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_PAGE(PageTail(page), page);
  31         VM_BUG_ON_PAGE(atomic_read(&page->_count), page);
  32         set_page_count(page, 1);
  33 }
  34
  35 static 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          */
  49         VM_BUG_ON_PAGE(atomic_read(&page->first_page->_count) <= 0, page);
  50         if (get_page_head)
  51                 atomic_inc(&page->first_page->_count);
  52         get_huge_page_tail(page);
  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  */
  60 static 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                  */
  74                 VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page);
  75                 atomic_inc(&page->_count);
  76         }
  77 }
  78
  79 extern unsigned long highest_memmap_pfn;
  80
  81 /*
  82  * in mm/vmscan.c:
  83  */
  84 extern int isolate_lru_page(struct page *page);
  85 extern void putback_lru_page(struct page *page);
  86 extern unsigned long zone_reclaimable_pages(struct zone *zone);
  87 extern bool zone_reclaimable(struct zone *zone);
  88
  89 /*
  90  * in mm/rmap.c:
  91  */
  92 extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
  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 extern int user_min_free_kbytes;
 103
 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  */
 116 struct 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 */
 123         bool sync;                      /* Synchronous migration */
 124         bool ignore_skip_hint;          /* Scan blocks even if marked skip */
 125         bool finished_update_free;      /* True when the zone cached pfns are
 126                                          * no longer being updated
 127                                          */
 128         bool finished_update_migrate;
 129
 130         int order;                      /* order a direct compactor needs */
 131         int migratetype;                /* MOVABLE, RECLAIMABLE etc */
 132         struct zone *zone;
 133         bool contended;                 /* True if a lock was contended */
 134 };
 135
 136 unsigned long
 137 isolate_freepages_range(struct compact_control *cc,
 138                         unsigned long start_pfn, unsigned long end_pfn);
 139 unsigned long
 140 isolate_migratepages_range(struct zone *zone, struct compact_control *cc,
 141         unsigned long low_pfn, unsigned long end_pfn, bool unevictable);
 142
 143 #endif
 144
 145 /*
 146  * This function returns the order of a free page in the buddy system. In
 147  * general, page_zone(page)->lock must be held by the caller to prevent the
 148  * page from being allocated in parallel and returning garbage as the order.
 149  * If a caller does not hold page_zone(page)->lock, it must guarantee that the
 150  * page cannot be allocated or merged in parallel.
 151  */
 152 static inline unsigned long page_order(struct page *page)
 153 {
 154         /* PageBuddy() must be checked by the caller */
 155         return page_private(page);
 156 }
 157
 158 /* mm/util.c */
 159 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
 160                 struct vm_area_struct *prev, struct rb_node *rb_parent);
 161
 162 #ifdef CONFIG_MMU
 163 extern long __mlock_vma_pages_range(struct vm_area_struct *vma,
 164                 unsigned long start, unsigned long end, int *nonblocking);
 165 extern void munlock_vma_pages_range(struct vm_area_struct *vma,
 166                         unsigned long start, unsigned long end);
 167 static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
 168 {
 169         munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
 170 }
 171
 172 /*
 173  * Called only in fault path, to determine if a new page is being
 174  * mapped into a LOCKED vma.  If it is, mark page as mlocked.
 175  */
 176 static inline int mlocked_vma_newpage(struct vm_area_struct *vma,
 177                                     struct page *page)
 178 {
 179         VM_BUG_ON_PAGE(PageLRU(page), page);
 180
 181         if (likely((vma->vm_flags & (VM_LOCKED | VM_SPECIAL)) != VM_LOCKED))
 182                 return 0;
 183
 184         if (!TestSetPageMlocked(page)) {
 185                 mod_zone_page_state(page_zone(page), NR_MLOCK,
 186                                     hpage_nr_pages(page));
 187                 count_vm_event(UNEVICTABLE_PGMLOCKED);
 188         }
 189         return 1;
 190 }
 191
 192 /*
 193  * must be called with vma's mmap_sem held for read or write, and page locked.
 194  */
 195 extern void mlock_vma_page(struct page *page);
 196 extern unsigned int munlock_vma_page(struct page *page);
 197
 198 /*
 199  * Clear the page's PageMlocked().  This can be useful in a situation where
 200  * we want to unconditionally remove a page from the pagecache -- e.g.,
 201  * on truncation or freeing.
 202  *
 203  * It is legal to call this function for any page, mlocked or not.
 204  * If called for a page that is still mapped by mlocked vmas, all we do
 205  * is revert to lazy LRU behaviour -- semantics are not broken.
 206  */
 207 extern void clear_page_mlock(struct page *page);
 208
 209 /*
 210  * mlock_migrate_page - called only from migrate_page_copy() to
 211  * migrate the Mlocked page flag; update statistics.
