[linux.git] / mm / hugetlb_cgroup.c

/*
 *
 * Copyright IBM Corporation, 2012
 * Author Aneesh Kumar K.V <[email protected]>
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms of version 2.1 of the GNU Lesser General Public License
 * as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it would be useful, but
 * WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
 *
 */

#include <linux/cgroup.h>
#include <linux/page_counter.h>
#include <linux/slab.h>
#include <linux/hugetlb.h>
#include <linux/hugetlb_cgroup.h>

struct hugetlb_cgroup {
	struct cgroup_subsys_state css;
	/*
	 * the counter to account for hugepages from hugetlb.
	 */
	struct page_counter hugepage[HUGE_MAX_HSTATE];
};

#define MEMFILE_PRIVATE(x, val)	(((x) << 16) | (val))
#define MEMFILE_IDX(val)	(((val) >> 16) & 0xffff)
#define MEMFILE_ATTR(val)	((val) & 0xffff)

static struct hugetlb_cgroup *root_h_cgroup __read_mostly;

static inline
struct hugetlb_cgroup *hugetlb_cgroup_from_css(struct cgroup_subsys_state *s)
{
	return s ? container_of(s, struct hugetlb_cgroup, css) : NULL;
}

static inline
struct hugetlb_cgroup *hugetlb_cgroup_from_task(struct task_struct *task)
{
	return hugetlb_cgroup_from_css(task_css(task, hugetlb_cgrp_id));
}

static inline bool hugetlb_cgroup_is_root(struct hugetlb_cgroup *h_cg)
{
	return (h_cg == root_h_cgroup);
}

static inline struct hugetlb_cgroup *
parent_hugetlb_cgroup(struct hugetlb_cgroup *h_cg)
{
	return hugetlb_cgroup_from_css(h_cg->css.parent);
}

static inline bool hugetlb_cgroup_have_usage(struct hugetlb_cgroup *h_cg)
{
	int idx;

	for (idx = 0; idx < hugetlb_max_hstate; idx++) {
		if (page_counter_read(&h_cg->hugepage[idx]))
			return true;
	}
	return false;
}

static void hugetlb_cgroup_init(struct hugetlb_cgroup *h_cgroup,
				struct hugetlb_cgroup *parent_h_cgroup)
{
	int idx;

	for (idx = 0; idx < HUGE_MAX_HSTATE; idx++) {
		struct page_counter *counter = &h_cgroup->hugepage[idx];
		struct page_counter *parent = NULL;
		unsigned long limit;
		int ret;

		if (parent_h_cgroup)
			parent = &parent_h_cgroup->hugepage[idx];
		page_counter_init(counter, parent);

		limit = round_down(PAGE_COUNTER_MAX,
				   1 << huge_page_order(&hstates[idx]));
		ret = page_counter_set_max(counter, limit);
		VM_BUG_ON(ret);
	}
}

static struct cgroup_subsys_state *
hugetlb_cgroup_css_alloc(struct cgroup_subsys_state *parent_css)
{
	struct hugetlb_cgroup *parent_h_cgroup = hugetlb_cgroup_from_css(parent_css);
	struct hugetlb_cgroup *h_cgroup;

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

	if (!parent_h_cgroup)
		root_h_cgroup = h_cgroup;

	hugetlb_cgroup_init(h_cgroup, parent_h_cgroup);
	return &h_cgroup->css;
}

static void hugetlb_cgroup_css_free(struct cgroup_subsys_state *css)
{
	struct hugetlb_cgroup *h_cgroup;

	h_cgroup = hugetlb_cgroup_from_css(css);
	kfree(h_cgroup);
}


/*
 * Should be called with hugetlb_lock held.
 * Since we are holding hugetlb_lock, pages cannot get moved from
 * active list or uncharged from the cgroup, So no need to get
 * page reference and test for page active here. This function
 * cannot fail.
 */
static void hugetlb_cgroup_move_parent(int idx, struct hugetlb_cgroup *h_cg,
				       struct page *page)
{
	unsigned int nr_pages;
	struct page_counter *counter;
	struct hugetlb_cgroup *page_hcg;
	struct hugetlb_cgroup *parent = parent_hugetlb_cgroup(h_cg);

	page_hcg = hugetlb_cgroup_from_page(page);
	/*
	 * We can have pages in active list without any cgroup
	 * ie, hugepage with less than 3 pages. We can safely
	 * ignore those pages.
	 */
	if (!page_hcg || page_hcg != h_cg)
		goto out;

