mm: page_alloc: do not treat a zone that cannot be used for dirty pages as "full"
[deliverable/linux.git] / mm / internal.h
CommitLineData
1da177e4
LT
1/* 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 */
0f8053a5
NP
11#ifndef __MM_INTERNAL_H
12#define __MM_INTERNAL_H
13
29f175d1 14#include <linux/fs.h>
0f8053a5 15#include <linux/mm.h>
1da177e4 16
42b77728
JB
17void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
18 unsigned long floor, unsigned long ceiling);
19
7835e98b 20static inline void set_page_count(struct page *page, int v)
77a8a788 21{
7835e98b
NP
22 atomic_set(&page->_count, v);
23}
24
29f175d1
FF
25extern int __do_page_cache_readahead(struct address_space *mapping,
26 struct file *filp, pgoff_t offset, unsigned long nr_to_read,
27 unsigned long lookahead_size);
28
29/*
30 * Submit IO for the read-ahead request in file_ra_state.
31 */
32static inline unsigned long ra_submit(struct file_ra_state *ra,
33 struct address_space *mapping, struct file *filp)
34{
35 return __do_page_cache_readahead(mapping, filp,
36 ra->start, ra->size, ra->async_size);
37}
38
7835e98b
NP
39/*
40 * Turn a non-refcounted page (->_count == 0) into refcounted with
41 * a count of one.
42 */
43static inline void set_page_refcounted(struct page *page)
44{
309381fe
SL
45 VM_BUG_ON_PAGE(PageTail(page), page);
46 VM_BUG_ON_PAGE(atomic_read(&page->_count), page);
77a8a788 47 set_page_count(page, 1);
77a8a788
NP
48}
49
70b50f94
AA
50static inline void __get_page_tail_foll(struct page *page,
51 bool get_page_head)
52{
53 /*
54 * If we're getting a tail page, the elevated page->_count is
55 * required only in the head page and we will elevate the head
56 * page->_count and tail page->_mapcount.
57 *
58 * We elevate page_tail->_mapcount for tail pages to force
59 * page_tail->_count to be zero at all times to avoid getting
60 * false positives from get_page_unless_zero() with
61 * speculative page access (like in
62 * page_cache_get_speculative()) on tail pages.
63 */
309381fe 64 VM_BUG_ON_PAGE(atomic_read(&page->first_page->_count) <= 0, page);
70b50f94
AA
65 if (get_page_head)
66 atomic_inc(&page->first_page->_count);
c728852f 67 get_huge_page_tail(page);
70b50f94
AA
68}
69
70/*
71 * This is meant to be called as the FOLL_GET operation of
72 * follow_page() and it must be called while holding the proper PT
73 * lock while the pte (or pmd_trans_huge) is still mapping the page.
74 */
75static inline void get_page_foll(struct page *page)
76{
77 if (unlikely(PageTail(page)))
78 /*
79 * This is safe only because
80 * __split_huge_page_refcount() can't run under
81 * get_page_foll() because we hold the proper PT lock.
82 */
83 __get_page_tail_foll(page, true);
84 else {
85 /*
86 * Getting a normal page or the head of a compound page
87 * requires to already have an elevated page->_count.
88 */
309381fe 89 VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page);
70b50f94
AA
90 atomic_inc(&page->_count);
91 }
92}
93
03f6462a
HD
94extern unsigned long highest_memmap_pfn;
95
894bc310
LS
96/*
97 * in mm/vmscan.c:
98 */
62695a84 99extern int isolate_lru_page(struct page *page);
894bc310 100extern void putback_lru_page(struct page *page);
6e543d57 101extern bool zone_reclaimable(struct zone *zone);
62695a84 102
6219049a
BL
103/*
104 * in mm/rmap.c:
105 */
106extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
107
894bc310
LS
108/*
109 * in mm/page_alloc.c
110 */
0c0a4a51 111extern void __free_pages_bootmem(struct page *page, unsigned int order);
20a0307c 112extern void prep_compound_page(struct page *page, unsigned long order);
8d22ba1b
WF
113#ifdef CONFIG_MEMORY_FAILURE
114extern bool is_free_buddy_page(struct page *page);
115#endif
42aa83cb 116extern int user_min_free_kbytes;
20a0307c 117
ff9543fd
MN
118#if defined CONFIG_COMPACTION || defined CONFIG_CMA
119
120/*
121 * in mm/compaction.c
122 */
123/*
124 * compact_control is used to track pages being migrated and the free pages
125 * they are being migrated to during memory compaction. The free_pfn starts
126 * at the end of a zone and migrate_pfn begins at the start. Movable pages
127 * are moved to the end of a zone during a compaction run and the run
128 * completes when free_pfn <= migrate_pfn
129 */
130struct compact_control {
131 struct list_head freepages; /* List of free pages to migrate to */
132 struct list_head migratepages; /* List of pages being migrated */
133 unsigned long nr_freepages; /* Number of isolated free pages */
134 unsigned long nr_migratepages; /* Number of pages to migrate */
135 unsigned long free_pfn; /* isolate_freepages search base */
136 unsigned long migrate_pfn; /* isolate_migratepages search base */
e0b9daeb 137 enum migrate_mode mode; /* Async or sync migration mode */
