mm/internal.h

   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  */
  11 #ifndef __MM_INTERNAL_H
  12 #define __MM_INTERNAL_H
  13
  14 #include <linux/mm.h>
  15
  16 void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
  17                 unsigned long floor, unsigned long ceiling);
  18
  19 static inline void set_page_count(struct page *page, int v)
  20 {
  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  */
  28 static inline void set_page_refcounted(struct page *page)
  29 {
  30         VM_BUG_ON(PageTail(page));
  31         VM_BUG_ON(atomic_read(&page->_count));
  32         set_page_count(page, 1);
  33 }
  34
  35 static inline void __put_page(struct page *page)
  36 {
  37         atomic_dec(&page->_count);
  38 }
  39
  40 static 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  */
  67 static 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
  86 extern unsigned long highest_memmap_pfn;
  87
  88 /*
  89  * in mm/vmscan.c:
  90  */
  91 extern int isolate_lru_page(struct page *page);
  92 extern void putback_lru_page(struct page *page);
  93
  94 /*
  95  * in mm/page_alloc.c
  96  */
  97 extern void __free_pages_bootmem(struct page *page, unsigned int order);
  98 extern void prep_compound_page(struct page *page, unsigned long order);
  99 #ifdef CONFIG_MEMORY_FAILURE
 100 extern bool is_free_buddy_page(struct page *page);
 101 #endif
 102
 103
 104 /*
 105  * function for dealing with page's order in buddy system.
 106  * zone->lock is already acquired when we use these.
 107  * So, we don't need atomic page->flags operations here.
 108  */
 109 static inline unsigned long page_order(struct page *page)
 110 {
 111         /* PageBuddy() must be checked by the caller */
 112         return page_private(page);
 113 }
 114
 115 /* mm/util.c */
 116 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
 117                 struct vm_area_struct *prev, struct rb_node *rb_parent);
 118
 119 #ifdef CONFIG_MMU
 120 extern long mlock_vma_pages_range(struct vm_area_struct *vma,
 121                         unsigned long start, unsigned long end);
 122 extern void munlock_vma_pages_range(struct vm_area_struct *vma,
 123                         unsigned long start, unsigned long end);
 124 static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
 125 {
 126         munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
 127 }
 128
 129 /*
 130  * Called only in fault path via page_evictable() for a new page
 131  * to determine if it's being mapped into a LOCKED vma.
 132  * If so, mark page as mlocked.
 133  */
 134 static inline int is_mlocked_vma(struct vm_area_struct *vma, struct page *page)
 135 {
 136         VM_BUG_ON(PageLRU(page));
 137
 138         if (likely((vma->vm_flags & (VM_LOCKED | VM_SPECIAL)) != VM_LOCKED))
 139                 return 0;
 140
 141         if (!TestSetPageMlocked(page)) {
 142                 inc_zone_page_state(page, NR_MLOCK);
 143                 count_vm_event(UNEVICTABLE_PGMLOCKED);
 144         }
 145         return 1;
 146 }
 147
 148 /*
 149  * must be called with vma's mmap_sem held for read or write, and page locked.
 150  */
 151 extern void mlock_vma_page(struct page *page);
 152 extern void munlock_vma_page(struct page *page);
 153
 154 /*
 155  * Clear the page's PageMlocked().  This can be useful in a situation where
 156  * we want to unconditionally remove a page from the pagecache -- e.g.,
 157  * on truncation or freeing.
 158  *
 159  * It is legal to call this function for any page, mlocked or not.
 160  * If called for a page that is still mapped by mlocked vmas, all we do
 161  * is revert to lazy LRU behaviour -- semantics are not broken.
 162  */
 163 extern void __clear_page_mlock(struct page *page);
 164 static inline void clear_page_mlock(struct page *page)
 165 {
 166         if (unlikely(TestClearPageMlocked(page)))
 167                 __clear_page_mlock(page);
 168 }
 169
 170 /*
 171  * mlock_migrate_page - called only from migrate_page_copy() to
 172  * migrate the Mlocked page flag; update statistics.
