blob: 3b79a5c9427a827e7646dbaf0abfe3b516c60a81 [file] [log] [blame]
Thomas Gleixner2874c5f2019-05-27 08:55:01 +02001/* SPDX-License-Identifier: GPL-2.0-or-later */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002/* internal.h: mm/ internal definitions
3 *
4 * Copyright (C) 2004 Red Hat, Inc. All Rights Reserved.
5 * Written by David Howells (dhowells@redhat.com)
Linus Torvalds1da177e2005-04-16 15:20:36 -07006 */
Nick Piggin0f8053a2006-03-22 00:08:33 -08007#ifndef __MM_INTERNAL_H
8#define __MM_INTERNAL_H
9
Fabian Frederick29f175d2014-04-07 15:37:55 -070010#include <linux/fs.h>
Nick Piggin0f8053a2006-03-22 00:08:33 -080011#include <linux/mm.h>
Kirill A. Shutemove9b61f12016-01-15 16:54:10 -080012#include <linux/pagemap.h>
Vlastimil Babkaedf14cd2016-03-15 14:55:56 -070013#include <linux/tracepoint-defs.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070014
Mel Gormandd56b042015-11-06 16:28:43 -080015/*
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|\
Michal Hockodcda9b02017-07-12 14:36:45 -070022 __GFP_NOWARN|__GFP_RETRY_MAYFAIL|__GFP_NOFAIL|\
Mel Gormane838a452016-06-24 14:49:37 -070023 __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC|\
24 __GFP_ATOMIC)
Mel Gormandd56b042015-11-06 16:28:43 -080025
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
Nicholas Piggin62906022016-12-25 13:00:30 +100035void page_writeback_init(void);
36
Matthew Wilcox (Oracle)64601002021-05-07 11:17:34 -040037static 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
Linus Torvalds512b79312021-11-06 14:08:17 -070044void __acct_reclaim_writeback(pg_data_t *pgdat, struct folio *folio,
45 int nr_throttled);
46static inline void acct_reclaim_writeback(struct folio *folio)
47{
48 pg_data_t *pgdat = folio_pgdat(folio);
Mel Gorman8cd7c582021-11-05 13:42:25 -070049 int nr_throttled = atomic_read(&pgdat->nr_writeback_throttled);
50
51 if (nr_throttled)
Linus Torvalds512b79312021-11-06 14:08:17 -070052 __acct_reclaim_writeback(pgdat, folio, nr_throttled);
Mel Gorman8cd7c582021-11-05 13:42:25 -070053}
54
Mel Gormand818fca2021-11-05 13:42:29 -070055static 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
Souptick Joarder2b740302018-08-23 17:01:36 -070064vm_fault_t do_swap_page(struct vm_fault *vmf);
Matthew Wilcox (Oracle)575ced12020-12-08 01:25:39 -050065void folio_rotate_reclaimable(struct folio *folio);
Matthew Wilcox (Oracle)269ccca32021-01-15 23:34:16 -050066bool __folio_end_writeback(struct folio *folio);
Ebru Akagunduz8a966ed2016-07-26 15:25:03 -070067
Jan Beulich42b77722008-07-23 21:27:10 -070068void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
69 unsigned long floor, unsigned long ceiling);
Qi Zheng03c4f202021-11-05 13:38:38 -070070void pmd_install(struct mm_struct *mm, pmd_t *pmd, pgtable_t *pte);
Jan Beulich42b77722008-07-23 21:27:10 -070071
Minchan Kim9c276cc2019-09-25 16:49:08 -070072static inline bool can_madv_lru_vma(struct vm_area_struct *vma)
Kirill A. Shutemov23519072017-02-22 15:46:39 -080073{
74 return !(vma->vm_flags & (VM_LOCKED|VM_HUGETLB|VM_PFNMAP));
75}
76
Michal Hockoaac45362016-03-25 14:20:24 -070077void 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
David Howells7b3df3b2020-10-15 20:06:24 -070082void do_page_cache_ra(struct readahead_control *, unsigned long nr_to_read,
83 unsigned long lookahead_size);
Matthew Wilcox (Oracle)fcd9ae42021-04-07 21:18:55 +010084void force_page_cache_ra(struct readahead_control *, unsigned long nr);
David Howells7b3df3b2020-10-15 20:06:24 -070085static inline void force_page_cache_readahead(struct address_space *mapping,
86 struct file *file, pgoff_t index, unsigned long nr_to_read)
87{
Matthew Wilcox (Oracle)fcd9ae42021-04-07 21:18:55 +010088 DEFINE_READAHEAD(ractl, file, &file->f_ra, mapping, index);
89 force_page_cache_ra(&ractl, nr_to_read);
David Howells7b3df3b2020-10-15 20:06:24 -070090}
Fabian Frederick29f175d2014-04-07 15:37:55 -070091
Matthew Wilcox (Oracle)5c211ba2021-02-25 17:15:56 -080092unsigned find_lock_entries(struct address_space *mapping, pgoff_t start,
93 pgoff_t end, struct pagevec *pvec, pgoff_t *indices);
94
Yang Shi1eb62342020-04-01 21:06:20 -070095/**
Matthew Wilcox (Oracle)3eed3ef2021-05-14 15:04:28 -040096 * folio_evictable - Test whether a folio is evictable.
