blob: d80300392a194f20b4864191ab0ecaba831cb1aa [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
Matthew Wilcox (Oracle)0e499ed2020-09-01 23:17:50 -040015struct folio_batch;
16
Mel Gormandd56b042015-11-06 16:28:43 -080017/*
18 * The set of flags that only affect watermark checking and reclaim
19 * behaviour. This is used by the MM to obey the caller constraints
20 * about IO, FS and watermark checking while ignoring placement
21 * hints such as HIGHMEM usage.
22 */
23#define GFP_RECLAIM_MASK (__GFP_RECLAIM|__GFP_HIGH|__GFP_IO|__GFP_FS|\
Michal Hockodcda9b02017-07-12 14:36:45 -070024 __GFP_NOWARN|__GFP_RETRY_MAYFAIL|__GFP_NOFAIL|\
Mel Gormane838a452016-06-24 14:49:37 -070025 __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC|\
Michal Hocko704687d2022-01-14 14:07:11 -080026 __GFP_ATOMIC|__GFP_NOLOCKDEP)
Mel Gormandd56b042015-11-06 16:28:43 -080027
28/* The GFP flags allowed during early boot */
29#define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_RECLAIM|__GFP_IO|__GFP_FS))
30
31/* Control allocation cpuset and node placement constraints */
32#define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE)
33
34/* Do not use these with a slab allocator */
35#define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
36
Nicholas Piggin62906022016-12-25 13:00:30 +100037void page_writeback_init(void);
38
Matthew Wilcox (Oracle)64601002021-05-07 11:17:34 -040039static inline void *folio_raw_mapping(struct folio *folio)
40{
41 unsigned long mapping = (unsigned long)folio->mapping;
42
43 return (void *)(mapping & ~PAGE_MAPPING_FLAGS);
44}
45
Linus Torvalds512b79312021-11-06 14:08:17 -070046void __acct_reclaim_writeback(pg_data_t *pgdat, struct folio *folio,
47 int nr_throttled);
48static inline void acct_reclaim_writeback(struct folio *folio)
49{
50 pg_data_t *pgdat = folio_pgdat(folio);
Mel Gorman8cd7c582021-11-05 13:42:25 -070051 int nr_throttled = atomic_read(&pgdat->nr_writeback_throttled);
52
53 if (nr_throttled)
Linus Torvalds512b79312021-11-06 14:08:17 -070054 __acct_reclaim_writeback(pgdat, folio, nr_throttled);
Mel Gorman8cd7c582021-11-05 13:42:25 -070055}
56
Mel Gormand818fca2021-11-05 13:42:29 -070057static inline void wake_throttle_isolated(pg_data_t *pgdat)
58{
59 wait_queue_head_t *wqh;
60
61 wqh = &pgdat->reclaim_wait[VMSCAN_THROTTLE_ISOLATED];
62 if (waitqueue_active(wqh))
63 wake_up(wqh);
64}
65
Souptick Joarder2b740302018-08-23 17:01:36 -070066vm_fault_t do_swap_page(struct vm_fault *vmf);
Matthew Wilcox (Oracle)575ced12020-12-08 01:25:39 -050067void folio_rotate_reclaimable(struct folio *folio);
Matthew Wilcox (Oracle)269ccca32021-01-15 23:34:16 -050068bool __folio_end_writeback(struct folio *folio);
Ebru Akagunduz8a966ed2016-07-26 15:25:03 -070069
Jan Beulich42b77722008-07-23 21:27:10 -070070void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
71 unsigned long floor, unsigned long ceiling);
Qi Zheng03c4f202021-11-05 13:38:38 -070072void pmd_install(struct mm_struct *mm, pmd_t *pmd, pgtable_t *pte);
Jan Beulich42b77722008-07-23 21:27:10 -070073
Minchan Kim9c276cc2019-09-25 16:49:08 -070074static inline bool can_madv_lru_vma(struct vm_area_struct *vma)
Kirill A. Shutemov23519072017-02-22 15:46:39 -080075{
76 return !(vma->vm_flags & (VM_LOCKED|VM_HUGETLB|VM_PFNMAP));
77}
78
Matthew Wilcox (Oracle)35066592021-11-28 14:53:35 -050079struct zap_details;
Michal Hockoaac45362016-03-25 14:20:24 -070080void unmap_page_range(struct mmu_gather *tlb,
81 struct vm_area_struct *vma,
82 unsigned long addr, unsigned long end,
83 struct zap_details *details);
84
David Howells7b3df3b2020-10-15 20:06:24 -070085void do_page_cache_ra(struct readahead_control *, unsigned long nr_to_read,
86 unsigned long lookahead_size);
Matthew Wilcox (Oracle)fcd9ae42021-04-07 21:18:55 +010087void force_page_cache_ra(struct readahead_control *, unsigned long nr);
David Howells7b3df3b2020-10-15 20:06:24 -070088static inline void force_page_cache_readahead(struct address_space *mapping,
89 struct file *file, pgoff_t index, unsigned long nr_to_read)
90{
Matthew Wilcox (Oracle)fcd9ae42021-04-07 21:18:55 +010091 DEFINE_READAHEAD(ractl, file, &file->f_ra, mapping, index);
92 force_page_cache_ra(&ractl, nr_to_read);
David Howells7b3df3b2020-10-15 20:06:24 -070093}
Fabian Frederick29f175d2014-04-07 15:37:55 -070094
Matthew Wilcox (Oracle)5c211ba2021-02-25 17:15:56 -080095unsigned find_lock_entries(struct address_space *mapping, pgoff_t start,
Matthew Wilcox (Oracle)51dcbda2021-12-07 14:15:07 -050096 pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices);
Matthew Wilcox (Oracle)0e499ed2020-09-01 23:17:50 -040097unsigned find_get_entries(struct address_space *mapping, pgoff_t start,
98 pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices);
Matthew Wilcox (Oracle)78f42662021-07-28 15:52:34 -040099void filemap_free_folio(struct address_space *mapping, struct folio *folio);
Matthew Wilcox (Oracle)1e84a3d2021-12-02 16:01:55 -0500100int truncate_inode_folio(struct address_space *mapping, struct folio *folio);
Matthew Wilcox (Oracle)b9a8a412020-05-27 17:59:22 -0400101bool truncate_inode_partial_folio(struct folio *folio, loff_t start,
102 loff_t end);
Matthew Wilcox (Oracle)5c211ba2021-02-25 17:15:56 -0800103
Yang Shi1eb62342020-04-01 21:06:20 -0700104/**
Matthew Wilcox (Oracle)3eed3ef2021-05-14 15:04:28 -0400105 * folio_evictable - Test whether a folio is evictable.
