Merge tag 'timers-urgent-2021-11-14' of git://git.kernel.org/pub/scm/linux/kernel...
[linux-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
aac45363
MH
77void unmap_page_range(struct mmu_gather *tlb,
78 struct vm_area_struct *vma,
79 unsigned long addr, unsigned long end,
80 struct zap_details *details);
81
7b3df3b9
DH
82void do_page_cache_ra(struct readahead_control *, unsigned long nr_to_read,
83 unsigned long lookahead_size);
fcd9ae4f 84void force_page_cache_ra(struct readahead_control *, unsigned long nr);
7b3df3b9
DH
85static inline void force_page_cache_readahead(struct address_space *mapping,
86 struct file *file, pgoff_t index, unsigned long nr_to_read)
87{
fcd9ae4f
MWO
88 DEFINE_READAHEAD(ractl, file, &file->f_ra, mapping, index);
89 force_page_cache_ra(&ractl, nr_to_read);
7b3df3b9 90}
29f175d1 91
5c211ba2
MWO
92unsigned find_lock_entries(struct address_space *mapping, pgoff_t start,
93 pgoff_t end, struct pagevec *pvec, pgoff_t *indices);
94
1eb6234e 95/**
3eed3ef5
MWO
96 * folio_evictable - Test whether a folio is evictable.
97 * @folio: The folio to test.
1eb6234e 98 *
3eed3ef5
MWO
99 * Test whether @folio is evictable -- i.e., should be placed on
100 * active/inactive lists vs unevictable list.
1eb6234e 101 *
3eed3ef5
MWO
102 * Reasons folio might not be evictable:
103 * 1. folio's mapping marked unevictable
104 * 2. One of the pages in the folio is part of an mlocked VMA
1eb6234e 105 */
3eed3ef5
MWO
106static inline bool folio_evictable(struct folio *folio)
107{
108 bool ret;
109
110 /* Prevent address_space of inode and swap cache from being freed */
111 rcu_read_lock();
112 ret = !mapping_unevictable(folio_mapping(folio)) &&
113 !folio_test_mlocked(folio);
114 rcu_read_unlock();
115 return ret;
116}
117
1eb6234e
YS
118static inline bool page_evictable(struct page *page)
119{
120 bool ret;
121
122 /* Prevent address_space of inode and swap cache from being freed */
123 rcu_read_lock();
124 ret = !mapping_unevictable(page_mapping(page)) && !PageMlocked(page);
125 rcu_read_unlock();
126 return ret;
127}
128
7835e98b 129/*
0139aa7b 130 * Turn a non-refcounted page (->_refcount == 0) into refcounted with
7835e98b
NP
131 * a count of one.
132 */
133static inline void set_page_refcounted(struct page *page)
134{
309381fe 135 VM_BUG_ON_PAGE(PageTail(page), page);
fe896d18 136 VM_BUG_ON_PAGE(page_ref_count(page), page);
77a8a788 137 set_page_count(page, 1);
77a8a788
NP
138}
139
03f6462a
HD
140extern unsigned long highest_memmap_pfn;
141
c73322d0
JW
142/*
143 * Maximum number of reclaim retries without progress before the OOM
144 * killer is consider the only way forward.
145 */
146#define MAX_RECLAIM_RETRIES 16
147
894bc310
LS
148/*
149 * in mm/vmscan.c:
150 */
62695a84 151extern int isolate_lru_page(struct page *page);
894bc310 152extern void putback_lru_page(struct page *page);
c3f4a9a2 153extern void reclaim_throttle(pg_data_t *pgdat, enum vmscan_throttle_state reason);
62695a84 154
6219049a
BL
155/*
156 * in mm/rmap.c:
157 */
158extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
159
494c1dfe
WL
160/*
161 * in mm/memcontrol.c:
162 */
163extern bool cgroup_memory_nokmem;
164
894bc310
LS
165/*
166 * in mm/page_alloc.c
167 */
3c605096 168
1a6d53a1
VB
169/*
170 * Structure for holding the mostly immutable allocation parameters passed
171 * between functions involved in allocations, including the alloc_pages*
172 * family of functions.
