ksm.c 55.4 KB
Newer Older
H
Hugh Dickins 已提交
1
/*
I
Izik Eidus 已提交
2 3 4 5 6
 * Memory merging support.
 *
 * This code enables dynamic sharing of identical pages found in different
 * memory areas, even if they are not shared by fork()
 *
I
Izik Eidus 已提交
7
 * Copyright (C) 2008-2009 Red Hat, Inc.
I
Izik Eidus 已提交
8 9 10 11
 * Authors:
 *	Izik Eidus
 *	Andrea Arcangeli
 *	Chris Wright
I
Izik Eidus 已提交
12
 *	Hugh Dickins
I
Izik Eidus 已提交
13 14
 *
 * This work is licensed under the terms of the GNU GPL, version 2.
H
Hugh Dickins 已提交
15 16 17
 */

#include <linux/errno.h>
I
Izik Eidus 已提交
18 19
#include <linux/mm.h>
#include <linux/fs.h>
H
Hugh Dickins 已提交
20
#include <linux/mman.h>
I
Izik Eidus 已提交
21 22 23 24 25 26 27 28 29 30 31
#include <linux/sched.h>
#include <linux/rwsem.h>
#include <linux/pagemap.h>
#include <linux/rmap.h>
#include <linux/spinlock.h>
#include <linux/jhash.h>
#include <linux/delay.h>
#include <linux/kthread.h>
#include <linux/wait.h>
#include <linux/slab.h>
#include <linux/rbtree.h>
32
#include <linux/memory.h>
I
Izik Eidus 已提交
33
#include <linux/mmu_notifier.h>
34
#include <linux/swap.h>
H
Hugh Dickins 已提交
35
#include <linux/ksm.h>
L
Lai Jiangshan 已提交
36
#include <linux/hash.h>
37
#include <linux/freezer.h>
38
#include <linux/oom.h>
H
Hugh Dickins 已提交
39

I
Izik Eidus 已提交
40
#include <asm/tlbflush.h>
H
Hugh Dickins 已提交
41
#include "internal.h"
I
Izik Eidus 已提交
42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86

/*
 * A few notes about the KSM scanning process,
 * to make it easier to understand the data structures below:
 *
 * In order to reduce excessive scanning, KSM sorts the memory pages by their
 * contents into a data structure that holds pointers to the pages' locations.
 *
 * Since the contents of the pages may change at any moment, KSM cannot just
 * insert the pages into a normal sorted tree and expect it to find anything.
 * Therefore KSM uses two data structures - the stable and the unstable tree.
 *
 * The stable tree holds pointers to all the merged pages (ksm pages), sorted
 * by their contents.  Because each such page is write-protected, searching on
 * this tree is fully assured to be working (except when pages are unmapped),
 * and therefore this tree is called the stable tree.
 *
 * In addition to the stable tree, KSM uses a second data structure called the
 * unstable tree: this tree holds pointers to pages which have been found to
 * be "unchanged for a period of time".  The unstable tree sorts these pages
 * by their contents, but since they are not write-protected, KSM cannot rely
 * upon the unstable tree to work correctly - the unstable tree is liable to
 * be corrupted as its contents are modified, and so it is called unstable.
 *
 * KSM solves this problem by several techniques:
 *
 * 1) The unstable tree is flushed every time KSM completes scanning all
 *    memory areas, and then the tree is rebuilt again from the beginning.
 * 2) KSM will only insert into the unstable tree, pages whose hash value
 *    has not changed since the previous scan of all memory areas.
 * 3) The unstable tree is a RedBlack Tree - so its balancing is based on the
 *    colors of the nodes and not on their contents, assuring that even when
 *    the tree gets "corrupted" it won't get out of balance, so scanning time
 *    remains the same (also, searching and inserting nodes in an rbtree uses
 *    the same algorithm, so we have no overhead when we flush and rebuild).
 * 4) KSM never flushes the stable tree, which means that even if it were to
 *    take 10 attempts to find a page in the unstable tree, once it is found,
 *    it is secured in the stable tree.  (When we scan a new page, we first
 *    compare it against the stable tree, and then against the unstable tree.)
 */

/**
 * struct mm_slot - ksm information per mm that is being scanned
 * @link: link to the mm_slots hash list
 * @mm_list: link into the mm_slots list, rooted in ksm_mm_head
H
Hugh Dickins 已提交
87
 * @rmap_list: head for this mm_slot's singly-linked list of rmap_items
I
Izik Eidus 已提交
88 89 90 91 92
 * @mm: the mm that this information is valid for
 */
struct mm_slot {
	struct hlist_node link;
	struct list_head mm_list;
H
Hugh Dickins 已提交
93
	struct rmap_item *rmap_list;
I
Izik Eidus 已提交
94 95 96 97 98 99 100
	struct mm_struct *mm;
};

/**
 * struct ksm_scan - cursor for scanning
 * @mm_slot: the current mm_slot we are scanning
 * @address: the next address inside that to be scanned
H
Hugh Dickins 已提交
101
 * @rmap_list: link to the next rmap to be scanned in the rmap_list
I
Izik Eidus 已提交
102 103 104 105 106 107 108
 * @seqnr: count of completed full scans (needed when removing unstable node)
 *
 * There is only the one ksm_scan instance of this cursor structure.
 */
struct ksm_scan {
	struct mm_slot *mm_slot;
	unsigned long address;
H
Hugh Dickins 已提交
109
	struct rmap_item **rmap_list;
I
Izik Eidus 已提交
110 111 112
	unsigned long seqnr;
};

H
Hugh Dickins 已提交
113 114 115 116
/**
 * struct stable_node - node of the stable rbtree
 * @node: rb node of this ksm page in the stable tree
 * @hlist: hlist head of rmap_items using this ksm page
117
 * @kpfn: page frame number of this ksm page
H
Hugh Dickins 已提交
118 119 120 121
 */
struct stable_node {
	struct rb_node node;
	struct hlist_head hlist;
122
	unsigned long kpfn;
H
Hugh Dickins 已提交
123 124
};

I
Izik Eidus 已提交
125 126
/**
 * struct rmap_item - reverse mapping item for virtual addresses
H
Hugh Dickins 已提交
127
 * @rmap_list: next rmap_item in mm_slot's singly-linked rmap_list
H
Hugh Dickins 已提交
128
 * @anon_vma: pointer to anon_vma for this mm,address, when in stable tree
I
Izik Eidus 已提交
129 130 131
 * @mm: the memory structure this rmap_item is pointing into
 * @address: the virtual address this rmap_item tracks (+ flags in low bits)
 * @oldchecksum: previous checksum of the page at that virtual address
H
Hugh Dickins 已提交
132 133 134
 * @node: rb node of this rmap_item in the unstable tree
 * @head: pointer to stable_node heading this list in the stable tree
 * @hlist: link into hlist of rmap_items hanging off that stable_node
I
Izik Eidus 已提交
135 136
 */
struct rmap_item {
H
Hugh Dickins 已提交
137
	struct rmap_item *rmap_list;
H
Hugh Dickins 已提交
138
	struct anon_vma *anon_vma;	/* when stable */
I
Izik Eidus 已提交
139 140
	struct mm_struct *mm;
	unsigned long address;		/* + low bits used for flags below */
H
Hugh Dickins 已提交
141
	unsigned int oldchecksum;	/* when unstable */
I
Izik Eidus 已提交
142
	union {
H
Hugh Dickins 已提交
143 144 145 146 147
		struct rb_node node;	/* when node of unstable tree */
		struct {		/* when listed from stable tree */
			struct stable_node *head;
			struct hlist_node hlist;
		};
I
Izik Eidus 已提交
148 149 150 151
	};
};

#define SEQNR_MASK	0x0ff	/* low bits of unstable tree seqnr */
H
Hugh Dickins 已提交
152 153
#define UNSTABLE_FLAG	0x100	/* is a node of the unstable tree */
#define STABLE_FLAG	0x200	/* is listed from the stable tree */
I
Izik Eidus 已提交
154 155 156 157 158

/* The stable and unstable tree heads */
static struct rb_root root_stable_tree = RB_ROOT;
static struct rb_root root_unstable_tree = RB_ROOT;

L
Lai Jiangshan 已提交
159 160 161
#define MM_SLOTS_HASH_SHIFT 10
#define MM_SLOTS_HASH_HEADS (1 << MM_SLOTS_HASH_SHIFT)
static struct hlist_head mm_slots_hash[MM_SLOTS_HASH_HEADS];
I
Izik Eidus 已提交
162 163 164 165 166 167 168 169 170

static struct mm_slot ksm_mm_head = {
	.mm_list = LIST_HEAD_INIT(ksm_mm_head.mm_list),
};
static struct ksm_scan ksm_scan = {
	.mm_slot = &ksm_mm_head,
};

static struct kmem_cache *rmap_item_cache;
H
Hugh Dickins 已提交
171
static struct kmem_cache *stable_node_cache;
I
Izik Eidus 已提交
172 173 174
static struct kmem_cache *mm_slot_cache;

/* The number of nodes in the stable tree */
175
static unsigned long ksm_pages_shared;
I
Izik Eidus 已提交
176

H
Hugh Dickins 已提交
177
/* The number of page slots additionally sharing those nodes */
178
static unsigned long ksm_pages_sharing;
I
Izik Eidus 已提交
179

180 181 182 183 184 185
/* The number of nodes in the unstable tree */
static unsigned long ksm_pages_unshared;

/* The number of rmap_items in use: to calculate pages_volatile */
static unsigned long ksm_rmap_items;

I
Izik Eidus 已提交
186
/* Number of pages ksmd should scan in one batch */
187
static unsigned int ksm_thread_pages_to_scan = 100;
I
Izik Eidus 已提交
188 189

/* Milliseconds ksmd should sleep between batches */
H
Hugh Dickins 已提交
190
static unsigned int ksm_thread_sleep_millisecs = 20;
I
Izik Eidus 已提交
191 192 193 194

#define KSM_RUN_STOP	0
#define KSM_RUN_MERGE	1
#define KSM_RUN_UNMERGE	2
195
static unsigned int ksm_run = KSM_RUN_STOP;
I
Izik Eidus 已提交
196 197 198 199 200 201 202 203 204 205 206 207 208 209 210

static DECLARE_WAIT_QUEUE_HEAD(ksm_thread_wait);
static DEFINE_MUTEX(ksm_thread_mutex);
static DEFINE_SPINLOCK(ksm_mmlist_lock);

#define KSM_KMEM_CACHE(__struct, __flags) kmem_cache_create("ksm_"#__struct,\
		sizeof(struct __struct), __alignof__(struct __struct),\
		(__flags), NULL)

static int __init ksm_slab_init(void)
{
	rmap_item_cache = KSM_KMEM_CACHE(rmap_item, 0);
	if (!rmap_item_cache)
		goto out;

H
Hugh Dickins 已提交
211 212 213 214
	stable_node_cache = KSM_KMEM_CACHE(stable_node, 0);
	if (!stable_node_cache)
		goto out_free1;

I
Izik Eidus 已提交
215 216
	mm_slot_cache = KSM_KMEM_CACHE(mm_slot, 0);
	if (!mm_slot_cache)
H
Hugh Dickins 已提交
217
		goto out_free2;
I
Izik Eidus 已提交
218 219 220

	return 0;

