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         struct page **page;             /* Page captured of requested size */
 139 };
 140
 141 unsigned long
 142 isolate_freepages_range(struct compact_control *cc,
 143                         unsigned long start_pfn, unsigned long end_pfn);
 144 unsigned long
 145 isolate_migratepages_range(struct zone *zone, struct compact_control *cc,
 146         unsigned long low_pfn, unsigned long end_pfn, bool unevictable);
 147
 148 #endif
 149
 150 /*
 151  * function for dealing with page's order in buddy system.
 152  * zone->lock is already acquired when we use these.
 153  * So, we don't need atomic page->flags operations here.
 154  */
 155 static inline unsigned long page_order(struct page *page)
 156 {
 157         /* PageBuddy() must be checked by the caller */
 158         return page_private(page);
 159 }
 160
 161 /* mm/util.c */
 162 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
 163                 struct vm_area_struct *prev, struct rb_node *rb_parent);
 164
 165 #ifdef CONFIG_MMU
 166 extern long mlock_vma_pages_range(struct vm_area_struct *vma,
 167                         unsigned long start, unsigned long end);
 168 extern void munlock_vma_pages_range(struct vm_area_struct *vma,
 169                         unsigned long start, unsigned long end);
 170 static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
 171 {
 172         munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
 173 }
 174
 175 /*
 176  * Called only in fault path, to determine if a new page is being
 177  * mapped into a LOCKED vma.  If it is, mark page as mlocked.
 178  */
 179 static inline int mlocked_vma_newpage(struct vm_area_struct *vma,
 180                                     struct page *page)
 181 {
 182         VM_BUG_ON(PageLRU(page));
 183
 184         if (likely((vma->vm_flags & (VM_LOCKED | VM_SPECIAL)) != VM_LOCKED))
 185                 return 0;
 186
 187         if (!TestSetPageMlocked(page)) {
 188                 mod_zone_page_state(page_zone(page), NR_MLOCK,
 189                                     hpage_nr_pages(page));
 190                 count_vm_event(UNEVICTABLE_PGMLOCKED);
 191         }
 192         return 1;
 193 }
 194
 195 /*
 196  * must be called with vma's mmap_sem held for read or write, and page locked.
 197  */
 198 extern void mlock_vma_page(struct page *page);
 199 extern void munlock_vma_page(struct page *page);
 200
 201 /*
 202  * Clear the page's PageMlocked().  This can be useful in a situation where
 203  * we want to unconditionally remove a page from the pagecache -- e.g.,
 204  * on truncation or freeing.
 205  *
 206  * It is legal to call this function for any page, mlocked or not.
 207  * If called for a page that is still mapped by mlocked vmas, all we do
 208  * is revert to lazy LRU behaviour -- semantics are not broken.
 209  */
 210 extern void clear_page_mlock(struct page *page);
 211
 212 /*
 213  * mlock_migrate_page - called only from migrate_page_copy() to
 214  * migrate the Mlocked page flag; update statistics.
 215  */
 216 static inline void mlock_migrate_page(struct page *newpage, struct page *page)
 217 {
 218         if (TestClearPageMlocked(page)) {
 219                 unsigned long flags;
 220                 int nr_pages = hpage_nr_pages(page);
 221
 222                 local_irq_save(flags);
 223                 __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
 224                 SetPageMlocked(newpage);
 225                 __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
 226                 local_irq_restore(flags);
 227         }
 228 }
 229
 230 extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
 231
 232 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 233 extern unsigned long vma_address(struct page *page,
 234                                  struct vm_area_struct *vma);
 235 #endif
 236 #else /* !CONFIG_MMU */
 237 static inline int mlocked_vma_newpage(struct vm_area_struct *v, struct page *p)
 238 {
 239         return 0;
 240 }
 241 static inline void clear_page_mlock(struct page *page) { }
 242 static inline void mlock_vma_page(struct page *page) { }
 243 static inline void mlock_migrate_page(struct page *new, struct page *old) { }
 244
 245 #endif /* !CONFIG_MMU */
 246
 247 /*
 248  * Return the mem_map entry representing the 'offset' subpage within
 249  * the maximally aligned gigantic page 'base'.  Handle any discontiguity
 250  * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
 251  */
 252 static inline struct page *mem_map_offset(struct page *base, int offset)
 253 {
 254         if (unlikely(offset >= MAX_ORDER_NR_PAGES))
 255                 return pfn_to_page(page_to_pfn(base) + offset);
 256         return base + offset;
 257 }
 258
 259 /*
 260  * Iterator over all subpages within the maximally aligned gigantic
 261  * page 'base'.  Handle any discontiguity in the mem_map.
