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
 112 /*
 113  * Locate the struct page for both the matching buddy in our
 114  * pair (buddy1) and the combined O(n+1) page they form (page).
 115  *
 116  * 1) Any buddy B1 will have an order O twin B2 which satisfies
 117  * the following equation:
 118  *     B2 = B1 ^ (1 << O)
 119  * For example, if the starting buddy (buddy2) is #8 its order
 120  * 1 buddy is #10:
 121  *     B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
 122  *
 123  * 2) Any buddy B will have an order O+1 parent P which
 124  * satisfies the following equation:
 125  *     P = B & ~(1 << O)
 126  *
 127  * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
 128  */
 129 static inline unsigned long
 130 __find_buddy_index(unsigned long page_idx, unsigned int order)
 131 {
 132         return page_idx ^ (1 << order);
 133 }
 134
 135 extern int __isolate_free_page(struct page *page, unsigned int order);
 136 extern void __free_pages_bootmem(struct page *page, unsigned int order);
 137 extern void prep_compound_page(struct page *page, unsigned long order);
 138 #ifdef CONFIG_MEMORY_FAILURE
 139 extern bool is_free_buddy_page(struct page *page);
 140 #endif
 141 extern int user_min_free_kbytes;
 142
 143 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
 144
 145 /*
 146  * in mm/compaction.c
 147  */
 148 /*
 149  * compact_control is used to track pages being migrated and the free pages
 150  * they are being migrated to during memory compaction. The free_pfn starts
 151  * at the end of a zone and migrate_pfn begins at the start. Movable pages
 152  * are moved to the end of a zone during a compaction run and the run
 153  * completes when free_pfn <= migrate_pfn
 154  */
 155 struct compact_control {
 156         struct list_head freepages;     /* List of free pages to migrate to */
 157         struct list_head migratepages;  /* List of pages being migrated */
 158         unsigned long nr_freepages;     /* Number of isolated free pages */
 159         unsigned long nr_migratepages;  /* Number of pages to migrate */
 160         unsigned long free_pfn;         /* isolate_freepages search base */
 161         unsigned long migrate_pfn;      /* isolate_migratepages search base */
 162         enum migrate_mode mode;         /* Async or sync migration mode */
 163         bool ignore_skip_hint;          /* Scan blocks even if marked skip */
 164         int order;                      /* order a direct compactor needs */
 165         const gfp_t gfp_mask;           /* gfp mask of a direct compactor */
 166         const int alloc_flags;          /* alloc flags of a direct compactor */
 167         const int classzone_idx;        /* zone index of a direct compactor */
 168         struct zone *zone;
 169         int contended;                  /* Signal need_sched() or lock
 170                                          * contention detected during
 171                                          * compaction
 172                                          */
 173 };
 174
 175 unsigned long
 176 isolate_freepages_range(struct compact_control *cc,
 177                         unsigned long start_pfn, unsigned long end_pfn);
 178 unsigned long
 179 isolate_migratepages_range(struct compact_control *cc,
 180                            unsigned long low_pfn, unsigned long end_pfn);
 181
 182 #endif
 183
 184 /*
 185  * This function returns the order of a free page in the buddy system. In
 186  * general, page_zone(page)->lock must be held by the caller to prevent the
 187  * page from being allocated in parallel and returning garbage as the order.
 188  * If a caller does not hold page_zone(page)->lock, it must guarantee that the
 189  * page cannot be allocated or merged in parallel. Alternatively, it must
 190  * handle invalid values gracefully, and use page_order_unsafe() below.
 191  */
 192 static inline unsigned long page_order(struct page *page)
 193 {
 194         /* PageBuddy() must be checked by the caller */
 195         return page_private(page);
 196 }
 197
 198 /*
 199  * Like page_order(), but for callers who cannot afford to hold the zone lock.
 200  * PageBuddy() should be checked first by the caller to minimize race window,
 201  * and invalid values must be handled gracefully.
