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
  29 /* The GFP flags allowed during early boot */
  30 #define GFP_BOOT_MASK (__GFP_BITS_MASK & ~(__GFP_RECLAIM|__GFP_IO|__GFP_FS))
  31
  32 /* Control allocation cpuset and node placement constraints */
  33 #define GFP_CONSTRAINT_MASK (__GFP_HARDWALL|__GFP_THISNODE)
  34
  35 /* Do not use these with a slab allocator */
  36 #define GFP_SLAB_BUG_MASK (__GFP_DMA32|__GFP_HIGHMEM|~__GFP_BITS_MASK)
  37
  38 void free_pgtables(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
  39                 unsigned long floor, unsigned long ceiling);
  40
  41 void unmap_page_range(struct mmu_gather *tlb,
  42                              struct vm_area_struct *vma,
  43                              unsigned long addr, unsigned long end,
  44                              struct zap_details *details);
  45
  46 extern int __do_page_cache_readahead(struct address_space *mapping,
  47                 struct file *filp, pgoff_t offset, unsigned long nr_to_read,
  48                 unsigned long lookahead_size);
  49
  50 /*
  51  * Submit IO for the read-ahead request in file_ra_state.
  52  */
  53 static inline unsigned long ra_submit(struct file_ra_state *ra,
  54                 struct address_space *mapping, struct file *filp)
  55 {
  56         return __do_page_cache_readahead(mapping, filp,
  57                                         ra->start, ra->size, ra->async_size);
  58 }
  59
  60 /*
  61  * Turn a non-refcounted page (->_refcount == 0) into refcounted with
  62  * a count of one.
  63  */
  64 static inline void set_page_refcounted(struct page *page)
  65 {
  66         VM_BUG_ON_PAGE(PageTail(page), page);
  67         VM_BUG_ON_PAGE(page_ref_count(page), page);
  68         set_page_count(page, 1);
  69 }
  70
  71 extern unsigned long highest_memmap_pfn;
  72
  73 /*
  74  * in mm/vmscan.c:
  75  */
  76 extern int isolate_lru_page(struct page *page);
  77 extern void putback_lru_page(struct page *page);
  78 extern bool zone_reclaimable(struct zone *zone);
  79
  80 /*
  81  * in mm/rmap.c:
  82  */
  83 extern pmd_t *mm_find_pmd(struct mm_struct *mm, unsigned long address);
  84
  85 /*
  86  * in mm/page_alloc.c
  87  */
  88
  89 /*
  90  * Structure for holding the mostly immutable allocation parameters passed
  91  * between functions involved in allocations, including the alloc_pages*
  92  * family of functions.
  93  *
  94  * nodemask, migratetype and high_zoneidx are initialized only once in
  95  * __alloc_pages_nodemask() and then never change.
  96  *
  97  * zonelist, preferred_zone and classzone_idx are set first in
  98  * __alloc_pages_nodemask() for the fast path, and might be later changed
  99  * in __alloc_pages_slowpath(). All other functions pass the whole strucure
 100  * by a const pointer.
 101  */
 102 struct alloc_context {
 103         struct zonelist *zonelist;
 104         nodemask_t *nodemask;
 105         struct zoneref *preferred_zoneref;
 106         int migratetype;
 107         enum zone_type high_zoneidx;
 108         bool spread_dirty_pages;
 109 };
 110
 111 #define ac_classzone_idx(ac) zonelist_zone_idx(ac->preferred_zoneref)
 112
 113 /*
 114  * Locate the struct page for both the matching buddy in our
 115  * pair (buddy1) and the combined O(n+1) page they form (page).
