blob: 8562de0a5197d31e3621d10db8b0c01d23f44f83 [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);
Nick Piggin62695a82008-10-18 20:26:09 -070093
Lee Schermerhorn894bc312008-10-18 20:26:39 -070094/*
Bob Liu62190492012-12-11 16:00:37 -080095 * in mm/rmap.c:
96 */
97extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
98
99/*
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700100 * in mm/page_alloc.c
101 */
Yasunori Goto0c0a4a52008-04-28 02:13:34 -0700102extern void __free_pages_bootmem(struct page *page, unsigned int order);
Wu Fengguang20a03072009-06-16 15:32:22 -0700103extern void prep_compound_page(struct page *page, unsigned long order);
Wu Fengguang8d22ba12009-12-16 12:19:58 +0100104#ifdef CONFIG_MEMORY_FAILURE
105extern bool is_free_buddy_page(struct page *page);
106#endif
Wu Fengguang20a03072009-06-16 15:32:22 -0700107
Michal Nazarewiczff9543f2011-12-29 13:09:50 +0100108#if defined CONFIG_COMPACTION || defined CONFIG_CMA
109
110/*
111 * in mm/compaction.c
112 */
113/*
114 * compact_control is used to track pages being migrated and the free pages
115 * they are being migrated to during memory compaction. The free_pfn starts
116 * at the end of a zone and migrate_pfn begins at the start. Movable pages
117 * are moved to the end of a zone during a compaction run and the run
118 * completes when free_pfn <= migrate_pfn
119 */
120struct compact_control {
121 struct list_head freepages; /* List of free pages to migrate to */
122 struct list_head migratepages; /* List of pages being migrated */
123 unsigned long nr_freepages; /* Number of isolated free pages */
124 unsigned long nr_migratepages; /* Number of pages to migrate */
125 unsigned long free_pfn; /* isolate_freepages search base */
126 unsigned long migrate_pfn; /* isolate_migratepages search base */
Linus Torvalds68e3e922012-06-03 20:05:57 -0700127 bool sync; /* Synchronous migration */
Mel Gormanbb13ffe2012-10-08 16:32:41 -0700128 bool ignore_skip_hint; /* Scan blocks even if marked skip */
Mel Gormanc89511a2012-10-08 16:32:45 -0700129 bool finished_update_free; /* True when the zone cached pfns are
130 * no longer being updated
131 */
132 bool finished_update_migrate;
Michal Nazarewiczff9543f2011-12-29 13:09:50 +0100133
134 int order; /* order a direct compactor needs */
135 int migratetype; /* MOVABLE, RECLAIMABLE etc */
136 struct zone *zone;
Shaohua Lie64c5232012-10-08 16:32:27 -0700137 bool contended; /* True if a lock was contended */
Michal Nazarewiczff9543f2011-12-29 13:09:50 +0100138};
139
140unsigned long
Mel Gormanbb13ffe2012-10-08 16:32:41 -0700141isolate_freepages_range(struct compact_control *cc,
142 unsigned long start_pfn, unsigned long end_pfn);
Michal Nazarewiczff9543f2011-12-29 13:09:50 +0100143unsigned long
144isolate_migratepages_range(struct zone *zone, struct compact_control *cc,
Minchan Kime46a2872012-10-08 16:33:48 -0700145 unsigned long low_pfn, unsigned long end_pfn, bool unevictable);
Michal Nazarewiczff9543f2011-12-29 13:09:50 +0100146
147#endif
Nick Piggin0f8053a2006-03-22 00:08:33 -0800148
Mel Gorman48f13bf2007-10-16 01:26:10 -0700149/*
150 * function for dealing with page's order in buddy system.
151 * zone->lock is already acquired when we use these.
152 * So, we don't need atomic page->flags operations here.
