filemap: Convert filemap_get_read_batch() to use a folio_batch
[linux-2.6-block.git] / mm / internal.h
CommitLineData
2874c5fd 1/* SPDX-License-Identifier: GPL-2.0-or-later */
1da177e4
LT
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)
1da177e4 6 */
0f8053a5
NP
7#ifndef __MM_INTERNAL_H
8#define __MM_INTERNAL_H
9
29f175d1 10#include <linux/fs.h>
0f8053a5 11#include <linux/mm.h>
e9b61f19 12#include <linux/pagemap.h>
edf14cdb 13#include <linux/tracepoint-defs.h>
1da177e4 14
dd56b046
MG
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|\
dcda9b04 22 __GFP_NOWARN|__GFP_RETRY_MAYFAIL|__GFP_NOFAIL|\
e838a45f
MG
23 __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC|\
24 __GFP_ATOMIC)
dd56b046
MG
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
62906027
NP
35void page_writeback_init(void);
36
64601000
MWO
37static inline void *folio_raw_mapping(struct folio *folio)
38{
39 unsigned long mapping = (unsigned long)folio->mapping;
40
41 return (void *)(mapping & ~PAGE_MAPPING_FLAGS);
42}
43
512b7931 44void __acct_reclaim_writeback(pg_data_t *pgdat, struct folio *folio,
8cd7c588 45 int nr_throttled);
512b7931 46static inline void acct_reclaim_writeback(struct folio *folio)
8cd7c588 47{
512b7931 48 pg_data_t *pgdat = folio_pgdat(folio);
8cd7c588
MG
49 int nr_throttled = atomic_read(&pgdat->nr_writeback_throttled);
50
51 if (nr_throttled)
512b7931 52 __acct_reclaim_writeback(pgdat, folio, nr_throttled);
8cd7c588
MG
53}
54
d818fca1
MG
55static inline void wake_throttle_isolated(pg_data_t *pgdat)
56{
57 wait_queue_head_t *wqh;
58
59 wqh = &pgdat->reclaim_wait[VMSCAN_THROTTLE_ISOLATED];
60 if (waitqueue_active(wqh))
61 wake_up(wqh);
62}
63
2b740303 64vm_fault_t do_swap_page(struct vm_fault *vmf);
575ced1c 65void folio_rotate_reclaimable(struct folio *folio);
269ccca3 66bool __folio_end_writeback(struct folio *folio);
8a966ed7 67
42b77728
JB
68void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
69 unsigned long floor, unsigned long ceiling);
03c4f204 70void pmd_install(struct mm_struct *mm, pmd_t *pmd, pgtable_t *pte);
42b77728 71
9c276cc6 72static inline bool can_madv_lru_vma(struct vm_area_struct *vma)
23519073
KS
73{
74 return !(vma->vm_flags & (VM_LOCKED|VM_HUGETLB|VM_PFNMAP));
75}
76
3506659e 77struct zap_details;
aac45363
MH
78void unmap_page_range(struct mmu_gather *tlb,
79 struct vm_area_struct *vma,
80 unsigned long addr, unsigned long end,
81 struct zap_details *details);
82
7b3df3b9
DH
83void do_page_cache_ra(struct readahead_control *, unsigned long nr_to_read,
84 unsigned long lookahead_size);
fcd9ae4f 85void force_page_cache_ra(struct readahead_control *, unsigned long nr);
7b3df3b9
DH
86static inline void force_page_cache_readahead(struct address_space *mapping,
87 struct file *file, pgoff_t index, unsigned long nr_to_read)
88{
fcd9ae4f
MWO
89 DEFINE_READAHEAD(ractl, file, &file->f_ra, mapping, index);
90 force_page_cache_ra(&ractl, nr_to_read);
7b3df3b9 91}
29f175d1 92
5c211ba2
MWO
93unsigned find_lock_entries(struct address_space *mapping, pgoff_t start,
94 pgoff_t end, struct pagevec *pvec, pgoff_t *indices);
78f42660 95void filemap_free_folio(struct address_space *mapping, struct folio *folio);
1e84a3d9 96int truncate_inode_folio(struct address_space *mapping, struct folio *folio);
5c211ba2 97
1eb6234e 98/**
3eed3ef5
MWO
99 * folio_evictable - Test whether a folio is evictable.
100 * @folio: The folio to test.
