfd;
};
136

137 138 139
/*
 * cgroup_event represents events which userspace want to receive.
 */
140
struct mem_cgroup_event {
141
	/*
142
	 * memcg which the event belongs to.
143
	 */
144
	struct mem_cgroup *memcg;
145 146 147 148 149 150 151 152
	/*
	 * 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;
153 154 155 156 157
	/*
	 * 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.
	 */
158
	int (*register_event)(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
159
			      struct eventfd_ctx *eventfd, const char *args);
160 161 162 163 164
	/*
	 * 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.
	 */
165
	void (*unregister_event)(struct mem_cgroup *memcg,
166
				 struct eventfd_ctx *eventfd);
167 168 169 170 171 172
	/*
	 * All fields below needed to unregister event when
	 * userspace closes eventfd.
	 */
	poll_table pt;
	wait_queue_head_t *wqh;
173
	wait_queue_entry_t wait;
174 175 176
	struct work_struct remove;
};

177 178
static void mem_cgroup_threshold(struct mem_cgroup *memcg);
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg);
179

180 181
/* Stuffs for move charges at task migration. */
/*
182
 * Types of charges to be moved.
183
 */
184 185 186
#define MOVE_ANON	0x1U
#define MOVE_FILE	0x2U
#define MOVE_MASK	(MOVE_ANON | MOVE_FILE)
187

188 189
/* "mc" and its members are protected by cgroup_mutex */
static struct move_charge_struct {
190
	spinlock_t	  lock; /* for from, to */
191
	struct mm_struct  *mm;
192 193
	struct mem_cgroup *from;
	struct mem_cgroup *to;
194
	unsigned long flags;
195
	unsigned long precharge;
196
	unsigned long moved_charge;
197
	unsigned long moved_swap;
198 199 200
	struct task_struct *moving_task;	/* a task moving charges */
	wait_queue_head_t waitq;		/* a waitq for other context */
} mc = {
201
	.lock = __SPIN_LOCK_UNLOCKED(mc.lock),
202 203
	.waitq = __WAIT_QUEUE_HEAD_INITIALIZER(mc.waitq),
};
204

205 206 207 208
/*
 * Maximum loops in mem_cgroup_hierarchical_reclaim(), used for soft
 * limit reclaim to prevent infinite loops, if they ever occur.
 */
209
#define	MEM_CGROUP_MAX_RECLAIM_LOOPS		100
210
#define	MEM_CGROUP_MAX_SOFT_LIMIT_RECLAIM_LOOPS	2
211

212 213
enum charge_type {
	MEM_CGROUP_CHARGE_TYPE_CACHE = 0,
214
	MEM_CGROUP_CHARGE_TYPE_ANON,
K
KAMEZAWA Hiroyuki 已提交
215
	MEM_CGROUP_CHARGE_TYPE_SWAPOUT,	/* for accounting swapcache */
K
KAMEZAWA Hiroyuki 已提交
216
	MEM_CGROUP_CHARGE_TYPE_DROP,	/* a page was unused swap cache */
217 218 219
	NR_CHARGE_TYPE,
};

220
/* for encoding cft->private value on file */
G
Glauber Costa 已提交
221 222 223 224
enum res_type {
	_MEM,
	_MEMSWAP,
	_OOM_TYPE,
225
	_KMEM,
V
Vladimir Davydov 已提交
226
	_TCP,
G
Glauber Costa 已提交
227 228
};

229 230
#define MEMFILE_PRIVATE(x, val)	((x) << 16 | (val))
#define MEMFILE_TYPE(val)	((val) >> 16 & 0xffff)
231
#define MEMFILE_ATTR(val)	((val) & 0xffff)
K
KAMEZAWA Hiroyuki 已提交
232 233
/* Used for OOM nofiier */
#define OOM_CONTROL		(0)
234

235 236 237 238 239 240 241 242 243 244 245 246 247
/* 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;
}

248 249 250 251 252
static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
{
	return (memcg == root_mem_cgroup);
}

253
#ifndef CONFIG_SLOB
254
/*
255
 * This will be the memcg's index in each cache's ->memcg_params.memcg_caches.
L
Li Zefan 已提交
256 257 258 259 260
 * 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.
261
 *
262 263
 * The current size of the caches array is stored in memcg_nr_cache_ids. It
 * will double each time we have to increase it.
264
 */
265 266
static DEFINE_IDA(memcg_cache_ida);
int memcg_nr_cache_ids;
267

268 269 270 271 272 273 274 275 276 277 278 279 280
/* 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);
}

281 282 283 284 285 286
/*
 * 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
287
 * MAX_SIZE should be as large as the number of cgrp_ids. Ideally, we could get
288 289
 * 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
290
 * cgrp_id space is not getting any smaller, and we don't have to necessarily
291 292 293
 * increase ours as well if it increases.
 */
#define MEMCG_CACHES_MIN_SIZE 4
L
Li Zefan 已提交
294
#define MEMCG_CACHES_MAX_SIZE MEM_CGROUP_ID_MAX
295

296 297 298 299 300 301
/*
 * 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
 */
302
DEFINE_STATIC_KEY_FALSE(memcg_kmem_enabled_key);
303
EXPORT_SYMBOL(memcg_kmem_enabled_key);
304

305 306
struct workqueue_struct *memcg_kmem_cache_wq;

307
#endif /* !CONFIG_SLOB */
308

309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325
/**
 * 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.
 */
struct cgroup_subsys_state *mem_cgroup_css_from_page(struct page *page)
{
	struct mem_cgroup *memcg;

	memcg = page->mem_cgroup;

326
	if (!memcg || !cgroup_subsys_on_dfl(memory_cgrp_subsys))
327 328 329 330 331
		memcg = root_mem_cgroup;

	return &memcg->css;
}

332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359
/**
 * 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;
}

360 361
static struct mem_cgroup_per_node *
mem_cgroup_page_nodeinfo(struct mem_cgroup *memcg, struct page *page)
362
{
363
	int nid = page_to_nid(page);
364

365
	return memcg->nodeinfo[nid];
366 367
}

368 369
static struct mem_cgroup_tree_per_node *
soft_limit_tree_node(int nid)
370
{
371
	return soft_limit_tree.rb_tree_per_node[nid];
372 373
}

374
static struct mem_cgroup_tree_per_node *
375 376 377 378
soft_limit_tree_from_page(struct page *page)
{
	int nid = page_to_nid(page);

379
	return soft_limit_tree.rb_tree_per_node[nid];
380 381
}

382 383
static void __mem_cgroup_insert_exceeded(struct mem_cgroup_per_node *mz,
					 struct mem_cgroup_tree_per_node *mctz,
384
					 unsigned long new_usage_in_excess)
385 386 387
{
	struct rb_node **p = &mctz->rb_root.rb_node;
	struct rb_node *parent = NULL;
388
	struct mem_cgroup_per_node *mz_node;
389 390 391 392 393 394 395 396 397

	if (mz->on_tree)
		return;

	mz->usage_in_excess = new_usage_in_excess;
	if (!mz->usage_in_excess)
		return;
	while (*p) {
		parent = *p;
398
		mz_node = rb_entry(parent, struct mem_cgroup_per_node,
399 400 401 402 403 404 405 406 407 408 409 410 411 412 413
					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;
}

414 415
static void __mem_cgroup_remove_exceeded(struct mem_cgroup_per_node *mz,
					 struct mem_cgroup_tree_per_node *mctz)
416 417 418 419 420 421 422
{
	if (!mz->on_tree)
		return;
	rb_erase(&mz->tree_node, &mctz->rb_root);
	mz->on_tree = false;
}

423 424
static void mem_cgroup_remove_exceeded(struct mem_cgroup_per_node *mz,
				       struct mem_cgroup_tree_per_node *mctz)
425
{
426 427 428
	unsigned long flags;

	spin_lock_irqsave(&mctz->lock, flags);
429
	__mem_cgroup_remove_exceeded(mz, mctz);
430
	spin_unlock_irqrestore(&mctz->lock, flags);
431 432
}

433 434 435
static unsigned long soft_limit_excess(struct mem_cgroup *memcg)
{
	unsigned long nr_pages = page_counter_read(&memcg->memory);
436
	unsigned long soft_limit = READ_ONCE(memcg->soft_limit);
437 438 439 440 441 442 443
	unsigned long excess = 0;

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

	return excess;
}
444 445 446

static void mem_cgroup_update_tree(struct mem_cgroup *memcg, struct page *page)
{
447
	unsigned long excess;
448 449
	struct mem_cgroup_per_node *mz;
	struct mem_cgroup_tree_per_node *mctz;
450

451
	mctz = soft_limit_tree_from_page(page);
452 453
	if (!mctz)
		return;
454 455 456 457 458
	/*
	 * Necessary to update all ancestors when hierarchy is used.
	 * because their event counter is not touched.
	 */
	for (; memcg; memcg = parent_mem_cgroup(memcg)) {
459
		mz = mem_cgroup_page_nodeinfo(memcg, page);
460
		excess = soft_limit_excess(memcg);
461 462 463 464 465
		/*
		 * We have to update the tree if mz is on RB-tree or
		 * mem is over its softlimit.
		 */
		if (excess || mz->on_tree) {
466 467 468
			unsigned long flags;

			spin_lock_irqsave(&mctz->lock, flags);
469 470
			/* if on-tree, remove it */
			if (mz->on_tree)
471
				__mem_cgroup_remove_exceeded(mz, mctz);
472 473 474 475
			/*
			 * Insert again. mz->usage_in_excess will be updated.
			 * If excess is 0, no tree ops.
			 */
476
			__mem_cgroup_insert_exceeded(mz, mctz, excess);
477
			spin_unlock_irqrestore(&mctz->lock, flags);
478 479 480 481 482 483
		}
	}
}

static void mem_cgroup_remove_from_trees(struct mem_cgroup *memcg)
{
484 485 486
	struct mem_cgroup_tree_per_node *mctz;
	struct mem_cgroup_per_node *mz;
	int nid;
487

488
	for_each_node(nid) {
489 490
		mz = mem_cgroup_nodeinfo(memcg, nid);
		mctz = soft_limit_tree_node(nid);
491 492
		if (mctz)
			mem_cgroup_remove_exceeded(mz, mctz);
493 494 495
	}
}

496 497
static struct mem_cgroup_per_node *
__mem_cgroup_largest_soft_limit_node(struct mem_cgroup_tree_per_node *mctz)
498 499
{
	struct rb_node *rightmost = NULL;
500
	struct mem_cgroup_per_node *mz;
501 502 503 504 505 506 507

retry:
	mz = NULL;
	rightmost = rb_last(&mctz->rb_root);
	if (!rightmost)
		goto done;		/* Nothing to reclaim from */

508
	mz = rb_entry(rightmost, struct mem_cgroup_per_node, tree_node);
509 510 511 512 513
	/*
	 * 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.
	 */
514
	__mem_cgroup_remove_exceeded(mz, mctz);
515
	if (!soft_limit_excess(mz->memcg) ||
516
	    !css_tryget_online(&mz->memcg->css))
517 518 519 520 521
		goto retry;
done:
	return mz;
}

522 523
static struct mem_cgroup_per_node *
mem_cgroup_largest_soft_limit_node(struct mem_cgroup_tree_per_node *mctz)
524
{
525
	struct mem_cgroup_per_node *mz;
526

527
	spin_lock_irq(&mctz->lock);
528
	mz = __mem_cgroup_largest_soft_limit_node(mctz);
529
	spin_unlock_irq(&mctz->lock);
530 531 532
	return mz;
}

533
/*
534 535
 * Return page count for single (non recursive) @memcg.
 *
536 537 538 539 540
 * 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
541
 * a periodic synchronization of counter in memcg's counter.
542 543 544 545 546 547 548 549 550
 *
 * 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
551
 * common workload, threshold and synchronization as vmstat[] should be
552 553
 * implemented.
 */
554

555 556
static unsigned long memcg_sum_events(struct mem_cgroup *memcg,
				      enum memcg_event_item event)
557 558 559 560
{
	unsigned long val = 0;
	int cpu;

561
	for_each_possible_cpu(cpu)
562
		val += per_cpu(memcg->stat->events[event], cpu);
563 564 565
	return val;
}

566
static void mem_cgroup_charge_statistics(struct mem_cgroup *memcg,
567
					 struct page *page,
568
					 bool compound, int nr_pages)
569
{
570 571 572 573
	/*
	 * Here, RSS means 'mapped anon' and anon's SwapCache. Shmem/tmpfs is
	 * counted as CACHE even if it's on ANON LRU.
	 */
574
	if (PageAnon(page))
575
		__this_cpu_add(memcg->stat->count[MEMCG_RSS], nr_pages);
576
	else {
577
		__this_cpu_add(memcg->stat->count[MEMCG_CACHE], nr_pages);
578
		if (PageSwapBacked(page))
579
			__this_cpu_add(memcg->stat->count[NR_SHMEM], nr_pages);
580
	}
581

582 583
	if (compound) {
		VM_BUG_ON_PAGE(!PageTransHuge(page), page);
584
		__this_cpu_add(memcg->stat->count[MEMCG_RSS_HUGE], nr_pages);
585
	}
586

587 588
	/* pagein of a big page is an event. So, ignore page size */
	if (nr_pages > 0)
589
		__this_cpu_inc(memcg->stat->events[PGPGIN]);
590
	else {
591
		__this_cpu_inc(memcg->stat->events[PGPGOUT]);
592 593
		nr_pages = -nr_pages; /* for event */
	}
594

595
	__this_cpu_add(memcg->stat->nr_page_events, nr_pages);
596 597
}

598 599
unsigned long mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
					   int nid, unsigned int lru_mask)
600
{
601
	struct lruvec *lruvec = mem_cgroup_lruvec(NODE_DATA(nid), memcg);
602
	unsigned long nr = 0;
603
	enum lru_list lru;
604

605
	VM_BUG_ON((unsigned)nid >= nr_node_ids);
606

607 608 609
	for_each_lru(lru) {
		if (!(BIT(lru) & lru_mask))
			continue;
610
		nr += mem_cgroup_get_lru_size(lruvec, lru);
611 612
	}
	return nr;
613
}
614

615
static unsigned long mem_cgroup_nr_lru_pages(struct mem_cgroup *memcg,
616
			unsigned int lru_mask)
617
{
618
	unsigned long nr = 0;
619
	int nid;
620

621
	for_each_node_state(nid, N_MEMORY)
622 623
		nr += mem_cgroup_node_nr_lru_pages(memcg, nid, lru_mask);
	return nr;
624 625
}

626 627
static bool mem_cgroup_event_ratelimit(struct mem_cgroup *memcg,
				       enum mem_cgroup_events_target target)
628 629 630
{
	unsigned long val, next;

631
	val = __this_cpu_read(memcg->stat->nr_page_events);
632
	next = __this_cpu_read(memcg->stat->targets[target]);
633
	/* from time_after() in jiffies.h */
634 635 636 637 638
	if ((long)next - (long)val < 0) {
		switch (target) {
		case MEM_CGROUP_TARGET_THRESH:
			next = val + THRESHOLDS_EVENTS_TARGET;
			break;
639 640 641
		case MEM_CGROUP_TARGET_SOFTLIMIT:
			next = val + SOFTLIMIT_EVENTS_TARGET;
			break;
642 643 644 645 646 647 648 649
		case MEM_CGROUP_TARGET_NUMAINFO:
			next = val + NUMAINFO_EVENTS_TARGET;
			break;
		default:
			break;
		}
		__this_cpu_write(memcg->stat->targets[target], next);
		return true;
650
	}
651
	return false;
652 653 654 655 656 657
}

/*
 * Check events in order.
 *
 */
658
static void memcg_check_events(struct mem_cgroup *memcg, struct page *page)
659 660
{
	/* threshold event is triggered in finer grain than soft limit */
661 662
	if (unlikely(mem_cgroup_event_ratelimit(memcg,
						MEM_CGROUP_TARGET_THRESH))) {
663
		bool do_softlimit;
664
		bool do_numainfo __maybe_unused;
665

666 667
		do_softlimit = mem_cgroup_event_ratelimit(memcg,
						MEM_CGROUP_TARGET_SOFTLIMIT);
668 669 670 671
#if MAX_NUMNODES > 1
		do_numainfo = mem_cgroup_event_ratelimit(memcg,
						MEM_CGROUP_TARGET_NUMAINFO);
#endif
672
		mem_cgroup_threshold(memcg);
673 674
		if (unlikely(do_softlimit))
			mem_cgroup_update_tree(memcg, page);
675
#if MAX_NUMNODES > 1
676
		if (unlikely(do_numainfo))
677
			atomic_inc(&memcg->numainfo_events);
678
#endif
679
	}
680 681
}

682
struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p)
683
{
684 685 686 687 688 689 690 691
	/*
	 * 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;

692
	return mem_cgroup_from_css(task_css(p, memory_cgrp_id));
693
}
M
Michal Hocko 已提交
694
EXPORT_SYMBOL(mem_cgroup_from_task);
695

696
static struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
697
{
698
	struct mem_cgroup *memcg = NULL;
699

700 701
	rcu_read_lock();
	do {
702 703 704 705 706 707
		/*
		 * 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))
708
			memcg = root_mem_cgroup;
709 710 711 712 713
		else {
			memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
			if (unlikely(!memcg))
				memcg = root_mem_cgroup;
		}
714
	} while (!css_tryget_online(&memcg->css));
715
	rcu_read_unlock();
716
	return memcg;
717 718
}

719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735
/**
 * 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.
 */
736
struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *root,
737
				   struct mem_cgroup *prev,
738
				   struct mem_cgroup_reclaim_cookie *reclaim)
K
KAMEZAWA Hiroyuki 已提交
739
{
M
Michal Hocko 已提交
740
	struct mem_cgroup_reclaim_iter *uninitialized_var(iter);
741
	struct cgroup_subsys_state *css = NULL;
742
	struct mem_cgroup *memcg = NULL;
743
	struct mem_cgroup *pos = NULL;
744

745 746
	if (mem_cgroup_disabled())
		return NULL;
747

748 749
	if (!root)
		root = root_mem_cgroup;
K
KAMEZAWA Hiroyuki 已提交
750

751
	if (prev && !reclaim)
752
		pos = prev;
K
KAMEZAWA Hiroyuki 已提交
753

754 755
	if (!root->use_hierarchy && root != root_mem_cgroup) {
		if (prev)
756
			goto out;
757
		return root;
758
	}
K
KAMEZAWA Hiroyuki 已提交
759

760
	rcu_read_lock();
M
Michal Hocko 已提交
761

762
	if (reclaim) {
763
		struct mem_cgroup_per_node *mz;
764

765
		mz = mem_cgroup_nodeinfo(root, reclaim->pgdat->node_id);
766 767 768 769 770
		iter = &mz->iter[reclaim->priority];

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

771
		while (1) {
772
			pos = READ_ONCE(iter->position);
773 774
			if (!pos || css_tryget(&pos->css))
				break;
775
			/*
776 777 778 779 780 781
			 * 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.
782
			 */
783 784
			(void)cmpxchg(&iter->position, pos, NULL);
		}
785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801
	}

	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;
802
		}
K
KAMEZAWA Hiroyuki 已提交
803

804 805 806 807 808 809
		/*
		 * 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
810

811 812
		if (css == &root->css)
			break;
K
KAMEZAWA Hiroyuki 已提交
813

814 815
		if (css_tryget(css))
			break;
816

817
		memcg = NULL;
818
	}
819 820 821

	if (reclaim) {
		/*
822 823 824
		 * 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.
825
		 */
826 827
		(void)cmpxchg(&iter->position, pos, memcg);

828 829 830 831 832 833 834
		if (pos)
			css_put(&pos->css);

		if (!memcg)
			iter->generation++;
		else if (!prev)
			reclaim->generation = iter->generation;
835
	}
836

837 838
out_unlock:
	rcu_read_unlock();
839
out:
840 841 842
	if (prev && prev != root)
		css_put(&prev->css);

843
	return memcg;
K
KAMEZAWA Hiroyuki 已提交
844
}
K
KAMEZAWA Hiroyuki 已提交
845

846 847 848 849 850 851 852
/**
 * 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)
853 854 855 856 857 858
{
	if (!root)
		root = root_mem_cgroup;
	if (prev && prev != root)
		css_put(&prev->css);
}
K
KAMEZAWA Hiroyuki 已提交
859

860 861 862 863
static void invalidate_reclaim_iterators(struct mem_cgroup *dead_memcg)
{
	struct mem_cgroup *memcg = dead_memcg;
	struct mem_cgroup_reclaim_iter *iter;
864 865
	struct mem_cgroup_per_node *mz;
	int nid;
866 867 868 869
	int i;

	while ((memcg = parent_mem_cgroup(memcg))) {
		for_each_node(nid) {
870 871 872 873 874
			mz = mem_cgroup_nodeinfo(memcg, nid);
			for (i = 0; i <= DEF_PRIORITY; i++) {
				iter = &mz->iter[i];
				cmpxchg(&iter->position,
					dead_memcg, NULL);
875 876 877 878 879
			}
		}
	}
}

880 881 882 883 884 885
/*
 * 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)		\
886
	for (iter = mem_cgroup_iter(root, NULL, NULL);	\
887
	     iter != NULL;				\
888
	     iter = mem_cgroup_iter(root, iter, NULL))
889

890
#define for_each_mem_cgroup(iter)			\
891
	for (iter = mem_cgroup_iter(NULL, NULL, NULL);	\
892
	     iter != NULL;				\
893
	     iter = mem_cgroup_iter(NULL, iter, NULL))
K
KAMEZAWA Hiroyuki 已提交
894

895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931
/**
 * mem_cgroup_scan_tasks - iterate over tasks of a memory cgroup hierarchy
 * @memcg: hierarchy root
 * @fn: function to call for each task
 * @arg: argument passed to @fn
 *
 * This function iterates over tasks attached to @memcg or to any of its
 * descendants and calls @fn for each task. If @fn returns a non-zero
 * value, the function breaks the iteration loop and returns the value.
 * Otherwise, it will iterate over all tasks and return 0.
 *
 * This function must not be called for the root memory cgroup.
 */
int mem_cgroup_scan_tasks(struct mem_cgroup *memcg,
			  int (*fn)(struct task_struct *, void *), void *arg)
{
	struct mem_cgroup *iter;
	int ret = 0;

	BUG_ON(memcg == root_mem_cgroup);

	for_each_mem_cgroup_tree(iter, memcg) {
		struct css_task_iter it;
		struct task_struct *task;

		css_task_iter_start(&iter->css, &it);
		while (!ret && (task = css_task_iter_next(&it)))
			ret = fn(task, arg);
		css_task_iter_end(&it);
		if (ret) {
			mem_cgroup_iter_break(memcg, iter);
			break;
		}
	}
	return ret;
}