97 * @folio: The folio to test.
Yang Shi1eb62342020-04-01 21:06:20 -070098 *
Matthew Wilcox (Oracle)3eed3ef2021-05-14 15:04:28 -040099 * Test whether @folio is evictable -- i.e., should be placed on
100 * active/inactive lists vs unevictable list.
Yang Shi1eb62342020-04-01 21:06:20 -0700101 *
Matthew Wilcox (Oracle)3eed3ef2021-05-14 15:04:28 -0400102 * 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
Yang Shi1eb62342020-04-01 21:06:20 -0700105 */
Matthew Wilcox (Oracle)3eed3ef2021-05-14 15:04:28 -0400106static 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
Yang Shi1eb62342020-04-01 21:06:20 -0700118static 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
Nick Piggin7835e982006-03-22 00:08:40 -0800129/*
Joonsoo Kim0139aa72016-05-19 17:10:49 -0700130 * Turn a non-refcounted page (->_refcount == 0) into refcounted with
Nick Piggin7835e982006-03-22 00:08:40 -0800131 * a count of one.
132 */
133static inline void set_page_refcounted(struct page *page)
134{
Sasha Levin309381fea2014-01-23 15:52:54 -0800135 VM_BUG_ON_PAGE(PageTail(page), page);
Joonsoo Kimfe896d12016-03-17 14:19:26 -0700136 VM_BUG_ON_PAGE(page_ref_count(page), page);
Nick Piggin77a8a782006-01-06 00:10:57 -0800137 set_page_count(page, 1);
Nick Piggin77a8a782006-01-06 00:10:57 -0800138}
139
Hugh Dickins03f64622009-09-21 17:03:35 -0700140extern unsigned long highest_memmap_pfn;
141
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700142/*
Johannes Weinerc73322d2017-05-03 14:51:51 -0700143 * 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
148/*
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700149 * in mm/vmscan.c:
150 */
Nick Piggin62695a82008-10-18 20:26:09 -0700151extern int isolate_lru_page(struct page *page);
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700152extern void putback_lru_page(struct page *page);
Mel Gormanc3f4a9a2021-11-05 13:42:42 -0700153extern void reclaim_throttle(pg_data_t *pgdat, enum vmscan_throttle_state reason);
Nick Piggin62695a82008-10-18 20:26:09 -0700154
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700155/*
Bob Liu62190492012-12-11 16:00:37 -0800156 * in mm/rmap.c:
157 */
158extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
159
160/*
Waiman Long494c1df2021-06-28 19:37:38 -0700161 * in mm/memcontrol.c:
162 */
163extern bool cgroup_memory_nokmem;
164
165/*
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700166 * in mm/page_alloc.c
167 */
Joonsoo Kim3c605092014-11-13 15:19:21 -0800168
169/*
Vlastimil Babka1a6d53a2015-02-11 15:25:44 -0800170 * Structure for holding the mostly immutable allocation parameters passed
171 * between functions involved in allocations, including the alloc_pages*
172 * family of functions.
