Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* internal.h: mm/ internal definitions |
2 | * | |
3 | * Copyright (C) 2004 Red Hat, Inc. All Rights Reserved. | |
4 | * Written by David Howells (dhowells@redhat.com) | |
5 | * | |
6 | * This program is free software; you can redistribute it and/or | |
7 | * modify it under the terms of the GNU General Public License | |
8 | * as published by the Free Software Foundation; either version | |
9 | * 2 of the License, or (at your option) any later version. | |
10 | */ | |
0f8053a5 NP |
11 | #ifndef __MM_INTERNAL_H |
12 | #define __MM_INTERNAL_H | |
13 | ||
29f175d1 | 14 | #include <linux/fs.h> |
0f8053a5 | 15 | #include <linux/mm.h> |
1da177e4 | 16 | |
42b77728 JB |
17 | void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma, |
18 | unsigned long floor, unsigned long ceiling); | |
19 | ||
7835e98b | 20 | static inline void set_page_count(struct page *page, int v) |
77a8a788 | 21 | { |
7835e98b NP |
22 | atomic_set(&page->_count, v); |
23 | } | |
24 | ||
29f175d1 FF |
25 | extern int __do_page_cache_readahead(struct address_space *mapping, |
26 | struct file *filp, pgoff_t offset, unsigned long nr_to_read, | |
27 | unsigned long lookahead_size); | |
28 | ||
29 | /* | |
30 | * Submit IO for the read-ahead request in file_ra_state. | |
31 | */ | |
32 | static inline unsigned long ra_submit(struct file_ra_state *ra, | |
33 | struct address_space *mapping, struct file *filp) | |
34 | { | |
35 | return __do_page_cache_readahead(mapping, filp, | |
36 | ra->start, ra->size, ra->async_size); | |
37 | } | |
38 | ||
7835e98b NP |
39 | /* |
40 | * Turn a non-refcounted page (->_count == 0) into refcounted with | |
41 | * a count of one. | |
42 | */ | |
43 | static inline void set_page_refcounted(struct page *page) | |
44 | { | |
309381fe SL |
45 | VM_BUG_ON_PAGE(PageTail(page), page); |
46 | VM_BUG_ON_PAGE(atomic_read(&page->_count), page); | |
77a8a788 | 47 | set_page_count(page, 1); |
77a8a788 NP |
48 | } |
49 | ||
70b50f94 AA |
50 | static inline void __get_page_tail_foll(struct page *page, |
51 | bool get_page_head) | |
52 | { | |
53 | /* | |
54 | * If we're getting a tail page, the elevated page->_count is | |
55 | * required only in the head page and we will elevate the head | |
56 | * page->_count and tail page->_mapcount. | |
57 | * | |
58 | * We elevate page_tail->_mapcount for tail pages to force | |
59 | * page_tail->_count to be zero at all times to avoid getting | |
60 | * false positives from get_page_unless_zero() with | |
61 | * speculative page access (like in | |
62 | * page_cache_get_speculative()) on tail pages. | |
63 | */ | |
309381fe | 64 | VM_BUG_ON_PAGE(atomic_read(&page->first_page->_count) <= 0, page); |
70b50f94 AA |
65 | if (get_page_head) |
66 | atomic_inc(&page->first_page->_count); | |
c728852f | 67 | get_huge_page_tail(page); |
70b50f94 AA |
68 | } |
69 | ||
70 | /* | |
71 | * This is meant to be called as the FOLL_GET operation of | |
72 | * follow_page() and it must be called while holding the proper PT | |
73 | * lock while the pte (or pmd_trans_huge) is still mapping the page. | |
74 | */ | |
75 | static inline void get_page_foll(struct page *page) | |
76 | { | |
77 | if (unlikely(PageTail(page))) | |
78 | /* | |
79 | * This is safe only because | |
80 | * __split_huge_page_refcount() can't run under | |
81 | * get_page_foll() because we hold the proper PT lock. | |
