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 __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(atomic_read(&page->first_page->_count) <= 0);
  50         VM_BUG_ON(atomic_read(&page->_count) != 0);
  51         VM_BUG_ON(page_mapcount(page) < 0);
  52         if (get_page_head)
  53                 atomic_inc(&page->first_page->_count);
  54         atomic_inc(&page->_mapcount);
  55 }
  56
  57 /*
  58  * This is meant to be called as the FOLL_GET operation of
  59  * follow_page() and it must be called while holding the proper PT
  60  * lock while the pte (or pmd_trans_huge) is still mapping the page.
  61  */
  62 static inline void get_page_foll(struct page *page)
  63 {
  64         if (unlikely(PageTail(page)))
  65                 /*
  66                  * This is safe only because
  67                  * __split_huge_page_refcount() can't run under
  68                  * get_page_foll() because we hold the proper PT lock.
  69                  */
  70                 __get_page_tail_foll(page, true);
  71         else {
  72                 /*
  73                  * Getting a normal page or the head of a compound page
  74                  * requires to already have an elevated page->_count.
  75                  */
  76                 VM_BUG_ON(atomic_read(&page->_count) <= 0);
  77                 atomic_inc(&page->_count);
  78         }
  79 }
  80
  81 extern unsigned long highest_memmap_pfn;
  82
  83 /*
  84  * in mm/vmscan.c:
  85  */
  86 extern int isolate_lru_page(struct page *page);
  87 extern void putback_lru_page(struct page *page);
  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
 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  * function for dealing with page's order in buddy system.
 146  * zone->lock is already acquired when we use these.
 147  * So, we don't need atomic page->flags operations here.
 148  */
 149 static inline unsigned long page_order(struct page *page)
 150 {
 151         /* PageBuddy() must be checked by the caller */
 152         return page_private(page);
 153 }
 154
 155 /* mm/util.c */
 156 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
 157                 struct vm_area_struct *prev, struct rb_node *rb_parent);
 158
 159 #ifdef CONFIG_MMU
 160 extern long __mlock_vma_pages_range(struct vm_area_struct *vma,
 161                 unsigned long start, unsigned long end, int *nonblocking);
 162 extern void munlock_vma_pages_range(struct vm_area_struct *vma,
 163                         unsigned long start, unsigned long end);
 164 static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
 165 {
 166         munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
 167 }
 168
 169 /*
 170  * Called only in fault path, to determine if a new page is being
 171  * mapped into a LOCKED vma.  If it is, mark page as mlocked.
 172  */
 173 static inline int mlocked_vma_newpage(struct vm_area_struct *vma,
 174                                     struct page *page)
 175 {
 176         VM_BUG_ON(PageLRU(page));
 177
 178         if (likely((vma->vm_flags & (VM_LOCKED | VM_SPECIAL)) != VM_LOCKED))
 179                 return 0;
 180
 181         if (!TestSetPageMlocked(page)) {
 182                 mod_zone_page_state(page_zone(page), NR_MLOCK,
 183                                     hpage_nr_pages(page));
 184                 count_vm_event(UNEVICTABLE_PGMLOCKED);
 185         }
 186         return 1;
 187 }
 188
 189 /*
 190  * must be called with vma's mmap_sem held for read or write, and page locked.
 191  */
 192 extern void mlock_vma_page(struct page *page);
 193 extern unsigned int munlock_vma_page(struct page *page);
 194
 195 /*
 196  * Clear the page's PageMlocked().  This can be useful in a situation where
 197  * we want to unconditionally remove a page from the pagecache -- e.g.,
 198  * on truncation or freeing.
 199  *
 200  * It is legal to call this function for any page, mlocked or not.
 201  * If called for a page that is still mapped by mlocked vmas, all we do
 202  * is revert to lazy LRU behaviour -- semantics are not broken.
 203  */
 204 extern void clear_page_mlock(struct page *page);
 205
 206 /*
 207  * mlock_migrate_page - called only from migrate_page_copy() to
 208  * migrate the Mlocked page flag; update statistics.
 209  */
 210 static inline void mlock_migrate_page(struct page *newpage, struct page *page)
 211 {
 212         if (TestClearPageMlocked(page)) {
 213                 unsigned long flags;
 214                 int nr_pages = hpage_nr_pages(page);
 215
 216                 local_irq_save(flags);
 217                 __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
 218                 SetPageMlocked(newpage);
 219                 __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
 220                 local_irq_restore(flags);
 221         }
 222 }
 223
 224 extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
 225
 226 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 227 extern unsigned long vma_address(struct page *page,
 228                                  struct vm_area_struct *vma);
 229 #endif
 230 #else /* !CONFIG_MMU */
 231 static inline int mlocked_vma_newpage(struct vm_area_struct *v, struct page *p)
 232 {
 233         return 0;
 234 }
 235 static inline void clear_page_mlock(struct page *page) { }
 236 static inline void mlock_vma_page(struct page *page) { }
 237 static inline void mlock_migrate_page(struct page *new, struct page *old) { }
 238
 239 #endif /* !CONFIG_MMU */
 240
 241 /*
 242  * Return the mem_map entry representing the 'offset' subpage within
 243  * the maximally aligned gigantic page 'base'.  Handle any discontiguity
 244  * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
 245  */
 246 static inline struct page *mem_map_offset(struct page *base, int offset)
 247 {
 248         if (unlikely(offset >= MAX_ORDER_NR_PAGES))
 249                 return pfn_to_page(page_to_pfn(base) + offset);
 250         return base + offset;
 251 }
 252
 253 /*
 254  * Iterator over all subpages within the maximally aligned gigantic
 255  * page 'base'.  Handle any discontiguity in the mem_map.
