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_NEED_MULTIPLE_NODES      /* Don't use mapnrs, do it properly */
  29extern unsigned long max_mapnr;
  30
  31static inline void set_max_mapnr(unsigned long limit)
  32{
  33        max_mapnr = limit;
  34}
  35#else
  36static inline void set_max_mapnr(unsigned long limit) { }
  37#endif
  38
  39extern unsigned long totalram_pages;
  40extern void * high_memory;
  41extern int page_cluster;
  42
  43#ifdef CONFIG_SYSCTL
  44extern int sysctl_legacy_va_layout;
  45#else
  46#define sysctl_legacy_va_layout 0
  47#endif
  48
  49#include <asm/page.h>
  50#include <asm/pgtable.h>
  51#include <asm/processor.h>
  52
  53extern unsigned long sysctl_user_reserve_kbytes;
  54extern unsigned long sysctl_admin_reserve_kbytes;
  55
  56#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
  57
  58/* to align the pointer to the (next) page boundary */
  59#define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
  60
  61/* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
  62#define PAGE_ALIGNED(addr)      IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
  63
  64/*
  65 * Linux kernel virtual memory manager primitives.
  66 * The idea being to have a "virtual" mm in the same way
  67 * we have a virtual fs - giving a cleaner interface to the
  68 * mm details, and allowing different kinds of memory mappings
  69 * (from shared memory to executable loading to arbitrary
  70 * mmap() functions).
  71 */
  72
  73extern struct kmem_cache *vm_area_cachep;
  74
  75#ifndef CONFIG_MMU
  76extern struct rb_root nommu_region_tree;
  77extern struct rw_semaphore nommu_region_sem;
  78
  79extern unsigned int kobjsize(const void *objp);
  80#endif
  81
  82/*
  83 * vm_flags in vm_area_struct, see mm_types.h.
  84 */
  85#define VM_NONE         0x00000000
  86
  87#define VM_READ         0x00000001      /* currently active flags */
  88#define VM_WRITE        0x00000002
  89#define VM_EXEC         0x00000004
  90#define VM_SHARED       0x00000008
  91
  92/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
  93#define VM_MAYREAD      0x00000010      /* limits for mprotect() etc */
  94#define VM_MAYWRITE     0x00000020
  95#define VM_MAYEXEC      0x00000040
  96#define VM_MAYSHARE     0x00000080
  97
  98#define VM_GROWSDOWN    0x00000100      /* general info on the segment */
  99#define VM_PFNMAP       0x00000400      /* Page-ranges managed without "struct page", just pure PFN */
 100#define VM_DENYWRITE    0x00000800      /* ETXTBSY on write attempts.. */
 101
 102#define VM_LOCKED       0x00002000
 103#define VM_IO           0x00004000      /* Memory mapped I/O or similar */
 104
 105                                        /* Used by sys_madvise() */
 106#define VM_SEQ_READ     0x00008000      /* App will access data sequentially */
 107#define VM_RAND_READ    0x00010000      /* App will not benefit from clustered reads */
 108
 109#define VM_DONTCOPY     0x00020000      /* Do not copy this vma on fork */
 110#define VM_DONTEXPAND   0x00040000      /* Cannot expand with mremap() */
 111#define VM_ACCOUNT      0x00100000      /* Is a VM accounted object */
 112#define VM_NORESERVE    0x00200000      /* should the VM suppress accounting */
 113#define VM_HUGETLB      0x00400000      /* Huge TLB Page VM */
 114#define VM_NONLINEAR    0x00800000      /* Is non-linear (remap_file_pages) */
 115#define VM_ARCH_1       0x01000000      /* Architecture-specific flag */
 116#define VM_DONTDUMP     0x04000000      /* Do not include in the core dump */
 117
 118#define VM_MIXEDMAP     0x10000000      /* Can contain "struct page" and pure PFN pages */
 119#define VM_HUGEPAGE     0x20000000      /* MADV_HUGEPAGE marked this vma */
 120#define VM_NOHUGEPAGE   0x40000000      /* MADV_NOHUGEPAGE marked this vma */
 121#define VM_MERGEABLE    0x80000000      /* KSM may merge identical pages */
 122
 123#if defined(CONFIG_X86)
 124# define VM_PAT         VM_ARCH_1       /* PAT reserves whole VMA at once (x86) */
 125#elif defined(CONFIG_PPC)
 126# define VM_SAO         VM_ARCH_1       /* Strong Access Ordering (powerpc) */
 127#elif defined(CONFIG_PARISC)
 128# define VM_GROWSUP     VM_ARCH_1
 129#elif defined(CONFIG_METAG)
 130# define VM_GROWSUP     VM_ARCH_1
 131#elif defined(CONFIG_IA64)
 132# define VM_GROWSUP     VM_ARCH_1
 133#elif !defined(CONFIG_MMU)
 134# define VM_MAPPED_COPY VM_ARCH_1       /* T if mapped copy of data (nommu mmap) */
 135#endif
 136
 137#ifndef VM_GROWSUP
 138# define VM_GROWSUP     VM_NONE
 139#endif
 140
 141/* Bits set in the VMA until the stack is in its final location */
 142#define VM_STACK_INCOMPLETE_SETUP       (VM_RAND_READ | VM_SEQ_READ)
 143
 144#ifndef VM_STACK_DEFAULT_FLAGS          /* arch can override this */
 145#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
 146#endif
 147
 148#ifdef CONFIG_STACK_GROWSUP
 149#define VM_STACK_FLAGS  (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
 150#else
 151#define VM_STACK_FLAGS  (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
 152#endif
 153
 154/*
 155 * Special vmas that are non-mergable, non-mlock()able.
 156 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
 157 */
 158#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP)
 159
 160/*
 161 * mapping from the currently active vm_flags protection bits (the
 162 * low four bits) to a page protection mask..
 163 */
 164extern pgprot_t protection_map[16];
 165
 166#define FAULT_FLAG_WRITE        0x01    /* Fault was a write access */
 167#define FAULT_FLAG_NONLINEAR    0x02    /* Fault was via a nonlinear mapping */
 168#define FAULT_FLAG_MKWRITE      0x04    /* Fault was mkwrite of existing pte */
 169#define FAULT_FLAG_ALLOW_RETRY  0x08    /* Retry fault if blocking */
 170#define FAULT_FLAG_RETRY_NOWAIT 0x10    /* Don't drop mmap_sem and wait when retrying */
 171#define FAULT_FLAG_KILLABLE     0x20    /* The fault task is in SIGKILL killable region */
 172#define FAULT_FLAG_TRIED        0x40    /* second try */
 173
 174/*
 175 * vm_fault is filled by the the pagefault handler and passed to the vma's
 176 * ->fault function. The vma's ->fault is responsible for returning a bitmask
 177 * of VM_FAULT_xxx flags that give details about how the fault was handled.
 178 *
 179 * pgoff should be used in favour of virtual_address, if possible. If pgoff
 180 * is used, one may implement ->remap_pages to get nonlinear mapping support.
 181 */
 182struct vm_fault {
 183        unsigned int flags;             /* FAULT_FLAG_xxx flags */
 184        pgoff_t pgoff;                  /* Logical page offset based on vma */
 185        void __user *virtual_address;   /* Faulting virtual address */
 186
 187        struct page *page;              /* ->fault handlers should return a
 188                                         * page here, unless VM_FAULT_NOPAGE
 189                                         * is set (which is also implied by
 190                                         * VM_FAULT_ERROR).
 191                                         */
 192};
 193
 194/*
 195 * These are the virtual MM functions - opening of an area, closing and
 196 * unmapping it (needed to keep files on disk up-to-date etc), pointer
 197 * to the functions called when a no-page or a wp-page exception occurs. 
