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 void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
  40                 unsigned long floor, unsigned long ceiling);
  41
  42 void unmap_page_range(struct mmu_gather *tlb,
  43                              struct vm_area_struct *vma,
  44                              unsigned long addr, unsigned long end,
  45                              struct zap_details *details);
  46
  47 extern int __do_page_cache_readahead(struct address_space *mapping,
  48                 struct file *filp, pgoff_t offset, unsigned long nr_to_read,
  49                 unsigned long lookahead_size);
  50
  51 /*
  52  * Submit IO for the read-ahead request in file_ra_state.
  53  */
  54 static inline unsigned long ra_submit(struct file_ra_state *ra,
  55                 struct address_space *mapping, struct file *filp)
  56 {
  57         return __do_page_cache_readahead(mapping, filp,
  58                                         ra->start, ra->size, ra->async_size);
  59 }
  60
  61 /*
  62  * Turn a non-refcounted page (->_refcount == 0) into refcounted with
  63  * a count of one.
  64  */
  65 static inline void set_page_refcounted(struct page *page)
  66 {
  67         VM_BUG_ON_PAGE(PageTail(page), page);
  68         VM_BUG_ON_PAGE(page_ref_count(page), page);
  69         set_page_count(page, 1);
  70 }
  71
  72 extern unsigned long highest_memmap_pfn;
  73
  74 /*
  75  * in mm/vmscan.c:
  76  */
  77 extern int isolate_lru_page(struct page *page);
  78 extern void putback_lru_page(struct page *page);
  79 extern bool zone_reclaimable(struct zone *zone);
  80
  81 /*
  82  * in mm/rmap.c:
  83  */
  84 extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
  85
  86 /*
  87  * in mm/page_alloc.c
  88  */
  89
  90 /*
  91  * Structure for holding the mostly immutable allocation parameters passed
  92  * between functions involved in allocations, including the alloc_pages*
  93  * family of functions.
  94  *
  95  * nodemask, migratetype and high_zoneidx are initialized only once in
  96  * __alloc_pages_nodemask() and then never change.
  97  *
  98  * zonelist, preferred_zone and classzone_idx are set first in
  99  * __alloc_pages_nodemask() for the fast path, and might be later changed
 100  * in __alloc_pages_slowpath(). All other functions pass the whole strucure
 101  * by a const pointer.
 102  */
 103 struct alloc_context {
 104         struct zonelist *zonelist;
 105         nodemask_t *nodemask;
 106         struct zoneref *preferred_zoneref;
 107         int migratetype;
 108         enum zone_type high_zoneidx;
 109         bool spread_dirty_pages;
 110 };
 111
 112 #define ac_classzone_idx(ac) zonelist_zone_idx(ac->preferred_zoneref)
 113
 114 /*
 115  * Locate the struct page for both the matching buddy in our
 116  * pair (buddy1) and the combined O(n+1) page they form (page).
