[thirdparty/linux.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>
#include <linux/page_table_check.h>
#include <linux/rcupdate.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.
 */

#ifdef CONFIG_SPARSEMEM
#define PAGE_EXT_INVALID       (0x1)
#endif

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

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

unsigned long page_ext_size;

static unsigned long total_usage;
static struct page_ext *lookup_page_ext(const struct page *page);

bool early_page_ext __meminitdata;
static int __init setup_early_page_ext(char *str)
{
	early_page_ext = true;
	return 0;
}
early_param("early_page_ext", setup_early_page_ext);

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()) {
			if (page_ext_ops[i]->need_shared_flags) {
				page_ext_size = sizeof(struct page_ext);
				break;
			}
		}
	}

	for (i = 0; i < entries; i++) {
		if (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;
}

/**
 * page_ext_get() - Get the extended information for a page.
 * @page: The page we're interested in.
 *
 * Ensures that the page_ext will remain valid until page_ext_put()
 * is called.
 *
 * Return: NULL if no page_ext exists for this page.
 * Context: Any context.  Caller may not sleep until they have called
 * page_ext_put().
 */
struct page_ext *page_ext_get(struct page *page)
{
	struct page_ext *page_ext;

	rcu_read_lock();
	page_ext = lookup_page_ext(page);
	if (!page_ext) {
		rcu_read_unlock();
		return NULL;
	}

	return page_ext;
}

/**
 * page_ext_put() - Working with page extended information is done.
 * @page_ext: Page extended information received from page_ext_get().
 *
 * The page extended information of the page may not be valid after this
 * function is called.
 *
 * Return: None.
 * Context: Any context with corresponding page_ext_get() is called.
 */
void page_ext_put(struct page_ext *page_ext)
{
	if (unlikely(!page_ext))
		return;

	rcu_read_unlock();
}
#ifndef CONFIG_SPARSEMEM


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

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

	WARN_ON_ONCE(!rcu_read_lock_held());
	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_SPARSEMEM */
static bool page_ext_invalid(struct page_ext *page_ext)
{
	return !page_ext || (((unsigned long)page_ext & PAGE_EXT_INVALID) == PAGE_EXT_INVALID);
}

static 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);
	struct page_ext *page_ext = READ_ONCE(section->page_ext);

	WARN_ON_ONCE(!rcu_read_lock_held());
	/*
	 * 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 (page_ext_invalid(page_ext))
		return NULL;
	return get_entry(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 = READ_ONCE(ms->page_ext);
	/*
	 * page_ext here can be valid while doing the roll back
	 * operation in online_page_ext().
	 */
	if (page_ext_invalid(base))
		base = (void *)base - PAGE_EXT_INVALID;
	WRITE_ONCE(ms->page_ext, NULL);

	base = get_entry(base, pfn);
	free_page_ext(base);
}

static void __invalidate_page_ext(unsigned long pfn)
{
	struct mem_section *ms;
	void *val;

	ms = __pfn_to_section(pfn);
	if (!ms || !ms->page_ext)
		return;
	val = (void *)ms->page_ext + PAGE_EXT_INVALID;
	WRITE_ONCE(ms->page_ext, val);
}

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_online(nid));
	}

	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)
{
	unsigned long start, end, pfn;

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

	/*
	 * Freeing of page_ext is done in 3 steps to avoid
	 * use-after-free of it:
	 * 1) Traverse all the sections and mark their page_ext
	 *    as invalid.
	 * 2) Wait for all the existing users of page_ext who
	 *    started before invalidation to finish.
	 * 3) Free the page_ext.
	 */
	for (pfn = start; pfn < end; pfn += PAGES_PER_SECTION)
		__invalidate_page_ext(pfn);

	synchronize_rcu();