932
/**
933
 * mem_cgroup_page_lruvec - return lruvec for isolating/putting an LRU page
934
 * @page: the page
935
 * @zone: zone of the page
936 937 938 939
 *
 * 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.
940
 */
M
Mel Gorman 已提交
941
struct lruvec *mem_cgroup_page_lruvec(struct page *page, struct pglist_data *pgdat)
K
KAMEZAWA Hiroyuki 已提交
942
{
943
	struct mem_cgroup_per_node *mz;
944
	struct mem_cgroup *memcg;
945
	struct lruvec *lruvec;
946

947
	if (mem_cgroup_disabled()) {
M
Mel Gorman 已提交
948
		lruvec = &pgdat->lruvec;
949 950
		goto out;
	}
951

952
	memcg = page->mem_cgroup;
953
	/*
954
	 * Swapcache readahead pages are added to the LRU - and
955
	 * possibly migrated - before they are charged.
956
	 */
957 958
	if (!memcg)
		memcg = root_mem_cgroup;
959

960
	mz = mem_cgroup_page_nodeinfo(memcg, page);
961 962 963 964 965 966 967
	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.
	 */
M
Mel Gorman 已提交
968 969
	if (unlikely(lruvec->pgdat != pgdat))
		lruvec->pgdat = pgdat;
970
	return lruvec;
K
KAMEZAWA Hiroyuki 已提交
971
}
972

973
/**
974 975 976
 * 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
977
 * @zid: zone id of the accounted pages
978
 * @nr_pages: positive when adding or negative when removing
979
 *
980 981 982
 * This function must be called under lru_lock, just before a page is added
 * to or just after a page is removed from an lru list (that ordering being
 * so as to allow it to check that lru_size 0 is consistent with list_empty).
983
 */
984
void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
985
				int zid, int nr_pages)
986
{
987
	struct mem_cgroup_per_node *mz;
988
	unsigned long *lru_size;
989
	long size;
990 991 992 993

	if (mem_cgroup_disabled())
		return;

994
	mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
995
	lru_size = &mz->lru_zone_size[zid][lru];
996 997 998 999 1000

	if (nr_pages < 0)
		*lru_size += nr_pages;

	size = *lru_size;
1001 1002 1003
	if (WARN_ONCE(size < 0,
		"%s(%p, %d, %d): lru_size %ld\n",
		__func__, lruvec, lru, nr_pages, size)) {
1004 1005 1006 1007 1008 1009
		VM_BUG_ON(1);
		*lru_size = 0;
	}

	if (nr_pages > 0)
		*lru_size += nr_pages;
K
KAMEZAWA Hiroyuki 已提交
1010
}
1011

1012
bool task_in_mem_cgroup(struct task_struct *task, struct mem_cgroup *memcg)
1013
{
1014
	struct mem_cgroup *task_memcg;
1015
	struct task_struct *p;
1016
	bool ret;
1017

1018
	p = find_lock_task_mm(task);
1019
	if (p) {
1020
		task_memcg = get_mem_cgroup_from_mm(p->mm);
1021 1022 1023 1024 1025 1026 1027
		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.
		 */
1028
		rcu_read_lock();
1029 1030
		task_memcg = mem_cgroup_from_task(task);
		css_get(&task_memcg->css);
1031
		rcu_read_unlock();
1032
	}
1033 1034
	ret = mem_cgroup_is_descendant(task_memcg, memcg);
	css_put(&task_memcg->css);
1035 1036 1037
	return ret;
}

1038
/**
1039
 * mem_cgroup_margin - calculate chargeable space of a memory cgroup
W
Wanpeng Li 已提交
1040
 * @memcg: the memory cgroup
1041
 *
1042
 * Returns the maximum amount of memory @mem can be charged with, in
1043
 * pages.
1044
 */
1045
static unsigned long mem_cgroup_margin(struct mem_cgroup *memcg)
1046
{
1047 1048 1049
	unsigned long margin = 0;
	unsigned long count;
	unsigned long limit;
1050

1051
	count = page_counter_read(&memcg->memory);
1052
	limit = READ_ONCE(memcg->memory.limit);
1053 1054 1055
	if (count < limit)
		margin = limit - count;

1056
	if (do_memsw_account()) {
1057
		count = page_counter_read(&memcg->memsw);
1058
		limit = READ_ONCE(memcg->memsw.limit);
1059 1060
		if (count <= limit)
			margin = min(margin, limit - count);
1061 1062
		else
			margin = 0;
1063 1064 1065
	}

	return margin;
1066 1067
}

1068
/*
Q
Qiang Huang 已提交
1069
 * A routine for checking "mem" is under move_account() or not.
1070
 *
Q
Qiang Huang 已提交
1071 1072 1073
 * 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".
1074
 */
1075
static bool mem_cgroup_under_move(struct mem_cgroup *memcg)
1076
{
1077 1078
	struct mem_cgroup *from;
	struct mem_cgroup *to;
1079
	bool ret = false;
1080 1081 1082 1083 1084 1085 1086 1087 1088
	/*
	 * 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;
1089

1090 1091
	ret = mem_cgroup_is_descendant(from, memcg) ||
		mem_cgroup_is_descendant(to, memcg);
1092 1093
unlock:
	spin_unlock(&mc.lock);
1094 1095 1096
	return ret;
}

1097
static bool mem_cgroup_wait_acct_move(struct mem_cgroup *memcg)
1098 1099
{
	if (mc.moving_task && current != mc.moving_task) {
1100
		if (mem_cgroup_under_move(memcg)) {
1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112
			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;
}

1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134
unsigned int memcg1_stats[] = {
	MEMCG_CACHE,
	MEMCG_RSS,
	MEMCG_RSS_HUGE,
	NR_SHMEM,
	NR_FILE_MAPPED,
	NR_FILE_DIRTY,
	NR_WRITEBACK,
	MEMCG_SWAP,
};

static const char *const memcg1_stat_names[] = {
	"cache",
	"rss",
	"rss_huge",
	"shmem",
	"mapped_file",
	"dirty",
	"writeback",
	"swap",
};

1135
#define K(x) ((x) << (PAGE_SHIFT-10))
1136
/**
1137
 * mem_cgroup_print_oom_info: Print OOM information relevant to memory controller.
1138 1139 1140 1141 1142 1143 1144 1145
 * @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)
{
1146 1147
	struct mem_cgroup *iter;
	unsigned int i;
1148 1149 1150

	rcu_read_lock();

1151 1152 1153 1154 1155 1156 1157 1158
	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
1159
	pr_cont_cgroup_path(memcg->css.cgroup);
1160
	pr_cont("\n");
1161 1162 1163

	rcu_read_unlock();

1164 1165 1166 1167 1168 1169 1170 1171 1172
	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);
1173 1174

	for_each_mem_cgroup_tree(iter, memcg) {
T
Tejun Heo 已提交
1175 1176
		pr_info("Memory cgroup stats for ");
		pr_cont_cgroup_path(iter->css.cgroup);
1177 1178
		pr_cont(":");

1179 1180
		for (i = 0; i < ARRAY_SIZE(memcg1_stats); i++) {
			if (memcg1_stats[i] == MEMCG_SWAP && !do_swap_account)
1181
				continue;
1182
			pr_cont(" %s:%luKB", memcg1_stat_names[i],
1183
				K(memcg_page_state(iter, memcg1_stats[i])));
1184 1185 1186 1187 1188 1189 1190 1191
		}

		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");
	}
1192 1193
}

1194 1195 1196 1197
/*
 * This function returns the number of memcg under hierarchy tree. Returns
 * 1(self count) if no children.
 */
1198
static int mem_cgroup_count_children(struct mem_cgroup *memcg)
1199 1200
{
	int num = 0;
K
KAMEZAWA Hiroyuki 已提交
1201 1202
	struct mem_cgroup *iter;

1203
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
1204
		num++;
1205 1206 1207
	return num;
}

D
David Rientjes 已提交
1208 1209 1210
/*
 * Return the memory (and swap, if configured) limit for a memcg.
 */
1211
unsigned long mem_cgroup_get_limit(struct mem_cgroup *memcg)
D
David Rientjes 已提交
1212
{
1213
	unsigned long limit;
1214

1215
	limit = memcg->memory.limit;
1216
	if (mem_cgroup_swappiness(memcg)) {
1217
		unsigned long memsw_limit;
1218
		unsigned long swap_limit;
1219

1220
		memsw_limit = memcg->memsw.limit;
1221 1222 1223
		swap_limit = memcg->swap.limit;
		swap_limit = min(swap_limit, (unsigned long)total_swap_pages);
		limit = min(limit + swap_limit, memsw_limit);
1224 1225
	}
	return limit;
D
David Rientjes 已提交
1226 1227
}

1228
static bool mem_cgroup_out_of_memory(struct mem_cgroup *memcg, gfp_t gfp_mask,
1229
				     int order)
1230
{
1231 1232 1233
	struct oom_control oc = {
		.zonelist = NULL,
		.nodemask = NULL,
1234
		.memcg = memcg,
1235 1236 1237
		.gfp_mask = gfp_mask,
		.order = order,
	};
1238
	bool ret;
1239

1240
	mutex_lock(&oom_lock);
1241
	ret = out_of_memory(&oc);
1242
	mutex_unlock(&oom_lock);
1243
	return ret;
1244 1245
}

1246 1247
#if MAX_NUMNODES > 1

1248 1249
/**
 * test_mem_cgroup_node_reclaimable
W
Wanpeng Li 已提交
1250
 * @memcg: the target memcg
1251 1252 1253 1254 1255 1256 1257
 * @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.
 */
1258
static bool test_mem_cgroup_node_reclaimable(struct mem_cgroup *memcg,
1259 1260
		int nid, bool noswap)
{
1261
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_FILE))
1262 1263 1264
		return true;
	if (noswap || !total_swap_pages)
		return false;
1265
	if (mem_cgroup_node_nr_lru_pages(memcg, nid, LRU_ALL_ANON))
1266 1267 1268 1269
		return true;
	return false;

}
1270 1271 1272 1273 1274 1275 1276

/*
 * 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.
 *
 */
1277
static void mem_cgroup_may_update_nodemask(struct mem_cgroup *memcg)
1278 1279
{
	int nid;
1280 1281 1282 1283
	/*
	 * numainfo_events > 0 means there was at least NUMAINFO_EVENTS_TARGET
	 * pagein/pageout changes since the last update.
	 */
1284
	if (!atomic_read(&memcg->numainfo_events))
1285
		return;
1286
	if (atomic_inc_return(&memcg->numainfo_updating) > 1)
1287 1288 1289
		return;

	/* make a nodemask where this memcg uses memory from */
1290
	memcg->scan_nodes = node_states[N_MEMORY];
1291

1292
	for_each_node_mask(nid, node_states[N_MEMORY]) {
1293

1294 1295
		if (!test_mem_cgroup_node_reclaimable(memcg, nid, false))
			node_clear(nid, memcg->scan_nodes);
1296
	}
1297

1298 1299
	atomic_set(&memcg->numainfo_events, 0);
	atomic_set(&memcg->numainfo_updating, 0);
1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313
}

/*
 * 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.
 */
1314
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
1315 1316 1317
{
	int node;

1318 1319
	mem_cgroup_may_update_nodemask(memcg);
	node = memcg->last_scanned_node;
1320

1321
	node = next_node_in(node, memcg->scan_nodes);
1322
	/*
1323 1324 1325
	 * mem_cgroup_may_update_nodemask might have seen no reclaimmable pages
	 * last time it really checked all the LRUs due to rate limiting.
	 * Fallback to the current node in that case for simplicity.
1326 1327 1328 1329
	 */
	if (unlikely(node == MAX_NUMNODES))
		node = numa_node_id();

1330
	memcg->last_scanned_node = node;
1331 1332 1333
	return node;
}
#else
1334
int mem_cgroup_select_victim_node(struct mem_cgroup *memcg)
1335 1336 1337 1338 1339
{
	return 0;
}
#endif

1340
static int mem_cgroup_soft_reclaim(struct mem_cgroup *root_memcg,
1341
				   pg_data_t *pgdat,
1342 1343 1344 1345 1346 1347 1348 1349 1350
				   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 = {
1351
		.pgdat = pgdat,
1352 1353 1354
		.priority = 0,
	};

1355
	excess = soft_limit_excess(root_memcg);
1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380

	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;
		}
1381
		total += mem_cgroup_shrink_node(victim, gfp_mask, false,
1382
					pgdat, &nr_scanned);
1383
		*total_scanned += nr_scanned;
1384
		if (!soft_limit_excess(root_memcg))
1385
			break;
1386
	}
1387 1388
	mem_cgroup_iter_break(root_memcg, victim);
	return total;
1389 1390
}

1391 1392 1393 1394 1395 1396
#ifdef CONFIG_LOCKDEP
static struct lockdep_map memcg_oom_lock_dep_map = {
	.name = "memcg_oom_lock",
};
#endif

1397 1398
static DEFINE_SPINLOCK(memcg_oom_lock);

K
KAMEZAWA Hiroyuki 已提交
1399 1400 1401 1402
/*
 * Check OOM-Killer is already running under our hierarchy.
 * If someone is running, return false.
 */
1403
static bool mem_cgroup_oom_trylock(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
1404
{
1405
	struct mem_cgroup *iter, *failed = NULL;
1406

1407 1408
	spin_lock(&memcg_oom_lock);

1409
	for_each_mem_cgroup_tree(iter, memcg) {
1410
		if (iter->oom_lock) {
1411 1412 1413 1414 1415
			/*
			 * this subtree of our hierarchy is already locked
			 * so we cannot give a lock.
			 */
			failed = iter;
1416 1417
			mem_cgroup_iter_break(memcg, iter);
			break;
1418 1419
		} else
			iter->oom_lock = true;
K
KAMEZAWA Hiroyuki 已提交
1420
	}
K
KAMEZAWA Hiroyuki 已提交
1421

1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432
	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;
1433
		}
1434 1435
	} else
		mutex_acquire(&memcg_oom_lock_dep_map, 0, 1, _RET_IP_);
1436 1437 1438 1439

	spin_unlock(&memcg_oom_lock);

	return !failed;
1440
}
1441

1442
static void mem_cgroup_oom_unlock(struct mem_cgroup *memcg)
1443
{
K
KAMEZAWA Hiroyuki 已提交
1444 1445
	struct mem_cgroup *iter;

1446
	spin_lock(&memcg_oom_lock);
1447
	mutex_release(&memcg_oom_lock_dep_map, 1, _RET_IP_);
1448
	for_each_mem_cgroup_tree(iter, memcg)
1449
		iter->oom_lock = false;
1450
	spin_unlock(&memcg_oom_lock);
1451 1452
}

1453
static void mem_cgroup_mark_under_oom(struct mem_cgroup *memcg)
1454 1455 1456
{
	struct mem_cgroup *iter;

1457
	spin_lock(&memcg_oom_lock);
1458
	for_each_mem_cgroup_tree(iter, memcg)
1459 1460
		iter->under_oom++;
	spin_unlock(&memcg_oom_lock);
1461 1462
}

1463
static void mem_cgroup_unmark_under_oom(struct mem_cgroup *memcg)
1464 1465 1466
{
	struct mem_cgroup *iter;

K
KAMEZAWA Hiroyuki 已提交
1467 1468
	/*
	 * When a new child is created while the hierarchy is under oom,
1469
	 * mem_cgroup_oom_lock() may not be called. Watch for underflow.
K
KAMEZAWA Hiroyuki 已提交
1470
	 */
1471
	spin_lock(&memcg_oom_lock);
1472
	for_each_mem_cgroup_tree(iter, memcg)
1473 1474 1475
		if (iter->under_oom > 0)
			iter->under_oom--;
	spin_unlock(&memcg_oom_lock);
1476 1477
}

K
KAMEZAWA Hiroyuki 已提交
1478 1479
static DECLARE_WAIT_QUEUE_HEAD(memcg_oom_waitq);

K
KAMEZAWA Hiroyuki 已提交
1480
struct oom_wait_info {
1481
	struct mem_cgroup *memcg;
1482
	wait_queue_entry_t	wait;
K
KAMEZAWA Hiroyuki 已提交
1483 1484
};

1485
static int memcg_oom_wake_function(wait_queue_entry_t *wait,
K
KAMEZAWA Hiroyuki 已提交
1486 1487
	unsigned mode, int sync, void *arg)
{
1488 1489
	struct mem_cgroup *wake_memcg = (struct mem_cgroup *)arg;
	struct mem_cgroup *oom_wait_memcg;
K
KAMEZAWA Hiroyuki 已提交
1490 1491 1492
	struct oom_wait_info *oom_wait_info;

	oom_wait_info = container_of(wait, struct oom_wait_info, wait);
1493
	oom_wait_memcg = oom_wait_info->memcg;
K
KAMEZAWA Hiroyuki 已提交
1494

1495 1496
	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
1497 1498 1499 1500
		return 0;
	return autoremove_wake_function(wait, mode, sync, arg);
}

1501
static void memcg_oom_recover(struct mem_cgroup *memcg)
1502
{
1503 1504 1505 1506 1507 1508 1509 1510 1511
	/*
	 * 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)
1512
		__wake_up(&memcg_oom_waitq, TASK_NORMAL, 0, memcg);
1513 1514
}

1515
static void mem_cgroup_oom(struct mem_cgroup *memcg, gfp_t mask, int order)
1516
{
1517
	if (!current->memcg_may_oom)
1518
		return;
K
KAMEZAWA Hiroyuki 已提交
1519
	/*
1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531
	 * 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
1532
	 */
1533
	css_get(&memcg->css);
T
Tejun Heo 已提交
1534 1535 1536
	current->memcg_in_oom = memcg;
	current->memcg_oom_gfp_mask = mask;
	current->memcg_oom_order = order;
1537 1538 1539 1540
}

/**
 * mem_cgroup_oom_synchronize - complete memcg OOM handling
1541
 * @handle: actually kill/wait or just clean up the OOM state
1542
 *
1543 1544
 * This has to be called at the end of a page fault if the memcg OOM
 * handler was enabled.
1545
 *
1546
 * Memcg supports userspace OOM handling where failed allocations must
1547 1548 1549 1550
 * 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
1551
 * the end of the page fault to complete the OOM handling.
1552 1553
 *
 * Returns %true if an ongoing memcg OOM situation was detected and
1554
 * completed, %false otherwise.
1555
 */
1556
bool mem_cgroup_oom_synchronize(bool handle)
1557
{
T
Tejun Heo 已提交
1558
	struct mem_cgroup *memcg = current->memcg_in_oom;
1559
	struct oom_wait_info owait;
1560
	bool locked;
1561 1562 1563

	/* OOM is global, do not handle */
	if (!memcg)
1564
		return false;
1565

1566
	if (!handle)
1567
		goto cleanup;
1568 1569 1570 1571 1572

	owait.memcg = memcg;
	owait.wait.flags = 0;
	owait.wait.func = memcg_oom_wake_function;
	owait.wait.private = current;
1573
	INIT_LIST_HEAD(&owait.wait.entry);
K
KAMEZAWA Hiroyuki 已提交
1574

1575
	prepare_to_wait(&memcg_oom_waitq, &owait.wait, TASK_KILLABLE);
1576 1577 1578 1579 1580 1581 1582 1583 1584 1585
	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
1586 1587
		mem_cgroup_out_of_memory(memcg, current->memcg_oom_gfp_mask,
					 current->memcg_oom_order);
1588
	} else {
1589
		schedule();
1590 1591 1592 1593 1594
		mem_cgroup_unmark_under_oom(memcg);
		finish_wait(&memcg_oom_waitq, &owait.wait);
	}

	if (locked) {
1595 1596 1597 1598 1599 1600 1601 1602
		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);
	}
1603
cleanup:
T
Tejun Heo 已提交
1604
	current->memcg_in_oom = NULL;
1605
	css_put(&memcg->css);
K
KAMEZAWA Hiroyuki 已提交
1606
	return true;
1607 1608
}

1609
/**
1610 1611
 * lock_page_memcg - lock a page->mem_cgroup binding
 * @page: the page
1612
 *
1613 1614
 * This function protects unlocked LRU pages from being moved to
 * another cgroup and stabilizes their page->mem_cgroup binding.
1615
 */
J
Johannes Weiner 已提交
1616
void lock_page_memcg(struct page *page)
1617 1618
{
	struct mem_cgroup *memcg;
1619
	unsigned long flags;
1620

1621 1622 1623 1624 1625
	/*
	 * 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.
	 */
1626 1627 1628
	rcu_read_lock();

	if (mem_cgroup_disabled())
J
Johannes Weiner 已提交
1629
		return;
1630
again:
1631
	memcg = page->mem_cgroup;
1632
	if (unlikely(!memcg))
J
Johannes Weiner 已提交
1633
		return;
1634

Q
Qiang Huang 已提交
1635
	if (atomic_read(&memcg->moving_account) <= 0)
J
Johannes Weiner 已提交
1636
		return;
1637

1638
	spin_lock_irqsave(&memcg->move_lock, flags);
1639
	if (memcg != page->mem_cgroup) {
1640
		spin_unlock_irqrestore(&memcg->move_lock, flags);
1641 1642
		goto again;
	}
1643 1644 1645 1646

	/*
	 * When charge migration first begins, we can have locked and
	 * unlocked page stat updates happening concurrently.  Track
1647
	 * the task who has the lock for unlock_page_memcg().
1648 1649 1650
	 */
	memcg->move_lock_task = current;
	memcg->move_lock_flags = flags;
1651

J
Johannes Weiner 已提交
1652
	return;
1653
}
1654
EXPORT_SYMBOL(lock_page_memcg);
1655

1656
/**
1657
 * unlock_page_memcg - unlock a page->mem_cgroup binding
J
Johannes Weiner 已提交
1658
 * @page: the page
1659
 */
J
Johannes Weiner 已提交
1660
void unlock_page_memcg(struct page *page)
1661
{
J
Johannes Weiner 已提交
1662 1663
	struct mem_cgroup *memcg = page->mem_cgroup;

1664 1665 1666 1667 1668 1669 1670 1671
	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);
	}
1672

1673
	rcu_read_unlock();
1674
}
1675
EXPORT_SYMBOL(unlock_page_memcg);
1676

1677 1678 1679 1680
/*
 * size of first charge trial. "32" comes from vmscan.c's magic value.
 * TODO: maybe necessary to use big numbers in big irons.
 */
1681
#define CHARGE_BATCH	32U
1682 1683
struct memcg_stock_pcp {
	struct mem_cgroup *cached; /* this never be root cgroup */
1684
	unsigned int nr_pages;
1685
	struct work_struct work;
1686
	unsigned long flags;
1687
#define FLUSHING_CACHED_CHARGE	0
1688 1689
};
static DEFINE_PER_CPU(struct memcg_stock_pcp, memcg_stock);
1690
static DEFINE_MUTEX(percpu_charge_mutex);
1691

1692 1693 1694 1695 1696 1697 1698 1699 1700 1701
/**
 * 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.
1702
 */
1703
static bool consume_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
1704 1705
{
	struct memcg_stock_pcp *stock;
1706
	unsigned long flags;
1707
	bool ret = false;
1708