173 *
97a225e6 174 * nodemask, migratetype and highest_zoneidx are initialized only once in
84172f4b 175 * __alloc_pages() and then never change.
1a6d53a1 176 *
97a225e6 177 * zonelist, preferred_zone and highest_zoneidx are set first in
84172f4b 178 * __alloc_pages() for the fast path, and might be later changed
68956ccb 179 * in __alloc_pages_slowpath(). All other functions pass the whole structure
1a6d53a1
VB
180 * by a const pointer.
181 */
182struct alloc_context {
183 struct zonelist *zonelist;
184 nodemask_t *nodemask;
c33d6c06 185 struct zoneref *preferred_zoneref;
1a6d53a1 186 int migratetype;
97a225e6
JK
187
188 /*
189 * highest_zoneidx represents highest usable zone index of
190 * the allocation request. Due to the nature of the zone,
191 * memory on lower zone than the highest_zoneidx will be
192 * protected by lowmem_reserve[highest_zoneidx].
193 *
194 * highest_zoneidx is also used by reclaim/compaction to limit
195 * the target zone since higher zone than this index cannot be
196 * usable for this allocation request.
197 */
198 enum zone_type highest_zoneidx;
c9ab0c4f 199 bool spread_dirty_pages;
1a6d53a1
VB
200};
201
3c605096
JK
202/*
203 * Locate the struct page for both the matching buddy in our
204 * pair (buddy1) and the combined O(n+1) page they form (page).
205 *
206 * 1) Any buddy B1 will have an order O twin B2 which satisfies
207 * the following equation:
208 * B2 = B1 ^ (1 << O)
209 * For example, if the starting buddy (buddy2) is #8 its order
210 * 1 buddy is #10:
211 * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
212 *
213 * 2) Any buddy B will have an order O+1 parent P which
214 * satisfies the following equation:
215 * P = B & ~(1 << O)
216 *
217 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
218 */
219static inline unsigned long
76741e77 220__find_buddy_pfn(unsigned long page_pfn, unsigned int order)
3c605096 221{
76741e77 222 return page_pfn ^ (1 << order);
3c605096
JK
223}
224
7cf91a98
JK
225extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn,
226 unsigned long end_pfn, struct zone *zone);
227
228static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn,
229 unsigned long end_pfn, struct zone *zone)
230{
231 if (zone->contiguous)
232 return pfn_to_page(start_pfn);
233
234 return __pageblock_pfn_to_page(start_pfn, end_pfn, zone);
235}
236
3c605096 237extern int __isolate_free_page(struct page *page, unsigned int order);
624f58d8
AD
238extern void __putback_isolated_page(struct page *page, unsigned int order,
239 int mt);
7c2ee349 240extern void memblock_free_pages(struct page *page, unsigned long pfn,
d70ddd7a 241 unsigned int order);
a9cd410a 242extern void __free_pages_core(struct page *page, unsigned int order);
d00181b9 243extern void prep_compound_page(struct page *page, unsigned int order);
46f24fd8
JK
244extern void post_alloc_hook(struct page *page, unsigned int order,
245 gfp_t gfp_flags);
42aa83cb 246extern int user_min_free_kbytes;
20a0307c 247
44042b44 248extern void free_unref_page(struct page *page, unsigned int order);
0966aeb4
MWO
249extern void free_unref_page_list(struct list_head *list);
250
04f8cfea 251extern void zone_pcp_update(struct zone *zone, int cpu_online);
68265390 252extern void zone_pcp_reset(struct zone *zone);
ec6e8c7e
VB
253extern void zone_pcp_disable(struct zone *zone);
254extern void zone_pcp_enable(struct zone *zone);
68265390 255
c803b3c8
MR
256extern void *memmap_alloc(phys_addr_t size, phys_addr_t align,
257 phys_addr_t min_addr,
258 int nid, bool exact_nid);
259
ff9543fd
MN
260#if defined CONFIG_COMPACTION || defined CONFIG_CMA
261
262/*
263 * in mm/compaction.c
264 */
265/*
266 * compact_control is used to track pages being migrated and the free pages
267 * they are being migrated to during memory compaction. The free_pfn starts
268 * at the end of a zone and migrate_pfn begins at the start. Movable pages
269 * are moved to the end of a zone during a compaction run and the run
270 * completes when free_pfn <= migrate_pfn
271 */
272struct compact_control {
273 struct list_head freepages; /* List of free pages to migrate to */
274 struct list_head migratepages; /* List of pages being migrated */
c5fbd937
MG
275 unsigned int nr_freepages; /* Number of isolated free pages */
276 unsigned int nr_migratepages; /* Number of pages to migrate */
ff9543fd 277 unsigned long free_pfn; /* isolate_freepages search base */
c2ad7a1f
OS
278 /*
279 * Acts as an in/out parameter to page isolation for migration.