153 */
154static inline unsigned long page_order(struct page *page)
155{
KAMEZAWA Hiroyuki572438f2010-10-26 14:22:08 -0700156 /* PageBuddy() must be checked by the caller */
Mel Gorman48f13bf2007-10-16 01:26:10 -0700157 return page_private(page);
158}
Alexander van Heukelumb5a0e012008-02-23 15:24:06 -0800159
Namhyung Kim6038def2011-05-24 17:11:22 -0700160/* mm/util.c */
161void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
162 struct vm_area_struct *prev, struct rb_node *rb_parent);
163
Hugh Dickinsaf8e3352009-12-14 17:58:59 -0800164#ifdef CONFIG_MMU
Michel Lespinassecea10a12013-02-22 16:32:44 -0800165extern long __mlock_vma_pages_range(struct vm_area_struct *vma,
166 unsigned long start, unsigned long end, int *nonblocking);
Hugh Dickinsaf8e3352009-12-14 17:58:59 -0800167extern void munlock_vma_pages_range(struct vm_area_struct *vma,
168 unsigned long start, unsigned long end);
169static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
170{
171 munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
172}
173
Nick Pigginb291f002008-10-18 20:26:44 -0700174/*
Hugh Dickins39b5f292012-10-08 16:33:18 -0700175 * Called only in fault path, to determine if a new page is being
176 * mapped into a LOCKED vma. If it is, mark page as mlocked.
Nick Pigginb291f002008-10-18 20:26:44 -0700177 */
Ying Han096a7cf2012-05-29 15:06:25 -0700178static inline int mlocked_vma_newpage(struct vm_area_struct *vma,
179 struct page *page)
Nick Pigginb291f002008-10-18 20:26:44 -0700180{
181 VM_BUG_ON(PageLRU(page));
182
183 if (likely((vma->vm_flags & (VM_LOCKED | VM_SPECIAL)) != VM_LOCKED))
184 return 0;
185
Nick Piggin5344b7e2008-10-18 20:26:51 -0700186 if (!TestSetPageMlocked(page)) {
David Rientjes8449d212012-10-08 16:34:06 -0700187 mod_zone_page_state(page_zone(page), NR_MLOCK,
188 hpage_nr_pages(page));
Nick Piggin5344b7e2008-10-18 20:26:51 -0700189 count_vm_event(UNEVICTABLE_PGMLOCKED);
190 }
Nick Pigginb291f002008-10-18 20:26:44 -0700191 return 1;
192}
193
194/*
Hugh Dickins73848b42009-12-14 17:59:22 -0800195 * must be called with vma's mmap_sem held for read or write, and page locked.
Nick Pigginb291f002008-10-18 20:26:44 -0700196 */
197extern void mlock_vma_page(struct page *page);
Michel Lespinasseff6a6da2013-02-27 17:02:44 -0800198extern unsigned int munlock_vma_page(struct page *page);
Nick Pigginb291f002008-10-18 20:26:44 -0700199
200/*
201 * Clear the page's PageMlocked(). This can be useful in a situation where
202 * we want to unconditionally remove a page from the pagecache -- e.g.,
203 * on truncation or freeing.
204 *
205 * It is legal to call this function for any page, mlocked or not.
206 * If called for a page that is still mapped by mlocked vmas, all we do
207 * is revert to lazy LRU behaviour -- semantics are not broken.
208 */
Hugh Dickinse6c509f2012-10-08 16:33:19 -0700209extern void clear_page_mlock(struct page *page);
Nick Pigginb291f002008-10-18 20:26:44 -0700210
211/*
212 * mlock_migrate_page - called only from migrate_page_copy() to
Nick Piggin5344b7e2008-10-18 20:26:51 -0700213 * migrate the Mlocked page flag; update statistics.