 116  *
 117  * 1) Any buddy B1 will have an order O twin B2 which satisfies
 118  * the following equation:
 119  *     B2 = B1 ^ (1 << O)
 120  * For example, if the starting buddy (buddy2) is #8 its order
 121  * 1 buddy is #10:
 122  *     B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
 123  *
 124  * 2) Any buddy B will have an order O+1 parent P which
 125  * satisfies the following equation:
 126  *     P = B & ~(1 << O)
 127  *
 128  * Assumption: *_mem_map is contiguous at least up to MAX_ORDER
 129  */
 130 static inline unsigned long
 131 __find_buddy_index(unsigned long page_idx, unsigned int order)
 132 {
 133         return page_idx ^ (1 << order);
 134 }
 135
 136 extern struct page *__pageblock_pfn_to_page(unsigned long start_pfn,
 137                                 unsigned long end_pfn, struct zone *zone);
 138
 139 static inline struct page *pageblock_pfn_to_page(unsigned long start_pfn,
 140                                 unsigned long end_pfn, struct zone *zone)
 141 {
 142         if (zone->contiguous)
 143                 return pfn_to_page(start_pfn);
 144
 145         return __pageblock_pfn_to_page(start_pfn, end_pfn, zone);
 146 }
 147
 148 extern int __isolate_free_page(struct page *page, unsigned int order);
 149 extern void __free_pages_bootmem(struct page *page, unsigned long pfn,
 150                                         unsigned int order);
 151 extern void prep_compound_page(struct page *page, unsigned int order);
 152 extern int user_min_free_kbytes;
 153
 154 #if defined CONFIG_COMPACTION || defined CONFIG_CMA
 155
 156 /*
 157  * in mm/compaction.c
 158  */
 159 /*
 160  * compact_control is used to track pages being migrated and the free pages
 161  * they are being migrated to during memory compaction. The free_pfn starts
 162  * at the end of a zone and migrate_pfn begins at the start. Movable pages
 163  * are moved to the end of a zone during a compaction run and the run
 164  * completes when free_pfn <= migrate_pfn
 165  */
 166 struct compact_control {
 167         struct list_head freepages;     /* List of free pages to migrate to */
 168         struct list_head migratepages;  /* List of pages being migrated */
 169         unsigned long nr_freepages;     /* Number of isolated free pages */
 170         unsigned long nr_migratepages;  /* Number of pages to migrate */
 171         unsigned long free_pfn;         /* isolate_freepages search base */
 172         unsigned long migrate_pfn;      /* isolate_migratepages search base */
 173         unsigned long last_migrated_pfn;/* Not yet flushed page being freed */
 174         enum migrate_mode mode;         /* Async or sync migration mode */
 175         bool ignore_skip_hint;          /* Scan blocks even if marked skip */
 176         bool direct_compaction;         /* False from kcompactd or /proc/... */
 177         int order;                      /* order a direct compactor needs */
 178         const gfp_t gfp_mask;           /* gfp mask of a direct compactor */
 179         const unsigned int alloc_flags; /* alloc flags of a direct compactor */
 180         const int classzone_idx;        /* zone index of a direct compactor */
 181         struct zone *zone;
 182         int contended;                  /* Signal need_sched() or lock
 183                                          * contention detected during
 184                                          * compaction
 185                                          */
 186 };
 187
 188 unsigned long
 189 isolate_freepages_range(struct compact_control *cc,
 190                         unsigned long start_pfn, unsigned long end_pfn);
 191 unsigned long
 192 isolate_migratepages_range(struct compact_control *cc,
 193                            unsigned long low_pfn, unsigned long end_pfn);
 194 int find_suitable_fallback(struct free_area *area, unsigned int order,
 195                         int migratetype, bool only_stealable, bool *can_steal);
 196
 197 #endif
 198
 199 /*
 200  * This function returns the order of a free page in the buddy system. In
 201  * general, page_zone(page)->lock must be held by the caller to prevent the
 202  * page from being allocated in parallel and returning garbage as the order.
 203  * If a caller does not hold page_zone(page)->lock, it must guarantee that the
 204  * page cannot be allocated or merged in parallel. Alternatively, it must
 205  * handle invalid values gracefully, and use page_order_unsafe() below.
 206  */
 207 static inline unsigned int page_order(struct page *page)
 208 {
 209         /* PageBuddy() must be checked by the caller */
 210         return page_private(page);
 211 }
 212
 213 /*
 214  * Like page_order(), but for callers who cannot afford to hold the zone lock.
 215  * PageBuddy() should be checked first by the caller to minimize race window,
 216  * and invalid values must be handled gracefully.
 217  *
 218  * READ_ONCE is used so that if the caller assigns the result into a local
 219  * variable and e.g. tests it for valid range before using, the compiler cannot
 220  * decide to remove the variable and inline the page_private(page) multiple
 221  * times, potentially observing different values in the tests and the actual
 222  * use of the result.
 223  */
 224 #define page_order_unsafe(page)         READ_ONCE(page_private(page))
 225
 226 static inline bool is_cow_mapping(vm_flags_t flags)
 227 {
 228         return (flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
 229 }
 230
 231 /*
 232  * These three helpers classifies VMAs for virtual memory accounting.
 233  */
 234
 235 /*
 236  * Executable code area - executable, not writable, not stack
 237  */
 238 static inline bool is_exec_mapping(vm_flags_t flags)
 239 {
 240         return (flags & (VM_EXEC | VM_WRITE | VM_STACK)) == VM_EXEC;
 241 }
 242
 243 /*
 244  * Stack area - atomatically grows in one direction
 245  *
 246  * VM_GROWSUP / VM_GROWSDOWN VMAs are always private anonymous:
 247  * do_mmap() forbids all other combinations.