 212  */
 213 static inline void mlock_migrate_page(struct page *newpage, struct page *page)
 214 {
 215         if (TestClearPageMlocked(page)) {
 216                 unsigned long flags;
 217                 int nr_pages = hpage_nr_pages(page);
 218
 219                 local_irq_save(flags);
 220                 __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
 221                 SetPageMlocked(newpage);
 222                 __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
 223                 local_irq_restore(flags);
 224         }
 225 }
 226
 227 extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
 228
 229 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 230 extern unsigned long vma_address(struct page *page,
 231                                  struct vm_area_struct *vma);
 232 #endif
 233 #else /* !CONFIG_MMU */
 234 static inline int mlocked_vma_newpage(struct vm_area_struct *v, struct page *p)
 235 {
 236         return 0;
 237 }
 238 static inline void clear_page_mlock(struct page *page) { }
 239 static inline void mlock_vma_page(struct page *page) { }
 240 static inline void mlock_migrate_page(struct page *new, struct page *old) { }
 241
 242 #endif /* !CONFIG_MMU */
 243
 244 /*
 245  * Return the mem_map entry representing the 'offset' subpage within
 246  * the maximally aligned gigantic page 'base'.  Handle any discontiguity
 247  * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
 248  */
 249 static inline struct page *mem_map_offset(struct page *base, int offset)
 250 {
 251         if (unlikely(offset >= MAX_ORDER_NR_PAGES))
 252                 return pfn_to_page(page_to_pfn(base) + offset);
 253         return base + offset;
 254 }
 255
 256 /*
 257  * Iterator over all subpages within the maximally aligned gigantic
 258  * page 'base'.  Handle any discontiguity in the mem_map.
 259  */
 260 static inline struct page *mem_map_next(struct page *iter,
 261                                                 struct page *base, int offset)
 262 {
 263         if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
 264                 unsigned long pfn = page_to_pfn(base) + offset;
 265                 if (!pfn_valid(pfn))
 266                         return NULL;
 267                 return pfn_to_page(pfn);
 268         }
 269         return iter + 1;
 270 }
 271
 272 /*
 273  * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
 274  * so all functions starting at paging_init should be marked __init
 275  * in those cases. SPARSEMEM, however, allows for memory hotplug,
 276  * and alloc_bootmem_node is not used.
 277  */
 278 #ifdef CONFIG_SPARSEMEM
 279 #define __paginginit __meminit
 280 #else
 281 #define __paginginit __init
 282 #endif
 283
 284 /* Memory initialisation debug and verification */
 285 enum mminit_level {
 286         MMINIT_WARNING,
 287         MMINIT_VERIFY,
 288         MMINIT_TRACE
 289 };
 290
 291 #ifdef CONFIG_DEBUG_MEMORY_INIT
 292
 293 extern int mminit_loglevel;
 294
 295 #define mminit_dprintk(level, prefix, fmt, arg...) \
 296 do { \
 297         if (level < mminit_loglevel) { \
 298                 printk(level <= MMINIT_WARNING ? KERN_WARNING : KERN_DEBUG); \
 299                 printk(KERN_CONT "mminit::" prefix " " fmt, ##arg); \
 300         } \
 301 } while (0)
 302
 303 extern void mminit_verify_pageflags_layout(void);
 304 extern void mminit_verify_page_links(struct page *page,
 305                 enum zone_type zone, unsigned long nid, unsigned long pfn);
 306 extern void mminit_verify_zonelist(void);
 307
 308 #else
 309
 310 static inline void mminit_dprintk(enum mminit_level level,
 311                                 const char *prefix, const char *fmt, ...)
 312 {
 313 }
 314
 315 static inline void mminit_verify_pageflags_layout(void)
 316 {
 317 }
 318
 319 static inline void mminit_verify_page_links(struct page *page,
 320                 enum zone_type zone, unsigned long nid, unsigned long pfn)
 321 {
 322 }
 323
 324 static inline void mminit_verify_zonelist(void)
 325 {
 326 }
 327 #endif /* CONFIG_DEBUG_MEMORY_INIT */
 328
 329 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
 330 #if defined(CONFIG_SPARSEMEM)
 331 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 332                                 unsigned long *end_pfn);
 333 #else
 334 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 335                                 unsigned long *end_pfn)
 336 {
 337 }
 338 #endif /* CONFIG_SPARSEMEM */
 339
 340 #define ZONE_RECLAIM_NOSCAN     -2
 341 #define ZONE_RECLAIM_FULL       -1
 342 #define ZONE_RECLAIM_SOME       0
 343 #define ZONE_RECLAIM_SUCCESS    1
 344
 345 extern int hwpoison_filter(struct page *p);
 346
 347 extern u32 hwpoison_filter_dev_major;
 348 extern u32 hwpoison_filter_dev_minor;
 349 extern u64 hwpoison_filter_flags_mask;
 350 extern u64 hwpoison_filter_flags_value;
 351 extern u64 hwpoison_filter_memcg;
 352 extern u32 hwpoison_filter_enable;
 353
 354 extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
 355         unsigned long, unsigned long,
 356         unsigned long, unsigned long);
 357
 358 extern void set_pageblock_order(void);
 359 unsigned long reclaim_clean_pages_from_list(struct zone *zone,
 360                                             struct list_head *page_list);
 361 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
 362 #define ALLOC_WMARK_MIN         WMARK_MIN
 363 #define ALLOC_WMARK_LOW         WMARK_LOW
 364 #define ALLOC_WMARK_HIGH        WMARK_HIGH
 365 #define ALLOC_NO_WATERMARKS     0x04 /* don't check watermarks at all */
 366
 367 /* Mask to get the watermark bits */
 368 #define ALLOC_WMARK_MASK        (ALLOC_NO_WATERMARKS-1)
 369
 370 #define ALLOC_HARDER            0x10 /* try to alloc harder */
 371 #define ALLOC_HIGH              0x20 /* __GFP_HIGH set */
 372 #define ALLOC_CPUSET            0x40 /* check for correct cpuset */
 373 #define ALLOC_CMA               0x80 /* allow allocations from CMA areas */
 374
 375 #endif  /* __MM_INTERNAL_H */