ksm.c 54.4 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>
H
Hugh Dickins 已提交
38

I
Izik Eidus 已提交
39
#include <asm/tlbflush.h>
H
Hugh Dickins 已提交
40
#include "internal.h"
I
Izik Eidus 已提交
41 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

/*
 * 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
86
 * @rmap_list: head for this mm_slot's singly-linked list of rmap_items
I
Izik Eidus 已提交
87 88 89 90 91
 * @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
92
	struct rmap_item *rmap_list;
I
Izik Eidus 已提交
93 94 95 96 97 98 99
	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
100
 * @rmap_list: link to the next rmap to be scanned in the rmap_list
I
Izik Eidus 已提交
101 102 103 104 105 106 107
 * @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
108
	struct rmap_item **rmap_list;
I
Izik Eidus 已提交
109 110 111
	unsigned long seqnr;
};

H
Hugh Dickins 已提交
112 113 114 115
/**
 * 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
116
 * @kpfn: page frame number of this ksm page
H
Hugh Dickins 已提交
117 118 119 120
 */
struct stable_node {
	struct rb_node node;
	struct hlist_head hlist;
121
	unsigned long kpfn;
H
Hugh Dickins 已提交
122 123
};

I
Izik Eidus 已提交
124 125
/**
 * struct rmap_item - reverse mapping item for virtual addresses
H
Hugh Dickins 已提交
126
 * @rmap_list: next rmap_item in mm_slot's singly-linked rmap_list
H
Hugh Dickins 已提交
127
 * @anon_vma: pointer to anon_vma for this mm,address, when in stable tree
I
Izik Eidus 已提交
128 129 130
 * @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
131 132 133
 * @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
134 135
 */
struct rmap_item {
H
Hugh Dickins 已提交
136
	struct rmap_item *rmap_list;
H
Hugh Dickins 已提交
137
	struct anon_vma *anon_vma;	/* when stable */
I
Izik Eidus 已提交
138 139
	struct mm_struct *mm;
	unsigned long address;		/* + low bits used for flags below */
H
Hugh Dickins 已提交
140
	unsigned int oldchecksum;	/* when unstable */
I
Izik Eidus 已提交
141
	union {
H
Hugh Dickins 已提交
142 143 144 145 146
		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
147 148 149 150
	};
};

#define SEQNR_MASK	0x0ff	/* low bits of unstable tree seqnr */
H
Hugh Dickins 已提交
151 152
#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
153 154 155 156 157

/* 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
158 159 160
#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
161 162 163 164 165 166 167 168 169

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
170
static struct kmem_cache *stable_node_cache;
I
Izik Eidus 已提交
171 172 173
static struct kmem_cache *mm_slot_cache;

/* The number of nodes in the stable tree */
174
static unsigned long ksm_pages_shared;
I
Izik Eidus 已提交
175

H
Hugh Dickins 已提交
176
/* The number of page slots additionally sharing those nodes */
177
static unsigned long ksm_pages_sharing;
I
Izik Eidus 已提交
178

179 180 181 182 183 184
/* 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
185
/* Number of pages ksmd should scan in one batch */
186
static unsigned int ksm_thread_pages_to_scan = 100;
I
Izik Eidus 已提交
187 188

/* Milliseconds ksmd should sleep between batches */
H
Hugh Dickins 已提交
189
static unsigned int ksm_thread_sleep_millisecs = 20;
I
Izik Eidus 已提交
190 191 192 193

#define KSM_RUN_STOP	0
#define KSM_RUN_MERGE	1
#define KSM_RUN_UNMERGE	2
194
static unsigned int ksm_run = KSM_RUN_STOP;
I
Izik Eidus 已提交
195 196 197 198 199 200 201 202 203 204 205 206 207 208 209

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
210 211 212 213
	stable_node_cache = KSM_KMEM_CACHE(stable_node, 0);
	if (!stable_node_cache)
		goto out_free1;

I
Izik Eidus 已提交
214 215
	mm_slot_cache = KSM_KMEM_CACHE(mm_slot, 0);
	if (!mm_slot_cache)
H
Hugh Dickins 已提交
216
		goto out_free2;
I
Izik Eidus 已提交
217 218 219

	return 0;

H
Hugh Dickins 已提交
220 221 222
out_free2:
	kmem_cache_destroy(stable_node_cache);
out_free1:
I
Izik Eidus 已提交
223 224 225 226 227 228 229 230
	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
231
	kmem_cache_destroy(stable_node_cache);
I
Izik Eidus 已提交
232 233 234 235 236 237
	kmem_cache_destroy(rmap_item_cache);
	mm_slot_cache = NULL;
}

static inline struct rmap_item *alloc_rmap_item(void)
{
238 239 240 241 242 243
	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
244 245 246 247
}

static inline void free_rmap_item(struct rmap_item *rmap_item)
{
248
	ksm_rmap_items--;
I
Izik Eidus 已提交
249 250 251 252
	rmap_item->mm = NULL;	/* debug safety */
	kmem_cache_free(rmap_item_cache, rmap_item);
}

H
Hugh Dickins 已提交
253 254 255 256 257 258 259 260 261 262
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
263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280
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
281
	bucket = &mm_slots_hash[hash_ptr(mm, MM_SLOTS_HASH_SHIFT)];
I
Izik Eidus 已提交
282 283 284 285 286 287 288 289 290 291 292 293
	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
294
	bucket = &mm_slots_hash[hash_ptr(mm, MM_SLOTS_HASH_SHIFT)];
I
Izik Eidus 已提交
295 296 297 298 299 300 301 302 303
	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
304 305 306 307 308 309 310 311 312 313 314 315 316
/*
 * 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
317 318 319 320 321 322 323 324 325 326 327
/*
 * 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
328
static int break_ksm(struct vm_area_struct *vma, unsigned long addr)
I
Izik Eidus 已提交
329 330
{
	struct page *page;
H
Hugh Dickins 已提交
331
	int ret = 0;
I
Izik Eidus 已提交
332 333 334 335

	do {
		cond_resched();
		page = follow_page(vma, addr, FOLL_GET);
336
		if (IS_ERR_OR_NULL(page))
I
Izik Eidus 已提交
337 338 339 340 341 342 343
			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
344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373
	} 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
374 375
}