173 *
Joonsoo Kim97a225e2020-06-03 15:59:01 -0700174 * nodemask, migratetype and highest_zoneidx are initialized only once in
Matthew Wilcox (Oracle)84172f42021-04-29 23:01:15 -0700175 * __alloc_pages() and then never change.
Vlastimil Babka1a6d53a2015-02-11 15:25:44 -0800176 *
Joonsoo Kim97a225e2020-06-03 15:59:01 -0700177 * zonelist, preferred_zone and highest_zoneidx are set first in
Matthew Wilcox (Oracle)84172f42021-04-29 23:01:15 -0700178 * __alloc_pages() for the fast path, and might be later changed
Ethon Paul68956cc2020-06-04 16:49:31 -0700179 * in __alloc_pages_slowpath(). All other functions pass the whole structure
Vlastimil Babka1a6d53a2015-02-11 15:25:44 -0800180 * by a const pointer.
181 */
182struct alloc_context {
183 struct zonelist *zonelist;
184 nodemask_t *nodemask;
Mel Gormanc33d6c02016-05-19 17:14:10 -0700185 struct zoneref *preferred_zoneref;
Vlastimil Babka1a6d53a2015-02-11 15:25:44 -0800186 int migratetype;
Joonsoo Kim97a225e2020-06-03 15:59:01 -0700187
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;
Mel Gormanc9ab0c42015-11-06 16:28:12 -0800199 bool spread_dirty_pages;
Vlastimil Babka1a6d53a2015-02-11 15:25:44 -0800200};
201
202/*
Joonsoo Kim3c605092014-11-13 15:19:21 -0800203 * 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
Vlastimil Babka76741e72017-02-22 15:41:48 -0800220__find_buddy_pfn(unsigned long page_pfn, unsigned int order)
Joonsoo Kim3c605092014-11-13 15:19:21 -0800221{
Vlastimil Babka76741e72017-02-22 15:41:48 -0800222 return page_pfn ^ (1 << order);
Joonsoo Kim3c605092014-11-13 15:19:21 -0800223}
224
Joonsoo Kim7cf91a92016-03-15 14:57:51 -0700225extern 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
Joonsoo Kim3c605092014-11-13 15:19:21 -0800237extern int __isolate_free_page(struct page *page, unsigned int order);
Alexander Duyck624f58d2020-04-06 20:04:53 -0700238extern void __putback_isolated_page(struct page *page, unsigned int order,
239 int mt);
Mike Rapoport7c2ee342018-10-30 15:09:36 -0700240extern void memblock_free_pages(struct page *page, unsigned long pfn,
Mel Gormand70ddd72015-06-30 14:56:52 -0700241 unsigned int order);
Arun KSa9cd4102019-03-05 15:42:14 -0800242extern void __free_pages_core(struct page *page, unsigned int order);
Kirill A. Shutemovd00181b2015-11-06 16:29:57 -0800243extern void prep_compound_page(struct page *page, unsigned int order);
Joonsoo Kim46f24fd2016-07-26 15:23:58 -0700244extern void post_alloc_hook(struct page *page, unsigned int order,
245 gfp_t gfp_flags);
Han Pingtian42aa83c2014-01-23 15:53:28 -0800246extern int user_min_free_kbytes;
Wu Fengguang20a03072009-06-16 15:32:22 -0700247
Mel Gorman44042b42021-06-28 19:43:08 -0700248extern void free_unref_page(struct page *page, unsigned int order);
Matthew Wilcox (Oracle)0966aeb2020-12-14 19:08:02 -0800249extern void free_unref_page_list(struct list_head *list);
250
Mel Gorman04f8cfe2021-06-28 19:42:15 -0700251extern void zone_pcp_update(struct zone *zone, int cpu_online);
Mel Gorman68265392019-11-30 17:55:15 -0800252extern void zone_pcp_reset(struct zone *zone);
Vlastimil Babkaec6e8c7e2020-12-14 19:10:59 -0800253extern void zone_pcp_disable(struct zone *zone);
254extern void zone_pcp_enable(struct zone *zone);
Mel Gorman68265392019-11-30 17:55:15 -0800255
Mike Rapoportc803b3c2021-09-02 14:58:02 -0700256extern void *memmap_alloc(phys_addr_t size, phys_addr_t align,
257 phys_addr_t min_addr,
258 int nid, bool exact_nid);
259
Michal Nazarewiczff9543f2011-12-29 13:09:50 +0100260#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 */
Mel Gormanc5fbd932019-03-05 15:44:25 -0800275 unsigned int nr_freepages; /* Number of isolated free pages */
