linux/include/linux/mm.h
<<
>>
Prefs
   1#ifndef _LINUX_MM_H
   2#define _LINUX_MM_H
   3
   4#include <linux/errno.h>
   5
   6#ifdef __KERNEL__
   7
   8#include <linux/gfp.h>
   9#include <linux/bug.h>
  10#include <linux/list.h>
  11#include <linux/mmzone.h>
  12#include <linux/rbtree.h>
  13#include <linux/atomic.h>
  14#include <linux/debug_locks.h>
  15#include <linux/mm_types.h>
  16#include <linux/range.h>
  17#include <linux/pfn.h>
  18#include <linux/bit_spinlock.h>
  19#include <linux/shrinker.h>
  20
  21struct mempolicy;
  22struct anon_vma;
  23struct anon_vma_chain;
  24struct file_ra_state;
  25struct user_struct;
  26struct writeback_control;
  27
  28#ifndef CONFIG_DISCONTIGMEM          /* Don't use mapnrs, do it properly */
  29extern unsigned long max_mapnr;
  30#endif
  31
  32extern unsigned long num_physpages;
  33extern unsigned long totalram_pages;
  34extern void * high_memory;
  35extern int page_cluster;
  36
  37#ifdef CONFIG_SYSCTL
  38extern int sysctl_legacy_va_layout;
  39#else
  40#define sysctl_legacy_va_layout 0
  41#endif
  42
  43#include <asm/page.h>
  44#include <asm/pgtable.h>
  45#include <asm/processor.h>
  46
  47#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
  48
  49/* to align the pointer to the (next) page boundary */
  50#define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
  51
  52/*
  53 * Linux kernel virtual memory manager primitives.
  54 * The idea being to have a "virtual" mm in the same way
  55 * we have a virtual fs - giving a cleaner interface to the
  56 * mm details, and allowing different kinds of memory mappings
  57 * (from shared memory to executable loading to arbitrary
  58 * mmap() functions).
  59 */
  60
  61extern struct kmem_cache *vm_area_cachep;
  62
  63#ifndef CONFIG_MMU
  64extern struct rb_root nommu_region_tree;
  65extern struct rw_semaphore nommu_region_sem;
  66
  67extern unsigned int kobjsize(const void *objp);
  68#endif
  69
  70/*
  71 * vm_flags in vm_area_struct, see mm_types.h.
  72 */
  73#define VM_NONE         0x00000000
  74
  75#define VM_READ         0x00000001      /* currently active flags */
  76#define VM_WRITE        0x00000002
  77#define VM_EXEC         0x00000004
  78#define VM_SHARED       0x00000008
  79
  80/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
  81#define VM_MAYREAD      0x00000010      /* limits for mprotect() etc */
  82#define VM_MAYWRITE     0x00000020
  83#define VM_MAYEXEC      0x00000040
  84#define VM_MAYSHARE     0x00000080
  85
  86#define VM_GROWSDOWN    0x00000100      /* general info on the segment */
  87#define VM_PFNMAP       0x00000400      /* Page-ranges managed without "struct page", just pure PFN */
  88#define VM_DENYWRITE    0x00000800      /* ETXTBSY on write attempts.. */
  89
  90#define VM_LOCKED       0x00002000
  91#define VM_IO           0x00004000      /* Memory mapped I/O or similar */
  92
  93                                        /* Used by sys_madvise() */
  94#define VM_SEQ_READ     0x00008000      /* App will access data sequentially */
  95#define VM_RAND_READ    0x00010000      /* App will not benefit from clustered reads */
  96
  97#define VM_DONTCOPY     0x00020000      /* Do not copy this vma on fork */
  98#define VM_DONTEXPAND   0x00040000      /* Cannot expand with mremap() */
  99#define VM_ACCOUNT      0x00100000      /* Is a VM accounted object */
 100#define VM_NORESERVE    0x00200000      /* should the VM suppress accounting */
 101#define VM_HUGETLB      0x00400000      /* Huge TLB Page VM */
 102#define VM_NONLINEAR    0x00800000      /* Is non-linear (remap_file_pages) */
 103#define VM_ARCH_1       0x01000000      /* Architecture-specific flag */
 104#define VM_DONTDUMP     0x04000000      /* Do not include in the core dump */
 105
 106#define VM_MIXEDMAP     0x10000000      /* Can contain "struct page" and pure PFN pages */
 107#define VM_HUGEPAGE     0x20000000      /* MADV_HUGEPAGE marked this vma */
 108#define VM_NOHUGEPAGE   0x40000000      /* MADV_NOHUGEPAGE marked this vma */
 109#define VM_MERGEABLE    0x80000000      /* KSM may merge identical pages */
 110
 111#if defined(CONFIG_X86)
 112# define VM_PAT         VM_ARCH_1       /* PAT reserves whole VMA at once (x86) */
 113#elif defined(CONFIG_PPC)
 114# define VM_SAO         VM_ARCH_1       /* Strong Access Ordering (powerpc) */
 115#elif defined(CONFIG_PARISC)
 116# define VM_GROWSUP     VM_ARCH_1
 117#elif defined(CONFIG_IA64)
 118# define VM_GROWSUP     VM_ARCH_1
 119#elif !defined(CONFIG_MMU)
 120# define VM_MAPPED_COPY VM_ARCH_1       /* T if mapped copy of data (nommu mmap) */
 121#endif
 122
 123#ifndef VM_GROWSUP
 124# define VM_GROWSUP     VM_NONE
 125#endif
 126
 127/* Bits set in the VMA until the stack is in its final location */
 128#define VM_STACK_INCOMPLETE_SETUP       (VM_RAND_READ | VM_SEQ_READ)
 129
 130#ifndef VM_STACK_DEFAULT_FLAGS          /* arch can override this */
 131#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
 132#endif
 133
 134#ifdef CONFIG_STACK_GROWSUP
 135#define VM_STACK_FLAGS  (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
 136#else
 137#define VM_STACK_FLAGS  (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
 138#endif
 139
 140#define VM_READHINTMASK                 (VM_SEQ_READ | VM_RAND_READ)
 141#define VM_ClearReadHint(v)             (v)->vm_flags &= ~VM_READHINTMASK
 142#define VM_NormalReadHint(v)            (!((v)->vm_flags & VM_READHINTMASK))
 143#define VM_SequentialReadHint(v)        ((v)->vm_flags & VM_SEQ_READ)
 144#define VM_RandomReadHint(v)            ((v)->vm_flags & VM_RAND_READ)
 145
 146/*
 147 * Special vmas that are non-mergable, non-mlock()able.
 148 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
 149 */
 150#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP)
 151
 152/*
 153 * mapping from the currently active vm_flags protection bits (the
 154 * low four bits) to a page protection mask..
 155 */
 156extern pgprot_t protection_map[16];
 157
 158#define FAULT_FLAG_WRITE        0x01    /* Fault was a write access */
 159#define FAULT_FLAG_NONLINEAR    0x02    /* Fault was via a nonlinear mapping */
 160#define FAULT_FLAG_MKWRITE      0x04    /* Fault was mkwrite of existing pte */
 161#define FAULT_FLAG_ALLOW_RETRY  0x08    /* Retry fault if blocking */
 162#define FAULT_FLAG_RETRY_NOWAIT 0x10    /* Don't drop mmap_sem and wait when retrying */
 163#define FAULT_FLAG_KILLABLE     0x20    /* The fault task is in SIGKILL killable region */
 164#define FAULT_FLAG_TRIED        0x40    /* second try */
 165
 166/*
 167 * vm_fault is filled by the the pagefault handler and passed to the vma's
 168 * ->fault function. The vma's ->fault is responsible for returning a bitmask
 169 * of VM_FAULT_xxx flags that give details about how the fault was handled.
