hugetlb.c 41.9 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
#include <linux/mutex.h>
17

David Gibson's avatar
David Gibson committed
18
19
20
21
#include <asm/page.h>
#include <asm/pgtable.h>

#include <linux/hugetlb.h>
22
#include "internal.h"
Linus Torvalds's avatar
Linus Torvalds committed
23
24

const unsigned long hugetlb_zero = 0, hugetlb_infinity = ~0UL;
25
static unsigned long nr_huge_pages, free_huge_pages, resv_huge_pages;
26
static unsigned long surplus_huge_pages;
27
static unsigned long nr_overcommit_huge_pages;
Linus Torvalds's avatar
Linus Torvalds committed
28
unsigned long max_huge_pages;
29
unsigned long sysctl_overcommit_huge_pages;
Linus Torvalds's avatar
Linus Torvalds committed
30
31
32
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];
33
static unsigned int surplus_huge_pages_node[MAX_NUMNODES];
34
35
static gfp_t htlb_alloc_mask = GFP_HIGHUSER;
unsigned long hugepages_treat_as_movable;
36
static int hugetlb_next_nid;
37

38
39
40
41
/*
 * Protects updates to hugepage_freelists, nr_huge_pages, and free_huge_pages
 */
static DEFINE_SPINLOCK(hugetlb_lock);
42

43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
/*
 * Convert the address within this vma to the page offset within
 * the mapping, in base page units.
 */
static pgoff_t vma_page_offset(struct vm_area_struct *vma,
				unsigned long address)
{
	return ((address - vma->vm_start) >> PAGE_SHIFT) +
					(vma->vm_pgoff >> PAGE_SHIFT);
}

/*
 * Convert the address within this vma to the page offset within
 * the mapping, in pagecache page units; huge pages here.
 */
static pgoff_t vma_pagecache_offset(struct vm_area_struct *vma,
					unsigned long address)
{
	return ((address - vma->vm_start) >> HPAGE_SHIFT) +
			(vma->vm_pgoff >> (HPAGE_SHIFT - PAGE_SHIFT));
}

65
66
67
#define HPAGE_RESV_OWNER    (1UL << (BITS_PER_LONG - 1))
#define HPAGE_RESV_UNMAPPED (1UL << (BITS_PER_LONG - 2))
#define HPAGE_RESV_MASK (HPAGE_RESV_OWNER | HPAGE_RESV_UNMAPPED)
68
69
70
71
72
73
74
75
76
77
/*
 * These helpers are used to track how many pages are reserved for
 * faults in a MAP_PRIVATE mapping. Only the process that called mmap()
 * is guaranteed to have their future faults succeed.
 *
 * With the exception of reset_vma_resv_huge_pages() which is called at fork(),
 * the reserve counters are updated with the hugetlb_lock held. It is safe
 * to reset the VMA at fork() time as it is not in use yet and there is no
 * chance of the global counters getting corrupted as a result of the values.
 */
78
79
80
81
82
83
84
85
86
87
88
static unsigned long get_vma_private_data(struct vm_area_struct *vma)
{
	return (unsigned long)vma->vm_private_data;
}

static void set_vma_private_data(struct vm_area_struct *vma,
							unsigned long value)
{
	vma->vm_private_data = (void *)value;
}

89
90
91
92
static unsigned long vma_resv_huge_pages(struct vm_area_struct *vma)
{
	VM_BUG_ON(!is_vm_hugetlb_page(vma));
	if (!(vma->vm_flags & VM_SHARED))
93
		return get_vma_private_data(vma) & ~HPAGE_RESV_MASK;
94
95
96
97
98
99
100
101
102
	return 0;
}

static void set_vma_resv_huge_pages(struct vm_area_struct *vma,
							unsigned long reserve)
{
	VM_BUG_ON(!is_vm_hugetlb_page(vma));
	VM_BUG_ON(vma->vm_flags & VM_SHARED);

103
104
	set_vma_private_data(vma,
		(get_vma_private_data(vma) & HPAGE_RESV_MASK) | reserve);
105
106
107
108
109
}

static void set_vma_resv_flags(struct vm_area_struct *vma, unsigned long flags)
{
	VM_BUG_ON(!is_vm_hugetlb_page(vma));
110
111
112
	VM_BUG_ON(vma->vm_flags & VM_SHARED);

	set_vma_private_data(vma, get_vma_private_data(vma) | flags);
113
114
115
116
117
}

static int is_vma_resv_set(struct vm_area_struct *vma, unsigned long flag)
{
	VM_BUG_ON(!is_vm_hugetlb_page(vma));
118
119

	return (get_vma_private_data(vma) & flag) != 0;
120
121
122
123
124
125
126
127
128
129
130
131
132
}