Nick Pigginb291f002008-10-18 20:26:44 -0700214 */
215static inline void mlock_migrate_page(struct page *newpage, struct page *page)
216{
Nick Piggin5344b7e2008-10-18 20:26:51 -0700217 if (TestClearPageMlocked(page)) {
218 unsigned long flags;
Mel Gormanb32967f2012-11-19 12:35:47 +0000219 int nr_pages = hpage_nr_pages(page);
Nick Piggin5344b7e2008-10-18 20:26:51 -0700220
221 local_irq_save(flags);
Mel Gormanb32967f2012-11-19 12:35:47 +0000222 __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
Nick Pigginb291f002008-10-18 20:26:44 -0700223 SetPageMlocked(newpage);
Mel Gormanb32967f2012-11-19 12:35:47 +0000224 __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
Nick Piggin5344b7e2008-10-18 20:26:51 -0700225 local_irq_restore(flags);
226 }
Nick Pigginb291f002008-10-18 20:26:44 -0700227}
228
Mel Gormanb32967f2012-11-19 12:35:47 +0000229extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
230
Andrea Arcangeli71e3aac2011-01-13 15:46:52 -0800231#ifdef CONFIG_TRANSPARENT_HUGEPAGE
232extern unsigned long vma_address(struct page *page,
233 struct vm_area_struct *vma);
234#endif
Hugh Dickinsaf8e3352009-12-14 17:58:59 -0800235#else /* !CONFIG_MMU */
Ying Han096a7cf2012-05-29 15:06:25 -0700236static inline int mlocked_vma_newpage(struct vm_area_struct *v, struct page *p)
Nick Pigginb291f002008-10-18 20:26:44 -0700237{
238 return 0;
239}
240static inline void clear_page_mlock(struct page *page) { }
241static inline void mlock_vma_page(struct page *page) { }
242static inline void mlock_migrate_page(struct page *new, struct page *old) { }
243
Hugh Dickinsaf8e3352009-12-14 17:58:59 -0800244#endif /* !CONFIG_MMU */
Lee Schermerhorn894bc312008-10-18 20:26:39 -0700245
Alexander van Heukelumb5a0e012008-02-23 15:24:06 -0800246/*
Andy Whitcroft69d177c2008-11-06 12:53:26 -0800247 * Return the mem_map entry representing the 'offset' subpage within
248 * the maximally aligned gigantic page 'base'. Handle any discontiguity
249 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
250 */
251static inline struct page *mem_map_offset(struct page *base, int offset)
252{
253 if (unlikely(offset >= MAX_ORDER_NR_PAGES))
254 return pfn_to_page(page_to_pfn(base) + offset);
255 return base + offset;
256}
257
258/*
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300259 * Iterator over all subpages within the maximally aligned gigantic
Andy Whitcroft69d177c2008-11-06 12:53:26 -0800260 * page 'base'. Handle any discontiguity in the mem_map.
261 */
262static inline struct page *mem_map_next(struct page *iter,
263 struct page *base, int offset)
264{
265 if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
266 unsigned long pfn = page_to_pfn(base) + offset;
267 if (!pfn_valid(pfn))
268 return NULL;
269 return pfn_to_page(pfn);
270 }
271 return iter + 1;
272}
273
274/*
Alexander van Heukelumb5a0e012008-02-23 15:24:06 -0800275 * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
276 * so all functions starting at paging_init should be marked __init
277 * in those cases. SPARSEMEM, however, allows for memory hotplug,
278 * and alloc_bootmem_node is not used.
279 */
280#ifdef CONFIG_SPARSEMEM
281#define __paginginit __meminit
282#else
283#define __paginginit __init
284#endif
285
Mel Gorman6b74ab92008-07-23 21:26:49 -0700286/* Memory initialisation debug and verification */
287enum mminit_level {
288 MMINIT_WARNING,
289 MMINIT_VERIFY,
290 MMINIT_TRACE
291};
292
293#ifdef CONFIG_DEBUG_MEMORY_INIT
294
295extern int mminit_loglevel;
296
297#define mminit_dprintk(level, prefix, fmt, arg...) \
298do { \
299 if (level < mminit_loglevel) { \
300 printk(level <= MMINIT_WARNING ? KERN_WARNING : KERN_DEBUG); \
301 printk(KERN_CONT "mminit::" prefix " " fmt, ##arg); \
302 } \
303} while (0)
304
Mel Gorman708614e2008-07-23 21:26:51 -0700305extern void mminit_verify_pageflags_layout(void);
306extern void mminit_verify_page_links(struct page *page,
307 enum zone_type zone, unsigned long nid, unsigned long pfn);
Mel Gorman68ad8df2008-07-23 21:26:52 -0700308extern void mminit_verify_zonelist(void);
Mel Gorman708614e2008-07-23 21:26:51 -0700309
Mel Gorman6b74ab92008-07-23 21:26:49 -0700310#else
311
312static inline void mminit_dprintk(enum mminit_level level,
313 const char *prefix, const char *fmt, ...)
