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/list.h>
  10#include <linux/mmzone.h>
  11#include <linux/rbtree.h>
  12#include <linux/prio_tree.h>
  13#include <linux/debug_locks.h>
  14#include <linux/mm_types.h>
  15
  16struct mempolicy;
  17struct anon_vma;
  18struct file_ra_state;
  19struct user_struct;
  20struct writeback_control;
  21struct rlimit;
  22
  23#ifndef CONFIG_DISCONTIGMEM          /* Don't use mapnrs, do it properly */
  24extern unsigned long max_mapnr;
  25#endif
  26
  27extern unsigned long num_physpages;
  28extern unsigned long totalram_pages;
  29extern void * high_memory;
  30extern int page_cluster;
  31
  32#ifdef CONFIG_SYSCTL
  33extern int sysctl_legacy_va_layout;
  34#else
  35#define sysctl_legacy_va_layout 0
  36#endif
  37
  38#include <asm/page.h>
  39#include <asm/pgtable.h>
  40#include <asm/processor.h>
  41
  42#define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
  43
  44/* to align the pointer to the (next) page boundary */
  45#define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
  46
  47/*
  48 * Linux kernel virtual memory manager primitives.
  49 * The idea being to have a "virtual" mm in the same way
  50 * we have a virtual fs - giving a cleaner interface to the
  51 * mm details, and allowing different kinds of memory mappings
  52 * (from shared memory to executable loading to arbitrary
  53 * mmap() functions).
  54 */
  55
  56extern struct kmem_cache *vm_area_cachep;
  57
  58#ifndef CONFIG_MMU
  59extern struct rb_root nommu_region_tree;
  60extern struct rw_semaphore nommu_region_sem;
  61
  62extern unsigned int kobjsize(const void *objp);
  63#endif
  64
  65/*
  66 * vm_flags in vm_area_struct, see mm_types.h.
  67 */
  68#define VM_READ         0x00000001      /* currently active flags */
  69#define VM_WRITE        0x00000002
  70#define VM_EXEC         0x00000004
  71#define VM_SHARED       0x00000008
  72
  73/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
  74#define VM_MAYREAD      0x00000010      /* limits for mprotect() etc */
  75#define VM_MAYWRITE     0x00000020
  76#define VM_MAYEXEC      0x00000040
  77#define VM_MAYSHARE     0x00000080
  78
  79#define VM_GROWSDOWN    0x00000100      /* general info on the segment */
  80#define VM_GROWSUP      0x00000200
  81#define VM_PFNMAP       0x00000400      /* Page-ranges managed without "struct page", just pure PFN */
  82#define VM_DENYWRITE    0x00000800      /* ETXTBSY on write attempts.. */
  83
  84#define VM_EXECUTABLE   0x00001000
  85#define VM_LOCKED       0x00002000
  86#define VM_IO           0x00004000      /* Memory mapped I/O or similar */
  87
  88                                        /* Used by sys_madvise() */
  89#define VM_SEQ_READ     0x00008000      /* App will access data sequentially */
  90#define VM_RAND_READ    0x00010000      /* App will not benefit from clustered reads */
  91
  92#define VM_DONTCOPY     0x00020000      /* Do not copy this vma on fork */
  93#define VM_DONTEXPAND   0x00040000      /* Cannot expand with mremap() */
  94#define VM_RESERVED     0x00080000      /* Count as reserved_vm like IO */
  95#define VM_ACCOUNT      0x00100000      /* Is a VM accounted object */
  96#define VM_NORESERVE    0x00200000      /* should the VM suppress accounting */
  97#define VM_HUGETLB      0x00400000      /* Huge TLB Page VM */
  98#define VM_NONLINEAR    0x00800000      /* Is non-linear (remap_file_pages) */
  99#define VM_MAPPED_COPY  0x01000000      /* T if mapped copy of data (nommu mmap) */
 100#define VM_INSERTPAGE   0x02000000      /* The vma has had "vm_insert_page()" done on it */
 101#define VM_ALWAYSDUMP   0x04000000      /* Always include in core dumps */
 102
 103#define VM_CAN_NONLINEAR 0x08000000     /* Has ->fault & does nonlinear pages */
 104#define VM_MIXEDMAP     0x10000000      /* Can contain "struct page" and pure PFN pages */
 105#define VM_SAO          0x20000000      /* Strong Access Ordering (powerpc) */
 106#define VM_PFN_AT_MMAP  0x40000000      /* PFNMAP vma that is fully mapped at mmap time */
 107#define VM_MERGEABLE    0x80000000      /* KSM may merge identical pages */
 108
 109#ifndef VM_STACK_DEFAULT_FLAGS          /* arch can override this */
 110#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
 111#endif
 112
 113#ifdef CONFIG_STACK_GROWSUP
 114#define VM_STACK_FLAGS  (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
 115#else
 116#define VM_STACK_FLAGS  (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
 117#endif
 118
 119#define VM_READHINTMASK                 (VM_SEQ_READ | VM_RAND_READ)
 120#define VM_ClearReadHint(v)             (v)->vm_flags &= ~VM_READHINTMASK
 121#define VM_NormalReadHint(v)            (!((v)->vm_flags & VM_READHINTMASK))
 122#define VM_SequentialReadHint(v)        ((v)->vm_flags & VM_SEQ_READ)
 123#define VM_RandomReadHint(v)            ((v)->vm_flags & VM_RAND_READ)
 124
 125/*
 126 * special vmas that are non-mergable, non-mlock()able
 127 */
 128#define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP)
 129
 130/*
 131 * mapping from the currently active vm_flags protection bits (the
 132 * low four bits) to a page protection mask..
 133 */
 134extern pgprot_t protection_map[16];
 135
 136#define FAULT_FLAG_WRITE        0x01    /* Fault was a write access */
 137#define FAULT_FLAG_NONLINEAR    0x02    /* Fault was via a nonlinear mapping */
 138#define FAULT_FLAG_MKWRITE      0x04    /* Fault was mkwrite of existing pte */
 139
 140/*
 141 * This interface is used by x86 PAT code to identify a pfn mapping that is
 142 * linear over entire vma. This is to optimize PAT code that deals with
 143 * marking the physical region with a particular prot. This is not for generic
 144 * mm use. Note also that this check will not work if the pfn mapping is
 145 * linear for a vma starting at physical address 0. In which case PAT code
 146 * falls back to slow path of reserving physical range page by page.
