[mirror_ubuntu-bionic-kernel.git] / mm / hugetlb.c

/*
 * Generic hugetlb support.
 * (C) William Irwin, April 2004
 */
#include <linux/gfp.h>
#include <linux/list.h>
#include <linux/init.h>
#include <linux/module.h>
#include <linux/mm.h>
#include <linux/sysctl.h>
#include <linux/highmem.h>
#include <linux/nodemask.h>
#include <linux/pagemap.h>
#include <asm/page.h>
#include <asm/pgtable.h>

#include <linux/hugetlb.h>

const unsigned long hugetlb_zero = 0, hugetlb_infinity = ~0UL;
static unsigned long nr_huge_pages, free_huge_pages;
unsigned long max_huge_pages;
static struct list_head hugepage_freelists[MAX_NUMNODES];
static unsigned int nr_huge_pages_node[MAX_NUMNODES];
static unsigned int free_huge_pages_node[MAX_NUMNODES];

/*
 * Protects updates to hugepage_freelists, nr_huge_pages, and free_huge_pages
 */
static DEFINE_SPINLOCK(hugetlb_lock);

static void enqueue_huge_page(struct page *page)
{
	int nid = page_to_nid(page);
	list_add(&page->lru, &hugepage_freelists[nid]);
	free_huge_pages++;
	free_huge_pages_node[nid]++;
}

static struct page *dequeue_huge_page(void)
{
	int nid = numa_node_id();
	struct page *page = NULL;

	if (list_empty(&hugepage_freelists[nid])) {
		for (nid = 0; nid < MAX_NUMNODES; ++nid)
			if (!list_empty(&hugepage_freelists[nid]))
				break;
	}
	if (nid >= 0 && nid < MAX_NUMNODES &&
	    !list_empty(&hugepage_freelists[nid])) {
		page = list_entry(hugepage_freelists[nid].next,
				  struct page, lru);
		list_del(&page->lru);
		free_huge_pages--;
		free_huge_pages_node[nid]--;
	}
	return page;
}

static struct page *alloc_fresh_huge_page(void)
{
	static int nid = 0;
	struct page *page;
	page = alloc_pages_node(nid, GFP_HIGHUSER|__GFP_COMP|__GFP_NOWARN,
					HUGETLB_PAGE_ORDER);
	nid = (nid + 1) % num_online_nodes();
	if (page) {
		spin_lock(&hugetlb_lock);
		nr_huge_pages++;
		nr_huge_pages_node[page_to_nid(page)]++;
		spin_unlock(&hugetlb_lock);
	}
	return page;
}

void free_huge_page(struct page *page)
{
	BUG_ON(page_count(page));

	INIT_LIST_HEAD(&page->lru);
	page[1].mapping = NULL;

	spin_lock(&hugetlb_lock);
	enqueue_huge_page(page);
	spin_unlock(&hugetlb_lock);
}

struct page *alloc_huge_page(void)
{
	struct page *page;
	int i;

	spin_lock(&hugetlb_lock);
	page = dequeue_huge_page();
	if (!page) {
		spin_unlock(&hugetlb_lock);
		return NULL;
	}
	spin_unlock(&hugetlb_lock);
	set_page_count(page, 1);
	page[1].mapping = (void *)free_huge_page;
	for (i = 0; i < (HPAGE_SIZE/PAGE_SIZE); ++i)
		clear_highpage(&page[i]);
	return page;
}

static int __init hugetlb_init(void)
{
	unsigned long i;
	struct page *page;

	if (HPAGE_SHIFT == 0)
		return 0;

	for (i = 0; i < MAX_NUMNODES; ++i)
		INIT_LIST_HEAD(&hugepage_freelists[i]);

	for (i = 0; i < max_huge_pages; ++i) {
		page = alloc_fresh_huge_page();
		if (!page)
			break;
		spin_lock(&hugetlb_lock);
		enqueue_huge_page(page);
		spin_unlock(&hugetlb_lock);
	}
	max_huge_pages = free_huge_pages = nr_huge_pages = i;
	printk("Total HugeTLB memory allocated, %ld\n", free_huge_pages);
	return 0;
}
module_init(hugetlb_init);

static int __init hugetlb_setup(char *s)
{
	if (sscanf(s, "%lu", &max_huge_pages) <= 0)
		max_huge_pages = 0;
	return 1;
}
__setup("hugepages=", hugetlb_setup);

#ifdef CONFIG_SYSCTL
static void update_and_free_page(struct page *page)
{
	int i;
	nr_huge_pages--;
	nr_huge_pages_node[page_zone(page)->zone_pgdat->node_id]--;
	for (i = 0; i < (HPAGE_SIZE / PAGE_SIZE); i++) {
		page[i].flags &= ~(1 << PG_locked | 1 << PG_error | 1 << PG_referenced |
				1 << PG_dirty | 1 << PG_active | 1 << PG_reserved |
				1 << PG_private | 1<< PG_writeback);
		set_page_count(&page[i], 0);
	}
	set_page_count(page, 1);
	__free_pages(page, HUGETLB_PAGE_ORDER);
}