106 * @folio: The folio to test.
Yang Shi1eb62342020-04-01 21:06:20 -0700107 *
Matthew Wilcox (Oracle)3eed3ef2021-05-14 15:04:28 -0400108 * Test whether @folio is evictable -- i.e., should be placed on
109 * active/inactive lists vs unevictable list.
Yang Shi1eb62342020-04-01 21:06:20 -0700110 *
Matthew Wilcox (Oracle)3eed3ef2021-05-14 15:04:28 -0400111 * Reasons folio might not be evictable:
112 * 1. folio's mapping marked unevictable
113 * 2. One of the pages in the folio is part of an mlocked VMA
Yang Shi1eb62342020-04-01 21:06:20 -0700114 */
Matthew Wilcox (Oracle)3eed3ef2021-05-14 15:04:28 -0400115static inline bool folio_evictable(struct folio *folio)
116{
117 bool ret;
118
119 /* Prevent address_space of inode and swap cache from being freed */
120 rcu_read_lock();
121 ret = !mapping_unevictable(folio_mapping(folio)) &&
122 !folio_test_mlocked(folio);
123 rcu_read_unlock();
124 return ret;
125}
126
Yang Shi1eb62342020-04-01 21:06:20 -0700127static inline bool page_evictable(struct page *page)
128{
129 bool ret;
130
131 /* Prevent address_space of inode and swap cache from being freed */
132 rcu_read_lock();
133 ret = !mapping_unevictable(page_mapping(page)) && !PageMlocked(page);
134 rcu_read_unlock();
135 return ret;
136}
137
Nick Piggin7835e982006-03-22 00:08:40 -0800138/*
Joonsoo Kim0139aa72016-05-19 17:10:49 -0700139 * Turn a non-refcounted page (->_refcount == 0) into refcounted with
Nick Piggin7835e982006-03-22 00:08:40 -0800140 * a count of one.
141 */
142static inline void set_page_refcounted(struct page *page)
143{
Sasha Levin309381fea2014-01-23 15:52:54 -0800144 VM_BUG_ON_PAGE(PageTail(page), page);
Joonsoo Kimfe896d12016-03-17 14:19:26 -0700145 VM_BUG_ON_PAGE(page_ref_count(page), page);
Nick Piggin77a8a782006-01-06 00:10:57 -0800146 set_page_count(page, 1);
Nick Piggin77a8a782006-01-06 00:10:57 -0800147}
148
Hugh Dickins03f64622009-09-21 17:03:35 -0700149extern unsigned long highest_memmap_pfn;
150
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700151/*
Johannes Weinerc73322d2017-05-03 14:51:51 -0700152 * Maximum number of reclaim retries without progress before the OOM
153 * killer is consider the only way forward.
154 */
155#define MAX_RECLAIM_RETRIES 16
156
157/*
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700158 * in mm/vmscan.c:
159 */
Nick Piggin62695a82008-10-18 20:26:09 -0700160extern int isolate_lru_page(struct page *page);
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700161extern void putback_lru_page(struct page *page);
Mel Gormanc3f4a9a2021-11-05 13:42:42 -0700162extern void reclaim_throttle(pg_data_t *pgdat, enum vmscan_throttle_state reason);
Nick Piggin62695a82008-10-18 20:26:09 -0700163
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700164/*
Bob Liu62190492012-12-11 16:00:37 -0800165 * in mm/rmap.c:
166 */
167extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
168
169/*
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700170 * in mm/page_alloc.c
171 */
Joonsoo Kim3c605092014-11-13 15:19:21 -0800172
173/*
Vlastimil Babka1a6d53a2015-02-11 15:25:44 -0800174 * Structure for holding the mostly immutable allocation parameters passed
175 * between functions involved in allocations, including the alloc_pages*
176 * family of functions.
177 *
Joonsoo Kim97a225e2020-06-03 15:59:01 -0700178 * nodemask, migratetype and highest_zoneidx are initialized only once in
Matthew Wilcox (Oracle)84172f42021-04-29 23:01:15 -0700179 * __alloc_pages() and then never change.
