huge_memory.c 73.7 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14
/*
 *  Copyright (C) 2009  Red Hat, Inc.
 *
 *  This work is licensed under the terms of the GNU GPL, version 2. See
 *  the COPYING file in the top-level directory.
 */

#include <linux/mm.h>
#include <linux/sched.h>
#include <linux/highmem.h>
#include <linux/hugetlb.h>
#include <linux/mmu_notifier.h>
#include <linux/rmap.h>
#include <linux/swap.h>
15
#include <linux/shrinker.h>
A
Andrea Arcangeli 已提交
16 17 18
#include <linux/mm_inline.h>
#include <linux/kthread.h>
#include <linux/khugepaged.h>
19
#include <linux/freezer.h>
A
Andrea Arcangeli 已提交
20
#include <linux/mman.h>
R
Ralf Baechle 已提交
21
#include <linux/pagemap.h>
22
#include <linux/migrate.h>
23
#include <linux/hashtable.h>
24

25 26 27 28
#include <asm/tlb.h>
#include <asm/pgalloc.h>
#include "internal.h"

A
Andrea Arcangeli 已提交
29 30 31 32 33 34 35
/*
 * By default transparent hugepage support is enabled for all mappings
 * and khugepaged scans all mappings. Defrag is only invoked by
 * khugepaged hugepage allocations and by page faults inside
 * MADV_HUGEPAGE regions to avoid the risk of slowing down short lived
 * allocations.
 */
36
unsigned long transparent_hugepage_flags __read_mostly =
37
#ifdef CONFIG_TRANSPARENT_HUGEPAGE_ALWAYS
A
Andrea Arcangeli 已提交
38
	(1<<TRANSPARENT_HUGEPAGE_FLAG)|
39 40 41 42
#endif
#ifdef CONFIG_TRANSPARENT_HUGEPAGE_MADVISE
	(1<<TRANSPARENT_HUGEPAGE_REQ_MADV_FLAG)|
#endif
A
Andrea Arcangeli 已提交
43
	(1<<TRANSPARENT_HUGEPAGE_DEFRAG_FLAG)|
44 45
	(1<<TRANSPARENT_HUGEPAGE_DEFRAG_KHUGEPAGED_FLAG)|
	(1<<TRANSPARENT_HUGEPAGE_USE_ZERO_PAGE_FLAG);
A
Andrea Arcangeli 已提交
46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67

/* default scan 8*512 pte (or vmas) every 30 second */
static unsigned int khugepaged_pages_to_scan __read_mostly = HPAGE_PMD_NR*8;
static unsigned int khugepaged_pages_collapsed;
static unsigned int khugepaged_full_scans;
static unsigned int khugepaged_scan_sleep_millisecs __read_mostly = 10000;
/* during fragmentation poll the hugepage allocator once every minute */
static unsigned int khugepaged_alloc_sleep_millisecs __read_mostly = 60000;
static struct task_struct *khugepaged_thread __read_mostly;
static DEFINE_MUTEX(khugepaged_mutex);
static DEFINE_SPINLOCK(khugepaged_mm_lock);
static DECLARE_WAIT_QUEUE_HEAD(khugepaged_wait);
/*
 * default collapse hugepages if there is at least one pte mapped like
 * it would have happened if the vma was large enough during page
 * fault.
 */
static unsigned int khugepaged_max_ptes_none __read_mostly = HPAGE_PMD_NR-1;

static int khugepaged(void *none);
static int khugepaged_slab_init(void);

68 69 70
#define MM_SLOTS_HASH_BITS 10
static __read_mostly DEFINE_HASHTABLE(mm_slots_hash, MM_SLOTS_HASH_BITS);

A
Andrea Arcangeli 已提交
71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96
static struct kmem_cache *mm_slot_cache __read_mostly;

/**
 * struct mm_slot - hash lookup from mm to mm_slot
 * @hash: hash collision list
 * @mm_node: khugepaged scan list headed in khugepaged_scan.mm_head
 * @mm: the mm that this information is valid for
 */
struct mm_slot {
	struct hlist_node hash;
	struct list_head mm_node;
	struct mm_struct *mm;
};

/**
 * struct khugepaged_scan - cursor for scanning
 * @mm_head: the head of the mm list to scan
 * @mm_slot: the current mm_slot we are scanning
 * @address: the next address inside that to be scanned
 *
 * There is only the one khugepaged_scan instance of this cursor structure.
 */
struct khugepaged_scan {
	struct list_head mm_head;
	struct mm_slot *mm_slot;
	unsigned long address;
97 98
};
static struct khugepaged_scan khugepaged_scan = {
A
Andrea Arcangeli 已提交
99 100 101
	.mm_head = LIST_HEAD_INIT(khugepaged_scan.mm_head),
};

102 103 104 105 106 107 108

static int set_recommended_min_free_kbytes(void)
{
	struct zone *zone;
	int nr_zones = 0;
	unsigned long recommended_min;

109
	if (!khugepaged_enabled())
110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138
		return 0;

	for_each_populated_zone(zone)
		nr_zones++;

	/* Make sure at least 2 hugepages are free for MIGRATE_RESERVE */
	recommended_min = pageblock_nr_pages * nr_zones * 2;

	/*
	 * Make sure that on average at least two pageblocks are almost free
	 * of another type, one for a migratetype to fall back to and a
	 * second to avoid subsequent fallbacks of other types There are 3
	 * MIGRATE_TYPES we care about.
	 */
	recommended_min += pageblock_nr_pages * nr_zones *
			   MIGRATE_PCPTYPES * MIGRATE_PCPTYPES;

	/* don't ever allow to reserve more than 5% of the lowmem */
	recommended_min = min(recommended_min,
			      (unsigned long) nr_free_buffer_pages() / 20);
	recommended_min <<= (PAGE_SHIFT-10);

	if (recommended_min > min_free_kbytes)
		min_free_kbytes = recommended_min;
	setup_per_zone_wmarks();
	return 0;
}
late_initcall(set_recommended_min_free_kbytes);

A
Andrea Arcangeli 已提交
139 140 141 142 143 144 145 146 147 148 149 150 151
static int start_khugepaged(void)
{
	int err = 0;
	if (khugepaged_enabled()) {
		if (!khugepaged_thread)
			khugepaged_thread = kthread_run(khugepaged, NULL,
							"khugepaged");
		if (unlikely(IS_ERR(khugepaged_thread))) {
			printk(KERN_ERR
			       "khugepaged: kthread_run(khugepaged) failed\n");
			err = PTR_ERR(khugepaged_thread);
			khugepaged_thread = NULL;
		}
152 153

		if (!list_empty(&khugepaged_scan.mm_head))
A
Andrea Arcangeli 已提交
154
			wake_up_interruptible(&khugepaged_wait);
155 156

		set_recommended_min_free_kbytes();
157 158 159 160
	} else if (khugepaged_thread) {
		kthread_stop(khugepaged_thread);
		khugepaged_thread = NULL;
	}
161

A
Andrea Arcangeli 已提交
162 163
	return err;
}
164

165
static atomic_t huge_zero_refcount;
166
static struct page *huge_zero_page __read_mostly;
167

168
static inline bool is_huge_zero_page(struct page *page)
169
{
170
	return ACCESS_ONCE(huge_zero_page) == page;
171
}
172

173 174
static inline bool is_huge_zero_pmd(pmd_t pmd)
{
175
	return is_huge_zero_page(pmd_page(pmd));
176 177
}

178
static struct page *get_huge_zero_page(void)
179 180 181 182
{
	struct page *zero_page;
retry:
	if (likely(atomic_inc_not_zero(&huge_zero_refcount)))
183
		return ACCESS_ONCE(huge_zero_page);
184 185

	zero_page = alloc_pages((GFP_TRANSHUGE | __GFP_ZERO) & ~__GFP_MOVABLE,
186
			HPAGE_PMD_ORDER);
187 188
	if (!zero_page) {
		count_vm_event(THP_ZERO_PAGE_ALLOC_FAILED);
189
		return NULL;
190 191
	}
	count_vm_event(THP_ZERO_PAGE_ALLOC);
192
	preempt_disable();
193
	if (cmpxchg(&huge_zero_page, NULL, zero_page)) {
194 195 196 197 198 199 200 201
		preempt_enable();
		__free_page(zero_page);
		goto retry;
	}

	/* We take additional reference here. It will be put back by shrinker */
	atomic_set(&huge_zero_refcount, 2);
	preempt_enable();
202
	return ACCESS_ONCE(huge_zero_page);
203 204
}

205
static void put_huge_zero_page(void)
206
{
207 208 209 210 211
	/*
	 * Counter should never go to zero here. Only shrinker can put
	 * last reference.
	 */
	BUG_ON(atomic_dec_and_test(&huge_zero_refcount));
212 213
}

214 215
static int shrink_huge_zero_page(struct shrinker *shrink,
		struct shrink_control *sc)
216
{
217 218 219 220 221
	if (!sc->nr_to_scan)
		/* we can free zero page only if last reference remains */
		return atomic_read(&huge_zero_refcount) == 1 ? HPAGE_PMD_NR : 0;

	if (atomic_cmpxchg(&huge_zero_refcount, 1, 0) == 1) {
222 223 224
		struct page *zero_page = xchg(&huge_zero_page, NULL);
		BUG_ON(zero_page == NULL);
		__free_page(zero_page);
225 226 227
	}

	return 0;
228 229
}

230 231 232 233 234
static struct shrinker huge_zero_page_shrinker = {
	.shrink = shrink_huge_zero_page,
	.seeks = DEFAULT_SEEKS,
};

235
#ifdef CONFIG_SYSFS
A
Andrea Arcangeli 已提交
236

237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284
static ssize_t double_flag_show(struct kobject *kobj,
				struct kobj_attribute *attr, char *buf,
				enum transparent_hugepage_flag enabled,
				enum transparent_hugepage_flag req_madv)
{
	if (test_bit(enabled, &transparent_hugepage_flags)) {
		VM_BUG_ON(test_bit(req_madv, &transparent_hugepage_flags));
		return sprintf(buf, "[always] madvise never\n");
	} else if (test_bit(req_madv, &transparent_hugepage_flags))
		return sprintf(buf, "always [madvise] never\n");
	else
		return sprintf(buf, "always madvise [never]\n");
}
static ssize_t double_flag_store(struct kobject *kobj,
				 struct kobj_attribute *attr,
				 const char *buf, size_t count,
				 enum transparent_hugepage_flag enabled,
				 enum transparent_hugepage_flag req_madv)
{
	if (!memcmp("always", buf,
		    min(sizeof("always")-1, count))) {
		set_bit(enabled, &transparent_hugepage_flags);
		clear_bit(req_madv, &transparent_hugepage_flags);
	} else if (!memcmp("madvise", buf,
			   min(sizeof("madvise")-1, count))) {
		clear_bit(enabled, &transparent_hugepage_flags);
		set_bit(req_madv, &transparent_hugepage_flags);
	} else if (!memcmp("never", buf,
			   min(sizeof("never")-1, count))) {
		clear_bit(enabled, &transparent_hugepage_flags);
		clear_bit(req_madv, &transparent_hugepage_flags);
	} else
		return -EINVAL;

	return count;
}

static ssize_t enabled_show(struct kobject *kobj,
			    struct kobj_attribute *attr, char *buf)
{
	return double_flag_show(kobj, attr, buf,
				TRANSPARENT_HUGEPAGE_FLAG,
				TRANSPARENT_HUGEPAGE_REQ_MADV_FLAG);
}
static ssize_t enabled_store(struct kobject *kobj,
			     struct kobj_attribute *attr,
			     const char *buf, size_t count)
{
A
Andrea Arcangeli 已提交
285 286 287 288 289 290 291
	ssize_t ret;

	ret = double_flag_store(kobj, attr, buf, count,
				TRANSPARENT_HUGEPAGE_FLAG,
				TRANSPARENT_HUGEPAGE_REQ_MADV_FLAG);

	if (ret > 0) {
292 293 294 295 296 297
		int err;

		mutex_lock(&khugepaged_mutex);
		err = start_khugepaged();
		mutex_unlock(&khugepaged_mutex);

A
Andrea Arcangeli 已提交
298 299 300 301 302
		if (err)
			ret = err;
	}

	return ret;
303 304 305 306 307 308 309 310
}
static struct kobj_attribute enabled_attr =
	__ATTR(enabled, 0644, enabled_show, enabled_store);

static ssize_t single_flag_show(struct kobject *kobj,
				struct kobj_attribute *attr, char *buf,
				enum transparent_hugepage_flag flag)
{
311 312
	return sprintf(buf, "%d\n",
		       !!test_bit(flag, &transparent_hugepage_flags));
313
}
314

315 316 317 318 319
static ssize_t single_flag_store(struct kobject *kobj,
				 struct kobj_attribute *attr,
				 const char *buf, size_t count,
				 enum transparent_hugepage_flag flag)
{
320 321 322 323 324 325 326 327 328 329
	unsigned long value;
	int ret;

	ret = kstrtoul(buf, 10, &value);
	if (ret < 0)
		return ret;
	if (value > 1)
		return -EINVAL;

	if (value)
330
		set_bit(flag, &transparent_hugepage_flags);
331
	else
332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359
		clear_bit(flag, &transparent_hugepage_flags);

	return count;
}

/*
 * Currently defrag only disables __GFP_NOWAIT for allocation. A blind
 * __GFP_REPEAT is too aggressive, it's never worth swapping tons of
 * memory just to allocate one more hugepage.
 */
static ssize_t defrag_show(struct kobject *kobj,
			   struct kobj_attribute *attr, char *buf)
{
	return double_flag_show(kobj, attr, buf,
				TRANSPARENT_HUGEPAGE_DEFRAG_FLAG,
				TRANSPARENT_HUGEPAGE_DEFRAG_REQ_MADV_FLAG);
}
static ssize_t defrag_store(struct kobject *kobj,
			    struct kobj_attribute *attr,
			    const char *buf, size_t count)
{
	return double_flag_store(kobj, attr, buf, count,
				 TRANSPARENT_HUGEPAGE_DEFRAG_FLAG,
				 TRANSPARENT_HUGEPAGE_DEFRAG_REQ_MADV_FLAG);
}
static struct kobj_attribute defrag_attr =
	__ATTR(defrag, 0644, defrag_show, defrag_store);

