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/fs.h>
  15 #include <linux/mm.h>
  16
  17 void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
  18                 unsigned long floor, unsigned long ceiling);
  19
  20 static inline void set_page_count(struct page *page, int v)
  21 {
  22         atomic_set(&page->_count, v);
  23 }
  24
  25 extern int __do_page_cache_readahead(struct address_space *mapping,
  26                 struct file *filp, pgoff_t offset, unsigned long nr_to_read,
  27                 unsigned long lookahead_size);
  28
  29 /*
  30  * Submit IO for the read-ahead request in file_ra_state.
  31  */
  32 static inline unsigned long ra_submit(struct file_ra_state *ra,
  33                 struct address_space *mapping, struct file *filp)
  34 {
  35         return __do_page_cache_readahead(mapping, filp,
  36                                         ra->start, ra->size, ra->async_size);
  37 }
  38
  39 /*
  40  * Turn a non-refcounted page (->_count == 0) into refcounted with
  41  * a count of one.
  42  */
  43 static inline void set_page_refcounted(struct page *page)
  44 {
  45         VM_BUG_ON(PageTail(page));
  46         VM_BUG_ON(atomic_read(&page->_count));
  47         set_page_count(page, 1);
  48 }
  49
  50 static inline void __get_page_tail_foll(struct page *page,
  51                                         bool get_page_head)
  52 {
  53         /*
  54          * If we're getting a tail page, the elevated page->_count is
  55          * required only in the head page and we will elevate the head
  56          * page->_count and tail page->_mapcount.
  57          *
  58          * We elevate page_tail->_mapcount for tail pages to force
  59          * page_tail->_count to be zero at all times to avoid getting
  60          * false positives from get_page_unless_zero() with
  61          * speculative page access (like in
  62          * page_cache_get_speculative()) on tail pages.
  63          */
  64         VM_BUG_ON(atomic_read(&page->first_page->_count) <= 0);
  65         VM_BUG_ON(atomic_read(&page->_count) != 0);
  66         VM_BUG_ON(page_mapcount(page) < 0);
  67         if (get_page_head)
  68                 atomic_inc(&page->first_page->_count);
  69         atomic_inc(&page->_mapcount);
  70 }
  71
  72 /*
  73  * This is meant to be called as the FOLL_GET operation of
  74  * follow_page() and it must be called while holding the proper PT
  75  * lock while the pte (or pmd_trans_huge) is still mapping the page.
  76  */
  77 static inline void get_page_foll(struct page *page)
  78 {
  79         if (unlikely(PageTail(page)))
  80                 /*
  81                  * This is safe only because
  82                  * __split_huge_page_refcount() can't run under
  83                  * get_page_foll() because we hold the proper PT lock.
  84                  */
  85                 __get_page_tail_foll(page, true);
  86         else {
  87                 /*
  88                  * Getting a normal page or the head of a compound page
  89                  * requires to already have an elevated page->_count.
  90                  */
  91                 VM_BUG_ON(atomic_read(&page->_count) <= 0);
  92                 atomic_inc(&page->_count);
  93         }
  94 }
  95
  96 extern unsigned long highest_memmap_pfn;
  97
  98 /*
  99  * in mm/vmscan.c:
 100  */
 101 extern int isolate_lru_page(struct page *page);
 102 extern void putback_lru_page(struct page *page);
 103 extern bool zone_reclaimable(struct zone *zone);
 104
 105 /*
 106  * in mm/rmap.c:
 107  */
 108 extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
 109
 110 /*
 111  * in mm/page_alloc.c
 112  */
 113 extern void __free_pages_bootmem(struct page *page, unsigned int order);
 114 extern void prep_compound_page(struct page *page, unsigned long order);
 115 #ifdef CONFIG_MEMORY_FAILURE
 116 extern bool is_free_buddy_page(struct page *page);
 117 #endif
 118
 119 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
 120
 121 /*
 122  * in mm/compaction.c
 123  */
 124 /*
 125  * compact_control is used to track pages being migrated and the free pages
 126  * they are being migrated to during memory compaction. The free_pfn starts
 127  * at the end of a zone and migrate_pfn begins at the start. Movable pages
 128  * are moved to the end of a zone during a compaction run and the run
 129  * completes when free_pfn <= migrate_pfn
 130  */
 131 struct compact_control {
 132         struct list_head freepages;     /* List of free pages to migrate to */
 133         struct list_head migratepages;  /* List of pages being migrated */
 134         unsigned long nr_freepages;     /* Number of isolated free pages */
 135         unsigned long nr_migratepages;  /* Number of pages to migrate */
 136         unsigned long free_pfn;         /* isolate_freepages search base */
 137         unsigned long migrate_pfn;      /* isolate_migratepages search base */
 138         enum migrate_mode mode;         /* Async or sync migration mode */
 139         bool ignore_skip_hint;          /* Scan blocks even if marked skip */
 140         bool finished_update_free;      /* True when the zone cached pfns are
 141                                          * no longer being updated
 142                                          */
 143         bool finished_update_migrate;
 144
 145         int order;                      /* order a direct compactor needs */
 146         int migratetype;                /* MOVABLE, RECLAIMABLE etc */
 147         struct zone *zone;
 148         bool contended;                 /* True if a lock was contended, or
 149                                          * need_resched() true during async
 150                                          * compaction
 151                                          */
 152 };
 153
 154 unsigned long
 155 isolate_freepages_range(struct compact_control *cc,
 156                         unsigned long start_pfn, unsigned long end_pfn);
 157 unsigned long
 158 isolate_migratepages_range(struct zone *zone, struct compact_control *cc,
 159         unsigned long low_pfn, unsigned long end_pfn, bool unevictable);
 160
 161 #endif
 162
 163 /*
 164  * function for dealing with page's order in buddy system.
