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 force_page_cache_readahead(struct address_space
*, struct file
*,
53 pgoff_t index
, unsigned long nr_to_read
);
54 void __do_page_cache_readahead(struct address_space
*, struct file
*,
55 pgoff_t index
, unsigned long nr_to_read
,
56 unsigned long lookahead_size
);
59 * Submit IO for the read-ahead request in file_ra_state.
61 static inline void ra_submit(struct file_ra_state
*ra
,
62 struct address_space
*mapping
, struct file
*filp
)
64 __do_page_cache_readahead(mapping
, filp
,
65 ra
->start
, ra
->size
, ra
->async_size
);
69 * page_evictable - test whether a page is evictable
70 * @page: the page to test
72 * Test whether page is evictable--i.e., should be placed on active/inactive
73 * lists vs unevictable list.
75 * Reasons page might not be evictable:
76 * (1) page's mapping marked unevictable
77 * (2) page is part of an mlocked VMA
80 static inline bool page_evictable(struct page
*page
)
84 /* Prevent address_space of inode and swap cache from being freed */
86 ret
= !mapping_unevictable(page_mapping(page
)) && !PageMlocked(page
);
92 * Turn a non-refcounted page (->_refcount == 0) into refcounted with
95 static inline void set_page_refcounted(struct page
*page
)
97 VM_BUG_ON_PAGE(PageTail(page
), page
);
98 VM_BUG_ON_PAGE(page_ref_count(page
), page
);
99 set_page_count(page
, 1);
102 extern unsigned long highest_memmap_pfn
;
105 * Maximum number of reclaim retries without progress before the OOM
106 * killer is consider the only way forward.
108 #define MAX_RECLAIM_RETRIES 16
113 extern int isolate_lru_page(struct page
*page
);
114 extern void putback_lru_page(struct page
*page
);
119 extern pmd_t
*mm_find_pmd(struct mm_struct
*mm
, unsigned long address
);
126 * Structure for holding the mostly immutable allocation parameters passed
127 * between functions involved in allocations, including the alloc_pages*
128 * family of functions.
130 * nodemask, migratetype and highest_zoneidx are initialized only once in
131 * __alloc_pages_nodemask() and then never change.
133 * zonelist, preferred_zone and highest_zoneidx are set first in
134 * __alloc_pages_nodemask() for the fast path, and might be later changed
135 * in __alloc_pages_slowpath(). All other functions pass the whole structure
136 * by a const pointer.
138 struct alloc_context
{
139 struct zonelist
*zonelist
;
140 nodemask_t
*nodemask
;
141 struct zoneref
*preferred_zoneref
;
145 * highest_zoneidx represents highest usable zone index of
146 * the allocation request. Due to the nature of the zone,
147 * memory on lower zone than the highest_zoneidx will be
148 * protected by lowmem_reserve[highest_zoneidx].
150 * highest_zoneidx is also used by reclaim/compaction to limit
151 * the target zone since higher zone than this index cannot be
152 * usable for this allocation request.
154 enum zone_type highest_zoneidx
;
155 bool spread_dirty_pages
;
159 * Locate the struct page for both the matching buddy in our
160 * pair (buddy1) and the combined O(n+1) page they form (page).
162 * 1) Any buddy B1 will have an order O twin B2 which satisfies
163 * the following equation:
165 * For example, if the starting buddy (buddy2) is #8 its order
167 * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
169 * 2) Any buddy B will have an order O+1 parent P which
170 * satisfies the following equation:
173 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
175 static inline unsigned long
176 __find_buddy_pfn(unsigned long page_pfn
, unsigned int order
)
178 return page_pfn
^ (1 << order
);
181 extern struct page
*__pageblock_pfn_to_page(unsigned long start_pfn
,
182 unsigned long end_pfn
, struct zone
*zone
);
184 static inline struct page
*pageblock_pfn_to_page(unsigned long start_pfn
,
185 unsigned long end_pfn
, struct zone
*zone
)
187 if (zone
->contiguous
)
188 return pfn_to_page(start_pfn
);
190 return __pageblock_pfn_to_page(start_pfn
, end_pfn
, zone
);
193 extern int __isolate_free_page(struct page
*page
, unsigned int order
);
194 extern void __putback_isolated_page(struct page
*page
, unsigned int order
,
196 extern void memblock_free_pages(struct page
*page
, unsigned long pfn
,
198 extern void __free_pages_core(struct page
*page
, unsigned int order
);
199 extern void prep_compound_page(struct page
*page
, unsigned int order
);
200 extern void post_alloc_hook(struct page
*page
, unsigned int order
,
202 extern int user_min_free_kbytes
;
204 extern void zone_pcp_update(struct zone
