]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blame - mm/internal.h
memcg: fix css reference leak and endless loop in mem_cgroup_iter
[mirror_ubuntu-jammy-kernel.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
14#include <linux/mm.h>
1da177e4 15
42b77728
JB
16void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
17 unsigned long floor, unsigned long ceiling);
18
7835e98b 19static inline void set_page_count(struct page *page, int v)
77a8a788 20{
7835e98b
NP
21 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{
309381fe
SL
30 VM_BUG_ON_PAGE(PageTail(page), page);
31 VM_BUG_ON_PAGE(atomic_read(&page->_count), page);
77a8a788 32 set_page_count(page, 1);
77a8a788
NP
33}
34
70b50f94
AA
35static inline void __get_page_tail_foll(struct page *page,
36 bool get_page_head)
37{
38 /*
39 * If we're getting a tail page, the elevated page->_count is
40 * required only in the head page and we will elevate the head
41 * page->_count and tail page->_mapcount.
42 *
43 * We elevate page_tail->_mapcount for tail pages to force
44 * page_tail->_count to be zero at all times to avoid getting
45 * false positives from get_page_unless_zero() with
46 * speculative page access (like in
47 * page_cache_get_speculative()) on tail pages.
48 */
309381fe 49 VM_BUG_ON_PAGE(atomic_read(&page->first_page->_count) <= 0, page);
70b50f94
AA
50 if (get_page_head)
51 atomic_inc(&page->first_page->_count);
c728852f 52 get_huge_page_tail(page);
70b50f94
AA
53}
54
55/*
56 * This is meant to be called as the FOLL_GET operation of
57 * follow_page() and it must be called while holding the proper PT
58 * lock while the pte (or pmd_trans_huge) is still mapping the page.
59 */
60static inline void get_page_foll(struct page *page)
61{
62 if (unlikely(PageTail(page)))
63 /*
64 * This is safe only because
65 * __split_huge_page_refcount() can't run under
66 * get_page_foll() because we hold the proper PT lock.
67 */
68 __get_page_tail_foll(page, true);
69 else {
70 /*
71 * Getting a normal page or the head of a compound page
72 * requires to already have an elevated page->_count.
73 */
309381fe 74 VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page);
70b50f94
AA
75 atomic_inc(&page->_count);
76 }
77}
78
03f6462a
HD
79extern unsigned long highest_memmap_pfn;
80
894bc310
LS
81/*
82 * in mm/vmscan.c:
83 */
62695a84 84extern int isolate_lru_page(struct page *page);
894bc310 85extern void putback_lru_page(struct page *page);
6e543d57
LD
86extern unsigned long zone_reclaimable_pages(struct zone *zone);
87extern bool zone_reclaimable(struct zone *zone);
62695a84 88
6219049a
BL
89/*
90 * in mm/rmap.c:
91 */
92extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
93
894bc310
LS
94/*
95 * in mm/page_alloc.c
96 */
0c0a4a51 97extern void __free_pages_bootmem(struct page *page, unsigned int order);
20a0307c 98extern void prep_compound_page(struct page *page, unsigned long order);
8d22ba1b
WF
99#ifdef CONFIG_MEMORY_FAILURE
100extern bool is_free_buddy_page(struct page *page);
101#endif
42aa83cb 102extern int user_min_free_kbytes;
20a0307c 103
ff9543fd
MN
104#if defined CONFIG_COMPACTION || defined CONFIG_CMA
105
106/*
107 * in mm/compaction.c
108 */
109/*
110 * compact_control is used to track pages being migrated and the free pages
111 * they are being migrated to during memory compaction. The free_pfn starts
112 * at the end of a zone and migrate_pfn begins at the start. Movable pages
113 * are moved to the end of a zone during a compaction run and the run
114 * completes when free_pfn <= migrate_pfn
115 */
116struct compact_control {
117 struct list_head freepages; /* List of free pages to migrate to */
118 struct list_head migratepages; /* List of pages being migrated */
119 unsigned long nr_freepages; /* Number of isolated free pages */
120 unsigned long nr_migratepages; /* Number of pages to migrate */
121 unsigned long free_pfn; /* isolate_freepages search base */
122 unsigned long migrate_pfn; /* isolate_migratepages search base */
68e3e926 123 bool sync; /* Synchronous migration */
bb13ffeb 124 bool ignore_skip_hint; /* Scan blocks even if marked skip */
c89511ab
MG
125 bool finished_update_free; /* True when the zone cached pfns are
126 * no longer being updated
127 */
128 bool finished_update_migrate;
ff9543fd
MN
129
130 int order; /* order a direct compactor needs */
131 int migratetype; /* MOVABLE, RECLAIMABLE etc */
132 struct zone *zone;
e64c5237 133 bool contended; /* True if a lock was contended */
ff9543fd
MN
134};
135
136unsigned long
bb13ffeb
MG
137isolate_freepages_range(struct compact_control *cc,
138 unsigned long start_pfn, unsigned long end_pfn);
ff9543fd
MN
139unsigned long
140isolate_migratepages_range(struct zone *zone, struct compact_control *cc,
e46a2879 141 unsigned long low_pfn, unsigned long end_pfn, bool unevictable);
ff9543fd
MN
142
143#endif
0f8053a5 144
48f13bf3
MG
145/*
146 * function for dealing with page's order in buddy system.
