Commit | Line | Data |
---|---|---|
2874c5fd | 1 | /* SPDX-License-Identifier: GPL-2.0-or-later */ |
1da177e4 LT |
2 | /* internal.h: mm/ internal definitions |
3 | * | |
4 | * Copyright (C) 2004 Red Hat, Inc. All Rights Reserved. | |
5 | * Written by David Howells (dhowells@redhat.com) | |
1da177e4 | 6 | */ |
0f8053a5 NP |
7 | #ifndef __MM_INTERNAL_H |
8 | #define __MM_INTERNAL_H | |
9 | ||
29f175d1 | 10 | #include <linux/fs.h> |
0f8053a5 | 11 | #include <linux/mm.h> |
e9b61f19 | 12 | #include <linux/pagemap.h> |
2aff7a47 | 13 | #include <linux/rmap.h> |
edf14cdb | 14 | #include <linux/tracepoint-defs.h> |
1da177e4 | 15 | |
0e499ed3 MWO |
16 | struct folio_batch; |
17 | ||
dd56b046 MG |
18 | /* |
19 | * The set of flags that only affect watermark checking and reclaim | |
20 | * behaviour. This is used by the MM to obey the caller constraints | |
21 | * about IO, FS and watermark checking while ignoring placement | |
22 | * hints such as HIGHMEM usage. | |
23 | */ | |
24 | #define GFP_RECLAIM_MASK (__GFP_RECLAIM|__GFP_HIGH|__GFP_IO|__GFP_FS|\ | |
dcda9b04 | 25 | __GFP_NOWARN|__GFP_RETRY_MAYFAIL|__GFP_NOFAIL|\ |
e838a45f | 26 | __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC|\ |
2973d822 | 27 | __GFP_NOLOCKDEP) |
dd56b046 MG |
28 | |
29 | /* The GFP flags allowed during early boot */ | |
30 | #define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_RECLAIM|__GFP_IO|__GFP_FS)) | |
31 | ||
32 | /* Control allocation cpuset and node placement constraints */ | |
33 | #define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE) | |
34 | ||
35 | /* Do not use these with a slab allocator */ | |
36 | #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK) | |
37 | ||
3f913fc5 QZ |
38 | /* |
39 | * Different from WARN_ON_ONCE(), no warning will be issued | |
40 | * when we specify __GFP_NOWARN. | |
41 | */ | |
42 | #define WARN_ON_ONCE_GFP(cond, gfp) ({ \ | |
43 | static bool __section(".data.once") __warned; \ | |
44 | int __ret_warn_once = !!(cond); \ | |
45 | \ | |
46 | if (unlikely(!(gfp & __GFP_NOWARN) && __ret_warn_once && !__warned)) { \ | |
47 | __warned = true; \ | |
48 | WARN_ON(1); \ | |
49 | } \ | |
50 | unlikely(__ret_warn_once); \ | |
51 | }) | |
52 | ||
62906027 NP |
53 | void page_writeback_init(void); |
54 | ||
eec20426 MWO |
55 | /* |
56 | * If a 16GB hugetlb folio were mapped by PTEs of all of its 4kB pages, | |
e78a13fd | 57 | * its nr_pages_mapped would be 0x400000: choose the ENTIRELY_MAPPED bit |
eec20426 MWO |
58 | * above that range, instead of 2*(PMD_SIZE/PAGE_SIZE). Hugetlb currently |
59 | * leaves nr_pages_mapped at 0, but avoid surprise if it participates later. | |
60 | */ | |
e78a13fd DH |
61 | #define ENTIRELY_MAPPED 0x800000 |
62 | #define FOLIO_PAGES_MAPPED (ENTIRELY_MAPPED - 1) | |
eec20426 | 63 | |
1279aa06 KW |
64 | /* |
65 | * Flags passed to __show_mem() and show_free_areas() to suppress output in | |
66 | * various contexts. | |
67 | */ | |
68 | #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */ | |
69 | ||
eec20426 MWO |
70 | /* |
71 | * How many individual pages have an elevated _mapcount. Excludes | |
72 | * the folio's entire_mapcount. | |
73 | */ | |
74 | static inline int folio_nr_pages_mapped(struct folio *folio) | |
75 | { | |
76 | return atomic_read(&folio->_nr_pages_mapped) & FOLIO_PAGES_MAPPED; | |
77 | } | |
78 | ||
64601000 MWO |
79 | static inline void *folio_raw_mapping(struct folio *folio) |
80 | { | |
81 | unsigned long mapping = (unsigned long)folio->mapping; | |
82 | ||
83 | return (void *)(mapping & ~PAGE_MAPPING_FLAGS); | |
84 | } | |
85 | ||
ac96cc4d BS |
86 | #ifdef CONFIG_MMU |
87 | ||
88 | /* Flags for folio_pte_batch(). */ | |
89 | typedef int __bitwise fpb_t; | |
90 | ||
91 | /* Compare PTEs after pte_mkclean(), ignoring the dirty bit. */ | |
92 | #define FPB_IGNORE_DIRTY ((__force fpb_t)BIT(0)) | |
93 | ||
94 | /* Compare PTEs after pte_clear_soft_dirty(), ignoring the soft-dirty bit. */ | |
95 | #define FPB_IGNORE_SOFT_DIRTY ((__force fpb_t)BIT(1)) | |
96 | ||
97 | static inline pte_t __pte_batch_clear_ignored(pte_t pte, fpb_t flags) | |
98 | { | |
99 | if (flags & FPB_IGNORE_DIRTY) | |
100 | pte = pte_mkclean(pte); | |
101 | if (likely(flags & FPB_IGNORE_SOFT_DIRTY)) | |
102 | pte = pte_clear_soft_dirty(pte); | |
103 | return pte_wrprotect(pte_mkold(pte)); | |
104 | } | |
105 | ||
106 | /** | |
107 | * folio_pte_batch - detect a PTE batch for a large folio | |
108 | * @folio: The large folio to detect a PTE batch for. | |
109 | * @addr: The user virtual address the first page is mapped at. | |
110 | * @start_ptep: Page table pointer for the first entry. | |
111 | * @pte: Page table entry for the first page. | |
112 | * @max_nr: The maximum number of table entries to consider. | |
113 | * @flags: Flags to modify the PTE batch semantics. | |
114 | * @any_writable: Optional pointer to indicate whether any entry except the | |
115 | * first one is writable. | |
116 | * | |
117 | * Detect a PTE batch: consecutive (present) PTEs that map consecutive | |
118 | * pages of the same large folio. | |
119 | * | |
120 | * All PTEs inside a PTE batch have the same PTE bits set, excluding the PFN, | |
121 | * the accessed bit, writable bit, dirty bit (with FPB_IGNORE_DIRTY) and | |
122 | * soft-dirty bit (with FPB_IGNORE_SOFT_DIRTY). | |
123 | * | |
124 | * start_ptep must map any page of the folio. max_nr must be at least one and | |
125 | * must be limited by the caller so scanning cannot exceed a single page table. | |
126 | * | |
127 | * Return: the number of table entries in the batch. | |
128 | */ | |
129 | static inline int folio_pte_batch(struct folio *folio, unsigned long addr, | |
130 | pte_t *start_ptep, pte_t pte, int max_nr, fpb_t flags, | |
131 | bool *any_writable) | |
132 | { | |
133 | unsigned long folio_end_pfn = folio_pfn(folio) + folio_nr_pages(folio); | |
134 | const pte_t *end_ptep = start_ptep + max_nr; | |
135 | pte_t expected_pte, *ptep; | |
136 | bool writable; | |
137 | int nr; | |
138 | ||
139 | if (any_writable) | |
140 | *any_writable = false; | |
141 | ||
142 | VM_WARN_ON_FOLIO(!pte_present(pte), folio); | |
143 | VM_WARN_ON_FOLIO(!folio_test_large(folio) || max_nr < 1, folio); | |
144 | VM_WARN_ON_FOLIO(page_folio(pfn_to_page(pte_pfn(pte))) != folio, folio); | |
145 | ||
146 | nr = pte_batch_hint(start_ptep, pte); | |
147 | expected_pte = __pte_batch_clear_ignored(pte_advance_pfn(pte, nr), flags); | |
148 | ptep = start_ptep + nr; | |
149 | ||
150 | while (ptep < end_ptep) { | |
151 | pte = ptep_get(ptep); | |
152 | if (any_writable) | |
153 | writable = !!pte_write(pte); | |
154 | pte = __pte_batch_clear_ignored(pte, flags); | |
155 | ||
156 | if (!pte_same(pte, expected_pte)) | |
157 | break; | |
158 | ||
159 | /* | |
160 | * Stop immediately once we reached the end of the folio. In | |
161 | * corner cases the next PFN might fall into a different | |
162 | * folio. | |
163 | */ | |
164 | if (pte_pfn(pte) >= folio_end_pfn) | |
165 | break; | |
166 | ||
167 | if (any_writable) | |
168 | *any_writable |= writable; | |
169 | ||
170 | nr = pte_batch_hint(ptep, pte); | |
171 | expected_pte = pte_advance_pfn(expected_pte, nr); | |
172 | ptep += nr; | |
173 | } | |
174 | ||
175 | return min(ptep - start_ptep, max_nr); | |
176 | } | |
177 | #endif /* CONFIG_MMU */ | |
178 | ||
512b7931 | 179 | void __acct_reclaim_writeback(pg_data_t *pgdat, struct folio *folio, |
8cd7c588 | 180 | int nr_throttled); |
512b7931 | 181 | static inline void acct_reclaim_writeback(struct folio *folio) |
