memcontrol.c 192.3 KB
Newer Older
1
// SPDX-License-Identifier: GPL-2.0-or-later
B
Balbir Singh 已提交
2 3 4 5 6
/* memcontrol.c - Memory Controller
 *
 * Copyright IBM Corporation, 2007
 * Author Balbir Singh <balbir@linux.vnet.ibm.com>
 *
7 8 9
 * Copyright 2007 OpenVZ SWsoft Inc
 * Author: Pavel Emelianov <xemul@openvz.org>
 *
10 11 12 13
 * Memory thresholds
 * Copyright (C) 2009 Nokia Corporation
 * Author: Kirill A. Shutemov
 *
14 15 16 17
 * Kernel Memory Controller
 * Copyright (C) 2012 Parallels Inc. and Google Inc.
 * Authors: Glauber Costa and Suleiman Souhlal
 *
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
#include <linux/page_counter.h>
B
Balbir Singh 已提交
26 27
#include <linux/memcontrol.h>
#include <linux/cgroup.h>
28
#include <linux/pagewalk.h>
29
#include <linux/sched/mm.h>
30
#include <linux/shmem_fs.h>
31
#include <linux/hugetlb.h>
K
KAMEZAWA Hiroyuki 已提交
32
#include <linux/pagemap.h>
33
#include <linux/vm_event_item.h>
34
#include <linux/smp.h>
35
#include <linux/page-flags.h>
36
#include <linux/backing-dev.h>
37 38
#include <linux/bit_spinlock.h>
#include <linux/rcupdate.h>
39
#include <linux/limits.h>
40
#include <linux/export.h>
41
#include <linux/mutex.h>
42
#include <linux/rbtree.h>
43
#include <linux/slab.h>
44
#include <linux/swap.h>
45
#include <linux/swapops.h>
46
#include <linux/spinlock.h>
47
#include <linux/eventfd.h>
48
#include <linux/poll.h>
49
#include <linux/sort.h>
50
#include <linux/fs.h>
51
#include <linux/seq_file.h>
52
#include <linux/vmpressure.h>
53
#include <linux/mm_inline.h>
54
#include <linux/swap_cgroup.h>
55
#include <linux/cpu.h>
56
#include <linux/oom.h>
57
#include <linux/lockdep.h>
58
#include <linux/file.h>
59
#include <linux/tracehook.h>
60
#include <linux/psi.h>
61
#include <linux/seq_buf.h>
K
KAMEZAWA Hiroyuki 已提交
62
#include "internal.h"
G
Glauber Costa 已提交
63
#include <net/sock.h>
M
Michal Hocko 已提交
64
#include <net/ip.h>
65
#include "slab.h"
B
Balbir Singh 已提交
66

67
#include <linux/uaccess.h>
68

69 70
#include <trace/events/vmscan.h>

71 72
struct cgroup_subsys memory_cgrp_subsys __read_mostly;
EXPORT_SYMBOL(memory_cgrp_subsys);
73

74 75
struct mem_cgroup *root_mem_cgroup __read_mostly;

76 77 78
/* Active memory cgroup to use from an interrupt context */
DEFINE_PER_CPU(struct mem_cgroup *, int_active_memcg);

79 80 81
/* Socket memory accounting disabled? */
static bool cgroup_memory_nosocket;

82 83 84
/* Kernel memory accounting disabled? */
static bool cgroup_memory_nokmem;

85
/* Whether the swap controller is active */
A
Andrew Morton 已提交
86
#ifdef CONFIG_MEMCG_SWAP
87
bool cgroup_memory_noswap __read_mostly;
88
#else
89
#define cgroup_memory_noswap		1
90
#endif
91

92 93 94 95
#ifdef CONFIG_CGROUP_WRITEBACK
static DECLARE_WAIT_QUEUE_HEAD(memcg_cgwb_frn_waitq);
#endif

96 97 98
/* Whether legacy memory+swap accounting is active */
static bool do_memsw_account(void)
{
99
	return !cgroup_subsys_on_dfl(memory_cgrp_subsys) && !cgroup_memory_noswap;
100 101
}

102 103
#define THRESHOLDS_EVENTS_TARGET 128
#define SOFTLIMIT_EVENTS_TARGET 1024
104

105 106 107 108 109
/*
 * Cgroups above their limits are maintained in a RB-Tree, independent of
 * their hierarchy representation
 */

110
struct mem_cgroup_tree_per_node {
111
	struct rb_root rb_root;
112
	struct rb_node *rb_rightmost;
113 114 115 116 117 118 119 120 121
	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
122 123 124 125 126
/* for OOM */
struct mem_cgroup_eventfd_list {
	struct list_head list;
	struct eventfd_ctx *eventfd;
};
127

128 129 130
/*
 * cgroup_event represents events which userspace want to receive.
 */
131
struct mem_cgroup_event {
132
	/*
133
	 * memcg which the event belongs to.
134
	 */
135
	struct mem_cgroup *memcg;
136 137 138 139 140 141 142 143
	/*
	 * 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;
144 145 146 147 148
	/*
	 * 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.
	 */
149
	int (*register_event)(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
150
			      struct eventfd_ctx *eventfd, const char *args);
151 152 153 154 155
	/*
	 * 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.
	 */
156
	void (*unregister_event)(struct mem_cgroup *memcg,
157
				 struct eventfd_ctx *eventfd);
158 159 160 161 162 163
	/*
	 * All fields below needed to unregister event when
	 * userspace closes eventfd.
	 */
	poll_table pt;
	wait_queue_head_t *wqh;
164
	wait_queue_entry_t wait;
165 166 167
	struct work_struct remove;
};

168 169
static void mem_cgroup_threshold(struct mem_cgroup *memcg);
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg);
170

171 172
/* Stuffs for move charges at task migration. */
/*
173
 * Types of charges to be moved.
174
 */
175 176 177
#define MOVE_ANON	0x1U
#define MOVE_FILE	0x2U
#define MOVE_MASK	(MOVE_ANON | MOVE_FILE)
178

179 180
/* "mc" and its members are protected by cgroup_mutex */
static struct move_charge_struct {
181
	spinlock_t	  lock; /* for from, to */
182
	struct mm_struct  *mm;
183 184
	struct mem_cgroup *from;
	struct mem_cgroup *to;
185
	unsigned long flags;
186
	unsigned long precharge;
187
	unsigned long moved_charge;
188
	unsigned long moved_swap;
189 190 191
	struct task_struct *moving_task;	/* a task moving charges */
	wait_queue_head_t waitq;		/* a waitq for other context */
} mc = {
192
	.lock = __SPIN_LOCK_UNLOCKED(mc.lock),
193 194
	.waitq = __WAIT_QUEUE_HEAD_INITIALIZER(mc.waitq),
};
195

196 197 198 199
/*
 * Maximum loops in mem_cgroup_hierarchical_reclaim(), used for soft
 * limit reclaim to prevent infinite loops, if they ever occur.
 */
200
#define	MEM_CGROUP_MAX_RECLAIM_LOOPS		100
201
#define	MEM_CGROUP_MAX_SOFT_LIMIT_RECLAIM_LOOPS	2
202

203
/* for encoding cft->private value on file */
G
Glauber Costa 已提交
204 205 206 207
enum res_type {
	_MEM,
	_MEMSWAP,
	_OOM_TYPE,
208
	_KMEM,
V
Vladimir Davydov 已提交
209
	_TCP,
G
Glauber Costa 已提交
210 211
};

212 213
#define MEMFILE_PRIVATE(x, val)	((x) << 16 | (val))
#define MEMFILE_TYPE(val)	((val) >> 16 & 0xffff)
214
#define MEMFILE_ATTR(val)	((val) & 0xffff)
K
KAMEZAWA Hiroyuki 已提交
215 216
/* Used for OOM nofiier */
#define OOM_CONTROL		(0)
217

218 219 220 221 222 223 224 225 226 227 228 229 230 231 232
/*
 * 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)		\
	for (iter = mem_cgroup_iter(root, NULL, NULL);	\
	     iter != NULL;				\
	     iter = mem_cgroup_iter(root, iter, NULL))

#define for_each_mem_cgroup(iter)			\
	for (iter = mem_cgroup_iter(NULL, NULL, NULL);	\
	     iter != NULL;				\
	     iter = mem_cgroup_iter(NULL, iter, NULL))

233 234 235 236 237 238
static inline bool should_force_charge(void)
{
	return tsk_is_oom_victim(current) || fatal_signal_pending(current) ||
		(current->flags & PF_EXITING);
}

239 240 241 242 243 244 245 246 247 248 249 250 251
/* 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;
}

252
#ifdef CONFIG_MEMCG_KMEM
R
Roman Gushchin 已提交
253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344
extern spinlock_t css_set_lock;

static void obj_cgroup_release(struct percpu_ref *ref)
{
	struct obj_cgroup *objcg = container_of(ref, struct obj_cgroup, refcnt);
	struct mem_cgroup *memcg;
	unsigned int nr_bytes;
	unsigned int nr_pages;
	unsigned long flags;

	/*
	 * At this point all allocated objects are freed, and
	 * objcg->nr_charged_bytes can't have an arbitrary byte value.
	 * However, it can be PAGE_SIZE or (x * PAGE_SIZE).
	 *
	 * The following sequence can lead to it:
	 * 1) CPU0: objcg == stock->cached_objcg
	 * 2) CPU1: we do a small allocation (e.g. 92 bytes),
	 *          PAGE_SIZE bytes are charged
	 * 3) CPU1: a process from another memcg is allocating something,
	 *          the stock if flushed,
	 *          objcg->nr_charged_bytes = PAGE_SIZE - 92
	 * 5) CPU0: we do release this object,
	 *          92 bytes are added to stock->nr_bytes
	 * 6) CPU0: stock is flushed,
	 *          92 bytes are added to objcg->nr_charged_bytes
	 *
	 * In the result, nr_charged_bytes == PAGE_SIZE.
	 * This page will be uncharged in obj_cgroup_release().
	 */
	nr_bytes = atomic_read(&objcg->nr_charged_bytes);
	WARN_ON_ONCE(nr_bytes & (PAGE_SIZE - 1));
	nr_pages = nr_bytes >> PAGE_SHIFT;

	spin_lock_irqsave(&css_set_lock, flags);
	memcg = obj_cgroup_memcg(objcg);
	if (nr_pages)
		__memcg_kmem_uncharge(memcg, nr_pages);
	list_del(&objcg->list);
	mem_cgroup_put(memcg);
	spin_unlock_irqrestore(&css_set_lock, flags);

	percpu_ref_exit(ref);
	kfree_rcu(objcg, rcu);
}

static struct obj_cgroup *obj_cgroup_alloc(void)
{
	struct obj_cgroup *objcg;
	int ret;

	objcg = kzalloc(sizeof(struct obj_cgroup), GFP_KERNEL);
	if (!objcg)
		return NULL;

	ret = percpu_ref_init(&objcg->refcnt, obj_cgroup_release, 0,
			      GFP_KERNEL);
	if (ret) {
		kfree(objcg);
		return NULL;
	}
	INIT_LIST_HEAD(&objcg->list);
	return objcg;
}

static void memcg_reparent_objcgs(struct mem_cgroup *memcg,
				  struct mem_cgroup *parent)
{
	struct obj_cgroup *objcg, *iter;

	objcg = rcu_replace_pointer(memcg->objcg, NULL, true);

	spin_lock_irq(&css_set_lock);

	/* Move active objcg to the parent's list */
	xchg(&objcg->memcg, parent);
	css_get(&parent->css);
	list_add(&objcg->list, &parent->objcg_list);

	/* Move already reparented objcgs to the parent's list */
	list_for_each_entry(iter, &memcg->objcg_list, list) {
		css_get(&parent->css);
		xchg(&iter->memcg, parent);
		css_put(&memcg->css);
	}
	list_splice(&memcg->objcg_list, &parent->objcg_list);

	spin_unlock_irq(&css_set_lock);

	percpu_ref_kill(&objcg->refcnt);
}

345
/*
346
 * This will be used as a shrinker list's index.
L
Li Zefan 已提交
347 348 349 350 351
 * 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.
352
 *
353 354
 * The current size of the caches array is stored in memcg_nr_cache_ids. It
 * will double each time we have to increase it.
355
 */
356 357
static DEFINE_IDA(memcg_cache_ida);
int memcg_nr_cache_ids;
358

359 360 361 362 363 364 365 366 367 368 369 370 371
/* 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);
}

372 373 374 375 376 377
/*
 * 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
378
 * MAX_SIZE should be as large as the number of cgrp_ids. Ideally, we could get
379 380
 * 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
381
 * cgrp_id space is not getting any smaller, and we don't have to necessarily
382 383 384
 * increase ours as well if it increases.
 */
#define MEMCG_CACHES_MIN_SIZE 4
L
Li Zefan 已提交
385
#define MEMCG_CACHES_MAX_SIZE MEM_CGROUP_ID_MAX
386

387 388
/*
 * A lot of the calls to the cache allocation functions are expected to be
389
 * inlined by the compiler. Since the calls to memcg_slab_pre_alloc_hook() are
390 391 392
 * 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
 */
393
DEFINE_STATIC_KEY_FALSE(memcg_kmem_enabled_key);
394
EXPORT_SYMBOL(memcg_kmem_enabled_key);
395
#endif
396

397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419
static int memcg_shrinker_map_size;
static DEFINE_MUTEX(memcg_shrinker_map_mutex);

static void memcg_free_shrinker_map_rcu(struct rcu_head *head)
{
	kvfree(container_of(head, struct memcg_shrinker_map, rcu));
}

static int memcg_expand_one_shrinker_map(struct mem_cgroup *memcg,
					 int size, int old_size)
{
	struct memcg_shrinker_map *new, *old;
	int nid;

	lockdep_assert_held(&memcg_shrinker_map_mutex);

	for_each_node(nid) {
		old = rcu_dereference_protected(
			mem_cgroup_nodeinfo(memcg, nid)->shrinker_map, true);
		/* Not yet online memcg */
		if (!old)
			return 0;

420
		new = kvmalloc_node(sizeof(*new) + size, GFP_KERNEL, nid);
421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463
		if (!new)
			return -ENOMEM;

		/* Set all old bits, clear all new bits */
		memset(new->map, (int)0xff, old_size);
		memset((void *)new->map + old_size, 0, size - old_size);

		rcu_assign_pointer(memcg->nodeinfo[nid]->shrinker_map, new);
		call_rcu(&old->rcu, memcg_free_shrinker_map_rcu);
	}

	return 0;
}

static void memcg_free_shrinker_maps(struct mem_cgroup *memcg)
{
	struct mem_cgroup_per_node *pn;
	struct memcg_shrinker_map *map;
	int nid;

	if (mem_cgroup_is_root(memcg))
		return;

	for_each_node(nid) {
		pn = mem_cgroup_nodeinfo(memcg, nid);
		map = rcu_dereference_protected(pn->shrinker_map, true);
		if (map)
			kvfree(map);
		rcu_assign_pointer(pn->shrinker_map, NULL);
	}
}

static int memcg_alloc_shrinker_maps(struct mem_cgroup *memcg)
{
	struct memcg_shrinker_map *map;
	int nid, size, ret = 0;

	if (mem_cgroup_is_root(memcg))
		return 0;

	mutex_lock(&memcg_shrinker_map_mutex);
	size = memcg_shrinker_map_size;
	for_each_node(nid) {
464
		map = kvzalloc_node(sizeof(*map) + size, GFP_KERNEL, nid);
465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494
		if (!map) {
			memcg_free_shrinker_maps(memcg);
			ret = -ENOMEM;
			break;
		}
		rcu_assign_pointer(memcg->nodeinfo[nid]->shrinker_map, map);
	}
	mutex_unlock(&memcg_shrinker_map_mutex);

	return ret;
}

int memcg_expand_shrinker_maps(int new_id)
{
	int size, old_size, ret = 0;
	struct mem_cgroup *memcg;

	size = DIV_ROUND_UP(new_id + 1, BITS_PER_LONG) * sizeof(unsigned long);
	old_size = memcg_shrinker_map_size;
	if (size <= old_size)
		return 0;

	mutex_lock(&memcg_shrinker_map_mutex);
	if (!root_mem_cgroup)
		goto unlock;

	for_each_mem_cgroup(memcg) {
		if (mem_cgroup_is_root(memcg))
			continue;
		ret = memcg_expand_one_shrinker_map(memcg, size, old_size);
495 496
		if (ret) {
			mem_cgroup_iter_break(NULL, memcg);
497
			goto unlock;
498
		}
499 500 501 502 503 504 505
	}
unlock:
	if (!ret)
		memcg_shrinker_map_size = size;
	mutex_unlock(&memcg_shrinker_map_mutex);
	return ret;
}
506 507 508 509 510 511 512 513

void memcg_set_shrinker_bit(struct mem_cgroup *memcg, int nid, int shrinker_id)
{
	if (shrinker_id >= 0 && memcg && !mem_cgroup_is_root(memcg)) {
		struct memcg_shrinker_map *map;

		rcu_read_lock();
		map = rcu_dereference(memcg->nodeinfo[nid]->shrinker_map);
514 515
		/* Pairs with smp mb in shrink_slab() */
		smp_mb__before_atomic();
516 517 518 519 520
		set_bit(shrinker_id, map->map);
		rcu_read_unlock();
	}
}

521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537
/**
 * 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;

538
	if (!memcg || !cgroup_subsys_on_dfl(memory_cgrp_subsys))
539 540 541 542 543
		memcg = root_mem_cgroup;

	return &memcg->css;
}

544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562
/**
 * 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();
563
	memcg = page->mem_cgroup;
564

565 566 567 568 569 570 571 572
	/*
	 * The lowest bit set means that memcg isn't a valid
	 * memcg pointer, but a obj_cgroups pointer.
	 * In this case the page is shared and doesn't belong
	 * to any specific memory cgroup.
	 */
	if ((unsigned long) memcg & 0x1UL)
		memcg = NULL;
573

574 575 576 577 578 579 580 581
	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;
}

582 583
static struct mem_cgroup_per_node *
mem_cgroup_page_nodeinfo(struct mem_cgroup *memcg, struct page *page)
584
{
585
	int nid = page_to_nid(page);
586

587
	return memcg->nodeinfo[nid];
588 589
}

590 591
static struct mem_cgroup_tree_per_node *
soft_limit_tree_node(int nid)
592
{
593
	return soft_limit_tree.rb_tree_per_node[nid];
594 595
}

596
static struct mem_cgroup_tree_per_node *
597 598 599 600
soft_limit_tree_from_page(struct page *page)
{
	int nid = page_to_nid(page);

601
	return soft_limit_tree.rb_tree_per_node[nid];
602 603
}

604 605
static void __mem_cgroup_insert_exceeded(struct mem_cgroup_per_node *mz,
					 struct mem_cgroup_tree_per_node *mctz,
606
					 unsigned long new_usage_in_excess)
607 608 609
{
	struct rb_node **p = &mctz->rb_root.rb_node;
	struct rb_node *parent = NULL;
610
	struct mem_cgroup_per_node *mz_node;
611
	bool rightmost = true;
612 613 614 615 616 617 618 619 620

	if (mz->on_tree)
		return;

	mz->usage_in_excess = new_usage_in_excess;
	if (!mz->usage_in_excess)
		return;
	while (*p) {
		parent = *p;
621
		mz_node = rb_entry(parent, struct mem_cgroup_per_node,
622
					tree_node);
623
		if (mz->usage_in_excess < mz_node->usage_in_excess) {
624
			p = &(*p)->rb_left;
625 626 627
			rightmost = false;
		}

628 629 630 631 632 633 634
		/*
		 * 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;
	}
635 636 637 638

	if (rightmost)
		mctz->rb_rightmost = &mz->tree_node;

639 640 641 642 643
	rb_link_node(&mz->tree_node, parent, p);
	rb_insert_color(&mz->tree_node, &mctz->rb_root);
	mz->on_tree = true;
}

644 645
static void __mem_cgroup_remove_exceeded(struct mem_cgroup_per_node *mz,
					 struct mem_cgroup_tree_per_node *mctz)
646 647 648
{
	if (!mz->on_tree)
		return;
649 650 651 652

	if (&mz->tree_node == mctz->rb_rightmost)
		mctz->rb_rightmost = rb_prev(&mz->tree_node);

653 654 655 656
	rb_erase(&mz->tree_node, &mctz->rb_root);
	mz->on_tree = false;
}

657 658
static void mem_cgroup_remove_exceeded(struct mem_cgroup_per_node *mz,
				       struct mem_cgroup_tree_per_node *mctz)
659
{
660 661 662
	unsigned long flags;

	spin_lock_irqsave(&mctz->lock, flags);
663
	__mem_cgroup_remove_exceeded(mz, mctz);
664
	spin_unlock_irqrestore(&mctz->lock, flags);
665 666
}

667 668 669
static unsigned long soft_limit_excess(struct mem_cgroup *memcg)
{
	unsigned long nr_pages = page_counter_read(&memcg->memory);
670
	unsigned long soft_limit = READ_ONCE(memcg->soft_limit);
671 672 673 674 675 676 677
	unsigned long excess = 0;

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

	return excess;
}
678 679 680

static void mem_cgroup_update_tree(struct mem_cgroup *memcg, struct page *page)
{
681
	unsigned long excess;
682 683
	struct mem_cgroup_per_node *mz;
	struct mem_cgroup_tree_per_node *mctz;
684

685
	mctz = soft_limit_tree_from_page(page);
686 687
	if (!mctz)
		return;
688 689 690 691 692
	/*
	 * Necessary to update all ancestors when hierarchy is used.
	 * because their event counter is not touched.
	 */
	for (; memcg; memcg = parent_mem_cgroup(memcg)) {
693
		mz = mem_cgroup_page_nodeinfo(memcg, page);
694
		excess = soft_limit_excess(memcg);
695 696 697 698 699
		/*
		 * We have to update the tree if mz is on RB-tree or
		 * mem is over its softlimit.
		 */
		if (excess || mz->on_tree) {
700 701 702
			unsigned long flags;

			spin_lock_irqsave(&mctz->lock, flags);
703 704
			/* if on-tree, remove it */
			if (mz->on_tree)
705
				__mem_cgroup_remove_exceeded(mz, mctz);
706 707 708 709
			/*
			 * Insert again. mz->usage_in_excess will be updated.
			 * If excess is 0, no tree ops.
			 */
710
			__mem_cgroup_insert_exceeded(mz, mctz, excess);
711
			spin_unlock_irqrestore(&mctz->lock, flags);
712 713 714 715 716 717
		}
	}
}

static void mem_cgroup_remove_from_trees(struct mem_cgroup *memcg)
{
718 719 720
	struct mem_cgroup_tree_per_node *mctz;
	struct mem_cgroup_per_node *mz;
	int nid;
721

722
	for_each_node(nid) {
723 724
		mz = mem_cgroup_nodeinfo(memcg, nid);
		mctz = soft_limit_tree_node(nid);
725 726
		if (mctz)
			mem_cgroup_remove_exceeded(mz, mctz);
727 728 729
	}
}

730 731
static struct mem_cgroup_per_node *
__mem_cgroup_largest_soft_limit_node(struct mem_cgroup_tree_per_node *mctz)
732
{
733
	struct mem_cgroup_per_node *mz;
734 735 736

retry:
	mz = NULL;
737
	if (!mctz->rb_rightmost)
738 739
		goto done;		/* Nothing to reclaim from */

740 741
	mz = rb_entry(mctz->rb_rightmost,
		      struct mem_cgroup_per_node, tree_node);
742 743 744 745 746
	/*
	 * 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.
	 */
747
	__mem_cgroup_remove_exceeded(mz, mctz);
748
	if (!soft_limit_excess(mz->memcg) ||
S
Shakeel Butt 已提交
749
	    !css_tryget(&mz->memcg->css))
750 751 752 753 754
		goto retry;
done:
	return mz;
}

755 756
static struct mem_cgroup_per_node *
mem_cgroup_largest_soft_limit_node(struct mem_cgroup_tree_per_node *mctz)
757
{
758
	struct mem_cgroup_per_node *mz;
759

760
	spin_lock_irq(&mctz->lock);
761
	mz = __mem_cgroup_largest_soft_limit_node(mctz);
762
	spin_unlock_irq(&mctz->lock);
763 764 765
	return mz;
}

766 767 768 769 770 771 772 773
/**
 * __mod_memcg_state - update cgroup memory statistics
 * @memcg: the memory cgroup
 * @idx: the stat item - can be enum memcg_stat_item or enum node_stat_item
 * @val: delta to add to the counter, can be negative
 */
void __mod_memcg_state(struct mem_cgroup *memcg, int idx, int val)
{
774
	long x, threshold = MEMCG_CHARGE_BATCH;
775 776 777 778

	if (mem_cgroup_disabled())
		return;

779
	if (memcg_stat_item_in_bytes(idx))
780 781
		threshold <<= PAGE_SHIFT;

782
	x = val + __this_cpu_read(memcg->vmstats_percpu->stat[idx]);
783
	if (unlikely(abs(x) > threshold)) {
784 785
		struct mem_cgroup *mi;

786 787 788 789 790
		/*
		 * Batch local counters to keep them in sync with
		 * the hierarchical ones.
		 */
		__this_cpu_add(memcg->vmstats_local->stat[idx], x);
791 792
		for (mi = memcg; mi; mi = parent_mem_cgroup(mi))
			atomic_long_add(x, &mi->vmstats[idx]);
793 794 795 796 797
		x = 0;
	}
	__this_cpu_write(memcg->vmstats_percpu->stat[idx], x);
}

798 799 800 801 802 803 804 805 806 807 808
static struct mem_cgroup_per_node *
parent_nodeinfo(struct mem_cgroup_per_node *pn, int nid)
{
	struct mem_cgroup *parent;

	parent = parent_mem_cgroup(pn->memcg);
	if (!parent)
		return NULL;
	return mem_cgroup_nodeinfo(parent, nid);
}

809 810
void __mod_memcg_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
			      int val)
811 812
{
	struct mem_cgroup_per_node *pn;
813
	struct mem_cgroup *memcg;
814
	long x, threshold = MEMCG_CHARGE_BATCH;
815 816

	pn = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
817
	memcg = pn->memcg;
818 819

	/* Update memcg */
820
	__mod_memcg_state(memcg, idx, val);
821

822 823 824
	/* Update lruvec */
	__this_cpu_add(pn->lruvec_stat_local->count[idx], val);

825 826 827
	if (vmstat_item_in_bytes(idx))
		threshold <<= PAGE_SHIFT;

828
	x = val + __this_cpu_read(pn->lruvec_stat_cpu->count[idx]);
829
	if (unlikely(abs(x) > threshold)) {
830
		pg_data_t *pgdat = lruvec_pgdat(lruvec);
831 832 833 834
		struct mem_cgroup_per_node *pi;

		for (pi = pn; pi; pi = parent_nodeinfo(pi, pgdat->node_id))
			atomic_long_add(x, &pi->lruvec_stat[idx]);
835 836 837 838 839
		x = 0;
	}
	__this_cpu_write(pn->lruvec_stat_cpu->count[idx], x);
}

840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860
/**
 * __mod_lruvec_state - update lruvec memory statistics
 * @lruvec: the lruvec
 * @idx: the stat item
 * @val: delta to add to the counter, can be negative
 *
 * The lruvec is the intersection of the NUMA node and a cgroup. This
 * function updates the all three counters that are affected by a
 * change of state at this level: per-node, per-cgroup, per-lruvec.
 */
void __mod_lruvec_state(struct lruvec *lruvec, enum node_stat_item idx,
			int val)
{
	/* Update node */
	__mod_node_page_state(lruvec_pgdat(lruvec), idx, val);

	/* Update memcg and lruvec */
	if (!mem_cgroup_disabled())
		__mod_memcg_lruvec_state(lruvec, idx, val);
}

861 862
void __mod_lruvec_slab_state(void *p, enum node_stat_item idx, int val)
{
863
	pg_data_t *pgdat = page_pgdat(virt_to_page(p));
864 865 866 867
	struct mem_cgroup *memcg;
	struct lruvec *lruvec;

	rcu_read_lock();
868
	memcg = mem_cgroup_from_obj(p);
869

870 871 872 873 874 875 876
	/*
	 * Untracked pages have no memcg, no lruvec. Update only the
	 * node. If we reparent the slab objects to the root memcg,
	 * when we free the slab object, we need to update the per-memcg
	 * vmstats to keep it correct for the root memcg.
	 */
	if (!memcg) {
877 878
		__mod_node_page_state(pgdat, idx, val);
	} else {
879
		lruvec = mem_cgroup_lruvec(memcg, pgdat);
880 881 882 883 884
		__mod_lruvec_state(lruvec, idx, val);
	}
	rcu_read_unlock();
}

885 886 887 888 889 890 891 892 893 894 895
void mod_memcg_obj_state(void *p, int idx, int val)
{
	struct mem_cgroup *memcg;

	rcu_read_lock();
	memcg = mem_cgroup_from_obj(p);
	if (memcg)
		mod_memcg_state(memcg, idx, val);
	rcu_read_unlock();
}

896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911
/**
 * __count_memcg_events - account VM events in a cgroup
 * @memcg: the memory cgroup
 * @idx: the event item
 * @count: the number of events that occured
 */
void __count_memcg_events(struct mem_cgroup *memcg, enum vm_event_item idx,
			  unsigned long count)
{
	unsigned long x;

	if (mem_cgroup_disabled())
		return;

	x = count + __this_cpu_read(memcg->vmstats_percpu->events[idx]);
	if (unlikely(x > MEMCG_CHARGE_BATCH)) {
912 913
		struct mem_cgroup *mi;

914 915 916 917 918
		/*
		 * Batch local counters to keep them in sync with
		 * the hierarchical ones.
		 */
		__this_cpu_add(memcg->vmstats_local->events[idx], x);
919 920
		for (mi = memcg; mi; mi = parent_mem_cgroup(mi))
			atomic_long_add(x, &mi->vmevents[idx]);
921 922 923 924 925
		x = 0;
	}
	__this_cpu_write(memcg->vmstats_percpu->events[idx], x);
}

926
static unsigned long memcg_events(struct mem_cgroup *memcg, int event)
927
{
928
	return atomic_long_read(&memcg->vmevents[event]);
929 930
}

931 932
static unsigned long memcg_events_local(struct mem_cgroup *memcg, int event)
{
933 934 935 936 937 938
	long x = 0;
	int cpu;

	for_each_possible_cpu(cpu)
		x += per_cpu(memcg->vmstats_local->events[event], cpu);
	return x;
939 940
}

941
static void mem_cgroup_charge_statistics(struct mem_cgroup *memcg,
942
					 struct page *page,
943
					 int nr_pages)
944
{
945 946
	/* pagein of a big page is an event. So, ignore page size */
	if (nr_pages > 0)
947
		__count_memcg_events(memcg, PGPGIN, 1);
948
	else {
949
		__count_memcg_events(memcg, PGPGOUT, 1);
950 951
		nr_pages = -nr_pages; /* for event */
	}
952

953
	__this_cpu_add(memcg->vmstats_percpu->nr_page_events, nr_pages);
954 955
}

956 957
static bool mem_cgroup_event_ratelimit(struct mem_cgroup *memcg,
				       enum mem_cgroup_events_target target)
958 959 960
{
	unsigned long val, next;

961 962
	val = __this_cpu_read(memcg->vmstats_percpu->nr_page_events);
	next = __this_cpu_read(memcg->vmstats_percpu->targets[target]);
963
	/* from time_after() in jiffies.h */
964
	if ((long)(next - val) < 0) {
965 966 967 968
		switch (target) {
		case MEM_CGROUP_TARGET_THRESH:
			next = val + THRESHOLDS_EVENTS_TARGET;
			break;
969 970 971
		case MEM_CGROUP_TARGET_SOFTLIMIT:
			next = val + SOFTLIMIT_EVENTS_TARGET;
			break;
972 973 974
		default:
			break;
		}
975
		__this_cpu_write(memcg->vmstats_percpu->targets[target], next);
976
		return true;
977
	}
978
	return false;
979 980 981 982 983 984
}

