1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /* internal.h: mm/ internal definitions
4 * Copyright (C) 2004 Red Hat, Inc. All Rights Reserved.
7 #ifndef __MM_INTERNAL_H
8 #define __MM_INTERNAL_H
12 #include <linux/pagemap.h>
13 #include <linux/rmap.h>
14 #include <linux/tracepoint-defs.h>
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.
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_ATOMIC|__GFP_NOLOCKDEP)
29 /* The GFP flags allowed during early boot */
30 #define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_RECLAIM|__GFP_IO|__GFP_FS))
32 /* Control allocation cpuset and node placement constraints */
33 #define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE)
35 /* Do not use these with a slab allocator */
36 #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
39 * Different from WARN_ON_ONCE(), no warning will be issued
40 * when we specify __GFP_NOWARN.
42 #define WARN_ON_ONCE_GFP(cond, gfp) ({ \
43 static bool __section(".data.once") __warned; \
44 int __ret_warn_once = !!(cond); \
46 if (unlikely(!(gfp & __GFP_NOWARN) && __ret_warn_once && !__warned)) { \
50 unlikely(__ret_warn_once); \
53 void page_writeback_init(void);
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 COMPOUND_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.
61 #define COMPOUND_MAPPED 0x800000
62 #define FOLIO_PAGES_MAPPED (COMPOUND_MAPPED - 1)
65 * How many individual pages have an elevated _mapcount. Excludes
66 * the folio's entire_mapcount.
68 static inline int folio_nr_pages_mapped(struct folio *folio)
70 return atomic_read(&folio->_nr_pages_mapped) & FOLIO_PAGES_MAPPED;
73 static inline void *folio_raw_mapping(struct folio *folio)
75 unsigned long mapping = (unsigned long)folio->mapping;
77 return (void *)(mapping & ~PAGE_MAPPING_FLAGS);
80 void __acct_reclaim_writeback(pg_data_t *pgdat, struct folio *folio,
82 static inline void acct_reclaim_writeback(struct folio *folio)
84 pg_data_t *pgdat = folio_pgdat(folio);
85 int nr_throttled = atomic_read(&pgdat->nr_writeback_throttled);
88 __acct_reclaim_writeback(pgdat, folio, nr_throttled);
91 static inline void wake_throttle_isolated(pg_data_t *pgdat)
93 wait_queue_head_t *wqh;
95 wqh = &pgdat->reclaim_wait[VMSCAN_THROTTLE_ISOLATED];
96 if (waitqueue_active(wqh))
100 vm_fault_t do_swap_page(struct vm_fault *vmf);
101 void folio_rotate_reclaimable(struct folio *folio);
102 bool __folio_end_writeback(struct folio *folio);
103 void deactivate_file_folio(struct folio *folio);
104 void folio_activate(struct folio *folio);
106 void free_pgtables(struct mmu_gather *tlb, struct maple_tree *mt,
107 struct vm_area_struct *start_vma, unsigned long floor,
108 unsigned long ceiling);
109 void pmd_install(struct mm_struct *mm, pmd_t *pmd, pgtable_t *pte);
112 void unmap_page_range(struct mmu_gather *tlb,
113 struct vm_area_struct *vma,
114 unsigned long addr, unsigned long end,
115 struct zap_details *details);
117 void page_cache_ra_order(struct readahead_control *, struct file_ra_state *,
119 void force_page_cache_ra(struct readahead_control *, unsigned long nr);
120 static inline void force_page_cache_readahead(struct address_space *mapping,
121 struct file *file, pgoff_t index, unsigned long nr_to_read)
123 DEFINE_READAHEAD(ractl, file, &file->f_ra, mapping, index);
124 force_page_cache_ra(&ractl, nr_to_read);
127 unsigned find_lock_entries(struct address_space *mapping, pgoff_t *start,
128 pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices);
129 unsigned find_get_entries(struct address_space *mapping, pgoff_t *start,
130 pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices);
131 void filemap_free_folio(struct address_space *mapping, struct folio *folio);
132 int truncate_inode_folio(struct address_space *mapping, struct folio *folio);
133 bool truncate_inode_partial_folio(struct folio *folio, loff_t start,
135 long invalidate_inode_page(struct page *page);
136 unsigned long invalidate_mapping_pagevec(struct address_space *mapping,
137 pgoff_t start, pgoff_t end, unsigned long *nr_pagevec);
140 * folio_evictable - Test whether a folio is evictable.