1709
	if (nr_pages > CHARGE_BATCH)
1710
		return ret;
1711

1712 1713 1714
	local_irq_save(flags);

	stock = this_cpu_ptr(&memcg_stock);
1715
	if (memcg == stock->cached && stock->nr_pages >= nr_pages) {
1716
		stock->nr_pages -= nr_pages;
1717 1718
		ret = true;
	}
1719 1720 1721

	local_irq_restore(flags);

1722 1723 1724 1725
	return ret;
}

/*
1726
 * Returns stocks cached in percpu and reset cached information.
1727 1728 1729 1730 1731
 */
static void drain_stock(struct memcg_stock_pcp *stock)
{
	struct mem_cgroup *old = stock->cached;

1732
	if (stock->nr_pages) {
1733
		page_counter_uncharge(&old->memory, stock->nr_pages);
1734
		if (do_memsw_account())
1735
			page_counter_uncharge(&old->memsw, stock->nr_pages);
1736
		css_put_many(&old->css, stock->nr_pages);
1737
		stock->nr_pages = 0;
1738 1739 1740 1741 1742 1743
	}
	stock->cached = NULL;
}

static void drain_local_stock(struct work_struct *dummy)
{
1744 1745 1746 1747 1748 1749
	struct memcg_stock_pcp *stock;
	unsigned long flags;

	local_irq_save(flags);

	stock = this_cpu_ptr(&memcg_stock);
1750
	drain_stock(stock);
1751
	clear_bit(FLUSHING_CACHED_CHARGE, &stock->flags);
1752 1753

	local_irq_restore(flags);
1754 1755 1756
}

/*
1757
 * Cache charges(val) to local per_cpu area.
1758
 * This will be consumed by consume_stock() function, later.
1759
 */
1760
static void refill_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
1761
{
1762 1763 1764 1765
	struct memcg_stock_pcp *stock;
	unsigned long flags;

	local_irq_save(flags);
1766

1767
	stock = this_cpu_ptr(&memcg_stock);
1768
	if (stock->cached != memcg) { /* reset if necessary */
1769
		drain_stock(stock);
1770
		stock->cached = memcg;
1771
	}
1772
	stock->nr_pages += nr_pages;
1773 1774

	local_irq_restore(flags);
1775 1776 1777
}

/*
1778
 * Drains all per-CPU charge caches for given root_memcg resp. subtree
1779
 * of the hierarchy under it.
1780
 */
1781
static void drain_all_stock(struct mem_cgroup *root_memcg)
1782
{
1783
	int cpu, curcpu;
1784

1785 1786 1787
	/* If someone's already draining, avoid adding running more workers. */
	if (!mutex_trylock(&percpu_charge_mutex))
		return;
1788 1789
	/* Notify other cpus that system-wide "drain" is running */
	get_online_cpus();
1790
	curcpu = get_cpu();
1791 1792
	for_each_online_cpu(cpu) {
		struct memcg_stock_pcp *stock = &per_cpu(memcg_stock, cpu);
1793
		struct mem_cgroup *memcg;
1794

1795 1796
		memcg = stock->cached;
		if (!memcg || !stock->nr_pages)
1797
			continue;
1798
		if (!mem_cgroup_is_descendant(memcg, root_memcg))
1799
			continue;
1800 1801 1802 1803 1804 1805
		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);
		}
1806
	}
1807
	put_cpu();
A
Andrew Morton 已提交
1808
	put_online_cpus();
1809
	mutex_unlock(&percpu_charge_mutex);
1810 1811
}

1812
static int memcg_hotplug_cpu_dead(unsigned int cpu)
1813 1814 1815 1816 1817
{
	struct memcg_stock_pcp *stock;

	stock = &per_cpu(memcg_stock, cpu);
	drain_stock(stock);
1818
	return 0;
1819 1820
}

1821 1822 1823 1824 1825 1826 1827
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;
1828
		mem_cgroup_event(memcg, MEMCG_HIGH);
1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840
		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);
}

1841 1842 1843 1844 1845 1846 1847
/*
 * 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;
1848
	struct mem_cgroup *memcg;
1849 1850 1851 1852

	if (likely(!nr_pages))
		return;

1853 1854
	memcg = get_mem_cgroup_from_mm(current->mm);
	reclaim_high(memcg, nr_pages, GFP_KERNEL);
1855 1856 1857 1858
	css_put(&memcg->css);
	current->memcg_nr_pages_over_high = 0;
}

1859 1860
static int try_charge(struct mem_cgroup *memcg, gfp_t gfp_mask,
		      unsigned int nr_pages)
1861
{
1862
	unsigned int batch = max(CHARGE_BATCH, nr_pages);
1863
	int nr_retries = MEM_CGROUP_RECLAIM_RETRIES;
1864
	struct mem_cgroup *mem_over_limit;
1865
	struct page_counter *counter;
1866
	unsigned long nr_reclaimed;
1867 1868
	bool may_swap = true;
	bool drained = false;
1869

1870
	if (mem_cgroup_is_root(memcg))
1871
		return 0;
1872
retry:
1873
	if (consume_stock(memcg, nr_pages))
1874
		return 0;
1875

1876
	if (!do_memsw_account() ||
1877 1878
	    page_counter_try_charge(&memcg->memsw, batch, &counter)) {
		if (page_counter_try_charge(&memcg->memory, batch, &counter))
1879
			goto done_restock;
1880
		if (do_memsw_account())
1881 1882
			page_counter_uncharge(&memcg->memsw, batch);
		mem_over_limit = mem_cgroup_from_counter(counter, memory);
1883
	} else {
1884
		mem_over_limit = mem_cgroup_from_counter(counter, memsw);
1885
		may_swap = false;
1886
	}
1887

1888 1889 1890 1891
	if (batch > nr_pages) {
		batch = nr_pages;
		goto retry;
	}
1892

1893 1894 1895 1896 1897 1898 1899 1900 1901
	/*
	 * 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))
1902
		goto force;
1903

1904 1905 1906 1907 1908 1909 1910 1911 1912
	/*
	 * Prevent unbounded recursion when reclaim operations need to
	 * allocate memory. This might exceed the limits temporarily,
	 * but we prefer facilitating memory reclaim and getting back
	 * under the limit over triggering OOM kills in these cases.
	 */
	if (unlikely(current->flags & PF_MEMALLOC))
		goto force;

1913 1914 1915
	if (unlikely(task_in_memcg_oom(current)))
		goto nomem;

1916
	if (!gfpflags_allow_blocking(gfp_mask))
1917
		goto nomem;
1918

1919
	mem_cgroup_event(mem_over_limit, MEMCG_MAX);
1920

1921 1922
	nr_reclaimed = try_to_free_mem_cgroup_pages(mem_over_limit, nr_pages,
						    gfp_mask, may_swap);
1923

1924
	if (mem_cgroup_margin(mem_over_limit) >= nr_pages)
1925
		goto retry;
1926

1927
	if (!drained) {
1928
		drain_all_stock(mem_over_limit);
1929 1930 1931 1932
		drained = true;
		goto retry;
	}

1933 1934
	if (gfp_mask & __GFP_NORETRY)
		goto nomem;
1935 1936 1937 1938 1939 1940 1941 1942 1943
	/*
	 * 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.
	 */
1944
	if (nr_reclaimed && nr_pages <= (1 << PAGE_ALLOC_COSTLY_ORDER))
1945 1946 1947 1948 1949 1950 1951 1952
		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;

1953 1954 1955
	if (nr_retries--)
		goto retry;

1956
	if (gfp_mask & __GFP_NOFAIL)
1957
		goto force;
1958

1959
	if (fatal_signal_pending(current))
1960
		goto force;
1961

1962
	mem_cgroup_event(mem_over_limit, MEMCG_OOM);
1963

1964 1965
	mem_cgroup_oom(mem_over_limit, gfp_mask,
		       get_order(nr_pages * PAGE_SIZE));
1966
nomem:
1967
	if (!(gfp_mask & __GFP_NOFAIL))
1968
		return -ENOMEM;
1969 1970 1971 1972 1973 1974 1975
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);
1976
	if (do_memsw_account())
1977 1978 1979 1980
		page_counter_charge(&memcg->memsw, nr_pages);
	css_get_many(&memcg->css, nr_pages);

	return 0;
1981 1982

done_restock:
1983
	css_get_many(&memcg->css, batch);
1984 1985
	if (batch > nr_pages)
		refill_stock(memcg, batch - nr_pages);
1986

1987
	/*
1988 1989
	 * If the hierarchy is above the normal consumption range, schedule
	 * reclaim on returning to userland.  We can perform reclaim here
1990
	 * if __GFP_RECLAIM but let's always punt for simplicity and so that
1991 1992 1993 1994
	 * 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.
1995 1996
	 */
	do {
1997
		if (page_counter_read(&memcg->memory) > memcg->high) {
1998 1999 2000 2001 2002
			/* Don't bother a random interrupted task */
			if (in_interrupt()) {
				schedule_work(&memcg->high_work);
				break;
			}
V
Vladimir Davydov 已提交
2003
			current->memcg_nr_pages_over_high += batch;
2004 2005 2006
			set_notify_resume(current);
			break;
		}
2007
	} while ((memcg = parent_mem_cgroup(memcg)));
2008 2009

	return 0;
2010
}
2011

2012
static void cancel_charge(struct mem_cgroup *memcg, unsigned int nr_pages)
2013
{
2014 2015 2016
	if (mem_cgroup_is_root(memcg))
		return;

2017
	page_counter_uncharge(&memcg->memory, nr_pages);
2018
	if (do_memsw_account())
2019
		page_counter_uncharge(&memcg->memsw, nr_pages);
2020

2021
	css_put_many(&memcg->css, nr_pages);
2022 2023
}

2024 2025 2026 2027
static void lock_page_lru(struct page *page, int *isolated)
{
	struct zone *zone = page_zone(page);

2028
	spin_lock_irq(zone_lru_lock(zone));
2029 2030 2031
	if (PageLRU(page)) {
		struct lruvec *lruvec;

M
Mel Gorman 已提交
2032
		lruvec = mem_cgroup_page_lruvec(page, zone->zone_pgdat);
2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046
		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;

M
Mel Gorman 已提交
2047
		lruvec = mem_cgroup_page_lruvec(page, zone->zone_pgdat);
2048 2049 2050 2051
		VM_BUG_ON_PAGE(PageLRU(page), page);
		SetPageLRU(page);
		add_page_to_lru_list(page, lruvec, page_lru(page));
	}
2052
	spin_unlock_irq(zone_lru_lock(zone));
2053 2054
}

2055
static void commit_charge(struct page *page, struct mem_cgroup *memcg,
2056
			  bool lrucare)
2057
{
2058
	int isolated;
2059

2060
	VM_BUG_ON_PAGE(page->mem_cgroup, page);
2061 2062 2063 2064 2065

	/*
	 * 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.
	 */
2066 2067
	if (lrucare)
		lock_page_lru(page, &isolated);
2068

2069 2070
	/*
	 * Nobody should be changing or seriously looking at
2071
	 * page->mem_cgroup at this point:
2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082
	 *
	 * - 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
	 */
2083
	page->mem_cgroup = memcg;
2084

2085 2086
	if (lrucare)
		unlock_page_lru(page, isolated);
2087
}
2088

2089
#ifndef CONFIG_SLOB
2090
static int memcg_alloc_cache_id(void)
2091
{
2092 2093 2094
	int id, size;
	int err;

2095
	id = ida_simple_get(&memcg_cache_ida,
2096 2097 2098
			    0, MEMCG_CACHES_MAX_SIZE, GFP_KERNEL);
	if (id < 0)
		return id;
2099

2100
	if (id < memcg_nr_cache_ids)
2101 2102 2103 2104 2105 2106
		return id;

	/*
	 * There's no space for the new id in memcg_caches arrays,
	 * so we have to grow them.
	 */
2107
	down_write(&memcg_cache_ids_sem);
2108 2109

	size = 2 * (id + 1);
2110 2111 2112 2113 2114
	if (size < MEMCG_CACHES_MIN_SIZE)
		size = MEMCG_CACHES_MIN_SIZE;
	else if (size > MEMCG_CACHES_MAX_SIZE)
		size = MEMCG_CACHES_MAX_SIZE;

2115
	err = memcg_update_all_caches(size);
2116 2117
	if (!err)
		err = memcg_update_all_list_lrus(size);
2118 2119 2120 2121 2122
	if (!err)
		memcg_nr_cache_ids = size;

	up_write(&memcg_cache_ids_sem);

2123
	if (err) {
2124
		ida_simple_remove(&memcg_cache_ida, id);
2125 2126 2127 2128 2129 2130 2131
		return err;
	}
	return id;
}

static void memcg_free_cache_id(int id)
{
2132
	ida_simple_remove(&memcg_cache_ida, id);
2133 2134
}

2135
struct memcg_kmem_cache_create_work {
2136 2137 2138 2139 2140
	struct mem_cgroup *memcg;
	struct kmem_cache *cachep;
	struct work_struct work;
};

2141
static void memcg_kmem_cache_create_func(struct work_struct *w)
2142
{
2143 2144
	struct memcg_kmem_cache_create_work *cw =
		container_of(w, struct memcg_kmem_cache_create_work, work);
2145 2146
	struct mem_cgroup *memcg = cw->memcg;
	struct kmem_cache *cachep = cw->cachep;
2147

2148
	memcg_create_kmem_cache(memcg, cachep);
2149

2150
	css_put(&memcg->css);
2151 2152 2153 2154 2155 2156
	kfree(cw);
}

/*
 * Enqueue the creation of a per-memcg kmem_cache.
 */
2157 2158
static void __memcg_schedule_kmem_cache_create(struct mem_cgroup *memcg,
					       struct kmem_cache *cachep)
2159
{
2160
	struct memcg_kmem_cache_create_work *cw;
2161

2162
	cw = kmalloc(sizeof(*cw), GFP_NOWAIT);
2163
	if (!cw)
2164
		return;
2165 2166

	css_get(&memcg->css);
2167 2168 2169

	cw->memcg = memcg;
	cw->cachep = cachep;
2170
	INIT_WORK(&cw->work, memcg_kmem_cache_create_func);
2171

2172
	queue_work(memcg_kmem_cache_wq, &cw->work);
2173 2174
}

2175 2176
static void memcg_schedule_kmem_cache_create(struct mem_cgroup *memcg,
					     struct kmem_cache *cachep)
2177 2178 2179 2180
{
	/*
	 * We need to stop accounting when we kmalloc, because if the
	 * corresponding kmalloc cache is not yet created, the first allocation
2181
	 * in __memcg_schedule_kmem_cache_create will recurse.
2182 2183 2184 2185 2186 2187 2188
	 *
	 * 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.
	 */
2189
	current->memcg_kmem_skip_account = 1;
2190
	__memcg_schedule_kmem_cache_create(memcg, cachep);
2191
	current->memcg_kmem_skip_account = 0;
2192
}
2193

2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204
static inline bool memcg_kmem_bypass(void)
{
	if (in_interrupt() || !current->mm || (current->flags & PF_KTHREAD))
		return true;
	return false;
}

/**
 * memcg_kmem_get_cache: select the correct per-memcg cache for allocation
 * @cachep: the original global kmem cache
 *
2205 2206 2207
 * 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.
 *
2208 2209 2210
 * If the cache does not exist yet, if we are the first user of it, we
 * create it asynchronously in a workqueue and let the current allocation
 * go through with the original cache.
2211
 *
2212 2213 2214 2215
 * This function takes a reference to the cache it returns to assure it
 * won't get destroyed while we are working with it. Once the caller is
 * done with it, memcg_kmem_put_cache() must be called to release the
 * reference.
2216
 */
2217
struct kmem_cache *memcg_kmem_get_cache(struct kmem_cache *cachep)
2218 2219
{
	struct mem_cgroup *memcg;
2220
	struct kmem_cache *memcg_cachep;
2221
	int kmemcg_id;
2222

2223
	VM_BUG_ON(!is_root_cache(cachep));
2224

2225
	if (memcg_kmem_bypass())
V
Vladimir Davydov 已提交
2226 2227
		return cachep;

2228
	if (current->memcg_kmem_skip_account)
2229 2230
		return cachep;

2231
	memcg = get_mem_cgroup_from_mm(current->mm);
2232
	kmemcg_id = READ_ONCE(memcg->kmemcg_id);
2233
	if (kmemcg_id < 0)
2234
		goto out;
2235

2236
	memcg_cachep = cache_from_memcg_idx(cachep, kmemcg_id);
2237 2238
	if (likely(memcg_cachep))
		return memcg_cachep;
2239 2240 2241 2242 2243 2244 2245 2246 2247

	/*
	 * 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
2248 2249 2250
	 * memcg_create_kmem_cache, this means no further allocation
	 * could happen with the slab_mutex held. So it's better to
	 * defer everything.
2251
	 */
2252
	memcg_schedule_kmem_cache_create(memcg, cachep);
2253
out:
2254
	css_put(&memcg->css);
2255
	return cachep;
2256 2257
}

2258 2259 2260 2261 2262
/**
 * memcg_kmem_put_cache: drop reference taken by memcg_kmem_get_cache
 * @cachep: the cache returned by memcg_kmem_get_cache
 */
void memcg_kmem_put_cache(struct kmem_cache *cachep)
2263 2264
{
	if (!is_root_cache(cachep))
2265
		css_put(&cachep->memcg_params.memcg->css);
2266 2267
}

2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278
/**
 * memcg_kmem_charge: charge a kmem page
 * @page: page to charge
 * @gfp: reclaim mode
 * @order: allocation order
 * @memcg: memory cgroup to charge
 *
 * Returns 0 on success, an error code on failure.
 */
int memcg_kmem_charge_memcg(struct page *page, gfp_t gfp, int order,
			    struct mem_cgroup *memcg)
2279
{
2280 2281
	unsigned int nr_pages = 1 << order;
	struct page_counter *counter;
2282 2283
	int ret;

2284
	ret = try_charge(memcg, gfp, nr_pages);
2285
	if (ret)
2286
		return ret;
2287 2288 2289 2290 2291

	if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) &&
	    !page_counter_try_charge(&memcg->kmem, nr_pages, &counter)) {
		cancel_charge(memcg, nr_pages);
		return -ENOMEM;
2292 2293
	}

2294
	page->mem_cgroup = memcg;
2295

2296
	return 0;
2297 2298
}

2299 2300 2301 2302 2303 2304 2305 2306 2307
/**
 * memcg_kmem_charge: charge a kmem page to the current memory cgroup
 * @page: page to charge
 * @gfp: reclaim mode
 * @order: allocation order
 *
 * Returns 0 on success, an error code on failure.
 */
int memcg_kmem_charge(struct page *page, gfp_t gfp, int order)
2308
{
2309
	struct mem_cgroup *memcg;
2310
	int ret = 0;
2311

2312 2313 2314
	if (memcg_kmem_bypass())
		return 0;

2315
	memcg = get_mem_cgroup_from_mm(current->mm);
2316
	if (!mem_cgroup_is_root(memcg)) {
2317
		ret = memcg_kmem_charge_memcg(page, gfp, order, memcg);
2318 2319 2320
		if (!ret)
			__SetPageKmemcg(page);
	}
2321
	css_put(&memcg->css);
2322
	return ret;
2323
}
2324 2325 2326 2327 2328 2329
/**
 * memcg_kmem_uncharge: uncharge a kmem page
 * @page: page to uncharge
 * @order: allocation order
 */
void memcg_kmem_uncharge(struct page *page, int order)
2330
{
2331
	struct mem_cgroup *memcg = page->mem_cgroup;
2332
	unsigned int nr_pages = 1 << order;
2333 2334 2335 2336

	if (!memcg)
		return;

2337
	VM_BUG_ON_PAGE(mem_cgroup_is_root(memcg), page);
2338

2339 2340 2341
	if (!cgroup_subsys_on_dfl(memory_cgrp_subsys))
		page_counter_uncharge(&memcg->kmem, nr_pages);

2342
	page_counter_uncharge(&memcg->memory, nr_pages);
2343
	if (do_memsw_account())
2344
		page_counter_uncharge(&memcg->memsw, nr_pages);
2345

2346
	page->mem_cgroup = NULL;
2347 2348 2349 2350 2351

	/* slab pages do not have PageKmemcg flag set */
	if (PageKmemcg(page))
		__ClearPageKmemcg(page);

2352
	css_put_many(&memcg->css, nr_pages);
2353
}
2354
#endif /* !CONFIG_SLOB */
2355

2356 2357 2358 2359
#ifdef CONFIG_TRANSPARENT_HUGEPAGE

/*
 * Because tail pages are not marked as "used", set it. We're under
2360
 * zone_lru_lock and migration entries setup in all page mappings.
2361
 */
2362
void mem_cgroup_split_huge_fixup(struct page *head)
2363
{
2364
	int i;
2365

2366 2367
	if (mem_cgroup_disabled())
		return;
2368

2369
	for (i = 1; i < HPAGE_PMD_NR; i++)
2370
		head[i].mem_cgroup = head->mem_cgroup;
2371

2372
	__this_cpu_sub(head->mem_cgroup->stat->count[MEMCG_RSS_HUGE],
2373
		       HPAGE_PMD_NR);
2374
}
2375
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
2376

A
Andrew Morton 已提交
2377
#ifdef CONFIG_MEMCG_SWAP
2378
static void mem_cgroup_swap_statistics(struct mem_cgroup *memcg,
2379
				       int nr_entries)
K
KAMEZAWA Hiroyuki 已提交
2380
{
2381
	this_cpu_add(memcg->stat->count[MEMCG_SWAP], nr_entries);
K
KAMEZAWA Hiroyuki 已提交
2382
}
2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394

/**
 * 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.
 *
2395
 * The caller must have charged to @to, IOW, called page_counter_charge() about
2396 2397 2398
 * both res and memsw, and called css_get().
 */
static int mem_cgroup_move_swap_account(swp_entry_t entry,
2399
				struct mem_cgroup *from, struct mem_cgroup *to)
2400 2401 2402
{
	unsigned short old_id, new_id;

L
Li Zefan 已提交
2403 2404
	old_id = mem_cgroup_id(from);
	new_id = mem_cgroup_id(to);
2405 2406

	if (swap_cgroup_cmpxchg(entry, old_id, new_id) == old_id) {
2407 2408
		mem_cgroup_swap_statistics(from, -1);
		mem_cgroup_swap_statistics(to, 1);
2409 2410 2411 2412 2413 2414
		return 0;
	}
	return -EINVAL;
}
#else
static inline int mem_cgroup_move_swap_account(swp_entry_t entry,
2415
				struct mem_cgroup *from, struct mem_cgroup *to)
2416 2417 2418
{
	return -EINVAL;
}
2419
#endif
K
KAMEZAWA Hiroyuki 已提交
2420

