Loading...
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/fs.h>
15#include <linux/mm.h>
16
17void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
18 unsigned long floor, unsigned long ceiling);
19
20static inline void set_page_count(struct page *page, int v)
21{
22 atomic_set(&page->_count, v);
23}
24
25extern 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 */
32static 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
39/*
40 * Turn a non-refcounted page (->_count == 0) into refcounted with
41 * a count of one.
42 */
43static inline void set_page_refcounted(struct page *page)
44{
45 VM_BUG_ON_PAGE(PageTail(page), page);
46 VM_BUG_ON_PAGE(atomic_read(&page->_count), page);
47 set_page_count(page, 1);
48}
49
50static 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 */
64 VM_BUG_ON_PAGE(atomic_read(&page->first_page->_count) <= 0, page);
65 if (get_page_head)
66 atomic_inc(&page->first_page->_count);
67 get_huge_page_tail(page);
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 */
75static 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 */
89 VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page);
90 atomic_inc(&page->_count);
91 }
92}
93
94extern unsigned long highest_memmap_pfn;
95
96/*
97 * in mm/vmscan.c:
98 */
99extern int isolate_lru_page(struct page *page);
100extern void putback_lru_page(struct page *page);
101extern bool zone_reclaimable(struct zone *zone);
102
103/*
104 * in mm/rmap.c:
105 */
106extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
107
108/*
109 * in mm/page_alloc.c
110 */
111extern void __free_pages_bootmem(struct page *page, unsigned int order);
112extern void prep_compound_page(struct page *page, unsigned long order);
113#ifdef CONFIG_MEMORY_FAILURE
114extern bool is_free_buddy_page(struct page *page);
115#endif
116extern int user_min_free_kbytes;
117
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 */
130struct 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 */
137 bool sync; /* Synchronous migration */
138 bool ignore_skip_hint; /* Scan blocks even if marked skip */
139 bool finished_update_free; /* True when the zone cached pfns are
140 * no longer being updated
141 */
142 bool finished_update_migrate;
143
144 int order; /* order a direct compactor needs */
145 int migratetype; /* MOVABLE, RECLAIMABLE etc */
146 struct zone *zone;
147 bool contended; /* True if a lock was contended */
148};
149
150unsigned long
151isolate_freepages_range(struct compact_control *cc,
152 unsigned long start_pfn, unsigned long end_pfn);
153unsigned long
154isolate_migratepages_range(struct zone *zone, struct compact_control *cc,
155 unsigned long low_pfn, unsigned long end_pfn, bool unevictable);
156
157#endif
158
159/*
160 * This function returns the order of a free page in the buddy system. In
161 * general, page_zone(page)->lock must be held by the caller to prevent the
162 * page from being allocated in parallel and returning garbage as the order.
163 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
164 * page cannot be allocated or merged in parallel.
165 */
166static inline unsigned long page_order(struct page *page)
167{
168 /* PageBuddy() must be checked by the caller */
169 return page_private(page);
170}
171
172/* mm/util.c */
173void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
174 struct vm_area_struct *prev, struct rb_node *rb_parent);
175
176#ifdef CONFIG_MMU
177extern long __mlock_vma_pages_range(struct vm_area_struct *vma,
178 unsigned long start, unsigned long end, int *nonblocking);
179extern void munlock_vma_pages_range(struct vm_area_struct *vma,
180 unsigned long start, unsigned long end);
181static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
182{
183 munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
184}
185
186/*
187 * Called only in fault path, to determine if a new page is being
188 * mapped into a LOCKED vma. If it is, mark page as mlocked.
189 */
190static inline int mlocked_vma_newpage(struct vm_area_struct *vma,
191 struct page *page)
192{
193 VM_BUG_ON_PAGE(PageLRU(page), page);
194
195 if (likely((vma->vm_flags & (VM_LOCKED | VM_SPECIAL)) != VM_LOCKED))
196 return 0;
197
198 if (!TestSetPageMlocked(page)) {
199 mod_zone_page_state(page_zone(page), NR_MLOCK,
200 hpage_nr_pages(page));
201 count_vm_event(UNEVICTABLE_PGMLOCKED);
202 }
203 return 1;
204}
205
206/*
207 * must be called with vma's mmap_sem held for read or write, and page locked.
