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|\
28 /* The GFP flags allowed during early boot */
29 #define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_RECLAIM|__GFP_IO|__GFP_FS))
31 /* Control allocation cpuset and node placement constraints */
32 #define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE)
34 /* Do not use these with a slab allocator */
35 #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
37 void free_pgtables(struct mmu_gather
*tlb
, struct vm_area_struct
*start_vma
,
38 unsigned long floor
, unsigned long ceiling
);
40 static inline void set_page_count(struct page
*page
, int v
)
42 atomic_set(&page
->_count
, v
);
45 extern int __do_page_cache_readahead(struct address_space
*mapping
,
46 struct file
*filp
, pgoff_t offset
, unsigned long nr_to_read
,
47 unsigned long lookahead_size
);
50 * Submit IO for the read-ahead request in file_ra_state.
52 static inline unsigned long ra_submit(struct file_ra_state
*ra
,
53 struct address_space
*mapping
, struct file
*filp
)
55 return __do_page_cache_readahead(mapping
, filp
,
56 ra
->start
, ra
->size
, ra
->async_size
);
60 * Turn a non-refcounted page (->_count == 0) into refcounted with
63 static inline void set_page_refcounted(struct page
*page
)
65 VM_BUG_ON_PAGE(PageTail(page
), page
);
66 VM_BUG_ON_PAGE(atomic_read(&page
->_count
), page
);
67 set_page_count(page
, 1);
70 static inline void __get_page_tail_foll(struct page
*page
,
74 * If we're getting a tail page, the elevated page->_count is
75 * required only in the head page and we will elevate the head
76 * page->_count and tail page->_mapcount.
78 * We elevate page_tail->_mapcount for tail pages to force
79 * page_tail->_count to be zero at all times to avoid getting
80 * false positives from get_page_unless_zero() with
81 * speculative page access (like in
82 * page_cache_get_speculative()) on tail pages.
84 VM_BUG_ON_PAGE(atomic_read(&compound_head(page
)->_count
) <= 0, page
);
86 atomic_inc(&compound_head(page
)->_count
);
87 get_huge_page_tail(page
);
91 * This is meant to be called as the FOLL_GET operation of
92 * follow_page() and it must be called while holding the proper PT
93 * lock while the pte (or pmd_trans_huge) is still mapping the page.
95 static inline void get_page_foll(struct page
*page
)
97 if (unlikely(PageTail(page
)))
99 * This is safe only because
100 * __split_huge_page_refcount() can't run under
101 * get_page_foll() because we hold the proper PT lock.
103 __get_page_tail_foll(page
, true);
106 * Getting a normal page or the head of a compound page
107 * requires to already have an elevated page->_count.
109 VM_BUG_ON_PAGE(atomic_read(&page
->_count
) <= 0, page
);
110 atomic_inc(&page
->_count
);
114 extern unsigned long highest_memmap_pfn
;
119 extern int isolate_lru_page(struct page
*page
);
120 extern void putback_lru_page(struct page
*page
);
121 extern bool zone_reclaimable(struct zone
*zone
);
126 extern pmd_t
*mm_find_pmd(struct mm_struct
*mm
, unsigned long address
);
133 * Structure for holding the mostly immutable allocation parameters passed
134 * between functions involved in allocations, including the alloc_pages*
135 * family of functions.
137 * nodemask, migratetype and high_zoneidx are initialized only once in
138 * __alloc_pages_nodemask() and then never change.
140 * zonelist, preferred_zone and classzone_idx are set first in
141 * __alloc_pages_nodemask() for the fast path, and might be later changed
142 * in __alloc_pages_slowpath(). All other functions pass the whole strucure
143 * by a const pointer.