/* Decrement the reserved pages in the hugepage pool by one */
static void decrement_hugepage_resv_vma(struct vm_area_struct *vma)
{
	if (vma->vm_flags & VM_SHARED) {
		/* Shared mappings always use reserves */
		resv_huge_pages--;
	} else {
		/*
		 * Only the process that called mmap() has reserves for
		 * private mappings.
		 */
133
134
		if (is_vma_resv_set(vma, HPAGE_RESV_OWNER)) {
			unsigned long flags, reserve;
135
			resv_huge_pages--;
136
137
			flags = (unsigned long)vma->vm_private_data &
							HPAGE_RESV_MASK;
138
			reserve = (unsigned long)vma->vm_private_data - 1;
139
			vma->vm_private_data = (void *)(reserve | flags);
140
141
142
143
		}
	}
}

144
/* Reset counters to 0 and clear all HPAGE_RESV_* flags */
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
void reset_vma_resv_huge_pages(struct vm_area_struct *vma)
{
	VM_BUG_ON(!is_vm_hugetlb_page(vma));
	if (!(vma->vm_flags & VM_SHARED))
		vma->vm_private_data = (void *)0;
}

/* Returns true if the VMA has associated reserve pages */
static int vma_has_private_reserves(struct vm_area_struct *vma)
{
	if (vma->vm_flags & VM_SHARED)
		return 0;
	if (!vma_resv_huge_pages(vma))
		return 0;
	return 1;
}

162
163
164
165
166
167
168
static void clear_huge_page(struct page *page, unsigned long addr)
{
	int i;

	might_sleep();
	for (i = 0; i < (HPAGE_SIZE/PAGE_SIZE); i++) {
		cond_resched();
169
		clear_user_highpage(page + i, addr + i * PAGE_SIZE);
170
171
172
173
	}
}

static void copy_huge_page(struct page *dst, struct page *src,
174
			   unsigned long addr, struct vm_area_struct *vma)
175
176
177
178
179
180
{
	int i;

	might_sleep();
	for (i = 0; i < HPAGE_SIZE/PAGE_SIZE; i++) {
		cond_resched();
181
		copy_user_highpage(dst + i, src + i, addr + i*PAGE_SIZE, vma);
182
183
184
	}
}

Linus Torvalds's avatar
Linus Torvalds committed
185
186
187
188
189
190
191
192
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]++;
}

193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
static struct page *dequeue_huge_page(void)
{
	int nid;
	struct page *page = NULL;

	for (nid = 0; nid < MAX_NUMNODES; ++nid) {
		if (!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]--;
			break;
		}
	}
	return page;
}

static struct page *dequeue_huge_page_vma(struct vm_area_struct *vma,
212
				unsigned long address, int avoid_reserve)
Linus Torvalds's avatar
Linus Torvalds committed
213
{
214
	int nid;
Linus Torvalds's avatar
Linus Torvalds committed
215
	struct page *page = NULL;
216
	struct mempolicy *mpol;
217
	nodemask_t *nodemask;
218
	struct zonelist *zonelist = huge_zonelist(vma, address,
219
					htlb_alloc_mask, &mpol, &nodemask);
220
221
	struct zone *zone;
	struct zoneref *z;
Linus Torvalds's avatar
Linus Torvalds committed
222

223
224
225
226
227
228
229
230
231
	/*
	 * A child process with MAP_PRIVATE mappings created by their parent
	 * have no page reserves. This check ensures that reservations are
	 * not "stolen". The child may still get SIGKILLed
	 */
	if (!vma_has_private_reserves(vma) &&
			free_huge_pages - resv_huge_pages == 0)
		return NULL;

232
233
234
235
	/* If reserves cannot be used, ensure enough pages are in the pool */
	if (avoid_reserve && free_huge_pages - resv_huge_pages == 0)
		return NULL;

236
237
	for_each_zone_zonelist_nodemask(zone, z, zonelist,
						MAX_NR_ZONES - 1, nodemask) {
238
239
		nid = zone_to_nid(zone);
		if (cpuset_zone_allowed_softwall(zone, htlb_alloc_mask) &&
240
241
242
243
244
245
		    !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]--;
246
247
248

			if (!avoid_reserve)
				decrement_hugepage_resv_vma(vma);
249

Ken Chen's avatar
Ken Chen committed
250
			break;
251
		}
Linus Torvalds's avatar
Linus Torvalds committed
252
	}
253
	mpol_cond_put(mpol);
Linus Torvalds's avatar
Linus Torvalds committed
254
255
256
	return page;
}