 147 */
 148static inline int is_linear_pfn_mapping(struct vm_area_struct *vma)
 149{
 150        return (vma->vm_flags & VM_PFN_AT_MMAP);
 151}
 152
 153static inline int is_pfn_mapping(struct vm_area_struct *vma)
 154{
 155        return (vma->vm_flags & VM_PFNMAP);
 156}
 157
 158/*
 159 * vm_fault is filled by the the pagefault handler and passed to the vma's
 160 * ->fault function. The vma's ->fault is responsible for returning a bitmask
 161 * of VM_FAULT_xxx flags that give details about how the fault was handled.
 162 *
 163 * pgoff should be used in favour of virtual_address, if possible. If pgoff
 164 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
 165 * mapping support.
 166 */
 167struct vm_fault {
 168        unsigned int flags;             /* FAULT_FLAG_xxx flags */
 169        pgoff_t pgoff;                  /* Logical page offset based on vma */
 170        void __user *virtual_address;   /* Faulting virtual address */
 171
 172        struct page *page;              /* ->fault handlers should return a
 173                                         * page here, unless VM_FAULT_NOPAGE
 174                                         * is set (which is also implied by
 175                                         * VM_FAULT_ERROR).
 176                                         */
 177};
 178
 179/*
 180 * These are the virtual MM functions - opening of an area, closing and
 181 * unmapping it (needed to keep files on disk up-to-date etc), pointer
 182 * to the functions called when a no-page or a wp-page exception occurs. 
 183 */
 184struct vm_operations_struct {
 185        void (*open)(struct vm_area_struct * area);
 186        void (*close)(struct vm_area_struct * area);
 187        int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
 188
 189        /* notification that a previously read-only page is about to become
 190         * writable, if an error is returned it will cause a SIGBUS */
 191        int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
 192
 193        /* called by access_process_vm when get_user_pages() fails, typically
 194         * for use by special VMAs that can switch between memory and hardware
 195         */
 196        int (*access)(struct vm_area_struct *vma, unsigned long addr,
 197                      void *buf, int len, int write);
 198#ifdef CONFIG_NUMA
 199        /*
 200         * set_policy() op must add a reference to any non-NULL @new mempolicy
 201         * to hold the policy upon return.  Caller should pass NULL @new to
 202         * remove a policy and fall back to surrounding context--i.e. do not
 203         * install a MPOL_DEFAULT policy, nor the task or system default
 204         * mempolicy.
 205         */
 206        int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
 207
 208        /*
 209         * get_policy() op must add reference [mpol_get()] to any policy at
 210         * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
 211         * in mm/mempolicy.c will do this automatically.
 212         * get_policy() must NOT add a ref if the policy at (vma,addr) is not
 213         * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
 214         * If no [shared/vma] mempolicy exists at the addr, get_policy() op
 215         * must return NULL--i.e., do not "fallback" to task or system default
 216         * policy.
 217         */
 218        struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
 219                                        unsigned long addr);
 220        int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
 221                const nodemask_t *to, unsigned long flags);
 222#endif
 223};
 224
 225struct mmu_gather;
 226struct inode;
 227
 228#define page_private(page)              ((page)->private)
 229#define set_page_private(page, v)       ((page)->private = (v))
 230
 231/*
 232 * FIXME: take this include out, include page-flags.h in
 233 * files which need it (119 of them)
 234 */
 235#include <linux/page-flags.h>
 236
 237/*
 238 * Methods to modify the page usage count.
 239 *
 240 * What counts for a page usage:
 241 * - cache mapping   (page->mapping)
 242 * - private data    (page->private)
 243 * - page mapped in a task's page tables, each mapping
 244 *   is counted separately
 245 *
 246 * Also, many kernel routines increase the page count before a critical
 247 * routine so they can be sure the page doesn't go away from under them.
 248 */
 249
 250/*
 251 * Drop a ref, return true if the refcount fell to zero (the page has no users)
 252 */
 253static inline int put_page_testzero(struct page *page)
 254{
 255        VM_BUG_ON(atomic_read(&page->_count) == 0);
 256        return atomic_dec_and_test(&page->_count);
 257}
 258
 259/*
 260 * Try to grab a ref unless the page has a refcount of zero, return false if
 261 * that is the case.
 262 */
 263static inline int get_page_unless_zero(struct page *page)
 264{
 265        return atomic_inc_not_zero(&page->_count);
 266}
 267
 268/* Support for virtually mapped pages */
 269struct page *vmalloc_to_page(const void *addr);
 270unsigned long vmalloc_to_pfn(const void *addr);
 271
 272/*
 273 * Determine if an address is within the vmalloc range
 274 *
 275 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
 276 * is no special casing required.
 277 */
 278static inline int is_vmalloc_addr(const void *x)
 279{
 280#ifdef CONFIG_MMU
 281        unsigned long addr = (unsigned long)x;
 282
 283        return addr >= VMALLOC_START && addr < VMALLOC_END;
 284#else
 285        return 0;
 286#endif
 287}
 288#ifdef CONFIG_MMU
 289extern int is_vmalloc_or_module_addr(const void *x);
 290#else
 291static inline int is_vmalloc_or_module_addr(const void *x)
 292{
 293        return 0;
 294}
 295#endif
 296
 297static inline struct page *compound_head(struct page *page)
 298{
 299        if (unlikely(PageTail(page)))
 300                return page->first_page;
 301        return page;
 302}
 303
 304static inline int page_count(struct page *page)
 305{
 306        return atomic_read(&compound_head(page)->_count);
 307}
 308
 309static inline void get_page(struct page *page)
 310{
 311        page = compound_head(page);
 312        VM_BUG_ON(atomic_read(&page->_count) == 0);
 313        atomic_inc(&page->_count);
 314}
 315
 316static inline struct page *virt_to_head_page(const void *x)
 317{
 318        struct page *page = virt_to_page(x);
 319        return compound_head(page);
 320}
 321
 322/*
 323 * Setup the page count before being freed into the page allocator for
 324 * the first time (boot or memory hotplug)
 325 */
 326static inline void init_page_count(struct page *page)
 327{
 328        atomic_set(&page->_count, 1);
 329}
 330
 331void put_page(struct page *page);
 332void put_pages_list(struct list_head *pages);
 333
 334void split_page(struct page *page, unsigned int order);
 335
 336/*
 337 * Compound pages have a destructor function.  Provide a
 338 * prototype for that function and accessor functions.
