hugetlb.c 13.2 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1
2
3
4
5
6
7
8
9
10
11
12
/*
 * 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>
David Gibson's avatar
David Gibson committed
13
#include <linux/pagemap.h>
14
#include <linux/mempolicy.h>
15
#include <linux/cpuset.h>
16

David Gibson's avatar
David Gibson committed
17
18
19
20
#include <asm/page.h>
#include <asm/pgtable.h>

#include <linux/hugetlb.h>
Linus Torvalds's avatar
Linus Torvalds committed
21
22
23
24
25
26
27

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];
28
29
30
31

/*
 * Protects updates to hugepage_freelists, nr_huge_pages, and free_huge_pages
 */
Linus Torvalds's avatar
Linus Torvalds committed
32
33
34
35
36
37
38
39
40
41
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]++;
}

42
43
static struct page *dequeue_huge_page(struct vm_area_struct *vma,
				unsigned long address)
Linus Torvalds's avatar
Linus Torvalds committed
44
45
46
{
	int nid = numa_node_id();
	struct page *page = NULL;
47
	struct zonelist *zonelist = huge_zonelist(vma, address);
48
	struct zone **z;
Linus Torvalds's avatar
Linus Torvalds committed
49

50
51
	for (z = zonelist->zones; *z; z++) {
		nid = (*z)->zone_pgdat->node_id;
52
53
		if (cpuset_zone_allowed(*z, GFP_HIGHUSER) &&
		    !list_empty(&hugepage_freelists[nid]))
54
			break;
Linus Torvalds's avatar
Linus Torvalds committed
55
	}
56
57

	if (*z) {
Linus Torvalds's avatar
Linus Torvalds committed
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
		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) {
75
		spin_lock(&hugetlb_lock);
Linus Torvalds's avatar
Linus Torvalds committed
76
77
		nr_huge_pages++;
		nr_huge_pages_node[page_to_nid(page)]++;
78
		spin_unlock(&hugetlb_lock);
Linus Torvalds's avatar
Linus Torvalds committed
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
	}
	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);
}

95
struct page *alloc_huge_page(struct vm_area_struct *vma, unsigned long addr)
Linus Torvalds's avatar
Linus Torvalds committed
96
97
98
99
100
{
	struct page *page;
	int i;

	spin_lock(&hugetlb_lock);
101
	page = dequeue_huge_page(vma, addr);
Linus Torvalds's avatar
Linus Torvalds committed
102
103
104
105
106
107
108
109
	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)
110
		clear_user_highpage(&page[i], addr);
Linus Torvalds's avatar
Linus Torvalds committed
111
112
113
114
115
116
117
118
	return page;
}

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

119
120
121
	if (HPAGE_SHIFT == 0)
		return 0;

Linus Torvalds's avatar
Linus Torvalds committed
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
	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) {
204
		struct page *page = dequeue_huge_page(NULL, 0);
Linus Torvalds's avatar
Linus Torvalds committed
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
		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,
};

271
272
static pte_t make_huge_pte(struct vm_area_struct *vma, struct page *page,
				int writable)
David Gibson's avatar
David Gibson committed
273
274
275
{
	pte_t entry;

276
	if (writable) {
David Gibson's avatar
David Gibson committed
277
278
279
280
281
282
283
284
285
286
287
		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;
}

288
289
290
291
292
293
294
295
296
297
298
299
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);
}


David Gibson's avatar
David Gibson committed
300
301
302
303
304
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;
305
	unsigned long addr;
306
307
308
	int cow;

	cow = (vma->vm_flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
David Gibson's avatar
David Gibson committed
309

310
	for (addr = vma->vm_start; addr < vma->vm_end; addr += HPAGE_SIZE) {
311
312
313
		src_pte = huge_pte_offset(src, addr);
		if (!src_pte)
			continue;
David Gibson's avatar
David Gibson committed
314
315
316
		dst_pte = huge_pte_alloc(dst, addr);
		if (!dst_pte)
			goto nomem;
317
		spin_lock(&dst->page_table_lock);
318
		spin_lock(&src->page_table_lock);
319
		if (!pte_none(*src_pte)) {
320
321
			if (cow)
				ptep_set_wrprotect(src, addr, src_pte);
322
323
324
			entry = *src_pte;
			ptepage = pte_page(entry);
			get_page(ptepage);
325
			add_mm_counter(dst, file_rss, HPAGE_SIZE / PAGE_SIZE);
326
327
328
			set_huge_pte_at(dst, addr, dst_pte, entry);
		}
		spin_unlock(&src->page_table_lock);
329
		spin_unlock(&dst->page_table_lock);
David Gibson's avatar
David Gibson committed
330
331
332
333
334
335
336
337
338
339
340
341
	}
	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;
342
	pte_t *ptep;
David Gibson's avatar
David Gibson committed
343
344
345
346
347
348
349
	pte_t pte;
	struct page *page;

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

350
351
	spin_lock(&mm->page_table_lock);

352
353
354
	/* Update high watermark before we lower rss */
	update_hiwater_rss(mm);