1eb6234e 101 *
3eed3ef5
MWO
102 * Test whether @folio is evictable -- i.e., should be placed on
103 * active/inactive lists vs unevictable list.
1eb6234e 104 *
3eed3ef5
MWO
105 * Reasons folio might not be evictable:
106 * 1. folio's mapping marked unevictable
107 * 2. One of the pages in the folio is part of an mlocked VMA
1eb6234e 108 */
3eed3ef5
MWO
109static inline bool folio_evictable(struct folio *folio)
110{
111 bool ret;
112
113 /* Prevent address_space of inode and swap cache from being freed */
114 rcu_read_lock();
115 ret = !mapping_unevictable(folio_mapping(folio)) &&
116 !folio_test_mlocked(folio);
117 rcu_read_unlock();
118 return ret;
119}
120
1eb6234e
YS
121static inline bool page_evictable(struct page *page)
122{
123 bool ret;
124
125 /* Prevent address_space of inode and swap cache from being freed */
126 rcu_read_lock();
127 ret = !mapping_unevictable(page_mapping(page)) && !PageMlocked(page);
128 rcu_read_unlock();
129 return ret;
130}
131
7835e98b 132/*
0139aa7b 133 * Turn a non-refcounted page (->_refcount == 0) into refcounted with
7835e98b
NP
134 * a count of one.
135 */
136static inline void set_page_refcounted(struct page *page)
137{
309381fe 138 VM_BUG_ON_PAGE(PageTail(page), page);
fe896d18 139 VM_BUG_ON_PAGE(page_ref_count(page), page);
77a8a788 140 set_page_count(page, 1);
77a8a788
NP
141}
142
03f6462a
HD
143extern unsigned long highest_memmap_pfn;
144
c73322d0
JW
145/*
146 * Maximum number of reclaim retries without progress before the OOM
147 * killer is consider the only way forward.
148 */
149#define MAX_RECLAIM_RETRIES 16
150
894bc310
LS
151/*
152 * in mm/vmscan.c:
153 */
62695a84 154extern int isolate_lru_page(struct page *page);
894bc310 155extern void putback_lru_page(struct page *page);
c3f4a9a2 156extern void reclaim_throttle(pg_data_t *pgdat, enum vmscan_throttle_state reason);
62695a84 157
6219049a
BL
158/*
159 * in mm/rmap.c:
160 */
161extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
162
494c1dfe
WL
163/*
164 * in mm/memcontrol.c:
165 */
166extern bool cgroup_memory_nokmem;
167
894bc310
LS
168/*
169 * in mm/page_alloc.c
170 */
3c605096 171
1a6d53a1
VB
172/*
173 * Structure for holding the mostly immutable allocation parameters passed
174 * between functions involved in allocations, including the alloc_pages*
175 * family of functions.
176 *
97a225e6 177 * nodemask, migratetype and highest_zoneidx are initialized only once in
84172f4b 178 * __alloc_pages() and then never change.
1a6d53a1 179 *
97a225e6 180 * zonelist, preferred_zone and highest_zoneidx are set first in
84172f4b 181 * __alloc_pages() for the fast path, and might be later changed
68956ccb 182 * in __alloc_pages_slowpath(). All other functions pass the whole structure
1a6d53a1
VB
183 * by a const pointer.
184 */
185struct alloc_context {
186 struct zonelist *zonelist;
187 nodemask_t *nodemask;
c33d6c06 188 struct zoneref *preferred_zoneref;
1a6d53a1 189 int migratetype;
97a225e6
JK
190
191 /*
192 * highest_zoneidx represents highest usable zone index of
193 * the allocation request. Due to the nature of the zone,
194 * memory on lower zone than the highest_zoneidx will be
195 * protected by lowmem_reserve[highest_zoneidx].
196 *
197 * highest_zoneidx is also used by reclaim/compaction to limit
198 * the target zone since higher zone than this index cannot be
199 * usable for this allocation request.
200 */
201 enum zone_type highest_zoneidx;
c9ab0c4f 202 bool spread_dirty_pages;
1a6d53a1
VB
203};
204
3c605096
JK
205/*
206 * Locate the struct page for both the matching buddy in our
207 * pair (buddy1) and the combined O(n+1) page they form (page).
