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.
5 * Written by David Howells (dhowells@redhat.com)
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)
38 void page_writeback_init(void);
40 static inline void *folio_raw_mapping(struct folio *folio)
42 unsigned long mapping = (unsigned long)folio->mapping;
44 return (void *)(mapping & ~PAGE_MAPPING_FLAGS);
47 void __acct_reclaim_writeback(pg_data_t *pgdat, struct folio *folio,
49 static inline void acct_reclaim_writeback(struct folio *folio)
51 pg_data_t *pgdat = folio_pgdat(folio);
52 int nr_throttled = atomic_read(&pgdat->nr_writeback_throttled);
55 __acct_reclaim_writeback(pgdat, folio, nr_throttled);
58 static inline void wake_throttle_isolated(pg_data_t *pgdat)
60 wait_queue_head_t *wqh;
62 wqh = &pgdat->reclaim_wait[VMSCAN_THROTTLE_ISOLATED];
63 if (waitqueue_active(wqh))
67 vm_fault_t do_swap_page(struct vm_fault *vmf);
68 void folio_rotate_reclaimable(struct folio *folio);
69 bool __folio_end_writeback(struct folio *folio);
70 void deactivate_file_folio(struct folio *folio);
72 void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
73 unsigned long floor, unsigned long ceiling);
74 void pmd_install(struct mm_struct *mm, pmd_t *pmd, pgtable_t *pte);
77 void unmap_page_range(struct mmu_gather *tlb,
78 struct vm_area_struct *vma,
79 unsigned long addr, unsigned long end,
80 struct zap_details *details);
82 void page_cache_ra_order(struct readahead_control *, struct file_ra_state *,
84 void force_page_cache_ra(struct readahead_control *, unsigned long nr);
85 static inline void force_page_cache_readahead(struct address_space *mapping,
86 struct file *file, pgoff_t index, unsigned long nr_to_read)
88 DEFINE_READAHEAD(ractl, file, &file->f_ra, mapping, index);
89 force_page_cache_ra(&ractl, nr_to_read);
92 unsigned find_lock_entries(struct address_space *mapping, pgoff_t start,
93 pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices);
94 unsigned find_get_entries(struct address_space *mapping, pgoff_t start,
95 pgoff_t end, struct folio_batch *fbatch, pgoff_t *indices);
96 void filemap_free_folio(struct address_space *mapping, struct folio *folio);
97 int truncate_inode_folio(struct address_space *mapping, struct folio *folio);
98 bool truncate_inode_partial_folio(struct folio *folio, loff_t start,
100 long invalidate_inode_page(struct page *page);
101 unsigned long invalidate_mapping_pagevec(struct address_space *mapping,
102 pgoff_t start, pgoff_t end, unsigned long *nr_pagevec);
105 * folio_evictable - Test whether a folio is evictable.
106 * @folio: The folio to test.
108 * Test whether @folio is evictable -- i.e., should be placed on
109 * active/inactive lists vs unevictable list.
111 * Reasons folio might not be evictable:
112 * 1. folio's mapping marked unevictable
113 * 2. One of the pages in the folio is part of an mlocked VMA
115 static inline bool folio_evictable(struct folio *folio)
119 /* Prevent address_space of inode and swap cache from being freed */
121 ret = !mapping_unevictable(folio_mapping(folio)) &&
122 !folio_test_mlocked(folio);
127 static inline bool page_evictable(struct page *page)
131 /* Prevent address_space of inode and swap cache from being freed */
133 ret = !mapping_unevictable(page_mapping(page)) && !PageMlocked(page);
139 * Turn a non-refcounted page (->_refcount == 0) into refcounted with
142 static inline void set_page_refcounted(struct page *page)
144 VM_BUG_ON_PAGE(PageTail(page), page);
145 VM_BUG_ON_PAGE(page_ref_count(page), page);
146 set_page_count(page, 1);
149 extern unsigned long highest_memmap_pfn;
152 * Maximum number of reclaim retries without progress before the OOM
153 * killer is consider the only way forward.
