[linux.git] / mm / page_cgroup.c

#include <linux/mm.h>
#include <linux/mmzone.h>
#include <linux/bootmem.h>
#include <linux/bit_spinlock.h>
#include <linux/page_cgroup.h>
#include <linux/hash.h>
#include <linux/slab.h>
#include <linux/memory.h>
#include <linux/vmalloc.h>
#include <linux/cgroup.h>
#include <linux/swapops.h>
#include <linux/kmemleak.h>

static unsigned long total_usage;

#if !defined(CONFIG_SPARSEMEM)


void __meminit pgdat_page_cgroup_init(struct pglist_data *pgdat)
{
	pgdat->node_page_cgroup = NULL;
}

struct page_cgroup *lookup_page_cgroup(struct page *page)
{
	unsigned long pfn = page_to_pfn(page);
	unsigned long offset;
	struct page_cgroup *base;

	base = NODE_DATA(page_to_nid(page))->node_page_cgroup;
#ifdef CONFIG_DEBUG_VM
	/*
	 * The sanity checks the page allocator does upon freeing a
	 * page can reach here before the page_cgroup arrays are
	 * allocated when feeding a range of pages to the allocator
	 * for the first time during bootup or memory hotplug.
	 */
	if (unlikely(!base))
		return NULL;
#endif
	offset = pfn - NODE_DATA(page_to_nid(page))->node_start_pfn;
	return base + offset;
}

static int __init alloc_node_page_cgroup(int nid)
{
	struct page_cgroup *base;
	unsigned long table_size;
	unsigned long nr_pages;

	nr_pages = NODE_DATA(nid)->node_spanned_pages;
	if (!nr_pages)
		return 0;

	table_size = sizeof(struct page_cgroup) * nr_pages;

	base = memblock_virt_alloc_try_nid_nopanic(
			table_size, PAGE_SIZE, __pa(MAX_DMA_ADDRESS),
			BOOTMEM_ALLOC_ACCESSIBLE, nid);
	if (!base)
		return -ENOMEM;
	NODE_DATA(nid)->node_page_cgroup = base;
	total_usage += table_size;
	return 0;
}

void __init page_cgroup_init_flatmem(void)
{

	int nid, fail;

	if (mem_cgroup_disabled())
		return;

	for_each_online_node(nid)  {
		fail = alloc_node_page_cgroup(nid);
		if (fail)
			goto fail;
	}
	printk(KERN_INFO "allocated %ld bytes of page_cgroup\n", total_usage);
	printk(KERN_INFO "please try 'cgroup_disable=memory' option if you"
	" don't want memory cgroups\n");
	return;
fail:
	printk(KERN_CRIT "allocation of page_cgroup failed.\n");
	printk(KERN_CRIT "please try 'cgroup_disable=memory' boot option\n");
	panic("Out of memory");
}

#else /* CONFIG_FLAT_NODE_MEM_MAP */

struct page_cgroup *lookup_page_cgroup(struct page *page)
{
	unsigned long pfn = page_to_pfn(page);
	struct mem_section *section = __pfn_to_section(pfn);
#ifdef CONFIG_DEBUG_VM
	/*
	 * The sanity checks the page allocator does upon freeing a
	 * page can reach here before the page_cgroup arrays are
	 * allocated when feeding a range of pages to the allocator
	 * for the first time during bootup or memory hotplug.
	 */
	if (!section->page_cgroup)
		return NULL;
#endif
	return section->page_cgroup + pfn;
}

static void *__meminit alloc_page_cgroup(size_t size, int nid)
{
	gfp_t flags = GFP_KERNEL | __GFP_ZERO | __GFP_NOWARN;
	void *addr = NULL;

	addr = alloc_pages_exact_nid(nid, size, flags);
	if (addr) {
		kmemleak_alloc(addr, size, 1, flags);
		return addr;
	}

	if (node_state(nid, N_HIGH_MEMORY))
		addr = vzalloc_node(size, nid);
	else
		addr = vzalloc(size);

	return addr;
}

static int __meminit init_section_page_cgroup(unsigned long pfn, int nid)
{
	struct mem_section *section;
	struct page_cgroup *base;
	unsigned long table_size;

	section = __pfn_to_section(pfn);

	if (section->page_cgroup)
		return 0;

	table_size = sizeof(struct page_cgroup) * PAGES_PER_SECTION;
	base = alloc_page_cgroup(table_size, nid);

	/*
	 * The value stored in section->page_cgroup is (base - pfn)
	 * and it does not point to the memory block allocated above,
	 * causing kmemleak false positives.
	 */
	kmemleak_not_leak(base);

	if (!base) {
		printk(KERN_ERR "page cgroup allocation failure\n");
		return -ENOMEM;
	}

