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|\
30 /* The GFP flags allowed during early boot */
31 #define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_RECLAIM|__GFP_IO|__GFP_FS))
33 /* Control allocation cpuset and node placement constraints */
34 #define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE)
36 /* Do not use these with a slab allocator */
37 #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
39 void page_writeback_init(void);
41 int do_swap_page(struct vm_fault *vmf);
43 void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
44 unsigned long floor, unsigned long ceiling);
46 void unmap_page_range(struct mmu_gather *tlb,
47 struct vm_area_struct *vma,
48 unsigned long addr, unsigned long end,
49 struct zap_details *details);
51 extern int __do_page_cache_readahead(struct address_space *mapping,
52 struct file *filp, pgoff_t offset, unsigned long nr_to_read,
53 unsigned long lookahead_size);
56 * Submit IO for the read-ahead request in file_ra_state.
58 static inline unsigned long ra_submit(struct file_ra_state *ra,
59 struct address_space *mapping, struct file *filp)
61 return __do_page_cache_readahead(mapping, filp,
62 ra->start, ra->size, ra->async_size);
66 * Turn a non-refcounted page (->_refcount == 0) into refcounted with
69 static inline void set_page_refcounted(struct page *page)
71 VM_BUG_ON_PAGE(PageTail(page), page);
72 VM_BUG_ON_PAGE(page_ref_count(page), page);
73 set_page_count(page, 1);
76 extern unsigned long highest_memmap_pfn;
81 extern int isolate_lru_page(struct page *page);
82 extern void putback_lru_page(struct page *page);
83 extern bool pgdat_reclaimable(struct pglist_data *pgdat);
88 extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
95 * Structure for holding the mostly immutable allocation parameters passed
96 * between functions involved in allocations, including the alloc_pages*
97 * family of functions.
99 * nodemask, migratetype and high_zoneidx are initialized only once in
100 * __alloc_pages_nodemask() and then never change.
102 * zonelist, preferred_zone and classzone_idx are set first in
103 * __alloc_pages_nodemask() for the fast path, and might be later changed
104 * in __alloc_pages_slowpath(). All other functions pass the whole strucure
105 * by a const pointer.
107 struct alloc_context {
108 struct zonelist *zonelist;
109 nodemask_t *nodemask;
110 struct zoneref *preferred_zoneref;
112 enum zone_type high_zoneidx;
113 bool spread_dirty_pages;
116 #define ac_classzone_idx(ac) zonelist_zone_idx(ac->preferred_zoneref)
119 * Locate the struct page for both the matching buddy in our
120 * pair (buddy1) and the combined O(n+1) page they form (page).
122 * 1) Any buddy B1 will have an order O twin B2 which satisfies
123 * the following equation:
125 * For example, if the starting buddy (buddy2) is #8 its order
127 * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
129 * 2) Any buddy B will have an order O+1 parent P which
130 * satisfies the following equation:
133 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
135 static inline unsigned long
136 __find_buddy_index(unsigned long page_idx, unsigned int order)
138 return page_idx ^ (1 << order);
141 extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn,
142 unsigned long end_pfn, struct zone *zone);
144 static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn,
145 unsigned long end_pfn, struct zone *zone)
147 if (zone->contiguous)
148 return pfn_to_page(start_pfn);
150 return __pageblock_pfn_to_page(start_pfn, end_pfn, zone);
153 extern int __isolate_free_page(struct page *page, unsigned int order);
154 extern void __free_pages_bootmem(struct page *page, unsigned long pfn,
156 extern void prep_compound_page(struct page *page, unsigned int order);
157 extern void post_alloc_hook(struct page *page, unsigned int order,
159 extern int user_min_free_kbytes;
161 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
167 * compact_control is used to track pages being migrated and the free pages
168 * they are being migrated to during memory compaction. The free_pfn starts
169 * at the end of a zone and migrate_pfn begins at the start. Movable pages
170 * are moved to the end of a zone during a compaction run and the run
171 * completes when free_pfn <= migrate_pfn
173 struct compact_control {
174 struct list_head freepages; /* List of free pages to migrate to */
175 struct list_head migratepages; /* List of pages being migrated */
176 unsigned long nr_freepages; /* Number of isolated free pages */
177 unsigned long nr_migratepages; /* Number of pages to migrate */
178 unsigned long free_pfn; /* isolate_freepages search base */
179 unsigned long migrate_pfn; /* isolate_migratepages search base */
180 unsigned long last_migrated_pfn;/* Not yet flushed page being freed */
181 enum migrate_mode mode; /* Async or sync migration mode */
182 bool ignore_skip_hint; /* Scan blocks even if marked skip */
183 bool ignore_block_suitable; /* Scan blocks considered unsuitable */
184 bool direct_compaction; /* False from kcompactd or /proc/... */
185 bool whole_zone; /* Whole zone should/has been scanned */
186 int order; /* order a direct compactor needs */
187 const gfp_t gfp_mask; /* gfp mask of a direct compactor */
188 const unsigned int alloc_flags; /* alloc flags of a direct compactor */
189 const int classzone_idx; /* zone index of a direct compactor */
191 bool contended; /* Signal lock or sched contention */
195 isolate_freepages_range(struct compact_control *cc,
196 unsigned long start_pfn, unsigned long end_pfn);
198 isolate_migratepages_range(struct compact_control *cc,
199 unsigned long low_pfn, unsigned long end_pfn);
200 int find_suitable_fallback(struct free_area *area, unsigned int order,
201 int migratetype, bool only_stealable, bool *can_steal);
206 * This function returns the order of a free page in the buddy system. In
207 * general, page_zone(page)->lock must be held by the caller to prevent the
208 * page from being allocated in parallel and returning garbage as the order.