8cd7c588 | 182 | { |
512b7931 | 183 | pg_data_t *pgdat = folio_pgdat(folio); |
8cd7c588 MG |
184 | int nr_throttled = atomic_read(&pgdat->nr_writeback_throttled); |
185 | ||
186 | if (nr_throttled) | |
512b7931 | 187 | __acct_reclaim_writeback(pgdat, folio, nr_throttled); |
8cd7c588 MG |
188 | } |
189 | ||
d818fca1 MG |
190 | static inline void wake_throttle_isolated(pg_data_t *pgdat) |
191 | { | |
192 | wait_queue_head_t *wqh; | |
193 | ||
194 | wqh = &pgdat->reclaim_wait[VMSCAN_THROTTLE_ISOLATED]; | |
195 | if (waitqueue_active(wqh)) | |
196 | wake_up(wqh); | |
197 | } | |
198 | ||
997f0ecb | 199 | vm_fault_t vmf_anon_prepare(struct vm_fault *vmf); |
2b740303 | 200 | vm_fault_t do_swap_page(struct vm_fault *vmf); |
575ced1c | 201 | void folio_rotate_reclaimable(struct folio *folio); |
2580d554 | 202 | bool __folio_end_writeback(struct folio *folio); |
261b6840 | 203 | void deactivate_file_folio(struct folio *folio); |
018ee47f | 204 | void folio_activate(struct folio *folio); |
8a966ed7 | 205 | |
fd892593 | 206 | void free_pgtables(struct mmu_gather *tlb, struct ma_state *mas, |
763ecb03 | 207 | struct vm_area_struct *start_vma, unsigned long floor, |
98e51a22 | 208 | unsigned long ceiling, bool mm_wr_locked); |
03c4f204 | 209 | void pmd_install(struct mm_struct *mm, pmd_t *pmd, pgtable_t *pte); |
42b77728 | 210 | |
3506659e | 211 | struct zap_details; |
aac45363 MH |
212 | void unmap_page_range(struct mmu_gather *tlb, |
213 | struct vm_area_struct *vma, | |
214 | unsigned long addr, unsigned long end, | |
215 | struct zap_details *details); | |
216 | ||
56a4d67c MWO |
217 | void page_cache_ra_order(struct readahead_control *, struct file_ra_state *, |
218 | unsigned int order); | |
fcd9ae4f | 219 | void force_page_cache_ra(struct readahead_control *, unsigned long nr); |
7b3df3b9 DH |
220 | static inline void force_page_cache_readahead(struct address_space *mapping, |
221 | struct file *file, pgoff_t index, unsigned long nr_to_read) | |
222 | { | |
fcd9ae4f MWO |
223 | DEFINE_READAHEAD(ractl, file, &file->f_ra, mapping, index); |
224 | force_page_cache_ra(&ractl, nr_to_read); | |
7b3df3b9 | 225 | } |
29f175d1 | 226 | |
3392ca12 | 227 | unsigned find_lock_entries(struct address_space *mapping, pgoff_t *start, |
51dcbdac | 228 | pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices); |
9fb6beea | 229 | unsigned find_get_entries(struct address_space *mapping, pgoff_t *start, |
0e499ed3 | 230 | pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices); |
78f42660 | 231 | void filemap_free_folio(struct address_space *mapping, struct folio *folio); |
1e84a3d9 | 232 | int truncate_inode_folio(struct address_space *mapping, struct folio *folio); |
b9a8a419 MWO |
233 | bool truncate_inode_partial_folio(struct folio *folio, loff_t start, |
234 | loff_t end); | |
1e12cbb9 | 235 | long mapping_evict_folio(struct address_space *mapping, struct folio *folio); |
1a0fc811 MWO |
236 | unsigned long mapping_try_invalidate(struct address_space *mapping, |
237 | pgoff_t start, pgoff_t end, unsigned long *nr_failed); | |
5c211ba2 | 238 | |
1eb6234e | 239 | /** |
3eed3ef5 MWO |
240 | * folio_evictable - Test whether a folio is evictable. |
241 | * @folio: The folio to test. | |
1eb6234e | 242 | * |
3eed3ef5 MWO |
243 | * Test whether @folio is evictable -- i.e., should be placed on |
244 | * active/inactive lists vs unevictable list. | |
1eb6234e | 245 | * |
3eed3ef5 MWO |
246 | * Reasons folio might not be evictable: |
247 | * 1. folio's mapping marked unevictable | |
248 | * 2. One of the pages in the folio is part of an mlocked VMA | |
1eb6234e | 249 | */ |
3eed3ef5 MWO |
250 | static inline bool folio_evictable(struct folio *folio) |
251 | { | |
252 | bool ret; | |
253 | ||
254 | /* Prevent address_space of inode and swap cache from being freed */ | |
255 | rcu_read_lock(); | |
256 | ret = !mapping_unevictable(folio_mapping(folio)) && | |
257 | !folio_test_mlocked(folio); | |
258 | rcu_read_unlock(); | |
259 | return ret; | |
260 | } | |
261 | ||
7835e98b | 262 | /* |
0139aa7b | 263 | * Turn a non-refcounted page (->_refcount == 0) into refcounted with |
7835e98b NP |
264 | * a count of one. |
265 | */ | |
266 | static inline void set_page_refcounted(struct page *page) | |
267 | { | |
309381fe | 268 | VM_BUG_ON_PAGE(PageTail(page), page); |
fe896d18 | 269 | VM_BUG_ON_PAGE(page_ref_count(page), page); |
77a8a788 | 270 | set_page_count(page, 1); |
77a8a788 NP |
271 | } |
272 | ||
0201ebf2 DH |
273 | /* |
274 | * Return true if a folio needs ->release_folio() calling upon it. | |
275 | */ | |
276 | static inline bool folio_needs_release(struct folio *folio) | |
277 | { | |
b4fa966f DH |
278 | struct address_space *mapping = folio_mapping(folio); |
279 | ||
280 | return folio_has_private(folio) || | |
281 | (mapping && mapping_release_always(mapping)); | |
0201ebf2 DH |
282 | } |
283 | ||
03f6462a HD |
284 | extern unsigned long highest_memmap_pfn; |
285 | ||
c73322d0 JW |
286 | /* |
287 | * Maximum number of reclaim retries without progress before the OOM | |
288 | * killer is consider the only way forward. | |
289 | */ | |
290 | #define MAX_RECLAIM_RETRIES 16 | |
291 | ||
894bc310 LS |
292 | /* |
293 | * in mm/vmscan.c: | |
294 | */ | |
f7f9c00d | 295 | bool isolate_lru_page(struct page *page); |
be2d5756 | 296 | bool folio_isolate_lru(struct folio *folio); |
ca6d60f3 MWO |
297 | void putback_lru_page(struct page *page); |
298 | void folio_putback_lru(struct folio *folio); | |
c3f4a9a2 | 299 | extern void reclaim_throttle(pg_data_t *pgdat, enum vmscan_throttle_state reason); |
62695a84 | 300 | |
6219049a BL |
301 | /* |
302 | * in mm/rmap.c: | |
303 | */ | |
50722804 | 304 | pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address); |
6219049a | 305 | |
894bc310 LS |
306 | /* |
307 | * in mm/page_alloc.c | |
308 | */ | |
eb8589b4 | 309 | #define K(x) ((x) << (PAGE_SHIFT-10)) |
3c605096 | 310 | |
9420f89d MRI |
311 | extern char * const zone_names[MAX_NR_ZONES]; |
312 | ||
f2fc4b44 MRI |
313 | /* perform sanity checks on struct pages being allocated or freed */ |
314 | DECLARE_STATIC_KEY_MAYBE(CONFIG_DEBUG_VM, check_pages_enabled); | |
315 | ||
e95d372c KW |
316 | extern int min_free_kbytes; |
317 | ||
318 | void setup_per_zone_wmarks(void); | |
319 | void calculate_min_free_kbytes(void); | |
320 | int __meminit init_per_zone_wmark_min(void); | |
321 | void page_alloc_sysctl_init(void); | |
f2fc4b44 | 322 | |
1a6d53a1 VB |
323 | /* |
324 | * Structure for holding the mostly immutable allocation parameters passed | |
325 | * between functions involved in allocations, including the alloc_pages* | |
326 | * family of functions. | |
327 | * | |
97a225e6 | 328 | * nodemask, migratetype and highest_zoneidx are initialized only once in |
84172f4b | 329 | * __alloc_pages() and then never change. |
1a6d53a1 | 330 | * |
97a225e6 | 331 | * zonelist, preferred_zone and highest_zoneidx are set first in |
84172f4b | 332 | * __alloc_pages() for the fast path, and might be later changed |
68956ccb | 333 | * in __alloc_pages_slowpath(). All other functions pass the whole structure |
1a6d53a1 VB |
334 | * by a const pointer. |
335 | */ | |
336 | struct alloc_context { | |
337 | struct zonelist *zonelist; | |
338 | nodemask_t *nodemask; | |
c33d6c06 | 339 | struct zoneref *preferred_zoneref; |
1a6d53a1 | 340 | int migratetype; |
97a225e6 JK |
341 | |
342 | /* | |
343 | * highest_zoneidx represents highest usable zone index of | |