257
258
259
260
261
262
263
264
265
266
267
268
static void update_and_free_page(struct page *page)
{
	int i;
	nr_huge_pages--;
	nr_huge_pages_node[page_to_nid(page)]--;
	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_compound_page_dtor(page, NULL);
	set_page_refcounted(page);
269
	arch_release_hugepage(page);
270
271
272
	__free_pages(page, HUGETLB_PAGE_ORDER);
}

273
274
static void free_huge_page(struct page *page)
{
275
	int nid = page_to_nid(page);
276
	struct address_space *mapping;
277

278
	mapping = (struct address_space *) page_private(page);
279
	set_page_private(page, 0);
280
	BUG_ON(page_count(page));
281
282
283
	INIT_LIST_HEAD(&page->lru);

	spin_lock(&hugetlb_lock);
284
285
286
287
288
289
290
	if (surplus_huge_pages_node[nid]) {
		update_and_free_page(page);
		surplus_huge_pages--;
		surplus_huge_pages_node[nid]--;
	} else {
		enqueue_huge_page(page);
	}
291
	spin_unlock(&hugetlb_lock);
292
	if (mapping)
293
		hugetlb_put_quota(mapping, 1);
294
295
}

296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
/*
 * Increment or decrement surplus_huge_pages.  Keep node-specific counters
 * balanced by operating on them in a round-robin fashion.
 * Returns 1 if an adjustment was made.
 */
static int adjust_pool_surplus(int delta)
{
	static int prev_nid;
	int nid = prev_nid;
	int ret = 0;

	VM_BUG_ON(delta != -1 && delta != 1);
	do {
		nid = next_node(nid, node_online_map);
		if (nid == MAX_NUMNODES)
			nid = first_node(node_online_map);

		/* To shrink on this node, there must be a surplus page */
		if (delta < 0 && !surplus_huge_pages_node[nid])
			continue;
		/* Surplus cannot exceed the total number of pages */
		if (delta > 0 && surplus_huge_pages_node[nid] >=
						nr_huge_pages_node[nid])
			continue;

		surplus_huge_pages += delta;
		surplus_huge_pages_node[nid] += delta;
		ret = 1;
		break;
	} while (nid != prev_nid);

	prev_nid = nid;
	return ret;
}

331
static struct page *alloc_fresh_huge_page_node(int nid)
Linus Torvalds's avatar
Linus Torvalds committed
332
333
{
	struct page *page;
334

335
	page = alloc_pages_node(nid,
336
337
		htlb_alloc_mask|__GFP_COMP|__GFP_THISNODE|
						__GFP_REPEAT|__GFP_NOWARN,
338
		HUGETLB_PAGE_ORDER);
Linus Torvalds's avatar
Linus Torvalds committed
339
	if (page) {
340
341
		if (arch_prepare_hugepage(page)) {
			__free_pages(page, HUGETLB_PAGE_ORDER);
342
			return NULL;
343
		}
344
		set_compound_page_dtor(page, free_huge_page);
345
		spin_lock(&hugetlb_lock);
Linus Torvalds's avatar
Linus Torvalds committed
346
		nr_huge_pages++;
347
		nr_huge_pages_node[nid]++;
348
		spin_unlock(&hugetlb_lock);
349
		put_page(page); /* free it into the hugepage allocator */
Linus Torvalds's avatar
Linus Torvalds committed
350
	}
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384

	return page;
}

static int alloc_fresh_huge_page(void)
{
	struct page *page;
	int start_nid;
	int next_nid;
	int ret = 0;

	start_nid = hugetlb_next_nid;

	do {
		page = alloc_fresh_huge_page_node(hugetlb_next_nid);
		if (page)
			ret = 1;
		/*
		 * Use a helper variable to find the next node and then
		 * copy it back to hugetlb_next_nid afterwards:
		 * otherwise there's a window in which a racer might
		 * pass invalid nid MAX_NUMNODES to alloc_pages_node.
		 * But we don't need to use a spin_lock here: it really
		 * doesn't matter if occasionally a racer chooses the
		 * same nid as we do.  Move nid forward in the mask even
		 * if we just successfully allocated a hugepage so that
		 * the next caller gets hugepages on the next node.
		 */
		next_nid = next_node(hugetlb_next_nid, node_online_map);
		if (next_nid == MAX_NUMNODES)
			next_nid = first_node(node_online_map);
		hugetlb_next_nid = next_nid;
	} while (!page && hugetlb_next_nid != start_nid);

