xref: /openbmc/linux/mm/internal.h (revision 95e9fd10)
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 start_free_pfn;	/* where we started the search */
122 	unsigned long migrate_pfn;	/* isolate_migratepages search base */
123 	bool sync;			/* Synchronous migration */
124 	bool wrapped;			/* Order > 0 compactions are
125 					   incremental, once free_pfn
126 					   and migrate_pfn meet, we restart
127 					   from the top of the zone;
128 					   remember we wrapped around. */
129 
130 	int order;			/* order a direct compactor needs */
131 	int migratetype;		/* MOVABLE, RECLAIMABLE etc */
132 	struct zone *zone;
133 	bool *contended;		/* True if a lock was contended */
134 };
135 
136 unsigned long
137 isolate_freepages_range(unsigned long start_pfn, unsigned long end_pfn);
138 unsigned long
139 isolate_migratepages_range(struct zone *zone, struct compact_control *cc,
140 			   unsigned long low_pfn, unsigned long end_pfn);
141 
142 #endif
143 
144 /*
145  * function for dealing with page's order in buddy system.
146  * zone->lock is already acquired when we use these.
147  * So, we don't need atomic page->flags operations here.
148  */
149 static inline unsigned long page_order(struct page *page)
150 {
151 	/* PageBuddy() must be checked by the caller */
152 	return page_private(page);
153 }
154 
155 /* mm/util.c */
156 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
157 		struct vm_area_struct *prev, struct rb_node *rb_parent);
158 
159 #ifdef CONFIG_MMU
160 extern long mlock_vma_pages_range(struct vm_area_struct *vma,
161 			unsigned long start, unsigned long end);
162 extern void munlock_vma_pages_range(struct vm_area_struct *vma,
163 			unsigned long start, unsigned long end);
164 static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
165 {
166 	munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
167 }
168 
169 /*
170  * Called only in fault path via page_evictable() for a new page
171  * to determine if it's being mapped into a LOCKED vma.
172  * If so, mark page as mlocked.
173  */
174 static inline int mlocked_vma_newpage(struct vm_area_struct *vma,
175 				    struct page *page)
176 {
177 	VM_BUG_ON(PageLRU(page));
178 
179 	if (likely((vma->vm_flags & (VM_LOCKED | VM_SPECIAL)) != VM_LOCKED))
180 		return 0;
181 
182 	if (!TestSetPageMlocked(page)) {
183 		inc_zone_page_state(page, NR_MLOCK);
184 		count_vm_event(UNEVICTABLE_PGMLOCKED);
185 	}
186 	return 1;
187 }
188 
189 /*
190  * must be called with vma's mmap_sem held for read or write, and page locked.
191  */
192 extern void mlock_vma_page(struct page *page);
193 extern void munlock_vma_page(struct page *page);
194 
195 /*
196  * Clear the page's PageMlocked().  This can be useful in a situation where
197  * we want to unconditionally remove a page from the pagecache -- e.g.,
198  * on truncation or freeing.
199  *
200  * It is legal to call this function for any page, mlocked or not.
201  * If called for a page that is still mapped by mlocked vmas, all we do
202  * is revert to lazy LRU behaviour -- semantics are not broken.
203  */
204 extern void __clear_page_mlock(struct page *page);
205 static inline void clear_page_mlock(struct page *page)
206 {
207 	if (unlikely(TestClearPageMlocked(page)))
208 		__clear_page_mlock(page);
209 }
210 
211 /*
212  * mlock_migrate_page - called only from migrate_page_copy() to
213  * migrate the Mlocked page flag; update statistics.