 117  *
 118  * 1) Any buddy B1 will have an order O twin B2 which satisfies
 119  * the following equation:
 120  *     B2 = B1 ^ (1 << O)
 121  * For example, if the starting buddy (buddy2) is #8 its order
 122  * 1 buddy is #10:
 123  *     B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
 124  *
 125  * 2) Any buddy B will have an order O+1 parent P which
 126  * satisfies the following equation:
 127  *     P = B & ~(1 << O)
 128  *
 129  * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
 130  */
 131 static inline unsigned long
 132 __find_buddy_index(unsigned long page_idx, unsigned int order)
 133 {
 134         return page_idx ^ (1 << order);
 135 }
 136
 137 extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn,
 138                                 unsigned long end_pfn, struct zone *zone);
 139
 140 static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn,
 141                                 unsigned long end_pfn, struct zone *zone)
 142 {
 143         if (zone->contiguous)
 144                 return pfn_to_page(start_pfn);
 145
 146         return __pageblock_pfn_to_page(start_pfn, end_pfn, zone);
 147 }
 148
 149 extern int __isolate_free_page(struct page *page, unsigned int order);
 150 extern void __free_pages_bootmem(struct page *page, unsigned long pfn,
 151                                         unsigned int order);
 152 extern void prep_compound_page(struct page *page, unsigned int order);
 153 extern int user_min_free_kbytes;
 154
 155 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
 156
 157 /*
 158  * in mm/compaction.c
 159  */
 160 /*
 161  * compact_control is used to track pages being migrated and the free pages
 162  * they are being migrated to during memory compaction. The free_pfn starts
 163  * at the end of a zone and migrate_pfn begins at the start. Movable pages
 164  * are moved to the end of a zone during a compaction run and the run
 165  * completes when free_pfn <= migrate_pfn
 166  */
 167 struct compact_control {
 168         struct list_head freepages;     /* List of free pages to migrate to */
 169         struct list_head migratepages;  /* List of pages being migrated */
 170         unsigned long nr_freepages;     /* Number of isolated free pages */
 171         unsigned long nr_migratepages;  /* Number of pages to migrate */
 172         unsigned long free_pfn;         /* isolate_freepages search base */
 173         unsigned long migrate_pfn;      /* isolate_migratepages search base */
 174         unsigned long last_migrated_pfn;/* Not yet flushed page being freed */
 175         enum migrate_mode mode;         /* Async or sync migration mode */
 176         bool ignore_skip_hint;          /* Scan blocks even if marked skip */
 177         bool direct_compaction;         /* False from kcompactd or /proc/... */
 178         bool whole_zone;                /* Whole zone has been scanned */
 179         int order;                      /* order a direct compactor needs */
 180         const gfp_t gfp_mask;           /* gfp mask of a direct compactor */
 181         const unsigned int alloc_flags; /* alloc flags of a direct compactor */
 182         const int classzone_idx;        /* zone index of a direct compactor */
 183         struct zone *zone;
 184         int contended;                  /* Signal need_sched() or lock
 185                                          * contention detected during
 186                                          * compaction
 187                                          */
 188 };
 189
 190 unsigned long
 191 isolate_freepages_range(struct compact_control *cc,
 192                         unsigned long start_pfn, unsigned long end_pfn);
 193 unsigned long
 194 isolate_migratepages_range(struct compact_control *cc,
 195                            unsigned long low_pfn, unsigned long end_pfn);
 196 int find_suitable_fallback(struct free_area *area, unsigned int order,
 197                         int migratetype, bool only_stealable, bool *can_steal);
 198
 199 #endif
 200
 201 /*
 202  * This function returns the order of a free page in the buddy system. In
 203  * general, page_zone(page)->lock must be held by the caller to prevent the
 204  * page from being allocated in parallel and returning garbage as the order.
 205  * If a caller does not hold page_zone(page)->lock, it must guarantee that the
 206  * page cannot be allocated or merged in parallel. Alternatively, it must
 207  * handle invalid values gracefully, and use page_order_unsafe() below.
 208  */
 209 static inline unsigned int page_order(struct page *page)
 210 {
 211         /* PageBuddy() must be checked by the caller */
 212         return page_private(page);
 213 }
 214
 215 /*
 216  * Like page_order(), but for callers who cannot afford to hold the zone lock.
 217  * PageBuddy() should be checked first by the caller to minimize race window,
 218  * and invalid values must be handled gracefully.
 219  *
 220  * READ_ONCE is used so that if the caller assigns the result into a local
 221  * variable and e.g. tests it for valid range before using, the compiler cannot
 222  * decide to remove the variable and inline the page_private(page) multiple
 223  * times, potentially observing different values in the tests and the actual
 224  * use of the result.
 225  */
 226 #define page_order_unsafe(page)         READ_ONCE(page_private(page))
 227
 228 static inline bool is_cow_mapping(vm_flags_t flags)
 229 {
 230         return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
 231 }
 232
 233 /*
 234  * These three helpers classifies VMAs for virtual memory accounting.
 235  */
 236
 237 /*
 238  * Executable code area - executable, not writable, not stack
 239  */
 240 static inline bool is_exec_mapping(vm_flags_t flags)
 241 {
 242         return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC;
 243 }
 244
 245 /*
 246  * Stack area - atomatically grows in one direction
 247  *
 248  * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
 249  * do_mmap() forbids all other combinations.