344 | * the allocation request. Due to the nature of the zone, | |
345 | * memory on lower zone than the highest_zoneidx will be | |
346 | * protected by lowmem_reserve[highest_zoneidx]. | |
347 | * | |
348 | * highest_zoneidx is also used by reclaim/compaction to limit | |
349 | * the target zone since higher zone than this index cannot be | |
350 | * usable for this allocation request. | |
351 | */ | |
352 | enum zone_type highest_zoneidx; | |
c9ab0c4f | 353 | bool spread_dirty_pages; |
1a6d53a1 VB |
354 | }; |
355 | ||
8170ac47 ZY |
356 | /* |
357 | * This function returns the order of a free page in the buddy system. In | |
358 | * general, page_zone(page)->lock must be held by the caller to prevent the | |
359 | * page from being allocated in parallel and returning garbage as the order. | |
360 | * If a caller does not hold page_zone(page)->lock, it must guarantee that the | |
361 | * page cannot be allocated or merged in parallel. Alternatively, it must | |
362 | * handle invalid values gracefully, and use buddy_order_unsafe() below. | |
363 | */ | |
364 | static inline unsigned int buddy_order(struct page *page) | |
365 | { | |
366 | /* PageBuddy() must be checked by the caller */ | |
367 | return page_private(page); | |
368 | } | |
369 | ||
370 | /* | |
371 | * Like buddy_order(), but for callers who cannot afford to hold the zone lock. | |
372 | * PageBuddy() should be checked first by the caller to minimize race window, | |
373 | * and invalid values must be handled gracefully. | |
374 | * | |
375 | * READ_ONCE is used so that if the caller assigns the result into a local | |
376 | * variable and e.g. tests it for valid range before using, the compiler cannot | |
377 | * decide to remove the variable and inline the page_private(page) multiple | |
378 | * times, potentially observing different values in the tests and the actual | |
379 | * use of the result. | |
380 | */ | |
381 | #define buddy_order_unsafe(page) READ_ONCE(page_private(page)) | |
382 | ||
383 | /* | |
384 | * This function checks whether a page is free && is the buddy | |
385 | * we can coalesce a page and its buddy if | |
386 | * (a) the buddy is not in a hole (check before calling!) && | |
387 | * (b) the buddy is in the buddy system && | |
388 | * (c) a page and its buddy have the same order && | |
389 | * (d) a page and its buddy are in the same zone. | |
390 | * | |
391 | * For recording whether a page is in the buddy system, we set PageBuddy. | |
392 | * Setting, clearing, and testing PageBuddy is serialized by zone->lock. | |
393 | * | |
394 | * For recording page's order, we use page_private(page). | |
395 | */ | |
396 | static inline bool page_is_buddy(struct page *page, struct page *buddy, | |
397 | unsigned int order) | |
398 | { | |
399 | if (!page_is_guard(buddy) && !PageBuddy(buddy)) | |
400 | return false; | |
401 | ||
402 | if (buddy_order(buddy) != order) | |
403 | return false; | |
404 | ||
405 | /* | |
406 | * zone check is done late to avoid uselessly calculating | |
407 | * zone/node ids for pages that could never merge. | |
408 | */ | |
409 | if (page_zone_id(page) != page_zone_id(buddy)) | |
410 | return false; | |
411 | ||
412 | VM_BUG_ON_PAGE(page_count(buddy) != 0, buddy); | |
413 | ||
414 | return true; | |
415 | } | |
416 | ||
3c605096 JK |
417 | /* |
418 | * Locate the struct page for both the matching buddy in our | |
419 | * pair (buddy1) and the combined O(n+1) page they form (page). | |
420 | * | |
421 | * 1) Any buddy B1 will have an order O twin B2 which satisfies | |
422 | * the following equation: | |
423 | * B2 = B1 ^ (1 << O) | |
424 | * For example, if the starting buddy (buddy2) is #8 its order | |
425 | * 1 buddy is #10: | |
426 | * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10 | |
427 | * | |
428 | * 2) Any buddy B will have an order O+1 parent P which | |
429 | * satisfies the following equation: | |
430 | * P = B & ~(1 << O) | |
431 | * | |
5e0a760b | 432 | * Assumption: *_mem_map is contiguous at least up to MAX_PAGE_ORDER |
3c605096 JK |
433 | */ |
434 | static inline unsigned long | |
76741e77 | 435 | __find_buddy_pfn(unsigned long page_pfn, unsigned int order) |
3c605096 | 436 | { |
76741e77 | 437 | return page_pfn ^ (1 << order); |
3c605096 JK |
438 | } |
439 | ||
8170ac47 ZY |
440 | /* |
441 | * Find the buddy of @page and validate it. | |
442 | * @page: The input page | |
443 | * @pfn: The pfn of the page, it saves a call to page_to_pfn() when the | |
444 | * function is used in the performance-critical __free_one_page(). | |
445 | * @order: The order of the page | |
446 | * @buddy_pfn: The output pointer to the buddy pfn, it also saves a call to | |
447 | * page_to_pfn(). | |
448 | * | |
449 | * The found buddy can be a non PageBuddy, out of @page's zone, or its order is | |
450 | * not the same as @page. The validation is necessary before use it. | |
451 | * | |
452 | * Return: the found buddy page or NULL if not found. | |
453 | */ | |
454 | static inline struct page *find_buddy_page_pfn(struct page *page, | |
455 | unsigned long pfn, unsigned int order, unsigned long *buddy_pfn) | |
456 | { | |
457 | unsigned long __buddy_pfn = __find_buddy_pfn(pfn, order); | |
458 | struct page *buddy; | |
459 | ||
460 | buddy = page + (__buddy_pfn - pfn); | |
461 | if (buddy_pfn) | |
462 | *buddy_pfn = __buddy_pfn; | |
463 | ||
464 | if (page_is_buddy(page, buddy, order)) | |
465 | return buddy; | |
466 | return NULL; | |
467 | } | |
468 | ||
7cf91a98 JK |
469 | extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn, |
470 | unsigned long end_pfn, struct zone *zone); | |
471 | ||
472 | static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn, | |
473 | unsigned long end_pfn, struct zone *zone) | |
474 | { | |
475 | if (zone->contiguous) | |
476 | return pfn_to_page(start_pfn); | |
477 | ||
478 | return __pageblock_pfn_to_page(start_pfn, end_pfn, zone); | |
479 | } | |
480 | ||
904d5857 KW |
481 | void set_zone_contiguous(struct zone *zone); |
482 | ||
483 | static inline void clear_zone_contiguous(struct zone *zone) | |
484 | { | |
485 | zone->contiguous = false; | |
486 | } | |
487 | ||
3c605096 | 488 | extern int __isolate_free_page(struct page *page, unsigned int order); |
624f58d8 AD |
489 | extern void __putback_isolated_page(struct page *page, unsigned int order, |
490 | int mt); | |
7c2ee349 | 491 | extern void memblock_free_pages(struct page *page, unsigned long pfn, |
d70ddd7a | 492 | unsigned int order); |
a9cd410a | 493 | extern void __free_pages_core(struct page *page, unsigned int order); |
9420f89d | 494 | |
1e3be485 TS |
495 | /* |
496 | * This will have no effect, other than possibly generating a warning, if the | |
497 | * caller passes in a non-large folio. | |
498 | */ | |
499 | static inline void folio_set_order(struct folio *folio, unsigned int order) | |
500 | { | |
501 | if (WARN_ON_ONCE(!order || !folio_test_large(folio))) | |
502 | return; | |
503 | ||
ebc1baf5 | 504 | folio->_flags_1 = (folio->_flags_1 & ~0xffUL) | order; |
1e3be485 TS |
505 | #ifdef CONFIG_64BIT |
506 | folio->_folio_nr_pages = 1U << order; | |
507 | #endif | |
508 | } | |
509 | ||
8dc4a8f1 MWO |
510 | void folio_undo_large_rmappable(struct folio *folio); |
511 | ||
23e48832 HD |
512 | static inline struct folio *page_rmappable_folio(struct page *page) |
513 | { | |
514 | struct folio *folio = (struct folio *)page; | |
515 | ||
85edc15a MWO |
516 | if (folio && folio_test_large(folio)) |
517 | folio_set_large_rmappable(folio); | |
