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/mm.h>
15
16void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
17 unsigned long floor, unsigned long ceiling);
18
19static inline void set_page_count(struct page *page, int v)
20{
21 atomic_set(&page->_count, v);
22}
23
24/*
25 * Turn a non-refcounted page (->_count == 0) into refcounted with
26 * a count of one.
27 */
28static inline void set_page_refcounted(struct page *page)
29{
30 VM_BUG_ON(PageTail(page));
31 VM_BUG_ON(atomic_read(&page->_count));
32 set_page_count(page, 1);
33}
34
35static inline void __put_page(struct page *page)
36{
37 atomic_dec(&page->_count);
38}
39
40extern unsigned long highest_memmap_pfn;
41
42/*
43 * in mm/vmscan.c:
44 */
45extern int isolate_lru_page(struct page *page);
46extern void putback_lru_page(struct page *page);
47
48/*
49 * in mm/page_alloc.c
50 */
51extern void __free_pages_bootmem(struct page *page, unsigned int order);
52extern void prep_compound_page(struct page *page, unsigned long order);
53#ifdef CONFIG_MEMORY_FAILURE
54extern bool is_free_buddy_page(struct page *page);
55#endif
56
57
58/*
59 * function for dealing with page's order in buddy system.
60 * zone->lock is already acquired when we use these.
61 * So, we don't need atomic page->flags operations here.
62 */
63static inline unsigned long page_order(struct page *page)
64{
65 /* PageBuddy() must be checked by the caller */
66 return page_private(page);
67}
68
69/* mm/util.c */
70void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
71 struct vm_area_struct *prev, struct rb_node *rb_parent);
72
73#ifdef CONFIG_MMU
74extern long mlock_vma_pages_range(struct vm_area_struct *vma,
75 unsigned long start, unsigned long end);
76extern void munlock_vma_pages_range(struct vm_area_struct *vma,
77 unsigned long start, unsigned long end);
78static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
79{
80 munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
81}
82
83/*
84 * Called only in fault path via page_evictable() for a new page
85 * to determine if it's being mapped into a LOCKED vma.
86 * If so, mark page as mlocked.
87 */
88static inline int is_mlocked_vma(struct vm_area_struct *vma, struct page *page)
89{
90 VM_BUG_ON(PageLRU(page));
91
92 if (likely((vma->vm_flags & (VM_LOCKED | VM_SPECIAL)) != VM_LOCKED))
93 return 0;
94
95 if (!TestSetPageMlocked(page)) {
96 inc_zone_page_state(page, NR_MLOCK);
97 count_vm_event(UNEVICTABLE_PGMLOCKED);
98 }
99 return 1;
100}
101
102/*
103 * must be called with vma's mmap_sem held for read or write, and page locked.
104 */
105extern void mlock_vma_page(struct page *page);
106extern void munlock_vma_page(struct page *page);
107
108/*
109 * Clear the page's PageMlocked(). This can be useful in a situation where
110 * we want to unconditionally remove a page from the pagecache -- e.g.,
111 * on truncation or freeing.
112 *
113 * It is legal to call this function for any page, mlocked or not.
114 * If called for a page that is still mapped by mlocked vmas, all we do
115 * is revert to lazy LRU behaviour -- semantics are not broken.
116 */
117extern void __clear_page_mlock(struct page *page);
118static inline void clear_page_mlock(struct page *page)
119{
120 if (unlikely(TestClearPageMlocked(page)))
121 __clear_page_mlock(page);
122}
123
124/*
125 * mlock_migrate_page - called only from migrate_page_copy() to
126 * migrate the Mlocked page flag; update statistics.
127 */
128static inline void mlock_migrate_page(struct page *newpage, struct page *page)
129{
130 if (TestClearPageMlocked(page)) {
131 unsigned long flags;
132
133 local_irq_save(flags);
134 __dec_zone_page_state(page, NR_MLOCK);
135 SetPageMlocked(newpage);
136 __inc_zone_page_state(newpage, NR_MLOCK);
137 local_irq_restore(flags);
138 }
139}
140
141#ifdef CONFIG_TRANSPARENT_HUGEPAGE
142extern unsigned long vma_address(struct page *page,
143 struct vm_area_struct *vma);
144#endif
145#else /* !CONFIG_MMU */
146static inline int is_mlocked_vma(struct vm_area_struct *v, struct page *p)
147{
148 return 0;
149}
150static inline void clear_page_mlock(struct page *page) { }
151static inline void mlock_vma_page(struct page *page) { }
152static inline void mlock_migrate_page(struct page *new, struct page *old) { }
153
154#endif /* !CONFIG_MMU */
155
156/*
157 * Return the mem_map entry representing the 'offset' subpage within
158 * the maximally aligned gigantic page 'base'. Handle any discontiguity
159 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
160 */
161static inline struct page *mem_map_offset(struct page *base, int offset)
162{
163 if (unlikely(offset >= MAX_ORDER_NR_PAGES))
164 return pfn_to_page(page_to_pfn(base) + offset);
165 return base + offset;
166}
167
168/*
169 * Iterator over all subpages within the maximally aligned gigantic
170 * page 'base'. Handle any discontiguity in the mem_map.
171 */
172static inline struct page *mem_map_next(struct page *iter,
173 struct page *base, int offset)
174{
175 if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
176 unsigned long pfn = page_to_pfn(base) + offset;
177 if (!pfn_valid(pfn))
178 return NULL;
179 return pfn_to_page(pfn);
180 }
181 return iter + 1;
182}
183
184/*
185 * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
186 * so all functions starting at paging_init should be marked __init
187 * in those cases. SPARSEMEM, however, allows for memory hotplug,
188 * and alloc_bootmem_node is not used.
189 */
190#ifdef CONFIG_SPARSEMEM
191#define __paginginit __meminit
192#else
193#define __paginginit __init
194#endif
195
196/* Memory initialisation debug and verification */
197enum mminit_level {
198 MMINIT_WARNING,
199 MMINIT_VERIFY,
200 MMINIT_TRACE
201};
202
203#ifdef CONFIG_DEBUG_MEMORY_INIT
204
205extern int mminit_loglevel;
206
207#define mminit_dprintk(level, prefix, fmt, arg...) \
208do { \
209 if (level < mminit_loglevel) { \
210 printk(level <= MMINIT_WARNING ? KERN_WARNING : KERN_DEBUG); \
211 printk(KERN_CONT "mminit::" prefix " " fmt, ##arg); \
212 } \
213} while (0)
214
215extern void mminit_verify_pageflags_layout(void);
216extern void mminit_verify_page_links(struct page *page,
217 enum zone_type zone, unsigned long nid, unsigned long pfn);
218extern void mminit_verify_zonelist(void);
219
220#else
221
222static inline void mminit_dprintk(enum mminit_level level,
223 const char *prefix, const char *fmt, ...)
224{
225}
226
227static inline void mminit_verify_pageflags_layout(void)
228{
229}
230
231static inline void mminit_verify_page_links(struct page *page,
232 enum zone_type zone, unsigned long nid, unsigned long pfn)
233{
234}
235
236static inline void mminit_verify_zonelist(void)
237{
238}
239#endif /* CONFIG_DEBUG_MEMORY_INIT */
240
241/* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
242#if defined(CONFIG_SPARSEMEM)
243extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
244 unsigned long *end_pfn);
245#else
246static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
247 unsigned long *end_pfn)
248{
249}
250#endif /* CONFIG_SPARSEMEM */
251
252#define ZONE_RECLAIM_NOSCAN -2
253#define ZONE_RECLAIM_FULL -1
254#define ZONE_RECLAIM_SOME 0
255#define ZONE_RECLAIM_SUCCESS 1
256#endif
257
258extern int hwpoison_filter(struct page *p);
259
260extern u32 hwpoison_filter_dev_major;
261extern u32 hwpoison_filter_dev_minor;
262extern u64 hwpoison_filter_flags_mask;
263extern u64 hwpoison_filter_flags_value;
264extern u64 hwpoison_filter_memcg;
265extern u32 hwpoison_filter_enable;
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/khugepaged.h>
12#include <linux/mm.h>
13#include <linux/mm_inline.h>
14#include <linux/pagemap.h>
15#include <linux/pagewalk.h>
16#include <linux/rmap.h>
17#include <linux/swap.h>
18#include <linux/swapops.h>
19#include <linux/swap_cgroup.h>
20#include <linux/tracepoint-defs.h>
21
22/* Internal core VMA manipulation functions. */
23#include "vma.h"
24
25struct folio_batch;
26
27/*
28 * The set of flags that only affect watermark checking and reclaim
29 * behaviour. This is used by the MM to obey the caller constraints
30 * about IO, FS and watermark checking while ignoring placement
31 * hints such as HIGHMEM usage.