376
static void break_cow(struct rmap_item *rmap_item)
I
Izik Eidus 已提交
377
{
378 379
	struct mm_struct *mm = rmap_item->mm;
	unsigned long addr = rmap_item->address;
I
Izik Eidus 已提交
380 381
	struct vm_area_struct *vma;

H
Hugh Dickins 已提交
382 383 384 385
	/*
	 * 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.
	 */
386
	put_anon_vma(rmap_item->anon_vma);
H
Hugh Dickins 已提交
387

H
Hugh Dickins 已提交
388
	down_read(&mm->mmap_sem);
H
Hugh Dickins 已提交
389 390
	if (ksm_test_exit(mm))
		goto out;
I
Izik Eidus 已提交
391 392
	vma = find_vma(mm, addr);
	if (!vma || vma->vm_start > addr)
H
Hugh Dickins 已提交
393
		goto out;
I
Izik Eidus 已提交
394
	if (!(vma->vm_flags & VM_MERGEABLE) || !vma->anon_vma)
H
Hugh Dickins 已提交
395
		goto out;
I
Izik Eidus 已提交
396
	break_ksm(vma, addr);
H
Hugh Dickins 已提交
397
out:
I
Izik Eidus 已提交
398 399 400
	up_read(&mm->mmap_sem);
}

A
Andrea Arcangeli 已提交
401 402 403
static struct page *page_trans_compound_anon(struct page *page)
{
	if (PageTransCompound(page)) {
404
		struct page *head = compound_trans_head(page);
A
Andrea Arcangeli 已提交
405
		/*
406 407
		 * head may actually be splitted and freed from under
		 * us but it's ok here.
A
Andrea Arcangeli 已提交
408 409 410 411 412 413 414
		 */
		if (PageAnon(head))
			return head;
	}
	return NULL;
}

I
Izik Eidus 已提交
415 416 417 418 419 420 421 422
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);
H
Hugh Dickins 已提交
423 424
	if (ksm_test_exit(mm))
		goto out;
I
Izik Eidus 已提交
425 426 427 428 429 430 431
	vma = find_vma(mm, addr);
	if (!vma || vma->vm_start > addr)
		goto out;
	if (!(vma->vm_flags & VM_MERGEABLE) || !vma->anon_vma)
		goto out;

	page = follow_page(vma, addr, FOLL_GET);
432
	if (IS_ERR_OR_NULL(page))
I
Izik Eidus 已提交
433
		goto out;
A
Andrea Arcangeli 已提交
434
	if (PageAnon(page) || page_trans_compound_anon(page)) {
I
Izik Eidus 已提交
435 436 437 438 439 440 441 442 443 444
		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
445 446 447 448 449 450 451 452 453 454
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--;
455
		put_anon_vma(rmap_item->anon_vma);
H
Hugh Dickins 已提交
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
		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;

498
	page = pfn_to_page(stable_node->kpfn);
H
Hugh Dickins 已提交
499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517
	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
518 519 520 521 522 523
/*
 * 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
524 525
	if (rmap_item->address & STABLE_FLAG) {
		struct stable_node *stable_node;
H
Hugh Dickins 已提交
526
		struct page *page;
I
Izik Eidus 已提交
527

H
Hugh Dickins 已提交
528
		stable_node = rmap_item->head;
H
Hugh Dickins 已提交
529 530 531
		page = get_ksm_page(stable_node);
		if (!page)
			goto out;
H
Hugh Dickins 已提交
532

H
Hugh Dickins 已提交
533
		lock_page(page);
H
Hugh Dickins 已提交
534
		hlist_del(&rmap_item->hlist);
H
Hugh Dickins 已提交
535 536
		unlock_page(page);
		put_page(page);
537

H
Hugh Dickins 已提交
538 539 540
		if (stable_node->hlist.first)
			ksm_pages_sharing--;
		else
H
Hugh Dickins 已提交
541
			ksm_pages_shared--;
I
Izik Eidus 已提交
542

543
		put_anon_vma(rmap_item->anon_vma);
544
		rmap_item->address &= PAGE_MASK;
I
Izik Eidus 已提交
545

H
Hugh Dickins 已提交
546
	} else if (rmap_item->address & UNSTABLE_FLAG) {
I
Izik Eidus 已提交
547 548
		unsigned char age;
		/*
H
Hugh Dickins 已提交
549
		 * Usually ksmd can and must skip the rb_erase, because
I
Izik Eidus 已提交
550
		 * root_unstable_tree was already reset to RB_ROOT.
H
Hugh Dickins 已提交
551 552 553
		 * 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
554 555
		 */
		age = (unsigned char)(ksm_scan.seqnr - rmap_item->address);
556
		BUG_ON(age > 1);
I
Izik Eidus 已提交
557 558
		if (!age)
			rb_erase(&rmap_item->node, &root_unstable_tree);
559

560
		ksm_pages_unshared--;
561
		rmap_item->address &= PAGE_MASK;
I
Izik Eidus 已提交
562
	}
H
Hugh Dickins 已提交
563
out:
I
Izik Eidus 已提交
564 565 566 567
	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
568
				       struct rmap_item **rmap_list)
I
Izik Eidus 已提交
569
{
H
Hugh Dickins 已提交
570 571 572
	while (*rmap_list) {
		struct rmap_item *rmap_item = *rmap_list;
		*rmap_list = rmap_item->rmap_list;
I
Izik Eidus 已提交
573 574 575 576 577 578 579 580 581 582 583 584
		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
585 586 587 588 589
 *
 * 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
590
 */
H
Hugh Dickins 已提交
591 592
static int unmerge_ksm_pages(struct vm_area_struct *vma,
			     unsigned long start, unsigned long end)
I
Izik Eidus 已提交
593 594
{
	unsigned long addr;
H
Hugh Dickins 已提交
595
	int err = 0;
I
Izik Eidus 已提交
596

H
Hugh Dickins 已提交
597
	for (addr = start; addr < end && !err; addr += PAGE_SIZE) {
H
Hugh Dickins 已提交
598 599
		if (ksm_test_exit(vma->vm_mm))
			break;
H
Hugh Dickins 已提交
600 601 602 603 604 605
		if (signal_pending(current))
			err = -ERESTARTSYS;
		else
			err = break_ksm(vma, addr);
	}
	return err;
I
Izik Eidus 已提交
606 607
}

