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 bool zone_reclaimable(struct zone *zone);
  87
  88 /*
  89  * in mm/rmap.c:
  90  */
  91 extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
  92
  93 /*
  94  * in mm/page_alloc.c
  95  */
  96 extern void __free_pages_bootmem(struct page *page, unsigned int order);
  97 extern void prep_compound_page(struct page *page, unsigned long order);
  98 #ifdef CONFIG_MEMORY_FAILURE
  99 extern bool is_free_buddy_page(struct page *page);
 100 #endif
 101 extern int user_min_free_kbytes;
 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 migrate_pfn;      /* isolate_migratepages search base */
 122         bool sync;                      /* Synchronous migration */
 123         bool ignore_skip_hint;          /* Scan blocks even if marked skip */
 124         bool finished_update_free;      /* True when the zone cached pfns are
 125                                          * no longer being updated
 126                                          */
 127         bool finished_update_migrate;
 128
 129         int order;                      /* order a direct compactor needs */
 130         int migratetype;                /* MOVABLE, RECLAIMABLE etc */
 131         struct zone *zone;
 132         bool contended;                 /* True if a lock was contended */
 133 };
 134
 135 unsigned long
 136 isolate_freepages_range(struct compact_control *cc,
 137                         unsigned long start_pfn, unsigned long end_pfn);
 138 unsigned long
 139 isolate_migratepages_range(struct zone *zone, struct compact_control *cc,
 140         unsigned long low_pfn, unsigned long end_pfn, bool unevictable);
 141
 142 #endif
 143
 144 /*
 145  * This function returns the order of a free page in the buddy system. In
 146  * general, page_zone(page)->lock must be held by the caller to prevent the
 147  * page from being allocated in parallel and returning garbage as the order.
 148  * If a caller does not hold page_zone(page)->lock, it must guarantee that the
 149  * page cannot be allocated or merged in parallel.
 150  */
 151 static inline unsigned long page_order(struct page *page)
 152 {
 153         /* PageBuddy() must be checked by the caller */
 154         return page_private(page);
 155 }
 156
 157 /* mm/util.c */
 158 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
 159                 struct vm_area_struct *prev, struct rb_node *rb_parent);
 160
 161 #ifdef CONFIG_MMU
 162 extern long __mlock_vma_pages_range(struct vm_area_struct *vma,
 163                 unsigned long start, unsigned long end, int *nonblocking);
 164 extern void munlock_vma_pages_range(struct vm_area_struct *vma,
 165                         unsigned long start, unsigned long end);
 166 static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
 167 {
 168         munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
 169 }
 170
 171 /*
 172  * Called only in fault path, to determine if a new page is being
 173  * mapped into a LOCKED vma.  If it is, mark page as mlocked.
 174  */
 175 static inline int mlocked_vma_newpage(struct vm_area_struct *vma,
 176                                     struct page *page)
 177 {
 178         VM_BUG_ON_PAGE(PageLRU(page), page);
 179
 180         if (likely((vma->vm_flags & (VM_LOCKED | VM_SPECIAL)) != VM_LOCKED))
 181                 return 0;
 182
 183         if (!TestSetPageMlocked(page)) {
 184                 mod_zone_page_state(page_zone(page), NR_MLOCK,
 185                                     hpage_nr_pages(page));
 186                 count_vm_event(UNEVICTABLE_PGMLOCKED);
 187         }
 188         return 1;
 189 }
 190
 191 /*
 192  * must be called with vma's mmap_sem held for read or write, and page locked.
 193  */
 194 extern void mlock_vma_page(struct page *page);
 195 extern unsigned int munlock_vma_page(struct page *page);
 196
 197 /*
 198  * Clear the page's PageMlocked().  This can be useful in a situation where
 199  * we want to unconditionally remove a page from the pagecache -- e.g.,
 200  * on truncation or freeing.
 201  *
 202  * It is legal to call this function for any page, mlocked or not.
 203  * If called for a page that is still mapped by mlocked vmas, all we do
 204  * is revert to lazy LRU behaviour -- semantics are not broken.
 205  */
 206 extern void clear_page_mlock(struct page *page);
 207
 208 /*
 209  * mlock_migrate_page - called only from migrate_page_copy() to
 210  * migrate the Mlocked page flag; update statistics.