	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);
		break;
	case MEM_CANCEL_ONLINE:
		offline_page_ext(mn->start_pfn,
				mn->nr_pages);
		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, DEFAULT_CALLBACK_PRI);
	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>
df4e817b	11	#include <linux/page_table_check.h>
b1d5488a	12	#include <linux/rcupdate.h>
eefa864b JK	13
	14	/*
	15	* struct page extension
	16	*
	17	* This is the feature to manage memory for extended data per page.
	18	*
	19	* Until now, we must modify struct page itself to store extra data per page.
	20	* This requires rebuilding the kernel and it is really time consuming process.
	21	* And, sometimes, rebuild is impossible due to third party module dependency.
	22	* At last, enlarging struct page could cause un-wanted system behaviour change.
	23	*
	24	* This feature is intended to overcome above mentioned problems. This feature
	25	* allocates memory for extended data per page in certain place rather than
	26	* the struct page itself. This memory can be accessed by the accessor
	27	* functions provided by this code. During the boot process, it checks whether
	28	* allocation of huge chunk of memory is needed or not. If not, it avoids
	29	* allocating memory at all. With this advantage, we can include this feature
	30	* into the kernel in default and can avoid rebuild and solve related problems.
	31	*
	32	* To help these things to work well, there are two callbacks for clients. One
	33	* is the need callback which is mandatory if user wants to avoid useless
	34	* memory allocation at boot-time. The other is optional, init callback, which
	35	* is used to do proper initialization after memory is allocated.
	36	*
	37	* The need callback is used to decide whether extended memory allocation is
	38	* needed or not. Sometimes users want to deactivate some features in this
8958b249	39	* boot and extra memory would be unnecessary. In this case, to avoid
eefa864b JK	40	* allocating huge chunk of memory, each clients represent their need of
	41	* extra memory through the need callback. If one of the need callbacks
	42	* returns true, it means that someone needs extra memory so that
	43	* page extension core should allocates memory for page extension. If
	44	* none of need callbacks return true, memory isn't needed at all in this boot
	45	* and page extension core can skip to allocate memory. As result,
	46	* none of memory is wasted.
	47	*
980ac167 JK	48	* When need callback returns true, page_ext checks if there is a request for
	49	* extra memory through size in struct page_ext_operations. If it is non-zero,
	50	* extra space is allocated for each page_ext entry and offset is returned to
	51	* user through offset in struct page_ext_operations.
	52	*
eefa864b JK	53	* The init callback is used to do proper initialization after page extension
	54	* is completely initialized. In sparse memory system, extra memory is
	55	* allocated some time later than memmap is allocated. In other words, lifetime
	56	* of memory for page extension isn't same with memmap for struct page.
	57	* Therefore, clients can't store extra data until page extension is
	58	* initialized, even if pages are allocated and used freely. This could
	59	* cause inadequate state of extra data per page, so, to prevent it, client
	60	* can utilize this callback to initialize the state of it correctly.