 198 */
 199struct vm_operations_struct {
 200        void (*open)(struct vm_area_struct * area);
 201        void (*close)(struct vm_area_struct * area);
 202        int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
 203
 204        /* notification that a previously read-only page is about to become
 205         * writable, if an error is returned it will cause a SIGBUS */
 206        int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
 207
 208        /* called by access_process_vm when get_user_pages() fails, typically
 209         * for use by special VMAs that can switch between memory and hardware
 210         */
 211        int (*access)(struct vm_area_struct *vma, unsigned long addr,
 212                      void *buf, int len, int write);
 213#ifdef CONFIG_NUMA
 214        /*
 215         * set_policy() op must add a reference to any non-NULL @new mempolicy
 216         * to hold the policy upon return.  Caller should pass NULL @new to
 217         * remove a policy and fall back to surrounding context--i.e. do not
 218         * install a MPOL_DEFAULT policy, nor the task or system default
 219         * mempolicy.
 220         */
 221        int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
 222
 223        /*
 224         * get_policy() op must add reference [mpol_get()] to any policy at
 225         * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
 226         * in mm/mempolicy.c will do this automatically.
 227         * get_policy() must NOT add a ref if the policy at (vma,addr) is not
 228         * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
 229         * If no [shared/vma] mempolicy exists at the addr, get_policy() op
 230         * must return NULL--i.e., do not "fallback" to task or system default
 231         * policy.
 232         */
 233        struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
 234                                        unsigned long addr);
 235        int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
 236                const nodemask_t *to, unsigned long flags);
 237#endif
 238        /* called by sys_remap_file_pages() to populate non-linear mapping */
 239        int (*remap_pages)(struct vm_area_struct *vma, unsigned long addr,
 240                           unsigned long size, pgoff_t pgoff);
 241};
 242
 243struct mmu_gather;
 244struct inode;
 245
 246#define page_private(page)              ((page)->private)
 247#define set_page_private(page, v)       ((page)->private = (v))
 248
 249/* It's valid only if the page is free path or free_list */
 250static inline void set_freepage_migratetype(struct page *page, int migratetype)
 251{
 252        page->index = migratetype;
 253}
 254
 255/* It's valid only if the page is free path or free_list */
 256static inline int get_freepage_migratetype(struct page *page)
 257{
 258        return page->index;
 259}
 260
 261/*
 262 * FIXME: take this include out, include page-flags.h in
 263 * files which need it (119 of them)
 264 */
 265#include <linux/page-flags.h>
 266#include <linux/huge_mm.h>
 267
 268/*
 269 * Methods to modify the page usage count.
 270 *
 271 * What counts for a page usage:
 272 * - cache mapping   (page->mapping)
 273 * - private data    (page->private)
 274 * - page mapped in a task's page tables, each mapping
 275 *   is counted separately
 276 *
 277 * Also, many kernel routines increase the page count before a critical
 278 * routine so they can be sure the page doesn't go away from under them.
 279 */
 280
 281/*
 282 * Drop a ref, return true if the refcount fell to zero (the page has no users)
 283 */
 284static inline int put_page_testzero(struct page *page)
 285{
 286        VM_BUG_ON(atomic_read(&page->_count) == 0);
 287        return atomic_dec_and_test(&page->_count);
 288}
 289
 290/*
 291 * Try to grab a ref unless the page has a refcount of zero, return false if
 292 * that is the case.
 293 */
 294static inline int get_page_unless_zero(struct page *page)
 295{
 296        return atomic_inc_not_zero(&page->_count);
 297}
 298
 299extern int page_is_ram(unsigned long pfn);
 300
 301/* Support for virtually mapped pages */
 302struct page *vmalloc_to_page(const void *addr);
 303unsigned long vmalloc_to_pfn(const void *addr);
 304
 305/*
 306 * Determine if an address is within the vmalloc range
 307 *
 308 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
 309 * is no special casing required.
 310 */
 311static inline int is_vmalloc_addr(const void *x)
 312{
 313#ifdef CONFIG_MMU
 314        unsigned long addr = (unsigned long)x;
 315
 316        return addr >= VMALLOC_START && addr < VMALLOC_END;
 317#else
 318        return 0;
 319#endif
 320}
 321#ifdef CONFIG_MMU
 322extern int is_vmalloc_or_module_addr(const void *x);
 323#else
 324static inline int is_vmalloc_or_module_addr(const void *x)
 325{
 326        return 0;
 327}
 328#endif
 329
 330static inline void compound_lock(struct page *page)
 331{
 332#ifdef CONFIG_TRANSPARENT_HUGEPAGE
 333        VM_BUG_ON(PageSlab(page));
 334        bit_spin_lock(PG_compound_lock, &page->flags);
 335#endif
 336}
 337
 338static inline void compound_unlock(struct page *page)
 339{
 340#ifdef CONFIG_TRANSPARENT_HUGEPAGE
 341        VM_BUG_ON(PageSlab(page));
 342        bit_spin_unlock(PG_compound_lock, &page->flags);
 343#endif
 344}
 345
 346static inline unsigned long compound_lock_irqsave(struct page *page)
 347{
 348        unsigned long uninitialized_var(flags);
 349#ifdef CONFIG_TRANSPARENT_HUGEPAGE
 350        local_irq_save(flags);
 351        compound_lock(page);
 352#endif
 353        return flags;
 354}
 355
 356static inline void compound_unlock_irqrestore(struct page *page,
 357                                              unsigned long flags)
 358{
 359#ifdef CONFIG_TRANSPARENT_HUGEPAGE
 360        compound_unlock(page);
 361        local_irq_restore(flags);
 362#endif
 363}
 364
 365static inline struct page *compound_head(struct page *page)
 366{
 367        if (unlikely(PageTail(page)))
 368                return page->first_page;
 369        return page;
 370}
 371
 372/*
 373 * The atomic page->_mapcount, starts from -1: so that transitions
 374 * both from it and to it can be tracked, using atomic_inc_and_test
 375 * and atomic_add_negative(-1).
 376 */
 377static inline void page_mapcount_reset(struct page *page)
 378{
 379        atomic_set(&(page)->_mapcount, -1);
 380}
 381
 382static inline int page_mapcount(struct page *page)
 383{
 384        return atomic_read(&(page)->_mapcount) + 1;
 385}
 386
 387static inline int page_count(struct page *page)
 388{
 389        return atomic_read(&compound_head(page)->_count);
 390}
 391
 392static inline void get_huge_page_tail(struct page *page)
 393{
 394        /*
 395         * __split_huge_page_refcount() cannot run
 396         * from under us.
 397         */
 398        VM_BUG_ON(page_mapcount(page) < 0);
 399        VM_BUG_ON(atomic_read(&page->_count) != 0);
 400        atomic_inc(&page->_mapcount);
 401}
 402
 403extern bool __get_page_tail(struct page *page);
 404
 405static inline void get_page(struct page *page)
 406{
 407        if (unlikely(PageTail(page)))
 408                if (likely(__get_page_tail(page)))
 409                        return;
 410        /*
 411         * Getting a normal page or the head of a compound page
 412         * requires to already have an elevated page->_count.
 413         */
 414        VM_BUG_ON(atomic_read(&page->_count) <= 0);
 415        atomic_inc(&page->_count);
 416}
 417
 418static inline struct page *virt_to_head_page(const void *x)
 419{
 420        struct page *page = virt_to_page(x);
 421        return compound_head(page);
 422}
 423
 424/*
 425 * Setup the page count before being freed into the page allocator for
 426 * the first time (boot or memory hotplug)
 427 */
 428static inline void init_page_count(struct page *page)
 429{
 430        atomic_set(&page->_count, 1);
 431}
 432
 433/*
 434 * PageBuddy() indicate that the page is free and in the buddy system
 435 * (see mm/page_alloc.c).
 436 *
 437 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
 438 * -2 so that an underflow of the page_mapcount() won't be mistaken
 439 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
 440 * efficiently by most CPU architectures.
 441 */
 442#define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
 443
 444static inline int PageBuddy(struct page *page)
 445{
 446        return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
 447}
 448
 449static inline void __SetPageBuddy(struct page *page)
 450{
 451        VM_BUG_ON(atomic_read(&page->_mapcount) != -1);
 452        atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
 453}
 454
 455static inline void __ClearPageBuddy(struct page *page)
 456{
 457        VM_BUG_ON(!PageBuddy(page));
 458        atomic_set(&page->_mapcount, -1);
 459}
 460
 461void put_page(struct page *page);
 462void put_pages_list(struct list_head *pages);
 463
 464void split_page(struct page *page, unsigned int order);
 465int split_free_page(struct page *page);
 466
 467/*
 468 * Compound pages have a destructor function.  Provide a
 469 * prototype for that function and accessor functions.