	nr_pages = 1 << compound_order(page);
	if (!parent) {
		parent = root_h_cgroup;
		/* root has no limit */
		page_counter_charge(&parent->hugepage[idx], nr_pages);
	}
	counter = &h_cg->hugepage[idx];
	/* Take the pages off the local counter */
	page_counter_cancel(counter, nr_pages);

	set_hugetlb_cgroup(page, parent);
out:
	return;
}

/*
 * Force the hugetlb cgroup to empty the hugetlb resources by moving them to
 * the parent cgroup.
 */
static void hugetlb_cgroup_css_offline(struct cgroup_subsys_state *css)
{
	struct hugetlb_cgroup *h_cg = hugetlb_cgroup_from_css(css);
	struct hstate *h;
	struct page *page;
	int idx = 0;

	do {
		for_each_hstate(h) {
			spin_lock(&hugetlb_lock);
			list_for_each_entry(page, &h->hugepage_activelist, lru)
				hugetlb_cgroup_move_parent(idx, h_cg, page);

			spin_unlock(&hugetlb_lock);
			idx++;
		}
		cond_resched();
	} while (hugetlb_cgroup_have_usage(h_cg));
}

int hugetlb_cgroup_charge_cgroup(int idx, unsigned long nr_pages,
				 struct hugetlb_cgroup **ptr)
{
	int ret = 0;
	struct page_counter *counter;
	struct hugetlb_cgroup *h_cg = NULL;

	if (hugetlb_cgroup_disabled())
		goto done;
	/*
	 * We don't charge any cgroup if the compound page have less
	 * than 3 pages.
	 */
	if (huge_page_order(&hstates[idx]) < HUGETLB_CGROUP_MIN_ORDER)
		goto done;
again:
	rcu_read_lock();
	h_cg = hugetlb_cgroup_from_task(current);
	if (!css_tryget_online(&h_cg->css)) {
		rcu_read_unlock();
		goto again;
	}
	rcu_read_unlock();

	if (!page_counter_try_charge(&h_cg->hugepage[idx], nr_pages, &counter))
		ret = -ENOMEM;
	css_put(&h_cg->css);
done:
	*ptr = h_cg;
	return ret;
}

/* Should be called with hugetlb_lock held */
void hugetlb_cgroup_commit_charge(int idx, unsigned long nr_pages,
				  struct hugetlb_cgroup *h_cg,
				  struct page *page)
{
	if (hugetlb_cgroup_disabled() || !h_cg)
		return;

	set_hugetlb_cgroup(page, h_cg);
	return;
}

/*
 * Should be called with hugetlb_lock held
 */
void hugetlb_cgroup_uncharge_page(int idx, unsigned long nr_pages,
				  struct page *page)
{
	struct hugetlb_cgroup *h_cg;

	if (hugetlb_cgroup_disabled())
		return;
	lockdep_assert_held(&hugetlb_lock);
	h_cg = hugetlb_cgroup_from_page(page);
	if (unlikely(!h_cg))
		return;
	set_hugetlb_cgroup(page, NULL);
	page_counter_uncharge(&h_cg->hugepage[idx], nr_pages);
	return;
}

void hugetlb_cgroup_uncharge_cgroup(int idx, unsigned long nr_pages,
				    struct hugetlb_cgroup *h_cg)
{
	if (hugetlb_cgroup_disabled() || !h_cg)
		return;

	if (huge_page_order(&hstates[idx]) < HUGETLB_CGROUP_MIN_ORDER)
		return;

	page_counter_uncharge(&h_cg->hugepage[idx], nr_pages);
	return;
}

enum {
	RES_USAGE,
	RES_LIMIT,
	RES_MAX_USAGE,
	RES_FAILCNT,
};

static u64 hugetlb_cgroup_read_u64(struct cgroup_subsys_state *css,
				   struct cftype *cft)
{
	struct page_counter *counter;
	struct hugetlb_cgroup *h_cg = hugetlb_cgroup_from_css(css);

	counter = &h_cg->hugepage[MEMFILE_IDX(cft->private)];

	switch (MEMFILE_ATTR(cft->private)) {
	case RES_USAGE:
		return (u64)page_counter_read(counter) * PAGE_SIZE;
	case RES_LIMIT:
		return (u64)counter->max * PAGE_SIZE;
	case RES_MAX_USAGE:
		return (u64)counter->watermark * PAGE_SIZE;
	case RES_FAILCNT:
		return counter->failcnt;
	default:
		BUG();
	}
}

