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
);
123 * Structure for holding the mostly immutable allocation parameters passed
124 * between functions involved in allocations, including the alloc_pages*
125 * family of functions.
127 * nodemask, migratetype and highest_zoneidx are initialized only once in
128 * __alloc_pages() and then never change.
130 * zonelist, preferred_zone and highest_zoneidx are set first in
131 * __alloc_pages() for the fast path, and might be later changed
132 * in __alloc_pages_slowpath(). All other functions pass the whole structure
133 * by a const pointer.
135 struct alloc_context
{
136 struct zonelist
*zonelist
;
137 nodemask_t
*nodemask
;
138 struct zoneref
*preferred_zoneref
;
142 * highest_zoneidx represents highest usable zone index of
143 * the allocation request. Due to the nature of the zone,
144 * memory on lower zone than the highest_zoneidx will be
145 * protected by lowmem_reserve[highest_zoneidx].
147 * highest_zoneidx is also used by reclaim/compaction to limit
148 * the target zone since higher zone than this index cannot be
149 * usable for this allocation request.
151 enum zone_type highest_zoneidx
;
152 bool spread_dirty_pages
;
156 * Locate the struct page for both the matching buddy in our
157 * pair (buddy1) and the combined O(n+1) page they form (page).
159 * 1) Any buddy B1 will have an order O twin B2 which satisfies
160 * the following equation:
162 * For example, if the starting buddy (buddy2) is #8 its order
164 * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
166 * 2) Any buddy B will have an order O+1 parent P which
167 * satisfies the following equation:
170 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
172 static inline unsigned long
173 __find_buddy_pfn(unsigned long page_pfn
, unsigned int order
)
175 return page_pfn
^ (1 << order
);
178 extern struct page
*__pageblock_pfn_to_page(unsigned long start_pfn
,
179 unsigned long end_pfn
, struct zone
*zone
);
181 static inline struct page
*pageblock_pfn_to_page(unsigned long start_pfn
,
182 unsigned long end_pfn
, struct zone
*zone
)
184 if (zone
->contiguous
)
185 return pfn_to_page(start_pfn
);
187 return __pageblock_pfn_to_page(start_pfn
, end_pfn
, zone
);
190 extern int __isolate_free_page(struct page
*page
, unsigned int order
);
191 extern void __putback_isolated_page(struct page
*page
, unsigned int order
,
193 extern void memblock_free_pages(struct page
*page
, unsigned long pfn
,
195 extern void __free_pages_core(struct page
*page
, unsigned int order
);
196 extern void prep_compound_page(struct page
*page
, unsigned int order
);
197 extern void post_alloc_hook(struct page
*page
, unsigned int order
,
199 extern int user_min_free_kbytes
;
201 extern void free_unref_page(struct page
*page
);
202 extern void free_unref_page_list(struct list_head
*list
);
204 extern void zone_pcp_update(struct zone
*zone
);
205 extern void zone_pcp_reset(struct zone
*zone
);
206 extern void zone_pcp_disable(struct zone
*zone
);
207 extern void zone_pcp_enable(struct zone
*zone
);
209 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
215 * compact_control is used to track pages being migrated and the free pages
216 * they are being migrated to during memory compaction. The free_pfn starts
217 * at the end of a zone and migrate_pfn begins at the start. Movable pages
218 * are moved to the end of a zone during a compaction run and the run
219 * completes when free_pfn <= migrate_pfn
221 struct compact_control
{
222 struct list_head freepages
; /* List of free pages to migrate to */
223 struct list_head migratepages
; /* List of pages being migrated */
224 unsigned int nr_freepages
; /* Number of isolated free pages */
225 unsigned int nr_migratepages
; /* Number of pages to migrate */
226 unsigned long free_pfn
; /* isolate_freepages search base */
228 * Acts as an in/out parameter to page isolation for migration.
229 * isolate_migratepages uses it as a search base.
230 * isolate_migratepages_block will update the value to the next pfn
231 * after the last isolated one.
