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;
  21
  22#ifndef CONFIG_DISCONTIGMEM          /* Don't use mapnrs, do it properly */
  23extern unsigned long max_mapnr;
  24#endif
  25
  26extern unsigned long num_physpages;
  27extern void * high_memory;
  28extern int page_cluster;
  29
  30#ifdef CONFIG_SYSCTL
  31extern int sysctl_legacy_va_layout;
  32#else
  33#define sysctl_legacy_va_layout 0
  34#endif
  35
  36extern unsigned long mmap_min_addr;
  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/*
  59 * This struct defines the per-mm list of VMAs for uClinux. If CONFIG_MMU is
  60 * disabled, then there's a single shared list of VMAs maintained by the
  61 * system, and mm's subscribe to these individually
  62 */
  63struct vm_list_struct {
  64        struct vm_list_struct   *next;
  65        struct vm_area_struct   *vma;
  66};
  67
  68#ifndef CONFIG_MMU
  69extern struct rb_root nommu_vma_tree;
  70extern struct rw_semaphore nommu_vma_sem;
  71
  72extern unsigned int kobjsize(const void *objp);
  73#endif
  74
  75/*
  76 * vm_flags in vm_area_struct, see mm_types.h.
  77 */
  78#define VM_READ         0x00000001      /* currently active flags */
  79#define VM_WRITE        0x00000002
  80#define VM_EXEC         0x00000004
  81#define VM_SHARED       0x00000008
  82
  83/* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
  84#define VM_MAYREAD      0x00000010      /* limits for mprotect() etc */
  85#define VM_MAYWRITE     0x00000020
  86#define VM_MAYEXEC      0x00000040
  87#define VM_MAYSHARE     0x00000080
  88
  89#define VM_GROWSDOWN    0x00000100      /* general info on the segment */
  90#define VM_GROWSUP      0x00000200
  91#define VM_PFNMAP       0x00000400      /* Page-ranges managed without "struct page", just pure PFN */
  92#define VM_DENYWRITE    0x00000800      /* ETXTBSY on write attempts.. */
  93
  94#define VM_EXECUTABLE   0x00001000
  95#define VM_LOCKED       0x00002000
  96#define VM_IO           0x00004000      /* Memory mapped I/O or similar */
  97
  98                                        /* Used by sys_madvise() */
  99#define VM_SEQ_READ     0x00008000      /* App will access data sequentially */
 100#define VM_RAND_READ    0x00010000      /* App will not benefit from clustered reads */
 101
 102#define VM_DONTCOPY     0x00020000      /* Do not copy this vma on fork */
 103#define VM_DONTEXPAND   0x00040000      /* Cannot expand with mremap() */
 104#define VM_RESERVED     0x00080000      /* Count as reserved_vm like IO */
 105#define VM_ACCOUNT      0x00100000      /* Is a VM accounted object */
 106#define VM_NORESERVE    0x00200000      /* should the VM suppress accounting */
 107#define VM_HUGETLB      0x00400000      /* Huge TLB Page VM */
 108#define VM_NONLINEAR    0x00800000      /* Is non-linear (remap_file_pages) */
 109#define VM_MAPPED_COPY  0x01000000      /* T if mapped copy of data (nommu mmap) */
 110#define VM_INSERTPAGE   0x02000000      /* The vma has had "vm_insert_page()" done on it */
 111#define VM_ALWAYSDUMP   0x04000000      /* Always include in core dumps */
 112
 113#define VM_CAN_NONLINEAR 0x08000000     /* Has ->fault & does nonlinear pages */
 114#define VM_MIXEDMAP     0x10000000      /* Can contain "struct page" and pure PFN pages */
 115#define VM_SAO          0x20000000      /* Strong Access Ordering (powerpc) */
 116
 117#ifndef VM_STACK_DEFAULT_FLAGS          /* arch can override this */
 118#define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
 119#endif
 120
 121#ifdef CONFIG_STACK_GROWSUP
 122#define VM_STACK_FLAGS  (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
 123#else
 124#define VM_STACK_FLAGS  (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
 125#endif
 126
 127#define VM_READHINTMASK                 (VM_SEQ_READ | VM_RAND_READ)
 128#define VM_ClearReadHint(v)             (v)->vm_flags &= ~VM_READHINTMASK
 129#define VM_NormalReadHint(v)            (!((v)->vm_flags & VM_READHINTMASK))
 130#define VM_SequentialReadHint(v)        ((v)->vm_flags & VM_SEQ_READ)
 131#define VM_RandomReadHint(v)            ((v)->vm_flags & VM_RAND_READ)
 132
 133/*
 134 * mapping from the currently active vm_flags protection bits (the
 135 * low four bits) to a page protection mask..
 136 */
 137extern pgprot_t protection_map[16];
 138
 139#define FAULT_FLAG_WRITE        0x01    /* Fault was a write access */
 140#define FAULT_FLAG_NONLINEAR    0x02    /* Fault was via a nonlinear mapping */
 141#define FAULT_FLAG_MKWRITE      0x04    /* Fault was mkwrite of existing pte */
 142
 143
 144/*
 145 * vm_fault is filled by the the pagefault handler and passed to the vma's
 146 * ->fault function. The vma's ->fault is responsible for returning a bitmask
 147 * of VM_FAULT_xxx flags that give details about how the fault was handled.
 148 *
 149 * pgoff should be used in favour of virtual_address, if possible. If pgoff
 150 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
 151 * mapping support.
