[linux.git] / mm / page_ext.c

#include <linux/mm.h>
#include <linux/mmzone.h>
#include <linux/bootmem.h>
#include <linux/page_ext.h>
#include <linux/memory.h>
#include <linux/vmalloc.h>
#include <linux/kmemleak.h>
#include <linux/page_owner.h>
#include <linux/page_idle.h>

/*
 * struct page extension
 *
 * This is the feature to manage memory for extended data per page.
 *
 * Until now, we must modify struct page itself to store extra data per page.
 * This requires rebuilding the kernel and it is really time consuming process.
 * And, sometimes, rebuild is impossible due to third party module dependency.
 * At last, enlarging struct page could cause un-wanted system behaviour change.
 *
 * This feature is intended to overcome above mentioned problems. This feature
 * allocates memory for extended data per page in certain place rather than
 * the struct page itself. This memory can be accessed by the accessor
 * functions provided by this code. During the boot process, it checks whether
 * allocation of huge chunk of memory is needed or not. If not, it avoids
 * allocating memory at all. With this advantage, we can include this feature
 * into the kernel in default and can avoid rebuild and solve related problems.
 *
 * To help these things to work well, there are two callbacks for clients. One
 * is the need callback which is mandatory if user wants to avoid useless
 * memory allocation at boot-time. The other is optional, init callback, which
 * is used to do proper initialization after memory is allocated.
 *
 * The need callback is used to decide whether extended memory allocation is
 * needed or not. Sometimes users want to deactivate some features in this
 * boot and extra memory would be unneccessary. In this case, to avoid
 * allocating huge chunk of memory, each clients represent their need of
 * extra memory through the need callback. If one of the need callbacks
 * returns true, it means that someone needs extra memory so that
 * page extension core should allocates memory for page extension. If
 * none of need callbacks return true, memory isn't needed at all in this boot
 * and page extension core can skip to allocate memory. As result,
 * none of memory is wasted.
 *
 * When need callback returns true, page_ext checks if there is a request for
 * extra memory through size in struct page_ext_operations. If it is non-zero,
 * extra space is allocated for each page_ext entry and offset is returned to
 * user through offset in struct page_ext_operations.
 *
 * The init callback is used to do proper initialization after page extension
 * is completely initialized. In sparse memory system, extra memory is
 * allocated some time later than memmap is allocated. In other words, lifetime
 * of memory for page extension isn't same with memmap for struct page.
 * Therefore, clients can't store extra data until page extension is
 * initialized, even if pages are allocated and used freely. This could
 * cause inadequate state of extra data per page, so, to prevent it, client
 * can utilize this callback to initialize the state of it correctly.
 */

static struct page_ext_operations *page_ext_ops[] = {
	&debug_guardpage_ops,
#ifdef CONFIG_PAGE_OWNER
	&page_owner_ops,
#endif
#if defined(CONFIG_IDLE_PAGE_TRACKING) && !defined(CONFIG_64BIT)
	&page_idle_ops,
#endif
};

static unsigned long total_usage;
static unsigned long extra_mem;

static bool __init invoke_need_callbacks(void)
{
	int i;
	int entries = ARRAY_SIZE(page_ext_ops);
	bool need = false;

	for (i = 0; i < entries; i++) {
		if (page_ext_ops[i]->need && page_ext_ops[i]->need()) {
			page_ext_ops[i]->offset = sizeof(struct page_ext) +
						extra_mem;
			extra_mem += page_ext_ops[i]->size;
			need = true;
		}
	}

	return need;
}

static void __init invoke_init_callbacks(void)
{
	int i;
	int entries = ARRAY_SIZE(page_ext_ops);

	for (i = 0; i < entries; i++) {
		if (page_ext_ops[i]->init)
			page_ext_ops[i]->init();
	}
}

static unsigned long get_entry_size(void)
{
	return sizeof(struct page_ext) + extra_mem;
}

static inline struct page_ext *get_entry(void *base, unsigned long index)
{
	return base + get_entry_size() * index;
}