#ifdef CONFIG_HIGHMEM
static void try_to_free_low(unsigned long count)
{
	int i, nid;
	for (i = 0; i < MAX_NUMNODES; ++i) {
		struct page *page, *next;
		list_for_each_entry_safe(page, next, &hugepage_freelists[i], lru) {
			if (PageHighMem(page))
				continue;
			list_del(&page->lru);
			update_and_free_page(page);
			nid = page_zone(page)->zone_pgdat->node_id;
			free_huge_pages--;
			free_huge_pages_node[nid]--;
			if (count >= nr_huge_pages)
				return;
		}
	}
}
#else
static inline void try_to_free_low(unsigned long count)
{
}
#endif

static unsigned long set_max_huge_pages(unsigned long count)
{
	while (count > nr_huge_pages) {
		struct page *page = alloc_fresh_huge_page();
		if (!page)
			return nr_huge_pages;
		spin_lock(&hugetlb_lock);
		enqueue_huge_page(page);
		spin_unlock(&hugetlb_lock);
	}
	if (count >= nr_huge_pages)
		return nr_huge_pages;

	spin_lock(&hugetlb_lock);
	try_to_free_low(count);
	while (count < nr_huge_pages) {
		struct page *page = dequeue_huge_page();
		if (!page)
			break;
		update_and_free_page(page);
	}
	spin_unlock(&hugetlb_lock);
	return nr_huge_pages;
}

int hugetlb_sysctl_handler(struct ctl_table *table, int write,
			   struct file *file, void __user *buffer,
			   size_t *length, loff_t *ppos)
{
	proc_doulongvec_minmax(table, write, file, buffer, length, ppos);
	max_huge_pages = set_max_huge_pages(max_huge_pages);
	return 0;
}
#endif /* CONFIG_SYSCTL */

int hugetlb_report_meminfo(char *buf)
{
	return sprintf(buf,
			"HugePages_Total: %5lu\n"
			"HugePages_Free:  %5lu\n"
			"Hugepagesize:    %5lu kB\n",
			nr_huge_pages,
			free_huge_pages,
			HPAGE_SIZE/1024);
}

int hugetlb_report_node_meminfo(int nid, char *buf)
{
	return sprintf(buf,
		"Node %d HugePages_Total: %5u\n"
		"Node %d HugePages_Free:  %5u\n",
		nid, nr_huge_pages_node[nid],
		nid, free_huge_pages_node[nid]);
}

int is_hugepage_mem_enough(size_t size)
{
	return (size + ~HPAGE_MASK)/HPAGE_SIZE <= free_huge_pages;
}

/* Return the number pages of memory we physically have, in PAGE_SIZE units. */
unsigned long hugetlb_total_pages(void)
{
	return nr_huge_pages * (HPAGE_SIZE / PAGE_SIZE);
}

/*
 * We cannot handle pagefaults against hugetlb pages at all.  They cause
 * handle_mm_fault() to try to instantiate regular-sized pages in the
 * hugegpage VMA.  do_page_fault() is supposed to trap this, so BUG is we get
 * this far.
 */
static struct page *hugetlb_nopage(struct vm_area_struct *vma,
				unsigned long address, int *unused)
{
	BUG();
	return NULL;
}

struct vm_operations_struct hugetlb_vm_ops = {
	.nopage = hugetlb_nopage,
};

static pte_t make_huge_pte(struct vm_area_struct *vma, struct page *page,
				int writable)
{
	pte_t entry;

	if (writable) {
		entry =
		    pte_mkwrite(pte_mkdirty(mk_pte(page, vma->vm_page_prot)));
	} else {
		entry = pte_wrprotect(mk_pte(page, vma->vm_page_prot));
	}
	entry = pte_mkyoung(entry);
	entry = pte_mkhuge(entry);

	return entry;
}

static void set_huge_ptep_writable(struct vm_area_struct *vma,
				   unsigned long address, pte_t *ptep)
{
	pte_t entry;

	entry = pte_mkwrite(pte_mkdirty(*ptep));
	ptep_set_access_flags(vma, address, ptep, entry, 1);
	update_mmu_cache(vma, address, entry);
	lazy_mmu_prot_update(entry);
}


int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
			    struct vm_area_struct *vma)
{
	pte_t *src_pte, *dst_pte, entry;
	struct page *ptepage;
	unsigned long addr;
	int cow;

	cow = (vma->vm_flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;

	for (addr = vma->vm_start; addr < vma->vm_end; addr += HPAGE_SIZE) {
		src_pte = huge_pte_offset(src, addr);
		if (!src_pte)
			continue;
		dst_pte = huge_pte_alloc(dst, addr);
		if (!dst_pte)
			goto nomem;
		spin_lock(&dst->page_table_lock);
		spin_lock(&src->page_table_lock);
		if (!pte_none(*src_pte)) {
			if (cow)
				ptep_set_wrprotect(src, addr, src_pte);
			entry = *src_pte;
			ptepage = pte_page(entry);
			get_page(ptepage);
			add_mm_counter(dst, file_rss, HPAGE_SIZE / PAGE_SIZE);
			set_huge_pte_at(dst, addr, dst_pte, entry);
		}
		spin_unlock(&src->page_table_lock);
		spin_unlock(&dst->page_table_lock);
	}
	return 0;

nomem:
	return -ENOMEM;
}

void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
			  unsigned long end)
{
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
	pte_t *ptep;
	pte_t pte;
	struct page *page;

