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
16 #include <linux/pagemap.h>
17 #include <linux/tracepoint-defs.h>
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.
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)
29 /* The GFP flags allowed during early boot */
30 #define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_RECLAIM|__GFP_IO|__GFP_FS))
32 /* Control allocation cpuset and node placement constraints */
33 #define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE)
35 /* Do not use these with a slab allocator */
36 #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
38 extern int do_swap_page(struct mm_struct *mm, struct vm_area_struct *vma,
39 unsigned long address, pte_t *page_table, pmd_t *pmd,
40 unsigned int flags, pte_t orig_pte);
42 void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
43 unsigned long floor, unsigned long ceiling);
45 void unmap_page_range(struct mmu_gather *tlb,
46 struct vm_area_struct *vma,
47 unsigned long addr, unsigned long end,
48 struct zap_details *details);
50 static inline void set_page_count(struct page *page, int v)
52 atomic_set(&page->_count, v);
55 extern int __do_page_cache_readahead(struct address_space *mapping,
56 struct file *filp, pgoff_t offset, unsigned long nr_to_read,
57 unsigned long lookahead_size);
60 * Submit IO for the read-ahead request in file_ra_state.
62 static inline unsigned long ra_submit(struct file_ra_state *ra,
63 struct address_space *mapping, struct file *filp)
65 return __do_page_cache_readahead(mapping, filp,
66 ra->start, ra->size, ra->async_size);
70 * Turn a non-refcounted page (->_count == 0) into refcounted with
73 static inline void set_page_refcounted(struct page *page)
75 VM_BUG_ON_PAGE(PageTail(page), page);
76 VM_BUG_ON_PAGE(atomic_read(&page->_count), page);
77 set_page_count(page, 1);
80 extern unsigned long highest_memmap_pfn;
85 extern int isolate_lru_page(struct page *page);
86 extern void putback_lru_page(struct page *page);
87 extern bool zone_reclaimable(struct zone *zone);
92 extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
99 * Structure for holding the mostly immutable allocation parameters passed
100 * between functions involved in allocations, including the alloc_pages*
101 * family of functions.
103 * nodemask, migratetype and high_zoneidx are initialized only once in
104 * __alloc_pages_nodemask() and then never change.
106 * zonelist, preferred_zone and classzone_idx are set first in
107 * __alloc_pages_nodemask() for the fast path, and might be later changed
108 * in __alloc_pages_slowpath(). All other functions pass the whole strucure
109 * by a const pointer.
111 struct alloc_context {
112 struct zonelist *zonelist;
113 nodemask_t *nodemask;
114 struct zone *preferred_zone;
117 enum zone_type high_zoneidx;
118 bool spread_dirty_pages;
122 * Locate the struct page for both the matching buddy in our
123 * pair (buddy1) and the combined O(n+1) page they form (page).
125 * 1) Any buddy B1 will have an order O twin B2 which satisfies
126 * the following equation:
128 * For example, if the starting buddy (buddy2) is #8 its order
130 * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
132 * 2) Any buddy B will have an order O+1 parent P which
133 * satisfies the following equation:
136 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
138 static inline unsigned long
139 __find_buddy_index(unsigned long page_idx, unsigned int order)
141 return page_idx ^ (1 << order);
144 extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn,
145 unsigned long end_pfn, struct zone *zone);
147 static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn,
148 unsigned long end_pfn, struct zone *zone)
150 if (zone->contiguous)
151 return pfn_to_page(start_pfn);
153 return __pageblock_pfn_to_page(start_pfn, end_pfn, zone);
156 extern int __isolate_free_page(struct page *page, unsigned int order);
157 extern void __free_pages_bootmem(struct page *page, unsigned long pfn,
159 extern void prep_compound_page(struct page *page, unsigned int order);
160 extern int user_min_free_kbytes;
162 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
168 * compact_control is used to track pages being migrated and the free pages
169 * they are being migrated to during memory compaction. The free_pfn starts
170 * at the end of a zone and migrate_pfn begins at the start. Movable pages
171 * are moved to the end of a zone during a compaction run and the run
172 * completes when free_pfn <= migrate_pfn
174 struct compact_control {
175 struct list_head freepages; /* List of free pages to migrate to */
176 struct list_head migratepages; /* List of pages being migrated */
177 unsigned long nr_freepages; /* Number of isolated free pages */
178 unsigned long nr_migratepages; /* Number of pages to migrate */
179 unsigned long free_pfn; /* isolate_freepages search base */
180 unsigned long migrate_pfn; /* isolate_migratepages search base */
181 unsigned long last_migrated_pfn;/* Not yet flushed page being freed */
182 enum migrate_mode mode; /* Async or sync migration mode */
183 bool ignore_skip_hint; /* Scan blocks even if marked skip */
184 bool direct_compaction; /* False from kcompactd or /proc/... */
185 int order; /* order a direct compactor needs */
186 const gfp_t gfp_mask; /* gfp mask of a direct compactor */
187 const int alloc_flags; /* alloc flags of a direct compactor */
188 const int classzone_idx; /* zone index of a direct compactor */
190 int contended; /* Signal need_sched() or lock
191 * contention detected during
197 isolate_freepages_range(struct compact_control *cc,
198 unsigned long start_pfn, unsigned long end_pfn);
200 isolate_migratepages_range(struct compact_control *cc,
201 unsigned long low_pfn, unsigned long end_pfn);
202 int find_suitable_fallback(struct free_area *area, unsigned int order,
203 int migratetype, bool only_stealable, bool *can_steal);
208 * This function returns the order of a free page in the buddy system. In
209 * general, page_zone(page)->lock must be held by the caller to prevent the
210 * page from being allocated in parallel and returning garbage as the order.
