Root/
1 | /* |
2 | * Memory merging support. |
3 | * |
4 | * This code enables dynamic sharing of identical pages found in different |
5 | * memory areas, even if they are not shared by fork() |
6 | * |
7 | * Copyright (C) 2008-2009 Red Hat, Inc. |
8 | * Authors: |
9 | * Izik Eidus |
10 | * Andrea Arcangeli |
11 | * Chris Wright |
12 | * Hugh Dickins |
13 | * |
14 | * This work is licensed under the terms of the GNU GPL, version 2. |
15 | */ |
16 | |
17 | #include <linux/errno.h> |
18 | #include <linux/mm.h> |
19 | #include <linux/fs.h> |
20 | #include <linux/mman.h> |
21 | #include <linux/sched.h> |
22 | #include <linux/rwsem.h> |
23 | #include <linux/pagemap.h> |
24 | #include <linux/rmap.h> |
25 | #include <linux/spinlock.h> |
26 | #include <linux/jhash.h> |
27 | #include <linux/delay.h> |
28 | #include <linux/kthread.h> |
29 | #include <linux/wait.h> |
30 | #include <linux/slab.h> |
31 | #include <linux/rbtree.h> |
32 | #include <linux/memory.h> |
33 | #include <linux/mmu_notifier.h> |
34 | #include <linux/swap.h> |
35 | #include <linux/ksm.h> |
36 | #include <linux/hash.h> |
37 | #include <linux/freezer.h> |
38 | |
39 | #include <asm/tlbflush.h> |
40 | #include "internal.h" |
41 | |
42 | /* |
43 | * A few notes about the KSM scanning process, |
44 | * to make it easier to understand the data structures below: |
45 | * |
46 | * In order to reduce excessive scanning, KSM sorts the memory pages by their |
47 | * contents into a data structure that holds pointers to the pages' locations. |
48 | * |
49 | * Since the contents of the pages may change at any moment, KSM cannot just |
50 | * insert the pages into a normal sorted tree and expect it to find anything. |
51 | * Therefore KSM uses two data structures - the stable and the unstable tree. |
52 | * |
53 | * The stable tree holds pointers to all the merged pages (ksm pages), sorted |
54 | * by their contents. Because each such page is write-protected, searching on |
55 | * this tree is fully assured to be working (except when pages are unmapped), |
56 | * and therefore this tree is called the stable tree. |
57 | * |
58 | * In addition to the stable tree, KSM uses a second data structure called the |
59 | * unstable tree: this tree holds pointers to pages which have been found to |
60 | * be "unchanged for a period of time". The unstable tree sorts these pages |
61 | * by their contents, but since they are not write-protected, KSM cannot rely |
62 | * upon the unstable tree to work correctly - the unstable tree is liable to |
63 | * be corrupted as its contents are modified, and so it is called unstable. |
64 | * |
65 | * KSM solves this problem by several techniques: |
66 | * |
67 | * 1) The unstable tree is flushed every time KSM completes scanning all |
68 | * memory areas, and then the tree is rebuilt again from the beginning. |
69 | * 2) KSM will only insert into the unstable tree, pages whose hash value |
70 | * has not changed since the previous scan of all memory areas. |
71 | * 3) The unstable tree is a RedBlack Tree - so its balancing is based on the |
72 | * colors of the nodes and not on their contents, assuring that even when |
73 | * the tree gets "corrupted" it won't get out of balance, so scanning time |
74 | * remains the same (also, searching and inserting nodes in an rbtree uses |
75 | * the same algorithm, so we have no overhead when we flush and rebuild). |
76 | * 4) KSM never flushes the stable tree, which means that even if it were to |
77 | * take 10 attempts to find a page in the unstable tree, once it is found, |
78 | * it is secured in the stable tree. (When we scan a new page, we first |
79 | * compare it against the stable tree, and then against the unstable tree.) |
80 | */ |
81 | |
82 | /** |
83 | * struct mm_slot - ksm information per mm that is being scanned |
84 | * @link: link to the mm_slots hash list |
85 | * @mm_list: link into the mm_slots list, rooted in ksm_mm_head |
86 | * @rmap_list: head for this mm_slot's singly-linked list of rmap_items |
87 | * @mm: the mm that this information is valid for |
88 | */ |
89 | struct mm_slot { |
90 | struct hlist_node link; |
91 | struct list_head mm_list; |
92 | struct rmap_item *rmap_list; |
93 | struct mm_struct *mm; |
94 | }; |
95 | |
96 | /** |
97 | * struct ksm_scan - cursor for scanning |
98 | * @mm_slot: the current mm_slot we are scanning |
99 | * @address: the next address inside that to be scanned |
100 | * @rmap_list: link to the next rmap to be scanned in the rmap_list |
101 | * @seqnr: count of completed full scans (needed when removing unstable node) |
102 | * |
103 | * There is only the one ksm_scan instance of this cursor structure. |
104 | */ |
105 | struct ksm_scan { |
106 | struct mm_slot *mm_slot; |
107 | unsigned long address; |
108 | struct rmap_item **rmap_list; |
109 | unsigned long seqnr; |
110 | }; |
111 | |
112 | /** |
113 | * struct stable_node - node of the stable rbtree |
114 | * @node: rb node of this ksm page in the stable tree |
115 | * @hlist: hlist head of rmap_items using this ksm page |
116 | * @kpfn: page frame number of this ksm page |
117 | */ |
118 | struct stable_node { |
119 | struct rb_node node; |
120 | struct hlist_head hlist; |
121 | unsigned long kpfn; |
122 | }; |
123 | |
124 | /** |
125 | * struct rmap_item - reverse mapping item for virtual addresses |
126 | * @rmap_list: next rmap_item in mm_slot's singly-linked rmap_list |
127 | * @anon_vma: pointer to anon_vma for this mm,address, when in stable tree |
128 | * @mm: the memory structure this rmap_item is pointing into |
129 | * @address: the virtual address this rmap_item tracks (+ flags in low bits) |
130 | * @oldchecksum: previous checksum of the page at that virtual address |
131 | * @node: rb node of this rmap_item in the unstable tree |
132 | * @head: pointer to stable_node heading this list in the stable tree |
133 | * @hlist: link into hlist of rmap_items hanging off that stable_node |
134 | */ |
135 | struct rmap_item { |
136 | struct rmap_item *rmap_list; |
137 | struct anon_vma *anon_vma; /* when stable */ |
138 | struct mm_struct *mm; |
139 | unsigned long address; /* + low bits used for flags below */ |
140 | unsigned int oldchecksum; /* when unstable */ |
141 | union { |
142 | struct rb_node node; /* when node of unstable tree */ |
143 | struct { /* when listed from stable tree */ |
144 | struct stable_node *head; |
145 | struct hlist_node hlist; |
146 | }; |
147 | }; |
148 | }; |
149 | |
150 | #define SEQNR_MASK 0x0ff /* low bits of unstable tree seqnr */ |
151 | #define UNSTABLE_FLAG 0x100 /* is a node of the unstable tree */ |
152 | #define STABLE_FLAG 0x200 /* is listed from the stable tree */ |
153 | |
154 | /* The stable and unstable tree heads */ |
155 | static struct rb_root root_stable_tree = RB_ROOT; |
156 | static struct rb_root root_unstable_tree = RB_ROOT; |
157 | |
158 | #define MM_SLOTS_HASH_SHIFT 10 |
159 | #define MM_SLOTS_HASH_HEADS (1 << MM_SLOTS_HASH_SHIFT) |
160 | static struct hlist_head mm_slots_hash[MM_SLOTS_HASH_HEADS]; |
161 | |
162 | static struct mm_slot ksm_mm_head = { |
163 | .mm_list = LIST_HEAD_INIT(ksm_mm_head.mm_list), |
164 | }; |
165 | static struct ksm_scan ksm_scan = { |
166 | .mm_slot = &ksm_mm_head, |
167 | }; |
168 | |
169 | static struct kmem_cache *rmap_item_cache; |
170 | static struct kmem_cache *stable_node_cache; |
171 | static struct kmem_cache *mm_slot_cache; |
172 | |
173 | /* The number of nodes in the stable tree */ |
174 | static unsigned long ksm_pages_shared; |
175 | |
176 | /* The number of page slots additionally sharing those nodes */ |
177 | static unsigned long ksm_pages_sharing; |
178 | |
179 | /* The number of nodes in the unstable tree */ |
180 | static unsigned long ksm_pages_unshared; |
181 | |
182 | /* The number of rmap_items in use: to calculate pages_volatile */ |
183 | static unsigned long ksm_rmap_items; |
184 | |
185 | /* Number of pages ksmd should scan in one batch */ |
186 | static unsigned int ksm_thread_pages_to_scan = 100; |
187 | |
188 | /* Milliseconds ksmd should sleep between batches */ |
189 | static unsigned int ksm_thread_sleep_millisecs = 20; |
190 | |
191 | #define KSM_RUN_STOP 0 |
192 | #define KSM_RUN_MERGE 1 |
193 | #define KSM_RUN_UNMERGE 2 |
194 | static unsigned int ksm_run = KSM_RUN_STOP; |
195 | |
196 | static DECLARE_WAIT_QUEUE_HEAD(ksm_thread_wait); |
197 | static DEFINE_MUTEX(ksm_thread_mutex); |
198 | static DEFINE_SPINLOCK(ksm_mmlist_lock); |
199 | |
200 | #define KSM_KMEM_CACHE(__struct, __flags) kmem_cache_create("ksm_"#__struct,\ |
201 | sizeof(struct __struct), __alignof__(struct __struct),\ |
202 | (__flags), NULL) |
203 | |
204 | static int __init ksm_slab_init(void) |
205 | { |
206 | rmap_item_cache = KSM_KMEM_CACHE(rmap_item, 0); |
207 | if (!rmap_item_cache) |
208 | goto out; |
209 | |