32 */
33#define GFP_RECLAIM_MASK (__GFP_RECLAIM|__GFP_HIGH|__GFP_IO|__GFP_FS|\
34 __GFP_NOWARN|__GFP_RETRY_MAYFAIL|__GFP_NOFAIL|\
35 __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC|\
36 __GFP_NOLOCKDEP)
37
38/* The GFP flags allowed during early boot */
39#define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_RECLAIM|__GFP_IO|__GFP_FS))
40
41/* Control allocation cpuset and node placement constraints */
42#define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE)
43
44/* Do not use these with a slab allocator */
45#define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
46
47/*
48 * Different from WARN_ON_ONCE(), no warning will be issued
49 * when we specify __GFP_NOWARN.
50 */
51#define WARN_ON_ONCE_GFP(cond, gfp) ({ \
52 static bool __section(".data..once") __warned; \
53 int __ret_warn_once = !!(cond); \
54 \
55 if (unlikely(!(gfp & __GFP_NOWARN) && __ret_warn_once && !__warned)) { \
56 __warned = true; \
57 WARN_ON(1); \
58 } \
59 unlikely(__ret_warn_once); \
60})
61
62void page_writeback_init(void);
63
64/*
65 * If a 16GB hugetlb folio were mapped by PTEs of all of its 4kB pages,
66 * its nr_pages_mapped would be 0x400000: choose the ENTIRELY_MAPPED bit
67 * above that range, instead of 2*(PMD_SIZE/PAGE_SIZE). Hugetlb currently
68 * leaves nr_pages_mapped at 0, but avoid surprise if it participates later.
69 */
70#define ENTIRELY_MAPPED 0x800000
71#define FOLIO_PAGES_MAPPED (ENTIRELY_MAPPED - 1)
72
73/*
74 * Flags passed to __show_mem() and show_free_areas() to suppress output in
75 * various contexts.
76 */
77#define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
78
79/*
80 * How many individual pages have an elevated _mapcount. Excludes
81 * the folio's entire_mapcount.
82 *
83 * Don't use this function outside of debugging code.
84 */
85static inline int folio_nr_pages_mapped(const struct folio *folio)
86{
87 return atomic_read(&folio->_nr_pages_mapped) & FOLIO_PAGES_MAPPED;
88}
89
90/*
91 * Retrieve the first entry of a folio based on a provided entry within the
92 * folio. We cannot rely on folio->swap as there is no guarantee that it has
93 * been initialized. Used for calling arch_swap_restore()
94 */
95static inline swp_entry_t folio_swap(swp_entry_t entry,
96 const struct folio *folio)
97{
98 swp_entry_t swap = {
99 .val = ALIGN_DOWN(entry.val, folio_nr_pages(folio)),
100 };
101
102 return swap;
103}
104
105static inline void *folio_raw_mapping(const struct folio *folio)
106{
107 unsigned long mapping = (unsigned long)folio->mapping;
108
109 return (void *)(mapping & ~PAGE_MAPPING_FLAGS);
110}
111
112/*
113 * This is a file-backed mapping, and is about to be memory mapped - invoke its
114 * mmap hook and safely handle error conditions. On error, VMA hooks will be
115 * mutated.
116 *
117 * @file: File which backs the mapping.
118 * @vma: VMA which we are mapping.
119 *
120 * Returns: 0 if success, error otherwise.
121 */
122static inline int mmap_file(struct file *file, struct vm_area_struct *vma)
123{
124 int err = call_mmap(file, vma);
125
126 if (likely(!err))
127 return 0;
128
129 /*
130 * OK, we tried to call the file hook for mmap(), but an error
131 * arose. The mapping is in an inconsistent state and we most not invoke
132 * any further hooks on it.
133 */
134 vma->vm_ops = &vma_dummy_vm_ops;
135
136 return err;
137}
138
139/*
140 * If the VMA has a close hook then close it, and since closing it might leave
141 * it in an inconsistent state which makes the use of any hooks suspect, clear
142 * them down by installing dummy empty hooks.
143 */
144static inline void vma_close(struct vm_area_struct *vma)
145{
146 if (vma->vm_ops && vma->vm_ops->close) {
147 vma->vm_ops->close(vma);
148
149 /*
150 * The mapping is in an inconsistent state, and no further hooks
151 * may be invoked upon it.
152 */
153 vma->vm_ops = &vma_dummy_vm_ops;
154 }
155}
156
157#ifdef CONFIG_MMU
158
159/* Flags for folio_pte_batch(). */
160typedef int __bitwise fpb_t;
161
162/* Compare PTEs after pte_mkclean(), ignoring the dirty bit. */
163#define FPB_IGNORE_DIRTY ((__force fpb_t)BIT(0))
164
165/* Compare PTEs after pte_clear_soft_dirty(), ignoring the soft-dirty bit. */
166#define FPB_IGNORE_SOFT_DIRTY ((__force fpb_t)BIT(1))
167
168static inline pte_t __pte_batch_clear_ignored(pte_t pte, fpb_t flags)
169{
170 if (flags & FPB_IGNORE_DIRTY)
171 pte = pte_mkclean(pte);
172 if (likely(flags & FPB_IGNORE_SOFT_DIRTY))
173 pte = pte_clear_soft_dirty(pte);
174 return pte_wrprotect(pte_mkold(pte));
175}
176
177/**
178 * folio_pte_batch - detect a PTE batch for a large folio
179 * @folio: The large folio to detect a PTE batch for.
180 * @addr: The user virtual address the first page is mapped at.
181 * @start_ptep: Page table pointer for the first entry.
182 * @pte: Page table entry for the first page.
183 * @max_nr: The maximum number of table entries to consider.
184 * @flags: Flags to modify the PTE batch semantics.
185 * @any_writable: Optional pointer to indicate whether any entry except the
186 * first one is writable.
187 * @any_young: Optional pointer to indicate whether any entry except the
188 * first one is young.
189 * @any_dirty: Optional pointer to indicate whether any entry except the
190 * first one is dirty.
191 *
192 * Detect a PTE batch: consecutive (present) PTEs that map consecutive
193 * pages of the same large folio.
194 *
195 * All PTEs inside a PTE batch have the same PTE bits set, excluding the PFN,
196 * the accessed bit, writable bit, dirty bit (with FPB_IGNORE_DIRTY) and
197 * soft-dirty bit (with FPB_IGNORE_SOFT_DIRTY).
198 *
199 * start_ptep must map any page of the folio. max_nr must be at least one and
200 * must be limited by the caller so scanning cannot exceed a single page table.
201 *
202 * Return: the number of table entries in the batch.
203 */
204static inline int folio_pte_batch(struct folio *folio, unsigned long addr,
205 pte_t *start_ptep, pte_t pte, int max_nr, fpb_t flags,
206 bool *any_writable, bool *any_young, bool *any_dirty)
207{
208 unsigned long folio_end_pfn = folio_pfn(folio) + folio_nr_pages(folio);
209 const pte_t *end_ptep = start_ptep + max_nr;
210 pte_t expected_pte, *ptep;
211 bool writable, young, dirty;
212 int nr;
213
214 if (any_writable)
215 *any_writable = false;
216 if (any_young)
217 *any_young = false;
218 if (any_dirty)
219 *any_dirty = false;
220
221 VM_WARN_ON_FOLIO(!pte_present(pte), folio);
222 VM_WARN_ON_FOLIO(!folio_test_large(folio) || max_nr < 1, folio);
223 VM_WARN_ON_FOLIO(page_folio(pfn_to_page(pte_pfn(pte))) != folio, folio);
224
225 nr = pte_batch_hint(start_ptep, pte);
226 expected_pte = __pte_batch_clear_ignored(pte_advance_pfn(pte, nr), flags);
227 ptep = start_ptep + nr;
228
229 while (ptep < end_ptep) {
230 pte = ptep_get(ptep);
231 if (any_writable)
232 writable = !!pte_write(pte);
233 if (any_young)
234 young = !!pte_young(pte);
235 if (any_dirty)
236 dirty = !!pte_dirty(pte);
237 pte = __pte_batch_clear_ignored(pte, flags);
238
239 if (!pte_same(pte, expected_pte))
240 break;
241
242 /*
243 * Stop immediately once we reached the end of the folio. In
244 * corner cases the next PFN might fall into a different
245 * folio.
