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 extern unsigned int __do_page_cache_readahead(struct address_space
*mapping
,
53 struct file
*filp
, pgoff_t offset
, unsigned long nr_to_read
,
54 unsigned long lookahead_size
);
57 * Submit IO for the read-ahead request in file_ra_state.
59 static inline unsigned long ra_submit(struct file_ra_state
*ra
,
60 struct address_space
*mapping
, struct file
*filp
)
62 return __do_page_cache_readahead(mapping
, filp
,
63 ra
->start
, ra
->size
, ra
->async_size
);
67 * page_evictable - test whether a page is evictable
68 * @page: the page to test
70 * Test whether page is evictable--i.e., should be placed on active/inactive
71 * lists vs unevictable list.
73 * Reasons page might not be evictable:
74 * (1) page's mapping marked unevictable
75 * (2) page is part of an mlocked VMA
78 static inline bool page_evictable(struct page
*page
)
82 /* Prevent address_space of inode and swap cache from being freed */
84 ret
= !mapping_unevictable(page_mapping(page
)) && !PageMlocked(page
);
90 * Turn a non-refcounted page (->_refcount == 0) into refcounted with
93 static inline void set_page_refcounted(struct page
*page
)
95 VM_BUG_ON_PAGE(PageTail(page
), page
);
96 VM_BUG_ON_PAGE(page_ref_count(page
), page
);
97 set_page_count(page
, 1);
100 extern unsigned long highest_memmap_pfn
;
103 * Maximum number of reclaim retries without progress before the OOM
104 * killer is consider the only way forward.
106 #define MAX_RECLAIM_RETRIES 16
111 extern int isolate_lru_page(struct page
*page
);
112 extern void putback_lru_page(struct page
*page
);
117 extern pmd_t
*mm_find_pmd(struct mm_struct
*mm
, unsigned long address
);
124 * Structure for holding the mostly immutable allocation parameters passed
125 * between functions involved in allocations, including the alloc_pages*
126 * family of functions.
128 * nodemask, migratetype and high_zoneidx are initialized only once in
129 * __alloc_pages_nodemask() and then never change.
131 * zonelist, preferred_zone and classzone_idx are set first in
132 * __alloc_pages_nodemask() for the fast path, and might be later changed
133 * in __alloc_pages_slowpath(). All other functions pass the whole strucure
134 * by a const pointer.
136 struct alloc_context
{
137 struct zonelist
*zonelist
;
138 nodemask_t
*nodemask
;
139 struct zoneref
*preferred_zoneref
;
141 enum zone_type high_zoneidx
;
142 bool spread_dirty_pages
;
145 #define ac_classzone_idx(ac) zonelist_zone_idx(ac->preferred_zoneref)
148 * Locate the struct page for both the matching buddy in our
149 * pair (buddy1) and the combined O(n+1) page they form (page).
151 * 1) Any buddy B1 will have an order O twin B2 which satisfies
152 * the following equation:
154 * For example, if the starting buddy (buddy2) is #8 its order
156 * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
158 * 2) Any buddy B will have an order O+1 parent P which
159 * satisfies the following equation:
162 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
164 static inline unsigned long
165 __find_buddy_pfn(unsigned long page_pfn
, unsigned int order
)
167 return page_pfn
^ (1 << order
);
170 extern struct page
*__pageblock_pfn_to_page(unsigned long start_pfn
,
171 unsigned long end_pfn
, struct zone
*zone
);
173 static inline struct page
*pageblock_pfn_to_page(unsigned long start_pfn
,
174 unsigned long end_pfn
, struct zone
*zone
)
176 if (zone
->contiguous
)
177 return pfn_to_page(start_pfn
);
179 return __pageblock_pfn_to_page(start_pfn
, end_pfn
, zone
);
182 extern int __isolate_free_page(struct page
*page
, unsigned int order
);
183 extern void memblock_free_pages(struct page
*page
, unsigned long pfn
,
185 extern void __free_pages_core(struct page
*page
, unsigned int order
);
186 extern void prep_compound_page(struct page
*page
, unsigned int order
);
187 extern void post_alloc_hook(struct page
*page
, unsigned int order
,
189 extern int user_min_free_kbytes
;
191 extern void zone_pcp_update(struct zone
*zone
);
192 extern void zone_pcp_reset(struct zone
*zone
);
194 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
200 * compact_control is used to track pages being migrated and the free pages
201 * they are being migrated to during memory compaction. The free_pfn starts
202 * at the end of a zone and migrate_pfn begins at the start. Movable pages
203 * are moved to the end of a zone during a compaction run and the run
204 * completes when free_pfn <= migrate_pfn
206 struct compact_control
{
207 struct list_head freepages
; /* List of free pages to migrate to */
208 struct list_head migratepages
; /* List of pages being migrated */