155 #define MAX_RECLAIM_RETRIES 16
158 * in mm/early_ioremap.c
160 pgprot_t __init early_memremap_pgprot_adjust(resource_size_t phys_addr,
161 unsigned long size, pgprot_t prot);
166 int isolate_lru_page(struct page *page);
167 int folio_isolate_lru(struct folio *folio);
168 void putback_lru_page(struct page *page);
169 void folio_putback_lru(struct folio *folio);
170 extern void reclaim_throttle(pg_data_t *pgdat, enum vmscan_throttle_state reason);
175 extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
182 * Structure for holding the mostly immutable allocation parameters passed
183 * between functions involved in allocations, including the alloc_pages*
184 * family of functions.
186 * nodemask, migratetype and highest_zoneidx are initialized only once in
187 * __alloc_pages() and then never change.
189 * zonelist, preferred_zone and highest_zoneidx are set first in
190 * __alloc_pages() for the fast path, and might be later changed
191 * in __alloc_pages_slowpath(). All other functions pass the whole structure
192 * by a const pointer.
194 struct alloc_context {
195 struct zonelist *zonelist;
196 nodemask_t *nodemask;
197 struct zoneref *preferred_zoneref;
201 * highest_zoneidx represents highest usable zone index of
202 * the allocation request. Due to the nature of the zone,
203 * memory on lower zone than the highest_zoneidx will be
204 * protected by lowmem_reserve[highest_zoneidx].
206 * highest_zoneidx is also used by reclaim/compaction to limit
207 * the target zone since higher zone than this index cannot be
208 * usable for this allocation request.
210 enum zone_type highest_zoneidx;
211 bool spread_dirty_pages;
215 * This function returns the order of a free page in the buddy system. In
216 * general, page_zone(page)->lock must be held by the caller to prevent the
217 * page from being allocated in parallel and returning garbage as the order.
218 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
219 * page cannot be allocated or merged in parallel. Alternatively, it must
220 * handle invalid values gracefully, and use buddy_order_unsafe() below.
222 static inline unsigned int buddy_order(struct page *page)
224 /* PageBuddy() must be checked by the caller */
225 return page_private(page);
229 * Like buddy_order(), but for callers who cannot afford to hold the zone lock.
230 * PageBuddy() should be checked first by the caller to minimize race window,
231 * and invalid values must be handled gracefully.
233 * READ_ONCE is used so that if the caller assigns the result into a local
234 * variable and e.g. tests it for valid range before using, the compiler cannot
235 * decide to remove the variable and inline the page_private(page) multiple
236 * times, potentially observing different values in the tests and the actual
239 #define buddy_order_unsafe(page) READ_ONCE(page_private(page))
242 * This function checks whether a page is free && is the buddy
243 * we can coalesce a page and its buddy if
244 * (a) the buddy is not in a hole (check before calling!) &&
245 * (b) the buddy is in the buddy system &&
246 * (c) a page and its buddy have the same order &&
247 * (d) a page and its buddy are in the same zone.
249 * For recording whether a page is in the buddy system, we set PageBuddy.
250 * Setting, clearing, and testing PageBuddy is serialized by zone->lock.
252 * For recording page's order, we use page_private(page).
254 static inline bool page_is_buddy(struct page *page, struct page *buddy,
257 if (!page_is_guard(buddy) && !PageBuddy(buddy))
260 if (buddy_order(buddy) != order)
264 * zone check is done late to avoid uselessly calculating
265 * zone/node ids for pages that could never merge.
267 if (page_zone_id(page) != page_zone_id(buddy))
270 VM_BUG_ON_PAGE(page_count(buddy) != 0, buddy);
276 * Locate the struct page for both the matching buddy in our
277 * pair (buddy1) and the combined O(n+1) page they form (page).
279 * 1) Any buddy B1 will have an order O twin B2 which satisfies
280 * the following equation:
282 * For example, if the starting buddy (buddy2) is #8 its order
284 * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
286 * 2) Any buddy B will have an order O+1 parent P which
287 * satisfies the following equation:
290 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
292 static inline unsigned long
293 __find_buddy_pfn(unsigned long page_pfn, unsigned int order)
295 return page_pfn ^ (1 << order);
299 * Find the buddy of @page and validate it.
300 * @page: The input page
301 * @pfn: The pfn of the page, it saves a call to page_to_pfn() when the
302 * function is used in the performance-critical __free_one_page().
303 * @order: The order of the page
304 * @buddy_pfn: The output pointer to the buddy pfn, it also saves a call to
307 * The found buddy can be a non PageBuddy, out of @page's zone, or its order is
308 * not the same as @page. The validation is necessary before use it.
