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/tracepoint-defs.h>
16 * The set of flags that only affect watermark checking and reclaim
17 * behaviour. This is used by the MM to obey the caller constraints
18 * about IO, FS and watermark checking while ignoring placement
19 * hints such as HIGHMEM usage.
21 #define GFP_RECLAIM_MASK (__GFP_RECLAIM|__GFP_HIGH|__GFP_IO|__GFP_FS|\
22 __GFP_NOWARN|__GFP_RETRY_MAYFAIL|__GFP_NOFAIL|\
23 __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC|\
26 /* The GFP flags allowed during early boot */
27 #define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_RECLAIM|__GFP_IO|__GFP_FS))
29 /* Control allocation cpuset and node placement constraints */
30 #define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE)
32 /* Do not use these with a slab allocator */
33 #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
35 void page_writeback_init(void);
37 vm_fault_t do_swap_page(struct vm_fault *vmf);
39 void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
40 unsigned long floor, unsigned long ceiling);
42 static inline bool can_madv_lru_vma(struct vm_area_struct *vma)
44 return !(vma->vm_flags & (VM_LOCKED|VM_HUGETLB|VM_PFNMAP));
47 void unmap_page_range(struct mmu_gather *tlb,
48 struct vm_area_struct *vma,
49 unsigned long addr, unsigned long end,
50 struct zap_details *details);
52 void do_page_cache_ra(struct readahead_control *, unsigned long nr_to_read,
53 unsigned long lookahead_size);
54 void force_page_cache_ra(struct readahead_control *, unsigned long nr);
55 static inline void force_page_cache_readahead(struct address_space *mapping,
56 struct file *file, pgoff_t index, unsigned long nr_to_read)
58 DEFINE_READAHEAD(ractl, file, &file->f_ra, mapping, index);
59 force_page_cache_ra(&ractl, nr_to_read);
62 unsigned find_lock_entries(struct address_space *mapping, pgoff_t start,
63 pgoff_t end, struct pagevec *pvec, pgoff_t *indices);
66 * page_evictable - test whether a page is evictable
67 * @page: the page to test
69 * Test whether page is evictable--i.e., should be placed on active/inactive
70 * lists vs unevictable list.
72 * Reasons page might not be evictable:
73 * (1) page's mapping marked unevictable
74 * (2) page is part of an mlocked VMA
77 static inline bool page_evictable(struct page *page)
81 /* Prevent address_space of inode and swap cache from being freed */
83 ret = !mapping_unevictable(page_mapping(page)) && !PageMlocked(page);
89 * Turn a non-refcounted page (->_refcount == 0) into refcounted with
92 static inline void set_page_refcounted(struct page *page)
94 VM_BUG_ON_PAGE(PageTail(page), page);
95 VM_BUG_ON_PAGE(page_ref_count(page), page);
96 set_page_count(page, 1);
99 extern unsigned long highest_memmap_pfn;
102 * Maximum number of reclaim retries without progress before the OOM
103 * killer is consider the only way forward.
105 #define MAX_RECLAIM_RETRIES 16
110 extern int isolate_lru_page(struct page *page);
111 extern void putback_lru_page(struct page *page);
116 extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
119 * in mm/memcontrol.c:
121 extern bool cgroup_memory_nokmem;
128 * Structure for holding the mostly immutable allocation parameters passed
129 * between functions involved in allocations, including the alloc_pages*
130 * family of functions.
132 * nodemask, migratetype and highest_zoneidx are initialized only once in
133 * __alloc_pages() and then never change.
135 * zonelist, preferred_zone and highest_zoneidx are set first in
136 * __alloc_pages() for the fast path, and might be later changed
137 * in __alloc_pages_slowpath(). All other functions pass the whole structure
138 * by a const pointer.
140 struct alloc_context {
141 struct zonelist *zonelist;
142 nodemask_t *nodemask;
143 struct zoneref *preferred_zoneref;
147 * highest_zoneidx represents highest usable zone index of
148 * the allocation request. Due to the nature of the zone,
149 * memory on lower zone than the highest_zoneidx will be
150 * protected by lowmem_reserve[highest_zoneidx].