	/*
	 * The passed "pfn" may not be aligned to SECTION.  For the calculation
	 * we need to apply a mask.
	 */
	pfn &= PAGE_SECTION_MASK;
	section->page_cgroup = base - pfn;
	total_usage += table_size;
	return 0;
}
#ifdef CONFIG_MEMORY_HOTPLUG
static void free_page_cgroup(void *addr)
{
	if (is_vmalloc_addr(addr)) {
		vfree(addr);
	} else {
		struct page *page = virt_to_page(addr);
		size_t table_size =
			sizeof(struct page_cgroup) * PAGES_PER_SECTION;

		BUG_ON(PageReserved(page));
		free_pages_exact(addr, table_size);
	}
}

static void __free_page_cgroup(unsigned long pfn)
{
	struct mem_section *ms;
	struct page_cgroup *base;

	ms = __pfn_to_section(pfn);
	if (!ms || !ms->page_cgroup)
		return;
	base = ms->page_cgroup + pfn;
	free_page_cgroup(base);
	ms->page_cgroup = NULL;
}

static int __meminit online_page_cgroup(unsigned long start_pfn,
				unsigned long nr_pages,
				int nid)
{
	unsigned long start, end, pfn;
	int fail = 0;

	start = SECTION_ALIGN_DOWN(start_pfn);
	end = SECTION_ALIGN_UP(start_pfn + nr_pages);

	if (nid == -1) {
		/*
		 * In this case, "nid" already exists and contains valid memory.
		 * "start_pfn" passed to us is a pfn which is an arg for
		 * online__pages(), and start_pfn should exist.
		 */
		nid = pfn_to_nid(start_pfn);
		VM_BUG_ON(!node_state(nid, N_ONLINE));
	}

	for (pfn = start; !fail && pfn < end; pfn += PAGES_PER_SECTION) {
		if (!pfn_present(pfn))
			continue;
		fail = init_section_page_cgroup(pfn, nid);
	}
	if (!fail)
		return 0;

	/* rollback */
	for (pfn = start; pfn < end; pfn += PAGES_PER_SECTION)
		__free_page_cgroup(pfn);

	return -ENOMEM;
}

static int __meminit offline_page_cgroup(unsigned long start_pfn,
				unsigned long nr_pages, int nid)
{
	unsigned long start, end, pfn;

	start = SECTION_ALIGN_DOWN(start_pfn);
	end = SECTION_ALIGN_UP(start_pfn + nr_pages);

	for (pfn = start; pfn < end; pfn += PAGES_PER_SECTION)
		__free_page_cgroup(pfn);
	return 0;

}

static int __meminit page_cgroup_callback(struct notifier_block *self,
			       unsigned long action, void *arg)
{
	struct memory_notify *mn = arg;
	int ret = 0;
	switch (action) {
	case MEM_GOING_ONLINE:
		ret = online_page_cgroup(mn->start_pfn,
				   mn->nr_pages, mn->status_change_nid);
		break;
	case MEM_OFFLINE:
		offline_page_cgroup(mn->start_pfn,
				mn->nr_pages, mn->status_change_nid);
		break;
	case MEM_CANCEL_ONLINE:
		offline_page_cgroup(mn->start_pfn,
				mn->nr_pages, mn->status_change_nid);
		break;
	case MEM_GOING_OFFLINE:
		break;
	case MEM_ONLINE:
	case MEM_CANCEL_OFFLINE:
		break;
	}

	return notifier_from_errno(ret);
}

#endif

void __init page_cgroup_init(void)
{
	unsigned long pfn;
	int nid;

	if (mem_cgroup_disabled())
		return;

	for_each_node_state(nid, N_MEMORY) {
		unsigned long start_pfn, end_pfn;

		start_pfn = node_start_pfn(nid);
		end_pfn = node_end_pfn(nid);
		/*
		 * start_pfn and end_pfn may not be aligned to SECTION and the
		 * page->flags of out of node pages are not initialized.  So we
		 * scan [start_pfn, the biggest section's pfn < end_pfn) here.
		 */
		for (pfn = start_pfn;
		     pfn < end_pfn;
                     pfn = ALIGN(pfn + 1, PAGES_PER_SECTION)) {

			if (!pfn_valid(pfn))
				continue;
			/*
			 * Nodes's pfns can be overlapping.
			 * We know some arch can have a nodes layout such as
			 * -------------pfn-------------->
			 * N0 | N1 | N2 | N0 | N1 | N2|....
			 */
			if (pfn_to_nid(pfn) != nid)
				continue;
			if (init_section_page_cgroup(pfn, nid))
				goto oom;
		}
	}
	hotplug_memory_notifier(page_cgroup_callback, 0);
	printk(KERN_INFO "allocated %ld bytes of page_cgroup\n", total_usage);
	printk(KERN_INFO "please try 'cgroup_disable=memory' option if you "
			 "don't want memory cgroups\n");
	return;
oom:
	printk(KERN_CRIT "try 'cgroup_disable=memory' boot option\n");
	panic("Out of memory");
}

void __meminit pgdat_page_cgroup_init(struct pglist_data *pgdat)
{
	return;
}

#endif


#ifdef CONFIG_MEMCG_SWAP

static DEFINE_MUTEX(swap_cgroup_mutex);
struct swap_cgroup_ctrl {
	struct page **map;
	unsigned long length;
	spinlock_t	lock;
};

static struct swap_cgroup_ctrl swap_cgroup_ctrl[MAX_SWAPFILES];

struct swap_cgroup {
	unsigned short		id;
};
#define SC_PER_PAGE	(PAGE_SIZE/sizeof(struct swap_cgroup))

