1/* internal.h: mm/ internal definitions
2 *
3 * Copyright (C) 2004 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
5 *
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.
10 */
11#ifndef __MM_INTERNAL_H
12#define __MM_INTERNAL_H
13
14#include <linux/fs.h>
15#include <linux/mm.h>
16
17void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
18		unsigned long floor, unsigned long ceiling);
19
20static inline void set_page_count(struct page *page, int v)
21{
22	atomic_set(&page->_count, v);
23}
24
25extern int __do_page_cache_readahead(struct address_space *mapping,
26		struct file *filp, pgoff_t offset, unsigned long nr_to_read,
27		unsigned long lookahead_size);
28
29/*
30 * Submit IO for the read-ahead request in file_ra_state.
31 */
32static inline unsigned long ra_submit(struct file_ra_state *ra,
33		struct address_space *mapping, struct file *filp)
34{
35	return __do_page_cache_readahead(mapping, filp,
36					ra->start, ra->size, ra->async_size);
37}
38
39/*
40 * Turn a non-refcounted page (->_count == 0) into refcounted with
41 * a count of one.
42 */
43static inline void set_page_refcounted(struct page *page)
44{
45	VM_BUG_ON_PAGE(PageTail(page), page);
46	VM_BUG_ON_PAGE(atomic_read(&page->_count), page);
47	set_page_count(page, 1);
48}
49
50static inline void __get_page_tail_foll(struct page *page,
51					bool get_page_head)
52{
53	/*
54	 * If we're getting a tail page, the elevated page->_count is
55	 * required only in the head page and we will elevate the head
56	 * page->_count and tail page->_mapcount.
57	 *
58	 * We elevate page_tail->_mapcount for tail pages to force
59	 * page_tail->_count to be zero at all times to avoid getting
60	 * false positives from get_page_unless_zero() with
61	 * speculative page access (like in
62	 * page_cache_get_speculative()) on tail pages.
63	 */
64	VM_BUG_ON_PAGE(atomic_read(&page->first_page->_count) <= 0, page);
65	if (get_page_head)
66		atomic_inc(&page->first_page->_count);
67	get_huge_page_tail(page);
68}
69
70/*
71 * This is meant to be called as the FOLL_GET operation of
72 * follow_page() and it must be called while holding the proper PT
73 * lock while the pte (or pmd_trans_huge) is still mapping the page.
74 */
75static inline void get_page_foll(struct page *page)
76{
77	if (unlikely(PageTail(page)))
78		/*
79		 * This is safe only because
80		 * __split_huge_page_refcount() can't run under
81		 * get_page_foll() because we hold the proper PT lock.
82		 */
83		__get_page_tail_foll(page, true);
84	else {
85		/*
86		 * Getting a normal page or the head of a compound page
87		 * requires to already have an elevated page->_count.
88		 */
89		VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page);
90		atomic_inc(&page->_count);
91	}
92}
93
94extern unsigned long highest_memmap_pfn;
95
96/*
97 * in mm/vmscan.c:
98 */
99extern int isolate_lru_page(struct page *page);
100extern void putback_lru_page(struct page *page);
101extern bool zone_reclaimable(struct zone *zone);
102
103/*
104 * in mm/rmap.c:
105 */
106extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
107
108/*
109 * in mm/page_alloc.c
110 */
111
112/*
113 * Structure for holding the mostly immutable allocation parameters passed
114 * between functions involved in allocations, including the alloc_pages*
115 * family of functions.
116 *
117 * nodemask, migratetype and high_zoneidx are initialized only once in
118 * __alloc_pages_nodemask() and then never change.
119 *
120 * zonelist, preferred_zone and classzone_idx are set first in
121 * __alloc_pages_nodemask() for the fast path, and might be later changed
122 * in __alloc_pages_slowpath(). All other functions pass the whole strucure
123 * by a const pointer.
124 */
125struct alloc_context {
126	struct zonelist *zonelist;
127	nodemask_t *nodemask;
128	struct zone *preferred_zone;
129	int classzone_idx;
130	int migratetype;
131	enum zone_type high_zoneidx;
132};
133
134/*
135 * Locate the struct page for both the matching buddy in our
136 * pair (buddy1) and the combined O(n+1) page they form (page).