 170 *
 171 * pgoff should be used in favour of virtual_address, if possible. If pgoff
 172 * is used, one may implement ->remap_pages to get nonlinear mapping support.
 173 */
 174struct vm_fault {
 175        unsigned int flags;             /* FAULT_FLAG_xxx flags */
 176        pgoff_t pgoff;                  /* Logical page offset based on vma */
 177        void __user *virtual_address;   /* Faulting virtual address */
 178
 179        struct page *page;              /* ->fault handlers should return a
 180                                         * page here, unless VM_FAULT_NOPAGE
 181                                         * is set (which is also implied by
 182                                         * VM_FAULT_ERROR).
 183                                         */
 184};
 185
 186/*
 187 * These are the virtual MM functions - opening of an area, closing and
 188 * unmapping it (needed to keep files on disk up-to-date etc), pointer
 189 * to the functions called when a no-page or a wp-page exception occurs. 
 190 */
 191struct vm_operations_struct {
 192        void (*open)(struct vm_area_struct * area);
 193        void (*close)(struct vm_area_struct * area);
 194        int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
 195
 196        /* notification that a previously read-only page is about to become
 197         * writable, if an error is returned it will cause a SIGBUS */
 198        int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
 199
 200        /* called by access_process_vm when get_user_pages() fails, typically
 201         * for use by special VMAs that can switch between memory and hardware
 202         */
 203        int (*access)(struct vm_area_struct *vma, unsigned long addr,
 204                      void *buf, int len, int write);
 205#ifdef CONFIG_NUMA
 206        /*
 207         * set_policy() op must add a reference to any non-NULL @new mempolicy
 208         * to hold the policy upon return.  Caller should pass NULL @new to
 209         * remove a policy and fall back to surrounding context--i.e. do not
 210         * install a MPOL_DEFAULT policy, nor the task or system default
 211         * mempolicy.
 212         */
 213        int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
 214
 215        /*
 216         * get_policy() op must add reference [mpol_get()] to any policy at
 217         * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
 218         * in mm/mempolicy.c will do this automatically.
 219         * get_policy() must NOT add a ref if the policy at (vma,addr) is not
 220         * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
 221         * If no [shared/vma] mempolicy exists at the addr, get_policy() op
 222         * must return NULL--i.e., do not "fallback" to task or system default
 223         * policy.
 224         */
 225        struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
 226                                        unsigned long addr);
 227        int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
 228                const nodemask_t *to, unsigned long flags);
 229#endif
 230        /* called by sys_remap_file_pages() to populate non-linear mapping */
 231        int (*remap_pages)(struct vm_area_struct *vma, unsigned long addr,
 232                           unsigned long size, pgoff_t pgoff);
 233};
 234
 235struct mmu_gather;
 236struct inode;
 237
 238#define page_private(page)              ((page)->private)
 239#define set_page_private(page, v)       ((page)->private = (v))
 240
 241/* It's valid only if the page is free path or free_list */
 242static inline void set_freepage_migratetype(struct page *page, int migratetype)
 243{
 244        page->index = migratetype;
 245}
 246
 247/* It's valid only if the page is free path or free_list */
 248static inline int get_freepage_migratetype(struct page *page)
 249{
 250        return page->index;
 251}
 252
 253/*
 254 * FIXME: take this include out, include page-flags.h in
 255 * files which need it (119 of them)
 256 */
 257#include <linux/page-flags.h>
 258#include <linux/huge_mm.h>
 259
 260/*
 261 * Methods to modify the page usage count.
 262 *
 263 * What counts for a page usage:
 264 * - cache mapping   (page->mapping)
 265 * - private data    (page->private)
 266 * - page mapped in a task's page tables, each mapping
 267 *   is counted separately
 268 *
 269 * Also, many kernel routines increase the page count before a critical
 270 * routine so they can be sure the page doesn't go away from under them.
 271 */
 272
 273/*
 274 * Drop a ref, return true if the refcount fell to zero (the page has no users)
 275 */
 276static inline int put_page_testzero(struct page *page)
 277{
 278        VM_BUG_ON(atomic_read(&page->_count) == 0);
 279        return atomic_dec_and_test(&page->_count);
 280}
 281
 282/*
 283 * Try to grab a ref unless the page has a refcount of zero, return false if
 284 * that is the case.
 285 */
 286static inline int get_page_unless_zero(struct page *page)
 287{
 288        return atomic_inc_not_zero(&page->_count);
 289}
 290
 291extern int page_is_ram(unsigned long pfn);
 292
 293/* Support for virtually mapped pages */
 294struct page *vmalloc_to_page(const void *addr);
 295unsigned long vmalloc_to_pfn(const void *addr);
 296
 297/*
 298 * Determine if an address is within the vmalloc range
 299 *
 300 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
 301 * is no special casing required.
 302 */
 303static inline int is_vmalloc_addr(const void *x)
 304{
 305#ifdef CONFIG_MMU
 306        unsigned long addr = (unsigned long)x;
 307
 308        return addr >= VMALLOC_START && addr < VMALLOC_END;
 309#else
 310        return 0;
 311#endif
 312}
 313#ifdef CONFIG_MMU
 314extern int is_vmalloc_or_module_addr(const void *x);
 315#else
 316static inline int is_vmalloc_or_module_addr(const void *x)
 317{
 318        return 0;
 319}
 320#endif
 321
 322static inline void compound_lock(struct page *page)
 323{
 324#ifdef CONFIG_TRANSPARENT_HUGEPAGE
 325        VM_BUG_ON(PageSlab(page));
 326        bit_spin_lock(PG_compound_lock, &page->flags);
 327#endif
 328}
 329
 330static inline void compound_unlock(struct page *page)
 331{
 332#ifdef CONFIG_TRANSPARENT_HUGEPAGE
 333        VM_BUG_ON(PageSlab(page));
 334        bit_spin_unlock(PG_compound_lock, &page->flags);
 335#endif
 336}
 337
 338static inline unsigned long compound_lock_irqsave(struct page *page)
 339{
 340        unsigned long uninitialized_var(flags);
 341#ifdef CONFIG_TRANSPARENT_HUGEPAGE
 342        local_irq_save(flags);
 343        compound_lock(page);
 344#endif
 345        return flags;
 346}
 347
 348static inline void compound_unlock_irqrestore(struct page *page,
 349                                              unsigned long flags)
 350{
 351#ifdef CONFIG_TRANSPARENT_HUGEPAGE
 352        compound_unlock(page);
 353        local_irq_restore(flags);
 354#endif
 355}
 356
 357static inline struct page *compound_head(struct page *page)
 358{
 359        if (unlikely(PageTail(page)))
 360                return page->first_page;
 361        return page;
 362}
 363
 364/*
 365 * The atomic page->_mapcount, starts from -1: so that transitions
 366 * both from it and to it can be tracked, using atomic_inc_and_test
 367 * and atomic_add_negative(-1).
 368 */
 369static inline void reset_page_mapcount(struct page *page)
 370{
 371        atomic_set(&(page)->_mapcount, -1);
 372}
 373
 374static inline int page_mapcount(struct page *page)
 375{
 376        return atomic_read(&(page)->_mapcount) + 1;
 377}
 378
 379static inline int page_count(struct page *page)
 380{
 381        return atomic_read(&compound_head(page)->_count);
 382}
 383
 384static inline void get_huge_page_tail(struct page *page)
 385{
 386        /*
 387         * __split_huge_page_refcount() cannot run
 388         * from under us.
 389         */
 390        VM_BUG_ON(page_mapcount(page) < 0);
 391        VM_BUG_ON(atomic_read(&page->_count) != 0);
 392        atomic_inc(&page->_mapcount);
 393}
 394
 395extern bool __get_page_tail(struct page *page);
 396
 397static inline void get_page(struct page *page)
 398{
 399        if (unlikely(PageTail(page)))
 400                if (likely(__get_page_tail(page)))
 401                        return;
 402        /*
 403         * Getting a normal page or the head of a compound page
 404         * requires to already have an elevated page->_count.