 250  */
 251 static inline bool is_stack_mapping(vm_flags_t flags)
 252 {
 253         return (flags & VM_STACK) == VM_STACK;
 254 }
 255
 256 /*
 257  * Data area - private, writable, not stack
 258  */
 259 static inline bool is_data_mapping(vm_flags_t flags)
 260 {
 261         return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE;
 262 }
 263
 264 /* mm/util.c */
 265 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
 266                 struct vm_area_struct *prev, struct rb_node *rb_parent);
 267
 268 #ifdef CONFIG_MMU
 269 extern long populate_vma_page_range(struct vm_area_struct *vma,
 270                 unsigned long start, unsigned long end, int *nonblocking);
 271 extern void munlock_vma_pages_range(struct vm_area_struct *vma,
 272                         unsigned long start, unsigned long end);
 273 static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
 274 {
 275         munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
 276 }
 277
 278 /*
 279  * must be called with vma's mmap_sem held for read or write, and page locked.
 280  */
 281 extern void mlock_vma_page(struct page *page);
 282 extern unsigned int munlock_vma_page(struct page *page);
 283
 284 /*
 285  * Clear the page's PageMlocked().  This can be useful in a situation where
 286  * we want to unconditionally remove a page from the pagecache -- e.g.,
 287  * on truncation or freeing.
 288  *
 289  * It is legal to call this function for any page, mlocked or not.
 290  * If called for a page that is still mapped by mlocked vmas, all we do
 291  * is revert to lazy LRU behaviour -- semantics are not broken.
 292  */
 293 extern void clear_page_mlock(struct page *page);
 294
 295 /*
 296  * mlock_migrate_page - called only from migrate_misplaced_transhuge_page()
 297  * (because that does not go through the full procedure of migration ptes):
 298  * to migrate the Mlocked page flag; update statistics.
 299  */
 300 static inline void mlock_migrate_page(struct page *newpage, struct page *page)
 301 {
 302         if (TestClearPageMlocked(page)) {
 303                 int nr_pages = hpage_nr_pages(page);
 304
 305                 /* Holding pmd lock, no change in irq context: __mod is safe */
 306                 __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
 307                 SetPageMlocked(newpage);
 308                 __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
 309         }
 310 }
 311
 312 extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
 313
 314 /*
 315  * At what user virtual address is page expected in @vma?
 316  */
 317 static inline unsigned long
 318 __vma_address(struct page *page, struct vm_area_struct *vma)
 319 {
 320         pgoff_t pgoff = page_to_pgoff(page);
 321         return vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
 322 }
 323
 324 static inline unsigned long
 325 vma_address(struct page *page, struct vm_area_struct *vma)
 326 {
 327         unsigned long address = __vma_address(page, vma);
 328
 329         /* page should be within @vma mapping range */
 330         VM_BUG_ON_VMA(address < vma->vm_start || address >= vma->vm_end, vma);
 331
 332         return address;
 333 }
 334
 335 #else /* !CONFIG_MMU */
 336 static inline void clear_page_mlock(struct page *page) { }
 337 static inline void mlock_vma_page(struct page *page) { }
 338 static inline void mlock_migrate_page(struct page *new, struct page *old) { }
 339
 340 #endif /* !CONFIG_MMU */
 341
 342 /*
 343  * Return the mem_map entry representing the 'offset' subpage within
 344  * the maximally aligned gigantic page 'base'.  Handle any discontiguity
 345  * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
 346  */
 347 static inline struct page *mem_map_offset(struct page *base, int offset)
 348 {
 349         if (unlikely(offset >= MAX_ORDER_NR_PAGES))
 350                 return nth_page(base, offset);
 351         return base + offset;
 352 }
 353
 354 /*
 355  * Iterator over all subpages within the maximally aligned gigantic
 356  * page 'base'.  Handle any discontiguity in the mem_map.
