[linux-2.6-block.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;
	struct zonelist *zonelist = NODE_DATA(nid)->node_zonelists;
	struct zone **z;

	for (z = zonelist->zones; *z; z++) {
		nid = (*z)->zone_pgdat->node_id;
		if (!list_empty(&hugepage_freelists[nid]))
			break;
	}

	if (*z) {
		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;
96df9333 CL	43	struct zonelist *zonelist = NODE_DATA(nid)->node_zonelists;
96df9333 CL	44	struct zone **z;
1da177e4	45
96df9333 CL	46	for (z = zonelist->zones; *z; z++) {
	47	nid = (*z)->zone_pgdat->node_id;
	48	if (!list_empty(&hugepage_freelists[nid]))
	49	break;
1da177e4	50	}
96df9333 CL	51
96df9333 CL	52	if (*z) {
1da177e4 LT	53	page = list_entry(hugepage_freelists[nid].next,
	54	struct page, lru);
	55	list_del(&page->lru);
	56	free_huge_pages--;
	57	free_huge_pages_node[nid]--;
	58	}
	59	return page;
	60	}
	61
	62	static struct page *alloc_fresh_huge_page(void)
	63	{
	64	static int nid = 0;
	65	struct page *page;
	66	page = alloc_pages_node(nid, GFP_HIGHUSER\|__GFP_COMP\|__GFP_NOWARN,
	67	HUGETLB_PAGE_ORDER);
	68	nid = (nid + 1) % num_online_nodes();
	69	if (page) {
0bd0f9fb	70	spin_lock(&hugetlb_lock);
1da177e4 LT	71	nr_huge_pages++;
1da177e4 LT	72	nr_huge_pages_node[page_to_nid(page)]++;
0bd0f9fb	73	spin_unlock(&hugetlb_lock);
1da177e4 LT	74	}
	75	return page;
	76	}
	77
	78	void free_huge_page(struct page *page)
	79	{
	80	BUG_ON(page_count(page));
	81
	82	INIT_LIST_HEAD(&page->lru);
	83	page[1].mapping = NULL;
	84
	85	spin_lock(&hugetlb_lock);
	86	enqueue_huge_page(page);
	87	spin_unlock(&hugetlb_lock);
	88	}
	89
	90	struct page *alloc_huge_page(void)
	91	{
	92	struct page *page;
	93	int i;
	94
	95	spin_lock(&hugetlb_lock);
	96	page = dequeue_huge_page();
	97	if (!page) {
	98	spin_unlock(&hugetlb_lock);
	99	return NULL;
	100	}
	101	spin_unlock(&hugetlb_lock);
	102	set_page_count(page, 1);
	103	page[1].mapping = (void *)free_huge_page;
	104	for (i = 0; i < (HPAGE_SIZE/PAGE_SIZE); ++i)
	105	clear_highpage(&page[i]);
	106	return page;
	107	}
	108
	109	static int __init hugetlb_init(void)
	110	{
	111	unsigned long i;
	112	struct page *page;
	113
3c726f8d BH	114	if (HPAGE_SHIFT == 0)
	115	return 0;
	116
1da177e4 LT	117	for (i = 0; i < MAX_NUMNODES; ++i)
	118	INIT_LIST_HEAD(&hugepage_freelists[i]);
	119
	120	for (i = 0; i < max_huge_pages; ++i) {
	121	page = alloc_fresh_huge_page();
	122	if (!page)
	123	break;
	124	spin_lock(&hugetlb_lock);
	125	enqueue_huge_page(page);
	126	spin_unlock(&hugetlb_lock);
	127	}
	128	max_huge_pages = free_huge_pages = nr_huge_pages = i;
	129	printk("Total HugeTLB memory allocated, %ld\n", free_huge_pages);
	130	return 0;
	131	}
	132	module_init(hugetlb_init);
	133
	134	static int __init hugetlb_setup(char *s)
	135	{
	136	if (sscanf(s, "%lu", &max_huge_pages) <= 0)
	137	max_huge_pages = 0;
	138	return 1;
	139	}
	140	__setup("hugepages=", hugetlb_setup);
	141
	142	#ifdef CONFIG_SYSCTL
	143	static void update_and_free_page(struct page *page)
	144	{
	145	int i;
	146	nr_huge_pages--;
	147	nr_huge_pages_node[page_zone(page)->zone_pgdat->node_id]--;
