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_METAG)
 118# define VM_GROWSUP     VM_ARCH_1
 119#elif defined(CONFIG_IA64)
 120# define VM_GROWSUP     VM_ARCH_1
 121#elif !defined(CONFIG_MMU)
 122# define VM_MAPPED_COPY VM_ARCH_1       /* T if mapped copy of data (nommu mmap) */
 123#endif
 124
 125#ifndef VM_GROWSUP
 126# define VM_GROWSUP     VM_NONE
 127#endif
 128
 129/* Bits set in the VMA until the stack is in its final location */
 130#define VM_STACK_INCOMPLETE_SETUP       (VM_RAND_READ | VM_SEQ_READ)
 131
 132#ifndef VM_STACK_DEFAULT_FLAGS          /* arch can override this */
 133#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
 134#endif
 135
 136#ifdef CONFIG_STACK_GROWSUP
 137#define VM_STACK_FLAGS  (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
 138#else
 139#define VM_STACK_FLAGS  (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
 140#endif
 141
 142#define VM_READHINTMASK                 (VM_SEQ_READ | VM_RAND_READ)
 143#define VM_ClearReadHint(v)             (v)->vm_flags &= ~VM_READHINTMASK
 144#define VM_NormalReadHint(v)            (!((v)->vm_flags & VM_READHINTMASK))
 145#define VM_SequentialReadHint(v)        ((v)->vm_flags & VM_SEQ_READ)
 146#define VM_RandomReadHint(v)            ((v)->vm_flags & VM_RAND_READ)
 147
 148/*
 149 * Special vmas that are non-mergable, non-mlock()able.
 150 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
 151 */
 152#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP)
 153
 154/*
 155 * mapping from the currently active vm_flags protection bits (the
 156 * low four bits) to a page protection mask..
 157 */
 158extern pgprot_t protection_map[16];
 159
 160#define FAULT_FLAG_WRITE        0x01    /* Fault was a write access */
 161#define FAULT_FLAG_NONLINEAR    0x02    /* Fault was via a nonlinear mapping */
 162#define FAULT_FLAG_MKWRITE      0x04    /* Fault was mkwrite of existing pte */
 163#define FAULT_FLAG_ALLOW_RETRY  0x08    /* Retry fault if blocking */
 164#define FAULT_FLAG_RETRY_NOWAIT 0x10    /* Don't drop mmap_sem and wait when retrying */
 165#define FAULT_FLAG_KILLABLE     0x20    /* The fault task is in SIGKILL killable region */
 166#define FAULT_FLAG_TRIED        0x40    /* second try */
 167
 168/*
 169 * vm_fault is filled by the the pagefault handler and passed to the vma's
 170 * ->fault function. The vma's ->fault is responsible for returning a bitmask
 171 * of VM_FAULT_xxx flags that give details about how the fault was handled.
 172 *
 173 * pgoff should be used in favour of virtual_address, if possible. If pgoff
 174 * is used, one may implement ->remap_pages to get nonlinear mapping support.
 175 */
 176struct vm_fault {
 177        unsigned int flags;             /* FAULT_FLAG_xxx flags */
 178        pgoff_t pgoff;                  /* Logical page offset based on vma */
 179        void __user *virtual_address;   /* Faulting virtual address */
 180
 181        struct page *page;              /* ->fault handlers should return a
 182                                         * page here, unless VM_FAULT_NOPAGE
 183                                         * is set (which is also implied by
 184                                         * VM_FAULT_ERROR).
 185                                         */
 186};
 187
 188/*
 189 * These are the virtual MM functions - opening of an area, closing and
 190 * unmapping it (needed to keep files on disk up-to-date etc), pointer
 191 * to the functions called when a no-page or a wp-page exception occurs. 
 192 */
 193struct vm_operations_struct {
 194        void (*open)(struct vm_area_struct * area);
 195        void (*close)(struct vm_area_struct * area);
 196        int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
 197
 198        /* notification that a previously read-only page is about to become
 199         * writable, if an error is returned it will cause a SIGBUS */
 200        int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
 201
 202        /* called by access_process_vm when get_user_pages() fails, typically
 203         * for use by special VMAs that can switch between memory and hardware
 204         */
 205        int (*access)(struct vm_area_struct *vma, unsigned long addr,
 206                      void *buf, int len, int write);
 207#ifdef CONFIG_NUMA
 208        /*
 209         * set_policy() op must add a reference to any non-NULL @new mempolicy
 210         * to hold the policy upon return.  Caller should pass NULL @new to
 211         * remove a policy and fall back to surrounding context--i.e. do not
 212         * install a MPOL_DEFAULT policy, nor the task or system default
 213         * mempolicy.
 214         */
 215        int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
 216
 217        /*
 218         * get_policy() op must add reference [mpol_get()] to any policy at
 219         * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
 220         * in mm/mempolicy.c will do this automatically.
 221         * get_policy() must NOT add a ref if the policy at (vma,addr) is not
 222         * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
 223         * If no [shared/vma] mempolicy exists at the addr, get_policy() op
 224         * must return NULL--i.e., do not "fallback" to task or system default
 225         * policy.
 226         */
 227        struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
 228                                        unsigned long addr);
 229        int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
 230                const nodemask_t *to, unsigned long flags);
 231#endif
 232        /* called by sys_remap_file_pages() to populate non-linear mapping */
 233        int (*remap_pages)(struct vm_area_struct *vma, unsigned long addr,
 234                           unsigned long size, pgoff_t pgoff);
 235};
 236
 237struct mmu_gather;
 238struct inode;
 239
 240#define page_private(page)              ((page)->private)
 241#define set_page_private(page, v)       ((page)->private = (v))
 242
 243/* It's valid only if the page is free path or free_list */
 244static inline void set_freepage_migratetype(struct page *page, int migratetype)
 245{
 246        page->index = migratetype;
 247}
 248
 249/* It's valid only if the page is free path or free_list */
 250static inline int get_freepage_migratetype(struct page *page)
 251{
 252        return page->index;
 253}
 254
 255/*
 256 * FIXME: take this include out, include page-flags.h in
 257 * files which need it (119 of them)
 258 */
 259#include <linux/page-flags.h>
 260#include <linux/huge_mm.h>
 261
 262/*
 263 * Methods to modify the page usage count.
 264 *
 265 * What counts for a page usage:
 266 * - cache mapping   (page->mapping)
 267 * - private data    (page->private)
 268 * - page mapped in a task's page tables, each mapping
 269 *   is counted separately
 270 *
 271 * Also, many kernel routines increase the page count before a critical
 272 * routine so they can be sure the page doesn't go away from under them.
 273 */
 274
 275/*
 276 * Drop a ref, return true if the refcount fell to zero (the page has no users)
 277 */
 278static inline int put_page_testzero(struct page *page)
 279{
 280        VM_BUG_ON(atomic_read(&page->_count) == 0);
 281        return atomic_dec_and_test(&page->_count);
 282}
 283
 284/*
 285 * Try to grab a ref unless the page has a refcount of zero, return false if
 286 * that is the case.