211 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
212 * page cannot be allocated or merged in parallel. Alternatively, it must
213 * handle invalid values gracefully, and use page_order_unsafe() below.
215 static inline unsigned int page_order(struct page *page)
217 /* PageBuddy() must be checked by the caller */
218 return page_private(page);
222 * Like page_order(), but for callers who cannot afford to hold the zone lock.
223 * PageBuddy() should be checked first by the caller to minimize race window,
224 * and invalid values must be handled gracefully.
226 * READ_ONCE is used so that if the caller assigns the result into a local
227 * variable and e.g. tests it for valid range before using, the compiler cannot
228 * decide to remove the variable and inline the page_private(page) multiple
229 * times, potentially observing different values in the tests and the actual
232 #define page_order_unsafe(page) READ_ONCE(page_private(page))
234 static inline bool is_cow_mapping(vm_flags_t flags)
236 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
240 * These three helpers classifies VMAs for virtual memory accounting.
244 * Executable code area - executable, not writable, not stack
246 static inline bool is_exec_mapping(vm_flags_t flags)
248 return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC;
252 * Stack area - atomatically grows in one direction
254 * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
255 * do_mmap() forbids all other combinations.
257 static inline bool is_stack_mapping(vm_flags_t flags)
259 return (flags & VM_STACK) == VM_STACK;
263 * Data area - private, writable, not stack
265 static inline bool is_data_mapping(vm_flags_t flags)
267 return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE;
271 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
272 struct vm_area_struct *prev, struct rb_node *rb_parent);
275 extern long populate_vma_page_range(struct vm_area_struct *vma,
276 unsigned long start, unsigned long end, int *nonblocking);
277 extern void munlock_vma_pages_range(struct vm_area_struct *vma,
278 unsigned long start, unsigned long end);
279 static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
281 munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
285 * must be called with vma's mmap_sem held for read or write, and page locked.
287 extern void mlock_vma_page(struct page *page);
288 extern unsigned int munlock_vma_page(struct page *page);
291 * Clear the page's PageMlocked(). This can be useful in a situation where
292 * we want to unconditionally remove a page from the pagecache -- e.g.,
293 * on truncation or freeing.
295 * It is legal to call this function for any page, mlocked or not.
296 * If called for a page that is still mapped by mlocked vmas, all we do
297 * is revert to lazy LRU behaviour -- semantics are not broken.
299 extern void clear_page_mlock(struct page *page);
302 * mlock_migrate_page - called only from migrate_misplaced_transhuge_page()
303 * (because that does not go through the full procedure of migration ptes):
304 * to migrate the Mlocked page flag; update statistics.
306 static inline void mlock_migrate_page(struct page *newpage, struct page *page)
308 if (TestClearPageMlocked(page)) {
309 int nr_pages = hpage_nr_pages(page);
311 /* Holding pmd lock, no change in irq context: __mod is safe */
312 __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
313 SetPageMlocked(newpage);
314 __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
318 extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
321 * At what user virtual address is page expected in @vma?