246 */
247 if (pte_pfn(pte) >= folio_end_pfn)
248 break;
249
250 if (any_writable)
251 *any_writable |= writable;
252 if (any_young)
253 *any_young |= young;
254 if (any_dirty)
255 *any_dirty |= dirty;
256
257 nr = pte_batch_hint(ptep, pte);
258 expected_pte = pte_advance_pfn(expected_pte, nr);
259 ptep += nr;
260 }
261
262 return min(ptep - start_ptep, max_nr);
263}
264
265/**
266 * pte_move_swp_offset - Move the swap entry offset field of a swap pte
267 * forward or backward by delta
268 * @pte: The initial pte state; is_swap_pte(pte) must be true and
269 * non_swap_entry() must be false.
270 * @delta: The direction and the offset we are moving; forward if delta
271 * is positive; backward if delta is negative
272 *
273 * Moves the swap offset, while maintaining all other fields, including
274 * swap type, and any swp pte bits. The resulting pte is returned.
275 */
276static inline pte_t pte_move_swp_offset(pte_t pte, long delta)
277{
278 swp_entry_t entry = pte_to_swp_entry(pte);
279 pte_t new = __swp_entry_to_pte(__swp_entry(swp_type(entry),
280 (swp_offset(entry) + delta)));
281
282 if (pte_swp_soft_dirty(pte))
283 new = pte_swp_mksoft_dirty(new);
284 if (pte_swp_exclusive(pte))
285 new = pte_swp_mkexclusive(new);
286 if (pte_swp_uffd_wp(pte))
287 new = pte_swp_mkuffd_wp(new);
288
289 return new;
290}
291
292
293/**
294 * pte_next_swp_offset - Increment the swap entry offset field of a swap pte.
295 * @pte: The initial pte state; is_swap_pte(pte) must be true and
296 * non_swap_entry() must be false.
297 *
298 * Increments the swap offset, while maintaining all other fields, including
299 * swap type, and any swp pte bits. The resulting pte is returned.
300 */
301static inline pte_t pte_next_swp_offset(pte_t pte)
302{
303 return pte_move_swp_offset(pte, 1);
304}
305
306/**
307 * swap_pte_batch - detect a PTE batch for a set of contiguous swap entries
308 * @start_ptep: Page table pointer for the first entry.
309 * @max_nr: The maximum number of table entries to consider.
310 * @pte: Page table entry for the first entry.
311 *
312 * Detect a batch of contiguous swap entries: consecutive (non-present) PTEs
313 * containing swap entries all with consecutive offsets and targeting the same
314 * swap type, all with matching swp pte bits.
315 *
316 * max_nr must be at least one and must be limited by the caller so scanning
317 * cannot exceed a single page table.
318 *
319 * Return: the number of table entries in the batch.
320 */
321static inline int swap_pte_batch(pte_t *start_ptep, int max_nr, pte_t pte)
322{
323 pte_t expected_pte = pte_next_swp_offset(pte);
324 const pte_t *end_ptep = start_ptep + max_nr;
325 swp_entry_t entry = pte_to_swp_entry(pte);
326 pte_t *ptep = start_ptep + 1;
327 unsigned short cgroup_id;
328
329 VM_WARN_ON(max_nr < 1);
330 VM_WARN_ON(!is_swap_pte(pte));
331 VM_WARN_ON(non_swap_entry(entry));
332
333 cgroup_id = lookup_swap_cgroup_id(entry);
334 while (ptep < end_ptep) {
335 pte = ptep_get(ptep);
336
337 if (!pte_same(pte, expected_pte))
338 break;
339 if (lookup_swap_cgroup_id(pte_to_swp_entry(pte)) != cgroup_id)
340 break;
341 expected_pte = pte_next_swp_offset(expected_pte);
342 ptep++;
343 }
344
345 return ptep - start_ptep;
346}
347#endif /* CONFIG_MMU */
348
349void __acct_reclaim_writeback(pg_data_t *pgdat, struct folio *folio,
350 int nr_throttled);
351static inline void acct_reclaim_writeback(struct folio *folio)
352{
353 pg_data_t *pgdat = folio_pgdat(folio);
354 int nr_throttled = atomic_read(&pgdat->nr_writeback_throttled);
355
356 if (nr_throttled)
357 __acct_reclaim_writeback(pgdat, folio, nr_throttled);
358}
359
360static inline void wake_throttle_isolated(pg_data_t *pgdat)
361{
362 wait_queue_head_t *wqh;
363
364 wqh = &pgdat->reclaim_wait[VMSCAN_THROTTLE_ISOLATED];
365 if (waitqueue_active(wqh))
366 wake_up(wqh);
367}
368
369vm_fault_t __vmf_anon_prepare(struct vm_fault *vmf);
370static inline vm_fault_t vmf_anon_prepare(struct vm_fault *vmf)
371{
372 vm_fault_t ret = __vmf_anon_prepare(vmf);
373
374 if (unlikely(ret & VM_FAULT_RETRY))
375 vma_end_read(vmf->vma);
376 return ret;
377}
378
379vm_fault_t do_swap_page(struct vm_fault *vmf);
380void folio_rotate_reclaimable(struct folio *folio);
381bool __folio_end_writeback(struct folio *folio);
382void deactivate_file_folio(struct folio *folio);
383void folio_activate(struct folio *folio);
384
385void free_pgtables(struct mmu_gather *tlb, struct ma_state *mas,
386 struct vm_area_struct *start_vma, unsigned long floor,
387 unsigned long ceiling, bool mm_wr_locked);
388void pmd_install(struct mm_struct *mm, pmd_t *pmd, pgtable_t *pte);
389
390struct zap_details;
391void unmap_page_range(struct mmu_gather *tlb,
392 struct vm_area_struct *vma,
393 unsigned long addr, unsigned long end,
394 struct zap_details *details);
395
396void page_cache_ra_order(struct readahead_control *, struct file_ra_state *,
397 unsigned int order);
398void force_page_cache_ra(struct readahead_control *, unsigned long nr);
399static inline void force_page_cache_readahead(struct address_space *mapping,
400 struct file *file, pgoff_t index, unsigned long nr_to_read)
401{
402 DEFINE_READAHEAD(ractl, file, &file->f_ra, mapping, index);
403 force_page_cache_ra(&ractl, nr_to_read);
404}
405
406unsigned find_lock_entries(struct address_space *mapping, pgoff_t *start,
407 pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices);
408unsigned find_get_entries(struct address_space *mapping, pgoff_t *start,
409 pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices);
410void filemap_free_folio(struct address_space *mapping, struct folio *folio);
411int truncate_inode_folio(struct address_space *mapping, struct folio *folio);
412bool truncate_inode_partial_folio(struct folio *folio, loff_t start,
413 loff_t end);
414long mapping_evict_folio(struct address_space *mapping, struct folio *folio);
415unsigned long mapping_try_invalidate(struct address_space *mapping,
416 pgoff_t start, pgoff_t end, unsigned long *nr_failed);
417
418/**
419 * folio_evictable - Test whether a folio is evictable.
420 * @folio: The folio to test.
421 *
422 * Test whether @folio is evictable -- i.e., should be placed on
423 * active/inactive lists vs unevictable list.
424 *
425 * Reasons folio might not be evictable:
426 * 1. folio's mapping marked unevictable
427 * 2. One of the pages in the folio is part of an mlocked VMA
428 */
429static inline bool folio_evictable(struct folio *folio)
430{
431 bool ret;
432
433 /* Prevent address_space of inode and swap cache from being freed */
434 rcu_read_lock();
435 ret = !mapping_unevictable(folio_mapping(folio)) &&
436 !folio_test_mlocked(folio);
437 rcu_read_unlock();
438 return ret;
439}
440
441/*
442 * Turn a non-refcounted page (->_refcount == 0) into refcounted with
443 * a count of one.
