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 #include <linux/pagemap.h>
  17 #include <linux/tracepoint-defs.h>
  18
  19 /*
  20  * The set of flags that only affect watermark checking and reclaim
  21  * behaviour. This is used by the MM to obey the caller constraints
  22  * about IO, FS and watermark checking while ignoring placement
  23  * hints such as HIGHMEM usage.
  24  */
  25 #define GFP_RECLAIM_MASK (__GFP_RECLAIM|__GFP_HIGH|__GFP_IO|__GFP_FS|\
  26                         __GFP_NOWARN|__GFP_REPEAT|__GFP_NOFAIL|\
  27                         __GFP_NORETRY|__GFP_MEMALLOC|__GFP_NOMEMALLOC|\
  28                         __GFP_ATOMIC)
  29
  30 /* The GFP flags allowed during early boot */
  31 #define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_RECLAIM|__GFP_IO|__GFP_FS))
  32
  33 /* Control allocation cpuset and node placement constraints */
  34 #define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE)
  35
  36 /* Do not use these with a slab allocator */
  37 #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
  38
  39 int do_swap_page(struct fault_env *fe, pte_t orig_pte);
  40
  41 void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
  42                 unsigned long floor, unsigned long ceiling);
  43
  44 void unmap_page_range(struct mmu_gather *tlb,
  45                              struct vm_area_struct *vma,
  46                              unsigned long addr, unsigned long end,
  47                              struct zap_details *details);
  48
  49 extern int __do_page_cache_readahead(struct address_space *mapping,
  50                 struct file *filp, pgoff_t offset, unsigned long nr_to_read,
  51                 unsigned long lookahead_size);
  52
  53 /*
  54  * Submit IO for the read-ahead request in file_ra_state.
  55  */
  56 static inline unsigned long ra_submit(struct file_ra_state *ra,
  57                 struct address_space *mapping, struct file *filp)
  58 {
  59         return __do_page_cache_readahead(mapping, filp,
  60                                         ra->start, ra->size, ra->async_size);
  61 }
  62
  63 /*
  64  * Turn a non-refcounted page (->_refcount == 0) into refcounted with
  65  * a count of one.
  66  */
  67 static inline void set_page_refcounted(struct page *page)
  68 {
  69         VM_BUG_ON_PAGE(PageTail(page), page);
  70         VM_BUG_ON_PAGE(page_ref_count(page), page);
  71         set_page_count(page, 1);
  72 }
  73
  74 extern unsigned long highest_memmap_pfn;
  75
  76 /*
  77  * in mm/vmscan.c:
  78  */
  79 extern int isolate_lru_page(struct page *page);
  80 extern void putback_lru_page(struct page *page);
  81 extern bool pgdat_reclaimable(struct pglist_data *pgdat);
  82
  83 /*
  84  * in mm/rmap.c:
  85  */
  86 extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
  87
  88 /*
  89  * in mm/page_alloc.c
  90  */
  91
  92 /*
  93  * Structure for holding the mostly immutable allocation parameters passed
  94  * between functions involved in allocations, including the alloc_pages*
  95  * family of functions.
  96  *
  97  * nodemask, migratetype and high_zoneidx are initialized only once in
  98  * __alloc_pages_nodemask() and then never change.
  99  *
 100  * zonelist, preferred_zone and classzone_idx are set first in
 101  * __alloc_pages_nodemask() for the fast path, and might be later changed
 102  * in __alloc_pages_slowpath(). All other functions pass the whole strucure
 103  * by a const pointer.
 104  */
 105 struct alloc_context {
 106         struct zonelist *zonelist;
 107         nodemask_t *nodemask;
 108         struct zoneref *preferred_zoneref;
 109         int migratetype;
 110         enum zone_type high_zoneidx;
 111         bool spread_dirty_pages;
 112 };
 113
 114 #define ac_classzone_idx(ac) zonelist_zone_idx(ac->preferred_zoneref)
 115
 116 /*
 117  * Locate the struct page for both the matching buddy in our
 118  * pair (buddy1) and the combined O(n+1) page they form (page).