 152 */
 153struct vm_fault {
 154        unsigned int flags;             /* FAULT_FLAG_xxx flags */
 155        pgoff_t pgoff;                  /* Logical page offset based on vma */
 156        void __user *virtual_address;   /* Faulting virtual address */
 157
 158        struct page *page;              /* ->fault handlers should return a
 159                                         * page here, unless VM_FAULT_NOPAGE
 160                                         * is set (which is also implied by
 161                                         * VM_FAULT_ERROR).
 162                                         */
 163};
 164
 165/*
 166 * These are the virtual MM functions - opening of an area, closing and
 167 * unmapping it (needed to keep files on disk up-to-date etc), pointer
 168 * to the functions called when a no-page or a wp-page exception occurs. 
 169 */
 170struct vm_operations_struct {
 171        void (*open)(struct vm_area_struct * area);
 172        void (*close)(struct vm_area_struct * area);
 173        int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
 174
 175        /* notification that a previously read-only page is about to become
 176         * writable, if an error is returned it will cause a SIGBUS */
 177        int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
 178
 179        /* called by access_process_vm when get_user_pages() fails, typically
 180         * for use by special VMAs that can switch between memory and hardware
 181         */
 182        int (*access)(struct vm_area_struct *vma, unsigned long addr,
 183                      void *buf, int len, int write);
 184#ifdef CONFIG_NUMA
 185        /*
 186         * set_policy() op must add a reference to any non-NULL @new mempolicy
 187         * to hold the policy upon return.  Caller should pass NULL @new to
 188         * remove a policy and fall back to surrounding context--i.e. do not
 189         * install a MPOL_DEFAULT policy, nor the task or system default
 190         * mempolicy.
 191         */
 192        int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
 193
 194        /*
 195         * get_policy() op must add reference [mpol_get()] to any policy at
 196         * (vma,addr) marked as MPOL_SHARED.  The shared policy infrastructure
 197         * in mm/mempolicy.c will do this automatically.
 198         * get_policy() must NOT add a ref if the policy at (vma,addr) is not
 199         * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
 200         * If no [shared/vma] mempolicy exists at the addr, get_policy() op
 201         * must return NULL--i.e., do not "fallback" to task or system default
 202         * policy.
 203         */
 204        struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
 205                                        unsigned long addr);
 206        int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
 207                const nodemask_t *to, unsigned long flags);
 208#endif
 209};
 210
 211struct mmu_gather;
 212struct inode;
 213
 214#define page_private(page)              ((page)->private)
 215#define set_page_private(page, v)       ((page)->private = (v))
 216
 217/*
 218 * FIXME: take this include out, include page-flags.h in
 219 * files which need it (119 of them)
 220 */
 221#include <linux/page-flags.h>
 222
 223#ifdef CONFIG_DEBUG_VM
 224#define VM_BUG_ON(cond) BUG_ON(cond)
 225#else
 226#define VM_BUG_ON(condition) do { } while(0)
 227#endif
 228
 229/*
 230 * Methods to modify the page usage count.
 231 *
 232 * What counts for a page usage:
 233 * - cache mapping   (page->mapping)
 234 * - private data    (page->private)
 235 * - page mapped in a task's page tables, each mapping
 236 *   is counted separately
 237 *
 238 * Also, many kernel routines increase the page count before a critical
 239 * routine so they can be sure the page doesn't go away from under them.
 240 */
 241
 242/*
 243 * Drop a ref, return true if the refcount fell to zero (the page has no users)
 244 */
 245static inline int put_page_testzero(struct page *page)
 246{
 247        VM_BUG_ON(atomic_read(&page->_count) == 0);
 248        return atomic_dec_and_test(&page->_count);
 249}
 250
 251/*
 252 * Try to grab a ref unless the page has a refcount of zero, return false if
 253 * that is the case.
 254 */
 255static inline int get_page_unless_zero(struct page *page)
 256{
 257        return atomic_inc_not_zero(&page->_count);
 258}
 259
 260/* Support for virtually mapped pages */
 261struct page *vmalloc_to_page(const void *addr);
 262unsigned long vmalloc_to_pfn(const void *addr);
 263
 264/*
 265 * Determine if an address is within the vmalloc range
 266 *
 267 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
 268 * is no special casing required.
 269 */
 270static inline int is_vmalloc_addr(const void *x)
 271{
 272#ifdef CONFIG_MMU
 273        unsigned long addr = (unsigned long)x;
 274
 275        return addr >= VMALLOC_START && addr < VMALLOC_END;
 276#else
 277        return 0;
 278#endif
 279}
 280
 281static inline struct page *compound_head(struct page *page)
 282{
 283        if (unlikely(PageTail(page)))
 284                return page->first_page;
 285        return page;
 286}
 287
 288static inline int page_count(struct page *page)
 289{
 290        return atomic_read(&compound_head(page)->_count);
 291}
 292
 293static inline void get_page(struct page *page)
 294{
 295        page = compound_head(page);
 296        VM_BUG_ON(atomic_read(&page->_count) == 0);
 297        atomic_inc(&page->_count);
 298}
 299
 300static inline struct page *virt_to_head_page(const void *x)
 301{
 302        struct page *page = virt_to_page(x);
 303        return compound_head(page);
 304}
 305
 306/*
 307 * Setup the page count before being freed into the page allocator for
 308 * the first time (boot or memory hotplug)
 309 */
 310static inline void init_page_count(struct page *page)
 311{
 312        atomic_set(&page->_count, 1);
 313}
 314
 315void put_page(struct page *page);
 316void put_pages_list(struct list_head *pages);
 317
 318void split_page(struct page *page, unsigned int order);
 319
 320/*
 321 * Compound pages have a destructor function.  Provide a
 322 * prototype for that function and accessor functions.
 323 * These are _only_ valid on the head of a PG_compound page.