360 361 362 363 364 365 366 367 368 369 370 371 372 373
static ssize_t use_zero_page_show(struct kobject *kobj,
		struct kobj_attribute *attr, char *buf)
{
	return single_flag_show(kobj, attr, buf,
				TRANSPARENT_HUGEPAGE_USE_ZERO_PAGE_FLAG);
}
static ssize_t use_zero_page_store(struct kobject *kobj,
		struct kobj_attribute *attr, const char *buf, size_t count)
{
	return single_flag_store(kobj, attr, buf, count,
				 TRANSPARENT_HUGEPAGE_USE_ZERO_PAGE_FLAG);
}
static struct kobj_attribute use_zero_page_attr =
	__ATTR(use_zero_page, 0644, use_zero_page_show, use_zero_page_store);
374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394
#ifdef CONFIG_DEBUG_VM
static ssize_t debug_cow_show(struct kobject *kobj,
				struct kobj_attribute *attr, char *buf)
{
	return single_flag_show(kobj, attr, buf,
				TRANSPARENT_HUGEPAGE_DEBUG_COW_FLAG);
}
static ssize_t debug_cow_store(struct kobject *kobj,
			       struct kobj_attribute *attr,
			       const char *buf, size_t count)
{
	return single_flag_store(kobj, attr, buf, count,
				 TRANSPARENT_HUGEPAGE_DEBUG_COW_FLAG);
}
static struct kobj_attribute debug_cow_attr =
	__ATTR(debug_cow, 0644, debug_cow_show, debug_cow_store);
#endif /* CONFIG_DEBUG_VM */

static struct attribute *hugepage_attr[] = {
	&enabled_attr.attr,
	&defrag_attr.attr,
395
	&use_zero_page_attr.attr,
396 397 398 399 400 401 402 403
#ifdef CONFIG_DEBUG_VM
	&debug_cow_attr.attr,
#endif
	NULL,
};

static struct attribute_group hugepage_attr_group = {
	.attrs = hugepage_attr,
A
Andrea Arcangeli 已提交
404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566
};

static ssize_t scan_sleep_millisecs_show(struct kobject *kobj,
					 struct kobj_attribute *attr,
					 char *buf)
{
	return sprintf(buf, "%u\n", khugepaged_scan_sleep_millisecs);
}

static ssize_t scan_sleep_millisecs_store(struct kobject *kobj,
					  struct kobj_attribute *attr,
					  const char *buf, size_t count)
{
	unsigned long msecs;
	int err;

	err = strict_strtoul(buf, 10, &msecs);
	if (err || msecs > UINT_MAX)
		return -EINVAL;

	khugepaged_scan_sleep_millisecs = msecs;
	wake_up_interruptible(&khugepaged_wait);

	return count;
}
static struct kobj_attribute scan_sleep_millisecs_attr =
	__ATTR(scan_sleep_millisecs, 0644, scan_sleep_millisecs_show,
	       scan_sleep_millisecs_store);

static ssize_t alloc_sleep_millisecs_show(struct kobject *kobj,
					  struct kobj_attribute *attr,
					  char *buf)
{
	return sprintf(buf, "%u\n", khugepaged_alloc_sleep_millisecs);
}

static ssize_t alloc_sleep_millisecs_store(struct kobject *kobj,
					   struct kobj_attribute *attr,
					   const char *buf, size_t count)
{
	unsigned long msecs;
	int err;

	err = strict_strtoul(buf, 10, &msecs);
	if (err || msecs > UINT_MAX)
		return -EINVAL;

	khugepaged_alloc_sleep_millisecs = msecs;
	wake_up_interruptible(&khugepaged_wait);

	return count;
}
static struct kobj_attribute alloc_sleep_millisecs_attr =
	__ATTR(alloc_sleep_millisecs, 0644, alloc_sleep_millisecs_show,
	       alloc_sleep_millisecs_store);

static ssize_t pages_to_scan_show(struct kobject *kobj,
				  struct kobj_attribute *attr,
				  char *buf)
{
	return sprintf(buf, "%u\n", khugepaged_pages_to_scan);
}
static ssize_t pages_to_scan_store(struct kobject *kobj,
				   struct kobj_attribute *attr,
				   const char *buf, size_t count)
{
	int err;
	unsigned long pages;

	err = strict_strtoul(buf, 10, &pages);
	if (err || !pages || pages > UINT_MAX)
		return -EINVAL;

	khugepaged_pages_to_scan = pages;

	return count;
}
static struct kobj_attribute pages_to_scan_attr =
	__ATTR(pages_to_scan, 0644, pages_to_scan_show,
	       pages_to_scan_store);

static ssize_t pages_collapsed_show(struct kobject *kobj,
				    struct kobj_attribute *attr,
				    char *buf)
{
	return sprintf(buf, "%u\n", khugepaged_pages_collapsed);
}
static struct kobj_attribute pages_collapsed_attr =
	__ATTR_RO(pages_collapsed);

static ssize_t full_scans_show(struct kobject *kobj,
			       struct kobj_attribute *attr,
			       char *buf)
{
	return sprintf(buf, "%u\n", khugepaged_full_scans);
}
static struct kobj_attribute full_scans_attr =
	__ATTR_RO(full_scans);

static ssize_t khugepaged_defrag_show(struct kobject *kobj,
				      struct kobj_attribute *attr, char *buf)
{
	return single_flag_show(kobj, attr, buf,
				TRANSPARENT_HUGEPAGE_DEFRAG_KHUGEPAGED_FLAG);
}
static ssize_t khugepaged_defrag_store(struct kobject *kobj,
				       struct kobj_attribute *attr,
				       const char *buf, size_t count)
{
	return single_flag_store(kobj, attr, buf, count,
				 TRANSPARENT_HUGEPAGE_DEFRAG_KHUGEPAGED_FLAG);
}
static struct kobj_attribute khugepaged_defrag_attr =
	__ATTR(defrag, 0644, khugepaged_defrag_show,
	       khugepaged_defrag_store);

/*
 * max_ptes_none controls if khugepaged should collapse hugepages over
 * any unmapped ptes in turn potentially increasing the memory
 * footprint of the vmas. When max_ptes_none is 0 khugepaged will not
 * reduce the available free memory in the system as it
 * runs. Increasing max_ptes_none will instead potentially reduce the
 * free memory in the system during the khugepaged scan.
 */
static ssize_t khugepaged_max_ptes_none_show(struct kobject *kobj,
					     struct kobj_attribute *attr,
					     char *buf)
{
	return sprintf(buf, "%u\n", khugepaged_max_ptes_none);
}
static ssize_t khugepaged_max_ptes_none_store(struct kobject *kobj,
					      struct kobj_attribute *attr,
					      const char *buf, size_t count)
{
	int err;
	unsigned long max_ptes_none;

	err = strict_strtoul(buf, 10, &max_ptes_none);
	if (err || max_ptes_none > HPAGE_PMD_NR-1)
		return -EINVAL;

	khugepaged_max_ptes_none = max_ptes_none;

	return count;
}
static struct kobj_attribute khugepaged_max_ptes_none_attr =
	__ATTR(max_ptes_none, 0644, khugepaged_max_ptes_none_show,
	       khugepaged_max_ptes_none_store);

static struct attribute *khugepaged_attr[] = {
	&khugepaged_defrag_attr.attr,
	&khugepaged_max_ptes_none_attr.attr,
	&pages_to_scan_attr.attr,
	&pages_collapsed_attr.attr,
	&full_scans_attr.attr,
	&scan_sleep_millisecs_attr.attr,
	&alloc_sleep_millisecs_attr.attr,
	NULL,
};

static struct attribute_group khugepaged_attr_group = {
	.attrs = khugepaged_attr,
	.name = "khugepaged",
567 568
};

S
Shaohua Li 已提交
569
static int __init hugepage_init_sysfs(struct kobject **hugepage_kobj)
570 571 572
{
	int err;

S
Shaohua Li 已提交
573 574
	*hugepage_kobj = kobject_create_and_add("transparent_hugepage", mm_kobj);
	if (unlikely(!*hugepage_kobj)) {
575
		printk(KERN_ERR "hugepage: failed to create transparent hugepage kobject\n");
S
Shaohua Li 已提交
576
		return -ENOMEM;
A
Andrea Arcangeli 已提交
577 578
	}

S
Shaohua Li 已提交
579
	err = sysfs_create_group(*hugepage_kobj, &hugepage_attr_group);
A
Andrea Arcangeli 已提交
580
	if (err) {
581
		printk(KERN_ERR "hugepage: failed to register transparent hugepage group\n");
S
Shaohua Li 已提交
582
		goto delete_obj;
A
Andrea Arcangeli 已提交
583 584
	}

S
Shaohua Li 已提交
585
	err = sysfs_create_group(*hugepage_kobj, &khugepaged_attr_group);
A
Andrea Arcangeli 已提交
586
	if (err) {
587
		printk(KERN_ERR "hugepage: failed to register transparent hugepage group\n");
S
Shaohua Li 已提交
588
		goto remove_hp_group;
A
Andrea Arcangeli 已提交
589
	}
S
Shaohua Li 已提交
590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629

	return 0;

remove_hp_group:
	sysfs_remove_group(*hugepage_kobj, &hugepage_attr_group);
delete_obj:
	kobject_put(*hugepage_kobj);
	return err;
}

static void __init hugepage_exit_sysfs(struct kobject *hugepage_kobj)
{
	sysfs_remove_group(hugepage_kobj, &khugepaged_attr_group);
	sysfs_remove_group(hugepage_kobj, &hugepage_attr_group);
	kobject_put(hugepage_kobj);
}
#else
static inline int hugepage_init_sysfs(struct kobject **hugepage_kobj)
{
	return 0;
}

static inline void hugepage_exit_sysfs(struct kobject *hugepage_kobj)
{
}
#endif /* CONFIG_SYSFS */

static int __init hugepage_init(void)
{
	int err;
	struct kobject *hugepage_kobj;

	if (!has_transparent_hugepage()) {
		transparent_hugepage_flags = 0;
		return -EINVAL;
	}

	err = hugepage_init_sysfs(&hugepage_kobj);
	if (err)
		return err;
A
Andrea Arcangeli 已提交
630 631 632 633 634

	err = khugepaged_slab_init();
	if (err)
		goto out;

635 636
	register_shrinker(&huge_zero_page_shrinker);

637 638 639 640 641 642 643 644
	/*
	 * By default disable transparent hugepages on smaller systems,
	 * where the extra memory used could hurt more than TLB overhead
	 * is likely to save.  The admin can still enable it through /sys.
	 */
	if (totalram_pages < (512 << (20 - PAGE_SHIFT)))
		transparent_hugepage_flags = 0;

A
Andrea Arcangeli 已提交
645 646
	start_khugepaged();

S
Shaohua Li 已提交
647
	return 0;
A
Andrea Arcangeli 已提交
648
out:
S
Shaohua Li 已提交
649
	hugepage_exit_sysfs(hugepage_kobj);
A
Andrea Arcangeli 已提交
650
	return err;
651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685
}
module_init(hugepage_init)

static int __init setup_transparent_hugepage(char *str)
{
	int ret = 0;
	if (!str)
		goto out;
	if (!strcmp(str, "always")) {
		set_bit(TRANSPARENT_HUGEPAGE_FLAG,
			&transparent_hugepage_flags);
		clear_bit(TRANSPARENT_HUGEPAGE_REQ_MADV_FLAG,
			  &transparent_hugepage_flags);
		ret = 1;
	} else if (!strcmp(str, "madvise")) {
		clear_bit(TRANSPARENT_HUGEPAGE_FLAG,
			  &transparent_hugepage_flags);
		set_bit(TRANSPARENT_HUGEPAGE_REQ_MADV_FLAG,
			&transparent_hugepage_flags);
		ret = 1;
	} else if (!strcmp(str, "never")) {
		clear_bit(TRANSPARENT_HUGEPAGE_FLAG,
			  &transparent_hugepage_flags);
		clear_bit(TRANSPARENT_HUGEPAGE_REQ_MADV_FLAG,
			  &transparent_hugepage_flags);
		ret = 1;
	}
out:
	if (!ret)
		printk(KERN_WARNING
		       "transparent_hugepage= cannot parse, ignored\n");
	return ret;
}
__setup("transparent_hugepage=", setup_transparent_hugepage);

686
pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma)
687 688 689 690 691 692
{
	if (likely(vma->vm_flags & VM_WRITE))
		pmd = pmd_mkwrite(pmd);
	return pmd;
}

B
Bob Liu 已提交
693 694 695 696 697 698 699 700 701
static inline pmd_t mk_huge_pmd(struct page *page, struct vm_area_struct *vma)
{
	pmd_t entry;
	entry = mk_pmd(page, vma->vm_page_prot);
	entry = maybe_pmd_mkwrite(pmd_mkdirty(entry), vma);
	entry = pmd_mkhuge(entry);
	return entry;
}

702 703 704 705 706 707 708 709 710
static int __do_huge_pmd_anonymous_page(struct mm_struct *mm,
					struct vm_area_struct *vma,
					unsigned long haddr, pmd_t *pmd,
					struct page *page)
{
	pgtable_t pgtable;