static DEFINE_MUTEX(hugetlb_limit_mutex);

static ssize_t hugetlb_cgroup_write(struct kernfs_open_file *of,
				    char *buf, size_t nbytes, loff_t off)
{
	int ret, idx;
	unsigned long nr_pages;
	struct hugetlb_cgroup *h_cg = hugetlb_cgroup_from_css(of_css(of));

	if (hugetlb_cgroup_is_root(h_cg)) /* Can't set limit on root */
		return -EINVAL;

	buf = strstrip(buf);
	ret = page_counter_memparse(buf, "-1", &nr_pages);
	if (ret)
		return ret;

	idx = MEMFILE_IDX(of_cft(of)->private);
	nr_pages = round_down(nr_pages, 1 << huge_page_order(&hstates[idx]));

	switch (MEMFILE_ATTR(of_cft(of)->private)) {
	case RES_LIMIT:
		mutex_lock(&hugetlb_limit_mutex);
		ret = page_counter_set_max(&h_cg->hugepage[idx], nr_pages);
		mutex_unlock(&hugetlb_limit_mutex);
		break;
	default:
		ret = -EINVAL;
		break;
	}
	return ret ?: nbytes;
}

static ssize_t hugetlb_cgroup_reset(struct kernfs_open_file *of,
				    char *buf, size_t nbytes, loff_t off)
{
	int ret = 0;
	struct page_counter *counter;
	struct hugetlb_cgroup *h_cg = hugetlb_cgroup_from_css(of_css(of));

	counter = &h_cg->hugepage[MEMFILE_IDX(of_cft(of)->private)];

	switch (MEMFILE_ATTR(of_cft(of)->private)) {
	case RES_MAX_USAGE:
		page_counter_reset_watermark(counter);
		break;
	case RES_FAILCNT:
		counter->failcnt = 0;
		break;
	default:
		ret = -EINVAL;
		break;
	}
	return ret ?: nbytes;
}

static char *mem_fmt(char *buf, int size, unsigned long hsize)
{
	if (hsize >= (1UL << 30))
		snprintf(buf, size, "%luGB", hsize >> 30);
	else if (hsize >= (1UL << 20))
		snprintf(buf, size, "%luMB", hsize >> 20);
	else
		snprintf(buf, size, "%luKB", hsize >> 10);
	return buf;
}

static void __init __hugetlb_cgroup_file_init(int idx)
{
	char buf[32];
	struct cftype *cft;
	struct hstate *h = &hstates[idx];

	/* format the size */
	mem_fmt(buf, 32, huge_page_size(h));

	/* Add the limit file */
	cft = &h->cgroup_files[0];
	snprintf(cft->name, MAX_CFTYPE_NAME, "%s.limit_in_bytes", buf);
	cft->private = MEMFILE_PRIVATE(idx, RES_LIMIT);
	cft->read_u64 = hugetlb_cgroup_read_u64;
	cft->write = hugetlb_cgroup_write;

	/* Add the usage file */
	cft = &h->cgroup_files[1];
	snprintf(cft->name, MAX_CFTYPE_NAME, "%s.usage_in_bytes", buf);
	cft->private = MEMFILE_PRIVATE(idx, RES_USAGE);
	cft->read_u64 = hugetlb_cgroup_read_u64;

	/* Add the MAX usage file */
	cft = &h->cgroup_files[2];
	snprintf(cft->name, MAX_CFTYPE_NAME, "%s.max_usage_in_bytes", buf);
	cft->private = MEMFILE_PRIVATE(idx, RES_MAX_USAGE);
	cft->write = hugetlb_cgroup_reset;
	cft->read_u64 = hugetlb_cgroup_read_u64;

	/* Add the failcntfile */
	cft = &h->cgroup_files[3];
	snprintf(cft->name, MAX_CFTYPE_NAME, "%s.failcnt", buf);
	cft->private  = MEMFILE_PRIVATE(idx, RES_FAILCNT);
	cft->write = hugetlb_cgroup_reset;
	cft->read_u64 = hugetlb_cgroup_read_u64;

	/* NULL terminate the last cft */
	cft = &h->cgroup_files[4];
	memset(cft, 0, sizeof(*cft));