 287 */
 288static inline int get_page_unless_zero(struct page *page)
 289{
 290        return atomic_inc_not_zero(&page->_count);
 291}
 292
 293extern int page_is_ram(unsigned long pfn);
 294
 295/* Support for virtually mapped pages */
 296struct page *vmalloc_to_page(const void *addr);
 297unsigned long vmalloc_to_pfn(const void *addr);
 298
 299/*
 300 * Determine if an address is within the vmalloc range
 301 *
 302 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
 303 * is no special casing required.
 304 */
 305static inline int is_vmalloc_addr(const void *x)
 306{
 307#ifdef CONFIG_MMU
 308        unsigned long addr = (unsigned long)x;
 309
 310        return addr >= VMALLOC_START && addr < VMALLOC_END;
 311#else
 312        return 0;
 313#endif
 314}
 315#ifdef CONFIG_MMU
 316extern int is_vmalloc_or_module_addr(const void *x);
 317#else
 318static inline int is_vmalloc_or_module_addr(const void *x)
 319{
 320        return 0;
 321}
 322#endif
 323
 324static inline void compound_lock(struct page *page)
 325{
 326#ifdef CONFIG_TRANSPARENT_HUGEPAGE
 327        VM_BUG_ON(PageSlab(page));
 328        bit_spin_lock(PG_compound_lock, &page->flags);
 329#endif
 330}
 331
 332static inline void compound_unlock(struct page *page)
 333{
 334#ifdef CONFIG_TRANSPARENT_HUGEPAGE
 335        VM_BUG_ON(PageSlab(page));
 336        bit_spin_unlock(PG_compound_lock, &page->flags);
 337#endif
 338}
 339
 340static inline unsigned long compound_lock_irqsave(struct page *page)
 341{
 342        unsigned long uninitialized_var(flags);
 343#ifdef CONFIG_TRANSPARENT_HUGEPAGE
 344        local_irq_save(flags);
 345        compound_lock(page);
 346#endif
 347        return flags;
 348}
 349
 350static inline void compound_unlock_irqrestore(struct page *page,
 351                                              unsigned long flags)
 352{
 353#ifdef CONFIG_TRANSPARENT_HUGEPAGE
 354        compound_unlock(page);
 355        local_irq_restore(flags);
 356#endif
 357}
 358
 359static inline struct page *compound_head(struct page *page)
 360{
 361        if (unlikely(PageTail(page)))
 362                return page->first_page;
 363        return page;
 364}
 365
 366/*
 367 * The atomic page->_mapcount, starts from -1: so that transitions
 368 * both from it and to it can be tracked, using atomic_inc_and_test
 369 * and atomic_add_negative(-1).
 370 */
 371static inline void page_mapcount_reset(struct page *page)
 372{
 373        atomic_set(&(page)->_mapcount, -1);
 374}
 375
 376static inline int page_mapcount(struct page *page)
 377{
 378        return atomic_read(&(page)->_mapcount) + 1;
 379}
 380
 381static inline int page_count(struct page *page)
 382{
 383        return atomic_read(&compound_head(page)->_count);
 384}
 385
 386static inline void get_huge_page_tail(struct page *page)
 387{
 388        /*
 389         * __split_huge_page_refcount() cannot run
 390         * from under us.
 391         */
 392        VM_BUG_ON(page_mapcount(page) < 0);
 393        VM_BUG_ON(atomic_read(&page->_count) != 0);
 394        atomic_inc(&page->_mapcount);
 395}
 396
 397extern bool __get_page_tail(struct page *page);
 398
 399static inline void get_page(struct page *page)
 400{
 401        if (unlikely(PageTail(page)))
 402                if (likely(__get_page_tail(page)))
 403                        return;
 404        /*
 405         * Getting a normal page or the head of a compound page
 406         * requires to already have an elevated page->_count.
 407         */
 408        VM_BUG_ON(atomic_read(&page->_count) <= 0);
 409        atomic_inc(&page->_count);
 410}
 411
 412static inline struct page *virt_to_head_page(const void *x)
 413{
 414        struct page *page = virt_to_page(x);
 415        return compound_head(page);
 416}
 417
 418/*
 419 * Setup the page count before being freed into the page allocator for
 420 * the first time (boot or memory hotplug)
 421 */
 422static inline void init_page_count(struct page *page)
 423{
 424        atomic_set(&page->_count, 1);
 425}
 426
 427/*
 428 * PageBuddy() indicate that the page is free and in the buddy system
 429 * (see mm/page_alloc.c).
 430 *
 431 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
 432 * -2 so that an underflow of the page_mapcount() won't be mistaken
 433 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
 434 * efficiently by most CPU architectures.
 435 */
 436#define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
 437
 438static inline int PageBuddy(struct page *page)
 439{
 440        return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
 441}
 442
 443static inline void __SetPageBuddy(struct page *page)
 444{
 445        VM_BUG_ON(atomic_read(&page->_mapcount) != -1);
 446        atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
 447}
 448
 449static inline void __ClearPageBuddy(struct page *page)
 450{
 451        VM_BUG_ON(!PageBuddy(page));
 452        atomic_set(&page->_mapcount, -1);
 453}
 454
 455void put_page(struct page *page);
 456void put_pages_list(struct list_head *pages);
 457
 458void split_page(struct page *page, unsigned int order);
 459int split_free_page(struct page *page);
 460
 461/*
 462 * Compound pages have a destructor function.  Provide a
 463 * prototype for that function and accessor functions.
 464 * These are _only_ valid on the head of a PG_compound page.
 465 */
 466typedef void compound_page_dtor(struct page *);
 467
 468static inline void set_compound_page_dtor(struct page *page,
 469                                                compound_page_dtor *dtor)
 470{
 471        page[1].lru.next = (void *)dtor;
 472}
 473
 474static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
 475{
 476        return (compound_page_dtor *)page[1].lru.next;
 477}
 478
 479static inline int compound_order(struct page *page)
 480{
 481        if (!PageHead(page))
 482                return 0;
 483        return (unsigned long)page[1].lru.prev;
 484}
 485
 486static inline int compound_trans_order(struct page *page)
 487{
 488        int order;
 489        unsigned long flags;
 490
 491        if (!PageHead(page))
 492                return 0;
 493
 494        flags = compound_lock_irqsave(page);
 495        order = compound_order(page);
 496        compound_unlock_irqrestore(page, flags);
 497        return order;
 498}
 499
 500static inline void set_compound_order(struct page *page, unsigned long order)
 501{
 502        page[1].lru.prev = (void *)order;
 503}
 504
 505#ifdef CONFIG_MMU
 506/*
 507 * Do pte_mkwrite, but only if the vma says VM_WRITE.  We do this when
 508 * servicing faults for write access.  In the normal case, do always want
 509 * pte_mkwrite.  But get_user_pages can cause write faults for mappings
 510 * that do not have writing enabled, when used by access_process_vm.