210 | stable_node_cache = KSM_KMEM_CACHE(stable_node, 0); |
211 | if (!stable_node_cache) |
212 | goto out_free1; |
213 | |
214 | mm_slot_cache = KSM_KMEM_CACHE(mm_slot, 0); |
215 | if (!mm_slot_cache) |
216 | goto out_free2; |
217 | |
218 | return 0; |
219 | |
220 | out_free2: |
221 | kmem_cache_destroy(stable_node_cache); |
222 | out_free1: |
223 | kmem_cache_destroy(rmap_item_cache); |
224 | out: |
225 | return -ENOMEM; |
226 | } |
227 | |
228 | static void __init ksm_slab_free(void) |
229 | { |
230 | kmem_cache_destroy(mm_slot_cache); |
231 | kmem_cache_destroy(stable_node_cache); |
232 | kmem_cache_destroy(rmap_item_cache); |
233 | mm_slot_cache = NULL; |
234 | } |
235 | |
236 | static inline struct rmap_item *alloc_rmap_item(void) |
237 | { |
238 | struct rmap_item *rmap_item; |
239 | |
240 | rmap_item = kmem_cache_zalloc(rmap_item_cache, GFP_KERNEL); |
241 | if (rmap_item) |
242 | ksm_rmap_items++; |
243 | return rmap_item; |
244 | } |
245 | |
246 | static inline void free_rmap_item(struct rmap_item *rmap_item) |
247 | { |
248 | ksm_rmap_items--; |
249 | rmap_item->mm = NULL; /* debug safety */ |
250 | kmem_cache_free(rmap_item_cache, rmap_item); |
251 | } |
252 | |
253 | static inline struct stable_node *alloc_stable_node(void) |
254 | { |
255 | return kmem_cache_alloc(stable_node_cache, GFP_KERNEL); |
256 | } |
257 | |
258 | static inline void free_stable_node(struct stable_node *stable_node) |
259 | { |
260 | kmem_cache_free(stable_node_cache, stable_node); |
261 | } |
262 | |
263 | static inline struct mm_slot *alloc_mm_slot(void) |
264 | { |
265 | if (!mm_slot_cache) /* initialization failed */ |
266 | return NULL; |
267 | return kmem_cache_zalloc(mm_slot_cache, GFP_KERNEL); |
268 | } |
269 | |
270 | static inline void free_mm_slot(struct mm_slot *mm_slot) |
271 | { |
272 | kmem_cache_free(mm_slot_cache, mm_slot); |
273 | } |
274 | |
275 | static struct mm_slot *get_mm_slot(struct mm_struct *mm) |
276 | { |
277 | struct mm_slot *mm_slot; |
278 | struct hlist_head *bucket; |
279 | struct hlist_node *node; |
280 | |
281 | bucket = &mm_slots_hash[hash_ptr(mm, MM_SLOTS_HASH_SHIFT)]; |
282 | hlist_for_each_entry(mm_slot, node, bucket, link) { |
283 | if (mm == mm_slot->mm) |
284 | return mm_slot; |
285 | } |
286 | return NULL; |
287 | } |
288 | |
289 | static void insert_to_mm_slots_hash(struct mm_struct *mm, |
290 | struct mm_slot *mm_slot) |
291 | { |
292 | struct hlist_head *bucket; |
293 | |
294 | bucket = &mm_slots_hash[hash_ptr(mm, MM_SLOTS_HASH_SHIFT)]; |
295 | mm_slot->mm = mm; |
296 | hlist_add_head(&mm_slot->link, bucket); |
297 | } |
298 | |
299 | static inline int in_stable_tree(struct rmap_item *rmap_item) |
300 | { |
301 | return rmap_item->address & STABLE_FLAG; |
302 | } |
303 | |
304 | static void hold_anon_vma(struct rmap_item *rmap_item, |
305 | struct anon_vma *anon_vma) |
306 | { |
307 | rmap_item->anon_vma = anon_vma; |
308 | get_anon_vma(anon_vma); |
309 | } |
310 | |
311 | static void ksm_drop_anon_vma(struct rmap_item *rmap_item) |
312 | { |
313 | struct anon_vma *anon_vma = rmap_item->anon_vma; |
314 | |
315 | drop_anon_vma(anon_vma); |
316 | } |
317 | |
318 | /* |
319 | * ksmd, and unmerge_and_remove_all_rmap_items(), must not touch an mm's |
320 | * page tables after it has passed through ksm_exit() - which, if necessary, |
321 | * takes mmap_sem briefly to serialize against them. ksm_exit() does not set |
322 | * a special flag: they can just back out as soon as mm_users goes to zero. |
323 | * ksm_test_exit() is used throughout to make this test for exit: in some |
324 | * places for correctness, in some places just to avoid unnecessary work. |
325 | */ |
326 | static inline bool ksm_test_exit(struct mm_struct *mm) |
327 | { |
328 | return atomic_read(&mm->mm_users) == 0; |
329 | } |
330 | |
331 | /* |
332 | * We use break_ksm to break COW on a ksm page: it's a stripped down |
333 | * |
334 | * if (get_user_pages(current, mm, addr, 1, 1, 1, &page, NULL) == 1) |
335 | * put_page(page); |
336 | * |
337 | * but taking great care only to touch a ksm page, in a VM_MERGEABLE vma, |
338 | * in case the application has unmapped and remapped mm,addr meanwhile. |
339 | * Could a ksm page appear anywhere else? Actually yes, in a VM_PFNMAP |
340 | * mmap of /dev/mem or /dev/kmem, where we would not want to touch it. |
341 | */ |
342 | static int break_ksm(struct vm_area_struct *vma, unsigned long addr) |
343 | { |
344 | struct page *page; |
345 | int ret = 0; |
346 | |
347 | do { |
348 | cond_resched(); |
349 | page = follow_page(vma, addr, FOLL_GET); |
350 | if (IS_ERR_OR_NULL(page)) |
351 | break; |
352 | if (PageKsm(page)) |
353 | ret = handle_mm_fault(vma->vm_mm, vma, addr, |
354 | FAULT_FLAG_WRITE); |
355 | else |
356 | ret = VM_FAULT_WRITE; |
357 | put_page(page); |
358 | } while (!(ret & (VM_FAULT_WRITE | VM_FAULT_SIGBUS | VM_FAULT_OOM))); |
359 | /* |
360 | * We must loop because handle_mm_fault() may back out if there's |
361 | * any difficulty e.g. if pte accessed bit gets updated concurrently. |
362 | * |
363 | * VM_FAULT_WRITE is what we have been hoping for: it indicates that |
364 | * COW has been broken, even if the vma does not permit VM_WRITE; |
365 | * but note that a concurrent fault might break PageKsm for us. |
366 | * |
367 | * VM_FAULT_SIGBUS could occur if we race with truncation of the |
368 | * backing file, which also invalidates anonymous pages: that's |
369 | * okay, that truncation will have unmapped the PageKsm for us. |
370 | * |
371 | * VM_FAULT_OOM: at the time of writing (late July 2009), setting |
372 | * aside mem_cgroup limits, VM_FAULT_OOM would only be set if the |
373 | * current task has TIF_MEMDIE set, and will be OOM killed on return |
374 | * to user; and ksmd, having no mm, would never be chosen for that. |
375 | * |
376 | * But if the mm is in a limited mem_cgroup, then the fault may fail |
377 | * with VM_FAULT_OOM even if the current task is not TIF_MEMDIE; and |
378 | * even ksmd can fail in this way - though it's usually breaking ksm |
379 | * just to undo a merge it made a moment before, so unlikely to oom. |
380 | * |
381 | * That's a pity: we might therefore have more kernel pages allocated |
382 | * than we're counting as nodes in the stable tree; but ksm_do_scan |
383 | * will retry to break_cow on each pass, so should recover the page |
384 | * in due course. The important thing is to not let VM_MERGEABLE |
385 | * be cleared while any such pages might remain in the area. |
386 | */ |
387 | return (ret & VM_FAULT_OOM) ? -ENOMEM : 0; |
388 | } |
389 | |
390 | static void break_cow(struct rmap_item *rmap_item) |
391 | { |
392 | struct mm_struct *mm = rmap_item->mm; |
393 | unsigned long addr = rmap_item->address; |
394 | struct vm_area_struct *vma; |
395 | |
396 | /* |
397 | * It is not an accident that whenever we want to break COW |
398 | * to undo, we also need to drop a reference to the anon_vma. |
399 | */ |
400 | ksm_drop_anon_vma(rmap_item); |
401 | |
402 | down_read(&mm->mmap_sem); |
403 | if (ksm_test_exit(mm)) |
404 | goto out; |
405 | vma = find_vma(mm, addr); |
406 | if (!vma || vma->vm_start > addr) |
407 | goto out; |
408 | if (!(vma->vm_flags & VM_MERGEABLE) || !vma->anon_vma) |
409 | goto out; |
410 | break_ksm(vma, addr); |
411 | out: |
412 | up_read(&mm->mmap_sem); |
413 | } |
414 | |
415 | static struct page *page_trans_compound_anon(struct page *page) |
416 | { |
417 | if (PageTransCompound(page)) { |
418 | struct page *head = compound_trans_head(page); |
419 | /* |
420 | * head may actually be splitted and freed from under |
421 | * us but it's ok here. |
422 | */ |
423 | if (PageAnon(head)) |
424 | return head; |
425 | } |
426 | return NULL; |
427 | } |
428 | |
429 | static struct page *get_mergeable_page(struct rmap_item *rmap_item) |
430 | { |
431 | struct mm_struct *mm = rmap_item->mm; |
432 | unsigned long addr = rmap_item->address; |
433 | struct vm_area_struct *vma; |
434 | struct page *page; |
435 | |
436 | down_read(&mm->mmap_sem); |
437 | if (ksm_test_exit(mm)) |
438 | goto out; |
439 | vma = find_vma(mm, addr); |
440 | if (!vma || vma->vm_start > addr) |
441 | goto out; |
442 | if (!(vma->vm_flags & VM_MERGEABLE) || !vma->anon_vma) |
443 | goto out; |
444 | |
445 | page = follow_page(vma, addr, FOLL_GET); |
446 | if (IS_ERR_OR_NULL(page)) |
447 | goto out; |
448 | if (PageAnon(page) || page_trans_compound_anon(page)) { |
449 | flush_anon_page(vma, page, addr); |
450 | flush_dcache_page(page); |
451 | } else { |
452 | put_page(page); |
453 | out: page = NULL; |
454 | } |
455 | up_read(&mm->mmap_sem); |
456 | return page; |
457 | } |
458 | |
459 | static void remove_node_from_stable_tree(struct stable_node *stable_node) |
460 | { |
461 | struct rmap_item *rmap_item; |
462 | struct hlist_node *hlist; |
463 | |