	WARN_ON(cgroup_add_legacy_cftypes(&hugetlb_cgrp_subsys,
					  h->cgroup_files));
}

void __init hugetlb_cgroup_file_init(void)
{
	struct hstate *h;

	for_each_hstate(h) {
		/*
		 * Add cgroup control files only if the huge page consists
		 * of more than two normal pages. This is because we use
		 * page[2].private for storing cgroup details.
		 */
		if (huge_page_order(h) >= HUGETLB_CGROUP_MIN_ORDER)
			__hugetlb_cgroup_file_init(hstate_index(h));
	}
}

/*
 * hugetlb_lock will make sure a parallel cgroup rmdir won't happen
 * when we migrate hugepages
 */
void hugetlb_cgroup_migrate(struct page *oldhpage, struct page *newhpage)
{
	struct hugetlb_cgroup *h_cg;
	struct hstate *h = page_hstate(oldhpage);

	if (hugetlb_cgroup_disabled())
		return;

	VM_BUG_ON_PAGE(!PageHuge(oldhpage), oldhpage);
	spin_lock(&hugetlb_lock);
	h_cg = hugetlb_cgroup_from_page(oldhpage);
	set_hugetlb_cgroup(oldhpage, NULL);

	/* move the h_cg details to new cgroup */
	set_hugetlb_cgroup(newhpage, h_cg);
	list_move(&newhpage->lru, &h->hugepage_activelist);
	spin_unlock(&hugetlb_lock);
	return;
}