	WARN_ON(!is_vm_hugetlb_page(vma));
	BUG_ON(start & ~HPAGE_MASK);
	BUG_ON(end & ~HPAGE_MASK);

	spin_lock(&mm->page_table_lock);

	/* Update high watermark before we lower rss */
	update_hiwater_rss(mm);

	for (address = start; address < end; address += HPAGE_SIZE) {
		ptep = huge_pte_offset(mm, address);
		if (!ptep)
			continue;

		pte = huge_ptep_get_and_clear(mm, address, ptep);
		if (pte_none(pte))
			continue;

		page = pte_page(pte);
		put_page(page);
		add_mm_counter(mm, file_rss, (int) -(HPAGE_SIZE / PAGE_SIZE));
	}

	spin_unlock(&mm->page_table_lock);
	flush_tlb_range(vma, start, end);
}

static struct page *find_or_alloc_huge_page(struct address_space *mapping,
				unsigned long idx, int shared)
{
	struct page *page;
	int err;

retry:
	page = find_lock_page(mapping, idx);
	if (page)
		goto out;

	if (hugetlb_get_quota(mapping))
		goto out;
	page = alloc_huge_page();
	if (!page) {
		hugetlb_put_quota(mapping);
		goto out;
	}

	if (shared) {
		err = add_to_page_cache(page, mapping, idx, GFP_KERNEL);
		if (err) {
			put_page(page);
			hugetlb_put_quota(mapping);
			if (err == -EEXIST)
				goto retry;
			page = NULL;
		}
	} else {
		/* Caller expects a locked page */
		lock_page(page);
	}
out:
	return page;
}

static int hugetlb_cow(struct mm_struct *mm, struct vm_area_struct *vma,
			unsigned long address, pte_t *ptep, pte_t pte)
{
	struct page *old_page, *new_page;
	int i, avoidcopy;

	old_page = pte_page(pte);

	/* If no-one else is actually using this page, avoid the copy
	 * and just make the page writable */
	avoidcopy = (page_count(old_page) == 1);
	if (avoidcopy) {
		set_huge_ptep_writable(vma, address, ptep);
		return VM_FAULT_MINOR;
	}

	page_cache_get(old_page);
	new_page = alloc_huge_page();

	if (!new_page) {
		page_cache_release(old_page);

		/* Logically this is OOM, not a SIGBUS, but an OOM
		 * could cause the kernel to go killing other
		 * processes which won't help the hugepage situation
		 * at all (?) */
		return VM_FAULT_SIGBUS;
	}

	spin_unlock(&mm->page_table_lock);
	for (i = 0; i < HPAGE_SIZE/PAGE_SIZE; i++)
		copy_user_highpage(new_page + i, old_page + i,
				   address + i*PAGE_SIZE);
	spin_lock(&mm->page_table_lock);

	ptep = huge_pte_offset(mm, address & HPAGE_MASK);
	if (likely(pte_same(*ptep, pte))) {
		/* Break COW */
		set_huge_pte_at(mm, address, ptep,
				make_huge_pte(vma, new_page, 1));
		/* Make the old page be freed below */
		new_page = old_page;
	}
	page_cache_release(new_page);
	page_cache_release(old_page);
	return VM_FAULT_MINOR;
}

int hugetlb_no_page(struct mm_struct *mm, struct vm_area_struct *vma,
			unsigned long address, pte_t *ptep, int write_access)
{
	int ret = VM_FAULT_SIGBUS;
	unsigned long idx;
	unsigned long size;
	struct page *page;
	struct address_space *mapping;
	pte_t new_pte;

	mapping = vma->vm_file->f_mapping;
	idx = ((address - vma->vm_start) >> HPAGE_SHIFT)
		+ (vma->vm_pgoff >> (HPAGE_SHIFT - PAGE_SHIFT));

	/*
	 * Use page lock to guard against racing truncation
	 * before we get page_table_lock.
	 */
	page = find_or_alloc_huge_page(mapping, idx,
			vma->vm_flags & VM_SHARED);
	if (!page)
		goto out;

	BUG_ON(!PageLocked(page));

	spin_lock(&mm->page_table_lock);
	size = i_size_read(mapping->host) >> HPAGE_SHIFT;
	if (idx >= size)
		goto backout;

	ret = VM_FAULT_MINOR;
	if (!pte_none(*ptep))
		goto backout;

	add_mm_counter(mm, file_rss, HPAGE_SIZE / PAGE_SIZE);
	new_pte = make_huge_pte(vma, page, ((vma->vm_flags & VM_WRITE)
				&& (vma->vm_flags & VM_SHARED)));
	set_huge_pte_at(mm, address, ptep, new_pte);

	if (write_access && !(vma->vm_flags & VM_SHARED)) {
		/* Optimization, do the COW without a second fault */
		ret = hugetlb_cow(mm, vma, address, ptep, new_pte);
	}

	spin_unlock(&mm->page_table_lock);
	unlock_page(page);
out:
	return ret;

backout:
	spin_unlock(&mm->page_table_lock);
	hugetlb_put_quota(mapping);
	unlock_page(page);
	put_page(page);
	goto out;
}