H
Hugh Dickins 已提交
221 222 223
out_free2:
	kmem_cache_destroy(stable_node_cache);
out_free1:
I
Izik Eidus 已提交
224 225 226 227 228 229 230 231
	kmem_cache_destroy(rmap_item_cache);
out:
	return -ENOMEM;
}

static void __init ksm_slab_free(void)
{
	kmem_cache_destroy(mm_slot_cache);
H
Hugh Dickins 已提交
232
	kmem_cache_destroy(stable_node_cache);
I
Izik Eidus 已提交
233 234 235 236 237 238
	kmem_cache_destroy(rmap_item_cache);
	mm_slot_cache = NULL;
}

static inline struct rmap_item *alloc_rmap_item(void)
{
239 240 241 242 243 244
	struct rmap_item *rmap_item;

	rmap_item = kmem_cache_zalloc(rmap_item_cache, GFP_KERNEL);
	if (rmap_item)
		ksm_rmap_items++;
	return rmap_item;
I
Izik Eidus 已提交
245 246 247 248
}

static inline void free_rmap_item(struct rmap_item *rmap_item)
{
249
	ksm_rmap_items--;
I
Izik Eidus 已提交
250 251 252 253
	rmap_item->mm = NULL;	/* debug safety */
	kmem_cache_free(rmap_item_cache, rmap_item);
}

H
Hugh Dickins 已提交
254 255 256 257 258 259 260 261 262 263
static inline struct stable_node *alloc_stable_node(void)
{
	return kmem_cache_alloc(stable_node_cache, GFP_KERNEL);
}

static inline void free_stable_node(struct stable_node *stable_node)
{
	kmem_cache_free(stable_node_cache, stable_node);
}

I
Izik Eidus 已提交
264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281
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;
	struct hlist_head *bucket;
	struct hlist_node *node;

L
Lai Jiangshan 已提交
282
	bucket = &mm_slots_hash[hash_ptr(mm, MM_SLOTS_HASH_SHIFT)];
I
Izik Eidus 已提交
283 284 285 286 287 288 289 290 291 292 293 294
	hlist_for_each_entry(mm_slot, node, bucket, link) {
		if (mm == mm_slot->mm)
			return mm_slot;
	}
	return NULL;
}

static void insert_to_mm_slots_hash(struct mm_struct *mm,
				    struct mm_slot *mm_slot)
{
	struct hlist_head *bucket;

L
Lai Jiangshan 已提交
295
	bucket = &mm_slots_hash[hash_ptr(mm, MM_SLOTS_HASH_SHIFT)];
I
Izik Eidus 已提交
296 297 298 299 300 301 302 303 304
	mm_slot->mm = mm;
	hlist_add_head(&mm_slot->link, bucket);
}

static inline int in_stable_tree(struct rmap_item *rmap_item)
{
	return rmap_item->address & STABLE_FLAG;
}

H
Hugh Dickins 已提交
305 306 307 308 309 310 311 312 313 314 315 316 317
/*
 * ksmd, and unmerge_and_remove_all_rmap_items(), must not touch an mm's
 * page tables after it has passed through ksm_exit() - which, if necessary,
 * takes mmap_sem briefly to serialize against them.  ksm_exit() does not set
 * a special flag: they can just back out as soon as mm_users goes to zero.
 * ksm_test_exit() is used throughout to make this test for exit: in some
 * places for correctness, in some places just to avoid unnecessary work.
 */
static inline bool ksm_test_exit(struct mm_struct *mm)
{
	return atomic_read(&mm->mm_users) == 0;
}

I
Izik Eidus 已提交
318 319 320 321 322 323 324 325 326 327 328
/*
 * We use break_ksm to break COW on a ksm page: it's a stripped down
 *
 *	if (get_user_pages(current, mm, addr, 1, 1, 1, &page, NULL) == 1)
 *		put_page(page);
 *
 * but taking great care only to touch a ksm page, in a VM_MERGEABLE vma,
 * in case the application has unmapped and remapped mm,addr meanwhile.
 * Could a ksm page appear anywhere else?  Actually yes, in a VM_PFNMAP
 * mmap of /dev/mem or /dev/kmem, where we would not want to touch it.
 */
H
Hugh Dickins 已提交
329
static int break_ksm(struct vm_area_struct *vma, unsigned long addr)
I
Izik Eidus 已提交
330 331
{
	struct page *page;
H
Hugh Dickins 已提交
332
	int ret = 0;
I
Izik Eidus 已提交
333 334 335 336

	do {
		cond_resched();
		page = follow_page(vma, addr, FOLL_GET);
337
		if (IS_ERR_OR_NULL(page))
I
Izik Eidus 已提交
338 339 340 341 342 343 344
			break;
		if (PageKsm(page))
			ret = handle_mm_fault(vma->vm_mm, vma, addr,
							FAULT_FLAG_WRITE);
		else
			ret = VM_FAULT_WRITE;
		put_page(page);
H
Hugh Dickins 已提交
345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374
	} while (!(ret & (VM_FAULT_WRITE | VM_FAULT_SIGBUS | VM_FAULT_OOM)));
	/*
	 * We must loop because handle_mm_fault() may back out if there's
	 * any difficulty e.g. if pte accessed bit gets updated concurrently.
	 *
	 * VM_FAULT_WRITE is what we have been hoping for: it indicates that
	 * COW has been broken, even if the vma does not permit VM_WRITE;
	 * but note that a concurrent fault might break PageKsm for us.
	 *
	 * VM_FAULT_SIGBUS could occur if we race with truncation of the
	 * backing file, which also invalidates anonymous pages: that's
	 * okay, that truncation will have unmapped the PageKsm for us.
	 *
	 * VM_FAULT_OOM: at the time of writing (late July 2009), setting
	 * aside mem_cgroup limits, VM_FAULT_OOM would only be set if the
	 * current task has TIF_MEMDIE set, and will be OOM killed on return
	 * to user; and ksmd, having no mm, would never be chosen for that.
	 *
	 * But if the mm is in a limited mem_cgroup, then the fault may fail
	 * with VM_FAULT_OOM even if the current task is not TIF_MEMDIE; and
	 * even ksmd can fail in this way - though it's usually breaking ksm
	 * just to undo a merge it made a moment before, so unlikely to oom.
	 *
	 * That's a pity: we might therefore have more kernel pages allocated
	 * than we're counting as nodes in the stable tree; but ksm_do_scan
	 * will retry to break_cow on each pass, so should recover the page
	 * in due course.  The important thing is to not let VM_MERGEABLE
	 * be cleared while any such pages might remain in the area.
	 */
	return (ret & VM_FAULT_OOM) ? -ENOMEM : 0;
I
Izik Eidus 已提交
375 376
}

377 378 379 380 381 382 383 384 385 386 387 388 389 390
static struct vm_area_struct *find_mergeable_vma(struct mm_struct *mm,
		unsigned long addr)
{
	struct vm_area_struct *vma;
	if (ksm_test_exit(mm))
		return NULL;
	vma = find_vma(mm, addr);
	if (!vma || vma->vm_start > addr)
		return NULL;
	if (!(vma->vm_flags & VM_MERGEABLE) || !vma->anon_vma)
		return NULL;
	return vma;
}

391
static void break_cow(struct rmap_item *rmap_item)
I
Izik Eidus 已提交
392
{
393 394
	struct mm_struct *mm = rmap_item->mm;
	unsigned long addr = rmap_item->address;
I
Izik Eidus 已提交
395 396
	struct vm_area_struct *vma;

H
Hugh Dickins 已提交
397 398 399 400
	/*
	 * It is not an accident that whenever we want to break COW
	 * to undo, we also need to drop a reference to the anon_vma.
	 */
401
	put_anon_vma(rmap_item->anon_vma);
H
Hugh Dickins 已提交
402

H
Hugh Dickins 已提交
403
	down_read(&mm->mmap_sem);
404 405 406
	vma = find_mergeable_vma(mm, addr);
	if (vma)
		break_ksm(vma, addr);
I
Izik Eidus 已提交
407 408 409
	up_read(&mm->mmap_sem);
}

A
Andrea Arcangeli 已提交
410 411 412
static struct page *page_trans_compound_anon(struct page *page)
{
	if (PageTransCompound(page)) {
413
		struct page *head = compound_trans_head(page);
A
Andrea Arcangeli 已提交
414
		/*
415 416
		 * head may actually be splitted and freed from under
		 * us but it's ok here.
A
Andrea Arcangeli 已提交
417 418 419 420 421 422 423
		 */
		if (PageAnon(head))
			return head;
	}
	return NULL;
}

I
Izik Eidus 已提交
424 425 426 427 428 429 430 431
static struct page *get_mergeable_page(struct rmap_item *rmap_item)
{
	struct mm_struct *mm = rmap_item->mm;
	unsigned long addr = rmap_item->address;
	struct vm_area_struct *vma;
	struct page *page;

	down_read(&mm->mmap_sem);
432 433
	vma = find_mergeable_vma(mm, addr);
	if (!vma)
I
Izik Eidus 已提交
434 435 436
		goto out;

	page = follow_page(vma, addr, FOLL_GET);
437
	if (IS_ERR_OR_NULL(page))
I
Izik Eidus 已提交
438
		goto out;
A
Andrea Arcangeli 已提交
439
	if (PageAnon(page) || page_trans_compound_anon(page)) {
I
Izik Eidus 已提交
440 441 442 443 444 445 446 447 448 449
		flush_anon_page(vma, page, addr);
		flush_dcache_page(page);
	} else {
		put_page(page);
out:		page = NULL;
	}
	up_read(&mm->mmap_sem);
	return page;
}

H
Hugh Dickins 已提交
450 451 452 453 454 455 456 457 458 459
static void remove_node_from_stable_tree(struct stable_node *stable_node)
{
	struct rmap_item *rmap_item;
	struct hlist_node *hlist;

	hlist_for_each_entry(rmap_item, hlist, &stable_node->hlist, hlist) {
		if (rmap_item->hlist.next)
			ksm_pages_sharing--;
		else
			ksm_pages_shared--;
460
		put_anon_vma(rmap_item->anon_vma);
H
Hugh Dickins 已提交
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
		rmap_item->address &= PAGE_MASK;
		cond_resched();
	}

	rb_erase(&stable_node->node, &root_stable_tree);
	free_stable_node(stable_node);
}

/*
 * get_ksm_page: checks if the page indicated by the stable node
 * is still its ksm page, despite having held no reference to it.
 * In which case we can trust the content of the page, and it
 * returns the gotten page; but if the page has now been zapped,
 * remove the stale node from the stable tree and return NULL.
 *
 * You would expect the stable_node to hold a reference to the ksm page.
 * But if it increments the page's count, swapping out has to wait for
 * ksmd to come around again before it can free the page, which may take
 * seconds or even minutes: much too unresponsive.  So instead we use a
 * "keyhole reference": access to the ksm page from the stable node peeps
 * out through its keyhole to see if that page still holds the right key,
 * pointing back to this stable node.  This relies on freeing a PageAnon
 * page to reset its page->mapping to NULL, and relies on no other use of
 * a page to put something that might look like our key in page->mapping.
 *
 * include/linux/pagemap.h page_cache_get_speculative() is a good reference,
 * but this is different - made simpler by ksm_thread_mutex being held, but
 * interesting for assuming that no other use of the struct page could ever
 * put our expected_mapping into page->mapping (or a field of the union which
 * coincides with page->mapping).  The RCU calls are not for KSM at all, but
 * to keep the page_count protocol described with page_cache_get_speculative.
 *
 * Note: it is possible that get_ksm_page() will return NULL one moment,
 * then page the next, if the page is in between page_freeze_refs() and
 * page_unfreeze_refs(): this shouldn't be a problem anywhere, the page
 * is on its way to being freed; but it is an anomaly to bear in mind.
 */
static struct page *get_ksm_page(struct stable_node *stable_node)
{
	struct page *page;
	void *expected_mapping;