209 unsigned int nr_freepages
; /* Number of isolated free pages */
210 unsigned int nr_migratepages
; /* Number of pages to migrate */
211 unsigned long free_pfn
; /* isolate_freepages search base */
212 unsigned long migrate_pfn
; /* isolate_migratepages search base */
213 unsigned long fast_start_pfn
; /* a pfn to start linear scan from */
215 unsigned long total_migrate_scanned
;
216 unsigned long total_free_scanned
;
217 unsigned short fast_search_fail
;/* failures to use free list searches */
218 short search_order
; /* order to start a fast search at */
219 const gfp_t gfp_mask
; /* gfp mask of a direct compactor */
220 int order
; /* order a direct compactor needs */
221 int migratetype
; /* migratetype of direct compactor */
222 const unsigned int alloc_flags
; /* alloc flags of a direct compactor */
223 const int classzone_idx
; /* zone index of a direct compactor */
224 enum migrate_mode mode
; /* Async or sync migration mode */
225 bool ignore_skip_hint
; /* Scan blocks even if marked skip */
226 bool no_set_skip_hint
; /* Don't mark blocks for skipping */
227 bool ignore_block_suitable
; /* Scan blocks considered unsuitable */
228 bool direct_compaction
; /* False from kcompactd or /proc/... */
229 bool whole_zone
; /* Whole zone should/has been scanned */
230 bool contended
; /* Signal lock or sched contention */
231 bool rescan
; /* Rescanning the same pageblock */
235 * Used in direct compaction when a page should be taken from the freelists
236 * immediately when one is created during the free path.
238 struct capture_control
{
239 struct compact_control
*cc
;
244 isolate_freepages_range(struct compact_control
*cc
,
245 unsigned long start_pfn
, unsigned long end_pfn
);
247 isolate_migratepages_range(struct compact_control
*cc
,
248 unsigned long low_pfn
, unsigned long end_pfn
);
249 int find_suitable_fallback(struct free_area
*area
, unsigned int order
,
250 int migratetype
, bool only_stealable
, bool *can_steal
);
255 * This function returns the order of a free page in the buddy system. In
256 * general, page_zone(page)->lock must be held by the caller to prevent the
257 * page from being allocated in parallel and returning garbage as the order.
258 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
259 * page cannot be allocated or merged in parallel. Alternatively, it must
260 * handle invalid values gracefully, and use page_order_unsafe() below.
262 static inline unsigned int page_order(struct page
*page
)
264 /* PageBuddy() must be checked by the caller */
265 return page_private(page
);
269 * Like page_order(), but for callers who cannot afford to hold the zone lock.
270 * PageBuddy() should be checked first by the caller to minimize race window,
271 * and invalid values must be handled gracefully.
273 * READ_ONCE is used so that if the caller assigns the result into a local
274 * variable and e.g. tests it for valid range before using, the compiler cannot
275 * decide to remove the variable and inline the page_private(page) multiple
276 * times, potentially observing different values in the tests and the actual
279 #define page_order_unsafe(page) READ_ONCE(page_private(page))
281 static inline bool is_cow_mapping(vm_flags_t flags
)
283 return (flags
& (VM_SHARED
| VM_MAYWRITE
)) == VM_MAYWRITE
;
287 * These three helpers classifies VMAs for virtual memory accounting.
291 * Executable code area - executable, not writable, not stack
293 static inline bool is_exec_mapping(vm_flags_t flags
)
295 return (flags
& (VM_EXEC
| VM_WRITE
| VM_STACK
)) == VM_EXEC
;
299 * Stack area - atomatically grows in one direction
301 * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
302 * do_mmap() forbids all other combinations.
304 static inline bool is_stack_mapping(vm_flags_t flags
)
306 return (flags
& VM_STACK
) == VM_STACK
;
310 * Data area - private, writable, not stack
312 static inline bool is_data_mapping(vm_flags_t flags
)
314 return (flags
& (VM_WRITE
| VM_SHARED
| VM_STACK
)) == VM_WRITE
;
318 void __vma_link_list(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
319 struct vm_area_struct
*prev
);
320 void __vma_unlink_list(struct mm_struct
*mm
, struct vm_area_struct
*vma
);
323 extern long populate_vma_page_range(struct vm_area_struct
*vma
,
324 unsigned long start
, unsigned long end
, int *nonblocking
);
325 extern void munlock_vma_pages_range(struct vm_area_struct
*vma
,
326 unsigned long start
, unsigned long end
);
327 static inline void munlock_vma_pages_all(struct vm_area_struct
*vma
)
329 munlock_vma_pages_range(vma
, vma
->vm_start
, vma
->vm_end
);
333 * must be called with vma's mmap_sem held for read or write, and page locked.