464 | hlist_for_each_entry(rmap_item, hlist, &stable_node->hlist, hlist) { |
465 | if (rmap_item->hlist.next) |
466 | ksm_pages_sharing--; |
467 | else |
468 | ksm_pages_shared--; |
469 | ksm_drop_anon_vma(rmap_item); |
470 | rmap_item->address &= PAGE_MASK; |
471 | cond_resched(); |
472 | } |
473 | |
474 | rb_erase(&stable_node->node, &root_stable_tree); |
475 | free_stable_node(stable_node); |
476 | } |
477 | |
478 | /* |
479 | * get_ksm_page: checks if the page indicated by the stable node |
480 | * is still its ksm page, despite having held no reference to it. |
481 | * In which case we can trust the content of the page, and it |
482 | * returns the gotten page; but if the page has now been zapped, |
483 | * remove the stale node from the stable tree and return NULL. |
484 | * |
485 | * You would expect the stable_node to hold a reference to the ksm page. |
486 | * But if it increments the page's count, swapping out has to wait for |
487 | * ksmd to come around again before it can free the page, which may take |
488 | * seconds or even minutes: much too unresponsive. So instead we use a |
489 | * "keyhole reference": access to the ksm page from the stable node peeps |
490 | * out through its keyhole to see if that page still holds the right key, |
491 | * pointing back to this stable node. This relies on freeing a PageAnon |
492 | * page to reset its page->mapping to NULL, and relies on no other use of |
493 | * a page to put something that might look like our key in page->mapping. |
494 | * |
495 | * include/linux/pagemap.h page_cache_get_speculative() is a good reference, |
496 | * but this is different - made simpler by ksm_thread_mutex being held, but |
497 | * interesting for assuming that no other use of the struct page could ever |
498 | * put our expected_mapping into page->mapping (or a field of the union which |
499 | * coincides with page->mapping). The RCU calls are not for KSM at all, but |
500 | * to keep the page_count protocol described with page_cache_get_speculative. |
501 | * |
502 | * Note: it is possible that get_ksm_page() will return NULL one moment, |
503 | * then page the next, if the page is in between page_freeze_refs() and |
504 | * page_unfreeze_refs(): this shouldn't be a problem anywhere, the page |
505 | * is on its way to being freed; but it is an anomaly to bear in mind. |
506 | */ |
507 | static struct page *get_ksm_page(struct stable_node *stable_node) |
508 | { |
509 | struct page *page; |
510 | void *expected_mapping; |
511 | |
512 | page = pfn_to_page(stable_node->kpfn); |
513 | expected_mapping = (void *)stable_node + |
514 | (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM); |
515 | rcu_read_lock(); |
516 | if (page->mapping != expected_mapping) |
517 | goto stale; |
518 | if (!get_page_unless_zero(page)) |
519 | goto stale; |
520 | if (page->mapping != expected_mapping) { |
521 | put_page(page); |
522 | goto stale; |
523 | } |
524 | rcu_read_unlock(); |
525 | return page; |
526 | stale: |
527 | rcu_read_unlock(); |
528 | remove_node_from_stable_tree(stable_node); |
529 | return NULL; |
530 | } |
531 | |
532 | /* |
533 | * Removing rmap_item from stable or unstable tree. |
534 | * This function will clean the information from the stable/unstable tree. |
535 | */ |
536 | static void remove_rmap_item_from_tree(struct rmap_item *rmap_item) |
537 | { |
538 | if (rmap_item->address & STABLE_FLAG) { |
539 | struct stable_node *stable_node; |
540 | struct page *page; |
541 | |
542 | stable_node = rmap_item->head; |
543 | page = get_ksm_page(stable_node); |
544 | if (!page) |
545 | goto out; |
546 | |
547 | lock_page(page); |
548 | hlist_del(&rmap_item->hlist); |
549 | unlock_page(page); |
550 | put_page(page); |
551 | |
552 | if (stable_node->hlist.first) |
553 | ksm_pages_sharing--; |
554 | else |
555 | ksm_pages_shared--; |
556 | |
557 | ksm_drop_anon_vma(rmap_item); |
558 | rmap_item->address &= PAGE_MASK; |
559 | |
560 | } else if (rmap_item->address & UNSTABLE_FLAG) { |
561 | unsigned char age; |
562 | /* |
563 | * Usually ksmd can and must skip the rb_erase, because |
564 | * root_unstable_tree was already reset to RB_ROOT. |
565 | * But be careful when an mm is exiting: do the rb_erase |
566 | * if this rmap_item was inserted by this scan, rather |
567 | * than left over from before. |
568 | */ |
569 | age = (unsigned char)(ksm_scan.seqnr - rmap_item->address); |
570 | BUG_ON(age > 1); |
571 | if (!age) |
572 | rb_erase(&rmap_item->node, &root_unstable_tree); |
573 | |
574 | ksm_pages_unshared--; |
575 | rmap_item->address &= PAGE_MASK; |
576 | } |
577 | out: |
578 | cond_resched(); /* we're called from many long loops */ |
579 | } |
580 | |
581 | static void remove_trailing_rmap_items(struct mm_slot *mm_slot, |
582 | struct rmap_item **rmap_list) |
583 | { |
584 | while (*rmap_list) { |
585 | struct rmap_item *rmap_item = *rmap_list; |
586 | *rmap_list = rmap_item->rmap_list; |
587 | remove_rmap_item_from_tree(rmap_item); |
588 | free_rmap_item(rmap_item); |
589 | } |
590 | } |
591 | |
592 | /* |
593 | * Though it's very tempting to unmerge in_stable_tree(rmap_item)s rather |
594 | * than check every pte of a given vma, the locking doesn't quite work for |
595 | * that - an rmap_item is assigned to the stable tree after inserting ksm |
596 | * page and upping mmap_sem. Nor does it fit with the way we skip dup'ing |
597 | * rmap_items from parent to child at fork time (so as not to waste time |
598 | * if exit comes before the next scan reaches it). |
599 | * |
600 | * Similarly, although we'd like to remove rmap_items (so updating counts |
601 | * and freeing memory) when unmerging an area, it's easier to leave that |
602 | * to the next pass of ksmd - consider, for example, how ksmd might be |
603 | * in cmp_and_merge_page on one of the rmap_items we would be removing. |
604 | */ |
605 | static int unmerge_ksm_pages(struct vm_area_struct *vma, |
606 | unsigned long start, unsigned long end) |
607 | { |
608 | unsigned long addr; |
609 | int err = 0; |
610 | |
611 | for (addr = start; addr < end && !err; addr += PAGE_SIZE) { |
612 | if (ksm_test_exit(vma->vm_mm)) |
613 | break; |
614 | if (signal_pending(current)) |
615 | err = -ERESTARTSYS; |
616 | else |
617 | err = break_ksm(vma, addr); |
618 | } |
619 | return err; |
620 | } |
621 | |
622 | #ifdef CONFIG_SYSFS |
623 | /* |
624 | * Only called through the sysfs control interface: |
625 | */ |
626 | static int unmerge_and_remove_all_rmap_items(void) |
627 | { |
628 | struct mm_slot *mm_slot; |
629 | struct mm_struct *mm; |
630 | struct vm_area_struct *vma; |
631 | int err = 0; |
632 | |
633 | spin_lock(&ksm_mmlist_lock); |
634 | ksm_scan.mm_slot = list_entry(ksm_mm_head.mm_list.next, |
635 | struct mm_slot, mm_list); |
636 | spin_unlock(&ksm_mmlist_lock); |
637 | |
638 | for (mm_slot = ksm_scan.mm_slot; |
639 | mm_slot != &ksm_mm_head; mm_slot = ksm_scan.mm_slot) { |
640 | mm = mm_slot->mm; |
641 | down_read(&mm->mmap_sem); |
642 | for (vma = mm->mmap; vma; vma = vma->vm_next) { |
643 | if (ksm_test_exit(mm)) |
644 | break; |
645 | if (!(vma->vm_flags & VM_MERGEABLE) || !vma->anon_vma) |
646 | continue; |
647 | err = unmerge_ksm_pages(vma, |
648 | vma->vm_start, vma->vm_end); |
649 | if (err) |
650 | goto error; |
651 | } |
652 | |
653 | remove_trailing_rmap_items(mm_slot, &mm_slot->rmap_list); |
654 | |
655 | spin_lock(&ksm_mmlist_lock); |
656 | ksm_scan.mm_slot = list_entry(mm_slot->mm_list.next, |
657 | struct mm_slot, mm_list); |
658 | if (ksm_test_exit(mm)) { |
659 | hlist_del(&mm_slot->link); |
660 | list_del(&mm_slot->mm_list); |
661 | spin_unlock(&ksm_mmlist_lock); |
662 | |
663 | free_mm_slot(mm_slot); |
664 | clear_bit(MMF_VM_MERGEABLE, &mm->flags); |
665 | up_read(&mm->mmap_sem); |
666 | mmdrop(mm); |
667 | } else { |
668 | spin_unlock(&ksm_mmlist_lock); |
669 | up_read(&mm->mmap_sem); |
670 | } |
671 | } |
672 | |
673 | ksm_scan.seqnr = 0; |
674 | return 0; |
675 | |
676 | error: |
677 | up_read(&mm->mmap_sem); |
678 | spin_lock(&ksm_mmlist_lock); |
679 | ksm_scan.mm_slot = &ksm_mm_head; |
680 | spin_unlock(&ksm_mmlist_lock); |
681 | return err; |
682 | } |
683 | #endif /* CONFIG_SYSFS */ |
684 | |
685 | static u32 calc_checksum(struct page *page) |
686 | { |
687 | u32 checksum; |
688 | void *addr = kmap_atomic(page, KM_USER0); |
689 | checksum = jhash2(addr, PAGE_SIZE / 4, 17); |
690 | kunmap_atomic(addr, KM_USER0); |
691 | return checksum; |
692 | } |
693 | |
694 | static int memcmp_pages(struct page *page1, struct page *page2) |
695 | { |
696 | char *addr1, *addr2; |
697 | int ret; |
698 | |
699 | addr1 = kmap_atomic(page1, KM_USER0); |
700 | addr2 = kmap_atomic(page2, KM_USER1); |
701 | ret = memcmp(addr1, addr2, PAGE_SIZE); |
702 | kunmap_atomic(addr2, KM_USER1); |
703 | kunmap_atomic(addr1, KM_USER0); |
704 | return ret; |
705 | } |
706 | |
