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 * Locate the struct page for both the matching buddy in our
114 * pair (buddy1) and the combined O(n+1) page they form (page).
115 *
116 * 1) Any buddy B1 will have an order O twin B2 which satisfies
117 * the following equation:
118 *     B2 = B1 ^ (1 << O)
119 * For example, if the starting buddy (buddy2) is #8 its order
120 * 1 buddy is #10:
121 *     B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
122 *
123 * 2) Any buddy B will have an order O+1 parent P which
124 * satisfies the following equation:
125 *     P = B & ~(1 << O)
126 *
127 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
128 */
129static inline unsigned long
130__find_buddy_index(unsigned long page_idx, unsigned int order)
131{
132	return page_idx ^ (1 << order);
133}
134
135extern int __isolate_free_page(struct page *page, unsigned int order);
136extern void __free_pages_bootmem(struct page *page, unsigned int order);
137extern void prep_compound_page(struct page *page, unsigned long order);
138#ifdef CONFIG_MEMORY_FAILURE
139extern bool is_free_buddy_page(struct page *page);
140#endif
141extern int user_min_free_kbytes;
142
143#if defined CONFIG_COMPACTION || defined CONFIG_CMA
144
145/*
146 * in mm/compaction.c
147 */
148/*
149 * compact_control is used to track pages being migrated and the free pages
150 * they are being migrated to during memory compaction. The free_pfn starts
151 * at the end of a zone and migrate_pfn begins at the start. Movable pages
152 * are moved to the end of a zone during a compaction run and the run
153 * completes when free_pfn <= migrate_pfn
154 */
155struct compact_control {
156	struct list_head freepages;	/* List of free pages to migrate to */
157	struct list_head migratepages;	/* List of pages being migrated */
158	unsigned long nr_freepages;	/* Number of isolated free pages */
159	unsigned long nr_migratepages;	/* Number of pages to migrate */
160	unsigned long free_pfn;		/* isolate_freepages search base */
161	unsigned long migrate_pfn;	/* isolate_migratepages search base */
162	enum migrate_mode mode;		/* Async or sync migration mode */
163	bool ignore_skip_hint;		/* Scan blocks even if marked skip */
164	bool finished_update_free;	/* True when the zone cached pfns are
165					 * no longer being updated
166					 */
167	bool finished_update_migrate;
168
169	int order;			/* order a direct compactor needs */
170	const gfp_t gfp_mask;		/* gfp mask of a direct compactor */
171	struct zone *zone;
172	int contended;			/* Signal need_sched() or lock
173					 * contention detected during
174					 * compaction
175					 */
176};
177
178unsigned long
179isolate_freepages_range(struct compact_control *cc,
180			unsigned long start_pfn, unsigned long end_pfn);
181unsigned long
182isolate_migratepages_range(struct compact_control *cc,
183			   unsigned long low_pfn, unsigned long end_pfn);
184
185#endif
186
187/*
188 * This function returns the order of a free page in the buddy system. In
189 * general, page_zone(page)->lock must be held by the caller to prevent the
190 * page from being allocated in parallel and returning garbage as the order.
191 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
192 * page cannot be allocated or merged in parallel. Alternatively, it must
193 * handle invalid values gracefully, and use page_order_unsafe() below.
194 */
195static inline unsigned long page_order(struct page *page)
196{
197	/* PageBuddy() must be checked by the caller */
198	return page_private(page);
199}
200
201/*
202 * Like page_order(), but for callers who cannot afford to hold the zone lock.
203 * PageBuddy() should be checked first by the caller to minimize race window,
204 * and invalid values must be handled gracefully.
205 *
206 * ACCESS_ONCE is used so that if the caller assigns the result into a local
207 * variable and e.g. tests it for valid range before using, the compiler cannot
208 * decide to remove the variable and inline the page_private(page) multiple
209 * times, potentially observing different values in the tests and the actual
210 * use of the result.
211 */
212#define page_order_unsafe(page)		ACCESS_ONCE(page_private(page))
213
214static inline bool is_cow_mapping(vm_flags_t flags)
215{
216	return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
217}
218
219/* mm/util.c */
220void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
221		struct vm_area_struct *prev, struct rb_node *rb_parent);
222
223#ifdef CONFIG_MMU
224extern long __mlock_vma_pages_range(struct vm_area_struct *vma,
225		unsigned long start, unsigned long end, int *nonblocking);
226extern void munlock_vma_pages_range(struct vm_area_struct *vma,
227			unsigned long start, unsigned long end);
228static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
229{
230	munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
231}
232
233/*
234 * must be called with vma's mmap_sem held for read or write, and page locked.
235 */
236extern void mlock_vma_page(struct page *page);
237extern unsigned int munlock_vma_page(struct page *page);
238
239/*
240 * Clear the page's PageMlocked().  This can be useful in a situation where
241 * we want to unconditionally remove a page from the pagecache -- e.g.,
242 * on truncation or freeing.
243 *
244 * It is legal to call this function for any page, mlocked or not.
245 * If called for a page that is still mapped by mlocked vmas, all we do
246 * is revert to lazy LRU behaviour -- semantics are not broken.
247 */
248extern void clear_page_mlock(struct page *page);
249
250/*
251 * mlock_migrate_page - called only from migrate_page_copy() to
252 * migrate the Mlocked page flag; update statistics.