147 * zone->lock is already acquired when we use these.
148 * So, we don't need atomic page->flags operations here.
149 */
150static inline unsigned long page_order(struct page *page)
151{
572438f9 152 /* PageBuddy() must be checked by the caller */
48f13bf3
MG
153 return page_private(page);
154}
b5a0e011 155
6038def0
NK
156/* mm/util.c */
157void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
158 struct vm_area_struct *prev, struct rb_node *rb_parent);
159
af8e3354 160#ifdef CONFIG_MMU
cea10a19
ML
161extern long __mlock_vma_pages_range(struct vm_area_struct *vma,
162 unsigned long start, unsigned long end, int *nonblocking);
af8e3354
HD
163extern void munlock_vma_pages_range(struct vm_area_struct *vma,
164 unsigned long start, unsigned long end);
165static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
166{
167 munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
168}
169
b291f000 170/*
39b5f29a
HD
171 * Called only in fault path, to determine if a new page is being
172 * mapped into a LOCKED vma. If it is, mark page as mlocked.
b291f000 173 */
096a7cf4
YH
174static inline int mlocked_vma_newpage(struct vm_area_struct *vma,
175 struct page *page)
b291f000 176{
309381fe 177 VM_BUG_ON_PAGE(PageLRU(page), page);
b291f000
NP
178
179 if (likely((vma->vm_flags & (VM_LOCKED | VM_SPECIAL)) != VM_LOCKED))
180 return 0;
181
5344b7e6 182 if (!TestSetPageMlocked(page)) {
8449d21f
DR
183 mod_zone_page_state(page_zone(page), NR_MLOCK,
184 hpage_nr_pages(page));
5344b7e6
NP
185 count_vm_event(UNEVICTABLE_PGMLOCKED);
186 }
b291f000
NP
187 return 1;
188}
189
190/*
73848b46 191 * must be called with vma's mmap_sem held for read or write, and page locked.
b291f000
NP
192 */
193extern void mlock_vma_page(struct page *page);
ff6a6da6 194extern unsigned int munlock_vma_page(struct page *page);
b291f000
NP
195
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 */
e6c509f8 205extern void clear_page_mlock(struct page *page);
b291f000
NP
206
207/*
208 * mlock_migrate_page - called only from migrate_page_copy() to
5344b7e6 209 * migrate the Mlocked page flag; update statistics.
b291f000
NP
210 */
211static inline void mlock_migrate_page(struct page *newpage, struct page *page)
212{
5344b7e6
NP
213 if (TestClearPageMlocked(page)) {
214 unsigned long flags;
b32967ff 215 int nr_pages = hpage_nr_pages(page);
5344b7e6
NP
216
217 local_irq_save(flags);
b32967ff 218 __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
b291f000 219 SetPageMlocked(newpage);
b32967ff 220 __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
5344b7e6
NP
221 local_irq_restore(flags);
222 }
b291f000
NP
223}
224
b32967ff
MG
225extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
226
71e3aac0
AA
227#ifdef CONFIG_TRANSPARENT_HUGEPAGE
228extern unsigned long vma_address(struct page *page,
229 struct vm_area_struct *vma);
230#endif
af8e3354 231#else /* !CONFIG_MMU */
096a7cf4 232static inline int mlocked_vma_newpage(struct vm_area_struct *v, struct page *p)
b291f000
NP
233{
234 return 0;
235}
236static inline void clear_page_mlock(struct page *page) { }
237static inline void mlock_vma_page(struct page *page) { }
238static inline void mlock_migrate_page(struct page *new, struct page *old) { }
239
af8e3354 240#endif /* !CONFIG_MMU */
894bc310 241
69d177c2
AW
242/*
243 * Return the mem_map entry representing the 'offset' subpage within
244 * the maximally aligned gigantic page 'base'. Handle any discontiguity
245 * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
246 */
247static inline struct page *mem_map_offset(struct page *base, int offset)
248{
249 if (unlikely(offset >= MAX_ORDER_NR_PAGES))
250 return pfn_to_page(page_to_pfn(base) + offset);
251 return base + offset;
252}
253
254/*
25985edc 255 * Iterator over all subpages within the maximally aligned gigantic
69d177c2
AW
256 * page 'base'. Handle any discontiguity in the mem_map.