707 | static inline int pages_identical(struct page *page1, struct page *page2) |
708 | { |
709 | return !memcmp_pages(page1, page2); |
710 | } |
711 | |
712 | static int write_protect_page(struct vm_area_struct *vma, struct page *page, |
713 | pte_t *orig_pte) |
714 | { |
715 | struct mm_struct *mm = vma->vm_mm; |
716 | unsigned long addr; |
717 | pte_t *ptep; |
718 | spinlock_t *ptl; |
719 | int swapped; |
720 | int err = -EFAULT; |
721 | |
722 | addr = page_address_in_vma(page, vma); |
723 | if (addr == -EFAULT) |
724 | goto out; |
725 | |
726 | BUG_ON(PageTransCompound(page)); |
727 | ptep = page_check_address(page, mm, addr, &ptl, 0); |
728 | if (!ptep) |
729 | goto out; |
730 | |
731 | if (pte_write(*ptep) || pte_dirty(*ptep)) { |
732 | pte_t entry; |
733 | |
734 | swapped = PageSwapCache(page); |
735 | flush_cache_page(vma, addr, page_to_pfn(page)); |
736 | /* |
737 | * Ok this is tricky, when get_user_pages_fast() run it doesnt |
738 | * take any lock, therefore the check that we are going to make |
739 | * with the pagecount against the mapcount is racey and |
740 | * O_DIRECT can happen right after the check. |
741 | * So we clear the pte and flush the tlb before the check |
742 | * this assure us that no O_DIRECT can happen after the check |
743 | * or in the middle of the check. |
744 | */ |
745 | entry = ptep_clear_flush(vma, addr, ptep); |
746 | /* |
747 | * Check that no O_DIRECT or similar I/O is in progress on the |
748 | * page |
749 | */ |
750 | if (page_mapcount(page) + 1 + swapped != page_count(page)) { |
751 | set_pte_at(mm, addr, ptep, entry); |
752 | goto out_unlock; |
753 | } |
754 | if (pte_dirty(entry)) |
755 | set_page_dirty(page); |
756 | entry = pte_mkclean(pte_wrprotect(entry)); |
757 | set_pte_at_notify(mm, addr, ptep, entry); |
758 | } |
759 | *orig_pte = *ptep; |
760 | err = 0; |
761 | |
762 | out_unlock: |
763 | pte_unmap_unlock(ptep, ptl); |
764 | out: |
765 | return err; |
766 | } |
767 | |
768 | /** |
769 | * replace_page - replace page in vma by new ksm page |
770 | * @vma: vma that holds the pte pointing to page |
771 | * @page: the page we are replacing by kpage |
772 | * @kpage: the ksm page we replace page by |
773 | * @orig_pte: the original value of the pte |
774 | * |
775 | * Returns 0 on success, -EFAULT on failure. |
776 | */ |
777 | static int replace_page(struct vm_area_struct *vma, struct page *page, |
778 | struct page *kpage, pte_t orig_pte) |
779 | { |
780 | struct mm_struct *mm = vma->vm_mm; |
781 | pgd_t *pgd; |
782 | pud_t *pud; |
783 | pmd_t *pmd; |
784 | pte_t *ptep; |
785 | spinlock_t *ptl; |
786 | unsigned long addr; |
787 | int err = -EFAULT; |
788 | |
789 | addr = page_address_in_vma(page, vma); |
790 | if (addr == -EFAULT) |
791 | goto out; |
792 | |
793 | pgd = pgd_offset(mm, addr); |
794 | if (!pgd_present(*pgd)) |
795 | goto out; |
796 | |
797 | pud = pud_offset(pgd, addr); |
798 | if (!pud_present(*pud)) |
799 | goto out; |
800 | |
801 | pmd = pmd_offset(pud, addr); |
802 | BUG_ON(pmd_trans_huge(*pmd)); |
803 | if (!pmd_present(*pmd)) |
804 | goto out; |
805 | |
806 | ptep = pte_offset_map_lock(mm, pmd, addr, &ptl); |
807 | if (!pte_same(*ptep, orig_pte)) { |
808 | pte_unmap_unlock(ptep, ptl); |
809 | goto out; |
810 | } |
811 | |
812 | get_page(kpage); |
813 | page_add_anon_rmap(kpage, vma, addr); |
814 | |
815 | flush_cache_page(vma, addr, pte_pfn(*ptep)); |
816 | ptep_clear_flush(vma, addr, ptep); |
817 | set_pte_at_notify(mm, addr, ptep, mk_pte(kpage, vma->vm_page_prot)); |
818 | |
819 | page_remove_rmap(page); |
820 | if (!page_mapped(page)) |
821 | try_to_free_swap(page); |
822 | put_page(page); |
823 | |
824 | pte_unmap_unlock(ptep, ptl); |
825 | err = 0; |
826 | out: |
827 | return err; |
828 | } |
829 | |
830 | static int page_trans_compound_anon_split(struct page *page) |
831 | { |
832 | int ret = 0; |
833 | struct page *transhuge_head = page_trans_compound_anon(page); |
834 | if (transhuge_head) { |
835 | /* Get the reference on the head to split it. */ |
836 | if (get_page_unless_zero(transhuge_head)) { |
837 | /* |
838 | * Recheck we got the reference while the head |
839 | * was still anonymous. |
840 | */ |
841 | if (PageAnon(transhuge_head)) |
842 | ret = split_huge_page(transhuge_head); |
843 | else |
844 | /* |
845 | * Retry later if split_huge_page run |
846 | * from under us. |
847 | */ |
848 | ret = 1; |
849 | put_page(transhuge_head); |
850 | } else |
851 | /* Retry later if split_huge_page run from under us. */ |
852 | ret = 1; |
853 | } |
854 | return ret; |
855 | } |
856 | |
857 | /* |
858 | * try_to_merge_one_page - take two pages and merge them into one |
859 | * @vma: the vma that holds the pte pointing to page |
860 | * @page: the PageAnon page that we want to replace with kpage |
861 | * @kpage: the PageKsm page that we want to map instead of page, |
862 | * or NULL the first time when we want to use page as kpage. |
863 | * |
864 | * This function returns 0 if the pages were merged, -EFAULT otherwise. |
865 | */ |
866 | static int try_to_merge_one_page(struct vm_area_struct *vma, |
867 | struct page *page, struct page *kpage) |
868 | { |
869 | pte_t orig_pte = __pte(0); |
870 | int err = -EFAULT; |
871 | |
872 | if (page == kpage) /* ksm page forked */ |
873 | return 0; |
874 | |
875 | if (!(vma->vm_flags & VM_MERGEABLE)) |
876 | goto out; |
877 | if (PageTransCompound(page) && page_trans_compound_anon_split(page)) |
878 | goto out; |
879 | BUG_ON(PageTransCompound(page)); |
880 | if (!PageAnon(page)) |
881 | goto out; |
882 | |
883 | /* |
884 | * We need the page lock to read a stable PageSwapCache in |
885 | * write_protect_page(). We use trylock_page() instead of |
886 | * lock_page() because we don't want to wait here - we |
887 | * prefer to continue scanning and merging different pages, |
888 | * then come back to this page when it is unlocked. |
889 | */ |
890 | if (!trylock_page(page)) |
891 | goto out; |
892 | /* |
893 | * If this anonymous page is mapped only here, its pte may need |
894 | * to be write-protected. If it's mapped elsewhere, all of its |
895 | * ptes are necessarily already write-protected. But in either |
896 | * case, we need to lock and check page_count is not raised. |
897 | */ |
898 | if (write_protect_page(vma, page, &orig_pte) == 0) { |
899 | if (!kpage) { |
900 | /* |
901 | * While we hold page lock, upgrade page from |
902 | * PageAnon+anon_vma to PageKsm+NULL stable_node: |
903 | * stable_tree_insert() will update stable_node. |
904 | */ |
905 | set_page_stable_node(page, NULL); |
906 | mark_page_accessed(page); |
907 | err = 0; |
908 | } else if (pages_identical(page, kpage)) |
909 | err = replace_page(vma, page, kpage, orig_pte); |
910 | } |
911 | |
912 | if ((vma->vm_flags & VM_LOCKED) && kpage && !err) { |
913 | munlock_vma_page(page); |
914 | if (!PageMlocked(kpage)) { |
915 | unlock_page(page); |
916 | lock_page(kpage); |
917 | mlock_vma_page(kpage); |
918 | page = kpage; /* for final unlock */ |
919 | } |
920 | } |
921 | |
922 | unlock_page(page); |
923 | out: |
924 | return err; |
925 | } |
926 | |
927 | /* |
928 | * try_to_merge_with_ksm_page - like try_to_merge_two_pages, |
929 | * but no new kernel page is allocated: kpage must already be a ksm page. |
930 | * |
931 | * This function returns 0 if the pages were merged, -EFAULT otherwise. |
932 | */ |
933 | static int try_to_merge_with_ksm_page(struct rmap_item *rmap_item, |
934 | struct page *page, struct page *kpage) |
935 | { |
936 | struct mm_struct *mm = rmap_item->mm; |
937 | struct vm_area_struct *vma; |
938 | int err = -EFAULT; |
939 | |
940 | down_read(&mm->mmap_sem); |
941 | if (ksm_test_exit(mm)) |
942 | goto out; |
943 | vma = find_vma(mm, rmap_item->address); |
944 | if (!vma || vma->vm_start > rmap_item->address) |
945 | goto out; |
946 | |
947 | err = try_to_merge_one_page(vma, page, kpage); |
948 | if (err) |
949 | goto out; |
950 | |
951 | /* Must get reference to anon_vma while still holding mmap_sem */ |
952 | hold_anon_vma(rmap_item, vma->anon_vma); |
953 | out: |
954 | up_read(&mm->mmap_sem); |
955 | return err; |
956 | } |
957 | |
958 | /* |
959 | * try_to_merge_two_pages - take two identical pages and prepare them |
960 | * to be merged into one page. |
961 | * |
962 | * This function returns the kpage if we successfully merged two identical |
963 | * pages into one ksm page, NULL otherwise. |
964 | * |
965 | * Note that this function upgrades page to ksm page: if one of the pages |
966 | * is already a ksm page, try_to_merge_with_ksm_page should be used. |
967 | */ |
968 | static struct page *try_to_merge_two_pages(struct rmap_item *rmap_item, |
969 | struct page *page, |
970 | struct rmap_item *tree_rmap_item, |
971 | struct page *tree_page) |
972 | { |
973 | int err; |
974 | |
975 | err = try_to_merge_with_ksm_page(rmap_item, page, NULL); |
