[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];
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) {
		nr_huge_pages++;
		nr_huge_pages_node[page_to_nid(page)]++;
	}
	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;

	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);
}
EXPORT_SYMBOL(hugetlb_total_pages);

/*
 * 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)
{
	pte_t entry;

	if (vma->vm_flags & VM_WRITE) {
		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;
}

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;

	for (addr = vma->vm_start; addr < vma->vm_end; addr += HPAGE_SIZE) {
		dst_pte = huge_pte_alloc(dst, addr);
		if (!dst_pte)
			goto nomem;
		spin_lock(&src->page_table_lock);
		src_pte = huge_pte_offset(src, addr);
		if (src_pte && !pte_none(*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);
	}
	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);

	/* 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)
			/* This can happen on truncate, or if an
			 * mmap() is aborted due to an error before
			 * the prefault */
			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));
	}
	flush_tlb_range(vma, start, end);
}

void zap_hugepage_range(struct vm_area_struct *vma,
			unsigned long start, unsigned long length)
{
	struct mm_struct *mm = vma->vm_mm;

	spin_lock(&mm->page_table_lock);
	unmap_hugepage_range(vma, start, start + length);
	spin_unlock(&mm->page_table_lock);
}

int hugetlb_prefault(struct address_space *mapping, struct vm_area_struct *vma)
{
	struct mm_struct *mm = current->mm;
	unsigned long addr;
	int ret = 0;

	WARN_ON(!is_vm_hugetlb_page(vma));
	BUG_ON(vma->vm_start & ~HPAGE_MASK);
	BUG_ON(vma->vm_end & ~HPAGE_MASK);

	hugetlb_prefault_arch_hook(mm);

	spin_lock(&mm->page_table_lock);
	for (addr = vma->vm_start; addr < vma->vm_end; addr += HPAGE_SIZE) {
		unsigned long idx;
		pte_t *pte = huge_pte_alloc(mm, addr);
		struct page *page;

		if (!pte) {
			ret = -ENOMEM;
			goto out;
		}

		idx = ((addr - vma->vm_start) >> HPAGE_SHIFT)
			+ (vma->vm_pgoff >> (HPAGE_SHIFT - PAGE_SHIFT));
		page = find_get_page(mapping, idx);
		if (!page) {
			/* charge the fs quota first */
			if (hugetlb_get_quota(mapping)) {
				ret = -ENOMEM;
				goto out;
			}
			page = alloc_huge_page();
			if (!page) {
				hugetlb_put_quota(mapping);
				ret = -ENOMEM;
				goto out;
			}
			ret = add_to_page_cache(page, mapping, idx, GFP_ATOMIC);
			if (! ret) {
				unlock_page(page);
			} else {
				hugetlb_put_quota(mapping);
				free_huge_page(page);
				goto out;
			}
		}
		add_mm_counter(mm, file_rss, HPAGE_SIZE / PAGE_SIZE);
		set_huge_pte_at(mm, addr, pte, make_huge_pte(vma, page));
	}
out:
	spin_unlock(&mm->page_table_lock);
	return ret;
}

/*
 * On ia64 at least, it is possible to receive a hugetlb fault from a
 * stale zero entry left in the TLB from earlier hardware prefetching.
 * Low-level arch code should already have flushed the stale entry as
 * part of its fault handling, but we do need to accept this minor fault
 * and return successfully.  Whereas the "normal" case is that this is
 * an access to a hugetlb page which has been truncated off since mmap.
 */
int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
			unsigned long address, int write_access)
{
	int ret = VM_FAULT_SIGBUS;
	pte_t *pte;

	spin_lock(&mm->page_table_lock);
	pte = huge_pte_offset(mm, address);
	if (pte && !pte_none(*pte))
		ret = VM_FAULT_MINOR;
	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;

	BUG_ON(!is_vm_hugetlb_page(vma));

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

		if (pages) {
			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);

			/* the hugetlb file might have been truncated */
			if (!pte || pte_none(*pte)) {
				remainder = 0;
				if (!i)
					i = -EFAULT;
				break;
			}

			page = &pte_page(*pte)[vpfn % (HPAGE_SIZE/PAGE_SIZE)];

			WARN_ON(!PageCompound(page));