503
	page = pfn_to_page(stable_node->kpfn);
H
Hugh Dickins 已提交
504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522
	expected_mapping = (void *)stable_node +
				(PAGE_MAPPING_ANON | PAGE_MAPPING_KSM);
	rcu_read_lock();
	if (page->mapping != expected_mapping)
		goto stale;
	if (!get_page_unless_zero(page))
		goto stale;
	if (page->mapping != expected_mapping) {
		put_page(page);
		goto stale;
	}
	rcu_read_unlock();
	return page;
stale:
	rcu_read_unlock();
	remove_node_from_stable_tree(stable_node);
	return NULL;
}

I
Izik Eidus 已提交
523 524 525 526 527 528
/*
 * Removing rmap_item from stable or unstable tree.
 * This function will clean the information from the stable/unstable tree.
 */
static void remove_rmap_item_from_tree(struct rmap_item *rmap_item)
{
H
Hugh Dickins 已提交
529 530
	if (rmap_item->address & STABLE_FLAG) {
		struct stable_node *stable_node;
H
Hugh Dickins 已提交
531
		struct page *page;
I
Izik Eidus 已提交
532

H
Hugh Dickins 已提交
533
		stable_node = rmap_item->head;
H
Hugh Dickins 已提交
534 535 536
		page = get_ksm_page(stable_node);
		if (!page)
			goto out;
H
Hugh Dickins 已提交
537

H
Hugh Dickins 已提交
538
		lock_page(page);
H
Hugh Dickins 已提交
539
		hlist_del(&rmap_item->hlist);
H
Hugh Dickins 已提交
540 541
		unlock_page(page);
		put_page(page);
542

H
Hugh Dickins 已提交
543 544 545
		if (stable_node->hlist.first)
			ksm_pages_sharing--;
		else
H
Hugh Dickins 已提交
546
			ksm_pages_shared--;
I
Izik Eidus 已提交
547

548
		put_anon_vma(rmap_item->anon_vma);
549
		rmap_item->address &= PAGE_MASK;
I
Izik Eidus 已提交
550

H
Hugh Dickins 已提交
551
	} else if (rmap_item->address & UNSTABLE_FLAG) {
I
Izik Eidus 已提交
552 553
		unsigned char age;
		/*
H
Hugh Dickins 已提交
554
		 * Usually ksmd can and must skip the rb_erase, because
I
Izik Eidus 已提交
555
		 * root_unstable_tree was already reset to RB_ROOT.
H
Hugh Dickins 已提交
556 557 558
		 * But be careful when an mm is exiting: do the rb_erase
		 * if this rmap_item was inserted by this scan, rather
		 * than left over from before.
I
Izik Eidus 已提交
559 560
		 */
		age = (unsigned char)(ksm_scan.seqnr - rmap_item->address);
561
		BUG_ON(age > 1);
I
Izik Eidus 已提交
562 563
		if (!age)
			rb_erase(&rmap_item->node, &root_unstable_tree);
564

565
		ksm_pages_unshared--;
566
		rmap_item->address &= PAGE_MASK;
I
Izik Eidus 已提交
567
	}
H
Hugh Dickins 已提交
568
out:
I
Izik Eidus 已提交
569 570 571 572
	cond_resched();		/* we're called from many long loops */
}

static void remove_trailing_rmap_items(struct mm_slot *mm_slot,
H
Hugh Dickins 已提交
573
				       struct rmap_item **rmap_list)
I
Izik Eidus 已提交
574
{
H
Hugh Dickins 已提交
575 576 577
	while (*rmap_list) {
		struct rmap_item *rmap_item = *rmap_list;
		*rmap_list = rmap_item->rmap_list;
I
Izik Eidus 已提交
578 579 580 581 582 583 584 585 586 587 588 589
		remove_rmap_item_from_tree(rmap_item);
		free_rmap_item(rmap_item);
	}
}

/*
 * Though it's very tempting to unmerge in_stable_tree(rmap_item)s rather
 * than check every pte of a given vma, the locking doesn't quite work for
 * that - an rmap_item is assigned to the stable tree after inserting ksm
 * page and upping mmap_sem.  Nor does it fit with the way we skip dup'ing
 * rmap_items from parent to child at fork time (so as not to waste time
 * if exit comes before the next scan reaches it).
H
Hugh Dickins 已提交
590 591 592 593 594
 *
 * Similarly, although we'd like to remove rmap_items (so updating counts
 * and freeing memory) when unmerging an area, it's easier to leave that
 * to the next pass of ksmd - consider, for example, how ksmd might be
 * in cmp_and_merge_page on one of the rmap_items we would be removing.
I
Izik Eidus 已提交
595
 */
H
Hugh Dickins 已提交
596 597
static int unmerge_ksm_pages(struct vm_area_struct *vma,
			     unsigned long start, unsigned long end)
I
Izik Eidus 已提交
598 599
{
	unsigned long addr;
H
Hugh Dickins 已提交
600
	int err = 0;
I
Izik Eidus 已提交
601

H
Hugh Dickins 已提交
602
	for (addr = start; addr < end && !err; addr += PAGE_SIZE) {
H
Hugh Dickins 已提交
603 604
		if (ksm_test_exit(vma->vm_mm))
			break;
H
Hugh Dickins 已提交
605 606 607 608 609 610
		if (signal_pending(current))
			err = -ERESTARTSYS;
		else
			err = break_ksm(vma, addr);
	}
	return err;
I
Izik Eidus 已提交
611 612
}

H
Hugh Dickins 已提交
613 614 615 616
#ifdef CONFIG_SYSFS
/*
 * Only called through the sysfs control interface:
 */
H
Hugh Dickins 已提交
617
static int unmerge_and_remove_all_rmap_items(void)
I
Izik Eidus 已提交
618 619 620 621
{
	struct mm_slot *mm_slot;
	struct mm_struct *mm;
	struct vm_area_struct *vma;
H
Hugh Dickins 已提交
622 623 624
	int err = 0;

	spin_lock(&ksm_mmlist_lock);
H
Hugh Dickins 已提交
625
	ksm_scan.mm_slot = list_entry(ksm_mm_head.mm_list.next,
H
Hugh Dickins 已提交
626 627
						struct mm_slot, mm_list);
	spin_unlock(&ksm_mmlist_lock);
I
Izik Eidus 已提交
628

H
Hugh Dickins 已提交
629 630
	for (mm_slot = ksm_scan.mm_slot;
			mm_slot != &ksm_mm_head; mm_slot = ksm_scan.mm_slot) {
I
Izik Eidus 已提交
631 632 633
		mm = mm_slot->mm;
		down_read(&mm->mmap_sem);
		for (vma = mm->mmap; vma; vma = vma->vm_next) {
H
Hugh Dickins 已提交
634 635
			if (ksm_test_exit(mm))
				break;
I
Izik Eidus 已提交
636 637
			if (!(vma->vm_flags & VM_MERGEABLE) || !vma->anon_vma)
				continue;
H
Hugh Dickins 已提交
638 639
			err = unmerge_ksm_pages(vma,
						vma->vm_start, vma->vm_end);
H
Hugh Dickins 已提交
640 641
			if (err)
				goto error;
I
Izik Eidus 已提交
642
		}
H
Hugh Dickins 已提交
643

H
Hugh Dickins 已提交
644
		remove_trailing_rmap_items(mm_slot, &mm_slot->rmap_list);
H
Hugh Dickins 已提交
645 646

		spin_lock(&ksm_mmlist_lock);
H
Hugh Dickins 已提交
647
		ksm_scan.mm_slot = list_entry(mm_slot->mm_list.next,
H
Hugh Dickins 已提交
648
						struct mm_slot, mm_list);
H
Hugh Dickins 已提交
649 650 651 652 653 654 655 656 657 658 659 660 661
		if (ksm_test_exit(mm)) {
			hlist_del(&mm_slot->link);
			list_del(&mm_slot->mm_list);
			spin_unlock(&ksm_mmlist_lock);

			free_mm_slot(mm_slot);
			clear_bit(MMF_VM_MERGEABLE, &mm->flags);
			up_read(&mm->mmap_sem);
			mmdrop(mm);
		} else {
			spin_unlock(&ksm_mmlist_lock);
			up_read(&mm->mmap_sem);
		}
I
Izik Eidus 已提交
662 663
	}

H
Hugh Dickins 已提交
664
	ksm_scan.seqnr = 0;
H
Hugh Dickins 已提交
665 666 667 668
	return 0;

error:
	up_read(&mm->mmap_sem);
I
Izik Eidus 已提交
669
	spin_lock(&ksm_mmlist_lock);
H
Hugh Dickins 已提交
670
	ksm_scan.mm_slot = &ksm_mm_head;
I
Izik Eidus 已提交
671
	spin_unlock(&ksm_mmlist_lock);
H
Hugh Dickins 已提交
672
	return err;
I
Izik Eidus 已提交
673
}
H
Hugh Dickins 已提交
674
#endif /* CONFIG_SYSFS */
I
Izik Eidus 已提交
675 676 677 678

static u32 calc_checksum(struct page *page)
{
	u32 checksum;
679
	void *addr = kmap_atomic(page);
I
Izik Eidus 已提交
680
	checksum = jhash2(addr, PAGE_SIZE / 4, 17);
681
	kunmap_atomic(addr);
I
Izik Eidus 已提交
682 683 684 685 686 687 688 689
	return checksum;
}

static int memcmp_pages(struct page *page1, struct page *page2)
{
	char *addr1, *addr2;
	int ret;

690 691
	addr1 = kmap_atomic(page1);
	addr2 = kmap_atomic(page2);
I
Izik Eidus 已提交
692
	ret = memcmp(addr1, addr2, PAGE_SIZE);
693 694
	kunmap_atomic(addr2);
	kunmap_atomic(addr1);
I
Izik Eidus 已提交
695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711
	return ret;
}

static inline int pages_identical(struct page *page1, struct page *page2)
{
	return !memcmp_pages(page1, page2);
}

static int write_protect_page(struct vm_area_struct *vma, struct page *page,
			      pte_t *orig_pte)
{
	struct mm_struct *mm = vma->vm_mm;
	unsigned long addr;
	pte_t *ptep;
	spinlock_t *ptl;
	int swapped;
	int err = -EFAULT;
712 713
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
I
Izik Eidus 已提交
714 715 716 717 718

	addr = page_address_in_vma(page, vma);
	if (addr == -EFAULT)
		goto out;

A
Andrea Arcangeli 已提交
719
	BUG_ON(PageTransCompound(page));
720 721 722 723 724

	mmun_start = addr;
	mmun_end   = addr + PAGE_SIZE;
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);