276 unsigned int nr_migratepages; /* Number of pages to migrate */
Michal Nazarewiczff9543f2011-12-29 13:09:50 +0100277 unsigned long free_pfn; /* isolate_freepages search base */
Oscar Salvadorc2ad7a12021-05-04 18:35:17 -0700278 /*
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;
Mel Gorman70b44592019-03-05 15:44:54 -0800285 unsigned long fast_start_pfn; /* a pfn to start linear scan from */
Mel Gormanc5943b92019-03-05 15:44:28 -0800286 struct zone *zone;
287 unsigned long total_migrate_scanned;
288 unsigned long total_free_scanned;
Mel Gormandbe2d4e2019-03-05 15:45:31 -0800289 unsigned short fast_search_fail;/* failures to use free list searches */
290 short search_order; /* order to start a fast search at */
Vlastimil Babkaf25ba6d2017-05-08 15:54:30 -0700291 const gfp_t gfp_mask; /* gfp mask of a direct compactor */
292 int order; /* order a direct compactor needs */
Vlastimil Babkad39773a2017-05-08 15:54:46 -0700293 int migratetype; /* migratetype of direct compactor */
Vlastimil Babkaf25ba6d2017-05-08 15:54:30 -0700294 const unsigned int alloc_flags; /* alloc flags of a direct compactor */
Joonsoo Kim97a225e2020-06-03 15:59:01 -0700295 const int highest_zoneidx; /* zone index of a direct compactor */
David Rientjese0b9dae2014-06-04 16:08:28 -0700296 enum migrate_mode mode; /* Async or sync migration mode */
Mel Gormanbb13ffe2012-10-08 16:32:41 -0700297 bool ignore_skip_hint; /* Scan blocks even if marked skip */
Vlastimil Babka2583d672017-11-17 15:26:38 -0800298 bool no_set_skip_hint; /* Don't mark blocks for skipping */
Vlastimil Babka9f7e3382016-10-07 17:00:37 -0700299 bool ignore_block_suitable; /* Scan blocks considered unsuitable */
Vlastimil Babkaaccf6242016-03-17 14:18:15 -0700300 bool direct_compaction; /* False from kcompactd or /proc/... */
Nitin Guptafacdaa92020-08-11 18:31:00 -0700301 bool proactive_compaction; /* kcompactd proactive compaction */
Vlastimil Babka06ed2992016-10-07 16:57:35 -0700302 bool whole_zone; /* Whole zone should/has been scanned */
Vlastimil Babkac3486f52016-07-28 15:49:30 -0700303 bool contended; /* Signal lock or sched contention */
Mel Gorman804d3122019-03-05 15:45:07 -0800304 bool rescan; /* Rescanning the same pageblock */
Rik van Rielb06eda02020-04-01 21:10:28 -0700305 bool alloc_contig; /* alloc_contig_range allocation */
Michal Nazarewiczff9543f2011-12-29 13:09:50 +0100306};
307
Mel Gorman5e1f0f02019-03-05 15:45:41 -0800308/*
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
Michal Nazarewiczff9543f2011-12-29 13:09:50 +0100317unsigned long
Mel Gormanbb13ffe2012-10-08 16:32:41 -0700318isolate_freepages_range(struct compact_control *cc,
319 unsigned long start_pfn, unsigned long end_pfn);
Oscar Salvadorc2ad7a12021-05-04 18:35:17 -0700320int
Vlastimil Babkaedc2ca62014-10-09 15:27:09 -0700321isolate_migratepages_range(struct compact_control *cc,
322 unsigned long low_pfn, unsigned long end_pfn);
Mel Gormanffd8f252021-06-30 18:53:53 -0700323#endif
Joonsoo Kim2149cda2015-04-14 15:45:21 -0700324int find_suitable_fallback(struct free_area *area, unsigned int order,
325 int migratetype, bool only_stealable, bool *can_steal);
Michal Nazarewiczff9543f2011-12-29 13:09:50 +0100326
Mel Gorman48f13bf2007-10-16 01:26:10 -0700327/*
Mel Gorman6c144662014-01-23 15:53:38 -0800328 * 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
Vlastimil Babka99c0fd52014-10-09 15:27:23 -0700332 * page cannot be allocated or merged in parallel. Alternatively, it must
Matthew Wilcox (Oracle)ab130f912020-10-15 20:10:15 -0700333 * handle invalid values gracefully, and use buddy_order_unsafe() below.