208 */
209extern void mlock_vma_page(struct page *page);
210extern unsigned int munlock_vma_page(struct page *page);
211
212/*
213 * Clear the page's PageMlocked(). This can be useful in a situation where
214 * we want to unconditionally remove a page from the pagecache -- e.g.,
215 * on truncation or freeing.
216 *
217 * It is legal to call this function for any page, mlocked or not.
218 * If called for a page that is still mapped by mlocked vmas, all we do
219 * is revert to lazy LRU behaviour -- semantics are not broken.
220 */
221extern void clear_page_mlock(struct page *page);
222
223/*
224 * mlock_migrate_page - called only from migrate_page_copy() to
225 * migrate the Mlocked page flag; update statistics.
226 */
227static inline void mlock_migrate_page(struct page *newpage, struct page *page)
228{
229 if (TestClearPageMlocked(page)) {
230 unsigned long flags;
231 int nr_pages = hpage_nr_pages(page);
232
233 local_irq_save(flags);
234 __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
235 SetPageMlocked(newpage);
236 __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
237 local_irq_restore(flags);
238 }
239}
240
241extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
242
243#ifdef CONFIG_TRANSPARENT_HUGEPAGE
244extern unsigned long vma_address(struct page *page,
245 struct vm_area_struct *vma);
246#endif
247#else /* !CONFIG_MMU */
248static inline int mlocked_vma_newpage(struct vm_area_struct *v, struct page *p)
249{
250 return 0;
251}
252static inline void clear_page_mlock(struct page *page) { }
253static inline void mlock_vma_page(struct page *page) { }
254static inline void mlock_migrate_page(struct page *new, struct page *old) { }
255
256#endif /* !CONFIG_MMU */
257
258/*
259 * Return the mem_map entry representing the 'offset' subpage within
260 * the maximally aligned gigantic page 'base'. Handle any discontiguity
261 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
262 */
263static inline struct page *mem_map_offset(struct page *base, int offset)
264{
265 if (unlikely(offset >= MAX_ORDER_NR_PAGES))
266 return pfn_to_page(page_to_pfn(base) + offset);
267 return base + offset;
268}
269
270/*
271 * Iterator over all subpages within the maximally aligned gigantic
272 * page 'base'. Handle any discontiguity in the mem_map.
273 */
274static inline struct page *mem_map_next(struct page *iter,
275 struct page *base, int offset)
276{
277 if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
278 unsigned long pfn = page_to_pfn(base) + offset;
279 if (!pfn_valid(pfn))
280 return NULL;
281 return pfn_to_page(pfn);
282 }
283 return iter + 1;
284}
285
286/*
287 * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
288 * so all functions starting at paging_init should be marked __init
289 * in those cases. SPARSEMEM, however, allows for memory hotplug,
290 * and alloc_bootmem_node is not used.
291 */
292#ifdef CONFIG_SPARSEMEM
293#define __paginginit __meminit
294#else
295#define __paginginit __init
296#endif
297
298/* Memory initialisation debug and verification */
299enum mminit_level {
300 MMINIT_WARNING,
301 MMINIT_VERIFY,
302 MMINIT_TRACE
303};
304
305#ifdef CONFIG_DEBUG_MEMORY_INIT
306
307extern int mminit_loglevel;
308
309#define mminit_dprintk(level, prefix, fmt, arg...) \
310do { \
311 if (level < mminit_loglevel) { \
312 printk(level <= MMINIT_WARNING ? KERN_WARNING : KERN_DEBUG); \
313 printk(KERN_CONT "mminit::" prefix " " fmt, ##arg); \
314 } \
315} while (0)
316
317extern void mminit_verify_pageflags_layout(void);
318extern void mminit_verify_page_links(struct page *page,
319 enum zone_type zone, unsigned long nid, unsigned long pfn);
320extern void mminit_verify_zonelist(void);
321
322#else
323
324static inline void mminit_dprintk(enum mminit_level level,
325 const char *prefix, const char *fmt, ...)