 405         */
 406        VM_BUG_ON(atomic_read(&page->_count) <= 0);
 407        atomic_inc(&page->_count);
 408}
 409
 410static inline struct page *virt_to_head_page(const void *x)
 411{
 412        struct page *page = virt_to_page(x);
 413        return compound_head(page);
 414}
 415
 416/*
 417 * Setup the page count before being freed into the page allocator for
 418 * the first time (boot or memory hotplug)
 419 */
 420static inline void init_page_count(struct page *page)
 421{
 422        atomic_set(&page->_count, 1);
 423}
 424
 425/*
 426 * PageBuddy() indicate that the page is free and in the buddy system
 427 * (see mm/page_alloc.c).
 428 *
 429 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
 430 * -2 so that an underflow of the page_mapcount() won't be mistaken
 431 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
 432 * efficiently by most CPU architectures.
 433 */
 434#define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
 435
 436static inline int PageBuddy(struct page *page)
 437{
 438        return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
 439}
 440
 441static inline void __SetPageBuddy(struct page *page)
 442{
 443        VM_BUG_ON(atomic_read(&page->_mapcount) != -1);
 444        atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
 445}
 446
 447static inline void __ClearPageBuddy(struct page *page)
 448{
 449        VM_BUG_ON(!PageBuddy(page));
 450        atomic_set(&page->_mapcount, -1);
 451}
 452
 453void put_page(struct page *page);
 454void put_pages_list(struct list_head *pages);
 455
 456void split_page(struct page *page, unsigned int order);
 457int split_free_page(struct page *page);
 458
 459/*
 460 * Compound pages have a destructor function.  Provide a
 461 * prototype for that function and accessor functions.
 462 * These are _only_ valid on the head of a PG_compound page.
 463 */
 464typedef void compound_page_dtor(struct page *);
 465
 466static inline void set_compound_page_dtor(struct page *page,
 467                                                compound_page_dtor *dtor)
 468{
 469        page[1].lru.next = (void *)dtor;
 470}
 471
 472static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
 473{
 474        return (compound_page_dtor *)page[1].lru.next;
 475}
 476
 477static inline int compound_order(struct page *page)
 478{
 479        if (!PageHead(page))
 480                return 0;
 481        return (unsigned long)page[1].lru.prev;
 482}
 483
 484static inline int compound_trans_order(struct page *page)
 485{
 486        int order;
 487        unsigned long flags;
 488
 489        if (!PageHead(page))
 490                return 0;
 491
 492        flags = compound_lock_irqsave(page);
 493        order = compound_order(page);
 494        compound_unlock_irqrestore(page, flags);
 495        return order;
 496}
 497
 498static inline void set_compound_order(struct page *page, unsigned long order)
 499{
 500        page[1].lru.prev = (void *)order;
 501}
 502
 503#ifdef CONFIG_MMU
 504/*
 505 * Do pte_mkwrite, but only if the vma says VM_WRITE.  We do this when
 506 * servicing faults for write access.  In the normal case, do always want
 507 * pte_mkwrite.  But get_user_pages can cause write faults for mappings
 508 * that do not have writing enabled, when used by access_process_vm.
 509 */
 510static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
 511{
 512        if (likely(vma->vm_flags & VM_WRITE))
 513                pte = pte_mkwrite(pte);
 514        return pte;
 515}
 516#endif
 517
 518/*
 519 * Multiple processes may "see" the same page. E.g. for untouched
 520 * mappings of /dev/null, all processes see the same page full of
 521 * zeroes, and text pages of executables and shared libraries have
 522 * only one copy in memory, at most, normally.
 523 *
 524 * For the non-reserved pages, page_count(page) denotes a reference count.
 525 *   page_count() == 0 means the page is free. page->lru is then used for
 526 *   freelist management in the buddy allocator.
 527 *   page_count() > 0  means the page has been allocated.
 528 *
 529 * Pages are allocated by the slab allocator in order to provide memory
 530 * to kmalloc and kmem_cache_alloc. In this case, the management of the
 531 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
 532 * unless a particular usage is carefully commented. (the responsibility of
 533 * freeing the kmalloc memory is the caller's, of course).
 534 *
 535 * A page may be used by anyone else who does a __get_free_page().
 536 * In this case, page_count still tracks the references, and should only
 537 * be used through the normal accessor functions. The top bits of page->flags
 538 * and page->virtual store page management information, but all other fields
 539 * are unused and could be used privately, carefully. The management of this
 540 * page is the responsibility of the one who allocated it, and those who have
 541 * subsequently been given references to it.
 542 *
 543 * The other pages (we may call them "pagecache pages") are completely
 544 * managed by the Linux memory manager: I/O, buffers, swapping etc.
 545 * The following discussion applies only to them.
 546 *
 547 * A pagecache page contains an opaque `private' member, which belongs to the
 548 * page's address_space. Usually, this is the address of a circular list of
 549 * the page's disk buffers. PG_private must be set to tell the VM to call
 550 * into the filesystem to release these pages.
 551 *
 552 * A page may belong to an inode's memory mapping. In this case, page->mapping
 553 * is the pointer to the inode, and page->index is the file offset of the page,
 554 * in units of PAGE_CACHE_SIZE.
 555 *
 556 * If pagecache pages are not associated with an inode, they are said to be
 557 * anonymous pages. These may become associated with the swapcache, and in that
 558 * case PG_swapcache is set, and page->private is an offset into the swapcache.
 559 *
 560 * In either case (swapcache or inode backed), the pagecache itself holds one
 561 * reference to the page. Setting PG_private should also increment the
 562 * refcount. The each user mapping also has a reference to the page.
 563 *
 564 * The pagecache pages are stored in a per-mapping radix tree, which is
 565 * rooted at mapping->page_tree, and indexed by offset.
 566 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
 567 * lists, we instead now tag pages as dirty/writeback in the radix tree.
 568 *
 569 * All pagecache pages may be subject to I/O:
 570 * - inode pages may need to be read from disk,
 571 * - inode pages which have been modified and are MAP_SHARED may need
 572 *   to be written back to the inode on disk,
 573 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
 574 *   modified may need to be swapped out to swap space and (later) to be read
 575 *   back into memory.
 576 */
 577
 578/*
 579 * The zone field is never updated after free_area_init_core()
 580 * sets it, so none of the operations on it need to be atomic.
 581 */
 582
 583
 584/*
 585 * page->flags layout:
 586 *
 587 * There are three possibilities for how page->flags get
 588 * laid out.  The first is for the normal case, without
 589 * sparsemem.  The second is for sparsemem when there is
 590 * plenty of space for node and section.  The last is when
 591 * we have run out of space and have to fall back to an
 592 * alternate (slower) way of determining the node.
 593 *
 594 * No sparsemem or sparsemem vmemmap: |       NODE     | ZONE | ... | FLAGS |
 595 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
 596 * classic sparse no space for node:  | SECTION |     ZONE    | ... | FLAGS |
 597 */
 598#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
 599#define SECTIONS_WIDTH          SECTIONS_SHIFT
 600#else
 601#define SECTIONS_WIDTH          0
 602#endif
 603
 604#define ZONES_WIDTH             ZONES_SHIFT
 605
 606#if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
 607#define NODES_WIDTH             NODES_SHIFT
 608#else
 609#ifdef CONFIG_SPARSEMEM_VMEMMAP
 610#error "Vmemmap: No space for nodes field in page flags"
 611#endif
 612#define NODES_WIDTH             0
 613#endif
 614
 615/* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
 616#define SECTIONS_PGOFF          ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
 617#define NODES_PGOFF             (SECTIONS_PGOFF - NODES_WIDTH)
 618#define ZONES_PGOFF             (NODES_PGOFF - ZONES_WIDTH)
 619
 620/*
 621 * We are going to use the flags for the page to node mapping if its in
 622 * there.  This includes the case where there is no node, so it is implicit.