Mel Gorman48f13bf2007-10-16 01:26:10 -0700334 */
Matthew Wilcox (Oracle)ab130f912020-10-15 20:10:15 -0700335static inline unsigned int buddy_order(struct page *page)
Mel Gorman48f13bf2007-10-16 01:26:10 -0700336{
KAMEZAWA Hiroyuki572438f2010-10-26 14:22:08 -0700337 /* PageBuddy() must be checked by the caller */
Mel Gorman48f13bf2007-10-16 01:26:10 -0700338 return page_private(page);
339}
Alexander van Heukelumb5a0e012008-02-23 15:24:06 -0800340
Vlastimil Babka99c0fd52014-10-09 15:27:23 -0700341/*
Matthew Wilcox (Oracle)ab130f912020-10-15 20:10:15 -0700342 * Like buddy_order(), but for callers who cannot afford to hold the zone lock.
Vlastimil Babka99c0fd52014-10-09 15:27:23 -0700343 * PageBuddy() should be checked first by the caller to minimize race window,
344 * and invalid values must be handled gracefully.
345 *
Jason Low4db0c3c2015-04-15 16:14:08 -0700346 * READ_ONCE is used so that if the caller assigns the result into a local
Vlastimil Babka99c0fd52014-10-09 15:27:23 -0700347 * 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 */
Matthew Wilcox (Oracle)ab130f912020-10-15 20:10:15 -0700352#define buddy_order_unsafe(page) READ_ONCE(page_private(page))
Vlastimil Babka99c0fd52014-10-09 15:27:23 -0700353
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800354/*
355 * These three helpers classifies VMAs for virtual memory accounting.
356 */
357
358/*
359 * Executable code area - executable, not writable, not stack
360 */
Konstantin Khlebnikovd977d562016-02-02 16:57:43 -0800361static inline bool is_exec_mapping(vm_flags_t flags)
362{
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800363 return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC;
Konstantin Khlebnikovd977d562016-02-02 16:57:43 -0800364}
365
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800366/*
Ingo Molnarf0953a12021-05-06 18:06:47 -0700367 * Stack area - automatically grows in one direction
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800368 *
369 * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
370 * do_mmap() forbids all other combinations.
371 */
Konstantin Khlebnikovd977d562016-02-02 16:57:43 -0800372static inline bool is_stack_mapping(vm_flags_t flags)
373{
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800374 return (flags & VM_STACK) == VM_STACK;
Konstantin Khlebnikovd977d562016-02-02 16:57:43 -0800375}
376
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800377/*
378 * Data area - private, writable, not stack
379 */
Konstantin Khlebnikovd977d562016-02-02 16:57:43 -0800380static inline bool is_data_mapping(vm_flags_t flags)
381{
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800382 return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE;
Konstantin Khlebnikovd977d562016-02-02 16:57:43 -0800383}
384
Namhyung Kim6038def2011-05-24 17:11:22 -0700385/* mm/util.c */
386void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
Wei Yangaba6dfb2019-11-30 17:50:53 -0800387 struct vm_area_struct *prev);
Wei Yang1b9fc5b22019-11-30 17:50:49 -0800388void __vma_unlink_list(struct mm_struct *mm, struct vm_area_struct *vma);
Namhyung Kim6038def2011-05-24 17:11:22 -0700389
Hugh Dickinsaf8e3352009-12-14 17:58:59 -0800390#ifdef CONFIG_MMU
Kirill A. Shutemovfc05f562015-04-14 15:44:39 -0700391extern long populate_vma_page_range(struct vm_area_struct *vma,
David Hildenbranda78f1cc2021-06-30 18:52:23 -0700392 unsigned long start, unsigned long end, int *locked);
David Hildenbrand4ca9b3852021-06-30 18:52:28 -0700393extern long faultin_vma_page_range(struct vm_area_struct *vma,
394 unsigned long start, unsigned long end,
395 bool write, int *locked);
Hugh Dickinsaf8e3352009-12-14 17:58:59 -0800396extern 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
Nick Pigginb291f002008-10-18 20:26:44 -0700403/*
Michel Lespinassec1e8d7c2020-06-08 21:33:54 -0700404 * must be called with vma's mmap_lock held for read or write, and page locked.