Vlastimil Babka1a6d53a2015-02-11 15:25:44 -0800180 *
Joonsoo Kim97a225e2020-06-03 15:59:01 -0700181 * zonelist, preferred_zone and highest_zoneidx are set first in
Matthew Wilcox (Oracle)84172f42021-04-29 23:01:15 -0700182 * __alloc_pages() for the fast path, and might be later changed
Ethon Paul68956cc2020-06-04 16:49:31 -0700183 * in __alloc_pages_slowpath(). All other functions pass the whole structure
Vlastimil Babka1a6d53a2015-02-11 15:25:44 -0800184 * by a const pointer.
185 */
186struct alloc_context {
187 struct zonelist *zonelist;
188 nodemask_t *nodemask;
Mel Gormanc33d6c02016-05-19 17:14:10 -0700189 struct zoneref *preferred_zoneref;
Vlastimil Babka1a6d53a2015-02-11 15:25:44 -0800190 int migratetype;
Joonsoo Kim97a225e2020-06-03 15:59:01 -0700191
192 /*
193 * highest_zoneidx represents highest usable zone index of
194 * the allocation request. Due to the nature of the zone,
195 * memory on lower zone than the highest_zoneidx will be
196 * protected by lowmem_reserve[highest_zoneidx].
197 *
198 * highest_zoneidx is also used by reclaim/compaction to limit
199 * the target zone since higher zone than this index cannot be
200 * usable for this allocation request.
201 */
202 enum zone_type highest_zoneidx;
Mel Gormanc9ab0c42015-11-06 16:28:12 -0800203 bool spread_dirty_pages;
Vlastimil Babka1a6d53a2015-02-11 15:25:44 -0800204};
205
206/*
Joonsoo Kim3c605092014-11-13 15:19:21 -0800207 * Locate the struct page for both the matching buddy in our
208 * pair (buddy1) and the combined O(n+1) page they form (page).
209 *
210 * 1) Any buddy B1 will have an order O twin B2 which satisfies
211 * the following equation:
212 * B2 = B1 ^ (1 << O)
213 * For example, if the starting buddy (buddy2) is #8 its order
214 * 1 buddy is #10:
215 * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
216 *
217 * 2) Any buddy B will have an order O+1 parent P which
218 * satisfies the following equation:
219 * P = B & ~(1 << O)
220 *
221 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
222 */
223static inline unsigned long
Vlastimil Babka76741e72017-02-22 15:41:48 -0800224__find_buddy_pfn(unsigned long page_pfn, unsigned int order)
Joonsoo Kim3c605092014-11-13 15:19:21 -0800225{
Vlastimil Babka76741e72017-02-22 15:41:48 -0800226 return page_pfn ^ (1 << order);
Joonsoo Kim3c605092014-11-13 15:19:21 -0800227}
228
Joonsoo Kim7cf91a92016-03-15 14:57:51 -0700229extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn,
230 unsigned long end_pfn, struct zone *zone);
231
232static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn,
233 unsigned long end_pfn, struct zone *zone)
234{
235 if (zone->contiguous)
236 return pfn_to_page(start_pfn);
237
238 return __pageblock_pfn_to_page(start_pfn, end_pfn, zone);
239}
240
Joonsoo Kim3c605092014-11-13 15:19:21 -0800241extern int __isolate_free_page(struct page *page, unsigned int order);
Alexander Duyck624f58d2020-04-06 20:04:53 -0700242extern void __putback_isolated_page(struct page *page, unsigned int order,
243 int mt);
Mike Rapoport7c2ee342018-10-30 15:09:36 -0700244extern void memblock_free_pages(struct page *page, unsigned long pfn,
Mel Gormand70ddd72015-06-30 14:56:52 -0700245 unsigned int order);
Arun KSa9cd4102019-03-05 15:42:14 -0800246extern void __free_pages_core(struct page *page, unsigned int order);
Kirill A. Shutemovd00181b2015-11-06 16:29:57 -0800247extern void prep_compound_page(struct page *page, unsigned int order);
Joonsoo Kim46f24fd2016-07-26 15:23:58 -0700248extern void post_alloc_hook(struct page *page, unsigned int order,
249 gfp_t gfp_flags);
Han Pingtian42aa83c2014-01-23 15:53:28 -0800250extern int user_min_free_kbytes;
Wu Fengguang20a03072009-06-16 15:32:22 -0700251
Mel Gorman44042b42021-06-28 19:43:08 -0700252extern void free_unref_page(struct page *page, unsigned int order);
Matthew Wilcox (Oracle)0966aeb2020-12-14 19:08:02 -0800253extern void free_unref_page_list(struct list_head *list);
254
Mel Gorman04f8cfe2021-06-28 19:42:15 -0700255extern void zone_pcp_update(struct zone *zone, int cpu_online);
Mel Gorman68265392019-11-30 17:55:15 -0800256extern void zone_pcp_reset(struct zone *zone);
Vlastimil Babkaec6e8c7e2020-12-14 19:10:59 -0800257extern void zone_pcp_disable(struct zone *zone);
258extern void zone_pcp_enable(struct zone *zone);
Mel Gorman68265392019-11-30 17:55:15 -0800259
Mike Rapoportc803b3c2021-09-02 14:58:02 -0700260extern void *memmap_alloc(phys_addr_t size, phys_addr_t align,
261 phys_addr_t min_addr,
262 int nid, bool exact_nid);
263
Michal Nazarewiczff9543f2011-12-29 13:09:50 +0100264#if defined CONFIG_COMPACTION || defined CONFIG_CMA
265
266/*
267 * in mm/compaction.c
268 */
269/*
270 * compact_control is used to track pages being migrated and the free pages
271 * they are being migrated to during memory compaction. The free_pfn starts
272 * at the end of a zone and migrate_pfn begins at the start. Movable pages
273 * are moved to the end of a zone during a compaction run and the run
274 * completes when free_pfn <= migrate_pfn
275 */
276struct compact_control {
277 struct list_head freepages; /* List of free pages to migrate to */
278 struct list_head migratepages; /* List of pages being migrated */
Mel Gormanc5fbd932019-03-05 15:44:25 -0800279 unsigned int nr_freepages; /* Number of isolated free pages */
280 unsigned int nr_migratepages; /* Number of pages to migrate */
Michal Nazarewiczff9543f2011-12-29 13:09:50 +0100281 unsigned long free_pfn; /* isolate_freepages search base */
Oscar Salvadorc2ad7a12021-05-04 18:35:17 -0700282 /*
283 * Acts as an in/out parameter to page isolation for migration.