208 *
209 * 1) Any buddy B1 will have an order O twin B2 which satisfies
210 * the following equation:
211 * B2 = B1 ^ (1 << O)
212 * For example, if the starting buddy (buddy2) is #8 its order
213 * 1 buddy is #10:
214 * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
215 *
216 * 2) Any buddy B will have an order O+1 parent P which
217 * satisfies the following equation:
218 * P = B & ~(1 << O)
219 *
220 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
221 */
222static inline unsigned long
76741e77 223__find_buddy_pfn(unsigned long page_pfn, unsigned int order)
3c605096 224{
76741e77 225 return page_pfn ^ (1 << order);
3c605096
JK
226}
227
7cf91a98
JK
228extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn,
229 unsigned long end_pfn, struct zone *zone);
230
231static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn,
232 unsigned long end_pfn, struct zone *zone)
233{
234 if (zone->contiguous)
235 return pfn_to_page(start_pfn);
236
237 return __pageblock_pfn_to_page(start_pfn, end_pfn, zone);
238}
239
3c605096 240extern int __isolate_free_page(struct page *page, unsigned int order);
624f58d8
AD
241extern void __putback_isolated_page(struct page *page, unsigned int order,
242 int mt);
7c2ee349 243extern void memblock_free_pages(struct page *page, unsigned long pfn,
d70ddd7a 244 unsigned int order);
a9cd410a 245extern void __free_pages_core(struct page *page, unsigned int order);
d00181b9 246extern void prep_compound_page(struct page *page, unsigned int order);
46f24fd8
JK
247extern void post_alloc_hook(struct page *page, unsigned int order,
248 gfp_t gfp_flags);
42aa83cb 249extern int user_min_free_kbytes;
20a0307c 250
44042b44 251extern void free_unref_page(struct page *page, unsigned int order);
0966aeb4
MWO
252extern void free_unref_page_list(struct list_head *list);
253
04f8cfea 254extern void zone_pcp_update(struct zone *zone, int cpu_online);
68265390 255extern void zone_pcp_reset(struct zone *zone);
ec6e8c7e
VB
256extern void zone_pcp_disable(struct zone *zone);
257extern void zone_pcp_enable(struct zone *zone);
68265390 258
c803b3c8
MR
259extern void *memmap_alloc(phys_addr_t size, phys_addr_t align,
260 phys_addr_t min_addr,
261 int nid, bool exact_nid);
262
ff9543fd
MN
263#if defined CONFIG_COMPACTION || defined CONFIG_CMA
264
265/*
266 * in mm/compaction.c
267 */
268/*
269 * compact_control is used to track pages being migrated and the free pages
270 * they are being migrated to during memory compaction. The free_pfn starts
271 * at the end of a zone and migrate_pfn begins at the start. Movable pages
272 * are moved to the end of a zone during a compaction run and the run
273 * completes when free_pfn <= migrate_pfn
274 */
275struct compact_control {
276 struct list_head freepages; /* List of free pages to migrate to */
277 struct list_head migratepages; /* List of pages being migrated */
c5fbd937
MG
278 unsigned int nr_freepages; /* Number of isolated free pages */
279 unsigned int nr_migratepages; /* Number of pages to migrate */
ff9543fd 280 unsigned long free_pfn; /* isolate_freepages search base */
c2ad7a1f
OS
281 /*
282 * Acts as an in/out parameter to page isolation for migration.
283 * isolate_migratepages uses it as a search base.
284 * isolate_migratepages_block will update the value to the next pfn
285 * after the last isolated one.
286 */
287 unsigned long migrate_pfn;
70b44595 288 unsigned long fast_start_pfn; /* a pfn to start linear scan from */
c5943b9c
MG
289 struct zone *zone;
290 unsigned long total_migrate_scanned;
291 unsigned long total_free_scanned;
dbe2d4e4
MG
292 unsigned short fast_search_fail;/* failures to use free list searches */
293 short search_order; /* order to start a fast search at */
f25ba6dc
VB
294 const gfp_t gfp_mask; /* gfp mask of a direct compactor */
295 int order; /* order a direct compactor needs */
d39773a0 296 int migratetype; /* migratetype of direct compactor */
f25ba6dc 297 const unsigned int alloc_flags; /* alloc flags of a direct compactor */
97a225e6 298 const int highest_zoneidx; /* zone index of a direct compactor */
e0b9daeb 299 enum migrate_mode mode; /* Async or sync migration mode */
bb13ffeb 300 bool ignore_skip_hint; /* Scan blocks even if marked skip */
2583d671 301 bool no_set_skip_hint; /* Don't mark blocks for skipping */
9f7e3387 302 bool ignore_block_suitable; /* Scan blocks considered unsuitable */
accf6242 303 bool direct_compaction; /* False from kcompactd or /proc/... */
facdaa91 304 bool proactive_compaction; /* kcompactd proactive compaction */
06ed2998 305 bool whole_zone; /* Whole zone should/has been scanned */
c3486f53 306 bool contended; /* Signal lock or sched contention */
804d3121 307 bool rescan; /* Rescanning the same pageblock */
b06eda09 308 bool alloc_contig; /* alloc_contig_range allocation */
ff9543fd
MN
309};
310
5e1f0f09
MG
311/*
312 * Used in direct compaction when a page should be taken from the freelists
313 * immediately when one is created during the free path.