2421
static DEFINE_MUTEX(memcg_limit_mutex);
2422

2423
static int mem_cgroup_resize_limit(struct mem_cgroup *memcg,
2424
				   unsigned long limit)
2425
{
2426 2427 2428
	unsigned long curusage;
	unsigned long oldusage;
	bool enlarge = false;
2429
	int retry_count;
2430
	int ret;
2431 2432 2433 2434 2435 2436

	/*
	 * 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.
	 */
2437 2438
	retry_count = MEM_CGROUP_RECLAIM_RETRIES *
		      mem_cgroup_count_children(memcg);
2439

2440
	oldusage = page_counter_read(&memcg->memory);
2441

2442
	do {
2443 2444 2445 2446
		if (signal_pending(current)) {
			ret = -EINTR;
			break;
		}
2447 2448 2449 2450

		mutex_lock(&memcg_limit_mutex);
		if (limit > memcg->memsw.limit) {
			mutex_unlock(&memcg_limit_mutex);
2451
			ret = -EINVAL;
2452 2453
			break;
		}
2454 2455 2456 2457
		if (limit > memcg->memory.limit)
			enlarge = true;
		ret = page_counter_limit(&memcg->memory, limit);
		mutex_unlock(&memcg_limit_mutex);
2458 2459 2460 2461

		if (!ret)
			break;

2462 2463
		try_to_free_mem_cgroup_pages(memcg, 1, GFP_KERNEL, true);

2464
		curusage = page_counter_read(&memcg->memory);
2465
		/* Usage is reduced ? */
A
Andrew Morton 已提交
2466
		if (curusage >= oldusage)
2467 2468 2469
			retry_count--;
		else
			oldusage = curusage;
2470 2471
	} while (retry_count);

2472 2473
	if (!ret && enlarge)
		memcg_oom_recover(memcg);
2474

2475 2476 2477
	return ret;
}

L
Li Zefan 已提交
2478
static int mem_cgroup_resize_memsw_limit(struct mem_cgroup *memcg,
2479
					 unsigned long limit)
2480
{
2481 2482 2483
	unsigned long curusage;
	unsigned long oldusage;
	bool enlarge = false;
2484
	int retry_count;
2485
	int ret;
2486

2487
	/* see mem_cgroup_resize_res_limit */
2488 2489 2490 2491 2492 2493
	retry_count = MEM_CGROUP_RECLAIM_RETRIES *
		      mem_cgroup_count_children(memcg);

	oldusage = page_counter_read(&memcg->memsw);

	do {
2494 2495 2496 2497
		if (signal_pending(current)) {
			ret = -EINTR;
			break;
		}
2498 2499 2500 2501

		mutex_lock(&memcg_limit_mutex);
		if (limit < memcg->memory.limit) {
			mutex_unlock(&memcg_limit_mutex);
2502 2503 2504
			ret = -EINVAL;
			break;
		}
2505 2506 2507 2508
		if (limit > memcg->memsw.limit)
			enlarge = true;
		ret = page_counter_limit(&memcg->memsw, limit);
		mutex_unlock(&memcg_limit_mutex);
2509 2510 2511 2512

		if (!ret)
			break;

2513 2514
		try_to_free_mem_cgroup_pages(memcg, 1, GFP_KERNEL, false);

2515
		curusage = page_counter_read(&memcg->memsw);
2516
		/* Usage is reduced ? */
2517
		if (curusage >= oldusage)
2518
			retry_count--;
2519 2520
		else
			oldusage = curusage;
2521 2522
	} while (retry_count);

2523 2524
	if (!ret && enlarge)
		memcg_oom_recover(memcg);
2525

2526 2527 2528
	return ret;
}

2529
unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
2530 2531 2532 2533
					    gfp_t gfp_mask,
					    unsigned long *total_scanned)
{
	unsigned long nr_reclaimed = 0;
2534
	struct mem_cgroup_per_node *mz, *next_mz = NULL;
2535 2536
	unsigned long reclaimed;
	int loop = 0;
2537
	struct mem_cgroup_tree_per_node *mctz;
2538
	unsigned long excess;
2539 2540 2541 2542 2543
	unsigned long nr_scanned;

	if (order > 0)
		return 0;

2544
	mctz = soft_limit_tree_node(pgdat->node_id);
2545 2546 2547 2548 2549 2550

	/*
	 * Do not even bother to check the largest node if the root
	 * is empty. Do it lockless to prevent lock bouncing. Races
	 * are acceptable as soft limit is best effort anyway.
	 */
2551
	if (!mctz || RB_EMPTY_ROOT(&mctz->rb_root))
2552 2553
		return 0;

2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567
	/*
	 * 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;
2568
		reclaimed = mem_cgroup_soft_reclaim(mz->memcg, pgdat,
2569 2570 2571
						    gfp_mask, &nr_scanned);
		nr_reclaimed += reclaimed;
		*total_scanned += nr_scanned;
2572
		spin_lock_irq(&mctz->lock);
2573
		__mem_cgroup_remove_exceeded(mz, mctz);
2574 2575 2576 2577 2578 2579

		/*
		 * If we failed to reclaim anything from this memory cgroup
		 * it is time to move on to the next cgroup
		 */
		next_mz = NULL;
2580 2581 2582
		if (!reclaimed)
			next_mz = __mem_cgroup_largest_soft_limit_node(mctz);

2583
		excess = soft_limit_excess(mz->memcg);
2584 2585 2586 2587 2588 2589 2590 2591 2592
		/*
		 * 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 */
2593
		__mem_cgroup_insert_exceeded(mz, mctz, excess);
2594
		spin_unlock_irq(&mctz->lock);
2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611
		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;
}

2612 2613 2614 2615 2616 2617
/*
 * 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.
 */
2618 2619
static inline bool memcg_has_children(struct mem_cgroup *memcg)
{
2620 2621 2622 2623 2624 2625
	bool ret;

	rcu_read_lock();
	ret = css_next_child(NULL, &memcg->css);
	rcu_read_unlock();
	return ret;
2626 2627
}

2628
/*
2629
 * Reclaims as many pages from the given memcg as possible.
2630 2631 2632 2633 2634 2635 2636
 *
 * 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;

2637 2638
	/* we call try-to-free pages for make this cgroup empty */
	lru_add_drain_all();
2639
	/* try to free all pages in this cgroup */
2640
	while (nr_retries && page_counter_read(&memcg->memory)) {
2641
		int progress;
2642

2643 2644 2645
		if (signal_pending(current))
			return -EINTR;

2646 2647
		progress = try_to_free_mem_cgroup_pages(memcg, 1,
							GFP_KERNEL, true);
2648
		if (!progress) {
2649
			nr_retries--;
2650
			/* maybe some writeback is necessary */
2651
			congestion_wait(BLK_RW_ASYNC, HZ/10);
2652
		}
2653 2654

	}
2655 2656

	return 0;
2657 2658
}

2659 2660 2661
static ssize_t mem_cgroup_force_empty_write(struct kernfs_open_file *of,
					    char *buf, size_t nbytes,
					    loff_t off)
2662
{
2663
	struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of));
2664

2665 2666
	if (mem_cgroup_is_root(memcg))
		return -EINVAL;
2667
	return mem_cgroup_force_empty(memcg) ?: nbytes;
2668 2669
}

2670 2671
static u64 mem_cgroup_hierarchy_read(struct cgroup_subsys_state *css,
				     struct cftype *cft)
2672
{
2673
	return mem_cgroup_from_css(css)->use_hierarchy;
2674 2675
}

2676 2677
static int mem_cgroup_hierarchy_write(struct cgroup_subsys_state *css,
				      struct cftype *cft, u64 val)
2678 2679
{
	int retval = 0;
2680
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
T
Tejun Heo 已提交
2681
	struct mem_cgroup *parent_memcg = mem_cgroup_from_css(memcg->css.parent);
2682

2683
	if (memcg->use_hierarchy == val)
2684
		return 0;
2685

2686
	/*
2687
	 * If parent's use_hierarchy is set, we can't make any modifications
2688 2689 2690 2691 2692 2693
	 * 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.
	 */
2694
	if ((!parent_memcg || !parent_memcg->use_hierarchy) &&
2695
				(val == 1 || val == 0)) {
2696
		if (!memcg_has_children(memcg))
2697
			memcg->use_hierarchy = val;
2698 2699 2700 2701
		else
			retval = -EBUSY;
	} else
		retval = -EINVAL;
2702

2703 2704 2705
	return retval;
}

2706
static void tree_stat(struct mem_cgroup *memcg, unsigned long *stat)
2707 2708
{
	struct mem_cgroup *iter;
2709
	int i;
2710

2711
	memset(stat, 0, sizeof(*stat) * MEMCG_NR_STAT);
2712

2713 2714
	for_each_mem_cgroup_tree(iter, memcg) {
		for (i = 0; i < MEMCG_NR_STAT; i++)
2715
			stat[i] += memcg_page_state(iter, i);
2716
	}
2717 2718
}

2719
static void tree_events(struct mem_cgroup *memcg, unsigned long *events)
2720 2721
{
	struct mem_cgroup *iter;
2722
	int i;
2723

2724
	memset(events, 0, sizeof(*events) * MEMCG_NR_EVENTS);
2725

2726 2727
	for_each_mem_cgroup_tree(iter, memcg) {
		for (i = 0; i < MEMCG_NR_EVENTS; i++)
2728
			events[i] += memcg_sum_events(iter, i);
2729
	}
2730 2731
}

2732
static unsigned long mem_cgroup_usage(struct mem_cgroup *memcg, bool swap)
2733
{
2734
	unsigned long val = 0;
2735

2736
	if (mem_cgroup_is_root(memcg)) {
2737 2738 2739
		struct mem_cgroup *iter;

		for_each_mem_cgroup_tree(iter, memcg) {
2740 2741
			val += memcg_page_state(iter, MEMCG_CACHE);
			val += memcg_page_state(iter, MEMCG_RSS);
2742
			if (swap)
2743
				val += memcg_page_state(iter, MEMCG_SWAP);
2744
		}
2745
	} else {
2746
		if (!swap)
2747
			val = page_counter_read(&memcg->memory);
2748
		else
2749
			val = page_counter_read(&memcg->memsw);
2750
	}
2751
	return val;
2752 2753
}

2754 2755 2756 2757 2758 2759 2760
enum {
	RES_USAGE,
	RES_LIMIT,
	RES_MAX_USAGE,
	RES_FAILCNT,
	RES_SOFT_LIMIT,
};
2761

2762
static u64 mem_cgroup_read_u64(struct cgroup_subsys_state *css,
2763
			       struct cftype *cft)
B
Balbir Singh 已提交
2764
{
2765
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
2766
	struct page_counter *counter;
2767

2768
	switch (MEMFILE_TYPE(cft->private)) {
2769
	case _MEM:
2770 2771
		counter = &memcg->memory;
		break;
2772
	case _MEMSWAP:
2773 2774
		counter = &memcg->memsw;
		break;
2775
	case _KMEM:
2776
		counter = &memcg->kmem;
2777
		break;
V
Vladimir Davydov 已提交
2778
	case _TCP:
2779
		counter = &memcg->tcpmem;
V
Vladimir Davydov 已提交
2780
		break;
2781 2782 2783
	default:
		BUG();
	}
2784 2785 2786 2787

	switch (MEMFILE_ATTR(cft->private)) {
	case RES_USAGE:
		if (counter == &memcg->memory)
2788
			return (u64)mem_cgroup_usage(memcg, false) * PAGE_SIZE;
2789
		if (counter == &memcg->memsw)
2790
			return (u64)mem_cgroup_usage(memcg, true) * PAGE_SIZE;
2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802
		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
2803
}
2804

2805
#ifndef CONFIG_SLOB
2806
static int memcg_online_kmem(struct mem_cgroup *memcg)
2807 2808 2809
{
	int memcg_id;

2810 2811 2812
	if (cgroup_memory_nokmem)
		return 0;

2813
	BUG_ON(memcg->kmemcg_id >= 0);
2814
	BUG_ON(memcg->kmem_state);
2815

2816
	memcg_id = memcg_alloc_cache_id();
2817 2818
	if (memcg_id < 0)
		return memcg_id;
2819

2820
	static_branch_inc(&memcg_kmem_enabled_key);
2821
	/*
2822
	 * A memory cgroup is considered kmem-online as soon as it gets
V
Vladimir Davydov 已提交
2823
	 * kmemcg_id. Setting the id after enabling static branching will
2824 2825 2826
	 * guarantee no one starts accounting before all call sites are
	 * patched.
	 */
V
Vladimir Davydov 已提交
2827
	memcg->kmemcg_id = memcg_id;
2828
	memcg->kmem_state = KMEM_ONLINE;
2829
	INIT_LIST_HEAD(&memcg->kmem_caches);
2830 2831

	return 0;
2832 2833
}

2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864 2865 2866
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().
	 */
2867
	rcu_read_lock(); /* can be called from css_free w/o cgroup_mutex */
2868 2869 2870 2871 2872 2873 2874
	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;
	}
2875 2876
	rcu_read_unlock();

2877 2878 2879 2880 2881 2882 2883
	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)
{
2884 2885 2886 2887
	/* css_alloc() failed, offlining didn't happen */
	if (unlikely(memcg->kmem_state == KMEM_ONLINE))
		memcg_offline_kmem(memcg);

2888 2889 2890 2891 2892 2893
	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));
	}
}
2894
#else
2895
static int memcg_online_kmem(struct mem_cgroup *memcg)
2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906
{
	return 0;
}
static void memcg_offline_kmem(struct mem_cgroup *memcg)
{
}
static void memcg_free_kmem(struct mem_cgroup *memcg)
{
}
#endif /* !CONFIG_SLOB */

2907
static int memcg_update_kmem_limit(struct mem_cgroup *memcg,
2908
				   unsigned long limit)
2909
{
2910
	int ret;
2911 2912 2913 2914 2915

	mutex_lock(&memcg_limit_mutex);
	ret = page_counter_limit(&memcg->kmem, limit);
	mutex_unlock(&memcg_limit_mutex);
	return ret;
2916
}
2917

V
Vladimir Davydov 已提交
2918 2919 2920 2921 2922 2923
static int memcg_update_tcp_limit(struct mem_cgroup *memcg, unsigned long limit)
{
	int ret;

	mutex_lock(&memcg_limit_mutex);

2924
	ret = page_counter_limit(&memcg->tcpmem, limit);
V
Vladimir Davydov 已提交
2925 2926 2927
	if (ret)
		goto out;

2928
	if (!memcg->tcpmem_active) {
V
Vladimir Davydov 已提交
2929 2930 2931
		/*
		 * The active flag needs to be written after the static_key
		 * update. This is what guarantees that the socket activation
2932 2933 2934
		 * function is the last one to run. See mem_cgroup_sk_alloc()
		 * for details, and note that we don't mark any socket as
		 * belonging to this memcg until that flag is up.
V
Vladimir Davydov 已提交
2935 2936 2937 2938 2939 2940
		 *
		 * We need to do this, because static_keys will span multiple
		 * sites, but we can't control their order. If we mark a socket
		 * as accounted, but the accounting functions are not patched in
		 * yet, we'll lose accounting.
		 *
2941
		 * We never race with the readers in mem_cgroup_sk_alloc(),
V
Vladimir Davydov 已提交
2942 2943 2944 2945
		 * because when this value change, the code to process it is not
		 * patched in yet.
		 */
		static_branch_inc(&memcg_sockets_enabled_key);
2946
		memcg->tcpmem_active = true;
V
Vladimir Davydov 已提交
2947 2948 2949 2950 2951 2952
	}
out:
	mutex_unlock(&memcg_limit_mutex);
	return ret;
}

2953 2954 2955 2956
/*
 * The user of this function is...
 * RES_LIMIT.
 */
2957 2958
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
2959
{
2960
	struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of));
2961
	unsigned long nr_pages;
2962 2963
	int ret;

2964
	buf = strstrip(buf);
2965
	ret = page_counter_memparse(buf, "-1", &nr_pages);
2966 2967
	if (ret)
		return ret;
2968

2969
	switch (MEMFILE_ATTR(of_cft(of)->private)) {
2970
	case RES_LIMIT:
2971 2972 2973 2974
		if (mem_cgroup_is_root(memcg)) { /* Can't set limit on root */
			ret = -EINVAL;
			break;
		}
2975 2976 2977
		switch (MEMFILE_TYPE(of_cft(of)->private)) {
		case _MEM:
			ret = mem_cgroup_resize_limit(memcg, nr_pages);
2978
			break;
2979 2980
		case _MEMSWAP:
			ret = mem_cgroup_resize_memsw_limit(memcg, nr_pages);
2981
			break;
2982 2983 2984
		case _KMEM:
			ret = memcg_update_kmem_limit(memcg, nr_pages);
			break;
V
Vladimir Davydov 已提交
2985 2986 2987
		case _TCP:
			ret = memcg_update_tcp_limit(memcg, nr_pages);
			break;
2988
		}
2989
		break;
2990 2991 2992
	case RES_SOFT_LIMIT:
		memcg->soft_limit = nr_pages;
		ret = 0;
2993 2994
		break;
	}
2995
	return ret ?: nbytes;
B
Balbir Singh 已提交
2996 2997
}

2998 2999
static ssize_t mem_cgroup_reset(struct kernfs_open_file *of, char *buf,
				size_t nbytes, loff_t off)
3000
{
3001
	struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of));
3002
	struct page_counter *counter;
3003

3004 3005 3006 3007 3008 3009 3010 3011 3012 3013
	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;
V
Vladimir Davydov 已提交
3014
	case _TCP:
3015
		counter = &memcg->tcpmem;
V
Vladimir Davydov 已提交
3016
		break;
3017 3018 3019
	default:
		BUG();
	}
3020

3021
	switch (MEMFILE_ATTR(of_cft(of)->private)) {
3022
	case RES_MAX_USAGE:
3023
		page_counter_reset_watermark(counter);
3024 3025
		break;
	case RES_FAILCNT:
3026
		counter->failcnt = 0;
3027
		break;
3028 3029
	default:
		BUG();
3030
	}
3031

3032
	return nbytes;
3033 3034
}

3035
static u64 mem_cgroup_move_charge_read(struct cgroup_subsys_state *css,
3036 3037
					struct cftype *cft)
{
3038
	return mem_cgroup_from_css(css)->move_charge_at_immigrate;
3039 3040
}

3041
#ifdef CONFIG_MMU
3042
static int mem_cgroup_move_charge_write(struct cgroup_subsys_state *css,
3043 3044
					struct cftype *cft, u64 val)
{
3045
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
3046

3047
	if (val & ~MOVE_MASK)
3048
		return -EINVAL;
3049

3050
	/*
3051 3052 3053 3054
	 * 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.
3055
	 */
3056
	memcg->move_charge_at_immigrate = val;
3057 3058
	return 0;
}
3059
#else
3060
static int mem_cgroup_move_charge_write(struct cgroup_subsys_state *css,
3061 3062 3063 3064 3065
					struct cftype *cft, u64 val)
{
	return -ENOSYS;
}
#endif
3066

3067
#ifdef CONFIG_NUMA
3068
static int memcg_numa_stat_show(struct seq_file *m, void *v)
3069
{
3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081
	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;
3082
	int nid;
3083
	unsigned long nr;
3084
	struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(m));
3085

3086 3087 3088 3089 3090 3091 3092 3093 3094
	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');
3095 3096
	}

3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111
	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');
3112 3113 3114 3115 3116 3117
	}

	return 0;
}
#endif /* CONFIG_NUMA */

3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129 3130 3131 3132
/* Universal VM events cgroup1 shows, original sort order */
unsigned int memcg1_events[] = {
	PGPGIN,
	PGPGOUT,
	PGFAULT,
	PGMAJFAULT,
};

static const char *const memcg1_event_names[] = {
	"pgpgin",
	"pgpgout",
	"pgfault",
	"pgmajfault",
};

3133
static int memcg_stat_show(struct seq_file *m, void *v)
3134
{
3135
	struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(m));
3136
	unsigned long memory, memsw;
3137 3138
	struct mem_cgroup *mi;
	unsigned int i;
3139

3140
	BUILD_BUG_ON(ARRAY_SIZE(memcg1_stat_names) != ARRAY_SIZE(memcg1_stats));
3141 3142
	BUILD_BUG_ON(ARRAY_SIZE(mem_cgroup_lru_names) != NR_LRU_LISTS);

3143 3144
	for (i = 0; i < ARRAY_SIZE(memcg1_stats); i++) {
		if (memcg1_stats[i] == MEMCG_SWAP && !do_memsw_account())
3145
			continue;
3146
		seq_printf(m, "%s %lu\n", memcg1_stat_names[i],
3147
			   memcg_page_state(memcg, memcg1_stats[i]) *
3148
			   PAGE_SIZE);
3149
	}
L
Lee Schermerhorn 已提交
3150

3151 3152
	for (i = 0; i < ARRAY_SIZE(memcg1_events); i++)
		seq_printf(m, "%s %lu\n", memcg1_event_names[i],
3153
			   memcg_sum_events(memcg, memcg1_events[i]));
3154 3155 3156 3157 3158

	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
3159
	/* Hierarchical information */
3160 3161 3162 3163
	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);
3164
	}
3165 3166
	seq_printf(m, "hierarchical_memory_limit %llu\n",
		   (u64)memory * PAGE_SIZE);
3167
	if (do_memsw_account())
3168 3169
		seq_printf(m, "hierarchical_memsw_limit %llu\n",
			   (u64)memsw * PAGE_SIZE);
K
KOSAKI Motohiro 已提交
3170

3171
	for (i = 0; i < ARRAY_SIZE(memcg1_stats); i++) {
3172
		unsigned long long val = 0;
3173

3174
		if (memcg1_stats[i] == MEMCG_SWAP && !do_memsw_account())
3175
			continue;
3176
		for_each_mem_cgroup_tree(mi, memcg)
3177
			val += memcg_page_state(mi, memcg1_stats[i]) *
3178 3179
			PAGE_SIZE;
		seq_printf(m, "total_%s %llu\n", memcg1_stat_names[i], val);
3180 3181
	}

3182
	for (i = 0; i < ARRAY_SIZE(memcg1_events); i++) {
3183 3184 3185
		unsigned long long val = 0;

		for_each_mem_cgroup_tree(mi, memcg)
3186
			val += memcg_sum_events(mi, memcg1_events[i]);
3187
		seq_printf(m, "total_%s %llu\n", memcg1_event_names[i], val);
3188 3189 3190 3191 3192 3193 3194 3195
	}

	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);
3196
	}
K
KAMEZAWA Hiroyuki 已提交
3197

K
KOSAKI Motohiro 已提交
3198 3199
#ifdef CONFIG_DEBUG_VM
	{
3200 3201
		pg_data_t *pgdat;
		struct mem_cgroup_per_node *mz;
3202
		struct zone_reclaim_stat *rstat;
K
KOSAKI Motohiro 已提交
3203 3204 3205
		unsigned long recent_rotated[2] = {0, 0};
		unsigned long recent_scanned[2] = {0, 0};

