Linux Audio

Check our new training course

Loading...
v5.4
  1/* SPDX-License-Identifier: GPL-2.0-or-later */
  2/* internal.h: mm/ internal definitions
  3 *
  4 * Copyright (C) 2004 Red Hat, Inc. All Rights Reserved.
  5 * Written by David Howells (dhowells@redhat.com)
  6 */
  7#ifndef __MM_INTERNAL_H
  8#define __MM_INTERNAL_H
  9
 10#include <linux/fs.h>
 11#include <linux/mm.h>
 12#include <linux/pagemap.h>
 13#include <linux/tracepoint-defs.h>
 14
 15/*
 16 * The set of flags that only affect watermark checking and reclaim
 17 * behaviour. This is used by the MM to obey the caller constraints
 18 * about IO, FS and watermark checking while ignoring placement
 19 * hints such as HIGHMEM usage.
 20 */
 21#define GFP_RECLAIM_MASK (__GFP_RECLAIM|__GFP_HIGH|__GFP_IO|__GFP_FS|\
 22			__GFP_NOWARN|__GFP_RETRY_MAYFAIL|__GFP_NOFAIL|\
 23			__GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC|\
 24			__GFP_ATOMIC)
 25
 26/* The GFP flags allowed during early boot */
 27#define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_RECLAIM|__GFP_IO|__GFP_FS))
 28
 29/* Control allocation cpuset and node placement constraints */
 30#define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE)
 31
 32/* Do not use these with a slab allocator */
 33#define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
 34
 35void page_writeback_init(void);
 36
 37vm_fault_t do_swap_page(struct vm_fault *vmf);
 38
 39void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
 40		unsigned long floor, unsigned long ceiling);
 41
 42static inline bool can_madv_lru_vma(struct vm_area_struct *vma)
 43{
 44	return !(vma->vm_flags & (VM_LOCKED|VM_HUGETLB|VM_PFNMAP));
 45}
 46
 47void unmap_page_range(struct mmu_gather *tlb,
 48			     struct vm_area_struct *vma,
 49			     unsigned long addr, unsigned long end,
 50			     struct zap_details *details);
 51
 52extern unsigned int __do_page_cache_readahead(struct address_space *mapping,
 53		struct file *filp, pgoff_t offset, unsigned long nr_to_read,
 54		unsigned long lookahead_size);
 
 
 
 
 
 
 
 55
 56/*
 57 * Submit IO for the read-ahead request in file_ra_state.
 
 
 
 
 
 
 
 
 
 
 
 
 58 */
 59static inline unsigned long ra_submit(struct file_ra_state *ra,
 60		struct address_space *mapping, struct file *filp)
 61{
 62	return __do_page_cache_readahead(mapping, filp,
 63					ra->start, ra->size, ra->async_size);
 
 
 
 
 
 64}
 65
 66/*
 67 * Turn a non-refcounted page (->_refcount == 0) into refcounted with
 68 * a count of one.
 69 */
 70static inline void set_page_refcounted(struct page *page)
 71{
 72	VM_BUG_ON_PAGE(PageTail(page), page);
 73	VM_BUG_ON_PAGE(page_ref_count(page), page);
 74	set_page_count(page, 1);
 75}
 76
 77extern unsigned long highest_memmap_pfn;
 78
 79/*
 80 * Maximum number of reclaim retries without progress before the OOM
 81 * killer is consider the only way forward.
 82 */
 83#define MAX_RECLAIM_RETRIES 16
 84
 85/*
 86 * in mm/vmscan.c:
 87 */
 88extern int isolate_lru_page(struct page *page);
 89extern void putback_lru_page(struct page *page);
 90
 91/*
 92 * in mm/rmap.c:
 93 */
 94extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
 95
 96/*
 
 
 
 
 
 97 * in mm/page_alloc.c
 98 */
 99
100/*
101 * Structure for holding the mostly immutable allocation parameters passed
102 * between functions involved in allocations, including the alloc_pages*
103 * family of functions.
104 *
105 * nodemask, migratetype and high_zoneidx are initialized only once in
106 * __alloc_pages_nodemask() and then never change.
107 *
108 * zonelist, preferred_zone and classzone_idx are set first in
109 * __alloc_pages_nodemask() for the fast path, and might be later changed
110 * in __alloc_pages_slowpath(). All other functions pass the whole strucure
111 * by a const pointer.
112 */
113struct alloc_context {
114	struct zonelist *zonelist;
115	nodemask_t *nodemask;
116	struct zoneref *preferred_zoneref;
117	int migratetype;
118	enum zone_type high_zoneidx;
 
 
 
 
 
 
 
 
 
 
 
119	bool spread_dirty_pages;
120};
121
122#define ac_classzone_idx(ac) zonelist_zone_idx(ac->preferred_zoneref)
123
124/*
125 * Locate the struct page for both the matching buddy in our
126 * pair (buddy1) and the combined O(n+1) page they form (page).
127 *
128 * 1) Any buddy B1 will have an order O twin B2 which satisfies
129 * the following equation:
130 *     B2 = B1 ^ (1 << O)
131 * For example, if the starting buddy (buddy2) is #8 its order
132 * 1 buddy is #10:
133 *     B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
134 *
135 * 2) Any buddy B will have an order O+1 parent P which
136 * satisfies the following equation:
137 *     P = B & ~(1 << O)
138 *
139 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
140 */
141static inline unsigned long
142__find_buddy_pfn(unsigned long page_pfn, unsigned int order)
143{
144	return page_pfn ^ (1 << order);
145}
146
147extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn,
148				unsigned long end_pfn, struct zone *zone);
149
150static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn,
151				unsigned long end_pfn, struct zone *zone)
152{
153	if (zone->contiguous)
154		return pfn_to_page(start_pfn);
155
156	return __pageblock_pfn_to_page(start_pfn, end_pfn, zone);
157}
158
159extern int __isolate_free_page(struct page *page, unsigned int order);
 