280 * isolate_migratepages uses it as a search base.
281 * isolate_migratepages_block will update the value to the next pfn
282 * after the last isolated one.
283 */
284 unsigned long migrate_pfn;
70b44595 285 unsigned long fast_start_pfn; /* a pfn to start linear scan from */
c5943b9c
MG
286 struct zone *zone;
287 unsigned long total_migrate_scanned;
288 unsigned long total_free_scanned;
dbe2d4e4
MG
289 unsigned short fast_search_fail;/* failures to use free list searches */
290 short search_order; /* order to start a fast search at */
f25ba6dc
VB
291 const gfp_t gfp_mask; /* gfp mask of a direct compactor */
292 int order; /* order a direct compactor needs */
d39773a0 293 int migratetype; /* migratetype of direct compactor */
f25ba6dc 294 const unsigned int alloc_flags; /* alloc flags of a direct compactor */
97a225e6 295 const int highest_zoneidx; /* zone index of a direct compactor */
e0b9daeb 296 enum migrate_mode mode; /* Async or sync migration mode */
bb13ffeb 297 bool ignore_skip_hint; /* Scan blocks even if marked skip */
2583d671 298 bool no_set_skip_hint; /* Don't mark blocks for skipping */
9f7e3387 299 bool ignore_block_suitable; /* Scan blocks considered unsuitable */
accf6242 300 bool direct_compaction; /* False from kcompactd or /proc/... */
facdaa91 301 bool proactive_compaction; /* kcompactd proactive compaction */
06ed2998 302 bool whole_zone; /* Whole zone should/has been scanned */
c3486f53 303 bool contended; /* Signal lock or sched contention */
804d3121 304 bool rescan; /* Rescanning the same pageblock */
b06eda09 305 bool alloc_contig; /* alloc_contig_range allocation */
ff9543fd
MN
306};
307
5e1f0f09
MG
308/*
309 * Used in direct compaction when a page should be taken from the freelists
310 * immediately when one is created during the free path.
311 */
312struct capture_control {
313 struct compact_control *cc;
314 struct page *page;
315};
316
ff9543fd 317unsigned long
bb13ffeb
MG
318isolate_freepages_range(struct compact_control *cc,
319 unsigned long start_pfn, unsigned long end_pfn);
c2ad7a1f 320int
edc2ca61
VB
321isolate_migratepages_range(struct compact_control *cc,
322 unsigned long low_pfn, unsigned long end_pfn);
ffd8f251 323#endif
2149cdae
JK
324int find_suitable_fallback(struct free_area *area, unsigned int order,
325 int migratetype, bool only_stealable, bool *can_steal);
ff9543fd 326
48f13bf3 327/*
6c14466c
MG
328 * This function returns the order of a free page in the buddy system. In
329 * general, page_zone(page)->lock must be held by the caller to prevent the
330 * page from being allocated in parallel and returning garbage as the order.