335 extern void mlock_vma_page(struct page
*page
);
336 extern unsigned int munlock_vma_page(struct page
*page
);
339 * Clear the page's PageMlocked(). This can be useful in a situation where
340 * we want to unconditionally remove a page from the pagecache -- e.g.,
341 * on truncation or freeing.
343 * It is legal to call this function for any page, mlocked or not.
344 * If called for a page that is still mapped by mlocked vmas, all we do
345 * is revert to lazy LRU behaviour -- semantics are not broken.
347 extern void clear_page_mlock(struct page
*page
);
350 * mlock_migrate_page - called only from migrate_misplaced_transhuge_page()
351 * (because that does not go through the full procedure of migration ptes):
352 * to migrate the Mlocked page flag; update statistics.
354 static inline void mlock_migrate_page(struct page
*newpage
, struct page
*page
)
356 if (TestClearPageMlocked(page
)) {
357 int nr_pages
= hpage_nr_pages(page
);
359 /* Holding pmd lock, no change in irq context: __mod is safe */
360 __mod_zone_page_state(page_zone(page
), NR_MLOCK
, -nr_pages
);
361 SetPageMlocked(newpage
);
362 __mod_zone_page_state(page_zone(newpage
), NR_MLOCK
, nr_pages
);
366 extern pmd_t
maybe_pmd_mkwrite(pmd_t pmd
, struct vm_area_struct
*vma
);
369 * At what user virtual address is page expected in @vma?
371 static inline unsigned long
372 __vma_address(struct page
*page
, struct vm_area_struct
*vma
)
374 pgoff_t pgoff
= page_to_pgoff(page
);
375 return vma
->vm_start
+ ((pgoff
- vma
->vm_pgoff
) << PAGE_SHIFT
);
378 static inline unsigned long
379 vma_address(struct page
*page
, struct vm_area_struct
*vma
)
381 unsigned long start
, end
;
383 start
= __vma_address(page
, vma
);
384 end
= start
+ PAGE_SIZE
* (hpage_nr_pages(page
) - 1);
386 /* page should be within @vma mapping range */
387 VM_BUG_ON_VMA(end
< vma
->vm_start
|| start
>= vma
->vm_end
, vma
);
389 return max(start
, vma
->vm_start
);
392 static inline struct file
*maybe_unlock_mmap_for_io(struct vm_fault
*vmf
,
395 int flags
= vmf
->flags
;
401 * FAULT_FLAG_RETRY_NOWAIT means we don't want to wait on page locks or
402 * anything, so we only pin the file and drop the mmap_sem if only
403 * FAULT_FLAG_ALLOW_RETRY is set, while this is the first attempt.
405 if (fault_flag_allow_retry_first(flags
) &&
406 !(flags
& FAULT_FLAG_RETRY_NOWAIT
)) {
407 fpin
= get_file(vmf
->vma
->vm_file
);
408 up_read(&vmf
->vma
->vm_mm
->mmap_sem
);
413 #else /* !CONFIG_MMU */
414 static inline void clear_page_mlock(struct page
*page
) { }
415 static inline void mlock_vma_page(struct page
*page
) { }
416 static inline void mlock_migrate_page(struct page
*new, struct page
*old
) { }
418 #endif /* !CONFIG_MMU */
421 * Return the mem_map entry representing the 'offset' subpage within
422 * the maximally aligned gigantic page 'base'. Handle any discontiguity
423 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
425 static inline struct page
*mem_map_offset(struct page
*base
, int offset
)
427 if (unlikely(offset
>= MAX_ORDER_NR_PAGES
))
428 return nth_page(base
, offset
);
429 return base
+ offset
;
433 * Iterator over all subpages within the maximally aligned gigantic
434 * page 'base'. Handle any discontiguity in the mem_map.