 165  * zone->lock is already acquired when we use these.
 166  * So, we don't need atomic page->flags operations here.
 167  */
 168 static inline unsigned long page_order(struct page *page)
 169 {
 170         /* PageBuddy() must be checked by the caller */
 171         return page_private(page);
 172 }
 173
 174 /* mm/util.c */
 175 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
 176                 struct vm_area_struct *prev, struct rb_node *rb_parent);
 177
 178 #ifdef CONFIG_MMU
 179 extern long __mlock_vma_pages_range(struct vm_area_struct *vma,
 180                 unsigned long start, unsigned long end, int *nonblocking);
 181 extern void munlock_vma_pages_range(struct vm_area_struct *vma,
 182                         unsigned long start, unsigned long end);
 183 static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
 184 {
 185         munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
 186 }
 187
 188 /*
 189  * Called only in fault path, to determine if a new page is being
 190  * mapped into a LOCKED vma.  If it is, mark page as mlocked.
 191  */
 192 static inline int mlocked_vma_newpage(struct vm_area_struct *vma,
 193                                     struct page *page)
 194 {
 195         VM_BUG_ON(PageLRU(page));
 196
 197         if (likely((vma->vm_flags & (VM_LOCKED | VM_SPECIAL)) != VM_LOCKED))
 198                 return 0;
 199
 200         if (!TestSetPageMlocked(page)) {
 201                 mod_zone_page_state(page_zone(page), NR_MLOCK,
 202                                     hpage_nr_pages(page));
 203                 count_vm_event(UNEVICTABLE_PGMLOCKED);
 204         }
 205         return 1;
 206 }
 207
 208 /*
 209  * must be called with vma's mmap_sem held for read or write, and page locked.
 210  */
 211 extern void mlock_vma_page(struct page *page);
 212 extern unsigned int munlock_vma_page(struct page *page);
 213
 214 /*
 215  * Clear the page's PageMlocked().  This can be useful in a situation where
 216  * we want to unconditionally remove a page from the pagecache -- e.g.,
 217  * on truncation or freeing.
 218  *
 219  * It is legal to call this function for any page, mlocked or not.
 220  * If called for a page that is still mapped by mlocked vmas, all we do
 221  * is revert to lazy LRU behaviour -- semantics are not broken.
 222  */
 223 extern void clear_page_mlock(struct page *page);
 224
 225 /*
 226  * mlock_migrate_page - called only from migrate_page_copy() to
 227  * migrate the Mlocked page flag; update statistics.
 228  */
 229 static inline void mlock_migrate_page(struct page *newpage, struct page *page)
 230 {
 231         if (TestClearPageMlocked(page)) {
 232                 unsigned long flags;
 233                 int nr_pages = hpage_nr_pages(page);
 234
 235                 local_irq_save(flags);
 236                 __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
 237                 SetPageMlocked(newpage);
 238                 __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
 239                 local_irq_restore(flags);
 240         }
 241 }
 242
 243 extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
 244
 245 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 246 extern unsigned long vma_address(struct page *page,
 247                                  struct vm_area_struct *vma);
 248 #endif
 249 #else /* !CONFIG_MMU */
 250 static inline int mlocked_vma_newpage(struct vm_area_struct *v, struct page *p)
 251 {
 252         return 0;
 253 }
 254 static inline void clear_page_mlock(struct page *page) { }
 255 static inline void mlock_vma_page(struct page *page) { }
 256 static inline void mlock_migrate_page(struct page *new, struct page *old) { }
 257
 258 #endif /* !CONFIG_MMU */
 259
 260 /*
 261  * Return the mem_map entry representing the 'offset' subpage within
 262  * the maximally aligned gigantic page 'base'.  Handle any discontiguity
 263  * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
 264  */
 265 static inline struct page *mem_map_offset(struct page *base, int offset)
 266 {
 267         if (unlikely(offset >= MAX_ORDER_NR_PAGES))
 268                 return pfn_to_page(page_to_pfn(base) + offset);
 269         return base + offset;
 270 }
 271
 272 /*
 273  * Iterator over all subpages within the maximally aligned gigantic
 274  * page 'base'.  Handle any discontiguity in the mem_map.
