Thomas Gleixner | 2874c5f | 2019-05-27 08:55:01 +0200 | [diff] [blame] | 1 | /* SPDX-License-Identifier: GPL-2.0-or-later */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2 | /* internal.h: mm/ internal definitions |
| 3 | * |
| 4 | * Copyright (C) 2004 Red Hat, Inc. All Rights Reserved. |
| 5 | * Written by David Howells (dhowells@redhat.com) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 6 | */ |
Nick Piggin | 0f8053a | 2006-03-22 00:08:33 -0800 | [diff] [blame] | 7 | #ifndef __MM_INTERNAL_H |
| 8 | #define __MM_INTERNAL_H |
| 9 | |
Fabian Frederick | 29f175d | 2014-04-07 15:37:55 -0700 | [diff] [blame] | 10 | #include <linux/fs.h> |
Nick Piggin | 0f8053a | 2006-03-22 00:08:33 -0800 | [diff] [blame] | 11 | #include <linux/mm.h> |
Kirill A. Shutemov | e9b61f1 | 2016-01-15 16:54:10 -0800 | [diff] [blame] | 12 | #include <linux/pagemap.h> |
Vlastimil Babka | edf14cd | 2016-03-15 14:55:56 -0700 | [diff] [blame] | 13 | #include <linux/tracepoint-defs.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 14 | |
Matthew Wilcox (Oracle) | 0e499ed | 2020-09-01 23:17:50 -0400 | [diff] [blame] | 15 | struct folio_batch; |
| 16 | |
Mel Gorman | dd56b04 | 2015-11-06 16:28:43 -0800 | [diff] [blame] | 17 | /* |
| 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 Hocko | dcda9b0 | 2017-07-12 14:36:45 -0700 | [diff] [blame] | 24 | __GFP_NOWARN|__GFP_RETRY_MAYFAIL|__GFP_NOFAIL|\ |
Mel Gorman | e838a45 | 2016-06-24 14:49:37 -0700 | [diff] [blame] | 25 | __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC|\ |
Michal Hocko | 704687d | 2022-01-14 14:07:11 -0800 | [diff] [blame] | 26 | __GFP_ATOMIC|__GFP_NOLOCKDEP) |
Mel Gorman | dd56b04 | 2015-11-06 16:28:43 -0800 | [diff] [blame] | 27 | |
| 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 Piggin | 6290602 | 2016-12-25 13:00:30 +1000 | [diff] [blame] | 37 | void page_writeback_init(void); |
| 38 | |
Matthew Wilcox (Oracle) | 6460100 | 2021-05-07 11:17:34 -0400 | [diff] [blame] | 39 | static 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 Torvalds | 512b7931 | 2021-11-06 14:08:17 -0700 | [diff] [blame] | 46 | void __acct_reclaim_writeback(pg_data_t *pgdat, struct folio *folio, |
| 47 | int nr_throttled); |
| 48 | static inline void acct_reclaim_writeback(struct folio *folio) |
| 49 | { |
| 50 | pg_data_t *pgdat = folio_pgdat(folio); |
Mel Gorman | 8cd7c58 | 2021-11-05 13:42:25 -0700 | [diff] [blame] | 51 | int nr_throttled = atomic_read(&pgdat->nr_writeback_throttled); |
| 52 | |
| 53 | if (nr_throttled) |
Linus Torvalds | 512b7931 | 2021-11-06 14:08:17 -0700 | [diff] [blame] | 54 | __acct_reclaim_writeback(pgdat, folio, nr_throttled); |
Mel Gorman | 8cd7c58 | 2021-11-05 13:42:25 -0700 | [diff] [blame] | 55 | } |
| 56 | |
Mel Gorman | d818fca | 2021-11-05 13:42:29 -0700 | [diff] [blame] | 57 | static 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 Joarder | 2b74030 | 2018-08-23 17:01:36 -0700 | [diff] [blame] | 66 | vm_fault_t do_swap_page(struct vm_fault *vmf); |
Matthew Wilcox (Oracle) | 575ced1 | 2020-12-08 01:25:39 -0500 | [diff] [blame] | 67 | void folio_rotate_reclaimable(struct folio *folio); |
Matthew Wilcox (Oracle) | 269ccca3 | 2021-01-15 23:34:16 -0500 | [diff] [blame] | 68 | bool __folio_end_writeback(struct folio *folio); |
Ebru Akagunduz | 8a966ed | 2016-07-26 15:25:03 -0700 | [diff] [blame] | 69 | |
Jan Beulich | 42b7772 | 2008-07-23 21:27:10 -0700 | [diff] [blame] | 70 | void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma, |
| 71 | unsigned long floor, unsigned long ceiling); |
Qi Zheng | 03c4f20 | 2021-11-05 13:38:38 -0700 | [diff] [blame] | 72 | void pmd_install(struct mm_struct *mm, pmd_t *pmd, pgtable_t *pte); |
Jan Beulich | 42b7772 | 2008-07-23 21:27:10 -0700 | [diff] [blame] | 73 | |
Minchan Kim | 9c276cc | 2019-09-25 16:49:08 -0700 | [diff] [blame] | 74 | static inline bool can_madv_lru_vma(struct vm_area_struct *vma) |
Kirill A. Shutemov | 2351907 | 2017-02-22 15:46:39 -0800 | [diff] [blame] | 75 | { |
| 76 | return !(vma->vm_flags & (VM_LOCKED|VM_HUGETLB|VM_PFNMAP)); |
| 77 | } |
| 78 | |
Matthew Wilcox (Oracle) | 3506659 | 2021-11-28 14:53:35 -0500 | [diff] [blame] | 79 | struct zap_details; |
Michal Hocko | aac4536 | 2016-03-25 14:20:24 -0700 | [diff] [blame] | 80 | void 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 Howells | 7b3df3b | 2020-10-15 20:06:24 -0700 | [diff] [blame] | 85 | void do_page_cache_ra(struct readahead_control *, unsigned long nr_to_read, |
| 86 | unsigned long lookahead_size); |
Matthew Wilcox (Oracle) | fcd9ae4 | 2021-04-07 21:18:55 +0100 | [diff] [blame] | 87 | void force_page_cache_ra(struct readahead_control *, unsigned long nr); |
David Howells | 7b3df3b | 2020-10-15 20:06:24 -0700 | [diff] [blame] | 88 | static 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) | fcd9ae4 | 2021-04-07 21:18:55 +0100 | [diff] [blame] | 91 | DEFINE_READAHEAD(ractl, file, &file->f_ra, mapping, index); |
| 92 | force_page_cache_ra(&ractl, nr_to_read); |