331 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
99c0fd5e 332 * page cannot be allocated or merged in parallel. Alternatively, it must
ab130f91 333 * handle invalid values gracefully, and use buddy_order_unsafe() below.
48f13bf3 334 */
ab130f91 335static inline unsigned int buddy_order(struct page *page)
48f13bf3 336{
572438f9 337 /* PageBuddy() must be checked by the caller */
48f13bf3
MG
338 return page_private(page);
339}
b5a0e011 340
99c0fd5e 341/*
ab130f91 342 * Like buddy_order(), but for callers who cannot afford to hold the zone lock.
99c0fd5e
VB
343 * PageBuddy() should be checked first by the caller to minimize race window,
344 * and invalid values must be handled gracefully.
345 *
4db0c3c2 346 * READ_ONCE is used so that if the caller assigns the result into a local
99c0fd5e
VB
347 * variable and e.g. tests it for valid range before using, the compiler cannot
348 * decide to remove the variable and inline the page_private(page) multiple
349 * times, potentially observing different values in the tests and the actual
350 * use of the result.
351 */
ab130f91 352#define buddy_order_unsafe(page) READ_ONCE(page_private(page))
99c0fd5e 353
30bdbb78
KK
354/*
355 * These three helpers classifies VMAs for virtual memory accounting.
356 */
357
358/*
359 * Executable code area - executable, not writable, not stack
360 */
d977d56c
KK
361static inline bool is_exec_mapping(vm_flags_t flags)
362{
30bdbb78 363 return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC;
d977d56c
KK
364}
365
30bdbb78 366/*
f0953a1b 367 * Stack area - automatically grows in one direction
30bdbb78
KK
368 *
369 * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
370 * do_mmap() forbids all other combinations.
371 */
d977d56c
KK
372static inline bool is_stack_mapping(vm_flags_t flags)
373{
30bdbb78 374 return (flags & VM_STACK) == VM_STACK;
d977d56c
KK
375}
376
30bdbb78
KK
377/*
378 * Data area - private, writable, not stack
379 */
d977d56c
KK
380static inline bool is_data_mapping(vm_flags_t flags)
381{
30bdbb78 382 return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE;
d977d56c
KK
383}
384
6038def0
NK
385/* mm/util.c */
386void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
aba6dfb7 387 struct vm_area_struct *prev);
1b9fc5b2 388void __vma_unlink_list(struct mm_struct *mm, struct vm_area_struct *vma);
6038def0 389
af8e3354 390#ifdef CONFIG_MMU
fc05f566 391extern long populate_vma_page_range(struct vm_area_struct *vma,
a78f1ccd 392 unsigned long start, unsigned long end, int *locked);
4ca9b385
DH
393extern long faultin_vma_page_range(struct vm_area_struct *vma,
394 unsigned long start, unsigned long end,
395 bool write, int *locked);
af8e3354
HD
396extern void munlock_vma_pages_range(struct vm_area_struct *vma,
397 unsigned long start, unsigned long end);
398static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
399{
400 munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
401}
402
b291f000 403/*
c1e8d7c6 404 * must be called with vma's mmap_lock held for read or write, and page locked.
b291f000
NP
405 */
406extern void mlock_vma_page(struct page *page);
ff6a6da6 407extern unsigned int munlock_vma_page(struct page *page);
b291f000 408
6aeb2542
MR
409extern int mlock_future_check(struct mm_struct *mm, unsigned long flags,
410 unsigned long len);
411
b291f000
NP
412/*
413 * Clear the page's PageMlocked(). This can be useful in a situation where
414 * we want to unconditionally remove a page from the pagecache -- e.g.,
415 * on truncation or freeing.
416 *
417 * It is legal to call this function for any page, mlocked or not.
418 * If called for a page that is still mapped by mlocked vmas, all we do
419 * is revert to lazy LRU behaviour -- semantics are not broken.