310 * Return: the found buddy page or NULL if not found.
312 static inline struct page *find_buddy_page_pfn(struct page *page,
313 unsigned long pfn, unsigned int order, unsigned long *buddy_pfn)
315 unsigned long __buddy_pfn = __find_buddy_pfn(pfn, order);
318 buddy = page + (__buddy_pfn - pfn);
320 *buddy_pfn = __buddy_pfn;
322 if (page_is_buddy(page, buddy, order))
327 extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn,
328 unsigned long end_pfn, struct zone *zone);
330 static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn,
331 unsigned long end_pfn, struct zone *zone)
333 if (zone->contiguous)
334 return pfn_to_page(start_pfn);
336 return __pageblock_pfn_to_page(start_pfn, end_pfn, zone);
339 extern int __isolate_free_page(struct page *page, unsigned int order);
340 extern void __putback_isolated_page(struct page *page, unsigned int order,
342 extern void memblock_free_pages(struct page *page, unsigned long pfn,
344 extern void __free_pages_core(struct page *page, unsigned int order);
345 extern void prep_compound_page(struct page *page, unsigned int order);
346 extern void post_alloc_hook(struct page *page, unsigned int order,
348 extern int user_min_free_kbytes;
350 extern void free_unref_page(struct page *page, unsigned int order);
351 extern void free_unref_page_list(struct list_head *list);
353 extern void zone_pcp_update(struct zone *zone, int cpu_online);
354 extern void zone_pcp_reset(struct zone *zone);
355 extern void zone_pcp_disable(struct zone *zone);
356 extern void zone_pcp_enable(struct zone *zone);
358 extern void *memmap_alloc(phys_addr_t size, phys_addr_t align,
359 phys_addr_t min_addr,
360 int nid, bool exact_nid);
362 void split_free_page(struct page *free_page,
363 int order, unsigned long split_pfn_offset);
365 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
371 * compact_control is used to track pages being migrated and the free pages
372 * they are being migrated to during memory compaction. The free_pfn starts
373 * at the end of a zone and migrate_pfn begins at the start. Movable pages
374 * are moved to the end of a zone during a compaction run and the run
375 * completes when free_pfn <= migrate_pfn
377 struct compact_control {
378 struct list_head freepages; /* List of free pages to migrate to */
379 struct list_head migratepages; /* List of pages being migrated */
380 unsigned int nr_freepages; /* Number of isolated free pages */
381 unsigned int nr_migratepages; /* Number of pages to migrate */
382 unsigned long free_pfn; /* isolate_freepages search base */
384 * Acts as an in/out parameter to page isolation for migration.
385 * isolate_migratepages uses it as a search base.
386 * isolate_migratepages_block will update the value to the next pfn
387 * after the last isolated one.
389 unsigned long migrate_pfn;
390 unsigned long fast_start_pfn; /* a pfn to start linear scan from */
392 unsigned long total_migrate_scanned;
393 unsigned long total_free_scanned;
394 unsigned short fast_search_fail;/* failures to use free list searches */
395 short search_order; /* order to start a fast search at */
396 const gfp_t gfp_mask; /* gfp mask of a direct compactor */
397 int order; /* order a direct compactor needs */
398 int migratetype; /* migratetype of direct compactor */
399 const unsigned int alloc_flags; /* alloc flags of a direct compactor */
400 const int highest_zoneidx; /* zone index of a direct compactor */
401 enum migrate_mode mode; /* Async or sync migration mode */
402 bool ignore_skip_hint; /* Scan blocks even if marked skip */
403 bool no_set_skip_hint; /* Don't mark blocks for skipping */
404 bool ignore_block_suitable; /* Scan blocks considered unsuitable */
405 bool direct_compaction; /* False from kcompactd or /proc/... */
406 bool proactive_compaction; /* kcompactd proactive compaction */
407 bool whole_zone; /* Whole zone should/has been scanned */
408 bool contended; /* Signal lock contention */
409 bool rescan; /* Rescanning the same pageblock */
410 bool alloc_contig; /* alloc_contig_range allocation */
414 * Used in direct compaction when a page should be taken from the freelists
415 * immediately when one is created during the free path.
