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

// SPDX-License-Identifier: GPL-2.0
#include <linux/mm.h>
#include <linux/mmzone.h>
#include <linux/memblock.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[] = {
#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(const 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;
	/*
	 * 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;
	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_alloc_try_nid(
			table_size, PAGE_SIZE, __pa(MAX_DMA_ADDRESS),
			MEMBLOCK_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(const struct page *page)
{
	unsigned long pfn = page_to_pfn(page);
	struct mem_section *section = __pfn_to_section(pfn);
	/*
	 * 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;
	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;
	}

	addr = vzalloc_node(size, nid);

	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));
		kmemleak_free(addr);
		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 == NUMA_NO_NODE) {
		/*
		 * 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|....
			 */
			if (pfn_to_nid(pfn) != nid)
				continue;
			if (init_section_page_ext(pfn, nid))
				goto oom;
			cond_resched();
		}
	}
	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
b2441318	1	// SPDX-License-Identifier: GPL-2.0
eefa864b JK	2	#include <linux/mm.h>
eefa864b JK	3	#include <linux/mmzone.h>
57c8a661	4	#include <linux/memblock.h>
eefa864b JK	5	#include <linux/page_ext.h>
	6	#include <linux/memory.h>
	7	#include <linux/vmalloc.h>
	8	#include <linux/kmemleak.h>
48c96a36	9	#include <linux/page_owner.h>
33c3fc71	10	#include <linux/page_idle.h>
eefa864b JK	11
	12	/*
	13	* struct page extension
	14	*
	15	* This is the feature to manage memory for extended data per page.
	16	*
	17	* Until now, we must modify struct page itself to store extra data per page.
	18	* This requires rebuilding the kernel and it is really time consuming process.
	19	* And, sometimes, rebuild is impossible due to third party module dependency.
	20	* At last, enlarging struct page could cause un-wanted system behaviour change.
	21	*
	22	* This feature is intended to overcome above mentioned problems. This feature
	23	* allocates memory for extended data per page in certain place rather than
	24	* the struct page itself. This memory can be accessed by the accessor
	25	* functions provided by this code. During the boot process, it checks whether
	26	* allocation of huge chunk of memory is needed or not. If not, it avoids
	27	* allocating memory at all. With this advantage, we can include this feature
	28	* into the kernel in default and can avoid rebuild and solve related problems.
	29	*
	30	* To help these things to work well, there are two callbacks for clients. One
	31	* is the need callback which is mandatory if user wants to avoid useless
	32	* memory allocation at boot-time. The other is optional, init callback, which
	33	* is used to do proper initialization after memory is allocated.
	34	*
	35	* The need callback is used to decide whether extended memory allocation is
	36	* needed or not. Sometimes users want to deactivate some features in this
	37	* boot and extra memory would be unneccessary. In this case, to avoid
	38	* allocating huge chunk of memory, each clients represent their need of
	39	* extra memory through the need callback. If one of the need callbacks
	40	* returns true, it means that someone needs extra memory so that
	41	* page extension core should allocates memory for page extension. If
	42	* none of need callbacks return true, memory isn't needed at all in this boot
	43	* and page extension core can skip to allocate memory. As result,
	44	* none of memory is wasted.
	45	*
980ac167 JK	46	* When need callback returns true, page_ext checks if there is a request for
	47	* extra memory through size in struct page_ext_operations. If it is non-zero,
	48	* extra space is allocated for each page_ext entry and offset is returned to
	49	* user through offset in struct page_ext_operations.
	50	*
eefa864b JK	51	* The init callback is used to do proper initialization after page extension
	52	* is completely initialized. In sparse memory system, extra memory is
	53	* allocated some time later than memmap is allocated. In other words, lifetime
	54	* of memory for page extension isn't same with memmap for struct page.
	55	* Therefore, clients can't store extra data until page extension is
	56	* initialized, even if pages are allocated and used freely. This could
	57	* cause inadequate state of extra data per page, so, to prevent it, client
	58	* can utilize this callback to initialize the state of it correctly.
	59	*/
	60
	61	static struct page_ext_operations *page_ext_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
10ed6341	120	struct page_ext lookup_page_ext(const struct page page)
eefa864b JK	121	{
eefa864b JK	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;
eefa864b JK	127	/*
	128	* The sanity checks the page allocator does upon freeing a
	129	* page can reach here before the page_ext arrays are
	130	* allocated when feeding a range of pages to the allocator
	131	* for the first time during bootup or memory hotplug.
	132	*/
	133	if (unlikely(!base))
	134	return NULL;
0b06bb3f	135	index = pfn - round_down(node_start_pfn(page_to_nid(page)),
eefa864b	136	MAX_ORDER_NR_PAGES);
980ac167	137	return get_entry(base, index);
eefa864b JK	138	}
	139
	140	static int __init alloc_node_page_ext(int nid)
	141	{
	142	struct page_ext *base;
	143	unsigned long table_size;
	144	unsigned long nr_pages;
	145
	146	nr_pages = NODE_DATA(nid)->node_spanned_pages;
	147	if (!nr_pages)
	148	return 0;
	149
	150	/*
	151	* Need extra space if node range is not aligned with
	152	* MAX_ORDER_NR_PAGES. When page allocator's buddy algorithm
	153	* checks buddy's status, range could be out of exact node range.