 470 * These are _only_ valid on the head of a PG_compound page.
 471 */
 472typedef void compound_page_dtor(struct page *);
 473
 474static inline void set_compound_page_dtor(struct page *page,
 475                                                compound_page_dtor *dtor)
 476{
 477        page[1].lru.next = (void *)dtor;
 478}
 479
 480static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
 481{
 482        return (compound_page_dtor *)page[1].lru.next;
 483}
 484
 485static inline int compound_order(struct page *page)
 486{
 487        if (!PageHead(page))
 488                return 0;
 489        return (unsigned long)page[1].lru.prev;
 490}
 491
 492static inline int compound_trans_order(struct page *page)
 493{
 494        int order;
 495        unsigned long flags;
 496
 497        if (!PageHead(page))
 498                return 0;
 499
 500        flags = compound_lock_irqsave(page);
 501        order = compound_order(page);
 502        compound_unlock_irqrestore(page, flags);
 503        return order;
 504}
 505
 506static inline void set_compound_order(struct page *page, unsigned long order)
 507{
 508        page[1].lru.prev = (void *)order;
 509}
 510
 511#ifdef CONFIG_MMU
 512/*
 513 * Do pte_mkwrite, but only if the vma says VM_WRITE.  We do this when
 514 * servicing faults for write access.  In the normal case, do always want
 515 * pte_mkwrite.  But get_user_pages can cause write faults for mappings
 516 * that do not have writing enabled, when used by access_process_vm.
 517 */
 518static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
 519{
 520        if (likely(vma->vm_flags & VM_WRITE))
 521                pte = pte_mkwrite(pte);
 522        return pte;
 523}
 524#endif
 525
 526/*
 527 * Multiple processes may "see" the same page. E.g. for untouched
 528 * mappings of /dev/null, all processes see the same page full of
 529 * zeroes, and text pages of executables and shared libraries have
 530 * only one copy in memory, at most, normally.
 531 *
 532 * For the non-reserved pages, page_count(page) denotes a reference count.
 533 *   page_count() == 0 means the page is free. page->lru is then used for
 534 *   freelist management in the buddy allocator.
 535 *   page_count() > 0  means the page has been allocated.
 536 *
 537 * Pages are allocated by the slab allocator in order to provide memory
 538 * to kmalloc and kmem_cache_alloc. In this case, the management of the
 539 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
 540 * unless a particular usage is carefully commented. (the responsibility of
 541 * freeing the kmalloc memory is the caller's, of course).
 542 *
 543 * A page may be used by anyone else who does a __get_free_page().
 544 * In this case, page_count still tracks the references, and should only
 545 * be used through the normal accessor functions. The top bits of page->flags
 546 * and page->virtual store page management information, but all other fields
 547 * are unused and could be used privately, carefully. The management of this
 548 * page is the responsibility of the one who allocated it, and those who have
 549 * subsequently been given references to it.
 550 *
 551 * The other pages (we may call them "pagecache pages") are completely
 552 * managed by the Linux memory manager: I/O, buffers, swapping etc.
 553 * The following discussion applies only to them.
 554 *
 555 * A pagecache page contains an opaque `private' member, which belongs to the
 556 * page's address_space. Usually, this is the address of a circular list of
 557 * the page's disk buffers. PG_private must be set to tell the VM to call
 558 * into the filesystem to release these pages.
 559 *
 560 * A page may belong to an inode's memory mapping. In this case, page->mapping
 561 * is the pointer to the inode, and page->index is the file offset of the page,
 562 * in units of PAGE_CACHE_SIZE.
 563 *
 564 * If pagecache pages are not associated with an inode, they are said to be
 565 * anonymous pages. These may become associated with the swapcache, and in that
 566 * case PG_swapcache is set, and page->private is an offset into the swapcache.
 567 *
 568 * In either case (swapcache or inode backed), the pagecache itself holds one
 569 * reference to the page. Setting PG_private should also increment the
 570 * refcount. The each user mapping also has a reference to the page.
 571 *
 572 * The pagecache pages are stored in a per-mapping radix tree, which is
 573 * rooted at mapping->page_tree, and indexed by offset.
 574 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
 575 * lists, we instead now tag pages as dirty/writeback in the radix tree.
 576 *
 577 * All pagecache pages may be subject to I/O:
 578 * - inode pages may need to be read from disk,
 579 * - inode pages which have been modified and are MAP_SHARED may need
 580 *   to be written back to the inode on disk,
 581 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
 582 *   modified may need to be swapped out to swap space and (later) to be read
 583 *   back into memory.
 584 */
 585
 586/*
 587 * The zone field is never updated after free_area_init_core()
 588 * sets it, so none of the operations on it need to be atomic.
 589 */
 590
 591/* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_NID] | ... | FLAGS | */
 592#define SECTIONS_PGOFF          ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
 593#define NODES_PGOFF             (SECTIONS_PGOFF - NODES_WIDTH)
 594#define ZONES_PGOFF             (NODES_PGOFF - ZONES_WIDTH)
 595#define LAST_NID_PGOFF          (ZONES_PGOFF - LAST_NID_WIDTH)
 596
 597/*
 598 * Define the bit shifts to access each section.  For non-existent
 599 * sections we define the shift as 0; that plus a 0 mask ensures
 600 * the compiler will optimise away reference to them.
 601 */
 602#define SECTIONS_PGSHIFT        (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
 603#define NODES_PGSHIFT           (NODES_PGOFF * (NODES_WIDTH != 0))
 604#define ZONES_PGSHIFT           (ZONES_PGOFF * (ZONES_WIDTH != 0))
 605#define LAST_NID_PGSHIFT        (LAST_NID_PGOFF * (LAST_NID_WIDTH != 0))
 606
 607/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
 608#ifdef NODE_NOT_IN_PAGE_FLAGS
 609#define ZONEID_SHIFT            (SECTIONS_SHIFT + ZONES_SHIFT)
 610#define ZONEID_PGOFF            ((SECTIONS_PGOFF < ZONES_PGOFF)? \
 611                                                SECTIONS_PGOFF : ZONES_PGOFF)
 612#else
 613#define ZONEID_SHIFT            (NODES_SHIFT + ZONES_SHIFT)
 614#define ZONEID_PGOFF            ((NODES_PGOFF < ZONES_PGOFF)? \
 615                                                NODES_PGOFF : ZONES_PGOFF)
 616#endif
 617
 618#define ZONEID_PGSHIFT          (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
 619
 620#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
 621#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
 622#endif
 623
 624#define ZONES_MASK              ((1UL << ZONES_WIDTH) - 1)
 625#define NODES_MASK              ((1UL << NODES_WIDTH) - 1)
 626#define SECTIONS_MASK           ((1UL << SECTIONS_WIDTH) - 1)
 627#define LAST_NID_MASK           ((1UL << LAST_NID_WIDTH) - 1)
 628#define ZONEID_MASK             ((1UL << ZONEID_SHIFT) - 1)
 629
 630static inline enum zone_type page_zonenum(const struct page *page)
 631{
 632        return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
 633}
 634
 635#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
 636#define SECTION_IN_PAGE_FLAGS
 637#endif
 638
 639/*
 640 * The identification function is only used by the buddy allocator for
 641 * determining if two pages could be buddies. We are not really
 642 * identifying a zone since we could be using a the section number
 643 * id if we have not node id available in page flags.
 644 * We guarantee only that it will return the same value for two
 645 * combinable pages in a zone.