209 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
210 * page cannot be allocated or merged in parallel. Alternatively, it must
211 * handle invalid values gracefully, and use page_order_unsafe() below.
213 static inline unsigned int page_order(struct page *page)
215 /* PageBuddy() must be checked by the caller */
216 return page_private(page);
220 * Like page_order(), but for callers who cannot afford to hold the zone lock.
221 * PageBuddy() should be checked first by the caller to minimize race window,
222 * and invalid values must be handled gracefully.
224 * READ_ONCE is used so that if the caller assigns the result into a local
225 * variable and e.g. tests it for valid range before using, the compiler cannot
226 * decide to remove the variable and inline the page_private(page) multiple
227 * times, potentially observing different values in the tests and the actual
230 #define page_order_unsafe(page) READ_ONCE(page_private(page))
232 static inline bool is_cow_mapping(vm_flags_t flags)
234 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
238 * These three helpers classifies VMAs for virtual memory accounting.
242 * Executable code area - executable, not writable, not stack
244 static inline bool is_exec_mapping(vm_flags_t flags)
246 return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC;
250 * Stack area - atomatically grows in one direction
252 * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
253 * do_mmap() forbids all other combinations.
255 static inline bool is_stack_mapping(vm_flags_t flags)
257 return (flags & VM_STACK) == VM_STACK;
261 * Data area - private, writable, not stack
263 static inline bool is_data_mapping(vm_flags_t flags)
265 return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE;
269 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
270 struct vm_area_struct *prev, struct rb_node *rb_parent);
273 extern long populate_vma_page_range(struct vm_area_struct *vma,
274 unsigned long start, unsigned long end, int *nonblocking);
275 extern void munlock_vma_pages_range(struct vm_area_struct *vma,
276 unsigned long start, unsigned long end);
277 static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
279 munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
283 * must be called with vma's mmap_sem held for read or write, and page locked.
285 extern void mlock_vma_page(struct page *page);
286 extern unsigned int munlock_vma_page(struct page *page);
289 * Clear the page's PageMlocked(). This can be useful in a situation where
290 * we want to unconditionally remove a page from the pagecache -- e.g.,
291 * on truncation or freeing.
293 * It is legal to call this function for any page, mlocked or not.
294 * If called for a page that is still mapped by mlocked vmas, all we do
295 * is revert to lazy LRU behaviour -- semantics are not broken.
297 extern void clear_page_mlock(struct page *page);
300 * mlock_migrate_page - called only from migrate_misplaced_transhuge_page()
301 * (because that does not go through the full procedure of migration ptes):
302 * to migrate the Mlocked page flag; update statistics.
304 static inline void mlock_migrate_page(struct page *newpage, struct page *page)
306 if (TestClearPageMlocked(page)) {
307 int nr_pages = hpage_nr_pages(page);
309 /* Holding pmd lock, no change in irq context: __mod is safe */
310 __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
311 SetPageMlocked(newpage);
312 __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
316 extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
319 * At what user virtual address is page expected in @vma?