	VM_BUG_ON(!PageCompound(page));
	pgtable = pte_alloc_one(mm, haddr);
711
	if (unlikely(!pgtable))
712 713 714
		return VM_FAULT_OOM;

	clear_huge_page(page, haddr, HPAGE_PMD_NR);
715 716 717 718 719
	/*
	 * The memory barrier inside __SetPageUptodate makes sure that
	 * clear_huge_page writes become visible before the set_pmd_at()
	 * write.
	 */
720 721 722 723 724
	__SetPageUptodate(page);

	spin_lock(&mm->page_table_lock);
	if (unlikely(!pmd_none(*pmd))) {
		spin_unlock(&mm->page_table_lock);
A
Andrea Arcangeli 已提交
725
		mem_cgroup_uncharge_page(page);
726 727 728 729
		put_page(page);
		pte_free(mm, pgtable);
	} else {
		pmd_t entry;
B
Bob Liu 已提交
730
		entry = mk_huge_pmd(page, vma);
731 732
		page_add_new_anon_rmap(page, vma, haddr);
		set_pmd_at(mm, haddr, pmd, entry);
733
		pgtable_trans_huge_deposit(mm, pmd, pgtable);
734
		add_mm_counter(mm, MM_ANONPAGES, HPAGE_PMD_NR);
735
		mm->nr_ptes++;
736 737 738
		spin_unlock(&mm->page_table_lock);
	}

739
	return 0;
740 741
}

742
static inline gfp_t alloc_hugepage_gfpmask(int defrag, gfp_t extra_gfp)
743
{
744
	return (GFP_TRANSHUGE & ~(defrag ? 0 : __GFP_WAIT)) | extra_gfp;
745 746 747 748
}

static inline struct page *alloc_hugepage_vma(int defrag,
					      struct vm_area_struct *vma,
749 750
					      unsigned long haddr, int nd,
					      gfp_t extra_gfp)
751
{
752
	return alloc_pages_vma(alloc_hugepage_gfpmask(defrag, extra_gfp),
753
			       HPAGE_PMD_ORDER, vma, haddr, nd);
754 755 756
}

#ifndef CONFIG_NUMA
757 758
static inline struct page *alloc_hugepage(int defrag)
{
759
	return alloc_pages(alloc_hugepage_gfpmask(defrag, 0),
760 761
			   HPAGE_PMD_ORDER);
}
762
#endif
763

764
static bool set_huge_zero_page(pgtable_t pgtable, struct mm_struct *mm,
765
		struct vm_area_struct *vma, unsigned long haddr, pmd_t *pmd,
766
		struct page *zero_page)
767 768
{
	pmd_t entry;
769 770
	if (!pmd_none(*pmd))
		return false;
771
	entry = mk_pmd(zero_page, vma->vm_page_prot);
772 773 774
	entry = pmd_wrprotect(entry);
	entry = pmd_mkhuge(entry);
	set_pmd_at(mm, haddr, pmd, entry);
775
	pgtable_trans_huge_deposit(mm, pmd, pgtable);
776
	mm->nr_ptes++;
777
	return true;
778 779
}

780 781 782 783 784 785 786 787 788 789 790
int do_huge_pmd_anonymous_page(struct mm_struct *mm, struct vm_area_struct *vma,
			       unsigned long address, pmd_t *pmd,
			       unsigned int flags)
{
	struct page *page;
	unsigned long haddr = address & HPAGE_PMD_MASK;
	pte_t *pte;

	if (haddr >= vma->vm_start && haddr + HPAGE_PMD_SIZE <= vma->vm_end) {
		if (unlikely(anon_vma_prepare(vma)))
			return VM_FAULT_OOM;
A
Andrea Arcangeli 已提交
791 792
		if (unlikely(khugepaged_enter(vma)))
			return VM_FAULT_OOM;
793 794
		if (!(flags & FAULT_FLAG_WRITE) &&
				transparent_hugepage_use_zero_page()) {
795
			pgtable_t pgtable;
796
			struct page *zero_page;
797
			bool set;
798 799 800
			pgtable = pte_alloc_one(mm, haddr);
			if (unlikely(!pgtable))
				return VM_FAULT_OOM;
801 802
			zero_page = get_huge_zero_page();
			if (unlikely(!zero_page)) {
803 804 805 806
				pte_free(mm, pgtable);
				count_vm_event(THP_FAULT_FALLBACK);
				goto out;
			}
807
			spin_lock(&mm->page_table_lock);
808
			set = set_huge_zero_page(pgtable, mm, vma, haddr, pmd,
809
					zero_page);
810
			spin_unlock(&mm->page_table_lock);
811 812 813 814
			if (!set) {
				pte_free(mm, pgtable);
				put_huge_zero_page();
			}
815 816
			return 0;
		}
817
		page = alloc_hugepage_vma(transparent_hugepage_defrag(vma),
818
					  vma, haddr, numa_node_id(), 0);
819 820
		if (unlikely(!page)) {
			count_vm_event(THP_FAULT_FALLBACK);
821
			goto out;
822 823
		}
		count_vm_event(THP_FAULT_ALLOC);
A
Andrea Arcangeli 已提交
824 825 826 827
		if (unlikely(mem_cgroup_newpage_charge(page, mm, GFP_KERNEL))) {
			put_page(page);
			goto out;
		}
828 829 830 831 832 833
		if (unlikely(__do_huge_pmd_anonymous_page(mm, vma, haddr, pmd,
							  page))) {
			mem_cgroup_uncharge_page(page);
			put_page(page);
			goto out;
		}
834

835
		return 0;
836 837 838 839 840 841 842
	}
out:
	/*
	 * Use __pte_alloc instead of pte_alloc_map, because we can't
	 * run pte_offset_map on the pmd, if an huge pmd could
	 * materialize from under us from a different thread.
	 */
843 844
	if (unlikely(pmd_none(*pmd)) &&
	    unlikely(__pte_alloc(mm, vma, pmd, address)))
845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881
		return VM_FAULT_OOM;
	/* if an huge pmd materialized from under us just retry later */
	if (unlikely(pmd_trans_huge(*pmd)))
		return 0;
	/*
	 * A regular pmd is established and it can't morph into a huge pmd
	 * from under us anymore at this point because we hold the mmap_sem
	 * read mode and khugepaged takes it in write mode. So now it's
	 * safe to run pte_offset_map().
	 */
	pte = pte_offset_map(pmd, address);
	return handle_pte_fault(mm, vma, address, pte, pmd, flags);
}

int copy_huge_pmd(struct mm_struct *dst_mm, struct mm_struct *src_mm,
		  pmd_t *dst_pmd, pmd_t *src_pmd, unsigned long addr,
		  struct vm_area_struct *vma)
{
	struct page *src_page;
	pmd_t pmd;
	pgtable_t pgtable;
	int ret;

	ret = -ENOMEM;
	pgtable = pte_alloc_one(dst_mm, addr);
	if (unlikely(!pgtable))
		goto out;

	spin_lock(&dst_mm->page_table_lock);
	spin_lock_nested(&src_mm->page_table_lock, SINGLE_DEPTH_NESTING);

	ret = -EAGAIN;
	pmd = *src_pmd;
	if (unlikely(!pmd_trans_huge(pmd))) {
		pte_free(dst_mm, pgtable);
		goto out_unlock;
	}
882 883 884 885 886 887
	/*
	 * mm->page_table_lock is enough to be sure that huge zero pmd is not
	 * under splitting since we don't split the page itself, only pmd to
	 * a page table.
	 */
	if (is_huge_zero_pmd(pmd)) {
888
		struct page *zero_page;
889
		bool set;
890 891 892 893 894
		/*
		 * get_huge_zero_page() will never allocate a new page here,
		 * since we already have a zero page to copy. It just takes a
		 * reference.
		 */
895
		zero_page = get_huge_zero_page();
896
		set = set_huge_zero_page(pgtable, dst_mm, vma, addr, dst_pmd,
897
				zero_page);
898
		BUG_ON(!set); /* unexpected !pmd_none(dst_pmd) */
899 900 901
		ret = 0;
		goto out_unlock;
	}
902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919
	if (unlikely(pmd_trans_splitting(pmd))) {
		/* split huge page running from under us */
		spin_unlock(&src_mm->page_table_lock);
		spin_unlock(&dst_mm->page_table_lock);
		pte_free(dst_mm, pgtable);

		wait_split_huge_page(vma->anon_vma, src_pmd); /* src_vma */
		goto out;
	}
	src_page = pmd_page(pmd);
	VM_BUG_ON(!PageHead(src_page));
	get_page(src_page);
	page_dup_rmap(src_page);
	add_mm_counter(dst_mm, MM_ANONPAGES, HPAGE_PMD_NR);

	pmdp_set_wrprotect(src_mm, addr, src_pmd);
	pmd = pmd_mkold(pmd_wrprotect(pmd));
	set_pmd_at(dst_mm, addr, dst_pmd, pmd);
920
	pgtable_trans_huge_deposit(dst_mm, dst_pmd, pgtable);
921
	dst_mm->nr_ptes++;
922 923 924 925 926 927 928 929 930

	ret = 0;
out_unlock:
	spin_unlock(&src_mm->page_table_lock);
	spin_unlock(&dst_mm->page_table_lock);
out:
	return ret;
}

931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952
void huge_pmd_set_accessed(struct mm_struct *mm,
			   struct vm_area_struct *vma,
			   unsigned long address,
			   pmd_t *pmd, pmd_t orig_pmd,
			   int dirty)
{
	pmd_t entry;
	unsigned long haddr;

	spin_lock(&mm->page_table_lock);
	if (unlikely(!pmd_same(*pmd, orig_pmd)))
		goto unlock;

	entry = pmd_mkyoung(orig_pmd);
	haddr = address & HPAGE_PMD_MASK;
	if (pmdp_set_access_flags(vma, haddr, pmd, entry, dirty))
		update_mmu_cache_pmd(vma, address, pmd);

unlock:
	spin_unlock(&mm->page_table_lock);
}

953 954
static int do_huge_pmd_wp_zero_page_fallback(struct mm_struct *mm,
		struct vm_area_struct *vma, unsigned long address,
955
		pmd_t *pmd, pmd_t orig_pmd, unsigned long haddr)
956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983
{
	pgtable_t pgtable;
	pmd_t _pmd;
	struct page *page;
	int i, ret = 0;
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */

	page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, address);
	if (!page) {
		ret |= VM_FAULT_OOM;
		goto out;
	}

	if (mem_cgroup_newpage_charge(page, mm, GFP_KERNEL)) {
		put_page(page);
		ret |= VM_FAULT_OOM;
		goto out;
	}

	clear_user_highpage(page, address);
	__SetPageUptodate(page);

	mmun_start = haddr;
	mmun_end   = haddr + HPAGE_PMD_SIZE;
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);

	spin_lock(&mm->page_table_lock);
984 985 986
	if (unlikely(!pmd_same(*pmd, orig_pmd)))
		goto out_free_page;

987 988 989
	pmdp_clear_flush(vma, haddr, pmd);
	/* leave pmd empty until pte is filled */

990
	pgtable = pgtable_trans_huge_withdraw(mm, pmd);
991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010
	pmd_populate(mm, &_pmd, pgtable);

	for (i = 0; i < HPAGE_PMD_NR; i++, haddr += PAGE_SIZE) {
		pte_t *pte, entry;
		if (haddr == (address & PAGE_MASK)) {
			entry = mk_pte(page, vma->vm_page_prot);
			entry = maybe_mkwrite(pte_mkdirty(entry), vma);
			page_add_new_anon_rmap(page, vma, haddr);
		} else {
			entry = pfn_pte(my_zero_pfn(haddr), vma->vm_page_prot);
			entry = pte_mkspecial(entry);
		}
		pte = pte_offset_map(&_pmd, haddr);
		VM_BUG_ON(!pte_none(*pte));
		set_pte_at(mm, haddr, pte, entry);
		pte_unmap(pte);
	}
	smp_wmb(); /* make pte visible before pmd */
	pmd_populate(mm, pmd, pgtable);
	spin_unlock(&mm->page_table_lock);
1011
	put_huge_zero_page();
1012 1013 1014 1015 1016 1017 1018
	inc_mm_counter(mm, MM_ANONPAGES);

	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);

	ret |= VM_FAULT_WRITE;
out:
	return ret;
1019 1020 1021 1022 1023 1024
out_free_page:
	spin_unlock(&mm->page_table_lock);
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
	mem_cgroup_uncharge_page(page);
	put_page(page);
	goto out;
1025 1026
}

1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037
static int do_huge_pmd_wp_page_fallback(struct mm_struct *mm,
					struct vm_area_struct *vma,
					unsigned long address,
					pmd_t *pmd, pmd_t orig_pmd,
					struct page *page,
					unsigned long haddr)
{
	pgtable_t pgtable;
	pmd_t _pmd;
	int ret = 0, i;
	struct page **pages;
1038 1039
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
1040 1041 1042 1043 1044 1045 1046 1047 1048

	pages = kmalloc(sizeof(struct page *) * HPAGE_PMD_NR,
			GFP_KERNEL);
	if (unlikely(!pages)) {
		ret |= VM_FAULT_OOM;
		goto out;
	}

	for (i = 0; i < HPAGE_PMD_NR; i++) {
1049 1050
		pages[i] = alloc_page_vma_node(GFP_HIGHUSER_MOVABLE |
					       __GFP_OTHER_NODE,
1051
					       vma, address, page_to_nid(page));
A
Andrea Arcangeli 已提交
1052 1053 1054 1055
		if (unlikely(!pages[i] ||
			     mem_cgroup_newpage_charge(pages[i], mm,
						       GFP_KERNEL))) {
			if (pages[i])
1056
				put_page(pages[i]);
A
Andrea Arcangeli 已提交
1057 1058 1059 1060 1061 1062
			mem_cgroup_uncharge_start();
			while (--i >= 0) {
				mem_cgroup_uncharge_page(pages[i]);
				put_page(pages[i]);
			}
			mem_cgroup_uncharge_end();
1063 1064 1065 1066 1067 1068 1069 1070
			kfree(pages);
			ret |= VM_FAULT_OOM;
			goto out;
		}
	}

	for (i = 0; i < HPAGE_PMD_NR; i++) {
		copy_user_highpage(pages[i], page + i,
1071
				   haddr + PAGE_SIZE * i, vma);
1072 1073 1074 1075
		__SetPageUptodate(pages[i]);
		cond_resched();
	}

