xref: /linux/mm/internal.h (revision 26b0d14106954ae46d2f4f7eec3481828a210f7d)
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  */
11 #ifndef __MM_INTERNAL_H
12 #define __MM_INTERNAL_H
13 
14 #include <linux/mm.h>
15 
16 void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
17 		unsigned long floor, unsigned long ceiling);
18 
19 static inline void set_page_count(struct page *page, int v)
20 {
21 	atomic_set(&page->_count, v);
22 }
23 
24 /*
25  * Turn a non-refcounted page (->_count == 0) into refcounted with
26  * a count of one.
27  */
28 static inline void set_page_refcounted(struct page *page)
29 {
30 	VM_BUG_ON(PageTail(page));
31 	VM_BUG_ON(atomic_read(&page->_count));
32 	set_page_count(page, 1);
33 }
34 
35 static inline void __put_page(struct page *page)
36 {
37 	atomic_dec(&page->_count);
38 }
39 
40 static inline void __get_page_tail_foll(struct page *page,
41 					bool get_page_head)
42 {
43 	/*
44 	 * If we're getting a tail page, the elevated page->_count is
45 	 * required only in the head page and we will elevate the head
46 	 * page->_count and tail page->_mapcount.
47 	 *
48 	 * We elevate page_tail->_mapcount for tail pages to force
49 	 * page_tail->_count to be zero at all times to avoid getting
50 	 * false positives from get_page_unless_zero() with
51 	 * speculative page access (like in
52 	 * page_cache_get_speculative()) on tail pages.
53 	 */
54 	VM_BUG_ON(atomic_read(&page->first_page->_count) <= 0);
55 	VM_BUG_ON(atomic_read(&page->_count) != 0);
56 	VM_BUG_ON(page_mapcount(page) < 0);
57 	if (get_page_head)
58 		atomic_inc(&page->first_page->_count);
59 	atomic_inc(&page->_mapcount);
60 }
61 
62 /*
63  * This is meant to be called as the FOLL_GET operation of
64  * follow_page() and it must be called while holding the proper PT
65  * lock while the pte (or pmd_trans_huge) is still mapping the page.
66  */
67 static inline void get_page_foll(struct page *page)
68 {
69 	if (unlikely(PageTail(page)))
70 		/*
71 		 * This is safe only because
72 		 * __split_huge_page_refcount() can't run under
73 		 * get_page_foll() because we hold the proper PT lock.
74 		 */
75 		__get_page_tail_foll(page, true);
76 	else {
77 		/*
78 		 * Getting a normal page or the head of a compound page
79 		 * requires to already have an elevated page->_count.
80 		 */
81 		VM_BUG_ON(atomic_read(&page->_count) <= 0);
82 		atomic_inc(&page->_count);
83 	}
84 }
85 
86 extern unsigned long highest_memmap_pfn;
87 
88 /*
89  * in mm/vmscan.c:
90  */
91 extern int isolate_lru_page(struct page *page);
92 extern void putback_lru_page(struct page *page);
93 
94 /*
95  * in mm/page_alloc.c
96  */
97 extern void __free_pages_bootmem(struct page *page, unsigned int order);
98 extern void prep_compound_page(struct page *page, unsigned long order);
99 #ifdef CONFIG_MEMORY_FAILURE
100 extern bool is_free_buddy_page(struct page *page);
101 #endif
102 
103 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
104 
105 /*
106  * in mm/compaction.c
107  */
108 /*
109  * compact_control is used to track pages being migrated and the free pages
110  * they are being migrated to during memory compaction. The free_pfn starts
111  * at the end of a zone and migrate_pfn begins at the start. Movable pages
112  * are moved to the end of a zone during a compaction run and the run
113  * completes when free_pfn <= migrate_pfn
114  */
115 struct compact_control {
116 	struct list_head freepages;	/* List of free pages to migrate to */
117 	struct list_head migratepages;	/* List of pages being migrated */
118 	unsigned long nr_freepages;	/* Number of isolated free pages */
119 	unsigned long nr_migratepages;	/* Number of pages to migrate */
120 	unsigned long free_pfn;		/* isolate_freepages search base */
121 	unsigned long migrate_pfn;	/* isolate_migratepages search base */
122 	bool sync;			/* Synchronous migration */
123 
124 	int order;			/* order a direct compactor needs */
125 	int migratetype;		/* MOVABLE, RECLAIMABLE etc */
126 	struct zone *zone;
127 };
128 
129 unsigned long
130 isolate_freepages_range(unsigned long start_pfn, unsigned long end_pfn);
131 unsigned long
132 isolate_migratepages_range(struct zone *zone, struct compact_control *cc,
133 			   unsigned long low_pfn, unsigned long end_pfn);
134 
135 #endif
136 
137 /*
138  * function for dealing with page's order in buddy system.