145 struct alloc_context
{
146 struct zonelist
*zonelist
;
147 nodemask_t
*nodemask
;
148 struct zone
*preferred_zone
;
151 enum zone_type high_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_index(unsigned long page_idx
, unsigned int order
)
175 return page_idx
^ (1 << order
);
178 extern int __isolate_free_page(struct page
*page
, unsigned int order
);
179 extern void __free_pages_bootmem(struct page
*page
, unsigned long pfn
,
181 extern void prep_compound_page(struct page
*page
, unsigned int order
);
182 extern void put_page_freelist(struct page
*page
);
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 bool direct_compaction
; /* False from kcompactd or /proc/... */
211 int order
; /* order a direct compactor needs */
212 const gfp_t gfp_mask
; /* gfp mask of a direct compactor */
213 const int alloc_flags
; /* alloc flags of a direct compactor */
214 const int classzone_idx
; /* zone index of a direct compactor */
216 int contended
; /* Signal need_sched() or lock
217 * contention detected during
223 isolate_freepages_range(struct compact_control
*cc
,
224 unsigned long start_pfn
, unsigned long end_pfn
);
226 isolate_migratepages_range(struct compact_control
*cc
,
227 unsigned long low_pfn
, unsigned long end_pfn
);
228 int find_suitable_fallback(struct free_area
*area
, unsigned int order
,
229 int migratetype
, bool only_stealable
, bool *can_steal
);
234 * This function returns the order of a free page in the buddy system. In
235 * general, page_zone(page)->lock must be held by the caller to prevent the
236 * page from being allocated in parallel and returning garbage as the order.
237 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
238 * page cannot be allocated or merged in parallel. Alternatively, it must
239 * handle invalid values gracefully, and use page_order_unsafe() below.
241 static inline unsigned int page_order(struct page
*page
)
243 /* PageBuddy() must be checked by the caller */
244 return page_private(page
);
248 * Like page_order(), but for callers who cannot afford to hold the zone lock.
249 * PageBuddy() should be checked first by the caller to minimize race window,
250 * and invalid values must be handled gracefully.
252 * READ_ONCE is used so that if the caller assigns the result into a local
253 * variable and e.g. tests it for valid range before using, the compiler cannot
254 * decide to remove the variable and inline the page_private(page) multiple
255 * times, potentially observing different values in the tests and the actual
258 #define page_order_unsafe(page) READ_ONCE(page_private(page))
260 static inline bool is_cow_mapping(vm_flags_t flags
)
262 return (flags
& (VM_SHARED
| VM_MAYWRITE
)) == VM_MAYWRITE
;
266 void __vma_link_list(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
267 struct vm_area_struct
*prev
, struct rb_node
*rb_parent
);
270 extern long populate_vma_page_range(struct vm_area_struct
*vma
,
271 unsigned long start
, unsigned long end
, int *nonblocking
);
272 extern void munlock_vma_pages_range(struct vm_area_struct
*vma
,
273 unsigned long start
, unsigned long end
);
274 static inline void munlock_vma_pages_all(struct vm_area_struct
*vma
)
276 munlock_vma_pages_range(vma
, vma
->vm_start
, vma
->vm_end
);
280 * must be called with vma's mmap_sem held for read or write, and page locked.
282 extern void mlock_vma_page(struct page
*page
);
283 extern unsigned int munlock_vma_page(struct page
*page
);
286 * Clear the page's PageMlocked(). This can be useful in a situation where
287 * we want to unconditionally remove a page from the pagecache -- e.g.,
288 * on truncation or freeing.
290 * It is legal to call this function for any page, mlocked or not.
291 * If called for a page that is still mapped by mlocked vmas, all we do
292 * is revert to lazy LRU behaviour -- semantics are not broken.
294 extern void clear_page_mlock(struct page
*page
);
297 * mlock_migrate_page - called only from migrate_misplaced_transhuge_page()
298 * (because that does not go through the full procedure of migration ptes):
299 * to migrate the Mlocked page flag; update statistics.
301 static inline void mlock_migrate_page(struct page
*newpage
, struct page
*page
)
303 if (TestClearPageMlocked(page
)) {
304 int nr_pages
= hpage_nr_pages(page
);
306 /* Holding pmd lock, no change in irq context: __mod is safe */
307 __mod_zone_page_state(page_zone(page
), NR_MLOCK
, -nr_pages
);
308 SetPageMlocked(newpage
);
309 __mod_zone_page_state(page_zone(newpage
), NR_MLOCK
, nr_pages
);
313 extern pmd_t
maybe_pmd_mkwrite(pmd_t pmd
, struct vm_area_struct
*vma
);
315 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
316 extern unsigned long vma_address(struct page
*page
,
317 struct vm_area_struct
*vma
);
319 #else /* !CONFIG_MMU */
320 static inline void clear_page_mlock(struct page
*page
) { }
321 static inline void mlock_vma_page(struct page
*page
) { }
322 static inline void mlock_migrate_page(struct page
*new, struct page
*old
) { }
324 #endif /* !CONFIG_MMU */
327 * Return the mem_map entry representing the 'offset' subpage within
328 * the maximally aligned gigantic page 'base'. Handle any discontiguity
329 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
331 static inline struct page
*mem_map_offset(struct page
*base
, int offset
)
333 if (unlikely(offset
>= MAX_ORDER_NR_PAGES
))
334 return nth_page(base
, offset
);
335 return base
+ offset
;
339 * Iterator over all subpages within the maximally aligned gigantic
340 * page 'base'. Handle any discontiguity in the mem_map.
