blob: e6bd35182daee1226b684464fd202df86762c636 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/* 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 */
Nick Piggin0f8053a2006-03-22 00:08:33 -080011#ifndef __MM_INTERNAL_H
12#define __MM_INTERNAL_H
13
Fabian Frederick29f175d2014-04-07 15:37:55 -070014#include <linux/fs.h>
Nick Piggin0f8053a2006-03-22 00:08:33 -080015#include <linux/mm.h>
Kirill A. Shutemove9b61f12016-01-15 16:54:10 -080016#include <linux/pagemap.h>
Vlastimil Babkaedf14cd2016-03-15 14:55:56 -070017#include <linux/tracepoint-defs.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070018
Mel Gormandd56b042015-11-06 16:28:43 -080019/*
20 * The set of flags that only affect watermark checking and reclaim
21 * behaviour. This is used by the MM to obey the caller constraints
22 * about IO, FS and watermark checking while ignoring placement
23 * hints such as HIGHMEM usage.
24 */
25#define GFP_RECLAIM_MASK (__GFP_RECLAIM|__GFP_HIGH|__GFP_IO|__GFP_FS|\
Michal Hockodcda9b02017-07-12 14:36:45 -070026 __GFP_NOWARN|__GFP_RETRY_MAYFAIL|__GFP_NOFAIL|\
Mel Gormane838a452016-06-24 14:49:37 -070027 __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC|\
28 __GFP_ATOMIC)
Mel Gormandd56b042015-11-06 16:28:43 -080029
30/* The GFP flags allowed during early boot */
31#define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_RECLAIM|__GFP_IO|__GFP_FS))
32
33/* Control allocation cpuset and node placement constraints */
34#define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE)
35
36/* Do not use these with a slab allocator */
37#define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
38
Nicholas Piggin62906022016-12-25 13:00:30 +100039void page_writeback_init(void);
40
Jan Kara29943022016-12-14 15:07:16 -080041int do_swap_page(struct vm_fault *vmf);
Ebru Akagunduz8a966ed2016-07-26 15:25:03 -070042
Jan Beulich42b77722008-07-23 21:27:10 -070043void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
44 unsigned long floor, unsigned long ceiling);
45
Kirill A. Shutemov23519072017-02-22 15:46:39 -080046static inline bool can_madv_dontneed_vma(struct vm_area_struct *vma)
47{
48 return !(vma->vm_flags & (VM_LOCKED|VM_HUGETLB|VM_PFNMAP));
49}
50
Michal Hockoaac45362016-03-25 14:20:24 -070051void unmap_page_range(struct mmu_gather *tlb,
52 struct vm_area_struct *vma,
53 unsigned long addr, unsigned long end,
54 struct zap_details *details);
55
Fabian Frederick29f175d2014-04-07 15:37:55 -070056extern int __do_page_cache_readahead(struct address_space *mapping,
57 struct file *filp, pgoff_t offset, unsigned long nr_to_read,
58 unsigned long lookahead_size);
59
60/*
61 * Submit IO for the read-ahead request in file_ra_state.
62 */
63static inline unsigned long ra_submit(struct file_ra_state *ra,
64 struct address_space *mapping, struct file *filp)
65{
66 return __do_page_cache_readahead(mapping, filp,
67 ra->start, ra->size, ra->async_size);
68}
69
Nick Piggin7835e982006-03-22 00:08:40 -080070/*
Joonsoo Kim0139aa72016-05-19 17:10:49 -070071 * Turn a non-refcounted page (->_refcount == 0) into refcounted with
Nick Piggin7835e982006-03-22 00:08:40 -080072 * a count of one.
