blob: f5369cca34bdce447daa49e4d089c8eb1812200b [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/* internal.h: mm/ internal definitions
2 *
3 * Copyright (C) 2004 Red Hat, Inc. All Rights Reserved.
4 * Written by David Howells (dhowells@redhat.com)
5 *
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.
10 */
Nick Piggin0f8053a2006-03-22 00:08:33 -080011#ifndef __MM_INTERNAL_H
12#define __MM_INTERNAL_H
13
14#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070015
Jan Beulich42b77722008-07-23 21:27:10 -070016void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
17 unsigned long floor, unsigned long ceiling);
18
Nick Piggin7835e982006-03-22 00:08:40 -080019static inline void set_page_count(struct page *page, int v)
Nick Piggin77a8a782006-01-06 00:10:57 -080020{
Nick Piggin7835e982006-03-22 00:08:40 -080021 atomic_set(&page->_count, v);
22}
23
24/*
25 * Turn a non-refcounted page (->_count == 0) into refcounted with
26 * a count of one.
27 */
28static inline void set_page_refcounted(struct page *page)
29{
Qi Yongae1276b2008-02-04 22:29:27 -080030 VM_BUG_ON(PageTail(page));
Nick Piggin725d7042006-09-25 23:30:55 -070031 VM_BUG_ON(atomic_read(&page->_count));
Nick Piggin77a8a782006-01-06 00:10:57 -080032 set_page_count(page, 1);
Nick Piggin77a8a782006-01-06 00:10:57 -080033}
34
Nick Piggin0f8053a2006-03-22 00:08:33 -080035static inline void __put_page(struct page *page)
36{
37 atomic_dec(&page->_count);
38}
39
Andrea Arcangeli70b50f92011-11-02 13:36:59 -070040static inline void __get_page_tail_foll(struct page *page,
41 bool get_page_head)
42{
43 /*
44 * If we're getting a tail page, the elevated page->_count is
45 * required only in the head page and we will elevate the head
46 * page->_count and tail page->_mapcount.
47 *
48 * We elevate page_tail->_mapcount for tail pages to force
49 * page_tail->_count to be zero at all times to avoid getting
50 * false positives from get_page_unless_zero() with
51 * speculative page access (like in
52 * page_cache_get_speculative()) on tail pages.
53 */
54 VM_BUG_ON(atomic_read(&page->first_page->_count) <= 0);
55 VM_BUG_ON(atomic_read(&page->_count) != 0);
56 VM_BUG_ON(page_mapcount(page) < 0);
57 if (get_page_head)
58 atomic_inc(&page->first_page->_count);
59 atomic_inc(&page->_mapcount);
60}
61
62/*
63 * This is meant to be called as the FOLL_GET operation of
64 * follow_page() and it must be called while holding the proper PT
65 * lock while the pte (or pmd_trans_huge) is still mapping the page.
66 */
67static inline void get_page_foll(struct page *page)
68{
69 if (unlikely(PageTail(page)))
70 /*
71 * This is safe only because
72 * __split_huge_page_refcount() can't run under
73 * get_page_foll() because we hold the proper PT lock.
74 */
75 __get_page_tail_foll(page, true);
76 else {
77 /*
78 * Getting a normal page or the head of a compound page
79 * requires to already have an elevated page->_count.