 119  *
 120  * 1) Any buddy B1 will have an order O twin B2 which satisfies
 121  * the following equation:
 122  *     B2 = B1 ^ (1 << O)
 123  * For example, if the starting buddy (buddy2) is #8 its order
 124  * 1 buddy is #10:
 125  *     B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
 126  *
 127  * 2) Any buddy B will have an order O+1 parent P which
 128  * satisfies the following equation:
 129  *     P = B & ~(1 << O)
 130  *
 131  * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
 132  */
 133 static inline unsigned long
 134 __find_buddy_index(unsigned long page_idx, unsigned int order)
 135 {
 136         return page_idx ^ (1 << order);
 137 }
 138
 139 extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn,
 140                                 unsigned long end_pfn, struct zone *zone);
 141
 142 static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn,
 143                                 unsigned long end_pfn, struct zone *zone)
 144 {
 145         if (zone->contiguous)
 146                 return pfn_to_page(start_pfn);
 147
 148         return __pageblock_pfn_to_page(start_pfn, end_pfn, zone);
 149 }
 150
 151 extern int __isolate_free_page(struct page *page, unsigned int order);
 152 extern void __free_pages_bootmem(struct page *page, unsigned long pfn,
 153                                         unsigned int order);
 154 extern void prep_compound_page(struct page *page, unsigned int order);
 155 extern void post_alloc_hook(struct page *page, unsigned int order,
 156                                         gfp_t gfp_flags);
 157 extern int user_min_free_kbytes;
 158
 159 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
 160
 161 /*
 162  * in mm/compaction.c
 163  */
 164 /*
 165  * compact_control is used to track pages being migrated and the free pages
 166  * they are being migrated to during memory compaction. The free_pfn starts
 167  * at the end of a zone and migrate_pfn begins at the start. Movable pages
 168  * are moved to the end of a zone during a compaction run and the run
 169  * completes when free_pfn <= migrate_pfn
 170  */
 171 struct compact_control {
 172         struct list_head freepages;     /* List of free pages to migrate to */
 173         struct list_head migratepages;  /* List of pages being migrated */
 174         unsigned long nr_freepages;     /* Number of isolated free pages */
 175         unsigned long nr_migratepages;  /* Number of pages to migrate */
 176         unsigned long free_pfn;         /* isolate_freepages search base */
 177         unsigned long migrate_pfn;      /* isolate_migratepages search base */
 178         unsigned long last_migrated_pfn;/* Not yet flushed page being freed */
 179         enum migrate_mode mode;         /* Async or sync migration mode */
 180         bool ignore_skip_hint;          /* Scan blocks even if marked skip */
 181         bool direct_compaction;         /* False from kcompactd or /proc/... */
 182         bool whole_zone;                /* Whole zone has been scanned */
 183         int order;                      /* order a direct compactor needs */
 184         const gfp_t gfp_mask;           /* gfp mask of a direct compactor */
 185         const unsigned int alloc_flags; /* alloc flags of a direct compactor */
 186         const int classzone_idx;        /* zone index of a direct compactor */
 187         struct zone *zone;
 188         bool contended;                 /* Signal lock or sched contention */
 189 };
 190
 191 unsigned long
 192 isolate_freepages_range(struct compact_control *cc,
 193                         unsigned long start_pfn, unsigned long end_pfn);
 194 unsigned long
 195 isolate_migratepages_range(struct compact_control *cc,
 196                            unsigned long low_pfn, unsigned long end_pfn);
 197 int find_suitable_fallback(struct free_area *area, unsigned int order,
 198                         int migratetype, bool only_stealable, bool *can_steal);
 199
 200 #endif
 201
 202 /*
 203  * This function returns the order of a free page in the buddy system. In
 204  * general, page_zone(page)->lock must be held by the caller to prevent the
 205  * page from being allocated in parallel and returning garbage as the order.
 206  * If a caller does not hold page_zone(page)->lock, it must guarantee that the
 207  * page cannot be allocated or merged in parallel. Alternatively, it must
 208  * handle invalid values gracefully, and use page_order_unsafe() below.
 209  */
 210 static inline unsigned int page_order(struct page *page)
 211 {
 212         /* PageBuddy() must be checked by the caller */
 213         return page_private(page);
 214 }
 215
 216 /*
 217  * Like page_order(), but for callers who cannot afford to hold the zone lock.
 218  * PageBuddy() should be checked first by the caller to minimize race window,
 219  * and invalid values must be handled gracefully.
 220  *
 221  * READ_ONCE is used so that if the caller assigns the result into a local
 222  * variable and e.g. tests it for valid range before using, the compiler cannot
 223  * decide to remove the variable and inline the page_private(page) multiple
 224  * times, potentially observing different values in the tests and the actual
 225  * use of the result.
 226  */
 227 #define page_order_unsafe(page)         READ_ONCE(page_private(page))
 228
 229 static inline bool is_cow_mapping(vm_flags_t flags)
 230 {
 231         return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
 232 }
 233
 234 /*
 235  * These three helpers classifies VMAs for virtual memory accounting.