73 */
74static inline void set_page_refcounted(struct page *page)
75{
Sasha Levin309381fea2014-01-23 15:52:54 -080076 VM_BUG_ON_PAGE(PageTail(page), page);
Joonsoo Kimfe896d12016-03-17 14:19:26 -070077 VM_BUG_ON_PAGE(page_ref_count(page), page);
Nick Piggin77a8a782006-01-06 00:10:57 -080078 set_page_count(page, 1);
Nick Piggin77a8a782006-01-06 00:10:57 -080079}
80
Hugh Dickins03f64622009-09-21 17:03:35 -070081extern unsigned long highest_memmap_pfn;
82
Lee Schermerhorn894bc312008-10-18 20:26:39 -070083/*
Johannes Weinerc73322d2017-05-03 14:51:51 -070084 * Maximum number of reclaim retries without progress before the OOM
85 * killer is consider the only way forward.
86 */
87#define MAX_RECLAIM_RETRIES 16
88
89/*
Lee Schermerhorn894bc312008-10-18 20:26:39 -070090 * in mm/vmscan.c:
91 */
Nick Piggin62695a82008-10-18 20:26:09 -070092extern int isolate_lru_page(struct page *page);
Lee Schermerhorn894bc312008-10-18 20:26:39 -070093extern void putback_lru_page(struct page *page);
Nick Piggin62695a82008-10-18 20:26:09 -070094
Lee Schermerhorn894bc312008-10-18 20:26:39 -070095/*
Bob Liu62190492012-12-11 16:00:37 -080096 * in mm/rmap.c:
97 */
98extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
99
100/*
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700101 * in mm/page_alloc.c
102 */
Joonsoo Kim3c605092014-11-13 15:19:21 -0800103
104/*
Vlastimil Babka1a6d53a2015-02-11 15:25:44 -0800105 * Structure for holding the mostly immutable allocation parameters passed
106 * between functions involved in allocations, including the alloc_pages*
107 * family of functions.
108 *
109 * nodemask, migratetype and high_zoneidx are initialized only once in
110 * __alloc_pages_nodemask() and then never change.
111 *
112 * zonelist, preferred_zone and classzone_idx are set first in
113 * __alloc_pages_nodemask() for the fast path, and might be later changed
114 * in __alloc_pages_slowpath(). All other functions pass the whole strucure
115 * by a const pointer.
116 */
117struct alloc_context {
118 struct zonelist *zonelist;
119 nodemask_t *nodemask;
Mel Gormanc33d6c02016-05-19 17:14:10 -0700120 struct zoneref *preferred_zoneref;
Vlastimil Babka1a6d53a2015-02-11 15:25:44 -0800121 int migratetype;
122 enum zone_type high_zoneidx;
Mel Gormanc9ab0c42015-11-06 16:28:12 -0800123 bool spread_dirty_pages;
Vlastimil Babka1a6d53a2015-02-11 15:25:44 -0800124};
125
Mel Gorman93ea9962016-05-19 17:14:13 -0700126#define ac_classzone_idx(ac) zonelist_zone_idx(ac->preferred_zoneref)
127
Vlastimil Babka1a6d53a2015-02-11 15:25:44 -0800128/*
Joonsoo Kim3c605092014-11-13 15:19:21 -0800129 * Locate the struct page for both the matching buddy in our
130 * pair (buddy1) and the combined O(n+1) page they form (page).