1076 1077 1078 1079
	mmun_start = haddr;
	mmun_end   = haddr + HPAGE_PMD_SIZE;
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);

1080 1081 1082 1083 1084
	spin_lock(&mm->page_table_lock);
	if (unlikely(!pmd_same(*pmd, orig_pmd)))
		goto out_free_pages;
	VM_BUG_ON(!PageHead(page));

1085
	pmdp_clear_flush(vma, haddr, pmd);
1086 1087
	/* leave pmd empty until pte is filled */

1088
	pgtable = pgtable_trans_huge_withdraw(mm, pmd);
1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107
	pmd_populate(mm, &_pmd, pgtable);

	for (i = 0; i < HPAGE_PMD_NR; i++, haddr += PAGE_SIZE) {
		pte_t *pte, entry;
		entry = mk_pte(pages[i], vma->vm_page_prot);
		entry = maybe_mkwrite(pte_mkdirty(entry), vma);
		page_add_new_anon_rmap(pages[i], vma, haddr);
		pte = pte_offset_map(&_pmd, haddr);
		VM_BUG_ON(!pte_none(*pte));
		set_pte_at(mm, haddr, pte, entry);
		pte_unmap(pte);
	}
	kfree(pages);

	smp_wmb(); /* make pte visible before pmd */
	pmd_populate(mm, pmd, pgtable);
	page_remove_rmap(page);
	spin_unlock(&mm->page_table_lock);

1108 1109
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);

1110 1111 1112 1113 1114 1115 1116 1117
	ret |= VM_FAULT_WRITE;
	put_page(page);

out:
	return ret;

out_free_pages:
	spin_unlock(&mm->page_table_lock);
1118
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
A
Andrea Arcangeli 已提交
1119 1120 1121
	mem_cgroup_uncharge_start();
	for (i = 0; i < HPAGE_PMD_NR; i++) {
		mem_cgroup_uncharge_page(pages[i]);
1122
		put_page(pages[i]);
A
Andrea Arcangeli 已提交
1123 1124
	}
	mem_cgroup_uncharge_end();
1125 1126 1127 1128 1129 1130 1131 1132
	kfree(pages);
	goto out;
}

int do_huge_pmd_wp_page(struct mm_struct *mm, struct vm_area_struct *vma,
			unsigned long address, pmd_t *pmd, pmd_t orig_pmd)
{
	int ret = 0;
1133
	struct page *page = NULL, *new_page;
1134
	unsigned long haddr;
1135 1136
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
1137 1138

	VM_BUG_ON(!vma->anon_vma);
1139 1140 1141
	haddr = address & HPAGE_PMD_MASK;
	if (is_huge_zero_pmd(orig_pmd))
		goto alloc;
1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152
	spin_lock(&mm->page_table_lock);
	if (unlikely(!pmd_same(*pmd, orig_pmd)))
		goto out_unlock;

	page = pmd_page(orig_pmd);
	VM_BUG_ON(!PageCompound(page) || !PageHead(page));
	if (page_mapcount(page) == 1) {
		pmd_t entry;
		entry = pmd_mkyoung(orig_pmd);
		entry = maybe_pmd_mkwrite(pmd_mkdirty(entry), vma);
		if (pmdp_set_access_flags(vma, haddr, pmd, entry,  1))
1153
			update_mmu_cache_pmd(vma, address, pmd);
1154 1155 1156 1157 1158
		ret |= VM_FAULT_WRITE;
		goto out_unlock;
	}
	get_page(page);
	spin_unlock(&mm->page_table_lock);
1159
alloc:
1160 1161
	if (transparent_hugepage_enabled(vma) &&
	    !transparent_hugepage_debug_cow())
1162
		new_page = alloc_hugepage_vma(transparent_hugepage_defrag(vma),
1163
					      vma, haddr, numa_node_id(), 0);
1164 1165 1166 1167
	else
		new_page = NULL;

	if (unlikely(!new_page)) {
1168
		count_vm_event(THP_FAULT_FALLBACK);
1169 1170
		if (is_huge_zero_pmd(orig_pmd)) {
			ret = do_huge_pmd_wp_zero_page_fallback(mm, vma,
1171
					address, pmd, orig_pmd, haddr);
1172 1173 1174 1175 1176 1177 1178
		} else {
			ret = do_huge_pmd_wp_page_fallback(mm, vma, address,
					pmd, orig_pmd, page, haddr);
			if (ret & VM_FAULT_OOM)
				split_huge_page(page);
			put_page(page);
		}
1179 1180
		goto out;
	}
1181
	count_vm_event(THP_FAULT_ALLOC);
1182

A
Andrea Arcangeli 已提交
1183 1184
	if (unlikely(mem_cgroup_newpage_charge(new_page, mm, GFP_KERNEL))) {
		put_page(new_page);
1185 1186 1187 1188
		if (page) {
			split_huge_page(page);
			put_page(page);
		}
A
Andrea Arcangeli 已提交
1189 1190 1191 1192
		ret |= VM_FAULT_OOM;
		goto out;
	}

1193 1194 1195 1196
	if (is_huge_zero_pmd(orig_pmd))
		clear_huge_page(new_page, haddr, HPAGE_PMD_NR);
	else
		copy_user_huge_page(new_page, page, haddr, vma, HPAGE_PMD_NR);
1197 1198
	__SetPageUptodate(new_page);

1199 1200 1201 1202
	mmun_start = haddr;
	mmun_end   = haddr + HPAGE_PMD_SIZE;
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);

1203
	spin_lock(&mm->page_table_lock);
1204 1205
	if (page)
		put_page(page);
A
Andrea Arcangeli 已提交
1206
	if (unlikely(!pmd_same(*pmd, orig_pmd))) {
1207
		spin_unlock(&mm->page_table_lock);
A
Andrea Arcangeli 已提交
1208
		mem_cgroup_uncharge_page(new_page);
1209
		put_page(new_page);
1210
		goto out_mn;
A
Andrea Arcangeli 已提交
1211
	} else {
1212
		pmd_t entry;
B
Bob Liu 已提交
1213
		entry = mk_huge_pmd(new_page, vma);
1214
		pmdp_clear_flush(vma, haddr, pmd);
1215 1216
		page_add_new_anon_rmap(new_page, vma, haddr);
		set_pmd_at(mm, haddr, pmd, entry);
1217
		update_mmu_cache_pmd(vma, address, pmd);
1218
		if (is_huge_zero_pmd(orig_pmd)) {
1219
			add_mm_counter(mm, MM_ANONPAGES, HPAGE_PMD_NR);
1220 1221
			put_huge_zero_page();
		} else {
1222 1223 1224 1225
			VM_BUG_ON(!PageHead(page));
			page_remove_rmap(page);
			put_page(page);
		}
1226 1227 1228
		ret |= VM_FAULT_WRITE;
	}
	spin_unlock(&mm->page_table_lock);
1229 1230
out_mn:
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
1231 1232
out:
	return ret;
1233 1234 1235
out_unlock:
	spin_unlock(&mm->page_table_lock);
	return ret;
1236 1237
}

1238
struct page *follow_trans_huge_pmd(struct vm_area_struct *vma,
1239 1240 1241 1242
				   unsigned long addr,
				   pmd_t *pmd,
				   unsigned int flags)
{
1243
	struct mm_struct *mm = vma->vm_mm;
1244 1245 1246 1247 1248 1249 1250
	struct page *page = NULL;

	assert_spin_locked(&mm->page_table_lock);

	if (flags & FOLL_WRITE && !pmd_write(*pmd))
		goto out;

1251 1252 1253 1254
	/* Avoid dumping huge zero page */
	if ((flags & FOLL_DUMP) && is_huge_zero_pmd(*pmd))
		return ERR_PTR(-EFAULT);

1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267
	page = pmd_page(*pmd);
	VM_BUG_ON(!PageHead(page));
	if (flags & FOLL_TOUCH) {
		pmd_t _pmd;
		/*
		 * We should set the dirty bit only for FOLL_WRITE but
		 * for now the dirty bit in the pmd is meaningless.
		 * And if the dirty bit will become meaningful and
		 * we'll only set it with FOLL_WRITE, an atomic
		 * set_bit will be required on the pmd to set the
		 * young bit, instead of the current set_pmd_at.
		 */
		_pmd = pmd_mkyoung(pmd_mkdirty(*pmd));
1268 1269 1270
		if (pmdp_set_access_flags(vma, addr & HPAGE_PMD_MASK,
					  pmd, _pmd,  1))
			update_mmu_cache_pmd(vma, addr, pmd);
1271
	}
1272 1273 1274 1275 1276 1277 1278 1279
	if ((flags & FOLL_MLOCK) && (vma->vm_flags & VM_LOCKED)) {
		if (page->mapping && trylock_page(page)) {
			lru_add_drain();
			if (page->mapping)
				mlock_vma_page(page);
			unlock_page(page);
		}
	}
1280 1281 1282
	page += (addr & ~HPAGE_PMD_MASK) >> PAGE_SHIFT;
	VM_BUG_ON(!PageCompound(page));
	if (flags & FOLL_GET)
1283
		get_page_foll(page);
1284 1285 1286 1287 1288

out:
	return page;
}

1289
/* NUMA hinting page fault entry point for trans huge pmds */
1290 1291
int do_huge_pmd_numa_page(struct mm_struct *mm, struct vm_area_struct *vma,
				unsigned long addr, pmd_t pmd, pmd_t *pmdp)
1292
{
1293
	struct page *page;
1294
	unsigned long haddr = addr & HPAGE_PMD_MASK;
1295
	int target_nid;
1296
	int current_nid = -1;
1297
	bool migrated;
1298 1299 1300 1301 1302 1303

	spin_lock(&mm->page_table_lock);
	if (unlikely(!pmd_same(pmd, *pmdp)))
		goto out_unlock;

	page = pmd_page(pmd);
1304
	get_page(page);
1305 1306 1307 1308
	current_nid = page_to_nid(page);
	count_vm_numa_event(NUMA_HINT_FAULTS);
	if (current_nid == numa_node_id())
		count_vm_numa_event(NUMA_HINT_FAULTS_LOCAL);
1309 1310

	target_nid = mpol_misplaced(page, vma, haddr);
1311 1312
	if (target_nid == -1) {
		put_page(page);
1313
		goto clear_pmdnuma;
1314
	}
1315

1316 1317 1318
	/* Acquire the page lock to serialise THP migrations */
	spin_unlock(&mm->page_table_lock);
	lock_page(page);
1319

1320
	/* Confirm the PTE did not while locked */
1321
	spin_lock(&mm->page_table_lock);
1322 1323 1324
	if (unlikely(!pmd_same(pmd, *pmdp))) {
		unlock_page(page);
		put_page(page);
1325
		goto out_unlock;
1326 1327
	}
	spin_unlock(&mm->page_table_lock);
1328

1329 1330
	/* Migrate the THP to the requested node */
	migrated = migrate_misplaced_transhuge_page(mm, vma,
1331 1332 1333
				pmdp, pmd, addr, page, target_nid);
	if (!migrated)
		goto check_same;
1334

1335
	task_numa_fault(target_nid, HPAGE_PMD_NR, true);
1336 1337
	return 0;

1338 1339 1340 1341
check_same:
	spin_lock(&mm->page_table_lock);
	if (unlikely(!pmd_same(pmd, *pmdp)))
		goto out_unlock;
1342
clear_pmdnuma:
1343 1344 1345 1346 1347 1348
	pmd = pmd_mknonnuma(pmd);
	set_pmd_at(mm, haddr, pmdp, pmd);
	VM_BUG_ON(pmd_numa(*pmdp));
	update_mmu_cache_pmd(vma, addr, pmdp);
out_unlock:
	spin_unlock(&mm->page_table_lock);
1349
	if (current_nid != -1)
1350
		task_numa_fault(current_nid, HPAGE_PMD_NR, false);
1351 1352 1353
	return 0;
}

1354
int zap_huge_pmd(struct mmu_gather *tlb, struct vm_area_struct *vma,
S
Shaohua Li 已提交
1355
		 pmd_t *pmd, unsigned long addr)
1356 1357 1358
{
	int ret = 0;

1359 1360 1361
	if (__pmd_trans_huge_lock(pmd, vma) == 1) {
		struct page *page;
		pgtable_t pgtable;
1362
		pmd_t orig_pmd;
1363 1364 1365 1366 1367 1368
		/*
		 * For architectures like ppc64 we look at deposited pgtable
		 * when calling pmdp_get_and_clear. So do the
		 * pgtable_trans_huge_withdraw after finishing pmdp related
		 * operations.
		 */
1369
		orig_pmd = pmdp_get_and_clear(tlb->mm, addr, pmd);
1370
		tlb_remove_pmd_tlb_entry(tlb, pmd, addr);
1371
		pgtable = pgtable_trans_huge_withdraw(tlb->mm, pmd);
1372 1373 1374
		if (is_huge_zero_pmd(orig_pmd)) {
			tlb->mm->nr_ptes--;
			spin_unlock(&tlb->mm->page_table_lock);
1375
			put_huge_zero_page();
1376 1377 1378 1379 1380 1381 1382 1383 1384 1385
		} else {
			page = pmd_page(orig_pmd);
			page_remove_rmap(page);
			VM_BUG_ON(page_mapcount(page) < 0);
			add_mm_counter(tlb->mm, MM_ANONPAGES, -HPAGE_PMD_NR);
			VM_BUG_ON(!PageHead(page));
			tlb->mm->nr_ptes--;
			spin_unlock(&tlb->mm->page_table_lock);
			tlb_remove_page(tlb, page);
		}
1386 1387 1388
		pte_free(tlb->mm, pgtable);
		ret = 1;
	}
1389 1390 1391
	return ret;
}

1392 1393 1394 1395 1396 1397
int mincore_huge_pmd(struct vm_area_struct *vma, pmd_t *pmd,
		unsigned long addr, unsigned long end,
		unsigned char *vec)
{
	int ret = 0;