82 | */ | |
83 | __get_page_tail_foll(page, true); | |
84 | else { | |
85 | /* | |
86 | * Getting a normal page or the head of a compound page | |
87 | * requires to already have an elevated page->_count. | |
88 | */ | |
309381fe | 89 | VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page); |
70b50f94 AA |
90 | atomic_inc(&page->_count); |
91 | } | |
92 | } | |
93 | ||
03f6462a HD |
94 | extern unsigned long highest_memmap_pfn; |
95 | ||
894bc310 LS |
96 | /* |
97 | * in mm/vmscan.c: | |
98 | */ | |
62695a84 | 99 | extern int isolate_lru_page(struct page *page); |
894bc310 | 100 | extern void putback_lru_page(struct page *page); |
6e543d57 | 101 | extern bool zone_reclaimable(struct zone *zone); |
62695a84 | 102 | |
6219049a BL |
103 | /* |
104 | * in mm/rmap.c: | |
105 | */ | |
106 | extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address); | |
107 | ||
894bc310 LS |
108 | /* |
109 | * in mm/page_alloc.c | |
110 | */ | |
0c0a4a51 | 111 | extern void __free_pages_bootmem(struct page *page, unsigned int order); |
20a0307c | 112 | extern void prep_compound_page(struct page *page, unsigned long order); |
8d22ba1b WF |
113 | #ifdef CONFIG_MEMORY_FAILURE |
114 | extern bool is_free_buddy_page(struct page *page); | |
115 | #endif | |
42aa83cb | 116 | extern int user_min_free_kbytes; |
20a0307c | 117 | |
ff9543fd MN |
118 | #if defined CONFIG_COMPACTION || defined CONFIG_CMA |
119 | ||
120 | /* | |
121 | * in mm/compaction.c | |
122 | */ | |
123 | /* | |
124 | * compact_control is used to track pages being migrated and the free pages | |
125 | * they are being migrated to during memory compaction. The free_pfn starts | |
126 | * at the end of a zone and migrate_pfn begins at the start. Movable pages | |
127 | * are moved to the end of a zone during a compaction run and the run | |
128 | * completes when free_pfn <= migrate_pfn | |
129 | */ | |
130 | struct compact_control { | |
131 | struct list_head freepages; /* List of free pages to migrate to */ | |
132 | struct list_head migratepages; /* List of pages being migrated */ | |
133 | unsigned long nr_freepages; /* Number of isolated free pages */ | |
134 | unsigned long nr_migratepages; /* Number of pages to migrate */ | |
135 | unsigned long free_pfn; /* isolate_freepages search base */ | |
136 | unsigned long migrate_pfn; /* isolate_migratepages search base */ | |
e0b9daeb | 137 | enum migrate_mode mode; /* Async or sync migration mode */ |
bb13ffeb | 138 | bool ignore_skip_hint; /* Scan blocks even if marked skip */ |
c89511ab MG |
139 | bool finished_update_free; /* True when the zone cached pfns are |
140 | * no longer being updated | |
141 | */ | |
142 | bool finished_update_migrate; | |
ff9543fd MN |
143 | |
144 | int order; /* order a direct compactor needs */ | |
6d7ce559 | 145 | const gfp_t gfp_mask; /* gfp mask of a direct compactor */ |
ff9543fd | 146 | struct zone *zone; |
1f9efdef VB |
147 | int contended; /* Signal need_sched() or lock |
148 | * contention detected during | |
be976572 VB |
149 | * compaction |
150 | */ | |
ff9543fd MN |
151 | }; |
152 | ||
153 | unsigned long | |
bb13ffeb MG |
154 | isolate_freepages_range(struct compact_control *cc, |
155 | unsigned long start_pfn, unsigned long end_pfn); | |
ff9543fd | 156 | unsigned long |
edc2ca61 VB |
157 | isolate_migratepages_range(struct compact_control *cc, |
