[mirror_ubuntu-jammy-kernel.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 unnecessary. 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.
 */

#if defined(CONFIG_PAGE_IDLE_FLAG) && !defined(CONFIG_64BIT)
static bool need_page_idle(void)
{
	return true;
}
struct page_ext_operations page_idle_ops = {
	.need = need_page_idle,
};
#endif

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

unsigned long page_ext_size = sizeof(struct page_ext);

static unsigned long total_usage;

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 = page_ext_size;
			page_ext_size += 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();
	}
}

#ifndef CONFIG_SPARSEMEM
void __init page_ext_init_flatmem_late(void)
{
	invoke_init_callbacks();
}
#endif

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

#ifndef 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 = page_ext_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);
	return;

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

#else /* CONFIG_FLATMEM */

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 = page_ext_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 - page_ext_size * pfn;
	total_usage += table_size;
	return 0;
}

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 = page_ext_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)
		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);
}

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
8958b249	37	* boot and extra memory would be unnecessary. In this case, to avoid
eefa864b JK	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
1c676e0d SP	61	#if defined(CONFIG_PAGE_IDLE_FLAG) && !defined(CONFIG_64BIT)
	62	static bool need_page_idle(void)
	63	{
	64	return true;
	65	}
	66	struct page_ext_operations page_idle_ops = {
	67	.need = need_page_idle,
	68	};
	69	#endif
	70
eefa864b	71	static struct page_ext_operations *page_ext_ops[] = {
48c96a36 JK	72	#ifdef CONFIG_PAGE_OWNER
	73	&page_owner_ops,
	74	#endif
1c676e0d	75	#if defined(CONFIG_PAGE_IDLE_FLAG) && !defined(CONFIG_64BIT)
33c3fc71 VD	76	&page_idle_ops,
33c3fc71 VD	77	#endif
eefa864b JK	78	};
eefa864b JK	79
5556cfe8 VB	80	unsigned long page_ext_size = sizeof(struct page_ext);
5556cfe8 VB	81
eefa864b JK	82	static unsigned long total_usage;
	83
	84	static bool __init invoke_need_callbacks(void)
	85	{
	86	int i;
	87	int entries = ARRAY_SIZE(page_ext_ops);
980ac167	88	bool need = false;
eefa864b JK	89
eefa864b JK	90	for (i = 0; i < entries; i++) {
980ac167	91	if (page_ext_ops[i]->need && page_ext_ops[i]->need()) {
5556cfe8 VB	92	page_ext_ops[i]->offset = page_ext_size;
5556cfe8 VB	93	page_ext_size += page_ext_ops[i]->size;
980ac167 JK	94	need = true;
980ac167 JK	95	}
eefa864b JK	96	}
eefa864b JK	97
980ac167	98	return need;
eefa864b JK	99	}
	100
	101	static void __init invoke_init_callbacks(void)
	102	{
	103	int i;
	104	int entries = ARRAY_SIZE(page_ext_ops);
	105
	106	for (i = 0; i < entries; i++) {
	107	if (page_ext_ops[i]->init)
	108	page_ext_ops[i]->init();
	109	}
	110	}
	111
7fb7ab6d ZH	112	#ifndef CONFIG_SPARSEMEM
	113	void __init page_ext_init_flatmem_late(void)
	114	{
	115	invoke_init_callbacks();
	116	}
	117	#endif
	118
980ac167 JK	119	static inline struct page_ext get_entry(void base, unsigned long index)
980ac167 JK	120	{
5556cfe8	121	return base + page_ext_size * index;
980ac167 JK	122	}
980ac167 JK	123
7fb7ab6d	124	#ifndef CONFIG_SPARSEMEM
eefa864b JK	125
	126
	127	void __meminit pgdat_page_ext_init(struct pglist_data *pgdat)
	128	{
	129	pgdat->node_page_ext = NULL;
	130	}
	131
10ed6341	132	struct page_ext lookup_page_ext(const struct page page)
eefa864b JK	133	{
eefa864b JK	134	unsigned long pfn = page_to_pfn(page);
0b06bb3f	135	unsigned long index;