233 unsigned long migrate_pfn
;
234 unsigned long fast_start_pfn
; /* a pfn to start linear scan from */
236 unsigned long total_migrate_scanned
;
237 unsigned long total_free_scanned
;
238 unsigned short fast_search_fail
;/* failures to use free list searches */
239 short search_order
; /* order to start a fast search at */
240 const gfp_t gfp_mask
; /* gfp mask of a direct compactor */
241 int order
; /* order a direct compactor needs */
242 int migratetype
; /* migratetype of direct compactor */
243 const unsigned int alloc_flags
; /* alloc flags of a direct compactor */
244 const int highest_zoneidx
; /* zone index of a direct compactor */
245 enum migrate_mode mode
; /* Async or sync migration mode */
246 bool ignore_skip_hint
; /* Scan blocks even if marked skip */
247 bool no_set_skip_hint
; /* Don't mark blocks for skipping */
248 bool ignore_block_suitable
; /* Scan blocks considered unsuitable */
249 bool direct_compaction
; /* False from kcompactd or /proc/... */
250 bool proactive_compaction
; /* kcompactd proactive compaction */
251 bool whole_zone
; /* Whole zone should/has been scanned */
252 bool contended
; /* Signal lock or sched contention */
253 bool rescan
; /* Rescanning the same pageblock */
254 bool alloc_contig
; /* alloc_contig_range allocation */
258 * Used in direct compaction when a page should be taken from the freelists
259 * immediately when one is created during the free path.
261 struct capture_control
{
262 struct compact_control
*cc
;
267 isolate_freepages_range(struct compact_control
*cc
,
268 unsigned long start_pfn
, unsigned long end_pfn
);
270 isolate_migratepages_range(struct compact_control
*cc
,
271 unsigned long low_pfn
, unsigned long end_pfn
);
272 int find_suitable_fallback(struct free_area
*area
, unsigned int order
,
273 int migratetype
, bool only_stealable
, bool *can_steal
);
278 * This function returns the order of a free page in the buddy system. In
279 * general, page_zone(page)->lock must be held by the caller to prevent the
280 * page from being allocated in parallel and returning garbage as the order.
281 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
282 * page cannot be allocated or merged in parallel. Alternatively, it must
283 * handle invalid values gracefully, and use buddy_order_unsafe() below.
285 static inline unsigned int buddy_order(struct page
*page
)
287 /* PageBuddy() must be checked by the caller */
288 return page_private(page
);
292 * Like buddy_order(), but for callers who cannot afford to hold the zone lock.
293 * PageBuddy() should be checked first by the caller to minimize race window,
294 * and invalid values must be handled gracefully.
296 * READ_ONCE is used so that if the caller assigns the result into a local
297 * variable and e.g. tests it for valid range before using, the compiler cannot
298 * decide to remove the variable and inline the page_private(page) multiple
299 * times, potentially observing different values in the tests and the actual
302 #define buddy_order_unsafe(page) READ_ONCE(page_private(page))
305 * These three helpers classifies VMAs for virtual memory accounting.
309 * Executable code area - executable, not writable, not stack
311 static inline bool is_exec_mapping(vm_flags_t flags
)
313 return (flags
& (VM_EXEC
| VM_WRITE
| VM_STACK
)) == VM_EXEC
;
317 * Stack area - automatically grows in one direction
319 * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
320 * do_mmap() forbids all other combinations.
322 static inline bool is_stack_mapping(vm_flags_t flags
)
324 return (flags
& VM_STACK
) == VM_STACK
;
328 * Data area - private, writable, not stack
330 static inline bool is_data_mapping(vm_flags_t flags
)
332 return (flags
& (VM_WRITE
| VM_SHARED
| VM_STACK
)) == VM_WRITE
;
336 void __vma_link_list(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
337 struct vm_area_struct
*prev
);
338 void __vma_unlink_list(struct mm_struct
*mm
, struct vm_area_struct
*vma
);
341 extern long populate_vma_page_range(struct vm_area_struct
*vma
,
342 unsigned long start
, unsigned long end
, int *nonblocking
);
343 extern void munlock_vma_pages_range(struct vm_area_struct
*vma
,
344 unsigned long start
, unsigned long end
);
345 static inline void munlock_vma_pages_all(struct vm_area_struct
*vma
)
347 munlock_vma_pages_range(vma
, vma
->vm_start
, vma
->vm_end
);
351 * must be called with vma's mmap_lock held for read or write, and page locked.
353 extern void mlock_vma_page(struct page
*page
);
354 extern unsigned int munlock_vma_page(struct page
*page
);
357 * Clear the page's PageMlocked(). This can be useful in a situation where
358 * we want to unconditionally remove a page from the pagecache -- e.g.,
359 * on truncation or freeing.
361 * It is legal to call this function for any page, mlocked or not.
362 * If called for a page that is still mapped by mlocked vmas, all we do
363 * is revert to lazy LRU behaviour -- semantics are not broken.
365 extern void clear_page_mlock(struct page
*page
);
368 * mlock_migrate_page - called only from migrate_misplaced_transhuge_page()
369 * (because that does not go through the full procedure of migration ptes):
370 * to migrate the Mlocked page flag; update statistics.