80 */
81 VM_BUG_ON(atomic_read(&page->_count) <= 0);
82 atomic_inc(&page->_count);
83 }
84}
85
Hugh Dickins03f64622009-09-21 17:03:35 -070086extern unsigned long highest_memmap_pfn;
87
Lee Schermerhorn894bc312008-10-18 20:26:39 -070088/*
89 * in mm/vmscan.c:
90 */
Nick Piggin62695a82008-10-18 20:26:09 -070091extern int isolate_lru_page(struct page *page);
Lee Schermerhorn894bc312008-10-18 20:26:39 -070092extern void putback_lru_page(struct page *page);
Lisa Due0935212013-09-11 14:22:36 -070093extern unsigned long zone_reclaimable_pages(struct zone *zone);
94extern bool zone_reclaimable(struct zone *zone);
Nick Piggin62695a82008-10-18 20:26:09 -070095
Lee Schermerhorn894bc312008-10-18 20:26:39 -070096/*
97 * in mm/page_alloc.c
98 */
Yasunori Goto0c0a4a52008-04-28 02:13:34 -070099extern void __free_pages_bootmem(struct page *page, unsigned int order);
Wu Fengguang20a03072009-06-16 15:32:22 -0700100extern void prep_compound_page(struct page *page, unsigned long order);
Wu Fengguang8d22ba12009-12-16 12:19:58 +0100101#ifdef CONFIG_MEMORY_FAILURE
102extern bool is_free_buddy_page(struct page *page);
103#endif
Wu Fengguang20a03072009-06-16 15:32:22 -0700104
Michal Nazarewicz02ff1de2011-12-29 13:09:50 +0100105#if defined CONFIG_COMPACTION || defined CONFIG_CMA
106
107/*
108 * in mm/compaction.c
109 */
110/*
111 * compact_control is used to track pages being migrated and the free pages
112 * they are being migrated to during memory compaction. The free_pfn starts
113 * at the end of a zone and migrate_pfn begins at the start. Movable pages
114 * are moved to the end of a zone during a compaction run and the run
115 * completes when free_pfn <= migrate_pfn
116 */
117struct compact_control {
118 struct list_head freepages; /* List of free pages to migrate to */
119 struct list_head migratepages; /* List of pages being migrated */
120 unsigned long nr_freepages; /* Number of isolated free pages */
121 unsigned long nr_migratepages; /* Number of pages to migrate */
122 unsigned long free_pfn; /* isolate_freepages search base */
123 unsigned long migrate_pfn; /* isolate_migratepages search base */
124 bool sync; /* Synchronous migration */
Mel Gorman6b506642012-10-08 16:32:41 -0700125 bool ignore_skip_hint; /* Scan blocks even if marked skip */
Mel Gormand58f1f22012-10-08 16:32:45 -0700126 bool finished_update_free; /* True when the zone cached pfns are
127 * no longer being updated
128 */
129 bool finished_update_migrate;
Michal Nazarewicz02ff1de2011-12-29 13:09:50 +0100130
131 int order; /* order a direct compactor needs */
132 int migratetype; /* MOVABLE, RECLAIMABLE etc */
133 struct zone *zone;
Shaohua Li3cbf3512012-10-08 16:32:27 -0700134 bool contended; /* True if a lock was contended */
Michal Nazarewicz02ff1de2011-12-29 13:09:50 +0100135};
136
137unsigned long
Mel Gorman6b506642012-10-08 16:32:41 -0700138isolate_freepages_range(struct compact_control *cc,
139 unsigned long start_pfn, unsigned long end_pfn);
Michal Nazarewicz02ff1de2011-12-29 13:09:50 +0100140unsigned long
141isolate_migratepages_range(struct zone *zone, struct compact_control *cc,
Minchan Kimd05b4522012-10-08 16:33:48 -0700142 unsigned long low_pfn, unsigned long end_pfn, bool unevictable);
Michal Nazarewicz02ff1de2011-12-29 13:09:50 +0100143
144#endif
Nick Piggin0f8053a2006-03-22 00:08:33 -0800145
Mel Gorman48f13bf2007-10-16 01:26:10 -0700146/*
147 * function for dealing with page's order in buddy system.
148 * zone->lock is already acquired when we use these.
149 * So, we don't need atomic page->flags operations here.
150 */
151static inline unsigned long page_order(struct page *page)
152{
KAMEZAWA Hiroyuki572438f2010-10-26 14:22:08 -0700153 /* PageBuddy() must be checked by the caller */
Mel Gorman48f13bf2007-10-16 01:26:10 -0700154 return page_private(page);
155}
Alexander van Heukelumb5a0e012008-02-23 15:24:06 -0800156
Namhyung Kim6038def2011-05-24 17:11:22 -0700157/* mm/util.c */
158void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
159 struct vm_area_struct *prev, struct rb_node *rb_parent);
160
Hugh Dickinsaf8e3352009-12-14 17:58:59 -0800161#ifdef CONFIG_MMU
162extern long mlock_vma_pages_range(struct vm_area_struct *vma,
163 unsigned long start, unsigned long end);
164extern void munlock_vma_pages_range(struct vm_area_struct *vma,
165 unsigned long start, unsigned long end);
166static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
167{
168 munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
169}
170
Nick Pigginb291f002008-10-18 20:26:44 -0700171/*
172 * Called only in fault path via page_evictable() for a new page
173 * to determine if it's being mapped into a LOCKED vma.
174 * If so, mark page as mlocked.