 256  */
 257 static inline struct page *mem_map_next(struct page *iter,
 258                                                 struct page *base, int offset)
 259 {
 260         if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
 261                 unsigned long pfn = page_to_pfn(base) + offset;
 262                 if (!pfn_valid(pfn))
 263                         return NULL;
 264                 return pfn_to_page(pfn);
 265         }
 266         return iter + 1;
 267 }
 268
 269 /*
 270  * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
 271  * so all functions starting at paging_init should be marked __init
 272  * in those cases. SPARSEMEM, however, allows for memory hotplug,
 273  * and alloc_bootmem_node is not used.
 274  */
 275 #ifdef CONFIG_SPARSEMEM
 276 #define __paginginit __meminit
 277 #else
 278 #define __paginginit __init
 279 #endif
 280
 281 /* Memory initialisation debug and verification */
 282 enum mminit_level {
 283         MMINIT_WARNING,
 284         MMINIT_VERIFY,
 285         MMINIT_TRACE
 286 };
 287
 288 #ifdef CONFIG_DEBUG_MEMORY_INIT
 289
 290 extern int mminit_loglevel;
 291
 292 #define mminit_dprintk(level, prefix, fmt, arg...) \
 293 do { \
 294         if (level < mminit_loglevel) { \
 295                 printk(level <= MMINIT_WARNING ? KERN_WARNING : KERN_DEBUG); \
 296                 printk(KERN_CONT "mminit::" prefix " " fmt, ##arg); \
 297         } \
 298 } while (0)
 299
 300 extern void mminit_verify_pageflags_layout(void);
 301 extern void mminit_verify_page_links(struct page *page,
 302                 enum zone_type zone, unsigned long nid, unsigned long pfn);
 303 extern void mminit_verify_zonelist(void);
 304
 305 #else
 306
 307 static inline void mminit_dprintk(enum mminit_level level,
 308                                 const char *prefix, const char *fmt, ...)
 309 {
 310 }
 311
 312 static inline void mminit_verify_pageflags_layout(void)
 313 {
 314 }
 315
 316 static inline void mminit_verify_page_links(struct page *page,
 317                 enum zone_type zone, unsigned long nid, unsigned long pfn)
 318 {
 319 }
 320
 321 static inline void mminit_verify_zonelist(void)
 322 {
 323 }
 324 #endif /* CONFIG_DEBUG_MEMORY_INIT */
 325
 326 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
 327 #if defined(CONFIG_SPARSEMEM)
 328 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 329                                 unsigned long *end_pfn);
 330 #else
 331 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 332                                 unsigned long *end_pfn)
 333 {
 334 }
 335 #endif /* CONFIG_SPARSEMEM */
 336
 337 #define ZONE_RECLAIM_NOSCAN     -2
 338 #define ZONE_RECLAIM_FULL       -1
 339 #define ZONE_RECLAIM_SOME       0
 340 #define ZONE_RECLAIM_SUCCESS    1
 341
 342 extern int hwpoison_filter(struct page *p);
 343
 344 extern u32 hwpoison_filter_dev_major;
 345 extern u32 hwpoison_filter_dev_minor;
 346 extern u64 hwpoison_filter_flags_mask;
 347 extern u64 hwpoison_filter_flags_value;
 348 extern u64 hwpoison_filter_memcg;
 349 extern u32 hwpoison_filter_enable;
 350
 351 extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
 352         unsigned long, unsigned long,
 353         unsigned long, unsigned long);
 354
 355 extern void set_pageblock_order(void);
 356 unsigned long reclaim_clean_pages_from_list(struct zone *zone,
 357                                             struct list_head *page_list);
 358 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
 359 #define ALLOC_WMARK_MIN         WMARK_MIN
 360 #define ALLOC_WMARK_LOW         WMARK_LOW
 361 #define ALLOC_WMARK_HIGH        WMARK_HIGH
 362 #define ALLOC_NO_WATERMARKS     0x04 /* don't check watermarks at all */
 363
 364 /* Mask to get the watermark bits */
 365 #define ALLOC_WMARK_MASK        (ALLOC_NO_WATERMARKS-1)
 366
 367 #define ALLOC_HARDER            0x10 /* try to alloc harder */
 368 #define ALLOC_HIGH              0x20 /* __GFP_HIGH set */
 369 #define ALLOC_CPUSET            0x40 /* check for correct cpuset */
 370 #define ALLOC_CMA               0x80 /* allow allocations from CMA areas */
 371
 372 #endif  /* __MM_INTERNAL_H */