 339 * These are _only_ valid on the head of a PG_compound page.
 340 */
 341typedef void compound_page_dtor(struct page *);
 342
 343static inline void set_compound_page_dtor(struct page *page,
 344                                                compound_page_dtor *dtor)
 345{
 346        page[1].lru.next = (void *)dtor;
 347}
 348
 349static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
 350{
 351        return (compound_page_dtor *)page[1].lru.next;
 352}
 353
 354static inline int compound_order(struct page *page)
 355{
 356        if (!PageHead(page))
 357                return 0;
 358        return (unsigned long)page[1].lru.prev;
 359}
 360
 361static inline void set_compound_order(struct page *page, unsigned long order)
 362{
 363        page[1].lru.prev = (void *)order;
 364}
 365
 366/*
 367 * Multiple processes may "see" the same page. E.g. for untouched
 368 * mappings of /dev/null, all processes see the same page full of
 369 * zeroes, and text pages of executables and shared libraries have
 370 * only one copy in memory, at most, normally.
 371 *
 372 * For the non-reserved pages, page_count(page) denotes a reference count.
 373 *   page_count() == 0 means the page is free. page->lru is then used for
 374 *   freelist management in the buddy allocator.
 375 *   page_count() > 0  means the page has been allocated.
 376 *
 377 * Pages are allocated by the slab allocator in order to provide memory
 378 * to kmalloc and kmem_cache_alloc. In this case, the management of the
 379 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
 380 * unless a particular usage is carefully commented. (the responsibility of
 381 * freeing the kmalloc memory is the caller's, of course).
 382 *
 383 * A page may be used by anyone else who does a __get_free_page().
 384 * In this case, page_count still tracks the references, and should only
 385 * be used through the normal accessor functions. The top bits of page->flags
 386 * and page->virtual store page management information, but all other fields
 387 * are unused and could be used privately, carefully. The management of this
 388 * page is the responsibility of the one who allocated it, and those who have
 389 * subsequently been given references to it.
 390 *
 391 * The other pages (we may call them "pagecache pages") are completely
 392 * managed by the Linux memory manager: I/O, buffers, swapping etc.
 393 * The following discussion applies only to them.
 394 *
 395 * A pagecache page contains an opaque `private' member, which belongs to the
 396 * page's address_space. Usually, this is the address of a circular list of
 397 * the page's disk buffers. PG_private must be set to tell the VM to call
 398 * into the filesystem to release these pages.
 399 *
 400 * A page may belong to an inode's memory mapping. In this case, page->mapping
 401 * is the pointer to the inode, and page->index is the file offset of the page,
 402 * in units of PAGE_CACHE_SIZE.
 403 *
 404 * If pagecache pages are not associated with an inode, they are said to be
 405 * anonymous pages. These may become associated with the swapcache, and in that
 406 * case PG_swapcache is set, and page->private is an offset into the swapcache.
 407 *
 408 * In either case (swapcache or inode backed), the pagecache itself holds one
 409 * reference to the page. Setting PG_private should also increment the
 410 * refcount. The each user mapping also has a reference to the page.
 411 *
 412 * The pagecache pages are stored in a per-mapping radix tree, which is
 413 * rooted at mapping->page_tree, and indexed by offset.
 414 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
 415 * lists, we instead now tag pages as dirty/writeback in the radix tree.
 416 *
 417 * All pagecache pages may be subject to I/O:
 418 * - inode pages may need to be read from disk,
 419 * - inode pages which have been modified and are MAP_SHARED may need
 420 *   to be written back to the inode on disk,
 421 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
 422 *   modified may need to be swapped out to swap space and (later) to be read
 423 *   back into memory.
 424 */
 425
 426/*
 427 * The zone field is never updated after free_area_init_core()
 428 * sets it, so none of the operations on it need to be atomic.
 429 */
 430
 431
 432/*
 433 * page->flags layout:
 434 *
 435 * There are three possibilities for how page->flags get
 436 * laid out.  The first is for the normal case, without
 437 * sparsemem.  The second is for sparsemem when there is
 438 * plenty of space for node and section.  The last is when
 439 * we have run out of space and have to fall back to an
 440 * alternate (slower) way of determining the node.
 441 *
 442 * No sparsemem or sparsemem vmemmap: |       NODE     | ZONE | ... | FLAGS |
 443 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
 444 * classic sparse no space for node:  | SECTION |     ZONE    | ... | FLAGS |
 445 */
 446#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
 447#define SECTIONS_WIDTH          SECTIONS_SHIFT
 448#else
 449#define SECTIONS_WIDTH          0
 450#endif
 451
 452#define ZONES_WIDTH             ZONES_SHIFT
 453
 454#if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
 455#define NODES_WIDTH             NODES_SHIFT
 456#else
 457#ifdef CONFIG_SPARSEMEM_VMEMMAP
 458#error "Vmemmap: No space for nodes field in page flags"
 459#endif
 460#define NODES_WIDTH             0
 461#endif
 462
 463/* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
 464#define SECTIONS_PGOFF          ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
 465#define NODES_PGOFF             (SECTIONS_PGOFF - NODES_WIDTH)
 466#define ZONES_PGOFF             (NODES_PGOFF - ZONES_WIDTH)
 467
 468/*
 469 * We are going to use the flags for the page to node mapping if its in
 470 * there.  This includes the case where there is no node, so it is implicit.