131 *
132 * 1) Any buddy B1 will have an order O twin B2 which satisfies
133 * the following equation:
134 * B2 = B1 ^ (1 << O)
135 * For example, if the starting buddy (buddy2) is #8 its order
136 * 1 buddy is #10:
137 * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
138 *
139 * 2) Any buddy B will have an order O+1 parent P which
140 * satisfies the following equation:
141 * P = B & ~(1 << O)
142 *
143 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
144 */
145static inline unsigned long
Vlastimil Babka76741e72017-02-22 15:41:48 -0800146__find_buddy_pfn(unsigned long page_pfn, unsigned int order)
Joonsoo Kim3c605092014-11-13 15:19:21 -0800147{
Vlastimil Babka76741e72017-02-22 15:41:48 -0800148 return page_pfn ^ (1 << order);
Joonsoo Kim3c605092014-11-13 15:19:21 -0800149}
150
Joonsoo Kim7cf91a92016-03-15 14:57:51 -0700151extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn,
152 unsigned long end_pfn, struct zone *zone);
153
154static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn,
155 unsigned long end_pfn, struct zone *zone)
156{
157 if (zone->contiguous)
158 return pfn_to_page(start_pfn);
159
160 return __pageblock_pfn_to_page(start_pfn, end_pfn, zone);
161}
162
Joonsoo Kim3c605092014-11-13 15:19:21 -0800163extern int __isolate_free_page(struct page *page, unsigned int order);
Mel Gormand70ddd72015-06-30 14:56:52 -0700164extern void __free_pages_bootmem(struct page *page, unsigned long pfn,
165 unsigned int order);
Kirill A. Shutemovd00181b2015-11-06 16:29:57 -0800166extern void prep_compound_page(struct page *page, unsigned int order);
Joonsoo Kim46f24fd2016-07-26 15:23:58 -0700167extern void post_alloc_hook(struct page *page, unsigned int order,
168 gfp_t gfp_flags);
Han Pingtian42aa83c2014-01-23 15:53:28 -0800169extern int user_min_free_kbytes;
Wu Fengguang20a03072009-06-16 15:32:22 -0700170
Michal Nazarewiczff9543f2011-12-29 13:09:50 +0100171#if defined CONFIG_COMPACTION || defined CONFIG_CMA
172
173/*
174 * in mm/compaction.c
175 */
176/*
177 * compact_control is used to track pages being migrated and the free pages
178 * they are being migrated to during memory compaction. The free_pfn starts
179 * at the end of a zone and migrate_pfn begins at the start. Movable pages
180 * are moved to the end of a zone during a compaction run and the run
181 * completes when free_pfn <= migrate_pfn
182 */
183struct compact_control {
184 struct list_head freepages; /* List of free pages to migrate to */
185 struct list_head migratepages; /* List of pages being migrated */
Vlastimil Babkaf25ba6d2017-05-08 15:54:30 -0700186 struct zone *zone;
Michal Nazarewiczff9543f2011-12-29 13:09:50 +0100187 unsigned long nr_freepages; /* Number of isolated free pages */
188 unsigned long nr_migratepages; /* Number of pages to migrate */
David Rientjes7f354a52017-02-22 15:44:50 -0800189 unsigned long total_migrate_scanned;
190 unsigned long total_free_scanned;
Michal Nazarewiczff9543f2011-12-29 13:09:50 +0100191 unsigned long free_pfn; /* isolate_freepages search base */
192 unsigned long migrate_pfn; /* isolate_migratepages search base */
Joonsoo Kim1a167182015-09-08 15:03:59 -0700193 unsigned long last_migrated_pfn;/* Not yet flushed page being freed */
Vlastimil Babkaf25ba6d2017-05-08 15:54:30 -0700194 const gfp_t gfp_mask; /* gfp mask of a direct compactor */
195 int order; /* order a direct compactor needs */
Vlastimil Babkad39773a2017-05-08 15:54:46 -0700196 int migratetype; /* migratetype of direct compactor */
Vlastimil Babkaf25ba6d2017-05-08 15:54:30 -0700197 const unsigned int alloc_flags; /* alloc flags of a direct compactor */
198 const int classzone_idx; /* zone index of a direct compactor */
David Rientjese0b9dae2014-06-04 16:08:28 -0700199 enum migrate_mode mode; /* Async or sync migration mode */
