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
15 #include <linux/rmap.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);
26 * Turn a non-refcounted page (->_count == 0) into refcounted with
29 static inline void set_page_refcounted(struct page *page)
31 VM_BUG_ON(PageTail(page));
32 VM_BUG_ON(atomic_read(&page->_count));
33 set_page_count(page, 1);
36 static inline void __put_page(struct page *page)
38 atomic_dec(&page->_count);
41 static inline void __get_page_tail_foll(struct page *page,
45 * If we're getting a tail page, the elevated page->_count is
46 * required only in the head page and we will elevate the head
47 * page->_count and tail page->_mapcount.
49 * We elevate page_tail->_mapcount for tail pages to force
50 * page_tail->_count to be zero at all times to avoid getting
51 * false positives from get_page_unless_zero() with
52 * speculative page access (like in
53 * page_cache_get_speculative()) on tail pages.
55 VM_BUG_ON(atomic_read(&page->first_page->_count) <= 0);
56 VM_BUG_ON(atomic_read(&page->_count) != 0);
57 VM_BUG_ON(page_mapcount(page) < 0);
59 atomic_inc(&page->first_page->_count);
60 atomic_inc(&page->_mapcount);
64 * This is meant to be called as the FOLL_GET operation of
65 * follow_page() and it must be called while holding the proper PT
66 * lock while the pte (or pmd_trans_huge) is still mapping the page.
68 static inline void get_page_foll(struct page *page)
70 if (unlikely(PageTail(page)))
72 * This is safe only because
73 * __split_huge_page_refcount() can't run under
74 * get_page_foll() because we hold the proper PT lock.
76 __get_page_tail_foll(page, true);
79 * Getting a normal page or the head of a compound page
80 * requires to already have an elevated page->_count.
82 VM_BUG_ON(atomic_read(&page->_count) <= 0);
83 atomic_inc(&page->_count);
87 extern unsigned long highest_memmap_pfn;
92 extern int isolate_lru_page(struct page *page);
93 extern void putback_lru_page(struct page *page);
98 extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
103 extern void __free_pages_bootmem(struct page *page, unsigned int order);
104 extern void prep_compound_page(struct page *page, unsigned long order);
105 #ifdef CONFIG_MEMORY_FAILURE
106 extern bool is_free_buddy_page(struct page *page);
109 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
115 * compact_control is used to track pages being migrated and the free pages
116 * they are being migrated to during memory compaction. The free_pfn starts
117 * at the end of a zone and migrate_pfn begins at the start. Movable pages
118 * are moved to the end of a zone during a compaction run and the run
119 * completes when free_pfn <= migrate_pfn
121 struct compact_control {
122 struct list_head freepages; /* List of free pages to migrate to */
123 struct list_head migratepages; /* List of pages being migrated */
124 unsigned long nr_freepages; /* Number of isolated free pages */
125 unsigned long nr_migratepages; /* Number of pages to migrate */
126 unsigned long free_pfn; /* isolate_freepages search base */
127 unsigned long migrate_pfn; /* isolate_migratepages search base */
128 bool sync; /* Synchronous migration */
129 bool ignore_skip_hint; /* Scan blocks even if marked skip */
130 bool finished_update_free; /* True when the zone cached pfns are
131 * no longer being updated
133 bool finished_update_migrate;
135 int order; /* order a direct compactor needs */
136 int migratetype; /* MOVABLE, RECLAIMABLE etc */
138 bool contended; /* True if a lock was contended */
139 struct page **page; /* Page captured of requested size */
143 isolate_freepages_range(struct compact_control *cc,
144 unsigned long start_pfn, unsigned long end_pfn);
146 isolate_migratepages_range(struct zone *zone, struct compact_control *cc,
147 unsigned long low_pfn, unsigned long end_pfn, bool unevictable);
152 * function for dealing with page's order in buddy system.
153 * zone->lock is already acquired when we use these.
154 * So, we don't need atomic page->flags operations here.