417 struct capture_control {
418 struct compact_control *cc;
423 isolate_freepages_range(struct compact_control *cc,
424 unsigned long start_pfn, unsigned long end_pfn);
426 isolate_migratepages_range(struct compact_control *cc,
427 unsigned long low_pfn, unsigned long end_pfn);
429 int __alloc_contig_migrate_range(struct compact_control *cc,
430 unsigned long start, unsigned long end);
432 int find_suitable_fallback(struct free_area *area, unsigned int order,
433 int migratetype, bool only_stealable, bool *can_steal);
436 * These three helpers classifies VMAs for virtual memory accounting.
440 * Executable code area - executable, not writable, not stack
442 static inline bool is_exec_mapping(vm_flags_t flags)
444 return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC;
448 * Stack area - automatically grows in one direction
450 * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
451 * do_mmap() forbids all other combinations.
453 static inline bool is_stack_mapping(vm_flags_t flags)
455 return (flags & VM_STACK) == VM_STACK;
459 * Data area - private, writable, not stack
461 static inline bool is_data_mapping(vm_flags_t flags)
463 return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE;
467 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
468 struct vm_area_struct *prev);
469 void __vma_unlink_list(struct mm_struct *mm, struct vm_area_struct *vma);
470 struct anon_vma *folio_anon_vma(struct folio *folio);
473 void unmap_mapping_folio(struct folio *folio);
474 extern long populate_vma_page_range(struct vm_area_struct *vma,
475 unsigned long start, unsigned long end, int *locked);
476 extern long faultin_vma_page_range(struct vm_area_struct *vma,
477 unsigned long start, unsigned long end,
478 bool write, int *locked);
479 extern int mlock_future_check(struct mm_struct *mm, unsigned long flags,
482 * mlock_vma_page() and munlock_vma_page():
483 * should be called with vma's mmap_lock held for read or write,
484 * under page table lock for the pte/pmd being added or removed.
486 * mlock is usually called at the end of page_add_*_rmap(),
487 * munlock at the end of page_remove_rmap(); but new anon
488 * pages are managed by lru_cache_add_inactive_or_unevictable()
489 * calling mlock_new_page().
491 * @compound is used to include pmd mappings of THPs, but filter out
492 * pte mappings of THPs, which cannot be consistently counted: a pte
493 * mapping of the THP head cannot be distinguished by the page alone.
495 void mlock_folio(struct folio *folio);
496 static inline void mlock_vma_folio(struct folio *folio,
497 struct vm_area_struct *vma, bool compound)
500 * The VM_SPECIAL check here serves two purposes.
501 * 1) VM_IO check prevents migration from double-counting during mlock.
502 * 2) Although mmap_region() and mlock_fixup() take care that VM_LOCKED
503 * is never left set on a VM_SPECIAL vma, there is an interval while
504 * file->f_op->mmap() is using vm_insert_page(s), when VM_LOCKED may
505 * still be set while VM_SPECIAL bits are added: so ignore it then.
507 if (unlikely((vma->vm_flags & (VM_LOCKED|VM_SPECIAL)) == VM_LOCKED) &&
508 (compound || !folio_test_large(folio)))
512 static inline void mlock_vma_page(struct page *page,
513 struct vm_area_struct *vma, bool compound)
515 mlock_vma_folio(page_folio(page), vma, compound);
518 void munlock_page(struct page *page);
519 static inline void munlock_vma_page(struct page *page,
520 struct vm_area_struct *vma, bool compound)
522 if (unlikely(vma->vm_flags & VM_LOCKED) &&
523 (compound || !PageTransCompound(page)))
526 void mlock_new_page(struct page *page);
527 bool need_mlock_page_drain(int cpu);
528 void mlock_page_drain_local(void);
529 void mlock_page_drain_remote(int cpu);
531 extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
534 * Return the start of user virtual address at the specific offset within
537 static inline unsigned long
538 vma_pgoff_address(pgoff_t pgoff, unsigned long nr_pages,
539 struct vm_area_struct *vma)
541 unsigned long address;
543 if (pgoff >= vma->vm_pgoff) {
544 address = vma->vm_start +
545 ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
546 /* Check for address beyond vma (or wrapped through 0?) */
547 if (address < vma->vm_start || address >= vma->vm_end)
549 } else if (pgoff + nr_pages - 1 >= vma->vm_pgoff) {
550 /* Test above avoids possibility of wrap to 0 on 32-bit */
551 address = vma->vm_start;
559 * Return the start of user virtual address of a page within a vma.
