Loading...
1/* internal.h: mm/ internal definitions
2 *
3 * Copyright (C) 2004 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
5 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
10 */
11#ifndef __MM_INTERNAL_H
12#define __MM_INTERNAL_H
13
14#include <linux/fs.h>
15#include <linux/mm.h>
16#include <linux/pagemap.h>
17#include <linux/tracepoint-defs.h>
18
19/*
20 * The set of flags that only affect watermark checking and reclaim
21 * behaviour. This is used by the MM to obey the caller constraints
22 * about IO, FS and watermark checking while ignoring placement
23 * hints such as HIGHMEM usage.
24 */
25#define GFP_RECLAIM_MASK (__GFP_RECLAIM|__GFP_HIGH|__GFP_IO|__GFP_FS|\
26 __GFP_NOWARN|__GFP_REPEAT|__GFP_NOFAIL|\
27 __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC)
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
38void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
39 unsigned long floor, unsigned long ceiling);
40
41void unmap_page_range(struct mmu_gather *tlb,
42 struct vm_area_struct *vma,
43 unsigned long addr, unsigned long end,
44 struct zap_details *details);
45
46extern int __do_page_cache_readahead(struct address_space *mapping,
47 struct file *filp, pgoff_t offset, unsigned long nr_to_read,
48 unsigned long lookahead_size);
49
50/*
51 * Submit IO for the read-ahead request in file_ra_state.
52 */
53static inline unsigned long ra_submit(struct file_ra_state *ra,
54 struct address_space *mapping, struct file *filp)
55{
56 return __do_page_cache_readahead(mapping, filp,
57 ra->start, ra->size, ra->async_size);
58}
59
60/*
61 * Turn a non-refcounted page (->_count == 0) into refcounted with
62 * a count of one.
63 */
64static inline void set_page_refcounted(struct page *page)
65{
66 VM_BUG_ON_PAGE(PageTail(page), page);
67 VM_BUG_ON_PAGE(page_ref_count(page), page);
68 set_page_count(page, 1);
69}
70
71extern unsigned long highest_memmap_pfn;
72
73/*
74 * in mm/vmscan.c:
75 */
76extern int isolate_lru_page(struct page *page);
77extern void putback_lru_page(struct page *page);
78extern bool zone_reclaimable(struct zone *zone);
79
80/*
81 * in mm/rmap.c:
82 */
83extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
84
85/*
86 * in mm/page_alloc.c
87 */
88
89/*
90 * Structure for holding the mostly immutable allocation parameters passed
91 * between functions involved in allocations, including the alloc_pages*
92 * family of functions.
93 *
94 * nodemask, migratetype and high_zoneidx are initialized only once in
95 * __alloc_pages_nodemask() and then never change.
96 *
97 * zonelist, preferred_zone and classzone_idx are set first in
98 * __alloc_pages_nodemask() for the fast path, and might be later changed
99 * in __alloc_pages_slowpath(). All other functions pass the whole strucure
100 * by a const pointer.
101 */
102struct alloc_context {
103 struct zonelist *zonelist;
104 nodemask_t *nodemask;
105 struct zone *preferred_zone;
106 int classzone_idx;
107 int migratetype;
108 enum zone_type high_zoneidx;
109 bool spread_dirty_pages;
110};
111
112/*
113 * Locate the struct page for both the matching buddy in our
114 * pair (buddy1) and the combined O(n+1) page they form (page).
115 *
116 * 1) Any buddy B1 will have an order O twin B2 which satisfies
117 * the following equation:
118 * B2 = B1 ^ (1 << O)
119 * For example, if the starting buddy (buddy2) is #8 its order
120 * 1 buddy is #10:
121 * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
122 *
123 * 2) Any buddy B will have an order O+1 parent P which
124 * satisfies the following equation:
125 * P = B & ~(1 << O)
126 *
127 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
128 */
129static inline unsigned long
130__find_buddy_index(unsigned long page_idx, unsigned int order)
131{
132 return page_idx ^ (1 << order);
133}
134
135extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn,
136 unsigned long end_pfn, struct zone *zone);
137
138static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn,
139 unsigned long end_pfn, struct zone *zone)
140{
141 if (zone->contiguous)
142 return pfn_to_page(start_pfn);
143
144 return __pageblock_pfn_to_page(start_pfn, end_pfn, zone);
145}
146
147extern int __isolate_free_page(struct page *page, unsigned int order);
148extern void __free_pages_bootmem(struct page *page, unsigned long pfn,
149 unsigned int order);
150extern void prep_compound_page(struct page *page, unsigned int order);
151extern int user_min_free_kbytes;
152
153#if defined CONFIG_COMPACTION || defined CONFIG_CMA
154
155/*
156 * in mm/compaction.c
157 */
158/*
159 * compact_control is used to track pages being migrated and the free pages
160 * they are being migrated to during memory compaction. The free_pfn starts
161 * at the end of a zone and migrate_pfn begins at the start. Movable pages
162 * are moved to the end of a zone during a compaction run and the run
163 * completes when free_pfn <= migrate_pfn
164 */
165struct compact_control {
166 struct list_head freepages; /* List of free pages to migrate to */
167 struct list_head migratepages; /* List of pages being migrated */
168 unsigned long nr_freepages; /* Number of isolated free pages */
169 unsigned long nr_migratepages; /* Number of pages to migrate */
170 unsigned long free_pfn; /* isolate_freepages search base */
171 unsigned long migrate_pfn; /* isolate_migratepages search base */
172 unsigned long last_migrated_pfn;/* Not yet flushed page being freed */
173 enum migrate_mode mode; /* Async or sync migration mode */
174 bool ignore_skip_hint; /* Scan blocks even if marked skip */
175 bool direct_compaction; /* False from kcompactd or /proc/... */
176 int order; /* order a direct compactor needs */
177 const gfp_t gfp_mask; /* gfp mask of a direct compactor */
178 const int alloc_flags; /* alloc flags of a direct compactor */
179 const int classzone_idx; /* zone index of a direct compactor */
180 struct zone *zone;
181 int contended; /* Signal need_sched() or lock
182 * contention detected during
183 * compaction
184 */
185};
186
187unsigned long
188isolate_freepages_range(struct compact_control *cc,
189 unsigned long start_pfn, unsigned long end_pfn);
190unsigned long
191isolate_migratepages_range(struct compact_control *cc,
192 unsigned long low_pfn, unsigned long end_pfn);
193int find_suitable_fallback(struct free_area *area, unsigned int order,
194 int migratetype, bool only_stealable, bool *can_steal);
195
196#endif
197
198/*
199 * This function returns the order of a free page in the buddy system. In
200 * general, page_zone(page)->lock must be held by the caller to prevent the
201 * page from being allocated in parallel and returning garbage as the order.
202 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
203 * page cannot be allocated or merged in parallel. Alternatively, it must
204 * handle invalid values gracefully, and use page_order_unsafe() below.
205 */
206static inline unsigned int page_order(struct page *page)
207{
208 /* PageBuddy() must be checked by the caller */
209 return page_private(page);
210}
211
212/*
213 * Like page_order(), but for callers who cannot afford to hold the zone lock.
214 * PageBuddy() should be checked first by the caller to minimize race window,
215 * and invalid values must be handled gracefully.
216 *
217 * READ_ONCE is used so that if the caller assigns the result into a local
218 * variable and e.g. tests it for valid range before using, the compiler cannot
219 * decide to remove the variable and inline the page_private(page) multiple
220 * times, potentially observing different values in the tests and the actual
221 * use of the result.
222 */
223#define page_order_unsafe(page) READ_ONCE(page_private(page))
224
225static inline bool is_cow_mapping(vm_flags_t flags)
226{
227 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
228}
229
230/*
231 * These three helpers classifies VMAs for virtual memory accounting.