1398 1399 1400 1401 1402
	if (__pmd_trans_huge_lock(pmd, vma) == 1) {
		/*
		 * All logical pages in the range are present
		 * if backed by a huge page.
		 */
1403
		spin_unlock(&vma->vm_mm->page_table_lock);
1404 1405 1406
		memset(vec, 1, (end - addr) >> PAGE_SHIFT);
		ret = 1;
	}
1407 1408 1409 1410

	return ret;
}

1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435
int move_huge_pmd(struct vm_area_struct *vma, struct vm_area_struct *new_vma,
		  unsigned long old_addr,
		  unsigned long new_addr, unsigned long old_end,
		  pmd_t *old_pmd, pmd_t *new_pmd)
{
	int ret = 0;
	pmd_t pmd;

	struct mm_struct *mm = vma->vm_mm;

	if ((old_addr & ~HPAGE_PMD_MASK) ||
	    (new_addr & ~HPAGE_PMD_MASK) ||
	    old_end - old_addr < HPAGE_PMD_SIZE ||
	    (new_vma->vm_flags & VM_NOHUGEPAGE))
		goto out;

	/*
	 * The destination pmd shouldn't be established, free_pgtables()
	 * should have release it.
	 */
	if (WARN_ON(!pmd_none(*new_pmd))) {
		VM_BUG_ON(pmd_trans_huge(*new_pmd));
		goto out;
	}

1436 1437 1438 1439 1440
	ret = __pmd_trans_huge_lock(old_pmd, vma);
	if (ret == 1) {
		pmd = pmdp_get_and_clear(mm, old_addr, old_pmd);
		VM_BUG_ON(!pmd_none(*new_pmd));
		set_pmd_at(mm, new_addr, new_pmd, pmd);
1441 1442 1443 1444 1445 1446
		spin_unlock(&mm->page_table_lock);
	}
out:
	return ret;
}

1447
int change_huge_pmd(struct vm_area_struct *vma, pmd_t *pmd,
1448
		unsigned long addr, pgprot_t newprot, int prot_numa)
1449 1450 1451 1452
{
	struct mm_struct *mm = vma->vm_mm;
	int ret = 0;

1453 1454 1455
	if (__pmd_trans_huge_lock(pmd, vma) == 1) {
		pmd_t entry;
		entry = pmdp_get_and_clear(mm, addr, pmd);
1456
		if (!prot_numa) {
1457
			entry = pmd_modify(entry, newprot);
1458 1459
			BUG_ON(pmd_write(entry));
		} else {
1460 1461 1462 1463 1464 1465 1466 1467
			struct page *page = pmd_page(*pmd);

			/* only check non-shared pages */
			if (page_mapcount(page) == 1 &&
			    !pmd_numa(*pmd)) {
				entry = pmd_mknuma(entry);
			}
		}
1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485
		set_pmd_at(mm, addr, pmd, entry);
		spin_unlock(&vma->vm_mm->page_table_lock);
		ret = 1;
	}

	return ret;
}

/*
 * Returns 1 if a given pmd maps a stable (not under splitting) thp.
 * Returns -1 if it maps a thp under splitting. Returns 0 otherwise.
 *
 * Note that if it returns 1, this routine returns without unlocking page
 * table locks. So callers must unlock them.
 */
int __pmd_trans_huge_lock(pmd_t *pmd, struct vm_area_struct *vma)
{
	spin_lock(&vma->vm_mm->page_table_lock);
1486 1487
	if (likely(pmd_trans_huge(*pmd))) {
		if (unlikely(pmd_trans_splitting(*pmd))) {
1488
			spin_unlock(&vma->vm_mm->page_table_lock);
1489
			wait_split_huge_page(vma->anon_vma, pmd);
1490
			return -1;
1491
		} else {
1492 1493 1494
			/* Thp mapped by 'pmd' is stable, so we can
			 * handle it as it is. */
			return 1;
1495
		}
1496 1497 1498
	}
	spin_unlock(&vma->vm_mm->page_table_lock);
	return 0;
1499 1500
}

1501 1502 1503 1504 1505 1506 1507 1508 1509 1510
pmd_t *page_check_address_pmd(struct page *page,
			      struct mm_struct *mm,
			      unsigned long address,
			      enum page_check_address_pmd_flag flag)
{
	pmd_t *pmd, *ret = NULL;

	if (address & ~HPAGE_PMD_MASK)
		goto out;

B
Bob Liu 已提交
1511 1512
	pmd = mm_find_pmd(mm, address);
	if (!pmd)
1513 1514 1515 1516 1517
		goto out;
	if (pmd_none(*pmd))
		goto out;
	if (pmd_page(*pmd) != page)
		goto out;
1518 1519 1520 1521 1522 1523 1524 1525 1526 1527
	/*
	 * split_vma() may create temporary aliased mappings. There is
	 * no risk as long as all huge pmd are found and have their
	 * splitting bit set before __split_huge_page_refcount
	 * runs. Finding the same huge pmd more than once during the
	 * same rmap walk is not a problem.
	 */
	if (flag == PAGE_CHECK_ADDRESS_PMD_NOTSPLITTING_FLAG &&
	    pmd_trans_splitting(*pmd))
		goto out;
1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543
	if (pmd_trans_huge(*pmd)) {
		VM_BUG_ON(flag == PAGE_CHECK_ADDRESS_PMD_SPLITTING_FLAG &&
			  !pmd_trans_splitting(*pmd));
		ret = pmd;
	}
out:
	return ret;
}

static int __split_huge_page_splitting(struct page *page,
				       struct vm_area_struct *vma,
				       unsigned long address)
{
	struct mm_struct *mm = vma->vm_mm;
	pmd_t *pmd;
	int ret = 0;
1544 1545 1546
	/* For mmu_notifiers */
	const unsigned long mmun_start = address;
	const unsigned long mmun_end   = address + HPAGE_PMD_SIZE;
1547

1548
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);
1549 1550 1551 1552 1553 1554 1555 1556
	spin_lock(&mm->page_table_lock);
	pmd = page_check_address_pmd(page, mm, address,
				     PAGE_CHECK_ADDRESS_PMD_NOTSPLITTING_FLAG);
	if (pmd) {
		/*
		 * We can't temporarily set the pmd to null in order
		 * to split it, the pmd must remain marked huge at all
		 * times or the VM won't take the pmd_trans_huge paths
1557
		 * and it won't wait on the anon_vma->root->rwsem to
1558 1559
		 * serialize against split_huge_page*.
		 */
1560
		pmdp_splitting_flush(vma, address, pmd);
1561 1562 1563
		ret = 1;
	}
	spin_unlock(&mm->page_table_lock);
1564
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
1565 1566 1567 1568

	return ret;
}

1569 1570
static void __split_huge_page_refcount(struct page *page,
				       struct list_head *list)
1571 1572 1573
{
	int i;
	struct zone *zone = page_zone(page);
1574
	struct lruvec *lruvec;
1575
	int tail_count = 0;
1576 1577 1578

	/* prevent PageLRU to go away from under us, and freeze lru stats */
	spin_lock_irq(&zone->lru_lock);
1579 1580
	lruvec = mem_cgroup_page_lruvec(page, zone);

1581
	compound_lock(page);
1582 1583
	/* complete memcg works before add pages to LRU */
	mem_cgroup_split_huge_fixup(page);
1584

1585
	for (i = HPAGE_PMD_NR - 1; i >= 1; i--) {
1586 1587
		struct page *page_tail = page + i;

1588 1589 1590 1591 1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605 1606 1607 1608
		/* tail_page->_mapcount cannot change */
		BUG_ON(page_mapcount(page_tail) < 0);
		tail_count += page_mapcount(page_tail);
		/* check for overflow */
		BUG_ON(tail_count < 0);
		BUG_ON(atomic_read(&page_tail->_count) != 0);
		/*
		 * tail_page->_count is zero and not changing from
		 * under us. But get_page_unless_zero() may be running
		 * from under us on the tail_page. If we used
		 * atomic_set() below instead of atomic_add(), we
		 * would then run atomic_set() concurrently with
		 * get_page_unless_zero(), and atomic_set() is
		 * implemented in C not using locked ops. spin_unlock
		 * on x86 sometime uses locked ops because of PPro
		 * errata 66, 92, so unless somebody can guarantee
		 * atomic_set() here would be safe on all archs (and
		 * not only on x86), it's safer to use atomic_add().
		 */
		atomic_add(page_mapcount(page) + page_mapcount(page_tail) + 1,
			   &page_tail->_count);
1609 1610 1611 1612

		/* after clearing PageTail the gup refcount can be released */
		smp_mb();

1613 1614 1615 1616 1617 1618
		/*
		 * retain hwpoison flag of the poisoned tail page:
		 *   fix for the unsuitable process killed on Guest Machine(KVM)
		 *   by the memory-failure.
		 */
		page_tail->flags &= ~PAGE_FLAGS_CHECK_AT_PREP | __PG_HWPOISON;
1619 1620 1621 1622 1623 1624 1625
		page_tail->flags |= (page->flags &
				     ((1L << PG_referenced) |
				      (1L << PG_swapbacked) |
				      (1L << PG_mlocked) |
				      (1L << PG_uptodate)));
		page_tail->flags |= (1L << PG_dirty);

1626
		/* clear PageTail before overwriting first_page */
1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647
		smp_wmb();

		/*
		 * __split_huge_page_splitting() already set the
		 * splitting bit in all pmd that could map this
		 * hugepage, that will ensure no CPU can alter the
		 * mapcount on the head page. The mapcount is only
		 * accounted in the head page and it has to be
		 * transferred to all tail pages in the below code. So
		 * for this code to be safe, the split the mapcount
		 * can't change. But that doesn't mean userland can't
		 * keep changing and reading the page contents while
		 * we transfer the mapcount, so the pmd splitting
		 * status is achieved setting a reserved bit in the
		 * pmd, not by clearing the present bit.
		*/
		page_tail->_mapcount = page->_mapcount;

		BUG_ON(page_tail->mapping);
		page_tail->mapping = page->mapping;

1648
		page_tail->index = page->index + i;
1649
		page_nid_xchg_last(page_tail, page_nid_last(page));
1650 1651 1652 1653 1654 1655

		BUG_ON(!PageAnon(page_tail));
		BUG_ON(!PageUptodate(page_tail));
		BUG_ON(!PageDirty(page_tail));
		BUG_ON(!PageSwapBacked(page_tail));

1656
		lru_add_page_tail(page, page_tail, lruvec, list);
1657
	}
1658 1659
	atomic_sub(tail_count, &page->_count);
	BUG_ON(atomic_read(&page->_count) <= 0);
1660

1661
	__mod_zone_page_state(zone, NR_ANON_TRANSPARENT_HUGEPAGES, -1);
1662 1663
	__mod_zone_page_state(zone, NR_ANON_PAGES, HPAGE_PMD_NR);

1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701
	ClearPageCompound(page);
	compound_unlock(page);
	spin_unlock_irq(&zone->lru_lock);

	for (i = 1; i < HPAGE_PMD_NR; i++) {
		struct page *page_tail = page + i;
		BUG_ON(page_count(page_tail) <= 0);
		/*
		 * Tail pages may be freed if there wasn't any mapping
		 * like if add_to_swap() is running on a lru page that
		 * had its mapping zapped. And freeing these pages
		 * requires taking the lru_lock so we do the put_page
		 * of the tail pages after the split is complete.
		 */
		put_page(page_tail);
	}

	/*
	 * Only the head page (now become a regular page) is required
	 * to be pinned by the caller.
	 */
	BUG_ON(page_count(page) <= 0);
}

static int __split_huge_page_map(struct page *page,
				 struct vm_area_struct *vma,
				 unsigned long address)
{
	struct mm_struct *mm = vma->vm_mm;
	pmd_t *pmd, _pmd;
	int ret = 0, i;
	pgtable_t pgtable;
	unsigned long haddr;

	spin_lock(&mm->page_table_lock);
	pmd = page_check_address_pmd(page, mm, address,
				     PAGE_CHECK_ADDRESS_PMD_SPLITTING_FLAG);
	if (pmd) {
1702
		pgtable = pgtable_trans_huge_withdraw(mm, pmd);
1703 1704
		pmd_populate(mm, &_pmd, pgtable);

1705 1706
		haddr = address;
		for (i = 0; i < HPAGE_PMD_NR; i++, haddr += PAGE_SIZE) {
1707 1708 1709 1710 1711 1712 1713 1714 1715 1716
			pte_t *pte, entry;
			BUG_ON(PageCompound(page+i));
			entry = mk_pte(page + i, vma->vm_page_prot);
			entry = maybe_mkwrite(pte_mkdirty(entry), vma);
			if (!pmd_write(*pmd))
				entry = pte_wrprotect(entry);
			else
				BUG_ON(page_mapcount(page) != 1);
			if (!pmd_young(*pmd))
				entry = pte_mkold(entry);
1717 1718
			if (pmd_numa(*pmd))
				entry = pte_mknuma(entry);
1719 1720 1721 1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751
			pte = pte_offset_map(&_pmd, haddr);
			BUG_ON(!pte_none(*pte));
			set_pte_at(mm, haddr, pte, entry);
			pte_unmap(pte);
		}

		smp_wmb(); /* make pte visible before pmd */
		/*
		 * Up to this point the pmd is present and huge and
		 * userland has the whole access to the hugepage
		 * during the split (which happens in place). If we
		 * overwrite the pmd with the not-huge version
		 * pointing to the pte here (which of course we could
		 * if all CPUs were bug free), userland could trigger
		 * a small page size TLB miss on the small sized TLB
		 * while the hugepage TLB entry is still established
		 * in the huge TLB. Some CPU doesn't like that. See
		 * http://support.amd.com/us/Processor_TechDocs/41322.pdf,
		 * Erratum 383 on page 93. Intel should be safe but is
		 * also warns that it's only safe if the permission
		 * and cache attributes of the two entries loaded in
		 * the two TLB is identical (which should be the case
		 * here). But it is generally safer to never allow
		 * small and huge TLB entries for the same virtual
		 * address to be loaded simultaneously. So instead of
		 * doing "pmd_populate(); flush_tlb_range();" we first
		 * mark the current pmd notpresent (atomically because
		 * here the pmd_trans_huge and pmd_trans_splitting
		 * must remain set at all times on the pmd until the
		 * split is complete for this pmd), then we flush the
		 * SMP TLB and finally we write the non-huge version
		 * of the pmd entry with pmd_populate.
		 */
G
Gerald Schaefer 已提交
1752
		pmdp_invalidate(vma, address, pmd);
1753 1754 1755 1756 1757 1758 1759 1760
		pmd_populate(mm, pmd, pgtable);
		ret = 1;
	}
	spin_unlock(&mm->page_table_lock);

	return ret;
}

1761
/* must be called with anon_vma->root->rwsem held */
1762
static void __split_huge_page(struct page *page,
1763 1764
			      struct anon_vma *anon_vma,
			      struct list_head *list)
1765 1766
{
	int mapcount, mapcount2;
1767
	pgoff_t pgoff = page->index << (PAGE_CACHE_SHIFT - PAGE_SHIFT);
1768 1769 1770 1771 1772 1773
	struct anon_vma_chain *avc;