152 * highest_zoneidx is also used by reclaim/compaction to limit
153 * the target zone since higher zone than this index cannot be
154 * usable for this allocation request.
156 enum zone_type highest_zoneidx;
157 bool spread_dirty_pages;
161 * Locate the struct page for both the matching buddy in our
162 * pair (buddy1) and the combined O(n+1) page they form (page).
164 * 1) Any buddy B1 will have an order O twin B2 which satisfies
165 * the following equation:
167 * For example, if the starting buddy (buddy2) is #8 its order
169 * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
171 * 2) Any buddy B will have an order O+1 parent P which
172 * satisfies the following equation:
175 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
177 static inline unsigned long
178 __find_buddy_pfn(unsigned long page_pfn, unsigned int order)
180 return page_pfn ^ (1 << order);
183 extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn,
184 unsigned long end_pfn, struct zone *zone);
186 static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn,
187 unsigned long end_pfn, struct zone *zone)
189 if (zone->contiguous)
190 return pfn_to_page(start_pfn);
192 return __pageblock_pfn_to_page(start_pfn, end_pfn, zone);
195 extern int __isolate_free_page(struct page *page, unsigned int order);
196 extern void __putback_isolated_page(struct page *page, unsigned int order,
198 extern void memblock_free_pages(struct page *page, unsigned long pfn,
200 extern void __free_pages_core(struct page *page, unsigned int order);
201 extern void prep_compound_page(struct page *page, unsigned int order);
202 extern void post_alloc_hook(struct page *page, unsigned int order,
204 extern int user_min_free_kbytes;
206 extern void free_unref_page(struct page *page, unsigned int order);
207 extern void free_unref_page_list(struct list_head *list);
209 extern void zone_pcp_update(struct zone *zone, int cpu_online);
210 extern void zone_pcp_reset(struct zone *zone);
211 extern void zone_pcp_disable(struct zone *zone);
212 extern void zone_pcp_enable(struct zone *zone);
214 extern void *memmap_alloc(phys_addr_t size, phys_addr_t align,
215 phys_addr_t min_addr,
216 int nid, bool exact_nid);
218 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
224 * compact_control is used to track pages being migrated and the free pages
225 * they are being migrated to during memory compaction. The free_pfn starts
226 * at the end of a zone and migrate_pfn begins at the start. Movable pages
227 * are moved to the end of a zone during a compaction run and the run
228 * completes when free_pfn <= migrate_pfn
230 struct compact_control {
231 struct list_head freepages; /* List of free pages to migrate to */
232 struct list_head migratepages; /* List of pages being migrated */
233 unsigned int nr_freepages; /* Number of isolated free pages */
234 unsigned int nr_migratepages; /* Number of pages to migrate */
235 unsigned long free_pfn; /* isolate_freepages search base */
237 * Acts as an in/out parameter to page isolation for migration.
238 * isolate_migratepages uses it as a search base.
239 * isolate_migratepages_block will update the value to the next pfn
240 * after the last isolated one.
242 unsigned long migrate_pfn;
243 unsigned long fast_start_pfn; /* a pfn to start linear scan from */
245 unsigned long total_migrate_scanned;
246 unsigned long total_free_scanned;
247 unsigned short fast_search_fail;/* failures to use free list searches */
248 short search_order; /* order to start a fast search at */
249 const gfp_t gfp_mask; /* gfp mask of a direct compactor */
250 int order; /* order a direct compactor needs */
251 int migratetype; /* migratetype of direct compactor */
252 const unsigned int alloc_flags; /* alloc flags of a direct compactor */
253 const int highest_zoneidx; /* zone index of a direct compactor */
254 enum migrate_mode mode; /* Async or sync migration mode */
255 bool ignore_skip_hint; /* Scan blocks even if marked skip */
256 bool no_set_skip_hint; /* Don't mark blocks for skipping */
257 bool ignore_block_suitable; /* Scan blocks considered unsuitable */
258 bool direct_compaction; /* False from kcompactd or /proc/... */
259 bool proactive_compaction; /* kcompactd proactive compaction */
260 bool whole_zone; /* Whole zone should/has been scanned */
261 bool contended; /* Signal lock or sched contention */
262 bool rescan; /* Rescanning the same pageblock */
263 bool alloc_contig; /* alloc_contig_range allocation */
267 * Used in direct compaction when a page should be taken from the freelists
268 * immediately when one is created during the free path.