326{
327}
328
329static inline void mminit_verify_pageflags_layout(void)
330{
331}
332
333static inline void mminit_verify_page_links(struct page *page,
334 enum zone_type zone, unsigned long nid, unsigned long pfn)
335{
336}
337
338static inline void mminit_verify_zonelist(void)
339{
340}
341#endif /* CONFIG_DEBUG_MEMORY_INIT */
342
343/* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
344#if defined(CONFIG_SPARSEMEM)
345extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
346 unsigned long *end_pfn);
347#else
348static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
349 unsigned long *end_pfn)
350{
351}
352#endif /* CONFIG_SPARSEMEM */
353
354#define ZONE_RECLAIM_NOSCAN -2
355#define ZONE_RECLAIM_FULL -1
356#define ZONE_RECLAIM_SOME 0
357#define ZONE_RECLAIM_SUCCESS 1
358
359extern int hwpoison_filter(struct page *p);
360
361extern u32 hwpoison_filter_dev_major;
362extern u32 hwpoison_filter_dev_minor;
363extern u64 hwpoison_filter_flags_mask;
364extern u64 hwpoison_filter_flags_value;
365extern u64 hwpoison_filter_memcg;
366extern u32 hwpoison_filter_enable;
367
368extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
369 unsigned long, unsigned long,
370 unsigned long, unsigned long);
371
372extern void set_pageblock_order(void);
373unsigned long reclaim_clean_pages_from_list(struct zone *zone,
374 struct list_head *page_list);
375/* The ALLOC_WMARK bits are used as an index to zone->watermark */
376#define ALLOC_WMARK_MIN WMARK_MIN
377#define ALLOC_WMARK_LOW WMARK_LOW
378#define ALLOC_WMARK_HIGH WMARK_HIGH
379#define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
380
381/* Mask to get the watermark bits */
382#define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
383
384#define ALLOC_HARDER 0x10 /* try to alloc harder */
385#define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
386#define ALLOC_CPUSET 0x40 /* check for correct cpuset */
387#define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
388#define ALLOC_FAIR 0x100 /* fair zone allocation */
389
390#endif /* __MM_INTERNAL_H */
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/fs.h>
15#include <linux/mm.h>
16#include <linux/pagemap.h>
17#include <linux/tracepoint-defs.h>
18
19/*
20 * The set of flags that only affect watermark checking and reclaim
21 * behaviour. This is used by the MM to obey the caller constraints
22 * about IO, FS and watermark checking while ignoring placement
23 * hints such as HIGHMEM usage.
24 */
25#define GFP_RECLAIM_MASK (__GFP_RECLAIM|__GFP_HIGH|__GFP_IO|__GFP_FS|\
26 __GFP_NOWARN|__GFP_REPEAT|__GFP_NOFAIL|\
27 __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC)
28
29/* The GFP flags allowed during early boot */
30#define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_RECLAIM|__GFP_IO|__GFP_FS))
31
32/* Control allocation cpuset and node placement constraints */
33#define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE)
34
35/* Do not use these with a slab allocator */
36#define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
37
38void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
39 unsigned long floor, unsigned long ceiling);
40
41void unmap_page_range(struct mmu_gather *tlb,
42 struct vm_area_struct *vma,
43 unsigned long addr, unsigned long end,
44 struct zap_details *details);
45
46extern int __do_page_cache_readahead(struct address_space *mapping,
47 struct file *filp, pgoff_t offset, unsigned long nr_to_read,
48 unsigned long lookahead_size);
49
50/*
51 * Submit IO for the read-ahead request in file_ra_state.
52 */
53static inline unsigned long ra_submit(struct file_ra_state *ra,
54 struct address_space *mapping, struct file *filp)
55{
56 return __do_page_cache_readahead(mapping, filp,
57 ra->start, ra->size, ra->async_size);
58}
59
60/*
61 * Turn a non-refcounted page (->_count == 0) into refcounted with
62 * a count of one.
63 */
64static inline void set_page_refcounted(struct page *page)
65{
66 VM_BUG_ON_PAGE(PageTail(page), page);
67 VM_BUG_ON_PAGE(page_ref_count(page), page);
68 set_page_count(page, 1);
69}
70
71extern unsigned long highest_memmap_pfn;
72
73/*
74 * in mm/vmscan.c:
75 */
76extern int isolate_lru_page(struct page *page);
77extern void putback_lru_page(struct page *page);
78extern bool zone_reclaimable(struct zone *zone);
79
80/*
81 * in mm/rmap.c:
82 */
83extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
84
85/*
86 * in mm/page_alloc.c
87 */
88
89/*
90 * Structure for holding the mostly immutable allocation parameters passed
91 * between functions involved in allocations, including the alloc_pages*
92 * family of functions.