Mel Gormanbb13ffe2012-10-08 16:32:41 -0700200 bool ignore_skip_hint; /* Scan blocks even if marked skip */
Vlastimil Babka2583d672017-11-17 15:26:38 -0800201 bool no_set_skip_hint; /* Don't mark blocks for skipping */
Vlastimil Babka9f7e3382016-10-07 17:00:37 -0700202 bool ignore_block_suitable; /* Scan blocks considered unsuitable */
Vlastimil Babkaaccf6242016-03-17 14:18:15 -0700203 bool direct_compaction; /* False from kcompactd or /proc/... */
Vlastimil Babka06ed2992016-10-07 16:57:35 -0700204 bool whole_zone; /* Whole zone should/has been scanned */
Vlastimil Babkac3486f52016-07-28 15:49:30 -0700205 bool contended; /* Signal lock or sched contention */
Vlastimil Babkabaf6a9a2017-05-08 15:54:52 -0700206 bool finishing_block; /* Finishing current pageblock */
Michal Nazarewiczff9543f2011-12-29 13:09:50 +0100207};
208
209unsigned long
Mel Gormanbb13ffe2012-10-08 16:32:41 -0700210isolate_freepages_range(struct compact_control *cc,
211 unsigned long start_pfn, unsigned long end_pfn);
Michal Nazarewiczff9543f2011-12-29 13:09:50 +0100212unsigned long
Vlastimil Babkaedc2ca62014-10-09 15:27:09 -0700213isolate_migratepages_range(struct compact_control *cc,
214 unsigned long low_pfn, unsigned long end_pfn);
Joonsoo Kim2149cda2015-04-14 15:45:21 -0700215int find_suitable_fallback(struct free_area *area, unsigned int order,
216 int migratetype, bool only_stealable, bool *can_steal);
Michal Nazarewiczff9543f2011-12-29 13:09:50 +0100217
218#endif
Nick Piggin0f8053a2006-03-22 00:08:33 -0800219
Mel Gorman48f13bf2007-10-16 01:26:10 -0700220/*
Mel Gorman6c144662014-01-23 15:53:38 -0800221 * This function returns the order of a free page in the buddy system. In
222 * general, page_zone(page)->lock must be held by the caller to prevent the
223 * page from being allocated in parallel and returning garbage as the order.
224 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
Vlastimil Babka99c0fd52014-10-09 15:27:23 -0700225 * page cannot be allocated or merged in parallel. Alternatively, it must
226 * handle invalid values gracefully, and use page_order_unsafe() below.
Mel Gorman48f13bf2007-10-16 01:26:10 -0700227 */
Kirill A. Shutemovd00181b2015-11-06 16:29:57 -0800228static inline unsigned int page_order(struct page *page)
Mel Gorman48f13bf2007-10-16 01:26:10 -0700229{
KAMEZAWA Hiroyuki572438f2010-10-26 14:22:08 -0700230 /* PageBuddy() must be checked by the caller */
Mel Gorman48f13bf2007-10-16 01:26:10 -0700231 return page_private(page);
232}
Alexander van Heukelumb5a0e012008-02-23 15:24:06 -0800233
Vlastimil Babka99c0fd52014-10-09 15:27:23 -0700234/*
235 * Like page_order(), but for callers who cannot afford to hold the zone lock.
236 * PageBuddy() should be checked first by the caller to minimize race window,
237 * and invalid values must be handled gracefully.
238 *
Jason Low4db0c3c2015-04-15 16:14:08 -0700239 * READ_ONCE is used so that if the caller assigns the result into a local
Vlastimil Babka99c0fd52014-10-09 15:27:23 -0700240 * variable and e.g. tests it for valid range before using, the compiler cannot
241 * decide to remove the variable and inline the page_private(page) multiple
242 * times, potentially observing different values in the tests and the actual
243 * use of the result.
244 */
Jason Low4db0c3c2015-04-15 16:14:08 -0700245#define page_order_unsafe(page) READ_ONCE(page_private(page))
Vlastimil Babka99c0fd52014-10-09 15:27:23 -0700246
Kirill A. Shutemov4bbd4c72014-06-04 16:08:10 -0700247static inline bool is_cow_mapping(vm_flags_t flags)
248{
249 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
250}
251
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800252/*
253 * These three helpers classifies VMAs for virtual memory accounting.