 511 */
 512static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
 513{
 514        if (likely(vma->vm_flags & VM_WRITE))
 515                pte = pte_mkwrite(pte);
 516        return pte;
 517}
 518#endif
 519
 520/*
 521 * Multiple processes may "see" the same page. E.g. for untouched
 522 * mappings of /dev/null, all processes see the same page full of
 523 * zeroes, and text pages of executables and shared libraries have
 524 * only one copy in memory, at most, normally.
 525 *
 526 * For the non-reserved pages, page_count(page) denotes a reference count.
 527 *   page_count() == 0 means the page is free. page->lru is then used for
 528 *   freelist management in the buddy allocator.
 529 *   page_count() > 0  means the page has been allocated.
 530 *
 531 * Pages are allocated by the slab allocator in order to provide memory
 532 * to kmalloc and kmem_cache_alloc. In this case, the management of the
 533 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
 534 * unless a particular usage is carefully commented. (the responsibility of
 535 * freeing the kmalloc memory is the caller's, of course).
 536 *
 537 * A page may be used by anyone else who does a __get_free_page().
 538 * In this case, page_count still tracks the references, and should only
 539 * be used through the normal accessor functions. The top bits of page->flags
 540 * and page->virtual store page management information, but all other fields
 541 * are unused and could be used privately, carefully. The management of this
 542 * page is the responsibility of the one who allocated it, and those who have
 543 * subsequently been given references to it.
 544 *
 545 * The other pages (we may call them "pagecache pages") are completely
 546 * managed by the Linux memory manager: I/O, buffers, swapping etc.
 547 * The following discussion applies only to them.
 548 *
 549 * A pagecache page contains an opaque `private' member, which belongs to the
 550 * page's address_space. Usually, this is the address of a circular list of
 551 * the page's disk buffers. PG_private must be set to tell the VM to call
 552 * into the filesystem to release these pages.
 553 *
 554 * A page may belong to an inode's memory mapping. In this case, page->mapping
 555 * is the pointer to the inode, and page->index is the file offset of the page,
 556 * in units of PAGE_CACHE_SIZE.
 557 *
 558 * If pagecache pages are not associated with an inode, they are said to be
 559 * anonymous pages. These may become associated with the swapcache, and in that
 560 * case PG_swapcache is set, and page->private is an offset into the swapcache.
 561 *
 562 * In either case (swapcache or inode backed), the pagecache itself holds one
 563 * reference to the page. Setting PG_private should also increment the
 564 * refcount. The each user mapping also has a reference to the page.
 565 *
 566 * The pagecache pages are stored in a per-mapping radix tree, which is
 567 * rooted at mapping->page_tree, and indexed by offset.
 568 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
 569 * lists, we instead now tag pages as dirty/writeback in the radix tree.
 570 *
 571 * All pagecache pages may be subject to I/O:
 572 * - inode pages may need to be read from disk,
 573 * - inode pages which have been modified and are MAP_SHARED may need
 574 *   to be written back to the inode on disk,
 575 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
 576 *   modified may need to be swapped out to swap space and (later) to be read
 577 *   back into memory.
 578 */
 579
 580/*
 581 * The zone field is never updated after free_area_init_core()
 582 * sets it, so none of the operations on it need to be atomic.
 583 */
 584
 585/* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_NID] | ... | FLAGS | */
 586#define SECTIONS_PGOFF          ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
 587#define NODES_PGOFF             (SECTIONS_PGOFF - NODES_WIDTH)
 588#define ZONES_PGOFF             (NODES_PGOFF - ZONES_WIDTH)
 589#define LAST_NID_PGOFF          (ZONES_PGOFF - LAST_NID_WIDTH)
 590
 591/*
 592 * Define the bit shifts to access each section.  For non-existent
 593 * sections we define the shift as 0; that plus a 0 mask ensures
 594 * the compiler will optimise away reference to them.
 595 */
 596#define SECTIONS_PGSHIFT        (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
 597#define NODES_PGSHIFT           (NODES_PGOFF * (NODES_WIDTH != 0))
 598#define ZONES_PGSHIFT           (ZONES_PGOFF * (ZONES_WIDTH != 0))
 599#define LAST_NID_PGSHIFT        (LAST_NID_PGOFF * (LAST_NID_WIDTH != 0))
 600
 601/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
 602#ifdef NODE_NOT_IN_PAGE_FLAGS
 603#define ZONEID_SHIFT            (SECTIONS_SHIFT + ZONES_SHIFT)
 604#define ZONEID_PGOFF            ((SECTIONS_PGOFF < ZONES_PGOFF)? \
 605                                                SECTIONS_PGOFF : ZONES_PGOFF)
 606#else
 607#define ZONEID_SHIFT            (NODES_SHIFT + ZONES_SHIFT)
 608#define ZONEID_PGOFF            ((NODES_PGOFF < ZONES_PGOFF)? \
 609                                                NODES_PGOFF : ZONES_PGOFF)
 610#endif
 611
 612#define ZONEID_PGSHIFT          (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
 613
 614#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
 615#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
 616#endif
 617
 618#define ZONES_MASK              ((1UL << ZONES_WIDTH) - 1)
 619#define NODES_MASK              ((1UL << NODES_WIDTH) - 1)
 620#define SECTIONS_MASK           ((1UL << SECTIONS_WIDTH) - 1)
 621#define LAST_NID_MASK           ((1UL << LAST_NID_WIDTH) - 1)
 622#define ZONEID_MASK             ((1UL << ZONEID_SHIFT) - 1)
 623
 624static inline enum zone_type page_zonenum(const struct page *page)
 625{
 626        return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
 627}
 628
 629#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
 630#define SECTION_IN_PAGE_FLAGS
 631#endif
 632
 633/*
 634 * The identification function is only used by the buddy allocator for
 635 * determining if two pages could be buddies. We are not really
 636 * identifying a zone since we could be using a the section number
 637 * id if we have not node id available in page flags.
 638 * We guarantee only that it will return the same value for two
 639 * combinable pages in a zone.