struct cgroup_subsys hugetlb_cgrp_subsys = {
	.css_alloc	= hugetlb_cgroup_css_alloc,
	.css_offline	= hugetlb_cgroup_css_offline,
	.css_free	= hugetlb_cgroup_css_free,
};
Commit	Line	Data
2bc64a20 AK	1	/*
	2	*
	3	* Copyright IBM Corporation, 2012
	4	* Author Aneesh Kumar K.V <[email protected]>
	5	*
	6	* This program is free software; you can redistribute it and/or modify it
	7	* under the terms of version 2.1 of the GNU Lesser General Public License
	8	* as published by the Free Software Foundation.
	9	*
	10	* This program is distributed in the hope that it would be useful, but
	11	* WITHOUT ANY WARRANTY; without even the implied warranty of
	12	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
	13	*
	14	*/
	15
	16	#include <linux/cgroup.h>
71f87bee	17	#include <linux/page_counter.h>
2bc64a20 AK	18	#include <linux/slab.h>
	19	#include <linux/hugetlb.h>
	20	#include <linux/hugetlb_cgroup.h>
	21
	22	struct hugetlb_cgroup {
	23	struct cgroup_subsys_state css;
	24	/*
	25	* the counter to account for hugepages from hugetlb.
	26	*/
71f87bee	27	struct page_counter hugepage[HUGE_MAX_HSTATE];
2bc64a20 AK	28	};
2bc64a20 AK	29
abb8206c AK	30	#define MEMFILE_PRIVATE(x, val) (((x) << 16) \| (val))
	31	#define MEMFILE_IDX(val) (((val) >> 16) & 0xffff)
	32	#define MEMFILE_ATTR(val) ((val) & 0xffff)
	33
2bc64a20 AK	34	static struct hugetlb_cgroup *root_h_cgroup __read_mostly;
	35
	36	static inline
	37	struct hugetlb_cgroup hugetlb_cgroup_from_css(struct cgroup_subsys_state s)
	38	{
a7c6d554	39	return s ? container_of(s, struct hugetlb_cgroup, css) : NULL;
2bc64a20 AK	40	}
2bc64a20 AK	41
2bc64a20 AK	42	static inline
	43	struct hugetlb_cgroup hugetlb_cgroup_from_task(struct task_struct task)
	44	{
073219e9	45	return hugetlb_cgroup_from_css(task_css(task, hugetlb_cgrp_id));
2bc64a20 AK	46	}
	47
	48	static inline bool hugetlb_cgroup_is_root(struct hugetlb_cgroup *h_cg)
	49	{
	50	return (h_cg == root_h_cgroup);
	51	}
	52
3f798518 TH	53	static inline struct hugetlb_cgroup *
3f798518 TH	54	parent_hugetlb_cgroup(struct hugetlb_cgroup *h_cg)
2bc64a20	55	{
5c9d535b	56	return hugetlb_cgroup_from_css(h_cg->css.parent);
2bc64a20 AK	57	}
2bc64a20 AK	58
3f798518	59	static inline bool hugetlb_cgroup_have_usage(struct hugetlb_cgroup *h_cg)
2bc64a20 AK	60	{
2bc64a20 AK	61	int idx;
2bc64a20 AK	62
2bc64a20 AK	63	for (idx = 0; idx < hugetlb_max_hstate; idx++) {
71f87bee	64	if (page_counter_read(&h_cg->hugepage[idx]))
2bc64a20 AK	65	return true;
	66	}
	67	return false;
	68	}
	69
297880f4 DR	70	static void hugetlb_cgroup_init(struct hugetlb_cgroup *h_cgroup,
	71	struct hugetlb_cgroup *parent_h_cgroup)
	72	{
	73	int idx;
	74
	75	for (idx = 0; idx < HUGE_MAX_HSTATE; idx++) {
	76	struct page_counter *counter = &h_cgroup->hugepage[idx];
	77	struct page_counter *parent = NULL;
	78	unsigned long limit;
	79	int ret;
	80
	81	if (parent_h_cgroup)
	82	parent = &parent_h_cgroup->hugepage[idx];
	83	page_counter_init(counter, parent);
	84
	85	limit = round_down(PAGE_COUNTER_MAX,
	86	1 << huge_page_order(&hstates[idx]));
bbec2e15	87	ret = page_counter_set_max(counter, limit);
297880f4 DR	88	VM_BUG_ON(ret);
	89	}
	90	}
	91
eb95419b TH	92	static struct cgroup_subsys_state *
eb95419b TH	93	hugetlb_cgroup_css_alloc(struct cgroup_subsys_state *parent_css)
2bc64a20	94	{
eb95419b TH	95	struct hugetlb_cgroup *parent_h_cgroup = hugetlb_cgroup_from_css(parent_css);
eb95419b TH	96	struct hugetlb_cgroup *h_cgroup;
2bc64a20 AK	97
	98	h_cgroup = kzalloc(sizeof(*h_cgroup), GFP_KERNEL);
	99	if (!h_cgroup)
	100	return ERR_PTR(-ENOMEM);
	101
297880f4	102	if (!parent_h_cgroup)
2bc64a20	103	root_h_cgroup = h_cgroup;
297880f4 DR	104
297880f4 DR	105	hugetlb_cgroup_init(h_cgroup, parent_h_cgroup);