#if !defined(CONFIG_SPARSEMEM)


void __meminit pgdat_page_ext_init(struct pglist_data *pgdat)
{
	pgdat->node_page_ext = NULL;
}

struct page_ext *lookup_page_ext(struct page *page)
{
	unsigned long pfn = page_to_pfn(page);
	unsigned long index;
	struct page_ext *base;

	base = NODE_DATA(page_to_nid(page))->node_page_ext;
#if defined(CONFIG_DEBUG_VM)
	/*
	 * The sanity checks the page allocator does upon freeing a
	 * page can reach here before the page_ext 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
	index = pfn - round_down(node_start_pfn(page_to_nid(page)),
					MAX_ORDER_NR_PAGES);
	return get_entry(base, index);
}

static int __init alloc_node_page_ext(int nid)
{
	struct page_ext *base;
	unsigned long table_size;
	unsigned long nr_pages;

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

	/*
	 * Need extra space if node range is not aligned with
	 * MAX_ORDER_NR_PAGES. When page allocator's buddy algorithm
	 * checks buddy's status, range could be out of exact node range.
	 */
	if (!IS_ALIGNED(node_start_pfn(nid), MAX_ORDER_NR_PAGES) ||
		!IS_ALIGNED(node_end_pfn(nid), MAX_ORDER_NR_PAGES))
		nr_pages += MAX_ORDER_NR_PAGES;

	table_size = get_entry_size() * 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_ext = base;
	total_usage += table_size;
	return 0;
}

void __init page_ext_init_flatmem(void)
{

	int nid, fail;

	if (!invoke_need_callbacks())
		return;

	for_each_online_node(nid)  {
		fail = alloc_node_page_ext(nid);
		if (fail)
			goto fail;
	}
	pr_info("allocated %ld bytes of page_ext\n", total_usage);
	invoke_init_callbacks();
	return;

fail:
	pr_crit("allocation of page_ext failed.\n");
	panic("Out of memory");
}

#else /* CONFIG_FLAT_NODE_MEM_MAP */

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

static void *__meminit alloc_page_ext(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_ext(unsigned long pfn, int nid)
{
	struct mem_section *section;
	struct page_ext *base;
	unsigned long table_size;

	section = __pfn_to_section(pfn);

	if (section->page_ext)
		return 0;

	table_size = get_entry_size() * PAGES_PER_SECTION;
	base = alloc_page_ext(table_size, nid);

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

	if (!base) {
		pr_err("page ext 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_ext = (void *)base - get_entry_size() * pfn;
	total_usage += table_size;
	return 0;
}
#ifdef CONFIG_MEMORY_HOTPLUG
static void free_page_ext(void *addr)
{
	if (is_vmalloc_addr(addr)) {
		vfree(addr);
	} else {
		struct page *page = virt_to_page(addr);
		size_t table_size;

		table_size = get_entry_size() * PAGES_PER_SECTION;

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

static void __free_page_ext(unsigned long pfn)
{
	struct mem_section *ms;
	struct page_ext *base;

	ms = __pfn_to_section(pfn);
	if (!ms || !ms->page_ext)
		return;
	base = get_entry(ms->page_ext, pfn);
	free_page_ext(base);
	ms->page_ext = NULL;
}

static int __meminit online_page_ext(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_ext(pfn, nid);
	}
	if (!fail)
		return 0;

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

	return -ENOMEM;
}

static int __meminit offline_page_ext(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_ext(pfn);
	return 0;

}

static int __meminit page_ext_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_ext(mn->start_pfn,
				   mn->nr_pages, mn->status_change_nid);
		break;
	case MEM_OFFLINE:
		offline_page_ext(mn->start_pfn,
				mn->nr_pages, mn->status_change_nid);
		break;
	case MEM_CANCEL_ONLINE:
		offline_page_ext(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_ext_init(void)
{
	unsigned long pfn;
	int nid;

	if (!invoke_need_callbacks())
		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|....
			 *
			 * Take into account DEFERRED_STRUCT_PAGE_INIT.
			 */
			if (early_pfn_to_nid(pfn) != nid)
				continue;
			if (init_section_page_ext(pfn, nid))
				goto oom;
		}
	}
	hotplug_memory_notifier(page_ext_callback, 0);
	pr_info("allocated %ld bytes of page_ext\n", total_usage);
	invoke_init_callbacks();
	return;