 646 */
 647static inline int page_zone_id(struct page *page)
 648{
 649        return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
 650}
 651
 652static inline int zone_to_nid(struct zone *zone)
 653{
 654#ifdef CONFIG_NUMA
 655        return zone->node;
 656#else
 657        return 0;
 658#endif
 659}
 660
 661#ifdef NODE_NOT_IN_PAGE_FLAGS
 662extern int page_to_nid(const struct page *page);
 663#else
 664static inline int page_to_nid(const struct page *page)
 665{
 666        return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
 667}
 668#endif
 669
 670#ifdef CONFIG_NUMA_BALANCING
 671#ifdef LAST_NID_NOT_IN_PAGE_FLAGS
 672static inline int page_nid_xchg_last(struct page *page, int nid)
 673{
 674        return xchg(&page->_last_nid, nid);
 675}
 676
 677static inline int page_nid_last(struct page *page)
 678{
 679        return page->_last_nid;
 680}
 681static inline void page_nid_reset_last(struct page *page)
 682{
 683        page->_last_nid = -1;
 684}
 685#else
 686static inline int page_nid_last(struct page *page)
 687{
 688        return (page->flags >> LAST_NID_PGSHIFT) & LAST_NID_MASK;
 689}
 690
 691extern int page_nid_xchg_last(struct page *page, int nid);
 692
 693static inline void page_nid_reset_last(struct page *page)
 694{
 695        int nid = (1 << LAST_NID_SHIFT) - 1;
 696
 697        page->flags &= ~(LAST_NID_MASK << LAST_NID_PGSHIFT);
 698        page->flags |= (nid & LAST_NID_MASK) << LAST_NID_PGSHIFT;
 699}
 700#endif /* LAST_NID_NOT_IN_PAGE_FLAGS */
 701#else
 702static inline int page_nid_xchg_last(struct page *page, int nid)
 703{
 704        return page_to_nid(page);
 705}
 706
 707static inline int page_nid_last(struct page *page)
 708{
 709        return page_to_nid(page);
 710}
 711
 712static inline void page_nid_reset_last(struct page *page)
 713{
 714}
 715#endif
 716
 717static inline struct zone *page_zone(const struct page *page)
 718{
 719        return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
 720}
 721
 722#ifdef SECTION_IN_PAGE_FLAGS
 723static inline void set_page_section(struct page *page, unsigned long section)
 724{
 725        page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
 726        page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
 727}
 728
 729static inline unsigned long page_to_section(const struct page *page)
 730{
 731        return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
 732}
 733#endif
 734
 735static inline void set_page_zone(struct page *page, enum zone_type zone)
 736{
 737        page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
 738        page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
 739}
 740
 741static inline void set_page_node(struct page *page, unsigned long node)
 742{
 743        page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
 744        page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
 745}
 746
 747static inline void set_page_links(struct page *page, enum zone_type zone,
 748        unsigned long node, unsigned long pfn)
 749{
 750        set_page_zone(page, zone);
 751        set_page_node(page, node);
 752#ifdef SECTION_IN_PAGE_FLAGS
 753        set_page_section(page, pfn_to_section_nr(pfn));
 754#endif
 755}
 756
 757/*
 758 * Some inline functions in vmstat.h depend on page_zone()
 759 */
 760#include <linux/vmstat.h>
 761
 762static __always_inline void *lowmem_page_address(const struct page *page)
 763{
 764        return __va(PFN_PHYS(page_to_pfn(page)));
 765}
 766
 767#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
 768#define HASHED_PAGE_VIRTUAL
 769#endif
 770
 771#if defined(WANT_PAGE_VIRTUAL)
 772#define page_address(page) ((page)->virtual)
 773#define set_page_address(page, address)                 \
 774        do {                                            \
 775                (page)->virtual = (address);            \
 776        } while(0)
 777#define page_address_init()  do { } while(0)
 778#endif
 779
 780#if defined(HASHED_PAGE_VIRTUAL)
 781void *page_address(const struct page *page);
 782void set_page_address(struct page *page, void *virtual);
 783void page_address_init(void);
 784#endif
 785
 786#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
 787#define page_address(page) lowmem_page_address(page)
 788#define set_page_address(page, address)  do { } while(0)
 789#define page_address_init()  do { } while(0)
 790#endif
 791
 792/*
 793 * On an anonymous page mapped into a user virtual memory area,
 794 * page->mapping points to its anon_vma, not to a struct address_space;
 795 * with the PAGE_MAPPING_ANON bit set to distinguish it.  See rmap.h.
 796 *
 797 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
 798 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
 799 * and then page->mapping points, not to an anon_vma, but to a private
 800 * structure which KSM associates with that merged page.  See ksm.h.
 801 *
 802 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
 803 *
 804 * Please note that, confusingly, "page_mapping" refers to the inode
 805 * address_space which maps the page from disk; whereas "page_mapped"
 806 * refers to user virtual address space into which the page is mapped.
 807 */
 808#define PAGE_MAPPING_ANON       1
 809#define PAGE_MAPPING_KSM        2
 810#define PAGE_MAPPING_FLAGS      (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
 811
 812extern struct address_space *page_mapping(struct page *page);
 813
 814/* Neutral page->mapping pointer to address_space or anon_vma or other */
 815static inline void *page_rmapping(struct page *page)
 816{
 817        return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
 818}
 819
 820extern struct address_space *__page_file_mapping(struct page *);
 821
 822static inline
 823struct address_space *page_file_mapping(struct page *page)
 824{
 825        if (unlikely(PageSwapCache(page)))
 826                return __page_file_mapping(page);
 827
 828        return page->mapping;
 829}
 830
 831static inline int PageAnon(struct page *page)
 832{
 833        return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
 834}
 835
 836/*
 837 * Return the pagecache index of the passed page.  Regular pagecache pages
 838 * use ->index whereas swapcache pages use ->private
 839 */
 840static inline pgoff_t page_index(struct page *page)
 841{
 842        if (unlikely(PageSwapCache(page)))
 843                return page_private(page);
 844        return page->index;
 845}
 846
 847extern pgoff_t __page_file_index(struct page *page);
 848
 849/*
 850 * Return the file index of the page. Regular pagecache pages use ->index
 851 * whereas swapcache pages use swp_offset(->private)
 852 */
 853static inline pgoff_t page_file_index(struct page *page)
 854{
 855        if (unlikely(PageSwapCache(page)))
 856                return __page_file_index(page);
 857
 858        return page->index;
 859}
 860
 861/*
 862 * Return true if this page is mapped into pagetables.
 863 */
 864static inline int page_mapped(struct page *page)
 865{
 866        return atomic_read(&(page)->_mapcount) >= 0;
 867}
 868
 869/*
 870 * Different kinds of faults, as returned by handle_mm_fault().
 871 * Used to decide whether a process gets delivered SIGBUS or
 872 * just gets major/minor fault counters bumped up.
 873 */
 874
 875#define VM_FAULT_MINOR  0 /* For backwards compat. Remove me quickly. */
 876
 877#define VM_FAULT_OOM    0x0001
 878#define VM_FAULT_SIGBUS 0x0002
 879#define VM_FAULT_MAJOR  0x0004
 880#define VM_FAULT_WRITE  0x0008  /* Special case for get_user_pages */
 881#define VM_FAULT_HWPOISON 0x0010        /* Hit poisoned small page */
 882#define VM_FAULT_HWPOISON_LARGE 0x0020  /* Hit poisoned large page. Index encoded in upper bits */
 883
 884#define VM_FAULT_NOPAGE 0x0100  /* ->fault installed the pte, not return page */
 885#define VM_FAULT_LOCKED 0x0200  /* ->fault locked the returned page */
 886#define VM_FAULT_RETRY  0x0400  /* ->fault blocked, must retry */
 887
 888#define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
 889
 890#define VM_FAULT_ERROR  (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
 891                         VM_FAULT_HWPOISON_LARGE)
 892
 893/* Encode hstate index for a hwpoisoned large page */
 894#define VM_FAULT_SET_HINDEX(x) ((x) << 12)
 895#define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
 896
 897/*
 898 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
 899 */
 900extern void pagefault_out_of_memory(void);
 901
 902#define offset_in_page(p)       ((unsigned long)(p) & ~PAGE_MASK)
 903
 904/*
 905 * Flags passed to show_mem() and show_free_areas() to suppress output in
 906 * various contexts.