93 *
94 * nodemask, migratetype and high_zoneidx are initialized only once in
95 * __alloc_pages_nodemask() and then never change.
96 *
97 * zonelist, preferred_zone and classzone_idx are set first in
98 * __alloc_pages_nodemask() for the fast path, and might be later changed
99 * in __alloc_pages_slowpath(). All other functions pass the whole strucure
100 * by a const pointer.
101 */
102struct alloc_context {
103 struct zonelist *zonelist;
104 nodemask_t *nodemask;
105 struct zone *preferred_zone;
106 int classzone_idx;
107 int migratetype;
108 enum zone_type high_zoneidx;
109 bool spread_dirty_pages;
110};
111
112/*
113 * Locate the struct page for both the matching buddy in our
114 * pair (buddy1) and the combined O(n+1) page they form (page).
115 *
116 * 1) Any buddy B1 will have an order O twin B2 which satisfies
117 * the following equation:
118 * B2 = B1 ^ (1 << O)
119 * For example, if the starting buddy (buddy2) is #8 its order
120 * 1 buddy is #10:
121 * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
122 *
123 * 2) Any buddy B will have an order O+1 parent P which
124 * satisfies the following equation:
125 * P = B & ~(1 << O)
126 *
127 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
128 */
129static inline unsigned long
130__find_buddy_index(unsigned long page_idx, unsigned int order)
131{
132 return page_idx ^ (1 << order);
133}
134
135extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn,
136 unsigned long end_pfn, struct zone *zone);
137
138static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn,
139 unsigned long end_pfn, struct zone *zone)
140{
141 if (zone->contiguous)
142 return pfn_to_page(start_pfn);
143
144 return __pageblock_pfn_to_page(start_pfn, end_pfn, zone);
145}
146
147extern int __isolate_free_page(struct page *page, unsigned int order);
148extern void __free_pages_bootmem(struct page *page, unsigned long pfn,
149 unsigned int order);
150extern void prep_compound_page(struct page *page, unsigned int order);
151extern int user_min_free_kbytes;
152
153#if defined CONFIG_COMPACTION || defined CONFIG_CMA
154
155/*
156 * in mm/compaction.c
157 */
158/*
159 * compact_control is used to track pages being migrated and the free pages
160 * they are being migrated to during memory compaction. The free_pfn starts
161 * at the end of a zone and migrate_pfn begins at the start. Movable pages
162 * are moved to the end of a zone during a compaction run and the run
163 * completes when free_pfn <= migrate_pfn
164 */
165struct compact_control {
166 struct list_head freepages; /* List of free pages to migrate to */
167 struct list_head migratepages; /* List of pages being migrated */
168 unsigned long nr_freepages; /* Number of isolated free pages */
169 unsigned long nr_migratepages; /* Number of pages to migrate */
170 unsigned long free_pfn; /* isolate_freepages search base */
171 unsigned long migrate_pfn; /* isolate_migratepages search base */
172 unsigned long last_migrated_pfn;/* Not yet flushed page being freed */
173 enum migrate_mode mode; /* Async or sync migration mode */
174 bool ignore_skip_hint; /* Scan blocks even if marked skip */
175 bool direct_compaction; /* False from kcompactd or /proc/... */
176 int order; /* order a direct compactor needs */
177 const gfp_t gfp_mask; /* gfp mask of a direct compactor */
178 const int alloc_flags; /* alloc flags of a direct compactor */
179 const int classzone_idx; /* zone index of a direct compactor */
180 struct zone *zone;
181 int contended; /* Signal need_sched() or lock
182 * contention detected during
183 * compaction
184 */
185};
186
187unsigned long
188isolate_freepages_range(struct compact_control *cc,
189 unsigned long start_pfn, unsigned long end_pfn);
190unsigned long
191isolate_migratepages_range(struct compact_control *cc,
192 unsigned long low_pfn, unsigned long end_pfn);
193int find_suitable_fallback(struct free_area *area, unsigned int order,
194 int migratetype, bool only_stealable, bool *can_steal);
195
196#endif
197
198/*
199 * This function returns the order of a free page in the buddy system. In
200 * general, page_zone(page)->lock must be held by the caller to prevent the
201 * page from being allocated in parallel and returning garbage as the order.
202 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
203 * page cannot be allocated or merged in parallel. Alternatively, it must
204 * handle invalid values gracefully, and use page_order_unsafe() below.