 623 */
 624#if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
 625#define NODE_NOT_IN_PAGE_FLAGS
 626#endif
 627
 628/*
 629 * Define the bit shifts to access each section.  For non-existent
 630 * sections we define the shift as 0; that plus a 0 mask ensures
 631 * the compiler will optimise away reference to them.
 632 */
 633#define SECTIONS_PGSHIFT        (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
 634#define NODES_PGSHIFT           (NODES_PGOFF * (NODES_WIDTH != 0))
 635#define ZONES_PGSHIFT           (ZONES_PGOFF * (ZONES_WIDTH != 0))
 636
 637/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
 638#ifdef NODE_NOT_IN_PAGE_FLAGS
 639#define ZONEID_SHIFT            (SECTIONS_SHIFT + ZONES_SHIFT)
 640#define ZONEID_PGOFF            ((SECTIONS_PGOFF < ZONES_PGOFF)? \
 641                                                SECTIONS_PGOFF : ZONES_PGOFF)
 642#else
 643#define ZONEID_SHIFT            (NODES_SHIFT + ZONES_SHIFT)
 644#define ZONEID_PGOFF            ((NODES_PGOFF < ZONES_PGOFF)? \
 645                                                NODES_PGOFF : ZONES_PGOFF)
 646#endif
 647
 648#define ZONEID_PGSHIFT          (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
 649
 650#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
 651#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
 652#endif
 653
 654#define ZONES_MASK              ((1UL << ZONES_WIDTH) - 1)
 655#define NODES_MASK              ((1UL << NODES_WIDTH) - 1)
 656#define SECTIONS_MASK           ((1UL << SECTIONS_WIDTH) - 1)
 657#define ZONEID_MASK             ((1UL << ZONEID_SHIFT) - 1)
 658
 659static inline enum zone_type page_zonenum(const struct page *page)
 660{
 661        return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
 662}
 663
 664/*
 665 * The identification function is only used by the buddy allocator for
 666 * determining if two pages could be buddies. We are not really
 667 * identifying a zone since we could be using a the section number
 668 * id if we have not node id available in page flags.
 669 * We guarantee only that it will return the same value for two
 670 * combinable pages in a zone.
 671 */
 672static inline int page_zone_id(struct page *page)
 673{
 674        return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
 675}
 676
 677static inline int zone_to_nid(struct zone *zone)
 678{
 679#ifdef CONFIG_NUMA
 680        return zone->node;
 681#else
 682        return 0;
 683#endif
 684}
 685
 686#ifdef NODE_NOT_IN_PAGE_FLAGS
 687extern int page_to_nid(const struct page *page);
 688#else
 689static inline int page_to_nid(const struct page *page)
 690{
 691        return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
 692}
 693#endif
 694
 695#ifdef CONFIG_NUMA_BALANCING
 696static inline int page_xchg_last_nid(struct page *page, int nid)
 697{
 698        return xchg(&page->_last_nid, nid);
 699}
 700
 701static inline int page_last_nid(struct page *page)
 702{
 703        return page->_last_nid;
 704}
 705static inline void reset_page_last_nid(struct page *page)
 706{
 707        page->_last_nid = -1;
 708}
 709#else
 710static inline int page_xchg_last_nid(struct page *page, int nid)
 711{
 712        return page_to_nid(page);
 713}
 714
 715static inline int page_last_nid(struct page *page)
 716{
 717        return page_to_nid(page);
 718}
 719
 720static inline void reset_page_last_nid(struct page *page)
 721{
 722}
 723#endif
 724
 725static inline struct zone *page_zone(const struct page *page)
 726{
 727        return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
 728}
 729
 730#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
 731static inline void set_page_section(struct page *page, unsigned long section)
 732{
 733        page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
 734        page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
 735}
 736
 737static inline unsigned long page_to_section(const struct page *page)
 738{
 739        return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
 740}
 741#endif
 742
 743static inline void set_page_zone(struct page *page, enum zone_type zone)
 744{
 745        page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
 746        page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
 747}
 748
 749static inline void set_page_node(struct page *page, unsigned long node)
 750{
 751        page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
 752        page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
 753}
 754
 755static inline void set_page_links(struct page *page, enum zone_type zone,
 756        unsigned long node, unsigned long pfn)
 757{
 758        set_page_zone(page, zone);
 759        set_page_node(page, node);
 760#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
 761        set_page_section(page, pfn_to_section_nr(pfn));
 762#endif
 763}
 764
 765/*
 766 * Some inline functions in vmstat.h depend on page_zone()
 767 */
 768#include <linux/vmstat.h>
 769
 770static __always_inline void *lowmem_page_address(const struct page *page)
 771{
 772        return __va(PFN_PHYS(page_to_pfn(page)));
 773}
 774
 775#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
 776#define HASHED_PAGE_VIRTUAL
 777#endif
 778
 779#if defined(WANT_PAGE_VIRTUAL)
 780#define page_address(page) ((page)->virtual)
 781#define set_page_address(page, address)                 \
 782        do {                                            \
 783                (page)->virtual = (address);            \
 784        } while(0)
 785#define page_address_init()  do { } while(0)
 786#endif
 787
 788#if defined(HASHED_PAGE_VIRTUAL)
 789void *page_address(const struct page *page);
 790void set_page_address(struct page *page, void *virtual);
 791void page_address_init(void);
 792#endif
 793
 794#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
 795#define page_address(page) lowmem_page_address(page)
 796#define set_page_address(page, address)  do { } while(0)
 797#define page_address_init()  do { } while(0)
 798#endif
 799
 800/*
 801 * On an anonymous page mapped into a user virtual memory area,
 802 * page->mapping points to its anon_vma, not to a struct address_space;
 803 * with the PAGE_MAPPING_ANON bit set to distinguish it.  See rmap.h.
 804 *
 805 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
 806 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
 807 * and then page->mapping points, not to an anon_vma, but to a private
 808 * structure which KSM associates with that merged page.  See ksm.h.
 809 *
 810 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
 811 *
 812 * Please note that, confusingly, "page_mapping" refers to the inode
 813 * address_space which maps the page from disk; whereas "page_mapped"
 814 * refers to user virtual address space into which the page is mapped.
 815 */
 816#define PAGE_MAPPING_ANON       1
 817#define PAGE_MAPPING_KSM        2
 818#define PAGE_MAPPING_FLAGS      (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
 819
 820extern struct address_space swapper_space;
 821static inline struct address_space *page_mapping(struct page *page)
 822{
 823        struct address_space *mapping = page->mapping;
 824
 825        VM_BUG_ON(PageSlab(page));
 826        if (unlikely(PageSwapCache(page)))
 827                mapping = &swapper_space;
 828        else if ((unsigned long)mapping & PAGE_MAPPING_ANON)
 829                mapping = NULL;
 830        return mapping;
 831}
 832
 833/* Neutral page->mapping pointer to address_space or anon_vma or other */
 834static inline void *page_rmapping(struct page *page)
 835{
 836        return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
 837}
 838
 839extern struct address_space *__page_file_mapping(struct page *);
 840
 841static inline
 842struct address_space *page_file_mapping(struct page *page)
 843{
 844        if (unlikely(PageSwapCache(page)))
 845                return __page_file_mapping(page);
 846
 847        return page->mapping;
 848}
 849
 850static inline int PageAnon(struct page *page)
 851{
 852        return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
 853}
 854
 855/*
 856 * Return the pagecache index of the passed page.  Regular pagecache pages
 857 * use ->index whereas swapcache pages use ->private
 858 */
 859static inline pgoff_t page_index(struct page *page)
 860{
 861        if (unlikely(PageSwapCache(page)))
 862                return page_private(page);
 863        return page->index;
 864}
 865
 866extern pgoff_t __page_file_index(struct page *page);
 867
 868/*
 869 * Return the file index of the page. Regular pagecache pages use ->index
 870 * whereas swapcache pages use swp_offset(->private)
 871 */
 872static inline pgoff_t page_file_index(struct page *page)
 873{
 874        if (unlikely(PageSwapCache(page)))
 875                return __page_file_index(page);
 876
 877        return page->index;
 878}
 879
 880/*
 881 * Return true if this page is mapped into pagetables.