254 */
255
256/*
257 * Executable code area - executable, not writable, not stack
258 */
Konstantin Khlebnikovd977d562016-02-02 16:57:43 -0800259static inline bool is_exec_mapping(vm_flags_t flags)
260{
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800261 return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC;
Konstantin Khlebnikovd977d562016-02-02 16:57:43 -0800262}
263
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800264/*
265 * Stack area - atomatically grows in one direction
266 *
267 * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
268 * do_mmap() forbids all other combinations.
269 */
Konstantin Khlebnikovd977d562016-02-02 16:57:43 -0800270static inline bool is_stack_mapping(vm_flags_t flags)
271{
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800272 return (flags & VM_STACK) == VM_STACK;
Konstantin Khlebnikovd977d562016-02-02 16:57:43 -0800273}
274
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800275/*
276 * Data area - private, writable, not stack
277 */
Konstantin Khlebnikovd977d562016-02-02 16:57:43 -0800278static inline bool is_data_mapping(vm_flags_t flags)
279{
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800280 return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE;
Konstantin Khlebnikovd977d562016-02-02 16:57:43 -0800281}
282
Namhyung Kim6038def2011-05-24 17:11:22 -0700283/* mm/util.c */
284void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
285 struct vm_area_struct *prev, struct rb_node *rb_parent);
286
Hugh Dickinsaf8e3352009-12-14 17:58:59 -0800287#ifdef CONFIG_MMU
Kirill A. Shutemovfc05f562015-04-14 15:44:39 -0700288extern long populate_vma_page_range(struct vm_area_struct *vma,
Michel Lespinassecea10a12013-02-22 16:32:44 -0800289 unsigned long start, unsigned long end, int *nonblocking);
Hugh Dickinsaf8e3352009-12-14 17:58:59 -0800290extern void munlock_vma_pages_range(struct vm_area_struct *vma,
291 unsigned long start, unsigned long end);
292static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
293{
294 munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
295}
296
Nick Pigginb291f002008-10-18 20:26:44 -0700297/*
Hugh Dickins73848b42009-12-14 17:59:22 -0800298 * must be called with vma's mmap_sem held for read or write, and page locked.
Nick Pigginb291f002008-10-18 20:26:44 -0700299 */
300extern void mlock_vma_page(struct page *page);
Michel Lespinasseff6a6da2013-02-27 17:02:44 -0800301extern unsigned int munlock_vma_page(struct page *page);
Nick Pigginb291f002008-10-18 20:26:44 -0700302
303/*
304 * Clear the page's PageMlocked(). This can be useful in a situation where
305 * we want to unconditionally remove a page from the pagecache -- e.g.,
306 * on truncation or freeing.
307 *
308 * It is legal to call this function for any page, mlocked or not.
309 * If called for a page that is still mapped by mlocked vmas, all we do
310 * is revert to lazy LRU behaviour -- semantics are not broken.
311 */
Hugh Dickinse6c509f2012-10-08 16:33:19 -0700312extern void clear_page_mlock(struct page *page);
Nick Pigginb291f002008-10-18 20:26:44 -0700313
314/*
Hugh Dickins51afb122015-11-05 18:49:37 -0800315 * mlock_migrate_page - called only from migrate_misplaced_transhuge_page()
316 * (because that does not go through the full procedure of migration ptes):
317 * to migrate the Mlocked page flag; update statistics.