H
Hugh Dickins 已提交
608 609 610 611
#ifdef CONFIG_SYSFS
/*
 * Only called through the sysfs control interface:
 */
H
Hugh Dickins 已提交
612
static int unmerge_and_remove_all_rmap_items(void)
I
Izik Eidus 已提交
613 614 615 616
{
	struct mm_slot *mm_slot;
	struct mm_struct *mm;
	struct vm_area_struct *vma;
H
Hugh Dickins 已提交
617 618 619
	int err = 0;

	spin_lock(&ksm_mmlist_lock);
H
Hugh Dickins 已提交
620
	ksm_scan.mm_slot = list_entry(ksm_mm_head.mm_list.next,
H
Hugh Dickins 已提交
621 622
						struct mm_slot, mm_list);
	spin_unlock(&ksm_mmlist_lock);
I
Izik Eidus 已提交
623

H
Hugh Dickins 已提交
624 625
	for (mm_slot = ksm_scan.mm_slot;
			mm_slot != &ksm_mm_head; mm_slot = ksm_scan.mm_slot) {
I
Izik Eidus 已提交
626 627 628
		mm = mm_slot->mm;
		down_read(&mm->mmap_sem);
		for (vma = mm->mmap; vma; vma = vma->vm_next) {
H
Hugh Dickins 已提交
629 630
			if (ksm_test_exit(mm))
				break;
I
Izik Eidus 已提交
631 632
			if (!(vma->vm_flags & VM_MERGEABLE) || !vma->anon_vma)
				continue;
H
Hugh Dickins 已提交
633 634
			err = unmerge_ksm_pages(vma,
						vma->vm_start, vma->vm_end);
H
Hugh Dickins 已提交
635 636
			if (err)
				goto error;
I
Izik Eidus 已提交
637
		}
H
Hugh Dickins 已提交
638

H
Hugh Dickins 已提交
639
		remove_trailing_rmap_items(mm_slot, &mm_slot->rmap_list);
H
Hugh Dickins 已提交
640 641

		spin_lock(&ksm_mmlist_lock);
H
Hugh Dickins 已提交
642
		ksm_scan.mm_slot = list_entry(mm_slot->mm_list.next,
H
Hugh Dickins 已提交
643
						struct mm_slot, mm_list);
H
Hugh Dickins 已提交
644 645 646 647 648 649 650 651 652 653 654 655 656
		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
657 658
	}

H
Hugh Dickins 已提交
659
	ksm_scan.seqnr = 0;
H
Hugh Dickins 已提交
660 661 662 663
	return 0;

error:
	up_read(&mm->mmap_sem);
I
Izik Eidus 已提交
664
	spin_lock(&ksm_mmlist_lock);
H
Hugh Dickins 已提交
665
	ksm_scan.mm_slot = &ksm_mm_head;
I
Izik Eidus 已提交
666
	spin_unlock(&ksm_mmlist_lock);
H
Hugh Dickins 已提交
667
	return err;
I
Izik Eidus 已提交
668
}
H
Hugh Dickins 已提交
669
#endif /* CONFIG_SYSFS */
I
Izik Eidus 已提交
670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711

static u32 calc_checksum(struct page *page)
{
	u32 checksum;
	void *addr = kmap_atomic(page, KM_USER0);
	checksum = jhash2(addr, PAGE_SIZE / 4, 17);
	kunmap_atomic(addr, KM_USER0);
	return checksum;
}

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

	addr1 = kmap_atomic(page1, KM_USER0);
	addr2 = kmap_atomic(page2, KM_USER1);
	ret = memcmp(addr1, addr2, PAGE_SIZE);
	kunmap_atomic(addr2, KM_USER1);
	kunmap_atomic(addr1, KM_USER0);
	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;

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

A
Andrea Arcangeli 已提交
712
	BUG_ON(PageTransCompound(page));
I
Izik Eidus 已提交
713 714 715 716
	ptep = page_check_address(page, mm, addr, &ptl, 0);
	if (!ptep)
		goto out;

717
	if (pte_write(*ptep) || pte_dirty(*ptep)) {
I
Izik Eidus 已提交
718 719 720 721 722
		pte_t entry;

		swapped = PageSwapCache(page);
		flush_cache_page(vma, addr, page_to_pfn(page));
		/*
L
Lucas De Marchi 已提交
723
		 * Ok this is tricky, when get_user_pages_fast() run it doesn't
I
Izik Eidus 已提交
724 725 726 727 728 729 730 731 732 733 734 735
		 * 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
		 */
736
		if (page_mapcount(page) + 1 + swapped != page_count(page)) {
737
			set_pte_at(mm, addr, ptep, entry);
I
Izik Eidus 已提交
738 739
			goto out_unlock;
		}
740 741 742
		if (pte_dirty(entry))
			set_page_dirty(page);
		entry = pte_mkclean(pte_wrprotect(entry));
I
Izik Eidus 已提交
743 744 745 746 747 748 749 750 751 752 753 754 755
		set_pte_at_notify(mm, addr, ptep, entry);
	}
	*orig_pte = *ptep;
	err = 0;

out_unlock:
	pte_unmap_unlock(ptep, ptl);
out:
	return err;
}

/**
 * replace_page - replace page in vma by new ksm page
756 757 758
 * @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
759 760 761 762
 * @orig_pte: the original value of the pte
 *
 * Returns 0 on success, -EFAULT on failure.
 */
763 764
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
765 766 767 768 769 770 771 772 773 774
{
	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;

775
	addr = page_address_in_vma(page, vma);
I
Izik Eidus 已提交
776 777 778 779 780 781 782 783 784 785 786 787
	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
788
	BUG_ON(pmd_trans_huge(*pmd));
I
Izik Eidus 已提交
789 790 791 792 793 794 795 796 797
	if (!pmd_present(*pmd))
		goto out;

	ptep = pte_offset_map_lock(mm, pmd, addr, &ptl);
	if (!pte_same(*ptep, orig_pte)) {
		pte_unmap_unlock(ptep, ptl);
		goto out;
	}