 471 */
 472#if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
 473#define NODE_NOT_IN_PAGE_FLAGS
 474#endif
 475
 476#ifndef PFN_SECTION_SHIFT
 477#define PFN_SECTION_SHIFT 0
 478#endif
 479
 480/*
 481 * Define the bit shifts to access each section.  For non-existant
 482 * sections we define the shift as 0; that plus a 0 mask ensures
 483 * the compiler will optimise away reference to them.
 484 */
 485#define SECTIONS_PGSHIFT        (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
 486#define NODES_PGSHIFT           (NODES_PGOFF * (NODES_WIDTH != 0))
 487#define ZONES_PGSHIFT           (ZONES_PGOFF * (ZONES_WIDTH != 0))
 488
 489/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
 490#ifdef NODE_NOT_IN_PAGEFLAGS
 491#define ZONEID_SHIFT            (SECTIONS_SHIFT + ZONES_SHIFT)
 492#define ZONEID_PGOFF            ((SECTIONS_PGOFF < ZONES_PGOFF)? \
 493                                                SECTIONS_PGOFF : ZONES_PGOFF)
 494#else
 495#define ZONEID_SHIFT            (NODES_SHIFT + ZONES_SHIFT)
 496#define ZONEID_PGOFF            ((NODES_PGOFF < ZONES_PGOFF)? \
 497                                                NODES_PGOFF : ZONES_PGOFF)
 498#endif
 499
 500#define ZONEID_PGSHIFT          (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
 501
 502#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
 503#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
 504#endif
 505
 506#define ZONES_MASK              ((1UL << ZONES_WIDTH) - 1)
 507#define NODES_MASK              ((1UL << NODES_WIDTH) - 1)
 508#define SECTIONS_MASK           ((1UL << SECTIONS_WIDTH) - 1)
 509#define ZONEID_MASK             ((1UL << ZONEID_SHIFT) - 1)
 510
 511static inline enum zone_type page_zonenum(struct page *page)
 512{
 513        return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
 514}
 515
 516/*
 517 * The identification function is only used by the buddy allocator for
 518 * determining if two pages could be buddies. We are not really
 519 * identifying a zone since we could be using a the section number
 520 * id if we have not node id available in page flags.
 521 * We guarantee only that it will return the same value for two
 522 * combinable pages in a zone.
 523 */
 524static inline int page_zone_id(struct page *page)
 525{
 526        return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
 527}
 528
 529static inline int zone_to_nid(struct zone *zone)
 530{
 531#ifdef CONFIG_NUMA
 532        return zone->node;
 533#else
 534        return 0;
 535#endif
 536}
 537
 538#ifdef NODE_NOT_IN_PAGE_FLAGS
 539extern int page_to_nid(struct page *page);
 540#else
 541static inline int page_to_nid(struct page *page)
 542{
 543        return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
 544}
 545#endif
 546
 547static inline struct zone *page_zone(struct page *page)
 548{
 549        return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
 550}
 551
 552#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
 553static inline unsigned long page_to_section(struct page *page)
 554{
 555        return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
 556}
 557#endif
 558
 559static inline void set_page_zone(struct page *page, enum zone_type zone)
 560{
 561        page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
 562        page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
 563}
 564
 565static inline void set_page_node(struct page *page, unsigned long node)
 566{
 567        page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
 568        page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
 569}
 570
 571static inline void set_page_section(struct page *page, unsigned long section)
 572{
 573        page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
 574        page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
 575}
 576
 577static inline void set_page_links(struct page *page, enum zone_type zone,
 578        unsigned long node, unsigned long pfn)
 579{
 580        set_page_zone(page, zone);
 581        set_page_node(page, node);
 582        set_page_section(page, pfn_to_section_nr(pfn));
 583}
 584
 585/*
 586 * Some inline functions in vmstat.h depend on page_zone()
 587 */
 588#include <linux/vmstat.h>
 589
 590static __always_inline void *lowmem_page_address(struct page *page)
 591{
 592        return __va(page_to_pfn(page) << PAGE_SHIFT);
 593}
 594
 595#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
 596#define HASHED_PAGE_VIRTUAL
 597#endif
 598
 599#if defined(WANT_PAGE_VIRTUAL)
 600#define page_address(page) ((page)->virtual)
 601#define set_page_address(page, address)                 \
 602        do {                                            \
 603                (page)->virtual = (address);            \
 604        } while(0)
 605#define page_address_init()  do { } while(0)
 606#endif
 607
 608#if defined(HASHED_PAGE_VIRTUAL)
 609void *page_address(struct page *page);
 610void set_page_address(struct page *page, void *virtual);
 611void page_address_init(void);
 612#endif
 613
 614#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
 615#define page_address(page) lowmem_page_address(page)
 616#define set_page_address(page, address)  do { } while(0)
 617#define page_address_init()  do { } while(0)
 618#endif
 619
 620/*
 621 * On an anonymous page mapped into a user virtual memory area,
 622 * page->mapping points to its anon_vma, not to a struct address_space;
 623 * with the PAGE_MAPPING_ANON bit set to distinguish it.
 624 *
 625 * Please note that, confusingly, "page_mapping" refers to the inode
 626 * address_space which maps the page from disk; whereas "page_mapped"
 627 * refers to user virtual address space into which the page is mapped.
 628 */
 629#define PAGE_MAPPING_ANON       1
 630
 631extern struct address_space swapper_space;
 632static inline struct address_space *page_mapping(struct page *page)
 633{
 634        struct address_space *mapping = page->mapping;
 635
 636        VM_BUG_ON(PageSlab(page));
 637#ifdef CONFIG_SWAP
 638        if (unlikely(PageSwapCache(page)))
 639                mapping = &swapper_space;
 640        else
 641#endif
 642        if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
 643                mapping = NULL;
 644        return mapping;
 645}
 646
 647static inline int PageAnon(struct page *page)
 648{
 649        return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
 650}
 651
 652/*
 653 * Return the pagecache index of the passed page.  Regular pagecache pages
 654 * use ->index whereas swapcache pages use ->private
 655 */
 656static inline pgoff_t page_index(struct page *page)
 657{
 658        if (unlikely(PageSwapCache(page)))
 659                return page_private(page);
 660        return page->index;
 661}
 662
 663/*
 664 * The atomic page->_mapcount, like _count, starts from -1:
 665 * so that transitions both from it and to it can be tracked,
 666 * using atomic_inc_and_test and atomic_add_negative(-1).