I
Izik Eidus 已提交
725 726
	ptep = page_check_address(page, mm, addr, &ptl, 0);
	if (!ptep)
727
		goto out_mn;
I
Izik Eidus 已提交
728

729
	if (pte_write(*ptep) || pte_dirty(*ptep)) {
I
Izik Eidus 已提交
730 731 732 733 734
		pte_t entry;

		swapped = PageSwapCache(page);
		flush_cache_page(vma, addr, page_to_pfn(page));
		/*
L
Lucas De Marchi 已提交
735
		 * Ok this is tricky, when get_user_pages_fast() run it doesn't
I
Izik Eidus 已提交
736 737 738 739 740 741 742 743 744 745 746 747
		 * take any lock, therefore the check that we are going to make
		 * with the pagecount against the mapcount is racey and
		 * O_DIRECT can happen right after the check.
		 * So we clear the pte and flush the tlb before the check
		 * this assure us that no O_DIRECT can happen after the check
		 * or in the middle of the check.
		 */
		entry = ptep_clear_flush(vma, addr, ptep);
		/*
		 * Check that no O_DIRECT or similar I/O is in progress on the
		 * page
		 */
748
		if (page_mapcount(page) + 1 + swapped != page_count(page)) {
749
			set_pte_at(mm, addr, ptep, entry);
I
Izik Eidus 已提交
750 751
			goto out_unlock;
		}
752 753 754
		if (pte_dirty(entry))
			set_page_dirty(page);
		entry = pte_mkclean(pte_wrprotect(entry));
I
Izik Eidus 已提交
755 756 757 758 759 760 761
		set_pte_at_notify(mm, addr, ptep, entry);
	}
	*orig_pte = *ptep;
	err = 0;

out_unlock:
	pte_unmap_unlock(ptep, ptl);
762 763
out_mn:
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
I
Izik Eidus 已提交
764 765 766 767 768 769
out:
	return err;
}

/**
 * replace_page - replace page in vma by new ksm page
770 771 772
 * @vma:      vma that holds the pte pointing to page
 * @page:     the page we are replacing by kpage
 * @kpage:    the ksm page we replace page by
I
Izik Eidus 已提交
773 774 775 776
 * @orig_pte: the original value of the pte
 *
 * Returns 0 on success, -EFAULT on failure.
 */
777 778
static int replace_page(struct vm_area_struct *vma, struct page *page,
			struct page *kpage, pte_t orig_pte)
I
Izik Eidus 已提交
779 780 781 782 783 784 785 786 787
{
	struct mm_struct *mm = vma->vm_mm;
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *ptep;
	spinlock_t *ptl;
	unsigned long addr;
	int err = -EFAULT;
788 789
	unsigned long mmun_start;	/* For mmu_notifiers */
	unsigned long mmun_end;		/* For mmu_notifiers */
I
Izik Eidus 已提交
790

791
	addr = page_address_in_vma(page, vma);
I
Izik Eidus 已提交
792 793 794 795 796 797 798 799 800 801 802 803
	if (addr == -EFAULT)
		goto out;

	pgd = pgd_offset(mm, addr);
	if (!pgd_present(*pgd))
		goto out;

	pud = pud_offset(pgd, addr);
	if (!pud_present(*pud))
		goto out;

	pmd = pmd_offset(pud, addr);
A
Andrea Arcangeli 已提交
804
	BUG_ON(pmd_trans_huge(*pmd));
I
Izik Eidus 已提交
805 806 807
	if (!pmd_present(*pmd))
		goto out;

808 809 810 811
	mmun_start = addr;
	mmun_end   = addr + PAGE_SIZE;
	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);

I
Izik Eidus 已提交
812 813 814
	ptep = pte_offset_map_lock(mm, pmd, addr, &ptl);
	if (!pte_same(*ptep, orig_pte)) {
		pte_unmap_unlock(ptep, ptl);
815
		goto out_mn;
I
Izik Eidus 已提交
816 817
	}

818
	get_page(kpage);
H
Hugh Dickins 已提交
819
	page_add_anon_rmap(kpage, vma, addr);
I
Izik Eidus 已提交
820 821 822

	flush_cache_page(vma, addr, pte_pfn(*ptep));
	ptep_clear_flush(vma, addr, ptep);
823
	set_pte_at_notify(mm, addr, ptep, mk_pte(kpage, vma->vm_page_prot));
I
Izik Eidus 已提交
824

825
	page_remove_rmap(page);
826 827
	if (!page_mapped(page))
		try_to_free_swap(page);
828
	put_page(page);
I
Izik Eidus 已提交
829 830 831

	pte_unmap_unlock(ptep, ptl);
	err = 0;
832 833
out_mn:
	mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
I
Izik Eidus 已提交
834 835 836 837
out:
	return err;
}

A
Andrea Arcangeli 已提交
838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864
static int page_trans_compound_anon_split(struct page *page)
{
	int ret = 0;
	struct page *transhuge_head = page_trans_compound_anon(page);
	if (transhuge_head) {
		/* Get the reference on the head to split it. */
		if (get_page_unless_zero(transhuge_head)) {
			/*
			 * Recheck we got the reference while the head
			 * was still anonymous.
			 */
			if (PageAnon(transhuge_head))
				ret = split_huge_page(transhuge_head);
			else
				/*
				 * Retry later if split_huge_page run
				 * from under us.
				 */
				ret = 1;
			put_page(transhuge_head);
		} else
			/* Retry later if split_huge_page run from under us. */
			ret = 1;
	}
	return ret;
}

I
Izik Eidus 已提交
865 866
/*
 * try_to_merge_one_page - take two pages and merge them into one
867 868
 * @vma: the vma that holds the pte pointing to page
 * @page: the PageAnon page that we want to replace with kpage
869 870
 * @kpage: the PageKsm page that we want to map instead of page,
 *         or NULL the first time when we want to use page as kpage.
I
Izik Eidus 已提交
871 872 873 874
 *
 * This function returns 0 if the pages were merged, -EFAULT otherwise.
 */
static int try_to_merge_one_page(struct vm_area_struct *vma,
875
				 struct page *page, struct page *kpage)
I
Izik Eidus 已提交
876 877 878 879
{
	pte_t orig_pte = __pte(0);
	int err = -EFAULT;

H
Hugh Dickins 已提交
880 881 882
	if (page == kpage)			/* ksm page forked */
		return 0;

I
Izik Eidus 已提交
883 884
	if (!(vma->vm_flags & VM_MERGEABLE))
		goto out;
A
Andrea Arcangeli 已提交
885 886 887
	if (PageTransCompound(page) && page_trans_compound_anon_split(page))
		goto out;
	BUG_ON(PageTransCompound(page));
888
	if (!PageAnon(page))
I
Izik Eidus 已提交
889 890 891 892 893 894 895 896 897
		goto out;

	/*
	 * We need the page lock to read a stable PageSwapCache in
	 * write_protect_page().  We use trylock_page() instead of
	 * lock_page() because we don't want to wait here - we
	 * prefer to continue scanning and merging different pages,
	 * then come back to this page when it is unlocked.
	 */
898
	if (!trylock_page(page))
899
		goto out;
I
Izik Eidus 已提交
900 901 902 903 904 905
	/*
	 * If this anonymous page is mapped only here, its pte may need
	 * to be write-protected.  If it's mapped elsewhere, all of its
	 * ptes are necessarily already write-protected.  But in either
	 * case, we need to lock and check page_count is not raised.
	 */
906 907 908 909 910 911 912 913 914 915 916 917 918
	if (write_protect_page(vma, page, &orig_pte) == 0) {
		if (!kpage) {
			/*
			 * While we hold page lock, upgrade page from
			 * PageAnon+anon_vma to PageKsm+NULL stable_node:
			 * stable_tree_insert() will update stable_node.
			 */
			set_page_stable_node(page, NULL);
			mark_page_accessed(page);
			err = 0;
		} else if (pages_identical(page, kpage))
			err = replace_page(vma, page, kpage, orig_pte);
	}
I
Izik Eidus 已提交
919

920
	if ((vma->vm_flags & VM_LOCKED) && kpage && !err) {
H
Hugh Dickins 已提交
921
		munlock_vma_page(page);
H
Hugh Dickins 已提交
922 923 924 925 926 927 928
		if (!PageMlocked(kpage)) {
			unlock_page(page);
			lock_page(kpage);
			mlock_vma_page(kpage);
			page = kpage;		/* for final unlock */
		}
	}
H
Hugh Dickins 已提交
929

930
	unlock_page(page);
I
Izik Eidus 已提交
931 932 933 934
out:
	return err;
}

H
Hugh Dickins 已提交
935 936 937
/*
 * try_to_merge_with_ksm_page - like try_to_merge_two_pages,
 * but no new kernel page is allocated: kpage must already be a ksm page.
938 939
 *
 * This function returns 0 if the pages were merged, -EFAULT otherwise.
H
Hugh Dickins 已提交
940
 */
941 942
static int try_to_merge_with_ksm_page(struct rmap_item *rmap_item,
				      struct page *page, struct page *kpage)
H
Hugh Dickins 已提交
943
{
944
	struct mm_struct *mm = rmap_item->mm;
H
Hugh Dickins 已提交
945 946 947
	struct vm_area_struct *vma;
	int err = -EFAULT;

948 949
	down_read(&mm->mmap_sem);
	if (ksm_test_exit(mm))
H
Hugh Dickins 已提交
950
		goto out;
951 952
	vma = find_vma(mm, rmap_item->address);
	if (!vma || vma->vm_start > rmap_item->address)
H
Hugh Dickins 已提交
953 954
		goto out;

955
	err = try_to_merge_one_page(vma, page, kpage);
H
Hugh Dickins 已提交
956 957 958 959
	if (err)
		goto out;

	/* Must get reference to anon_vma while still holding mmap_sem */
960 961
	rmap_item->anon_vma = vma->anon_vma;
	get_anon_vma(vma->anon_vma);
H
Hugh Dickins 已提交
962
out:
963
	up_read(&mm->mmap_sem);
H
Hugh Dickins 已提交
964 965 966
	return err;
}

I
Izik Eidus 已提交
967 968 969 970
/*
 * try_to_merge_two_pages - take two identical pages and prepare them
 * to be merged into one page.
 *
971 972
 * This function returns the kpage if we successfully merged two identical
 * pages into one ksm page, NULL otherwise.
I
Izik Eidus 已提交
973
 *
974
 * Note that this function upgrades page to ksm page: if one of the pages
I
Izik Eidus 已提交
975 976
 * is already a ksm page, try_to_merge_with_ksm_page should be used.
 */
977 978 979 980
static struct page *try_to_merge_two_pages(struct rmap_item *rmap_item,
					   struct page *page,
					   struct rmap_item *tree_rmap_item,
					   struct page *tree_page)
I
Izik Eidus 已提交
981
{
982
	int err;
I
Izik Eidus 已提交
983

984
	err = try_to_merge_with_ksm_page(rmap_item, page, NULL);
I
Izik Eidus 已提交
985
	if (!err) {
986
		err = try_to_merge_with_ksm_page(tree_rmap_item,
987
							tree_page, page);
I
Izik Eidus 已提交
988
		/*
H
Hugh Dickins 已提交
989 990
		 * If that fails, we have a ksm page with only one pte
		 * pointing to it: so break it.
I
Izik Eidus 已提交
991
		 */
H
Hugh Dickins 已提交
992
		if (err)
993
			break_cow(rmap_item);
I
Izik Eidus 已提交
994
	}
995
	return err ? NULL : page;
I
Izik Eidus 已提交
996 997 998
}