253 */
254static inline void mlock_migrate_page(struct page *newpage, struct page *page)
255{
256	if (TestClearPageMlocked(page)) {
257		unsigned long flags;
258		int nr_pages = hpage_nr_pages(page);
259
260		local_irq_save(flags);
261		__mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
262		SetPageMlocked(newpage);
263		__mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
264		local_irq_restore(flags);
265	}
266}
267
268extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
269
270#ifdef CONFIG_TRANSPARENT_HUGEPAGE
271extern unsigned long vma_address(struct page *page,
272				 struct vm_area_struct *vma);
273#endif
274#else /* !CONFIG_MMU */
275static inline void clear_page_mlock(struct page *page) { }
276static inline void mlock_vma_page(struct page *page) { }
277static inline void mlock_migrate_page(struct page *new, struct page *old) { }
278
279#endif /* !CONFIG_MMU */
280
281/*
282 * Return the mem_map entry representing the 'offset' subpage within
283 * the maximally aligned gigantic page 'base'.  Handle any discontiguity
284 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
285 */
286static inline struct page *mem_map_offset(struct page *base, int offset)
287{
288	if (unlikely(offset >= MAX_ORDER_NR_PAGES))
289		return nth_page(base, offset);
290	return base + offset;
291}
292
293/*
294 * Iterator over all subpages within the maximally aligned gigantic
295 * page 'base'.  Handle any discontiguity in the mem_map.
296 */
297static inline struct page *mem_map_next(struct page *iter,
298						struct page *base, int offset)
299{
300	if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
301		unsigned long pfn = page_to_pfn(base) + offset;
302		if (!pfn_valid(pfn))
303			return NULL;
304		return pfn_to_page(pfn);
305	}
306	return iter + 1;
307}
308
309/*
310 * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
311 * so all functions starting at paging_init should be marked __init
312 * in those cases. SPARSEMEM, however, allows for memory hotplug,
313 * and alloc_bootmem_node is not used.
314 */
315#ifdef CONFIG_SPARSEMEM
316#define __paginginit __meminit
317#else
318#define __paginginit __init
319#endif
320
321/* Memory initialisation debug and verification */
322enum mminit_level {
323	MMINIT_WARNING,
324	MMINIT_VERIFY,
325	MMINIT_TRACE
326};
327
328#ifdef CONFIG_DEBUG_MEMORY_INIT
329
330extern int mminit_loglevel;
331
332#define mminit_dprintk(level, prefix, fmt, arg...) \
333do { \
334	if (level < mminit_loglevel) { \
335		printk(level <= MMINIT_WARNING ? KERN_WARNING : KERN_DEBUG); \
336		printk(KERN_CONT "mminit::" prefix " " fmt, ##arg); \
337	} \
338} while (0)
339
340extern void mminit_verify_pageflags_layout(void);
341extern void mminit_verify_page_links(struct page *page,
342		enum zone_type zone, unsigned long nid, unsigned long pfn);
343extern void mminit_verify_zonelist(void);
344
345#else
346
347static inline void mminit_dprintk(enum mminit_level level,
348				const char *prefix, const char *fmt, ...)
349{
350}
351
352static inline void mminit_verify_pageflags_layout(void)
353{
354}
355
356static inline void mminit_verify_page_links(struct page *page,
357		enum zone_type zone, unsigned long nid, unsigned long pfn)
358{
359}
360
361static inline void mminit_verify_zonelist(void)
362{
363}
364#endif /* CONFIG_DEBUG_MEMORY_INIT */
365
366/* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
367#if defined(CONFIG_SPARSEMEM)
368extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
369				unsigned long *end_pfn);
370#else
371static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
372				unsigned long *end_pfn)
373{
374}
375#endif /* CONFIG_SPARSEMEM */
376
377#define ZONE_RECLAIM_NOSCAN	-2
378#define ZONE_RECLAIM_FULL	-1
379#define ZONE_RECLAIM_SOME	0
380#define ZONE_RECLAIM_SUCCESS	1
381
382extern int hwpoison_filter(struct page *p);
383
384extern u32 hwpoison_filter_dev_major;
385extern u32 hwpoison_filter_dev_minor;
386extern u64 hwpoison_filter_flags_mask;
387extern u64 hwpoison_filter_flags_value;
388extern u64 hwpoison_filter_memcg;
389extern u32 hwpoison_filter_enable;
390
391extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
392        unsigned long, unsigned long,
393        unsigned long, unsigned long);
394
395extern void set_pageblock_order(void);
396unsigned long reclaim_clean_pages_from_list(struct zone *zone,
397					    struct list_head *page_list);
398/* The ALLOC_WMARK bits are used as an index to zone->watermark */
399#define ALLOC_WMARK_MIN		WMARK_MIN
400#define ALLOC_WMARK_LOW		WMARK_LOW
401#define ALLOC_WMARK_HIGH	WMARK_HIGH
402#define ALLOC_NO_WATERMARKS	0x04 /* don't check watermarks at all */
403
404/* Mask to get the watermark bits */
405#define ALLOC_WMARK_MASK	(ALLOC_NO_WATERMARKS-1)
406
407#define ALLOC_HARDER		0x10 /* try to alloc harder */
408#define ALLOC_HIGH		0x20 /* __GFP_HIGH set */
409#define ALLOC_CPUSET		0x40 /* check for correct cpuset */
410#define ALLOC_CMA		0x80 /* allow allocations from CMA areas */
411#define ALLOC_FAIR		0x100 /* fair zone allocation */
412
413#endif	/* __MM_INTERNAL_H */
414