2bc64a20 AK	106	return &h_cgroup->css;
	107	}
	108
eb95419b	109	static void hugetlb_cgroup_css_free(struct cgroup_subsys_state *css)
2bc64a20 AK	110	{
	111	struct hugetlb_cgroup *h_cgroup;
	112
eb95419b	113	h_cgroup = hugetlb_cgroup_from_css(css);
2bc64a20 AK	114	kfree(h_cgroup);
	115	}
	116
da1def55 AK	117
	118	/*
	119	* Should be called with hugetlb_lock held.
	120	* Since we are holding hugetlb_lock, pages cannot get moved from
	121	* active list or uncharged from the cgroup, So no need to get
	122	* page reference and test for page active here. This function
	123	* cannot fail.
	124	*/
3f798518	125	static void hugetlb_cgroup_move_parent(int idx, struct hugetlb_cgroup *h_cg,
da1def55 AK	126	struct page *page)
da1def55 AK	127	{
71f87bee JW	128	unsigned int nr_pages;
71f87bee JW	129	struct page_counter *counter;
da1def55	130	struct hugetlb_cgroup *page_hcg;
3f798518	131	struct hugetlb_cgroup *parent = parent_hugetlb_cgroup(h_cg);
da1def55 AK	132
	133	page_hcg = hugetlb_cgroup_from_page(page);
	134	/*
	135	* We can have pages in active list without any cgroup
	136	* ie, hugepage with less than 3 pages. We can safely
	137	* ignore those pages.
	138	*/
	139	if (!page_hcg \|\| page_hcg != h_cg)
	140	goto out;
	141
71f87bee	142	nr_pages = 1 << compound_order(page);
da1def55 AK	143	if (!parent) {
	144	parent = root_h_cgroup;
	145	/* root has no limit */
71f87bee	146	page_counter_charge(&parent->hugepage[idx], nr_pages);
da1def55 AK	147	}
da1def55 AK	148	counter = &h_cg->hugepage[idx];
71f87bee JW	149	/* Take the pages off the local counter */
71f87bee JW	150	page_counter_cancel(counter, nr_pages);
da1def55 AK	151
	152	set_hugetlb_cgroup(page, parent);
	153	out:
	154	return;
	155	}
	156
	157	/*
	158	* Force the hugetlb cgroup to empty the hugetlb resources by moving them to
	159	* the parent cgroup.
	160	*/
eb95419b	161	static void hugetlb_cgroup_css_offline(struct cgroup_subsys_state *css)
2bc64a20	162	{
eb95419b	163	struct hugetlb_cgroup *h_cg = hugetlb_cgroup_from_css(css);
da1def55 AK	164	struct hstate *h;
da1def55 AK	165	struct page *page;
9d093cb1	166	int idx = 0;
da1def55 AK	167
da1def55 AK	168	do {
da1def55 AK	169	for_each_hstate(h) {
	170	spin_lock(&hugetlb_lock);
	171	list_for_each_entry(page, &h->hugepage_activelist, lru)
3f798518	172	hugetlb_cgroup_move_parent(idx, h_cg, page);
da1def55 AK	173
	174	spin_unlock(&hugetlb_lock);
	175	idx++;
	176	}
	177	cond_resched();
3f798518	178	} while (hugetlb_cgroup_have_usage(h_cg));
2bc64a20 AK	179	}
2bc64a20 AK	180
6d76dcf4 AK	181	int hugetlb_cgroup_charge_cgroup(int idx, unsigned long nr_pages,
	182	struct hugetlb_cgroup **ptr)
	183	{
	184	int ret = 0;
71f87bee	185	struct page_counter *counter;
6d76dcf4	186	struct hugetlb_cgroup *h_cg = NULL;
6d76dcf4 AK	187
	188	if (hugetlb_cgroup_disabled())
	189	goto done;
	190	/*
	191	* We don't charge any cgroup if the compound page have less
	192	* than 3 pages.
	193	*/
	194	if (huge_page_order(&hstates[idx]) < HUGETLB_CGROUP_MIN_ORDER)
	195	goto done;
	196	again:
	197	rcu_read_lock();
	198	h_cg = hugetlb_cgroup_from_task(current);
ec903c0c	199	if (!css_tryget_online(&h_cg->css)) {
6d76dcf4 AK	200	rcu_read_unlock();
	201	goto again;
	202	}
	203	rcu_read_unlock();
	204
6071ca52 JW	205	if (!page_counter_try_charge(&h_cg->hugepage[idx], nr_pages, &counter))
6071ca52 JW	206	ret = -ENOMEM;
6d76dcf4 AK	207	css_put(&h_cg->css);
	208	done:
	209	*ptr = h_cg;
	210	return ret;
	211	}
	212
94ae8ba7	213	/* Should be called with hugetlb_lock held */
6d76dcf4 AK	214	void hugetlb_cgroup_commit_charge(int idx, unsigned long nr_pages,
	215	struct hugetlb_cgroup *h_cg,
	216	struct page *page)
	217	{
	218	if (hugetlb_cgroup_disabled() \|\| !h_cg)
