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 void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
39 unsigned long floor, unsigned long ceiling);
41 static inline void set_page_count(struct page *page, int v)
43 atomic_set(&page->_count, v);
46 extern 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);
51 * Submit IO for the read-ahead request in file_ra_state.
53 static inline unsigned long ra_submit(struct file_ra_state *ra,
54 struct address_space *mapping, struct file *filp)
56 return __do_page_cache_readahead(mapping, filp,
57 ra->start, ra->size, ra->async_size);
61 * Turn a non-refcounted page (->_count == 0) into refcounted with
64 static inline void set_page_refcounted(struct page *page)
66 VM_BUG_ON_PAGE(PageTail(page), page);
67 VM_BUG_ON_PAGE(atomic_read(&page->_count), page);
68 set_page_count(page, 1);
71 extern unsigned long highest_memmap_pfn;
76 extern int isolate_lru_page(struct page *page);
77 extern void putback_lru_page(struct page *page);
78 extern bool zone_reclaimable(struct zone *zone);
83 extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
90 * Structure for holding the mostly immutable allocation parameters passed
91 * between functions involved in allocations, including the alloc_pages*
92 * family of functions.
94 * nodemask, migratetype and high_zoneidx are initialized only once in
95 * __alloc_pages_nodemask() and then never change.
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.
102 struct alloc_context {
103 struct zonelist *zonelist;
104 nodemask_t *nodemask;
105 struct zone *preferred_zone;
108 enum zone_type high_zoneidx;
109 bool spread_dirty_pages;
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 long pfn,
138 extern void prep_compound_page(struct page *page, unsigned int order);
139 #ifdef CONFIG_MEMORY_FAILURE
140 extern bool is_free_buddy_page(struct page *page);
142 extern int user_min_free_kbytes;
144 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
150 * compact_control is used to track pages being migrated and the free pages
151 * they are being migrated to during memory compaction. The free_pfn starts
152 * at the end of a zone and migrate_pfn begins at the start. Movable pages
153 * are moved to the end of a zone during a compaction run and the run
154 * completes when free_pfn <= migrate_pfn
156 struct compact_control {
157 struct list_head freepages; /* List of free pages to migrate to */
158 struct list_head migratepages; /* List of pages being migrated */
159 unsigned long nr_freepages; /* Number of isolated free pages */
160 unsigned long nr_migratepages; /* Number of pages to migrate */
161 unsigned long free_pfn; /* isolate_freepages search base */
162 unsigned long migrate_pfn; /* isolate_migratepages search base */
163 unsigned long last_migrated_pfn;/* Not yet flushed page being freed */
164 enum migrate_mode mode; /* Async or sync migration mode */
165 bool ignore_skip_hint; /* Scan blocks even if marked skip */
166 int order; /* order a direct compactor needs */
167 const gfp_t gfp_mask; /* gfp mask of a direct compactor */
168 const int alloc_flags; /* alloc flags of a direct compactor */
169 const int classzone_idx; /* zone index of a direct compactor */
171 int contended; /* Signal need_sched() or lock
172 * contention detected during
178 isolate_freepages_range(struct compact_control *cc,
179 unsigned long start_pfn, unsigned long end_pfn);
181 isolate_migratepages_range(struct compact_control *cc,
182 unsigned long low_pfn, unsigned long end_pfn);
183 int find_suitable_fallback(struct free_area *area, unsigned int order,
184 int migratetype, bool only_stealable, bool *can_steal);
189 * This function returns the order of a free page in the buddy system. In
190 * general, page_zone(page)->lock must be held by the caller to prevent the
191 * page from being allocated in parallel and returning garbage as the order.
192 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
193 * page cannot be allocated or merged in parallel. Alternatively, it must
194 * handle invalid values gracefully, and use page_order_unsafe() below.
196 static inline unsigned int page_order(struct page *page)
198 /* PageBuddy() must be checked by the caller */
199 return page_private(page);
203 * Like page_order(), but for callers who cannot afford to hold the zone lock.
204 * PageBuddy() should be checked first by the caller to minimize race window,
205 * and invalid values must be handled gracefully.
207 * READ_ONCE is used so that if the caller assigns the result into a local
208 * variable and e.g. tests it for valid range before using, the compiler cannot
209 * decide to remove the variable and inline the page_private(page) multiple
210 * times, potentially observing different values in the tests and the actual
213 #define page_order_unsafe(page) READ_ONCE(page_private(page))
215 static inline bool is_cow_mapping(vm_flags_t flags)
217 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
221 * These three helpers classifies VMAs for virtual memory accounting.
225 * Executable code area - executable, not writable, not stack
227 static inline bool is_exec_mapping(vm_flags_t flags)
229 return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC;
233 * Stack area - atomatically grows in one direction
235 * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
236 * do_mmap() forbids all other combinations.
238 static inline bool is_stack_mapping(vm_flags_t flags)
240 return (flags & VM_STACK) == VM_STACK;
244 * Data area - private, writable, not stack
246 static inline bool is_data_mapping(vm_flags_t flags)
248 return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE;
252 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
253 struct vm_area_struct *prev, struct rb_node *rb_parent);
256 extern long populate_vma_page_range(struct vm_area_struct *vma,
257 unsigned long start, unsigned long end, int *nonblocking);
258 extern void munlock_vma_pages_range(struct vm_area_struct *vma,
259 unsigned long start, unsigned long end);
260 static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
262 munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
266 * must be called with vma's mmap_sem held for read or write, and page locked.
268 extern void mlock_vma_page(struct page *page);
269 extern unsigned int munlock_vma_page(struct page *page);
272 * Clear the page's PageMlocked(). This can be useful in a situation where
273 * we want to unconditionally remove a page from the pagecache -- e.g.,
274 * on truncation or freeing.