420 */
e6c509f8 421extern void clear_page_mlock(struct page *page);
b291f000 422
f55e1014 423extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
b32967ff 424
e9b61f19 425/*
494334e4
HD
426 * At what user virtual address is page expected in vma?
427 * Returns -EFAULT if all of the page is outside the range of vma.
428 * If page is a compound head, the entire compound page is considered.
e9b61f19
KS
429 */
430static inline unsigned long
494334e4 431vma_address(struct page *page, struct vm_area_struct *vma)
e9b61f19 432{
494334e4
HD
433 pgoff_t pgoff;
434 unsigned long address;
435
436 VM_BUG_ON_PAGE(PageKsm(page), page); /* KSM page->index unusable */
437 pgoff = page_to_pgoff(page);
438 if (pgoff >= vma->vm_pgoff) {
439 address = vma->vm_start +
440 ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
441 /* Check for address beyond vma (or wrapped through 0?) */
442 if (address < vma->vm_start || address >= vma->vm_end)
443 address = -EFAULT;
444 } else if (PageHead(page) &&
445 pgoff + compound_nr(page) - 1 >= vma->vm_pgoff) {
446 /* Test above avoids possibility of wrap to 0 on 32-bit */
447 address = vma->vm_start;
448 } else {
449 address = -EFAULT;
450 }
451 return address;
e9b61f19
KS
452}
453
494334e4
HD
454/*
455 * Then at what user virtual address will none of the page be found in vma?
456 * Assumes that vma_address() already returned a good starting address.
457 * If page is a compound head, the entire compound page is considered.
458 */
e9b61f19 459static inline unsigned long
494334e4 460vma_address_end(struct page *page, struct vm_area_struct *vma)
e9b61f19 461{
494334e4
HD
462 pgoff_t pgoff;
463 unsigned long address;
464
465 VM_BUG_ON_PAGE(PageKsm(page), page); /* KSM page->index unusable */
466 pgoff = page_to_pgoff(page) + compound_nr(page);
467 address = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
468 /* Check for address beyond vma (or wrapped through 0?) */
469 if (address < vma->vm_start || address > vma->vm_end)
470 address = vma->vm_end;
471 return address;
e9b61f19
KS
472}
473
89b15332
JW
474static inline struct file *maybe_unlock_mmap_for_io(struct vm_fault *vmf,
475 struct file *fpin)
476{
477 int flags = vmf->flags;
478
479 if (fpin)
480 return fpin;
481
482 /*
483 * FAULT_FLAG_RETRY_NOWAIT means we don't want to wait on page locks or
c1e8d7c6 484 * anything, so we only pin the file and drop the mmap_lock if only
4064b982 485 * FAULT_FLAG_ALLOW_RETRY is set, while this is the first attempt.
89b15332 486 */
4064b982
PX
487 if (fault_flag_allow_retry_first(flags) &&
488 !(flags & FAULT_FLAG_RETRY_NOWAIT)) {
89b15332 489 fpin = get_file(vmf->vma->vm_file);
d8ed45c5 490 mmap_read_unlock(vmf->vma->vm_mm);
89b15332
JW
491 }
492 return fpin;
493}
494
af8e3354 495#else /* !CONFIG_MMU */
b291f000
NP
496static inline void clear_page_mlock(struct page *page) { }
497static inline void mlock_vma_page(struct page *page) { }
4ad0ae8c
NP
498static inline void vunmap_range_noflush(unsigned long start, unsigned long end)
499{
500}
af8e3354 501#endif /* !CONFIG_MMU */
894bc310 502
69d177c2
AW
503/*
504 * Return the mem_map entry representing the 'offset' subpage within
505 * the maximally aligned gigantic page 'base'. Handle any discontiguity
506 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
507 */
508static inline struct page *mem_map_offset(struct page *base, int offset)
509{
510 if (unlikely(offset >= MAX_ORDER_NR_PAGES))
bc7f84c0 511 return nth_page(base, offset);
69d177c2
AW
512 return base + offset;
513}
514
515/*
25985edc 516 * Iterator over all subpages within the maximally aligned gigantic
69d177c2
AW
517 * page 'base'. Handle any discontiguity in the mem_map.