976 | if (!err) { |
977 | err = try_to_merge_with_ksm_page(tree_rmap_item, |
978 | tree_page, page); |
979 | /* |
980 | * If that fails, we have a ksm page with only one pte |
981 | * pointing to it: so break it. |
982 | */ |
983 | if (err) |
984 | break_cow(rmap_item); |
985 | } |
986 | return err ? NULL : page; |
987 | } |
988 | |
989 | /* |
990 | * stable_tree_search - search for page inside the stable tree |
991 | * |
992 | * This function checks if there is a page inside the stable tree |
993 | * with identical content to the page that we are scanning right now. |
994 | * |
995 | * This function returns the stable tree node of identical content if found, |
996 | * NULL otherwise. |
997 | */ |
998 | static struct page *stable_tree_search(struct page *page) |
999 | { |
1000 | struct rb_node *node = root_stable_tree.rb_node; |
1001 | struct stable_node *stable_node; |
1002 | |
1003 | stable_node = page_stable_node(page); |
1004 | if (stable_node) { /* ksm page forked */ |
1005 | get_page(page); |
1006 | return page; |
1007 | } |
1008 | |
1009 | while (node) { |
1010 | struct page *tree_page; |
1011 | int ret; |
1012 | |
1013 | cond_resched(); |
1014 | stable_node = rb_entry(node, struct stable_node, node); |
1015 | tree_page = get_ksm_page(stable_node); |
1016 | if (!tree_page) |
1017 | return NULL; |
1018 | |
1019 | ret = memcmp_pages(page, tree_page); |
1020 | |
1021 | if (ret < 0) { |
1022 | put_page(tree_page); |
1023 | node = node->rb_left; |
1024 | } else if (ret > 0) { |
1025 | put_page(tree_page); |
1026 | node = node->rb_right; |
1027 | } else |
1028 | return tree_page; |
1029 | } |
1030 | |
1031 | return NULL; |
1032 | } |
1033 | |
1034 | /* |
1035 | * stable_tree_insert - insert rmap_item pointing to new ksm page |
1036 | * into the stable tree. |
1037 | * |
1038 | * This function returns the stable tree node just allocated on success, |
1039 | * NULL otherwise. |
1040 | */ |
1041 | static struct stable_node *stable_tree_insert(struct page *kpage) |
1042 | { |
1043 | struct rb_node **new = &root_stable_tree.rb_node; |
1044 | struct rb_node *parent = NULL; |
1045 | struct stable_node *stable_node; |
1046 | |
1047 | while (*new) { |
1048 | struct page *tree_page; |
1049 | int ret; |
1050 | |
1051 | cond_resched(); |
1052 | stable_node = rb_entry(*new, struct stable_node, node); |
1053 | tree_page = get_ksm_page(stable_node); |
1054 | if (!tree_page) |
1055 | return NULL; |
1056 | |
1057 | ret = memcmp_pages(kpage, tree_page); |
1058 | put_page(tree_page); |
1059 | |
1060 | parent = *new; |
1061 | if (ret < 0) |
1062 | new = &parent->rb_left; |
1063 | else if (ret > 0) |
1064 | new = &parent->rb_right; |
1065 | else { |
1066 | /* |
1067 | * It is not a bug that stable_tree_search() didn't |
1068 | * find this node: because at that time our page was |
1069 | * not yet write-protected, so may have changed since. |
1070 | */ |
1071 | return NULL; |
1072 | } |
1073 | } |
1074 | |
1075 | stable_node = alloc_stable_node(); |
1076 | if (!stable_node) |
1077 | return NULL; |
1078 | |
1079 | rb_link_node(&stable_node->node, parent, new); |
1080 | rb_insert_color(&stable_node->node, &root_stable_tree); |
1081 | |
1082 | INIT_HLIST_HEAD(&stable_node->hlist); |
1083 | |
1084 | stable_node->kpfn = page_to_pfn(kpage); |
1085 | set_page_stable_node(kpage, stable_node); |
1086 | |
1087 | return stable_node; |
1088 | } |
1089 | |
1090 | /* |
1091 | * unstable_tree_search_insert - search for identical page, |
1092 | * else insert rmap_item into the unstable tree. |
1093 | * |
1094 | * This function searches for a page in the unstable tree identical to the |
1095 | * page currently being scanned; and if no identical page is found in the |
1096 | * tree, we insert rmap_item as a new object into the unstable tree. |
1097 | * |
1098 | * This function returns pointer to rmap_item found to be identical |
1099 | * to the currently scanned page, NULL otherwise. |
1100 | * |
1101 | * This function does both searching and inserting, because they share |
1102 | * the same walking algorithm in an rbtree. |
1103 | */ |
1104 | static |
1105 | struct rmap_item *unstable_tree_search_insert(struct rmap_item *rmap_item, |
1106 | struct page *page, |
1107 | struct page **tree_pagep) |
1108 | |
1109 | { |
1110 | struct rb_node **new = &root_unstable_tree.rb_node; |
1111 | struct rb_node *parent = NULL; |
1112 | |
1113 | while (*new) { |
1114 | struct rmap_item *tree_rmap_item; |
1115 | struct page *tree_page; |
1116 | int ret; |
1117 | |
1118 | cond_resched(); |
1119 | tree_rmap_item = rb_entry(*new, struct rmap_item, node); |
1120 | tree_page = get_mergeable_page(tree_rmap_item); |
1121 | if (IS_ERR_OR_NULL(tree_page)) |
1122 | return NULL; |
1123 | |
1124 | /* |
1125 | * Don't substitute a ksm page for a forked page. |
1126 | */ |
1127 | if (page == tree_page) { |
1128 | put_page(tree_page); |
1129 | return NULL; |
1130 | } |
1131 | |
1132 | ret = memcmp_pages(page, tree_page); |
1133 | |
1134 | parent = *new; |
1135 | if (ret < 0) { |
1136 | put_page(tree_page); |
1137 | new = &parent->rb_left; |
1138 | } else if (ret > 0) { |
1139 | put_page(tree_page); |
1140 | new = &parent->rb_right; |
1141 | } else { |
1142 | *tree_pagep = tree_page; |
1143 | return tree_rmap_item; |
1144 | } |
1145 | } |
1146 | |
1147 | rmap_item->address |= UNSTABLE_FLAG; |
1148 | rmap_item->address |= (ksm_scan.seqnr & SEQNR_MASK); |
1149 | rb_link_node(&rmap_item->node, parent, new); |
1150 | rb_insert_color(&rmap_item->node, &root_unstable_tree); |
1151 | |
1152 | ksm_pages_unshared++; |
1153 | return NULL; |
1154 | } |
1155 | |
1156 | /* |
1157 | * stable_tree_append - add another rmap_item to the linked list of |
1158 | * rmap_items hanging off a given node of the stable tree, all sharing |
1159 | * the same ksm page. |
1160 | */ |
1161 | static void stable_tree_append(struct rmap_item *rmap_item, |
1162 | struct stable_node *stable_node) |
1163 | { |
1164 | rmap_item->head = stable_node; |
1165 | rmap_item->address |= STABLE_FLAG; |
1166 | hlist_add_head(&rmap_item->hlist, &stable_node->hlist); |
1167 | |
1168 | if (rmap_item->hlist.next) |
1169 | ksm_pages_sharing++; |
1170 | else |
1171 | ksm_pages_shared++; |
1172 | } |
1173 | |
1174 | /* |
1175 | * cmp_and_merge_page - first see if page can be merged into the stable tree; |
1176 | * if not, compare checksum to previous and if it's the same, see if page can |
1177 | * be inserted into the unstable tree, or merged with a page already there and |
1178 | * both transferred to the stable tree. |
1179 | * |
1180 | * @page: the page that we are searching identical page to. |
1181 | * @rmap_item: the reverse mapping into the virtual address of this page |
1182 | */ |
1183 | static void cmp_and_merge_page(struct page *page, struct rmap_item *rmap_item) |
1184 | { |
1185 | struct rmap_item *tree_rmap_item; |
1186 | struct page *tree_page = NULL; |
1187 | struct stable_node *stable_node; |
1188 | struct page *kpage; |
1189 | unsigned int checksum; |
1190 | int err; |
1191 | |
1192 | remove_rmap_item_from_tree(rmap_item); |
1193 | |
1194 | /* We first start with searching the page inside the stable tree */ |
1195 | kpage = stable_tree_search(page); |
1196 | if (kpage) { |
1197 | err = try_to_merge_with_ksm_page(rmap_item, page, kpage); |
1198 | if (!err) { |
1199 | /* |
1200 | * The page was successfully merged: |
1201 | * add its rmap_item to the stable tree. |
1202 | */ |
1203 | lock_page(kpage); |
1204 | stable_tree_append(rmap_item, page_stable_node(kpage)); |
1205 | unlock_page(kpage); |
1206 | } |
1207 | put_page(kpage); |
1208 | return; |
1209 | } |
1210 | |
1211 | /* |
1212 | * If the hash value of the page has changed from the last time |
1213 | * we calculated it, this page is changing frequently: therefore we |
1214 | * don't want to insert it in the unstable tree, and we don't want |
1215 | * to waste our time searching for something identical to it there. |
1216 | */ |
1217 | checksum = calc_checksum(page); |
1218 | if (rmap_item->oldchecksum != checksum) { |
1219 | rmap_item->oldchecksum = checksum; |
1220 | return; |
1221 | } |
1222 | |
1223 | tree_rmap_item = |
1224 | unstable_tree_search_insert(rmap_item, page, &tree_page); |
1225 | if (tree_rmap_item) { |
1226 | kpage = try_to_merge_two_pages(rmap_item, page, |
1227 | tree_rmap_item, tree_page); |
1228 | put_page(tree_page); |
1229 | /* |
1230 | * As soon as we merge this page, we want to remove the |
1231 | * rmap_item of the page we have merged with from the unstable |
1232 | * tree, and insert it instead as new node in the stable tree. |
1233 | */ |
1234 | if (kpage) { |
1235 | remove_rmap_item_from_tree(tree_rmap_item); |
1236 | |
1237 | lock_page(kpage); |
1238 | stable_node = stable_tree_insert(kpage); |
1239 | if (stable_node) { |