158 | unsigned long low_pfn, unsigned long end_pfn); | |
ff9543fd MN |
159 | |
160 | #endif | |
0f8053a5 | 161 | |
48f13bf3 | 162 | /* |
6c14466c MG |
163 | * This function returns the order of a free page in the buddy system. In |
164 | * general, page_zone(page)->lock must be held by the caller to prevent the | |
165 | * page from being allocated in parallel and returning garbage as the order. | |
166 | * If a caller does not hold page_zone(page)->lock, it must guarantee that the | |
99c0fd5e VB |
167 | * page cannot be allocated or merged in parallel. Alternatively, it must |
168 | * handle invalid values gracefully, and use page_order_unsafe() below. | |
48f13bf3 MG |
169 | */ |
170 | static inline unsigned long page_order(struct page *page) | |
171 | { | |
572438f9 | 172 | /* PageBuddy() must be checked by the caller */ |
48f13bf3 MG |
173 | return page_private(page); |
174 | } | |
b5a0e011 | 175 | |
99c0fd5e VB |
176 | /* |
177 | * Like page_order(), but for callers who cannot afford to hold the zone lock. | |
178 | * PageBuddy() should be checked first by the caller to minimize race window, | |
179 | * and invalid values must be handled gracefully. | |
180 | * | |
181 | * ACCESS_ONCE is used so that if the caller assigns the result into a local | |
182 | * variable and e.g. tests it for valid range before using, the compiler cannot | |
183 | * decide to remove the variable and inline the page_private(page) multiple | |
184 | * times, potentially observing different values in the tests and the actual | |
185 | * use of the result. | |
186 | */ | |
187 | #define page_order_unsafe(page) ACCESS_ONCE(page_private(page)) | |
188 | ||
4bbd4c77 KS |
189 | static inline bool is_cow_mapping(vm_flags_t flags) |
190 | { | |
191 | return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE; | |
192 | } | |
193 | ||
6038def0 NK |
194 | /* mm/util.c */ |
195 | void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma, | |
196 | struct vm_area_struct *prev, struct rb_node *rb_parent); | |
197 | ||
af8e3354 | 198 | #ifdef CONFIG_MMU |
cea10a19 ML |
199 | extern long __mlock_vma_pages_range(struct vm_area_struct *vma, |
200 | unsigned long start, unsigned long end, int *nonblocking); | |
af8e3354 HD |
201 | extern void munlock_vma_pages_range(struct vm_area_struct *vma, |
202 | unsigned long start, unsigned long end); | |
203 | static inline void munlock_vma_pages_all(struct vm_area_struct *vma) | |
204 | { | |
205 | munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end); | |
206 | } | |
207 | ||
b291f000 | 208 | /* |
73848b46 | 209 | * must be called with vma's mmap_sem held for read or write, and page locked. |
b291f000 NP |
210 | */ |
211 | extern void mlock_vma_page(struct page *page); | |
ff6a6da6 | 212 | extern unsigned int munlock_vma_page(struct page *page); |
b291f000 NP |
213 | |
214 | /* | |
215 | * Clear the page's PageMlocked(). This can be useful in a situation where | |
216 | * we want to unconditionally remove a page from the pagecache -- e.g., | |
217 | * on truncation or freeing. | |
218 | * | |
219 | * It is legal to call this function for any page, mlocked or not. | |
220 | * If called for a page that is still mapped by mlocked vmas, all we do | |
221 | * is revert to lazy LRU behaviour -- semantics are not broken. | |
222 | */ | |
e6c509f8 | 223 | extern void clear_page_mlock(struct page *page); |