214  */
215 static inline void mlock_migrate_page(struct page *newpage, struct page *page)
216 {
217 	if (TestClearPageMlocked(page)) {
218 		unsigned long flags;
219 
220 		local_irq_save(flags);
221 		__dec_zone_page_state(page, NR_MLOCK);
222 		SetPageMlocked(newpage);
223 		__inc_zone_page_state(newpage, NR_MLOCK);
224 		local_irq_restore(flags);
225 	}
226 }
227 
228 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
229 extern unsigned long vma_address(struct page *page,
230 				 struct vm_area_struct *vma);
231 #endif
232 #else /* !CONFIG_MMU */
233 static inline int mlocked_vma_newpage(struct vm_area_struct *v, struct page *p)
234 {
235 	return 0;
236 }
237 static inline void clear_page_mlock(struct page *page) { }
238 static inline void mlock_vma_page(struct page *page) { }
239 static inline void mlock_migrate_page(struct page *new, struct page *old) { }
240 
241 #endif /* !CONFIG_MMU */
242 
243 /*
244  * Return the mem_map entry representing the 'offset' subpage within
245  * the maximally aligned gigantic page 'base'.  Handle any discontiguity
246  * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
247  */
248 static inline struct page *mem_map_offset(struct page *base, int offset)
249 {
250 	if (unlikely(offset >= MAX_ORDER_NR_PAGES))
251 		return pfn_to_page(page_to_pfn(base) + offset);
252 	return base + offset;
253 }
254 
255 /*
256  * Iterator over all subpages within the maximally aligned gigantic
257  * page 'base'.  Handle any discontiguity in the mem_map.
258  */
259 static inline struct page *mem_map_next(struct page *iter,
260 						struct page *base, int offset)
261 {
262 	if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
263 		unsigned long pfn = page_to_pfn(base) + offset;
264 		if (!pfn_valid(pfn))
265 			return NULL;
266 		return pfn_to_page(pfn);
267 	}
268 	return iter + 1;
269 }
270 
271 /*
272  * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
273  * so all functions starting at paging_init should be marked __init
274  * in those cases. SPARSEMEM, however, allows for memory hotplug,
275  * and alloc_bootmem_node is not used.
276  */
277 #ifdef CONFIG_SPARSEMEM
278 #define __paginginit __meminit
279 #else
280 #define __paginginit __init
281 #endif
282 
283 /* Memory initialisation debug and verification */
284 enum mminit_level {
285 	MMINIT_WARNING,
286 	MMINIT_VERIFY,
287 	MMINIT_TRACE
288 };
289 
290 #ifdef CONFIG_DEBUG_MEMORY_INIT
291 
292 extern int mminit_loglevel;
293 
294 #define mminit_dprintk(level, prefix, fmt, arg...) \
295 do { \
296 	if (level < mminit_loglevel) { \
297 		printk(level <= MMINIT_WARNING ? KERN_WARNING : KERN_DEBUG); \
298 		printk(KERN_CONT "mminit::" prefix " " fmt, ##arg); \
299 	} \
300 } while (0)
301 
302 extern void mminit_verify_pageflags_layout(void);
303 extern void mminit_verify_page_links(struct page *page,
304 		enum zone_type zone, unsigned long nid, unsigned long pfn);
305 extern void mminit_verify_zonelist(void);
306 
307 #else
308 
309 static inline void mminit_dprintk(enum mminit_level level,
310 				const char *prefix, const char *fmt, ...)
311 {
312 }
313 
314 static inline void mminit_verify_pageflags_layout(void)
315 {
316 }
317 
318 static inline void mminit_verify_page_links(struct page *page,
319 		enum zone_type zone, unsigned long nid, unsigned long pfn)
320 {
321 }
322 
323 static inline void mminit_verify_zonelist(void)
324 {
325 }
326 #endif /* CONFIG_DEBUG_MEMORY_INIT */
327 
328 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
329 #if defined(CONFIG_SPARSEMEM)
330 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
331 				unsigned long *end_pfn);
332 #else
333 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
334 				unsigned long *end_pfn)
335 {
336 }
337 #endif /* CONFIG_SPARSEMEM */
338 
339 #define ZONE_RECLAIM_NOSCAN	-2
340 #define ZONE_RECLAIM_FULL	-1
341 #define ZONE_RECLAIM_SOME	0
342 #define ZONE_RECLAIM_SUCCESS	1
343 #endif
344 
345 extern int hwpoison_filter(struct page *p);
346 
347 extern u32 hwpoison_filter_dev_major;
348 extern u32 hwpoison_filter_dev_minor;
349 extern u64 hwpoison_filter_flags_mask;
350 extern u64 hwpoison_filter_flags_value;
351 extern u64 hwpoison_filter_memcg;
352 extern u32 hwpoison_filter_enable;
353 
354 extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
355         unsigned long, unsigned long,
356         unsigned long, unsigned long);
357 
358 extern void set_pageblock_order(void);
359