385
386
387
388
389
	if (ret)
		count_vm_event(HTLB_BUDDY_PGALLOC);
	else
		count_vm_event(HTLB_BUDDY_PGALLOC_FAIL);

390
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
391
392
}

393
394
395
396
static struct page *alloc_buddy_huge_page(struct vm_area_struct *vma,
						unsigned long address)
{
	struct page *page;
397
	unsigned int nid;
398

399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
	/*
	 * Assume we will successfully allocate the surplus page to
	 * prevent racing processes from causing the surplus to exceed
	 * overcommit
	 *
	 * This however introduces a different race, where a process B
	 * tries to grow the static hugepage pool while alloc_pages() is
	 * called by process A. B will only examine the per-node
	 * counters in determining if surplus huge pages can be
	 * converted to normal huge pages in adjust_pool_surplus(). A
	 * won't be able to increment the per-node counter, until the
	 * lock is dropped by B, but B doesn't drop hugetlb_lock until
	 * no more huge pages can be converted from surplus to normal
	 * state (and doesn't try to convert again). Thus, we have a
	 * case where a surplus huge page exists, the pool is grown, and
	 * the surplus huge page still exists after, even though it
	 * should just have been converted to a normal huge page. This
	 * does not leak memory, though, as the hugepage will be freed
	 * once it is out of use. It also does not allow the counters to
	 * go out of whack in adjust_pool_surplus() as we don't modify
	 * the node values until we've gotten the hugepage and only the
	 * per-node value is checked there.
	 */
	spin_lock(&hugetlb_lock);
	if (surplus_huge_pages >= nr_overcommit_huge_pages) {
		spin_unlock(&hugetlb_lock);
		return NULL;
	} else {
		nr_huge_pages++;
		surplus_huge_pages++;
	}
	spin_unlock(&hugetlb_lock);

432
433
	page = alloc_pages(htlb_alloc_mask|__GFP_COMP|
					__GFP_REPEAT|__GFP_NOWARN,
434
					HUGETLB_PAGE_ORDER);
435
436

	spin_lock(&hugetlb_lock);
437
	if (page) {
438
439
440
441
442
443
		/*
		 * This page is now managed by the hugetlb allocator and has
		 * no users -- drop the buddy allocator's reference.
		 */
		put_page_testzero(page);
		VM_BUG_ON(page_count(page));
444
		nid = page_to_nid(page);
445
		set_compound_page_dtor(page, free_huge_page);
446
447
448
449
450
		/*
		 * We incremented the global counters already
		 */
		nr_huge_pages_node[nid]++;
		surplus_huge_pages_node[nid]++;
451
		__count_vm_event(HTLB_BUDDY_PGALLOC);
452
453
454
	} else {
		nr_huge_pages--;
		surplus_huge_pages--;
455
		__count_vm_event(HTLB_BUDDY_PGALLOC_FAIL);
456
	}
457
	spin_unlock(&hugetlb_lock);
458
459
460
461

	return page;
}

462
463
464
465
466
467
468
469
470
471
472
473
/*
 * Increase the hugetlb pool such that it can accomodate a reservation
 * of size 'delta'.
 */
static int gather_surplus_pages(int delta)
{
	struct list_head surplus_list;
	struct page *page, *tmp;
	int ret, i;
	int needed, allocated;

	needed = (resv_huge_pages + delta) - free_huge_pages;
474
475
	if (needed <= 0) {
		resv_huge_pages += delta;
476
		return 0;
477
	}
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514

	allocated = 0;
	INIT_LIST_HEAD(&surplus_list);

	ret = -ENOMEM;
retry:
	spin_unlock(&hugetlb_lock);
	for (i = 0; i < needed; i++) {
		page = alloc_buddy_huge_page(NULL, 0);
		if (!page) {
			/*
			 * We were not able to allocate enough pages to
			 * satisfy the entire reservation so we free what
			 * we've allocated so far.
			 */
			spin_lock(&hugetlb_lock);
			needed = 0;
			goto free;
		}

		list_add(&page->lru, &surplus_list);
	}
	allocated += needed;

	/*
	 * After retaking hugetlb_lock, we need to recalculate 'needed'
	 * because either resv_huge_pages or free_huge_pages may have changed.
	 */
	spin_lock(&hugetlb_lock);
	needed = (resv_huge_pages + delta) - (free_huge_pages + allocated);
	if (needed > 0)
		goto retry;