284 * isolate_migratepages uses it as a search base.
285 * isolate_migratepages_block will update the value to the next pfn
286 * after the last isolated one.
287 */
288 unsigned long migrate_pfn;
Mel Gorman70b44592019-03-05 15:44:54 -0800289 unsigned long fast_start_pfn; /* a pfn to start linear scan from */
Mel Gormanc5943b92019-03-05 15:44:28 -0800290 struct zone *zone;
291 unsigned long total_migrate_scanned;
292 unsigned long total_free_scanned;
Mel Gormandbe2d4e2019-03-05 15:45:31 -0800293 unsigned short fast_search_fail;/* failures to use free list searches */
294 short search_order; /* order to start a fast search at */
Vlastimil Babkaf25ba6d2017-05-08 15:54:30 -0700295 const gfp_t gfp_mask; /* gfp mask of a direct compactor */
296 int order; /* order a direct compactor needs */
Vlastimil Babkad39773a2017-05-08 15:54:46 -0700297 int migratetype; /* migratetype of direct compactor */
Vlastimil Babkaf25ba6d2017-05-08 15:54:30 -0700298 const unsigned int alloc_flags; /* alloc flags of a direct compactor */
Joonsoo Kim97a225e2020-06-03 15:59:01 -0700299 const int highest_zoneidx; /* zone index of a direct compactor */
David Rientjese0b9dae2014-06-04 16:08:28 -0700300 enum migrate_mode mode; /* Async or sync migration mode */
Mel Gormanbb13ffe2012-10-08 16:32:41 -0700301 bool ignore_skip_hint; /* Scan blocks even if marked skip */
Vlastimil Babka2583d672017-11-17 15:26:38 -0800302 bool no_set_skip_hint; /* Don't mark blocks for skipping */
Vlastimil Babka9f7e3382016-10-07 17:00:37 -0700303 bool ignore_block_suitable; /* Scan blocks considered unsuitable */
Vlastimil Babkaaccf6242016-03-17 14:18:15 -0700304 bool direct_compaction; /* False from kcompactd or /proc/... */
Nitin Guptafacdaa92020-08-11 18:31:00 -0700305 bool proactive_compaction; /* kcompactd proactive compaction */
Vlastimil Babka06ed2992016-10-07 16:57:35 -0700306 bool whole_zone; /* Whole zone should/has been scanned */
Vlastimil Babkac3486f52016-07-28 15:49:30 -0700307 bool contended; /* Signal lock or sched contention */
Mel Gorman804d3122019-03-05 15:45:07 -0800308 bool rescan; /* Rescanning the same pageblock */
Rik van Rielb06eda02020-04-01 21:10:28 -0700309 bool alloc_contig; /* alloc_contig_range allocation */
Michal Nazarewiczff9543f2011-12-29 13:09:50 +0100310};
311
Mel Gorman5e1f0f02019-03-05 15:45:41 -0800312/*
313 * Used in direct compaction when a page should be taken from the freelists
314 * immediately when one is created during the free path.
315 */
316struct capture_control {
317 struct compact_control *cc;
318 struct page *page;
319};
320
Michal Nazarewiczff9543f2011-12-29 13:09:50 +0100321unsigned long
Mel Gormanbb13ffe2012-10-08 16:32:41 -0700322isolate_freepages_range(struct compact_control *cc,
323 unsigned long start_pfn, unsigned long end_pfn);
Oscar Salvadorc2ad7a12021-05-04 18:35:17 -0700324int
Vlastimil Babkaedc2ca62014-10-09 15:27:09 -0700325isolate_migratepages_range(struct compact_control *cc,
326 unsigned long low_pfn, unsigned long end_pfn);
Mel Gormanffd8f252021-06-30 18:53:53 -0700327#endif
Joonsoo Kim2149cda2015-04-14 15:45:21 -0700328int find_suitable_fallback(struct free_area *area, unsigned int order,
329 int migratetype, bool only_stealable, bool *can_steal);
Michal Nazarewiczff9543f2011-12-29 13:09:50 +0100330
Mel Gorman48f13bf2007-10-16 01:26:10 -0700331/*
Mel Gorman6c144662014-01-23 15:53:38 -0800332 * This function returns the order of a free page in the buddy system. In
333 * general, page_zone(page)->lock must be held by the caller to prevent the
334 * page from being allocated in parallel and returning garbage as the order.
335 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
Vlastimil Babka99c0fd52014-10-09 15:27:23 -0700336 * page cannot be allocated or merged in parallel. Alternatively, it must
Matthew Wilcox (Oracle)ab130f912020-10-15 20:10:15 -0700337 * handle invalid values gracefully, and use buddy_order_unsafe() below.