David Howells | 7b3df3b | 2020-10-15 20:06:24 -0700 | [diff] [blame] | 93 | } |
Fabian Frederick | 29f175d | 2014-04-07 15:37:55 -0700 | [diff] [blame] | 94 | |
Matthew Wilcox (Oracle) | 5c211ba | 2021-02-25 17:15:56 -0800 | [diff] [blame] | 95 | unsigned find_lock_entries(struct address_space *mapping, pgoff_t start, |
Matthew Wilcox (Oracle) | 51dcbda | 2021-12-07 14:15:07 -0500 | [diff] [blame] | 96 | pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices); |
Matthew Wilcox (Oracle) | 0e499ed | 2020-09-01 23:17:50 -0400 | [diff] [blame] | 97 | unsigned find_get_entries(struct address_space *mapping, pgoff_t start, |
| 98 | pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices); |
Matthew Wilcox (Oracle) | 78f4266 | 2021-07-28 15:52:34 -0400 | [diff] [blame] | 99 | void filemap_free_folio(struct address_space *mapping, struct folio *folio); |
Matthew Wilcox (Oracle) | 1e84a3d | 2021-12-02 16:01:55 -0500 | [diff] [blame] | 100 | int truncate_inode_folio(struct address_space *mapping, struct folio *folio); |
Matthew Wilcox (Oracle) | b9a8a41 | 2020-05-27 17:59:22 -0400 | [diff] [blame] | 101 | bool truncate_inode_partial_folio(struct folio *folio, loff_t start, |
| 102 | loff_t end); |
Matthew Wilcox (Oracle) | 5c211ba | 2021-02-25 17:15:56 -0800 | [diff] [blame] | 103 | |
Yang Shi | 1eb6234 | 2020-04-01 21:06:20 -0700 | [diff] [blame] | 104 | /** |
Matthew Wilcox (Oracle) | 3eed3ef | 2021-05-14 15:04:28 -0400 | [diff] [blame] | 105 | * folio_evictable - Test whether a folio is evictable. |
| 106 | * @folio: The folio to test. |
Yang Shi | 1eb6234 | 2020-04-01 21:06:20 -0700 | [diff] [blame] | 107 | * |
Matthew Wilcox (Oracle) | 3eed3ef | 2021-05-14 15:04:28 -0400 | [diff] [blame] | 108 | * Test whether @folio is evictable -- i.e., should be placed on |
| 109 | * active/inactive lists vs unevictable list. |
Yang Shi | 1eb6234 | 2020-04-01 21:06:20 -0700 | [diff] [blame] | 110 | * |
Matthew Wilcox (Oracle) | 3eed3ef | 2021-05-14 15:04:28 -0400 | [diff] [blame] | 111 | * 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 Shi | 1eb6234 | 2020-04-01 21:06:20 -0700 | [diff] [blame] | 114 | */ |
Matthew Wilcox (Oracle) | 3eed3ef | 2021-05-14 15:04:28 -0400 | [diff] [blame] | 115 | static 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 Shi | 1eb6234 | 2020-04-01 21:06:20 -0700 | [diff] [blame] | 127 | static 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 Piggin | 7835e98 | 2006-03-22 00:08:40 -0800 | [diff] [blame] | 138 | /* |
Joonsoo Kim | 0139aa7 | 2016-05-19 17:10:49 -0700 | [diff] [blame] | 139 | * Turn a non-refcounted page (->_refcount == 0) into refcounted with |
Nick Piggin | 7835e98 | 2006-03-22 00:08:40 -0800 | [diff] [blame] | 140 | * a count of one. |
| 141 | */ |
| 142 | static inline void set_page_refcounted(struct page *page) |
| 143 | { |
Sasha Levin | 309381fea | 2014-01-23 15:52:54 -0800 | [diff] [blame] | 144 | VM_BUG_ON_PAGE(PageTail(page), page); |
Joonsoo Kim | fe896d1 | 2016-03-17 14:19:26 -0700 | [diff] [blame] | 145 | VM_BUG_ON_PAGE(page_ref_count(page), page); |
Nick Piggin | 77a8a78 | 2006-01-06 00:10:57 -0800 | [diff] [blame] | 146 | set_page_count(page, 1); |
Nick Piggin | 77a8a78 | 2006-01-06 00:10:57 -0800 | [diff] [blame] | 147 | } |
| 148 | |
Hugh Dickins | 03f6462 | 2009-09-21 17:03:35 -0700 | [diff] [blame] | 149 | extern unsigned long highest_memmap_pfn; |
| 150 | |
Lee Schermerhorn | 894bc31 | 2008-10-18 20:26:39 -0700 | [diff] [blame] | 151 | /* |
Johannes Weiner | c73322d | 2017-05-03 14:51:51 -0700 | [diff] [blame] | 152 | * 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 Schermerhorn | 894bc31 | 2008-10-18 20:26:39 -0700 | [diff] [blame] | 158 | * in mm/vmscan.c: |
| 159 | */ |
Nick Piggin | 62695a8 | 2008-10-18 20:26:09 -0700 | [diff] [blame] | 160 | extern int isolate_lru_page(struct page *page); |
Lee Schermerhorn | 894bc31 | 2008-10-18 20:26:39 -0700 | [diff] [blame] | 161 | extern void putback_lru_page(struct page *page); |
Mel Gorman | c3f4a9a | 2021-11-05 13:42:42 -0700 | [diff] [blame] | 162 | extern void reclaim_throttle(pg_data_t *pgdat, enum vmscan_throttle_state reason); |
Nick Piggin | 62695a8 | 2008-10-18 20:26:09 -0700 | [diff] [blame] | 163 | |
Lee Schermerhorn | 894bc31 | 2008-10-18 20:26:39 -0700 | [diff] [blame] | 164 | /* |
Bob Liu | 6219049 | 2012-12-11 16:00:37 -0800 | [diff] [blame] | 165 | * in mm/rmap.c: |
| 166 | */ |
| 167 | extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address); |
| 168 | |
| 169 | /* |
Lee Schermerhorn | 894bc31 | 2008-10-18 20:26:39 -0700 | [diff] [blame] | 170 | * in mm/page_alloc.c |
| 171 | */ |
Joonsoo Kim | 3c60509 | 2014-11-13 15:19:21 -0800 | [diff] [blame] | 172 | |
| 173 | /* |
Vlastimil Babka | 1a6d53a | 2015-02-11 15:25:44 -0800 | [diff] [blame] | 174 | * 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 Kim | 97a225e | 2020-06-03 15:59:01 -0700 | [diff] [blame] | 178 | * nodemask, migratetype and highest_zoneidx are initialized only once in |