*zone
);
205 extern void zone_pcp_reset(struct zone
*zone
);
207 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
213 * compact_control is used to track pages being migrated and the free pages
214 * they are being migrated to during memory compaction. The free_pfn starts
215 * at the end of a zone and migrate_pfn begins at the start. Movable pages
216 * are moved to the end of a zone during a compaction run and the run
217 * completes when free_pfn <= migrate_pfn
219 struct compact_control
{
220 struct list_head freepages
; /* List of free pages to migrate to */
221 struct list_head migratepages
; /* List of pages being migrated */
222 unsigned int nr_freepages
; /* Number of isolated free pages */
223 unsigned int nr_migratepages
; /* Number of pages to migrate */
224 unsigned long free_pfn
; /* isolate_freepages search base */
225 unsigned long migrate_pfn
; /* isolate_migratepages search base */
226 unsigned long fast_start_pfn
; /* a pfn to start linear scan from */
228 unsigned long total_migrate_scanned
;
229 unsigned long total_free_scanned
;
230 unsigned short fast_search_fail
;/* failures to use free list searches */
231 short search_order
; /* order to start a fast search at */
232 const gfp_t gfp_mask
; /* gfp mask of a direct compactor */
233 int order
; /* order a direct compactor needs */
234 int migratetype
; /* migratetype of direct compactor */
235 const unsigned int alloc_flags
; /* alloc flags of a direct compactor */
236 const int highest_zoneidx
; /* zone index of a direct compactor */
237 enum migrate_mode mode
; /* Async or sync migration mode */
238 bool ignore_skip_hint
; /* Scan blocks even if marked skip */
239 bool no_set_skip_hint
; /* Don't mark blocks for skipping */
240 bool ignore_block_suitable
; /* Scan blocks considered unsuitable */
241 bool direct_compaction
; /* False from kcompactd or /proc/... */
242 bool proactive_compaction
; /* kcompactd proactive compaction */
243 bool whole_zone
; /* Whole zone should/has been scanned */
244 bool contended
; /* Signal lock or sched contention */
245 bool rescan
; /* Rescanning the same pageblock */
246 bool alloc_contig
; /* alloc_contig_range allocation */
250 * Used in direct compaction when a page should be taken from the freelists
251 * immediately when one is created during the free path.
253 struct capture_control
{
254 struct compact_control
*cc
;
259 isolate_freepages_range(struct compact_control
*cc
,
260 unsigned long start_pfn
, unsigned long end_pfn
);
262 isolate_migratepages_range(struct compact_control
*cc
,
263 unsigned long low_pfn
, unsigned long end_pfn
);
264 int find_suitable_fallback(struct free_area
*area
, unsigned int order
,
265 int migratetype
, bool only_stealable
, bool *can_steal
);
270 * This function returns the order of a free page in the buddy system. In
271 * general, page_zone(page)->lock must be held by the caller to prevent the
272 * page from being allocated in parallel and returning garbage as the order.
273 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
274 * page cannot be allocated or merged in parallel. Alternatively, it must
275 * handle invalid values gracefully, and use page_order_unsafe() below.
277 static inline unsigned int page_order(struct page
*page
)
279 /* PageBuddy() must be checked by the caller */
280 return page_private(page
);
284 * Like page_order(), but for callers who cannot afford to hold the zone lock.
285 * PageBuddy() should be checked first by the caller to minimize race window,
286 * and invalid values must be handled gracefully.
288 * READ_ONCE is used so that if the caller assigns the result into a local
289 * variable and e.g. tests it for valid range before using, the compiler cannot
290 * decide to remove the variable and inline the page_private(page) multiple
291 * times, potentially observing different values in the tests and the actual
294 #define page_order_unsafe(page) READ_ONCE(page_private(page))
296 static inline bool is_cow_mapping(vm_flags_t flags
)
298 return (flags
& (VM_SHARED
| VM_MAYWRITE
)) == VM_MAYWRITE
;
302 * These three helpers classifies VMAs for virtual memory accounting.
306 * Executable code area - executable, not writable, not stack
308 static inline bool is_exec_mapping(vm_flags_t flags
)
310 return (flags
& (VM_EXEC
| VM_WRITE
| VM_STACK
)) == VM_EXEC
;
314 * Stack area - atomatically grows in one direction
316 * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
317 * do_mmap() forbids all other combinations.