560 * Returns -EFAULT if all of the page is outside the range of vma.
561 * If page is a compound head, the entire compound page is considered.
563 static inline unsigned long
564 vma_address(struct page *page, struct vm_area_struct *vma)
566 VM_BUG_ON_PAGE(PageKsm(page), page); /* KSM page->index unusable */
567 return vma_pgoff_address(page_to_pgoff(page), compound_nr(page), vma);
571 * Then at what user virtual address will none of the range be found in vma?
572 * Assumes that vma_address() already returned a good starting address.
574 static inline unsigned long vma_address_end(struct page_vma_mapped_walk *pvmw)
576 struct vm_area_struct *vma = pvmw->vma;
578 unsigned long address;
580 /* Common case, plus ->pgoff is invalid for KSM */
581 if (pvmw->nr_pages == 1)
582 return pvmw->address + PAGE_SIZE;
584 pgoff = pvmw->pgoff + pvmw->nr_pages;
585 address = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
586 /* Check for address beyond vma (or wrapped through 0?) */
587 if (address < vma->vm_start || address > vma->vm_end)
588 address = vma->vm_end;
592 static inline struct file *maybe_unlock_mmap_for_io(struct vm_fault *vmf,
595 int flags = vmf->flags;
601 * FAULT_FLAG_RETRY_NOWAIT means we don't want to wait on page locks or
602 * anything, so we only pin the file and drop the mmap_lock if only
603 * FAULT_FLAG_ALLOW_RETRY is set, while this is the first attempt.
605 if (fault_flag_allow_retry_first(flags) &&
606 !(flags & FAULT_FLAG_RETRY_NOWAIT)) {
607 fpin = get_file(vmf->vma->vm_file);
608 mmap_read_unlock(vmf->vma->vm_mm);
612 #else /* !CONFIG_MMU */
613 static inline void unmap_mapping_folio(struct folio *folio) { }
614 static inline void mlock_vma_page(struct page *page,
615 struct vm_area_struct *vma, bool compound) { }
616 static inline void munlock_vma_page(struct page *page,
617 struct vm_area_struct *vma, bool compound) { }
618 static inline void mlock_new_page(struct page *page) { }
619 static inline bool need_mlock_page_drain(int cpu) { return false; }
620 static inline void mlock_page_drain_local(void) { }
621 static inline void mlock_page_drain_remote(int cpu) { }
622 static inline void vunmap_range_noflush(unsigned long start, unsigned long end)
625 #endif /* !CONFIG_MMU */
628 * Return the mem_map entry representing the 'offset' subpage within
629 * the maximally aligned gigantic page 'base'. Handle any discontiguity
630 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
632 static inline struct page *mem_map_offset(struct page *base, int offset)
634 if (unlikely(offset >= MAX_ORDER_NR_PAGES))
635 return nth_page(base, offset);
636 return base + offset;
640 * Iterator over all subpages within the maximally aligned gigantic
641 * page 'base'. Handle any discontiguity in the mem_map.
643 static inline struct page *mem_map_next(struct page *iter,
644 struct page *base, int offset)
646 if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
647 unsigned long pfn = page_to_pfn(base) + offset;
650 return pfn_to_page(pfn);
655 /* Memory initialisation debug and verification */
662 #ifdef CONFIG_DEBUG_MEMORY_INIT
664 extern int mminit_loglevel;
666 #define mminit_dprintk(level, prefix, fmt, arg...) \
668 if (level < mminit_loglevel) { \
669 if (level <= MMINIT_WARNING) \
670 pr_warn("mminit::" prefix " " fmt, ##arg); \
672 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
676 extern void mminit_verify_pageflags_layout(void);
677 extern void mminit_verify_zonelist(void);
680 static inline void mminit_dprintk(enum mminit_level level,
681 const char *prefix, const char *fmt, ...)