139  * zone->lock is already acquired when we use these.
140  * So, we don't need atomic page->flags operations here.
141  */
142 static inline unsigned long page_order(struct page *page)
143 {
144 	/* PageBuddy() must be checked by the caller */
145 	return page_private(page);
146 }
147 
148 /* mm/util.c */
149 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
150 		struct vm_area_struct *prev, struct rb_node *rb_parent);
151 
152 #ifdef CONFIG_MMU
153 extern long mlock_vma_pages_range(struct vm_area_struct *vma,
154 			unsigned long start, unsigned long end);
155 extern void munlock_vma_pages_range(struct vm_area_struct *vma,
156 			unsigned long start, unsigned long end);
157 static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
158 {
159 	munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
160 }
161 
162 /*
163  * Called only in fault path via page_evictable() for a new page
164  * to determine if it's being mapped into a LOCKED vma.
165  * If so, mark page as mlocked.
166  */
167 static inline int mlocked_vma_newpage(struct vm_area_struct *vma,
168 				    struct page *page)
169 {
170 	VM_BUG_ON(PageLRU(page));
171 
172 	if (likely((vma->vm_flags & (VM_LOCKED | VM_SPECIAL)) != VM_LOCKED))
173 		return 0;
174 
175 	if (!TestSetPageMlocked(page)) {
176 		inc_zone_page_state(page, NR_MLOCK);
177 		count_vm_event(UNEVICTABLE_PGMLOCKED);
178 	}
179 	return 1;
180 }
181 
182 /*
183  * must be called with vma's mmap_sem held for read or write, and page locked.
184  */
185 extern void mlock_vma_page(struct page *page);
186 extern void munlock_vma_page(struct page *page);
187 
188 /*
189  * Clear the page's PageMlocked().  This can be useful in a situation where
190  * we want to unconditionally remove a page from the pagecache -- e.g.,
191  * on truncation or freeing.
192  *
193  * It is legal to call this function for any page, mlocked or not.
194  * If called for a page that is still mapped by mlocked vmas, all we do
195  * is revert to lazy LRU behaviour -- semantics are not broken.
196  */
197 extern void __clear_page_mlock(struct page *page);
198 static inline void clear_page_mlock(struct page *page)
199 {
200 	if (unlikely(TestClearPageMlocked(page)))
201 		__clear_page_mlock(page);
202 }
203 
204 /*
205  * mlock_migrate_page - called only from migrate_page_copy() to
206  * migrate the Mlocked page flag; update statistics.
207  */
208 static inline void mlock_migrate_page(struct page *newpage, struct page *page)
209 {
210 	if (TestClearPageMlocked(page)) {
211 		unsigned long flags;
212 
213 		local_irq_save(flags);
214 		__dec_zone_page_state(page, NR_MLOCK);
215 		SetPageMlocked(newpage);
216 		__inc_zone_page_state(newpage, NR_MLOCK);
217 		local_irq_restore(flags);
218 	}
219 }
220 
221 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
222 extern unsigned long vma_address(struct page *page,
223 				 struct vm_area_struct *vma);
224 #endif
225 #else /* !CONFIG_MMU */
226 static inline int mlocked_vma_newpage(struct vm_area_struct *v, struct page *p)
227 {
228 	return 0;
229 }
230 static inline void clear_page_mlock(struct page *page) { }
231 static inline void mlock_vma_page(struct page *page) { }
232 static inline void mlock_migrate_page(struct page *new, struct page *old) { }
233 
234 #endif /* !CONFIG_MMU */
235 
236 /*
237  * Return the mem_map entry representing the 'offset' subpage within
238  * the maximally aligned gigantic page 'base'.  Handle any discontiguity
239  * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
240  */
241 static inline struct page *mem_map_offset(struct page *base, int offset)
242 {
243 	if (unlikely(offset >= MAX_ORDER_NR_PAGES))
244 		return pfn_to_page(page_to_pfn(base) + offset);
245 	return base + offset;
246 }
247 
248 /*
249  * Iterator over all subpages within the maximally aligned gigantic
250  * page 'base'.  Handle any discontiguity in the mem_map.