Nick Pigginb291f002008-10-18 20:26:44 -0700318 */
319static inline void mlock_migrate_page(struct page *newpage, struct page *page)
320{
Nick Piggin5344b7e2008-10-18 20:26:51 -0700321 if (TestClearPageMlocked(page)) {
Mel Gormanb32967f2012-11-19 12:35:47 +0000322 int nr_pages = hpage_nr_pages(page);
Nick Piggin5344b7e2008-10-18 20:26:51 -0700323
Hugh Dickins51afb122015-11-05 18:49:37 -0800324 /* Holding pmd lock, no change in irq context: __mod is safe */
Mel Gormanb32967f2012-11-19 12:35:47 +0000325 __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
Nick Pigginb291f002008-10-18 20:26:44 -0700326 SetPageMlocked(newpage);
Mel Gormanb32967f2012-11-19 12:35:47 +0000327 __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
Nick Piggin5344b7e2008-10-18 20:26:51 -0700328 }
Nick Pigginb291f002008-10-18 20:26:44 -0700329}
330
Mel Gormanb32967f2012-11-19 12:35:47 +0000331extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
332
Kirill A. Shutemove9b61f12016-01-15 16:54:10 -0800333/*
334 * At what user virtual address is page expected in @vma?
335 */
336static inline unsigned long
337__vma_address(struct page *page, struct vm_area_struct *vma)
338{
339 pgoff_t pgoff = page_to_pgoff(page);
340 return vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
341}
342
343static inline unsigned long
344vma_address(struct page *page, struct vm_area_struct *vma)
345{
Kirill A. Shutemova8fa41ad2017-02-24 14:57:54 -0800346 unsigned long start, end;
347
348 start = __vma_address(page, vma);
349 end = start + PAGE_SIZE * (hpage_nr_pages(page) - 1);
Kirill A. Shutemove9b61f12016-01-15 16:54:10 -0800350
351 /* page should be within @vma mapping range */
Kirill A. Shutemova8fa41ad2017-02-24 14:57:54 -0800352 VM_BUG_ON_VMA(end < vma->vm_start || start >= vma->vm_end, vma);
Kirill A. Shutemove9b61f12016-01-15 16:54:10 -0800353
Kirill A. Shutemova8fa41ad2017-02-24 14:57:54 -0800354 return max(start, vma->vm_start);
Kirill A. Shutemove9b61f12016-01-15 16:54:10 -0800355}
356
Hugh Dickinsaf8e3352009-12-14 17:58:59 -0800357#else /* !CONFIG_MMU */
Nick Pigginb291f002008-10-18 20:26:44 -0700358static inline void clear_page_mlock(struct page *page) { }
359static inline void mlock_vma_page(struct page *page) { }
360static inline void mlock_migrate_page(struct page *new, struct page *old) { }
361
Hugh Dickinsaf8e3352009-12-14 17:58:59 -0800362#endif /* !CONFIG_MMU */
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700363
Alexander van Heukelumb5a0e012008-02-23 15:24:06 -0800364/*
Andy Whitcroft69d177c2008-11-06 12:53:26 -0800365 * Return the mem_map entry representing the 'offset' subpage within
366 * the maximally aligned gigantic page 'base'. Handle any discontiguity
367 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
368 */
369static inline struct page *mem_map_offset(struct page *base, int offset)
370{
371 if (unlikely(offset >= MAX_ORDER_NR_PAGES))
Fabian Frederickbc7f84c2014-08-06 16:05:17 -0700372 return nth_page(base, offset);
Andy Whitcroft69d177c2008-11-06 12:53:26 -0800373 return base + offset;
374}
375
376/*
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300377 * Iterator over all subpages within the maximally aligned gigantic
Andy Whitcroft69d177c2008-11-06 12:53:26 -0800378 * page 'base'. Handle any discontiguity in the mem_map.
379 */
380static inline struct page *mem_map_next(struct page *iter,
381 struct page *base, int offset)
382{
383 if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
384 unsigned long pfn = page_to_pfn(base) + offset;
385 if (!pfn_valid(pfn))
386 return NULL;
387 return pfn_to_page(pfn);
388 }
389 return iter + 1;
390}
391
392/*
Alexander van Heukelumb5a0e012008-02-23 15:24:06 -0800393 * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
394 * so all functions starting at paging_init should be marked __init
395 * in those cases. SPARSEMEM, however, allows for memory hotplug,
396 * and alloc_bootmem_node is not used.