 882 */
 883static inline int page_mapped(struct page *page)
 884{
 885        return atomic_read(&(page)->_mapcount) >= 0;
 886}
 887
 888/*
 889 * Different kinds of faults, as returned by handle_mm_fault().
 890 * Used to decide whether a process gets delivered SIGBUS or
 891 * just gets major/minor fault counters bumped up.
 892 */
 893
 894#define VM_FAULT_MINOR  0 /* For backwards compat. Remove me quickly. */
 895
 896#define VM_FAULT_OOM    0x0001
 897#define VM_FAULT_SIGBUS 0x0002
 898#define VM_FAULT_MAJOR  0x0004
 899#define VM_FAULT_WRITE  0x0008  /* Special case for get_user_pages */
 900#define VM_FAULT_HWPOISON 0x0010        /* Hit poisoned small page */
 901#define VM_FAULT_HWPOISON_LARGE 0x0020  /* Hit poisoned large page. Index encoded in upper bits */
 902
 903#define VM_FAULT_NOPAGE 0x0100  /* ->fault installed the pte, not return page */
 904#define VM_FAULT_LOCKED 0x0200  /* ->fault locked the returned page */
 905#define VM_FAULT_RETRY  0x0400  /* ->fault blocked, must retry */
 906
 907#define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
 908
 909#define VM_FAULT_ERROR  (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
 910                         VM_FAULT_HWPOISON_LARGE)
 911
 912/* Encode hstate index for a hwpoisoned large page */
 913#define VM_FAULT_SET_HINDEX(x) ((x) << 12)
 914#define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
 915
 916/*
 917 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
 918 */
 919extern void pagefault_out_of_memory(void);
 920
 921#define offset_in_page(p)       ((unsigned long)(p) & ~PAGE_MASK)
 922
 923/*
 924 * Flags passed to show_mem() and show_free_areas() to suppress output in
 925 * various contexts.
 926 */
 927#define SHOW_MEM_FILTER_NODES   (0x0001u)       /* filter disallowed nodes */
 928
 929extern void show_free_areas(unsigned int flags);
 930extern bool skip_free_areas_node(unsigned int flags, int nid);
 931
 932int shmem_zero_setup(struct vm_area_struct *);
 933
 934extern int can_do_mlock(void);
 935extern int user_shm_lock(size_t, struct user_struct *);
 936extern void user_shm_unlock(size_t, struct user_struct *);
 937
 938/*
 939 * Parameter block passed down to zap_pte_range in exceptional cases.
 940 */
 941struct zap_details {
 942        struct vm_area_struct *nonlinear_vma;   /* Check page->index if set */
 943        struct address_space *check_mapping;    /* Check page->mapping if set */
 944        pgoff_t first_index;                    /* Lowest page->index to unmap */
 945        pgoff_t last_index;                     /* Highest page->index to unmap */
 946};
 947
 948struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
 949                pte_t pte);
 950
 951int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
 952                unsigned long size);
 953void zap_page_range(struct vm_area_struct *vma, unsigned long address,
 954                unsigned long size, struct zap_details *);
 955void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
 956                unsigned long start, unsigned long end);
 957
 958/**
 959 * mm_walk - callbacks for walk_page_range
 960 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
 961 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
 962 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
 963 *             this handler is required to be able to handle
 964 *             pmd_trans_huge() pmds.  They may simply choose to
 965 *             split_huge_page() instead of handling it explicitly.
 966 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
 967 * @pte_hole: if set, called for each hole at all levels
 968 * @hugetlb_entry: if set, called for each hugetlb entry
 969 *                 *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
 970 *                            is used.
 971 *
 972 * (see walk_page_range for more details)
 973 */
 974struct mm_walk {
 975        int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
 976        int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
 977        int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
 978        int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
 979        int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
 980        int (*hugetlb_entry)(pte_t *, unsigned long,
 981                             unsigned long, unsigned long, struct mm_walk *);
 982        struct mm_struct *mm;
 983        void *private;
 984};
 985
 986int walk_page_range(unsigned long addr, unsigned long end,
 987                struct mm_walk *walk);
 988void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
 989                unsigned long end, unsigned long floor, unsigned long ceiling);
 990int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
 991                        struct vm_area_struct *vma);
 992void unmap_mapping_range(struct address_space *mapping,
 993                loff_t const holebegin, loff_t const holelen, int even_cows);
 994int follow_pfn(struct vm_area_struct *vma, unsigned long address,
 995        unsigned long *pfn);
 996int follow_phys(struct vm_area_struct *vma, unsigned long address,
 997                unsigned int flags, unsigned long *prot, resource_size_t *phys);
 998int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
 999                        void *buf, int len, int write);
1000
1001static inline void unmap_shared_mapping_range(struct address_space *mapping,
1002                loff_t const holebegin, loff_t const holelen)
1003{
1004        unmap_mapping_range(mapping, holebegin, holelen, 0);
1005}
1006
1007extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new);
1008extern void truncate_setsize(struct inode *inode, loff_t newsize);
1009void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1010int truncate_inode_page(struct address_space *mapping, struct page *page);
1011int generic_error_remove_page(struct address_space *mapping, struct page *page);
1012int invalidate_inode_page(struct page *page);
1013
1014#ifdef CONFIG_MMU
1015extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1016                        unsigned long address, unsigned int flags);
1017extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1018                            unsigned long address, unsigned int fault_flags);
1019#else
1020static inline int handle_mm_fault(struct mm_struct *mm,
1021                        struct vm_area_struct *vma, unsigned long address,
1022                        unsigned int flags)
1023{
1024        /* should never happen if there's no MMU */
1025        BUG();
1026        return VM_FAULT_SIGBUS;
1027}
1028static inline int fixup_user_fault(struct task_struct *tsk,
1029                struct mm_struct *mm, unsigned long address,
1030                unsigned int fault_flags)
1031{
1032        /* should never happen if there's no MMU */
1033        BUG();
1034        return -EFAULT;
1035}
1036#endif
1037
1038extern int make_pages_present(unsigned long addr, unsigned long end);
1039extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1040extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1041                void *buf, int len, int write);
1042
1043int __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1044                     unsigned long start, int len, unsigned int foll_flags,
1045                     struct page **pages, struct vm_area_struct **vmas,
1046                     int *nonblocking);
1047int get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1048                        unsigned long start, int nr_pages, int write, int force,
1049                        struct page **pages, struct vm_area_struct **vmas);
1050int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1051                        struct page **pages);
1052struct kvec;
1053int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1054                        struct page **pages);
1055int get_kernel_page(unsigned long start, int write, struct page **pages);
1056struct page *get_dump_page(unsigned long addr);
1057
1058extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1059extern void do_invalidatepage(struct page *page, unsigned long offset);
1060
1061int __set_page_dirty_nobuffers(struct page *page);
1062int __set_page_dirty_no_writeback(struct page *page);
1063int redirty_page_for_writepage(struct writeback_control *wbc,
1064                                struct page *page);
1065void account_page_dirtied(struct page *page, struct address_space *mapping);
1066void account_page_writeback(struct page *page);
1067int set_page_dirty(struct page *page);
1068int set_page_dirty_lock(struct page *page);
1069int clear_page_dirty_for_io(struct page *page);
1070
1071/* Is the vma a continuation of the stack vma above it? */
1072static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1073{
1074        return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1075}
1076
1077static inline int stack_guard_page_start(struct vm_area_struct *vma,
1078                                             unsigned long addr)
1079{
1080        return (vma->vm_flags & VM_GROWSDOWN) &&
1081                (vma->vm_start == addr) &&
1082                !vma_growsdown(vma->vm_prev, addr);
1083}
1084
1085/* Is the vma a continuation of the stack vma below it? */
1086static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1087{
1088        return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1089}
1090
1091static inline int stack_guard_page_end(struct vm_area_struct *vma,
1092                                           unsigned long addr)
1093{
1094        return (vma->vm_flags & VM_GROWSUP) &&
1095                (vma->vm_end == addr) &&
1096                !vma_growsup(vma->vm_next, addr);
1097}
1098
1099extern pid_t
1100vm_is_stack(struct task_struct *task, struct vm_area_struct *vma, int in_group);
1101
1102extern unsigned long move_page_tables(struct vm_area_struct *vma,
1103                unsigned long old_addr, struct vm_area_struct *new_vma,
1104                unsigned long new_addr, unsigned long len,
1105                bool need_rmap_locks);
1106extern unsigned long do_mremap(unsigned long addr,
1107                               unsigned long old_len, unsigned long new_len,
1108                               unsigned long flags, unsigned long new_addr);
1109extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1110                              unsigned long end, pgprot_t newprot,
1111                              int dirty_accountable, int prot_numa);
1112extern int mprotect_fixup(struct vm_area_struct *vma,
1113                          struct vm_area_struct **pprev, unsigned long start,
1114                          unsigned long end, unsigned long newflags);
1115
1116/*
1117 * doesn't attempt to fault and will return short.