321 static inline unsigned long
322 __vma_address(struct page *page, struct vm_area_struct *vma)
324 pgoff_t pgoff = page_to_pgoff(page);
325 return vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
328 static inline unsigned long
329 vma_address(struct page *page, struct vm_area_struct *vma)
331 unsigned long address = __vma_address(page, vma);
333 /* page should be within @vma mapping range */
334 VM_BUG_ON_VMA(address < vma->vm_start || address >= vma->vm_end, vma);
339 #else /* !CONFIG_MMU */
340 static inline void clear_page_mlock(struct page *page) { }
341 static inline void mlock_vma_page(struct page *page) { }
342 static inline void mlock_migrate_page(struct page *new, struct page *old) { }
344 #endif /* !CONFIG_MMU */
347 * Return the mem_map entry representing the 'offset' subpage within
348 * the maximally aligned gigantic page 'base'. Handle any discontiguity
349 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
351 static inline struct page *mem_map_offset(struct page *base, int offset)
353 if (unlikely(offset >= MAX_ORDER_NR_PAGES))
354 return nth_page(base, offset);
355 return base + offset;
359 * Iterator over all subpages within the maximally aligned gigantic
360 * page 'base'. Handle any discontiguity in the mem_map.
362 static inline struct page *mem_map_next(struct page *iter,
363 struct page *base, int offset)
365 if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
366 unsigned long pfn = page_to_pfn(base) + offset;
369 return pfn_to_page(pfn);
375 * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
376 * so all functions starting at paging_init should be marked __init
377 * in those cases. SPARSEMEM, however, allows for memory hotplug,
378 * and alloc_bootmem_node is not used.
380 #ifdef CONFIG_SPARSEMEM
381 #define __paginginit __meminit
383 #define __paginginit __init
386 /* Memory initialisation debug and verification */
393 #ifdef CONFIG_DEBUG_MEMORY_INIT
395 extern int mminit_loglevel;
397 #define mminit_dprintk(level, prefix, fmt, arg...) \
399 if (level < mminit_loglevel) { \
400 if (level <= MMINIT_WARNING) \
401 pr_warn("mminit::" prefix " " fmt, ##arg); \
403 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
407 extern void mminit_verify_pageflags_layout(void);
408 extern void mminit_verify_zonelist(void);
411 static inline void mminit_dprintk(enum mminit_level level,
412 const char *prefix, const char *fmt, ...)
416 static inline void mminit_verify_pageflags_layout(void)
420 static inline void mminit_verify_zonelist(void)
423 #endif /* CONFIG_DEBUG_MEMORY_INIT */
425 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
426 #if defined(CONFIG_SPARSEMEM)
427 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
428 unsigned long *end_pfn);
430 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
431 unsigned long *end_pfn)
434 #endif /* CONFIG_SPARSEMEM */
436 #define NODE_RECLAIM_NOSCAN -2
437 #define NODE_RECLAIM_FULL -1
438 #define NODE_RECLAIM_SOME 0
439 #define NODE_RECLAIM_SUCCESS 1
441 extern int hwpoison_filter(struct page *p);
443 extern u32 hwpoison_filter_dev_major;
444 extern u32 hwpoison_filter_dev_minor;
445 extern u64 hwpoison_filter_flags_mask;
446 extern u64 hwpoison_filter_flags_value;
447 extern u64 hwpoison_filter_memcg;
448 extern u32 hwpoison_filter_enable;
450 extern unsigned long __must_check vm_mmap_pgoff(struct file *, unsigned long,
451 unsigned long, unsigned long,
452 unsigned long, unsigned long);
454 extern void set_pageblock_order(void);
455 unsigned long reclaim_clean_pages_from_list(struct zone *zone,
456 struct list_head *page_list);
457 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
458 #define ALLOC_WMARK_MIN WMARK_MIN
459 #define ALLOC_WMARK_LOW WMARK_LOW
460 #define ALLOC_WMARK_HIGH WMARK_HIGH
461 #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
463 /* Mask to get the watermark bits */
464 #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
466 #define ALLOC_HARDER 0x10 /* try to alloc harder */
467 #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
468 #define ALLOC_CPUSET 0x40 /* check for correct cpuset */
469 #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
472 struct tlbflush_unmap_batch;
474 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
475 void try_to_unmap_flush(void);
476 void try_to_unmap_flush_dirty(void);
478 static inline void try_to_unmap_flush(void)
481 static inline void try_to_unmap_flush_dirty(void)
485 #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
487 extern const struct trace_print_flags pageflag_names[];
488 extern const struct trace_print_flags vmaflag_names[];
489 extern const struct trace_print_flags gfpflag_names[];
491 #endif /* __MM_INTERNAL_H */