 667 */
 668static inline void reset_page_mapcount(struct page *page)
 669{
 670        atomic_set(&(page)->_mapcount, -1);
 671}
 672
 673static inline int page_mapcount(struct page *page)
 674{
 675        return atomic_read(&(page)->_mapcount) + 1;
 676}
 677
 678/*
 679 * Return true if this page is mapped into pagetables.
 680 */
 681static inline int page_mapped(struct page *page)
 682{
 683        return atomic_read(&(page)->_mapcount) >= 0;
 684}
 685
 686/*
 687 * Different kinds of faults, as returned by handle_mm_fault().
 688 * Used to decide whether a process gets delivered SIGBUS or
 689 * just gets major/minor fault counters bumped up.
 690 */
 691
 692#define VM_FAULT_MINOR  0 /* For backwards compat. Remove me quickly. */
 693
 694#define VM_FAULT_OOM    0x0001
 695#define VM_FAULT_SIGBUS 0x0002
 696#define VM_FAULT_MAJOR  0x0004
 697#define VM_FAULT_WRITE  0x0008  /* Special case for get_user_pages */
 698#define VM_FAULT_HWPOISON 0x0010        /* Hit poisoned page */
 699
 700#define VM_FAULT_NOPAGE 0x0100  /* ->fault installed the pte, not return page */
 701#define VM_FAULT_LOCKED 0x0200  /* ->fault locked the returned page */
 702
 703#define VM_FAULT_ERROR  (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON)
 704
 705/*
 706 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
 707 */
 708extern void pagefault_out_of_memory(void);
 709
 710#define offset_in_page(p)       ((unsigned long)(p) & ~PAGE_MASK)
 711
 712extern void show_free_areas(void);
 713
 714int shmem_lock(struct file *file, int lock, struct user_struct *user);
 715struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags);
 716int shmem_zero_setup(struct vm_area_struct *);
 717
 718#ifndef CONFIG_MMU
 719extern unsigned long shmem_get_unmapped_area(struct file *file,
 720                                             unsigned long addr,
 721                                             unsigned long len,
 722                                             unsigned long pgoff,
 723                                             unsigned long flags);
 724#endif
 725
 726extern int can_do_mlock(void);
 727extern int user_shm_lock(size_t, struct user_struct *);
 728extern void user_shm_unlock(size_t, struct user_struct *);
 729
 730/*
 731 * Parameter block passed down to zap_pte_range in exceptional cases.
 732 */
 733struct zap_details {
 734        struct vm_area_struct *nonlinear_vma;   /* Check page->index if set */
 735        struct address_space *check_mapping;    /* Check page->mapping if set */
 736        pgoff_t first_index;                    /* Lowest page->index to unmap */
 737        pgoff_t last_index;                     /* Highest page->index to unmap */
 738        spinlock_t *i_mmap_lock;                /* For unmap_mapping_range: */
 739        unsigned long truncate_count;           /* Compare vm_truncate_count */
 740};
 741
 742struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
 743                pte_t pte);
 744
 745int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
 746                unsigned long size);
 747unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
 748                unsigned long size, struct zap_details *);
 749unsigned long unmap_vmas(struct mmu_gather **tlb,
 750                struct vm_area_struct *start_vma, unsigned long start_addr,
 751                unsigned long end_addr, unsigned long *nr_accounted,
 752                struct zap_details *);
 753
 754/**
 755 * mm_walk - callbacks for walk_page_range
 756 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
 757 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
 758 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
 759 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
 760 * @pte_hole: if set, called for each hole at all levels
 761 *
 762 * (see walk_page_range for more details)
 763 */
 764struct mm_walk {
 765        int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
 766        int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
 767        int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
 768        int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
 769        int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
 770        struct mm_struct *mm;
 771        void *private;
 772};
 773
 774int walk_page_range(unsigned long addr, unsigned long end,
 775                struct mm_walk *walk);
 776void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
 777                unsigned long end, unsigned long floor, unsigned long ceiling);
 778int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
 779                        struct vm_area_struct *vma);
 780void unmap_mapping_range(struct address_space *mapping,
 781                loff_t const holebegin, loff_t const holelen, int even_cows);
 782int follow_pfn(struct vm_area_struct *vma, unsigned long address,
 783        unsigned long *pfn);
 784int follow_phys(struct vm_area_struct *vma, unsigned long address,
 785                unsigned int flags, unsigned long *prot, resource_size_t *phys);
 786int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
 787                        void *buf, int len, int write);
 788
 789static inline void unmap_shared_mapping_range(struct address_space *mapping,
 790                loff_t const holebegin, loff_t const holelen)
 791{
 792        unmap_mapping_range(mapping, holebegin, holelen, 0);
 793}
 794
 795extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new);
 796extern int vmtruncate(struct inode *inode, loff_t offset);
 797extern int vmtruncate_range(struct inode *inode, loff_t offset, loff_t end);
 798
 799int truncate_inode_page(struct address_space *mapping, struct page *page);
 800int generic_error_remove_page(struct address_space *mapping, struct page *page);
 801
 802int invalidate_inode_page(struct page *page);
 803
 804#ifdef CONFIG_MMU
 805extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
 806                        unsigned long address, unsigned int flags);
 807#else
 808static inline int handle_mm_fault(struct mm_struct *mm,
 809                        struct vm_area_struct *vma, unsigned long address,
 810                        unsigned int flags)
 811{
 812        /* should never happen if there's no MMU */
 813        BUG();
 814        return VM_FAULT_SIGBUS;
 815}
 816#endif
 817
 818extern int make_pages_present(unsigned long addr, unsigned long end);
 819extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
 820
 821int get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
 822                        unsigned long start, int nr_pages, int write, int force,
 823                        struct page **pages, struct vm_area_struct **vmas);
 824int get_user_pages_fast(unsigned long start, int nr_pages, int write,
 825                        struct page **pages);
 826struct page *get_dump_page(unsigned long addr);
 827
 828extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
 829extern void do_invalidatepage(struct page *page, unsigned long offset);
 830
 831int __set_page_dirty_nobuffers(struct page *page);
 832int __set_page_dirty_no_writeback(struct page *page);
 833int redirty_page_for_writepage(struct writeback_control *wbc,
 834                                struct page *page);
 835void account_page_dirtied(struct page *page, struct address_space *mapping);
 836int set_page_dirty(struct page *page);
 837int set_page_dirty_lock(struct page *page);
 838int clear_page_dirty_for_io(struct page *page);
 839
 840extern unsigned long move_page_tables(struct vm_area_struct *vma,
 841                unsigned long old_addr, struct vm_area_struct *new_vma,
 842                unsigned long new_addr, unsigned long len);
 843extern unsigned long do_mremap(unsigned long addr,
 844                               unsigned long old_len, unsigned long new_len,
 845                               unsigned long flags, unsigned long new_addr);
 846extern int mprotect_fixup(struct vm_area_struct *vma,
 847                          struct vm_area_struct **pprev, unsigned long start,
 848                          unsigned long end, unsigned long newflags);
 849
 850/*
 851 * doesn't attempt to fault and will return short.
