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 __put_page(struct page *page)
  36 {
  37         atomic_dec(&page->_count);
  38 }
  39
  40 static inline void __get_page_tail_foll(struct page *page,
  41                                         bool get_page_head)
  42 {
  43         /*
  44          * If we're getting a tail page, the elevated page->_count is
  45          * required only in the head page and we will elevate the head
  46          * page->_count and tail page->_mapcount.
  47          *
  48          * We elevate page_tail->_mapcount for tail pages to force
  49          * page_tail->_count to be zero at all times to avoid getting
  50          * false positives from get_page_unless_zero() with
  51          * speculative page access (like in
  52          * page_cache_get_speculative()) on tail pages.
  53          */
  54         VM_BUG_ON(atomic_read(&page->first_page->_count) <= 0);
  55         VM_BUG_ON(atomic_read(&page->_count) != 0);
  56         VM_BUG_ON(page_mapcount(page) < 0);
  57         if (get_page_head)
  58                 atomic_inc(&page->first_page->_count);
  59         atomic_inc(&page->_mapcount);
  60 }
  61
  62 /*
  63  * This is meant to be called as the FOLL_GET operation of
  64  * follow_page() and it must be called while holding the proper PT
  65  * lock while the pte (or pmd_trans_huge) is still mapping the page.
  66  */
  67 static inline void get_page_foll(struct page *page)
  68 {
  69         if (unlikely(PageTail(page)))
  70                 /*
  71                  * This is safe only because
  72                  * __split_huge_page_refcount() can't run under
  73                  * get_page_foll() because we hold the proper PT lock.
  74                  */
  75                 __get_page_tail_foll(page, true);
  76         else {
  77                 /*
  78                  * Getting a normal page or the head of a compound page
  79                  * requires to already have an elevated page->_count.
  80                  */
  81                 VM_BUG_ON(atomic_read(&page->_count) <= 0);
  82                 atomic_inc(&page->_count);
  83         }
  84 }
  85
  86 extern unsigned long highest_memmap_pfn;
  87
  88 /*
  89  * in mm/vmscan.c:
  90  */
  91 extern int isolate_lru_page(struct page *page);
  92 extern void putback_lru_page(struct page *page);
  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
 125         int order;                      /* order a direct compactor needs */
 126         int migratetype;                /* MOVABLE, RECLAIMABLE etc */
 127         struct zone *zone;
 128         bool contended;                 /* True if a lock was contended */
 129         struct page **page;             /* Page captured of requested size */
 130 };
 131
 132 unsigned long
 133 isolate_freepages_range(struct compact_control *cc,
 134                         unsigned long start_pfn, unsigned long end_pfn);
 135 unsigned long
 136 isolate_migratepages_range(struct zone *zone, struct compact_control *cc,
 137                            unsigned long low_pfn, unsigned long end_pfn);
 138
 139 #endif
 140
 141 /*
 142  * function for dealing with page's order in buddy system.
 143  * zone->lock is already acquired when we use these.
 144  * So, we don't need atomic page->flags operations here.
 145  */
 146 static inline unsigned long page_order(struct page *page)
 147 {
 148         /* PageBuddy() must be checked by the caller */
 149         return page_private(page);
 150 }
 151
 152 /* mm/util.c */
 153 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
 154                 struct vm_area_struct *prev, struct rb_node *rb_parent);
 155
 156 #ifdef CONFIG_MMU
 157 extern long mlock_vma_pages_range(struct vm_area_struct *vma,
 158                         unsigned long start, unsigned long end);
 159 extern void munlock_vma_pages_range(struct vm_area_struct *vma,
 160                         unsigned long start, unsigned long end);
 161 static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
 162 {
 163         munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
 164 }
 165
 166 /*
 167  * Called only in fault path via page_evictable() for a new page
 168  * to determine if it's being mapped into a LOCKED vma.
 169  * If so, mark page as mlocked.
 170  */
 171 static inline int mlocked_vma_newpage(struct vm_area_struct *vma,
 172                                     struct page *page)
 173 {
 174         VM_BUG_ON(PageLRU(page));
 175
 176         if (likely((vma->vm_flags & (VM_LOCKED | VM_SPECIAL)) != VM_LOCKED))
 177                 return 0;
 178
 179         if (!TestSetPageMlocked(page)) {
 180                 inc_zone_page_state(page, NR_MLOCK);
 181                 count_vm_event(UNEVICTABLE_PGMLOCKED);
 182         }
 183         return 1;
 184 }
 185
 186 /*
 187  * must be called with vma's mmap_sem held for read or write, and page locked.
 188  */
 189 extern void mlock_vma_page(struct page *page);
 190 extern void munlock_vma_page(struct page *page);
 191
 192 /*
 193  * Clear the page's PageMlocked().  This can be useful in a situation where
 194  * we want to unconditionally remove a page from the pagecache -- e.g.,
 195  * on truncation or freeing.
 196  *
 197  * It is legal to call this function for any page, mlocked or not.
 198  * If called for a page that is still mapped by mlocked vmas, all we do
 199  * is revert to lazy LRU behaviour -- semantics are not broken.