 640 */
 641static inline int page_zone_id(struct page *page)
 642{
 643        return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
 644}
 645
 646static inline int zone_to_nid(struct zone *zone)
 647{
 648#ifdef CONFIG_NUMA
 649        return zone->node;
 650#else
 651        return 0;
 652#endif
 653}
 654
 655#ifdef NODE_NOT_IN_PAGE_FLAGS
 656extern int page_to_nid(const struct page *page);
 657#else
 658static inline int page_to_nid(const struct page *page)
 659{
 660        return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
 661}
 662#endif
 663
 664#ifdef CONFIG_NUMA_BALANCING
 665#ifdef LAST_NID_NOT_IN_PAGE_FLAGS
 666static inline int page_nid_xchg_last(struct page *page, int nid)
 667{
 668        return xchg(&page->_last_nid, nid);
 669}
 670
 671static inline int page_nid_last(struct page *page)
 672{
 673        return page->_last_nid;
 674}
 675static inline void page_nid_reset_last(struct page *page)
 676{
 677        page->_last_nid = -1;
 678}
 679#else
 680static inline int page_nid_last(struct page *page)
 681{
 682        return (page->flags >> LAST_NID_PGSHIFT) & LAST_NID_MASK;
 683}
 684
 685extern int page_nid_xchg_last(struct page *page, int nid);
 686
 687static inline void page_nid_reset_last(struct page *page)
 688{
 689        int nid = (1 << LAST_NID_SHIFT) - 1;
 690
 691        page->flags &= ~(LAST_NID_MASK << LAST_NID_PGSHIFT);
 692        page->flags |= (nid & LAST_NID_MASK) << LAST_NID_PGSHIFT;
 693}
 694#endif /* LAST_NID_NOT_IN_PAGE_FLAGS */
 695#else
 696static inline int page_nid_xchg_last(struct page *page, int nid)
 697{
 698        return page_to_nid(page);
 699}
 700
 701static inline int page_nid_last(struct page *page)
 702{
 703        return page_to_nid(page);
 704}
 705
 706static inline void page_nid_reset_last(struct page *page)
 707{
 708}
 709#endif
 710
 711static inline struct zone *page_zone(const struct page *page)
 712{
 713        return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
 714}
 715
 716#ifdef SECTION_IN_PAGE_FLAGS
 717static inline void set_page_section(struct page *page, unsigned long section)
 718{
 719        page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
 720        page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
 721}
 722
 723static inline unsigned long page_to_section(const struct page *page)
 724{
 725        return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
 726}
 727#endif
 728
 729static inline void set_page_zone(struct page *page, enum zone_type zone)
 730{
 731        page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
 732        page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
 733}
 734
 735static inline void set_page_node(struct page *page, unsigned long node)
 736{
 737        page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
 738        page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
 739}
 740
 741static inline void set_page_links(struct page *page, enum zone_type zone,
 742        unsigned long node, unsigned long pfn)
 743{
 744        set_page_zone(page, zone);
 745        set_page_node(page, node);
 746#ifdef SECTION_IN_PAGE_FLAGS
 747        set_page_section(page, pfn_to_section_nr(pfn));
 748#endif
 749}
 750
 751/*
 752 * Some inline functions in vmstat.h depend on page_zone()
 753 */
 754#include <linux/vmstat.h>
 755
 756static __always_inline void *lowmem_page_address(const struct page *page)
 757{
 758        return __va(PFN_PHYS(page_to_pfn(page)));
 759}
 760
 761#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
 762#define HASHED_PAGE_VIRTUAL
 763#endif
 764
 765#if defined(WANT_PAGE_VIRTUAL)
 766#define page_address(page) ((page)->virtual)
 767#define set_page_address(page, address)                 \
 768        do {                                            \
 769                (page)->virtual = (address);            \
 770        } while(0)
 771#define page_address_init()  do { } while(0)
 772#endif
 773
 774#if defined(HASHED_PAGE_VIRTUAL)
 775void *page_address(const struct page *page);
 776void set_page_address(struct page *page, void *virtual);
 777void page_address_init(void);
 778#endif
 779
 780#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
 781#define page_address(page) lowmem_page_address(page)
 782#define set_page_address(page, address)  do { } while(0)
 783#define page_address_init()  do { } while(0)
 784#endif
 785
 786/*
 787 * On an anonymous page mapped into a user virtual memory area,
 788 * page->mapping points to its anon_vma, not to a struct address_space;
 789 * with the PAGE_MAPPING_ANON bit set to distinguish it.  See rmap.h.
 790 *
 791 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
 792 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
 793 * and then page->mapping points, not to an anon_vma, but to a private
 794 * structure which KSM associates with that merged page.  See ksm.h.
 795 *
 796 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
 797 *
 798 * Please note that, confusingly, "page_mapping" refers to the inode
 799 * address_space which maps the page from disk; whereas "page_mapped"
 800 * refers to user virtual address space into which the page is mapped.
 801 */
 802#define PAGE_MAPPING_ANON       1
 803#define PAGE_MAPPING_KSM        2
 804#define PAGE_MAPPING_FLAGS      (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
 805
 806extern struct address_space *page_mapping(struct page *page);
 807
 808/* Neutral page->mapping pointer to address_space or anon_vma or other */
 809static inline void *page_rmapping(struct page *page)
 810{
 811        return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
 812}
 813
 814extern struct address_space *__page_file_mapping(struct page *);
 815
 816static inline
 817struct address_space *page_file_mapping(struct page *page)
 818{
 819        if (unlikely(PageSwapCache(page)))
 820                return __page_file_mapping(page);
 821
 822        return page->mapping;
 823}
 824
 825static inline int PageAnon(struct page *page)
 826{
 827        return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
 828}
 829
 830/*
 831 * Return the pagecache index of the passed page.  Regular pagecache pages
 832 * use ->index whereas swapcache pages use ->private
 833 */
 834static inline pgoff_t page_index(struct page *page)
 835{
 836        if (unlikely(PageSwapCache(page)))
 837                return page_private(page);
 838        return page->index;
 839}
 840
 841extern pgoff_t __page_file_index(struct page *page);
 842
 843/*
 844 * Return the file index of the page. Regular pagecache pages use ->index
 845 * whereas swapcache pages use swp_offset(->private)
 846 */
 847static inline pgoff_t page_file_index(struct page *page)
 848{
 849        if (unlikely(PageSwapCache(page)))
 850                return __page_file_index(page);
 851
 852        return page->index;
 853}
 854
 855/*
 856 * Return true if this page is mapped into pagetables.
 857 */
 858static inline int page_mapped(struct page *page)
 859{
 860        return atomic_read(&(page)->_mapcount) >= 0;
 861}
 862
 863/*
 864 * Different kinds of faults, as returned by handle_mm_fault().
 865 * Used to decide whether a process gets delivered SIGBUS or
 866 * just gets major/minor fault counters bumped up.
 867 */
 868
 869#define VM_FAULT_MINOR  0 /* For backwards compat. Remove me quickly. */
 870
 871#define VM_FAULT_OOM    0x0001
 872#define VM_FAULT_SIGBUS 0x0002
 873#define VM_FAULT_MAJOR  0x0004
 874#define VM_FAULT_WRITE  0x0008  /* Special case for get_user_pages */
 875#define VM_FAULT_HWPOISON 0x0010        /* Hit poisoned small page */
 876#define VM_FAULT_HWPOISON_LARGE 0x0020  /* Hit poisoned large page. Index encoded in upper bits */
 877
 878#define VM_FAULT_NOPAGE 0x0100  /* ->fault installed the pte, not return page */
 879#define VM_FAULT_LOCKED 0x0200  /* ->fault locked the returned page */
 880#define VM_FAULT_RETRY  0x0400  /* ->fault blocked, must retry */
 881
 882#define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
 883
 884#define VM_FAULT_ERROR  (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
 885                         VM_FAULT_HWPOISON_LARGE)
 886
 887/* Encode hstate index for a hwpoisoned large page */
 888#define VM_FAULT_SET_HINDEX(x) ((x) << 12)
 889#define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
 890
 891/*
 892 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
 893 */
 894extern void pagefault_out_of_memory(void);
 895
 896#define offset_in_page(p)       ((unsigned long)(p) & ~PAGE_MASK)
 897
 898/*
 899 * Flags passed to show_mem() and show_free_areas() to suppress output in
 900 * various contexts.