 
160extern void memblock_free_pages(struct page *page, unsigned long pfn,
161					unsigned int order);
162extern void __free_pages_core(struct page *page, unsigned int order);
163extern void prep_compound_page(struct page *page, unsigned int order);
164extern void post_alloc_hook(struct page *page, unsigned int order,
165					gfp_t gfp_flags);
166extern int user_min_free_kbytes;
167
 
 
 
 
 
 
 
 
168#if defined CONFIG_COMPACTION || defined CONFIG_CMA
169
170/*
171 * in mm/compaction.c
172 */
173/*
174 * compact_control is used to track pages being migrated and the free pages
175 * they are being migrated to during memory compaction. The free_pfn starts
176 * at the end of a zone and migrate_pfn begins at the start. Movable pages
177 * are moved to the end of a zone during a compaction run and the run
178 * completes when free_pfn <= migrate_pfn
179 */
180struct compact_control {
181	struct list_head freepages;	/* List of free pages to migrate to */
182	struct list_head migratepages;	/* List of pages being migrated */
183	unsigned int nr_freepages;	/* Number of isolated free pages */
184	unsigned int nr_migratepages;	/* Number of pages to migrate */
185	unsigned long free_pfn;		/* isolate_freepages search base */
186	unsigned long migrate_pfn;	/* isolate_migratepages search base */
 
 
 
 
 
 
187	unsigned long fast_start_pfn;	/* a pfn to start linear scan from */
188	struct zone *zone;
189	unsigned long total_migrate_scanned;
190	unsigned long total_free_scanned;
191	unsigned short fast_search_fail;/* failures to use free list searches */
192	short search_order;		/* order to start a fast search at */
193	const gfp_t gfp_mask;		/* gfp mask of a direct compactor */
194	int order;			/* order a direct compactor needs */
195	int migratetype;		/* migratetype of direct compactor */
196	const unsigned int alloc_flags;	/* alloc flags of a direct compactor */
197	const int classzone_idx;	/* zone index of a direct compactor */
198	enum migrate_mode mode;		/* Async or sync migration mode */
199	bool ignore_skip_hint;		/* Scan blocks even if marked skip */
200	bool no_set_skip_hint;		/* Don't mark blocks for skipping */
201	bool ignore_block_suitable;	/* Scan blocks considered unsuitable */
202	bool direct_compaction;		/* False from kcompactd or /proc/... */
 
203	bool whole_zone;		/* Whole zone should/has been scanned */
204	bool contended;			/* Signal lock or sched contention */
205	bool rescan;			/* Rescanning the same pageblock */
 
206};
207
208/*
209 * Used in direct compaction when a page should be taken from the freelists
210 * immediately when one is created during the free path.
211 */
212struct capture_control {
213	struct compact_control *cc;
214	struct page *page;
215};
216
217unsigned long
218isolate_freepages_range(struct compact_control *cc,
219			unsigned long start_pfn, unsigned long end_pfn);
220unsigned long
221isolate_migratepages_range(struct compact_control *cc,
222			   unsigned long low_pfn, unsigned long end_pfn);
 
223int find_suitable_fallback(struct free_area *area, unsigned int order,
224			int migratetype, bool only_stealable, bool *can_steal);
225
226#endif
227
228/*
229 * This function returns the order of a free page in the buddy system. In
230 * general, page_zone(page)->lock must be held by the caller to prevent the
231 * page from being allocated in parallel and returning garbage as the order.
232 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
233 * page cannot be allocated or merged in parallel. Alternatively, it must
234 * handle invalid values gracefully, and use page_order_unsafe() below.
235 */
236static inline unsigned int page_order(struct page *page)
237{
238	/* PageBuddy() must be checked by the caller */
239	return page_private(page);
240}
241
242/*
243 * Like page_order(), but for callers who cannot afford to hold the zone lock.
244 * PageBuddy() should be checked first by the caller to minimize race window,
245 * and invalid values must be handled gracefully.
246 *
247 * READ_ONCE is used so that if the caller assigns the result into a local
248 * variable and e.g. tests it for valid range before using, the compiler cannot
249 * decide to remove the variable and inline the page_private(page) multiple
250 * times, potentially observing different values in the tests and the actual
251 * use of the result.
252 */
253#define page_order_unsafe(page)		READ_ONCE(page_private(page))
254
255static inline bool is_cow_mapping(vm_flags_t flags)
256{
257	return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
258}
259
260/*
261 * These three helpers classifies VMAs for virtual memory accounting.
262 */
263
264/*
265 * Executable code area - executable, not writable, not stack
266 */
267static inline bool is_exec_mapping(vm_flags_t flags)
268{
269	return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC;
270}
271
272/*
273 * Stack area - atomatically grows in one direction
274 *
275 * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
276 * do_mmap() forbids all other combinations.
277 */
278static inline bool is_stack_mapping(vm_flags_t flags)
279{
280	return (flags & VM_STACK) == VM_STACK;
281}
282
283/*
284 * Data area - private, writable, not stack
285 */
286static inline bool is_data_mapping(vm_flags_t flags)
287{
288	return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE;
289}
290
291/* mm/util.c */
292void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
293		struct vm_area_struct *prev, struct rb_node *rb_parent);
 