319 static inline bool is_stack_mapping(vm_flags_t flags
)
321 return (flags
& VM_STACK
) == VM_STACK
;
325 * Data area - private, writable, not stack
327 static inline bool is_data_mapping(vm_flags_t flags
)
329 return (flags
& (VM_WRITE
| VM_SHARED
| VM_STACK
)) == VM_WRITE
;
333 void __vma_link_list(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
334 struct vm_area_struct
*prev
);
335 void __vma_unlink_list(struct mm_struct
*mm
, struct vm_area_struct
*vma
);
338 extern long populate_vma_page_range(struct vm_area_struct
*vma
,
339 unsigned long start
, unsigned long end
, int *nonblocking
);
340 extern void munlock_vma_pages_range(struct vm_area_struct
*vma
,
341 unsigned long start
, unsigned long end
);
342 static inline void munlock_vma_pages_all(struct vm_area_struct
*vma
)
344 munlock_vma_pages_range(vma
, vma
->vm_start
, vma
->vm_end
);
348 * must be called with vma's mmap_lock held for read or write, and page locked.
350 extern void mlock_vma_page(struct page
*page
);
351 extern unsigned int munlock_vma_page(struct page
*page
);
354 * Clear the page's PageMlocked(). This can be useful in a situation where
355 * we want to unconditionally remove a page from the pagecache -- e.g.,
356 * on truncation or freeing.
358 * It is legal to call this function for any page, mlocked or not.
359 * If called for a page that is still mapped by mlocked vmas, all we do
360 * is revert to lazy LRU behaviour -- semantics are not broken.
362 extern void clear_page_mlock(struct page
*page
);
365 * mlock_migrate_page - called only from migrate_misplaced_transhuge_page()
366 * (because that does not go through the full procedure of migration ptes):
367 * to migrate the Mlocked page flag; update statistics.
369 static inline void mlock_migrate_page(struct page
*newpage
, struct page
*page
)
371 if (TestClearPageMlocked(page
)) {
372 int nr_pages
= thp_nr_pages(page
);
374 /* Holding pmd lock, no change in irq context: __mod is safe */
375 __mod_zone_page_state(page_zone(page
), NR_MLOCK
, -nr_pages
);
376 SetPageMlocked(newpage
);
377 __mod_zone_page_state(page_zone(newpage
), NR_MLOCK
, nr_pages
);
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?
386 static inline unsigned long
387 __vma_address(struct page
*page
, struct vm_area_struct
*vma
)
389 pgoff_t pgoff
= page_to_pgoff(page
);
390 return vma
->vm_start
+ ((pgoff
- vma
->vm_pgoff
) << PAGE_SHIFT
);
393 static inline unsigned long
394 vma_address(struct page
*page
, struct vm_area_struct
*vma
)
396 unsigned long start
, end
;
398 start
= __vma_address(page
, vma
);
399 end
= start
+ thp_size(page
) - PAGE_SIZE
;
401 /* page should be within @vma mapping range */
402 VM_BUG_ON_VMA(end
< vma
->vm_start
|| start
>= vma
->vm_end
, vma
);
404 return max(start
, vma
->vm_start
);
407 static inline struct file
*maybe_unlock_mmap_for_io(struct vm_fault
*vmf
,
410 int flags
= vmf
->flags
;
416 * FAULT_FLAG_RETRY_NOWAIT means we don't want to wait on page locks or
417 * anything, so we only pin the file and drop the mmap_lock if only
418 * FAULT_FLAG_ALLOW_RETRY is set, while this is the first attempt.
420 if (fault_flag_allow_retry_first(flags
) &&
421 !(flags
& FAULT_FLAG_RETRY_NOWAIT
)) {
422 fpin
= get_file(vmf
->vma
->vm_file
);
423 mmap_read_unlock(vmf
->vma
->vm_mm
);
428 #else /* !CONFIG_MMU */
429 static inline void clear_page_mlock(struct page
*page
) { }
430 static inline void mlock_vma_page(struct page
*page
) { }
431 static inline void mlock_migrate_page(struct page
*new, struct page
*old
) { }
433 #endif /* !CONFIG_MMU */
436 * Return the mem_map entry representing the 'offset' subpage within
437 * the maximally aligned gigantic page 'base'. Handle any discontiguity
438 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
440 static inline struct page
*mem_map_offset(struct page
*base
, int offset
)
442 if (unlikely(offset
>= MAX_ORDER_NR_PAGES
))
443 return nth_page(base
, offset
);
444 return base
+ offset
;
448 * Iterator over all subpages within the maximally aligned gigantic
449 * page 'base'. Handle any discontiguity in the mem_map.