 901 */
 902#define SHOW_MEM_FILTER_NODES   (0x0001u)       /* filter disallowed nodes */
 903
 904extern void show_free_areas(unsigned int flags);
 905extern bool skip_free_areas_node(unsigned int flags, int nid);
 906
 907int shmem_zero_setup(struct vm_area_struct *);
 908
 909extern int can_do_mlock(void);
 910extern int user_shm_lock(size_t, struct user_struct *);
 911extern void user_shm_unlock(size_t, struct user_struct *);
 912
 913/*
 914 * Parameter block passed down to zap_pte_range in exceptional cases.
 915 */
 916struct zap_details {
 917        struct vm_area_struct *nonlinear_vma;   /* Check page->index if set */
 918        struct address_space *check_mapping;    /* Check page->mapping if set */
 919        pgoff_t first_index;                    /* Lowest page->index to unmap */
 920        pgoff_t last_index;                     /* Highest page->index to unmap */
 921};
 922
 923struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
 924                pte_t pte);
 925
 926int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
 927                unsigned long size);
 928void zap_page_range(struct vm_area_struct *vma, unsigned long address,
 929                unsigned long size, struct zap_details *);
 930void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
 931                unsigned long start, unsigned long end);
 932
 933/**
 934 * mm_walk - callbacks for walk_page_range
 935 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
 936 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
 937 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
 938 *             this handler is required to be able to handle
 939 *             pmd_trans_huge() pmds.  They may simply choose to
 940 *             split_huge_page() instead of handling it explicitly.
 941 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
 942 * @pte_hole: if set, called for each hole at all levels
 943 * @hugetlb_entry: if set, called for each hugetlb entry
 944 *                 *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
 945 *                            is used.
 946 *
 947 * (see walk_page_range for more details)
 948 */
 949struct mm_walk {
 950        int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
 951        int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
 952        int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
 953        int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
 954        int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
 955        int (*hugetlb_entry)(pte_t *, unsigned long,
 956                             unsigned long, unsigned long, struct mm_walk *);
 957        struct mm_struct *mm;
 958        void *private;
 959};
 960
 961int walk_page_range(unsigned long addr, unsigned long end,
 962                struct mm_walk *walk);
 963void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
 964                unsigned long end, unsigned long floor, unsigned long ceiling);
 965int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
 966                        struct vm_area_struct *vma);
 967void unmap_mapping_range(struct address_space *mapping,
 968                loff_t const holebegin, loff_t const holelen, int even_cows);
 969int follow_pfn(struct vm_area_struct *vma, unsigned long address,
 970        unsigned long *pfn);
 971int follow_phys(struct vm_area_struct *vma, unsigned long address,
 972                unsigned int flags, unsigned long *prot, resource_size_t *phys);
 973int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
 974                        void *buf, int len, int write);
 975
 976static inline void unmap_shared_mapping_range(struct address_space *mapping,
 977                loff_t const holebegin, loff_t const holelen)
 978{
 979        unmap_mapping_range(mapping, holebegin, holelen, 0);
 980}
 981
 982extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new);
 983extern void truncate_setsize(struct inode *inode, loff_t newsize);
 984void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
 985int truncate_inode_page(struct address_space *mapping, struct page *page);
 986int generic_error_remove_page(struct address_space *mapping, struct page *page);
 987int invalidate_inode_page(struct page *page);
 988
 989#ifdef CONFIG_MMU
 990extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
 991                        unsigned long address, unsigned int flags);
 992extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
 993                            unsigned long address, unsigned int fault_flags);
 994#else
 995static inline int handle_mm_fault(struct mm_struct *mm,
 996                        struct vm_area_struct *vma, unsigned long address,
 997                        unsigned int flags)
 998{
 999        /* should never happen if there's no MMU */
1000        BUG();
1001        return VM_FAULT_SIGBUS;
1002}
1003static inline int fixup_user_fault(struct task_struct *tsk,
1004                struct mm_struct *mm, unsigned long address,
1005                unsigned int fault_flags)
1006{
1007        /* should never happen if there's no MMU */
1008        BUG();
1009        return -EFAULT;
1010}
1011#endif
1012
1013extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1014extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1015                void *buf, int len, int write);
1016
1017long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1018                      unsigned long start, unsigned long nr_pages,
1019                      unsigned int foll_flags, struct page **pages,
1020                      struct vm_area_struct **vmas, int *nonblocking);
1021long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1022                    unsigned long start, unsigned long nr_pages,
1023                    int write, int force, struct page **pages,
1024                    struct vm_area_struct **vmas);
1025int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1026                        struct page **pages);
1027struct kvec;
1028int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1029                        struct page **pages);
1030int get_kernel_page(unsigned long start, int write, struct page **pages);
1031struct page *get_dump_page(unsigned long addr);
1032
1033extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1034extern void do_invalidatepage(struct page *page, unsigned long offset);
1035
1036int __set_page_dirty_nobuffers(struct page *page);
1037int __set_page_dirty_no_writeback(struct page *page);
1038int redirty_page_for_writepage(struct writeback_control *wbc,
1039                                struct page *page);
1040void account_page_dirtied(struct page *page, struct address_space *mapping);
1041void account_page_writeback(struct page *page);
1042int set_page_dirty(struct page *page);
1043int set_page_dirty_lock(struct page *page);
1044int clear_page_dirty_for_io(struct page *page);
1045
1046/* Is the vma a continuation of the stack vma above it? */
1047static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1048{
1049        return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1050}
1051
1052static inline int stack_guard_page_start(struct vm_area_struct *vma,
1053                                             unsigned long addr)
1054{
1055        return (vma->vm_flags & VM_GROWSDOWN) &&
1056                (vma->vm_start == addr) &&
1057                !vma_growsdown(vma->vm_prev, addr);
1058}
1059
1060/* Is the vma a continuation of the stack vma below it? */
1061static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1062{
1063        return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1064}
1065
1066static inline int stack_guard_page_end(struct vm_area_struct *vma,
1067                                           unsigned long addr)
1068{
1069        return (vma->vm_flags & VM_GROWSUP) &&
1070                (vma->vm_end == addr) &&
1071                !vma_growsup(vma->vm_next, addr);
1072}
1073
1074extern pid_t
1075vm_is_stack(struct task_struct *task, struct vm_area_struct *vma, int in_group);
1076
1077extern unsigned long move_page_tables(struct vm_area_struct *vma,
1078                unsigned long old_addr, struct vm_area_struct *new_vma,
1079                unsigned long new_addr, unsigned long len,
1080                bool need_rmap_locks);
1081extern unsigned long do_mremap(unsigned long addr,
1082                               unsigned long old_len, unsigned long new_len,
1083                               unsigned long flags, unsigned long new_addr);
1084extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1085                              unsigned long end, pgprot_t newprot,
1086                              int dirty_accountable, int prot_numa);
1087extern int mprotect_fixup(struct vm_area_struct *vma,
1088                          struct vm_area_struct **pprev, unsigned long start,
1089                          unsigned long end, unsigned long newflags);
1090
1091/*
1092 * doesn't attempt to fault and will return short.