	219	return;
	220
6d76dcf4	221	set_hugetlb_cgroup(page, h_cg);
6d76dcf4 AK	222	return;
	223	}
	224
	225	/*
	226	* Should be called with hugetlb_lock held
	227	*/
	228	void hugetlb_cgroup_uncharge_page(int idx, unsigned long nr_pages,
	229	struct page *page)
	230	{
	231	struct hugetlb_cgroup *h_cg;
6d76dcf4 AK	232
	233	if (hugetlb_cgroup_disabled())
	234	return;
7ea8574e	235	lockdep_assert_held(&hugetlb_lock);
6d76dcf4 AK	236	h_cg = hugetlb_cgroup_from_page(page);
	237	if (unlikely(!h_cg))
	238	return;
	239	set_hugetlb_cgroup(page, NULL);
71f87bee	240	page_counter_uncharge(&h_cg->hugepage[idx], nr_pages);
6d76dcf4 AK	241	return;
	242	}
	243
	244	void hugetlb_cgroup_uncharge_cgroup(int idx, unsigned long nr_pages,
	245	struct hugetlb_cgroup *h_cg)
	246	{
6d76dcf4 AK	247	if (hugetlb_cgroup_disabled() \|\| !h_cg)
	248	return;
	249
	250	if (huge_page_order(&hstates[idx]) < HUGETLB_CGROUP_MIN_ORDER)
	251	return;
	252
71f87bee	253	page_counter_uncharge(&h_cg->hugepage[idx], nr_pages);
6d76dcf4 AK	254	return;
	255	}
	256
71f87bee JW	257	enum {
	258	RES_USAGE,
	259	RES_LIMIT,
	260	RES_MAX_USAGE,
	261	RES_FAILCNT,
	262	};
	263
716f479d TH	264	static u64 hugetlb_cgroup_read_u64(struct cgroup_subsys_state *css,
716f479d TH	265	struct cftype *cft)
abb8206c	266	{
71f87bee	267	struct page_counter *counter;
182446d0	268	struct hugetlb_cgroup *h_cg = hugetlb_cgroup_from_css(css);
abb8206c	269
71f87bee	270	counter = &h_cg->hugepage[MEMFILE_IDX(cft->private)];
abb8206c	271
71f87bee JW	272	switch (MEMFILE_ATTR(cft->private)) {
	273	case RES_USAGE:
	274	return (u64)page_counter_read(counter) * PAGE_SIZE;
	275	case RES_LIMIT:
bbec2e15	276	return (u64)counter->max * PAGE_SIZE;
71f87bee JW	277	case RES_MAX_USAGE:
	278	return (u64)counter->watermark * PAGE_SIZE;
	279	case RES_FAILCNT:
	280	return counter->failcnt;
	281	default:
	282	BUG();
	283	}
abb8206c AK	284	}
abb8206c AK	285
71f87bee JW	286	static DEFINE_MUTEX(hugetlb_limit_mutex);
71f87bee JW	287
451af504 TH	288	static ssize_t hugetlb_cgroup_write(struct kernfs_open_file *of,
451af504 TH	289	char *buf, size_t nbytes, loff_t off)
abb8206c	290	{
71f87bee JW	291	int ret, idx;
71f87bee JW	292	unsigned long nr_pages;
451af504	293	struct hugetlb_cgroup *h_cg = hugetlb_cgroup_from_css(of_css(of));
abb8206c	294
71f87bee JW	295	if (hugetlb_cgroup_is_root(h_cg)) /* Can't set limit on root */
	296	return -EINVAL;
	297
451af504	298	buf = strstrip(buf);
650c5e56	299	ret = page_counter_memparse(buf, "-1", &nr_pages);
71f87bee JW	300	if (ret)
	301	return ret;
	302
451af504	303	idx = MEMFILE_IDX(of_cft(of)->private);
297880f4	304	nr_pages = round_down(nr_pages, 1 << huge_page_order(&hstates[idx]));
abb8206c	305
71f87bee	306	switch (MEMFILE_ATTR(of_cft(of)->private)) {
abb8206c	307	case RES_LIMIT:
71f87bee	308	mutex_lock(&hugetlb_limit_mutex);
bbec2e15	309	ret = page_counter_set_max(&h_cg->hugepage[idx], nr_pages);
71f87bee	310	mutex_unlock(&hugetlb_limit_mutex);
abb8206c AK	311	break;
	312	default:
	313	ret = -EINVAL;
	314	break;
	315	}
451af504	316	return ret ?: nbytes;
abb8206c AK	317	}
abb8206c AK	318
6770c64e TH	319	static ssize_t hugetlb_cgroup_reset(struct kernfs_open_file *of,
6770c64e TH	320	char *buf, size_t nbytes, loff_t off)
abb8206c	321	{
71f87bee JW	322	int ret = 0;
71f87bee JW	323	struct page_counter *counter;
6770c64e	324	struct hugetlb_cgroup *h_cg = hugetlb_cgroup_from_css(of_css(of));
abb8206c	325
71f87bee	326	counter = &h_cg->hugepage[MEMFILE_IDX(of_cft(of)->private)];
abb8206c	327
71f87bee	328	switch (MEMFILE_ATTR(of_cft(of)->private)) {
abb8206c	329	case RES_MAX_USAGE:
71f87bee	330	page_counter_reset_watermark(counter);
abb8206c AK	331	break;