1240 | stable_tree_append(tree_rmap_item, stable_node); |
1241 | stable_tree_append(rmap_item, stable_node); |
1242 | } |
1243 | unlock_page(kpage); |
1244 | |
1245 | /* |
1246 | * If we fail to insert the page into the stable tree, |
1247 | * we will have 2 virtual addresses that are pointing |
1248 | * to a ksm page left outside the stable tree, |
1249 | * in which case we need to break_cow on both. |
1250 | */ |
1251 | if (!stable_node) { |
1252 | break_cow(tree_rmap_item); |
1253 | break_cow(rmap_item); |
1254 | } |
1255 | } |
1256 | } |
1257 | } |
1258 | |
1259 | static struct rmap_item *get_next_rmap_item(struct mm_slot *mm_slot, |
1260 | struct rmap_item **rmap_list, |
1261 | unsigned long addr) |
1262 | { |
1263 | struct rmap_item *rmap_item; |
1264 | |
1265 | while (*rmap_list) { |
1266 | rmap_item = *rmap_list; |
1267 | if ((rmap_item->address & PAGE_MASK) == addr) |
1268 | return rmap_item; |
1269 | if (rmap_item->address > addr) |
1270 | break; |
1271 | *rmap_list = rmap_item->rmap_list; |
1272 | remove_rmap_item_from_tree(rmap_item); |
1273 | free_rmap_item(rmap_item); |
1274 | } |
1275 | |
1276 | rmap_item = alloc_rmap_item(); |
1277 | if (rmap_item) { |
1278 | /* It has already been zeroed */ |
1279 | rmap_item->mm = mm_slot->mm; |
1280 | rmap_item->address = addr; |
1281 | rmap_item->rmap_list = *rmap_list; |
1282 | *rmap_list = rmap_item; |
1283 | } |
1284 | return rmap_item; |
1285 | } |
1286 | |
1287 | static struct rmap_item *scan_get_next_rmap_item(struct page **page) |
1288 | { |
1289 | struct mm_struct *mm; |
1290 | struct mm_slot *slot; |
1291 | struct vm_area_struct *vma; |
1292 | struct rmap_item *rmap_item; |
1293 | |
1294 | if (list_empty(&ksm_mm_head.mm_list)) |
1295 | return NULL; |
1296 | |
1297 | slot = ksm_scan.mm_slot; |
1298 | if (slot == &ksm_mm_head) { |
1299 | /* |
1300 | * A number of pages can hang around indefinitely on per-cpu |
1301 | * pagevecs, raised page count preventing write_protect_page |
1302 | * from merging them. Though it doesn't really matter much, |
1303 | * it is puzzling to see some stuck in pages_volatile until |
1304 | * other activity jostles them out, and they also prevented |
1305 | * LTP's KSM test from succeeding deterministically; so drain |
1306 | * them here (here rather than on entry to ksm_do_scan(), |
1307 | * so we don't IPI too often when pages_to_scan is set low). |
1308 | */ |
1309 | lru_add_drain_all(); |
1310 | |
1311 | root_unstable_tree = RB_ROOT; |
1312 | |
1313 | spin_lock(&ksm_mmlist_lock); |
1314 | slot = list_entry(slot->mm_list.next, struct mm_slot, mm_list); |
1315 | ksm_scan.mm_slot = slot; |
1316 | spin_unlock(&ksm_mmlist_lock); |
1317 | next_mm: |
1318 | ksm_scan.address = 0; |
1319 | ksm_scan.rmap_list = &slot->rmap_list; |
1320 | } |
1321 | |
1322 | mm = slot->mm; |
1323 | down_read(&mm->mmap_sem); |
1324 | if (ksm_test_exit(mm)) |
1325 | vma = NULL; |
1326 | else |
1327 | vma = find_vma(mm, ksm_scan.address); |
1328 | |
1329 | for (; vma; vma = vma->vm_next) { |
1330 | if (!(vma->vm_flags & VM_MERGEABLE)) |
1331 | continue; |
1332 | if (ksm_scan.address < vma->vm_start) |
1333 | ksm_scan.address = vma->vm_start; |
1334 | if (!vma->anon_vma) |
1335 | ksm_scan.address = vma->vm_end; |
1336 | |
1337 | while (ksm_scan.address < vma->vm_end) { |
1338 | if (ksm_test_exit(mm)) |
1339 | break; |
1340 | *page = follow_page(vma, ksm_scan.address, FOLL_GET); |
1341 | if (IS_ERR_OR_NULL(*page)) { |
1342 | ksm_scan.address += PAGE_SIZE; |
1343 | cond_resched(); |
1344 | continue; |
1345 | } |
1346 | if (PageAnon(*page) || |
1347 | page_trans_compound_anon(*page)) { |
1348 | flush_anon_page(vma, *page, ksm_scan.address); |
1349 | flush_dcache_page(*page); |
1350 | rmap_item = get_next_rmap_item(slot, |
1351 | ksm_scan.rmap_list, ksm_scan.address); |
1352 | if (rmap_item) { |
1353 | ksm_scan.rmap_list = |
1354 | &rmap_item->rmap_list; |
1355 | ksm_scan.address += PAGE_SIZE; |
1356 | } else |
1357 | put_page(*page); |
1358 | up_read(&mm->mmap_sem); |
1359 | return rmap_item; |
1360 | } |
1361 | put_page(*page); |
1362 | ksm_scan.address += PAGE_SIZE; |
1363 | cond_resched(); |
1364 | } |
1365 | } |
1366 | |
1367 | if (ksm_test_exit(mm)) { |
1368 | ksm_scan.address = 0; |
1369 | ksm_scan.rmap_list = &slot->rmap_list; |
1370 | } |
1371 | /* |
1372 | * Nuke all the rmap_items that are above this current rmap: |
1373 | * because there were no VM_MERGEABLE vmas with such addresses. |
1374 | */ |
1375 | remove_trailing_rmap_items(slot, ksm_scan.rmap_list); |
1376 | |
1377 | spin_lock(&ksm_mmlist_lock); |
1378 | ksm_scan.mm_slot = list_entry(slot->mm_list.next, |
1379 | struct mm_slot, mm_list); |
1380 | if (ksm_scan.address == 0) { |
1381 | /* |
1382 | * We've completed a full scan of all vmas, holding mmap_sem |
1383 | * throughout, and found no VM_MERGEABLE: so do the same as |
1384 | * __ksm_exit does to remove this mm from all our lists now. |
1385 | * This applies either when cleaning up after __ksm_exit |
1386 | * (but beware: we can reach here even before __ksm_exit), |
1387 | * or when all VM_MERGEABLE areas have been unmapped (and |
1388 | * mmap_sem then protects against race with MADV_MERGEABLE). |
1389 | */ |
1390 | hlist_del(&slot->link); |
1391 | list_del(&slot->mm_list); |
1392 | spin_unlock(&ksm_mmlist_lock); |
1393 | |
1394 | free_mm_slot(slot); |
1395 | clear_bit(MMF_VM_MERGEABLE, &mm->flags); |
1396 | up_read(&mm->mmap_sem); |
1397 | mmdrop(mm); |
1398 | } else { |
1399 | spin_unlock(&ksm_mmlist_lock); |
1400 | up_read(&mm->mmap_sem); |
1401 | } |
1402 | |
1403 | /* Repeat until we've completed scanning the whole list */ |
1404 | slot = ksm_scan.mm_slot; |
1405 | if (slot != &ksm_mm_head) |
1406 | goto next_mm; |
1407 | |
1408 | ksm_scan.seqnr++; |
1409 | return NULL; |
1410 | } |
1411 | |
1412 | /** |
1413 | * ksm_do_scan - the ksm scanner main worker function. |
1414 | * @scan_npages - number of pages we want to scan before we return. |
1415 | */ |
1416 | static void ksm_do_scan(unsigned int scan_npages) |
1417 | { |
1418 | struct rmap_item *rmap_item; |
1419 | struct page *uninitialized_var(page); |
1420 | |
1421 | while (scan_npages-- && likely(!freezing(current))) { |
1422 | cond_resched(); |
1423 | rmap_item = scan_get_next_rmap_item(&page); |
1424 | if (!rmap_item) |
1425 | return; |
1426 | if (!PageKsm(page) || !in_stable_tree(rmap_item)) |
1427 | cmp_and_merge_page(page, rmap_item); |
1428 | put_page(page); |
1429 | } |
1430 | } |
1431 | |
1432 | static int ksmd_should_run(void) |
1433 | { |
1434 | return (ksm_run & KSM_RUN_MERGE) && !list_empty(&ksm_mm_head.mm_list); |
1435 | } |
1436 | |
1437 | static int ksm_scan_thread(void *nothing) |
1438 | { |
1439 | set_freezable(); |
1440 | set_user_nice(current, 5); |
1441 | |
1442 | while (!kthread_should_stop()) { |
1443 | mutex_lock(&ksm_thread_mutex); |
1444 | if (ksmd_should_run()) |
1445 | ksm_do_scan(ksm_thread_pages_to_scan); |
1446 | mutex_unlock(&ksm_thread_mutex); |
1447 | |
1448 | try_to_freeze(); |
1449 | |
1450 | if (ksmd_should_run()) { |
1451 | schedule_timeout_interruptible( |
1452 | msecs_to_jiffies(ksm_thread_sleep_millisecs)); |
1453 | } else { |
1454 | wait_event_freezable(ksm_thread_wait, |
1455 | ksmd_should_run() || kthread_should_stop()); |
1456 | } |
1457 | } |
1458 | return 0; |
1459 | } |
1460 | |
1461 | int ksm_madvise(struct vm_area_struct *vma, unsigned long start, |
1462 | unsigned long end, int advice, unsigned long *vm_flags) |
1463 | { |
1464 | struct mm_struct *mm = vma->vm_mm; |
1465 | int err; |
1466 | |
1467 | switch (advice) { |
1468 | case MADV_MERGEABLE: |
1469 | /* |
1470 | * Be somewhat over-protective for now! |
1471 | */ |
1472 | if (*vm_flags & (VM_MERGEABLE | VM_SHARED | VM_MAYSHARE | |
1473 | VM_PFNMAP | VM_IO | VM_DONTEXPAND | |
1474 | VM_RESERVED | VM_HUGETLB | VM_INSERTPAGE | |
1475 | VM_NONLINEAR | VM_MIXEDMAP | VM_SAO)) |
1476 | return 0; /* just ignore the advice */ |
1477 | |
1478 | if (!test_bit(MMF_VM_MERGEABLE, &mm->flags)) { |
1479 | err = __ksm_enter(mm); |
1480 | if (err) |
1481 | return err; |
1482 | } |
1483 | |
1484 | *vm_flags |= VM_MERGEABLE; |
1485 | break; |
1486 | |
1487 | case MADV_UNMERGEABLE: |
1488 | if (!(*vm_flags & VM_MERGEABLE)) |
1489 | return 0; /* just ignore the advice */ |
1490 | |
1491 | if (vma->anon_vma) { |
1492 | err = unmerge_ksm_pages(vma, start, end); |
1493 | if (err) |
1494 | return err; |
1495 | } |
1496 | |
1497 | *vm_flags &= ~VM_MERGEABLE; |
1498 | break; |
1499 | } |
1500 | |
1501 | return 0; |
1502 | } |
1503 | |
1504 | int __ksm_enter(struct mm_struct *mm) |
1505 | { |
1506 | struct mm_slot *mm_slot; |
1507 | int needs_wakeup; |
1508 | |
1509 | mm_slot = alloc_mm_slot(); |
1510 | if (!mm_slot) |
1511 | return -ENOMEM; |
1512 | |
1513 | /* Check ksm_run too? Would need tighter locking */ |