/*
999
 * stable_tree_search - search for page inside the stable tree
I
Izik Eidus 已提交
1000 1001 1002 1003
 *
 * This function checks if there is a page inside the stable tree
 * with identical content to the page that we are scanning right now.
 *
H
Hugh Dickins 已提交
1004
 * This function returns the stable tree node of identical content if found,
I
Izik Eidus 已提交
1005 1006
 * NULL otherwise.
 */
1007
static struct page *stable_tree_search(struct page *page)
I
Izik Eidus 已提交
1008 1009
{
	struct rb_node *node = root_stable_tree.rb_node;
H
Hugh Dickins 已提交
1010
	struct stable_node *stable_node;
I
Izik Eidus 已提交
1011

1012 1013 1014
	stable_node = page_stable_node(page);
	if (stable_node) {			/* ksm page forked */
		get_page(page);
1015
		return page;
1016 1017
	}

I
Izik Eidus 已提交
1018
	while (node) {
H
Hugh Dickins 已提交
1019
		struct page *tree_page;
I
Izik Eidus 已提交
1020 1021
		int ret;

1022
		cond_resched();
H
Hugh Dickins 已提交
1023
		stable_node = rb_entry(node, struct stable_node, node);
H
Hugh Dickins 已提交
1024 1025 1026
		tree_page = get_ksm_page(stable_node);
		if (!tree_page)
			return NULL;
I
Izik Eidus 已提交
1027

H
Hugh Dickins 已提交
1028
		ret = memcmp_pages(page, tree_page);
I
Izik Eidus 已提交
1029

H
Hugh Dickins 已提交
1030 1031
		if (ret < 0) {
			put_page(tree_page);
I
Izik Eidus 已提交
1032
			node = node->rb_left;
H
Hugh Dickins 已提交
1033 1034
		} else if (ret > 0) {
			put_page(tree_page);
I
Izik Eidus 已提交
1035
			node = node->rb_right;
H
Hugh Dickins 已提交
1036
		} else
1037
			return tree_page;
I
Izik Eidus 已提交
1038 1039 1040 1041 1042 1043 1044 1045 1046
	}

	return NULL;
}

/*
 * stable_tree_insert - insert rmap_item pointing to new ksm page
 * into the stable tree.
 *
H
Hugh Dickins 已提交
1047 1048
 * This function returns the stable tree node just allocated on success,
 * NULL otherwise.
I
Izik Eidus 已提交
1049
 */
H
Hugh Dickins 已提交
1050
static struct stable_node *stable_tree_insert(struct page *kpage)
I
Izik Eidus 已提交
1051 1052 1053
{
	struct rb_node **new = &root_stable_tree.rb_node;
	struct rb_node *parent = NULL;
H
Hugh Dickins 已提交
1054
	struct stable_node *stable_node;
I
Izik Eidus 已提交
1055 1056

	while (*new) {
H
Hugh Dickins 已提交
1057
		struct page *tree_page;
I
Izik Eidus 已提交
1058 1059
		int ret;

1060
		cond_resched();
H
Hugh Dickins 已提交
1061
		stable_node = rb_entry(*new, struct stable_node, node);
H
Hugh Dickins 已提交
1062 1063 1064
		tree_page = get_ksm_page(stable_node);
		if (!tree_page)
			return NULL;
I
Izik Eidus 已提交
1065

H
Hugh Dickins 已提交
1066 1067
		ret = memcmp_pages(kpage, tree_page);
		put_page(tree_page);
I
Izik Eidus 已提交
1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083

		parent = *new;
		if (ret < 0)
			new = &parent->rb_left;
		else if (ret > 0)
			new = &parent->rb_right;
		else {
			/*
			 * It is not a bug that stable_tree_search() didn't
			 * find this node: because at that time our page was
			 * not yet write-protected, so may have changed since.
			 */
			return NULL;
		}
	}

H
Hugh Dickins 已提交
1084 1085 1086
	stable_node = alloc_stable_node();
	if (!stable_node)
		return NULL;
I
Izik Eidus 已提交
1087

H
Hugh Dickins 已提交
1088 1089 1090 1091 1092
	rb_link_node(&stable_node->node, parent, new);
	rb_insert_color(&stable_node->node, &root_stable_tree);

	INIT_HLIST_HEAD(&stable_node->hlist);

1093
	stable_node->kpfn = page_to_pfn(kpage);
1094 1095
	set_page_stable_node(kpage, stable_node);

H
Hugh Dickins 已提交
1096
	return stable_node;
I
Izik Eidus 已提交
1097 1098 1099
}

/*
1100 1101
 * unstable_tree_search_insert - search for identical page,
 * else insert rmap_item into the unstable tree.
I
Izik Eidus 已提交
1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112
 *
 * This function searches for a page in the unstable tree identical to the
 * page currently being scanned; and if no identical page is found in the
 * tree, we insert rmap_item as a new object into the unstable tree.
 *
 * This function returns pointer to rmap_item found to be identical
 * to the currently scanned page, NULL otherwise.
 *
 * This function does both searching and inserting, because they share
 * the same walking algorithm in an rbtree.
 */
1113 1114 1115 1116 1117
static
struct rmap_item *unstable_tree_search_insert(struct rmap_item *rmap_item,
					      struct page *page,
					      struct page **tree_pagep)

I
Izik Eidus 已提交
1118 1119 1120 1121 1122 1123
{
	struct rb_node **new = &root_unstable_tree.rb_node;
	struct rb_node *parent = NULL;

	while (*new) {
		struct rmap_item *tree_rmap_item;
1124
		struct page *tree_page;
I
Izik Eidus 已提交
1125 1126
		int ret;

H
Hugh Dickins 已提交
1127
		cond_resched();
I
Izik Eidus 已提交
1128
		tree_rmap_item = rb_entry(*new, struct rmap_item, node);
1129
		tree_page = get_mergeable_page(tree_rmap_item);
1130
		if (IS_ERR_OR_NULL(tree_page))
I
Izik Eidus 已提交
1131 1132 1133
			return NULL;

		/*
1134
		 * Don't substitute a ksm page for a forked page.
I
Izik Eidus 已提交
1135
		 */
1136 1137
		if (page == tree_page) {
			put_page(tree_page);
I
Izik Eidus 已提交
1138 1139 1140
			return NULL;
		}

1141
		ret = memcmp_pages(page, tree_page);
I
Izik Eidus 已提交
1142 1143 1144

		parent = *new;
		if (ret < 0) {
1145
			put_page(tree_page);
I
Izik Eidus 已提交
1146 1147
			new = &parent->rb_left;
		} else if (ret > 0) {
1148
			put_page(tree_page);
I
Izik Eidus 已提交
1149 1150
			new = &parent->rb_right;
		} else {
1151
			*tree_pagep = tree_page;
I
Izik Eidus 已提交
1152 1153 1154 1155
			return tree_rmap_item;
		}
	}

H
Hugh Dickins 已提交
1156
	rmap_item->address |= UNSTABLE_FLAG;
I
Izik Eidus 已提交
1157 1158 1159 1160
	rmap_item->address |= (ksm_scan.seqnr & SEQNR_MASK);
	rb_link_node(&rmap_item->node, parent, new);
	rb_insert_color(&rmap_item->node, &root_unstable_tree);

1161
	ksm_pages_unshared++;
I
Izik Eidus 已提交
1162 1163 1164 1165 1166 1167 1168 1169 1170
	return NULL;
}

/*
 * stable_tree_append - add another rmap_item to the linked list of
 * rmap_items hanging off a given node of the stable tree, all sharing
 * the same ksm page.
 */
static void stable_tree_append(struct rmap_item *rmap_item,
H
Hugh Dickins 已提交
1171
			       struct stable_node *stable_node)
I
Izik Eidus 已提交
1172
{
H
Hugh Dickins 已提交
1173
	rmap_item->head = stable_node;
I
Izik Eidus 已提交
1174
	rmap_item->address |= STABLE_FLAG;
H
Hugh Dickins 已提交
1175
	hlist_add_head(&rmap_item->hlist, &stable_node->hlist);
H
Hugh Dickins 已提交
1176

H
Hugh Dickins 已提交
1177 1178 1179 1180
	if (rmap_item->hlist.next)
		ksm_pages_sharing++;
	else
		ksm_pages_shared++;
I
Izik Eidus 已提交
1181 1182 1183
}

/*
H
Hugh Dickins 已提交
1184 1185 1186 1187
 * cmp_and_merge_page - first see if page can be merged into the stable tree;
 * if not, compare checksum to previous and if it's the same, see if page can
 * be inserted into the unstable tree, or merged with a page already there and
 * both transferred to the stable tree.
I
Izik Eidus 已提交
1188 1189 1190 1191 1192 1193 1194
 *
 * @page: the page that we are searching identical page to.
 * @rmap_item: the reverse mapping into the virtual address of this page
 */
static void cmp_and_merge_page(struct page *page, struct rmap_item *rmap_item)
{
	struct rmap_item *tree_rmap_item;
1195
	struct page *tree_page = NULL;
H
Hugh Dickins 已提交
1196
	struct stable_node *stable_node;
1197
	struct page *kpage;
I
Izik Eidus 已提交
1198 1199 1200
	unsigned int checksum;
	int err;

1201
	remove_rmap_item_from_tree(rmap_item);
I
Izik Eidus 已提交
1202 1203

	/* We first start with searching the page inside the stable tree */
1204 1205
	kpage = stable_tree_search(page);
	if (kpage) {
1206
		err = try_to_merge_with_ksm_page(rmap_item, page, kpage);
I
Izik Eidus 已提交
1207 1208 1209 1210 1211
		if (!err) {
			/*
			 * The page was successfully merged:
			 * add its rmap_item to the stable tree.
			 */
H
Hugh Dickins 已提交
1212
			lock_page(kpage);
1213
			stable_tree_append(rmap_item, page_stable_node(kpage));
H
Hugh Dickins 已提交
1214
			unlock_page(kpage);
I
Izik Eidus 已提交
1215
		}
1216
		put_page(kpage);
I
Izik Eidus 已提交
1217 1218 1219 1220
		return;
	}

	/*
H
Hugh Dickins 已提交
1221 1222 1223 1224
	 * If the hash value of the page has changed from the last time
	 * we calculated it, this page is changing frequently: therefore we
	 * don't want to insert it in the unstable tree, and we don't want
	 * to waste our time searching for something identical to it there.
I
Izik Eidus 已提交
1225 1226 1227 1228 1229 1230 1231
	 */
	checksum = calc_checksum(page);
	if (rmap_item->oldchecksum != checksum) {
		rmap_item->oldchecksum = checksum;
		return;
	}

1232 1233
	tree_rmap_item =
		unstable_tree_search_insert(rmap_item, page, &tree_page);
I
Izik Eidus 已提交
1234
	if (tree_rmap_item) {
1235 1236 1237
		kpage = try_to_merge_two_pages(rmap_item, page,
						tree_rmap_item, tree_page);
		put_page(tree_page);
I
Izik Eidus 已提交
1238 1239 1240 1241 1242
		/*
		 * As soon as we merge this page, we want to remove the
		 * rmap_item of the page we have merged with from the unstable
		 * tree, and insert it instead as new node in the stable tree.
		 */
1243
		if (kpage) {
1244
			remove_rmap_item_from_tree(tree_rmap_item);
1245