Nick Pigginb291f002008-10-18 20:26:44 -0700405 */
406extern void mlock_vma_page(struct page *page);
Michel Lespinasseff6a6da2013-02-27 17:02:44 -0800407extern unsigned int munlock_vma_page(struct page *page);
Nick Pigginb291f002008-10-18 20:26:44 -0700408
Mike Rapoport6aeb2542021-07-07 18:07:50 -0700409extern int mlock_future_check(struct mm_struct *mm, unsigned long flags,
410 unsigned long len);
411
Nick Pigginb291f002008-10-18 20:26:44 -0700412/*
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 */
Hugh Dickinse6c509f2012-10-08 16:33:19 -0700421extern void clear_page_mlock(struct page *page);
Nick Pigginb291f002008-10-18 20:26:44 -0700422
Linus Torvaldsf55e1012017-11-29 09:01:01 -0800423extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
Mel Gormanb32967f2012-11-19 12:35:47 +0000424
Kirill A. Shutemove9b61f12016-01-15 16:54:10 -0800425/*
Hugh Dickins494334e2021-06-15 18:23:56 -0700426 * 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.
Kirill A. Shutemove9b61f12016-01-15 16:54:10 -0800429 */
430static inline unsigned long
Kirill A. Shutemove9b61f12016-01-15 16:54:10 -0800431vma_address(struct page *page, struct vm_area_struct *vma)
432{
Hugh Dickins494334e2021-06-15 18:23:56 -0700433 pgoff_t pgoff;
434 unsigned long address;
Kirill A. Shutemova8fa41ad2017-02-24 14:57:54 -0800435
Hugh Dickins494334e2021-06-15 18:23:56 -0700436 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;
452}
Kirill A. Shutemove9b61f12016-01-15 16:54:10 -0800453
Hugh Dickins494334e2021-06-15 18:23:56 -0700454/*
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 */
459static inline unsigned long
460vma_address_end(struct page *page, struct vm_area_struct *vma)
461{
462 pgoff_t pgoff;
463 unsigned long address;
Kirill A. Shutemove9b61f12016-01-15 16:54:10 -0800464
Hugh Dickins494334e2021-06-15 18:23:56 -0700465 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;
Kirill A. Shutemove9b61f12016-01-15 16:54:10 -0800472}
473
Johannes Weiner89b15332019-11-30 17:50:22 -0800474static 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
Michel Lespinassec1e8d7c2020-06-08 21:33:54 -0700484 * anything, so we only pin the file and drop the mmap_lock if only
Peter Xu4064b982020-04-01 21:08:45 -0700485 * FAULT_FLAG_ALLOW_RETRY is set, while this is the first attempt.