/*
 * Check events in order.
 *
 */
985
static void memcg_check_events(struct mem_cgroup *memcg, struct page *page)
986 987
{
	/* threshold event is triggered in finer grain than soft limit */
988 989
	if (unlikely(mem_cgroup_event_ratelimit(memcg,
						MEM_CGROUP_TARGET_THRESH))) {
990
		bool do_softlimit;
991

992 993
		do_softlimit = mem_cgroup_event_ratelimit(memcg,
						MEM_CGROUP_TARGET_SOFTLIMIT);
994
		mem_cgroup_threshold(memcg);
995 996
		if (unlikely(do_softlimit))
			mem_cgroup_update_tree(memcg, page);
997
	}
998 999
}

1000
struct mem_cgroup *mem_cgroup_from_task(struct task_struct *p)
1001
{
1002 1003 1004 1005 1006 1007 1008 1009
	/*
	 * 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;

1010
	return mem_cgroup_from_css(task_css(p, memory_cgrp_id));
1011
}
M
Michal Hocko 已提交
1012
EXPORT_SYMBOL(mem_cgroup_from_task);
1013

1014 1015 1016 1017 1018 1019 1020 1021 1022
/**
 * get_mem_cgroup_from_mm: Obtain a reference on given mm_struct's memcg.
 * @mm: mm from which memcg should be extracted. It can be NULL.
 *
 * Obtain a reference on mm->memcg and returns it if successful. Otherwise
 * root_mem_cgroup is returned. However if mem_cgroup is disabled, NULL is
 * returned.
 */
struct mem_cgroup *get_mem_cgroup_from_mm(struct mm_struct *mm)
1023
{
1024 1025 1026 1027
	struct mem_cgroup *memcg;

	if (mem_cgroup_disabled())
		return NULL;
1028

1029 1030
	rcu_read_lock();
	do {
1031 1032 1033 1034 1035 1036
		/*
		 * 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))
1037
			memcg = root_mem_cgroup;
1038 1039 1040 1041 1042
		else {
			memcg = mem_cgroup_from_task(rcu_dereference(mm->owner));
			if (unlikely(!memcg))
				memcg = root_mem_cgroup;
		}
1043
	} while (!css_tryget(&memcg->css));
1044
	rcu_read_unlock();
1045
	return memcg;
1046
}
1047 1048
EXPORT_SYMBOL(get_mem_cgroup_from_mm);

1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063
/**
 * get_mem_cgroup_from_page: Obtain a reference on given page's memcg.
 * @page: page from which memcg should be extracted.
 *
 * Obtain a reference on page->memcg and returns it if successful. Otherwise
 * root_mem_cgroup is returned.
 */
struct mem_cgroup *get_mem_cgroup_from_page(struct page *page)
{
	struct mem_cgroup *memcg = page->mem_cgroup;

	if (mem_cgroup_disabled())
		return NULL;

	rcu_read_lock();
S
Shakeel Butt 已提交
1064 1065
	/* Page should not get uncharged and freed memcg under us. */
	if (!memcg || WARN_ON_ONCE(!css_tryget(&memcg->css)))
1066 1067 1068 1069 1070 1071
		memcg = root_mem_cgroup;
	rcu_read_unlock();
	return memcg;
}
EXPORT_SYMBOL(get_mem_cgroup_from_page);

1072
static __always_inline struct mem_cgroup *active_memcg(void)
1073
{
1074 1075 1076 1077 1078
	if (in_interrupt())
		return this_cpu_read(int_active_memcg);
	else
		return current->active_memcg;
}
1079

1080 1081 1082
static __always_inline struct mem_cgroup *get_active_memcg(void)
{
	struct mem_cgroup *memcg;
1083

1084 1085
	rcu_read_lock();
	memcg = active_memcg();
1086 1087 1088
	/* remote memcg must hold a ref. */
	if (memcg && WARN_ON_ONCE(!css_tryget(&memcg->css)))
		memcg = root_mem_cgroup;
1089 1090 1091 1092 1093
	rcu_read_unlock();

	return memcg;
}

1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106
static __always_inline bool memcg_kmem_bypass(void)
{
	/* Allow remote memcg charging from any context. */
	if (unlikely(active_memcg()))
		return false;

	/* Memcg to charge can't be determined. */
	if (in_interrupt() || !current->mm || (current->flags & PF_KTHREAD))
		return true;

	return false;
}

1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117
/**
 * If active memcg is set, do not fallback to current->mm->memcg.
 */
static __always_inline struct mem_cgroup *get_mem_cgroup_from_current(void)
{
	if (memcg_kmem_bypass())
		return NULL;

	if (unlikely(active_memcg()))
		return get_active_memcg();

1118 1119
	return get_mem_cgroup_from_mm(current->mm);
}
1120

1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133
/**
 * 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.
 *
1134 1135 1136
 * Reclaimers can specify a node in @reclaim to divide up the memcgs
 * in the hierarchy among all concurrent reclaimers operating on the
 * same node.
1137
 */
1138
struct mem_cgroup *mem_cgroup_iter(struct mem_cgroup *root,
1139
				   struct mem_cgroup *prev,
1140
				   struct mem_cgroup_reclaim_cookie *reclaim)
K
KAMEZAWA Hiroyuki 已提交
1141
{
1142
	struct mem_cgroup_reclaim_iter *iter;
1143
	struct cgroup_subsys_state *css = NULL;
1144
	struct mem_cgroup *memcg = NULL;
1145
	struct mem_cgroup *pos = NULL;
1146

1147 1148
	if (mem_cgroup_disabled())
		return NULL;
1149

1150 1151
	if (!root)
		root = root_mem_cgroup;
K
KAMEZAWA Hiroyuki 已提交
1152

1153
	if (prev && !reclaim)
1154
		pos = prev;
K
KAMEZAWA Hiroyuki 已提交
1155

1156 1157
	if (!root->use_hierarchy && root != root_mem_cgroup) {
		if (prev)
1158
			goto out;
1159
		return root;
1160
	}
K
KAMEZAWA Hiroyuki 已提交
1161

1162
	rcu_read_lock();
M
Michal Hocko 已提交
1163

1164
	if (reclaim) {
1165
		struct mem_cgroup_per_node *mz;
1166

1167
		mz = mem_cgroup_nodeinfo(root, reclaim->pgdat->node_id);
1168
		iter = &mz->iter;
1169 1170 1171 1172

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

1173
		while (1) {
1174
			pos = READ_ONCE(iter->position);
1175 1176
			if (!pos || css_tryget(&pos->css))
				break;
1177
			/*
1178 1179 1180 1181 1182 1183
			 * 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.
1184
			 */
1185 1186
			(void)cmpxchg(&iter->position, pos, NULL);
		}
1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203
	}

	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;
1204
		}
K
KAMEZAWA Hiroyuki 已提交
1205

1206 1207 1208 1209 1210 1211
		/*
		 * 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
1212

1213 1214
		if (css == &root->css)
			break;
K
KAMEZAWA Hiroyuki 已提交
1215

1216 1217
		if (css_tryget(css))
			break;
1218

1219
		memcg = NULL;
1220
	}
1221 1222 1223

	if (reclaim) {
		/*
1224 1225 1226
		 * 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.
1227
		 */
1228 1229
		(void)cmpxchg(&iter->position, pos, memcg);

1230 1231 1232 1233 1234 1235 1236
		if (pos)
			css_put(&pos->css);

		if (!memcg)
			iter->generation++;
		else if (!prev)
			reclaim->generation = iter->generation;
1237
	}
1238

1239 1240
out_unlock:
	rcu_read_unlock();
1241
out:
1242 1243 1244
	if (prev && prev != root)
		css_put(&prev->css);

1245
	return memcg;
K
KAMEZAWA Hiroyuki 已提交
1246
}
K
KAMEZAWA Hiroyuki 已提交
1247

1248 1249 1250 1251 1252 1253 1254
/**
 * 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)
1255 1256 1257 1258 1259 1260
{
	if (!root)
		root = root_mem_cgroup;
	if (prev && prev != root)
		css_put(&prev->css);
}
K
KAMEZAWA Hiroyuki 已提交
1261

1262 1263
static void __invalidate_reclaim_iterators(struct mem_cgroup *from,
					struct mem_cgroup *dead_memcg)
1264 1265
{
	struct mem_cgroup_reclaim_iter *iter;
1266 1267
	struct mem_cgroup_per_node *mz;
	int nid;
1268

1269 1270
	for_each_node(nid) {
		mz = mem_cgroup_nodeinfo(from, nid);
1271 1272
		iter = &mz->iter;
		cmpxchg(&iter->position, dead_memcg, NULL);
1273 1274 1275
	}
}

1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296
static void invalidate_reclaim_iterators(struct mem_cgroup *dead_memcg)
{
	struct mem_cgroup *memcg = dead_memcg;
	struct mem_cgroup *last;

	do {
		__invalidate_reclaim_iterators(memcg, dead_memcg);
		last = memcg;
	} while ((memcg = parent_mem_cgroup(memcg)));

	/*
	 * When cgruop1 non-hierarchy mode is used,
	 * parent_mem_cgroup() does not walk all the way up to the
	 * cgroup root (root_mem_cgroup). So we have to handle
	 * dead_memcg from cgroup root separately.
	 */
	if (last != root_mem_cgroup)
		__invalidate_reclaim_iterators(root_mem_cgroup,
						dead_memcg);
}

1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321
/**
 * 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;

1322
		css_task_iter_start(&iter->css, CSS_TASK_ITER_PROCS, &it);
1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333
		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;
}

1334
/**
1335
 * mem_cgroup_page_lruvec - return lruvec for isolating/putting an LRU page
1336
 * @page: the page
1337
 * @pgdat: pgdat of the page
1338
 *
1339 1340
 * This function relies on page->mem_cgroup being stable - see the
 * access rules in commit_charge().
1341
 */
M
Mel Gorman 已提交
1342
struct lruvec *mem_cgroup_page_lruvec(struct page *page, struct pglist_data *pgdat)
K
KAMEZAWA Hiroyuki 已提交
1343
{
1344
	struct mem_cgroup_per_node *mz;
1345
	struct mem_cgroup *memcg;
1346
	struct lruvec *lruvec;
1347

1348
	if (mem_cgroup_disabled()) {
1349
		lruvec = &pgdat->__lruvec;
1350 1351
		goto out;
	}
1352

1353
	memcg = page->mem_cgroup;
1354
	/*
1355
	 * Swapcache readahead pages are added to the LRU - and
1356
	 * possibly migrated - before they are charged.
1357
	 */
1358 1359
	if (!memcg)
		memcg = root_mem_cgroup;
1360

1361
	mz = mem_cgroup_page_nodeinfo(memcg, page);
1362 1363 1364 1365 1366 1367 1368
	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
1369 1370
	if (unlikely(lruvec->pgdat != pgdat))
		lruvec->pgdat = pgdat;
1371
	return lruvec;
K
KAMEZAWA Hiroyuki 已提交
1372
}
1373

1374
/**
1375 1376 1377
 * 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
1378
 * @zid: zone id of the accounted pages
1379
 * @nr_pages: positive when adding or negative when removing
1380
 *
1381 1382 1383
 * 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).
1384
 */
1385
void mem_cgroup_update_lru_size(struct lruvec *lruvec, enum lru_list lru,
1386
				int zid, int nr_pages)
1387
{
1388
	struct mem_cgroup_per_node *mz;
1389
	unsigned long *lru_size;
1390
	long size;
1391 1392 1393 1394

	if (mem_cgroup_disabled())
		return;

1395
	mz = container_of(lruvec, struct mem_cgroup_per_node, lruvec);
1396
	lru_size = &mz->lru_zone_size[zid][lru];
1397 1398 1399 1400 1401

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

	size = *lru_size;
1402 1403 1404
	if (WARN_ONCE(size < 0,
		"%s(%p, %d, %d): lru_size %ld\n",
		__func__, lruvec, lru, nr_pages, size)) {
1405 1406 1407 1408 1409 1410
		VM_BUG_ON(1);
		*lru_size = 0;
	}

	if (nr_pages > 0)
		*lru_size += nr_pages;
K
KAMEZAWA Hiroyuki 已提交
1411
}
1412

1413
/**
1414
 * mem_cgroup_margin - calculate chargeable space of a memory cgroup
W
Wanpeng Li 已提交
1415
 * @memcg: the memory cgroup
1416
 *
1417
 * Returns the maximum amount of memory @mem can be charged with, in
1418
 * pages.
1419
 */
1420
static unsigned long mem_cgroup_margin(struct mem_cgroup *memcg)
1421
{
1422 1423 1424
	unsigned long margin = 0;
	unsigned long count;
	unsigned long limit;
1425

1426
	count = page_counter_read(&memcg->memory);
1427
	limit = READ_ONCE(memcg->memory.max);
1428 1429 1430
	if (count < limit)
		margin = limit - count;

1431
	if (do_memsw_account()) {
1432
		count = page_counter_read(&memcg->memsw);
1433
		limit = READ_ONCE(memcg->memsw.max);
1434
		if (count < limit)
1435
			margin = min(margin, limit - count);
1436 1437
		else
			margin = 0;
1438 1439 1440
	}

	return margin;
1441 1442
}

1443
/*
Q
Qiang Huang 已提交
1444
 * A routine for checking "mem" is under move_account() or not.
1445
 *
Q
Qiang Huang 已提交
1446 1447 1448
 * 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".
1449
 */
1450
static bool mem_cgroup_under_move(struct mem_cgroup *memcg)
1451
{
1452 1453
	struct mem_cgroup *from;
	struct mem_cgroup *to;
1454
	bool ret = false;
1455 1456 1457 1458 1459 1460 1461 1462 1463
	/*
	 * 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;
1464

1465 1466
	ret = mem_cgroup_is_descendant(from, memcg) ||
		mem_cgroup_is_descendant(to, memcg);
1467 1468
unlock:
	spin_unlock(&mc.lock);
1469 1470 1471
	return ret;
}

1472
static bool mem_cgroup_wait_acct_move(struct mem_cgroup *memcg)
1473 1474
{
	if (mc.moving_task && current != mc.moving_task) {
1475
		if (mem_cgroup_under_move(memcg)) {
1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487
			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;
}

1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551
struct memory_stat {
	const char *name;
	unsigned int ratio;
	unsigned int idx;
};

static struct memory_stat memory_stats[] = {
	{ "anon", PAGE_SIZE, NR_ANON_MAPPED },
	{ "file", PAGE_SIZE, NR_FILE_PAGES },
	{ "kernel_stack", 1024, NR_KERNEL_STACK_KB },
	{ "percpu", 1, MEMCG_PERCPU_B },
	{ "sock", PAGE_SIZE, MEMCG_SOCK },
	{ "shmem", PAGE_SIZE, NR_SHMEM },
	{ "file_mapped", PAGE_SIZE, NR_FILE_MAPPED },
	{ "file_dirty", PAGE_SIZE, NR_FILE_DIRTY },
	{ "file_writeback", PAGE_SIZE, NR_WRITEBACK },
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
	/*
	 * The ratio will be initialized in memory_stats_init(). Because
	 * on some architectures, the macro of HPAGE_PMD_SIZE is not
	 * constant(e.g. powerpc).
	 */
	{ "anon_thp", 0, NR_ANON_THPS },
#endif
	{ "inactive_anon", PAGE_SIZE, NR_INACTIVE_ANON },
	{ "active_anon", PAGE_SIZE, NR_ACTIVE_ANON },
	{ "inactive_file", PAGE_SIZE, NR_INACTIVE_FILE },
	{ "active_file", PAGE_SIZE, NR_ACTIVE_FILE },
	{ "unevictable", PAGE_SIZE, NR_UNEVICTABLE },

	/*
	 * Note: The slab_reclaimable and slab_unreclaimable must be
	 * together and slab_reclaimable must be in front.
	 */
	{ "slab_reclaimable", 1, NR_SLAB_RECLAIMABLE_B },
	{ "slab_unreclaimable", 1, NR_SLAB_UNRECLAIMABLE_B },

	/* The memory events */
	{ "workingset_refault_anon", 1, WORKINGSET_REFAULT_ANON },
	{ "workingset_refault_file", 1, WORKINGSET_REFAULT_FILE },
	{ "workingset_activate_anon", 1, WORKINGSET_ACTIVATE_ANON },
	{ "workingset_activate_file", 1, WORKINGSET_ACTIVATE_FILE },
	{ "workingset_restore_anon", 1, WORKINGSET_RESTORE_ANON },
	{ "workingset_restore_file", 1, WORKINGSET_RESTORE_FILE },
	{ "workingset_nodereclaim", 1, WORKINGSET_NODERECLAIM },
};

static int __init memory_stats_init(void)
{
	int i;

	for (i = 0; i < ARRAY_SIZE(memory_stats); i++) {
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
		if (memory_stats[i].idx == NR_ANON_THPS)
			memory_stats[i].ratio = HPAGE_PMD_SIZE;
#endif
		VM_BUG_ON(!memory_stats[i].ratio);
		VM_BUG_ON(memory_stats[i].idx >= MEMCG_NR_STAT);
	}

	return 0;
}
pure_initcall(memory_stats_init);

1552 1553 1554 1555
static char *memory_stat_format(struct mem_cgroup *memcg)
{
	struct seq_buf s;
	int i;
1556

1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571
	seq_buf_init(&s, kmalloc(PAGE_SIZE, GFP_KERNEL), PAGE_SIZE);
	if (!s.buffer)
		return NULL;

	/*
	 * 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:
	 */

1572 1573
	for (i = 0; i < ARRAY_SIZE(memory_stats); i++) {
		u64 size;
1574

1575 1576 1577
		size = memcg_page_state(memcg, memory_stats[i].idx);
		size *= memory_stats[i].ratio;
		seq_buf_printf(&s, "%s %llu\n", memory_stats[i].name, size);
1578

1579 1580 1581 1582 1583 1584
		if (unlikely(memory_stats[i].idx == NR_SLAB_UNRECLAIMABLE_B)) {
			size = memcg_page_state(memcg, NR_SLAB_RECLAIMABLE_B) +
			       memcg_page_state(memcg, NR_SLAB_UNRECLAIMABLE_B);
			seq_buf_printf(&s, "slab %llu\n", size);
		}
	}
1585 1586 1587

	/* Accumulated memory events */

1588 1589 1590 1591 1592 1593
	seq_buf_printf(&s, "%s %lu\n", vm_event_name(PGFAULT),
		       memcg_events(memcg, PGFAULT));
	seq_buf_printf(&s, "%s %lu\n", vm_event_name(PGMAJFAULT),
		       memcg_events(memcg, PGMAJFAULT));
	seq_buf_printf(&s, "%s %lu\n",  vm_event_name(PGREFILL),
		       memcg_events(memcg, PGREFILL));
1594 1595 1596 1597 1598 1599
	seq_buf_printf(&s, "pgscan %lu\n",
		       memcg_events(memcg, PGSCAN_KSWAPD) +
		       memcg_events(memcg, PGSCAN_DIRECT));
	seq_buf_printf(&s, "pgsteal %lu\n",
		       memcg_events(memcg, PGSTEAL_KSWAPD) +
		       memcg_events(memcg, PGSTEAL_DIRECT));
1600 1601 1602 1603 1604 1605 1606 1607
	seq_buf_printf(&s, "%s %lu\n", vm_event_name(PGACTIVATE),
		       memcg_events(memcg, PGACTIVATE));
	seq_buf_printf(&s, "%s %lu\n", vm_event_name(PGDEACTIVATE),
		       memcg_events(memcg, PGDEACTIVATE));
	seq_buf_printf(&s, "%s %lu\n", vm_event_name(PGLAZYFREE),
		       memcg_events(memcg, PGLAZYFREE));
	seq_buf_printf(&s, "%s %lu\n", vm_event_name(PGLAZYFREED),
		       memcg_events(memcg, PGLAZYFREED));
1608 1609

#ifdef CONFIG_TRANSPARENT_HUGEPAGE
1610
	seq_buf_printf(&s, "%s %lu\n", vm_event_name(THP_FAULT_ALLOC),
1611
		       memcg_events(memcg, THP_FAULT_ALLOC));
1612
	seq_buf_printf(&s, "%s %lu\n", vm_event_name(THP_COLLAPSE_ALLOC),
1613 1614 1615 1616 1617 1618 1619 1620
		       memcg_events(memcg, THP_COLLAPSE_ALLOC));
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */

	/* The above should easily fit into one page */
	WARN_ON_ONCE(seq_buf_has_overflowed(&s));

	return s.buffer;
}
1621

1622
#define K(x) ((x) << (PAGE_SHIFT-10))
1623
/**
1624 1625
 * mem_cgroup_print_oom_context: Print OOM information relevant to
 * memory controller.
1626 1627 1628 1629 1630 1631
 * @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
 */
1632
void mem_cgroup_print_oom_context(struct mem_cgroup *memcg, struct task_struct *p)
1633 1634 1635
{
	rcu_read_lock();

1636 1637 1638 1639 1640
	if (memcg) {
		pr_cont(",oom_memcg=");
		pr_cont_cgroup_path(memcg->css.cgroup);
	} else
		pr_cont(",global_oom");
1641
	if (p) {
1642
		pr_cont(",task_memcg=");
1643 1644
		pr_cont_cgroup_path(task_cgroup(p, memory_cgrp_id));
	}
1645
	rcu_read_unlock();
1646 1647 1648 1649 1650 1651 1652 1653 1654
}

/**
 * mem_cgroup_print_oom_meminfo: Print OOM memory information relevant to
 * memory controller.
 * @memcg: The memory cgroup that went over limit
 */
void mem_cgroup_print_oom_meminfo(struct mem_cgroup *memcg)
{
1655
	char *buf;
1656

1657 1658
	pr_info("memory: usage %llukB, limit %llukB, failcnt %lu\n",
		K((u64)page_counter_read(&memcg->memory)),
1659
		K((u64)READ_ONCE(memcg->memory.max)), memcg->memory.failcnt);
1660 1661 1662
	if (cgroup_subsys_on_dfl(memory_cgrp_subsys))
		pr_info("swap: usage %llukB, limit %llukB, failcnt %lu\n",
			K((u64)page_counter_read(&memcg->swap)),
1663
			K((u64)READ_ONCE(memcg->swap.max)), memcg->swap.failcnt);
1664 1665 1666 1667 1668 1669 1670
	else {
		pr_info("memory+swap: usage %llukB, limit %llukB, failcnt %lu\n",
			K((u64)page_counter_read(&memcg->memsw)),
			K((u64)memcg->memsw.max), memcg->memsw.failcnt);
		pr_info("kmem: usage %llukB, limit %llukB, failcnt %lu\n",
			K((u64)page_counter_read(&memcg->kmem)),
			K((u64)memcg->kmem.max), memcg->kmem.failcnt);
1671
	}
1672 1673 1674 1675 1676 1677 1678 1679 1680

	pr_info("Memory cgroup stats for ");
	pr_cont_cgroup_path(memcg->css.cgroup);
	pr_cont(":");
	buf = memory_stat_format(memcg);
	if (!buf)
		return;
	pr_info("%s", buf);
	kfree(buf);
1681 1682
}

D
David Rientjes 已提交
1683 1684 1685
/*
 * Return the memory (and swap, if configured) limit for a memcg.
 */
1686
unsigned long mem_cgroup_get_max(struct mem_cgroup *memcg)
D
David Rientjes 已提交
1687
{
1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700
	unsigned long max = READ_ONCE(memcg->memory.max);

	if (cgroup_subsys_on_dfl(memory_cgrp_subsys)) {
		if (mem_cgroup_swappiness(memcg))
			max += min(READ_ONCE(memcg->swap.max),
				   (unsigned long)total_swap_pages);
	} else { /* v1 */
		if (mem_cgroup_swappiness(memcg)) {
			/* Calculate swap excess capacity from memsw limit */
			unsigned long swap = READ_ONCE(memcg->memsw.max) - max;

			max += min(swap, (unsigned long)total_swap_pages);
		}
1701
	}
1702
	return max;
D
David Rientjes 已提交
1703 1704
}

1705 1706 1707 1708 1709
unsigned long mem_cgroup_size(struct mem_cgroup *memcg)
{
	return page_counter_read(&memcg->memory);
}

1710
static bool mem_cgroup_out_of_memory(struct mem_cgroup *memcg, gfp_t gfp_mask,
1711
				     int order)
1712
{
1713 1714 1715
	struct oom_control oc = {
		.zonelist = NULL,
		.nodemask = NULL,
1716
		.memcg = memcg,
1717 1718 1719
		.gfp_mask = gfp_mask,
		.order = order,
	};
1720
	bool ret = true;
1721

1722 1723
	if (mutex_lock_killable(&oom_lock))
		return true;
1724 1725 1726 1727

	if (mem_cgroup_margin(memcg) >= (1 << order))
		goto unlock;

1728 1729 1730 1731 1732
	/*
	 * A few threads which were not waiting at mutex_lock_killable() can
	 * fail to bail out. Therefore, check again after holding oom_lock.
	 */
	ret = should_force_charge() || out_of_memory(&oc);
1733 1734

unlock:
1735
	mutex_unlock(&oom_lock);
1736
	return ret;
1737 1738
}

1739
static int mem_cgroup_soft_reclaim(struct mem_cgroup *root_memcg,
1740
				   pg_data_t *pgdat,
1741 1742 1743 1744 1745 1746 1747 1748 1749
				   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 = {
1750
		.pgdat = pgdat,
1751 1752
	};

1753
	excess = soft_limit_excess(root_memcg);
1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778

	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;
		}
1779
		total += mem_cgroup_shrink_node(victim, gfp_mask, false,
1780
					pgdat, &nr_scanned);
1781
		*total_scanned += nr_scanned;
1782
		if (!soft_limit_excess(root_memcg))
1783
			break;
1784
	}
1785 1786
	mem_cgroup_iter_break(root_memcg, victim);
	return total;
1787 1788
}

1789 1790 1791 1792 1793 1794
#ifdef CONFIG_LOCKDEP
static struct lockdep_map memcg_oom_lock_dep_map = {
	.name = "memcg_oom_lock",
};
#endif

1795 1796
static DEFINE_SPINLOCK(memcg_oom_lock);

K
KAMEZAWA Hiroyuki 已提交
1797 1798 1799 1800
/*
 * Check OOM-Killer is already running under our hierarchy.
 * If someone is running, return false.
 */
1801
static bool mem_cgroup_oom_trylock(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
1802
{
1803
	struct mem_cgroup *iter, *failed = NULL;
1804

1805 1806
	spin_lock(&memcg_oom_lock);

1807
	for_each_mem_cgroup_tree(iter, memcg) {
1808
		if (iter->oom_lock) {
1809 1810 1811 1812 1813
			/*
			 * this subtree of our hierarchy is already locked
			 * so we cannot give a lock.
			 */
			failed = iter;
1814 1815
			mem_cgroup_iter_break(memcg, iter);
			break;
1816 1817
		} else
			iter->oom_lock = true;
K
KAMEZAWA Hiroyuki 已提交
1818
	}
K
KAMEZAWA Hiroyuki 已提交
1819

1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830
	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;
1831
		}
1832 1833
	} else
		mutex_acquire(&memcg_oom_lock_dep_map, 0, 1, _RET_IP_);
1834 1835 1836 1837

	spin_unlock(&memcg_oom_lock);

	return !failed;
1838
}
1839

1840
static void mem_cgroup_oom_unlock(struct mem_cgroup *memcg)
1841
{
K
KAMEZAWA Hiroyuki 已提交
1842 1843
	struct mem_cgroup *iter;

1844
	spin_lock(&memcg_oom_lock);
1845
	mutex_release(&memcg_oom_lock_dep_map, _RET_IP_);
1846
	for_each_mem_cgroup_tree(iter, memcg)
1847
		iter->oom_lock = false;
1848
	spin_unlock(&memcg_oom_lock);
1849 1850
}

1851
static void mem_cgroup_mark_under_oom(struct mem_cgroup *memcg)
1852 1853 1854
{
	struct mem_cgroup *iter;

1855
	spin_lock(&memcg_oom_lock);
1856
	for_each_mem_cgroup_tree(iter, memcg)
1857 1858
		iter->under_oom++;
	spin_unlock(&memcg_oom_lock);
1859 1860
}

1861
static void mem_cgroup_unmark_under_oom(struct mem_cgroup *memcg)
1862 1863 1864
{
	struct mem_cgroup *iter;

K
KAMEZAWA Hiroyuki 已提交
1865
	/*
1866 1867
	 * Be careful about under_oom underflows becase a child memcg
	 * could have been added after mem_cgroup_mark_under_oom.
K
KAMEZAWA Hiroyuki 已提交
1868
	 */
1869
	spin_lock(&memcg_oom_lock);
1870
	for_each_mem_cgroup_tree(iter, memcg)
1871 1872 1873
		if (iter->under_oom > 0)
			iter->under_oom--;
	spin_unlock(&memcg_oom_lock);
1874 1875
}

K
KAMEZAWA Hiroyuki 已提交
1876 1877
static DECLARE_WAIT_QUEUE_HEAD(memcg_oom_waitq);

K
KAMEZAWA Hiroyuki 已提交
1878
struct oom_wait_info {
1879
	struct mem_cgroup *memcg;
1880
	wait_queue_entry_t	wait;
K
KAMEZAWA Hiroyuki 已提交
1881 1882
};

1883
static int memcg_oom_wake_function(wait_queue_entry_t *wait,
K
KAMEZAWA Hiroyuki 已提交
1884 1885
	unsigned mode, int sync, void *arg)
{
1886 1887
	struct mem_cgroup *wake_memcg = (struct mem_cgroup *)arg;
	struct mem_cgroup *oom_wait_memcg;
K
KAMEZAWA Hiroyuki 已提交
1888 1889 1890
	struct oom_wait_info *oom_wait_info;

	oom_wait_info = container_of(wait, struct oom_wait_info, wait);
1891
	oom_wait_memcg = oom_wait_info->memcg;
K
KAMEZAWA Hiroyuki 已提交
1892

1893 1894
	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
1895 1896 1897 1898
		return 0;
	return autoremove_wake_function(wait, mode, sync, arg);
}

1899
static void memcg_oom_recover(struct mem_cgroup *memcg)
1900
{
1901 1902 1903 1904 1905 1906 1907 1908 1909
	/*
	 * 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)
1910
		__wake_up(&memcg_oom_waitq, TASK_NORMAL, 0, memcg);
1911 1912
}

1913 1914 1915 1916 1917 1918 1919 1920
enum oom_status {
	OOM_SUCCESS,
	OOM_FAILED,
	OOM_ASYNC,
	OOM_SKIPPED
};

static enum oom_status mem_cgroup_oom(struct mem_cgroup *memcg, gfp_t mask, int order)
1921
{
1922 1923 1924
	enum oom_status ret;
	bool locked;

1925 1926 1927
	if (order > PAGE_ALLOC_COSTLY_ORDER)
		return OOM_SKIPPED;

1928 1929
	memcg_memory_event(memcg, MEMCG_OOM);