232 */
233
234/*
235 * Executable code area - executable, not writable, not stack
236 */
237static inline bool is_exec_mapping(vm_flags_t flags)
238{
239 return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC;
240}
241
242/*
243 * Stack area - atomatically grows in one direction
244 *
245 * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
246 * do_mmap() forbids all other combinations.
247 */
248static inline bool is_stack_mapping(vm_flags_t flags)
249{
250 return (flags & VM_STACK) == VM_STACK;
251}
252
253/*
254 * Data area - private, writable, not stack
255 */
256static inline bool is_data_mapping(vm_flags_t flags)
257{
258 return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE;
259}
260
261/* mm/util.c */
262void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
263 struct vm_area_struct *prev, struct rb_node *rb_parent);
264
265#ifdef CONFIG_MMU
266extern long populate_vma_page_range(struct vm_area_struct *vma,
267 unsigned long start, unsigned long end, int *nonblocking);
268extern void munlock_vma_pages_range(struct vm_area_struct *vma,
269 unsigned long start, unsigned long end);
270static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
271{
272 munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
273}
274
275/*
276 * must be called with vma's mmap_sem held for read or write, and page locked.
277 */
278extern void mlock_vma_page(struct page *page);
279extern unsigned int munlock_vma_page(struct page *page);
280
281/*
282 * Clear the page's PageMlocked(). This can be useful in a situation where
283 * we want to unconditionally remove a page from the pagecache -- e.g.,
284 * on truncation or freeing.
285 *
286 * It is legal to call this function for any page, mlocked or not.
287 * If called for a page that is still mapped by mlocked vmas, all we do
288 * is revert to lazy LRU behaviour -- semantics are not broken.
289 */
290extern void clear_page_mlock(struct page *page);
291
292/*
293 * mlock_migrate_page - called only from migrate_misplaced_transhuge_page()
294 * (because that does not go through the full procedure of migration ptes):
295 * to migrate the Mlocked page flag; update statistics.
296 */
297static inline void mlock_migrate_page(struct page *newpage, struct page *page)
298{
299 if (TestClearPageMlocked(page)) {
300 int nr_pages = hpage_nr_pages(page);
301
302 /* Holding pmd lock, no change in irq context: __mod is safe */
303 __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
304 SetPageMlocked(newpage);
305 __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
306 }
307}
308
309extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
310
311/*
312 * At what user virtual address is page expected in @vma?
313 */
314static inline unsigned long
315__vma_address(struct page *page, struct vm_area_struct *vma)
316{
317 pgoff_t pgoff = page_to_pgoff(page);
318 return vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
319}
320
321static inline unsigned long
322vma_address(struct page *page, struct vm_area_struct *vma)
323{
324 unsigned long address = __vma_address(page, vma);
325
326 /* page should be within @vma mapping range */
327 VM_BUG_ON_VMA(address < vma->vm_start || address >= vma->vm_end, vma);
328
329 return address;
330}
331
332#else /* !CONFIG_MMU */
333static inline void clear_page_mlock(struct page *page) { }
334static inline void mlock_vma_page(struct page *page) { }
335static inline void mlock_migrate_page(struct page *new, struct page *old) { }
336
337#endif /* !CONFIG_MMU */
338
339/*
340 * Return the mem_map entry representing the 'offset' subpage within
341 * the maximally aligned gigantic page 'base'. Handle any discontiguity
342 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
343 */
344static inline struct page *mem_map_offset(struct page *base, int offset)
345{
346 if (unlikely(offset >= MAX_ORDER_NR_PAGES))
347 return nth_page(base, offset);
348 return base + offset;
349}
350
351/*
352 * Iterator over all subpages within the maximally aligned gigantic
353 * page 'base'. Handle any discontiguity in the mem_map.
354 */
355static inline struct page *mem_map_next(struct page *iter,
356 struct page *base, int offset)
357{
358 if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
359 unsigned long pfn = page_to_pfn(base) + offset;
360 if (!pfn_valid(pfn))
361 return NULL;
362 return pfn_to_page(pfn);
363 }
364 return iter + 1;
365}
366
367/*
368 * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
369 * so all functions starting at paging_init should be marked __init
370 * in those cases. SPARSEMEM, however, allows for memory hotplug,
371 * and alloc_bootmem_node is not used.
372 */
373#ifdef CONFIG_SPARSEMEM
374#define __paginginit __meminit
375#else
376#define __paginginit __init
377#endif
378
379/* Memory initialisation debug and verification */
380enum mminit_level {
381 MMINIT_WARNING,
382 MMINIT_VERIFY,
383 MMINIT_TRACE
384};
385
386#ifdef CONFIG_DEBUG_MEMORY_INIT
387
388extern int mminit_loglevel;
389
390#define mminit_dprintk(level, prefix, fmt, arg...) \
391do { \
392 if (level < mminit_loglevel) { \
393 if (level <= MMINIT_WARNING) \
394 pr_warn("mminit::" prefix " " fmt, ##arg); \
395 else \
396 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
397 } \
398} while (0)
399
400extern void mminit_verify_pageflags_layout(void);
401extern void mminit_verify_zonelist(void);
402#else
403
404static inline void mminit_dprintk(enum mminit_level level,
405 const char *prefix, const char *fmt, ...)
406{
407}
408
409static inline void mminit_verify_pageflags_layout(void)
410{
411}
412
413static inline void mminit_verify_zonelist(void)
414{
415}
416#endif /* CONFIG_DEBUG_MEMORY_INIT */
417
418/* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
419#if defined(CONFIG_SPARSEMEM)
420extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
421 unsigned long *end_pfn);
422#else
423static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
424 unsigned long *end_pfn)
425{
426}
427#endif /* CONFIG_SPARSEMEM */
428
429#define ZONE_RECLAIM_NOSCAN -2
430#define ZONE_RECLAIM_FULL -1
431#define ZONE_RECLAIM_SOME 0
432#define ZONE_RECLAIM_SUCCESS 1
433
434extern int hwpoison_filter(struct page *p);
435
436extern u32 hwpoison_filter_dev_major;
437extern u32 hwpoison_filter_dev_minor;
438extern u64 hwpoison_filter_flags_mask;
439extern u64 hwpoison_filter_flags_value;
440extern u64 hwpoison_filter_memcg;
441extern u32 hwpoison_filter_enable;
442
443extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
444 unsigned long, unsigned long,
445 unsigned long, unsigned long);
446
447extern void set_pageblock_order(void);
448unsigned long reclaim_clean_pages_from_list(struct zone *zone,
449 struct list_head *page_list);
450/* The ALLOC_WMARK bits are used as an index to zone->watermark */
451#define ALLOC_WMARK_MIN WMARK_MIN
452#define ALLOC_WMARK_LOW WMARK_LOW
453#define ALLOC_WMARK_HIGH WMARK_HIGH
454#define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
455
456/* Mask to get the watermark bits */
457#define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
458
459#define ALLOC_HARDER 0x10 /* try to alloc harder */
460#define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
461#define ALLOC_CPUSET 0x40 /* check for correct cpuset */
462#define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
463#define ALLOC_FAIR 0x100 /* fair zone allocation */
464
465enum ttu_flags;
466struct tlbflush_unmap_batch;
467
468#ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
469void try_to_unmap_flush(void);
470void try_to_unmap_flush_dirty(void);
471#else
472static inline void try_to_unmap_flush(void)
473{
474}
475static inline void try_to_unmap_flush_dirty(void)
476{
477}
478
479#endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
480
481extern const struct trace_print_flags pageflag_names[];
482extern const struct trace_print_flags vmaflag_names[];
483extern const struct trace_print_flags gfpflag_names[];
484
485#endif /* __MM_INTERNAL_H */
1/* SPDX-License-Identifier: GPL-2.0-or-later */
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)
6 */
7#ifndef __MM_INTERNAL_H
8#define __MM_INTERNAL_H
9
10#include <linux/fs.h>
11#include <linux/mm.h>
12#include <linux/pagemap.h>
13#include <linux/rmap.h>
14#include <linux/tracepoint-defs.h>
15
16struct folio_batch;
17
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|\
25 __GFP_NOWARN|__GFP_RETRY_MAYFAIL|__GFP_NOFAIL|\
26 __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC|\
27 __GFP_NOLOCKDEP)
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
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
53void page_writeback_init(void);
54
55/*
56 * If a 16GB hugetlb folio were mapped by PTEs of all of its 4kB pages,
57 * its nr_pages_mapped would be 0x400000: choose the ENTIRELY_MAPPED bit
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 */
61#define ENTIRELY_MAPPED 0x800000
62#define FOLIO_PAGES_MAPPED (ENTIRELY_MAPPED - 1)
63
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
70/*
71 * How many individual pages have an elevated _mapcount. Excludes
72 * the folio's entire_mapcount.
