1 /* internal.h: mm/ internal definitions
3 * Copyright (C) 2004 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
11 #ifndef __MM_INTERNAL_H
12 #define __MM_INTERNAL_H
18 * The set of flags that only affect watermark checking and reclaim
19 * behaviour. This is used by the MM to obey the caller constraints
20 * about IO, FS and watermark checking while ignoring placement
21 * hints such as HIGHMEM usage.
23 #define GFP_RECLAIM_MASK (__GFP_RECLAIM|__GFP_HIGH|__GFP_IO|__GFP_FS|\
24 __GFP_NOWARN|__GFP_REPEAT|__GFP_NOFAIL|\
25 __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC)
27 /* The GFP flags allowed during early boot */
28 #define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_RECLAIM|__GFP_IO|__GFP_FS))
30 /* Control allocation cpuset and node placement constraints */
31 #define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE)
33 /* Do not use these with a slab allocator */
34 #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
36 void free_pgtables(struct mmu_gather
*tlb
, struct vm_area_struct
*start_vma
,
37 unsigned long floor
, unsigned long ceiling
);
39 static inline void set_page_count(struct page
*page
, int v
)
41 atomic_set(&page
->_count
, v
);
44 extern int __do_page_cache_readahead(struct address_space
*mapping
,
45 struct file
*filp
, pgoff_t offset
, unsigned long nr_to_read
,
46 unsigned long lookahead_size
);
49 * Submit IO for the read-ahead request in file_ra_state.
51 static inline unsigned long ra_submit(struct file_ra_state
*ra
,
52 struct address_space
*mapping
, struct file
*filp
)
54 return __do_page_cache_readahead(mapping
, filp
,
55 ra
->start
, ra
->size
, ra
->async_size
);
59 * Turn a non-refcounted page (->_count == 0) into refcounted with
62 static inline void set_page_refcounted(struct page
*page
)
64 VM_BUG_ON_PAGE(PageTail(page
), page
);
65 VM_BUG_ON_PAGE(atomic_read(&page
->_count
), page
);
66 set_page_count(page
, 1);
69 static inline void __get_page_tail_foll(struct page
*page
,
73 * If we're getting a tail page, the elevated page->_count is
74 * required only in the head page and we will elevate the head
75 * page->_count and tail page->_mapcount.
77 * We elevate page_tail->_mapcount for tail pages to force
78 * page_tail->_count to be zero at all times to avoid getting
79 * false positives from get_page_unless_zero() with
80 * speculative page access (like in
81 * page_cache_get_speculative()) on tail pages.
83 VM_BUG_ON_PAGE(atomic_read(&compound_head(page
)->_count
) <= 0, page
);
85 atomic_inc(&compound_head(page
)->_count
);
86 get_huge_page_tail(page
);
90 * This is meant to be called as the FOLL_GET operation of
91 * follow_page() and it must be called while holding the proper PT
92 * lock while the pte (or pmd_trans_huge) is still mapping the page.
94 static inline void get_page_foll(struct page
*page
)
96 if (unlikely(PageTail(page
)))
98 * This is safe only because
99 * __split_huge_page_refcount() can't run under
100 * get_page_foll() because we hold the proper PT lock.
102 __get_page_tail_foll(page
, true);
105 * Getting a normal page or the head of a compound page
106 * requires to already have an elevated page->_count.
108 VM_BUG_ON_PAGE(atomic_read(&page
->_count
) <= 0, page
);
109 atomic_inc(&page
->_count
);
113 extern unsigned long highest_memmap_pfn
;
118 extern int isolate_lru_page(struct page
*page
);
119 extern void putback_lru_page(struct page
*page
);
120 extern bool zone_reclaimable(struct zone
*zone
);
122 #define lru_to_page(_head) (list_entry((_head)->prev, struct page, lru))
127 extern pmd_t
*mm_find_pmd(struct mm_struct
*mm
, unsigned long address
);
134 * Structure for holding the mostly immutable allocation parameters passed
135 * between functions involved in allocations, including the alloc_pages*
136 * family of functions.
138 * nodemask, migratetype and high_zoneidx are initialized only once in
139 * __alloc_pages_nodemask() and then never change.
141 * zonelist, preferred_zone and classzone_idx are set first in
142 * __alloc_pages_nodemask() for the fast path, and might be later changed
143 * in __alloc_pages_slowpath(). All other functions pass the whole strucure
144 * by a const pointer.
146 struct alloc_context
{
147 struct zonelist
*zonelist
;
148 nodemask_t
*nodemask
;
149 struct zone
*preferred_zone
;
152 enum zone_type high_zoneidx
;
153 bool spread_dirty_pages
;
157 * Locate the struct page for both the matching buddy in our
158 * pair (buddy1) and the combined O(n+1) page they form (page).