oom:
	panic("Out of memory");
}

void __meminit pgdat_page_ext_init(struct pglist_data *pgdat)
{
}

#endif
Commit	Line	Data
eefa864b JK	1	#include <linux/mm.h>
	2	#include <linux/mmzone.h>
	3	#include <linux/bootmem.h>
	4	#include <linux/page_ext.h>
	5	#include <linux/memory.h>
	6	#include <linux/vmalloc.h>
	7	#include <linux/kmemleak.h>
48c96a36	8	#include <linux/page_owner.h>
33c3fc71	9	#include <linux/page_idle.h>
eefa864b JK	10
	11	/*
	12	* struct page extension
	13	*
	14	* This is the feature to manage memory for extended data per page.
	15	*
	16	* Until now, we must modify struct page itself to store extra data per page.
	17	* This requires rebuilding the kernel and it is really time consuming process.
	18	* And, sometimes, rebuild is impossible due to third party module dependency.
	19	* At last, enlarging struct page could cause un-wanted system behaviour change.
	20	*
	21	* This feature is intended to overcome above mentioned problems. This feature
	22	* allocates memory for extended data per page in certain place rather than
	23	* the struct page itself. This memory can be accessed by the accessor
	24	* functions provided by this code. During the boot process, it checks whether
	25	* allocation of huge chunk of memory is needed or not. If not, it avoids
	26	* allocating memory at all. With this advantage, we can include this feature
	27	* into the kernel in default and can avoid rebuild and solve related problems.
	28	*
	29	* To help these things to work well, there are two callbacks for clients. One
	30	* is the need callback which is mandatory if user wants to avoid useless
	31	* memory allocation at boot-time. The other is optional, init callback, which
	32	* is used to do proper initialization after memory is allocated.
	33	*
	34	* The need callback is used to decide whether extended memory allocation is
	35	* needed or not. Sometimes users want to deactivate some features in this
	36	* boot and extra memory would be unneccessary. In this case, to avoid
	37	* allocating huge chunk of memory, each clients represent their need of
	38	* extra memory through the need callback. If one of the need callbacks
	39	* returns true, it means that someone needs extra memory so that
	40	* page extension core should allocates memory for page extension. If
	41	* none of need callbacks return true, memory isn't needed at all in this boot
	42	* and page extension core can skip to allocate memory. As result,
	43	* none of memory is wasted.
	44	*
980ac167 JK	45	* When need callback returns true, page_ext checks if there is a request for
	46	* extra memory through size in struct page_ext_operations. If it is non-zero,
	47	* extra space is allocated for each page_ext entry and offset is returned to
	48	* user through offset in struct page_ext_operations.
	49	*
eefa864b JK	50	* The init callback is used to do proper initialization after page extension
	51	* is completely initialized. In sparse memory system, extra memory is
	52	* allocated some time later than memmap is allocated. In other words, lifetime
	53	* of memory for page extension isn't same with memmap for struct page.
	54	* Therefore, clients can't store extra data until page extension is
	55	* initialized, even if pages are allocated and used freely. This could
	56	* cause inadequate state of extra data per page, so, to prevent it, client