314 */
315struct capture_control {
316 struct compact_control *cc;
317 struct page *page;
318};
319
ff9543fd 320unsigned long
bb13ffeb
MG
321isolate_freepages_range(struct compact_control *cc,
322 unsigned long start_pfn, unsigned long end_pfn);
c2ad7a1f 323int
edc2ca61
VB
324isolate_migratepages_range(struct compact_control *cc,
325 unsigned long low_pfn, unsigned long end_pfn);
ffd8f251 326#endif
2149cdae
JK
327int find_suitable_fallback(struct free_area *area, unsigned int order,
328 int migratetype, bool only_stealable, bool *can_steal);
ff9543fd 329
48f13bf3 330/*
6c14466c
MG
331 * This function returns the order of a free page in the buddy system. In
332 * general, page_zone(page)->lock must be held by the caller to prevent the
333 * page from being allocated in parallel and returning garbage as the order.
334 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
99c0fd5e 335 * page cannot be allocated or merged in parallel. Alternatively, it must
ab130f91 336 * handle invalid values gracefully, and use buddy_order_unsafe() below.
48f13bf3 337 */
ab130f91 338static inline unsigned int buddy_order(struct page *page)
48f13bf3 339{
572438f9 340 /* PageBuddy() must be checked by the caller */
48f13bf3
MG
341 return page_private(page);
342}
b5a0e011 343
99c0fd5e 344/*
ab130f91 345 * Like buddy_order(), but for callers who cannot afford to hold the zone lock.
99c0fd5e
VB
346 * PageBuddy() should be checked first by the caller to minimize race window,
347 * and invalid values must be handled gracefully.
348 *
4db0c3c2 349 * READ_ONCE is used so that if the caller assigns the result into a local
99c0fd5e
VB
350 * variable and e.g. tests it for valid range before using, the compiler cannot
351 * decide to remove the variable and inline the page_private(page) multiple
352 * times, potentially observing different values in the tests and the actual
353 * use of the result.
354 */
ab130f91 355#define buddy_order_unsafe(page) READ_ONCE(page_private(page))
99c0fd5e 356
30bdbb78
KK
357/*
358 * These three helpers classifies VMAs for virtual memory accounting.
359 */
360
361/*
362 * Executable code area - executable, not writable, not stack
363 */
d977d56c
KK
364static inline bool is_exec_mapping(vm_flags_t flags)
365{
30bdbb78 366 return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC;
d977d56c
KK
367}
368
30bdbb78 369/*
f0953a1b 370 * Stack area - automatically grows in one direction
30bdbb78
KK
371 *
372 * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
373 * do_mmap() forbids all other combinations.
374 */
d977d56c
KK
375static inline bool is_stack_mapping(vm_flags_t flags)
376{
30bdbb78 377 return (flags & VM_STACK) == VM_STACK;
d977d56c
KK
378}
379
30bdbb78
KK
380/*
381 * Data area - private, writable, not stack
382 */
d977d56c
KK
383static inline bool is_data_mapping(vm_flags_t flags)
384{
30bdbb78 385 return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE;
d977d56c
KK
386}
387
6038def0
NK
388/* mm/util.c */
389void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
aba6dfb7 390 struct vm_area_struct *prev);
1b9fc5b2 391void __vma_unlink_list(struct mm_struct *mm, struct vm_area_struct *vma);
6038def0 392
af8e3354 393#ifdef CONFIG_MMU
3506659e 394void unmap_mapping_folio(struct folio *folio);
fc05f566 395extern long populate_vma_page_range(struct vm_area_struct *vma,
a78f1ccd 396 unsigned long start, unsigned long end, int *locked);
4ca9b385
DH
397extern long faultin_vma_page_range(struct vm_area_struct *vma,
398 unsigned long start, unsigned long end,
399 bool write, int *locked);
af8e3354
HD
400extern void munlock_vma_pages_range(struct vm_area_struct *vma,
401 unsigned long start, unsigned long end);
402static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
403{
404 munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
405}
406
b291f000 407/*
c1e8d7c6 408 * must be called with vma's mmap_lock held for read or write, and page locked.
