[linux-2.6-block.git] / mm / hugetlb_cgroup.c

/*
 *
 * Copyright IBM Corporation, 2012
 * Author Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
 *
 * 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 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;
	int idx;

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

	if (parent_h_cgroup) {
		for (idx = 0; idx < HUGE_MAX_HSTATE; idx++)
			page_counter_init(&h_cgroup->hugepage[idx],
					  &parent_h_cgroup->hugepage[idx]);
	} else {
		root_h_cgroup = h_cgroup;
		for (idx = 0; idx < HUGE_MAX_HSTATE; idx++)
			page_counter_init(&h_cgroup->hugepage[idx], NULL);
	}
	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();

	ret = page_counter_try_charge(&h_cg->hugepage[idx], nr_pages, &counter);
	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->limit * 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);

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