 324 */
 325typedef void compound_page_dtor(struct page *);
 326
 327static inline void set_compound_page_dtor(struct page *page,
 328                                                compound_page_dtor *dtor)
 329{
 330        page[1].lru.next = (void *)dtor;
 331}
 332
 333static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
 334{
 335        return (compound_page_dtor *)page[1].lru.next;
 336}
 337
 338static inline int compound_order(struct page *page)
 339{
 340        if (!PageHead(page))
 341                return 0;
 342        return (unsigned long)page[1].lru.prev;
 343}
 344
 345static inline void set_compound_order(struct page *page, unsigned long order)
 346{
 347        page[1].lru.prev = (void *)order;
 348}
 349
 350/*
 351 * Multiple processes may "see" the same page. E.g. for untouched
 352 * mappings of /dev/null, all processes see the same page full of
 353 * zeroes, and text pages of executables and shared libraries have
 354 * only one copy in memory, at most, normally.
 355 *
 356 * For the non-reserved pages, page_count(page) denotes a reference count.
 357 *   page_count() == 0 means the page is free. page->lru is then used for
 358 *   freelist management in the buddy allocator.
 359 *   page_count() > 0  means the page has been allocated.
 360 *
 361 * Pages are allocated by the slab allocator in order to provide memory
 362 * to kmalloc and kmem_cache_alloc. In this case, the management of the
 363 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
 364 * unless a particular usage is carefully commented. (the responsibility of
 365 * freeing the kmalloc memory is the caller's, of course).
 366 *
 367 * A page may be used by anyone else who does a __get_free_page().
 368 * In this case, page_count still tracks the references, and should only
 369 * be used through the normal accessor functions. The top bits of page->flags
 370 * and page->virtual store page management information, but all other fields
 371 * are unused and could be used privately, carefully. The management of this
 372 * page is the responsibility of the one who allocated it, and those who have
 373 * subsequently been given references to it.
 374 *
 375 * The other pages (we may call them "pagecache pages") are completely
 376 * managed by the Linux memory manager: I/O, buffers, swapping etc.
 377 * The following discussion applies only to them.
 378 *
 379 * A pagecache page contains an opaque `private' member, which belongs to the
 380 * page's address_space. Usually, this is the address of a circular list of
 381 * the page's disk buffers. PG_private must be set to tell the VM to call
 382 * into the filesystem to release these pages.
 383 *
 384 * A page may belong to an inode's memory mapping. In this case, page->mapping
 385 * is the pointer to the inode, and page->index is the file offset of the page,
 386 * in units of PAGE_CACHE_SIZE.
 387 *
 388 * If pagecache pages are not associated with an inode, they are said to be
 389 * anonymous pages. These may become associated with the swapcache, and in that
 390 * case PG_swapcache is set, and page->private is an offset into the swapcache.
 391 *
 392 * In either case (swapcache or inode backed), the pagecache itself holds one
 393 * reference to the page. Setting PG_private should also increment the
 394 * refcount. The each user mapping also has a reference to the page.
 395 *
 396 * The pagecache pages are stored in a per-mapping radix tree, which is
 397 * rooted at mapping->page_tree, and indexed by offset.
 398 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
 399 * lists, we instead now tag pages as dirty/writeback in the radix tree.
 400 *
 401 * All pagecache pages may be subject to I/O:
 402 * - inode pages may need to be read from disk,
 403 * - inode pages which have been modified and are MAP_SHARED may need
 404 *   to be written back to the inode on disk,
 405 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
 406 *   modified may need to be swapped out to swap space and (later) to be read
 407 *   back into memory.
 408 */
 409
 410/*
 411 * The zone field is never updated after free_area_init_core()
 412 * sets it, so none of the operations on it need to be atomic.
 413 */
 414
 415
 416/*
 417 * page->flags layout:
 418 *
 419 * There are three possibilities for how page->flags get
 420 * laid out.  The first is for the normal case, without
 421 * sparsemem.  The second is for sparsemem when there is
 422 * plenty of space for node and section.  The last is when
 423 * we have run out of space and have to fall back to an
 424 * alternate (slower) way of determining the node.
 425 *
 426 * No sparsemem or sparsemem vmemmap: |       NODE     | ZONE | ... | FLAGS |
 427 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
 428 * classic sparse no space for node:  | SECTION |     ZONE    | ... | FLAGS |
 429 */
 430#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
 431#define SECTIONS_WIDTH          SECTIONS_SHIFT
 432#else
 433#define SECTIONS_WIDTH          0
 434#endif
 435
 436#define ZONES_WIDTH             ZONES_SHIFT
 437
 438#if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
 439#define NODES_WIDTH             NODES_SHIFT
 440#else
 441#ifdef CONFIG_SPARSEMEM_VMEMMAP
 442#error "Vmemmap: No space for nodes field in page flags"
 443#endif
 444#define NODES_WIDTH             0
 445#endif
 446
 447/* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
 448#define SECTIONS_PGOFF          ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
 449#define NODES_PGOFF             (SECTIONS_PGOFF - NODES_WIDTH)
 450#define ZONES_PGOFF             (NODES_PGOFF - ZONES_WIDTH)
 451
 452/*
 453 * We are going to use the flags for the page to node mapping if its in
 454 * there.  This includes the case where there is no node, so it is implicit.
 455 */
 456#if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
 457#define NODE_NOT_IN_PAGE_FLAGS
 458#endif
 459
 460#ifndef PFN_SECTION_SHIFT
 461#define PFN_SECTION_SHIFT 0
 462#endif
 463
 464/*
 465 * Define the bit shifts to access each section.  For non-existant
 466 * sections we define the shift as 0; that plus a 0 mask ensures
 467 * the compiler will optimise away reference to them.