23e48832 HD |
518 | return folio; |
519 | } | |
520 | ||
9420f89d MRI |
521 | static inline void prep_compound_head(struct page *page, unsigned int order) |
522 | { | |
523 | struct folio *folio = (struct folio *)page; | |
524 | ||
1e3be485 | 525 | folio_set_order(folio, order); |
9420f89d MRI |
526 | atomic_set(&folio->_entire_mapcount, -1); |
527 | atomic_set(&folio->_nr_pages_mapped, 0); | |
528 | atomic_set(&folio->_pincount, 0); | |
b7b098cf MWO |
529 | if (order > 1) |
530 | INIT_LIST_HEAD(&folio->_deferred_list); | |
9420f89d MRI |
531 | } |
532 | ||
533 | static inline void prep_compound_tail(struct page *head, int tail_idx) | |
534 | { | |
535 | struct page *p = head + tail_idx; | |
536 | ||
537 | p->mapping = TAIL_MAPPING; | |
538 | set_compound_head(p, head); | |
539 | set_page_private(p, 0); | |
540 | } | |
541 | ||
d00181b9 | 542 | extern void prep_compound_page(struct page *page, unsigned int order); |
9420f89d | 543 | |
46f24fd8 JK |
544 | extern void post_alloc_hook(struct page *page, unsigned int order, |
545 | gfp_t gfp_flags); | |
733aea0b ZY |
546 | extern bool free_pages_prepare(struct page *page, unsigned int order); |
547 | ||
42aa83cb | 548 | extern int user_min_free_kbytes; |
20a0307c | 549 | |
90491d87 MWO |
550 | void free_unref_page(struct page *page, unsigned int order); |
551 | void free_unref_folios(struct folio_batch *fbatch); | |
0966aeb4 | 552 | |
68265390 | 553 | extern void zone_pcp_reset(struct zone *zone); |
ec6e8c7e VB |
554 | extern void zone_pcp_disable(struct zone *zone); |
555 | extern void zone_pcp_enable(struct zone *zone); | |
9420f89d | 556 | extern void zone_pcp_init(struct zone *zone); |
68265390 | 557 | |
c803b3c8 MR |
558 | extern void *memmap_alloc(phys_addr_t size, phys_addr_t align, |
559 | phys_addr_t min_addr, | |
560 | int nid, bool exact_nid); | |
561 | ||
e95d372c KW |
562 | void memmap_init_range(unsigned long, int, unsigned long, unsigned long, |
563 | unsigned long, enum meminit_context, struct vmem_altmap *, int); | |
b2c9e2fb | 564 | |
ff9543fd MN |
565 | #if defined CONFIG_COMPACTION || defined CONFIG_CMA |
566 | ||
567 | /* | |
568 | * in mm/compaction.c | |
569 | */ | |
570 | /* | |
571 | * compact_control is used to track pages being migrated and the free pages | |
572 | * they are being migrated to during memory compaction. The free_pfn starts | |
573 | * at the end of a zone and migrate_pfn begins at the start. Movable pages | |
574 | * are moved to the end of a zone during a compaction run and the run | |
575 | * completes when free_pfn <= migrate_pfn | |
576 | */ | |
577 | struct compact_control { | |
733aea0b | 578 | struct list_head freepages[NR_PAGE_ORDERS]; /* List of free pages to migrate to */ |
ff9543fd | 579 | struct list_head migratepages; /* List of pages being migrated */ |
c5fbd937 MG |
580 | unsigned int nr_freepages; /* Number of isolated free pages */ |
581 | unsigned int nr_migratepages; /* Number of pages to migrate */ | |
ff9543fd | 582 | unsigned long free_pfn; /* isolate_freepages search base */ |
c2ad7a1f OS |
583 | /* |
584 | * Acts as an in/out parameter to page isolation for migration. | |
585 | * isolate_migratepages uses it as a search base. | |
586 | * isolate_migratepages_block will update the value to the next pfn | |
587 | * after the last isolated one. | |
588 | */ | |
589 | unsigned long migrate_pfn; | |
70b44595 | 590 | unsigned long fast_start_pfn; /* a pfn to start linear scan from */ |
c5943b9c MG |
591 | struct zone *zone; |
592 | unsigned long total_migrate_scanned; | |
593 | unsigned long total_free_scanned; | |
dbe2d4e4 MG |
594 | unsigned short fast_search_fail;/* failures to use free list searches */ |
595 | short search_order; /* order to start a fast search at */ | |
f25ba6dc VB |
596 | const gfp_t gfp_mask; /* gfp mask of a direct compactor */ |
597 | int order; /* order a direct compactor needs */ | |
d39773a0 | 598 | int migratetype; /* migratetype of direct compactor */ |
f25ba6dc | 599 | const unsigned int alloc_flags; /* alloc flags of a direct compactor */ |
97a225e6 | 600 | const int highest_zoneidx; /* zone index of a direct compactor */ |
e0b9daeb | 601 | enum migrate_mode mode; /* Async or sync migration mode */ |
bb13ffeb | 602 | bool ignore_skip_hint; /* Scan blocks even if marked skip */ |
2583d671 | 603 | bool no_set_skip_hint; /* Don't mark blocks for skipping */ |
9f7e3387 | 604 | bool ignore_block_suitable; /* Scan blocks considered unsuitable */ |
accf6242 | 605 | bool direct_compaction; /* False from kcompactd or /proc/... */ |
facdaa91 | 606 | bool proactive_compaction; /* kcompactd proactive compaction */ |
06ed2998 | 607 | bool whole_zone; /* Whole zone should/has been scanned */ |
d56c1584 | 608 | bool contended; /* Signal lock contention */ |
48731c84 MG |
609 | bool finish_pageblock; /* Scan the remainder of a pageblock. Used |
610 | * when there are potentially transient | |
611 | * isolation or migration failures to | |
612 | * ensure forward progress. | |
613 | */ | |
b06eda09 | 614 | bool alloc_contig; /* alloc_contig_range allocation */ |
ff9543fd MN |
615 | }; |
616 | ||
5e1f0f09 MG |
617 | /* |
618 | * Used in direct compaction when a page should be taken from the freelists | |
619 | * immediately when one is created during the free path. | |
620 | */ | |
621 | struct capture_control { | |
622 | struct compact_control *cc; | |
623 | struct page *page; | |
624 | }; | |
625 | ||
ff9543fd | 626 | unsigned long |
bb13ffeb MG |
627 | isolate_freepages_range(struct compact_control *cc, |
628 | unsigned long start_pfn, unsigned long end_pfn); | |
c2ad7a1f | 629 | int |
edc2ca61 VB |
630 | isolate_migratepages_range(struct compact_control *cc, |
631 | unsigned long low_pfn, unsigned long end_pfn); | |
b2c9e2fb ZY |
632 | |
633 | int __alloc_contig_migrate_range(struct compact_control *cc, | |
c8b36003 RC |
634 | unsigned long start, unsigned long end, |
635 | int migratetype); | |
9420f89d MRI |
636 | |
637 | /* Free whole pageblock and set its migration type to MIGRATE_CMA. */ | |
638 | void init_cma_reserved_pageblock(struct page *page); | |
639 | ||
640 | #endif /* CONFIG_COMPACTION || CONFIG_CMA */ | |
641 | ||
2149cdae JK |
642 | int find_suitable_fallback(struct free_area *area, unsigned int order, |
643 | int migratetype, bool only_stealable, bool *can_steal); | |
ff9543fd | 644 | |
62f31bd4 MRI |
645 | static inline bool free_area_empty(struct free_area *area, int migratetype) |
646 | { | |
647 | return list_empty(&area->free_list[migratetype]); | |
648 | } | |
649 | ||
30bdbb78 KK |
650 | /* |
651 | * These three helpers classifies VMAs for virtual memory accounting. | |
652 | */ | |
653 | ||
654 | /* | |
655 | * Executable code area - executable, not writable, not stack | |
656 | */ | |
d977d56c KK |
657 | static inline bool is_exec_mapping(vm_flags_t flags) |
658 | { | |
30bdbb78 | 659 | return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC; |
d977d56c KK |
660 | } |
661 | ||
30bdbb78 | 662 | /* |
00547ef7 | 663 | * Stack area (including shadow stacks) |
30bdbb78 KK |
664 | * |
665 | * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous: | |
666 | * do_mmap() forbids all other combinations. | |
667 | */ | |
d977d56c KK |
668 | static inline bool is_stack_mapping(vm_flags_t flags) |
669 | { | |
00547ef7 | 670 | return ((flags & VM_STACK) == VM_STACK) || (flags & VM_SHADOW_STACK); |
d977d56c KK |
671 | } |
672 | ||
30bdbb78 KK |
673 | /* |