eefa864b JK	136	struct page_ext *base;
	137
	138	base = NODE_DATA(page_to_nid(page))->node_page_ext;
eefa864b JK	139	/*
	140	* The sanity checks the page allocator does upon freeing a
	141	* page can reach here before the page_ext arrays are
	142	* allocated when feeding a range of pages to the allocator
	143	* for the first time during bootup or memory hotplug.
	144	*/
	145	if (unlikely(!base))
	146	return NULL;
0b06bb3f	147	index = pfn - round_down(node_start_pfn(page_to_nid(page)),
eefa864b	148	MAX_ORDER_NR_PAGES);
980ac167	149	return get_entry(base, index);
eefa864b JK	150	}
	151
	152	static int __init alloc_node_page_ext(int nid)
	153	{
	154	struct page_ext *base;
	155	unsigned long table_size;
	156	unsigned long nr_pages;
	157
	158	nr_pages = NODE_DATA(nid)->node_spanned_pages;
	159	if (!nr_pages)
	160	return 0;
	161
	162	/*
	163	* Need extra space if node range is not aligned with
	164	* MAX_ORDER_NR_PAGES. When page allocator's buddy algorithm
	165	* checks buddy's status, range could be out of exact node range.
	166	*/
	167	if (!IS_ALIGNED(node_start_pfn(nid), MAX_ORDER_NR_PAGES) \|\|
	168	!IS_ALIGNED(node_end_pfn(nid), MAX_ORDER_NR_PAGES))
	169	nr_pages += MAX_ORDER_NR_PAGES;
	170
5556cfe8	171	table_size = page_ext_size * nr_pages;
eefa864b	172
26fb3dae	173	base = memblock_alloc_try_nid(
eefa864b	174	table_size, PAGE_SIZE, __pa(MAX_DMA_ADDRESS),
97ad1087	175	MEMBLOCK_ALLOC_ACCESSIBLE, nid);
eefa864b JK	176	if (!base)
	177	return -ENOMEM;
	178	NODE_DATA(nid)->node_page_ext = base;
	179	total_usage += table_size;
	180	return 0;
	181	}
	182
	183	void __init page_ext_init_flatmem(void)
	184	{
	185
	186	int nid, fail;
	187
	188	if (!invoke_need_callbacks())
	189	return;
	190
	191	for_each_online_node(nid) {
	192	fail = alloc_node_page_ext(nid);
	193	if (fail)
	194	goto fail;
	195	}
	196	pr_info("allocated %ld bytes of page_ext\n", total_usage);
eefa864b JK	197	return;
	198
	199	fail:
	200	pr_crit("allocation of page_ext failed.\n");
	201	panic("Out of memory");
	202	}
	203
43b02ba9	204	#else /* CONFIG_FLATMEM */
eefa864b	205
10ed6341	206	struct page_ext lookup_page_ext(const struct page page)
eefa864b JK	207	{
	208	unsigned long pfn = page_to_pfn(page);
	209	struct mem_section *section = __pfn_to_section(pfn);
eefa864b JK	210	/*
	211	* The sanity checks the page allocator does upon freeing a
	212	* page can reach here before the page_ext arrays are
	213	* allocated when feeding a range of pages to the allocator
	214	* for the first time during bootup or memory hotplug.
	215	*/
	216	if (!section->page_ext)
	217	return NULL;
980ac167	218	return get_entry(section->page_ext, pfn);
eefa864b JK	219	}
	220
	221	static void *__meminit alloc_page_ext(size_t size, int nid)
	222	{
	223	gfp_t flags = GFP_KERNEL \| __GFP_ZERO \| __GFP_NOWARN;
	224	void *addr = NULL;
	225
	226	addr = alloc_pages_exact_nid(nid, size, flags);
	227	if (addr) {
	228	kmemleak_alloc(addr, size, 1, flags);
	229	return addr;
	230	}
	231
b95046b0	232	addr = vzalloc_node(size, nid);
eefa864b JK	233
	234	return addr;
	235	}
	236
	237	static int __meminit init_section_page_ext(unsigned long pfn, int nid)
	238	{
	239	struct mem_section *section;
	240	struct page_ext *base;
	241	unsigned long table_size;
	242
	243	section = __pfn_to_section(pfn);
	244
	245	if (section->page_ext)
	246	return 0;
	247
5556cfe8	248	table_size = page_ext_size * PAGES_PER_SECTION;
eefa864b JK	249	base = alloc_page_ext(table_size, nid);
	250
	251	/*
	252	* The value stored in section->page_ext is (base - pfn)
	253	* and it does not point to the memory block allocated above,
	254	* causing kmemleak false positives.
	255	*/
	256	kmemleak_not_leak(base);
	257
	258	if (!base) {
	259	pr_err("page ext allocation failure\n");