 202  *
 203  * ACCESS_ONCE is used so that if the caller assigns the result into a local
 204  * variable and e.g. tests it for valid range before using, the compiler cannot
 205  * decide to remove the variable and inline the page_private(page) multiple
 206  * times, potentially observing different values in the tests and the actual
 207  * use of the result.
 208  */
 209 #define page_order_unsafe(page)         ACCESS_ONCE(page_private(page))
 210
 211 static inline bool is_cow_mapping(vm_flags_t flags)
 212 {
 213         return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
 214 }
 215
 216 /* mm/util.c */
 217 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
 218                 struct vm_area_struct *prev, struct rb_node *rb_parent);
 219
 220 #ifdef CONFIG_MMU
 221 extern long __mlock_vma_pages_range(struct vm_area_struct *vma,
 222                 unsigned long start, unsigned long end, int *nonblocking);
 223 extern void munlock_vma_pages_range(struct vm_area_struct *vma,
 224                         unsigned long start, unsigned long end);
 225 static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
 226 {
 227         munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
 228 }
 229
 230 /*
 231  * must be called with vma's mmap_sem held for read or write, and page locked.
 232  */
 233 extern void mlock_vma_page(struct page *page);
 234 extern unsigned int munlock_vma_page(struct page *page);
 235
 236 /*
 237  * Clear the page's PageMlocked().  This can be useful in a situation where
 238  * we want to unconditionally remove a page from the pagecache -- e.g.,
 239  * on truncation or freeing.
 240  *
 241  * It is legal to call this function for any page, mlocked or not.
 242  * If called for a page that is still mapped by mlocked vmas, all we do
 243  * is revert to lazy LRU behaviour -- semantics are not broken.
 244  */
 245 extern void clear_page_mlock(struct page *page);
 246
 247 /*
 248  * mlock_migrate_page - called only from migrate_page_copy() to
 249  * migrate the Mlocked page flag; update statistics.
 250  */
 251 static inline void mlock_migrate_page(struct page *newpage, struct page *page)
 252 {
 253         if (TestClearPageMlocked(page)) {
 254                 unsigned long flags;
 255                 int nr_pages = hpage_nr_pages(page);
 256
 257                 local_irq_save(flags);
 258                 __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
 259                 SetPageMlocked(newpage);
 260                 __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
 261                 local_irq_restore(flags);
 262         }
 263 }
 264
 265 extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
 266
 267 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 268 extern unsigned long vma_address(struct page *page,
 269                                  struct vm_area_struct *vma);
 270 #endif
 271 #else /* !CONFIG_MMU */
 272 static inline void clear_page_mlock(struct page *page) { }
 273 static inline void mlock_vma_page(struct page *page) { }
 274 static inline void mlock_migrate_page(struct page *new, struct page *old) { }
 275
 276 #endif /* !CONFIG_MMU */
 277
 278 /*
 279  * Return the mem_map entry representing the 'offset' subpage within
 280  * the maximally aligned gigantic page 'base'.  Handle any discontiguity
 281  * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
 282  */
 283 static inline struct page *mem_map_offset(struct page *base, int offset)
 284 {
 285         if (unlikely(offset >= MAX_ORDER_NR_PAGES))
 286                 return nth_page(base, offset);
 287         return base + offset;
 288 }
 289
 290 /*
 291  * Iterator over all subpages within the maximally aligned gigantic
 292  * page 'base'.  Handle any discontiguity in the mem_map.
 293  */
 294 static inline struct page *mem_map_next(struct page *iter,
 295                                                 struct page *base, int offset)
 296 {
 297         if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
 298                 unsigned long pfn = page_to_pfn(base) + offset;
 299                 if (!pfn_valid(pfn))
 300                         return NULL;
 301                 return pfn_to_page(pfn);
 302         }
 303         return iter + 1;
 304 }
 305
 306 /*
 307  * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
 308  * so all functions starting at paging_init should be marked __init
 309  * in those cases. SPARSEMEM, however, allows for memory hotplug,
 310  * and alloc_bootmem_node is not used.