73 */
74static inline int folio_nr_pages_mapped(struct folio *folio)
75{
76 return atomic_read(&folio->_nr_pages_mapped) & FOLIO_PAGES_MAPPED;
77}
78
79static 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
86void __acct_reclaim_writeback(pg_data_t *pgdat, struct folio *folio,
87 int nr_throttled);
88static inline void acct_reclaim_writeback(struct folio *folio)
89{
90 pg_data_t *pgdat = folio_pgdat(folio);
91 int nr_throttled = atomic_read(&pgdat->nr_writeback_throttled);
92
93 if (nr_throttled)
94 __acct_reclaim_writeback(pgdat, folio, nr_throttled);
95}
96
97static inline void wake_throttle_isolated(pg_data_t *pgdat)
98{
99 wait_queue_head_t *wqh;
100
101 wqh = &pgdat->reclaim_wait[VMSCAN_THROTTLE_ISOLATED];
102 if (waitqueue_active(wqh))
103 wake_up(wqh);
104}
105
106vm_fault_t do_swap_page(struct vm_fault *vmf);
107void folio_rotate_reclaimable(struct folio *folio);
108bool __folio_end_writeback(struct folio *folio);
109void deactivate_file_folio(struct folio *folio);
110void folio_activate(struct folio *folio);
111
112void free_pgtables(struct mmu_gather *tlb, struct ma_state *mas,
113 struct vm_area_struct *start_vma, unsigned long floor,
114 unsigned long ceiling, bool mm_wr_locked);
115void pmd_install(struct mm_struct *mm, pmd_t *pmd, pgtable_t *pte);
116
117struct zap_details;
118void unmap_page_range(struct mmu_gather *tlb,
119 struct vm_area_struct *vma,
120 unsigned long addr, unsigned long end,
121 struct zap_details *details);
122
123void page_cache_ra_order(struct readahead_control *, struct file_ra_state *,
124 unsigned int order);
125void force_page_cache_ra(struct readahead_control *, unsigned long nr);
126static inline void force_page_cache_readahead(struct address_space *mapping,
127 struct file *file, pgoff_t index, unsigned long nr_to_read)
128{
129 DEFINE_READAHEAD(ractl, file, &file->f_ra, mapping, index);
130 force_page_cache_ra(&ractl, nr_to_read);
131}
132
133unsigned find_lock_entries(struct address_space *mapping, pgoff_t *start,
134 pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices);
135unsigned find_get_entries(struct address_space *mapping, pgoff_t *start,
136 pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices);
137void filemap_free_folio(struct address_space *mapping, struct folio *folio);
138int truncate_inode_folio(struct address_space *mapping, struct folio *folio);
139bool truncate_inode_partial_folio(struct folio *folio, loff_t start,
140 loff_t end);
141long mapping_evict_folio(struct address_space *mapping, struct folio *folio);
142unsigned long mapping_try_invalidate(struct address_space *mapping,
143 pgoff_t start, pgoff_t end, unsigned long *nr_failed);
144
145/**
146 * folio_evictable - Test whether a folio is evictable.
147 * @folio: The folio to test.
148 *
149 * Test whether @folio is evictable -- i.e., should be placed on
150 * active/inactive lists vs unevictable list.
151 *
152 * Reasons folio might not be evictable:
153 * 1. folio's mapping marked unevictable
154 * 2. One of the pages in the folio is part of an mlocked VMA
155 */
156static inline bool folio_evictable(struct folio *folio)
157{
158 bool ret;
159
160 /* Prevent address_space of inode and swap cache from being freed */
161 rcu_read_lock();
162 ret = !mapping_unevictable(folio_mapping(folio)) &&
163 !folio_test_mlocked(folio);
164 rcu_read_unlock();
165 return ret;
166}
167
168/*
169 * Turn a non-refcounted page (->_refcount == 0) into refcounted with
170 * a count of one.
171 */
172static inline void set_page_refcounted(struct page *page)
173{
174 VM_BUG_ON_PAGE(PageTail(page), page);
175 VM_BUG_ON_PAGE(page_ref_count(page), page);
176 set_page_count(page, 1);
177}
178
179/*
180 * Return true if a folio needs ->release_folio() calling upon it.
181 */
182static inline bool folio_needs_release(struct folio *folio)
183{
184 struct address_space *mapping = folio_mapping(folio);
185
186 return folio_has_private(folio) ||
187 (mapping && mapping_release_always(mapping));
188}
189
190extern unsigned long highest_memmap_pfn;
191
192/*
193 * Maximum number of reclaim retries without progress before the OOM
194 * killer is consider the only way forward.
195 */
196#define MAX_RECLAIM_RETRIES 16
197
198/*
199 * in mm/vmscan.c:
200 */
201bool isolate_lru_page(struct page *page);
202bool folio_isolate_lru(struct folio *folio);
203void putback_lru_page(struct page *page);
204void folio_putback_lru(struct folio *folio);
205extern void reclaim_throttle(pg_data_t *pgdat, enum vmscan_throttle_state reason);
206
207/*
208 * in mm/rmap.c:
209 */
210pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
211
212/*
213 * in mm/page_alloc.c
214 */
215#define K(x) ((x) << (PAGE_SHIFT-10))
216
217extern char * const zone_names[MAX_NR_ZONES];
218
219/* perform sanity checks on struct pages being allocated or freed */
220DECLARE_STATIC_KEY_MAYBE(CONFIG_DEBUG_VM, check_pages_enabled);
221
222extern int min_free_kbytes;
223
224void setup_per_zone_wmarks(void);
225void calculate_min_free_kbytes(void);
226int __meminit init_per_zone_wmark_min(void);
227void page_alloc_sysctl_init(void);
228
229/*
230 * Structure for holding the mostly immutable allocation parameters passed
231 * between functions involved in allocations, including the alloc_pages*
232 * family of functions.
233 *
234 * nodemask, migratetype and highest_zoneidx are initialized only once in
235 * __alloc_pages() and then never change.
236 *
237 * zonelist, preferred_zone and highest_zoneidx are set first in
238 * __alloc_pages() for the fast path, and might be later changed
239 * in __alloc_pages_slowpath(). All other functions pass the whole structure
240 * by a const pointer.