b291f000 NP |
224 | |
225 | /* | |
226 | * mlock_migrate_page - called only from migrate_page_copy() to | |
5344b7e6 | 227 | * migrate the Mlocked page flag; update statistics. |
b291f000 NP |
228 | */ |
229 | static inline void mlock_migrate_page(struct page *newpage, struct page *page) | |
230 | { | |
5344b7e6 NP |
231 | if (TestClearPageMlocked(page)) { |
232 | unsigned long flags; | |
b32967ff | 233 | int nr_pages = hpage_nr_pages(page); |
5344b7e6 NP |
234 | |
235 | local_irq_save(flags); | |
b32967ff | 236 | __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages); |
b291f000 | 237 | SetPageMlocked(newpage); |
b32967ff | 238 | __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages); |
5344b7e6 NP |
239 | local_irq_restore(flags); |
240 | } | |
b291f000 NP |
241 | } |
242 | ||
b32967ff MG |
243 | extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma); |
244 | ||
71e3aac0 AA |
245 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
246 | extern unsigned long vma_address(struct page *page, | |
247 | struct vm_area_struct *vma); | |
248 | #endif | |
af8e3354 | 249 | #else /* !CONFIG_MMU */ |
b291f000 NP |
250 | static inline void clear_page_mlock(struct page *page) { } |
251 | static inline void mlock_vma_page(struct page *page) { } | |
252 | static inline void mlock_migrate_page(struct page *new, struct page *old) { } | |
253 | ||
af8e3354 | 254 | #endif /* !CONFIG_MMU */ |
894bc310 | 255 | |
69d177c2 AW |
256 | /* |
257 | * Return the mem_map entry representing the 'offset' subpage within | |
258 | * the maximally aligned gigantic page 'base'. Handle any discontiguity | |
259 | * in the mem_map at MAX_ORDER_NR_PAGES boundaries. | |
260 | */ | |
261 | static inline struct page *mem_map_offset(struct page *base, int offset) | |
262 | { | |
263 | if (unlikely(offset >= MAX_ORDER_NR_PAGES)) | |
bc7f84c0 | 264 | return nth_page(base, offset); |
69d177c2 AW |
265 | return base + offset; |
266 | } | |
267 | ||
268 | /* | |
25985edc | 269 | * Iterator over all subpages within the maximally aligned gigantic |
69d177c2 AW |
270 | * page 'base'. Handle any discontiguity in the mem_map. |
271 | */ | |
272 | static inline struct page *mem_map_next(struct page *iter, | |
273 | struct page *base, int offset) | |
274 | { | |
275 | if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) { | |
276 | unsigned long pfn = page_to_pfn(base) + offset; | |
277 | if (!pfn_valid(pfn)) | |
278 | return NULL; | |
279 | return pfn_to_page(pfn); | |
280 | } | |
281 | return iter + 1; | |
282 | } | |
283 | ||
b5a0e011 AH |
284 | /* |
285 | * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node, | |
286 | * so all functions starting at paging_init should be marked __init | |
287 | * in those cases. SPARSEMEM, however, allows for memory hotplug, | |
288 | * and alloc_bootmem_node is not used. | |
289 | */ | |
290 | #ifdef CONFIG_SPARSEMEM | |
291 | #define __paginginit __meminit | |
292 | #else | |
293 | #define __paginginit __init | |
294 | #endif | |
295 | ||
6b74ab97 MG |
296 | /* Memory initialisation debug and verification */ |
297 | enum mminit_level { | |
298 | MMINIT_WARNING, | |
299 | MMINIT_VERIFY, | |
300 | MMINIT_TRACE | |
301 | }; | |
302 | ||
303 | #ifdef CONFIG_DEBUG_MEMORY_INIT | |
304 | ||
305 | extern int mminit_loglevel; | |
306 | ||
307 | #define mminit_dprintk(level, prefix, fmt, arg...) \ | |