Matthew Wilcox (Oracle) | 84172f4 | 2021-04-29 23:01:15 -0700 | [diff] [blame] | 179 | * __alloc_pages() and then never change. |
Vlastimil Babka | 1a6d53a | 2015-02-11 15:25:44 -0800 | [diff] [blame] | 180 | * |
Joonsoo Kim | 97a225e | 2020-06-03 15:59:01 -0700 | [diff] [blame] | 181 | * zonelist, preferred_zone and highest_zoneidx are set first in |
Matthew Wilcox (Oracle) | 84172f4 | 2021-04-29 23:01:15 -0700 | [diff] [blame] | 182 | * __alloc_pages() for the fast path, and might be later changed |
Ethon Paul | 68956cc | 2020-06-04 16:49:31 -0700 | [diff] [blame] | 183 | * in __alloc_pages_slowpath(). All other functions pass the whole structure |
Vlastimil Babka | 1a6d53a | 2015-02-11 15:25:44 -0800 | [diff] [blame] | 184 | * by a const pointer. |
| 185 | */ |
| 186 | struct alloc_context { |
| 187 | struct zonelist *zonelist; |
| 188 | nodemask_t *nodemask; |
Mel Gorman | c33d6c0 | 2016-05-19 17:14:10 -0700 | [diff] [blame] | 189 | struct zoneref *preferred_zoneref; |
Vlastimil Babka | 1a6d53a | 2015-02-11 15:25:44 -0800 | [diff] [blame] | 190 | int migratetype; |
Joonsoo Kim | 97a225e | 2020-06-03 15:59:01 -0700 | [diff] [blame] | 191 | |
| 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 Gorman | c9ab0c4 | 2015-11-06 16:28:12 -0800 | [diff] [blame] | 203 | bool spread_dirty_pages; |
Vlastimil Babka | 1a6d53a | 2015-02-11 15:25:44 -0800 | [diff] [blame] | 204 | }; |
| 205 | |
| 206 | /* |
Joonsoo Kim | 3c60509 | 2014-11-13 15:19:21 -0800 | [diff] [blame] | 207 | * 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 | */ |
| 223 | static inline unsigned long |
Vlastimil Babka | 76741e7 | 2017-02-22 15:41:48 -0800 | [diff] [blame] | 224 | __find_buddy_pfn(unsigned long page_pfn, unsigned int order) |
Joonsoo Kim | 3c60509 | 2014-11-13 15:19:21 -0800 | [diff] [blame] | 225 | { |
Vlastimil Babka | 76741e7 | 2017-02-22 15:41:48 -0800 | [diff] [blame] | 226 | return page_pfn ^ (1 << order); |
Joonsoo Kim | 3c60509 | 2014-11-13 15:19:21 -0800 | [diff] [blame] | 227 | } |
| 228 | |
Joonsoo Kim | 7cf91a9 | 2016-03-15 14:57:51 -0700 | [diff] [blame] | 229 | extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn, |
| 230 | unsigned long end_pfn, struct zone *zone); |
| 231 | |
| 232 | static 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 Kim | 3c60509 | 2014-11-13 15:19:21 -0800 | [diff] [blame] | 241 | extern int __isolate_free_page(struct page *page, unsigned int order); |
Alexander Duyck | 624f58d | 2020-04-06 20:04:53 -0700 | [diff] [blame] | 242 | extern void __putback_isolated_page(struct page *page, unsigned int order, |
| 243 | int mt); |
Mike Rapoport | 7c2ee34 | 2018-10-30 15:09:36 -0700 | [diff] [blame] | 244 | extern void memblock_free_pages(struct page *page, unsigned long pfn, |
Mel Gorman | d70ddd7 | 2015-06-30 14:56:52 -0700 | [diff] [blame] | 245 | unsigned int order); |
Arun KS | a9cd410 | 2019-03-05 15:42:14 -0800 | [diff] [blame] | 246 | extern void __free_pages_core(struct page *page, unsigned int order); |
Kirill A. Shutemov | d00181b | 2015-11-06 16:29:57 -0800 | [diff] [blame] | 247 | extern void prep_compound_page(struct page *page, unsigned int order); |
Joonsoo Kim | 46f24fd | 2016-07-26 15:23:58 -0700 | [diff] [blame] | 248 | extern void post_alloc_hook(struct page *page, unsigned int order, |
| 249 | gfp_t gfp_flags); |
Han Pingtian | 42aa83c | 2014-01-23 15:53:28 -0800 | [diff] [blame] | 250 | extern int user_min_free_kbytes; |
Wu Fengguang | 20a0307 | 2009-06-16 15:32:22 -0700 | [diff] [blame] | 251 | |
Mel Gorman | 44042b4 | 2021-06-28 19:43:08 -0700 | [diff] [blame] | 252 | extern void free_unref_page(struct page *page, unsigned int order); |
Matthew Wilcox (Oracle) | 0966aeb | 2020-12-14 19:08:02 -0800 | [diff] [blame] | 253 | extern void free_unref_page_list(struct list_head *list); |
| 254 | |
Mel Gorman | 04f8cfe | 2021-06-28 19:42:15 -0700 | [diff] [blame] | 255 | extern void zone_pcp_update(struct zone *zone, int cpu_online); |
Mel Gorman | 6826539 | 2019-11-30 17:55:15 -0800 | [diff] [blame] | 256 | extern void zone_pcp_reset(struct zone *zone); |
Vlastimil Babka | ec6e8c7e | 2020-12-14 19:10:59 -0800 | [diff] [blame] | 257 | extern void zone_pcp_disable(struct zone *zone); |
| 258 | extern void zone_pcp_enable(struct zone *zone); |
Mel Gorman | 6826539 | 2019-11-30 17:55:15 -0800 | [diff] [blame] | 259 | |
Mike Rapoport | c803b3c | 2021-09-02 14:58:02 -0700 | [diff] [blame] | 260 | extern 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 Nazarewicz | ff9543f | 2011-12-29 13:09:50 +0100 | [diff] [blame] | 264 | #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 | */ |
| 276 | struct 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 Gorman | c5fbd93 | 2019-03-05 15:44:25 -0800 | [diff] [blame] | 279 | unsigned int nr_freepages; /* Number of isolated free pages */ |
| 280 | unsigned int nr_migratepages; /* Number of pages to migrate */ |