David Gibson's avatar
David Gibson committed
355
	for (address = start; address < end; address += HPAGE_SIZE) {
356
		ptep = huge_pte_offset(mm, address);
357
		if (!ptep)
358
359
360
			continue;

		pte = huge_ptep_get_and_clear(mm, address, ptep);
David Gibson's avatar
David Gibson committed
361
362
		if (pte_none(pte))
			continue;
363

David Gibson's avatar
David Gibson committed
364
365
		page = pte_page(pte);
		put_page(page);
366
		add_mm_counter(mm, file_rss, (int) -(HPAGE_SIZE / PAGE_SIZE));
David Gibson's avatar
David Gibson committed
367
368
	}

Linus Torvalds's avatar
Linus Torvalds committed
369
	spin_unlock(&mm->page_table_lock);
370
	flush_tlb_range(vma, start, end);
Linus Torvalds's avatar
Linus Torvalds committed
371
}
David Gibson's avatar
David Gibson committed
372

373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
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);
390
	new_page = alloc_huge_page(vma, address);
391
392
393

	if (!new_page) {
		page_cache_release(old_page);
394
		return VM_FAULT_OOM;
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
	}

	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;
}

416
int hugetlb_no_page(struct mm_struct *mm, struct vm_area_struct *vma,
417
			unsigned long address, pte_t *ptep, int write_access)
418
419
{
	int ret = VM_FAULT_SIGBUS;
420
421
422
423
	unsigned long idx;
	unsigned long size;
	struct page *page;
	struct address_space *mapping;
424
	pte_t new_pte;
425
426
427
428
429
430
431
432
433

	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.
	 */
434
435
436
437
438
439
440
441
retry:
	page = find_lock_page(mapping, idx);
	if (!page) {
		if (hugetlb_get_quota(mapping))
			goto out;
		page = alloc_huge_page(vma, address);
		if (!page) {
			hugetlb_put_quota(mapping);
442
			ret = VM_FAULT_OOM;
443
444
			goto out;
		}
445

446
447
448
449
450
451
452
453
454
455
456
457
458
459
		if (vma->vm_flags & VM_SHARED) {
			int err;

			err = add_to_page_cache(page, mapping, idx, GFP_KERNEL);
			if (err) {
				put_page(page);
				hugetlb_put_quota(mapping);
				if (err == -EEXIST)
					goto retry;
				goto out;
			}
		} else
			lock_page(page);
	}
460

461
	spin_lock(&mm->page_table_lock);
462
463
464
465
466
	size = i_size_read(mapping->host) >> HPAGE_SHIFT;
	if (idx >= size)
		goto backout;

	ret = VM_FAULT_MINOR;
467
	if (!pte_none(*ptep))
468
469
470
		goto backout;

	add_mm_counter(mm, file_rss, HPAGE_SIZE / PAGE_SIZE);
471
472
473
474
475
476
477
478
479
	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);
	}

480
	spin_unlock(&mm->page_table_lock);
481
482
	unlock_page(page);
out:
483
	return ret;
484
485
486
487
488
489
490

backout:
	spin_unlock(&mm->page_table_lock);
	hugetlb_put_quota(mapping);
	unlock_page(page);
	put_page(page);
	goto out;
491
492
}

493
494
495
496
497
int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
			unsigned long address, int write_access)
{
	pte_t *ptep;
	pte_t entry;
498
	int ret;
499
500
501
502
503
504
505

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

	entry = *ptep;
	if (pte_none(entry))
506
		return hugetlb_no_page(mm, vma, address, ptep, write_access);
507

508
509
510
511
512
513
514
515
516
517
	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;
518
519
}

David Gibson's avatar
David Gibson committed
520
521
522
523
524
525
526
527
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;
528
	spin_lock(&mm->page_table_lock);
David Gibson's avatar
David Gibson committed
529
	while (vaddr < vma->vm_end && remainder) {
530
531
		pte_t *pte;
		struct page *page;
David Gibson's avatar
David Gibson committed
532

533
534
535
536
537
538
		/*
		 * 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);
David Gibson's avatar
David Gibson committed
539

540
541
		if (!pte || pte_none(*pte)) {
			int ret;
David Gibson's avatar
David Gibson committed
542

543
544
545
546
547
			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;
David Gibson's avatar
David Gibson committed
548

549
550
551
552
553
554
555
556
			remainder = 0;
			if (!i)
				i = -EFAULT;
			break;
		}

		if (pages) {
			page = &pte_page(*pte)[vpfn % (HPAGE_SIZE/PAGE_SIZE)];
David Gibson's avatar
David Gibson committed
557
558
559
560
561
562
563
564
565
566
567
568
			get_page(page);
			pages[i] = page;
		}

		if (vmas)
			vmas[i] = vma;

		vaddr += PAGE_SIZE;
		++vpfn;
		--remainder;
		++i;
	}
569
	spin_unlock(&mm->page_table_lock);
David Gibson's avatar
David Gibson committed
570
571
572
573
574
	*length = remainder;
	*position = vaddr;

	return i;
}