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 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
 104
 105 /*
 106  * in mm/compaction.c
 107  */
 108 /*
 109  * compact_control is used to track pages being migrated and the free pages
 110  * they are being migrated to during memory compaction. The free_pfn starts
 111  * at the end of a zone and migrate_pfn begins at the start. Movable pages
 112  * are moved to the end of a zone during a compaction run and the run
 113  * completes when free_pfn <= migrate_pfn
 114  */
 115 struct compact_control {
 116         struct list_head freepages;     /* List of free pages to migrate to */
 117         struct list_head migratepages;  /* List of pages being migrated */
 118         unsigned long nr_freepages;     /* Number of isolated free pages */
 119         unsigned long nr_migratepages;  /* Number of pages to migrate */
 120         unsigned long free_pfn;         /* isolate_freepages search base */
 121         unsigned long start_free_pfn;   /* where we started the search */
 122         unsigned long migrate_pfn;      /* isolate_migratepages search base */
 123         bool sync;                      /* Synchronous migration */
 124         bool wrapped;                   /* Order > 0 compactions are
 125                                            incremental, once free_pfn
 126                                            and migrate_pfn meet, we restart
 127                                            from the top of the zone;
 128                                            remember we wrapped around. */
 129
 130         int order;                      /* order a direct compactor needs */
 131         int migratetype;                /* MOVABLE, RECLAIMABLE etc */
 132         struct zone *zone;
 133 };
 134
 135 unsigned long
 136 isolate_freepages_range(unsigned long start_pfn, unsigned long end_pfn);
 137 unsigned long
 138 isolate_migratepages_range(struct zone *zone, struct compact_control *cc,
 139                            unsigned long low_pfn, unsigned long end_pfn);
 140
 141 #endif
 142
 143 /*
 144  * function for dealing with page's order in buddy system.
 145  * zone->lock is already acquired when we use these.
 146  * So, we don't need atomic page->flags operations here.
 147  */
 148 static inline unsigned long page_order(struct page *page)
 149 {
 150         /* PageBuddy() must be checked by the caller */
 151         return page_private(page);
 152 }
 153
 154 /* mm/util.c */
 155 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
 156                 struct vm_area_struct *prev, struct rb_node *rb_parent);
 157
 158 #ifdef CONFIG_MMU
 159 extern long mlock_vma_pages_range(struct vm_area_struct *vma,
 160                         unsigned long start, unsigned long end);
 161 extern void munlock_vma_pages_range(struct vm_area_struct *vma,
 162                         unsigned long start, unsigned long end);
 163 static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
 164 {
 165         munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
 166 }
 167
 168 /*
 169  * Called only in fault path via page_evictable() for a new page
 170  * to determine if it's being mapped into a LOCKED vma.
 171  * If so, mark page as mlocked.
 172  */
 173 static inline int mlocked_vma_newpage(struct vm_area_struct *vma,
 174                                     struct page *page)
 175 {
 176         VM_BUG_ON(PageLRU(page));
 177
 178         if (likely((vma->vm_flags & (VM_LOCKED | VM_SPECIAL)) != VM_LOCKED))
 179                 return 0;
 180
 181         if (!TestSetPageMlocked(page)) {
 182                 inc_zone_page_state(page, NR_MLOCK);
 183                 count_vm_event(UNEVICTABLE_PGMLOCKED);
 184         }
 185         return 1;
 186 }
 187
 188 /*
 189  * must be called with vma's mmap_sem held for read or write, and page locked.
 190  */
 191 extern void mlock_vma_page(struct page *page);
 192 extern void munlock_vma_page(struct page *page);
 193
 194 /*
 195  * Clear the page's PageMlocked().  This can be useful in a situation where
 196  * we want to unconditionally remove a page from the pagecache -- e.g.,
 197  * on truncation or freeing.
 198  *
 199  * It is legal to call this function for any page, mlocked or not.
 200  * If called for a page that is still mapped by mlocked vmas, all we do
 201  * is revert to lazy LRU behaviour -- semantics are not broken.