674 | * Data area - private, writable, not stack | |
675 | */ | |
d977d56c KK |
676 | static inline bool is_data_mapping(vm_flags_t flags) |
677 | { | |
30bdbb78 | 678 | return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE; |
d977d56c KK |
679 | } |
680 | ||
6038def0 | 681 | /* mm/util.c */ |
e05b3453 | 682 | struct anon_vma *folio_anon_vma(struct folio *folio); |
6038def0 | 683 | |
af8e3354 | 684 | #ifdef CONFIG_MMU |
3506659e | 685 | void unmap_mapping_folio(struct folio *folio); |
fc05f566 | 686 | extern long populate_vma_page_range(struct vm_area_struct *vma, |
a78f1ccd | 687 | unsigned long start, unsigned long end, int *locked); |
631426ba DH |
688 | extern long faultin_page_range(struct mm_struct *mm, unsigned long start, |
689 | unsigned long end, bool write, int *locked); | |
b0cc5e89 | 690 | extern bool mlock_future_ok(struct mm_struct *mm, unsigned long flags, |
3c54a298 | 691 | unsigned long bytes); |
28e56657 YF |
692 | |
693 | /* | |
694 | * NOTE: This function can't tell whether the folio is "fully mapped" in the | |
695 | * range. | |
696 | * "fully mapped" means all the pages of folio is associated with the page | |
697 | * table of range while this function just check whether the folio range is | |
be16dd76 | 698 | * within the range [start, end). Function caller needs to do page table |
28e56657 YF |
699 | * check if it cares about the page table association. |
700 | * | |
701 | * Typical usage (like mlock or madvise) is: | |
702 | * Caller knows at least 1 page of folio is associated with page table of VMA | |
703 | * and the range [start, end) is intersect with the VMA range. Caller wants | |
704 | * to know whether the folio is fully associated with the range. It calls | |
705 | * this function to check whether the folio is in the range first. Then checks | |
706 | * the page table to know whether the folio is fully mapped to the range. | |
707 | */ | |
708 | static inline bool | |
709 | folio_within_range(struct folio *folio, struct vm_area_struct *vma, | |
710 | unsigned long start, unsigned long end) | |
711 | { | |
712 | pgoff_t pgoff, addr; | |
dd05f5ec | 713 | unsigned long vma_pglen = vma_pages(vma); |
28e56657 YF |
714 | |
715 | VM_WARN_ON_FOLIO(folio_test_ksm(folio), folio); | |
716 | if (start > end) | |
717 | return false; | |
718 | ||
719 | if (start < vma->vm_start) | |
720 | start = vma->vm_start; | |
721 | ||
722 | if (end > vma->vm_end) | |
723 | end = vma->vm_end; | |
724 | ||
725 | pgoff = folio_pgoff(folio); | |
726 | ||
727 | /* if folio start address is not in vma range */ | |
728 | if (!in_range(pgoff, vma->vm_pgoff, vma_pglen)) | |
729 | return false; | |
730 | ||
731 | addr = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT); | |
732 | ||
733 | return !(addr < start || end - addr < folio_size(folio)); | |
734 | } | |
735 | ||
736 | static inline bool | |
737 | folio_within_vma(struct folio *folio, struct vm_area_struct *vma) | |
738 | { | |
739 | return folio_within_range(folio, vma, vma->vm_start, vma->vm_end); | |
740 | } | |
741 | ||
b291f000 | 742 | /* |
7efecffb | 743 | * mlock_vma_folio() and munlock_vma_folio(): |
cea86fe2 HD |
744 | * should be called with vma's mmap_lock held for read or write, |
745 | * under page table lock for the pte/pmd being added or removed. | |
b291f000 | 746 | * |
4a8ffab0 | 747 | * mlock is usually called at the end of folio_add_*_rmap_*(), munlock at |
4d8f7418 | 748 | * the end of folio_remove_rmap_*(); but new anon folios are managed by |
96f97c43 | 749 | * folio_add_lru_vma() calling mlock_new_folio(). |
b291f000 | 750 | */ |
dcc5d337 MWO |
751 | void mlock_folio(struct folio *folio); |
752 | static inline void mlock_vma_folio(struct folio *folio, | |
1acbc3f9 | 753 | struct vm_area_struct *vma) |
cea86fe2 | 754 | { |
c8263bd6 HD |
755 | /* |
756 | * The VM_SPECIAL check here serves two purposes. | |
757 | * 1) VM_IO check prevents migration from double-counting during mlock. | |
758 | * 2) Although mmap_region() and mlock_fixup() take care that VM_LOCKED | |
759 | * is never left set on a VM_SPECIAL vma, there is an interval while | |
760 | * file->f_op->mmap() is using vm_insert_page(s), when VM_LOCKED may | |
761 | * still be set while VM_SPECIAL bits are added: so ignore it then. | |
762 | */ | |
1acbc3f9 | 763 | if (unlikely((vma->vm_flags & (VM_LOCKED|VM_SPECIAL)) == VM_LOCKED)) |
dcc5d337 MWO |
764 | mlock_folio(folio); |
765 | } | |
766 | ||
96f97c43 | 767 | void munlock_folio(struct folio *folio); |
96f97c43 | 768 | static inline void munlock_vma_folio(struct folio *folio, |
1acbc3f9 | 769 | struct vm_area_struct *vma) |
cea86fe2 | 770 | { |
1acbc3f9 YF |
771 | /* |
772 | * munlock if the function is called. Ideally, we should only | |
773 | * do munlock if any page of folio is unmapped from VMA and | |
774 | * cause folio not fully mapped to VMA. | |
775 | * | |
776 | * But it's not easy to confirm that's the situation. So we | |
777 | * always munlock the folio and page reclaim will correct it | |
778 | * if it's wrong. | |
779 | */ | |
780 | if (unlikely(vma->vm_flags & VM_LOCKED)) | |
96f97c43 | 781 | munlock_folio(folio); |
cea86fe2 | 782 | } |
96f97c43 | 783 | |
96f97c43 LS |
784 | void mlock_new_folio(struct folio *folio); |
785 | bool need_mlock_drain(int cpu); | |
786 | void mlock_drain_local(void); | |
787 | void mlock_drain_remote(int cpu); | |
b291f000 | 788 | |
f55e1014 | 789 | extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma); |
b32967ff | 790 | |
e9b61f19 | 791 | /* |
6a8e0596 MS |
792 | * Return the start of user virtual address at the specific offset within |
793 | * a vma. | |
e9b61f19 KS |
794 | */ |
795 | static inline unsigned long | |
6a8e0596 MS |
796 | vma_pgoff_address(pgoff_t pgoff, unsigned long nr_pages, |
797 | struct vm_area_struct *vma) | |
e9b61f19 | 798 | { |
494334e4 HD |
799 | unsigned long address; |
800 | ||
494334e4 HD |
801 | if (pgoff >= vma->vm_pgoff) { |
802 | address = vma->vm_start + | |
803 | ((pgoff - vma->vm_pgoff) << PAGE_SHIFT); | |
804 | /* Check for address beyond vma (or wrapped through 0?) */ | |
805 | if (address < vma->vm_start || address >= vma->vm_end) | |
806 | address = -EFAULT; | |
6a8e0596 | 807 | } else if (pgoff + nr_pages - 1 >= vma->vm_pgoff) { |
494334e4 HD |
808 | /* Test above avoids possibility of wrap to 0 on 32-bit */ |
809 | address = vma->vm_start; | |
810 | } else { | |
811 | address = -EFAULT; | |
812 | } | |
813 | return address; | |
e9b61f19 KS |
814 | } |
815 | ||
6a8e0596 MS |
816 | /* |
817 | * Return the start of user virtual address of a page within a vma. | |
818 | * Returns -EFAULT if all of the page is outside the range of vma. | |
819 | * If page is a compound head, the entire compound page is considered. | |
820 | */ | |
821 | static inline unsigned long | |
822 | vma_address(struct page *page, struct vm_area_struct *vma) | |
823 | { | |
824 | VM_BUG_ON_PAGE(PageKsm(page), page); /* KSM page->index unusable */ | |
825 | return vma_pgoff_address(page_to_pgoff(page), compound_nr(page), vma); | |
826 | } | |
827 | ||
494334e4 | 828 | /* |
2aff7a47 | 829 | * Then at what user virtual address will none of the range be found in vma? |
494334e4 | 830 | * Assumes that vma_address() already returned a good starting address. |
494334e4 | 831 | */ |
2aff7a47 | 832 | static inline unsigned long vma_address_end(struct page_vma_mapped_walk *pvmw) |
e9b61f19 | 833 | { |
2aff7a47 | 834 | struct vm_area_struct *vma = pvmw->vma; |
494334e4 HD |
835 | pgoff_t pgoff; |