141 * @folio: The folio to test.
143 * Test whether @folio is evictable -- i.e., should be placed on
144 * active/inactive lists vs unevictable list.
146 * Reasons folio might not be evictable:
147 * 1. folio's mapping marked unevictable
148 * 2. One of the pages in the folio is part of an mlocked VMA
150 static inline bool folio_evictable(struct folio *folio)
154 /* Prevent address_space of inode and swap cache from being freed */
156 ret = !mapping_unevictable(folio_mapping(folio)) &&
157 !folio_test_mlocked(folio);
162 static inline bool page_evictable(struct page *page)
166 /* Prevent address_space of inode and swap cache from being freed */
168 ret = !mapping_unevictable(page_mapping(page)) && !PageMlocked(page);
174 * Turn a non-refcounted page (->_refcount == 0) into refcounted with
177 static inline void set_page_refcounted(struct page *page)
179 VM_BUG_ON_PAGE(PageTail(page), page);
180 VM_BUG_ON_PAGE(page_ref_count(page), page);
181 set_page_count(page, 1);
184 extern unsigned long highest_memmap_pfn;
187 * Maximum number of reclaim retries without progress before the OOM
188 * killer is consider the only way forward.
190 #define MAX_RECLAIM_RETRIES 16
193 * in mm/early_ioremap.c
195 pgprot_t __init early_memremap_pgprot_adjust(resource_size_t phys_addr,
196 unsigned long size, pgprot_t prot);
201 int isolate_lru_page(struct page *page);
202 int folio_isolate_lru(struct folio *folio);
203 void putback_lru_page(struct page *page);
204 void folio_putback_lru(struct folio *folio);
205 extern void reclaim_throttle(pg_data_t *pgdat, enum vmscan_throttle_state reason);
210 pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
217 * Structure for holding the mostly immutable allocation parameters passed
218 * between functions involved in allocations, including the alloc_pages*
219 * family of functions.
221 * nodemask, migratetype and highest_zoneidx are initialized only once in
222 * __alloc_pages() and then never change.
224 * zonelist, preferred_zone and highest_zoneidx are set first in
225 * __alloc_pages() for the fast path, and might be later changed
226 * in __alloc_pages_slowpath(). All other functions pass the whole structure
227 * by a const pointer.
229 struct alloc_context {
230 struct zonelist *zonelist;
231 nodemask_t *nodemask;
232 struct zoneref *preferred_zoneref;
236 * highest_zoneidx represents highest usable zone index of
237 * the allocation request. Due to the nature of the zone,
238 * memory on lower zone than the highest_zoneidx will be
239 * protected by lowmem_reserve[highest_zoneidx].
241 * highest_zoneidx is also used by reclaim/compaction to limit
242 * the target zone since higher zone than this index cannot be
243 * usable for this allocation request.
245 enum zone_type highest_zoneidx;
246 bool spread_dirty_pages;
250 * This function returns the order of a free page in the buddy system. In
251 * general, page_zone(page)->lock must be held by the caller to prevent the
252 * page from being allocated in parallel and returning garbage as the order.
253 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
254 * page cannot be allocated or merged in parallel. Alternatively, it must
255 * handle invalid values gracefully, and use buddy_order_unsafe() below.
257 static inline unsigned int buddy_order(struct page *page)
259 /* PageBuddy() must be checked by the caller */
260 return page_private(page);
264 * Like buddy_order(), but for callers who cannot afford to hold the zone lock.
265 * PageBuddy() should be checked first by the caller to minimize race window,
266 * and invalid values must be handled gracefully.