b291f000
NP
409 */
410extern void mlock_vma_page(struct page *page);
ff6a6da6 411extern unsigned int munlock_vma_page(struct page *page);
b291f000 412
6aeb2542
MR
413extern int mlock_future_check(struct mm_struct *mm, unsigned long flags,
414 unsigned long len);
415
b291f000
NP
416/*
417 * Clear the page's PageMlocked(). This can be useful in a situation where
418 * we want to unconditionally remove a page from the pagecache -- e.g.,
419 * on truncation or freeing.
420 *
421 * It is legal to call this function for any page, mlocked or not.
422 * If called for a page that is still mapped by mlocked vmas, all we do
423 * is revert to lazy LRU behaviour -- semantics are not broken.
424 */
e6c509f8 425extern void clear_page_mlock(struct page *page);
b291f000 426
f55e1014 427extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
b32967ff 428
e9b61f19 429/*
494334e4
HD
430 * At what user virtual address is page expected in vma?
431 * Returns -EFAULT if all of the page is outside the range of vma.
432 * If page is a compound head, the entire compound page is considered.
e9b61f19
KS
433 */
434static inline unsigned long
494334e4 435vma_address(struct page *page, struct vm_area_struct *vma)
e9b61f19 436{
494334e4
HD
437 pgoff_t pgoff;
438 unsigned long address;
439
440 VM_BUG_ON_PAGE(PageKsm(page), page); /* KSM page->index unusable */
441 pgoff = page_to_pgoff(page);
442 if (pgoff >= vma->vm_pgoff) {
443 address = vma->vm_start +
444 ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
445 /* Check for address beyond vma (or wrapped through 0?) */
446 if (address < vma->vm_start || address >= vma->vm_end)
447 address = -EFAULT;
448 } else if (PageHead(page) &&
449 pgoff + compound_nr(page) - 1 >= vma->vm_pgoff) {
450 /* Test above avoids possibility of wrap to 0 on 32-bit */
451 address = vma->vm_start;
452 } else {
453 address = -EFAULT;
454 }
455 return address;
e9b61f19
KS
456}
457
494334e4
HD
458/*
459 * Then at what user virtual address will none of the page be found in vma?
460 * Assumes that vma_address() already returned a good starting address.
461 * If page is a compound head, the entire compound page is considered.
462 */
e9b61f19 463static inline unsigned long
494334e4 464vma_address_end(struct page *page, struct vm_area_struct *vma)
e9b61f19 465{
494334e4
HD
466 pgoff_t pgoff;
467 unsigned long address;
468
469 VM_BUG_ON_PAGE(PageKsm(page), page); /* KSM page->index unusable */
470 pgoff = page_to_pgoff(page) + compound_nr(page);
471 address = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
472 /* Check for address beyond vma (or wrapped through 0?) */
473 if (address < vma->vm_start || address > vma->vm_end)
474 address = vma->vm_end;
475 return address;
e9b61f19
KS
476}
477
89b15332
JW
478static inline struct file *maybe_unlock_mmap_for_io(struct vm_fault *vmf,
479 struct file *fpin)
480{
481 int flags = vmf->flags;
482
483 if (fpin)
484 return fpin;
485
486 /*
487 * FAULT_FLAG_RETRY_NOWAIT means we don't want to wait on page locks or
c1e8d7c6 488 * anything, so we only pin the file and drop the mmap_lock if only
4064b982 489 * FAULT_FLAG_ALLOW_RETRY is set, while this is the first attempt.