			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];
	25	static DEFINE_SPINLOCK(hugetlb_lock);
	26
	27	static void enqueue_huge_page(struct page *page)
	28	{
	29	int nid = page_to_nid(page);
	30	list_add(&page->lru, &hugepage_freelists[nid]);
	31	free_huge_pages++;
	32	free_huge_pages_node[nid]++;
	33	}
	34
	35	static struct page *dequeue_huge_page(void)
	36	{
	37	int nid = numa_node_id();
	38	struct page *page = NULL;
	39
	40	if (list_empty(&hugepage_freelists[nid])) {
	41	for (nid = 0; nid < MAX_NUMNODES; ++nid)
	42	if (!list_empty(&hugepage_freelists[nid]))
	43	break;
	44	}
	45	if (nid >= 0 && nid < MAX_NUMNODES &&
	46	!list_empty(&hugepage_freelists[nid])) {
	47	page = list_entry(hugepage_freelists[nid].next,
	48	struct page, lru);
	49	list_del(&page->lru);
	50	free_huge_pages--;
	51	free_huge_pages_node[nid]--;
	52	}
	53	return page;
	54	}
	55
	56	static struct page *alloc_fresh_huge_page(void)
	57	{
	58	static int nid = 0;
	59	struct page *page;
	60	page = alloc_pages_node(nid, GFP_HIGHUSER\|__GFP_COMP\|__GFP_NOWARN,
	61	HUGETLB_PAGE_ORDER);
	62	nid = (nid + 1) % num_online_nodes();
	63	if (page) {
	64	nr_huge_pages++;
	65	nr_huge_pages_node[page_to_nid(page)]++;
	66	}
	67	return page;
	68	}
	69
	70	void free_huge_page(struct page *page)
	71	{
	72	BUG_ON(page_count(page));
	73
	74	INIT_LIST_HEAD(&page->lru);
	75	page[1].mapping = NULL;
	76
	77	spin_lock(&hugetlb_lock);
	78	enqueue_huge_page(page);
	79	spin_unlock(&hugetlb_lock);
	80	}
	81
82	struct page *alloc_huge_page(void)
83	{
84	struct page *page;
85	int i;
86
87	spin_lock(&hugetlb_lock);
88	page = dequeue_huge_page();
89	if (!page) {
90	spin_unlock(&hugetlb_lock);
91	return NULL;
92	}
93	spin_unlock(&hugetlb_lock);
94	set_page_count(page, 1);
95	page[1].mapping = (void *)free_huge_page;
96	for (i = 0; i < (HPAGE_SIZE/PAGE_SIZE); ++i)
97	clear_highpage(&page[i]);
98	return page;
99	}
100
101	static int __init hugetlb_init(void)
102	{
103	unsigned long i;
104	struct page *page;
105
106	for (i = 0; i < MAX_NUMNODES; ++i)
107	INIT_LIST_HEAD(&hugepage_freelists[i]);
108
109	for (i = 0; i < max_huge_pages; ++i) {
110	page = alloc_fresh_huge_page();
111	if (!page)
112	break;
113	spin_lock(&hugetlb_lock);
114	enqueue_huge_page(page);
115	spin_unlock(&hugetlb_lock);
116	}
117	max_huge_pages = free_huge_pages = nr_huge_pages = i;
118	printk("Total HugeTLB memory allocated, %ld\n", free_huge_pages);
119	return 0;
120	}
121	module_init(hugetlb_init);
122
123	static int __init hugetlb_setup(char *s)
124	{
125	if (sscanf(s, "%lu", &max_huge_pages) <= 0)
126	max_huge_pages = 0;
127	return 1;
128	}
129	__setup("hugepages=", hugetlb_setup);
130
131	#ifdef CONFIG_SYSCTL
132	static void update_and_free_page(struct page *page)
133	{
134	int i;
135	nr_huge_pages--;
136	nr_huge_pages_node[page_zone(page)->zone_pgdat->node_id]--;
137	for (i = 0; i < (HPAGE_SIZE / PAGE_SIZE); i++) {
138	page[i].flags &= ~(1 << PG_locked \| 1 << PG_error \| 1 << PG_referenced \|
139	1 << PG_dirty \| 1 << PG_active \| 1 << PG_reserved \|
140	1 << PG_private \| 1<< PG_writeback);
141	set_page_count(&page[i], 0);
142	}
143	set_page_count(page, 1);