372 static inline void mlock_migrate_page(struct page
*newpage
, struct page
*page
)
374 if (TestClearPageMlocked(page
)) {
375 int nr_pages
= thp_nr_pages(page
);
377 /* Holding pmd lock, no change in irq context: __mod is safe */
378 __mod_zone_page_state(page_zone(page
), NR_MLOCK
, -nr_pages
);
379 SetPageMlocked(newpage
);
380 __mod_zone_page_state(page_zone(newpage
), NR_MLOCK
, nr_pages
);
384 extern pmd_t
maybe_pmd_mkwrite(pmd_t pmd
, struct vm_area_struct
*vma
);
387 * At what user virtual address is page expected in vma?
388 * Returns -EFAULT if all of the page is outside the range of vma.
389 * If page is a compound head, the entire compound page is considered.
391 static inline unsigned long
392 vma_address(struct page
*page
, struct vm_area_struct
*vma
)
395 unsigned long address
;
397 VM_BUG_ON_PAGE(PageKsm(page
), page
); /* KSM page->index unusable */
398 pgoff
= page_to_pgoff(page
);
399 if (pgoff
>= vma
->vm_pgoff
) {
400 address
= vma
->vm_start
+
401 ((pgoff
- vma
->vm_pgoff
) << PAGE_SHIFT
);
402 /* Check for address beyond vma (or wrapped through 0?) */
403 if (address
< vma
->vm_start
|| address
>= vma
->vm_end
)
405 } else if (PageHead(page
) &&
406 pgoff
+ compound_nr(page
) - 1 >= vma
->vm_pgoff
) {
407 /* Test above avoids possibility of wrap to 0 on 32-bit */
408 address
= vma
->vm_start
;
416 * Then at what user virtual address will none of the page be found in vma?
417 * Assumes that vma_address() already returned a good starting address.
418 * If page is a compound head, the entire compound page is considered.
420 static inline unsigned long
421 vma_address_end(struct page
*page
, struct vm_area_struct
*vma
)
424 unsigned long address
;
426 VM_BUG_ON_PAGE(PageKsm(page
), page
); /* KSM page->index unusable */
427 pgoff
= page_to_pgoff(page
) + compound_nr(page
);
428 address
= vma
->vm_start
+ ((pgoff
- vma
->vm_pgoff
) << PAGE_SHIFT
);
429 /* Check for address beyond vma (or wrapped through 0?) */
430 if (address
< vma
->vm_start
|| address
> vma
->vm_end
)
431 address
= vma
->vm_end
;
435 static inline struct file
*maybe_unlock_mmap_for_io(struct vm_fault
*vmf
,
438 int flags
= vmf
->flags
;
444 * FAULT_FLAG_RETRY_NOWAIT means we don't want to wait on page locks or
445 * anything, so we only pin the file and drop the mmap_lock if only
446 * FAULT_FLAG_ALLOW_RETRY is set, while this is the first attempt.
448 if (fault_flag_allow_retry_first(flags
) &&
449 !(flags
& FAULT_FLAG_RETRY_NOWAIT
)) {
450 fpin
= get_file(vmf
->vma
->vm_file
);
451 mmap_read_unlock(vmf
->vma
->vm_mm
);
456 #else /* !CONFIG_MMU */
457 static inline void clear_page_mlock(struct page
*page
) { }
458 static inline void mlock_vma_page(struct page
*page
) { }
459 static inline void mlock_migrate_page(struct page
*new, struct page
*old
) { }
460 static inline void vunmap_range_noflush(unsigned long start
, unsigned long end
)
463 #endif /* !CONFIG_MMU */
466 * Return the mem_map entry representing the 'offset' subpage within
467 * the maximally aligned gigantic page 'base'. Handle any discontiguity
468 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
470 static inline struct page
*mem_map_offset(struct page
*base
, int offset
)
472 if (unlikely(offset
>= MAX_ORDER_NR_PAGES
))
473 return nth_page(base
, offset
);
474 return base
+ offset
;
478 * Iterator over all subpages within the maximally aligned gigantic
479 * page 'base'. Handle any discontiguity in the mem_map.
481 static inline struct page
*mem_map_next(struct page
*iter
,
482 struct page
*base
, int offset
)
484 if (unlikely((offset
& (MAX_ORDER_NR_PAGES
- 1)) == 0)) {
485 unsigned long pfn
= page_to_pfn(base
) + offset
;
488 return pfn_to_page(pfn
);
493 /* Memory initialisation debug and verification */
500 #ifdef CONFIG_DEBUG_MEMORY_INIT
502 extern int mminit_loglevel
;
504 #define mminit_dprintk(level, prefix, fmt, arg...) \
506 if (level < mminit_loglevel) { \
507 if (level <= MMINIT_WARNING) \
508 pr_warn("mminit::" prefix " " fmt, ##arg); \
510 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
514 extern void mminit_verify_pageflags_layout(void);
515 extern void mminit_verify_zonelist(void);
518 static inline void mminit_dprintk(enum mminit_level level
,
519 const char *prefix
, const char *fmt
, ...)