205 */
206static inline unsigned int page_order(struct page *page)
207{
208 /* PageBuddy() must be checked by the caller */
209 return page_private(page);
210}
211
212/*
213 * Like page_order(), but for callers who cannot afford to hold the zone lock.
214 * PageBuddy() should be checked first by the caller to minimize race window,
215 * and invalid values must be handled gracefully.
216 *
217 * READ_ONCE is used so that if the caller assigns the result into a local
218 * variable and e.g. tests it for valid range before using, the compiler cannot
219 * decide to remove the variable and inline the page_private(page) multiple
220 * times, potentially observing different values in the tests and the actual
221 * use of the result.
222 */
223#define page_order_unsafe(page) READ_ONCE(page_private(page))
224
225static inline bool is_cow_mapping(vm_flags_t flags)
226{
227 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
228}
229
230/*
231 * These three helpers classifies VMAs for virtual memory accounting.
232 */
233
234/*
235 * Executable code area - executable, not writable, not stack
236 */
237static inline bool is_exec_mapping(vm_flags_t flags)
238{
239 return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC;
240}
241
242/*
243 * Stack area - atomatically grows in one direction
244 *
245 * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
246 * do_mmap() forbids all other combinations.
247 */
248static inline bool is_stack_mapping(vm_flags_t flags)
249{
250 return (flags & VM_STACK) == VM_STACK;
251}
252
253/*
254 * Data area - private, writable, not stack
255 */
256static inline bool is_data_mapping(vm_flags_t flags)
257{
258 return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE;
259}
260
261/* mm/util.c */
262void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
263 struct vm_area_struct *prev, struct rb_node *rb_parent);
264
265#ifdef CONFIG_MMU
266extern long populate_vma_page_range(struct vm_area_struct *vma,
267 unsigned long start, unsigned long end, int *nonblocking);
268extern void munlock_vma_pages_range(struct vm_area_struct *vma,
269 unsigned long start, unsigned long end);
270static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
271{
272 munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
273}
274
275/*
276 * must be called with vma's mmap_sem held for read or write, and page locked.
277 */
278extern void mlock_vma_page(struct page *page);
279extern unsigned int munlock_vma_page(struct page *page);
280
281/*
282 * Clear the page's PageMlocked(). This can be useful in a situation where
283 * we want to unconditionally remove a page from the pagecache -- e.g.,
284 * on truncation or freeing.
285 *
286 * It is legal to call this function for any page, mlocked or not.
287 * If called for a page that is still mapped by mlocked vmas, all we do
288 * is revert to lazy LRU behaviour -- semantics are not broken.
289 */
290extern void clear_page_mlock(struct page *page);
291
292/*
293 * mlock_migrate_page - called only from migrate_misplaced_transhuge_page()
294 * (because that does not go through the full procedure of migration ptes):
295 * to migrate the Mlocked page flag; update statistics.
296 */
297static inline void mlock_migrate_page(struct page *newpage, struct page *page)
298{
299 if (TestClearPageMlocked(page)) {
300 int nr_pages = hpage_nr_pages(page);
301
302 /* Holding pmd lock, no change in irq context: __mod is safe */
303 __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
304 SetPageMlocked(newpage);
305 __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
306 }
307}
308
309extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
310
311/*
312 * At what user virtual address is page expected in @vma?
313 */
314static inline unsigned long
315__vma_address(struct page *page, struct vm_area_struct *vma)
316{
317 pgoff_t pgoff = page_to_pgoff(page);
318 return vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
319}
320
321static inline unsigned long
322vma_address(struct page *page, struct vm_area_struct *vma)
323{
324 unsigned long address = __vma_address(page, vma);
325
326 /* page should be within @vma mapping range */
327 VM_BUG_ON_VMA(address < vma->vm_start || address >= vma->vm_end, vma);
328
329 return address;
330}
331
332#else /* !CONFIG_MMU */
333static inline void clear_page_mlock(struct page *page) { }
334static inline void mlock_vma_page(struct page *page) { }
335static inline void mlock_migrate_page(struct page *new, struct page *old) { }
336
337#endif /* !CONFIG_MMU */
338
339/*
340 * Return the mem_map entry representing the 'offset' subpage within
341 * the maximally aligned gigantic page 'base'. Handle any discontiguity
342 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
343 */
344static inline struct page *mem_map_offset(struct page *base, int offset)
345{
346 if (unlikely(offset >= MAX_ORDER_NR_PAGES))
347 return nth_page(base, offset);
348 return base + offset;
349}
350
351/*
352 * Iterator over all subpages within the maximally aligned gigantic
353 * page 'base'. Handle any discontiguity in the mem_map.