	BUG_ON(!PageHead(page));
	BUG_ON(PageTail(page));

	mapcount = 0;
1774
	anon_vma_interval_tree_foreach(avc, &anon_vma->rb_root, pgoff, pgoff) {
1775 1776 1777 1778 1779
		struct vm_area_struct *vma = avc->vma;
		unsigned long addr = vma_address(page, vma);
		BUG_ON(is_vma_temporary_stack(vma));
		mapcount += __split_huge_page_splitting(page, vma, addr);
	}
1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792
	/*
	 * It is critical that new vmas are added to the tail of the
	 * anon_vma list. This guarantes that if copy_huge_pmd() runs
	 * and establishes a child pmd before
	 * __split_huge_page_splitting() freezes the parent pmd (so if
	 * we fail to prevent copy_huge_pmd() from running until the
	 * whole __split_huge_page() is complete), we will still see
	 * the newly established pmd of the child later during the
	 * walk, to be able to set it as pmd_trans_splitting too.
	 */
	if (mapcount != page_mapcount(page))
		printk(KERN_ERR "mapcount %d page_mapcount %d\n",
		       mapcount, page_mapcount(page));
1793 1794
	BUG_ON(mapcount != page_mapcount(page));

1795
	__split_huge_page_refcount(page, list);
1796 1797

	mapcount2 = 0;
1798
	anon_vma_interval_tree_foreach(avc, &anon_vma->rb_root, pgoff, pgoff) {
1799 1800 1801 1802 1803
		struct vm_area_struct *vma = avc->vma;
		unsigned long addr = vma_address(page, vma);
		BUG_ON(is_vma_temporary_stack(vma));
		mapcount2 += __split_huge_page_map(page, vma, addr);
	}
1804 1805 1806
	if (mapcount != mapcount2)
		printk(KERN_ERR "mapcount %d mapcount2 %d page_mapcount %d\n",
		       mapcount, mapcount2, page_mapcount(page));
1807 1808 1809
	BUG_ON(mapcount != mapcount2);
}

1810 1811 1812 1813 1814 1815 1816 1817
/*
 * Split a hugepage into normal pages. This doesn't change the position of head
 * page. If @list is null, tail pages will be added to LRU list, otherwise, to
 * @list. Both head page and tail pages will inherit mapping, flags, and so on
 * from the hugepage.
 * Return 0 if the hugepage is split successfully otherwise return 1.
 */
int split_huge_page_to_list(struct page *page, struct list_head *list)
1818 1819 1820 1821
{
	struct anon_vma *anon_vma;
	int ret = 1;

1822
	BUG_ON(is_huge_zero_page(page));
1823
	BUG_ON(!PageAnon(page));
1824 1825 1826 1827 1828 1829 1830 1831 1832

	/*
	 * The caller does not necessarily hold an mmap_sem that would prevent
	 * the anon_vma disappearing so we first we take a reference to it
	 * and then lock the anon_vma for write. This is similar to
	 * page_lock_anon_vma_read except the write lock is taken to serialise
	 * against parallel split or collapse operations.
	 */
	anon_vma = page_get_anon_vma(page);
1833 1834
	if (!anon_vma)
		goto out;
1835 1836
	anon_vma_lock_write(anon_vma);

1837 1838 1839 1840 1841
	ret = 0;
	if (!PageCompound(page))
		goto out_unlock;

	BUG_ON(!PageSwapBacked(page));
1842
	__split_huge_page(page, anon_vma, list);
1843
	count_vm_event(THP_SPLIT);
1844 1845 1846

	BUG_ON(PageCompound(page));
out_unlock:
1847
	anon_vma_unlock_write(anon_vma);
1848
	put_anon_vma(anon_vma);
1849 1850 1851 1852
out:
	return ret;
}

1853
#define VM_NO_THP (VM_SPECIAL|VM_MIXEDMAP|VM_HUGETLB|VM_SHARED|VM_MAYSHARE)
1854

1855 1856
int hugepage_madvise(struct vm_area_struct *vma,
		     unsigned long *vm_flags, int advice)
A
Andrea Arcangeli 已提交
1857
{
1858 1859
	struct mm_struct *mm = vma->vm_mm;

A
Andrea Arcangeli 已提交
1860 1861 1862 1863 1864
	switch (advice) {
	case MADV_HUGEPAGE:
		/*
		 * Be somewhat over-protective like KSM for now!
		 */
1865
		if (*vm_flags & (VM_HUGEPAGE | VM_NO_THP))
A
Andrea Arcangeli 已提交
1866
			return -EINVAL;
1867 1868
		if (mm->def_flags & VM_NOHUGEPAGE)
			return -EINVAL;
A
Andrea Arcangeli 已提交
1869 1870
		*vm_flags &= ~VM_NOHUGEPAGE;
		*vm_flags |= VM_HUGEPAGE;
1871 1872 1873 1874 1875 1876 1877
		/*
		 * If the vma become good for khugepaged to scan,
		 * register it here without waiting a page fault that
		 * may not happen any time soon.
		 */
		if (unlikely(khugepaged_enter_vma_merge(vma)))
			return -ENOMEM;
A
Andrea Arcangeli 已提交
1878 1879 1880 1881 1882
		break;
	case MADV_NOHUGEPAGE:
		/*
		 * Be somewhat over-protective like KSM for now!
		 */
1883
		if (*vm_flags & (VM_NOHUGEPAGE | VM_NO_THP))
A
Andrea Arcangeli 已提交
1884 1885 1886
			return -EINVAL;
		*vm_flags &= ~VM_HUGEPAGE;
		*vm_flags |= VM_NOHUGEPAGE;
1887 1888 1889 1890 1891
		/*
		 * Setting VM_NOHUGEPAGE will prevent khugepaged from scanning
		 * this vma even if we leave the mm registered in khugepaged if
		 * it got registered before VM_NOHUGEPAGE was set.
		 */
A
Andrea Arcangeli 已提交
1892 1893
		break;
	}
A
Andrea Arcangeli 已提交
1894 1895 1896 1897

	return 0;
}

A
Andrea Arcangeli 已提交
1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924
static int __init khugepaged_slab_init(void)
{
	mm_slot_cache = kmem_cache_create("khugepaged_mm_slot",
					  sizeof(struct mm_slot),
					  __alignof__(struct mm_slot), 0, NULL);
	if (!mm_slot_cache)
		return -ENOMEM;

	return 0;
}

static inline struct mm_slot *alloc_mm_slot(void)
{
	if (!mm_slot_cache)	/* initialization failed */
		return NULL;
	return kmem_cache_zalloc(mm_slot_cache, GFP_KERNEL);
}

static inline void free_mm_slot(struct mm_slot *mm_slot)
{
	kmem_cache_free(mm_slot_cache, mm_slot);
}

static struct mm_slot *get_mm_slot(struct mm_struct *mm)
{
	struct mm_slot *mm_slot;

1925
	hash_for_each_possible(mm_slots_hash, mm_slot, hash, (unsigned long)mm)
A
Andrea Arcangeli 已提交
1926 1927
		if (mm == mm_slot->mm)
			return mm_slot;
1928

A
Andrea Arcangeli 已提交
1929 1930 1931 1932 1933 1934 1935
	return NULL;
}

static void insert_to_mm_slots_hash(struct mm_struct *mm,
				    struct mm_slot *mm_slot)
{
	mm_slot->mm = mm;
1936
	hash_add(mm_slots_hash, &mm_slot->hash, (long)mm);
A
Andrea Arcangeli 已提交
1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985
}

static inline int khugepaged_test_exit(struct mm_struct *mm)
{
	return atomic_read(&mm->mm_users) == 0;
}

int __khugepaged_enter(struct mm_struct *mm)
{
	struct mm_slot *mm_slot;
	int wakeup;

	mm_slot = alloc_mm_slot();
	if (!mm_slot)
		return -ENOMEM;

	/* __khugepaged_exit() must not run from under us */
	VM_BUG_ON(khugepaged_test_exit(mm));
	if (unlikely(test_and_set_bit(MMF_VM_HUGEPAGE, &mm->flags))) {
		free_mm_slot(mm_slot);
		return 0;
	}

	spin_lock(&khugepaged_mm_lock);
	insert_to_mm_slots_hash(mm, mm_slot);
	/*
	 * Insert just behind the scanning cursor, to let the area settle
	 * down a little.
	 */
	wakeup = list_empty(&khugepaged_scan.mm_head);
	list_add_tail(&mm_slot->mm_node, &khugepaged_scan.mm_head);
	spin_unlock(&khugepaged_mm_lock);

	atomic_inc(&mm->mm_count);
	if (wakeup)
		wake_up_interruptible(&khugepaged_wait);

	return 0;
}

int khugepaged_enter_vma_merge(struct vm_area_struct *vma)
{
	unsigned long hstart, hend;
	if (!vma->anon_vma)
		/*
		 * Not yet faulted in so we will register later in the
		 * page fault if needed.
		 */
		return 0;
1986
	if (vma->vm_ops)
A
Andrea Arcangeli 已提交
1987 1988
		/* khugepaged not yet working on file or special mappings */
		return 0;
1989
	VM_BUG_ON(vma->vm_flags & VM_NO_THP);
A
Andrea Arcangeli 已提交
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
	hstart = (vma->vm_start + ~HPAGE_PMD_MASK) & HPAGE_PMD_MASK;
	hend = vma->vm_end & HPAGE_PMD_MASK;
	if (hstart < hend)
		return khugepaged_enter(vma);
	return 0;
}

void __khugepaged_exit(struct mm_struct *mm)
{
	struct mm_slot *mm_slot;
	int free = 0;

	spin_lock(&khugepaged_mm_lock);
	mm_slot = get_mm_slot(mm);
	if (mm_slot && khugepaged_scan.mm_slot != mm_slot) {
2005
		hash_del(&mm_slot->hash);
A
Andrea Arcangeli 已提交
2006 2007 2008
		list_del(&mm_slot->mm_node);
		free = 1;
	}
2009
	spin_unlock(&khugepaged_mm_lock);
A
Andrea Arcangeli 已提交
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025

	if (free) {
		clear_bit(MMF_VM_HUGEPAGE, &mm->flags);
		free_mm_slot(mm_slot);
		mmdrop(mm);
	} else if (mm_slot) {
		/*
		 * This is required to serialize against
		 * khugepaged_test_exit() (which is guaranteed to run
		 * under mmap sem read mode). Stop here (after we
		 * return all pagetables will be destroyed) until
		 * khugepaged has finished working on the pagetables
		 * under the mmap_sem.
		 */
		down_write(&mm->mmap_sem);
		up_write(&mm->mmap_sem);
2026
	}
A
Andrea Arcangeli 已提交
2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051
}

static void release_pte_page(struct page *page)
{
	/* 0 stands for page_is_file_cache(page) == false */
	dec_zone_page_state(page, NR_ISOLATED_ANON + 0);
	unlock_page(page);
	putback_lru_page(page);
}

static void release_pte_pages(pte_t *pte, pte_t *_pte)
{
	while (--_pte >= pte) {
		pte_t pteval = *_pte;
		if (!pte_none(pteval))
			release_pte_page(pte_page(pteval));
	}
}

static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
					unsigned long address,
					pte_t *pte)
{
	struct page *page;
	pte_t *_pte;
2052
	int referenced = 0, none = 0;
A
Andrea Arcangeli 已提交
2053 2054 2055 2056 2057 2058
	for (_pte = pte; _pte < pte+HPAGE_PMD_NR;
	     _pte++, address += PAGE_SIZE) {
		pte_t pteval = *_pte;
		if (pte_none(pteval)) {
			if (++none <= khugepaged_max_ptes_none)
				continue;
2059
			else
A
Andrea Arcangeli 已提交
2060 2061
				goto out;
		}
2062
		if (!pte_present(pteval) || !pte_write(pteval))
A
Andrea Arcangeli 已提交
2063 2064
			goto out;
		page = vm_normal_page(vma, address, pteval);
2065
		if (unlikely(!page))
A
Andrea Arcangeli 已提交
2066
			goto out;
2067

A
Andrea Arcangeli 已提交
2068 2069 2070 2071 2072
		VM_BUG_ON(PageCompound(page));
		BUG_ON(!PageAnon(page));
		VM_BUG_ON(!PageSwapBacked(page));

		/* cannot use mapcount: can't collapse if there's a gup pin */
2073
		if (page_count(page) != 1)
A
Andrea Arcangeli 已提交
2074 2075 2076 2077 2078 2079 2080
			goto out;
		/*
		 * We can do it before isolate_lru_page because the
		 * page can't be freed from under us. NOTE: PG_lock
		 * is needed to serialize against split_huge_page
		 * when invoked from the VM.
		 */
2081
		if (!trylock_page(page))
A
Andrea Arcangeli 已提交
2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096
			goto out;
		/*
		 * Isolate the page to avoid collapsing an hugepage
		 * currently in use by the VM.
		 */
		if (isolate_lru_page(page)) {
			unlock_page(page);
			goto out;
		}
		/* 0 stands for page_is_file_cache(page) == false */
		inc_zone_page_state(page, NR_ISOLATED_ANON + 0);
		VM_BUG_ON(!PageLocked(page));
		VM_BUG_ON(PageLRU(page));