H
Hugh Dickins 已提交
1246
			lock_page(kpage);
H
Hugh Dickins 已提交
1247 1248 1249 1250 1251
			stable_node = stable_tree_insert(kpage);
			if (stable_node) {
				stable_tree_append(tree_rmap_item, stable_node);
				stable_tree_append(rmap_item, stable_node);
			}
H
Hugh Dickins 已提交
1252
			unlock_page(kpage);
H
Hugh Dickins 已提交
1253

I
Izik Eidus 已提交
1254 1255 1256 1257 1258 1259
			/*
			 * If we fail to insert the page into the stable tree,
			 * we will have 2 virtual addresses that are pointing
			 * to a ksm page left outside the stable tree,
			 * in which case we need to break_cow on both.
			 */
H
Hugh Dickins 已提交
1260
			if (!stable_node) {
1261 1262
				break_cow(tree_rmap_item);
				break_cow(rmap_item);
I
Izik Eidus 已提交
1263 1264 1265 1266 1267 1268
			}
		}
	}
}

static struct rmap_item *get_next_rmap_item(struct mm_slot *mm_slot,
H
Hugh Dickins 已提交
1269
					    struct rmap_item **rmap_list,
I
Izik Eidus 已提交
1270 1271 1272 1273
					    unsigned long addr)
{
	struct rmap_item *rmap_item;

H
Hugh Dickins 已提交
1274 1275
	while (*rmap_list) {
		rmap_item = *rmap_list;
1276
		if ((rmap_item->address & PAGE_MASK) == addr)
I
Izik Eidus 已提交
1277 1278 1279
			return rmap_item;
		if (rmap_item->address > addr)
			break;
H
Hugh Dickins 已提交
1280
		*rmap_list = rmap_item->rmap_list;
I
Izik Eidus 已提交
1281 1282 1283 1284 1285 1286 1287 1288 1289
		remove_rmap_item_from_tree(rmap_item);
		free_rmap_item(rmap_item);
	}

	rmap_item = alloc_rmap_item();
	if (rmap_item) {
		/* It has already been zeroed */
		rmap_item->mm = mm_slot->mm;
		rmap_item->address = addr;
H
Hugh Dickins 已提交
1290 1291
		rmap_item->rmap_list = *rmap_list;
		*rmap_list = rmap_item;
I
Izik Eidus 已提交
1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307
	}
	return rmap_item;
}

static struct rmap_item *scan_get_next_rmap_item(struct page **page)
{
	struct mm_struct *mm;
	struct mm_slot *slot;
	struct vm_area_struct *vma;
	struct rmap_item *rmap_item;

	if (list_empty(&ksm_mm_head.mm_list))
		return NULL;

	slot = ksm_scan.mm_slot;
	if (slot == &ksm_mm_head) {
H
Hugh Dickins 已提交
1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319
		/*
		 * A number of pages can hang around indefinitely on per-cpu
		 * pagevecs, raised page count preventing write_protect_page
		 * from merging them.  Though it doesn't really matter much,
		 * it is puzzling to see some stuck in pages_volatile until
		 * other activity jostles them out, and they also prevented
		 * LTP's KSM test from succeeding deterministically; so drain
		 * them here (here rather than on entry to ksm_do_scan(),
		 * so we don't IPI too often when pages_to_scan is set low).
		 */
		lru_add_drain_all();

I
Izik Eidus 已提交
1320 1321 1322 1323 1324 1325
		root_unstable_tree = RB_ROOT;

		spin_lock(&ksm_mmlist_lock);
		slot = list_entry(slot->mm_list.next, struct mm_slot, mm_list);
		ksm_scan.mm_slot = slot;
		spin_unlock(&ksm_mmlist_lock);
1326 1327 1328 1329 1330 1331
		/*
		 * Although we tested list_empty() above, a racing __ksm_exit
		 * of the last mm on the list may have removed it since then.
		 */
		if (slot == &ksm_mm_head)
			return NULL;
I
Izik Eidus 已提交
1332 1333
next_mm:
		ksm_scan.address = 0;
H
Hugh Dickins 已提交
1334
		ksm_scan.rmap_list = &slot->rmap_list;
I
Izik Eidus 已提交
1335 1336 1337 1338
	}

	mm = slot->mm;
	down_read(&mm->mmap_sem);
H
Hugh Dickins 已提交
1339 1340 1341 1342 1343 1344
	if (ksm_test_exit(mm))
		vma = NULL;
	else
		vma = find_vma(mm, ksm_scan.address);

	for (; vma; vma = vma->vm_next) {
I
Izik Eidus 已提交
1345 1346 1347 1348 1349 1350 1351 1352
		if (!(vma->vm_flags & VM_MERGEABLE))
			continue;
		if (ksm_scan.address < vma->vm_start)
			ksm_scan.address = vma->vm_start;
		if (!vma->anon_vma)
			ksm_scan.address = vma->vm_end;

		while (ksm_scan.address < vma->vm_end) {
H
Hugh Dickins 已提交
1353 1354
			if (ksm_test_exit(mm))
				break;
I
Izik Eidus 已提交
1355
			*page = follow_page(vma, ksm_scan.address, FOLL_GET);
1356 1357 1358 1359 1360
			if (IS_ERR_OR_NULL(*page)) {
				ksm_scan.address += PAGE_SIZE;
				cond_resched();
				continue;
			}
A
Andrea Arcangeli 已提交
1361 1362
			if (PageAnon(*page) ||
			    page_trans_compound_anon(*page)) {
I
Izik Eidus 已提交
1363 1364 1365
				flush_anon_page(vma, *page, ksm_scan.address);
				flush_dcache_page(*page);
				rmap_item = get_next_rmap_item(slot,
H
Hugh Dickins 已提交
1366
					ksm_scan.rmap_list, ksm_scan.address);
I
Izik Eidus 已提交
1367
				if (rmap_item) {
H
Hugh Dickins 已提交
1368 1369
					ksm_scan.rmap_list =
							&rmap_item->rmap_list;
I
Izik Eidus 已提交
1370 1371 1372 1373 1374 1375
					ksm_scan.address += PAGE_SIZE;
				} else
					put_page(*page);
				up_read(&mm->mmap_sem);
				return rmap_item;
			}
1376
			put_page(*page);
I
Izik Eidus 已提交
1377 1378 1379 1380 1381
			ksm_scan.address += PAGE_SIZE;
			cond_resched();
		}
	}

H
Hugh Dickins 已提交
1382 1383
	if (ksm_test_exit(mm)) {
		ksm_scan.address = 0;
H
Hugh Dickins 已提交
1384
		ksm_scan.rmap_list = &slot->rmap_list;
H
Hugh Dickins 已提交
1385
	}
I
Izik Eidus 已提交
1386 1387 1388 1389
	/*
	 * Nuke all the rmap_items that are above this current rmap:
	 * because there were no VM_MERGEABLE vmas with such addresses.
	 */
H
Hugh Dickins 已提交
1390
	remove_trailing_rmap_items(slot, ksm_scan.rmap_list);
I
Izik Eidus 已提交
1391 1392

	spin_lock(&ksm_mmlist_lock);
1393 1394 1395 1396 1397 1398 1399
	ksm_scan.mm_slot = list_entry(slot->mm_list.next,
						struct mm_slot, mm_list);
	if (ksm_scan.address == 0) {
		/*
		 * We've completed a full scan of all vmas, holding mmap_sem
		 * throughout, and found no VM_MERGEABLE: so do the same as
		 * __ksm_exit does to remove this mm from all our lists now.
H
Hugh Dickins 已提交
1400 1401 1402 1403
		 * This applies either when cleaning up after __ksm_exit
		 * (but beware: we can reach here even before __ksm_exit),
		 * or when all VM_MERGEABLE areas have been unmapped (and
		 * mmap_sem then protects against race with MADV_MERGEABLE).
1404 1405 1406
		 */
		hlist_del(&slot->link);
		list_del(&slot->mm_list);
H
Hugh Dickins 已提交
1407 1408
		spin_unlock(&ksm_mmlist_lock);

1409 1410
		free_mm_slot(slot);
		clear_bit(MMF_VM_MERGEABLE, &mm->flags);
H
Hugh Dickins 已提交
1411 1412 1413 1414 1415
		up_read(&mm->mmap_sem);
		mmdrop(mm);
	} else {
		spin_unlock(&ksm_mmlist_lock);
		up_read(&mm->mmap_sem);
1416
	}
I
Izik Eidus 已提交
1417 1418

	/* Repeat until we've completed scanning the whole list */
1419
	slot = ksm_scan.mm_slot;
I
Izik Eidus 已提交
1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433
	if (slot != &ksm_mm_head)
		goto next_mm;

	ksm_scan.seqnr++;
	return NULL;
}

/**
 * ksm_do_scan  - the ksm scanner main worker function.
 * @scan_npages - number of pages we want to scan before we return.
 */
static void ksm_do_scan(unsigned int scan_npages)
{
	struct rmap_item *rmap_item;
1434
	struct page *uninitialized_var(page);
I
Izik Eidus 已提交
1435

1436
	while (scan_npages-- && likely(!freezing(current))) {
I
Izik Eidus 已提交
1437 1438 1439 1440 1441 1442 1443 1444 1445 1446
		cond_resched();
		rmap_item = scan_get_next_rmap_item(&page);
		if (!rmap_item)
			return;
		if (!PageKsm(page) || !in_stable_tree(rmap_item))
			cmp_and_merge_page(page, rmap_item);
		put_page(page);
	}
}

H
Hugh Dickins 已提交
1447 1448 1449 1450 1451
static int ksmd_should_run(void)
{
	return (ksm_run & KSM_RUN_MERGE) && !list_empty(&ksm_mm_head.mm_list);
}

I
Izik Eidus 已提交
1452 1453
static int ksm_scan_thread(void *nothing)
{
1454
	set_freezable();
I
Izik Eidus 已提交
1455
	set_user_nice(current, 5);
I
Izik Eidus 已提交
1456 1457

	while (!kthread_should_stop()) {
H
Hugh Dickins 已提交
1458 1459
		mutex_lock(&ksm_thread_mutex);
		if (ksmd_should_run())
I
Izik Eidus 已提交
1460
			ksm_do_scan(ksm_thread_pages_to_scan);
H
Hugh Dickins 已提交
1461 1462
		mutex_unlock(&ksm_thread_mutex);

1463 1464
		try_to_freeze();

H
Hugh Dickins 已提交
1465
		if (ksmd_should_run()) {
I
Izik Eidus 已提交
1466 1467 1468
			schedule_timeout_interruptible(
				msecs_to_jiffies(ksm_thread_sleep_millisecs));
		} else {
1469
			wait_event_freezable(ksm_thread_wait,
H
Hugh Dickins 已提交
1470
				ksmd_should_run() || kthread_should_stop());
I
Izik Eidus 已提交
1471 1472 1473 1474 1475
		}
	}
	return 0;
}