 275  */
 276 static inline struct page *mem_map_next(struct page *iter,
 277                                                 struct page *base, int offset)
 278 {
 279         if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
 280                 unsigned long pfn = page_to_pfn(base) + offset;
 281                 if (!pfn_valid(pfn))
 282                         return NULL;
 283                 return pfn_to_page(pfn);
 284         }
 285         return iter + 1;
 286 }
 287
 288 /*
 289  * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
 290  * so all functions starting at paging_init should be marked __init
 291  * in those cases. SPARSEMEM, however, allows for memory hotplug,
 292  * and alloc_bootmem_node is not used.
 293  */
 294 #ifdef CONFIG_SPARSEMEM
 295 #define __paginginit __meminit
 296 #else
 297 #define __paginginit __init
 298 #endif
 299
 300 /* Memory initialisation debug and verification */
 301 enum mminit_level {
 302         MMINIT_WARNING,
 303         MMINIT_VERIFY,
 304         MMINIT_TRACE
 305 };
 306
 307 #ifdef CONFIG_DEBUG_MEMORY_INIT
 308
 309 extern int mminit_loglevel;
 310
 311 #define mminit_dprintk(level, prefix, fmt, arg...) \
 312 do { \
 313         if (level < mminit_loglevel) { \
 314                 printk(level <= MMINIT_WARNING ? KERN_WARNING : KERN_DEBUG); \
 315                 printk(KERN_CONT "mminit::" prefix " " fmt, ##arg); \
 316         } \
 317 } while (0)
 318
 319 extern void mminit_verify_pageflags_layout(void);
 320 extern void mminit_verify_page_links(struct page *page,
 321                 enum zone_type zone, unsigned long nid, unsigned long pfn);
 322 extern void mminit_verify_zonelist(void);
 323
 324 #else
 325
 326 static inline void mminit_dprintk(enum mminit_level level,
 327                                 const char *prefix, const char *fmt, ...)
 328 {
 329 }
 330
 331 static inline void mminit_verify_pageflags_layout(void)
 332 {
 333 }
 334
 335 static inline void mminit_verify_page_links(struct page *page,
 336                 enum zone_type zone, unsigned long nid, unsigned long pfn)
 337 {
 338 }
 339
 340 static inline void mminit_verify_zonelist(void)
 341 {
 342 }
 343 #endif /* CONFIG_DEBUG_MEMORY_INIT */
 344
 345 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
 346 #if defined(CONFIG_SPARSEMEM)
 347 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 348                                 unsigned long *end_pfn);
 349 #else
 350 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 351                                 unsigned long *end_pfn)
 352 {
 353 }
 354 #endif /* CONFIG_SPARSEMEM */
 355
 356 #define ZONE_RECLAIM_NOSCAN     -2
 357 #define ZONE_RECLAIM_FULL       -1
 358 #define ZONE_RECLAIM_SOME       0
 359 #define ZONE_RECLAIM_SUCCESS    1
 360
 361 extern int hwpoison_filter(struct page *p);
 362
 363 extern u32 hwpoison_filter_dev_major;
 364 extern u32 hwpoison_filter_dev_minor;
 365 extern u64 hwpoison_filter_flags_mask;
 366 extern u64 hwpoison_filter_flags_value;
 367 extern u64 hwpoison_filter_memcg;
 368 extern u32 hwpoison_filter_enable;
 369
 370 extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
 371         unsigned long, unsigned long,
 372         unsigned long, unsigned long);
 373
 374 extern void set_pageblock_order(void);
 375 unsigned long reclaim_clean_pages_from_list(struct zone *zone,
 376                                             struct list_head *page_list);
 377 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
 378 #define ALLOC_WMARK_MIN         WMARK_MIN
 379 #define ALLOC_WMARK_LOW         WMARK_LOW
 380 #define ALLOC_WMARK_HIGH        WMARK_HIGH
 381 #define ALLOC_NO_WATERMARKS     0x04 /* don't check watermarks at all */
 382
 383 /* Mask to get the watermark bits */
 384 #define ALLOC_WMARK_MASK        (ALLOC_NO_WATERMARKS-1)
 385
 386 #define ALLOC_HARDER            0x10 /* try to alloc harder */
 387 #define ALLOC_HIGH              0x20 /* __GFP_HIGH set */
 388 #define ALLOC_CPUSET            0x40 /* check for correct cpuset */
 389 #define ALLOC_CMA               0x80 /* allow allocations from CMA areas */
 390 #define ALLOC_FAIR              0x100 /* fair zone allocation */
 391
 392 #endif  /* __MM_INTERNAL_H */