		/* If there is no mapped pte young don't collapse the page */
A
Andrea Arcangeli 已提交
2097 2098
		if (pte_young(pteval) || PageReferenced(page) ||
		    mmu_notifier_test_young(vma->vm_mm, address))
A
Andrea Arcangeli 已提交
2099 2100
			referenced = 1;
	}
2101 2102
	if (likely(referenced))
		return 1;
A
Andrea Arcangeli 已提交
2103
out:
2104 2105
	release_pte_pages(pte, _pte);
	return 0;
A
Andrea Arcangeli 已提交
2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146
}

static void __collapse_huge_page_copy(pte_t *pte, struct page *page,
				      struct vm_area_struct *vma,
				      unsigned long address,
				      spinlock_t *ptl)
{
	pte_t *_pte;
	for (_pte = pte; _pte < pte+HPAGE_PMD_NR; _pte++) {
		pte_t pteval = *_pte;
		struct page *src_page;

		if (pte_none(pteval)) {
			clear_user_highpage(page, address);
			add_mm_counter(vma->vm_mm, MM_ANONPAGES, 1);
		} else {
			src_page = pte_page(pteval);
			copy_user_highpage(page, src_page, address, vma);
			VM_BUG_ON(page_mapcount(src_page) != 1);
			release_pte_page(src_page);
			/*
			 * ptl mostly unnecessary, but preempt has to
			 * be disabled to update the per-cpu stats
			 * inside page_remove_rmap().
			 */
			spin_lock(ptl);
			/*
			 * paravirt calls inside pte_clear here are
			 * superfluous.
			 */
			pte_clear(vma->vm_mm, address, _pte);
			page_remove_rmap(src_page);
			spin_unlock(ptl);
			free_page_and_swap_cache(src_page);
		}

		address += PAGE_SIZE;
		page++;
	}
}

2147
static void khugepaged_alloc_sleep(void)
A
Andrea Arcangeli 已提交
2148
{
2149 2150 2151
	wait_event_freezable_timeout(khugepaged_wait, false,
			msecs_to_jiffies(khugepaged_alloc_sleep_millisecs));
}
A
Andrea Arcangeli 已提交
2152

2153 2154 2155 2156 2157 2158 2159 2160
#ifdef CONFIG_NUMA
static bool khugepaged_prealloc_page(struct page **hpage, bool *wait)
{
	if (IS_ERR(*hpage)) {
		if (!*wait)
			return false;

		*wait = false;
2161
		*hpage = NULL;
2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175
		khugepaged_alloc_sleep();
	} else if (*hpage) {
		put_page(*hpage);
		*hpage = NULL;
	}

	return true;
}

static struct page
*khugepaged_alloc_page(struct page **hpage, struct mm_struct *mm,
		       struct vm_area_struct *vma, unsigned long address,
		       int node)
{
2176
	VM_BUG_ON(*hpage);
2177 2178 2179 2180 2181 2182 2183 2184 2185 2186
	/*
	 * Allocate the page while the vma is still valid and under
	 * the mmap_sem read mode so there is no memory allocation
	 * later when we take the mmap_sem in write mode. This is more
	 * friendly behavior (OTOH it may actually hide bugs) to
	 * filesystems in userland with daemons allocating memory in
	 * the userland I/O paths.  Allocating memory with the
	 * mmap_sem in read mode is good idea also to allow greater
	 * scalability.
	 */
2187
	*hpage  = alloc_hugepage_vma(khugepaged_defrag(), vma, address,
2188
				      node, __GFP_OTHER_NODE);
2189 2190 2191 2192 2193 2194

	/*
	 * After allocating the hugepage, release the mmap_sem read lock in
	 * preparation for taking it in write mode.
	 */
	up_read(&mm->mmap_sem);
2195
	if (unlikely(!*hpage)) {
2196
		count_vm_event(THP_COLLAPSE_ALLOC_FAILED);
2197
		*hpage = ERR_PTR(-ENOMEM);
2198
		return NULL;
2199
	}
2200

2201
	count_vm_event(THP_COLLAPSE_ALLOC);
2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244
	return *hpage;
}
#else
static struct page *khugepaged_alloc_hugepage(bool *wait)
{
	struct page *hpage;

	do {
		hpage = alloc_hugepage(khugepaged_defrag());
		if (!hpage) {
			count_vm_event(THP_COLLAPSE_ALLOC_FAILED);
			if (!*wait)
				return NULL;

			*wait = false;
			khugepaged_alloc_sleep();
		} else
			count_vm_event(THP_COLLAPSE_ALLOC);
	} while (unlikely(!hpage) && likely(khugepaged_enabled()));

	return hpage;
}

static bool khugepaged_prealloc_page(struct page **hpage, bool *wait)
{
	if (!*hpage)
		*hpage = khugepaged_alloc_hugepage(wait);

	if (unlikely(!*hpage))
		return false;

	return true;
}

static struct page
*khugepaged_alloc_page(struct page **hpage, struct mm_struct *mm,
		       struct vm_area_struct *vma, unsigned long address,
		       int node)
{
	up_read(&mm->mmap_sem);
	VM_BUG_ON(!*hpage);
	return  *hpage;
}
2245 2246
#endif

B
Bob Liu 已提交
2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260
static bool hugepage_vma_check(struct vm_area_struct *vma)
{
	if ((!(vma->vm_flags & VM_HUGEPAGE) && !khugepaged_always()) ||
	    (vma->vm_flags & VM_NOHUGEPAGE))
		return false;

	if (!vma->anon_vma || vma->vm_ops)
		return false;
	if (is_vma_temporary_stack(vma))
		return false;
	VM_BUG_ON(vma->vm_flags & VM_NO_THP);
	return true;
}

2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273
static void collapse_huge_page(struct mm_struct *mm,
				   unsigned long address,
				   struct page **hpage,
				   struct vm_area_struct *vma,
				   int node)
{
	pmd_t *pmd, _pmd;
	pte_t *pte;
	pgtable_t pgtable;
	struct page *new_page;
	spinlock_t *ptl;
	int isolated;
	unsigned long hstart, hend;
2274 2275
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
2276 2277 2278 2279 2280 2281 2282 2283

	VM_BUG_ON(address & ~HPAGE_PMD_MASK);

	/* release the mmap_sem read lock. */
	new_page = khugepaged_alloc_page(hpage, mm, vma, address, node);
	if (!new_page)
		return;

2284
	if (unlikely(mem_cgroup_newpage_charge(new_page, mm, GFP_KERNEL)))
2285
		return;
A
Andrea Arcangeli 已提交
2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300

	/*
	 * Prevent all access to pagetables with the exception of
	 * gup_fast later hanlded by the ptep_clear_flush and the VM
	 * handled by the anon_vma lock + PG_lock.
	 */
	down_write(&mm->mmap_sem);
	if (unlikely(khugepaged_test_exit(mm)))
		goto out;

	vma = find_vma(mm, address);
	hstart = (vma->vm_start + ~HPAGE_PMD_MASK) & HPAGE_PMD_MASK;
	hend = vma->vm_end & HPAGE_PMD_MASK;
	if (address < hstart || address + HPAGE_PMD_SIZE > hend)
		goto out;
B
Bob Liu 已提交
2301
	if (!hugepage_vma_check(vma))
2302
		goto out;
B
Bob Liu 已提交
2303 2304
	pmd = mm_find_pmd(mm, address);
	if (!pmd)
A
Andrea Arcangeli 已提交
2305
		goto out;
B
Bob Liu 已提交
2306
	if (pmd_trans_huge(*pmd))
A
Andrea Arcangeli 已提交
2307 2308
		goto out;

2309
	anon_vma_lock_write(vma->anon_vma);
A
Andrea Arcangeli 已提交
2310 2311 2312 2313

	pte = pte_offset_map(pmd, address);
	ptl = pte_lockptr(mm, pmd);

2314 2315 2316
	mmun_start = address;
	mmun_end   = address + HPAGE_PMD_SIZE;
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);
A
Andrea Arcangeli 已提交
2317 2318 2319 2320 2321 2322 2323
	spin_lock(&mm->page_table_lock); /* probably unnecessary */
	/*
	 * After this gup_fast can't run anymore. This also removes
	 * any huge TLB entry from the CPU so we won't allow
	 * huge and small TLB entries for the same virtual address
	 * to avoid the risk of CPU bugs in that area.
	 */
2324
	_pmd = pmdp_clear_flush(vma, address, pmd);
A
Andrea Arcangeli 已提交
2325
	spin_unlock(&mm->page_table_lock);
2326
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
A
Andrea Arcangeli 已提交
2327 2328 2329 2330 2331 2332

	spin_lock(ptl);
	isolated = __collapse_huge_page_isolate(vma, address, pte);
	spin_unlock(ptl);

	if (unlikely(!isolated)) {
2333
		pte_unmap(pte);
A
Andrea Arcangeli 已提交
2334 2335
		spin_lock(&mm->page_table_lock);
		BUG_ON(!pmd_none(*pmd));
2336 2337 2338 2339 2340 2341
		/*
		 * We can only use set_pmd_at when establishing
		 * hugepmds and never for establishing regular pmds that
		 * points to regular pagetables. Use pmd_populate for that
		 */
		pmd_populate(mm, pmd, pmd_pgtable(_pmd));
A
Andrea Arcangeli 已提交
2342
		spin_unlock(&mm->page_table_lock);
2343
		anon_vma_unlock_write(vma->anon_vma);
2344
		goto out;
A
Andrea Arcangeli 已提交
2345 2346 2347 2348 2349 2350
	}

	/*
	 * All pages are isolated and locked so anon_vma rmap
	 * can't run anymore.
	 */
2351
	anon_vma_unlock_write(vma->anon_vma);
A
Andrea Arcangeli 已提交
2352 2353

	__collapse_huge_page_copy(pte, new_page, vma, address, ptl);
2354
	pte_unmap(pte);
A
Andrea Arcangeli 已提交
2355 2356 2357
	__SetPageUptodate(new_page);
	pgtable = pmd_pgtable(_pmd);

B
Bob Liu 已提交
2358
	_pmd = mk_huge_pmd(new_page, vma);
A
Andrea Arcangeli 已提交
2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370

	/*
	 * spin_lock() below is not the equivalent of smp_wmb(), so
	 * this is needed to avoid the copy_huge_page writes to become
	 * visible after the set_pmd_at() write.
	 */
	smp_wmb();

	spin_lock(&mm->page_table_lock);
	BUG_ON(!pmd_none(*pmd));
	page_add_new_anon_rmap(new_page, vma, address);
	set_pmd_at(mm, address, pmd, _pmd);
2371
	update_mmu_cache_pmd(vma, address, pmd);
2372
	pgtable_trans_huge_deposit(mm, pmd, pgtable);
A
Andrea Arcangeli 已提交
2373 2374 2375
	spin_unlock(&mm->page_table_lock);

	*hpage = NULL;
2376

A
Andrea Arcangeli 已提交
2377
	khugepaged_pages_collapsed++;
2378
out_up_write:
A
Andrea Arcangeli 已提交
2379
	up_write(&mm->mmap_sem);
2380 2381
	return;

2382
out:
2383
	mem_cgroup_uncharge_page(new_page);
2384
	goto out_up_write;
A
Andrea Arcangeli 已提交
2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397
}

static int khugepaged_scan_pmd(struct mm_struct *mm,
			       struct vm_area_struct *vma,
			       unsigned long address,
			       struct page **hpage)
{
	pmd_t *pmd;
	pte_t *pte, *_pte;
	int ret = 0, referenced = 0, none = 0;
	struct page *page;
	unsigned long _address;
	spinlock_t *ptl;
D
David Rientjes 已提交
2398
	int node = NUMA_NO_NODE;
A
Andrea Arcangeli 已提交
2399 2400 2401

	VM_BUG_ON(address & ~HPAGE_PMD_MASK);

B
Bob Liu 已提交
2402 2403
	pmd = mm_find_pmd(mm, address);
	if (!pmd)
A
Andrea Arcangeli 已提交
2404
		goto out;
B
Bob Liu 已提交
2405
	if (pmd_trans_huge(*pmd))
A
Andrea Arcangeli 已提交
2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422
		goto out;

	pte = pte_offset_map_lock(mm, pmd, address, &ptl);
	for (_address = address, _pte = pte; _pte < pte+HPAGE_PMD_NR;
	     _pte++, _address += PAGE_SIZE) {
		pte_t pteval = *_pte;
		if (pte_none(pteval)) {
			if (++none <= khugepaged_max_ptes_none)
				continue;
			else
				goto out_unmap;
		}
		if (!pte_present(pteval) || !pte_write(pteval))
			goto out_unmap;
		page = vm_normal_page(vma, _address, pteval);
		if (unlikely(!page))
			goto out_unmap;
2423 2424 2425 2426 2427
		/*
		 * Chose the node of the first page. This could
		 * be more sophisticated and look at more pages,
		 * but isn't for now.
		 */
D
David Rientjes 已提交
2428
		if (node == NUMA_NO_NODE)
2429
			node = page_to_nid(page);
A
Andrea Arcangeli 已提交
2430 2431 2432 2433 2434 2435
		VM_BUG_ON(PageCompound(page));
		if (!PageLRU(page) || PageLocked(page) || !PageAnon(page))
			goto out_unmap;
		/* cannot use mapcount: can't collapse if there's a gup pin */
		if (page_count(page) != 1)
			goto out_unmap;
A
Andrea Arcangeli 已提交
2436 2437
		if (pte_young(pteval) || PageReferenced(page) ||
		    mmu_notifier_test_young(vma->vm_mm, address))
A
Andrea Arcangeli 已提交
2438 2439 2440 2441 2442 2443
			referenced = 1;
	}
	if (referenced)
		ret = 1;
out_unmap:
	pte_unmap_unlock(pte, ptl);
2444 2445
	if (ret)
		/* collapse_huge_page will return with the mmap_sem released */
2446
		collapse_huge_page(mm, address, hpage, vma, node);
A
Andrea Arcangeli 已提交
2447 2448 2449 2450 2451 2452 2453 2454
out:
	return ret;
}

static void collect_mm_slot(struct mm_slot *mm_slot)
{
	struct mm_struct *mm = mm_slot->mm;