798
	get_page(kpage);
H
Hugh Dickins 已提交
799
	page_add_anon_rmap(kpage, vma, addr);
I
Izik Eidus 已提交
800 801 802

	flush_cache_page(vma, addr, pte_pfn(*ptep));
	ptep_clear_flush(vma, addr, ptep);
803
	set_pte_at_notify(mm, addr, ptep, mk_pte(kpage, vma->vm_page_prot));
I
Izik Eidus 已提交
804

805
	page_remove_rmap(page);
806 807
	if (!page_mapped(page))
		try_to_free_swap(page);
808
	put_page(page);
I
Izik Eidus 已提交
809 810 811 812 813 814 815

	pte_unmap_unlock(ptep, ptl);
	err = 0;
out:
	return err;
}

A
Andrea Arcangeli 已提交
816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842
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
843 844
/*
 * try_to_merge_one_page - take two pages and merge them into one
845 846
 * @vma: the vma that holds the pte pointing to page
 * @page: the PageAnon page that we want to replace with kpage
847 848
 * @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
849 850 851 852
 *
 * This function returns 0 if the pages were merged, -EFAULT otherwise.
 */
static int try_to_merge_one_page(struct vm_area_struct *vma,
853
				 struct page *page, struct page *kpage)
I
Izik Eidus 已提交
854 855 856 857
{
	pte_t orig_pte = __pte(0);
	int err = -EFAULT;

H
Hugh Dickins 已提交
858 859 860
	if (page == kpage)			/* ksm page forked */
		return 0;

I
Izik Eidus 已提交
861 862
	if (!(vma->vm_flags & VM_MERGEABLE))
		goto out;
A
Andrea Arcangeli 已提交
863 864 865
	if (PageTransCompound(page) && page_trans_compound_anon_split(page))
		goto out;
	BUG_ON(PageTransCompound(page));
866
	if (!PageAnon(page))
I
Izik Eidus 已提交
867 868 869 870 871 872 873 874 875
		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.
	 */
876
	if (!trylock_page(page))
877
		goto out;
I
Izik Eidus 已提交
878 879 880 881 882 883
	/*
	 * 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.
	 */
884 885 886 887 888 889 890 891 892 893 894 895 896
	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
897

898
	if ((vma->vm_flags & VM_LOCKED) && kpage && !err) {
H
Hugh Dickins 已提交
899
		munlock_vma_page(page);
H
Hugh Dickins 已提交
900 901 902 903 904 905 906
		if (!PageMlocked(kpage)) {
			unlock_page(page);
			lock_page(kpage);
			mlock_vma_page(kpage);
			page = kpage;		/* for final unlock */
		}
	}
H
Hugh Dickins 已提交
907

908
	unlock_page(page);
I
Izik Eidus 已提交
909 910 911 912
out:
	return err;
}

H
Hugh Dickins 已提交
913 914 915
/*
 * 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.
916 917
 *
 * This function returns 0 if the pages were merged, -EFAULT otherwise.
H
Hugh Dickins 已提交
918
 */
919 920
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
921
{
922
	struct mm_struct *mm = rmap_item->mm;
H
Hugh Dickins 已提交
923 924 925
	struct vm_area_struct *vma;
	int err = -EFAULT;

926 927
	down_read(&mm->mmap_sem);
	if (ksm_test_exit(mm))
H
Hugh Dickins 已提交
928
		goto out;
929 930
	vma = find_vma(mm, rmap_item->address);
	if (!vma || vma->vm_start > rmap_item->address)
H
Hugh Dickins 已提交
931 932
		goto out;

933
	err = try_to_merge_one_page(vma, page, kpage);
H
Hugh Dickins 已提交
934 935 936 937
	if (err)
		goto out;

	/* Must get reference to anon_vma while still holding mmap_sem */
938 939
	rmap_item->anon_vma = vma->anon_vma;
	get_anon_vma(vma->anon_vma);
H
Hugh Dickins 已提交
940
out:
941
	up_read(&mm->mmap_sem);
H
Hugh Dickins 已提交
942 943 944
	return err;
}

I
Izik Eidus 已提交
945 946 947 948
/*
 * try_to_merge_two_pages - take two identical pages and prepare them
 * to be merged into one page.
 *
949 950
 * This function returns the kpage if we successfully merged two identical
 * pages into one ksm page, NULL otherwise.
I
Izik Eidus 已提交
951
 *
952
 * Note that this function upgrades page to ksm page: if one of the pages
I
Izik Eidus 已提交
953 954
 * is already a ksm page, try_to_merge_with_ksm_page should be used.
 */
955 956 957 958
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
959
{
960
	int err;
I
Izik Eidus 已提交
961

962
	err = try_to_merge_with_ksm_page(rmap_item, page, NULL);
I
Izik Eidus 已提交
963
	if (!err) {
964
		err = try_to_merge_with_ksm_page(tree_rmap_item,
965
							tree_page, page);
I
Izik Eidus 已提交
966
		/*
H
Hugh Dickins 已提交
967 968
		 * If that fails, we have a ksm page with only one pte
		 * pointing to it: so break it.
I
Izik Eidus 已提交
969
		 */
H
Hugh Dickins 已提交
970
		if (err)
971
			break_cow(rmap_item);
I
Izik Eidus 已提交
972
	}
973
	return err ? NULL : page;
I
Izik Eidus 已提交
974 975 976
}

/*
977
 * stable_tree_search - search for page inside the stable tree
I
Izik Eidus 已提交
978 979 980 981
 *
 * 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
982
 * This function returns the stable tree node of identical content if found,
I
Izik Eidus 已提交
983 984
 * NULL otherwise.
 */
985
static struct page *stable_tree_search(struct page *page)
I
Izik Eidus 已提交
986 987
{
	struct rb_node *node = root_stable_tree.rb_node;
H
Hugh Dickins 已提交
988
	struct stable_node *stable_node;
I
Izik Eidus 已提交
989

990 991 992
	stable_node = page_stable_node(page);
	if (stable_node) {			/* ksm page forked */
		get_page(page);
993
		return page;
994 995
	}

I
Izik Eidus 已提交
996
	while (node) {
H
Hugh Dickins 已提交
997
		struct page *tree_page;
I
Izik Eidus 已提交
998 999
		int ret;

1000
		cond_resched();
H
Hugh Dickins 已提交
1001
		stable_node = rb_entry(node, struct stable_node, node);
H
Hugh Dickins 已提交
1002 1003 1004
		tree_page = get_ksm_page(stable_node);
		if (!tree_page)
			return NULL;
I
Izik Eidus 已提交
1005