1118 */
1119int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1120                          struct page **pages);
1121/*
1122 * per-process(per-mm_struct) statistics.
1123 */
1124static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1125{
1126        long val = atomic_long_read(&mm->rss_stat.count[member]);
1127
1128#ifdef SPLIT_RSS_COUNTING
1129        /*
1130         * counter is updated in asynchronous manner and may go to minus.
1131         * But it's never be expected number for users.
1132         */
1133        if (val < 0)
1134                val = 0;
1135#endif
1136        return (unsigned long)val;
1137}
1138
1139static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1140{
1141        atomic_long_add(value, &mm->rss_stat.count[member]);
1142}
1143
1144static inline void inc_mm_counter(struct mm_struct *mm, int member)
1145{
1146        atomic_long_inc(&mm->rss_stat.count[member]);
1147}
1148
1149static inline void dec_mm_counter(struct mm_struct *mm, int member)
1150{
1151        atomic_long_dec(&mm->rss_stat.count[member]);
1152}
1153
1154static inline unsigned long get_mm_rss(struct mm_struct *mm)
1155{
1156        return get_mm_counter(mm, MM_FILEPAGES) +
1157                get_mm_counter(mm, MM_ANONPAGES);
1158}
1159
1160static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1161{
1162        return max(mm->hiwater_rss, get_mm_rss(mm));
1163}
1164
1165static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1166{
1167        return max(mm->hiwater_vm, mm->total_vm);
1168}
1169
1170static inline void update_hiwater_rss(struct mm_struct *mm)
1171{
1172        unsigned long _rss = get_mm_rss(mm);
1173
1174        if ((mm)->hiwater_rss < _rss)
1175                (mm)->hiwater_rss = _rss;
1176}
1177
1178static inline void update_hiwater_vm(struct mm_struct *mm)
1179{
1180        if (mm->hiwater_vm < mm->total_vm)
1181                mm->hiwater_vm = mm->total_vm;
1182}
1183
1184static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1185                                         struct mm_struct *mm)
1186{
1187        unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1188
1189        if (*maxrss < hiwater_rss)
1190                *maxrss = hiwater_rss;
1191}
1192
1193#if defined(SPLIT_RSS_COUNTING)
1194void sync_mm_rss(struct mm_struct *mm);
1195#else
1196static inline void sync_mm_rss(struct mm_struct *mm)
1197{
1198}
1199#endif
1200
1201int vma_wants_writenotify(struct vm_area_struct *vma);
1202
1203extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1204                               spinlock_t **ptl);
1205static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1206                                    spinlock_t **ptl)
1207{
1208        pte_t *ptep;
1209        __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1210        return ptep;
1211}
1212
1213#ifdef __PAGETABLE_PUD_FOLDED
1214static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1215                                                unsigned long address)
1216{
1217        return 0;
1218}
1219#else
1220int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1221#endif
1222
1223#ifdef __PAGETABLE_PMD_FOLDED
1224static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1225                                                unsigned long address)
1226{
1227        return 0;
1228}
1229#else
1230int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1231#endif
1232
1233int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1234                pmd_t *pmd, unsigned long address);
1235int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1236
1237/*
1238 * The following ifdef needed to get the 4level-fixup.h header to work.
1239 * Remove it when 4level-fixup.h has been removed.
1240 */
1241#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1242static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1243{
1244        return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1245                NULL: pud_offset(pgd, address);
1246}
1247
1248static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1249{
1250        return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1251                NULL: pmd_offset(pud, address);
1252}
1253#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1254
1255#if USE_SPLIT_PTLOCKS
1256/*
1257 * We tuck a spinlock to guard each pagetable page into its struct page,
1258 * at page->private, with BUILD_BUG_ON to make sure that this will not
1259 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1260 * When freeing, reset page->mapping so free_pages_check won't complain.
1261 */
1262#define __pte_lockptr(page)     &((page)->ptl)
1263#define pte_lock_init(_page)    do {                                    \
1264        spin_lock_init(__pte_lockptr(_page));                           \
1265} while (0)
1266#define pte_lock_deinit(page)   ((page)->mapping = NULL)
1267#define pte_lockptr(mm, pmd)    ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
1268#else   /* !USE_SPLIT_PTLOCKS */
1269/*
1270 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1271 */
1272#define pte_lock_init(page)     do {} while (0)
1273#define pte_lock_deinit(page)   do {} while (0)
1274#define pte_lockptr(mm, pmd)    ({(void)(pmd); &(mm)->page_table_lock;})
1275#endif /* USE_SPLIT_PTLOCKS */
1276
1277static inline void pgtable_page_ctor(struct page *page)
1278{
1279        pte_lock_init(page);
1280        inc_zone_page_state(page, NR_PAGETABLE);
1281}
1282
1283static inline void pgtable_page_dtor(struct page *page)
1284{
1285        pte_lock_deinit(page);
1286        dec_zone_page_state(page, NR_PAGETABLE);
1287}
1288
1289#define pte_offset_map_lock(mm, pmd, address, ptlp)     \
1290({                                                      \
1291        spinlock_t *__ptl = pte_lockptr(mm, pmd);       \
1292        pte_t *__pte = pte_offset_map(pmd, address);    \
1293        *(ptlp) = __ptl;                                \
1294        spin_lock(__ptl);                               \
1295        __pte;                                          \
1296})
1297
1298#define pte_unmap_unlock(pte, ptl)      do {            \
1299        spin_unlock(ptl);                               \
1300        pte_unmap(pte);                                 \
1301} while (0)
1302
1303#define pte_alloc_map(mm, vma, pmd, address)                            \
1304        ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma,    \
1305                                                        pmd, address))? \
1306         NULL: pte_offset_map(pmd, address))
1307
1308#define pte_alloc_map_lock(mm, pmd, address, ptlp)      \
1309        ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL,   \
1310                                                        pmd, address))? \
1311                NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1312
1313#define pte_alloc_kernel(pmd, address)                  \
1314        ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1315                NULL: pte_offset_kernel(pmd, address))
1316
1317extern void free_area_init(unsigned long * zones_size);
1318extern void free_area_init_node(int nid, unsigned long * zones_size,
1319                unsigned long zone_start_pfn, unsigned long *zholes_size);
1320extern void free_initmem(void);
1321
1322#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1323/*
1324 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1325 * zones, allocate the backing mem_map and account for memory holes in a more
1326 * architecture independent manner. This is a substitute for creating the
1327 * zone_sizes[] and zholes_size[] arrays and passing them to
1328 * free_area_init_node()
1329 *
1330 * An architecture is expected to register range of page frames backed by
1331 * physical memory with memblock_add[_node]() before calling
1332 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1333 * usage, an architecture is expected to do something like
1334 *
1335 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1336 *                                                       max_highmem_pfn};
1337 * for_each_valid_physical_page_range()
1338 *      memblock_add_node(base, size, nid)
1339 * free_area_init_nodes(max_zone_pfns);
1340 *
1341 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1342 * registered physical page range.  Similarly
1343 * sparse_memory_present_with_active_regions() calls memory_present() for
1344 * each range when SPARSEMEM is enabled.