268 * READ_ONCE is used so that if the caller assigns the result into a local
269 * variable and e.g. tests it for valid range before using, the compiler cannot
270 * decide to remove the variable and inline the page_private(page) multiple
271 * times, potentially observing different values in the tests and the actual
274 #define buddy_order_unsafe(page) READ_ONCE(page_private(page))
277 * This function checks whether a page is free && is the buddy
278 * we can coalesce a page and its buddy if
279 * (a) the buddy is not in a hole (check before calling!) &&
280 * (b) the buddy is in the buddy system &&
281 * (c) a page and its buddy have the same order &&
282 * (d) a page and its buddy are in the same zone.
284 * For recording whether a page is in the buddy system, we set PageBuddy.
285 * Setting, clearing, and testing PageBuddy is serialized by zone->lock.
287 * For recording page's order, we use page_private(page).
289 static inline bool page_is_buddy(struct page *page, struct page *buddy,
292 if (!page_is_guard(buddy) && !PageBuddy(buddy))
295 if (buddy_order(buddy) != order)
299 * zone check is done late to avoid uselessly calculating
300 * zone/node ids for pages that could never merge.
302 if (page_zone_id(page) != page_zone_id(buddy))
305 VM_BUG_ON_PAGE(page_count(buddy) != 0, buddy);
311 * Locate the struct page for both the matching buddy in our
312 * pair (buddy1) and the combined O(n+1) page they form (page).
314 * 1) Any buddy B1 will have an order O twin B2 which satisfies
315 * the following equation:
317 * For example, if the starting buddy (buddy2) is #8 its order
319 * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
321 * 2) Any buddy B will have an order O+1 parent P which
322 * satisfies the following equation:
325 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
327 static inline unsigned long
328 __find_buddy_pfn(unsigned long page_pfn, unsigned int order)
330 return page_pfn ^ (1 << order);
334 * Find the buddy of @page and validate it.
335 * @page: The input page
336 * @pfn: The pfn of the page, it saves a call to page_to_pfn() when the
337 * function is used in the performance-critical __free_one_page().
338 * @order: The order of the page
339 * @buddy_pfn: The output pointer to the buddy pfn, it also saves a call to
342 * The found buddy can be a non PageBuddy, out of @page's zone, or its order is
343 * not the same as @page. The validation is necessary before use it.
345 * Return: the found buddy page or NULL if not found.
347 static inline struct page *find_buddy_page_pfn(struct page *page,
348 unsigned long pfn, unsigned int order, unsigned long *buddy_pfn)
350 unsigned long __buddy_pfn = __find_buddy_pfn(pfn, order);
353 buddy = page + (__buddy_pfn - pfn);
355 *buddy_pfn = __buddy_pfn;
357 if (page_is_buddy(page, buddy, order))
362 extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn,
363 unsigned long end_pfn, struct zone *zone);
365 static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn,
366 unsigned long end_pfn, struct zone *zone)
368 if (zone->contiguous)
369 return pfn_to_page(start_pfn);
371 return __pageblock_pfn_to_page(start_pfn, end_pfn, zone);
374 extern int __isolate_free_page(struct page *page, unsigned int order);
375 extern void __putback_isolated_page(struct page *page, unsigned int order,
377 extern void memblock_free_pages(struct page *page, unsigned long pfn,
379 extern void __free_pages_core(struct page *page, unsigned int order);
380 extern void prep_compound_page(struct page *page, unsigned int order);
381 extern void post_alloc_hook(struct page *page, unsigned int order,
383 extern int user_min_free_kbytes;
385 extern void free_unref_page(struct page *page, unsigned int order);
386 extern void free_unref_page_list(struct list_head *list);
388 extern void zone_pcp_reset(struct zone *zone);
389 extern void zone_pcp_disable(struct zone *zone);
390 extern void zone_pcp_enable(struct zone *zone);
392 extern void *memmap_alloc(phys_addr_t size, phys_addr_t align,
393 phys_addr_t min_addr,
394 int nid, bool exact_nid);
396 int split_free_page(struct page *free_page,
397 unsigned int order, unsigned long split_pfn_offset);
400 * This will have no effect, other than possibly generating a warning, if the
401 * caller passes in a non-large folio.