294
295#ifdef CONFIG_MMU
296extern long populate_vma_page_range(struct vm_area_struct *vma,
297		unsigned long start, unsigned long end, int *nonblocking);
 
 
 
298extern void munlock_vma_pages_range(struct vm_area_struct *vma,
299			unsigned long start, unsigned long end);
300static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
301{
302	munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
303}
304
305/*
306 * must be called with vma's mmap_sem held for read or write, and page locked.
307 */
308extern void mlock_vma_page(struct page *page);
309extern unsigned int munlock_vma_page(struct page *page);
310
 
 
 
311/*
312 * Clear the page's PageMlocked().  This can be useful in a situation where
313 * we want to unconditionally remove a page from the pagecache -- e.g.,
314 * on truncation or freeing.
315 *
316 * It is legal to call this function for any page, mlocked or not.
317 * If called for a page that is still mapped by mlocked vmas, all we do
318 * is revert to lazy LRU behaviour -- semantics are not broken.
319 */
320extern void clear_page_mlock(struct page *page);
321
322/*
323 * mlock_migrate_page - called only from migrate_misplaced_transhuge_page()
324 * (because that does not go through the full procedure of migration ptes):
325 * to migrate the Mlocked page flag; update statistics.
326 */
327static inline void mlock_migrate_page(struct page *newpage, struct page *page)
328{
329	if (TestClearPageMlocked(page)) {
330		int nr_pages = hpage_nr_pages(page);
331
332		/* Holding pmd lock, no change in irq context: __mod is safe */
333		__mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
334		SetPageMlocked(newpage);
335		__mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
336	}
337}
338
339extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
340
341/*
342 * At what user virtual address is page expected in @vma?
 
 
343 */
344static inline unsigned long
345__vma_address(struct page *page, struct vm_area_struct *vma)
346{
347	pgoff_t pgoff = page_to_pgoff(page);
348	return vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
349}
350
 
 
 
 
 
351static inline unsigned long
352vma_address(struct page *page, struct vm_area_struct *vma)
353{
354	unsigned long start, end;
 
355
356	start = __vma_address(page, vma);
357	end = start + PAGE_SIZE * (hpage_nr_pages(page) - 1);
358
359	/* page should be within @vma mapping range */
360	VM_BUG_ON_VMA(end < vma->vm_start || start >= vma->vm_end, vma);
361
362	return max(start, vma->vm_start);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
363}
364
365#else /* !CONFIG_MMU */
366static inline void clear_page_mlock(struct page *page) { }
367static inline void mlock_vma_page(struct page *page) { }
368static inline void mlock_migrate_page(struct page *new, struct page *old) { }
369
 