K
KAMEZAWA Hiroyuki 已提交
1930
	/*
1931 1932 1933 1934
	 * 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.
	 *
1935 1936 1937 1938
	 * cgroup1 allows disabling the OOM killer and waiting for outside
	 * handling until the charge can succeed; remember the context and put
	 * the task to sleep at the end of the page fault when all locks are
	 * released.
1939
	 *
1940 1941 1942 1943 1944 1945 1946
	 * On the other hand, in-kernel OOM killer allows for an async victim
	 * memory reclaim (oom_reaper) and that means that we are not solely
	 * relying on the oom victim to make a forward progress and we can
	 * invoke the oom killer here.
	 *
	 * Please note that mem_cgroup_out_of_memory might fail to find a
	 * victim and then we have to bail out from the charge path.
K
KAMEZAWA Hiroyuki 已提交
1947
	 */
1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958
	if (memcg->oom_kill_disable) {
		if (!current->in_user_fault)
			return OOM_SKIPPED;
		css_get(&memcg->css);
		current->memcg_in_oom = memcg;
		current->memcg_oom_gfp_mask = mask;
		current->memcg_oom_order = order;

		return OOM_ASYNC;
	}

1959 1960 1961 1962 1963 1964 1965 1966
	mem_cgroup_mark_under_oom(memcg);

	locked = mem_cgroup_oom_trylock(memcg);

	if (locked)
		mem_cgroup_oom_notify(memcg);

	mem_cgroup_unmark_under_oom(memcg);
1967
	if (mem_cgroup_out_of_memory(memcg, mask, order))
1968 1969 1970 1971 1972 1973
		ret = OOM_SUCCESS;
	else
		ret = OOM_FAILED;

	if (locked)
		mem_cgroup_oom_unlock(memcg);
1974

1975
	return ret;
1976 1977 1978 1979
}

/**
 * mem_cgroup_oom_synchronize - complete memcg OOM handling
1980
 * @handle: actually kill/wait or just clean up the OOM state
1981
 *
1982 1983
 * This has to be called at the end of a page fault if the memcg OOM
 * handler was enabled.
1984
 *
1985
 * Memcg supports userspace OOM handling where failed allocations must
1986 1987 1988 1989
 * 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
1990
 * the end of the page fault to complete the OOM handling.
1991 1992
 *
 * Returns %true if an ongoing memcg OOM situation was detected and
1993
 * completed, %false otherwise.
1994
 */
1995
bool mem_cgroup_oom_synchronize(bool handle)
1996
{
T
Tejun Heo 已提交
1997
	struct mem_cgroup *memcg = current->memcg_in_oom;
1998
	struct oom_wait_info owait;
1999
	bool locked;
2000 2001 2002

	/* OOM is global, do not handle */
	if (!memcg)
2003
		return false;
2004

2005
	if (!handle)
2006
		goto cleanup;
2007 2008 2009 2010 2011

	owait.memcg = memcg;
	owait.wait.flags = 0;
	owait.wait.func = memcg_oom_wake_function;
	owait.wait.private = current;
2012
	INIT_LIST_HEAD(&owait.wait.entry);
K
KAMEZAWA Hiroyuki 已提交
2013

2014
	prepare_to_wait(&memcg_oom_waitq, &owait.wait, TASK_KILLABLE);
2015 2016 2017 2018 2019 2020 2021 2022 2023 2024
	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
2025 2026
		mem_cgroup_out_of_memory(memcg, current->memcg_oom_gfp_mask,
					 current->memcg_oom_order);
2027
	} else {
2028
		schedule();
2029 2030 2031 2032 2033
		mem_cgroup_unmark_under_oom(memcg);
		finish_wait(&memcg_oom_waitq, &owait.wait);
	}

	if (locked) {
2034 2035 2036 2037 2038 2039 2040 2041
		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);
	}
2042
cleanup:
T
Tejun Heo 已提交
2043
	current->memcg_in_oom = NULL;
2044
	css_put(&memcg->css);
K
KAMEZAWA Hiroyuki 已提交
2045
	return true;
2046 2047
}

2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075
/**
 * mem_cgroup_get_oom_group - get a memory cgroup to clean up after OOM
 * @victim: task to be killed by the OOM killer
 * @oom_domain: memcg in case of memcg OOM, NULL in case of system-wide OOM
 *
 * Returns a pointer to a memory cgroup, which has to be cleaned up
 * by killing all belonging OOM-killable tasks.
 *
 * Caller has to call mem_cgroup_put() on the returned non-NULL memcg.
 */
struct mem_cgroup *mem_cgroup_get_oom_group(struct task_struct *victim,
					    struct mem_cgroup *oom_domain)
{
	struct mem_cgroup *oom_group = NULL;
	struct mem_cgroup *memcg;

	if (!cgroup_subsys_on_dfl(memory_cgrp_subsys))
		return NULL;

	if (!oom_domain)
		oom_domain = root_mem_cgroup;

	rcu_read_lock();

	memcg = mem_cgroup_from_task(victim);
	if (memcg == root_mem_cgroup)
		goto out;

2076 2077 2078 2079 2080 2081 2082 2083
	/*
	 * If the victim task has been asynchronously moved to a different
	 * memory cgroup, we might end up killing tasks outside oom_domain.
	 * In this case it's better to ignore memory.group.oom.
	 */
	if (unlikely(!mem_cgroup_is_descendant(memcg, oom_domain)))
		goto out;

2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111
	/*
	 * Traverse the memory cgroup hierarchy from the victim task's
	 * cgroup up to the OOMing cgroup (or root) to find the
	 * highest-level memory cgroup with oom.group set.
	 */
	for (; memcg; memcg = parent_mem_cgroup(memcg)) {
		if (memcg->oom_group)
			oom_group = memcg;

		if (memcg == oom_domain)
			break;
	}

	if (oom_group)
		css_get(&oom_group->css);
out:
	rcu_read_unlock();

	return oom_group;
}

void mem_cgroup_print_oom_group(struct mem_cgroup *memcg)
{
	pr_info("Tasks in ");
	pr_cont_cgroup_path(memcg->css.cgroup);
	pr_cont(" are going to be killed due to memory.oom.group set\n");
}

2112
/**
2113 2114
 * lock_page_memcg - lock a page->mem_cgroup binding
 * @page: the page
2115
 *
2116
 * This function protects unlocked LRU pages from being moved to
2117 2118 2119 2120 2121
 * another cgroup.
 *
 * It ensures lifetime of the returned memcg. Caller is responsible
 * for the lifetime of the page; __unlock_page_memcg() is available
 * when @page might get freed inside the locked section.
2122
 */
2123
struct mem_cgroup *lock_page_memcg(struct page *page)
2124
{
2125
	struct page *head = compound_head(page); /* rmap on tail pages */
2126
	struct mem_cgroup *memcg;
2127
	unsigned long flags;
2128

2129 2130 2131 2132
	/*
	 * 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.
2133 2134 2135 2136 2137 2138 2139
	 *
	 * The RCU lock also protects the memcg from being freed when
	 * the page state that is going to change is the only thing
	 * preventing the page itself from being freed. E.g. writeback
	 * doesn't hold a page reference and relies on PG_writeback to
	 * keep off truncation, migration and so forth.
         */
2140 2141 2142
	rcu_read_lock();

	if (mem_cgroup_disabled())
2143
		return NULL;
2144
again:
2145
	memcg = head->mem_cgroup;
2146
	if (unlikely(!memcg))
2147
		return NULL;
2148

Q
Qiang Huang 已提交
2149
	if (atomic_read(&memcg->moving_account) <= 0)
2150
		return memcg;
2151

2152
	spin_lock_irqsave(&memcg->move_lock, flags);
2153
	if (memcg != head->mem_cgroup) {
2154
		spin_unlock_irqrestore(&memcg->move_lock, flags);
2155 2156
		goto again;
	}
2157 2158 2159 2160

	/*
	 * When charge migration first begins, we can have locked and
	 * unlocked page stat updates happening concurrently.  Track
2161
	 * the task who has the lock for unlock_page_memcg().
2162 2163 2164
	 */
	memcg->move_lock_task = current;
	memcg->move_lock_flags = flags;
2165

2166
	return memcg;
2167
}
2168
EXPORT_SYMBOL(lock_page_memcg);
2169

2170
/**
2171 2172 2173 2174
 * __unlock_page_memcg - unlock and unpin a memcg
 * @memcg: the memcg
 *
 * Unlock and unpin a memcg returned by lock_page_memcg().
2175
 */
2176
void __unlock_page_memcg(struct mem_cgroup *memcg)
2177
{
2178 2179 2180 2181 2182 2183 2184 2185
	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);
	}
2186

2187
	rcu_read_unlock();
2188
}
2189 2190 2191 2192 2193 2194 2195

/**
 * unlock_page_memcg - unlock a page->mem_cgroup binding
 * @page: the page
 */
void unlock_page_memcg(struct page *page)
{
2196 2197 2198
	struct page *head = compound_head(page);

	__unlock_page_memcg(head->mem_cgroup);
2199
}
2200
EXPORT_SYMBOL(unlock_page_memcg);
2201

2202 2203
struct memcg_stock_pcp {
	struct mem_cgroup *cached; /* this never be root cgroup */
2204
	unsigned int nr_pages;
R
Roman Gushchin 已提交
2205 2206 2207 2208 2209 2210

#ifdef CONFIG_MEMCG_KMEM
	struct obj_cgroup *cached_objcg;
	unsigned int nr_bytes;
#endif

2211
	struct work_struct work;
2212
	unsigned long flags;
2213
#define FLUSHING_CACHED_CHARGE	0
2214 2215
};
static DEFINE_PER_CPU(struct memcg_stock_pcp, memcg_stock);
2216
static DEFINE_MUTEX(percpu_charge_mutex);
2217

R
Roman Gushchin 已提交
2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233
#ifdef CONFIG_MEMCG_KMEM
static void drain_obj_stock(struct memcg_stock_pcp *stock);
static bool obj_stock_flush_required(struct memcg_stock_pcp *stock,
				     struct mem_cgroup *root_memcg);

#else
static inline void drain_obj_stock(struct memcg_stock_pcp *stock)
{
}
static bool obj_stock_flush_required(struct memcg_stock_pcp *stock,
				     struct mem_cgroup *root_memcg)
{
	return false;
}
#endif

2234 2235 2236 2237 2238 2239 2240 2241 2242 2243
/**
 * 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.
2244
 */
2245
static bool consume_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
2246 2247
{
	struct memcg_stock_pcp *stock;
2248
	unsigned long flags;
2249
	bool ret = false;
2250

2251
	if (nr_pages > MEMCG_CHARGE_BATCH)
2252
		return ret;
2253

2254 2255 2256
	local_irq_save(flags);

	stock = this_cpu_ptr(&memcg_stock);
2257
	if (memcg == stock->cached && stock->nr_pages >= nr_pages) {
2258
		stock->nr_pages -= nr_pages;
2259 2260
		ret = true;
	}
2261 2262 2263

	local_irq_restore(flags);

2264 2265 2266 2267
	return ret;
}

/*
2268
 * Returns stocks cached in percpu and reset cached information.
2269 2270 2271 2272 2273
 */
static void drain_stock(struct memcg_stock_pcp *stock)
{
	struct mem_cgroup *old = stock->cached;

2274 2275 2276
	if (!old)
		return;

2277
	if (stock->nr_pages) {
2278
		page_counter_uncharge(&old->memory, stock->nr_pages);
2279
		if (do_memsw_account())
2280
			page_counter_uncharge(&old->memsw, stock->nr_pages);
2281
		stock->nr_pages = 0;
2282
	}
2283 2284

	css_put(&old->css);
2285 2286 2287 2288 2289
	stock->cached = NULL;
}

static void drain_local_stock(struct work_struct *dummy)
{
2290 2291 2292
	struct memcg_stock_pcp *stock;
	unsigned long flags;

2293 2294 2295 2296
	/*
	 * The only protection from memory hotplug vs. drain_stock races is
	 * that we always operate on local CPU stock here with IRQ disabled
	 */
2297 2298 2299
	local_irq_save(flags);

	stock = this_cpu_ptr(&memcg_stock);
R
Roman Gushchin 已提交
2300
	drain_obj_stock(stock);
2301
	drain_stock(stock);
2302
	clear_bit(FLUSHING_CACHED_CHARGE, &stock->flags);
2303 2304

	local_irq_restore(flags);
2305 2306 2307
}

/*
2308
 * Cache charges(val) to local per_cpu area.
2309
 * This will be consumed by consume_stock() function, later.
2310
 */
2311
static void refill_stock(struct mem_cgroup *memcg, unsigned int nr_pages)
2312
{
2313 2314 2315 2316
	struct memcg_stock_pcp *stock;
	unsigned long flags;

	local_irq_save(flags);
2317

2318
	stock = this_cpu_ptr(&memcg_stock);
2319
	if (stock->cached != memcg) { /* reset if necessary */
2320
		drain_stock(stock);
2321
		css_get(&memcg->css);
2322
		stock->cached = memcg;
2323
	}
2324
	stock->nr_pages += nr_pages;
2325

2326
	if (stock->nr_pages > MEMCG_CHARGE_BATCH)
2327 2328
		drain_stock(stock);

2329
	local_irq_restore(flags);
2330 2331 2332
}

/*
2333
 * Drains all per-CPU charge caches for given root_memcg resp. subtree
2334
 * of the hierarchy under it.
2335
 */
2336
static void drain_all_stock(struct mem_cgroup *root_memcg)
2337
{
2338
	int cpu, curcpu;
2339

2340 2341 2342
	/* If someone's already draining, avoid adding running more workers. */
	if (!mutex_trylock(&percpu_charge_mutex))
		return;
2343 2344 2345 2346 2347 2348
	/*
	 * Notify other cpus that system-wide "drain" is running
	 * We do not care about races with the cpu hotplug because cpu down
	 * as well as workers from this path always operate on the local
	 * per-cpu data. CPU up doesn't touch memcg_stock at all.
	 */
2349
	curcpu = get_cpu();
2350 2351
	for_each_online_cpu(cpu) {
		struct memcg_stock_pcp *stock = &per_cpu(memcg_stock, cpu);
2352
		struct mem_cgroup *memcg;
2353
		bool flush = false;
2354

2355
		rcu_read_lock();
2356
		memcg = stock->cached;
2357 2358 2359
		if (memcg && stock->nr_pages &&
		    mem_cgroup_is_descendant(memcg, root_memcg))
			flush = true;
R
Roman Gushchin 已提交
2360 2361
		if (obj_stock_flush_required(stock, root_memcg))
			flush = true;
2362 2363 2364 2365
		rcu_read_unlock();

		if (flush &&
		    !test_and_set_bit(FLUSHING_CACHED_CHARGE, &stock->flags)) {
2366 2367 2368 2369 2370
			if (cpu == curcpu)
				drain_local_stock(&stock->work);
			else
				schedule_work_on(cpu, &stock->work);
		}
2371
	}
2372
	put_cpu();
2373
	mutex_unlock(&percpu_charge_mutex);
2374 2375
}

2376
static int memcg_hotplug_cpu_dead(unsigned int cpu)
2377 2378
{
	struct memcg_stock_pcp *stock;
2379
	struct mem_cgroup *memcg, *mi;
2380 2381 2382

	stock = &per_cpu(memcg_stock, cpu);
	drain_stock(stock);
2383 2384 2385 2386 2387 2388 2389 2390

	for_each_mem_cgroup(memcg) {
		int i;

		for (i = 0; i < MEMCG_NR_STAT; i++) {
			int nid;
			long x;

2391
			x = this_cpu_xchg(memcg->vmstats_percpu->stat[i], 0);
2392
			if (x)
2393 2394
				for (mi = memcg; mi; mi = parent_mem_cgroup(mi))
					atomic_long_add(x, &memcg->vmstats[i]);
2395 2396 2397 2398 2399 2400 2401 2402 2403

			if (i >= NR_VM_NODE_STAT_ITEMS)
				continue;

			for_each_node(nid) {
				struct mem_cgroup_per_node *pn;

				pn = mem_cgroup_nodeinfo(memcg, nid);
				x = this_cpu_xchg(pn->lruvec_stat_cpu->count[i], 0);
2404
				if (x)
2405 2406 2407
					do {
						atomic_long_add(x, &pn->lruvec_stat[i]);
					} while ((pn = parent_nodeinfo(pn, nid)));
2408 2409 2410
			}
		}

2411
		for (i = 0; i < NR_VM_EVENT_ITEMS; i++) {
2412 2413
			long x;

2414
			x = this_cpu_xchg(memcg->vmstats_percpu->events[i], 0);
2415
			if (x)
2416 2417
				for (mi = memcg; mi; mi = parent_mem_cgroup(mi))
					atomic_long_add(x, &memcg->vmevents[i]);
2418 2419 2420
		}
	}

2421
	return 0;
2422 2423
}

2424 2425 2426
static unsigned long reclaim_high(struct mem_cgroup *memcg,
				  unsigned int nr_pages,
				  gfp_t gfp_mask)
2427
{
2428 2429
	unsigned long nr_reclaimed = 0;

2430
	do {
2431 2432
		unsigned long pflags;

2433 2434
		if (page_counter_read(&memcg->memory) <=
		    READ_ONCE(memcg->memory.high))
2435
			continue;
2436

2437
		memcg_memory_event(memcg, MEMCG_HIGH);
2438 2439

		psi_memstall_enter(&pflags);
2440 2441
		nr_reclaimed += try_to_free_mem_cgroup_pages(memcg, nr_pages,
							     gfp_mask, true);
2442
		psi_memstall_leave(&pflags);
2443 2444
	} while ((memcg = parent_mem_cgroup(memcg)) &&
		 !mem_cgroup_is_root(memcg));
2445 2446

	return nr_reclaimed;
2447 2448 2449 2450 2451 2452 2453
}

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

	memcg = container_of(work, struct mem_cgroup, high_work);
2454
	reclaim_high(memcg, MEMCG_CHARGE_BATCH, GFP_KERNEL);
2455 2456
}

2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470
/*
 * Clamp the maximum sleep time per allocation batch to 2 seconds. This is
 * enough to still cause a significant slowdown in most cases, while still
 * allowing diagnostics and tracing to proceed without becoming stuck.
 */
#define MEMCG_MAX_HIGH_DELAY_JIFFIES (2UL*HZ)

/*
 * When calculating the delay, we use these either side of the exponentiation to
 * maintain precision and scale to a reasonable number of jiffies (see the table
 * below.
 *
 * - MEMCG_DELAY_PRECISION_SHIFT: Extra precision bits while translating the
 *   overage ratio to a delay.
2471
 * - MEMCG_DELAY_SCALING_SHIFT: The number of bits to scale down the
2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509
 *   proposed penalty in order to reduce to a reasonable number of jiffies, and
 *   to produce a reasonable delay curve.
 *
 * MEMCG_DELAY_SCALING_SHIFT just happens to be a number that produces a
 * reasonable delay curve compared to precision-adjusted overage, not
 * penalising heavily at first, but still making sure that growth beyond the
 * limit penalises misbehaviour cgroups by slowing them down exponentially. For
 * example, with a high of 100 megabytes:
 *
 *  +-------+------------------------+
 *  | usage | time to allocate in ms |
 *  +-------+------------------------+
 *  | 100M  |                      0 |
 *  | 101M  |                      6 |
 *  | 102M  |                     25 |
 *  | 103M  |                     57 |
 *  | 104M  |                    102 |
 *  | 105M  |                    159 |
 *  | 106M  |                    230 |
 *  | 107M  |                    313 |
 *  | 108M  |                    409 |
 *  | 109M  |                    518 |
 *  | 110M  |                    639 |
 *  | 111M  |                    774 |
 *  | 112M  |                    921 |
 *  | 113M  |                   1081 |
 *  | 114M  |                   1254 |
 *  | 115M  |                   1439 |
 *  | 116M  |                   1638 |
 *  | 117M  |                   1849 |
 *  | 118M  |                   2000 |
 *  | 119M  |                   2000 |
 *  | 120M  |                   2000 |
 *  +-------+------------------------+
 */
 #define MEMCG_DELAY_PRECISION_SHIFT 20
 #define MEMCG_DELAY_SCALING_SHIFT 14

2510
static u64 calculate_overage(unsigned long usage, unsigned long high)
2511
{
2512
	u64 overage;
2513

2514 2515
	if (usage <= high)
		return 0;
2516

2517 2518 2519 2520 2521
	/*
	 * Prevent division by 0 in overage calculation by acting as if
	 * it was a threshold of 1 page
	 */
	high = max(high, 1UL);
2522

2523 2524 2525 2526
	overage = usage - high;
	overage <<= MEMCG_DELAY_PRECISION_SHIFT;
	return div64_u64(overage, high);
}
2527

2528 2529 2530
static u64 mem_find_max_overage(struct mem_cgroup *memcg)
{
	u64 overage, max_overage = 0;
2531

2532 2533
	do {
		overage = calculate_overage(page_counter_read(&memcg->memory),
2534
					    READ_ONCE(memcg->memory.high));
2535
		max_overage = max(overage, max_overage);
2536 2537 2538
	} while ((memcg = parent_mem_cgroup(memcg)) &&
		 !mem_cgroup_is_root(memcg));

2539 2540 2541
	return max_overage;
}

2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557
static u64 swap_find_max_overage(struct mem_cgroup *memcg)
{
	u64 overage, max_overage = 0;

	do {
		overage = calculate_overage(page_counter_read(&memcg->swap),
					    READ_ONCE(memcg->swap.high));
		if (overage)
			memcg_memory_event(memcg, MEMCG_SWAP_HIGH);
		max_overage = max(overage, max_overage);
	} while ((memcg = parent_mem_cgroup(memcg)) &&
		 !mem_cgroup_is_root(memcg));

	return max_overage;
}

2558 2559 2560 2561 2562 2563 2564 2565 2566 2567
/*
 * Get the number of jiffies that we should penalise a mischievous cgroup which
 * is exceeding its memory.high by checking both it and its ancestors.
 */
static unsigned long calculate_high_delay(struct mem_cgroup *memcg,
					  unsigned int nr_pages,
					  u64 max_overage)
{
	unsigned long penalty_jiffies;

2568 2569
	if (!max_overage)
		return 0;
2570 2571 2572 2573 2574 2575 2576 2577 2578

	/*
	 * We use overage compared to memory.high to calculate the number of
	 * jiffies to sleep (penalty_jiffies). Ideally this value should be
	 * fairly lenient on small overages, and increasingly harsh when the
	 * memcg in question makes it clear that it has no intention of stopping
	 * its crazy behaviour, so we exponentially increase the delay based on
	 * overage amount.
	 */
2579 2580 2581
	penalty_jiffies = max_overage * max_overage * HZ;
	penalty_jiffies >>= MEMCG_DELAY_PRECISION_SHIFT;
	penalty_jiffies >>= MEMCG_DELAY_SCALING_SHIFT;
2582 2583 2584 2585 2586 2587 2588 2589 2590

	/*
	 * Factor in the task's own contribution to the overage, such that four
	 * N-sized allocations are throttled approximately the same as one
	 * 4N-sized allocation.
	 *
	 * MEMCG_CHARGE_BATCH pages is nominal, so work out how much smaller or
	 * larger the current charge patch is than that.
	 */
2591
	return penalty_jiffies * nr_pages / MEMCG_CHARGE_BATCH;
2592 2593 2594 2595 2596 2597 2598 2599 2600 2601
}

/*
 * 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 long penalty_jiffies;
	unsigned long pflags;
2602
	unsigned long nr_reclaimed;
2603
	unsigned int nr_pages = current->memcg_nr_pages_over_high;
2604
	int nr_retries = MAX_RECLAIM_RETRIES;
2605
	struct mem_cgroup *memcg;
2606
	bool in_retry = false;
2607 2608 2609 2610 2611 2612 2613

	if (likely(!nr_pages))
		return;

	memcg = get_mem_cgroup_from_mm(current->mm);
	current->memcg_nr_pages_over_high = 0;

2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627
retry_reclaim:
	/*
	 * The allocating task should reclaim at least the batch size, but for
	 * subsequent retries we only want to do what's necessary to prevent oom
	 * or breaching resource isolation.
	 *
	 * This is distinct from memory.max or page allocator behaviour because
	 * memory.high is currently batched, whereas memory.max and the page
	 * allocator run every time an allocation is made.
	 */
	nr_reclaimed = reclaim_high(memcg,
				    in_retry ? SWAP_CLUSTER_MAX : nr_pages,
				    GFP_KERNEL);

2628 2629 2630 2631
	/*
	 * memory.high is breached and reclaim is unable to keep up. Throttle
	 * allocators proactively to slow down excessive growth.
	 */
2632 2633
	penalty_jiffies = calculate_high_delay(memcg, nr_pages,
					       mem_find_max_overage(memcg));
2634

2635 2636 2637
	penalty_jiffies += calculate_high_delay(memcg, nr_pages,
						swap_find_max_overage(memcg));

2638 2639 2640 2641 2642 2643 2644
	/*
	 * Clamp the max delay per usermode return so as to still keep the
	 * application moving forwards and also permit diagnostics, albeit
	 * extremely slowly.
	 */
	penalty_jiffies = min(penalty_jiffies, MEMCG_MAX_HIGH_DELAY_JIFFIES);

2645 2646 2647 2648 2649 2650 2651 2652 2653
	/*
	 * Don't sleep if the amount of jiffies this memcg owes us is so low
	 * that it's not even worth doing, in an attempt to be nice to those who
	 * go only a small amount over their memory.high value and maybe haven't
	 * been aggressively reclaimed enough yet.
	 */
	if (penalty_jiffies <= HZ / 100)
		goto out;

2654 2655 2656 2657 2658 2659 2660 2661 2662 2663
	/*
	 * If reclaim is making forward progress but we're still over
	 * memory.high, we want to encourage that rather than doing allocator
	 * throttling.
	 */
	if (nr_reclaimed || nr_retries--) {
		in_retry = true;
		goto retry_reclaim;
	}

2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674
	/*
	 * If we exit early, we're guaranteed to die (since
	 * schedule_timeout_killable sets TASK_KILLABLE). This means we don't
	 * need to account for any ill-begotten jiffies to pay them off later.
	 */
	psi_memstall_enter(&pflags);
	schedule_timeout_killable(penalty_jiffies);
	psi_memstall_leave(&pflags);

out:
	css_put(&memcg->css);
2675 2676
}

2677 2678
static int try_charge(struct mem_cgroup *memcg, gfp_t gfp_mask,
		      unsigned int nr_pages)
2679
{
2680
	unsigned int batch = max(MEMCG_CHARGE_BATCH, nr_pages);
2681
	int nr_retries = MAX_RECLAIM_RETRIES;
2682
	struct mem_cgroup *mem_over_limit;
2683
	struct page_counter *counter;
2684
	enum oom_status oom_status;
2685
	unsigned long nr_reclaimed;
2686 2687
	bool may_swap = true;
	bool drained = false;
2688
	unsigned long pflags;
2689

2690
	if (mem_cgroup_is_root(memcg))
2691
		return 0;
2692
retry:
2693
	if (consume_stock(memcg, nr_pages))
2694
		return 0;
2695

2696
	if (!do_memsw_account() ||
2697 2698
	    page_counter_try_charge(&memcg->memsw, batch, &counter)) {
		if (page_counter_try_charge(&memcg->memory, batch, &counter))
2699
			goto done_restock;
2700
		if (do_memsw_account())
2701 2702
			page_counter_uncharge(&memcg->memsw, batch);
		mem_over_limit = mem_cgroup_from_counter(counter, memory);
2703
	} else {
2704
		mem_over_limit = mem_cgroup_from_counter(counter, memsw);
2705
		may_swap = false;
2706
	}
2707

2708 2709 2710 2711
	if (batch > nr_pages) {
		batch = nr_pages;
		goto retry;
	}
2712

2713 2714 2715 2716 2717 2718 2719 2720 2721
	/*
	 * Memcg doesn't have a dedicated reserve for atomic
	 * allocations. But like the global atomic pool, we need to
	 * put the burden of reclaim on regular allocation requests
	 * and let these go through as privileged allocations.
	 */
	if (gfp_mask & __GFP_ATOMIC)
		goto force;

2722 2723 2724 2725 2726 2727
	/*
	 * 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.
	 */
2728
	if (unlikely(should_force_charge()))
2729
		goto force;
2730

2731 2732 2733 2734 2735 2736 2737 2738 2739
	/*
	 * 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;

2740 2741 2742
	if (unlikely(task_in_memcg_oom(current)))
		goto nomem;

2743
	if (!gfpflags_allow_blocking(gfp_mask))
2744
		goto nomem;
2745

2746
	memcg_memory_event(mem_over_limit, MEMCG_MAX);
2747

2748
	psi_memstall_enter(&pflags);
2749 2750
	nr_reclaimed = try_to_free_mem_cgroup_pages(mem_over_limit, nr_pages,
						    gfp_mask, may_swap);
2751
	psi_memstall_leave(&pflags);
2752

2753
	if (mem_cgroup_margin(mem_over_limit) >= nr_pages)
2754
		goto retry;
2755

2756
	if (!drained) {
2757
		drain_all_stock(mem_over_limit);
2758 2759 2760 2761
		drained = true;
		goto retry;
	}

2762 2763
	if (gfp_mask & __GFP_NORETRY)
		goto nomem;
2764 2765 2766 2767 2768 2769 2770 2771 2772
	/*
	 * 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.
	 */
2773
	if (nr_reclaimed && nr_pages <= (1 << PAGE_ALLOC_COSTLY_ORDER))
2774 2775 2776 2777 2778 2779 2780 2781
		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;

2782 2783 2784
	if (nr_retries--)
		goto retry;

2785
	if (gfp_mask & __GFP_RETRY_MAYFAIL)
2786 2787
		goto nomem;

2788
	if (gfp_mask & __GFP_NOFAIL)
2789
		goto force;
2790

2791
	if (fatal_signal_pending(current))
2792
		goto force;
2793

2794 2795 2796 2797 2798 2799
	/*
	 * keep retrying as long as the memcg oom killer is able to make
	 * a forward progress or bypass the charge if the oom killer
	 * couldn't make any progress.
	 */
	oom_status = mem_cgroup_oom(mem_over_limit, gfp_mask,
2800
		       get_order(nr_pages * PAGE_SIZE));
2801 2802
	switch (oom_status) {
	case OOM_SUCCESS:
2803
		nr_retries = MAX_RECLAIM_RETRIES;
2804 2805 2806 2807 2808 2809
		goto retry;
	case OOM_FAILED:
		goto force;
	default:
		goto nomem;
	}
2810
nomem:
2811
	if (!(gfp_mask & __GFP_NOFAIL))
2812
		return -ENOMEM;
2813 2814 2815 2816 2817 2818 2819
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);
2820
	if (do_memsw_account())
2821 2822 2823
		page_counter_charge(&memcg->memsw, nr_pages);

	return 0;
2824 2825 2826 2827

done_restock:
	if (batch > nr_pages)
		refill_stock(memcg, batch - nr_pages);
2828

2829
	/*
2830 2831
	 * If the hierarchy is above the normal consumption range, schedule
	 * reclaim on returning to userland.  We can perform reclaim here
2832
	 * if __GFP_RECLAIM but let's always punt for simplicity and so that
2833 2834 2835 2836
	 * 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.
2837 2838
	 */
	do {
2839 2840 2841 2842 2843 2844 2845 2846 2847 2848
		bool mem_high, swap_high;

		mem_high = page_counter_read(&memcg->memory) >
			READ_ONCE(memcg->memory.high);
		swap_high = page_counter_read(&memcg->swap) >
			READ_ONCE(memcg->swap.high);

		/* Don't bother a random interrupted task */
		if (in_interrupt()) {
			if (mem_high) {
2849 2850 2851
				schedule_work(&memcg->high_work);
				break;
			}
2852 2853 2854 2855 2856 2857 2858 2859 2860 2861 2862 2863 2864
			continue;
		}

		if (mem_high || swap_high) {
			/*
			 * The allocating tasks in this cgroup will need to do
			 * reclaim or be throttled to prevent further growth
			 * of the memory or swap footprints.
			 *
			 * Target some best-effort fairness between the tasks,
			 * and distribute reclaim work and delay penalties
			 * based on how much each task is actually allocating.
			 */
V
Vladimir Davydov 已提交
2865
			current->memcg_nr_pages_over_high += batch;
2866 2867 2868
			set_notify_resume(current);
			break;
		}
2869
	} while ((memcg = parent_mem_cgroup(memcg)));
2870 2871

	return 0;
2872
}
2873

2874
#if defined(CONFIG_MEMCG_KMEM) || defined(CONFIG_MMU)
2875
static void cancel_charge(struct mem_cgroup *memcg, unsigned int nr_pages)
2876
{
2877 2878 2879
	if (mem_cgroup_is_root(memcg))
		return;