Michal Nazarewicz | ff9543f | 2011-12-29 13:09:50 +0100 | [diff] [blame] | 281 | unsigned long free_pfn; /* isolate_freepages search base */ |
Oscar Salvador | c2ad7a1 | 2021-05-04 18:35:17 -0700 | [diff] [blame] | 282 | /* |
| 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 Gorman | 70b4459 | 2019-03-05 15:44:54 -0800 | [diff] [blame] | 289 | unsigned long fast_start_pfn; /* a pfn to start linear scan from */ |
Mel Gorman | c5943b9 | 2019-03-05 15:44:28 -0800 | [diff] [blame] | 290 | struct zone *zone; |
| 291 | unsigned long total_migrate_scanned; |
| 292 | unsigned long total_free_scanned; |
Mel Gorman | dbe2d4e | 2019-03-05 15:45:31 -0800 | [diff] [blame] | 293 | unsigned short fast_search_fail;/* failures to use free list searches */ |
| 294 | short search_order; /* order to start a fast search at */ |
Vlastimil Babka | f25ba6d | 2017-05-08 15:54:30 -0700 | [diff] [blame] | 295 | const gfp_t gfp_mask; /* gfp mask of a direct compactor */ |
| 296 | int order; /* order a direct compactor needs */ |
Vlastimil Babka | d39773a | 2017-05-08 15:54:46 -0700 | [diff] [blame] | 297 | int migratetype; /* migratetype of direct compactor */ |
Vlastimil Babka | f25ba6d | 2017-05-08 15:54:30 -0700 | [diff] [blame] | 298 | const unsigned int alloc_flags; /* alloc flags of a direct compactor */ |
Joonsoo Kim | 97a225e | 2020-06-03 15:59:01 -0700 | [diff] [blame] | 299 | const int highest_zoneidx; /* zone index of a direct compactor */ |
David Rientjes | e0b9dae | 2014-06-04 16:08:28 -0700 | [diff] [blame] | 300 | enum migrate_mode mode; /* Async or sync migration mode */ |
Mel Gorman | bb13ffe | 2012-10-08 16:32:41 -0700 | [diff] [blame] | 301 | bool ignore_skip_hint; /* Scan blocks even if marked skip */ |
Vlastimil Babka | 2583d67 | 2017-11-17 15:26:38 -0800 | [diff] [blame] | 302 | bool no_set_skip_hint; /* Don't mark blocks for skipping */ |
Vlastimil Babka | 9f7e338 | 2016-10-07 17:00:37 -0700 | [diff] [blame] | 303 | bool ignore_block_suitable; /* Scan blocks considered unsuitable */ |
Vlastimil Babka | accf624 | 2016-03-17 14:18:15 -0700 | [diff] [blame] | 304 | bool direct_compaction; /* False from kcompactd or /proc/... */ |
Nitin Gupta | facdaa9 | 2020-08-11 18:31:00 -0700 | [diff] [blame] | 305 | bool proactive_compaction; /* kcompactd proactive compaction */ |
Vlastimil Babka | 06ed299 | 2016-10-07 16:57:35 -0700 | [diff] [blame] | 306 | bool whole_zone; /* Whole zone should/has been scanned */ |
Vlastimil Babka | c3486f5 | 2016-07-28 15:49:30 -0700 | [diff] [blame] | 307 | bool contended; /* Signal lock or sched contention */ |
Mel Gorman | 804d312 | 2019-03-05 15:45:07 -0800 | [diff] [blame] | 308 | bool rescan; /* Rescanning the same pageblock */ |
Rik van Riel | b06eda0 | 2020-04-01 21:10:28 -0700 | [diff] [blame] | 309 | bool alloc_contig; /* alloc_contig_range allocation */ |
Michal Nazarewicz | ff9543f | 2011-12-29 13:09:50 +0100 | [diff] [blame] | 310 | }; |
| 311 | |
Mel Gorman | 5e1f0f0 | 2019-03-05 15:45:41 -0800 | [diff] [blame] | 312 | /* |
| 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 | */ |
| 316 | struct capture_control { |
| 317 | struct compact_control *cc; |
| 318 | struct page *page; |
| 319 | }; |
| 320 | |
Michal Nazarewicz | ff9543f | 2011-12-29 13:09:50 +0100 | [diff] [blame] | 321 | unsigned long |
Mel Gorman | bb13ffe | 2012-10-08 16:32:41 -0700 | [diff] [blame] | 322 | isolate_freepages_range(struct compact_control *cc, |
| 323 | unsigned long start_pfn, unsigned long end_pfn); |
Oscar Salvador | c2ad7a1 | 2021-05-04 18:35:17 -0700 | [diff] [blame] | 324 | int |
Vlastimil Babka | edc2ca6 | 2014-10-09 15:27:09 -0700 | [diff] [blame] | 325 | isolate_migratepages_range(struct compact_control *cc, |
| 326 | unsigned long low_pfn, unsigned long end_pfn); |
Mel Gorman | ffd8f25 | 2021-06-30 18:53:53 -0700 | [diff] [blame] | 327 | #endif |
Joonsoo Kim | 2149cda | 2015-04-14 15:45:21 -0700 | [diff] [blame] | 328 | int find_suitable_fallback(struct free_area *area, unsigned int order, |
| 329 | int migratetype, bool only_stealable, bool *can_steal); |
Michal Nazarewicz | ff9543f | 2011-12-29 13:09:50 +0100 | [diff] [blame] | 330 | |
Mel Gorman | 48f13bf | 2007-10-16 01:26:10 -0700 | [diff] [blame] | 331 | /* |
Mel Gorman | 6c14466 | 2014-01-23 15:53:38 -0800 | [diff] [blame] | 332 | * 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 Babka | 99c0fd5 | 2014-10-09 15:27:23 -0700 | [diff] [blame] | 336 | * page cannot be allocated or merged in parallel. Alternatively, it must |
Matthew Wilcox (Oracle) | ab130f91 | 2020-10-15 20:10:15 -0700 | [diff] [blame] | 337 | * handle invalid values gracefully, and use buddy_order_unsafe() below. |
Mel Gorman | 48f13bf | 2007-10-16 01:26:10 -0700 | [diff] [blame] | 338 | */ |
Matthew Wilcox (Oracle) | ab130f91 | 2020-10-15 20:10:15 -0700 | [diff] [blame] | 339 | static inline unsigned int buddy_order(struct page *page) |
Mel Gorman | 48f13bf | 2007-10-16 01:26:10 -0700 | [diff] [blame] | 340 | { |
KAMEZAWA Hiroyuki | 572438f | 2010-10-26 14:22:08 -0700 | [diff] [blame] | 341 | /* PageBuddy() must be checked by the caller */ |