403 static inline void folio_set_order(struct folio *folio, unsigned int order)
405 if (WARN_ON_ONCE(!folio_test_large(folio)))
408 folio->_folio_order = order;
411 * When hugetlb dissolves a folio, we need to clear the tail
412 * page, rather than setting nr_pages to 1.
414 folio->_folio_nr_pages = order ? 1U << order : 0;
418 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
424 * compact_control is used to track pages being migrated and the free pages
425 * they are being migrated to during memory compaction. The free_pfn starts
426 * at the end of a zone and migrate_pfn begins at the start. Movable pages
427 * are moved to the end of a zone during a compaction run and the run
428 * completes when free_pfn <= migrate_pfn
430 struct compact_control {
431 struct list_head freepages; /* List of free pages to migrate to */
432 struct list_head migratepages; /* List of pages being migrated */
433 unsigned int nr_freepages; /* Number of isolated free pages */
434 unsigned int nr_migratepages; /* Number of pages to migrate */
435 unsigned long free_pfn; /* isolate_freepages search base */
437 * Acts as an in/out parameter to page isolation for migration.
438 * isolate_migratepages uses it as a search base.
439 * isolate_migratepages_block will update the value to the next pfn
440 * after the last isolated one.
442 unsigned long migrate_pfn;
443 unsigned long fast_start_pfn; /* a pfn to start linear scan from */
445 unsigned long total_migrate_scanned;
446 unsigned long total_free_scanned;
447 unsigned short fast_search_fail;/* failures to use free list searches */
448 short search_order; /* order to start a fast search at */
449 const gfp_t gfp_mask; /* gfp mask of a direct compactor */
450 int order; /* order a direct compactor needs */
451 int migratetype; /* migratetype of direct compactor */
452 const unsigned int alloc_flags; /* alloc flags of a direct compactor */
453 const int highest_zoneidx; /* zone index of a direct compactor */
454 enum migrate_mode mode; /* Async or sync migration mode */
455 bool ignore_skip_hint; /* Scan blocks even if marked skip */
456 bool no_set_skip_hint; /* Don't mark blocks for skipping */
457 bool ignore_block_suitable; /* Scan blocks considered unsuitable */
458 bool direct_compaction; /* False from kcompactd or /proc/... */
459 bool proactive_compaction; /* kcompactd proactive compaction */
460 bool whole_zone; /* Whole zone should/has been scanned */
461 bool contended; /* Signal lock contention */
462 bool rescan; /* Rescanning the same pageblock */
463 bool alloc_contig; /* alloc_contig_range allocation */
467 * Used in direct compaction when a page should be taken from the freelists
468 * immediately when one is created during the free path.
470 struct capture_control {
471 struct compact_control *cc;
476 isolate_freepages_range(struct compact_control *cc,
477 unsigned long start_pfn, unsigned long end_pfn);
479 isolate_migratepages_range(struct compact_control *cc,
480 unsigned long low_pfn, unsigned long end_pfn);
482 int __alloc_contig_migrate_range(struct compact_control *cc,
483 unsigned long start, unsigned long end);
485 int find_suitable_fallback(struct free_area *area, unsigned int order,
486 int migratetype, bool only_stealable, bool *can_steal);
489 * These three helpers classifies VMAs for virtual memory accounting.
493 * Executable code area - executable, not writable, not stack
495 static inline bool is_exec_mapping(vm_flags_t flags)
497 return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC;
501 * Stack area - automatically grows in one direction
503 * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
504 * do_mmap() forbids all other combinations.