451 static inline struct page
*mem_map_next(struct page
*iter
,
452 struct page
*base
, int offset
)
454 if (unlikely((offset
& (MAX_ORDER_NR_PAGES
- 1)) == 0)) {
455 unsigned long pfn
= page_to_pfn(base
) + offset
;
458 return pfn_to_page(pfn
);
463 /* Memory initialisation debug and verification */
470 #ifdef CONFIG_DEBUG_MEMORY_INIT
472 extern int mminit_loglevel
;
474 #define mminit_dprintk(level, prefix, fmt, arg...) \
476 if (level < mminit_loglevel) { \
477 if (level <= MMINIT_WARNING) \
478 pr_warn("mminit::" prefix " " fmt, ##arg); \
480 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
484 extern void mminit_verify_pageflags_layout(void);
485 extern void mminit_verify_zonelist(void);
488 static inline void mminit_dprintk(enum mminit_level level
,
489 const char *prefix
, const char *fmt
, ...)
493 static inline void mminit_verify_pageflags_layout(void)
497 static inline void mminit_verify_zonelist(void)
500 #endif /* CONFIG_DEBUG_MEMORY_INIT */
502 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
503 #if defined(CONFIG_SPARSEMEM)
504 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn
,
505 unsigned long *end_pfn
);
507 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn
,
508 unsigned long *end_pfn
)
511 #endif /* CONFIG_SPARSEMEM */
513 #define NODE_RECLAIM_NOSCAN -2
514 #define NODE_RECLAIM_FULL -1
515 #define NODE_RECLAIM_SOME 0
516 #define NODE_RECLAIM_SUCCESS 1
519 extern int node_reclaim(struct pglist_data
*, gfp_t
, unsigned int);
521 static inline int node_reclaim(struct pglist_data
*pgdat
, gfp_t mask
,
524 return NODE_RECLAIM_NOSCAN
;
528 extern int hwpoison_filter(struct page
*p
);
530 extern u32 hwpoison_filter_dev_major
;
531 extern u32 hwpoison_filter_dev_minor
;
532 extern u64 hwpoison_filter_flags_mask
;
533 extern u64 hwpoison_filter_flags_value
;
534 extern u64 hwpoison_filter_memcg
;
535 extern u32 hwpoison_filter_enable
;
537 extern unsigned long __must_check
vm_mmap_pgoff(struct file
*, unsigned long,
538 unsigned long, unsigned long,
539 unsigned long, unsigned long);
541 extern void set_pageblock_order(void);
542 unsigned int reclaim_clean_pages_from_list(struct zone
*zone
,
543 struct list_head
*page_list
);
544 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
545 #define ALLOC_WMARK_MIN WMARK_MIN
546 #define ALLOC_WMARK_LOW WMARK_LOW
547 #define ALLOC_WMARK_HIGH WMARK_HIGH
548 #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
550 /* Mask to get the watermark bits */
551 #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
554 * Only MMU archs have async oom victim reclaim - aka oom_reaper so we
555 * cannot assume a reduced access to memory reserves is sufficient for
559 #define ALLOC_OOM 0x08
561 #define ALLOC_OOM ALLOC_NO_WATERMARKS
564 #define ALLOC_HARDER 0x10 /* try to alloc harder */
565 #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
566 #define ALLOC_CPUSET 0x40 /* check for correct cpuset */
567 #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
568 #ifdef CONFIG_ZONE_DMA32
569 #define ALLOC_NOFRAGMENT 0x100 /* avoid mixing pageblock types */
571 #define ALLOC_NOFRAGMENT 0x0
573 #define ALLOC_KSWAPD 0x800 /* allow waking of kswapd, __GFP_KSWAPD_RECLAIM set */
576 struct tlbflush_unmap_batch
;
580 * only for MM internal work items which do not depend on
581 * any allocations or locks which might depend on allocations
583 extern struct workqueue_struct
*mm_percpu_wq
;
585 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
586 void try_to_unmap_flush(void);
587 void try_to_unmap_flush_dirty(void);
588 void flush_tlb_batched_pending(struct mm_struct
*mm
);
590 static inline void try_to_unmap_flush(void)
593 static inline void try_to_unmap_flush_dirty(void)
596 static inline void flush_tlb_batched_pending(struct mm_struct
*mm
)
599 #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
601 extern const struct trace_print_flags pageflag_names
[];
602 extern const struct trace_print_flags vmaflag_names
[];
603 extern const struct trace_print_flags gfpflag_names
[];
605 static inline bool is_migrate_highatomic(enum migratetype migratetype
)
607 return migratetype
== MIGRATE_HIGHATOMIC
;
610 static inline bool is_migrate_highatomic_page(struct page
*page
)
612 return get_pageblock_migratetype(page
) == MIGRATE_HIGHATOMIC
;
615 void setup_zone_pageset(struct zone
*zone
);
617 struct migration_target_control
{
618 int nid
; /* preferred node id */
623 #endif /* __MM_INTERNAL_H */