156 static inline unsigned long page_order(struct page *page)
158 /* PageBuddy() must be checked by the caller */
159 return page_private(page);
163 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
164 struct vm_area_struct *prev, struct rb_node *rb_parent);
167 extern long mlock_vma_pages_range(struct vm_area_struct *vma,
168 unsigned long start, unsigned long end);
169 extern void munlock_vma_pages_range(struct vm_area_struct *vma,
170 unsigned long start, unsigned long end);
171 static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
173 munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
177 * Called only in fault path, to determine if a new page is being
178 * mapped into a LOCKED vma. If it is, mark page as mlocked.
180 static inline int mlocked_vma_newpage(struct vm_area_struct *vma,
183 VM_BUG_ON(PageLRU(page));
185 if (likely((vma->vm_flags & (VM_LOCKED | VM_SPECIAL)) != VM_LOCKED))
188 if (!TestSetPageMlocked(page)) {
189 mod_zone_page_state(page_zone(page), NR_MLOCK,
190 hpage_nr_pages(page));
191 count_vm_event(UNEVICTABLE_PGMLOCKED);
197 * must be called with vma's mmap_sem held for read or write, and page locked.
199 extern void mlock_vma_page(struct page *page);
200 extern void munlock_vma_page(struct page *page);
203 * Clear the page's PageMlocked(). This can be useful in a situation where
204 * we want to unconditionally remove a page from the pagecache -- e.g.,
205 * on truncation or freeing.
207 * It is legal to call this function for any page, mlocked or not.
208 * If called for a page that is still mapped by mlocked vmas, all we do
209 * is revert to lazy LRU behaviour -- semantics are not broken.
211 extern void clear_page_mlock(struct page *page);
214 * mlock_migrate_page - called only from migrate_page_copy() to
215 * migrate the Mlocked page flag; update statistics.
217 static inline void mlock_migrate_page(struct page *newpage, struct page *page)
219 if (TestClearPageMlocked(page)) {
222 local_irq_save(flags);
223 __dec_zone_page_state(page, NR_MLOCK);
224 SetPageMlocked(newpage);
225 __inc_zone_page_state(newpage, NR_MLOCK);
226 local_irq_restore(flags);
230 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
231 extern unsigned long vma_address(struct page *page,
232 struct vm_area_struct *vma);
234 #else /* !CONFIG_MMU */
235 static inline int mlocked_vma_newpage(struct vm_area_struct *v, struct page *p)
239 static inline void clear_page_mlock(struct page *page) { }
240 static inline void mlock_vma_page(struct page *page) { }
241 static inline void mlock_migrate_page(struct page *new, struct page *old) { }
243 #endif /* !CONFIG_MMU */
246 * Return the mem_map entry representing the 'offset' subpage within
247 * the maximally aligned gigantic page 'base'. Handle any discontiguity
248 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
250 static inline struct page *mem_map_offset(struct page *base, int offset)
252 if (unlikely(offset >= MAX_ORDER_NR_PAGES))
253 return pfn_to_page(page_to_pfn(base) + offset);
254 return base + offset;
258 * Iterator over all subpages within the maximally aligned gigantic
259 * page 'base'. Handle any discontiguity in the mem_map.
261 static inline struct page *mem_map_next(struct page *iter,
262 struct page *base, int offset)
264 if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
265 unsigned long pfn = page_to_pfn(base) + offset;
268 return pfn_to_page(pfn);
274 * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
275 * so all functions starting at paging_init should be marked __init
276 * in those cases. SPARSEMEM, however, allows for memory hotplug,
277 * and alloc_bootmem_node is not used.