 852 */
 853int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
 854                          struct page **pages);
 855
 856/*
 857 * A callback you can register to apply pressure to ageable caches.
 858 *
 859 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'.  It should
 860 * look through the least-recently-used 'nr_to_scan' entries and
 861 * attempt to free them up.  It should return the number of objects
 862 * which remain in the cache.  If it returns -1, it means it cannot do
 863 * any scanning at this time (eg. there is a risk of deadlock).
 864 *
 865 * The 'gfpmask' refers to the allocation we are currently trying to
 866 * fulfil.
 867 *
 868 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
 869 * querying the cache size, so a fastpath for that case is appropriate.
 870 */
 871struct shrinker {
 872        int (*shrink)(int nr_to_scan, gfp_t gfp_mask);
 873        int seeks;      /* seeks to recreate an obj */
 874
 875        /* These are for internal use */
 876        struct list_head list;
 877        long nr;        /* objs pending delete */
 878};
 879#define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
 880extern void register_shrinker(struct shrinker *);
 881extern void unregister_shrinker(struct shrinker *);
 882
 883int vma_wants_writenotify(struct vm_area_struct *vma);
 884
 885extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl);
 886
 887#ifdef __PAGETABLE_PUD_FOLDED
 888static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
 889                                                unsigned long address)
 890{
 891        return 0;
 892}
 893#else
 894int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
 895#endif
 896
 897#ifdef __PAGETABLE_PMD_FOLDED
 898static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
 899                                                unsigned long address)
 900{
 901        return 0;
 902}
 903#else
 904int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
 905#endif
 906
 907int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
 908int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
 909
 910/*
 911 * The following ifdef needed to get the 4level-fixup.h header to work.
 912 * Remove it when 4level-fixup.h has been removed.
 913 */
 914#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
 915static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
 916{
 917        return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
 918                NULL: pud_offset(pgd, address);
 919}
 920
 921static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
 922{
 923        return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
 924                NULL: pmd_offset(pud, address);
 925}
 926#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
 927
 928#if USE_SPLIT_PTLOCKS
 929/*
 930 * We tuck a spinlock to guard each pagetable page into its struct page,
 931 * at page->private, with BUILD_BUG_ON to make sure that this will not
 932 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
 933 * When freeing, reset page->mapping so free_pages_check won't complain.
 934 */
 935#define __pte_lockptr(page)     &((page)->ptl)
 936#define pte_lock_init(_page)    do {                                    \
 937        spin_lock_init(__pte_lockptr(_page));                           \
 938} while (0)
 939#define pte_lock_deinit(page)   ((page)->mapping = NULL)
 940#define pte_lockptr(mm, pmd)    ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
 941#else   /* !USE_SPLIT_PTLOCKS */
 942/*
 943 * We use mm->page_table_lock to guard all pagetable pages of the mm.
 944 */
 945#define pte_lock_init(page)     do {} while (0)
 946#define pte_lock_deinit(page)   do {} while (0)
 947#define pte_lockptr(mm, pmd)    ({(void)(pmd); &(mm)->page_table_lock;})
 948#endif /* USE_SPLIT_PTLOCKS */
 949
 950static inline void pgtable_page_ctor(struct page *page)
 951{
 952        pte_lock_init(page);
 953        inc_zone_page_state(page, NR_PAGETABLE);
 954}
 955
 956static inline void pgtable_page_dtor(struct page *page)
 957{
 958        pte_lock_deinit(page);
 959        dec_zone_page_state(page, NR_PAGETABLE);
 960}
 961
 962#define pte_offset_map_lock(mm, pmd, address, ptlp)     \
 963({                                                      \
 964        spinlock_t *__ptl = pte_lockptr(mm, pmd);       \
 965        pte_t *__pte = pte_offset_map(pmd, address);    \
 966        *(ptlp) = __ptl;                                \
 967        spin_lock(__ptl);                               \
 968        __pte;                                          \
 969})
 970
 971#define pte_unmap_unlock(pte, ptl)      do {            \
 972        spin_unlock(ptl);                               \
 973        pte_unmap(pte);                                 \
 974} while (0)
 975
 976#define pte_alloc_map(mm, pmd, address)                 \
 977        ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
 978                NULL: pte_offset_map(pmd, address))
 979
 980#define pte_alloc_map_lock(mm, pmd, address, ptlp)      \
 981        ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
 982                NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
 983
 984#define pte_alloc_kernel(pmd, address)                  \
 985        ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
 986                NULL: pte_offset_kernel(pmd, address))
 987
 988extern void free_area_init(unsigned long * zones_size);
 989extern void free_area_init_node(int nid, unsigned long * zones_size,
 990                unsigned long zone_start_pfn, unsigned long *zholes_size);
 991#ifdef CONFIG_ARCH_POPULATES_NODE_MAP
 992/*
 993 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
 994 * zones, allocate the backing mem_map and account for memory holes in a more
 995 * architecture independent manner. This is a substitute for creating the
 996 * zone_sizes[] and zholes_size[] arrays and passing them to
 997 * free_area_init_node()
 998 *
 999 * An architecture is expected to register range of page frames backed by
1000 * physical memory with add_active_range() before calling
1001 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1002 * usage, an architecture is expected to do something like
1003 *
1004 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1005 *                                                       max_highmem_pfn};
1006 * for_each_valid_physical_page_range()
1007 *      add_active_range(node_id, start_pfn, end_pfn)
1008 * free_area_init_nodes(max_zone_pfns);
1009 *
1010 * If the architecture guarantees that there are no holes in the ranges
1011 * registered with add_active_range(), free_bootmem_active_regions()
1012 * will call free_bootmem_node() for each registered physical page range.