89b15332 490 */
4064b982
PX
491 if (fault_flag_allow_retry_first(flags) &&
492 !(flags & FAULT_FLAG_RETRY_NOWAIT)) {
89b15332 493 fpin = get_file(vmf->vma->vm_file);
d8ed45c5 494 mmap_read_unlock(vmf->vma->vm_mm);
89b15332
JW
495 }
496 return fpin;
497}
af8e3354 498#else /* !CONFIG_MMU */
3506659e 499static inline void unmap_mapping_folio(struct folio *folio) { }
b291f000
NP
500static inline void clear_page_mlock(struct page *page) { }
501static inline void mlock_vma_page(struct page *page) { }
4ad0ae8c
NP
502static inline void vunmap_range_noflush(unsigned long start, unsigned long end)
503{
504}
af8e3354 505#endif /* !CONFIG_MMU */
894bc310 506
69d177c2
AW
507/*
508 * Return the mem_map entry representing the 'offset' subpage within
509 * the maximally aligned gigantic page 'base'. Handle any discontiguity
510 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
511 */
512static inline struct page *mem_map_offset(struct page *base, int offset)
513{
514 if (unlikely(offset >= MAX_ORDER_NR_PAGES))
bc7f84c0 515 return nth_page(base, offset);
69d177c2
AW
516 return base + offset;
517}
518
519/*
25985edc 520 * Iterator over all subpages within the maximally aligned gigantic
69d177c2
AW
521 * page 'base'. Handle any discontiguity in the mem_map.
522 */
523static inline struct page *mem_map_next(struct page *iter,
524 struct page *base, int offset)
525{
526 if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
527 unsigned long pfn = page_to_pfn(base) + offset;
528 if (!pfn_valid(pfn))
529 return NULL;
530 return pfn_to_page(pfn);
531 }
532 return iter + 1;
533}
534
6b74ab97
MG
535/* Memory initialisation debug and verification */
536enum mminit_level {
537 MMINIT_WARNING,
538 MMINIT_VERIFY,
539 MMINIT_TRACE
540};
541
542#ifdef CONFIG_DEBUG_MEMORY_INIT
543
544extern int mminit_loglevel;
545
546#define mminit_dprintk(level, prefix, fmt, arg...) \
547do { \
548 if (level < mminit_loglevel) { \
fc5199d1 549 if (level <= MMINIT_WARNING) \
1170532b 550 pr_warn("mminit::" prefix " " fmt, ##arg); \
fc5199d1
RV
551 else \
552 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
6b74ab97
MG
553 } \
554} while (0)
555
708614e6 556extern void mminit_verify_pageflags_layout(void);
68ad8df4 557extern void mminit_verify_zonelist(void);
6b74ab97
MG
558#else
559
560static inline void mminit_dprintk(enum mminit_level level,
561 const char *prefix, const char *fmt, ...)
562{
563}
564
708614e6
MG
565static inline void mminit_verify_pageflags_layout(void)
566{
567}
568
68ad8df4
MG
569static inline void mminit_verify_zonelist(void)
570{
571}
6b74ab97 572#endif /* CONFIG_DEBUG_MEMORY_INIT */
2dbb51c4
MG
573
574/* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
575#if defined(CONFIG_SPARSEMEM)
576extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
577 unsigned long *end_pfn);
578#else
579static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
580 unsigned long *end_pfn)
581{
582}
583#endif /* CONFIG_SPARSEMEM */
584
a5f5f91d
MG
585#define NODE_RECLAIM_NOSCAN -2
586#define NODE_RECLAIM_FULL -1
587#define NODE_RECLAIM_SOME 0
588#define NODE_RECLAIM_SUCCESS 1
7c116f2b 589
8b09549c
WY
590#ifdef CONFIG_NUMA
591extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int);
79c28a41 592extern int find_next_best_node(int node, nodemask_t *used_node_mask);
8b09549c
WY
593#else
594static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask,
595 unsigned int order)
596{
597 return NODE_RECLAIM_NOSCAN;
598}
79c28a41
DH
599static inline int find_next_best_node(int node, nodemask_t *used_node_mask)
600{
601 return NUMA_NO_NODE;
602}
8b09549c
WY
603#endif
604
31d3d348
WF
605extern int hwpoison_filter(struct page *p);
606
7c116f2b
WF
607extern u32 hwpoison_filter_dev_major;
608extern u32 hwpoison_filter_dev_minor;
478c5ffc
WF
609extern u64 hwpoison_filter_flags_mask;
610extern u64 hwpoison_filter_flags_value;