444 */
445static inline void set_page_refcounted(struct page *page)
446{
447 VM_BUG_ON_PAGE(PageTail(page), page);
448 VM_BUG_ON_PAGE(page_ref_count(page), page);
449 set_page_count(page, 1);
450}
451
452/*
453 * Return true if a folio needs ->release_folio() calling upon it.
454 */
455static inline bool folio_needs_release(struct folio *folio)
456{
457 struct address_space *mapping = folio_mapping(folio);
458
459 return folio_has_private(folio) ||
460 (mapping && mapping_release_always(mapping));
461}
462
463extern unsigned long highest_memmap_pfn;
464
465/*
466 * Maximum number of reclaim retries without progress before the OOM
467 * killer is consider the only way forward.
468 */
469#define MAX_RECLAIM_RETRIES 16
470
471/*
472 * in mm/vmscan.c:
473 */
474bool folio_isolate_lru(struct folio *folio);
475void folio_putback_lru(struct folio *folio);
476extern void reclaim_throttle(pg_data_t *pgdat, enum vmscan_throttle_state reason);
477
478/*
479 * in mm/rmap.c:
480 */
481pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
482
483/*
484 * in mm/page_alloc.c
485 */
486#define K(x) ((x) << (PAGE_SHIFT-10))
487
488extern char * const zone_names[MAX_NR_ZONES];
489
490/* perform sanity checks on struct pages being allocated or freed */
491DECLARE_STATIC_KEY_MAYBE(CONFIG_DEBUG_VM, check_pages_enabled);
492
493extern int min_free_kbytes;
494
495void setup_per_zone_wmarks(void);
496void calculate_min_free_kbytes(void);
497int __meminit init_per_zone_wmark_min(void);
498void page_alloc_sysctl_init(void);
499
500/*
501 * Structure for holding the mostly immutable allocation parameters passed
502 * between functions involved in allocations, including the alloc_pages*
503 * family of functions.
504 *
505 * nodemask, migratetype and highest_zoneidx are initialized only once in
506 * __alloc_pages() and then never change.
507 *
508 * zonelist, preferred_zone and highest_zoneidx are set first in
509 * __alloc_pages() for the fast path, and might be later changed
510 * in __alloc_pages_slowpath(). All other functions pass the whole structure
511 * by a const pointer.
512 */
513struct alloc_context {
514 struct zonelist *zonelist;
515 nodemask_t *nodemask;
516 struct zoneref *preferred_zoneref;
517 int migratetype;
518
519 /*
520 * highest_zoneidx represents highest usable zone index of
521 * the allocation request. Due to the nature of the zone,
522 * memory on lower zone than the highest_zoneidx will be
523 * protected by lowmem_reserve[highest_zoneidx].
524 *
525 * highest_zoneidx is also used by reclaim/compaction to limit
526 * the target zone since higher zone than this index cannot be
527 * usable for this allocation request.
528 */
529 enum zone_type highest_zoneidx;
530 bool spread_dirty_pages;
531};
532
533/*
534 * This function returns the order of a free page in the buddy system. In
535 * general, page_zone(page)->lock must be held by the caller to prevent the
536 * page from being allocated in parallel and returning garbage as the order.
537 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
538 * page cannot be allocated or merged in parallel. Alternatively, it must
539 * handle invalid values gracefully, and use buddy_order_unsafe() below.
540 */
541static inline unsigned int buddy_order(struct page *page)
542{
543 /* PageBuddy() must be checked by the caller */
544 return page_private(page);
545}
546
547/*
548 * Like buddy_order(), but for callers who cannot afford to hold the zone lock.
549 * PageBuddy() should be checked first by the caller to minimize race window,
550 * and invalid values must be handled gracefully.
551 *
552 * READ_ONCE is used so that if the caller assigns the result into a local
553 * variable and e.g. tests it for valid range before using, the compiler cannot
554 * decide to remove the variable and inline the page_private(page) multiple
555 * times, potentially observing different values in the tests and the actual
556 * use of the result.
557 */
558#define buddy_order_unsafe(page) READ_ONCE(page_private(page))
559
560/*
561 * This function checks whether a page is free && is the buddy
562 * we can coalesce a page and its buddy if
563 * (a) the buddy is not in a hole (check before calling!) &&
564 * (b) the buddy is in the buddy system &&
565 * (c) a page and its buddy have the same order &&
566 * (d) a page and its buddy are in the same zone.
567 *
568 * For recording whether a page is in the buddy system, we set PageBuddy.
569 * Setting, clearing, and testing PageBuddy is serialized by zone->lock.
570 *
571 * For recording page's order, we use page_private(page).
572 */
573static inline bool page_is_buddy(struct page *page, struct page *buddy,
574 unsigned int order)
575{
576 if (!page_is_guard(buddy) && !PageBuddy(buddy))
577 return false;
578
579 if (buddy_order(buddy) != order)
580 return false;
581
582 /*
583 * zone check is done late to avoid uselessly calculating
584 * zone/node ids for pages that could never merge.
585 */
586 if (page_zone_id(page) != page_zone_id(buddy))
587 return false;
588
589 VM_BUG_ON_PAGE(page_count(buddy) != 0, buddy);
590
591 return true;
592}
593
594/*
595 * Locate the struct page for both the matching buddy in our
596 * pair (buddy1) and the combined O(n+1) page they form (page).
597 *
598 * 1) Any buddy B1 will have an order O twin B2 which satisfies
599 * the following equation:
600 * B2 = B1 ^ (1 << O)
601 * For example, if the starting buddy (buddy2) is #8 its order
602 * 1 buddy is #10:
603 * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
604 *
605 * 2) Any buddy B will have an order O+1 parent P which
606 * satisfies the following equation:
607 * P = B & ~(1 << O)
608 *
609 * Assumption: *_mem_map is contiguous at least up to MAX_PAGE_ORDER
610 */
611static inline unsigned long
612__find_buddy_pfn(unsigned long page_pfn, unsigned int order)
613{
614 return page_pfn ^ (1 << order);
615}
616
617/*
618 * Find the buddy of @page and validate it.
619 * @page: The input page
620 * @pfn: The pfn of the page, it saves a call to page_to_pfn() when the
621 * function is used in the performance-critical __free_one_page().
622 * @order: The order of the page
623 * @buddy_pfn: The output pointer to the buddy pfn, it also saves a call to
624 * page_to_pfn().
625 *
626 * The found buddy can be a non PageBuddy, out of @page's zone, or its order is
627 * not the same as @page. The validation is necessary before use it.
628 *
629 * Return: the found buddy page or NULL if not found.
630 */
631static inline struct page *find_buddy_page_pfn(struct page *page,
632 unsigned long pfn, unsigned int order, unsigned long *buddy_pfn)
633{
634 unsigned long __buddy_pfn = __find_buddy_pfn(pfn, order);
635 struct page *buddy;
636
637 buddy = page + (__buddy_pfn - pfn);
638 if (buddy_pfn)
639 *buddy_pfn = __buddy_pfn;
640
641 if (page_is_buddy(page, buddy, order))
642 return buddy;
643 return NULL;
644}
645
646extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn,
647 unsigned long end_pfn, struct zone *zone);
648
649static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn,
650 unsigned long end_pfn, struct zone *zone)
651{
652 if (zone->contiguous)
653 return pfn_to_page(start_pfn);
654
655 return __pageblock_pfn_to_page(start_pfn, end_pfn, zone);
656}
657
658void set_zone_contiguous(struct zone *zone);
659
660static inline void clear_zone_contiguous(struct zone *zone)
661{
662 zone->contiguous = false;
663}
664
665extern int __isolate_free_page(struct page *page, unsigned int order);
666extern void __putback_isolated_page(struct page *page, unsigned int order,
667 int mt);
668extern void memblock_free_pages(struct page *page, unsigned long pfn,
669 unsigned int order);
670extern void __free_pages_core(struct page *page, unsigned int order,
671 enum meminit_context context);
672
673/*
674 * This will have no effect, other than possibly generating a warning, if the
675 * caller passes in a non-large folio.