 357  */
 358 static inline struct page *mem_map_next(struct page *iter,
 359                                                 struct page *base, int offset)
 360 {
 361         if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
 362                 unsigned long pfn = page_to_pfn(base) + offset;
 363                 if (!pfn_valid(pfn))
 364                         return NULL;
 365                 return pfn_to_page(pfn);
 366         }
 367         return iter + 1;
 368 }
 369
 370 /*
 371  * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
 372  * so all functions starting at paging_init should be marked __init
 373  * in those cases. SPARSEMEM, however, allows for memory hotplug,
 374  * and alloc_bootmem_node is not used.
 375  */
 376 #ifdef CONFIG_SPARSEMEM
 377 #define __paginginit __meminit
 378 #else
 379 #define __paginginit __init
 380 #endif
 381
 382 /* Memory initialisation debug and verification */
 383 enum mminit_level {
 384         MMINIT_WARNING,
 385         MMINIT_VERIFY,
 386         MMINIT_TRACE
 387 };
 388
 389 #ifdef CONFIG_DEBUG_MEMORY_INIT
 390
 391 extern int mminit_loglevel;
 392
 393 #define mminit_dprintk(level, prefix, fmt, arg...) \
 394 do { \
 395         if (level < mminit_loglevel) { \
 396                 if (level <= MMINIT_WARNING) \
 397                         pr_warn("mminit::" prefix " " fmt, ##arg);      \
 398                 else \
 399                         printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
 400         } \
 401 } while (0)
 402
 403 extern void mminit_verify_pageflags_layout(void);
 404 extern void mminit_verify_zonelist(void);
 405 #else
 406
 407 static inline void mminit_dprintk(enum mminit_level level,
 408                                 const char *prefix, const char *fmt, ...)
 409 {
 410 }
 411
 412 static inline void mminit_verify_pageflags_layout(void)
 413 {
 414 }
 415
 416 static inline void mminit_verify_zonelist(void)
 417 {
 418 }
 419 #endif /* CONFIG_DEBUG_MEMORY_INIT */
 420
 421 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
 422 #if defined(CONFIG_SPARSEMEM)
 423 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 424                                 unsigned long *end_pfn);
 425 #else
 426 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 427                                 unsigned long *end_pfn)
 428 {
 429 }
 430 #endif /* CONFIG_SPARSEMEM */
 431
 432 #define ZONE_RECLAIM_NOSCAN     -2
 433 #define ZONE_RECLAIM_FULL       -1
 434 #define ZONE_RECLAIM_SOME       0
 435 #define ZONE_RECLAIM_SUCCESS    1
 436
 437 extern int hwpoison_filter(struct page *p);
 438
 439 extern u32 hwpoison_filter_dev_major;
 440 extern u32 hwpoison_filter_dev_minor;
 441 extern u64 hwpoison_filter_flags_mask;
 442 extern u64 hwpoison_filter_flags_value;
 443 extern u64 hwpoison_filter_memcg;
 444 extern u32 hwpoison_filter_enable;
 445
 446 extern unsigned long  __must_check vm_mmap_pgoff(struct file *, unsigned long,
 447         unsigned long, unsigned long,
 448         unsigned long, unsigned long);
 449
 450 extern void set_pageblock_order(void);
 451 unsigned long reclaim_clean_pages_from_list(struct zone *zone,
 452                                             struct list_head *page_list);
 453 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
 454 #define ALLOC_WMARK_MIN         WMARK_MIN
 455 #define ALLOC_WMARK_LOW         WMARK_LOW
 456 #define ALLOC_WMARK_HIGH        WMARK_HIGH
 457 #define ALLOC_NO_WATERMARKS     0x04 /* don't check watermarks at all */
 458
 459 /* Mask to get the watermark bits */
 460 #define ALLOC_WMARK_MASK        (ALLOC_NO_WATERMARKS-1)
 461
 462 #define ALLOC_HARDER            0x10 /* try to alloc harder */
 463 #define ALLOC_HIGH              0x20 /* __GFP_HIGH set */
 464 #define ALLOC_CPUSET            0x40 /* check for correct cpuset */
 465 #define ALLOC_CMA               0x80 /* allow allocations from CMA areas */
 466 #define ALLOC_FAIR              0x100 /* fair zone allocation */
 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 */