370#endif /* !CONFIG_MMU */
371
372/*
373 * Return the mem_map entry representing the 'offset' subpage within
374 * the maximally aligned gigantic page 'base'.  Handle any discontiguity
375 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
376 */
377static inline struct page *mem_map_offset(struct page *base, int offset)
378{
379	if (unlikely(offset >= MAX_ORDER_NR_PAGES))
380		return nth_page(base, offset);
381	return base + offset;
382}
383
384/*
385 * Iterator over all subpages within the maximally aligned gigantic
386 * page 'base'.  Handle any discontiguity in the mem_map.
387 */
388static inline struct page *mem_map_next(struct page *iter,
389						struct page *base, int offset)
390{
391	if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
392		unsigned long pfn = page_to_pfn(base) + offset;
393		if (!pfn_valid(pfn))
394			return NULL;
395		return pfn_to_page(pfn);
396	}
397	return iter + 1;
398}
399
400/* Memory initialisation debug and verification */
401enum mminit_level {
402	MMINIT_WARNING,
403	MMINIT_VERIFY,
404	MMINIT_TRACE
405};
406
407#ifdef CONFIG_DEBUG_MEMORY_INIT
408
409extern int mminit_loglevel;
410
411#define mminit_dprintk(level, prefix, fmt, arg...) \
412do { \
413	if (level < mminit_loglevel) { \
414		if (level <= MMINIT_WARNING) \
415			pr_warn("mminit::" prefix " " fmt, ##arg);	\
416		else \
417			printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
418	} \
419} while (0)
420
421extern void mminit_verify_pageflags_layout(void);
422extern void mminit_verify_zonelist(void);
423#else
424
425static inline void mminit_dprintk(enum mminit_level level,
426				const char *prefix, const char *fmt, ...)
427{
428}
429
430static inline void mminit_verify_pageflags_layout(void)
431{
432}
433
434static inline void mminit_verify_zonelist(void)
435{
436}
437#endif /* CONFIG_DEBUG_MEMORY_INIT */
438
439/* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
440#if defined(CONFIG_SPARSEMEM)
441extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
442				unsigned long *end_pfn);
443#else
444static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
445				unsigned long *end_pfn)
446{
447}
448#endif /* CONFIG_SPARSEMEM */
449
450#define NODE_RECLAIM_NOSCAN	-2
451#define NODE_RECLAIM_FULL	-1
452#define NODE_RECLAIM_SOME	0
453#define NODE_RECLAIM_SUCCESS	1
454
455#ifdef CONFIG_NUMA
456extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int);
457#else
458static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask,
459				unsigned int order)
460{
461	return NODE_RECLAIM_NOSCAN;
462}
463#endif
464
465extern int hwpoison_filter(struct page *p);
466
467extern u32 hwpoison_filter_dev_major;
468extern u32 hwpoison_filter_dev_minor;
469extern u64 hwpoison_filter_flags_mask;
470extern u64 hwpoison_filter_flags_value;
471extern u64 hwpoison_filter_memcg;
472extern u32 hwpoison_filter_enable;
473
474extern unsigned long  __must_check vm_mmap_pgoff(struct file *, unsigned long,
475        unsigned long, unsigned long,
476        unsigned long, unsigned long);
477
478extern void set_pageblock_order(void);
479unsigned long reclaim_clean_pages_from_list(struct zone *zone,
480					    struct list_head *page_list);
481/* The ALLOC_WMARK bits are used as an index to zone->watermark */
482#define ALLOC_WMARK_MIN		WMARK_MIN
483#define ALLOC_WMARK_LOW		WMARK_LOW
484#define ALLOC_WMARK_HIGH	WMARK_HIGH
485#define ALLOC_NO_WATERMARKS	0x04 /* don't check watermarks at all */
486
487/* Mask to get the watermark bits */
488#define ALLOC_WMARK_MASK	(ALLOC_NO_WATERMARKS-1)
489
490/*
491 * Only MMU archs have async oom victim reclaim - aka oom_reaper so we
492 * cannot assume a reduced access to memory reserves is sufficient for
493 * !MMU
494 */
495#ifdef CONFIG_MMU
496#define ALLOC_OOM		0x08
497#else
498#define ALLOC_OOM		ALLOC_NO_WATERMARKS
499#endif
500
501#define ALLOC_HARDER		 0x10 /* try to alloc harder */
502#define ALLOC_HIGH		 0x20 /* __GFP_HIGH set */
503#define ALLOC_CPUSET		 0x40 /* check for correct cpuset */
504#define ALLOC_CMA		 0x80 /* allow allocations from CMA areas */
505#ifdef CONFIG_ZONE_DMA32
506#define ALLOC_NOFRAGMENT	0x100 /* avoid mixing pageblock types */
507#else
508#define ALLOC_NOFRAGMENT	  0x0
509#endif
510#define ALLOC_KSWAPD		0x200 /* allow waking of kswapd */
511
512enum ttu_flags;
513struct tlbflush_unmap_batch;
514
515
516/*
517 * only for MM internal work items which do not depend on
518 * any allocations or locks which might depend on allocations
519 */
520extern struct workqueue_struct *mm_percpu_wq;
521
522#ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
523void try_to_unmap_flush(void);
524void try_to_unmap_flush_dirty(void);
525void flush_tlb_batched_pending(struct mm_struct *mm);
526#else
527static inline void try_to_unmap_flush(void)
528{
529}
530static inline void try_to_unmap_flush_dirty(void)
531{
532}
533static inline void flush_tlb_batched_pending(struct mm_struct *mm)
534{
535}
536#endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
537
538extern const struct trace_print_flags pageflag_names[];
539extern const struct trace_print_flags vmaflag_names[];
540extern const struct trace_print_flags gfpflag_names[];
541
542static inline bool is_migrate_highatomic(enum migratetype migratetype)
543{
544	return migratetype == MIGRATE_HIGHATOMIC;
545}
546
547static inline bool is_migrate_highatomic_page(struct page *page)
548{
549	return get_pageblock_migratetype(page) == MIGRATE_HIGHATOMIC;
550}
551
552void setup_zone_pageset(struct zone *zone);
553extern struct page *alloc_new_node_page(struct page *page, unsigned long node);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
554#endif	/* __MM_INTERNAL_H */
v5.14.15
  1/* SPDX-License-Identifier: GPL-2.0-or-later */
  2/* internal.h: mm/ internal definitions
  3 *
  4 * Copyright (C) 2004 Red Hat, Inc. All Rights Reserved.
  5 * Written by David Howells (dhowells@redhat.com)
  6 */
  7#ifndef __MM_INTERNAL_H
  8#define __MM_INTERNAL_H
  9
 10#include <linux/fs.h>
 11#include <linux/mm.h>
 12#include <linux/pagemap.h>
 13#include <linux/tracepoint-defs.h>
 14
 15/*
 16 * The set of flags that only affect watermark checking and reclaim
 17 * behaviour. This is used by the MM to obey the caller constraints
 18 * about IO, FS and watermark checking while ignoring placement
 19 * hints such as HIGHMEM usage.
 20 */
 21#define GFP_RECLAIM_MASK (__GFP_RECLAIM|__GFP_HIGH|__GFP_IO|__GFP_FS|\
 22			__GFP_NOWARN|__GFP_RETRY_MAYFAIL|__GFP_NOFAIL|\
 23			__GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC|\
 24			__GFP_ATOMIC)
 25
 26/* The GFP flags allowed during early boot */
 27#define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_RECLAIM|__GFP_IO|__GFP_FS))
 28
 29/* Control allocation cpuset and node placement constraints */
 30#define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE)
 31
 32/* Do not use these with a slab allocator */
 33#define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
 34
 35void page_writeback_init(void);
 36
 37vm_fault_t do_swap_page(struct vm_fault *vmf);
 38
 39void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
 40		unsigned long floor, unsigned long ceiling);
 41
 42static inline bool can_madv_lru_vma(struct vm_area_struct *vma)
 43{
 44	return !(vma->vm_flags & (VM_LOCKED|VM_HUGETLB|VM_PFNMAP));
 45}
 46
 47void unmap_page_range(struct mmu_gather *tlb,
 48			     struct vm_area_struct *vma,
 49			     unsigned long addr, unsigned long end,
 50			     struct zap_details *details);
 51
 52void do_page_cache_ra(struct readahead_control *, unsigned long nr_to_read,
 
