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 * Structure for holding the mostly immutable allocation parameters passed
114 * between functions involved in allocations, including the alloc_pages*
115 * family of functions.
117 * nodemask, migratetype and high_zoneidx are initialized only once in
118 * __alloc_pages_nodemask() and then never change.
120 * zonelist, preferred_zone and classzone_idx are set first in
121 * __alloc_pages_nodemask() for the fast path, and might be later changed
122 * in __alloc_pages_slowpath(). All other functions pass the whole strucure
123 * by a const pointer.
125 struct alloc_context {
126 struct zonelist *zonelist;
127 nodemask_t *nodemask;
128 struct zone *preferred_zone;
131 enum zone_type high_zoneidx;
135 * Locate the struct page for both the matching buddy in our
136 * pair (buddy1) and the combined O(n+1) page they form (page).
138 * 1) Any buddy B1 will have an order O twin B2 which satisfies
139 * the following equation:
141 * For example, if the starting buddy (buddy2) is #8 its order
143 * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
145 * 2) Any buddy B will have an order O+1 parent P which
146 * satisfies the following equation:
149 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
151 static inline unsigned long
152 __find_buddy_index(unsigned long page_idx, unsigned int order)
154 return page_idx ^ (1 << order);
157 extern int __isolate_free_page(struct page *page, unsigned int order);
158 extern void __free_pages_bootmem(struct page *page, unsigned int order);
159 extern void prep_compound_page(struct page *page, unsigned long order);
160 #ifdef CONFIG_MEMORY_FAILURE
161 extern bool is_free_buddy_page(struct page *page);
163 extern int user_min_free_kbytes;
165 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
171 * compact_control is used to track pages being migrated and the free pages
172 * they are being migrated to during memory compaction. The free_pfn starts
173 * at the end of a zone and migrate_pfn begins at the start. Movable pages
174 * are moved to the end of a zone during a compaction run and the run
175 * completes when free_pfn <= migrate_pfn
177 struct compact_control {
178 struct list_head freepages; /* List of free pages to migrate to */
179 struct list_head migratepages; /* List of pages being migrated */
180 unsigned long nr_freepages; /* Number of isolated free pages */
181 unsigned long nr_migratepages; /* Number of pages to migrate */
182 unsigned long free_pfn; /* isolate_freepages search base */
183 unsigned long migrate_pfn; /* isolate_migratepages search base */
184 enum migrate_mode mode; /* Async or sync migration mode */
185 bool ignore_skip_hint; /* Scan blocks even if marked skip */
186 int order; /* order a direct compactor needs */
187 const gfp_t gfp_mask; /* gfp mask of a direct compactor */
188 const int alloc_flags; /* alloc flags of a direct compactor */
189 const int classzone_idx; /* zone index of a direct compactor */
191 int contended; /* Signal need_sched() or lock
192 * contention detected during
198 isolate_freepages_range(struct compact_control *cc,
199 unsigned long start_pfn, unsigned long end_pfn);
201 isolate_migratepages_range(struct compact_control *cc,
202 unsigned long low_pfn, unsigned long end_pfn);
203 int find_suitable_fallback(struct free_area *area, unsigned int order,
204 int migratetype, bool only_stealable, bool *can_steal);
209 * This function returns the order of a free page in the buddy system. In
210 * general, page_zone(page)->lock must be held by the caller to prevent the
211 * page from being allocated in parallel and returning garbage as the order.
212 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
213 * page cannot be allocated or merged in parallel. Alternatively, it must
214 * handle invalid values gracefully, and use page_order_unsafe() below.
216 static inline unsigned long page_order(struct page *page)
218 /* PageBuddy() must be checked by the caller */
219 return page_private(page);
223 * Like page_order(), but for callers who cannot afford to hold the zone lock.
224 * PageBuddy() should be checked first by the caller to minimize race window,
225 * and invalid values must be handled gracefully.
227 * READ_ONCE is used so that if the caller assigns the result into a local
228 * variable and e.g. tests it for valid range before using, the compiler cannot
229 * decide to remove the variable and inline the page_private(page) multiple
230 * times, potentially observing different values in the tests and the actual
233 #define page_order_unsafe(page) READ_ONCE(page_private(page))
235 static inline bool is_cow_mapping(vm_flags_t flags)
237 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
241 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
242 struct vm_area_struct *prev, struct rb_node *rb_parent);
245 extern long populate_vma_page_range(struct vm_area_struct *vma,
246 unsigned long start, unsigned long end, int *nonblocking);
247 extern void munlock_vma_pages_range(struct vm_area_struct *vma,
248 unsigned long start, unsigned long end);
249 static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
251 munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
255 * must be called with vma's mmap_sem held for read or write, and page locked.
257 extern void mlock_vma_page(struct page *page);
258 extern unsigned int munlock_vma_page(struct page *page);
261 * Clear the page's PageMlocked(). This can be useful in a situation where
262 * we want to unconditionally remove a page from the pagecache -- e.g.,
263 * on truncation or freeing.