 907 */
 908#define SHOW_MEM_FILTER_NODES           (0x0001u)       /* disallowed nodes */
 909#define SHOW_MEM_FILTER_PAGE_COUNT      (0x0002u)       /* page type count */
 910
 911extern void show_free_areas(unsigned int flags);
 912extern bool skip_free_areas_node(unsigned int flags, int nid);
 913
 914int shmem_zero_setup(struct vm_area_struct *);
 915
 916extern int can_do_mlock(void);
 917extern int user_shm_lock(size_t, struct user_struct *);
 918extern void user_shm_unlock(size_t, struct user_struct *);
 919
 920/*
 921 * Parameter block passed down to zap_pte_range in exceptional cases.
 922 */
 923struct zap_details {
 924        struct vm_area_struct *nonlinear_vma;   /* Check page->index if set */
 925        struct address_space *check_mapping;    /* Check page->mapping if set */
 926        pgoff_t first_index;                    /* Lowest page->index to unmap */
 927        pgoff_t last_index;                     /* Highest page->index to unmap */
 928};
 929
 930struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
 931                pte_t pte);
 932
 933int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
 934                unsigned long size);
 935void zap_page_range(struct vm_area_struct *vma, unsigned long address,
 936                unsigned long size, struct zap_details *);
 937void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
 938                unsigned long start, unsigned long end);
 939
 940/**
 941 * mm_walk - callbacks for walk_page_range
 942 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
 943 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
 944 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
 945 *             this handler is required to be able to handle
 946 *             pmd_trans_huge() pmds.  They may simply choose to
 947 *             split_huge_page() instead of handling it explicitly.
 948 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
 949 * @pte_hole: if set, called for each hole at all levels
 950 * @hugetlb_entry: if set, called for each hugetlb entry
 951 *                 *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
 952 *                            is used.
 953 *
 954 * (see walk_page_range for more details)
 955 */
 956struct mm_walk {
 957        int (*pgd_entry)(pgd_t *pgd, unsigned long addr,
 958                         unsigned long next, struct mm_walk *walk);
 959        int (*pud_entry)(pud_t *pud, unsigned long addr,
 960                         unsigned long next, struct mm_walk *walk);
 961        int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
 962                         unsigned long next, struct mm_walk *walk);
 963        int (*pte_entry)(pte_t *pte, unsigned long addr,
 964                         unsigned long next, struct mm_walk *walk);
 965        int (*pte_hole)(unsigned long addr, unsigned long next,
 966                        struct mm_walk *walk);
 967        int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
 968                             unsigned long addr, unsigned long next,
 969                             struct mm_walk *walk);
 970        struct mm_struct *mm;
 971        void *private;
 972};
 973
 974int walk_page_range(unsigned long addr, unsigned long end,
 975                struct mm_walk *walk);
 976void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
 977                unsigned long end, unsigned long floor, unsigned long ceiling);
 978int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
 979                        struct vm_area_struct *vma);
 980void unmap_mapping_range(struct address_space *mapping,
 981                loff_t const holebegin, loff_t const holelen, int even_cows);
 982int follow_pfn(struct vm_area_struct *vma, unsigned long address,
 983        unsigned long *pfn);
 984int follow_phys(struct vm_area_struct *vma, unsigned long address,
 985                unsigned int flags, unsigned long *prot, resource_size_t *phys);
 986int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
 987                        void *buf, int len, int write);
 988
 989static inline void unmap_shared_mapping_range(struct address_space *mapping,
 990                loff_t const holebegin, loff_t const holelen)
 991{
 992        unmap_mapping_range(mapping, holebegin, holelen, 0);
 993}
 994
 995extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new);
 996extern void truncate_setsize(struct inode *inode, loff_t newsize);
 997void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
 998int truncate_inode_page(struct address_space *mapping, struct page *page);
 999int generic_error_remove_page(struct address_space *mapping, struct page *page);
1000int invalidate_inode_page(struct page *page);
1001
1002#ifdef CONFIG_MMU
1003extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1004                        unsigned long address, unsigned int flags);
1005extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1006                            unsigned long address, unsigned int fault_flags);
1007#else
1008static inline int handle_mm_fault(struct mm_struct *mm,
1009                        struct vm_area_struct *vma, unsigned long address,
1010                        unsigned int flags)
1011{
1012        /* should never happen if there's no MMU */
1013        BUG();
1014        return VM_FAULT_SIGBUS;
1015}
1016static inline int fixup_user_fault(struct task_struct *tsk,
1017                struct mm_struct *mm, unsigned long address,
1018                unsigned int fault_flags)
1019{
1020        /* should never happen if there's no MMU */
1021        BUG();
1022        return -EFAULT;
1023}
1024#endif
1025
1026extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1027extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1028                void *buf, int len, int write);
1029
1030long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1031                      unsigned long start, unsigned long nr_pages,
1032                      unsigned int foll_flags, struct page **pages,
1033                      struct vm_area_struct **vmas, int *nonblocking);
1034long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1035                    unsigned long start, unsigned long nr_pages,
1036                    int write, int force, struct page **pages,
1037                    struct vm_area_struct **vmas);
1038int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1039                        struct page **pages);
1040struct kvec;
1041int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1042                        struct page **pages);
1043int get_kernel_page(unsigned long start, int write, struct page **pages);
1044struct page *get_dump_page(unsigned long addr);
1045
1046extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1047extern void do_invalidatepage(struct page *page, unsigned int offset,
1048                              unsigned int length);
1049
1050int __set_page_dirty_nobuffers(struct page *page);
1051int __set_page_dirty_no_writeback(struct page *page);
1052int redirty_page_for_writepage(struct writeback_control *wbc,
1053                                struct page *page);
1054void account_page_dirtied(struct page *page, struct address_space *mapping);
1055void account_page_writeback(struct page *page);
1056int set_page_dirty(struct page *page);
1057int set_page_dirty_lock(struct page *page);
1058int clear_page_dirty_for_io(struct page *page);
1059
1060/* Is the vma a continuation of the stack vma above it? */
1061static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1062{
1063        return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1064}
1065
1066static inline int stack_guard_page_start(struct vm_area_struct *vma,
1067                                             unsigned long addr)
1068{
1069        return (vma->vm_flags & VM_GROWSDOWN) &&
1070                (vma->vm_start == addr) &&
1071                !vma_growsdown(vma->vm_prev, addr);
1072}
1073
1074/* Is the vma a continuation of the stack vma below it? */
1075static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1076{
1077        return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1078}
1079
1080static inline int stack_guard_page_end(struct vm_area_struct *vma,
1081                                           unsigned long addr)
1082{
1083        return (vma->vm_flags & VM_GROWSUP) &&
1084                (vma->vm_end == addr) &&
1085                !vma_growsup(vma->vm_next, addr);
1086}
1087
1088extern pid_t
1089vm_is_stack(struct task_struct *task, struct vm_area_struct *vma, int in_group);
1090
1091extern unsigned long move_page_tables(struct vm_area_struct *vma,
1092                unsigned long old_addr, struct vm_area_struct *new_vma,
1093                unsigned long new_addr, unsigned long len,
1094                bool need_rmap_locks);
1095extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1096                              unsigned long end, pgprot_t newprot,
1097                              int dirty_accountable, int prot_numa);
1098extern int mprotect_fixup(struct vm_area_struct *vma,
1099                          struct vm_area_struct **pprev, unsigned long start,
1100                          unsigned long end, unsigned long newflags);
1101
1102/*
1103 * doesn't attempt to fault and will return short.
1104 */
1105int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1106                          struct page **pages);
1107/*
1108 * per-process(per-mm_struct) statistics.
1109 */
1110static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1111{
1112        long val = atomic_long_read(&mm->rss_stat.count[member]);
1113
1114#ifdef SPLIT_RSS_COUNTING
1115        /*
1116         * counter is updated in asynchronous manner and may go to minus.
1117         * But it's never be expected number for users.