1514 | needs_wakeup = list_empty(&ksm_mm_head.mm_list); |
1515 | |
1516 | spin_lock(&ksm_mmlist_lock); |
1517 | insert_to_mm_slots_hash(mm, mm_slot); |
1518 | /* |
1519 | * Insert just behind the scanning cursor, to let the area settle |
1520 | * down a little; when fork is followed by immediate exec, we don't |
1521 | * want ksmd to waste time setting up and tearing down an rmap_list. |
1522 | */ |
1523 | list_add_tail(&mm_slot->mm_list, &ksm_scan.mm_slot->mm_list); |
1524 | spin_unlock(&ksm_mmlist_lock); |
1525 | |
1526 | set_bit(MMF_VM_MERGEABLE, &mm->flags); |
1527 | atomic_inc(&mm->mm_count); |
1528 | |
1529 | if (needs_wakeup) |
1530 | wake_up_interruptible(&ksm_thread_wait); |
1531 | |
1532 | return 0; |
1533 | } |
1534 | |
1535 | void __ksm_exit(struct mm_struct *mm) |
1536 | { |
1537 | struct mm_slot *mm_slot; |
1538 | int easy_to_free = 0; |
1539 | |
1540 | /* |
1541 | * This process is exiting: if it's straightforward (as is the |
1542 | * case when ksmd was never running), free mm_slot immediately. |
1543 | * But if it's at the cursor or has rmap_items linked to it, use |
1544 | * mmap_sem to synchronize with any break_cows before pagetables |
1545 | * are freed, and leave the mm_slot on the list for ksmd to free. |
1546 | * Beware: ksm may already have noticed it exiting and freed the slot. |
1547 | */ |
1548 | |
1549 | spin_lock(&ksm_mmlist_lock); |
1550 | mm_slot = get_mm_slot(mm); |
1551 | if (mm_slot && ksm_scan.mm_slot != mm_slot) { |
1552 | if (!mm_slot->rmap_list) { |
1553 | hlist_del(&mm_slot->link); |
1554 | list_del(&mm_slot->mm_list); |
1555 | easy_to_free = 1; |
1556 | } else { |
1557 | list_move(&mm_slot->mm_list, |
1558 | &ksm_scan.mm_slot->mm_list); |
1559 | } |
1560 | } |
1561 | spin_unlock(&ksm_mmlist_lock); |
1562 | |
1563 | if (easy_to_free) { |
1564 | free_mm_slot(mm_slot); |
1565 | clear_bit(MMF_VM_MERGEABLE, &mm->flags); |
1566 | mmdrop(mm); |
1567 | } else if (mm_slot) { |
1568 | down_write(&mm->mmap_sem); |
1569 | up_write(&mm->mmap_sem); |
1570 | } |
1571 | } |
1572 | |
1573 | struct page *ksm_does_need_to_copy(struct page *page, |
1574 | struct vm_area_struct *vma, unsigned long address) |
1575 | { |
1576 | struct page *new_page; |
1577 | |
1578 | new_page = alloc_page_vma(GFP_HIGHUSER_MOVABLE, vma, address); |
1579 | if (new_page) { |
1580 | copy_user_highpage(new_page, page, address, vma); |
1581 | |
1582 | SetPageDirty(new_page); |
1583 | __SetPageUptodate(new_page); |
1584 | SetPageSwapBacked(new_page); |
1585 | __set_page_locked(new_page); |
1586 | |
1587 | if (page_evictable(new_page, vma)) |
1588 | lru_cache_add_lru(new_page, LRU_ACTIVE_ANON); |
1589 | else |
1590 | add_page_to_unevictable_list(new_page); |
1591 | } |
1592 | |
1593 | return new_page; |
1594 | } |
1595 | |
1596 | int page_referenced_ksm(struct page *page, struct mem_cgroup *memcg, |
1597 | unsigned long *vm_flags) |
1598 | { |
1599 | struct stable_node *stable_node; |
1600 | struct rmap_item *rmap_item; |
1601 | struct hlist_node *hlist; |
1602 | unsigned int mapcount = page_mapcount(page); |
1603 | int referenced = 0; |
1604 | int search_new_forks = 0; |
1605 | |
1606 | VM_BUG_ON(!PageKsm(page)); |
1607 | VM_BUG_ON(!PageLocked(page)); |
1608 | |
1609 | stable_node = page_stable_node(page); |
1610 | if (!stable_node) |
1611 | return 0; |
1612 | again: |
1613 | hlist_for_each_entry(rmap_item, hlist, &stable_node->hlist, hlist) { |
1614 | struct anon_vma *anon_vma = rmap_item->anon_vma; |
1615 | struct anon_vma_chain *vmac; |
1616 | struct vm_area_struct *vma; |
1617 | |
1618 | anon_vma_lock(anon_vma); |
1619 | list_for_each_entry(vmac, &anon_vma->head, same_anon_vma) { |
1620 | vma = vmac->vma; |
1621 | if (rmap_item->address < vma->vm_start || |
1622 | rmap_item->address >= vma->vm_end) |
1623 | continue; |
1624 | /* |
1625 | * Initially we examine only the vma which covers this |
1626 | * rmap_item; but later, if there is still work to do, |
1627 | * we examine covering vmas in other mms: in case they |
1628 | * were forked from the original since ksmd passed. |
1629 | */ |
1630 | if ((rmap_item->mm == vma->vm_mm) == search_new_forks) |
1631 | continue; |
1632 | |
1633 | if (memcg && !mm_match_cgroup(vma->vm_mm, memcg)) |
1634 | continue; |
1635 | |
1636 | referenced += page_referenced_one(page, vma, |
1637 | rmap_item->address, &mapcount, vm_flags); |
1638 | if (!search_new_forks || !mapcount) |
1639 | break; |
1640 | } |
1641 | anon_vma_unlock(anon_vma); |
1642 | if (!mapcount) |
1643 | goto out; |
1644 | } |
1645 | if (!search_new_forks++) |
1646 | goto again; |
1647 | out: |
1648 | return referenced; |
1649 | } |
1650 | |
1651 | int try_to_unmap_ksm(struct page *page, enum ttu_flags flags) |
1652 | { |
1653 | struct stable_node *stable_node; |
1654 | struct hlist_node *hlist; |
1655 | struct rmap_item *rmap_item; |
1656 | int ret = SWAP_AGAIN; |
1657 | int search_new_forks = 0; |
1658 | |
1659 | VM_BUG_ON(!PageKsm(page)); |
1660 | VM_BUG_ON(!PageLocked(page)); |
1661 | |
1662 | stable_node = page_stable_node(page); |
1663 | if (!stable_node) |
1664 | return SWAP_FAIL; |
1665 | again: |
1666 | hlist_for_each_entry(rmap_item, hlist, &stable_node->hlist, hlist) { |
1667 | struct anon_vma *anon_vma = rmap_item->anon_vma; |
1668 | struct anon_vma_chain *vmac; |
1669 | struct vm_area_struct *vma; |
1670 | |
1671 | anon_vma_lock(anon_vma); |
1672 | list_for_each_entry(vmac, &anon_vma->head, same_anon_vma) { |
1673 | vma = vmac->vma; |
1674 | if (rmap_item->address < vma->vm_start || |
1675 | rmap_item->address >= vma->vm_end) |
1676 | continue; |
1677 | /* |
1678 | * Initially we examine only the vma which covers this |
1679 | * rmap_item; but later, if there is still work to do, |
1680 | * we examine covering vmas in other mms: in case they |
1681 | * were forked from the original since ksmd passed. |
1682 | */ |
1683 | if ((rmap_item->mm == vma->vm_mm) == search_new_forks) |
1684 | continue; |
1685 | |
1686 | ret = try_to_unmap_one(page, vma, |
1687 | rmap_item->address, flags); |
1688 | if (ret != SWAP_AGAIN || !page_mapped(page)) { |
1689 | anon_vma_unlock(anon_vma); |
1690 | goto out; |
1691 | } |
1692 | } |
1693 | anon_vma_unlock(anon_vma); |
1694 | } |
1695 | if (!search_new_forks++) |
1696 | goto again; |
1697 | out: |
1698 | return ret; |
1699 | } |
1700 | |
1701 | #ifdef CONFIG_MIGRATION |
1702 | int rmap_walk_ksm(struct page *page, int (*rmap_one)(struct page *, |
1703 | struct vm_area_struct *, unsigned long, void *), void *arg) |
1704 | { |
1705 | struct stable_node *stable_node; |
1706 | struct hlist_node *hlist; |
1707 | struct rmap_item *rmap_item; |
1708 | int ret = SWAP_AGAIN; |
1709 | int search_new_forks = 0; |
1710 | |
1711 | VM_BUG_ON(!PageKsm(page)); |
1712 | VM_BUG_ON(!PageLocked(page)); |
1713 | |
1714 | stable_node = page_stable_node(page); |
1715 | if (!stable_node) |
1716 | return ret; |
1717 | again: |
1718 | hlist_for_each_entry(rmap_item, hlist, &stable_node->hlist, hlist) { |
1719 | struct anon_vma *anon_vma = rmap_item->anon_vma; |
1720 | struct anon_vma_chain *vmac; |
1721 | struct vm_area_struct *vma; |
1722 | |
1723 | anon_vma_lock(anon_vma); |
1724 | list_for_each_entry(vmac, &anon_vma->head, same_anon_vma) { |
1725 | vma = vmac->vma; |
1726 | if (rmap_item->address < vma->vm_start || |
1727 | rmap_item->address >= vma->vm_end) |
1728 | continue; |
1729 | /* |
1730 | * Initially we examine only the vma which covers this |
1731 | * rmap_item; but later, if there is still work to do, |
1732 | * we examine covering vmas in other mms: in case they |
1733 | * were forked from the original since ksmd passed. |
1734 | */ |
1735 | if ((rmap_item->mm == vma->vm_mm) == search_new_forks) |
1736 | continue; |
1737 | |
1738 | ret = rmap_one(page, vma, rmap_item->address, arg); |
1739 | if (ret != SWAP_AGAIN) { |
1740 | anon_vma_unlock(anon_vma); |
1741 | goto out; |
1742 | } |
1743 | } |
1744 | anon_vma_unlock(anon_vma); |
1745 | } |
1746 | if (!search_new_forks++) |
1747 | goto again; |
1748 | out: |
1749 | return ret; |
1750 | } |
1751 | |
1752 | void ksm_migrate_page(struct page *newpage, struct page *oldpage) |
1753 | { |
1754 | struct stable_node *stable_node; |
1755 | |
1756 | VM_BUG_ON(!PageLocked(oldpage)); |
1757 | VM_BUG_ON(!PageLocked(newpage)); |
1758 | VM_BUG_ON(newpage->mapping != oldpage->mapping); |
1759 | |
1760 | stable_node = page_stable_node(newpage); |
1761 | if (stable_node) { |
1762 | VM_BUG_ON(stable_node->kpfn != page_to_pfn(oldpage)); |
1763 | stable_node->kpfn = page_to_pfn(newpage); |
1764 | } |
1765 | } |
1766 | #endif /* CONFIG_MIGRATION */ |
1767 | |
1768 | #ifdef CONFIG_MEMORY_HOTREMOVE |
1769 | static struct stable_node *ksm_check_stable_tree(unsigned long start_pfn, |
1770 | unsigned long end_pfn) |
1771 | { |
1772 | struct rb_node *node; |
1773 | |
1774 | for (node = rb_first(&root_stable_tree); node; node = rb_next(node)) { |