 236  */
 237
 238 /*
 239  * Executable code area - executable, not writable, not stack
 240  */
 241 static inline bool is_exec_mapping(vm_flags_t flags)
 242 {
 243         return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC;
 244 }
 245
 246 /*
 247  * Stack area - atomatically grows in one direction
 248  *
 249  * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
 250  * do_mmap() forbids all other combinations.
 251  */
 252 static inline bool is_stack_mapping(vm_flags_t flags)
 253 {
 254         return (flags & VM_STACK) == VM_STACK;
 255 }
 256
 257 /*
 258  * Data area - private, writable, not stack
 259  */
 260 static inline bool is_data_mapping(vm_flags_t flags)
 261 {
 262         return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE;
 263 }
 264
 265 /* mm/util.c */
 266 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
 267                 struct vm_area_struct *prev, struct rb_node *rb_parent);
 268
 269 #ifdef CONFIG_MMU
 270 extern long populate_vma_page_range(struct vm_area_struct *vma,
 271                 unsigned long start, unsigned long end, int *nonblocking);
 272 extern void munlock_vma_pages_range(struct vm_area_struct *vma,
 273                         unsigned long start, unsigned long end);
 274 static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
 275 {
 276         munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
 277 }
 278
 279 /*
 280  * must be called with vma's mmap_sem held for read or write, and page locked.
 281  */
 282 extern void mlock_vma_page(struct page *page);
 283 extern unsigned int munlock_vma_page(struct page *page);
 284
 285 /*
 286  * Clear the page's PageMlocked().  This can be useful in a situation where
 287  * we want to unconditionally remove a page from the pagecache -- e.g.,
 288  * on truncation or freeing.
 289  *
 290  * It is legal to call this function for any page, mlocked or not.
 291  * If called for a page that is still mapped by mlocked vmas, all we do
 292  * is revert to lazy LRU behaviour -- semantics are not broken.
 293  */
 294 extern void clear_page_mlock(struct page *page);
 295
 296 /*
 297  * mlock_migrate_page - called only from migrate_misplaced_transhuge_page()
 298  * (because that does not go through the full procedure of migration ptes):
 299  * to migrate the Mlocked page flag; update statistics.
 300  */
 301 static inline void mlock_migrate_page(struct page *newpage, struct page *page)
 302 {
 303         if (TestClearPageMlocked(page)) {
 304                 int nr_pages = hpage_nr_pages(page);
 305
 306                 /* Holding pmd lock, no change in irq context: __mod is safe */
 307                 __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
 308                 SetPageMlocked(newpage);
 309                 __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
 310         }
 311 }
 312
 313 extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
 314
 315 /*
 316  * At what user virtual address is page expected in @vma?
 317  */
 318 static inline unsigned long
 319 __vma_address(struct page *page, struct vm_area_struct *vma)
 320 {
 321         pgoff_t pgoff = page_to_pgoff(page);
 322         return vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
 323 }
 324
 325 static inline unsigned long
 326 vma_address(struct page *page, struct vm_area_struct *vma)
 327 {
 328         unsigned long address = __vma_address(page, vma);
 329
 330         /* page should be within @vma mapping range */
 331         VM_BUG_ON_VMA(address < vma->vm_start || address >= vma->vm_end, vma);
 332
 333         return address;
 334 }
 335
 336 #else /* !CONFIG_MMU */
 337 static inline void clear_page_mlock(struct page *page) { }
 338 static inline void mlock_vma_page(struct page *page) { }
 339 static inline void mlock_migrate_page(struct page *new, struct page *old) { }
 340
 341 #endif /* !CONFIG_MMU */
 342
 343 /*
 344  * Return the mem_map entry representing the 'offset' subpage within
 345  * the maximally aligned gigantic page 'base'.  Handle any discontiguity
 346  * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
 347  */
 348 static inline struct page *mem_map_offset(struct page *base, int offset)
 349 {
 350         if (unlikely(offset >= MAX_ORDER_NR_PAGES))
 351                 return nth_page(base, offset);
 352         return base + offset;
 353 }
 354
 355 /*
 356  * Iterator over all subpages within the maximally aligned gigantic
 357  * page 'base'.  Handle any discontiguity in the mem_map.