160 * 1) Any buddy B1 will have an order O twin B2 which satisfies
161 * the following equation:
163 * For example, if the starting buddy (buddy2) is #8 its order
165 * B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
167 * 2) Any buddy B will have an order O+1 parent P which
168 * satisfies the following equation:
171 * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
173 static inline unsigned long
174 __find_buddy_index(unsigned long page_idx
, unsigned int order
)
176 return page_idx
^ (1 << order
);
179 extern int __isolate_free_page(struct page
*page
, unsigned int order
);
180 extern void __free_pages_bootmem(struct page
*page
, unsigned long pfn
,
182 extern void prep_compound_page(struct page
*page
, unsigned int order
);
183 #ifdef CONFIG_MEMORY_FAILURE
184 extern bool is_free_buddy_page(struct page
*page
);
186 extern int user_min_free_kbytes
;
188 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
194 * compact_control is used to track pages being migrated and the free pages
195 * they are being migrated to during memory compaction. The free_pfn starts
196 * at the end of a zone and migrate_pfn begins at the start. Movable pages
197 * are moved to the end of a zone during a compaction run and the run
198 * completes when free_pfn <= migrate_pfn
200 struct compact_control
{
201 struct list_head freepages
; /* List of free pages to migrate to */
202 struct list_head migratepages
; /* List of pages being migrated */
203 unsigned long nr_freepages
; /* Number of isolated free pages */
204 unsigned long nr_migratepages
; /* Number of pages to migrate */
205 unsigned long free_pfn
; /* isolate_freepages search base */
206 unsigned long migrate_pfn
; /* isolate_migratepages search base */
207 unsigned long last_migrated_pfn
;/* Not yet flushed page being freed */
208 enum migrate_mode mode
; /* Async or sync migration mode */
209 bool ignore_skip_hint
; /* Scan blocks even if marked skip */
210 int order
; /* order a direct compactor needs */
211 const gfp_t gfp_mask
; /* gfp mask of a direct compactor */
212 const int alloc_flags
; /* alloc flags of a direct compactor */
213 const int classzone_idx
; /* zone index of a direct compactor */
215 int contended
; /* Signal need_sched() or lock
216 * contention detected during
222 isolate_freepages_range(struct compact_control
*cc
,
223 unsigned long start_pfn
, unsigned long end_pfn
);
225 isolate_migratepages_range(struct compact_control
*cc
,
226 unsigned long low_pfn
, unsigned long end_pfn
);
227 int find_suitable_fallback(struct free_area
*area
, unsigned int order
,
228 int migratetype
, bool only_stealable
, bool *can_steal
);
233 * This function returns the order of a free page in the buddy system. In
234 * general, page_zone(page)->lock must be held by the caller to prevent the
235 * page from being allocated in parallel and returning garbage as the order.
236 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
237 * page cannot be allocated or merged in parallel. Alternatively, it must
238 * handle invalid values gracefully, and use page_order_unsafe() below.
240 static inline unsigned int page_order(struct page
*page
)
242 /* PageBuddy() must be checked by the caller */
243 return page_private(page
);
247 * Like page_order(), but for callers who cannot afford to hold the zone lock.
248 * PageBuddy() should be checked first by the caller to minimize race window,
249 * and invalid values must be handled gracefully.
251 * READ_ONCE is used so that if the caller assigns the result into a local
252 * variable and e.g. tests it for valid range before using, the compiler cannot
253 * decide to remove the variable and inline the page_private(page) multiple
254 * times, potentially observing different values in the tests and the actual
257 #define page_order_unsafe(page) READ_ONCE(page_private(page))
259 static inline bool is_cow_mapping(vm_flags_t flags
)
261 return (flags
& (VM_SHARED
| VM_MAYWRITE
)) == VM_MAYWRITE
;
265 void __vma_link_list(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
266 struct vm_area_struct
*prev
, struct rb_node
*rb_parent
);
269 extern long populate_vma_page_range(struct vm_area_struct
*vma
,
270 unsigned long start
, unsigned long end
, int *nonblocking
);
271 extern void munlock_vma_pages_range(struct vm_area_struct
*vma
,
272 unsigned long start
, unsigned long end
);
273 static inline void munlock_vma_pages_all(struct vm_area_struct
*vma
)
275 munlock_vma_pages_range(vma
, vma
->vm_start
, vma
->vm_end
);
279 * must be called with vma's mmap_sem held for read or write, and page locked.
281 extern void mlock_vma_page(struct page
*page
);
282 extern unsigned int munlock_vma_page(struct page
*page
);
285 * Clear the page's PageMlocked(). This can be useful in a situation where
286 * we want to unconditionally remove a page from the pagecache -- e.g.,
287 * on truncation or freeing.
289 * It is legal to call this function for any page, mlocked or not.
290 * If called for a page that is still mapped by mlocked vmas, all we do
291 * is revert to lazy LRU behaviour -- semantics are not broken.
293 extern void clear_page_mlock(struct page
*page
);
296 * mlock_migrate_page - called only from migrate_misplaced_transhuge_page()
297 * (because that does not go through the full procedure of migration ptes):
298 * to migrate the Mlocked page flag; update statistics.