H
Hugh Dickins 已提交
1006
		ret = memcmp_pages(page, tree_page);
I
Izik Eidus 已提交
1007

H
Hugh Dickins 已提交
1008 1009
		if (ret < 0) {
			put_page(tree_page);
I
Izik Eidus 已提交
1010
			node = node->rb_left;
H
Hugh Dickins 已提交
1011 1012
		} else if (ret > 0) {
			put_page(tree_page);
I
Izik Eidus 已提交
1013
			node = node->rb_right;
H
Hugh Dickins 已提交
1014
		} else
1015
			return tree_page;
I
Izik Eidus 已提交
1016 1017 1018 1019 1020 1021 1022 1023 1024
	}

	return NULL;
}

/*
 * stable_tree_insert - insert rmap_item pointing to new ksm page
 * into the stable tree.
 *
H
Hugh Dickins 已提交
1025 1026
 * This function returns the stable tree node just allocated on success,
 * NULL otherwise.
I
Izik Eidus 已提交
1027
 */
H
Hugh Dickins 已提交
1028
static struct stable_node *stable_tree_insert(struct page *kpage)
I
Izik Eidus 已提交
1029 1030 1031
{
	struct rb_node **new = &root_stable_tree.rb_node;
	struct rb_node *parent = NULL;
H
Hugh Dickins 已提交
1032
	struct stable_node *stable_node;
I
Izik Eidus 已提交
1033 1034

	while (*new) {
H
Hugh Dickins 已提交
1035
		struct page *tree_page;
I
Izik Eidus 已提交
1036 1037
		int ret;

1038
		cond_resched();
H
Hugh Dickins 已提交
1039
		stable_node = rb_entry(*new, struct stable_node, node);
H
Hugh Dickins 已提交
1040 1041 1042
		tree_page = get_ksm_page(stable_node);
		if (!tree_page)
			return NULL;
I
Izik Eidus 已提交
1043

H
Hugh Dickins 已提交
1044 1045
		ret = memcmp_pages(kpage, tree_page);
		put_page(tree_page);
I
Izik Eidus 已提交
1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061

		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
1062 1063 1064
	stable_node = alloc_stable_node();
	if (!stable_node)
		return NULL;
I
Izik Eidus 已提交
1065

H
Hugh Dickins 已提交
1066 1067 1068 1069 1070
	rb_link_node(&stable_node->node, parent, new);
	rb_insert_color(&stable_node->node, &root_stable_tree);

	INIT_HLIST_HEAD(&stable_node->hlist);

1071
	stable_node->kpfn = page_to_pfn(kpage);
1072 1073
	set_page_stable_node(kpage, stable_node);

H
Hugh Dickins 已提交
1074
	return stable_node;
I
Izik Eidus 已提交
1075 1076 1077
}

/*
1078 1079
 * unstable_tree_search_insert - search for identical page,
 * else insert rmap_item into the unstable tree.
I
Izik Eidus 已提交
1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090
 *
 * 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.
 */
1091 1092 1093 1094 1095
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
1096 1097 1098 1099 1100 1101
{
	struct rb_node **new = &root_unstable_tree.rb_node;
	struct rb_node *parent = NULL;

	while (*new) {
		struct rmap_item *tree_rmap_item;
1102
		struct page *tree_page;
I
Izik Eidus 已提交
1103 1104
		int ret;

H
Hugh Dickins 已提交
1105
		cond_resched();
I
Izik Eidus 已提交
1106
		tree_rmap_item = rb_entry(*new, struct rmap_item, node);
1107
		tree_page = get_mergeable_page(tree_rmap_item);
1108
		if (IS_ERR_OR_NULL(tree_page))
I
Izik Eidus 已提交
1109 1110 1111
			return NULL;

		/*
1112
		 * Don't substitute a ksm page for a forked page.
I
Izik Eidus 已提交
1113
		 */
1114 1115
		if (page == tree_page) {
			put_page(tree_page);
I
Izik Eidus 已提交
1116 1117 1118
			return NULL;
		}

1119
		ret = memcmp_pages(page, tree_page);
I
Izik Eidus 已提交
1120 1121 1122

		parent = *new;
		if (ret < 0) {
1123
			put_page(tree_page);
I
Izik Eidus 已提交
1124 1125
			new = &parent->rb_left;
		} else if (ret > 0) {
1126
			put_page(tree_page);
I
Izik Eidus 已提交
1127 1128
			new = &parent->rb_right;
		} else {
1129
			*tree_pagep = tree_page;
I
Izik Eidus 已提交
1130 1131 1132 1133
			return tree_rmap_item;
		}
	}

H
Hugh Dickins 已提交
1134
	rmap_item->address |= UNSTABLE_FLAG;
I
Izik Eidus 已提交
1135 1136 1137 1138
	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);

1139
	ksm_pages_unshared++;
I
Izik Eidus 已提交
1140 1141 1142 1143 1144 1145 1146 1147 1148
	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
1149
			       struct stable_node *stable_node)
I
Izik Eidus 已提交
1150
{
H
Hugh Dickins 已提交
1151
	rmap_item->head = stable_node;
I
Izik Eidus 已提交
1152
	rmap_item->address |= STABLE_FLAG;
H
Hugh Dickins 已提交
1153
	hlist_add_head(&rmap_item->hlist, &stable_node->hlist);
H
Hugh Dickins 已提交
1154

H
Hugh Dickins 已提交
1155 1156 1157 1158
	if (rmap_item->hlist.next)
		ksm_pages_sharing++;
	else
		ksm_pages_shared++;
I
Izik Eidus 已提交
1159 1160 1161
}

/*
H
Hugh Dickins 已提交
1162 1163 1164 1165
 * 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
1166 1167 1168 1169 1170 1171 1172
 *
 * @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;
1173
	struct page *tree_page = NULL;
H
Hugh Dickins 已提交
1174
	struct stable_node *stable_node;
1175
	struct page *kpage;
I
Izik Eidus 已提交
1176 1177 1178
	unsigned int checksum;
	int err;

