1 /* internal.h: mm/ internal definitions
3 * Copyright (C) 2004 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
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.
11 #ifndef __MM_INTERNAL_H
12 #define __MM_INTERNAL_H
17 void free_pgtables(struct mmu_gather
*tlb
, struct vm_area_struct
*start_vma
,
18 unsigned long floor
, unsigned long ceiling
);
20 static inline void set_page_count(struct page
*page
, int v
)
22 atomic_set(&page
->_count
, v
);
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
);
30 * Submit IO for the read-ahead request in file_ra_state.
32 static inline unsigned long ra_submit(struct file_ra_state
*ra
,
33 struct address_space
*mapping
, struct file
*filp
)
35 return __do_page_cache_readahead(mapping
, filp
,
36 ra
->start
, ra
->size
, ra
->async_size
);
40 * Turn a non-refcounted page (->_count == 0) into refcounted with
43 static inline void set_page_refcounted(struct page
*page
)
45 VM_BUG_ON_PAGE(PageTail(page
), page
);
46 VM_BUG_ON_PAGE(atomic_read(&page
->_count
), page
);
47 set_page_count(page
, 1);
50 static inline void __get_page_tail_foll(struct page
*page
,
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.
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.
64 VM_BUG_ON_PAGE(atomic_read(&page
->first_page
->_count
) <= 0, page
);
66 atomic_inc(&page
->first_page
->_count
);
67 get_huge_page_tail(page
);
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.
75 static inline void get_page_foll(struct page
*page
)
77 if (unlikely(PageTail(page
)))
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.
83 __get_page_tail_foll(page
, true);
86 * Getting a normal page or the head of a compound page
87 * requires to already have an elevated page->_count.
89 VM_BUG_ON_PAGE(atomic_read(&page
->_count
) <= 0, page
);
90 atomic_inc(&page
->_count
);
94 extern unsigned long highest_memmap_pfn
;
99 extern int isolate_lru_page(struct page
*page
);
100 extern void putback_lru_page(struct page
*page
);
101 extern bool zone_reclaimable(struct zone
*zone
);
106 extern pmd_t
*mm_find_pmd(struct mm_struct
*mm
, unsigned long address
);
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).
116 * 1) Any buddy B1 will have an order O twin B2 which satisfies
117 * the following equation:
119 * For example, if the starting buddy (buddy2) is #8 its order
121 * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
123 * 2) Any buddy B will have an order O+1 parent P which
124 * satisfies the following equation:
127 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
129 static inline unsigned long
130 __find_buddy_index(unsigned long page_idx
, unsigned int order
)
132 return page_idx
^ (1 << order
);
135 extern int __isolate_free_page(struct page
*page
, unsigned int order
);
136 extern void __free_pages_bootmem(struct page
*page
, unsigned int order
);
137 extern void prep_compound_page(struct page
*page
, unsigned long order
);
138 #ifdef CONFIG_MEMORY_FAILURE
139 extern bool is_free_buddy_page(struct page
*page
);
141 extern int user_min_free_kbytes
;
143 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
149 * compact_control is used to track pages being migrated and the free pages
150 * they are being migrated to during memory compaction. The free_pfn starts
151 * at the end of a zone and migrate_pfn begins at the start. Movable pages
152 * are moved to the end of a zone during a compaction run and the run
153 * completes when free_pfn <= migrate_pfn
155 struct compact_control
{
156 struct list_head freepages
; /* List of free pages to migrate to */
157 struct list_head migratepages
; /* List of pages being migrated */
158 unsigned long nr_freepages
; /* Number of isolated free pages */
159 unsigned long nr_migratepages
; /* Number of pages to migrate */
160 unsigned long free_pfn
; /* isolate_freepages search base */
161 unsigned long migrate_pfn
; /* isolate_migratepages search base */
162 enum migrate_mode mode
; /* Async or sync migration mode */
163 bool ignore_skip_hint
; /* Scan blocks even if marked skip */
164 bool finished_update_free
; /* True when the zone cached pfns are
165 * no longer being updated
167 bool finished_update_migrate
;
169 int order
; /* order a direct compactor needs */
170 const gfp_t gfp_mask
; /* gfp mask of a direct compactor */
172 int contended
; /* Signal need_sched() or lock
173 * contention detected during
179 isolate_freepages_range(struct compact_control
*cc
,
180 unsigned long start_pfn
, unsigned long end_pfn
);
182 isolate_migratepages_range(struct compact_control
*cc
,
183 unsigned long low_pfn
, unsigned long end_pfn
);
188 * This function returns the order of a free page in the buddy system. In
189 * general, page_zone(page)->lock must be held by the caller to prevent the
190 * page from being allocated in parallel and returning garbage as the order.