836 | unsigned long address; | |
837 | ||
2aff7a47 MWO |
838 | /* Common case, plus ->pgoff is invalid for KSM */ |
839 | if (pvmw->nr_pages == 1) | |
840 | return pvmw->address + PAGE_SIZE; | |
841 | ||
842 | pgoff = pvmw->pgoff + pvmw->nr_pages; | |
494334e4 HD |
843 | address = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT); |
844 | /* Check for address beyond vma (or wrapped through 0?) */ | |
845 | if (address < vma->vm_start || address > vma->vm_end) | |
846 | address = vma->vm_end; | |
847 | return address; | |
e9b61f19 KS |
848 | } |
849 | ||
89b15332 JW |
850 | static inline struct file *maybe_unlock_mmap_for_io(struct vm_fault *vmf, |
851 | struct file *fpin) | |
852 | { | |
853 | int flags = vmf->flags; | |
854 | ||
855 | if (fpin) | |
856 | return fpin; | |
857 | ||
858 | /* | |
859 | * FAULT_FLAG_RETRY_NOWAIT means we don't want to wait on page locks or | |
c1e8d7c6 | 860 | * anything, so we only pin the file and drop the mmap_lock if only |
4064b982 | 861 | * FAULT_FLAG_ALLOW_RETRY is set, while this is the first attempt. |
89b15332 | 862 | */ |
4064b982 PX |
863 | if (fault_flag_allow_retry_first(flags) && |
864 | !(flags & FAULT_FLAG_RETRY_NOWAIT)) { | |
89b15332 | 865 | fpin = get_file(vmf->vma->vm_file); |
0790e1e2 | 866 | release_fault_lock(vmf); |
89b15332 JW |
867 | } |
868 | return fpin; | |
869 | } | |
af8e3354 | 870 | #else /* !CONFIG_MMU */ |
3506659e | 871 | static inline void unmap_mapping_folio(struct folio *folio) { } |
96f97c43 LS |
872 | static inline void mlock_new_folio(struct folio *folio) { } |
873 | static inline bool need_mlock_drain(int cpu) { return false; } | |
874 | static inline void mlock_drain_local(void) { } | |
875 | static inline void mlock_drain_remote(int cpu) { } | |
4ad0ae8c NP |
876 | static inline void vunmap_range_noflush(unsigned long start, unsigned long end) |
877 | { | |
878 | } | |
af8e3354 | 879 | #endif /* !CONFIG_MMU */ |
894bc310 | 880 | |
6b74ab97 | 881 | /* Memory initialisation debug and verification */ |
9420f89d MRI |
882 | #ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT |
883 | DECLARE_STATIC_KEY_TRUE(deferred_pages); | |
884 | ||
885 | bool __init deferred_grow_zone(struct zone *zone, unsigned int order); | |
886 | #endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */ | |
887 | ||
6b74ab97 MG |
888 | enum mminit_level { |
889 | MMINIT_WARNING, | |
890 | MMINIT_VERIFY, | |
891 | MMINIT_TRACE | |
892 | }; | |
893 | ||
894 | #ifdef CONFIG_DEBUG_MEMORY_INIT | |
895 | ||
896 | extern int mminit_loglevel; | |
897 | ||
898 | #define mminit_dprintk(level, prefix, fmt, arg...) \ | |
899 | do { \ | |
900 | if (level < mminit_loglevel) { \ | |
fc5199d1 | 901 | if (level <= MMINIT_WARNING) \ |
1170532b | 902 | pr_warn("mminit::" prefix " " fmt, ##arg); \ |
fc5199d1 RV |
903 | else \ |
904 | printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \ | |
6b74ab97 MG |
905 | } \ |
906 | } while (0) | |
907 | ||
708614e6 | 908 | extern void mminit_verify_pageflags_layout(void); |
68ad8df4 | 909 | extern void mminit_verify_zonelist(void); |
6b74ab97 MG |
910 | #else |
911 | ||
912 | static inline void mminit_dprintk(enum mminit_level level, | |
913 | const char *prefix, const char *fmt, ...) | |
914 | { | |
915 | } | |
916 | ||
708614e6 MG |
917 | static inline void mminit_verify_pageflags_layout(void) |
918 | { | |
919 | } | |
920 | ||
68ad8df4 MG |
921 | static inline void mminit_verify_zonelist(void) |
922 | { | |
923 | } | |
6b74ab97 | 924 | #endif /* CONFIG_DEBUG_MEMORY_INIT */ |
2dbb51c4 | 925 | |
a5f5f91d MG |
926 | #define NODE_RECLAIM_NOSCAN -2 |
927 | #define NODE_RECLAIM_FULL -1 | |
928 | #define NODE_RECLAIM_SOME 0 | |
929 | #define NODE_RECLAIM_SUCCESS 1 | |
7c116f2b | 930 | |
8b09549c WY |
931 | #ifdef CONFIG_NUMA |
932 | extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int); | |
79c28a41 | 933 | extern int find_next_best_node(int node, nodemask_t *used_node_mask); |
8b09549c WY |
934 | #else |
935 | static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask, | |
936 | unsigned int order) | |
937 | { | |
938 | return NODE_RECLAIM_NOSCAN; | |
939 | } | |
79c28a41 DH |
940 | static inline int find_next_best_node(int node, nodemask_t *used_node_mask) |
941 | { | |
942 | return NUMA_NO_NODE; | |
943 | } | |
8b09549c WY |
944 | #endif |
945 | ||
60f272f6 | 946 | /* |
947 | * mm/memory-failure.c | |
948 | */ | |
31d3d348 WF |
949 | extern int hwpoison_filter(struct page *p); |
950 | ||
7c116f2b WF |
951 | extern u32 hwpoison_filter_dev_major; |
952 | extern u32 hwpoison_filter_dev_minor; | |
478c5ffc WF |
953 | extern u64 hwpoison_filter_flags_mask; |
954 | extern u64 hwpoison_filter_flags_value; | |
4fd466eb | 955 | extern u64 hwpoison_filter_memcg; |
1bfe5feb | 956 | extern u32 hwpoison_filter_enable; |
eb36c587 | 957 | |
dc0ef0df | 958 | extern unsigned long __must_check vm_mmap_pgoff(struct file *, unsigned long, |
eb36c587 | 959 | unsigned long, unsigned long, |
9fbeb5ab | 960 | unsigned long, unsigned long); |
ca57df79 XQ |
961 | |
962 | extern void set_pageblock_order(void); | |
2864f3d0 | 963 | unsigned long reclaim_pages(struct list_head *folio_list, bool ignore_references); |
730ec8c0 | 964 | unsigned int reclaim_clean_pages_from_list(struct zone *zone, |
4bf4f155 | 965 | struct list_head *folio_list); |
d95ea5d1 BZ |
966 | /* The ALLOC_WMARK bits are used as an index to zone->watermark */ |
967 | #define ALLOC_WMARK_MIN WMARK_MIN | |
968 | #define ALLOC_WMARK_LOW WMARK_LOW | |
969 | #define ALLOC_WMARK_HIGH WMARK_HIGH | |
970 | #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */ | |
971 | ||
972 | /* Mask to get the watermark bits */ | |
973 | #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1) | |
974 | ||
cd04ae1e MH |
975 | /* |
976 | * Only MMU archs have async oom victim reclaim - aka oom_reaper so we | |
977 | * cannot assume a reduced access to memory reserves is sufficient for | |
978 | * !MMU | |
979 | */ | |
980 | #ifdef CONFIG_MMU | |
981 | #define ALLOC_OOM 0x08 | |
982 | #else | |
983 | #define ALLOC_OOM ALLOC_NO_WATERMARKS | |
984 | #endif | |
985 | ||
1ebbb218 MG |
986 | #define ALLOC_NON_BLOCK 0x10 /* Caller cannot block. Allow access |
987 | * to 25% of the min watermark or | |
988 | * 62.5% if __GFP_HIGH is set. | |
989 | */ | |
524c4807 MG |
990 | #define ALLOC_MIN_RESERVE 0x20 /* __GFP_HIGH set. Allow access to 50% |
991 | * of the min watermark. | |
992 | */ | |
6bb15450 MG |
993 | #define ALLOC_CPUSET 0x40 /* check for correct cpuset */ |
994 | #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */ | |
995 | #ifdef CONFIG_ZONE_DMA32 | |
996 | #define ALLOC_NOFRAGMENT 0x100 /* avoid mixing pageblock types */ | |
997 | #else | |
998 | #define ALLOC_NOFRAGMENT 0x0 | |
999 | #endif | |
eb2e2b42 | 1000 | #define ALLOC_HIGHATOMIC 0x200 /* Allows access to MIGRATE_HIGHATOMIC */ |
736838e9 | 1001 | #define ALLOC_KSWAPD 0x800 /* allow waking of kswapd, __GFP_KSWAPD_RECLAIM set */ |
d95ea5d1 | 1002 | |
ab350885 | 1003 | /* Flags that allow allocations below the min watermark. */ |
1ebbb218 | 1004 | #define ALLOC_RESERVES (ALLOC_NON_BLOCK|ALLOC_MIN_RESERVE|ALLOC_HIGHATOMIC|ALLOC_OOM) |
ab350885 | 1005 | |
72b252ae MG |
1006 | enum ttu_flags; |
1007 | struct tlbflush_unmap_batch; | |
1008 | ||
ce612879 MH |
1009 | |
1010 | /* | |
1011 | * only for MM internal work items which do not depend on | |
1012 | * any allocations or locks which might depend on allocations | |
1013 | */ | |
1014 | extern struct workqueue_struct *mm_percpu_wq; | |