H
Hugh Dickins 已提交
1476 1477 1478 1479
int ksm_madvise(struct vm_area_struct *vma, unsigned long start,
		unsigned long end, int advice, unsigned long *vm_flags)
{
	struct mm_struct *mm = vma->vm_mm;
H
Hugh Dickins 已提交
1480
	int err;
H
Hugh Dickins 已提交
1481 1482 1483 1484 1485 1486 1487 1488

	switch (advice) {
	case MADV_MERGEABLE:
		/*
		 * Be somewhat over-protective for now!
		 */
		if (*vm_flags & (VM_MERGEABLE | VM_SHARED  | VM_MAYSHARE   |
				 VM_PFNMAP    | VM_IO      | VM_DONTEXPAND |
1489
				 VM_HUGETLB | VM_NONLINEAR | VM_MIXEDMAP))
H
Hugh Dickins 已提交
1490 1491
			return 0;		/* just ignore the advice */

1492 1493 1494 1495 1496
#ifdef VM_SAO
		if (*vm_flags & VM_SAO)
			return 0;
#endif

H
Hugh Dickins 已提交
1497 1498 1499 1500 1501
		if (!test_bit(MMF_VM_MERGEABLE, &mm->flags)) {
			err = __ksm_enter(mm);
			if (err)
				return err;
		}
H
Hugh Dickins 已提交
1502 1503 1504 1505 1506 1507 1508 1509

		*vm_flags |= VM_MERGEABLE;
		break;

	case MADV_UNMERGEABLE:
		if (!(*vm_flags & VM_MERGEABLE))
			return 0;		/* just ignore the advice */

H
Hugh Dickins 已提交
1510 1511 1512 1513 1514
		if (vma->anon_vma) {
			err = unmerge_ksm_pages(vma, start, end);
			if (err)
				return err;
		}
H
Hugh Dickins 已提交
1515 1516 1517 1518 1519 1520 1521 1522 1523 1524

		*vm_flags &= ~VM_MERGEABLE;
		break;
	}

	return 0;
}

int __ksm_enter(struct mm_struct *mm)
{
H
Hugh Dickins 已提交
1525 1526 1527 1528
	struct mm_slot *mm_slot;
	int needs_wakeup;

	mm_slot = alloc_mm_slot();
I
Izik Eidus 已提交
1529 1530 1531
	if (!mm_slot)
		return -ENOMEM;

H
Hugh Dickins 已提交
1532 1533 1534
	/* Check ksm_run too?  Would need tighter locking */
	needs_wakeup = list_empty(&ksm_mm_head.mm_list);

I
Izik Eidus 已提交
1535 1536 1537 1538 1539 1540 1541 1542 1543 1544
	spin_lock(&ksm_mmlist_lock);
	insert_to_mm_slots_hash(mm, mm_slot);
	/*
	 * Insert just behind the scanning cursor, to let the area settle
	 * down a little; when fork is followed by immediate exec, we don't
	 * want ksmd to waste time setting up and tearing down an rmap_list.
	 */
	list_add_tail(&mm_slot->mm_list, &ksm_scan.mm_slot->mm_list);
	spin_unlock(&ksm_mmlist_lock);

H
Hugh Dickins 已提交
1545
	set_bit(MMF_VM_MERGEABLE, &mm->flags);
H
Hugh Dickins 已提交
1546
	atomic_inc(&mm->mm_count);
H
Hugh Dickins 已提交
1547 1548 1549 1550

	if (needs_wakeup)
		wake_up_interruptible(&ksm_thread_wait);

H
Hugh Dickins 已提交
1551 1552 1553
	return 0;
}

1554
void __ksm_exit(struct mm_struct *mm)
H
Hugh Dickins 已提交
1555
{
1556
	struct mm_slot *mm_slot;
H
Hugh Dickins 已提交
1557
	int easy_to_free = 0;
1558

I
Izik Eidus 已提交
1559
	/*
H
Hugh Dickins 已提交
1560 1561 1562 1563 1564 1565
	 * This process is exiting: if it's straightforward (as is the
	 * case when ksmd was never running), free mm_slot immediately.
	 * But if it's at the cursor or has rmap_items linked to it, use
	 * mmap_sem to synchronize with any break_cows before pagetables
	 * are freed, and leave the mm_slot on the list for ksmd to free.
	 * Beware: ksm may already have noticed it exiting and freed the slot.
I
Izik Eidus 已提交
1566
	 */
H
Hugh Dickins 已提交
1567

1568 1569
	spin_lock(&ksm_mmlist_lock);
	mm_slot = get_mm_slot(mm);
H
Hugh Dickins 已提交
1570
	if (mm_slot && ksm_scan.mm_slot != mm_slot) {
H
Hugh Dickins 已提交
1571
		if (!mm_slot->rmap_list) {
H
Hugh Dickins 已提交
1572 1573 1574 1575 1576 1577 1578
			hlist_del(&mm_slot->link);
			list_del(&mm_slot->mm_list);
			easy_to_free = 1;
		} else {
			list_move(&mm_slot->mm_list,
				  &ksm_scan.mm_slot->mm_list);
		}
1579 1580 1581
	}
	spin_unlock(&ksm_mmlist_lock);

H
Hugh Dickins 已提交
1582 1583 1584 1585 1586 1587 1588 1589
	if (easy_to_free) {
		free_mm_slot(mm_slot);
		clear_bit(MMF_VM_MERGEABLE, &mm->flags);
		mmdrop(mm);
	} else if (mm_slot) {
		down_write(&mm->mmap_sem);
		up_write(&mm->mmap_sem);
	}
I
Izik Eidus 已提交
1590 1591
}

H
Hugh Dickins 已提交
1592 1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 1603 1604 1605
struct page *ksm_does_need_to_copy(struct page *page,
			struct vm_area_struct *vma, unsigned long address)
{
	struct page *new_page;

	new_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, address);
	if (new_page) {
		copy_user_highpage(new_page, page, address, vma);

		SetPageDirty(new_page);
		__SetPageUptodate(new_page);
		SetPageSwapBacked(new_page);
		__set_page_locked(new_page);

1606
		if (!mlocked_vma_newpage(vma, new_page))
H
Hugh Dickins 已提交
1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622
			lru_cache_add_lru(new_page, LRU_ACTIVE_ANON);
		else
			add_page_to_unevictable_list(new_page);
	}

	return new_page;
}

int page_referenced_ksm(struct page *page, struct mem_cgroup *memcg,
			unsigned long *vm_flags)
{
	struct stable_node *stable_node;
	struct rmap_item *rmap_item;
	struct hlist_node *hlist;
	unsigned int mapcount = page_mapcount(page);
	int referenced = 0;
H
Hugh Dickins 已提交
1623
	int search_new_forks = 0;
H
Hugh Dickins 已提交
1624 1625 1626 1627 1628 1629 1630

	VM_BUG_ON(!PageKsm(page));
	VM_BUG_ON(!PageLocked(page));

	stable_node = page_stable_node(page);
	if (!stable_node)
		return 0;
H
Hugh Dickins 已提交
1631
again:
H
Hugh Dickins 已提交
1632
	hlist_for_each_entry(rmap_item, hlist, &stable_node->hlist, hlist) {
H
Hugh Dickins 已提交
1633
		struct anon_vma *anon_vma = rmap_item->anon_vma;
1634
		struct anon_vma_chain *vmac;
H
Hugh Dickins 已提交
1635
		struct vm_area_struct *vma;
H
Hugh Dickins 已提交
1636

1637
		anon_vma_lock(anon_vma);
1638 1639
		anon_vma_interval_tree_foreach(vmac, &anon_vma->rb_root,
					       0, ULONG_MAX) {
1640
			vma = vmac->vma;
H
Hugh Dickins 已提交
1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654
			if (rmap_item->address < vma->vm_start ||
			    rmap_item->address >= vma->vm_end)
				continue;
			/*
			 * Initially we examine only the vma which covers this
			 * rmap_item; but later, if there is still work to do,
			 * we examine covering vmas in other mms: in case they
			 * were forked from the original since ksmd passed.
			 */
			if ((rmap_item->mm == vma->vm_mm) == search_new_forks)
				continue;

			if (memcg && !mm_match_cgroup(vma->vm_mm, memcg))
				continue;
H
Hugh Dickins 已提交
1655

H
Hugh Dickins 已提交
1656
			referenced += page_referenced_one(page, vma,
H
Hugh Dickins 已提交
1657
				rmap_item->address, &mapcount, vm_flags);
H
Hugh Dickins 已提交
1658 1659 1660
			if (!search_new_forks || !mapcount)
				break;
		}
1661
		anon_vma_unlock(anon_vma);
H
Hugh Dickins 已提交
1662 1663 1664
		if (!mapcount)
			goto out;
	}
H
Hugh Dickins 已提交
1665 1666
	if (!search_new_forks++)
		goto again;
H
Hugh Dickins 已提交
1667 1668 1669 1670 1671 1672 1673 1674 1675 1676
out:
	return referenced;
}

int try_to_unmap_ksm(struct page *page, enum ttu_flags flags)
{
	struct stable_node *stable_node;
	struct hlist_node *hlist;
	struct rmap_item *rmap_item;
	int ret = SWAP_AGAIN;
H
Hugh Dickins 已提交
1677
	int search_new_forks = 0;
H
Hugh Dickins 已提交
1678 1679 1680 1681 1682 1683 1684

	VM_BUG_ON(!PageKsm(page));
	VM_BUG_ON(!PageLocked(page));

	stable_node = page_stable_node(page);
	if (!stable_node)
		return SWAP_FAIL;
H
Hugh Dickins 已提交
1685
again:
H
Hugh Dickins 已提交
1686
	hlist_for_each_entry(rmap_item, hlist, &stable_node->hlist, hlist) {
H
Hugh Dickins 已提交
1687
		struct anon_vma *anon_vma = rmap_item->anon_vma;
1688
		struct anon_vma_chain *vmac;
H
Hugh Dickins 已提交
1689
		struct vm_area_struct *vma;
H
Hugh Dickins 已提交
1690

1691
		anon_vma_lock(anon_vma);
1692 1693
		anon_vma_interval_tree_foreach(vmac, &anon_vma->rb_root,
					       0, ULONG_MAX) {
1694
			vma = vmac->vma;
H
Hugh Dickins 已提交
1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709
			if (rmap_item->address < vma->vm_start ||
			    rmap_item->address >= vma->vm_end)
				continue;
			/*
			 * Initially we examine only the vma which covers this
			 * rmap_item; but later, if there is still work to do,
			 * we examine covering vmas in other mms: in case they
			 * were forked from the original since ksmd passed.
			 */
			if ((rmap_item->mm == vma->vm_mm) == search_new_forks)
				continue;

			ret = try_to_unmap_one(page, vma,
					rmap_item->address, flags);
			if (ret != SWAP_AGAIN || !page_mapped(page)) {
1710
				anon_vma_unlock(anon_vma);
H
Hugh Dickins 已提交
1711 1712 1713
				goto out;
			}
		}
1714
		anon_vma_unlock(anon_vma);
H
Hugh Dickins 已提交
1715
	}
H
Hugh Dickins 已提交
1716 1717
	if (!search_new_forks++)
		goto again;
H
Hugh Dickins 已提交
1718 1719 1720 1721
out:
	return ret;
}