 53		unsigned long lookahead_size);
 54void force_page_cache_ra(struct readahead_control *, unsigned long nr);
 55static inline void force_page_cache_readahead(struct address_space *mapping,
 56		struct file *file, pgoff_t index, unsigned long nr_to_read)
 57{
 58	DEFINE_READAHEAD(ractl, file, &file->f_ra, mapping, index);
 59	force_page_cache_ra(&ractl, nr_to_read);
 60}
 61
 62unsigned find_lock_entries(struct address_space *mapping, pgoff_t start,
 63		pgoff_t end, struct pagevec *pvec, pgoff_t *indices);
 64
 65/**
 66 * page_evictable - test whether a page is evictable
 67 * @page: the page to test
 68 *
 69 * Test whether page is evictable--i.e., should be placed on active/inactive
 70 * lists vs unevictable list.
 71 *
 72 * Reasons page might not be evictable:
 73 * (1) page's mapping marked unevictable
 74 * (2) page is part of an mlocked VMA
 75 *
 76 */
 77static inline bool page_evictable(struct page *page)
 
 78{
 79	bool ret;
 80
 81	/* Prevent address_space of inode and swap cache from being freed */
 82	rcu_read_lock();
 83	ret = !mapping_unevictable(page_mapping(page)) && !PageMlocked(page);
 84	rcu_read_unlock();
 85	return ret;
 86}
 87
 88/*
 89 * Turn a non-refcounted page (->_refcount == 0) into refcounted with
 90 * a count of one.
 91 */
 92static inline void set_page_refcounted(struct page *page)
 93{
 94	VM_BUG_ON_PAGE(PageTail(page), page);
 95	VM_BUG_ON_PAGE(page_ref_count(page), page);
 96	set_page_count(page, 1);
 97}
 98
 99extern unsigned long highest_memmap_pfn;
100
101/*
102 * Maximum number of reclaim retries without progress before the OOM
103 * killer is consider the only way forward.
104 */
105#define MAX_RECLAIM_RETRIES 16
106
107/*
108 * in mm/vmscan.c:
109 */
110extern int isolate_lru_page(struct page *page);
111extern void putback_lru_page(struct page *page);
112
113/*
114 * in mm/rmap.c:
115 */
116extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
117
118/*
119 * in mm/memcontrol.c:
120 */
121extern bool cgroup_memory_nokmem;
122
123/*
124 * in mm/page_alloc.c
125 */
126
127/*
128 * Structure for holding the mostly immutable allocation parameters passed
129 * between functions involved in allocations, including the alloc_pages*
130 * family of functions.
131 *
132 * nodemask, migratetype and highest_zoneidx are initialized only once in
133 * __alloc_pages() and then never change.
134 *
135 * zonelist, preferred_zone and highest_zoneidx are set first in
136 * __alloc_pages() for the fast path, and might be later changed
137 * in __alloc_pages_slowpath(). All other functions pass the whole structure
138 * by a const pointer.
139 */
140struct alloc_context {
141	struct zonelist *zonelist;
142	nodemask_t *nodemask;
143	struct zoneref *preferred_zoneref;
144	int migratetype;
145
146	/*
147	 * highest_zoneidx represents highest usable zone index of
148	 * the allocation request. Due to the nature of the zone,
149	 * memory on lower zone than the highest_zoneidx will be
150	 * protected by lowmem_reserve[highest_zoneidx].
151	 *
152	 * highest_zoneidx is also used by reclaim/compaction to limit
153	 * the target zone since higher zone than this index cannot be
154	 * usable for this allocation request.
155	 */
156	enum zone_type highest_zoneidx;
157	bool spread_dirty_pages;
158};
159
 