	/*
	 * The surplus_list now contains _at_least_ the number of extra pages
	 * needed to accomodate the reservation.  Add the appropriate number
	 * of pages to the hugetlb pool and free the extras back to the buddy
515
516
517
	 * allocator.  Commit the entire reservation here to prevent another
	 * process from stealing the pages as they are added to the pool but
	 * before they are reserved.
518
519
	 */
	needed += allocated;
520
	resv_huge_pages += delta;
521
522
	ret = 0;
free:
523
	/* Free the needed pages to the hugetlb pool */
524
	list_for_each_entry_safe(page, tmp, &surplus_list, lru) {
525
526
		if ((--needed) < 0)
			break;
527
		list_del(&page->lru);
528
529
530
531
532
533
534
535
		enqueue_huge_page(page);
	}

	/* Free unnecessary surplus pages to the buddy allocator */
	if (!list_empty(&surplus_list)) {
		spin_unlock(&hugetlb_lock);
		list_for_each_entry_safe(page, tmp, &surplus_list, lru) {
			list_del(&page->lru);
536
			/*
537
538
539
			 * The page has a reference count of zero already, so
			 * call free_huge_page directly instead of using
			 * put_page.  This must be done with hugetlb_lock
540
541
542
			 * unlocked which is safe because free_huge_page takes
			 * hugetlb_lock before deciding how to free the page.
			 */
543
			free_huge_page(page);
544
		}
545
		spin_lock(&hugetlb_lock);
546
547
548
549
550
551
552
553
554
555
	}

	return ret;
}

/*
 * When releasing a hugetlb pool reservation, any surplus pages that were
 * allocated to satisfy the reservation must be explicitly freed if they were
 * never used.
 */
556
static void return_unused_surplus_pages(unsigned long unused_resv_pages)
557
558
559
560
561
{
	static int nid = -1;
	struct page *page;
	unsigned long nr_pages;

562
563
564
565
566
567
568
569
	/*
	 * We want to release as many surplus pages as possible, spread
	 * evenly across all nodes. Iterate across all nodes until we
	 * can no longer free unreserved surplus pages. This occurs when
	 * the nodes with surplus pages have no free pages.
	 */
	unsigned long remaining_iterations = num_online_nodes();

570
571
572
	/* Uncommit the reservation */
	resv_huge_pages -= unused_resv_pages;

573
574
	nr_pages = min(unused_resv_pages, surplus_huge_pages);

575
	while (remaining_iterations-- && nr_pages) {
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
		nid = next_node(nid, node_online_map);
		if (nid == MAX_NUMNODES)
			nid = first_node(node_online_map);

		if (!surplus_huge_pages_node[nid])
			continue;

		if (!list_empty(&hugepage_freelists[nid])) {
			page = list_entry(hugepage_freelists[nid].next,
					  struct page, lru);
			list_del(&page->lru);
			update_and_free_page(page);
			free_huge_pages--;
			free_huge_pages_node[nid]--;
			surplus_huge_pages--;
			surplus_huge_pages_node[nid]--;
			nr_pages--;
593
			remaining_iterations = num_online_nodes();
594
595
596
597
		}
	}
}

598
static struct page *alloc_huge_page(struct vm_area_struct *vma,
599
				    unsigned long addr, int avoid_reserve)
Linus Torvalds's avatar
Linus Torvalds committed
600
{
601
	struct page *page;
602
603
604
605
606
607
608
609
610
	struct address_space *mapping = vma->vm_file->f_mapping;
	struct inode *inode = mapping->host;
	unsigned int chg = 0;

	/*
	 * Processes that did not create the mapping will have no reserves and
	 * will not have accounted against quota. Check that the quota can be
	 * made before satisfying the allocation
	 */
611
612
	if (!(vma->vm_flags & VM_SHARED) &&
			!is_vma_resv_set(vma, HPAGE_RESV_OWNER)) {
613
614
615
616
		chg = 1;
		if (hugetlb_get_quota(inode->i_mapping, chg))
			return ERR_PTR(-ENOSPC);
	}
Linus Torvalds's avatar
Linus Torvalds committed
617
618

	spin_lock(&hugetlb_lock);
619
	page = dequeue_huge_page_vma(vma, addr, avoid_reserve);
Linus Torvalds's avatar
Linus Torvalds committed
620
	spin_unlock(&hugetlb_lock);
621

Ken Chen's avatar
Ken Chen committed
622
	if (!page) {
623
		page = alloc_buddy_huge_page(vma, addr);
Ken Chen's avatar
Ken Chen committed
624
		if (!page) {
625
			hugetlb_put_quota(inode->i_mapping, chg);
Ken Chen's avatar
Ken Chen committed
626
627
628
			return ERR_PTR(-VM_FAULT_OOM);
		}
	}
629