523 static inline void mminit_verify_pageflags_layout(void)
527 static inline void mminit_verify_zonelist(void)
530 #endif /* CONFIG_DEBUG_MEMORY_INIT */
532 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
533 #if defined(CONFIG_SPARSEMEM)
534 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn
,
535 unsigned long *end_pfn
);
537 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn
,
538 unsigned long *end_pfn
)
541 #endif /* CONFIG_SPARSEMEM */
543 #define NODE_RECLAIM_NOSCAN -2
544 #define NODE_RECLAIM_FULL -1
545 #define NODE_RECLAIM_SOME 0
546 #define NODE_RECLAIM_SUCCESS 1
549 extern int node_reclaim(struct pglist_data
*, gfp_t
, unsigned int);
551 static inline int node_reclaim(struct pglist_data
*pgdat
, gfp_t mask
,
554 return NODE_RECLAIM_NOSCAN
;
558 extern int hwpoison_filter(struct page
*p
);
560 extern u32 hwpoison_filter_dev_major
;
561 extern u32 hwpoison_filter_dev_minor
;
562 extern u64 hwpoison_filter_flags_mask
;
563 extern u64 hwpoison_filter_flags_value
;
564 extern u64 hwpoison_filter_memcg
;
565 extern u32 hwpoison_filter_enable
;
567 extern unsigned long __must_check
vm_mmap_pgoff(struct file
*, unsigned long,
568 unsigned long, unsigned long,
569 unsigned long, unsigned long);
571 extern void set_pageblock_order(void);
572 unsigned int reclaim_clean_pages_from_list(struct zone
*zone
,
573 struct list_head
*page_list
);
574 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
575 #define ALLOC_WMARK_MIN WMARK_MIN
576 #define ALLOC_WMARK_LOW WMARK_LOW
577 #define ALLOC_WMARK_HIGH WMARK_HIGH
578 #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
580 /* Mask to get the watermark bits */
581 #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
584 * Only MMU archs have async oom victim reclaim - aka oom_reaper so we
585 * cannot assume a reduced access to memory reserves is sufficient for
589 #define ALLOC_OOM 0x08
591 #define ALLOC_OOM ALLOC_NO_WATERMARKS
594 #define ALLOC_HARDER 0x10 /* try to alloc harder */
595 #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
596 #define ALLOC_CPUSET 0x40 /* check for correct cpuset */
597 #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
598 #ifdef CONFIG_ZONE_DMA32
599 #define ALLOC_NOFRAGMENT 0x100 /* avoid mixing pageblock types */
601 #define ALLOC_NOFRAGMENT 0x0
603 #define ALLOC_KSWAPD 0x800 /* allow waking of kswapd, __GFP_KSWAPD_RECLAIM set */
606 struct tlbflush_unmap_batch
;
610 * only for MM internal work items which do not depend on
611 * any allocations or locks which might depend on allocations
613 extern struct workqueue_struct
*mm_percpu_wq
;
615 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
616 void try_to_unmap_flush(void);
617 void try_to_unmap_flush_dirty(void);
618 void flush_tlb_batched_pending(struct mm_struct
*mm
);
620 static inline void try_to_unmap_flush(void)
623 static inline void try_to_unmap_flush_dirty(void)
626 static inline void flush_tlb_batched_pending(struct mm_struct
*mm
)
629 #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
631 extern const struct trace_print_flags pageflag_names
[];
632 extern const struct trace_print_flags vmaflag_names
[];
633 extern const struct trace_print_flags gfpflag_names
[];
635 static inline bool is_migrate_highatomic(enum migratetype migratetype
)
637 return migratetype
== MIGRATE_HIGHATOMIC
;
640 static inline bool is_migrate_highatomic_page(struct page
*page
)
642 return get_pageblock_migratetype(page
) == MIGRATE_HIGHATOMIC
;
645 void setup_zone_pageset(struct zone
*zone
);
647 struct migration_target_control
{
648 int nid
; /* preferred node id */
657 int vmap_pages_range_noflush(unsigned long addr
, unsigned long end
,
658 pgprot_t prot
, struct page
**pages
, unsigned int page_shift
);
661 int vmap_pages_range_noflush(unsigned long addr
, unsigned long end
,
662 pgprot_t prot
, struct page
**pages
, unsigned int page_shift
)
668 void vunmap_range_noflush(unsigned long start
, unsigned long end
);
670 #endif /* __MM_INTERNAL_H */