506 static inline bool is_stack_mapping(vm_flags_t flags)
508 return (flags & VM_STACK) == VM_STACK;
512 * Data area - private, writable, not stack
514 static inline bool is_data_mapping(vm_flags_t flags)
516 return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE;
520 struct anon_vma *folio_anon_vma(struct folio *folio);
523 void unmap_mapping_folio(struct folio *folio);
524 extern long populate_vma_page_range(struct vm_area_struct *vma,
525 unsigned long start, unsigned long end, int *locked);
526 extern long faultin_vma_page_range(struct vm_area_struct *vma,
527 unsigned long start, unsigned long end,
528 bool write, int *locked);
529 extern int mlock_future_check(struct mm_struct *mm, unsigned long flags,
532 * mlock_vma_page() and munlock_vma_page():
533 * should be called with vma's mmap_lock held for read or write,
534 * under page table lock for the pte/pmd being added or removed.
536 * mlock is usually called at the end of page_add_*_rmap(), munlock at
537 * the end of page_remove_rmap(); but new anon folios are managed by
538 * folio_add_lru_vma() calling mlock_new_folio().
540 * @compound is used to include pmd mappings of THPs, but filter out
541 * pte mappings of THPs, which cannot be consistently counted: a pte
542 * mapping of the THP head cannot be distinguished by the page alone.
544 void mlock_folio(struct folio *folio);
545 static inline void mlock_vma_folio(struct folio *folio,
546 struct vm_area_struct *vma, bool compound)
549 * The VM_SPECIAL check here serves two purposes.
550 * 1) VM_IO check prevents migration from double-counting during mlock.
551 * 2) Although mmap_region() and mlock_fixup() take care that VM_LOCKED
552 * is never left set on a VM_SPECIAL vma, there is an interval while
553 * file->f_op->mmap() is using vm_insert_page(s), when VM_LOCKED may
554 * still be set while VM_SPECIAL bits are added: so ignore it then.
556 if (unlikely((vma->vm_flags & (VM_LOCKED|VM_SPECIAL)) == VM_LOCKED) &&
557 (compound || !folio_test_large(folio)))
561 static inline void mlock_vma_page(struct page *page,
562 struct vm_area_struct *vma, bool compound)
564 mlock_vma_folio(page_folio(page), vma, compound);
567 void munlock_folio(struct folio *folio);
569 static inline void munlock_vma_folio(struct folio *folio,
570 struct vm_area_struct *vma, bool compound)
572 if (unlikely(vma->vm_flags & VM_LOCKED) &&
573 (compound || !folio_test_large(folio)))
574 munlock_folio(folio);
577 static inline void munlock_vma_page(struct page *page,
578 struct vm_area_struct *vma, bool compound)
580 munlock_vma_folio(page_folio(page), vma, compound);
582 void mlock_new_folio(struct folio *folio);
583 bool need_mlock_drain(int cpu);
584 void mlock_drain_local(void);
585 void mlock_drain_remote(int cpu);
587 extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
590 * Return the start of user virtual address at the specific offset within
593 static inline unsigned long
594 vma_pgoff_address(pgoff_t pgoff, unsigned long nr_pages,
595 struct vm_area_struct *vma)
597 unsigned long address;
599 if (pgoff >= vma->vm_pgoff) {
600 address = vma->vm_start +
601 ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
602 /* Check for address beyond vma (or wrapped through 0?) */
603 if (address < vma->vm_start || address >= vma->vm_end)
605 } else if (pgoff + nr_pages - 1 >= vma->vm_pgoff) {
606 /* Test above avoids possibility of wrap to 0 on 32-bit */
607 address = vma->vm_start;
615 * Return the start of user virtual address of a page within a vma.
616 * Returns -EFAULT if all of the page is outside the range of vma.
617 * If page is a compound head, the entire compound page is considered.
619 static inline unsigned long
620 vma_address(struct page *page, struct vm_area_struct *vma)
622 VM_BUG_ON_PAGE(PageKsm(page), page); /* KSM page->index unusable */
623 return vma_pgoff_address(page_to_pgoff(page), compound_nr(page), vma);
627 * Then at what user virtual address will none of the range be found in vma?