int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
			unsigned long address, int write_access)
{
	pte_t *ptep;
	pte_t entry;
	int ret;

	ptep = huge_pte_alloc(mm, address);
	if (!ptep)
		return VM_FAULT_OOM;

	entry = *ptep;
	if (pte_none(entry))
		return hugetlb_no_page(mm, vma, address, ptep, write_access);

	ret = VM_FAULT_MINOR;

	spin_lock(&mm->page_table_lock);
	/* Check for a racing update before calling hugetlb_cow */
	if (likely(pte_same(entry, *ptep)))
		if (write_access && !pte_write(entry))
			ret = hugetlb_cow(mm, vma, address, ptep, entry);
	spin_unlock(&mm->page_table_lock);

	return ret;
}

int follow_hugetlb_page(struct mm_struct *mm, struct vm_area_struct *vma,
			struct page **pages, struct vm_area_struct **vmas,
			unsigned long *position, int *length, int i)
{
	unsigned long vpfn, vaddr = *position;
	int remainder = *length;

	vpfn = vaddr/PAGE_SIZE;
	spin_lock(&mm->page_table_lock);
	while (vaddr < vma->vm_end && remainder) {
		pte_t *pte;
		struct page *page;

		/*
		 * Some archs (sparc64, sh*) have multiple pte_ts to
		 * each hugepage.  We have to make * sure we get the
		 * first, for the page indexing below to work.
		 */
		pte = huge_pte_offset(mm, vaddr & HPAGE_MASK);

		if (!pte || pte_none(*pte)) {
			int ret;

			spin_unlock(&mm->page_table_lock);
			ret = hugetlb_fault(mm, vma, vaddr, 0);
			spin_lock(&mm->page_table_lock);
			if (ret == VM_FAULT_MINOR)
				continue;

			remainder = 0;
			if (!i)
				i = -EFAULT;
			break;
		}

		if (pages) {
			page = &pte_page(*pte)[vpfn % (HPAGE_SIZE/PAGE_SIZE)];
			get_page(page);
			pages[i] = page;
		}

		if (vmas)
			vmas[i] = vma;

		vaddr += PAGE_SIZE;
		++vpfn;
		--remainder;
		++i;
	}
	spin_unlock(&mm->page_table_lock);
	*length = remainder;
	*position = vaddr;