Mel Gorman48f13bf2007-10-16 01:26:10 -0700338 */
Matthew Wilcox (Oracle)ab130f912020-10-15 20:10:15 -0700339static inline unsigned int buddy_order(struct page *page)
Mel Gorman48f13bf2007-10-16 01:26:10 -0700340{
KAMEZAWA Hiroyuki572438f2010-10-26 14:22:08 -0700341 /* PageBuddy() must be checked by the caller */
Mel Gorman48f13bf2007-10-16 01:26:10 -0700342 return page_private(page);
343}
Alexander van Heukelumb5a0e012008-02-23 15:24:06 -0800344
Vlastimil Babka99c0fd52014-10-09 15:27:23 -0700345/*
Matthew Wilcox (Oracle)ab130f912020-10-15 20:10:15 -0700346 * Like buddy_order(), but for callers who cannot afford to hold the zone lock.
Vlastimil Babka99c0fd52014-10-09 15:27:23 -0700347 * PageBuddy() should be checked first by the caller to minimize race window,
348 * and invalid values must be handled gracefully.
349 *
Jason Low4db0c3c2015-04-15 16:14:08 -0700350 * READ_ONCE is used so that if the caller assigns the result into a local
Vlastimil Babka99c0fd52014-10-09 15:27:23 -0700351 * variable and e.g. tests it for valid range before using, the compiler cannot
352 * decide to remove the variable and inline the page_private(page) multiple
353 * times, potentially observing different values in the tests and the actual
354 * use of the result.
355 */
Matthew Wilcox (Oracle)ab130f912020-10-15 20:10:15 -0700356#define buddy_order_unsafe(page) READ_ONCE(page_private(page))
Vlastimil Babka99c0fd52014-10-09 15:27:23 -0700357
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800358/*
359 * These three helpers classifies VMAs for virtual memory accounting.
360 */
361
362/*
363 * Executable code area - executable, not writable, not stack
364 */
Konstantin Khlebnikovd977d562016-02-02 16:57:43 -0800365static inline bool is_exec_mapping(vm_flags_t flags)
366{
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800367 return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC;
Konstantin Khlebnikovd977d562016-02-02 16:57:43 -0800368}
369
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800370/*
Ingo Molnarf0953a12021-05-06 18:06:47 -0700371 * Stack area - automatically grows in one direction
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800372 *
373 * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
374 * do_mmap() forbids all other combinations.
375 */
Konstantin Khlebnikovd977d562016-02-02 16:57:43 -0800376static inline bool is_stack_mapping(vm_flags_t flags)
377{
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800378 return (flags & VM_STACK) == VM_STACK;
Konstantin Khlebnikovd977d562016-02-02 16:57:43 -0800379}
380
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800381/*
382 * Data area - private, writable, not stack
383 */
Konstantin Khlebnikovd977d562016-02-02 16:57:43 -0800384static inline bool is_data_mapping(vm_flags_t flags)
385{
Konstantin Khlebnikov30bdbb72016-02-02 16:57:46 -0800386 return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE;
Konstantin Khlebnikovd977d562016-02-02 16:57:43 -0800387}
388
Namhyung Kim6038def2011-05-24 17:11:22 -0700389/* mm/util.c */
390void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
Wei Yangaba6dfb2019-11-30 17:50:53 -0800391 struct vm_area_struct *prev);
Wei Yang1b9fc5b22019-11-30 17:50:49 -0800392void __vma_unlink_list(struct mm_struct *mm, struct vm_area_struct *vma);
Namhyung Kim6038def2011-05-24 17:11:22 -0700393
Hugh Dickinsaf8e3352009-12-14 17:58:59 -0800394#ifdef CONFIG_MMU
Matthew Wilcox (Oracle)35066592021-11-28 14:53:35 -0500395void unmap_mapping_folio(struct folio *folio);
Kirill A. Shutemovfc05f562015-04-14 15:44:39 -0700396extern long populate_vma_page_range(struct vm_area_struct *vma,
David Hildenbranda78f1cc2021-06-30 18:52:23 -0700397 unsigned long start, unsigned long end, int *locked);
David Hildenbrand4ca9b3852021-06-30 18:52:28 -0700398extern long faultin_vma_page_range(struct vm_area_struct *vma,
399 unsigned long start, unsigned long end,
400 bool write, int *locked);
Hugh Dickinsaf8e3352009-12-14 17:58:59 -0800401extern void munlock_vma_pages_range(struct vm_area_struct *vma,
402 unsigned long start, unsigned long end);
403static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
404{
405 munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
406}
407
Nick Pigginb291f002008-10-18 20:26:44 -0700408/*
Michel Lespinassec1e8d7c2020-06-08 21:33:54 -0700409 * must be called with vma's mmap_lock held for read or write, and page locked.
Nick Pigginb291f002008-10-18 20:26:44 -0700410 */
411extern void mlock_vma_page(struct page *page);
Michel Lespinasseff6a6da2013-02-27 17:02:44 -0800412extern unsigned int munlock_vma_page(struct page *page);
Nick Pigginb291f002008-10-18 20:26:44 -0700413
Mike Rapoport6aeb2542021-07-07 18:07:50 -0700414extern int mlock_future_check(struct mm_struct *mm, unsigned long flags,
415 unsigned long len);
416
Nick Pigginb291f002008-10-18 20:26:44 -0700417/*
418 * Clear the page's PageMlocked(). This can be useful in a situation where
419 * we want to unconditionally remove a page from the pagecache -- e.g.,
420 * on truncation or freeing.
421 *
422 * It is legal to call this function for any page, mlocked or not.