265 * It is legal to call this function for any page, mlocked or not.
266 * If called for a page that is still mapped by mlocked vmas, all we do
267 * is revert to lazy LRU behaviour -- semantics are not broken.
269 extern void clear_page_mlock(struct page *page);
272 * mlock_migrate_page - called only from migrate_page_copy() to
273 * migrate the Mlocked page flag; update statistics.
275 static inline void mlock_migrate_page(struct page *newpage, struct page *page)
277 if (TestClearPageMlocked(page)) {
279 int nr_pages = hpage_nr_pages(page);
281 local_irq_save(flags);
282 __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
283 SetPageMlocked(newpage);
284 __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
285 local_irq_restore(flags);
289 extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
291 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
292 extern unsigned long vma_address(struct page *page,
293 struct vm_area_struct *vma);
295 #else /* !CONFIG_MMU */
296 static inline void clear_page_mlock(struct page *page) { }
297 static inline void mlock_vma_page(struct page *page) { }
298 static inline void mlock_migrate_page(struct page *new, struct page *old) { }
300 #endif /* !CONFIG_MMU */
303 * Return the mem_map entry representing the 'offset' subpage within
304 * the maximally aligned gigantic page 'base'. Handle any discontiguity
305 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
307 static inline struct page *mem_map_offset(struct page *base, int offset)
309 if (unlikely(offset >= MAX_ORDER_NR_PAGES))
310 return nth_page(base, offset);
311 return base + offset;
315 * Iterator over all subpages within the maximally aligned gigantic
316 * page 'base'. Handle any discontiguity in the mem_map.
318 static inline struct page *mem_map_next(struct page *iter,
319 struct page *base, int offset)
321 if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
322 unsigned long pfn = page_to_pfn(base) + offset;
325 return pfn_to_page(pfn);
331 * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
332 * so all functions starting at paging_init should be marked __init
333 * in those cases. SPARSEMEM, however, allows for memory hotplug,
334 * and alloc_bootmem_node is not used.
336 #ifdef CONFIG_SPARSEMEM
337 #define __paginginit __meminit
339 #define __paginginit __init
342 /* Memory initialisation debug and verification */
349 #ifdef CONFIG_DEBUG_MEMORY_INIT
351 extern int mminit_loglevel;
353 #define mminit_dprintk(level, prefix, fmt, arg...) \
355 if (level < mminit_loglevel) { \
356 if (level <= MMINIT_WARNING) \
357 printk(KERN_WARNING "mminit::" prefix " " fmt, ##arg); \
359 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
363 extern void mminit_verify_pageflags_layout(void);
364 extern void mminit_verify_page_links(struct page *page,
365 enum zone_type zone, unsigned long nid, unsigned long pfn);
366 extern void mminit_verify_zonelist(void);
370 static inline void mminit_dprintk(enum mminit_level level,
371 const char *prefix, const char *fmt, ...)
375 static inline void mminit_verify_pageflags_layout(void)
379 static inline void mminit_verify_page_links(struct page *page,
380 enum zone_type zone, unsigned long nid, unsigned long pfn)
384 static inline void mminit_verify_zonelist(void)
387 #endif /* CONFIG_DEBUG_MEMORY_INIT */
389 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
390 #if defined(CONFIG_SPARSEMEM)
391 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
392 unsigned long *end_pfn);
394 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
395 unsigned long *end_pfn)
398 #endif /* CONFIG_SPARSEMEM */
400 #define ZONE_RECLAIM_NOSCAN -2
401 #define ZONE_RECLAIM_FULL -1
402 #define ZONE_RECLAIM_SOME 0
403 #define ZONE_RECLAIM_SUCCESS 1
405 extern int hwpoison_filter(struct page *p);
407 extern u32 hwpoison_filter_dev_major;
408 extern u32 hwpoison_filter_dev_minor;
409 extern u64 hwpoison_filter_flags_mask;
410 extern u64 hwpoison_filter_flags_value;
411 extern u64 hwpoison_filter_memcg;
412 extern u32 hwpoison_filter_enable;
414 extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
415 unsigned long, unsigned long,
416 unsigned long, unsigned long);
418 extern void set_pageblock_order(void);
419 unsigned long reclaim_clean_pages_from_list(struct zone *zone,
420 struct list_head *page_list);
421 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
422 #define ALLOC_WMARK_MIN WMARK_MIN
423 #define ALLOC_WMARK_LOW WMARK_LOW
424 #define ALLOC_WMARK_HIGH WMARK_HIGH
425 #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
427 /* Mask to get the watermark bits */
428 #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
430 #define ALLOC_HARDER 0x10 /* try to alloc harder */
431 #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
432 #define ALLOC_CPUSET 0x40 /* check for correct cpuset */
433 #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
434 #define ALLOC_FAIR 0x100 /* fair zone allocation */
436 #endif /* __MM_INTERNAL_H */