628 * Assumes that vma_address() already returned a good starting address.
630 static inline unsigned long vma_address_end(struct page_vma_mapped_walk *pvmw)
632 struct vm_area_struct *vma = pvmw->vma;
634 unsigned long address;
636 /* Common case, plus ->pgoff is invalid for KSM */
637 if (pvmw->nr_pages == 1)
638 return pvmw->address + PAGE_SIZE;
640 pgoff = pvmw->pgoff + pvmw->nr_pages;
641 address = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
642 /* Check for address beyond vma (or wrapped through 0?) */
643 if (address < vma->vm_start || address > vma->vm_end)
644 address = vma->vm_end;
648 static inline struct file *maybe_unlock_mmap_for_io(struct vm_fault *vmf,
651 int flags = vmf->flags;
657 * FAULT_FLAG_RETRY_NOWAIT means we don't want to wait on page locks or
658 * anything, so we only pin the file and drop the mmap_lock if only
659 * FAULT_FLAG_ALLOW_RETRY is set, while this is the first attempt.
661 if (fault_flag_allow_retry_first(flags) &&
662 !(flags & FAULT_FLAG_RETRY_NOWAIT)) {
663 fpin = get_file(vmf->vma->vm_file);
664 mmap_read_unlock(vmf->vma->vm_mm);
668 #else /* !CONFIG_MMU */
669 static inline void unmap_mapping_folio(struct folio *folio) { }
670 static inline void mlock_vma_page(struct page *page,
671 struct vm_area_struct *vma, bool compound) { }
672 static inline void munlock_vma_page(struct page *page,
673 struct vm_area_struct *vma, bool compound) { }
674 static inline void mlock_new_folio(struct folio *folio) { }
675 static inline bool need_mlock_drain(int cpu) { return false; }
676 static inline void mlock_drain_local(void) { }
677 static inline void mlock_drain_remote(int cpu) { }
678 static inline void vunmap_range_noflush(unsigned long start, unsigned long end)
681 #endif /* !CONFIG_MMU */
683 /* Memory initialisation debug and verification */
690 #ifdef CONFIG_DEBUG_MEMORY_INIT
692 extern int mminit_loglevel;
694 #define mminit_dprintk(level, prefix, fmt, arg...) \
696 if (level < mminit_loglevel) { \
697 if (level <= MMINIT_WARNING) \
698 pr_warn("mminit::" prefix " " fmt, ##arg); \
700 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
704 extern void mminit_verify_pageflags_layout(void);
705 extern void mminit_verify_zonelist(void);
708 static inline void mminit_dprintk(enum mminit_level level,
709 const char *prefix, const char *fmt, ...)
713 static inline void mminit_verify_pageflags_layout(void)
717 static inline void mminit_verify_zonelist(void)
720 #endif /* CONFIG_DEBUG_MEMORY_INIT */
722 #define NODE_RECLAIM_NOSCAN -2
723 #define NODE_RECLAIM_FULL -1
724 #define NODE_RECLAIM_SOME 0
725 #define NODE_RECLAIM_SUCCESS 1
728 extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int);
729 extern int find_next_best_node(int node, nodemask_t *used_node_mask);
731 static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask,
734 return NODE_RECLAIM_NOSCAN;
736 static inline int find_next_best_node(int node, nodemask_t *used_node_mask)
743 * mm/memory-failure.c
745 extern int hwpoison_filter(struct page *p);
747 extern u32 hwpoison_filter_dev_major;
748 extern u32 hwpoison_filter_dev_minor;
749 extern u64 hwpoison_filter_flags_mask;
750 extern u64 hwpoison_filter_flags_value;
751 extern u64 hwpoison_filter_memcg;
752 extern u32 hwpoison_filter_enable;
754 extern unsigned long __must_check vm_mmap_pgoff(struct file *, unsigned long,
755 unsigned long, unsigned long,
756 unsigned long, unsigned long);
758 extern void set_pageblock_order(void);
759 unsigned int reclaim_clean_pages_from_list(struct zone *zone,
760 struct list_head *page_list);