423 * If called for a page that is still mapped by mlocked vmas, all we do
424 * is revert to lazy LRU behaviour -- semantics are not broken.
425 */
Hugh Dickinse6c509f2012-10-08 16:33:19 -0700426extern void clear_page_mlock(struct page *page);
Nick Pigginb291f002008-10-18 20:26:44 -0700427
Linus Torvaldsf55e1012017-11-29 09:01:01 -0800428extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
Mel Gormanb32967f2012-11-19 12:35:47 +0000429
Kirill A. Shutemove9b61f12016-01-15 16:54:10 -0800430/*
Hugh Dickins494334e2021-06-15 18:23:56 -0700431 * At what user virtual address is page expected in vma?
432 * Returns -EFAULT if all of the page is outside the range of vma.
433 * If page is a compound head, the entire compound page is considered.
Kirill A. Shutemove9b61f12016-01-15 16:54:10 -0800434 */
435static inline unsigned long
Kirill A. Shutemove9b61f12016-01-15 16:54:10 -0800436vma_address(struct page *page, struct vm_area_struct *vma)
437{
Hugh Dickins494334e2021-06-15 18:23:56 -0700438 pgoff_t pgoff;
439 unsigned long address;
Kirill A. Shutemova8fa41ad2017-02-24 14:57:54 -0800440
Hugh Dickins494334e2021-06-15 18:23:56 -0700441 VM_BUG_ON_PAGE(PageKsm(page), page); /* KSM page->index unusable */
442 pgoff = page_to_pgoff(page);
443 if (pgoff >= vma->vm_pgoff) {
444 address = vma->vm_start +
445 ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
446 /* Check for address beyond vma (or wrapped through 0?) */
447 if (address < vma->vm_start || address >= vma->vm_end)
448 address = -EFAULT;
449 } else if (PageHead(page) &&
450 pgoff + compound_nr(page) - 1 >= vma->vm_pgoff) {
451 /* Test above avoids possibility of wrap to 0 on 32-bit */
452 address = vma->vm_start;
453 } else {
454 address = -EFAULT;
455 }
456 return address;
457}
Kirill A. Shutemove9b61f12016-01-15 16:54:10 -0800458
Hugh Dickins494334e2021-06-15 18:23:56 -0700459/*
460 * Then at what user virtual address will none of the page be found in vma?
461 * Assumes that vma_address() already returned a good starting address.
462 * If page is a compound head, the entire compound page is considered.
463 */
464static inline unsigned long
465vma_address_end(struct page *page, struct vm_area_struct *vma)
466{
467 pgoff_t pgoff;
468 unsigned long address;
Kirill A. Shutemove9b61f12016-01-15 16:54:10 -0800469
Hugh Dickins494334e2021-06-15 18:23:56 -0700470 VM_BUG_ON_PAGE(PageKsm(page), page); /* KSM page->index unusable */
471 pgoff = page_to_pgoff(page) + compound_nr(page);
472 address = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
473 /* Check for address beyond vma (or wrapped through 0?) */
474 if (address < vma->vm_start || address > vma->vm_end)
475 address = vma->vm_end;
476 return address;
Kirill A. Shutemove9b61f12016-01-15 16:54:10 -0800477}
478
Johannes Weiner89b15332019-11-30 17:50:22 -0800479static inline struct file *maybe_unlock_mmap_for_io(struct vm_fault *vmf,
480 struct file *fpin)
481{
482 int flags = vmf->flags;
483
484 if (fpin)
485 return fpin;
486
487 /*
488 * FAULT_FLAG_RETRY_NOWAIT means we don't want to wait on page locks or
Michel Lespinassec1e8d7c2020-06-08 21:33:54 -0700489 * anything, so we only pin the file and drop the mmap_lock if only
Peter Xu4064b982020-04-01 21:08:45 -0700490 * FAULT_FLAG_ALLOW_RETRY is set, while this is the first attempt.
Johannes Weiner89b15332019-11-30 17:50:22 -0800491 */
Peter Xu4064b982020-04-01 21:08:45 -0700492 if (fault_flag_allow_retry_first(flags) &&
493 !(flags & FAULT_FLAG_RETRY_NOWAIT)) {
Johannes Weiner89b15332019-11-30 17:50:22 -0800494 fpin = get_file(vmf->vma->vm_file);
Michel Lespinassed8ed45c2020-06-08 21:33:25 -0700495 mmap_read_unlock(vmf->vma->vm_mm);
Johannes Weiner89b15332019-11-30 17:50:22 -0800496 }
497 return fpin;
498}
Hugh Dickinsaf8e3352009-12-14 17:58:59 -0800499#else /* !CONFIG_MMU */
Matthew Wilcox (Oracle)35066592021-11-28 14:53:35 -0500500static inline void unmap_mapping_folio(struct folio *folio) { }
Nick Pigginb291f002008-10-18 20:26:44 -0700501static inline void clear_page_mlock(struct page *page) { }
502static inline void mlock_vma_page(struct page *page) { }
Nicholas Piggin4ad0ae82021-04-29 22:59:01 -0700503static inline void vunmap_range_noflush(unsigned long start, unsigned long end)
504{
505}
Hugh Dickinsaf8e3352009-12-14 17:58:59 -0800506#endif /* !CONFIG_MMU */
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700507
Alexander van Heukelumb5a0e012008-02-23 15:24:06 -0800508/*
Andy Whitcroft69d177c2008-11-06 12:53:26 -0800509 * Return the mem_map entry representing the 'offset' subpage within
510 * the maximally aligned gigantic page 'base'. Handle any discontiguity
511 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
512 */
513static inline struct page *mem_map_offset(struct page *base, int offset)
514{
515 if (unlikely(offset >= MAX_ORDER_NR_PAGES))
Fabian Frederickbc7f84c2014-08-06 16:05:17 -0700516 return nth_page(base, offset);
Andy Whitcroft69d177c2008-11-06 12:53:26 -0800517 return base + offset;
518}
519
520/*
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300521 * Iterator over all subpages within the maximally aligned gigantic
Andy Whitcroft69d177c2008-11-06 12:53:26 -0800522 * page 'base'. Handle any discontiguity in the mem_map.