 468 */
 469#define SECTIONS_PGSHIFT        (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
 470#define NODES_PGSHIFT           (NODES_PGOFF * (NODES_WIDTH != 0))
 471#define ZONES_PGSHIFT           (ZONES_PGOFF * (ZONES_WIDTH != 0))
 472
 473/* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allcator */
 474#ifdef NODE_NOT_IN_PAGEFLAGS
 475#define ZONEID_SHIFT            (SECTIONS_SHIFT + ZONES_SHIFT)
 476#define ZONEID_PGOFF            ((SECTIONS_PGOFF < ZONES_PGOFF)? \
 477                                                SECTIONS_PGOFF : ZONES_PGOFF)
 478#else
 479#define ZONEID_SHIFT            (NODES_SHIFT + ZONES_SHIFT)
 480#define ZONEID_PGOFF            ((NODES_PGOFF < ZONES_PGOFF)? \
 481                                                NODES_PGOFF : ZONES_PGOFF)
 482#endif
 483
 484#define ZONEID_PGSHIFT          (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
 485
 486#if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
 487#error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
 488#endif
 489
 490#define ZONES_MASK              ((1UL << ZONES_WIDTH) - 1)
 491#define NODES_MASK              ((1UL << NODES_WIDTH) - 1)
 492#define SECTIONS_MASK           ((1UL << SECTIONS_WIDTH) - 1)
 493#define ZONEID_MASK             ((1UL << ZONEID_SHIFT) - 1)
 494
 495static inline enum zone_type page_zonenum(struct page *page)
 496{
 497        return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
 498}
 499
 500/*
 501 * The identification function is only used by the buddy allocator for
 502 * determining if two pages could be buddies. We are not really
 503 * identifying a zone since we could be using a the section number
 504 * id if we have not node id available in page flags.
 505 * We guarantee only that it will return the same value for two
 506 * combinable pages in a zone.
 507 */
 508static inline int page_zone_id(struct page *page)
 509{
 510        return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
 511}
 512
 513static inline int zone_to_nid(struct zone *zone)
 514{
 515#ifdef CONFIG_NUMA
 516        return zone->node;
 517#else
 518        return 0;
 519#endif
 520}
 521
 522#ifdef NODE_NOT_IN_PAGE_FLAGS
 523extern int page_to_nid(struct page *page);
 524#else
 525static inline int page_to_nid(struct page *page)
 526{
 527        return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
 528}
 529#endif
 530
 531static inline struct zone *page_zone(struct page *page)
 532{
 533        return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
 534}
 535
 536#if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
 537static inline unsigned long page_to_section(struct page *page)
 538{
 539        return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
 540}
 541#endif
 542
 543static inline void set_page_zone(struct page *page, enum zone_type zone)
 544{
 545        page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
 546        page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
 547}
 548
 549static inline void set_page_node(struct page *page, unsigned long node)
 550{
 551        page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
 552        page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
 553}
 554
 555static inline void set_page_section(struct page *page, unsigned long section)
 556{
 557        page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
 558        page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
 559}
 560
 561static inline void set_page_links(struct page *page, enum zone_type zone,
 562        unsigned long node, unsigned long pfn)
 563{
 564        set_page_zone(page, zone);
 565        set_page_node(page, node);
 566        set_page_section(page, pfn_to_section_nr(pfn));
 567}
 568
 569/*
 570 * If a hint addr is less than mmap_min_addr change hint to be as
 571 * low as possible but still greater than mmap_min_addr
 572 */
 573static inline unsigned long round_hint_to_min(unsigned long hint)
 574{
 575        hint &= PAGE_MASK;
 576        if (((void *)hint != NULL) &&
 577            (hint < mmap_min_addr))
 578                return PAGE_ALIGN(mmap_min_addr);
 579        return hint;
 580}
 581
 582/*
 583 * Some inline functions in vmstat.h depend on page_zone()
 584 */
 585#include <linux/vmstat.h>
 586
 587static __always_inline void *lowmem_page_address(struct page *page)
 588{
 589        return __va(page_to_pfn(page) << PAGE_SHIFT);
 590}
 591
 592#if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
 593#define HASHED_PAGE_VIRTUAL
 594#endif
 595
 596#if defined(WANT_PAGE_VIRTUAL)
 597#define page_address(page) ((page)->virtual)
 598#define set_page_address(page, address)                 \
 599        do {                                            \
 600                (page)->virtual = (address);            \
 601        } while(0)
 602#define page_address_init()  do { } while(0)
 603#endif
 604
 605#if defined(HASHED_PAGE_VIRTUAL)
 606void *page_address(struct page *page);
 607void set_page_address(struct page *page, void *virtual);
 608void page_address_init(void);
 609#endif
 610
 611#if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
 612#define page_address(page) lowmem_page_address(page)
 613#define set_page_address(page, address)  do { } while(0)
 614#define page_address_init()  do { } while(0)
 615#endif
 616
 617/*
 618 * On an anonymous page mapped into a user virtual memory area,
 619 * page->mapping points to its anon_vma, not to a struct address_space;
 620 * with the PAGE_MAPPING_ANON bit set to distinguish it.
 621 *
 622 * Please note that, confusingly, "page_mapping" refers to the inode
 623 * address_space which maps the page from disk; whereas "page_mapped"
 624 * refers to user virtual address space into which the page is mapped.