	return i;
}
Commit	Line	Data
1da177e4 LT	1	/*
	2	* Generic hugetlb support.
	3	* (C) William Irwin, April 2004
	4	*/
	5	#include <linux/gfp.h>
	6	#include <linux/list.h>
	7	#include <linux/init.h>
	8	#include <linux/module.h>
	9	#include <linux/mm.h>
1da177e4 LT	10	#include <linux/sysctl.h>
	11	#include <linux/highmem.h>
	12	#include <linux/nodemask.h>
63551ae0 DG	13	#include <linux/pagemap.h>
	14	#include <asm/page.h>
	15	#include <asm/pgtable.h>
	16
	17	#include <linux/hugetlb.h>
1da177e4 LT	18
	19	const unsigned long hugetlb_zero = 0, hugetlb_infinity = ~0UL;
	20	static unsigned long nr_huge_pages, free_huge_pages;
	21	unsigned long max_huge_pages;
	22	static struct list_head hugepage_freelists[MAX_NUMNODES];
	23	static unsigned int nr_huge_pages_node[MAX_NUMNODES];
	24	static unsigned int free_huge_pages_node[MAX_NUMNODES];
0bd0f9fb EP	25
	26	/*
	27	* Protects updates to hugepage_freelists, nr_huge_pages, and free_huge_pages
	28	*/
1da177e4 LT	29	static DEFINE_SPINLOCK(hugetlb_lock);
	30
	31	static void enqueue_huge_page(struct page *page)
	32	{
	33	int nid = page_to_nid(page);
	34	list_add(&page->lru, &hugepage_freelists[nid]);
	35	free_huge_pages++;
	36	free_huge_pages_node[nid]++;
	37	}
	38
	39	static struct page *dequeue_huge_page(void)
	40	{
	41	int nid = numa_node_id();
	42	struct page *page = NULL;
	43
	44	if (list_empty(&hugepage_freelists[nid])) {
	45	for (nid = 0; nid < MAX_NUMNODES; ++nid)
	46	if (!list_empty(&hugepage_freelists[nid]))
	47	break;
	48	}
	49	if (nid >= 0 && nid < MAX_NUMNODES &&
	50	!list_empty(&hugepage_freelists[nid])) {
	51	page = list_entry(hugepage_freelists[nid].next,
	52	struct page, lru);
	53	list_del(&page->lru);
	54	free_huge_pages--;
	55	free_huge_pages_node[nid]--;
	56	}
	57	return page;
	58	}
	59
	60	static struct page *alloc_fresh_huge_page(void)
	61	{
	62	static int nid = 0;
	63	struct page *page;
	64	page = alloc_pages_node(nid, GFP_HIGHUSER\|__GFP_COMP\|__GFP_NOWARN,
	65	HUGETLB_PAGE_ORDER);
	66	nid = (nid + 1) % num_online_nodes();
	67	if (page) {
0bd0f9fb	68	spin_lock(&hugetlb_lock);
1da177e4 LT	69	nr_huge_pages++;
1da177e4 LT	70	nr_huge_pages_node[page_to_nid(page)]++;
0bd0f9fb	71	spin_unlock(&hugetlb_lock);
1da177e4 LT	72	}
	73	return page;
	74	}
	75
	76	void free_huge_page(struct page *page)
	77	{
	78	BUG_ON(page_count(page));
	79
	80	INIT_LIST_HEAD(&page->lru);
	81	page[1].mapping = NULL;
	82
	83	spin_lock(&hugetlb_lock);
	84	enqueue_huge_page(page);
	85	spin_unlock(&hugetlb_lock);
	86	}
	87
	88	struct page *alloc_huge_page(void)
	89	{
	90	struct page *page;
	91	int i;
	92
	93	spin_lock(&hugetlb_lock);
	94	page = dequeue_huge_page();
	95	if (!page) {
	96	spin_unlock(&hugetlb_lock);
	97	return NULL;
	98	}
	99	spin_unlock(&hugetlb_lock);
	100	set_page_count(page, 1);
	101	page[1].mapping = (void *)free_huge_page;
	102	for (i = 0; i < (HPAGE_SIZE/PAGE_SIZE); ++i)
	103	clear_highpage(&page[i]);
	104	return page;
	105	}
	106
	107	static int __init hugetlb_init(void)
	108	{
	109	unsigned long i;
	110	struct page *page;
	111
3c726f8d BH	112	if (HPAGE_SHIFT == 0)
	113	return 0;
	114
1da177e4 LT	115	for (i = 0; i < MAX_NUMNODES; ++i)
	116	INIT_LIST_HEAD(&hugepage_freelists[i]);
	117
	118	for (i = 0; i < max_huge_pages; ++i) {
	119	page = alloc_fresh_huge_page();
	120	if (!page)
	121	break;
	122	spin_lock(&hugetlb_lock);
	123	enqueue_huge_page(page);
	124	spin_unlock(&hugetlb_lock);
	125	}
	126	max_huge_pages = free_huge_pages = nr_huge_pages = i;
	127	printk("Total HugeTLB memory allocated, %ld\n", free_huge_pages);
	128	return 0;
	129	}
	130	module_init(hugetlb_init);
	131
	132	static int __init hugetlb_setup(char *s)
	133	{
	134	if (sscanf(s, "%lu", &max_huge_pages) <= 0)
	135	max_huge_pages = 0;
	136	return 1;
	137	}
	138	__setup("hugepages=", hugetlb_setup);
	139
	140	#ifdef CONFIG_SYSCTL
	141	static void update_and_free_page(struct page *page)
	142	{
	143	int i;
	144	nr_huge_pages--;
	145	nr_huge_pages_node[page_zone(page)->zone_pgdat->node_id]--;
	146	for (i = 0; i < (HPAGE_SIZE / PAGE_SIZE); i++) {
	147	page[i].flags &= ~(1 << PG_locked \| 1 << PG_error \| 1 << PG_referenced \|
	148	1 << PG_dirty \| 1 << PG_active \| 1 << PG_reserved \|
	149	1 << PG_private \| 1<< PG_writeback);
	150	set_page_count(&page[i], 0);