	154	*/
	155	if (!IS_ALIGNED(node_start_pfn(nid), MAX_ORDER_NR_PAGES) \|\|
	156	!IS_ALIGNED(node_end_pfn(nid), MAX_ORDER_NR_PAGES))
	157	nr_pages += MAX_ORDER_NR_PAGES;
	158
980ac167	159	table_size = get_entry_size() * nr_pages;
eefa864b	160
26fb3dae	161	base = memblock_alloc_try_nid(
eefa864b	162	table_size, PAGE_SIZE, __pa(MAX_DMA_ADDRESS),
97ad1087	163	MEMBLOCK_ALLOC_ACCESSIBLE, nid);
eefa864b JK	164	if (!base)
	165	return -ENOMEM;
	166	NODE_DATA(nid)->node_page_ext = base;
	167	total_usage += table_size;
	168	return 0;
	169	}
	170
	171	void __init page_ext_init_flatmem(void)
	172	{
	173
	174	int nid, fail;
	175
	176	if (!invoke_need_callbacks())
	177	return;
	178
	179	for_each_online_node(nid) {
	180	fail = alloc_node_page_ext(nid);
	181	if (fail)
	182	goto fail;
	183	}
	184	pr_info("allocated %ld bytes of page_ext\n", total_usage);
	185	invoke_init_callbacks();
	186	return;
	187
	188	fail:
	189	pr_crit("allocation of page_ext failed.\n");
	190	panic("Out of memory");
	191	}
	192
	193	#else /* CONFIG_FLAT_NODE_MEM_MAP */
	194
10ed6341	195	struct page_ext lookup_page_ext(const struct page page)
eefa864b JK	196	{
	197	unsigned long pfn = page_to_pfn(page);
	198	struct mem_section *section = __pfn_to_section(pfn);
eefa864b JK	199	/*
	200	* The sanity checks the page allocator does upon freeing a
	201	* page can reach here before the page_ext arrays are
	202	* allocated when feeding a range of pages to the allocator
	203	* for the first time during bootup or memory hotplug.
	204	*/
	205	if (!section->page_ext)
	206	return NULL;
980ac167	207	return get_entry(section->page_ext, pfn);
eefa864b JK	208	}
	209
	210	static void *__meminit alloc_page_ext(size_t size, int nid)
	211	{
	212	gfp_t flags = GFP_KERNEL \| __GFP_ZERO \| __GFP_NOWARN;
	213	void *addr = NULL;
	214
	215	addr = alloc_pages_exact_nid(nid, size, flags);
	216	if (addr) {
	217	kmemleak_alloc(addr, size, 1, flags);
	218	return addr;
	219	}
	220
b95046b0	221	addr = vzalloc_node(size, nid);
eefa864b JK	222
	223	return addr;
	224	}
	225
	226	static int __meminit init_section_page_ext(unsigned long pfn, int nid)
	227	{
	228	struct mem_section *section;
	229	struct page_ext *base;
	230	unsigned long table_size;
	231
	232	section = __pfn_to_section(pfn);
	233
	234	if (section->page_ext)
	235	return 0;
	236
980ac167	237	table_size = get_entry_size() * PAGES_PER_SECTION;
eefa864b JK	238	base = alloc_page_ext(table_size, nid);
	239
	240	/*
	241	* The value stored in section->page_ext is (base - pfn)
	242	* and it does not point to the memory block allocated above,
	243	* causing kmemleak false positives.
	244	*/
	245	kmemleak_not_leak(base);
	246
	247	if (!base) {
	248	pr_err("page ext allocation failure\n");
	249	return -ENOMEM;
	250	}
	251
	252	/*
	253	* The passed "pfn" may not be aligned to SECTION. For the calculation
	254	* we need to apply a mask.