 311  */
 312 #ifdef CONFIG_SPARSEMEM
 313 #define __paginginit __meminit
 314 #else
 315 #define __paginginit __init
 316 #endif
 317
 318 /* Memory initialisation debug and verification */
 319 enum mminit_level {
 320         MMINIT_WARNING,
 321         MMINIT_VERIFY,
 322         MMINIT_TRACE
 323 };
 324
 325 #ifdef CONFIG_DEBUG_MEMORY_INIT
 326
 327 extern int mminit_loglevel;
 328
 329 #define mminit_dprintk(level, prefix, fmt, arg...) \
 330 do { \
 331         if (level < mminit_loglevel) { \
 332                 printk(level <= MMINIT_WARNING ? KERN_WARNING : KERN_DEBUG); \
 333                 printk(KERN_CONT "mminit::" prefix " " fmt, ##arg); \
 334         } \
 335 } while (0)
 336
 337 extern void mminit_verify_pageflags_layout(void);
 338 extern void mminit_verify_page_links(struct page *page,
 339                 enum zone_type zone, unsigned long nid, unsigned long pfn);
 340 extern void mminit_verify_zonelist(void);
 341
 342 #else
 343
 344 static inline void mminit_dprintk(enum mminit_level level,
 345                                 const char *prefix, const char *fmt, ...)
 346 {
 347 }
 348
 349 static inline void mminit_verify_pageflags_layout(void)
 350 {
 351 }
 352
 353 static inline void mminit_verify_page_links(struct page *page,
 354                 enum zone_type zone, unsigned long nid, unsigned long pfn)
 355 {
 356 }
 357
 358 static inline void mminit_verify_zonelist(void)
 359 {
 360 }
 361 #endif /* CONFIG_DEBUG_MEMORY_INIT */
 362
 363 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
 364 #if defined(CONFIG_SPARSEMEM)
 365 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 366                                 unsigned long *end_pfn);
 367 #else
 368 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 369                                 unsigned long *end_pfn)
 370 {
 371 }
 372 #endif /* CONFIG_SPARSEMEM */
 373
 374 #define ZONE_RECLAIM_NOSCAN     -2
 375 #define ZONE_RECLAIM_FULL       -1
 376 #define ZONE_RECLAIM_SOME       0
 377 #define ZONE_RECLAIM_SUCCESS    1
 378
 379 extern int hwpoison_filter(struct page *p);
 380
 381 extern u32 hwpoison_filter_dev_major;
 382 extern u32 hwpoison_filter_dev_minor;
 383 extern u64 hwpoison_filter_flags_mask;
 384 extern u64 hwpoison_filter_flags_value;
 385 extern u64 hwpoison_filter_memcg;
 386 extern u32 hwpoison_filter_enable;
 387
 388 extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
 389         unsigned long, unsigned long,
 390         unsigned long, unsigned long);
 391
 392 extern void set_pageblock_order(void);
 393 unsigned long reclaim_clean_pages_from_list(struct zone *zone,
 394                                             struct list_head *page_list);
 395 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
 396 #define ALLOC_WMARK_MIN         WMARK_MIN
 397 #define ALLOC_WMARK_LOW         WMARK_LOW
 398 #define ALLOC_WMARK_HIGH        WMARK_HIGH
 399 #define ALLOC_NO_WATERMARKS     0x04 /* don't check watermarks at all */
 400
 401 /* Mask to get the watermark bits */
 402 #define ALLOC_WMARK_MASK        (ALLOC_NO_WATERMARKS-1)
 403
 404 #define ALLOC_HARDER            0x10 /* try to alloc harder */
 405 #define ALLOC_HIGH              0x20 /* __GFP_HIGH set */
 406 #define ALLOC_CPUSET            0x40 /* check for correct cpuset */
 407 #define ALLOC_CMA               0x80 /* allow allocations from CMA areas */
 408 #define ALLOC_FAIR              0x100 /* fair zone allocation */
 409
 410 #endif  /* __MM_INTERNAL_H */