270 struct capture_control {
271 struct compact_control *cc;
276 isolate_freepages_range(struct compact_control *cc,
277 unsigned long start_pfn, unsigned long end_pfn);
279 isolate_migratepages_range(struct compact_control *cc,
280 unsigned long low_pfn, unsigned long end_pfn);
282 int find_suitable_fallback(struct free_area *area, unsigned int order,
283 int migratetype, bool only_stealable, bool *can_steal);
286 * This function returns the order of a free page in the buddy system. In
287 * general, page_zone(page)->lock must be held by the caller to prevent the
288 * page from being allocated in parallel and returning garbage as the order.
289 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
290 * page cannot be allocated or merged in parallel. Alternatively, it must
291 * handle invalid values gracefully, and use buddy_order_unsafe() below.
293 static inline unsigned int buddy_order(struct page *page)
295 /* PageBuddy() must be checked by the caller */
296 return page_private(page);
300 * Like buddy_order(), but for callers who cannot afford to hold the zone lock.
301 * PageBuddy() should be checked first by the caller to minimize race window,
302 * and invalid values must be handled gracefully.
304 * READ_ONCE is used so that if the caller assigns the result into a local
305 * variable and e.g. tests it for valid range before using, the compiler cannot
306 * decide to remove the variable and inline the page_private(page) multiple
307 * times, potentially observing different values in the tests and the actual
310 #define buddy_order_unsafe(page) READ_ONCE(page_private(page))
313 * These three helpers classifies VMAs for virtual memory accounting.
317 * Executable code area - executable, not writable, not stack
319 static inline bool is_exec_mapping(vm_flags_t flags)
321 return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC;
325 * Stack area - automatically grows in one direction
327 * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
328 * do_mmap() forbids all other combinations.
330 static inline bool is_stack_mapping(vm_flags_t flags)
332 return (flags & VM_STACK) == VM_STACK;
336 * Data area - private, writable, not stack
338 static inline bool is_data_mapping(vm_flags_t flags)
340 return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE;
344 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
345 struct vm_area_struct *prev);
346 void __vma_unlink_list(struct mm_struct *mm, struct vm_area_struct *vma);
349 extern long populate_vma_page_range(struct vm_area_struct *vma,
350 unsigned long start, unsigned long end, int *locked);
351 extern long faultin_vma_page_range(struct vm_area_struct *vma,
352 unsigned long start, unsigned long end,
353 bool write, int *locked);
354 extern void munlock_vma_pages_range(struct vm_area_struct *vma,
355 unsigned long start, unsigned long end);
356 static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
358 munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
362 * must be called with vma's mmap_lock held for read or write, and page locked.
364 extern void mlock_vma_page(struct page *page);
365 extern unsigned int munlock_vma_page(struct page *page);
367 extern int mlock_future_check(struct mm_struct *mm, unsigned long flags,
371 * Clear the page's PageMlocked(). This can be useful in a situation where
372 * we want to unconditionally remove a page from the pagecache -- e.g.,
373 * on truncation or freeing.
375 * It is legal to call this function for any page, mlocked or not.
376 * If called for a page that is still mapped by mlocked vmas, all we do
377 * is revert to lazy LRU behaviour -- semantics are not broken.
379 extern void clear_page_mlock(struct page *page);
381 extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
384 * At what user virtual address is page expected in vma?
385 * Returns -EFAULT if all of the page is outside the range of vma.
386 * If page is a compound head, the entire compound page is considered.