1013 * Similarly sparse_memory_present_with_active_regions() calls
1014 * memory_present() for each range when SPARSEMEM is enabled.
1015 *
1016 * See mm/page_alloc.c for more information on each function exposed by
1017 * CONFIG_ARCH_POPULATES_NODE_MAP
1018 */
1019extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1020extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1021                                        unsigned long end_pfn);
1022extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
1023                                        unsigned long end_pfn);
1024extern void remove_all_active_ranges(void);
1025extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1026                                                unsigned long end_pfn);
1027extern void get_pfn_range_for_nid(unsigned int nid,
1028                        unsigned long *start_pfn, unsigned long *end_pfn);
1029extern unsigned long find_min_pfn_with_active_regions(void);
1030extern void free_bootmem_with_active_regions(int nid,
1031                                                unsigned long max_low_pfn);
1032typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
1033extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
1034extern void sparse_memory_present_with_active_regions(int nid);
1035#endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
1036
1037#if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \
1038    !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1039static inline int __early_pfn_to_nid(unsigned long pfn)
1040{
1041        return 0;
1042}
1043#else
1044/* please see mm/page_alloc.c */
1045extern int __meminit early_pfn_to_nid(unsigned long pfn);
1046#ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1047/* there is a per-arch backend function. */
1048extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1049#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1050#endif
1051
1052extern void set_dma_reserve(unsigned long new_dma_reserve);
1053extern void memmap_init_zone(unsigned long, int, unsigned long,
1054                                unsigned long, enum memmap_context);
1055extern void setup_per_zone_wmarks(void);
1056extern void calculate_zone_inactive_ratio(struct zone *zone);
1057extern void mem_init(void);
1058extern void __init mmap_init(void);
1059extern void show_mem(void);
1060extern void si_meminfo(struct sysinfo * val);
1061extern void si_meminfo_node(struct sysinfo *val, int nid);
1062extern int after_bootmem;
1063
1064#ifdef CONFIG_NUMA
1065extern void setup_per_cpu_pageset(void);
1066#else
1067static inline void setup_per_cpu_pageset(void) {}
1068#endif
1069
1070extern void zone_pcp_update(struct zone *zone);
1071
1072/* nommu.c */
1073extern atomic_long_t mmap_pages_allocated;
1074
1075/* prio_tree.c */
1076void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1077void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1078void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1079struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1080        struct prio_tree_iter *iter);
1081
1082#define vma_prio_tree_foreach(vma, iter, root, begin, end)      \
1083        for (prio_tree_iter_init(iter, root, begin, end), vma = NULL;   \
1084                (vma = vma_prio_tree_next(vma, iter)); )
1085
1086static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1087                                        struct list_head *list)
1088{
1089        vma->shared.vm_set.parent = NULL;
1090        list_add_tail(&vma->shared.vm_set.list, list);
1091}
1092
1093/* mmap.c */
1094extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1095extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
1096        unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1097extern struct vm_area_struct *vma_merge(struct mm_struct *,
1098        struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1099        unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1100        struct mempolicy *);
1101extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1102extern int split_vma(struct mm_struct *,
1103        struct vm_area_struct *, unsigned long addr, int new_below);
1104extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1105extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1106        struct rb_node **, struct rb_node *);
1107extern void unlink_file_vma(struct vm_area_struct *);
1108extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1109        unsigned long addr, unsigned long len, pgoff_t pgoff);
1110extern void exit_mmap(struct mm_struct *);
1111
1112extern int mm_take_all_locks(struct mm_struct *mm);
1113extern void mm_drop_all_locks(struct mm_struct *mm);
1114
1115#ifdef CONFIG_PROC_FS
1116/* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1117extern void added_exe_file_vma(struct mm_struct *mm);
1118extern void removed_exe_file_vma(struct mm_struct *mm);
1119#else
1120static inline void added_exe_file_vma(struct mm_struct *mm)
1121{}
1122
1123static inline void removed_exe_file_vma(struct mm_struct *mm)
1124{}
1125#endif /* CONFIG_PROC_FS */
1126
1127extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1128extern int install_special_mapping(struct mm_struct *mm,
1129                                   unsigned long addr, unsigned long len,
1130                                   unsigned long flags, struct page **pages);
1131
1132extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1133
1134extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1135        unsigned long len, unsigned long prot,
1136        unsigned long flag, unsigned long pgoff);
1137extern unsigned long mmap_region(struct file *file, unsigned long addr,
1138        unsigned long len, unsigned long flags,
1139        unsigned int vm_flags, unsigned long pgoff);
1140
1141static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1142        unsigned long len, unsigned long prot,
1143        unsigned long flag, unsigned long offset)
1144{
1145        unsigned long ret = -EINVAL;
1146        if ((offset + PAGE_ALIGN(len)) < offset)
1147                goto out;
1148        if (!(offset & ~PAGE_MASK))
1149                ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1150out:
1151        return ret;
1152}
1153
1154extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1155
1156extern unsigned long do_brk(unsigned long, unsigned long);
1157
1158/* filemap.c */
1159extern unsigned long page_unuse(struct page *);
1160extern void truncate_inode_pages(struct address_space *, loff_t);
1161extern void truncate_inode_pages_range(struct address_space *,
1162                                       loff_t lstart, loff_t lend);
1163
1164/* generic vm_area_ops exported for stackable file systems */
1165extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1166