1118         */
1119        if (val < 0)
1120                val = 0;
1121#endif
1122        return (unsigned long)val;
1123}
1124
1125static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1126{
1127        atomic_long_add(value, &mm->rss_stat.count[member]);
1128}
1129
1130static inline void inc_mm_counter(struct mm_struct *mm, int member)
1131{
1132        atomic_long_inc(&mm->rss_stat.count[member]);
1133}
1134
1135static inline void dec_mm_counter(struct mm_struct *mm, int member)
1136{
1137        atomic_long_dec(&mm->rss_stat.count[member]);
1138}
1139
1140static inline unsigned long get_mm_rss(struct mm_struct *mm)
1141{
1142        return get_mm_counter(mm, MM_FILEPAGES) +
1143                get_mm_counter(mm, MM_ANONPAGES);
1144}
1145
1146static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1147{
1148        return max(mm->hiwater_rss, get_mm_rss(mm));
1149}
1150
1151static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1152{
1153        return max(mm->hiwater_vm, mm->total_vm);
1154}
1155
1156static inline void update_hiwater_rss(struct mm_struct *mm)
1157{
1158        unsigned long _rss = get_mm_rss(mm);
1159
1160        if ((mm)->hiwater_rss < _rss)
1161                (mm)->hiwater_rss = _rss;
1162}
1163
1164static inline void update_hiwater_vm(struct mm_struct *mm)
1165{
1166        if (mm->hiwater_vm < mm->total_vm)
1167                mm->hiwater_vm = mm->total_vm;
1168}
1169
1170static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1171                                         struct mm_struct *mm)
1172{
1173        unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1174
1175        if (*maxrss < hiwater_rss)
1176                *maxrss = hiwater_rss;
1177}
1178
1179#if defined(SPLIT_RSS_COUNTING)
1180void sync_mm_rss(struct mm_struct *mm);
1181#else
1182static inline void sync_mm_rss(struct mm_struct *mm)
1183{
1184}
1185#endif
1186
1187int vma_wants_writenotify(struct vm_area_struct *vma);
1188
1189extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1190                               spinlock_t **ptl);
1191static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1192                                    spinlock_t **ptl)
1193{
1194        pte_t *ptep;
1195        __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1196        return ptep;
1197}
1198
1199#ifdef __PAGETABLE_PUD_FOLDED
1200static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1201                                                unsigned long address)
1202{
1203        return 0;
1204}
1205#else
1206int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1207#endif
1208
1209#ifdef __PAGETABLE_PMD_FOLDED
1210static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1211                                                unsigned long address)
1212{
1213        return 0;
1214}
1215#else
1216int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1217#endif
1218
1219int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1220                pmd_t *pmd, unsigned long address);
1221int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1222
1223/*
1224 * The following ifdef needed to get the 4level-fixup.h header to work.
1225 * Remove it when 4level-fixup.h has been removed.
1226 */
1227#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1228static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1229{
1230        return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1231                NULL: pud_offset(pgd, address);
1232}
1233
1234static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1235{
1236        return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1237                NULL: pmd_offset(pud, address);
1238}
1239#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1240
1241#if USE_SPLIT_PTLOCKS
1242/*
1243 * We tuck a spinlock to guard each pagetable page into its struct page,
1244 * at page->private, with BUILD_BUG_ON to make sure that this will not
1245 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1246 * When freeing, reset page->mapping so free_pages_check won't complain.
1247 */
1248#define __pte_lockptr(page)     &((page)->ptl)
1249#define pte_lock_init(_page)    do {                                    \
1250        spin_lock_init(__pte_lockptr(_page));                           \
1251} while (0)
1252#define pte_lock_deinit(page)   ((page)->mapping = NULL)
1253#define pte_lockptr(mm, pmd)    ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
1254#else   /* !USE_SPLIT_PTLOCKS */
1255/*
1256 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1257 */
1258#define pte_lock_init(page)     do {} while (0)
1259#define pte_lock_deinit(page)   do {} while (0)
1260#define pte_lockptr(mm, pmd)    ({(void)(pmd); &(mm)->page_table_lock;})
1261#endif /* USE_SPLIT_PTLOCKS */
1262
1263static inline void pgtable_page_ctor(struct page *page)
1264{
1265        pte_lock_init(page);
1266        inc_zone_page_state(page, NR_PAGETABLE);
1267}
1268
1269static inline void pgtable_page_dtor(struct page *page)
1270{
1271        pte_lock_deinit(page);
1272        dec_zone_page_state(page, NR_PAGETABLE);
1273}
1274
1275#define pte_offset_map_lock(mm, pmd, address, ptlp)     \
1276({                                                      \
1277        spinlock_t *__ptl = pte_lockptr(mm, pmd);       \
1278        pte_t *__pte = pte_offset_map(pmd, address);    \
1279        *(ptlp) = __ptl;                                \
1280        spin_lock(__ptl);                               \
1281        __pte;                                          \
1282})
1283
1284#define pte_unmap_unlock(pte, ptl)      do {            \
1285        spin_unlock(ptl);                               \
1286        pte_unmap(pte);                                 \
1287} while (0)
1288
1289#define pte_alloc_map(mm, vma, pmd, address)                            \
1290        ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma,    \
1291                                                        pmd, address))? \
1292         NULL: pte_offset_map(pmd, address))
1293
1294#define pte_alloc_map_lock(mm, pmd, address, ptlp)      \
1295        ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL,   \
1296                                                        pmd, address))? \
1297                NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1298
1299#define pte_alloc_kernel(pmd, address)                  \
1300        ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1301                NULL: pte_offset_kernel(pmd, address))
1302
1303extern void free_area_init(unsigned long * zones_size);
1304extern void free_area_init_node(int nid, unsigned long * zones_size,
1305                unsigned long zone_start_pfn, unsigned long *zholes_size);
1306extern void free_initmem(void);
1307
1308/*
1309 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1310 * into the buddy system. The freed pages will be poisoned with pattern
1311 * "poison" if it's within range [0, UCHAR_MAX].
1312 * Return pages freed into the buddy system.
1313 */
1314extern unsigned long free_reserved_area(void *start, void *end,
1315                                        int poison, char *s);
1316
1317#ifdef  CONFIG_HIGHMEM
1318/*
1319 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1320 * and totalram_pages.
1321 */
1322extern void free_highmem_page(struct page *page);
1323#endif
1324
1325extern void adjust_managed_page_count(struct page *page, long count);
1326extern void mem_init_print_info(const char *str);
1327
1328/* Free the reserved page into the buddy system, so it gets managed. */
1329static inline void __free_reserved_page(struct page *page)
1330{
1331        ClearPageReserved(page);
1332        init_page_count(page);
1333        __free_page(page);
1334}
1335
1336static inline void free_reserved_page(struct page *page)
1337{
1338        __free_reserved_page(page);
1339        adjust_managed_page_count(page, 1);
1340}
1341
1342static inline void mark_page_reserved(struct page *page)
1343{
1344        SetPageReserved(page);
1345        adjust_managed_page_count(page, -1);
1346}
1347
1348/*
1349 * Default method to free all the __init memory into the buddy system.
1350 * The freed pages will be poisoned with pattern "poison" if it's within
1351 * range [0, UCHAR_MAX].
1352 * Return pages freed into the buddy system.
1353 */
1354static inline unsigned long free_initmem_default(int poison)
1355{
1356        extern char __init_begin[], __init_end[];
1357
1358        return free_reserved_area(&__init_begin, &__init_end,
1359                                  poison, "unused kernel");
1360}
1361
1362static inline unsigned long get_num_physpages(void)
1363{
1364        int nid;
1365        unsigned long phys_pages = 0;
1366
1367        for_each_online_node(nid)
1368                phys_pages += node_present_pages(nid);
1369
1370        return phys_pages;
1371}
1372
1373#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1374/*
1375 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1376 * zones, allocate the backing mem_map and account for memory holes in a more
1377 * architecture independent manner. This is a substitute for creating the
1378 * zone_sizes[] and zholes_size[] arrays and passing them to
1379 * free_area_init_node()
1380 *
1381 * An architecture is expected to register range of page frames backed by
1382 * physical memory with memblock_add[_node]() before calling
1383 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1384 * usage, an architecture is expected to do something like
1385 *
1386 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1387 *                                                       max_highmem_pfn};
1388 * for_each_valid_physical_page_range()
1389 *      memblock_add_node(base, size, nid)
1390 * free_area_init_nodes(max_zone_pfns);
1391 *
1392 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1393 * registered physical page range.  Similarly
1394 * sparse_memory_present_with_active_regions() calls memory_present() for
1395 * each range when SPARSEMEM is enabled.