1179
	remove_rmap_item_from_tree(rmap_item);
I
Izik Eidus 已提交
1180 1181

	/* We first start with searching the page inside the stable tree */
1182 1183
	kpage = stable_tree_search(page);
	if (kpage) {
1184
		err = try_to_merge_with_ksm_page(rmap_item, page, kpage);
I
Izik Eidus 已提交
1185 1186 1187 1188 1189
		if (!err) {
			/*
			 * The page was successfully merged:
			 * add its rmap_item to the stable tree.
			 */
H
Hugh Dickins 已提交
1190
			lock_page(kpage);
1191
			stable_tree_append(rmap_item, page_stable_node(kpage));
H
Hugh Dickins 已提交
1192
			unlock_page(kpage);
I
Izik Eidus 已提交
1193
		}
1194
		put_page(kpage);
I
Izik Eidus 已提交
1195 1196 1197 1198
		return;
	}

	/*
H
Hugh Dickins 已提交
1199 1200 1201 1202
	 * 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
1203 1204 1205 1206 1207 1208 1209
	 */
	checksum = calc_checksum(page);
	if (rmap_item->oldchecksum != checksum) {
		rmap_item->oldchecksum = checksum;
		return;
	}

1210 1211
	tree_rmap_item =
		unstable_tree_search_insert(rmap_item, page, &tree_page);
I
Izik Eidus 已提交
1212
	if (tree_rmap_item) {
1213 1214 1215
		kpage = try_to_merge_two_pages(rmap_item, page,
						tree_rmap_item, tree_page);
		put_page(tree_page);
I
Izik Eidus 已提交
1216 1217 1218 1219 1220
		/*
		 * 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.
		 */
1221
		if (kpage) {
1222
			remove_rmap_item_from_tree(tree_rmap_item);
1223

H
Hugh Dickins 已提交
1224
			lock_page(kpage);
H
Hugh Dickins 已提交
1225 1226 1227 1228 1229
			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
1230
			unlock_page(kpage);
H
Hugh Dickins 已提交
1231

I
Izik Eidus 已提交
1232 1233 1234 1235 1236 1237
			/*
			 * 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
1238
			if (!stable_node) {
1239 1240
				break_cow(tree_rmap_item);
				break_cow(rmap_item);
I
Izik Eidus 已提交
1241 1242 1243 1244 1245 1246
			}
		}
	}
}

static struct rmap_item *get_next_rmap_item(struct mm_slot *mm_slot,
H
Hugh Dickins 已提交
1247
					    struct rmap_item **rmap_list,
I
Izik Eidus 已提交
1248 1249 1250 1251
					    unsigned long addr)
{
	struct rmap_item *rmap_item;

H
Hugh Dickins 已提交
1252 1253
	while (*rmap_list) {
		rmap_item = *rmap_list;
1254
		if ((rmap_item->address & PAGE_MASK) == addr)
I
Izik Eidus 已提交
1255 1256 1257
			return rmap_item;
		if (rmap_item->address > addr)
			break;
H
Hugh Dickins 已提交
1258
		*rmap_list = rmap_item->rmap_list;
I
Izik Eidus 已提交
1259 1260 1261 1262 1263 1264 1265 1266 1267
		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
1268 1269
		rmap_item->rmap_list = *rmap_list;
		*rmap_list = rmap_item;
I
Izik Eidus 已提交
1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285
	}
	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
1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297
		/*
		 * 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
1298 1299 1300 1301 1302 1303 1304 1305
		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);
next_mm:
		ksm_scan.address = 0;
H
Hugh Dickins 已提交
1306
		ksm_scan.rmap_list = &slot->rmap_list;
I
Izik Eidus 已提交
1307 1308 1309 1310
	}

	mm = slot->mm;
	down_read(&mm->mmap_sem);
H
Hugh Dickins 已提交
1311 1312 1313 1314 1315 1316
	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
1317 1318 1319 1320 1321 1322 1323 1324
		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
1325 1326
			if (ksm_test_exit(mm))
				break;
I
Izik Eidus 已提交
1327
			*page = follow_page(vma, ksm_scan.address, FOLL_GET);
1328 1329 1330 1331 1332
			if (IS_ERR_OR_NULL(*page)) {
				ksm_scan.address += PAGE_SIZE;
				cond_resched();
				continue;
			}
A
Andrea Arcangeli 已提交
1333 1334
			if (PageAnon(*page) ||
			    page_trans_compound_anon(*page)) {
I
Izik Eidus 已提交
1335 1336 1337
				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
1338
					ksm_scan.rmap_list, ksm_scan.address);
I
Izik Eidus 已提交
1339
				if (rmap_item) {
H
Hugh Dickins 已提交
1340 1341
					ksm_scan.rmap_list =
							&rmap_item->rmap_list;
I
Izik Eidus 已提交
1342 1343 1344 1345 1346 1347
					ksm_scan.address += PAGE_SIZE;
				} else
					put_page(*page);
				up_read(&mm->mmap_sem);
				return rmap_item;
			}
1348
			put_page(*page);
I
Izik Eidus 已提交
1349 1350 1351 1352 1353
			ksm_scan.address += PAGE_SIZE;
			cond_resched();
		}
	}

H
Hugh Dickins 已提交
1354 1355
	if (ksm_test_exit(mm)) {
		ksm_scan.address = 0;
H
Hugh Dickins 已提交
1356
		ksm_scan.rmap_list = &slot->rmap_list;
H
Hugh Dickins 已提交
1357
	}
I
Izik Eidus 已提交
1358 1359 1360 1361
	/*
	 * 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
1362
	remove_trailing_rmap_items(slot, ksm_scan.rmap_list);
I
Izik Eidus 已提交
1363 1364

	spin_lock(&ksm_mmlist_lock);
1365 1366 1367 1368 1369 1370 1371
	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
1372 1373 1374 1375
		 * 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).
1376 1377 1378
		 */
		hlist_del(&slot->link);
		list_del(&slot->mm_list);
H
Hugh Dickins 已提交
1379 1380
		spin_unlock(&ksm_mmlist_lock);

1381 1382
		free_mm_slot(slot);
		clear_bit(MMF_VM_MERGEABLE, &mm->flags);
H
Hugh Dickins 已提交
1383 1384 1385 1386 1387
		up_read(&mm->mmap_sem);
		mmdrop(mm);
	} else {
		spin_unlock(&ksm_mmlist_lock);
		up_read(&mm->mmap_sem);
1388
	}
I
Izik Eidus 已提交
1389 1390