 
160/*
161 * Locate the struct page for both the matching buddy in our
162 * pair (buddy1) and the combined O(n+1) page they form (page).
163 *
164 * 1) Any buddy B1 will have an order O twin B2 which satisfies
165 * the following equation:
166 *     B2 = B1 ^ (1 << O)
167 * For example, if the starting buddy (buddy2) is #8 its order
168 * 1 buddy is #10:
169 *     B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
170 *
171 * 2) Any buddy B will have an order O+1 parent P which
172 * satisfies the following equation:
173 *     P = B & ~(1 << O)
174 *
175 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
176 */
177static inline unsigned long
178__find_buddy_pfn(unsigned long page_pfn, unsigned int order)
179{
180	return page_pfn ^ (1 << order);
181}
182
183extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn,
184				unsigned long end_pfn, struct zone *zone);
185
186static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn,
187				unsigned long end_pfn, struct zone *zone)
188{
189	if (zone->contiguous)
190		return pfn_to_page(start_pfn);
191
192	return __pageblock_pfn_to_page(start_pfn, end_pfn, zone);
193}
194
195extern int __isolate_free_page(struct page *page, unsigned int order);
196extern void __putback_isolated_page(struct page *page, unsigned int order,
197				    int mt);
198extern void memblock_free_pages(struct page *page, unsigned long pfn,
199					unsigned int order);
200extern void __free_pages_core(struct page *page, unsigned int order);
201extern void prep_compound_page(struct page *page, unsigned int order);
202extern void post_alloc_hook(struct page *page, unsigned int order,
203					gfp_t gfp_flags);
204extern int user_min_free_kbytes;
205
206extern void free_unref_page(struct page *page, unsigned int order);
207extern void free_unref_page_list(struct list_head *list);
208
209extern void zone_pcp_update(struct zone *zone, int cpu_online);
210extern void zone_pcp_reset(struct zone *zone);
211extern void zone_pcp_disable(struct zone *zone);
212extern void zone_pcp_enable(struct zone *zone);
213
214#if defined CONFIG_COMPACTION || defined CONFIG_CMA
215
216/*
217 * in mm/compaction.c
218 */
219/*
220 * compact_control is used to track pages being migrated and the free pages
221 * they are being migrated to during memory compaction. The free_pfn starts
222 * at the end of a zone and migrate_pfn begins at the start. Movable pages
223 * are moved to the end of a zone during a compaction run and the run
224 * completes when free_pfn <= migrate_pfn
225 */
226struct compact_control {
227	struct list_head freepages;	/* List of free pages to migrate to */
228	struct list_head migratepages;	/* List of pages being migrated */
229	unsigned int nr_freepages;	/* Number of isolated free pages */
230	unsigned int nr_migratepages;	/* Number of pages to migrate */
231	unsigned long free_pfn;		/* isolate_freepages search base */
232	/*
233	 * Acts as an in/out parameter to page isolation for migration.
234	 * isolate_migratepages uses it as a search base.
235	 * isolate_migratepages_block will update the value to the next pfn
236	 * after the last isolated one.
237	 */
238	unsigned long migrate_pfn;
239	unsigned long fast_start_pfn;	/* a pfn to start linear scan from */
240	struct zone *zone;
241	unsigned long total_migrate_scanned;
242	unsigned long total_free_scanned;
243	unsigned short fast_search_fail;/* failures to use free list searches */
244	short search_order;		/* order to start a fast search at */
245	const gfp_t gfp_mask;		/* gfp mask of a direct compactor */
246	int order;			/* order a direct compactor needs */
247	int migratetype;		/* migratetype of direct compactor */
248	const unsigned int alloc_flags;	/* alloc flags of a direct compactor */
249	const int highest_zoneidx;	/* zone index of a direct compactor */
250	enum migrate_mode mode;		/* Async or sync migration mode */
251	bool ignore_skip_hint;		/* Scan blocks even if marked skip */
252	bool no_set_skip_hint;		/* Don't mark blocks for skipping */
253	bool ignore_block_suitable;	/* Scan blocks considered unsuitable */
254	bool direct_compaction;		/* False from kcompactd or /proc/... */
255	bool proactive_compaction;	/* kcompactd proactive compaction */
256	bool whole_zone;		/* Whole zone should/has been scanned */
257	bool contended;			/* Signal lock or sched contention */
258	bool rescan;			/* Rescanning the same pageblock */
259	bool alloc_contig;		/* alloc_contig_range allocation */
260};
261
262/*
263 * Used in direct compaction when a page should be taken from the freelists
264 * immediately when one is created during the free path.
265 */
266struct capture_control {
267	struct compact_control *cc;
268	struct page *page;
269};
270
271unsigned long
272isolate_freepages_range(struct compact_control *cc,
273			unsigned long start_pfn, unsigned long end_pfn);
274int
275isolate_migratepages_range(struct compact_control *cc,
276			   unsigned long low_pfn, unsigned long end_pfn);
277#endif
278int find_suitable_fallback(struct free_area *area, unsigned int order,
279			int migratetype, bool only_stealable, bool *can_steal);
280
 
 
281/*
282 * This function returns the order of a free page in the buddy system. In
283 * general, page_zone(page)->lock must be held by the caller to prevent the
284 * page from being allocated in parallel and returning garbage as the order.
285 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
286 * page cannot be allocated or merged in parallel. Alternatively, it must
287 * handle invalid values gracefully, and use buddy_order_unsafe() below.
288 */
289static inline unsigned int buddy_order(struct page *page)
290{
291	/* PageBuddy() must be checked by the caller */
292	return page_private(page);
293}
294
295/*
296 * Like buddy_order(), but for callers who cannot afford to hold the zone lock.
297 * PageBuddy() should be checked first by the caller to minimize race window,
298 * and invalid values must be handled gracefully.
299 *
300 * READ_ONCE is used so that if the caller assigns the result into a local
301 * variable and e.g. tests it for valid range before using, the compiler cannot
302 * decide to remove the variable and inline the page_private(page) multiple
303 * times, potentially observing different values in the tests and the actual
304 * use of the result.
305 */
306#define buddy_order_unsafe(page)	READ_ONCE(page_private(page))
 
 
 
 
 