342 static inline struct page
*mem_map_next(struct page
*iter
,
343 struct page
*base
, int offset
)
345 if (unlikely((offset
& (MAX_ORDER_NR_PAGES
- 1)) == 0)) {
346 unsigned long pfn
= page_to_pfn(base
) + offset
;
349 return pfn_to_page(pfn
);
355 * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
356 * so all functions starting at paging_init should be marked __init
357 * in those cases. SPARSEMEM, however, allows for memory hotplug,
358 * and alloc_bootmem_node is not used.
360 #ifdef CONFIG_SPARSEMEM
361 #define __paginginit __meminit
363 #define __paginginit __init
366 /* Memory initialisation debug and verification */
373 #ifdef CONFIG_DEBUG_MEMORY_INIT
375 extern int mminit_loglevel
;
377 #define mminit_dprintk(level, prefix, fmt, arg...) \
379 if (level < mminit_loglevel) { \
380 if (level <= MMINIT_WARNING) \
381 printk(KERN_WARNING "mminit::" prefix " " fmt, ##arg); \
383 printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
387 extern void mminit_verify_pageflags_layout(void);
388 extern void mminit_verify_zonelist(void);
391 static inline void mminit_dprintk(enum mminit_level level
,
392 const char *prefix
, const char *fmt
, ...)
396 static inline void mminit_verify_pageflags_layout(void)
400 static inline void mminit_verify_zonelist(void)
403 #endif /* CONFIG_DEBUG_MEMORY_INIT */
405 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
406 #if defined(CONFIG_SPARSEMEM)
407 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn
,
408 unsigned long *end_pfn
);
410 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn
,
411 unsigned long *end_pfn
)
414 #endif /* CONFIG_SPARSEMEM */
416 #define ZONE_RECLAIM_NOSCAN -2
417 #define ZONE_RECLAIM_FULL -1
418 #define ZONE_RECLAIM_SOME 0
419 #define ZONE_RECLAIM_SUCCESS 1
421 extern int hwpoison_filter(struct page
*p
);
423 extern u32 hwpoison_filter_dev_major
;
424 extern u32 hwpoison_filter_dev_minor
;
425 extern u64 hwpoison_filter_flags_mask
;
426 extern u64 hwpoison_filter_flags_value
;
427 extern u64 hwpoison_filter_memcg
;
428 extern u32 hwpoison_filter_enable
;
430 extern unsigned long vm_mmap_pgoff(struct file
*, unsigned long,
431 unsigned long, unsigned long,
432 unsigned long, unsigned long);
434 extern void set_pageblock_order(void);
435 unsigned long reclaim_clean_pages_from_list(struct zone
*zone
,
436 struct list_head
*page_list
);
437 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
438 #define ALLOC_WMARK_MIN WMARK_MIN
439 #define ALLOC_WMARK_LOW WMARK_LOW
440 #define ALLOC_WMARK_HIGH WMARK_HIGH
441 #define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
443 /* Mask to get the watermark bits */
444 #define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
446 #define ALLOC_HARDER 0x10 /* try to alloc harder */
447 #define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
448 #define ALLOC_CPUSET 0x40 /* check for correct cpuset */
449 #define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
450 #define ALLOC_FAIR 0x100 /* fair zone allocation */
453 struct tlbflush_unmap_batch
;
455 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
456 void try_to_unmap_flush(void);
457 void try_to_unmap_flush_dirty(void);
458 void flush_tlb_batched_pending(struct mm_struct
*mm
);
460 static inline void try_to_unmap_flush(void)
463 static inline void try_to_unmap_flush_dirty(void)
466 static inline void flush_tlb_batched_pending(struct mm_struct
*mm
)
469 #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
470 #endif /* __MM_INTERNAL_H */