	260	return -ENOMEM;
	261	}
	262
	263	/*
	264	* The passed "pfn" may not be aligned to SECTION. For the calculation
	265	* we need to apply a mask.
	266	*/
	267	pfn &= PAGE_SECTION_MASK;
5556cfe8	268	section->page_ext = (void )base - page_ext_size pfn;
eefa864b JK	269	total_usage += table_size;
	270	return 0;
	271	}
76af6a05	272
eefa864b JK	273	static void free_page_ext(void *addr)
	274	{
	275	if (is_vmalloc_addr(addr)) {
	276	vfree(addr);
	277	} else {
	278	struct page *page = virt_to_page(addr);
	279	size_t table_size;
	280
5556cfe8	281	table_size = page_ext_size * PAGES_PER_SECTION;
eefa864b JK	282
eefa864b JK	283	BUG_ON(PageReserved(page));
0c815854	284	kmemleak_free(addr);
eefa864b JK	285	free_pages_exact(addr, table_size);
	286	}
	287	}
	288
	289	static void __free_page_ext(unsigned long pfn)
	290	{
	291	struct mem_section *ms;
	292	struct page_ext *base;
	293
	294	ms = __pfn_to_section(pfn);
	295	if (!ms \|\| !ms->page_ext)
	296	return;
980ac167	297	base = get_entry(ms->page_ext, pfn);
eefa864b JK	298	free_page_ext(base);
	299	ms->page_ext = NULL;
	300	}
	301
	302	static int __meminit online_page_ext(unsigned long start_pfn,
	303	unsigned long nr_pages,
	304	int nid)
	305	{
	306	unsigned long start, end, pfn;
	307	int fail = 0;
	308
	309	start = SECTION_ALIGN_DOWN(start_pfn);
	310	end = SECTION_ALIGN_UP(start_pfn + nr_pages);
	311
98fa15f3	312	if (nid == NUMA_NO_NODE) {
eefa864b JK	313	/*
	314	* In this case, "nid" already exists and contains valid memory.
	315	* "start_pfn" passed to us is a pfn which is an arg for
	316	* online__pages(), and start_pfn should exist.
	317	*/
	318	nid = pfn_to_nid(start_pfn);
	319	VM_BUG_ON(!node_state(nid, N_ONLINE));
	320	}
	321
dccacf8d	322	for (pfn = start; !fail && pfn < end; pfn += PAGES_PER_SECTION)
eefa864b	323	fail = init_section_page_ext(pfn, nid);
eefa864b JK	324	if (!fail)
	325	return 0;
	326
	327	/* rollback */
	328	for (pfn = start; pfn < end; pfn += PAGES_PER_SECTION)
	329	__free_page_ext(pfn);
	330
	331	return -ENOMEM;
	332	}
	333
	334	static int __meminit offline_page_ext(unsigned long start_pfn,
	335	unsigned long nr_pages, int nid)
	336	{
	337	unsigned long start, end, pfn;
	338
	339	start = SECTION_ALIGN_DOWN(start_pfn);
	340	end = SECTION_ALIGN_UP(start_pfn + nr_pages);
	341
	342	for (pfn = start; pfn < end; pfn += PAGES_PER_SECTION)
	343	__free_page_ext(pfn);
	344	return 0;
	345
	346	}
	347
	348	static int __meminit page_ext_callback(struct notifier_block *self,
	349	unsigned long action, void *arg)
	350	{
	351	struct memory_notify *mn = arg;
	352	int ret = 0;
	353
	354	switch (action) {
	355	case MEM_GOING_ONLINE:
	356	ret = online_page_ext(mn->start_pfn,
	357	mn->nr_pages, mn->status_change_nid);
	358	break;
	359	case MEM_OFFLINE:
	360	offline_page_ext(mn->start_pfn,
	361	mn->nr_pages, mn->status_change_nid);
	362	break;
	363	case MEM_CANCEL_ONLINE:
	364	offline_page_ext(mn->start_pfn,
	365	mn->nr_pages, mn->status_change_nid);
	366	break;
	367	case MEM_GOING_OFFLINE:
	368	break;
	369	case MEM_ONLINE:
	370	case MEM_CANCEL_OFFLINE:
	371	break;
	372	}
	373
	374	return notifier_from_errno(ret);
	375	}
	376
eefa864b JK	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\|....
	405	*/
2f1ee091	406	if (pfn_to_nid(pfn) != nid)
eefa864b JK	407	continue;
	408	if (init_section_page_ext(pfn, nid))
	409	goto oom;
0fc542b7	410	cond_resched();
eefa864b JK	411	}
	412	}
	413	hotplug_memory_notifier(page_ext_callback, 0);
	414	pr_info("allocated %ld bytes of page_ext\n", total_usage);
	415	invoke_init_callbacks();
	416	return;
	417
	418	oom:
	419	panic("Out of memory");
	420	}
	421
	422	void __meminit pgdat_page_ext_init(struct pglist_data *pgdat)
	423	{
	424	}
	425
	426	#endif