323 static inline unsigned long
324 __vma_address(struct page *page, struct vm_area_struct *vma)
326 pgoff_t pgoff = page_to_pgoff(page);
327 return vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
330 static inline unsigned long
331 vma_address(struct page *page, struct vm_area_struct *vma)
333 unsigned long address = __vma_address(page, vma);
335 /* page should be within @vma mapping range */
336 VM_BUG_ON_VMA(address < vma->vm_start || address >= vma->vm_end, vma);
341 #else /* !CONFIG_MMU */
342 static inline void clear_page_mlock(struct page *page) { }
343 static inline void mlock_vma_page(struct page *page) { }
344 static inline void mlock_migrate_page(struct page *new, struct page *old) { }
346 #endif /* !CONFIG_MMU */
349 * Return the mem_map entry representing the 'offset' subpage within
350 * the maximally aligned gigantic page 'base'. Handle any discontiguity
351 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
353 static inline struct page *mem_map_offset(struct page *base, int offset)
355 if (unlikely(offset >= MAX_ORDER_NR_PAGES))
356 return nth_page(base, offset);
357 return base + offset;
361 * Iterator over all subpages within the maximally aligned gigantic
362 * page 'base'. Handle any discontiguity in the mem_map.
364 static inline struct page *mem_map_next(struct page *iter,
365 struct page *base, int offset)
367 if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
368 unsigned long pfn = page_to_pfn(base) + offset;
371 return pfn_to_page(pfn);
377 * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
378 * so all functions starting at paging_init should be marked __init
379 * in those cases. SPARSEMEM, however, allows for memory hotplug,
380 * and alloc_bootmem_node is not used.
382 #ifdef CONFIG_SPARSEMEM
383 #define __paginginit __meminit
385 #define __paginginit __init
388 /* Memory initialisation debug and verification */
395 #ifdef CONFIG_DEBUG_MEMORY_INIT
397 extern int mminit_loglevel;
399 #define mminit_dprintk(level, prefix, fmt, arg...) \
401 if (level < mminit_loglevel) { \
402 if (level <= MMINIT_WARNING) \
403 printk(KERN_WARNING "mminit::" prefix " " fmt, ##arg); \
405 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
409 extern void mminit_verify_pageflags_layout(void);
410 extern void mminit_verify_zonelist(void);
413 static inline void mminit_dprintk(enum mminit_level level,
414 const char *prefix, const char *fmt, ...)
418 static inline void mminit_verify_pageflags_layout(void)
422 static inline void mminit_verify_zonelist(void)
425 #endif /* CONFIG_DEBUG_MEMORY_INIT */
427 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
428 #if defined(CONFIG_SPARSEMEM)
429 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
430 unsigned long *end_pfn);
432 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
433 unsigned long *end_pfn)
436 #endif /* CONFIG_SPARSEMEM */
438 #define ZONE_RECLAIM_NOSCAN -2
439 #define ZONE_RECLAIM_FULL -1
440 #define ZONE_RECLAIM_SOME 0
441 #define ZONE_RECLAIM_SUCCESS 1
443 extern int hwpoison_filter(struct page *p);
445 extern u32 hwpoison_filter_dev_major;
446 extern u32 hwpoison_filter_dev_minor;
447 extern u64 hwpoison_filter_flags_mask;
448 extern u64 hwpoison_filter_flags_value;
449 extern u64 hwpoison_filter_memcg;
450 extern u32 hwpoison_filter_enable;
452 extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
453 unsigned long, unsigned long,
454 unsigned long, unsigned long);
456 extern void set_pageblock_order(void);
457 unsigned long reclaim_clean_pages_from_list(struct zone *zone,
458 struct list_head *page_list);
459 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
460 #define ALLOC_WMARK_MIN WMARK_MIN
461 #define ALLOC_WMARK_LOW WMARK_LOW
462 #define ALLOC_WMARK_HIGH WMARK_HIGH
463 #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
465 /* Mask to get the watermark bits */
466 #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
468 #define ALLOC_HARDER 0x10 /* try to alloc harder */
469 #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
470 #define ALLOC_CPUSET 0x40 /* check for correct cpuset */
471 #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
472 #define ALLOC_FAIR 0x100 /* fair zone allocation */
475 struct tlbflush_unmap_batch;
477 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
478 void try_to_unmap_flush(void);
479 void try_to_unmap_flush_dirty(void);
481 static inline void try_to_unmap_flush(void)
484 static inline void try_to_unmap_flush_dirty(void)
488 #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
490 extern const struct trace_print_flags pageflag_names[];
491 extern const struct trace_print_flags vmaflag_names[];
492 extern const struct trace_print_flags gfpflag_names[];
494 #endif /* __MM_INTERNAL_H */