	61	*/
	62
b1d5488a CTK	63	#ifdef CONFIG_SPARSEMEM
	64	#define PAGE_EXT_INVALID (0x1)
	65	#endif
	66
1c676e0d SP	67	#if defined(CONFIG_PAGE_IDLE_FLAG) && !defined(CONFIG_64BIT)
	68	static bool need_page_idle(void)
	69	{
	70	return true;
	71	}
cab0a7c1	72	static struct page_ext_operations page_idle_ops __initdata = {
1c676e0d	73	.need = need_page_idle,
6189eb82	74	.need_shared_flags = true,
1c676e0d SP	75	};
	76	#endif
	77
cab0a7c1	78	static struct page_ext_operations *page_ext_ops[] __initdata = {
48c96a36 JK	79	#ifdef CONFIG_PAGE_OWNER
	80	&page_owner_ops,
	81	#endif
1c676e0d	82	#if defined(CONFIG_PAGE_IDLE_FLAG) && !defined(CONFIG_64BIT)
33c3fc71 VD	83	&page_idle_ops,
33c3fc71 VD	84	#endif
df4e817b PT	85	#ifdef CONFIG_PAGE_TABLE_CHECK
	86	&page_table_check_ops,
	87	#endif
eefa864b JK	88	};
eefa864b JK	89
6189eb82	90	unsigned long page_ext_size;
5556cfe8	91
eefa864b	92	static unsigned long total_usage;
b1d5488a	93	static struct page_ext lookup_page_ext(const struct page page);
eefa864b	94
7ec7096b	95	bool early_page_ext __meminitdata;
c4f20f14 LZ	96	static int __init setup_early_page_ext(char *str)
	97	{
	98	early_page_ext = true;
	99	return 0;
	100	}
	101	early_param("early_page_ext", setup_early_page_ext);
	102
eefa864b JK	103	static bool __init invoke_need_callbacks(void)
	104	{
	105	int i;
	106	int entries = ARRAY_SIZE(page_ext_ops);
980ac167	107	bool need = false;
eefa864b JK	108
eefa864b JK	109	for (i = 0; i < entries; i++) {
6189eb82 PT	110	if (page_ext_ops[i]->need()) {
	111	if (page_ext_ops[i]->need_shared_flags) {
	112	page_ext_size = sizeof(struct page_ext);
	113	break;
	114	}
	115	}
	116	}
	117
	118	for (i = 0; i < entries; i++) {
	119	if (page_ext_ops[i]->need()) {
5556cfe8 VB	120	page_ext_ops[i]->offset = page_ext_size;
5556cfe8 VB	121	page_ext_size += page_ext_ops[i]->size;
980ac167 JK	122	need = true;
980ac167 JK	123	}
eefa864b JK	124	}
eefa864b JK	125
980ac167	126	return need;
eefa864b JK	127	}
	128
	129	static void __init invoke_init_callbacks(void)
	130	{
	131	int i;
	132	int entries = ARRAY_SIZE(page_ext_ops);
	133
	134	for (i = 0; i < entries; i++) {
	135	if (page_ext_ops[i]->init)
	136	page_ext_ops[i]->init();
	137	}
	138	}
	139
7fb7ab6d ZH	140	#ifndef CONFIG_SPARSEMEM
	141	void __init page_ext_init_flatmem_late(void)
	142	{
	143	invoke_init_callbacks();
	144	}
	145	#endif
	146
980ac167 JK	147	static inline struct page_ext get_entry(void base, unsigned long index)
980ac167 JK	148	{
5556cfe8	149	return base + page_ext_size * index;
980ac167 JK	150	}
980ac167 JK	151
b1d5488a CTK	152	/**
	153	* page_ext_get() - Get the extended information for a page.
	154	* @page: The page we're interested in.
	155	*
	156	* Ensures that the page_ext will remain valid until page_ext_put()
	157	* is called.
	158	*
	159	* Return: NULL if no page_ext exists for this page.
	160	* Context: Any context. Caller may not sleep until they have called
	161	* page_ext_put().
	162	*/
	163	struct page_ext page_ext_get(struct page page)
	164	{
	165	struct page_ext *page_ext;
	166
	167	rcu_read_lock();
	168	page_ext = lookup_page_ext(page);
	169	if (!page_ext) {
	170	rcu_read_unlock();
	171	return NULL;
	172	}
	173
	174	return page_ext;
	175	}
	176
	177	/**
	178	* page_ext_put() - Working with page extended information is done.
ed86b748	179	* @page_ext: Page extended information received from page_ext_get().
b1d5488a CTK	180	*