436 static inline struct page
*mem_map_next(struct page
*iter
,
437 struct page
*base
, int offset
)
439 if (unlikely((offset
& (MAX_ORDER_NR_PAGES
- 1)) == 0)) {
440 unsigned long pfn
= page_to_pfn(base
) + offset
;
443 return pfn_to_page(pfn
);
448 /* Memory initialisation debug and verification */
455 #ifdef CONFIG_DEBUG_MEMORY_INIT
457 extern int mminit_loglevel
;
459 #define mminit_dprintk(level, prefix, fmt, arg...) \
461 if (level < mminit_loglevel) { \
462 if (level <= MMINIT_WARNING) \
463 pr_warn("mminit::" prefix " " fmt, ##arg); \
465 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
469 extern void mminit_verify_pageflags_layout(void);
470 extern void mminit_verify_zonelist(void);
473 static inline void mminit_dprintk(enum mminit_level level
,
474 const char *prefix
, const char *fmt
, ...)
478 static inline void mminit_verify_pageflags_layout(void)
482 static inline void mminit_verify_zonelist(void)
485 #endif /* CONFIG_DEBUG_MEMORY_INIT */
487 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
488 #if defined(CONFIG_SPARSEMEM)
489 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn
,
490 unsigned long *end_pfn
);
492 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn
,
493 unsigned long *end_pfn
)
496 #endif /* CONFIG_SPARSEMEM */
498 #define NODE_RECLAIM_NOSCAN -2
499 #define NODE_RECLAIM_FULL -1
500 #define NODE_RECLAIM_SOME 0
501 #define NODE_RECLAIM_SUCCESS 1
504 extern int node_reclaim(struct pglist_data
*, gfp_t
, unsigned int);
506 static inline int node_reclaim(struct pglist_data
*pgdat
, gfp_t mask
,
509 return NODE_RECLAIM_NOSCAN
;
513 extern int hwpoison_filter(struct page
*p
);
515 extern u32 hwpoison_filter_dev_major
;
516 extern u32 hwpoison_filter_dev_minor
;
517 extern u64 hwpoison_filter_flags_mask
;
518 extern u64 hwpoison_filter_flags_value
;
519 extern u64 hwpoison_filter_memcg
;
520 extern u32 hwpoison_filter_enable
;
522 extern unsigned long __must_check
vm_mmap_pgoff(struct file
*, unsigned long,
523 unsigned long, unsigned long,
524 unsigned long, unsigned long);
526 extern void set_pageblock_order(void);
527 unsigned long reclaim_clean_pages_from_list(struct zone
*zone
,
528 struct list_head
*page_list
);
529 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
530 #define ALLOC_WMARK_MIN WMARK_MIN
531 #define ALLOC_WMARK_LOW WMARK_LOW
532 #define ALLOC_WMARK_HIGH WMARK_HIGH
533 #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
535 /* Mask to get the watermark bits */
536 #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
539 * Only MMU archs have async oom victim reclaim - aka oom_reaper so we
540 * cannot assume a reduced access to memory reserves is sufficient for
544 #define ALLOC_OOM 0x08
546 #define ALLOC_OOM ALLOC_NO_WATERMARKS
549 #define ALLOC_HARDER 0x10 /* try to alloc harder */
550 #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
551 #define ALLOC_CPUSET 0x40 /* check for correct cpuset */
552 #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
553 #ifdef CONFIG_ZONE_DMA32
554 #define ALLOC_NOFRAGMENT 0x100 /* avoid mixing pageblock types */
556 #define ALLOC_NOFRAGMENT 0x0
558 #define ALLOC_KSWAPD 0x200 /* allow waking of kswapd */
561 struct tlbflush_unmap_batch
;
565 * only for MM internal work items which do not depend on
566 * any allocations or locks which might depend on allocations
568 extern struct workqueue_struct
*mm_percpu_wq
;
570 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
571 void try_to_unmap_flush(void);
572 void try_to_unmap_flush_dirty(void);
573 void flush_tlb_batched_pending(struct mm_struct
*mm
);
575 static inline void try_to_unmap_flush(void)
578 static inline void try_to_unmap_flush_dirty(void)
581 static inline void flush_tlb_batched_pending(struct mm_struct
*mm
)
584 #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
586 extern const struct trace_print_flags pageflag_names
[];
587 extern const struct trace_print_flags vmaflag_names
[];
588 extern const struct trace_print_flags gfpflag_names
[];
590 static inline bool is_migrate_highatomic(enum migratetype migratetype
)
592 return migratetype
== MIGRATE_HIGHATOMIC
;
595 static inline bool is_migrate_highatomic_page(struct page
*page
)
597 return get_pageblock_migratetype(page
) == MIGRATE_HIGHATOMIC
;
600 void setup_zone_pageset(struct zone
*zone
);
601 extern struct page
*alloc_new_node_page(struct page
*page
, unsigned long node
);
602 #endif /* __MM_INTERNAL_H */