3206 3207 3208
		for_each_online_pgdat(pgdat) {
			mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
			rstat = &mz->lruvec.reclaim_stat;
K
KOSAKI Motohiro 已提交
3209

3210 3211 3212 3213 3214
			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];
		}
3215 3216 3217 3218
		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
3219 3220 3221
	}
#endif

3222 3223 3224
	return 0;
}

3225 3226
static u64 mem_cgroup_swappiness_read(struct cgroup_subsys_state *css,
				      struct cftype *cft)
K
KOSAKI Motohiro 已提交
3227
{
3228
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
K
KOSAKI Motohiro 已提交
3229

3230
	return mem_cgroup_swappiness(memcg);
K
KOSAKI Motohiro 已提交
3231 3232
}

3233 3234
static int mem_cgroup_swappiness_write(struct cgroup_subsys_state *css,
				       struct cftype *cft, u64 val)
K
KOSAKI Motohiro 已提交
3235
{
3236
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
K
KOSAKI Motohiro 已提交
3237

3238
	if (val > 100)
K
KOSAKI Motohiro 已提交
3239 3240
		return -EINVAL;

3241
	if (css->parent)
3242 3243 3244
		memcg->swappiness = val;
	else
		vm_swappiness = val;
3245

K
KOSAKI Motohiro 已提交
3246 3247 3248
	return 0;
}

3249 3250 3251
static void __mem_cgroup_threshold(struct mem_cgroup *memcg, bool swap)
{
	struct mem_cgroup_threshold_ary *t;
3252
	unsigned long usage;
3253 3254 3255 3256
	int i;

	rcu_read_lock();
	if (!swap)
3257
		t = rcu_dereference(memcg->thresholds.primary);
3258
	else
3259
		t = rcu_dereference(memcg->memsw_thresholds.primary);
3260 3261 3262 3263

	if (!t)
		goto unlock;

3264
	usage = mem_cgroup_usage(memcg, swap);
3265 3266

	/*
3267
	 * current_threshold points to threshold just below or equal to usage.
3268 3269 3270
	 * If it's not true, a threshold was crossed after last
	 * call of __mem_cgroup_threshold().
	 */
3271
	i = t->current_threshold;
3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294

	/*
	 * 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 */
3295
	t->current_threshold = i - 1;
3296 3297 3298 3299 3300 3301
unlock:
	rcu_read_unlock();
}

static void mem_cgroup_threshold(struct mem_cgroup *memcg)
{
3302 3303
	while (memcg) {
		__mem_cgroup_threshold(memcg, false);
3304
		if (do_memsw_account())
3305 3306 3307 3308
			__mem_cgroup_threshold(memcg, true);

		memcg = parent_mem_cgroup(memcg);
	}
3309 3310 3311 3312 3313 3314 3315
}

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

3316 3317 3318 3319 3320 3321 3322
	if (_a->threshold > _b->threshold)
		return 1;

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

	return 0;
3323 3324
}

3325
static int mem_cgroup_oom_notify_cb(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
3326 3327 3328
{
	struct mem_cgroup_eventfd_list *ev;

3329 3330
	spin_lock(&memcg_oom_lock);

3331
	list_for_each_entry(ev, &memcg->oom_notify, list)
K
KAMEZAWA Hiroyuki 已提交
3332
		eventfd_signal(ev->eventfd, 1);
3333 3334

	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
3335 3336 3337
	return 0;
}

3338
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
3339
{
K
KAMEZAWA Hiroyuki 已提交
3340 3341
	struct mem_cgroup *iter;

3342
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
3343
		mem_cgroup_oom_notify_cb(iter);
K
KAMEZAWA Hiroyuki 已提交
3344 3345
}

3346
static int __mem_cgroup_usage_register_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
3347
	struct eventfd_ctx *eventfd, const char *args, enum res_type type)
3348
{
3349 3350
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
3351 3352
	unsigned long threshold;
	unsigned long usage;
3353
	int i, size, ret;
3354

3355
	ret = page_counter_memparse(args, "-1", &threshold);
3356 3357 3358 3359
	if (ret)
		return ret;

	mutex_lock(&memcg->thresholds_lock);
3360

3361
	if (type == _MEM) {
3362
		thresholds = &memcg->thresholds;
3363
		usage = mem_cgroup_usage(memcg, false);
3364
	} else if (type == _MEMSWAP) {
3365
		thresholds = &memcg->memsw_thresholds;
3366
		usage = mem_cgroup_usage(memcg, true);
3367
	} else
3368 3369 3370
		BUG();

	/* Check if a threshold crossed before adding a new one */
3371
	if (thresholds->primary)
3372 3373
		__mem_cgroup_threshold(memcg, type == _MEMSWAP);

3374
	size = thresholds->primary ? thresholds->primary->size + 1 : 1;
3375 3376

	/* Allocate memory for new array of thresholds */
3377
	new = kmalloc(sizeof(*new) + size * sizeof(struct mem_cgroup_threshold),
3378
			GFP_KERNEL);
3379
	if (!new) {
3380 3381 3382
		ret = -ENOMEM;
		goto unlock;
	}
3383
	new->size = size;
3384 3385

	/* Copy thresholds (if any) to new array */
3386 3387
	if (thresholds->primary) {
		memcpy(new->entries, thresholds->primary->entries, (size - 1) *
3388
				sizeof(struct mem_cgroup_threshold));
3389 3390
	}

3391
	/* Add new threshold */
3392 3393
	new->entries[size - 1].eventfd = eventfd;
	new->entries[size - 1].threshold = threshold;
3394 3395

	/* Sort thresholds. Registering of new threshold isn't time-critical */
3396
	sort(new->entries, size, sizeof(struct mem_cgroup_threshold),
3397 3398 3399
			compare_thresholds, NULL);

	/* Find current threshold */
3400
	new->current_threshold = -1;
3401
	for (i = 0; i < size; i++) {
3402
		if (new->entries[i].threshold <= usage) {
3403
			/*
3404 3405
			 * new->current_threshold will not be used until
			 * rcu_assign_pointer(), so it's safe to increment
3406 3407
			 * it here.
			 */
3408
			++new->current_threshold;
3409 3410
		} else
			break;
3411 3412
	}

3413 3414 3415 3416 3417
	/* Free old spare buffer and save old primary buffer as spare */
	kfree(thresholds->spare);
	thresholds->spare = thresholds->primary;

	rcu_assign_pointer(thresholds->primary, new);
3418

3419
	/* To be sure that nobody uses thresholds */
3420 3421 3422 3423 3424 3425 3426 3427
	synchronize_rcu();

unlock:
	mutex_unlock(&memcg->thresholds_lock);

	return ret;
}

3428
static int mem_cgroup_usage_register_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
3429 3430
	struct eventfd_ctx *eventfd, const char *args)
{
3431
	return __mem_cgroup_usage_register_event(memcg, eventfd, args, _MEM);
T
Tejun Heo 已提交
3432 3433
}

3434
static int memsw_cgroup_usage_register_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
3435 3436
	struct eventfd_ctx *eventfd, const char *args)
{
3437
	return __mem_cgroup_usage_register_event(memcg, eventfd, args, _MEMSWAP);
T
Tejun Heo 已提交
3438 3439
}

3440
static void __mem_cgroup_usage_unregister_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
3441
	struct eventfd_ctx *eventfd, enum res_type type)
3442
{
3443 3444
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
3445
	unsigned long usage;
3446
	int i, j, size;
3447 3448

	mutex_lock(&memcg->thresholds_lock);
3449 3450

	if (type == _MEM) {
3451
		thresholds = &memcg->thresholds;
3452
		usage = mem_cgroup_usage(memcg, false);
3453
	} else if (type == _MEMSWAP) {
3454
		thresholds = &memcg->memsw_thresholds;
3455
		usage = mem_cgroup_usage(memcg, true);
3456
	} else
3457 3458
		BUG();

3459 3460 3461
	if (!thresholds->primary)
		goto unlock;

3462 3463 3464 3465
	/* Check if a threshold crossed before removing */
	__mem_cgroup_threshold(memcg, type == _MEMSWAP);

	/* Calculate new number of threshold */
3466 3467 3468
	size = 0;
	for (i = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd != eventfd)
3469 3470 3471
			size++;
	}

3472
	new = thresholds->spare;
3473

3474 3475
	/* Set thresholds array to NULL if we don't have thresholds */
	if (!size) {
3476 3477
		kfree(new);
		new = NULL;
3478
		goto swap_buffers;
3479 3480
	}

3481
	new->size = size;
3482 3483

	/* Copy thresholds and find current threshold */
3484 3485 3486
	new->current_threshold = -1;
	for (i = 0, j = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd == eventfd)
3487 3488
			continue;

3489
		new->entries[j] = thresholds->primary->entries[i];
3490
		if (new->entries[j].threshold <= usage) {
3491
			/*
3492
			 * new->current_threshold will not be used
3493 3494 3495
			 * until rcu_assign_pointer(), so it's safe to increment
			 * it here.
			 */
3496
			++new->current_threshold;
3497 3498 3499 3500
		}
		j++;
	}

3501
swap_buffers:
3502 3503
	/* Swap primary and spare array */
	thresholds->spare = thresholds->primary;
3504

3505
	rcu_assign_pointer(thresholds->primary, new);
3506

3507
	/* To be sure that nobody uses thresholds */
3508
	synchronize_rcu();
3509 3510 3511 3512 3513 3514

	/* If all events are unregistered, free the spare array */
	if (!new) {
		kfree(thresholds->spare);
		thresholds->spare = NULL;
	}
3515
unlock:
3516 3517
	mutex_unlock(&memcg->thresholds_lock);
}
3518

3519
static void mem_cgroup_usage_unregister_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
3520 3521
	struct eventfd_ctx *eventfd)
{
3522
	return __mem_cgroup_usage_unregister_event(memcg, eventfd, _MEM);
T
Tejun Heo 已提交
3523 3524
}

3525
static void memsw_cgroup_usage_unregister_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
3526 3527
	struct eventfd_ctx *eventfd)
{
3528
	return __mem_cgroup_usage_unregister_event(memcg, eventfd, _MEMSWAP);
T
Tejun Heo 已提交
3529 3530
}

3531
static int mem_cgroup_oom_register_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
3532
	struct eventfd_ctx *eventfd, const char *args)
K
KAMEZAWA Hiroyuki 已提交
3533 3534 3535 3536 3537 3538 3539
{
	struct mem_cgroup_eventfd_list *event;

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

3540
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
3541 3542 3543 3544 3545

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

	/* already in OOM ? */
3546
	if (memcg->under_oom)
K
KAMEZAWA Hiroyuki 已提交
3547
		eventfd_signal(eventfd, 1);
3548
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
3549 3550 3551 3552

	return 0;
}

3553
static void mem_cgroup_oom_unregister_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
3554
	struct eventfd_ctx *eventfd)
K
KAMEZAWA Hiroyuki 已提交
3555 3556 3557
{
	struct mem_cgroup_eventfd_list *ev, *tmp;

3558
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
3559

3560
	list_for_each_entry_safe(ev, tmp, &memcg->oom_notify, list) {
K
KAMEZAWA Hiroyuki 已提交
3561 3562 3563 3564 3565 3566
		if (ev->eventfd == eventfd) {
			list_del(&ev->list);
			kfree(ev);
		}
	}

3567
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
3568 3569
}

3570
static int mem_cgroup_oom_control_read(struct seq_file *sf, void *v)
3571
{
3572
	struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(sf));
3573

3574
	seq_printf(sf, "oom_kill_disable %d\n", memcg->oom_kill_disable);
3575
	seq_printf(sf, "under_oom %d\n", (bool)memcg->under_oom);
3576
	seq_printf(sf, "oom_kill %lu\n", memcg_sum_events(memcg, OOM_KILL));
3577 3578 3579
	return 0;
}

3580
static int mem_cgroup_oom_control_write(struct cgroup_subsys_state *css,
3581 3582
	struct cftype *cft, u64 val)
{
3583
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
3584 3585

	/* cannot set to root cgroup and only 0 and 1 are allowed */
3586
	if (!css->parent || !((val == 0) || (val == 1)))
3587 3588
		return -EINVAL;

3589
	memcg->oom_kill_disable = val;
3590
	if (!val)
3591
		memcg_oom_recover(memcg);
3592

3593 3594 3595
	return 0;
}

3596 3597 3598 3599 3600 3601 3602
#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
3603 3604 3605 3606 3607 3608 3609 3610 3611 3612
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);
}

3613 3614 3615 3616 3617
static void memcg_wb_domain_size_changed(struct mem_cgroup *memcg)
{
	wb_domain_size_changed(&memcg->cgwb_domain);
}

T
Tejun Heo 已提交
3618 3619 3620 3621 3622 3623 3624 3625 3626 3627
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;
}

3628 3629 3630
/**
 * mem_cgroup_wb_stats - retrieve writeback related stats from its memcg
 * @wb: bdi_writeback in question
3631 3632
 * @pfilepages: out parameter for number of file pages
 * @pheadroom: out parameter for number of allocatable pages according to memcg
3633 3634 3635
 * @pdirty: out parameter for number of dirty pages
 * @pwriteback: out parameter for number of pages under writeback
 *
3636 3637 3638
 * 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.
3639
 *
3640 3641 3642 3643 3644
 * 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.
3645
 */
3646 3647 3648
void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
			 unsigned long *pheadroom, unsigned long *pdirty,
			 unsigned long *pwriteback)
3649 3650 3651 3652
{
	struct mem_cgroup *memcg = mem_cgroup_from_css(wb->memcg_css);
	struct mem_cgroup *parent;

3653
	*pdirty = memcg_page_state(memcg, NR_FILE_DIRTY);
3654 3655

	/* this should eventually include NR_UNSTABLE_NFS */
3656
	*pwriteback = memcg_page_state(memcg, NR_WRITEBACK);
3657 3658 3659
	*pfilepages = mem_cgroup_nr_lru_pages(memcg, (1 << LRU_INACTIVE_FILE) |
						     (1 << LRU_ACTIVE_FILE));
	*pheadroom = PAGE_COUNTER_MAX;
3660 3661 3662 3663 3664

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

3665
		*pheadroom = min(*pheadroom, ceiling - min(ceiling, used));
3666 3667 3668 3669
		memcg = parent;
	}
}

T
Tejun Heo 已提交
3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680
#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)
{
}

3681 3682 3683 3684
static void memcg_wb_domain_size_changed(struct mem_cgroup *memcg)
{
}

3685 3686
#endif	/* CONFIG_CGROUP_WRITEBACK */

3687 3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699
/*
 * 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.
 */

3700 3701 3702 3703 3704
/*
 * Unregister event and free resources.
 *
 * Gets called from workqueue.
 */
3705
static void memcg_event_remove(struct work_struct *work)
3706
{
3707 3708
	struct mem_cgroup_event *event =
		container_of(work, struct mem_cgroup_event, remove);
3709
	struct mem_cgroup *memcg = event->memcg;
3710 3711 3712

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

3713
	event->unregister_event(memcg, event->eventfd);
3714 3715 3716 3717 3718 3719

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

	eventfd_ctx_put(event->eventfd);
	kfree(event);
3720
	css_put(&memcg->css);
3721 3722 3723 3724 3725 3726 3727
}

/*
 * Gets called on POLLHUP on eventfd when user closes it.
 *
 * Called with wqh->lock held and interrupts disabled.
 */
3728
static int memcg_event_wake(wait_queue_entry_t *wait, unsigned mode,
3729
			    int sync, void *key)
3730
{
3731 3732
	struct mem_cgroup_event *event =
		container_of(wait, struct mem_cgroup_event, wait);
3733
	struct mem_cgroup *memcg = event->memcg;
3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745
	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.
		 */
3746
		spin_lock(&memcg->event_list_lock);
3747 3748 3749 3750 3751 3752 3753 3754
		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);
		}
3755
		spin_unlock(&memcg->event_list_lock);
3756 3757 3758 3759 3760
	}

	return 0;
}

3761
static void memcg_event_ptable_queue_proc(struct file *file,
3762 3763
		wait_queue_head_t *wqh, poll_table *pt)
{
3764 3765
	struct mem_cgroup_event *event =
		container_of(pt, struct mem_cgroup_event, pt);
3766 3767 3768 3769 3770 3771

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

/*
3772 3773
 * DO NOT USE IN NEW FILES.
 *
3774 3775 3776 3777 3778
 * 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.
 */
3779 3780
static ssize_t memcg_write_event_control(struct kernfs_open_file *of,
					 char *buf, size_t nbytes, loff_t off)
3781
{
3782
	struct cgroup_subsys_state *css = of_css(of);
3783
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
3784
	struct mem_cgroup_event *event;
3785 3786 3787 3788
	struct cgroup_subsys_state *cfile_css;
	unsigned int efd, cfd;
	struct fd efile;
	struct fd cfile;
3789
	const char *name;
3790 3791 3792
	char *endp;
	int ret;

3793 3794 3795
	buf = strstrip(buf);

	efd = simple_strtoul(buf, &endp, 10);
3796 3797
	if (*endp != ' ')
		return -EINVAL;
3798
	buf = endp + 1;
3799

3800
	cfd = simple_strtoul(buf, &endp, 10);
3801 3802
	if ((*endp != ' ') && (*endp != '\0'))
		return -EINVAL;
3803
	buf = endp + 1;
3804 3805 3806 3807 3808

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

3809
	event->memcg = memcg;
3810
	INIT_LIST_HEAD(&event->list);
3811 3812 3813
	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);
3814 3815 3816 3817 3818 3819 3820 3821 3822 3823 3824 3825 3826 3827 3828 3829 3830 3831 3832 3833 3834 3835 3836 3837 3838

	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;

3839 3840 3841 3842 3843
	/*
	 * 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.
3844 3845
	 *
	 * DO NOT ADD NEW FILES.
3846
	 */
A
Al Viro 已提交
3847
	name = cfile.file->f_path.dentry->d_name.name;
3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858

	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
3859 3860
		event->register_event = memsw_cgroup_usage_register_event;
		event->unregister_event = memsw_cgroup_usage_unregister_event;
3861 3862 3863 3864 3865
	} else {
		ret = -EINVAL;
		goto out_put_cfile;
	}

3866
	/*
3867 3868 3869
	 * 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.
3870
	 */
A
Al Viro 已提交
3871
	cfile_css = css_tryget_online_from_dir(cfile.file->f_path.dentry->d_parent,
3872
					       &memory_cgrp_subsys);
3873
	ret = -EINVAL;
3874
	if (IS_ERR(cfile_css))
3875
		goto out_put_cfile;
3876 3877
	if (cfile_css != css) {
		css_put(cfile_css);
3878
		goto out_put_cfile;
3879
	}
3880

3881
	ret = event->register_event(memcg, event->eventfd, buf);
3882 3883 3884 3885 3886
	if (ret)
		goto out_put_css;

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

3887 3888 3889
	spin_lock(&memcg->event_list_lock);
	list_add(&event->list, &memcg->event_list);
	spin_unlock(&memcg->event_list_lock);
3890 3891 3892 3893

	fdput(cfile);
	fdput(efile);

3894
	return nbytes;
3895 3896

out_put_css:
3897
	css_put(css);
3898 3899 3900 3901 3902 3903 3904 3905 3906 3907 3908 3909
out_put_cfile:
	fdput(cfile);
out_put_eventfd:
	eventfd_ctx_put(event->eventfd);
out_put_efile:
	fdput(efile);
out_kfree:
	kfree(event);

	return ret;
}

3910
static struct cftype mem_cgroup_legacy_files[] = {
B
Balbir Singh 已提交
3911
	{
3912
		.name = "usage_in_bytes",
3913
		.private = MEMFILE_PRIVATE(_MEM, RES_USAGE),
3914
		.read_u64 = mem_cgroup_read_u64,
B
Balbir Singh 已提交
3915
	},
3916 3917
	{
		.name = "max_usage_in_bytes",
3918
		.private = MEMFILE_PRIVATE(_MEM, RES_MAX_USAGE),
3919
		.write = mem_cgroup_reset,
3920
		.read_u64 = mem_cgroup_read_u64,
3921
	},
B
Balbir Singh 已提交
3922
	{
3923
		.name = "limit_in_bytes",
3924
		.private = MEMFILE_PRIVATE(_MEM, RES_LIMIT),
3925
		.write = mem_cgroup_write,
3926
		.read_u64 = mem_cgroup_read_u64,
B
Balbir Singh 已提交
3927
	},
3928 3929 3930
	{
		.name = "soft_limit_in_bytes",
		.private = MEMFILE_PRIVATE(_MEM, RES_SOFT_LIMIT),
3931
		.write = mem_cgroup_write,
3932
		.read_u64 = mem_cgroup_read_u64,
3933
	},
B
Balbir Singh 已提交
3934 3935
	{
		.name = "failcnt",
3936
		.private = MEMFILE_PRIVATE(_MEM, RES_FAILCNT),
3937
		.write = mem_cgroup_reset,
3938
		.read_u64 = mem_cgroup_read_u64,
B
Balbir Singh 已提交
3939
	},
3940 3941
	{
		.name = "stat",
3942
		.seq_show = memcg_stat_show,
3943
	},
3944 3945
	{
		.name = "force_empty",
3946
		.write = mem_cgroup_force_empty_write,
3947
	},
3948 3949 3950 3951 3952
	{
		.name = "use_hierarchy",
		.write_u64 = mem_cgroup_hierarchy_write,
		.read_u64 = mem_cgroup_hierarchy_read,
	},
3953
	{
3954
		.name = "cgroup.event_control",		/* XXX: for compat */
3955
		.write = memcg_write_event_control,
3956
		.flags = CFTYPE_NO_PREFIX | CFTYPE_WORLD_WRITABLE,
3957
	},
K
KOSAKI Motohiro 已提交
3958 3959 3960 3961 3962
	{
		.name = "swappiness",
		.read_u64 = mem_cgroup_swappiness_read,
		.write_u64 = mem_cgroup_swappiness_write,
	},
3963 3964 3965 3966 3967
	{
		.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
3968 3969
	{
		.name = "oom_control",
3970
		.seq_show = mem_cgroup_oom_control_read,
3971
		.write_u64 = mem_cgroup_oom_control_write,
K
KAMEZAWA Hiroyuki 已提交
3972 3973
		.private = MEMFILE_PRIVATE(_OOM_TYPE, OOM_CONTROL),
	},
3974 3975 3976
	{
		.name = "pressure_level",
	},
3977 3978 3979
#ifdef CONFIG_NUMA
	{
		.name = "numa_stat",
3980
		.seq_show = memcg_numa_stat_show,
3981 3982
	},
#endif
3983 3984 3985
	{
		.name = "kmem.limit_in_bytes",
		.private = MEMFILE_PRIVATE(_KMEM, RES_LIMIT),
3986
		.write = mem_cgroup_write,
3987
		.read_u64 = mem_cgroup_read_u64,
3988 3989 3990 3991
	},
	{
		.name = "kmem.usage_in_bytes",
		.private = MEMFILE_PRIVATE(_KMEM, RES_USAGE),
3992
		.read_u64 = mem_cgroup_read_u64,
3993 3994 3995 3996
	},
	{
		.name = "kmem.failcnt",
		.private = MEMFILE_PRIVATE(_KMEM, RES_FAILCNT),
3997
		.write = mem_cgroup_reset,
3998
		.read_u64 = mem_cgroup_read_u64,
3999 4000 4001 4002
	},
	{
		.name = "kmem.max_usage_in_bytes",
		.private = MEMFILE_PRIVATE(_KMEM, RES_MAX_USAGE),
4003
		.write = mem_cgroup_reset,
4004
		.read_u64 = mem_cgroup_read_u64,
4005
	},
4006 4007 4008
#ifdef CONFIG_SLABINFO
	{
		.name = "kmem.slabinfo",
4009 4010 4011
		.seq_start = memcg_slab_start,
		.seq_next = memcg_slab_next,
		.seq_stop = memcg_slab_stop,
4012
		.seq_show = memcg_slab_show,
4013 4014
	},
#endif
V
Vladimir Davydov 已提交
4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037
	{
		.name = "kmem.tcp.limit_in_bytes",
		.private = MEMFILE_PRIVATE(_TCP, RES_LIMIT),
		.write = mem_cgroup_write,
		.read_u64 = mem_cgroup_read_u64,
	},
	{
		.name = "kmem.tcp.usage_in_bytes",
		.private = MEMFILE_PRIVATE(_TCP, RES_USAGE),
		.read_u64 = mem_cgroup_read_u64,
	},
	{
		.name = "kmem.tcp.failcnt",
		.private = MEMFILE_PRIVATE(_TCP, RES_FAILCNT),
		.write = mem_cgroup_reset,
		.read_u64 = mem_cgroup_read_u64,
	},
	{
		.name = "kmem.tcp.max_usage_in_bytes",
		.private = MEMFILE_PRIVATE(_TCP, RES_MAX_USAGE),
		.write = mem_cgroup_reset,
		.read_u64 = mem_cgroup_read_u64,
	},
4038
	{ },	/* terminate */
4039
};
4040