523 */
524static inline struct page *mem_map_next(struct page *iter,
525 struct page *base, int offset)
526{
527 if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
528 unsigned long pfn = page_to_pfn(base) + offset;
529 if (!pfn_valid(pfn))
530 return NULL;
531 return pfn_to_page(pfn);
532 }
533 return iter + 1;
534}
535
Mel Gorman6b74ab92008-07-23 21:26:49 -0700536/* Memory initialisation debug and verification */
537enum mminit_level {
538 MMINIT_WARNING,
539 MMINIT_VERIFY,
540 MMINIT_TRACE
541};
542
543#ifdef CONFIG_DEBUG_MEMORY_INIT
544
545extern int mminit_loglevel;
546
547#define mminit_dprintk(level, prefix, fmt, arg...) \
548do { \
549 if (level < mminit_loglevel) { \
Rasmus Villemoesfc5199d2015-02-12 15:00:02 -0800550 if (level <= MMINIT_WARNING) \
Joe Perches11705322016-03-17 14:19:50 -0700551 pr_warn("mminit::" prefix " " fmt, ##arg); \
Rasmus Villemoesfc5199d2015-02-12 15:00:02 -0800552 else \
553 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
Mel Gorman6b74ab92008-07-23 21:26:49 -0700554 } \
555} while (0)
556
Mel Gorman708614e2008-07-23 21:26:51 -0700557extern void mminit_verify_pageflags_layout(void);
Mel Gorman68ad8df2008-07-23 21:26:52 -0700558extern void mminit_verify_zonelist(void);
Mel Gorman6b74ab92008-07-23 21:26:49 -0700559#else
560
561static inline void mminit_dprintk(enum mminit_level level,
562 const char *prefix, const char *fmt, ...)
563{
564}
565
Mel Gorman708614e2008-07-23 21:26:51 -0700566static inline void mminit_verify_pageflags_layout(void)
567{
568}
569
Mel Gorman68ad8df2008-07-23 21:26:52 -0700570static inline void mminit_verify_zonelist(void)
571{
572}
Mel Gorman6b74ab92008-07-23 21:26:49 -0700573#endif /* CONFIG_DEBUG_MEMORY_INIT */
Mel Gorman2dbb51c2008-07-23 21:26:52 -0700574
575/* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
576#if defined(CONFIG_SPARSEMEM)
577extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
578 unsigned long *end_pfn);
579#else
580static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
581 unsigned long *end_pfn)
582{
583}
584#endif /* CONFIG_SPARSEMEM */
585
Mel Gormana5f5f912016-07-28 15:46:32 -0700586#define NODE_RECLAIM_NOSCAN -2
587#define NODE_RECLAIM_FULL -1
588#define NODE_RECLAIM_SOME 0
589#define NODE_RECLAIM_SUCCESS 1
Wu Fengguang7c116f22009-12-16 12:19:59 +0100590
Wei Yang8b095492018-12-28 00:34:36 -0800591#ifdef CONFIG_NUMA
592extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int);
Dave Hansen79c28a42021-09-02 14:59:06 -0700593extern int find_next_best_node(int node, nodemask_t *used_node_mask);
Wei Yang8b095492018-12-28 00:34:36 -0800594#else
595static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask,
596 unsigned int order)
597{
598 return NODE_RECLAIM_NOSCAN;
599}
Dave Hansen79c28a42021-09-02 14:59:06 -0700600static inline int find_next_best_node(int node, nodemask_t *used_node_mask)
601{
602 return NUMA_NO_NODE;
603}
Wei Yang8b095492018-12-28 00:34:36 -0800604#endif
605
Wu Fengguang31d3d342009-12-16 12:19:59 +0100606extern int hwpoison_filter(struct page *p);
607
Wu Fengguang7c116f22009-12-16 12:19:59 +0100608extern u32 hwpoison_filter_dev_major;
609extern u32 hwpoison_filter_dev_minor;
Wu Fengguang478c5ff2009-12-16 12:19:59 +0100610extern u64 hwpoison_filter_flags_mask;
611extern u64 hwpoison_filter_flags_value;
Andi Kleen4fd466e2009-12-16 12:19:59 +0100612extern u64 hwpoison_filter_memcg;
Haicheng Li1bfe5fe2009-12-16 12:19:59 +0100613extern u32 hwpoison_filter_enable;
Al Viroeb36c582012-05-30 20:17:35 -0400614
Michal Hockodc0ef0d2016-05-23 16:25:27 -0700615extern unsigned long __must_check vm_mmap_pgoff(struct file *, unsigned long,
Al Viroeb36c582012-05-30 20:17:35 -0400616 unsigned long, unsigned long,
Michal Hocko9fbeb5a2016-05-23 16:25:30 -0700617 unsigned long, unsigned long);
Xishi Qiuca57df72012-07-31 16:43:19 -0700618
619extern void set_pageblock_order(void);
Maninder Singh730ec8c2020-06-03 16:01:18 -0700620unsigned int reclaim_clean_pages_from_list(struct zone *zone,
Minchan Kim02c6de82012-10-08 16:31:55 -0700621 struct list_head *page_list);
Bartlomiej Zolnierkiewiczd95ea5d2012-10-08 16:32:05 -0700622/* The ALLOC_WMARK bits are used as an index to zone->watermark */