	57	* can utilize this callback to initialize the state of it correctly.
	58	*/
	59
	60	static struct page_ext_operations *page_ext_ops[] = {
e30825f1	61	&debug_guardpage_ops,
48c96a36 JK	62	#ifdef CONFIG_PAGE_OWNER
	63	&page_owner_ops,
	64	#endif
33c3fc71 VD	65	#if defined(CONFIG_IDLE_PAGE_TRACKING) && !defined(CONFIG_64BIT)
	66	&page_idle_ops,
	67	#endif
eefa864b JK	68	};
	69
	70	static unsigned long total_usage;
980ac167	71	static unsigned long extra_mem;
eefa864b JK	72
	73	static bool __init invoke_need_callbacks(void)
	74	{
	75	int i;
	76	int entries = ARRAY_SIZE(page_ext_ops);
980ac167	77	bool need = false;
eefa864b JK	78
eefa864b JK	79	for (i = 0; i < entries; i++) {
980ac167 JK	80	if (page_ext_ops[i]->need && page_ext_ops[i]->need()) {
	81	page_ext_ops[i]->offset = sizeof(struct page_ext) +
	82	extra_mem;
	83	extra_mem += page_ext_ops[i]->size;
	84	need = true;
	85	}
eefa864b JK	86	}
eefa864b JK	87
980ac167	88	return need;
eefa864b JK	89	}
	90
	91	static void __init invoke_init_callbacks(void)
	92	{
	93	int i;
	94	int entries = ARRAY_SIZE(page_ext_ops);
	95
	96	for (i = 0; i < entries; i++) {
	97	if (page_ext_ops[i]->init)
	98	page_ext_ops[i]->init();
	99	}
	100	}
	101
980ac167 JK	102	static unsigned long get_entry_size(void)
	103	{
	104	return sizeof(struct page_ext) + extra_mem;
	105	}
	106
	107	static inline struct page_ext get_entry(void base, unsigned long index)
	108	{
	109	return base + get_entry_size() * index;
	110	}
	111
eefa864b JK	112	#if !defined(CONFIG_SPARSEMEM)
	113
	114
	115	void __meminit pgdat_page_ext_init(struct pglist_data *pgdat)
	116	{
	117	pgdat->node_page_ext = NULL;
	118	}
	119
	120	struct page_ext lookup_page_ext(struct page page)
	121	{
	122	unsigned long pfn = page_to_pfn(page);
0b06bb3f	123	unsigned long index;
eefa864b JK	124	struct page_ext *base;
	125
	126	base = NODE_DATA(page_to_nid(page))->node_page_ext;
bd33ef36	127	#if defined(CONFIG_DEBUG_VM)
eefa864b JK	128	/*
	129	* The sanity checks the page allocator does upon freeing a
	130	* page can reach here before the page_ext arrays are
	131	* allocated when feeding a range of pages to the allocator
	132	* for the first time during bootup or memory hotplug.
	133	*/
	134	if (unlikely(!base))
	135	return NULL;
	136	#endif
0b06bb3f	137	index = pfn - round_down(node_start_pfn(page_to_nid(page)),
eefa864b	138	MAX_ORDER_NR_PAGES);
980ac167	139	return get_entry(base, index);
eefa864b JK	140	}
	141
	142	static int __init alloc_node_page_ext(int nid)
	143	{
	144	struct page_ext *base;
	145	unsigned long table_size;
	146	unsigned long nr_pages;
	147
	148	nr_pages = NODE_DATA(nid)->node_spanned_pages;
	149	if (!nr_pages)
	150	return 0;
	151
	152	/*
	153	* Need extra space if node range is not aligned with
	154	* MAX_ORDER_NR_PAGES. When page allocator's buddy algorithm
	155	* checks buddy's status, range could be out of exact node range.
	156	*/
	157	if (!IS_ALIGNED(node_start_pfn(nid), MAX_ORDER_NR_PAGES) \|\|
	158	!IS_ALIGNED(node_end_pfn(nid), MAX_ORDER_NR_PAGES))
	159	nr_pages += MAX_ORDER_NR_PAGES;
	160
980ac167	161	table_size = get_entry_size() * nr_pages;
eefa864b JK	162
	163	base = memblock_virt_alloc_try_nid_nopanic(
	164	table_size, PAGE_SIZE, __pa(MAX_DMA_ADDRESS),
	165	BOOTMEM_ALLOC_ACCESSIBLE, nid);
	166	if (!base)
	167	return -ENOMEM;
	168	NODE_DATA(nid)->node_page_ext = base;
	169	total_usage += table_size;