1396 *
1397 * See mm/page_alloc.c for more information on each function exposed by
1398 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1399 */
1400extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1401unsigned long node_map_pfn_alignment(void);
1402unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1403                                                unsigned long end_pfn);
1404extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1405                                                unsigned long end_pfn);
1406extern void get_pfn_range_for_nid(unsigned int nid,
1407                        unsigned long *start_pfn, unsigned long *end_pfn);
1408extern unsigned long find_min_pfn_with_active_regions(void);
1409extern void free_bootmem_with_active_regions(int nid,
1410                                                unsigned long max_low_pfn);
1411extern void sparse_memory_present_with_active_regions(int nid);
1412
1413#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1414
1415#if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1416    !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1417static inline int __early_pfn_to_nid(unsigned long pfn)
1418{
1419        return 0;
1420}
1421#else
1422/* please see mm/page_alloc.c */
1423extern int __meminit early_pfn_to_nid(unsigned long pfn);
1424#ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1425/* there is a per-arch backend function. */
1426extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1427#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1428#endif
1429
1430extern void set_dma_reserve(unsigned long new_dma_reserve);
1431extern void memmap_init_zone(unsigned long, int, unsigned long,
1432                                unsigned long, enum memmap_context);
1433extern void setup_per_zone_wmarks(void);
1434extern int __meminit init_per_zone_wmark_min(void);
1435extern void mem_init(void);
1436extern void __init mmap_init(void);
1437extern void show_mem(unsigned int flags);
1438extern void si_meminfo(struct sysinfo * val);
1439extern void si_meminfo_node(struct sysinfo *val, int nid);
1440
1441extern __printf(3, 4)
1442void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1443
1444extern void setup_per_cpu_pageset(void);
1445
1446extern void zone_pcp_update(struct zone *zone);
1447extern void zone_pcp_reset(struct zone *zone);
1448
1449/* page_alloc.c */
1450extern int min_free_kbytes;
1451
1452/* nommu.c */
1453extern atomic_long_t mmap_pages_allocated;
1454extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1455
1456/* interval_tree.c */
1457void vma_interval_tree_insert(struct vm_area_struct *node,
1458                              struct rb_root *root);
1459void vma_interval_tree_insert_after(struct vm_area_struct *node,
1460                                    struct vm_area_struct *prev,
1461                                    struct rb_root *root);
1462void vma_interval_tree_remove(struct vm_area_struct *node,
1463                              struct rb_root *root);
1464struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1465                                unsigned long start, unsigned long last);
1466struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1467                                unsigned long start, unsigned long last);
1468
1469#define vma_interval_tree_foreach(vma, root, start, last)               \
1470        for (vma = vma_interval_tree_iter_first(root, start, last);     \
1471             vma; vma = vma_interval_tree_iter_next(vma, start, last))
1472
1473static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1474                                        struct list_head *list)
1475{
1476        list_add_tail(&vma->shared.nonlinear, list);
1477}
1478
1479void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1480                                   struct rb_root *root);
1481void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1482                                   struct rb_root *root);
1483struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1484        struct rb_root *root, unsigned long start, unsigned long last);
1485struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1486        struct anon_vma_chain *node, unsigned long start, unsigned long last);
1487#ifdef CONFIG_DEBUG_VM_RB
1488void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1489#endif
1490
1491#define anon_vma_interval_tree_foreach(avc, root, start, last)           \
1492        for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1493             avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1494
1495/* mmap.c */
1496extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1497extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1498        unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1499extern struct vm_area_struct *vma_merge(struct mm_struct *,
1500        struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1501        unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1502        struct mempolicy *);
1503extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1504extern int split_vma(struct mm_struct *,
1505        struct vm_area_struct *, unsigned long addr, int new_below);
1506extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1507extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1508        struct rb_node **, struct rb_node *);
1509extern void unlink_file_vma(struct vm_area_struct *);
1510extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1511        unsigned long addr, unsigned long len, pgoff_t pgoff,
1512        bool *need_rmap_locks);
1513extern void exit_mmap(struct mm_struct *);
1514
1515extern int mm_take_all_locks(struct mm_struct *mm);
1516extern void mm_drop_all_locks(struct mm_struct *mm);
1517
1518extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1519extern struct file *get_mm_exe_file(struct mm_struct *mm);
1520
1521extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1522extern int install_special_mapping(struct mm_struct *mm,
1523                                   unsigned long addr, unsigned long len,
1524                                   unsigned long flags, struct page **pages);
1525
1526extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1527
1528extern unsigned long mmap_region(struct file *file, unsigned long addr,
1529        unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
1530extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1531        unsigned long len, unsigned long prot, unsigned long flags,
1532        unsigned long pgoff, unsigned long *populate);
1533extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1534
1535#ifdef CONFIG_MMU
1536extern int __mm_populate(unsigned long addr, unsigned long len,
1537                         int ignore_errors);
1538static inline void mm_populate(unsigned long addr, unsigned long len)
1539{
1540        /* Ignore errors */
1541        (void) __mm_populate(addr, len, 1);
1542}
1543#else
1544static inline void mm_populate(unsigned long addr, unsigned long len) {}
1545#endif
1546
1547/* These take the mm semaphore themselves */
1548extern unsigned long vm_brk(unsigned long, unsigned long);
1549extern int vm_munmap(unsigned long, size_t);
1550extern unsigned long vm_mmap(struct file *, unsigned long,
1551        unsigned long, unsigned long,
1552        unsigned long, unsigned long);
1553
1554struct vm_unmapped_area_info {
1555#define VM_UNMAPPED_AREA_TOPDOWN 1
1556        unsigned long flags;
1557        unsigned long length;
1558        unsigned long low_limit;
1559        unsigned long high_limit;
1560        unsigned long align_mask;
1561        unsigned long align_offset;
1562};
1563
1564extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
1565extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
1566
1567/*
1568 * Search for an unmapped address range.
1569 *
1570 * We are looking for a range that:
1571 * - does not intersect with any VMA;
1572 * - is contained within the [low_limit, high_limit) interval;
1573 * - is at least the desired size.