 262  */
 263 static inline struct page *mem_map_next(struct page *iter,
 264                                                 struct page *base, int offset)
 265 {
 266         if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
 267                 unsigned long pfn = page_to_pfn(base) + offset;
 268                 if (!pfn_valid(pfn))
 269                         return NULL;
 270                 return pfn_to_page(pfn);
 271         }
 272         return iter + 1;
 273 }
 274
 275 /*
 276  * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
 277  * so all functions starting at paging_init should be marked __init
 278  * in those cases. SPARSEMEM, however, allows for memory hotplug,
 279  * and alloc_bootmem_node is not used.
 280  */
 281 #ifdef CONFIG_SPARSEMEM
 282 #define __paginginit __meminit
 283 #else
 284 #define __paginginit __init
 285 #endif
 286
 287 /* Memory initialisation debug and verification */
 288 enum mminit_level {
 289         MMINIT_WARNING,
 290         MMINIT_VERIFY,
 291         MMINIT_TRACE
 292 };
 293
 294 #ifdef CONFIG_DEBUG_MEMORY_INIT
 295
 296 extern int mminit_loglevel;
 297
 298 #define mminit_dprintk(level, prefix, fmt, arg...) \
 299 do { \
 300         if (level < mminit_loglevel) { \
 301                 printk(level <= MMINIT_WARNING ? KERN_WARNING : KERN_DEBUG); \
 302                 printk(KERN_CONT "mminit::" prefix " " fmt, ##arg); \
 303         } \
 304 } while (0)
 305
 306 extern void mminit_verify_pageflags_layout(void);
 307 extern void mminit_verify_page_links(struct page *page,
 308                 enum zone_type zone, unsigned long nid, unsigned long pfn);
 309 extern void mminit_verify_zonelist(void);
 310
 311 #else
 312
 313 static inline void mminit_dprintk(enum mminit_level level,
 314                                 const char *prefix, const char *fmt, ...)
 315 {
 316 }
 317
 318 static inline void mminit_verify_pageflags_layout(void)
 319 {
 320 }
 321
 322 static inline void mminit_verify_page_links(struct page *page,
 323                 enum zone_type zone, unsigned long nid, unsigned long pfn)
 324 {
 325 }
 326
 327 static inline void mminit_verify_zonelist(void)
 328 {
 329 }
 330 #endif /* CONFIG_DEBUG_MEMORY_INIT */
 331
 332 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
 333 #if defined(CONFIG_SPARSEMEM)
 334 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 335                                 unsigned long *end_pfn);
 336 #else
 337 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 338                                 unsigned long *end_pfn)
 339 {
 340 }
 341 #endif /* CONFIG_SPARSEMEM */
 342
 343 #define ZONE_RECLAIM_NOSCAN     -2
 344 #define ZONE_RECLAIM_FULL       -1
 345 #define ZONE_RECLAIM_SOME       0
 346 #define ZONE_RECLAIM_SUCCESS    1
 347
 348 extern int hwpoison_filter(struct page *p);
 349
 350 extern u32 hwpoison_filter_dev_major;
 351 extern u32 hwpoison_filter_dev_minor;
 352 extern u64 hwpoison_filter_flags_mask;
 353 extern u64 hwpoison_filter_flags_value;
 354 extern u64 hwpoison_filter_memcg;
 355 extern u32 hwpoison_filter_enable;
 356
 357 extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
 358         unsigned long, unsigned long,
 359         unsigned long, unsigned long);
 360
 361 extern void set_pageblock_order(void);
 362 unsigned long reclaim_clean_pages_from_list(struct zone *zone,
 363                                             struct list_head *page_list);
 364 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
 365 #define ALLOC_WMARK_MIN         WMARK_MIN
 366 #define ALLOC_WMARK_LOW         WMARK_LOW
 367 #define ALLOC_WMARK_HIGH        WMARK_HIGH
 368 #define ALLOC_NO_WATERMARKS     0x04 /* don't check watermarks at all */
 369
 370 /* Mask to get the watermark bits */
 371 #define ALLOC_WMARK_MASK        (ALLOC_NO_WATERMARKS-1)
 372
 373 #define ALLOC_HARDER            0x10 /* try to alloc harder */
 374 #define ALLOC_HIGH              0x20 /* __GFP_HIGH set */
 375 #define ALLOC_CPUSET            0x40 /* check for correct cpuset */
 376 #define ALLOC_CMA               0x80 /* allow allocations from CMA areas */
 377
 378 #endif  /* __MM_INTERNAL_H */