2880
	page_counter_uncharge(&memcg->memory, nr_pages);
2881
	if (do_memsw_account())
2882
		page_counter_uncharge(&memcg->memsw, nr_pages);
2883
}
2884
#endif
2885

2886
static void commit_charge(struct page *page, struct mem_cgroup *memcg)
2887
{
2888
	VM_BUG_ON_PAGE(page->mem_cgroup, page);
2889
	/*
2890
	 * Any of the following ensures page->mem_cgroup stability:
2891
	 *
2892 2893 2894 2895
	 * - the page lock
	 * - LRU isolation
	 * - lock_page_memcg()
	 * - exclusive reference
2896
	 */
2897
	page->mem_cgroup = memcg;
2898
}
2899

2900
#ifdef CONFIG_MEMCG_KMEM
2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914 2915 2916 2917 2918 2919 2920
int memcg_alloc_page_obj_cgroups(struct page *page, struct kmem_cache *s,
				 gfp_t gfp)
{
	unsigned int objects = objs_per_slab_page(s, page);
	void *vec;

	vec = kcalloc_node(objects, sizeof(struct obj_cgroup *), gfp,
			   page_to_nid(page));
	if (!vec)
		return -ENOMEM;

	if (cmpxchg(&page->obj_cgroups, NULL,
		    (struct obj_cgroup **) ((unsigned long)vec | 0x1UL)))
		kfree(vec);
	else
		kmemleak_not_leak(vec);

	return 0;
}

2921 2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934 2935
/*
 * Returns a pointer to the memory cgroup to which the kernel object is charged.
 *
 * The caller must ensure the memcg lifetime, e.g. by taking rcu_read_lock(),
 * cgroup_mutex, etc.
 */
struct mem_cgroup *mem_cgroup_from_obj(void *p)
{
	struct page *page;

	if (mem_cgroup_disabled())
		return NULL;

	page = virt_to_head_page(p);

2936 2937 2938 2939 2940 2941 2942 2943 2944 2945 2946
	/*
	 * If page->mem_cgroup is set, it's either a simple mem_cgroup pointer
	 * or a pointer to obj_cgroup vector. In the latter case the lowest
	 * bit of the pointer is set.
	 * The page->mem_cgroup pointer can be asynchronously changed
	 * from NULL to (obj_cgroup_vec | 0x1UL), but can't be changed
	 * from a valid memcg pointer to objcg vector or back.
	 */
	if (!page->mem_cgroup)
		return NULL;

2947
	/*
2948 2949 2950
	 * Slab objects are accounted individually, not per-page.
	 * Memcg membership data for each individual object is saved in
	 * the page->obj_cgroups.
2951
	 */
2952 2953 2954 2955 2956 2957
	if (page_has_obj_cgroups(page)) {
		struct obj_cgroup *objcg;
		unsigned int off;

		off = obj_to_index(page->slab_cache, page, p);
		objcg = page_obj_cgroups(page)[off];
2958 2959 2960 2961
		if (objcg)
			return obj_cgroup_memcg(objcg);

		return NULL;
2962
	}
2963 2964 2965 2966 2967

	/* All other pages use page->mem_cgroup */
	return page->mem_cgroup;
}

R
Roman Gushchin 已提交
2968 2969 2970 2971 2972
__always_inline struct obj_cgroup *get_obj_cgroup_from_current(void)
{
	struct obj_cgroup *objcg = NULL;
	struct mem_cgroup *memcg;

2973 2974 2975
	if (memcg_kmem_bypass())
		return NULL;

R
Roman Gushchin 已提交
2976
	rcu_read_lock();
2977 2978
	if (unlikely(active_memcg()))
		memcg = active_memcg();
R
Roman Gushchin 已提交
2979 2980 2981 2982 2983 2984 2985
	else
		memcg = mem_cgroup_from_task(current);

	for (; memcg != root_mem_cgroup; memcg = parent_mem_cgroup(memcg)) {
		objcg = rcu_dereference(memcg->objcg);
		if (objcg && obj_cgroup_tryget(objcg))
			break;
2986
		objcg = NULL;
R
Roman Gushchin 已提交
2987 2988 2989 2990 2991 2992
	}
	rcu_read_unlock();

	return objcg;
}

2993
static int memcg_alloc_cache_id(void)
2994
{
2995 2996 2997
	int id, size;
	int err;

2998
	id = ida_simple_get(&memcg_cache_ida,
2999 3000 3001
			    0, MEMCG_CACHES_MAX_SIZE, GFP_KERNEL);
	if (id < 0)
		return id;
3002

3003
	if (id < memcg_nr_cache_ids)
3004 3005 3006 3007 3008 3009
		return id;

	/*
	 * There's no space for the new id in memcg_caches arrays,
	 * so we have to grow them.
	 */
3010
	down_write(&memcg_cache_ids_sem);
3011 3012

	size = 2 * (id + 1);
3013 3014 3015 3016 3017
	if (size < MEMCG_CACHES_MIN_SIZE)
		size = MEMCG_CACHES_MIN_SIZE;
	else if (size > MEMCG_CACHES_MAX_SIZE)
		size = MEMCG_CACHES_MAX_SIZE;

3018
	err = memcg_update_all_list_lrus(size);
3019 3020 3021 3022 3023
	if (!err)
		memcg_nr_cache_ids = size;

	up_write(&memcg_cache_ids_sem);

3024
	if (err) {
3025
		ida_simple_remove(&memcg_cache_ida, id);
3026 3027 3028 3029 3030 3031 3032
		return err;
	}
	return id;
}

static void memcg_free_cache_id(int id)
{
3033
	ida_simple_remove(&memcg_cache_ida, id);
3034 3035
}

3036
/**
3037
 * __memcg_kmem_charge: charge a number of kernel pages to a memcg
3038
 * @memcg: memory cgroup to charge
3039
 * @gfp: reclaim mode
3040
 * @nr_pages: number of pages to charge
3041 3042 3043
 *
 * Returns 0 on success, an error code on failure.
 */
3044 3045
int __memcg_kmem_charge(struct mem_cgroup *memcg, gfp_t gfp,
			unsigned int nr_pages)
3046
{
3047
	struct page_counter *counter;
3048 3049
	int ret;

3050
	ret = try_charge(memcg, gfp, nr_pages);
3051
	if (ret)
3052
		return ret;
3053 3054 3055

	if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) &&
	    !page_counter_try_charge(&memcg->kmem, nr_pages, &counter)) {
3056 3057 3058 3059 3060 3061 3062 3063 3064 3065

		/*
		 * Enforce __GFP_NOFAIL allocation because callers are not
		 * prepared to see failures and likely do not have any failure
		 * handling code.
		 */
		if (gfp & __GFP_NOFAIL) {
			page_counter_charge(&memcg->kmem, nr_pages);
			return 0;
		}
3066 3067
		cancel_charge(memcg, nr_pages);
		return -ENOMEM;
3068
	}
3069
	return 0;
3070 3071
}

3072 3073 3074 3075 3076 3077 3078 3079 3080 3081
/**
 * __memcg_kmem_uncharge: uncharge a number of kernel pages from a memcg
 * @memcg: memcg to uncharge
 * @nr_pages: number of pages to uncharge
 */
void __memcg_kmem_uncharge(struct mem_cgroup *memcg, unsigned int nr_pages)
{
	if (!cgroup_subsys_on_dfl(memory_cgrp_subsys))
		page_counter_uncharge(&memcg->kmem, nr_pages);

3082
	refill_stock(memcg, nr_pages);
3083 3084
}

3085
/**
3086
 * __memcg_kmem_charge_page: charge a kmem page to the current memory cgroup
3087 3088 3089 3090 3091 3092
 * @page: page to charge
 * @gfp: reclaim mode
 * @order: allocation order
 *
 * Returns 0 on success, an error code on failure.
 */
3093
int __memcg_kmem_charge_page(struct page *page, gfp_t gfp, int order)
3094
{
3095
	struct mem_cgroup *memcg;
3096
	int ret = 0;
3097

3098
	memcg = get_mem_cgroup_from_current();
3099
	if (memcg && !mem_cgroup_is_root(memcg)) {
3100
		ret = __memcg_kmem_charge(memcg, gfp, 1 << order);
3101 3102
		if (!ret) {
			page->mem_cgroup = memcg;
3103
			__SetPageKmemcg(page);
3104
			return 0;
3105
		}
3106
		css_put(&memcg->css);
3107
	}
3108
	return ret;
3109
}
3110

3111
/**
3112
 * __memcg_kmem_uncharge_page: uncharge a kmem page
3113 3114 3115
 * @page: page to uncharge
 * @order: allocation order
 */
3116
void __memcg_kmem_uncharge_page(struct page *page, int order)
3117
{
3118
	struct mem_cgroup *memcg = page->mem_cgroup;
3119
	unsigned int nr_pages = 1 << order;
3120 3121 3122 3123

	if (!memcg)
		return;

3124
	VM_BUG_ON_PAGE(mem_cgroup_is_root(memcg), page);
3125
	__memcg_kmem_uncharge(memcg, nr_pages);
3126
	page->mem_cgroup = NULL;
3127
	css_put(&memcg->css);
3128 3129 3130 3131

	/* slab pages do not have PageKmemcg flag set */
	if (PageKmemcg(page))
		__ClearPageKmemcg(page);
3132
}
R
Roman Gushchin 已提交
3133 3134 3135 3136 3137 3138 3139 3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183 3184 3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243

static bool consume_obj_stock(struct obj_cgroup *objcg, unsigned int nr_bytes)
{
	struct memcg_stock_pcp *stock;
	unsigned long flags;
	bool ret = false;

	local_irq_save(flags);

	stock = this_cpu_ptr(&memcg_stock);
	if (objcg == stock->cached_objcg && stock->nr_bytes >= nr_bytes) {
		stock->nr_bytes -= nr_bytes;
		ret = true;
	}

	local_irq_restore(flags);

	return ret;
}

static void drain_obj_stock(struct memcg_stock_pcp *stock)
{
	struct obj_cgroup *old = stock->cached_objcg;

	if (!old)
		return;

	if (stock->nr_bytes) {
		unsigned int nr_pages = stock->nr_bytes >> PAGE_SHIFT;
		unsigned int nr_bytes = stock->nr_bytes & (PAGE_SIZE - 1);

		if (nr_pages) {
			rcu_read_lock();
			__memcg_kmem_uncharge(obj_cgroup_memcg(old), nr_pages);
			rcu_read_unlock();
		}

		/*
		 * The leftover is flushed to the centralized per-memcg value.
		 * On the next attempt to refill obj stock it will be moved
		 * to a per-cpu stock (probably, on an other CPU), see
		 * refill_obj_stock().
		 *
		 * How often it's flushed is a trade-off between the memory
		 * limit enforcement accuracy and potential CPU contention,
		 * so it might be changed in the future.
		 */
		atomic_add(nr_bytes, &old->nr_charged_bytes);
		stock->nr_bytes = 0;
	}

	obj_cgroup_put(old);
	stock->cached_objcg = NULL;
}

static bool obj_stock_flush_required(struct memcg_stock_pcp *stock,
				     struct mem_cgroup *root_memcg)
{
	struct mem_cgroup *memcg;

	if (stock->cached_objcg) {
		memcg = obj_cgroup_memcg(stock->cached_objcg);
		if (memcg && mem_cgroup_is_descendant(memcg, root_memcg))
			return true;
	}

	return false;
}

static void refill_obj_stock(struct obj_cgroup *objcg, unsigned int nr_bytes)
{
	struct memcg_stock_pcp *stock;
	unsigned long flags;

	local_irq_save(flags);

	stock = this_cpu_ptr(&memcg_stock);
	if (stock->cached_objcg != objcg) { /* reset if necessary */
		drain_obj_stock(stock);
		obj_cgroup_get(objcg);
		stock->cached_objcg = objcg;
		stock->nr_bytes = atomic_xchg(&objcg->nr_charged_bytes, 0);
	}
	stock->nr_bytes += nr_bytes;

	if (stock->nr_bytes > PAGE_SIZE)
		drain_obj_stock(stock);

	local_irq_restore(flags);
}

int obj_cgroup_charge(struct obj_cgroup *objcg, gfp_t gfp, size_t size)
{
	struct mem_cgroup *memcg;
	unsigned int nr_pages, nr_bytes;
	int ret;

	if (consume_obj_stock(objcg, size))
		return 0;

	/*
	 * In theory, memcg->nr_charged_bytes can have enough
	 * pre-charged bytes to satisfy the allocation. However,
	 * flushing memcg->nr_charged_bytes requires two atomic
	 * operations, and memcg->nr_charged_bytes can't be big,
	 * so it's better to ignore it and try grab some new pages.
	 * memcg->nr_charged_bytes will be flushed in
	 * refill_obj_stock(), called from this function or
	 * independently later.
	 */
	rcu_read_lock();
3244
retry:
R
Roman Gushchin 已提交
3245
	memcg = obj_cgroup_memcg(objcg);
3246 3247
	if (unlikely(!css_tryget(&memcg->css)))
		goto retry;
R
Roman Gushchin 已提交
3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268
	rcu_read_unlock();

	nr_pages = size >> PAGE_SHIFT;
	nr_bytes = size & (PAGE_SIZE - 1);

	if (nr_bytes)
		nr_pages += 1;

	ret = __memcg_kmem_charge(memcg, gfp, nr_pages);
	if (!ret && nr_bytes)
		refill_obj_stock(objcg, PAGE_SIZE - nr_bytes);

	css_put(&memcg->css);
	return ret;
}

void obj_cgroup_uncharge(struct obj_cgroup *objcg, size_t size)
{
	refill_obj_stock(objcg, size);
}

3269
#endif /* CONFIG_MEMCG_KMEM */
3270

3271 3272 3273 3274
#ifdef CONFIG_TRANSPARENT_HUGEPAGE

/*
 * Because tail pages are not marked as "used", set it. We're under
3275
 * pgdat->lru_lock and migration entries setup in all page mappings.
3276
 */
3277
void mem_cgroup_split_huge_fixup(struct page *head)
3278
{
3279
	struct mem_cgroup *memcg = head->mem_cgroup;
3280
	int i;
3281

3282 3283
	if (mem_cgroup_disabled())
		return;
3284

3285 3286 3287 3288
	for (i = 1; i < HPAGE_PMD_NR; i++) {
		css_get(&memcg->css);
		head[i].mem_cgroup = memcg;
	}
3289
}
3290
#endif /* CONFIG_TRANSPARENT_HUGEPAGE */
3291

A
Andrew Morton 已提交
3292
#ifdef CONFIG_MEMCG_SWAP
3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303
/**
 * 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.
 *
3304
 * The caller must have charged to @to, IOW, called page_counter_charge() about
3305 3306 3307
 * both res and memsw, and called css_get().
 */
static int mem_cgroup_move_swap_account(swp_entry_t entry,
3308
				struct mem_cgroup *from, struct mem_cgroup *to)
3309 3310 3311
{
	unsigned short old_id, new_id;

L
Li Zefan 已提交
3312 3313
	old_id = mem_cgroup_id(from);
	new_id = mem_cgroup_id(to);
3314 3315

	if (swap_cgroup_cmpxchg(entry, old_id, new_id) == old_id) {
3316 3317
		mod_memcg_state(from, MEMCG_SWAP, -1);
		mod_memcg_state(to, MEMCG_SWAP, 1);
3318 3319 3320 3321 3322 3323
		return 0;
	}
	return -EINVAL;
}
#else
static inline int mem_cgroup_move_swap_account(swp_entry_t entry,
3324
				struct mem_cgroup *from, struct mem_cgroup *to)
3325 3326 3327
{
	return -EINVAL;
}
3328
#endif
K
KAMEZAWA Hiroyuki 已提交
3329

3330
static DEFINE_MUTEX(memcg_max_mutex);
3331

3332 3333
static int mem_cgroup_resize_max(struct mem_cgroup *memcg,
				 unsigned long max, bool memsw)
3334
{
3335
	bool enlarge = false;
3336
	bool drained = false;
3337
	int ret;
3338 3339
	bool limits_invariant;
	struct page_counter *counter = memsw ? &memcg->memsw : &memcg->memory;
3340

3341
	do {
3342 3343 3344 3345
		if (signal_pending(current)) {
			ret = -EINTR;
			break;
		}
3346

3347
		mutex_lock(&memcg_max_mutex);
3348 3349
		/*
		 * Make sure that the new limit (memsw or memory limit) doesn't
3350
		 * break our basic invariant rule memory.max <= memsw.max.
3351
		 */
3352
		limits_invariant = memsw ? max >= READ_ONCE(memcg->memory.max) :
3353
					   max <= memcg->memsw.max;
3354
		if (!limits_invariant) {
3355
			mutex_unlock(&memcg_max_mutex);
3356 3357 3358
			ret = -EINVAL;
			break;
		}
3359
		if (max > counter->max)
3360
			enlarge = true;
3361 3362
		ret = page_counter_set_max(counter, max);
		mutex_unlock(&memcg_max_mutex);
3363 3364 3365 3366

		if (!ret)
			break;

3367 3368 3369 3370 3371 3372
		if (!drained) {
			drain_all_stock(memcg);
			drained = true;
			continue;
		}

3373 3374 3375 3376 3377 3378
		if (!try_to_free_mem_cgroup_pages(memcg, 1,
					GFP_KERNEL, !memsw)) {
			ret = -EBUSY;
			break;
		}
	} while (true);
3379

3380 3381
	if (!ret && enlarge)
		memcg_oom_recover(memcg);
3382

3383 3384 3385
	return ret;
}

3386
unsigned long mem_cgroup_soft_limit_reclaim(pg_data_t *pgdat, int order,
3387 3388 3389 3390
					    gfp_t gfp_mask,
					    unsigned long *total_scanned)
{
	unsigned long nr_reclaimed = 0;
3391
	struct mem_cgroup_per_node *mz, *next_mz = NULL;
3392 3393
	unsigned long reclaimed;
	int loop = 0;
3394
	struct mem_cgroup_tree_per_node *mctz;
3395
	unsigned long excess;
3396 3397 3398 3399 3400
	unsigned long nr_scanned;

	if (order > 0)
		return 0;

3401
	mctz = soft_limit_tree_node(pgdat->node_id);
3402 3403 3404 3405 3406 3407

	/*
	 * 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.
	 */
3408
	if (!mctz || RB_EMPTY_ROOT(&mctz->rb_root))
3409 3410
		return 0;

3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424
	/*
	 * 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;
3425
		reclaimed = mem_cgroup_soft_reclaim(mz->memcg, pgdat,
3426 3427 3428
						    gfp_mask, &nr_scanned);
		nr_reclaimed += reclaimed;
		*total_scanned += nr_scanned;
3429
		spin_lock_irq(&mctz->lock);
3430
		__mem_cgroup_remove_exceeded(mz, mctz);
3431 3432 3433 3434 3435 3436

		/*
		 * If we failed to reclaim anything from this memory cgroup
		 * it is time to move on to the next cgroup
		 */
		next_mz = NULL;
3437 3438 3439
		if (!reclaimed)
			next_mz = __mem_cgroup_largest_soft_limit_node(mctz);

3440
		excess = soft_limit_excess(mz->memcg);
3441 3442 3443 3444 3445 3446 3447 3448 3449
		/*
		 * 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 */
3450
		__mem_cgroup_insert_exceeded(mz, mctz, excess);
3451
		spin_unlock_irq(&mctz->lock);
3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468
		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;
}

3469 3470 3471 3472
/*
 * 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
3473
 * hierarchy.  Testing use_hierarchy is the caller's responsibility.
3474
 */
3475 3476
static inline bool memcg_has_children(struct mem_cgroup *memcg)
{
3477 3478 3479 3480 3481 3482
	bool ret;

	rcu_read_lock();
	ret = css_next_child(NULL, &memcg->css);
	rcu_read_unlock();
	return ret;
3483 3484
}

3485
/*
3486
 * Reclaims as many pages from the given memcg as possible.
3487 3488 3489 3490 3491
 *
 * Caller is responsible for holding css reference for memcg.
 */
static int mem_cgroup_force_empty(struct mem_cgroup *memcg)
{
3492
	int nr_retries = MAX_RECLAIM_RETRIES;
3493

3494 3495
	/* we call try-to-free pages for make this cgroup empty */
	lru_add_drain_all();
3496 3497 3498

	drain_all_stock(memcg);

3499
	/* try to free all pages in this cgroup */
3500
	while (nr_retries && page_counter_read(&memcg->memory)) {
3501
		int progress;
3502

3503 3504 3505
		if (signal_pending(current))
			return -EINTR;

3506 3507
		progress = try_to_free_mem_cgroup_pages(memcg, 1,
							GFP_KERNEL, true);
3508
		if (!progress) {
3509
			nr_retries--;
3510
			/* maybe some writeback is necessary */
3511
			congestion_wait(BLK_RW_ASYNC, HZ/10);
3512
		}
3513 3514

	}
3515 3516

	return 0;
3517 3518
}

3519 3520 3521
static ssize_t mem_cgroup_force_empty_write(struct kernfs_open_file *of,
					    char *buf, size_t nbytes,
					    loff_t off)
3522
{
3523
	struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of));
3524

3525 3526
	if (mem_cgroup_is_root(memcg))
		return -EINVAL;
3527
	return mem_cgroup_force_empty(memcg) ?: nbytes;
3528 3529
}

3530 3531
static u64 mem_cgroup_hierarchy_read(struct cgroup_subsys_state *css,
				     struct cftype *cft)
3532
{
3533
	return mem_cgroup_from_css(css)->use_hierarchy;
3534 3535
}

3536 3537
static int mem_cgroup_hierarchy_write(struct cgroup_subsys_state *css,
				      struct cftype *cft, u64 val)
3538 3539
{
	int retval = 0;
3540
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
T
Tejun Heo 已提交
3541
	struct mem_cgroup *parent_memcg = mem_cgroup_from_css(memcg->css.parent);
3542

3543
	if (memcg->use_hierarchy == val)
3544
		return 0;
3545

3546
	/*
3547
	 * If parent's use_hierarchy is set, we can't make any modifications
3548 3549 3550 3551 3552 3553
	 * 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.
	 */
3554
	if ((!parent_memcg || !parent_memcg->use_hierarchy) &&
3555
				(val == 1 || val == 0)) {
3556
		if (!memcg_has_children(memcg))
3557
			memcg->use_hierarchy = val;
3558 3559 3560 3561
		else
			retval = -EBUSY;
	} else
		retval = -EINVAL;
3562

3563 3564 3565
	return retval;
}

3566
static unsigned long mem_cgroup_usage(struct mem_cgroup *memcg, bool swap)
3567
{
3568
	unsigned long val;
3569

3570
	if (mem_cgroup_is_root(memcg)) {
3571
		val = memcg_page_state(memcg, NR_FILE_PAGES) +
3572
			memcg_page_state(memcg, NR_ANON_MAPPED);
3573 3574
		if (swap)
			val += memcg_page_state(memcg, MEMCG_SWAP);
3575
	} else {
3576
		if (!swap)
3577
			val = page_counter_read(&memcg->memory);
3578
		else
3579
			val = page_counter_read(&memcg->memsw);
3580
	}
3581
	return val;
3582 3583
}

3584 3585 3586 3587 3588 3589 3590
enum {
	RES_USAGE,
	RES_LIMIT,
	RES_MAX_USAGE,
	RES_FAILCNT,
	RES_SOFT_LIMIT,
};
3591

3592
static u64 mem_cgroup_read_u64(struct cgroup_subsys_state *css,
3593
			       struct cftype *cft)
B
Balbir Singh 已提交
3594
{
3595
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
3596
	struct page_counter *counter;
3597

3598
	switch (MEMFILE_TYPE(cft->private)) {
3599
	case _MEM:
3600 3601
		counter = &memcg->memory;
		break;
3602
	case _MEMSWAP:
3603 3604
		counter = &memcg->memsw;
		break;
3605
	case _KMEM:
3606
		counter = &memcg->kmem;
3607
		break;
V
Vladimir Davydov 已提交
3608
	case _TCP:
3609
		counter = &memcg->tcpmem;
V
Vladimir Davydov 已提交
3610
		break;
3611 3612 3613
	default:
		BUG();
	}
3614 3615 3616 3617

	switch (MEMFILE_ATTR(cft->private)) {
	case RES_USAGE:
		if (counter == &memcg->memory)
3618
			return (u64)mem_cgroup_usage(memcg, false) * PAGE_SIZE;
3619
		if (counter == &memcg->memsw)
3620
			return (u64)mem_cgroup_usage(memcg, true) * PAGE_SIZE;
3621 3622
		return (u64)page_counter_read(counter) * PAGE_SIZE;
	case RES_LIMIT:
3623
		return (u64)counter->max * PAGE_SIZE;
3624 3625 3626 3627 3628 3629 3630 3631 3632
	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
3633
}
3634

3635
static void memcg_flush_percpu_vmstats(struct mem_cgroup *memcg)
3636
{
3637
	unsigned long stat[MEMCG_NR_STAT] = {0};
3638 3639 3640 3641
	struct mem_cgroup *mi;
	int node, cpu, i;

	for_each_online_cpu(cpu)
3642
		for (i = 0; i < MEMCG_NR_STAT; i++)
3643
			stat[i] += per_cpu(memcg->vmstats_percpu->stat[i], cpu);
3644 3645

	for (mi = memcg; mi; mi = parent_mem_cgroup(mi))
3646
		for (i = 0; i < MEMCG_NR_STAT; i++)
3647 3648 3649 3650 3651 3652
			atomic_long_add(stat[i], &mi->vmstats[i]);

	for_each_node(node) {
		struct mem_cgroup_per_node *pn = memcg->nodeinfo[node];
		struct mem_cgroup_per_node *pi;

3653
		for (i = 0; i < NR_VM_NODE_STAT_ITEMS; i++)
3654 3655 3656
			stat[i] = 0;

		for_each_online_cpu(cpu)
3657
			for (i = 0; i < NR_VM_NODE_STAT_ITEMS; i++)
3658 3659
				stat[i] += per_cpu(
					pn->lruvec_stat_cpu->count[i], cpu);
3660 3661

		for (pi = pn; pi; pi = parent_nodeinfo(pi, node))
3662
			for (i = 0; i < NR_VM_NODE_STAT_ITEMS; i++)
3663 3664 3665 3666
				atomic_long_add(stat[i], &pi->lruvec_stat[i]);
	}
}

3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677
static void memcg_flush_percpu_vmevents(struct mem_cgroup *memcg)
{
	unsigned long events[NR_VM_EVENT_ITEMS];
	struct mem_cgroup *mi;
	int cpu, i;

	for (i = 0; i < NR_VM_EVENT_ITEMS; i++)
		events[i] = 0;

	for_each_online_cpu(cpu)
		for (i = 0; i < NR_VM_EVENT_ITEMS; i++)
3678 3679
			events[i] += per_cpu(memcg->vmstats_percpu->events[i],
					     cpu);
3680 3681 3682 3683 3684 3685

	for (mi = memcg; mi; mi = parent_mem_cgroup(mi))
		for (i = 0; i < NR_VM_EVENT_ITEMS; i++)
			atomic_long_add(events[i], &mi->vmevents[i]);
}

3686
#ifdef CONFIG_MEMCG_KMEM
3687
static int memcg_online_kmem(struct mem_cgroup *memcg)
3688
{
R
Roman Gushchin 已提交
3689
	struct obj_cgroup *objcg;
3690 3691
	int memcg_id;

3692 3693 3694
	if (cgroup_memory_nokmem)
		return 0;

3695
	BUG_ON(memcg->kmemcg_id >= 0);
3696
	BUG_ON(memcg->kmem_state);
3697

3698
	memcg_id = memcg_alloc_cache_id();
3699 3700
	if (memcg_id < 0)
		return memcg_id;
3701

R
Roman Gushchin 已提交
3702 3703 3704 3705 3706 3707 3708 3709
	objcg = obj_cgroup_alloc();
	if (!objcg) {
		memcg_free_cache_id(memcg_id);
		return -ENOMEM;
	}
	objcg->memcg = memcg;
	rcu_assign_pointer(memcg->objcg, objcg);

3710 3711
	static_branch_enable(&memcg_kmem_enabled_key);

3712
	/*
3713
	 * A memory cgroup is considered kmem-online as soon as it gets
V
Vladimir Davydov 已提交
3714
	 * kmemcg_id. Setting the id after enabling static branching will
3715 3716 3717
	 * guarantee no one starts accounting before all call sites are
	 * patched.
	 */
V
Vladimir Davydov 已提交
3718
	memcg->kmemcg_id = memcg_id;
3719
	memcg->kmem_state = KMEM_ONLINE;
3720 3721

	return 0;
3722 3723
}

3724 3725 3726 3727 3728 3729 3730 3731
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;
3732

3733 3734 3735 3736 3737 3738
	memcg->kmem_state = KMEM_ALLOCATED;

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

R
Roman Gushchin 已提交
3739
	memcg_reparent_objcgs(memcg, parent);
3740 3741 3742 3743

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

3744 3745 3746 3747 3748 3749 3750 3751
	/*
	 * 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().
	 */
3752
	rcu_read_lock(); /* can be called from css_free w/o cgroup_mutex */
3753 3754 3755 3756 3757 3758 3759
	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;
	}
3760 3761
	rcu_read_unlock();

3762
	memcg_drain_all_list_lrus(kmemcg_id, parent);
3763 3764 3765 3766 3767 3768

	memcg_free_cache_id(kmemcg_id);
}

static void memcg_free_kmem(struct mem_cgroup *memcg)
{
3769 3770 3771
	/* css_alloc() failed, offlining didn't happen */
	if (unlikely(memcg->kmem_state == KMEM_ONLINE))
		memcg_offline_kmem(memcg);
3772
}
3773
#else
3774
static int memcg_online_kmem(struct mem_cgroup *memcg)
3775 3776 3777 3778 3779 3780 3781 3782 3783
{
	return 0;
}
static void memcg_offline_kmem(struct mem_cgroup *memcg)
{
}
static void memcg_free_kmem(struct mem_cgroup *memcg)
{
}
3784
#endif /* CONFIG_MEMCG_KMEM */
3785

3786 3787
static int memcg_update_kmem_max(struct mem_cgroup *memcg,
				 unsigned long max)
3788
{
3789
	int ret;
3790

3791 3792 3793
	mutex_lock(&memcg_max_mutex);
	ret = page_counter_set_max(&memcg->kmem, max);
	mutex_unlock(&memcg_max_mutex);
3794
	return ret;
3795
}
3796

3797
static int memcg_update_tcp_max(struct mem_cgroup *memcg, unsigned long max)
V
Vladimir Davydov 已提交
3798 3799 3800
{
	int ret;

3801
	mutex_lock(&memcg_max_mutex);
V
Vladimir Davydov 已提交
3802

3803
	ret = page_counter_set_max(&memcg->tcpmem, max);
V
Vladimir Davydov 已提交
3804 3805 3806
	if (ret)
		goto out;

3807
	if (!memcg->tcpmem_active) {
V
Vladimir Davydov 已提交
3808 3809 3810
		/*
		 * The active flag needs to be written after the static_key
		 * update. This is what guarantees that the socket activation
3811 3812 3813
		 * 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
3814 3815 3816 3817 3818 3819
		 *
		 * 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.
		 *
3820
		 * We never race with the readers in mem_cgroup_sk_alloc(),
V
Vladimir Davydov 已提交
3821 3822 3823 3824
		 * because when this value change, the code to process it is not
		 * patched in yet.
		 */
		static_branch_inc(&memcg_sockets_enabled_key);
3825
		memcg->tcpmem_active = true;
V
Vladimir Davydov 已提交
3826 3827
	}
out:
3828
	mutex_unlock(&memcg_max_mutex);
V
Vladimir Davydov 已提交
3829 3830 3831
	return ret;
}