abb8206c AK	332	case RES_FAILCNT:
71f87bee	333	counter->failcnt = 0;
abb8206c AK	334	break;
	335	default:
	336	ret = -EINVAL;
	337	break;
	338	}
6770c64e	339	return ret ?: nbytes;
abb8206c AK	340	}
	341
	342	static char mem_fmt(char buf, int size, unsigned long hsize)
	343	{
	344	if (hsize >= (1UL << 30))
	345	snprintf(buf, size, "%luGB", hsize >> 30);
	346	else if (hsize >= (1UL << 20))
	347	snprintf(buf, size, "%luMB", hsize >> 20);
	348	else
	349	snprintf(buf, size, "%luKB", hsize >> 10);
	350	return buf;
	351	}
	352
7179e7bf	353	static void __init __hugetlb_cgroup_file_init(int idx)
abb8206c AK	354	{
	355	char buf[32];
	356	struct cftype *cft;
	357	struct hstate *h = &hstates[idx];
	358
	359	/* format the size */
	360	mem_fmt(buf, 32, huge_page_size(h));
	361
	362	/* Add the limit file */
	363	cft = &h->cgroup_files[0];
	364	snprintf(cft->name, MAX_CFTYPE_NAME, "%s.limit_in_bytes", buf);
	365	cft->private = MEMFILE_PRIVATE(idx, RES_LIMIT);
716f479d	366	cft->read_u64 = hugetlb_cgroup_read_u64;
451af504	367	cft->write = hugetlb_cgroup_write;
abb8206c AK	368
	369	/* Add the usage file */
	370	cft = &h->cgroup_files[1];
	371	snprintf(cft->name, MAX_CFTYPE_NAME, "%s.usage_in_bytes", buf);
	372	cft->private = MEMFILE_PRIVATE(idx, RES_USAGE);
716f479d	373	cft->read_u64 = hugetlb_cgroup_read_u64;
abb8206c AK	374
	375	/* Add the MAX usage file */
	376	cft = &h->cgroup_files[2];
	377	snprintf(cft->name, MAX_CFTYPE_NAME, "%s.max_usage_in_bytes", buf);
	378	cft->private = MEMFILE_PRIVATE(idx, RES_MAX_USAGE);
6770c64e	379	cft->write = hugetlb_cgroup_reset;
716f479d	380	cft->read_u64 = hugetlb_cgroup_read_u64;
abb8206c AK	381
	382	/* Add the failcntfile */
	383	cft = &h->cgroup_files[3];
	384	snprintf(cft->name, MAX_CFTYPE_NAME, "%s.failcnt", buf);
	385	cft->private = MEMFILE_PRIVATE(idx, RES_FAILCNT);
6770c64e	386	cft->write = hugetlb_cgroup_reset;
716f479d	387	cft->read_u64 = hugetlb_cgroup_read_u64;
abb8206c AK	388
	389	/* NULL terminate the last cft */
	390	cft = &h->cgroup_files[4];
	391	memset(cft, 0, sizeof(*cft));
	392
2cf669a5 TH	393	WARN_ON(cgroup_add_legacy_cftypes(&hugetlb_cgrp_subsys,
2cf669a5 TH	394	h->cgroup_files));
7179e7bf JW	395	}
	396
	397	void __init hugetlb_cgroup_file_init(void)
	398	{
	399	struct hstate *h;
	400
	401	for_each_hstate(h) {
	402	/*
	403	* Add cgroup control files only if the huge page consists
	404	* of more than two normal pages. This is because we use
1d798ca3	405	* page[2].private for storing cgroup details.
7179e7bf JW	406	*/
	407	if (huge_page_order(h) >= HUGETLB_CGROUP_MIN_ORDER)
	408	__hugetlb_cgroup_file_init(hstate_index(h));
	409	}
abb8206c AK	410	}
abb8206c AK	411
75754681 AK	412	/*
	413	* hugetlb_lock will make sure a parallel cgroup rmdir won't happen
	414	* when we migrate hugepages
	415	*/
8e6ac7fa AK	416	void hugetlb_cgroup_migrate(struct page oldhpage, struct page newhpage)
	417	{
	418	struct hugetlb_cgroup *h_cg;
94ae8ba7	419	struct hstate *h = page_hstate(oldhpage);
8e6ac7fa AK	420
	421	if (hugetlb_cgroup_disabled())
	422	return;
	423
309381fe	424	VM_BUG_ON_PAGE(!PageHuge(oldhpage), oldhpage);
8e6ac7fa AK	425	spin_lock(&hugetlb_lock);
	426	h_cg = hugetlb_cgroup_from_page(oldhpage);
	427	set_hugetlb_cgroup(oldhpage, NULL);
8e6ac7fa AK	428
	429	/* move the h_cg details to new cgroup */
	430	set_hugetlb_cgroup(newhpage, h_cg);
94ae8ba7	431	list_move(&newhpage->lru, &h->hugepage_activelist);
8e6ac7fa	432	spin_unlock(&hugetlb_lock);
8e6ac7fa AK	433	return;
	434	}
	435
073219e9	436	struct cgroup_subsys hugetlb_cgrp_subsys = {
92fb9748 TH	437	.css_alloc = hugetlb_cgroup_css_alloc,
	438	.css_offline = hugetlb_cgroup_css_offline,
	439	.css_free = hugetlb_cgroup_css_free,
2bc64a20	440	};