676 */
677static inline void folio_set_order(struct folio *folio, unsigned int order)
678{
679 if (WARN_ON_ONCE(!order || !folio_test_large(folio)))
680 return;
681
682 folio->_flags_1 = (folio->_flags_1 & ~0xffUL) | order;
683#ifdef CONFIG_64BIT
684 folio->_folio_nr_pages = 1U << order;
685#endif
686}
687
688bool __folio_unqueue_deferred_split(struct folio *folio);
689static inline bool folio_unqueue_deferred_split(struct folio *folio)
690{
691 if (folio_order(folio) <= 1 || !folio_test_large_rmappable(folio))
692 return false;
693
694 /*
695 * At this point, there is no one trying to add the folio to
696 * deferred_list. If folio is not in deferred_list, it's safe
697 * to check without acquiring the split_queue_lock.
698 */
699 if (data_race(list_empty(&folio->_deferred_list)))
700 return false;
701
702 return __folio_unqueue_deferred_split(folio);
703}
704
705static inline struct folio *page_rmappable_folio(struct page *page)
706{
707 struct folio *folio = (struct folio *)page;
708
709 if (folio && folio_test_large(folio))
710 folio_set_large_rmappable(folio);
711 return folio;
712}
713
714static inline void prep_compound_head(struct page *page, unsigned int order)
715{
716 struct folio *folio = (struct folio *)page;
717
718 folio_set_order(folio, order);
719 atomic_set(&folio->_large_mapcount, -1);
720 atomic_set(&folio->_entire_mapcount, -1);
721 atomic_set(&folio->_nr_pages_mapped, 0);
722 atomic_set(&folio->_pincount, 0);
723 if (order > 1)
724 INIT_LIST_HEAD(&folio->_deferred_list);
725}
726
727static inline void prep_compound_tail(struct page *head, int tail_idx)
728{
729 struct page *p = head + tail_idx;
730
731 p->mapping = TAIL_MAPPING;
732 set_compound_head(p, head);
733 set_page_private(p, 0);
734}
735
736extern void prep_compound_page(struct page *page, unsigned int order);
737
738extern void post_alloc_hook(struct page *page, unsigned int order,
739 gfp_t gfp_flags);
740extern bool free_pages_prepare(struct page *page, unsigned int order);
741
742extern int user_min_free_kbytes;
743
744void free_unref_page(struct page *page, unsigned int order);
745void free_unref_folios(struct folio_batch *fbatch);
746
747extern void zone_pcp_reset(struct zone *zone);
748extern void zone_pcp_disable(struct zone *zone);
749extern void zone_pcp_enable(struct zone *zone);
750extern void zone_pcp_init(struct zone *zone);
751
752extern void *memmap_alloc(phys_addr_t size, phys_addr_t align,
753 phys_addr_t min_addr,
754 int nid, bool exact_nid);
755
756void memmap_init_range(unsigned long, int, unsigned long, unsigned long,
757 unsigned long, enum meminit_context, struct vmem_altmap *, int);
758
759#if defined CONFIG_COMPACTION || defined CONFIG_CMA
760
761/*
762 * in mm/compaction.c
763 */
764/*
765 * compact_control is used to track pages being migrated and the free pages
766 * they are being migrated to during memory compaction. The free_pfn starts
767 * at the end of a zone and migrate_pfn begins at the start. Movable pages
768 * are moved to the end of a zone during a compaction run and the run
769 * completes when free_pfn <= migrate_pfn
770 */
771struct compact_control {
772 struct list_head freepages[NR_PAGE_ORDERS]; /* List of free pages to migrate to */
773 struct list_head migratepages; /* List of pages being migrated */
774 unsigned int nr_freepages; /* Number of isolated free pages */
775 unsigned int nr_migratepages; /* Number of pages to migrate */
776 unsigned long free_pfn; /* isolate_freepages search base */
777 /*
778 * Acts as an in/out parameter to page isolation for migration.
779 * isolate_migratepages uses it as a search base.
780 * isolate_migratepages_block will update the value to the next pfn
781 * after the last isolated one.
782 */
783 unsigned long migrate_pfn;
784 unsigned long fast_start_pfn; /* a pfn to start linear scan from */
785 struct zone *zone;
786 unsigned long total_migrate_scanned;
787 unsigned long total_free_scanned;
788 unsigned short fast_search_fail;/* failures to use free list searches */
789 short search_order; /* order to start a fast search at */
790 const gfp_t gfp_mask; /* gfp mask of a direct compactor */
791 int order; /* order a direct compactor needs */
792 int migratetype; /* migratetype of direct compactor */
793 const unsigned int alloc_flags; /* alloc flags of a direct compactor */
794 const int highest_zoneidx; /* zone index of a direct compactor */
795 enum migrate_mode mode; /* Async or sync migration mode */
796 bool ignore_skip_hint; /* Scan blocks even if marked skip */
797 bool no_set_skip_hint; /* Don't mark blocks for skipping */
798 bool ignore_block_suitable; /* Scan blocks considered unsuitable */
799 bool direct_compaction; /* False from kcompactd or /proc/... */
800 bool proactive_compaction; /* kcompactd proactive compaction */
801 bool whole_zone; /* Whole zone should/has been scanned */
802 bool contended; /* Signal lock contention */
803 bool finish_pageblock; /* Scan the remainder of a pageblock. Used
804 * when there are potentially transient
805 * isolation or migration failures to
806 * ensure forward progress.
807 */
808 bool alloc_contig; /* alloc_contig_range allocation */
809};
810
811/*
812 * Used in direct compaction when a page should be taken from the freelists
813 * immediately when one is created during the free path.
814 */
815struct capture_control {
816 struct compact_control *cc;
817 struct page *page;
818};
819
820unsigned long
821isolate_freepages_range(struct compact_control *cc,
822 unsigned long start_pfn, unsigned long end_pfn);
823int
824isolate_migratepages_range(struct compact_control *cc,
825 unsigned long low_pfn, unsigned long end_pfn);
826
827int __alloc_contig_migrate_range(struct compact_control *cc,
828 unsigned long start, unsigned long end,
829 int migratetype);
830
831/* Free whole pageblock and set its migration type to MIGRATE_CMA. */
832void init_cma_reserved_pageblock(struct page *page);
833
834#endif /* CONFIG_COMPACTION || CONFIG_CMA */
835
836int find_suitable_fallback(struct free_area *area, unsigned int order,
837 int migratetype, bool only_stealable, bool *can_steal);
838
839static inline bool free_area_empty(struct free_area *area, int migratetype)
840{
841 return list_empty(&area->free_list[migratetype]);
842}
843
844/* mm/util.c */
845struct anon_vma *folio_anon_vma(const struct folio *folio);
846
847#ifdef CONFIG_MMU
848void unmap_mapping_folio(struct folio *folio);
849extern long populate_vma_page_range(struct vm_area_struct *vma,
850 unsigned long start, unsigned long end, int *locked);
851extern long faultin_page_range(struct mm_struct *mm, unsigned long start,
852 unsigned long end, bool write, int *locked);
853extern bool mlock_future_ok(struct mm_struct *mm, unsigned long flags,
854 unsigned long bytes);
855
856/*
857 * NOTE: This function can't tell whether the folio is "fully mapped" in the
858 * range.
859 * "fully mapped" means all the pages of folio is associated with the page
860 * table of range while this function just check whether the folio range is
861 * within the range [start, end). Function caller needs to do page table
862 * check if it cares about the page table association.
863 *
864 * Typical usage (like mlock or madvise) is:
865 * Caller knows at least 1 page of folio is associated with page table of VMA
866 * and the range [start, end) is intersect with the VMA range. Caller wants
867 * to know whether the folio is fully associated with the range. It calls
868 * this function to check whether the folio is in the range first. Then checks
869 * the page table to know whether the folio is fully mapped to the range.
870 */
871static inline bool
872folio_within_range(struct folio *folio, struct vm_area_struct *vma,
873 unsigned long start, unsigned long end)
874{
875 pgoff_t pgoff, addr;
876 unsigned long vma_pglen = vma_pages(vma);
877
878 VM_WARN_ON_FOLIO(folio_test_ksm(folio), folio);
879 if (start > end)
880 return false;
881
882 if (start < vma->vm_start)
883 start = vma->vm_start;
884
885 if (end > vma->vm_end)
886 end = vma->vm_end;
887
888 pgoff = folio_pgoff(folio);
889
890 /* if folio start address is not in vma range */
891 if (!in_range(pgoff, vma->vm_pgoff, vma_pglen))
892 return false;
893
894 addr = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
895
896 return !(addr < start || end - addr < folio_size(folio));
897}
898
899static inline bool
900folio_within_vma(struct folio *folio, struct vm_area_struct *vma)
901{
902 return folio_within_range(folio, vma, vma->vm_start, vma->vm_end);
903}
904
905/*
906 * mlock_vma_folio() and munlock_vma_folio():
907 * should be called with vma's mmap_lock held for read or write,
908 * under page table lock for the pte/pmd being added or removed.