685 static inline void mminit_verify_pageflags_layout(void)
689 static inline void mminit_verify_zonelist(void)
692 #endif /* CONFIG_DEBUG_MEMORY_INIT */
694 #define NODE_RECLAIM_NOSCAN -2
695 #define NODE_RECLAIM_FULL -1
696 #define NODE_RECLAIM_SOME 0
697 #define NODE_RECLAIM_SUCCESS 1
700 extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int);
701 extern int find_next_best_node(int node, nodemask_t *used_node_mask);
703 static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask,
706 return NODE_RECLAIM_NOSCAN;
708 static inline int find_next_best_node(int node, nodemask_t *used_node_mask)
715 * mm/memory-failure.c
717 extern int hwpoison_filter(struct page *p);
719 extern u32 hwpoison_filter_dev_major;
720 extern u32 hwpoison_filter_dev_minor;
721 extern u64 hwpoison_filter_flags_mask;
722 extern u64 hwpoison_filter_flags_value;
723 extern u64 hwpoison_filter_memcg;
724 extern u32 hwpoison_filter_enable;
726 #ifdef CONFIG_MEMORY_FAILURE
727 void clear_hwpoisoned_pages(struct page *memmap, int nr_pages);
729 static inline void clear_hwpoisoned_pages(struct page *memmap, int nr_pages)
734 extern unsigned long __must_check vm_mmap_pgoff(struct file *, unsigned long,
735 unsigned long, unsigned long,
736 unsigned long, unsigned long);
738 extern void set_pageblock_order(void);
739 unsigned int reclaim_clean_pages_from_list(struct zone *zone,
740 struct list_head *page_list);
741 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
742 #define ALLOC_WMARK_MIN WMARK_MIN
743 #define ALLOC_WMARK_LOW WMARK_LOW
744 #define ALLOC_WMARK_HIGH WMARK_HIGH
745 #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
747 /* Mask to get the watermark bits */
748 #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
751 * Only MMU archs have async oom victim reclaim - aka oom_reaper so we
752 * cannot assume a reduced access to memory reserves is sufficient for
756 #define ALLOC_OOM 0x08
758 #define ALLOC_OOM ALLOC_NO_WATERMARKS
761 #define ALLOC_HARDER 0x10 /* try to alloc harder */
762 #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
763 #define ALLOC_CPUSET 0x40 /* check for correct cpuset */
764 #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
765 #ifdef CONFIG_ZONE_DMA32
766 #define ALLOC_NOFRAGMENT 0x100 /* avoid mixing pageblock types */
768 #define ALLOC_NOFRAGMENT 0x0
770 #define ALLOC_KSWAPD 0x800 /* allow waking of kswapd, __GFP_KSWAPD_RECLAIM set */
773 struct tlbflush_unmap_batch;
777 * only for MM internal work items which do not depend on
778 * any allocations or locks which might depend on allocations
780 extern struct workqueue_struct *mm_percpu_wq;
782 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
783 void try_to_unmap_flush(void);
784 void try_to_unmap_flush_dirty(void);
785 void flush_tlb_batched_pending(struct mm_struct *mm);
787 static inline void try_to_unmap_flush(void)
790 static inline void try_to_unmap_flush_dirty(void)
793 static inline void flush_tlb_batched_pending(struct mm_struct *mm)
796 #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
798 extern const struct trace_print_flags pageflag_names[];
799 extern const struct trace_print_flags vmaflag_names[];
800 extern const struct trace_print_flags gfpflag_names[];
802 static inline bool is_migrate_highatomic(enum migratetype migratetype)
804 return migratetype == MIGRATE_HIGHATOMIC;
807 static inline bool is_migrate_highatomic_page(struct page *page)
809 return get_pageblock_migratetype(page) == MIGRATE_HIGHATOMIC;
812 void setup_zone_pageset(struct zone *zone);
814 struct migration_target_control {
815 int nid; /* preferred node id */
824 int vmap_pages_range_noflush(unsigned long addr, unsigned long end,
825 pgprot_t prot, struct page **pages, unsigned int page_shift);
828 int vmap_pages_range_noflush(unsigned long addr, unsigned long end,
829 pgprot_t prot, struct page **pages, unsigned int page_shift)
835 void vunmap_range_noflush(unsigned long start, unsigned long end);
837 int numa_migrate_prep(struct page *page, struct vm_area_struct *vma,
838 unsigned long addr, int page_nid, int *flags);
840 void free_zone_device_page(struct page *page);
845 struct folio *try_grab_folio(struct page *page, int refs, unsigned int flags);
847 DECLARE_PER_CPU(struct per_cpu_nodestat, boot_nodestats);
849 #endif /* __MM_INTERNAL_H */