279 #ifdef CONFIG_SPARSEMEM
280 #define __paginginit __meminit
282 #define __paginginit __init
285 /* Memory initialisation debug and verification */
292 #ifdef CONFIG_DEBUG_MEMORY_INIT
294 extern int mminit_loglevel;
296 #define mminit_dprintk(level, prefix, fmt, arg...) \
298 if (level < mminit_loglevel) { \
299 printk(level <= MMINIT_WARNING ? KERN_WARNING : KERN_DEBUG); \
300 printk(KERN_CONT "mminit::" prefix " " fmt, ##arg); \
304 extern void mminit_verify_pageflags_layout(void);
305 extern void mminit_verify_page_links(struct page *page,
306 enum zone_type zone, unsigned long nid, unsigned long pfn);
307 extern void mminit_verify_zonelist(void);
311 static inline void mminit_dprintk(enum mminit_level level,
312 const char *prefix, const char *fmt, ...)
316 static inline void mminit_verify_pageflags_layout(void)
320 static inline void mminit_verify_page_links(struct page *page,
321 enum zone_type zone, unsigned long nid, unsigned long pfn)
325 static inline void mminit_verify_zonelist(void)
328 #endif /* CONFIG_DEBUG_MEMORY_INIT */
330 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
331 #if defined(CONFIG_SPARSEMEM)
332 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
333 unsigned long *end_pfn);
335 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
336 unsigned long *end_pfn)
339 #endif /* CONFIG_SPARSEMEM */
341 #define ZONE_RECLAIM_NOSCAN -2
342 #define ZONE_RECLAIM_FULL -1
343 #define ZONE_RECLAIM_SOME 0
344 #define ZONE_RECLAIM_SUCCESS 1
346 extern int hwpoison_filter(struct page *p);
348 extern u32 hwpoison_filter_dev_major;
349 extern u32 hwpoison_filter_dev_minor;
350 extern u64 hwpoison_filter_flags_mask;
351 extern u64 hwpoison_filter_flags_value;
352 extern u64 hwpoison_filter_memcg;
353 extern u32 hwpoison_filter_enable;
355 extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
356 unsigned long, unsigned long,
357 unsigned long, unsigned long);
359 extern void set_pageblock_order(void);
360 unsigned long reclaim_clean_pages_from_list(struct zone *zone,
361 struct list_head *page_list);
362 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
363 #define ALLOC_WMARK_MIN WMARK_MIN
364 #define ALLOC_WMARK_LOW WMARK_LOW
365 #define ALLOC_WMARK_HIGH WMARK_HIGH
366 #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
368 /* Mask to get the watermark bits */
369 #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
371 #define ALLOC_HARDER 0x10 /* try to alloc harder */
372 #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
373 #define ALLOC_CPUSET 0x40 /* check for correct cpuset */
374 #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
377 * Unnecessary readahead harms performance, especially for SSD devices, where
378 * large reads are significantly more expensive than small ones.
379 * These implements simple swap random access detection. In swap page fault: if
380 * the page is found in swapcache, decrease a counter in the vma, otherwise we
381 * need to perform sync swapin and the counter is increased. Optionally swapin
382 * will perform readahead if the counter is below a threshold.
385 #define SWAPRA_MISS_THRESHOLD (100)
386 #define SWAPRA_MAX_MISS ((SWAPRA_MISS_THRESHOLD) * 10)
387 static inline void swap_cache_hit(struct vm_area_struct *vma)
389 if (vma && vma->anon_vma)
390 atomic_dec_if_positive(&vma->anon_vma->swapra_miss);
393 static inline void swap_cache_miss(struct vm_area_struct *vma)
395 if (!vma || !vma->anon_vma)
397 if (atomic_read(&vma->anon_vma->swapra_miss) < SWAPRA_MAX_MISS)
398 atomic_inc(&vma->anon_vma->swapra_miss);
401 static inline int swap_cache_skip_readahead(struct vm_area_struct *vma)
403 if (!vma || !vma->anon_vma)
405 return atomic_read(&vma->anon_vma->swapra_miss) >
406 SWAPRA_MISS_THRESHOLD;
409 static inline void swap_cache_hit(struct vm_area_struct *vma)
413 static inline void swap_cache_miss(struct vm_area_struct *vma)
417 static inline int swap_cache_skip_readahead(struct vm_area_struct *vma)
421 #endif /* CONFIG_SWAP */
423 #endif /* __MM_INTERNAL_H */