1167/* mm/page-writeback.c */
1168int write_one_page(struct page *page, int wait);
1169void task_dirty_inc(struct task_struct *tsk);
1170
1171/* readahead.c */
1172#define VM_MAX_READAHEAD        128     /* kbytes */
1173#define VM_MIN_READAHEAD        16      /* kbytes (includes current page) */
1174
1175int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1176                        pgoff_t offset, unsigned long nr_to_read);
1177
1178void page_cache_sync_readahead(struct address_space *mapping,
1179                               struct file_ra_state *ra,
1180                               struct file *filp,
1181                               pgoff_t offset,
1182                               unsigned long size);
1183
1184void page_cache_async_readahead(struct address_space *mapping,
1185                                struct file_ra_state *ra,
1186                                struct file *filp,
1187                                struct page *pg,
1188                                pgoff_t offset,
1189                                unsigned long size);
1190
1191unsigned long max_sane_readahead(unsigned long nr);
1192unsigned long ra_submit(struct file_ra_state *ra,
1193                        struct address_space *mapping,
1194                        struct file *filp);
1195
1196/* Do stack extension */
1197extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1198#ifdef CONFIG_IA64
1199extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1200#endif
1201extern int expand_stack_downwards(struct vm_area_struct *vma,
1202                                  unsigned long address);
1203
1204/* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
1205extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1206extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1207                                             struct vm_area_struct **pprev);
1208
1209/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1210   NULL if none.  Assume start_addr < end_addr. */
1211static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1212{
1213        struct vm_area_struct * vma = find_vma(mm,start_addr);
1214
1215        if (vma && end_addr <= vma->vm_start)
1216                vma = NULL;
1217        return vma;
1218}
1219
1220static inline unsigned long vma_pages(struct vm_area_struct *vma)
1221{
1222        return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1223}
1224
1225pgprot_t vm_get_page_prot(unsigned long vm_flags);
1226struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1227int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1228                        unsigned long pfn, unsigned long size, pgprot_t);
1229int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1230int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1231                        unsigned long pfn);
1232int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1233                        unsigned long pfn);
1234
1235struct page *follow_page(struct vm_area_struct *, unsigned long address,
1236                        unsigned int foll_flags);
1237#define FOLL_WRITE      0x01    /* check pte is writable */
1238#define FOLL_TOUCH      0x02    /* mark page accessed */
1239#define FOLL_GET        0x04    /* do get_page on page */
1240#define FOLL_DUMP       0x08    /* give error on hole if it would be zero */
1241#define FOLL_FORCE      0x10    /* get_user_pages read/write w/o permission */
1242
1243typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1244                        void *data);
1245extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1246                               unsigned long size, pte_fn_t fn, void *data);
1247
1248#ifdef CONFIG_PROC_FS
1249void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1250#else
1251static inline void vm_stat_account(struct mm_struct *mm,
1252                        unsigned long flags, struct file *file, long pages)
1253{
1254}
1255#endif /* CONFIG_PROC_FS */
1256
1257#ifdef CONFIG_DEBUG_PAGEALLOC
1258extern int debug_pagealloc_enabled;
1259
1260extern void kernel_map_pages(struct page *page, int numpages, int enable);
1261
1262static inline void enable_debug_pagealloc(void)
1263{
1264        debug_pagealloc_enabled = 1;
1265}
1266#ifdef CONFIG_HIBERNATION
1267extern bool kernel_page_present(struct page *page);
1268#endif /* CONFIG_HIBERNATION */
1269#else
1270static inline void
1271kernel_map_pages(struct page *page, int numpages, int enable) {}
1272static inline void enable_debug_pagealloc(void)
1273{
1274}
1275#ifdef CONFIG_HIBERNATION
1276static inline bool kernel_page_present(struct page *page) { return true; }
1277#endif /* CONFIG_HIBERNATION */
1278#endif
1279
1280extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1281#ifdef  __HAVE_ARCH_GATE_AREA
1282int in_gate_area_no_task(unsigned long addr);
1283int in_gate_area(struct task_struct *task, unsigned long addr);
1284#else
1285int in_gate_area_no_task(unsigned long addr);
1286#define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1287#endif  /* __HAVE_ARCH_GATE_AREA */
1288
1289int drop_caches_sysctl_handler(struct ctl_table *, int,
1290                                        void __user *, size_t *, loff_t *);
1291unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
1292                        unsigned long lru_pages);
1293
1294#ifndef CONFIG_MMU
1295#define randomize_va_space 0
1296#else
1297extern int randomize_va_space;
1298#endif
1299
1300const char * arch_vma_name(struct vm_area_struct *vma);
1301void print_vma_addr(char *prefix, unsigned long rip);
1302
1303struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1304pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1305pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1306pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1307pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1308void *vmemmap_alloc_block(unsigned long size, int node);
1309void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1310int vmemmap_populate_basepages(struct page *start_page,
1311                                                unsigned long pages, int node);
1312int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
1313void vmemmap_populate_print_last(void);
1314
1315extern int account_locked_memory(struct mm_struct *mm, struct rlimit *rlim,
1316                                 size_t size);
1317extern void refund_locked_memory(struct mm_struct *mm, size_t size);
1318
1319extern void memory_failure(unsigned long pfn, int trapno);
1320extern int __memory_failure(unsigned long pfn, int trapno, int ref);
1321extern int sysctl_memory_failure_early_kill;
1322extern int sysctl_memory_failure_recovery;
1323extern atomic_long_t mce_bad_pages;
1324
1325#endif /* __KERNEL__ */
1326#endif /* _LINUX_MM_H */
1327