1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738 1739 1740
#ifdef CONFIG_MIGRATION
int rmap_walk_ksm(struct page *page, int (*rmap_one)(struct page *,
		  struct vm_area_struct *, unsigned long, void *), void *arg)
{
	struct stable_node *stable_node;
	struct hlist_node *hlist;
	struct rmap_item *rmap_item;
	int ret = SWAP_AGAIN;
	int search_new_forks = 0;

	VM_BUG_ON(!PageKsm(page));
	VM_BUG_ON(!PageLocked(page));

	stable_node = page_stable_node(page);
	if (!stable_node)
		return ret;
again:
	hlist_for_each_entry(rmap_item, hlist, &stable_node->hlist, hlist) {
		struct anon_vma *anon_vma = rmap_item->anon_vma;
1741
		struct anon_vma_chain *vmac;
1742 1743
		struct vm_area_struct *vma;

1744
		anon_vma_lock(anon_vma);
1745 1746
		anon_vma_interval_tree_foreach(vmac, &anon_vma->rb_root,
					       0, ULONG_MAX) {
1747
			vma = vmac->vma;
1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761
			if (rmap_item->address < vma->vm_start ||
			    rmap_item->address >= vma->vm_end)
				continue;
			/*
			 * Initially we examine only the vma which covers this
			 * rmap_item; but later, if there is still work to do,
			 * we examine covering vmas in other mms: in case they
			 * were forked from the original since ksmd passed.
			 */
			if ((rmap_item->mm == vma->vm_mm) == search_new_forks)
				continue;

			ret = rmap_one(page, vma, rmap_item->address, arg);
			if (ret != SWAP_AGAIN) {
1762
				anon_vma_unlock(anon_vma);
1763 1764 1765
				goto out;
			}
		}
1766
		anon_vma_unlock(anon_vma);
1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783
	}
	if (!search_new_forks++)
		goto again;
out:
	return ret;
}

void ksm_migrate_page(struct page *newpage, struct page *oldpage)
{
	struct stable_node *stable_node;

	VM_BUG_ON(!PageLocked(oldpage));
	VM_BUG_ON(!PageLocked(newpage));
	VM_BUG_ON(newpage->mapping != oldpage->mapping);

	stable_node = page_stable_node(newpage);
	if (stable_node) {
1784 1785
		VM_BUG_ON(stable_node->kpfn != page_to_pfn(oldpage));
		stable_node->kpfn = page_to_pfn(newpage);
1786 1787 1788 1789
	}
}
#endif /* CONFIG_MIGRATION */

1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817
#ifdef CONFIG_MEMORY_HOTREMOVE
static struct stable_node *ksm_check_stable_tree(unsigned long start_pfn,
						 unsigned long end_pfn)
{
	struct rb_node *node;

	for (node = rb_first(&root_stable_tree); node; node = rb_next(node)) {
		struct stable_node *stable_node;

		stable_node = rb_entry(node, struct stable_node, node);
		if (stable_node->kpfn >= start_pfn &&
		    stable_node->kpfn < end_pfn)
			return stable_node;
	}
	return NULL;
}

static int ksm_memory_callback(struct notifier_block *self,
			       unsigned long action, void *arg)
{
	struct memory_notify *mn = arg;
	struct stable_node *stable_node;

	switch (action) {
	case MEM_GOING_OFFLINE:
		/*
		 * Keep it very simple for now: just lock out ksmd and
		 * MADV_UNMERGEABLE while any memory is going offline.
1818 1819 1820 1821 1822
		 * mutex_lock_nested() is necessary because lockdep was alarmed
		 * that here we take ksm_thread_mutex inside notifier chain
		 * mutex, and later take notifier chain mutex inside
		 * ksm_thread_mutex to unlock it.   But that's safe because both
		 * are inside mem_hotplug_mutex.
1823
		 */
1824
		mutex_lock_nested(&ksm_thread_mutex, SINGLE_DEPTH_NESTING);
1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845
		break;

	case MEM_OFFLINE:
		/*
		 * Most of the work is done by page migration; but there might
		 * be a few stable_nodes left over, still pointing to struct
		 * pages which have been offlined: prune those from the tree.
		 */
		while ((stable_node = ksm_check_stable_tree(mn->start_pfn,
					mn->start_pfn + mn->nr_pages)) != NULL)
			remove_node_from_stable_tree(stable_node);
		/* fallthrough */

	case MEM_CANCEL_OFFLINE:
		mutex_unlock(&ksm_thread_mutex);
		break;
	}
	return NOTIFY_OK;
}
#endif /* CONFIG_MEMORY_HOTREMOVE */

H
Hugh Dickins 已提交
1846 1847 1848 1849 1850
#ifdef CONFIG_SYSFS
/*
 * This all compiles without CONFIG_SYSFS, but is a waste of space.
 */

I
Izik Eidus 已提交
1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 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
#define KSM_ATTR_RO(_name) \
	static struct kobj_attribute _name##_attr = __ATTR_RO(_name)
#define KSM_ATTR(_name) \
	static struct kobj_attribute _name##_attr = \
		__ATTR(_name, 0644, _name##_show, _name##_store)

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

static ssize_t 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;

	ksm_thread_sleep_millisecs = msecs;

	return count;
}
KSM_ATTR(sleep_millisecs);

static ssize_t pages_to_scan_show(struct kobject *kobj,
				  struct kobj_attribute *attr, char *buf)
{
	return sprintf(buf, "%u\n", ksm_thread_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 nr_pages;

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

	ksm_thread_pages_to_scan = nr_pages;

	return count;
}
KSM_ATTR(pages_to_scan);

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

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

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

	/*
	 * KSM_RUN_MERGE sets ksmd running, and 0 stops it running.
	 * KSM_RUN_UNMERGE stops it running and unmerges all rmap_items,
1924 1925
	 * breaking COW to free the pages_shared (but leaves mm_slots
	 * on the list for when ksmd may be set running again).
I
Izik Eidus 已提交
1926 1927 1928 1929 1930
	 */

	mutex_lock(&ksm_thread_mutex);
	if (ksm_run != flags) {
		ksm_run = flags;
H
Hugh Dickins 已提交
1931
		if (flags & KSM_RUN_UNMERGE) {
1932 1933 1934
			int oom_score_adj;

			oom_score_adj = test_set_oom_score_adj(OOM_SCORE_ADJ_MAX);
H
Hugh Dickins 已提交
1935
			err = unmerge_and_remove_all_rmap_items();
1936 1937
			compare_swap_oom_score_adj(OOM_SCORE_ADJ_MAX,
								oom_score_adj);
H
Hugh Dickins 已提交
1938 1939 1940 1941 1942
			if (err) {
				ksm_run = KSM_RUN_STOP;
				count = err;
			}
		}
I
Izik Eidus 已提交
1943 1944 1945 1946 1947 1948 1949 1950 1951 1952
	}
	mutex_unlock(&ksm_thread_mutex);

	if (flags & KSM_RUN_MERGE)
		wake_up_interruptible(&ksm_thread_wait);

	return count;
}
KSM_ATTR(run);

1953 1954 1955 1956 1957 1958 1959 1960 1961 1962
static ssize_t pages_shared_show(struct kobject *kobj,
				 struct kobj_attribute *attr, char *buf)
{
	return sprintf(buf, "%lu\n", ksm_pages_shared);
}
KSM_ATTR_RO(pages_shared);

static ssize_t pages_sharing_show(struct kobject *kobj,
				  struct kobj_attribute *attr, char *buf)
{
H
Hugh Dickins 已提交
1963
	return sprintf(buf, "%lu\n", ksm_pages_sharing);
1964 1965 1966
}
KSM_ATTR_RO(pages_sharing);

1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997
static ssize_t pages_unshared_show(struct kobject *kobj,
				   struct kobj_attribute *attr, char *buf)
{
	return sprintf(buf, "%lu\n", ksm_pages_unshared);
}
KSM_ATTR_RO(pages_unshared);

static ssize_t pages_volatile_show(struct kobject *kobj,
				   struct kobj_attribute *attr, char *buf)
{
	long ksm_pages_volatile;

	ksm_pages_volatile = ksm_rmap_items - ksm_pages_shared
				- ksm_pages_sharing - ksm_pages_unshared;
	/*
	 * It was not worth any locking to calculate that statistic,
	 * but it might therefore sometimes be negative: conceal that.
	 */
	if (ksm_pages_volatile < 0)
		ksm_pages_volatile = 0;
	return sprintf(buf, "%ld\n", ksm_pages_volatile);
}
KSM_ATTR_RO(pages_volatile);

static ssize_t full_scans_show(struct kobject *kobj,
			       struct kobj_attribute *attr, char *buf)
{
	return sprintf(buf, "%lu\n", ksm_scan.seqnr);
}
KSM_ATTR_RO(full_scans);

I
Izik Eidus 已提交
1998 1999 2000 2001
static struct attribute *ksm_attrs[] = {
	&sleep_millisecs_attr.attr,
	&pages_to_scan_attr.attr,
	&run_attr.attr,
2002 2003
	&pages_shared_attr.attr,
	&pages_sharing_attr.attr,
2004 2005 2006
	&pages_unshared_attr.attr,
	&pages_volatile_attr.attr,
	&full_scans_attr.attr,
I
Izik Eidus 已提交
2007 2008 2009 2010 2011 2012 2013
	NULL,
};

static struct attribute_group ksm_attr_group = {
	.attrs = ksm_attrs,
	.name = "ksm",
};
H
Hugh Dickins 已提交
2014
#endif /* CONFIG_SYSFS */
I
Izik Eidus 已提交
2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028

static int __init ksm_init(void)
{
	struct task_struct *ksm_thread;
	int err;

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

	ksm_thread = kthread_run(ksm_scan_thread, NULL, "ksmd");
	if (IS_ERR(ksm_thread)) {
		printk(KERN_ERR "ksm: creating kthread failed\n");
		err = PTR_ERR(ksm_thread);
L
Lai Jiangshan 已提交
2029
		goto out_free;
I
Izik Eidus 已提交
2030 2031
	}

H
Hugh Dickins 已提交
2032
#ifdef CONFIG_SYSFS
I
Izik Eidus 已提交
2033 2034 2035
	err = sysfs_create_group(mm_kobj, &ksm_attr_group);
	if (err) {
		printk(KERN_ERR "ksm: register sysfs failed\n");
H
Hugh Dickins 已提交
2036
		kthread_stop(ksm_thread);
L
Lai Jiangshan 已提交
2037
		goto out_free;
I
Izik Eidus 已提交
2038
	}
H
Hugh Dickins 已提交
2039 2040 2041
#else
	ksm_run = KSM_RUN_MERGE;	/* no way for user to start it */

H
Hugh Dickins 已提交
2042
#endif /* CONFIG_SYSFS */
I
Izik Eidus 已提交
2043

2044 2045 2046 2047 2048 2049 2050
#ifdef CONFIG_MEMORY_HOTREMOVE
	/*
	 * Choose a high priority since the callback takes ksm_thread_mutex:
	 * later callbacks could only be taking locks which nest within that.
	 */
	hotplug_memory_notifier(ksm_memory_callback, 100);
#endif
I
Izik Eidus 已提交
2051 2052
	return 0;

L
Lai Jiangshan 已提交
2053
out_free:
I
Izik Eidus 已提交
2054 2055 2056
	ksm_slab_free();
out:
	return err;
H
Hugh Dickins 已提交
2057
}
I
Izik Eidus 已提交
2058
module_init(ksm_init)