175 */
176static inline int is_mlocked_vma(struct vm_area_struct *vma, struct page *page)
177{
178 VM_BUG_ON(PageLRU(page));
179
180 if (likely((vma->vm_flags & (VM_LOCKED | VM_SPECIAL)) != VM_LOCKED))
181 return 0;
182
Nick Piggin5344b7e2008-10-18 20:26:51 -0700183 if (!TestSetPageMlocked(page)) {
184 inc_zone_page_state(page, NR_MLOCK);
185 count_vm_event(UNEVICTABLE_PGMLOCKED);
186 }
Nick Pigginb291f002008-10-18 20:26:44 -0700187 return 1;
188}
189
190/*
Hugh Dickins73848b42009-12-14 17:59:22 -0800191 * must be called with vma's mmap_sem held for read or write, and page locked.
Nick Pigginb291f002008-10-18 20:26:44 -0700192 */
193extern void mlock_vma_page(struct page *page);
Hugh Dickins73848b42009-12-14 17:59:22 -0800194extern void munlock_vma_page(struct page *page);
Nick Pigginb291f002008-10-18 20:26:44 -0700195
196/*
197 * Clear the page's PageMlocked(). This can be useful in a situation where
198 * we want to unconditionally remove a page from the pagecache -- e.g.,
199 * on truncation or freeing.
200 *
201 * It is legal to call this function for any page, mlocked or not.
202 * If called for a page that is still mapped by mlocked vmas, all we do
203 * is revert to lazy LRU behaviour -- semantics are not broken.
204 */
205extern void __clear_page_mlock(struct page *page);
206static inline void clear_page_mlock(struct page *page)
207{
208 if (unlikely(TestClearPageMlocked(page)))
209 __clear_page_mlock(page);
210}
211
212/*
213 * mlock_migrate_page - called only from migrate_page_copy() to
Nick Piggin5344b7e2008-10-18 20:26:51 -0700214 * migrate the Mlocked page flag; update statistics.
Nick Pigginb291f002008-10-18 20:26:44 -0700215 */
216static inline void mlock_migrate_page(struct page *newpage, struct page *page)
217{
Nick Piggin5344b7e2008-10-18 20:26:51 -0700218 if (TestClearPageMlocked(page)) {
219 unsigned long flags;
220
221 local_irq_save(flags);
222 __dec_zone_page_state(page, NR_MLOCK);
Nick Pigginb291f002008-10-18 20:26:44 -0700223 SetPageMlocked(newpage);
Nick Piggin5344b7e2008-10-18 20:26:51 -0700224 __inc_zone_page_state(newpage, NR_MLOCK);
225 local_irq_restore(flags);
226 }
Nick Pigginb291f002008-10-18 20:26:44 -0700227}
228
Andrea Arcangeli71e3aac2011-01-13 15:46:52 -0800229#ifdef CONFIG_TRANSPARENT_HUGEPAGE
230extern unsigned long vma_address(struct page *page,
231 struct vm_area_struct *vma);
232#endif
Hugh Dickinsaf8e3352009-12-14 17:58:59 -0800233#else /* !CONFIG_MMU */
Nick Pigginb291f002008-10-18 20:26:44 -0700234static inline int is_mlocked_vma(struct vm_area_struct *v, struct page *p)
235{
236 return 0;
237}
238static inline void clear_page_mlock(struct page *page) { }
239static inline void mlock_vma_page(struct page *page) { }
240static inline void mlock_migrate_page(struct page *new, struct page *old) { }
241
Hugh Dickinsaf8e3352009-12-14 17:58:59 -0800242#endif /* !CONFIG_MMU */
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700243
Alexander van Heukelumb5a0e012008-02-23 15:24:06 -0800244/*
Andy Whitcroft69d177c2008-11-06 12:53:26 -0800245 * Return the mem_map entry representing the 'offset' subpage within
246 * the maximally aligned gigantic page 'base'. Handle any discontiguity
247 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
248 */
249static inline struct page *mem_map_offset(struct page *base, int offset)
250{
251 if (unlikely(offset >= MAX_ORDER_NR_PAGES))
252 return pfn_to_page(page_to_pfn(base) + offset);
253 return base + offset;
254}
255
256/*
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300257 * Iterator over all subpages within the maximally aligned gigantic
Andy Whitcroft69d177c2008-11-06 12:53:26 -0800258 * page 'base'. Handle any discontiguity in the mem_map.
259 */
260static inline struct page *mem_map_next(struct page *iter,
261 struct page *base, int offset)
262{
263 if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
264 unsigned long pfn = page_to_pfn(base) + offset;
265 if (!pfn_valid(pfn))
266 return NULL;
267 return pfn_to_page(pfn);
268 }
269 return iter + 1;
270}
271
272/*
Alexander van Heukelumb5a0e012008-02-23 15:24:06 -0800273 * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
274 * so all functions starting at paging_init should be marked __init
275 * in those cases. SPARSEMEM, however, allows for memory hotplug,
276 * and alloc_bootmem_node is not used.