909 *
910 * mlock is usually called at the end of folio_add_*_rmap_*(), munlock at
911 * the end of folio_remove_rmap_*(); but new anon folios are managed by
912 * folio_add_lru_vma() calling mlock_new_folio().
913 */
914void mlock_folio(struct folio *folio);
915static inline void mlock_vma_folio(struct folio *folio,
916 struct vm_area_struct *vma)
917{
918 /*
919 * The VM_SPECIAL check here serves two purposes.
920 * 1) VM_IO check prevents migration from double-counting during mlock.
921 * 2) Although mmap_region() and mlock_fixup() take care that VM_LOCKED
922 * is never left set on a VM_SPECIAL vma, there is an interval while
923 * file->f_op->mmap() is using vm_insert_page(s), when VM_LOCKED may
924 * still be set while VM_SPECIAL bits are added: so ignore it then.
925 */
926 if (unlikely((vma->vm_flags & (VM_LOCKED|VM_SPECIAL)) == VM_LOCKED))
927 mlock_folio(folio);
928}
929
930void munlock_folio(struct folio *folio);
931static inline void munlock_vma_folio(struct folio *folio,
932 struct vm_area_struct *vma)
933{
934 /*
935 * munlock if the function is called. Ideally, we should only
936 * do munlock if any page of folio is unmapped from VMA and
937 * cause folio not fully mapped to VMA.
938 *
939 * But it's not easy to confirm that's the situation. So we
940 * always munlock the folio and page reclaim will correct it
941 * if it's wrong.
942 */
943 if (unlikely(vma->vm_flags & VM_LOCKED))
944 munlock_folio(folio);
945}
946
947void mlock_new_folio(struct folio *folio);
948bool need_mlock_drain(int cpu);
949void mlock_drain_local(void);
950void mlock_drain_remote(int cpu);
951
952extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
953
954/**
955 * vma_address - Find the virtual address a page range is mapped at
956 * @vma: The vma which maps this object.
957 * @pgoff: The page offset within its object.
958 * @nr_pages: The number of pages to consider.
959 *
960 * If any page in this range is mapped by this VMA, return the first address
961 * where any of these pages appear. Otherwise, return -EFAULT.
962 */
963static inline unsigned long vma_address(const struct vm_area_struct *vma,
964 pgoff_t pgoff, unsigned long nr_pages)
965{
966 unsigned long address;
967
968 if (pgoff >= vma->vm_pgoff) {
969 address = vma->vm_start +
970 ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
971 /* Check for address beyond vma (or wrapped through 0?) */
972 if (address < vma->vm_start || address >= vma->vm_end)
973 address = -EFAULT;
974 } else if (pgoff + nr_pages - 1 >= vma->vm_pgoff) {
975 /* Test above avoids possibility of wrap to 0 on 32-bit */
976 address = vma->vm_start;
977 } else {
978 address = -EFAULT;
979 }
980 return address;
981}
982
983/*
984 * Then at what user virtual address will none of the range be found in vma?
985 * Assumes that vma_address() already returned a good starting address.
986 */
987static inline unsigned long vma_address_end(struct page_vma_mapped_walk *pvmw)
988{
989 struct vm_area_struct *vma = pvmw->vma;
990 pgoff_t pgoff;
991 unsigned long address;
992
993 /* Common case, plus ->pgoff is invalid for KSM */
994 if (pvmw->nr_pages == 1)
995 return pvmw->address + PAGE_SIZE;
996
997 pgoff = pvmw->pgoff + pvmw->nr_pages;
998 address = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
999 /* Check for address beyond vma (or wrapped through 0?) */
1000 if (address < vma->vm_start || address > vma->vm_end)
1001 address = vma->vm_end;
1002 return address;
1003}
1004
1005static inline struct file *maybe_unlock_mmap_for_io(struct vm_fault *vmf,
1006 struct file *fpin)
1007{
1008 int flags = vmf->flags;
1009
1010 if (fpin)
1011 return fpin;
1012
1013 /*
1014 * FAULT_FLAG_RETRY_NOWAIT means we don't want to wait on page locks or
1015 * anything, so we only pin the file and drop the mmap_lock if only
1016 * FAULT_FLAG_ALLOW_RETRY is set, while this is the first attempt.
1017 */
1018 if (fault_flag_allow_retry_first(flags) &&
1019 !(flags & FAULT_FLAG_RETRY_NOWAIT)) {
1020 fpin = get_file(vmf->vma->vm_file);
1021 release_fault_lock(vmf);
1022 }
1023 return fpin;
1024}
1025#else /* !CONFIG_MMU */
1026static inline void unmap_mapping_folio(struct folio *folio) { }
1027static inline void mlock_new_folio(struct folio *folio) { }
1028static inline bool need_mlock_drain(int cpu) { return false; }
1029static inline void mlock_drain_local(void) { }
1030static inline void mlock_drain_remote(int cpu) { }
1031static inline void vunmap_range_noflush(unsigned long start, unsigned long end)
1032{
1033}
1034#endif /* !CONFIG_MMU */
1035
1036/* Memory initialisation debug and verification */
1037#ifdef CONFIG_DEFERRED_STRUCT_PAGE_INIT
1038DECLARE_STATIC_KEY_TRUE(deferred_pages);
1039
1040bool __init deferred_grow_zone(struct zone *zone, unsigned int order);
1041#endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */
1042
1043enum mminit_level {
1044 MMINIT_WARNING,
1045 MMINIT_VERIFY,
1046 MMINIT_TRACE
1047};
1048
1049#ifdef CONFIG_DEBUG_MEMORY_INIT
1050
1051extern int mminit_loglevel;
1052
1053#define mminit_dprintk(level, prefix, fmt, arg...) \
1054do { \
1055 if (level < mminit_loglevel) { \
1056 if (level <= MMINIT_WARNING) \
1057 pr_warn("mminit::" prefix " " fmt, ##arg); \
1058 else \
1059 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
1060 } \
1061} while (0)
1062
1063extern void mminit_verify_pageflags_layout(void);
1064extern void mminit_verify_zonelist(void);
1065#else
1066
1067static inline void mminit_dprintk(enum mminit_level level,
1068 const char *prefix, const char *fmt, ...)