314{
315}
316
Mel Gorman708614e2008-07-23 21:26:51 -0700317static inline void mminit_verify_pageflags_layout(void)
318{
319}
320
321static inline void mminit_verify_page_links(struct page *page,
322 enum zone_type zone, unsigned long nid, unsigned long pfn)
323{
324}
Mel Gorman68ad8df2008-07-23 21:26:52 -0700325
326static inline void mminit_verify_zonelist(void)
327{
328}
Mel Gorman6b74ab92008-07-23 21:26:49 -0700329#endif /* CONFIG_DEBUG_MEMORY_INIT */
Mel Gorman2dbb51c2008-07-23 21:26:52 -0700330
331/* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
332#if defined(CONFIG_SPARSEMEM)
333extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
334 unsigned long *end_pfn);
335#else
336static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
337 unsigned long *end_pfn)
338{
339}
340#endif /* CONFIG_SPARSEMEM */
341
Mel Gormanfa5e0842009-06-16 15:33:22 -0700342#define ZONE_RECLAIM_NOSCAN -2
343#define ZONE_RECLAIM_FULL -1
344#define ZONE_RECLAIM_SOME 0
345#define ZONE_RECLAIM_SUCCESS 1
Wu Fengguang7c116f22009-12-16 12:19:59 +0100346
Wu Fengguang31d3d342009-12-16 12:19:59 +0100347extern int hwpoison_filter(struct page *p);
348
Wu Fengguang7c116f22009-12-16 12:19:59 +0100349extern u32 hwpoison_filter_dev_major;
350extern u32 hwpoison_filter_dev_minor;
Wu Fengguang478c5ff2009-12-16 12:19:59 +0100351extern u64 hwpoison_filter_flags_mask;
352extern u64 hwpoison_filter_flags_value;
Andi Kleen4fd466e2009-12-16 12:19:59 +0100353extern u64 hwpoison_filter_memcg;
Haicheng Li1bfe5fe2009-12-16 12:19:59 +0100354extern u32 hwpoison_filter_enable;
Al Viroeb36c582012-05-30 20:17:35 -0400355
356extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
357 unsigned long, unsigned long,
358 unsigned long, unsigned long);
Xishi Qiuca57df72012-07-31 16:43:19 -0700359
360extern void set_pageblock_order(void);
Minchan Kim02c6de82012-10-08 16:31:55 -0700361unsigned long reclaim_clean_pages_from_list(struct zone *zone,
362 struct list_head *page_list);
Bartlomiej Zolnierkiewiczd95ea5d2012-10-08 16:32:05 -0700363/* The ALLOC_WMARK bits are used as an index to zone->watermark */
364#define ALLOC_WMARK_MIN WMARK_MIN
365#define ALLOC_WMARK_LOW WMARK_LOW
366#define ALLOC_WMARK_HIGH WMARK_HIGH
367#define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
368
369/* Mask to get the watermark bits */
370#define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
371
372#define ALLOC_HARDER 0x10 /* try to alloc harder */
373#define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
374#define ALLOC_CPUSET 0x40 /* check for correct cpuset */
375#define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
376
Michel Lespinassedb971412012-10-08 16:29:34 -0700377#endif /* __MM_INTERNAL_H */