bb13ffeb 138 bool ignore_skip_hint; /* Scan blocks even if marked skip */
c89511ab
MG
139 bool finished_update_free; /* True when the zone cached pfns are
140 * no longer being updated
141 */
142 bool finished_update_migrate;
ff9543fd
MN
143
144 int order; /* order a direct compactor needs */
145 int migratetype; /* MOVABLE, RECLAIMABLE etc */
146 struct zone *zone;
e64c5237 147 bool contended; /* True if a lock was contended */
ff9543fd
MN
148};
149
150unsigned long
bb13ffeb
MG
151isolate_freepages_range(struct compact_control *cc,
152 unsigned long start_pfn, unsigned long end_pfn);
ff9543fd
MN
153unsigned long
154isolate_migratepages_range(struct zone *zone, struct compact_control *cc,
e46a2879 155 unsigned long low_pfn, unsigned long end_pfn, bool unevictable);
ff9543fd
MN
156
157#endif
0f8053a5 158
48f13bf3 159/*
6c14466c
MG
160 * This function returns the order of a free page in the buddy system. In
161 * general, page_zone(page)->lock must be held by the caller to prevent the
162 * page from being allocated in parallel and returning garbage as the order.
163 * If a caller does not hold page_zone(page)->lock, it must guarantee that the
164 * page cannot be allocated or merged in parallel.
48f13bf3
MG
165 */
166static inline unsigned long page_order(struct page *page)
167{
572438f9 168 /* PageBuddy() must be checked by the caller */
48f13bf3
MG
169 return page_private(page);
170}
b5a0e011 171
4bbd4c77
KS
172static inline bool is_cow_mapping(vm_flags_t flags)
173{
174 return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
175}
176
6038def0
NK
177/* mm/util.c */
178void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
179 struct vm_area_struct *prev, struct rb_node *rb_parent);
180
af8e3354 181#ifdef CONFIG_MMU
cea10a19
ML
182extern long __mlock_vma_pages_range(struct vm_area_struct *vma,
183 unsigned long start, unsigned long end, int *nonblocking);
af8e3354
HD
184extern void munlock_vma_pages_range(struct vm_area_struct *vma,
185 unsigned long start, unsigned long end);
186static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
187{
188 munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
189}
190
b291f000 191/*
73848b46 192 * must be called with vma's mmap_sem held for read or write, and page locked.
b291f000
NP
193 */
194extern void mlock_vma_page(struct page *page);
ff6a6da6 195extern unsigned int munlock_vma_page(struct page *page);
b291f000
NP
196
197/*
198 * Clear the page's PageMlocked(). This can be useful in a situation where
199 * we want to unconditionally remove a page from the pagecache -- e.g.,
200 * on truncation or freeing.
201 *
202 * It is legal to call this function for any page, mlocked or not.
203 * If called for a page that is still mapped by mlocked vmas, all we do
204 * is revert to lazy LRU behaviour -- semantics are not broken.
205 */
e6c509f8 206extern void clear_page_mlock(struct page *page);
b291f000
NP
207
208/*
209 * mlock_migrate_page - called only from migrate_page_copy() to
5344b7e6 210 * migrate the Mlocked page flag; update statistics.
b291f000
NP
211 */
212static inline void mlock_migrate_page(struct page *newpage, struct page *page)
213{
5344b7e6
NP
214 if (TestClearPageMlocked(page)) {
215 unsigned long flags;
b32967ff 216 int nr_pages = hpage_nr_pages(page);
5344b7e6
NP
217
218 local_irq_save(flags);
b32967ff 219 __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
b291f000 220 SetPageMlocked(newpage);
b32967ff 221 __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
5344b7e6
NP
222 local_irq_restore(flags);
223 }
b291f000
NP
224}
225
b32967ff
MG
226extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
227
71e3aac0
AA
228#ifdef CONFIG_TRANSPARENT_HUGEPAGE
229extern unsigned long vma_address(struct page *page,
230 struct vm_area_struct *vma);
231#endif
af8e3354 232#else /* !CONFIG_MMU */
b291f000
NP
233static inline void clear_page_mlock(struct page *page) { }
234static inline void mlock_vma_page(struct page *page) { }
235static inline void mlock_migrate_page(struct page *new, struct page *old) { }
236
af8e3354 237#endif /* !CONFIG_MMU */
894bc310 238
69d177c2
AW
239/*
240 * Return the mem_map entry representing the 'offset' subpage within
241 * the maximally aligned gigantic page 'base'. Handle any discontiguity
242 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
243 */
244static inline struct page *mem_map_offset(struct page *base, int offset)
245{
246 if (unlikely(offset >= MAX_ORDER_NR_PAGES))
247 return pfn_to_page(page_to_pfn(base) + offset);
248 return base + offset;
249}
250
251/*
25985edc 252 * Iterator over all subpages within the maximally aligned gigantic
69d177c2
AW
253 * page 'base'. Handle any discontiguity in the mem_map.