1069{
1070}
1071
1072static inline void mminit_verify_pageflags_layout(void)
1073{
1074}
1075
1076static inline void mminit_verify_zonelist(void)
1077{
1078}
1079#endif /* CONFIG_DEBUG_MEMORY_INIT */
1080
1081#define NODE_RECLAIM_NOSCAN -2
1082#define NODE_RECLAIM_FULL -1
1083#define NODE_RECLAIM_SOME 0
1084#define NODE_RECLAIM_SUCCESS 1
1085
1086#ifdef CONFIG_NUMA
1087extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int);
1088extern int find_next_best_node(int node, nodemask_t *used_node_mask);
1089#else
1090static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask,
1091 unsigned int order)
1092{
1093 return NODE_RECLAIM_NOSCAN;
1094}
1095static inline int find_next_best_node(int node, nodemask_t *used_node_mask)
1096{
1097 return NUMA_NO_NODE;
1098}
1099#endif
1100
1101/*
1102 * mm/memory-failure.c
1103 */
1104#ifdef CONFIG_MEMORY_FAILURE
1105int unmap_poisoned_folio(struct folio *folio, unsigned long pfn, bool must_kill);
1106void shake_folio(struct folio *folio);
1107extern int hwpoison_filter(struct page *p);
1108
1109extern u32 hwpoison_filter_dev_major;
1110extern u32 hwpoison_filter_dev_minor;
1111extern u64 hwpoison_filter_flags_mask;
1112extern u64 hwpoison_filter_flags_value;
1113extern u64 hwpoison_filter_memcg;
1114extern u32 hwpoison_filter_enable;
1115#define MAGIC_HWPOISON 0x48575053U /* HWPS */
1116void SetPageHWPoisonTakenOff(struct page *page);
1117void ClearPageHWPoisonTakenOff(struct page *page);
1118bool take_page_off_buddy(struct page *page);
1119bool put_page_back_buddy(struct page *page);
1120struct task_struct *task_early_kill(struct task_struct *tsk, int force_early);
1121void add_to_kill_ksm(struct task_struct *tsk, const struct page *p,
1122 struct vm_area_struct *vma, struct list_head *to_kill,
1123 unsigned long ksm_addr);
1124unsigned long page_mapped_in_vma(const struct page *page,
1125 struct vm_area_struct *vma);
1126
1127#else
1128static inline int unmap_poisoned_folio(struct folio *folio, unsigned long pfn, bool must_kill)
1129{
1130 return -EBUSY;
1131}
1132#endif
1133
1134extern unsigned long __must_check vm_mmap_pgoff(struct file *, unsigned long,
1135 unsigned long, unsigned long,
1136 unsigned long, unsigned long);
1137
1138extern void set_pageblock_order(void);
1139struct folio *alloc_migrate_folio(struct folio *src, unsigned long private);
1140unsigned long reclaim_pages(struct list_head *folio_list);
1141unsigned int reclaim_clean_pages_from_list(struct zone *zone,
1142 struct list_head *folio_list);
1143/* The ALLOC_WMARK bits are used as an index to zone->watermark */
1144#define ALLOC_WMARK_MIN WMARK_MIN
1145#define ALLOC_WMARK_LOW WMARK_LOW
1146#define ALLOC_WMARK_HIGH WMARK_HIGH
1147#define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
1148
1149/* Mask to get the watermark bits */
1150#define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
1151
1152/*
1153 * Only MMU archs have async oom victim reclaim - aka oom_reaper so we
1154 * cannot assume a reduced access to memory reserves is sufficient for
1155 * !MMU
1156 */
1157#ifdef CONFIG_MMU
1158#define ALLOC_OOM 0x08
1159#else
1160#define ALLOC_OOM ALLOC_NO_WATERMARKS
1161#endif
1162
1163#define ALLOC_NON_BLOCK 0x10 /* Caller cannot block. Allow access
1164 * to 25% of the min watermark or
1165 * 62.5% if __GFP_HIGH is set.
1166 */
1167#define ALLOC_MIN_RESERVE 0x20 /* __GFP_HIGH set. Allow access to 50%
1168 * of the min watermark.
1169 */
1170#define ALLOC_CPUSET 0x40 /* check for correct cpuset */
1171#define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
1172#ifdef CONFIG_ZONE_DMA32
1173#define ALLOC_NOFRAGMENT 0x100 /* avoid mixing pageblock types */
1174#else
1175#define ALLOC_NOFRAGMENT 0x0
1176#endif
1177#define ALLOC_HIGHATOMIC 0x200 /* Allows access to MIGRATE_HIGHATOMIC */
1178#define ALLOC_KSWAPD 0x800 /* allow waking of kswapd, __GFP_KSWAPD_RECLAIM set */
1179
1180/* Flags that allow allocations below the min watermark. */
1181#define ALLOC_RESERVES (ALLOC_NON_BLOCK|ALLOC_MIN_RESERVE|ALLOC_HIGHATOMIC|ALLOC_OOM)
1182
1183enum ttu_flags;
1184struct tlbflush_unmap_batch;
1185
1186
1187/*
1188 * only for MM internal work items which do not depend on
1189 * any allocations or locks which might depend on allocations
1190 */
1191extern struct workqueue_struct *mm_percpu_wq;
1192
1193#ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
1194void try_to_unmap_flush(void);
1195void try_to_unmap_flush_dirty(void);
1196void flush_tlb_batched_pending(struct mm_struct *mm);
1197#else
1198static inline void try_to_unmap_flush(void)
1199{
1200}
1201static inline void try_to_unmap_flush_dirty(void)
1202{
1203}
1204static inline void flush_tlb_batched_pending(struct mm_struct *mm)
1205{
1206}
1207#endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
1208
1209extern const struct trace_print_flags pageflag_names[];
1210extern const struct trace_print_flags vmaflag_names[];
1211extern const struct trace_print_flags gfpflag_names[];
1212
1213static inline bool is_migrate_highatomic(enum migratetype migratetype)
1214{
1215 return migratetype == MIGRATE_HIGHATOMIC;
1216}
1217
1218void setup_zone_pageset(struct zone *zone);
1219
1220struct migration_target_control {
1221 int nid; /* preferred node id */
1222 nodemask_t *nmask;
1223 gfp_t gfp_mask;
1224 enum migrate_reason reason;
1225};
1226
1227/*
1228 * mm/filemap.c
1229 */
1230size_t splice_folio_into_pipe(struct pipe_inode_info *pipe,
1231 struct folio *folio, loff_t fpos, size_t size);
1232
1233/*
1234 * mm/vmalloc.c
1235 */
1236#ifdef CONFIG_MMU
1237void __init vmalloc_init(void);
1238int __must_check vmap_pages_range_noflush(unsigned long addr, unsigned long end,
1239 pgprot_t prot, struct page **pages, unsigned int page_shift);
1240unsigned int get_vm_area_page_order(struct vm_struct *vm);
1241#else
1242static inline void vmalloc_init(void)
1243{
1244}
1245
1246static inline
1247int __must_check vmap_pages_range_noflush(unsigned long addr, unsigned long end,
1248 pgprot_t prot, struct page **pages, unsigned int page_shift)
1249{
1250 return -EINVAL;
1251}
1252#endif
1253
1254int __must_check __vmap_pages_range_noflush(unsigned long addr,
1255 unsigned long end, pgprot_t prot,
1256 struct page **pages, unsigned int page_shift);
1257
1258void vunmap_range_noflush(unsigned long start, unsigned long end);
1259
1260void __vunmap_range_noflush(unsigned long start, unsigned long end);
1261
1262int numa_migrate_check(struct folio *folio, struct vm_fault *vmf,
1263 unsigned long addr, int *flags, bool writable,
1264 int *last_cpupid);
1265
1266void free_zone_device_folio(struct folio *folio);
1267int migrate_device_coherent_folio(struct folio *folio);
1268
1269struct vm_struct *__get_vm_area_node(unsigned long size,
1270 unsigned long align, unsigned long shift,
1271 unsigned long flags, unsigned long start,
1272 unsigned long end, int node, gfp_t gfp_mask,
1273 const void *caller);
1274
1275/*
1276 * mm/gup.c
1277 */
1278int __must_check try_grab_folio(struct folio *folio, int refs,
1279 unsigned int flags);
1280
1281/*
1282 * mm/huge_memory.c
1283 */
1284void touch_pud(struct vm_area_struct *vma, unsigned long addr,
1285 pud_t *pud, bool write);
1286void touch_pmd(struct vm_area_struct *vma, unsigned long addr,
1287 pmd_t *pmd, bool write);
1288
1289/*
1290 * Parses a string with mem suffixes into its order. Useful to parse kernel
1291 * parameters.
1292 */
1293static inline int get_order_from_str(const char *size_str,
1294 unsigned long valid_orders)
1295{
1296 unsigned long size;
1297 char *endptr;
1298 int order;
1299
1300 size = memparse(size_str, &endptr);
1301
1302 if (!is_power_of_2(size))
1303 return -EINVAL;
1304 order = get_order(size);
1305 if (BIT(order) & ~valid_orders)
1306 return -EINVAL;
1307
1308 return order;
1309}
1310
1311enum {
1312 /* mark page accessed */
1313 FOLL_TOUCH = 1 << 16,
1314 /* a retry, previous pass started an IO */
1315 FOLL_TRIED = 1 << 17,
1316 /* we are working on non-current tsk/mm */
1317 FOLL_REMOTE = 1 << 18,
1318 /* pages must be released via unpin_user_page */
1319 FOLL_PIN = 1 << 19,
1320 /* gup_fast: prevent fall-back to slow gup */
1321 FOLL_FAST_ONLY = 1 << 20,
1322 /* allow unlocking the mmap lock */
1323 FOLL_UNLOCKABLE = 1 << 21,
1324 /* VMA lookup+checks compatible with MADV_POPULATE_(READ|WRITE) */
1325 FOLL_MADV_POPULATE = 1 << 22,
1326};
1327
1328#define INTERNAL_GUP_FLAGS (FOLL_TOUCH | FOLL_TRIED | FOLL_REMOTE | FOLL_PIN | \
1329 FOLL_FAST_ONLY | FOLL_UNLOCKABLE | \
1330 FOLL_MADV_POPULATE)
1331
1332/*
1333 * Indicates for which pages that are write-protected in the page table,
1334 * whether GUP has to trigger unsharing via FAULT_FLAG_UNSHARE such that the
1335 * GUP pin will remain consistent with the pages mapped into the page tables
1336 * of the MM.