 625 */
 626#define PAGE_MAPPING_ANON       1
 627
 628extern struct address_space swapper_space;
 629static inline struct address_space *page_mapping(struct page *page)
 630{
 631        struct address_space *mapping = page->mapping;
 632
 633        VM_BUG_ON(PageSlab(page));
 634#ifdef CONFIG_SWAP
 635        if (unlikely(PageSwapCache(page)))
 636                mapping = &swapper_space;
 637        else
 638#endif
 639        if (unlikely((unsigned long)mapping & PAGE_MAPPING_ANON))
 640                mapping = NULL;
 641        return mapping;
 642}
 643
 644static inline int PageAnon(struct page *page)
 645{
 646        return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
 647}
 648
 649/*
 650 * Return the pagecache index of the passed page.  Regular pagecache pages
 651 * use ->index whereas swapcache pages use ->private
 652 */
 653static inline pgoff_t page_index(struct page *page)
 654{
 655        if (unlikely(PageSwapCache(page)))
 656                return page_private(page);
 657        return page->index;
 658}
 659
 660/*
 661 * The atomic page->_mapcount, like _count, starts from -1:
 662 * so that transitions both from it and to it can be tracked,
 663 * using atomic_inc_and_test and atomic_add_negative(-1).
 664 */
 665static inline void reset_page_mapcount(struct page *page)
 666{
 667        atomic_set(&(page)->_mapcount, -1);
 668}
 669
 670static inline int page_mapcount(struct page *page)
 671{
 672        return atomic_read(&(page)->_mapcount) + 1;
 673}
 674
 675/*
 676 * Return true if this page is mapped into pagetables.
 677 */
 678static inline int page_mapped(struct page *page)
 679{
 680        return atomic_read(&(page)->_mapcount) >= 0;
 681}
 682
 683/*
 684 * Different kinds of faults, as returned by handle_mm_fault().
 685 * Used to decide whether a process gets delivered SIGBUS or
 686 * just gets major/minor fault counters bumped up.
 687 */
 688
 689#define VM_FAULT_MINOR  0 /* For backwards compat. Remove me quickly. */
 690
 691#define VM_FAULT_OOM    0x0001
 692#define VM_FAULT_SIGBUS 0x0002
 693#define VM_FAULT_MAJOR  0x0004
 694#define VM_FAULT_WRITE  0x0008  /* Special case for get_user_pages */
 695
 696#define VM_FAULT_NOPAGE 0x0100  /* ->fault installed the pte, not return page */
 697#define VM_FAULT_LOCKED 0x0200  /* ->fault locked the returned page */
 698
 699#define VM_FAULT_ERROR  (VM_FAULT_OOM | VM_FAULT_SIGBUS)
 700
 701#define offset_in_page(p)       ((unsigned long)(p) & ~PAGE_MASK)
 702
 703extern void show_free_areas(void);
 704
 705#ifdef CONFIG_SHMEM
 706int shmem_lock(struct file *file, int lock, struct user_struct *user);
 707#else
 708static inline int shmem_lock(struct file *file, int lock,
 709                             struct user_struct *user)
 710{
 711        return 0;
 712}
 713#endif
 714struct file *shmem_file_setup(char *name, loff_t size, unsigned long flags);
 715
 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 generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
 783                        void *buf, int len, int write);
 784
 785static inline void unmap_shared_mapping_range(struct address_space *mapping,
 786                loff_t const holebegin, loff_t const holelen)
 787{
 788        unmap_mapping_range(mapping, holebegin, holelen, 0);
 789}
 790
 791extern int vmtruncate(struct inode * inode, loff_t offset);
 792extern int vmtruncate_range(struct inode * inode, loff_t offset, loff_t end);
 793
 794#ifdef CONFIG_MMU
 795extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
 796                        unsigned long address, int write_access);
 797#else
 798static inline int handle_mm_fault(struct mm_struct *mm,
 799                        struct vm_area_struct *vma, unsigned long address,
 800                        int write_access)
 801{
 802        /* should never happen if there's no MMU */
 803        BUG();
 804        return VM_FAULT_SIGBUS;
 805}
 806#endif
 807
 808extern int make_pages_present(unsigned long addr, unsigned long end);
 809extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
 810
 811int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, unsigned long start,
 812                int len, int write, int force, struct page **pages, struct vm_area_struct **vmas);
 813
 814extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
 815extern void do_invalidatepage(struct page *page, unsigned long offset);
 816
 817int __set_page_dirty_nobuffers(struct page *page);
 818int __set_page_dirty_no_writeback(struct page *page);
 819int redirty_page_for_writepage(struct writeback_control *wbc,
 820                                struct page *page);
 821int set_page_dirty(struct page *page);
 822int set_page_dirty_lock(struct page *page);
 823int clear_page_dirty_for_io(struct page *page);
 824
 825extern unsigned long move_page_tables(struct vm_area_struct *vma,
 826                unsigned long old_addr, struct vm_area_struct *new_vma,
 827                unsigned long new_addr, unsigned long len);
 828extern unsigned long do_mremap(unsigned long addr,
 829                               unsigned long old_len, unsigned long new_len,
 830                               unsigned long flags, unsigned long new_addr);
 831extern int mprotect_fixup(struct vm_area_struct *vma,
 832                          struct vm_area_struct **pprev, unsigned long start,
 833                          unsigned long end, unsigned long newflags);
 834
 835/*
 836 * get_user_pages_fast provides equivalent functionality to get_user_pages,
 837 * operating on current and current->mm (force=0 and doesn't return any vmas).
 838 *
 839 * get_user_pages_fast may take mmap_sem and page tables, so no assumptions
 840 * can be made about locking. get_user_pages_fast is to be implemented in a
 841 * way that is advantageous (vs get_user_pages()) when the user memory area is
 842 * already faulted in and present in ptes. However if the pages have to be
 843 * faulted in, it may turn out to be slightly slower).