137 *
138 * 1) Any buddy B1 will have an order O twin B2 which satisfies
139 * the following equation:
140 *     B2 = B1 ^ (1 << O)
141 * For example, if the starting buddy (buddy2) is #8 its order
142 * 1 buddy is #10:
143 *     B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
144 *
145 * 2) Any buddy B will have an order O+1 parent P which
146 * satisfies the following equation:
147 *     P = B & ~(1 << O)
148 *
149 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
150 */
151static inline unsigned long
152__find_buddy_index(unsigned long page_idx, unsigned int order)
153{
154	return page_idx ^ (1 << order);
155}
156
157extern int __isolate_free_page(struct page *page, unsigned int order);
158extern void __free_pages_bootmem(struct page *page, unsigned long pfn,
159					unsigned int order);
160extern void prep_compound_page(struct page *page, unsigned int order);
161#ifdef CONFIG_MEMORY_FAILURE
162extern bool is_free_buddy_page(struct page *page);
163#endif
164extern int user_min_free_kbytes;
165
166#if defined CONFIG_COMPACTION || defined CONFIG_CMA
167
168/*
169 * in mm/compaction.c
170 */
171/*
172 * compact_control is used to track pages being migrated and the free pages
173 * they are being migrated to during memory compaction. The free_pfn starts
174 * at the end of a zone and migrate_pfn begins at the start. Movable pages
175 * are moved to the end of a zone during a compaction run and the run
176 * completes when free_pfn <= migrate_pfn
177 */
178struct compact_control {
179	struct list_head freepages;	/* List of free pages to migrate to */
180	struct list_head migratepages;	/* List of pages being migrated */
181	unsigned long nr_freepages;	/* Number of isolated free pages */
182	unsigned long nr_migratepages;	/* Number of pages to migrate */
183	unsigned long free_pfn;		/* isolate_freepages search base */
184	unsigned long migrate_pfn;	/* isolate_migratepages search base */
185	enum migrate_mode mode;		/* Async or sync migration mode */
186	bool ignore_skip_hint;		/* Scan blocks even if marked skip */
187	int order;			/* order a direct compactor needs */
188	const gfp_t gfp_mask;		/* gfp mask of a direct compactor */
189	const int alloc_flags;		/* alloc flags of a direct compactor */
190	const int classzone_idx;	/* zone index of a direct compactor */
191	struct zone *zone;
192	int contended;			/* Signal need_sched() or lock
193					 * contention detected during
194					 * compaction
195					 */
196};
197
198unsigned long
199isolate_freepages_range(struct compact_control *cc,
200			unsigned long start_pfn, unsigned long end_pfn);
201unsigned long
202isolate_migratepages_range(struct compact_control *cc,
203			   unsigned long low_pfn, unsigned long end_pfn);
204int find_suitable_fallback(struct free_area *area, unsigned int order,
205			int migratetype, bool only_stealable, bool *can_steal);
206
207#endif
208
209/*
210 * This function returns the order of a free page in the buddy system. In
211 * general, page_zone(page)->lock must be held by the caller to prevent the
212 * page from being allocated in parallel and returning garbage as the order.
213 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
214 * page cannot be allocated or merged in parallel. Alternatively, it must
215 * handle invalid values gracefully, and use page_order_unsafe() below.
216 */
217static inline unsigned int page_order(struct page *page)
218{
219	/* PageBuddy() must be checked by the caller */
220	return page_private(page);
221}
222
223/*
224 * Like page_order(), but for callers who cannot afford to hold the zone lock.
225 * PageBuddy() should be checked first by the caller to minimize race window,
226 * and invalid values must be handled gracefully.
227 *
228 * READ_ONCE is used so that if the caller assigns the result into a local
229 * variable and e.g. tests it for valid range before using, the compiler cannot
230 * decide to remove the variable and inline the page_private(page) multiple
231 * times, potentially observing different values in the tests and the actual
232 * use of the result.
233 */
234#define page_order_unsafe(page)		READ_ONCE(page_private(page))
235
236static inline bool is_cow_mapping(vm_flags_t flags)
237{
238	return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
239}
240
241/* mm/util.c */
242void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
243		struct vm_area_struct *prev, struct rb_node *rb_parent);
244
245#ifdef CONFIG_MMU
246extern long populate_vma_page_range(struct vm_area_struct *vma,
247		unsigned long start, unsigned long end, int *nonblocking);
248extern void munlock_vma_pages_range(struct vm_area_struct *vma,
249			unsigned long start, unsigned long end);
250static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
251{
252	munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
253}
254
255/*
256 * must be called with vma's mmap_sem held for read or write, and page locked.
257 */
258extern void mlock_vma_page(struct page *page);
259extern unsigned int munlock_vma_page(struct page *page);
260
261/*
262 * Clear the page's PageMlocked().  This can be useful in a situation where
263 * we want to unconditionally remove a page from the pagecache -- e.g.,
264 * on truncation or freeing.
265 *
266 * It is legal to call this function for any page, mlocked or not.
267 * If called for a page that is still mapped by mlocked vmas, all we do
268 * is revert to lazy LRU behaviour -- semantics are not broken.
269 */
270extern void clear_page_mlock(struct page *page);
271
272/*
273 * mlock_migrate_page - called only from migrate_page_copy() to
274 * migrate the Mlocked page flag; update statistics.