	148	for (i = 0; i < (HPAGE_SIZE / PAGE_SIZE); i++) {
	149	page[i].flags &= ~(1 << PG_locked \| 1 << PG_error \| 1 << PG_referenced \|
	150	1 << PG_dirty \| 1 << PG_active \| 1 << PG_reserved \|
	151	1 << PG_private \| 1<< PG_writeback);
	152	set_page_count(&page[i], 0);
	153	}
	154	set_page_count(page, 1);
	155	__free_pages(page, HUGETLB_PAGE_ORDER);
	156	}
	157
	158	#ifdef CONFIG_HIGHMEM
	159	static void try_to_free_low(unsigned long count)
	160	{
	161	int i, nid;
	162	for (i = 0; i < MAX_NUMNODES; ++i) {
	163	struct page page, next;
	164	list_for_each_entry_safe(page, next, &hugepage_freelists[i], lru) {
	165	if (PageHighMem(page))
	166	continue;
	167	list_del(&page->lru);
	168	update_and_free_page(page);
	169	nid = page_zone(page)->zone_pgdat->node_id;
	170	free_huge_pages--;
	171	free_huge_pages_node[nid]--;
	172	if (count >= nr_huge_pages)
	173	return;
	174	}
	175	}
	176	}
	177	#else
	178	static inline void try_to_free_low(unsigned long count)
	179	{
	180	}
181	#endif
182
183	static unsigned long set_max_huge_pages(unsigned long count)
184	{
185	while (count > nr_huge_pages) {
186	struct page *page = alloc_fresh_huge_page();
187	if (!page)
188	return nr_huge_pages;
189	spin_lock(&hugetlb_lock);
190	enqueue_huge_page(page);
191	spin_unlock(&hugetlb_lock);
192	}
193	if (count >= nr_huge_pages)
194	return nr_huge_pages;
195
196	spin_lock(&hugetlb_lock);
197	try_to_free_low(count);
198	while (count < nr_huge_pages) {
199	struct page *page = dequeue_huge_page();
200	if (!page)
201	break;
202	update_and_free_page(page);
203	}
204	spin_unlock(&hugetlb_lock);
205	return nr_huge_pages;
206	}
207
208	int hugetlb_sysctl_handler(struct ctl_table *table, int write,
209	struct file file, void __user buffer,
210	size_t length, loff_t ppos)
211	{
212	proc_doulongvec_minmax(table, write, file, buffer, length, ppos);
213	max_huge_pages = set_max_huge_pages(max_huge_pages);
214	return 0;
215	}
216	#endif /* CONFIG_SYSCTL */
217
218	int hugetlb_report_meminfo(char *buf)
219	{
220	return sprintf(buf,
221	"HugePages_Total: %5lu\n"
222	"HugePages_Free: %5lu\n"
223	"Hugepagesize: %5lu kB\n",
224	nr_huge_pages,
225	free_huge_pages,
226	HPAGE_SIZE/1024);
227	}
228
229	int hugetlb_report_node_meminfo(int nid, char *buf)
230	{
231	return sprintf(buf,
232	"Node %d HugePages_Total: %5u\n"
233	"Node %d HugePages_Free: %5u\n",
234	nid, nr_huge_pages_node[nid],
235	nid, free_huge_pages_node[nid]);
236	}
237
238	int is_hugepage_mem_enough(size_t size)
239	{
240	return (size + ~HPAGE_MASK)/HPAGE_SIZE <= free_huge_pages;
241	}
242
243	/* Return the number pages of memory we physically have, in PAGE_SIZE units. */
244	unsigned long hugetlb_total_pages(void)
245	{
246	return nr_huge_pages * (HPAGE_SIZE / PAGE_SIZE);
247	}
1da177e4 LT	248
	249	/*
	250	* We cannot handle pagefaults against hugetlb pages at all. They cause
	251	* handle_mm_fault() to try to instantiate regular-sized pages in the
	252	* hugegpage VMA. do_page_fault() is supposed to trap this, so BUG is we get
	253	* this far.