	170	return 0;
	171	}
	172
	173	void __init page_ext_init_flatmem(void)
	174	{
	175
	176	int nid, fail;
	177
	178	if (!invoke_need_callbacks())
	179	return;
	180
	181	for_each_online_node(nid) {
	182	fail = alloc_node_page_ext(nid);
	183	if (fail)
	184	goto fail;
	185	}
	186	pr_info("allocated %ld bytes of page_ext\n", total_usage);
	187	invoke_init_callbacks();
	188	return;
	189
	190	fail:
	191	pr_crit("allocation of page_ext failed.\n");
	192	panic("Out of memory");
	193	}
	194
	195	#else /* CONFIG_FLAT_NODE_MEM_MAP */
	196
	197	struct page_ext lookup_page_ext(struct page page)
	198	{
	199	unsigned long pfn = page_to_pfn(page);
	200	struct mem_section *section = __pfn_to_section(pfn);
bd33ef36	201	#if defined(CONFIG_DEBUG_VM)
eefa864b JK	202	/*
	203	* The sanity checks the page allocator does upon freeing a
	204	* page can reach here before the page_ext arrays are
	205	* allocated when feeding a range of pages to the allocator
	206	* for the first time during bootup or memory hotplug.
	207	*/
	208	if (!section->page_ext)
	209	return NULL;
	210	#endif
980ac167	211	return get_entry(section->page_ext, pfn);
eefa864b JK	212	}
	213
	214	static void *__meminit alloc_page_ext(size_t size, int nid)
	215	{
	216	gfp_t flags = GFP_KERNEL \| __GFP_ZERO \| __GFP_NOWARN;
	217	void *addr = NULL;
	218
	219	addr = alloc_pages_exact_nid(nid, size, flags);
	220	if (addr) {
	221	kmemleak_alloc(addr, size, 1, flags);
	222	return addr;
	223	}
	224
	225	if (node_state(nid, N_HIGH_MEMORY))
	226	addr = vzalloc_node(size, nid);
	227	else
	228	addr = vzalloc(size);
	229
	230	return addr;
	231	}
	232
	233	static int __meminit init_section_page_ext(unsigned long pfn, int nid)
	234	{
	235	struct mem_section *section;
	236	struct page_ext *base;
	237	unsigned long table_size;
	238
	239	section = __pfn_to_section(pfn);
	240
	241	if (section->page_ext)
	242	return 0;
	243
980ac167	244	table_size = get_entry_size() * PAGES_PER_SECTION;
eefa864b JK	245	base = alloc_page_ext(table_size, nid);
	246
	247	/*
	248	* The value stored in section->page_ext is (base - pfn)
	249	* and it does not point to the memory block allocated above,
	250	* causing kmemleak false positives.
	251	*/
	252	kmemleak_not_leak(base);
	253
	254	if (!base) {
	255	pr_err("page ext allocation failure\n");
	256	return -ENOMEM;
	257	}
	258
	259	/*
	260	* The passed "pfn" may not be aligned to SECTION. For the calculation
	261	* we need to apply a mask.
	262	*/
	263	pfn &= PAGE_SECTION_MASK;
980ac167	264	section->page_ext = (void )base - get_entry_size() pfn;
eefa864b JK	265	total_usage += table_size;
	266	return 0;
	267	}
	268	#ifdef CONFIG_MEMORY_HOTPLUG
	269	static void free_page_ext(void *addr)
	270	{
	271	if (is_vmalloc_addr(addr)) {
	272	vfree(addr);
	273	} else {
	274	struct page *page = virt_to_page(addr);
	275	size_t table_size;
	276
980ac167	277	table_size = get_entry_size() * PAGES_PER_SECTION;
eefa864b JK	278
	279	BUG_ON(PageReserved(page));
	280	free_pages_exact(addr, table_size);
	281	}
	282	}
	283
	284	static void __free_page_ext(unsigned long pfn)
	285	{
	286	struct mem_section *ms;
	287	struct page_ext *base;
	288
	289	ms = __pfn_to_section(pfn);
	290	if (!ms \|\| !ms->page_ext)
	291	return;
980ac167	292	base = get_entry(ms->page_ext, pfn);
eefa864b JK	293	free_page_ext(base);
	294	ms->page_ext = NULL;
	295	}
	296
	297	static int __meminit online_page_ext(unsigned long start_pfn,