 844 */
 845int get_user_pages_fast(unsigned long start, int nr_pages, int write,
 846                        struct page **pages);
 847
 848/*
 849 * A callback you can register to apply pressure to ageable caches.
 850 *
 851 * 'shrink' is passed a count 'nr_to_scan' and a 'gfpmask'.  It should
 852 * look through the least-recently-used 'nr_to_scan' entries and
 853 * attempt to free them up.  It should return the number of objects
 854 * which remain in the cache.  If it returns -1, it means it cannot do
 855 * any scanning at this time (eg. there is a risk of deadlock).
 856 *
 857 * The 'gfpmask' refers to the allocation we are currently trying to
 858 * fulfil.
 859 *
 860 * Note that 'shrink' will be passed nr_to_scan == 0 when the VM is
 861 * querying the cache size, so a fastpath for that case is appropriate.
 862 */
 863struct shrinker {
 864        int (*shrink)(int nr_to_scan, gfp_t gfp_mask);
 865        int seeks;      /* seeks to recreate an obj */
 866
 867        /* These are for internal use */
 868        struct list_head list;
 869        long nr;        /* objs pending delete */
 870};
 871#define DEFAULT_SEEKS 2 /* A good number if you don't know better. */
 872extern void register_shrinker(struct shrinker *);
 873extern void unregister_shrinker(struct shrinker *);
 874
 875int vma_wants_writenotify(struct vm_area_struct *vma);
 876
 877extern pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, spinlock_t **ptl);
 878
 879#ifdef __PAGETABLE_PUD_FOLDED
 880static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
 881                                                unsigned long address)
 882{
 883        return 0;
 884}
 885#else
 886int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
 887#endif
 888
 889#ifdef __PAGETABLE_PMD_FOLDED
 890static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
 891                                                unsigned long address)
 892{
 893        return 0;
 894}
 895#else
 896int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
 897#endif
 898
 899int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
 900int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
 901
 902/*
 903 * The following ifdef needed to get the 4level-fixup.h header to work.
 904 * Remove it when 4level-fixup.h has been removed.
 905 */
 906#if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
 907static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
 908{
 909        return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
 910                NULL: pud_offset(pgd, address);
 911}
 912
 913static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
 914{
 915        return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
 916                NULL: pmd_offset(pud, address);
 917}
 918#endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
 919
 920#if NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS
 921/*
 922 * We tuck a spinlock to guard each pagetable page into its struct page,
 923 * at page->private, with BUILD_BUG_ON to make sure that this will not
 924 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
 925 * When freeing, reset page->mapping so free_pages_check won't complain.
 926 */
 927#define __pte_lockptr(page)     &((page)->ptl)
 928#define pte_lock_init(_page)    do {                                    \
 929        spin_lock_init(__pte_lockptr(_page));                           \
 930} while (0)
 931#define pte_lock_deinit(page)   ((page)->mapping = NULL)
 932#define pte_lockptr(mm, pmd)    ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
 933#else
 934/*
 935 * We use mm->page_table_lock to guard all pagetable pages of the mm.
 936 */
 937#define pte_lock_init(page)     do {} while (0)
 938#define pte_lock_deinit(page)   do {} while (0)
 939#define pte_lockptr(mm, pmd)    ({(void)(pmd); &(mm)->page_table_lock;})
 940#endif /* NR_CPUS < CONFIG_SPLIT_PTLOCK_CPUS */
 941
 942static inline void pgtable_page_ctor(struct page *page)
 943{
 944        pte_lock_init(page);
 945        inc_zone_page_state(page, NR_PAGETABLE);
 946}
 947
 948static inline void pgtable_page_dtor(struct page *page)
 949{
 950        pte_lock_deinit(page);
 951        dec_zone_page_state(page, NR_PAGETABLE);
 952}
 953
 954#define pte_offset_map_lock(mm, pmd, address, ptlp)     \
 955({                                                      \
 956        spinlock_t *__ptl = pte_lockptr(mm, pmd);       \
 957        pte_t *__pte = pte_offset_map(pmd, address);    \
 958        *(ptlp) = __ptl;                                \
 959        spin_lock(__ptl);                               \
 960        __pte;                                          \
 961})
 962
 963#define pte_unmap_unlock(pte, ptl)      do {            \
 964        spin_unlock(ptl);                               \
 965        pte_unmap(pte);                                 \
 966} while (0)
 967
 968#define pte_alloc_map(mm, pmd, address)                 \
 969        ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
 970                NULL: pte_offset_map(pmd, address))
 971
 972#define pte_alloc_map_lock(mm, pmd, address, ptlp)      \
 973        ((unlikely(!pmd_present(*(pmd))) && __pte_alloc(mm, pmd, address))? \
 974                NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
 975
 976#define pte_alloc_kernel(pmd, address)                  \
 977        ((unlikely(!pmd_present(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
 978                NULL: pte_offset_kernel(pmd, address))
 979
 980extern void free_area_init(unsigned long * zones_size);
 981extern void free_area_init_node(int nid, unsigned long * zones_size,
 982                unsigned long zone_start_pfn, unsigned long *zholes_size);
 983#ifdef CONFIG_ARCH_POPULATES_NODE_MAP
 984/*
 985 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
 986 * zones, allocate the backing mem_map and account for memory holes in a more
 987 * architecture independent manner. This is a substitute for creating the
 988 * zone_sizes[] and zholes_size[] arrays and passing them to
 989 * free_area_init_node()
 990 *
 991 * An architecture is expected to register range of page frames backed by
 992 * physical memory with add_active_range() before calling
 993 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
 994 * usage, an architecture is expected to do something like
 995 *
 996 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
 997 *                                                       max_highmem_pfn};
 998 * for_each_valid_physical_page_range()
 999 *      add_active_range(node_id, start_pfn, end_pfn)
1000 * free_area_init_nodes(max_zone_pfns);
1001 *
1002 * If the architecture guarantees that there are no holes in the ranges
1003 * registered with add_active_range(), free_bootmem_active_regions()
1004 * will call free_bootmem_node() for each registered physical page range.