623#define ALLOC_WMARK_MIN WMARK_MIN
624#define ALLOC_WMARK_LOW WMARK_LOW
625#define ALLOC_WMARK_HIGH WMARK_HIGH
626#define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
627
628/* Mask to get the watermark bits */
629#define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
630
Michal Hockocd04ae12017-09-06 16:24:50 -0700631/*
632 * Only MMU archs have async oom victim reclaim - aka oom_reaper so we
633 * cannot assume a reduced access to memory reserves is sufficient for
634 * !MMU
635 */
636#ifdef CONFIG_MMU
637#define ALLOC_OOM 0x08
638#else
639#define ALLOC_OOM ALLOC_NO_WATERMARKS
640#endif
641
Mel Gorman6bb15452018-12-28 00:35:41 -0800642#define ALLOC_HARDER 0x10 /* try to alloc harder */
643#define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
644#define ALLOC_CPUSET 0x40 /* check for correct cpuset */
645#define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
646#ifdef CONFIG_ZONE_DMA32
647#define ALLOC_NOFRAGMENT 0x100 /* avoid mixing pageblock types */
648#else
649#define ALLOC_NOFRAGMENT 0x0
650#endif
Mateusz Nosek736838e2020-04-01 21:09:47 -0700651#define ALLOC_KSWAPD 0x800 /* allow waking of kswapd, __GFP_KSWAPD_RECLAIM set */
Bartlomiej Zolnierkiewiczd95ea5d2012-10-08 16:32:05 -0700652
Mel Gorman72b252a2015-09-04 15:47:32 -0700653enum ttu_flags;
654struct tlbflush_unmap_batch;
655
Michal Hockoce612872017-04-07 16:05:05 -0700656
657/*
658 * only for MM internal work items which do not depend on
659 * any allocations or locks which might depend on allocations
660 */
661extern struct workqueue_struct *mm_percpu_wq;
662
Mel Gorman72b252a2015-09-04 15:47:32 -0700663#ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
664void try_to_unmap_flush(void);
Mel Gormand950c942015-09-04 15:47:35 -0700665void try_to_unmap_flush_dirty(void);
Mel Gorman3ea27712017-08-02 13:31:52 -0700666void flush_tlb_batched_pending(struct mm_struct *mm);
Mel Gorman72b252a2015-09-04 15:47:32 -0700667#else
668static inline void try_to_unmap_flush(void)
669{
670}
Mel Gormand950c942015-09-04 15:47:35 -0700671static inline void try_to_unmap_flush_dirty(void)
672{
673}
Mel Gorman3ea27712017-08-02 13:31:52 -0700674static inline void flush_tlb_batched_pending(struct mm_struct *mm)
675{
676}
Mel Gorman72b252a2015-09-04 15:47:32 -0700677#endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
Vlastimil Babkaedf14cd2016-03-15 14:55:56 -0700678
679extern const struct trace_print_flags pageflag_names[];
680extern const struct trace_print_flags vmaflag_names[];
681extern const struct trace_print_flags gfpflag_names[];
682
Xishi Qiua6ffdc02017-05-03 14:52:52 -0700683static inline bool is_migrate_highatomic(enum migratetype migratetype)
684{
685 return migratetype == MIGRATE_HIGHATOMIC;
686}
687
688static inline bool is_migrate_highatomic_page(struct page *page)
689{
690 return get_pageblock_migratetype(page) == MIGRATE_HIGHATOMIC;
691}
692
Michal Hocko72675e12017-09-06 16:20:24 -0700693void setup_zone_pageset(struct zone *zone);
Joonsoo Kim19fc7be2020-08-11 18:37:25 -0700694
695struct migration_target_control {
696 int nid; /* preferred node id */
697 nodemask_t *nmask;
698 gfp_t gfp_mask;
699};
700
Nicholas Pigginb67177e2021-04-29 22:58:53 -0700701/*
702 * mm/vmalloc.c
703 */
Nicholas Piggin4ad0ae82021-04-29 22:59:01 -0700704#ifdef CONFIG_MMU
Nicholas Pigginb67177e2021-04-29 22:58:53 -0700705int vmap_pages_range_noflush(unsigned long addr, unsigned long end,
706 pgprot_t prot, struct page **pages, unsigned int page_shift);
Nicholas Piggin4ad0ae82021-04-29 22:59:01 -0700707#else
708static inline
709int vmap_pages_range_noflush(unsigned long addr, unsigned long end,
710 pgprot_t prot, struct page **pages, unsigned int page_shift)
711{
712 return -EINVAL;
713}
714#endif
715
716void vunmap_range_noflush(unsigned long start, unsigned long end);
Nicholas Pigginb67177e2021-04-29 22:58:53 -0700717
Yang Shif4c0d832021-06-30 18:51:39 -0700718int numa_migrate_prep(struct page *page, struct vm_area_struct *vma,
719 unsigned long addr, int page_nid, int *flags);
720
Michel Lespinassedb971412012-10-08 16:29:34 -0700721#endif /* __MM_INTERNAL_H */