	/* Repeat until we've completed scanning the whole list */
1391
	slot = ksm_scan.mm_slot;
I
Izik Eidus 已提交
1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405
	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;
1406
	struct page *uninitialized_var(page);
I
Izik Eidus 已提交
1407

1408
	while (scan_npages-- && likely(!freezing(current))) {
I
Izik Eidus 已提交
1409 1410 1411 1412 1413 1414 1415 1416 1417 1418
		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
1419 1420 1421 1422 1423
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
1424 1425
static int ksm_scan_thread(void *nothing)
{
1426
	set_freezable();
I
Izik Eidus 已提交
1427
	set_user_nice(current, 5);
I
Izik Eidus 已提交
1428 1429

	while (!kthread_should_stop()) {
H
Hugh Dickins 已提交
1430 1431
		mutex_lock(&ksm_thread_mutex);
		if (ksmd_should_run())
I
Izik Eidus 已提交
1432
			ksm_do_scan(ksm_thread_pages_to_scan);
H
Hugh Dickins 已提交
1433 1434
		mutex_unlock(&ksm_thread_mutex);

1435 1436
		try_to_freeze();

H
Hugh Dickins 已提交
1437
		if (ksmd_should_run()) {
I
Izik Eidus 已提交
1438 1439 1440
			schedule_timeout_interruptible(
				msecs_to_jiffies(ksm_thread_sleep_millisecs));
		} else {
1441
			wait_event_freezable(ksm_thread_wait,
H
Hugh Dickins 已提交
1442
				ksmd_should_run() || kthread_should_stop());
I
Izik Eidus 已提交
1443 1444 1445 1446 1447
		}
	}
	return 0;
}

H
Hugh Dickins 已提交
1448 1449 1450 1451
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
1452
	int err;
H
Hugh Dickins 已提交
1453 1454 1455 1456 1457 1458 1459 1460 1461

	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 |
				 VM_RESERVED  | VM_HUGETLB | VM_INSERTPAGE |
H
Hugh Dickins 已提交
1462
				 VM_NONLINEAR | VM_MIXEDMAP | VM_SAO))
H
Hugh Dickins 已提交
1463 1464
			return 0;		/* just ignore the advice */

H
Hugh Dickins 已提交
1465 1466 1467 1468 1469
		if (!test_bit(MMF_VM_MERGEABLE, &mm->flags)) {
			err = __ksm_enter(mm);
			if (err)
				return err;
		}
H
Hugh Dickins 已提交
1470 1471 1472 1473 1474 1475 1476 1477

		*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
1478 1479 1480 1481 1482
		if (vma->anon_vma) {
			err = unmerge_ksm_pages(vma, start, end);
			if (err)
				return err;
		}
H
Hugh Dickins 已提交
1483 1484 1485 1486 1487 1488 1489 1490 1491 1492

		*vm_flags &= ~VM_MERGEABLE;
		break;
	}

	return 0;
}

int __ksm_enter(struct mm_struct *mm)
{
H
Hugh Dickins 已提交
1493 1494 1495 1496
	struct mm_slot *mm_slot;
	int needs_wakeup;

	mm_slot = alloc_mm_slot();
I
Izik Eidus 已提交
1497 1498 1499
	if (!mm_slot)
		return -ENOMEM;

H
Hugh Dickins 已提交
1500 1501 1502
	/* Check ksm_run too?  Would need tighter locking */
	needs_wakeup = list_empty(&ksm_mm_head.mm_list);

I
Izik Eidus 已提交
1503 1504 1505 1506 1507 1508 1509 1510 1511 1512
	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
1513
	set_bit(MMF_VM_MERGEABLE, &mm->flags);
H
Hugh Dickins 已提交
1514
	atomic_inc(&mm->mm_count);
H
Hugh Dickins 已提交
1515 1516 1517 1518

	if (needs_wakeup)
		wake_up_interruptible(&ksm_thread_wait);

H
Hugh Dickins 已提交
1519 1520 1521
	return 0;
}

1522
void __ksm_exit(struct mm_struct *mm)
H
Hugh Dickins 已提交
1523
{
1524
	struct mm_slot *mm_slot;
H
Hugh Dickins 已提交
1525
	int easy_to_free = 0;
1526

I
Izik Eidus 已提交
1527
	/*
H
Hugh Dickins 已提交
1528 1529 1530 1531 1532 1533
	 * 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
1534
	 */
H
Hugh Dickins 已提交
1535

1536 1537
	spin_lock(&ksm_mmlist_lock);
	mm_slot = get_mm_slot(mm);
H
Hugh Dickins 已提交
1538
	if (mm_slot && ksm_scan.mm_slot != mm_slot) {
H
Hugh Dickins 已提交
1539
		if (!mm_slot->rmap_list) {
H
Hugh Dickins 已提交
1540 1541 1542 1543 1544 1545 1546
			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);
		}
1547 1548 1549
	}
	spin_unlock(&ksm_mmlist_lock);

H
Hugh Dickins 已提交
1550 1551 1552 1553 1554 1555 1556 1557
	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
1558 1559
}

H
Hugh Dickins 已提交
1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588 1589 1590
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);

		if (page_evictable(new_page, vma))
			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
1591
	int search_new_forks = 0;
H
Hugh Dickins 已提交
1592 1593 1594 1595 1596 1597 1598

	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
1599
again:
H
Hugh Dickins 已提交
1600
	hlist_for_each_entry(rmap_item, hlist, &stable_node->hlist, hlist) {
H
Hugh Dickins 已提交
1601
		struct anon_vma *anon_vma = rmap_item->anon_vma;
1602
		struct anon_vma_chain *vmac;
H
Hugh Dickins 已提交
1603
		struct vm_area_struct *vma;
H
Hugh Dickins 已提交
1604

1605
		anon_vma_lock(anon_vma);
1606 1607
		list_for_each_entry(vmac, &anon_vma->head, same_anon_vma) {
			vma = vmac->vma;
H
Hugh Dickins 已提交
1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621
			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
1622