144	__free_pages(page, HUGETLB_PAGE_ORDER);
145	}
146
147	#ifdef CONFIG_HIGHMEM
148	static void try_to_free_low(unsigned long count)
149	{
150	int i, nid;
151	for (i = 0; i < MAX_NUMNODES; ++i) {
152	struct page page, next;
153	list_for_each_entry_safe(page, next, &hugepage_freelists[i], lru) {
154	if (PageHighMem(page))
155	continue;
156	list_del(&page->lru);
157	update_and_free_page(page);
158	nid = page_zone(page)->zone_pgdat->node_id;
159	free_huge_pages--;
160	free_huge_pages_node[nid]--;
161	if (count >= nr_huge_pages)
162	return;
163	}
164	}
165	}
166	#else
167	static inline void try_to_free_low(unsigned long count)
168	{
169	}
170	#endif
171
172	static unsigned long set_max_huge_pages(unsigned long count)
173	{
174	while (count > nr_huge_pages) {
175	struct page *page = alloc_fresh_huge_page();
176	if (!page)
177	return nr_huge_pages;
178	spin_lock(&hugetlb_lock);
179	enqueue_huge_page(page);
180	spin_unlock(&hugetlb_lock);
181	}
182	if (count >= nr_huge_pages)
183	return nr_huge_pages;
184
185	spin_lock(&hugetlb_lock);
186	try_to_free_low(count);
187	while (count < nr_huge_pages) {
188	struct page *page = dequeue_huge_page();
189	if (!page)
190	break;
191	update_and_free_page(page);
192	}
193	spin_unlock(&hugetlb_lock);
194	return nr_huge_pages;
195	}
196
197	int hugetlb_sysctl_handler(struct ctl_table *table, int write,
198	struct file file, void __user buffer,
199	size_t length, loff_t ppos)
200	{
201	proc_doulongvec_minmax(table, write, file, buffer, length, ppos);
202	max_huge_pages = set_max_huge_pages(max_huge_pages);
203	return 0;
204	}
205	#endif /* CONFIG_SYSCTL */
206
207	int hugetlb_report_meminfo(char *buf)
208	{
209	return sprintf(buf,
210	"HugePages_Total: %5lu\n"
211	"HugePages_Free: %5lu\n"
212	"Hugepagesize: %5lu kB\n",
213	nr_huge_pages,
214	free_huge_pages,
215	HPAGE_SIZE/1024);
216	}
217
218	int hugetlb_report_node_meminfo(int nid, char *buf)
219	{
220	return sprintf(buf,
221	"Node %d HugePages_Total: %5u\n"
222	"Node %d HugePages_Free: %5u\n",
223	nid, nr_huge_pages_node[nid],
224	nid, free_huge_pages_node[nid]);
225	}
226
227	int is_hugepage_mem_enough(size_t size)
228	{
229	return (size + ~HPAGE_MASK)/HPAGE_SIZE <= free_huge_pages;
230	}
231
232	/* Return the number pages of memory we physically have, in PAGE_SIZE units. */
233	unsigned long hugetlb_total_pages(void)
234	{
235	return nr_huge_pages * (HPAGE_SIZE / PAGE_SIZE);
236	}
237	EXPORT_SYMBOL(hugetlb_total_pages);
238
239	/*
240	* We cannot handle pagefaults against hugetlb pages at all. They cause
241	* handle_mm_fault() to try to instantiate regular-sized pages in the
242	* hugegpage VMA. do_page_fault() is supposed to trap this, so BUG is we get
243	* this far.
244	*/
245	static struct page hugetlb_nopage(struct vm_area_struct vma,
246	unsigned long address, int *unused)
247	{
248	BUG();
249	return NULL;
250	}
251
252	struct vm_operations_struct hugetlb_vm_ops = {
253	.nopage = hugetlb_nopage,
254	};
255
63551ae0 DG	256	static pte_t make_huge_pte(struct vm_area_struct vma, struct page page)
	257	{
	258	pte_t entry;
	259
	260	if (vma->vm_flags & VM_WRITE) {