1574 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1575 */
1576static inline unsigned long
1577vm_unmapped_area(struct vm_unmapped_area_info *info)
1578{
1579        if (!(info->flags & VM_UNMAPPED_AREA_TOPDOWN))
1580                return unmapped_area(info);
1581        else
1582                return unmapped_area_topdown(info);
1583}
1584
1585/* truncate.c */
1586extern void truncate_inode_pages(struct address_space *, loff_t);
1587extern void truncate_inode_pages_range(struct address_space *,
1588                                       loff_t lstart, loff_t lend);
1589
1590/* generic vm_area_ops exported for stackable file systems */
1591extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1592extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1593
1594/* mm/page-writeback.c */
1595int write_one_page(struct page *page, int wait);
1596void task_dirty_inc(struct task_struct *tsk);
1597
1598/* readahead.c */
1599#define VM_MAX_READAHEAD        128     /* kbytes */
1600#define VM_MIN_READAHEAD        16      /* kbytes (includes current page) */
1601
1602int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1603                        pgoff_t offset, unsigned long nr_to_read);
1604
1605void page_cache_sync_readahead(struct address_space *mapping,
1606                               struct file_ra_state *ra,
1607                               struct file *filp,
1608                               pgoff_t offset,
1609                               unsigned long size);
1610
1611void page_cache_async_readahead(struct address_space *mapping,
1612                                struct file_ra_state *ra,
1613                                struct file *filp,
1614                                struct page *pg,
1615                                pgoff_t offset,
1616                                unsigned long size);
1617
1618unsigned long max_sane_readahead(unsigned long nr);
1619unsigned long ra_submit(struct file_ra_state *ra,
1620                        struct address_space *mapping,
1621                        struct file *filp);
1622
1623/* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1624extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1625
1626/* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1627extern int expand_downwards(struct vm_area_struct *vma,
1628                unsigned long address);
1629#if VM_GROWSUP
1630extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1631#else
1632  #define expand_upwards(vma, address) do { } while (0)
1633#endif
1634
1635/* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
1636extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1637extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1638                                             struct vm_area_struct **pprev);
1639
1640/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1641   NULL if none.  Assume start_addr < end_addr. */
1642static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1643{
1644        struct vm_area_struct * vma = find_vma(mm,start_addr);
1645
1646        if (vma && end_addr <= vma->vm_start)
1647                vma = NULL;
1648        return vma;
1649}
1650
1651static inline unsigned long vma_pages(struct vm_area_struct *vma)
1652{
1653        return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1654}
1655
1656/* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1657static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
1658                                unsigned long vm_start, unsigned long vm_end)
1659{
1660        struct vm_area_struct *vma = find_vma(mm, vm_start);
1661
1662        if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
1663                vma = NULL;
1664
1665        return vma;
1666}
1667
1668#ifdef CONFIG_MMU
1669pgprot_t vm_get_page_prot(unsigned long vm_flags);
1670#else
1671static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1672{
1673        return __pgprot(0);
1674}
1675#endif
1676
1677#ifdef CONFIG_ARCH_USES_NUMA_PROT_NONE
1678unsigned long change_prot_numa(struct vm_area_struct *vma,
1679                        unsigned long start, unsigned long end);
1680#endif
1681
1682struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1683int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1684                        unsigned long pfn, unsigned long size, pgprot_t);
1685int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1686int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1687                        unsigned long pfn);
1688int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1689                        unsigned long pfn);
1690int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
1691
1692
1693struct page *follow_page_mask(struct vm_area_struct *vma,
1694                              unsigned long address, unsigned int foll_flags,
1695                              unsigned int *page_mask);
1696
1697static inline struct page *follow_page(struct vm_area_struct *vma,
1698                unsigned long address, unsigned int foll_flags)
1699{
1700        unsigned int unused_page_mask;
1701        return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
1702}
1703
1704#define FOLL_WRITE      0x01    /* check pte is writable */
1705#define FOLL_TOUCH      0x02    /* mark page accessed */
1706#define FOLL_GET        0x04    /* do get_page on page */
1707#define FOLL_DUMP       0x08    /* give error on hole if it would be zero */
1708#define FOLL_FORCE      0x10    /* get_user_pages read/write w/o permission */
1709#define FOLL_NOWAIT     0x20    /* if a disk transfer is needed, start the IO
1710                                 * and return without waiting upon it */
1711#define FOLL_MLOCK      0x40    /* mark page as mlocked */
1712#define FOLL_SPLIT      0x80    /* don't return transhuge pages, split them */
1713#define FOLL_HWPOISON   0x100   /* check page is hwpoisoned */
1714#define FOLL_NUMA       0x200   /* force NUMA hinting page fault */
1715#define FOLL_MIGRATION  0x400   /* wait for page to replace migration entry */
1716
1717typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1718                        void *data);
1719extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1720                               unsigned long size, pte_fn_t fn, void *data);
1721
1722#ifdef CONFIG_PROC_FS
1723void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1724#else
1725static inline void vm_stat_account(struct mm_struct *mm,
1726                        unsigned long flags, struct file *file, long pages)
1727{
1728        mm->total_vm += pages;
1729}
1730#endif /* CONFIG_PROC_FS */
1731
1732#ifdef CONFIG_DEBUG_PAGEALLOC
1733extern void kernel_map_pages(struct page *page, int numpages, int enable);
1734#ifdef CONFIG_HIBERNATION
1735extern bool kernel_page_present(struct page *page);
1736#endif /* CONFIG_HIBERNATION */
1737#else
1738static inline void
1739kernel_map_pages(struct page *page, int numpages, int enable) {}
1740#ifdef CONFIG_HIBERNATION
1741static inline bool kernel_page_present(struct page *page) { return true; }
1742#endif /* CONFIG_HIBERNATION */
1743#endif
1744
1745extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
1746#ifdef  __HAVE_ARCH_GATE_AREA
1747int in_gate_area_no_mm(unsigned long addr);
1748int in_gate_area(struct mm_struct *mm, unsigned long addr);
1749#else
1750int in_gate_area_no_mm(unsigned long addr);
1751#define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
1752#endif  /* __HAVE_ARCH_GATE_AREA */
1753
1754#ifdef CONFIG_SYSCTL
1755extern int sysctl_drop_caches;
1756int drop_caches_sysctl_handler(struct ctl_table *, int,
1757                                        void __user *, size_t *, loff_t *);
1758#endif
1759
1760unsigned long shrink_slab(struct shrink_control *shrink,
1761                          unsigned long nr_pages_scanned,
1762                          unsigned long lru_pages);
1763
1764#ifndef CONFIG_MMU
1765#define randomize_va_space 0
1766#else
1767extern int randomize_va_space;
1768#endif
1769
1770const char * arch_vma_name(struct vm_area_struct *vma);
1771void print_vma_addr(char *prefix, unsigned long rip);
1772
1773void sparse_mem_maps_populate_node(struct page **map_map,
1774                                   unsigned long pnum_begin,
1775                                   unsigned long pnum_end,
1776                                   unsigned long map_count,
1777                                   int nodeid);
1778
1779struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1780pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1781pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1782pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1783pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1784void *vmemmap_alloc_block(unsigned long size, int node);
1785void *vmemmap_alloc_block_buf(unsigned long size, int node);
1786void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1787int vmemmap_populate_basepages(unsigned long start, unsigned long end,
1788                               int node);
1789int vmemmap_populate(unsigned long start, unsigned long end, int node);
1790void vmemmap_populate_print_last(void);
1791#ifdef CONFIG_MEMORY_HOTPLUG
1792void vmemmap_free(unsigned long start, unsigned long end);
1793#endif
1794void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
1795                                  unsigned long size);
1796
1797enum mf_flags {
1798        MF_COUNT_INCREASED = 1 << 0,
1799        MF_ACTION_REQUIRED = 1 << 1,
1800        MF_MUST_KILL = 1 << 2,
1801};
1802extern int memory_failure(unsigned long pfn, int trapno, int flags);
1803extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
1804extern int unpoison_memory(unsigned long pfn);
1805extern int sysctl_memory_failure_early_kill;
1806extern int sysctl_memory_failure_recovery;
1807extern void shake_page(struct page *p, int access);
1808extern atomic_long_t num_poisoned_pages;
1809extern int soft_offline_page(struct page *page, int flags);
1810
1811extern void dump_page(struct page *page);
1812
1813#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1814extern void clear_huge_page(struct page *page,
1815                            unsigned long addr,
1816                            unsigned int pages_per_huge_page);
1817extern void copy_user_huge_page(struct page *dst, struct page *src,
1818                                unsigned long addr, struct vm_area_struct *vma,
1819                                unsigned int pages_per_huge_page);
1820#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1821
1822#ifdef CONFIG_DEBUG_PAGEALLOC
1823extern unsigned int _debug_guardpage_minorder;
1824
1825static inline unsigned int debug_guardpage_minorder(void)
1826{
1827        return _debug_guardpage_minorder;
1828}
1829
1830static inline bool page_is_guard(struct page *page)
1831{
1832        return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags);
1833}
1834#else
1835static inline unsigned int debug_guardpage_minorder(void) { return 0; }
1836static inline bool page_is_guard(struct page *page) { return false; }
1837#endif /* CONFIG_DEBUG_PAGEALLOC */
1838
1839#if MAX_NUMNODES > 1
1840void __init setup_nr_node_ids(void);
1841#else
1842static inline void setup_nr_node_ids(void) {}
1843#endif
1844
1845#endif /* __KERNEL__ */
1846#endif /* _LINUX_MM_H */
1847