3832 3833 3834 3835
/*
 * The user of this function is...
 * RES_LIMIT.
 */
3836 3837
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
3838
{
3839
	struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of));
3840
	unsigned long nr_pages;
3841 3842
	int ret;

3843
	buf = strstrip(buf);
3844
	ret = page_counter_memparse(buf, "-1", &nr_pages);
3845 3846
	if (ret)
		return ret;
3847

3848
	switch (MEMFILE_ATTR(of_cft(of)->private)) {
3849
	case RES_LIMIT:
3850 3851 3852 3853
		if (mem_cgroup_is_root(memcg)) { /* Can't set limit on root */
			ret = -EINVAL;
			break;
		}
3854 3855
		switch (MEMFILE_TYPE(of_cft(of)->private)) {
		case _MEM:
3856
			ret = mem_cgroup_resize_max(memcg, nr_pages, false);
3857
			break;
3858
		case _MEMSWAP:
3859
			ret = mem_cgroup_resize_max(memcg, nr_pages, true);
3860
			break;
3861
		case _KMEM:
3862 3863 3864
			pr_warn_once("kmem.limit_in_bytes is deprecated and will be removed. "
				     "Please report your usecase to linux-mm@kvack.org if you "
				     "depend on this functionality.\n");
3865
			ret = memcg_update_kmem_max(memcg, nr_pages);
3866
			break;
V
Vladimir Davydov 已提交
3867
		case _TCP:
3868
			ret = memcg_update_tcp_max(memcg, nr_pages);
V
Vladimir Davydov 已提交
3869
			break;
3870
		}
3871
		break;
3872 3873 3874
	case RES_SOFT_LIMIT:
		memcg->soft_limit = nr_pages;
		ret = 0;
3875 3876
		break;
	}
3877
	return ret ?: nbytes;
B
Balbir Singh 已提交
3878 3879
}

3880 3881
static ssize_t mem_cgroup_reset(struct kernfs_open_file *of, char *buf,
				size_t nbytes, loff_t off)
3882
{
3883
	struct mem_cgroup *memcg = mem_cgroup_from_css(of_css(of));
3884
	struct page_counter *counter;
3885

3886 3887 3888 3889 3890 3891 3892 3893 3894 3895
	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
3896
	case _TCP:
3897
		counter = &memcg->tcpmem;
V
Vladimir Davydov 已提交
3898
		break;
3899 3900 3901
	default:
		BUG();
	}
3902

3903
	switch (MEMFILE_ATTR(of_cft(of)->private)) {
3904
	case RES_MAX_USAGE:
3905
		page_counter_reset_watermark(counter);
3906 3907
		break;
	case RES_FAILCNT:
3908
		counter->failcnt = 0;
3909
		break;
3910 3911
	default:
		BUG();
3912
	}
3913

3914
	return nbytes;
3915 3916
}

3917
static u64 mem_cgroup_move_charge_read(struct cgroup_subsys_state *css,
3918 3919
					struct cftype *cft)
{
3920
	return mem_cgroup_from_css(css)->move_charge_at_immigrate;
3921 3922
}

3923
#ifdef CONFIG_MMU
3924
static int mem_cgroup_move_charge_write(struct cgroup_subsys_state *css,
3925 3926
					struct cftype *cft, u64 val)
{
3927
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
3928

3929
	if (val & ~MOVE_MASK)
3930
		return -EINVAL;
3931

3932
	/*
3933 3934 3935 3936
	 * 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.
3937
	 */
3938
	memcg->move_charge_at_immigrate = val;
3939 3940
	return 0;
}
3941
#else
3942
static int mem_cgroup_move_charge_write(struct cgroup_subsys_state *css,
3943 3944 3945 3946 3947
					struct cftype *cft, u64 val)
{
	return -ENOSYS;
}
#endif
3948

3949
#ifdef CONFIG_NUMA
3950 3951 3952 3953 3954 3955

#define LRU_ALL_FILE (BIT(LRU_INACTIVE_FILE) | BIT(LRU_ACTIVE_FILE))
#define LRU_ALL_ANON (BIT(LRU_INACTIVE_ANON) | BIT(LRU_ACTIVE_ANON))
#define LRU_ALL	     ((1 << NR_LRU_LISTS) - 1)

static unsigned long mem_cgroup_node_nr_lru_pages(struct mem_cgroup *memcg,
3956
				int nid, unsigned int lru_mask, bool tree)
3957
{
3958
	struct lruvec *lruvec = mem_cgroup_lruvec(memcg, NODE_DATA(nid));
3959 3960 3961 3962 3963 3964 3965 3966
	unsigned long nr = 0;
	enum lru_list lru;

	VM_BUG_ON((unsigned)nid >= nr_node_ids);

	for_each_lru(lru) {
		if (!(BIT(lru) & lru_mask))
			continue;
3967 3968 3969 3970
		if (tree)
			nr += lruvec_page_state(lruvec, NR_LRU_BASE + lru);
		else
			nr += lruvec_page_state_local(lruvec, NR_LRU_BASE + lru);
3971 3972 3973 3974 3975
	}
	return nr;
}

static unsigned long mem_cgroup_nr_lru_pages(struct mem_cgroup *memcg,
3976 3977
					     unsigned int lru_mask,
					     bool tree)
3978 3979 3980 3981 3982 3983 3984
{
	unsigned long nr = 0;
	enum lru_list lru;

	for_each_lru(lru) {
		if (!(BIT(lru) & lru_mask))
			continue;
3985 3986 3987 3988
		if (tree)
			nr += memcg_page_state(memcg, NR_LRU_BASE + lru);
		else
			nr += memcg_page_state_local(memcg, NR_LRU_BASE + lru);
3989 3990 3991 3992
	}
	return nr;
}

3993
static int memcg_numa_stat_show(struct seq_file *m, void *v)
3994
{
3995 3996 3997 3998 3999 4000 4001 4002 4003 4004 4005 4006
	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;
4007
	int nid;
4008
	struct mem_cgroup *memcg = mem_cgroup_from_seq(m);
4009

4010
	for (stat = stats; stat < stats + ARRAY_SIZE(stats); stat++) {
4011 4012 4013 4014 4015 4016 4017
		seq_printf(m, "%s=%lu", stat->name,
			   mem_cgroup_nr_lru_pages(memcg, stat->lru_mask,
						   false));
		for_each_node_state(nid, N_MEMORY)
			seq_printf(m, " N%d=%lu", nid,
				   mem_cgroup_node_nr_lru_pages(memcg, nid,
							stat->lru_mask, false));
4018
		seq_putc(m, '\n');
4019 4020
	}

4021
	for (stat = stats; stat < stats + ARRAY_SIZE(stats); stat++) {
4022 4023 4024 4025 4026 4027 4028 4029

		seq_printf(m, "hierarchical_%s=%lu", stat->name,
			   mem_cgroup_nr_lru_pages(memcg, stat->lru_mask,
						   true));
		for_each_node_state(nid, N_MEMORY)
			seq_printf(m, " N%d=%lu", nid,
				   mem_cgroup_node_nr_lru_pages(memcg, nid,
							stat->lru_mask, true));
4030
		seq_putc(m, '\n');
4031 4032 4033 4034 4035 4036
	}

	return 0;
}
#endif /* CONFIG_NUMA */

4037
static const unsigned int memcg1_stats[] = {
4038
	NR_FILE_PAGES,
4039
	NR_ANON_MAPPED,
4040 4041 4042
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
	NR_ANON_THPS,
#endif
4043 4044 4045 4046 4047 4048 4049 4050 4051 4052
	NR_SHMEM,
	NR_FILE_MAPPED,
	NR_FILE_DIRTY,
	NR_WRITEBACK,
	MEMCG_SWAP,
};

static const char *const memcg1_stat_names[] = {
	"cache",
	"rss",
4053
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
4054
	"rss_huge",
4055
#endif
4056 4057 4058 4059 4060 4061 4062
	"shmem",
	"mapped_file",
	"dirty",
	"writeback",
	"swap",
};

4063
/* Universal VM events cgroup1 shows, original sort order */
4064
static const unsigned int memcg1_events[] = {
4065 4066 4067 4068 4069 4070
	PGPGIN,
	PGPGOUT,
	PGFAULT,
	PGMAJFAULT,
};

4071
static int memcg_stat_show(struct seq_file *m, void *v)
4072
{
4073
	struct mem_cgroup *memcg = mem_cgroup_from_seq(m);
4074
	unsigned long memory, memsw;
4075 4076
	struct mem_cgroup *mi;
	unsigned int i;
4077

4078
	BUILD_BUG_ON(ARRAY_SIZE(memcg1_stat_names) != ARRAY_SIZE(memcg1_stats));
4079

4080
	for (i = 0; i < ARRAY_SIZE(memcg1_stats); i++) {
4081 4082
		unsigned long nr;

4083
		if (memcg1_stats[i] == MEMCG_SWAP && !do_memsw_account())
4084
			continue;
4085 4086 4087 4088 4089 4090
		nr = memcg_page_state_local(memcg, memcg1_stats[i]);
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
		if (memcg1_stats[i] == NR_ANON_THPS)
			nr *= HPAGE_PMD_NR;
#endif
		seq_printf(m, "%s %lu\n", memcg1_stat_names[i], nr * PAGE_SIZE);
4091
	}
L
Lee Schermerhorn 已提交
4092

4093
	for (i = 0; i < ARRAY_SIZE(memcg1_events); i++)
4094
		seq_printf(m, "%s %lu\n", vm_event_name(memcg1_events[i]),
4095
			   memcg_events_local(memcg, memcg1_events[i]));
4096 4097

	for (i = 0; i < NR_LRU_LISTS; i++)
4098
		seq_printf(m, "%s %lu\n", lru_list_name(i),
4099
			   memcg_page_state_local(memcg, NR_LRU_BASE + i) *
4100
			   PAGE_SIZE);
4101

K
KAMEZAWA Hiroyuki 已提交
4102
	/* Hierarchical information */
4103 4104
	memory = memsw = PAGE_COUNTER_MAX;
	for (mi = memcg; mi; mi = parent_mem_cgroup(mi)) {
4105 4106
		memory = min(memory, READ_ONCE(mi->memory.max));
		memsw = min(memsw, READ_ONCE(mi->memsw.max));
4107
	}
4108 4109
	seq_printf(m, "hierarchical_memory_limit %llu\n",
		   (u64)memory * PAGE_SIZE);
4110
	if (do_memsw_account())
4111 4112
		seq_printf(m, "hierarchical_memsw_limit %llu\n",
			   (u64)memsw * PAGE_SIZE);
K
KOSAKI Motohiro 已提交
4113

4114
	for (i = 0; i < ARRAY_SIZE(memcg1_stats); i++) {
4115 4116
		unsigned long nr;

4117
		if (memcg1_stats[i] == MEMCG_SWAP && !do_memsw_account())
4118
			continue;
4119 4120 4121 4122 4123
		nr = memcg_page_state(memcg, memcg1_stats[i]);
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
		if (memcg1_stats[i] == NR_ANON_THPS)
			nr *= HPAGE_PMD_NR;
#endif
4124
		seq_printf(m, "total_%s %llu\n", memcg1_stat_names[i],
4125
						(u64)nr * PAGE_SIZE);
4126 4127
	}

4128
	for (i = 0; i < ARRAY_SIZE(memcg1_events); i++)
4129 4130
		seq_printf(m, "total_%s %llu\n",
			   vm_event_name(memcg1_events[i]),
4131
			   (u64)memcg_events(memcg, memcg1_events[i]));
4132

4133
	for (i = 0; i < NR_LRU_LISTS; i++)
4134
		seq_printf(m, "total_%s %llu\n", lru_list_name(i),
4135 4136
			   (u64)memcg_page_state(memcg, NR_LRU_BASE + i) *
			   PAGE_SIZE);
K
KAMEZAWA Hiroyuki 已提交
4137

K
KOSAKI Motohiro 已提交
4138 4139
#ifdef CONFIG_DEBUG_VM
	{
4140 4141
		pg_data_t *pgdat;
		struct mem_cgroup_per_node *mz;
4142 4143
		unsigned long anon_cost = 0;
		unsigned long file_cost = 0;
K
KOSAKI Motohiro 已提交
4144

4145 4146
		for_each_online_pgdat(pgdat) {
			mz = mem_cgroup_nodeinfo(memcg, pgdat->node_id);
K
KOSAKI Motohiro 已提交
4147

4148 4149
			anon_cost += mz->lruvec.anon_cost;
			file_cost += mz->lruvec.file_cost;
4150
		}
4151 4152
		seq_printf(m, "anon_cost %lu\n", anon_cost);
		seq_printf(m, "file_cost %lu\n", file_cost);
K
KOSAKI Motohiro 已提交
4153 4154 4155
	}
#endif

4156 4157 4158
	return 0;
}

4159 4160
static u64 mem_cgroup_swappiness_read(struct cgroup_subsys_state *css,
				      struct cftype *cft)
K
KOSAKI Motohiro 已提交
4161
{
4162
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
K
KOSAKI Motohiro 已提交
4163

4164
	return mem_cgroup_swappiness(memcg);
K
KOSAKI Motohiro 已提交
4165 4166
}

4167 4168
static int mem_cgroup_swappiness_write(struct cgroup_subsys_state *css,
				       struct cftype *cft, u64 val)
K
KOSAKI Motohiro 已提交
4169
{
4170
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
K
KOSAKI Motohiro 已提交
4171

4172
	if (val > 100)
K
KOSAKI Motohiro 已提交
4173 4174
		return -EINVAL;

4175
	if (css->parent)
4176 4177 4178
		memcg->swappiness = val;
	else
		vm_swappiness = val;
4179

K
KOSAKI Motohiro 已提交
4180 4181 4182
	return 0;
}

4183 4184 4185
static void __mem_cgroup_threshold(struct mem_cgroup *memcg, bool swap)
{
	struct mem_cgroup_threshold_ary *t;
4186
	unsigned long usage;
4187 4188 4189 4190
	int i;

	rcu_read_lock();
	if (!swap)
4191
		t = rcu_dereference(memcg->thresholds.primary);
4192
	else
4193
		t = rcu_dereference(memcg->memsw_thresholds.primary);
4194 4195 4196 4197

	if (!t)
		goto unlock;

4198
	usage = mem_cgroup_usage(memcg, swap);
4199 4200

	/*
4201
	 * current_threshold points to threshold just below or equal to usage.
4202 4203 4204
	 * If it's not true, a threshold was crossed after last
	 * call of __mem_cgroup_threshold().
	 */
4205
	i = t->current_threshold;
4206 4207 4208 4209 4210 4211 4212 4213 4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227 4228

	/*
	 * 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 */
4229
	t->current_threshold = i - 1;
4230 4231 4232 4233 4234 4235
unlock:
	rcu_read_unlock();
}

static void mem_cgroup_threshold(struct mem_cgroup *memcg)
{
4236 4237
	while (memcg) {
		__mem_cgroup_threshold(memcg, false);
4238
		if (do_memsw_account())
4239 4240 4241 4242
			__mem_cgroup_threshold(memcg, true);

		memcg = parent_mem_cgroup(memcg);
	}
4243 4244 4245 4246 4247 4248 4249
}

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

4250 4251 4252 4253 4254 4255 4256
	if (_a->threshold > _b->threshold)
		return 1;

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

	return 0;
4257 4258
}

4259
static int mem_cgroup_oom_notify_cb(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
4260 4261 4262
{
	struct mem_cgroup_eventfd_list *ev;

4263 4264
	spin_lock(&memcg_oom_lock);

4265
	list_for_each_entry(ev, &memcg->oom_notify, list)
K
KAMEZAWA Hiroyuki 已提交
4266
		eventfd_signal(ev->eventfd, 1);
4267 4268

	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4269 4270 4271
	return 0;
}

4272
static void mem_cgroup_oom_notify(struct mem_cgroup *memcg)
K
KAMEZAWA Hiroyuki 已提交
4273
{
K
KAMEZAWA Hiroyuki 已提交
4274 4275
	struct mem_cgroup *iter;

4276
	for_each_mem_cgroup_tree(iter, memcg)
K
KAMEZAWA Hiroyuki 已提交
4277
		mem_cgroup_oom_notify_cb(iter);
K
KAMEZAWA Hiroyuki 已提交
4278 4279
}

4280
static int __mem_cgroup_usage_register_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
4281
	struct eventfd_ctx *eventfd, const char *args, enum res_type type)
4282
{
4283 4284
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
4285 4286
	unsigned long threshold;
	unsigned long usage;
4287
	int i, size, ret;
4288

4289
	ret = page_counter_memparse(args, "-1", &threshold);
4290 4291 4292 4293
	if (ret)
		return ret;

	mutex_lock(&memcg->thresholds_lock);
4294

4295
	if (type == _MEM) {
4296
		thresholds = &memcg->thresholds;
4297
		usage = mem_cgroup_usage(memcg, false);
4298
	} else if (type == _MEMSWAP) {
4299
		thresholds = &memcg->memsw_thresholds;
4300
		usage = mem_cgroup_usage(memcg, true);
4301
	} else
4302 4303 4304
		BUG();

	/* Check if a threshold crossed before adding a new one */
4305
	if (thresholds->primary)
4306 4307
		__mem_cgroup_threshold(memcg, type == _MEMSWAP);

4308
	size = thresholds->primary ? thresholds->primary->size + 1 : 1;
4309 4310

	/* Allocate memory for new array of thresholds */
4311
	new = kmalloc(struct_size(new, entries, size), GFP_KERNEL);
4312
	if (!new) {
4313 4314 4315
		ret = -ENOMEM;
		goto unlock;
	}
4316
	new->size = size;
4317 4318

	/* Copy thresholds (if any) to new array */
4319 4320 4321
	if (thresholds->primary)
		memcpy(new->entries, thresholds->primary->entries,
		       flex_array_size(new, entries, size - 1));
4322

4323
	/* Add new threshold */
4324 4325
	new->entries[size - 1].eventfd = eventfd;
	new->entries[size - 1].threshold = threshold;
4326 4327

	/* Sort thresholds. Registering of new threshold isn't time-critical */
4328
	sort(new->entries, size, sizeof(*new->entries),
4329 4330 4331
			compare_thresholds, NULL);

	/* Find current threshold */
4332
	new->current_threshold = -1;
4333
	for (i = 0; i < size; i++) {
4334
		if (new->entries[i].threshold <= usage) {
4335
			/*
4336 4337
			 * new->current_threshold will not be used until
			 * rcu_assign_pointer(), so it's safe to increment
4338 4339
			 * it here.
			 */
4340
			++new->current_threshold;
4341 4342
		} else
			break;
4343 4344
	}

4345 4346 4347 4348 4349
	/* Free old spare buffer and save old primary buffer as spare */
	kfree(thresholds->spare);
	thresholds->spare = thresholds->primary;

	rcu_assign_pointer(thresholds->primary, new);
4350

4351
	/* To be sure that nobody uses thresholds */
4352 4353 4354 4355 4356 4357 4358 4359
	synchronize_rcu();

unlock:
	mutex_unlock(&memcg->thresholds_lock);

	return ret;
}

4360
static int mem_cgroup_usage_register_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
4361 4362
	struct eventfd_ctx *eventfd, const char *args)
{
4363
	return __mem_cgroup_usage_register_event(memcg, eventfd, args, _MEM);
T
Tejun Heo 已提交
4364 4365
}

4366
static int memsw_cgroup_usage_register_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
4367 4368
	struct eventfd_ctx *eventfd, const char *args)
{
4369
	return __mem_cgroup_usage_register_event(memcg, eventfd, args, _MEMSWAP);
T
Tejun Heo 已提交
4370 4371
}

4372
static void __mem_cgroup_usage_unregister_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
4373
	struct eventfd_ctx *eventfd, enum res_type type)
4374
{
4375 4376
	struct mem_cgroup_thresholds *thresholds;
	struct mem_cgroup_threshold_ary *new;
4377
	unsigned long usage;
4378
	int i, j, size, entries;
4379 4380

	mutex_lock(&memcg->thresholds_lock);
4381 4382

	if (type == _MEM) {
4383
		thresholds = &memcg->thresholds;
4384
		usage = mem_cgroup_usage(memcg, false);
4385
	} else if (type == _MEMSWAP) {
4386
		thresholds = &memcg->memsw_thresholds;
4387
		usage = mem_cgroup_usage(memcg, true);
4388
	} else
4389 4390
		BUG();

4391 4392 4393
	if (!thresholds->primary)
		goto unlock;

4394 4395 4396 4397
	/* Check if a threshold crossed before removing */
	__mem_cgroup_threshold(memcg, type == _MEMSWAP);

	/* Calculate new number of threshold */
4398
	size = entries = 0;
4399 4400
	for (i = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd != eventfd)
4401
			size++;
4402 4403
		else
			entries++;
4404 4405
	}

4406
	new = thresholds->spare;
4407

4408 4409 4410 4411
	/* If no items related to eventfd have been cleared, nothing to do */
	if (!entries)
		goto unlock;

4412 4413
	/* Set thresholds array to NULL if we don't have thresholds */
	if (!size) {
4414 4415
		kfree(new);
		new = NULL;
4416
		goto swap_buffers;
4417 4418
	}

4419
	new->size = size;
4420 4421

	/* Copy thresholds and find current threshold */
4422 4423 4424
	new->current_threshold = -1;
	for (i = 0, j = 0; i < thresholds->primary->size; i++) {
		if (thresholds->primary->entries[i].eventfd == eventfd)
4425 4426
			continue;

4427
		new->entries[j] = thresholds->primary->entries[i];
4428
		if (new->entries[j].threshold <= usage) {
4429
			/*
4430
			 * new->current_threshold will not be used
4431 4432 4433
			 * until rcu_assign_pointer(), so it's safe to increment
			 * it here.
			 */
4434
			++new->current_threshold;
4435 4436 4437 4438
		}
		j++;
	}

4439
swap_buffers:
4440 4441
	/* Swap primary and spare array */
	thresholds->spare = thresholds->primary;
4442

4443
	rcu_assign_pointer(thresholds->primary, new);
4444

4445
	/* To be sure that nobody uses thresholds */
4446
	synchronize_rcu();
4447 4448 4449 4450 4451 4452

	/* If all events are unregistered, free the spare array */
	if (!new) {
		kfree(thresholds->spare);
		thresholds->spare = NULL;
	}
4453
unlock:
4454 4455
	mutex_unlock(&memcg->thresholds_lock);
}
4456

4457
static void mem_cgroup_usage_unregister_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
4458 4459
	struct eventfd_ctx *eventfd)
{
4460
	return __mem_cgroup_usage_unregister_event(memcg, eventfd, _MEM);
T
Tejun Heo 已提交
4461 4462
}

4463
static void memsw_cgroup_usage_unregister_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
4464 4465
	struct eventfd_ctx *eventfd)
{
4466
	return __mem_cgroup_usage_unregister_event(memcg, eventfd, _MEMSWAP);
T
Tejun Heo 已提交
4467 4468
}

4469
static int mem_cgroup_oom_register_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
4470
	struct eventfd_ctx *eventfd, const char *args)
K
KAMEZAWA Hiroyuki 已提交
4471 4472 4473 4474 4475 4476 4477
{
	struct mem_cgroup_eventfd_list *event;

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

4478
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4479 4480 4481 4482 4483

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

	/* already in OOM ? */
4484
	if (memcg->under_oom)
K
KAMEZAWA Hiroyuki 已提交
4485
		eventfd_signal(eventfd, 1);
4486
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4487 4488 4489 4490

	return 0;
}

4491
static void mem_cgroup_oom_unregister_event(struct mem_cgroup *memcg,
T
Tejun Heo 已提交
4492
	struct eventfd_ctx *eventfd)
K
KAMEZAWA Hiroyuki 已提交
4493 4494 4495
{
	struct mem_cgroup_eventfd_list *ev, *tmp;

4496
	spin_lock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4497

4498
	list_for_each_entry_safe(ev, tmp, &memcg->oom_notify, list) {
K
KAMEZAWA Hiroyuki 已提交
4499 4500 4501 4502 4503 4504
		if (ev->eventfd == eventfd) {
			list_del(&ev->list);
			kfree(ev);
		}
	}

4505
	spin_unlock(&memcg_oom_lock);
K
KAMEZAWA Hiroyuki 已提交
4506 4507
}

4508
static int mem_cgroup_oom_control_read(struct seq_file *sf, void *v)
4509
{
4510
	struct mem_cgroup *memcg = mem_cgroup_from_seq(sf);
4511

4512
	seq_printf(sf, "oom_kill_disable %d\n", memcg->oom_kill_disable);
4513
	seq_printf(sf, "under_oom %d\n", (bool)memcg->under_oom);
R
Roman Gushchin 已提交
4514 4515
	seq_printf(sf, "oom_kill %lu\n",
		   atomic_long_read(&memcg->memory_events[MEMCG_OOM_KILL]));
4516 4517 4518
	return 0;
}

4519
static int mem_cgroup_oom_control_write(struct cgroup_subsys_state *css,
4520 4521
	struct cftype *cft, u64 val)
{
4522
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
4523 4524

	/* cannot set to root cgroup and only 0 and 1 are allowed */
4525
	if (!css->parent || !((val == 0) || (val == 1)))
4526 4527
		return -EINVAL;

4528
	memcg->oom_kill_disable = val;
4529
	if (!val)
4530
		memcg_oom_recover(memcg);
4531

4532 4533 4534
	return 0;
}

4535 4536
#ifdef CONFIG_CGROUP_WRITEBACK

4537 4538
#include <trace/events/writeback.h>

T
Tejun Heo 已提交
4539 4540 4541 4542 4543 4544 4545 4546 4547 4548
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);
}

4549 4550 4551 4552 4553
static void memcg_wb_domain_size_changed(struct mem_cgroup *memcg)
{
	wb_domain_size_changed(&memcg->cgwb_domain);
}

T
Tejun Heo 已提交
4554 4555 4556 4557 4558 4559 4560 4561 4562 4563
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;
}

4564 4565 4566 4567 4568 4569
/*
 * idx can be of type enum memcg_stat_item or node_stat_item.
 * Keep in sync with memcg_exact_page().
 */
static unsigned long memcg_exact_page_state(struct mem_cgroup *memcg, int idx)
{
4570
	long x = atomic_long_read(&memcg->vmstats[idx]);
4571 4572 4573
	int cpu;

	for_each_online_cpu(cpu)
4574
		x += per_cpu_ptr(memcg->vmstats_percpu, cpu)->stat[idx];
4575 4576 4577 4578 4579
	if (x < 0)
		x = 0;
	return x;
}

4580 4581 4582
/**
 * mem_cgroup_wb_stats - retrieve writeback related stats from its memcg
 * @wb: bdi_writeback in question
4583 4584
 * @pfilepages: out parameter for number of file pages
 * @pheadroom: out parameter for number of allocatable pages according to memcg
4585 4586 4587
 * @pdirty: out parameter for number of dirty pages
 * @pwriteback: out parameter for number of pages under writeback
 *
4588 4589 4590
 * 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.
4591
 *
4592 4593 4594 4595 4596
 * 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.
4597
 */
4598 4599 4600
void mem_cgroup_wb_stats(struct bdi_writeback *wb, unsigned long *pfilepages,
			 unsigned long *pheadroom, unsigned long *pdirty,
			 unsigned long *pwriteback)
4601 4602 4603 4604
{
	struct mem_cgroup *memcg = mem_cgroup_from_css(wb->memcg_css);
	struct mem_cgroup *parent;

4605
	*pdirty = memcg_exact_page_state(memcg, NR_FILE_DIRTY);
4606

4607
	*pwriteback = memcg_exact_page_state(memcg, NR_WRITEBACK);
4608 4609
	*pfilepages = memcg_exact_page_state(memcg, NR_INACTIVE_FILE) +
			memcg_exact_page_state(memcg, NR_ACTIVE_FILE);
4610
	*pheadroom = PAGE_COUNTER_MAX;
4611 4612

	while ((parent = parent_mem_cgroup(memcg))) {
4613
		unsigned long ceiling = min(READ_ONCE(memcg->memory.max),
4614
					    READ_ONCE(memcg->memory.high));
4615 4616
		unsigned long used = page_counter_read(&memcg->memory);

4617
		*pheadroom = min(*pheadroom, ceiling - min(ceiling, used));
4618 4619 4620 4621
		memcg = parent;
	}
}

4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642 4643 4644 4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663 4664 4665 4666 4667 4668 4669 4670 4671 4672 4673 4674 4675
/*
 * Foreign dirty flushing
 *
 * There's an inherent mismatch between memcg and writeback.  The former
 * trackes ownership per-page while the latter per-inode.  This was a
 * deliberate design decision because honoring per-page ownership in the
 * writeback path is complicated, may lead to higher CPU and IO overheads
 * and deemed unnecessary given that write-sharing an inode across
 * different cgroups isn't a common use-case.
 *
 * Combined with inode majority-writer ownership switching, this works well
 * enough in most cases but there are some pathological cases.  For
 * example, let's say there are two cgroups A and B which keep writing to
 * different but confined parts of the same inode.  B owns the inode and
 * A's memory is limited far below B's.  A's dirty ratio can rise enough to
 * trigger balance_dirty_pages() sleeps but B's can be low enough to avoid
 * triggering background writeback.  A will be slowed down without a way to
 * make writeback of the dirty pages happen.
 *
 * Conditions like the above can lead to a cgroup getting repatedly and
 * severely throttled after making some progress after each
 * dirty_expire_interval while the underyling IO device is almost
 * completely idle.
 *
 * Solving this problem completely requires matching the ownership tracking
 * granularities between memcg and writeback in either direction.  However,
 * the more egregious behaviors can be avoided by simply remembering the
 * most recent foreign dirtying events and initiating remote flushes on
 * them when local writeback isn't enough to keep the memory clean enough.
 *
 * The following two functions implement such mechanism.  When a foreign
 * page - a page whose memcg and writeback ownerships don't match - is
 * dirtied, mem_cgroup_track_foreign_dirty() records the inode owning
 * bdi_writeback on the page owning memcg.  When balance_dirty_pages()
 * decides that the memcg needs to sleep due to high dirty ratio, it calls
 * mem_cgroup_flush_foreign() which queues writeback on the recorded
 * foreign bdi_writebacks which haven't expired.  Both the numbers of
 * recorded bdi_writebacks and concurrent in-flight foreign writebacks are
 * limited to MEMCG_CGWB_FRN_CNT.
 *
 * The mechanism only remembers IDs and doesn't hold any object references.
 * As being wrong occasionally doesn't matter, updates and accesses to the
 * records are lockless and racy.
 */
void mem_cgroup_track_foreign_dirty_slowpath(struct page *page,
					     struct bdi_writeback *wb)
{
	struct mem_cgroup *memcg = page->mem_cgroup;
	struct memcg_cgwb_frn *frn;
	u64 now = get_jiffies_64();
	u64 oldest_at = now;
	int oldest = -1;
	int i;

4676 4677
	trace_track_foreign_dirty(page, wb);