630
631
	set_page_refcounted(page);
	set_page_private(page, (unsigned long) mapping);
632
633

	return page;
634
635
}

Linus Torvalds's avatar
Linus Torvalds committed
636
637
638
639
static int __init hugetlb_init(void)
{
	unsigned long i;

640
641
642
	if (HPAGE_SHIFT == 0)
		return 0;

Linus Torvalds's avatar
Linus Torvalds committed
643
644
645
	for (i = 0; i < MAX_NUMNODES; ++i)
		INIT_LIST_HEAD(&hugepage_freelists[i]);

646
647
	hugetlb_next_nid = first_node(node_online_map);

Linus Torvalds's avatar
Linus Torvalds committed
648
	for (i = 0; i < max_huge_pages; ++i) {
649
		if (!alloc_fresh_huge_page())
Linus Torvalds's avatar
Linus Torvalds committed
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
			break;
	}
	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);

666
667
668
669
670
671
672
673
674
675
676
static unsigned int cpuset_mems_nr(unsigned int *array)
{
	int node;
	unsigned int nr = 0;

	for_each_node_mask(node, cpuset_current_mems_allowed)
		nr += array[node];

	return nr;
}

Linus Torvalds's avatar
Linus Torvalds committed
677
678
679
680
#ifdef CONFIG_SYSCTL
#ifdef CONFIG_HIGHMEM
static void try_to_free_low(unsigned long count)
{
681
682
	int i;

Linus Torvalds's avatar
Linus Torvalds committed
683
684
685
	for (i = 0; i < MAX_NUMNODES; ++i) {
		struct page *page, *next;
		list_for_each_entry_safe(page, next, &hugepage_freelists[i], lru) {
686
687
			if (count >= nr_huge_pages)
				return;
Linus Torvalds's avatar
Linus Torvalds committed
688
689
690
691
692
			if (PageHighMem(page))
				continue;
			list_del(&page->lru);
			update_and_free_page(page);
			free_huge_pages--;
693
			free_huge_pages_node[page_to_nid(page)]--;
Linus Torvalds's avatar
Linus Torvalds committed
694
695
696
697
698
699
700
701
702
		}
	}
}
#else
static inline void try_to_free_low(unsigned long count)
{
}
#endif

703
#define persistent_huge_pages (nr_huge_pages - surplus_huge_pages)
Linus Torvalds's avatar
Linus Torvalds committed
704
705
static unsigned long set_max_huge_pages(unsigned long count)
{
706
	unsigned long min_count, ret;
Linus Torvalds's avatar
Linus Torvalds committed
707

708
709
710
711
	/*
	 * Increase the pool size
	 * First take pages out of surplus state.  Then make up the
	 * remaining difference by allocating fresh huge pages.
712
713
714
715
716
717
	 *
	 * We might race with alloc_buddy_huge_page() here and be unable
	 * to convert a surplus huge page to a normal huge page. That is
	 * not critical, though, it just means the overall size of the
	 * pool might be one hugepage larger than it needs to be, but
	 * within all the constraints specified by the sysctls.
718
	 */
Linus Torvalds's avatar
Linus Torvalds committed
719
	spin_lock(&hugetlb_lock);
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
	while (surplus_huge_pages && count > persistent_huge_pages) {
		if (!adjust_pool_surplus(-1))
			break;
	}

	while (count > persistent_huge_pages) {
		/*
		 * If this allocation races such that we no longer need the
		 * page, free_huge_page will handle it by freeing the page
		 * and reducing the surplus.
		 */
		spin_unlock(&hugetlb_lock);
		ret = alloc_fresh_huge_page();
		spin_lock(&hugetlb_lock);
		if (!ret)
			goto out;

	}

	/*
	 * Decrease the pool size
	 * First return free pages to the buddy allocator (being careful
	 * to keep enough around to satisfy reservations).  Then place
	 * pages into surplus state as needed so the pool will shrink
	 * to the desired size as pages become free.
745
746
747
748
749
750
751
752
	 *
	 * By placing pages into the surplus state independent of the
	 * overcommit value, we are allowing the surplus pool size to
	 * exceed overcommit. There are few sane options here. Since
	 * alloc_buddy_huge_page() is checking the global counter,
	 * though, we'll note that we're not allowed to exceed surplus
	 * and won't grow the pool anywhere else. Not until one of the
	 * sysctls are changed, or the surplus pages go out of use.
753
	 */
754
755
	min_count = resv_huge_pages + nr_huge_pages - free_huge_pages;
	min_count = max(count, min_count);
756
757
	try_to_free_low(min_count);
	while (min_count < persistent_huge_pages) {
758
		struct page *page = dequeue_huge_page();
Linus Torvalds's avatar
Linus Torvalds committed
759
760
761
762
		if (!page)
			break;
		update_and_free_page(page);
	}
763
764
765
766
767
768
	while (count < persistent_huge_pages) {
		if (!adjust_pool_surplus(1))
			break;
	}
out:
	ret = persistent_huge_pages;
Linus Torvalds's avatar
Linus Torvalds committed
769
	spin_unlock(&hugetlb_lock);
770
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
771
772
773
774
775
776
777
778
779
780
}