/*
 * SwapCgroup implements "lookup" and "exchange" operations.
 * In typical usage, this swap_cgroup is accessed via memcg's charge/uncharge
 * against SwapCache. At swap_free(), this is accessed directly from swap.
 *
 * This means,
 *  - we have no race in "exchange" when we're accessed via SwapCache because
 *    SwapCache(and its swp_entry) is under lock.
 *  - When called via swap_free(), there is no user of this entry and no race.
 * Then, we don't need lock around "exchange".
 *
 * TODO: we can push these buffers out to HIGHMEM.
 */

/*
 * allocate buffer for swap_cgroup.
 */
static int swap_cgroup_prepare(int type)
{
	struct page *page;
	struct swap_cgroup_ctrl *ctrl;
	unsigned long idx, max;

	ctrl = &swap_cgroup_ctrl[type];

	for (idx = 0; idx < ctrl->length; idx++) {
		page = alloc_page(GFP_KERNEL | __GFP_ZERO);
		if (!page)
			goto not_enough_page;
		ctrl->map[idx] = page;
	}
	return 0;
not_enough_page:
	max = idx;
	for (idx = 0; idx < max; idx++)
		__free_page(ctrl->map[idx]);

	return -ENOMEM;
}

static struct swap_cgroup *lookup_swap_cgroup(swp_entry_t ent,
					struct swap_cgroup_ctrl **ctrlp)
{
	pgoff_t offset = swp_offset(ent);
	struct swap_cgroup_ctrl *ctrl;
	struct page *mappage;
	struct swap_cgroup *sc;

	ctrl = &swap_cgroup_ctrl[swp_type(ent)];
	if (ctrlp)
		*ctrlp = ctrl;

	mappage = ctrl->map[offset / SC_PER_PAGE];
	sc = page_address(mappage);
	return sc + offset % SC_PER_PAGE;
}

/**
 * swap_cgroup_cmpxchg - cmpxchg mem_cgroup's id for this swp_entry.
 * @ent: swap entry to be cmpxchged
 * @old: old id
 * @new: new id
 *
 * Returns old id at success, 0 at failure.
 * (There is no mem_cgroup using 0 as its id)
 */
unsigned short swap_cgroup_cmpxchg(swp_entry_t ent,
					unsigned short old, unsigned short new)
{
	struct swap_cgroup_ctrl *ctrl;
	struct swap_cgroup *sc;
	unsigned long flags;
	unsigned short retval;

	sc = lookup_swap_cgroup(ent, &ctrl);

	spin_lock_irqsave(&ctrl->lock, flags);
	retval = sc->id;
	if (retval == old)
		sc->id = new;
	else
		retval = 0;
	spin_unlock_irqrestore(&ctrl->lock, flags);
	return retval;
}

/**
 * swap_cgroup_record - record mem_cgroup for this swp_entry.
 * @ent: swap entry to be recorded into
 * @id: mem_cgroup to be recorded
 *
 * Returns old value at success, 0 at failure.
 * (Of course, old value can be 0.)
 */
unsigned short swap_cgroup_record(swp_entry_t ent, unsigned short id)
{
	struct swap_cgroup_ctrl *ctrl;
	struct swap_cgroup *sc;
	unsigned short old;
	unsigned long flags;

	sc = lookup_swap_cgroup(ent, &ctrl);

	spin_lock_irqsave(&ctrl->lock, flags);
	old = sc->id;
	sc->id = id;
	spin_unlock_irqrestore(&ctrl->lock, flags);

	return old;
}

/**
 * lookup_swap_cgroup_id - lookup mem_cgroup id tied to swap entry
 * @ent: swap entry to be looked up.
 *
 * Returns ID of mem_cgroup at success. 0 at failure. (0 is invalid ID)
 */
unsigned short lookup_swap_cgroup_id(swp_entry_t ent)
{
	return lookup_swap_cgroup(ent, NULL)->id;
}

int swap_cgroup_swapon(int type, unsigned long max_pages)
{
	void *array;
	unsigned long array_size;
	unsigned long length;
	struct swap_cgroup_ctrl *ctrl;

	if (!do_swap_account)
		return 0;

	length = DIV_ROUND_UP(max_pages, SC_PER_PAGE);
	array_size = length * sizeof(void *);

	array = vzalloc(array_size);
	if (!array)
		goto nomem;