241 */
242struct alloc_context {
243 struct zonelist *zonelist;
244 nodemask_t *nodemask;
245 struct zoneref *preferred_zoneref;
246 int migratetype;
247
248 /*
249 * highest_zoneidx represents highest usable zone index of
250 * the allocation request. Due to the nature of the zone,
251 * memory on lower zone than the highest_zoneidx will be
252 * protected by lowmem_reserve[highest_zoneidx].
253 *
254 * highest_zoneidx is also used by reclaim/compaction to limit
255 * the target zone since higher zone than this index cannot be
256 * usable for this allocation request.
257 */
258 enum zone_type highest_zoneidx;
259 bool spread_dirty_pages;
260};
261
262/*
263 * This function returns the order of a free page in the buddy system. In
264 * general, page_zone(page)->lock must be held by the caller to prevent the
265 * page from being allocated in parallel and returning garbage as the order.
266 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
267 * page cannot be allocated or merged in parallel. Alternatively, it must
268 * handle invalid values gracefully, and use buddy_order_unsafe() below.
269 */
270static inline unsigned int buddy_order(struct page *page)
271{
272 /* PageBuddy() must be checked by the caller */
273 return page_private(page);
274}
275
276/*
277 * Like buddy_order(), but for callers who cannot afford to hold the zone lock.
278 * PageBuddy() should be checked first by the caller to minimize race window,
279 * and invalid values must be handled gracefully.
280 *
281 * READ_ONCE is used so that if the caller assigns the result into a local
282 * variable and e.g. tests it for valid range before using, the compiler cannot
283 * decide to remove the variable and inline the page_private(page) multiple
284 * times, potentially observing different values in the tests and the actual
285 * use of the result.
286 */
287#define buddy_order_unsafe(page) READ_ONCE(page_private(page))
288
289/*
290 * This function checks whether a page is free && is the buddy
291 * we can coalesce a page and its buddy if
292 * (a) the buddy is not in a hole (check before calling!) &&
293 * (b) the buddy is in the buddy system &&
294 * (c) a page and its buddy have the same order &&
295 * (d) a page and its buddy are in the same zone.
296 *
297 * For recording whether a page is in the buddy system, we set PageBuddy.
298 * Setting, clearing, and testing PageBuddy is serialized by zone->lock.
299 *
300 * For recording page's order, we use page_private(page).
301 */
302static inline bool page_is_buddy(struct page *page, struct page *buddy,
303 unsigned int order)
304{
305 if (!page_is_guard(buddy) && !PageBuddy(buddy))
306 return false;
307
308 if (buddy_order(buddy) != order)
309 return false;
310
311 /*
312 * zone check is done late to avoid uselessly calculating
313 * zone/node ids for pages that could never merge.
314 */
315 if (page_zone_id(page) != page_zone_id(buddy))
316 return false;
317
318 VM_BUG_ON_PAGE(page_count(buddy) != 0, buddy);
319
320 return true;
321}
322
323/*
324 * Locate the struct page for both the matching buddy in our
325 * pair (buddy1) and the combined O(n+1) page they form (page).
326 *
327 * 1) Any buddy B1 will have an order O twin B2 which satisfies
328 * the following equation:
329 * B2 = B1 ^ (1 << O)
330 * For example, if the starting buddy (buddy2) is #8 its order
331 * 1 buddy is #10:
332 * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
333 *
334 * 2) Any buddy B will have an order O+1 parent P which
335 * satisfies the following equation:
336 * P = B & ~(1 << O)
337 *
338 * Assumption: *_mem_map is contiguous at least up to MAX_PAGE_ORDER
339 */
340static inline unsigned long
341__find_buddy_pfn(unsigned long page_pfn, unsigned int order)
342{
343 return page_pfn ^ (1 << order);
344}
345
346/*
347 * Find the buddy of @page and validate it.
348 * @page: The input page
349 * @pfn: The pfn of the page, it saves a call to page_to_pfn() when the
350 * function is used in the performance-critical __free_one_page().
351 * @order: The order of the page
352 * @buddy_pfn: The output pointer to the buddy pfn, it also saves a call to
353 * page_to_pfn().
354 *
355 * The found buddy can be a non PageBuddy, out of @page's zone, or its order is
356 * not the same as @page. The validation is necessary before use it.
357 *
358 * Return: the found buddy page or NULL if not found.
359 */
360static inline struct page *find_buddy_page_pfn(struct page *page,
361 unsigned long pfn, unsigned int order, unsigned long *buddy_pfn)
362{
363 unsigned long __buddy_pfn = __find_buddy_pfn(pfn, order);
364 struct page *buddy;
365
366 buddy = page + (__buddy_pfn - pfn);
367 if (buddy_pfn)
368 *buddy_pfn = __buddy_pfn;
369
370 if (page_is_buddy(page, buddy, order))
371 return buddy;
372 return NULL;
373}
374
375extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn,
376 unsigned long end_pfn, struct zone *zone);
377
378static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn,
379 unsigned long end_pfn, struct zone *zone)
380{
381 if (zone->contiguous)
382 return pfn_to_page(start_pfn);
383
384 return __pageblock_pfn_to_page(start_pfn, end_pfn, zone);
385}
386
387void set_zone_contiguous(struct zone *zone);
388
389static inline void clear_zone_contiguous(struct zone *zone)
390{
391 zone->contiguous = false;
392}
393
394extern int __isolate_free_page(struct page *page, unsigned int order);
395extern void __putback_isolated_page(struct page *page, unsigned int order,
396 int mt);
397extern void memblock_free_pages(struct page *page, unsigned long pfn,
398 unsigned int order);
399extern void __free_pages_core(struct page *page, unsigned int order);
400
401/*
402 * This will have no effect, other than possibly generating a warning, if the
403 * caller passes in a non-large folio.
404 */
405static inline void folio_set_order(struct folio *folio, unsigned int order)
406{
407 if (WARN_ON_ONCE(!order || !folio_test_large(folio)))
408 return;
409
410 folio->_flags_1 = (folio->_flags_1 & ~0xffUL) | order;
411#ifdef CONFIG_64BIT
412 folio->_folio_nr_pages = 1U << order;
413#endif
414}
415
416void folio_undo_large_rmappable(struct folio *folio);
417
418static inline struct folio *page_rmappable_folio(struct page *page)
419{
420 struct folio *folio = (struct folio *)page;
421
422 if (folio && folio_order(folio) > 1)
423 folio_prep_large_rmappable(folio);
424 return folio;
425}
426
427static inline void prep_compound_head(struct page *page, unsigned int order)
428{
429 struct folio *folio = (struct folio *)page;
430
431 folio_set_order(folio, order);
432 atomic_set(&folio->_entire_mapcount, -1);
433 atomic_set(&folio->_nr_pages_mapped, 0);
434 atomic_set(&folio->_pincount, 0);
435}
436
437static inline void prep_compound_tail(struct page *head, int tail_idx)
438{
439 struct page *p = head + tail_idx;
440
441 p->mapping = TAIL_MAPPING;
442 set_compound_head(p, head);
443 set_page_private(p, 0);
444}
445
446extern void prep_compound_page(struct page *page, unsigned int order);
447
448extern void post_alloc_hook(struct page *page, unsigned int order,
449 gfp_t gfp_flags);
450extern int user_min_free_kbytes;
451
452extern void free_unref_page(struct page *page, unsigned int order);
453extern void free_unref_page_list(struct list_head *list);
454
455extern void zone_pcp_reset(struct zone *zone);
456extern void zone_pcp_disable(struct zone *zone);
457extern void zone_pcp_enable(struct zone *zone);
458extern void zone_pcp_init(struct zone *zone);
459
460extern void *memmap_alloc(phys_addr_t size, phys_addr_t align,
461 phys_addr_t min_addr,
462 int nid, bool exact_nid);
463
464void memmap_init_range(unsigned long, int, unsigned long, unsigned long,
465 unsigned long, enum meminit_context, struct vmem_altmap *, int);
466
467
468int split_free_page(struct page *free_page,
469 unsigned int order, unsigned long split_pfn_offset);
470
471#if defined CONFIG_COMPACTION || defined CONFIG_CMA
472
473/*
474 * in mm/compaction.c
475 */
476/*
477 * compact_control is used to track pages being migrated and the free pages
478 * they are being migrated to during memory compaction. The free_pfn starts
479 * at the end of a zone and migrate_pfn begins at the start. Movable pages
480 * are moved to the end of a zone during a compaction run and the run
481 * completes when free_pfn <= migrate_pfn
482 */
483struct compact_control {
484 struct list_head freepages; /* List of free pages to migrate to */
485 struct list_head migratepages; /* List of pages being migrated */
486 unsigned int nr_freepages; /* Number of isolated free pages */
487 unsigned int nr_migratepages; /* Number of pages to migrate */
488 unsigned long free_pfn; /* isolate_freepages search base */
489 /*
490 * Acts as an in/out parameter to page isolation for migration.