Mel Gorman | 48f13bf | 2007-10-16 01:26:10 -0700 | [diff] [blame] | 342 | return page_private(page); |
| 343 | } |
Alexander van Heukelum | b5a0e01 | 2008-02-23 15:24:06 -0800 | [diff] [blame] | 344 | |
Vlastimil Babka | 99c0fd5 | 2014-10-09 15:27:23 -0700 | [diff] [blame] | 345 | /* |
Matthew Wilcox (Oracle) | ab130f91 | 2020-10-15 20:10:15 -0700 | [diff] [blame] | 346 | * Like buddy_order(), but for callers who cannot afford to hold the zone lock. |
Vlastimil Babka | 99c0fd5 | 2014-10-09 15:27:23 -0700 | [diff] [blame] | 347 | * PageBuddy() should be checked first by the caller to minimize race window, |
| 348 | * and invalid values must be handled gracefully. |
| 349 | * |
Jason Low | 4db0c3c | 2015-04-15 16:14:08 -0700 | [diff] [blame] | 350 | * READ_ONCE is used so that if the caller assigns the result into a local |
Vlastimil Babka | 99c0fd5 | 2014-10-09 15:27:23 -0700 | [diff] [blame] | 351 | * 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) | ab130f91 | 2020-10-15 20:10:15 -0700 | [diff] [blame] | 356 | #define buddy_order_unsafe(page) READ_ONCE(page_private(page)) |
Vlastimil Babka | 99c0fd5 | 2014-10-09 15:27:23 -0700 | [diff] [blame] | 357 | |
Konstantin Khlebnikov | 30bdbb7 | 2016-02-02 16:57:46 -0800 | [diff] [blame] | 358 | /* |
| 359 | * These three helpers classifies VMAs for virtual memory accounting. |
| 360 | */ |
| 361 | |
| 362 | /* |
| 363 | * Executable code area - executable, not writable, not stack |
| 364 | */ |
Konstantin Khlebnikov | d977d56 | 2016-02-02 16:57:43 -0800 | [diff] [blame] | 365 | static inline bool is_exec_mapping(vm_flags_t flags) |
| 366 | { |
Konstantin Khlebnikov | 30bdbb7 | 2016-02-02 16:57:46 -0800 | [diff] [blame] | 367 | return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC; |
Konstantin Khlebnikov | d977d56 | 2016-02-02 16:57:43 -0800 | [diff] [blame] | 368 | } |
| 369 | |
Konstantin Khlebnikov | 30bdbb7 | 2016-02-02 16:57:46 -0800 | [diff] [blame] | 370 | /* |
Ingo Molnar | f0953a1 | 2021-05-06 18:06:47 -0700 | [diff] [blame] | 371 | * Stack area - automatically grows in one direction |
Konstantin Khlebnikov | 30bdbb7 | 2016-02-02 16:57:46 -0800 | [diff] [blame] | 372 | * |
| 373 | * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous: |
| 374 | * do_mmap() forbids all other combinations. |
| 375 | */ |
Konstantin Khlebnikov | d977d56 | 2016-02-02 16:57:43 -0800 | [diff] [blame] | 376 | static inline bool is_stack_mapping(vm_flags_t flags) |
| 377 | { |
Konstantin Khlebnikov | 30bdbb7 | 2016-02-02 16:57:46 -0800 | [diff] [blame] | 378 | return (flags & VM_STACK) == VM_STACK; |
Konstantin Khlebnikov | d977d56 | 2016-02-02 16:57:43 -0800 | [diff] [blame] | 379 | } |
| 380 | |
Konstantin Khlebnikov | 30bdbb7 | 2016-02-02 16:57:46 -0800 | [diff] [blame] | 381 | /* |
| 382 | * Data area - private, writable, not stack |
| 383 | */ |
Konstantin Khlebnikov | d977d56 | 2016-02-02 16:57:43 -0800 | [diff] [blame] | 384 | static inline bool is_data_mapping(vm_flags_t flags) |
| 385 | { |
Konstantin Khlebnikov | 30bdbb7 | 2016-02-02 16:57:46 -0800 | [diff] [blame] | 386 | return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE; |
Konstantin Khlebnikov | d977d56 | 2016-02-02 16:57:43 -0800 | [diff] [blame] | 387 | } |
| 388 | |
Namhyung Kim | 6038def | 2011-05-24 17:11:22 -0700 | [diff] [blame] | 389 | /* mm/util.c */ |
| 390 | void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma, |
Wei Yang | aba6dfb | 2019-11-30 17:50:53 -0800 | [diff] [blame] | 391 | struct vm_area_struct *prev); |
Wei Yang | 1b9fc5b2 | 2019-11-30 17:50:49 -0800 | [diff] [blame] | 392 | void __vma_unlink_list(struct mm_struct *mm, struct vm_area_struct *vma); |
Namhyung Kim | 6038def | 2011-05-24 17:11:22 -0700 | [diff] [blame] | 393 | |
Hugh Dickins | af8e335 | 2009-12-14 17:58:59 -0800 | [diff] [blame] | 394 | #ifdef CONFIG_MMU |
Matthew Wilcox (Oracle) | 3506659 | 2021-11-28 14:53:35 -0500 | [diff] [blame] | 395 | void unmap_mapping_folio(struct folio *folio); |
Kirill A. Shutemov | fc05f56 | 2015-04-14 15:44:39 -0700 | [diff] [blame] | 396 | extern long populate_vma_page_range(struct vm_area_struct *vma, |
David Hildenbrand | a78f1cc | 2021-06-30 18:52:23 -0700 | [diff] [blame] | 397 | unsigned long start, unsigned long end, int *locked); |
David Hildenbrand | 4ca9b385 | 2021-06-30 18:52:28 -0700 | [diff] [blame] | 398 | extern long faultin_vma_page_range(struct vm_area_struct *vma, |
| 399 | unsigned long start, unsigned long end, |
| 400 | bool write, int *locked); |
Hugh Dickins | af8e335 | 2009-12-14 17:58:59 -0800 | [diff] [blame] | 401 | extern void munlock_vma_pages_range(struct vm_area_struct *vma, |
| 402 | unsigned long start, unsigned long end); |
| 403 | static 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 Piggin | b291f00 | 2008-10-18 20:26:44 -0700 | [diff] [blame] | 408 | /* |