	ctrl = &swap_cgroup_ctrl[type];
	mutex_lock(&swap_cgroup_mutex);
	ctrl->length = length;
	ctrl->map = array;
	spin_lock_init(&ctrl->lock);
	if (swap_cgroup_prepare(type)) {
		/* memory shortage */
		ctrl->map = NULL;
		ctrl->length = 0;
		mutex_unlock(&swap_cgroup_mutex);
		vfree(array);
		goto nomem;
	}
	mutex_unlock(&swap_cgroup_mutex);

	return 0;
nomem:
	printk(KERN_INFO "couldn't allocate enough memory for swap_cgroup.\n");
	printk(KERN_INFO
		"swap_cgroup can be disabled by swapaccount=0 boot option\n");
	return -ENOMEM;
}

void swap_cgroup_swapoff(int type)
{
	struct page **map;
	unsigned long i, length;
	struct swap_cgroup_ctrl *ctrl;

	if (!do_swap_account)
		return;

	mutex_lock(&swap_cgroup_mutex);
	ctrl = &swap_cgroup_ctrl[type];
	map = ctrl->map;
	length = ctrl->length;
	ctrl->map = NULL;
	ctrl->length = 0;
	mutex_unlock(&swap_cgroup_mutex);

	if (map) {
		for (i = 0; i < length; i++) {
			struct page *page = map[i];
			if (page)
				__free_page(page);
		}
		vfree(map);
	}
}