277 */
278#ifdef CONFIG_SPARSEMEM
279#define __paginginit __meminit
280#else
281#define __paginginit __init
282#endif
283
Mel Gorman6b74ab92008-07-23 21:26:49 -0700284/* Memory initialisation debug and verification */
285enum mminit_level {
286 MMINIT_WARNING,
287 MMINIT_VERIFY,
288 MMINIT_TRACE
289};
290
291#ifdef CONFIG_DEBUG_MEMORY_INIT
292
293extern int mminit_loglevel;
294
295#define mminit_dprintk(level, prefix, fmt, arg...) \
296do { \
297 if (level < mminit_loglevel) { \
298 printk(level <= MMINIT_WARNING ? KERN_WARNING : KERN_DEBUG); \
299 printk(KERN_CONT "mminit::" prefix " " fmt, ##arg); \
300 } \
301} while (0)
302
Mel Gorman708614e2008-07-23 21:26:51 -0700303extern void mminit_verify_pageflags_layout(void);
304extern void mminit_verify_page_links(struct page *page,
305 enum zone_type zone, unsigned long nid, unsigned long pfn);
Mel Gorman68ad8df2008-07-23 21:26:52 -0700306extern void mminit_verify_zonelist(void);
Mel Gorman708614e2008-07-23 21:26:51 -0700307
Mel Gorman6b74ab92008-07-23 21:26:49 -0700308#else
309
310static inline void mminit_dprintk(enum mminit_level level,
311 const char *prefix, const char *fmt, ...)
312{
313}
314
Mel Gorman708614e2008-07-23 21:26:51 -0700315static inline void mminit_verify_pageflags_layout(void)
316{
317}
318
319static inline void mminit_verify_page_links(struct page *page,
320 enum zone_type zone, unsigned long nid, unsigned long pfn)
321{
322}
Mel Gorman68ad8df2008-07-23 21:26:52 -0700323
324static inline void mminit_verify_zonelist(void)
325{
326}
Mel Gorman6b74ab92008-07-23 21:26:49 -0700327#endif /* CONFIG_DEBUG_MEMORY_INIT */
Mel Gorman2dbb51c2008-07-23 21:26:52 -0700328
329/* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
330#if defined(CONFIG_SPARSEMEM)
331extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
332 unsigned long *end_pfn);
333#else
334static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
335 unsigned long *end_pfn)
336{
337}
338#endif /* CONFIG_SPARSEMEM */
339
Mel Gormanfa5e0842009-06-16 15:33:22 -0700340#define ZONE_RECLAIM_NOSCAN -2
341#define ZONE_RECLAIM_FULL -1
342#define ZONE_RECLAIM_SOME 0
343#define ZONE_RECLAIM_SUCCESS 1
Nick Piggin0f8053a2006-03-22 00:08:33 -0800344#endif
Wu Fengguang7c116f22009-12-16 12:19:59 +0100345
Wu Fengguang31d3d342009-12-16 12:19:59 +0100346extern int hwpoison_filter(struct page *p);
347
Wu Fengguang7c116f22009-12-16 12:19:59 +0100348extern u32 hwpoison_filter_dev_major;
349extern u32 hwpoison_filter_dev_minor;
Wu Fengguang478c5ff2009-12-16 12:19:59 +0100350extern u64 hwpoison_filter_flags_mask;
351extern u64 hwpoison_filter_flags_value;
Andi Kleen4fd466e2009-12-16 12:19:59 +0100352extern u64 hwpoison_filter_memcg;
Haicheng Li1bfe5fe2009-12-16 12:19:59 +0100353extern u32 hwpoison_filter_enable;
Bartlomiej Zolnierkiewicz19644312012-10-08 16:32:05 -0700354
355/* The ALLOC_WMARK bits are used as an index to zone->watermark */
356#define ALLOC_WMARK_MIN WMARK_MIN
357#define ALLOC_WMARK_LOW WMARK_LOW
358#define ALLOC_WMARK_HIGH WMARK_HIGH
359#define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
360
361/* Mask to get the watermark bits */
362#define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
363
364#define ALLOC_HARDER 0x10 /* try to alloc harder */
365#define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
366#define ALLOC_CPUSET 0x40 /* check for correct cpuset */
367#define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
Minchan Kimaab50f12012-10-08 16:31:55 -0700368
369unsigned long reclaim_clean_pages_from_list(struct zone *zone,
370 struct list_head *page_list);