H
Hugh Dickins 已提交
1623
			referenced += page_referenced_one(page, vma,
H
Hugh Dickins 已提交
1624
				rmap_item->address, &mapcount, vm_flags);
H
Hugh Dickins 已提交
1625 1626 1627
			if (!search_new_forks || !mapcount)
				break;
		}
1628
		anon_vma_unlock(anon_vma);
H
Hugh Dickins 已提交
1629 1630 1631
		if (!mapcount)
			goto out;
	}
H
Hugh Dickins 已提交
1632 1633
	if (!search_new_forks++)
		goto again;
H
Hugh Dickins 已提交
1634 1635 1636 1637 1638 1639 1640 1641 1642 1643
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
1644
	int search_new_forks = 0;
H
Hugh Dickins 已提交
1645 1646 1647 1648 1649 1650 1651

	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
1652
again:
H
Hugh Dickins 已提交
1653
	hlist_for_each_entry(rmap_item, hlist, &stable_node->hlist, hlist) {
H
Hugh Dickins 已提交
1654
		struct anon_vma *anon_vma = rmap_item->anon_vma;
1655
		struct anon_vma_chain *vmac;
H
Hugh Dickins 已提交
1656
		struct vm_area_struct *vma;
H
Hugh Dickins 已提交
1657

1658
		anon_vma_lock(anon_vma);
1659 1660
		list_for_each_entry(vmac, &anon_vma->head, same_anon_vma) {
			vma = vmac->vma;
H
Hugh Dickins 已提交
1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675
			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)) {
1676
				anon_vma_unlock(anon_vma);
H
Hugh Dickins 已提交
1677 1678 1679
				goto out;
			}
		}
1680
		anon_vma_unlock(anon_vma);
H
Hugh Dickins 已提交
1681
	}
H
Hugh Dickins 已提交
1682 1683
	if (!search_new_forks++)
		goto again;
H
Hugh Dickins 已提交
1684 1685 1686 1687
out:
	return ret;
}

1688 1689 1690 1691 1692 1693 1694 1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706
#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;
1707
		struct anon_vma_chain *vmac;
1708 1709
		struct vm_area_struct *vma;

1710
		anon_vma_lock(anon_vma);
1711 1712
		list_for_each_entry(vmac, &anon_vma->head, same_anon_vma) {
			vma = vmac->vma;
1713 1714 1715 1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726
			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) {
1727
				anon_vma_unlock(anon_vma);
1728 1729 1730
				goto out;
			}
		}
1731
		anon_vma_unlock(anon_vma);
1732 1733 1734 1735 1736 1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 1748
	}
	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) {
1749 1750
		VM_BUG_ON(stable_node->kpfn != page_to_pfn(oldpage));
		stable_node->kpfn = page_to_pfn(newpage);
1751 1752 1753 1754
	}
}
#endif /* CONFIG_MIGRATION */

1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782
#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.
1783 1784 1785 1786 1787
		 * 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.
1788
		 */
1789
		mutex_lock_nested(&ksm_thread_mutex, SINGLE_DEPTH_NESTING);
1790 1791 1792 1793 1794 1795 1796 1797 1798 1799 1800 1801 1802 1803 1804 1805 1806 1807 1808 1809 1810
		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
1811 1812 1813 1814 1815
#ifdef CONFIG_SYSFS
/*
 * This all compiles without CONFIG_SYSFS, but is a waste of space.
 */

I
Izik Eidus 已提交
1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 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
#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,
1889 1890
	 * 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
1891 1892 1893 1894 1895
	 */

	mutex_lock(&ksm_thread_mutex);
	if (ksm_run != flags) {
		ksm_run = flags;
H
Hugh Dickins 已提交
1896
		if (flags & KSM_RUN_UNMERGE) {
H
Hugh Dickins 已提交
1897
			current->flags |= PF_OOM_ORIGIN;
H
Hugh Dickins 已提交
1898
			err = unmerge_and_remove_all_rmap_items();
H
Hugh Dickins 已提交
1899
			current->flags &= ~PF_OOM_ORIGIN;
H
Hugh Dickins 已提交
1900 1901 1902 1903 1904
			if (err) {
				ksm_run = KSM_RUN_STOP;
				count = err;
			}
		}
I
Izik Eidus 已提交
1905 1906 1907 1908 1909 1910 1911 1912 1913 1914
	}
	mutex_unlock(&ksm_thread_mutex);

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

	return count;
}
KSM_ATTR(run);

1915 1916 1917 1918 1919 1920 1921 1922 1923 1924
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
1925
	return sprintf(buf, "%lu\n", ksm_pages_sharing);
1926 1927 1928
}
KSM_ATTR_RO(pages_sharing);

1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958 1959
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
1960 1961 1962 1963
static struct attribute *ksm_attrs[] = {
	&sleep_millisecs_attr.attr,
	&pages_to_scan_attr.attr,
	&run_attr.attr,
1964 1965
	&pages_shared_attr.attr,
	&pages_sharing_attr.attr,
1966 1967 1968
	&pages_unshared_attr.attr,
	&pages_volatile_attr.attr,
	&full_scans_attr.attr,
I
Izik Eidus 已提交
1969 1970 1971 1972 1973 1974 1975
	NULL,
};

static struct attribute_group ksm_attr_group = {
	.attrs = ksm_attrs,
	.name = "ksm",
};
H
Hugh Dickins 已提交
1976
#endif /* CONFIG_SYSFS */
I
Izik Eidus 已提交
1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990

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
1991
		goto out_free;
I
Izik Eidus 已提交
1992 1993
	}

H
Hugh Dickins 已提交
1994
#ifdef CONFIG_SYSFS
I
Izik Eidus 已提交
1995 1996 1997
	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
1998
		kthread_stop(ksm_thread);
L
Lai Jiangshan 已提交
1999
		goto out_free;
I
Izik Eidus 已提交
2000
	}
H
Hugh Dickins 已提交
2001 2002 2003
#else
	ksm_run = KSM_RUN_MERGE;	/* no way for user to start it */

H
Hugh Dickins 已提交
2004
#endif /* CONFIG_SYSFS */
I
Izik Eidus 已提交
2005

2006 2007 2008 2009 2010 2011 2012
#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
2013 2014
	return 0;

L
Lai Jiangshan 已提交
2015
out_free:
I
Izik Eidus 已提交
2016 2017 2018
	ksm_slab_free();
out:
	return err;
H
Hugh Dickins 已提交
2019
}
I
Izik Eidus 已提交
2020
module_init(ksm_init)