#endif
Commit	Line	Data
52d4b9ac KH	1	#include <linux/mm.h>
	2	#include <linux/mmzone.h>
	3	#include <linux/bootmem.h>
	4	#include <linux/bit_spinlock.h>
	5	#include <linux/page_cgroup.h>
	6	#include <linux/hash.h>
94b6da5a	7	#include <linux/slab.h>
52d4b9ac	8	#include <linux/memory.h>
4c821042	9	#include <linux/vmalloc.h>
94b6da5a	10	#include <linux/cgroup.h>
27a7faa0	11	#include <linux/swapops.h>
7952f988	12	#include <linux/kmemleak.h>
52d4b9ac	13
52d4b9ac KH	14	static unsigned long total_usage;
	15
	16	#if !defined(CONFIG_SPARSEMEM)
	17
	18
31168481	19	void __meminit pgdat_page_cgroup_init(struct pglist_data *pgdat)
52d4b9ac KH	20	{
	21	pgdat->node_page_cgroup = NULL;
	22	}
	23
	24	struct page_cgroup lookup_page_cgroup(struct page page)
	25	{
	26	unsigned long pfn = page_to_pfn(page);
	27	unsigned long offset;
	28	struct page_cgroup *base;
	29
	30	base = NODE_DATA(page_to_nid(page))->node_page_cgroup;
00c54c0b JW	31	#ifdef CONFIG_DEBUG_VM
	32	/*
	33	* The sanity checks the page allocator does upon freeing a
	34	* page can reach here before the page_cgroup arrays are
	35	* allocated when feeding a range of pages to the allocator
	36	* for the first time during bootup or memory hotplug.
	37	*/
52d4b9ac KH	38	if (unlikely(!base))
52d4b9ac KH	39	return NULL;
00c54c0b	40	#endif
52d4b9ac KH	41	offset = pfn - NODE_DATA(page_to_nid(page))->node_start_pfn;
	42	return base + offset;
	43	}
	44
	45	static int __init alloc_node_page_cgroup(int nid)
	46	{
6b208e3f	47	struct page_cgroup *base;
52d4b9ac	48	unsigned long table_size;
6b208e3f	49	unsigned long nr_pages;
52d4b9ac	50
52d4b9ac	51	nr_pages = NODE_DATA(nid)->node_spanned_pages;
653d22c0 KH	52	if (!nr_pages)
	53	return 0;
	54
52d4b9ac	55	table_size = sizeof(struct page_cgroup) * nr_pages;
ca371c0d	56
0d036e9e GS	57	base = memblock_virt_alloc_try_nid_nopanic(
	58	table_size, PAGE_SIZE, __pa(MAX_DMA_ADDRESS),
	59	BOOTMEM_ALLOC_ACCESSIBLE, nid);
ca371c0d	60	if (!base)
52d4b9ac	61	return -ENOMEM;
52d4b9ac KH	62	NODE_DATA(nid)->node_page_cgroup = base;
	63	total_usage += table_size;
	64	return 0;
	65	}
	66
ca371c0d	67	void __init page_cgroup_init_flatmem(void)
52d4b9ac KH	68	{
	69
	70	int nid, fail;
	71
f8d66542	72	if (mem_cgroup_disabled())
94b6da5a KH	73	return;
94b6da5a KH	74
52d4b9ac KH	75	for_each_online_node(nid) {
	76	fail = alloc_node_page_cgroup(nid);
	77	if (fail)
	78	goto fail;
	79	}
	80	printk(KERN_INFO "allocated %ld bytes of page_cgroup\n", total_usage);
8ca739e3 RD	81	printk(KERN_INFO "please try 'cgroup_disable=memory' option if you"
8ca739e3 RD	82	" don't want memory cgroups\n");
52d4b9ac KH	83	return;
52d4b9ac KH	84	fail:
8ca739e3 RD	85	printk(KERN_CRIT "allocation of page_cgroup failed.\n");
8ca739e3 RD	86	printk(KERN_CRIT "please try 'cgroup_disable=memory' boot option\n");
52d4b9ac KH	87	panic("Out of memory");
	88	}
	89
	90	#else /* CONFIG_FLAT_NODE_MEM_MAP */
	91
	92	struct page_cgroup lookup_page_cgroup(struct page page)
	93	{
	94	unsigned long pfn = page_to_pfn(page);
	95	struct mem_section *section = __pfn_to_section(pfn);
00c54c0b JW	96	#ifdef CONFIG_DEBUG_VM
	97	/*
	98	* The sanity checks the page allocator does upon freeing a
	99	* page can reach here before the page_cgroup arrays are
	100	* allocated when feeding a range of pages to the allocator
	101	* for the first time during bootup or memory hotplug.
	102	*/
d69b042f BS	103	if (!section->page_cgroup)
d69b042f BS	104	return NULL;
00c54c0b	105	#endif
52d4b9ac KH	106	return section->page_cgroup + pfn;
	107	}
	108
268433b8	109	static void *__meminit alloc_page_cgroup(size_t size, int nid)
dde79e00	110	{
6b208e3f	111	gfp_t flags = GFP_KERNEL \| __GFP_ZERO \| __GFP_NOWARN;
dde79e00 MH	112	void *addr = NULL;
dde79e00 MH	113
ff7ee93f SR	114	addr = alloc_pages_exact_nid(nid, size, flags);
	115	if (addr) {
	116	kmemleak_alloc(addr, size, 1, flags);
dde79e00	117	return addr;
ff7ee93f	118	}
dde79e00 MH	119
dde79e00 MH	120	if (node_state(nid, N_HIGH_MEMORY))
6b208e3f	121	addr = vzalloc_node(size, nid);
dde79e00	122	else
6b208e3f	123	addr = vzalloc(size);
dde79e00 MH	124