518 */
519static inline struct page *mem_map_next(struct page *iter,
520 struct page *base, int offset)
521{
522 if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
523 unsigned long pfn = page_to_pfn(base) + offset;
524 if (!pfn_valid(pfn))
525 return NULL;
526 return pfn_to_page(pfn);
527 }
528 return iter + 1;
529}
530
6b74ab97
MG
531/* Memory initialisation debug and verification */
532enum mminit_level {
533 MMINIT_WARNING,
534 MMINIT_VERIFY,
535 MMINIT_TRACE
536};
537
538#ifdef CONFIG_DEBUG_MEMORY_INIT
539
540extern int mminit_loglevel;
541
542#define mminit_dprintk(level, prefix, fmt, arg...) \
543do { \
544 if (level < mminit_loglevel) { \
fc5199d1 545 if (level <= MMINIT_WARNING) \
1170532b 546 pr_warn("mminit::" prefix " " fmt, ##arg); \
fc5199d1
RV
547 else \
548 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
6b74ab97
MG
549 } \
550} while (0)
551
708614e6 552extern void mminit_verify_pageflags_layout(void);
68ad8df4 553extern void mminit_verify_zonelist(void);
6b74ab97
MG
554#else
555
556static inline void mminit_dprintk(enum mminit_level level,
557 const char *prefix, const char *fmt, ...)
558{
559}
560
708614e6
MG
561static inline void mminit_verify_pageflags_layout(void)
562{
563}
564
68ad8df4
MG
565static inline void mminit_verify_zonelist(void)
566{
567}
6b74ab97 568#endif /* CONFIG_DEBUG_MEMORY_INIT */
2dbb51c4
MG
569
570/* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
571#if defined(CONFIG_SPARSEMEM)
572extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
573 unsigned long *end_pfn);
574#else
575static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
576 unsigned long *end_pfn)
577{
578}
579#endif /* CONFIG_SPARSEMEM */
580
a5f5f91d
MG
581#define NODE_RECLAIM_NOSCAN -2
582#define NODE_RECLAIM_FULL -1
583#define NODE_RECLAIM_SOME 0
584#define NODE_RECLAIM_SUCCESS 1
7c116f2b 585
8b09549c
WY
586#ifdef CONFIG_NUMA
587extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int);
79c28a41 588extern int find_next_best_node(int node, nodemask_t *used_node_mask);
8b09549c
WY
589#else
590static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask,
591 unsigned int order)
592{
593 return NODE_RECLAIM_NOSCAN;
594}
79c28a41
DH
595static inline int find_next_best_node(int node, nodemask_t *used_node_mask)
596{
597 return NUMA_NO_NODE;
598}
8b09549c
WY
599#endif
600
31d3d348
WF
601extern int hwpoison_filter(struct page *p);
602
7c116f2b
WF
603extern u32 hwpoison_filter_dev_major;
604extern u32 hwpoison_filter_dev_minor;
478c5ffc
WF
605extern u64 hwpoison_filter_flags_mask;
606extern u64 hwpoison_filter_flags_value;
4fd466eb 607extern u64 hwpoison_filter_memcg;
1bfe5feb 608extern u32 hwpoison_filter_enable;
eb36c587 609
dc0ef0df 610extern unsigned long __must_check vm_mmap_pgoff(struct file *, unsigned long,
eb36c587 611 unsigned long, unsigned long,
9fbeb5ab 612 unsigned long, unsigned long);
ca57df79
XQ
613
614extern void set_pageblock_order(void);
730ec8c0 615unsigned int reclaim_clean_pages_from_list(struct zone *zone,
02c6de8d 616 struct list_head *page_list);
d95ea5d1
BZ
617/* The ALLOC_WMARK bits are used as an index to zone->watermark */
618#define ALLOC_WMARK_MIN WMARK_MIN
619#define ALLOC_WMARK_LOW WMARK_LOW