 358  */
 359 static inline struct page *mem_map_next(struct page *iter,
 360                                                 struct page *base, int offset)
 361 {
 362         if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
 363                 unsigned long pfn = page_to_pfn(base) + offset;
 364                 if (!pfn_valid(pfn))
 365                         return NULL;
 366                 return pfn_to_page(pfn);
 367         }
 368         return iter + 1;
 369 }
 370
 371 /*
 372  * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
 373  * so all functions starting at paging_init should be marked __init
 374  * in those cases. SPARSEMEM, however, allows for memory hotplug,
 375  * and alloc_bootmem_node is not used.
 376  */
 377 #ifdef CONFIG_SPARSEMEM
 378 #define __paginginit __meminit
 379 #else
 380 #define __paginginit __init
 381 #endif
 382
 383 /* Memory initialisation debug and verification */
 384 enum mminit_level {
 385         MMINIT_WARNING,
 386         MMINIT_VERIFY,
 387         MMINIT_TRACE
 388 };
 389
 390 #ifdef CONFIG_DEBUG_MEMORY_INIT
 391
 392 extern int mminit_loglevel;
 393
 394 #define mminit_dprintk(level, prefix, fmt, arg...) \
 395 do { \
 396         if (level < mminit_loglevel) { \
 397                 if (level <= MMINIT_WARNING) \
 398                         pr_warn("mminit::" prefix " " fmt, ##arg);      \
 399                 else \
 400                         printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
 401         } \
 402 } while (0)
 403
 404 extern void mminit_verify_pageflags_layout(void);
 405 extern void mminit_verify_zonelist(void);
 406 #else
 407
 408 static inline void mminit_dprintk(enum mminit_level level,
 409                                 const char *prefix, const char *fmt, ...)
 410 {
 411 }
 412
 413 static inline void mminit_verify_pageflags_layout(void)
 414 {
 415 }
 416
 417 static inline void mminit_verify_zonelist(void)
 418 {
 419 }
 420 #endif /* CONFIG_DEBUG_MEMORY_INIT */
 421
 422 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
 423 #if defined(CONFIG_SPARSEMEM)
 424 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 425                                 unsigned long *end_pfn);
 426 #else
 427 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 428                                 unsigned long *end_pfn)
 429 {
 430 }
 431 #endif /* CONFIG_SPARSEMEM */
 432
 433 #define NODE_RECLAIM_NOSCAN     -2
 434 #define NODE_RECLAIM_FULL       -1
 435 #define NODE_RECLAIM_SOME       0
 436 #define NODE_RECLAIM_SUCCESS    1
 437
 438 extern int hwpoison_filter(struct page *p);
 439
 440 extern u32 hwpoison_filter_dev_major;
 441 extern u32 hwpoison_filter_dev_minor;
 442 extern u64 hwpoison_filter_flags_mask;
 443 extern u64 hwpoison_filter_flags_value;
 444 extern u64 hwpoison_filter_memcg;
 445 extern u32 hwpoison_filter_enable;
 446
 447 extern unsigned long  __must_check vm_mmap_pgoff(struct file *, unsigned long,
 448         unsigned long, unsigned long,
 449         unsigned long, unsigned long);
 450
 451 extern void set_pageblock_order(void);
 452 unsigned long reclaim_clean_pages_from_list(struct zone *zone,
 453                                             struct list_head *page_list);
 454 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
 455 #define ALLOC_WMARK_MIN         WMARK_MIN
 456 #define ALLOC_WMARK_LOW         WMARK_LOW
 457 #define ALLOC_WMARK_HIGH        WMARK_HIGH
 458 #define ALLOC_NO_WATERMARKS     0x04 /* don't check watermarks at all */
 459
 460 /* Mask to get the watermark bits */
 461 #define ALLOC_WMARK_MASK        (ALLOC_NO_WATERMARKS-1)
 462
 463 #define ALLOC_HARDER            0x10 /* try to alloc harder */
 464 #define ALLOC_HIGH              0x20 /* __GFP_HIGH set */
 465 #define ALLOC_CPUSET            0x40 /* check for correct cpuset */
 466 #define ALLOC_CMA               0x80 /* allow allocations from CMA areas */
 467
 468 enum ttu_flags;
 469 struct tlbflush_unmap_batch;
 470
 471 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
 472 void try_to_unmap_flush(void);
 473 void try_to_unmap_flush_dirty(void);
 474 #else
 475 static inline void try_to_unmap_flush(void)
 476 {
 477 }
 478 static inline void try_to_unmap_flush_dirty(void)
 479 {
 480 }
 481
 482 #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
 483
 484 extern const struct trace_print_flags pageflag_names[];
 485 extern const struct trace_print_flags vmaflag_names[];
 486 extern const struct trace_print_flags gfpflag_names[];
 487
 488 #endif  /* __MM_INTERNAL_H */