1345 *
1346 * See mm/page_alloc.c for more information on each function exposed by
1347 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1348 */
1349extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1350unsigned long node_map_pfn_alignment(void);
1351unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1352                                                unsigned long end_pfn);
1353extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1354                                                unsigned long end_pfn);
1355extern void get_pfn_range_for_nid(unsigned int nid,
1356                        unsigned long *start_pfn, unsigned long *end_pfn);
1357extern unsigned long find_min_pfn_with_active_regions(void);
1358extern void free_bootmem_with_active_regions(int nid,
1359                                                unsigned long max_low_pfn);
1360extern void sparse_memory_present_with_active_regions(int nid);
1361
1362#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1363
1364#if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1365    !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1366static inline int __early_pfn_to_nid(unsigned long pfn)
1367{
1368        return 0;
1369}
1370#else
1371/* please see mm/page_alloc.c */
1372extern int __meminit early_pfn_to_nid(unsigned long pfn);
1373#ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1374/* there is a per-arch backend function. */
1375extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1376#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1377#endif
1378
1379extern void set_dma_reserve(unsigned long new_dma_reserve);
1380extern void memmap_init_zone(unsigned long, int, unsigned long,
1381                                unsigned long, enum memmap_context);
1382extern void setup_per_zone_wmarks(void);
1383extern int __meminit init_per_zone_wmark_min(void);
1384extern void mem_init(void);
1385extern void __init mmap_init(void);
1386extern void show_mem(unsigned int flags);
1387extern void si_meminfo(struct sysinfo * val);
1388extern void si_meminfo_node(struct sysinfo *val, int nid);
1389extern int after_bootmem;
1390
1391extern __printf(3, 4)
1392void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1393
1394extern void setup_per_cpu_pageset(void);
1395
1396extern void zone_pcp_update(struct zone *zone);
1397extern void zone_pcp_reset(struct zone *zone);
1398
1399/* nommu.c */
1400extern atomic_long_t mmap_pages_allocated;
1401extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1402
1403/* interval_tree.c */
1404void vma_interval_tree_insert(struct vm_area_struct *node,
1405                              struct rb_root *root);
1406void vma_interval_tree_insert_after(struct vm_area_struct *node,
1407                                    struct vm_area_struct *prev,
1408                                    struct rb_root *root);
1409void vma_interval_tree_remove(struct vm_area_struct *node,
1410                              struct rb_root *root);
1411struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1412                                unsigned long start, unsigned long last);
1413struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1414                                unsigned long start, unsigned long last);
1415
1416#define vma_interval_tree_foreach(vma, root, start, last)               \
1417        for (vma = vma_interval_tree_iter_first(root, start, last);     \
1418             vma; vma = vma_interval_tree_iter_next(vma, start, last))
1419
1420static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1421                                        struct list_head *list)
1422{
1423        list_add_tail(&vma->shared.nonlinear, list);
1424}
1425
1426void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1427                                   struct rb_root *root);
1428void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1429                                   struct rb_root *root);
1430struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1431        struct rb_root *root, unsigned long start, unsigned long last);
1432struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1433        struct anon_vma_chain *node, unsigned long start, unsigned long last);
1434#ifdef CONFIG_DEBUG_VM_RB
1435void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1436#endif
1437
1438#define anon_vma_interval_tree_foreach(avc, root, start, last)           \
1439        for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1440             avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1441
1442/* mmap.c */
1443extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1444extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1445        unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1446extern struct vm_area_struct *vma_merge(struct mm_struct *,
1447        struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1448        unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1449        struct mempolicy *);
1450extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1451extern int split_vma(struct mm_struct *,
1452        struct vm_area_struct *, unsigned long addr, int new_below);
1453extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1454extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1455        struct rb_node **, struct rb_node *);
1456extern void unlink_file_vma(struct vm_area_struct *);
1457extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1458        unsigned long addr, unsigned long len, pgoff_t pgoff,
1459        bool *need_rmap_locks);
1460extern void exit_mmap(struct mm_struct *);
1461
1462extern int mm_take_all_locks(struct mm_struct *mm);
1463extern void mm_drop_all_locks(struct mm_struct *mm);
1464
1465extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1466extern struct file *get_mm_exe_file(struct mm_struct *mm);
1467
1468extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1469extern int install_special_mapping(struct mm_struct *mm,
1470                                   unsigned long addr, unsigned long len,
1471                                   unsigned long flags, struct page **pages);
1472
1473extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1474
1475extern unsigned long mmap_region(struct file *file, unsigned long addr,
1476        unsigned long len, unsigned long flags,
1477        vm_flags_t vm_flags, unsigned long pgoff);
1478extern unsigned long do_mmap_pgoff(struct file *, unsigned long,
1479        unsigned long, unsigned long,
1480        unsigned long, unsigned long);
1481extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1482
1483/* These take the mm semaphore themselves */
1484extern unsigned long vm_brk(unsigned long, unsigned long);
1485extern int vm_munmap(unsigned long, size_t);
1486extern unsigned long vm_mmap(struct file *, unsigned long,
1487        unsigned long, unsigned long,
1488        unsigned long, unsigned long);
1489
1490struct vm_unmapped_area_info {
1491#define VM_UNMAPPED_AREA_TOPDOWN 1
1492        unsigned long flags;
1493        unsigned long length;
1494        unsigned long low_limit;
1495        unsigned long high_limit;
1496        unsigned long align_mask;
1497        unsigned long align_offset;
1498};
1499
1500extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
1501extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
1502
1503/*
1504 * Search for an unmapped address range.
1505 *
1506 * We are looking for a range that:
1507 * - does not intersect with any VMA;
1508 * - is contained within the [low_limit, high_limit) interval;
1509 * - is at least the desired size.