 200  */
 201 extern void __clear_page_mlock(struct page *page);
 202 static inline void clear_page_mlock(struct page *page)
 203 {
 204         if (unlikely(TestClearPageMlocked(page)))
 205                 __clear_page_mlock(page);
 206 }
 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
 217                 local_irq_save(flags);
 218                 __dec_zone_page_state(page, NR_MLOCK);
 219                 SetPageMlocked(newpage);
 220                 __inc_zone_page_state(newpage, NR_MLOCK);
 221                 local_irq_restore(flags);
 222         }
 223 }
 224
 225 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
 226 extern unsigned long vma_address(struct page *page,
 227                                  struct vm_area_struct *vma);
 228 #endif
 229 #else /* !CONFIG_MMU */
 230 static inline int mlocked_vma_newpage(struct vm_area_struct *v, struct page *p)
 231 {
 232         return 0;
 233 }
 234 static inline void clear_page_mlock(struct page *page) { }
 235 static inline void mlock_vma_page(struct page *page) { }
 236 static inline void mlock_migrate_page(struct page *new, struct page *old) { }
 237
 238 #endif /* !CONFIG_MMU */
 239
 240 /*
 241  * Return the mem_map entry representing the 'offset' subpage within
 242  * the maximally aligned gigantic page 'base'.  Handle any discontiguity
 243  * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
 244  */
 245 static inline struct page *mem_map_offset(struct page *base, int offset)
 246 {
 247         if (unlikely(offset >= MAX_ORDER_NR_PAGES))
 248                 return pfn_to_page(page_to_pfn(base) + offset);
 249         return base + offset;
 250 }
 251
 252 /*
 253  * Iterator over all subpages within the maximally aligned gigantic
 254  * page 'base'.  Handle any discontiguity in the mem_map.
 255  */
 256 static inline struct page *mem_map_next(struct page *iter,
 257                                                 struct page *base, int offset)
 258 {
 259         if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
 260                 unsigned long pfn = page_to_pfn(base) + offset;
 261                 if (!pfn_valid(pfn))
 262                         return NULL;
 263                 return pfn_to_page(pfn);
 264         }
 265         return iter + 1;
 266 }
 267
 268 /*
 269  * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
 270  * so all functions starting at paging_init should be marked __init
 271  * in those cases. SPARSEMEM, however, allows for memory hotplug,
 272  * and alloc_bootmem_node is not used.
 273  */
 274 #ifdef CONFIG_SPARSEMEM
 275 #define __paginginit __meminit
 276 #else
 277 #define __paginginit __init
 278 #endif
 279
 280 /* Memory initialisation debug and verification */
 281 enum mminit_level {
 282         MMINIT_WARNING,
 283         MMINIT_VERIFY,
 284         MMINIT_TRACE
 285 };
 286
 287 #ifdef CONFIG_DEBUG_MEMORY_INIT
 288
 289 extern int mminit_loglevel;
 290
 291 #define mminit_dprintk(level, prefix, fmt, arg...) \
 292 do { \
 293         if (level < mminit_loglevel) { \
 294                 printk(level <= MMINIT_WARNING ? KERN_WARNING : KERN_DEBUG); \
 295                 printk(KERN_CONT "mminit::" prefix " " fmt, ##arg); \
 296         } \
 297 } while (0)
 298
 299 extern void mminit_verify_pageflags_layout(void);
 300 extern void mminit_verify_page_links(struct page *page,
 301                 enum zone_type zone, unsigned long nid, unsigned long pfn);
 302 extern void mminit_verify_zonelist(void);
 303
 304 #else
 305
 306 static inline void mminit_dprintk(enum mminit_level level,
 307                                 const char *prefix, const char *fmt, ...)
 308 {
 309 }
 310
 311 static inline void mminit_verify_pageflags_layout(void)
 312 {
 313 }
 314
 315 static inline void mminit_verify_page_links(struct page *page,
 316                 enum zone_type zone, unsigned long nid, unsigned long pfn)
 317 {
 318 }
 319
 320 static inline void mminit_verify_zonelist(void)
 321 {
 322 }
 323 #endif /* CONFIG_DEBUG_MEMORY_INIT */
 324
 325 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
 326 #if defined(CONFIG_SPARSEMEM)
 327 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 328                                 unsigned long *end_pfn);
 329 #else
 330 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 331                                 unsigned long *end_pfn)
 332 {
 333 }
 334 #endif /* CONFIG_SPARSEMEM */
 335
 336 #define ZONE_RECLAIM_NOSCAN     -2
 337 #define ZONE_RECLAIM_FULL       -1
 338 #define ZONE_RECLAIM_SOME       0
 339 #define ZONE_RECLAIM_SUCCESS    1
 340
 341 extern int hwpoison_filter(struct page *p);
 342
 343 extern u32 hwpoison_filter_dev_major;
 344 extern u32 hwpoison_filter_dev_minor;
 345 extern u64 hwpoison_filter_flags_mask;
 346 extern u64 hwpoison_filter_flags_value;
 347 extern u64 hwpoison_filter_memcg;
 348 extern u32 hwpoison_filter_enable;
 349
 350 extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
 351         unsigned long, unsigned long,
 352         unsigned long, unsigned long);
 353
 354 extern void set_pageblock_order(void);
 355 unsigned long reclaim_clean_pages_from_list(struct zone *zone,
 356                                             struct list_head *page_list);
 357 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
 358 #define ALLOC_WMARK_MIN         WMARK_MIN
 359 #define ALLOC_WMARK_LOW         WMARK_LOW
 360 #define ALLOC_WMARK_HIGH        WMARK_HIGH
 361 #define ALLOC_NO_WATERMARKS     0x04 /* don't check watermarks at all */
 362
 363 /* Mask to get the watermark bits */
 364 #define ALLOC_WMARK_MASK        (ALLOC_NO_WATERMARKS-1)
 365
 366 #define ALLOC_HARDER            0x10 /* try to alloc harder */
 367 #define ALLOC_HIGH              0x20 /* __GFP_HIGH set */
 368 #define ALLOC_CPUSET            0x40 /* check for correct cpuset */
 369 #define ALLOC_CMA               0x80 /* allow allocations from CMA areas */
 370
 371 #endif  /* __MM_INTERNAL_H */