491 * isolate_migratepages uses it as a search base.
492 * isolate_migratepages_block will update the value to the next pfn
493 * after the last isolated one.
494 */
495 unsigned long migrate_pfn;
496 unsigned long fast_start_pfn; /* a pfn to start linear scan from */
497 struct zone *zone;
498 unsigned long total_migrate_scanned;
499 unsigned long total_free_scanned;
500 unsigned short fast_search_fail;/* failures to use free list searches */
501 short search_order; /* order to start a fast search at */
502 const gfp_t gfp_mask; /* gfp mask of a direct compactor */
503 int order; /* order a direct compactor needs */
504 int migratetype; /* migratetype of direct compactor */
505 const unsigned int alloc_flags; /* alloc flags of a direct compactor */
506 const int highest_zoneidx; /* zone index of a direct compactor */
507 enum migrate_mode mode; /* Async or sync migration mode */
508 bool ignore_skip_hint; /* Scan blocks even if marked skip */
509 bool no_set_skip_hint; /* Don't mark blocks for skipping */
510 bool ignore_block_suitable; /* Scan blocks considered unsuitable */
511 bool direct_compaction; /* False from kcompactd or /proc/... */
512 bool proactive_compaction; /* kcompactd proactive compaction */
513 bool whole_zone; /* Whole zone should/has been scanned */
514 bool contended; /* Signal lock contention */
515 bool finish_pageblock; /* Scan the remainder of a pageblock. Used
516 * when there are potentially transient
517 * isolation or migration failures to
518 * ensure forward progress.
519 */
520 bool alloc_contig; /* alloc_contig_range allocation */
521};
522
523/*
524 * Used in direct compaction when a page should be taken from the freelists
525 * immediately when one is created during the free path.
526 */
527struct capture_control {
528 struct compact_control *cc;
529 struct page *page;
530};
531
532unsigned long
533isolate_freepages_range(struct compact_control *cc,
534 unsigned long start_pfn, unsigned long end_pfn);
535int
536isolate_migratepages_range(struct compact_control *cc,
537 unsigned long low_pfn, unsigned long end_pfn);
538
539int __alloc_contig_migrate_range(struct compact_control *cc,
540 unsigned long start, unsigned long end);
541
542/* Free whole pageblock and set its migration type to MIGRATE_CMA. */
543void init_cma_reserved_pageblock(struct page *page);
544
545#endif /* CONFIG_COMPACTION || CONFIG_CMA */
546
547int find_suitable_fallback(struct free_area *area, unsigned int order,
548 int migratetype, bool only_stealable, bool *can_steal);
549
550static inline bool free_area_empty(struct free_area *area, int migratetype)
551{
552 return list_empty(&area->free_list[migratetype]);
553}
554
555/*
556 * These three helpers classifies VMAs for virtual memory accounting.
557 */
558
559/*
560 * Executable code area - executable, not writable, not stack
561 */
562static inline bool is_exec_mapping(vm_flags_t flags)
563{
564 return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC;
565}
566
567/*
568 * Stack area (including shadow stacks)
569 *
570 * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
571 * do_mmap() forbids all other combinations.
572 */
573static inline bool is_stack_mapping(vm_flags_t flags)
574{
575 return ((flags & VM_STACK) == VM_STACK) || (flags & VM_SHADOW_STACK);
576}
577
578/*
579 * Data area - private, writable, not stack
580 */
581static inline bool is_data_mapping(vm_flags_t flags)
582{
583 return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE;
584}
585
586/* mm/util.c */
587struct anon_vma *folio_anon_vma(struct folio *folio);
588
589#ifdef CONFIG_MMU
590void unmap_mapping_folio(struct folio *folio);
591extern long populate_vma_page_range(struct vm_area_struct *vma,
592 unsigned long start, unsigned long end, int *locked);
593extern long faultin_vma_page_range(struct vm_area_struct *vma,
594 unsigned long start, unsigned long end,
595 bool write, int *locked);
596extern bool mlock_future_ok(struct mm_struct *mm, unsigned long flags,
597 unsigned long bytes);
598
599/*
600 * NOTE: This function can't tell whether the folio is "fully mapped" in the
601 * range.
602 * "fully mapped" means all the pages of folio is associated with the page
603 * table of range while this function just check whether the folio range is
604 * within the range [start, end). Function caller needs to do page table
605 * check if it cares about the page table association.
606 *
607 * Typical usage (like mlock or madvise) is:
608 * Caller knows at least 1 page of folio is associated with page table of VMA
609 * and the range [start, end) is intersect with the VMA range. Caller wants
610 * to know whether the folio is fully associated with the range. It calls
611 * this function to check whether the folio is in the range first. Then checks
612 * the page table to know whether the folio is fully mapped to the range.
613 */
614static inline bool
615folio_within_range(struct folio *folio, struct vm_area_struct *vma,
616 unsigned long start, unsigned long end)
617{
618 pgoff_t pgoff, addr;
619 unsigned long vma_pglen = vma_pages(vma);
620
621 VM_WARN_ON_FOLIO(folio_test_ksm(folio), folio);
622 if (start > end)
623 return false;
624
625 if (start < vma->vm_start)
626 start = vma->vm_start;
627
628 if (end > vma->vm_end)
629 end = vma->vm_end;
630
631 pgoff = folio_pgoff(folio);
632
633 /* if folio start address is not in vma range */
634 if (!in_range(pgoff, vma->vm_pgoff, vma_pglen))
635 return false;
636
637 addr = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
638
639 return !(addr < start || end - addr < folio_size(folio));
640}
641
642static inline bool
643folio_within_vma(struct folio *folio, struct vm_area_struct *vma)
644{
645 return folio_within_range(folio, vma, vma->vm_start, vma->vm_end);
646}
647
648/*
649 * mlock_vma_folio() and munlock_vma_folio():
650 * should be called with vma's mmap_lock held for read or write,
651 * under page table lock for the pte/pmd being added or removed.
652 *
653 * mlock is usually called at the end of folio_add_*_rmap_*(), munlock at
654 * the end of folio_remove_rmap_*(); but new anon folios are managed by
655 * folio_add_lru_vma() calling mlock_new_folio().
656 */
657void mlock_folio(struct folio *folio);
658static inline void mlock_vma_folio(struct folio *folio,
659 struct vm_area_struct *vma)
660{
661 /*
662 * The VM_SPECIAL check here serves two purposes.
663 * 1) VM_IO check prevents migration from double-counting during mlock.