191 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
192 * page cannot be allocated or merged in parallel. Alternatively, it must
193 * handle invalid values gracefully, and use page_order_unsafe() below.
195 static inline unsigned long page_order(struct page
*page
)
197 /* PageBuddy() must be checked by the caller */
198 return page_private(page
);
202 * Like page_order(), but for callers who cannot afford to hold the zone lock.
203 * PageBuddy() should be checked first by the caller to minimize race window,
204 * and invalid values must be handled gracefully.
206 * ACCESS_ONCE is used so that if the caller assigns the result into a local
207 * variable and e.g. tests it for valid range before using, the compiler cannot
208 * decide to remove the variable and inline the page_private(page) multiple
209 * times, potentially observing different values in the tests and the actual
212 #define page_order_unsafe(page) ACCESS_ONCE(page_private(page))
214 static inline bool is_cow_mapping(vm_flags_t flags
)
216 return (flags
& (VM_SHARED
| VM_MAYWRITE
)) == VM_MAYWRITE
;
220 void __vma_link_list(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
221 struct vm_area_struct
*prev
, struct rb_node
*rb_parent
);
224 extern long __mlock_vma_pages_range(struct vm_area_struct
*vma
,
225 unsigned long start
, unsigned long end
, int *nonblocking
);
226 extern void munlock_vma_pages_range(struct vm_area_struct
*vma
,
227 unsigned long start
, unsigned long end
);
228 static inline void munlock_vma_pages_all(struct vm_area_struct
*vma
)
230 munlock_vma_pages_range(vma
, vma
->vm_start
, vma
->vm_end
);
234 * must be called with vma's mmap_sem held for read or write, and page locked.
236 extern void mlock_vma_page(struct page
*page
);
237 extern unsigned int munlock_vma_page(struct page
*page
);
240 * Clear the page's PageMlocked(). This can be useful in a situation where
241 * we want to unconditionally remove a page from the pagecache -- e.g.,
242 * on truncation or freeing.
244 * It is legal to call this function for any page, mlocked or not.
245 * If called for a page that is still mapped by mlocked vmas, all we do
246 * is revert to lazy LRU behaviour -- semantics are not broken.
248 extern void clear_page_mlock(struct page
*page
);
251 * mlock_migrate_page - called only from migrate_page_copy() to
252 * migrate the Mlocked page flag; update statistics.
254 static inline void mlock_migrate_page(struct page
*newpage
, struct page
*page
)
256 if (TestClearPageMlocked(page
)) {
258 int nr_pages
= hpage_nr_pages(page
);
260 local_irq_save(flags
);
261 __mod_zone_page_state(page_zone(page
), NR_MLOCK
, -nr_pages
);
262 SetPageMlocked(newpage
);
263 __mod_zone_page_state(page_zone(newpage
), NR_MLOCK
, nr_pages
);
264 local_irq_restore(flags
);
268 extern pmd_t
maybe_pmd_mkwrite(pmd_t pmd
, struct vm_area_struct
*vma
);
270 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
271 extern unsigned long vma_address(struct page
*page
,
272 struct vm_area_struct
*vma
);
274 #else /* !CONFIG_MMU */
275 static inline void clear_page_mlock(struct page
*page
) { }
276 static inline void mlock_vma_page(struct page
*page
) { }
277 static inline void mlock_migrate_page(struct page
*new, struct page
*old
) { }
279 #endif /* !CONFIG_MMU */
282 * Return the mem_map entry representing the 'offset' subpage within
283 * the maximally aligned gigantic page 'base'. Handle any discontiguity
284 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
286 static inline struct page
*mem_map_offset(struct page
*base
, int offset
)
288 if (unlikely(offset
>= MAX_ORDER_NR_PAGES
))
289 return nth_page(base
, offset
);
290 return base
+ offset
;
294 * Iterator over all subpages within the maximally aligned gigantic
295 * page 'base'. Handle any discontiguity in the mem_map.