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;
}
781
782
783
784
785
786
787
788
789
790
791
792
793

int hugetlb_treat_movable_handler(struct ctl_table *table, int write,
			struct file *file, void __user *buffer,
			size_t *length, loff_t *ppos)
{
	proc_dointvec(table, write, file, buffer, length, ppos);
	if (hugepages_treat_as_movable)
		htlb_alloc_mask = GFP_HIGHUSER_MOVABLE;
	else
		htlb_alloc_mask = GFP_HIGHUSER;
	return 0;
}

794
795
796
797
798
int hugetlb_overcommit_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);
799
800
	spin_lock(&hugetlb_lock);
	nr_overcommit_huge_pages = sysctl_overcommit_huge_pages;
801
802
803
804
	spin_unlock(&hugetlb_lock);
	return 0;
}

Linus Torvalds's avatar
Linus Torvalds committed
805
806
807
808
809
810
811
#endif /* CONFIG_SYSCTL */

int hugetlb_report_meminfo(char *buf)
{
	return sprintf(buf,
			"HugePages_Total: %5lu\n"
			"HugePages_Free:  %5lu\n"
812
			"HugePages_Rsvd:  %5lu\n"
813
			"HugePages_Surp:  %5lu\n"
Linus Torvalds's avatar
Linus Torvalds committed
814
815
816
			"Hugepagesize:    %5lu kB\n",
			nr_huge_pages,
			free_huge_pages,
817
			resv_huge_pages,
818
			surplus_huge_pages,
Linus Torvalds's avatar
Linus Torvalds committed
819
820
821
822
823
824
825
			HPAGE_SIZE/1024);
}

int hugetlb_report_node_meminfo(int nid, char *buf)
{
	return sprintf(buf,
		"Node %d HugePages_Total: %5u\n"
826
827
		"Node %d HugePages_Free:  %5u\n"
		"Node %d HugePages_Surp:  %5u\n",
Linus Torvalds's avatar
Linus Torvalds committed
828
		nid, nr_huge_pages_node[nid],
829
830
		nid, free_huge_pages_node[nid],
		nid, surplus_huge_pages_node[nid]);
Linus Torvalds's avatar
Linus Torvalds committed
831
832
833
834
835
836
837
838
}

/* 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);
}

839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
static int hugetlb_acct_memory(long delta)
{
	int ret = -ENOMEM;

	spin_lock(&hugetlb_lock);
	/*
	 * When cpuset is configured, it breaks the strict hugetlb page
	 * reservation as the accounting is done on a global variable. Such
	 * reservation is completely rubbish in the presence of cpuset because
	 * the reservation is not checked against page availability for the
	 * current cpuset. Application can still potentially OOM'ed by kernel
	 * with lack of free htlb page in cpuset that the task is in.
	 * Attempt to enforce strict accounting with cpuset is almost
	 * impossible (or too ugly) because cpuset is too fluid that
	 * task or memory node can be dynamically moved between cpusets.
	 *
	 * The change of semantics for shared hugetlb mapping with cpuset is
	 * undesirable. However, in order to preserve some of the semantics,
	 * we fall back to check against current free page availability as
	 * a best attempt and hopefully to minimize the impact of changing
	 * semantics that cpuset has.
	 */
	if (delta > 0) {
		if (gather_surplus_pages(delta) < 0)
			goto out;

		if (delta > cpuset_mems_nr(free_huge_pages_node)) {
			return_unused_surplus_pages(delta);
			goto out;
		}
	}

	ret = 0;
	if (delta < 0)
		return_unused_surplus_pages((unsigned long) -delta);

out:
	spin_unlock(&hugetlb_lock);
	return ret;
}

880
881
882
883
884
885
886
static void hugetlb_vm_op_close(struct vm_area_struct *vma)
{
	unsigned long reserve = vma_resv_huge_pages(vma);
	if (reserve)
		hugetlb_acct_memory(-reserve);
}