1005 * Similarly sparse_memory_present_with_active_regions() calls
1006 * memory_present() for each range when SPARSEMEM is enabled.
1007 *
1008 * See mm/page_alloc.c for more information on each function exposed by
1009 * CONFIG_ARCH_POPULATES_NODE_MAP
1010 */
1011extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1012extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1013                                        unsigned long end_pfn);
1014extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
1015                                        unsigned long end_pfn);
1016extern void push_node_boundaries(unsigned int nid, unsigned long start_pfn,
1017                                        unsigned long end_pfn);
1018extern void remove_all_active_ranges(void);
1019extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1020                                                unsigned long end_pfn);
1021extern void get_pfn_range_for_nid(unsigned int nid,
1022                        unsigned long *start_pfn, unsigned long *end_pfn);
1023extern unsigned long find_min_pfn_with_active_regions(void);
1024extern void free_bootmem_with_active_regions(int nid,
1025                                                unsigned long max_low_pfn);
1026typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
1027extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
1028extern void sparse_memory_present_with_active_regions(int nid);
1029#endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
1030
1031#if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \
1032    !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1033static inline int __early_pfn_to_nid(unsigned long pfn)
1034{
1035        return 0;
1036}
1037#else
1038/* please see mm/page_alloc.c */
1039extern int __meminit early_pfn_to_nid(unsigned long pfn);
1040#ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1041/* there is a per-arch backend function. */
1042extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1043#endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1044#endif
1045
1046extern void set_dma_reserve(unsigned long new_dma_reserve);
1047extern void memmap_init_zone(unsigned long, int, unsigned long,
1048                                unsigned long, enum memmap_context);
1049extern void setup_per_zone_pages_min(void);
1050extern void mem_init(void);
1051extern void show_mem(void);
1052extern void si_meminfo(struct sysinfo * val);
1053extern void si_meminfo_node(struct sysinfo *val, int nid);
1054extern int after_bootmem;
1055
1056#ifdef CONFIG_NUMA
1057extern void setup_per_cpu_pageset(void);
1058#else
1059static inline void setup_per_cpu_pageset(void) {}
1060#endif
1061
1062/* prio_tree.c */
1063void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1064void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1065void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1066struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1067        struct prio_tree_iter *iter);
1068
1069#define vma_prio_tree_foreach(vma, iter, root, begin, end)      \
1070        for (prio_tree_iter_init(iter, root, begin, end), vma = NULL;   \
1071                (vma = vma_prio_tree_next(vma, iter)); )
1072
1073static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1074                                        struct list_head *list)
1075{
1076        vma->shared.vm_set.parent = NULL;
1077        list_add_tail(&vma->shared.vm_set.list, list);
1078}
1079
1080/* mmap.c */
1081extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1082extern void vma_adjust(struct vm_area_struct *vma, unsigned long start,
1083        unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1084extern struct vm_area_struct *vma_merge(struct mm_struct *,
1085        struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1086        unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1087        struct mempolicy *);
1088extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1089extern int split_vma(struct mm_struct *,
1090        struct vm_area_struct *, unsigned long addr, int new_below);
1091extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1092extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1093        struct rb_node **, struct rb_node *);
1094extern void unlink_file_vma(struct vm_area_struct *);
1095extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1096        unsigned long addr, unsigned long len, pgoff_t pgoff);
1097extern void exit_mmap(struct mm_struct *);
1098
1099extern int mm_take_all_locks(struct mm_struct *mm);
1100extern void mm_drop_all_locks(struct mm_struct *mm);
1101
1102#ifdef CONFIG_PROC_FS
1103/* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1104extern void added_exe_file_vma(struct mm_struct *mm);
1105extern void removed_exe_file_vma(struct mm_struct *mm);
1106#else
1107static inline void added_exe_file_vma(struct mm_struct *mm)
1108{}
1109
1110static inline void removed_exe_file_vma(struct mm_struct *mm)
1111{}
1112#endif /* CONFIG_PROC_FS */
1113
1114extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1115extern int install_special_mapping(struct mm_struct *mm,
1116                                   unsigned long addr, unsigned long len,
1117                                   unsigned long flags, struct page **pages);
1118
1119extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1120
1121extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1122        unsigned long len, unsigned long prot,
1123        unsigned long flag, unsigned long pgoff);
1124extern unsigned long mmap_region(struct file *file, unsigned long addr,
1125        unsigned long len, unsigned long flags,
1126        unsigned int vm_flags, unsigned long pgoff,
1127        int accountable);
1128
1129static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1130        unsigned long len, unsigned long prot,
1131        unsigned long flag, unsigned long offset)
1132{
1133        unsigned long ret = -EINVAL;
1134        if ((offset + PAGE_ALIGN(len)) < offset)
1135                goto out;
1136        if (!(offset & ~PAGE_MASK))
1137                ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1138out:
1139        return ret;
1140}
1141
1142extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1143
1144extern unsigned long do_brk(unsigned long, unsigned long);
1145
1146/* filemap.c */
1147extern unsigned long page_unuse(struct page *);
1148extern void truncate_inode_pages(struct address_space *, loff_t);