1093 */
1094int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1095                          struct page **pages);
1096/*
1097 * per-process(per-mm_struct) statistics.
1098 */
1099static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1100{
1101        long val = atomic_long_read(&mm->rss_stat.count[member]);
1102
1103#ifdef SPLIT_RSS_COUNTING
1104        /*
1105         * counter is updated in asynchronous manner and may go to minus.
1106         * But it's never be expected number for users.
1107         */
1108        if (val < 0)
1109                val = 0;
1110#endif
1111        return (unsigned long)val;
1112}
1113
1114static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1115{
1116        atomic_long_add(value, &mm->rss_stat.count[member]);
1117}
1118
1119static inline void inc_mm_counter(struct mm_struct *mm, int member)
1120{
1121        atomic_long_inc(&mm->rss_stat.count[member]);
1122}
1123
1124static inline void dec_mm_counter(struct mm_struct *mm, int member)
1125{
1126        atomic_long_dec(&mm->rss_stat.count[member]);
1127}
1128
1129static inline unsigned long get_mm_rss(struct mm_struct *mm)
1130{
1131        return get_mm_counter(mm, MM_FILEPAGES) +
1132                get_mm_counter(mm, MM_ANONPAGES);
1133}
1134
1135static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1136{
1137        return max(mm->hiwater_rss, get_mm_rss(mm));
1138}
1139
1140static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1141{
1142        return max(mm->hiwater_vm, mm->total_vm);
1143}
1144
1145static inline void update_hiwater_rss(struct mm_struct *mm)
1146{
1147        unsigned long _rss = get_mm_rss(mm);
1148
1149        if ((mm)->hiwater_rss < _rss)
1150                (mm)->hiwater_rss = _rss;
1151}
1152
1153static inline void update_hiwater_vm(struct mm_struct *mm)
1154{
1155        if (mm->hiwater_vm < mm->total_vm)
1156                mm->hiwater_vm = mm->total_vm;
1157}
1158
1159static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1160                                         struct mm_struct *mm)
1161{
1162        unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1163
1164        if (*maxrss < hiwater_rss)
1165                *maxrss = hiwater_rss;
1166}
1167
1168#if defined(SPLIT_RSS_COUNTING)
1169void sync_mm_rss(struct mm_struct *mm);
1170#else
1171static inline void sync_mm_rss(struct mm_struct *mm)
1172{
1173}
1174#endif
1175
1176int vma_wants_writenotify(struct vm_area_struct *vma);
1177
1178extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1179                               spinlock_t **ptl);
1180static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1181                                    spinlock_t **ptl)
1182{
1183        pte_t *ptep;
1184        __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1185        return ptep;
1186}
1187
1188#ifdef __PAGETABLE_PUD_FOLDED
1189static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1190                                                unsigned long address)
1191{
1192        return 0;
1193}
1194#else
1195int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1196#endif
1197
1198#ifdef __PAGETABLE_PMD_FOLDED
1199static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1200                                                unsigned long address)
1201{
1202        return 0;
1203}
1204#else
1205int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1206#endif
1207
1208int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1209                pmd_t *pmd, unsigned long address);
1210int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1211
1212/*
1213 * The following ifdef needed to get the 4level-fixup.h header to work.
1214 * Remove it when 4level-fixup.h has been removed.
1215 */
1216#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1217static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1218{
1219        return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1220                NULL: pud_offset(pgd, address);
1221}
1222
1223static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1224{
1225        return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1226                NULL: pmd_offset(pud, address);
1227}
1228#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1229
1230#if USE_SPLIT_PTLOCKS
1231/*
1232 * We tuck a spinlock to guard each pagetable page into its struct page,
1233 * at page->private, with BUILD_BUG_ON to make sure that this will not
1234 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1235 * When freeing, reset page->mapping so free_pages_check won't complain.
1236 */
1237#define __pte_lockptr(page)     &((page)->ptl)
1238#define pte_lock_init(_page)    do {                                    \
1239        spin_lock_init(__pte_lockptr(_page));                           \
1240} while (0)
1241#define pte_lock_deinit(page)   ((page)->mapping = NULL)
1242#define pte_lockptr(mm, pmd)    ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
1243#else   /* !USE_SPLIT_PTLOCKS */
1244/*
1245 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1246 */
1247#define pte_lock_init(page)     do {} while (0)
1248#define pte_lock_deinit(page)   do {} while (0)
1249#define pte_lockptr(mm, pmd)    ({(void)(pmd); &(mm)->page_table_lock;})
1250#endif /* USE_SPLIT_PTLOCKS */
1251
1252static inline void pgtable_page_ctor(struct page *page)
1253{
1254        pte_lock_init(page);
1255        inc_zone_page_state(page, NR_PAGETABLE);
1256}
1257
1258static inline void pgtable_page_dtor(struct page *page)
1259{
1260        pte_lock_deinit(page);
1261        dec_zone_page_state(page, NR_PAGETABLE);
1262}
1263
1264#define pte_offset_map_lock(mm, pmd, address, ptlp)     \
1265({                                                      \
1266        spinlock_t *__ptl = pte_lockptr(mm, pmd);       \
1267        pte_t *__pte = pte_offset_map(pmd, address);    \
1268        *(ptlp) = __ptl;                                \
1269        spin_lock(__ptl);                               \
1270        __pte;                                          \
1271})
1272
1273#define pte_unmap_unlock(pte, ptl)      do {            \
1274        spin_unlock(ptl);                               \
1275        pte_unmap(pte);                                 \
1276} while (0)
1277
1278#define pte_alloc_map(mm, vma, pmd, address)                            \
1279        ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma,    \
1280                                                        pmd, address))? \
1281         NULL: pte_offset_map(pmd, address))
1282
1283#define pte_alloc_map_lock(mm, pmd, address, ptlp)      \
1284        ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL,   \
1285                                                        pmd, address))? \
1286                NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1287
1288#define pte_alloc_kernel(pmd, address)                  \
1289        ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1290                NULL: pte_offset_kernel(pmd, address))
1291
1292extern void free_area_init(unsigned long * zones_size);
1293extern void free_area_init_node(int nid, unsigned long * zones_size,
1294                unsigned long zone_start_pfn, unsigned long *zholes_size);
1295extern void free_initmem(void);
1296
1297#ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1298/*
1299 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1300 * zones, allocate the backing mem_map and account for memory holes in a more
1301 * architecture independent manner. This is a substitute for creating the
1302 * zone_sizes[] and zholes_size[] arrays and passing them to
1303 * free_area_init_node()
1304 *
1305 * An architecture is expected to register range of page frames backed by
1306 * physical memory with memblock_add[_node]() before calling
1307 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1308 * usage, an architecture is expected to do something like
1309 *
1310 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1311 *                                                       max_highmem_pfn};
1312 * for_each_valid_physical_page_range()
1313 *      memblock_add_node(base, size, nid)
1314 * free_area_init_nodes(max_zone_pfns);
1315 *
1316 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1317 * registered physical page range.  Similarly
1318 * sparse_memory_present_with_active_regions() calls memory_present() for
1319 * each range when SPARSEMEM is enabled.