	151	}
	152	set_page_count(page, 1);
	153	__free_pages(page, HUGETLB_PAGE_ORDER);
	154	}
	155
	156	#ifdef CONFIG_HIGHMEM
	157	static void try_to_free_low(unsigned long count)
	158	{
	159	int i, nid;
	160	for (i = 0; i < MAX_NUMNODES; ++i) {
	161	struct page page, next;
	162	list_for_each_entry_safe(page, next, &hugepage_freelists[i], lru) {
	163	if (PageHighMem(page))
	164	continue;
	165	list_del(&page->lru);
	166	update_and_free_page(page);
	167	nid = page_zone(page)->zone_pgdat->node_id;
	168	free_huge_pages--;
	169	free_huge_pages_node[nid]--;
	170	if (count >= nr_huge_pages)
	171	return;
	172	}
	173	}
	174	}
	175	#else
	176	static inline void try_to_free_low(unsigned long count)
	177	{
	178	}
179	#endif
180
181	static unsigned long set_max_huge_pages(unsigned long count)
182	{
183	while (count > nr_huge_pages) {
184	struct page *page = alloc_fresh_huge_page();
185	if (!page)
186	return nr_huge_pages;
187	spin_lock(&hugetlb_lock);
188	enqueue_huge_page(page);
189	spin_unlock(&hugetlb_lock);
190	}
191	if (count >= nr_huge_pages)
192	return nr_huge_pages;
193
194	spin_lock(&hugetlb_lock);
195	try_to_free_low(count);
196	while (count < nr_huge_pages) {
197	struct page *page = dequeue_huge_page();
198	if (!page)
199	break;
200	update_and_free_page(page);
201	}
202	spin_unlock(&hugetlb_lock);
203	return nr_huge_pages;
204	}
205
206	int hugetlb_sysctl_handler(struct ctl_table *table, int write,
207	struct file file, void __user buffer,
208	size_t length, loff_t ppos)
209	{
210	proc_doulongvec_minmax(table, write, file, buffer, length, ppos);
211	max_huge_pages = set_max_huge_pages(max_huge_pages);
212	return 0;
213	}
214	#endif /* CONFIG_SYSCTL */
215
216	int hugetlb_report_meminfo(char *buf)
217	{
218	return sprintf(buf,
219	"HugePages_Total: %5lu\n"
220	"HugePages_Free: %5lu\n"
221	"Hugepagesize: %5lu kB\n",
222	nr_huge_pages,
223	free_huge_pages,
224	HPAGE_SIZE/1024);
225	}
226
227	int hugetlb_report_node_meminfo(int nid, char *buf)
228	{
229	return sprintf(buf,
230	"Node %d HugePages_Total: %5u\n"
231	"Node %d HugePages_Free: %5u\n",
232	nid, nr_huge_pages_node[nid],
233	nid, free_huge_pages_node[nid]);
234	}
235
236	int is_hugepage_mem_enough(size_t size)
237	{
238	return (size + ~HPAGE_MASK)/HPAGE_SIZE <= free_huge_pages;
239	}
240
241	/* Return the number pages of memory we physically have, in PAGE_SIZE units. */
242	unsigned long hugetlb_total_pages(void)
243	{
244	return nr_huge_pages * (HPAGE_SIZE / PAGE_SIZE);
245	}
1da177e4 LT	246
	247	/*
	248	* We cannot handle pagefaults against hugetlb pages at all. They cause
	249	* handle_mm_fault() to try to instantiate regular-sized pages in the
	250	* hugegpage VMA. do_page_fault() is supposed to trap this, so BUG is we get
	251	* this far.
	252	*/
	253	static struct page hugetlb_nopage(struct vm_area_struct vma,
	254	unsigned long address, int *unused)
	255	{
	256	BUG();
	257	return NULL;
	258	}
	259
	260	struct vm_operations_struct hugetlb_vm_ops = {
	261	.nopage = hugetlb_nopage,
	262	};
	263
1e8f889b DG	264	static pte_t make_huge_pte(struct vm_area_struct vma, struct page page,
1e8f889b DG	265	int writable)
63551ae0 DG	266	{
	267	pte_t entry;
	268
1e8f889b	269	if (writable) {
63551ae0 DG	270	entry =
	271	pte_mkwrite(pte_mkdirty(mk_pte(page, vma->vm_page_prot)));
	272	} else {
	273	entry = pte_wrprotect(mk_pte(page, vma->vm_page_prot));
	274	}
	275	entry = pte_mkyoung(entry);
	276	entry = pte_mkhuge(entry);
	277
	278	return entry;
	279	}
	280
1e8f889b DG	281	static void set_huge_ptep_writable(struct vm_area_struct *vma,
	282	unsigned long address, pte_t *ptep)
	283	{
	284	pte_t entry;
	285
	286	entry = pte_mkwrite(pte_mkdirty(*ptep));
	287	ptep_set_access_flags(vma, address, ptep, entry, 1);
	288	update_mmu_cache(vma, address, entry);
	289	lazy_mmu_prot_update(entry);
	290	}
	291
	292
63551ae0 DG	293	int copy_hugetlb_page_range(struct mm_struct dst, struct mm_struct src,
	294	struct vm_area_struct *vma)
	295	{
	296	pte_t src_pte, dst_pte, entry;
	297	struct page *ptepage;
1c59827d	298	unsigned long addr;
1e8f889b DG	299	int cow;
	300
	301	cow = (vma->vm_flags & (VM_SHARED \| VM_MAYWRITE)) == VM_MAYWRITE;
63551ae0	302
1c59827d	303	for (addr = vma->vm_start; addr < vma->vm_end; addr += HPAGE_SIZE) {