Michel Lespinasse | c1e8d7c | 2020-06-08 21:33:54 -0700 | [diff] [blame] | 409 | * must be called with vma's mmap_lock held for read or write, and page locked. |
Nick Piggin | b291f00 | 2008-10-18 20:26:44 -0700 | [diff] [blame] | 410 | */ |
| 411 | extern void mlock_vma_page(struct page *page); |
Michel Lespinasse | ff6a6da | 2013-02-27 17:02:44 -0800 | [diff] [blame] | 412 | extern unsigned int munlock_vma_page(struct page *page); |
Nick Piggin | b291f00 | 2008-10-18 20:26:44 -0700 | [diff] [blame] | 413 | |
Mike Rapoport | 6aeb254 | 2021-07-07 18:07:50 -0700 | [diff] [blame] | 414 | extern int mlock_future_check(struct mm_struct *mm, unsigned long flags, |
| 415 | unsigned long len); |
| 416 | |
Nick Piggin | b291f00 | 2008-10-18 20:26:44 -0700 | [diff] [blame] | 417 | /* |
| 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 Dickins | e6c509f | 2012-10-08 16:33:19 -0700 | [diff] [blame] | 426 | extern void clear_page_mlock(struct page *page); |
Nick Piggin | b291f00 | 2008-10-18 20:26:44 -0700 | [diff] [blame] | 427 | |
Linus Torvalds | f55e101 | 2017-11-29 09:01:01 -0800 | [diff] [blame] | 428 | extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma); |
Mel Gorman | b32967f | 2012-11-19 12:35:47 +0000 | [diff] [blame] | 429 | |
Kirill A. Shutemov | e9b61f1 | 2016-01-15 16:54:10 -0800 | [diff] [blame] | 430 | /* |
Hugh Dickins | 494334e | 2021-06-15 18:23:56 -0700 | [diff] [blame] | 431 | * 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. Shutemov | e9b61f1 | 2016-01-15 16:54:10 -0800 | [diff] [blame] | 434 | */ |
| 435 | static inline unsigned long |
Kirill A. Shutemov | e9b61f1 | 2016-01-15 16:54:10 -0800 | [diff] [blame] | 436 | vma_address(struct page *page, struct vm_area_struct *vma) |
| 437 | { |
Hugh Dickins | 494334e | 2021-06-15 18:23:56 -0700 | [diff] [blame] | 438 | pgoff_t pgoff; |
| 439 | unsigned long address; |
Kirill A. Shutemov | a8fa41ad | 2017-02-24 14:57:54 -0800 | [diff] [blame] | 440 | |
Hugh Dickins | 494334e | 2021-06-15 18:23:56 -0700 | [diff] [blame] | 441 | 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. Shutemov | e9b61f1 | 2016-01-15 16:54:10 -0800 | [diff] [blame] | 458 | |
Hugh Dickins | 494334e | 2021-06-15 18:23:56 -0700 | [diff] [blame] | 459 | /* |
| 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 | */ |
| 464 | static inline unsigned long |
| 465 | vma_address_end(struct page *page, struct vm_area_struct *vma) |
| 466 | { |
| 467 | pgoff_t pgoff; |
| 468 | unsigned long address; |
Kirill A. Shutemov | e9b61f1 | 2016-01-15 16:54:10 -0800 | [diff] [blame] | 469 | |
Hugh Dickins | 494334e | 2021-06-15 18:23:56 -0700 | [diff] [blame] | 470 | 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. Shutemov | e9b61f1 | 2016-01-15 16:54:10 -0800 | [diff] [blame] | 477 | } |
| 478 | |
Johannes Weiner | 89b1533 | 2019-11-30 17:50:22 -0800 | [diff] [blame] | 479 | static 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 Lespinasse | c1e8d7c | 2020-06-08 21:33:54 -0700 | [diff] [blame] | 489 | * anything, so we only pin the file and drop the mmap_lock if only |
Peter Xu | 4064b98 | 2020-04-01 21:08:45 -0700 | [diff] [blame] | 490 | * FAULT_FLAG_ALLOW_RETRY is set, while this is the first attempt. |
Johannes Weiner | 89b1533 | 2019-11-30 17:50:22 -0800 | [diff] [blame] | 491 | */ |
Peter Xu | 4064b98 | 2020-04-01 21:08:45 -0700 | [diff] [blame] | 492 | if (fault_flag_allow_retry_first(flags) && |
| 493 | !(flags & FAULT_FLAG_RETRY_NOWAIT)) { |
Johannes Weiner | 89b1533 | 2019-11-30 17:50:22 -0800 | [diff] [blame] | 494 | fpin = get_file(vmf->vma->vm_file); |
Michel Lespinasse | d8ed45c | 2020-06-08 21:33:25 -0700 | [diff] [blame] | 495 | mmap_read_unlock(vmf->vma->vm_mm); |
Johannes Weiner | 89b1533 | 2019-11-30 17:50:22 -0800 | [diff] [blame] | 496 | } |
| 497 | return fpin; |
| 498 | } |
Hugh Dickins | af8e335 | 2009-12-14 17:58:59 -0800 | [diff] [blame] | 499 | #else /* !CONFIG_MMU */ |
Matthew Wilcox (Oracle) | 3506659 | 2021-11-28 14:53:35 -0500 | [diff] [blame] | 500 | static inline void unmap_mapping_folio(struct folio *folio) { } |
Nick Piggin | b291f00 | 2008-10-18 20:26:44 -0700 | [diff] [blame] | 501 | static inline void clear_page_mlock(struct page *page) { } |
| 502 | static inline void mlock_vma_page(struct page *page) { } |
Nicholas Piggin | 4ad0ae8 | 2021-04-29 22:59:01 -0700 | [diff] [blame] | 503 | static inline void vunmap_range_noflush(unsigned long start, unsigned long end) |
| 504 | { |
| 505 | } |
Hugh Dickins | af8e335 | 2009-12-14 17:58:59 -0800 | [diff] [blame] | 506 | #endif /* !CONFIG_MMU */ |
Lee Schermerhorn | 894bc31 | 2008-10-18 20:26:39 -0700 | [diff] [blame] | 507 | |
Alexander van Heukelum | b5a0e01 | 2008-02-23 15:24:06 -0800 | [diff] [blame] | 508 | /* |
Andy Whitcroft | 69d177c | 2008-11-06 12:53:26 -0800 | [diff] [blame] | 509 | * 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 | */ |
| 513 | static inline struct page *mem_map_offset(struct page *base, int offset) |
| 514 | { |
| 515 | if (unlikely(offset >= MAX_ORDER_NR_PAGES)) |