	261	entry =
	262	pte_mkwrite(pte_mkdirty(mk_pte(page, vma->vm_page_prot)));
	263	} else {
	264	entry = pte_wrprotect(mk_pte(page, vma->vm_page_prot));
	265	}
	266	entry = pte_mkyoung(entry);
	267	entry = pte_mkhuge(entry);
	268
	269	return entry;
	270	}
	271
	272	int copy_hugetlb_page_range(struct mm_struct dst, struct mm_struct src,
	273	struct vm_area_struct *vma)
	274	{
	275	pte_t src_pte, dst_pte, entry;
	276	struct page *ptepage;
1c59827d	277	unsigned long addr;
63551ae0	278
1c59827d	279	for (addr = vma->vm_start; addr < vma->vm_end; addr += HPAGE_SIZE) {
63551ae0 DG	280	dst_pte = huge_pte_alloc(dst, addr);
	281	if (!dst_pte)
	282	goto nomem;
1c59827d	283	spin_lock(&src->page_table_lock);
63551ae0	284	src_pte = huge_pte_offset(src, addr);
1c59827d HD	285	if (src_pte && !pte_none(*src_pte)) {
	286	entry = *src_pte;
	287	ptepage = pte_page(entry);
	288	get_page(ptepage);
4294621f	289	add_mm_counter(dst, file_rss, HPAGE_SIZE / PAGE_SIZE);
1c59827d HD	290	set_huge_pte_at(dst, addr, dst_pte, entry);
	291	}
	292	spin_unlock(&src->page_table_lock);
63551ae0 DG	293	}
	294	return 0;
	295
	296	nomem:
	297	return -ENOMEM;
	298	}
	299
	300	void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
	301	unsigned long end)
	302	{
	303	struct mm_struct *mm = vma->vm_mm;
	304	unsigned long address;
c7546f8f	305	pte_t *ptep;
63551ae0 DG	306	pte_t pte;
	307	struct page *page;
	308
	309	WARN_ON(!is_vm_hugetlb_page(vma));
	310	BUG_ON(start & ~HPAGE_MASK);
	311	BUG_ON(end & ~HPAGE_MASK);
	312
365e9c87 HD	313	/* Update high watermark before we lower rss */
	314	update_hiwater_rss(mm);
	315
63551ae0	316	for (address = start; address < end; address += HPAGE_SIZE) {
c7546f8f DG	317	ptep = huge_pte_offset(mm, address);
	318	if (! ptep)
	319	/* This can happen on truncate, or if an
	320	* mmap() is aborted due to an error before
	321	* the prefault */
	322	continue;
	323
	324	pte = huge_ptep_get_and_clear(mm, address, ptep);
63551ae0 DG	325	if (pte_none(pte))
63551ae0 DG	326	continue;
c7546f8f	327
63551ae0 DG	328	page = pte_page(pte);
63551ae0 DG	329	put_page(page);
4294621f	330	add_mm_counter(mm, file_rss, (int) -(HPAGE_SIZE / PAGE_SIZE));
63551ae0	331	}
63551ae0 DG	332	flush_tlb_range(vma, start, end);
	333	}
	334
1da177e4 LT	335	void zap_hugepage_range(struct vm_area_struct *vma,
	336	unsigned long start, unsigned long length)
	337	{
	338	struct mm_struct *mm = vma->vm_mm;
	339
	340	spin_lock(&mm->page_table_lock);
	341	unmap_hugepage_range(vma, start, start + length);
	342	spin_unlock(&mm->page_table_lock);
	343	}
63551ae0 DG	344
	345	int hugetlb_prefault(struct address_space mapping, struct vm_area_struct vma)
	346	{
	347	struct mm_struct *mm = current->mm;
	348	unsigned long addr;
	349	int ret = 0;
	350
	351	WARN_ON(!is_vm_hugetlb_page(vma));
	352	BUG_ON(vma->vm_start & ~HPAGE_MASK);
	353	BUG_ON(vma->vm_end & ~HPAGE_MASK);
	354
	355	hugetlb_prefault_arch_hook(mm);
	356
	357	spin_lock(&mm->page_table_lock);
	358	for (addr = vma->vm_start; addr < vma->vm_end; addr += HPAGE_SIZE) {
	359	unsigned long idx;
	360	pte_t *pte = huge_pte_alloc(mm, addr);
	361	struct page *page;
	362