300 static inline void mlock_migrate_page(struct page
*newpage
, struct page
*page
)
302 if (TestClearPageMlocked(page
)) {
303 int nr_pages
= hpage_nr_pages(page
);
305 /* Holding pmd lock, no change in irq context: __mod is safe */
306 __mod_zone_page_state(page_zone(page
), NR_MLOCK
, -nr_pages
);
307 SetPageMlocked(newpage
);
308 __mod_zone_page_state(page_zone(newpage
), NR_MLOCK
, nr_pages
);
312 extern pmd_t
maybe_pmd_mkwrite(pmd_t pmd
, struct vm_area_struct
*vma
);
314 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
315 extern unsigned long vma_address(struct page
*page
,
316 struct vm_area_struct
*vma
);
318 #else /* !CONFIG_MMU */
319 static inline void clear_page_mlock(struct page
*page
) { }
320 static inline void mlock_vma_page(struct page
*page
) { }
321 static inline void mlock_migrate_page(struct page
*new, struct page
*old
) { }
323 #endif /* !CONFIG_MMU */
326 * Return the mem_map entry representing the 'offset' subpage within
327 * the maximally aligned gigantic page 'base'. Handle any discontiguity
328 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
330 static inline struct page
*mem_map_offset(struct page
*base
, int offset
)
332 if (unlikely(offset
>= MAX_ORDER_NR_PAGES
))
333 return nth_page(base
, offset
);
334 return base
+ offset
;
338 * Iterator over all subpages within the maximally aligned gigantic
339 * page 'base'. Handle any discontiguity in the mem_map.
341 static inline struct page
*mem_map_next(struct page
*iter
,
342 struct page
*base
, int offset
)
344 if (unlikely((offset
& (MAX_ORDER_NR_PAGES
- 1)) == 0)) {
345 unsigned long pfn
= page_to_pfn(base
) + offset
;
348 return pfn_to_page(pfn
);
354 * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
355 * so all functions starting at paging_init should be marked __init
356 * in those cases. SPARSEMEM, however, allows for memory hotplug,
357 * and alloc_bootmem_node is not used.
359 #ifdef CONFIG_SPARSEMEM
360 #define __paginginit __meminit
362 #define __paginginit __init
365 /* Memory initialisation debug and verification */
372 #ifdef CONFIG_DEBUG_MEMORY_INIT
374 extern int mminit_loglevel
;
376 #define mminit_dprintk(level, prefix, fmt, arg...) \
378 if (level < mminit_loglevel) { \
379 if (level <= MMINIT_WARNING) \
380 printk(KERN_WARNING "mminit::" prefix " " fmt, ##arg); \
382 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
386 extern void mminit_verify_pageflags_layout(void);
387 extern void mminit_verify_zonelist(void);
390 static inline void mminit_dprintk(enum mminit_level level
,
391 const char *prefix
, const char *fmt
, ...)
395 static inline void mminit_verify_pageflags_layout(void)
399 static inline void mminit_verify_zonelist(void)
402 #endif /* CONFIG_DEBUG_MEMORY_INIT */
404 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
405 #if defined(CONFIG_SPARSEMEM)
406 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn
,
407 unsigned long *end_pfn
);
409 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn
,
410 unsigned long *end_pfn
)
413 #endif /* CONFIG_SPARSEMEM */
415 #define ZONE_RECLAIM_NOSCAN -2
416 #define ZONE_RECLAIM_FULL -1
417 #define ZONE_RECLAIM_SOME 0
418 #define ZONE_RECLAIM_SUCCESS 1
420 extern int hwpoison_filter(struct page
*p
);
422 extern u32 hwpoison_filter_dev_major
;
423 extern u32 hwpoison_filter_dev_minor
;
424 extern u64 hwpoison_filter_flags_mask
;
425 extern u64 hwpoison_filter_flags_value
;
426 extern u64 hwpoison_filter_memcg
;
427 extern u32 hwpoison_filter_enable
;
429 extern unsigned long vm_mmap_pgoff(struct file
*, unsigned long,
430 unsigned long, unsigned long,
431 unsigned long, unsigned long);
433 extern void set_pageblock_order(void);
434 unsigned long reclaim_clean_pages_from_list(struct zone
*zone
,
435 struct list_head
*page_list
);
436 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
437 #define ALLOC_WMARK_MIN WMARK_MIN
438 #define ALLOC_WMARK_LOW WMARK_LOW
439 #define ALLOC_WMARK_HIGH WMARK_HIGH
440 #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
442 /* Mask to get the watermark bits */
443 #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
445 #define ALLOC_HARDER 0x10 /* try to alloc harder */
446 #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
447 #define ALLOC_CPUSET 0x40 /* check for correct cpuset */
448 #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
449 #define ALLOC_FAIR 0x100 /* fair zone allocation */
452 struct tlbflush_unmap_batch
;
454 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
455 void try_to_unmap_flush(void);
456 void try_to_unmap_flush_dirty(void);
458 static inline void try_to_unmap_flush(void)
461 static inline void try_to_unmap_flush_dirty(void)
465 #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
466 #endif /* __MM_INTERNAL_H */