Johannes Weiner89b15332019-11-30 17:50:22 -0800486 */
Peter Xu4064b982020-04-01 21:08:45 -0700487 if (fault_flag_allow_retry_first(flags) &&
488 !(flags & FAULT_FLAG_RETRY_NOWAIT)) {
Johannes Weiner89b15332019-11-30 17:50:22 -0800489 fpin = get_file(vmf->vma->vm_file);
Michel Lespinassed8ed45c2020-06-08 21:33:25 -0700490 mmap_read_unlock(vmf->vma->vm_mm);
Johannes Weiner89b15332019-11-30 17:50:22 -0800491 }
492 return fpin;
493}
494
Hugh Dickinsaf8e3352009-12-14 17:58:59 -0800495#else /* !CONFIG_MMU */
Nick Pigginb291f002008-10-18 20:26:44 -0700496static inline void clear_page_mlock(struct page *page) { }
497static inline void mlock_vma_page(struct page *page) { }
Nicholas Piggin4ad0ae82021-04-29 22:59:01 -0700498static inline void vunmap_range_noflush(unsigned long start, unsigned long end)
499{
500}
Hugh Dickinsaf8e3352009-12-14 17:58:59 -0800501#endif /* !CONFIG_MMU */
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700502
Alexander van Heukelumb5a0e012008-02-23 15:24:06 -0800503/*
Andy Whitcroft69d177c2008-11-06 12:53:26 -0800504 * 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))
Fabian Frederickbc7f84c2014-08-06 16:05:17 -0700511 return nth_page(base, offset);
Andy Whitcroft69d177c2008-11-06 12:53:26 -0800512 return base + offset;
513}
514
515/*
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300516 * Iterator over all subpages within the maximally aligned gigantic
Andy Whitcroft69d177c2008-11-06 12:53:26 -0800517 * 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
Mel Gorman6b74ab92008-07-23 21:26:49 -0700531/* 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) { \
Rasmus Villemoesfc5199d2015-02-12 15:00:02 -0800545 if (level <= MMINIT_WARNING) \
Joe Perches11705322016-03-17 14:19:50 -0700546 pr_warn("mminit::" prefix " " fmt, ##arg); \
Rasmus Villemoesfc5199d2015-02-12 15:00:02 -0800547 else \
548 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
Mel Gorman6b74ab92008-07-23 21:26:49 -0700549 } \
550} while (0)
551
Mel Gorman708614e2008-07-23 21:26:51 -0700552extern void mminit_verify_pageflags_layout(void);
Mel Gorman68ad8df2008-07-23 21:26:52 -0700553extern void mminit_verify_zonelist(void);
Mel Gorman6b74ab92008-07-23 21:26:49 -0700554#else
555
556static inline void mminit_dprintk(enum mminit_level level,
557 const char *prefix, const char *fmt, ...)
558{
559}
560
Mel Gorman708614e2008-07-23 21:26:51 -0700561static inline void mminit_verify_pageflags_layout(void)
562{
563}
564
Mel Gorman68ad8df2008-07-23 21:26:52 -0700565static inline void mminit_verify_zonelist(void)
566{
567}
Mel Gorman6b74ab92008-07-23 21:26:49 -0700568#endif /* CONFIG_DEBUG_MEMORY_INIT */
Mel Gorman2dbb51c2008-07-23 21:26:52 -0700569
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
Mel Gormana5f5f912016-07-28 15:46:32 -0700581#define NODE_RECLAIM_NOSCAN -2
582#define NODE_RECLAIM_FULL -1
583#define NODE_RECLAIM_SOME 0
584#define NODE_RECLAIM_SUCCESS 1
Wu Fengguang7c116f22009-12-16 12:19:59 +0100585
Wei Yang8b095492018-12-28 00:34:36 -0800586#ifdef CONFIG_NUMA
587extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int);
Dave Hansen79c28a42021-09-02 14:59:06 -0700588extern int find_next_best_node(int node, nodemask_t *used_node_mask);
Wei Yang8b095492018-12-28 00:34:36 -0800589#else
590static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask,
591 unsigned int order)
592{
593 return NODE_RECLAIM_NOSCAN;
594}
Dave Hansen79c28a42021-09-02 14:59:06 -0700595static inline int find_next_best_node(int node, nodemask_t *used_node_mask)
596{
597 return NUMA_NO_NODE;
598}
Wei Yang8b095492018-12-28 00:34:36 -0800599#endif
600
Wu Fengguang31d3d342009-12-16 12:19:59 +0100601extern int hwpoison_filter(struct page *p);