307
308/*
309 * These three helpers classifies VMAs for virtual memory accounting.
310 */
311
312/*
313 * Executable code area - executable, not writable, not stack
314 */
315static inline bool is_exec_mapping(vm_flags_t flags)
316{
317	return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC;
318}
319
320/*
321 * Stack area - automatically grows in one direction
322 *
323 * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
324 * do_mmap() forbids all other combinations.
325 */
326static inline bool is_stack_mapping(vm_flags_t flags)
327{
328	return (flags & VM_STACK) == VM_STACK;
329}
330
331/*
332 * Data area - private, writable, not stack
333 */
334static inline bool is_data_mapping(vm_flags_t flags)
335{
336	return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE;
337}
338
339/* mm/util.c */
340void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
341		struct vm_area_struct *prev);
342void __vma_unlink_list(struct mm_struct *mm, struct vm_area_struct *vma);
343
344#ifdef CONFIG_MMU
345extern long populate_vma_page_range(struct vm_area_struct *vma,
346		unsigned long start, unsigned long end, int *locked);
347extern long faultin_vma_page_range(struct vm_area_struct *vma,
348				   unsigned long start, unsigned long end,
349				   bool write, int *locked);
350extern void munlock_vma_pages_range(struct vm_area_struct *vma,
351			unsigned long start, unsigned long end);
352static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
353{
354	munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
355}
356
357/*
358 * must be called with vma's mmap_lock held for read or write, and page locked.
359 */
360extern void mlock_vma_page(struct page *page);
361extern unsigned int munlock_vma_page(struct page *page);
362
363extern int mlock_future_check(struct mm_struct *mm, unsigned long flags,
364			      unsigned long len);
365
366/*
367 * Clear the page's PageMlocked().  This can be useful in a situation where
368 * we want to unconditionally remove a page from the pagecache -- e.g.,
369 * on truncation or freeing.
370 *
371 * It is legal to call this function for any page, mlocked or not.
372 * If called for a page that is still mapped by mlocked vmas, all we do
373 * is revert to lazy LRU behaviour -- semantics are not broken.
374 */
375extern void clear_page_mlock(struct page *page);
376
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
377extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
378
379/*
380 * At what user virtual address is page expected in vma?
381 * Returns -EFAULT if all of the page is outside the range of vma.
382 * If page is a compound head, the entire compound page is considered.
383 */
384static inline unsigned long
385vma_address(struct page *page, struct vm_area_struct *vma)
386{
387	pgoff_t pgoff;
388	unsigned long address;
389
390	VM_BUG_ON_PAGE(PageKsm(page), page);	/* KSM page->index unusable */
391	pgoff = page_to_pgoff(page);
392	if (pgoff >= vma->vm_pgoff) {
393		address = vma->vm_start +
394			((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
395		/* Check for address beyond vma (or wrapped through 0?) */
396		if (address < vma->vm_start || address >= vma->vm_end)
397			address = -EFAULT;
398	} else if (PageHead(page) &&
399		   pgoff + compound_nr(page) - 1 >= vma->vm_pgoff) {
400		/* Test above avoids possibility of wrap to 0 on 32-bit */
401		address = vma->vm_start;
402	} else {
403		address = -EFAULT;
404	}
405	return address;
406}
407
408/*
409 * Then at what user virtual address will none of the page be found in vma?
410 * Assumes that vma_address() already returned a good starting address.
411 * If page is a compound head, the entire compound page is considered.
412 */
413static inline unsigned long
414vma_address_end(struct page *page, struct vm_area_struct *vma)
415{
416	pgoff_t pgoff;
417	unsigned long address;
418
419	VM_BUG_ON_PAGE(PageKsm(page), page);	/* KSM page->index unusable */
420	pgoff = page_to_pgoff(page) + compound_nr(page);
421	address = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
422	/* Check for address beyond vma (or wrapped through 0?) */
423	if (address < vma->vm_start || address > vma->vm_end)
424		address = vma->vm_end;
425	return address;
426}
427
428static inline struct file *maybe_unlock_mmap_for_io(struct vm_fault *vmf,
429						    struct file *fpin)
430{
431	int flags = vmf->flags;
432
433	if (fpin)
434		return fpin;
435
436	/*
437	 * FAULT_FLAG_RETRY_NOWAIT means we don't want to wait on page locks or
438	 * anything, so we only pin the file and drop the mmap_lock if only
439	 * FAULT_FLAG_ALLOW_RETRY is set, while this is the first attempt.
440	 */
441	if (fault_flag_allow_retry_first(flags) &&
442	    !(flags & FAULT_FLAG_RETRY_NOWAIT)) {
443		fpin = get_file(vmf->vma->vm_file);
444		mmap_read_unlock(vmf->vma->vm_mm);
445	}
446	return fpin;
447}
448
449#else /* !CONFIG_MMU */
450static inline void clear_page_mlock(struct page *page) { }
451static inline void mlock_vma_page(struct page *page) { }
452static inline void vunmap_range_noflush(unsigned long start, unsigned long end)
453{
454}
455#endif /* !CONFIG_MMU */
456
457/*
458 * Return the mem_map entry representing the 'offset' subpage within
459 * the maximally aligned gigantic page 'base'.  Handle any discontiguity
460 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
461 */
462static inline struct page *mem_map_offset(struct page *base, int offset)
463{
464	if (unlikely(offset >= MAX_ORDER_NR_PAGES))
465		return nth_page(base, offset);
466	return base + offset;
467}
468
469/*
470 * Iterator over all subpages within the maximally aligned gigantic
471 * page 'base'.  Handle any discontiguity in the mem_map.
472 */