	254	*/
	255	static struct page hugetlb_nopage(struct vm_area_struct vma,
	256	unsigned long address, int *unused)
	257	{
	258	BUG();
	259	return NULL;
	260	}
	261
	262	struct vm_operations_struct hugetlb_vm_ops = {
	263	.nopage = hugetlb_nopage,
	264	};
	265
1e8f889b DG	266	static pte_t make_huge_pte(struct vm_area_struct vma, struct page page,
1e8f889b DG	267	int writable)
63551ae0 DG	268	{
	269	pte_t entry;
	270
1e8f889b	271	if (writable) {
63551ae0 DG	272	entry =
	273	pte_mkwrite(pte_mkdirty(mk_pte(page, vma->vm_page_prot)));
	274	} else {
	275	entry = pte_wrprotect(mk_pte(page, vma->vm_page_prot));
	276	}
	277	entry = pte_mkyoung(entry);
	278	entry = pte_mkhuge(entry);
	279
	280	return entry;
	281	}
	282
1e8f889b DG	283	static void set_huge_ptep_writable(struct vm_area_struct *vma,
	284	unsigned long address, pte_t *ptep)
	285	{
	286	pte_t entry;
	287
	288	entry = pte_mkwrite(pte_mkdirty(*ptep));
	289	ptep_set_access_flags(vma, address, ptep, entry, 1);
	290	update_mmu_cache(vma, address, entry);
	291	lazy_mmu_prot_update(entry);
	292	}
	293
	294
63551ae0 DG	295	int copy_hugetlb_page_range(struct mm_struct dst, struct mm_struct src,
	296	struct vm_area_struct *vma)
	297	{
	298	pte_t src_pte, dst_pte, entry;
	299	struct page *ptepage;
1c59827d	300	unsigned long addr;
1e8f889b DG	301	int cow;
	302
	303	cow = (vma->vm_flags & (VM_SHARED \| VM_MAYWRITE)) == VM_MAYWRITE;
63551ae0	304
1c59827d	305	for (addr = vma->vm_start; addr < vma->vm_end; addr += HPAGE_SIZE) {
c74df32c HD	306	src_pte = huge_pte_offset(src, addr);
	307	if (!src_pte)
	308	continue;
63551ae0 DG	309	dst_pte = huge_pte_alloc(dst, addr);
	310	if (!dst_pte)
	311	goto nomem;
c74df32c	312	spin_lock(&dst->page_table_lock);
1c59827d	313	spin_lock(&src->page_table_lock);
c74df32c	314	if (!pte_none(*src_pte)) {
1e8f889b DG	315	if (cow)
1e8f889b DG	316	ptep_set_wrprotect(src, addr, src_pte);
1c59827d HD	317	entry = *src_pte;
	318	ptepage = pte_page(entry);
	319	get_page(ptepage);
4294621f	320	add_mm_counter(dst, file_rss, HPAGE_SIZE / PAGE_SIZE);
1c59827d HD	321	set_huge_pte_at(dst, addr, dst_pte, entry);
	322	}
	323	spin_unlock(&src->page_table_lock);
c74df32c	324	spin_unlock(&dst->page_table_lock);
63551ae0 DG	325	}
	326	return 0;
	327
	328	nomem:
	329	return -ENOMEM;
	330	}
	331
	332	void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
	333	unsigned long end)
	334	{
	335	struct mm_struct *mm = vma->vm_mm;
	336	unsigned long address;
c7546f8f	337	pte_t *ptep;
63551ae0 DG	338	pte_t pte;
	339	struct page *page;
	340
	341	WARN_ON(!is_vm_hugetlb_page(vma));
	342	BUG_ON(start & ~HPAGE_MASK);
	343	BUG_ON(end & ~HPAGE_MASK);
	344
508034a3 HD	345	spin_lock(&mm->page_table_lock);
508034a3 HD	346
365e9c87 HD	347	/* Update high watermark before we lower rss */
	348	update_hiwater_rss(mm);
	349
63551ae0	350	for (address = start; address < end; address += HPAGE_SIZE) {
c7546f8f	351	ptep = huge_pte_offset(mm, address);
4c887265	352	if (!ptep)
c7546f8f DG	353	continue;
	354
	355	pte = huge_ptep_get_and_clear(mm, address, ptep);
63551ae0 DG	356	if (pte_none(pte))
63551ae0 DG	357	continue;
c7546f8f	358
63551ae0 DG	359	page = pte_page(pte);
63551ae0 DG	360	put_page(page);
4294621f	361	add_mm_counter(mm, file_rss, (int) -(HPAGE_SIZE / PAGE_SIZE));