c74df32c HD	304	src_pte = huge_pte_offset(src, addr);
	305	if (!src_pte)
	306	continue;
63551ae0 DG	307	dst_pte = huge_pte_alloc(dst, addr);
	308	if (!dst_pte)
	309	goto nomem;
c74df32c	310	spin_lock(&dst->page_table_lock);
1c59827d	311	spin_lock(&src->page_table_lock);
c74df32c	312	if (!pte_none(*src_pte)) {
1e8f889b DG	313	if (cow)
1e8f889b DG	314	ptep_set_wrprotect(src, addr, src_pte);
1c59827d HD	315	entry = *src_pte;
	316	ptepage = pte_page(entry);
	317	get_page(ptepage);
4294621f	318	add_mm_counter(dst, file_rss, HPAGE_SIZE / PAGE_SIZE);
1c59827d HD	319	set_huge_pte_at(dst, addr, dst_pte, entry);
	320	}
	321	spin_unlock(&src->page_table_lock);
c74df32c	322	spin_unlock(&dst->page_table_lock);
63551ae0 DG	323	}
	324	return 0;
	325
	326	nomem:
	327	return -ENOMEM;
	328	}
	329
	330	void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
	331	unsigned long end)
	332	{
	333	struct mm_struct *mm = vma->vm_mm;
	334	unsigned long address;
c7546f8f	335	pte_t *ptep;
63551ae0 DG	336	pte_t pte;
	337	struct page *page;
	338
	339	WARN_ON(!is_vm_hugetlb_page(vma));
	340	BUG_ON(start & ~HPAGE_MASK);
	341	BUG_ON(end & ~HPAGE_MASK);
	342
508034a3 HD	343	spin_lock(&mm->page_table_lock);
508034a3 HD	344
365e9c87 HD	345	/* Update high watermark before we lower rss */
	346	update_hiwater_rss(mm);
	347
63551ae0	348	for (address = start; address < end; address += HPAGE_SIZE) {
c7546f8f	349	ptep = huge_pte_offset(mm, address);
4c887265	350	if (!ptep)
c7546f8f DG	351	continue;
	352
	353	pte = huge_ptep_get_and_clear(mm, address, ptep);
63551ae0 DG	354	if (pte_none(pte))
63551ae0 DG	355	continue;
c7546f8f	356
63551ae0 DG	357	page = pte_page(pte);
63551ae0 DG	358	put_page(page);
4294621f	359	add_mm_counter(mm, file_rss, (int) -(HPAGE_SIZE / PAGE_SIZE));
63551ae0	360	}
63551ae0	361
1da177e4	362	spin_unlock(&mm->page_table_lock);
508034a3	363	flush_tlb_range(vma, start, end);
1da177e4	364	}
63551ae0	365
85ef47f7	366	static struct page find_or_alloc_huge_page(struct address_space mapping,
1e8f889b	367	unsigned long idx, int shared)
63551ae0	368	{
4c887265 AL	369	struct page *page;
4c887265 AL	370	int err;
4c887265 AL	371
	372	retry:
	373	page = find_lock_page(mapping, idx);
	374	if (page)
	375	goto out;
	376
4c887265 AL	377	if (hugetlb_get_quota(mapping))
	378	goto out;
	379	page = alloc_huge_page();
	380	if (!page) {
	381	hugetlb_put_quota(mapping);
	382	goto out;
	383	}
63551ae0	384
1e8f889b DG	385	if (shared) {
	386	err = add_to_page_cache(page, mapping, idx, GFP_KERNEL);
	387	if (err) {
	388	put_page(page);
	389	hugetlb_put_quota(mapping);
	390	if (err == -EEXIST)
	391	goto retry;
	392	page = NULL;
	393	}
	394	} else {
	395	/* Caller expects a locked page */
	396	lock_page(page);
63551ae0 DG	397	}
63551ae0 DG	398	out:
4c887265	399	return page;
63551ae0 DG	400	}
63551ae0 DG	401
1e8f889b DG	402	static int hugetlb_cow(struct mm_struct mm, struct vm_area_struct vma,
	403	unsigned long address, pte_t *ptep, pte_t pte)
	404	{
	405	struct page old_page, new_page;
	406	int i, avoidcopy;
	407
	408	old_page = pte_page(pte);
	409
	410	/* If no-one else is actually using this page, avoid the copy
	411	* and just make the page writable */
	412	avoidcopy = (page_count(old_page) == 1);
	413	if (avoidcopy) {
	414	set_huge_ptep_writable(vma, address, ptep);
	415	return VM_FAULT_MINOR;
	416	}
	417
	418	page_cache_get(old_page);
	419	new_page = alloc_huge_page();
	420
	421	if (!new_page) {
	422	page_cache_release(old_page);
	423
	424	/* Logically this is OOM, not a SIGBUS, but an OOM
	425	* could cause the kernel to go killing other
	426	* processes which won't help the hugepage situation
	427	* at all (?) */
	428	return VM_FAULT_SIGBUS;
	429	}
	430
	431	spin_unlock(&mm->page_table_lock);
	432	for (i = 0; i < HPAGE_SIZE/PAGE_SIZE; i++)
	433	copy_user_highpage(new_page + i, old_page + i,
	434	address + i*PAGE_SIZE);
	435	spin_lock(&mm->page_table_lock);
	436
	437	ptep = huge_pte_offset(mm, address & HPAGE_MASK);
	438	if (likely(pte_same(*ptep, pte))) {
	439	/* Break COW */
	440	set_huge_pte_at(mm, address, ptep,
	441	make_huge_pte(vma, new_page, 1));
	442	/* Make the old page be freed below */
	443	new_page = old_page;
	444	}
	445	page_cache_release(new_page);
	446	page_cache_release(old_page);
	447	return VM_FAULT_MINOR;