1320 *
1321 * See mm/page_alloc.c for more information on each function exposed by
1322 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1323 */
1324extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1325unsigned long node_map_pfn_alignment(void);
1326unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1327                                                unsigned long end_pfn);
1328extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1329                                                unsigned long end_pfn);
1330extern void get_pfn_range_for_nid(unsigned int nid,
1331                        unsigned long *start_pfn, unsigned long *end_pfn);
1332extern unsigned long find_min_pfn_with_active_regions(void);
1333extern void free_bootmem_with_active_regions(int nid,
1334                                                unsigned long max_low_pfn);
1335extern void sparse_memory_present_with_active_regions(int nid);
1336
1337#endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1338
1339#if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1340    !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1341static inline int __early_pfn_to_nid(unsigned long pfn)
1342{
1343        return 0;
1344}
1345#else
1346/* please see mm/page_alloc.c */
1347extern int __meminit early_pfn_to_nid(unsigned long pfn);
1348#ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1349/* there is a per-arch backend function. */
1350extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1351#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1352#endif
1353
1354extern void set_dma_reserve(unsigned long new_dma_reserve);
1355extern void memmap_init_zone(unsigned long, int, unsigned long,
1356                                unsigned long, enum memmap_context);
1357extern void setup_per_zone_wmarks(void);
1358extern int __meminit init_per_zone_wmark_min(void);
1359extern void mem_init(void);
1360extern void __init mmap_init(void);
1361extern void show_mem(unsigned int flags);
1362extern void si_meminfo(struct sysinfo * val);
1363extern void si_meminfo_node(struct sysinfo *val, int nid);
1364
1365extern __printf(3, 4)
1366void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1367
1368extern void setup_per_cpu_pageset(void);
1369
1370extern void zone_pcp_update(struct zone *zone);
1371extern void zone_pcp_reset(struct zone *zone);
1372
1373/* page_alloc.c */
1374extern int min_free_kbytes;
1375
1376/* nommu.c */
1377extern atomic_long_t mmap_pages_allocated;
1378extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1379
1380/* interval_tree.c */
1381void vma_interval_tree_insert(struct vm_area_struct *node,
1382                              struct rb_root *root);
1383void vma_interval_tree_insert_after(struct vm_area_struct *node,
1384                                    struct vm_area_struct *prev,
1385                                    struct rb_root *root);
1386void vma_interval_tree_remove(struct vm_area_struct *node,
1387                              struct rb_root *root);
1388struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1389                                unsigned long start, unsigned long last);
1390struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1391                                unsigned long start, unsigned long last);
1392
1393#define vma_interval_tree_foreach(vma, root, start, last)               \
1394        for (vma = vma_interval_tree_iter_first(root, start, last);     \
1395             vma; vma = vma_interval_tree_iter_next(vma, start, last))
1396
1397static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1398                                        struct list_head *list)
1399{
1400        list_add_tail(&vma->shared.nonlinear, list);
1401}
1402
1403void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1404                                   struct rb_root *root);
1405void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1406                                   struct rb_root *root);
1407struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1408        struct rb_root *root, unsigned long start, unsigned long last);
1409struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1410        struct anon_vma_chain *node, unsigned long start, unsigned long last);
1411#ifdef CONFIG_DEBUG_VM_RB
1412void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1413#endif
1414
1415#define anon_vma_interval_tree_foreach(avc, root, start, last)           \
1416        for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1417             avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1418
1419/* mmap.c */
1420extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1421extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1422        unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1423extern struct vm_area_struct *vma_merge(struct mm_struct *,
1424        struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1425        unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1426        struct mempolicy *);
1427extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1428extern int split_vma(struct mm_struct *,
1429        struct vm_area_struct *, unsigned long addr, int new_below);
1430extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1431extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1432        struct rb_node **, struct rb_node *);
1433extern void unlink_file_vma(struct vm_area_struct *);
1434extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1435        unsigned long addr, unsigned long len, pgoff_t pgoff,
1436        bool *need_rmap_locks);
1437extern void exit_mmap(struct mm_struct *);
1438
1439extern int mm_take_all_locks(struct mm_struct *mm);
1440extern void mm_drop_all_locks(struct mm_struct *mm);
1441
1442extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1443extern struct file *get_mm_exe_file(struct mm_struct *mm);
1444
1445extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1446extern int install_special_mapping(struct mm_struct *mm,
1447                                   unsigned long addr, unsigned long len,
1448                                   unsigned long flags, struct page **pages);
1449
1450extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1451
1452extern unsigned long mmap_region(struct file *file, unsigned long addr,
1453        unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
1454extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1455        unsigned long len, unsigned long prot, unsigned long flags,
1456        unsigned long pgoff, unsigned long *populate);
1457extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1458
1459#ifdef CONFIG_MMU
1460extern int __mm_populate(unsigned long addr, unsigned long len,
1461                         int ignore_errors);
1462static inline void mm_populate(unsigned long addr, unsigned long len)
1463{
1464        /* Ignore errors */
1465        (void) __mm_populate(addr, len, 1);
1466}
1467#else
1468static inline void mm_populate(unsigned long addr, unsigned long len) {}
1469#endif
1470
1471/* These take the mm semaphore themselves */
1472extern unsigned long vm_brk(unsigned long, unsigned long);
1473extern int vm_munmap(unsigned long, size_t);
1474extern unsigned long vm_mmap(struct file *, unsigned long,
1475        unsigned long, unsigned long,
1476        unsigned long, unsigned long);
1477
1478struct vm_unmapped_area_info {
1479#define VM_UNMAPPED_AREA_TOPDOWN 1
1480        unsigned long flags;
1481        unsigned long length;
1482        unsigned long low_limit;
1483        unsigned long high_limit;
1484        unsigned long align_mask;
1485        unsigned long align_offset;
1486};
1487
1488extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
1489extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
1490
1491/*
1492 * Search for an unmapped address range.
