Loading...
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 */
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/mm.h>
15
16void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
17 unsigned long floor, unsigned long ceiling);
18
19static inline void set_page_count(struct page *page, int v)
20{
21 atomic_set(&page->_count, v);
22}
23
24/*
25 * Turn a non-refcounted page (->_count == 0) into refcounted with
26 * a count of one.
27 */
28static inline void set_page_refcounted(struct page *page)
29{
30 VM_BUG_ON(PageTail(page));
31 VM_BUG_ON(atomic_read(&page->_count));
32 set_page_count(page, 1);
33}
34
35static inline void __put_page(struct page *page)
36{
37 atomic_dec(&page->_count);
38}
39
40static inline void __get_page_tail_foll(struct page *page,
41 bool get_page_head)
42{
43 /*
44 * If we're getting a tail page, the elevated page->_count is
45 * required only in the head page and we will elevate the head
46 * page->_count and tail page->_mapcount.
47 *
48 * We elevate page_tail->_mapcount for tail pages to force
49 * page_tail->_count to be zero at all times to avoid getting
50 * false positives from get_page_unless_zero() with
51 * speculative page access (like in
52 * page_cache_get_speculative()) on tail pages.
53 */
54 VM_BUG_ON(atomic_read(&page->first_page->_count) <= 0);
55 VM_BUG_ON(atomic_read(&page->_count) != 0);
56 VM_BUG_ON(page_mapcount(page) < 0);
57 if (get_page_head)
58 atomic_inc(&page->first_page->_count);
59 atomic_inc(&page->_mapcount);
60}
61
62/*
63 * This is meant to be called as the FOLL_GET operation of
64 * follow_page() and it must be called while holding the proper PT
65 * lock while the pte (or pmd_trans_huge) is still mapping the page.
66 */
67static inline void get_page_foll(struct page *page)
68{
69 if (unlikely(PageTail(page)))
70 /*
71 * This is safe only because
72 * __split_huge_page_refcount() can't run under
73 * get_page_foll() because we hold the proper PT lock.
74 */
75 __get_page_tail_foll(page, true);
76 else {
77 /*
78 * Getting a normal page or the head of a compound page
79 * requires to already have an elevated page->_count.
80 */
81 VM_BUG_ON(atomic_read(&page->_count) <= 0);
82 atomic_inc(&page->_count);
83 }
84}
85
86extern unsigned long highest_memmap_pfn;
87
88/*
89 * in mm/vmscan.c:
90 */
91extern int isolate_lru_page(struct page *page);
92extern void putback_lru_page(struct page *page);
93
94/*
95 * in mm/page_alloc.c
96 */
97extern void __free_pages_bootmem(struct page *page, unsigned int order);
98extern void prep_compound_page(struct page *page, unsigned long order);
99#ifdef CONFIG_MEMORY_FAILURE
100extern bool is_free_buddy_page(struct page *page);
101#endif
102
103#if defined CONFIG_COMPACTION || defined CONFIG_CMA
104
105/*
106 * in mm/compaction.c
107 */
108/*
109 * compact_control is used to track pages being migrated and the free pages
110 * they are being migrated to during memory compaction. The free_pfn starts
111 * at the end of a zone and migrate_pfn begins at the start. Movable pages
112 * are moved to the end of a zone during a compaction run and the run
113 * completes when free_pfn <= migrate_pfn
114 */
115struct compact_control {
116 struct list_head freepages; /* List of free pages to migrate to */
117 struct list_head migratepages; /* List of pages being migrated */
118 unsigned long nr_freepages; /* Number of isolated free pages */
119 unsigned long nr_migratepages; /* Number of pages to migrate */
120 unsigned long free_pfn; /* isolate_freepages search base */
121 unsigned long migrate_pfn; /* isolate_migratepages search base */
122 bool sync; /* Synchronous migration */
123
124 int order; /* order a direct compactor needs */
125 int migratetype; /* MOVABLE, RECLAIMABLE etc */
126 struct zone *zone;
127};
128
129unsigned long
130isolate_freepages_range(unsigned long start_pfn, unsigned long end_pfn);
131unsigned long
132isolate_migratepages_range(struct zone *zone, struct compact_control *cc,
133 unsigned long low_pfn, unsigned long end_pfn);
134
135#endif
136
137/*
138 * function for dealing with page's order in buddy system.
139 * zone->lock is already acquired when we use these.
140 * So, we don't need atomic page->flags operations here.
141 */
142static inline unsigned long page_order(struct page *page)
143{
144 /* PageBuddy() must be checked by the caller */
145 return page_private(page);
146}
147
148/* mm/util.c */
149void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
150 struct vm_area_struct *prev, struct rb_node *rb_parent);
151
152#ifdef CONFIG_MMU
153extern long mlock_vma_pages_range(struct vm_area_struct *vma,
154 unsigned long start, unsigned long end);
155extern void munlock_vma_pages_range(struct vm_area_struct *vma,
156 unsigned long start, unsigned long end);
157static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
158{
159 munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
160}
161
162/*
163 * Called only in fault path via page_evictable() for a new page
164 * to determine if it's being mapped into a LOCKED vma.
165 * If so, mark page as mlocked.
166 */
167static inline int mlocked_vma_newpage(struct vm_area_struct *vma,
168 struct page *page)
169{
170 VM_BUG_ON(PageLRU(page));
171
172 if (likely((vma->vm_flags & (VM_LOCKED | VM_SPECIAL)) != VM_LOCKED))
173 return 0;
174
175 if (!TestSetPageMlocked(page)) {
176 inc_zone_page_state(page, NR_MLOCK);
177 count_vm_event(UNEVICTABLE_PGMLOCKED);
178 }
179 return 1;
180}
181
182/*
183 * must be called with vma's mmap_sem held for read or write, and page locked.