	181	* The page extended information of the page may not be valid after this
	182	* function is called.
	183	*
	184	* Return: None.
	185	* Context: Any context with corresponding page_ext_get() is called.
	186	*/
	187	void page_ext_put(struct page_ext *page_ext)
	188	{
	189	if (unlikely(!page_ext))
	190	return;
	191
	192	rcu_read_unlock();
	193	}
7fb7ab6d	194	#ifndef CONFIG_SPARSEMEM
eefa864b JK	195
	196
	197	void __meminit pgdat_page_ext_init(struct pglist_data *pgdat)
	198	{
	199	pgdat->node_page_ext = NULL;
	200	}
	201
b1d5488a	202	static struct page_ext lookup_page_ext(const struct page page)
eefa864b JK	203	{
eefa864b JK	204	unsigned long pfn = page_to_pfn(page);
0b06bb3f	205	unsigned long index;
eefa864b JK	206	struct page_ext *base;
eefa864b JK	207
b1d5488a	208	WARN_ON_ONCE(!rcu_read_lock_held());
eefa864b	209	base = NODE_DATA(page_to_nid(page))->node_page_ext;
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 (unlikely(!base))
	217	return NULL;
0b06bb3f	218	index = pfn - round_down(node_start_pfn(page_to_nid(page)),
eefa864b	219	MAX_ORDER_NR_PAGES);
980ac167	220	return get_entry(base, index);
eefa864b JK	221	}
	222
	223	static int __init alloc_node_page_ext(int nid)
	224	{
	225	struct page_ext *base;
	226	unsigned long table_size;
	227	unsigned long nr_pages;
	228
	229	nr_pages = NODE_DATA(nid)->node_spanned_pages;
	230	if (!nr_pages)
	231	return 0;
	232
	233	/*
	234	* Need extra space if node range is not aligned with
	235	* MAX_ORDER_NR_PAGES. When page allocator's buddy algorithm
	236	* checks buddy's status, range could be out of exact node range.
	237	*/
	238	if (!IS_ALIGNED(node_start_pfn(nid), MAX_ORDER_NR_PAGES) \|\|
	239	!IS_ALIGNED(node_end_pfn(nid), MAX_ORDER_NR_PAGES))
	240	nr_pages += MAX_ORDER_NR_PAGES;
	241
5556cfe8	242	table_size = page_ext_size * nr_pages;
eefa864b	243
26fb3dae	244	base = memblock_alloc_try_nid(
eefa864b	245	table_size, PAGE_SIZE, __pa(MAX_DMA_ADDRESS),
97ad1087	246	MEMBLOCK_ALLOC_ACCESSIBLE, nid);
eefa864b JK	247	if (!base)
	248	return -ENOMEM;
	249	NODE_DATA(nid)->node_page_ext = base;
	250	total_usage += table_size;
	251	return 0;
	252	}
	253
	254	void __init page_ext_init_flatmem(void)
	255	{
	256
	257	int nid, fail;
	258
	259	if (!invoke_need_callbacks())
	260	return;
	261
	262	for_each_online_node(nid) {
	263	fail = alloc_node_page_ext(nid);
	264	if (fail)
	265	goto fail;
	266	}
	267	pr_info("allocated %ld bytes of page_ext\n", total_usage);
eefa864b JK	268	return;
	269
	270	fail:
	271	pr_crit("allocation of page_ext failed.\n");
	272	panic("Out of memory");
	273	}
	274
d1fea155	275	#else /* CONFIG_SPARSEMEM */
b1d5488a CTK	276	static bool page_ext_invalid(struct page_ext *page_ext)
	277	{
	278	return !page_ext \|\| (((unsigned long)page_ext & PAGE_EXT_INVALID) == PAGE_EXT_INVALID);
	279	}
eefa864b	280
b1d5488a	281	static struct page_ext lookup_page_ext(const struct page page)
eefa864b JK	282	{
	283	unsigned long pfn = page_to_pfn(page);
	284	struct mem_section *section = __pfn_to_section(pfn);
b1d5488a CTK	285	struct page_ext *page_ext = READ_ONCE(section->page_ext);
	286
	287	WARN_ON_ONCE(!rcu_read_lock_held());
eefa864b JK	288	/*
	289	* The sanity checks the page allocator does upon freeing a
	290	* page can reach here before the page_ext arrays are
	291	* allocated when feeding a range of pages to the allocator
	292	* for the first time during bootup or memory hotplug.
	293	*/