275 */
276static inline void mlock_migrate_page(struct page *newpage, struct page *page)
277{
278	if (TestClearPageMlocked(page)) {
279		unsigned long flags;
280		int nr_pages = hpage_nr_pages(page);
281
282		local_irq_save(flags);
283		__mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
284		SetPageMlocked(newpage);
285		__mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
286		local_irq_restore(flags);
287	}
288}
289
290extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
291
292#ifdef CONFIG_TRANSPARENT_HUGEPAGE
293extern unsigned long vma_address(struct page *page,
294				 struct vm_area_struct *vma);
295#endif
296#else /* !CONFIG_MMU */
297static inline void clear_page_mlock(struct page *page) { }
298static inline void mlock_vma_page(struct page *page) { }
299static inline void mlock_migrate_page(struct page *new, struct page *old) { }
300
301#endif /* !CONFIG_MMU */
302
303/*
304 * Return the mem_map entry representing the 'offset' subpage within
305 * the maximally aligned gigantic page 'base'.  Handle any discontiguity
306 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
307 */
308static inline struct page *mem_map_offset(struct page *base, int offset)
309{
310	if (unlikely(offset >= MAX_ORDER_NR_PAGES))
311		return nth_page(base, offset);
312	return base + offset;
313}
314
315/*
316 * Iterator over all subpages within the maximally aligned gigantic
317 * page 'base'.  Handle any discontiguity in the mem_map.
318 */
319static inline struct page *mem_map_next(struct page *iter,
320						struct page *base, int offset)
321{
322	if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
323		unsigned long pfn = page_to_pfn(base) + offset;
324		if (!pfn_valid(pfn))
325			return NULL;
326		return pfn_to_page(pfn);
327	}
328	return iter + 1;
329}
330
331/*
332 * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
333 * so all functions starting at paging_init should be marked __init
334 * in those cases. SPARSEMEM, however, allows for memory hotplug,
335 * and alloc_bootmem_node is not used.
336 */
337#ifdef CONFIG_SPARSEMEM
338#define __paginginit __meminit
339#else
340#define __paginginit __init
341#endif
342
343/* Memory initialisation debug and verification */
344enum mminit_level {
345	MMINIT_WARNING,
346	MMINIT_VERIFY,
347	MMINIT_TRACE
348};
349
350#ifdef CONFIG_DEBUG_MEMORY_INIT
351
352extern int mminit_loglevel;
353
354#define mminit_dprintk(level, prefix, fmt, arg...) \
355do { \
356	if (level < mminit_loglevel) { \
357		if (level <= MMINIT_WARNING) \
358			printk(KERN_WARNING "mminit::" prefix " " fmt, ##arg); \
359		else \
360			printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
361	} \
362} while (0)
363
364extern void mminit_verify_pageflags_layout(void);
365extern void mminit_verify_page_links(struct page *page,
366		enum zone_type zone, unsigned long nid, unsigned long pfn);
367extern void mminit_verify_zonelist(void);
368
369#else
370
371static inline void mminit_dprintk(enum mminit_level level,
372				const char *prefix, const char *fmt, ...)
373{
374}
375
376static inline void mminit_verify_pageflags_layout(void)
377{
378}
379
380static inline void mminit_verify_page_links(struct page *page,
381		enum zone_type zone, unsigned long nid, unsigned long pfn)
382{
383}
384
385static inline void mminit_verify_zonelist(void)
386{
387}
388#endif /* CONFIG_DEBUG_MEMORY_INIT */
389
390/* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
391#if defined(CONFIG_SPARSEMEM)
392extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
393				unsigned long *end_pfn);
394#else
395static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
396				unsigned long *end_pfn)
397{
398}
399#endif /* CONFIG_SPARSEMEM */
400
401#define ZONE_RECLAIM_NOSCAN	-2
402#define ZONE_RECLAIM_FULL	-1
403#define ZONE_RECLAIM_SOME	0
404#define ZONE_RECLAIM_SUCCESS	1
405
406extern int hwpoison_filter(struct page *p);
407
408extern u32 hwpoison_filter_dev_major;
409extern u32 hwpoison_filter_dev_minor;
410extern u64 hwpoison_filter_flags_mask;
411extern u64 hwpoison_filter_flags_value;
412extern u64 hwpoison_filter_memcg;
413extern u32 hwpoison_filter_enable;
414
415extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
416        unsigned long, unsigned long,
417        unsigned long, unsigned long);
418
419extern void set_pageblock_order(void);
420unsigned long reclaim_clean_pages_from_list(struct zone *zone,
421					    struct list_head *page_list);
422/* The ALLOC_WMARK bits are used as an index to zone->watermark */
423#define ALLOC_WMARK_MIN		WMARK_MIN
424#define ALLOC_WMARK_LOW		WMARK_LOW
425#define ALLOC_WMARK_HIGH	WMARK_HIGH
426#define ALLOC_NO_WATERMARKS	0x04 /* don't check watermarks at all */
427
428/* Mask to get the watermark bits */
429#define ALLOC_WMARK_MASK	(ALLOC_NO_WATERMARKS-1)
430
431#define ALLOC_HARDER		0x10 /* try to alloc harder */
432#define ALLOC_HIGH		0x20 /* __GFP_HIGH set */
433#define ALLOC_CPUSET		0x40 /* check for correct cpuset */
434#define ALLOC_CMA		0x80 /* allow allocations from CMA areas */
435#define ALLOC_FAIR		0x100 /* fair zone allocation */
436
437#endif	/* __MM_INTERNAL_H */
438