Fabian Frederick | bc7f84c | 2014-08-06 16:05:17 -0700 | [diff] [blame] | 516 | return nth_page(base, offset); |
Andy Whitcroft | 69d177c | 2008-11-06 12:53:26 -0800 | [diff] [blame] | 517 | return base + offset; |
| 518 | } |
| 519 | |
| 520 | /* |
Lucas De Marchi | 25985ed | 2011-03-30 22:57:33 -0300 | [diff] [blame] | 521 | * Iterator over all subpages within the maximally aligned gigantic |
Andy Whitcroft | 69d177c | 2008-11-06 12:53:26 -0800 | [diff] [blame] | 522 | * page 'base'. Handle any discontiguity in the mem_map. |
| 523 | */ |
| 524 | static 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 Gorman | 6b74ab9 | 2008-07-23 21:26:49 -0700 | [diff] [blame] | 536 | /* Memory initialisation debug and verification */ |
| 537 | enum mminit_level { |
| 538 | MMINIT_WARNING, |
| 539 | MMINIT_VERIFY, |
| 540 | MMINIT_TRACE |
| 541 | }; |
| 542 | |
| 543 | #ifdef CONFIG_DEBUG_MEMORY_INIT |
| 544 | |
| 545 | extern int mminit_loglevel; |
| 546 | |
| 547 | #define mminit_dprintk(level, prefix, fmt, arg...) \ |
| 548 | do { \ |
| 549 | if (level < mminit_loglevel) { \ |
Rasmus Villemoes | fc5199d | 2015-02-12 15:00:02 -0800 | [diff] [blame] | 550 | if (level <= MMINIT_WARNING) \ |
Joe Perches | 1170532 | 2016-03-17 14:19:50 -0700 | [diff] [blame] | 551 | pr_warn("mminit::" prefix " " fmt, ##arg); \ |
Rasmus Villemoes | fc5199d | 2015-02-12 15:00:02 -0800 | [diff] [blame] | 552 | else \ |
| 553 | printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \ |
Mel Gorman | 6b74ab9 | 2008-07-23 21:26:49 -0700 | [diff] [blame] | 554 | } \ |
| 555 | } while (0) |
| 556 | |
Mel Gorman | 708614e | 2008-07-23 21:26:51 -0700 | [diff] [blame] | 557 | extern void mminit_verify_pageflags_layout(void); |
Mel Gorman | 68ad8df | 2008-07-23 21:26:52 -0700 | [diff] [blame] | 558 | extern void mminit_verify_zonelist(void); |
Mel Gorman | 6b74ab9 | 2008-07-23 21:26:49 -0700 | [diff] [blame] | 559 | #else |
| 560 | |
| 561 | static inline void mminit_dprintk(enum mminit_level level, |
| 562 | const char *prefix, const char *fmt, ...) |
| 563 | { |
| 564 | } |
| 565 | |
Mel Gorman | 708614e | 2008-07-23 21:26:51 -0700 | [diff] [blame] | 566 | static inline void mminit_verify_pageflags_layout(void) |
| 567 | { |
| 568 | } |
| 569 | |
Mel Gorman | 68ad8df | 2008-07-23 21:26:52 -0700 | [diff] [blame] | 570 | static inline void mminit_verify_zonelist(void) |
| 571 | { |
| 572 | } |
Mel Gorman | 6b74ab9 | 2008-07-23 21:26:49 -0700 | [diff] [blame] | 573 | #endif /* CONFIG_DEBUG_MEMORY_INIT */ |
Mel Gorman | 2dbb51c | 2008-07-23 21:26:52 -0700 | [diff] [blame] | 574 | |
| 575 | /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */ |
| 576 | #if defined(CONFIG_SPARSEMEM) |
| 577 | extern void mminit_validate_memmodel_limits(unsigned long *start_pfn, |
| 578 | unsigned long *end_pfn); |
| 579 | #else |
| 580 | static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn, |
| 581 | unsigned long *end_pfn) |
| 582 | { |
| 583 | } |
| 584 | #endif /* CONFIG_SPARSEMEM */ |
| 585 | |
Mel Gorman | a5f5f91 | 2016-07-28 15:46:32 -0700 | [diff] [blame] | 586 | #define NODE_RECLAIM_NOSCAN -2 |
| 587 | #define NODE_RECLAIM_FULL -1 |
| 588 | #define NODE_RECLAIM_SOME 0 |
| 589 | #define NODE_RECLAIM_SUCCESS 1 |
Wu Fengguang | 7c116f2 | 2009-12-16 12:19:59 +0100 | [diff] [blame] | 590 | |
Wei Yang | 8b09549 | 2018-12-28 00:34:36 -0800 | [diff] [blame] | 591 | #ifdef CONFIG_NUMA |
| 592 | extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int); |
Dave Hansen | 79c28a4 | 2021-09-02 14:59:06 -0700 | [diff] [blame] | 593 | extern int find_next_best_node(int node, nodemask_t *used_node_mask); |
Wei Yang | 8b09549 | 2018-12-28 00:34:36 -0800 | [diff] [blame] | 594 | #else |
| 595 | static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask, |
| 596 | unsigned int order) |
| 597 | { |
| 598 | return NODE_RECLAIM_NOSCAN; |
| 599 | } |
Dave Hansen | 79c28a4 | 2021-09-02 14:59:06 -0700 | [diff] [blame] | 600 | static inline int find_next_best_node(int node, nodemask_t *used_node_mask) |
| 601 | { |
| 602 | return NUMA_NO_NODE; |
| 603 | } |
Wei Yang | 8b09549 | 2018-12-28 00:34:36 -0800 | [diff] [blame] | 604 | #endif |
| 605 | |
Wu Fengguang | 31d3d34 | 2009-12-16 12:19:59 +0100 | [diff] [blame] | 606 | extern int hwpoison_filter(struct page *p); |
| 607 | |
Wu Fengguang | 7c116f2 | 2009-12-16 12:19:59 +0100 | [diff] [blame] | 608 | extern u32 hwpoison_filter_dev_major; |
| 609 | extern u32 hwpoison_filter_dev_minor; |
Wu Fengguang | 478c5ff | 2009-12-16 12:19:59 +0100 | [diff] [blame] | 610 | extern u64 hwpoison_filter_flags_mask; |
| 611 | extern u64 hwpoison_filter_flags_value; |
Andi Kleen | 4fd466e | 2009-12-16 12:19:59 +0100 | [diff] [blame] | 612 | extern u64 hwpoison_filter_memcg; |
Haicheng Li | 1bfe5fe | 2009-12-16 12:19:59 +0100 | [diff] [blame] | 613 | extern u32 hwpoison_filter_enable; |
Al Viro | eb36c58 | 2012-05-30 20:17:35 -0400 | [diff] [blame] | 614 | |