	125	return addr;
	126	}
	127
37573e8c	128	static int __meminit init_section_page_cgroup(unsigned long pfn, int nid)
52d4b9ac	129	{
6b3ae58e	130	struct mem_section *section;
6b208e3f	131	struct page_cgroup *base;
52d4b9ac	132	unsigned long table_size;
52d4b9ac	133
6b208e3f	134	section = __pfn_to_section(pfn);
6b3ae58e JW	135
	136	if (section->page_cgroup)
	137	return 0;
	138
6b3ae58e	139	table_size = sizeof(struct page_cgroup) * PAGES_PER_SECTION;
dde79e00 MH	140	base = alloc_page_cgroup(table_size, nid);
dde79e00 MH	141
6b3ae58e JW	142	/*
	143	* The value stored in section->page_cgroup is (base - pfn)
	144	* and it does not point to the memory block allocated above,
	145	* causing kmemleak false positives.
	146	*/
	147	kmemleak_not_leak(base);
52d4b9ac KH	148
	149	if (!base) {
	150	printk(KERN_ERR "page cgroup allocation failure\n");
	151	return -ENOMEM;
	152	}
	153
37573e8c KH	154	/*
	155	* The passed "pfn" may not be aligned to SECTION. For the calculation
	156	* we need to apply a mask.
	157	*/
	158	pfn &= PAGE_SECTION_MASK;
52d4b9ac KH	159	section->page_cgroup = base - pfn;
	160	total_usage += table_size;
	161	return 0;
	162	}
	163	#ifdef CONFIG_MEMORY_HOTPLUG
0efc8eb9 BL	164	static void free_page_cgroup(void *addr)
	165	{
	166	if (is_vmalloc_addr(addr)) {
	167	vfree(addr);
	168	} else {
	169	struct page *page = virt_to_page(addr);
	170	size_t table_size =
	171	sizeof(struct page_cgroup) * PAGES_PER_SECTION;
	172
	173	BUG_ON(PageReserved(page));
	174	free_pages_exact(addr, table_size);
	175	}
	176	}
	177
d20199e1	178	static void __free_page_cgroup(unsigned long pfn)
52d4b9ac KH	179	{
	180	struct mem_section *ms;
	181	struct page_cgroup *base;
	182
	183	ms = __pfn_to_section(pfn);
	184	if (!ms \|\| !ms->page_cgroup)
	185	return;
	186	base = ms->page_cgroup + pfn;
dde79e00 MH	187	free_page_cgroup(base);
dde79e00 MH	188	ms->page_cgroup = NULL;
52d4b9ac KH	189	}
52d4b9ac KH	190
d20199e1 RK	191	static int __meminit online_page_cgroup(unsigned long start_pfn,
	192	unsigned long nr_pages,
	193	int nid)
52d4b9ac KH	194	{
	195	unsigned long start, end, pfn;
	196	int fail = 0;
	197
1bb36fbd DK	198	start = SECTION_ALIGN_DOWN(start_pfn);
1bb36fbd DK	199	end = SECTION_ALIGN_UP(start_pfn + nr_pages);
52d4b9ac	200
37573e8c KH	201	if (nid == -1) {
	202	/*
	203	* In this case, "nid" already exists and contains valid memory.
	204	* "start_pfn" passed to us is a pfn which is an arg for
	205	* online__pages(), and start_pfn should exist.
	206	*/
	207	nid = pfn_to_nid(start_pfn);
	208	VM_BUG_ON(!node_state(nid, N_ONLINE));
	209	}
	210
52d4b9ac KH	211	for (pfn = start; !fail && pfn < end; pfn += PAGES_PER_SECTION) {
	212	if (!pfn_present(pfn))
	213	continue;
37573e8c	214	fail = init_section_page_cgroup(pfn, nid);
52d4b9ac KH	215	}
	216	if (!fail)
	217	return 0;
	218
	219	/* rollback */
	220	for (pfn = start; pfn < end; pfn += PAGES_PER_SECTION)
	221	__free_page_cgroup(pfn);
	222
	223	return -ENOMEM;
	224	}
	225
d20199e1 RK	226	static int __meminit offline_page_cgroup(unsigned long start_pfn,
d20199e1 RK	227	unsigned long nr_pages, int nid)
52d4b9ac KH	228	{
	229	unsigned long start, end, pfn;
	230
1bb36fbd DK	231	start = SECTION_ALIGN_DOWN(start_pfn);
1bb36fbd DK	232	end = SECTION_ALIGN_UP(start_pfn + nr_pages);
52d4b9ac KH	233
	234	for (pfn = start; pfn < end; pfn += PAGES_PER_SECTION)
	235	__free_page_cgroup(pfn);
	236	return 0;
	237
	238	}
	239
31168481	240	static int __meminit page_cgroup_callback(struct notifier_block *self,
52d4b9ac KH	241	unsigned long action, void *arg)
	242	{
	243	struct memory_notify *mn = arg;
	244	int ret = 0;
	245	switch (action) {
	246	case MEM_GOING_ONLINE:
	247	ret = online_page_cgroup(mn->start_pfn,
	248	mn->nr_pages, mn->status_change_nid);
	249	break;
52d4b9ac KH	250	case MEM_OFFLINE:
	251	offline_page_cgroup(mn->start_pfn,
	252	mn->nr_pages, mn->status_change_nid);
	253	break;
dc19f9db	254	case MEM_CANCEL_ONLINE:
7c72eb32 WC	255	offline_page_cgroup(mn->start_pfn,
	256	mn->nr_pages, mn->status_change_nid);
	257	break;
52d4b9ac KH	258	case MEM_GOING_OFFLINE:
	259	break;
	260	case MEM_ONLINE:
	261	case MEM_CANCEL_OFFLINE:
	262	break;
	263	}
dc19f9db	264
5fda1bd5	265	return notifier_from_errno(ret);
52d4b9ac KH	266	}
	267
	268	#endif
	269
	270	void __init page_cgroup_init(void)
	271	{
	272	unsigned long pfn;
37573e8c	273	int nid;
52d4b9ac	274
f8d66542	275	if (mem_cgroup_disabled())
94b6da5a KH	276	return;
94b6da5a KH	277
31aaea4a	278	for_each_node_state(nid, N_MEMORY) {
37573e8c KH	279	unsigned long start_pfn, end_pfn;
	280
	281	start_pfn = node_start_pfn(nid);
	282	end_pfn = node_end_pfn(nid);
	283	/*
	284	* start_pfn and end_pfn may not be aligned to SECTION and the
	285	* page->flags of out of node pages are not initialized. So we
	286	* scan [start_pfn, the biggest section's pfn < end_pfn) here.
	287	*/
	288	for (pfn = start_pfn;
	289	pfn < end_pfn;
	290	pfn = ALIGN(pfn + 1, PAGES_PER_SECTION)) {
	291
	292	if (!pfn_valid(pfn))
	293	continue;
	294	/*
	295	* Nodes's pfns can be overlapping.
	296	* We know some arch can have a nodes layout such as
	297	* -------------pfn-------------->
	298	* N0 \| N1 \| N2 \| N0 \| N1 \| N2\|....
	299	*/
	300	if (pfn_to_nid(pfn) != nid)
	301	continue;
	302	if (init_section_page_cgroup(pfn, nid))
	303	goto oom;
	304	}
52d4b9ac	305	}
37573e8c	306	hotplug_memory_notifier(page_cgroup_callback, 0);
52d4b9ac	307	printk(KERN_INFO "allocated %ld bytes of page_cgroup\n", total_usage);
37573e8c KH	308	printk(KERN_INFO "please try 'cgroup_disable=memory' option if you "
	309	"don't want memory cgroups\n");
	310	return;
	311	oom:
	312	printk(KERN_CRIT "try 'cgroup_disable=memory' boot option\n");
	313	panic("Out of memory");
52d4b9ac KH	314	}
52d4b9ac KH	315
31168481	316	void __meminit pgdat_page_cgroup_init(struct pglist_data *pgdat)
52d4b9ac KH	317	{
	318	return;
	319	}
	320
	321	#endif
27a7faa0 KH	322
27a7faa0 KH	323
c255a458	324	#ifdef CONFIG_MEMCG_SWAP
27a7faa0 KH	325
	326	static DEFINE_MUTEX(swap_cgroup_mutex);
	327	struct swap_cgroup_ctrl {
	328	struct page **map;
	329	unsigned long length;
e9e58a4e	330	spinlock_t lock;
27a7faa0 KH	331	};
27a7faa0 KH	332
61600f57	333	static struct swap_cgroup_ctrl swap_cgroup_ctrl[MAX_SWAPFILES];
27a7faa0	334
27a7faa0	335	struct swap_cgroup {
a3b2d692	336	unsigned short id;
27a7faa0 KH	337	};
27a7faa0 KH	338	#define SC_PER_PAGE (PAGE_SIZE/sizeof(struct swap_cgroup))
27a7faa0 KH	339
	340	/*
	341	* SwapCgroup implements "lookup" and "exchange" operations.
	342	* In typical usage, this swap_cgroup is accessed via memcg's charge/uncharge
	343	* against SwapCache. At swap_free(), this is accessed directly from swap.
	344	*
	345	* This means,
	346	* - we have no race in "exchange" when we're accessed via SwapCache because
	347	* SwapCache(and its swp_entry) is under lock.
	348	* - When called via swap_free(), there is no user of this entry and no race.
	349	* Then, we don't need lock around "exchange".
	350	*
	351	* TODO: we can push these buffers out to HIGHMEM.
	352	*/
	353
	354	/*
	355	* allocate buffer for swap_cgroup.
	356	*/
	357	static int swap_cgroup_prepare(int type)
	358	{
	359	struct page *page;
	360	struct swap_cgroup_ctrl *ctrl;
	361	unsigned long idx, max;
	362
27a7faa0 KH	363	ctrl = &swap_cgroup_ctrl[type];
	364
	365	for (idx = 0; idx < ctrl->length; idx++) {
	366	page = alloc_page(GFP_KERNEL \| __GFP_ZERO);
	367	if (!page)
	368	goto not_enough_page;
	369	ctrl->map[idx] = page;
	370	}
	371	return 0;
	372	not_enough_page:
	373	max = idx;
	374	for (idx = 0; idx < max; idx++)
	375	__free_page(ctrl->map[idx]);
	376
	377	return -ENOMEM;
	378	}
	379
9fb4b7cc BL	380	static struct swap_cgroup *lookup_swap_cgroup(swp_entry_t ent,
	381	struct swap_cgroup_ctrl **ctrlp)
	382	{
	383	pgoff_t offset = swp_offset(ent);
	384	struct swap_cgroup_ctrl *ctrl;
	385	struct page *mappage;
c09ff089	386	struct swap_cgroup *sc;
9fb4b7cc BL	387
	388	ctrl = &swap_cgroup_ctrl[swp_type(ent)];
	389	if (ctrlp)
	390	*ctrlp = ctrl;
	391
	392	mappage = ctrl->map[offset / SC_PER_PAGE];
c09ff089 HD	393	sc = page_address(mappage);
c09ff089 HD	394	return sc + offset % SC_PER_PAGE;