4678 4679 4680 4681 4682 4683 4684 4685 4686 4687 4688 4689 4690 4691 4692 4693 4694 4695 4696 4697 4698 4699 4700 4701 4702 4703 4704 4705 4706 4707 4708 4709 4710 4711 4712 4713 4714 4715 4716 4717 4718 4719 4720 4721 4722 4723 4724 4725 4726 4727 4728 4729 4730 4731 4732 4733 4734 4735 4736 4737
	/*
	 * Pick the slot to use.  If there is already a slot for @wb, keep
	 * using it.  If not replace the oldest one which isn't being
	 * written out.
	 */
	for (i = 0; i < MEMCG_CGWB_FRN_CNT; i++) {
		frn = &memcg->cgwb_frn[i];
		if (frn->bdi_id == wb->bdi->id &&
		    frn->memcg_id == wb->memcg_css->id)
			break;
		if (time_before64(frn->at, oldest_at) &&
		    atomic_read(&frn->done.cnt) == 1) {
			oldest = i;
			oldest_at = frn->at;
		}
	}

	if (i < MEMCG_CGWB_FRN_CNT) {
		/*
		 * Re-using an existing one.  Update timestamp lazily to
		 * avoid making the cacheline hot.  We want them to be
		 * reasonably up-to-date and significantly shorter than
		 * dirty_expire_interval as that's what expires the record.
		 * Use the shorter of 1s and dirty_expire_interval / 8.
		 */
		unsigned long update_intv =
			min_t(unsigned long, HZ,
			      msecs_to_jiffies(dirty_expire_interval * 10) / 8);

		if (time_before64(frn->at, now - update_intv))
			frn->at = now;
	} else if (oldest >= 0) {
		/* replace the oldest free one */
		frn = &memcg->cgwb_frn[oldest];
		frn->bdi_id = wb->bdi->id;
		frn->memcg_id = wb->memcg_css->id;
		frn->at = now;
	}
}

/* issue foreign writeback flushes for recorded foreign dirtying events */
void mem_cgroup_flush_foreign(struct bdi_writeback *wb)
{
	struct mem_cgroup *memcg = mem_cgroup_from_css(wb->memcg_css);
	unsigned long intv = msecs_to_jiffies(dirty_expire_interval * 10);
	u64 now = jiffies_64;
	int i;

	for (i = 0; i < MEMCG_CGWB_FRN_CNT; i++) {
		struct memcg_cgwb_frn *frn = &memcg->cgwb_frn[i];

		/*
		 * If the record is older than dirty_expire_interval,
		 * writeback on it has already started.  No need to kick it
		 * off again.  Also, don't start a new one if there's
		 * already one in flight.
		 */
		if (time_after64(frn->at, now - intv) &&
		    atomic_read(&frn->done.cnt) == 1) {
			frn->at = 0;
4738
			trace_flush_foreign(wb, frn->bdi_id, frn->memcg_id);
4739 4740 4741 4742 4743 4744 4745
			cgroup_writeback_by_id(frn->bdi_id, frn->memcg_id, 0,
					       WB_REASON_FOREIGN_FLUSH,
					       &frn->done);
		}
	}
}

T
Tejun Heo 已提交
4746 4747 4748 4749 4750 4751 4752 4753 4754 4755 4756
#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)
{
}

4757 4758 4759 4760
static void memcg_wb_domain_size_changed(struct mem_cgroup *memcg)
{
}

4761 4762
#endif	/* CONFIG_CGROUP_WRITEBACK */

4763 4764 4765 4766 4767 4768 4769 4770 4771 4772 4773 4774 4775
/*
 * 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.
 */

4776 4777 4778 4779 4780
/*
 * Unregister event and free resources.
 *
 * Gets called from workqueue.
 */
4781
static void memcg_event_remove(struct work_struct *work)
4782
{
4783 4784
	struct mem_cgroup_event *event =
		container_of(work, struct mem_cgroup_event, remove);
4785
	struct mem_cgroup *memcg = event->memcg;
4786 4787 4788

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

4789
	event->unregister_event(memcg, event->eventfd);
4790 4791 4792 4793 4794 4795

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

	eventfd_ctx_put(event->eventfd);
	kfree(event);
4796
	css_put(&memcg->css);
4797 4798 4799
}

/*
4800
 * Gets called on EPOLLHUP on eventfd when user closes it.
4801 4802 4803
 *
 * Called with wqh->lock held and interrupts disabled.
 */
4804
static int memcg_event_wake(wait_queue_entry_t *wait, unsigned mode,
4805
			    int sync, void *key)
4806
{
4807 4808
	struct mem_cgroup_event *event =
		container_of(wait, struct mem_cgroup_event, wait);
4809
	struct mem_cgroup *memcg = event->memcg;
A
Al Viro 已提交
4810
	__poll_t flags = key_to_poll(key);
4811

4812
	if (flags & EPOLLHUP) {
4813 4814 4815 4816 4817 4818 4819 4820 4821
		/*
		 * 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.
		 */
4822
		spin_lock(&memcg->event_list_lock);
4823 4824 4825 4826 4827 4828 4829 4830
		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);
		}
4831
		spin_unlock(&memcg->event_list_lock);
4832 4833 4834 4835 4836
	}

	return 0;
}

4837
static void memcg_event_ptable_queue_proc(struct file *file,
4838 4839
		wait_queue_head_t *wqh, poll_table *pt)
{
4840 4841
	struct mem_cgroup_event *event =
		container_of(pt, struct mem_cgroup_event, pt);
4842 4843 4844 4845 4846 4847

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

/*
4848 4849
 * DO NOT USE IN NEW FILES.
 *
4850 4851 4852 4853 4854
 * 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.
 */
4855 4856
static ssize_t memcg_write_event_control(struct kernfs_open_file *of,
					 char *buf, size_t nbytes, loff_t off)
4857
{
4858
	struct cgroup_subsys_state *css = of_css(of);
4859
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
4860
	struct mem_cgroup_event *event;
4861 4862 4863 4864
	struct cgroup_subsys_state *cfile_css;
	unsigned int efd, cfd;
	struct fd efile;
	struct fd cfile;
4865
	const char *name;
4866 4867 4868
	char *endp;
	int ret;

4869 4870 4871
	buf = strstrip(buf);

	efd = simple_strtoul(buf, &endp, 10);
4872 4873
	if (*endp != ' ')
		return -EINVAL;
4874
	buf = endp + 1;
4875

4876
	cfd = simple_strtoul(buf, &endp, 10);
4877 4878
	if ((*endp != ' ') && (*endp != '\0'))
		return -EINVAL;
4879
	buf = endp + 1;
4880 4881 4882 4883 4884

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

4885
	event->memcg = memcg;
4886
	INIT_LIST_HEAD(&event->list);
4887 4888 4889
	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);
4890 4891 4892 4893 4894 4895 4896 4897 4898 4899 4900 4901 4902 4903 4904 4905 4906 4907 4908 4909 4910 4911 4912 4913 4914

	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;

4915 4916 4917 4918 4919
	/*
	 * 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.
4920 4921
	 *
	 * DO NOT ADD NEW FILES.
4922
	 */
A
Al Viro 已提交
4923
	name = cfile.file->f_path.dentry->d_name.name;
4924 4925 4926 4927 4928 4929 4930 4931 4932 4933 4934

	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
4935 4936
		event->register_event = memsw_cgroup_usage_register_event;
		event->unregister_event = memsw_cgroup_usage_unregister_event;
4937 4938 4939 4940 4941
	} else {
		ret = -EINVAL;
		goto out_put_cfile;
	}

4942
	/*
4943 4944 4945
	 * 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.
4946
	 */
A
Al Viro 已提交
4947
	cfile_css = css_tryget_online_from_dir(cfile.file->f_path.dentry->d_parent,
4948
					       &memory_cgrp_subsys);
4949
	ret = -EINVAL;
4950
	if (IS_ERR(cfile_css))
4951
		goto out_put_cfile;
4952 4953
	if (cfile_css != css) {
		css_put(cfile_css);
4954
		goto out_put_cfile;
4955
	}
4956

4957
	ret = event->register_event(memcg, event->eventfd, buf);
4958 4959 4960
	if (ret)
		goto out_put_css;

4961
	vfs_poll(efile.file, &event->pt);
4962

4963 4964 4965
	spin_lock(&memcg->event_list_lock);
	list_add(&event->list, &memcg->event_list);
	spin_unlock(&memcg->event_list_lock);
4966 4967 4968 4969

	fdput(cfile);
	fdput(efile);

4970
	return nbytes;
4971 4972

out_put_css:
4973
	css_put(css);
4974 4975 4976 4977 4978 4979 4980 4981 4982 4983 4984 4985
out_put_cfile:
	fdput(cfile);
out_put_eventfd:
	eventfd_ctx_put(event->eventfd);
out_put_efile:
	fdput(efile);
out_kfree:
	kfree(event);

	return ret;
}

4986
static struct cftype mem_cgroup_legacy_files[] = {
B
Balbir Singh 已提交
4987
	{
4988
		.name = "usage_in_bytes",
4989
		.private = MEMFILE_PRIVATE(_MEM, RES_USAGE),
4990
		.read_u64 = mem_cgroup_read_u64,
B
Balbir Singh 已提交
4991
	},
4992 4993
	{
		.name = "max_usage_in_bytes",
4994
		.private = MEMFILE_PRIVATE(_MEM, RES_MAX_USAGE),
4995
		.write = mem_cgroup_reset,
4996
		.read_u64 = mem_cgroup_read_u64,
4997
	},
B
Balbir Singh 已提交
4998
	{
4999
		.name = "limit_in_bytes",
5000
		.private = MEMFILE_PRIVATE(_MEM, RES_LIMIT),
5001
		.write = mem_cgroup_write,
5002
		.read_u64 = mem_cgroup_read_u64,
B
Balbir Singh 已提交
5003
	},
5004 5005 5006
	{
		.name = "soft_limit_in_bytes",
		.private = MEMFILE_PRIVATE(_MEM, RES_SOFT_LIMIT),
5007
		.write = mem_cgroup_write,
5008
		.read_u64 = mem_cgroup_read_u64,
5009
	},
B
Balbir Singh 已提交
5010 5011
	{
		.name = "failcnt",
5012
		.private = MEMFILE_PRIVATE(_MEM, RES_FAILCNT),
5013
		.write = mem_cgroup_reset,
5014
		.read_u64 = mem_cgroup_read_u64,
B
Balbir Singh 已提交
5015
	},
5016 5017
	{
		.name = "stat",
5018
		.seq_show = memcg_stat_show,
5019
	},
5020 5021
	{
		.name = "force_empty",
5022
		.write = mem_cgroup_force_empty_write,
5023
	},
5024 5025 5026 5027 5028
	{
		.name = "use_hierarchy",
		.write_u64 = mem_cgroup_hierarchy_write,
		.read_u64 = mem_cgroup_hierarchy_read,
	},
5029
	{
5030
		.name = "cgroup.event_control",		/* XXX: for compat */
5031
		.write = memcg_write_event_control,
5032
		.flags = CFTYPE_NO_PREFIX | CFTYPE_WORLD_WRITABLE,
5033
	},
K
KOSAKI Motohiro 已提交
5034 5035 5036 5037 5038
	{
		.name = "swappiness",
		.read_u64 = mem_cgroup_swappiness_read,
		.write_u64 = mem_cgroup_swappiness_write,
	},
5039 5040 5041 5042 5043
	{
		.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
5044 5045
	{
		.name = "oom_control",
5046
		.seq_show = mem_cgroup_oom_control_read,
5047
		.write_u64 = mem_cgroup_oom_control_write,
K
KAMEZAWA Hiroyuki 已提交
5048 5049
		.private = MEMFILE_PRIVATE(_OOM_TYPE, OOM_CONTROL),
	},
5050 5051 5052
	{
		.name = "pressure_level",
	},
5053 5054 5055
#ifdef CONFIG_NUMA
	{
		.name = "numa_stat",
5056
		.seq_show = memcg_numa_stat_show,
5057 5058
	},
#endif
5059 5060 5061
	{
		.name = "kmem.limit_in_bytes",
		.private = MEMFILE_PRIVATE(_KMEM, RES_LIMIT),
5062
		.write = mem_cgroup_write,
5063
		.read_u64 = mem_cgroup_read_u64,
5064 5065 5066 5067
	},
	{
		.name = "kmem.usage_in_bytes",
		.private = MEMFILE_PRIVATE(_KMEM, RES_USAGE),
5068
		.read_u64 = mem_cgroup_read_u64,
5069 5070 5071 5072
	},
	{
		.name = "kmem.failcnt",
		.private = MEMFILE_PRIVATE(_KMEM, RES_FAILCNT),
5073
		.write = mem_cgroup_reset,
5074
		.read_u64 = mem_cgroup_read_u64,
5075 5076 5077 5078
	},
	{
		.name = "kmem.max_usage_in_bytes",
		.private = MEMFILE_PRIVATE(_KMEM, RES_MAX_USAGE),
5079
		.write = mem_cgroup_reset,
5080
		.read_u64 = mem_cgroup_read_u64,
5081
	},
5082 5083
#if defined(CONFIG_MEMCG_KMEM) && \
	(defined(CONFIG_SLAB) || defined(CONFIG_SLUB_DEBUG))
5084 5085
	{
		.name = "kmem.slabinfo",
5086
		.seq_show = memcg_slab_show,
5087 5088
	},
#endif
V
Vladimir Davydov 已提交
5089 5090 5091 5092 5093 5094 5095 5096 5097 5098 5099 5100 5101 5102 5103 5104 5105 5106 5107 5108 5109 5110 5111
	{
		.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,
	},
5112
	{ },	/* terminate */
5113
};
5114

5115 5116 5117 5118 5119 5120 5121 5122
/*
 * 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.
 *
5123
 * However, there usually are many references to the offline CSS after
5124 5125 5126 5127 5128 5129 5130 5131 5132 5133 5134 5135 5136 5137 5138 5139 5140
 * 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);

5141 5142 5143 5144 5145 5146 5147 5148
static void mem_cgroup_id_remove(struct mem_cgroup *memcg)
{
	if (memcg->id.id > 0) {
		idr_remove(&mem_cgroup_idr, memcg->id.id);
		memcg->id.id = 0;
	}
}

5149 5150
static void __maybe_unused mem_cgroup_id_get_many(struct mem_cgroup *memcg,
						  unsigned int n)
5151
{
5152
	refcount_add(n, &memcg->id.ref);
5153 5154
}

5155
static void mem_cgroup_id_put_many(struct mem_cgroup *memcg, unsigned int n)
5156
{
5157
	if (refcount_sub_and_test(n, &memcg->id.ref)) {
5158
		mem_cgroup_id_remove(memcg);
5159 5160 5161 5162 5163 5164

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

5165 5166 5167 5168 5169
static inline void mem_cgroup_id_put(struct mem_cgroup *memcg)
{
	mem_cgroup_id_put_many(memcg, 1);
}

5170 5171 5172 5173 5174 5175 5176 5177 5178 5179 5180 5181
/**
 * 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);
}

5182
static int alloc_mem_cgroup_per_node_info(struct mem_cgroup *memcg, int node)
5183 5184
{
	struct mem_cgroup_per_node *pn;
5185
	int tmp = node;
5186 5187 5188 5189 5190 5191 5192 5193
	/*
	 * 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.
	 */
5194 5195
	if (!node_state(node, N_NORMAL_MEMORY))
		tmp = -1;
5196
	pn = kzalloc_node(sizeof(*pn), GFP_KERNEL, tmp);
5197 5198
	if (!pn)
		return 1;
5199

5200 5201
	pn->lruvec_stat_local = alloc_percpu_gfp(struct lruvec_stat,
						 GFP_KERNEL_ACCOUNT);
5202 5203 5204 5205 5206
	if (!pn->lruvec_stat_local) {
		kfree(pn);
		return 1;
	}

5207 5208
	pn->lruvec_stat_cpu = alloc_percpu_gfp(struct lruvec_stat,
					       GFP_KERNEL_ACCOUNT);
5209
	if (!pn->lruvec_stat_cpu) {
5210
		free_percpu(pn->lruvec_stat_local);
5211 5212 5213 5214
		kfree(pn);
		return 1;
	}

5215 5216 5217 5218 5219
	lruvec_init(&pn->lruvec);
	pn->usage_in_excess = 0;
	pn->on_tree = false;
	pn->memcg = memcg;

5220
	memcg->nodeinfo[node] = pn;
5221 5222 5223
	return 0;
}

5224
static void free_mem_cgroup_per_node_info(struct mem_cgroup *memcg, int node)
5225
{
5226 5227
	struct mem_cgroup_per_node *pn = memcg->nodeinfo[node];

M
Michal Hocko 已提交
5228 5229 5230
	if (!pn)
		return;

5231
	free_percpu(pn->lruvec_stat_cpu);
5232
	free_percpu(pn->lruvec_stat_local);
5233
	kfree(pn);
5234 5235
}

5236
static void __mem_cgroup_free(struct mem_cgroup *memcg)
5237
{
5238
	int node;
5239

5240
	for_each_node(node)
5241
		free_mem_cgroup_per_node_info(memcg, node);
5242
	free_percpu(memcg->vmstats_percpu);
5243
	free_percpu(memcg->vmstats_local);
5244
	kfree(memcg);
5245
}
5246

5247 5248 5249
static void mem_cgroup_free(struct mem_cgroup *memcg)
{
	memcg_wb_domain_exit(memcg);
5250 5251 5252 5253
	/*
	 * Flush percpu vmstats and vmevents to guarantee the value correctness
	 * on parent's and all ancestor levels.
	 */
5254
	memcg_flush_percpu_vmstats(memcg);
5255
	memcg_flush_percpu_vmevents(memcg);
5256 5257 5258
	__mem_cgroup_free(memcg);
}

5259
static struct mem_cgroup *mem_cgroup_alloc(void)
B
Balbir Singh 已提交
5260
{
5261
	struct mem_cgroup *memcg;
5262
	unsigned int size;
5263
	int node;
5264
	int __maybe_unused i;
5265
	long error = -ENOMEM;
B
Balbir Singh 已提交
5266

5267 5268 5269 5270
	size = sizeof(struct mem_cgroup);
	size += nr_node_ids * sizeof(struct mem_cgroup_per_node *);

	memcg = kzalloc(size, GFP_KERNEL);
5271
	if (!memcg)
5272
		return ERR_PTR(error);
5273

5274 5275 5276
	memcg->id.id = idr_alloc(&mem_cgroup_idr, NULL,
				 1, MEM_CGROUP_ID_MAX,
				 GFP_KERNEL);
5277 5278
	if (memcg->id.id < 0) {
		error = memcg->id.id;
5279
		goto fail;
5280
	}
5281

5282 5283
	memcg->vmstats_local = alloc_percpu_gfp(struct memcg_vmstats_percpu,
						GFP_KERNEL_ACCOUNT);
5284 5285 5286
	if (!memcg->vmstats_local)
		goto fail;

5287 5288
	memcg->vmstats_percpu = alloc_percpu_gfp(struct memcg_vmstats_percpu,
						 GFP_KERNEL_ACCOUNT);
5289
	if (!memcg->vmstats_percpu)
5290
		goto fail;
5291

B
Bob Liu 已提交
5292
	for_each_node(node)
5293
		if (alloc_mem_cgroup_per_node_info(memcg, node))
5294
			goto fail;
5295

5296 5297
	if (memcg_wb_domain_init(memcg, GFP_KERNEL))
		goto fail;
5298

5299
	INIT_WORK(&memcg->high_work, high_work_func);
5300 5301 5302
	INIT_LIST_HEAD(&memcg->oom_notify);
	mutex_init(&memcg->thresholds_lock);
	spin_lock_init(&memcg->move_lock);
5303
	vmpressure_init(&memcg->vmpressure);
5304 5305
	INIT_LIST_HEAD(&memcg->event_list);
	spin_lock_init(&memcg->event_list_lock);
5306
	memcg->socket_pressure = jiffies;
5307
#ifdef CONFIG_MEMCG_KMEM
V
Vladimir Davydov 已提交
5308
	memcg->kmemcg_id = -1;
R
Roman Gushchin 已提交
5309
	INIT_LIST_HEAD(&memcg->objcg_list);
V
Vladimir Davydov 已提交
5310
#endif
5311 5312
#ifdef CONFIG_CGROUP_WRITEBACK
	INIT_LIST_HEAD(&memcg->cgwb_list);
5313 5314 5315
	for (i = 0; i < MEMCG_CGWB_FRN_CNT; i++)
		memcg->cgwb_frn[i].done =
			__WB_COMPLETION_INIT(&memcg_cgwb_frn_waitq);
5316 5317 5318 5319 5320
#endif
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
	spin_lock_init(&memcg->deferred_split_queue.split_queue_lock);
	INIT_LIST_HEAD(&memcg->deferred_split_queue.split_queue);
	memcg->deferred_split_queue.split_queue_len = 0;
5321
#endif
5322
	idr_replace(&mem_cgroup_idr, memcg, memcg->id.id);
5323 5324
	return memcg;
fail:
5325
	mem_cgroup_id_remove(memcg);
5326
	__mem_cgroup_free(memcg);
5327
	return ERR_PTR(error);
5328 5329
}

5330 5331
static struct cgroup_subsys_state * __ref
mem_cgroup_css_alloc(struct cgroup_subsys_state *parent_css)
5332
{
5333
	struct mem_cgroup *parent = mem_cgroup_from_css(parent_css);
5334
	struct mem_cgroup *memcg, *old_memcg;
5335
	long error = -ENOMEM;
5336

5337
	old_memcg = set_active_memcg(parent);
5338
	memcg = mem_cgroup_alloc();
5339
	set_active_memcg(old_memcg);
5340 5341
	if (IS_ERR(memcg))
		return ERR_CAST(memcg);
5342

5343
	page_counter_set_high(&memcg->memory, PAGE_COUNTER_MAX);
5344
	memcg->soft_limit = PAGE_COUNTER_MAX;
5345
	page_counter_set_high(&memcg->swap, PAGE_COUNTER_MAX);
5346 5347 5348 5349
	if (parent) {
		memcg->swappiness = mem_cgroup_swappiness(parent);
		memcg->oom_kill_disable = parent->oom_kill_disable;
	}
5350 5351 5352 5353 5354 5355
	if (!parent) {
		page_counter_init(&memcg->memory, NULL);
		page_counter_init(&memcg->swap, NULL);
		page_counter_init(&memcg->kmem, NULL);
		page_counter_init(&memcg->tcpmem, NULL);
	} else if (parent->use_hierarchy) {
5356
		memcg->use_hierarchy = true;
5357
		page_counter_init(&memcg->memory, &parent->memory);
5358
		page_counter_init(&memcg->swap, &parent->swap);
5359
		page_counter_init(&memcg->kmem, &parent->kmem);
5360
		page_counter_init(&memcg->tcpmem, &parent->tcpmem);
5361
	} else {
5362 5363 5364 5365
		page_counter_init(&memcg->memory, &root_mem_cgroup->memory);
		page_counter_init(&memcg->swap, &root_mem_cgroup->swap);
		page_counter_init(&memcg->kmem, &root_mem_cgroup->kmem);
		page_counter_init(&memcg->tcpmem, &root_mem_cgroup->tcpmem);
5366 5367 5368 5369 5370
		/*
		 * Deeper hierachy with use_hierarchy == false doesn't make
		 * much sense so let cgroup subsystem know about this
		 * unfortunate state in our controller.
		 */
5371
		if (parent != root_mem_cgroup)
5372
			memory_cgrp_subsys.broken_hierarchy = true;
5373
	}
5374

5375 5376 5377 5378 5379 5380
	/* The following stuff does not apply to the root */
	if (!parent) {
		root_mem_cgroup = memcg;
		return &memcg->css;
	}

5381
	error = memcg_online_kmem(memcg);
5382 5383
	if (error)
		goto fail;
5384

5385
	if (cgroup_subsys_on_dfl(memory_cgrp_subsys) && !cgroup_memory_nosocket)
5386
		static_branch_inc(&memcg_sockets_enabled_key);
5387

5388 5389
	return &memcg->css;
fail:
5390
	mem_cgroup_id_remove(memcg);
5391
	mem_cgroup_free(memcg);
5392
	return ERR_PTR(error);
5393 5394
}

5395
static int mem_cgroup_css_online(struct cgroup_subsys_state *css)
5396
{
5397 5398
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);

5399 5400 5401 5402 5403 5404 5405 5406 5407 5408
	/*
	 * A memcg must be visible for memcg_expand_shrinker_maps()
	 * by the time the maps are allocated. So, we allocate maps
	 * here, when for_each_mem_cgroup() can't skip it.
	 */
	if (memcg_alloc_shrinker_maps(memcg)) {
		mem_cgroup_id_remove(memcg);
		return -ENOMEM;
	}

5409
	/* Online state pins memcg ID, memcg ID pins CSS */
5410
	refcount_set(&memcg->id.ref, 1);
5411
	css_get(css);
5412
	return 0;
B
Balbir Singh 已提交
5413 5414
}

5415
static void mem_cgroup_css_offline(struct cgroup_subsys_state *css)
5416
{
5417
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
5418
	struct mem_cgroup_event *event, *tmp;
5419 5420 5421 5422 5423 5424

	/*
	 * Unregister events and notify userspace.
	 * Notify userspace about cgroup removing only after rmdir of cgroup
	 * directory to avoid race between userspace and kernelspace.
	 */
5425 5426
	spin_lock(&memcg->event_list_lock);
	list_for_each_entry_safe(event, tmp, &memcg->event_list, list) {
5427 5428 5429
		list_del_init(&event->list);
		schedule_work(&event->remove);
	}
5430
	spin_unlock(&memcg->event_list_lock);
5431

R
Roman Gushchin 已提交
5432
	page_counter_set_min(&memcg->memory, 0);
5433
	page_counter_set_low(&memcg->memory, 0);
5434

5435
	memcg_offline_kmem(memcg);
5436
	wb_memcg_offline(memcg);
5437

5438 5439
	drain_all_stock(memcg);

5440
	mem_cgroup_id_put(memcg);
5441 5442
}

5443 5444 5445 5446 5447 5448 5449
static void mem_cgroup_css_released(struct cgroup_subsys_state *css)
{
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);

	invalidate_reclaim_iterators(memcg);
}

5450
static void mem_cgroup_css_free(struct cgroup_subsys_state *css)
B
Balbir Singh 已提交
5451
{
5452
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);
5453
	int __maybe_unused i;
5454

5455 5456 5457 5458
#ifdef CONFIG_CGROUP_WRITEBACK
	for (i = 0; i < MEMCG_CGWB_FRN_CNT; i++)
		wb_wait_for_completion(&memcg->cgwb_frn[i].done);
#endif
5459
	if (cgroup_subsys_on_dfl(memory_cgrp_subsys) && !cgroup_memory_nosocket)
5460
		static_branch_dec(&memcg_sockets_enabled_key);
5461

5462
	if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && memcg->tcpmem_active)
V
Vladimir Davydov 已提交
5463
		static_branch_dec(&memcg_sockets_enabled_key);
5464

5465 5466 5467
	vmpressure_cleanup(&memcg->vmpressure);
	cancel_work_sync(&memcg->high_work);
	mem_cgroup_remove_from_trees(memcg);
5468
	memcg_free_shrinker_maps(memcg);
5469
	memcg_free_kmem(memcg);
5470
	mem_cgroup_free(memcg);
B
Balbir Singh 已提交
5471 5472
}

5473 5474 5475 5476 5477 5478 5479 5480 5481 5482 5483 5484 5485 5486 5487 5488 5489
/**
 * 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);

5490 5491 5492 5493
	page_counter_set_max(&memcg->memory, PAGE_COUNTER_MAX);
	page_counter_set_max(&memcg->swap, PAGE_COUNTER_MAX);
	page_counter_set_max(&memcg->kmem, PAGE_COUNTER_MAX);
	page_counter_set_max(&memcg->tcpmem, PAGE_COUNTER_MAX);
R
Roman Gushchin 已提交
5494
	page_counter_set_min(&memcg->memory, 0);
5495
	page_counter_set_low(&memcg->memory, 0);
5496
	page_counter_set_high(&memcg->memory, PAGE_COUNTER_MAX);
5497
	memcg->soft_limit = PAGE_COUNTER_MAX;
5498
	page_counter_set_high(&memcg->swap, PAGE_COUNTER_MAX);
5499
	memcg_wb_domain_size_changed(memcg);
5500 5501
}

5502
#ifdef CONFIG_MMU
5503
/* Handlers for move charge at task migration. */
5504
static int mem_cgroup_do_precharge(unsigned long count)
5505
{
5506
	int ret;
5507

5508 5509
	/* Try a single bulk charge without reclaim first, kswapd may wake */
	ret = try_charge(mc.to, GFP_KERNEL & ~__GFP_DIRECT_RECLAIM, count);
5510
	if (!ret) {
5511 5512 5513
		mc.precharge += count;
		return ret;
	}
5514

5515
	/* Try charges one by one with reclaim, but do not retry */
5516
	while (count--) {
5517
		ret = try_charge(mc.to, GFP_KERNEL | __GFP_NORETRY, 1);
5518 5519
		if (ret)
			return ret;
5520
		mc.precharge++;
5521
		cond_resched();
5522
	}
5523
	return 0;
5524 5525 5526 5527
}

union mc_target {
	struct page	*page;
5528
	swp_entry_t	ent;
5529 5530 5531
};

enum mc_target_type {
5532
	MC_TARGET_NONE = 0,
5533
	MC_TARGET_PAGE,
5534
	MC_TARGET_SWAP,
5535
	MC_TARGET_DEVICE,
5536 5537
};

D
Daisuke Nishimura 已提交
5538 5539
static struct page *mc_handle_present_pte(struct vm_area_struct *vma,
						unsigned long addr, pte_t ptent)
5540
{
5541
	struct page *page = vm_normal_page(vma, addr, ptent);
5542

D
Daisuke Nishimura 已提交
5543 5544 5545
	if (!page || !page_mapped(page))
		return NULL;
	if (PageAnon(page)) {
5546
		if (!(mc.flags & MOVE_ANON))
D
Daisuke Nishimura 已提交
5547
			return NULL;
5548 5549 5550 5551
	} else {
		if (!(mc.flags & MOVE_FILE))
			return NULL;
	}
D
Daisuke Nishimura 已提交
5552 5553 5554 5555 5556 5557
	if (!get_page_unless_zero(page))
		return NULL;

	return page;
}

5558
#if defined(CONFIG_SWAP) || defined(CONFIG_DEVICE_PRIVATE)
D
Daisuke Nishimura 已提交
5559
static struct page *mc_handle_swap_pte(struct vm_area_struct *vma,
5560
			pte_t ptent, swp_entry_t *entry)
D
Daisuke Nishimura 已提交
5561 5562 5563 5564
{
	struct page *page = NULL;
	swp_entry_t ent = pte_to_swp_entry(ptent);

5565
	if (!(mc.flags & MOVE_ANON))
D
Daisuke Nishimura 已提交
5566
		return NULL;
5567 5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578 5579 5580 5581 5582 5583

	/*
	 * Handle MEMORY_DEVICE_PRIVATE which are ZONE_DEVICE page belonging to
	 * a device and because they are not accessible by CPU they are store
	 * as special swap entry in the CPU page table.
	 */
	if (is_device_private_entry(ent)) {
		page = device_private_entry_to_page(ent);
		/*
		 * MEMORY_DEVICE_PRIVATE means ZONE_DEVICE page and which have
		 * a refcount of 1 when free (unlike normal page)
		 */
		if (!page_ref_add_unless(page, 1, 1))
			return NULL;
		return page;
	}

5584 5585 5586
	if (non_swap_entry(ent))
		return NULL;

5587 5588 5589 5590
	/*
	 * Because lookup_swap_cache() updates some statistics counter,
	 * we call find_get_page() with swapper_space directly.
	 */
5591
	page = find_get_page(swap_address_space(ent), swp_offset(ent));
5592
	entry->val = ent.val;
D
Daisuke Nishimura 已提交
5593 5594 5595

	return page;
}
5596 5597
#else
static struct page *mc_handle_swap_pte(struct vm_area_struct *vma,
5598
			pte_t ptent, swp_entry_t *entry)
5599 5600 5601 5602
{
	return NULL;
}
#endif
D
Daisuke Nishimura 已提交
5603