 248  */
 249 static inline bool is_stack_mapping(vm_flags_t flags)
 250 {
 251         return (flags & VM_STACK) == VM_STACK;
 252 }
 253
 254 /*
 255  * Data area - private, writable, not stack
 256  */
 257 static inline bool is_data_mapping(vm_flags_t flags)
 258 {
 259         return (flags & (VM_WRITE | VM_SHARED | VM_STACK)) == VM_WRITE;
 260 }
 261
 262 /* mm/util.c */
 263 void __vma_link_list(struct mm_struct *mm, struct vm_area_struct *vma,
 264                 struct vm_area_struct *prev, struct rb_node *rb_parent);
 265
 266 #ifdef CONFIG_MMU
 267 extern long populate_vma_page_range(struct vm_area_struct *vma,
 268                 unsigned long start, unsigned long end, int *nonblocking);
 269 extern void munlock_vma_pages_range(struct vm_area_struct *vma,
 270                         unsigned long start, unsigned long end);
 271 static inline void munlock_vma_pages_all(struct vm_area_struct *vma)
 272 {
 273         munlock_vma_pages_range(vma, vma->vm_start, vma->vm_end);
 274 }
 275
 276 /*
 277  * must be called with vma's mmap_sem held for read or write, and page locked.
 278  */
 279 extern void mlock_vma_page(struct page *page);
 280 extern unsigned int munlock_vma_page(struct page *page);
 281
 282 /*
 283  * Clear the page's PageMlocked().  This can be useful in a situation where
 284  * we want to unconditionally remove a page from the pagecache -- e.g.,
 285  * on truncation or freeing.
 286  *
 287  * It is legal to call this function for any page, mlocked or not.
 288  * If called for a page that is still mapped by mlocked vmas, all we do
 289  * is revert to lazy LRU behaviour -- semantics are not broken.
 290  */
 291 extern void clear_page_mlock(struct page *page);
 292
 293 /*
 294  * mlock_migrate_page - called only from migrate_misplaced_transhuge_page()
 295  * (because that does not go through the full procedure of migration ptes):
 296  * to migrate the Mlocked page flag; update statistics.
 297  */
 298 static inline void mlock_migrate_page(struct page *newpage, struct page *page)
 299 {
 300         if (TestClearPageMlocked(page)) {
 301                 int nr_pages = hpage_nr_pages(page);
 302
 303                 /* Holding pmd lock, no change in irq context: __mod is safe */
 304                 __mod_zone_page_state(page_zone(page), NR_MLOCK, -nr_pages);
 305                 SetPageMlocked(newpage);
 306                 __mod_zone_page_state(page_zone(newpage), NR_MLOCK, nr_pages);
 307         }
 308 }
 309
 310 extern pmd_t maybe_pmd_mkwrite(pmd_t pmd, struct vm_area_struct *vma);
 311
 312 /*
 313  * At what user virtual address is page expected in @vma?
 314  */
 315 static inline unsigned long
 316 __vma_address(struct page *page, struct vm_area_struct *vma)
 317 {
 318         pgoff_t pgoff = page_to_pgoff(page);
 319         return vma->vm_start + ((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
 320 }
 321
 322 static inline unsigned long
 323 vma_address(struct page *page, struct vm_area_struct *vma)
 324 {
 325         unsigned long address = __vma_address(page, vma);
 326
 327         /* page should be within @vma mapping range */
 328         VM_BUG_ON_VMA(address < vma->vm_start || address >= vma->vm_end, vma);
 329
 330         return address;
 331 }
 332
 333 #else /* !CONFIG_MMU */
 334 static inline void clear_page_mlock(struct page *page) { }
 335 static inline void mlock_vma_page(struct page *page) { }
 336 static inline void mlock_migrate_page(struct page *new, struct page *old) { }
 337
 338 #endif /* !CONFIG_MMU */
 339
 340 /*
 341  * Return the mem_map entry representing the 'offset' subpage within
 342  * the maximally aligned gigantic page 'base'.  Handle any discontiguity
 343  * in the mem_map at MAX_ORDER_NR_PAGES boundaries.
 344  */
 345 static inline struct page *mem_map_offset(struct page *base, int offset)
 346 {
 347         if (unlikely(offset >= MAX_ORDER_NR_PAGES))
 348                 return nth_page(base, offset);
 349         return base + offset;
 350 }
 351
 352 /*
 353  * Iterator over all subpages within the maximally aligned gigantic
 354  * page 'base'.  Handle any discontiguity in the mem_map.
 355  */
 356 static inline struct page *mem_map_next(struct page *iter,
 357                                                 struct page *base, int offset)
 358 {
 359         if (unlikely((offset & (MAX_ORDER_NR_PAGES - 1)) == 0)) {
 360                 unsigned long pfn = page_to_pfn(base) + offset;
 361                 if (!pfn_valid(pfn))
 362                         return NULL;
 363                 return pfn_to_page(pfn);
 364         }
 365         return iter + 1;
 366 }
 367
 368 /*
 369  * FLATMEM and DISCONTIGMEM configurations use alloc_bootmem_node,
 370  * so all functions starting at paging_init should be marked __init
 371  * in those cases. SPARSEMEM, however, allows for memory hotplug,
 372  * and alloc_bootmem_node is not used.