664 * 2) Although mmap_region() and mlock_fixup() take care that VM_LOCKED
665 * is never left set on a VM_SPECIAL vma, there is an interval while
666 * file->f_op->mmap() is using vm_insert_page(s), when VM_LOCKED may
667 * still be set while VM_SPECIAL bits are added: so ignore it then.
668 */
669 if (unlikely((vma->vm_flags & (VM_LOCKED|VM_SPECIAL)) == VM_LOCKED))
670 mlock_folio(folio);
671}
672
673void munlock_folio(struct folio *folio);
674static inline void munlock_vma_folio(struct folio *folio,
675 struct vm_area_struct *vma)
676{
677 /*
678 * munlock if the function is called. Ideally, we should only
679 * do munlock if any page of folio is unmapped from VMA and
680 * cause folio not fully mapped to VMA.
681 *
682 * But it's not easy to confirm that's the situation. So we
683 * always munlock the folio and page reclaim will correct it
684 * if it's wrong.
685 */
686 if (unlikely(vma->vm_flags & VM_LOCKED))
687 munlock_folio(folio);
688}
689
690void mlock_new_folio(struct folio *folio);
691bool need_mlock_drain(int cpu);
692void mlock_drain_local(void);
693void mlock_drain_remote(int cpu);
694
695extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
696
697/*
698 * Return the start of user virtual address at the specific offset within
699 * a vma.
700 */
701static inline unsigned long
702vma_pgoff_address(pgoff_t pgoff, unsigned long nr_pages,
703 struct vm_area_struct *vma)
704{
705 unsigned long address;
706
707 if (pgoff >= vma->vm_pgoff) {
708 address = vma->vm_start +
709 ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
710 /* Check for address beyond vma (or wrapped through 0?) */
711 if (address < vma->vm_start || address >= vma->vm_end)
712 address = -EFAULT;
713 } else if (pgoff + nr_pages - 1 >= vma->vm_pgoff) {
714 /* Test above avoids possibility of wrap to 0 on 32-bit */
715 address = vma->vm_start;
716 } else {
717 address = -EFAULT;
718 }
719 return address;
720}
721
722/*
723 * Return the start of user virtual address of a page within a vma.
724 * Returns -EFAULT if all of the page is outside the range of vma.
725 * If page is a compound head, the entire compound page is considered.
726 */
727static inline unsigned long
728vma_address(struct page *page, struct vm_area_struct *vma)
729{
730 VM_BUG_ON_PAGE(PageKsm(page), page); /* KSM page->index unusable */
731 return vma_pgoff_address(page_to_pgoff(page), compound_nr(page), vma);
732}
733
734/*
735 * Then at what user virtual address will none of the range be found in vma?
736 * Assumes that vma_address() already returned a good starting address.
737 */
738static inline unsigned long vma_address_end(struct page_vma_mapped_walk *pvmw)
739{
740 struct vm_area_struct *vma = pvmw->vma;
741 pgoff_t pgoff;
742 unsigned long address;
743
744 /* Common case, plus ->pgoff is invalid for KSM */
745 if (pvmw->nr_pages == 1)
746 return pvmw->address + PAGE_SIZE;
747
748 pgoff = pvmw->pgoff + pvmw->nr_pages;
749 address = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
750 /* Check for address beyond vma (or wrapped through 0?) */
751 if (address < vma->vm_start || address > vma->vm_end)
752 address = vma->vm_end;
753 return address;
754}
755
756static inline struct file *maybe_unlock_mmap_for_io(struct vm_fault *vmf,
757 struct file *fpin)
758{
759 int flags = vmf->flags;
760
761 if (fpin)
762 return fpin;
763
764 /*
765 * FAULT_FLAG_RETRY_NOWAIT means we don't want to wait on page locks or
766 * anything, so we only pin the file and drop the mmap_lock if only
767 * FAULT_FLAG_ALLOW_RETRY is set, while this is the first attempt.
768 */
769 if (fault_flag_allow_retry_first(flags) &&
770 !(flags & FAULT_FLAG_RETRY_NOWAIT)) {
771 fpin = get_file(vmf->vma->vm_file);
772 release_fault_lock(vmf);
773 }
774 return fpin;
775}
776#else /* !CONFIG_MMU */
777static inline void unmap_mapping_folio(struct folio *folio) { }
778static inline void mlock_new_folio(struct folio *folio) { }
779static inline bool need_mlock_drain(int cpu) { return false; }
780static inline void mlock_drain_local(void) { }
781static inline void mlock_drain_remote(int cpu) { }
782static inline void vunmap_range_noflush(unsigned long start, unsigned long end)
783{
784}
785#endif /* !CONFIG_MMU */
786
787/* Memory initialisation debug and verification */
788#ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
789DECLARE_STATIC_KEY_TRUE(deferred_pages);
790
791bool __init deferred_grow_zone(struct zone *zone, unsigned int order);
792#endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */
793
794enum mminit_level {
795 MMINIT_WARNING,
796 MMINIT_VERIFY,
797 MMINIT_TRACE
798};
799
800#ifdef CONFIG_DEBUG_MEMORY_INIT
801
802extern int mminit_loglevel;
803
804#define mminit_dprintk(level, prefix, fmt, arg...) \
805do { \
806 if (level < mminit_loglevel) { \
807 if (level <= MMINIT_WARNING) \
808 pr_warn("mminit::" prefix " " fmt, ##arg); \
809 else \
810 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
811 } \
812} while (0)
813
814extern void mminit_verify_pageflags_layout(void);
815extern void mminit_verify_zonelist(void);
816#else
817
818static inline void mminit_dprintk(enum mminit_level level,
819 const char *prefix, const char *fmt, ...)
820{
821}
822
823static inline void mminit_verify_pageflags_layout(void)
824{
825}
826
827static inline void mminit_verify_zonelist(void)
828{
829}
830#endif /* CONFIG_DEBUG_MEMORY_INIT */
831
832#define NODE_RECLAIM_NOSCAN -2
833#define NODE_RECLAIM_FULL -1
834#define NODE_RECLAIM_SOME 0
835#define NODE_RECLAIM_SUCCESS 1
836
837#ifdef CONFIG_NUMA
838extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int);
839extern int find_next_best_node(int node, nodemask_t *used_node_mask);
840#else
841static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask,
842 unsigned int order)
843{
844 return NODE_RECLAIM_NOSCAN;
845}
846static inline int find_next_best_node(int node, nodemask_t *used_node_mask)
847{
848 return NUMA_NO_NODE;
849}
850#endif
851
852/*
853 * mm/memory-failure.c
854 */
855extern int hwpoison_filter(struct page *p);
856
857extern u32 hwpoison_filter_dev_major;
858extern u32 hwpoison_filter_dev_minor;
859extern u64 hwpoison_filter_flags_mask;
860extern u64 hwpoison_filter_flags_value;
861extern u64 hwpoison_filter_memcg;
862extern u32 hwpoison_filter_enable;
863
864extern unsigned long __must_check vm_mmap_pgoff(struct file *, unsigned long,
865 unsigned long, unsigned long,
866 unsigned long, unsigned long);
867
868extern void set_pageblock_order(void);
869unsigned long reclaim_pages(struct list_head *folio_list);
870unsigned int reclaim_clean_pages_from_list(struct zone *zone,
871 struct list_head *folio_list);
872/* The ALLOC_WMARK bits are used as an index to zone->watermark */
873#define ALLOC_WMARK_MIN WMARK_MIN
874#define ALLOC_WMARK_LOW WMARK_LOW
875#define ALLOC_WMARK_HIGH WMARK_HIGH
876#define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
877
878/* Mask to get the watermark bits */
879#define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
880
881/*
882 * Only MMU archs have async oom victim reclaim - aka oom_reaper so we
883 * cannot assume a reduced access to memory reserves is sufficient for
884 * !MMU
885 */
886#ifdef CONFIG_MMU
887#define ALLOC_OOM 0x08
888#else
889#define ALLOC_OOM ALLOC_NO_WATERMARKS
890#endif
891
892#define ALLOC_NON_BLOCK 0x10 /* Caller cannot block. Allow access
893 * to 25% of the min watermark or
894 * 62.5% if __GFP_HIGH is set.