1015 | ||
72b252ae MG |
1016 | #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH |
1017 | void try_to_unmap_flush(void); | |
d950c947 | 1018 | void try_to_unmap_flush_dirty(void); |
3ea27719 | 1019 | void flush_tlb_batched_pending(struct mm_struct *mm); |
72b252ae MG |
1020 | #else |
1021 | static inline void try_to_unmap_flush(void) | |
1022 | { | |
1023 | } | |
d950c947 MG |
1024 | static inline void try_to_unmap_flush_dirty(void) |
1025 | { | |
1026 | } | |
3ea27719 MG |
1027 | static inline void flush_tlb_batched_pending(struct mm_struct *mm) |
1028 | { | |
1029 | } | |
72b252ae | 1030 | #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */ |
edf14cdb VB |
1031 | |
1032 | extern const struct trace_print_flags pageflag_names[]; | |
4c85c0be | 1033 | extern const struct trace_print_flags pagetype_names[]; |
edf14cdb VB |
1034 | extern const struct trace_print_flags vmaflag_names[]; |
1035 | extern const struct trace_print_flags gfpflag_names[]; | |
1036 | ||
a6ffdc07 XQ |
1037 | static inline bool is_migrate_highatomic(enum migratetype migratetype) |
1038 | { | |
1039 | return migratetype == MIGRATE_HIGHATOMIC; | |
1040 | } | |
1041 | ||
72675e13 | 1042 | void setup_zone_pageset(struct zone *zone); |
19fc7bed JK |
1043 | |
1044 | struct migration_target_control { | |
1045 | int nid; /* preferred node id */ | |
1046 | nodemask_t *nmask; | |
1047 | gfp_t gfp_mask; | |
e42dfe4e | 1048 | enum migrate_reason reason; |
19fc7bed JK |
1049 | }; |
1050 | ||
07073eb0 DH |
1051 | /* |
1052 | * mm/filemap.c | |
1053 | */ | |
1054 | size_t splice_folio_into_pipe(struct pipe_inode_info *pipe, | |
1055 | struct folio *folio, loff_t fpos, size_t size); | |
1056 | ||
b67177ec NP |
1057 | /* |
1058 | * mm/vmalloc.c | |
1059 | */ | |
4ad0ae8c | 1060 | #ifdef CONFIG_MMU |
b6714911 | 1061 | void __init vmalloc_init(void); |
d905ae2b | 1062 | int __must_check vmap_pages_range_noflush(unsigned long addr, unsigned long end, |
b67177ec | 1063 | pgprot_t prot, struct page **pages, unsigned int page_shift); |
4ad0ae8c | 1064 | #else |
b6714911 MRI |
1065 | static inline void vmalloc_init(void) |
1066 | { | |
1067 | } | |
1068 | ||
4ad0ae8c | 1069 | static inline |
d905ae2b | 1070 | int __must_check vmap_pages_range_noflush(unsigned long addr, unsigned long end, |
4ad0ae8c NP |
1071 | pgprot_t prot, struct page **pages, unsigned int page_shift) |
1072 | { | |
1073 | return -EINVAL; | |
1074 | } | |
1075 | #endif | |
1076 | ||
d905ae2b AP |
1077 | int __must_check __vmap_pages_range_noflush(unsigned long addr, |
1078 | unsigned long end, pgprot_t prot, | |
1079 | struct page **pages, unsigned int page_shift); | |
b073d7f8 | 1080 | |
4ad0ae8c | 1081 | void vunmap_range_noflush(unsigned long start, unsigned long end); |
b67177ec | 1082 | |
b073d7f8 AP |
1083 | void __vunmap_range_noflush(unsigned long start, unsigned long end); |
1084 | ||
f8fd525b | 1085 | int numa_migrate_prep(struct folio *folio, struct vm_fault *vmf, |
f4c0d836 YS |
1086 | unsigned long addr, int page_nid, int *flags); |
1087 | ||
27674ef6 | 1088 | void free_zone_device_page(struct page *page); |
b05a79d4 | 1089 | int migrate_device_coherent_page(struct page *page); |
27674ef6 | 1090 | |
ece1ed7b MWO |
1091 | /* |
1092 | * mm/gup.c | |
1093 | */ | |
1094 | struct folio *try_grab_folio(struct page *page, int refs, unsigned int flags); | |
7ce154fe | 1095 | int __must_check try_grab_page(struct page *page, unsigned int flags); |
ece1ed7b | 1096 | |
8b9c1cc0 DH |
1097 | /* |
1098 | * mm/huge_memory.c | |
1099 | */ | |
1100 | struct page *follow_trans_huge_pmd(struct vm_area_struct *vma, | |
1101 | unsigned long addr, pmd_t *pmd, | |
1102 | unsigned int flags); | |
1103 | ||
adb20b0c LS |
1104 | /* |
1105 | * mm/mmap.c | |
1106 | */ | |
93bf5d4a LS |
1107 | struct vm_area_struct *vma_merge_extend(struct vma_iterator *vmi, |
1108 | struct vm_area_struct *vma, | |
1109 | unsigned long delta); | |
adb20b0c | 1110 | |
2c224108 JG |
1111 | enum { |
1112 | /* mark page accessed */ | |
1113 | FOLL_TOUCH = 1 << 16, | |
1114 | /* a retry, previous pass started an IO */ | |
1115 | FOLL_TRIED = 1 << 17, | |
1116 | /* we are working on non-current tsk/mm */ | |
1117 | FOLL_REMOTE = 1 << 18, | |
1118 | /* pages must be released via unpin_user_page */ | |
1119 | FOLL_PIN = 1 << 19, | |
1120 | /* gup_fast: prevent fall-back to slow gup */ | |
1121 | FOLL_FAST_ONLY = 1 << 20, | |
1122 | /* allow unlocking the mmap lock */ | |
1123 | FOLL_UNLOCKABLE = 1 << 21, | |
631426ba DH |
1124 | /* VMA lookup+checks compatible with MADV_POPULATE_(READ|WRITE) */ |
1125 | FOLL_MADV_POPULATE = 1 << 22, | |
2c224108 JG |
1126 | }; |
1127 | ||
0f20bba1 | 1128 | #define INTERNAL_GUP_FLAGS (FOLL_TOUCH | FOLL_TRIED | FOLL_REMOTE | FOLL_PIN | \ |
631426ba DH |
1129 | FOLL_FAST_ONLY | FOLL_UNLOCKABLE | \ |
1130 | FOLL_MADV_POPULATE) | |
0f20bba1 | 1131 | |
63b60512 JG |
1132 | /* |
1133 | * Indicates for which pages that are write-protected in the page table, | |
1134 | * whether GUP has to trigger unsharing via FAULT_FLAG_UNSHARE such that the | |
1135 | * GUP pin will remain consistent with the pages mapped into the page tables | |
1136 | * of the MM. | |
1137 | * | |
1138 | * Temporary unmapping of PageAnonExclusive() pages or clearing of | |
1139 | * PageAnonExclusive() has to protect against concurrent GUP: | |
1140 | * * Ordinary GUP: Using the PT lock | |
1141 | * * GUP-fast and fork(): mm->write_protect_seq | |
1142 | * * GUP-fast and KSM or temporary unmapping (swap, migration): see | |
e3b4b137 | 1143 | * folio_try_share_anon_rmap_*() |
63b60512 JG |
1144 | * |
1145 | * Must be called with the (sub)page that's actually referenced via the | |
1146 | * page table entry, which might not necessarily be the head page for a | |
1147 | * PTE-mapped THP. | |
1148 | * | |
1149 | * If the vma is NULL, we're coming from the GUP-fast path and might have | |
1150 | * to fallback to the slow path just to lookup the vma. | |
1151 | */ | |
1152 | static inline bool gup_must_unshare(struct vm_area_struct *vma, | |
1153 | unsigned int flags, struct page *page) | |
1154 | { | |
1155 | /* | |
1156 | * FOLL_WRITE is implicitly handled correctly as the page table entry | |
1157 | * has to be writable -- and if it references (part of) an anonymous | |
1158 | * folio, that part is required to be marked exclusive. | |
1159 | */ | |
1160 | if ((flags & (FOLL_WRITE | FOLL_PIN)) != FOLL_PIN) | |
1161 | return false; | |
1162 | /* | |
1163 | * Note: PageAnon(page) is stable until the page is actually getting | |
1164 | * freed. | |
1165 | */ | |
1166 | if (!PageAnon(page)) { | |
1167 | /* | |
1168 | * We only care about R/O long-term pining: R/O short-term | |
1169 | * pinning does not have the semantics to observe successive | |
1170 | * changes through the process page tables. | |
1171 | */ | |
1172 | if (!(flags & FOLL_LONGTERM)) | |
1173 | return false; | |
1174 | ||
1175 | /* We really need the vma ... */ | |
1176 | if (!vma) | |
1177 | return true; | |
1178 | ||
1179 | /* | |
1180 | * ... because we only care about writable private ("COW") | |
1181 | * mappings where we have to break COW early. | |
1182 | */ | |
1183 | return is_cow_mapping(vma->vm_flags); | |
1184 | } | |
1185 | ||
e3b4b137 | 1186 | /* Paired with a memory barrier in folio_try_share_anon_rmap_*(). */ |
63b60512 JG |
1187 | if (IS_ENABLED(CONFIG_HAVE_FAST_GUP)) |
1188 | smp_rmb(); | |
1189 | ||
5805192c DH |
1190 | /* |
1191 | * During GUP-fast we might not get called on the head page for a | |