257 */
258static inline struct page *mem_map_next(struct page *iter,
259 struct page *base, int offset)
260{
261 if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
262 unsigned long pfn = page_to_pfn(base) + offset;
263 if (!pfn_valid(pfn))
264 return NULL;
265 return pfn_to_page(pfn);
266 }
267 return iter + 1;
268}
269
b5a0e011
AH
270/*
271 * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
272 * so all functions starting at paging_init should be marked __init
273 * in those cases. SPARSEMEM, however, allows for memory hotplug,
274 * and alloc_bootmem_node is not used.
275 */
276#ifdef CONFIG_SPARSEMEM
277#define __paginginit __meminit
278#else
279#define __paginginit __init
280#endif
281
6b74ab97
MG
282/* Memory initialisation debug and verification */
283enum mminit_level {
284 MMINIT_WARNING,
285 MMINIT_VERIFY,
286 MMINIT_TRACE
287};
288
289#ifdef CONFIG_DEBUG_MEMORY_INIT
290
291extern int mminit_loglevel;
292
293#define mminit_dprintk(level, prefix, fmt, arg...) \
294do { \
295 if (level < mminit_loglevel) { \
296 printk(level <= MMINIT_WARNING ? KERN_WARNING : KERN_DEBUG); \
297 printk(KERN_CONT "mminit::" prefix " " fmt, ##arg); \
298 } \
299} while (0)
300
708614e6
MG
301extern void mminit_verify_pageflags_layout(void);
302extern void mminit_verify_page_links(struct page *page,
303 enum zone_type zone, unsigned long nid, unsigned long pfn);
68ad8df4 304extern void mminit_verify_zonelist(void);
708614e6 305
6b74ab97
MG
306#else
307
308static inline void mminit_dprintk(enum mminit_level level,
309 const char *prefix, const char *fmt, ...)
310{
311}
312
708614e6
MG
313static inline void mminit_verify_pageflags_layout(void)
314{
315}
316
317static inline void mminit_verify_page_links(struct page *page,
318 enum zone_type zone, unsigned long nid, unsigned long pfn)
319{
320}
68ad8df4
MG
321
322static inline void mminit_verify_zonelist(void)
323{
324}
6b74ab97 325#endif /* CONFIG_DEBUG_MEMORY_INIT */
2dbb51c4
MG
326
327/* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
328#if defined(CONFIG_SPARSEMEM)
329extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
330 unsigned long *end_pfn);
331#else
332static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
333 unsigned long *end_pfn)
334{
335}
336#endif /* CONFIG_SPARSEMEM */
337
fa5e084e
MG
338#define ZONE_RECLAIM_NOSCAN -2
339#define ZONE_RECLAIM_FULL -1
340#define ZONE_RECLAIM_SOME 0
341#define ZONE_RECLAIM_SUCCESS 1
7c116f2b 342
31d3d348
WF
343extern int hwpoison_filter(struct page *p);
344
7c116f2b
WF
345extern u32 hwpoison_filter_dev_major;
346extern u32 hwpoison_filter_dev_minor;
478c5ffc
WF
347extern u64 hwpoison_filter_flags_mask;
348extern u64 hwpoison_filter_flags_value;
4fd466eb 349extern u64 hwpoison_filter_memcg;
1bfe5feb 350extern u32 hwpoison_filter_enable;
eb36c587
AV
351
352extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
353 unsigned long, unsigned long,
354 unsigned long, unsigned long);
ca57df79
XQ
355
356extern void set_pageblock_order(void);
02c6de8d
MK
357unsigned long reclaim_clean_pages_from_list(struct zone *zone,
358 struct list_head *page_list);
d95ea5d1
BZ
359/* The ALLOC_WMARK bits are used as an index to zone->watermark */
360#define ALLOC_WMARK_MIN WMARK_MIN
361#define ALLOC_WMARK_LOW WMARK_LOW
362#define ALLOC_WMARK_HIGH WMARK_HIGH
363#define ALLOC_NO_WATERMARKS 0x04 /* don't check watermarks at all */
364
365/* Mask to get the watermark bits */
366#define ALLOC_WMARK_MASK (ALLOC_NO_WATERMARKS-1)
367
368#define ALLOC_HARDER 0x10 /* try to alloc harder */
369#define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
370#define ALLOC_CPUSET 0x40 /* check for correct cpuset */
371#define ALLOC_CMA 0x80 /* allow allocations from CMA areas */
372
db971418 373#endif /* __MM_INTERNAL_H */