1337 *
1338 * Temporary unmapping of PageAnonExclusive() pages or clearing of
1339 * PageAnonExclusive() has to protect against concurrent GUP:
1340 * * Ordinary GUP: Using the PT lock
1341 * * GUP-fast and fork(): mm->write_protect_seq
1342 * * GUP-fast and KSM or temporary unmapping (swap, migration): see
1343 * folio_try_share_anon_rmap_*()
1344 *
1345 * Must be called with the (sub)page that's actually referenced via the
1346 * page table entry, which might not necessarily be the head page for a
1347 * PTE-mapped THP.
1348 *
1349 * If the vma is NULL, we're coming from the GUP-fast path and might have
1350 * to fallback to the slow path just to lookup the vma.
1351 */
1352static inline bool gup_must_unshare(struct vm_area_struct *vma,
1353 unsigned int flags, struct page *page)
1354{
1355 /*
1356 * FOLL_WRITE is implicitly handled correctly as the page table entry
1357 * has to be writable -- and if it references (part of) an anonymous
1358 * folio, that part is required to be marked exclusive.
1359 */
1360 if ((flags & (FOLL_WRITE | FOLL_PIN)) != FOLL_PIN)
1361 return false;
1362 /*
1363 * Note: PageAnon(page) is stable until the page is actually getting
1364 * freed.
1365 */
1366 if (!PageAnon(page)) {
1367 /*
1368 * We only care about R/O long-term pining: R/O short-term
1369 * pinning does not have the semantics to observe successive
1370 * changes through the process page tables.
1371 */
1372 if (!(flags & FOLL_LONGTERM))
1373 return false;
1374
1375 /* We really need the vma ... */
1376 if (!vma)
1377 return true;
1378
1379 /*
1380 * ... because we only care about writable private ("COW")
1381 * mappings where we have to break COW early.
1382 */
1383 return is_cow_mapping(vma->vm_flags);
1384 }
1385
1386 /* Paired with a memory barrier in folio_try_share_anon_rmap_*(). */
1387 if (IS_ENABLED(CONFIG_HAVE_GUP_FAST))
1388 smp_rmb();
1389
1390 /*
1391 * Note that KSM pages cannot be exclusive, and consequently,
1392 * cannot get pinned.
1393 */
1394 return !PageAnonExclusive(page);
1395}
1396
1397extern bool mirrored_kernelcore;
1398extern bool memblock_has_mirror(void);
1399
1400static __always_inline void vma_set_range(struct vm_area_struct *vma,
1401 unsigned long start, unsigned long end,
1402 pgoff_t pgoff)
1403{
1404 vma->vm_start = start;
1405 vma->vm_end = end;
1406 vma->vm_pgoff = pgoff;
1407}
1408
1409static inline bool vma_soft_dirty_enabled(struct vm_area_struct *vma)
1410{
1411 /*
1412 * NOTE: we must check this before VM_SOFTDIRTY on soft-dirty
1413 * enablements, because when without soft-dirty being compiled in,
1414 * VM_SOFTDIRTY is defined as 0x0, then !(vm_flags & VM_SOFTDIRTY)
1415 * will be constantly true.
1416 */
1417 if (!IS_ENABLED(CONFIG_MEM_SOFT_DIRTY))
1418 return false;
1419
1420 /*
1421 * Soft-dirty is kind of special: its tracking is enabled when the
1422 * vma flags not set.
1423 */
1424 return !(vma->vm_flags & VM_SOFTDIRTY);
1425}
1426
1427static inline bool pmd_needs_soft_dirty_wp(struct vm_area_struct *vma, pmd_t pmd)
1428{
1429 return vma_soft_dirty_enabled(vma) && !pmd_soft_dirty(pmd);
1430}
1431
1432static inline bool pte_needs_soft_dirty_wp(struct vm_area_struct *vma, pte_t pte)
1433{
1434 return vma_soft_dirty_enabled(vma) && !pte_soft_dirty(pte);
1435}
1436
1437void __meminit __init_single_page(struct page *page, unsigned long pfn,
1438 unsigned long zone, int nid);
1439
1440/* shrinker related functions */
1441unsigned long shrink_slab(gfp_t gfp_mask, int nid, struct mem_cgroup *memcg,
1442 int priority);
1443
1444#ifdef CONFIG_64BIT
1445static inline int can_do_mseal(unsigned long flags)
1446{
1447 if (flags)
1448 return -EINVAL;
1449
1450 return 0;
1451}
1452
1453#else
1454static inline int can_do_mseal(unsigned long flags)
1455{
1456 return -EPERM;
1457}
1458#endif
1459
1460#ifdef CONFIG_SHRINKER_DEBUG
1461static inline __printf(2, 0) int shrinker_debugfs_name_alloc(
1462 struct shrinker *shrinker, const char *fmt, va_list ap)
1463{
1464 shrinker->name = kvasprintf_const(GFP_KERNEL, fmt, ap);
1465
1466 return shrinker->name ? 0 : -ENOMEM;
1467}
1468
1469static inline void shrinker_debugfs_name_free(struct shrinker *shrinker)
1470{
1471 kfree_const(shrinker->name);
1472 shrinker->name = NULL;
1473}
1474
1475extern int shrinker_debugfs_add(struct shrinker *shrinker);
1476extern struct dentry *shrinker_debugfs_detach(struct shrinker *shrinker,
1477 int *debugfs_id);
1478extern void shrinker_debugfs_remove(struct dentry *debugfs_entry,
1479 int debugfs_id);
1480#else /* CONFIG_SHRINKER_DEBUG */
1481static inline int shrinker_debugfs_add(struct shrinker *shrinker)
1482{
1483 return 0;
1484}
1485static inline int shrinker_debugfs_name_alloc(struct shrinker *shrinker,
1486 const char *fmt, va_list ap)
1487{
1488 return 0;
1489}
1490static inline void shrinker_debugfs_name_free(struct shrinker *shrinker)
1491{
1492}
1493static inline struct dentry *shrinker_debugfs_detach(struct shrinker *shrinker,
1494 int *debugfs_id)
1495{
1496 *debugfs_id = -1;
1497 return NULL;
1498}
1499static inline void shrinker_debugfs_remove(struct dentry *debugfs_entry,
1500 int debugfs_id)
1501{
1502}
1503#endif /* CONFIG_SHRINKER_DEBUG */
1504
1505/* Only track the nodes of mappings with shadow entries */
1506void workingset_update_node(struct xa_node *node);
1507extern struct list_lru shadow_nodes;
1508#define mapping_set_update(xas, mapping) do { \
1509 if (!dax_mapping(mapping) && !shmem_mapping(mapping)) { \
1510 xas_set_update(xas, workingset_update_node); \
1511 xas_set_lru(xas, &shadow_nodes); \
1512 } \
1513} while (0)
1514
1515/* mremap.c */
1516unsigned long move_page_tables(struct vm_area_struct *vma,
1517 unsigned long old_addr, struct vm_area_struct *new_vma,
1518 unsigned long new_addr, unsigned long len,
1519 bool need_rmap_locks, bool for_stack);
1520
1521#ifdef CONFIG_UNACCEPTED_MEMORY
1522void accept_page(struct page *page);
1523#else /* CONFIG_UNACCEPTED_MEMORY */
1524static inline void accept_page(struct page *page)
1525{
1526}
1527#endif /* CONFIG_UNACCEPTED_MEMORY */
1528
1529/* pagewalk.c */
1530int walk_page_range_mm(struct mm_struct *mm, unsigned long start,
1531 unsigned long end, const struct mm_walk_ops *ops,
1532 void *private);
1533
1534#endif /* __MM_INTERNAL_H */