4fd466eb 611extern u64 hwpoison_filter_memcg;
1bfe5feb 612extern u32 hwpoison_filter_enable;
eb36c587 613
dc0ef0df 614extern unsigned long __must_check vm_mmap_pgoff(struct file *, unsigned long,
eb36c587 615 unsigned long, unsigned long,
9fbeb5ab 616 unsigned long, unsigned long);
ca57df79
XQ
617
618extern void set_pageblock_order(void);
730ec8c0 619unsigned int reclaim_clean_pages_from_list(struct zone *zone,
02c6de8d 620 struct list_head *page_list);
d95ea5d1
BZ
621/* The ALLOC_WMARK bits are used as an index to zone->watermark */
622#define ALLOC_WMARK_MIN WMARK_MIN
623#define ALLOC_WMARK_LOW WMARK_LOW
624#define ALLOC_WMARK_HIGH WMARK_HIGH
625#define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
626
627/* Mask to get the watermark bits */
628#define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
629
cd04ae1e
MH
630/*
631 * Only MMU archs have async oom victim reclaim - aka oom_reaper so we
632 * cannot assume a reduced access to memory reserves is sufficient for
633 * !MMU
634 */
635#ifdef CONFIG_MMU
636#define ALLOC_OOM 0x08
637#else
638#define ALLOC_OOM ALLOC_NO_WATERMARKS
639#endif
640
6bb15450
MG
641#define ALLOC_HARDER 0x10 /* try to alloc harder */
642#define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
643#define ALLOC_CPUSET 0x40 /* check for correct cpuset */
644#define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
645#ifdef CONFIG_ZONE_DMA32
646#define ALLOC_NOFRAGMENT 0x100 /* avoid mixing pageblock types */
647#else
648#define ALLOC_NOFRAGMENT 0x0
649#endif
736838e9 650#define ALLOC_KSWAPD 0x800 /* allow waking of kswapd, __GFP_KSWAPD_RECLAIM set */
d95ea5d1 651
72b252ae
MG
652enum ttu_flags;
653struct tlbflush_unmap_batch;
654
ce612879
MH
655
656/*
657 * only for MM internal work items which do not depend on
658 * any allocations or locks which might depend on allocations
659 */
660extern struct workqueue_struct *mm_percpu_wq;
661
72b252ae
MG
662#ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
663void try_to_unmap_flush(void);
d950c947 664void try_to_unmap_flush_dirty(void);
3ea27719 665void flush_tlb_batched_pending(struct mm_struct *mm);
72b252ae
MG
666#else
667static inline void try_to_unmap_flush(void)
668{
669}
d950c947
MG
670static inline void try_to_unmap_flush_dirty(void)
671{
672}
3ea27719
MG
673static inline void flush_tlb_batched_pending(struct mm_struct *mm)
674{
675}
72b252ae 676#endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
edf14cdb
VB
677
678extern const struct trace_print_flags pageflag_names[];
679extern const struct trace_print_flags vmaflag_names[];
680extern const struct trace_print_flags gfpflag_names[];
681
a6ffdc07
XQ
682static inline bool is_migrate_highatomic(enum migratetype migratetype)
683{
684 return migratetype == MIGRATE_HIGHATOMIC;
685}
686
687static inline bool is_migrate_highatomic_page(struct page *page)
688{
689 return get_pageblock_migratetype(page) == MIGRATE_HIGHATOMIC;
690}
691
72675e13 692void setup_zone_pageset(struct zone *zone);
19fc7bed
JK
693
694struct migration_target_control {
695 int nid; /* preferred node id */
696 nodemask_t *nmask;
697 gfp_t gfp_mask;
698};
699
b67177ec
NP
700/*
701 * mm/vmalloc.c
702 */
4ad0ae8c 703#ifdef CONFIG_MMU
b67177ec
NP
704int vmap_pages_range_noflush(unsigned long addr, unsigned long end,
705 pgprot_t prot, struct page **pages, unsigned int page_shift);
4ad0ae8c
NP
706#else
707static inline
708int vmap_pages_range_noflush(unsigned long addr, unsigned long end,
709 pgprot_t prot, struct page **pages, unsigned int page_shift)
710{
711 return -EINVAL;
712}
713#endif
714
715void vunmap_range_noflush(unsigned long start, unsigned long end);
b67177ec 716
f4c0d836
YS
717int numa_migrate_prep(struct page *page, struct vm_area_struct *vma,
718 unsigned long addr, int page_nid, int *flags);
719
db971418 720#endif /* __MM_INTERNAL_H */