1493 *
1494 * We are looking for a range that:
1495 * - does not intersect with any VMA;
1496 * - is contained within the [low_limit, high_limit) interval;
1497 * - is at least the desired size.
1498 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1499 */
1500static inline unsigned long
1501vm_unmapped_area(struct vm_unmapped_area_info *info)
1502{
1503        if (!(info->flags & VM_UNMAPPED_AREA_TOPDOWN))
1504                return unmapped_area(info);
1505        else
1506                return unmapped_area_topdown(info);
1507}
1508
1509/* truncate.c */
1510extern void truncate_inode_pages(struct address_space *, loff_t);
1511extern void truncate_inode_pages_range(struct address_space *,
1512                                       loff_t lstart, loff_t lend);
1513
1514/* generic vm_area_ops exported for stackable file systems */
1515extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1516extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1517
1518/* mm/page-writeback.c */
1519int write_one_page(struct page *page, int wait);
1520void task_dirty_inc(struct task_struct *tsk);
1521
1522/* readahead.c */
1523#define VM_MAX_READAHEAD        128     /* kbytes */
1524#define VM_MIN_READAHEAD        16      /* kbytes (includes current page) */
1525
1526int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1527                        pgoff_t offset, unsigned long nr_to_read);
1528
1529void page_cache_sync_readahead(struct address_space *mapping,
1530                               struct file_ra_state *ra,
1531                               struct file *filp,
1532                               pgoff_t offset,
1533                               unsigned long size);
1534
1535void page_cache_async_readahead(struct address_space *mapping,
1536                                struct file_ra_state *ra,
1537                                struct file *filp,
1538                                struct page *pg,
1539                                pgoff_t offset,
1540                                unsigned long size);
1541
1542unsigned long max_sane_readahead(unsigned long nr);
1543unsigned long ra_submit(struct file_ra_state *ra,
1544                        struct address_space *mapping,
1545                        struct file *filp);
1546
1547/* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1548extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1549
1550/* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1551extern int expand_downwards(struct vm_area_struct *vma,
1552                unsigned long address);
1553#if VM_GROWSUP
1554extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1555#else
1556  #define expand_upwards(vma, address) do { } while (0)
1557#endif
1558
1559/* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
1560extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1561extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1562                                             struct vm_area_struct **pprev);
1563
1564/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1565   NULL if none.  Assume start_addr < end_addr. */
1566static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1567{
1568        struct vm_area_struct * vma = find_vma(mm,start_addr);
1569
1570        if (vma && end_addr <= vma->vm_start)
1571                vma = NULL;
1572        return vma;
1573}
1574
1575static inline unsigned long vma_pages(struct vm_area_struct *vma)
1576{
1577        return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1578}
1579
1580/* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1581static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
1582                                unsigned long vm_start, unsigned long vm_end)
1583{
1584        struct vm_area_struct *vma = find_vma(mm, vm_start);
1585
1586        if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
1587                vma = NULL;
1588
1589        return vma;
1590}
1591
1592#ifdef CONFIG_MMU
1593pgprot_t vm_get_page_prot(unsigned long vm_flags);
1594#else
1595static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1596{
1597        return __pgprot(0);
1598}
1599#endif
1600
1601#ifdef CONFIG_ARCH_USES_NUMA_PROT_NONE
1602unsigned long change_prot_numa(struct vm_area_struct *vma,
1603                        unsigned long start, unsigned long end);
1604#endif
1605
1606struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1607int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1608                        unsigned long pfn, unsigned long size, pgprot_t);
1609int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1610int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1611                        unsigned long pfn);
1612int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1613                        unsigned long pfn);
1614int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
1615
1616
1617struct page *follow_page_mask(struct vm_area_struct *vma,
1618                              unsigned long address, unsigned int foll_flags,
1619                              unsigned int *page_mask);
1620
1621static inline struct page *follow_page(struct vm_area_struct *vma,
1622                unsigned long address, unsigned int foll_flags)
1623{
1624        unsigned int unused_page_mask;
1625        return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
1626}
1627
1628#define FOLL_WRITE      0x01    /* check pte is writable */
1629#define FOLL_TOUCH      0x02    /* mark page accessed */
1630#define FOLL_GET        0x04    /* do get_page on page */
1631#define FOLL_DUMP       0x08    /* give error on hole if it would be zero */
1632#define FOLL_FORCE      0x10    /* get_user_pages read/write w/o permission */
1633#define FOLL_NOWAIT     0x20    /* if a disk transfer is needed, start the IO
1634                                 * and return without waiting upon it */
1635#define FOLL_MLOCK      0x40    /* mark page as mlocked */
1636#define FOLL_SPLIT      0x80    /* don't return transhuge pages, split them */
1637#define FOLL_HWPOISON   0x100   /* check page is hwpoisoned */
1638#define FOLL_NUMA       0x200   /* force NUMA hinting page fault */
1639#define FOLL_MIGRATION  0x400   /* wait for page to replace migration entry */
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#ifdef CONFIG_MEMORY_HOTPLUG
1712void vmemmap_free(struct page *memmap, unsigned long nr_pages);
1713#endif
1714void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
1715                                  unsigned long size);
1716
1717enum mf_flags {
1718        MF_COUNT_INCREASED = 1 << 0,
1719        MF_ACTION_REQUIRED = 1 << 1,
1720        MF_MUST_KILL = 1 << 2,
1721};
1722extern int memory_failure(unsigned long pfn, int trapno, int flags);
1723extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
1724extern int unpoison_memory(unsigned long pfn);
1725extern int sysctl_memory_failure_early_kill;
1726extern int sysctl_memory_failure_recovery;
1727extern void shake_page(struct page *p, int access);
1728extern atomic_long_t num_poisoned_pages;
1729extern int soft_offline_page(struct page *page, int flags);
1730
1731extern void dump_page(struct page *page);
1732
1733#if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1734extern void clear_huge_page(struct page *page,
1735                            unsigned long addr,
1736                            unsigned int pages_per_huge_page);
1737extern void copy_user_huge_page(struct page *dst, struct page *src,
1738                                unsigned long addr, struct vm_area_struct *vma,
1739                                unsigned int pages_per_huge_page);
1740#endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1741
1742#ifdef CONFIG_DEBUG_PAGEALLOC
1743extern unsigned int _debug_guardpage_minorder;
1744
1745static inline unsigned int debug_guardpage_minorder(void)
1746{
1747        return _debug_guardpage_minorder;
1748}
1749
1750static inline bool page_is_guard(struct page *page)
1751{
1752        return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags);
1753}
1754#else
1755static inline unsigned int debug_guardpage_minorder(void) { return 0; }
1756static inline bool page_is_guard(struct page *page) { return false; }
1757#endif /* CONFIG_DEBUG_PAGEALLOC */
1758
1759#endif /* __KERNEL__ */
1760#endif /* _LINUX_MM_H */
1761
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.