b1d5488a	294	if (page_ext_invalid(page_ext))
eefa864b	295	return NULL;
b1d5488a	296	return get_entry(page_ext, pfn);
eefa864b JK	297	}
	298
	299	static void *__meminit alloc_page_ext(size_t size, int nid)
	300	{
	301	gfp_t flags = GFP_KERNEL \| __GFP_ZERO \| __GFP_NOWARN;
	302	void *addr = NULL;
	303
	304	addr = alloc_pages_exact_nid(nid, size, flags);
	305	if (addr) {
	306	kmemleak_alloc(addr, size, 1, flags);
	307	return addr;
	308	}
	309
b95046b0	310	addr = vzalloc_node(size, nid);
eefa864b JK	311
	312	return addr;
	313	}
	314
	315	static int __meminit init_section_page_ext(unsigned long pfn, int nid)
	316	{
	317	struct mem_section *section;
	318	struct page_ext *base;
	319	unsigned long table_size;
	320
	321	section = __pfn_to_section(pfn);
	322
	323	if (section->page_ext)
	324	return 0;
	325
5556cfe8	326	table_size = page_ext_size * PAGES_PER_SECTION;
eefa864b JK	327	base = alloc_page_ext(table_size, nid);
	328
	329	/*
	330	* The value stored in section->page_ext is (base - pfn)
	331	* and it does not point to the memory block allocated above,
	332	* causing kmemleak false positives.
	333	*/
	334	kmemleak_not_leak(base);
	335
	336	if (!base) {
	337	pr_err("page ext allocation failure\n");
	338	return -ENOMEM;
	339	}
	340
	341	/*
	342	* The passed "pfn" may not be aligned to SECTION. For the calculation
	343	* we need to apply a mask.
	344	*/
	345	pfn &= PAGE_SECTION_MASK;
5556cfe8	346	section->page_ext = (void )base - page_ext_size pfn;
eefa864b JK	347	total_usage += table_size;
	348	return 0;
	349	}
76af6a05	350
eefa864b JK	351	static void free_page_ext(void *addr)
	352	{
	353	if (is_vmalloc_addr(addr)) {
	354	vfree(addr);
	355	} else {
	356	struct page *page = virt_to_page(addr);
	357	size_t table_size;
	358
5556cfe8	359	table_size = page_ext_size * PAGES_PER_SECTION;
eefa864b JK	360
eefa864b JK	361	BUG_ON(PageReserved(page));
0c815854	362	kmemleak_free(addr);
eefa864b JK	363	free_pages_exact(addr, table_size);
	364	}
	365	}
	366
	367	static void __free_page_ext(unsigned long pfn)
	368	{
	369	struct mem_section *ms;
	370	struct page_ext *base;
	371
	372	ms = __pfn_to_section(pfn);
	373	if (!ms \|\| !ms->page_ext)
	374	return;
b1d5488a CTK	375
	376	base = READ_ONCE(ms->page_ext);
	377	/*
	378	* page_ext here can be valid while doing the roll back
	379	* operation in online_page_ext().
	380	*/
	381	if (page_ext_invalid(base))
	382	base = (void *)base - PAGE_EXT_INVALID;
	383	WRITE_ONCE(ms->page_ext, NULL);
	384
	385	base = get_entry(base, pfn);
eefa864b	386	free_page_ext(base);
b1d5488a CTK	387	}
	388
	389	static void __invalidate_page_ext(unsigned long pfn)
	390	{
	391	struct mem_section *ms;
	392	void *val;
	393
	394	ms = __pfn_to_section(pfn);
	395	if (!ms \|\| !ms->page_ext)
	396	return;
	397	val = (void *)ms->page_ext + PAGE_EXT_INVALID;
	398	WRITE_ONCE(ms->page_ext, val);
eefa864b JK	399	}
	400
	401	static int __meminit online_page_ext(unsigned long start_pfn,
	402	unsigned long nr_pages,
	403	int nid)
	404	{
	405	unsigned long start, end, pfn;
	406	int fail = 0;
	407
	408	start = SECTION_ALIGN_DOWN(start_pfn);
	409	end = SECTION_ALIGN_UP(start_pfn + nr_pages);
	410
98fa15f3	411	if (nid == NUMA_NO_NODE) {
eefa864b JK	412	/*
	413	* In this case, "nid" already exists and contains valid memory.
	414	* "start_pfn" passed to us is a pfn which is an arg for
	415	* online__pages(), and start_pfn should exist.
	416	*/