63551ae0	362	}
63551ae0	363
1da177e4	364	spin_unlock(&mm->page_table_lock);
508034a3	365	flush_tlb_range(vma, start, end);
1da177e4	366	}
63551ae0	367
85ef47f7	368	static struct page find_or_alloc_huge_page(struct address_space mapping,
1e8f889b	369	unsigned long idx, int shared)
63551ae0	370	{
4c887265 AL	371	struct page *page;
4c887265 AL	372	int err;
4c887265 AL	373
	374	retry:
	375	page = find_lock_page(mapping, idx);
	376	if (page)
	377	goto out;
	378
4c887265 AL	379	if (hugetlb_get_quota(mapping))
	380	goto out;
	381	page = alloc_huge_page();
	382	if (!page) {
	383	hugetlb_put_quota(mapping);
	384	goto out;
	385	}
63551ae0	386
1e8f889b DG	387	if (shared) {
	388	err = add_to_page_cache(page, mapping, idx, GFP_KERNEL);
	389	if (err) {
	390	put_page(page);
	391	hugetlb_put_quota(mapping);
	392	if (err == -EEXIST)
	393	goto retry;
	394	page = NULL;
	395	}
	396	} else {
	397	/* Caller expects a locked page */
	398	lock_page(page);
63551ae0 DG	399	}
63551ae0 DG	400	out:
4c887265	401	return page;
63551ae0 DG	402	}
63551ae0 DG	403
1e8f889b DG	404	static int hugetlb_cow(struct mm_struct mm, struct vm_area_struct vma,
	405	unsigned long address, pte_t *ptep, pte_t pte)
	406	{
	407	struct page old_page, new_page;
	408	int i, avoidcopy;
	409
	410	old_page = pte_page(pte);
	411
	412	/* If no-one else is actually using this page, avoid the copy
	413	* and just make the page writable */
	414	avoidcopy = (page_count(old_page) == 1);
	415	if (avoidcopy) {
	416	set_huge_ptep_writable(vma, address, ptep);
	417	return VM_FAULT_MINOR;
	418	}
	419
	420	page_cache_get(old_page);
	421	new_page = alloc_huge_page();
	422
	423	if (!new_page) {
	424	page_cache_release(old_page);
	425
	426	/* Logically this is OOM, not a SIGBUS, but an OOM
	427	* could cause the kernel to go killing other
	428	* processes which won't help the hugepage situation
	429	* at all (?) */
	430	return VM_FAULT_SIGBUS;
	431	}
	432
	433	spin_unlock(&mm->page_table_lock);
	434	for (i = 0; i < HPAGE_SIZE/PAGE_SIZE; i++)
	435	copy_user_highpage(new_page + i, old_page + i,
	436	address + i*PAGE_SIZE);
	437	spin_lock(&mm->page_table_lock);
	438
	439	ptep = huge_pte_offset(mm, address & HPAGE_MASK);
	440	if (likely(pte_same(*ptep, pte))) {
	441	/* Break COW */
	442	set_huge_pte_at(mm, address, ptep,
	443	make_huge_pte(vma, new_page, 1));
	444	/* Make the old page be freed below */
	445	new_page = old_page;
	446	}
	447	page_cache_release(new_page);
	448	page_cache_release(old_page);
	449	return VM_FAULT_MINOR;
	450	}
	451
86e5216f	452	int hugetlb_no_page(struct mm_struct mm, struct vm_area_struct vma,
1e8f889b	453	unsigned long address, pte_t *ptep, int write_access)
ac9b9c66 HD	454	{
ac9b9c66 HD	455	int ret = VM_FAULT_SIGBUS;
4c887265 AL	456	unsigned long idx;
4c887265 AL	457	unsigned long size;
4c887265 AL	458	struct page *page;
4c887265 AL	459	struct address_space *mapping;
1e8f889b	460	pte_t new_pte;
4c887265	461
4c887265 AL	462	mapping = vma->vm_file->f_mapping;
	463	idx = ((address - vma->vm_start) >> HPAGE_SHIFT)
	464	+ (vma->vm_pgoff >> (HPAGE_SHIFT - PAGE_SHIFT));
	465
	466	/*
	467	* Use page lock to guard against racing truncation
	468	* before we get page_table_lock.