4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066
/*
 * Private memory cgroup IDR
 *
 * Swap-out records and page cache shadow entries need to store memcg
 * references in constrained space, so we maintain an ID space that is
 * limited to 16 bit (MEM_CGROUP_ID_MAX), limiting the total number of
 * memory-controlled cgroups to 64k.
 *
 * However, there usually are many references to the oflline CSS after
 * the cgroup has been destroyed, such as page cache or reclaimable
 * slab objects, that don't need to hang on to the ID. We want to keep
 * those dead CSS from occupying IDs, or we might quickly exhaust the
 * relatively small ID space and prevent the creation of new cgroups
 * even when there are much fewer than 64k cgroups - possibly none.
 *
 * Maintain a private 16-bit ID space for memcg, and allow the ID to
 * be freed and recycled when it's no longer needed, which is usually
 * when the CSS is offlined.
 *
 * The only exception to that are records of swapped out tmpfs/shmem
 * pages that need to be attributed to live ancestors on swapin. But
 * those references are manageable from userspace.
 */

static DEFINE_IDR(mem_cgroup_idr);

4067
static void mem_cgroup_id_get_many(struct mem_cgroup *memcg, unsigned int n)
4068
{
4069
	VM_BUG_ON(atomic_read(&memcg->id.ref) <= 0);
4070
	atomic_add(n, &memcg->id.ref);
4071 4072
}

4073
static void mem_cgroup_id_put_many(struct mem_cgroup *memcg, unsigned int n)
4074
{
4075
	VM_BUG_ON(atomic_read(&memcg->id.ref) < n);
4076
	if (atomic_sub_and_test(n, &memcg->id.ref)) {
4077 4078 4079 4080 4081 4082 4083 4084
		idr_remove(&mem_cgroup_idr, memcg->id.id);
		memcg->id.id = 0;

		/* Memcg ID pins CSS */
		css_put(&memcg->css);
	}
}

4085 4086 4087 4088 4089 4090 4091 4092 4093 4094
static inline void mem_cgroup_id_get(struct mem_cgroup *memcg)
{
	mem_cgroup_id_get_many(memcg, 1);
}

static inline void mem_cgroup_id_put(struct mem_cgroup *memcg)
{
	mem_cgroup_id_put_many(memcg, 1);
}

4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105 4106
/**
 * mem_cgroup_from_id - look up a memcg from a memcg id
 * @id: the memcg id to look up
 *
 * Caller must hold rcu_read_lock().
 */
struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
{
	WARN_ON_ONCE(!rcu_read_lock_held());
	return idr_find(&mem_cgroup_idr, id);
}

4107
static int alloc_mem_cgroup_per_node_info(struct mem_cgroup *memcg, int node)
4108 4109
{
	struct mem_cgroup_per_node *pn;
4110
	int tmp = node;
4111 4112 4113 4114 4115 4116 4117 4118
	/*
	 * 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.
	 */
4119 4120
	if (!node_state(node, N_NORMAL_MEMORY))
		tmp = -1;
4121
	pn = kzalloc_node(sizeof(*pn), GFP_KERNEL, tmp);
4122 4123
	if (!pn)
		return 1;
4124

4125 4126 4127 4128 4129
	lruvec_init(&pn->lruvec);
	pn->usage_in_excess = 0;
	pn->on_tree = false;
	pn->memcg = memcg;

4130
	memcg->nodeinfo[node] = pn;
4131 4132 4133
	return 0;
}

4134
static void free_mem_cgroup_per_node_info(struct mem_cgroup *memcg, int node)
4135
{
4136
	kfree(memcg->nodeinfo[node]);
4137 4138
}

4139
static void __mem_cgroup_free(struct mem_cgroup *memcg)
4140
{
4141
	int node;
4142

4143
	for_each_node(node)
4144
		free_mem_cgroup_per_node_info(memcg, node);
4145
	free_percpu(memcg->stat);
4146
	kfree(memcg);
4147
}
4148

4149 4150 4151 4152 4153 4154
static void mem_cgroup_free(struct mem_cgroup *memcg)
{
	memcg_wb_domain_exit(memcg);
	__mem_cgroup_free(memcg);
}

4155
static struct mem_cgroup *mem_cgroup_alloc(void)
B
Balbir Singh 已提交
4156
{
4157
	struct mem_cgroup *memcg;
4158
	size_t size;
4159
	int node;
B
Balbir Singh 已提交
4160

4161 4162 4163 4164
	size = sizeof(struct mem_cgroup);
	size += nr_node_ids * sizeof(struct mem_cgroup_per_node *);

	memcg = kzalloc(size, GFP_KERNEL);
4165
	if (!memcg)
4166 4167
		return NULL;

4168 4169 4170 4171 4172 4173
	memcg->id.id = idr_alloc(&mem_cgroup_idr, NULL,
				 1, MEM_CGROUP_ID_MAX,
				 GFP_KERNEL);
	if (memcg->id.id < 0)
		goto fail;

4174 4175 4176
	memcg->stat = alloc_percpu(struct mem_cgroup_stat_cpu);
	if (!memcg->stat)
		goto fail;
4177

B
Bob Liu 已提交
4178
	for_each_node(node)
4179
		if (alloc_mem_cgroup_per_node_info(memcg, node))
4180
			goto fail;
4181

4182 4183
	if (memcg_wb_domain_init(memcg, GFP_KERNEL))
		goto fail;
4184

4185
	INIT_WORK(&memcg->high_work, high_work_func);
4186 4187 4188 4189
	memcg->last_scanned_node = MAX_NUMNODES;
	INIT_LIST_HEAD(&memcg->oom_notify);
	mutex_init(&memcg->thresholds_lock);
	spin_lock_init(&memcg->move_lock);
4190
	vmpressure_init(&memcg->vmpressure);
4191 4192
	INIT_LIST_HEAD(&memcg->event_list);
	spin_lock_init(&memcg->event_list_lock);
4193
	memcg->socket_pressure = jiffies;
4194
#ifndef CONFIG_SLOB
V
Vladimir Davydov 已提交
4195 4196
	memcg->kmemcg_id = -1;
#endif
4197 4198 4199
#ifdef CONFIG_CGROUP_WRITEBACK
	INIT_LIST_HEAD(&memcg->cgwb_list);
#endif
4200
	idr_replace(&mem_cgroup_idr, memcg, memcg->id.id);
4201 4202
	return memcg;
fail:
4203 4204
	if (memcg->id.id > 0)
		idr_remove(&mem_cgroup_idr, memcg->id.id);
4205
	__mem_cgroup_free(memcg);
4206
	return NULL;
4207 4208
}

4209 4210
static struct cgroup_subsys_state * __ref
mem_cgroup_css_alloc(struct cgroup_subsys_state *parent_css)
4211
{
4212 4213 4214
	struct mem_cgroup *parent = mem_cgroup_from_css(parent_css);
	struct mem_cgroup *memcg;
	long error = -ENOMEM;
4215

4216 4217 4218
	memcg = mem_cgroup_alloc();
	if (!memcg)
		return ERR_PTR(error);
4219

4220 4221 4222 4223 4224 4225 4226 4227
	memcg->high = PAGE_COUNTER_MAX;
	memcg->soft_limit = PAGE_COUNTER_MAX;
	if (parent) {
		memcg->swappiness = mem_cgroup_swappiness(parent);
		memcg->oom_kill_disable = parent->oom_kill_disable;
	}
	if (parent && parent->use_hierarchy) {
		memcg->use_hierarchy = true;
4228
		page_counter_init(&memcg->memory, &parent->memory);
4229
		page_counter_init(&memcg->swap, &parent->swap);
4230 4231
		page_counter_init(&memcg->memsw, &parent->memsw);
		page_counter_init(&memcg->kmem, &parent->kmem);
4232
		page_counter_init(&memcg->tcpmem, &parent->tcpmem);
4233
	} else {
4234
		page_counter_init(&memcg->memory, NULL);
4235
		page_counter_init(&memcg->swap, NULL);
4236 4237
		page_counter_init(&memcg->memsw, NULL);
		page_counter_init(&memcg->kmem, NULL);
4238
		page_counter_init(&memcg->tcpmem, NULL);
4239 4240 4241 4242 4243
		/*
		 * Deeper hierachy with use_hierarchy == false doesn't make
		 * much sense so let cgroup subsystem know about this
		 * unfortunate state in our controller.
		 */
4244
		if (parent != root_mem_cgroup)
4245
			memory_cgrp_subsys.broken_hierarchy = true;
4246
	}
4247

4248 4249 4250 4251 4252 4253
	/* The following stuff does not apply to the root */
	if (!parent) {
		root_mem_cgroup = memcg;
		return &memcg->css;
	}

4254
	error = memcg_online_kmem(memcg);
4255 4256
	if (error)
		goto fail;
4257

4258
	if (cgroup_subsys_on_dfl(memory_cgrp_subsys) && !cgroup_memory_nosocket)
4259
		static_branch_inc(&memcg_sockets_enabled_key);
4260

4261 4262 4263
	return &memcg->css;
fail:
	mem_cgroup_free(memcg);
4264
	return ERR_PTR(-ENOMEM);
4265 4266
}

4267
static int mem_cgroup_css_online(struct cgroup_subsys_state *css)
4268
{
4269 4270
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);

4271
	/* Online state pins memcg ID, memcg ID pins CSS */
4272
	atomic_set(&memcg->id.ref, 1);
4273
	css_get(css);
4274
	return 0;
B
Balbir Singh 已提交
4275 4276
}

4277
static void mem_cgroup_css_offline(struct cgroup_subsys_state *css)
4278
{
4279
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
4280
	struct mem_cgroup_event *event, *tmp;
4281 4282 4283 4284 4285 4286

	/*
	 * Unregister events and notify userspace.
	 * Notify userspace about cgroup removing only after rmdir of cgroup
	 * directory to avoid race between userspace and kernelspace.
	 */
4287 4288
	spin_lock(&memcg->event_list_lock);
	list_for_each_entry_safe(event, tmp, &memcg->event_list, list) {
4289 4290 4291
		list_del_init(&event->list);
		schedule_work(&event->remove);
	}
4292
	spin_unlock(&memcg->event_list_lock);
4293

4294
	memcg_offline_kmem(memcg);
4295
	wb_memcg_offline(memcg);
4296 4297

	mem_cgroup_id_put(memcg);
4298 4299
}

4300 4301 4302 4303 4304 4305 4306
static void mem_cgroup_css_released(struct cgroup_subsys_state *css)
{
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);

	invalidate_reclaim_iterators(memcg);
}

4307
static void mem_cgroup_css_free(struct cgroup_subsys_state *css)
B
Balbir Singh 已提交
4308
{
4309
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
4310

4311
	if (cgroup_subsys_on_dfl(memory_cgrp_subsys) && !cgroup_memory_nosocket)
4312
		static_branch_dec(&memcg_sockets_enabled_key);
4313

4314
	if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_active)
V
Vladimir Davydov 已提交
4315
		static_branch_dec(&memcg_sockets_enabled_key);
4316

4317 4318 4319
	vmpressure_cleanup(&memcg->vmpressure);
	cancel_work_sync(&memcg->high_work);
	mem_cgroup_remove_from_trees(memcg);
4320
	memcg_free_kmem(memcg);
4321
	mem_cgroup_free(memcg);
B
Balbir Singh 已提交
4322 4323
}

4324 4325 4326 4327 4328 4329 4330 4331 4332 4333 4334 4335 4336 4337 4338 4339 4340
/**
 * 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);

4341 4342 4343 4344 4345
	page_counter_limit(&memcg->memory, PAGE_COUNTER_MAX);
	page_counter_limit(&memcg->swap, PAGE_COUNTER_MAX);
	page_counter_limit(&memcg->memsw, PAGE_COUNTER_MAX);
	page_counter_limit(&memcg->kmem, PAGE_COUNTER_MAX);
	page_counter_limit(&memcg->tcpmem, PAGE_COUNTER_MAX);
4346 4347
	memcg->low = 0;
	memcg->high = PAGE_COUNTER_MAX;
4348
	memcg->soft_limit = PAGE_COUNTER_MAX;
4349
	memcg_wb_domain_size_changed(memcg);
4350 4351
}

4352
#ifdef CONFIG_MMU
4353
/* Handlers for move charge at task migration. */
4354
static int mem_cgroup_do_precharge(unsigned long count)
4355
{
4356
	int ret;
4357

4358 4359
	/* Try a single bulk charge without reclaim first, kswapd may wake */
	ret = try_charge(mc.to, GFP_KERNEL & ~__GFP_DIRECT_RECLAIM, count);
4360
	if (!ret) {
4361 4362 4363
		mc.precharge += count;
		return ret;
	}
4364

4365
	/* Try charges one by one with reclaim, but do not retry */
4366
	while (count--) {
4367
		ret = try_charge(mc.to, GFP_KERNEL | __GFP_NORETRY, 1);
4368 4369
		if (ret)
			return ret;
4370
		mc.precharge++;
4371
		cond_resched();
4372
	}
4373
	return 0;
4374 4375 4376 4377
}

union mc_target {
	struct page	*page;
4378
	swp_entry_t	ent;
4379 4380 4381
};

enum mc_target_type {
4382
	MC_TARGET_NONE = 0,
4383
	MC_TARGET_PAGE,
4384
	MC_TARGET_SWAP,
4385 4386
};

D
Daisuke Nishimura 已提交
4387 4388
static struct page *mc_handle_present_pte(struct vm_area_struct *vma,
						unsigned long addr, pte_t ptent)
4389
{
D
Daisuke Nishimura 已提交
4390
	struct page *page = vm_normal_page(vma, addr, ptent);
4391

D
Daisuke Nishimura 已提交
4392 4393 4394
	if (!page || !page_mapped(page))
		return NULL;
	if (PageAnon(page)) {
4395
		if (!(mc.flags & MOVE_ANON))
D
Daisuke Nishimura 已提交
4396
			return NULL;
4397 4398 4399 4400
	} else {
		if (!(mc.flags & MOVE_FILE))
			return NULL;
	}
D
Daisuke Nishimura 已提交
4401 4402 4403 4404 4405 4406
	if (!get_page_unless_zero(page))
		return NULL;

	return page;
}

4407
#ifdef CONFIG_SWAP
D
Daisuke Nishimura 已提交
4408
static struct page *mc_handle_swap_pte(struct vm_area_struct *vma,
4409
			pte_t ptent, swp_entry_t *entry)
D
Daisuke Nishimura 已提交
4410 4411 4412 4413
{
	struct page *page = NULL;
	swp_entry_t ent = pte_to_swp_entry(ptent);

4414
	if (!(mc.flags & MOVE_ANON) || non_swap_entry(ent))
D
Daisuke Nishimura 已提交
4415
		return NULL;
4416 4417 4418 4419
	/*
	 * Because lookup_swap_cache() updates some statistics counter,
	 * we call find_get_page() with swapper_space directly.
	 */
4420
	page = find_get_page(swap_address_space(ent), swp_offset(ent));
4421
	if (do_memsw_account())
D
Daisuke Nishimura 已提交
4422 4423 4424 4425
		entry->val = ent.val;

	return page;
}
4426 4427
#else
static struct page *mc_handle_swap_pte(struct vm_area_struct *vma,
4428
			pte_t ptent, swp_entry_t *entry)
4429 4430 4431 4432
{
	return NULL;
}
#endif
D
Daisuke Nishimura 已提交
4433

4434 4435 4436 4437 4438 4439 4440 4441 4442
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;
4443
	if (!(mc.flags & MOVE_FILE))
4444 4445 4446
		return NULL;

	mapping = vma->vm_file->f_mapping;
4447
	pgoff = linear_page_index(vma, addr);
4448 4449

	/* page is moved even if it's not RSS of this task(page-faulted). */
4450 4451
#ifdef CONFIG_SWAP
	/* shmem/tmpfs may report page out on swap: account for that too. */
4452 4453 4454 4455
	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);
4456
			if (do_memsw_account())
4457
				*entry = swp;
4458 4459
			page = find_get_page(swap_address_space(swp),
					     swp_offset(swp));
4460 4461 4462 4463 4464
		}
	} else
		page = find_get_page(mapping, pgoff);
#else
	page = find_get_page(mapping, pgoff);
4465
#endif
4466 4467 4468
	return page;
}

4469 4470 4471
/**
 * mem_cgroup_move_account - move account of the page
 * @page: the page
4472
 * @compound: charge the page as compound or small page
4473 4474 4475
 * @from: mem_cgroup which the page is moved from.
 * @to:	mem_cgroup which the page is moved to. @from != @to.
 *
4476
 * The caller must make sure the page is not on LRU (isolate_page() is useful.)
4477 4478 4479 4480 4481
 *
 * 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,
4482
				   bool compound,
4483 4484 4485 4486
				   struct mem_cgroup *from,
				   struct mem_cgroup *to)
{
	unsigned long flags;
4487
	unsigned int nr_pages = compound ? hpage_nr_pages(page) : 1;
4488
	int ret;
4489
	bool anon;
4490 4491 4492

	VM_BUG_ON(from == to);
	VM_BUG_ON_PAGE(PageLRU(page), page);
4493
	VM_BUG_ON(compound && !PageTransHuge(page));
4494 4495

	/*
4496
	 * Prevent mem_cgroup_migrate() from looking at
4497
	 * page->mem_cgroup of its source page while we change it.
4498
	 */
4499
	ret = -EBUSY;
4500 4501 4502 4503 4504 4505 4506
	if (!trylock_page(page))
		goto out;

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

4507 4508
	anon = PageAnon(page);

4509 4510
	spin_lock_irqsave(&from->move_lock, flags);

4511
	if (!anon && page_mapped(page)) {
4512 4513
		__this_cpu_sub(from->stat->count[NR_FILE_MAPPED], nr_pages);
		__this_cpu_add(to->stat->count[NR_FILE_MAPPED], nr_pages);
4514 4515
	}

4516 4517
	/*
	 * move_lock grabbed above and caller set from->moving_account, so
4518
	 * mod_memcg_page_state will serialize updates to PageDirty.
4519 4520 4521 4522 4523 4524
	 * 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)) {
4525
			__this_cpu_sub(from->stat->count[NR_FILE_DIRTY],
4526
				       nr_pages);
4527
			__this_cpu_add(to->stat->count[NR_FILE_DIRTY],
4528 4529 4530 4531
				       nr_pages);
		}
	}

4532
	if (PageWriteback(page)) {
4533 4534
		__this_cpu_sub(from->stat->count[NR_WRITEBACK], nr_pages);
		__this_cpu_add(to->stat->count[NR_WRITEBACK], nr_pages);
4535 4536 4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547 4548 4549
	}

	/*
	 * 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();
4550
	mem_cgroup_charge_statistics(to, page, compound, nr_pages);
4551
	memcg_check_events(to, page);
4552
	mem_cgroup_charge_statistics(from, page, compound, -nr_pages);
4553 4554 4555 4556 4557 4558 4559 4560
	memcg_check_events(from, page);
	local_irq_enable();
out_unlock:
	unlock_page(page);
out:
	return ret;
}

4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579
/**
 * get_mctgt_type - get target type of moving charge
 * @vma: the vma the pte to be checked belongs
 * @addr: the address corresponding to the pte to be checked
 * @ptent: the pte to be checked
 * @target: the pointer the target page or swap ent will be stored(can be NULL)
 *
 * 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).
 *   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.
 *
 * Called with pte lock held.
 */

4580
static enum mc_target_type get_mctgt_type(struct vm_area_struct *vma,
D
Daisuke Nishimura 已提交
4581 4582 4583
		unsigned long addr, pte_t ptent, union mc_target *target)
{
	struct page *page = NULL;
4584
	enum mc_target_type ret = MC_TARGET_NONE;
D
Daisuke Nishimura 已提交
4585 4586 4587 4588 4589
	swp_entry_t ent = { .val = 0 };

	if (pte_present(ptent))
		page = mc_handle_present_pte(vma, addr, ptent);
	else if (is_swap_pte(ptent))
4590
		page = mc_handle_swap_pte(vma, ptent, &ent);
4591
	else if (pte_none(ptent))
4592
		page = mc_handle_file_pte(vma, addr, ptent, &ent);
D
Daisuke Nishimura 已提交
4593 4594

	if (!page && !ent.val)
4595
		return ret;
4596 4597
	if (page) {
		/*
4598
		 * Do only loose check w/o serialization.
4599
		 * mem_cgroup_move_account() checks the page is valid or
4600
		 * not under LRU exclusion.
4601
		 */
4602
		if (page->mem_cgroup == mc.from) {
4603 4604 4605 4606 4607 4608 4609
			ret = MC_TARGET_PAGE;
			if (target)
				target->page = page;
		}
		if (!ret || !target)
			put_page(page);
	}
D
Daisuke Nishimura 已提交
4610 4611
	/* 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
4612
	    mem_cgroup_id(mc.from) == lookup_swap_cgroup_id(ent)) {
4613 4614 4615
		ret = MC_TARGET_SWAP;
		if (target)
			target->ent = ent;
4616 4617 4618 4619
	}
	return ret;
}