5604 5605 5606 5607 5608
static struct page *mc_handle_file_pte(struct vm_area_struct *vma,
			unsigned long addr, pte_t ptent, swp_entry_t *entry)
{
	if (!vma->vm_file) /* anonymous vma */
		return NULL;
5609
	if (!(mc.flags & MOVE_FILE))
5610 5611 5612
		return NULL;

	/* page is moved even if it's not RSS of this task(page-faulted). */
5613
	/* shmem/tmpfs may report page out on swap: account for that too. */
5614 5615
	return find_get_incore_page(vma->vm_file->f_mapping,
			linear_page_index(vma, addr));
5616 5617
}

5618 5619 5620
/**
 * mem_cgroup_move_account - move account of the page
 * @page: the page
5621
 * @compound: charge the page as compound or small page
5622 5623 5624
 * @from: mem_cgroup which the page is moved from.
 * @to:	mem_cgroup which the page is moved to. @from != @to.
 *
5625
 * The caller must make sure the page is not on LRU (isolate_page() is useful.)
5626 5627 5628 5629 5630
 *
 * 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,
5631
				   bool compound,
5632 5633 5634
				   struct mem_cgroup *from,
				   struct mem_cgroup *to)
{
5635 5636
	struct lruvec *from_vec, *to_vec;
	struct pglist_data *pgdat;
5637
	unsigned int nr_pages = compound ? thp_nr_pages(page) : 1;
5638 5639 5640 5641
	int ret;

	VM_BUG_ON(from == to);
	VM_BUG_ON_PAGE(PageLRU(page), page);
5642
	VM_BUG_ON(compound && !PageTransHuge(page));
5643 5644

	/*
5645
	 * Prevent mem_cgroup_migrate() from looking at
5646
	 * page->mem_cgroup of its source page while we change it.
5647
	 */
5648
	ret = -EBUSY;
5649 5650 5651 5652 5653 5654 5655
	if (!trylock_page(page))
		goto out;

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

5656
	pgdat = page_pgdat(page);
5657 5658
	from_vec = mem_cgroup_lruvec(from, pgdat);
	to_vec = mem_cgroup_lruvec(to, pgdat);
5659

5660
	lock_page_memcg(page);
5661

5662 5663 5664 5665
	if (PageAnon(page)) {
		if (page_mapped(page)) {
			__mod_lruvec_state(from_vec, NR_ANON_MAPPED, -nr_pages);
			__mod_lruvec_state(to_vec, NR_ANON_MAPPED, nr_pages);
5666
			if (PageTransHuge(page)) {
5667 5668
				__dec_lruvec_state(from_vec, NR_ANON_THPS);
				__inc_lruvec_state(to_vec, NR_ANON_THPS);
5669 5670
			}

5671 5672
		}
	} else {
5673 5674 5675 5676 5677 5678 5679 5680
		__mod_lruvec_state(from_vec, NR_FILE_PAGES, -nr_pages);
		__mod_lruvec_state(to_vec, NR_FILE_PAGES, nr_pages);

		if (PageSwapBacked(page)) {
			__mod_lruvec_state(from_vec, NR_SHMEM, -nr_pages);
			__mod_lruvec_state(to_vec, NR_SHMEM, nr_pages);
		}

5681 5682 5683 5684
		if (page_mapped(page)) {
			__mod_lruvec_state(from_vec, NR_FILE_MAPPED, -nr_pages);
			__mod_lruvec_state(to_vec, NR_FILE_MAPPED, nr_pages);
		}
5685

5686 5687
		if (PageDirty(page)) {
			struct address_space *mapping = page_mapping(page);
5688

5689
			if (mapping_can_writeback(mapping)) {
5690 5691 5692 5693 5694
				__mod_lruvec_state(from_vec, NR_FILE_DIRTY,
						   -nr_pages);
				__mod_lruvec_state(to_vec, NR_FILE_DIRTY,
						   nr_pages);
			}
5695 5696 5697
		}
	}

5698
	if (PageWriteback(page)) {
5699 5700
		__mod_lruvec_state(from_vec, NR_WRITEBACK, -nr_pages);
		__mod_lruvec_state(to_vec, NR_WRITEBACK, nr_pages);
5701 5702 5703
	}

	/*
5704 5705
	 * All state has been migrated, let's switch to the new memcg.
	 *
5706
	 * It is safe to change page->mem_cgroup here because the page
5707 5708 5709 5710 5711 5712 5713 5714
	 * is referenced, charged, isolated, and locked: we can't race
	 * with (un)charging, migration, LRU putback, or anything else
	 * that would rely on a stable page->mem_cgroup.
	 *
	 * Note that lock_page_memcg is a memcg lock, not a page lock,
	 * to save space. As soon as we switch page->mem_cgroup to a
	 * new memcg that isn't locked, the above state can change
	 * concurrently again. Make sure we're truly done with it.
5715
	 */
5716
	smp_mb();
5717

5718 5719 5720 5721
	css_get(&to->css);
	css_put(&from->css);

	page->mem_cgroup = to;
5722

5723
	__unlock_page_memcg(from);
5724 5725 5726 5727

	ret = 0;

	local_irq_disable();
5728
	mem_cgroup_charge_statistics(to, page, nr_pages);
5729
	memcg_check_events(to, page);
5730
	mem_cgroup_charge_statistics(from, page, -nr_pages);
5731 5732 5733 5734 5735 5736 5737 5738
	memcg_check_events(from, page);
	local_irq_enable();
out_unlock:
	unlock_page(page);
out:
	return ret;
}

5739 5740 5741 5742 5743 5744 5745 5746 5747 5748 5749 5750 5751 5752 5753
/**
 * 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.
5754 5755
 *   3(MC_TARGET_DEVICE): like MC_TARGET_PAGE  but page is MEMORY_DEVICE_PRIVATE
 *     (so ZONE_DEVICE page and thus not on the lru).
5756 5757 5758
 *     For now we such page is charge like a regular page would be as for all
 *     intent and purposes it is just special memory taking the place of a
 *     regular page.
5759 5760
 *
 *     See Documentations/vm/hmm.txt and include/linux/hmm.h
5761 5762 5763 5764
 *
 * Called with pte lock held.
 */

5765
static enum mc_target_type get_mctgt_type(struct vm_area_struct *vma,
D
Daisuke Nishimura 已提交
5766 5767 5768
		unsigned long addr, pte_t ptent, union mc_target *target)
{
	struct page *page = NULL;
5769
	enum mc_target_type ret = MC_TARGET_NONE;
D
Daisuke Nishimura 已提交
5770 5771 5772 5773 5774
	swp_entry_t ent = { .val = 0 };

	if (pte_present(ptent))
		page = mc_handle_present_pte(vma, addr, ptent);
	else if (is_swap_pte(ptent))
5775
		page = mc_handle_swap_pte(vma, ptent, &ent);
5776
	else if (pte_none(ptent))
5777
		page = mc_handle_file_pte(vma, addr, ptent, &ent);
D
Daisuke Nishimura 已提交
5778 5779

	if (!page && !ent.val)
5780
		return ret;
5781 5782
	if (page) {
		/*
5783
		 * Do only loose check w/o serialization.
5784
		 * mem_cgroup_move_account() checks the page is valid or
5785
		 * not under LRU exclusion.
5786
		 */
5787
		if (page->mem_cgroup == mc.from) {
5788
			ret = MC_TARGET_PAGE;
5789
			if (is_device_private_page(page))
5790
				ret = MC_TARGET_DEVICE;
5791 5792 5793 5794 5795 5796
			if (target)
				target->page = page;
		}
		if (!ret || !target)
			put_page(page);
	}
5797 5798 5799 5800 5801
	/*
	 * There is a swap entry and a page doesn't exist or isn't charged.
	 * But we cannot move a tail-page in a THP.
	 */
	if (ent.val && !ret && (!page || !PageTransCompound(page)) &&
L
Li Zefan 已提交
5802
	    mem_cgroup_id(mc.from) == lookup_swap_cgroup_id(ent)) {
5803 5804 5805
		ret = MC_TARGET_SWAP;
		if (target)
			target->ent = ent;
5806 5807 5808 5809
	}
	return ret;
}

5810 5811
#ifdef CONFIG_TRANSPARENT_HUGEPAGE
/*
5812 5813
 * We don't consider PMD mapped swapping or file mapped pages because THP does
 * not support them for now.
5814 5815 5816 5817 5818 5819 5820 5821
 * 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;

5822 5823 5824 5825 5826
	if (unlikely(is_swap_pmd(pmd))) {
		VM_BUG_ON(thp_migration_supported() &&
				  !is_pmd_migration_entry(pmd));
		return ret;
	}
5827
	page = pmd_page(pmd);
5828
	VM_BUG_ON_PAGE(!page || !PageHead(page), page);
5829
	if (!(mc.flags & MOVE_ANON))
5830
		return ret;
5831
	if (page->mem_cgroup == mc.from) {
5832 5833 5834 5835 5836 5837 5838 5839 5840 5841 5842 5843 5844 5845 5846 5847
		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

5848 5849 5850 5851
static int mem_cgroup_count_precharge_pte_range(pmd_t *pmd,
					unsigned long addr, unsigned long end,
					struct mm_walk *walk)
{
5852
	struct vm_area_struct *vma = walk->vma;
5853 5854 5855
	pte_t *pte;
	spinlock_t *ptl;

5856 5857
	ptl = pmd_trans_huge_lock(pmd, vma);
	if (ptl) {
5858 5859
		/*
		 * Note their can not be MC_TARGET_DEVICE for now as we do not
5860 5861
		 * support transparent huge page with MEMORY_DEVICE_PRIVATE but
		 * this might change.
5862
		 */
5863 5864
		if (get_mctgt_type_thp(vma, addr, *pmd, NULL) == MC_TARGET_PAGE)
			mc.precharge += HPAGE_PMD_NR;
5865
		spin_unlock(ptl);
5866
		return 0;
5867
	}
5868

5869 5870
	if (pmd_trans_unstable(pmd))
		return 0;
5871 5872
	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
	for (; addr != end; pte++, addr += PAGE_SIZE)
5873
		if (get_mctgt_type(vma, addr, *pte, NULL))
5874 5875 5876 5877
			mc.precharge++;	/* increment precharge temporarily */
	pte_unmap_unlock(pte - 1, ptl);
	cond_resched();

5878 5879 5880
	return 0;
}

5881 5882 5883 5884
static const struct mm_walk_ops precharge_walk_ops = {
	.pmd_entry	= mem_cgroup_count_precharge_pte_range,
};

5885 5886 5887 5888
static unsigned long mem_cgroup_count_precharge(struct mm_struct *mm)
{
	unsigned long precharge;

5889
	mmap_read_lock(mm);
5890
	walk_page_range(mm, 0, mm->highest_vm_end, &precharge_walk_ops, NULL);
5891
	mmap_read_unlock(mm);
5892 5893 5894 5895 5896 5897 5898 5899 5900

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

	return precharge;
}

static int mem_cgroup_precharge_mc(struct mm_struct *mm)
{
5901 5902 5903 5904 5905
	unsigned long precharge = mem_cgroup_count_precharge(mm);

	VM_BUG_ON(mc.moving_task);
	mc.moving_task = current;
	return mem_cgroup_do_precharge(precharge);
5906 5907
}

5908 5909
/* cancels all extra charges on mc.from and mc.to, and wakes up all waiters. */
static void __mem_cgroup_clear_mc(void)
5910
{
5911 5912 5913
	struct mem_cgroup *from = mc.from;
	struct mem_cgroup *to = mc.to;

5914
	/* we must uncharge all the leftover precharges from mc.to */
5915
	if (mc.precharge) {
5916
		cancel_charge(mc.to, mc.precharge);
5917 5918 5919 5920 5921 5922 5923
		mc.precharge = 0;
	}
	/*
	 * we didn't uncharge from mc.from at mem_cgroup_move_account(), so
	 * we must uncharge here.
	 */
	if (mc.moved_charge) {
5924
		cancel_charge(mc.from, mc.moved_charge);
5925
		mc.moved_charge = 0;
5926
	}
5927 5928 5929
	/* we must fixup refcnts and charges */
	if (mc.moved_swap) {
		/* uncharge swap account from the old cgroup */
5930
		if (!mem_cgroup_is_root(mc.from))
5931
			page_counter_uncharge(&mc.from->memsw, mc.moved_swap);
5932

5933 5934
		mem_cgroup_id_put_many(mc.from, mc.moved_swap);

5935
		/*
5936 5937
		 * we charged both to->memory and to->memsw, so we
		 * should uncharge to->memory.
5938
		 */
5939
		if (!mem_cgroup_is_root(mc.to))
5940 5941
			page_counter_uncharge(&mc.to->memory, mc.moved_swap);

5942 5943
		mc.moved_swap = 0;
	}
5944 5945 5946 5947 5948 5949 5950
	memcg_oom_recover(from);
	memcg_oom_recover(to);
	wake_up_all(&mc.waitq);
}

static void mem_cgroup_clear_mc(void)
{
5951 5952
	struct mm_struct *mm = mc.mm;

5953 5954 5955 5956 5957 5958
	/*
	 * we must clear moving_task before waking up waiters at the end of
	 * task migration.
	 */
	mc.moving_task = NULL;
	__mem_cgroup_clear_mc();
5959
	spin_lock(&mc.lock);
5960 5961
	mc.from = NULL;
	mc.to = NULL;
5962
	mc.mm = NULL;
5963
	spin_unlock(&mc.lock);
5964 5965

	mmput(mm);
5966 5967
}

5968
static int mem_cgroup_can_attach(struct cgroup_taskset *tset)
5969
{
5970
	struct cgroup_subsys_state *css;
5971
	struct mem_cgroup *memcg = NULL; /* unneeded init to make gcc happy */
5972
	struct mem_cgroup *from;
5973
	struct task_struct *leader, *p;
5974
	struct mm_struct *mm;
5975
	unsigned long move_flags;
5976
	int ret = 0;
5977

5978 5979
	/* charge immigration isn't supported on the default hierarchy */
	if (cgroup_subsys_on_dfl(memory_cgrp_subsys))
5980 5981
		return 0;

5982 5983 5984 5985 5986 5987 5988
	/*
	 * 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;
5989
	cgroup_taskset_for_each_leader(leader, css, tset) {
5990 5991
		WARN_ON_ONCE(p);
		p = leader;
5992
		memcg = mem_cgroup_from_css(css);
5993 5994 5995 5996
	}
	if (!p)
		return 0;

5997 5998 5999 6000 6001 6002 6003 6004 6005
	/*
	 * 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;

6006 6007 6008 6009 6010 6011 6012 6013 6014 6015 6016 6017 6018 6019 6020 6021
	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);
6022
		mc.mm = mm;
6023 6024 6025 6026 6027 6028 6029 6030 6031
		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();
6032 6033
	} else {
		mmput(mm);
6034 6035 6036 6037
	}
	return ret;
}

6038
static void mem_cgroup_cancel_attach(struct cgroup_taskset *tset)
6039
{
6040 6041
	if (mc.to)
		mem_cgroup_clear_mc();
6042 6043
}

6044 6045 6046
static int mem_cgroup_move_charge_pte_range(pmd_t *pmd,
				unsigned long addr, unsigned long end,
				struct mm_walk *walk)
6047
{
6048
	int ret = 0;
6049
	struct vm_area_struct *vma = walk->vma;
6050 6051
	pte_t *pte;
	spinlock_t *ptl;
6052 6053 6054
	enum mc_target_type target_type;
	union mc_target target;
	struct page *page;
6055

6056 6057
	ptl = pmd_trans_huge_lock(pmd, vma);
	if (ptl) {
6058
		if (mc.precharge < HPAGE_PMD_NR) {
6059
			spin_unlock(ptl);
6060 6061 6062 6063 6064 6065
			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)) {
6066
				if (!mem_cgroup_move_account(page, true,
6067
							     mc.from, mc.to)) {
6068 6069 6070 6071 6072 6073
					mc.precharge -= HPAGE_PMD_NR;
					mc.moved_charge += HPAGE_PMD_NR;
				}
				putback_lru_page(page);
			}
			put_page(page);
6074 6075 6076 6077 6078 6079 6080 6081
		} else if (target_type == MC_TARGET_DEVICE) {
			page = target.page;
			if (!mem_cgroup_move_account(page, true,
						     mc.from, mc.to)) {
				mc.precharge -= HPAGE_PMD_NR;
				mc.moved_charge += HPAGE_PMD_NR;
			}
			put_page(page);
6082
		}
6083
		spin_unlock(ptl);
6084
		return 0;
6085 6086
	}

6087 6088
	if (pmd_trans_unstable(pmd))
		return 0;
6089 6090 6091 6092
retry:
	pte = pte_offset_map_lock(vma->vm_mm, pmd, addr, &ptl);
	for (; addr != end; addr += PAGE_SIZE) {
		pte_t ptent = *(pte++);
6093
		bool device = false;
6094
		swp_entry_t ent;
6095 6096 6097 6098

		if (!mc.precharge)
			break;

6099
		switch (get_mctgt_type(vma, addr, ptent, &target)) {
6100 6101
		case MC_TARGET_DEVICE:
			device = true;
J
Joe Perches 已提交
6102
			fallthrough;
6103 6104
		case MC_TARGET_PAGE:
			page = target.page;
6105 6106 6107 6108 6109 6110 6111 6112
			/*
			 * 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;
6113
			if (!device && isolate_lru_page(page))
6114
				goto put;
6115 6116
			if (!mem_cgroup_move_account(page, false,
						mc.from, mc.to)) {
6117
				mc.precharge--;
6118 6119
				/* we uncharge from mc.from later. */
				mc.moved_charge++;
6120
			}
6121 6122
			if (!device)
				putback_lru_page(page);
6123
put:			/* get_mctgt_type() gets the page */
6124 6125
			put_page(page);
			break;
6126 6127
		case MC_TARGET_SWAP:
			ent = target.ent;
6128
			if (!mem_cgroup_move_swap_account(ent, mc.from, mc.to)) {
6129
				mc.precharge--;
6130 6131
				mem_cgroup_id_get_many(mc.to, 1);
				/* we fixup other refcnts and charges later. */
6132 6133
				mc.moved_swap++;
			}
6134
			break;
6135 6136 6137 6138 6139 6140 6141 6142 6143 6144 6145 6146 6147 6148
		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.
		 */
6149
		ret = mem_cgroup_do_precharge(1);
6150 6151 6152 6153 6154 6155 6156
		if (!ret)
			goto retry;
	}

	return ret;
}

6157 6158 6159 6160
static const struct mm_walk_ops charge_walk_ops = {
	.pmd_entry	= mem_cgroup_move_charge_pte_range,
};

6161
static void mem_cgroup_move_charge(void)
6162 6163
{
	lru_add_drain_all();
6164
	/*
6165 6166 6167
	 * 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.
6168 6169 6170
	 */
	atomic_inc(&mc.from->moving_account);
	synchronize_rcu();
6171
retry:
6172
	if (unlikely(!mmap_read_trylock(mc.mm))) {
6173
		/*
6174
		 * Someone who are holding the mmap_lock might be waiting in
6175 6176 6177 6178 6179 6180 6181 6182 6183
		 * 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;
	}
6184 6185 6186 6187
	/*
	 * When we have consumed all precharges and failed in doing
	 * additional charge, the page walk just aborts.
	 */
6188 6189
	walk_page_range(mc.mm, 0, mc.mm->highest_vm_end, &charge_walk_ops,
			NULL);
6190

6191
	mmap_read_unlock(mc.mm);
6192
	atomic_dec(&mc.from->moving_account);
6193 6194
}

6195
static void mem_cgroup_move_task(void)
B
Balbir Singh 已提交
6196
{
6197 6198
	if (mc.to) {
		mem_cgroup_move_charge();
6199
		mem_cgroup_clear_mc();
6200
	}
B
Balbir Singh 已提交
6201
}
6202
#else	/* !CONFIG_MMU */
6203
static int mem_cgroup_can_attach(struct cgroup_taskset *tset)
6204 6205 6206
{
	return 0;
}
6207
static void mem_cgroup_cancel_attach(struct cgroup_taskset *tset)
6208 6209
{
}
6210
static void mem_cgroup_move_task(void)
6211 6212 6213
{
}
#endif
B
Balbir Singh 已提交
6214

6215 6216
/*
 * Cgroup retains root cgroups across [un]mount cycles making it necessary
6217 6218
 * to verify whether we're attached to the default hierarchy on each mount
 * attempt.
6219
 */
6220
static void mem_cgroup_bind(struct cgroup_subsys_state *root_css)
6221 6222
{
	/*
6223
	 * use_hierarchy is forced on the default hierarchy.  cgroup core
6224 6225 6226
	 * guarantees that @root doesn't have any children, so turning it
	 * on for the root memcg is enough.
	 */
6227
	if (cgroup_subsys_on_dfl(memory_cgrp_subsys))
6228 6229 6230
		root_mem_cgroup->use_hierarchy = true;
	else
		root_mem_cgroup->use_hierarchy = false;
6231 6232
}

6233 6234 6235 6236 6237 6238 6239 6240 6241 6242
static int seq_puts_memcg_tunable(struct seq_file *m, unsigned long value)
{
	if (value == PAGE_COUNTER_MAX)
		seq_puts(m, "max\n");
	else
		seq_printf(m, "%llu\n", (u64)value * PAGE_SIZE);

	return 0;
}

6243 6244 6245
static u64 memory_current_read(struct cgroup_subsys_state *css,
			       struct cftype *cft)
{
6246 6247 6248
	struct mem_cgroup *memcg = mem_cgroup_from_css(css);

	return (u64)page_counter_read(&memcg->memory) * PAGE_SIZE;
6249 6250
}

R
Roman Gushchin 已提交
6251 6252
static int memory_min_show(struct seq_file *m, void *v)
{
6253 6254
	return seq_puts_memcg_tunable(m,
		READ_ONCE(mem_cgroup_from_seq(m)->memory.min));
R
Roman Gushchin 已提交
6255 6256 6257 6258 6259 6260 6261 6262 6263 6264 6265 6266 6267 6268 6269 6270 6271 6272 6273
}

static ssize_t memory_min_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 min;
	int err;

	buf = strstrip(buf);
	err = page_counter_memparse(buf, "max", &min);
	if (err)
		return err;

	page_counter_set_min(&memcg->memory, min);

	return nbytes;
}

6274 6275
static int memory_low_show(struct seq_file *m, void *v)
{
6276 6277
	return seq_puts_memcg_tunable(m,
		READ_ONCE(mem_cgroup_from_seq(m)->memory.low));
6278 6279 6280 6281 6282 6283 6284 6285 6286 6287
}

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);
6288
	err = page_counter_memparse(buf, "max", &low);
6289 6290 6291
	if (err)
		return err;

6292
	page_counter_set_low(&memcg->memory, low);
6293 6294 6295 6296 6297 6298

	return nbytes;
}

static int memory_high_show(struct seq_file *m, void *v)
{
6299 6300
	return seq_puts_memcg_tunable(m,
		READ_ONCE(mem_cgroup_from_seq(m)->memory.high));
6301 6302 6303 6304 6305 6306
}

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));
6307
	unsigned int nr_retries = MAX_RECLAIM_RETRIES;
6308
	bool drained = false;
6309 6310 6311 6312
	unsigned long high;
	int err;

	buf = strstrip(buf);
6313
	err = page_counter_memparse(buf, "max", &high);
6314 6315 6316
	if (err)
		return err;

6317 6318
	page_counter_set_high(&memcg->memory, high);

6319 6320 6321 6322 6323 6324 6325 6326 6327 6328 6329 6330 6331 6332 6333 6334 6335 6336 6337 6338 6339 6340
	for (;;) {
		unsigned long nr_pages = page_counter_read(&memcg->memory);
		unsigned long reclaimed;

		if (nr_pages <= high)
			break;

		if (signal_pending(current))
			break;

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

		reclaimed = try_to_free_mem_cgroup_pages(memcg, nr_pages - high,
							 GFP_KERNEL, true);

		if (!reclaimed && !nr_retries--)
			break;
	}
6341

6342
	memcg_wb_domain_size_changed(memcg);
6343 6344 6345 6346 6347
	return nbytes;
}

static int memory_max_show(struct seq_file *m, void *v)
{
6348 6349
	return seq_puts_memcg_tunable(m,
		READ_ONCE(mem_cgroup_from_seq(m)->memory.max));
6350 6351 6352 6353 6354 6355
}

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));
6356
	unsigned int nr_reclaims = MAX_RECLAIM_RETRIES;
6357
	bool drained = false;
6358 6359 6360 6361
	unsigned long max;
	int err;

	buf = strstrip(buf);
6362
	err = page_counter_memparse(buf, "max", &max);
6363 6364 6365
	if (err)
		return err;

6366
	xchg(&memcg->memory.max, max);
6367 6368 6369 6370 6371 6372 6373

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

		if (nr_pages <= max)
			break;

6374
		if (signal_pending(current))
6375 6376 6377 6378 6379 6380 6381 6382 6383 6384 6385 6386 6387 6388 6389
			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;
		}

6390
		memcg_memory_event(memcg, MEMCG_OOM);
6391 6392 6393
		if (!mem_cgroup_out_of_memory(memcg, GFP_KERNEL, 0))
			break;
	}
6394

6395
	memcg_wb_domain_size_changed(memcg);
6396 6397 6398
	return nbytes;
}

6399 6400 6401 6402 6403 6404 6405 6406 6407 6408
static void __memory_events_show(struct seq_file *m, atomic_long_t *events)
{
	seq_printf(m, "low %lu\n", atomic_long_read(&events[MEMCG_LOW]));
	seq_printf(m, "high %lu\n", atomic_long_read(&events[MEMCG_HIGH]));
	seq_printf(m, "max %lu\n", atomic_long_read(&events[MEMCG_MAX]));
	seq_printf(m, "oom %lu\n", atomic_long_read(&events[MEMCG_OOM]));
	seq_printf(m, "oom_kill %lu\n",
		   atomic_long_read(&events[MEMCG_OOM_KILL]));
}

6409 6410
static int memory_events_show(struct seq_file *m, void *v)
{
6411
	struct mem_cgroup *memcg = mem_cgroup_from_seq(m);
6412

6413 6414 6415 6416 6417 6418 6419
	__memory_events_show(m, memcg->memory_events);
	return 0;
}

static int memory_events_local_show(struct seq_file *m, void *v)
{
	struct mem_cgroup *memcg = mem_cgroup_from_seq(m);
6420

6421
	__memory_events_show(m, memcg->memory_events_local);
6422 6423 6424
	return 0;
}

6425 6426
static int memory_stat_show(struct seq_file *m, void *v)
{
6427
	struct mem_cgroup *memcg = mem_cgroup_from_seq(m);
6428
	char *buf;
6429

6430 6431 6432 6433 6434
	buf = memory_stat_format(memcg);
	if (!buf)
		return -ENOMEM;
	seq_puts(m, buf);
	kfree(buf);
6435 6436 6437
	return 0;
}

6438 6439 6440 6441 6442 6443 6444 6445 6446 6447 6448 6449 6450 6451 6452 6453 6454 6455 6456 6457 6458 6459 6460 6461 6462 6463 6464 6465 6466
#ifdef CONFIG_NUMA
static int memory_numa_stat_show(struct seq_file *m, void *v)
{
	int i;
	struct mem_cgroup *memcg = mem_cgroup_from_seq(m);

	for (i = 0; i < ARRAY_SIZE(memory_stats); i++) {
		int nid;

		if (memory_stats[i].idx >= NR_VM_NODE_STAT_ITEMS)
			continue;

		seq_printf(m, "%s", memory_stats[i].name);
		for_each_node_state(nid, N_MEMORY) {
			u64 size;
			struct lruvec *lruvec;

			lruvec = mem_cgroup_lruvec(memcg, NODE_DATA(nid));
			size = lruvec_page_state(lruvec, memory_stats[i].idx);
			size *= memory_stats[i].ratio;
			seq_printf(m, " N%d=%llu", nid, size);
		}
		seq_putc(m, '\n');
	}

	return 0;
}
#endif

6467 6468
static int memory_oom_group_show(struct seq_file *m, void *v)
{
6469
	struct mem_cgroup *memcg = mem_cgroup_from_seq(m);
6470 6471 6472 6473 6474 6475 6476 6477 6478 6479 6480 6481 6482 6483 6484 6485 6486 6487 6488 6489 6490 6491 6492 6493 6494 6495 6496 6497

	seq_printf(m, "%d\n", memcg->oom_group);

	return 0;
}

static ssize_t memory_oom_group_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));
	int ret, oom_group;

	buf = strstrip(buf);
	if (!buf)
		return -EINVAL;

	ret = kstrtoint(buf, 0, &oom_group);
	if (ret)
		return ret;

	if (oom_group != 0 && oom_group != 1)
		return -EINVAL;

	memcg->oom_group = oom_group;

	return nbytes;
}

6498 6499 6500
static struct cftype memory_files[] = {
	{
		.name = "current",
6501
		.flags = CFTYPE_NOT_ON_ROOT,
6502 6503
		.read_u64 = memory_current_read,
	},
R
Roman Gushchin 已提交
6504 6505 6506 6507 6508 6509
	{
		.name = "min",
		.flags = CFTYPE_NOT_ON_ROOT,
		.seq_show = memory_min_show,
		.write = memory_min_write,
	},
6510 6511 6512 6513 6514 6515 6516 6517 6518 6519 6520 6521 6522 6523 6524 6525 6526 6527 6528 6529 6530
	{
		.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,
6531
		.file_offset = offsetof(struct mem_cgroup, events_file),
6532 6533
		.seq_show = memory_events_show,
	},
6534 6535 6536 6537 6538 6539
	{
		.name = "events.local",
		.flags = CFTYPE_NOT_ON_ROOT,
		.file_offset = offsetof(struct mem_cgroup, events_local_file),
		.seq_show = memory_events_local_show,
	},
6540 6541 6542 6543
	{
		.name = "stat",
		.seq_show = memory_stat_show,
	},
6544 6545 6546 6547 6548 6549
#ifdef CONFIG_NUMA
	{
		.name = "numa_stat",
		.seq_show = memory_numa_stat_show,
	},
#endif
6550 6551 6552 6553 6554 6555
	{
		.name = "oom.group",
		.flags = CFTYPE_NOT_ON_ROOT | CFTYPE_NS_DELEGATABLE,
		.seq_show = memory_oom_group_show,
		.write = memory_oom_group_write,
	},
6556 6557 6558
	{ }	/* terminate */
};

6559
struct cgroup_subsys memory_cgrp_subsys = {
6560
	.css_alloc = mem_cgroup_css_alloc,
6561
	.css_online = mem_cgroup_css_online,
6562
	.css_offline = mem_cgroup_css_offline,
6563
	.css_released = mem_cgroup_css_released,
6564
	.css_free = mem_cgroup_css_free,
6565
	.css_reset = mem_cgroup_css_reset,
6566 6567
	.can_attach = mem_cgroup_can_attach,
	.cancel_attach = mem_cgroup_cancel_attach,
6568
	.post_attach = mem_cgroup_move_task,
6569
	.bind = mem_cgroup_bind,
6570 6571
	.dfl_cftypes = memory_files,
	.legacy_cftypes = mem_cgroup_legacy_files,
6572
	.early_init = 0,
B
Balbir Singh 已提交
6573
};
6574