2455
	VM_BUG_ON(NR_CPUS != 1 && !spin_is_locked(&khugepaged_mm_lock));
A
Andrea Arcangeli 已提交
2456 2457 2458

	if (khugepaged_test_exit(mm)) {
		/* free mm_slot */
2459
		hash_del(&mm_slot->hash);
A
Andrea Arcangeli 已提交
2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475
		list_del(&mm_slot->mm_node);

		/*
		 * Not strictly needed because the mm exited already.
		 *
		 * clear_bit(MMF_VM_HUGEPAGE, &mm->flags);
		 */

		/* khugepaged_mm_lock actually not necessary for the below */
		free_mm_slot(mm_slot);
		mmdrop(mm);
	}
}

static unsigned int khugepaged_scan_mm_slot(unsigned int pages,
					    struct page **hpage)
2476 2477
	__releases(&khugepaged_mm_lock)
	__acquires(&khugepaged_mm_lock)
A
Andrea Arcangeli 已提交
2478 2479 2480 2481 2482 2483 2484
{
	struct mm_slot *mm_slot;
	struct mm_struct *mm;
	struct vm_area_struct *vma;
	int progress = 0;

	VM_BUG_ON(!pages);
2485
	VM_BUG_ON(NR_CPUS != 1 && !spin_is_locked(&khugepaged_mm_lock));
A
Andrea Arcangeli 已提交
2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512

	if (khugepaged_scan.mm_slot)
		mm_slot = khugepaged_scan.mm_slot;
	else {
		mm_slot = list_entry(khugepaged_scan.mm_head.next,
				     struct mm_slot, mm_node);
		khugepaged_scan.address = 0;
		khugepaged_scan.mm_slot = mm_slot;
	}
	spin_unlock(&khugepaged_mm_lock);

	mm = mm_slot->mm;
	down_read(&mm->mmap_sem);
	if (unlikely(khugepaged_test_exit(mm)))
		vma = NULL;
	else
		vma = find_vma(mm, khugepaged_scan.address);

	progress++;
	for (; vma; vma = vma->vm_next) {
		unsigned long hstart, hend;

		cond_resched();
		if (unlikely(khugepaged_test_exit(mm))) {
			progress++;
			break;
		}
B
Bob Liu 已提交
2513 2514
		if (!hugepage_vma_check(vma)) {
skip:
A
Andrea Arcangeli 已提交
2515 2516 2517 2518 2519
			progress++;
			continue;
		}
		hstart = (vma->vm_start + ~HPAGE_PMD_MASK) & HPAGE_PMD_MASK;
		hend = vma->vm_end & HPAGE_PMD_MASK;
2520 2521 2522 2523
		if (hstart >= hend)
			goto skip;
		if (khugepaged_scan.address > hend)
			goto skip;
A
Andrea Arcangeli 已提交
2524 2525
		if (khugepaged_scan.address < hstart)
			khugepaged_scan.address = hstart;
2526
		VM_BUG_ON(khugepaged_scan.address & ~HPAGE_PMD_MASK);
A
Andrea Arcangeli 已提交
2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554

		while (khugepaged_scan.address < hend) {
			int ret;
			cond_resched();
			if (unlikely(khugepaged_test_exit(mm)))
				goto breakouterloop;

			VM_BUG_ON(khugepaged_scan.address < hstart ||
				  khugepaged_scan.address + HPAGE_PMD_SIZE >
				  hend);
			ret = khugepaged_scan_pmd(mm, vma,
						  khugepaged_scan.address,
						  hpage);
			/* move to next address */
			khugepaged_scan.address += HPAGE_PMD_SIZE;
			progress += HPAGE_PMD_NR;
			if (ret)
				/* we released mmap_sem so break loop */
				goto breakouterloop_mmap_sem;
			if (progress >= pages)
				goto breakouterloop;
		}
	}
breakouterloop:
	up_read(&mm->mmap_sem); /* exit_mmap will destroy ptes after this */
breakouterloop_mmap_sem:

	spin_lock(&khugepaged_mm_lock);
2555
	VM_BUG_ON(khugepaged_scan.mm_slot != mm_slot);
A
Andrea Arcangeli 已提交
2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590
	/*
	 * Release the current mm_slot if this mm is about to die, or
	 * if we scanned all vmas of this mm.
	 */
	if (khugepaged_test_exit(mm) || !vma) {
		/*
		 * Make sure that if mm_users is reaching zero while
		 * khugepaged runs here, khugepaged_exit will find
		 * mm_slot not pointing to the exiting mm.
		 */
		if (mm_slot->mm_node.next != &khugepaged_scan.mm_head) {
			khugepaged_scan.mm_slot = list_entry(
				mm_slot->mm_node.next,
				struct mm_slot, mm_node);
			khugepaged_scan.address = 0;
		} else {
			khugepaged_scan.mm_slot = NULL;
			khugepaged_full_scans++;
		}

		collect_mm_slot(mm_slot);
	}

	return progress;
}

static int khugepaged_has_work(void)
{
	return !list_empty(&khugepaged_scan.mm_head) &&
		khugepaged_enabled();
}

static int khugepaged_wait_event(void)
{
	return !list_empty(&khugepaged_scan.mm_head) ||
2591
		kthread_should_stop();
A
Andrea Arcangeli 已提交
2592 2593
}

2594
static void khugepaged_do_scan(void)
A
Andrea Arcangeli 已提交
2595
{
2596
	struct page *hpage = NULL;
A
Andrea Arcangeli 已提交
2597 2598
	unsigned int progress = 0, pass_through_head = 0;
	unsigned int pages = khugepaged_pages_to_scan;
2599
	bool wait = true;
A
Andrea Arcangeli 已提交
2600 2601 2602 2603

	barrier(); /* write khugepaged_pages_to_scan to local stack */

	while (progress < pages) {
2604
		if (!khugepaged_prealloc_page(&hpage, &wait))
2605
			break;
2606

2607
		cond_resched();
A
Andrea Arcangeli 已提交
2608

2609 2610 2611
		if (unlikely(kthread_should_stop() || freezing(current)))
			break;

A
Andrea Arcangeli 已提交
2612 2613 2614 2615 2616 2617
		spin_lock(&khugepaged_mm_lock);
		if (!khugepaged_scan.mm_slot)
			pass_through_head++;
		if (khugepaged_has_work() &&
		    pass_through_head < 2)
			progress += khugepaged_scan_mm_slot(pages - progress,
2618
							    &hpage);
A
Andrea Arcangeli 已提交
2619 2620 2621 2622 2623
		else
			progress = pages;
		spin_unlock(&khugepaged_mm_lock);
	}

2624 2625
	if (!IS_ERR_OR_NULL(hpage))
		put_page(hpage);
2626 2627
}

2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639 2640 2641 2642 2643 2644 2645
static void khugepaged_wait_work(void)
{
	try_to_freeze();

	if (khugepaged_has_work()) {
		if (!khugepaged_scan_sleep_millisecs)
			return;

		wait_event_freezable_timeout(khugepaged_wait,
					     kthread_should_stop(),
			msecs_to_jiffies(khugepaged_scan_sleep_millisecs));
		return;
	}

	if (khugepaged_enabled())
		wait_event_freezable(khugepaged_wait, khugepaged_wait_event());
}

A
Andrea Arcangeli 已提交
2646 2647 2648 2649
static int khugepaged(void *none)
{
	struct mm_slot *mm_slot;

2650
	set_freezable();
A
Andrea Arcangeli 已提交
2651 2652
	set_user_nice(current, 19);

X
Xiao Guangrong 已提交
2653 2654 2655 2656
	while (!kthread_should_stop()) {
		khugepaged_do_scan();
		khugepaged_wait_work();
	}
A
Andrea Arcangeli 已提交
2657 2658 2659 2660 2661 2662 2663 2664 2665 2666

	spin_lock(&khugepaged_mm_lock);
	mm_slot = khugepaged_scan.mm_slot;
	khugepaged_scan.mm_slot = NULL;
	if (mm_slot)
		collect_mm_slot(mm_slot);
	spin_unlock(&khugepaged_mm_lock);
	return 0;
}

2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677
static void __split_huge_zero_page_pmd(struct vm_area_struct *vma,
		unsigned long haddr, pmd_t *pmd)
{
	struct mm_struct *mm = vma->vm_mm;
	pgtable_t pgtable;
	pmd_t _pmd;
	int i;

	pmdp_clear_flush(vma, haddr, pmd);
	/* leave pmd empty until pte is filled */

2678
	pgtable = pgtable_trans_huge_withdraw(mm, pmd);
2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691
	pmd_populate(mm, &_pmd, pgtable);

	for (i = 0; i < HPAGE_PMD_NR; i++, haddr += PAGE_SIZE) {
		pte_t *pte, entry;
		entry = pfn_pte(my_zero_pfn(haddr), vma->vm_page_prot);
		entry = pte_mkspecial(entry);
		pte = pte_offset_map(&_pmd, haddr);
		VM_BUG_ON(!pte_none(*pte));
		set_pte_at(mm, haddr, pte, entry);
		pte_unmap(pte);
	}
	smp_wmb(); /* make pte visible before pmd */
	pmd_populate(mm, pmd, pgtable);
2692
	put_huge_zero_page();
2693 2694
}

2695 2696
void __split_huge_page_pmd(struct vm_area_struct *vma, unsigned long address,
		pmd_t *pmd)
2697 2698
{
	struct page *page;
2699
	struct mm_struct *mm = vma->vm_mm;
2700 2701 2702
	unsigned long haddr = address & HPAGE_PMD_MASK;
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
2703 2704

	BUG_ON(vma->vm_start > haddr || vma->vm_end < haddr + HPAGE_PMD_SIZE);
2705

2706 2707 2708
	mmun_start = haddr;
	mmun_end   = haddr + HPAGE_PMD_SIZE;
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);
2709 2710 2711
	spin_lock(&mm->page_table_lock);
	if (unlikely(!pmd_trans_huge(*pmd))) {
		spin_unlock(&mm->page_table_lock);
2712 2713 2714 2715 2716 2717 2718
		mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
		return;
	}
	if (is_huge_zero_pmd(*pmd)) {
		__split_huge_zero_page_pmd(vma, haddr, pmd);
		spin_unlock(&mm->page_table_lock);
		mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
2719 2720 2721 2722 2723 2724
		return;
	}
	page = pmd_page(*pmd);
	VM_BUG_ON(!page_count(page));
	get_page(page);
	spin_unlock(&mm->page_table_lock);
2725
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
2726 2727 2728 2729 2730 2731

	split_huge_page(page);

	put_page(page);
	BUG_ON(pmd_trans_huge(*pmd));
}
2732

2733 2734 2735 2736 2737 2738 2739 2740 2741 2742
void split_huge_page_pmd_mm(struct mm_struct *mm, unsigned long address,
		pmd_t *pmd)
{
	struct vm_area_struct *vma;

	vma = find_vma(mm, address);
	BUG_ON(vma == NULL);
	split_huge_page_pmd(vma, address, pmd);
}

2743 2744 2745 2746 2747 2748 2749
static void split_huge_page_address(struct mm_struct *mm,
				    unsigned long address)
{
	pmd_t *pmd;

	VM_BUG_ON(!(address & ~HPAGE_PMD_MASK));

B
Bob Liu 已提交
2750 2751
	pmd = mm_find_pmd(mm, address);
	if (!pmd)
2752 2753 2754 2755 2756
		return;
	/*
	 * Caller holds the mmap_sem write mode, so a huge pmd cannot
	 * materialize from under us.
	 */
2757
	split_huge_page_pmd_mm(mm, address, pmd);
2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799
}

void __vma_adjust_trans_huge(struct vm_area_struct *vma,
			     unsigned long start,
			     unsigned long end,
			     long adjust_next)
{
	/*
	 * If the new start address isn't hpage aligned and it could
	 * previously contain an hugepage: check if we need to split
	 * an huge pmd.
	 */
	if (start & ~HPAGE_PMD_MASK &&
	    (start & HPAGE_PMD_MASK) >= vma->vm_start &&
	    (start & HPAGE_PMD_MASK) + HPAGE_PMD_SIZE <= vma->vm_end)
		split_huge_page_address(vma->vm_mm, start);

	/*
	 * If the new end address isn't hpage aligned and it could
	 * previously contain an hugepage: check if we need to split
	 * an huge pmd.
	 */
	if (end & ~HPAGE_PMD_MASK &&
	    (end & HPAGE_PMD_MASK) >= vma->vm_start &&
	    (end & HPAGE_PMD_MASK) + HPAGE_PMD_SIZE <= vma->vm_end)
		split_huge_page_address(vma->vm_mm, end);

	/*
	 * If we're also updating the vma->vm_next->vm_start, if the new
	 * vm_next->vm_start isn't page aligned and it could previously
	 * contain an hugepage: check if we need to split an huge pmd.
	 */
	if (adjust_next > 0) {
		struct vm_area_struct *next = vma->vm_next;
		unsigned long nstart = next->vm_start;
		nstart += adjust_next << PAGE_SHIFT;
		if (nstart & ~HPAGE_PMD_MASK &&
		    (nstart & HPAGE_PMD_MASK) >= next->vm_start &&
		    (nstart & HPAGE_PMD_MASK) + HPAGE_PMD_SIZE <= next->vm_end)
			split_huge_page_address(next->vm_mm, nstart);
	}
}