473static inline struct page *mem_map_next(struct page *iter,
474						struct page *base, int offset)
475{
476	if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
477		unsigned long pfn = page_to_pfn(base) + offset;
478		if (!pfn_valid(pfn))
479			return NULL;
480		return pfn_to_page(pfn);
481	}
482	return iter + 1;
483}
484
485/* Memory initialisation debug and verification */
486enum mminit_level {
487	MMINIT_WARNING,
488	MMINIT_VERIFY,
489	MMINIT_TRACE
490};
491
492#ifdef CONFIG_DEBUG_MEMORY_INIT
493
494extern int mminit_loglevel;
495
496#define mminit_dprintk(level, prefix, fmt, arg...) \
497do { \
498	if (level < mminit_loglevel) { \
499		if (level <= MMINIT_WARNING) \
500			pr_warn("mminit::" prefix " " fmt, ##arg);	\
501		else \
502			printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
503	} \
504} while (0)
505
506extern void mminit_verify_pageflags_layout(void);
507extern void mminit_verify_zonelist(void);
508#else
509
510static inline void mminit_dprintk(enum mminit_level level,
511				const char *prefix, const char *fmt, ...)
512{
513}
514
515static inline void mminit_verify_pageflags_layout(void)
516{
517}
518
519static inline void mminit_verify_zonelist(void)
520{
521}
522#endif /* CONFIG_DEBUG_MEMORY_INIT */
523
524/* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
525#if defined(CONFIG_SPARSEMEM)
526extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
527				unsigned long *end_pfn);
528#else
529static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
530				unsigned long *end_pfn)
531{
532}
533#endif /* CONFIG_SPARSEMEM */
534
535#define NODE_RECLAIM_NOSCAN	-2
536#define NODE_RECLAIM_FULL	-1
537#define NODE_RECLAIM_SOME	0
538#define NODE_RECLAIM_SUCCESS	1
539
540#ifdef CONFIG_NUMA
541extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int);
542#else
543static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask,
544				unsigned int order)
545{
546	return NODE_RECLAIM_NOSCAN;
547}
548#endif
549
550extern int hwpoison_filter(struct page *p);
551
552extern u32 hwpoison_filter_dev_major;
553extern u32 hwpoison_filter_dev_minor;
554extern u64 hwpoison_filter_flags_mask;
555extern u64 hwpoison_filter_flags_value;
556extern u64 hwpoison_filter_memcg;
557extern u32 hwpoison_filter_enable;
558
559extern unsigned long  __must_check vm_mmap_pgoff(struct file *, unsigned long,
560        unsigned long, unsigned long,
561        unsigned long, unsigned long);
562
563extern void set_pageblock_order(void);
564unsigned int reclaim_clean_pages_from_list(struct zone *zone,
565					    struct list_head *page_list);
566/* The ALLOC_WMARK bits are used as an index to zone->watermark */
567#define ALLOC_WMARK_MIN		WMARK_MIN
568#define ALLOC_WMARK_LOW		WMARK_LOW
569#define ALLOC_WMARK_HIGH	WMARK_HIGH
570#define ALLOC_NO_WATERMARKS	0x04 /* don't check watermarks at all */
571
572/* Mask to get the watermark bits */
573#define ALLOC_WMARK_MASK	(ALLOC_NO_WATERMARKS-1)
574
575/*
576 * Only MMU archs have async oom victim reclaim - aka oom_reaper so we
577 * cannot assume a reduced access to memory reserves is sufficient for
578 * !MMU
579 */
580#ifdef CONFIG_MMU
581#define ALLOC_OOM		0x08
582#else
583#define ALLOC_OOM		ALLOC_NO_WATERMARKS
584#endif
585
586#define ALLOC_HARDER		 0x10 /* try to alloc harder */
587#define ALLOC_HIGH		 0x20 /* __GFP_HIGH set */
588#define ALLOC_CPUSET		 0x40 /* check for correct cpuset */
589#define ALLOC_CMA		 0x80 /* allow allocations from CMA areas */
590#ifdef CONFIG_ZONE_DMA32
591#define ALLOC_NOFRAGMENT	0x100 /* avoid mixing pageblock types */
592#else
593#define ALLOC_NOFRAGMENT	  0x0
594#endif
595#define ALLOC_KSWAPD		0x800 /* allow waking of kswapd, __GFP_KSWAPD_RECLAIM set */
596
597enum ttu_flags;
598struct tlbflush_unmap_batch;
599
600
601/*
602 * only for MM internal work items which do not depend on
603 * any allocations or locks which might depend on allocations
604 */
605extern struct workqueue_struct *mm_percpu_wq;
606
607#ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
608void try_to_unmap_flush(void);
609void try_to_unmap_flush_dirty(void);
610void flush_tlb_batched_pending(struct mm_struct *mm);
611#else
612static inline void try_to_unmap_flush(void)
613{
614}
615static inline void try_to_unmap_flush_dirty(void)
616{
617}
618static inline void flush_tlb_batched_pending(struct mm_struct *mm)
619{
620}
621#endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
622
623extern const struct trace_print_flags pageflag_names[];
624extern const struct trace_print_flags vmaflag_names[];
625extern const struct trace_print_flags gfpflag_names[];
626
627static inline bool is_migrate_highatomic(enum migratetype migratetype)
628{
629	return migratetype == MIGRATE_HIGHATOMIC;
630}
631
632static inline bool is_migrate_highatomic_page(struct page *page)
633{
634	return get_pageblock_migratetype(page) == MIGRATE_HIGHATOMIC;
635}
636
637void setup_zone_pageset(struct zone *zone);
638
639struct migration_target_control {
640	int nid;		/* preferred node id */
641	nodemask_t *nmask;
642	gfp_t gfp_mask;
643};
644
645/*
646 * mm/vmalloc.c
647 */
648#ifdef CONFIG_MMU
649int vmap_pages_range_noflush(unsigned long addr, unsigned long end,
650                pgprot_t prot, struct page **pages, unsigned int page_shift);
651#else
652static inline
653int vmap_pages_range_noflush(unsigned long addr, unsigned long end,
654                pgprot_t prot, struct page **pages, unsigned int page_shift)
655{
656	return -EINVAL;
657}
658#endif
659
660void vunmap_range_noflush(unsigned long start, unsigned long end);
661
662int numa_migrate_prep(struct page *page, struct vm_area_struct *vma,
663		      unsigned long addr, int page_nid, int *flags);
664
665#endif	/* __MM_INTERNAL_H */