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/rmap.c:
  96  */
  97 extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
  98
  99 /*
 100  * in mm/page_alloc.c
 101  */
 102 extern void __free_pages_bootmem(struct page *page, unsigned int order);
 103 extern void prep_compound_page(struct page *page, unsigned long order);
 104 #ifdef CONFIG_MEMORY_FAILURE
 105 extern bool is_free_buddy_page(struct page *page);
 106 #endif
 107
 108 #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  */
 120 struct 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 */
 127         bool sync;                      /* Synchronous migration */
 128         bool ignore_skip_hint;          /* Scan blocks even if marked skip */
 129         bool finished_update_free;      /* True when the zone cached pfns are
 130                                          * no longer being updated
 131                                          */
 132         bool finished_update_migrate;
 133
 134         int order;                      /* order a direct compactor needs */
 135         int migratetype;                /* MOVABLE, RECLAIMABLE etc */
 136         struct zone *zone;
 137         bool contended;                 /* True if a lock was contended */
 138 };
 139
 140 unsigned long
 141 isolate_freepages_range(struct compact_control *cc,
 142                         unsigned long start_pfn, unsigned long end_pfn);
 143 unsigned long
 144 isolate_migratepages_range(struct zone *zone, struct compact_control *cc,
 145         unsigned long low_pfn, unsigned long end_pfn, bool unevictable);
 146
 147 #endif
 148
 149 /*
 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  */
 154 static inline unsigned long page_order(struct page *page)
 155 {
 156         /* PageBuddy() must be checked by the caller */
 157         return page_private(page);
 158 }
 159
 160 /* mm/util.c */
 161 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
 162                 struct vm_area_struct *prev, struct rb_node *rb_parent);
 163
 164 #ifdef CONFIG_MMU
 165 extern long mlock_vma_pages_range(struct vm_area_struct *vma,
 166                         unsigned long start, unsigned long end);
 167 extern void munlock_vma_pages_range(struct vm_area_struct *vma,
 168                         unsigned long start, unsigned long end);
 169 static 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
 174 /*
 175  * 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.
 177  */
 178 static inline int mlocked_vma_newpage(struct vm_area_struct *vma,
 179                                     struct page *page)
 180 {
 181         VM_BUG_ON(PageLRU(page));
 182
 183         if (likely((vma->vm_flags & (VM_LOCKED | VM_SPECIAL)) != VM_LOCKED))
 184                 return 0;
 185
 186         if (!TestSetPageMlocked(page)) {
 187                 mod_zone_page_state(page_zone(page), NR_MLOCK,
 188                                     hpage_nr_pages(page));
 189                 count_vm_event(UNEVICTABLE_PGMLOCKED);
 190         }
 191         return 1;
 192 }
 193
 194 /*
 195  * must be called with vma's mmap_sem held for read or write, and page locked.
 196  */
 197 extern void mlock_vma_page(struct page *page);
 198 extern void munlock_vma_page(struct page *page);
 199
 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  */
 209 extern void clear_page_mlock(struct page *page);
 210
 211 /*
 212  * mlock_migrate_page - called only from migrate_page_copy() to
 213  * migrate the Mlocked page flag; update statistics.
 214  */
 215 static inline void mlock_migrate_page(struct page *newpage, struct page *page)
 216 {
 217         if (TestClearPageMlocked(page)) {
 218                 unsigned long flags;
 219                 int nr_pages = hpage_nr_pages(page);
 220
 221                 local_irq_save(flags);
 222                 __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
 223                 SetPageMlocked(newpage);
 224                 __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
 225                 local_irq_restore(flags);
 226         }
 227 }
 228
 229 extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
 230
 231 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 232 extern unsigned long vma_address(struct page *page,
 233                                  struct vm_area_struct *vma);
 234 #endif
 235 #else /* !CONFIG_MMU */
 236 static inline int mlocked_vma_newpage(struct vm_area_struct *v, struct page *p)
 237 {
 238         return 0;
 239 }
 240 static inline void clear_page_mlock(struct page *page) { }
 241 static inline void mlock_vma_page(struct page *page) { }
 242 static inline void mlock_migrate_page(struct page *new, struct page *old) { }
 243
 244 #endif /* !CONFIG_MMU */
 245
 246 /*
 247  * 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  */
 251 static 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 /*
 259  * Iterator over all subpages within the maximally aligned gigantic
 260  * page 'base'.  Handle any discontiguity in the mem_map.
 261  */
 262 static 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 /*
 275  * 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
 286 /* Memory initialisation debug and verification */
 287 enum mminit_level {
 288         MMINIT_WARNING,
 289         MMINIT_VERIFY,
 290         MMINIT_TRACE
 291 };
 292
 293 #ifdef CONFIG_DEBUG_MEMORY_INIT
 294
 295 extern int mminit_loglevel;
 296
 297 #define mminit_dprintk(level, prefix, fmt, arg...) \
 298 do { \
 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
 305 extern void mminit_verify_pageflags_layout(void);
 306 extern void mminit_verify_page_links(struct page *page,
 307                 enum zone_type zone, unsigned long nid, unsigned long pfn);
 308 extern void mminit_verify_zonelist(void);
 309
 310 #else
 311
 312 static inline void mminit_dprintk(enum mminit_level level,
 313                                 const char *prefix, const char *fmt, ...)
 314 {
 315 }
 316
 317 static inline void mminit_verify_pageflags_layout(void)
 318 {
 319 }
 320
 321 static inline void mminit_verify_page_links(struct page *page,
 322                 enum zone_type zone, unsigned long nid, unsigned long pfn)
 323 {
 324 }
 325
 326 static inline void mminit_verify_zonelist(void)
 327 {
 328 }
 329 #endif /* CONFIG_DEBUG_MEMORY_INIT */
 330
 331 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
 332 #if defined(CONFIG_SPARSEMEM)
 333 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 334                                 unsigned long *end_pfn);
 335 #else
 336 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 337                                 unsigned long *end_pfn)
 338 {
 339 }
 340 #endif /* CONFIG_SPARSEMEM */
 341
 342 #define ZONE_RECLAIM_NOSCAN     -2
 343 #define ZONE_RECLAIM_FULL       -1
 344 #define ZONE_RECLAIM_SOME       0
 345 #define ZONE_RECLAIM_SUCCESS    1
 346
 347 extern int hwpoison_filter(struct page *p);
 348
 349 extern u32 hwpoison_filter_dev_major;
 350 extern u32 hwpoison_filter_dev_minor;
 351 extern u64 hwpoison_filter_flags_mask;
 352 extern u64 hwpoison_filter_flags_value;
 353 extern u64 hwpoison_filter_memcg;
 354 extern u32 hwpoison_filter_enable;
 355
 356 extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
 357         unsigned long, unsigned long,
 358         unsigned long, unsigned long);
 359
 360 extern void set_pageblock_order(void);
 361 unsigned long reclaim_clean_pages_from_list(struct zone *zone,
 362                                             struct list_head *page_list);
 363 /* 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
 377 #endif  /* __MM_INTERNAL_H */