1510 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1511 */
1512static inline unsigned long
1513vm_unmapped_area(struct vm_unmapped_area_info *info)
1514{
1515        if (!(info->flags & VM_UNMAPPED_AREA_TOPDOWN))
1516                return unmapped_area(info);
1517        else
1518                return unmapped_area_topdown(info);
1519}
1520
1521/* truncate.c */
1522extern void truncate_inode_pages(struct address_space *, loff_t);
1523extern void truncate_inode_pages_range(struct address_space *,
1524                                       loff_t lstart, loff_t lend);
1525
1526/* generic vm_area_ops exported for stackable file systems */
1527extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1528extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1529
1530/* mm/page-writeback.c */
1531int write_one_page(struct page *page, int wait);
1532void task_dirty_inc(struct task_struct *tsk);
1533
1534/* readahead.c */
1535#define VM_MAX_READAHEAD        128     /* kbytes */
1536#define VM_MIN_READAHEAD        16      /* kbytes (includes current page) */
1537
1538int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1539                        pgoff_t offset, unsigned long nr_to_read);
1540
1541void page_cache_sync_readahead(struct address_space *mapping,
1542                               struct file_ra_state *ra,
1543                               struct file *filp,
1544                               pgoff_t offset,
1545                               unsigned long size);
1546
1547void page_cache_async_readahead(struct address_space *mapping,
1548                                struct file_ra_state *ra,
1549                                struct file *filp,
1550                                struct page *pg,
1551                                pgoff_t offset,
1552                                unsigned long size);
1553
1554unsigned long max_sane_readahead(unsigned long nr);
1555unsigned long ra_submit(struct file_ra_state *ra,
1556                        struct address_space *mapping,
1557                        struct file *filp);
1558
1559/* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1560extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1561
1562/* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1563extern int expand_downwards(struct vm_area_struct *vma,
1564                unsigned long address);
1565#if VM_GROWSUP
1566extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1567#else
1568  #define expand_upwards(vma, address) do { } while (0)
1569#endif
1570
1571/* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
1572extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1573extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1574                                             struct vm_area_struct **pprev);
1575
1576/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1577   NULL if none.  Assume start_addr < end_addr. */
1578static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1579{
1580        struct vm_area_struct * vma = find_vma(mm,start_addr);
1581
1582        if (vma && end_addr <= vma->vm_start)
1583                vma = NULL;
1584        return vma;
1585}
1586
1587static inline unsigned long vma_pages(struct vm_area_struct *vma)
1588{
1589        return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1590}
1591
1592/* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1593static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
1594                                unsigned long vm_start, unsigned long vm_end)
1595{
1596        struct vm_area_struct *vma = find_vma(mm, vm_start);
1597
1598        if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
1599                vma = NULL;
1600
1601        return vma;
1602}
1603
1604#ifdef CONFIG_MMU
1605pgprot_t vm_get_page_prot(unsigned long vm_flags);
1606#else
1607static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1608{
1609        return __pgprot(0);
1610}
1611#endif
1612
1613#ifdef CONFIG_ARCH_USES_NUMA_PROT_NONE
1614unsigned long change_prot_numa(struct vm_area_struct *vma,
1615                        unsigned long start, unsigned long end);
1616#endif
1617
1618struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1619int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1620                        unsigned long pfn, unsigned long size, pgprot_t);
1621int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1622int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1623                        unsigned long pfn);
1624int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1625                        unsigned long pfn);
1626
1627struct page *follow_page(struct vm_area_struct *, unsigned long address,
1628                        unsigned int foll_flags);
1629#define FOLL_WRITE      0x01    /* check pte is writable */
1630#define FOLL_TOUCH      0x02    /* mark page accessed */
1631#define FOLL_GET        0x04    /* do get_page on page */
1632#define FOLL_DUMP       0x08    /* give error on hole if it would be zero */
1633#define FOLL_FORCE      0x10    /* get_user_pages read/write w/o permission */
1634#define FOLL_NOWAIT     0x20    /* if a disk transfer is needed, start the IO
1635                                 * and return without waiting upon it */
1636#define FOLL_MLOCK      0x40    /* mark page as mlocked */
1637#define FOLL_SPLIT      0x80    /* don't return transhuge pages, split them */
1638#define FOLL_HWPOISON   0x100   /* check page is hwpoisoned */
1639#define FOLL_NUMA       0x200   /* force NUMA hinting page fault */
1640
1641typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1642                        void *data);
1643extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1644                               unsigned long size, pte_fn_t fn, void *data);
1645
1646#ifdef CONFIG_PROC_FS
1647void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1648#else
1649static inline void vm_stat_account(struct mm_struct *mm,
1650                        unsigned long flags, struct file *file, long pages)
1651{
1652        mm->total_vm += pages;
1653}
1654#endif /* CONFIG_PROC_FS */
1655
1656#ifdef CONFIG_DEBUG_PAGEALLOC
1657extern void kernel_map_pages(struct page *page, int numpages, int enable);
1658#ifdef CONFIG_HIBERNATION
1659extern bool kernel_page_present(struct page *page);
1660#endif /* CONFIG_HIBERNATION */
1661#else
1662static inline void
1663kernel_map_pages(struct page *page, int numpages, int enable) {}
1664#ifdef CONFIG_HIBERNATION
1665static inline bool kernel_page_present(struct page *page) { return true; }
1666#endif /* CONFIG_HIBERNATION */
1667#endif
1668
1669extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
1670#ifdef  __HAVE_ARCH_GATE_AREA
1671int in_gate_area_no_mm(unsigned long addr);
1672int in_gate_area(struct mm_struct *mm, unsigned long addr);
1673#else
1674int in_gate_area_no_mm(unsigned long addr);
1675#define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
1676#endif  /* __HAVE_ARCH_GATE_AREA */
1677
1678int drop_caches_sysctl_handler(struct ctl_table *, int,
1679                                        void __user *, size_t *, loff_t *);
1680unsigned long shrink_slab(struct shrink_control *shrink,
1681                          unsigned long nr_pages_scanned,
1682                          unsigned long lru_pages);
1683
1684#ifndef CONFIG_MMU
1685#define randomize_va_space 0
1686#else
1687extern int randomize_va_space;
1688#endif
1689
1690const char * arch_vma_name(struct vm_area_struct *vma);
1691void print_vma_addr(char *prefix, unsigned long rip);
1692
1693void sparse_mem_maps_populate_node(struct page **map_map,
1694                                   unsigned long pnum_begin,
1695                                   unsigned long pnum_end,
1696                                   unsigned long map_count,
1697                                   int nodeid);
1698
1699struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1700pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1701pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1702pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1703pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1704void *vmemmap_alloc_block(unsigned long size, int node);
1705void *vmemmap_alloc_block_buf(unsigned long size, int node);
1706void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1707int vmemmap_populate_basepages(struct page *start_page,
1708                                                unsigned long pages, int node);
1709int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
1710void vmemmap_populate_print_last(void);
1711
1712
1713enum mf_flags {
1714        MF_COUNT_INCREASED = 1 << 0,
1715        MF_ACTION_REQUIRED = 1 << 1,
1716        MF_MUST_KILL = 1 << 2,
1717};
1718extern int memory_failure(unsigned long pfn, int trapno, int flags);
1719extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
1720extern int unpoison_memory(unsigned long pfn);
1721extern int sysctl_memory_failure_early_kill;
1722extern int sysctl_memory_failure_recovery;
1723extern void shake_page(struct page *p, int access);
1724extern atomic_long_t mce_bad_pages;
1725extern int soft_offline_page(struct page *page, int flags);
1726
1727extern void dump_page(struct page *page);
1728
1729#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1730extern void clear_huge_page(struct page *page,
1731                            unsigned long addr,
1732                            unsigned int pages_per_huge_page);
1733extern void copy_user_huge_page(struct page *dst, struct page *src,
1734                                unsigned long addr, struct vm_area_struct *vma,
1735                                unsigned int pages_per_huge_page);
1736#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1737
1738#ifdef CONFIG_DEBUG_PAGEALLOC
1739extern unsigned int _debug_guardpage_minorder;
1740
1741static inline unsigned int debug_guardpage_minorder(void)
1742{
1743        return _debug_guardpage_minorder;
1744}
1745
1746static inline bool page_is_guard(struct page *page)
1747{
1748        return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags);
1749}
1750#else
1751static inline unsigned int debug_guardpage_minorder(void) { return 0; }
1752static inline bool page_is_guard(struct page *page) { return false; }
1753#endif /* CONFIG_DEBUG_PAGEALLOC */
1754
1755#endif /* __KERNEL__ */
1756#endif /* _LINUX_MM_H */
1757
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.