354 */
355static inline struct page *mem_map_next(struct page *iter,
356 struct page *base, int offset)
357{
358 if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
359 unsigned long pfn = page_to_pfn(base) + offset;
360 if (!pfn_valid(pfn))
361 return NULL;
362 return pfn_to_page(pfn);
363 }
364 return iter + 1;
365}
366
367/*
368 * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
369 * so all functions starting at paging_init should be marked __init
370 * in those cases. SPARSEMEM, however, allows for memory hotplug,
371 * and alloc_bootmem_node is not used.
372 */
373#ifdef CONFIG_SPARSEMEM
374#define __paginginit __meminit
375#else
376#define __paginginit __init
377#endif
378
379/* Memory initialisation debug and verification */
380enum mminit_level {
381 MMINIT_WARNING,
382 MMINIT_VERIFY,
383 MMINIT_TRACE
384};
385
386#ifdef CONFIG_DEBUG_MEMORY_INIT
387
388extern int mminit_loglevel;
389
390#define mminit_dprintk(level, prefix, fmt, arg...) \
391do { \
392 if (level < mminit_loglevel) { \
393 if (level <= MMINIT_WARNING) \
394 pr_warn("mminit::" prefix " " fmt, ##arg); \
395 else \
396 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
397 } \
398} while (0)
399
400extern void mminit_verify_pageflags_layout(void);
401extern void mminit_verify_zonelist(void);
402#else
403
404static inline void mminit_dprintk(enum mminit_level level,
405 const char *prefix, const char *fmt, ...)
406{
407}
408
409static inline void mminit_verify_pageflags_layout(void)
410{
411}
412
413static inline void mminit_verify_zonelist(void)
414{
415}
416#endif /* CONFIG_DEBUG_MEMORY_INIT */
417
418/* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
419#if defined(CONFIG_SPARSEMEM)
420extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
421 unsigned long *end_pfn);
422#else
423static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
424 unsigned long *end_pfn)
425{
426}
427#endif /* CONFIG_SPARSEMEM */
428
429#define ZONE_RECLAIM_NOSCAN -2
430#define ZONE_RECLAIM_FULL -1
431#define ZONE_RECLAIM_SOME 0
432#define ZONE_RECLAIM_SUCCESS 1
433
434extern int hwpoison_filter(struct page *p);
435
436extern u32 hwpoison_filter_dev_major;
437extern u32 hwpoison_filter_dev_minor;
438extern u64 hwpoison_filter_flags_mask;
439extern u64 hwpoison_filter_flags_value;
440extern u64 hwpoison_filter_memcg;
441extern u32 hwpoison_filter_enable;
442
443extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
444 unsigned long, unsigned long,
445 unsigned long, unsigned long);
446
447extern void set_pageblock_order(void);
448unsigned long reclaim_clean_pages_from_list(struct zone *zone,
449 struct list_head *page_list);
450/* The ALLOC_WMARK bits are used as an index to zone->watermark */
451#define ALLOC_WMARK_MIN WMARK_MIN
452#define ALLOC_WMARK_LOW WMARK_LOW
453#define ALLOC_WMARK_HIGH WMARK_HIGH
454#define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
455
456/* Mask to get the watermark bits */
457#define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
458
459#define ALLOC_HARDER 0x10 /* try to alloc harder */
460#define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
461#define ALLOC_CPUSET 0x40 /* check for correct cpuset */
462#define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
463#define ALLOC_FAIR 0x100 /* fair zone allocation */
464
465enum ttu_flags;
466struct tlbflush_unmap_batch;
467
468#ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
469void try_to_unmap_flush(void);
470void try_to_unmap_flush_dirty(void);
471#else
472static inline void try_to_unmap_flush(void)
473{
474}
475static inline void try_to_unmap_flush_dirty(void)
476{
477}
478
479#endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
480
481extern const struct trace_print_flags pageflag_names[];
482extern const struct trace_print_flags vmaflag_names[];
483extern const struct trace_print_flags gfpflag_names[];
484
485#endif /* __MM_INTERNAL_H */