	298	unsigned long nr_pages,
	299	int nid)
	300	{
	301	unsigned long start, end, pfn;
	302	int fail = 0;
	303
	304	start = SECTION_ALIGN_DOWN(start_pfn);
	305	end = SECTION_ALIGN_UP(start_pfn + nr_pages);
	306
	307	if (nid == -1) {
	308	/*
	309	* In this case, "nid" already exists and contains valid memory.
	310	* "start_pfn" passed to us is a pfn which is an arg for
	311	* online__pages(), and start_pfn should exist.
	312	*/
	313	nid = pfn_to_nid(start_pfn);
	314	VM_BUG_ON(!node_state(nid, N_ONLINE));
	315	}
	316
	317	for (pfn = start; !fail && pfn < end; pfn += PAGES_PER_SECTION) {
	318	if (!pfn_present(pfn))
	319	continue;
	320	fail = init_section_page_ext(pfn, nid);
	321	}
	322	if (!fail)
	323	return 0;
	324
	325	/* rollback */
	326	for (pfn = start; pfn < end; pfn += PAGES_PER_SECTION)
	327	__free_page_ext(pfn);
	328
	329	return -ENOMEM;
	330	}
	331
	332	static int __meminit offline_page_ext(unsigned long start_pfn,
	333	unsigned long nr_pages, int nid)
	334	{
	335	unsigned long start, end, pfn;
	336
	337	start = SECTION_ALIGN_DOWN(start_pfn);
	338	end = SECTION_ALIGN_UP(start_pfn + nr_pages);
	339
	340	for (pfn = start; pfn < end; pfn += PAGES_PER_SECTION)
	341	__free_page_ext(pfn);
	342	return 0;
	343
	344	}
	345
	346	static int __meminit page_ext_callback(struct notifier_block *self,
	347	unsigned long action, void *arg)
	348	{
	349	struct memory_notify *mn = arg;
	350	int ret = 0;
	351
	352	switch (action) {
	353	case MEM_GOING_ONLINE:
	354	ret = online_page_ext(mn->start_pfn,
	355	mn->nr_pages, mn->status_change_nid);
	356	break;
357	case MEM_OFFLINE:
358	offline_page_ext(mn->start_pfn,
359	mn->nr_pages, mn->status_change_nid);
360	break;
361	case MEM_CANCEL_ONLINE:
362	offline_page_ext(mn->start_pfn,
363	mn->nr_pages, mn->status_change_nid);
364	break;
365	case MEM_GOING_OFFLINE:
366	break;
367	case MEM_ONLINE:
368	case MEM_CANCEL_OFFLINE:
369	break;
370	}
371
372	return notifier_from_errno(ret);
373	}
374
375	#endif
376
377	void __init page_ext_init(void)
378	{
379	unsigned long pfn;
380	int nid;
381
382	if (!invoke_need_callbacks())
383	return;
384
385	for_each_node_state(nid, N_MEMORY) {
386	unsigned long start_pfn, end_pfn;
387
388	start_pfn = node_start_pfn(nid);
389	end_pfn = node_end_pfn(nid);
390	/*
391	* start_pfn and end_pfn may not be aligned to SECTION and the
392	* page->flags of out of node pages are not initialized. So we
393	* scan [start_pfn, the biggest section's pfn < end_pfn) here.
394	*/
395	for (pfn = start_pfn; pfn < end_pfn;
396	pfn = ALIGN(pfn + 1, PAGES_PER_SECTION)) {
397
398	if (!pfn_valid(pfn))
399	continue;
400	/*
401	* Nodes's pfns can be overlapping.
402	* We know some arch can have a nodes layout such as
403	* -------------pfn-------------->
404	* N0 \| N1 \| N2 \| N0 \| N1 \| N2\|....
fe53ca54 YS	405	*
fe53ca54 YS	406	* Take into account DEFERRED_STRUCT_PAGE_INIT.
eefa864b	407	*/
fe53ca54	408	if (early_pfn_to_nid(pfn) != nid)
eefa864b JK	409	continue;
	410	if (init_section_page_ext(pfn, nid))
	411	goto oom;
	412	}
	413	}
	414	hotplug_memory_notifier(page_ext_callback, 0);
	415	pr_info("allocated %ld bytes of page_ext\n", total_usage);
	416	invoke_init_callbacks();
	417	return;
	418
	419	oom:
	420	panic("Out of memory");
	421	}
	422
	423	void __meminit pgdat_page_ext_init(struct pglist_data *pgdat)
	424	{
	425	}
	426
	427	#endif