1775 | struct stable_node *stable_node; |
1776 | |
1777 | stable_node = rb_entry(node, struct stable_node, node); |
1778 | if (stable_node->kpfn >= start_pfn && |
1779 | stable_node->kpfn < end_pfn) |
1780 | return stable_node; |
1781 | } |
1782 | return NULL; |
1783 | } |
1784 | |
1785 | static int ksm_memory_callback(struct notifier_block *self, |
1786 | unsigned long action, void *arg) |
1787 | { |
1788 | struct memory_notify *mn = arg; |
1789 | struct stable_node *stable_node; |
1790 | |
1791 | switch (action) { |
1792 | case MEM_GOING_OFFLINE: |
1793 | /* |
1794 | * Keep it very simple for now: just lock out ksmd and |
1795 | * MADV_UNMERGEABLE while any memory is going offline. |
1796 | * mutex_lock_nested() is necessary because lockdep was alarmed |
1797 | * that here we take ksm_thread_mutex inside notifier chain |
1798 | * mutex, and later take notifier chain mutex inside |
1799 | * ksm_thread_mutex to unlock it. But that's safe because both |
1800 | * are inside mem_hotplug_mutex. |
1801 | */ |
1802 | mutex_lock_nested(&ksm_thread_mutex, SINGLE_DEPTH_NESTING); |
1803 | break; |
1804 | |
1805 | case MEM_OFFLINE: |
1806 | /* |
1807 | * Most of the work is done by page migration; but there might |
1808 | * be a few stable_nodes left over, still pointing to struct |
1809 | * pages which have been offlined: prune those from the tree. |
1810 | */ |
1811 | while ((stable_node = ksm_check_stable_tree(mn->start_pfn, |
1812 | mn->start_pfn + mn->nr_pages)) != NULL) |
1813 | remove_node_from_stable_tree(stable_node); |
1814 | /* fallthrough */ |
1815 | |
1816 | case MEM_CANCEL_OFFLINE: |
1817 | mutex_unlock(&ksm_thread_mutex); |
1818 | break; |
1819 | } |
1820 | return NOTIFY_OK; |
1821 | } |
1822 | #endif /* CONFIG_MEMORY_HOTREMOVE */ |
1823 | |
1824 | #ifdef CONFIG_SYSFS |
1825 | /* |
1826 | * This all compiles without CONFIG_SYSFS, but is a waste of space. |
1827 | */ |
1828 | |
1829 | #define KSM_ATTR_RO(_name) \ |
1830 | static struct kobj_attribute _name##_attr = __ATTR_RO(_name) |
1831 | #define KSM_ATTR(_name) \ |
1832 | static struct kobj_attribute _name##_attr = \ |
1833 | __ATTR(_name, 0644, _name##_show, _name##_store) |
1834 | |
1835 | static ssize_t sleep_millisecs_show(struct kobject *kobj, |
1836 | struct kobj_attribute *attr, char *buf) |
1837 | { |
1838 | return sprintf(buf, "%u\n", ksm_thread_sleep_millisecs); |
1839 | } |
1840 | |
1841 | static ssize_t sleep_millisecs_store(struct kobject *kobj, |
1842 | struct kobj_attribute *attr, |
1843 | const char *buf, size_t count) |
1844 | { |
1845 | unsigned long msecs; |
1846 | int err; |
1847 | |
1848 | err = strict_strtoul(buf, 10, &msecs); |
1849 | if (err || msecs > UINT_MAX) |
1850 | return -EINVAL; |
1851 | |
1852 | ksm_thread_sleep_millisecs = msecs; |
1853 | |
1854 | return count; |
1855 | } |
1856 | KSM_ATTR(sleep_millisecs); |
1857 | |
1858 | static ssize_t pages_to_scan_show(struct kobject *kobj, |
1859 | struct kobj_attribute *attr, char *buf) |
1860 | { |
1861 | return sprintf(buf, "%u\n", ksm_thread_pages_to_scan); |
1862 | } |
1863 | |
1864 | static ssize_t pages_to_scan_store(struct kobject *kobj, |
1865 | struct kobj_attribute *attr, |
1866 | const char *buf, size_t count) |
1867 | { |
1868 | int err; |
1869 | unsigned long nr_pages; |
1870 | |
1871 | err = strict_strtoul(buf, 10, &nr_pages); |
1872 | if (err || nr_pages > UINT_MAX) |
1873 | return -EINVAL; |
1874 | |
1875 | ksm_thread_pages_to_scan = nr_pages; |
1876 | |
1877 | return count; |
1878 | } |
1879 | KSM_ATTR(pages_to_scan); |
1880 | |
1881 | static ssize_t run_show(struct kobject *kobj, struct kobj_attribute *attr, |
1882 | char *buf) |
1883 | { |
1884 | return sprintf(buf, "%u\n", ksm_run); |
1885 | } |
1886 | |
1887 | static ssize_t run_store(struct kobject *kobj, struct kobj_attribute *attr, |
1888 | const char *buf, size_t count) |
1889 | { |
1890 | int err; |
1891 | unsigned long flags; |
1892 | |
1893 | err = strict_strtoul(buf, 10, &flags); |
1894 | if (err || flags > UINT_MAX) |
1895 | return -EINVAL; |
1896 | if (flags > KSM_RUN_UNMERGE) |
1897 | return -EINVAL; |
1898 | |
1899 | /* |
1900 | * KSM_RUN_MERGE sets ksmd running, and 0 stops it running. |
1901 | * KSM_RUN_UNMERGE stops it running and unmerges all rmap_items, |
1902 | * breaking COW to free the pages_shared (but leaves mm_slots |
1903 | * on the list for when ksmd may be set running again). |
1904 | */ |
1905 | |
1906 | mutex_lock(&ksm_thread_mutex); |
1907 | if (ksm_run != flags) { |
1908 | ksm_run = flags; |
1909 | if (flags & KSM_RUN_UNMERGE) { |
1910 | current->flags |= PF_OOM_ORIGIN; |
1911 | err = unmerge_and_remove_all_rmap_items(); |
1912 | current->flags &= ~PF_OOM_ORIGIN; |
1913 | if (err) { |
1914 | ksm_run = KSM_RUN_STOP; |
1915 | count = err; |
1916 | } |
1917 | } |
1918 | } |
1919 | mutex_unlock(&ksm_thread_mutex); |
1920 | |
1921 | if (flags & KSM_RUN_MERGE) |
1922 | wake_up_interruptible(&ksm_thread_wait); |
1923 | |
1924 | return count; |
1925 | } |
1926 | KSM_ATTR(run); |
1927 | |
1928 | static ssize_t pages_shared_show(struct kobject *kobj, |
1929 | struct kobj_attribute *attr, char *buf) |
1930 | { |
1931 | return sprintf(buf, "%lu\n", ksm_pages_shared); |
1932 | } |
1933 | KSM_ATTR_RO(pages_shared); |
1934 | |
1935 | static ssize_t pages_sharing_show(struct kobject *kobj, |
1936 | struct kobj_attribute *attr, char *buf) |
1937 | { |
1938 | return sprintf(buf, "%lu\n", ksm_pages_sharing); |
1939 | } |
1940 | KSM_ATTR_RO(pages_sharing); |
1941 | |
1942 | static ssize_t pages_unshared_show(struct kobject *kobj, |
1943 | struct kobj_attribute *attr, char *buf) |
1944 | { |
1945 | return sprintf(buf, "%lu\n", ksm_pages_unshared); |
1946 | } |
1947 | KSM_ATTR_RO(pages_unshared); |
1948 | |
1949 | static ssize_t pages_volatile_show(struct kobject *kobj, |
1950 | struct kobj_attribute *attr, char *buf) |
1951 | { |
1952 | long ksm_pages_volatile; |
1953 | |
1954 | ksm_pages_volatile = ksm_rmap_items - ksm_pages_shared |
1955 | - ksm_pages_sharing - ksm_pages_unshared; |
1956 | /* |
1957 | * It was not worth any locking to calculate that statistic, |
1958 | * but it might therefore sometimes be negative: conceal that. |
1959 | */ |
1960 | if (ksm_pages_volatile < 0) |
1961 | ksm_pages_volatile = 0; |
1962 | return sprintf(buf, "%ld\n", ksm_pages_volatile); |
1963 | } |
1964 | KSM_ATTR_RO(pages_volatile); |
1965 | |
1966 | static ssize_t full_scans_show(struct kobject *kobj, |
1967 | struct kobj_attribute *attr, char *buf) |
1968 | { |
1969 | return sprintf(buf, "%lu\n", ksm_scan.seqnr); |
1970 | } |
1971 | KSM_ATTR_RO(full_scans); |
1972 | |
1973 | static struct attribute *ksm_attrs[] = { |
1974 | &sleep_millisecs_attr.attr, |
1975 | &pages_to_scan_attr.attr, |
1976 | &run_attr.attr, |
1977 | &pages_shared_attr.attr, |
1978 | &pages_sharing_attr.attr, |
1979 | &pages_unshared_attr.attr, |
1980 | &pages_volatile_attr.attr, |
1981 | &full_scans_attr.attr, |
1982 | NULL, |
1983 | }; |
1984 | |
1985 | static struct attribute_group ksm_attr_group = { |
1986 | .attrs = ksm_attrs, |
1987 | .name = "ksm", |
1988 | }; |
1989 | #endif /* CONFIG_SYSFS */ |
1990 | |
1991 | static int __init ksm_init(void) |
1992 | { |
1993 | struct task_struct *ksm_thread; |
1994 | int err; |
1995 | |
1996 | err = ksm_slab_init(); |
1997 | if (err) |
1998 | goto out; |
1999 | |
2000 | ksm_thread = kthread_run(ksm_scan_thread, NULL, "ksmd"); |
2001 | if (IS_ERR(ksm_thread)) { |
2002 | printk(KERN_ERR "ksm: creating kthread failed\n"); |
2003 | err = PTR_ERR(ksm_thread); |
2004 | goto out_free; |
2005 | } |
2006 | |
2007 | #ifdef CONFIG_SYSFS |
2008 | err = sysfs_create_group(mm_kobj, &ksm_attr_group); |
2009 | if (err) { |
2010 | printk(KERN_ERR "ksm: register sysfs failed\n"); |
2011 | kthread_stop(ksm_thread); |
2012 | goto out_free; |
2013 | } |
2014 | #else |
2015 | ksm_run = KSM_RUN_MERGE; /* no way for user to start it */ |
2016 | |
2017 | #endif /* CONFIG_SYSFS */ |
2018 | |
2019 | #ifdef CONFIG_MEMORY_HOTREMOVE |
2020 | /* |
2021 | * Choose a high priority since the callback takes ksm_thread_mutex: |
2022 | * later callbacks could only be taking locks which nest within that. |
2023 | */ |
2024 | hotplug_memory_notifier(ksm_memory_callback, 100); |
2025 | #endif |
2026 | return 0; |
2027 | |
2028 | out_free: |
2029 | ksm_slab_free(); |
2030 | out: |
2031 | return err; |
2032 | } |
2033 | module_init(ksm_init) |
2034 |
Branches:
ben-wpan
ben-wpan-stefan
javiroman/ks7010
jz-2.6.34
jz-2.6.34-rc5
jz-2.6.34-rc6
jz-2.6.34-rc7
jz-2.6.35
jz-2.6.36
jz-2.6.37
jz-2.6.38
jz-2.6.39
jz-3.0
jz-3.1
jz-3.11
jz-3.12
jz-3.13
jz-3.15
jz-3.16
jz-3.18-dt
jz-3.2
jz-3.3
jz-3.4
jz-3.5
jz-3.6
jz-3.6-rc2-pwm
jz-3.9
jz-3.9-clk
jz-3.9-rc8
jz47xx
jz47xx-2.6.38
master
Tags:
od-2011-09-04
od-2011-09-18
v2.6.34-rc5
v2.6.34-rc6
v2.6.34-rc7
v3.9