 211  */
 212 static inline void mlock_migrate_page(struct page *newpage, struct page *page)
 213 {
 214         if (TestClearPageMlocked(page)) {
 215                 unsigned long flags;
 216                 int nr_pages = hpage_nr_pages(page);
 217
 218                 local_irq_save(flags);
 219                 __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
 220                 SetPageMlocked(newpage);
 221                 __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
 222                 local_irq_restore(flags);
 223         }
 224 }
 225
 226 extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
 227
 228 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 229 extern unsigned long vma_address(struct page *page,
 230                                  struct vm_area_struct *vma);
 231 #endif
 232 #else /* !CONFIG_MMU */
 233 static inline int mlocked_vma_newpage(struct vm_area_struct *v, struct page *p)
 234 {
 235         return 0;
 236 }
 237 static inline void clear_page_mlock(struct page *page) { }
 238 static inline void mlock_vma_page(struct page *page) { }
 239 static inline void mlock_migrate_page(struct page *new, struct page *old) { }
 240
 241 #endif /* !CONFIG_MMU */
 242
 243 /*
 244  * Return the mem_map entry representing the 'offset' subpage within
 245  * the maximally aligned gigantic page 'base'.  Handle any discontiguity
 246  * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
 247  */
 248 static inline struct page *mem_map_offset(struct page *base, int offset)
 249 {
 250         if (unlikely(offset >= MAX_ORDER_NR_PAGES))
 251                 return pfn_to_page(page_to_pfn(base) + offset);
 252         return base + offset;
 253 }
 254
 255 /*
 256  * Iterator over all subpages within the maximally aligned gigantic
 257  * page 'base'.  Handle any discontiguity in the mem_map.
 258  */
 259 static inline struct page *mem_map_next(struct page *iter,
 260                                                 struct page *base, int offset)
 261 {
 262         if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
 263                 unsigned long pfn = page_to_pfn(base) + offset;
 264                 if (!pfn_valid(pfn))
 265                         return NULL;
 266                 return pfn_to_page(pfn);
 267         }
 268         return iter + 1;
 269 }
 270
 271 /*
 272  * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
 273  * so all functions starting at paging_init should be marked __init
 274  * in those cases. SPARSEMEM, however, allows for memory hotplug,
 275  * and alloc_bootmem_node is not used.
 276  */
 277 #ifdef CONFIG_SPARSEMEM
 278 #define __paginginit __meminit
 279 #else
 280 #define __paginginit __init
 281 #endif
 282
 283 /* Memory initialisation debug and verification */
 284 enum mminit_level {
 285         MMINIT_WARNING,
 286         MMINIT_VERIFY,
 287         MMINIT_TRACE
 288 };
 289
 290 #ifdef CONFIG_DEBUG_MEMORY_INIT
 291
 292 extern int mminit_loglevel;
 293
 294 #define mminit_dprintk(level, prefix, fmt, arg...) \
 295 do { \
 296         if (level < mminit_loglevel) { \
 297                 printk(level <= MMINIT_WARNING ? KERN_WARNING : KERN_DEBUG); \
 298                 printk(KERN_CONT "mminit::" prefix " " fmt, ##arg); \
 299         } \
 300 } while (0)
 301
 302 extern void mminit_verify_pageflags_layout(void);
 303 extern void mminit_verify_page_links(struct page *page,
 304                 enum zone_type zone, unsigned long nid, unsigned long pfn);
 305 extern void mminit_verify_zonelist(void);
 306
 307 #else
 308
 309 static inline void mminit_dprintk(enum mminit_level level,
 310                                 const char *prefix, const char *fmt, ...)
 311 {
 312 }
 313
 314 static inline void mminit_verify_pageflags_layout(void)
 315 {
 316 }
 317
 318 static inline void mminit_verify_page_links(struct page *page,
 319                 enum zone_type zone, unsigned long nid, unsigned long pfn)
 320 {
 321 }
 322
 323 static inline void mminit_verify_zonelist(void)
 324 {
 325 }
 326 #endif /* CONFIG_DEBUG_MEMORY_INIT */
 327
 328 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
 329 #if defined(CONFIG_SPARSEMEM)
 330 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 331                                 unsigned long *end_pfn);
 332 #else
 333 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 334                                 unsigned long *end_pfn)
 335 {
 336 }
 337 #endif /* CONFIG_SPARSEMEM */
 338
 339 #define ZONE_RECLAIM_NOSCAN     -2
 340 #define ZONE_RECLAIM_FULL       -1
 341 #define ZONE_RECLAIM_SOME       0
 342 #define ZONE_RECLAIM_SUCCESS    1
 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);
 358 unsigned long reclaim_clean_pages_from_list(struct zone *zone,
 359                                             struct list_head *page_list);
 360 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
 361 #define ALLOC_WMARK_MIN         WMARK_MIN
 362 #define ALLOC_WMARK_LOW         WMARK_LOW
 363 #define ALLOC_WMARK_HIGH        WMARK_HIGH
 364 #define ALLOC_NO_WATERMARKS     0x04 /* don't check watermarks at all */
 365
 366 /* Mask to get the watermark bits */
 367 #define ALLOC_WMARK_MASK        (ALLOC_NO_WATERMARKS-1)
 368
 369 #define ALLOC_HARDER            0x10 /* try to alloc harder */
 370 #define ALLOC_HIGH              0x20 /* __GFP_HIGH set */
 371 #define ALLOC_CPUSET            0x40 /* check for correct cpuset */
 372 #define ALLOC_CMA               0x80 /* allow allocations from CMA areas */
 373
 374 #endif  /* __MM_INTERNAL_H */