Michal Hocko | dc0ef0d | 2016-05-23 16:25:27 -0700 | [diff] [blame] | 615 | extern unsigned long __must_check vm_mmap_pgoff(struct file *, unsigned long, |
Al Viro | eb36c58 | 2012-05-30 20:17:35 -0400 | [diff] [blame] | 616 | unsigned long, unsigned long, |
Michal Hocko | 9fbeb5a | 2016-05-23 16:25:30 -0700 | [diff] [blame] | 617 | unsigned long, unsigned long); |
Xishi Qiu | ca57df7 | 2012-07-31 16:43:19 -0700 | [diff] [blame] | 618 | |
| 619 | extern void set_pageblock_order(void); |
Maninder Singh | 730ec8c | 2020-06-03 16:01:18 -0700 | [diff] [blame] | 620 | unsigned int reclaim_clean_pages_from_list(struct zone *zone, |
Minchan Kim | 02c6de8 | 2012-10-08 16:31:55 -0700 | [diff] [blame] | 621 | struct list_head *page_list); |
Bartlomiej Zolnierkiewicz | d95ea5d | 2012-10-08 16:32:05 -0700 | [diff] [blame] | 622 | /* 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 Hocko | cd04ae1 | 2017-09-06 16:24:50 -0700 | [diff] [blame] | 631 | /* |
| 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 Gorman | 6bb1545 | 2018-12-28 00:35:41 -0800 | [diff] [blame] | 642 | #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 Nosek | 736838e | 2020-04-01 21:09:47 -0700 | [diff] [blame] | 651 | #define ALLOC_KSWAPD 0x800 /* allow waking of kswapd, __GFP_KSWAPD_RECLAIM set */ |
Bartlomiej Zolnierkiewicz | d95ea5d | 2012-10-08 16:32:05 -0700 | [diff] [blame] | 652 | |
Mel Gorman | 72b252a | 2015-09-04 15:47:32 -0700 | [diff] [blame] | 653 | enum ttu_flags; |
| 654 | struct tlbflush_unmap_batch; |
| 655 | |
Michal Hocko | ce61287 | 2017-04-07 16:05:05 -0700 | [diff] [blame] | 656 | |
| 657 | /* |
| 658 | * only for MM internal work items which do not depend on |
| 659 | * any allocations or locks which might depend on allocations |
| 660 | */ |
| 661 | extern struct workqueue_struct *mm_percpu_wq; |
| 662 | |
Mel Gorman | 72b252a | 2015-09-04 15:47:32 -0700 | [diff] [blame] | 663 | #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH |
| 664 | void try_to_unmap_flush(void); |
Mel Gorman | d950c94 | 2015-09-04 15:47:35 -0700 | [diff] [blame] | 665 | void try_to_unmap_flush_dirty(void); |
Mel Gorman | 3ea2771 | 2017-08-02 13:31:52 -0700 | [diff] [blame] | 666 | void flush_tlb_batched_pending(struct mm_struct *mm); |
Mel Gorman | 72b252a | 2015-09-04 15:47:32 -0700 | [diff] [blame] | 667 | #else |
| 668 | static inline void try_to_unmap_flush(void) |
| 669 | { |
| 670 | } |
Mel Gorman | d950c94 | 2015-09-04 15:47:35 -0700 | [diff] [blame] | 671 | static inline void try_to_unmap_flush_dirty(void) |
| 672 | { |
| 673 | } |
Mel Gorman | 3ea2771 | 2017-08-02 13:31:52 -0700 | [diff] [blame] | 674 | static inline void flush_tlb_batched_pending(struct mm_struct *mm) |
| 675 | { |
| 676 | } |
Mel Gorman | 72b252a | 2015-09-04 15:47:32 -0700 | [diff] [blame] | 677 | #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */ |
Vlastimil Babka | edf14cd | 2016-03-15 14:55:56 -0700 | [diff] [blame] | 678 | |
| 679 | extern const struct trace_print_flags pageflag_names[]; |
| 680 | extern const struct trace_print_flags vmaflag_names[]; |
| 681 | extern const struct trace_print_flags gfpflag_names[]; |
| 682 | |
Xishi Qiu | a6ffdc0 | 2017-05-03 14:52:52 -0700 | [diff] [blame] | 683 | static inline bool is_migrate_highatomic(enum migratetype migratetype) |
| 684 | { |
| 685 | return migratetype == MIGRATE_HIGHATOMIC; |
| 686 | } |
| 687 | |
| 688 | static inline bool is_migrate_highatomic_page(struct page *page) |
| 689 | { |
| 690 | return get_pageblock_migratetype(page) == MIGRATE_HIGHATOMIC; |
| 691 | } |
| 692 | |
Michal Hocko | 72675e1 | 2017-09-06 16:20:24 -0700 | [diff] [blame] | 693 | void setup_zone_pageset(struct zone *zone); |
Joonsoo Kim | 19fc7be | 2020-08-11 18:37:25 -0700 | [diff] [blame] | 694 | |
| 695 | struct migration_target_control { |
| 696 | int nid; /* preferred node id */ |
| 697 | nodemask_t *nmask; |
| 698 | gfp_t gfp_mask; |
| 699 | }; |
| 700 | |
Nicholas Piggin | b67177e | 2021-04-29 22:58:53 -0700 | [diff] [blame] | 701 | /* |
| 702 | * mm/vmalloc.c |
| 703 | */ |
Nicholas Piggin | 4ad0ae8 | 2021-04-29 22:59:01 -0700 | [diff] [blame] | 704 | #ifdef CONFIG_MMU |
Nicholas Piggin | b67177e | 2021-04-29 22:58:53 -0700 | [diff] [blame] | 705 | int vmap_pages_range_noflush(unsigned long addr, unsigned long end, |
| 706 | pgprot_t prot, struct page **pages, unsigned int page_shift); |
Nicholas Piggin | 4ad0ae8 | 2021-04-29 22:59:01 -0700 | [diff] [blame] | 707 | #else |
| 708 | static inline |
| 709 | int 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 | |
| 716 | void vunmap_range_noflush(unsigned long start, unsigned long end); |
Nicholas Piggin | b67177e | 2021-04-29 22:58:53 -0700 | [diff] [blame] | 717 | |
Yang Shi | f4c0d83 | 2021-06-30 18:51:39 -0700 | [diff] [blame] | 718 | int numa_migrate_prep(struct page *page, struct vm_area_struct *vma, |
| 719 | unsigned long addr, int page_nid, int *flags); |
| 720 | |
Michel Lespinasse | db97141 | 2012-10-08 16:29:34 -0700 | [diff] [blame] | 721 | #endif /* __MM_INTERNAL_H */ |