	417	nid = pfn_to_nid(start_pfn);
30a51400	418	VM_BUG_ON(!node_online(nid));
eefa864b JK	419	}
eefa864b JK	420
dccacf8d	421	for (pfn = start; !fail && pfn < end; pfn += PAGES_PER_SECTION)
eefa864b	422	fail = init_section_page_ext(pfn, nid);
eefa864b JK	423	if (!fail)
	424	return 0;
	425
	426	/* rollback */
	427	for (pfn = start; pfn < end; pfn += PAGES_PER_SECTION)
	428	__free_page_ext(pfn);
	429
	430	return -ENOMEM;
	431	}
	432
	433	static int __meminit offline_page_ext(unsigned long start_pfn,
7b5a0b66	434	unsigned long nr_pages)
eefa864b JK	435	{
	436	unsigned long start, end, pfn;
	437
	438	start = SECTION_ALIGN_DOWN(start_pfn);
	439	end = SECTION_ALIGN_UP(start_pfn + nr_pages);
	440
b1d5488a CTK	441	/*
	442	* Freeing of page_ext is done in 3 steps to avoid
	443	* use-after-free of it:
	444	* 1) Traverse all the sections and mark their page_ext
	445	* as invalid.
	446	* 2) Wait for all the existing users of page_ext who
	447	* started before invalidation to finish.
	448	* 3) Free the page_ext.
	449	*/
	450	for (pfn = start; pfn < end; pfn += PAGES_PER_SECTION)
	451	__invalidate_page_ext(pfn);
	452
	453	synchronize_rcu();
	454
eefa864b JK	455	for (pfn = start; pfn < end; pfn += PAGES_PER_SECTION)
	456	__free_page_ext(pfn);
	457	return 0;
	458
	459	}
	460
	461	static int __meminit page_ext_callback(struct notifier_block *self,
	462	unsigned long action, void *arg)
	463	{
	464	struct memory_notify *mn = arg;
	465	int ret = 0;
	466
	467	switch (action) {
	468	case MEM_GOING_ONLINE:
	469	ret = online_page_ext(mn->start_pfn,
	470	mn->nr_pages, mn->status_change_nid);
	471	break;
	472	case MEM_OFFLINE:
	473	offline_page_ext(mn->start_pfn,
7b5a0b66	474	mn->nr_pages);
eefa864b JK	475	break;
	476	case MEM_CANCEL_ONLINE:
	477	offline_page_ext(mn->start_pfn,
7b5a0b66	478	mn->nr_pages);
eefa864b JK	479	break;
	480	case MEM_GOING_OFFLINE:
	481	break;
	482	case MEM_ONLINE:
	483	case MEM_CANCEL_OFFLINE:
	484	break;
	485	}
	486
	487	return notifier_from_errno(ret);
	488	}
	489
eefa864b JK	490	void __init page_ext_init(void)
	491	{
	492	unsigned long pfn;
	493	int nid;
	494
	495	if (!invoke_need_callbacks())
	496	return;
	497
	498	for_each_node_state(nid, N_MEMORY) {
	499	unsigned long start_pfn, end_pfn;
	500
	501	start_pfn = node_start_pfn(nid);
	502	end_pfn = node_end_pfn(nid);
	503	/*
	504	* start_pfn and end_pfn may not be aligned to SECTION and the
	505	* page->flags of out of node pages are not initialized. So we
	506	* scan [start_pfn, the biggest section's pfn < end_pfn) here.
	507	*/
	508	for (pfn = start_pfn; pfn < end_pfn;
	509	pfn = ALIGN(pfn + 1, PAGES_PER_SECTION)) {
	510
	511	if (!pfn_valid(pfn))
	512	continue;
	513	/*
	514	* Nodes's pfns can be overlapping.
	515	* We know some arch can have a nodes layout such as
	516	* -------------pfn-------------->
	517	* N0 \| N1 \| N2 \| N0 \| N1 \| N2\|....
	518	*/
2f1ee091	519	if (pfn_to_nid(pfn) != nid)
eefa864b JK	520	continue;
	521	if (init_section_page_ext(pfn, nid))
	522	goto oom;
0fc542b7	523	cond_resched();
eefa864b JK	524	}
eefa864b JK	525	}
1eeaa4fd	526	hotplug_memory_notifier(page_ext_callback, DEFAULT_CALLBACK_PRI);
eefa864b JK	527	pr_info("allocated %ld bytes of page_ext\n", total_usage);
	528	invoke_init_callbacks();
	529	return;
	530
	531	oom:
	532	panic("Out of memory");
	533	}
	534
	535	void __meminit pgdat_page_ext_init(struct pglist_data *pgdat)
	536	{
	537	}
	538
	539	#endif