276 * It is legal to call this function for any page, mlocked or not.
277 * If called for a page that is still mapped by mlocked vmas, all we do
278 * is revert to lazy LRU behaviour -- semantics are not broken.
280 extern void clear_page_mlock(struct page *page);
283 * mlock_migrate_page - called only from migrate_misplaced_transhuge_page()
284 * (because that does not go through the full procedure of migration ptes):
285 * to migrate the Mlocked page flag; update statistics.
287 static inline void mlock_migrate_page(struct page *newpage, struct page *page)
289 if (TestClearPageMlocked(page)) {
290 int nr_pages = hpage_nr_pages(page);
292 /* Holding pmd lock, no change in irq context: __mod is safe */
293 __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
294 SetPageMlocked(newpage);
295 __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
299 extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
302 * At what user virtual address is page expected in @vma?
304 static inline unsigned long
305 __vma_address(struct page *page, struct vm_area_struct *vma)
307 pgoff_t pgoff = page_to_pgoff(page);
308 return vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
311 static inline unsigned long
312 vma_address(struct page *page, struct vm_area_struct *vma)
314 unsigned long address = __vma_address(page, vma);
316 /* page should be within @vma mapping range */
317 VM_BUG_ON_VMA(address < vma->vm_start || address >= vma->vm_end, vma);
322 #else /* !CONFIG_MMU */
323 static inline void clear_page_mlock(struct page *page) { }
324 static inline void mlock_vma_page(struct page *page) { }
325 static inline void mlock_migrate_page(struct page *new, struct page *old) { }
327 #endif /* !CONFIG_MMU */
330 * Return the mem_map entry representing the 'offset' subpage within
331 * the maximally aligned gigantic page 'base'. Handle any discontiguity
332 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
334 static inline struct page *mem_map_offset(struct page *base, int offset)
336 if (unlikely(offset >= MAX_ORDER_NR_PAGES))
337 return nth_page(base, offset);
338 return base + offset;
342 * Iterator over all subpages within the maximally aligned gigantic
343 * page 'base'. Handle any discontiguity in the mem_map.
345 static inline struct page *mem_map_next(struct page *iter,
346 struct page *base, int offset)
348 if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
349 unsigned long pfn = page_to_pfn(base) + offset;
352 return pfn_to_page(pfn);
358 * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
359 * so all functions starting at paging_init should be marked __init
360 * in those cases. SPARSEMEM, however, allows for memory hotplug,
361 * and alloc_bootmem_node is not used.
363 #ifdef CONFIG_SPARSEMEM
364 #define __paginginit __meminit
366 #define __paginginit __init
369 /* Memory initialisation debug and verification */
376 #ifdef CONFIG_DEBUG_MEMORY_INIT
378 extern int mminit_loglevel;
380 #define mminit_dprintk(level, prefix, fmt, arg...) \
382 if (level < mminit_loglevel) { \
383 if (level <= MMINIT_WARNING) \
384 printk(KERN_WARNING "mminit::" prefix " " fmt, ##arg); \
386 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
390 extern void mminit_verify_pageflags_layout(void);
391 extern void mminit_verify_zonelist(void);
394 static inline void mminit_dprintk(enum mminit_level level,
395 const char *prefix, const char *fmt, ...)
399 static inline void mminit_verify_pageflags_layout(void)
403 static inline void mminit_verify_zonelist(void)
406 #endif /* CONFIG_DEBUG_MEMORY_INIT */
408 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
409 #if defined(CONFIG_SPARSEMEM)
410 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
411 unsigned long *end_pfn);
413 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
414 unsigned long *end_pfn)
417 #endif /* CONFIG_SPARSEMEM */
419 #define ZONE_RECLAIM_NOSCAN -2
420 #define ZONE_RECLAIM_FULL -1
421 #define ZONE_RECLAIM_SOME 0
422 #define ZONE_RECLAIM_SUCCESS 1
424 extern int hwpoison_filter(struct page *p);
426 extern u32 hwpoison_filter_dev_major;
427 extern u32 hwpoison_filter_dev_minor;
428 extern u64 hwpoison_filter_flags_mask;
429 extern u64 hwpoison_filter_flags_value;
430 extern u64 hwpoison_filter_memcg;
431 extern u32 hwpoison_filter_enable;
433 extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
434 unsigned long, unsigned long,
435 unsigned long, unsigned long);
437 extern void set_pageblock_order(void);
438 unsigned long reclaim_clean_pages_from_list(struct zone *zone,
439 struct list_head *page_list);
440 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
441 #define ALLOC_WMARK_MIN WMARK_MIN
442 #define ALLOC_WMARK_LOW WMARK_LOW
443 #define ALLOC_WMARK_HIGH WMARK_HIGH
444 #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
446 /* Mask to get the watermark bits */
447 #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
449 #define ALLOC_HARDER 0x10 /* try to alloc harder */
450 #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
451 #define ALLOC_CPUSET 0x40 /* check for correct cpuset */
452 #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
453 #define ALLOC_FAIR 0x100 /* fair zone allocation */
456 struct tlbflush_unmap_batch;
458 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
459 void try_to_unmap_flush(void);
460 void try_to_unmap_flush_dirty(void);
462 static inline void try_to_unmap_flush(void)
465 static inline void try_to_unmap_flush_dirty(void)
469 #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
471 extern const struct trace_print_flags pageflag_names[];
472 extern const struct trace_print_flags vmaflag_names[];
473 extern const struct trace_print_flags gfpflag_names[];
475 #endif /* __MM_INTERNAL_H */