895 */
896#define ALLOC_MIN_RESERVE 0x20 /* __GFP_HIGH set. Allow access to 50%
897 * of the min watermark.
898 */
899#define ALLOC_CPUSET 0x40 /* check for correct cpuset */
900#define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
901#ifdef CONFIG_ZONE_DMA32
902#define ALLOC_NOFRAGMENT 0x100 /* avoid mixing pageblock types */
903#else
904#define ALLOC_NOFRAGMENT 0x0
905#endif
906#define ALLOC_HIGHATOMIC 0x200 /* Allows access to MIGRATE_HIGHATOMIC */
907#define ALLOC_KSWAPD 0x800 /* allow waking of kswapd, __GFP_KSWAPD_RECLAIM set */
908
909/* Flags that allow allocations below the min watermark. */
910#define ALLOC_RESERVES (ALLOC_NON_BLOCK|ALLOC_MIN_RESERVE|ALLOC_HIGHATOMIC|ALLOC_OOM)
911
912enum ttu_flags;
913struct tlbflush_unmap_batch;
914
915
916/*
917 * only for MM internal work items which do not depend on
918 * any allocations or locks which might depend on allocations
919 */
920extern struct workqueue_struct *mm_percpu_wq;
921
922#ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
923void try_to_unmap_flush(void);
924void try_to_unmap_flush_dirty(void);
925void flush_tlb_batched_pending(struct mm_struct *mm);
926#else
927static inline void try_to_unmap_flush(void)
928{
929}
930static inline void try_to_unmap_flush_dirty(void)
931{
932}
933static inline void flush_tlb_batched_pending(struct mm_struct *mm)
934{
935}
936#endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
937
938extern const struct trace_print_flags pageflag_names[];
939extern const struct trace_print_flags pagetype_names[];
940extern const struct trace_print_flags vmaflag_names[];
941extern const struct trace_print_flags gfpflag_names[];
942
943static inline bool is_migrate_highatomic(enum migratetype migratetype)
944{
945 return migratetype == MIGRATE_HIGHATOMIC;
946}
947
948static inline bool is_migrate_highatomic_page(struct page *page)
949{
950 return get_pageblock_migratetype(page) == MIGRATE_HIGHATOMIC;
951}
952
953void setup_zone_pageset(struct zone *zone);
954
955struct migration_target_control {
956 int nid; /* preferred node id */
957 nodemask_t *nmask;
958 gfp_t gfp_mask;
959};
960
961/*
962 * mm/filemap.c
963 */
964size_t splice_folio_into_pipe(struct pipe_inode_info *pipe,
965 struct folio *folio, loff_t fpos, size_t size);
966
967/*
968 * mm/vmalloc.c
969 */
970#ifdef CONFIG_MMU
971void __init vmalloc_init(void);
972int __must_check vmap_pages_range_noflush(unsigned long addr, unsigned long end,
973 pgprot_t prot, struct page **pages, unsigned int page_shift);
974#else
975static inline void vmalloc_init(void)
976{
977}
978
979static inline
980int __must_check vmap_pages_range_noflush(unsigned long addr, unsigned long end,
981 pgprot_t prot, struct page **pages, unsigned int page_shift)
982{
983 return -EINVAL;
984}
985#endif
986
987int __must_check __vmap_pages_range_noflush(unsigned long addr,
988 unsigned long end, pgprot_t prot,
989 struct page **pages, unsigned int page_shift);
990
991void vunmap_range_noflush(unsigned long start, unsigned long end);
992
993void __vunmap_range_noflush(unsigned long start, unsigned long end);
994
995int numa_migrate_prep(struct folio *folio, struct vm_area_struct *vma,
996 unsigned long addr, int page_nid, int *flags);
997
998void free_zone_device_page(struct page *page);
999int migrate_device_coherent_page(struct page *page);
1000
1001/*
1002 * mm/gup.c
1003 */
1004struct folio *try_grab_folio(struct page *page, int refs, unsigned int flags);
1005int __must_check try_grab_page(struct page *page, unsigned int flags);
1006
1007/*
1008 * mm/huge_memory.c
1009 */
1010struct page *follow_trans_huge_pmd(struct vm_area_struct *vma,
1011 unsigned long addr, pmd_t *pmd,
1012 unsigned int flags);
1013
1014/*
1015 * mm/mmap.c
1016 */
1017struct vm_area_struct *vma_merge_extend(struct vma_iterator *vmi,
1018 struct vm_area_struct *vma,
1019 unsigned long delta);
1020
1021enum {
1022 /* mark page accessed */
1023 FOLL_TOUCH = 1 << 16,
1024 /* a retry, previous pass started an IO */
1025 FOLL_TRIED = 1 << 17,
1026 /* we are working on non-current tsk/mm */
1027 FOLL_REMOTE = 1 << 18,
1028 /* pages must be released via unpin_user_page */
1029 FOLL_PIN = 1 << 19,
1030 /* gup_fast: prevent fall-back to slow gup */
1031 FOLL_FAST_ONLY = 1 << 20,
1032 /* allow unlocking the mmap lock */
1033 FOLL_UNLOCKABLE = 1 << 21,
1034};
1035
1036#define INTERNAL_GUP_FLAGS (FOLL_TOUCH | FOLL_TRIED | FOLL_REMOTE | FOLL_PIN | \
1037 FOLL_FAST_ONLY | FOLL_UNLOCKABLE)
1038
1039/*
1040 * Indicates for which pages that are write-protected in the page table,
1041 * whether GUP has to trigger unsharing via FAULT_FLAG_UNSHARE such that the
1042 * GUP pin will remain consistent with the pages mapped into the page tables
1043 * of the MM.
1044 *
1045 * Temporary unmapping of PageAnonExclusive() pages or clearing of
1046 * PageAnonExclusive() has to protect against concurrent GUP:
1047 * * Ordinary GUP: Using the PT lock
1048 * * GUP-fast and fork(): mm->write_protect_seq
1049 * * GUP-fast and KSM or temporary unmapping (swap, migration): see
1050 * folio_try_share_anon_rmap_*()
1051 *
1052 * Must be called with the (sub)page that's actually referenced via the
1053 * page table entry, which might not necessarily be the head page for a
1054 * PTE-mapped THP.
1055 *
1056 * If the vma is NULL, we're coming from the GUP-fast path and might have
1057 * to fallback to the slow path just to lookup the vma.
1058 */
1059static inline bool gup_must_unshare(struct vm_area_struct *vma,
1060 unsigned int flags, struct page *page)
1061{
1062 /*
1063 * FOLL_WRITE is implicitly handled correctly as the page table entry
1064 * has to be writable -- and if it references (part of) an anonymous
1065 * folio, that part is required to be marked exclusive.
1066 */
1067 if ((flags & (FOLL_WRITE | FOLL_PIN)) != FOLL_PIN)
1068 return false;
1069 /*
1070 * Note: PageAnon(page) is stable until the page is actually getting
1071 * freed.
1072 */
1073 if (!PageAnon(page)) {
1074 /*
1075 * We only care about R/O long-term pining: R/O short-term
1076 * pinning does not have the semantics to observe successive
1077 * changes through the process page tables.
