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         const gfp_t gfp_mask;           /* gfp mask of a direct compactor */
 146         struct zone *zone;
 147         int contended;                  /* Signal need_sched() or lock
 148                                          * contention detected during
 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 compact_control *cc,
 158                            unsigned long low_pfn, unsigned long end_pfn);
 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. Alternatively, it must
 168  * handle invalid values gracefully, and use page_order_unsafe() below.
 169  */
 170 static inline unsigned long page_order(struct page *page)
 171 {
 172         /* PageBuddy() must be checked by the caller */
 173         return page_private(page);
 174 }
 175
 176 /*
 177  * Like page_order(), but for callers who cannot afford to hold the zone lock.
 178  * PageBuddy() should be checked first by the caller to minimize race window,
 179  * and invalid values must be handled gracefully.
 180  *
 181  * ACCESS_ONCE is used so that if the caller assigns the result into a local
 182  * variable and e.g. tests it for valid range before using, the compiler cannot
 183  * decide to remove the variable and inline the page_private(page) multiple
 184  * times, potentially observing different values in the tests and the actual
 185  * use of the result.
 186  */
 187 #define page_order_unsafe(page)         ACCESS_ONCE(page_private(page))
 188
 189 static inline bool is_cow_mapping(vm_flags_t flags)
 190 {
 191         return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
 192 }
 193
 194 /* mm/util.c */
 195 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
 196                 struct vm_area_struct *prev, struct rb_node *rb_parent);
 197
 198 #ifdef CONFIG_MMU
 199 extern long __mlock_vma_pages_range(struct vm_area_struct *vma,
 200                 unsigned long start, unsigned long end, int *nonblocking);
 201 extern void munlock_vma_pages_range(struct vm_area_struct *vma,
 202                         unsigned long start, unsigned long end);
 203 static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
 204 {
 205         munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
 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 void clear_page_mlock(struct page *page) { }
 251 static inline void mlock_vma_page(struct page *page) { }
 252 static inline void mlock_migrate_page(struct page *new, struct page *old) { }
 253
 254 #endif /* !CONFIG_MMU */
 255
 256 /*
 257  * Return the mem_map entry representing the 'offset' subpage within
 258  * the maximally aligned gigantic page 'base'.  Handle any discontiguity
 259  * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
 260  */
 261 static inline struct page *mem_map_offset(struct page *base, int offset)
 262 {
 263         if (unlikely(offset >= MAX_ORDER_NR_PAGES))
 264                 return nth_page(base, offset);
 265         return base + offset;
 266 }
 267
 268 /*
 269  * Iterator over all subpages within the maximally aligned gigantic
 270  * page 'base'.  Handle any discontiguity in the mem_map.
 271  */
 272 static inline struct page *mem_map_next(struct page *iter,
 273                                                 struct page *base, int offset)
 274 {
 275         if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
 276                 unsigned long pfn = page_to_pfn(base) + offset;
 277                 if (!pfn_valid(pfn))
 278                         return NULL;
 279                 return pfn_to_page(pfn);
 280         }
 281         return iter + 1;
 282 }
 283
 284 /*
 285  * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
 286  * so all functions starting at paging_init should be marked __init
 287  * in those cases. SPARSEMEM, however, allows for memory hotplug,
 288  * and alloc_bootmem_node is not used.
 289  */
 290 #ifdef CONFIG_SPARSEMEM
 291 #define __paginginit __meminit
 292 #else
 293 #define __paginginit __init
 294 #endif
 295
 296 /* Memory initialisation debug and verification */
 297 enum mminit_level {
 298         MMINIT_WARNING,
 299         MMINIT_VERIFY,
 300         MMINIT_TRACE
 301 };
 302
 303 #ifdef CONFIG_DEBUG_MEMORY_INIT
 304
 305 extern int mminit_loglevel;
 306
 307 #define mminit_dprintk(level, prefix, fmt, arg...) \
 308 do { \
 309         if (level < mminit_loglevel) { \
 310                 printk(level <= MMINIT_WARNING ? KERN_WARNING : KERN_DEBUG); \
 311                 printk(KERN_CONT "mminit::" prefix " " fmt, ##arg); \
 312         } \
 313 } while (0)
 314
 315 extern void mminit_verify_pageflags_layout(void);
 316 extern void mminit_verify_page_links(struct page *page,
 317                 enum zone_type zone, unsigned long nid, unsigned long pfn);
 318 extern void mminit_verify_zonelist(void);
 319
 320 #else
 321
 322 static inline void mminit_dprintk(enum mminit_level level,
 323                                 const char *prefix, const char *fmt, ...)
 324 {
 325 }
 326
 327 static inline void mminit_verify_pageflags_layout(void)
 328 {
 329 }
 330
 331 static inline void mminit_verify_page_links(struct page *page,
 332                 enum zone_type zone, unsigned long nid, unsigned long pfn)
 333 {
 334 }
 335
 336 static inline void mminit_verify_zonelist(void)
 337 {
 338 }
 339 #endif /* CONFIG_DEBUG_MEMORY_INIT */
 340
 341 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
 342 #if defined(CONFIG_SPARSEMEM)
 343 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 344                                 unsigned long *end_pfn);
 345 #else
 346 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 347                                 unsigned long *end_pfn)
 348 {
 349 }
 350 #endif /* CONFIG_SPARSEMEM */
 351
 352 #define ZONE_RECLAIM_NOSCAN     -2
 353 #define ZONE_RECLAIM_FULL       -1
 354 #define ZONE_RECLAIM_SOME       0
 355 #define ZONE_RECLAIM_SUCCESS    1
 356
 357 extern int hwpoison_filter(struct page *p);
 358
 359 extern u32 hwpoison_filter_dev_major;
 360 extern u32 hwpoison_filter_dev_minor;
 361 extern u64 hwpoison_filter_flags_mask;
 362 extern u64 hwpoison_filter_flags_value;
 363 extern u64 hwpoison_filter_memcg;
 364 extern u32 hwpoison_filter_enable;
 365
 366 extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
 367         unsigned long, unsigned long,
 368         unsigned long, unsigned long);
 369
 370 extern void set_pageblock_order(void);
 371 unsigned long reclaim_clean_pages_from_list(struct zone *zone,
 372                                             struct list_head *page_list);
 373 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
 374 #define ALLOC_WMARK_MIN         WMARK_MIN
 375 #define ALLOC_WMARK_LOW         WMARK_LOW
 376 #define ALLOC_WMARK_HIGH        WMARK_HIGH
 377 #define ALLOC_NO_WATERMARKS     0x04 /* don't check watermarks at all */
 378
 379 /* Mask to get the watermark bits */
 380 #define ALLOC_WMARK_MASK        (ALLOC_NO_WATERMARKS-1)
 381
 382 #define ALLOC_HARDER            0x10 /* try to alloc harder */
 383 #define ALLOC_HIGH              0x20 /* __GFP_HIGH set */
 384 #define ALLOC_CPUSET            0x40 /* check for correct cpuset */
 385 #define ALLOC_CMA               0x80 /* allow allocations from CMA areas */
 386 #define ALLOC_FAIR              0x100 /* fair zone allocation */
 387
 388 #endif  /* __MM_INTERNAL_H */