388 static inline unsigned long
389 vma_address(struct page *page, struct vm_area_struct *vma)
392 unsigned long address;
394 VM_BUG_ON_PAGE(PageKsm(page), page); /* KSM page->index unusable */
395 pgoff = page_to_pgoff(page);
396 if (pgoff >= vma->vm_pgoff) {
397 address = vma->vm_start +
398 ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
399 /* Check for address beyond vma (or wrapped through 0?) */
400 if (address < vma->vm_start || address >= vma->vm_end)
402 } else if (PageHead(page) &&
403 pgoff + compound_nr(page) - 1 >= vma->vm_pgoff) {
404 /* Test above avoids possibility of wrap to 0 on 32-bit */
405 address = vma->vm_start;
413 * Then at what user virtual address will none of the page be found in vma?
414 * Assumes that vma_address() already returned a good starting address.
415 * If page is a compound head, the entire compound page is considered.
417 static inline unsigned long
418 vma_address_end(struct page *page, struct vm_area_struct *vma)
421 unsigned long address;
423 VM_BUG_ON_PAGE(PageKsm(page), page); /* KSM page->index unusable */
424 pgoff = page_to_pgoff(page) + compound_nr(page);
425 address = vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
426 /* Check for address beyond vma (or wrapped through 0?) */
427 if (address < vma->vm_start || address > vma->vm_end)
428 address = vma->vm_end;
432 static inline struct file *maybe_unlock_mmap_for_io(struct vm_fault *vmf,
435 int flags = vmf->flags;
441 * FAULT_FLAG_RETRY_NOWAIT means we don't want to wait on page locks or
442 * anything, so we only pin the file and drop the mmap_lock if only
443 * FAULT_FLAG_ALLOW_RETRY is set, while this is the first attempt.
445 if (fault_flag_allow_retry_first(flags) &&
446 !(flags & FAULT_FLAG_RETRY_NOWAIT)) {
447 fpin = get_file(vmf->vma->vm_file);
448 mmap_read_unlock(vmf->vma->vm_mm);
453 #else /* !CONFIG_MMU */
454 static inline void clear_page_mlock(struct page *page) { }
455 static inline void mlock_vma_page(struct page *page) { }
456 static inline void vunmap_range_noflush(unsigned long start, unsigned long end)
459 #endif /* !CONFIG_MMU */
462 * Return the mem_map entry representing the 'offset' subpage within
463 * the maximally aligned gigantic page 'base'. Handle any discontiguity
464 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
466 static inline struct page *mem_map_offset(struct page *base, int offset)
468 if (unlikely(offset >= MAX_ORDER_NR_PAGES))
469 return nth_page(base, offset);
470 return base + offset;
474 * Iterator over all subpages within the maximally aligned gigantic
475 * page 'base'. Handle any discontiguity in the mem_map.
477 static inline struct page *mem_map_next(struct page *iter,
478 struct page *base, int offset)
480 if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
481 unsigned long pfn = page_to_pfn(base) + offset;
484 return pfn_to_page(pfn);
489 /* Memory initialisation debug and verification */
496 #ifdef CONFIG_DEBUG_MEMORY_INIT
498 extern int mminit_loglevel;
500 #define mminit_dprintk(level, prefix, fmt, arg...) \
502 if (level < mminit_loglevel) { \
503 if (level <= MMINIT_WARNING) \
504 pr_warn("mminit::" prefix " " fmt, ##arg); \
506 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
510 extern void mminit_verify_pageflags_layout(void);
511 extern void mminit_verify_zonelist(void);
514 static inline void mminit_dprintk(enum mminit_level level,
515 const char *prefix, const char *fmt, ...)