1078 */
1079 if (!(flags & FOLL_LONGTERM))
1080 return false;
1081
1082 /* We really need the vma ... */
1083 if (!vma)
1084 return true;
1085
1086 /*
1087 * ... because we only care about writable private ("COW")
1088 * mappings where we have to break COW early.
1089 */
1090 return is_cow_mapping(vma->vm_flags);
1091 }
1092
1093 /* Paired with a memory barrier in folio_try_share_anon_rmap_*(). */
1094 if (IS_ENABLED(CONFIG_HAVE_FAST_GUP))
1095 smp_rmb();
1096
1097 /*
1098 * During GUP-fast we might not get called on the head page for a
1099 * hugetlb page that is mapped using cont-PTE, because GUP-fast does
1100 * not work with the abstracted hugetlb PTEs that always point at the
1101 * head page. For hugetlb, PageAnonExclusive only applies on the head
1102 * page (as it cannot be partially COW-shared), so lookup the head page.
1103 */
1104 if (unlikely(!PageHead(page) && PageHuge(page)))
1105 page = compound_head(page);
1106
1107 /*
1108 * Note that PageKsm() pages cannot be exclusive, and consequently,
1109 * cannot get pinned.
1110 */
1111 return !PageAnonExclusive(page);
1112}
1113
1114extern bool mirrored_kernelcore;
1115extern bool memblock_has_mirror(void);
1116
1117static inline bool vma_soft_dirty_enabled(struct vm_area_struct *vma)
1118{
1119 /*
1120 * NOTE: we must check this before VM_SOFTDIRTY on soft-dirty
1121 * enablements, because when without soft-dirty being compiled in,
1122 * VM_SOFTDIRTY is defined as 0x0, then !(vm_flags & VM_SOFTDIRTY)
1123 * will be constantly true.
1124 */
1125 if (!IS_ENABLED(CONFIG_MEM_SOFT_DIRTY))
1126 return false;
1127
1128 /*
1129 * Soft-dirty is kind of special: its tracking is enabled when the
1130 * vma flags not set.
1131 */
1132 return !(vma->vm_flags & VM_SOFTDIRTY);
1133}
1134
1135static inline void vma_iter_config(struct vma_iterator *vmi,
1136 unsigned long index, unsigned long last)
1137{
1138 __mas_set_range(&vmi->mas, index, last - 1);
1139}
1140
1141/*
1142 * VMA Iterator functions shared between nommu and mmap
1143 */
1144static inline int vma_iter_prealloc(struct vma_iterator *vmi,
1145 struct vm_area_struct *vma)
1146{
1147 return mas_preallocate(&vmi->mas, vma, GFP_KERNEL);
1148}
1149
1150static inline void vma_iter_clear(struct vma_iterator *vmi)
1151{
1152 mas_store_prealloc(&vmi->mas, NULL);
1153}
1154
1155static inline struct vm_area_struct *vma_iter_load(struct vma_iterator *vmi)
1156{
1157 return mas_walk(&vmi->mas);
1158}
1159
1160/* Store a VMA with preallocated memory */
1161static inline void vma_iter_store(struct vma_iterator *vmi,
1162 struct vm_area_struct *vma)
1163{
1164
1165#if defined(CONFIG_DEBUG_VM_MAPLE_TREE)
1166 if (MAS_WARN_ON(&vmi->mas, vmi->mas.status != ma_start &&
1167 vmi->mas.index > vma->vm_start)) {
1168 pr_warn("%lx > %lx\n store vma %lx-%lx\n into slot %lx-%lx\n",
1169 vmi->mas.index, vma->vm_start, vma->vm_start,
1170 vma->vm_end, vmi->mas.index, vmi->mas.last);
1171 }
1172 if (MAS_WARN_ON(&vmi->mas, vmi->mas.status != ma_start &&
1173 vmi->mas.last < vma->vm_start)) {
1174 pr_warn("%lx < %lx\nstore vma %lx-%lx\ninto slot %lx-%lx\n",
1175 vmi->mas.last, vma->vm_start, vma->vm_start, vma->vm_end,
1176 vmi->mas.index, vmi->mas.last);
1177 }
1178#endif
1179
1180 if (vmi->mas.status != ma_start &&
1181 ((vmi->mas.index > vma->vm_start) || (vmi->mas.last < vma->vm_start)))
1182 vma_iter_invalidate(vmi);
1183
1184 __mas_set_range(&vmi->mas, vma->vm_start, vma->vm_end - 1);
1185 mas_store_prealloc(&vmi->mas, vma);
1186}
1187
1188static inline int vma_iter_store_gfp(struct vma_iterator *vmi,
1189 struct vm_area_struct *vma, gfp_t gfp)
1190{
1191 if (vmi->mas.status != ma_start &&
1192 ((vmi->mas.index > vma->vm_start) || (vmi->mas.last < vma->vm_start)))
1193 vma_iter_invalidate(vmi);
1194
1195 __mas_set_range(&vmi->mas, vma->vm_start, vma->vm_end - 1);
1196 mas_store_gfp(&vmi->mas, vma, gfp);
1197 if (unlikely(mas_is_err(&vmi->mas)))
1198 return -ENOMEM;
1199
1200 return 0;
1201}
1202
1203/*
1204 * VMA lock generalization
1205 */
1206struct vma_prepare {
1207 struct vm_area_struct *vma;
1208 struct vm_area_struct *adj_next;
1209 struct file *file;
1210 struct address_space *mapping;
1211 struct anon_vma *anon_vma;
1212 struct vm_area_struct *insert;
1213 struct vm_area_struct *remove;
1214 struct vm_area_struct *remove2;
1215};
1216
1217void __meminit __init_single_page(struct page *page, unsigned long pfn,
1218 unsigned long zone, int nid);
1219
1220/* shrinker related functions */
1221unsigned long shrink_slab(gfp_t gfp_mask, int nid, struct mem_cgroup *memcg,
1222 int priority);
1223
1224#ifdef CONFIG_SHRINKER_DEBUG
1225static inline __printf(2, 0) int shrinker_debugfs_name_alloc(
1226 struct shrinker *shrinker, const char *fmt, va_list ap)
1227{
1228 shrinker->name = kvasprintf_const(GFP_KERNEL, fmt, ap);
1229
1230 return shrinker->name ? 0 : -ENOMEM;
1231}
1232
1233static inline void shrinker_debugfs_name_free(struct shrinker *shrinker)
1234{
1235 kfree_const(shrinker->name);
1236 shrinker->name = NULL;
1237}
1238
1239extern int shrinker_debugfs_add(struct shrinker *shrinker);
1240extern struct dentry *shrinker_debugfs_detach(struct shrinker *shrinker,
1241 int *debugfs_id);
1242extern void shrinker_debugfs_remove(struct dentry *debugfs_entry,
1243 int debugfs_id);
1244#else /* CONFIG_SHRINKER_DEBUG */
1245static inline int shrinker_debugfs_add(struct shrinker *shrinker)
1246{
1247 return 0;
1248}
1249static inline int shrinker_debugfs_name_alloc(struct shrinker *shrinker,
1250 const char *fmt, va_list ap)
1251{
1252 return 0;
1253}
1254static inline void shrinker_debugfs_name_free(struct shrinker *shrinker)
1255{
1256}
1257static inline struct dentry *shrinker_debugfs_detach(struct shrinker *shrinker,
1258 int *debugfs_id)
1259{
1260 *debugfs_id = -1;
1261 return NULL;
1262}
1263static inline void shrinker_debugfs_remove(struct dentry *debugfs_entry,
1264 int debugfs_id)
1265{
1266}
1267#endif /* CONFIG_SHRINKER_DEBUG */
1268
1269#endif /* __MM_INTERNAL_H */