254 */
255static inline struct page *mem_map_next(struct page *iter,
256 struct page *base, int offset)
257{
258 if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
259 unsigned long pfn = page_to_pfn(base) + offset;
260 if (!pfn_valid(pfn))
261 return NULL;
262 return pfn_to_page(pfn);
263 }
264 return iter + 1;
265}
266
b5a0e011
AH
267/*
268 * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
269 * so all functions starting at paging_init should be marked __init
270 * in those cases. SPARSEMEM, however, allows for memory hotplug,
271 * and alloc_bootmem_node is not used.
272 */
273#ifdef CONFIG_SPARSEMEM
274#define __paginginit __meminit
275#else
276#define __paginginit __init
277#endif
278
6b74ab97
MG
279/* Memory initialisation debug and verification */
280enum mminit_level {
281 MMINIT_WARNING,
282 MMINIT_VERIFY,
283 MMINIT_TRACE
284};
285
286#ifdef CONFIG_DEBUG_MEMORY_INIT
287
288extern int mminit_loglevel;
289
290#define mminit_dprintk(level, prefix, fmt, arg...) \
291do { \
292 if (level < mminit_loglevel) { \
293 printk(level <= MMINIT_WARNING ? KERN_WARNING : KERN_DEBUG); \
294 printk(KERN_CONT "mminit::" prefix " " fmt, ##arg); \
295 } \
296} while (0)
297
708614e6
MG
298extern void mminit_verify_pageflags_layout(void);
299extern void mminit_verify_page_links(struct page *page,
300 enum zone_type zone, unsigned long nid, unsigned long pfn);
68ad8df4 301extern void mminit_verify_zonelist(void);
708614e6 302
6b74ab97
MG
303#else
304
305static inline void mminit_dprintk(enum mminit_level level,
306 const char *prefix, const char *fmt, ...)
307{
308}
309
708614e6
MG
310static inline void mminit_verify_pageflags_layout(void)
311{
312}
313
314static inline void mminit_verify_page_links(struct page *page,
315 enum zone_type zone, unsigned long nid, unsigned long pfn)
316{
317}
68ad8df4
MG
318
319static inline void mminit_verify_zonelist(void)
320{
321}
6b74ab97 322#endif /* CONFIG_DEBUG_MEMORY_INIT */
2dbb51c4
MG
323
324/* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
325#if defined(CONFIG_SPARSEMEM)
326extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
327 unsigned long *end_pfn);
328#else
329static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
330 unsigned long *end_pfn)
331{
332}
333#endif /* CONFIG_SPARSEMEM */
334
fa5e084e
MG
335#define ZONE_RECLAIM_NOSCAN -2
336#define ZONE_RECLAIM_FULL -1
337#define ZONE_RECLAIM_SOME 0
338#define ZONE_RECLAIM_SUCCESS 1
7c116f2b 339
31d3d348
WF
340extern int hwpoison_filter(struct page *p);
341
7c116f2b
WF
342extern u32 hwpoison_filter_dev_major;
343extern u32 hwpoison_filter_dev_minor;
478c5ffc
WF
344extern u64 hwpoison_filter_flags_mask;
345extern u64 hwpoison_filter_flags_value;
4fd466eb 346extern u64 hwpoison_filter_memcg;
1bfe5feb 347extern u32 hwpoison_filter_enable;
eb36c587
AV
348
349extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
350 unsigned long, unsigned long,
351 unsigned long, unsigned long);
ca57df79
XQ
352
353extern void set_pageblock_order(void);
02c6de8d
MK
354unsigned long reclaim_clean_pages_from_list(struct zone *zone,
355 struct list_head *page_list);
d95ea5d1
BZ
356/* The ALLOC_WMARK bits are used as an index to zone->watermark */
357#define ALLOC_WMARK_MIN WMARK_MIN
358#define ALLOC_WMARK_LOW WMARK_LOW
359#define ALLOC_WMARK_HIGH WMARK_HIGH
360#define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
361
362/* Mask to get the watermark bits */
363#define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
364
365#define ALLOC_HARDER 0x10 /* try to alloc harder */
366#define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
367#define ALLOC_CPUSET 0x40 /* check for correct cpuset */
368#define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
3a025760 369#define ALLOC_FAIR 0x100 /* fair zone allocation */
d95ea5d1 370
db971418 371#endif /* __MM_INTERNAL_H */
This page took 0.895328 seconds and 5 git commands to generate.