519 static inline void mminit_verify_pageflags_layout(void)
523 static inline void mminit_verify_zonelist(void)
526 #endif /* CONFIG_DEBUG_MEMORY_INIT */
528 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
529 #if defined(CONFIG_SPARSEMEM)
530 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
531 unsigned long *end_pfn);
533 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
534 unsigned long *end_pfn)
537 #endif /* CONFIG_SPARSEMEM */
539 #define NODE_RECLAIM_NOSCAN -2
540 #define NODE_RECLAIM_FULL -1
541 #define NODE_RECLAIM_SOME 0
542 #define NODE_RECLAIM_SUCCESS 1
545 extern int node_reclaim(struct pglist_data *, gfp_t, unsigned int);
546 extern int find_next_best_node(int node, nodemask_t *used_node_mask);
548 static inline int node_reclaim(struct pglist_data *pgdat, gfp_t mask,
551 return NODE_RECLAIM_NOSCAN;
553 static inline int find_next_best_node(int node, nodemask_t *used_node_mask)
559 extern int hwpoison_filter(struct page *p);
561 extern u32 hwpoison_filter_dev_major;
562 extern u32 hwpoison_filter_dev_minor;
563 extern u64 hwpoison_filter_flags_mask;
564 extern u64 hwpoison_filter_flags_value;
565 extern u64 hwpoison_filter_memcg;
566 extern u32 hwpoison_filter_enable;
568 extern unsigned long __must_check vm_mmap_pgoff(struct file *, unsigned long,
569 unsigned long, unsigned long,
570 unsigned long, unsigned long);
572 extern void set_pageblock_order(void);
573 unsigned int reclaim_clean_pages_from_list(struct zone *zone,
574 struct list_head *page_list);
575 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
576 #define ALLOC_WMARK_MIN WMARK_MIN
577 #define ALLOC_WMARK_LOW WMARK_LOW
578 #define ALLOC_WMARK_HIGH WMARK_HIGH
579 #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
581 /* Mask to get the watermark bits */
582 #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
585 * Only MMU archs have async oom victim reclaim - aka oom_reaper so we
586 * cannot assume a reduced access to memory reserves is sufficient for
590 #define ALLOC_OOM 0x08
592 #define ALLOC_OOM ALLOC_NO_WATERMARKS
595 #define ALLOC_HARDER 0x10 /* try to alloc harder */
596 #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
597 #define ALLOC_CPUSET 0x40 /* check for correct cpuset */
598 #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
599 #ifdef CONFIG_ZONE_DMA32
600 #define ALLOC_NOFRAGMENT 0x100 /* avoid mixing pageblock types */
602 #define ALLOC_NOFRAGMENT 0x0
604 #define ALLOC_KSWAPD 0x800 /* allow waking of kswapd, __GFP_KSWAPD_RECLAIM set */
607 struct tlbflush_unmap_batch;
611 * only for MM internal work items which do not depend on
612 * any allocations or locks which might depend on allocations
614 extern struct workqueue_struct *mm_percpu_wq;
616 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
617 void try_to_unmap_flush(void);
618 void try_to_unmap_flush_dirty(void);
619 void flush_tlb_batched_pending(struct mm_struct *mm);
621 static inline void try_to_unmap_flush(void)
624 static inline void try_to_unmap_flush_dirty(void)
627 static inline void flush_tlb_batched_pending(struct mm_struct *mm)
630 #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
632 extern const struct trace_print_flags pageflag_names[];
633 extern const struct trace_print_flags vmaflag_names[];
634 extern const struct trace_print_flags gfpflag_names[];
636 static inline bool is_migrate_highatomic(enum migratetype migratetype)
638 return migratetype == MIGRATE_HIGHATOMIC;
641 static inline bool is_migrate_highatomic_page(struct page *page)
643 return get_pageblock_migratetype(page) == MIGRATE_HIGHATOMIC;
646 void setup_zone_pageset(struct zone *zone);
648 struct migration_target_control {
649 int nid; /* preferred node id */
658 int vmap_pages_range_noflush(unsigned long addr, unsigned long end,
659 pgprot_t prot, struct page **pages, unsigned int page_shift);
662 int vmap_pages_range_noflush(unsigned long addr, unsigned long end,
663 pgprot_t prot, struct page **pages, unsigned int page_shift)
669 void vunmap_range_noflush(unsigned long start, unsigned long end);
671 int numa_migrate_prep(struct page *page, struct vm_area_struct *vma,
672 unsigned long addr, int page_nid, int *flags);
674 #endif /* __MM_INTERNAL_H */