184 */
185extern void mlock_vma_page(struct page *page);
186extern void munlock_vma_page(struct page *page);
187
188/*
189 * Clear the page's PageMlocked(). This can be useful in a situation where
190 * we want to unconditionally remove a page from the pagecache -- e.g.,
191 * on truncation or freeing.
192 *
193 * It is legal to call this function for any page, mlocked or not.
194 * If called for a page that is still mapped by mlocked vmas, all we do
195 * is revert to lazy LRU behaviour -- semantics are not broken.
196 */
197extern void __clear_page_mlock(struct page *page);
198static inline void clear_page_mlock(struct page *page)
199{
200 if (unlikely(TestClearPageMlocked(page)))
201 __clear_page_mlock(page);
202}
203
204/*
205 * mlock_migrate_page - called only from migrate_page_copy() to
206 * migrate the Mlocked page flag; update statistics.
207 */
208static inline void mlock_migrate_page(struct page *newpage, struct page *page)
209{
210 if (TestClearPageMlocked(page)) {
211 unsigned long flags;
212
213 local_irq_save(flags);
214 __dec_zone_page_state(page, NR_MLOCK);
215 SetPageMlocked(newpage);
216 __inc_zone_page_state(newpage, NR_MLOCK);
217 local_irq_restore(flags);
218 }
219}
220
221#ifdef CONFIG_TRANSPARENT_HUGEPAGE
222extern unsigned long vma_address(struct page *page,
223 struct vm_area_struct *vma);
224#endif
225#else /* !CONFIG_MMU */
226static inline int mlocked_vma_newpage(struct vm_area_struct *v, struct page *p)
227{
228 return 0;
229}
230static inline void clear_page_mlock(struct page *page) { }
231static inline void mlock_vma_page(struct page *page) { }
232static inline void mlock_migrate_page(struct page *new, struct page *old) { }
233
234#endif /* !CONFIG_MMU */
235
236/*
237 * Return the mem_map entry representing the 'offset' subpage within
238 * the maximally aligned gigantic page 'base'. Handle any discontiguity
239 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
240 */
241static inline struct page *mem_map_offset(struct page *base, int offset)
242{
243 if (unlikely(offset >= MAX_ORDER_NR_PAGES))
244 return pfn_to_page(page_to_pfn(base) + offset);
245 return base + offset;
246}
247
248/*
249 * Iterator over all subpages within the maximally aligned gigantic
250 * page 'base'. Handle any discontiguity in the mem_map.
251 */
252static inline struct page *mem_map_next(struct page *iter,
253 struct page *base, int offset)
254{
255 if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
256 unsigned long pfn = page_to_pfn(base) + offset;
257 if (!pfn_valid(pfn))
258 return NULL;
259 return pfn_to_page(pfn);
260 }
261 return iter + 1;
262}
263
264/*
265 * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
266 * so all functions starting at paging_init should be marked __init
267 * in those cases. SPARSEMEM, however, allows for memory hotplug,
268 * and alloc_bootmem_node is not used.
269 */
270#ifdef CONFIG_SPARSEMEM
271#define __paginginit __meminit
272#else
273#define __paginginit __init
274#endif
275
276/* Memory initialisation debug and verification */
277enum mminit_level {
278 MMINIT_WARNING,
279 MMINIT_VERIFY,
280 MMINIT_TRACE
281};
282
283#ifdef CONFIG_DEBUG_MEMORY_INIT
284
285extern int mminit_loglevel;
286
287#define mminit_dprintk(level, prefix, fmt, arg...) \
288do { \
289 if (level < mminit_loglevel) { \
290 printk(level <= MMINIT_WARNING ? KERN_WARNING : KERN_DEBUG); \
291 printk(KERN_CONT "mminit::" prefix " " fmt, ##arg); \
292 } \
293} while (0)
294
295extern void mminit_verify_pageflags_layout(void);
296extern void mminit_verify_page_links(struct page *page,
297 enum zone_type zone, unsigned long nid, unsigned long pfn);
298extern void mminit_verify_zonelist(void);
299
300#else
301
302static inline void mminit_dprintk(enum mminit_level level,
303 const char *prefix, const char *fmt, ...)
304{
305}
306
307static inline void mminit_verify_pageflags_layout(void)
308{
309}
310
311static inline void mminit_verify_page_links(struct page *page,
312 enum zone_type zone, unsigned long nid, unsigned long pfn)
313{
314}
315
316static inline void mminit_verify_zonelist(void)
317{
318}
319#endif /* CONFIG_DEBUG_MEMORY_INIT */
320
321/* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
322#if defined(CONFIG_SPARSEMEM)
323extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
324 unsigned long *end_pfn);
325#else
326static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
327 unsigned long *end_pfn)
328{
329}
330#endif /* CONFIG_SPARSEMEM */
331
332#define ZONE_RECLAIM_NOSCAN -2
333#define ZONE_RECLAIM_FULL -1
334#define ZONE_RECLAIM_SOME 0
335#define ZONE_RECLAIM_SUCCESS 1
336#endif
337
338extern int hwpoison_filter(struct page *p);
339
340extern u32 hwpoison_filter_dev_major;
341extern u32 hwpoison_filter_dev_minor;
342extern u64 hwpoison_filter_flags_mask;
343extern u64 hwpoison_filter_flags_value;
344extern u64 hwpoison_filter_memcg;
345extern u32 hwpoison_filter_enable;
346
347extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
348 unsigned long, unsigned long,
349 unsigned long, unsigned long);
350
351extern void set_pageblock_order(void);