	363	if (!pte) {
	364	ret = -ENOMEM;
	365	goto out;
	366	}
63551ae0 DG	367
	368	idx = ((addr - vma->vm_start) >> HPAGE_SHIFT)
	369	+ (vma->vm_pgoff >> (HPAGE_SHIFT - PAGE_SHIFT));
	370	page = find_get_page(mapping, idx);
	371	if (!page) {
	372	/* charge the fs quota first */
	373	if (hugetlb_get_quota(mapping)) {
	374	ret = -ENOMEM;
	375	goto out;
	376	}
	377	page = alloc_huge_page();
	378	if (!page) {
	379	hugetlb_put_quota(mapping);
	380	ret = -ENOMEM;
	381	goto out;
	382	}
	383	ret = add_to_page_cache(page, mapping, idx, GFP_ATOMIC);
	384	if (! ret) {
	385	unlock_page(page);
	386	} else {
	387	hugetlb_put_quota(mapping);
	388	free_huge_page(page);
	389	goto out;
	390	}
	391	}
4294621f	392	add_mm_counter(mm, file_rss, HPAGE_SIZE / PAGE_SIZE);
63551ae0 DG	393	set_huge_pte_at(mm, addr, pte, make_huge_pte(vma, page));
	394	}
	395	out:
	396	spin_unlock(&mm->page_table_lock);
	397	return ret;
	398	}
	399
ac9b9c66 HD	400	/*
	401	* On ia64 at least, it is possible to receive a hugetlb fault from a
	402	* stale zero entry left in the TLB from earlier hardware prefetching.
	403	* Low-level arch code should already have flushed the stale entry as
	404	* part of its fault handling, but we do need to accept this minor fault
	405	* and return successfully. Whereas the "normal" case is that this is
	406	* an access to a hugetlb page which has been truncated off since mmap.
	407	*/
	408	int hugetlb_fault(struct mm_struct mm, struct vm_area_struct vma,
	409	unsigned long address, int write_access)
	410	{
	411	int ret = VM_FAULT_SIGBUS;
	412	pte_t *pte;
	413
	414	spin_lock(&mm->page_table_lock);
	415	pte = huge_pte_offset(mm, address);
	416	if (pte && !pte_none(*pte))
	417	ret = VM_FAULT_MINOR;
	418	spin_unlock(&mm->page_table_lock);
	419	return ret;
	420	}
	421
63551ae0 DG	422	int follow_hugetlb_page(struct mm_struct mm, struct vm_area_struct vma,
	423	struct page pages, struct vm_area_struct vmas,
	424	unsigned long position, int length, int i)
	425	{
	426	unsigned long vpfn, vaddr = *position;
	427	int remainder = *length;
	428
	429	BUG_ON(!is_vm_hugetlb_page(vma));
	430
	431	vpfn = vaddr/PAGE_SIZE;
1c59827d	432	spin_lock(&mm->page_table_lock);
63551ae0 DG	433	while (vaddr < vma->vm_end && remainder) {
	434
	435	if (pages) {
	436	pte_t *pte;
	437	struct page *page;
	438
	439	/* Some archs (sparc64, sh*) have multiple
	440	* pte_ts to each hugepage. We have to make
	441	* sure we get the first, for the page
	442	* indexing below to work. */
	443	pte = huge_pte_offset(mm, vaddr & HPAGE_MASK);
	444
1c59827d HD	445	/* the hugetlb file might have been truncated */
	446	if (!pte \|\| pte_none(*pte)) {
	447	remainder = 0;
	448	if (!i)
	449	i = -EFAULT;
	450	break;
	451	}
63551ae0 DG	452
	453	page = &pte_page(*pte)[vpfn % (HPAGE_SIZE/PAGE_SIZE)];
	454
	455	WARN_ON(!PageCompound(page));
	456
	457	get_page(page);
	458	pages[i] = page;
	459	}
	460
	461	if (vmas)
	462	vmas[i] = vma;
	463
	464	vaddr += PAGE_SIZE;
	465	++vpfn;
	466	--remainder;
	467	++i;
	468	}
1c59827d	469	spin_unlock(&mm->page_table_lock);
63551ae0 DG	470	*length = remainder;
	471	*position = vaddr;
	472
	473	return i;
	474	}