620#define ALLOC_WMARK_HIGH WMARK_HIGH
621#define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
622
623/* Mask to get the watermark bits */
624#define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
625
cd04ae1e
MH
626/*
627 * Only MMU archs have async oom victim reclaim - aka oom_reaper so we
628 * cannot assume a reduced access to memory reserves is sufficient for
629 * !MMU
630 */
631#ifdef CONFIG_MMU
632#define ALLOC_OOM 0x08
633#else
634#define ALLOC_OOM ALLOC_NO_WATERMARKS
635#endif
636
6bb15450
MG
637#define ALLOC_HARDER 0x10 /* try to alloc harder */
638#define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
639#define ALLOC_CPUSET 0x40 /* check for correct cpuset */
640#define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
641#ifdef CONFIG_ZONE_DMA32
642#define ALLOC_NOFRAGMENT 0x100 /* avoid mixing pageblock types */
643#else
644#define ALLOC_NOFRAGMENT 0x0
645#endif
736838e9 646#define ALLOC_KSWAPD 0x800 /* allow waking of kswapd, __GFP_KSWAPD_RECLAIM set */
d95ea5d1 647
72b252ae
MG
648enum ttu_flags;
649struct tlbflush_unmap_batch;
650
ce612879
MH
651
652/*
653 * only for MM internal work items which do not depend on
654 * any allocations or locks which might depend on allocations
655 */
656extern struct workqueue_struct *mm_percpu_wq;
657
72b252ae
MG
658#ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
659void try_to_unmap_flush(void);
d950c947 660void try_to_unmap_flush_dirty(void);
3ea27719 661void flush_tlb_batched_pending(struct mm_struct *mm);
72b252ae
MG
662#else
663static inline void try_to_unmap_flush(void)
664{
665}
d950c947
MG
666static inline void try_to_unmap_flush_dirty(void)
667{
668}
3ea27719
MG
669static inline void flush_tlb_batched_pending(struct mm_struct *mm)
670{
671}
72b252ae 672#endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
edf14cdb
VB
673
674extern const struct trace_print_flags pageflag_names[];
675extern const struct trace_print_flags vmaflag_names[];
676extern const struct trace_print_flags gfpflag_names[];
677
a6ffdc07
XQ
678static inline bool is_migrate_highatomic(enum migratetype migratetype)
679{
680 return migratetype == MIGRATE_HIGHATOMIC;
681}
682
683static inline bool is_migrate_highatomic_page(struct page *page)
684{
685 return get_pageblock_migratetype(page) == MIGRATE_HIGHATOMIC;
686}
687
72675e13 688void setup_zone_pageset(struct zone *zone);
19fc7bed
JK
689
690struct migration_target_control {
691 int nid; /* preferred node id */
692 nodemask_t *nmask;
693 gfp_t gfp_mask;
694};
695
b67177ec
NP
696/*
697 * mm/vmalloc.c
698 */
4ad0ae8c 699#ifdef CONFIG_MMU
b67177ec
NP
700int vmap_pages_range_noflush(unsigned long addr, unsigned long end,
701 pgprot_t prot, struct page **pages, unsigned int page_shift);
4ad0ae8c
NP
702#else
703static inline
704int vmap_pages_range_noflush(unsigned long addr, unsigned long end,
705 pgprot_t prot, struct page **pages, unsigned int page_shift)
706{
707 return -EINVAL;
708}
709#endif
710
711void vunmap_range_noflush(unsigned long start, unsigned long end);
b67177ec 712
f4c0d836
YS
713int numa_migrate_prep(struct page *page, struct vm_area_struct *vma,
714 unsigned long addr, int page_nid, int *flags);
715
db971418 716#endif /* __MM_INTERNAL_H */