 202  */
 203 extern void __clear_page_mlock(struct page *page);
 204 static inline void clear_page_mlock(struct page *page)
 205 {
 206         if (unlikely(TestClearPageMlocked(page)))
 207                 __clear_page_mlock(page);
 208 }
 209
 210 /*
 211  * mlock_migrate_page - called only from migrate_page_copy() to
 212  * migrate the Mlocked page flag; update statistics.
 213  */
 214 static inline void mlock_migrate_page(struct page *newpage, struct page *page)
 215 {
 216         if (TestClearPageMlocked(page)) {
 217                 unsigned long flags;
 218
 219                 local_irq_save(flags);
 220                 __dec_zone_page_state(page, NR_MLOCK);
 221                 SetPageMlocked(newpage);
 222                 __inc_zone_page_state(newpage, NR_MLOCK);
 223                 local_irq_restore(flags);
 224         }
 225 }
 226
 227 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 228 extern unsigned long vma_address(struct page *page,
 229                                  struct vm_area_struct *vma);
 230 #endif
 231 #else /* !CONFIG_MMU */
 232 static inline int mlocked_vma_newpage(struct vm_area_struct *v, struct page *p)
 233 {
 234         return 0;
 235 }
 236 static inline void clear_page_mlock(struct page *page) { }
 237 static inline void mlock_vma_page(struct page *page) { }
 238 static inline void mlock_migrate_page(struct page *new, struct page *old) { }
 239
 240 #endif /* !CONFIG_MMU */
 241
 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  */
 247 static 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 /*
 255  * Iterator over all subpages within the maximally aligned gigantic
 256  * page 'base'.  Handle any discontiguity in the mem_map.
 257  */
 258 static 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
 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
 282 /* Memory initialisation debug and verification */
 283 enum mminit_level {
 284         MMINIT_WARNING,
 285         MMINIT_VERIFY,
 286         MMINIT_TRACE
 287 };
 288
 289 #ifdef CONFIG_DEBUG_MEMORY_INIT
 290
 291 extern int mminit_loglevel;
 292
 293 #define mminit_dprintk(level, prefix, fmt, arg...) \
 294 do { \
 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
 301 extern void mminit_verify_pageflags_layout(void);
 302 extern void mminit_verify_page_links(struct page *page,
 303                 enum zone_type zone, unsigned long nid, unsigned long pfn);
 304 extern void mminit_verify_zonelist(void);
 305
 306 #else
 307
 308 static inline void mminit_dprintk(enum mminit_level level,
 309                                 const char *prefix, const char *fmt, ...)
 310 {
 311 }
 312
 313 static inline void mminit_verify_pageflags_layout(void)
 314 {
 315 }
 316
 317 static inline void mminit_verify_page_links(struct page *page,
 318                 enum zone_type zone, unsigned long nid, unsigned long pfn)
 319 {
 320 }
 321
 322 static inline void mminit_verify_zonelist(void)
 323 {
 324 }
 325 #endif /* CONFIG_DEBUG_MEMORY_INIT */
 326
 327 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
 328 #if defined(CONFIG_SPARSEMEM)
 329 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 330                                 unsigned long *end_pfn);
 331 #else
 332 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 333                                 unsigned long *end_pfn)
 334 {
 335 }
 336 #endif /* CONFIG_SPARSEMEM */
 337
 338 #define ZONE_RECLAIM_NOSCAN     -2
 339 #define ZONE_RECLAIM_FULL       -1
 340 #define ZONE_RECLAIM_SOME       0
 341 #define ZONE_RECLAIM_SUCCESS    1
 342 #endif
 343
 344 extern int hwpoison_filter(struct page *p);
 345
 346 extern u32 hwpoison_filter_dev_major;
 347 extern u32 hwpoison_filter_dev_minor;
 348 extern u64 hwpoison_filter_flags_mask;
 349 extern u64 hwpoison_filter_flags_value;
 350 extern u64 hwpoison_filter_memcg;
 351 extern u32 hwpoison_filter_enable;
 352
 353 extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
 354         unsigned long, unsigned long,
 355         unsigned long, unsigned long);
 356
 357 extern void set_pageblock_order(void);