9fb4b7cc BL	395	}
9fb4b7cc BL	396
02491447 DN	397	/**
02491447 DN	398	* swap_cgroup_cmpxchg - cmpxchg mem_cgroup's id for this swp_entry.
dad7557e	399	* @ent: swap entry to be cmpxchged
02491447 DN	400	* @old: old id
	401	* @new: new id
	402	*
	403	* Returns old id at success, 0 at failure.
25985edc	404	* (There is no mem_cgroup using 0 as its id)
02491447 DN	405	*/
	406	unsigned short swap_cgroup_cmpxchg(swp_entry_t ent,
	407	unsigned short old, unsigned short new)
	408	{
02491447	409	struct swap_cgroup_ctrl *ctrl;
02491447	410	struct swap_cgroup *sc;
e9e58a4e KH	411	unsigned long flags;
e9e58a4e KH	412	unsigned short retval;
02491447	413
9fb4b7cc	414	sc = lookup_swap_cgroup(ent, &ctrl);
02491447	415
e9e58a4e KH	416	spin_lock_irqsave(&ctrl->lock, flags);
	417	retval = sc->id;
	418	if (retval == old)
	419	sc->id = new;
02491447	420	else
e9e58a4e KH	421	retval = 0;
	422	spin_unlock_irqrestore(&ctrl->lock, flags);
	423	return retval;
02491447 DN	424	}
02491447 DN	425
27a7faa0 KH	426	/**
	427	* swap_cgroup_record - record mem_cgroup for this swp_entry.
	428	* @ent: swap entry to be recorded into
dad7557e	429	* @id: mem_cgroup to be recorded
27a7faa0	430	*
a3b2d692 KH	431	* Returns old value at success, 0 at failure.
a3b2d692 KH	432	* (Of course, old value can be 0.)
27a7faa0	433	*/
a3b2d692	434	unsigned short swap_cgroup_record(swp_entry_t ent, unsigned short id)
27a7faa0	435	{
27a7faa0	436	struct swap_cgroup_ctrl *ctrl;
27a7faa0	437	struct swap_cgroup *sc;
a3b2d692	438	unsigned short old;
e9e58a4e	439	unsigned long flags;
27a7faa0	440
9fb4b7cc	441	sc = lookup_swap_cgroup(ent, &ctrl);
27a7faa0	442
e9e58a4e KH	443	spin_lock_irqsave(&ctrl->lock, flags);
	444	old = sc->id;
	445	sc->id = id;
	446	spin_unlock_irqrestore(&ctrl->lock, flags);
27a7faa0 KH	447
	448	return old;
	449	}
	450
	451	/**
9fb4b7cc	452	* lookup_swap_cgroup_id - lookup mem_cgroup id tied to swap entry
27a7faa0 KH	453	* @ent: swap entry to be looked up.
27a7faa0 KH	454	*
b3ff8a2f	455	* Returns ID of mem_cgroup at success. 0 at failure. (0 is invalid ID)
27a7faa0	456	*/
9fb4b7cc	457	unsigned short lookup_swap_cgroup_id(swp_entry_t ent)
27a7faa0	458	{
9fb4b7cc	459	return lookup_swap_cgroup(ent, NULL)->id;
27a7faa0 KH	460	}
	461
	462	int swap_cgroup_swapon(int type, unsigned long max_pages)
	463	{
	464	void *array;
	465	unsigned long array_size;
	466	unsigned long length;
	467	struct swap_cgroup_ctrl *ctrl;
	468
	469	if (!do_swap_account)
	470	return 0;
	471
33278f7f	472	length = DIV_ROUND_UP(max_pages, SC_PER_PAGE);
27a7faa0 KH	473	array_size = length * sizeof(void *);
27a7faa0 KH	474
8c1fec1b	475	array = vzalloc(array_size);
27a7faa0 KH	476	if (!array)
	477	goto nomem;
	478
27a7faa0 KH	479	ctrl = &swap_cgroup_ctrl[type];
	480	mutex_lock(&swap_cgroup_mutex);
	481	ctrl->length = length;
	482	ctrl->map = array;
e9e58a4e	483	spin_lock_init(&ctrl->lock);
27a7faa0 KH	484	if (swap_cgroup_prepare(type)) {
	485	/* memory shortage */
	486	ctrl->map = NULL;
	487	ctrl->length = 0;
27a7faa0	488	mutex_unlock(&swap_cgroup_mutex);
6a5b18d2	489	vfree(array);
27a7faa0 KH	490	goto nomem;
	491	}
	492	mutex_unlock(&swap_cgroup_mutex);
	493
27a7faa0 KH	494	return 0;
	495	nomem:
	496	printk(KERN_INFO "couldn't allocate enough memory for swap_cgroup.\n");
	497	printk(KERN_INFO
00a66d29	498	"swap_cgroup can be disabled by swapaccount=0 boot option\n");
27a7faa0 KH	499	return -ENOMEM;
	500	}
	501
	502	void swap_cgroup_swapoff(int type)
	503	{
6a5b18d2 NK	504	struct page **map;
6a5b18d2 NK	505	unsigned long i, length;
27a7faa0 KH	506	struct swap_cgroup_ctrl *ctrl;
	507
	508	if (!do_swap_account)
	509	return;
	510
	511	mutex_lock(&swap_cgroup_mutex);
	512	ctrl = &swap_cgroup_ctrl[type];
6a5b18d2 NK	513	map = ctrl->map;
	514	length = ctrl->length;
	515	ctrl->map = NULL;
	516	ctrl->length = 0;
	517	mutex_unlock(&swap_cgroup_mutex);
	518
	519	if (map) {
	520	for (i = 0; i < length; i++) {
	521	struct page *page = map[i];
27a7faa0 KH	522	if (page)
	523	__free_page(page);
	524	}
6a5b18d2	525	vfree(map);
27a7faa0	526	}
27a7faa0 KH	527	}
	528
	529	#endif