4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632
#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);
4633
	VM_BUG_ON_PAGE(!page || !PageHead(page), page);
4634
	if (!(mc.flags & MOVE_ANON))
4635
		return ret;
4636
	if (page->mem_cgroup == mc.from) {
4637 4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650 4651 4652
		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

4653 4654 4655 4656
static int mem_cgroup_count_precharge_pte_range(pmd_t *pmd,
					unsigned long addr, unsigned long end,
					struct mm_walk *walk)
{
4657
	struct vm_area_struct *vma = walk->vma;
4658 4659 4660
	pte_t *pte;
	spinlock_t *ptl;

4661 4662
	ptl = pmd_trans_huge_lock(pmd, vma);
	if (ptl) {
4663 4664
		if (get_mctgt_type_thp(vma, addr, *pmd, NULL) == MC_TARGET_PAGE)
			mc.precharge += HPAGE_PMD_NR;
4665
		spin_unlock(ptl);
4666
		return 0;
4667
	}
4668

4669 4670
	if (pmd_trans_unstable(pmd))
		return 0;
4671 4672
	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
	for (; addr != end; pte++, addr += PAGE_SIZE)
4673
		if (get_mctgt_type(vma, addr, *pte, NULL))
4674 4675 4676 4677
			mc.precharge++;	/* increment precharge temporarily */
	pte_unmap_unlock(pte - 1, ptl);
	cond_resched();

4678 4679 4680
	return 0;
}

4681 4682 4683 4684
static unsigned long mem_cgroup_count_precharge(struct mm_struct *mm)
{
	unsigned long precharge;

4685 4686 4687 4688
	struct mm_walk mem_cgroup_count_precharge_walk = {
		.pmd_entry = mem_cgroup_count_precharge_pte_range,
		.mm = mm,
	};
4689
	down_read(&mm->mmap_sem);
4690 4691
	walk_page_range(0, mm->highest_vm_end,
			&mem_cgroup_count_precharge_walk);
4692
	up_read(&mm->mmap_sem);
4693 4694 4695 4696 4697 4698 4699 4700 4701

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

	return precharge;
}

static int mem_cgroup_precharge_mc(struct mm_struct *mm)
{
4702 4703 4704 4705 4706
	unsigned long precharge = mem_cgroup_count_precharge(mm);

	VM_BUG_ON(mc.moving_task);
	mc.moving_task = current;
	return mem_cgroup_do_precharge(precharge);
4707 4708
}

4709 4710
/* cancels all extra charges on mc.from and mc.to, and wakes up all waiters. */
static void __mem_cgroup_clear_mc(void)
4711
{
4712 4713 4714
	struct mem_cgroup *from = mc.from;
	struct mem_cgroup *to = mc.to;

4715
	/* we must uncharge all the leftover precharges from mc.to */
4716
	if (mc.precharge) {
4717
		cancel_charge(mc.to, mc.precharge);
4718 4719 4720 4721 4722 4723 4724
		mc.precharge = 0;
	}
	/*
	 * we didn't uncharge from mc.from at mem_cgroup_move_account(), so
	 * we must uncharge here.
	 */
	if (mc.moved_charge) {
4725
		cancel_charge(mc.from, mc.moved_charge);
4726
		mc.moved_charge = 0;
4727
	}
4728 4729 4730
	/* we must fixup refcnts and charges */
	if (mc.moved_swap) {
		/* uncharge swap account from the old cgroup */
4731
		if (!mem_cgroup_is_root(mc.from))
4732
			page_counter_uncharge(&mc.from->memsw, mc.moved_swap);
4733

4734 4735
		mem_cgroup_id_put_many(mc.from, mc.moved_swap);

4736
		/*
4737 4738
		 * we charged both to->memory and to->memsw, so we
		 * should uncharge to->memory.
4739
		 */
4740
		if (!mem_cgroup_is_root(mc.to))
4741 4742
			page_counter_uncharge(&mc.to->memory, mc.moved_swap);

4743 4744
		mem_cgroup_id_get_many(mc.to, mc.moved_swap);
		css_put_many(&mc.to->css, mc.moved_swap);
4745

4746 4747
		mc.moved_swap = 0;
	}
4748 4749 4750 4751 4752 4753 4754
	memcg_oom_recover(from);
	memcg_oom_recover(to);
	wake_up_all(&mc.waitq);
}

static void mem_cgroup_clear_mc(void)
{
4755 4756
	struct mm_struct *mm = mc.mm;

4757 4758 4759 4760 4761 4762
	/*
	 * we must clear moving_task before waking up waiters at the end of
	 * task migration.
	 */
	mc.moving_task = NULL;
	__mem_cgroup_clear_mc();
4763
	spin_lock(&mc.lock);
4764 4765
	mc.from = NULL;
	mc.to = NULL;
4766
	mc.mm = NULL;
4767
	spin_unlock(&mc.lock);
4768 4769

	mmput(mm);
4770 4771
}

4772
static int mem_cgroup_can_attach(struct cgroup_taskset *tset)
4773
{
4774
	struct cgroup_subsys_state *css;
4775
	struct mem_cgroup *memcg = NULL; /* unneeded init to make gcc happy */
4776
	struct mem_cgroup *from;
4777
	struct task_struct *leader, *p;
4778
	struct mm_struct *mm;
4779
	unsigned long move_flags;
4780
	int ret = 0;
4781

4782 4783
	/* charge immigration isn't supported on the default hierarchy */
	if (cgroup_subsys_on_dfl(memory_cgrp_subsys))
4784 4785
		return 0;

4786 4787 4788 4789 4790 4791 4792
	/*
	 * 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;
4793
	cgroup_taskset_for_each_leader(leader, css, tset) {
4794 4795
		WARN_ON_ONCE(p);
		p = leader;
4796
		memcg = mem_cgroup_from_css(css);
4797 4798 4799 4800
	}
	if (!p)
		return 0;

4801 4802 4803 4804 4805 4806 4807 4808 4809
	/*
	 * 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;

4810 4811 4812 4813 4814 4815 4816 4817 4818 4819 4820 4821 4822 4823 4824 4825
	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);
4826
		mc.mm = mm;
4827 4828 4829 4830 4831 4832 4833 4834 4835
		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();
4836 4837
	} else {
		mmput(mm);
4838 4839 4840 4841
	}
	return ret;
}

4842
static void mem_cgroup_cancel_attach(struct cgroup_taskset *tset)
4843
{
4844 4845
	if (mc.to)
		mem_cgroup_clear_mc();
4846 4847
}

4848 4849 4850
static int mem_cgroup_move_charge_pte_range(pmd_t *pmd,
				unsigned long addr, unsigned long end,
				struct mm_walk *walk)
4851
{
4852
	int ret = 0;
4853
	struct vm_area_struct *vma = walk->vma;
4854 4855
	pte_t *pte;
	spinlock_t *ptl;
4856 4857 4858
	enum mc_target_type target_type;
	union mc_target target;
	struct page *page;
4859

4860 4861
	ptl = pmd_trans_huge_lock(pmd, vma);
	if (ptl) {
4862
		if (mc.precharge < HPAGE_PMD_NR) {
4863
			spin_unlock(ptl);
4864 4865 4866 4867 4868 4869
			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)) {
4870
				if (!mem_cgroup_move_account(page, true,
4871
							     mc.from, mc.to)) {
4872 4873 4874 4875 4876 4877 4878
					mc.precharge -= HPAGE_PMD_NR;
					mc.moved_charge += HPAGE_PMD_NR;
				}
				putback_lru_page(page);
			}
			put_page(page);
		}
4879
		spin_unlock(ptl);
4880
		return 0;
4881 4882
	}

4883 4884
	if (pmd_trans_unstable(pmd))
		return 0;
4885 4886 4887 4888
retry:
	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
	for (; addr != end; addr += PAGE_SIZE) {
		pte_t ptent = *(pte++);
4889
		swp_entry_t ent;
4890 4891 4892 4893

		if (!mc.precharge)
			break;

4894
		switch (get_mctgt_type(vma, addr, ptent, &target)) {
4895 4896
		case MC_TARGET_PAGE:
			page = target.page;
4897 4898 4899 4900 4901 4902 4903 4904
			/*
			 * 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;
4905 4906
			if (isolate_lru_page(page))
				goto put;
4907 4908
			if (!mem_cgroup_move_account(page, false,
						mc.from, mc.to)) {
4909
				mc.precharge--;
4910 4911
				/* we uncharge from mc.from later. */
				mc.moved_charge++;
4912 4913
			}
			putback_lru_page(page);
4914
put:			/* get_mctgt_type() gets the page */
4915 4916
			put_page(page);
			break;
4917 4918
		case MC_TARGET_SWAP:
			ent = target.ent;
4919
			if (!mem_cgroup_move_swap_account(ent, mc.from, mc.to)) {
4920
				mc.precharge--;
4921 4922 4923
				/* we fixup refcnts and charges later. */
				mc.moved_swap++;
			}
4924
			break;
4925 4926 4927 4928 4929 4930 4931 4932 4933 4934 4935 4936 4937 4938
		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.
		 */
4939
		ret = mem_cgroup_do_precharge(1);
4940 4941 4942 4943 4944 4945 4946
		if (!ret)
			goto retry;
	}

	return ret;
}

4947
static void mem_cgroup_move_charge(void)
4948
{
4949 4950
	struct mm_walk mem_cgroup_move_charge_walk = {
		.pmd_entry = mem_cgroup_move_charge_pte_range,
4951
		.mm = mc.mm,
4952
	};
4953 4954

	lru_add_drain_all();
4955
	/*
4956 4957 4958
	 * Signal lock_page_memcg() 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.
4959 4960 4961
	 */
	atomic_inc(&mc.from->moving_account);
	synchronize_rcu();
4962
retry:
4963
	if (unlikely(!down_read_trylock(&mc.mm->mmap_sem))) {
4964 4965 4966 4967 4968 4969 4970 4971 4972 4973 4974
		/*
		 * 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;
	}
4975 4976 4977 4978
	/*
	 * When we have consumed all precharges and failed in doing
	 * additional charge, the page walk just aborts.
	 */
4979 4980
	walk_page_range(0, mc.mm->highest_vm_end, &mem_cgroup_move_charge_walk);

4981
	up_read(&mc.mm->mmap_sem);
4982
	atomic_dec(&mc.from->moving_account);
4983 4984
}

4985
static void mem_cgroup_move_task(void)
B
Balbir Singh 已提交
4986
{
4987 4988
	if (mc.to) {
		mem_cgroup_move_charge();
4989
		mem_cgroup_clear_mc();
4990
	}
B
Balbir Singh 已提交
4991
}
4992
#else	/* !CONFIG_MMU */
4993
static int mem_cgroup_can_attach(struct cgroup_taskset *tset)
4994 4995 4996
{
	return 0;
}
4997
static void mem_cgroup_cancel_attach(struct cgroup_taskset *tset)
4998 4999
{
}
5000
static void mem_cgroup_move_task(void)
5001 5002 5003
{
}
#endif
B
Balbir Singh 已提交
5004

5005 5006
/*
 * Cgroup retains root cgroups across [un]mount cycles making it necessary
5007 5008
 * to verify whether we're attached to the default hierarchy on each mount
 * attempt.
5009
 */
5010
static void mem_cgroup_bind(struct cgroup_subsys_state *root_css)
5011 5012
{
	/*
5013
	 * use_hierarchy is forced on the default hierarchy.  cgroup core
5014 5015 5016
	 * guarantees that @root doesn't have any children, so turning it
	 * on for the root memcg is enough.
	 */
5017
	if (cgroup_subsys_on_dfl(memory_cgrp_subsys))
5018 5019 5020
		root_mem_cgroup->use_hierarchy = true;
	else
		root_mem_cgroup->use_hierarchy = false;
5021 5022
}

5023 5024 5025
static u64 memory_current_read(struct cgroup_subsys_state *css,
			       struct cftype *cft)
{
5026 5027 5028
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);

	return (u64)page_counter_read(&memcg->memory) * PAGE_SIZE;
5029 5030 5031 5032 5033
}

static int memory_low_show(struct seq_file *m, void *v)
{
	struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(m));
5034
	unsigned long low = READ_ONCE(memcg->low);
5035 5036

	if (low == PAGE_COUNTER_MAX)
5037
		seq_puts(m, "max\n");
5038 5039 5040 5041 5042 5043 5044 5045 5046 5047 5048 5049 5050 5051
	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);
5052
	err = page_counter_memparse(buf, "max", &low);
5053 5054 5055 5056 5057 5058 5059 5060 5061 5062 5063
	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));
5064
	unsigned long high = READ_ONCE(memcg->high);
5065 5066

	if (high == PAGE_COUNTER_MAX)
5067
		seq_puts(m, "max\n");
5068 5069 5070 5071 5072 5073 5074 5075 5076 5077
	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));
5078
	unsigned long nr_pages;
5079 5080 5081 5082
	unsigned long high;
	int err;

	buf = strstrip(buf);
5083
	err = page_counter_memparse(buf, "max", &high);
5084 5085 5086 5087 5088
	if (err)
		return err;

	memcg->high = high;

5089 5090 5091 5092 5093
	nr_pages = page_counter_read(&memcg->memory);
	if (nr_pages > high)
		try_to_free_mem_cgroup_pages(memcg, nr_pages - high,
					     GFP_KERNEL, true);

5094
	memcg_wb_domain_size_changed(memcg);
5095 5096 5097 5098 5099 5100
	return nbytes;
}

static int memory_max_show(struct seq_file *m, void *v)
{
	struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(m));
5101
	unsigned long max = READ_ONCE(memcg->memory.limit);
5102 5103

	if (max == PAGE_COUNTER_MAX)
5104
		seq_puts(m, "max\n");
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));
5115 5116
	unsigned int nr_reclaims = MEM_CGROUP_RECLAIM_RETRIES;
	bool drained = false;
5117 5118 5119 5120
	unsigned long max;
	int err;

	buf = strstrip(buf);
5121
	err = page_counter_memparse(buf, "max", &max);
5122 5123 5124
	if (err)
		return err;

5125 5126 5127 5128 5129 5130 5131 5132 5133 5134 5135 5136 5137 5138 5139 5140 5141 5142 5143 5144 5145 5146 5147 5148 5149 5150
	xchg(&memcg->memory.limit, max);

	for (;;) {
		unsigned long nr_pages = page_counter_read(&memcg->memory);

		if (nr_pages <= max)
			break;

		if (signal_pending(current)) {
			err = -EINTR;
			break;
		}

		if (!drained) {
			drain_all_stock(memcg);
			drained = true;
			continue;
		}

		if (nr_reclaims) {
			if (!try_to_free_mem_cgroup_pages(memcg, nr_pages - max,
							  GFP_KERNEL, true))
				nr_reclaims--;
			continue;
		}

5151
		mem_cgroup_event(memcg, MEMCG_OOM);
5152 5153 5154
		if (!mem_cgroup_out_of_memory(memcg, GFP_KERNEL, 0))
			break;
	}
5155

5156
	memcg_wb_domain_size_changed(memcg);
5157 5158 5159 5160 5161 5162 5163
	return nbytes;
}

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

5164 5165 5166 5167
	seq_printf(m, "low %lu\n", memcg_sum_events(memcg, MEMCG_LOW));
	seq_printf(m, "high %lu\n", memcg_sum_events(memcg, MEMCG_HIGH));
	seq_printf(m, "max %lu\n", memcg_sum_events(memcg, MEMCG_MAX));
	seq_printf(m, "oom %lu\n", memcg_sum_events(memcg, MEMCG_OOM));
5168
	seq_printf(m, "oom_kill %lu\n", memcg_sum_events(memcg, OOM_KILL));
5169 5170 5171 5172

	return 0;
}

5173 5174 5175
static int memory_stat_show(struct seq_file *m, void *v)
{
	struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(m));
5176 5177
	unsigned long stat[MEMCG_NR_STAT];
	unsigned long events[MEMCG_NR_EVENTS];
5178 5179 5180 5181 5182 5183 5184 5185 5186 5187 5188 5189 5190
	int i;

	/*
	 * Provide statistics on the state of the memory subsystem as
	 * well as cumulative event counters that show past behavior.
	 *
	 * This list is ordered following a combination of these gradients:
	 * 1) generic big picture -> specifics and details
	 * 2) reflecting userspace activity -> reflecting kernel heuristics
	 *
	 * Current memory state:
	 */

5191 5192 5193
	tree_stat(memcg, stat);
	tree_events(memcg, events);

5194
	seq_printf(m, "anon %llu\n",
5195
		   (u64)stat[MEMCG_RSS] * PAGE_SIZE);
5196
	seq_printf(m, "file %llu\n",
5197
		   (u64)stat[MEMCG_CACHE] * PAGE_SIZE);
5198
	seq_printf(m, "kernel_stack %llu\n",
5199
		   (u64)stat[MEMCG_KERNEL_STACK_KB] * 1024);
5200
	seq_printf(m, "slab %llu\n",
5201 5202
		   (u64)(stat[NR_SLAB_RECLAIMABLE] +
			 stat[NR_SLAB_UNRECLAIMABLE]) * PAGE_SIZE);
5203
	seq_printf(m, "sock %llu\n",
5204
		   (u64)stat[MEMCG_SOCK] * PAGE_SIZE);
5205

5206
	seq_printf(m, "shmem %llu\n",
5207
		   (u64)stat[NR_SHMEM] * PAGE_SIZE);
5208
	seq_printf(m, "file_mapped %llu\n",
5209
		   (u64)stat[NR_FILE_MAPPED] * PAGE_SIZE);
5210
	seq_printf(m, "file_dirty %llu\n",
5211
		   (u64)stat[NR_FILE_DIRTY] * PAGE_SIZE);
5212
	seq_printf(m, "file_writeback %llu\n",
5213
		   (u64)stat[NR_WRITEBACK] * PAGE_SIZE);
5214 5215 5216 5217 5218 5219 5220 5221 5222 5223 5224

	for (i = 0; i < NR_LRU_LISTS; i++) {
		struct mem_cgroup *mi;
		unsigned long val = 0;

		for_each_mem_cgroup_tree(mi, memcg)
			val += mem_cgroup_nr_lru_pages(mi, BIT(i));
		seq_printf(m, "%s %llu\n",
			   mem_cgroup_lru_names[i], (u64)val * PAGE_SIZE);
	}

5225
	seq_printf(m, "slab_reclaimable %llu\n",
5226
		   (u64)stat[NR_SLAB_RECLAIMABLE] * PAGE_SIZE);
5227
	seq_printf(m, "slab_unreclaimable %llu\n",
5228
		   (u64)stat[NR_SLAB_UNRECLAIMABLE] * PAGE_SIZE);
5229

5230 5231
	/* Accumulated memory events */

5232 5233
	seq_printf(m, "pgfault %lu\n", events[PGFAULT]);
	seq_printf(m, "pgmajfault %lu\n", events[PGMAJFAULT]);
5234

5235 5236 5237 5238 5239 5240 5241 5242 5243 5244
	seq_printf(m, "pgrefill %lu\n", events[PGREFILL]);
	seq_printf(m, "pgscan %lu\n", events[PGSCAN_KSWAPD] +
		   events[PGSCAN_DIRECT]);
	seq_printf(m, "pgsteal %lu\n", events[PGSTEAL_KSWAPD] +
		   events[PGSTEAL_DIRECT]);
	seq_printf(m, "pgactivate %lu\n", events[PGACTIVATE]);
	seq_printf(m, "pgdeactivate %lu\n", events[PGDEACTIVATE]);
	seq_printf(m, "pglazyfree %lu\n", events[PGLAZYFREE]);
	seq_printf(m, "pglazyfreed %lu\n", events[PGLAZYFREED]);

5245
	seq_printf(m, "workingset_refault %lu\n",
5246
		   stat[WORKINGSET_REFAULT]);
5247
	seq_printf(m, "workingset_activate %lu\n",
5248
		   stat[WORKINGSET_ACTIVATE]);
5249
	seq_printf(m, "workingset_nodereclaim %lu\n",
5250
		   stat[WORKINGSET_NODERECLAIM]);
5251

5252 5253 5254
	return 0;
}

5255 5256 5257
static struct cftype memory_files[] = {
	{
		.name = "current",
5258
		.flags = CFTYPE_NOT_ON_ROOT,
5259 5260 5261 5262 5263 5264 5265 5266 5267 5268 5269 5270 5271 5272 5273 5274 5275 5276 5277 5278 5279 5280 5281
		.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,
5282
		.file_offset = offsetof(struct mem_cgroup, events_file),
5283 5284
		.seq_show = memory_events_show,
	},
5285 5286 5287 5288 5289
	{
		.name = "stat",
		.flags = CFTYPE_NOT_ON_ROOT,
		.seq_show = memory_stat_show,
	},
5290 5291 5292
	{ }	/* terminate */
};

5293
struct cgroup_subsys memory_cgrp_subsys = {
5294
	.css_alloc = mem_cgroup_css_alloc,
5295
	.css_online = mem_cgroup_css_online,
5296
	.css_offline = mem_cgroup_css_offline,
5297
	.css_released = mem_cgroup_css_released,
5298
	.css_free = mem_cgroup_css_free,
5299
	.css_reset = mem_cgroup_css_reset,
5300 5301
	.can_attach = mem_cgroup_can_attach,
	.cancel_attach = mem_cgroup_cancel_attach,
5302
	.post_attach = mem_cgroup_move_task,
5303
	.bind = mem_cgroup_bind,
5304 5305
	.dfl_cftypes = memory_files,
	.legacy_cftypes = mem_cgroup_legacy_files,
5306
	.early_init = 0,
B
Balbir Singh 已提交
5307
};
5308

5309 5310 5311 5312 5313 5314 5315 5316 5317 5318 5319 5320 5321 5322 5323 5324 5325 5326 5327 5328 5329 5330
/**
 * 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
5331
	if (page_counter_read(&memcg->memory) >= memcg->low)
5332 5333 5334 5335 5336 5337 5338 5339
		return false;

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

		if (memcg == root_mem_cgroup)
			break;

M
Michal Hocko 已提交
5340
		if (page_counter_read(&memcg->memory) >= memcg->low)
5341 5342 5343 5344 5345
			return false;
	}
	return true;
}