397 */
398#ifdef CONFIG_SPARSEMEM
399#define __paginginit __meminit
400#else
401#define __paginginit __init
402#endif
403
Mel Gorman6b74ab92008-07-23 21:26:49 -0700404/* Memory initialisation debug and verification */
405enum mminit_level {
406 MMINIT_WARNING,
407 MMINIT_VERIFY,
408 MMINIT_TRACE
409};
410
411#ifdef CONFIG_DEBUG_MEMORY_INIT
412
413extern int mminit_loglevel;
414
415#define mminit_dprintk(level, prefix, fmt, arg...) \
416do { \
417 if (level < mminit_loglevel) { \
Rasmus Villemoesfc5199d2015-02-12 15:00:02 -0800418 if (level <= MMINIT_WARNING) \
Joe Perches11705322016-03-17 14:19:50 -0700419 pr_warn("mminit::" prefix " " fmt, ##arg); \
Rasmus Villemoesfc5199d2015-02-12 15:00:02 -0800420 else \
421 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
Mel Gorman6b74ab92008-07-23 21:26:49 -0700422 } \
423} while (0)
424
Mel Gorman708614e2008-07-23 21:26:51 -0700425extern void mminit_verify_pageflags_layout(void);
Mel Gorman68ad8df2008-07-23 21:26:52 -0700426extern void mminit_verify_zonelist(void);
Mel Gorman6b74ab92008-07-23 21:26:49 -0700427#else
428
429static inline void mminit_dprintk(enum mminit_level level,
430 const char *prefix, const char *fmt, ...)
431{
432}
433
Mel Gorman708614e2008-07-23 21:26:51 -0700434static inline void mminit_verify_pageflags_layout(void)
435{
436}
437
Mel Gorman68ad8df2008-07-23 21:26:52 -0700438static inline void mminit_verify_zonelist(void)
439{
440}
Mel Gorman6b74ab92008-07-23 21:26:49 -0700441#endif /* CONFIG_DEBUG_MEMORY_INIT */
Mel Gorman2dbb51c2008-07-23 21:26:52 -0700442
443/* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
444#if defined(CONFIG_SPARSEMEM)
445extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
446 unsigned long *end_pfn);
447#else
448static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
449 unsigned long *end_pfn)
450{
451}
452#endif /* CONFIG_SPARSEMEM */
453
Mel Gormana5f5f912016-07-28 15:46:32 -0700454#define NODE_RECLAIM_NOSCAN -2
455#define NODE_RECLAIM_FULL -1
456#define NODE_RECLAIM_SOME 0
457#define NODE_RECLAIM_SUCCESS 1
Wu Fengguang7c116f22009-12-16 12:19:59 +0100458
Wu Fengguang31d3d342009-12-16 12:19:59 +0100459extern int hwpoison_filter(struct page *p);
460
Wu Fengguang7c116f22009-12-16 12:19:59 +0100461extern u32 hwpoison_filter_dev_major;
462extern u32 hwpoison_filter_dev_minor;
Wu Fengguang478c5ff2009-12-16 12:19:59 +0100463extern u64 hwpoison_filter_flags_mask;
464extern u64 hwpoison_filter_flags_value;
Andi Kleen4fd466e2009-12-16 12:19:59 +0100465extern u64 hwpoison_filter_memcg;
Haicheng Li1bfe5fe2009-12-16 12:19:59 +0100466extern u32 hwpoison_filter_enable;
Al Viroeb36c582012-05-30 20:17:35 -0400467
Michal Hockodc0ef0d2016-05-23 16:25:27 -0700468extern unsigned long __must_check vm_mmap_pgoff(struct file *, unsigned long,
Al Viroeb36c582012-05-30 20:17:35 -0400469 unsigned long, unsigned long,
Michal Hocko9fbeb5a2016-05-23 16:25:30 -0700470 unsigned long, unsigned long);
Xishi Qiuca57df72012-07-31 16:43:19 -0700471
472extern void set_pageblock_order(void);
