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