5346 5347 5348 5349 5350 5351
/**
 * 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
5352
 * @compound: charge the page as compound or small page
5353 5354 5355 5356 5357 5358 5359 5360 5361 5362 5363 5364
 *
 * 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,
5365 5366
			  gfp_t gfp_mask, struct mem_cgroup **memcgp,
			  bool compound)
5367 5368
{
	struct mem_cgroup *memcg = NULL;
5369
	unsigned int nr_pages = compound ? hpage_nr_pages(page) : 1;
5370 5371 5372 5373 5374 5375 5376 5377 5378 5379 5380 5381 5382
	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.
		 */
5383
		VM_BUG_ON_PAGE(!PageLocked(page), page);
5384
		if (page->mem_cgroup)
5385
			goto out;
5386

5387
		if (do_swap_account) {
5388 5389 5390 5391 5392 5393 5394 5395 5396
			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();
		}
5397 5398 5399 5400 5401 5402 5403 5404 5405 5406 5407 5408 5409 5410 5411 5412 5413 5414
	}

	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
5415
 * @compound: charge the page as compound or small page
5416 5417 5418 5419 5420 5421 5422 5423 5424 5425 5426 5427
 *
 * 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,
5428
			      bool lrucare, bool compound)
5429
{
5430
	unsigned int nr_pages = compound ? hpage_nr_pages(page) : 1;
5431 5432 5433 5434 5435 5436 5437 5438 5439 5440 5441 5442 5443 5444

	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;

5445 5446 5447
	commit_charge(page, memcg, lrucare);

	local_irq_disable();
5448
	mem_cgroup_charge_statistics(memcg, page, compound, nr_pages);
5449 5450
	memcg_check_events(memcg, page);
	local_irq_enable();
5451

5452
	if (do_memsw_account() && PageSwapCache(page)) {
5453 5454 5455 5456 5457 5458
		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.
		 */
5459
		mem_cgroup_uncharge_swap(entry, nr_pages);
5460 5461 5462 5463 5464 5465 5466
	}
}

/**
 * mem_cgroup_cancel_charge - cancel a page charge
 * @page: page to charge
 * @memcg: memcg to charge the page to
5467
 * @compound: charge the page as compound or small page
5468 5469 5470
 *
 * Cancel a charge transaction started by mem_cgroup_try_charge().
 */
5471 5472
void mem_cgroup_cancel_charge(struct page *page, struct mem_cgroup *memcg,
		bool compound)
5473
{
5474
	unsigned int nr_pages = compound ? hpage_nr_pages(page) : 1;
5475 5476 5477 5478 5479 5480 5481 5482 5483 5484 5485 5486 5487 5488

	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);
}

5489 5490
static void uncharge_batch(struct mem_cgroup *memcg, unsigned long pgpgout,
			   unsigned long nr_anon, unsigned long nr_file,
5491 5492
			   unsigned long nr_kmem, unsigned long nr_huge,
			   unsigned long nr_shmem, struct page *dummy_page)
5493
{
5494
	unsigned long nr_pages = nr_anon + nr_file + nr_kmem;
5495 5496
	unsigned long flags;

5497
	if (!mem_cgroup_is_root(memcg)) {
5498
		page_counter_uncharge(&memcg->memory, nr_pages);
5499
		if (do_memsw_account())
5500
			page_counter_uncharge(&memcg->memsw, nr_pages);
5501 5502
		if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && nr_kmem)
			page_counter_uncharge(&memcg->kmem, nr_kmem);
5503 5504
		memcg_oom_recover(memcg);
	}
5505 5506

	local_irq_save(flags);
5507 5508 5509 5510
	__this_cpu_sub(memcg->stat->count[MEMCG_RSS], nr_anon);
	__this_cpu_sub(memcg->stat->count[MEMCG_CACHE], nr_file);
	__this_cpu_sub(memcg->stat->count[MEMCG_RSS_HUGE], nr_huge);
	__this_cpu_sub(memcg->stat->count[NR_SHMEM], nr_shmem);
5511
	__this_cpu_add(memcg->stat->events[PGPGOUT], pgpgout);
5512
	__this_cpu_add(memcg->stat->nr_page_events, nr_pages);
5513 5514
	memcg_check_events(memcg, dummy_page);
	local_irq_restore(flags);
5515 5516

	if (!mem_cgroup_is_root(memcg))
5517
		css_put_many(&memcg->css, nr_pages);
5518 5519 5520 5521 5522
}

static void uncharge_list(struct list_head *page_list)
{
	struct mem_cgroup *memcg = NULL;
5523
	unsigned long nr_shmem = 0;
5524 5525 5526
	unsigned long nr_anon = 0;
	unsigned long nr_file = 0;
	unsigned long nr_huge = 0;
5527
	unsigned long nr_kmem = 0;
5528 5529 5530 5531
	unsigned long pgpgout = 0;
	struct list_head *next;
	struct page *page;

5532 5533 5534 5535
	/*
	 * Note that the list can be a single page->lru; hence the
	 * do-while loop instead of a simple list_for_each_entry().
	 */
5536 5537 5538 5539 5540 5541
	next = page_list->next;
	do {
		page = list_entry(next, struct page, lru);
		next = page->lru.next;

		VM_BUG_ON_PAGE(PageLRU(page), page);
5542
		VM_BUG_ON_PAGE(!PageHWPoison(page) && page_count(page), page);
5543

5544
		if (!page->mem_cgroup)
5545 5546 5547 5548
			continue;

		/*
		 * Nobody should be changing or seriously looking at
5549
		 * page->mem_cgroup at this point, we have fully
5550
		 * exclusive access to the page.
5551 5552
		 */

5553
		if (memcg != page->mem_cgroup) {
5554
			if (memcg) {
5555
				uncharge_batch(memcg, pgpgout, nr_anon, nr_file,
5556 5557 5558
					       nr_kmem, nr_huge, nr_shmem, page);
				pgpgout = nr_anon = nr_file = nr_kmem = 0;
				nr_huge = nr_shmem = 0;
5559
			}
5560
			memcg = page->mem_cgroup;
5561 5562
		}

5563 5564
		if (!PageKmemcg(page)) {
			unsigned int nr_pages = 1;
5565

5566 5567 5568 5569 5570 5571
			if (PageTransHuge(page)) {
				nr_pages <<= compound_order(page);
				nr_huge += nr_pages;
			}
			if (PageAnon(page))
				nr_anon += nr_pages;
5572
			else {
5573
				nr_file += nr_pages;
5574 5575 5576
				if (PageSwapBacked(page))
					nr_shmem += nr_pages;
			}
5577
			pgpgout++;
5578
		} else {
5579
			nr_kmem += 1 << compound_order(page);
5580 5581
			__ClearPageKmemcg(page);
		}
5582

5583
		page->mem_cgroup = NULL;
5584 5585 5586
	} while (next != page_list);

	if (memcg)
5587
		uncharge_batch(memcg, pgpgout, nr_anon, nr_file,
5588
			       nr_kmem, nr_huge, nr_shmem, page);
5589 5590
}

5591 5592 5593 5594 5595 5596 5597 5598 5599 5600 5601 5602
/**
 * 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;

5603
	/* Don't touch page->lru of any random page, pre-check: */
5604
	if (!page->mem_cgroup)
5605 5606
		return;

5607 5608 5609
	INIT_LIST_HEAD(&page->lru);
	uncharge_list(&page->lru);
}
5610

5611 5612 5613 5614 5615 5616 5617 5618 5619 5620 5621
/**
 * 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;
5622

5623 5624
	if (!list_empty(page_list))
		uncharge_list(page_list);
5625 5626 5627
}

/**
5628 5629 5630
 * mem_cgroup_migrate - charge a page's replacement
 * @oldpage: currently circulating page
 * @newpage: replacement page
5631
 *
5632 5633
 * Charge @newpage as a replacement page for @oldpage. @oldpage will
 * be uncharged upon free.
5634 5635 5636
 *
 * Both pages must be locked, @newpage->mapping must be set up.
 */
5637
void mem_cgroup_migrate(struct page *oldpage, struct page *newpage)
5638
{
5639
	struct mem_cgroup *memcg;
5640 5641
	unsigned int nr_pages;
	bool compound;
5642
	unsigned long flags;
5643 5644 5645 5646

	VM_BUG_ON_PAGE(!PageLocked(oldpage), oldpage);
	VM_BUG_ON_PAGE(!PageLocked(newpage), newpage);
	VM_BUG_ON_PAGE(PageAnon(oldpage) != PageAnon(newpage), newpage);
5647 5648
	VM_BUG_ON_PAGE(PageTransHuge(oldpage) != PageTransHuge(newpage),
		       newpage);
5649 5650 5651 5652 5653

	if (mem_cgroup_disabled())
		return;

	/* Page cache replacement: new page already charged? */
5654
	if (newpage->mem_cgroup)
5655 5656
		return;

5657
	/* Swapcache readahead pages can get replaced before being charged */
5658
	memcg = oldpage->mem_cgroup;
5659
	if (!memcg)
5660 5661
		return;

5662 5663 5664 5665 5666 5667 5668 5669
	/* Force-charge the new page. The old one will be freed soon */
	compound = PageTransHuge(newpage);
	nr_pages = compound ? hpage_nr_pages(newpage) : 1;

	page_counter_charge(&memcg->memory, nr_pages);
	if (do_memsw_account())
		page_counter_charge(&memcg->memsw, nr_pages);
	css_get_many(&memcg->css, nr_pages);
5670

5671
	commit_charge(newpage, memcg, false);
5672

5673
	local_irq_save(flags);
5674 5675
	mem_cgroup_charge_statistics(memcg, newpage, compound, nr_pages);
	memcg_check_events(memcg, newpage);
5676
	local_irq_restore(flags);
5677 5678
}

5679
DEFINE_STATIC_KEY_FALSE(memcg_sockets_enabled_key);
5680 5681
EXPORT_SYMBOL(memcg_sockets_enabled_key);

5682
void mem_cgroup_sk_alloc(struct sock *sk)
5683 5684 5685
{
	struct mem_cgroup *memcg;

5686 5687 5688 5689 5690
	if (!mem_cgroup_sockets_enabled)
		return;

	/*
	 * Socket cloning can throw us here with sk_memcg already
5691 5692 5693 5694 5695 5696 5697 5698 5699 5700 5701 5702 5703 5704 5705
	 * 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);
5706 5707
	if (memcg == root_mem_cgroup)
		goto out;
5708
	if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && !memcg->tcpmem_active)
5709 5710
		goto out;
	if (css_tryget_online(&memcg->css))
5711
		sk->sk_memcg = memcg;
5712
out:
5713 5714 5715
	rcu_read_unlock();
}

5716
void mem_cgroup_sk_free(struct sock *sk)
5717
{
5718 5719
	if (sk->sk_memcg)
		css_put(&sk->sk_memcg->css);
5720 5721 5722 5723 5724 5725 5726 5727 5728 5729 5730 5731
}

/**
 * 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)
{
5732
	gfp_t gfp_mask = GFP_KERNEL;
5733

5734
	if (!cgroup_subsys_on_dfl(memory_cgrp_subsys)) {
5735
		struct page_counter *fail;
5736

5737 5738
		if (page_counter_try_charge(&memcg->tcpmem, nr_pages, &fail)) {
			memcg->tcpmem_pressure = 0;
5739 5740
			return true;
		}
5741 5742
		page_counter_charge(&memcg->tcpmem, nr_pages);
		memcg->tcpmem_pressure = 1;
5743
		return false;
5744
	}
5745

5746 5747 5748 5749
	/* Don't block in the packet receive path */
	if (in_softirq())
		gfp_mask = GFP_NOWAIT;

5750 5751
	this_cpu_add(memcg->stat->count[MEMCG_SOCK], nr_pages);

5752 5753 5754 5755
	if (try_charge(memcg, gfp_mask, nr_pages) == 0)
		return true;

	try_charge(memcg, gfp_mask|__GFP_NOFAIL, nr_pages);
5756 5757 5758 5759 5760 5761 5762 5763 5764 5765
	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)
{
5766
	if (!cgroup_subsys_on_dfl(memory_cgrp_subsys)) {
5767
		page_counter_uncharge(&memcg->tcpmem, nr_pages);
5768 5769
		return;
	}
5770

5771 5772
	this_cpu_sub(memcg->stat->count[MEMCG_SOCK], nr_pages);

5773 5774
	page_counter_uncharge(&memcg->memory, nr_pages);
	css_put_many(&memcg->css, nr_pages);
5775 5776
}

5777 5778 5779 5780 5781 5782 5783 5784 5785
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;
5786 5787
		if (!strcmp(token, "nokmem"))
			cgroup_memory_nokmem = true;
5788 5789 5790 5791
	}
	return 0;
}
__setup("cgroup.memory=", cgroup_memory);
5792

5793
/*
5794 5795
 * subsys_initcall() for memory controller.
 *
5796 5797 5798 5799
 * Some parts like memcg_hotplug_cpu_dead() 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.
5800 5801 5802
 */
static int __init mem_cgroup_init(void)
{
5803 5804
	int cpu, node;

5805 5806 5807
#ifndef CONFIG_SLOB
	/*
	 * Kmem cache creation is mostly done with the slab_mutex held,
5808 5809 5810
	 * so use a workqueue with limited concurrency to avoid stalling
	 * all worker threads in case lots of cgroups are created and
	 * destroyed simultaneously.
5811
	 */
5812 5813
	memcg_kmem_cache_wq = alloc_workqueue("memcg_kmem_cache", 0, 1);
	BUG_ON(!memcg_kmem_cache_wq);
5814 5815
#endif

5816 5817
	cpuhp_setup_state_nocalls(CPUHP_MM_MEMCQ_DEAD, "mm/memctrl:dead", NULL,
				  memcg_hotplug_cpu_dead);
5818 5819 5820 5821 5822 5823 5824 5825 5826 5827 5828

	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;

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

5829 5830
		rtpn->rb_root = RB_ROOT;
		spin_lock_init(&rtpn->lock);
5831 5832 5833
		soft_limit_tree.rb_tree_per_node[node] = rtpn;
	}

5834 5835 5836
	return 0;
}
subsys_initcall(mem_cgroup_init);
5837 5838

#ifdef CONFIG_MEMCG_SWAP
5839 5840 5841 5842 5843 5844 5845 5846 5847 5848 5849 5850 5851 5852 5853 5854 5855 5856
static struct mem_cgroup *mem_cgroup_id_get_online(struct mem_cgroup *memcg)
{
	while (!atomic_inc_not_zero(&memcg->id.ref)) {
		/*
		 * The root cgroup cannot be destroyed, so it's refcount must
		 * always be >= 1.
		 */
		if (WARN_ON_ONCE(memcg == root_mem_cgroup)) {
			VM_BUG_ON(1);
			break;
		}
		memcg = parent_mem_cgroup(memcg);
		if (!memcg)
			memcg = root_mem_cgroup;
	}
	return memcg;
}

5857 5858 5859 5860 5861 5862 5863 5864 5865
/**
 * 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)
{
5866
	struct mem_cgroup *memcg, *swap_memcg;
5867 5868 5869 5870 5871
	unsigned short oldid;

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

5872
	if (!do_memsw_account())
5873 5874 5875 5876 5877 5878 5879 5880
		return;

	memcg = page->mem_cgroup;

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

5881 5882 5883 5884 5885 5886
	/*
	 * In case the memcg owning these pages has been offlined and doesn't
	 * have an ID allocated to it anymore, charge the closest online
	 * ancestor for the swap instead and transfer the memory+swap charge.
	 */
	swap_memcg = mem_cgroup_id_get_online(memcg);
5887
	oldid = swap_cgroup_record(entry, mem_cgroup_id(swap_memcg), 1);
5888
	VM_BUG_ON_PAGE(oldid, page);
5889
	mem_cgroup_swap_statistics(swap_memcg, 1);
5890 5891 5892 5893 5894 5895

	page->mem_cgroup = NULL;

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

5896 5897 5898 5899 5900 5901
	if (memcg != swap_memcg) {
		if (!mem_cgroup_is_root(swap_memcg))
			page_counter_charge(&swap_memcg->memsw, 1);
		page_counter_uncharge(&memcg->memsw, 1);
	}

5902 5903 5904 5905 5906 5907 5908
	/*
	 * 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());
5909
	mem_cgroup_charge_statistics(memcg, page, false, -1);
5910
	memcg_check_events(memcg, page);
5911 5912 5913

	if (!mem_cgroup_is_root(memcg))
		css_put(&memcg->css);
5914 5915
}

5916 5917
/**
 * mem_cgroup_try_charge_swap - try charging swap space for a page
5918 5919 5920
 * @page: page being added to swap
 * @entry: swap entry to charge
 *
5921
 * Try to charge @page's memcg for the swap space at @entry.
5922 5923 5924 5925 5926
 *
 * Returns 0 on success, -ENOMEM on failure.
 */
int mem_cgroup_try_charge_swap(struct page *page, swp_entry_t entry)
{
5927
	unsigned int nr_pages = hpage_nr_pages(page);
5928
	struct page_counter *counter;
5929
	struct mem_cgroup *memcg;
5930 5931 5932 5933 5934 5935 5936 5937 5938 5939 5940
	unsigned short oldid;

	if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) || !do_swap_account)
		return 0;

	memcg = page->mem_cgroup;

	/* Readahead page, never charged */
	if (!memcg)
		return 0;

5941 5942
	memcg = mem_cgroup_id_get_online(memcg);

5943
	if (!mem_cgroup_is_root(memcg) &&
5944
	    !page_counter_try_charge(&memcg->swap, nr_pages, &counter)) {
5945
		mem_cgroup_id_put(memcg);
5946
		return -ENOMEM;
5947
	}
5948

5949 5950 5951 5952
	/* Get references for the tail pages, too */
	if (nr_pages > 1)
		mem_cgroup_id_get_many(memcg, nr_pages - 1);
	oldid = swap_cgroup_record(entry, mem_cgroup_id(memcg), nr_pages);
5953
	VM_BUG_ON_PAGE(oldid, page);
5954
	mem_cgroup_swap_statistics(memcg, nr_pages);
5955 5956 5957 5958

	return 0;
}

5959
/**
5960
 * mem_cgroup_uncharge_swap - uncharge swap space
5961
 * @entry: swap entry to uncharge
5962
 * @nr_pages: the amount of swap space to uncharge
5963
 */
5964
void mem_cgroup_uncharge_swap(swp_entry_t entry, unsigned int nr_pages)
5965 5966 5967 5968
{
	struct mem_cgroup *memcg;
	unsigned short id;

5969
	if (!do_swap_account)
5970 5971
		return;

5972
	id = swap_cgroup_record(entry, 0, nr_pages);
5973
	rcu_read_lock();
5974
	memcg = mem_cgroup_from_id(id);
5975
	if (memcg) {
5976 5977
		if (!mem_cgroup_is_root(memcg)) {
			if (cgroup_subsys_on_dfl(memory_cgrp_subsys))
5978
				page_counter_uncharge(&memcg->swap, nr_pages);
5979
			else
5980
				page_counter_uncharge(&memcg->memsw, nr_pages);
5981
		}
5982 5983
		mem_cgroup_swap_statistics(memcg, -nr_pages);
		mem_cgroup_id_put_many(memcg, nr_pages);
5984 5985 5986 5987
	}
	rcu_read_unlock();
}

5988 5989 5990 5991 5992 5993 5994 5995 5996 5997 5998 5999 6000
long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg)
{
	long nr_swap_pages = get_nr_swap_pages();

	if (!do_swap_account || !cgroup_subsys_on_dfl(memory_cgrp_subsys))
		return nr_swap_pages;
	for (; memcg != root_mem_cgroup; memcg = parent_mem_cgroup(memcg))
		nr_swap_pages = min_t(long, nr_swap_pages,
				      READ_ONCE(memcg->swap.limit) -
				      page_counter_read(&memcg->swap));
	return nr_swap_pages;
}

6001 6002 6003 6004 6005 6006 6007 6008 6009 6010 6011 6012 6013 6014 6015 6016 6017 6018 6019 6020 6021 6022
bool mem_cgroup_swap_full(struct page *page)
{
	struct mem_cgroup *memcg;

	VM_BUG_ON_PAGE(!PageLocked(page), page);

	if (vm_swap_full())
		return true;
	if (!do_swap_account || !cgroup_subsys_on_dfl(memory_cgrp_subsys))
		return false;

	memcg = page->mem_cgroup;
	if (!memcg)
		return false;

	for (; memcg != root_mem_cgroup; memcg = parent_mem_cgroup(memcg))
		if (page_counter_read(&memcg->swap) * 2 >= memcg->swap.limit)
			return true;

	return false;
}

6023 6024 6025 6026 6027 6028 6029 6030 6031 6032 6033 6034 6035 6036 6037 6038 6039
/* 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);

6040 6041 6042 6043 6044 6045 6046 6047 6048 6049 6050 6051 6052 6053 6054 6055 6056 6057 6058 6059 6060 6061 6062 6063 6064 6065 6066 6067 6068 6069 6070 6071 6072 6073 6074 6075 6076 6077 6078 6079 6080 6081 6082 6083 6084 6085 6086 6087 6088 6089 6090 6091 6092 6093 6094 6095 6096
static u64 swap_current_read(struct cgroup_subsys_state *css,
			     struct cftype *cft)
{
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);

	return (u64)page_counter_read(&memcg->swap) * PAGE_SIZE;
}

static int swap_max_show(struct seq_file *m, void *v)
{
	struct mem_cgroup *memcg = mem_cgroup_from_css(seq_css(m));
	unsigned long max = READ_ONCE(memcg->swap.limit);

	if (max == PAGE_COUNTER_MAX)
		seq_puts(m, "max\n");
	else
		seq_printf(m, "%llu\n", (u64)max * PAGE_SIZE);

	return 0;
}

static ssize_t swap_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);
	err = page_counter_memparse(buf, "max", &max);
	if (err)
		return err;

	mutex_lock(&memcg_limit_mutex);
	err = page_counter_limit(&memcg->swap, max);
	mutex_unlock(&memcg_limit_mutex);
	if (err)
		return err;

	return nbytes;
}

static struct cftype swap_files[] = {
	{
		.name = "swap.current",
		.flags = CFTYPE_NOT_ON_ROOT,
		.read_u64 = swap_current_read,
	},
	{
		.name = "swap.max",
		.flags = CFTYPE_NOT_ON_ROOT,
		.seq_show = swap_max_show,
		.write = swap_max_write,
	},
	{ }	/* terminate */
};

6097 6098 6099 6100 6101 6102 6103 6104 6105 6106 6107 6108 6109 6110 6111 6112 6113 6114 6115 6116 6117 6118 6119 6120 6121 6122 6123 6124 6125 6126 6127
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;
6128 6129
		WARN_ON(cgroup_add_dfl_cftypes(&memory_cgrp_subsys,
					       swap_files));
6130 6131 6132 6133 6134 6135 6136 6137
		WARN_ON(cgroup_add_legacy_cftypes(&memory_cgrp_subsys,
						  memsw_cgroup_files));
	}
	return 0;
}
subsys_initcall(mem_cgroup_swap_init);

#endif /* CONFIG_MEMCG_SWAP */