602
Wu Fengguang7c116f22009-12-16 12:19:59 +0100603extern u32 hwpoison_filter_dev_major;
604extern u32 hwpoison_filter_dev_minor;
Wu Fengguang478c5ff2009-12-16 12:19:59 +0100605extern u64 hwpoison_filter_flags_mask;
606extern u64 hwpoison_filter_flags_value;
Andi Kleen4fd466e2009-12-16 12:19:59 +0100607extern u64 hwpoison_filter_memcg;
Haicheng Li1bfe5fe2009-12-16 12:19:59 +0100608extern u32 hwpoison_filter_enable;
Al Viroeb36c582012-05-30 20:17:35 -0400609
Michal Hockodc0ef0d2016-05-23 16:25:27 -0700610extern unsigned long __must_check vm_mmap_pgoff(struct file *, unsigned long,
Al Viroeb36c582012-05-30 20:17:35 -0400611 unsigned long, unsigned long,
Michal Hocko9fbeb5a2016-05-23 16:25:30 -0700612 unsigned long, unsigned long);
Xishi Qiuca57df72012-07-31 16:43:19 -0700613
614extern void set_pageblock_order(void);
Maninder Singh730ec8c2020-06-03 16:01:18 -0700615unsigned int reclaim_clean_pages_from_list(struct zone *zone,
Minchan Kim02c6de82012-10-08 16:31:55 -0700616 struct list_head *page_list);
Bartlomiej Zolnierkiewiczd95ea5d2012-10-08 16:32:05 -0700617/* 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
Michal Hockocd04ae12017-09-06 16:24:50 -0700626/*
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
Mel Gorman6bb15452018-12-28 00:35:41 -0800637#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
Mateusz Nosek736838e2020-04-01 21:09:47 -0700646#define ALLOC_KSWAPD 0x800 /* allow waking of kswapd, __GFP_KSWAPD_RECLAIM set */
Bartlomiej Zolnierkiewiczd95ea5d2012-10-08 16:32:05 -0700647
Mel Gorman72b252a2015-09-04 15:47:32 -0700648enum ttu_flags;
649struct tlbflush_unmap_batch;
650
Michal Hockoce612872017-04-07 16:05:05 -0700651
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
Mel Gorman72b252a2015-09-04 15:47:32 -0700658#ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
659void try_to_unmap_flush(void);
Mel Gormand950c942015-09-04 15:47:35 -0700660void try_to_unmap_flush_dirty(void);
Mel Gorman3ea27712017-08-02 13:31:52 -0700661void flush_tlb_batched_pending(struct mm_struct *mm);
Mel Gorman72b252a2015-09-04 15:47:32 -0700662#else
663static inline void try_to_unmap_flush(void)
664{
665}
Mel Gormand950c942015-09-04 15:47:35 -0700666static inline void try_to_unmap_flush_dirty(void)
667{
668}
Mel Gorman3ea27712017-08-02 13:31:52 -0700669static inline void flush_tlb_batched_pending(struct mm_struct *mm)
670{
671}
Mel Gorman72b252a2015-09-04 15:47:32 -0700672#endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
Vlastimil Babkaedf14cd2016-03-15 14:55:56 -0700673
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
Xishi Qiua6ffdc02017-05-03 14:52:52 -0700678static 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
Michal Hocko72675e12017-09-06 16:20:24 -0700688void setup_zone_pageset(struct zone *zone);
Joonsoo Kim19fc7be2020-08-11 18:37:25 -0700689
690struct migration_target_control {
691 int nid; /* preferred node id */
692 nodemask_t *nmask;
693 gfp_t gfp_mask;
694};
695
Nicholas Pigginb67177e2021-04-29 22:58:53 -0700696/*
697 * mm/vmalloc.c
698 */
Nicholas Piggin4ad0ae82021-04-29 22:59:01 -0700699#ifdef CONFIG_MMU
Nicholas Pigginb67177e2021-04-29 22:58:53 -0700700int vmap_pages_range_noflush(unsigned long addr, unsigned long end,
701 pgprot_t prot, struct page **pages, unsigned int page_shift);
Nicholas Piggin4ad0ae82021-04-29 22:59:01 -0700702#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);
Nicholas Pigginb67177e2021-04-29 22:58:53 -0700712
Yang Shif4c0d832021-06-30 18:51:39 -0700713int numa_migrate_prep(struct page *page, struct vm_area_struct *vma,
714 unsigned long addr, int page_nid, int *flags);
715
Michel Lespinassedb971412012-10-08 16:29:34 -0700716#endif /* __MM_INTERNAL_H */