Linus Torvalds's avatar
Linus Torvalds committed
887
888
889
890
891
892
/*
 * 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.
 */
Nick Piggin's avatar
Nick Piggin committed
893
static int hugetlb_vm_op_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Linus Torvalds's avatar
Linus Torvalds committed
894
895
{
	BUG();
Nick Piggin's avatar
Nick Piggin committed
896
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
897
898
899
}

struct vm_operations_struct hugetlb_vm_ops = {
Nick Piggin's avatar
Nick Piggin committed
900
	.fault = hugetlb_vm_op_fault,
901
	.close = hugetlb_vm_op_close,
Linus Torvalds's avatar
Linus Torvalds committed
902
903
};

904
905
static pte_t make_huge_pte(struct vm_area_struct *vma, struct page *page,
				int writable)
David Gibson's avatar
David Gibson committed
906
907
908
{
	pte_t entry;

909
	if (writable) {
David Gibson's avatar
David Gibson committed
910
911
912
		entry =
		    pte_mkwrite(pte_mkdirty(mk_pte(page, vma->vm_page_prot)));
	} else {
913
		entry = huge_pte_wrprotect(mk_pte(page, vma->vm_page_prot));
David Gibson's avatar
David Gibson committed
914
915
916
917
918
919
920
	}
	entry = pte_mkyoung(entry);
	entry = pte_mkhuge(entry);

	return entry;
}

921
922
923
924
925
static void set_huge_ptep_writable(struct vm_area_struct *vma,
				   unsigned long address, pte_t *ptep)
{
	pte_t entry;

926
927
	entry = pte_mkwrite(pte_mkdirty(huge_ptep_get(ptep)));
	if (huge_ptep_set_access_flags(vma, address, ptep, entry, 1)) {
928
929
		update_mmu_cache(vma, address, entry);
	}
930
931
932
}


David Gibson's avatar
David Gibson committed
933
934
935
936
937
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;
938
	unsigned long addr;
939
940
941
	int cow;

	cow = (vma->vm_flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
David Gibson's avatar
David Gibson committed
942

943
	for (addr = vma->vm_start; addr < vma->vm_end; addr += HPAGE_SIZE) {
944
945
946
		src_pte = huge_pte_offset(src, addr);
		if (!src_pte)
			continue;
David Gibson's avatar
David Gibson committed
947
948
949
		dst_pte = huge_pte_alloc(dst, addr);
		if (!dst_pte)
			goto nomem;
950
951
952
953
954

		/* If the pagetables are shared don't copy or take references */
		if (dst_pte == src_pte)
			continue;

955
		spin_lock(&dst->page_table_lock);
Nick Piggin's avatar
Nick Piggin committed
956
		spin_lock_nested(&src->page_table_lock, SINGLE_DEPTH_NESTING);
957
		if (!huge_pte_none(huge_ptep_get(src_pte))) {
958
			if (cow)
959
960
				huge_ptep_set_wrprotect(src, addr, src_pte);
			entry = huge_ptep_get(src_pte);
961
962
963
964
965
			ptepage = pte_page(entry);
			get_page(ptepage);
			set_huge_pte_at(dst, addr, dst_pte, entry);
		}
		spin_unlock(&src->page_table_lock);
966
		spin_unlock(&dst->page_table_lock);
David Gibson's avatar
David Gibson committed
967
968
969
970
971
972
973
	}
	return 0;

nomem:
	return -ENOMEM;
}

974
void __unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
975
			    unsigned long end, struct page *ref_page)
David Gibson's avatar
David Gibson committed
976
977
978
{
	struct mm_struct *mm = vma->vm_mm;
	unsigned long address;
979
	pte_t *ptep;
David Gibson's avatar
David Gibson committed
980
981
	pte_t pte;
	struct page *page;
982
	struct page *tmp;
983
984
985
986
987
	/*
	 * A page gathering list, protected by per file i_mmap_lock. The
	 * lock is used to avoid list corruption from multiple unmapping
	 * of the same page since we are using page->lru.
	 */
988
	LIST_HEAD(page_list);
David Gibson's avatar
David Gibson committed
989
990
991
992
993

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

994
	spin_lock(&mm->page_table_lock);
David Gibson's avatar
David Gibson committed
995
	for (address = start; address < end; address += HPAGE_SIZE) {
996
		ptep = huge_pte_offset(mm, address);
997
		if (!ptep)
998
999
			continue;

1000
		if (huge_pmd_unshare(mm, &address, ptep))