Minchan Kim02c6de82012-10-08 16:31:55 -0700473unsigned long reclaim_clean_pages_from_list(struct zone *zone,
474 struct list_head *page_list);
Bartlomiej Zolnierkiewiczd95ea5d2012-10-08 16:32:05 -0700475/* The ALLOC_WMARK bits are used as an index to zone->watermark */
476#define ALLOC_WMARK_MIN WMARK_MIN
477#define ALLOC_WMARK_LOW WMARK_LOW
478#define ALLOC_WMARK_HIGH WMARK_HIGH
479#define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
480
481/* Mask to get the watermark bits */
482#define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
483
Michal Hockocd04ae12017-09-06 16:24:50 -0700484/*
485 * Only MMU archs have async oom victim reclaim - aka oom_reaper so we
486 * cannot assume a reduced access to memory reserves is sufficient for
487 * !MMU
488 */
489#ifdef CONFIG_MMU
490#define ALLOC_OOM 0x08
491#else
492#define ALLOC_OOM ALLOC_NO_WATERMARKS
493#endif
494
Bartlomiej Zolnierkiewiczd95ea5d2012-10-08 16:32:05 -0700495#define ALLOC_HARDER 0x10 /* try to alloc harder */
496#define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
497#define ALLOC_CPUSET 0x40 /* check for correct cpuset */
498#define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
499
Mel Gorman72b252a2015-09-04 15:47:32 -0700500enum ttu_flags;
501struct tlbflush_unmap_batch;
502
Michal Hockoce612872017-04-07 16:05:05 -0700503
504/*
505 * only for MM internal work items which do not depend on
506 * any allocations or locks which might depend on allocations
507 */
508extern struct workqueue_struct *mm_percpu_wq;
509
Mel Gorman72b252a2015-09-04 15:47:32 -0700510#ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
511void try_to_unmap_flush(void);
Mel Gormand950c942015-09-04 15:47:35 -0700512void try_to_unmap_flush_dirty(void);
Mel Gorman3ea27712017-08-02 13:31:52 -0700513void flush_tlb_batched_pending(struct mm_struct *mm);
Mel Gorman72b252a2015-09-04 15:47:32 -0700514#else
515static inline void try_to_unmap_flush(void)
516{
517}
Mel Gormand950c942015-09-04 15:47:35 -0700518static inline void try_to_unmap_flush_dirty(void)
519{
520}
Mel Gorman3ea27712017-08-02 13:31:52 -0700521static inline void flush_tlb_batched_pending(struct mm_struct *mm)
522{
523}
Mel Gorman72b252a2015-09-04 15:47:32 -0700524#endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
Vlastimil Babkaedf14cd2016-03-15 14:55:56 -0700525
526extern const struct trace_print_flags pageflag_names[];
527extern const struct trace_print_flags vmaflag_names[];
528extern const struct trace_print_flags gfpflag_names[];
529
Xishi Qiua6ffdc02017-05-03 14:52:52 -0700530static inline bool is_migrate_highatomic(enum migratetype migratetype)
531{
532 return migratetype == MIGRATE_HIGHATOMIC;
533}
534
535static inline bool is_migrate_highatomic_page(struct page *page)
536{
537 return get_pageblock_migratetype(page) == MIGRATE_HIGHATOMIC;
538}
539
Michal Hocko72675e12017-09-06 16:20:24 -0700540void setup_zone_pageset(struct zone *zone);
Michel Lespinassedb971412012-10-08 16:29:34 -0700541#endif /* __MM_INTERNAL_H */