761 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
762 #define ALLOC_WMARK_MIN WMARK_MIN
763 #define ALLOC_WMARK_LOW WMARK_LOW
764 #define ALLOC_WMARK_HIGH WMARK_HIGH
765 #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
767 /* Mask to get the watermark bits */
768 #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
771 * Only MMU archs have async oom victim reclaim - aka oom_reaper so we
772 * cannot assume a reduced access to memory reserves is sufficient for
776 #define ALLOC_OOM 0x08
778 #define ALLOC_OOM ALLOC_NO_WATERMARKS
781 #define ALLOC_HARDER 0x10 /* try to alloc harder */
782 #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
783 #define ALLOC_CPUSET 0x40 /* check for correct cpuset */
784 #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
785 #ifdef CONFIG_ZONE_DMA32
786 #define ALLOC_NOFRAGMENT 0x100 /* avoid mixing pageblock types */
788 #define ALLOC_NOFRAGMENT 0x0
790 #define ALLOC_KSWAPD 0x800 /* allow waking of kswapd, __GFP_KSWAPD_RECLAIM set */
793 struct tlbflush_unmap_batch;
797 * only for MM internal work items which do not depend on
798 * any allocations or locks which might depend on allocations
800 extern struct workqueue_struct *mm_percpu_wq;
802 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
803 void try_to_unmap_flush(void);
804 void try_to_unmap_flush_dirty(void);
805 void flush_tlb_batched_pending(struct mm_struct *mm);
807 static inline void try_to_unmap_flush(void)
810 static inline void try_to_unmap_flush_dirty(void)
813 static inline void flush_tlb_batched_pending(struct mm_struct *mm)
816 #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
818 extern const struct trace_print_flags pageflag_names[];
819 extern const struct trace_print_flags vmaflag_names[];
820 extern const struct trace_print_flags gfpflag_names[];
822 static inline bool is_migrate_highatomic(enum migratetype migratetype)
824 return migratetype == MIGRATE_HIGHATOMIC;
827 static inline bool is_migrate_highatomic_page(struct page *page)
829 return get_pageblock_migratetype(page) == MIGRATE_HIGHATOMIC;
832 void setup_zone_pageset(struct zone *zone);
834 struct migration_target_control {
835 int nid; /* preferred node id */
844 int vmap_pages_range_noflush(unsigned long addr, unsigned long end,
845 pgprot_t prot, struct page **pages, unsigned int page_shift);
848 int vmap_pages_range_noflush(unsigned long addr, unsigned long end,
849 pgprot_t prot, struct page **pages, unsigned int page_shift)
855 int __vmap_pages_range_noflush(unsigned long addr, unsigned long end,
856 pgprot_t prot, struct page **pages,
857 unsigned int page_shift);
859 void vunmap_range_noflush(unsigned long start, unsigned long end);
861 void __vunmap_range_noflush(unsigned long start, unsigned long end);
863 int numa_migrate_prep(struct page *page, struct vm_area_struct *vma,
864 unsigned long addr, int page_nid, int *flags);
866 void free_zone_device_page(struct page *page);
867 int migrate_device_coherent_page(struct page *page);
872 struct folio *try_grab_folio(struct page *page, int refs, unsigned int flags);
874 extern bool mirrored_kernelcore;
876 static inline bool vma_soft_dirty_enabled(struct vm_area_struct *vma)
879 * NOTE: we must check this before VM_SOFTDIRTY on soft-dirty
880 * enablements, because when without soft-dirty being compiled in,
881 * VM_SOFTDIRTY is defined as 0x0, then !(vm_flags & VM_SOFTDIRTY)
882 * will be constantly true.
884 if (!IS_ENABLED(CONFIG_MEM_SOFT_DIRTY))
888 * Soft-dirty is kind of special: its tracking is enabled when the
891 return !(vma->vm_flags & VM_SOFTDIRTY);
894 #endif /* __MM_INTERNAL_H */