297 static inline struct page
*mem_map_next(struct page
*iter
,
298 struct page
*base
, int offset
)
300 if (unlikely((offset
& (MAX_ORDER_NR_PAGES
- 1)) == 0)) {
301 unsigned long pfn
= page_to_pfn(base
) + offset
;
304 return pfn_to_page(pfn
);
310 * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
311 * so all functions starting at paging_init should be marked __init
312 * in those cases. SPARSEMEM, however, allows for memory hotplug,
313 * and alloc_bootmem_node is not used.
315 #ifdef CONFIG_SPARSEMEM
316 #define __paginginit __meminit
318 #define __paginginit __init
321 /* Memory initialisation debug and verification */
328 #ifdef CONFIG_DEBUG_MEMORY_INIT
330 extern int mminit_loglevel
;
332 #define mminit_dprintk(level, prefix, fmt, arg...) \
334 if (level < mminit_loglevel) { \
335 printk(level <= MMINIT_WARNING ? KERN_WARNING : KERN_DEBUG); \
336 printk(KERN_CONT "mminit::" prefix " " fmt, ##arg); \
340 extern void mminit_verify_pageflags_layout(void);
341 extern void mminit_verify_page_links(struct page
*page
,
342 enum zone_type zone
, unsigned long nid
, unsigned long pfn
);
343 extern void mminit_verify_zonelist(void);
347 static inline void mminit_dprintk(enum mminit_level level
,
348 const char *prefix
, const char *fmt
, ...)
352 static inline void mminit_verify_pageflags_layout(void)
356 static inline void mminit_verify_page_links(struct page
*page
,
357 enum zone_type zone
, unsigned long nid
, unsigned long pfn
)
361 static inline void mminit_verify_zonelist(void)
364 #endif /* CONFIG_DEBUG_MEMORY_INIT */
366 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
367 #if defined(CONFIG_SPARSEMEM)
368 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn
,
369 unsigned long *end_pfn
);
371 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn
,
372 unsigned long *end_pfn
)
375 #endif /* CONFIG_SPARSEMEM */
377 #define ZONE_RECLAIM_NOSCAN -2
378 #define ZONE_RECLAIM_FULL -1
379 #define ZONE_RECLAIM_SOME 0
380 #define ZONE_RECLAIM_SUCCESS 1
382 extern int hwpoison_filter(struct page
*p
);
384 extern u32 hwpoison_filter_dev_major
;
385 extern u32 hwpoison_filter_dev_minor
;
386 extern u64 hwpoison_filter_flags_mask
;
387 extern u64 hwpoison_filter_flags_value
;
388 extern u64 hwpoison_filter_memcg
;
389 extern u32 hwpoison_filter_enable
;
391 extern unsigned long vm_mmap_pgoff(struct file
*, unsigned long,
392 unsigned long, unsigned long,
393 unsigned long, unsigned long);
395 extern void set_pageblock_order(void);
396 unsigned long reclaim_clean_pages_from_list(struct zone
*zone
,
397 struct list_head
*page_list
);
398 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
399 #define ALLOC_WMARK_MIN WMARK_MIN
400 #define ALLOC_WMARK_LOW WMARK_LOW
401 #define ALLOC_WMARK_HIGH WMARK_HIGH
402 #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
404 /* Mask to get the watermark bits */
405 #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
407 #define ALLOC_HARDER 0x10 /* try to alloc harder */
408 #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
409 #define ALLOC_CPUSET 0x40 /* check for correct cpuset */
410 #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
411 #define ALLOC_FAIR 0x100 /* fair zone allocation */
413 #endif /* __MM_INTERNAL_H */