6575 6576 6577 6578 6579 6580 6581 6582 6583 6584 6585 6586 6587 6588 6589 6590 6591 6592 6593 6594 6595 6596 6597 6598 6599 6600 6601 6602 6603 6604
/*
 * This function calculates an individual cgroup's effective
 * protection which is derived from its own memory.min/low, its
 * parent's and siblings' settings, as well as the actual memory
 * distribution in the tree.
 *
 * The following rules apply to the effective protection values:
 *
 * 1. At the first level of reclaim, effective protection is equal to
 *    the declared protection in memory.min and memory.low.
 *
 * 2. To enable safe delegation of the protection configuration, at
 *    subsequent levels the effective protection is capped to the
 *    parent's effective protection.
 *
 * 3. To make complex and dynamic subtrees easier to configure, the
 *    user is allowed to overcommit the declared protection at a given
 *    level. If that is the case, the parent's effective protection is
 *    distributed to the children in proportion to how much protection
 *    they have declared and how much of it they are utilizing.
 *
 *    This makes distribution proportional, but also work-conserving:
 *    if one cgroup claims much more protection than it uses memory,
 *    the unused remainder is available to its siblings.
 *
 * 4. Conversely, when the declared protection is undercommitted at a
 *    given level, the distribution of the larger parental protection
 *    budget is NOT proportional. A cgroup's protection from a sibling
 *    is capped to its own memory.min/low setting.
 *
6605 6606 6607 6608 6609 6610 6611 6612 6613 6614 6615 6616
 * 5. However, to allow protecting recursive subtrees from each other
 *    without having to declare each individual cgroup's fixed share
 *    of the ancestor's claim to protection, any unutilized -
 *    "floating" - protection from up the tree is distributed in
 *    proportion to each cgroup's *usage*. This makes the protection
 *    neutral wrt sibling cgroups and lets them compete freely over
 *    the shared parental protection budget, but it protects the
 *    subtree as a whole from neighboring subtrees.
 *
 * Note that 4. and 5. are not in conflict: 4. is about protecting
 * against immediate siblings whereas 5. is about protecting against
 * neighboring subtrees.
6617 6618
 */
static unsigned long effective_protection(unsigned long usage,
6619
					  unsigned long parent_usage,
6620 6621 6622 6623 6624
					  unsigned long setting,
					  unsigned long parent_effective,
					  unsigned long siblings_protected)
{
	unsigned long protected;
6625
	unsigned long ep;
6626 6627 6628 6629 6630 6631 6632 6633 6634 6635 6636 6637 6638 6639 6640 6641 6642 6643 6644 6645 6646 6647 6648 6649 6650 6651 6652 6653 6654 6655

	protected = min(usage, setting);
	/*
	 * If all cgroups at this level combined claim and use more
	 * protection then what the parent affords them, distribute
	 * shares in proportion to utilization.
	 *
	 * We are using actual utilization rather than the statically
	 * claimed protection in order to be work-conserving: claimed
	 * but unused protection is available to siblings that would
	 * otherwise get a smaller chunk than what they claimed.
	 */
	if (siblings_protected > parent_effective)
		return protected * parent_effective / siblings_protected;

	/*
	 * Ok, utilized protection of all children is within what the
	 * parent affords them, so we know whatever this child claims
	 * and utilizes is effectively protected.
	 *
	 * If there is unprotected usage beyond this value, reclaim
	 * will apply pressure in proportion to that amount.
	 *
	 * If there is unutilized protection, the cgroup will be fully
	 * shielded from reclaim, but we do return a smaller value for
	 * protection than what the group could enjoy in theory. This
	 * is okay. With the overcommit distribution above, effective
	 * protection is always dependent on how memory is actually
	 * consumed among the siblings anyway.
	 */
6656 6657 6658 6659 6660 6661 6662 6663 6664 6665 6666 6667 6668
	ep = protected;

	/*
	 * If the children aren't claiming (all of) the protection
	 * afforded to them by the parent, distribute the remainder in
	 * proportion to the (unprotected) memory of each cgroup. That
	 * way, cgroups that aren't explicitly prioritized wrt each
	 * other compete freely over the allowance, but they are
	 * collectively protected from neighboring trees.
	 *
	 * We're using unprotected memory for the weight so that if
	 * some cgroups DO claim explicit protection, we don't protect
	 * the same bytes twice.
6669 6670 6671 6672
	 *
	 * Check both usage and parent_usage against the respective
	 * protected values. One should imply the other, but they
	 * aren't read atomically - make sure the division is sane.
6673 6674 6675
	 */
	if (!(cgrp_dfl_root.flags & CGRP_ROOT_MEMORY_RECURSIVE_PROT))
		return ep;
6676 6677 6678
	if (parent_effective > siblings_protected &&
	    parent_usage > siblings_protected &&
	    usage > protected) {
6679 6680 6681 6682 6683 6684 6685 6686 6687 6688
		unsigned long unclaimed;

		unclaimed = parent_effective - siblings_protected;
		unclaimed *= usage - protected;
		unclaimed /= parent_usage - siblings_protected;

		ep += unclaimed;
	}

	return ep;
6689 6690
}

6691
/**
R
Roman Gushchin 已提交
6692
 * mem_cgroup_protected - check if memory consumption is in the normal range
6693
 * @root: the top ancestor of the sub-tree being checked
6694 6695
 * @memcg: the memory cgroup to check
 *
6696 6697
 * WARNING: This function is not stateless! It can only be used as part
 *          of a top-down tree iteration, not for isolated queries.
6698
 */
6699 6700
void mem_cgroup_calculate_protection(struct mem_cgroup *root,
				     struct mem_cgroup *memcg)
6701
{
6702
	unsigned long usage, parent_usage;
6703 6704
	struct mem_cgroup *parent;

6705
	if (mem_cgroup_disabled())
6706
		return;
6707

6708 6709
	if (!root)
		root = root_mem_cgroup;
6710 6711 6712 6713 6714 6715 6716 6717

	/*
	 * Effective values of the reclaim targets are ignored so they
	 * can be stale. Have a look at mem_cgroup_protection for more
	 * details.
	 * TODO: calculation should be more robust so that we do not need
	 * that special casing.
	 */
6718
	if (memcg == root)
6719
		return;
6720

6721
	usage = page_counter_read(&memcg->memory);
R
Roman Gushchin 已提交
6722
	if (!usage)
6723
		return;
R
Roman Gushchin 已提交
6724 6725

	parent = parent_mem_cgroup(memcg);
6726 6727
	/* No parent means a non-hierarchical mode on v1 memcg */
	if (!parent)
6728
		return;
6729

6730
	if (parent == root) {
6731
		memcg->memory.emin = READ_ONCE(memcg->memory.min);
6732
		memcg->memory.elow = READ_ONCE(memcg->memory.low);
6733
		return;
R
Roman Gushchin 已提交
6734 6735
	}

6736 6737
	parent_usage = page_counter_read(&parent->memory);

6738
	WRITE_ONCE(memcg->memory.emin, effective_protection(usage, parent_usage,
6739 6740
			READ_ONCE(memcg->memory.min),
			READ_ONCE(parent->memory.emin),
6741
			atomic_long_read(&parent->memory.children_min_usage)));
6742

6743
	WRITE_ONCE(memcg->memory.elow, effective_protection(usage, parent_usage,
6744 6745
			READ_ONCE(memcg->memory.low),
			READ_ONCE(parent->memory.elow),
6746
			atomic_long_read(&parent->memory.children_low_usage)));
6747 6748
}

6749
/**
6750
 * mem_cgroup_charge - charge a newly allocated page to a cgroup
6751 6752 6753 6754 6755 6756 6757
 * @page: page to charge
 * @mm: mm context of the victim
 * @gfp_mask: reclaim mode
 *
 * Try to charge @page to the memcg that @mm belongs to, reclaiming
 * pages according to @gfp_mask if necessary.
 *
6758
 * Returns 0 on success. Otherwise, an error code is returned.
6759
 */
6760
int mem_cgroup_charge(struct page *page, struct mm_struct *mm, gfp_t gfp_mask)
6761
{
6762
	unsigned int nr_pages = thp_nr_pages(page);
6763 6764 6765 6766 6767 6768 6769
	struct mem_cgroup *memcg = NULL;
	int ret = 0;

	if (mem_cgroup_disabled())
		goto out;

	if (PageSwapCache(page)) {
6770 6771 6772
		swp_entry_t ent = { .val = page_private(page), };
		unsigned short id;

6773 6774 6775
		/*
		 * Every swap fault against a single page tries to charge the
		 * page, bail as early as possible.  shmem_unuse() encounters
6776 6777
		 * already charged pages, too.  page->mem_cgroup is protected
		 * by the page lock, which serializes swap cache removal, which
6778 6779
		 * in turn serializes uncharging.
		 */
6780
		VM_BUG_ON_PAGE(!PageLocked(page), page);
6781
		if (compound_head(page)->mem_cgroup)
6782
			goto out;
6783

6784 6785 6786 6787 6788 6789
		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();
6790 6791 6792 6793 6794 6795
	}

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

	ret = try_charge(memcg, gfp_mask, nr_pages);
6796 6797
	if (ret)
		goto out_put;
6798

6799
	css_get(&memcg->css);
6800
	commit_charge(page, memcg);
6801 6802

	local_irq_disable();
6803
	mem_cgroup_charge_statistics(memcg, page, nr_pages);
6804 6805
	memcg_check_events(memcg, page);
	local_irq_enable();
6806

6807 6808 6809 6810 6811 6812 6813 6814 6815 6816 6817 6818 6819
	/*
	 * Cgroup1's unified memory+swap counter has been charged with the
	 * new swapcache page, finish the transfer by uncharging the swap
	 * slot. The swap slot would also get uncharged when it dies, but
	 * it can stick around indefinitely and we'd count the page twice
	 * the entire time.
	 *
	 * Cgroup2 has separate resource counters for memory and swap,
	 * so this is a non-issue here. Memory and swap charge lifetimes
	 * correspond 1:1 to page and swap slot lifetimes: we charge the
	 * page to memory here, and uncharge swap when the slot is freed.
	 */
	if (do_memsw_account() && PageSwapCache(page)) {
6820 6821 6822 6823 6824 6825
		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.
		 */
6826
		mem_cgroup_uncharge_swap(entry, nr_pages);
6827 6828
	}

6829 6830 6831 6832
out_put:
	css_put(&memcg->css);
out:
	return ret;
6833 6834
}

6835 6836
struct uncharge_gather {
	struct mem_cgroup *memcg;
6837
	unsigned long nr_pages;
6838 6839 6840 6841 6842 6843
	unsigned long pgpgout;
	unsigned long nr_kmem;
	struct page *dummy_page;
};

static inline void uncharge_gather_clear(struct uncharge_gather *ug)
6844
{
6845 6846 6847 6848 6849
	memset(ug, 0, sizeof(*ug));
}

static void uncharge_batch(const struct uncharge_gather *ug)
{
6850 6851
	unsigned long flags;

6852
	if (!mem_cgroup_is_root(ug->memcg)) {
6853
		page_counter_uncharge(&ug->memcg->memory, ug->nr_pages);
6854
		if (do_memsw_account())
6855
			page_counter_uncharge(&ug->memcg->memsw, ug->nr_pages);
6856 6857 6858
		if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && ug->nr_kmem)
			page_counter_uncharge(&ug->memcg->kmem, ug->nr_kmem);
		memcg_oom_recover(ug->memcg);
6859
	}
6860 6861

	local_irq_save(flags);
6862
	__count_memcg_events(ug->memcg, PGPGOUT, ug->pgpgout);
6863
	__this_cpu_add(ug->memcg->vmstats_percpu->nr_page_events, ug->nr_pages);
6864
	memcg_check_events(ug->memcg, ug->dummy_page);
6865
	local_irq_restore(flags);
6866 6867 6868

	/* drop reference from uncharge_page */
	css_put(&ug->memcg->css);
6869 6870 6871 6872
}

static void uncharge_page(struct page *page, struct uncharge_gather *ug)
{
6873 6874
	unsigned long nr_pages;

6875 6876 6877 6878 6879 6880 6881 6882 6883 6884 6885 6886 6887 6888 6889 6890 6891
	VM_BUG_ON_PAGE(PageLRU(page), page);

	if (!page->mem_cgroup)
		return;

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

	if (ug->memcg != page->mem_cgroup) {
		if (ug->memcg) {
			uncharge_batch(ug);
			uncharge_gather_clear(ug);
		}
		ug->memcg = page->mem_cgroup;
6892 6893 6894

		/* pairs with css_put in uncharge_batch */
		css_get(&ug->memcg->css);
6895 6896
	}

6897 6898
	nr_pages = compound_nr(page);
	ug->nr_pages += nr_pages;
6899

6900
	if (!PageKmemcg(page)) {
6901 6902
		ug->pgpgout++;
	} else {
6903
		ug->nr_kmem += nr_pages;
6904 6905 6906 6907 6908
		__ClearPageKmemcg(page);
	}

	ug->dummy_page = page;
	page->mem_cgroup = NULL;
6909
	css_put(&ug->memcg->css);
6910 6911 6912 6913
}

static void uncharge_list(struct list_head *page_list)
{
6914
	struct uncharge_gather ug;
6915
	struct list_head *next;
6916 6917

	uncharge_gather_clear(&ug);
6918

6919 6920 6921 6922
	/*
	 * Note that the list can be a single page->lru; hence the
	 * do-while loop instead of a simple list_for_each_entry().
	 */
6923 6924
	next = page_list->next;
	do {
6925 6926
		struct page *page;

6927 6928 6929
		page = list_entry(next, struct page, lru);
		next = page->lru.next;

6930
		uncharge_page(page, &ug);
6931 6932
	} while (next != page_list);

6933 6934
	if (ug.memcg)
		uncharge_batch(&ug);
6935 6936
}

6937 6938 6939 6940
/**
 * mem_cgroup_uncharge - uncharge a page
 * @page: page to uncharge
 *
6941
 * Uncharge a page previously charged with mem_cgroup_charge().
6942 6943 6944
 */
void mem_cgroup_uncharge(struct page *page)
{
6945 6946
	struct uncharge_gather ug;

6947 6948 6949
	if (mem_cgroup_disabled())
		return;

6950
	/* Don't touch page->lru of any random page, pre-check: */
6951
	if (!page->mem_cgroup)
6952 6953
		return;

6954 6955 6956
	uncharge_gather_clear(&ug);
	uncharge_page(page, &ug);
	uncharge_batch(&ug);
6957
}
6958

6959 6960 6961 6962 6963
/**
 * mem_cgroup_uncharge_list - uncharge a list of page
 * @page_list: list of pages to uncharge
 *
 * Uncharge a list of pages previously charged with
6964
 * mem_cgroup_charge().
6965 6966 6967 6968 6969
 */
void mem_cgroup_uncharge_list(struct list_head *page_list)
{
	if (mem_cgroup_disabled())
		return;
6970

6971 6972
	if (!list_empty(page_list))
		uncharge_list(page_list);
6973 6974 6975
}

/**
6976 6977 6978
 * mem_cgroup_migrate - charge a page's replacement
 * @oldpage: currently circulating page
 * @newpage: replacement page
6979
 *
6980 6981
 * Charge @newpage as a replacement page for @oldpage. @oldpage will
 * be uncharged upon free.
6982 6983 6984
 *
 * Both pages must be locked, @newpage->mapping must be set up.
 */
6985
void mem_cgroup_migrate(struct page *oldpage, struct page *newpage)
6986
{
6987
	struct mem_cgroup *memcg;
6988
	unsigned int nr_pages;
6989
	unsigned long flags;
6990 6991 6992 6993

	VM_BUG_ON_PAGE(!PageLocked(oldpage), oldpage);
	VM_BUG_ON_PAGE(!PageLocked(newpage), newpage);
	VM_BUG_ON_PAGE(PageAnon(oldpage) != PageAnon(newpage), newpage);
6994 6995
	VM_BUG_ON_PAGE(PageTransHuge(oldpage) != PageTransHuge(newpage),
		       newpage);
6996 6997 6998 6999 7000

	if (mem_cgroup_disabled())
		return;

	/* Page cache replacement: new page already charged? */
7001
	if (newpage->mem_cgroup)
7002 7003
		return;

7004
	/* Swapcache readahead pages can get replaced before being charged */
7005
	memcg = oldpage->mem_cgroup;
7006
	if (!memcg)
7007 7008
		return;

7009
	/* Force-charge the new page. The old one will be freed soon */
7010
	nr_pages = thp_nr_pages(newpage);
7011 7012 7013 7014

	page_counter_charge(&memcg->memory, nr_pages);
	if (do_memsw_account())
		page_counter_charge(&memcg->memsw, nr_pages);
7015

7016
	css_get(&memcg->css);
7017
	commit_charge(newpage, memcg);
7018

7019
	local_irq_save(flags);
7020
	mem_cgroup_charge_statistics(memcg, newpage, nr_pages);
7021
	memcg_check_events(memcg, newpage);
7022
	local_irq_restore(flags);
7023 7024
}

7025
DEFINE_STATIC_KEY_FALSE(memcg_sockets_enabled_key);
7026 7027
EXPORT_SYMBOL(memcg_sockets_enabled_key);

7028
void mem_cgroup_sk_alloc(struct sock *sk)
7029 7030 7031
{
	struct mem_cgroup *memcg;

7032 7033 7034
	if (!mem_cgroup_sockets_enabled)
		return;

7035 7036 7037 7038
	/* Do not associate the sock with unrelated interrupted task's memcg. */
	if (in_interrupt())
		return;

7039 7040
	rcu_read_lock();
	memcg = mem_cgroup_from_task(current);
7041 7042
	if (memcg == root_mem_cgroup)
		goto out;
7043
	if (!cgroup_subsys_on_dfl(memory_cgrp_subsys) && !memcg->tcpmem_active)
7044
		goto out;
S
Shakeel Butt 已提交
7045
	if (css_tryget(&memcg->css))
7046
		sk->sk_memcg = memcg;
7047
out:
7048 7049 7050
	rcu_read_unlock();
}

7051
void mem_cgroup_sk_free(struct sock *sk)
7052
{
7053 7054
	if (sk->sk_memcg)
		css_put(&sk->sk_memcg->css);
7055 7056 7057 7058 7059 7060 7061 7062 7063 7064 7065 7066
}

/**
 * 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)
{
7067
	gfp_t gfp_mask = GFP_KERNEL;
7068

7069
	if (!cgroup_subsys_on_dfl(memory_cgrp_subsys)) {
7070
		struct page_counter *fail;
7071

7072 7073
		if (page_counter_try_charge(&memcg->tcpmem, nr_pages, &fail)) {
			memcg->tcpmem_pressure = 0;
7074 7075
			return true;
		}
7076 7077
		page_counter_charge(&memcg->tcpmem, nr_pages);
		memcg->tcpmem_pressure = 1;
7078
		return false;
7079
	}
7080

7081 7082 7083 7084
	/* Don't block in the packet receive path */
	if (in_softirq())
		gfp_mask = GFP_NOWAIT;

7085
	mod_memcg_state(memcg, MEMCG_SOCK, nr_pages);
7086

7087 7088 7089 7090
	if (try_charge(memcg, gfp_mask, nr_pages) == 0)
		return true;

	try_charge(memcg, gfp_mask|__GFP_NOFAIL, nr_pages);
7091 7092 7093 7094 7095
	return false;
}

/**
 * mem_cgroup_uncharge_skmem - uncharge socket memory
M
Mike Rapoport 已提交
7096 7097
 * @memcg: memcg to uncharge
 * @nr_pages: number of pages to uncharge
7098 7099 7100
 */
void mem_cgroup_uncharge_skmem(struct mem_cgroup *memcg, unsigned int nr_pages)
{
7101
	if (!cgroup_subsys_on_dfl(memory_cgrp_subsys)) {
7102
		page_counter_uncharge(&memcg->tcpmem, nr_pages);
7103 7104
		return;
	}
7105

7106
	mod_memcg_state(memcg, MEMCG_SOCK, -nr_pages);
7107

7108
	refill_stock(memcg, nr_pages);
7109 7110
}

7111 7112 7113 7114 7115 7116 7117 7118 7119
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;
7120 7121
		if (!strcmp(token, "nokmem"))
			cgroup_memory_nokmem = true;
7122 7123 7124 7125
	}
	return 0;
}
__setup("cgroup.memory=", cgroup_memory);
7126

7127
/*
7128 7129
 * subsys_initcall() for memory controller.
 *
7130 7131 7132 7133
 * 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.
7134 7135 7136
 */
static int __init mem_cgroup_init(void)
{
7137 7138
	int cpu, node;

7139 7140
	cpuhp_setup_state_nocalls(CPUHP_MM_MEMCQ_DEAD, "mm/memctrl:dead", NULL,
				  memcg_hotplug_cpu_dead);
7141 7142 7143 7144 7145 7146 7147 7148 7149 7150 7151

	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);

7152
		rtpn->rb_root = RB_ROOT;
7153
		rtpn->rb_rightmost = NULL;
7154
		spin_lock_init(&rtpn->lock);
7155 7156 7157
		soft_limit_tree.rb_tree_per_node[node] = rtpn;
	}

7158 7159 7160
	return 0;
}
subsys_initcall(mem_cgroup_init);
7161 7162

#ifdef CONFIG_MEMCG_SWAP
7163 7164
static struct mem_cgroup *mem_cgroup_id_get_online(struct mem_cgroup *memcg)
{
7165
	while (!refcount_inc_not_zero(&memcg->id.ref)) {
7166 7167 7168 7169 7170 7171 7172 7173 7174 7175 7176 7177 7178 7179 7180
		/*
		 * 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;
}

7181 7182 7183 7184 7185 7186 7187 7188 7189
/**
 * 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)
{
7190
	struct mem_cgroup *memcg, *swap_memcg;
7191
	unsigned int nr_entries;
7192 7193 7194 7195 7196
	unsigned short oldid;

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

7197
	if (cgroup_subsys_on_dfl(memory_cgrp_subsys))
7198 7199 7200 7201 7202 7203 7204 7205
		return;

	memcg = page->mem_cgroup;

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

7206 7207 7208 7209 7210 7211
	/*
	 * 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);
7212
	nr_entries = thp_nr_pages(page);
7213 7214 7215 7216 7217
	/* Get references for the tail pages, too */
	if (nr_entries > 1)
		mem_cgroup_id_get_many(swap_memcg, nr_entries - 1);
	oldid = swap_cgroup_record(entry, mem_cgroup_id(swap_memcg),
				   nr_entries);
7218
	VM_BUG_ON_PAGE(oldid, page);
7219
	mod_memcg_state(swap_memcg, MEMCG_SWAP, nr_entries);
7220 7221 7222 7223

	page->mem_cgroup = NULL;

	if (!mem_cgroup_is_root(memcg))
7224
		page_counter_uncharge(&memcg->memory, nr_entries);
7225

7226
	if (!cgroup_memory_noswap && memcg != swap_memcg) {
7227
		if (!mem_cgroup_is_root(swap_memcg))
7228 7229
			page_counter_charge(&swap_memcg->memsw, nr_entries);
		page_counter_uncharge(&memcg->memsw, nr_entries);
7230 7231
	}

7232 7233
	/*
	 * Interrupts should be disabled here because the caller holds the
M
Matthew Wilcox 已提交
7234
	 * i_pages lock which is taken with interrupts-off. It is
7235
	 * important here to have the interrupts disabled because it is the
M
Matthew Wilcox 已提交
7236
	 * only synchronisation we have for updating the per-CPU variables.
7237 7238
	 */
	VM_BUG_ON(!irqs_disabled());
7239
	mem_cgroup_charge_statistics(memcg, page, -nr_entries);
7240
	memcg_check_events(memcg, page);
7241

7242
	css_put(&memcg->css);
7243 7244
}

7245 7246
/**
 * mem_cgroup_try_charge_swap - try charging swap space for a page
7247 7248 7249
 * @page: page being added to swap
 * @entry: swap entry to charge
 *
7250
 * Try to charge @page's memcg for the swap space at @entry.
7251 7252 7253 7254 7255
 *
 * Returns 0 on success, -ENOMEM on failure.
 */
int mem_cgroup_try_charge_swap(struct page *page, swp_entry_t entry)
{
7256
	unsigned int nr_pages = thp_nr_pages(page);
7257
	struct page_counter *counter;
7258
	struct mem_cgroup *memcg;
7259 7260
	unsigned short oldid;

7261
	if (!cgroup_subsys_on_dfl(memory_cgrp_subsys))
7262 7263 7264 7265 7266 7267 7268 7269
		return 0;

	memcg = page->mem_cgroup;

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

7270 7271
	if (!entry.val) {
		memcg_memory_event(memcg, MEMCG_SWAP_FAIL);
7272
		return 0;
7273
	}
7274

7275 7276
	memcg = mem_cgroup_id_get_online(memcg);

7277
	if (!cgroup_memory_noswap && !mem_cgroup_is_root(memcg) &&
7278
	    !page_counter_try_charge(&memcg->swap, nr_pages, &counter)) {
7279 7280
		memcg_memory_event(memcg, MEMCG_SWAP_MAX);
		memcg_memory_event(memcg, MEMCG_SWAP_FAIL);
7281
		mem_cgroup_id_put(memcg);
7282
		return -ENOMEM;
7283
	}
7284

7285 7286 7287 7288
	/* 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);
7289
	VM_BUG_ON_PAGE(oldid, page);
7290
	mod_memcg_state(memcg, MEMCG_SWAP, nr_pages);
7291 7292 7293 7294

	return 0;
}

7295
/**
7296
 * mem_cgroup_uncharge_swap - uncharge swap space
7297
 * @entry: swap entry to uncharge
7298
 * @nr_pages: the amount of swap space to uncharge
7299
 */
7300
void mem_cgroup_uncharge_swap(swp_entry_t entry, unsigned int nr_pages)
7301 7302 7303 7304
{
	struct mem_cgroup *memcg;
	unsigned short id;

7305
	id = swap_cgroup_record(entry, 0, nr_pages);
7306
	rcu_read_lock();
7307
	memcg = mem_cgroup_from_id(id);
7308
	if (memcg) {
7309
		if (!cgroup_memory_noswap && !mem_cgroup_is_root(memcg)) {
7310
			if (cgroup_subsys_on_dfl(memory_cgrp_subsys))
7311
				page_counter_uncharge(&memcg->swap, nr_pages);
7312
			else
7313
				page_counter_uncharge(&memcg->memsw, nr_pages);
7314
		}
7315
		mod_memcg_state(memcg, MEMCG_SWAP, -nr_pages);
7316
		mem_cgroup_id_put_many(memcg, nr_pages);
7317 7318 7319 7320
	}
	rcu_read_unlock();
}

7321 7322 7323 7324
long mem_cgroup_get_nr_swap_pages(struct mem_cgroup *memcg)
{
	long nr_swap_pages = get_nr_swap_pages();

7325
	if (cgroup_memory_noswap || !cgroup_subsys_on_dfl(memory_cgrp_subsys))
7326 7327 7328
		return nr_swap_pages;
	for (; memcg != root_mem_cgroup; memcg = parent_mem_cgroup(memcg))
		nr_swap_pages = min_t(long, nr_swap_pages,
7329
				      READ_ONCE(memcg->swap.max) -
7330 7331 7332 7333
				      page_counter_read(&memcg->swap));
	return nr_swap_pages;
}

7334 7335 7336 7337 7338 7339 7340 7341
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;
7342
	if (cgroup_memory_noswap || !cgroup_subsys_on_dfl(memory_cgrp_subsys))
7343 7344 7345 7346 7347 7348
		return false;

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

7349 7350 7351 7352 7353
	for (; memcg != root_mem_cgroup; memcg = parent_mem_cgroup(memcg)) {
		unsigned long usage = page_counter_read(&memcg->swap);

		if (usage * 2 >= READ_ONCE(memcg->swap.high) ||
		    usage * 2 >= READ_ONCE(memcg->swap.max))
7354
			return true;
7355
	}
7356 7357 7358 7359

	return false;
}

7360
static int __init setup_swap_account(char *s)
7361 7362
{
	if (!strcmp(s, "1"))
7363
		cgroup_memory_noswap = 0;
7364
	else if (!strcmp(s, "0"))
7365
		cgroup_memory_noswap = 1;
7366 7367
	return 1;
}
7368
__setup("swapaccount=", setup_swap_account);
7369

7370 7371 7372 7373 7374 7375 7376 7377
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;
}

7378 7379 7380 7381 7382 7383 7384 7385 7386 7387 7388 7389 7390 7391 7392 7393 7394 7395 7396 7397 7398 7399 7400
static int swap_high_show(struct seq_file *m, void *v)
{
	return seq_puts_memcg_tunable(m,
		READ_ONCE(mem_cgroup_from_seq(m)->swap.high));
}

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

	buf = strstrip(buf);
	err = page_counter_memparse(buf, "max", &high);
	if (err)
		return err;

	page_counter_set_high(&memcg->swap, high);

	return nbytes;
}

7401 7402
static int swap_max_show(struct seq_file *m, void *v)
{
7403 7404
	return seq_puts_memcg_tunable(m,
		READ_ONCE(mem_cgroup_from_seq(m)->swap.max));
7405 7406 7407 7408 7409 7410 7411 7412 7413 7414 7415 7416 7417 7418
}

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;

7419
	xchg(&memcg->swap.max, max);
7420 7421 7422 7423

	return nbytes;
}

7424 7425
static int swap_events_show(struct seq_file *m, void *v)
{
7426
	struct mem_cgroup *memcg = mem_cgroup_from_seq(m);
7427

7428 7429
	seq_printf(m, "high %lu\n",
		   atomic_long_read(&memcg->memory_events[MEMCG_SWAP_HIGH]));
7430 7431 7432 7433 7434 7435 7436 7437
	seq_printf(m, "max %lu\n",
		   atomic_long_read(&memcg->memory_events[MEMCG_SWAP_MAX]));
	seq_printf(m, "fail %lu\n",
		   atomic_long_read(&memcg->memory_events[MEMCG_SWAP_FAIL]));

	return 0;
}

7438 7439 7440 7441 7442 7443
static struct cftype swap_files[] = {
	{
		.name = "swap.current",
		.flags = CFTYPE_NOT_ON_ROOT,
		.read_u64 = swap_current_read,
	},
7444 7445 7446 7447 7448 7449
	{
		.name = "swap.high",
		.flags = CFTYPE_NOT_ON_ROOT,
		.seq_show = swap_high_show,
		.write = swap_high_write,
	},
7450 7451 7452 7453 7454 7455
	{
		.name = "swap.max",
		.flags = CFTYPE_NOT_ON_ROOT,
		.seq_show = swap_max_show,
		.write = swap_max_write,
	},
7456 7457 7458 7459 7460 7461
	{
		.name = "swap.events",
		.flags = CFTYPE_NOT_ON_ROOT,
		.file_offset = offsetof(struct mem_cgroup, swap_events_file),
		.seq_show = swap_events_show,
	},
7462 7463 7464
	{ }	/* terminate */
};

7465
static struct cftype memsw_files[] = {
7466 7467 7468 7469 7470 7471 7472 7473 7474 7475 7476 7477 7478 7479 7480 7481 7482 7483 7484 7485 7486 7487 7488 7489 7490 7491
	{
		.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 */
};

7492 7493 7494 7495 7496 7497 7498
/*
 * If mem_cgroup_swap_init() is implemented as a subsys_initcall()
 * instead of a core_initcall(), this could mean cgroup_memory_noswap still
 * remains set to false even when memcg is disabled via "cgroup_disable=memory"
 * boot parameter. This may result in premature OOPS inside
 * mem_cgroup_get_nr_swap_pages() function in corner cases.
 */
7499 7500
static int __init mem_cgroup_swap_init(void)
{
7501 7502 7503 7504 7505
	/* No memory control -> no swap control */
	if (mem_cgroup_disabled())
		cgroup_memory_noswap = true;

	if (cgroup_memory_noswap)
7506 7507 7508 7509 7510
		return 0;

	WARN_ON(cgroup_add_dfl_cftypes(&memory_cgrp_subsys, swap_files));
	WARN_ON(cgroup_add_legacy_cftypes(&memory_cgrp_subsys, memsw_files));

7511 7512
	return 0;
}
7513
core_initcall(mem_cgroup_swap_init);
7514 7515

#endif /* CONFIG_MEMCG_SWAP */