251  */
252 static inline struct page *mem_map_next(struct page *iter,
253 						struct page *base, int offset)
254 {
255 	if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
256 		unsigned long pfn = page_to_pfn(base) + offset;
257 		if (!pfn_valid(pfn))
258 			return NULL;
259 		return pfn_to_page(pfn);
260 	}
261 	return iter + 1;
262 }
263 
264 /*
265  * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
266  * so all functions starting at paging_init should be marked __init
267  * in those cases. SPARSEMEM, however, allows for memory hotplug,
268  * and alloc_bootmem_node is not used.
269  */
270 #ifdef CONFIG_SPARSEMEM
271 #define __paginginit __meminit
272 #else
273 #define __paginginit __init
274 #endif
275 
276 /* Memory initialisation debug and verification */
277 enum mminit_level {
278 	MMINIT_WARNING,
279 	MMINIT_VERIFY,
280 	MMINIT_TRACE
281 };
282 
283 #ifdef CONFIG_DEBUG_MEMORY_INIT
284 
285 extern int mminit_loglevel;
286 
287 #define mminit_dprintk(level, prefix, fmt, arg...) \
288 do { \
289 	if (level < mminit_loglevel) { \
290 		printk(level <= MMINIT_WARNING ? KERN_WARNING : KERN_DEBUG); \
291 		printk(KERN_CONT "mminit::" prefix " " fmt, ##arg); \
292 	} \
293 } while (0)
294 
295 extern void mminit_verify_pageflags_layout(void);
296 extern void mminit_verify_page_links(struct page *page,
297 		enum zone_type zone, unsigned long nid, unsigned long pfn);
298 extern void mminit_verify_zonelist(void);
299 
300 #else
301 
302 static inline void mminit_dprintk(enum mminit_level level,
303 				const char *prefix, const char *fmt, ...)
304 {
305 }
306 
307 static inline void mminit_verify_pageflags_layout(void)
308 {
309 }
310 
311 static inline void mminit_verify_page_links(struct page *page,
312 		enum zone_type zone, unsigned long nid, unsigned long pfn)
313 {
314 }
315 
316 static inline void mminit_verify_zonelist(void)
317 {
318 }
319 #endif /* CONFIG_DEBUG_MEMORY_INIT */
320 
321 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
322 #if defined(CONFIG_SPARSEMEM)
323 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
324 				unsigned long *end_pfn);
325 #else
326 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
327 				unsigned long *end_pfn)
328 {
329 }
330 #endif /* CONFIG_SPARSEMEM */
331 
332 #define ZONE_RECLAIM_NOSCAN	-2
333 #define ZONE_RECLAIM_FULL	-1
334 #define ZONE_RECLAIM_SOME	0
335 #define ZONE_RECLAIM_SUCCESS	1
336 #endif
337 
338 extern int hwpoison_filter(struct page *p);
339 
340 extern u32 hwpoison_filter_dev_major;
341 extern u32 hwpoison_filter_dev_minor;
342 extern u64 hwpoison_filter_flags_mask;
343 extern u64 hwpoison_filter_flags_value;
344 extern u64 hwpoison_filter_memcg;
345 extern u32 hwpoison_filter_enable;
346 
347 extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
348         unsigned long, unsigned long,
349         unsigned long, unsigned long);
350