	255	*/
	256	pfn &= PAGE_SECTION_MASK;
980ac167	257	section->page_ext = (void )base - get_entry_size() pfn;
eefa864b JK	258	total_usage += table_size;
	259	return 0;
	260	}
	261	#ifdef CONFIG_MEMORY_HOTPLUG
	262	static void free_page_ext(void *addr)
	263	{
	264	if (is_vmalloc_addr(addr)) {
	265	vfree(addr);
	266	} else {
	267	struct page *page = virt_to_page(addr);
	268	size_t table_size;
	269
980ac167	270	table_size = get_entry_size() * PAGES_PER_SECTION;
eefa864b JK	271
eefa864b JK	272	BUG_ON(PageReserved(page));
0c815854	273	kmemleak_free(addr);
eefa864b JK	274	free_pages_exact(addr, table_size);
	275	}
	276	}
	277
	278	static void __free_page_ext(unsigned long pfn)
	279	{
	280	struct mem_section *ms;
	281	struct page_ext *base;
	282
	283	ms = __pfn_to_section(pfn);
	284	if (!ms \|\| !ms->page_ext)
	285	return;
980ac167	286	base = get_entry(ms->page_ext, pfn);
eefa864b JK	287	free_page_ext(base);
	288	ms->page_ext = NULL;
	289	}
	290
	291	static int __meminit online_page_ext(unsigned long start_pfn,
	292	unsigned long nr_pages,
	293	int nid)
	294	{
	295	unsigned long start, end, pfn;
	296	int fail = 0;
	297
	298	start = SECTION_ALIGN_DOWN(start_pfn);
	299	end = SECTION_ALIGN_UP(start_pfn + nr_pages);
	300
98fa15f3	301	if (nid == NUMA_NO_NODE) {
eefa864b JK	302	/*
	303	* In this case, "nid" already exists and contains valid memory.
	304	* "start_pfn" passed to us is a pfn which is an arg for
	305	* online__pages(), and start_pfn should exist.
	306	*/
	307	nid = pfn_to_nid(start_pfn);
	308	VM_BUG_ON(!node_state(nid, N_ONLINE));
	309	}
	310
	311	for (pfn = start; !fail && pfn < end; pfn += PAGES_PER_SECTION) {
	312	if (!pfn_present(pfn))
	313	continue;
	314	fail = init_section_page_ext(pfn, nid);
	315	}
	316	if (!fail)
	317	return 0;
	318
	319	/* rollback */
	320	for (pfn = start; pfn < end; pfn += PAGES_PER_SECTION)
	321	__free_page_ext(pfn);
	322
	323	return -ENOMEM;
	324	}
	325
	326	static int __meminit offline_page_ext(unsigned long start_pfn,
	327	unsigned long nr_pages, int nid)
	328	{
	329	unsigned long start, end, pfn;
	330
	331	start = SECTION_ALIGN_DOWN(start_pfn);
	332	end = SECTION_ALIGN_UP(start_pfn + nr_pages);
	333
	334	for (pfn = start; pfn < end; pfn += PAGES_PER_SECTION)
	335	__free_page_ext(pfn);
	336	return 0;
	337
	338	}
	339
	340	static int __meminit page_ext_callback(struct notifier_block *self,
	341	unsigned long action, void *arg)
	342	{
	343	struct memory_notify *mn = arg;
	344	int ret = 0;
	345
	346	switch (action) {
	347	case MEM_GOING_ONLINE:
	348	ret = online_page_ext(mn->start_pfn,
	349	mn->nr_pages, mn->status_change_nid);
	350	break;
	351	case MEM_OFFLINE:
	352	offline_page_ext(mn->start_pfn,
	353	mn->nr_pages, mn->status_change_nid);
	354	break;
	355	case MEM_CANCEL_ONLINE:
	356	offline_page_ext(mn->start_pfn,
	357	mn->nr_pages, mn->status_change_nid);
	358	break;
	359	case MEM_GOING_OFFLINE:
	360	break;
	361	case MEM_ONLINE:
	362	case MEM_CANCEL_OFFLINE:
	363	break;
	364	}
	365
366	return notifier_from_errno(ret);
367	}
368
369	#endif
370
371	void __init page_ext_init(void)
372	{
373	unsigned long pfn;
374	int nid;
375
376	if (!invoke_need_callbacks())
377	return;
378
379	for_each_node_state(nid, N_MEMORY) {
380	unsigned long start_pfn, end_pfn;
381
382	start_pfn = node_start_pfn(nid);
383	end_pfn = node_end_pfn(nid);
384	/*
385	* start_pfn and end_pfn may not be aligned to SECTION and the
386	* page->flags of out of node pages are not initialized. So we
387	* scan [start_pfn, the biggest section's pfn < end_pfn) here.
388	*/
389	for (pfn = start_pfn; pfn < end_pfn;
390	pfn = ALIGN(pfn + 1, PAGES_PER_SECTION)) {
391
392	if (!pfn_valid(pfn))
393	continue;
394	/*
395	* Nodes's pfns can be overlapping.
396	* We know some arch can have a nodes layout such as
397	* -------------pfn-------------->
398	* N0 \| N1 \| N2 \| N0 \| N1 \| N2\|....
399	*/
2f1ee091	400	if (pfn_to_nid(pfn) != nid)
eefa864b JK	401	continue;
	402	if (init_section_page_ext(pfn, nid))
	403	goto oom;
0fc542b7	404	cond_resched();
eefa864b JK	405	}
	406	}
	407	hotplug_memory_notifier(page_ext_callback, 0);
	408	pr_info("allocated %ld bytes of page_ext\n", total_usage);
	409	invoke_init_callbacks();
	410	return;
	411
	412	oom:
	413	panic("Out of memory");
	414	}
	415
	416	void __meminit pgdat_page_ext_init(struct pglist_data *pgdat)
	417	{
	418	}
	419
	420	#endif