 373  */
 374 #ifdef CONFIG_SPARSEMEM
 375 #define __paginginit __meminit
 376 #else
 377 #define __paginginit __init
 378 #endif
 379
 380 /* Memory initialisation debug and verification */
 381 enum mminit_level {
 382         MMINIT_WARNING,
 383         MMINIT_VERIFY,
 384         MMINIT_TRACE
 385 };
 386
 387 #ifdef CONFIG_DEBUG_MEMORY_INIT
 388
 389 extern int mminit_loglevel;
 390
 391 #define mminit_dprintk(level, prefix, fmt, arg...) \
 392 do { \
 393         if (level < mminit_loglevel) { \
 394                 if (level <= MMINIT_WARNING) \
 395                         pr_warn("mminit::" prefix " " fmt, ##arg);      \
 396                 else \
 397                         printk(KERN_DEBUG "mminit::" prefix " " fmt, ##arg); \
 398         } \
 399 } while (0)
 400
 401 extern void mminit_verify_pageflags_layout(void);
 402 extern void mminit_verify_zonelist(void);
 403 #else
 404
 405 static inline void mminit_dprintk(enum mminit_level level,
 406                                 const char *prefix, const char *fmt, ...)
 407 {
 408 }
 409
 410 static inline void mminit_verify_pageflags_layout(void)
 411 {
 412 }
 413
 414 static inline void mminit_verify_zonelist(void)
 415 {
 416 }
 417 #endif /* CONFIG_DEBUG_MEMORY_INIT */
 418
 419 /* mminit_validate_memmodel_limits is independent of CONFIG_DEBUG_MEMORY_INIT */
 420 #if defined(CONFIG_SPARSEMEM)
 421 extern void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 422                                 unsigned long *end_pfn);
 423 #else
 424 static inline void mminit_validate_memmodel_limits(unsigned long *start_pfn,
 425                                 unsigned long *end_pfn)
 426 {
 427 }
 428 #endif /* CONFIG_SPARSEMEM */
 429
 430 #define ZONE_RECLAIM_NOSCAN     -2
 431 #define ZONE_RECLAIM_FULL       -1
 432 #define ZONE_RECLAIM_SOME       0
 433 #define ZONE_RECLAIM_SUCCESS    1
 434
 435 extern int hwpoison_filter(struct page *p);
 436
 437 extern u32 hwpoison_filter_dev_major;
 438 extern u32 hwpoison_filter_dev_minor;
 439 extern u64 hwpoison_filter_flags_mask;
 440 extern u64 hwpoison_filter_flags_value;
 441 extern u64 hwpoison_filter_memcg;
 442 extern u32 hwpoison_filter_enable;
 443
 444 extern unsigned long vm_mmap_pgoff(struct file *, unsigned long,
 445         unsigned long, unsigned long,
 446         unsigned long, unsigned long);
 447
 448 extern void set_pageblock_order(void);
 449 unsigned long reclaim_clean_pages_from_list(struct zone *zone,
 450                                             struct list_head *page_list);
 451 /* The ALLOC_WMARK bits are used as an index to zone->watermark */
 452 #define ALLOC_WMARK_MIN         WMARK_MIN
 453 #define ALLOC_WMARK_LOW         WMARK_LOW
 454 #define ALLOC_WMARK_HIGH        WMARK_HIGH
 455 #define ALLOC_NO_WATERMARKS     0x04 /* don't check watermarks at all */
 456
 457 /* Mask to get the watermark bits */
 458 #define ALLOC_WMARK_MASK        (ALLOC_NO_WATERMARKS-1)
 459
 460 #define ALLOC_HARDER            0x10 /* try to alloc harder */
 461 #define ALLOC_HIGH              0x20 /* __GFP_HIGH set */
 462 #define ALLOC_CPUSET            0x40 /* check for correct cpuset */
 463 #define ALLOC_CMA               0x80 /* allow allocations from CMA areas */
 464 #define ALLOC_FAIR              0x100 /* fair zone allocation */
 465
 466 enum ttu_flags;
 467 struct tlbflush_unmap_batch;
 468
 469 #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
 470 void try_to_unmap_flush(void);
 471 void try_to_unmap_flush_dirty(void);
 472 #else
 473 static inline void try_to_unmap_flush(void)
 474 {
 475 }
 476 static inline void try_to_unmap_flush_dirty(void)
 477 {
 478 }
 479
 480 #endif /* CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH */
 481
 482 extern const struct trace_print_flags pageflag_names[];
 483 extern const struct trace_print_flags vmaflag_names[];
 484 extern const struct trace_print_flags gfpflag_names[];
 485
 486 #endif  /* __MM_INTERNAL_H */