	448	}
	449
86e5216f	450	int hugetlb_no_page(struct mm_struct mm, struct vm_area_struct vma,
1e8f889b	451	unsigned long address, pte_t *ptep, int write_access)
ac9b9c66 HD	452	{
ac9b9c66 HD	453	int ret = VM_FAULT_SIGBUS;
4c887265 AL	454	unsigned long idx;
4c887265 AL	455	unsigned long size;
4c887265 AL	456	struct page *page;
4c887265 AL	457	struct address_space *mapping;
1e8f889b	458	pte_t new_pte;
4c887265	459
4c887265 AL	460	mapping = vma->vm_file->f_mapping;
	461	idx = ((address - vma->vm_start) >> HPAGE_SHIFT)
	462	+ (vma->vm_pgoff >> (HPAGE_SHIFT - PAGE_SHIFT));
	463
	464	/*
	465	* Use page lock to guard against racing truncation
	466	* before we get page_table_lock.
	467	*/
1e8f889b DG	468	page = find_or_alloc_huge_page(mapping, idx,
1e8f889b DG	469	vma->vm_flags & VM_SHARED);
4c887265 AL	470	if (!page)
4c887265 AL	471	goto out;
ac9b9c66	472
1e8f889b DG	473	BUG_ON(!PageLocked(page));
1e8f889b DG	474
ac9b9c66	475	spin_lock(&mm->page_table_lock);
4c887265 AL	476	size = i_size_read(mapping->host) >> HPAGE_SHIFT;
	477	if (idx >= size)
	478	goto backout;
	479
	480	ret = VM_FAULT_MINOR;
86e5216f	481	if (!pte_none(*ptep))
4c887265 AL	482	goto backout;
	483
	484	add_mm_counter(mm, file_rss, HPAGE_SIZE / PAGE_SIZE);
1e8f889b DG	485	new_pte = make_huge_pte(vma, page, ((vma->vm_flags & VM_WRITE)
	486	&& (vma->vm_flags & VM_SHARED)));
	487	set_huge_pte_at(mm, address, ptep, new_pte);
	488
	489	if (write_access && !(vma->vm_flags & VM_SHARED)) {
	490	/* Optimization, do the COW without a second fault */
	491	ret = hugetlb_cow(mm, vma, address, ptep, new_pte);
	492	}
	493
ac9b9c66	494	spin_unlock(&mm->page_table_lock);
4c887265 AL	495	unlock_page(page);
4c887265 AL	496	out:
ac9b9c66	497	return ret;
4c887265 AL	498
	499	backout:
	500	spin_unlock(&mm->page_table_lock);
	501	hugetlb_put_quota(mapping);
	502	unlock_page(page);
	503	put_page(page);
	504	goto out;
ac9b9c66 HD	505	}
ac9b9c66 HD	506
86e5216f AL	507	int hugetlb_fault(struct mm_struct mm, struct vm_area_struct vma,
	508	unsigned long address, int write_access)
	509	{
	510	pte_t *ptep;
	511	pte_t entry;
1e8f889b	512	int ret;
86e5216f AL	513
	514	ptep = huge_pte_alloc(mm, address);
	515	if (!ptep)
	516	return VM_FAULT_OOM;
	517
	518	entry = *ptep;
	519	if (pte_none(entry))
1e8f889b	520	return hugetlb_no_page(mm, vma, address, ptep, write_access);
86e5216f	521
1e8f889b DG	522	ret = VM_FAULT_MINOR;
	523
	524	spin_lock(&mm->page_table_lock);
	525	/* Check for a racing update before calling hugetlb_cow */
	526	if (likely(pte_same(entry, *ptep)))
	527	if (write_access && !pte_write(entry))
	528	ret = hugetlb_cow(mm, vma, address, ptep, entry);
	529	spin_unlock(&mm->page_table_lock);
	530
	531	return ret;
86e5216f AL	532	}
86e5216f AL	533
63551ae0 DG	534	int follow_hugetlb_page(struct mm_struct mm, struct vm_area_struct vma,
	535	struct page pages, struct vm_area_struct vmas,
	536	unsigned long position, int length, int i)
	537	{
	538	unsigned long vpfn, vaddr = *position;
	539	int remainder = *length;
	540
63551ae0	541	vpfn = vaddr/PAGE_SIZE;
1c59827d	542	spin_lock(&mm->page_table_lock);
63551ae0	543	while (vaddr < vma->vm_end && remainder) {
4c887265 AL	544	pte_t *pte;
4c887265 AL	545	struct page *page;
63551ae0	546
4c887265 AL	547	/*
	548	* Some archs (sparc64, sh*) have multiple pte_ts to
	549	* each hugepage. We have to make * sure we get the
	550	* first, for the page indexing below to work.
	551	*/
	552	pte = huge_pte_offset(mm, vaddr & HPAGE_MASK);
63551ae0	553
4c887265 AL	554	if (!pte \|\| pte_none(*pte)) {
4c887265 AL	555	int ret;
63551ae0	556
4c887265 AL	557	spin_unlock(&mm->page_table_lock);
	558	ret = hugetlb_fault(mm, vma, vaddr, 0);
	559	spin_lock(&mm->page_table_lock);
	560	if (ret == VM_FAULT_MINOR)
	561	continue;
63551ae0	562
4c887265 AL	563	remainder = 0;
	564	if (!i)
	565	i = -EFAULT;
	566	break;
	567	}
	568
	569	if (pages) {
	570	page = &pte_page(*pte)[vpfn % (HPAGE_SIZE/PAGE_SIZE)];
63551ae0 DG	571	get_page(page);
	572	pages[i] = page;
	573	}
	574
	575	if (vmas)
	576	vmas[i] = vma;
	577
	578	vaddr += PAGE_SIZE;
	579	++vpfn;
	580	--remainder;
	581	++i;
	582	}
1c59827d	583	spin_unlock(&mm->page_table_lock);
63551ae0 DG	584	*length = remainder;
	585	*position = vaddr;
	586
	587	return i;
	588	}