 173  */
 174 static inline void mlock_migrate_page(struct page *newpage, struct page *page)
 175 {
 176         if (TestClearPageMlocked(page)) {
 177                 unsigned long flags;
 178
 179                 local_irq_save(flags);
 180                 __dec_zone_page_state(page, NR_MLOCK);
 181                 SetPageMlocked(newpage);
 182                 __inc_zone_page_state(newpage, NR_MLOCK);
 183                 local_irq_restore(flags);
 184         }
 185 }
 186
 187 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 188 extern unsigned long vma_address(struct page *page,
 189                                  struct vm_area_struct *vma);
 190 #endif
 191 #else /* !CONFIG_MMU */
 192 static inline int is_mlocked_vma(struct vm_area_struct *v, struct page *p)
 193 {
 194         return 0;
 195 }
 196 static inline void clear_page_mlock(struct page *page) { }
 197 static inline void mlock_vma_page(struct page *page) { }
 198 static inline void mlock_migrate_page(struct page *new, struct page *old) { }
 199
 200 #endif /* !CONFIG_MMU */
 201
 202 /*
 203  * Return the mem_map entry representing the 'offset' subpage within
 204  * the maximally aligned gigantic page 'base'.  Handle any discontiguity
 205  * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
 206  */
 207 static inline struct page *mem_map_offset(struct page *base, int offset)
 208 {
 209         if (unlikely(offset >= MAX_ORDER_NR_PAGES))
 210                 return pfn_to_page(page_to_pfn(base) + offset);
 211         return base + offset;
 212 }
 213
 214 /*
 215  * Iterator over all subpages within the maximally aligned gigantic
 216  * page 'base'.  Handle any discontiguity in the mem_map.
 217  */
 218 static inline struct page *mem_map_next(struct page *iter,
 219                                                 struct page *base, int offset)
 220 {
 221         if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
 222                 unsigned long pfn = page_to_pfn(base) + offset;
 223                 if (!pfn_valid(pfn))
 224                         return NULL;
 225                 return pfn_to_page(pfn);
 226         }
 227         return iter + 1;
 228 }
 229
 230 /*
 231  * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
 232  * so all functions starting at paging_init should be marked __init
 233  * in those cases. SPARSEMEM, however, allows for memory hotplug,
 234  * and alloc_bootmem_node is not used.
 235  */
 236 #ifdef CONFIG_SPARSEMEM
 237 #define __paginginit __meminit
 238 #else
 239 #define __paginginit __init
 240 #endif
 241
 242 /* Memory initialisation debug and verification */
 243 enum mminit_level {
 244         MMINIT_WARNING,
 245         MMINIT_VERIFY,
 246         MMINIT_TRACE
 247 };
 248
 249 #ifdef CONFIG_DEBUG_MEMORY_INIT
 250
 251 extern int mminit_loglevel;
 252
 253 #define mminit_dprintk(level, prefix, fmt, arg...) \
 254 do { \
 255         if (level < mminit_loglevel) { \
 256                 printk(level <= MMINIT_WARNING ? KERN_WARNING : KERN_DEBUG); \
 257                 printk(KERN_CONT "mminit::" prefix " " fmt, ##arg); \
 258         } \
 259 } while (0)
 260
 261 extern void mminit_verify_pageflags_layout(void);
 262 extern void mminit_verify_page_links(struct page *page,
 263                 enum zone_type zone, unsigned long nid, unsigned long pfn);
 264 extern void mminit_verify_zonelist(void);
 265
 266 #else
 267
 268 static inline void mminit_dprintk(enum mminit_level level,
 269                                 const char *prefix, const char *fmt, ...)
 270 {
 271 }
 272
 273 static inline void mminit_verify_pageflags_layout(void)
 274 {
 275 }
 276
 277 static inline void mminit_verify_page_links(struct page *page,
 278                 enum zone_type zone, unsigned long nid, unsigned long pfn)
 279 {
 280 }
 281
 282 static inline void mminit_verify_zonelist(void)
 283 {
 284 }
 285 #endif /* CONFIG_DEBUG_MEMORY_INIT */
 286
 287 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
 288 #if defined(CONFIG_SPARSEMEM)
 289 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 290                                 unsigned long *end_pfn);
 291 #else
 292 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 293                                 unsigned long *end_pfn)
 294 {
 295 }
 296 #endif /* CONFIG_SPARSEMEM */
 297
 298 #define ZONE_RECLAIM_NOSCAN     -2
 299 #define ZONE_RECLAIM_FULL       -1
 300 #define ZONE_RECLAIM_SOME       0
 301 #define ZONE_RECLAIM_SUCCESS    1
 302 #endif
 303
 304 extern int hwpoison_filter(struct page *p);
 305
 306 extern u32 hwpoison_filter_dev_major;
 307 extern u32 hwpoison_filter_dev_minor;
 308 extern u64 hwpoison_filter_flags_mask;
 309 extern u64 hwpoison_filter_flags_value;
 310 extern u64 hwpoison_filter_memcg;
 311 extern u32 hwpoison_filter_enable;