308 | do { \ | |
309 | if (level < mminit_loglevel) { \ | |
310 | printk(level <= MMINIT_WARNING ? KERN_WARNING : KERN_DEBUG); \ | |
311 | printk(KERN_CONT "mminit::" prefix " " fmt, ##arg); \ | |
312 | } \ | |
313 | } while (0) | |
314 | ||
708614e6 MG |
315 | extern void mminit_verify_pageflags_layout(void); |
316 | extern void mminit_verify_page_links(struct page *page, | |
317 | enum zone_type zone, unsigned long nid, unsigned long pfn); | |
68ad8df4 | 318 | extern void mminit_verify_zonelist(void); |
708614e6 | 319 | |
6b74ab97 MG |
320 | #else |
321 | ||
322 | static inline void mminit_dprintk(enum mminit_level level, | |
323 | const char *prefix, const char *fmt, ...) | |
324 | { | |
325 | } | |
326 | ||
708614e6 MG |
327 | static inline void mminit_verify_pageflags_layout(void) |
328 | { | |
329 | } | |
330 | ||
331 | static inline void mminit_verify_page_links(struct page *page, | |
332 | enum zone_type zone, unsigned long nid, unsigned long pfn) | |
333 | { | |
334 | } | |
68ad8df4 MG |
335 | |
336 | static inline void mminit_verify_zonelist(void) | |
337 | { | |
338 | } | |
6b74ab97 | 339 | #endif /* CONFIG_DEBUG_MEMORY_INIT */ |
2dbb51c4 MG |
340 | |
341 | /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */ | |
342 | #if defined(CONFIG_SPARSEMEM) | |
343 | extern void mminit_validate_memmodel_limits(unsigned long *start_pfn, | |
344 | unsigned long *end_pfn); | |
345 | #else | |
346 | static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn, | |
347 | unsigned long *end_pfn) | |
348 | { | |
349 | } | |
350 | #endif /* CONFIG_SPARSEMEM */ | |
351 | ||
fa5e084e MG |
352 | #define ZONE_RECLAIM_NOSCAN -2 |
353 | #define ZONE_RECLAIM_FULL -1 | |
354 | #define ZONE_RECLAIM_SOME 0 | |
355 | #define ZONE_RECLAIM_SUCCESS 1 | |
7c116f2b | 356 | |
31d3d348 WF |
357 | extern int hwpoison_filter(struct page *p); |
358 | ||
7c116f2b WF |
359 | extern u32 hwpoison_filter_dev_major; |
360 | extern u32 hwpoison_filter_dev_minor; | |
478c5ffc WF |
361 | extern u64 hwpoison_filter_flags_mask; |
362 | extern u64 hwpoison_filter_flags_value; | |
4fd466eb | 363 | extern u64 hwpoison_filter_memcg; |
1bfe5feb | 364 | extern u32 hwpoison_filter_enable; |
eb36c587 AV |
365 | |
366 | extern unsigned long vm_mmap_pgoff(struct file *, unsigned long, | |
367 | unsigned long, unsigned long, | |
368 | unsigned long, unsigned long); | |
ca57df79 XQ |
369 | |
370 | extern void set_pageblock_order(void); | |
02c6de8d MK |
371 | unsigned long reclaim_clean_pages_from_list(struct zone *zone, |
372 | struct list_head *page_list); | |
d95ea5d1 BZ |
373 | /* The ALLOC_WMARK bits are used as an index to zone->watermark */ |
374 | #define ALLOC_WMARK_MIN WMARK_MIN | |
375 | #define ALLOC_WMARK_LOW WMARK_LOW | |
376 | #define ALLOC_WMARK_HIGH WMARK_HIGH | |
377 | #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */ | |
378 | ||
379 | /* Mask to get the watermark bits */ | |
380 | #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1) | |
381 | ||
382 | #define ALLOC_HARDER 0x10 /* try to alloc harder */ | |
383 | #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */ | |
384 | #define ALLOC_CPUSET 0x40 /* check for correct cpuset */ | |
385 | #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */ | |
3a025760 | 386 | #define ALLOC_FAIR 0x100 /* fair zone allocation */ |
d95ea5d1 | 387 | |
db971418 | 388 | #endif /* __MM_INTERNAL_H */ |