1192 | * hugetlb page that is mapped using cont-PTE, because GUP-fast does | |
1193 | * not work with the abstracted hugetlb PTEs that always point at the | |
1194 | * head page. For hugetlb, PageAnonExclusive only applies on the head | |
1195 | * page (as it cannot be partially COW-shared), so lookup the head page. | |
1196 | */ | |
1197 | if (unlikely(!PageHead(page) && PageHuge(page))) | |
1198 | page = compound_head(page); | |
1199 | ||
63b60512 JG |
1200 | /* |
1201 | * Note that PageKsm() pages cannot be exclusive, and consequently, | |
1202 | * cannot get pinned. | |
1203 | */ | |
1204 | return !PageAnonExclusive(page); | |
1205 | } | |
ece1ed7b | 1206 | |
902c2d91 | 1207 | extern bool mirrored_kernelcore; |
0db31d63 | 1208 | extern bool memblock_has_mirror(void); |
902c2d91 | 1209 | |
412c6ef9 YD |
1210 | static __always_inline void vma_set_range(struct vm_area_struct *vma, |
1211 | unsigned long start, unsigned long end, | |
1212 | pgoff_t pgoff) | |
1213 | { | |
1214 | vma->vm_start = start; | |
1215 | vma->vm_end = end; | |
1216 | vma->vm_pgoff = pgoff; | |
1217 | } | |
1218 | ||
76aefad6 PX |
1219 | static inline bool vma_soft_dirty_enabled(struct vm_area_struct *vma) |
1220 | { | |
1221 | /* | |
1222 | * NOTE: we must check this before VM_SOFTDIRTY on soft-dirty | |
1223 | * enablements, because when without soft-dirty being compiled in, | |
1224 | * VM_SOFTDIRTY is defined as 0x0, then !(vm_flags & VM_SOFTDIRTY) | |
1225 | * will be constantly true. | |
1226 | */ | |
1227 | if (!IS_ENABLED(CONFIG_MEM_SOFT_DIRTY)) | |
1228 | return false; | |
1229 | ||
1230 | /* | |
1231 | * Soft-dirty is kind of special: its tracking is enabled when the | |
1232 | * vma flags not set. | |
1233 | */ | |
1234 | return !(vma->vm_flags & VM_SOFTDIRTY); | |
1235 | } | |
1236 | ||
53bee98d LH |
1237 | static inline void vma_iter_config(struct vma_iterator *vmi, |
1238 | unsigned long index, unsigned long last) | |
1239 | { | |
53bee98d LH |
1240 | __mas_set_range(&vmi->mas, index, last - 1); |
1241 | } | |
1242 | ||
b62b633e LH |
1243 | /* |
1244 | * VMA Iterator functions shared between nommu and mmap | |
1245 | */ | |
b5df0922 LH |
1246 | static inline int vma_iter_prealloc(struct vma_iterator *vmi, |
1247 | struct vm_area_struct *vma) | |
b62b633e | 1248 | { |
b5df0922 | 1249 | return mas_preallocate(&vmi->mas, vma, GFP_KERNEL); |
b62b633e LH |
1250 | } |
1251 | ||
b5df0922 | 1252 | static inline void vma_iter_clear(struct vma_iterator *vmi) |
b62b633e | 1253 | { |
b62b633e LH |
1254 | mas_store_prealloc(&vmi->mas, NULL); |
1255 | } | |
1256 | ||
1257 | static inline struct vm_area_struct *vma_iter_load(struct vma_iterator *vmi) | |
1258 | { | |
1259 | return mas_walk(&vmi->mas); | |
1260 | } | |
1261 | ||
1262 | /* Store a VMA with preallocated memory */ | |
1263 | static inline void vma_iter_store(struct vma_iterator *vmi, | |
1264 | struct vm_area_struct *vma) | |
1265 | { | |
1266 | ||
1267 | #if defined(CONFIG_DEBUG_VM_MAPLE_TREE) | |
067311d3 | 1268 | if (MAS_WARN_ON(&vmi->mas, vmi->mas.status != ma_start && |
36bd9310 LH |
1269 | vmi->mas.index > vma->vm_start)) { |
1270 | pr_warn("%lx > %lx\n store vma %lx-%lx\n into slot %lx-%lx\n", | |
1271 | vmi->mas.index, vma->vm_start, vma->vm_start, | |
1272 | vma->vm_end, vmi->mas.index, vmi->mas.last); | |
b62b633e | 1273 | } |
067311d3 | 1274 | if (MAS_WARN_ON(&vmi->mas, vmi->mas.status != ma_start && |
36bd9310 LH |
1275 | vmi->mas.last < vma->vm_start)) { |
1276 | pr_warn("%lx < %lx\nstore vma %lx-%lx\ninto slot %lx-%lx\n", | |
1277 | vmi->mas.last, vma->vm_start, vma->vm_start, vma->vm_end, | |
1278 | vmi->mas.index, vmi->mas.last); | |
b62b633e LH |
1279 | } |
1280 | #endif | |
1281 | ||
067311d3 | 1282 | if (vmi->mas.status != ma_start && |
b62b633e LH |
1283 | ((vmi->mas.index > vma->vm_start) || (vmi->mas.last < vma->vm_start))) |
1284 | vma_iter_invalidate(vmi); | |
1285 | ||
b5df0922 | 1286 | __mas_set_range(&vmi->mas, vma->vm_start, vma->vm_end - 1); |
b62b633e LH |
1287 | mas_store_prealloc(&vmi->mas, vma); |
1288 | } | |
1289 | ||
1290 | static inline int vma_iter_store_gfp(struct vma_iterator *vmi, | |
1291 | struct vm_area_struct *vma, gfp_t gfp) | |
1292 | { | |
067311d3 | 1293 | if (vmi->mas.status != ma_start && |
b62b633e LH |
1294 | ((vmi->mas.index > vma->vm_start) || (vmi->mas.last < vma->vm_start))) |
1295 | vma_iter_invalidate(vmi); | |
1296 | ||
b5df0922 | 1297 | __mas_set_range(&vmi->mas, vma->vm_start, vma->vm_end - 1); |
b62b633e LH |
1298 | mas_store_gfp(&vmi->mas, vma, gfp); |
1299 | if (unlikely(mas_is_err(&vmi->mas))) | |
1300 | return -ENOMEM; | |
1301 | ||
1302 | return 0; | |
1303 | } | |
440703e0 LH |
1304 | |
1305 | /* | |
1306 | * VMA lock generalization | |
1307 | */ | |
1308 | struct vma_prepare { | |
1309 | struct vm_area_struct *vma; | |
1310 | struct vm_area_struct *adj_next; | |
1311 | struct file *file; | |
1312 | struct address_space *mapping; | |
1313 | struct anon_vma *anon_vma; | |
1314 | struct vm_area_struct *insert; | |
1315 | struct vm_area_struct *remove; | |
1316 | struct vm_area_struct *remove2; | |
1317 | }; | |
3ee0aa9f | 1318 | |
fde1c4ec UA |
1319 | void __meminit __init_single_page(struct page *page, unsigned long pfn, |
1320 | unsigned long zone, int nid); | |
1321 | ||
3ee0aa9f | 1322 | /* shrinker related functions */ |
96f7b2b9 QZ |
1323 | unsigned long shrink_slab(gfp_t gfp_mask, int nid, struct mem_cgroup *memcg, |
1324 | int priority); | |
3ee0aa9f QZ |
1325 | |
1326 | #ifdef CONFIG_SHRINKER_DEBUG | |
f04eba13 LM |
1327 | static inline __printf(2, 0) int shrinker_debugfs_name_alloc( |
1328 | struct shrinker *shrinker, const char *fmt, va_list ap) | |
c42d50ae QZ |
1329 | { |
1330 | shrinker->name = kvasprintf_const(GFP_KERNEL, fmt, ap); | |
1331 | ||
1332 | return shrinker->name ? 0 : -ENOMEM; | |
1333 | } | |
1334 | ||
1335 | static inline void shrinker_debugfs_name_free(struct shrinker *shrinker) | |
1336 | { | |
1337 | kfree_const(shrinker->name); | |
1338 | shrinker->name = NULL; | |
1339 | } | |
1340 | ||
3ee0aa9f QZ |
1341 | extern int shrinker_debugfs_add(struct shrinker *shrinker); |
1342 | extern struct dentry *shrinker_debugfs_detach(struct shrinker *shrinker, | |
1343 | int *debugfs_id); | |
1344 | extern void shrinker_debugfs_remove(struct dentry *debugfs_entry, | |
1345 | int debugfs_id); | |
1346 | #else /* CONFIG_SHRINKER_DEBUG */ | |
1347 | static inline int shrinker_debugfs_add(struct shrinker *shrinker) | |
1348 | { | |
1349 | return 0; | |
1350 | } | |
c42d50ae QZ |
1351 | static inline int shrinker_debugfs_name_alloc(struct shrinker *shrinker, |
1352 | const char *fmt, va_list ap) | |
1353 | { | |
1354 | return 0; | |
1355 | } | |
1356 | static inline void shrinker_debugfs_name_free(struct shrinker *shrinker) | |
1357 | { | |
1358 | } | |
3ee0aa9f QZ |
1359 | static inline struct dentry *shrinker_debugfs_detach(struct shrinker *shrinker, |
1360 | int *debugfs_id) | |
1361 | { | |
1362 | *debugfs_id = -1; | |
1363 | return NULL; | |
1364 | } | |
1365 | static inline void shrinker_debugfs_remove(struct dentry *debugfs_entry, | |
1366 | int debugfs_id) | |
1367 | { | |
1368 | } | |
1369 | #endif /* CONFIG_SHRINKER_DEBUG */ | |
1370 | ||
b64e74e9 CH |
1371 | /* Only track the nodes of mappings with shadow entries */ |
1372 | void workingset_update_node(struct xa_node *node); | |
1373 | extern struct list_lru shadow_nodes; | |
1374 | ||
db971418 | 1375 | #endif /* __MM_INTERNAL_H */ |