	469	*/
1e8f889b DG	470	page = find_or_alloc_huge_page(mapping, idx,
1e8f889b DG	471	vma->vm_flags & VM_SHARED);
4c887265 AL	472	if (!page)
4c887265 AL	473	goto out;
ac9b9c66	474
1e8f889b DG	475	BUG_ON(!PageLocked(page));
1e8f889b DG	476
ac9b9c66	477	spin_lock(&mm->page_table_lock);
4c887265 AL	478	size = i_size_read(mapping->host) >> HPAGE_SHIFT;
	479	if (idx >= size)
	480	goto backout;
	481
	482	ret = VM_FAULT_MINOR;
86e5216f	483	if (!pte_none(*ptep))
4c887265 AL	484	goto backout;
	485
	486	add_mm_counter(mm, file_rss, HPAGE_SIZE / PAGE_SIZE);
1e8f889b DG	487	new_pte = make_huge_pte(vma, page, ((vma->vm_flags & VM_WRITE)
	488	&& (vma->vm_flags & VM_SHARED)));
	489	set_huge_pte_at(mm, address, ptep, new_pte);
	490
	491	if (write_access && !(vma->vm_flags & VM_SHARED)) {
	492	/* Optimization, do the COW without a second fault */
	493	ret = hugetlb_cow(mm, vma, address, ptep, new_pte);
	494	}
	495
ac9b9c66	496	spin_unlock(&mm->page_table_lock);
4c887265 AL	497	unlock_page(page);
4c887265 AL	498	out:
ac9b9c66	499	return ret;
4c887265 AL	500
	501	backout:
	502	spin_unlock(&mm->page_table_lock);
	503	hugetlb_put_quota(mapping);
	504	unlock_page(page);
	505	put_page(page);
	506	goto out;
ac9b9c66 HD	507	}
ac9b9c66 HD	508
86e5216f AL	509	int hugetlb_fault(struct mm_struct mm, struct vm_area_struct vma,
	510	unsigned long address, int write_access)
	511	{
	512	pte_t *ptep;
	513	pte_t entry;
1e8f889b	514	int ret;
86e5216f AL	515
	516	ptep = huge_pte_alloc(mm, address);
	517	if (!ptep)
	518	return VM_FAULT_OOM;
	519
	520	entry = *ptep;
	521	if (pte_none(entry))
1e8f889b	522	return hugetlb_no_page(mm, vma, address, ptep, write_access);
86e5216f	523
1e8f889b DG	524	ret = VM_FAULT_MINOR;
	525
	526	spin_lock(&mm->page_table_lock);
	527	/* Check for a racing update before calling hugetlb_cow */
	528	if (likely(pte_same(entry, *ptep)))
	529	if (write_access && !pte_write(entry))
	530	ret = hugetlb_cow(mm, vma, address, ptep, entry);
	531	spin_unlock(&mm->page_table_lock);
	532
	533	return ret;
86e5216f AL	534	}
86e5216f AL	535
63551ae0 DG	536	int follow_hugetlb_page(struct mm_struct mm, struct vm_area_struct vma,
	537	struct page pages, struct vm_area_struct vmas,
	538	unsigned long position, int length, int i)
	539	{
	540	unsigned long vpfn, vaddr = *position;
	541	int remainder = *length;
	542
63551ae0	543	vpfn = vaddr/PAGE_SIZE;
1c59827d	544	spin_lock(&mm->page_table_lock);
63551ae0	545	while (vaddr < vma->vm_end && remainder) {
4c887265 AL	546	pte_t *pte;
4c887265 AL	547	struct page *page;
63551ae0	548
4c887265 AL	549	/*
	550	* Some archs (sparc64, sh*) have multiple pte_ts to
	551	* each hugepage. We have to make * sure we get the
	552	* first, for the page indexing below to work.
	553	*/
	554	pte = huge_pte_offset(mm, vaddr & HPAGE_MASK);
63551ae0	555
4c887265 AL	556	if (!pte \|\| pte_none(*pte)) {
4c887265 AL	557	int ret;
63551ae0	558
4c887265 AL	559	spin_unlock(&mm->page_table_lock);
	560	ret = hugetlb_fault(mm, vma, vaddr, 0);
	561	spin_lock(&mm->page_table_lock);
	562	if (ret == VM_FAULT_MINOR)
	563	continue;
63551ae0	564
4c887265 AL	565	remainder = 0;
	566	if (!i)
	567	i = -EFAULT;
	568	break;
	569	}
	570
	571	if (pages) {
	572	page = &pte_page(*pte)[vpfn % (HPAGE_SIZE/PAGE_SIZE)];
63551ae0 DG	573	get_page(page);
	574	pages[i] = page;
	575	}
	576
	577	if (vmas)
	578	vmas[i] = vma;
	579
	580	vaddr += PAGE_SIZE;
	581	++vpfn;
	582	--remainder;
	583	++i;
	584	}
1c59827d	585	spin_unlock(&mm->page_table_lock);
63551ae0 DG	586	*length = remainder;
	587	*position = vaddr;
	588
	589	return i;
	590	}