1149extern void truncate_inode_pages_range(struct address_space *,
1150                                       loff_t lstart, loff_t lend);
1151
1152/* generic vm_area_ops exported for stackable file systems */
1153extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1154
1155/* mm/page-writeback.c */
1156int write_one_page(struct page *page, int wait);
1157
1158/* readahead.c */
1159#define VM_MAX_READAHEAD        128     /* kbytes */
1160#define VM_MIN_READAHEAD        16      /* kbytes (includes current page) */
1161
1162int do_page_cache_readahead(struct address_space *mapping, struct file *filp,
1163                        pgoff_t offset, unsigned long nr_to_read);
1164int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1165                        pgoff_t offset, unsigned long nr_to_read);
1166
1167void page_cache_sync_readahead(struct address_space *mapping,
1168                               struct file_ra_state *ra,
1169                               struct file *filp,
1170                               pgoff_t offset,
1171                               unsigned long size);
1172
1173void page_cache_async_readahead(struct address_space *mapping,
1174                                struct file_ra_state *ra,
1175                                struct file *filp,
1176                                struct page *pg,
1177                                pgoff_t offset,
1178                                unsigned long size);
1179
1180unsigned long max_sane_readahead(unsigned long nr);
1181
1182/* Do stack extension */
1183extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1184#ifdef CONFIG_IA64
1185extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1186#endif
1187extern int expand_stack_downwards(struct vm_area_struct *vma,
1188                                  unsigned long address);
1189
1190/* Look up the first VMA which satisfies  addr < vm_end,  NULL if none. */
1191extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1192extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1193                                             struct vm_area_struct **pprev);
1194
1195/* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1196   NULL if none.  Assume start_addr < end_addr. */
1197static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1198{
1199        struct vm_area_struct * vma = find_vma(mm,start_addr);
1200
1201        if (vma && end_addr <= vma->vm_start)
1202                vma = NULL;
1203        return vma;
1204}
1205
1206static inline unsigned long vma_pages(struct vm_area_struct *vma)
1207{
1208        return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1209}
1210
1211pgprot_t vm_get_page_prot(unsigned long vm_flags);
1212struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1213int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1214                        unsigned long pfn, unsigned long size, pgprot_t);
1215int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1216int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1217                        unsigned long pfn);
1218int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1219                        unsigned long pfn);
1220
1221struct page *follow_page(struct vm_area_struct *, unsigned long address,
1222                        unsigned int foll_flags);
1223#define FOLL_WRITE      0x01    /* check pte is writable */
1224#define FOLL_TOUCH      0x02    /* mark page accessed */
1225#define FOLL_GET        0x04    /* do get_page on page */
1226#define FOLL_ANON       0x08    /* give ZERO_PAGE if no pgtable */
1227
1228typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1229                        void *data);
1230extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1231                               unsigned long size, pte_fn_t fn, void *data);
1232
1233#ifdef CONFIG_PROC_FS
1234void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1235#else
1236static inline void vm_stat_account(struct mm_struct *mm,
1237                        unsigned long flags, struct file *file, long pages)
1238{
1239}
1240#endif /* CONFIG_PROC_FS */
1241
1242#ifdef CONFIG_DEBUG_PAGEALLOC
1243extern int debug_pagealloc_enabled;
1244
1245extern void kernel_map_pages(struct page *page, int numpages, int enable);
1246
1247static inline void enable_debug_pagealloc(void)
1248{
1249        debug_pagealloc_enabled = 1;
1250}
1251#ifdef CONFIG_HIBERNATION
1252extern bool kernel_page_present(struct page *page);
1253#endif /* CONFIG_HIBERNATION */
1254#else
1255static inline void
1256kernel_map_pages(struct page *page, int numpages, int enable) {}
1257static inline void enable_debug_pagealloc(void)
1258{
1259}
1260#ifdef CONFIG_HIBERNATION
1261static inline bool kernel_page_present(struct page *page) { return true; }
1262#endif /* CONFIG_HIBERNATION */
1263#endif
1264
1265extern struct vm_area_struct *get_gate_vma(struct task_struct *tsk);
1266#ifdef  __HAVE_ARCH_GATE_AREA
1267int in_gate_area_no_task(unsigned long addr);
1268int in_gate_area(struct task_struct *task, unsigned long addr);
1269#else
1270int in_gate_area_no_task(unsigned long addr);
1271#define in_gate_area(task, addr) ({(void)task; in_gate_area_no_task(addr);})
1272#endif  /* __HAVE_ARCH_GATE_AREA */
1273
1274int drop_caches_sysctl_handler(struct ctl_table *, int, struct file *,
1275                                        void __user *, size_t *, loff_t *);
1276unsigned long shrink_slab(unsigned long scanned, gfp_t gfp_mask,
1277                        unsigned long lru_pages);
1278
1279#ifndef CONFIG_MMU
1280#define randomize_va_space 0
1281#else
1282extern int randomize_va_space;
1283#endif
1284
1285const char * arch_vma_name(struct vm_area_struct *vma);
1286void print_vma_addr(char *prefix, unsigned long rip);
1287
1288struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1289pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1290pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1291pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1292pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1293void *vmemmap_alloc_block(unsigned long size, int node);
1294void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1295int vmemmap_populate_basepages(struct page *start_page,
1296                                                unsigned long pages, int node);
1297int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
1298void vmemmap_populate_print_last(void);
1299
1300#endif /* __KERNEL__ */
1301#endif /* _LINUX_MM_H */
1302
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.